xref: /netbsd-src/external/gpl3/gcc/dist/NEWS (revision 3f351f34c6d827cf017cdcff3543f6ec0c88b420)
1This file contains information about GCC releases which has been generated
2automatically from the online release notes.  It covers releases of GCC
3(and the former EGCS project) since EGCS 1.0, on the line of development
4that led to GCC 3. For information on GCC 2.8.1 and older releases of GCC 2,
5see ONEWS.
6
7======================================================================
8http://gcc.gnu.org/gcc-12/index.html
9
10                             GCC 12 Release Series
11
12   May 8, 2023
13
14   The GCC developers are pleased to announce the release of GCC 12.3.
15
16   This release is a bug-fix release, containing fixes for regressions in
17   GCC 12.2 relative to previous releases of GCC.
18
19Release History
20
21   GCC 12.3
22          May 8, 2023 ([1]changes, [2]documentation)
23
24   GCC 12.2
25          Aug 19, 2022 ([3]changes, [4]documentation)
26
27   GCC 12.1
28          May 6, 2022 ([5]changes, [6]documentation)
29
30References and Acknowledgements
31
32   GCC used to stand for the GNU C Compiler, but since the compiler
33   supports several other languages aside from C, it now stands for the
34   GNU Compiler Collection.
35
36   A list of [7]successful builds is updated as new information becomes
37   available.
38
39   The GCC developers would like to thank the numerous people that have
40   contributed new features, improvements, bug fixes, and other changes as
41   well as test results to GCC. This [8]amazing group of volunteers is
42   what makes GCC successful.
43
44   For additional information about GCC please refer to the [9]GCC project
45   web site or contact the [10]GCC development mailing list.
46
47   To obtain GCC please use [11]our mirror sites or [12]our version
48   control system.
49
50
51    For questions related to the use of GCC, please consult these web
52    pages and the [13]GCC manuals. If that fails, the
53    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
54    web pages and the development of GCC are welcome on our developer
55    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
56    archives.
57
58   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
59   distribution of this entire article is permitted in any medium,
60   provided this notice is preserved.
61
62   These pages are [18]maintained by the GCC team. Last modified
63   2023-05-08.
64
65References
66
67   1. http://gcc.gnu.org/gcc-12/changes.html
68   2. http://gcc.gnu.org/onlinedocs/12.3.0/
69   3. http://gcc.gnu.org/gcc-12/changes.html
70   4. http://gcc.gnu.org/onlinedocs/12.2.0/
71   5. http://gcc.gnu.org/gcc-12/changes.html
72   6. http://gcc.gnu.org/onlinedocs/12.1.0/
73   7. http://gcc.gnu.org/gcc-12/buildstat.html
74   8. http://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Contributors.html
75   9. http://gcc.gnu.org/index.html
76  10. mailto:gcc@gcc.gnu.org
77  11. http://gcc.gnu.org/mirrors.html
78  12. http://gcc.gnu.org/git.html
79  13. https://gcc.gnu.org/onlinedocs/
80  14. mailto:gcc-help@gcc.gnu.org
81  15. mailto:gcc@gcc.gnu.org
82  16. https://gcc.gnu.org/lists.html
83  17. https://www.fsf.org/
84  18. https://gcc.gnu.org/about.html
85======================================================================
86http://gcc.gnu.org/gcc-12/changes.html
87
88                             GCC 12 Release Series
89                        Changes, New Features, and Fixes
90
91   This page is a "brief" summary of some of the huge number of
92   improvements in GCC 12. You may also want to check out our [1]Porting
93   to GCC 12 page and the [2]full GCC documentation.
94
95Caveats
96
97     * An ABI incompatibility between C and C++ when passing or returning
98       by value certain aggregates containing zero width bit-fields has
99       been discovered on various targets. As mentioned in [3]PR102024,
100       since the [4]PR42217 fix in GCC 4.5 the C++ front-end has been
101       removing zero width bit-fields from the internal representation of
102       the aggregates after the layout of those aggregates, but the C
103       front-end kept them, so passing e.g. struct S { float a; int : 0;
104       float b; } or struct T { float c; int : 0; } by value could differ
105       between C and C++. Starting with GCC 12 the C++ front-end no longer
106       removes those bit-fields from the internal representation and per
107       clarified psABI some targets have been changed, so that they either
108       ignore those bit-fields in the argument passing by value decisions
109       in both C and C++, or they always take them into account. x86-64,
110       ARM and AArch64 will always ignore them (so there is a C ABI
111       incompatibility between GCC 11 and earlier with GCC 12 or later),
112       PowerPC64 ELFv2 and S/390 always take them into account (so there
113       is a C++ ABI incompatibility, GCC 4.4 and earlier compatible with
114       GCC 12 or later, incompatible with GCC 4.5 through GCC 11). RISC-V
115       has changed the handling of these already starting with GCC 10. As
116       the ABI requires, MIPS takes them into account handling function
117       return values so there is a C++ ABI incompatibility with GCC 4.5
118       through 11. For function arguments on MIPS, refer to [5]the MIPS
119       specific entry. GCC 12 on the above targets will report such
120       incompatibilities as warnings or other diagnostics unless
121       -Wno-psabi is used.
122     * C: Computed gotos require a pointer type now.
123     * C++: Two non-standard std::pair constructors have been deprecated.
124       These allowed the use of an rvalue and a literal 0 to construct a
125       pair containing a move-only type and a pointer. The nullptr keyword
126       should be used to initialize the pointer member instead of a
127       literal 0, as this is portable to other C++ implementations.
128     * The configuration option --enable-libstdcxx-allocator no longer
129       supports the bitmap, mt, and pool arguments. Those configurations
130       had been broken for some time.
131     * D: Building and bootstrapping GDC, the D compiler, now requires a
132       working GDC compiler (GCC version 9.1 or later) and D runtime
133       library, libphobos, as the D front end is written in D. On some
134       targets, libphobos isn't enabled by default, but compiles and works
135       if --enable-libphobos is used. Other targets may require a more
136       recent minimum version of GCC to bootstrap. Specifics are
137       documented for affected targets in the [6]manual for installing
138       GCC.
139     * Fortran: OpenMP code using the omp_lib.h include file can no longer
140       be compiled with -std=f95 but now requires at least -std=f2003.
141       Alternatively, use the omp_lib module, which still supports
142       -std=f95 and is recommended to be used instead in general.
143     * OpenMP offloading to Intel MIC has been deprecated and will be
144       removed in a future release.
145     * The cr16 target with the cr16-*-* configuration has been obsoleted
146       and will be removed in a future release.
147     * The hppa[12]*-*-hpux10* and hppa[12]*-*-hpux11* configurations
148       targeting 32-bit PA-RISC with HP-UX have been obsoleted and will be
149       removed in a future release.
150     * The m32c*-*-rtems* configuration has been obsoleted and will be
151       removed in a future release.
152     * The support for the m32r-*-linux*, m32rle-*-linux*,
153       m68k*-*-openbsd* and vax-*-openbsd* configurations has been
154       removed.
155     * STABS: Support for emitting the STABS debugging format is
156       deprecated and will be removed in the next release. All ports now
157       default to emit DWARF (version 2 or later) debugging info or are
158       obsoleted.
159     * The optimization level -Ofast now implies
160       -fno-semantic-interposition.
161
162General Improvements
163
164     * Vectorization is enabled at -O2 which is now equivalent to the
165       original -O2 -ftree-vectorize -fvect-cost-model=very-cheap. Note
166       that default vectorizer cost model has been changed which used to
167       behave as -fvect-cost-model=cheap were specified.
168     * GCC now supports the [7]ShadowCallStack sanitizer, which can be
169       enabled using the command-line option
170       [8]-fsanitize=shadow-call-stack. This sanitizer currently only
171       works on AArch64 targets and it requires an environment in which
172       all code has been compiled with -ffixed-r18. Its primary initial
173       user is the Linux kernel.
174
175New Languages and Language specific improvements
176
177     * OpenMP
178          + OpenMP 5.0 support has been extended: The close map modifier
179            and the affinity clause are now supported. In addition,
180            Fortran gained the following features which were available in
181            C and C++ before: declare variant is now available, depobj,
182            mutexinoutset and iterator can now also be used with the
183            depend clause, defaultmap has been updated for OpenMP 5.0, and
184            the loop directive and combined directives involving the
185            master directive have been added.
186          + The following OpenMP 5.1 features have been added: support for
187            expressing OpenMP directives as C++ 11 attributes, the masked
188            and scope constructs, the nothing and error directives, and
189            using primary with the proc_bind clause and OMP_PROC_BIND
190            environment variable, the reproducible and unconstrained
191            modifiers to the order clause, and, for C/C++ only, the align
192            and allocator modifiers to the allocate clause and the atomic
193            extensions are now available. The OMP_PLACE environment
194            variable supports the OpenMP 5.1 features. In addition, the
195            OMP_NUM_TEAMS and OMP_TEAMS_THREAD_LIMIT environment variables
196            and their associated API routines are now supported as well as
197            the memory-allocation routines added for Fortran and extended
198            for C/C++ in OpenMP 5.1. In Fortran code, strictly structured
199            blocks can be used.
200          + The [9]OpenMP Implementation Status can be found in the
201            libgomp manual.
202     * Version 2.6 of the [10]OpenACC specification continues to be
203       maintained and improved in the C, C++ and Fortran compilers. See
204       the [11]implementation status section on the OpenACC wiki page and
205       the [12]run-time library documentation for further information. In
206       addition to general performance tuning and bug fixing, new features
207       include:
208          + OpenACC worker parallelism for [13]AMD GPUs (already for a
209            long time supported for [14]Nvidia GPUs).
210          + Data privatization/sharing at the OpenACC gang level.
211          + Considerable improvements for the experimental OpenACC
212            'kernels' decomposition ([15]--param
213            openacc-kernels=decompose).
214          + A new warning flag [16]-Wopenacc-parallelism to warn about
215            potentially suboptimal choices related to OpenACC parallelism.
216     * The offload target code generation for OpenMP and OpenACC can now
217       be better adjusted using the new [17]-foffload-options= flag and
218       the pre-existing but now documented [18]-foffload= flag.
219
220  Ada
221
222     * Ada 2022
223          + Added the -gnat2022 flag to indicate strict Ada 2022
224            compliance. The old -gnat2020 flag is now deprecated.
225          + Support for Big Numbers (Annex G) has seen continuous
226            improvements. It is now considered complete. It is compatible
227            with SPARK, i.e. can be used from SPARK code.
228          + Continuous improvements to the Ada 2022 standard since GCC 11.
229          + Greatly improved compile time support. More functions can now
230            have the with Static aspect and can be used in more contexts.
231     * Ada 2022 extensions. The use of the -gnatX flag is necessary to
232       access these features as they are not considered stable or
233       standard.
234          + Fixed lower bound for unconstrained arrays.
235               o type Matrix is array (Natural range 0 .. <>, Natural
236                 range 0 .. <>) of Integer; is now valid.
237               o Subtypes can also specify a lower bound: subtype String_1
238                 is String (1 .. <>);. Boundaries from slices will "slide"
239                 to the correct lower bound of the subtype.
240          + Generalized Object.Operand notation. The following code is now
241            valid V.Add_Element(42);, with V being a vector, for example.
242          + Additional when constructs. Keywords return, goto and raise
243            can now use when in addition to the existing exit when. The
244            following expression is therefore now valid raise
245            Constraint_Error with "Element is null" when Element = null;
246          + Pattern matching
247               o The case statement has been extended to cover records and
248                 arrays as well as finer grained casing on scalar types.
249                 In the future it is expected to provide more compile time
250                 guarantees when accessing discriminated fields. Case
251                 exhaustion is supported for pattern matching. An example
252                 would be
253type Sign is (Neg, Zero, Pos);
254
255function Multiply (S1, S2 : Sign) return Sign is
256  (case (S1, S2) is
257     when (Neg, Neg) | (Pos, Pos) => Pos,
258     when (Zero, <>) | (<>, Zero) => Zero,
259     when (Neg, Pos) | (Pos, Neg) => Neg);
260
261     * gnatfind and gnatxref, which were already deprecated, have been
262       removed.
263     * Greatly expanded code covered by contracts. Thanks to this work,
264       there are now several Ada standard libraries fully proven in SPARK
265       which means they have no runtime nor logical errors. They are
266       mostly numeric and string handling libraries.
267     * Enable return-slot optimization for Pure functions.
268     * General optimizations, improvements and additions to the standard
269       library. Performance, correctness and in some cases stability was
270       improved. Memory pools have also seen some minor enhancements.
271     * Improvements to embedded-RTOS targets such as RTEMS, VxWorks and
272       QNX. Older targets were removed or cleaned.
273     * Added some [19]hardening features.
274
275  C family
276
277     * Support for __builtin_shufflevector compatible with the clang
278       language extension was added.
279     * Support for attribute unavailable was added.
280     * A new built-in function, __builtin_assoc_barrier, was added. It can
281       be used to inhibit re-association of floating-point expressions.
282     * Support for __builtin_dynamic_object_size compatible with the clang
283       language extension was added.
284     * New warnings:
285          + [20]-Wbidi-chars warns about potentially misleading UTF-8
286            bidirectional control characters. The default is
287            -Wbidi-chars=unpaired ([21]PR103026)
288          + [22]-Warray-compare warns about comparisons between two
289            operands of array type ([23]PR97573)
290     * Enhancements to existing warnings:
291          + [24]-Wattributes has been extended so that it's possible to
292            use -Wno-attributes=ns::attr or -Wno-attributes=ns:: to
293            suppress warnings about unknown scoped attributes (in C++11
294            and C2X). Similarly, #pragma GCC diagnostic ignored_attributes
295            "vendor::attr" can be used to achieve the same effect
296            ([25]PR101940)
297
298  C
299
300     * Some new features from the upcoming C2X revision of the ISO C
301       standard are supported with -std=c2x and -std=gnu2x. Some of these
302       features are also supported as extensions when compiling for older
303       language versions. In addition to the features listed, some
304       features previously supported as extensions and now added to the C
305       standard are enabled by default in C2X mode and not diagnosed with
306       -std=c2x -Wpedantic.
307          + Digit separators (as in C++) are supported for C2X.
308          + The #elifdef and #elifndef preprocessing directives are now
309            supported.
310          + The printf and scanf format checking with [26]-Wformat now
311            supports the %b format specified by C2X for binary integers,
312            and the %B format recommended by C2X for printf.
313
314  C++
315
316     * Several C++23 features have been implemented:
317          + [27]P1938R3, if consteval ([28]PR100974)
318          + [29]P0849R8, auto(x): decay-copy in the language
319            ([30]PR103049)
320          + [31]P2242R3, Non-literal variables (and labels and gotos) in
321            constexpr functions ([32]PR102612)
322          + [33]P2334R1, Support for preprocessing directives elifdef and
323            elifndef ([34]PR102616)
324          + [35]P2360R0, Extend init-statement to allow alias-declaration
325            ([36]PR102617)
326          + [37]P2128R6, Multidimensional subscript operator
327          + [38]DR 2397, auto specifier for pointers and references to
328            arrays ([39]PR100975)
329     * Several C++ Defect Reports have been resolved, e.g.:
330          + [40]DR 960, Covariant functions and lvalue/rvalue references
331          + [41]DR 1227, Mixing immediate and non-immediate contexts in
332            deduction failure
333          + [42]DR 1315, Restrictions on non-type template arguments in
334            partial specializations
335          + [43]DR 2082, Referring to parameters in unevaluated operands
336            of default arguments
337          + [44]DR 2351, void{}
338          + [45]DR 2374, Overly permissive specification of enum
339            direct-list-initialization
340          + [46]DR 2397, auto specifier for pointers and references to
341            arrays
342          + [47]DR 2446, Questionable type-dependency of concept-ids
343     * New command-line option -fimplicit-constexpr can be used to make
344       inline functions implicitly constexpr ([48]git)
345     * New command-line option -ffold-simple-inlines can be used to fold
346       calls to certain trivial inline functions (currently std::move,
347       std::forward, std::addressof and std::as_const). In contrast to
348       inlining such calls, folding means that no intermediate code or
349       debug information will be generated for them; this minimizes the
350       abstraction penalty incurred for using these functions versus using
351       the fundamental operations from which they're defined (e.g.
352       std::move versus static_cast). This flag is enabled by default when
353       -fno-inline is not active.
354     * Deduction guides can be declared at class scope ([49]PR79501)
355     * [50]-Wuninitialized warns about using uninitialized variables in
356       member initializer lists ([51]PR19808)
357     * [52]-Wint-in-bool-context is now disabled when instantiating a
358       template ([53]git)
359     * Stricter checking of attributes on friend declarations: if a friend
360       declaration has an attribute, that declaration must be a
361       definition. Moreover, a C++11 attribute cannot appear in the middle
362       of the decl-specifier-seq. ([54]PR99032)
363     * New warning options for C++ language mismatches:
364       -Wc++11-extensions, -Wc++14-extensions, -Wc++17-extensions,
365       -Wc++20-extensions, and -Wc++23-extensions. They are enabled by
366       default and can be used to control existing pedwarns about
367       occurrences of new C++ constructs in code using an old C++ standard
368       dialect.
369     * New warning [55]-Wmissing-requires warns about missing requires
370       ([56]git)
371     * The existing std::is_constant_evaluated in if warning was extended
372       to warn in more cases ([57]PR100995)
373     * [58]-Waddress has been enhanced so that it now warns about, for
374       instance, comparing the address of a nonstatic member function to
375       null ([59]PR102103)
376     * Errors about narrowing are no longer hidden if they occur in system
377       headers
378     * Ordered comparison of null pointers is now rejected ([60]PR99701)
379     * Anonymous structs with bases are now rejected ([61]git)
380     * The compiler rejects taking the address of an immediate member
381       function ([62]PR102753)
382     * The compiler has support for C++20
383       __cpp_lib_is_pointer_interconvertible and
384       __cpp_lib_is_layout_compatible to help the C++ library implement
385       [63]P0466, Layout-compatibility and Pointer-interconvertibility
386       Traits ([64]PR101539)
387     * Memory usage of constraint subsumption has been improved
388       ([65]PR100828)
389     * constinit thread_local variables are optimized better
390       ([66]PR101786)
391     * Support for C++17 std::hardware_destructive_interference_size was
392       added, along with the [67]-Winterference-size warning ([68]git)
393     * Many bugs in the CTAD handling have been fixed ([69]PR101344,
394       [70]PR101883, [71]PR89062, [72]PR101233, [73]PR88252, [74]PR86439,
395       [75]PR98832, [76]PR102933 ...)
396     * Two-stage name lookup for dependent operator expressions has been
397       corrected ([77]PR51577)
398     * Several issues with constrained variable templates have been fixed
399       ([78]PR98486)
400     * The compiler performs less instantiating when doing speculative
401       constant evaluation ([79]git)
402     * Various diagnostic improvements; e.g., a more precise caret
403       location for pointer-to-member expressions
404     * The new -fconstexpr-fp-except flag allows IEC559 floating point
405       exceptions in constant-expressions.
406
407    Runtime Library (libstdc++)
408
409     * Improved experimental C++20 support, including:
410          + std::vector, std::basic_string, std::optional, and
411            std::variant can be used in constexpr functions.
412          + std::make_shared for arrays with default initialization, and
413            std::atomic<std::shared_ptr<T>>.
414          + Layout-compatibility and pointer-interconvertibility traits.
415     * Improved experimental C++23 support, including:
416          + Monadic operations for std::optional.
417          + std::expected
418          + std::move_only_function
419          + <spanstream>
420          + std::basic_string::resize_and_overwrite
421          + std::unique_ptr can be used in constexpr functions.
422          + <stacktrace> (not built by default, requires linking to an
423            extra library).
424          + <stdatomic.h>
425          + std::invoke_r
426          + constexpr std::type_info::operator==
427
428  D
429
430     * New features:
431          + Support for the D programming language has been updated to
432            version 2.100.1 of the language and run-time library. Full
433            changelog for this release and previous releases can be found
434            on the [80]dlang.org website.
435          + On supported targets, the __traits(compiles) expression can
436            now be used to determine whether a target-specific built-in is
437            available without error during CTFE ([81]PR101127).
438          + Functions annotated with pragma(inline, true) are now compiled
439            into every module where they are used from ([82]PR106563).
440          + Partial support for directly importing C99 sources into a D
441            compilation ([83]ImportC) has been added to the language. A
442            notable missing feature is support for preprocessing C
443            imports, which can be worked around by preprocessing all C
444            sources used for importing ahead of time.
445     * New language options:
446          + -fcheck=, enables or disables the code generation of specific
447            run-time contract checks.
448          + -fcheckaction=, controls the run-time behavior on an assert,
449            array bounds check, or final switch contract failure. The
450            default is -fcheckaction=throw.
451          + -fdump-c++-spec=, dumps all compiled extern(C++) declarations
452            as C++ code to the given file. The supplementary option
453            -fdump-c++-spec-verbose turns on emission of comments for
454            ignored declarations in the generated spec.
455          + -fextern-std=, controls which C++ standard extern(C++)
456            declarations are compiled to be compatible with. The default
457            is -fextern-std=c++17.
458          + -fpreview=, added to enable upcoming D language features in
459            the compiler.
460          + -frevert=, added to revert D language changes to support older
461            D codebases that need more time to transition.
462          + -fsave-mixins=, saves mixins expanded at compile-time to a
463            file.
464     * Deprecated and removed features:
465          + The -Wtemplates compiler switch has been removed, as it had
466            been superceded by -ftransition=templates, which more
467            accurately reports on which templates have been instantiated.
468          + The -ftransition=dip25 and -ftransition=dip1000 compiler
469            switches have been renamed to -fpreview=dip25 and
470            -fpreview=dip1000.
471
472  Fortran
473
474     * WG5/N1942, "TS 29113 Further Interoperability of Fortran with C",
475       is now fully supported. In addition to implementing previously
476       missing functionality, such as support for character arguments of
477       length greater than one in functions marked bind(c) and gaps in the
478       handling for assumed-rank arrays, numerous other bugs have been
479       fixed, and an extensive set of new conformance test cases has been
480       added.
481     * GCC 12 now uses OPERATION as the name of the function to the
482       CO_REDUCE intrinsic for the pairwise reduction, thus conforming to
483       the Fortran 2018 standard. Previous versions used OPERATOR which
484       conforms to TS 18508.
485     * On POWER systems which support it, the -mabi=ieeelongdouble option
486       now selects the IEEE 128-bit floating point format for
487       REAL(KIND=16). R16_IBM and R16_IEEE have been added to the
488       -fconvert option, the CONVERT specifier of the OPEN statement and
489       the GFORTRAN_CONVERT_UNIT environment variable.
490
491  Go
492
493     * GCC 12 provides a complete implementation of the Go 1.18 user
494       packages.
495     * Although Go 1.18 includes support for generic programming, that
496       support is not yet available in GCC.
497
498libgccjit
499
500     * The libgccjit API gained 30 new entry points:
501          + 17 new "reflection" entry points for querying functions and
502            types ([84]LIBGCCJIT_ABI_16)
503          + [85]gcc_jit_lvalue_set_tls_model for supporting thread-local
504            variables ([86]LIBGCCJIT_ABI_17)
505          + [87]gcc_jit_lvalue_set_link_section for setting the link
506            section of global variables, analogous to
507            [88]__attribute__((section(".section")))
508            ([89]LIBGCCJIT_ABI_18)
509          + 4 new entry points for initializing global variables and
510            creating constructors for rvalues ([90]LIBGCCJIT_ABI_19)
511          + Support for sized integer types, including 128-bit integers
512            and helper functions for such types ([91]LIBGCCJIT_ABI_20)
513          + [92]gcc_jit_context_new_bitcast for reinterpreting the bits of
514            an rvalue as a different type ([93]LIBGCCJIT_ABI_21)
515          + [94]gcc_jit_lvalue_set_register_name for setting a specific
516            register for a variable ([95]LIBGCCJIT_ABI_22)
517          + [96]gcc_jit_context_set_bool_print_errors_to_stderr
518            ([97]LIBGCCJIT_ABI_23)
519          + 2 new entry points for setting the alignment of a variable
520            ([98]LIBGCCJIT_ABI_24)
521     * libgccjit has gained support for the use of various atomic builtins
522       ([99]PR96066, [100]PR96067)
523     * [101]gcc_jit_context_new_cast is now able to handle truncation and
524       extension between different integer types ([102]PR95498)
525
526New Targets and Target Specific Improvements
527
528  AArch64 & arm
529
530     * Newer revisions of the Arm Architecture are supported as arguments
531       to the -march option: armv8.7-a, armv8.8-a, armv9-a.
532     * The Arm Cortex-A510 CPU is now supported through the cortex-a510
533       argument to the -mcpu and -mtune options.
534     * GCC can now auto-vectorize operations performing sign-differing
535       dot-product operations, taking advantage of instructions in the
536       Advanced SIMD (AArch64/AArch32) and SVE (AArch64) instruction sets.
537
538  AArch64
539
540     * A number of new CPUs are supported through the -mcpu and -mtune
541       options (GCC identifiers in parentheses).
542          + Ampere-1 (ampere1).
543          + Arm Cortex-A710 (cortex-a710).
544          + Arm Cortex-X2 (cortex-x2).
545     * The 64-byte atomic load/store intrinsics to accelerator memory from
546       the [103]2020 Arm Architecture extensions are supported through the
547       +ls64 option extension.
548     * Initial code generation support is supported for hardware
549       instructions used to accelerate the memcpy,memmove and memset
550       standard functions. These instructions can be generated when
551       compiling with the +mopsoption extension.
552     * The ACLE Advanced SIMD intrinsics accessible through the arm_neon.h
553       header have been significantly reimplemented and generate
554       higher-performing code than previous GCC versions.
555     * The option -mtune=neoverse-512tvb is added to tune for Arm Neoverse
556       cores that have a total vector bandwidth of 512 bits. Please refer
557       to the documentation for more details.
558
559  AMD Radeon (GCN)
560
561     * Debug experience with ROCGDB has been improved.
562     * Support for the type __int128_t/integer(kind=16) was added.
563     * For offloading, the limitation of using only one wavefront per
564       compute unit (CU) has been lifted. Up to 40 workgroups per CU and
565       16 wavefronts per workgroup are supported (up to a limit of 40
566       wavefronts in total, per CU). Additionally, the number of used
567       wavefronts and workgroups was tuned for performance.
568
569  arm
570
571     * Support is added for accessing the stack canary value via the TLS
572       register through the -fstack-protector-guard=tls and
573       -mstack-protector-guard-offset= options. This intended for use in
574       Linux kernel development. Please refer to the documentation for
575       more details.
576
577  BPF
578
579     * Support for CO-RE (compile-once, run-everywhere) has been added to
580       the BPF backend. CO-RE allows to compile portable BPF programs that
581       are able to run among different versions of the Linux kernel.
582
583  IA-32/x86-64
584
585     * New ISA extension support for Intel AVX512-FP16 was added.
586       AVX512-FP16 intrinsics are available via the -mavx512fp16 compiler
587       switch.
588     * For both C and C++ the _Float16 type is supported on x86 systems
589       with SSE2 enabled. Without {-mavx512fp16}, all operations will be
590       emulated in software and float instructions.
591     * Mitigation against straight line speculation (SLS) for function
592       return and indirect jump is supported via
593       -mharden-sls=[none|all|return|indirect-jmp].
594     * Add CS prefix to call and jmp to indirect thunk with branch target
595       in r8-r15 registers via -mindirect-branch-cs-prefix.
596     * Always use global offset table (GOT) to access external data and
597       function symbols when the new -mno-direct-extern-access
598       command-line option is specified.
599
600  LoongArch
601
602     * Support for the LoongArch architecture instruction set has been
603       added.
604     * The Loongson CPU codename LA464 and LoongArch 64-bit generic CPU
605       codename loongarch64 are supported through the -march= and -mtune=
606       options (GCC identifiers in parentheses).
607          + Loongson LA464 core (la464).
608          + LoongArch 64-bit generic core (loongarch64).
609
610  MIPS
611
612     * The ABI passing arguments containing zero-width fields (for
613       example, C/C++ zero-width bit-fields, GNU C/C++ zero-length arrays,
614       and GNU C empty structs) has changed. Now a zero-width field will
615       not prevent an aligned 64-bit floating-point field next to it from
616       being passed through FPR. This is compatible with LLVM, but
617       incompatible with previous GCC releases. GCC 12 on MIPS will report
618       such incompatibilities as an inform unless -Wno-psabi is used.
619     * The ABI returning values containing C++17 empty bases has changed.
620       Now an empty base will not prevent an aggregate containing only one
621       or two floating-point fields from being returned through FPR. This
622       is compatible with GCC 6 and earlier, but incompatible with GCC 7
623       through 11. GCC 12 on MIPS will report such incompatibilities as an
624       inform unless -Wno-psabi is used.
625
626  NVPTX
627
628     * The -march flag has been added. The -misa flag is now considered an
629       alias of the -march flag.
630     * Support for PTX ISA target architectures sm_53, sm_70, sm_75 and
631       sm_80 has been added. These can be specified using the -march flag.
632     * The default PTX ISA target architecture has been set back to sm_30,
633       to fix support for sm_30 boards.
634     * The -march-map flag has been added. The -march-map value will be
635       mapped to an valid -march flag value. For instance,
636       -march-map=sm_50 maps to -march=sm_35. This can be used to specify
637       that generated code is to be executed on a board with at least some
638       specific compute capability, without having to know the valid
639       values for the -march flag.
640     * The -mptx flag has been added to specify the PTX ISA version for
641       the generated code; permitted values are 3.1 (matches previous GCC
642       versions), 6.0, 6.3, and 7.0. If not specified, the used version is
643       the minimal version required for -march but at least 6.0.
644     * An mptx-3.1 multilib was added. This allows using older drivers
645       which do not support PTX ISA version 6.0.
646     * The new __PTX_SM__ predefined macro allows code to check the PTX
647       ISA target architecture being targeted by the compiler.
648     * The new __PTX_ISA_VERSION_MAJOR__ and __PTX_ISA_VERSION_MINOR__
649       predefined macros allows code to check the PTX ISA version being
650       targeted by the compiler.
651
652  PowerPC / PowerPC64 / RS6000
653
654     * The internal implementation of Power's target-specific built-in
655       functions has been rewritten to be easier and less error-prone to
656       maintain. Every attempt has been made to ensure that the new
657       behavior matches the old behavior, but inevitably some bugs can be
658       expected. Please report any problems via [104]GCC Bugzilla.
659     * The built-in functions __builtin_get_texasr, __builtin_get_texasru,
660       __builtin_get_tfhar, __builtin_get_tfiar, __builtin_set_texasr,
661       __builtin_set_texasru, __builtin_set_tfhar, and __builtin_set_tfiar
662       now behave as documented in all supported configurations. On prior
663       releases, the arguments and return values of these functions were
664       treated as unsigned long long instead of as unsigned long, when the
665       options -m32 -mpowerpc64 were in effect.
666     * The overloaded built-in functions vec_cntlz_lsbb and vec_cnttz_lsbb
667       now behave as documented. On prior releases, these built-in
668       functions had incorrect semantics on little-endian targets.
669
670  PRU
671
672     * The [105]__regio_symbol variable qualifier has been added. It
673       allows easier access in C programs to the __R30 and __R31 CPU I/O
674       registers.
675
676  RISC-V
677
678     * Default ISA spec version was bump to 20191213, more detail see this
679       [106]announcement
680     * New ISA extension support for zba, zbb, zbc, zbs was added.
681     * New ISA extension support for vector and scalar crypto was added,
682       only support architecture testing marco and -march= parsing.
683     * The option -mtune=thead-c906 is added to tune for T-HEAD c906
684       cores.
685     * libstdc++ no longer attempts to detect built-in atomics.
686       Distributions that have out-of-tree workarounds for -latomic should
687       check their ABIs again.
688
689Operating Systems
690
691Improvements to Static Analyzer
692
693     * The analyzer has gained a
694       [107]-Wanalyzer-use-of-uninitialized-value warning, similar to
695       [108]-Wuninitialized and [109]-Wmaybe-uninitialized, but based on
696       an interprocedural path-sensitive analysis ([110]PR95006).
697       Such warnings are not disabled by the new
698       [111]-ftrivial-auto-var-init (see below), as the latter is
699       considered a mitigation option.
700     * [112]-Wanalyzer-write-to-const and
701       [113]-Wanalyzer-write-to-string-literal will now check for
702       [114]__attribute__ ((access, ....)) on calls to externally-defined
703       functions, and complain about read-only regions pointed to by
704       arguments marked with a write_only or read_write attribute
705       ([115]PR104793).
706     * The analyzer's "taint" mode, activated by
707       [116]-fanalyzer-checker=taint (in addition to [117]-fanalyzer), has
708       gained four new taint-based warnings:
709          + [118]-Wanalyzer-tainted-allocation-size for e.g.
710            attacker-controlled malloc and alloca,
711          + [119]-Wanalyzer-tainted-divisor for detecting where an
712            attacker can inject a divide-by-zero,
713          + [120]-Wanalyzer-tainted-offset for attacker-controlled pointer
714            offsets,
715          + [121]-Wanalyzer-tainted-size for attacker-controlled values
716            being used as a size parameter to calls to memset or to
717            functions marked with [122]__attribute__ ((access, ....)).
718       The existing [123]-Wanalyzer-tainted-array-index has been reworded
719       to talk about "attacker-controlled" rather than "tainted" values,
720       for consistency with the new warnings.
721       A new [124]__attribute__ ((tainted_args)) has been added to the C
722       and C++ front ends, usable on functions, and on function pointer
723       callback fields in structs. The analyzer's taint mode will treat
724       all parameters and buffers pointed to by parameters of such
725       functions as being attacker-controlled, such as for annotating
726       system calls in an operating system kernel as being an "attack
727       surface".
728     * The analyzer now respects [125]__attribute__((const)): it will
729       treat such functions as returning the same value when given the
730       same inputs ([126]PR104434), and as having no side effects
731       ([127]PR104576).
732     * The analyzer is now able to split its analysis into multiple
733       execution paths in places where there isn't a split in the control
734       flow graph. For example, it now handles realloc calls by splitting
735       the execution path into three possible outcomes for the call:
736          + failure, returning NULL
737          + success, growing the buffer in-place without moving it
738          + success, allocating a new buffer, copying the content of the
739            old buffer to it, and freeing the old buffer
740     * The analyzer's interprocedural path exploration logic is now able
741       to track calls through function pointers.
742     * The analyzer now makes the assumption that if we know PTR is
743       non-NULL, then (PTR + OFFSET) is also non-NULL. This isn't strictly
744       true, but eliminates false positives in practice ([128]PR101962).
745     * The analyzer has gained some initial support for inline assembler
746       code. This is extremely limited, and is purely to help suppress
747       false positives when analyzing the Linux kernel, which makes heavy
748       use of inline assembler ([129]PR101570).
749     * The way the analyzer tracks the state of memory along an execution
750       path has been improved in various ways for GCC 12:
751          + An optimization for representing bulk updates to memory (e.g.
752            zero fills) has been removed as it never worked well. In GCC
753            12 it has been replaced with a simpler and more accurate
754            approach, eliminating many false positives ([130]PR95006).
755          + Various optimizations have been added, speeding up the
756            analysis on a particularly problematic source file from 4
757            minutes down to 17 seconds ([131]PR104943, [132]PR104954, and
758            [133]PR104955).
759          + The analyzer now tracks the sizes of dynamically-allocated
760            regions, both on the heap (via malloc etc) and stack (via
761            alloca), though none of the analyzer warnings make use of this
762            yet in GCC 12.
763     * The analyzer's handling of switch statements has been rewritten,
764       fixing various bugs.
765
766Other significant improvements
767
768  Eliminating uninitialized variables
769
770     * GCC can now [134]initialize all stack variables implicitly,
771       including padding. This is intended to eliminate all classes of
772       uninitialized stack variable flaws. Lack of explicit initialization
773       will still warn when [135]-Wuninitialized is active. For best
774       debugging, use of the new command-line option
775       [136]-ftrivial-auto-var-init=pattern can be used to fill variables
776       with a repeated 0xFE pattern, which tends to illuminate many bugs
777       (e.g. pointers receive invalid addresses, sizes and indices are
778       very large). For best production results, the new command-line
779       option [137]-ftrivial-auto-var-init=zero can be used to fill
780       variables with 0x00, which tends to provide a safer state for bugs
781       (e.g. pointers are NULL, strings are NUL filled, and sizes and
782       indices are 0).
783
784  Debugging formats
785
786     * GCC can now generate debugging information in [138]CTF, a
787       lightweight debugging format that provides information about C
788       types and the association between functions and data symbols and
789       types. This format is designed to be embedded in ELF files and to
790       be very compact and simple. A new command-line option -gctf enables
791       the generation of CTF.
792     * GCC can now generate debugging information in BTF. This is a
793       debugging format mainly used in BPF programs and the Linux kernel.
794       The compiler can generate BTF for any target, when enabled with the
795       command-line option -gbtf
796
797GCC 12.1
798
799   This is the [139]list of problem reports (PRs) from GCC's bug tracking
800   system that are known to be fixed in the 12.1 release. This list might
801   not be complete (that is, it is possible that some PRs that have been
802   fixed are not listed here).
803
804GCC 12.2
805
806   This is the [140]list of problem reports (PRs) from GCC's bug tracking
807   system that are known to be fixed in the 12.2 release. This list might
808   not be complete (that is, it is possible that some PRs that have been
809   fixed are not listed here).
810
811  Target Specific Changes
812
813    LoongArch
814
815     * The default setting of -m[check|no-check]-zero-division is changed
816       for optimized code. Now -mno-check-zero-division is the default for
817       all optimization levels but -O0 and -Og. The old behavior can be
818       obtained by explicitly passing -mcheck-zero-division to GCC.
819
820GCC 12.3
821
822  Target Specific Changes
823
824  x86-64
825
826     * GCC now supports AMD CPUs based on the znver4 core via
827       -march=znver4. The switch makes GCC consider using 512 bit vectors
828       when auto-vectorizing.
829
830   This is the [141]list of problem reports (PRs) from GCC's bug tracking
831   system that are known to be fixed in the 12.3 release. This list might
832   not be complete (that is, it is possible that some PRs that have been
833   fixed are not listed here).
834
835GCC 12.4
836
837   Note: GCC 12.4 has not been released yet, so this section is a
838   work-in-progress.
839
840   This is the [142]list of problem reports (PRs) from GCC's bug tracking
841   system that are known to be fixed in the 12.4 release. This list might
842   not be complete (that is, it is possible that some PRs that have been
843   fixed are not listed here).
844
845
846    For questions related to the use of GCC, please consult these web
847    pages and the [143]GCC manuals. If that fails, the
848    [144]gcc-help@gcc.gnu.org mailing list might help. Comments on these
849    web pages and the development of GCC are welcome on our developer
850    list at [145]gcc@gcc.gnu.org. All of [146]our lists have public
851    archives.
852
853   Copyright (C) [147]Free Software Foundation, Inc. Verbatim copying and
854   distribution of this entire article is permitted in any medium,
855   provided this notice is preserved.
856
857   These pages are [148]maintained by the GCC team. Last modified
858   2023-05-08.
859
860References
861
862   1. http://gcc.gnu.org/gcc-12/porting_to.html
863   2. http://gcc.gnu.org/onlinedocs/index.html#current
864   3. https://gcc.gnu.org/PR102024
865   4. https://gcc.gnu.org/PR42217
866   5. http://gcc.gnu.org/gcc-12/changes.html#mips_zero_width_fields
867   6. https://gcc.gnu.org/install/specific.html
868   7. https://clang.llvm.org/docs/ShadowCallStack.html
869   8. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Instrumentation-Options.html#index-fsanitize_003dshadow-call-stack
870   9. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/libgomp/OpenMP-Implementation-Status.html
871  10. https://www.openacc.org/
872  11. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-12
873  12. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/libgomp/Enabling-OpenACC.html
874  13. http://gcc.gnu.org/gcc-12/changes.html#amdgcn
875  14. http://gcc.gnu.org/gcc-12/changes.html#nvptx
876  15. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-param
877  16. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wopenacc-parallelism
878  17. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/C-Dialect-Options.html#index-foffload-options
879  18. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/C-Dialect-Options.html#index-foffload
880  19. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gnat_rm/Security-Hardening-Features.html#Security-Hardening-Features
881  20. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wbidi-chars
882  21. https://gcc.gnu.org/PR103026
883  22. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Warray-compare
884  23. https://gcc.gnu.org/PR97573
885  24. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wattributes
886  25. https://gcc.gnu.org/PR101940
887  26. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wformat
888  27. https://wg21.link/p1938
889  28. https://gcc.gnu.org/PR100974
890  29. https://wg21.link/p0849
891  30. https://gcc.gnu.org/PR103049
892  31. https://wg21.link/p2242
893  32. https://gcc.gnu.org/PR102612
894  33. https://wg21.link/p2334
895  34. https://gcc.gnu.org/PR102616
896  35. https://wg21.link/p2360
897  36. https://gcc.gnu.org/PR102617
898  37. https://wg21.link/p2128
899  38. https://wg21.link/cwg2397
900  39. https://gcc.gnu.org/PR100975
901  40. https://wg21.link/cwg960
902  41. https://wg21.link/cwg1227
903  42. https://wg21.link/cwg1315
904  43. https://wg21.link/cwg2082
905  44. https://wg21.link/cwg2351
906  45. https://wg21.link/cwg2374
907  46. https://wg21.link/cwg2397
908  47. https://wg21.link/cwg2446
909  48. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=87c2080b
910  49. https://gcc.gnu.org/PR79501
911  50. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wuninitialized
912  51. https://gcc.gnu.org/PR19808
913  52. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wint-in-bool-context
914  53. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=3a2b12bc
915  54. https://gcc.gnu.org/PR99032
916  55. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wmissing-requires
917  56. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=e18e56c7
918  57. https://gcc.gnu.org/PR100995
919  58. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Waddress
920  59. https://gcc.gnu.org/PR102103
921  60. https://gcc.gnu.org/PR99701
922  61. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=3ead06c1
923  62. https://gcc.gnu.org/PR102753
924  63. https://wg21.link/p0466
925  64. https://gcc.gnu.org/PR101539
926  65. https://gcc.gnu.org/PR100828
927  66. https://gcc.gnu.org/PR101786
928  67. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Winterference-size
929  68. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=76b75018
930  69. https://gcc.gnu.org/PR101344
931  70. https://gcc.gnu.org/PR101883
932  71. https://gcc.gnu.org/PR89062
933  72. https://gcc.gnu.org/PR101233
934  73. https://gcc.gnu.org/PR88252
935  74. https://gcc.gnu.org/PR86439
936  75. https://gcc.gnu.org/PR98832
937  76. https://gcc.gnu.org/PR102933
938  77. https://gcc.gnu.org/PR51577
939  78. https://gcc.gnu.org/PR98486
940  79. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=1595fe44
941  80. https://dlang.org/changelog/2.100.1.html
942  81. https://gcc.gnu.org/PR101127
943  82. https://gcc.gnu.org/PR106563
944  83. https://dlang.org/spec/importc.html
945  84. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-16
946  85. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_lvalue_set_tls_model
947  86. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-17
948  87. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_lvalue_set_link_section
949  88. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Variable-Attributes.html#index-section-variable-attribute
950  89. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-18
951  90. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-19
952  91. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-20
953  92. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_context_new_bitcast
954  93. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-21
955  94. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_lvalue_set_register_name
956  95. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-22
957  96. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/contexts.html#c.gcc_jit_context_set_bool_print_errors_to_stderr
958  97. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-23
959  98. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/compatibility.html#libgccjit-abi-24
960  99. https://gcc.gnu.org/PR96066
961 100. https://gcc.gnu.org/PR96067
962 101. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/jit/topics/expressions.html#c.gcc_jit_context_new_cast
963 102. https://gcc.gnu.org/PR95498
964 103. https://community.arm.com/arm-community-blogs/b/architectures-and-processors-blog/posts/arm-a-profile-architecture-developments-2020
965 104. https://gcc.gnu.org/bugzilla/
966 105. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Named-Address-Spaces.html#PRU-Named-Address-Spaces
967 106. https://groups.google.com/a/groups.riscv.org/g/sw-dev/c/aE1ZeHHCYf4
968 107. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-use-of-uninitialized-value
969 108. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wuninitialized
970 109. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wmaybe-uninitialized
971 110. https://gcc.gnu.org/PR95006
972 111. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init
973 112. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-const
974 113. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-string-literal
975 114. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html
976 115. https://gcc.gnu.org/PR104793
977 116. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer-checker
978 117. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-fanalyzer
979 118. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-allocation-size
980 119. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-divisor
981 120. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-offset
982 121. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-size
983 122. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html
984 123. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-tainted-array-index
985 124. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html#index-tainted_005fargs-function-attribute
986 125. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Common-Function-Attributes.html#index-const-function-attribute
987 126. https://gcc.gnu.org/PR104434
988 127. https://gcc.gnu.org/PR104576
989 128. https://gcc.gnu.org/PR101962
990 129. https://gcc.gnu.org/PR101570
991 130. https://gcc.gnu.org/PR95006
992 131. https://gcc.gnu.org/PR104943
993 132. https://gcc.gnu.org/PR104954
994 133. https://gcc.gnu.org/PR104955
995 134. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init
996 135. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Warning-Options.html#index-Wuninitialized
997 136. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init
998 137. https://gcc.gnu.org/onlinedocs/gcc-12.1.0/gcc/Optimize-Options.html#index-ftrivial-auto-var-init
999 138. https://ctfstd.org/
1000 139. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.0
1001 140. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.2
1002 141. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.3
1003 142. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=12.4
1004 143. https://gcc.gnu.org/onlinedocs/
1005 144. mailto:gcc-help@gcc.gnu.org
1006 145. mailto:gcc@gcc.gnu.org
1007 146. https://gcc.gnu.org/lists.html
1008 147. https://www.fsf.org/
1009 148. https://gcc.gnu.org/about.html
1010======================================================================
1011http://gcc.gnu.org/gcc-11/index.html
1012
1013                             GCC 11 Release Series
1014
1015   April 21, 2022
1016
1017   The GCC developers are pleased to announce the release of GCC 11.3.
1018
1019   This release is a bug-fix release, containing fixes for regressions in
1020   GCC 11.2 relative to previous releases of GCC.
1021
1022Release History
1023
1024   GCC 11.3
1025          April 21, 2022 ([1]changes, [2]documentation)
1026
1027   GCC 11.2
1028          July 28, 2021 ([3]changes, [4]documentation)
1029
1030   GCC 11.1
1031          April 27, 2021 ([5]changes, [6]documentation)
1032
1033References and Acknowledgements
1034
1035   GCC used to stand for the GNU C Compiler, but since the compiler
1036   supports several other languages aside from C, it now stands for the
1037   GNU Compiler Collection.
1038
1039   A list of [7]successful builds is updated as new information becomes
1040   available.
1041
1042   The GCC developers would like to thank the numerous people that have
1043   contributed new features, improvements, bug fixes, and other changes as
1044   well as test results to GCC. This [8]amazing group of volunteers is
1045   what makes GCC successful.
1046
1047   For additional information about GCC please refer to the [9]GCC project
1048   web site or contact the [10]GCC development mailing list.
1049
1050   To obtain GCC please use [11]our mirror sites or [12]our version
1051   control system.
1052
1053
1054    For questions related to the use of GCC, please consult these web
1055    pages and the [13]GCC manuals. If that fails, the
1056    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1057    web pages and the development of GCC are welcome on our developer
1058    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
1059    archives.
1060
1061   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
1062   distribution of this entire article is permitted in any medium,
1063   provided this notice is preserved.
1064
1065   These pages are [18]maintained by the GCC team. Last modified
1066   2022-10-26.
1067
1068References
1069
1070   1. http://gcc.gnu.org/gcc-11/changes.html
1071   2. http://gcc.gnu.org/onlinedocs/11.3.0/
1072   3. http://gcc.gnu.org/gcc-11/changes.html
1073   4. http://gcc.gnu.org/onlinedocs/11.2.0/
1074   5. http://gcc.gnu.org/gcc-11/changes.html
1075   6. http://gcc.gnu.org/onlinedocs/11.1.0/
1076   7. http://gcc.gnu.org/gcc-11/buildstat.html
1077   8. http://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Contributors.html
1078   9. http://gcc.gnu.org/index.html
1079  10. mailto:gcc@gcc.gnu.org
1080  11. http://gcc.gnu.org/mirrors.html
1081  12. http://gcc.gnu.org/git.html
1082  13. https://gcc.gnu.org/onlinedocs/
1083  14. mailto:gcc-help@gcc.gnu.org
1084  15. mailto:gcc@gcc.gnu.org
1085  16. https://gcc.gnu.org/lists.html
1086  17. https://www.fsf.org/
1087  18. https://gcc.gnu.org/about.html
1088======================================================================
1089http://gcc.gnu.org/gcc-11/changes.html
1090
1091                             GCC 11 Release Series
1092                        Changes, New Features, and Fixes
1093
1094   This page is a "brief" summary of some of the huge number of
1095   improvements in GCC 11. You may also want to check out our [1]Porting
1096   to GCC 11 page and the [2]full GCC documentation.
1097
1098Caveats
1099
1100     * The default mode for C++ is now -std=gnu++17 instead of
1101       -std=gnu++14. Note that [3]C++17 changes to template template
1102       parameter matching can be disabled independently of other features
1103       with -fno-new-ttp-matching.
1104     * When building GCC itself, the host compiler must now support C++11,
1105       rather than C++98. In particular bootstrapping GCC 11 using an
1106       older version of GCC requires a binary of GCC 4.8 or later, rather
1107       than of GCC 3.4 or later as was the case for bootstrapping GCC 10.
1108     * Naming and location of auxiliary and dump output files changed. If
1109       you compile multiple input files in a single command, if you enable
1110       Link Time Optimization, or if you use -dumpbase, -dumpdir,
1111       -save-temps=*, and you expect any file other than the primary
1112       output file(s) to be created as a side effect, watch out for
1113       improvements and a few surprises. See [4]the patch, particularly
1114       its textual description, for more details about the changes.
1115     * -gsplit-dwarf no longer enables debug info generation on its own
1116       but requires a separate -g for this.
1117     * The libstdc++ configure option --enable-cheaders=c_std is
1118       deprecated and will be removed in a future release. It should be
1119       possible to use --enable-cheaders=c_global (the default) with no
1120       change in behaviour.
1121     * The front end for compiling BRIG format of Heterogeneous System
1122       Architecture Intermediate Language (HSAIL) has been deprecated and
1123       will likely be removed in a future release.
1124     * Some short options of the gcov tool have been renamed: -i to -j and
1125       -j to -H.
1126
1127General Improvements
1128
1129     * [5]ThreadSanitizer improvements to support alternative runtimes and
1130       environments. The [6]Linux Kernel Concurrency Sanitizer (KCSAN) is
1131       now supported.
1132          + Add --param tsan-distinguish-volatile to optionally emit
1133            instrumentation distinguishing volatile accesses.
1134          + Add --param tsan-instrument-func-entry-exit to optionally
1135            control if function entries and exits should be instrumented.
1136     * In previous releases of GCC, the "column numbers" emitted in
1137       diagnostics were actually a count of bytes from the start of the
1138       source line. This could be problematic, both because of:
1139          + multibyte characters (requiring more than one byte to encode),
1140            and
1141          + multicolumn characters (requiring more than one column to
1142            display in a monospace font)
1143       For example, the character p ("GREEK SMALL LETTER PI (U+03C0)")
1144       occupies one column, and its UTF-8 encoding requires two bytes; the
1145       character &#x1F642; ("SLIGHTLY SMILING FACE (U+1F642)") occupies
1146       two columns, and its UTF-8 encoding requires four bytes.
1147       In GCC 11 the column numbers default to being column numbers,
1148       respecting multi-column characters. The old behavior can be
1149       restored using a new option [7]-fdiagnostics-column-unit=byte.
1150       There is also a new option [8]-fdiagnostics-column-origin=,
1151       allowing the pre-existing default of the left-hand column being
1152       column 1 to be overridden if desired (e.g. for 0-based columns).
1153       The output of [9]-fdiagnostics-format=json has been extended to
1154       supply both byte counts and column numbers for all source
1155       locations.
1156       Additionally, in previous releases of GCC, tab characters in the
1157       source would be emitted verbatim when quoting source code, but be
1158       prefixed with whitespace or line number information, leading to
1159       misalignments in the resulting output when compared with the actual
1160       source. Tab characters are now printed as an appropriate number of
1161       spaces, using the [10]-ftabstop option (which defaults to 8 spaces
1162       per tab stop).
1163     * Introduce [11]Hardware-assisted AddressSanitizer support. This
1164       sanitizer currently only works for the AArch64 target. It helps
1165       debug address problems similarly to [12]AddressSanitizer but is
1166       based on partial hardware assistance and provides probabilistic
1167       protection to use less RAM at run time. [13]Hardware-assisted
1168       AddressSanitizer is not production-ready for user space, and is
1169       provided mainly for use compiling the Linux Kernel.
1170       To use this sanitizer the command line arguments are:
1171          + -fsanitize=hwaddress to instrument userspace code.
1172          + -fsanitize=kernel-hwaddress to instrument kernel code.
1173     * For targets that produce DWARF debugging information GCC now
1174       defaults to [14]DWARF version 5 (with the exception of VxWorks and
1175       Darwin/Mac OS X which default to version 2 and AIX which defaults
1176       to version 4). This can produce up to 25% more compact debug
1177       information compared to earlier versions.
1178       To take full advantage of DWARF version 5 GCC needs to be built
1179       against binutils version 2.35.2 or higher. When GCC is built
1180       against earlier versions of binutils GCC will still emit DWARF
1181       version 5 for most debuginfo data, but will generate version 4
1182       debug line tables (even when explicitly given -gdwarf-5).
1183       The following debug information consumers can process DWARF version
1184       5:
1185          + GDB 8.0, or higher
1186          + valgrind 3.17.0
1187          + elfutils 0.172, or higher (for use with systemtap,
1188            dwarves/pahole, perf and libabigail)
1189          + dwz 0.14
1190       Programs embedding libbacktrace are urged to upgrade to the version
1191       shipping with GCC 11.
1192       To make GCC 11 generate an older DWARF version use -g together with
1193       -gdwarf-2, -gdwarf-3 or -gdwarf-4.
1194     * Vectorizer improvements:
1195          + The straight-line code vectorizer now considers the whole
1196            function when vectorizing and can handle opportunities
1197            crossing CFG merges and backedges.
1198     * A series of conditional expressions that compare the same variable
1199       can be transformed into a switch statement if each of them contains
1200       a comparison expression. Example:
1201        int IsHTMLWhitespace(int aChar) {
1202          return aChar == 0x0009 || aChar == 0x000A ||
1203                 aChar == 0x000C || aChar == 0x000D ||
1204                 aChar == 0x0020;
1205        }
1206
1207       This statement can be transformed into a switch statement and then
1208       expanded into a bit-test.
1209     * New command-line options:
1210          + [15]-fbit-tests, enabled by default, can be used to enable or
1211            disable switch expansion using bit-tests.
1212     * Inter-procedural optimization improvements:
1213          + A new IPA-modref pass was added to track side effects of
1214            function calls and improve precision of points-to-analysis.
1215            The pass can be controlled by the [16]-fipa-modref option.
1216          + The identical code folding pass (controlled by [17]-fipa-icf)
1217            was significantly improved to increase the number of unified
1218            functions and to reduce compile-time memory use.
1219          + IPA-CP (Interprocedural constant propagation) heuristics
1220            improved its estimation of potential usefulness of known loop
1221            bounds and strides by taking the estimated frequency of these
1222            loops into account.
1223     * Link-time optimization improvements:
1224          + The LTO bytecode format was optimized for smaller object files
1225            and faster streaming.
1226          + Memory allocation of the linking stage was improved to reduce
1227            peak memory use.
1228     * Profile driven optimization improvements:
1229          + Using [18]-fprofile-values, was improved by tracking more
1230            target values for e.g. indirect calls.
1231          + GCOV data file format outputs smaller files by representing
1232            zero counters in a more compact way.
1233
1234New Languages and Language specific improvements
1235
1236     * GCC 11 adds support for non-rectangular loop nests in OpenMP
1237       constructs and the allocator routines of [19]OpenMP 5.0, including
1238       initial allocate clause support in C/C++. The OMP_TARGET_OFFLOAD
1239       environment variable and the active-levels routines are now
1240       supported. For C/C++, the declare variant and map support has been
1241       extended. For Fortran, OpenMP 4.5 is now fully supported and OpenMP
1242       5.0 support has been extended, including the following features
1243       which were before only available in C and C++: order(concurrent),
1244       device_type, memorder-clauses for flush, lastprivate with
1245       conditional modifier, atomic construct and reduction clause
1246       extensions of OpenMP 5.0, if clause with simd and cancel modifiers,
1247       target data without map clause, and limited support for the
1248       requires construct.
1249     * Version 2.6 of the [20]OpenACC specification continues to be
1250       maintained and improved in the C, C++ and Fortran compilers. See
1251       the [21]implementation status section on the OpenACC wiki page and
1252       the [22]run-time library documentation for further information.
1253
1254  C family
1255
1256     * New attributes:
1257          + The [23]no_stack_protector attribute has been added to mark
1258            functions which should not be instrumented with stack
1259            protection (-fstack-protector).
1260          + The existing [24]malloc attribute has been extended so that it
1261            can be used to identify allocator/deallocator API pairs. A
1262            pair of new [25]-Wmismatched-dealloc and
1263            [26]-Wmismatched-new-delete warnings will complain about
1264            mismatched calls, and [27]-Wfree-nonheap-object about
1265            deallocation calls with pointers not obtained from allocation
1266            functions. Additionally, the static analyzer will use these
1267            attributes when checking for leaks, double-frees,
1268            use-after-frees, and similar issues.
1269     * New warnings:
1270          + [28]-Wmismatched-dealloc, enabled by default, warns about
1271            calls to deallocation functions with pointers returned from
1272            mismatched allocation functions.
1273          + [29]-Wsizeof-array-div, enabled by -Wall, warns about
1274            divisions of two sizeof operators when the first one is
1275            applied to an array and the divisor does not equal the size of
1276            the array element.
1277          + [30]-Wstringop-overread, enabled by default, warns about calls
1278            to string functions reading past the end of the arrays passed
1279            to them as arguments. In prior GCC releases most instances of
1280            his warning are diagnosed by -Wstringop-overflow.
1281          + [31]-Wtsan, enabled by default, warns about unsupported
1282            features in ThreadSanitizer (currently
1283            std::atomic_thread_fence).
1284     * Enhancements to existing warnings:
1285          + [32]-Wfree-nonheap-object detects many more instances of calls
1286            to deallocation functions with pointers not returned from a
1287            dynamic memory allocation function.
1288          + [33]-Wmaybe-uninitialized diagnoses passing pointers or
1289            references to uninitialized memory to functions taking
1290            const-qualified arguments.
1291          + [34]-Wuninitialized detects reads from uninitialized
1292            dynamically allocated memory.
1293     * For ELF targets that support the GNU or FreeBSD OSABIs, the used
1294       attribute will now save the symbol declaration it is applied to
1295       from linker garbage collection.
1296       To support this behavior, used symbols that have not been placed in
1297       specific sections (e.g. with the section attribute, or the
1298       -f{function,data}-sections options) will be placed in new, unique
1299       sections.
1300       This functionality requires Binutils version 2.36 or later.
1301
1302  C
1303
1304     * Several new features from the upcoming C2X revision of the ISO C
1305       standard are supported with -std=c2x and -std=gnu2x. Some of these
1306       features are also supported as extensions when compiling for older
1307       language versions. In addition to the features listed, some
1308       features previously supported as extensions and now added to the C
1309       standard are enabled by default in C2X mode and not diagnosed with
1310       -std=c2x -Wpedantic.
1311          + The BOOL_MAX and BOOL_WIDTH macros are provided in <limits.h>.
1312          + As in C++, function definitions no longer need to give names
1313            for unused function parameters.
1314          + The expansions of the true and false macros in <stdbool.h>
1315            have changed so that they have type bool.
1316          + The [[nodiscard]] standard attribute is now supported.
1317          + The __has_c_attribute preprocessor operator is now supported.
1318          + Macros INFINITY, NAN, FLT_SNAN, DBL_SNAN, LDBL_SNAN,
1319            DEC_INFINITY, DEC_NAN, and corresponding signaling NaN macros
1320            for _FloatN, _FloatNx and _DecimalN types, are provided in
1321            <float.h>. There are also corresponding built-in functions
1322            __builtin_nansdN for decimal signaling NaNs.
1323          + Macros FLT_IS_IEC_60559, DBL_IS_IEC_60559 and
1324            LDBL_IS_IEC_60559 are provided in <float.h>.
1325          + The feature test macro __STDC_WANT_IEC_60559_EXT__ is
1326            supported by <float.h>.
1327          + Labels may appear before declarations and at the end of a
1328            compound statement.
1329     * New warnings:
1330          + [35]-Warray-parameter, enabled by -Wall, warns about
1331            redeclarations of functions with ordinary array arguments
1332            declared using inconsistent forms. The warning also enables
1333            the detection of the likely out of bounds accesses in calls to
1334            such functions with smaller arrays.
1335          + [36]-Wvla-parameter, enabled by -Wall, warns redeclarations of
1336            functions with variable length array arguments declared using
1337            inconsistent forms or with mismatched bounds. The warning also
1338            enables the detection of the likely out of bounds accesses in
1339            calls to such functions with smaller arrays.
1340
1341  C++
1342
1343     * The default mode has been changed to -std=gnu++17.
1344     * Several C++20 features have been implemented:
1345          + the compiler now supports consteval virtual functions
1346          + P2082R1, Fixing CTAD for aggregates
1347          + P0593R6, Pseudo-destructors end object lifetimes
1348          + P1907R1, Inconsistencies with non-type template parameters
1349            (complete implementation)
1350          + P1975R0, Fixing the wording of parenthesized
1351            aggregate-initialization
1352          + P1009R2, Array size deduction in new-expressions
1353          + P1099R5, using enum
1354          + Modules, Requires -fmodules-ts and some aspects are
1355            incomplete. Refer to [37]C++ 20 Status
1356     * The C++ front end has experimental support for some of the upcoming
1357       C++23 draft features with the -std=c++23, -std=gnu++23, -std=c++2b
1358       or -std=gnu++2b flags, including
1359          + P0330R8, Literal Suffix for (signed) size_t.
1360       For a full list of new features, see [38]the C++ status page.
1361     * Several C++ Defect Reports have been resolved, e.g.:
1362          + DR 625, Use of auto as a template-argument
1363          + DR 1512, Pointer comparison vs qualification conversions
1364          + DR 1722, Should lambda to function pointer conversion function
1365            be noexcept?
1366          + DR 1914, Duplicate standard attributes
1367          + DR 2032, Default template-arguments of variable templates
1368          + DR 2289, Uniqueness of decomposition declaration names
1369          + DR 2237, Can a template-id name a constructor?
1370          + DR 2303, Partial ordering and recursive variadic inheritance
1371          + DR 2369, Ordering between constraints and substitution
1372          + DR 2450, braced-init-list as a template-argument
1373     * G++ now performs better access checking in templates ([39]PR41437).
1374     * reinterpret_casts in constexpr evaluation are now checked more
1375       completely ([40]PR95307).
1376     * The diagnostic for static_assert has been improved: the compiler
1377       now shows the expression including its template arguments (if there
1378       were any), and can point to the failing clause if the condition
1379       comprised of any logical AND operators ([41]PR97518).
1380     * New warnings:
1381          + [42]-Wctad-maybe-unsupported, disabled by default, warns about
1382            performing class template argument deduction on a type with no
1383            deduction guides.
1384          + [43]-Wrange-loop-construct, enabled by -Wall, warns when a
1385            range-based for-loop is creating unnecessary and expensive
1386            copies.
1387          + [44]-Wdeprecated-enum-enum-conversion, enabled by default in
1388            C++20, warns about deprecated arithmetic conversions on
1389            operands of enumeration types, as outlined in
1390            [depr.arith.conv.enum].
1391          + [45]-Wdeprecated-enum-float-conversion, enabled by default in
1392            C++20, warns about deprecated arithmetic conversions on
1393            operands where one is of enumeration type and the other is of
1394            a floating-point type, as outlined in [depr.arith.conv.enum].
1395          + [46]-Wmismatched-new-delete, enabled by -Wall, warns about
1396            calls to C++ operator delete with pointers returned from
1397            mismatched forms of operator new or from other mismatched
1398            allocation functions.
1399          + [47]-Wvexing-parse, enabled by default, warns about the most
1400            vexing parse rule: the cases when a declaration looks like a
1401            variable definition, but the C++ language requires it to be
1402            interpreted as a function declaration.
1403     * Enhancements to existing warnings:
1404          + [48]-Wnonnull considers the implicit this argument of every
1405            C++ nonstatic member function to have been implicitly declared
1406            with attribute nonnull and triggers warnings for calls where
1407            the pointer is null.
1408
1409    Runtime Library (libstdc++)
1410
1411     * Improved C++17 support, including:
1412          + std::from_chars and std::to_chars for floating-point types.
1413     * Improved experimental C++20 support, including:
1414          + Calendar additions to <chrono>. Thanks to Cassio Neri for
1415            optimizations.
1416          + std::bit_cast
1417          + std::source_location
1418          + Atomic wait and notify operations.
1419          + <barrier>, <latch>, and <semaphore>
1420          + <syncstream>
1421          + Efficient access to basic_stringbuf's buffer.
1422          + Heterogeneous lookup in unordered containers.
1423     * Experimental C++23 support, including:
1424          + contains member functions for strings, thanks to Paul Fee.
1425          + std::to_underlying, std::is_scoped_enum
1426     * Experimental support for Data-Parallel Types (simd) from the
1427       Parallelism 2 TS, thanks to Matthias Kretz.
1428     * Faster std::uniform_int_distribution, thanks to Daniel Lemire.
1429
1430  D
1431
1432     * New features:
1433          + A new bottom type typeof(*null) has been added to represent
1434            run-time errors and non-terminating functions. This also
1435            introduces a new standard alias for the type named noreturn,
1436            and is implicitly imported into every module.
1437          + Printf-like and scanf-like functions are now detected by
1438            prefixing them with pragma(printf) for printf-like functions
1439            or pragma(scanf) for scanf-like functions.
1440          + The __traits() expression now supports the extensions
1441            isDeprecated, isDisabled, isFuture, isModule, isPackage,
1442            child, isReturnOnStack, isZeroInit, getTargetInfo,
1443            getLocation, hasPostblit, isCopyable, getVisibility, and
1444            totype.
1445          + An expression-based contract syntax has been added to the
1446            language.
1447          + Function literals can now return a value by reference with the
1448            ref keyword.
1449          + A new syntax is available to declare aliases to function types
1450            using the alias syntax based on the assignment operator.
1451          + New types __c_complex_float, __c_complex_double,
1452            __c_complex_real, and __c_wchar_t have been added for
1453            interfacing with C and C++ code, and are available from the
1454            core.stdc.config module.
1455          + User-defined attributes can now be used to annotate enum
1456            members, alias declarations, and function parameters.
1457          + Templates alias parameters can now be instantiated with basic
1458            types such as int or void function().
1459          + The mixin construct can now be used as types in the form
1460            mixin(string) var.
1461          + The mixin construct can now take an argument list, same as
1462            pragma(msg).
1463     * New intrinsics:
1464          + Bitwise rotate intrinsics core.bitop.rol and core.bitop.ror
1465            have been added.
1466          + Byte swap intrinsic core.bitop.byteswap for swapping bytes in
1467            a 2-byte ushort has been added.
1468          + Math intrinsics available from core.math now have overloads
1469            for float and double types.
1470          + Volatile intrinsics core.volatile.volatileLoad and
1471            core.volatile.volatileStore have been moved from the
1472            core.bitop module.
1473     * New attributes:
1474          + The following GCC attributes are now recognized and available
1475            from the gcc.attributes module with short-hand aliases for
1476            convenience:
1477               o @attribute("alloc_size", arguments) or
1478                 @alloc_size(arguments).
1479               o @attribute("always_inline") or @always_inline.
1480               o @attribute("used") or @used.
1481               o @attribute("optimize", arguments) or
1482                 @optimize(arguments).
1483               o @attribute("cold") or @cold.
1484               o @attribute("noplt") or @noplt.
1485               o @attribute("target_clones", arguments) or
1486                 @target_clones(arguments).
1487               o @attribute("no_icf") or @no_icf.
1488               o @attribute("noipa") or @noipa.
1489               o @attribute("symver", arguments) or @symver(arguments).
1490          + New aliases have been added to gcc.attributes for
1491            compatibility with ldc.attributes.
1492               o The @allocSize(arguments) attribute is the same as
1493                 @alloc_size(arguments), but uses a 0-based index for
1494                 function arguments.
1495               o The @assumeUsed attribute is an alias for
1496                 @attribute("used").
1497               o The @fastmath attribute is an alias for
1498                 @optimize("Ofast").
1499               o The @naked attribute is an alias for @attribute("naked").
1500                 This attribute may not be available on all targets.
1501               o The @restrict attribute has been added to specify that a
1502                 function parameter is to be restrict-qualified in the C99
1503                 sense of the term.
1504               o The @optStrategy(strategy) attribute is an alias for
1505                 @optimize("O0") when the strategy is "none", otherwise
1506                 @optimize("Os") for the "optsize" and "minsize"
1507                 strategies.
1508               o The @polly attribute is an alias for
1509                 @optimize("loop-parallelize-all").
1510               o The @section(name) attribute is an alias for
1511                 @attribute("section", name).
1512               o The @target(arguments) attribute is an alias for
1513                 attribute("target", arguments).
1514               o The @weak attribute is an alias for @attribute("weak").
1515     * New language options:
1516          + -fweak-templates, added to control whether declarations that
1517            can be defined in multiple objects should be emitted as weak
1518            symbols. The default is to emit all symbols with extern
1519            linkage as weak, unless the target lacks support for weak
1520            symbols.
1521          + -Wdeprecated, this option is now enabled by default.
1522          + -Wextra, this option now turns on all warnings that are not
1523            part of the core D language front-end - -Waddress,
1524            -Wcast-result, -Wunknown-pragmas.
1525          + -Wvarargs, added to turn on warnings about questionable usage
1526            of the va_start intrinsic.
1527     * Deprecated and removed features:
1528          + Compiler-recognized attributes are now made available from the
1529            gcc.attributes module, the former module gcc.attribute has
1530            been deprecated and will be removed in a future release.
1531          + The @attribute("alias") attribute has been removed, as it had
1532            been superseded by pragma(mangle).
1533          + The @attribute("forceinline") attribute has been removed and
1534            renamed to @always_inline.
1535          + __vector types that are not supported in hardware are now
1536            rejected at compile-time. Previously all vector types were
1537            accepted by the compiler and emulated when target support was
1538            absent.
1539          + The extern(Pascal) linkage attribute has been removed.
1540          + The deprecation phase for -ftransition=import and
1541            -ftransition=checkimports is finished. These switches no
1542            longer have an effect and are now removed. Symbols that are
1543            not visible in a particular scope will no longer be found by
1544            the compiler.
1545          + It is now an error to use private variables selectively
1546            imported from other modules. Due to a bug, some imported
1547            private members were visible from other modules, violating the
1548            specification.
1549          + The -fweak compiler switch has been removed, as it existed
1550            only for testing.
1551
1552  Fortran
1553
1554     * Added DEPRECATED to !GCC$'s attributes directive.
1555
1556  Go
1557
1558     * GCC 11 provides a complete implementation of the Go 1.16.3 user
1559       packages.
1560
1561libgccjit
1562
1563     * libgccjit was marked as merely "Alpha" quality when [49]originally
1564       added in GCC 5. Given that we have maintained [50]API and ABI
1565       compatibility since then and it is in use by various projects, we
1566       have removed that caveat.
1567     * libgccjit can now be built for MinGW
1568     * The libgccjit API gained 10 new entry points:
1569          + [51]gcc_jit_global_set_initializer
1570          + 9 entrypoints for [52]directly embedding asm statements into a
1571            compile, analogous to inline asm in the C front end
1572
1573New Targets and Target Specific Improvements
1574
1575  AArch64 & arm
1576
1577     * A number of new CPUs are supported through arguments to the -mcpu
1578       and -mtune options in both the arm and aarch64 backends (GCC
1579       identifiers in parentheses):
1580          + Arm Cortex-A78 (cortex-a78).
1581          + Arm Cortex-A78AE (cortex-a78ae).
1582          + Arm Cortex-A78C (cortex-a78c).
1583          + Arm Cortex-X1 (cortex-x1).
1584          + Arm Neoverse V1 (neoverse-v1).
1585          + Arm Neoverse N2 (neoverse-n2).
1586     * GCC can now auto-vectorize operations performing addition,
1587       subtraction, multiplication and the accumulate/subtract variants on
1588       complex numbers, taking advantage of the Advanced SIMD instructions
1589       in the Armv8.3-a (AArch64/AArch32), SVE (AArch64), SVE2 (AArch64)
1590       and MVE (AArch32 M-profile) instruction sets.
1591
1592  AArch64
1593
1594     * In addition to the above, the following AArch64-only CPUs are now
1595       supported:
1596          + Fujitsu A64FX (a64fx).
1597          + Arm Cortex-R82 (cortex-r82).
1598     * The AArch64 Armv8-R architecture is now supported through the
1599       -march=armv8-r option.
1600     * Mitigation against the [53]Straight-line Speculation vulnerability
1601       is supported with the -mharden-sls= option. Please refer to the
1602       documentation for usage instructions.
1603     * The availability of Advanced SIMD intrinsics available through the
1604       arm_neon.h header is improved and GCC 11 supports the full set of
1605       intrinsics defined by ACLE Q3 2020.
1606
1607  AMD Radeon (GCN)
1608
1609     * Initial support for gfx908 GPUs has been added.
1610
1611  arm
1612
1613     * Initial auto-vectorization is now available when targeting the MVE
1614       instruction set.
1615     * GCC can now make use of the Low Overhead Branch instruction in
1616       Armv8.1-M to optimize loop counters and checks.
1617     * The -mcpu=cortex-m55 option now supports the extensions +nomve and
1618       +nomve.fp to control generation of MVE and MVE floating-point
1619       instructions.
1620
1621  IA-32/x86-64
1622
1623     * New ISA extension support for Intel TSXLDTRK was added to GCC.
1624       TSXLDTRK intrinsics are available via the -mtsxldtrk compiler
1625       switch.
1626     * New ISA extension support for Intel SERIALIZE was added to GCC.
1627       SERIALIZE intrinsics are available via the -mserialize compiler
1628       switch.
1629     * New ISA extension support for Intel HRESET was added to GCC. HRESET
1630       intrinsics are available via the -mhreset compiler switch.
1631     * New ISA extension support for Intel UINTR was added to GCC. UINTR
1632       intrinsics are available via the -muintr compiler switch.
1633     * New ISA extension support for Intel KEYLOCKER was added to GCC.
1634       KEYLOCKER intrinsics are available via the -mkeylocker compiler
1635       switch.
1636     * New ISA extension support for Intel AMX-TILE, AMX-INT8, AMX-BF16
1637       was added to GCC. AMX-TILE, AMX-INT8, AMX-BF16 intrinsics are
1638       available via the -mamx-tile, -mamx-int8, -mamx-bf16 compiler
1639       switches.
1640     * New ISA extension support for Intel AVX-VNNI was added to GCC.
1641       AVX-VNNI intrinsics are available via the -mavxvnni compiler
1642       switch.
1643     * GCC now supports the Intel CPU named Sapphire Rapids through
1644       -march=sapphirerapids. The switch enables the MOVDIRI, MOVDIR64B,
1645       AVX512VP2INTERSECT, ENQCMD, CLDEMOTE, SERIALIZE, PTWRITE, WAITPKG,
1646       TSXLDTRK, AMT-TILE, AMX-INT8, AMX-BF16, and AVX-VNNI ISA
1647       extensions.
1648     * GCC now supports the Intel CPU named Alderlake through
1649       -march=alderlake. The switch enables the CLDEMOTE, PTWRITE,
1650       WAITPKG, SERIALIZE, KEYLOCKER, AVX-VNNI, and HRESET ISA extensions.
1651     * GCC now supports the Intel CPU named Rocketlake through
1652       -march=rocketlake. Rocket Lake is based on Icelake client and minus
1653       SGX.
1654     * GCC now supports AMD CPUs based on the znver3 core via
1655       -march=znver3.
1656     * GCC now supports micro-architecture levels defined in the x86-64
1657       psABI via -march=x86-64-v2, -march=x86-64-v3 and -march=x86-64-v4.
1658
1659  Nios II
1660
1661     * The options -mcustom-insn=N no longer produce compiler warnings if
1662       the custom instruction is not generated due to missing optimization
1663       options such as -fno-math-errno, -ffinite-math-only, or
1664       -funsafe-math-optimizations. These warnings were not consistently
1665       emitted for all custom instructions.
1666     * The -mcustom-fpu-cfg=fph2 option has been added to enable the
1667       custom instructions supported by the Nios II Floating Point
1668       Hardware 2 Component.
1669
1670  NVPTX
1671
1672     * The -misa default has changed from sm_30 to sm_35.
1673     * The -m32 compiler switch has been removed.
1674     * The -msoft-stack-reserve-local format has been fixed. Previously,
1675       it accepted -msoft-stack-reserve-local<n>. It now accepts
1676       -msoft-stack-reserve-local=<n>.
1677
1678  S/390, System z, IBM Z Systems
1679
1680     * The behavior when compiling with -fexcess-precision=standard (e.g.,
1681       implied by -std=c99) on s390(x) targets can now be controlled at
1682       configure time with the flag --enable-s390-excess-float-precision.
1683       When enabled, GCC will maintain previous behavior and evaluate
1684       float expressions in double precision, which aligns with the
1685       definition of float_t as double. With the flag disabled, GCC will
1686       always evaluate float expressions in single precision. In native
1687       builds and cross compiles that have target libc headers, GCC will
1688       by default match the definition of float_t in the installed glibc.
1689
1690  RISC-V
1691
1692     * Support address sanitizer for RISC-V.
1693     * Support big-endian for RISC-V, thanks to Marcus Comstedt.
1694     * Implement new style of architecture extension test macros: each
1695       architecture extension has a corresponding feature test macro,
1696       which can be used to test its existence and version information.
1697     * Legacy architecture extension test macros like __riscv_atomic are
1698       deprecated, but will still be supported for at least 2 release
1699       cycles.
1700     * Support IFUNC for riscv*-*-linux*.
1701     * Add new option -misa-spec=* to control ISA spec version. This
1702       controls the default version of each extensions. It defaults to
1703       2.2.
1704     * Introduce the --with-multilib-generator configure time option. This
1705       allows for flexible config multi-lib settings. Its syntax is the
1706       same as RISC-V's multilib-generator.
1707     * Extend the sytax for multilib-generator, support expansion operator
1708       * to reduce the complexity of complicated multi-lib re-use rules.
1709     * Support -mcpu=* option aligned with RISC-V clang/LLVM. It sets the
1710       pipeline model and architecture extensions, like -mtune=* plus
1711       -march=*.
1712     * Support for TLS stack protector canary access, thanks to Cooper Qu.
1713     * Support __builtin_thread_pointer for RISC-V.
1714     * Introduce shorten_memrefs optimization, which reduces the code size
1715       for memory access, thanks to Craig Blackmore.
1716
1717Operating Systems
1718
1719  AIX
1720
1721     * GCC for AIX can be built as a 64 bit application and the runtime is
1722       built as FAT libraries containing both 32 bit and 64 bit objects.
1723     * Support AIX Vector Extended ABI with -mabi=vec-extabi.
1724     * Thread-Local uninitiated data placed in local common section.
1725     * Use thread-safe access in ctype.
1726     * Link with libc128.a when long-double-128 enabled.
1727
1728Improvements to Static Analyzer
1729
1730     * The implementation of how program state is tracked within
1731       [54]-fanalyzer has been completely rewritten for GCC 11, fixing
1732       numerous bugs, and allowing for the analyzer to scale to larger C
1733       source files.
1734     * The analysis of allocations and deallocations has been generalized
1735       beyond malloc and free.
1736          + As preliminary work towards eventually supporting C++, the
1737            malloc/free checking will also check new/delete and
1738            new[]/delete[]. However, C++ is not yet properly supported by
1739            [55]-fanalyzer (for example, exception-handling is
1740            unimplemented).
1741          + As noted above, the existing [56]malloc attribute has been
1742            extended so that it can be used to identify
1743            allocator/deallocator API pairs. The analyzer will use these
1744            attributes when checking for leaks, double-frees,
1745            use-after-frees, and similar issues.
1746          + A new [57]-Wanalyzer-mismatching-deallocation warning has been
1747            added, covering such mismatches as using scalar delete rather
1748            vector delete[].
1749     * The analyzer has gained warnings
1750       [58]-Wanalyzer-shift-count-negative,
1751       [59]-Wanalyzer-shift-count-overflow, [60]-Wanalyzer-write-to-const,
1752       and [61]-Wanalyzer-write-to-string-literal, all enabled by default
1753       when [62]-fanalyzer is enabled.
1754     * The analyzer can now be extended by GCC plugins, allowing for
1755       domain-specific path-sensitive warnings. An example of using a
1756       [63]GCC plugin to check for misuses of CPython's global interpreter
1757       lock can be seen in the test suite
1758     * The analyzer has gained new debugging options
1759       [64]-fdump-analyzer-json and [65]-fno-analyzer-feasibility.
1760
1761Other significant improvements
1762
1763     * GCC has gained a new environment variable
1764       [66]GCC_EXTRA_DIAGNOSTIC_OUTPUT which can be used by IDEs to
1765       request machine-readable fix-it hints without needing to adjust
1766       build flags.
1767
1768GCC 11.1
1769
1770   This is the [67]list of problem reports (PRs) from GCC's bug tracking
1771   system that are known to be fixed in the 11.1 release. This list might
1772   not be complete (that is, it is possible that some PRs that have been
1773   fixed are not listed here).
1774
1775GCC 11.2
1776
1777   This is the [68]list of problem reports (PRs) from GCC's bug tracking
1778   system that are known to be fixed in the 11.2 release. This list might
1779   not be complete (that is, it is possible that some PRs that have been
1780   fixed are not listed here).
1781
1782GCC 11.3
1783
1784   This is the [69]list of problem reports (PRs) from GCC's bug tracking
1785   system that are known to be fixed in the 11.3 release. This list might
1786   not be complete (that is, it is possible that some PRs that have been
1787   fixed are not listed here).
1788
1789  Target Specific Changes
1790
1791    x86-64
1792
1793     * The x86-64 ABI of passing and returning structure with a 64-bit
1794       single precision vector changed in GCC 11.1 when 3DNOW is disabled.
1795       Disabling 3DNOW no longer changes how they are passed nor returned.
1796       This ABI change is now diagnosed with -Wpsabi.
1797     * Mitigation against straight line speculation (SLS) for function
1798       return and indirect jump is supported via
1799       -mharden-sls=[none|all|return|indirect-jmp].
1800     * Add CS prefix to call and jmp to indirect thunk with branch target
1801       in r8-r15 registers via -mindirect-branch-cs-prefix.
1802
1803
1804    For questions related to the use of GCC, please consult these web
1805    pages and the [70]GCC manuals. If that fails, the
1806    [71]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1807    web pages and the development of GCC are welcome on our developer
1808    list at [72]gcc@gcc.gnu.org. All of [73]our lists have public
1809    archives.
1810
1811   Copyright (C) [74]Free Software Foundation, Inc. Verbatim copying and
1812   distribution of this entire article is permitted in any medium,
1813   provided this notice is preserved.
1814
1815   These pages are [75]maintained by the GCC team. Last modified
1816   2023-02-02.
1817
1818References
1819
1820   1. http://gcc.gnu.org/gcc-11/porting_to.html
1821   2. http://gcc.gnu.org/onlinedocs/index.html#current
1822   3. https://wg21.link/p0522r0
1823   4. https://gcc.gnu.org/pipermail/gcc-patches/2020-May/546494.html
1824   5. https://github.com/google/sanitizers/wiki/ThreadSanitizerCppManual
1825   6. https://www.kernel.org/doc/html/latest/dev-tools/kcsan.html
1826   7. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-column-unit
1827   8. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-column-origin
1828   9. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format
1829  10. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Preprocessor-Options.html#index-ftabstop
1830  11. https://clang.llvm.org/docs/HardwareAssistedAddressSanitizerDesign.html
1831  12. https://github.com/google/sanitizers/wiki/AddressSanitizer
1832  13. https://clang.llvm.org/docs/HardwareAssistedAddressSanitizerDesign.html
1833  14. https://dwarfstd.org/doc/DWARF5.pdf
1834  15. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Code-Gen-Options.html#index-fno-bit-tests
1835  16. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fipa-modref
1836  17. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fipa-ocf
1837  18. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Optimize-Options.html#index-fprofile-values
1838  19. https://www.openmp.org/specifications/
1839  20. https://www.openacc.org/
1840  21. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-11
1841  22. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/libgomp/Enabling-OpenACC.html
1842  23. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-no_005fstack_005fprotector-function-attribute
1843  24. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-malloc-function-attribute
1844  25. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmismatched-dealloc
1845  26. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-new-delete
1846  27. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wfree-nonheap-object
1847  28. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmismatched-dealloc
1848  29. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wsizeof-array-div
1849  30. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wstringop-overread
1850  31. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wtsan
1851  32. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wfree-nonheap-object
1852  33. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wmaybe-uninitialized
1853  34. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wuninitialized
1854  35. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Warray-parameter
1855  36. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wvla-parameter
1856  37. http://gcc.gnu.org/projects/cxx-status.html#cxx20
1857  38. http://gcc.gnu.org/projects/cxx-status.html#cxx23
1858  39. https://gcc.gnu.org/PR41437
1859  40. https://gcc.gnu.org/PR95307
1860  41. https://gcc.gnu.org/PR97518
1861  42. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wctad-maybe-unsupported
1862  43. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wrange-loop-construct
1863  44. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-enum-enum-conversion
1864  45. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-enum-float-conversion
1865  46. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-new-delete
1866  47. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wvexing-parse
1867  48. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Warning-Options.html#index-Wnonnull
1868  49. http://gcc.gnu.org/gcc-5/changes.html#jit
1869  50. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/compatibility.html
1870  51. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/expressions.html#c.gcc_jit_global_set_initializer
1871  52. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/jit/topics/asm.html
1872  53. https://developer.arm.com/documentation/102587/0102/Straight-line-speculation-frequently-asked-questions
1873  54. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html
1874  55. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html
1875  56. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Common-Function-Attributes.html#index-malloc-function-attribute
1876  57. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-mismatching-deallocation
1877  58. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-shift-count-negative
1878  59. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-shift-count-overflow
1879  60. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-const
1880  61. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-Wanalyzer-write-to-string-literal
1881  62. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html
1882  63. https://gcc.gnu.org/git/?p=gcc.git;a=commitdiff;h=66dde7bc64b75d4a338266333c9c490b12d49825
1883  64. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-fdump-analyzer-json
1884  65. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Static-Analyzer-Options.html#index-fno-analyzer-feasibility
1885  66. https://gcc.gnu.org/onlinedocs/gcc-11.1.0/gcc/Environment-Variables.html#index-GCC_005fEXTRA_005fDIAGNOSTIC_005fOUTPUT
1886  67. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.0
1887  68. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.2
1888  69. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=11.3
1889  70. https://gcc.gnu.org/onlinedocs/
1890  71. mailto:gcc-help@gcc.gnu.org
1891  72. mailto:gcc@gcc.gnu.org
1892  73. https://gcc.gnu.org/lists.html
1893  74. https://www.fsf.org/
1894  75. https://gcc.gnu.org/about.html
1895======================================================================
1896http://gcc.gnu.org/gcc-10/index.html
1897
1898                             GCC 10 Release Series
1899
1900   June 28, 2022
1901
1902   The [1]GNU project and the GCC developers are pleased to announce the
1903   release of GCC 10.4.
1904
1905   This release is a bug-fix release, containing fixes for regressions in
1906   GCC 10.3 relative to previous releases of GCC.
1907
1908Release History
1909
1910   GCC 10.4
1911          June 28, 2022 ([2]changes, [3]documentation)
1912
1913   GCC 10.3
1914          April 8, 2021 ([4]changes, [5]documentation)
1915
1916   GCC 10.2
1917          July 23, 2020 ([6]changes, [7]documentation)
1918
1919   GCC 10.1
1920          May 7, 2020 ([8]changes, [9]documentation)
1921
1922References and Acknowledgements
1923
1924   GCC used to stand for the GNU C Compiler, but since the compiler
1925   supports several other languages aside from C, it now stands for the
1926   GNU Compiler Collection.
1927
1928   A list of [10]successful builds is updated as new information becomes
1929   available.
1930
1931   The GCC developers would like to thank the numerous people that have
1932   contributed new features, improvements, bug fixes, and other changes as
1933   well as test results to GCC. This [11]amazing group of volunteers is
1934   what makes GCC successful.
1935
1936   For additional information about GCC please refer to the [12]GCC
1937   project web site or contact the [13]GCC development mailing list.
1938
1939   To obtain GCC please use [14]our mirror sites or [15]our version
1940   control system.
1941
1942
1943    For questions related to the use of GCC, please consult these web
1944    pages and the [16]GCC manuals. If that fails, the
1945    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1946    web pages and the development of GCC are welcome on our developer
1947    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
1948    archives.
1949
1950   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
1951   distribution of this entire article is permitted in any medium,
1952   provided this notice is preserved.
1953
1954   These pages are [21]maintained by the GCC team. Last modified
1955   2022-10-26.
1956
1957References
1958
1959   1. http://www.gnu.org/
1960   2. http://gcc.gnu.org/gcc-10/changes.html
1961   3. http://gcc.gnu.org/onlinedocs/10.4.0/
1962   4. http://gcc.gnu.org/gcc-10/changes.html
1963   5. http://gcc.gnu.org/onlinedocs/10.3.0/
1964   6. http://gcc.gnu.org/gcc-10/changes.html
1965   7. http://gcc.gnu.org/onlinedocs/10.2.0/
1966   8. http://gcc.gnu.org/gcc-10/changes.html
1967   9. http://gcc.gnu.org/onlinedocs/10.1.0/
1968  10. http://gcc.gnu.org/gcc-10/buildstat.html
1969  11. http://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Contributors.html
1970  12. http://gcc.gnu.org/index.html
1971  13. mailto:gcc@gcc.gnu.org
1972  14. http://gcc.gnu.org/mirrors.html
1973  15. http://gcc.gnu.org/git.html
1974  16. https://gcc.gnu.org/onlinedocs/
1975  17. mailto:gcc-help@gcc.gnu.org
1976  18. mailto:gcc@gcc.gnu.org
1977  19. https://gcc.gnu.org/lists.html
1978  20. https://www.fsf.org/
1979  21. https://gcc.gnu.org/about.html
1980======================================================================
1981http://gcc.gnu.org/gcc-10/changes.html
1982
1983                             GCC 10 Release Series
1984                        Changes, New Features, and Fixes
1985
1986   This page is a "brief" summary of some of the huge number of
1987   improvements in GCC 10. You may also want to check out our [1]Porting
1988   to GCC 10 page and the [2]full GCC documentation.
1989
1990Caveats
1991
1992     * An ABI incompatibility between C++14 and C++17 has been fixed. On
1993       some targets a class with a zero-sized subobject would be passed
1994       incorrectly when compiled as C++17 or C++20. See the [3]C++ notes
1995       below for more details.
1996     * The deprecated Profile Mode and array_allocator extensions have
1997       been removed from libstdc++.
1998     * The non-standard std::__is_nullptr_t type trait is deprecated and
1999       will be removed from libstdc++ in a future release. The standard
2000       trait std::is_null_pointer should be instead.
2001     * The minimum version of the [4]MPFR library required for building
2002       GCC has been increased to version 3.1.0 (released 2011-10-03).
2003     * The automatic template instantiation at link time (-frepo) has been
2004       removed.
2005     * The --param allow-store-data-races internal parameter has been
2006       removed in favor of a new official option -fallow-store-data-races.
2007       While default behavior is unchanged and the new option allows to
2008       correctly maintain a per compilation unit setting across link-time
2009       optimization, alteration of the default via --param
2010       allow-store-data-races will now be diagnosed and build systems have
2011       to be adjusted accordingly.
2012     * Offloading to Heterogeneous System Architecture Intermediate
2013       Language (HSAIL) has been deprecated and will likely be removed in
2014       a future release.
2015     * The type of the std::iterator base class of
2016       std::istreambuf_iterator was changed in C++98 mode to be consistent
2017       with C++11 and later standards. See the [5]libstdc++ notes below
2018       for more details.
2019
2020General Improvements
2021
2022     * New built-in functions:
2023          + The [6]__has_builtin built-in preprocessor operator can be
2024            used to query support for built-in functions provided by GCC
2025            and other compilers that support it.
2026          + __builtin_roundeven for the corresponding function from
2027            ISO/IEC TS 18661.
2028     * New command-line options:
2029          + [7]-fallocation-dce removes unneeded pairs of new and delete
2030            operators.
2031          + [8]-fprofile-partial-training can now be used to inform the
2032            compiler that code paths not covered by the training run
2033            should not be optimized for size.
2034          + [9]-fprofile-reproducible controls level of reproducibility of
2035            profile gathered by [10]-fprofile-generate. This makes it
2036            possible to rebuild program with same outcome which is useful,
2037            for example, for distribution packages.
2038          + [11]-fprofile-prefix-path can be used in combination with
2039            -fprofile-generate=profile_dir and -fprofile-use=profile_dir
2040            to inform GCC where the base directory of build source tree is
2041            in case it differs between instrumentation and optimized
2042            builds.
2043          + [12]-fanalyzer enables a new static analysis pass and
2044            associated warnings. This pass performs a time-consuming
2045            exploration of paths through the code in the hope of detecting
2046            various common errors, such as double-free bugs. This option
2047            should be regarded as experimental in this release. In
2048            particular, analysis of non-C code is unlikely to work.
2049     * Inter-procedural optimization improvements:
2050          + The inter-procedural scalar replacement of aggregates
2051            (IPA-SRA) pass was re-implemented to work at link-time and can
2052            now also remove computing and returning unused return values.
2053          + [13]-finline-functions is now enabled at -O2 and was retuned
2054            for better code size versus runtime performance trade-offs.
2055            Inliner heuristics was also significantly sped up to avoid
2056            negative impact to -flto -O2 compile times.
2057          + Inliner heuristics and function cloning can now use
2058            value-range information to predict effectivity of individual
2059            transformations.
2060          + During link-time optimization the C++ One Definition Rule is
2061            used to increase precision of type based alias analysis.
2062     * Link-time optimization improvements:
2063          + A new binary [14]lto-dump has been added. It dumps various
2064            information about LTO bytecode object files.
2065          + The parallel phase of the LTO can automatically detect a
2066            running make's jobserver or fall back to number of available
2067            cores.
2068          + The LTO bytecode can be compressed with the [15]zstd
2069            algorithm. The configure script automatically detects zstd
2070            support.
2071          + Most --param values can now be specified at translation unit
2072            granularity. This includes all parameters controlling the
2073            inliner and other inter-procedural optimizations. Unlike
2074            earlier releases, GCC 10 will ignore parameters controlling
2075            optimizations specified at link-time and apply parameters
2076            specified at compile-time in the same manner as done for
2077            optimization flags.
2078     * Profile driven optimization improvements:
2079          + Profile maintenance during compilation and hot/cold code
2080            partitioning have been improved.
2081          + Using [16]-fprofile-values, an instrumented binary can track
2082            multiple values (up to 4) for e.g. indirect calls and provide
2083            more precise profile information.
2084
2085New Languages and Language-Specific Improvements
2086
2087     * Version 2.6 of the [17]OpenACC specification is now supported by
2088       the C, C++ and Fortran compilers. See the [18]implementation status
2089       section on the OpenACC wiki page and the [19]run-time library
2090       documentation for further information.
2091     * GCC 10 adds a number of newly implemented [20]OpenMP 5.0 features
2092       such as conditional lastprivate clause, scan and loop directives,
2093       order(concurrent) and use_device_addr clauses support, if clause on
2094       simd construct, and partial support for the declare variant
2095       directive, getting closer to full support of the OpenMP 5.0
2096       standard.
2097     * OpenMP and OpenACC now support [21]offloading to AMD Radeon (GCN)
2098       GPUs; supported are the third-generation Fiji (fiji) and the
2099       fifth-generation VEGA 10/VEGA 20 (gfx900 or gfx906).
2100
2101  C family
2102
2103     * New attributes:
2104          + The access function and type attribute has been added to
2105            describe how a function accesses objects passed to it by
2106            pointer or reference, and to associate such arguments with
2107            integer arguments denoting the objects' sizes. The attribute
2108            is used to enable the detection of invalid accesses by
2109            user-defined functions, such as those diagnosed by
2110            -Wstringop-overflow.
2111          + The symver attribute can be used to bind symbols to specific
2112            version nodes on ELF platforms. This is preferred to using
2113            inline assembly with GNU as symver directive because the
2114            latter is not compatible with link-time optimizations.
2115     * New warnings:
2116          + [22]-Wstring-compare, enabled by -Wextra, warns about equality
2117            and inequality expressions between zero and the result of a
2118            call to either strcmp and strncmp that evaluate to a constant
2119            as a result of the length of one argument being greater than
2120            the size of the array pointed to by the other.
2121          + [23]-Wzero-length-bounds, enabled by -Warray-bounds, warns
2122            about accesses to elements of zero-length arrays that might
2123            overlap other members of the same object.
2124     * Enhancements to existing warnings:
2125          + [24]-Warray-bounds detects more out-of-bounds accesses to
2126            member arrays as well as accesses to elements of zero-length
2127            arrays.
2128          + [25]-Wformat-overflow makes full use of string length
2129            information computed by the strlen optimization pass.
2130          + [26]-Wrestrict detects overlapping accesses to dynamically
2131            allocated objects.
2132          + [27]-Wreturn-local-addr diagnoses more instances of return
2133            statements returning addresses of automatic variables.
2134          + [28]-Wstringop-overflow detects more out-of-bounds stores to
2135            member arrays including zero-length arrays, dynamically
2136            allocated objects and variable length arrays, as well as more
2137            instances of reads of unterminated character arrays by string
2138            built-in functions. The warning also detects out-of-bounds
2139            accesses by calls to user-defined functions declared with the
2140            new attribute access.
2141          + [29]-Warith-conversion re-enables warnings from -Wconversion,
2142            -Wfloat-conversion, and -Wsign-conversion that are now off by
2143            default for an expression where the result of an arithmetic
2144            operation will not fit in the target type due to promotion,
2145            but the operands of the expression do fit in the target type.
2146     * Extended characters in identifiers may now be specified directly in
2147       the input encoding (UTF-8, by default), in addition to the UCN
2148       syntax (\uNNNN or \UNNNNNNNN) that is already supported:
2149
2150static const int p = 3;
2151int get_na�ve_pi() {
2152  return p;
2153}
2154
2155  C
2156
2157     * Several new features from the upcoming C2X revision of the ISO C
2158       standard are supported with -std=c2x and -std=gnu2x. Some of these
2159       features are also supported as extensions when compiling for older
2160       language versions. In addition to the features listed, some
2161       features previously supported as extensions and now added to the C
2162       standard are enabled by default in C2X mode and not diagnosed with
2163       -std=c2x -Wpedantic.
2164          + The [[]] attribute syntax is supported, as in C++. Existing
2165            attributes can be used with this syntax in forms such as
2166            [[gnu::const]]. The standard attributes [[deprecated]],
2167            [[fallthrough]] and [[maybe_unused]] are supported.
2168          + UTF-8 character constants using the u8'' syntax are supported.
2169          + <float.h> defines macros FLT_NORM_MAX, DBL_NORM_MAX and
2170            LDBL_NORM_MAX.
2171          + When decimal floating-point arithmetic is supported, <float.h>
2172            defines macros DEC32_TRUE_MIN, DEC64_TRUE_MIN and
2173            DEC128_TRUE_MIN, in addition to the macros that were
2174            previously only defined if __STDC_WANT_DEC_FP__ was defined
2175            before including <float.h>.
2176          + In C2X mode, empty parentheses in a function definition give
2177            that function a type with a prototype for subsequent calls;
2178            other old-style function definitions are diagnosed by default
2179            in C2X mode.
2180          + The strftime format checking supports the %OB and %Ob formats.
2181          + In C2X mode, -fno-fp-int-builtin-inexact is enabled by
2182            default.
2183     * GCC now defaults to -fno-common. As a result, global variable
2184       accesses are more efficient on various targets. In C, global
2185       variables with multiple tentative definitions now result in linker
2186       errors. With -fcommon such definitions are silently merged during
2187       linking.
2188
2189  C++
2190
2191     * Several C++20 features have been implemented:
2192          + Concepts, including P0734R0, P0857R0, P1084R2, P1141R2,
2193            P0848R3, P1616R1, P1452R2
2194          + P1668R1, Permitting Unevaluated inline-assembly in constexpr
2195            Functions
2196          + P1161R3, Deprecate a[b,c]
2197          + P0848R3, Conditionally Trivial Special Member Functions
2198          + P1091R3, Extending structured bindings
2199          + P1143R2, Adding the constinit keyword
2200          + P1152R4, Deprecating volatile
2201          + P0388R4, Permit conversions to arrays of unknown bound
2202          + P0784R7, constexpr new
2203          + P1301R4, [[nodiscard("with reason")]]
2204          + P1814R0, class template argument deduction for alias templates
2205          + P1816R0, class template argument deduction for aggregates
2206          + P0960R3, Parenthesized initialization of aggregates
2207          + P1331R2, Allow trivial default initialization in constexpr
2208            contexts
2209          + P1327R1, Allowing dynamic_cast and polymorphic typeid in
2210            constexpr contexts
2211          + P0912R5, Coroutines (requires -fcoroutines)
2212     * Several C++ Defect Reports have been resolved, e.g.:
2213          + DR 1560, lvalue-to-rvalue conversion in ?:
2214          + DR 1813, __is_standard_layout for a class with repeated bases
2215          + DR 2094, volatile scalars are trivially copyable,
2216          + DR 2096, constraints on literal unions
2217          + DR 2413, typename in conversion-function-ids
2218          + DR 2352, Similar types and reference binding
2219          + DR 1601, Promotion of enumeration with fixed underlying type
2220          + DR 330, Qualification conversions and pointers to arrays of
2221            pointers
2222          + DR 1307, Overload resolution based on size of array
2223            initializer-list
2224          + DR 1710, Missing template keyword in class-or-decltype
2225     * New warnings:
2226          + [30]-Wmismatched-tags, disabled by default, warns about
2227            declarations of structs, classes, and class templates and
2228            their specializations with a class-key that does not match
2229            either the definition or the first declaration if no
2230            definition is provided. The option is provided to ease
2231            portability to Windows-based compilers.
2232          + [31]-Wredundant-tags, disabled by default, warns about
2233            redundant class-key and enum-key in contexts where the key can
2234            be eliminated without causing an syntactic ambiguity.
2235     * G++ can now detect modifying constant objects in constexpr
2236       evaluation (which is undefined behavior).
2237     * G++ no longer emits bogus -Wsign-conversion warnings with explicit
2238       casts.
2239     * Narrowing is now detected in more contexts (e.g., case values).
2240     * Memory consumption of the compiler has been reduced in constexpr
2241       evaluation.
2242     * The noexcept-specifier is now properly treated as a complete-class
2243       context as per [class.mem].
2244     * The attribute deprecated can now be used on namespaces too.
2245     * The ABI of passing and returning certain C++ classes by value
2246       changed on several targets in GCC 10, including [32]AArch64,
2247       [33]ARM, [34]PowerPC ELFv2, [35]S/390 and [36]Itanium. These
2248       changes affect classes with a zero-sized subobject (an empty base
2249       class, or data member with the [[no_unique_address]] attribute)
2250       where all other non-static data members have the same type (this is
2251       called a "homogeneous aggregate" in some ABI specifications, or if
2252       there is only one such member, a "single element"). In -std=c++17
2253       and -std=c++20 modes, classes with an empty base class were not
2254       considered to have a single element or to be a homogeneous
2255       aggregate, and so could be passed differently (in the wrong
2256       registers or at the wrong stack address). This could make code
2257       compiled with -std=c++17 and -std=c++14 ABI incompatible. This has
2258       been corrected and the empty bases are ignored in those ABI
2259       decisions, so functions compiled with -std=c++14 and -std=c++17 are
2260       now ABI compatible again. Example: struct empty {}; struct S :
2261       empty { float f; }; void f(S);. Similarly, in classes containing
2262       non-static data members with empty class types using the C++20
2263       [[no_unique_address]] attribute, those members weren't ignored in
2264       the ABI argument passing decisions as they should be. Both of these
2265       ABI changes are now diagnosed with -Wpsabi.
2266
2267    Runtime Library (libstdc++)
2268
2269     * Improved experimental C++2a support, including:
2270          + Library concepts in <concepts> and <iterator>.
2271          + Constrained algorithms in <ranges>, <algorithm>, and <memory>
2272            (thanks to Patrick Palka).
2273          + New algorithms shift_left and shift_right (thanks to Patrick
2274            Palka).
2275          + std::span (thanks to JeanHeyd Meneide).
2276          + Three-way comparisons in <compare> and throughout the library.
2277          + Constexpr support in <algorithm> and elsewhere (thanks to
2278            Edward Smith-Rowland).
2279          + <stop_token> and std::jthread (thanks to Thomas Rodgers).
2280          + std::atomic_ref and std::atomic<floating point>.
2281          + Integer comparison functions (cmp_equal, cmp_less etc.).
2282          + std::ssize, std::to_array.
2283          + std::construct_at, std::destroy, constexpr std::allocator.
2284          + Mathematical constants in <numbers>.
2285     * Support for RDSEED in std::random_device.
2286     * Reduced header dependencies, leading to faster compilation for some
2287       code.
2288     * The std::iterator base class of std::istreambuf_iterator was
2289       changed in C++98 mode to be consistent with C++11 and later
2290       standards. This is expected to have no noticeable effect except in
2291       the unlikely case of a class which has potentially overlapping
2292       subobjects of type std::istreambuf_iterator<C> and another iterator
2293       type with a std::iterator<input_iterator_tag, C, ...> base class.
2294       The layout of such a type might change when compiled as C++98.
2295       [37]Bug 92285 has more details and concrete examples.
2296
2297  D
2298
2299     * Support for static foreach has been implemented.
2300     * Aliases can now be created directly from any __traits that return
2301       symbols or tuples. Previously, an AliasSeq was necessary in order
2302       to alias their return.
2303     * It is now possible to detect the language ABI specified for a
2304       struct, class, or interface using __traits(getLinkage, ...).
2305     * Support for core.math.toPrec intrinsics has been added. These
2306       intrinsics guarantee the rounding to specific floating-point
2307       precisions at specified points in the code.
2308     * Support for pragma(inline) has been implemented. Previously the
2309       pragma was recognized, but had no effect on the compilation.
2310     * Optional parentheses in asm operands are deprecated and will be
2311       removed in a future release.
2312     * All content imported files are now included in the make dependency
2313       list when compiling with -M.
2314     * Compiler recognized attributes provided by the gcc.attribute module
2315       will now take effect when applied to function prototypes as well as
2316       when applied to full function declarations.
2317     * Added a --enable-libphobos-checking configure option to control
2318       whether run-time checks are compiled into the D runtime library.
2319     * Added a --with-libphobos-druntime-only configure option to indicate
2320       whether to build only the core D runtime library, or both the core
2321       and standard libraries into libphobos.
2322
2323  Fortran
2324
2325     * use_device_addr of version 5.0 of the [38]OpenMP specification is
2326       now supported. Note that otherwise OpenMP 4.5 is partially
2327       supported in the Fortran compiler; the largest missing item is
2328       structure element mapping.
2329     * The default buffer size for I/O using unformatted files has been
2330       increased to 1048576. The buffer size for can now be set at runtime
2331       via the environment variables GFORTRAN_FORMATTED_BUFFER_SIZE and
2332       GFORTRAN_UNFORMATTED_BUFFER_SIZE for formatted and unformatted
2333       files, respectively.
2334     * Mismatches between actual and dummy argument lists in a single file
2335       are now rejected with an error. Use the new option
2336       -fallow-argument-mismatch to turn these errors into warnings; this
2337       option is implied with -std=legacy. -Wargument-mismatch has been
2338       removed.
2339     * The handling of a BOZ literal constant has been reworked to provide
2340       better conformance to the Fortran 2008 and 2018 standards. In these
2341       Fortran standards, a BOZ literal constant is a typeless and
2342       kindless entity. As a part of the rework, documented and
2343       undocumented extensions to the Fortran standard now emit errors
2344       during compilation. Some of these extensions are permitted with the
2345       -fallow-invalid-boz option, which degrades the error to a warning
2346       and the code is compiled as with older gfortran.
2347     * At any optimization level except-Os, gfortran now uses inline
2348       packing for arguments instead of calling a library routine. If the
2349       source contains a large number of arguments that need to be
2350       repacked, code size or time for compilation can become excessive.
2351       If that is the case, -fno-inline-arg-packing can be used to disable
2352       inline argument packing.
2353     * Legacy extensions:
2354          + For formatted input/output, if the explicit widths after the
2355            data-edit descriptors I, F and G have been omitted, default
2356            widths are used.
2357          + A blank format item at the end of a format specification, i.e.
2358            nothing following the final comma, is allowed. Use the option
2359            -fdec-blank-format-item; this option is implied with -fdec.
2360          + The existing support for AUTOMATIC and STATIC attributes has
2361            been extended to allow variables with the AUTOMATIC attribute
2362            to be used in EQUIVALENCE statements. Use -fdec-static; this
2363            option is implied by -fdec.
2364          + Allow character literals in assignments and DATA statements
2365            for numeric (INTEGER, REAL, or COMPLEX) or LOGICAL variables.
2366            Use the option -fdec-char-conversions; this option is implied
2367            with -fdec.
2368          + DEC comparisons, i.e. allow Hollerith constants to be used in
2369            comparisons with INTEGER, REAL, COMPLEX and CHARACTER
2370            expressions. Use the option -fdec.
2371     * Character type names in errors and warnings now include len in
2372       addition to kind; * is used for assumed length. The kind is omitted
2373       if it is the default kind. Examples: CHARACTER(12), CHARACTER(6,4).
2374     * CO_BROADCAST now supports derived type variables including objects
2375       with allocatable components. In this case, the optional arguments
2376       STAT= and ERRMSG= are currently ignored.
2377     * The handling of module and submodule names has been reworked to
2378       allow the full 63-character length mandated by the standard.
2379       Previously symbol names were truncated if the combined length of
2380       module, submodule, and function name exceeded 126 characters. This
2381       change therefore breaks the ABI, but only for cases where this 126
2382       character limit was exceeded.
2383
2384  Go
2385
2386     * GCC 10 provides a complete implementation of the Go 1.14.6 user
2387       packages.
2388
2389libgccjit
2390
2391     * The libgccjit API gained four new entry points:
2392          + [39]gcc_jit_version_major, [40]gcc_jit_version_minor, and
2393            [41]gcc_jit_version_patchlevel for programmatically checking
2394            the libgccjit version from client code, and
2395          + [42]gcc_jit_context_new_bitfield
2396
2397New Targets and Target Specific Improvements
2398
2399  AArch64 & arm
2400
2401     * The AArch64 and arm ports now support condition flag output
2402       constraints in inline assembly, as indicated by the
2403       __GCC_ASM_FLAG_OUTPUTS__. On arm this feature is only available for
2404       A32 and T32 targets. Please refer to the documentation for more
2405       details.
2406
2407  AArch64
2408
2409     * There have been several improvements related to the Scalable Vector
2410       Extension (SVE):
2411          + The SVE ACLE types and intrinsics are now supported. They can
2412            be accessed using the header file arm_sve.h.
2413          + It is now possible to create fixed-length SVE types using the
2414            arm_sve_vector_bits attribute. For example:
2415#if __ARM_FEATURE_SVE_BITS==512
2416typedef svint32_t vec512 __attribute__((arm_sve_vector_bits(512)));
2417typedef svbool_t pred512 __attribute__((arm_sve_vector_bits(512)));
2418#endif
2419          + -mlow-precision-div, -mlow-precision-sqrt and
2420            -mlow-precision-recip-sqrt now work for SVE.
2421          + -msve-vector-bits=128 now generates vector-length-specific
2422            code for little-endian targets. It continues to generate
2423            vector-length-agnostic code for big-endian targets, just as
2424            previous releases did for all targets.
2425          + The vectorizer is now able to use extending loads and
2426            truncating stores, including gather loads and scatter stores.
2427          + The vectorizer now compares the cost of vectorizing with SVE
2428            and vectorizing with Advanced SIMD and tries to pick the best
2429            one. Previously it would always use SVE if possible.
2430          + If a vector loop uses Advanced SIMD rather than SVE, the
2431            vectorizer now considers using SVE to vectorize the left-over
2432            elements (the "scalar tail" or "epilog").
2433          + Besides these specific points, there have been many general
2434            improvements to the way that the vectorizer uses SVE.
2435     * The -mbranch-protection=pac-ret option now accepts the optional
2436       argument +b-key extension to perform return address signing with
2437       the B-key instead of the A-key.
2438     * The option -moutline-atomics has been added to aid deployment of
2439       the Large System Extensions (LSE) on GNU/Linux systems built with a
2440       baseline architecture targeting Armv8-A. When the option is
2441       specified code is emitted to detect the presence of LSE
2442       instructions at runtime and use them for standard atomic
2443       operations. For more information please refer to the documentation.
2444     * The Transactional Memory Extension is now supported through ACLE
2445       intrinsics. It can be enabled through the +tme option extension
2446       (for example, -march=armv8.5-a+tme).
2447     * A number of features from Armv8.5-A are now supported through ACLE
2448       intrinsics. These include:
2449          + The random number instructions that can be enabled through the
2450            (already present in GCC 9.1) +rng option extension.
2451          + Floating-point intrinsics to round to integer instructions
2452            from Armv8.5-A when targeting -march=armv8.5-a or later.
2453          + Memory Tagging Extension intrinsics enabled through the
2454            +memtag option extension.
2455     * Similarly, the following Armv8.6-A features are now supported
2456       through ACLE intrinsics:
2457          + The bfloat16 extension. This extension is enabled
2458            automatically when Armv8.6-A is selected (such as by
2459            -march=armv8.6-a). It can also be enabled for Armv8.2-A and
2460            later using the +bf16 option extension.
2461          + The Matrix Multiply extension. This extension is split into
2462            three parts, one for each supported data type:
2463               o Support for 8-bit integer matrix multiply instructions.
2464                 This extension is enabled automatically when Armv8.6-A is
2465                 selected. It can also be enabled for Armv8.2-A and later
2466                 using the +i8mm option extension.
2467               o Support for 32-bit floating-point matrix multiply
2468                 instructions. This extension can be enabled using the
2469                 +f32mm option extension, which also has the effect of
2470                 enabling SVE.
2471               o Support for 64-bit floating-point matrix multiply
2472                 instructions. This extension can be enabled using the
2473                 +f64mm option extension, which likewise has the effect of
2474                 enabling SVE.
2475     * SVE2 is now supported through ACLE intrinsics and (to a limited
2476       extent) through autovectorization. It can be enabled through the
2477       +sve2 option extension (for example, -march=armv8.5-a+sve2).
2478       Additional extensions can be enabled through +sve2-sm4, +sve2-aes,
2479       +sve2-sha3 and +sve2-bitperm.
2480     * Support has been added for the following processors (GCC
2481       identifiers in parentheses):
2482          + Arm Cortex-A77 (cortex-a77).
2483          + Arm Cortex-A76AE (cortex-a76ae).
2484          + Arm Cortex-A65 (cortex-a65).
2485          + Arm Cortex-A65AE (cortex-a65ae).
2486          + Arm Cortex-A34 (cortex-a34).
2487          + Marvell ThunderX3 (thunderx3t110).
2488       The GCC identifiers can be used as arguments to the -mcpu or -mtune
2489       options, for example: -mcpu=cortex-a77 or -mtune=cortex-a65ae or as
2490       arguments to the equivalent target attributes and pragmas.
2491
2492  arm
2493
2494     * Support for the FDPIC ABI has been added. It uses 64-bit function
2495       descriptors to represent pointers to functions, and enables code
2496       sharing on MMU-less systems. The corresponding target triple is
2497       arm-uclinuxfdpiceabi, and the C library is uclibc-ng.
2498     * Support has been added for the Arm EABI on NetBSD through the
2499       arm*-*-netbsdelf-*eabi* triplet.
2500     * The handling of 64-bit integer operations has been significantly
2501       reworked and improved leading to improved performance and reduced
2502       stack usage when using 64-bit integral data types. The option
2503       -mneon-for-64bits is now deprecated and will be removed in a future
2504       release.
2505     * Support has been added for the following processors (GCC
2506       identifiers in parentheses):
2507          + Arm Cortex-A77 (cortex-a77).
2508          + Arm Cortex-A76AE (cortex-a76ae).
2509          + Arm Cortex-M35P (cortex-m35p).
2510          + Arm Cortex-M55 (cortex-m55).
2511       The GCC identifiers can be used as arguments to the -mcpu or -mtune
2512       options, for example: -mcpu=cortex-a77 or -mtune=cortex-m35p.
2513     * Support has been extended for the ACLE [43]data-processing
2514       intrinsics to include 32-bit SIMD, saturating arithmetic, 16-bit
2515       multiplication and other related intrinsics aimed at DSP algorithm
2516       optimization.
2517     * Support for -mpure-code in Thumb-1 (v6m) has been added: this
2518       M-profile feature is no longer restricted to targets with MOVT. For
2519       example, -mcpu=cortex-m0 now supports this option.
2520     * Support for the [44]Armv8.1-M Mainline Architecture has been added.
2521          + Armv8.1-M Mainline can be enabled by using the
2522            -march=armv8.1-m.main command-line option.
2523     * Support for the [45]MVE beta ACLE intrinsics has been added. These
2524       intrinsics can be enabled by including the arm_mve.h header file
2525       and passing the +mve or +mve.fp option extensions (for example:
2526       -march=armv8.1-m.main+mve).
2527     * Support for the Custom Datapath Extension beta ACLE [46]intrinsics
2528       has been added.
2529     * Support for Armv8.1-M Mainline Security Extensions architecture has
2530       been added. The -mcmse option, when used in combination with an
2531       Armv8.1-M Mainline architecture (for example: -march=armv8.1-m.main
2532       -mcmse), now leads to the generation of improved code sequences
2533       when changing security states.
2534
2535  AMD Radeon (GCN)
2536
2537     * Code generation and in particular vectorization support have been
2538       much improved.
2539
2540  ARC
2541
2542     * The interrupt service routine functions save all used registers,
2543       including extension registers and auxiliary registers used by Zero
2544       Overhead Loops.
2545     * Improve code size by using multiple short instructions instead of a
2546       single long mov or ior instruction when its long immediate constant
2547       is known.
2548     * Fix usage of the accumulator register for ARC600.
2549     * Fix issues with uncached attribute.
2550     * Remove -mq-class option.
2551     * Improve 64-bit integer addition and subtraction operations.
2552
2553  AVR
2554
2555     * Support for the XMEGA-like devices
2556
2557     ATtiny202, ATtiny204, ATtiny402, ATtiny404, ATtiny406, ATtiny804,
2558     ATtiny806, ATtiny807, ATtiny1604, ATtiny1606, ATtiny1607, ATmega808,
2559     ATmega809, ATmega1608, ATmega1609, ATmega3208, ATmega3209,
2560     ATmega4808, ATmega4809
2561       has been added.
2562     * A new command-line option -nodevicespecs has been added. It allows
2563       to provide a custom device-specs file by means of
2564
2565     avr-gcc -nodevicespecs -specs=my-spec-file <options>
2566       and without the need to provide options -B and -mmcu=. See [47]AVR
2567       command-line options for details. This feature is also available in
2568       GCC 9.3+ and GCC 8.4+.
2569     * New command-line options -mdouble=[32,64] and -mlong-double=[32,64]
2570       have been added. They allow to choose the size (in bits) of the
2571       double and long double types, respectively. Whether or not the
2572       mentioned layouts are available, whether the options act as a
2573       multilib option, and the default for either option are controlled
2574       by the new [48]AVR configure options --with-double= and
2575       --with-long-double=.
2576     * A new configure option --with-libf7= has been added. It controls to
2577       which level avr-libgcc provides 64-bit floating point support by
2578       means of [49]Libf7.
2579     * A new configure option --with-double-comparison= has been added.
2580       It's unlikely you need to set this option by hand.
2581
2582  IA-32/x86-64
2583
2584     * Support to expand __builtin_roundeven into the appropriate SSE 4.1
2585       instruction has been added.
2586     * New ISA extension support for Intel ENQCMD was added to GCC. ENQCMD
2587       intrinsics are available via the -menqcmd compiler switch.
2588     * GCC now supports the Intel CPU named Cooperlake through
2589       -march=cooperlake. The switch enables the AVX512BF16 ISA
2590       extensions.
2591     * GCC now supports the Intel CPU named Tigerlake through
2592       -march=tigerlake. The switch enables the MOVDIRI MOVDIR64B
2593       AVX512VP2INTERSECT ISA extensions.
2594
2595  MIPS
2596
2597     * The mips*-*-linux* targets now mark object files with appropriate
2598       GNU-stack note, facilitating use of non-executable stack hardening
2599       on GNU/Linux. The soft-float targets have this feature enabled by
2600       default, while for hard-float targets it is required for GCC to be
2601       configured with --with-glibc-version=2.31 against glibc 2.31 or
2602       later.
2603
2604  PowerPC / PowerPC64 / RS6000
2605
2606     * Many vector builtins have been listed as deprecated in the
2607       [50]64-Bit ELF V2 ABI Specification for quite a number of years.
2608       The vector builtins listed in Tables A.8 through A.10 are now
2609       deprecated for GCC 10, and will likely be removed from support in
2610       GCC 11. Note that this does not result in any loss of function.
2611       These deprecated builtins generally provide somewhat nonsensical
2612       argument lists (for example, mixing signed, unsigned, and bool
2613       vector arguments arbitrarily), or are duplicate builtins that are
2614       inconsistent with the expected naming scheme. We expect that this
2615       will be unlikely to affect much if any code, and any required code
2616       changes will be trivial.
2617
2618  PRU
2619
2620     * A new back end targeting TI PRU I/O processors has been contributed
2621       to GCC.
2622
2623  RISC-V
2624
2625     * The riscv*-*-* targets now require GNU binutils version 2.30 or
2626       later, to support new assembly instructions produced by GCC.
2627
2628  V850
2629
2630     * The ABI for V850 nested functions has been changed. Previously the
2631       V850 port used %r20 for the static chain pointer, now the port uses
2632       %r19. This corrects a long standing latent bug in the v850 port
2633       where a call to a nested function would unexpectedly change the
2634       value in %r20.
2635
2636Operating Systems
2637
2638Improvements for plugin authors
2639
2640     * GCC diagnostics can now have a chain of events associated with
2641       them, describing a path through the code that triggers the problem.
2642       These can be printed by the diagnostics subsystem in various ways,
2643       controlled by the [51]-fdiagnostics-path-format option, or captured
2644       in JSON form via [52]-fdiagnostics-format=json.
2645     * GCC diagnostics can now be associated with [53]CWE weakness
2646       identifiers, which will appear on the standard error stream, and in
2647       the JSON output from [54]-fdiagnostics-format=json.
2648
2649Other significant improvements
2650
2651     * To allow inline expansion of both memcpy and memmove, the existing
2652       movmem instruction patterns used for non-overlapping memory copies
2653       have been renamed to cpymem. The movmem name is now used for
2654       overlapping memory moves, consistent with the library functions
2655       memcpy and memmove.
2656     * For many releases, when GCC emits a warning it prints the option
2657       controlling that warning. As of GCC 10, that option text is now a
2658       clickable hyperlink for the documentation of that option (assuming
2659       a [55]sufficiently capable terminal). This behavior can be
2660       controlled via a new [56]-fdiagnostics-urls option (along with
2661       various environment variables and heuristics documented with that
2662       option).
2663
2664GCC 10.1
2665
2666   This is the [57]list of problem reports (PRs) from GCC's bug tracking
2667   system that are known to be fixed in the 10.1 release. This list might
2668   not be complete (that is, it is possible that some PRs that have been
2669   fixed are not listed here).
2670
2671GCC 10.2
2672
2673   This is the [58]list of problem reports (PRs) from GCC's bug tracking
2674   system that are known to be fixed in the 10.2 release. This list might
2675   not be complete (that is, it is possible that some PRs that have been
2676   fixed are not listed here).
2677
2678GCC 10.3
2679
2680   This is the [59]list of problem reports (PRs) from GCC's bug tracking
2681   system that are known to be fixed in the 10.3 release. This list might
2682   not be complete (that is, it is possible that some PRs that have been
2683   fixed are not listed here).
2684
2685  Target Specific Changes
2686
2687    AArch64
2688
2689     * A bug with the Random Number intrinsics in the arm_acle.h header
2690       that resulted in an incorrect status result being returned has been
2691       fixed.
2692     * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune
2693       options are -mcpu=a64fx and -mtune=a64fx respectively. In
2694       particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and
2695       tunes the code for the A64FX. This includes tuning the SVE code,
2696       although by default the code is still length-agnostic and so works
2697       for all SVE implementations. Adding -msve-vector-bits=512 makes the
2698       code specific to 512-bit SVE.
2699
2700  x86-64
2701
2702     * GCC 10.3 supports AMD CPUs based on the znver3 core via
2703       -march=znver3.
2704
2705GCC 10.4
2706
2707   This is the [60]list of problem reports (PRs) from GCC's bug tracking
2708   system that are known to be fixed in the 10.4 release. This list might
2709   not be complete (that is, it is possible that some PRs that have been
2710   fixed are not listed here).
2711
2712  Target Specific Changes
2713
2714    x86-64
2715
2716     * The x86-64 ABI of passing and returning structures with a 64-bit
2717       integer vector changed in GCC 10.1 when MMX is disabled. Disabling
2718       MMX no longer changes how they are passed nor returned. This ABI
2719       change is now diagnosed with -Wpsabi.
2720
2721
2722    For questions related to the use of GCC, please consult these web
2723    pages and the [61]GCC manuals. If that fails, the
2724    [62]gcc-help@gcc.gnu.org mailing list might help. Comments on these
2725    web pages and the development of GCC are welcome on our developer
2726    list at [63]gcc@gcc.gnu.org. All of [64]our lists have public
2727    archives.
2728
2729   Copyright (C) [65]Free Software Foundation, Inc. Verbatim copying and
2730   distribution of this entire article is permitted in any medium,
2731   provided this notice is preserved.
2732
2733   These pages are [66]maintained by the GCC team. Last modified
2734   2023-03-11.
2735
2736References
2737
2738   1. http://gcc.gnu.org/gcc-10/porting_to.html
2739   2. http://gcc.gnu.org/onlinedocs/index.html#current
2740   3. http://gcc.gnu.org/gcc-10/changes.html#empty_base
2741   4. https://www.mpfr.org/
2742   5. http://gcc.gnu.org/gcc-10/changes.html#iterator_base
2743   6. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/cpp/_005f_005fhas_005fbuiltin.html#g_t_005f_005fhas_005fbuiltin
2744   7. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fno-allocation-dce
2745   8. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-partial-training
2746   9. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-reproducible
2747  10. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-generate
2748  11. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-prefix-path
2749  12. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Static-Analyzer-Options.html
2750  13. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-finline-functions
2751  14. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/lto-dump.html
2752  15. https://facebook.github.io/zstd/
2753  16. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-values
2754  17. https://www.openacc.org/
2755  18. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-10
2756  19. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/libgomp/#toc-Enabling-OpenACC-1
2757  20. https://www.openmp.org/specifications/
2758  21. https://gcc.gnu.org/wiki/Offloading
2759  22. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstring-compare
2760  23. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wzero-length-bounds
2761  24. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warray-bounds
2762  25. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wformat-overflow
2763  26. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wrestrict
2764  27. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wreturn-local-addr
2765  28. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstringop-overflow
2766  29. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warith-conversion
2767  30. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-tags
2768  31. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-tags
2769  32. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94383
2770  33. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94711
2771  34. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94707
2772  35. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94704
2773  36. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94706
2774  37. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92285
2775  38. https://www.openmp.org/specifications/
2776  39. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_major
2777  40. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_minor
2778  41. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_patchlevel
2779  42. https://gcc.gnu.org/onlinedocs/jit/topics/types.html#c.gcc_jit_context_new_bitfield
2780  43. https://developer.arm.com/documentation/101028/0009/Data-processing-intrinsics
2781  44. https://developer.arm.com/Architectures/M-Profile%20Architecture
2782  45. https://developer.arm.com/architectures/instruction-sets/intrinsics/
2783  46. https://developer.arm.com/documentation/101028/0010/Custom-Datapath-Extension
2784  47. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/AVR-Options.html#index-nodevicespecs
2785  48. https://gcc.gnu.org/install/configure.html#avr
2786  49. https://gcc.gnu.org/wiki/avr-gcc#Libf7
2787  50. https://openpowerfoundation.org/?resource_lib=64-bit-elf-v2-abi-specification-power-architecture
2788  51. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-path-format
2789  52. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format
2790  53. https://cwe.mitre.org/
2791  54. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format
2792  55. https://gist.github.com/egmontkob/eb114294efbcd5adb1944c9f3cb5feda
2793  56. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-urls
2794  57. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.0
2795  58. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.2
2796  59. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.3
2797  60. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.4
2798  61. https://gcc.gnu.org/onlinedocs/
2799  62. mailto:gcc-help@gcc.gnu.org
2800  63. mailto:gcc@gcc.gnu.org
2801  64. https://gcc.gnu.org/lists.html
2802  65. https://www.fsf.org/
2803  66. https://gcc.gnu.org/about.html
2804======================================================================
2805http://gcc.gnu.org/gcc-9/index.html
2806
2807                              GCC 9 Release Series
2808
2809   (This release series is no longer supported.)
2810
2811   May 27, 2022
2812
2813   The [1]GNU project and the GCC developers are pleased to announce the
2814   release of GCC 9.5.
2815
2816   This release is a bug-fix release, containing fixes for regressions in
2817   GCC 9.4 relative to previous releases of GCC.
2818
2819Release History
2820
2821   GCC 9.5
2822          May 27, 2022 ([2]changes, [3]documentation)
2823
2824   GCC 9.4
2825          June 1, 2021 ([4]changes, [5]documentation)
2826
2827   GCC 9.3
2828          Mar 12, 2020 ([6]changes, [7]documentation)
2829
2830   GCC 9.2
2831          Aug 12, 2019 ([8]changes, [9]documentation)
2832
2833   GCC 9.1
2834          May 3, 2019 ([10]changes, [11]documentation)
2835
2836References and Acknowledgements
2837
2838   GCC used to stand for the GNU C Compiler, but since the compiler
2839   supports several other languages aside from C, it now stands for the
2840   GNU Compiler Collection.
2841
2842   A list of [12]successful builds is updated as new information becomes
2843   available.
2844
2845   The GCC developers would like to thank the numerous people that have
2846   contributed new features, improvements, bug fixes, and other changes as
2847   well as test results to GCC. This [13]amazing group of volunteers is
2848   what makes GCC successful.
2849
2850   For additional information about GCC please refer to the [14]GCC
2851   project web site or contact the [15]GCC development mailing list.
2852
2853   To obtain GCC please use [16]our mirror sites or [17]our version
2854   control system.
2855
2856
2857    For questions related to the use of GCC, please consult these web
2858    pages and the [18]GCC manuals. If that fails, the
2859    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
2860    web pages and the development of GCC are welcome on our developer
2861    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
2862    archives.
2863
2864   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
2865   distribution of this entire article is permitted in any medium,
2866   provided this notice is preserved.
2867
2868   These pages are [23]maintained by the GCC team. Last modified
2869   2022-10-26.
2870
2871References
2872
2873   1. http://www.gnu.org/
2874   2. http://gcc.gnu.org/gcc-9/changes.html
2875   3. http://gcc.gnu.org/onlinedocs/9.5.0/
2876   4. http://gcc.gnu.org/gcc-9/changes.html
2877   5. http://gcc.gnu.org/onlinedocs/9.4.0/
2878   6. http://gcc.gnu.org/gcc-9/changes.html
2879   7. http://gcc.gnu.org/onlinedocs/9.3.0/
2880   8. http://gcc.gnu.org/gcc-9/changes.html
2881   9. http://gcc.gnu.org/onlinedocs/9.2.0/
2882  10. http://gcc.gnu.org/gcc-9/changes.html
2883  11. http://gcc.gnu.org/onlinedocs/9.1.0/
2884  12. http://gcc.gnu.org/gcc-9/buildstat.html
2885  13. http://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Contributors.html
2886  14. http://gcc.gnu.org/index.html
2887  15. mailto:gcc@gcc.gnu.org
2888  16. http://gcc.gnu.org/mirrors.html
2889  17. http://gcc.gnu.org/git.html
2890  18. https://gcc.gnu.org/onlinedocs/
2891  19. mailto:gcc-help@gcc.gnu.org
2892  20. mailto:gcc@gcc.gnu.org
2893  21. https://gcc.gnu.org/lists.html
2894  22. https://www.fsf.org/
2895  23. https://gcc.gnu.org/about.html
2896======================================================================
2897http://gcc.gnu.org/gcc-9/changes.html
2898
2899                              GCC 9 Release Series
2900                        Changes, New Features, and Fixes
2901
2902   This page is a "brief" summary of some of the huge number of
2903   improvements in GCC 9. You may also want to check out our [1]Porting to
2904   GCC 9 page and the [2]full GCC documentation.
2905
2906Caveats
2907
2908     * On Arm targets (arm*-*-*), [3]a bug in the implementation of the
2909       procedure call standard (AAPCS) in the GCC 6, 7 and 8 releases has
2910       been fixed: a structure containing a bit-field based on a 64-bit
2911       integral type and where no other element in a structure required
2912       64-bit alignment could be passed incorrectly to functions. This is
2913       an ABI change. If the option -Wpsabi is enabled (on by default) the
2914       compiler will emit a diagnostic note for code that might be
2915       affected.
2916     * Support for a number of older systems and recently unmaintained or
2917       untested target ports of GCC has been declared obsolete in GCC 9.
2918       Unless there is activity to revive them, the next release of GCC
2919       will have their sources permanently removed.
2920       The following ports for individual systems on particular
2921       architectures have been obsoleted:
2922          + Solaris 10 (*-*-solaris2.10). Details can be found in the
2923            [4]announcement.
2924          + Cell Broadband Engine SPU (spu*-*-*). Details can be found in
2925            the [5]announcement.
2926     * A change to the C++ std::rotate algorithm in GCC 9.1.0 can cause
2927       ABI incompatibilities with object files compiled with other
2928       versions of GCC. If the std::rotate algorithm is called with an
2929       empty range then it might cause a divide-by-zero error (as a SIGFPE
2930       signal) and crash. The change has been reverted for GCC 9.2.0 and
2931       future releases. For more details see [6]Bug 90920. The problem can
2932       be avoided by recompiling any objects that might call std::rotate
2933       with an empty range, so that the GCC 9.1.0 definition of
2934       std::rotate is not used.
2935     * The automatic template instantiation at link time ([7]-frepo) has
2936       been deprecated and will be removed in a future release.
2937     * The --with-default-libstdcxx-abi=gcc4-compatible configure option
2938       is broken in the 9.1 and 9.2 releases, producing a shared library
2939       with missing symbols (see [8]Bug 90361). As a workaround, configure
2940       without that option and build GCC as normal, then edit the
2941       installed <bits/c++config.h> headers to define the
2942       _GLIBCXX_USE_CXX11_ABI macro to 0.
2943
2944General Improvements
2945
2946   The following GCC command line options have been introduced or
2947   improved.
2948     * All command line options that take a byte-size argument accept
2949       64-bit integers as well as standard SI and IEC suffixes such as kb
2950       and KiB, MB and MiB, or GB and GiB denoting the corresponding
2951       multiples of bytes. See [9]Invoking GCC for more.
2952     * A new option [10]-flive-patching=[inline-only-static|inline-clone]
2953       generates code suitable for live patching. At the same time it
2954       provides multiple-level control over IPA optimizations. See the
2955       user guide for more details.
2956     * A new option, --completion, has been added to provide more fine
2957       option completion in a shell. It is intended to be used by
2958       Bash-completion.
2959     * GCC's diagnostics now print source code with a left margin showing
2960       line numbers, configurable with
2961       [11]-fno-diagnostics-show-line-numbers.
2962       GCC's diagnostics can also now label regions of the source code to
2963       show pertinent information, such as the types within an expression.
2964$ g++ t.cc
2965t.cc: In function 'int test(const shape&, const shape&)':
2966t.cc:15:4: error: no match for 'operator+' (operand types are 'boxed_value<doubl
2967e>' and 'boxed_value<double>')
2968   14 |   return (width(s1) * height(s1)
2969      |           ~~~~~~~~~~~~~~~~~~~~~~
2970      |                     |
2971      |                     boxed_value<[...]>
2972   15 |    + width(s2) * height(s2));
2973      |    ^ ~~~~~~~~~~~~~~~~~~~~~~
2974      |                |
2975      |                boxed_value<[...]>
2976
2977       These labels can be disabled via [12]-fno-diagnostics-show-labels.
2978     * A new option [13]-fdiagnostics-format=json has been introduced for
2979       emitting diagnostics in a machine-readable format.
2980     * The alignment-related options [14]-falign-functions,
2981       [15]-falign-labels, [16]-falign-loops, and [17]-falign-jumps
2982       received support for a secondary alignment (e.g.
2983       -falign-loops=n:m:n2:m2).
2984     * New pair of profiling options ([18]-fprofile-filter-files and
2985       [19]-fprofile-exclude-files) has been added. The options help to
2986       filter which source files are instrumented.
2987     * AddressSanitizer generates more compact redzones for automatic
2988       variables. That helps to reduce memory footprint of a sanitized
2989       binary.
2990     * Numerous improvements have been made to the output of
2991       [20]-fopt-info.
2992       Messages are now prefixed with optimized, missed, or note, rather
2993       than the old behavior of all being prefixed with note.
2994       The output from -fopt-info can now contain information on inlining
2995       decisions:
2996$ g++ -c inline.cc -O2 -fopt-info-inline-all
2997inline.cc:24:11: note: Considering inline candidate void foreach(T, T, void (*)(
2998E)) [with T = char**; E = char*]/2.
2999inline.cc:24:11: optimized:  Inlining void foreach(T, T, void (*)(E)) [with T =
3000char**; E = char*]/2 into int main(int, char**)/1.
3001inline.cc:19:12: missed:   not inlinable: void inline_me(char*)/0 -> int std::pu
3002ts(const char*)/3, function body not available
3003inline.cc:13:8: optimized:  Inlined void inline_me(char*)/4 into int main(int, c
3004har**)/1 which now has time 127.363637 and size 11, net change of +0.
3005Unit growth for small function inlining: 16->16 (0%)
3006
3007Inlined 2 calls, eliminated 1 functions
3008
3009
3010       The output from the vectorizer has been rationalized so that failed
3011       attempts to vectorize a loop are displayed in the form
3012    [LOOP-LOCATION]: couldn't vectorize this loop
3013    [PROBLEM-LOCATION]: because of [REASON]
3014
3015       rather than an exhaustive log of all decisions made by the
3016       vectorizer. For example:
3017$ gcc -c v.c -O3 -fopt-info-all-vec
3018v.c:7:3: missed: couldn't vectorize loop
3019v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" :  :  : "me
3020mory");
3021v.c:3:6: note: vectorized 0 loops in function.
3022v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" :  :  : "me
3023mory");
3024
3025       The old behavior can be obtained via a new -internals suboption of
3026       -fopt-info.
3027     * A new option, [21]-fsave-optimization-record has been added, which
3028       writes a SRCFILE.opt-record.json.gz file describing the
3029       optimization decisions made by GCC. This is similar to the output
3030       of -fopt-info, but with additional metadata such as the inlining
3031       chain, and profile information (if available).
3032     * Inter-procedural propagation of stack alignment can now be
3033       controlled by [22]-fipa-stack-alignment.
3034     * Propagation of addressability, readonly and writeonly flags on
3035       static variables can now be controlled by
3036       [23]-fipa-reference-addressable.
3037
3038   The following built-in functions have been introduced.
3039     * [24]__builtin_expect_with_probability to provide branch prediction
3040       probability hints to the optimizer.
3041     * [25]__builtin_has_attribute determines whether a function, type, or
3042       variable has been declared with some attribute.
3043     * [26]__builtin_speculation_safe_value can be used to help mitigate
3044       against unsafe speculative execution.
3045
3046   The following attributes have been introduced.
3047     * The [27]copy function attribute has been added. The attribute can
3048       also be applied to type definitions and to variable declarations.
3049
3050   A large number of improvements to code generation have been made,
3051   including but not limited to the following.
3052     * Switch expansion has been improved by using a different strategy
3053       (jump table, bit test, decision tree) for a subset of switch cases.
3054     * A linear function expression defined as a switch statement can be
3055       transformed by [28]-ftree-switch-conversion. For example:
3056
3057int
3058foo (int how)
3059{
3060  switch (how) {
3061    case 2: how = 205; break;
3062    case 3: how = 305; break;
3063    case 4: how = 405; break;
3064    case 5: how = 505; break;
3065    case 6: how = 605; break;
3066  }
3067  return how;
3068}
3069
3070       can be transformed into 100 * how + 5 (for values defined in the
3071       switch statement).
3072     * Inter-procedural optimization improvements:
3073          + Inliner defaults was tuned to better suits modern C++
3074            codebases especially when built with link time optimizations.
3075            New parameters max-inline-insns-small, max-inline-insns-size,
3076            uninlined-function-insns, uninlined-function-time,
3077            uninlined-thunk-insns, and uninlined-thunk-time were added.
3078          + Hot/cold partitioning is now more precise and aggressive.
3079          + Improved scalability for very large translation units
3080            (especially when link-time optimizing large programs).
3081     * Profile driven optimization improvements:
3082          + [29]-fprofile-use now enables [30]-fversion-loops-for-strides,
3083            [31]-floop-interchange, [32]-floop-unroll-and-jam,
3084            [33]-ftree-loop-distribution.
3085          + Streaming of counter histograms was removed. This reduces the
3086            size of profile files. Histogram is computed on the fly with
3087            link-time optimization. Parameter hot-bb-count-ws-permille was
3088            reduced from 999 to 990 to account for more precise
3089            histograms.
3090     * Link-time optimization improvements:
3091          + Types are now simplified prior streaming resulting in
3092            significant reductions of the LTO object files, link-time
3093            memory use, and improvements of link-time parallelism.
3094          + Default number of partitions (--param lto-partitions) was
3095            increased from 32 to 128 enabling effective use of CPUs with
3096            more than 32 hyperthreads. --param
3097            lto-max-streaming-parallelism can now be used to control
3098            number of streaming processes.
3099          + Warnings on C++ One Decl Rule violations (-Wodr) are now more
3100            informative and produce fewer redundant results.
3101       Overall compile time of Firefox 66 and LibreOffice 6.2.3 on an
3102       8-core machine was reduced by about 5% compared to GCC 8.3, and the
3103       size of LTO object files by 7%. LTO link-time improves by 11% on an
3104       8-core machine and scales significantly better for more parallel
3105       build environments. The serial stage of the link-time optimization
3106       is 28% faster consuming 20% less memory. The parallel stage now
3107       scales to up to 128 partitions rather than 32 and reduces memory
3108       use for every worker by 30%.
3109
3110   The following improvements to the gcov command-line utility have been
3111   made.
3112     * The gcov tool received a new option [34]--use-hotness-colors (-q)
3113       that can provide perf-like coloring of hot functions.
3114     * The gcov tool has changed its intermediate format to a new JSON
3115       format.
3116
3117New Languages and Language specific improvements
3118
3119   [35]OpenACC support in C, C++, and Fortran continues to be maintained
3120   and improved. Most of the OpenACC 2.5 specification is implemented. See
3121   the [36]implementation status section on the OpenACC wiki page for
3122   further information.
3123
3124  C family
3125
3126     * Version 5.0 of the [37]OpenMP specification is now partially
3127       supported in the C and C++ compilers. For details which features of
3128       OpenMP 5.0 are and which are not supported in the GCC 9 release see
3129       [38]this mail.
3130     * New extensions:
3131          + [39]__builtin_convertvector built-in for vector conversions
3132            has been added.
3133     * New warnings:
3134          + [40]-Waddress-of-packed-member, enabled by default, warns
3135            about an unaligned pointer value from the address of a packed
3136            member of a struct or union.
3137     * Enhancements to existing warnings:
3138          + [41]-Warray-bounds detects more instances of out-of-bounds
3139            indices.
3140          + [42]-Wattribute-alias also detects attribute mismatches
3141            between alias declarations and their targets, in addition to
3142            mismatches between their types.
3143          + [43]-Wformat-overflow and [44]-Wformat-truncation have been
3144            extended to all formatted input/output functions (where
3145            applicable) and enhanced to detect a subset of instances of
3146            reading past the end of unterminated constant character arrays
3147            in %s directives.
3148          + [45]-Wmissing-attributes detects instances of missing function
3149            attributes on declarations of aliases and weak references.
3150          + [46]-Wstringop-truncation also detects a subset of instances
3151            of reading past the end of unterminated constant character
3152            arrays,
3153     * If a macro is used with the wrong argument count, the C and C++
3154       front ends now show the definition of that macro via a note.
3155     * The spelling corrector now considers transposed letters, and the
3156       threshold for similarity has been tightened, to avoid nonsensical
3157       suggestions.
3158
3159  C
3160
3161     * There is now experimental support for -std=c2x, to select support
3162       for the upcoming C2X revision of the ISO C standard. This standard
3163       is in the early stages of development and the only feature
3164       supported in GCC 9 is _Static_assert with a single argument
3165       (support for _Static_assert with two arguments was added in C11 and
3166       GCC 4.6). There are also new options -std=gnu2x, for C2X with GNU
3167       extensions, and -Wc11-c2x-compat, to warn for uses of features
3168       added in C2X (such warnings are also enabled by use of -Wpedantic
3169       if not using -std=c2x or -std=gnu2x).
3170     * New warnings:
3171          + [47]-Wabsolute-value warns for calls to standard functions
3172            that compute the absolute value of an argument when a more
3173            appropriate standard function is available. For example,
3174            calling abs(3.14) triggers the warning because the appropriate
3175            function to call to compute the absolute value of a double
3176            argument is fabs. The option also triggers warnings when the
3177            argument in a call to such a function has an unsigned type.
3178            This warning can be suppressed with an explicit type cast and
3179            it is also enabled by -Wextra.
3180
3181  C++
3182
3183     * New warnings:
3184          + [48]-Wdeprecated-copy, implied by -Wextra, warns about the
3185            C++11 deprecation of implicitly declared copy constructor and
3186            assignment operator if one of them is user-provided.
3187            -Wdeprecated-copy-dtor also warns if the destructor is
3188            user-provided, as specified in C++11.
3189          + [49]-Winit-list-lifetime, on by default, warns about uses of
3190            std::initializer_list that are likely to result in a dangling
3191            pointer, such as returning or assigning from a temporary list.
3192          + [50]-Wredundant-move, implied by -Wextra, warns about
3193            redundant calls to std::move.
3194          + [51]-Wpessimizing-move, implied by -Wall, warns when a call to
3195            std::move prevents copy elision.
3196          + [52]-Wclass-conversion, on by default, warns when a conversion
3197            function will never be called due to the type it converts to.
3198     * The C++ front end has experimental support for some of the upcoming
3199       C++2a draft features with the -std=c++2a or -std=gnu++2a flags,
3200       including range-based for statements with initializer, default
3201       constructible and assignable stateless lambdas, lambdas in
3202       unevaluated contexts, language support for empty data members,
3203       allowing pack expansion in lambda init-capture, likely and unlikely
3204       attributes, class types in non-type template parameters, allowing
3205       virtual function calls in constant expressions, explicit(bool),
3206       std::is_constant_evaluated, nested inline namespaces, etc. For a
3207       full list of new features, see [53]the C++ status page.
3208     * The C++ front end now preserves source locations for literals,
3209       id-expression, and mem-initializer for longer. For example it is
3210       now able to pin-point the pertinent locations for bad
3211       initializations such as these
3212$ g++ -c bad-inits.cc
3213bad-inits.cc:10:14: error: cannot convert 'json' to 'int' in initialization
3214   10 |   { 3, json::object },
3215      |        ~~~~~~^~~~~~
3216      |              |
3217      |              json
3218bad-inits.cc:14:31: error: initializer-string for array of chars is too long [-f
3219permissive]
3220   14 | char buffers[3][5] = { "red", "green", "blue" };
3221      |                               ^~~~~~~
3222bad-inits.cc: In constructor 'X::X()':
3223bad-inits.cc:17:13: error: invalid conversion from 'int' to 'void*' [-fpermissiv
3224e]
3225   17 |   X() : one(42), two(42), three(42)
3226      |             ^~
3227      |             |
3228      |             int
3229
3230       rather than emitting the error at the final closing parenthesis or
3231       brace.
3232     * Error-reporting of overload resolution has been special-cased to
3233       make the case of a single failed candidate easier to read. For
3234       example:
3235$ g++ param-type-mismatch.cc
3236param-type-mismatch.cc: In function 'int test(int, const char*, float)':
3237param-type-mismatch.cc:8:32: error: cannot convert 'const char*' to 'const char*
3238*'
3239    8 |   return foo::member_1 (first, second, third);
3240      |                                ^~~~~~
3241      |                                |
3242      |                                const char*
3243param-type-mismatch.cc:3:46: note:   initializing argument 2 of 'static int foo:
3244:member_1(int, const char**, float)'
3245    3 |   static int member_1 (int one, const char **two, float three);
3246      |                                 ~~~~~~~~~~~~~^~~
3247
3248       highlights both the problematic argument, and the parameter that it
3249       can't be converted to.
3250     * Diagnostics involving binary operators now use color to distinguish
3251       the two operands, and label them separately (as per the example of
3252       source labelling above).
3253     * Diagnostics involving function calls now highlight the pertinent
3254       parameter of the declaration in more places.
3255$ g++ bad-conversion.cc
3256bad-conversion.cc: In function 'void caller()':
3257bad-conversion.cc:9:14: error: cannot convert 'bool' to 'void*'
3258    9 |   callee (0, false, 2);
3259      |              ^~~~~
3260      |              |
3261      |              bool
3262bad-conversion.cc:3:19: note:   initializing argument 2 of 'void callee(int, voi
3263d*, int)'
3264    3 | void callee (int, void *, int)
3265      |                   ^~~~~~
3266
3267     * The C++ front end's implementation of [54]-Wformat now shows
3268       precise locations within string literals, and underlines the
3269       pertinent arguments at bogus call sites (the C front end has been
3270       doing this since GCC 7). For example:
3271$ g++ -c bad-printf.cc -Wall
3272bad-printf.cc: In function 'void print_field(const char*, float, long int, long
3273int)':
3274bad-printf.cc:6:17: warning: field width specifier '*' expects argument of type
3275'int', but argument 3 has type 'long int' [-Wformat=]
3276    6 |   printf ("%s: %*ld ", fieldname, column - width, value);
3277      |                ~^~~               ~~~~~~~~~~~~~~
3278      |                 |                        |
3279      |                 int                      long int
3280bad-printf.cc:6:19: warning: format '%ld' expects argument of type 'long int', b
3281ut argument 4 has type 'double' [-Wformat=]
3282    6 |   printf ("%s: %*ld ", fieldname, column - width, value);
3283      |                ~~~^                               ~~~~~
3284      |                   |                               |
3285      |                   long int                        double
3286      |                %*f
3287
3288     * The C++ front end has gained new fix-it hints for forgetting the
3289       return *this; needed by various C++ operators:
3290$ g++ -c operator.cc
3291operator.cc: In member function 'boxed_ptr& boxed_ptr::operator=(const boxed_ptr
3292&)':
3293operator.cc:7:3: warning: no return statement in function returning non-void [-W
3294return-type]
3295    6 |     m_ptr = other.m_ptr;
3296  +++ |+    return *this;
3297    7 |   }
3298      |   ^
3299
3300       for when the compiler needs a typename:
3301$ g++ -c template.cc
3302template.cc:3:3: error: need 'typename' before 'Traits::type' because 'Traits' i
3303s a dependent scope
3304    3 |   Traits::type type;
3305      |   ^~~~~~
3306      |   typename
3307
3308       when trying to use an accessor member as if it were a data member:
3309$ g++ -c fncall.cc
3310fncall.cc: In function 'void hangman(const mystring&)':
3311fncall.cc:12:11: error: invalid use of member function 'int mystring::get_length
3312() const' (did you forget the '()' ?)
3313   12 |   if (str.get_length > 0)
3314      |       ~~~~^~~~~~~~~~
3315      |                     ()
3316
3317       for C++11's scoped enums:
3318$ g++ -c enums.cc
3319enums.cc: In function 'void json::test(const json::value&)':
3320enums.cc:12:26: error: 'STRING' was not declared in this scope; did you mean 'js
3321on::kind::STRING'?
3322   12 |     if (v.get_kind () == STRING)
3323      |                          ^~~~~~
3324      |                          json::kind::STRING
3325enums.cc:3:44: note: 'json::kind::STRING' declared here
3326    3 |   enum class kind { OBJECT, ARRAY, NUMBER, STRING, TRUE, FALSE, NULL_ };
3327      |                                            ^~~~~~
3328
3329       and a tweak to integrate the suggestions about misspelled members
3330       with that for accessors:
3331$ g++ -c accessor-fixit.cc
3332accessor-fixit.cc: In function 'int test(t*)':
3333accessor-fixit.cc:17:15: error: 'class t' has no member named 'ratio'; did you m
3334ean 'int t::m_ratio'? (accessible via 'int t::get_ratio() const')
3335   17 |   return ptr->ratio;
3336      |               ^~~~~
3337      |               get_ratio()
3338
3339       In addition, various diagnostics in the C++ front-end have been
3340       streamlined by consolidating the suggestion into the initial error,
3341       rather than emitting a follow-up note:
3342$ g++ typo.cc
3343typo.cc:5:13: error: 'BUFSIZE' was not declared in this scope; did you mean 'BUF
3344_SIZE'?
3345    5 | uint8_t buf[BUFSIZE];
3346      |             ^~~~~~~
3347      |             BUF_SIZE
3348
3349    Runtime Library (libstdc++)
3350
3351     * Improved support for C++17, including:
3352          + The C++17 implementation is no longer experimental.
3353          + Parallel algorithms and <execution> (requires [55]Thread
3354            Building Blocks 2018 or newer).
3355          + <memory_resource>.
3356          + Using the types and functions in <filesystem> does not require
3357            linking with -lstdc++fs now.
3358     * Improved experimental support for C++2a, including:
3359          + Type traits std::remove_cvref, std::unwrap_reference,
3360            std::unwrap_decay_ref, std::is_nothrow_convertible, and
3361            std::type_identity.
3362          + Headers <bit> and <version>.
3363          + Uniform container erasure (std::erase_if).
3364          + contains member of maps and sets.
3365          + String prefix and suffix checking (starts_with, ends_with).
3366          + Functions std::midpoint and std::lerp for interpolation.
3367          + std::bind_front.
3368          + std::visit<R>.
3369          + std::assume_aligned.
3370          + Uses-allocator construction utilities.
3371          + std::pmr::polymorphic_allocator<std::byte>.
3372          + Library support for char8_t type.
3373          + Destroying delete.
3374          + std::is_constant_evaluated() function.
3375     * Support for opening file streams with wide character paths on
3376       Windows
3377     * Incomplete support for the C++17 Filesystem library and the
3378       Filesystem TS on Windows.
3379     * Incomplete, experimental support for the Networking TS.
3380
3381  D
3382
3383     * Support for the D programming language has been added to GCC,
3384       implementing version 2.076 of the language and run-time library.
3385
3386  Fortran
3387
3388     * Asynchronous I/O is now fully supported. The program needs to be
3389       linked against the pthreads library to use it, otherwise the I/O is
3390       done synchronously. For systems which do not support POSIX
3391       condition variables, such as AIX, all I/O is still done
3392       synchronously.
3393     * The BACK argument for MINLOC and MAXLOC has been implemented.
3394     * The FINDLOC intrinsic function has been implemented.
3395     * The IS_CONTIGUOUS intrinsic function has been implemented.
3396     * Direct access to the real and imaginary parts of a complex variable
3397       via c%re and c%im has been implemented.
3398     * Type parameter inquiry via str%len and a%kind has been implemented.
3399     * C descriptors and the ISO_Fortran_binding.h source file have been
3400       implemented.
3401     * The MAX and MIN intrinsics are no longer guaranteed to return any
3402       particular value in case one of the arguments is a NaN. Note that
3403       this conforms to the Fortran standard and to what other Fortran
3404       compilers do. If there is a need to handle that case in some
3405       specific way, one needs to explicitly check for NaN's before
3406       calling MAX or MIN, e.g. by using the IEEE_IS_NAN function from the
3407       intrinsic module IEEE_ARITHMETIC.
3408     * A new command-line option [56]-fdec-include, set also by the
3409       [57]-fdec option, has been added to increase compatibility with
3410       legacy code. With this option, an INCLUDE directive is also parsed
3411       as a statement, which allows the directive to be spread across
3412       multiple source lines with line continuations.
3413     * A new [58]BUILTIN directive, has been added. The purpose of the
3414       directive is to provide an API between the GCC compiler and the GNU
3415       C Library which would define vector implementations of math
3416       routines.
3417
3418  Go
3419
3420     * GCC 9 provides a complete implementation of the Go 1.12.2 user
3421       packages.
3422
3423libgccjit
3424
3425     * The libgccjit API gained a new entry point:
3426       [59]gcc_jit_context_add_driver_option.
3427
3428New Targets and Target Specific Improvements
3429
3430  AArch64 & Arm
3431
3432     * Support has been added for the following processors (GCC
3433       identifiers in parentheses):
3434          + Arm Cortex-A76 (cortex-a76).
3435          + Arm Cortex-A55/Cortex-A76 DynamIQ big.LITTLE
3436            (cortex-a76.cortex-a55).
3437          + Arm Neoverse N1 (neoverse-n1).
3438       The GCC identifiers can be used as arguments to the -mcpu or -mtune
3439       options, for example: -mcpu=cortex-a76 or
3440       -mtune=cortex-a76.cortex-a55 or as arguments to the equivalent
3441       target attributes and pragmas.
3442     * The Armv8.3-A complex number instructions are now supported via
3443       intrinsics when the option -march=armv8.3-a or equivalent is
3444       specified. For the half-precision floating-point variants of these
3445       instructions use the architecture extension flag +fp16, e.g.
3446       -march=armv8.3-a+fp16.
3447       The intrinsics are defined by the ACLE specification.
3448     * The Armv8.5-A architecture is now supported through the
3449       -march=armv8.5-a option.
3450     * The Armv8.5-A architecture also adds some security features that
3451       are optional to all older architecture versions. These are now
3452       supported and only affect the assembler.
3453          + Speculation Barrier instruction through the -march=armv8-a+sb
3454            option.
3455          + Execution and Data Prediction Restriction instructions through
3456            the -march=armv8-a+predres option.
3457          + Speculative Store Bypass Safe instruction through the
3458            -march=armv8-a+ssbs option. This does not require a compiler
3459            option for Arm and thus -march=armv8-a+ssbs is an
3460            AArch64-specific option.
3461
3462      AArch64 specific
3463
3464     * Support has been added for the Arm Neoverse E1 processor
3465       (-mcpu=neoverse-e1).
3466     * The AArch64 port now has support for stack clash protection using
3467       the [60]-fstack-clash-protection option. The probing interval/guard
3468       size can be set by using --param
3469       stack-clash-protection-guard-size=12|16. The value of this
3470       parameter must be in bytes represented as a power of two. The two
3471       supported values for this parameter are 12 (for a 4KiB size, 2^12)
3472       and 16 (for a 64KiB size, 2^16). The default value is 16 (64Kb) and
3473       can be changed at configure time using the flag
3474       --with-stack-clash-protection-guard-size=12|16.
3475     * The option -msign-return-address= has been deprecated. This has
3476       been replaced by the new -mbranch-protection= option. This new
3477       option can now be used to enable the return address signing as well
3478       as the new Branch Target Identification feature of Armv8.5-A
3479       architecture. For more information on the arguments accepted by
3480       this option, please refer to [61]AArch64-Options.
3481     * The following optional extensions to Armv8.5-A architecture are now
3482       supported and only affect the assembler.
3483          + Random Number Generation instructions through the
3484            -march=armv8.5-a+rng option.
3485          + Memory Tagging Extension through the -march=armv8.5-a+memtag
3486            option.
3487
3488      Arm specific
3489
3490     * Support for the deprecated Armv2 and Armv3 architectures and their
3491       variants has been removed. Their corresponding -march values and
3492       the -mcpu options that used these architectures have been removed.
3493     * Support for the Armv5 and Armv5E architectures (which have no known
3494       implementations) has been removed. Note that Armv5T, Armv5TE and
3495       Armv5TEJ architectures remain supported.
3496     * Corrected FPU configurations for Cortex-R7 and Cortex-R8 when using
3497       their respective -mcpu options.
3498
3499  AMD GCN
3500
3501     * A new back end targeting AMD GCN GPUs has been contributed to GCC.
3502       The implementation is currently limited to compiling
3503       single-threaded, stand-alone programs. Future versions will add
3504       support for offloading multi-threaded kernels via OpenMP and
3505       OpenACC. The following devices are supported (GCC identifiers in
3506       parentheses):
3507          + Fiji (fiji).
3508          + Vega 10 (gfx900).
3509
3510  ARC
3511
3512     * LRA is now on by default for the ARC target. This can be controlled
3513       by -mlra.
3514     * Add support for frame code-density and branch-and-index
3515       instructions.
3516
3517  C-SKY
3518
3519     * A new back end targeting C-SKY V2 processors has been contributed
3520       to GCC.
3521
3522  IA-32/x86-64
3523
3524     * Support of Intel MPX (Memory Protection Extensions) has been
3525       removed.
3526     * New ISA extension support for Intel PTWRITE was added to GCC.
3527       PTWRITE intrinsics are available via the -mptwrite compiler switch.
3528     * GCC now supports the Intel CPU named Cascade Lake with AVX512
3529       extensions through -march=cascadelake. The switch enables the
3530       following ISA extensions: AVX512F, AVX512VL, AVX512CD, AVX512BW,
3531       AVX512DQ, AVX512VNNI.
3532
3533  MIPS
3534
3535     * The Loongson loongson-mmi and loongson-ext extensions have been
3536       split from loongson3a:
3537          + loongson-mmi contains the Loongson MMI (MultiMedia extensions
3538            Instructions).
3539          + loongson-ext contains the Loongson EXT (EXTensions
3540            instructions).
3541     * The Loongson EXT2 (EXTensions R2 instructions) are now supported.
3542          + loongson-ext2 contains the Loongson EXT2 instructions.
3543       Command-line options-m[no-]loongson-mmi, -m[no-]loongson-ext, and
3544       -m[no-]loongson-ext2 enable or disable those extensions.
3545     * Support has been added for the following processors (GCC
3546       identifiers in parentheses):
3547          + Loongson 3A1000 (gs464) which enables loongson-mmi,
3548            loongson-ext by default.
3549          + Loongson 3A2000/3A3000 (gs464e) which enables loongson-mmi,
3550            loongson-ext, loongson-ext2 by default.
3551          + Loongson 2K1000 (gs264e) which enables loongson-ext,
3552            loongson-ext2, msa by default.
3553       The GCC identifiers can be used as arguments to the -mcpu and
3554       -mtune options (as in -mcpu=gs464 or -mtune=gs464e) or as arguments
3555       to the equivalent target attributes and pragmas.
3556
3557  OpenRISC
3558
3559     * A new back end targeting OpenRISC processors has been contributed
3560       to GCC.
3561
3562  S/390, System z, IBM z Systems
3563
3564     * Support for the arch13 architecture has been added. When using the
3565       -march=arch13 option, the compiler will generate code making use of
3566       the new instructions introduced with the vector enhancement
3567       facility 2 and the miscellaneous instruction extension facility 2.
3568       The -mtune=arch13 option enables arch13 specific instruction
3569       scheduling without making use of new instructions.
3570     * Builtins for the new vector instructions have been added and can be
3571       enabled using the -mzvector option.
3572     * Support for ESA architecture machines g5 and g6 is deprecated since
3573       GCC 6.1.0 and has been removed now.
3574     * When compiling with -march=z14 or higher GCC emits alignments hints
3575       on the vector load/store instructions (8 or 16 byte).
3576     * Functions now have a default alignment of 16 bytes. This helps with
3577       branch prediction effects.
3578     * -mfentry is now supported. As well as the mcount mechanism the
3579       __fentry__ is called before the function prologue. However, since
3580       just a single instruction is required to call __fentry__ the call
3581       sequence imposes a smaller overhead than mcount (4 instructions).
3582       The produced code is compatible only with newer glibc versions,
3583       which provide the __fentry__ symbol and do not clobber r0 when
3584       resolving lazily bound functions. -mfentry is only supported when
3585       generating 64 bit code and does not work with nested C functions.
3586     * The -mnop-mcount option can be used to emit NOP instructions
3587       instead of an mcount or fentry call stub.
3588     * With the -mrecord-mcount option a __mcount_loc section is generated
3589       containing pointers to each profiling call stub. This is useful for
3590       automatically patching in and out calls.
3591
3592Operating Systems
3593
3594  Solaris
3595
3596     * g++ now unconditionally enables large file support when compiling
3597       32-bit code.
3598     * Support for the AddressSanitizer and UndefinedBehaviorSanitizer has
3599       been merged from LLVM. For the moment, this only works for 32-bit
3600       code on both SPARC and x86.
3601     * An initial port of the D runtime library has been completed on
3602       Solaris 11/x86. It requires the use of GNU as. Solaris 11/SPARC
3603       support is still work-in-progress.
3604
3605  Windows
3606
3607     * A C++ Microsoft ABI bitfield layout bug, [62]PR87137 has been
3608       fixed. A non-field declaration could cause the current bitfield
3609       allocation unit to be completed, incorrectly placing a following
3610       bitfield into a new allocation unit. The Microsoft ABI is selected
3611       for:
3612          + Mingw targets
3613          + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields
3614            option is specified, or __attribute__((ms_struct)) is used
3615          + SuperH targets when the -mhitachi option is specified, or
3616            __attribute__((renesas)) is used
3617
3618Improvements for plugin authors
3619
3620     * GCC's diagnostic subsystem now has a way to logically group
3621       together related diagnostics, auto_diagnostic_group. Such
3622       diagnostics will be nested by the output of
3623       [63]-fdiagnostics-format=json.
3624     * GCC now has a set of [64]user experience guidelines for GCC, with
3625       information and advice on implementing new diagnostics.
3626
3627Other significant improvements
3628
3629     * GCC's internal "selftest" suite now runs for C++ as well as C (in
3630       debug builds of the compiler).
3631
3632GCC 9.1
3633
3634   This is the [65]list of problem reports (PRs) from GCC's bug tracking
3635   system that are known to be fixed in the 9.1 release. This list might
3636   not be complete (that is, it is possible that some PRs that have been
3637   fixed are not listed here).
3638
3639GCC 9.2
3640
3641   This is the [66]list of problem reports (PRs) from GCC's bug tracking
3642   system that are known to be fixed in the 9.2 release. This list might
3643   not be complete (that is, it is possible that some PRs that have been
3644   fixed are not listed here).
3645
3646GCC 9.3
3647
3648   This is the [67]list of problem reports (PRs) from GCC's bug tracking
3649   system that are known to be fixed in the 9.3 release. This list might
3650   not be complete (that is, it is possible that some PRs that have been
3651   fixed are not listed here).
3652
3653GCC 9.4
3654
3655   This is the [68]list of problem reports (PRs) from GCC's bug tracking
3656   system that are known to be fixed in the 9.4 release. This list might
3657   not be complete (that is, it is possible that some PRs that have been
3658   fixed are not listed here).
3659
3660  Target Specific Changes
3661
3662    AArch64
3663
3664     * The option -moutline-atomics has been added to aid deployment of
3665       the Large System Extensions (LSE) on GNU/Linux systems built with a
3666       baseline architecture targeting Armv8-A. When the option is
3667       specified code is emitted to detect the presence of LSE
3668       instructions at run time and use them for standard atomic
3669       operations. For more information please refer to the documentation.
3670     * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune
3671       options are -mcpu=a64fx and -mtune=a64fx respectively. In
3672       particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and
3673       tunes the code for the A64FX. This includes tuning the SVE code,
3674       although by default the code is still length-agnostic and so works
3675       for all SVE implementations. Adding -msve-vector-bits=512 makes the
3676       code specific to 512-bit SVE.
3677
3678GCC 9.5
3679
3680   This is the [69]list of problem reports (PRs) from GCC's bug tracking
3681   system that are known to be fixed in the 9.5 release. This list might
3682   not be complete (that is, it is possible that some PRs that have been
3683   fixed are not listed here).
3684
3685
3686    For questions related to the use of GCC, please consult these web
3687    pages and the [70]GCC manuals. If that fails, the
3688    [71]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3689    web pages and the development of GCC are welcome on our developer
3690    list at [72]gcc@gcc.gnu.org. All of [73]our lists have public
3691    archives.
3692
3693   Copyright (C) [74]Free Software Foundation, Inc. Verbatim copying and
3694   distribution of this entire article is permitted in any medium,
3695   provided this notice is preserved.
3696
3697   These pages are [75]maintained by the GCC team. Last modified
3698   2023-02-22.
3699
3700References
3701
3702   1. http://gcc.gnu.org/gcc-9/porting_to.html
3703   2. http://gcc.gnu.org/onlinedocs/index.html#current
3704   3. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88469
3705   4. https://gcc.gnu.org/ml/gcc/2018-10/msg00139.html
3706   5. https://gcc.gnu.org/ml/gcc/2019-04/msg00023.html
3707   6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90920
3708   7. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-frepo
3709   8. https://gcc.gnu.org/PR90361
3710   9. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-GCC.html#Invoking-GCC
3711  10. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-flive-patching
3712  11. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-line-numbers
3713  12. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-labels
3714  13. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format
3715  14. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-functions
3716  15. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-labels
3717  16. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-loops
3718  17. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-jumps
3719  18. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-filter-files
3720  19. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-exclude-files
3721  20. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fopt-info
3722  21. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fsave-optimization-record
3723  22. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-stack-alignment
3724  23. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-reference-addressable
3725  24. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fexpect_005fwith_005fprobability
3726  25. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fhas_005fattribute-1
3727  26. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fspeculation_005fsafe_005fvalue-1
3728  27. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Common-Function-Attributes.html#index-copy-function-attribute
3729  28. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-switch-conversion
3730  29. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fprofile-use
3731  30. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fversion-loops-for-strides
3732  31. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-interchange
3733  32. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-unroll-and-jam
3734  33. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-loop-distribution
3735  34. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-Gcov.html#Invoking-Gcov
3736  35. https://www.openacc.org/
3737  36. https://gcc.gnu.org/wiki/OpenACC/Implementation%20Status#status-9
3738  37. https://www.openmp.org/specifications/
3739  38. https://gcc.gnu.org/ml/gcc-patches/2018-11/msg00628.html
3740  39. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Vector-Extensions.html#index-_005f_005fbuiltin_005fconvertvector
3741  40. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Waddress-of-packed-member
3742  41. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Warray-bounds
3743  42. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wattribute-alias
3744  43. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-overflow
3745  44. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-truncation
3746  45. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wmissing-attributes
3747  46. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wstringop-truncation
3748  47. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wabsolute-value
3749  48. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-copy
3750  49. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Winit-list-lifetime
3751  50. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-move
3752  51. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wpessimizing-move
3753  52. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-conversion
3754  53. http://gcc.gnu.org/projects/cxx-status.html#cxx2a
3755  54. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat
3756  55. https://github.com/oneapi-src/oneTBB
3757  56. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec-include
3758  57. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec
3759  58. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/BUILTIN-directive.html#BUILTIN-directive
3760  59. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/jit/topics/contexts.html#gcc_jit_context_add_driver_option
3761  60. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fstack-protector
3762  61. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/AArch64-Options.html#AArch64-Options
3763  62. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137
3764  63. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format
3765  64. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gccint/User-Experience-Guidelines.html
3766  65. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.0
3767  66. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.2
3768  67. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.3
3769  68. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.4
3770  69. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.5
3771  70. https://gcc.gnu.org/onlinedocs/
3772  71. mailto:gcc-help@gcc.gnu.org
3773  72. mailto:gcc@gcc.gnu.org
3774  73. https://gcc.gnu.org/lists.html
3775  74. https://www.fsf.org/
3776  75. https://gcc.gnu.org/about.html
3777======================================================================
3778http://gcc.gnu.org/gcc-8/index.html
3779
3780                              GCC 8 Release Series
3781
3782   (This release series is no longer supported.)
3783
3784   May 14, 2021
3785
3786   The [1]GNU project and the GCC developers are pleased to announce the
3787   release of GCC 8.5.
3788
3789   This release is a bug-fix release, containing fixes for regressions in
3790   GCC 8.4 relative to previous releases of GCC.
3791
3792Release History
3793
3794   GCC 8.5
3795          May 14, 2021 ([2]changes, [3]documentation)
3796
3797   GCC 8.4
3798          Mar 4, 2020 ([4]changes, [5]documentation)
3799
3800   GCC 8.3
3801          Feb 22, 2019 ([6]changes, [7]documentation)
3802
3803   GCC 8.2
3804          Jul 26, 2018 ([8]changes, [9]documentation)
3805
3806   GCC 8.1
3807          May 2, 2018 ([10]changes, [11]documentation)
3808
3809References and Acknowledgements
3810
3811   GCC used to stand for the GNU C Compiler, but since the compiler
3812   supports several other languages aside from C, it now stands for the
3813   GNU Compiler Collection.
3814
3815   A list of [12]successful builds is updated as new information becomes
3816   available.
3817
3818   The GCC developers would like to thank the numerous people that have
3819   contributed new features, improvements, bug fixes, and other changes as
3820   well as test results to GCC. This [13]amazing group of volunteers is
3821   what makes GCC successful.
3822
3823   For additional information about GCC please refer to the [14]GCC
3824   project web site or contact the [15]GCC development mailing list.
3825
3826   To obtain GCC please use [16]our mirror sites or [17]our version
3827   control system.
3828
3829
3830    For questions related to the use of GCC, please consult these web
3831    pages and the [18]GCC manuals. If that fails, the
3832    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3833    web pages and the development of GCC are welcome on our developer
3834    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
3835    archives.
3836
3837   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
3838   distribution of this entire article is permitted in any medium,
3839   provided this notice is preserved.
3840
3841   These pages are [23]maintained by the GCC team. Last modified
3842   2022-10-26.
3843
3844References
3845
3846   1. http://www.gnu.org/
3847   2. http://gcc.gnu.org/gcc-8/changes.html
3848   3. http://gcc.gnu.org/onlinedocs/8.5.0/
3849   4. http://gcc.gnu.org/gcc-8/changes.html
3850   5. http://gcc.gnu.org/onlinedocs/8.4.0/
3851   6. http://gcc.gnu.org/gcc-8/changes.html
3852   7. http://gcc.gnu.org/onlinedocs/8.3.0/
3853   8. http://gcc.gnu.org/gcc-8/changes.html
3854   9. http://gcc.gnu.org/onlinedocs/8.2.0/
3855  10. http://gcc.gnu.org/gcc-8/changes.html
3856  11. http://gcc.gnu.org/onlinedocs/8.1.0/
3857  12. http://gcc.gnu.org/gcc-8/buildstat.html
3858  13. http://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Contributors.html
3859  14. http://gcc.gnu.org/index.html
3860  15. mailto:gcc@gcc.gnu.org
3861  16. http://gcc.gnu.org/mirrors.html
3862  17. http://gcc.gnu.org/git.html
3863  18. https://gcc.gnu.org/onlinedocs/
3864  19. mailto:gcc-help@gcc.gnu.org
3865  20. mailto:gcc@gcc.gnu.org
3866  21. https://gcc.gnu.org/lists.html
3867  22. https://www.fsf.org/
3868  23. https://gcc.gnu.org/about.html
3869======================================================================
3870http://gcc.gnu.org/gcc-8/changes.html
3871
3872                              GCC 8 Release Series
3873                        Changes, New Features, and Fixes
3874
3875   This page is a "brief" summary of some of the huge number of
3876   improvements in GCC 8. You may also want to check out our [1]Porting to
3877   GCC 8 page and the [2]full GCC documentation.
3878
3879Caveats
3880
3881     * Support for the obsolete SDB/coff debug info format has been
3882       removed. The option -gcoff no longer does anything.
3883     * The Cilk+ extensions to the C and C++ languages have been removed.
3884     * The MPX extensions to the C and C++ languages have been deprecated
3885       and will be removed in a future release.
3886     * The extension allowing arithmetic on std::atomic<void*> and types
3887       like std::atomic<R(*)()> has been deprecated.
3888     * The non-standard C++0x std::copy_exception function was removed.
3889       std::make_exception_ptr should be used instead.
3890     * Support for the powerpc*-*-*spe* target ports which have been
3891       recently unmaintained and untested in GCC has been declared
3892       obsolete in GCC 8 as [3]announced. Unless there is activity to
3893       revive them, the next release of GCC will have their sources
3894       permanently removed.
3895
3896General Improvements
3897
3898     * Inter-procedural optimization improvements:
3899          + Reworked run-time estimation metrics leading to more realistic
3900            guesses driving inliner and cloning heuristics.
3901          + The ipa-pure-const pass is extended to propagate the malloc
3902            attribute, and the corresponding warning option
3903            -Wsuggest-attribute=malloc emits a diagnostic for functions
3904            which can be annotated with the malloc attribute.
3905     * Profile driven optimization improvements:
3906          + New infrastructure for representing profiles (both statically
3907            guessed and profile feedback) which allows propagation of
3908            additional information about the reliability of the profile.
3909          + A number of improvements in the profile updating code solving
3910            problems found by new verification code.
3911          + Static detection of code which is not executed in a valid run
3912            of the program. This includes paths which trigger undefined
3913            behavior as well as calls to functions declared with the cold
3914            attribute. Newly the noreturn attribute does not imply all
3915            effects of cold to differentiate between exit (which is
3916            noreturn) and abort (which is in addition not executed in
3917            valid runs).
3918          + -freorder-blocks-and-partition, a pass splitting function
3919            bodies into hot and cold regions, is now enabled by default at
3920            -O2 and higher for x86 and x86-64.
3921     * Link-time optimization improvements:
3922          + We have significantly improved debug information on ELF
3923            targets using DWARF by properly preserving language-specific
3924            information. This allows for example the libstdc++
3925            pretty-printers to work with LTO optimized executables.
3926     * A new option -fcf-protection=[full|branch|return|none] is
3927       introduced to perform code instrumentation to increase program
3928       security by checking that target addresses of control-flow transfer
3929       instructions (such as indirect function call, function return,
3930       indirect jump) are valid. Currently the instrumentation is
3931       supported on x86 GNU/Linux targets only. See the user guide for
3932       further information about the option syntax and section "New
3933       Targets and Target Specific Improvements" for IA-32/x86-64 for more
3934       details.
3935     * The -gcolumn-info option is now enabled by default. It includes
3936       column information in addition to just filenames and line numbers
3937       in DWARF debugging information.
3938     * The polyhedral-based loop nest optimization pass
3939       -floop-nest-optimize has been overhauled. It's still considered
3940       experimental and may not result in any runtime improvements.
3941     * Two new classical loop nest optimization passes have been added.
3942       -floop-unroll-and-jam performs outer loop unrolling and fusing of
3943       the inner loop copies. -floop-interchange exchanges loops in a loop
3944       nest to improve data locality. Both passes are enabled by default
3945       at -O3 and above.
3946     * The classic loop nest optimization pass -ftree-loop-distribution
3947       has been improved and enabled by default at -O3 and above. It
3948       supports loop nest distribution in some restricted scenarios; it
3949       also supports cancellable innermost loop distribution with loop
3950       versioning under run-time alias checks.
3951     * The new option -fstack-clash-protection causes the compiler to
3952       insert probes whenever stack space is allocated statically or
3953       dynamically to reliably detect stack overflows and thus mitigate
3954       the attack vector that relies on jumping over a stack guard page as
3955       provided by the operating system.
3956     * A new pragma GCC unroll has been implemented in the C family of
3957       languages, as well as Fortran and Ada, so as to make it possible
3958       for the user to have a finer-grained control over the loop
3959       unrolling optimization.
3960     * GCC has been enhanced to detect more instances of meaningless or
3961       mutually exclusive attribute specifications and handle such
3962       conflicts more consistently. Mutually exclusive attribute
3963       specifications are ignored with a warning regardless of whether
3964       they appear on the same declaration or on distinct declarations of
3965       the same entity. For example, because the noreturn attribute on the
3966       second declaration below is mutually exclusive with the malloc
3967       attribute on the first, it is ignored and a warning is issued.
3968>
3969      void* __attribute__ ((malloc)) f (unsigned);
3970      void* __attribute__ ((noreturn)) f (unsigned);
3971
3972      warning: ignoring attribute 'noreturn' because it conflicts with attribute
3973 'malloc' [-Wattributes]
3974     * The gcov tool can distinguish functions that begin on a same line
3975       in a source file. This can be a different template instantiation or
3976       a class constructor:
3977
3978File 'ins.C'
3979Lines executed:100.00% of 8
3980Creating 'ins.C.gcov'
3981
3982        -:    0:Source:ins.C
3983        -:    0:Graph:ins.gcno
3984        -:    0:Data:ins.gcda
3985        -:    0:Runs:1
3986        -:    0:Programs:1
3987        -:    1:template<class T>
3988        -:    2:class Foo
3989        -:    3:{
3990        -:    4: public:
3991        2:    5:   Foo(): b (1000) {}
3992------------------
3993Foo<char>::Foo():
3994        1:    5:   Foo(): b (1000) {}
3995------------------
3996Foo<int>::Foo():
3997        1:    5:   Foo(): b (1000) {}
3998------------------
3999        2:    6:   void inc () { b++; }
4000------------------
4001Foo<char>::inc():
4002        1:    6:   void inc () { b++; }
4003------------------
4004Foo<int>::inc():
4005        1:    6:   void inc () { b++; }
4006------------------
4007        -:    7:
4008        -:    8:  private:
4009        -:    9:   int b;
4010        -:   10:};
4011        -:   11:
4012        1:   12:int main(int argc, char **argv)
4013        -:   13:{
4014        1:   14:  Foo<int> a;
4015        1:   15:  Foo<char> b;
4016        -:   16:
4017        1:   17:  a.inc ();
4018        1:   18:  b.inc ();
4019        1:   19:}
4020
4021     * The gcov tool has more accurate numbers for execution of lines in a
4022       source file.
4023     * The gcov tool can use TERM colors to provide more readable output.
4024     * AddressSanitizer gained a new pair of sanitization options,
4025       -fsanitize=pointer-compare and -fsanitize=pointer-subtract, which
4026       warn about subtraction (or comparison) of pointers that point to a
4027       different memory object:
4028
4029int
4030main ()
4031{
4032  /* Heap allocated memory.  */
4033  char *heap1 = (char *)__builtin_malloc (42);
4034  char *heap2 = (char *)__builtin_malloc (42);
4035  if (heap1 > heap2)
4036      return 1;
4037
4038  return 0;
4039}
4040
4041==17465==ERROR: AddressSanitizer: invalid-pointer-pair: 0x604000000010 0x6040000
404200050
4043    #0 0x40070f in main /tmp/pointer-compare.c:7
4044    #1 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86)
4045    #2 0x400629 in _start (/tmp/a.out+0x400629)
4046
40470x604000000010 is located 0 bytes inside of 42-byte region [0x604000000010,0x604
404800000003a)
4049allocated by thread T0 here:
4050    #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan
4051_malloc_linux.cc:86
4052    #1 0x4006ea in main /tmp/pointer-compare.c:5
4053    #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86)
4054
40550x604000000050 is located 0 bytes inside of 42-byte region [0x604000000050,0x604
405600000007a)
4057allocated by thread T0 here:
4058    #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan
4059_malloc_linux.cc:86
4060    #1 0x4006f8 in main /tmp/pointer-compare.c:6
4061    #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86)
4062
4063SUMMARY: AddressSanitizer: invalid-pointer-pair /tmp/pointer-compare.c:7 in main
4064
4065     * The store merging pass has been enhanced to handle bit-fields and
4066       not just constant stores, but also data copying from adjacent
4067       memory locations into other adjacent memory locations, including
4068       bitwise logical operations on the data. The pass can also handle
4069       byte swapping into memory locations.
4070     * The undefined behavior sanitizer gained two new options included in
4071       -fsanitize=undefined: -fsanitize=builtin which diagnoses at run
4072       time invalid arguments to __builtin_clz or __builtin_ctz prefixed
4073       builtins, and -fsanitize=pointer-overflow which performs cheap run
4074       time tests for pointer wrapping.
4075     * A new attribute no_sanitize can be applied to functions to instruct
4076       the compiler not to do sanitization of the options provided as
4077       arguments to the attribute. Acceptable values for no_sanitize match
4078       those acceptable by the -fsanitize command-line option.
4079
4080void __attribute__ ((no_sanitize ("alignment", "object-size")))
4081f () { /* Do something. */; }
4082
4083New Languages and Language specific improvements
4084
4085  Ada
4086
4087     * For its internal exception handling used on the host for error
4088       recovery in the front-end, the compiler now relies on the native
4089       exception handling mechanism of the host platform, which should be
4090       more efficient than the former mechanism.
4091
4092  BRIG (HSAIL)
4093
4094   In this release cycle, the focus for the BRIGFE was on stabilization
4095   and performance improvements. Also a couple of completely new features
4096   were added.
4097     * Improved support for function and module scope group segment
4098       variables. PRM specs define function and module scope group segment
4099       variables as an experimental feature. However, PRM test suite uses
4100       them. Now group segment is handled by separate book keeping of
4101       module scope and function (kernel) offsets. Each function has a
4102       "frame" in the group segment offset to which is given as an
4103       argument, similar to traditional call stack frame handling.
4104     * Reduce the number of type conversions due to the untyped HSAIL
4105       registers. Instead of always representing the HSAIL's untyped
4106       registers as unsigned int, the gccbrig now pre-analyzes the BRIG
4107       code and builds the register variables as a type used the most when
4108       storing or reading data to/from each register. This reduces the
4109       number of total casts which cannot be always optimized away.
4110     * Support for BRIG_KIND_NONE directives.
4111     * Made -O3 the default optimization level for BRIGFE.
4112     * Fixed illegal addresses generated from address expressions which
4113       refer only to offset 0.
4114     * Fixed a bug with reg+offset addressing on 32b segments. In 'large'
4115       mode, the offset is treated as 32-bit unless it's in global,
4116       read-only or kernarg address space.
4117     * Fixed a crash caused sometimes by calls with more than 4 arguments.
4118     * Fixed a mis-execution issue with kernels that have both unexpanded
4119       ID functions and calls to subfunctions.
4120     * Treat HSAIL barrier builtins as setjmp/longjump style functions to
4121       avoid illegal optimizations.
4122     * Ensure per WI copies of private variables are aligned correctly.
4123     * libhsail-rt: Assume the host runtime allocates the work group
4124       memory.
4125
4126  C family
4127
4128     * New command-line options have been added for the C and C++
4129       compilers:
4130          + [4]-Wmultistatement-macros warns about unsafe macros expanding
4131            to multiple statements used as a body of a statement such as
4132            if, else, while, switch, or for.
4133          + [5]-Wstringop-truncation warns for calls to bounded string
4134            manipulation functions such as strncat, strncpy, and stpncpy
4135            that might either truncate the copied string or leave the
4136            destination unchanged. For example, the following call to
4137            strncat is diagnosed because it appends just three of the four
4138            characters from the source string.
4139void append (char *buf, size_t bufsize)
4140{
4141    strncat (buf, ".txt", 3);
4142}
4143warning: 'strncat' output truncated copying 3 bytes from a string of length 4 [-
4144Wstringop-truncation]
4145            Similarly, in the following example, the call to strncpy
4146            specifies the size of the destination buffer as the bound. If
4147            the length of the source string is equal to or greater than
4148            this size the result of the copy will not be NUL-terminated.
4149            Therefore, the call is also diagnosed. To avoid the warning,
4150            specify sizeof buf - 1 as the bound and set the last element
4151            of the buffer to NUL.
4152void copy (const char *s)
4153{
4154    char buf[80];
4155    strncpy (buf, s, sizeof buf);
4156    ...
4157}
4158warning: 'strncpy' specified bound 80 equals destination size [-Wstringop-trunca
4159tion]
4160            The -Wstringop-truncation option is included in -Wall.
4161            Note that due to GCC bug [6]82944, defining strncat, strncpy,
4162            or stpncpy as a macro in a system header as some
4163            implementations do, suppresses the warning.
4164          + [7]-Wif-not-aligned controls warnings issued in response to
4165            invalid uses of objects declared with attribute
4166            [8]warn_if_not_aligned.
4167            The -Wif-not-aligned option is included in -Wall.
4168          + [9]-Wmissing-attributes warns when a declaration of a function
4169            is missing one or more attributes that a related function is
4170            declared with and whose absence may adversely affect the
4171            correctness or efficiency of generated code. For example, in
4172            C++, the warning is issued when an explicit specialization of
4173            a primary template declared with attribute alloc_align,
4174            alloc_size, assume_aligned, format, format_arg, malloc, or
4175            nonnull is declared without it. Attributes deprecated, error,
4176            and warning suppress the warning.
4177            The -Wmissing-attributes option is included in -Wall.
4178          + [10]-Wpacked-not-aligned warns when a struct or union declared
4179            with attribute packed defines a member with an explicitly
4180            specified alignment greater than 1. Such a member will wind up
4181            under-aligned. For example, a warning will be issued for the
4182            definition of struct A in the following:
4183struct __attribute__ ((aligned (8)))
4184S8 { char a[8]; };
4185
4186struct __attribute__ ((packed)) A
4187{
4188    struct S8 s8;
4189};
4190warning: alignment 1 of 'struct S' is less than 8 [-Wpacked-not-aligned]
4191            The -Wpacked-not-aligned option is included in -Wall.
4192          + -Wcast-function-type warns when a function pointer is cast to
4193            an incompatible function pointer. This warning is enabled by
4194            -Wextra.
4195          + -Wsizeof-pointer-div warns for suspicious divisions of the
4196            size of a pointer by the size of the elements it points to,
4197            which looks like the usual way to compute the array size but
4198            won't work out correctly with pointers. This warning is
4199            enabled by -Wall.
4200          + -Wcast-align=strict warns whenever a pointer is cast such that
4201            the required alignment of the target is increased. For
4202            example, warn if a char * is cast to an int * regardless of
4203            the target machine.
4204          + -fprofile-abs-path creates absolute path names in the .gcno
4205            files. This allows gcov to find the correct sources in
4206            projects where compilations occur with different working
4207            directories.
4208     * -fno-strict-overflow is now mapped to -fwrapv -fwrapv-pointer and
4209       signed integer overflow is now undefined by default at all
4210       optimization levels. Using -fsanitize=signed-integer-overflow is
4211       now the preferred way to audit code, -Wstrict-overflow is
4212       deprecated.
4213     * The [11]-Warray-bounds option has been improved to detect more
4214       instances of out-of-bounds array indices and pointer offsets. For
4215       example, negative or excessive indices into flexible array members
4216       and string literals are detected.
4217     * The [12]-Wrestrict option introduced in GCC 7 has been enhanced to
4218       detect many more instances of overlapping accesses to objects via
4219       restrict-qualified arguments to standard memory and string
4220       manipulation functions such as memcpy and strcpy. For example, the
4221       strcpy call in the function below attempts to truncate the string
4222       by replacing its initial characters with the last four. However,
4223       because the function writes the terminating NUL into a[4], the
4224       copies overlap and the call is diagnosed.
4225void f (void)
4226{
4227    char a[] = "abcd1234";
4228    strcpy (a, a + 4);
4229    ...
4230}
4231warning: 'strcpy' accessing 5 bytes at offsets 0 and 4 overlaps 1 byte at offset
4232 4 [-Wrestrict]
4233       The -Wrestrict option is included in -Wall.
4234     * Several optimizer enhancements have enabled improvements to the
4235       [13]-Wformat-overflow and [14]-Wformat-truncation options. The
4236       warnings detect more instances of buffer overflow and truncation
4237       than in GCC 7 and are better at avoiding certain kinds of false
4238       positives.
4239     * When reporting mismatching argument types at a function call, the C
4240       and C++ compilers now underline both the argument and the pertinent
4241       parameter in the declaration.
4242$ gcc arg-type-mismatch.cc
4243arg-type-mismatch.cc: In function 'int caller(int, int, float)':
4244arg-type-mismatch.cc:5:24: error: invalid conversion from 'int' to 'const char*'
4245 [-fpermissive]
4246   return callee(first, second, third);
4247                        ^~~~~~
4248arg-type-mismatch.cc:1:40: note:   initializing argument 2 of 'int callee(int, c
4249onst char*, float)'
4250 extern int callee(int one, const char *two, float three);
4251                            ~~~~~~~~~~~~^~~
4252
4253     * When reporting on unrecognized identifiers, the C and C++ compilers
4254       will now emit fix-it hints suggesting #include directives for
4255       various headers in the C and C++ standard libraries.
4256$ gcc incomplete.c
4257incomplete.c: In function 'test':
4258incomplete.c:3:10: error: 'NULL' undeclared (first use in this function)
4259   return NULL;
4260          ^~~~
4261incomplete.c:3:10: note: 'NULL' is defined in header '<stddef.h>'; did you forge
4262t to '#include <stddef.h>'?
4263incomplete.c:1:1:
4264+#include <stddef.h>
4265 const char *test(void)
4266incomplete.c:3:10:
4267   return NULL;
4268          ^~~~
4269incomplete.c:3:10: note: each undeclared identifier is reported only once for ea
4270ch function it appears in
4271
4272$ gcc incomplete.cc
4273incomplete.cc:1:6: error: 'string' in namespace 'std' does not name a type
4274 std::string s("hello world");
4275      ^~~~~~
4276incomplete.cc:1:1: note: 'std::string' is defined in header '<string>'; did you
4277forget to '#include <string>'?
4278+#include <string>
4279 std::string s("hello world");
4280 ^~~
4281
4282     * The C and C++ compilers now use more intuitive locations when
4283       reporting on missing semicolons, and offer fix-it hints:
4284$ gcc t.c
4285t.c: In function 'test':
4286t.c:3:12: error: expected ';' before '}' token
4287   return 42
4288            ^
4289            ;
4290 }
4291 ~
4292
4293     * When reporting on missing '}' and ')' tokens, the C and C++
4294       compilers will now highlight the corresponding '{' and '(' token,
4295       issuing a 'note' if it's on a separate line:
4296$ gcc unclosed.c
4297unclosed.c: In function 'log_when_out_of_range':
4298unclosed.c:12:50: error: expected ')' before '{' token
4299       && (temperature < MIN || temperature > MAX) {
4300                                                  ^~
4301                                                  )
4302unclosed.c:11:6: note: to match this '('
4303   if (logging_enabled && check_range ()
4304      ^
4305
4306       or highlighting it directly if it's on the same line:
4307$ gcc unclosed-2.c
4308unclosed-2.c: In function 'test':
4309unclosed-2.c:8:45: error: expected ')' before '{' token
4310   if (temperature < MIN || temperature > MAX {
4311      ~                                      ^~
4312                                             )
4313
4314       They will also emit fix-it hints.
4315
4316  C++
4317
4318     * GCC 8 (-fabi-version=12) has a couple of corrections to the calling
4319       convention, which changes the ABI for some uncommon code:
4320          + Passing an empty class as an argument now takes up no space on
4321            x86_64, as required by the psABI.
4322          + Passing or returning a class with only deleted copy and move
4323            constructors now uses the same calling convention as a class
4324            with a non-trivial copy or move constructor. This only affects
4325            C++17 mode, as in earlier standards passing or returning such
4326            a class was impossible.
4327          + WARNING: In GCC 8.1 the second change mistakenly also affects
4328            classes with a deleted copy constructor and defaulted trivial
4329            move constructor (bug [15]c++/86094). This issue is fixed in
4330            GCC 8.2 (-fabi-version=13).
4331       You can test whether these changes affect your code with -Wabi=11
4332       (or -Wabi=12 in GCC 8.2 for the third issue); if these changes are
4333       problematic for your project, the GCC 7 ABI can be selected with
4334       -fabi-version=11.
4335     * The value of the C++11 alignof operator has been corrected to match
4336       C _Alignof (minimum alignment) rather than GNU __alignof__
4337       (preferred alignment); on ia32 targets this means that
4338       alignof(double) is now 4 rather than 8. Code that wants the
4339       preferred alignment should use __alignof__ instead.
4340     * New command-line options have been added for the C++ compiler to
4341       control warnings:
4342          + [16]-Wclass-memaccess warns when objects of non-trivial class
4343            types are manipulated in potentially unsafe ways by raw memory
4344            functions such as memcpy, or realloc. The warning helps detect
4345            calls that bypass user-defined constructors or copy-assignment
4346            operators, corrupt virtual table pointers, data members of
4347            const-qualified types or references, or member pointers. The
4348            warning also detects calls that would bypass access controls
4349            to data members. For example, a call such as:
4350        memcpy (&std::cout, &std::cerr, sizeof std::cout);
4351            results in
4352        warning: 'void* memcpy(void*, const void*, long unsigned int)' writing t
4353o an object of type 'std::ostream' {aka 'class std::basic_ostream<char>'} with n
4354o trivial copy-assignment [-Wclass-memaccess]
4355            The -Wclass-memaccess option is included in -Wall.
4356     * The C++ front end has experimental support for some of the upcoming
4357       C++2a draft features with the -std=c++2a or -std=gnu++2a flags,
4358       including designated initializers, default member initializers for
4359       bit-fields, __VA_OPT__ (except that #__VA_OPT__ is unsupported),
4360       lambda [=, this] captures, etc. For a full list of new features,
4361       see [17]the C++ status page.
4362     * When reporting on attempts to access private fields of a class or
4363       struct, the C++ compiler will now offer fix-it hints showing how to
4364       use an accessor function to get at the field in question, if one
4365       exists.
4366$ gcc accessor.cc
4367accessor.cc: In function 'void test(foo*)':
4368accessor.cc:12:12: error: 'double foo::m_ratio' is private within this context
4369   if (ptr->m_ratio >= 0.5)
4370            ^~~~~~~
4371accessor.cc:7:10: note: declared private here
4372   double m_ratio;
4373          ^~~~~~~
4374accessor.cc:12:12: note: field 'double foo::m_ratio' can be accessed via 'double
4375 foo::get_ratio() const'
4376   if (ptr->m_ratio >= 0.5)
4377            ^~~~~~~
4378            get_ratio()
4379
4380     * The C++ compiler can now give you a hint if you use a macro before
4381       it was defined (e.g. if you mess up the order of your #include
4382       directives):
4383$ gcc ordering.cc
4384ordering.cc:2:24: error: expected ';' at end of member declaration
4385   virtual void clone() const OVERRIDE { }
4386                        ^~~~~
4387                             ;
4388ordering.cc:2:30: error: 'OVERRIDE' does not name a type
4389   virtual void clone() const OVERRIDE { }
4390                              ^~~~~~~~
4391ordering.cc:2:30: note: the macro 'OVERRIDE' had not yet been defined
4392In file included from ordering.cc:5:
4393c++11-compat.h:2: note: it was later defined here
4394 #define OVERRIDE override
4395
4396
4397     * The -Wold-style-cast diagnostic can now emit fix-it hints telling
4398       you when you can use a static_cast, const_cast, or
4399       reinterpret_cast.
4400$ gcc -c old-style-cast-fixits.cc -Wold-style-cast
4401old-style-cast-fixits.cc: In function 'void test(void*)':
4402old-style-cast-fixits.cc:5:19: warning: use of old-style cast to 'struct foo*' [
4403-Wold-style-cast]
4404   foo *f = (foo *)ptr;
4405                   ^~~
4406            ----------
4407            static_cast<foo *> (ptr)
4408
4409     * When reporting on problems within extern "C" linkage
4410       specifications, the C++ compiler will now display the location of
4411       the start of the extern "C".
4412$ gcc -c extern-c.cc
4413extern-c.cc:3:1: error: template with C linkage
4414 template <typename T> void test (void);
4415 ^~~~~~~~
4416In file included from extern-c.cc:1:
4417unclosed.h:1:1: note: 'extern "C"' linkage started here
4418 extern "C" {
4419 ^~~~~~~~~~
4420extern-c.cc:3:39: error: expected '}' at end of input
4421 template <typename T> void test (void);
4422                                       ^
4423In file included from extern-c.cc:1:
4424unclosed.h:1:12: note: to match this '{'
4425 extern "C" {
4426            ^
4427
4428     * When reporting on mismatching template types, the C++ compiler will
4429       now use color to highlight the mismatching parts of the template,
4430       and will elide the parameters that are common between two
4431       mismatching templates, printing [...] instead:
4432$ gcc templates.cc
4433templates.cc: In function 'void test()':
4434templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl
4435e>' to 'vector<int>'
4436   fn_1(vector<double> ());
4437        ^~~~~~~~~~~~~~~~~
4438templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<[...]
4439,double>' to 'map<[...],int>'
4440   fn_2(map<int, double>());
4441        ^~~~~~~~~~~~~~~~~~
4442
4443       Those [...] elided parameters can be seen using -fno-elide-type:
4444$ gcc templates.cc -fno-elide-type
4445templates.cc: In function 'void test()':
4446templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl
4447e>' to 'vector<int>'
4448   fn_1(vector<double> ());
4449        ^~~~~~~~~~~~~~~~~
4450templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<int,d
4451ouble>' to 'map<int,int>'
4452   fn_2(map<int, double>());
4453        ^~~~~~~~~~~~~~~~~~
4454
4455       The C++ compiler has also gained an option
4456       -fdiagnostics-show-template-tree which visualizes such mismatching
4457       templates in a hierarchical form:
4458$ gcc templates-2.cc -fdiagnostics-show-template-tree
4459templates-2.cc: In function 'void test()':
4460templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou
4461ble>' to 'vector<int>'
4462  vector<
4463    [double != int]>
4464   fn_1(vector<double> ());
4465        ^~~~~~~~~~~~~~~~~
4466templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve
4467ctor<double> >()' from 'map<map<[...],vector<double>>,vector<double>>' to 'map<m
4468ap<[...],vector<float>>,vector<float>>'
4469  map<
4470    map<
4471      [...],
4472      vector<
4473        [double != float]>>,
4474    vector<
4475      [double != float]>>
4476   fn_2(map<map<int, vector<double>>, vector<double>> ());
4477        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4478
4479       which again works with -fno-elide-type:
4480$ gcc templates-2.cc -fdiagnostics-show-template-tree -fno-elide-type
4481templates-2.cc: In function 'void test()':
4482templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou
4483ble>' to 'vector<int>'
4484  vector<
4485    [double != int]>
4486   fn_1(vector<double> ());
4487        ^~~~~~~~~~~~~~~~~
4488templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve
4489ctor<double> >()' from 'map<map<int,vector<double>>,vector<double>>' to 'map<map
4490<int,vector<float>>,vector<float>>'
4491  map<
4492    map<
4493      int,
4494      vector<
4495        [double != float]>>,
4496    vector<
4497      [double != float]>>
4498   fn_2(map<map<int, vector<double>>, vector<double>> ());
4499        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4500
4501     * Flowing off the end of a non-void function is considered
4502       unreachable and may be subject to optimization on that basis. As a
4503       result of this change, -Wreturn-type warnings are enabled by
4504       default for C++.
4505
4506    Runtime Library (libstdc++)
4507
4508     * Improved experimental support for C++17, including the following
4509       features:
4510          + Deduction guides to support class template argument deduction.
4511          + std::filesystem implementation.
4512          + std::char_traits<char> and std::char_traits<wchar_t> are
4513            usable in constant expressions.
4514          + std::to_chars and std::from_chars (for integers only, not for
4515            floating point types).
4516     * Experimental support for C++2a: std::to_address (thanks to Glen
4517       Fernandes) and std::endian.
4518     * On GNU/Linux, std::random_device::entropy() accesses the kernel's
4519       entropy count for the random device, if known (thanks to Xi
4520       Ruoyao).
4521     * Support for std::experimental::source_location.
4522     * AddressSanitizer integration for std::vector, detecting
4523       out-of-range accesses to the unused capacity of a vector.
4524     * Extensions __gnu_cxx::airy_ai and __gnu_cxx::airy_bi added to the
4525       Mathematical Special Functions.
4526
4527  Fortran
4528
4529     * The main version of libfortran has been changed to 5.
4530     * Parameterized derived types, a major feature of Fortran 2003, have
4531       been implemented.
4532     * Partial support is provided for Fortran 2018 teams, which are
4533       hierarchical subsets of images that execute independently of other
4534       image subsets.
4535     * The maximum rank for arrays has been increased to 15, conforming to
4536       the Fortran 2008 standard.
4537     * Transformational intrinsics are now fully supported in
4538       initialization expressions.
4539     * New flag -fc-prototypes to write C prototypes for BIND(C)
4540       procedures and variables.
4541     * If -fmax-stack-var-size is honored if given together with -Ofast,
4542       -fstack-arrays is no longer set in that case.
4543     * New options -fdefault-real-16 and -fdefault-real-10 to control the
4544       default kind of REAL variables.
4545     * A warning is now issued if an array subscript inside a DO loop
4546       could lead to an out-of-bounds-access. The new option
4547       -Wdo-subscript, enabled by -Wextra, warns about this even if the
4548       compiler can not prove that the code will be executed.
4549     * The Fortran front end now attempts to interchange loops if it is
4550       deemed profitable. So far, this is restricted to FORALL and DO
4551       CONCURRENT statements with multiple indices. This behavior be
4552       controlled with the new flag -ffrontend-loop-interchange, which is
4553       enabled with optimization by default. The
4554       -Wfrontend-loop-interchange option warns about such occurrences.
4555     * When an actual argument contains too few elements for a dummy
4556       argument, an error is now issued. The -std=legacy option can be
4557       used to still compile such code.
4558     * The RECL= argument to OPEN and INQUIRE statements now allows 64-bit
4559       integers, making records larger than 2GiB possible.
4560     * The GFORTRAN_DEFAULT_RECL environment variable no longer has any
4561       effect. The record length for preconnected units is now larger than
4562       any practical limit, same as for sequential access units opened
4563       without an explicit RECL= specifier.
4564     * Character variables longer than HUGE(0) elements are now possible
4565       on 64-bit targets. Note that this changes the procedure call ABI
4566       for all procedures with character arguments on 64-bit targets, as
4567       the type of the hidden character length argument has changed. The
4568       hidden character length argument is now of type INTEGER(C_SIZE_T).
4569     * Partial support is provided for Fortran 2018 teams, which are
4570       hierarchical subsets of images that execute independently of other
4571       image subsets.
4572
4573  Go
4574
4575     * GCC 8 provides a complete implementation of the Go 1.10.1 user
4576       packages.
4577     * The garbage collector is now fully concurrent. As before, values
4578       stored on the stack are scanned conservatively, but value stored in
4579       the heap are scanned precisely.
4580     * Escape analysis is fully implemented and enabled by default in the
4581       Go front end. This significantly reduces the number of heap
4582       allocations by allocating values on the stack instead.
4583
4584libgccjit
4585
4586   The libgccjit API gained four new entry points:
4587     * [18]gcc_jit_type_get_vector and
4588     * [19]gcc_jit_context_new_rvalue_from_vector for working with
4589       vectors,
4590     * [20]gcc_jit_type_get_aligned
4591     * [21]gcc_jit_function_get_address
4592
4593   The C code generated by [22]gcc_jit_context_dump_reproducer_to_file is
4594   now easier-to-read.
4595
4596New Targets and Target Specific Improvements
4597
4598  AArch64
4599
4600     * The Armv8.4-A architecture is now supported. It can be used by
4601       specifying the -march=armv8.4-a option.
4602     * The Dot Product instructions are now supported as an optional
4603       extension to the Armv8.2-A architecture and newer and are mandatory
4604       on Armv8.4-A. The extension can be used by specifying the +dotprod
4605       architecture extension. E.g. -march=armv8.2-a+dotprod.
4606     * The Armv8-A +crypto extension has now been split into two
4607       extensions for finer grained control:
4608          + +aes which contains the Armv8-A AES crytographic instructions.
4609          + +sha2 which contains the Armv8-A SHA2 and SHA1 cryptographic
4610            instructions.
4611       Using +crypto will now enable these two extensions.
4612     * New Armv8.4-A FP16 Floating Point Multiplication Variant
4613       instructions have been added. These instructions are mandatory in
4614       Armv8.4-A but available as an optional extension to Armv8.2-A and
4615       Armv8.3-A. The new extension can be used by specifying the +fp16fml
4616       architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A
4617       the instructions can be enabled by specifying +fp16.
4618     * New cryptographic instructions have been added as optional
4619       extensions to Armv8.2-A and newer. These instructions can be
4620       enabled with:
4621          + +sha3 New SHA3 and SHA2 instructions from Armv8.4-A. This
4622            implies +sha2.
4623          + +sm4 New SM3 and SM4 instructions from Armv8.4-A.
4624     * The Scalable Vector Extension (SVE) is now supported as an optional
4625       extension to the Armv8.2-A architecture and newer. This support
4626       includes automatic vectorization with SVE instructions, but it does
4627       not yet include the SVE Arm C Language Extensions (ACLE). It can be
4628       enabled by specifying the +sve architecture extension (for example,
4629       -march=armv8.2-a+sve). By default, the generated code works with
4630       all vector lengths, but it can be made specific to N-bit vectors
4631       using -msve-vector-bits=N.
4632     * Support has been added for the following processors (GCC
4633       identifiers in parentheses):
4634          + Arm Cortex-A75 (cortex-a75).
4635          + Arm Cortex-A55 (cortex-a55).
4636          + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE
4637            (cortex-a75.cortex-a55).
4638       The GCC identifiers can be used as arguments to the -mcpu or -mtune
4639       options, for example: -mcpu=cortex-a75 or -mtune=cortex-a75 or as
4640       arguments to the equivalent target attributes and pragmas.
4641
4642  ARC
4643
4644     * Added support for:
4645          + Fast interrupts.
4646          + Naked functions.
4647          + aux variable attributes.
4648          + uncached type qualifier.
4649          + Secure functions via sjli instruction.
4650     * New exception handling implementation.
4651     * Revamped trampoline implementation.
4652     * Refactored small data feature implementation, controlled via the -G
4653       command-line option.
4654     * New support for reduced register set ARC architecture
4655       configurations, controlled via the -mrf16 command-line option.
4656     * Refurbished and improved support for zero overhead loops.
4657       Introduced -mlpc-width command-line option to control the width of
4658       the lp_count register.
4659
4660  ARM
4661
4662     * The -mfpu option now takes a new option setting of -mfpu=auto. When
4663       set to this the floating-point and SIMD settings are derived from
4664       the settings of the -mcpu or -march options. The internal CPU
4665       configurations have been updated with information about the
4666       permitted floating-point configurations supported. See the user
4667       guide for further information about the extended option syntax for
4668       controlling architectural extensions via the -march option.
4669       -mfpu=auto is now the default setting unless the compiler has been
4670       configured with an explicit --with-fpu option.
4671     * The -march and -mcpu options now accept optional extensions to the
4672       architecture or CPU option, allowing the user to enable or disable
4673       any such extensions supported by that architecture or CPU such as
4674       (but not limited to) floating-point and AdvancedSIMD. For example:
4675       the option -mcpu=cortex-a53+nofp will generate code for the
4676       Cortex-A53 processor with no floating-point support. This, in
4677       combination with the new -mfpu=auto option, provides a
4678       straightforward way of specifying a valid build target through a
4679       single -mcpu or -march option. The -mtune option accepts the same
4680       arguments as -mcpu but only the CPU name has an effect on tuning.
4681       The architecture extensions do not have any effect. For details of
4682       what extensions a particular architecture or CPU option supports
4683       please refer to the [23]documentation.
4684     * The -mstructure-size-boundary option has been deprecated and will
4685       be removed in a future release.
4686     * The default link behavior for Armv6 and Armv7-R targets has been
4687       changed to produce BE8 format when generating big-endian images. A
4688       new flag -mbe32 can be used to force the linker to produce legacy
4689       BE32 format images. There is no change of behavior for Armv6-M and
4690       other Armv7 or later targets: these already defaulted to BE8
4691       format. This change brings GCC into alignment with other compilers
4692       for the ARM architecture.
4693     * The Armv8-R architecture is now supported. It can be used by
4694       specifying the -march=armv8-r option.
4695     * The Armv8.3-A architecture is now supported. It can be used by
4696       specifying the -march=armv8.3-a option.
4697     * The Armv8.4-A architecture is now supported. It can be used by
4698       specifying the -march=armv8.4-a option.
4699     * The Dot Product instructions are now supported as an optional
4700       extension to the Armv8.2-A architecture and newer and are mandatory
4701       on Armv8.4-A. The extension can be used by specifying the +dotprod
4702       architecture extension. E.g. -march=armv8.2-a+dotprod.
4703     * Support for setting extensions and architectures using the GCC
4704       target pragma and attribute has been added. It can be used by
4705       specifying #pragma GCC target ("arch=..."), #pragma GCC target
4706       ("+extension"), __attribute__((target("arch=..."))) or
4707       __attribute__((target("+extension"))).
4708     * New Armv8.4-A FP16 Floating Point Multiplication Variant
4709       instructions have been added. These instructions are mandatory in
4710       Armv8.4-A but available as an optional extension to Armv8.2-A and
4711       Armv8.3-A. The new extension can be used by specifying the +fp16fml
4712       architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A
4713       the instructions can be enabled by specifying +fp16.
4714     * Support has been added for the following processors (GCC
4715       identifiers in parentheses):
4716          + Arm Cortex-A75 (cortex-a75).
4717          + Arm Cortex-A55 (cortex-a55).
4718          + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE
4719            (cortex-a75.cortex-a55).
4720          + Arm Cortex-R52 for Armv8-R (cortex-r52).
4721       The GCC identifiers can be used as arguments to the -mcpu or -mtune
4722       options, for example: -mcpu=cortex-a75 or -mtune=cortex-r52 or as
4723       arguments to the equivalent target attributes and pragmas.
4724
4725  AVR
4726
4727     * The AVR port now supports the following XMEGA-like devices:
4728
4729     ATtiny212, ATtiny214, ATtiny412, ATtiny414, ATtiny416, ATtiny417,
4730     ATtiny814, ATtiny816, ATtiny817, ATtiny1614, ATtiny1616, ATtiny1617,
4731     ATtiny3214, ATtiny3216, ATtiny3217
4732       The new devices are listed under [24]-mmcu=avrxmega3.
4733          + These devices see flash memory in the RAM address space, so
4734            that features like PROGMEM and __flash are not needed any more
4735            (as opposed to other AVR families for which read-only data
4736            will be located in RAM except special, non-standard features
4737            are used to locate and access such data). This requires that
4738            the compiler is used with Binutils 2.29 or newer so that
4739            [25]read-only data will be located in flash memory.
4740          + A new command-line option -mshort-calls is supported. This
4741            option is used internally for multilib selection of the
4742            avrxmega3 variants. It is not an optimization option. Do not
4743            set it by hand.
4744     * The compiler now generates [26]efficient interrupt service routine
4745       (ISR) prologues and epilogues. This is achieved by using the new
4746       [27]AVR pseudo instruction __gcc_isr which is supported and
4747       resolved by the GNU assembler.
4748          + As the __gcc_isr pseudo-instruction will be resolved by the
4749            assembler, inline assembly is transparent to the process. This
4750            means that when inline assembly uses an instruction like INC
4751            that clobbers the condition code, then the assembler will
4752            detect this and generate an appropriate ISR prologue /
4753            epilogue chunk to save / restore SREG as needed.
4754          + A new command-line option -mno-gas-isr-prologues disables the
4755            generation of the __gcc_isr pseudo instruction. Any non-naked
4756            ISR will save and restore SREG, tmp_reg and zero_reg, no
4757            matter whether the respective register is clobbered or used.
4758          + The feature is turned on per default for all optimization
4759            levels except for -O0 and -Og. It is explicitly enabled by
4760            means of option -mgas-isr-prologues.
4761          + Support has been added for a new [28]AVR function attribute
4762            no_gccisr. It can be used to disable __gcc_isr pseudo
4763            instruction generation for individual ISRs.
4764          + This optimization is only available if GCC is configured with
4765            GNU Binutils 2.29 or newer; or at least with a version of
4766            Binutils that implements feature [29]PR21683.
4767     * The compiler no more saves / restores registers in main; the effect
4768       is the same as if attribute OS_task was specified for main. This
4769       optimization can be switched off by the new command-line option
4770       -mno-main-is-OS_task.
4771
4772  IA-32/x86-64
4773
4774     * The x86 port now supports the naked function attribute.
4775     * Better tuning for znver1 and Intel Core based CPUs.
4776     * Vectorization cost metrics has been reworked leading to significant
4777       improvements on some benchmarks.
4778     * GCC now supports the Intel CPU named Cannonlake through
4779       -march=cannonlake. The switch enables the AVX512VBMI, AVX512IFMA
4780       and SHA ISA extensions.
4781     * GCC now supports the Intel CPU named Icelake through
4782       -march=icelake. The switch enables the AVX512VNNI, GFNI, VAES,
4783       AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID and AVX512VPOPCNTDQ
4784       ISA extensions.
4785     * GCC now supports the Intel Control-flow Enforcement Technology
4786       (CET) extension through -fcf-protection option.
4787
4788  NDS32
4789
4790     * New command-line options -mext-perf, -mext-perf2, and -mext-string
4791       have been added for performance extension instructions.
4792
4793  Nios II
4794
4795     * The Nios II back end has been improved to generate better-optimized
4796       code. Changes include switching to LRA, more accurate cost models,
4797       and more compact code for addressing static variables.
4798     * New command-line options -mgprel-sec= and -mr0rel-sec= have been
4799       added.
4800     * The stack-smashing protection options are now enabled on Nios II.
4801
4802  PA-RISC
4803
4804     * The default call ABI on 32-bit linux has been changed from callee
4805       copies to caller copies. This affects objects larger than eight
4806       bytes passed by value. The goal is to improve compatibility with
4807       x86 and resolve issues with OpenMP.
4808     * Other PA-RISC targets are unchanged.
4809
4810  PowerPC / PowerPC64 / RS6000
4811
4812     * The PowerPC SPE support is split off to a separate powerpcspe port.
4813       The separate port is deprecated and might be removed in a future
4814       release.
4815     * The Paired Single support (as used on some PPC750 CPUs, -mpaired,
4816       powerpc*-*-linux*paired*) is deprecated and will be removed in a
4817       future release.
4818     * The Xilinx floating point support (-mxilinx-fpu,
4819       powerpc-xilinx-eabi*) is deprecated and will be removed in a future
4820       release.
4821     * Support for using big-endian AltiVec intrinsics on a little-endian
4822       target (-maltivec=be) is deprecated and will be removed in a future
4823       release.
4824
4825  Tile
4826
4827     * The TILE-Gx port is deprecated and will be removed in a future
4828       release.
4829
4830Operating Systems
4831
4832  Windows
4833
4834     * GCC on Microsoft Windows can now be configured via
4835       --enable-mingw-wildcard or --disable-mingw-wildcard to force a
4836       specific behavior for GCC itself with regards to supporting the
4837       wildcard character. Prior versions of GCC would follow the
4838       configuration of the MinGW runtime. This behavior can still be
4839       obtained by not using the above options or by using
4840       --enable-mingw-wildcard=platform.
4841
4842Improvements for plugin authors
4843
4844     * Plugins can now register a callback hook for when comments are
4845       encountered by the C and C++ compilers, e.g. allowing for plugins
4846       to handle documentation markup in code comments.
4847     * The gdbinit support script for debugging GCC now has a
4848       break-on-diagnostic command, providing an easy way to trigger a
4849       breakpoint whenever a diagnostic is emitted.
4850     * The API for creating fix-it hints now supports newlines, and for
4851       emitting mutually incompatible fix-it hints for one diagnostic.
4852
4853GCC 8.1
4854
4855   This is the [30]list of problem reports (PRs) from GCC's bug tracking
4856   system that are known to be fixed in the 8.1 release. This list might
4857   not be complete (that is, it is possible that some PRs that have been
4858   fixed are not listed here).
4859
4860GCC 8.2
4861
4862   This is the [31]list of problem reports (PRs) from GCC's bug tracking
4863   system that are known to be fixed in the 8.2 release. This list might
4864   not be complete (that is, it is possible that some PRs that have been
4865   fixed are not listed here).
4866
4867  General Improvements
4868
4869     * Fixed LTO link-time performance problems caused by an overflow in
4870       the partitioning algorithm while building large binaries.
4871
4872  Language Specific Changes
4873
4874    C++
4875
4876   GCC 8.2 fixed a bug introduced in GCC 8.1 affecting passing or
4877   returning of classes with a deleted copy constructor and defaulted
4878   trivial move constructor (bug [32]c++/86094). GCC 8.2 introduces
4879   -fabi-version=13 and makes it the default, ABI incompatibilities
4880   between GCC 8.1 and 8.2 can be reported with -Wabi=12. See [33]C++
4881   changes for more details.
4882
4883  Target Specific Changes
4884
4885    IA-32/x86-64
4886
4887     * -mtune=native performance regression [34]PR84413 on Intel Skylake
4888       processors has been fixed.
4889
4890GCC 8.3
4891
4892   This is the [35]list of problem reports (PRs) from GCC's bug tracking
4893   system that are known to be fixed in the 8.3 release. This list might
4894   not be complete (that is, it is possible that some PRs that have been
4895   fixed are not listed here).
4896
4897  Windows
4898
4899     * A C++ Microsoft ABI bitfield layout bug, [36]PR87137 has been
4900       fixed. A non-field declaration could cause the current bitfield
4901       allocation unit to be completed, incorrectly placing a following
4902       bitfield into a new allocation unit. The Microsoft ABI is selected
4903       for:
4904          + Mingw targets
4905          + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields
4906            option is specified, or __attribute__((ms_struct)) is used
4907          + SuperH targets when the -mhitachi option is specified, or
4908            __attribute__((renesas)) is used
4909       GCC 8 introduced additional cases of this defect, but rather than
4910       resolve only those regressions, we decided to resolve all the cases
4911       of this defect in single change.
4912
4913GCC 8.4
4914
4915   This is the [37]list of problem reports (PRs) from GCC's bug tracking
4916   system that are known to be fixed in the 8.4 release. This list might
4917   not be complete (that is, it is possible that some PRs that have been
4918   fixed are not listed here).
4919
4920GCC 8.5
4921
4922   This is the [38]list of problem reports (PRs) from GCC's bug tracking
4923   system that are known to be fixed in the 8.5 release. This list might
4924   not be complete (that is, it is possible that some PRs that have been
4925   fixed are not listed here).
4926
4927  Target Specific Changes
4928
4929    AArch64
4930
4931     * The option -moutline-atomics has been added to aid deployment of
4932       the Large System Extensions (LSE) on GNU/Linux systems built with a
4933       baseline architecture targeting Armv8-A. When the option is
4934       specified code is emitted to detect the presence of LSE
4935       instructions at run time and use them for standard atomic
4936       operations. For more information please refer to the documentation.
4937
4938
4939    For questions related to the use of GCC, please consult these web
4940    pages and the [39]GCC manuals. If that fails, the
4941    [40]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4942    web pages and the development of GCC are welcome on our developer
4943    list at [41]gcc@gcc.gnu.org. All of [42]our lists have public
4944    archives.
4945
4946   Copyright (C) [43]Free Software Foundation, Inc. Verbatim copying and
4947   distribution of this entire article is permitted in any medium,
4948   provided this notice is preserved.
4949
4950   These pages are [44]maintained by the GCC team. Last modified
4951   2023-01-11.
4952
4953References
4954
4955   1. http://gcc.gnu.org/gcc-8/porting_to.html
4956   2. http://gcc.gnu.org/onlinedocs/index.html#current
4957   3. https://gcc.gnu.org/ml/gcc/2018-04/msg00102.html
4958   4. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmultistatement-macros
4959   5. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wstringop-truncation
4960   6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82944
4961   7. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wif-not-aligned
4962   8. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Common-Variable-Attributes.html#index-warn_005fif_005fnot_005faligned-variable-attribute
4963   9. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmissing-attributes
4964  10. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wpacked-not-aligned
4965  11. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Warray-bounds
4966  12. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wrestrict
4967  13. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-overflow
4968  14. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-truncation
4969  15. https://gcc.gnu.org/PR86094
4970  16. https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess
4971  17. http://gcc.gnu.org/projects/cxx-status.html#cxx2a
4972  18. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_vector
4973  19. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/expressions.html#gcc_jit_context_new_rvalue_from_vector
4974  20. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_aligned
4975  21. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/function-pointers.html#gcc_jit_function_get_address
4976  22. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/contexts.html#gcc_jit_context_dump_reproducer_to_file
4977  23. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/ARM-Options.html#ARM-Options
4978  24. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Options.html
4979  25. https://sourceware.org/PR21472
4980  26. https://gcc.gnu.org/PR20296
4981  27. https://sourceware.org/binutils/docs-2.29/as/AVR-Pseudo-Instructions.html
4982  28. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Function-Attributes.html
4983  29. https://sourceware.org/PR21683
4984  30. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.0
4985  31. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.2
4986  32. https://gcc.gnu.org/PR86094
4987  33. http://gcc.gnu.org/gcc-8/changes.html#cxx
4988  34. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84413
4989  35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.3
4990  36. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137
4991  37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.4
4992  38. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.5
4993  39. https://gcc.gnu.org/onlinedocs/
4994  40. mailto:gcc-help@gcc.gnu.org
4995  41. mailto:gcc@gcc.gnu.org
4996  42. https://gcc.gnu.org/lists.html
4997  43. https://www.fsf.org/
4998  44. https://gcc.gnu.org/about.html
4999======================================================================
5000http://gcc.gnu.org/gcc-7/index.html
5001
5002                              GCC 7 Release Series
5003
5004   (This release series is no longer supported.)
5005
5006   Nov 14, 2019
5007
5008   The [1]GNU project and the GCC developers are pleased to announce the
5009   release of GCC 7.5.
5010
5011   This release is a bug-fix release, containing fixes for regressions in
5012   GCC 7.4 relative to previous releases of GCC.
5013
5014Release History
5015
5016   GCC 7.5
5017          Nov 14, 2019 ([2]changes, [3]documentation)
5018
5019   GCC 7.4
5020          Dec 6, 2018 ([4]changes, [5]documentation)
5021
5022   GCC 7.3
5023          Jan 25, 2018 ([6]changes, [7]documentation)
5024
5025   GCC 7.2
5026          Aug 14, 2017 ([8]changes, [9]documentation)
5027
5028   GCC 7.1
5029          May 2, 2017 ([10]changes, [11]documentation)
5030
5031References and Acknowledgements
5032
5033   GCC used to stand for the GNU C Compiler, but since the compiler
5034   supports several other languages aside from C, it now stands for the
5035   GNU Compiler Collection.
5036
5037   A list of [12]successful builds is updated as new information becomes
5038   available.
5039
5040   The GCC developers would like to thank the numerous people that have
5041   contributed new features, improvements, bug fixes, and other changes as
5042   well as test results to GCC. This [13]amazing group of volunteers is
5043   what makes GCC successful.
5044
5045   For additional information about GCC please refer to the [14]GCC
5046   project web site or contact the [15]GCC development mailing list.
5047
5048   To obtain GCC please use [16]our mirror sites or [17]our version
5049   control system.
5050
5051
5052    For questions related to the use of GCC, please consult these web
5053    pages and the [18]GCC manuals. If that fails, the
5054    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5055    web pages and the development of GCC are welcome on our developer
5056    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
5057    archives.
5058
5059   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
5060   distribution of this entire article is permitted in any medium,
5061   provided this notice is preserved.
5062
5063   These pages are [23]maintained by the GCC team. Last modified
5064   2022-10-26.
5065
5066References
5067
5068   1. http://www.gnu.org/
5069   2. http://gcc.gnu.org/gcc-7/changes.html
5070   3. http://gcc.gnu.org/onlinedocs/7.5.0/
5071   4. http://gcc.gnu.org/gcc-7/changes.html
5072   5. http://gcc.gnu.org/onlinedocs/7.4.0/
5073   6. http://gcc.gnu.org/gcc-7/changes.html
5074   7. http://gcc.gnu.org/onlinedocs/7.3.0/
5075   8. http://gcc.gnu.org/gcc-7/changes.html
5076   9. http://gcc.gnu.org/onlinedocs/7.2.0/
5077  10. http://gcc.gnu.org/gcc-7/changes.html
5078  11. http://gcc.gnu.org/onlinedocs/7.1.0/
5079  12. http://gcc.gnu.org/gcc-7/buildstat.html
5080  13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
5081  14. http://gcc.gnu.org/index.html
5082  15. mailto:gcc@gcc.gnu.org
5083  16. http://gcc.gnu.org/mirrors.html
5084  17. http://gcc.gnu.org/git.html
5085  18. https://gcc.gnu.org/onlinedocs/
5086  19. mailto:gcc-help@gcc.gnu.org
5087  20. mailto:gcc@gcc.gnu.org
5088  21. https://gcc.gnu.org/lists.html
5089  22. https://www.fsf.org/
5090  23. https://gcc.gnu.org/about.html
5091======================================================================
5092http://gcc.gnu.org/gcc-7/changes.html
5093
5094                              GCC 7 Release Series
5095                        Changes, New Features, and Fixes
5096
5097   This page is a brief summary of some of the huge number of improvements
5098   in GCC 7. For more information, see the [1]Porting to GCC 7 page and
5099   the [2]full GCC documentation.
5100
5101Caveats
5102
5103     * GCC now uses [3]LRA (a new local register allocator) by default for
5104       new targets.
5105     * The non-standard C++0x type traits has_trivial_default_constructor,
5106       has_trivial_copy_constructor and has_trivial_copy_assign have been
5107       removed.
5108     * The libstdc++ [4]Profile Mode has been deprecated and will be
5109       removed in a future version.
5110     * The Cilk+ extensions to the C and C++ languages have been
5111       deprecated.
5112     * On ARM targets (arm*-*-*), [5]a bug introduced in GCC 5 that
5113       affects conformance to the procedure call standard (AAPCS) has been
5114       fixed. The bug affects some C++ code where class objects are passed
5115       by value to functions and could result in incorrect or inconsistent
5116       code being generated. This is an ABI change. If the option -Wpsabi
5117       is enabled (on by default) the compiler will emit a diagnostic note
5118       for code that might be affected.
5119
5120General Optimizer Improvements
5121
5122     * GCC 7 can determine the return value or range of return values of
5123       some calls to the sprintf family of functions and make it available
5124       to other optimization passes. Some calls to the snprintf function
5125       with a zero size argument can be folded into constants. This
5126       optimization is included in -O1 and can be selectively controlled
5127       by the -fprintf-return-value option.
5128     * A new store merging pass has been added. It merges constant stores
5129       to adjacent memory locations into fewer, wider, stores. It is
5130       enabled by the -fstore-merging option and at the -O2 optimization
5131       level or higher (and -Os).
5132     * A new code hoisting optimization has been added to the partial
5133       redundancy elimination pass. It attempts to move evaluation of
5134       expressions executed on all paths to the function exit as early as
5135       possible. This primarily helps improve code size, but can improve
5136       the speed of the generated code as well. It is enabled by the
5137       -fcode-hoisting option and at the -O2 optimization level or higher
5138       (and -Os).
5139     * A new interprocedural bitwise constant propagation optimization has
5140       been added, which propagates knowledge about which bits of
5141       variables are known to be zero (including pointer alignment
5142       information) across the call graph. It is enabled by the
5143       -fipa-bit-cp option if -fipa-cp is enabled as well, and is enabled
5144       at the -O2 optimization level and higher (and -Os). This
5145       optimization supersedes interprocedural alignment propagation of
5146       GCC 6, and therefore the option -fipa-cp-alignment is now
5147       deprecated and ignored.
5148     * A new interprocedural value range propagation optimization has been
5149       added, which propagates integral range information across the call
5150       graph when variable values can be proven to be within those ranges.
5151       It is enabled by the -fipa-vrp option and at the -O2 optimization
5152       level and higher (and -Os).
5153     * A new loop splitting optimization pass has been added. Certain
5154       loops which contain a condition that is always true on one side of
5155       the iteration space and always false on the other are split into
5156       two loops, such that each of the two new loops iterates on just one
5157       side of the iteration space and the condition does not need to be
5158       checked inside of the loop. It is enabled by the -fsplit-loops
5159       option and at the -O3 optimization level or higher.
5160     * The shrink-wrapping optimization can now separate portions of
5161       prologues and epilogues to improve performance if some of the work
5162       done traditionally by prologues and epilogues is not needed on
5163       certain paths. This is controlled by the -fshrink-wrap-separate
5164       option, enabled by default. It requires target support, which is
5165       currently only implemented in the PowerPC and AArch64 ports.
5166     * AddressSanitizer gained a new sanitization option,
5167       -fsanitize-address-use-after-scope, which enables sanitization of
5168       variables whose address is taken and used after a scope where the
5169       variable is defined:
5170
5171int
5172main (int argc, char **argv)
5173{
5174  char *ptr;
5175    {
5176      char my_char;
5177      ptr = &my_char;
5178    }
5179
5180  *ptr = 123;
5181  return *ptr;
5182}
5183
5184==28882==ERROR: AddressSanitizer: stack-use-after-scope on address 0x7fffb8dba99
51850 at pc 0x0000004006d5 bp 0x7fffb8dba960 sp 0x7fffb8dba958
5186WRITE of size 1 at 0x7fffb8dba990 thread T0
5187    #0 0x4006d4 in main /tmp/use-after-scope-1.c:10
5188    #1 0x7f9c71943290 in __libc_start_main (/lib64/libc.so.6+0x20290)
5189    #2 0x400739 in _start (/tmp/a.out+0x400739)
5190
5191Address 0x7fffb8dba990 is located in stack of thread T0 at offset 32 in frame
5192    #0 0x40067f in main /tmp/use-after-scope-1.c:3
5193
5194  This frame has 1 object(s):
5195    [32, 33) 'my_char' <== Memory access at offset 32 is inside this variable
5196
5197       The option is enabled by default with -fsanitize=address and
5198       disabled by default with -fsanitize=kernel-address. Compared to the
5199       LLVM compiler, where the option already exists, the implementation
5200       in the GCC compiler has some improvements and advantages:
5201          + Complex uses of gotos and case labels are properly handled and
5202            should not report any false positive or false negatives.
5203          + C++ temporaries are sanitized.
5204          + Sanitization can handle invalid memory stores that are
5205            optimized out by the LLVM compiler when optimization is
5206            enabled.
5207     * The -fsanitize=signed-integer-overflow suboption of the
5208       UndefinedBehavior Sanitizer now diagnoses arithmetic overflows even
5209       on arithmetic operations with generic vectors.
5210     * Version 5 of the DWARF debugging information standard is supported
5211       through the -gdwarf-5 option. The DWARF version 4 debugging
5212       information remains the default until consumers of debugging
5213       information are adjusted.
5214
5215New Languages and Language specific improvements
5216
5217   OpenACC support in C, C++, and Fortran continues to be maintained and
5218   improved. See the [6]OpenACC and [7]Offloading wiki pages for further
5219   information.
5220
5221  Ada
5222
5223     * On mainstream native platforms, Ada programs no longer require the
5224       stack to be made executable in order to run properly.
5225
5226  BRIG (HSAIL)
5227
5228   Support for processing BRIG 1.0 files was added in this release. BRIG
5229   is a binary format for HSAIL (Heterogeneous System Architecture
5230   Intermediate Language). The BRIG front end can be used for implementing
5231   HSAIL "finalizers" (compilation of HSAIL to a native ISA) for
5232   GCC-supported targets. An implementation of an HSAIL runtime library,
5233   libhsail-rt is also included.
5234
5235  C family
5236
5237     * New command-line options have been added for the C and C++
5238       compilers:
5239          + -Wimplicit-fallthrough warns when a switch case falls through.
5240            This warning has five different levels. The compiler is able
5241            to parse a wide range of fallthrough comments, depending on
5242            the level. It also handles control-flow statements, such as
5243            ifs. It's possible to suppress the warning by either adding a
5244            fallthrough comment, or by using a null statement:
5245            __attribute__ ((fallthrough)); (C, C++), or [[fallthrough]];
5246            (C++17), or [[gnu::fallthrough]]; (C++11/C++14). This warning
5247            is enabled by -Wextra.
5248          + -Wpointer-compare warns when a pointer is compared with a zero
5249            character constant. Such code is now invalid in C++11 and GCC
5250            rejects it. This warning is enabled by default.
5251          + -Wduplicated-branches warns when an if-else has identical
5252            branches.
5253          + -Wrestrict warns when an argument passed to a
5254            restrict-qualified parameter aliases with another argument.
5255          + -Wmemset-elt-size warns for memset calls, when the first
5256            argument references an array, and the third argument is a
5257            number equal to the number of elements of the array, but not
5258            the size of the array. This warning is enabled by -Wall.
5259          + -Wint-in-bool-context warns about suspicious uses of integer
5260            values where boolean values are expected. This warning is
5261            enabled by -Wall.
5262          + -Wswitch-unreachable warns when a switch statement has
5263            statements between the controlling expression and the first
5264            case label which will never be executed. This warning is
5265            enabled by default.
5266          + -Wexpansion-to-defined warns when defined is used outside #if.
5267            This warning is enabled by -Wextra or -Wpedantic.
5268          + -Wregister warns about uses of the register storage specifier.
5269            In C++17 this keyword has been removed and for C++17 this is a
5270            pedantic warning enabled by default. The warning is not
5271            emitted for the GNU Explicit Register Variables extension.
5272          + -Wvla-larger-than=N warns about unbounded uses of
5273            variable-length arrays, and about bounded uses of
5274            variable-length arrays whose bound can be larger than N bytes.
5275          + -Wduplicate-decl-specifier warns when a declaration has
5276            duplicate const, volatile, restrict or _Atomic specifier. This
5277            warning is enabled by -Wall.
5278     * GCC 6's C and C++ front ends were able to offer suggestions for
5279       misspelled field names:
5280
5281spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
5282you mean 'color'?
5283   return ptr->colour;
5284               ^~~~~~
5285
5286       GCC 7 greatly expands the scope of these suggestions. Firstly, it
5287       adds fix-it hints to such suggestions:
5288
5289spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
5290you mean 'color'?
5291   return ptr->colour;
5292               ^~~~~~
5293               color
5294
5295       The suggestions now cover many other things, such as misspelled
5296       function names:
5297
5298spellcheck-identifiers.c:11:3: warning: implicit declaration of function 'gtk_wi
5299dget_showall'; did you mean 'gtk_widget_show_all'? [-Wimplicit-function-declarat
5300ion]
5301   gtk_widget_showall (w);
5302   ^~~~~~~~~~~~~~~~~~
5303   gtk_widget_show_all
5304
5305       misspelled macro names and enum values:
5306
5307spellcheck-identifiers.cc:85:11: error: 'MAX_ITEM' undeclared here (not in a fun
5308ction); did you mean 'MAX_ITEMS'?
5309 int array[MAX_ITEM];
5310           ^~~~~~~~
5311           MAX_ITEMS
5312
5313       misspelled type names:
5314
5315spellcheck-typenames.c:7:14: error: unknown type name 'singed'; did you mean 'si
5316gned'?
5317 void test (singed char e);
5318            ^~~~~~
5319            signed
5320
5321       and, in the C front end, named initializers:
5322
5323test.c:7:20: error: 'struct s' has no member named 'colour'; did you mean 'color
5324'?
5325 struct s test = { .colour = 3 };
5326                    ^~~~~~
5327                    color
5328
5329     * The preprocessor can now offer suggestions for misspelled
5330       directives, e.g.:
5331
5332test.c:5:2: error:invalid preprocessing directive #endfi; did you mean #endif?
5333 #endfi
5334  ^~~~~
5335  endif
5336
5337     * Warnings about format strings now underline the pertinent part of
5338       the string, and can offer suggested fixes. In some cases, the
5339       pertinent argument is underlined.
5340
5341test.c:51:29: warning: format '%s' expects argument of type 'char *', but argume
5342nt 3 has type 'int' [-Wformat=]
5343   printf ("foo: %d  bar: %s baz: %d", 100, i + j, 102);
5344                          ~^                ~~~~~
5345                          %d
5346
5347     * The new -Wdangling-else command-line option has been split out of
5348       -Wparentheses and warns about dangling else.
5349     * The -Wshadow warning has been split into three variants.
5350       -Wshadow=global warns for any shadowing. This is the default when
5351       using -Wshadow without any argument. -Wshadow=local only warns for
5352       a local variable shadowing another local variable or parameter.
5353       -Wshadow=compatible-local only warns for a local variable shadowing
5354       another local variable or parameter whose type is compatible (in
5355       C++ compatible means that the type of the shadowing variable can be
5356       converted to that of the shadowed variable).
5357       The following example shows the different kinds of shadow warnings:
5358
5359enum operation { add, count };
5360struct container { int nr; };
5361
5362int
5363container_count (struct container c, int count)
5364{
5365  int r = 0;
5366  for (int count = 0; count > 0; count--)
5367    {
5368      struct container count = c;
5369      r += count.nr;
5370    }
5371  return r;
5372}
5373
5374       -Wshadow=compatible-local will warn for the parameter being
5375       shadowed with the same type:
5376
5377warn-test.c:8:12: warning: declaration of 'count' shadows a parameter [-Wshadow=
5378compatible-local]
5379   for (int count = 0; count > 0; count--)
5380            ^~~~~
5381warn-test.c:5:42: note: shadowed declaration is here
5382 container_count (struct container c, int count)
5383                                          ^~~~~
5384
5385       -Wshadow=local will warn for the above and for the shadowed
5386       declaration with incompatible type:
5387
5388warn-test.c:10:24: warning: declaration of 'count' shadows a previous local [-Ws
5389hadow=local]
5390       struct container count = c;
5391                        ^~~~~
5392warn-test.c:8:12: note: shadowed declaration is here
5393   for (int count = 0; count > 0; count--)
5394            ^~~~~
5395
5396       -Wshadow=global will warn for all of the above and the shadowing of
5397       the global declaration:
5398
5399warn-test.c:5:42: warning: declaration of 'count' shadows a global declaration [
5400-Wshadow]
5401 container_count (struct container c, int count)
5402                                          ^~~~~
5403warn-test.c:1:23: note: shadowed declaration is here
5404 enum operation { add, count };
5405                       ^~~~~
5406
5407     * GCC 7 contains a number of enhancements that help detect buffer
5408       overflow and other forms of invalid memory accesses.
5409          + The -Walloc-size-larger-than=size option detects calls to
5410            standard and user-defined memory allocation functions
5411            decorated with attribute alloc_size whose argument exceeds the
5412            specified size (PTRDIFF_MAX by default). The option also
5413            detects arithmetic overflow in the computation of the size in
5414            two-argument allocation functions like calloc where the total
5415            size is the product of the two arguments. Since calls with an
5416            excessive size cannot succeed they are typically the result of
5417            programming errors. Such bugs have been known to be the source
5418            of security vulnerabilities and a target of exploits.
5419            -Walloc-size-larger-than=PTRDIFF_MAX is included in -Wall.
5420            For example, the following call to malloc incorrectly tries to
5421            avoid passing a negative argument to the function and instead
5422            ends up unconditionally invoking it with an argument less than
5423            or equal to zero. Since after conversion to the type of the
5424            argument of the function (size_t) a negative argument results
5425            in a value in excess of the maximum PTRDIFF_MAX the call is
5426            diagnosed.
5427
5428void* f (int n)
5429{
5430  return malloc (n > 0 ? 0 : n);
5431}
5432
5433warning: argument 1 range [2147483648, 4294967295] exceeds maximum object size 2
5434147483647 [-Walloc-size-larger-than=]
5435
5436          + The -Walloc-zero option detects calls to standard and
5437            user-defined memory allocation functions decorated with
5438            attribute alloc_size with a zero argument. -Walloc-zero is not
5439            included in either -Wall or -Wextra and must be explicitly
5440            enabled.
5441          + The -Walloca option detects all calls to the alloca function
5442            in the program. -Walloca is not included in either -Wall or
5443            -Wextra and must be explicitly enabled.
5444          + The -Walloca-larger-than=size option detects calls to the
5445            alloca function whose argument either may exceed the specified
5446            size, or that is not known to be sufficiently constrained to
5447            avoid exceeding it. -Walloca-larger-than is not included in
5448            either -Wall or -Wextra and must be explicitly enabled.
5449            For example, compiling the following snippet with
5450            -Walloca-larger-than=1024 results in a warning because even
5451            though the code appears to call alloca only with sizes of 1kb
5452            and less, since n is signed, a negative value would result in
5453            a call to the function well in excess of the limit.
5454
5455void f (int n)
5456{
5457  char *d;
5458  if (n < 1025)
5459    d = alloca (n);
5460  else
5461    d = malloc (n);
5462  ...
5463}
5464
5465warning: argument to 'alloca may be too large due to conversion from 'int' to 'l
5466ong unsigned int' [-Walloca-larger-than=]
5467
5468            In contrast, a call to alloca that isn't bounded at all such
5469            as in the following function will elicit the warning below
5470            regardless of the size argument to the option.
5471
5472void f (size_t n)
5473{
5474  char *d = alloca (n);
5475  ...
5476}
5477
5478warning: unbounded use of 'alloca' [-Walloca-larger-than=]
5479
5480          + The -Wformat-overflow=level option detects certain and likely
5481            buffer overflow in calls to the sprintf family of formatted
5482            output functions. Although the option is enabled even without
5483            optimization it works best with -O2 and higher.
5484            For example, in the following snippet the call to sprintf is
5485            diagnosed because even though its output has been constrained
5486            using the modulo operation it could result in as many as three
5487            bytes if mday were negative. The solution is to either
5488            allocate a larger buffer or make sure the argument is not
5489            negative, for example by changing mday's type to unsigned or
5490            by making the type of the second operand of the modulo
5491            expression unsigned: 100U.
5492
5493void* f (int mday)
5494{
5495  char *buf = malloc (3);
5496  sprintf (buf, "%02i", mday % 100);
5497  return buf;
5498}
5499
5500warning: 'sprintf may write a terminating nul past the end of the destination [-
5501Wformat-overflow=]
5502note: 'sprintf' output between 3 and 4 bytes into a destination of size 3
5503
5504          + The -Wformat-truncation=level option detects certain and
5505            likely output truncation in calls to the snprintf family of
5506            formatted output functions. -Wformat-truncation=1 is included
5507            in -Wall and enabled without optimization but works best with
5508            -O2 and higher.
5509            For example, the following function attempts to format an
5510            integer between 0 and 255 in hexadecimal, including the 0x
5511            prefix, into a buffer of four characters. But since the
5512            function must always terminate output by the null character
5513            ('\0') such a buffer is only big enough to fit just one digit
5514            plus the prefix. Therefore the snprintf call is diagnosed. To
5515            avoid the warning either use a bigger buffer or handle the
5516            function's return value which indicates whether or not its
5517            output has been truncated.
5518
5519void f (unsigned x)
5520{
5521  char d[4];
5522  snprintf (d, sizeof d, "%#02x", x & 0xff);
5523  ...
5524}
5525
5526warning: 'snprintf' output may be truncated before the last format character [-W
5527format-truncation=]
5528note: 'snprintf' output between 3 and 5 bytes into a destination of size 4
5529
5530          + The -Wnonnull option has been enhanced to detect a broader set
5531            of cases of passing null pointers to functions that expect a
5532            non-null argument (those decorated with attribute nonnull). By
5533            taking advantage of optimizations the option can detect many
5534            more cases of the problem than in prior GCC versions.
5535          + The -Wstringop-overflow=type option detects buffer overflow in
5536            calls to string handling functions like memcpy and strcpy. The
5537            option relies on [8]Object Size Checking and has an effect
5538            similar to defining the _FORTIFY_SOURCE macro.
5539            -Wstringop-overflow=2 is enabled by default.
5540            For example, in the following snippet, because the call to
5541            strncat specifies a maximum that allows the function to write
5542            past the end of the destination, it is diagnosed. To correct
5543            the problem and avoid the overflow the function should be
5544            called with a size of at most sizeof d - strlen(d) - 1.
5545
5546void f (const char *fname)
5547{
5548  char d[8];
5549  strncpy (d, "/tmp/", sizeof d);
5550  strncat (d, fname, sizeof d);
5551  ...
5552}
5553
5554warning: specified bound 8 equals the size of the destination [-Wstringop-overfl
5555ow=]
5556
5557     * The <limits.h> header provided by GCC defines macros such as
5558       INT_WIDTH for the width in bits of integer types, if
5559       __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is
5560       included. The <stdint.h> header defines such macros as SIZE_WIDTH
5561       and INTMAX_WIDTH for the width of some standard typedef names for
5562       integer types, again if __STDC_WANT_IEC_60559_BFP_EXT__ is defined
5563       before the header is included; note that GCC's implementation of
5564       this header is only used for freestanding compilations, not hosted
5565       compilations, on most systems. These macros come from ISO/IEC TS
5566       18661-1:2014.
5567     * The <float.h> header provided by GCC defines the macro
5568       CR_DECIMAL_DIG, from ISO/IEC TS 18661-1:2014, if
5569       __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is
5570       included. This represents the number of decimal digits for which
5571       conversions between decimal character strings and binary formats,
5572       in both directions, are correctly rounded, and currently has the
5573       value of UINTMAX_MAX on all systems, reflecting that GCC's
5574       compile-time conversions are correctly rounded for any number of
5575       digits.
5576     * New __builtin_add_overflow_p, __builtin_sub_overflow_p,
5577       __builtin_mul_overflow_p built-in functions have been added. These
5578       work similarly to their siblings without the _p suffix, but do not
5579       actually store the result of the arithmetics anywhere, just return
5580       whether the operation would overflow. Calls to these built-ins with
5581       integer constant arguments evaluate to integer constants
5582       expressions.
5583       For example, in the following, c is assigned the result of a * b
5584       only if the multiplication does not overflow, otherwise it is
5585       assigned the value zero. The multiplication is performed at
5586       compile-time and without triggering a -Woverflow warning.
5587
5588enum {
5589  a = 12345678,
5590  b = 87654321,
5591  c = __builtin_mul_overflow_p (a, b, a) ? 0 : a * b
5592};
5593
5594  C
5595
5596     * The C front end now supports type names _FloatN for floating-point
5597       types with IEEE interchange formats and _FloatNx for floating-point
5598       types with IEEE extended formats. These type names come from
5599       ISO/IEC TS 18661-3:2015.
5600       The set of types supported depends on the target for which GCC is
5601       configured. Most targets support _Float32, _Float32x and _Float64.
5602       _Float128 is supported on targets where IEEE binary128 encoding was
5603       already supported as long double or __float128. _Float64x is
5604       supported on targets where a type with either binary128 or Intel
5605       extended precision format is available.
5606       Constants with these types are supported using suffixes fN, FN, fNx
5607       and FNx (e.g., 1.2f128 or 2.3F64x). Macros such as FLT128_MAX are
5608       defined in <float.h> if __STDC_WANT_IEC_60559_TYPES_EXT__ is
5609       defined before it is included.
5610       These new types are always distinct from each other and from float,
5611       double and long double, even if they have the same encoding.
5612       Complex types such as _Complex _Float128 are also supported.
5613       Type-generic built-in functions such as __builtin_isinf support the
5614       new types, and the following type-specific built-in functions have
5615       versions (suffixed fN or fNx) for the new types:
5616       __builtin_copysign, __builtin_fabs, __builtin_huge_val,
5617       __builtin_inf, __builtin_nan, __builtin_nans.
5618     * Compilation with -fopenmp is now compatible with the C11 _Atomic
5619       keyword.
5620
5621  C++
5622
5623     * The C++ front end has experimental support for all of the current
5624       C++17 draft with the -std=c++1z or -std=gnu++1z flags, including if
5625       constexpr, class template argument deduction, auto template
5626       parameters, and structured bindings. For a full list of new
5627       features, see [9]the C++ status page.
5628     * C++17 support for new of over-aligned types can be enabled in other
5629       modes with the -faligned-new flag.
5630     * The C++17 evaluation order requirements can be selected in other
5631       modes with the -fstrong-eval-order flag, or disabled in C++17 mode
5632       with -fno-strong-eval-order.
5633     * The default semantics of inherited constructors has changed in all
5634       modes, following [10]P0136. Essentially, overload resolution
5635       happens as if calling the inherited constructor directly, and the
5636       compiler fills in construction of the other bases and members as
5637       needed. Most uses should not need any changes. The old behavior can
5638       be restored with -fno-new-inheriting-ctors, or -fabi-version less
5639       than 11.
5640     * The resolution of DR 150 on matching of template template
5641       parameters, allowing default template arguments to make a template
5642       match a parameter, is currently enabled by default in C++17 mode
5643       only. The default can be overridden with -f{no-,}new-ttp-matching.
5644     * The C++ front end will now provide fix-it hints for some missing
5645       semicolons, allowing for automatic fixes by IDEs:
5646
5647test.cc:4:11: error: expected ';' after class definition
5648 class a {}
5649           ^
5650           ;
5651
5652     * -Waligned-new has been added to the C++ front end. It warns about
5653       new of type with extended alignment without -faligned-new.
5654
5655    Runtime Library (libstdc++)
5656
5657     * The type of exception thrown by iostreams, std::ios_base::failure,
5658       now uses the [11]cxx11 ABI.
5659     * Experimental support for C++17, including the following new
5660       features:
5661          + std::string_view;
5662          + std::any, std::optional, and std::variant;
5663          + std::invoke, std::is_invocable, std::is_nothrow_invocable, and
5664            invoke_result;
5665          + std::is_swappable, and std::is_nothrow_swappable;
5666          + std::apply, and std::make_from_tuple;
5667          + std::void_t, std::bool_constant, std::conjunction,
5668            std::disjunction, and std::negation;
5669          + Variable templates for type traits;
5670          + Mathematical Special Functions;
5671          + std::chrono::floor, std::chrono::ceil, std::chrono::round, and
5672            std::chrono::abs;
5673          + std::clamp, std::gcd, std::lcm, 3-dimensional std::hypot;
5674          + std::scoped_lock, std::shared_mutex,
5675            std::atomic<T>::is_always_lock_free;
5676          + std::sample, std::default_searcher, std::boyer_moore_searcher
5677            and std::boyer_moore_horspool_searcher;
5678          + Extraction and re-insertion of map and set nodes, try_emplace
5679            members for maps, and functions for accessing containers
5680            std::size, std::empty, and std::data;
5681          + std::shared_ptr support for arrays,
5682            std::shared_ptr<T>::weak_type,
5683            std::enable_shared_from_this<T>::weak_from_this(), and
5684            std::owner_less<void>;
5685          + std::byte;
5686          + std::as_const, std::not_fn,
5687            std::has_unique_object_representations, constexpr
5688            std::addressof.
5689       Thanks to Daniel Kr�gler, Tim Shen, Edward Smith-Rowland, and Ville
5690       Voutilainen for work on the C++17 support.
5691     * A new power-of-two rehashing policy for use with the _Hashtable
5692       internals, thanks to Fran�ois Dumont.
5693
5694  Fortran
5695
5696     * Support for a number of extensions for compatibility with legacy
5697       code with new flags:
5698          + -fdec-structure Support for DEC STRUCTURE and UNION
5699          + -fdec-intrinsic-ints Support for new integer intrinsics with
5700            B/I/J/K prefixes such as BABS, JIAND...
5701          + -fdec-math Support for additional math intrinsics, including
5702            COTAN and degree-valued trigonometric functions such as TAND,
5703            ASIND...
5704          + -fdec Enable the -fdec-* family of extensions.
5705     * New flag -finit-derived to allow default initialization of
5706       derived-type variables.
5707     * Improved DO loops with step equal to 1 or -1, generates faster code
5708       without a loop preheader. A new warning, -Wundefined-do-loop, warns
5709       when a loop iterates either to HUGE(i) (with step equal to 1), or
5710       to -HUGE(i) (with step equal to -1). Invalid behavior can be caught
5711       at run time with -fcheck=do enabled:
5712
5713program test
5714  implicit none
5715  integer(1) :: i
5716  do i = -HUGE(i)+10, -HUGE(i)-1, -1
5717    print *, i
5718  end do
5719end program test
5720
5721At line 8 of file do_check_12.f90
5722Fortran runtime error: Loop iterates infinitely
5723
5724     * Version 4.5 of the [12]OpenMP specification is now partially
5725       supported in the Fortran compiler; the largest missing item is
5726       structure element mapping.
5727     * User-defined derived-type input/output (UDTIO) is added.
5728     * Derived type coarrays with allocatable and pointer components are
5729       partially supported.
5730     * Non-constant stop codes and error stop codes (Fortran 2015
5731       feature).
5732     * Derived types with allocatable components of recursive type.
5733     * Intrinsic assignment to polymorphic variables.
5734     * Improved submodule support.
5735     * Improved diagnostics (polymorphic results in pure functions).
5736     * Coarray: Support for failed images (Fortan 2015 feature).
5737
5738  Go
5739
5740     * GCC 7 provides a complete implementation of the Go 1.8.1 user
5741       packages.
5742     * Compared to the Go 1.8.1 toolchain, the garbage collector is more
5743       conservative and less concurrent.
5744     * Escape analysis is available for experimental use via the
5745       -fgo-optimize-allocs option. The -fgo-debug-escape prints
5746       information useful for debugging escape analysis choices.
5747
5748  Java (GCJ)
5749
5750   The GCC Java front end and associated libjava runtime library have been
5751   removed from GCC.
5752
5753libgccjit
5754
5755   The libgccjit API gained support for marking calls as requiring
5756   tail-call optimization via a new entry point:
5757   [13]gcc_jit_rvalue_set_bool_require_tail_call.
5758
5759   libgccjit performs numerous checks at the API boundary, but if these
5760   succeed, it previously ignored errors and other diagnostics emitted
5761   within the core of GCC, and treated the compile of a gcc_jit_context as
5762   having succeeded. As of GCC 7 it now ensures that if any diagnostics
5763   are emitted, they are visible from the libgccjit API, and that the the
5764   context is flagged as having failed.
5765
5766New Targets and Target Specific Improvements
5767
5768  AArch64
5769
5770     * GCC has been updated to the latest revision of the procedure call
5771       standard (AAPCS64) to provide support for parameter passing when
5772       data types have been over-aligned.
5773     * The ARMv8.3-A architecture is now supported. It can be used by
5774       specifying the -march=armv8.3-a option.
5775     * The option -msign-return-address= is supported to enable return
5776       address protection using ARMv8.3-A Pointer Authentication
5777       Extensions. For more information on the arguments accepted by this
5778       option, please refer to [14]AArch64-Options.
5779     * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point
5780       Extensions are now supported. They can be used by specifying the
5781       -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit
5782       Floating-Point Extensions introduce new half-precision data
5783       processing floating-point instructions.
5784     * Support has been added for the following processors (GCC
5785       identifiers in parentheses): ARM Cortex-A73 (cortex-a73), Broadcom
5786       Vulcan (vulcan), Cavium ThunderX CN81xx (thunderxt81), Cavium
5787       ThunderX CN83xx (thunderxt83), Cavium ThunderX CN88xx
5788       (thunderxt88), Cavium ThunderX CN88xx pass 1.x (thunderxt88p1),
5789       Cavium ThunderX 2 CN99xx (thunderx2t99), Qualcomm Falkor (falkor).
5790       The GCC identifiers can be used as arguments to the -mcpu or -mtune
5791       options, for example: -mcpu=cortex-a73 or -mtune=vulcan or as
5792       arguments to the equivalent target attributes and pragmas.
5793
5794  ARC
5795
5796     * Added support for ARC HS and ARC EM processors.
5797     * Added support for ARC EM variation found in Intel QuarkSE SoCs.
5798     * Added support for NPS400 ARC700 based CPUs.
5799     * Thread Local Storage is now supported by ARC CPUs.
5800     * Fixed errors for ARC600 when using 32x16 multiplier option.
5801     * Fixed PIE for ARC CPUs.
5802     * New CPU templates are supported via multilib.
5803
5804  ARM
5805
5806     * Support for the ARMv5 and ARMv5E architectures has been deprecated
5807       (which have no known implementations) and will be removed in a
5808       future GCC release. Note that ARMv5T, ARMv5TE and ARMv5TEJ
5809       architectures remain supported. The values armv5 and armv5e of
5810       -march are thus deprecated.
5811     * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point
5812       Extensions are now supported. They can be used by specifying the
5813       -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit
5814       Floating-Point Extensions introduce new half-precision data
5815       processing floating-point instructions.
5816     * The ARMv8-M architecture is now supported in its two architecture
5817       profiles: ARMv8-M Baseline and ARMv8-M Mainline with its DSP and
5818       Floating-Point Extensions. They can be used by specifying the
5819       -march=armv8-m.base, armv8-m.main or armv8-m.main+dsp options.
5820     * Support has been added for the following processors (GCC
5821       identifiers in parentheses): ARM Cortex-A73 (cortex-a73), ARM
5822       Cortex-M23 (cortex-m23) and ARM Cortex-M33 (cortex-m33). The GCC
5823       identifiers can be used as arguments to the -mcpu or -mtune
5824       options, for example: -mcpu=cortex-a73 or -mtune=cortex-m33.
5825     * A new command-line option -mpure-code has been added. It does not
5826       allow constant data to be placed in code sections. This option is
5827       only available when generating non-PIC code for ARMv7-M targets.
5828     * Support for the ACLE Coprocessor Intrinsics has been added. This
5829       enables the generation of coprocessor instructions through the use
5830       of intrinsics such as cdp, ldc, and others.
5831     * The configure option --with-multilib-list now accepts the value
5832       rmprofile to build multilib libraries for a range of embedded
5833       targets. See our [15]installation instructions for details.
5834
5835  AVR
5836
5837     * On the reduced Tiny cores, the progmem [16]variable attribute is
5838       now properly supported. Respective read-only variables are located
5839       in flash memory in section .progmem.data. No special code is needed
5840       to access such variables; the compiler automatically adds an offset
5841       of 0x4000 to all addresses, which is needed to access variables in
5842       flash memory. As opposed to ordinary cores where it is sufficient
5843       to specify the progmem attribute with definitions, on the reduced
5844       Tiny cores the attribute also has to be specified with (external)
5845       declarations:
5846
5847extern const int array[] __attribute__((__progmem__));
5848
5849int get_value2 (void)
5850{
5851  /* Access via addresses array + 0x4004 and array + 0x4005. */
5852  return array[2];
5853}
5854
5855const int* get_address (unsigned idx)
5856{
5857  /* Returns array + 0x4000 + 2 * idx. */
5858  return &array[idx];
5859}
5860
5861     * A new command-line option -Wmisspelled-isr has been added. It turns
5862       off -- or turns into errors -- warnings that are reported for
5863       interrupt service routines (ISRs) which don't follow AVR-LibC's
5864       naming convention of prefixing ISR names with __vector.
5865     * __builtin_avr_nops(n) is a new [17]built-in function that inserts n
5866       NOP instructions into the instruction stream. n must be a value
5867       known at compile time.
5868
5869  IA-32/x86-64
5870
5871     * Support for the AVX-512 Fused Multiply Accumulation Packed Single
5872       precision (4FMAPS), AVX-512 Vector Neural Network Instructions Word
5873       variable precision (4VNNIW), AVX-512 Vector Population Count
5874       (VPOPCNTDQ) and Software Guard Extensions (SGX) ISA extensions has
5875       been added.
5876
5877  NVPTX
5878
5879     * OpenMP target regions can now be offloaded to NVidia PTX GPGPUs.
5880       See the [18]Offloading Wiki on how to configure it.
5881
5882  PowerPC / PowerPC64 / RS6000
5883
5884     * The PowerPC port now uses LRA by default.
5885     * GCC now diagnoses inline assembly that clobbers register r2. This
5886       has always been invalid code, and is no longer quietly tolerated.
5887     * The PowerPC port's support for ISA 3.0 (-mcpu=power9) has been
5888       enhanced to generate more of the new instructions by default, and
5889       to provide more built-in functions to generate code for other new
5890       instructions.
5891     * The configuration option --enable-gnu-indirect-function is now
5892       enabled by default on PowerPC GNU/Linux builds.
5893     * The PowerPC port will now allow 64-bit and 32-bit integer types to
5894       be allocated to the VSX vector registers (ISA 2.06 and above). In
5895       addition, on ISA 3.0, 16-bit and 8-bit integer types can be
5896       allocated in the vector registers. Previously, only 64-bit integer
5897       types were allowed in the traditional floating point registers.
5898     * New options -mstack-protector-guard=global,
5899       -mstack-protector-guard=tls, -mstack-protector-guard-reg=, and
5900       -mstack-protector-guard-offset= change how the stack protector gets
5901       the value to use as canary.
5902
5903  S/390, System z, IBM z Systems, IBM Z
5904
5905     * Support for the IBM z14 processor has been added. When using the
5906       -march=z14 option, the compiler will generate code making use of
5907       the new instructions introduced with the vector enhancement
5908       facility and the miscellaneous instruction extension facility 2.
5909       The -mtune=z14 option enables z14 specific instruction scheduling
5910       without making use of new instructions.
5911     * Builtins for the new vector instructions have been added and can be
5912       enabled using the -mzvector option.
5913
5914  RISC-V
5915
5916     * Support for the RISC-V instruction set has been added.
5917
5918  RX
5919
5920   Basic support for atomic built-in function has been added. It is
5921   currently implemented by flipping interrupts off and on as needed.
5922
5923  SH
5924
5925     * Support for SH5/SH64 has been removed.
5926     * Improved utilization of delay slots on SH2A.
5927     * Improved utilization of zero-displacement conditional branches.
5928     * The following deprecated options have been removed
5929          + -mcbranchdi
5930          + -mcmpeqdi
5931          + -minvalid-symbols
5932          + -msoft-atomic
5933          + -mspace
5934          + -madjust-unroll
5935     * Support for the following SH2A instructions has been added
5936          + mov.b @-Rm,R0
5937          + mov.w @-Rm,R0
5938          + mov.l @-Rm,R0
5939          + mov.b R0,@Rn+
5940          + mov.w R0,@Rn+
5941          + mov.l R0,@Rn+
5942
5943  SPARC
5944
5945     * The SPARC port now uses LRA by default.
5946     * Support for the new Subtract-Extended-with-Carry instruction
5947       available in SPARC M7 (Niagara 7) has been added.
5948
5949Operating Systems
5950
5951  AIX
5952
5953     * Visibility support has been enabled for AIX 7.1 and above.
5954
5955  Fuchsia
5956
5957     * Support has been added for the [19]Fuchsia OS.
5958
5959  RTEMS
5960
5961     * The ABI changes on ARM so that no short enums are used by default.
5962
5963Other significant improvements
5964
5965     * -fverbose-asm previously emitted information on the meanings of
5966       assembly expressions. This has been extended so that it now also
5967       prints comments showing the source lines that correspond to the
5968       assembly, making it easier to read the generated assembly
5969       (especially with larger functions). For example, given this C
5970       source file:
5971
5972int test (int n)
5973{
5974  int i;
5975  int total = 0;
5976
5977  for (i = 0; i < n; i++)
5978    total += i * i;
5979  return total;
5980}
5981
5982       -fverbose-asm now gives output similar to this for the function
5983       body (when compiling for x86_64, with -Os):
5984
5985       .text
5986       .globl  test
5987       .type   test, @@function
5988test:
5989.LFB0:
5990       .cfi_startproc
5991# example.c:4:   int total = 0;
5992       xorl    %eax, %eax      # <retval>
5993# example.c:6:   for (i = 0; i < n; i++)
5994       xorl    %edx, %edx      # i
5995.L2:
5996# example.c:6:   for (i = 0; i < n; i++)
5997       cmpl    %edi, %edx      # n, i
5998       jge     .L5     #,
5999# example.c:7:     total += i * i;
6000       movl    %edx, %ecx      # i, tmp92
6001       imull   %edx, %ecx      # i, tmp92
6002# example.c:6:   for (i = 0; i < n; i++)
6003       incl    %edx    # i
6004# example.c:7:     total += i * i;
6005       addl    %ecx, %eax      # tmp92, <retval>
6006       jmp     .L2     #
6007.L5:
6008# example.c:10: }
6009       ret
6010       .cfi_endproc
6011
6012     * Two new options have been added for printing fix-it hints:
6013          + -fdiagnostics-parseable-fixits allows for fix-it hints to be
6014            emitted in a machine-readable form, suitable for consumption
6015            by IDEs. For example, given:
6016
6017spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
6018you mean 'color'?
6019   return ptr->colour;
6020               ^~~~~~
6021               color
6022
6023            it will emit:
6024
6025fix-it:"spellcheck-fields.cc":{52:13-52:19}:"color"
6026
6027          + -fdiagnostics-generate-patch will print a patch in "unified"
6028            format after any diagnostics are printed, showing the result
6029            of applying all fix-it hints. For the above example it would
6030            emit:
6031
6032--- spellcheck-fields.cc
6033+++ spellcheck-fields.cc
6034@@ -49,5 +49,5 @@
6035
6036 color get_color(struct s *ptr)
6037 {
6038-  return ptr->colour;
6039+  return ptr->color;
6040 }
6041
6042     * The gcc and g++ driver programs will now provide suggestions for
6043       misspelled arguments to command-line options.
6044
6045$ gcc -c test.c -ftls-model=global-dinamic
6046gcc: error: unknown TLS model 'global-dinamic'
6047gcc: note: valid arguments to '-ftls-model=' are: global-dynamic initial-exec lo
6048cal-dynamic local-exec; did you mean 'global-dynamic'?
6049
6050     * The compiler will now provide suggestions for misspelled
6051       parameters.
6052
6053$ gcc -c test.c --param max-early-inliner-iteration=3
6054cc1: error: invalid --param name 'max-early-inliner-iteration'; did you mean 'ma
6055x-early-inliner-iterations'?
6056
6057     * Profile-guided optimization (PGO) instrumentation, as well as test
6058       coverage (GCOV), can newly instrument constructors (functions marks
6059       with __attribute__((constructor))), destructors and C++
6060       constructors (and destructors) of classes that are used as the type
6061       of a global variable.
6062     * A new option -fprofile-update=atomic prevents creation of corrupted
6063       profiles created during an instrumentation run (-fprofile=generate)
6064       of an application. The downside of the option is a speed penalty.
6065       Providing -pthread on the command line selects atomic profile
6066       updating (when supported by the target).
6067     * GCC's already extensive testsuite has gained some new capabilities,
6068       to further improve the reliability of the compiler:
6069          + GCC now has an internal unit-testing API and a suite of tests
6070            for programmatic self-testing of subsystems.
6071          + GCC's C front end has been extended so that it can parse dumps
6072            of GCC's internal representations, allowing for DejaGnu tests
6073            that more directly exercise specific optimization passes. This
6074            covers both the [20]GIMPLE representation (for testing
6075            higher-level optimizations) and the [21]RTL representation,
6076            allowing for more direct testing of lower-level details, such
6077            as register allocation and instruction selection.
6078
6079GCC 7.1
6080
6081   This is the [22]list of problem reports (PRs) from GCC's bug tracking
6082   system that are known to be fixed in the 7.1 release. This list might
6083   not be complete (that is, it is possible that some PRs that have been
6084   fixed are not listed here).
6085
6086GCC 7.2
6087
6088   This is the [23]list of problem reports (PRs) from GCC's bug tracking
6089   system that are known to be fixed in the 7.2 release. This list might
6090   not be complete (that is, it is possible that some PRs that have been
6091   fixed are not listed here).
6092
6093  Target Specific Changes
6094
6095    SPARC
6096
6097     * Support for the SPARC M8 processor has been added.
6098     * The switches -mfix-ut700 and -mfix-gr712rc have been added to work
6099       around an erratum in LEON3FT processors.
6100     * Use of the Floating-point Multiply Single to Double (FsMULd)
6101       instruction can now be controlled by the -mfsmuld and -fno-fsmuld
6102       options.
6103
6104  Operating Systems
6105
6106    RTEMS
6107
6108     * The Ada run-time support uses now thread-local storage (TLS).
6109     * Support for RISC-V has been added.
6110     * Support for 64-bit PowerPC using the ELFv2 ABI with 64-bit long
6111       double has been added.
6112
6113GCC 7.3
6114
6115   This is the [24]list of problem reports (PRs) from GCC's bug tracking
6116   system that are known to be fixed in the 7.3 release. This list might
6117   not be complete (that is, it is possible that some PRs that have been
6118   fixed are not listed here).
6119
6120  Target Specific Changes
6121
6122    SPARC
6123
6124     * Workarounds for the four [25]LEON3FT errata GRLIB-TN-0010..0013
6125       have been added. Relevant errata are activated by the target
6126       specific -mfix-ut699, -mfix-ut700 and -mfix-gr712rc switches.
6127
6128  Operating Systems
6129
6130    RTEMS
6131
6132     * Support has been added for Epiphany target.
6133
6134GCC 7.4
6135
6136   This is the [26]list of problem reports (PRs) from GCC's bug tracking
6137   system that are known to be fixed in the 7.4 release. This list might
6138   not be complete (that is, it is possible that some PRs that have been
6139   fixed are not listed here).
6140
6141GCC 7.5
6142
6143   This is the [27]list of problem reports (PRs) from GCC's bug tracking
6144   system that are known to be fixed in the 7.5 release. This list might
6145   not be complete (that is, it is possible that some PRs that have been
6146   fixed are not listed here).
6147
6148
6149    For questions related to the use of GCC, please consult these web
6150    pages and the [28]GCC manuals. If that fails, the
6151    [29]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6152    web pages and the development of GCC are welcome on our developer
6153    list at [30]gcc@gcc.gnu.org. All of [31]our lists have public
6154    archives.
6155
6156   Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and
6157   distribution of this entire article is permitted in any medium,
6158   provided this notice is preserved.
6159
6160   These pages are [33]maintained by the GCC team. Last modified
6161   2022-10-26.
6162
6163References
6164
6165   1. http://gcc.gnu.org/gcc-7/porting_to.html
6166   2. http://gcc.gnu.org/onlinedocs/index.html#current
6167   3. https://gcc.gnu.org/wiki/LRAIsDefault
6168   4. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/manual/profile_mode.html
6169   5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77728
6170   6. https://gcc.gnu.org/wiki/OpenACC
6171   7. https://gcc.gnu.org/wiki/Offloading
6172   8. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/Object-Size-Checking.html
6173   9. https://gcc.gnu.org/projects/cxx-status.html#cxx1z
6174  10. https://wg21.link/p0136
6175  11. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/using_dual_abi.html
6176  12. https://www.openmp.org/specifications/
6177  13. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/jit/topics/expressions.html#gcc_jit_rvalue_set_bool_require_tail_call
6178  14. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AArch64-Options.html#AArch64-Options
6179  15. https://gcc.gnu.org/install/configure.html
6180  16. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Variable-Attributes.html
6181  17. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Built-in-Functions.html
6182  18. https://gcc.gnu.org/wiki/Offloading
6183  19. https://fuchsia.googlesource.com/
6184  20. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/GIMPLE-Tests.html
6185  21. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/RTL-Tests.html
6186  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.0
6187  23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.2
6188  24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.3
6189  25. https://www.gaisler.com/index.php/information/app-tech-notes
6190  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.4
6191  27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.5
6192  28. https://gcc.gnu.org/onlinedocs/
6193  29. mailto:gcc-help@gcc.gnu.org
6194  30. mailto:gcc@gcc.gnu.org
6195  31. https://gcc.gnu.org/lists.html
6196  32. https://www.fsf.org/
6197  33. https://gcc.gnu.org/about.html
6198======================================================================
6199http://gcc.gnu.org/gcc-6/index.html
6200
6201                              GCC 6 Release Series
6202
6203   (This release series is no longer supported.)
6204
6205   October 26, 2018
6206
6207   The [1]GNU project and the GCC developers are pleased to announce the
6208   release of GCC 6.5.
6209
6210   This release is a bug-fix release, containing fixes for regressions in
6211   GCC 6.4 relative to previous releases of GCC.
6212
6213Release History
6214
6215   GCC 6.5
6216          October 26, 2018 ([2]changes, [3]documentation)
6217
6218   GCC 6.4
6219          July 4, 2017 ([4]changes, [5]documentation)
6220
6221   GCC 6.3
6222          December 21, 2016 ([6]changes, [7]documentation)
6223
6224   GCC 6.2
6225          August 22, 2016 ([8]changes, [9]documentation)
6226
6227   GCC 6.1
6228          April 27, 2016 ([10]changes, [11]documentation)
6229
6230References and Acknowledgements
6231
6232   GCC used to stand for the GNU C Compiler, but since the compiler
6233   supports several other languages aside from C, it now stands for the
6234   GNU Compiler Collection.
6235
6236   A list of [12]successful builds is updated as new information becomes
6237   available.
6238
6239   The GCC developers would like to thank the numerous people that have
6240   contributed new features, improvements, bug fixes, and other changes as
6241   well as test results to GCC. This [13]amazing group of volunteers is
6242   what makes GCC successful.
6243
6244   For additional information about GCC please refer to the [14]GCC
6245   project web site or contact the [15]GCC development mailing list.
6246
6247   To obtain GCC please use [16]our mirror sites or [17]our version
6248   control system.
6249
6250
6251    For questions related to the use of GCC, please consult these web
6252    pages and the [18]GCC manuals. If that fails, the
6253    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6254    web pages and the development of GCC are welcome on our developer
6255    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
6256    archives.
6257
6258   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
6259   distribution of this entire article is permitted in any medium,
6260   provided this notice is preserved.
6261
6262   These pages are [23]maintained by the GCC team. Last modified
6263   2022-10-26.
6264
6265References
6266
6267   1. http://www.gnu.org/
6268   2. http://gcc.gnu.org/gcc-6/changes.html
6269   3. http://gcc.gnu.org/onlinedocs/6.5.0/
6270   4. http://gcc.gnu.org/gcc-6/changes.html
6271   5. http://gcc.gnu.org/onlinedocs/6.4.0/
6272   6. http://gcc.gnu.org/gcc-6/changes.html
6273   7. http://gcc.gnu.org/onlinedocs/6.3.0/
6274   8. http://gcc.gnu.org/gcc-6/changes.html
6275   9. http://gcc.gnu.org/onlinedocs/6.2.0/
6276  10. http://gcc.gnu.org/gcc-6/changes.html
6277  11. http://gcc.gnu.org/onlinedocs/6.1.0/
6278  12. http://gcc.gnu.org/gcc-6/buildstat.html
6279  13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6280  14. http://gcc.gnu.org/index.html
6281  15. mailto:gcc@gcc.gnu.org
6282  16. http://gcc.gnu.org/mirrors.html
6283  17. http://gcc.gnu.org/git.html
6284  18. https://gcc.gnu.org/onlinedocs/
6285  19. mailto:gcc-help@gcc.gnu.org
6286  20. mailto:gcc@gcc.gnu.org
6287  21. https://gcc.gnu.org/lists.html
6288  22. https://www.fsf.org/
6289  23. https://gcc.gnu.org/about.html
6290======================================================================
6291http://gcc.gnu.org/gcc-6/changes.html
6292
6293                              GCC 6 Release Series
6294                        Changes, New Features, and Fixes
6295
6296   This page is a brief summary of some of the huge number of improvements
6297   in GCC 6. For more information, see the [1]Porting to GCC 6 page and
6298   the [2]full GCC documentation.
6299
6300Caveats
6301
6302     * The default mode for C++ is now -std=gnu++14 instead of
6303       -std=gnu++98.
6304     * Support for a number of older systems and recently unmaintained or
6305       untested target ports of GCC has been declared obsolete in GCC 6.
6306       Unless there is activity to revive them, the next release of GCC
6307       will have their sources permanently removed.
6308       The following ports for individual systems on particular
6309       architectures have been obsoleted:
6310          + SH5 / SH64 (sh64-*-*) as announced [3]here.
6311     * The AVR port requires binutils version 2.26.1 or later for the fix
6312       for [4]PR71151 to work.
6313     * The GCC 6.5 release has an accidental ABI incompatibility for
6314       nested std::pair objects, for more details see [5]PR 87822. The bug
6315       causes a layout change for pairs where the first member is also a
6316       pair, e.g. std::pair<std::pair<X, Y>, Z>. The GCC 6 release series
6317       is closed so the bug in GCC 6.5 will not be fixed upstream, but
6318       there is a patch in the bug report to allow it to be fixed by
6319       anybody packaging GCC 6.5 or installing it themselves.
6320
6321General Optimizer Improvements
6322
6323     * UndefinedBehaviorSanitizer gained a new sanitization option,
6324       -fsanitize=bounds-strict, which enables strict checking of array
6325       bounds. In particular, it enables -fsanitize=bounds as well as
6326       instrumentation of flexible array member-like arrays.
6327     * Type-based alias analysis now disambiguates accesses to different
6328       pointers. This improves precision of the alias oracle by about
6329       20-30% on higher-level C++ programs. Programs doing invalid type
6330       punning of pointer types may now need -fno-strict-aliasing to work
6331       correctly.
6332     * Alias analysis now correctly supports the weakref and alias
6333       attributes. This allows accessing both a variable and its alias in
6334       one translation unit which is common with link-time optimization.
6335     * Value range propagation now assumes that the this pointer in C++
6336       member functions is non-null. This eliminates common null pointer
6337       checks but also breaks some non-conforming code-bases (such as
6338       Qt-5, Chromium, KDevelop). As a temporary work-around
6339       -fno-delete-null-pointer-checks can be used. Wrong code can be
6340       identified by using -fsanitize=undefined.
6341     * Link-time optimization improvements:
6342          + warning and error attributes are now correctly preserved by
6343            declaration linking and thus -D_FORTIFY_SOURCE=2 is now
6344            supported with -flto.
6345          + Type merging was fixed to handle C and Fortran
6346            interoperability rules as defined by the Fortran 2008 language
6347            standard.
6348            As an exception, CHARACTER(KIND=C_CHAR) is not inter-operable
6349            with char in all cases because it is an array while char is
6350            scalar. INTEGER(KIND=C_SIGNED_CHAR) should be used instead. In
6351            general, this inter-operability cannot be implemented, for
6352            example on targets where the argument passing convention for
6353            arrays differs from scalars.
6354          + More type information is now preserved at link time, reducing
6355            the loss of accuracy of the type-based alias analysis compared
6356            to builds without link-time optimization.
6357          + Invalid type punning on global variables and declarations is
6358            now reported with -Wodr-type-mismatch.
6359          + The size of LTO object files was reduced by about 11%
6360            (measured by compiling Firefox 46.0).
6361          + Link-time parallelization (enabled using -flto=n) was
6362            significantly improved by decreasing the size of streamed data
6363            when partitioning programs. The size of streamed IL while
6364            compiling Firefox 46.0 was reduced by 66%.
6365          + The linker plugin was extended to pass information about the
6366            type of binary produced to the GCC back end. (That can also be
6367            controlled manually by -flinker-output.) This makes it
6368            possible to properly configure the code generator and support
6369            incremental linking. Incremental linking of LTO objects by gcc
6370            -r is now supported for plugin-enabled setups.
6371            There are two ways to perform incremental linking:
6372              1. Linking by ld -r will result in an object file with all
6373                 sections from individual object files mechanically
6374                 merged. This delays the actual link-time optimization to
6375                 the final linking step and thus permits whole program
6376                 optimization. Linking the final binary with such object
6377                 files is however slower.
6378              2. Linking by gcc -r will lead to link-time optimization and
6379                 emit the final binary into the object file. Linking such
6380                 an object file is fast but avoids any benefits from whole
6381                 program optimization.
6382            GCC 7 will support incremental link-time optimization with gcc
6383            -r.
6384     * Inter-procedural optimization improvements:
6385          + Basic jump threading is now performed before profile
6386            construction and inline analysis, resulting in more realistic
6387            size and time estimates that drive the heuristics of the
6388            inliner and function cloning passes.
6389          + Function cloning now more aggressively eliminates unused
6390            function parameters.
6391
6392New Languages and Language specific improvements
6393
6394   Compared to GCC 5, the GCC 6 release series includes a much improved
6395   implementation of the [6]OpenACC 2.0a specification. Highlights are:
6396     * In addition to single-threaded host-fallback execution, offloading
6397       is supported for nvptx (Nvidia GPUs) on x86_64 and PowerPC 64-bit
6398       little-endian GNU/Linux host systems. For nvptx offloading, with
6399       the OpenACC parallel construct, the execution model allows for an
6400       arbitrary number of gangs, up to 32 workers, and 32 vectors.
6401     * Initial support for parallelized execution of OpenACC kernels
6402       constructs:
6403          + Parallelization of a kernels region is switched on by
6404            -fopenacc combined with -O2 or higher.
6405          + Code is offloaded onto multiple gangs, but executes with just
6406            one worker, and a vector length of 1.
6407          + Directives inside a kernels region are not supported.
6408          + Loops with reductions can be parallelized.
6409          + Only kernels regions with one loop nest are parallelized.
6410          + Only the outer-most loop of a loop nest can be parallelized.
6411          + Loop nests containing sibling loops are not parallelized.
6412       Typically, using the OpenACC parallel construct gives much better
6413       performance, compared to the initial support of the OpenACC kernels
6414       construct.
6415     * The device_type clause is not supported. The bind and nohost
6416       clauses are not supported. The host_data directive is not supported
6417       in Fortran.
6418     * Nested parallelism (cf. CUDA dynamic parallelism) is not supported.
6419     * Usage of OpenACC constructs inside multithreaded contexts (such as
6420       created by OpenMP, or pthread programming) is not supported.
6421     * If a call to the acc_on_device function has a compile-time constant
6422       argument, the function call evaluates to a compile-time constant
6423       value only for C and C++ but not for Fortran.
6424
6425   See the [7]OpenACC and [8]Offloading wiki pages for further
6426   information.
6427
6428  C family
6429
6430     * Version 4.5 of the [9]OpenMP specification is now supported in the
6431       C and C++ compilers.
6432     * The C and C++ compilers now support attributes on enumerators. For
6433       instance, it is now possible to mark enumerators as deprecated:
6434
6435enum {
6436  newval,
6437  oldval __attribute__ ((deprecated ("too old")))
6438};
6439
6440     * Source locations for the C and C++ compilers are now tracked as
6441       ranges, rather than just points, making it easier to identify the
6442       subexpression of interest within a complicated expression. For
6443       example:
6444
6445test.cc: In function 'int test(int, int, foo, int, int)':
6446test.cc:5:16: error: no match for 'operator*' (operand types are 'int' and 'foo'
6447)
6448   return p + q * r * s + t;
6449              ~~^~~
6450
6451       In addition, there is now initial support for precise diagnostic
6452       locations within strings:
6453
6454format-strings.c:3:14: warning: field width specifier '*' expects a matching 'in
6455t' argument [-Wformat=]
6456   printf("%*d");
6457            ^
6458
6459     * Diagnostics can now contain "fix-it hints", which are displayed in
6460       context underneath the relevant source code. For example:
6461
6462fixits.c: In function 'bad_deref':
6463fixits.c:11:13: error: 'ptr' is a pointer; did you mean to use '->'?
6464   return ptr.x;
6465             ^
6466             ->
6467
6468     * The C and C++ compilers now offer suggestions for misspelled field
6469       names:
6470
6471spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
6472you mean 'color'?
6473   return ptr->colour;
6474               ^~~~~~
6475
6476     * New command-line options have been added for the C and C++
6477       compilers:
6478          + -Wshift-negative-value warns about left shifting a negative
6479            value.
6480          + -Wshift-overflow warns about left shift overflows. This
6481            warning is enabled by default. -Wshift-overflow=2 also warns
6482            about left-shifting 1 into the sign bit.
6483          + -Wtautological-compare warns if a self-comparison always
6484            evaluates to true or false. This warning is enabled by -Wall.
6485          + -Wnull-dereference warns if the compiler detects paths that
6486            trigger erroneous or undefined behavior due to dereferencing a
6487            null pointer. This option is only active when
6488            -fdelete-null-pointer-checks is active, which is enabled by
6489            optimizations in most targets. The precision of the warnings
6490            depends on the optimization options used.
6491          + -Wduplicated-cond warns about duplicated conditions in an
6492            if-else-if chain.
6493          + -Wmisleading-indentation warns about places where the
6494            indentation of the code gives a misleading idea of the block
6495            structure of the code to a human reader. For example, given
6496            [10]CVE-2014-1266:
6497
6498sslKeyExchange.c: In function 'SSLVerifySignedServerKeyExchange':
6499sslKeyExchange.c:629:3: warning: this 'if' clause does not guard... [-Wmisleadin
6500g-indentation]
6501    if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0)
6502    ^~
6503sslKeyExchange.c:631:5: note: ...this statement, but the latter is misleadingly
6504indented as if it is guarded by the 'if'
6505        goto fail;
6506        ^~~~
6507
6508            This warning is enabled by -Wall.
6509     * The C and C++ compilers now emit saner error messages if
6510       merge-conflict markers are present in a source file.
6511
6512test.c:3:1: error: version control conflict marker in file
6513 <<<<<<< HEAD
6514 ^~~~~~~
6515
6516  C
6517
6518     * It is possible to disable warnings when an initialized field of a
6519       structure or a union with side effects is being overridden when
6520       using designated initializers via a new warning option
6521       -Woverride-init-side-effects.
6522     * A new type attribute scalar_storage_order applying to structures
6523       and unions has been introduced. It specifies the storage order (aka
6524       endianness) in memory of scalar fields in structures or unions.
6525
6526  C++
6527
6528     * The default mode has been changed to -std=gnu++14.
6529     * [11]C++ Concepts are now supported when compiling with -fconcepts.
6530     * -flifetime-dse is more aggressive in dead-store elimination in
6531       situations where a memory store to a location precedes a
6532       constructor to that memory location.
6533     * G++ now supports [12]C++17 fold expressions, u8 character literals,
6534       extended static_assert, and nested namespace definitions.
6535     * G++ now allows constant evaluation for all non-type template
6536       arguments.
6537     * G++ now supports C++ Transactional Memory when compiling with
6538       -fgnu-tm.
6539
6540    Runtime Library (libstdc++)
6541
6542     * Extensions to the C++ Library to support mathematical special
6543       functions (ISO/IEC 29124:2010), thanks to Edward Smith-Rowland.
6544     * Experimental support for C++17, including the following new
6545       features:
6546          + std::uncaught_exceptions function (this is also available for
6547            -std=gnu++NN modes);
6548          + new member functions try_emplace and insert_or_assign for
6549            unique_key maps;
6550          + non-member functions std::size, std::empty, and std::data for
6551            accessing containers and arrays;
6552          + std::invoke;
6553          + std::shared_mutex;
6554          + std::void_t and std::bool_constant metaprogramming utilities.
6555       Thanks to Ville Voutilainen for contributing many of the C++17
6556       features.
6557     * An experimental implementation of the File System TS.
6558     * Experimental support for most features of the second version of the
6559       Library Fundamentals TS. This includes polymorphic memory resources
6560       and array support in shared_ptr, thanks to Fan You.
6561     * Some assertions checked by Debug Mode can now also be enabled by
6562       _GLIBCXX_ASSERTIONS. The subset of checks enabled by the new macro
6563       have less run-time overhead than the full _GLIBCXX_DEBUG checks and
6564       don't affect the library ABI, so can be enabled per-translation
6565       unit.
6566     * Timed mutex types are supported on more targets, including Darwin.
6567     * Improved std::locale support for DragonFly and FreeBSD, thanks to
6568       John Marino and Andreas Tobler.
6569
6570  Fortran
6571
6572     * Fortran 2008 SUBMODULE support.
6573     * Fortran 2015 EVENT_TYPE, EVENT_POST, EVENT_WAIT, and EVENT_QUERY
6574       support.
6575     * Improved support for Fortran 2003 deferred-length character
6576       variables.
6577     * Improved support for OpenMP and OpenACC.
6578     * The MATMUL intrinsic is now inlined for straightforward cases if
6579       front-end optimization is active. The maximum size for inlining can
6580       be set to n with the -finline-matmul-limit=n option and turned off
6581       with -finline-matmul-limit=0.
6582     * The -Wconversion-extra option will warn about REAL constants which
6583       have excess precision for their kind.
6584     * The -Winteger-division option has been added, which warns about
6585       divisions of integer constants which are truncated. This option is
6586       included in -Wall by default.
6587
6588libgccjit
6589
6590     * The driver code is now run in-process within libgccjit, providing a
6591       small speed-up of the compilation process.
6592     * The API has gained entrypoints for
6593          + [13]timing how long was spent in different parts of code,
6594          + [14]creating switch statements,
6595          + [15]allowing unreachable basic blocks in a function, and
6596          + [16]adding arbitrary command-line options to a compilation.
6597
6598New Targets and Target Specific Improvements
6599
6600  AArch64
6601
6602     * A number of AArch64-specific options have been added. The most
6603       important ones are summarised in this section; for more detailed
6604       information please refer to the documentation.
6605     * The command-line options -march=native, -mcpu=native and
6606       -mtune=native are now available on native AArch64 GNU/Linux
6607       systems. Specifying these options causes GCC to auto-detect the
6608       host CPU and choose the optimal setting for that system.
6609     * -fpic is now supported when generating code for the small code
6610       model (-mcmodel=small). The size of the global offset table (GOT)
6611       is limited to 28KiB under the LP64 SysV ABI, and 15KiB under the
6612       ILP32 SysV ABI.
6613     * The AArch64 port now supports target attributes and pragmas. Please
6614       refer to the [17]documentation for details of available attributes
6615       and pragmas as well as usage instructions.
6616     * Link-time optimization across translation units with different
6617       target-specific options is now supported.
6618     * The option -mtls-size= is now supported. It can be used to specify
6619       the bit size of TLS offsets, allowing GCC to generate better TLS
6620       instruction sequences.
6621     * The option -fno-plt is now fully functional.
6622     * The ARMv8.1-A architecture and the Large System Extensions are now
6623       supported. They can be used by specifying the -march=armv8.1-a
6624       option. Additionally, the +lse option extension can be used in a
6625       similar fashion to other option extensions. The Large System
6626       Extensions introduce new instructions that are used in the
6627       implementation of atomic operations.
6628     * The ACLE half-precision floating-point type __fp16 is now supported
6629       in the C and C++ languages.
6630     * The ARM Cortex-A35 processor is now supported via the
6631       -mcpu=cortex-a35 and -mtune=cortex-a35 options as well as the
6632       equivalent target attributes and pragmas.
6633     * The Qualcomm QDF24xx processor is now supported via the
6634       -mcpu=qdf24xx and -mtune=qdf24xx options as well as the equivalent
6635       target attributes and pragmas.
6636     * Code generation for the ARM Cortex-A57 processor is improved. Among
6637       general code generation improvements, a better algorithm is added
6638       for allocating registers to floating-point multiply-accumulate
6639       instructions offering increased performance when compiling with
6640       -mcpu=cortex-a57 or -mtune=cortex-a57.
6641     * Code generation for the ARM Cortex-A53 processor is improved. A
6642       more accurate instruction scheduling model for the processor is now
6643       used, and a number of compiler tuning parameters have been set to
6644       offer increased performance when compiling with -mcpu=cortex-a53 or
6645       -mtune=cortex-a53.
6646     * Code generation for the Samsung Exynos M1 processor is improved. A
6647       more accurate instruction scheduling model for the processor is now
6648       used, and a number of compiler tuning parameters have been set to
6649       offer increased performance when compiling with -mcpu=exynos-m1 or
6650       -mtune=exynos-m1.
6651     * Improvements in the generation of conditional branches and literal
6652       pools allow the compiler to compile functions of a large size.
6653       Constant pools are now placed into separate rodata sections. The
6654       new option -mpc-relative-literal-loads generates per-function
6655       literal pools, limiting the maximum size of functions to 1MiB.
6656     * Several correctness issues generating Advanced SIMD instructions
6657       for big-endian targets have been fixed resulting in improved code
6658       generation for ACLE intrinsics with -mbig-endian.
6659
6660  ARM
6661
6662     * Support for revisions of the ARM architecture prior to ARMv4t has
6663       been deprecated and will be removed in a future GCC release. The
6664       -mcpu and -mtune values that are deprecated are: arm2, arm250,
6665       arm3, arm6, arm60, arm600, arm610, arm620, arm7, arm7d, arm7di,
6666       arm70, arm700, arm700i, arm710, arm720, arm710c, arm7100, arm7500,
6667       arm7500fe, arm7m, arm7dm, arm7dmi, arm8, arm810, strongarm,
6668       strongarm110, strongarm1100, strongarm1110, fa526, fa626. The value
6669       arm7tdmi is still supported. The values of -march that are
6670       deprecated are: armv2,armv2a,armv3,armv3m,armv4.
6671     * The ARM port now supports target attributes and pragmas. Please
6672       refer to the [18]documentation for details of available attributes
6673       and pragmas as well as usage instructions.
6674     * Support has been added for the following processors (GCC
6675       identifiers in parentheses): ARM Cortex-A32 (cortex-a32), ARM
6676       Cortex-A35 (cortex-a35) and ARM Cortex-R8 (cortex-r8). The GCC
6677       identifiers can be used as arguments to the -mcpu or -mtune
6678       options, for example: -mcpu=cortex-a32 or -mtune=cortex-a35.
6679
6680  Heterogeneous Systems Architecture
6681
6682     * GCC can now generate HSAIL (Heterogeneous System Architecture
6683       Intermediate Language) for simple OpenMP device constructs if
6684       configured with --enable-offload-targets=hsa. A new libgomp plugin
6685       then runs the HSA GPU kernels implementing these constructs on HSA
6686       capable GPUs via a standard HSA run time.
6687       If the HSA compilation back end determines it cannot output HSAIL
6688       for a particular input, it gives a warning by default. These
6689       warnings can be suppressed with -Wno-hsa. To give a few examples,
6690       the HSA back end does not implement compilation of code using
6691       function pointers, automatic allocation of variable sized arrays,
6692       functions with variadic arguments as well as a number of other less
6693       common programming constructs.
6694       When compilation for HSA is enabled, the compiler attempts to
6695       compile composite OpenMP constructs
6696
6697#pragma omp target teams distribute parallel for
6698
6699       into parallel HSA GPU kernels.
6700
6701  IA-32/x86-64
6702
6703     * GCC now supports the Intel CPU named Skylake with AVX-512
6704       extensions through -march=skylake-avx512. The switch enables the
6705       following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW,
6706       AVX-512DQ.
6707     * Support for new AMD instructions monitorx and mwaitx has been
6708       added. This includes new intrinsic and built-in support. It is
6709       enabled through option -mmwaitx. The instructions monitorx and
6710       mwaitx implement the same functionality as the old monitor and
6711       mwait instructions. In addition, mwaitx adds a configurable timer.
6712       The timer value is received as third argument and stored in
6713       register %ebx.
6714     * x86-64 targets now allow stack realignment from a word-aligned
6715       stack pointer using the command-line option -mstackrealign or
6716       __attribute__ ((force_align_arg_pointer)). This allows functions
6717       compiled with a vector-aligned stack to be invoked from objects
6718       that keep only word-alignment.
6719     * Support for address spaces __seg_fs, __seg_gs, and __seg_tls. These
6720       can be used to access data via the %fs and %gs segments without
6721       having to resort to inline assembly. Please refer to the
6722       [19]documentation for usage instructions.
6723     * Support for AMD Zen (family 17h) processors is now available
6724       through the -march=znver1 and -mtune=znver1 options.
6725
6726  MeP
6727
6728     * Support for the MeP (mep-elf) architecture has been deprecated and
6729       will be removed in a future GCC release.
6730
6731  MSP430
6732
6733     * The MSP430 compiler now has the ability to automatically distribute
6734       code and data between low memory (addresses below 64K) and high
6735       memory. This only applies to parts that actually have both memory
6736       regions and only if the linker script for the part has been
6737       specifically set up to support this feature.
6738       A new attribute of either can be applied to both functions and
6739       data, and this tells the compiler to place the object into low
6740       memory if there is room and into high memory otherwise. Two other
6741       new attributes - lower and upper - can be used to explicitly state
6742       that an object should be placed in the specified memory region. If
6743       there is not enough left in that region the compilation will fail.
6744       Two new command-line options - -mcode-region=[lower|upper|either]
6745       and -mdata-region=[lower|upper|either] - can be used to tell the
6746       compiler what to do with objects that do not have one of these new
6747       attributes.
6748
6749  PowerPC / PowerPC64 / RS6000
6750
6751     * PowerPC64 now supports IEEE 128-bit floating-point using the
6752       __float128 data type. In GCC 6, this is not enabled by default, but
6753       you can enable it with -mfloat128. The IEEE 128-bit floating-point
6754       support requires the use of the VSX instruction set. IEEE 128-bit
6755       floating-point values are passed and returned as a single vector
6756       value. The software emulator for IEEE 128-bit floating-point
6757       support is only built on PowerPC GNU/Linux systems where the
6758       default CPU is at least power7. On future ISA 3.0 systems (POWER 9
6759       and later), you will be able to use the -mfloat128-hardware option
6760       to use the ISA 3.0 instructions that support IEEE 128-bit
6761       floating-point. An additional type (__ibm128) has been added to
6762       refer to the IBM extended double type that normally implements long
6763       double. This will allow for a future transition to implementing
6764       long double with IEEE 128-bit floating-point.
6765     * Basic support has been added for POWER9 hardware that will use the
6766       recently published OpenPOWER ISA 3.0 instructions. The following
6767       new switches are available:
6768          + -mcpu=power9: Implement all of the ISA 3.0 instructions
6769            supported by the compiler.
6770          + -mtune=power9: In the future, apply tuning for POWER9 systems.
6771            Currently, POWER8 tunings are used.
6772          + -mmodulo: Generate code using the ISA 3.0 integer instructions
6773            (modulus, count trailing zeros, array index support, integer
6774            multiply/add).
6775          + -mpower9-fusion: Generate code to suitably fuse instruction
6776            sequences for a POWER9 system.
6777          + -mpower9-dform: Generate code to use the new D-form
6778            (register+offset) memory instructions for the vector
6779            registers.
6780          + -mpower9-vector: Generate code using the new ISA 3.0 vector
6781            (VSX or Altivec) instructions.
6782          + -mpower9-minmax: Reserved for future development.
6783          + -mtoc-fusion: Keep TOC entries together to provide more fusion
6784            opportunities.
6785     * New constraints have been added to support IEEE 128-bit
6786       floating-point and ISA 3.0 instructions:
6787          + wb: Altivec register if -mpower9-dform is enabled.
6788          + we: VSX register if -mpower9-vector is enabled for 64-bit code
6789            generation.
6790          + wo: VSX register if -mpower9-vector is enabled.
6791          + wp: Reserved for future use if long double is implemented with
6792            IEEE 128-bit floating-point instead of IBM extended double.
6793          + wq: VSX register if -mfloat128 is enabled.
6794          + wF: Memory operand suitable for POWER9 fusion load/store.
6795          + wG: Memory operand suitable for TOC fusion memory references.
6796          + wL: Integer constant identifying the element number mfvsrld
6797            accesses within a vector.
6798     * Support has been added for __builtin_cpu_is() and
6799       __builtin_cpu_supports(), allowing for very fast access to
6800       AT_PLATFORM, AT_HWCAP, and AT_HWCAP2 values. This requires use of
6801       glibc 2.23 or later.
6802     * All hardware transactional memory builtins now correctly behave as
6803       memory barriers. Programmers can use #ifdef __TM_FENCE__ to
6804       determine whether their "old" compiler treats the builtins as
6805       barriers.
6806     * Split-stack support has been added for gccgo on PowerPC64 for both
6807       big- and little-endian (but not for 32-bit). The gold linker from
6808       at least binutils 2.25.1 must be available in the PATH when
6809       configuring and building gccgo to enable split stack. (The
6810       requirement for binutils 2.25.1 applies to PowerPC64 only.) The
6811       split-stack feature allows a small initial stack size to be
6812       allocated for each goroutine, which increases as needed.
6813     * GCC on PowerPC now supports the standard lround function.
6814     * A new configuration option ---with-advance-toolchain=at was added
6815       for PowerPC 64-bit GNU/Linux systems to use the header files,
6816       library files, and the dynamic linker from a specific Advance
6817       Toolchain release instead of the default versions that are provided
6818       by the GNU/Linux distribution. In general, this option is intended
6819       for the developers of GCC, and it is not intended for general use.
6820     * The "q", "S", "T", and "t" asm-constraints have been removed.
6821     * The "b", "B", "m", "M", and "W" format modifiers have been removed.
6822
6823  S/390, System z, IBM z Systems
6824
6825     * Support for the IBM z13 processor has been added. When using the
6826       -march=z13 option, the compiler will generate code making use of
6827       the new instructions and registers introduced with the vector
6828       extension facility. The -mtune=z13 option enables z13 specific
6829       instruction scheduling without making use of new instructions.
6830       Compiling code with -march=z13 reduces the default alignment of
6831       vector types bigger than 8 bytes to 8. This is an ABI change and
6832       care must be taken when linking modules compiled with different
6833       arch levels which interchange variables containing vector type
6834       values. For newly compiled code the GNU linker will emit a warning.
6835     * The -mzvector option enables a C/C++ language extension. This
6836       extension provides a new keyword vector which can be used to define
6837       vector type variables. (Note: This is not available when enforcing
6838       strict standard compliance e.g. with -std=c99. Either enable GNU
6839       extensions with e.g. -std=gnu99 or use __vector instead of vector.)
6840       Additionally a set of overloaded builtins is provided which is
6841       partially compatible to the PowerPC Altivec builtins. In order to
6842       make use of these builtins the vecintrin.h header file needs to be
6843       included.
6844     * The new command-line options -march=native, and -mtune=native are
6845       now available on native IBM z Systems. Specifying these options
6846       causes GCC to auto-detect the host CPU and choose the optimal
6847       setting for that system. If GCC is unable to detect the host CPU
6848       these options have no effect.
6849     * The IBM z Systems port now supports target attributes and pragmas.
6850       Please refer to the [20]documentation for details of available
6851       attributes and pragmas as well as usage instructions.
6852     * -fsplit-stack is now supported as part of the IBM z Systems port.
6853       This feature requires a recent gold linker to be used.
6854     * Support for the g5 and g6 -march=/-mtune= CPU level switches has
6855       been deprecated and will be removed in a future GCC release. -m31
6856       from now on defaults to -march=z900 if not specified otherwise.
6857       -march=native on a g5/g6 machine will default to -march=z900.
6858
6859  SH
6860
6861     * Support for SH5 / SH64 has been declared obsolete and will be
6862       removed in future releases.
6863     * Support for the FDPIC ABI has been added. It can be enabled using
6864       the new -mfdpic target option and --enable-fdpic configure option.
6865
6866  SPARC
6867
6868     * An ABI bug has been fixed in 64-bit mode. Unfortunately, this
6869       change will break binary compatibility with earlier releases for
6870       code it affects, but this should be pretty rare in practice. The
6871       conditions are: a 16-byte structure containing a double or a 8-byte
6872       vector in the second half is passed to a subprogram in slot #15,
6873       for example as 16th parameter if the first 15 ones have at most 8
6874       bytes. The double or vector was wrongly passed in floating-point
6875       register %d32 in lieu of on the stack as per the SPARC calling
6876       conventions.
6877
6878Operating Systems
6879
6880  AIX
6881
6882     * DWARF debugging support for AIX 7.1 has been enabled as an optional
6883       debugging format. A more recent Technology Level (TL) and GCC built
6884       with that level are required for full exploitation of DWARF
6885       debugging capabilities.
6886
6887  Linux
6888
6889     * Support for the [21]musl C library was added for the AArch64, ARM,
6890       MicroBlaze, MIPS, MIPS64, PowerPC, PowerPC64, SH, i386, x32 and
6891       x86_64 targets. It can be selected using the new -mmusl option in
6892       case musl is not the default libc. GCC defaults to musl libc if it
6893       is built with a target triplet matching the *-linux-musl* pattern.
6894
6895  RTEMS
6896
6897     * The RTEMS thread model implementation changed. Mutexes now use
6898       self-contained objects defined in newlib <sys/lock.h> instead of
6899       Classic API semaphores. The keys for thread specific data and the
6900       once function are directly defined via <pthread.h>. Self-contained
6901       condition variables are provided via newlib <sys/lock.h>. The RTEMS
6902       thread model also supports C++11 threads.
6903     * OpenMP support now uses self-contained objects provided by newlib
6904       <sys/lock.h> and offers a significantly better performance compared
6905       to the POSIX configuration of libgomp. It is possible to configure
6906       thread pools for each scheduler instance via the environment
6907       variable GOMP_RTEMS_THREAD_POOLS.
6908
6909  Solaris
6910
6911     * Solaris 12 is now fully supported. Minimal support had already been
6912       present in GCC 5.3.
6913     * Solaris 12 provides a full set of startup files (crt1.o, crti.o,
6914       crtn.o), which GCC now prefers over its own ones.
6915     * Position independent executables (PIE) are now supported on Solaris
6916       12.
6917     * Constructor priority is now supported on Solaris 12 with the system
6918       linker.
6919     * libvtv has been ported to Solaris 11 and up.
6920
6921  Windows
6922
6923     * The option -mstackrealign is now automatically activated in 32-bit
6924       mode whenever the use of SSE instructions is requested.
6925
6926Other significant improvements
6927
6928     * The gcc and g++ driver programs will now provide suggestions for
6929       misspelled command-line options.
6930
6931$ gcc -static-libfortran test.f95
6932gcc: error: unrecognized command line option '-static-libfortran'; did you mean
6933'-static-libgfortran'?
6934
6935     * The --enable-default-pie configure option enables generation of PIE
6936       by default.
6937
6938                                    GCC 6.2
6939
6940   This is the [22]list of problem reports (PRs) from GCC's bug tracking
6941   system that are known to be fixed in the 6.2 release. This list might
6942   not be complete (that is, it is possible that some PRs that have been
6943   fixed are not listed here).
6944
6945Target Specific Changes
6946
6947  SPARC
6948
6949     * Support for --with-cpu-32 and --with-cpu-64 configure options has
6950       been added on bi-architecture platforms.
6951     * Support for the SPARC M7 (Niagara 7) processor has been added.
6952     * Support for the VIS 4.0 instruction set has been added.
6953
6954                                    GCC 6.3
6955
6956   This is the [23]list of problem reports (PRs) from GCC's bug tracking
6957   system that are known to be fixed in the 6.3 release. This list might
6958   not be complete (that is, it is possible that some PRs that have been
6959   fixed are not listed here).
6960
6961Target Specific Changes
6962
6963  IA-32/x86-64
6964
6965     * Support for the [24]deprecated pcommit instruction has been
6966       removed.
6967
6968                                    GCC 6.4
6969
6970   This is the [25]list of problem reports (PRs) from GCC's bug tracking
6971   system that are known to be fixed in the 6.4 release. This list might
6972   not be complete (that is, it is possible that some PRs that have been
6973   fixed are not listed here).
6974
6975Operating Systems
6976
6977  RTEMS
6978
6979     * The ABI changes on ARM so that no short enums are used by default.
6980
6981                                    GCC 6.5
6982
6983   This is the [26]list of problem reports (PRs) from GCC's bug tracking
6984   system that are known to be fixed in the 6.5 release. This list might
6985   not be complete (that is, it is possible that some PRs that have been
6986   fixed are not listed here).
6987
6988
6989    For questions related to the use of GCC, please consult these web
6990    pages and the [27]GCC manuals. If that fails, the
6991    [28]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6992    web pages and the development of GCC are welcome on our developer
6993    list at [29]gcc@gcc.gnu.org. All of [30]our lists have public
6994    archives.
6995
6996   Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and
6997   distribution of this entire article is permitted in any medium,
6998   provided this notice is preserved.
6999
7000   These pages are [32]maintained by the GCC team. Last modified
7001   2023-02-20.
7002
7003References
7004
7005   1. http://gcc.gnu.org/gcc-6/porting_to.html
7006   2. http://gcc.gnu.org/onlinedocs/index.html#current
7007   3. https://gcc.gnu.org/ml/gcc/2015-08/msg00101.html
7008   4. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71151
7009   5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87822
7010   6. https://www.openacc.org/
7011   7. https://gcc.gnu.org/wiki/OpenACC
7012   8. https://gcc.gnu.org/wiki/Offloading
7013   9. https://www.openmp.org/specifications/
7014  10. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2014-1266
7015  11. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4377.pdf
7016  12. https://gcc.gnu.org/projects/cxx-status.html#cxx1z
7017  13. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/performance.html
7018  14. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/functions.html#gcc_jit_block_end_with_switch
7019  15. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_set_bool_allow_unreachable_blocks
7020  16. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_add_command_line_option
7021  17. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/AArch64-Function-Attributes.html#AArch64-Function-Attributes
7022  18. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/ARM-Function-Attributes.html#ARM-Function-Attributes
7023  19. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/Named-Address-Spaces.html#Named-Address-Spaces
7024  20. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/S_002f390-Function-Attributes.html#S_002f390-Function-Attributes
7025  21. http://www.musl-libc.org/
7026  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.2
7027  23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.3
7028  24. https://www.intel.com/content/www/us/en/developer/articles/technical/deprecate-pcommit-instruction.html
7029  25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.4
7030  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.5
7031  27. https://gcc.gnu.org/onlinedocs/
7032  28. mailto:gcc-help@gcc.gnu.org
7033  29. mailto:gcc@gcc.gnu.org
7034  30. https://gcc.gnu.org/lists.html
7035  31. https://www.fsf.org/
7036  32. https://gcc.gnu.org/about.html
7037======================================================================
7038http://gcc.gnu.org/gcc-5/index.html
7039
7040                              GCC 5 Release Series
7041
7042   (This release series is no longer supported.)
7043
7044   October 10, 2017
7045
7046   The [1]GNU project and the GCC developers are pleased to announce the
7047   release of GCC 5.5.
7048
7049   This release is a bug-fix release, containing fixes for regressions in
7050   GCC 5.4 relative to previous releases of GCC.
7051
7052Release History
7053
7054   GCC 5.5
7055          October 10, 2017 ([2]changes, [3]documentation)
7056
7057   GCC 5.4
7058          June 3, 2016 ([4]changes, [5]documentation)
7059
7060   GCC 5.3
7061          December 4, 2015 ([6]changes, [7]documentation)
7062
7063   GCC 5.2
7064          July 16, 2015 ([8]changes, [9]documentation)
7065
7066   GCC 5.1
7067          April 22, 2015 ([10]changes, [11]documentation)
7068
7069References and Acknowledgements
7070
7071   GCC used to stand for the GNU C Compiler, but since the compiler
7072   supports several other languages aside from C, it now stands for the
7073   GNU Compiler Collection.
7074
7075   A list of [12]successful builds is updated as new information becomes
7076   available.
7077
7078   The GCC developers would like to thank the numerous people that have
7079   contributed new features, improvements, bug fixes, and other changes as
7080   well as test results to GCC. This [13]amazing group of volunteers is
7081   what makes GCC successful.
7082
7083   For additional information about GCC please refer to the [14]GCC
7084   project web site or contact the [15]GCC development mailing list.
7085
7086   To obtain GCC please use [16]our mirror sites or [17]our version
7087   control system.
7088
7089
7090    For questions related to the use of GCC, please consult these web
7091    pages and the [18]GCC manuals. If that fails, the
7092    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
7093    web pages and the development of GCC are welcome on our developer
7094    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
7095    archives.
7096
7097   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
7098   distribution of this entire article is permitted in any medium,
7099   provided this notice is preserved.
7100
7101   These pages are [23]maintained by the GCC team. Last modified
7102   2022-10-26.
7103
7104References
7105
7106   1. http://www.gnu.org/
7107   2. http://gcc.gnu.org/gcc-5/changes.html
7108   3. http://gcc.gnu.org/onlinedocs/5.5.0/
7109   4. http://gcc.gnu.org/gcc-5/changes.html
7110   5. http://gcc.gnu.org/onlinedocs/5.4.0/
7111   6. http://gcc.gnu.org/gcc-5/changes.html
7112   7. http://gcc.gnu.org/onlinedocs/5.3.0/
7113   8. http://gcc.gnu.org/gcc-5/changes.html
7114   9. http://gcc.gnu.org/onlinedocs/5.2.0/
7115  10. http://gcc.gnu.org/gcc-5/changes.html
7116  11. http://gcc.gnu.org/onlinedocs/5.1.0/
7117  12. http://gcc.gnu.org/gcc-5/buildstat.html
7118  13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
7119  14. http://gcc.gnu.org/index.html
7120  15. mailto:gcc@gcc.gnu.org
7121  16. http://gcc.gnu.org/mirrors.html
7122  17. http://gcc.gnu.org/git.html
7123  18. https://gcc.gnu.org/onlinedocs/
7124  19. mailto:gcc-help@gcc.gnu.org
7125  20. mailto:gcc@gcc.gnu.org
7126  21. https://gcc.gnu.org/lists.html
7127  22. https://www.fsf.org/
7128  23. https://gcc.gnu.org/about.html
7129======================================================================
7130http://gcc.gnu.org/gcc-5/changes.html
7131
7132                              GCC 5 Release Series
7133                        Changes, New Features, and Fixes
7134
7135Caveats
7136
7137     * The default mode for C is now -std=gnu11 instead of -std=gnu89.
7138     * The C++ runtime library (libstdc++) uses a new ABI by default (see
7139       [1]below).
7140     * The Graphite framework for loop optimizations no longer requires
7141       the CLooG library, only ISL version 0.14 (recommended) or 0.12.2.
7142       The installation manual contains more information about
7143       requirements to build GCC.
7144     * The non-standard C++0x type traits has_trivial_default_constructor,
7145       has_trivial_copy_constructor and has_trivial_copy_assign have been
7146       deprecated and will be removed in a future version. The standard
7147       C++11 traits is_trivially_default_constructible,
7148       is_trivially_copy_constructible and is_trivially_copy_assignable
7149       should be used instead.
7150     * On AVR, support has been added for the devices
7151       ATtiny4/5/9/10/20/40. This requires Binutils 2.25 or newer.
7152     * The AVR port uses a new scheme to describe supported devices: For
7153       each supported device the compiler provides a device-specific
7154       [2]spec file. If the compiler is used together with AVR-LibC, this
7155       requires at least GCC 5.2 and a version of AVR-LibC which
7156       implements [3]feature #44574.
7157
7158General Optimizer Improvements
7159
7160     * Inter-procedural optimization improvements:
7161          + An Identical Code Folding (ICF) pass (controlled via
7162            -fipa-icf) has been added. Compared to the identical code
7163            folding performed by the Gold linker this pass does not
7164            require function sections. It also performs merging before
7165            inlining, so inter-procedural optimizations are aware of the
7166            code re-use. On the other hand not all unifications performed
7167            by a linker are doable by GCC which must honor aliasing
7168            information. During link-time optimization of Firefox, this
7169            pass unifies about 31000 functions, that is 14% overall.
7170          + The devirtualization pass was significantly improved by adding
7171            better support for speculative devirtualization and dynamic
7172            type detection. About 50% of virtual calls in Firefox are now
7173            speculatively devirtualized during link-time optimization.
7174          + A new comdat localization pass allows the linker to eliminate
7175            more dead code in presence of C++ inline functions.
7176          + Virtual tables are now optimized. Local aliases are used to
7177            reduce dynamic linking time of C++ virtual tables on ELF
7178            targets and data alignment has been reduced to limit data
7179            segment bloat.
7180          + A new -fno-semantic-interposition option can be used to
7181            improve code quality of shared libraries where interposition
7182            of exported symbols is not allowed.
7183          + Write-only variables are now detected and optimized out.
7184          + With profile feedback the function inliner can now bypass
7185            --param inline-insns-auto and --param inline-insns-single
7186            limits for hot calls.
7187          + The IPA reference pass was significantly sped up making it
7188            feasible to enable -fipa-reference with -fprofile-generate.
7189            This also solves a bottleneck seen when building Chromium with
7190            link-time optimization.
7191          + The symbol table and call-graph API was reworked to C++ and
7192            simplified.
7193          + The interprocedural propagation of constants now also
7194            propagates alignments of pointer parameters. This for example
7195            means that the vectorizer often does not need to generate loop
7196            prologues and epilogues to make up for potential
7197            misalignments.
7198     * Link-time optimization improvements:
7199          + One Definition Rule based merging of C++ types has been
7200            implemented. Type merging enables better devirtualization and
7201            alias analysis. Streaming extra information needed to merge
7202            types adds about 2-6% of memory size and object size increase.
7203            This can be controlled by -flto-odr-type-merging.
7204          + Command-line optimization and target options are now streamed
7205            on a per-function basis and honored by the link-time
7206            optimizer. This change makes link-time optimization a more
7207            transparent replacement of per-file optimizations. It is now
7208            possible to build projects that require different optimization
7209            settings for different translation units (such as -ffast-math,
7210            -mavx, or -finline). Contrary to earlier GCC releases, the
7211            optimization and target options passed on the link command
7212            line are ignored.
7213            Note that this applies only to those command-line options that
7214            can be passed to optimize and target attributes. Command-line
7215            options affecting global code generation (such as -fpic),
7216            warnings (such as -Wodr), optimizations affecting the way
7217            static variables are optimized (such as -fcommon), debug
7218            output (such as -g), and --param parameters can be applied
7219            only to the whole link-time optimization unit. In these cases,
7220            it is recommended to consistently use the same options at both
7221            compile time and link time.
7222          + GCC bootstrap now uses slim LTO object files.
7223          + Memory usage and link times were improved. Tree merging was
7224            sped up, memory usage of GIMPLE declarations and types was
7225            reduced, and, support for on-demand streaming of variable
7226            constructors was added.
7227     * Feedback directed optimization improvements:
7228          + A new auto-FDO mode uses profiles collected by low overhead
7229            profiling tools (perf) instead of more expensive program
7230            instrumentation (via -fprofile-generate). SPEC2006 benchmarks
7231            on x86-64 improve by 4.7% with auto-FDO and by 7.3% with
7232            traditional feedback directed optimization.
7233          + Profile precision was improved in presence of C++ inline and
7234            extern inline functions.
7235          + The new gcov-tool utility allows manipulating profiles.
7236          + Profiles are now more tolerant to source file changes (this
7237            can be controlled by --param profile-func-internal-id).
7238     * Register allocation improvements:
7239          + A new local register allocator (LRA) sub-pass, controlled by
7240            -flra-remat, implements control-flow sensitive global register
7241            rematerialization. Instead of spilling and restoring a
7242            register value, it is recalculated if it is profitable. The
7243            sub-pass improved SPEC2000 generated code by 1% and 0.5%
7244            correspondingly on ARM and x86-64.
7245          + Reuse of the PIC hard register, instead of using a fixed
7246            register, was implemented on x86/x86-64 targets. This improves
7247            generated PIC code performance as more hard registers can be
7248            used. Shared libraries can significantly benefit from this
7249            optimization. Currently it is switched on only for x86/x86-64
7250            targets. As RA infrastructure is already implemented for PIC
7251            register reuse, other targets might follow this in the future.
7252          + A simple form of inter-procedural RA was implemented. When it
7253            is known that a called function does not use caller-saved
7254            registers, save/restore code is not generated around the call
7255            for such registers. This optimization can be controlled by
7256            -fipa-ra
7257          + LRA is now much more effective at generating spills of general
7258            registers into vector registers instead of memory on
7259            architectures (e.g., modern Intel processors) where this is
7260            profitable.
7261     * UndefinedBehaviorSanitizer gained a few new sanitization options:
7262          + -fsanitize=float-divide-by-zero: detect floating-point
7263            division by zero;
7264          + -fsanitize=float-cast-overflow: check that the result of
7265            floating-point type to integer conversions do not overflow;
7266          + -fsanitize=bounds: enable instrumentation of array bounds and
7267            detect out-of-bounds accesses;
7268          + -fsanitize=alignment: enable alignment checking, detect
7269            various misaligned objects;
7270          + -fsanitize=object-size: enable object size checking, detect
7271            various out-of-bounds accesses.
7272          + -fsanitize=vptr: enable checking of C++ member function calls,
7273            member accesses and some conversions between pointers to base
7274            and derived classes, detect if the referenced object does not
7275            have the correct dynamic type.
7276     * Pointer Bounds Checker, a bounds violation detector, has been added
7277       and can be enabled via -fcheck-pointer-bounds. Memory accesses are
7278       instrumented with run-time checks of used pointers against their
7279       bounds to detect pointer bounds violations (overflows). The Pointer
7280       Bounds Checker is available on x86/x86-64 GNU/Linux targets with a
7281       new ISA extension Intel MPX support. See the Pointer Bounds Checker
7282       [4]Wiki page for more details.
7283
7284New Languages and Language specific improvements
7285
7286     * [5]OpenMP 4.0 specification offloading features are now supported
7287       by the C, C++, and Fortran compilers. Generic changes:
7288          + Infrastructure (suitable for any vendor).
7289          + Testsuite which covers offloading from the [6]OpenMP 4.0
7290            Examples document.
7291       Specific for upcoming Intel Xeon Phi products:
7292          + Run-time library.
7293          + Card emulator.
7294     * GCC 5 includes a preliminary implementation of the OpenACC 2.0a
7295       specification. OpenACC is intended for programming accelerator
7296       devices such as GPUs. See [7]the OpenACC wiki page for more
7297       information.
7298
7299  C family
7300
7301     * The default setting of the -fdiagnostics-color= command-line option
7302       is now [8]configurable when building GCC using configuration option
7303       --with-diagnostics-color=. The possible values are: never, always,
7304       auto and auto-if-env. The new default auto uses color only when the
7305       standard error is a terminal. The default in GCC 4.9 was
7306       auto-if-env, which is equivalent to auto if there is a non-empty
7307       GCC_COLORS environment variable, and never otherwise. As in GCC
7308       4.9, an empty GCC_COLORS variable in the environment will always
7309       disable colors, no matter what the default is or what command-line
7310       options are used.
7311     * A new command-line option -Wswitch-bool has been added for the C
7312       and C++ compilers, which warns whenever a switch statement has an
7313       index of boolean type.
7314     * A new command-line option -Wlogical-not-parentheses has been added
7315       for the C and C++ compilers, which warns about "logical not" used
7316       on the left hand side operand of a comparison.
7317     * A new command-line option -Wsizeof-array-argument has been added
7318       for the C and C++ compilers, which warns when the sizeof operator
7319       is applied to a parameter that has been declared as an array in a
7320       function definition.
7321     * A new command-line option -Wbool-compare has been added for the C
7322       and C++ compilers, which warns about boolean expressions compared
7323       with an integer value different from true/false.
7324     * Full support for Cilk Plus has been added to the GCC compiler. Cilk
7325       Plus is an extension to the C and C++ languages to support data and
7326       task parallelism.
7327     * A new attribute no_reorder prevents reordering of selected symbols
7328       against other such symbols or inline assembler. This enables to
7329       link-time optimize the Linux kernel without having to resort to
7330       -fno-toplevel-reorder that disables several optimizations.
7331     * New preprocessor constructs, __has_include and __has_include_next,
7332       to test the availability of headers have been added.
7333       This demonstrates a way to include the header <optional> only if it
7334       is available:
7335
7336#ifdef __has_include
7337#  if __has_include(<optional>)
7338#    include <optional>
7339#    define have_optional 1
7340#  elif __has_include(<experimental/optional>)
7341#    include <experimental/optional>
7342#    define have_optional 1
7343#    define experimental_optional
7344#  else
7345#    define have_optional 0
7346#  endif
7347#endif
7348
7349       The header search paths for __has_include and __has_include_next
7350       are equivalent to those of the standard directive #include and the
7351       extension #include_next respectively.
7352     * A new built-in function-like macro to determine the existence of an
7353       attribute, __has_attribute, has been added. The equivalent built-in
7354       macro __has_cpp_attribute was added to C++ to support
7355       [9]Feature-testing recommendations for C++. The macro
7356       __has_attribute is added to all C-like languages as an extension:
7357
7358int
7359#ifdef __has_attribute
7360#  if __has_attribute(__noinline__)
7361  __attribute__((__noinline__))
7362#  endif
7363#endif
7364foo(int x);
7365
7366       If an attribute exists, a nonzero constant integer is returned. For
7367       standardized C++ attributes a date is returned, otherwise the
7368       constant returned is 1. Both __has_attribute and
7369       __has_cpp_attribute will add underscores to an attribute name if
7370       necessary to resolve the name. For C++11 and onwards the attribute
7371       may be scoped.
7372     * A new set of built-in functions for arithmetics with overflow
7373       checking has been added: __builtin_add_overflow,
7374       __builtin_sub_overflow and __builtin_mul_overflow and for
7375       compatibility with clang also other variants. These builtins have
7376       two integral arguments (which don't need to have the same type),
7377       the arguments are extended to infinite precision signed type, +, -
7378       or * is performed on those, and the result is stored in an integer
7379       variable pointed to by the last argument. If the stored value is
7380       equal to the infinite precision result, the built-in functions
7381       return false, otherwise true. The type of the integer variable that
7382       will hold the result can be different from the types of the first
7383       two arguments. The following snippet demonstrates how this can be
7384       used in computing the size for the calloc function:
7385
7386void *
7387calloc (size_t x, size_t y)
7388{
7389  size_t sz;
7390  if (__builtin_mul_overflow (x, y, &sz))
7391    return NULL;
7392  void *ret = malloc (sz);
7393  if (ret) memset (res, 0, sz);
7394  return ret;
7395}
7396
7397       On e.g. i?86 or x86-64 the above will result in a mul instruction
7398       followed by a jump on overflow.
7399     * The option -fextended-identifiers is now enabled by default for
7400       C++, and for C99 and later C versions. Various bugs in the
7401       implementation of extended identifiers have been fixed.
7402
7403  C
7404
7405     * The default mode has been changed to -std=gnu11.
7406     * A new command-line option -Wc90-c99-compat has been added to warn
7407       about features not present in ISO C90, but present in ISO C99.
7408     * A new command-line option -Wc99-c11-compat has been added to warn
7409       about features not present in ISO C99, but present in ISO C11.
7410     * It is possible to disable warnings about conversions between
7411       pointers that have incompatible types via a new warning option
7412       -Wno-incompatible-pointer-types; warnings about implicit
7413       incompatible integer to pointer and pointer to integer conversions
7414       via a new warning option -Wno-int-conversion; and warnings about
7415       qualifiers on pointers being discarded via a new warning option
7416       -Wno-discarded-qualifiers.
7417     * To allow proper use of const qualifiers with multidimensional
7418       arrays, GCC will not warn about incompatible pointer types anymore
7419       for conversions between pointers to arrays with and without const
7420       qualifier (except when using -pedantic). Instead, a new warning is
7421       emitted only if the const qualifier is lost. This can be controlled
7422       with a new warning option -Wno-discarded-array-qualifiers.
7423     * The C front end now generates more precise caret diagnostics.
7424     * The -pg command-line option now only affects the current file in an
7425       LTO build.
7426
7427  C++
7428
7429     * G++ now supports [10]C++14 variable templates.
7430     * -Wnon-virtual-dtor doesn't warn anymore for final classes.
7431     * Excessive template instantiation depth is now a fatal error. This
7432       prevents excessive diagnostics that usually do not help to identify
7433       the problem.
7434     * G++ and libstdc++ now implement the feature-testing macros from
7435       [11]Feature-testing recommendations for C++.
7436     * G++ now allows typename in a template template parameter.
7437
7438template<template<typename> typename X> struct D; // OK
7439
7440     * G++ now supports [12]C++14 aggregates with non-static data member
7441       initializers.
7442
7443struct A { int i, j = i; };
7444A a = { 42 }; // a.j is also 42
7445
7446     * G++ now supports [13]C++14 extended constexpr.
7447
7448constexpr int f (int i)
7449{
7450  int j = 0;
7451  for (; i > 0; --i)
7452    ++j;
7453  return j;
7454}
7455
7456constexpr int i = f(42); // i is 42
7457
7458     * G++ now supports the [14]C++14 sized deallocation functions.
7459
7460void operator delete (void *, std::size_t) noexcept;
7461void operator delete[] (void *, std::size_t) noexcept;
7462
7463     * A new One Definition Rule violation warning (controlled by -Wodr)
7464       detects mismatches in type definitions and virtual table contents
7465       during link-time optimization.
7466     * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help
7467       developers to annotate programs with final specifiers (or anonymous
7468       namespaces) to improve code generation. These warnings can be used
7469       at compile time, but they are more useful in combination with
7470       link-time optimization.
7471     * G++ no longer supports [15]N3639 variable length arrays, as they
7472       were removed from the C++14 working paper prior to ratification.
7473       GNU VLAs are still supported, so VLA support is now the same in
7474       C++14 mode as in C++98 and C++11 modes.
7475     * G++ now allows passing a non-trivially-copyable class via C
7476       varargs, which is conditionally-supported with
7477       implementation-defined semantics in the standard. This uses the
7478       same calling convention as a normal value parameter.
7479     * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So
7480       various mangling bugs are fixed, but G++ will still emit aliases
7481       with the old, wrong mangling where feasible. -Wabi=2 will warn
7482       about differences between ABI version 2 and the current setting.
7483     * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely
7484       to be unaffected, but -Wabi=8 will warn about a non-static data
7485       member with type std::nullptr_t which changes position due to this
7486       change.
7487
7488    Runtime Library (libstdc++)
7489
7490     * A [16]Dual ABI is provided by the library. A new ABI is enabled by
7491       default. The old ABI is still supported and can be used by defining
7492       the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++
7493       standard library headers.
7494     * A new implementation of std::string is enabled by default, using
7495       the small string optimization instead of copy-on-write reference
7496       counting.
7497     * A new implementation of std::list is enabled by default, with an
7498       O(1) size() function;
7499     * [17]Full support for C++11, including the following new features:
7500          + std::deque and std::vector<bool> meet the allocator-aware
7501            container requirements;
7502          + movable and swappable iostream classes;
7503          + support for std::align and std::aligned_union;
7504          + type traits std::is_trivially_copyable,
7505            std::is_trivially_constructible, std::is_trivially_assignable
7506            etc.;
7507          + I/O manipulators std::put_time, std::get_time, std::hexfloat
7508            and std::defaultfloat;
7509          + generic locale-aware std::isblank;
7510          + locale facets for Unicode conversion;
7511          + atomic operations for std::shared_ptr;
7512          + std::notify_all_at_thread_exit() and functions for making
7513            futures ready at thread exit.
7514     * Support for the C++11 hexfloat manipulator changes how the num_put
7515       facet formats floating point types when
7516       ios_base::fixed|ios_base::scientific is set in a stream's fmtflags.
7517       This change affects all language modes, even though the C++98
7518       standard gave no special meaning to that combination of flags. To
7519       prevent the use of hexadecimal notation for floating point types
7520       use str.unsetf(std::ios_base::floatfield) to clear the relevant
7521       bits in str.flags().
7522     * [18]Full experimental support for C++14, including the following
7523       new features:
7524          + std::is_final type trait;
7525          + heterogeneous comparison lookup in associative containers.
7526          + global functions cbegin, cend, rbegin, rend, crbegin, and
7527            crend for range access to containers, arrays and initializer
7528            lists.
7529     * [19]Improved experimental support for the Library Fundamentals TS,
7530       including:
7531          + class std::experimental::any;
7532          + function template std::experimental::apply;
7533          + function template std::experimental::sample;
7534          + function template std::experimental::search and related
7535            searcher types;
7536          + variable templates for type traits;
7537          + function template std::experimental::not_fn.
7538     * New random number distributions logistic_distribution and
7539       uniform_on_sphere_distribution as extensions.
7540     * [20]GDB Xmethods for containers and std::unique_ptr.
7541
7542  Fortran
7543
7544     * Compatibility notice:
7545          + The version of the module files (.mod) has been incremented.
7546          + For free-form source files [21]-Werror=line-truncation is now
7547            enabled by default. Note that comments exceeding the line
7548            length are not diagnosed. (For fixed-form source code, the
7549            same warning is available but turned off by default, such that
7550            excess characters are ignored. -ffree-line-length-n and
7551            -ffixed-line-length-n can be used to modify the default line
7552            lengths of 132 and 72 columns, respectively.)
7553          + The -Wtabs option is now more sensible: with -Wtabs the
7554            compiler warns if it encounters tabs and with -Wno-tabs this
7555            warning is turned off. Before, -Wno-tabs warned and -Wtabs
7556            disabled the warning. As before, this warning is also enabled
7557            by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts
7558            options of -std=.
7559     * Incomplete support for colorizing diagnostics emitted by gfortran
7560       has been added. The option [22]-fdiagnostics-color controls when
7561       color is used in diagnostics. The default value of this option can
7562       be [23]configured when building GCC. The GCC_COLORS environment
7563       variable can be used to customize the colors or disable coloring
7564       completely. Sample diagnostics output:
7565      $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90
7566      test.f90:6:1:
7567
7568       0 continue
7569       1
7570      Error: Zero is not a valid statement label at (1)
7571      test.f90:9:6:
7572
7573         USE foo
7574            1
7575      Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only]
7576
7577     * The -Wuse-without-only option has been added to warn when a USE
7578       statement has no ONLY qualifier and thus implicitly imports all
7579       public entities of the used module.
7580     * Formatted READ and WRITE statements now work correctly in
7581       locale-aware programs. For more information and potential caveats,
7582       see [24]Section 5.3 Thread-safety of the runtime library in the
7583       manual.
7584     * [25]Fortran 2003:
7585          + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and
7586            IEEE_ARITHMETIC) are now supported.
7587     * [26]Fortran 2008:
7588          + [27]Coarrays: Full experimental support of Fortran 2008's
7589            coarrays with -fcoarray=lib except for allocatable/pointer
7590            components of derived-type coarrays. GCC currently only ships
7591            with a single-image library (libcaf_single), but multi-image
7592            support based on MPI and GASNet is provided by the libraries
7593            of the [28]OpenCoarrays project.
7594     * TS18508 Additional Parallel Features in Fortran:
7595          + Support for the collective intrinsic subroutines CO_MAX,
7596            CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added,
7597            including -fcoarray=lib support.
7598          + Support for the new atomic intrinsics has been added,
7599            including -fcoarray=lib support.
7600     * Fortran 2015:
7601          + Support for IMPLICIT NONE (external, type).
7602          + ERROR STOP is now permitted in pure procedures.
7603
7604  Go
7605
7606     * GCC 5 provides a complete implementation of the Go 1.4.2 release.
7607     * Building GCC 5 with Go enabled will install two new programs:
7608       [29]go and [30]gofmt.
7609
7610libgccjit
7611
7612   New in GCC 5 is the ability to build GCC as a shared library for
7613   embedding in other processes (such as interpreters), suitable for
7614   Just-In-Time compilation to machine code.
7615
7616   The shared library has a [31]C API and a [32]C++ wrapper API providing
7617   some "syntactic sugar". There are also bindings available from 3rd
7618   parties for [33]Python and for [34]D.
7619
7620   For example, this library can be used by interpreters for [35]compiling
7621   functions from bytecode to machine code.
7622
7623   The library can also be used for ahead-of-time compilation, enabling
7624   GCC to be plugged into a pre-existing front end. An example of using
7625   this to build a compiler for an esoteric language we'll refer to as
7626   "brainf" can be seen [36]here.
7627
7628   libgccjit is licensed under the GPLv3 (or at your option, any later
7629   version)
7630
7631   It should be regarded as experimental at this time.
7632
7633New Targets and Target Specific Improvements
7634
7635  Reporting stack usage
7636
7637     * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the
7638       -fstack-usage option.
7639
7640  AArch64
7641
7642     * Code generation for the ARM Cortex-A57 processor has been improved.
7643       A more accurate instruction scheduling model for the processor is
7644       now used, and a number of compiler tuning parameters have been set
7645       to offer increased performance when compiling with -mcpu=cortex-a57
7646       or -mtune=cortex-a57.
7647     * A workaround for the ARM Cortex-A53 erratum 835769 has been added
7648       and can be enabled by giving the -mfix-cortex-a53-835769 option.
7649       Alternatively it can be enabled by default by configuring GCC with
7650       the --enable-fix-cortex-a53-835769 option.
7651     * The optional cryptographic extensions to the ARMv8-A architecture
7652       are no longer enabled by default when specifying the
7653       -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53
7654       options. To enable these extensions add +crypto to the value of
7655       -mcpu or -march e.g. -mcpu=cortex-a53+crypto.
7656     * Support has been added for the following processors (GCC
7657       identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and
7658       initial support for its big.LITTLE combination with the ARM
7659       Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx),
7660       Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1).
7661       The GCC identifiers can be used as arguments to the -mcpu or -mtune
7662       options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53.
7663       Using -mcpu=cortex-a72 requires a version of GNU binutils that has
7664       support for the Cortex-A72.
7665     * The transitional options -mlra and -mno-lra have been removed. The
7666       AArch64 backend now uses the local register allocator (LRA) only.
7667
7668  ARM
7669
7670     * Thumb-1 assembly code is now generated in unified syntax. The new
7671       option -masm-syntax-unified specifies whether inline assembly code
7672       is using unified syntax. By default the option is off which means
7673       non-unified syntax is used. However this is subject to change in
7674       future releases. Eventually the non-unified syntax will be
7675       deprecated.
7676     * It is now a configure-time error to use the --with-cpu configure
7677       option with either of --with-tune or --with-arch.
7678     * Code generation for the ARM Cortex-A57 processor has been improved.
7679       A more accurate instruction scheduling model for the processor is
7680       now used, and a number of compiler tuning parameters have been set
7681       to offer increased performance when compiling with -mcpu=cortex-a57
7682       or -mtune=cortex-a57.
7683     * Support has been added for the following processors (GCC
7684       identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and
7685       initial support for its big.LITTLE combination with the ARM
7686       Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and
7687       initial support for its big.LITTLE combination with the ARM
7688       Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7),
7689       Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1).
7690       The GCC identifiers can be used as arguments to the -mcpu or -mtune
7691       options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53.
7692       Using -mcpu=cortex-a72 requires a version of GNU binutils that has
7693       support for the Cortex-A72.
7694     * The deprecated option -mwords-little-endian has been removed.
7695     * The options -mapcs, -mapcs-frame, -mtpcs-frame and
7696       -mtpcs-leaf-frame which are only applicable to the old ABI have
7697       been deprecated.
7698     * The transitional options -mlra and -mno-lra have been removed. The
7699       ARM backend now uses the local register allocator (LRA) only.
7700
7701  AVR
7702
7703     * The compiler no more supports individual devices like ATmega8.
7704       Specifying, say, -mmcu=atmega8 triggers the usage of the
7705       device-specific [37]spec file specs-atmega8 which is part of the
7706       installation and describes options for the sub-processes like
7707       compiler proper, assembler and linker. You can add support for a
7708       new device -mmcu=mydevice as follows:
7709         1. In an empty directory /someplace, create a new directory
7710            device-specs.
7711         2. Copy a device spec file from the installed device-specs
7712            folder, follow the comments in that file and then save it as
7713            /someplace/device-specs/specs-mydevice.
7714         3. Add -B /someplace -mmcu=mydevice to the compiler's
7715            command-line options. Notice that /someplace must specify an
7716            absolute path and that mydevice must not start with "avr".
7717         4. Provided you have a device-specific library libmydevice.a
7718            available, you can put it at /someplace, dito for a
7719            device-specific startup file crtmydevice.o.
7720       The contents of the device spec files depend on the compiler's
7721       configuration, in particular on --with-avrlibc=no and whether or
7722       not it is configured for RTEMS.
7723     * A new command-line option -nodevicelib has been added. It prevents
7724       the compiler from linking against AVR-LibC's device-specific
7725       library libdevice.a.
7726     * The following three command-line options have been added:
7727
7728        -mrmw
7729                Set if the device supports the read-modify-write
7730                instructions LAC, LAS, LAT and XCH.
7731
7732        -mn-flash=size
7733                Specify the flash size of the device in units of 64 KiB,
7734                rounded up to the next integer as needed. This option
7735                affects the availability of the [38]AVR address-spaces.
7736
7737        -mskip-bug
7738                Set if the device is affected by the respective silicon
7739                bug.
7740
7741       In general, you don't need to set these options by hand. The new
7742       device-specific spec file will set them as needed.
7743
7744  IA-32/x86-64
7745
7746     * New ISA extensions support AVX-512{BW,DQ,VL,IFMA,VBMI} of Intel's
7747       CPU codenamed Skylake Server was added to GCC. That includes inline
7748       assembly support, new intrinsics, and basic autovectorization.
7749       These new AVX-512 extensions are available via the following GCC
7750       switches: AVX-512 Vector Length EVEX feature: -mavx512vl, AVX-512
7751       Byte and Word instructions: -mavx512bw, AVX-512 Dword and Qword
7752       instructions: -mavx512dq, AVX-512 FMA-52 instructions: -mavx512ifma
7753       and for AVX-512 Vector Bit Manipulation Instructions: -mavx512vbmi.
7754     * New ISA extensions support Intel MPX was added to GCC. This new
7755       extension is available via the -mmpx compiler switch. Intel MPX is
7756       a set of processor features which, with compiler, run-time library
7757       and OS support, brings increased robustness to software by run-time
7758       checking pointer references against their bounds. In GCC Intel MPX
7759       is supported by Pointer Bounds Checker and libmpx run-time
7760       libraries.
7761     * The new -mrecord-mcount option for -pg generates a Linux kernel
7762       style table of pointers to mcount or __fentry__ calls at the
7763       beginning of functions. The new -mnop-mcount option in addition
7764       also generates nops in place of the __fentry__ or mcount call, so
7765       that a call per function can be later patched in. This can be used
7766       for low overhead tracing or hot code patching.
7767     * The new -malign-data option controls how GCC aligns variables.
7768       -malign-data=compat uses increased alignment compatible with GCC
7769       4.8 and earlier, -malign-data=abi uses alignment as specified by
7770       the psABI, and -malign-data=cacheline uses increased alignment to
7771       match the cache line size. -malign-data=compat is the default.
7772     * The new -mskip-rax-setup option skips setting up the RAX register
7773       when SSE is disabled and there are no variable arguments passed in
7774       vector registers. This can be used to optimize the Linux kernel.
7775
7776  MIPS
7777
7778     * MIPS Releases 3 and 5 are now directly supported. Use the
7779       command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5
7780       to enable code-generation for these processors.
7781     * The Imagination P5600 processor is now supported using the
7782       -march=p5600 command-line option.
7783     * The Cavium Octeon3 processor is now supported using the
7784       -march=octeon3 command-line option.
7785     * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6
7786       command-line options.
7787     * The o32 ABI has been modified and extended. The o32 64-bit
7788       floating-point register support is now obsolete and has been
7789       removed. It has been replaced by three ABI extensions FPXX, FP64A,
7790       and FP64. The meaning of the -mfp64 command-line option has
7791       changed. It is now used to enable the FP64A and FP64 ABI
7792       extensions.
7793          + The FPXX extension requires that code generated to access
7794            double-precision values use even-numbered registers. Code that
7795            adheres to this extension is link-compatible with all other
7796            o32 double-precision ABI variants and will execute correctly
7797            in all hardware FPU modes. The command-line options -mabi=32
7798            -mfpxx can be used to enable this extension. MIPS II is the
7799            minimum processor required.
7800          + The o32 FP64A extension requires that floating-point registers
7801            be 64-bit and odd-numbered single-precision registers are not
7802            allowed. Code that adheres to the o32 FP64A variant is
7803            link-compatible with all other o32 double-precision ABI
7804            variants. The command-line options -mabi=32 -mfp64
7805            -mno-odd-spreg can be used to enable this extension. MIPS32R2
7806            is the minimum processor required.
7807          + The o32 FP64 extension also requires that floating-point
7808            registers be 64-bit, but permits the use of single-precision
7809            registers. Code that adheres to the o32 FP64 variant is
7810            link-compatible with o32 FPXX and o32 FP64A variants only,
7811            i.e. it is not compatible with the original o32
7812            double-precision ABI. The command-line options -mabi=32 -mfp64
7813            -modd-spreg can be used to enable this extension. MIPS32R2 is
7814            the minimum processor required.
7815       The new ABI variants can be enabled by default using the configure
7816       time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32.
7817       It is strongly recommended that all vendors begin to set o32 FPXX
7818       as the default ABI. This will be required to run the generated code
7819       on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and
7820       MIPSR6 cores.
7821     * GCC will now pass all floating-point options to the assembler if
7822       GNU binutils 2.25 is used. As a result, any inline assembly code
7823       that uses hard-float instructions should be amended to include a
7824       .set directive to override the global assembler options when
7825       compiling for soft-float targets.
7826
7827  NDS32
7828
7829     * The variadic function ABI implementation is now compatible with
7830       past Andes toolchains where the caller uses registers to pass
7831       arguments and the callee is in charge of pushing them on stack.
7832     * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have
7833       been removed since they are not yet available in the nds32 port of
7834       GNU binutils.
7835     * A new option -mcmodel=[small|medium|large] supports varied code
7836       models on code generation. The -mgp-direct option became
7837       meaningless and can be discarded.
7838
7839  RX
7840
7841     * A new command line option -mno-allow-string-insns can be used to
7842       disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL,
7843       SWHILE and RMPA instructions. An erratum released by Renesas shows
7844       that it is unsafe to use these instructions on addresses within the
7845       I/O space of the processor. The new option can be used when the
7846       programmer is concerned that the I/O space might be accessed. The
7847       default is still to enable these instructions.
7848
7849  SH
7850
7851     * The compiler will now pass the appropriate --isa= option to the
7852       assembler.
7853     * The default handling for the GBR has been changed from call
7854       clobbered to call preserved. The old behavior can be reinstated by
7855       specifying the option -fcall-used-gbr.
7856     * Support for the SH4A fpchg instruction has been added which will be
7857       utilized when switching between single and double precision FPU
7858       modes.
7859     * The compiler no longer uses the __fpscr_values array for switching
7860       between single and double FPU precision modes on non-SH4A targets.
7861       Instead mode switching will now be performed by storing, modifying
7862       and reloading the FPSCR, so that other FPSCR bits are preserved
7863       across mode switches. The __fpscr_values array that is defined in
7864       libgcc is still present for backwards compatibility, but it will
7865       not be referenced by compiler generated code anymore.
7866     * New builtin functions __builtin_sh_get_fpscr and
7867       __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr
7868       function will mask the specified bits in such a way that the SZ, PR
7869       and FR mode bits will be preserved, while changing the other bits.
7870       These new functions do not reference the __fpscr_values array. The
7871       old functions __set_fpscr and __get_fpscr in libgcc which access
7872       the __fpscr_values array are still present for backwards
7873       compatibility, but their usage is highly discouraged.
7874     * Some improvements to code generated for __atomic built-in
7875       functions.
7876     * When compiling for SH2E the compiler will no longer force the usage
7877       of delay slots for conditional branch instructions bt and bf. The
7878       old behavior can be reinstated (e.g. to work around a hardware bug
7879       in the original SH7055) by specifying the new option
7880       -mcbranch-force-delay-slot.
7881
7882Operating Systems
7883
7884  AIX
7885
7886     * GCC now supports stabs debugging continuation lines to allow long
7887       stabs debug information without overflow that generates AIX linker
7888       errors.
7889
7890  DragonFly BSD
7891
7892     * GCC now supports the DragonFly BSD operating system.
7893
7894  FreeBSD
7895
7896     * GCC now supports the FreeBSD operating system for the arm port
7897       through the arm*-*-freebsd* target triplets.
7898
7899  VxWorks MILS
7900
7901     * GCC now supports the MILS (Multiple Independent Levels of Security)
7902       variant of WindRiver's VxWorks operating system for PowerPC
7903       targets.
7904
7905Other significant improvements
7906
7907     * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option
7908       to set the compiler to use.
7909
7910     * When the new command-line option -freport-bug is used, GCC
7911       automatically generates a developer-friendly reproducer whenever an
7912       internal compiler error is encountered.
7913
7914                                    GCC 5.2
7915
7916   This is the [39]list of problem reports (PRs) from GCC's bug tracking
7917   system that are known to be fixed in the 5.2 release. This list might
7918   not be complete (that is, it is possible that some PRs that have been
7919   fixed are not listed here).
7920
7921Target Specific Changes
7922
7923  IA-32/x86-64
7924
7925     * Support for new AMD instructions monitorx and mwaitx has been
7926       added. This includes new intrinsic and built-in support. It is
7927       enabled through option -mmwaitx. The instructions monitorx and
7928       mwaitx implement the same functionality as the old monitor and
7929       mwait instructions. In addition, mwaitx adds a configurable timer.
7930       The timer value is received as third argument and stored in
7931       register %ebx.
7932
7933  S/390, System z, IBM z Systems
7934
7935     * Support for the IBM z13 processor has been added. When using the
7936       -march=z13 option, the compiler will generate code making use of
7937       the new instructions and registers introduced with the vector
7938       extension facility. The -mtune=z13 option enables z13 specific
7939       instruction scheduling without making use of new instructions.
7940       Compiling code with -march=z13 reduces the default alignment of
7941       vector types bigger than 8 bytes to 8. This is an ABI change and
7942       care must be taken when linking modules compiled with different
7943       arch levels which interchange variables containing vector type
7944       values. For newly compiled code the GNU linker will emit a warning.
7945     * The -mzvector option enables a C/C++ language extension. This
7946       extension provides a new keyword vector which can be used to define
7947       vector type variables. (Note: This is not available when enforcing
7948       strict standard compliance e.g. with -std=c99. Either enable GNU
7949       extensions with e.g. -std=gnu99 or use __vector instead of vector.)
7950       Additionally a set of overloaded builtins is provided which is
7951       partially compatible to the PowerPC Altivec builtins. In order to
7952       make use of these builtins the vecintrin.h header file needs to be
7953       included.
7954
7955                                    GCC 5.3
7956
7957   This is the [40]list of problem reports (PRs) from GCC's bug tracking
7958   system that are known to be fixed in the 5.3 release. This list might
7959   not be complete (that is, it is possible that some PRs that have been
7960   fixed are not listed here).
7961
7962Target Specific Changes
7963
7964  IA-32/x86-64
7965
7966     * GCC now supports the Intel CPU named Skylake with AVX-512
7967       extensions through -march=skylake-avx512. The switch enables the
7968       following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW,
7969       AVX-512DQ.
7970
7971  S/390, System z, IBM z Systems
7972
7973     * With this version of GCC IBM z Systems support has been added to
7974       the GO runtime environment. GCC 5.3 has proven to be able to
7975       compile larger GO applications on IBM z Systems.
7976
7977                                    GCC 5.4
7978
7979   This is the [41]list of problem reports (PRs) from GCC's bug tracking
7980   system that are known to be fixed in the 5.4 release. This list might
7981   not be complete (that is, it is possible that some PRs that have been
7982   fixed are not listed here).
7983
7984                                    GCC 5.5
7985
7986   This is the [42]list of problem reports (PRs) from GCC's bug tracking
7987   system that are known to be fixed in the 5.5 release. This list might
7988   not be complete (that is, it is possible that some PRs that have been
7989   fixed are not listed here).
7990
7991Target Specific Changes
7992
7993  IA-32/x86-64
7994
7995     * Support for the [43]deprecated pcommit instruction has been
7996       removed.
7997
7998
7999    For questions related to the use of GCC, please consult these web
8000    pages and the [44]GCC manuals. If that fails, the
8001    [45]gcc-help@gcc.gnu.org mailing list might help. Comments on these
8002    web pages and the development of GCC are welcome on our developer
8003    list at [46]gcc@gcc.gnu.org. All of [47]our lists have public
8004    archives.
8005
8006   Copyright (C) [48]Free Software Foundation, Inc. Verbatim copying and
8007   distribution of this entire article is permitted in any medium,
8008   provided this notice is preserved.
8009
8010   These pages are [49]maintained by the GCC team. Last modified
8011   2023-02-20.
8012
8013References
8014
8015   1. http://gcc.gnu.org/gcc-5/changes.html#libstdcxx
8016   2. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html
8017   3. https://savannah.nongnu.org/bugs/?44574
8018   4. https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler
8019   5. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.pdf
8020   6. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.Examples.pdf
8021   7. https://gcc.gnu.org/wiki/OpenACC
8022   8. https://gcc.gnu.org/install/configure.html
8023   9. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations
8024  10. http://gcc.gnu.org/projects/cxx1y.html
8025  11. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations
8026  12. http://gcc.gnu.org/projects/cxx1y.html
8027  13. http://gcc.gnu.org/projects/cxx1y.html
8028  14. http://gcc.gnu.org/projects/cxx1y.html
8029  15. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html
8030  16. http://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html
8031  17. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011
8032  18. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014
8033  19. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014
8034  20. https://sourceware.org/gdb/current/onlinedocs/gdb#Xmethods-In-Python
8035  21. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html
8036  22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html
8037  23. https://gcc.gnu.org/install/configure.html
8038  24. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html
8039  25. https://gcc.gnu.org/wiki/Fortran2003Status
8040  26. https://gcc.gnu.org/wiki/Fortran2008Status
8041  27. https://gcc.gnu.org/wiki/Coarray
8042  28. http://www.opencoarrays.org/
8043  29. https://pkg.go.dev/cmd/go
8044  30. https://pkg.go.dev/cmd/gofmt
8045  31. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html
8046  32. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html
8047  33. https://github.com/davidmalcolm/pygccjit
8048  34. https://github.com/ibuclaw/gccjitd
8049  35. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html
8050  36. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html
8051  37. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html
8052  38. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html
8053  39. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2
8054  40. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3
8055  41. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.4
8056  42. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.5
8057  43. https://www.intel.com/content/www/us/en/developer/articles/technical/deprecate-pcommit-instruction.html
8058  44. https://gcc.gnu.org/onlinedocs/
8059  45. mailto:gcc-help@gcc.gnu.org
8060  46. mailto:gcc@gcc.gnu.org
8061  47. https://gcc.gnu.org/lists.html
8062  48. https://www.fsf.org/
8063  49. https://gcc.gnu.org/about.html
8064======================================================================
8065http://gcc.gnu.org/gcc-4.9/index.html
8066
8067                             GCC 4.9 Release Series
8068
8069   (This release series is no longer supported.)
8070
8071   Aug 3, 2016
8072
8073   The [1]GNU project and the GCC developers are pleased to announce the
8074   release of GCC 4.9.4.
8075
8076   This release is a bug-fix release, containing fixes for regressions in
8077   GCC 4.9.3 relative to previous releases of GCC.
8078
8079Release History
8080
8081   GCC 4.9.4
8082          Aug 3, 2016 ([2]changes, [3]documentation)
8083
8084   GCC 4.9.3
8085          June 26, 2015 ([4]changes, [5]documentation)
8086
8087   GCC 4.9.2
8088          October 30, 2014 ([6]changes, [7]documentation)
8089
8090   GCC 4.9.1
8091          July 16, 2014 ([8]changes, [9]documentation)
8092
8093   GCC 4.9.0
8094          April 22, 2014 ([10]changes, [11]documentation)
8095
8096References and Acknowledgements
8097
8098   GCC used to stand for the GNU C Compiler, but since the compiler
8099   supports several other languages aside from C, it now stands for the
8100   GNU Compiler Collection.
8101
8102   A list of [12]successful builds is updated as new information becomes
8103   available.
8104
8105   The GCC developers would like to thank the numerous people that have
8106   contributed new features, improvements, bug fixes, and other changes as
8107   well as test results to GCC. This [13]amazing group of volunteers is
8108   what makes GCC successful.
8109
8110   For additional information about GCC please refer to the [14]GCC
8111   project web site or contact the [15]GCC development mailing list.
8112
8113   To obtain GCC please use [16]our mirror sites or [17]our version
8114   control system.
8115
8116
8117    For questions related to the use of GCC, please consult these web
8118    pages and the [18]GCC manuals. If that fails, the
8119    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
8120    web pages and the development of GCC are welcome on our developer
8121    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
8122    archives.
8123
8124   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
8125   distribution of this entire article is permitted in any medium,
8126   provided this notice is preserved.
8127
8128   These pages are [23]maintained by the GCC team. Last modified
8129   2022-10-26.
8130
8131References
8132
8133   1. http://www.gnu.org/
8134   2. http://gcc.gnu.org/gcc-4.9/changes.html
8135   3. http://gcc.gnu.org/onlinedocs/4.9.4/
8136   4. http://gcc.gnu.org/gcc-4.9/changes.html
8137   5. http://gcc.gnu.org/onlinedocs/4.9.3/
8138   6. http://gcc.gnu.org/gcc-4.9/changes.html
8139   7. http://gcc.gnu.org/onlinedocs/4.9.2/
8140   8. http://gcc.gnu.org/gcc-4.9/changes.html
8141   9. http://gcc.gnu.org/onlinedocs/4.9.1/
8142  10. http://gcc.gnu.org/gcc-4.9/changes.html
8143  11. https://gcc.gnu.org/onlinedocs/4.9.0/
8144  12. http://gcc.gnu.org/gcc-4.9/buildstat.html
8145  13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
8146  14. http://gcc.gnu.org/index.html
8147  15. mailto:gcc@gcc.gnu.org
8148  16. http://gcc.gnu.org/mirrors.html
8149  17. http://gcc.gnu.org/git.html
8150  18. https://gcc.gnu.org/onlinedocs/
8151  19. mailto:gcc-help@gcc.gnu.org
8152  20. mailto:gcc@gcc.gnu.org
8153  21. https://gcc.gnu.org/lists.html
8154  22. https://www.fsf.org/
8155  23. https://gcc.gnu.org/about.html
8156======================================================================
8157http://gcc.gnu.org/gcc-4.9/changes.html
8158
8159                             GCC 4.9 Release Series
8160                        Changes, New Features, and Fixes
8161
8162Caveats
8163
8164     * The mudflap run time checker has been removed. The mudflap options
8165       remain, but do nothing.
8166     * Support for a number of older systems and recently unmaintained or
8167       untested target ports of GCC has been declared obsolete in GCC 4.9.
8168       Unless there is activity to revive them, the next release of GCC
8169       will have their sources permanently removed.
8170       The following ports for individual systems on particular
8171       architectures have been obsoleted:
8172          + Solaris 9 (*-*-solaris2.9). Details can be found in the
8173            [1]announcement.
8174     * On AArch64, the singleton vector types int64x1_t, uint64x1_t and
8175       float64x1_t exported by arm_neon.h are defined to be the same as
8176       their base types. This results in incorrect application of
8177       parameter passing rules to arguments of types int64x1_t and
8178       uint64x1_t, with respect to the AAPCS64 ABI specification. In
8179       addition, names of C++ functions with parameters of these types
8180       (including float64x1_t) are not mangled correctly. The current
8181       typedef declarations also unintentionally allow implicit casting
8182       between singleton vector types and their base types. These issues
8183       will be resolved in a near future release. See [2]PR60825 for more
8184       information.
8185
8186   More information on porting to GCC 4.9 from previous versions of GCC
8187   can be found in the [3]porting guide for this release.
8188
8189General Optimizer Improvements
8190
8191     * AddressSanitizer, a fast memory error detector, is now available on
8192       ARM.
8193     * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior
8194       detector, has been added and can be enabled via
8195       -fsanitize=undefined. Various computations will be instrumented to
8196       detect undefined behavior at runtime. UndefinedBehaviorSanitizer is
8197       currently available for the C and C++ languages.
8198     * Link-time optimization (LTO) improvements:
8199          + Type merging was rewritten. The new implementation is
8200            significantly faster and uses less memory.
8201          + Better partitioning algorithm resulting in less streaming
8202            during link time.
8203          + Early removal of virtual methods reduces the size of object
8204            files and improves link-time memory usage and compile time.
8205          + Function bodies are now loaded on-demand and released early
8206            improving overall memory usage at link time.
8207          + C++ hidden keyed methods can now be optimized out.
8208          + When using a linker plugin, compiling with the -flto option
8209            now generates slim object files (.o) which only contain
8210            intermediate language representation for LTO. Use
8211            -ffat-lto-objects to create files which contain additionally
8212            the object code. To generate static libraries suitable for LTO
8213            processing, use gcc-ar and gcc-ranlib; to list symbols from a
8214            slim object file use gcc-nm. (This requires that ar, ranlib
8215            and nm have been compiled with plugin support.)
8216       Memory usage building Firefox with debug enabled was reduced from
8217       15GB to 3.5GB; link time from 1700 seconds to 350 seconds.
8218     * Inter-procedural optimization improvements:
8219          + New type inheritance analysis module improving
8220            devirtualization. Devirtualization now takes into account
8221            anonymous name-spaces and the C++11 final keyword.
8222          + New speculative devirtualization pass (controlled by
8223            -fdevirtualize-speculatively.
8224          + Calls that were speculatively made direct are turned back to
8225            indirect where direct call is not cheaper.
8226          + Local aliases are introduced for symbols that are known to be
8227            semantically equivalent across shared libraries improving
8228            dynamic linking times.
8229     * Feedback directed optimization improvements:
8230          + Profiling of programs using C++ inline functions is now more
8231            reliable.
8232          + New time profiling determines typical order in which functions
8233            are executed.
8234          + A new function reordering pass (controlled by
8235            -freorder-functions) significantly reduces startup time of
8236            large applications. Until binutils support is completed, it is
8237            effective only with link-time optimization.
8238          + Feedback driven indirect call removal and devirtualization now
8239            handle cross-module calls when link-time optimization is
8240            enabled.
8241
8242New Languages and Language specific improvements
8243
8244     * Version 4.0 of the [4]OpenMP specification is now supported in the
8245       C and C++ compilers and starting with the 4.9.1 release also in the
8246       Fortran compiler. The new -fopenmp-simd option can be used to
8247       enable OpenMP's SIMD directives while ignoring other OpenMP
8248       directives. The new [5]-fsimd-cost-model= option permits to tune
8249       the vectorization cost model for loops annotated with OpenMP and
8250       Cilk Plus simd directives. -Wopenmp-simd warns when the current
8251       cost model overrides simd directives set by the user.
8252     * The -Wdate-time option has been added for the C, C++ and Fortran
8253       compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__
8254       macros are used. Those macros might prevent bit-wise-identical
8255       reproducible compilations.
8256
8257  Ada
8258
8259     * GNAT switched to Ada 2012 instead of Ada 2005 by default.
8260
8261  C family
8262
8263     * Support for colorizing diagnostics emitted by GCC has been added.
8264       The [6]-fdiagnostics-color=auto will enable it when outputting to
8265       terminals, -fdiagnostics-color=always unconditionally. The
8266       GCC_COLORS environment variable can be used to customize the colors
8267       or disable coloring. If GCC_COLORS variable is present in the
8268       environment, the default is -fdiagnostics-color=auto, otherwise
8269       -fdiagnostics-color=never.
8270       Sample diagnostics output:
8271    $ g++ -fdiagnostics-color=always -S -Wall test.C
8272    test.C: In function `int foo()':
8273    test.C:1:14: warning: no return statement in function returning non-void [-W
8274return-type]
8275     int foo () { }
8276                  ^
8277    test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use
8278 -ftemplate-depth= to increase the maximum) instantiating `struct X<100>'
8279     template <int N> struct X { static const int value = X<N-1>::value; }; temp
8280late struct X<1000>;
8281                                                  ^
8282    test.C:2:46:   recursively required from `const int X<999>::value'
8283    test.C:2:46:   required from `const int X<1000>::value'
8284    test.C:2:88:   required from here
8285
8286    test.C:2:46: error: incomplete type `X<100>' used in nested name specifier
8287
8288     * With the new [7]#pragma GCC ivdep, the user can assert that there
8289       are no loop-carried dependencies which would prevent concurrent
8290       execution of consecutive iterations using SIMD (single instruction
8291       multiple data) instructions.
8292     * Support for Cilk Plus has been added and can be enabled with the
8293       -fcilkplus option. Cilk Plus is an extension to the C and C++
8294       languages to support data and task parallelism. The present
8295       implementation follows ABI version 1.2; all features but _Cilk_for
8296       have been implemented.
8297
8298  C
8299
8300     * ISO C11 atomics (the _Atomic type specifier and qualifier and the
8301       <stdatomic.h> header) are now supported.
8302     * ISO C11 generic selections (_Generic keyword) are now supported.
8303     * ISO C11 thread-local storage (_Thread_local, similar to GNU C
8304       __thread) is now supported.
8305     * ISO C11 support is now at a similar level of completeness to ISO
8306       C99 support: substantially complete modulo bugs, extended
8307       identifiers (supported except for corner cases when
8308       -fextended-identifiers is used), floating-point issues (mainly but
8309       not entirely relating to optional C99 features from Annexes F and
8310       G) and the optional Annexes K (Bounds-checking interfaces) and L
8311       (Analyzability).
8312     * A new C extension __auto_type provides a subset of the
8313       functionality of C++11 auto in GNU C.
8314
8315  C++
8316
8317     * The G++ implementation of [8]C++1y return type deduction for normal
8318       functions has been updated to conform to [9]N3638, the proposal
8319       accepted into the working paper. Most notably, it adds
8320       decltype(auto) for getting decltype semantics rather than the
8321       template argument deduction semantics of plain auto:
8322
8323int& f();
8324         auto  i1 = f(); // int
8325decltype(auto) i2 = f(); // int&
8326
8327     * G++ supports [10]C++1y lambda capture initializers:
8328
8329[x = 42]{ ... };
8330
8331       Actually, they have been accepted since GCC 4.5, but now the
8332       compiler doesn't warn about them with -std=c++1y, and supports
8333       parenthesized and brace-enclosed initializers as well.
8334     * G++ supports [11]C++1y variable length arrays. G++ has supported
8335       GNU/C99-style VLAs for a long time, but now additionally supports
8336       initializers and lambda capture by reference. In C++1y mode G++
8337       will complain about VLA uses that are not permitted by the draft
8338       standard, such as forming a pointer to VLA type or applying sizeof
8339       to a VLA variable. Note that it now appears that VLAs will not be
8340       part of C++14, but will be part of a separate document and then
8341       perhaps C++17.
8342
8343void f(int n) {
8344  int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3
8345  [&a]{ for (int i : a) { cout << i << endl; } }();
8346  &a; // error, taking address of VLA
8347}
8348
8349     * G++ supports the [12]C++1y [[deprecated]] attribute modulo bugs in
8350       the underlying [[gnu::deprecated]] attribute. Classes and functions
8351       can be marked deprecated and a diagnostic message added:
8352
8353class A;
8354int bar(int n);
8355#if __cplusplus > 201103
8356class [[deprecated("A is deprecated in C++14; Use B instead")]] A;
8357[[deprecated("bar is unsafe; use foo() instead")]]
8358int bar(int n);
8359
8360int foo(int n);
8361class B;
8362#endif
8363A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead
8364int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo
8365o() instead
8366
8367     * G++ supports [13]C++1y digit separators. Long numeric literals can
8368       be subdivided with a single quote ' to enhance readability:
8369
8370int i = 1048576;
8371int j = 1'048'576;
8372int k = 0x10'0000;
8373int m = 0'004'000'000;
8374int n = 0b0001'0000'0000'0000'0000'0000;
8375
8376double x = 1.602'176'565e-19;
8377double y = 1.602'176'565e-1'9;
8378
8379     * G++ supports [14]C++1y generic (polymorphic) lambdas.
8380
8381// a functional object that will increment any type
8382auto incr = [](auto x) { return x++; };
8383
8384     * As a GNU extension, G++ supports explicit template parameter syntax
8385       for generic lambdas. This can be combined in the expected way with
8386       the standard auto syntax.
8387
8388// a functional object that will add two like-type objects
8389auto add = [] <typename T> (T a, T b) { return a + b; };
8390
8391     * G++ supports unconstrained generic functions as specified by �4.1.2
8392       and �5.1.1 of [15]N3889: Concepts Lite Specification. Briefly, auto
8393       may be used as a type-specifier in a parameter declaration of any
8394       function declarator in order to introduce an implicit function
8395       template parameter, akin to generic lambdas.
8396
8397// the following two function declarations are equivalent
8398auto incr(auto x) { return x++; }
8399template <typename T>
8400auto incr(T x) { return x++; }
8401
8402    Runtime Library (libstdc++)
8403
8404     * [16]Improved support for C++11, including:
8405          + support for <regex>;
8406          + The associative containers in <map> and <set> and the
8407            unordered associative containers in <unordered_map> and
8408            <unordered_set> meet the allocator-aware container
8409            requirements;
8410     * [17]Improved experimental support for the upcoming ISO C++
8411       standard, C++14, including:
8412          + fixing constexpr member functions without const;
8413          + implementation of the std::exchange() utility function;
8414          + addressing tuples by type;
8415          + implemention of std::make_unique;
8416          + implemention of std::shared_lock;
8417          + making std::result_of SFINAE-friendly;
8418          + adding operator() to std::integral_constant;
8419          + adding user-defined literals for standard library types
8420            std::basic_string, std::chrono::duration, and std::complex;
8421          + adding two range overloads to non-modifying sequence oprations
8422            std::equal and std::mismatch;
8423          + adding IO manipulators for quoted strings;
8424          + adding constexpr members to <utility>, <complex>, <chrono>,
8425            and some containers;
8426          + adding compile-time std::integer_sequence;
8427          + adding cleaner transformation traits;
8428          + making <functional>s operator functors easier to use and more
8429            generic;
8430     * An implementation of std::experimental::optional.
8431     * An implementation of std::experimental::string_view.
8432     * The non-standard function std::copy_exception has been deprecated
8433       and will be removed in a future version. std::make_exception_ptr
8434       should be used instead.
8435
8436  Fortran
8437
8438     * Compatibility notice:
8439          + Module files: The version of the module files (.mod) has been
8440            incremented; additionally, module files are now compressed.
8441            Fortran MODULEs compiled by earlier GCC versions have to be
8442            recompiled, when they are USEd by files compiled with GCC 4.9.
8443            GCC 4.9 is not able to read .mod files of earlier GCC
8444            versions; attempting to do so gives an error message. Note:
8445            The ABI of the produced assembler data itself has not changed:
8446            object files and libraries are fully compatible with older
8447            versions (except as stated below).
8448          + ABI changes:
8449               o The [18]argument passing ABI has changed for scalar dummy
8450                 arguments of type INTEGER, REAL, COMPLEX and LOGICAL,
8451                 which have both the VALUE and the OPTIONAL attributes.
8452               o To support finalization the virtual table associated with
8453                 polymorphic variables has changed. Code containing CLASS
8454                 should be recompiled, including all files which define
8455                 derived types involved in the type definition used by
8456                 polymorphic variables. (Note: Due to the incremented
8457                 module version, trying to mix old code with new code will
8458                 usually give an error message.)
8459          + GNU Fortran no longer deallocates allocatable variables or
8460            allocatable components of variables declared in the main
8461            program. Since Fortran 2008, the standard explicitly states
8462            that variables declared in the Fortran main program
8463            automatically have the SAVE attribute.
8464          + When opening files, the close-on-exec flag is set if the
8465            system supports such a feature. This is generally considered
8466            good practice these days, but if there is a need to pass file
8467            descriptors to child processes the parent process must now
8468            remember to clear the close-on-exec flag by calling fcntl(),
8469            e.g. via ISO_C_BINDING, before executing the child process.
8470     * The deprecated command-line option -fno-whole-file has been
8471       removed. (-fwhole-file is the default since GCC 4.6.)
8472       -fwhole-file/-fno-whole-file continue to be accepted but do not
8473       influence the code generation.
8474     * The compiler no longer unconditionally warns about DO loops with
8475       zero iterations. This warning is now controlled by the -Wzerotrip
8476       option, which is implied by -Wall.
8477     * The new NO_ARG_CHECK attribute of the [19]!GCC$ directive can be
8478       used to disable the type-kind-rank (TKR) argument check for a dummy
8479       argument. The feature is similar to ISO/IEC TS 29133:2012's
8480       TYPE(*), except that it additionally also disables the rank check.
8481       Variables with NO_ARG_CHECK have to be dummy arguments and may only
8482       be used as argument to ISO_C_BINDING's C_LOC and as actual argument
8483       to another NO_ARG_CHECK dummy argument; also the other constraints
8484       of TYPE(*) apply. The dummy arguments should be declared as scalar
8485       or assumed-size variable of type type(*) (recommended) - or of type
8486       integer, real, complex or logical. With NO_ARG_CHECK, a pointer to
8487       the data without further type or shape information is passed,
8488       similar to C's void*. Note that also TS 29113's
8489       type(*),dimension(..) accepts arguments of any type and rank;
8490       contrary to NO_ARG_CHECK assumed-rank arguments pass an array
8491       descriptor which contains the array shape and stride of the
8492       argument.
8493     * [20]Fortran 2003:
8494          + Finalization is now supported. It is currently only done for a
8495            subset of those situations in which it should occur.
8496          + Experimental support for scalar character components with
8497            deferred length (i.e. allocatable string length) in derived
8498            types has been added. (Deferred-length character variables are
8499            supported since GCC 4.6.)
8500     * [21]Fortran 2008:
8501          + When STOP or ERROR STOP are used to terminate the execution
8502            and any exception (but inexact) is signaling, a warning is
8503            printed to ERROR_UNIT, indicating which exceptions are
8504            signaling. The [22]-ffpe-summary= command-line option can be
8505            used to fine-tune for which exceptions the warning should be
8506            shown.
8507          + Rounding on input (READ) is now handled on systems where
8508            strtod honours the rounding mode. (For output, rounding is
8509            supported since GCC 4.5.) Note that for input, the compatible
8510            rounding mode is handled as nearest (i.e., rounding to an even
8511            least significant [cf. IEC 60559:1989] for a tie, while
8512            compatible rounds away from zero in that case).
8513
8514  Go
8515
8516     * GCC 4.9 provides a complete implementation of the Go 1.2.1 release.
8517
8518New Targets and Target Specific Improvements
8519
8520  AArch64
8521
8522     * The ARMv8-A crypto and CRC instructions are now supported through
8523       intrinsics. These are enabled when the architecture supports these
8524       and are available through the -march=armv8-a+crc and
8525       -march=armv8-a+crypto options.
8526     * Initial support for ILP32 has now been added to the compiler. This
8527       is now available through the command-line option -mabi=ilp32.
8528       Support for ILP32 is considered experimental as the ABI
8529       specification is still beta.
8530     * Coverage of more of the ISA including the SIMD extensions has been
8531       added. The Advanced SIMD intrinsics have also been improved.
8532     * The new local register allocator (LRA) is now on by default for the
8533       AArch64 backend.
8534     * The REE (Redundant extension elimination) pass has now been enabled
8535       by default for the AArch64 backend.
8536     * Tuning for the Cortex-A53 and Cortex-A57 has been improved.
8537     * Initial big.LITTLE tuning support for the combination of Cortex-A57
8538       and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53
8539       option.
8540     * A number of structural changes have been made to both the ARM and
8541       AArch64 backends to facilitate improved code-generation.
8542     * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769
8543       has been added and can be enabled by giving the
8544       -mfix-cortex-a53-835769 option. Alternatively it can be enabled by
8545       default by configuring GCC with the --enable-fix-cortex-a53-835769
8546       option.
8547
8548  ARC
8549
8550     * A port for Synopsys Designware ARC has been contributed by Embecosm
8551       and Synopsys Inc.
8552
8553  ARM
8554
8555     * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been
8556       disabled by default. This was found to generate better code in only
8557       a small number of cases. It can be turned back on with the
8558       -mneon-for-64bits option.
8559     * Further support for the ARMv8-A architecture, notably implementing
8560       the restriction around IT blocks in the Thumb32 instruction set has
8561       been added. The -mrestrict-it option can be used with
8562       -march=armv7-a or the -march=armv7ve options to make code
8563       generation fully compatible with the deprecated instructions in
8564       ARMv8-A.
8565     * Support has now been added for the ARMv7ve variant of the
8566       architecture. This can be used by the -march=armv7ve option.
8567     * The ARMv8-A crypto and CRC instructions are now supported through
8568       intrinsics and are available through the -march=armv8-a+crc and
8569       mfpu=crypto-neon-fp-armv8 options.
8570     * LRA is now on by default for the ARM target. This can be turned off
8571       using the -mno-lra option. This option is a purely transitionary
8572       command-line option and will be removed in a future release. We are
8573       interested in any bug reports regarding functional and performance
8574       regressions with LRA.
8575     * A new option -mslow-flash-data to improve performance of programs
8576       fetching data on slow flash memory has now been introduced for the
8577       ARMv7-M profile cores.
8578     * A new option -mpic-data-is-text-relative for targets that allows
8579       data segments to be relative to text segments has been added. This
8580       is on by default for all targets except VxWorks RTP.
8581     * A number of infrastructural changes have been made to both the ARM
8582       and AArch64 backends to facilitate improved code-generation.
8583     * GCC now supports Cortex-A12 and the Cortex-R7 through the
8584       -mcpu=cortex-a12 and -mcpu=cortex-r7 options.
8585     * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the
8586       -mcpu=cortex-a57 and -mcpu=cortex-a53 options.
8587     * Initial big.LITTLE tuning support for the combination of Cortex-A57
8588       and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53
8589       option. Similar support was added for the combination of Cortex-A15
8590       and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option.
8591     * Further performance optimizations for the Cortex-A15 and the
8592       Cortex-M4 have been added.
8593     * A number of code generation improvements for Thumb2 to reduce code
8594       size when compiling for the M-profile processors.
8595
8596  AVR
8597
8598     * A new command-line option -mfract-convert-truncate has been added.
8599       It allows compiler to use truncation instead of rounding towards
8600       zero for fractional fixed-point types.
8601
8602  IA-32/x86-64
8603
8604     * -mfpmath=sse is now implied by -ffast-math on all targets where
8605       SSE2 is supported.
8606     * Intel AVX-512 support was added to GCC. That includes inline
8607       assembly support, new registers and extending existing ones, new
8608       intrinsics (covered by corresponding testsuite), and basic
8609       autovectorization. AVX-512 instructions are available via the
8610       following GCC switches: AVX-512 foundation instructions: -mavx512f,
8611       AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and
8612       reciprocal instructions: -mavx512er, AVX-512 conflict detection
8613       instructions: -mavx512cd.
8614     * It is now possible to call x86 intrinsics from select functions in
8615       a file that are tagged with the corresponding target attribute
8616       without having to compile the entire file with the -mxxx option.
8617       This improves the usability of x86 intrinsics and is particularly
8618       useful when doing [23]Function Multiversioning.
8619     * GCC now supports the new Intel microarchitecture named Silvermont
8620       through -march=silvermont.
8621     * GCC now supports the new Intel microarchitecture named Broadwell
8622       through -march=broadwell.
8623     * Optimizing for other Intel microarchitectures have been renamed to
8624       -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell.
8625     * -march=generic has been retuned for better support of Intel core
8626       and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel
8627       Pentium-M, and Pentium4 based CPUs is no longer considered
8628       important for generic.
8629     * -mtune=intel can now be used to generate code running well on the
8630       most current Intel processors, which are Haswell and Silvermont for
8631       GCC 4.9.
8632     * Support to encode 32-bit assembly instructions in 16-bit format is
8633       now available through the -m16 command-line option.
8634     * Better inlining of memcpy and memset that is aware of value ranges
8635       and produces shorter alignment prologues.
8636     * -mno-accumulate-outgoing-args is now honored when unwind
8637       information is output. Argument accumulation is also now turned off
8638       for portions of programs optimized for size.
8639     * Support for new AMD family 15h processors (Excavator core) is now
8640       available through the -march=bdver4 and -mtune=bdver4 options.
8641
8642  MSP430
8643
8644     * A new command-line option -mcpu= has been added to the MSP430
8645       backend. This option is used to specify the ISA to be used.
8646       Accepted values are msp430 (the default), msp430x and msp430xv2.
8647       The ISA is no longer deduced from the -mmcu= option as there are
8648       far too many different MCU names. The -mmcu= option is still
8649       supported, and this is still used to select linker scripts and
8650       generate a C preprocessor symbol that will be recognised by the
8651       msp430.h header file.
8652
8653  NDS32
8654
8655     * A new nds32 port supports the 32-bit architecture from Andes
8656       Technology Corporation.
8657     * The port provides initial support for the V2, V3, V3m instruction
8658       set architectures.
8659
8660  Nios II
8661
8662     * A port for the Altera Nios II has been contributed by Mentor
8663       Graphics.
8664
8665  PowerPC / PowerPC64 / RS6000
8666
8667     * GCC now supports Power ISA 2.07, which includes support for
8668       Hardware Transactional Memory (HTM), Quadword atomics and several
8669       VMX and VSX additions, including Crypto, 64-bit integer, 128-bit
8670       integer and decimal integer operations.
8671     * Support for the POWER8 processor is now available through the
8672       -mcpu=power8 and -mtune=power8 options.
8673     * The libitm library has been modified to add a HTM fastpath that
8674       automatically uses POWER's HTM hardware instructions when it is
8675       executing on a HTM enabled processor.
8676     * Support for the new powerpc64le-linux platform has been added. It
8677       defaults to generating code that conforms to the ELFV2 ABI.
8678
8679  S/390, System z
8680
8681     * Support for the Transactional Execution Facility included with the
8682       IBM zEnterprise zEC12 processor has been added. A set of GCC style
8683       builtins as well as XLC style builtins are provided. The builtins
8684       are enabled by default when using the -march=zEC12 option but can
8685       explicitly be disabled with -mno-htm. Using the GCC builtins also
8686       libitm supports hardware transactions on S/390.
8687     * The hotpatch features allows to prepare functions for hotpatching.
8688       A certain amount of bytes is reserved before the function entry
8689       label plus a NOP is inserted at its very beginning to implement a
8690       backward jump when applying a patch. The feature can either be
8691       enabled per compilation unit via the command-line option -mhotpatch
8692       or per function using the hotpatch attribute.
8693     * The shrink wrap optimization is now supported on S/390 and enabled
8694       by default.
8695     * A major rework of the routines to determine which registers need to
8696       be saved and restored in function prologue/epilogue now allow to
8697       use floating point registers as save slots. This will happen for
8698       certain leaf function with -march=z10 or higher.
8699     * The LRA rtl pass replaces reload by default on S/390.
8700
8701  RX
8702
8703     * The port now allows to specify the RX100, RX200, and RX600
8704       processors with the command-line options -mcpu=rx100, -mcpu=rx200
8705       and -mcpu=rx600.
8706
8707  SH
8708
8709     * Minor improvements to code generated for integer arithmetic and
8710       code that involves the T bit.
8711     * Added support for the SH2A clips and clipu instructions. The
8712       compiler will now try to utilize them for min/max expressions such
8713       as max (-128, min (127, x)).
8714     * Added support for the cmp/str instruction through built-in
8715       functions such as __builtin_strlen. When not optimizing for size,
8716       the compiler will now expand calls to e.g. strlen as an inlined
8717       sequences which utilize the cmp/str instruction.
8718     * Improved code generated around volatile memory loads and stores.
8719     * The option -mcbranchdi has been deprecated. Specifying it will
8720       result in a warning and will not influence code generation.
8721     * The option -mcmpeqdi has been deprecated. Specifying it will result
8722       in a warning and will not influence code generation.
8723
8724GCC 4.9.1
8725
8726   This is the [24]list of problem reports (PRs) from GCC's bug tracking
8727   system that are known to be fixed in the 4.9.1 release. This list might
8728   not be complete (that is, it is possible that some PRs that have been
8729   fixed are not listed here).
8730
8731   Version 4.0 of the OpenMP specification is supported even in Fortran,
8732   not just C and C++.
8733
8734GCC 4.9.2
8735
8736   This is the [25]list of problem reports (PRs) from GCC's bug tracking
8737   system that are known to be fixed in the 4.9.2 release. This list might
8738   not be complete (that is, it is possible that some PRs that have been
8739   fixed are not listed here).
8740
8741GCC 4.9.3
8742
8743   This is the [26]list of problem reports (PRs) from GCC's bug tracking
8744   system that are known to be fixed in the 4.9.3 release. This list might
8745   not be complete (that is, it is possible that some PRs that have been
8746   fixed are not listed here).
8747
8748GCC 4.9.4
8749
8750   This is the [27]list of problem reports (PRs) from GCC's bug tracking
8751   system that are known to be fixed in the 4.9.4 release. This list might
8752   not be complete (that is, it is possible that some PRs that have been
8753   fixed are not listed here).
8754
8755
8756    For questions related to the use of GCC, please consult these web
8757    pages and the [28]GCC manuals. If that fails, the
8758    [29]gcc-help@gcc.gnu.org mailing list might help. Comments on these
8759    web pages and the development of GCC are welcome on our developer
8760    list at [30]gcc@gcc.gnu.org. All of [31]our lists have public
8761    archives.
8762
8763   Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and
8764   distribution of this entire article is permitted in any medium,
8765   provided this notice is preserved.
8766
8767   These pages are [33]maintained by the GCC team. Last modified
8768   2023-01-19.
8769
8770References
8771
8772   1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html
8773   2. https://gcc.gnu.org/PR60825
8774   3. https://gcc.gnu.org/gcc-4.9/porting_to.html
8775   4. https://www.openmp.org/specifications/
8776   5. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908
8777   6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252
8778   7. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html
8779   8. http://gcc.gnu.org/projects/cxx1y.html
8780   9. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html
8781  10. http://gcc.gnu.org/projects/cxx1y.html
8782  11. http://gcc.gnu.org/projects/cxx1y.html
8783  12. http://gcc.gnu.org/projects/cxx1y.html
8784  13. http://gcc.gnu.org/projects/cxx1y.html
8785  14. http://gcc.gnu.org/projects/cxx1y.html
8786  15. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf
8787  16. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011
8788  17. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014
8789  18. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html
8790  19. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html
8791  20. https://gcc.gnu.org/wiki/Fortran2003Status
8792  21. https://gcc.gnu.org/wiki/Fortran2008Status
8793  22. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html
8794  23. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html
8795  24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1
8796  25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2
8797  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3
8798  27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.4
8799  28. https://gcc.gnu.org/onlinedocs/
8800  29. mailto:gcc-help@gcc.gnu.org
8801  30. mailto:gcc@gcc.gnu.org
8802  31. https://gcc.gnu.org/lists.html
8803  32. https://www.fsf.org/
8804  33. https://gcc.gnu.org/about.html
8805======================================================================
8806http://gcc.gnu.org/gcc-4.8/index.html
8807
8808                             GCC 4.8 Release Series
8809
8810   (This release series is no longer supported.)
8811
8812   June 23, 2015
8813
8814   The [1]GNU project and the GCC developers are pleased to announce the
8815   release of GCC 4.8.5.
8816
8817   This release is a bug-fix release, containing fixes for regressions in
8818   GCC 4.8.4 relative to previous releases of GCC.
8819
8820Release History
8821
8822   GCC 4.8.5
8823          June 23, 2015 ([2]changes, [3]documentation)
8824
8825   GCC 4.8.4
8826          December 19, 2014 ([4]changes, [5]documentation)
8827
8828   GCC 4.8.3
8829          May 22, 2014 ([6]changes, [7]documentation)
8830
8831   GCC 4.8.2
8832          October 16, 2013 ([8]changes, [9]documentation)
8833
8834   GCC 4.8.1
8835          May 31, 2013 ([10]changes, [11]documentation)
8836
8837   GCC 4.8.0
8838          March 22, 2013 ([12]changes, [13]documentation)
8839
8840References and Acknowledgements
8841
8842   GCC used to stand for the GNU C Compiler, but since the compiler
8843   supports several other languages aside from C, it now stands for the
8844   GNU Compiler Collection.
8845
8846   A list of [14]successful builds is updated as new information becomes
8847   available.
8848
8849   The GCC developers would like to thank the numerous people that have
8850   contributed new features, improvements, bug fixes, and other changes as
8851   well as test results to GCC. This [15]amazing group of volunteers is
8852   what makes GCC successful.
8853
8854   For additional information about GCC please refer to the [16]GCC
8855   project web site or contact the [17]GCC development mailing list.
8856
8857   To obtain GCC please use [18]our mirror sites or [19]our version
8858   control system.
8859
8860
8861    For questions related to the use of GCC, please consult these web
8862    pages and the [20]GCC manuals. If that fails, the
8863    [21]gcc-help@gcc.gnu.org mailing list might help. Comments on these
8864    web pages and the development of GCC are welcome on our developer
8865    list at [22]gcc@gcc.gnu.org. All of [23]our lists have public
8866    archives.
8867
8868   Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and
8869   distribution of this entire article is permitted in any medium,
8870   provided this notice is preserved.
8871
8872   These pages are [25]maintained by the GCC team. Last modified
8873   2022-10-26.
8874
8875References
8876
8877   1. http://www.gnu.org/
8878   2. http://gcc.gnu.org/gcc-4.8/changes.html
8879   3. https://gcc.gnu.org/onlinedocs/4.8.5/
8880   4. http://gcc.gnu.org/gcc-4.8/changes.html
8881   5. https://gcc.gnu.org/onlinedocs/4.8.4/
8882   6. http://gcc.gnu.org/gcc-4.8/changes.html
8883   7. https://gcc.gnu.org/onlinedocs/4.8.3/
8884   8. http://gcc.gnu.org/gcc-4.8/changes.html
8885   9. https://gcc.gnu.org/onlinedocs/4.8.2/
8886  10. http://gcc.gnu.org/gcc-4.8/changes.html
8887  11. https://gcc.gnu.org/onlinedocs/4.8.1/
8888  12. http://gcc.gnu.org/gcc-4.8/changes.html
8889  13. https://gcc.gnu.org/onlinedocs/4.8.0/
8890  14. http://gcc.gnu.org/gcc-4.8/buildstat.html
8891  15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
8892  16. http://gcc.gnu.org/index.html
8893  17. mailto:gcc@gcc.gnu.org
8894  18. http://gcc.gnu.org/mirrors.html
8895  19. http://gcc.gnu.org/git.html
8896  20. https://gcc.gnu.org/onlinedocs/
8897  21. mailto:gcc-help@gcc.gnu.org
8898  22. mailto:gcc@gcc.gnu.org
8899  23. https://gcc.gnu.org/lists.html
8900  24. https://www.fsf.org/
8901  25. https://gcc.gnu.org/about.html
8902======================================================================
8903http://gcc.gnu.org/gcc-4.8/changes.html
8904
8905                             GCC 4.8 Release Series
8906                        Changes, New Features, and Fixes
8907
8908Caveats
8909
8910   GCC now uses C++ as its implementation language. This means that to
8911   build GCC from sources, you will need a C++ compiler that understands
8912   C++ 2003. For more details on the rationale and specific changes,
8913   please refer to the [1]C++ conversion page.
8914
8915   To enable the Graphite framework for loop optimizations you now need
8916   CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from
8917   the [2]GCC infrastructure directory. The installation manual contains
8918   more information about requirements to build GCC.
8919
8920   GCC now uses a more aggressive analysis to derive an upper bound for
8921   the number of iterations of loops using constraints imposed by language
8922   standards. This may cause non-conforming programs to no longer work as
8923   expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new
8924   option, -fno-aggressive-loop-optimizations, was added to disable this
8925   aggressive analysis. In some loops that have known constant number of
8926   iterations, but undefined behavior is known to occur in the loop before
8927   reaching or during the last iteration, GCC will warn about the
8928   undefined behavior in the loop instead of deriving lower upper bound of
8929   the number of iterations for the loop. The warning can be disabled with
8930   -Wno-aggressive-loop-optimizations.
8931
8932   On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules
8933   for the layout of vectors that could lead to wrong code being
8934   generated. Vectors larger than 8 bytes in size are now by default
8935   aligned to an 8-byte boundary. This is an ABI change: code that makes
8936   explicit use of vector types may be incompatible with binary objects
8937   built with older versions of GCC. Auto-vectorized code is not affected
8938   by this change.
8939
8940   On AVR, support has been removed for the command-line option
8941   -mshort-calls deprecated in GCC 4.7.
8942
8943   On AVR, the configure option --with-avrlibc supported since GCC 4.7.2
8944   is turned on per default for all non-RTEMS configurations. This option
8945   arranges for a better integration of [3]AVR Libc with avr-gcc. For
8946   technical details, see [4]PR54461. To turn off the option in non-RTEMS
8947   configurations, use --with-avrlibc=no. If the compiler is configured
8948   for RTEMS, the option is always turned off.
8949
8950   More information on porting to GCC 4.8 from previous versions of GCC
8951   can be found in the [5]porting guide for this release.
8952
8953General Optimizer Improvements (and Changes)
8954
8955     * DWARF4 is now the default when generating DWARF debug information.
8956       When -g is used on a platform that uses DWARF debugging
8957       information, GCC will now default to -gdwarf-4
8958       -fno-debug-types-section.
8959       GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information
8960       consumers support DWARF4 by default. Before GCC 4.8 the default
8961       version used was DWARF2. To make GCC 4.8 generate an older DWARF
8962       version use -g together with -gdwarf-2 or -gdwarf-3. The default
8963       for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf.
8964     * A new general optimization level, -Og, has been introduced. It
8965       addresses the need for fast compilation and a superior debugging
8966       experience while providing a reasonable level of run-time
8967       performance. Overall experience for development should be better
8968       than the default optimization level -O0.
8969     * A new option -ftree-partial-pre was added to control the partial
8970       redundancy elimination (PRE) optimization. This option is enabled
8971       by default at the -O3 optimization level, and it makes PRE more
8972       aggressive.
8973     * The option -fconserve-space has been removed; it was no longer
8974       useful on most targets since GCC supports putting variables into
8975       BSS without making them common.
8976     * The struct reorg and matrix reorg optimizations (command-line
8977       options -fipa-struct-reorg and -fipa-matrix-reorg) have been
8978       removed. They did not always work correctly, nor did they work with
8979       link-time optimization (LTO), hence were only applicable to
8980       programs consisting of a single translation unit.
8981     * Several scalability bottle-necks have been removed from GCC's
8982       optimization passes. Compilation of extremely large functions, e.g.
8983       due to the use of the flatten attribute in the "Eigen" C++ linear
8984       algebra templates library, is significantly faster than previous
8985       releases of GCC.
8986     * Link-time optimization (LTO) improvements:
8987          + LTO partitioning has been rewritten for better reliability and
8988            maintanibility. Several important bugs leading to link
8989            failures have been fixed.
8990     * Interprocedural optimization improvements:
8991          + A new symbol table has been implemented. It builds on existing
8992            callgraph and varpool modules and provide a new API. Unusual
8993            symbol visibilities and aliases are handled more consistently
8994            leading to, for example, more aggressive unreachable code
8995            removal with LTO.
8996          + The inline heuristic can now bypass limits on the size of of
8997            inlined functions when the inlining is particularly
8998            profitable. This happens, for example, when loop bounds or
8999            array strides get propagated.
9000          + Values passed through aggregates (either by value or
9001            reference) are now propagated at the inter-procedural level
9002            leading to better inlining decisions (for example in the case
9003            of Fortran array descriptors) and devirtualization.
9004     * [6]AddressSanitizer , a fast memory error detector, has been added
9005       and can be enabled via -fsanitize=address. Memory access
9006       instructions will be instrumented to detect heap-, stack-, and
9007       global-buffer overflow as well as use-after-free bugs. To get nicer
9008       stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is
9009       available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on
9010       x86-64 Darwin.
9011     * [7]ThreadSanitizer has been added and can be enabled via
9012       -fsanitize=thread. Instructions will be instrumented to detect data
9013       races. The ThreadSanitizer is available on x86-64 GNU/Linux.
9014     * A new local register allocator (LRA) has been implemented, which
9015       replaces the 26 year old reload pass and improves generated code
9016       quality. For now it is active on the IA-32 and x86-64 targets.
9017     * Support for transactional memory has been implemented on the
9018       following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and
9019       Alpha.
9020
9021New Languages and Language specific improvements
9022
9023  C family
9024
9025     * Each diagnostic emitted now includes the original source line and a
9026       caret '^' indicating the column. The option
9027       -fno-diagnostics-show-caret suppresses this information.
9028     * The option -ftrack-macro-expansion=2 is now enabled by default.
9029       This allows the compiler to display the macro expansion stack in
9030       diagnostics. Combined with the caret information, an example
9031       diagnostic showing these two features is:
9032
9033t.c:1:94: error: invalid operands to binary < (have `struct mystruct' and `float
9034')
9035 #define MYMAX(A,B)    __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _
9036_b = (B); __a < __b ? __b : __a; })
9037
9038              ^
9039t.c:7:7: note: in expansion of macro 'MYMAX'
9040   X = MYMAX(P, F);
9041       ^
9042
9043     * A new -Wsizeof-pointer-memaccess warning has been added (also
9044       enabled by -Wall) to warn about suspicious length parameters to
9045       certain string and memory built-in functions if the argument uses
9046       sizeof. This warning warns e.g. about memset (ptr, 0, sizeof
9047       (ptr)); if ptr is not an array, but a pointer, and suggests a
9048       possible fix, or about memcpy (&foo, ptr, sizeof (&foo));.
9049     * The new option -Wpedantic is an alias for -pedantic, which is now
9050       deprecated. The forms -Wno-pedantic, -Werror=pedantic, and
9051       -Wno-error=pedantic work in the same way as for any other -W
9052       option. One caveat is that -Werror=pedantic is not equivalent to
9053       -pedantic-errors, since the latter makes into errors some warnings
9054       that are not controlled by -Wpedantic, and the former only affects
9055       diagnostics that are disabled when using -Wno-pedantic.
9056     * The option -Wshadow no longer warns if a declaration shadows a
9057       function declaration, unless the former declares a function or
9058       pointer to function, because this is [8]a common and valid case in
9059       real-world code.
9060
9061  C++
9062
9063     * G++ now implements the [9]C++11 thread_local keyword; this differs
9064       from the GNU __thread keyword primarily in that it allows dynamic
9065       initialization and destruction semantics. Unfortunately, this
9066       support requires a run-time penalty for references to
9067       non-function-local thread_local variables defined in a different
9068       translation unit even if they don't need dynamic initialization, so
9069       users may want to continue to use __thread for TLS variables with
9070       static initialization semantics.
9071       If the programmer can be sure that no use of the variable in a
9072       non-defining TU needs to trigger dynamic initialization (either
9073       because the variable is statically initialized, or a use of the
9074       variable in the defining TU will be executed before any uses in
9075       another TU), they can avoid this overhead with the
9076       -fno-extern-tls-init option.
9077       OpenMP threadprivate variables now also support dynamic
9078       initialization and destruction by the same mechanism.
9079     * G++ now implements the [10]C++11 attribute syntax, e.g.
9080
9081[[noreturn]] void f();
9082
9083       and also the alignment specifier, e.g.
9084
9085alignas(double) int i;
9086
9087     * G++ now implements [11]C++11 inheriting constructors, e.g.
9088
9089struct A { A(int); };
9090struct B: A { using A::A; }; // defines B::B(int)
9091B b(42); // OK
9092
9093     * As of GCC 4.8.1, G++ implements the change to decltype semantics
9094       from [12]N3276.
9095
9096struct A f();
9097decltype(f()) g();    // OK, return type of f() is not required to be complete.
9098
9099     * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g.
9100
9101struct A { int f() &; };
9102int i = A().f();  // error, f() requires an lvalue object
9103
9104     * G++ now supports a -std=c++1y option for experimentation with
9105       features proposed for the next revision of the standard, expected
9106       around 2014. Currently the only difference from -std=c++11 is
9107       support for return type deduction in normal functions, as proposed
9108       in [14]N3386. Status of C++1y features in GCC 4.8 can be found
9109       [15]here.
9110     * The G++ namespace association extension, __attribute ((strong)),
9111       has been deprecated. Inline namespaces should be used instead.
9112     * G++ now supports a -fext-numeric-literal option to control whether
9113       GNU numeric literal suffixes are accepted as extensions or
9114       processed as C++11 user-defined numeric literal suffixes. The flag
9115       is on (use suffixes for GNU literals) by default for -std=gnu++*,
9116       and -std=c++98. The flag is off (use suffixes for user-defined
9117       literals) by default for -std=c++11 and later.
9118
9119    Runtime Library (libstdc++)
9120
9121     * [16]Improved experimental support for the new ISO C++ standard,
9122       C++11, including:
9123          + forward_list meets the allocator-aware container requirements;
9124          + this_thread::sleep_for(), this_thread::sleep_until() and
9125            this_thread::yield() are defined without requiring the
9126            configure option --enable-libstdcxx-time;
9127     * Improvements to <random>:
9128          + SSE optimized normal_distribution.
9129          + Use of hardware RNG instruction for random_device on new x86
9130            processors (requires the assembler to support the
9131            instruction.)
9132       and <ext/random>:
9133          + New random number engine simd_fast_mersenne_twister_engine
9134            with an optimized SSE implementation.
9135          + New random number distributions beta_distribution,
9136            normal_mv_distribution, rice_distribution,
9137            nakagami_distribution, pareto_distribution, k_distribution,
9138            arcsine_distribution, hoyt_distribution.
9139     * Added --disable-libstdcxx-verbose configure option to disable
9140       diagnostic messages issued when a process terminates abnormally.
9141       This may be useful for embedded systems to reduce the size of
9142       executables that link statically to the library.
9143
9144  Fortran
9145
9146     * Compatibility notice:
9147          + Module files: The version of module files (.mod) has been
9148            incremented. Fortran MODULEs compiled by earlier GCC versions
9149            have to be recompiled, when they are USEd by files compiled
9150            with GCC 4.8. GCC 4.8 is not able to read .mod files created
9151            by earlier versions; attempting to do so gives an error
9152            message.
9153            Note: The ABI of the produced assembler data itself has not
9154            changed; object files and libraries are fully compatible with
9155            older versions except as noted below.
9156          + ABI: Some internal names (used in the assembler/object file)
9157            have changed for symbols declared in the specification part of
9158            a module. If an affected module - or a file using it via use
9159            association - is recompiled, the module and all files which
9160            directly use such symbols have to be recompiled as well. This
9161            change only affects the following kind of module symbols:
9162               o Procedure pointers. Note: C-interoperable function
9163                 pointers (type(c_funptr)) are not affected nor are
9164                 procedure-pointer components.
9165               o Deferred-length character strings.
9166     * The [17]BACKTRACE intrinsic subroutine has been added. It shows a
9167       backtrace at an arbitrary place in user code; program execution
9168       continues normally afterwards.
9169     * The [18]-Wc-binding-type warning option has been added (disabled by
9170       default). It warns if the a variable might not be C interoperable;
9171       in particular, if the variable has been declared using an intrinsic
9172       type with default kind instead of using a kind parameter defined
9173       for C interoperability in the intrinsic ISO_C_Binding module.
9174       Before, this warning was always printed. The -Wc-binding-type
9175       option is enabled by -Wall.
9176     * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line
9177       options have been added, which diagnose when code is inserted for
9178       automatic (re)allocation of a variable during assignment. This
9179       option can be used to decide whether it is safe to use
9180       [20]-fno-realloc-lhs. Additionally, it can be used to find
9181       automatic (re)allocation in hot loops. (For arrays, replacing
9182       "var=" by "var(:)=" disables the automatic reallocation.)
9183     * The [21]-Wcompare-reals command-line option has been added. When
9184       this is set, warnings are issued when comparing REAL or COMPLEX
9185       types for equality and inequality; consider replacing a == b by
9186       abs(a-b) < eps with a suitable eps. -Wcompare-reals is enabled by
9187       -Wextra.
9188     * The [22]-Wtarget-lifetime command-line option has been added
9189       (enabled with -Wall), which warns if the pointer in a pointer
9190       assignment might outlive its target.
9191     * Reading floating point numbers which use "q" for the exponential
9192       (such as 4.0q0) is now supported as vendor extension for better
9193       compatibility with old data files. It is strongly recommended to
9194       use for I/O the equivalent but standard conforming "e" (such as
9195       4.0e0).
9196       (For Fortran source code, consider replacing the "q" in
9197       floating-point literals by a kind parameter (e.g. 4.0e0_qp with a
9198       suitable qp). Note that - in Fortran source code - replacing "q" by
9199       a simple "e" is not equivalent.)
9200     * The GFORTRAN_TMPDIR environment variable for specifying a
9201       non-default directory for files opened with STATUS="SCRATCH", is
9202       not used anymore. Instead gfortran checks the POSIX/GNU standard
9203       TMPDIR environment variable. If TMPDIR is not defined, gfortran
9204       falls back to other methods to determine the directory for
9205       temporary files as documented in the [23]user manual.
9206     * [24]Fortran 2003:
9207          + Support for unlimited polymorphic variables (CLASS(*)) has
9208            been added. Nonconstant character lengths are not yet
9209            supported.
9210     * [25]TS 29113:
9211          + Assumed types (TYPE(*)) are now supported.
9212          + Experimental support for assumed-rank arrays (dimension(..))
9213            has been added. Note that currently gfortran's own array
9214            descriptor is used, which is different from the one defined in
9215            TS29113, see [26]gfortran's header file or use the [27]Chasm
9216            Language Interoperability Tools.
9217
9218  Go
9219
9220     * GCC 4.8.2 provides a complete implementation of the Go 1.1.2
9221       release.
9222     * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1
9223       release. The library support is not quite complete.
9224     * Go has been tested on GNU/Linux and Solaris platforms for various
9225       processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may
9226       work on other platforms as well.
9227
9228New Targets and Target Specific Improvements
9229
9230  AArch64
9231
9232     * A new port has been added to support AArch64, the new 64-bit
9233       architecture from ARM. Note that this is a separate port from the
9234       existing 32-bit ARM port.
9235     * The port provides initial support for the Cortex-A53 and the
9236       Cortex-A57 processors with the command line options
9237       -mcpu=cortex-a53 and -mcpu=cortex-a57.
9238     * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769
9239       has been added and can be enabled by giving the
9240       -mfix-cortex-a53-835769 option. Alternatively it can be enabled by
9241       default by configuring GCC with the --enable-fix-cortex-a53-835769
9242       option.
9243
9244  ARM
9245
9246     * Initial support has been added for the AArch32 extensions defined
9247       in the ARMv8 architecture.
9248     * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs.
9249     * A new option, -mcpu=marvell-pj4, has been added to generate code
9250       for the Marvell PJ4 processor.
9251     * The compiler can now automatically generate the VFMA, VFMS, REVSH
9252       and REV16 instructions.
9253     * A new vectorizer cost model for Advanced SIMD configurations to
9254       improve the auto-vectorization strategies used.
9255     * The scheduler now takes into account the number of live registers
9256       to reduce the amount of spilling that can occur. This should
9257       improve code performance in large functions. The limit can be
9258       removed by using the option -fno-sched-pressure.
9259     * Improvements have been made to the Marvell iWMMX code generation
9260       and support for the iWMMX2 SIMD unit has been added. The option
9261       -mcpu=iwmmxt2 can be used to enable code generation for the latter.
9262     * A number of code generation improvements for Thumb2 to reduce code
9263       size when compiling for the M-profile processors.
9264     * The RTEMS (arm-rtems) port has been updated to use the EABI.
9265     * Code generation support for the old FPA and Maverick floating-point
9266       architectures has been removed. Ports that previously relied on
9267       these features have also been removed. This includes the targets:
9268          + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
9269          + arm*-*-elf (use arm*-*-eabi)
9270          + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
9271          + arm*-*-ecos-elf (no alternative)
9272          + arm*-*-freebsd (no alternative)
9273          + arm*-wince-pe* (no alternative).
9274
9275  AVR
9276
9277     * Support for the "Embedded C" fixed-point has been added. For
9278       details, see the [28]GCC wiki and the [29]user manual. The support
9279       is not complete.
9280     * A new print modifier %r for register operands in inline assembler
9281       is supported. It will print the raw register number without the
9282       register prefix 'r':
9283    /* Return the most significant byte of 'val', a 64-bit value.  */
9284
9285    unsigned char msb (long long val)
9286    {
9287      unsigned char c;
9288      __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val));
9289      return c;
9290    }
9291       The inline assembler in this example will generate code like
9292    mov r24, 8+7
9293       provided c is allocated to R24 and val is allocated to R8...R15.
9294       This works because the GNU assembler accepts plain register numbers
9295       without register prefix.
9296     * Static initializers with 3-byte symbols are supported now:
9297    extern const __memx char foo;
9298    const __memx void *pfoo = &foo;
9299       This requires at least Binutils 2.23.
9300
9301  IA-32/x86-64
9302
9303     * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with
9304       SSE extensions disabled. Since the x86-64 ABI requires 16 byte
9305       stack alignment, this is ABI incompatible and intended to be used
9306       in controlled environments where stack space is an important
9307       limitation. This option will lead to wrong code when functions
9308       compiled with 16 byte stack alignment (such as functions from a
9309       standard library) are called with misaligned stack. In this case,
9310       SSE instructions may lead to misaligned memory access traps. In
9311       addition, variable arguments will be handled incorrectly for 16
9312       byte aligned objects (including x87 long double and __int128),
9313       leading to wrong results. You must build all modules with
9314       -mpreferred-stack-boundary=3, including any libraries. This
9315       includes the system libraries and startup modules.
9316     * Support for the new Intel processor codename Broadwell with RDSEED,
9317       ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw,
9318       -mrdseed command-line options.
9319     * Support for the Intel RTM and HLE intrinsics, built-in functions
9320       and code generation is available via -mrtm and -mhle.
9321     * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets.
9322       Intrinsics and built-in functions are available via -mfxsr, -mxsave
9323       and -mxsaveopt respectively.
9324     * New -maddress-mode=[short|long] options for x32.
9325       -maddress-mode=short overrides default 64-bit addresses to 32-bit
9326       by emitting the 0x67 address-size override prefix. This is the
9327       default address mode for x32.
9328     * New built-in functions to detect run-time CPU type and ISA:
9329          + A built-in function __builtin_cpu_is has been added to detect
9330            if the run-time CPU is of a particular type. It returns a
9331            positive integer on a match and zero otherwise. It accepts one
9332            string literal argument, the CPU name. For example,
9333            __builtin_cpu_is("westmere") returns a positive integer if the
9334            run-time CPU is an Intel Core i7 Westmere processor. Please
9335            refer to the [30]user manual for the list of valid CPU names
9336            recognized.
9337          + A built-in function __builtin_cpu_supports has been added to
9338            detect if the run-time CPU supports a particular ISA feature.
9339            It returns a positive integer on a match and zero otherwise.
9340            It accepts one string literal argument, the ISA feature. For
9341            example, __builtin_cpu_supports("ssse3") returns a positive
9342            integer if the run-time CPU supports SSSE3 instructions.
9343            Please refer to the [31]user manual for the list of valid ISA
9344            names recognized.
9345       Caveat: If these built-in functions are called before any static
9346       constructors are invoked, like during IFUNC initialization, then
9347       the CPU detection initialization must be explicitly run using this
9348       newly provided built-in function, __builtin_cpu_init. The
9349       initialization needs to be done only once. For example, this is how
9350       the invocation would look like inside an IFUNC initializer:
9351    static void (*some_ifunc_resolver(void))(void)
9352    {
9353      __builtin_cpu_init();
9354      if (__builtin_cpu_is("amdfam10h") ...
9355      if (__builtin_cpu_supports("popcnt") ...
9356    }
9357
9358     * Function Multiversioning Support with G++:
9359       It is now possible to create multiple function versions each
9360       targeting a specific processor and/or ISA. Function versions have
9361       the same signature but different target attributes. For example,
9362       here is a program with function versions:
9363    __attribute__ ((target ("default")))
9364    int foo(void)
9365    {
9366      return 1;
9367    }
9368
9369    __attribute__ ((target ("sse4.2")))
9370    int foo(void)
9371    {
9372      return 2;
9373    }
9374
9375    int main (void)
9376    {
9377      int (*p) = &foo;
9378      assert ((*p)() == foo());
9379      return 0;
9380    }
9381
9382       Please refer to this [32]wiki for more information.
9383     * The x86 back end has been improved to allow option -fschedule-insns
9384       to work reliably. This option can be used to schedule instructions
9385       better and leads to improved performace in certain cases.
9386     * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437
9387       from the Mingw-w64 trunk.
9388     * Support for new AMD family 15h processors (Steamroller core) is now
9389       available through the -march=bdver3 and -mtune=bdver3 options.
9390     * Support for new AMD family 16h processors (Jaguar core) is now
9391       available through the -march=btver2 and -mtune=btver2 options.
9392
9393  FRV
9394
9395     * This target now supports the -fstack-usage command-line option.
9396
9397  MIPS
9398
9399     * GCC can now generate code specifically for the R4700, Broadcom XLP
9400       and MIPS 34kn processors. The associated -march options are
9401       -march=r4700, -march=xlp and -march=34kn respectively.
9402     * GCC now generates better DSP code for MIPS 74k cores thanks to
9403       further scheduling optimizations.
9404     * The MIPS port now supports the -fstack-check option.
9405     * GCC now passes the -mmcu and -mno-mcu options to the assembler.
9406     * Previous versions of GCC would silently accept -fpic and -fPIC for
9407       -mno-abicalls targets like mips*-elf. This combination was not
9408       intended or supported, and did not generate position-independent
9409       code. GCC 4.8 now reports an error when this combination is used.
9410
9411  PowerPC / PowerPC64 / RS6000
9412
9413     * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save,
9414       restore or update the VRSAVE register by default. The respective
9415       operating systems manage the VRSAVE register directly.
9416     * Large TOC support has been added for AIX through the command line
9417       option -mcmodel=large.
9418     * Native Thread-Local Storage support has been added for AIX.
9419     * VMX (Altivec) and VSX instruction sets now are enabled implicitly
9420       when targetting processors that support those hardware features on
9421       AIX 6.1 and above.
9422
9423  RX
9424
9425     * This target will now issue a warning message whenever multiple fast
9426       interrupt handlers are found in the same compilation unit. This
9427       feature can be turned off by the new
9428       -mno-warn-multiple-fast-interrupts command-line option.
9429
9430  S/390, System z
9431
9432     * Support for the IBM zEnterprise zEC12 processor has been added.
9433       When using the -march=zEC12 option, the compiler will generate code
9434       making use of the following new instructions:
9435          + load and trap instructions
9436          + 2 new compare and trap instructions
9437          + rotate and insert selected bits - without CC clobber
9438       The -mtune=zEC12 option enables zEC12 specific instruction
9439       scheduling without making use of new instructions.
9440     * Register pressure sensitive instruction scheduling is enabled by
9441       default.
9442     * The ifunc function attribute is enabled by default.
9443     * memcpy and memcmp invokations on big memory chunks or with run time
9444       lengths are not generated inline anymore when tuning for z10 or
9445       higher. The purpose is to make use of the IFUNC optimized versions
9446       in Glibc.
9447
9448  SH
9449
9450     * The default alignment settings have been reduced to be less
9451       aggressive. This results in more compact code for optimization
9452       levels other than -Os.
9453     * Improved support for the __atomic built-in functions:
9454          + A new option -matomic-model=model selects the model for the
9455            generated atomic sequences. The following models are
9456            supported:
9457
9458              soft-gusa
9459                      Software gUSA sequences (SH3* and SH4* only). On
9460                      SH4A targets this will now also partially utilize
9461                      the movco.l and movli.l instructions. This is the
9462                      default when the target is sh3*-*-linux* or
9463                      sh4*-*-linux*.
9464
9465              hard-llcs
9466                      Hardware movco.l / movli.l sequences (SH4A only).
9467
9468              soft-tcb
9469                      Software thread control block sequences.
9470
9471              soft-imask
9472                      Software interrupt flipping sequences (privileged
9473                      mode only). This is the default when the target is
9474                      sh1*-*-linux* or sh2*-*-linux*.
9475
9476              none
9477                      Generates function calls to the respective __atomic
9478                      built-in functions. This is the default for SH64
9479                      targets or when the target is not sh*-*-linux*.
9480
9481          + The option -msoft-atomic has been deprecated. It is now an
9482            alias for -matomic-model=soft-gusa.
9483          + A new option -mtas makes the compiler generate the tas.b
9484            instruction for the __atomic_test_and_set built-in function
9485            regardless of the selected atomic model.
9486          + The __sync functions in libgcc now reflect the selected atomic
9487            model when building the toolchain.
9488     * Added support for the mov.b and mov.w instructions with
9489       displacement addressing.
9490     * Added support for the SH2A instructions movu.b and movu.w.
9491     * Various improvements to code generated for integer arithmetic.
9492     * Improvements to conditional branches and code that involves the T
9493       bit. A new option -mzdcbranch tells the compiler to favor
9494       zero-displacement branches. This is enabled by default for SH4*
9495       targets.
9496     * The pref instruction will now be emitted by the __builtin_prefetch
9497       built-in function for SH3* targets.
9498     * The fmac instruction will now be emitted by the fmaf standard
9499       function and the __builtin_fmaf built-in function.
9500     * The -mfused-madd option has been deprecated in favor of the
9501       machine-independent -ffp-contract option. Notice that the fmac
9502       instruction will now be generated by default for expressions like a
9503       * b + c. This is due to the compiler default setting
9504       -ffp-contract=fast.
9505     * Added new options -mfsrra and -mfsca to allow the compiler using
9506       the fsrra and fsca instructions on targets other than SH4A (where
9507       they are already enabled by default).
9508     * Added support for the __builtin_bswap32 built-in function. It is
9509       now expanded as a sequence of swap.b and swap.w instructions
9510       instead of a library function call.
9511     * The behavior of the -mieee option has been fixed and the negative
9512       form -mno-ieee has been added to control the IEEE conformance of
9513       floating point comparisons. By default -mieee is now enabled and
9514       the option -ffinite-math-only implicitly sets -mno-ieee.
9515     * Added support for the built-in functions __builtin_thread_pointer
9516       and __builtin_set_thread_pointer. This assumes that GBR is used to
9517       hold the thread pointer of the current thread. Memory loads and
9518       stores relative to the address returned by __builtin_thread_pointer
9519       will now also utilize GBR based displacement address modes.
9520     * The -mdiv= option for targets other than SHmedia has been fixed and
9521       documented.
9522
9523  SPARC
9524
9525     * Added optimized instruction scheduling for Niagara4.
9526
9527  TILE-Gx
9528
9529     * Added support for the -mcmodel=MODEL command-line option. The
9530       models supported are small and large.
9531
9532  V850
9533
9534     * This target now supports the E3V5 architecture via the use of the
9535       new -mv850e3v5 command-line option. It also has experimental
9536       support for the e3v5 LOOP instruction which can be enabled via the
9537       new -mloop command-line option.
9538
9539  XStormy16
9540
9541     * This target now supports the -fstack-usage command-line option.
9542
9543Operating Systems
9544
9545  OpenBSD
9546
9547     * Support for OpenBSD/amd64 (x86_64-*-openbsd*) has been added and
9548       support for OpenBSD/i386 (i386-*-openbsd*) has been rejuvenated.
9549
9550  Windows (Cygwin)
9551
9552     * Executables are now linked against shared libgcc by default. The
9553       previous default was to link statically, which can still be done by
9554       explicitly specifying -static or static-libgcc on the command line.
9555       However it is strongly advised against, as it will cause problems
9556       for any application that makes use of DLLs compiled by GCC. It
9557       should be alright for a monolithic stand-alone application that
9558       only links against the Windows DLLs, but offers little or no
9559       benefit.
9560
9561GCC 4.8.1
9562
9563   This is the [33]list of problem reports (PRs) from GCC's bug tracking
9564   system that are known to be fixed in the 4.8.1 release. This list might
9565   not be complete (that is, it is possible that some PRs that have been
9566   fixed are not listed here).
9567
9568   The C++11 <chrono> std::chrono::system_clock and
9569   std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they
9570   both are now separate (never typedefs of each other), both use
9571   std::chrono::nanoseconds resolution, on most GNU/Linux configurations
9572   std::chrono::steady_clock is now finally monotonic, and both classes
9573   are mangled differently than in the previous GCC releases.
9574   std::chrono::system_clock::now() with std::chrono::microseconds resp.
9575   std::chrono::seconds resolution is still exported for backwards
9576   compatibility with default configured libstdc++. Note that libstdc++
9577   configured with --enable-libstdcxx-time= used to be ABI incompatible
9578   with default configured libstdc++ for those two classes and no ABI
9579   compatibility can be offered for those configurations, so any C++11
9580   code that uses those classes and has been compiled and linked against
9581   libstdc++ configured with the non-default --enable-libstdcxx-time=
9582   configuration option needs to be recompiled.
9583
9584GCC 4.8.2
9585
9586   This is the [34]list of problem reports (PRs) from GCC's bug tracking
9587   system that are known to be fixed in the 4.8.2 release. This list might
9588   not be complete (that is, it is possible that some PRs that have been
9589   fixed are not listed here).
9590
9591GCC 4.8.3
9592
9593   This is the [35]list of problem reports (PRs) from GCC's bug tracking
9594   system that are known to be fixed in the 4.8.3 release. This list might
9595   not be complete (that is, it is possible that some PRs that have been
9596   fixed are not listed here).
9597
9598   Support for the new powerpc64le-linux platform has been added. It
9599   defaults to generating code that conforms to the ELFV2 ABI.
9600
9601GCC 4.8.4
9602
9603   This is the [36]list of problem reports (PRs) from GCC's bug tracking
9604   system that are known to be fixed in the 4.8.4 release. This list might
9605   not be complete (that is, it is possible that some PRs that have been
9606   fixed are not listed here).
9607
9608GCC 4.8.5
9609
9610   This is the [37]list of problem reports (PRs) from GCC's bug tracking
9611   system that are known to be fixed in the 4.8.5 release. This list might
9612   not be complete (that is, it is possible that some PRs that have been
9613   fixed are not listed here).
9614
9615
9616    For questions related to the use of GCC, please consult these web
9617    pages and the [38]GCC manuals. If that fails, the
9618    [39]gcc-help@gcc.gnu.org mailing list might help. Comments on these
9619    web pages and the development of GCC are welcome on our developer
9620    list at [40]gcc@gcc.gnu.org. All of [41]our lists have public
9621    archives.
9622
9623   Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and
9624   distribution of this entire article is permitted in any medium,
9625   provided this notice is preserved.
9626
9627   These pages are [43]maintained by the GCC team. Last modified
9628   2022-11-05.
9629
9630References
9631
9632   1. https://gcc.gnu.org/wiki/cxx-conversion
9633   2. ftp://gcc.gnu.org/pub/gcc/infrastructure/
9634   3. http://www.nongnu.org/avr-libc/
9635   4. https://gcc.gnu.org/PR54461
9636   5. https://gcc.gnu.org/gcc-4.8/porting_to.html
9637   6. https://github.com/google/sanitizers
9638   7. https://code.google.com/archive/p/data-race-test/wikis/ThreadSanitizer.wiki
9639   8. https://lkml.org/lkml/2006/11/28/239
9640   9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
9641  10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
9642  11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
9643  12. https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf
9644  13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
9645  14. https://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html
9646  15. http://gcc.gnu.org/projects/cxx1y.html
9647  16. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011
9648  17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html
9649  18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
9650  19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
9651  20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
9652  21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
9653  22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
9654  23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html
9655  24. https://gcc.gnu.org/wiki/Fortran2003Status
9656  25. https://gcc.gnu.org/wiki/TS29113Status
9657  26. https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=libgfortran/libgfortran.h
9658  27. https://chasm-interop.sourceforge.net/
9659  28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support
9660  29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html
9661  30. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html
9662  31. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html
9663  32. https://gcc.gnu.org/wiki/FunctionMultiVersioning
9664  33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1
9665  34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2
9666  35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3
9667  36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4
9668  37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5
9669  38. https://gcc.gnu.org/onlinedocs/
9670  39. mailto:gcc-help@gcc.gnu.org
9671  40. mailto:gcc@gcc.gnu.org
9672  41. https://gcc.gnu.org/lists.html
9673  42. https://www.fsf.org/
9674  43. https://gcc.gnu.org/about.html
9675======================================================================
9676http://gcc.gnu.org/gcc-4.7/index.html
9677
9678                             GCC 4.7 Release Series
9679
9680   (This release series is no longer supported.)
9681
9682   June 12, 2014
9683
9684   The [1]GNU project and the GCC developers are pleased to announce the
9685   release of GCC 4.7.4.
9686
9687   This release is a bug-fix release, containing fixes for regressions in
9688   GCC 4.7.3 relative to previous releases of GCC.
9689
9690Release History
9691
9692   GCC 4.7.4
9693          June 12, 2014 ([2]changes, [3]documentation)
9694
9695   GCC 4.7.3
9696          April 11, 2013 ([4]changes, [5]documentation)
9697
9698   GCC 4.7.2
9699          September 20, 2012 ([6]changes, [7]documentation)
9700
9701   GCC 4.7.1
9702          June 14, 2012 ([8]changes, [9]documentation)
9703
9704   GCC 4.7.0
9705          March 22, 2012 ([10]changes, [11]documentation)
9706
9707References and Acknowledgements
9708
9709   GCC used to stand for the GNU C Compiler, but since the compiler
9710   supports several other languages aside from C, it now stands for the
9711   GNU Compiler Collection.
9712
9713   A list of [12]successful builds is updated as new information becomes
9714   available.
9715
9716   The GCC developers would like to thank the numerous people that have
9717   contributed new features, improvements, bug fixes, and other changes as
9718   well as test results to GCC. This [13]amazing group of volunteers is
9719   what makes GCC successful.
9720
9721   For additional information about GCC please refer to the [14]GCC
9722   project web site or contact the [15]GCC development mailing list.
9723
9724   To obtain GCC please use [16]our mirror sites or [17]our version
9725   control system.
9726
9727
9728    For questions related to the use of GCC, please consult these web
9729    pages and the [18]GCC manuals. If that fails, the
9730    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
9731    web pages and the development of GCC are welcome on our developer
9732    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
9733    archives.
9734
9735   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
9736   distribution of this entire article is permitted in any medium,
9737   provided this notice is preserved.
9738
9739   These pages are [23]maintained by the GCC team. Last modified
9740   2022-10-26.
9741
9742References
9743
9744   1. http://www.gnu.org/
9745   2. http://gcc.gnu.org/gcc-4.7/changes.html
9746   3. https://gcc.gnu.org/onlinedocs/4.7.4/
9747   4. http://gcc.gnu.org/gcc-4.7/changes.html
9748   5. https://gcc.gnu.org/onlinedocs/4.7.3/
9749   6. http://gcc.gnu.org/gcc-4.7/changes.html
9750   7. https://gcc.gnu.org/onlinedocs/4.7.2/
9751   8. http://gcc.gnu.org/gcc-4.7/changes.html
9752   9. https://gcc.gnu.org/onlinedocs/4.7.1/
9753  10. http://gcc.gnu.org/gcc-4.7/changes.html
9754  11. https://gcc.gnu.org/onlinedocs/4.7.0/
9755  12. http://gcc.gnu.org/gcc-4.7/buildstat.html
9756  13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
9757  14. http://gcc.gnu.org/index.html
9758  15. mailto:gcc@gcc.gnu.org
9759  16. http://gcc.gnu.org/mirrors.html
9760  17. http://gcc.gnu.org/git.html
9761  18. https://gcc.gnu.org/onlinedocs/
9762  19. mailto:gcc-help@gcc.gnu.org
9763  20. mailto:gcc@gcc.gnu.org
9764  21. https://gcc.gnu.org/lists.html
9765  22. https://www.fsf.org/
9766  23. https://gcc.gnu.org/about.html
9767======================================================================
9768http://gcc.gnu.org/gcc-4.7/changes.html
9769
9770                             GCC 4.7 Release Series
9771                        Changes, New Features, and Fixes
9772
9773Caveats
9774
9775     * The -fconserve-space flag has been deprecated. The flag had no
9776       effect for most targets: only targets without a global .bss section
9777       and without support for switchable sections. Furthermore, the flag
9778       only had an effect for G++, where it could result in wrong
9779       semantics (please refer to the GCC manual for further details). The
9780       flag will be removed in GCC 4.8
9781     * Support for a number of older systems and recently unmaintained or
9782       untested target ports of GCC has been declared obsolete in GCC 4.7.
9783       Unless there is activity to revive them, the next release of GCC
9784       will have their sources permanently removed.
9785       All GCC ports for the following processor architectures have been
9786       declared obsolete:
9787          + picoChip (picochip-*)
9788       The following ports for individual systems on particular
9789       architectures have been obsoleted:
9790          + IRIX 6.5 (mips-sgi-irix6.5)
9791          + MIPS OpenBSD (mips*-*-openbsd*)
9792          + Solaris 8 (*-*-solaris2.8). Details can be found in the
9793            [1]announcement.
9794          + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*)
9795     * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A,
9796       ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by
9797       default, which for some sources generates code that accesses memory
9798       on unaligned addresses. This requires the kernel of those systems
9799       to enable such accesses (controlled by CP15 register c1, refer to
9800       ARM documentation). Alternatively, or for compatibility with
9801       kernels where unaligned accesses are not supported, all code has to
9802       be compiled with -mno-unaligned-access. Upstream Linux kernel
9803       releases have automatically and unconditionally supported unaligned
9804       accesses as emitted by GCC due to this option being active since
9805       version 2.6.28.
9806     * Support on ARM for the legacy floating-point accelerator (FPA) and
9807       the mixed-endian floating-point format that it used has been
9808       obsoleted. The ports that still use this format have been obsoleted
9809       as well. Many legacy ARM ports already provide an alternative that
9810       uses the VFP floating-point format. The obsolete ports will be
9811       deleted in the next release.
9812       The obsolete ports with alternatives are:
9813          + arm*-*-rtems (use arm*-*-rtemseabi)
9814          + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
9815          + arm*-*-elf (use arm*-*-eabi)
9816          + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
9817       Note, however, that these alternatives are not binary compatible
9818       with their legacy counterparts (although some can support running
9819       legacy applications).
9820       The obsolete ports that currently lack a modern alternative are:
9821          + arm*-*-ecos-elf
9822          + arm*-*-freebsd
9823          + arm*-wince-pe*
9824       New ports that support more recent versions of the architecture are
9825       welcome.
9826     * Support for the Maverick co-processor on ARM has been obsoleted.
9827       Code to support it will be deleted in the next release.
9828     * Support has been removed for Unix International threads on Solaris
9829       2, so the --enable-threads=solaris configure option and the
9830       -threads compiler option don't work any longer.
9831     * Support has been removed for the Solaris BSD Compatibility Package,
9832       which lives in /usr/ucbinclude and /usr/ucblib. It has been removed
9833       from Solaris 11, and was only intended as a migration aid from
9834       SunOS 4 to SunOS 5. The -compat-bsd compiler option is not
9835       recognized any longer.
9836     * The AVR port's libgcc has been improved and its multilib structure
9837       has been enhanced. As a result, all objects contributing to an
9838       application must either be compiled with GCC versions up to 4.6.x
9839       or with GCC versions 4.7.1 or later. If the compiler is used with
9840       AVR Libc, you need a version that supports the new layout, i.e.
9841       implements [2]#35407.
9842     * The AVR port's -mshort-calls command-line option has been
9843       deprecated. It will be removed in the GCC 4.8 release. See -mrelax
9844       for a replacement.
9845     * The AVR port only references startup code that clears .bss and the
9846       common section resp. initializes the .data and .rodata section
9847       provided respective sections (or subsections thereof) are not
9848       empty, see [3]PR18145. Applications that put all static storage
9849       objects into non-standard sections and / or define all static
9850       storage objects in assembler modules, must reference __do_clear_bss
9851       resp. __do_copy_data by hand or undefine the symbol(s) by means of
9852       -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data.
9853     * The ARM port's -mwords-little-endian option has been deprecated. It
9854       will be removed in a future release.
9855     * Support has been removed for the NetWare x86 configuration
9856       obsoleted in GCC 4.6.
9857     * It is no longer possible to use the "l" constraint in MIPS16 asm
9858       statements.
9859     * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard
9860       library which affected the ABI in C++11 mode: a data member was
9861       added to std::list changing its size and altering the definitions
9862       of some member functions, and std::pair's move constructor was
9863       non-trivial which altered the calling convention for functions with
9864       std::pair arguments or return types. The ABI incompatibilities have
9865       been fixed for GCC version 4.7.2 but as a result C++11 code
9866       compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11
9867       code compiled with different GCC versions and with C++98/C++03 code
9868       compiled with any version.
9869     * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
9870       rules for the layout of vectors that could lead to wrong code being
9871       generated. Vectors larger than 8 bytes in size are now by default
9872       aligned to an 8-byte boundary. This is an ABI change: code that
9873       makes explicit use of vector types may be incompatible with binary
9874       objects built with older versions of GCC. Auto-vectorized code is
9875       not affected by this change. (This change affects GCC versions
9876       4.7.2 and later.)
9877     * More information on porting to GCC 4.7 from previous versions of
9878       GCC can be found in the [4]porting guide for this release.
9879
9880General Optimizer Improvements
9881
9882     * Support for a new parameter --param case-values-threshold=n was
9883       added to allow users to control the cutoff between doing switch
9884       statements as a series of if statements and using a jump table.
9885     * Link-time optimization (LTO) improvements:
9886          + Improved scalability and reduced memory usage. Link time
9887            optimization of Firefox now requires 3GB of RAM on a 64-bit
9888            system, while over 8GB was needed previously. Linking time has
9889            been improved, too. The serial stage of linking Firefox has
9890            been sped up by about a factor of 10.
9891          + Reduced size of object files and temporary storage used during
9892            linking.
9893          + Streaming performance (both outbound and inbound) has been
9894            improved.
9895          + ld -r is now supported with LTO.
9896          + Several bug fixes, especially in symbol table handling and
9897            merging.
9898     * Interprocedural optimization improvements:
9899          + Heuristics now take into account that after inlining code will
9900            be optimized out because of known values (or properties) of
9901            function parameters. For example:
9902void foo(int a)
9903{
9904  if (a > 10)
9905    ... huge code ...
9906}
9907void bar (void)
9908{
9909  foo (0);
9910}
9911
9912            The call of foo will be inlined into bar even when optimizing
9913            for code size. Constructs based on __builtin_constant_p are
9914            now understood by the inliner and code size estimates are
9915            evaluated a lot more realistically.
9916          + The representation of C++ virtual thunks and aliases (both
9917            implicit and defined via the alias attribute) has been
9918            re-engineered. Aliases no longer pose optimization barriers
9919            and calls to an alias can be inlined and otherwise optimized.
9920          + The inter-procedural constant propagation pass has been
9921            rewritten. It now performs generic function specialization.
9922            For example when compiling the following:
9923void foo(bool flag)
9924{
9925  if (flag)
9926    ... do something ...
9927  else
9928    ... do something else ...
9929}
9930void bar (void)
9931{
9932  foo (false);
9933  foo (true);
9934  foo (false);
9935  foo (true);
9936  foo (false);
9937  foo (true);
9938}
9939
9940            GCC will now produce two copies of foo. One with flag being
9941            true, while other with flag being false. This leads to
9942            performance improvements previously possible only by inlining
9943            all calls. Cloning causes a lot less code size growth.
9944     * A string length optimization pass has been added. It attempts to
9945       track string lengths and optimize various standard C string
9946       functions like strlen, strchr, strcpy, strcat, stpcpy and their
9947       _FORTIFY_SOURCE counterparts into faster alternatives. This pass is
9948       enabled by default at -O2 or above, unless optimizing for size, and
9949       can be disabled by the -fno-optimize-strlen option. The pass can
9950       e.g. optimize
9951char *bar (const char *a)
9952{
9953  size_t l = strlen (a) + 2;
9954  char *p = malloc (l); if (p == NULL) return p;
9955  strcpy (p, a); strcat (p, "/"); return p;
9956}
9957
9958       into:
9959char *bar (const char *a)
9960{
9961  size_t tmp = strlen (a);
9962  char *p = malloc (tmp + 2); if (p == NULL) return p;
9963  memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p;
9964}
9965
9966       or for hosted compilations where stpcpy is available in the runtime
9967       and headers provide its prototype, e.g.
9968void foo (char *a, const char *b, const char *c, const char *d)
9969{
9970  strcpy (a, b); strcat (a, c); strcat (a, d);
9971}
9972
9973       can be optimized into:
9974void foo (char *a, const char *b, const char *c, const char *d)
9975{
9976  strcpy (stpcpy (stpcpy (a, b), c), d);
9977}
9978
9979New Languages and Language specific improvements
9980
9981     * Version 3.1 of the OpenMP specification is now supported for the C,
9982       C++, and Fortran compilers.
9983
9984  Ada
9985
9986     * The command-line option -feliminate-unused-debug-types has been
9987       re-enabled by default, as it is for the other languages, leading to
9988       a reduction in debug info size of 12.5% and more for relevant
9989       cases, as well as to a small compilation speedup.
9990
9991  C family
9992
9993     * A new built-in, __builtin_assume_aligned, has been added, through
9994       which the compiler can be hinted about pointer alignment and can
9995       use it to improve generated code.
9996     * A new warning option -Wunused-local-typedefs was added for C, C++,
9997       Objective-C and Objective-C++. This warning diagnoses typedefs
9998       locally defined in a function, and otherwise not used.
9999     * A new experimental command-line option -ftrack-macro-expansion was
10000       added for C, C++, Objective-C, Objective-C++ and Fortran. It allows
10001       the compiler to emit diagnostic about the current macro expansion
10002       stack when a compilation error occurs in a macro expansion.
10003     * Experimental support for transactional memory has been added. It
10004       includes support in the compiler, as well as a supporting runtime
10005       library called libitm. To compile code with transactional memory
10006       constructs, use the -fgnu-tm option.
10007       Support is currently available for Alpha, ARM, PowerPC, SH, SPARC,
10008       and 32-bit/64-bit x86 platforms.
10009       For more details on transactional memory see [5]the GCC WiKi.
10010     * Support for atomic operations specifying the C++11/C11 memory model
10011       has been added. These new __atomic routines replace the existing
10012       __sync built-in routines.
10013       Atomic support is also available for memory blocks. Lock-free
10014       instructions will be used if a memory block is the same size and
10015       alignment as a supported integer type. Atomic operations which do
10016       not have lock-free support are left as function calls. A set of
10017       library functions is available on the GCC atomic wiki in the
10018       "External Atomics Library" section.
10019       For more details on the memory models and features, see the
10020       [6]atomic wiki.
10021     * When a binary operation is performed on vector types and one of the
10022       operands is a uniform vector, it is possible to replace the vector
10023       with the generating element. For example:
10024typedef int v4si __attribute__ ((vector_size (16)));
10025v4si res, a = {1,2,3,4};
10026int x;
10027
10028res = 2 + a;  /* means {2,2,2,2} + a  */
10029res = a - x;  /* means a - {x,x,x,x}  */
10030
10031  C
10032
10033     * There is support for some more features from the C11 revision of
10034       the ISO C standard. GCC now accepts the options -std=c11 and
10035       -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x.
10036          + Unicode strings (previously supported only with options such
10037            as -std=gnu11, now supported with -std=c11), and the
10038            predefined macros __STDC_UTF_16__ and __STDC_UTF_32__.
10039          + Nonreturning functions (_Noreturn and <stdnoreturn.h>).
10040          + Alignment support (_Alignas, _Alignof, max_align_t,
10041            <stdalign.h>).
10042          + A built-in function __builtin_complex is provided to support C
10043            library implementation of the CMPLX family of macros.
10044
10045  C++
10046
10047     * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat
10048       options, which are equivalent to -std=c++0x, -std=gnu++0x, and
10049       -Wc++0x-compat, respectively.
10050     * G++ now implements [7]C++11 extended friend syntax:
10051
10052template<class W>
10053class Q
10054{
10055  static const int I = 2;
10056public:
10057  friend W;
10058};
10059
10060struct B
10061{
10062  int ar[Q<B>::I];
10063};
10064
10065     * Thanks to Ville Voutilainen, G++ now implements [8]C++11 explicit
10066       override control.
10067
10068struct B {
10069  virtual void f() const final;
10070  virtual void f(int);
10071};
10072
10073struct D : B {
10074  void f() const;            // error: D::f attempts to override final B::f
10075  void f(long) override;     // error: doesn't override anything
10076  void f(int) override;      // ok
10077};
10078
10079struct E final { };
10080struct F: E { }; // error: deriving from final class
10081
10082     * G++ now implements [9]C++11 non-static data member initializers.
10083
10084struct A {
10085  int i = 42;
10086} a; // initializes a.i to 42
10087
10088     * Thanks to Ed Smith-Rowland, G++ now implements [10]C++11
10089       user-defined literals.
10090
10091// Not actually a good approximation.  :)
10092constexpr long double operator"" _degrees (long double d) { return d * 0.0175; }
10093long double pi = 180.0_degrees;
10094
10095     * G++ now implements [11]C++11 alias-declarations.
10096
10097template <class T> using Ptr = T*;
10098Ptr<int> ip;  // decltype(ip) is int*
10099
10100     * Thanks to Ville Voutilainen and Pedro Lamar�o, G++ now implements
10101       [12]C++11 delegating constructors.
10102
10103struct A {
10104  A(int);
10105  A(): A(42) { } // delegate to the A(int) constructor
10106};
10107
10108     * G++ now fully implements C++11 atomic classes rather than just
10109       integer derived classes.
10110
10111class POD {
10112  int a;
10113  int b;
10114};
10115std::atomic<POD> my_atomic_POD;
10116
10117     * G++ now sets the predefined macro __cplusplus to the correct value,
10118       199711L for C++98/03, and 201103L for C++11.
10119     * G++ now correctly implements the two-phase lookup rules such that
10120       an unqualified name used in a template must have an appropriate
10121       declaration found either in scope at the point of definition of the
10122       template or by argument-dependent lookup at the point of
10123       instantiation. As a result, code that relies on a second
10124       unqualified lookup at the point of instantiation to find functions
10125       declared after the template or in dependent bases will be rejected.
10126       The compiler will suggest ways to fix affected code, and using the
10127       -fpermissive compiler flag will allow the code to compile with a
10128       warning.
10129
10130template <class T>
10131void f() { g(T()); } // error, g(int) not found by argument-dependent lookup
10132void g(int) { } // fix by moving this declaration before the declaration of f
10133
10134template <class T>
10135struct A: T {
10136  // error, B::g(B) not found by argument-dependent lookup
10137  void f() { g(T()); } // fix by using this->g or A::g
10138};
10139
10140struct B { void g(B); };
10141
10142int main()
10143{
10144  f<int>();
10145  A<B>().f();
10146}
10147
10148     * G++ now properly re-uses stack space allocated for temporary
10149       objects when their lifetime ends, which can significantly lower
10150       stack consumption for some C++ functions. As a result of this, some
10151       code with undefined behavior will now break:
10152
10153const int &f(const int &i) { return i; }
10154....
10155const int &x = f(1);
10156const int &y = f(2);
10157
10158       Here, x refers to the temporary allocated to hold the 1 argument,
10159       which only lives until the end of the initialization; it
10160       immediately becomes a dangling reference. So the next statement
10161       re-uses the stack slot to hold the 2 argument, and users of x get
10162       that value instead.
10163       Note that this should not cause any change of behavior for
10164       temporaries of types with non-trivial destructors, as they are
10165       already destroyed at end of full-expression; the change is that now
10166       the storage is released as well.
10167     * A new command-line option -Wdelete-non-virtual-dtor has been added
10168       to warn when delete is used to destroy an instance of a class which
10169       has virtual functions and non-virtual destructor. It is unsafe to
10170       delete an instance of a derived class through a pointer to a base
10171       class if the base class does not have a virtual destructor. This
10172       warning is enabled by -Wall.
10173     * A new command-line option -Wzero-as-null-pointer-constant has been
10174       added to warn when a literal '0' is used as null pointer constant.
10175       It can be useful to facilitate the conversion to nullptr in C++11.
10176     * As per C++98, access-declarations are now deprecated by G++.
10177       Using-declarations are to be used instead. Furthermore, some
10178       efforts have been made to improve the support of class scope
10179       using-declarations. In particular, using-declarations referring to
10180       a dependent type now work as expected ([13]bug c++/14258).
10181     * The ELF symbol visibility of a template instantiation is now
10182       properly constrained by the visibility of its template arguments
10183       ([14]bug c++/35688).
10184
10185    Runtime Library (libstdc++)
10186
10187     * [15]Improved experimental support for the new ISO C++ standard,
10188       C++11, including:
10189          + using noexcept in most of the library;
10190          + implementations of pointer_traits, allocator_traits and
10191            scoped_allocator_adaptor;
10192          + uses-allocator construction for tuple;
10193          + vector meets the allocator-aware container requirements;
10194          + replacing monotonic_clock with steady_clock;
10195          + enabling the thread support library on most POSIX targets;
10196          + many small improvements to conform to the FDIS.
10197     * Added --enable-clocale=newlib configure option.
10198     * Debug Mode iterators for unordered associative containers.
10199     * Avoid polluting the global namespace and do not include <unistd.h>.
10200
10201  Fortran
10202
10203     * The compile flag [16]-fstack-arrays has been added, which causes
10204       all local arrays to be put on stack memory. For some programs this
10205       will improve the performance significantly. If your program uses
10206       very large local arrays, it is possible that you will have to
10207       extend your runtime limits for stack memory.
10208     * The [17]-Ofast flag now also implies [18]-fno-protect-parens and
10209       [19]-fstack-arrays.
10210     * Front-end optimizations can now be selected by the
10211       [20]-ffrontend-optimize option and deselected by the
10212       -fno-frontend-optimize option.
10213     * When front-end optimization removes a function call,
10214       [21]-Wfunction-elimination warns about that.
10215     * When performing front-end-optimization, the
10216       [22]-faggressive-function-elimination option allows the removal of
10217       duplicate function calls even for impure functions.
10218     * The flag [23]-Wreal-q-constant has been added, which warns if
10219       floating-point literals have been specified using q (such as
10220       1.0q0); the q marker is now supported as a vendor extension to
10221       denote quad precision (REAL(16) or, if not available, REAL(10)).
10222       Consider using a kind parameter (such as in 1.0_qp) instead, which
10223       can be obtained via [24]SELECTED_REAL_KIND.
10224     * The GFORTRAN_USE_STDERR environment variable has been removed. GNU
10225       Fortran now always prints error messages to standard error. If you
10226       wish to redirect standard error, please consult the manual for your
10227       OS, shell, batch environment etc. as appropriate.
10228     * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment
10229       variable have been removed. When encountering a serious error,
10230       gfortran will now always abort the program. Whether a core dump is
10231       generated depends on the user environment settings; see the ulimit
10232       -c setting for POSIX shells, limit coredumpsize for C shells, and
10233       the [25]WER user-mode dumps settings on Windows.
10234     * The [26]-fbacktrace option is now enabled by default. When
10235       encountering a fatal error, gfortran will attempt to print a
10236       backtrace to standard error before aborting. It can be disabled
10237       with -fno-backtrace. Note: On POSIX targets with the addr2line
10238       utility from GNU binutils, GNU Fortran can print a backtrace with
10239       function name, file name, line number information in addition to
10240       the addresses; otherwise only the addresses are printed.
10241     * [27]Fortran 2003:
10242          + Generic interface names which have the same name as derived
10243            types are now supported, which allows to write constructor
10244            functions. Note that Fortran does not support static
10245            constructor functions; only default initialization or an
10246            explicit structure-constructor initialization are available.
10247          + [28]Polymorphic (class) arrays are now supported.
10248     * [29]Fortran 2008:
10249          + Support for the DO CONCURRENT construct has been added, which
10250            allows the user to specify that individual loop iterations
10251            have no interdependencies.
10252          + [30]Coarrays: Full single-image support except for polymorphic
10253            coarrays. Additionally, preliminary support for multiple
10254            images via an MPI-based [31]coarray communication library has
10255            been added. Note: The library version is not yet usable as
10256            remote coarray access is not yet possible.
10257     * [32]TS 29113:
10258          + New flag [33]-std=f2008ts permits programs that are expected
10259            to conform to the Fortran 2008 standard and the draft
10260            Technical Specification (TS) 29113 on Further Interoperability
10261            of Fortran with C.
10262          + The OPTIONAL attribute is now allowed for dummy arguments of
10263            BIND(C) procedures.
10264          + The RANK intrinsic has been added.
10265          + The implementation of the ASYNCHRONOUS attribute in GCC is
10266            compatible with the candidate draft of TS 29113 (since GCC
10267            4.6).
10268
10269  Go
10270
10271     * GCC 4.7 implements the [34]Go 1 language standard. The library
10272       support in 4.7.0 is not quite complete, due to release timing.
10273       Release 4.7.1 includes complete support for Go 1. The Go library is
10274       from the Go 1.0.1 release.
10275     * Go has been tested on GNU/Linux and Solaris platforms. It may work
10276       on other platforms as well.
10277
10278New Targets and Target Specific Improvements
10279
10280  ARM
10281
10282     * GCC now supports the Cortex-A7 processor implementing the v7-a
10283       version of the architecture using the option -mcpu=cortex-a7.
10284     * The default vector size in auto-vectorization for NEON is now 128
10285       bits. If vectorization fails thusly, the vectorizer tries again
10286       with 64-bit vectors.
10287     * A new option -mvectorize-with-neon-double was added to allow users
10288       to change the vector size to 64 bits.
10289
10290  AVR
10291
10292     * GCC now supports the XMEGA architecture. This requires GNU binutils
10293       2.22 or later.
10294     * Support for the [35]named address spaces __flash, __flash1, ...,
10295       __flash5 and __memx has been added. These address spaces locate
10296       read-only data in flash memory and allow reading from flash memory
10297       by means of ordinary C code, i.e. without the need of (inline)
10298       assembler code:
10299
10300const __flash int values[] = { 42, 31 };
10301
10302int add_values (const __flash int *p, int i)
10303{
10304    return values[i] + *p;
10305}
10306
10307     * Support has been added for the AVR-specific configure option
10308       --with-avrlibc=yes in order to arrange for better integration of
10309       [36]AVR-Libc. This configure option is supported in avr-gcc 4.7.2
10310       and newer and will only take effect in non-RTEMS configurations. If
10311       avr-gcc is configured for RTEMS, the option will be ignored which
10312       is the same as specifying --with-avrlibc=no. See [37]PR54461 for
10313       more technical details.
10314     * Support for AVR-specific [38]built-in functions has been added.
10315     * Support has been added for the signed and unsigned 24-bit scalar
10316       integer types __int24 and __uint24.
10317     * New command-line options -maccumulate-args, -mbranch-cost=cost and
10318       -mstrict-X were added to allow better fine-tuning of code
10319       optimization.
10320     * The command-line option -fdata-sections now also takes affect on
10321       the section names of variables with the progmem attribute.
10322     * A new inline assembler print modifier %i to print a RAM address as
10323       I/O address has been added:
10324
10325#include <avr/io.h> /* Port Definitions from AVR-LibC */
10326
10327void set_portb (uint8_t value)
10328{
10329    asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory");
10330}
10331
10332       The offset between an I/O address and the RAM address for that I/O
10333       location is device-specific. This offset is taken into account when
10334       printing a RAM address with the %i modifier so that the address is
10335       suitable to be used as operand in an I/O command. The address must
10336       be a constant integer known at compile time.
10337     * The inline assembler constraint "R" to represent integers in the
10338       range -6 ... 5 has been removed without replacement.
10339     * Many optimizations to:
10340          + 64-bit integer arithmetic
10341          + Widening multiplication
10342          + Integer division by a constant
10343          + Avoid constant reloading in multi-byte instructions.
10344          + Micro-optimizations for special instruction sequences.
10345          + Generic built-in functions like __builtin_ffs*,
10346            __builtin_clz*, etc.
10347          + If-else decision trees generated by switch instructions
10348          + Merging of data located in flash memory
10349          + New libgcc variants for devices with 8-bit wide stack pointer
10350          + ...
10351     * Better documentation:
10352          + Handling of EIND and indirect jumps on devices with more than
10353            128 KiB of program memory.
10354          + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function
10355            registers.
10356          + Function attributes OS_main and OS_task.
10357          + AVR-specific built-in macros.
10358
10359  C6X
10360
10361     * Support has been added for the Texas Instruments C6X family of
10362       processors.
10363
10364  CR16
10365
10366     * Support has been added for National Semiconductor's CR16
10367       architecture.
10368
10369  Epiphany
10370
10371     * Support has been added for Adapteva's Epiphany architecture.
10372
10373  IA-32/x86-64
10374
10375     * Support for Intel AVX2 intrinsics, built-in functions and code
10376       generation is available via -mavx2.
10377     * Support for Intel BMI2 intrinsics, built-in functions and code
10378       generation is available via -mbmi2.
10379     * Implementation and automatic generation of __builtin_clz* using the
10380       lzcnt instruction is available via -mlzcnt.
10381     * Support for Intel FMA3 intrinsics and code generation is available
10382       via -mfma.
10383     * A new -mfsgsbase command-line option is available that makes GCC
10384       generate new segment register read/write instructions through
10385       dedicated built-ins.
10386     * Support for the new Intel rdrnd instruction is available via
10387       -mrdrnd.
10388     * Two additional AVX vector conversion instructions are available via
10389       -mf16c.
10390     * Support for new Intel processor codename IvyBridge with RDRND,
10391       FSGSBASE and F16C is available through -march=core-avx-i.
10392     * Support for the new Intel processor codename Haswell with AVX2,
10393       FMA, BMI, BMI2, LZCNT is available through -march=core-avx2.
10394     * Support for new AMD family 15h processors (Piledriver core) is now
10395       available through -march=bdver2 and -mtune=bdver2 options.
10396     * Support for [39]the x32 psABI is now available through the -mx32
10397       option.
10398     * Windows mingw targets are using the -mms-bitfields option by
10399       default.
10400     * Windows x86 targets are using the __thiscall calling convention for
10401       C++ class-member functions.
10402     * Support for the configure option --with-threads=posix for Windows
10403       mingw targets.
10404
10405  MIPS
10406
10407     * GCC now supports thread-local storage (TLS) for MIPS16. This
10408       requires GNU binutils 2.22 or later.
10409     * GCC can now generate code specifically for the Cavium Octeon+ and
10410       Octeon2 processors. The associated command-line options are
10411       -march=octeon+ and -march=octeon2 respectively. Both options
10412       require GNU binutils 2.22 or later.
10413     * GCC can now work around certain 24k errata, under the control of
10414       the command-line option -mfix-24k. These workarounds require GNU
10415       binutils 2.20 or later.
10416     * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build
10417       n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux
10418       toolchain that generates 32-bit code by default. Use the
10419       configure-time option --enable-targets=all to select these extra
10420       multilibs.
10421     * Passing -fno-delayed-branch now also stops the assembler from
10422       automatically filling delay slots.
10423
10424  PowerPC/PowerPC64
10425
10426     * Vectors of type vector long long or vector long are passed and
10427       returned using the same method as other vectors with the VSX
10428       instruction set. Previously GCC did not adhere to the ABI for
10429       128-bit vectors with 64-bit integer base types (PR 48857). This
10430       will also be fixed in the GCC 4.6.1 and 4.5.4 releases.
10431     * A new option -mno-pointers-to-nested-functions was added to allow
10432       AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify
10433       that the compiler should not load up the chain register (r11)
10434       before calling a function through a pointer. If you use this
10435       option, you cannot call nested functions through a pointer, or call
10436       other languages that might use the static chain.
10437     * A new option msave-toc-indirect was added to allow AIX
10438       32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we
10439       save the TOC in the prologue for indirect calls or generate the
10440       save inline. This can speed up some programs that call through a
10441       function pointer a lot, but it can slow down other functions that
10442       only call through a function pointer in exceptional cases.
10443     * The PowerPC port will now enable machine-specific built-in
10444       functions when the user switches the target machine using the
10445       #pragma GCC target or __attribute__ ((__target__ ("target"))) code
10446       sequences. In addition, the target macros are updated. However, due
10447       to the way the -save-temps switch is implemented, you won't see the
10448       effect of these additional macros being defined in preprocessor
10449       output.
10450
10451  SH
10452
10453     * A new option -msoft-atomic has been added. When it is specified,
10454       GCC will generate GNU/Linux-compatible gUSA atomic sequences for
10455       the new __atomic routines.
10456     * Since it is neither supported by GAS nor officially documented,
10457       code generation for little endian SH2A has been disabled.
10458       Specifying -ml with -m2a* will now result in a compiler error.
10459     * The defunct -mbranch-cost option has been fixed.
10460     * Some improvements to the generated code of:
10461          + Utilization of the tst #imm,R0 instruction.
10462          + Dynamic shift instructions on SH2A.
10463          + Integer absolute value calculations.
10464     * The -mdiv= option for targets other than SHmedia has been fixed and
10465       documented.
10466
10467  SPARC
10468
10469     * The option -mflat has been reinstated. When it is specified, the
10470       compiler will generate code for a single register window model.
10471       This is essentially a new implementation and the corresponding
10472       debugger support has been added to GDB 7.4.
10473     * Support for the options -mtune=native and -mcpu=native has been
10474       added on selected native platforms (GNU/Linux and Solaris).
10475     * Support for the SPARC T3 (Niagara 3) processor has been added.
10476     * VIS:
10477          + An intrinsics header visintrin.h has been added.
10478          + Builtin intrinsics for the VIS 1.0 edge handling and pixel
10479            compare instructions have been added.
10480          + The little-endian version of alignaddr is now supported.
10481          + When possible, VIS builtins are marked const, which should
10482            increase the compiler's ability to optimize VIS operations.
10483          + The compiler now properly tracks the %gsr register and how it
10484            behaves as an input for various VIS instructions.
10485          + Akin to fzero, the compiler can now generate fone instructions
10486            in order to set all of the bits of a floating-point register
10487            to 1.
10488          + The documentation for the VIS intrinsics in the GCC manual has
10489            been brought up to date and many inaccuracies were fixed.
10490          + Intrinsics for the VIS 2.0 bmask, bshuffle, and
10491            non-condition-code setting edge instructions have been added.
10492            Their availability is controlled by the new -mvis2 and
10493            -mno-vis2 options. They are enabled by default on
10494            UltraSPARC-III and later CPUs.
10495     * Support for UltraSPARC Fused Multiply-Add floating-point extensions
10496       has been added. These instructions are enabled by default on SPARC
10497       T3 (Niagara 3) and later CPUs.
10498
10499  TILE-Gx/TILEPro
10500
10501     * Support has been added for the Tilera TILE-Gx and TILEPro families
10502       of processors.
10503
10504Other significant improvements
10505
10506     * A new option (-grecord-gcc-switches) was added that appends
10507       compiler command-line options that might affect code generation to
10508       the DW_AT_producer attribute string in the DWARF debugging
10509       information.
10510     * GCC now supports various new GNU extensions to the DWARF debugging
10511       information format, like [40]entry value and [41]call site
10512       information, [42]typed DWARF stack or [43]a more compact macro
10513       representation. Support for these extensions has been added to GDB
10514       7.4. They can be disabled through the -gstrict-dwarf command-line
10515       option.
10516
10517GCC 4.7.1
10518
10519   This is the [44]list of problem reports (PRs) from GCC's bug tracking
10520   system that are known to be fixed in the 4.7.1 release. This list might
10521   not be complete (that is, it is possible that some PRs that have been
10522   fixed are not listed here).
10523
10524   The Go front end in the 4.7.1 release fully supports the [45]Go 1
10525   language standard.
10526
10527GCC 4.7.2
10528
10529   This is the [46]list of problem reports (PRs) from GCC's bug tracking
10530   system that are known to be fixed in the 4.7.2 release. This list might
10531   not be complete (that is, it is possible that some PRs that have been
10532   fixed are not listed here).
10533
10534GCC 4.7.3
10535
10536   This is the [47]list of problem reports (PRs) from GCC's bug tracking
10537   system that are known to be fixed in the 4.7.3 release. This list might
10538   not be complete (that is, it is possible that some PRs that have been
10539   fixed are not listed here).
10540
10541GCC 4.7.4
10542
10543   This is the [48]list of problem reports (PRs) from GCC's bug tracking
10544   system that are known to be fixed in the 4.7.4 release. This list might
10545   not be complete (that is, it is possible that some PRs that have been
10546   fixed are not listed here).
10547
10548
10549    For questions related to the use of GCC, please consult these web
10550    pages and the [49]GCC manuals. If that fails, the
10551    [50]gcc-help@gcc.gnu.org mailing list might help. Comments on these
10552    web pages and the development of GCC are welcome on our developer
10553    list at [51]gcc@gcc.gnu.org. All of [52]our lists have public
10554    archives.
10555
10556   Copyright (C) [53]Free Software Foundation, Inc. Verbatim copying and
10557   distribution of this entire article is permitted in any medium,
10558   provided this notice is preserved.
10559
10560   These pages are [54]maintained by the GCC team. Last modified
10561   2023-03-29.
10562
10563References
10564
10565   1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html
10566   2. http://savannah.nongnu.org/bugs/?35407
10567   3. https://gcc.gnu.org/PR18145
10568   4. https://gcc.gnu.org/gcc-4.7/porting_to.html
10569   5. https://gcc.gnu.org/wiki/TransactionalMemory
10570   6. https://gcc.gnu.org/wiki/Atomic/GCCMM
10571   7. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
10572   8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
10573   9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
10574  10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
10575  11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
10576  12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
10577  13. https://gcc.gnu.org/PR14258
10578  14. https://gcc.gnu.org/PR35688
10579  15. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011
10580  16. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
10581  17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689
10582  18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270
10583  19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
10584  20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275
10585  21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170
10586  22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270
10587  23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWreal-q-constant_007d-149
10588  24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html
10589  25. https://learn.microsoft.com/en-us/windows/win32/wer/collecting-user-mode-dumps
10590  26. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183
10591  27. https://gcc.gnu.org/wiki/Fortran2003Status
10592  28. https://gcc.gnu.org/wiki/OOP
10593  29. https://gcc.gnu.org/wiki/Fortran2008Status
10594  30. https://gcc.gnu.org/wiki/Coarray
10595  31. https://gcc.gnu.org/wiki/CoarrayLib
10596  32. https://gcc.gnu.org/wiki/TS29113Status
10597  33. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bstd_003d_007d_0040var_007bstd_007d-option-53
10598  34. https://go.dev/doc/go1
10599  35. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html
10600  36. http://www.nongnu.org/avr-libc/
10601  37. https://gcc.gnu.org/PR54461
10602  38. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built%5f002din-Functions.html
10603  39. https://sites.google.com/site/x32abi/
10604  40. https://dwarfstd.org/issues/100909.1.html
10605  41. https://dwarfstd.org/issues/100909.2.html
10606  42. https://dwarfstd.org/issues/140425.1.html
10607  43. https://dwarfstd.org/issues/110722.1.html
10608  44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1
10609  45. https://go.dev/doc/go1
10610  46. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2
10611  47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3
10612  48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4
10613  49. https://gcc.gnu.org/onlinedocs/
10614  50. mailto:gcc-help@gcc.gnu.org
10615  51. mailto:gcc@gcc.gnu.org
10616  52. https://gcc.gnu.org/lists.html
10617  53. https://www.fsf.org/
10618  54. https://gcc.gnu.org/about.html
10619======================================================================
10620http://gcc.gnu.org/gcc-4.6/index.html
10621
10622                             GCC 4.6 Release Series
10623
10624   (This release series is no longer supported.)
10625
10626   April 12, 2013
10627
10628   The [1]GNU project and the GCC developers are pleased to announce the
10629   release of GCC 4.6.4.
10630
10631   This release is a bug-fix release, containing fixes for regressions in
10632   GCC 4.6.3 relative to previous releases of GCC.
10633
10634Release History
10635
10636   GCC 4.6.4
10637          April 12, 2013 ([2]changes, [3]documentation)
10638
10639   GCC 4.6.3
10640          March 1, 2012 ([4]changes, [5]documentation)
10641
10642   GCC 4.6.2
10643          October 26, 2011 ([6]changes, [7]documentation)
10644
10645   GCC 4.6.1
10646          June 27, 2011 ([8]changes, [9]documentation)
10647
10648   GCC 4.6.0
10649          March 25, 2011 ([10]changes, [11]documentation)
10650
10651References and Acknowledgements
10652
10653   GCC used to stand for the GNU C Compiler, but since the compiler
10654   supports several other languages aside from C, it now stands for the
10655   GNU Compiler Collection.
10656
10657   A list of [12]successful builds is updated as new information becomes
10658   available.
10659
10660   The GCC developers would like to thank the numerous people that have
10661   contributed new features, improvements, bug fixes, and other changes as
10662   well as test results to GCC. This [13]amazing group of volunteers is
10663   what makes GCC successful.
10664
10665   For additional information about GCC please refer to the [14]GCC
10666   project web site or contact the [15]GCC development mailing list.
10667
10668   To obtain GCC please use [16]our mirror sites or [17]our version
10669   control system.
10670
10671
10672    For questions related to the use of GCC, please consult these web
10673    pages and the [18]GCC manuals. If that fails, the
10674    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
10675    web pages and the development of GCC are welcome on our developer
10676    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
10677    archives.
10678
10679   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
10680   distribution of this entire article is permitted in any medium,
10681   provided this notice is preserved.
10682
10683   These pages are [23]maintained by the GCC team. Last modified
10684   2022-10-26.
10685
10686References
10687
10688   1. http://www.gnu.org/
10689   2. http://gcc.gnu.org/gcc-4.6/changes.html
10690   3. https://gcc.gnu.org/onlinedocs/4.6.4/
10691   4. http://gcc.gnu.org/gcc-4.6/changes.html
10692   5. https://gcc.gnu.org/onlinedocs/4.6.3/
10693   6. http://gcc.gnu.org/gcc-4.6/changes.html
10694   7. https://gcc.gnu.org/onlinedocs/4.6.2/
10695   8. http://gcc.gnu.org/gcc-4.6/changes.html
10696   9. https://gcc.gnu.org/onlinedocs/4.6.1/
10697  10. http://gcc.gnu.org/gcc-4.6/changes.html
10698  11. https://gcc.gnu.org/onlinedocs/4.6.0/
10699  12. http://gcc.gnu.org/gcc-4.6/buildstat.html
10700  13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
10701  14. http://gcc.gnu.org/index.html
10702  15. mailto:gcc@gcc.gnu.org
10703  16. http://gcc.gnu.org/mirrors.html
10704  17. http://gcc.gnu.org/git.html
10705  18. https://gcc.gnu.org/onlinedocs/
10706  19. mailto:gcc-help@gcc.gnu.org
10707  20. mailto:gcc@gcc.gnu.org
10708  21. https://gcc.gnu.org/lists.html
10709  22. https://www.fsf.org/
10710  23. https://gcc.gnu.org/about.html
10711======================================================================
10712http://gcc.gnu.org/gcc-4.6/changes.html
10713
10714                             GCC 4.6 Release Series
10715                        Changes, New Features, and Fixes
10716
10717Caveats
10718
10719     * The options -b <machine> and -V <version> have been removed because
10720       they were unreliable. Instead, users should directly run
10721       <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to
10722       run a different version of gcc.
10723     * GCC now has stricter checks for invalid command-line options. In
10724       particular, when gcc was called to link object files rather than
10725       compile source code, it would previously accept and ignore all
10726       options starting with --, including linker options such as
10727       --as-needed and --export-dynamic, although such options would
10728       result in errors if any source code was compiled. Such options, if
10729       unknown to the compiler, are now rejected in all cases; if the
10730       intent was to pass them to the linker, options such as
10731       -Wl,--as-needed should be used.
10732     * Versions of the GNU C library up to and including 2.11.1 included
10733       an [1]incorrect implementation of the cproj function. GCC optimizes
10734       its builtin cproj according to the behavior specified and allowed
10735       by the ISO C99 standard. If you want to avoid discrepancies between
10736       the C library and GCC's builtin transformations when using cproj in
10737       your code, use GLIBC 2.12 or later. If you are using an older GLIBC
10738       and actually rely on the incorrect behavior of cproj, then you can
10739       disable GCC's transformations using -fno-builtin-cproj.
10740     * The C-only intermodule optimization framework (IMA, enabled by
10741       -combine) has been removed in favor of the new generic link-time
10742       optimization framework (LTO) introduced in [2]GCC 4.5.0.
10743     * GCC now ships with the LGPL-licensed libquadmath library, which
10744       provides quad-precision mathematical functions for targets with a
10745       __float128 datatype. __float128 is available for targets on 32-bit
10746       x86, x86-64 and Itanium architectures. The libquadmath library is
10747       automatically built on such targets when building the Fortran
10748       compiler.
10749     * New -Wunused-but-set-variable and -Wunused-but-set-parameter
10750       warnings were added for C, C++, Objective-C and Objective-C++.
10751       These warnings diagnose variables respective parameters which are
10752       only set in the code and never otherwise used. Usually such
10753       variables are useless and often even the value assigned to them is
10754       computed needlessly, sometimes expensively. The
10755       -Wunused-but-set-variable warning is enabled by default by -Wall
10756       flag and -Wunused-but-set-parameter by -Wall -Wextra flags.
10757     * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
10758       rules for the layout of vectors that could lead to wrong code being
10759       generated. Vectors larger than 8 bytes in size are now by default
10760       aligned to an 8-byte boundary. This is an ABI change: code that
10761       makes explicit use of vector types may be incompatible with binary
10762       objects built with older versions of GCC. Auto-vectorized code is
10763       not affected by this change. (This change affects GCC versions
10764       4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.)
10765     * On AVR, variables with the progmem attribute to locate data in
10766       flash memory must be qualified as const.
10767     * Support for a number of older systems and recently unmaintained or
10768       untested target ports of GCC has been declared obsolete in GCC 4.6.
10769       Unless there is activity to revive them, the next release of GCC
10770       will have their sources permanently removed.
10771       All GCC ports for the following processor architectures have been
10772       declared obsolete:
10773          + Argonaut ARC (arc-*)
10774          + National Semiconductor CRX (crx-*)
10775          + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*,
10776            m68hc12-*-*, m6812-*-*)
10777          + Sunplus S+core (score-*)
10778       The following ports for individual systems on particular
10779       architectures have been obsoleted:
10780          + Interix (i[34567]86-*-interix3*)
10781          + NetWare x86 (i[3456x]86-*-netware*)
10782          + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*)
10783          + MCore PE (mcore-*-pe*)
10784          + SH SymbianOS (sh*-*-symbianelf*)
10785          + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*)
10786          + M68K uClinux old ABI (m68k-*-uclinuxoldabi*)
10787          + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*,
10788            vax-*-netbsd*, but not *-*-netbsdelf*)
10789       The i[34567]86-*-pe alias for Cygwin targets has also been
10790       obsoleted; users should configure for i[34567]86-*-cygwin* instead.
10791       Certain configure options to control the set of libraries built
10792       with GCC on some targets have been obsoleted. On ARM targets, the
10793       options --disable-fpu, --disable-26bit, --disable-underscore,
10794       --disable-interwork, --disable-biendian and --disable-nofmult have
10795       been obsoleted. On MIPS targets, the options
10796       --disable-single-float, --disable-biendian and --disable-softfloat
10797       have been obsoleted.
10798     * Support has been removed for all the [3]configurations obsoleted in
10799       GCC 4.5.
10800     * More information on porting to GCC 4.6 from previous versions of
10801       GCC can be found in the [4]porting guide for this release.
10802
10803General Optimizer Improvements
10804
10805     * A new general optimization level, -Ofast, has been introduced. It
10806       combines the existing optimization level -O3 with options that can
10807       affect standards compliance but result in better optimized code.
10808       For example, -Ofast enables -ffast-math.
10809     * Link-time optimization improvements:
10810          + The [5]Scalable Whole Program Optimizer (WHOPR) project has
10811            stabilized to the point of being usable. It has become the
10812            default mode when using the LTO optimization model. Link time
10813            optimization can now split itself into multiple parallel
10814            compilations. Parallelism is controlled with -flto=n (where n
10815            specifies the number of compilations to execute in parallel).
10816            GCC can also cooperate with a GNU make job server by
10817            specifying the -flto=jobserver option and adding + to the
10818            beginning of the Makefile rule executing the linker.
10819            Classical LTO mode can be enforced by -flto-partition=none.
10820            This may result in small code quality improvements.
10821          + A large number of bugs were fixed. GCC itself, Mozilla Firefox
10822            and other large applications can be built with LTO enabled.
10823          + The linker plugin support improvements
10824               o Linker plugin is now enabled by default when the linker
10825                 is detected to have plugin support. This is the case for
10826                 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and
10827                 the Gold linker on ELF targets. Plugin support of the
10828                 Apple linker on Darwin is not compatible with GCC. The
10829                 linker plugin can also be controlled by the
10830                 -fuse-linker-plugin command-line option.
10831               o Resolution information from the linker plugin is used to
10832                 drive whole program assumptions. Use of the linker plugin
10833                 results in more aggressive optimization on binaries and
10834                 on shared libraries that use the hidden visibility
10835                 attribute. Consequently the use of -fwhole-program is not
10836                 necessary in addition to LTO.
10837          + Hidden symbols used from non-LTO objects now have to be
10838            explicitly annotated with externally_visible when the linker
10839            plugin is not used.
10840          + C++ inline functions and virtual tables are now privatized
10841            more aggressively, leading to better inter-procedural
10842            optimization and faster dynamic linking.
10843          + Memory usage and intermediate language streaming performance
10844            have been improved.
10845          + Static constructors and destructors from individual units are
10846            inlined into a single function. This can significantly improve
10847            startup times of large C++ applications where static
10848            constructors are very common. For example, static constructors
10849            are used when including the iostream header.
10850          + Support for the Ada language has been added.
10851     * Interprocedural optimization improvements
10852          + The interprocedural framework was re-tuned for link time
10853            optimization. Several scalability issues were resolved.
10854          + Improved auto-detection of const and pure functions. Newly,
10855            noreturn functions are auto-detected.
10856            The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is
10857            available that informs users when adding attributes to headers
10858            might improve code generation.
10859          + A number of inlining heuristic improvements. In particular:
10860               o Partial inlining is now supported and enabled by default
10861                 at -O2 and greater. The feature can be controlled via
10862                 -fpartial-inlining.
10863                 Partial inlining splits functions with short hot path to
10864                 return. This allows more aggressive inlining of the hot
10865                 path leading to better performance and often to code size
10866                 reductions (because cold parts of functions are not
10867                 duplicated).
10868               o Scalability for large compilation units was improved
10869                 significantly.
10870               o Inlining of callbacks is now more aggressive.
10871               o Virtual methods are considered for inlining when the
10872                 caller is inlined and devirtualization is then possible.
10873               o Inlining when optimizing for size (either in cold regions
10874                 of a program or when compiling with -Os) was improved to
10875                 better handle C++ programs with larger abstraction
10876                 penalty, leading to smaller and faster code.
10877          + The IPA reference optimization pass detecting global variables
10878            used or modified by functions was strengthened and sped up.
10879          + Functions whose address was taken are now optimized out when
10880            all references to them are dead.
10881          + A new inter-procedural static profile estimation pass detects
10882            functions that are executed once or unlikely to be executed.
10883            Unlikely executed functions are optimized for size. Functions
10884            executed once are optimized for size except for the inner
10885            loops.
10886          + On most targets with named section support, functions used
10887            only at startup (static constructors and main), functions used
10888            only at exit and functions detected to be cold are placed into
10889            separate text segment subsections. This extends the
10890            -freorder-functions feature and is controlled by the same
10891            switch. The goal is to improve the startup time of large C++
10892            programs.
10893            Proper function placement requires linker support. GNU ld
10894            2.21.51 on ELF targets was updated to place those functions
10895            together within the text section leading to better code
10896            locality and faster startup times of large C++ programs. The
10897            feature is also supported in the Apple linker. Support in the
10898            gold linker is planned.
10899     * A new switch -fstack-usage has been added. It makes the compiler
10900       output stack usage information for the program, on a per-function
10901       basis, in an auxiliary file.
10902     * A new switch -fcombine-stack-adjustments has been added. It can be
10903       used to enable or disable the compiler's stack-slot combining pass
10904       which before was enabled automatically at -O1 and above, but could
10905       not be controlled on its own.
10906     * A new switch -fstrict-volatile-bitfields has been added. Using it
10907       indicates that accesses to volatile bitfields should use a single
10908       access of the width of the field's type. This option can be useful
10909       for precisely defining and accessing memory-mapped peripheral
10910       registers from C or C++.
10911
10912Compile time and memory usage improvements
10913
10914     * Datastructures used by the dataflow framework in GCC were
10915       reorganized for better memory usage and more cache locality.
10916       Compile time is improved especially on units with large functions
10917       (possibly resulting from a lot of inlining) not fitting into the
10918       processor cache. The compile time of the GCC C compiler binary with
10919       link-time optimization went down by over 10% (benchmarked on x86-64
10920       target).
10921
10922New Languages and Language specific improvements
10923
10924  Ada
10925
10926     * Stack checking has been improved on selected architectures (Alpha,
10927       IA-32/x86-64, RS/6000 and SPARC): it now will detect stack
10928       overflows in all cases on these architectures.
10929     * Initial support for Ada 2012 has been added.
10930
10931  C family
10932
10933     * A new warning, enabled by -Wdouble-promotion, has been added that
10934       warns about cases where a value of type float is implicitly
10935       promoted to double. This is especially helpful for CPUs that handle
10936       the former in hardware, but emulate the latter in software.
10937     * A new function attribute leaf was introduced. This attribute allows
10938       better inter-procedural optimization across calls to functions that
10939       return to the current unit only via returning or exception
10940       handling. This is the case for most library functions that have no
10941       callbacks.
10942     * Support for a new data type __int128 for targets having wide enough
10943       machine-mode support.
10944     * The new function attribute callee_pop_aggregate allows to specify
10945       if the caller or callee is responsible for popping the aggregate
10946       return pointer value from the stack.
10947     * Support for selectively enabling and disabling warnings via #pragma
10948       GCC diagnostic has been added. For instance:
10949#pragma GCC diagnostic error "-Wuninitialized"
10950  foo(a);                       /* error is given for this one */
10951#pragma GCC diagnostic push
10952#pragma GCC diagnostic ignored "-Wuninitialized"
10953  foo(b);                       /* no diagnostic for this one */
10954#pragma GCC diagnostic pop
10955  foo(c);                       /* error is given for this one */
10956#pragma GCC diagnostic pop
10957  foo(d);                       /* depends on command-line options */
10958
10959     * The -fmax-errors=N option is now supported. Using this option
10960       causes the compiler to exit after N errors have been issued.
10961
10962  C
10963
10964     * There is now experimental support for some features from the
10965       upcoming C1X revision of the ISO C standard. This support may be
10966       selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions.
10967       Note that this support is experimental and may change incompatibly
10968       in future releases for consistency with changes to the C1X standard
10969       draft. The following features are newly supported as described in
10970       the N1539 draft of C1X (with changes agreed at the March 2011 WG14
10971       meeting); some other features were already supported with no
10972       compiler changes being needed, or have some support but not in full
10973       accord with N1539 (as amended).
10974          + Static assertions (_Static_assert keyword)
10975          + Typedef redefinition
10976          + New macros in <float.h>
10977          + Anonymous structures and unions
10978     * The new -fplan9-extensions option directs the compiler to support
10979       some extensions for anonymous struct fields which are implemented
10980       by the Plan 9 compiler. A pointer to a struct may be automatically
10981       converted to a pointer to an anonymous field when calling a
10982       function, in order to make the types match. An anonymous struct
10983       field whose type is a typedef name may be referred to using the
10984       typedef name.
10985
10986  C++
10987
10988     * Improved [7]experimental support for the upcoming C++0x ISO C++
10989       standard, including support for constexpr (thanks to Gabriel Dos
10990       Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide),
10991       noexcept, unrestricted unions, range-based for loops (thanks to
10992       Rodrigo Rivas Costa), opaque enum declarations (thanks also to
10993       Rodrigo), implicitly deleted functions and implicit move
10994       constructors.
10995     * When an extern declaration within a function does not match a
10996       declaration in the enclosing context, G++ now properly declares the
10997       name within the namespace of the function rather than the namespace
10998       which was open just before the function definition ([8]c++/43145).
10999     * GCC now warns by default when casting integers to larger pointer
11000       types. These warnings can be disabled with the option
11001       -Wno-int-to-pointer-cast, which is now also available in C++.
11002     * G++ no longer optimizes using the assumption that a value of
11003       enumeration type will fall within the range specified by the
11004       standard, since that assumption is easily violated with a
11005       conversion from integer type ([9]c++/43680). The old behavior can
11006       be restored with -fstrict-enums.
11007     * The new -fnothrow-opt flag changes the semantics of a throw()
11008       exception specification to match the proposed semantics of the
11009       noexcept specification: just call terminate if an exception tries
11010       to propagate out of a function with such an exception
11011       specification. This dramatically reduces or eliminates the code
11012       size overhead from adding the exception specification.
11013     * The new -Wnoexcept flag will suggest adding a noexcept qualifier to
11014       a function that the compiler can tell doesn't throw if it would
11015       change the value of a noexcept expression.
11016     * The -Wshadow option now warns if a local variable or type
11017       declaration shadows another type in C++. Note that the compiler
11018       will not warn if a local variable shadows a struct/class/enum, but
11019       will warn if it shadows an explicit typedef.
11020     * When an identifier is not found in the current scope, G++ now
11021       offers suggestions about which identifier might have been intended.
11022     * G++ now issues clearer diagnostics for missing semicolons after
11023       class, struct, and union definitions.
11024     * G++ now issues clearer diagnostics for missing semicolons after
11025       class member declarations.
11026     * G++ now issues clearer diagnostics when a colon is used in a place
11027       where a double-colon was intended.
11028     * G++ no longer accepts mutable on reference members ([10]c++/33558).
11029       Use -fpermissive to allow the old, non-conforming behaviour.
11030     * A few mangling fixes have been made, to attribute const/volatile on
11031       function pointer types, decltype of a plain decl, and use of a
11032       function parameter in the declaration of another parameter. By
11033       default the compiler still uses the old mangling, but emits aliases
11034       with the new mangling on targets that support strong aliases. Users
11035       can switch over entirely to the new mangling with -fabi-version=5
11036       or -fabi-version=0. -Wabi will now warn about code that uses the
11037       old mangling.
11038     * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified
11039       type to be default initialized unless the type has a user-declared
11040       default constructor. In 4.6.2 G++ implements the proposed
11041       resolution of [11]DR 253, so default initialization is allowed if
11042       it initializes all subobjects. Code that fails to compile can be
11043       fixed by providing an initializer e.g.
11044    struct A { A(); };
11045    struct B : A { int i; };
11046    const B b = B();
11047       Use -fpermissive to allow the old, non-conforming behaviour.
11048
11049    Runtime Library (libstdc++)
11050
11051     * [12]Improved experimental support for the upcoming ISO C++
11052       standard, C++0x, including using constexpr and nullptr.
11053     * Performance improvements to the [13]Debug Mode, thanks to Fran�ois
11054       Dumont.
11055     * Atomic operations used for reference-counting are annotated so that
11056       they can be understood by race detectors such as Helgrind, see
11057       [14]Data Race Hunting.
11058     * Most libstdc++ standard headers have been changed to no longer
11059       include the cstddef header as an implementation detail. Code that
11060       relied on that header being included as side-effect of including
11061       other standard headers will need to include cstddef explicitly.
11062
11063  Fortran
11064
11065     * On systems supporting the libquadmath library, GNU Fortran now also
11066       supports a quad-precision, kind=16 floating-point data type
11067       (REAL(16), COMPLEX(16)). As the data type is not fully supported in
11068       hardware, calculations might be one to two orders of magnitude
11069       slower than with the 4, 8 or 10 bytes floating-point data types.
11070       This change does not affect systems which support REAL(16) in
11071       hardware nor those which do not support libquadmath.
11072     * Much improved compile time for large array constructors.
11073     * In order to reduce execution time and memory consumption, use of
11074       temporary arrays in assignment expressions is avoided for many
11075       cases. The compiler now reverses loops in order to avoid generating
11076       a temporary array where possible.
11077     * Improved diagnostics, especially with -fwhole-file.
11078     * The -fwhole-file flag is now enabled by default. This improves code
11079       generation and diagnostics. It can be disabled using the deprecated
11080       -fno-whole-file flag.
11081     * Support the generation of Makefile dependencies via the [15]-M...
11082       flags of GCC; you may need to specify the -cpp option in addition.
11083       The dependencies take modules, Fortran's include, and CPP's
11084       #include into account. Note: Using -M for the module path is no
11085       longer supported, use -J instead.
11086     * The flag -Wconversion has been modified to only issue warnings
11087       where a conversion leads to information loss. This drastically
11088       reduces the number of warnings; -Wconversion is thus now enabled
11089       with -Wall. The flag -Wconversion-extra has been added and also
11090       warns about other conversions; -Wconversion-extra typically issues
11091       a huge number of warnings, most of which can be ignored.
11092     * A new command-line option -Wunused-dummy-argument warns about
11093       unused dummy arguments and is included in -Wall. Before,
11094       -Wunused-variable also warned about unused dummy arguments.
11095     * Fortran 2003 support has been extended:
11096          + Improved support for polymorphism between libraries and
11097            programs and for complicated inheritance patterns (cf.
11098            [16]object-oriented programming).
11099          + Experimental support of the ASSOCIATE construct.
11100          + In pointer assignments it is now possible to specify the lower
11101            bounds of the pointer and, for a rank-1 or a simply contiguous
11102            data-target, to remap the bounds.
11103          + Automatic (re)allocation: In intrinsic assignments to
11104            allocatable variables the left-hand side will be automatically
11105            allocated (if unallocated) or reallocated (if the shape or
11106            type parameter is different). To avoid the small performance
11107            penalty, you can use a(:) = ... instead of a = ... for arrays
11108            and character strings - or disable the feature using -std=f95
11109            or -fno-realloc-lhs.
11110          + Deferred type parameter: For scalar allocatable and pointer
11111            variables the character length can be deferred.
11112          + Namelist variables with allocatable and pointer attribute and
11113            nonconstant length type parameter are supported.
11114     * Fortran 2008 support has been extended:
11115          + Experimental [17]coarray support (for one image only, i.e.
11116            num_images() == 1); use the [18]-fcoarray=single flag to
11117            enable it.
11118          + The STOP and the new ERROR STOP statements now support all
11119            constant expressions.
11120          + Support for the CONTIGUOUS attribute.
11121          + Support for ALLOCATE with MOLD.
11122          + Support for the STORAGE_SIZE intrinsic inquiry function.
11123          + Support of the NORM2 and PARITY intrinsic functions.
11124          + The following bit intrinsics were added: POPCNT and POPPAR for
11125            counting the number of 1 bits and returning the parity; BGE,
11126            BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR
11127            for combined left and right shifts, MASKL and MASKR for simple
11128            left and right justified masks, MERGE_BITS for a bitwise merge
11129            using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations,
11130            and the transformational bit intrinsics IALL, IANY and
11131            IPARITY.
11132          + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine.
11133          + Support for the IMPURE attribute for procedures, which allows
11134            for ELEMENTAL procedures without the restrictions of PURE.
11135          + Null pointers (including NULL()) and not allocated variables
11136            can be used as actual argument to optional non-pointer,
11137            non-allocatable dummy arguments, denoting an absent argument.
11138          + Non-pointer variables with TARGET attribute can be used as
11139            actual argument to POINTER dummies with INTENT(IN)
11140          + Pointers including procedure pointers and those in a derived
11141            type (pointer components) can now be initialized by a target
11142            instead of only by NULL.
11143          + The EXIT statement (with construct-name) can now be used to
11144            leave not only the DO but also the ASSOCIATE, BLOCK, IF,
11145            SELECT CASE and SELECT TYPE constructs.
11146          + Internal procedures can now be used as actual argument.
11147          + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS
11148            and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV
11149            have been added; these arrays contain the supported kind
11150            values for the respective types.
11151          + The module procedures C_SIZEOF of the intrinsic module
11152            ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of
11153            ISO_FORTRAN_ENV have been implemented.
11154          + Minor changes: obsolescence diagnostics for ENTRY was added
11155            for -std=f2008; a line may start with a semicolon; for
11156            internal and module procedures END can be used instead of END
11157            SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes
11158            a RADIX argument; intrinsic types are supported for
11159            TYPE(intrinsic-type-spec); multiple type-bound procedures can
11160            be declared in a single PROCEDURE statement; implied-shape
11161            arrays are supported for named constants (PARAMETER). The
11162            transformational, three argument versions of BESSEL_JN and
11163            BESSEL_YN were added - the elemental, two-argument version had
11164            been added in GCC 4.4; note that the transformational
11165            functions use a recurrence algorithm.
11166
11167  Go
11168
11169   Support for the Go programming language has been added to GCC. It is
11170   not enabled by default when you build GCC; use the --enable-languages
11171   configure option to build it. The driver program for compiling Go code
11172   is gccgo.
11173
11174   Go is currently known to work on GNU/Linux and RTEMS. Solaris support
11175   is in progress. It may or may not work on other platforms.
11176
11177  Objective-C and Objective-C++
11178
11179     * The -fobjc-exceptions flag is now required to enable Objective-C
11180       exception and synchronization syntax (introduced by the keywords
11181       @try, @catch, @finally and @synchronized).
11182     * A number of Objective-C 2.0 features and extensions are now
11183       supported by GCC. These features are enabled by default; you can
11184       disable them by using the new -fobjc-std=objc1 command-line option.
11185     * The Objective-C 2.0 dot-syntax is now supported. It is an
11186       alternative syntax for using getters and setters; object.count is
11187       automatically converted into [object count] or [object setCount:
11188       ...] depending on context; for example if (object.count > 0) is
11189       automatically compiled into the equivalent of if ([object count] >
11190       0) while object.count = 0; is automatically compiled into the
11191       equivalent ot [object setCount: 0];. The dot-syntax can be used
11192       with instance and class objects and with any setters or getters, no
11193       matter if they are part of a declared property or not.
11194     * Objective-C 2.0 declared properties are now supported. They are
11195       declared using the new @property keyword, and are most commonly
11196       used in conjunction with the new Objective-C 2.0 dot-syntax. The
11197       nonatomic, readonly, readwrite, assign, retain, copy, setter and
11198       getter attributes are all supported. Marking declared properties
11199       with __attribute__ ((deprecated)) is supported too.
11200     * The Objective-C 2.0 @synthesize and @dynamic keywords are
11201       supported. @synthesize causes the compiler to automatically
11202       synthesize a declared property, while @dynamic is used to disable
11203       all warnings for a declared property for which no implementation is
11204       provided at compile time. Synthesizing declared properties requires
11205       runtime support in most useful cases; to be able to use it with the
11206       GNU runtime, appropriate helper functions have been added to the
11207       GNU Objective-C runtime ABI, and are implemented by the GNU
11208       Objective-C runtime library shipped with GCC.
11209     * The Objective-C 2.0 fast enumeration syntax is supported in
11210       Objective-C. This is currently not yet available in Objective-C++.
11211       Fast enumeration requires support in the runtime, and such support
11212       has been added to the GNU Objective-C runtime library (shipped with
11213       GCC).
11214     * The Objective-C 2.0 @optional keyword is supported. It allows you
11215       to mark methods or properties in a protocol as optional as opposed
11216       to required.
11217     * The Objective-C 2.0 @package keyword is supported. It has currently
11218       the same effect as the @public keyword.
11219     * Objective-C 2.0 method attributes are supported. Currently the
11220       supported attributes are deprecated, sentinel, noreturn and format.
11221     * Objective-C 2.0 method argument attributes are supported. The most
11222       widely used attribute is unused, to mark an argument as unused in
11223       the implementation.
11224     * Objective-C 2.0 class and protocol attributes are supported.
11225       Currently the only supported attribute is deprecated.
11226     * Objective-C 2.0 class extensions are supported. A class extension
11227       has the same syntax as a category declaration with no category
11228       name, and the methods and properties declared in it are added
11229       directly to the main class. It is mostly used as an alternative to
11230       a category to add methods to a class without advertising them in
11231       the public headers, with the advantage that for class extensions
11232       the compiler checks that all the privately declared methods are
11233       actually implemented.
11234     * As a result of these enhancements, GCC can now be used to build
11235       Objective-C and Objective-C++ software that uses Foundation and
11236       other important system frameworks with the NeXT runtime on Darwin 9
11237       and Darwin 10 (OSX 10.5 and 10.6).
11238     * Many bugs in the compiler have been fixed in this release; in
11239       particular, LTO can now be used when compiling Objective-C and
11240       Objective-C++ and the parser is much more robust in dealing with
11241       invalid code.
11242
11243    Runtime Library (libobjc)
11244
11245     * The GNU Objective-C runtime library now defines the macro
11246       __GNU_LIBOBJC__ (with a value that is increased at every release
11247       where there is any change to the API) in objc/objc.h, making it
11248       easy to determine if the GNU Objective-C runtime library is being
11249       used, and if so, which version. Previous versions of the GNU
11250       Objective-C runtime library (and other Objective-C runtime
11251       libraries such as the Apple one) do not define this macro.
11252     * A new Objective-C 2.0 API, almost identical to the one implemented
11253       by the Apple Objective-C runtime, has been implemented in the GNU
11254       Objective-C runtime library. The new API hides the internals of
11255       most runtime structures but provides a more extensive set of
11256       functions to operate on them. It is much easier, for example, to
11257       create or modify classes at runtime. The new API also makes it
11258       easier to port software from Apple to GNU as almost no changes
11259       should be required. The old API is still supported for backwards
11260       compatibility; including the old objc/objc-api.h header file
11261       automatically selects the old API, while including the new
11262       objc/runtime.h header file automatically selects the new API.
11263       Support for the old API is being phased out and upgrading the
11264       software to use the new API is strongly recommended. To check for
11265       the availability of the new API, the __GNU_LIBOBJC__ macro can be
11266       used as older versions of the GNU Objective-C runtime library,
11267       which do not support the new API, do not define such a macro.
11268     * Runtime support for @synchronized has been added.
11269     * Runtime support for Objective-C 2.0 synthesized property accessors
11270       has been added.
11271     * Runtime support for Objective-C 2.0 fast enumeration has been
11272       added.
11273
11274New Targets and Target Specific Improvements
11275
11276  ARM
11277
11278     * GCC now supports the Cortex-M4 processor implementing the v7-em
11279       version of the architecture using the option -mcpu=cortex-m4.
11280     * Scheduling descriptions for the Cortex-M4, the Neon and the
11281       floating point units of the Cortex-A9 and a pipeline description
11282       for the Cortex-A5 have been added.
11283     * Synchronization primitives such as __sync_fetch_and_add and friends
11284       are now inlined for supported architectures rather than calling
11285       into a kernel helper function.
11286     * SSA loop prefetching is enabled by default for the Cortex-A9 at
11287       -O3.
11288     * Several improvements were committed to improve code generation for
11289       the ARM architecture including a rewritten implementation for load
11290       and store multiples.
11291     * Several enhancements were committed to improve SIMD code generation
11292       for NEON by adding support for widening instructions, misaligned
11293       loads and stores, vector conditionals and support for 64 bit
11294       arithmetic.
11295     * Support was added for the Faraday cores fa526, fa606te, fa626te,
11296       fmp626te, fmp626 and fa726te and can be used with the respective
11297       names as parameters to the -mcpu= option.
11298     * Basic support was added for Cortex-A15 and is available through
11299       -mcpu=cortex-a15.
11300     * GCC for AAPCS configurations now more closely adheres to the AAPCS
11301       specification by enabling -fstrict-volatile-bitfields by default.
11302
11303  IA-32/x86-64
11304
11305     * The new -fsplit-stack option permits programs to use a
11306       discontiguous stack. This is useful for threaded programs, in that
11307       it is no longer necessary to specify the maximum stack size when
11308       creating a thread. This feature is currently only implemented for
11309       32-bit and 64-bit x86 GNU/Linux targets.
11310     * Support for emitting profiler counter calls before function
11311       prologues. This is enabled via a new command-line option -mfentry.
11312     * Optimization for the Intel Core 2 processors is now available
11313       through the -march=core2 and -mtune=core2 options.
11314     * Support for Intel Core i3/i5/i7 processors is now available through
11315       the -march=corei7 and -mtune=corei7 options.
11316     * Support for Intel Core i3/i5/i7 processors with AVX is now
11317       available through the -march=corei7-avx and -mtune=corei7-avx
11318       options.
11319     * Support for AMD Bobcat (family 14) processors is now available
11320       through the -march=btver1 and -mtune=btver1 options.
11321     * Support for AMD Bulldozer (family 15) processors is now available
11322       through the -march=bdver1 and -mtune=bdver1 options.
11323     * The default setting (when not optimizing for size) for 32-bit
11324       GNU/Linux and Darwin x86 targets has been changed to
11325       -fomit-frame-pointer. The default can be reverted to
11326       -fno-omit-frame-pointer by configuring GCC with the
11327       --enable-frame-pointer configure option.
11328     * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support
11329       __float128 on 32-bit and 64-bit x86 targets.
11330     * AVX floating-point arithmetic can now be enabled by default at
11331       configure time with the new --with-fpmath=avx option.
11332     * The SSA loop prefetching pass is enabled when using -O3 when
11333       optimizing for CPUs where prefetching is beneficial (AMD CPUs newer
11334       than K6).
11335     * Support for TBM (Trailing Bit Manipulation) built-in functions and
11336       code generation is available via -mtbm.
11337     * Support for AMD's BMI (Bit Manipulation) built-in functions and
11338       code generation is available via -mbmi.
11339
11340  MicroBlaze
11341
11342     * Support has been added for the Xilinx MicroBlaze softcore processor
11343       (microblaze-elf) embedded target. This configurable processor is
11344       supported on several Xilinx Spartan and Virtex FPGAs.
11345
11346  MIPS
11347
11348     * GCC now supports the Loongson 3A processor. Its canonical -march=
11349       and -mtune= name is loongson3a.
11350
11351  MN10300 / AM33
11352
11353     * The inline assembly register constraint "A" has been renamed "c".
11354       This constraint is used to select a floating-point register that
11355       can be used as the destination of a multiply-accumulate
11356       instruction.
11357     * New inline assembly register constraints "A" and "D" have been
11358       added. These constraint letters resolve to all general registers
11359       when compiling for AM33, and resolve to address registers only or
11360       data registers only when compiling for MN10300.
11361     * The MDR register is represented in the compiler. One can access the
11362       register via the "z" constraint in inline assembly. It can be
11363       marked as clobbered or used as a local register variable via the
11364       "mdr" name. The compiler uses the RETF instruction if the function
11365       does not modify the MDR register, so it is important that inline
11366       assembly properly annotate any usage of the register.
11367
11368  PowerPC/PowerPC64
11369
11370     * GCC now supports the Applied Micro Titan processor with
11371       -mcpu=titan.
11372     * The -mrecip option has been added, which indicates whether the
11373       reciprocal and reciprocal square root instructions should be used.
11374     * The -mveclibabi=mass option can be used to enable the compiler to
11375       autovectorize mathematical functions using the Mathematical
11376       Acceleration Subsystem library.
11377     * The -msingle-pic-base option has been added, which instructs the
11378       compiler to avoid loading the PIC base register in function
11379       prologues. The PIC base register must be initialized by the runtime
11380       system.
11381     * The -mblock-move-inline-limit option has been added, which enables
11382       the user to control the maximum size of inlined memcpy calls and
11383       similar.
11384     * PowerPC64 GNU/Linux support for applications requiring a large TOC
11385       section has been improved. A new command-line option,
11386       -mcmodel=MODEL, controls this feature; valid values for MODEL are
11387       small, medium, or large.
11388     * The Altivec builtin functions vec_ld and vec_st have been modified
11389       to generate the Altivec memory instructions LVX and STVX, even if
11390       the -mvsx option is used. In the initial GCC 4.5 release, these
11391       builtin functions were changed to generate VSX memory reference
11392       instructions instead of Altivec memory instructions, but there are
11393       differences between the two instructions. If the VSX instruction
11394       set is available, you can now use the new builtin functions
11395       vec_vsx_ld and vec_vsx_st which always generates the VSX memory
11396       instructions.
11397     * The GCC compiler on AIX now defaults to a process layout with a
11398       larger data space allowing larger programs to be compiled.
11399     * The GCC long double type on AIX 6.1 and above has reverted to 64
11400       bit double precision, matching the AIX XL compiler default, because
11401       of missing C99 symbols required by the GCC runtime.
11402     * The default processor scheduling model and tuning for PowerPC64
11403       GNU/Linux and for AIX 6.1 and above now is POWER7.
11404     * Starting with GCC 4.6.1, vectors of type vector long long or vector
11405       long are passed and returned in the same method as other vectors
11406       with the VSX instruction set. Previously the GCC compiler did not
11407       adhere to the ABI for 128-bit vectors with 64-bit integer base
11408       types (PR 48857). This is also fixed in the GCC 4.5.4 release.
11409
11410  S/390, zSeries and System z9/z10, IBM zEnterprise z196
11411
11412     * Support for the zEnterprise z196 processor has been added. When
11413       using the -march=z196 option, the compiler will generate code
11414       making use of the following instruction facilities:
11415          + Conditional load/store
11416          + Distinct-operands
11417          + Floating-point-extension
11418          + Interlocked-access
11419          + Population-count
11420       The -mtune=z196 option avoids the compare and branch instructions
11421       as well as the load address instruction with an index register as
11422       much as possible and performs instruction scheduling appropriate
11423       for the new out-of-order pipeline architecture.
11424     * When using the -m31 -mzarch options the generated code still
11425       conforms to the 32-bit ABI but uses the general purpose registers
11426       as 64-bit registers internally. This requires a Linux kernel saving
11427       the whole 64-bit registers when doing a context switch. Kernels
11428       providing that feature indicate that by the 'highgprs' string in
11429       /proc/cpuinfo.
11430     * The SSA loop prefetching pass is enabled when using -O3.
11431
11432  SPARC
11433
11434     * GCC now supports the LEON series of SPARC V8 processors. The code
11435       generated by the compiler can either be tuned to it by means of the
11436       --with-tune=leon configure option and -mtune=leon compilation
11437       option, or the compiler can be built for the sparc-leon-{elf,linux}
11438       and sparc-leon3-{elf,linux} targets directly.
11439     * GCC has stopped sign/zero-extending parameter registers in the
11440       callee for functions taking parameters with sub-word size in 32-bit
11441       mode, since this is redundant with the specification of the ABI.
11442       GCC has never done so in 64-bit mode since this is also redundant.
11443     * The command-line option -mfix-at697f has been added to enable the
11444       documented workaround for the single erratum of the Atmel AT697F
11445       processor.
11446
11447Operating Systems
11448
11449  Android
11450
11451     * GCC now supports the Bionic C library and provides a convenient way
11452       of building native libraries and applications for the Android
11453       platform. Refer to the documentation of the -mandroid and -mbionic
11454       options for details on building native code. At the moment, Android
11455       support is enabled only for ARM.
11456
11457  Darwin/Mac OS X
11458
11459     * General
11460          + Initial support for CFString types has been added.
11461            This allows GCC to build projects including the system Core
11462            Foundation frameworks. The GCC Objective-C family supports
11463            CFString "toll-free bridged" as per the Mac OS X system tools.
11464            CFString is also recognized in the context of format
11465            attributes and arguments (see the documentation for format
11466            attributes for limitations). At present, 8-bit character types
11467            are supported.
11468          + Object file size reduction.
11469            The Darwin zeroed memory allocators have been re-written to
11470            make more use of .zerofill sections. For non-debug code, this
11471            can reduce object file size significantly.
11472          + Objective-C family 64-bit support (NeXT ABI 2).
11473            Initial support has been added to support 64-bit Objective-C
11474            code using the Darwin/OS X native (NeXT) runtime. ABI version
11475            2 will be selected automatically when 64-bit code is built.
11476          + Objective-C family 32-bit ABI 1.
11477            For 32-bit code ABI 1 is also now also allowed. At present it
11478            must be selected manually using -fobjc-abi-version=1 where
11479            applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6).
11480     * x86 Architecture
11481          + The -mdynamic-no-pic option has been enabled.
11482            Code supporting -mdynamic-no-pic optimization has been added
11483            and is applicable to -m32 builds. The compiler bootstrap uses
11484            the option where appropriate.
11485          + The default value for -mtune= has been changed.
11486            Since Darwin systems are primarily Xeon, Core-2 or similar the
11487            default tuning has been changed to -mtune=core2.
11488          + Enable 128-bit long double (__float128) support on Darwin.
11489     * PPC Architecture
11490          + Darwin64 ABI.
11491            Several significant bugs have been fixed, such that GCC now
11492            produces code compatible with the Darwin64 PowerPC ABI.
11493          + libffi and boehm-gc.
11494            The Darwin ports of the libffi and boehm-gc libraries have
11495            been upgraded to include a Darwin64 implementation. This means
11496            that powerpc*-*-darwin9 platforms may now, for example, build
11497            Java applications with -m64 enabled.
11498          + Plug-in support has been enabled.
11499          + The -fsection-anchors option is now available although,
11500            presently, not heavily tested.
11501
11502  Solaris 2
11503
11504    New Features
11505
11506     * Support symbol versioning with the Sun linker.
11507     * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+.
11508     * Support thread-local storage (TLS) with the Sun assembler on
11509       Solaris 2/x86.
11510     * Support TLS on Solaris 8/9 if prerequisites are met.
11511     * Support COMDAT group with the GNU assembler and recent Sun linker.
11512     * Support the Sun assembler visibility syntax.
11513     * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp.
11514       -march=pentiumpro (Solaris 8/9).
11515     * Don't use SSE on Solaris 8/9 x86 by default.
11516     * Enable 128-bit long double (__float128) support on Solaris 2/x86.
11517
11518    ABI Change
11519
11520     * Change the ABI for returning 8-byte vectors like __m64 in MMX
11521       registers on Solaris 10+/x86 to match the Sun Studio 12.1+
11522       compilers. This is an incompatible change. If you use such types,
11523       you must either recompile all your code with the new compiler or
11524       use the new -mvect8-ret-in-mem option to remain compatible with
11525       previous versions of GCC and Sun Studio.
11526
11527  Windows x86/x86_64
11528
11529     * Initial support for decimal floating point.
11530     * Support for the __thiscall calling-convention.
11531     * Support for hot-patchable function prologues via the
11532       ms_hook_prologue attribute for x86_64 in addition to 32-bit x86.
11533     * Improvements of stack-probing and stack-allocation mechanisms.
11534     * Support of push/pop-macro pragma as preprocessor command.
11535       With #pragma push_macro("macro-name") the current definition of
11536       macro-name is saved and can be restored with #pragma
11537       pop_macro("macro-name") to its saved definition.
11538     * Enable 128-bit long double (__float128) support on MinGW and
11539       Cygwin.
11540
11541Other significant improvements
11542
11543  Installation changes
11544
11545     * An install-strip make target is provided that installs stripped
11546       executables, and may install libraries with unneeded or debugging
11547       sections stripped.
11548     * On Power7 systems, there is a potential problem if you build the
11549       GCC compiler with a host compiler using options that enable the VSX
11550       instruction set generation. If the host compiler has been patched
11551       so that the vec_ld and vec_st builtin functions generate Altivec
11552       memory instructions instead of VSX memory instructions, then you
11553       should be able to build the compiler with VSX instruction
11554       generation.
11555
11556Changes for GCC Developers
11557
11558   Note: these changes concern developers that develop GCC itself or
11559   software that integrates with GCC, such as plugins, and not the general
11560   GCC users.
11561     * The gengtype utility, which previously was internal to the GCC
11562       build process, has been enchanced to provide GC root information
11563       for plugins as necessary.
11564     * The old GC allocation interface of ggc_alloc and friends was
11565       replaced with a type-safe alternative.
11566
11567GCC 4.6.1
11568
11569   This is the [19]list of problem reports (PRs) from GCC's bug tracking
11570   system that are known to be fixed in the 4.6.1 release. This list might
11571   not be complete (that is, it is possible that some PRs that have been
11572   fixed are not listed here).
11573
11574GCC 4.6.2
11575
11576   This is the [20]list of problem reports (PRs) from GCC's bug tracking
11577   system that are known to be fixed in the 4.6.2 release. This list might
11578   not be complete (that is, it is possible that some PRs that have been
11579   fixed are not listed here).
11580
11581GCC 4.6.3
11582
11583   This is the [21]list of problem reports (PRs) from GCC's bug tracking
11584   system that are known to be fixed in the 4.6.3 release. This list might
11585   not be complete (that is, it is possible that some PRs that have been
11586   fixed are not listed here).
11587
11588GCC 4.6.4
11589
11590   This is the [22]list of problem reports (PRs) from GCC's bug tracking
11591   system that are known to be fixed in the 4.6.4 release. This list might
11592   not be complete (that is, it is possible that some PRs that have been
11593   fixed are not listed here).
11594
11595
11596    For questions related to the use of GCC, please consult these web
11597    pages and the [23]GCC manuals. If that fails, the
11598    [24]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11599    web pages and the development of GCC are welcome on our developer
11600    list at [25]gcc@gcc.gnu.org. All of [26]our lists have public
11601    archives.
11602
11603   Copyright (C) [27]Free Software Foundation, Inc. Verbatim copying and
11604   distribution of this entire article is permitted in any medium,
11605   provided this notice is preserved.
11606
11607   These pages are [28]maintained by the GCC team. Last modified
11608   2023-01-18.
11609
11610References
11611
11612   1. https://sourceware.org/bugzilla/show_bug.cgi?id=10401
11613   2. http://gcc.gnu.org/gcc-4.5/changes.html
11614   3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted
11615   4. http://gcc.gnu.org/gcc-4.6/porting_to.html
11616   5. http://gcc.gnu.org/projects/lto/whopr.pdf
11617   6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
11618   7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html
11619   8. https://gcc.gnu.org/PR43145
11620   9. https://gcc.gnu.org/PR43680
11621  10. https://gcc.gnu.org/PR33558
11622  11. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253
11623  12. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x
11624  13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html
11625  14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races
11626  15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html
11627  16. https://gcc.gnu.org/wiki/OOP
11628  17. https://gcc.gnu.org/wiki/Coarray
11629  18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233
11630  19. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1
11631  20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2
11632  21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3
11633  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4
11634  23. https://gcc.gnu.org/onlinedocs/
11635  24. mailto:gcc-help@gcc.gnu.org
11636  25. mailto:gcc@gcc.gnu.org
11637  26. https://gcc.gnu.org/lists.html
11638  27. https://www.fsf.org/
11639  28. https://gcc.gnu.org/about.html
11640======================================================================
11641http://gcc.gnu.org/gcc-4.5/index.html
11642
11643                             GCC 4.5 Release Series
11644
11645   (This release series is no longer supported.)
11646
11647   Jul 2, 2012
11648
11649   The [1]GNU project and the GCC developers are pleased to announce the
11650   release of GCC 4.5.4.
11651
11652   This release is a bug-fix release, containing fixes for regressions in
11653   GCC 4.5.3 relative to previous releases of GCC.
11654
11655Release History
11656
11657   GCC 4.5.4
11658          Jul 2, 2012 ([2]changes)
11659
11660   GCC 4.5.3
11661          Apr 28, 2011 ([3]changes)
11662
11663   GCC 4.5.2
11664          Dec 16, 2010 ([4]changes)
11665
11666   GCC 4.5.1
11667          Jul 31, 2010 ([5]changes)
11668
11669   GCC 4.5.0
11670          April 14, 2010 ([6]changes)
11671
11672References and Acknowledgements
11673
11674   GCC used to stand for the GNU C Compiler, but since the compiler
11675   supports several other languages aside from C, it now stands for the
11676   GNU Compiler Collection.
11677
11678   A list of [7]successful builds is updated as new information becomes
11679   available.
11680
11681   The GCC developers would like to thank the numerous people that have
11682   contributed new features, improvements, bug fixes, and other changes as
11683   well as test results to GCC. This [8]amazing group of volunteers is
11684   what makes GCC successful.
11685
11686   For additional information about GCC please refer to the [9]GCC project
11687   web site or contact the [10]GCC development mailing list.
11688
11689   To obtain GCC please use [11]our mirror sites or [12]our version
11690   control system.
11691
11692
11693    For questions related to the use of GCC, please consult these web
11694    pages and the [13]GCC manuals. If that fails, the
11695    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11696    web pages and the development of GCC are welcome on our developer
11697    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
11698    archives.
11699
11700   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
11701   distribution of this entire article is permitted in any medium,
11702   provided this notice is preserved.
11703
11704   These pages are [18]maintained by the GCC team. Last modified
11705   2022-10-26.
11706
11707References
11708
11709   1. http://www.gnu.org/
11710   2. http://gcc.gnu.org/gcc-4.5/changes.html
11711   3. http://gcc.gnu.org/gcc-4.5/changes.html
11712   4. http://gcc.gnu.org/gcc-4.5/changes.html
11713   5. http://gcc.gnu.org/gcc-4.5/changes.html
11714   6. http://gcc.gnu.org/gcc-4.5/changes.html
11715   7. http://gcc.gnu.org/gcc-4.5/buildstat.html
11716   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
11717   9. http://gcc.gnu.org/index.html
11718  10. mailto:gcc@gcc.gnu.org
11719  11. http://gcc.gnu.org/mirrors.html
11720  12. http://gcc.gnu.org/git.html
11721  13. https://gcc.gnu.org/onlinedocs/
11722  14. mailto:gcc-help@gcc.gnu.org
11723  15. mailto:gcc@gcc.gnu.org
11724  16. https://gcc.gnu.org/lists.html
11725  17. https://www.fsf.org/
11726  18. https://gcc.gnu.org/about.html
11727======================================================================
11728http://gcc.gnu.org/gcc-4.5/changes.html
11729
11730                             GCC 4.5 Release Series
11731                        Changes, New Features, and Fixes
11732
11733Caveats
11734
11735     * GCC now requires the [1]MPC library in order to build. See the
11736       [2]prerequisites page for version requirements.
11737     * Support for a number of older systems and recently unmaintained or
11738       untested target ports of GCC has been declared obsolete in GCC 4.5.
11739       Unless there is activity to revive them, the next release of GCC
11740       will have their sources permanently removed.
11741       The following ports for individual systems on particular
11742       architectures have been obsoleted:
11743          + IRIX releases before 6.5 (mips-sgi-irix5*,
11744            mips-sgi-irix6.[0-4])
11745          + Solaris 7 (*-*-solaris2.7)
11746          + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*,
11747            alpha-dec-osf5.0*)
11748          + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions
11749            can be found in the [3]announcement.
11750       Support for the classic POWER architecture implemented in the
11751       original RIOS and RIOS2 processors of the old IBM RS/6000 product
11752       line has been obsoleted in the rs6000 port. This does not affect
11753       the new generation Power and PowerPC architectures.
11754     * Support has been removed for all the [4]configurations obsoleted in
11755       GCC 4.4.
11756     * Support has been removed for the protoize and unprotoize utilities,
11757       obsoleted in GCC 4.4.
11758     * Support has been removed for tuning for Itanium1 (Merced) variants.
11759       Note that code tuned for Itanium2 should also run correctly on
11760       Itanium1.
11761     * GCC now generates unwind info also for epilogues. DWARF debuginfo
11762       generated by GCC now uses more features of DWARF3 than before, and
11763       also some DWARF4 features. GDB older than 7.0 is not able to handle
11764       either of these, so to debug GCC 4.5 generated binaries or
11765       libraries GDB 7.0 or later is needed. You can disable use of DWARF4
11766       features with the -gdwarf-3 -gstrict-dwarf options, or use
11767       -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but
11768       epilogue unwind info is emitted unconditionally whenever unwind
11769       info is emitted.
11770     * On x86 targets, code containing floating-point calculations may run
11771       significantly more slowly when compiled with GCC 4.5 in strict C99
11772       conformance mode than they did with earlier GCC versions. This is
11773       due to stricter standard conformance of the compiler and can be
11774       avoided by using the option -fexcess-precision=fast; also see
11775       [5]below.
11776     * The function attribute noinline no longer prevents GCC from cloning
11777       the function. A new attribute noclone has been introduced for this
11778       purpose. Cloning a function means that it is duplicated and the new
11779       copy is specialized for certain contexts (for example when a
11780       parameter is a known constant).
11781
11782General Optimizer Improvements
11783
11784     * The -save-temps now takes an optional argument. The -save-temps and
11785       -save-temps=cwd switches write the temporary files in the current
11786       working directory based on the original source file. The
11787       -save-temps=obj switch will write files into the directory
11788       specified with the -o option, and the intermediate filenames are
11789       based on the output file. This will allow the user to get the
11790       compiler intermediate files when doing parallel builds without two
11791       builds of the same filename located in different directories from
11792       interfering with each other.
11793     * Debugging dumps are now created in the same directory as the object
11794       file rather than in the current working directory. This allows the
11795       user to get debugging dumps when doing parallel builds without two
11796       builds of the same filename interfering with each other.
11797     * GCC has been integrated with the MPC library. This allows GCC to
11798       evaluate complex arithmetic at compile time [6]more accurately. It
11799       also allows GCC to evaluate calls to complex built-in math
11800       functions having constant arguments and replace them at compile
11801       time with their mathematically equivalent results. In doing so, GCC
11802       can generate correct results regardless of the math library
11803       implementation or floating point precision of the host platform.
11804       This also allows GCC to generate identical results regardless of
11805       whether one compiles in native or cross-compile configurations to a
11806       particular target. The following built-in functions take advantage
11807       of this new capability: cacos, cacosh, casin, casinh, catan,
11808       catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan,
11809       and ctanh. The float and long double variants of these functions
11810       (e.g. csinf and csinl) are also handled.
11811     * A new link-time optimizer has been added ([7]-flto). When this
11812       option is used, GCC generates a bytecode representation of each
11813       input file and writes it to specially-named sections in each object
11814       file. When the object files are linked together, all the function
11815       bodies are read from these named sections and instantiated as if
11816       they had been part of the same translation unit. This enables
11817       interprocedural optimizations to work across different files (and
11818       even different languages), potentially improving the performance of
11819       the generated code. To use the link-timer optimizer, -flto needs to
11820       be specified at compile time and during the final link. If the
11821       program does not require any symbols to be exported, it is possible
11822       to combine -flto and the experimental [8]-fwhopr with
11823       [9]-fwhole-program to allow the interprocedural optimizers to use
11824       more aggressive assumptions.
11825     * The automatic parallelization pass was enhanced to support
11826       parallelization of outer loops.
11827     * Automatic parallelization can be enabled as part of Graphite. In
11828       addition to -ftree-parallelize-loops=, specify
11829       -floop-parallelize-all to enable the Graphite-based optimization.
11830     * The infrastructure for optimizing based on [10]restrict qualified
11831       pointers has been rewritten and should result in code generation
11832       improvements. Optimizations based on restrict qualified pointers
11833       are now also available when using -fno-strict-aliasing.
11834     * There is a new optimization pass that attempts to change prototype
11835       of functions to avoid unused parameters, pass only relevant parts
11836       of structures and turn arguments passed by reference to arguments
11837       passed by value when possible. It is enabled by -O2 and above as
11838       well as -Os and can be manually invoked using the new command-line
11839       switch -fipa-sra.
11840     * GCC now optimize exception handling code. In particular cleanup
11841       regions that are proved to not have any effect are optimized out.
11842
11843New Languages and Language specific improvements
11844
11845  All languages
11846
11847     * The -fshow-column option is now on by default. This means error
11848       messages now have a column associated with them.
11849
11850  Ada
11851
11852     * Compilation of programs heavily using discriminated record types
11853       with variant parts has been sped up and generates more compact
11854       code.
11855     * Stack checking now works reasonably well on most plaforms. In some
11856       specific cases, stack overflows may still fail to be detected, but
11857       a compile-time warning will be issued for these cases.
11858
11859  C family
11860
11861     * If a header named in a #include directive is not found, the
11862       compiler exits immediately. This avoids a cascade of errors arising
11863       from declarations expected to be found in that header being
11864       missing.
11865     * A new built-in function __builtin_unreachable() has been added that
11866       tells the compiler that control will never reach that point. It may
11867       be used after asm statements that terminate by transferring control
11868       elsewhere, and in other places that are known to be unreachable.
11869     * The -Wlogical-op option now warns for logical expressions such as
11870       (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be
11871       mistakes. This option is disabled by default.
11872     * An asm goto feature has been added to allow asm statements that
11873       jump to C labels.
11874     * C++0x raw strings are supported for C++ and for C with -std=gnu99.
11875     * The deprecated attribute now takes an optional string argument, for
11876       example, __attribute__((deprecated("text string"))), that will be
11877       printed together with the deprecation warning.
11878
11879  C
11880
11881     * The -Wenum-compare option, which warns when comparing values of
11882       different enum types, now works for C. It formerly only worked for
11883       C++. This warning is enabled by -Wall. It may be avoided by using a
11884       type cast.
11885     * The -Wcast-qual option now warns about casts which are unsafe in
11886       that they permit const-correctness to be violated without further
11887       warnings. Specifically, it warns about cases where a qualifier is
11888       added when all the lower types are not const. For example, it warns
11889       about a cast from char ** to const char **.
11890     * The -Wc++-compat option is significantly improved. It issues new
11891       warnings for:
11892          + Using C++ reserved operator names as identifiers.
11893          + Conversions to enum types without explicit casts.
11894          + Using va_arg with an enum type.
11895          + Using different enum types in the two branches of ?:.
11896          + Using ++ or -- on a variable of enum type.
11897          + Using the same name as both a struct, union or enum tag and a
11898            typedef, unless the typedef refers to the tagged type itself.
11899          + Using a struct, union, or enum which is defined within another
11900            struct or union.
11901          + A struct field defined using a typedef if there is a field in
11902            the struct, or an enclosing struct, whose name is the typedef
11903            name.
11904          + Duplicate definitions at file scope.
11905          + Uninitialized const variables.
11906          + A global variable with an anonymous struct, union, or enum
11907            type.
11908          + Using a string constant to initialize a char array whose size
11909            is the length of the string.
11910     * The new -Wjump-misses-init option warns about cases where a goto or
11911       switch skips the initialization of a variable. This sort of branch
11912       is an error in C++ but not in C. This warning is enabled by
11913       -Wc++-compat.
11914     * GCC now ensures that a C99-conforming <stdint.h> is present on most
11915       targets, and uses information about the types in this header to
11916       implement the Fortran bindings to those types. GCC does not ensure
11917       the presence of such a header, and does not implement the Fortran
11918       bindings, on the following targets: NetBSD, VxWorks, VMS,
11919       SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF.
11920     * GCC now implements C90- and C99-conforming rules for constant
11921       expressions. This may cause warnings or errors for some code using
11922       expressions that can be folded to a constant but are not constant
11923       expressions as defined by ISO C.
11924     * All known target-independent C90 and C90 Amendment 1 conformance
11925       bugs, and all known target-independent C99 conformance bugs not
11926       related to floating point or extended identifiers, have been fixed.
11927     * The C decimal floating point support now includes support for the
11928       FLOAT_CONST_DECIMAL64 pragma.
11929     * The named address space feature from ISO/IEC TR 18037 is now
11930       supported. This is currently only implemented for the SPU
11931       processor.
11932
11933  C++
11934
11935     * Improved [11]experimental support for the upcoming C++0x ISO C++
11936       standard, including support for raw strings, lambda expressions and
11937       explicit type conversion operators.
11938     * When printing the name of a class template specialization, G++ will
11939       now omit any template arguments which come from default template
11940       arguments. This behavior (and the pretty-printing of function
11941       template specializations as template signature and arguments) can
11942       be disabled with the -fno-pretty-templates option.
11943     * Access control is now applied to typedef names used in a template,
11944       which may cause G++ to reject some ill-formed code that was
11945       accepted by earlier releases. The -fno-access-control option can be
11946       used as a temporary workaround until the code is corrected.
11947     * Compilation time for code that uses templates should now scale
11948       linearly with the number of instantiations rather than
11949       quadratically, as template instantiations are now looked up using
11950       hash tables.
11951     * Declarations of functions that look like builtin declarations of
11952       library functions are only considered to be redeclarations if they
11953       are declared with extern "C". This may cause problems with code
11954       that omits extern "C" on hand-written declarations of C library
11955       functions such as abort or memcpy. Such code is ill-formed, but was
11956       accepted by earlier releases.
11957     * Diagnostics that used to complain about passing non-POD types to
11958       ... or jumping past the declaration of a non-POD variable now check
11959       for triviality rather than PODness, as per C++0x.
11960     * In C++0x mode local and anonymous classes are now allowed as
11961       template arguments, and in declarations of variables and functions
11962       with linkage, so long as any such declaration that is used is also
11963       defined ([12]DR 757).
11964     * Labels may now have attributes, as has been permitted for a while
11965       in C. This is only permitted when the label definition and the
11966       attribute specifier is followed by a semicolon--i.e., the label
11967       applies to an empty statement. The only useful attribute for a
11968       label is unused.
11969     * G++ now implements [13]DR 176. Previously G++ did not support using
11970       the injected-class-name of a template base class as a type name,
11971       and lookup of the name found the declaration of the template in the
11972       enclosing scope. Now lookup of the name finds the
11973       injected-class-name, which can be used either as a type or as a
11974       template, depending on whether or not the name is followed by a
11975       template argument list. As a result of this change, some code that
11976       was previously accepted may be ill-formed because
11977         1. The injected-class-name is not accessible because it's from a
11978            private base, or
11979         2. The injected-class-name cannot be used as an argument for a
11980            template template parameter.
11981       In either of these cases, the code can be fixed by adding a
11982       nested-name-specifier to explicitly name the template. The first
11983       can be worked around with -fno-access-control; the second is only
11984       rejected with -pedantic.
11985     * A new standard mangling for SIMD vector types has been added, to
11986       avoid name clashes on systems with vectors of varying length. By
11987       default the compiler still uses the old mangling, but emits aliases
11988       with the new mangling on targets that support strong aliases. Users
11989       can switch over entirely to the new mangling with -fabi-version=4
11990       or -fabi-version=0. -Wabi will now warn about code that uses the
11991       old mangling.
11992     * The command-line option -ftemplate-depth-N is now written as
11993       -ftemplate-depth=N and the old form is deprecated.
11994     * Conversions between NULL and non-pointer types are now warned by
11995       default. The new option -Wno-conversion-null disables these
11996       warnings. Previously these warnings were only available when using
11997       -Wconversion explicitly.
11998
11999    Runtime Library (libstdc++)
12000
12001     * Improved experimental support for the upcoming ISO C++ standard,
12002       C++0x, including:
12003          + Support for <future>, <functional>, and <random>.
12004          + Existing facilities now exploit explicit operators and the
12005            newly implemented core C++0x features.
12006          + The header <cstdatomic> has been renamed to <atomic>.
12007     * An experimental [14]profile mode has been added. This is an
12008       implementation of many C++ standard library constructs with an
12009       additional analysis layer that gives performance improvement advice
12010       based on recognition of suboptimal usage patterns. For example,
12011#include <vector>
12012int main()
12013{
12014  std::vector<int> v;
12015  for (int k = 0; k < 1024; ++k)
12016    v.insert(v.begin(), k);
12017}
12018
12019       When instrumented via the profile mode, can return suggestions
12020       about the initial size and choice of the container used as follows:
12021vector-to-list: improvement = 5: call stack = 0x804842c ...
12022    : advice = change std::vector to std::list
12023vector-size: improvement = 3: call stack = 0x804842c ...
12024    : advice = change initial container size from 0 to 1024
12025
12026       These constructs can be substituted for the normal libstdc++
12027       constructs on a piecemeal basis, or all existing components can be
12028       transformed via the -D_GLIBCXX_PROFILE macro.
12029     * [15]Support for decimal floating-point arithmetic (aka ISO C++ TR
12030       24733) has been added. This support is in header file
12031       <decimal/decimal>, uses namespace std::decimal, and includes
12032       classes decimal32, decimal64, and decimal128.
12033     * Sources have been audited for application of function attributes
12034       nothrow, const, pure, and noreturn.
12035     * Python pretty-printers have been added for many standard library
12036       components that simplify the internal representation and present a
12037       more intuitive view of components when used with
12038       appropriately-advanced versions of GDB. For more information,
12039       please consult the more [16]detailed description.
12040     * The default behavior for comparing typeinfo names has changed, so
12041       in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero.
12042     * The new -static-libstdc++ option directs g++ to link the C++
12043       library statically, even if the default would normally be to link
12044       it dynamically.
12045
12046  Fortran
12047
12048     * The COMMON default padding has been changed - instead of adding the
12049       padding before a variable it is now added afterwards, which
12050       increases the compatibility with other vendors and helps to obtain
12051       the correct output in some cases. Cf. also the -falign-commons
12052       option ([17]added in 4.4).
12053     * The -finit-real= option now also supports the value snan for
12054       signaling not-a-number; to be effective, one additionally needs to
12055       enable trapping (e.g. via -ffpe-trap=). Note: Compile-time
12056       optimizations can turn a signaling NaN into a quiet one.
12057     * The new option -fcheck= has been added with the options bounds,
12058       array-temps, do, pointer, and recursive. The bounds and array-temps
12059       options are equivalent to -fbounds-check and
12060       -fcheck-array-temporaries. The do option checks for invalid
12061       modification of loop iteration variables, and the recursive option
12062       tests for recursive calls to subroutines/functions which are not
12063       marked as recursive. With pointer pointer association checks in
12064       calls are performed; however, neither undefined pointers nor
12065       pointers in expressions are handled. Using -fcheck=all enables all
12066       these run-time checks.
12067     * The run-time checking -fcheck=bounds now warns about invalid string
12068       lengths of character dummy arguments. Additionally, more
12069       compile-time checks have been added.
12070     * The new option [18]-fno-protect-parens has been added; if set, the
12071       compiler may reorder REAL and COMPLEX expressions without regard to
12072       parentheses.
12073     * GNU Fortran no longer links against libgfortranbegin. As before,
12074       MAIN__ (assembler symbol name) is the actual Fortran main program,
12075       which is invoked by the main function. However, main is now
12076       generated and put in the same object file as MAIN__. For the time
12077       being, libgfortranbegin still exists for backward compatibility.
12078       For details see the new [19]Mixed-Language Programming chapter in
12079       the manual.
12080     * The I/O library was restructured for performance and cleaner code.
12081     * Array assignments and WHERE are now run in parallel when OpenMP's
12082       WORKSHARE is used.
12083     * The experimental option -fwhole-file was added. The option allows
12084       whole-file checking of procedure arguments and allows for better
12085       optimizations. It can also be used with -fwhole-program, which is
12086       now also supported in gfortran.
12087     * More Fortran 2003 and Fortran 2008 mathematical functions can now
12088       be used as initialization expressions.
12089     * Some extended attributes such as STDCALL are now supported via the
12090       [20]GCC$ compiler directive.
12091     * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN
12092       intrinsic behaves now as if zero were always positive.
12093     * For legacy compatibiliy: On Cygwin and MinGW, the special files
12094       CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now
12095       supported.
12096     * Fortran 2003 support has been extended:
12097          + Procedure-pointer function results and procedure-pointer
12098            components (including PASS),
12099          + allocatable scalars (experimental),
12100          + DEFERRED type-bound procedures,
12101          + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements
12102            have been implemented.
12103          + The ALLOCATE statement supports type-specs and the SOURCE=
12104            argument.
12105          + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC
12106            type-bound procedure (i.e. as type-bound operators).
12107          + Rounding (ROUND=, RZ, ...) for output is now supported.
12108          + The INT_FAST{8,16,32,64,128}_T kind type parameters of the
12109            intrinsic module ISO_C_BINDING are now supported, except for
12110            the targets listed above as ones where GCC does not have
12111            <stdint.h> type information.
12112          + Extensible derived types with type-bound procedure or
12113            procedure pointer with PASS attribute now have to use CLASS in
12114            line with the Fortran 2003 standard; the workaround to use
12115            TYPE is no longer supported.
12116          + [21]Experimental, incomplete support for polymorphism,
12117            including CLASS, SELECT TYPE and dynamic dispatch of
12118            type-bound procedure calls. Some features do not work yet such
12119            as unlimited polymorphism (CLASS(*)).
12120     * Fortran 2008 support has been extended:
12121          + The OPEN statement now supports the NEWUNIT= option, which
12122            returns a unique file unit, thus preventing inadvertent use of
12123            the same unit in different parts of the program.
12124          + Support for unlimited format items has been added.
12125          + The INT{8,16,32} and REAL{32,64,128} kind type parameters of
12126            the intrinsic module ISO_FORTRAN_ENV are now supported.
12127          + Using complex arguments with TAN, SINH, COSH, TANH, ASIN,
12128            ACOS, and ATAN is now possible; the functions ASINH, ACOSH,
12129            and ATANH have been added (for real and complex arguments) and
12130            ATAN(Y,X) is now an alias for ATAN2(Y,X).
12131          + The BLOCK construct has been implemented.
12132
12133New Targets and Target Specific Improvements
12134
12135  AIX
12136
12137     * Full cross-toolchain support now available with GNU Binutils
12138
12139  ARM
12140
12141     * GCC now supports the Cortex-M0 and Cortex-A5 processors.
12142     * GCC now supports the ARM v7E-M architecture.
12143     * GCC now supports VFPv4-based FPUs and FPUs with
12144       single-precision-only VFP.
12145     * GCC has many improvements to optimization for other ARM processors,
12146       including scheduling support for the integer pipeline on Cortex-A9.
12147     * GCC now supports the IEEE 754-2008 half-precision floating-point
12148       type, and a variant ARM-specific half-precision type. This type is
12149       specified using __fp16, with the layout determined by
12150       -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and
12151       VFPv4 half-precision instructions will be used.
12152     * GCC now supports the variant of AAPCS that uses VFP registers for
12153       parameter passing and return values.
12154
12155  AVR
12156
12157     * The -mno-tablejump option has been removed because it has the same
12158       effect as the -fno-jump-tables option.
12159     * Added support for these new AVR devices:
12160          + ATmega8U2
12161          + ATmega16U2
12162          + ATmega32U2
12163
12164  IA-32/x86-64
12165
12166     * GCC now will set the default for -march= based on the configure
12167       target.
12168     * GCC now supports handling floating-point excess precision arising
12169       from use of the x87 floating-point unit in a way that conforms to
12170       ISO C99. This is enabled with -fexcess-precision=standard and with
12171       standards conformance options such as -std=c99, and may be disabled
12172       using -fexcess-precision=fast.
12173     * Support for the Intel Atom processor is now available through the
12174       -march=atom and -mtune=atom options.
12175     * A new -mcrc32 option is now available to enable crc32 intrinsics.
12176     * A new -mmovbe option is now available to enable GCC to use the
12177       movbe instruction to implement __builtin_bswap32 and
12178       __builtin_bswap64.
12179     * SSE math now can be enabled by default at configure time with the
12180       new --with-fpmath=sse option.
12181     * There is a new intrinsic header file, <x86intrin.h>. It should be
12182       included before using any IA-32/x86-64 intrinsics.
12183     * Support for the XOP, FMA4, and LWP instruction sets for the AMD
12184       Orochi processors are now available with the -mxop, -mfma4, and
12185       -mlwp options.
12186     * The -mabm option enables GCC to use the popcnt and lzcnt
12187       instructions on AMD processors.
12188     * The -mpopcnt option enables GCC to use the popcnt instructions on
12189       both AMD and Intel processors.
12190
12191  M68K/ColdFire
12192
12193     * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x
12194       and 5441x devices.
12195     * GCC now supports thread-local storage (TLS) on M68K and ColdFire
12196       processors.
12197
12198  MeP
12199
12200   Support has been added for the Toshiba Media embedded Processor (MeP,
12201   or mep-elf) embedded target.
12202
12203  MIPS
12204
12205     * GCC now supports MIPS 1004K processors.
12206     * GCC can now be configured with options --with-arch-32,
12207       --with-arch-64, --with-tune-32 and --with-tune-64 to control the
12208       default optimization separately for 32-bit and 64-bit modes.
12209     * MIPS targets now support an alternative _mcount interface, in which
12210       register $12 points to the function's save slot for register $31.
12211       This interface is selected by the -mcount-ra-address option; see
12212       the documentation for more details.
12213     * GNU/Linux targets can now generate read-only .eh_frame sections.
12214       This optimization requires GNU binutils 2.20 or above, and is only
12215       available if GCC is configured with a suitable version of binutils.
12216     * GNU/Linux targets can now attach special relocations to indirect
12217       calls, so that the linker can turn them into direct jumps or
12218       branches. This optimization requires GNU binutils 2.20 or later,
12219       and is automatically selected if GCC is configured with an
12220       appropriate version of binutils. It can be explicitly enabled or
12221       disabled using the -mrelax-pic-calls command-line option.
12222     * GCC now generates more heavily-optimized atomic operations on
12223       Octeon processors.
12224     * MIPS targets now support the -fstack-protector option.
12225     * GCC now supports an -msynci option, which specifies that synci is
12226       enough to flush the instruction cache, without help from the
12227       operating system. GCC uses this information to optimize
12228       automatically-generated cache flush operations, such as those used
12229       for nested functions in C. There is also a --with-synci
12230       configure-time option, which makes -msynci the default.
12231     * GCC supports four new function attributes for interrupt handlers:
12232       interrupt, use_shadow_register_set, keep_interrupts_masked and
12233       use_debug_exception_return. See the documentation for more details
12234       about these attributes.
12235
12236  RS/6000 (POWER/PowerPC)
12237
12238     * GCC now supports the Power ISA 2.06, which includes the VSX
12239       instructions that add vector 64-bit floating point support, new
12240       population count instructions, and conversions between floating
12241       point and unsigned types.
12242     * Support for the power7 processor is now available through the
12243       -mcpu=power7 and -mtune=power7.
12244     * GCC will now vectorize loops that contain simple math functions
12245       like copysign when generating code for altivec or VSX targets.
12246     * Support for the A2 processor is now available through the -mcpu=a2
12247       and -mtune=a2 options.
12248     * Support for the 476 processor is now available through the
12249       -mcpu={476,476fp} and -mtune={476,476fp} options.
12250     * Support for the e500mc64 processor is now available through the
12251       -mcpu=e500mc64 and -mtune=e500mc64 options.
12252     * GCC can now be configured with options --with-cpu-32,
12253       --with-cpu-64, --with-tune-32 and --with-tune-64 to control the
12254       default optimization separately for 32-bit and 64-bit modes.
12255     * Starting with GCC 4.5.4, vectors of type vector long long or vector
12256       long are passed and returned in the same method as other vectors
12257       with the VSX instruction set. Previously the GCC compiler did not
12258       adhere to the ABI for 128-bit vectors with 64-bit integer base
12259       types (PR 48857). This is also fixed in the GCC 4.6.1 release.
12260
12261  RX
12262
12263   Support has been added for the Renesas RX Processor (rx-elf) target.
12264
12265Operating Systems
12266
12267  Windows (Cygwin and MinGW)
12268
12269     * GCC now installs all the major language runtime libraries as DLLs
12270       when configured with the --enable-shared option.
12271     * GCC now makes use of the new support for aligned common variables
12272       in versions of binutils >= 2.20 to fix bugs in the support for SSE
12273       data types.
12274     * Improvements to the libffi support library increase the reliability
12275       of code generated by GCJ on all Windows platforms. Libgcj is
12276       enabled by default for the first time.
12277     * Libtool improvements simplify installation by placing the generated
12278       DLLs in the correct binaries directory.
12279     * Numerous other minor bugfixes and improvements, and substantial
12280       enhancements to the Fortran language support library.
12281
12282   >
12283
12284Other significant improvements
12285
12286  Plugins
12287
12288     * It is now possible to extend the compiler without having to modify
12289       its source code. A new option -fplugin=file.so tells GCC to load
12290       the shared object file.so and execute it as part of the compiler.
12291       The internal documentation describes the details on how plugins can
12292       interact with the compiler.
12293
12294  Installation changes
12295
12296     * The move to newer autotools changed default installation
12297       directories and switches to control them: The --with-datarootdir,
12298       --with-docdir, --with-pdfdir, and --with-htmldir switches are not
12299       used any more. Instead, you can now use --datarootdir, --docdir,
12300       --htmldir, and --pdfdir. The default installation directories have
12301       changed as follows according to the GNU Coding Standards:
12302
12303       datarootdir read-only architecture-independent data root [PREFIX/share]
12304       localedir   locale-specific message catalogs [DATAROOTDIR/locale]
12305       docdir      documentation root [DATAROOTDIR/doc/PACKAGE]
12306       htmldir     html documentation [DOCDIR]
12307       dvidir      dvi documentation [DOCDIR]
12308       pdfdir      pdf documentation [DOCDIR]
12309       psdir       ps documentation [DOCDIR]
12310       The following variables have new default values:
12311
12312       datadir read-only architecture-independent data [DATAROOTDIR]
12313       infodir info documentation [DATAROOTDIR/info]
12314       mandir  man documentation [DATAROOTDIR/man]
12315
12316GCC 4.5.1
12317
12318   This is the [22]list of problem reports (PRs) from GCC's bug tracking
12319   system that are known to be fixed in the 4.5.1 release. This list might
12320   not be complete (that is, it is possible that some PRs that have been
12321   fixed are not listed here).
12322
12323  All languages
12324
12325     * GCC's new link-time optimizer ([23]-flto) now also works on a few
12326       non-ELF targets:
12327          + Cygwin (*-cygwin*)
12328          + MinGW (*-mingw*)
12329          + Darwin on x86-64 (x86_64-apple-darwin*)
12330       LTO is not enabled by default for these targets. To enable LTO, you
12331       should configure with the --enable-lto option.
12332
12333GCC 4.5.2
12334
12335   This is the [24]list of problem reports (PRs) from GCC's bug tracking
12336   system that are known to be fixed in the 4.5.2 release. This list might
12337   not be complete (that is, it is possible that some PRs that have been
12338   fixed are not listed here).
12339
12340GCC 4.5.3
12341
12342   This is the [25]list of problem reports (PRs) from GCC's bug tracking
12343   system that are known to be fixed in the 4.5.3 release. This list might
12344   not be complete (that is, it is possible that some PRs that have been
12345   fixed are not listed here).
12346
12347   On the PowerPC compiler, the Altivec builtin functions vec_ld and
12348   vec_st have been modified to generate the Altivec memory instructions
12349   LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5
12350   release, these builtin functions were changed to generate VSX memory
12351   reference instructions instead of Altivec memory instructions, but
12352   there are differences between the two instructions. If the VSX
12353   instruction set is available, you can now use the new builtin functions
12354   vec_vsx_ld and vec_vsx_st which always generates the VSX memory
12355   instructions.
12356
12357GCC 4.5.4
12358
12359   This is the [26]list of problem reports (PRs) from GCC's bug tracking
12360   system that are known to be fixed in the 4.5.4 release. This list might
12361   not be complete (that is, it is possible that some PRs that have been
12362   fixed are not listed here).
12363
12364
12365    For questions related to the use of GCC, please consult these web
12366    pages and the [27]GCC manuals. If that fails, the
12367    [28]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12368    web pages and the development of GCC are welcome on our developer
12369    list at [29]gcc@gcc.gnu.org. All of [30]our lists have public
12370    archives.
12371
12372   Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and
12373   distribution of this entire article is permitted in any medium,
12374   provided this notice is preserved.
12375
12376   These pages are [32]maintained by the GCC team. Last modified
12377   2023-01-15.
12378
12379References
12380
12381   1. https://www.multiprecision.org/mpc/
12382   2. https://gcc.gnu.org/install/prerequisites.html
12383   3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html
12384   4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted
12385   5. http://gcc.gnu.org/gcc-4.5/changes.html#x86
12386   6. https://gcc.gnu.org/PR30789
12387   7. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
12388   8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802
12389   9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800
12390  10. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html
12391  11. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html
12392  12. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757
12393  13. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176
12394  14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html
12395  15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733
12396  16. https://sourceware.org/gdb/wiki/STLSupport
12397  17. http://gcc.gnu.org/gcc-4.4/changes.html
12398  18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
12399  19. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html
12400  20. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html
12401  21. https://gcc.gnu.org/wiki/OOP
12402  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1
12403  23. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
12404  24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2
12405  25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3
12406  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4
12407  27. https://gcc.gnu.org/onlinedocs/
12408  28. mailto:gcc-help@gcc.gnu.org
12409  29. mailto:gcc@gcc.gnu.org
12410  30. https://gcc.gnu.org/lists.html
12411  31. https://www.fsf.org/
12412  32. https://gcc.gnu.org/about.html
12413======================================================================
12414http://gcc.gnu.org/gcc-4.4/index.html
12415
12416                             GCC 4.4 Release Series
12417
12418   This release series is no longer maintained.
12419
12420   March 13, 2012
12421
12422   The [1]GNU project and the GCC developers are pleased to announce the
12423   release of GCC 4.4.7.
12424
12425   This release is a bug-fix release, containing fixes for regressions in
12426   GCC 4.4.6 relative to previous releases of GCC.
12427
12428Release History
12429
12430   GCC 4.4.7
12431          March 13, 2012 ([2]changes)
12432
12433   GCC 4.4.6
12434          April 16, 2011 ([3]changes)
12435
12436   GCC 4.4.5
12437          October 1, 2010 ([4]changes)
12438
12439   GCC 4.4.4
12440          April 29, 2010 ([5]changes)
12441
12442   GCC 4.4.3
12443          January 21, 2010 ([6]changes)
12444
12445   GCC 4.4.2
12446          October 15, 2009 ([7]changes)
12447
12448   GCC 4.4.1
12449          July 22, 2009 ([8]changes)
12450
12451   GCC 4.4.0
12452          April 21, 2009 ([9]changes)
12453
12454References and Acknowledgements
12455
12456   GCC used to stand for the GNU C Compiler, but since the compiler
12457   supports several other languages aside from C, it now stands for the
12458   GNU Compiler Collection.
12459
12460   A list of [10]successful builds is updated as new information becomes
12461   available.
12462
12463   The GCC developers would like to thank the numerous people that have
12464   contributed new features, improvements, bug fixes, and other changes as
12465   well as test results to GCC. This [11]amazing group of volunteers is
12466   what makes GCC successful.
12467
12468   For additional information about GCC please refer to the [12]GCC
12469   project web site or contact the [13]GCC development mailing list.
12470
12471   To obtain GCC please use [14]our mirror sites or [15]our version
12472   control system.
12473
12474
12475    For questions related to the use of GCC, please consult these web
12476    pages and the [16]GCC manuals. If that fails, the
12477    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12478    web pages and the development of GCC are welcome on our developer
12479    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
12480    archives.
12481
12482   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
12483   distribution of this entire article is permitted in any medium,
12484   provided this notice is preserved.
12485
12486   These pages are [21]maintained by the GCC team. Last modified
12487   2022-10-26.
12488
12489References
12490
12491   1. http://www.gnu.org/
12492   2. http://gcc.gnu.org/gcc-4.4/changes.html
12493   3. http://gcc.gnu.org/gcc-4.4/changes.html
12494   4. http://gcc.gnu.org/gcc-4.4/changes.html
12495   5. http://gcc.gnu.org/gcc-4.4/changes.html
12496   6. http://gcc.gnu.org/gcc-4.4/changes.html
12497   7. http://gcc.gnu.org/gcc-4.4/changes.html
12498   8. http://gcc.gnu.org/gcc-4.4/changes.html
12499   9. http://gcc.gnu.org/gcc-4.4/changes.html
12500  10. http://gcc.gnu.org/gcc-4.4/buildstat.html
12501  11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12502  12. http://gcc.gnu.org/index.html
12503  13. mailto:gcc@gcc.gnu.org
12504  14. http://gcc.gnu.org/mirrors.html
12505  15. http://gcc.gnu.org/git.html
12506  16. https://gcc.gnu.org/onlinedocs/
12507  17. mailto:gcc-help@gcc.gnu.org
12508  18. mailto:gcc@gcc.gnu.org
12509  19. https://gcc.gnu.org/lists.html
12510  20. https://www.fsf.org/
12511  21. https://gcc.gnu.org/about.html
12512======================================================================
12513http://gcc.gnu.org/gcc-4.4/changes.html
12514
12515                             GCC 4.4 Release Series
12516                        Changes, New Features, and Fixes
12517
12518   The latest release in the 4.4 release series is [1]GCC 4.4.7.
12519
12520Caveats
12521
12522     * __builtin_stdarg_start has been completely removed from GCC.
12523       Support for <varargs.h> had been deprecated since GCC 4.0. Use
12524       __builtin_va_start as a replacement.
12525     * Some of the errors issued by the C++ front end that could be
12526       downgraded to warnings in previous releases by using -fpermissive
12527       are now warnings by default. They can be converted into errors by
12528       using -pedantic-errors.
12529     * Use of the cpp assertion extension will now emit a warning when
12530       -Wdeprecated or -pedantic is used. This extension has been
12531       deprecated for many years, but never warned about.
12532     * Packed bit-fields of type char were not properly bit-packed on many
12533       targets prior to GCC 4.4. On these targets, the fix in GCC 4.4
12534       causes an ABI change. For example there is no longer a 4-bit
12535       padding between field a and b in this structure:
12536    struct foo
12537    {
12538      char a:4;
12539      char b:8;
12540    } __attribute__ ((packed));
12541       There is a new warning to help identify fields that are affected:
12542    foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4
12543       The warning can be disabled with -Wno-packed-bitfield-compat.
12544     * On ARM EABI targets, the C++ mangling of the va_list type has been
12545       changed to conform to the current revision of the EABI. This does
12546       not affect the libstdc++ library included with GCC.
12547     * The SCOUNT and POS bits of the MIPS DSP control register are now
12548       treated as global. Previous versions of GCC treated these fields as
12549       call-clobbered instead.
12550     * The MIPS port no longer recognizes the h asm constraint. It was
12551       necessary to remove this constraint in order to avoid generating
12552       unpredictable code sequences.
12553       One of the main uses of the h constraint was to extract the high
12554       part of a multiplication on 64-bit targets. For example:
12555    asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y));
12556       You can now achieve the same effect using 128-bit types:
12557    typedef unsigned int uint128_t __attribute__((mode(TI)));
12558    result = ((uint128_t) x * y) >> 64;
12559       The second sequence is better in many ways. For example, if x and y
12560       are constants, the compiler can perform the multiplication at
12561       compile time. If x and y are not constants, the compiler can
12562       schedule the runtime multiplication better than it can schedule an
12563       asm statement.
12564     * Support for a number of older systems and recently unmaintained or
12565       untested target ports of GCC has been declared obsolete in GCC 4.4.
12566       Unless there is activity to revive them, the next release of GCC
12567       will have their sources permanently removed.
12568       The following ports for individual systems on particular
12569       architectures have been obsoleted:
12570          + Generic a.out on IA32 and m68k (i[34567]86-*-aout*,
12571            m68k-*-aout*)
12572          + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*,
12573            armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*,
12574            sh-*-*). This does not affect other more specific targets
12575            using the COFF object format on those architectures, or the
12576            more specific H8300 and SH targets (h8300-*-rtems*,
12577            h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*,
12578            sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks).
12579          + 2BSD on PDP-11 (pdp11-*-bsd)
12580          + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*,
12581            powerpc-ibm-aix4.[12]*)
12582          + Tuning support for Itanium1 (Merced) variants. Note that code
12583            tuned for Itanium2 should also run correctly on Itanium1.
12584     * The protoize and unprotoize utilities have been obsoleted and will
12585       be removed in GCC 4.5. These utilities have not been installed by
12586       default since GCC 3.0.
12587     * Support has been removed for all the [2]configurations obsoleted in
12588       GCC 4.3.
12589     * Unknown -Wno-* options are now silently ignored by GCC if no other
12590       diagnostics are issued. If other diagnostics are issued, then GCC
12591       warns about the unknown options.
12592     * More information on porting to GCC 4.4 from previous versions of
12593       GCC can be found in the [3]porting guide for this release.
12594
12595General Optimizer Improvements
12596
12597     * A new command-line switch -findirect-inlining has been added. When
12598       turned on it allows the inliner to also inline indirect calls that
12599       are discovered to have known targets at compile time thanks to
12600       previous inlining.
12601     * A new command-line switch -ftree-switch-conversion has been added.
12602       This new pass turns simple initializations of scalar variables in
12603       switch statements into initializations from a static array, given
12604       that all the values are known at compile time and the ratio between
12605       the new array size and the original switch branches does not exceed
12606       the parameter --param switch-conversion-max-branch-ratio (default
12607       is eight).
12608     * A new command-line switch -ftree-builtin-call-dce has been added.
12609       This optimization eliminates unnecessary calls to certain builtin
12610       functions when the return value is not used, in cases where the
12611       calls can not be eliminated entirely because the function may set
12612       errno. This optimization is on by default at -O2 and above.
12613     * A new command-line switch -fconserve-stack directs the compiler to
12614       minimize stack usage even if it makes the generated code slower.
12615       This affects inlining decisions.
12616     * When the assembler supports it, the compiler will now emit unwind
12617       information using assembler .cfi directives. This makes it possible
12618       to use such directives in inline assembler code. The new option
12619       -fno-dwarf2-cfi-asm directs the compiler to not use .cfi
12620       directives.
12621     * The [4]Graphite branch has been merged. This merge has brought in a
12622       new framework for loop optimizations based on a polyhedral
12623       intermediate representation. These optimizations apply to all the
12624       languages supported by GCC. The following new code transformations
12625       are available in GCC 4.4:
12626          + -floop-interchange performs loop interchange transformations
12627            on loops. Interchanging two nested loops switches the inner
12628            and outer loops. For example, given a loop like:
12629          DO J = 1, M
12630            DO I = 1, N
12631              A(J, I) = A(J, I) * C
12632            ENDDO
12633          ENDDO
12634
12635            loop interchange will transform the loop as if the user had
12636            written:
12637          DO I = 1, N
12638            DO J = 1, M
12639              A(J, I) = A(J, I) * C
12640            ENDDO
12641          ENDDO
12642
12643            which can be beneficial when N is larger than the caches,
12644            because in Fortran, the elements of an array are stored in
12645            memory contiguously by column, and the original loop iterates
12646            over rows, potentially creating at each access a cache miss.
12647          + -floop-strip-mine performs loop strip mining transformations
12648            on loops. Strip mining splits a loop into two nested loops.
12649            The outer loop has strides equal to the strip size and the
12650            inner loop has strides of the original loop within a strip.
12651            For example, given a loop like:
12652          DO I = 1, N
12653            A(I) = A(I) + C
12654          ENDDO
12655
12656            loop strip mining will transform the loop as if the user had
12657            written:
12658          DO II = 1, N, 4
12659            DO I = II, min (II + 3, N)
12660              A(I) = A(I) + C
12661            ENDDO
12662          ENDDO
12663
12664          + -floop-block performs loop blocking transformations on loops.
12665            Blocking strip mines each loop in the loop nest such that the
12666            memory accesses of the element loops fit inside caches. For
12667            example, given a loop like:
12668          DO I = 1, N
12669            DO J = 1, M
12670              A(J, I) = B(I) + C(J)
12671            ENDDO
12672          ENDDO
12673
12674            loop blocking will transform the loop as if the user had
12675            written:
12676          DO II = 1, N, 64
12677            DO JJ = 1, M, 64
12678              DO I = II, min (II + 63, N)
12679                DO J = JJ, min (JJ + 63, M)
12680                  A(J, I) = B(I) + C(J)
12681                ENDDO
12682              ENDDO
12683            ENDDO
12684          ENDDO
12685
12686            which can be beneficial when M is larger than the caches,
12687            because the innermost loop will iterate over a smaller amount
12688            of data that can be kept in the caches.
12689     * A new register allocator has replaced the old one. It is called
12690       integrated register allocator (IRA) because coalescing, register
12691       live range splitting, and hard register preferencing are done
12692       on-the-fly during coloring. It also has better integration with the
12693       reload pass. IRA is a regional register allocator which uses modern
12694       Chaitin-Briggs coloring instead of Chow's priority coloring used in
12695       the old register allocator. More info about IRA internals and
12696       options can be found in the GCC manuals.
12697     * A new instruction scheduler and software pipeliner, based on the
12698       selective scheduling approach, has been added. The new pass
12699       performs instruction unification, register renaming, substitution
12700       through register copies, and speculation during scheduling. The
12701       software pipeliner is able to pipeline non-countable loops. The new
12702       pass is targeted at scheduling-eager in-order platforms. In GCC 4.4
12703       it is available for the Intel Itanium platform working by default
12704       as the second scheduling pass (after register allocation) at the
12705       -O3 optimization level.
12706     * When using -fprofile-generate with a multi-threaded program, the
12707       profile counts may be slightly wrong due to race conditions. The
12708       new -fprofile-correction option directs the compiler to apply
12709       heuristics to smooth out the inconsistencies. By default the
12710       compiler will give an error message when it finds an inconsistent
12711       profile.
12712     * The new -fprofile-dir=PATH option permits setting the directory
12713       where profile data files are stored when using -fprofile-generate
12714       and friends, and the directory used when reading profile data files
12715       using -fprofile-use and friends.
12716
12717New warning options
12718
12719     * The new -Wframe-larger-than=NUMBER option directs GCC to emit a
12720       warning if any stack frame is larger than NUMBER bytes. This may be
12721       used to help ensure that code fits within a limited amount of stack
12722       space.
12723     * The command-line option -Wlarger-than-N is now written as
12724       -Wlarger-than=N and the old form is deprecated.
12725     * The new -Wno-mudflap option disables warnings about constructs
12726       which can not be instrumented when using -fmudflap.
12727
12728New Languages and Language specific improvements
12729
12730     * Version 3.0 of the OpenMP specification is now supported for the C,
12731       C++, and Fortran compilers.
12732     * New character data types, per [5]TR 19769: New character types in
12733       C, are now supported for the C compiler in -std=gnu99 mode, as
12734       __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in
12735       -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too.
12736
12737  C family
12738
12739     * A new optimize attribute was added to allow programmers to change
12740       the optimization level and particular optimization options for an
12741       individual function. You can also change the optimization options
12742       via the GCC optimize pragma for functions defined after the pragma.
12743       The GCC push_options pragma and the GCC pop_options pragma allow
12744       you temporarily save and restore the options used. The GCC
12745       reset_options pragma restores the options to what was specified on
12746       the command line.
12747     * Uninitialized warnings do not require enabling optimization
12748       anymore, that is, -Wuninitialized can be used together with -O0.
12749       Nonetheless, the warnings given by -Wuninitialized will probably be
12750       more accurate if optimization is enabled.
12751     * -Wparentheses now warns about expressions such as (!x | y) and (!x
12752       & y). Using explicit parentheses, such as in ((!x) | y), silences
12753       this warning.
12754     * -Wsequence-point now warns within if, while,do while and for
12755       conditions, and within for begin/end expressions.
12756     * A new option -dU is available to dump definitions of preprocessor
12757       macros that are tested or expanded.
12758
12759  C++
12760
12761     * [6]Improved experimental support for the upcoming ISO C++ standard,
12762       C++0x. Including support for auto, inline namespaces, generalized
12763       initializer lists, defaulted and deleted functions, new character
12764       types, and scoped enums.
12765     * Those errors that may be downgraded to warnings to build legacy
12766       code now mention -fpermissive when -fdiagnostics-show-option is
12767       enabled.
12768     * -Wconversion now warns if the result of a static_cast to enumeral
12769       type is unspecified because the value is outside the range of the
12770       enumeral type.
12771     * -Wuninitialized now warns if a non-static reference or non-static
12772       const member appears in a class without constructors.
12773     * G++ now properly implements value-initialization, so objects with
12774       an initializer of () and an implicitly defined default constructor
12775       will be zero-initialized before the default constructor is called.
12776
12777    Runtime Library (libstdc++)
12778
12779     * Improved experimental support for the upcoming ISO C++ standard,
12780       C++0x, including:
12781          + Support for <chrono>, <condition_variable>, <cstdatomic>,
12782            <forward_list>, <initializer_list>, <mutex>, <ratio>,
12783            <system_error>, and <thread>.
12784          + unique_ptr, <algorithm> additions, exception propagation, and
12785            support for the new character types in <string> and <limits>.
12786          + Existing facilities now exploit initializer lists, defaulted
12787            and deleted functions, and the newly implemented core C++0x
12788            features.
12789          + Some standard containers are more efficient together with
12790            stateful allocators, i.e., no allocator is constructed on the
12791            fly at element construction time.
12792     * Experimental support for non-standard pointer types in containers.
12793     * The long standing libstdc++/30928 has been fixed for targets
12794       running glibc 2.10 or later.
12795     * As usual, many small and larger bug fixes, in particular quite a
12796       few corner cases in <locale>.
12797
12798  Fortran
12799
12800     * GNU Fortran now employs libcpp directly instead of using cc1 as an
12801       external preprocessor. The [7]-cpp option was added to allow manual
12802       invocation of the preprocessor without relying on filename
12803       extensions.
12804     * The [8]-Warray-temporaries option warns about array temporaries
12805       generated by the compiler, as an aid to optimization.
12806     * The [9]-fcheck-array-temporaries option has been added, printing a
12807       notification at run time, when an array temporary had to be created
12808       for an function argument. Contrary to -Warray-temporaries the
12809       warning is only printed if the array is noncontiguous.
12810     * Improved generation of DWARF debugging symbols
12811     * If using an intrinsic not part of the selected standard (via -std=
12812       and -fall-intrinsics) gfortran will now treat it as if this
12813       procedure were declared EXTERNAL and try to link to a user-supplied
12814       procedure. -Wintrinsics-std will warn whenever this happens. The
12815       now-useless option -Wnonstd-intrinsic was removed.
12816     * The flag -falign-commons has been added to control the alignment of
12817       variables in COMMON blocks, which is enabled by default in line
12818       with previous GCC version. Using -fno-align-commons one can force
12819       commons to be contiguous in memory as required by the Fortran
12820       standard, however, this slows down the memory access. The option
12821       -Walign-commons, which is enabled by default, warns when padding
12822       bytes were added for alignment. The proper solution is to sort the
12823       common objects by decreasing storage size, which avoids the
12824       alignment problems.
12825     * Fortran 2003 support has been extended:
12826          + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is
12827            now supported (except internal reads from/writes to wide
12828            strings). [10]-fbackslash now supports also \unnnn and
12829            \Unnnnnnnn to enter Unicode characters.
12830          + Asynchronous I/O (implemented as synchronous I/O) and the
12831            decimal=, size=, sign=, pad=, blank=, and delim= specifiers
12832            are now supported in I/O statements.
12833          + Support for Fortran 2003 structure constructors and for array
12834            constructor with typespec has been added.
12835          + Procedure Pointers (but not yet as component in derived types
12836            and as function results) are now supported.
12837          + Abstract types, type extension, and type-bound procedures
12838            (both PROCEDURE and GENERIC but not as operators). Note: As
12839            CLASS/polymorphyic types are not implemented, type-bound
12840            procedures with PASS accept as non-standard extension TYPE
12841            arguments.
12842     * Fortran 2008 support has been added:
12843          + The -std=f2008 option and support for the file extensions
12844            .f2008 and .F2008 has been added.
12845          + The g0 format descriptor is now supported.
12846          + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH,
12847            ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED
12848            are now available (some of them existed as GNU extension
12849            before). Note: The hyperbolic functions are not yet supporting
12850            complex arguments and the three- argument version of BESSEL_*N
12851            is not available.
12852          + The bit intrinsics LEADZ and TRAILZ have been added.
12853
12854  Java (GCJ)
12855
12856  Ada
12857
12858     * The Ada runtime now supports multilibs on many platforms including
12859       x86_64, SPARC and PowerPC. Their build is enabled by default.
12860
12861New Targets and Target Specific Improvements
12862
12863  ARM
12864
12865     * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and
12866       Cortex-R4F processors and has many other improvements to
12867       optimization for ARM processors.
12868     * GCC now supports the VFPv3 variant with 16 double-precision
12869       registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been
12870       renamed to -mfpu=vfpv3.
12871     * GCC now supports the -mfix-cortex-m3-ldrd option to work around an
12872       erratum on Cortex-M3 processors.
12873     * GCC now supports the __sync_* atomic operations for ARM EABI
12874       GNU/Linux.
12875     * The section anchors optimization is now enabled by default when
12876       optimizing for ARM.
12877     * GCC now uses a new EABI-compatible profiling interface for EABI
12878       targets. This requires a function __gnu_mcount_nc, which is
12879       provided by GNU libc versions 2.8 and later.
12880
12881  AVR
12882
12883     * The -mno-tablejump option has been deprecated because it has the
12884       same effect as the -fno-jump-tables option.
12885     * Added support for these new AVR devices:
12886          + ATA6289
12887          + ATtiny13A
12888          + ATtiny87
12889          + ATtiny167
12890          + ATtiny327
12891          + ATmega8C1
12892          + ATmega16C1
12893          + ATmega32C1
12894          + ATmega8M1
12895          + ATmega16M1
12896          + ATmega32M1
12897          + ATmega32U4
12898          + ATmega16HVB
12899          + ATmega4HVD
12900          + ATmega8HVD
12901          + ATmega64C1
12902          + ATmega64M1
12903          + ATmega16U4
12904          + ATmega32U6
12905          + ATmega128RFA1
12906          + AT90PWM81
12907          + AT90SCR100
12908          + M3000F
12909          + M3000S
12910          + M3001B
12911
12912  IA-32/x86-64
12913
12914     * Support for Intel AES built-in functions and code generation is
12915       available via -maes.
12916     * Support for Intel PCLMUL built-in function and code generation is
12917       available via -mpclmul.
12918     * Support for Intel AVX built-in functions and code generation is
12919       available via -mavx.
12920     * Automatically align the stack for local variables with alignment
12921       requirement.
12922     * GCC can now utilize the SVML library for vectorizing calls to a set
12923       of C99 functions if -mveclibabi=svml is specified and you link to
12924       an SVML ABI compatible library.
12925     * On x86-64, the ABI has been changed in the following cases to
12926       conform to the x86-64 ABI:
12927          + Passing/returning structures with flexible array member:
12928  struct foo
12929    {
12930      int i;
12931      int flex[];
12932    };
12933          + Passing/returning structures with complex float member:
12934  struct foo
12935    {
12936      int i;
12937      __complex__ float f;
12938    };
12939          + Passing/returning unions with long double member:
12940  union foo
12941    {
12942      int x;
12943      long double ld;
12944    };
12945       Code built with previous versions of GCC that uses any of these is
12946       not compatible with code built with GCC 4.4.0 or later.
12947     * A new target attribute was added to allow programmers to change the
12948       target options like -msse2 or -march=k8 for an individual function.
12949       You can also change the target options via the GCC target pragma
12950       for functions defined after the pragma.
12951     * GCC can now be configured with options --with-arch-32,
12952       --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and
12953       --with-tune-64 to control the default optimization separately for
12954       32-bit and 64-bit modes.
12955
12956  IA-32/IA64
12957
12958     * Support for __float128 (TFmode) IEEE quad type and corresponding
12959       TCmode IEEE complex quad type is available via the soft-fp library
12960       on IA-32/IA64 targets. This includes basic arithmetic operations
12961       (addition, subtraction, negation, multiplication and division) on
12962       __float128 real and TCmode complex values, the full set of IEEE
12963       comparisons between __float128 values, conversions to and from
12964       float, double and long double floating point types, as well as
12965       conversions to and from signed or unsigned integer, signed or
12966       unsigned long integer and signed or unsigned quad (TImode, IA64
12967       only) integer types. Additionally, all operations generate the full
12968       set of IEEE exceptions and support the full set of IEEE rounding
12969       modes.
12970
12971  M68K/ColdFire
12972
12973     * GCC now supports instruction scheduling for ColdFire V1, V3 and V4
12974       processors. (Scheduling support for ColdFire V2 processors was
12975       added in GCC 4.3.)
12976     * GCC now supports the -mxgot option to support programs requiring
12977       many GOT entries on ColdFire.
12978     * The m68k-*-linux-gnu target now builds multilibs by default.
12979
12980  MIPS
12981
12982     * MIPS Technologies have extended the original MIPS SVR4 ABI to
12983       include support for procedure linkage tables (PLTs) and copy
12984       relocations. These extensions allow GNU/Linux executables to use a
12985       significantly more efficient code model than the one defined by the
12986       original ABI.
12987       GCC support for this code model is available via a new command-line
12988       option, -mplt. There is also a new configure-time option,
12989       --with-mips-plt, to make -mplt the default.
12990       The new code model requires support from the assembler, the linker,
12991       and the runtime C library. This support is available in binutils
12992       2.19 and GLIBC 2.9.
12993     * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables
12994       and 32-bit GNU/Linux shared libraries. This feature requires GNU
12995       binutils 2.19 or above.
12996     * Support for RMI's XLR processor is now available through the
12997       -march=xlr and -mtune=xlr options.
12998     * 64-bit targets can now perform 128-bit multiplications inline,
12999       instead of relying on a libgcc function.
13000     * Native GNU/Linux toolchains now support -march=native and
13001       -mtune=native, which select the host processor.
13002     * GCC now supports the R10K, R12K, R14K and R16K processors. The
13003       canonical -march= and -mtune= names for these processors are
13004       r10000, r12000, r14000 and r16000 respectively.
13005     * GCC can now work around the side effects of speculative execution
13006       on R10K processors. Please see the documentation of the
13007       -mr10k-cache-barrier option for details.
13008     * Support for the MIPS64 Release 2 instruction set has been added.
13009       The option -march=mips64r2 enables generation of these
13010       instructions.
13011     * GCC now supports Cavium Networks' Octeon processor. This support is
13012       available through the -march=octeon and -mtune=octeon options.
13013     * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The
13014       canonical -march= and -mtune= names for these processors are
13015       loongson2e and loongson2f.
13016
13017  picochip
13018
13019   Picochip is a 16-bit processor. A typical picoChip contains over 250
13020   small cores, each with small amounts of memory. There are three
13021   processor variants (STAN, MEM and CTRL) with different instruction sets
13022   and memory configurations and they can be chosen using the -mae option.
13023
13024   This port is intended to be a "C" only port.
13025
13026  Power Architecture and PowerPC
13027
13028     * GCC now supports the e300c2, e300c3 and e500mc processors.
13029     * GCC now supports Xilinx processors with a single-precision FPU.
13030     * Decimal floating point is now supported for e500 processors.
13031
13032  S/390, zSeries and System z9/z10
13033
13034     * Support for the IBM System z10 EC/BC processor has been added. When
13035       using the -march=z10 option, the compiler will generate code making
13036       use of instructions provided by the General-Instruction-Extension
13037       Facility and the Execute-Extension Facility.
13038
13039  VxWorks
13040
13041     * GCC now supports the thread-local storage mechanism used on
13042       VxWorks.
13043
13044  Xtensa
13045
13046     * GCC now supports thread-local storage (TLS) for Xtensa processor
13047       configurations that include the Thread Pointer option. TLS also
13048       requires support from the assembler and linker; this support is
13049       provided in the GNU binutils beginning with version 2.19.
13050
13051Documentation improvements
13052
13053Other significant improvements
13054
13055GCC 4.4.1
13056
13057   This is the [11]list of problem reports (PRs) from GCC's bug tracking
13058   system that are known to be fixed in the 4.4.1 release. This list might
13059   not be complete (that is, it is possible that some PRs that have been
13060   fixed are not listed here).
13061
13062GCC 4.4.2
13063
13064   This is the [12]list of problem reports (PRs) from GCC's bug tracking
13065   system that are known to be fixed in the 4.4.2 release. This list might
13066   not be complete (that is, it is possible that some PRs that have been
13067   fixed are not listed here).
13068
13069GCC 4.4.3
13070
13071   This is the [13]list of problem reports (PRs) from GCC's bug tracking
13072   system that are known to be fixed in the 4.4.3 release. This list might
13073   not be complete (that is, it is possible that some PRs that have been
13074   fixed are not listed here).
13075
13076GCC 4.4.4
13077
13078   This is the [14]list of problem reports (PRs) from GCC's bug tracking
13079   system that are known to be fixed in the 4.4.4 release. This list might
13080   not be complete (that is, it is possible that some PRs that have been
13081   fixed are not listed here).
13082
13083GCC 4.4.5
13084
13085   This is the [15]list of problem reports (PRs) from GCC's bug tracking
13086   system that are known to be fixed in the 4.4.5 release. This list might
13087   not be complete (that is, it is possible that some PRs that have been
13088   fixed are not listed here).
13089
13090GCC 4.4.6
13091
13092   This is the [16]list of problem reports (PRs) from GCC's bug tracking
13093   system that are known to be fixed in the 4.4.6 release. This list might
13094   not be complete (that is, it is possible that some PRs that have been
13095   fixed are not listed here).
13096
13097GCC 4.4.7
13098
13099   This is the [17]list of problem reports (PRs) from GCC's bug tracking
13100   system that are known to be fixed in the 4.4.7 release. This list might
13101   not be complete (that is, it is possible that some PRs that have been
13102   fixed are not listed here).
13103
13104
13105    For questions related to the use of GCC, please consult these web
13106    pages and the [18]GCC manuals. If that fails, the
13107    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13108    web pages and the development of GCC are welcome on our developer
13109    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
13110    archives.
13111
13112   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
13113   distribution of this entire article is permitted in any medium,
13114   provided this notice is preserved.
13115
13116   These pages are [23]maintained by the GCC team. Last modified
13117   2022-11-01.
13118
13119References
13120
13121   1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7
13122   2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted
13123   3. http://gcc.gnu.org/gcc-4.4/porting_to.html
13124   4. https://gcc.gnu.org/wiki/Graphite
13125   5. https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf
13126   6. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html
13127   7. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html
13128   8. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125
13129   9. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221
13130  10. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34
13131  11. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1
13132  12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2
13133  13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3
13134  14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4
13135  15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5
13136  16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6
13137  17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7
13138  18. https://gcc.gnu.org/onlinedocs/
13139  19. mailto:gcc-help@gcc.gnu.org
13140  20. mailto:gcc@gcc.gnu.org
13141  21. https://gcc.gnu.org/lists.html
13142  22. https://www.fsf.org/
13143  23. https://gcc.gnu.org/about.html
13144======================================================================
13145http://gcc.gnu.org/gcc-4.3/index.html
13146
13147                             GCC 4.3 Release Series
13148
13149   (This release series is no longer supported.)
13150
13151   Jun 27, 2011
13152
13153   The [1]GNU project and the GCC developers are pleased to announce the
13154   release of GCC 4.3.6.
13155
13156   This release is a bug-fix release, containing fixes for regressions in
13157   GCC 4.3.5 relative to previous releases of GCC.
13158
13159Release History
13160
13161   GCC 4.3.6
13162          Jun 27, 2011 ([2]changes)
13163
13164   GCC 4.3.5
13165          May 22, 2010 ([3]changes)
13166
13167   GCC 4.3.4
13168          August 4, 2009 ([4]changes)
13169
13170   GCC 4.3.3
13171          January 24, 2009 ([5]changes)
13172
13173   GCC 4.3.2
13174          August 27, 2008 ([6]changes)
13175
13176   GCC 4.3.1
13177          June 6, 2008 ([7]changes)
13178
13179   GCC 4.3.0
13180          March 5, 2008 ([8]changes)
13181
13182References and Acknowledgements
13183
13184   GCC used to stand for the GNU C Compiler, but since the compiler
13185   supports several other languages aside from C, it now stands for the
13186   GNU Compiler Collection.
13187
13188   A list of [9]successful builds is updated as new information becomes
13189   available.
13190
13191   The GCC developers would like to thank the numerous people that have
13192   contributed new features, improvements, bug fixes, and other changes as
13193   well as test results to GCC. This [10]amazing group of volunteers is
13194   what makes GCC successful.
13195
13196   For additional information about GCC please refer to the [11]GCC
13197   project web site or contact the [12]GCC development mailing list.
13198
13199   To obtain GCC please use [13]our mirror sites or [14]our version
13200   control system.
13201
13202
13203    For questions related to the use of GCC, please consult these web
13204    pages and the [15]GCC manuals. If that fails, the
13205    [16]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13206    web pages and the development of GCC are welcome on our developer
13207    list at [17]gcc@gcc.gnu.org. All of [18]our lists have public
13208    archives.
13209
13210   Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and
13211   distribution of this entire article is permitted in any medium,
13212   provided this notice is preserved.
13213
13214   These pages are [20]maintained by the GCC team. Last modified
13215   2022-10-26.
13216
13217References
13218
13219   1. http://www.gnu.org/
13220   2. http://gcc.gnu.org/gcc-4.3/changes.html
13221   3. http://gcc.gnu.org/gcc-4.3/changes.html
13222   4. http://gcc.gnu.org/gcc-4.3/changes.html
13223   5. http://gcc.gnu.org/gcc-4.3/changes.html
13224   6. http://gcc.gnu.org/gcc-4.3/changes.html
13225   7. http://gcc.gnu.org/gcc-4.3/changes.html
13226   8. http://gcc.gnu.org/gcc-4.3/changes.html
13227   9. http://gcc.gnu.org/gcc-4.3/buildstat.html
13228  10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
13229  11. http://gcc.gnu.org/index.html
13230  12. mailto:gcc@gcc.gnu.org
13231  13. http://gcc.gnu.org/mirrors.html
13232  14. http://gcc.gnu.org/git.html
13233  15. https://gcc.gnu.org/onlinedocs/
13234  16. mailto:gcc-help@gcc.gnu.org
13235  17. mailto:gcc@gcc.gnu.org
13236  18. https://gcc.gnu.org/lists.html
13237  19. https://www.fsf.org/
13238  20. https://gcc.gnu.org/about.html
13239======================================================================
13240http://gcc.gnu.org/gcc-4.3/changes.html
13241
13242                             GCC 4.3 Release Series
13243                        Changes, New Features, and Fixes
13244
13245   The latest release in the 4.3 release series is [1]GCC 4.3.5.
13246
13247Caveats
13248
13249     * GCC requires the [2]GMP and [3]MPFR libraries for building all the
13250       various front-end languages it supports. See the [4]prerequisites
13251       page for version requirements.
13252     * ColdFire targets now treat long double as having the same format as
13253       double. In earlier versions of GCC, they used the 68881 long double
13254       format instead.
13255     * The m68k-uclinux target now uses the same calling conventions as
13256       m68k-linux-gnu. You can select the original calling conventions by
13257       configuring for m68k-uclinuxoldabi instead. Note that
13258       m68k-uclinuxoldabi also retains the original 80-bit long double on
13259       ColdFire targets.
13260     * The -fforce-mem option has been removed because it has had no
13261       effect in the last few GCC releases.
13262     * The i386 -msvr3-shlib option has been removed since it is no longer
13263       used.
13264     * Fastcall for i386 has been changed not to pass aggregate arguments
13265       in registers, following Microsoft compilers.
13266     * Support for the AOF assembler has been removed from the ARM back
13267       end; this affects only the targets arm-semi-aof and armel-semi-aof,
13268       which are no longer recognized. We removed these targets without a
13269       deprecation period because we discovered that they have been
13270       unusable since GCC 4.0.0.
13271     * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*)
13272       has been removed. This support had been deprecated since GCC 4.0.0.
13273     * Support for a number of older systems and recently unmaintained or
13274       untested target ports of GCC has been declared obsolete in GCC 4.3.
13275       Unless there is activity to revive them, the next release of GCC
13276       will have their sources permanently removed.
13277       All GCC ports for the following processor architectures have been
13278       declared obsolete:
13279          + Morpho MT (mt-*)
13280       The following aliases for processor architectures have been
13281       declared obsolete. Users should use the indicated generic target
13282       names instead, with compile-time options such as -mcpu or
13283       configure-time options such as --with-cpu to control the
13284       configuration more precisely.
13285          + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-*
13286            instead).
13287          + parisc*-*-* (use hppa*-*-* instead).
13288          + m680[012]0-*-* (use m68k-*-* instead).
13289       All GCC ports for the following operating systems have been
13290       declared obsolete:
13291          + BeOS (*-*-beos*)
13292          + kaOS (*-*-kaos*)
13293          + GNU/Linux using the a.out object format (*-*-linux*aout*)
13294          + GNU/Linux using version 1 of the GNU C Library
13295            (*-*-linux*libc1*)
13296          + Solaris versions before Solaris 7 (*-*-solaris2.[0-6],
13297            *-*-solaris2.[0-6].*)
13298          + Miscellaneous System V (*-*-sysv*)
13299          + WindISS (*-*-windiss*)
13300       Also, those for some individual systems on particular architectures
13301       have been obsoleted:
13302          + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*)
13303          + CRIS with a.out object format (cris-*-aout)
13304          + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*)
13305          + OSF/1 on PA-RISC (hppa1.1-*-osf*)
13306          + PRO on PA-RISC (hppa1.1-*-pro*)
13307          + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*,
13308            i[34567]86-sequent-sysv4*)
13309          + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*)
13310          + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host
13311            was previously [5]removed in 2001, leaving only the support
13312            for UWIN as a target now being deprecated)
13313          + ChorusOS on PowerPC (powerpc-*-chorusos*)
13314          + All VAX configurations apart from NetBSD and OpenBSD
13315            (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*)
13316     * The [6]-Wconversion option has been modified. Its purpose now is to
13317       warn for implicit conversions that may alter a value. This new
13318       behavior is available for both C and C++. Warnings about
13319       conversions between signed and unsigned integers can be disabled by
13320       using -Wno-sign-conversion. In C++, they are disabled by default
13321       unless -Wsign-conversion is explicitly requested. The old behavior
13322       of -Wconversion, that is, warn for prototypes causing a type
13323       conversion that is different from what would happen to the same
13324       argument in the absence of a prototype, has been moved to a new
13325       option -Wtraditional-conversion, which is only available for C.
13326     * The -m386, -m486, -mpentium and -mpentiumpro tuning options have
13327       been removed because they were deprecated for more than 3 GCC major
13328       releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or
13329       -mtune=pentiumpro as a replacement.
13330     * The -funsafe-math-optimizations option now automatically turns on
13331       -fno-trapping-math in addition to -fno-signed-zeros, as it enables
13332       reassociation and thus may introduce or remove traps.
13333     * The -ftree-vectorize option is now on by default under -O3. In
13334       order to generate code for a SIMD extension, it has to be enabled
13335       as well: use -maltivec for PowerPC platforms and -msse/-msse2 for
13336       i?86 and x86_64.
13337     * More information on porting to GCC 4.3 from previous versions of
13338       GCC can be found in the [7]porting guide for this release.
13339
13340General Optimizer Improvements
13341
13342     * The GCC middle-end has been integrated with the MPFR library. This
13343       allows GCC to evaluate and replace at compile-time calls to
13344       built-in math functions having constant arguments with their
13345       mathematically equivalent results. In making use of MPFR, GCC can
13346       generate correct results regardless of the math library
13347       implementation or floating point precision of the host platform.
13348       This also allows GCC to generate identical results regardless of
13349       whether one compiles in native or cross-compile configurations to a
13350       particular target. The following built-in functions take advantage
13351       of this new capability: acos, acosh, asin, asinh, atan2, atan,
13352       atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1,
13353       fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10,
13354       log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh,
13355       tan, tanh, tgamma, y0, y1 and yn. The float and long double
13356       variants of these functions (e.g. sinf and sinl) are also handled.
13357       The sqrt and cabs functions with constant arguments were already
13358       optimized in prior GCC releases. Now they also use MPFR.
13359     * A new forward propagation pass on RTL was added. The new pass
13360       replaces several slower transformations, resulting in compile-time
13361       improvements as well as better code generation in some cases.
13362     * A new command-line switch -frecord-gcc-switches has been added to
13363       GCC, although it is only enabled for some targets. The switch
13364       causes the command line that was used to invoke the compiler to be
13365       recorded into the object file that is being created. The exact
13366       format of this recording is target and binary file format
13367       dependent, but it usually takes the form of a note section
13368       containing ASCII text. The switch is related to the -fverbose-asm
13369       switch, but that one only records the information in the assembler
13370       output file as comments, so the information never reaches the
13371       object file.
13372     * The inliner heuristic is now aware of stack frame consumption. New
13373       command-line parameters --param large-stack-frame and --param
13374       large-stack-frame-growth can be used to limit stack frame size
13375       growth caused by inlining.
13376     * During feedback directed optimizations, the expected block size the
13377       memcpy, memset and bzero functions operate on is discovered and for
13378       cases of commonly used small sizes, specialized inline code is
13379       generated.
13380     * __builtin_expect no longer requires its argument to be a compile
13381       time constant.
13382     * Interprocedural optimization was reorganized to work on functions
13383       in SSA form. This enables more precise and cheaper dataflow
13384       analysis and makes writing interprocedural optimizations easier.
13385       The following improvements have been implemented on top of this
13386       framework:
13387          + Pre-inline optimization: Selected local optimization passes
13388            are run before the inliner (and other interprocedural passes)
13389            are executed. This significantly improves the accuracy of code
13390            growth estimates used by the inliner and reduces the overall
13391            memory footprint for large compilation units.
13392          + Early inlining (a simple bottom-up inliner pass inlining only
13393            functions whose body is smaller than the expected call
13394            overhead) is now executed with the early optimization passes,
13395            thus inlining already optimized function bodies into an
13396            unoptimized function that is subsequently optimized by early
13397            optimizers. This enables the compiler to quickly eliminate
13398            abstraction penalty in C++ programs.
13399          + Interprocedural constant propagation now operate on SSA form
13400            increasing accuracy of the analysis.
13401     * A new internal representation for GIMPLE statements has been
13402       contributed, resulting in compile-time memory savings.
13403     * The vectorizer was enhanced to support vectorization of outer
13404       loops, intra-iteration parallelism (loop-aware SLP), vectorization
13405       of strided accesses and loops with multiple data-types. Run-time
13406       dependency testing using loop versioning was added. The cost model,
13407       turned on by -fvect-cost-model, was developed.
13408
13409New Languages and Language specific improvements
13410
13411     * We have added new command-line options
13412       -finstrument-functions-exclude-function-list and
13413       -finstrument-functions-exclude-file-list. They provide more control
13414       over which functions are annotated by the -finstrument-functions
13415       option.
13416
13417  C family
13418
13419     * Implicit conversions between generic vector types are now only
13420       permitted when the two vectors in question have the same number of
13421       elements and compatible element types. (Note that the restriction
13422       involves compatible element types, not implicitly-convertible
13423       element types: thus, a vector type with element type int may not be
13424       implicitly converted to a vector type with element type unsigned
13425       int.) This restriction, which is in line with specifications for
13426       SIMD architectures such as AltiVec, may be relaxed using the flag
13427       -flax-vector-conversions. This flag is intended only as a
13428       compatibility measure and should not be used for new code.
13429     * -Warray-bounds has been added and is now enabled by default for
13430       -Wall . It produces warnings for array subscripts that can be
13431       determined at compile time to be always out of bounds.
13432       -Wno-array-bounds will disable the warning.
13433     * The constructor and destructor function attributes now accept
13434       optional priority arguments which control the order in which the
13435       constructor and destructor functions are run.
13436     * New [8]command-line options -Wtype-limits, -Wold-style-declaration,
13437       -Wmissing-parameter-type, -Wempty-body, -Wclobbered and
13438       -Wignored-qualifiers have been added for finer control of the
13439       diverse warnings enabled by -Wextra.
13440     * A new function attribute alloc_size has been added to mark up
13441       malloc style functions. For constant sized allocations this can be
13442       used to find out the size of the returned pointer using the
13443       __builtin_object_size() function for buffer overflow checking and
13444       similar. This supplements the already built-in malloc and calloc
13445       constant size handling.
13446     * Integer constants written in binary are now supported as a GCC
13447       extension. They consist of a prefix 0b or 0B, followed by a
13448       sequence of 0 and 1 digits.
13449     * A new predefined macro __COUNTER__ has been added. It expands to
13450       sequential integral values starting from 0. In conjunction with the
13451       ## operator, this provides a convenient means to generate unique
13452       identifiers.
13453     * A new command-line option -fdirectives-only has been added. It
13454       enables a special preprocessing mode which improves the performance
13455       of applications like distcc and ccache.
13456     * Fixed-point data types and operators have been added. They are
13457       based on Chapter 4 of the Embedded-C specification (n1169.pdf).
13458       Currently, only MIPS targets are supported.
13459     * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732,
13460       N1241, is now supported as a GCC extension to C for targets
13461       i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu,
13462       and x86_64-*-linux-gnu. The feature introduces new data types
13463       _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF,
13464       DD, and DL.
13465
13466  C++
13467
13468     * [9]Experimental support for the upcoming ISO C++ standard, C++0x.
13469     * -Wc++0x-compat has been added and is now enabled by default for
13470       -Wall. It produces warnings for constructs whose meaning differs
13471       between ISO C++ 1998 and C++0x.
13472     * The -Wparentheses option now works for C++ as it does for C. It
13473       warns if parentheses are omitted when operators with confusing
13474       precedence are nested. It also warns about ambiguous else
13475       statements. Since -Wparentheses is enabled by -Wall, this may cause
13476       additional warnings with existing C++ code which uses -Wall. These
13477       new warnings may be disabled by using -Wall -Wno-parentheses.
13478     * The -Wmissing-declarations now works for C++ as it does for C.
13479     * The -fvisibility-ms-compat flag was added, to make it easier to
13480       port larger projects using shared libraries from Microsoft's Visual
13481       Studio to ELF and Mach-O systems.
13482     * C++ attribute handling has been overhauled for template arguments
13483       (ie dependent types). In particular, __attribute__((aligned(T)));
13484       works for C++ types.
13485
13486    Runtime Library (libstdc++)
13487
13488     * [10]Experimental support for the upcoming ISO C++ standard, C++0x.
13489     * Support for TR1 mathematical special functions and regular
13490       expressions.
13491     * Default what implementations give more elaborate exception strings
13492       for bad_cast, bad_typeid, bad_exception, and bad_alloc.
13493     * Header dependencies have been streamlined, reducing unnecessary
13494       includes and pre-processed bloat.
13495     * Variadic template implementations of items in <tuple> and
13496       <functional>.
13497     * An experimental [11]parallel mode has been added. This is a
13498       parallel implementation of many C++ Standard library algorithms,
13499       like std::accumulate, std::for_each, std::transform, or std::sort,
13500       to give but four examples. These algorithms can be substituted for
13501       the normal (sequential) libstdc++ algorithms on a piecemeal basis,
13502       or all existing algorithms can be transformed via the
13503       -D_GLIBCXX_PARALLEL macro.
13504     * Debug mode versions of classes in <unordered_set> and
13505       <unordered_map>.
13506     * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are
13507       now <backward/hash_set> and <backward/hash_map>. This code:
13508    #include <ext/hash_set>
13509    __gnu_cxx::hash_set<int> s;
13510
13511       Can be transformed (in order of preference) to:
13512    #include <tr1/unordered_set>
13513    std::tr1::unordered_set<int> s;
13514
13515       or
13516    #include <backward/hash_set>
13517    __gnu_cxx::hash_set<int> s;
13518
13519       Similar transformations apply to __gnu_cxx::hash_map,
13520       __gnu_cxx::hash_multimap, __gnu_cxx::hash_set,
13521       __gnu_cxx::hash_multiset.
13522
13523  Fortran
13524
13525     * Due to the fact that the GMP and MPFR libraries are required for
13526       all languages, Fortran is no longer special in this regard and is
13527       available by default.
13528     * The [12]-fexternal-blas option has been added, which generates
13529       calls to BLAS routines for intrinsic matrix operations such as
13530       matmul rather than using the built-in algorithms.
13531     * Support to give a backtrace (compiler flag -fbacktrace or
13532       environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems
13533       only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a
13534       run-time error occured.
13535     * GNU Fortran now defines __GFORTRAN__ when it runs the C
13536       preprocessor (CPP).
13537     * The [13]-finit-local-zero, -finit-real, -finit-integer,
13538       -finit-character, and -finit-logical options have been added, which
13539       can be used to initialize local variables.
13540     * The intrinsic procedures [14]GAMMA and [15]LGAMMA have been added,
13541       which calculate the Gamma function and its logarithm. Use EXTERNAL
13542       gamma if you want to use your own gamma function.
13543     * GNU Fortran now regards the backslash character as literal (as
13544       required by the Fortran 2003 standard); using [16]-fbackslash GNU
13545       Fortran interprets backslashes as C-style escape characters.
13546     * The [17]interpretation of binary, octal and hexadecimal (BOZ)
13547       literal constants has been changed. Before they were always
13548       interpreted as integer; now they are bit-wise transferred as
13549       argument of INT, REAL, DBLE and CMPLX as required by the Fortran
13550       2003 standard, and for real and complex variables in DATA
13551       statements or when directly assigned to real and complex variables.
13552       Everywhere else and especially in expressions they are still
13553       regarded as integer constants.
13554     * Fortran 2003 support has been extended:
13555          + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE
13556          + Pointer intent
13557          + Intrinsic module ISO_ENV_FORTRAN
13558          + Interoperability with C (ISO C Bindings)
13559          + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER
13560            attribute)
13561          + Fortran 2003 BOZ
13562
13563  Java (GCJ)
13564
13565     * GCJ now uses the Eclipse Java compiler for its Java parsing needs.
13566       This enables the use of all 1.5 language features, and fixes most
13567       existing front end bugs.
13568     * libgcj now supports all 1.5 language features which require runtime
13569       support: foreach, enum, annotations, generics, and auto-boxing.
13570     * We've made many changes to the tools shipped with gcj.
13571          + The old jv-scan tool has been removed. This tool never really
13572            worked properly. There is no replacement.
13573          + gcjh has been rewritten. Some of its more obscure options no
13574            longer work, but are still recognized in an attempt at
13575            compatibility. gjavah is a new program with similar
13576            functionality but different command-line options.
13577          + grmic and grmiregistry have been rewritten. grmid has been
13578            added.
13579          + gjar replaces the old fastjar.
13580          + gjarsigner (used for signing jars), gkeytool (used for key
13581            management), gorbd (for CORBA), gserialver (computes
13582            serialization UIDs), and gtnameserv (also for CORBA) are now
13583            installed.
13584     * The ability to dump the contents of the java run time heap to a
13585       file for off-line analysis has been added. The heap dumps may be
13586       analyzed with the new gc-analyze tool. They may be generated on
13587       out-of-memory conditions or on demand and are controlled by the new
13588       run time class gnu.gcj.util.GCInfo.
13589     * java.util.TimeZone can now read files from /usr/share/zoneinfo to
13590       provide correct, updated, timezone information. This means that
13591       packagers no longer have to update libgcj when a time zone change
13592       is published.
13593
13594New Targets and Target Specific Improvements
13595
13596  IA-32/x86-64
13597
13598     * Tuning for Intel Core 2 processors is available via -mtune=core2
13599       and -march=core2.
13600     * Tuning for AMD Geode processors is available via -mtune=geode and
13601       -march=geode.
13602     * Code generation of block move (memcpy) and block set (memset) was
13603       rewritten. GCC can now pick the best algorithm (loop, unrolled
13604       loop, instruction with rep prefix or a library call) based on the
13605       size of the block being copied and the CPU being optimized for. A
13606       new option -minline-stringops-dynamically has been added. With this
13607       option string operations of unknown size are expanded such that
13608       small blocks are copied by in-line code, while for large blocks a
13609       library call is used. This results in faster code than
13610       -minline-all-stringops when the library implementation is capable
13611       of using cache hierarchy hints. The heuristic choosing the
13612       particular algorithm can be overwritten via -mstringop-strategy.
13613       Newly also memset of values different from 0 is inlined.
13614     * GCC no longer places the cld instruction before string operations.
13615       Both i386 and x86-64 ABI documents mandate the direction flag to be
13616       clear at the entry of a function. It is now invalid to set the flag
13617       in asm statement without reseting it afterward.
13618     * Support for SSSE3 built-in functions and code generation are
13619       available via -mssse3.
13620     * Support for SSE4.1 built-in functions and code generation are
13621       available via -msse4.1.
13622     * Support for SSE4.2 built-in functions and code generation are
13623       available via -msse4.2.
13624     * Both SSE4.1 and SSE4.2 support can be enabled via -msse4.
13625     * A new set of options -mpc32, -mpc64 and -mpc80 have been added to
13626       allow explicit control of x87 floating point precision.
13627     * Support for __float128 (TFmode) IEEE quad type and corresponding
13628       TCmode IEEE complex quad type is available via the soft-fp library
13629       on x86_64 targets. This includes basic arithmetic operations
13630       (addition, subtraction, negation, multiplication and division) on
13631       __float128 real and TCmode complex values, the full set of IEEE
13632       comparisons between __float128 values, conversions to and from
13633       float, double and long double floating point types, as well as
13634       conversions to and from signed or unsigned integer, signed or
13635       unsigned long integer and signed or unsigned quad (TImode) integer
13636       types. Additionally, all operations generate the full set of IEEE
13637       exceptions and support the full set of IEEE rounding modes.
13638     * GCC can now utilize the ACML library for vectorizing calls to a set
13639       of C99 functions on x86_64 if -mveclibabi=acml is specified and you
13640       link to an ACML ABI compatible library.
13641
13642  ARM
13643
13644     * Compiler and Library support for Thumb-2 and the ARMv7 architecture
13645       has been added.
13646
13647  CRIS
13648
13649    New features
13650
13651     * Compiler and Library support for the CRIS v32 architecture, as
13652       found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been
13653       added.
13654
13655    Configuration changes
13656
13657     * The cris-*-elf target now includes support for CRIS v32, including
13658       libraries, through the -march=v32 option.
13659     * A new crisv32-*-elf target defaults to generate code for CRIS v32.
13660     * A new crisv32-*-linux* target defaults to generate code for CRIS
13661       v32.
13662     * The cris-*-aout target has been obsoleted.
13663
13664    Improved support for built-in functions
13665
13666     * GCC can now use the lz and swapwbr instructions to implement the
13667       __builtin_clz, __builtin_ctz and __builtin_ffs family of functions.
13668     * __builtin_bswap32 is now implemented using the swapwb instruction,
13669       when available.
13670
13671  m68k and ColdFire
13672
13673    New features
13674
13675     * Support for several new ColdFire processors has been added. You can
13676       generate code for them using the new -mcpu option.
13677     * All targets now support ColdFire processors.
13678     * m68k-uclinux targets have improved support for C++ constructors and
13679       destructors, and for shared libraries.
13680     * It is now possible to set breakpoints on the first or last line of
13681       a function, even if there are no statements on that line.
13682
13683    Optimizations
13684
13685     * Support for sibling calls has been added.
13686     * More use is now made of the ColdFire mov3q instruction.
13687     * __builtin_clz is now implemented using the ff1 ColdFire
13688       instruction, when available.
13689     * GCC now honors the -m68010 option. 68010 code now uses clr rather
13690       than move to zero volatile memory.
13691     * 68020 targets and above can now use symbol(index.size*scale)
13692       addresses for indexed array accesses. Earlier compilers would
13693       always load the symbol into a base register first.
13694
13695    Configuration changes
13696
13697     * All m68k and ColdFire targets now allow the default processor to be
13698       set at configure time using --with-cpu.
13699     * A --with-arch configuration option has been added. This option
13700       allows you to restrict a target to ColdFire or non-ColdFire
13701       processors.
13702
13703    Preprocessor macros
13704
13705     * An __mcfv*__ macro is now defined for all ColdFire targets.
13706       (Earlier versions of GCC only defined __mcfv4e__.)
13707     * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added.
13708     * All targets now define __mc68010 and __mc68010__ when generating
13709       68010 code.
13710
13711    Command-line changes
13712
13713     * New command-line options -march, -mcpu, -mtune and -mhard-float
13714       have been added. These options apply to both m68k and ColdFire
13715       targets.
13716     * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative
13717       versions of -mshort, etc.
13718     * -fforce-addr has been removed. It is now ignored by the compiler.
13719
13720    Other improvements
13721
13722     * ColdFire targets now try to maintain a 4-byte-aligned stack where
13723       possible.
13724     * m68k-uclinux targets now try to avoid situations that lead to the
13725       load-time error: BINFMT_FLAT: reloc outside program.
13726
13727  MIPS
13728
13729    Changes to existing configurations
13730
13731     * libffi and libjava now support all three GNU/Linux ABIs: o32, n32
13732       and n64. Every GNU/Linux configuration now builds these libraries
13733       by default.
13734     * GNU/Linux configurations now generate -mno-shared code unless
13735       overridden by -fpic, -fPIC, -fpie or -fPIE.
13736     * mipsisa32*-linux-gnu configurations now generate hard-float code by
13737       default, just like other mipsisa32* and mips*-linux-gnu
13738       configurations. You can build a soft-float version of any
13739       mips*-linux-gnu configuration by passing --with-float=soft to
13740       configure.
13741     * mips-wrs-vxworks now supports run-time processes (RTPs).
13742
13743    Changes to existing command-line options
13744
13745     * The -march and -mtune options no longer accept 24k as a processor
13746       name. Please use 24kc, 24kf2_1 or 24kf1_1 instead.
13747     * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and
13748       34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The
13749       options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for
13750       24kx, 24kex and 34kx.
13751
13752    New configurations
13753
13754   GCC now supports the following configurations:
13755     * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by
13756       default. Earlier releases also recognized this configuration, but
13757       they treated it in the same way as mipsisa32*-linux-gnu*. Note that
13758       you can customize any mips*-linux-gnu* configuration to a
13759       particular ISA or processor by passing an appropriate --with-arch
13760       option to configure.
13761     * mipsisa*-sde-elf*, which provides compatibility with MIPS
13762       Technologies' SDE toolchains. The configuration uses the SDE
13763       libraries by default, but you can use it like other newlib-based
13764       ELF configurations by passing --with-newlib to configure. It is the
13765       only configuration besides mips64vr*-elf* to build MIPS16 as well
13766       as non-MIPS16 libraries.
13767     * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf*
13768       configuration, but uses the o32 and o64 ABIs instead of the 32-bit
13769       and 64-bit forms of the EABI.
13770
13771    New processors and application-specific extensions
13772
13773     * Support for the SmartMIPS ASE is available through the new
13774       -msmartmips option.
13775     * Support for revision 2 of the DSP ASE is available through the new
13776       -mdspr2 option. A new preprocessor macro called __mips_dsp_rev
13777       indicates the revision of the ASE in use.
13778     * Support for the 4KS and 74K families of processors is available
13779       through the -march and -mtune options.
13780
13781    Improved support for built-in functions
13782
13783     * GCC can now use load-linked, store-conditional and sync
13784       instructions to implement atomic built-in functions such as
13785       __sync_fetch_and_add. The memory reference must be 4 bytes wide for
13786       32-bit targets and either 4 or 8 bytes wide for 64-bit targets.
13787     * GCC can now use the clz and dclz instructions to implement the
13788       __builtin_ctz and __builtin_ffs families of functions.
13789     * There is a new __builtin___clear_cache function for flushing the
13790       instruction cache. GCC expands this function inline on MIPS32
13791       revision 2 targets, otherwise it calls the function specified by
13792       -mcache-flush-func.
13793
13794    MIPS16 improvements
13795
13796     * GCC can now compile objects that contain a mixture of MIPS16 and
13797       non-MIPS16 code. There are two new attributes, mips16 and nomips16,
13798       for specifying which mode a function should use.
13799     * A new option called -minterlink-mips16 makes non-MIPS16 code
13800       link-compatible with MIPS16 code.
13801     * After many bug fixes, the long-standing MIPS16 -mhard-float support
13802       should now work fairly reliably.
13803     * GCC can now use the MIPS16e save and restore instructions.
13804     * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled
13805       with -G0 -fsection-anchors is often smaller than code compiled with
13806       -G8. However, please note that you must usually compile all objects
13807       in your application with the same -G option; see the documentation
13808       of -G for details.
13809     * A new option called-mcode-readable specifies which instructions are
13810       allowed to load from the code segment. -mcode-readable=yes is the
13811       default and says that any instruction may load from the code
13812       segment. The other alternatives are -mcode-readable=pcrel, which
13813       says that only PC-relative MIPS16 instructions may load from the
13814       code segment, and -mcode-readable=no, which says that no
13815       instruction may do so. Please see the documentation for more
13816       details, including example uses.
13817
13818    Small-data improvements
13819
13820   There are three new options for controlling small data:
13821     * -mno-extern-sdata, which disables small-data accesses for
13822       externally-defined variables. Code compiled with -Gn
13823       -mno-extern-sdata will be link-compatible with any -G setting
13824       between -G0 and -Gn inclusive.
13825     * -mno-local-sdata, which disables the use of small-data sections for
13826       data that is not externally visible. This option can be a useful
13827       way of reducing small-data usage in less performance-critical parts
13828       of an application.
13829     * -mno-gpopt, which disables the use of the $gp register while still
13830       honoring the -G limit when placing externally-visible data. This
13831       option implies -mno-extern-sdata and -mno-local-sdata and it can be
13832       useful in situations where $gp does not necessarily hold the
13833       expected value.
13834
13835    Miscellaneous improvements
13836
13837     * There is a new option called -mbranch-cost for tweaking the
13838       perceived cost of branches.
13839     * If GCC is configured to use a version of GAS that supports the
13840       .gnu_attribute directive, it will use that directive to record
13841       certain properties of the output code. .gnu_attribute is new to GAS
13842       2.18.
13843     * There are two new function attributes, near and far, for overriding
13844       the command-line setting of -mlong-calls on a function-by-function
13845       basis.
13846     * -mfp64, which previously required a 64-bit target, now works with
13847       MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and
13848       mipsisa*-sde-elf* configurations provide suitable library support.
13849     * GCC now recognizes the -mdmx and -mmt options and passes them down
13850       to the assembler. It does nothing else with the options at present.
13851
13852  SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture
13853  (BEA)
13854
13855     * Support has been added for this new architecture.
13856
13857  RS6000 (POWER/PowerPC)
13858
13859     * Support for the PowerPC 750CL paired-single instructions has been
13860       added with a new powerpc-*-linux*paired* target configuration. It
13861       is enabled by an associated -mpaired option and can be accessed
13862       using new built-in functions.
13863     * Support for auto-detecting architecture and system configuration to
13864       auto-select processor optimization tuning.
13865     * Support for VMX on AIX 5.3 has been added.
13866     * Support for AIX Version 6.1 has been added.
13867
13868  S/390, zSeries and System z9
13869
13870     * Support for the IBM System z9 EC/BC processor (z9 GA3) has been
13871       added. When using the -march=z9-ec option, the compiler will
13872       generate code making use of instructions provided by the decimal
13873       floating point facility and the floating point conversion facility
13874       (pfpo). Besides the instructions used to implement decimal floating
13875       point operations these facilities also contain instructions to move
13876       between general purpose and floating point registers and to modify
13877       and copy the sign-bit of floating point values.
13878     * When the -march=z9-ec option is used the new
13879       -mhard-dfp/-mno-hard-dfp options can be used to specify whether the
13880       decimal floating point hardware instructions will be used or not.
13881       If none of them is given the hardware support is enabled by
13882       default.
13883     * The -mstack-guard option can now be omitted when using stack
13884       checking via -mstack-size in order to let GCC choose a sensible
13885       stack guard value according to the frame size of each function.
13886     * Various changes to improve performance of generated code have been
13887       implemented, including:
13888          + The condition code set by an add logical with carry
13889            instruction is now available for overflow checks like: a + b +
13890            carry < b.
13891          + The test data class instruction is now used to implement
13892            sign-bit and infinity checks of binary and decimal floating
13893            point numbers.
13894
13895  SPARC
13896
13897     * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been
13898       added.
13899
13900  Xtensa
13901
13902     * Stack unwinding for exception handling now uses by default a
13903       specialized version of DWARF unwinding. This is not
13904       binary-compatible with the setjmp/longjmp (sjlj) unwinding used for
13905       Xtensa with previous versions of GCC.
13906     * For Xtensa processors that include the Conditional Store option,
13907       the built-in functions for atomic memory access are now implemented
13908       using S32C1I instructions.
13909     * If the Xtensa NSA option is available, GCC will use it to implement
13910       the __builtin_ctz and __builtin_clz functions.
13911
13912Documentation improvements
13913
13914     * Existing libstdc++ documentation has been edited and restructured
13915       into a single DocBook XML manual. The results can be viewed online
13916       [18]here.
13917
13918Other significant improvements
13919
13920     * The compiler's --help command-line option has been extended so that
13921       it now takes an optional set of arguments. These arguments restrict
13922       the information displayed to specific classes of command-line
13923       options, and possibly only a subset of those options. It is also
13924       now possible to replace the descriptive text associated with each
13925       displayed option with an indication of its current value, or for
13926       binary options, whether it has been enabled or disabled.
13927       Here are some examples. The following will display all the options
13928       controlling warning messages:
13929      --help=warnings
13930
13931       Whereas this will display all the undocumented, target specific
13932       options:
13933      --help=target,undocumented
13934
13935       This sequence of commands will display the binary optimizations
13936       that are enabled by -O3:
13937      gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
13938      gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
13939      diff /tmp/O2-opts /tmp/O3-opts | grep enabled
13940
13941     * The configure options --with-pkgversion and --with-bugurl have been
13942       added. These allow distributors of GCC to include a
13943       distributor-specific string in manuals and --version output and to
13944       specify the URL for reporting bugs in their versions of GCC.
13945
13946GCC 4.3.1
13947
13948   This is the [19]list of problem reports (PRs) from GCC's bug tracking
13949   system that are known to be fixed in the 4.3.1 release. This list might
13950   not be complete (that is, it is possible that some PRs that have been
13951   fixed are not listed here).
13952
13953Target Specific Changes
13954
13955  IA-32/x86-64
13956
13957    ABI changes
13958
13959     * Starting with GCC 4.3.1, decimal floating point variables are
13960       aligned to their natural boundaries when they are passed on the
13961       stack for i386.
13962
13963    Command-line changes
13964
13965     * Starting with GCC 4.3.1, the -mcld option has been added to
13966       automatically generate a cld instruction in the prologue of
13967       functions that use string instructions. This option is used for
13968       backward compatibility on some operating systems and can be enabled
13969       by default for 32-bit x86 targets by configuring GCC with the
13970       --enable-cld configure option.
13971
13972GCC 4.3.2
13973
13974   This is the [20]list of problem reports (PRs) from GCC's bug tracking
13975   system that are known to be fixed in the 4.3.2 release. This list might
13976   not be complete (that is, it is possible that some PRs that have been
13977   fixed are not listed here).
13978
13979GCC 4.3.3
13980
13981   This is the [21]list of problem reports (PRs) from GCC's bug tracking
13982   system that are known to be fixed in the 4.3.3 release. This list might
13983   not be complete (that is, it is possible that some PRs that have been
13984   fixed are not listed here).
13985
13986GCC 4.3.4
13987
13988   This is the [22]list of problem reports (PRs) from GCC's bug tracking
13989   system that are known to be fixed in the 4.3.4 release. This list might
13990   not be complete (that is, it is possible that some PRs that have been
13991   fixed are not listed here).
13992
13993GCC 4.3.5
13994
13995   This is the [23]list of problem reports (PRs) from GCC's bug tracking
13996   system that are known to be fixed in the 4.3.5 release. This list might
13997   not be complete (that is, it is possible that some PRs that have been
13998   fixed are not listed here).
13999
14000GCC 4.3.6
14001
14002   This is the [24]list of problem reports (PRs) from GCC's bug tracking
14003   system that are known to be fixed in the 4.3.6 release. This list might
14004   not be complete (that is, it is possible that some PRs that have been
14005   fixed are not listed here).
14006
14007
14008    For questions related to the use of GCC, please consult these web
14009    pages and the [25]GCC manuals. If that fails, the
14010    [26]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14011    web pages and the development of GCC are welcome on our developer
14012    list at [27]gcc@gcc.gnu.org. All of [28]our lists have public
14013    archives.
14014
14015   Copyright (C) [29]Free Software Foundation, Inc. Verbatim copying and
14016   distribution of this entire article is permitted in any medium,
14017   provided this notice is preserved.
14018
14019   These pages are [30]maintained by the GCC team. Last modified
14020   2022-10-26.
14021
14022References
14023
14024   1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5
14025   2. https://gmplib.org/
14026   3. https://www.mpfr.org/
14027   4. https://gcc.gnu.org/install/prerequisites.html
14028   5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html
14029   6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
14030   7. http://gcc.gnu.org/gcc-4.3/porting_to.html
14031   8. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
14032   9. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
14033  10. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
14034  11. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html
14035  12. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options
14036  13. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167
14037  14. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html
14038  15. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html
14039  16. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html
14040  17. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html
14041  18. https://gcc.gnu.org/onlinedocs/libstdc++/
14042  19. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1
14043  20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2
14044  21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3
14045  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4
14046  23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5
14047  24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6
14048  25. https://gcc.gnu.org/onlinedocs/
14049  26. mailto:gcc-help@gcc.gnu.org
14050  27. mailto:gcc@gcc.gnu.org
14051  28. https://gcc.gnu.org/lists.html
14052  29. https://www.fsf.org/
14053  30. https://gcc.gnu.org/about.html
14054======================================================================
14055http://gcc.gnu.org/gcc-4.2/index.html
14056
14057                             GCC 4.2 Release Series
14058
14059   (This release series is no longer supported.)
14060
14061   May 19, 2008
14062
14063   The [1]GNU project and the GCC developers are pleased to announce the
14064   release of GCC 4.2.4.
14065
14066   This release is a bug-fix release, containing fixes for regressions in
14067   GCC 4.2.3 relative to previous releases of GCC.
14068
14069Release History
14070
14071   GCC 4.2.4
14072          May 19, 2008 ([2]changes)
14073
14074   GCC 4.2.3
14075          February 1, 2008 ([3]changes)
14076
14077   GCC 4.2.2
14078          October 7, 2007 ([4]changes)
14079
14080   GCC 4.2.1
14081          July 18, 2007 ([5]changes)
14082
14083   GCC 4.2.0
14084          May 13, 2007 ([6]changes)
14085
14086References and Acknowledgements
14087
14088   GCC used to stand for the GNU C Compiler, but since the compiler
14089   supports several other languages aside from C, it now stands for the
14090   GNU Compiler Collection.
14091
14092   A list of [7]successful builds is updated as new information becomes
14093   available.
14094
14095   The GCC developers would like to thank the numerous people that have
14096   contributed new features, improvements, bug fixes, and other changes as
14097   well as test results to GCC. This [8]amazing group of volunteers is
14098   what makes GCC successful.
14099
14100   For additional information about GCC please refer to the [9]GCC project
14101   web site or contact the [10]GCC development mailing list.
14102
14103   To obtain GCC please use [11]our mirror sites or [12]our version
14104   control system.
14105
14106
14107    For questions related to the use of GCC, please consult these web
14108    pages and the [13]GCC manuals. If that fails, the
14109    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14110    web pages and the development of GCC are welcome on our developer
14111    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
14112    archives.
14113
14114   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
14115   distribution of this entire article is permitted in any medium,
14116   provided this notice is preserved.
14117
14118   These pages are [18]maintained by the GCC team. Last modified
14119   2022-10-26.
14120
14121References
14122
14123   1. http://www.gnu.org/
14124   2. http://gcc.gnu.org/gcc-4.2/changes.html
14125   3. http://gcc.gnu.org/gcc-4.2/changes.html
14126   4. http://gcc.gnu.org/gcc-4.2/changes.html
14127   5. http://gcc.gnu.org/gcc-4.2/changes.html
14128   6. http://gcc.gnu.org/gcc-4.2/changes.html
14129   7. http://gcc.gnu.org/gcc-4.2/buildstat.html
14130   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
14131   9. http://gcc.gnu.org/index.html
14132  10. mailto:gcc@gcc.gnu.org
14133  11. http://gcc.gnu.org/mirrors.html
14134  12. http://gcc.gnu.org/git.html
14135  13. https://gcc.gnu.org/onlinedocs/
14136  14. mailto:gcc-help@gcc.gnu.org
14137  15. mailto:gcc@gcc.gnu.org
14138  16. https://gcc.gnu.org/lists.html
14139  17. https://www.fsf.org/
14140  18. https://gcc.gnu.org/about.html
14141======================================================================
14142http://gcc.gnu.org/gcc-4.2/changes.html
14143
14144                             GCC 4.2 Release Series
14145                        Changes, New Features, and Fixes
14146
14147Caveats
14148
14149     * GCC no longer accepts the -fshared-data option. This option has had
14150       no effect in any GCC 4 release; the targets to which the option
14151       used to apply had been removed before GCC 4.0.
14152
14153General Optimizer Improvements
14154
14155     * New command-line options specify the possible relationships among
14156       parameters and between parameters and global data. For example,
14157       -fargument-noalias-anything specifies that arguments do not alias
14158       any other storage.
14159       Each language will automatically use whatever option is required by
14160       the language standard. You should not need to use these options
14161       yourself.
14162
14163New Languages and Language specific improvements
14164
14165     * [1]OpenMP is now supported for the C, C++ and Fortran compilers.
14166     * New command-line options -fstrict-overflow and -Wstrict-overflow
14167       have been added. -fstrict-overflow tells the compiler that it may
14168       assume that the program follows the strict signed overflow
14169       semantics permitted for the language: for C and C++ this means that
14170       the compiler may assume that signed overflow does not occur. For
14171       example, a loop like
14172      for (i = 1; i > 0; i *= 2)
14173
14174       is presumably intended to continue looping until i overflows. With
14175       -fstrict-overflow, the compiler may assume that signed overflow
14176       will not occur, and transform this into an infinite loop.
14177       -fstrict-overflow is turned on by default at -O2, and may be
14178       disabled via -fno-strict-overflow. The -Wstrict-overflow option may
14179       be used to warn about cases where the compiler assumes that signed
14180       overflow will not occur. It takes five different levels:
14181       -Wstrict-overflow=1 to 5. See the [2]documentation for details.
14182       -Wstrict-overflow=1 is enabled by -Wall.
14183     * The new command-line option -fno-toplevel-reorder directs GCC to
14184       emit top-level functions, variables, and asm statements in the same
14185       order that they appear in the input file. This is intended to
14186       support existing code which relies on a particular ordering (for
14187       example, code which uses top-level asm statements to switch
14188       sections). For new code, it is generally better to use function and
14189       variable attributes. The -fno-toplevel-reorder option may be used
14190       for most cases which currently use -fno-unit-at-a-time. The
14191       -fno-unit-at-a-time option will be removed in some future version
14192       of GCC. If you know of a case which requires -fno-unit-at-a-time
14193       which is not fixed by -fno-toplevel-reorder, please open a bug
14194       report.
14195
14196  C family
14197
14198     * The pragma redefine_extname will now macro expand its tokens for
14199       compatibility with SunPRO.
14200     * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct
14201       GCC to handle inline functions as specified in the C99 standard. In
14202       preparation for this, GCC 4.2 will warn about any use of non-static
14203       inline functions in gnu99 or c99 mode. This new warning may be
14204       disabled with the new gnu_inline function attribute or the new
14205       -fgnu89-inline command-line option. Also, GCC 4.2 and later will
14206       define one of the preprocessor macros __GNUC_GNU_INLINE__ or
14207       __GNUC_STDC_INLINE__ to indicate the semantics of inline functions
14208       in the current compilation.
14209     * A new command-line option -Waddress has been added to warn about
14210       suspicious uses of memory addresses as, for example, using the
14211       address of a function in a conditional expression, and comparisons
14212       against the memory address of a string literal. This warning is
14213       enabled by -Wall.
14214
14215  C++
14216
14217     * C++ visibility handling has been overhauled.
14218       Restricted visiblity is propagated from classes to members, from
14219       functions to local statics, and from templates and template
14220       arguments to instantiations, unless the latter has explicitly
14221       declared visibility.
14222       The visibility attribute for a class must come between the
14223       class-key and the name, not after the closing brace.
14224       Attributes are now allowed for enums and elaborated-type-specifiers
14225       that only declare a type.
14226       Members of the anonymous namespace are now local to a particular
14227       translation unit, along with any other declarations which use them,
14228       though they are still treated as having external linkage for
14229       language semantics.
14230     * The (undocumented) extension which permitted templates with default
14231       arguments to be bound to template template parameters with fewer
14232       parameters has been removed. For example:
14233        template <template <typename> class C>
14234        void f(C<double>) {}
14235
14236        template <typename T, typename U = int>
14237        struct S {};
14238
14239        template void f(S<double>);
14240
14241       is no longer accepted by G++. The reason this code is not accepted
14242       is that S is a template with two parameters; therefore, it cannot
14243       be bound to C which has only one parameter.
14244     * The <?, >?, <?=, and >?= operators, deprecated in previous GCC
14245       releases, have been removed.
14246     * The command-line option -fconst-strings, deprecated in previous GCC
14247       releases, has been removed.
14248     * The configure variable enable-__cxa_atexit is now enabled by
14249       default for more targets. Enabling this variable is necessary in
14250       order for static destructors to be executed in the correct order,
14251       but it depends upon the presence of a non-standard C library in the
14252       target library in order to work. The variable is now enabled for
14253       more targets which are known to have suitable C libraries.
14254     * -Wextra will produce warnings for if statements with a semicolon as
14255       the only body, to catch code like:
14256         if (a);
14257            return 1;
14258         return 0;
14259
14260       To suppress the warning in valid cases, use { } instead.
14261     * The C++ front end now also produces strict aliasing warnings when
14262       -fstrict-aliasing -Wstrict-aliasing is in effect.
14263
14264    Runtime Library (libstdc++)
14265
14266     * Added support for TR1 <random>, <complex>, and C compatibility
14267       headers. In addition, a lock-free version of shared_ptr was
14268       contributed as part of Phillip Jordan's Google Summer of Code
14269       project on lock-free containers.
14270     * In association with the Summer of Code work on lock-free
14271       containers, the interface for atomic builtins was adjusted,
14272       creating simpler alternatives for non-threaded code paths. Also,
14273       usage was consolidated and all elements were moved from namespace
14274       std to namespace__gnu_cxx. Affected interfaces are the functions
14275       __exchange_and_add, __atomic_add, and the objects __mutex,
14276       __recursive_mutex, and __scoped_lock.
14277     * Support for versioning weak symbol names via namespace association
14278       was added. However, as this changes the names of exported symbols,
14279       this is turned off by default in the current ABI. Intrepid users
14280       can enable this feature by using
14281       --enable-symvers=gnu-versioned-namespace during configuration.
14282     * Revised, simplified, and expanded policy-based associative
14283       containers, including data types for tree and trie forms
14284       (basic_tree, tree, trie), lists (list_update), and both
14285       collision-chaining and probing hash-based containers
14286       (basic_hash_table, cc_hash_table, gp_hash_table). More details per
14287       the [3]documentation.
14288     * The implementation of the debug mode was modified, whereby the
14289       debug namespaces were nested inside of namespace std and namespace
14290       __gnu_cxx in order to resolve some long standing corner cases
14291       involving name lookup. Debug functionality from the policy-based
14292       data structures was consolidated and enabled with the single macro,
14293       _GLIBCXX_DEBUG. See PR 26142 for more information.
14294     * Added extensions for type traits: __conditional_type,
14295       __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if.
14296     * Added a typelist implementation for compile-time meta-programming.
14297       Elements for typelist construction and operation can be found
14298       within namespace __gnu_cxx::typelist.
14299     * Added a new allocator, __gnu_cxx::throw_allocator, for testing
14300       exception-safety.
14301     * Enabled library-wide visibility control, allowing -fvisibility to
14302       be used.
14303     * Consolidated all nested namespaces and the conversion of
14304       __gnu_internal implementation-private details to anonymous
14305       namespaces whenever possible.
14306     * Implemented LWG resolutions DR 431 and DR 538.
14307
14308  Fortran
14309
14310     * Support for allocatable components has been added (TR 15581 and
14311       Fortran 2003).
14312     * Support for the Fortran 2003 streaming IO extension has been added.
14313     * The GNU Fortran compiler now uses 4-byte record markers by default
14314       for unformatted files to be compatible with g77 and most other
14315       compilers. The implementation allows for records greater than 2 GB
14316       and is compatible with several other compilers. Older versions of
14317       gfortran used 8-byte record markers by default (on most systems).
14318       In order to change the length of the record markers, e.g. to read
14319       unformatted files created by older gfortran versions, the
14320       [4]-frecord-marker=8 option can be used.
14321
14322  Java (GCJ)
14323
14324     * A new command-line option -static-libgcj has been added for targets
14325       that use a linker compatible with GNU Binutils. As its name
14326       implies, this causes libgcj to be linked statically. In some cases
14327       this causes the resulting executable to start faster and use less
14328       memory than if the shared version of libgcj were used. However
14329       caution should be used as it can also cause essential parts of the
14330       library to be omitted. Some of these issues are discussed in:
14331       [5]https://gcc.gnu.org/wiki/Statically_linking_libgcj
14332     * fastjar is no longer bundled with GCC. To build libgcj, you will
14333       need either InfoZIP (both zip and unzip) or an external jar
14334       program. In the former case, the GCC build will install a jar shell
14335       script that is based on InfoZIP and provides the same functionality
14336       as fastjar.
14337
14338New Targets and Target Specific Improvements
14339
14340  IA-32/x86-64
14341
14342     * -mtune=generic can now be used to generate code running well on
14343       common x86 chips. This includes AMD Athlon, AMD Opteron, Intel
14344       Pentium-M, Intel Pentium 4 and Intel Core 2.
14345     * -mtune=native and -march=native will produce code optimized for the
14346       host architecture as detected using the cpuid instruction.
14347     * Added a new command-line option -fstackrealign and and
14348       __attribute__ ((force_align_arg_pointer)) to realign the stack at
14349       runtime. This allows functions compiled with a vector-aligned stack
14350       to be invoked from legacy objects that keep only word-alignment.
14351
14352  SPARC
14353
14354     * The default CPU setting has been changed from V7 to V9 in 32-bit
14355       mode on Solaris 7 and above. This is already the case in 64-bit
14356       mode. It can be overridden by specifying --with-cpu at configure
14357       time.
14358     * Back-end support of built-in functions for atomic memory access has
14359       been implemented.
14360     * Support for the Sun UltraSPARC T1 (Niagara) processor has been
14361       added.
14362
14363  M32C
14364
14365     * Various bug fixes have made some functions (notably, functions
14366       returning structures) incompatible with previous releases.
14367       Recompiling all libraries is recommended. Note that code quality
14368       has considerably improved since 4.1, making a recompile even more
14369       beneficial.
14370
14371  MIPS
14372
14373     * Added support for the Broadcom SB-1A core.
14374
14375  IA-64
14376
14377     * Added support for IA-64 data and control speculation. By default
14378       speculation is enabled only during second scheduler pass. A number
14379       of machine flags was introduced to control the usage of speculation
14380       for both scheduler passes.
14381
14382  HPPA
14383
14384     * Added Java language support (libffi and libjava) for 32-bit HP-UX
14385       11 target.
14386
14387Obsolete Systems
14388
14389Documentation improvements
14390
14391  PDF Documentation
14392
14393     * A make pdf target has been added to the top-level makefile,
14394       enabling automated production of PDF documentation files.
14395       (Front-ends external to GCC should modify their Make-lang.in file
14396       to add a lang.pdf: target.)
14397
14398Other significant improvements
14399
14400  Build system improvements
14401
14402     * All the components of the compiler are now bootstrapped by default.
14403       This improves the resilience to bugs in the system compiler or
14404       binary compatibility problems, as well as providing better testing
14405       of GCC 4.2 itself. In addition, if you build the compiler from a
14406       combined tree, the assembler, linker, etc. will also be
14407       bootstrapped (i.e. built with themselves).
14408       You can disable this behavior, and go back to the pre-GCC 4.2 set
14409       up, by configuring GCC with --disable-bootstrap.
14410     * The rules that configure follows to find target tools resemble more
14411       closely the locations that the built compiler will search. In
14412       addition, you can use the new configure option --with-target-tools
14413       to specify where to find the target tools used during the build,
14414       without affecting what the built compiler will use.
14415       This can be especially useful when building packages of GCC. For
14416       example, you may want to build GCC with GNU as or ld, even if the
14417       resulting compiler to work with the native assembler and linker. To
14418       do so, you can use --with-target-tools to point to the native
14419       tools.
14420
14421  Incompatible changes to the build system
14422
14423     * Front-ends external to GCC should modify their Make-lang.in file to
14424       replace double-colon rules (e.g. dvi::) with normal rules (like
14425       lang.dvi:). Front-end makefile hooks do not use double-colon rules
14426       anymore.
14427     * Up to GCC 4.1, a popular way to specify the target tools used
14428       during the build was to create directories named gas, binutils,
14429       etc. in the build tree, and create links to the tools from there.
14430       This does not work any more when the compiler is bootstrapped. The
14431       new configure option --with-target-tools provides a better way to
14432       achieve the same effect, and works for all native and cross
14433       settings.
14434
14435
14436    For questions related to the use of GCC, please consult these web
14437    pages and the [6]GCC manuals. If that fails, the
14438    [7]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14439    web pages and the development of GCC are welcome on our developer
14440    list at [8]gcc@gcc.gnu.org. All of [9]our lists have public archives.
14441
14442   Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and
14443   distribution of this entire article is permitted in any medium,
14444   provided this notice is preserved.
14445
14446   These pages are [11]maintained by the GCC team. Last modified
14447   2022-10-26.
14448
14449References
14450
14451   1. http://gcc.gnu.org/projects/gomp/
14452   2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
14453   3. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html
14454   4. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html
14455   5. https://gcc.gnu.org/wiki/Statically_linking_libgcj
14456   6. https://gcc.gnu.org/onlinedocs/
14457   7. mailto:gcc-help@gcc.gnu.org
14458   8. mailto:gcc@gcc.gnu.org
14459   9. https://gcc.gnu.org/lists.html
14460  10. https://www.fsf.org/
14461  11. https://gcc.gnu.org/about.html
14462======================================================================
14463http://gcc.gnu.org/gcc-4.1/index.html
14464
14465                             GCC 4.1 Release Series
14466
14467   (This release series is no longer supported.)
14468
14469   February 13, 2007
14470
14471   The [1]GNU project and the GCC developers are pleased to announce the
14472   release of GCC 4.1.2.
14473
14474   This release is a bug-fix release, containing fixes for regressions in
14475   GCC 4.1.1 relative to previous releases of GCC.
14476
14477Release History
14478
14479   GCC 4.1.2
14480          February 13, 2007 ([2]changes)
14481
14482   GCC 4.1.1
14483          May 24, 2006 ([3]changes)
14484
14485   GCC 4.1.0
14486          February 28, 2006 ([4]changes)
14487
14488References and Acknowledgements
14489
14490   GCC used to stand for the GNU C Compiler, but since the compiler
14491   supports several other languages aside from C, it now stands for the
14492   GNU Compiler Collection.
14493
14494   A list of [5]successful builds is updated as new information becomes
14495   available.
14496
14497   The GCC developers would like to thank the numerous people that have
14498   contributed new features, improvements, bug fixes, and other changes as
14499   well as test results to GCC. This [6]amazing group of volunteers is
14500   what makes GCC successful.
14501
14502   For additional information about GCC please refer to the [7]GCC project
14503   web site or contact the [8]GCC development mailing list.
14504
14505   To obtain GCC please use [9]our mirror sites or [10]our version control
14506   system.
14507
14508
14509    For questions related to the use of GCC, please consult these web
14510    pages and the [11]GCC manuals. If that fails, the
14511    [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14512    web pages and the development of GCC are welcome on our developer
14513    list at [13]gcc@gcc.gnu.org. All of [14]our lists have public
14514    archives.
14515
14516   Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and
14517   distribution of this entire article is permitted in any medium,
14518   provided this notice is preserved.
14519
14520   These pages are [16]maintained by the GCC team. Last modified
14521   2022-10-26.
14522
14523References
14524
14525   1. http://www.gnu.org/
14526   2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
14527   3. http://gcc.gnu.org/gcc-4.1/changes.html
14528   4. http://gcc.gnu.org/gcc-4.1/changes.html
14529   5. http://gcc.gnu.org/gcc-4.1/buildstat.html
14530   6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
14531   7. http://gcc.gnu.org/index.html
14532   8. mailto:gcc@gcc.gnu.org
14533   9. http://gcc.gnu.org/mirrors.html
14534  10. http://gcc.gnu.org/git.html
14535  11. https://gcc.gnu.org/onlinedocs/
14536  12. mailto:gcc-help@gcc.gnu.org
14537  13. mailto:gcc@gcc.gnu.org
14538  14. https://gcc.gnu.org/lists.html
14539  15. https://www.fsf.org/
14540  16. https://gcc.gnu.org/about.html
14541======================================================================
14542http://gcc.gnu.org/gcc-4.1/changes.html
14543
14544                             GCC 4.1 Release Series
14545                        Changes, New Features, and Fixes
14546
14547   The latest release in the 4.1 release series is [1]GCC 4.1.2.
14548
14549Caveats
14550
14551General Optimizer Improvements
14552
14553     * GCC now has infrastructure for inter-procedural optimizations and
14554       the following inter-procedural optimizations are implemented:
14555          + Profile guided inlining. When doing profile feedback guided
14556            optimization, GCC can now use the profile to make better
14557            informed decisions on whether inlining of a function is
14558            profitable or not. This means that GCC will no longer inline
14559            functions at call sites that are not executed very often, and
14560            that functions at hot call sites are more likely to be
14561            inlined.
14562            A new parameter min-inline-recursive-probability is also now
14563            available to throttle recursive inlining of functions with
14564            small average recursive depths.
14565          + Discovery of pure and const functions, a form of side-effects
14566            analysis. While older GCC releases could also discover such
14567            special functions, the new IPA-based pass runs earlier so that
14568            the results are available to more optimizers. The pass is also
14569            simply more powerful than the old one.
14570          + Analysis of references to static variables and type escape
14571            analysis, also forms of side-effects analysis. The results of
14572            these passes allow the compiler to be less conservative about
14573            call-clobbered variables and references. This results in more
14574            redundant loads being eliminated and in making static
14575            variables candidates for register promotion.
14576          + Improvement of RTL-based alias analysis. The results of type
14577            escape analysis are fed to the RTL type-based alias analyzer,
14578            allowing it to disambiguate more memory references.
14579          + Interprocedural constant propagation and function versioning.
14580            This pass looks for functions that are always called with the
14581            same constant value for one or more of the function arguments,
14582            and propagates those constants into those functions.
14583          + GCC will now eliminate static variables whose usage was
14584            optimized out.
14585          + -fwhole-program --combine can now be used to make all
14586            functions in program static allowing whole program
14587            optimization. As an exception, the main function and all
14588            functions marked with the new externally_visible attribute are
14589            kept global so that programs can link with runtime libraries.
14590     * GCC can now do a form of partial dead code elimination (PDCE) that
14591       allows code motion of expressions to the paths where the result of
14592       the expression is actually needed. This is not always a win, so the
14593       pass has been limited to only consider profitable cases. Here is an
14594       example:
14595    int foo (int *, int *);
14596    int
14597    bar (int d)
14598    {
14599      int a, b, c;
14600      b = d + 1;
14601      c = d + 2;
14602      a = b + c;
14603      if (d)
14604        {
14605          foo (&b, &c);
14606          a = b + c;
14607        }
14608      printf ("%d\n", a);
14609    }
14610
14611       The a = b + c can be sunk to right before the printf. Normal code
14612       sinking will not do this, it will sink the first one above into the
14613       else-branch of the conditional jump, which still gives you two
14614       copies of the code.
14615     * GCC now has a value range propagation pass. This allows the
14616       compiler to eliminate bounds checks and branches. The results of
14617       the pass can also be used to accurately compute branch
14618       probabilities.
14619     * The pass to convert PHI nodes to straight-line code (a form of
14620       if-conversion for GIMPLE) has been improved significantly. The two
14621       most significant improvements are an improved algorithm to
14622       determine the order in which the PHI nodes are considered, and an
14623       improvement that allow the pass to consider if-conversions of basic
14624       blocks with more than two predecessors.
14625     * Alias analysis improvements. GCC can now differentiate between
14626       different fields of structures in Tree-SSA's virtual operands form.
14627       This lets stores/loads from non-overlapping structure fields not
14628       conflict. A new algorithm to compute points-to sets was contributed
14629       that can allows GCC to see now that p->a and p->b, where p is a
14630       pointer to a structure, can never point to the same field.
14631     * Various enhancements to auto-vectorization:
14632          + Incrementally preserve SSA form when vectorizing.
14633          + Incrementally preserve loop-closed form when vectorizing.
14634          + Improvements to peeling for alignment: generate better code
14635            when the misalignment of an access is known at compile time,
14636            or when different accesses are known to have the same
14637            misalignment, even if the misalignment amount itself is
14638            unknown.
14639          + Consider dependence distance in the vectorizer.
14640          + Externalize generic parts of data reference analysis to make
14641            this analysis available to other passes.
14642          + Vectorization of conditional code.
14643          + Reduction support.
14644     * GCC can now partition functions in sections of hot and cold code.
14645       This can significantly improve performance due to better
14646       instruction cache locality. This feature works best together with
14647       profile feedback driven optimization.
14648     * A new pass to avoid saving of unneeded arguments to the stack in
14649       vararg functions if the compiler can prove that they will not be
14650       needed.
14651     * Transition of basic block profiling to tree level implementation
14652       has been completed. The new implementation should be considerably
14653       more reliable (hopefully avoiding profile mismatch errors when
14654       using -fprofile-use or -fbranch-probabilities) and can be used to
14655       drive higher level optimizations, such as inlining.
14656       The -ftree-based-profiling command-line option was removed and
14657       -fprofile-use now implies disabling old RTL level loop optimizer
14658       (-fno-loop-optimize). Speculative prefetching optimization
14659       (originally enabled by -fspeculative-prefetching) was removed.
14660
14661New Languages and Language specific improvements
14662
14663  C and Objective-C
14664
14665     * The old Bison-based C and Objective-C parser has been replaced by a
14666       new, faster hand-written recursive-descent parser.
14667
14668  Ada
14669
14670     * The build infrastructure for the Ada runtime library and tools has
14671       been changed to be better integrated with the rest of the build
14672       infrastructure of GCC. This should make doing cross builds of Ada a
14673       bit easier.
14674
14675  C++
14676
14677     * ARM-style name-injection of friend declarations is no longer the
14678       default. For example:
14679          struct S {
14680            friend void f();
14681          };
14682
14683          void g() { f(); }
14684       will not be accepted; instead a declaration of f will need to be
14685       present outside of the scope of S. The new -ffriend-injection
14686       option will enable the old behavior.
14687     * The (undocumented) extension which permitted templates with default
14688       arguments to be bound to template template parameters with fewer
14689       parameters has been deprecated, and will be removed in the next
14690       major release of G++. For example:
14691       template <template <typename> class C>
14692       void f(C<double>) {}
14693
14694       template <typename T, typename U = int>
14695       struct S {};
14696
14697       template void f(S<double>);
14698
14699       makes use of the deprecated extension. The reason this code is not
14700       valid ISO C++ is that S is a template with two parameters;
14701       therefore, it cannot be bound to C which has only one parameter.
14702
14703    Runtime Library (libstdc++)
14704
14705     * Optimization work:
14706          + A new implementation of std::search_n is provided, better
14707            performing in case of random access iterators.
14708          + Added further efficient specializations of istream functions,
14709            i.e., character array and string extractors.
14710          + Other smaller improvements throughout.
14711     * Policy-based associative containers, designed for high-performance,
14712       flexibility and semantic safety are delivered in ext/pb_assoc.
14713     * A versatile string class, __gnu_cxx::__versa_string, providing
14714       facilities conforming to the standard requirements for
14715       basic_string, is delivered in <ext/vstring.h>. In particular:
14716          + Two base classes are provided: the default one avoids
14717            reference counting and is optimized for short strings; the
14718            alternate one, still uses it while improving in a few low
14719            level areas (e.g., alignment). See vstring_fwd.h for some
14720            useful typedefs.
14721          + Various algorithms have been rewritten (e.g., replace), the
14722            code streamlined and simple optimizations added.
14723          + Option 3 of DR 431 is implemented for both available bases,
14724            thus improving the support for stateful allocators.
14725     * As usual, many bugs have been fixed (e.g., libstdc++/13583,
14726       libstdc++/23953) and LWG resolutions put into effect for the first
14727       time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1
14728       Issue 6.19). The implementation status of TR1 is now tracked in the
14729       docs in tr1.html.
14730
14731  Objective-C++
14732
14733     * A new language front end for Objective-C++ has been added. This
14734       language allows users to mix the object oriented features of
14735       Objective-C with those of C++.
14736
14737  Java (GCJ)
14738
14739     * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19
14740       features (plus some 0.20 bug-fixes)
14741          + Networking
14742               o The java.net.HttpURLConnection implementation no longer
14743                 buffers the entire response body in memory. This means
14744                 that response bodies larger than available memory can now
14745                 be handled.
14746          + (N)IO
14747               o NIO FileChannel.map implementation, fast bulk put
14748                 implementation for DirectByteBuffer (speeds up this
14749                 method 10x).
14750               o FileChannel.lock() and FileChannel.force() implemented.
14751          + XML
14752               o gnu.xml fix for nodes created outside a namespace
14753                 context.
14754               o Add support for output indenting and
14755                 cdata-section-elements output instruction in
14756                 xml.transform.
14757               o xml.xpath corrections for cases where elements/attributes
14758                 might have been created in non-namespace-aware mode.
14759                 Corrections to handling of XSL variables and minor
14760                 conformance updates.
14761          + AWT
14762               o GNU JAWT implementation, the AWT Native Interface, which
14763                 allows direct access to native screen resources from
14764                 within a Canvas's paint method. GNU Classpath Examples
14765                 comes with a Demo, see libjava/classpath/examples/README.
14766               o awt.datatransfer updated to 1.5 with support for
14767                 FlavorEvents. The gtk+ awt peers now allow copy/paste of
14768                 text, images, URIs/files and serialized objects with
14769                 other applications and tracking clipboard change events
14770                 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized
14771                 objects are supported). A GNU Classpath Examples
14772                 datatransfer Demo was added to show the new
14773                 functionality.
14774               o Split gtk+ awt peers event handling in two threads and
14775                 improve gdk lock handling (solves several awt lock ups).
14776               o Speed up awt Image loading.
14777               o Better gtk+ scrollbar peer implementation when using gtk+
14778                 >= 2.6.
14779               o Handle image loading errors correctly for gdkpixbuf and
14780                 MediaTracker.
14781               o Better handle GDK lock. Properly prefix gtkpeer native
14782                 functions (cp_gtk).
14783               o GdkGraphics2D has been updated to use Cairo 0.5.x or
14784                 higher.
14785               o BufferedImage and GtkImage rewrites. All image drawing
14786                 operations should now work correctly (flipping requires
14787                 gtk+ >= 2.6)
14788               o When gtk+ 2.6 or higher is installed the default log
14789                 handler will produce stack traces whenever a WARNING,
14790                 CRITICAL or ERROR message is produced.
14791          + Free Swing
14792               o The RepaintManager has been reworked for more efficient
14793                 painting, especially for large GUIs.
14794               o The layout manager OverlayLayout has been implemented,
14795                 the BoxLayout has been rewritten to make use of the
14796                 SizeRequirements utility class and caching for more
14797                 efficient layout.
14798               o Improved accessibility support.
14799               o Significant progress has been made in the implementation
14800                 of the javax.swing.plaf.metal package, with most UI
14801                 delegates in a working state now. Please test this with
14802                 your own applications and provide feedback that will help
14803                 us to improve this package.
14804               o The GUI demo (gnu.classpath.examples.swing.Demo) has been
14805                 extended to highlight various features in our Free Swing
14806                 implementation. And it includes a look and feel switcher
14807                 for Metal (default), Ocean and GNU themes.
14808               o The javax.swing.plaf.multi package is now implemented.
14809               o Editing and several key actions for JTree and JTable were
14810                 implemented.
14811               o Lots of icons and look and feel improvements for Free
14812                 Swing basic and metal themes were added. Try running the
14813                 GNU Classpath Swing Demo in examples
14814                 (gnu.classpath.examples.swing.Demo) with:
14815                 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee
14816                 l or
14817                 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee
14818                 l
14819               o Start of styled text capabilites for java.swing.text.
14820               o DefaultMutableTreeNode pre-order, post-order, depth-first
14821                 and breadth-first traversal enumerations implemented.
14822               o JInternalFrame colors and titlebar draw properly.
14823               o JTree is working up to par (icons, selection and keyboard
14824                 traversal).
14825               o JMenus were made more compatible in visual and
14826                 programmatic behavior.
14827               o JTable changeSelection and multiple selections
14828                 implemented.
14829               o JButton and JToggleButton change states work properly
14830                 now.
14831               o JFileChooser fixes.
14832               o revalidate() and repaint() fixes which make Free Swing
14833                 much more responsive.
14834               o MetalIconFactory implemented.
14835               o Free Swing Top-Level Compatibility. JFrame, JDialog,
14836                 JApplet, JInternalFrame, and JWindow are now 1.5
14837                 compatible in the sense that you can call add() and
14838                 setLayout() directly on them, which will have the same
14839                 effect as calling getContentPane().add() and
14840                 getContentPane().setLayout().
14841               o The JTree interface has been completed. JTrees now
14842                 recognizes mouse clicks and selections work.
14843               o BoxLayout works properly now.
14844               o Fixed GrayFilter to actually work.
14845               o Metal SplitPane implemented.
14846               o Lots of Free Swing text and editor stuff work now.
14847          + Free RMI and Corba
14848               o Andrew Watson, Vice President and Technical Director of
14849                 the Object Management Group, has officially assigned us
14850                 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will
14851                 mark remote classpath-specific system exceptions.
14852                 Obtaining the VMCID means that GNU Classpath now is a
14853                 recogniseable type of node in a highly interoperable
14854                 CORBA world.
14855               o GNU Classpath now includes the first working draft to
14856                 support the RMI over IIOP protocol. The current
14857                 implementation is capable of remote invocations,
14858                 transferring various Serializables and Externalizables
14859                 via RMI-IIOP protocol. It can flatten graphs and, at
14860                 least for the simple cases, is interoperable with 1.5
14861                 JDKs.
14862               o org.omg.PortableInterceptor and related functionality in
14863                 other packages is now implemented:
14864                    # The sever and client interceptors work as required
14865                      since 1.4.
14866                    # The IOR interceptor works as needed for 1.5.
14867               o The org.omg.DynamicAny package is completed and passes
14868                 the prepared tests.
14869               o The Portable Object Adapter should now support the output
14870                 of the recent IDL to java compilers. These compilers now
14871                 generate servants and not CORBA objects as before, making
14872                 the output depend on the existing POA implementation.
14873                 Completing POA means that such code can already be tried
14874                 to run on Classpath. Our POA is tested for the following
14875                 usager scenarios:
14876                    # POA converts servant to the CORBA object.
14877                    # Servant provides to the CORBA object.
14878                    # POA activates new CORBA object with the given Object
14879                      Id (byte array) that is later accessible for the
14880                      servant.
14881                    # During the first call, the ServantActivator provides
14882                      servant for this and all subsequent calls on the
14883                      current object.
14884                    # During each call, the ServantLocator provides
14885                      servant for this call only.
14886                    # ServantLocator or ServantActivator forwards call to
14887                      another server.
14888                    # POA has a single servant, responsible for all
14889                      objects.
14890                    # POA has a default servant, but some objects are
14891                      explicitly connected to they specific servants.
14892                 The POA is verified using tests from the former
14893                 cost.omg.org.
14894               o The CORBA implementation is now a working prototype that
14895                 should support features up to 1.3 inclusive. We invite
14896                 groups writing CORBA dependent applications to try
14897                 Classpath implementation, reporting any possible bugs.
14898                 The CORBA prototype is interoperable with Sun's
14899                 implementation v 1.4, transferring object references,
14900                 primitive types, narrow and wide strings, arrays,
14901                 structures, trees, abstract interfaces and value types
14902                 (feature of CORBA 2.3) between these two platforms.
14903                 Remote exceptions are transferred and handled correctly.
14904                 The stringified object references (IORs) from various
14905                 sources are parsed as required. The transient (for
14906                 current session) and permanent (till jre restart)
14907                 redirections work. Both Little and Big Endian encoded
14908                 messages are accepted. The implementation is verified
14909                 using tests from the former cost.omg.org. The current
14910                 release includes working examples (see the examples
14911                 directory), demonstrating the client-server
14912                 communication, using either CORBA Request or IDL-based
14913                 stub (usually generated by a IDL to java compiler). These
14914                 examples also show how to use the Classpath CORBA naming
14915                 service. The IDL to java compiler is not yet written, but
14916                 as our library must be compatible, it naturally accepts
14917                 the output of other idlj implementations.
14918          + Misc
14919               o Updated TimeZone data against Olson tzdata2005l.
14920               o Make zip and jar packages UTF-8 clean.
14921               o "native" code builds and compiles (warning free) on
14922                 Darwin and Solaris.
14923               o java.util.logging.FileHandler now rotates files.
14924               o Start of a generic JDWP framework in gnu/classpath/jdwp.
14925                 This is unfinished, but feedback (at classpath@gnu.org)
14926                 from runtime hackers is greatly appreciated. Although
14927                 most of the work is currently being done around gcj/gij
14928                 we want this framework to be as VM neutral as possible.
14929                 Early design is described in:
14930                 [2]https://gcc.gnu.org/ml/java/2005-05/msg00260.html
14931               o QT4 AWT peers, enable by giving configure
14932                 --enable-qt-peer. Included, but not ready for production
14933                 yet. They are explicitly disabled and not supported. But
14934                 if you want to help with the development of these new
14935                 features we are interested in feedback. You will have to
14936                 explicitly enable them to try them out (and they will
14937                 most likely contain bugs).
14938               o Documentation fixes all over the place. See
14939                 [3]https://developer.classpath.org/doc/
14940
14941New Targets and Target Specific Improvements
14942
14943  IA-32/x86-64
14944
14945     * The x86-64 medium model (that allows building applications whose
14946       data segment exceeds 4GB) was redesigned to match latest ABI draft.
14947       New implementation split large datastructures into separate segment
14948       improving performance of accesses to small datastructures and also
14949       allows linking of small model libraries into medium model programs
14950       as long as the libraries are not accessing the large datastructures
14951       directly. Medium model is also supported in position independent
14952       code now.
14953       The ABI change results in partial incompatibility among medium
14954       model objects. Linking medium model libraries (or objects) compiled
14955       with new compiler into medium model program compiled with older
14956       will likely result in exceeding ranges of relocations.
14957       Binutils 2.16.91 or newer are required for compiling medium model
14958       now.
14959
14960  RS6000 (POWER/PowerPC)
14961
14962     * The AltiVec vector primitives in <altivec.h> are now implemented in
14963       a way that puts a smaller burden on the preprocessor, instead
14964       processing the "overloading" in the front ends. This should benefit
14965       compilation speed on AltiVec vector code.
14966     * AltiVec initializers now are generated more efficiently.
14967     * The popcountb instruction available on POWER5 now is generated.
14968     * The floating point round to integer instructions available on
14969       POWER5+ now is generated.
14970     * Floating point divides can be synthesized using the floating point
14971       reciprocal estimate instructions.
14972     * Double precision floating point constants are initialized as single
14973       precision values if they can be represented exactly.
14974
14975  S/390, zSeries and System z9
14976
14977     * Support for the IBM System z9 109 processor has been added. When
14978       using the -march=z9-109 option, the compiler will generate code
14979       making use of instructions provided by the extended immediate
14980       facility.
14981     * Support for 128-bit IEEE floating point has been added. When using
14982       the -mlong-double-128 option, the compiler will map the long double
14983       data type to 128-bit IEEE floating point. Using this option
14984       constitutes an ABI change, and requires glibc support.
14985     * Various changes to improve performance of generated code have been
14986       implemented, including:
14987          + In functions that do not require a literal pool, register %r13
14988            (which is traditionally reserved as literal pool pointer), can
14989            now be freely used for other purposes by the compiler.
14990          + More precise tracking of register use allows the compiler to
14991            generate more efficient function prolog and epilog code in
14992            certain cases.
14993          + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING
14994            instructions are now used to implement C string functions.
14995          + The MOVE CHARACTER instruction with single byte overlap is now
14996            used to implement the memset function with non-zero fill byte.
14997          + The LOAD ZERO instructions are now used where appropriate.
14998          + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK,
14999            and INSERT IMMEDIATE instructions are now used more frequently
15000            to optimize bitfield operations.
15001          + The BRANCH ON COUNT instruction is now used more frequently.
15002            In particular, the fact that a loop contains a subroutine call
15003            no longer prevents the compiler from using this instruction.
15004          + The compiler is now aware that all shift and rotate
15005            instructions implicitly truncate the shift count to six bits.
15006     * Back-end support for the following generic features has been
15007       implemented:
15008          + The full set of [4]built-in functions for atomic memory
15009            access.
15010          + The -fstack-protector feature.
15011          + The optimization pass avoiding unnecessary stores of incoming
15012            argument registers in functions with variable argument list.
15013
15014  SPARC
15015
15016     * The default code model in 64-bit mode has been changed from
15017       Medium/Anywhere to Medium/Middle on Solaris.
15018     * TLS support is disabled by default on Solaris prior to release 10.
15019       It can be enabled on TLS-capable Solaris 9 versions (4/04 release
15020       and later) by specifying --enable-tls at configure time.
15021
15022  MorphoSys
15023
15024     * Support has been added for this new architecture.
15025
15026Obsolete Systems
15027
15028Documentation improvements
15029
15030Other significant improvements
15031
15032     * GCC can now emit code for protecting applications from
15033       stack-smashing attacks. The protection is realized by buffer
15034       overflow detection and reordering of stack variables to avoid
15035       pointer corruption.
15036     * Some built-in functions have been fortified to protect them against
15037       various buffer overflow (and format string) vulnerabilities.
15038       Compared to the mudflap bounds checking feature, the safe builtins
15039       have far smaller overhead. This means that programs built using
15040       safe builtins should not experience any measurable slowdown.
15041
15042GCC 4.1.2
15043
15044   This is the [5]list of problem reports (PRs) from GCC's bug tracking
15045   system that are known to be fixed in the 4.1.2 release. This list might
15046   not be complete (that is, it is possible that some PRs that have been
15047   fixed are not listed here).
15048
15049   When generating code for a shared library, GCC now recognizes that
15050   global functions may be replaced when the program runs. Therefore, it
15051   is now more conservative in deducing information from the bodies of
15052   functions. For example, in this example:
15053    void f() {}
15054    void g() {
15055     try { f(); }
15056     catch (...) {
15057       cout << "Exception";
15058     }
15059    }
15060
15061   G++ would previously have optimized away the catch clause, since it
15062   would have concluded that f cannot throw exceptions. Because users may
15063   replace f with another function in the main body of the program, this
15064   optimization is unsafe, and is no longer performed. If you wish G++ to
15065   continue to optimize as before, you must add a throw() clause to the
15066   declaration of f to make clear that it does not throw exceptions.
15067
15068
15069    For questions related to the use of GCC, please consult these web
15070    pages and the [6]GCC manuals. If that fails, the
15071    [7]gcc-help@gcc.gnu.org mailing list might help. Comments on these
15072    web pages and the development of GCC are welcome on our developer
15073    list at [8]gcc@gcc.gnu.org. All of [9]our lists have public archives.
15074
15075   Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and
15076   distribution of this entire article is permitted in any medium,
15077   provided this notice is preserved.
15078
15079   These pages are [11]maintained by the GCC team. Last modified
15080   2022-10-26.
15081
15082References
15083
15084   1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
15085   2. https://gcc.gnu.org/ml/java/2005-05/msg00260.html
15086   3. https://developer.classpath.org/doc/
15087   4. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html
15088   5. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2
15089   6. https://gcc.gnu.org/onlinedocs/
15090   7. mailto:gcc-help@gcc.gnu.org
15091   8. mailto:gcc@gcc.gnu.org
15092   9. https://gcc.gnu.org/lists.html
15093  10. https://www.fsf.org/
15094  11. https://gcc.gnu.org/about.html
15095======================================================================
15096http://gcc.gnu.org/gcc-4.0/index.html
15097
15098                             GCC 4.0 Release Series
15099
15100   (This release series is no longer supported.)
15101
15102   January 31, 2007
15103
15104   The [1]GNU project and the GCC developers are pleased to announce the
15105   release of GCC 4.0.4.
15106
15107   This release is a bug-fix release, containing fixes for regressions in
15108   GCC 4.0.3 relative to previous releases of GCC.
15109
15110Release History
15111
15112   GCC 4.0.4
15113          January 31, 2007 ([2]changes)
15114
15115   GCC 4.0.3
15116          March 10, 2006 ([3]changes)
15117
15118   GCC 4.0.2
15119          September 28, 2005 ([4]changes)
15120
15121   GCC 4.0.1
15122          July 7, 2005 ([5]changes)
15123
15124   GCC 4.0.0
15125          April 20, 2005 ([6]changes)
15126
15127References and Acknowledgements
15128
15129   GCC used to stand for the GNU C Compiler, but since the compiler
15130   supports several other languages aside from C, it now stands for the
15131   GNU Compiler Collection.
15132
15133   A list of [7]successful builds is updated as new information becomes
15134   available.
15135
15136   The GCC developers would like to thank the numerous people that have
15137   contributed new features, improvements, bug fixes, and other changes as
15138   well as test results to GCC. This [8]amazing group of volunteers is
15139   what makes GCC successful.
15140
15141   For additional information about GCC please refer to the [9]GCC project
15142   web site or contact the [10]GCC development mailing list.
15143
15144   To obtain GCC please use [11]our mirror sites, or [12]our version
15145   control system.
15146
15147
15148    For questions related to the use of GCC, please consult these web
15149    pages and the [13]GCC manuals. If that fails, the
15150    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
15151    web pages and the development of GCC are welcome on our developer
15152    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
15153    archives.
15154
15155   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
15156   distribution of this entire article is permitted in any medium,
15157   provided this notice is preserved.
15158
15159   These pages are [18]maintained by the GCC team. Last modified
15160   2022-10-26.
15161
15162References
15163
15164   1. http://www.gnu.org/
15165   2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
15166   3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3
15167   4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2
15168   5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1
15169   6. http://gcc.gnu.org/gcc-4.0/changes.html
15170   7. http://gcc.gnu.org/gcc-4.0/buildstat.html
15171   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
15172   9. http://gcc.gnu.org/index.html
15173  10. mailto:gcc@gcc.gnu.org
15174  11. http://gcc.gnu.org/mirrors.html
15175  12. http://gcc.gnu.org/git.html
15176  13. https://gcc.gnu.org/onlinedocs/
15177  14. mailto:gcc-help@gcc.gnu.org
15178  15. mailto:gcc@gcc.gnu.org
15179  16. https://gcc.gnu.org/lists.html
15180  17. https://www.fsf.org/
15181  18. https://gcc.gnu.org/about.html
15182======================================================================
15183http://gcc.gnu.org/gcc-4.0/changes.html
15184
15185                             GCC 4.0 Release Series
15186                        Changes, New Features, and Fixes
15187
15188   The latest release in the 4.0 release series is [1]GCC 4.0.4.
15189
15190Caveats
15191
15192     * GCC now generates location lists by default when compiling with
15193       debug info and optimization.
15194          + GDB 6.0 and older crashes when it sees location lists. GDB 6.1
15195            or later is needed to debug binaries containing location
15196            lists.
15197          + When you are trying to view a value of a variable in a part of
15198            a function where it has no location (for example when the
15199            variable is no longer used and thus its location was used for
15200            something else) GDB will say that it is not available.
15201       You can disable generating location lists by -fno-var-tracking.
15202     * GCC no longer accepts the -fwritable-strings option. Use named
15203       character arrays when you need a writable string.
15204     * The options -freduce-all-givs and -fmove-all-movables have been
15205       discontinued. They were used to circumvent a shortcoming in the
15206       heuristics of the old loop optimization code with respect to common
15207       Fortran constructs. The new (tree) loop optimizer works differently
15208       and doesn't need those work-arounds.
15209     * The graph-coloring register allocator, formerly enabled by the
15210       option -fnew-ra, has been discontinued.
15211     * -I- has been deprecated. -iquote is meant to replace the need for
15212       this option.
15213     * The MIPS -membedded-pic and -mrnames options have been removed.
15214     * All MIPS targets now require the GNU assembler. In particular, IRIX
15215       configurations can no longer use the MIPSpro assemblers, although
15216       they do still support the MIPSpro linkers.
15217     * The SPARC option -mflat has been removed.
15218     * English-language diagnostic messages will now use Unicode quotation
15219       marks in UTF-8 locales. (Non-English messages already used the
15220       quotes appropriate for the language in previous releases.) If your
15221       terminal does not support UTF-8 but you are using a UTF-8 locale
15222       (such locales are the default on many GNU/Linux systems) then you
15223       should set LC_CTYPE=C in the environment to disable that locale.
15224       Programs that parse diagnostics and expect plain ASCII
15225       English-language messages should set LC_ALL=C. See [2]Markus Kuhn's
15226       explanation of Unicode quotation marks for more information.
15227     * The specs file is no longer installed on most platforms. Most users
15228       will be totally unaffected. However, if you are accustomed to
15229       editing the specs file yourself, you will now have to use the
15230       -dumpspecs option to generate the specs file, and then edit the
15231       resulting file.
15232
15233General Optimizer Improvements
15234
15235     * The [3]tree ssa branch has been merged. This merge has brought in a
15236       completely new optimization framework based on a higher level
15237       intermediate representation than the existing RTL representation.
15238       Numerous new code transformations based on the new framework are
15239       available in GCC 4.0, including:
15240          + Scalar replacement of aggregates
15241          + Constant propagation
15242          + Value range propagation
15243          + Partial redundancy elimination
15244          + Load and store motion
15245          + Strength reduction
15246          + Dead store elimination
15247          + Dead and unreachable code elimination
15248          + [4]Autovectorization
15249          + Loop interchange
15250          + Tail recursion by accumulation
15251       Many of these passes outperform their counterparts from previous
15252       GCC releases.
15253     * [5]Swing Modulo Scheduling (SMS). An RTL level instruction
15254       scheduling optimization intended for loops that perform heavy
15255       computations.
15256
15257New Languages and Language specific improvements
15258
15259  C family
15260
15261     * The sentinel attribute has been added to GCC. This function
15262       attribute allows GCC to warn when variadic functions such as execl
15263       are not NULL terminated. See the GCC manual for a complete
15264       description of its behavior.
15265     * Given __attribute__((alias("target"))) it is now an error if target
15266       is not a symbol, defined in the same translation unit. This also
15267       applies to aliases created by #pragma weak alias=target. This is
15268       because it's meaningless to define an alias to an undefined symbol.
15269       On Solaris, the native assembler would have caught this error, but
15270       GNU as does not.
15271
15272  C and Objective-C
15273
15274     * The -Wstrict-aliasing=2 option has been added. This warning catches
15275       all unsafe cases, but it may also give a warning for some cases
15276       that are safe.
15277     * The cast-as-lvalue, conditional-expression-as-lvalue and
15278       compound-expression-as-lvalue extensions, which were deprecated in
15279       3.3.4 and 3.4, have been removed.
15280     * The -fwritable-strings option, which was deprecated in 3.4, has
15281       been removed.
15282     * #pragma pack() semantics have been brought closer to those used by
15283       other compilers. This also applies to C++.
15284     * Taking the address of a variable with register storage is invalid
15285       in C. GCC now issues an error instead of a warning.
15286     * Arrays of incomplete element type are invalid in C. GCC now issues
15287       an error for such arrays. Declarations such as extern struct s x[];
15288       (where struct s has not been defined) can be moved after the
15289       definition of struct s. Function parameters declared as arrays of
15290       incomplete type can instead be declared as pointers.
15291
15292  C++
15293
15294     * When compiling without optimizations (-O0), the C++ front end is
15295       much faster than in any previous versions of GCC. Independent
15296       testers have measured speed-ups up to 25% in real-world production
15297       code, compared to the 3.4 family (which was already the fastest
15298       version to date). Upgrading from older versions might show even
15299       bigger improvements.
15300     * ELF visibility attributes can now be applied to a class type, so
15301       that it affects every member function of a class at once, without
15302       having to specify each individually:
15303class __attribute__ ((visibility("hidden"))) Foo
15304{
15305   int foo1();
15306   void foo2();
15307};
15308       The syntax is deliberately similar to the __declspec() system used
15309       by Microsoft Windows based compilers, allowing cross-platform
15310       projects to easily reuse their existing macro system for denoting
15311       exports and imports. By explicitly marking internal classes never
15312       used outside a binary as hidden, one can completely avoid PLT
15313       indirection overheads during their usage by the compiler. You can
15314       find out more about the advantages of this at
15315       [6]https://www.akkadia.org/drepper/dsohowto.pdf
15316     * The -fvisibility-inlines-hidden option has been added which marks
15317       all inlineable functions as having hidden ELF visibility, thus
15318       removing their symbol and typeinfo from the exported symbol table
15319       of the output ELF binary. Using this option can reduce the exported
15320       symbol count of template-heavy code by up to 40% with no code
15321       change at all, thus notably improving link and load times for the
15322       binary as well as a reduction in size of up to 10%. Also, check the
15323       new [7]-fvisibility option.
15324     * The compiler now uses the library interface specified by the [8]C++
15325       ABI for thread-safe initialization of function-scope static
15326       variables. Most users should leave this alone, but embedded
15327       programmers may want to disable this by specifying
15328       -fno-threadsafe-statics for a small savings in code size.
15329     * Taking the address of an explicit register variable is no longer
15330       supported. Note that C++ allows taking the address of variables
15331       with register storage so this will continue to compile with a
15332       warning. For example, assuming that r0 is a machine register:
15333register int foo asm ("r0");
15334register int bar;
15335&foo; // error, no longer accepted
15336&bar; // OK, with a warning
15337     * G++ has an undocumented extension to virtual function covariancy
15338       rules that allowed the overrider to return a type that was
15339       implicitly convertable to the overridden function's return type.
15340       For instance a function returning void * could be overridden by a
15341       function returning T *. This is now deprecated and will be removed
15342       in a future release.
15343     * The G++ minimum and maximum operators (<? and >?) and their
15344       compound forms (<?=) and >?=) have been deprecated and will be
15345       removed in a future version. Code using these operators should be
15346       modified to use std::min and std::max instead.
15347     * Declaration of nested classes of class templates as friends are
15348       supported:
15349template <typename T> struct A {
15350  class B {};
15351};
15352class C {
15353  template <typename T> friend class A<T>::B;
15354};
15355       This complements the feature member functions of class templates as
15356       friends introduced in GCC 3.4.0.
15357     * When declaring a friend class using an unqualified name, classes
15358       outside the innermost non-class scope are not searched:
15359class A;
15360namespace N {
15361  class B {
15362    friend class A;   // Refer to N::A which has not been declared yet
15363                      // because name outside namespace N are not searched
15364    friend class ::A; // Refer to ::A
15365  };
15366}
15367       Hiding the friend name until declaration is still not implemented.
15368     * Friends of classes defined outside their namespace are correctly
15369       handled:
15370namespace N {
15371  class A;
15372}
15373class N::A {
15374  friend class B; // Refer to N::B in GCC 4.0.0
15375                  // but ::B in earlier versions of GCC
15376};
15377
15378    Runtime Library (libstdc++)
15379
15380     * Optimization work:
15381          + Added efficient specializations of istream functions for char
15382            and wchar_t.
15383          + Further performance tuning of strings, in particular wrt
15384            single-char append and getline.
15385          + iter_swap - and therefore most of the mutating algorithms -
15386            now makes an unqualified call to swap when the value_type of
15387            the two iterators is the same.
15388     * A large subset of the features in Technical Report 1 (TR1 for
15389       short) is experimentally delivered (i.e., no guarantees about the
15390       implementation are provided. In particular it is not promised that
15391       the library will remain link-compatible when code using TR1 is
15392       used):
15393          + General utilities such as reference_wrapper and shared_ptr.
15394          + Function objects, i.e., result_of, mem_fn, bind, function.
15395          + Support for metaprogramming.
15396          + New containers such as tuple, array, unordered_set,
15397            unordered_map, unordered_multiset, unordered_multimap.
15398     * As usual, many bugs have been fixed and LWG resolutions implemented
15399       for the first time (e.g., DR 409).
15400
15401  Java
15402
15403     * In order to prevent naming conflicts with other implementations of
15404       these tools, some GCJ binaries have been renamed:
15405          + rmic is now grmic,
15406          + rmiregistry is now grmiregistry, and
15407          + jar is now fastjar.
15408       In particular, these names were problematic for the jpackage.org
15409       packaging conventions which install symlinks in /usr/bin that point
15410       to the preferred versions of these tools.
15411     * The -findirect-dispatch argument to the compiler now works and
15412       generates code following a new "binary compatibility" ABI. Code
15413       compiled this way follows the binary compatibility rules of the
15414       Java Language Specification.
15415     * libgcj now has support for using GCJ as a JIT, using the
15416       gnu.gcj.jit family of system properties.
15417     * libgcj can now find a shared library corresponding to the bytecode
15418       representation of a class. See the documentation for the new
15419       gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system
15420       property.
15421     * There have been many improvements to the class library. Here are
15422       some highlights:
15423          + Much more of AWT and Swing exist.
15424          + Many new packages and classes were added, including
15425            java.util.regex, java.net.URI, javax.crypto,
15426            javax.crypto.interfaces, javax.crypto.spec, javax.net,
15427            javax.net.ssl, javax.security.auth,
15428            javax.security.auth.callback, javax.security.auth.login,
15429            javax.security.auth.x500, javax.security.sasl, org.ietf.jgss,
15430            javax.imageio, javax.imageio.event, javax.imageio.spi,
15431            javax.print, javax.print.attribute,
15432            javax.print.attribute.standard, javax.print.event, and
15433            javax.xml
15434          + Updated SAX and DOM, and imported GNU JAXP
15435
15436  Fortran
15437
15438     * A new [9]Fortran front end has replaced the aging GNU Fortran 77
15439       front end. The new front end supports Fortran 90 and Fortran 95. It
15440       may not yet be as stable as the old Fortran front end.
15441
15442  Ada
15443
15444     * Ada (with tasking and Zero Cost Exceptions) is now available on
15445       many more targets, including but not limited to: alpha-linux,
15446       hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux,
15447       s390x-linux, sparc-linux.
15448     * Some of the new Ada 2005 features are now implemented like
15449       Wide_Wide_Character and Ada.Containers.
15450     * Many bugs have been fixed, tools and documentation improved.
15451     * To compile Ada from the sources, install an older working Ada
15452       compiler and then use --enable-languages=ada at configuration time,
15453       since the Ada front end is not currently activated by default. See
15454       the [10]Installing GCC for details.
15455
15456New Targets and Target Specific Improvements
15457
15458  H8/300
15459
15460     * The frame layout has changed. In the new layout, the prologue of a
15461       function first saves registers and then allocate space for locals,
15462       resulting in an 1% improvement on code size.
15463
15464  IA-32/x86-64 (AMD64)
15465
15466     * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10,
15467       log1p, log2, logb and tan mathematical builtins (and their float
15468       and long double variants) are now implemented as inline x87
15469       intrinsics when using -ffast-math.
15470     * The ceil, floor, nearbyint, rint and trunc mathematical builtins
15471       (and their float and long double variants) are now implemented as
15472       inline x87 intrinsics when using -ffast-math.
15473     * The x87's fsincos instruction is now used automatically with
15474       -ffast-math when calculating both the sin and cos of the same
15475       argument.
15476     * Instruction selection for multiplication and division by constants
15477       has been improved.
15478
15479  IA-64
15480
15481     * Floating point division, integer division and sqrt are now inlined,
15482       resulting in significant performance improvements on some codes.
15483
15484  MIPS
15485
15486     * Division by zero checks now use conditional traps if the target
15487       processor supports them. This decreases code size by one word per
15488       division operation. The old behavior (branch and break) can be
15489       obtained either at configure time by passing --with-divide=breaks
15490       to configure or at runtime by passing -mdivide-breaks to GCC.
15491     * Support for MIPS64 paired-single instructions has been added. It is
15492       enabled by -mpaired-single and can be accessed using both the
15493       target-independent vector extensions and new MIPS-specific built-in
15494       functions.
15495     * Support for the MIPS-3D ASE has been added. It is enabled by
15496       -mips3d and provides new MIPS-3D-specific built-in functions.
15497     * The -mexplicit-relocs option now supports static n64 code (as is
15498       used, for example, in 64-bit linux kernels). -mexplicit-relocs
15499       should now be feature-complete and is enabled by default when GCC
15500       is configured to use a compatible assembler.
15501     * Support for the NEC VR4130 series has been added. This support
15502       includes the use of VR-specific instructions and a new VR4130
15503       scheduler. Full VR4130 support can be selected with -march=vr4130
15504       while code for any ISA can be tuned for the VR4130 using
15505       -mtune=vr4130. There is also a new -mvr4130-align option that
15506       produces better schedules at the cost of increased code size.
15507     * Support for the Broadcom SB-1 has been extended. There is now an
15508       SB-1 scheduler as well as support for the SB-1-specific
15509       paired-single instructions. Full SB-1 support can be selected with
15510       -march=sb1 while code for any ISA can be optimized for the SB-1
15511       using -mtune=sb1.
15512     * The compiler can now work around errata in R4000, R4400, VR4120 and
15513       VR4130 processors. These workarounds are enabled by -mfix-r4000,
15514       -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120
15515       and VR4130 workarounds need binutils 2.16 or above.
15516     * IRIX shared libraries are now installed into the standard library
15517       directories: o32 libraries go into lib/, n32 libraries go into
15518       lib32/ and n64 libraries go into lib64/.
15519     * The compiler supports a new -msym32 option. It can be used to
15520       optimize n64 code in which all symbols are known to have 32-bit
15521       values.
15522
15523  S/390 and zSeries
15524
15525     * New command-line options help to generate code intended to run in
15526       an environment where stack space is restricted, e.g. Linux kernel
15527       code:
15528          + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time
15529            warnings for single functions that require large or dynamic
15530            stack frames.
15531          + -mstack-size and -mstack-guard generate code that checks for
15532            stack overflow at run time.
15533          + -mpacked-stack generates code that reduces the stack frame
15534            size of many functions by reusing unneeded parts of the stack
15535            bias area.
15536     * The -msoft-float option now ensures that generated code never
15537       accesses floating point registers.
15538     * The s390x-ibm-tpf target now fully supports C++, including
15539       exceptions and threads.
15540     * Various changes to improve performance of the generated code have
15541       been implemented, including:
15542          + GCC now uses sibling calls where possible.
15543          + Condition code handling has been optimized, allowing GCC to
15544            omit redundant comparisons in certain cases.
15545          + The cost function guiding many optimizations has been refined
15546            to more accurately represent the z900 and z990 processors.
15547          + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW
15548            instructions are now used to avoid conditional branches in
15549            certain cases.
15550          + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to
15551            optimize address arithmetic required to access large stack
15552            frames.
15553          + GCC now makes more efficient use of memory-to-memory type
15554            instructions (MVC, CLC, ...).
15555          + More precise tracking of special register use allows better
15556            instruction scheduling, in particular of the function prologue
15557            and epilogue sequences.
15558          + The Java front end now generates inline code to implement
15559            integer division, instead of calling library routines.
15560
15561  SPARC
15562
15563     * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and
15564       -mf934 have been removed. They have been replaced with -mcpu=xxx.
15565     * The internal model used to estimate the relative cost of each
15566       instruction has been updated. It is expected to give better results
15567       on recent UltraSPARC processors.
15568     * Code generation for function prologues and epilogues has been
15569       improved, resulting in better scheduling and allowing multiple exit
15570       points in functions.
15571     * Support for Sun's Visual Instruction Set (VIS) has been enhanced.
15572       It is enabled by -mvis and provides new built-in functions for VIS
15573       instructions on UltraSPARC processors.
15574     * The option -mapp-regs has been turned on by default on Solaris too.
15575
15576  NetWare
15577
15578     * Novell NetWare (on ix86, no other hardware platform was ever really
15579       supported by this OS) has been re-enabled and the ABI supported by
15580       GCC has been brought into sync with that of MetroWerks CodeWarrior
15581       (the ABI previously supported was that of some Unix systems, which
15582       NetWare never tried to support).
15583
15584Obsolete Systems
15585
15586   Support for a number of older systems has been declared obsolete in GCC
15587   4.0. Unless there is activity to revive them, the next release of GCC
15588   will have their sources permanently removed.
15589
15590   All GCC ports for the following processor architectures have been
15591   declared obsolete:
15592     * Intel i860
15593     * Ubicom IP2022
15594     * National Semiconductor NS32K (ns32k)
15595     * Texas Instruments TMS320C[34]x
15596
15597   Also, those for some individual systems have been obsoleted:
15598     * SPARC family
15599          + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf,
15600            sparc86x-*-elf)
15601          + OpenBSD 32-bit (sparc-*-openbsd*)
15602
15603Documentation improvements
15604
15605Other significant improvements
15606
15607     * Location lists are now generated by default when compiling with
15608       debug info and optimization. Location lists provide more accurate
15609       debug info about locations of variables and they allow debugging
15610       code compiled with -fomit-frame-pointer.
15611     * The -fvisibility option has been added which allows the default ELF
15612       visibility of all symbols to be set per compilation and the new
15613       #pragma GCC visibility preprocessor command allows the setting of
15614       default ELF visibility for a region of code. Using
15615       -fvisibility=hidden especially in combination with the new
15616       -fvisibility-inlines-hidden can yield substantial improvements in
15617       output binary quality including avoiding PLT indirection overheads,
15618       reduction of the exported symbol count by up to 60% (with resultant
15619       improvements to link and load times), better scope for the
15620       optimizer to improve code and up to a 20% reduction in binary size.
15621       Using these options correctly yields a binary with a similar symbol
15622       count to a Windows DLL.
15623       Perhaps more importantly, this new feature finally allows (with
15624       careful planning) complete avoidance of symbol clashes when
15625       manually loading shared objects with RTLD_GLOBAL, thus finally
15626       solving problems many projects such as python were forced to use
15627       RTLD_LOCAL for (with its resulting issues for C++ correctness). You
15628       can find more information about using these options at
15629       [11]https://gcc.gnu.org/wiki/Visibility.
15630     __________________________________________________________________
15631
15632GCC 4.0.1
15633
15634   This is the [12]list of problem reports (PRs) from GCC's bug tracking
15635   system that are known to be fixed in the 4.0.1 release. This list might
15636   not be complete (that is, it is possible that some PRs that have been
15637   fixed are not listed here).
15638
15639GCC 4.0.2
15640
15641   This is the [13]list of problem reports (PRs) from GCC's bug tracking
15642   system that are known to be fixed in the 4.0.2 release. This list might
15643   not be complete (that is, it is possible that some PRs that have been
15644   fixed are not listed here).
15645
15646   Unfortunately, due to a release engineering failure, this release has a
15647   regression on Solaris that will affect some C++ programs. We suggest
15648   that Solaris users apply a [14]patch that corrects the problem. Users
15649   who do not wish to apply the patch should explicitly link C++ programs
15650   with the -pthreads option, even if they do not use threads. This
15651   problem has been corrected in the current 4.0 branch sources and will
15652   not be present in GCC 4.0.3.
15653
15654GCC 4.0.3
15655
15656   Starting with this release, the function getcontext is recognized by
15657   the compiler as having the same semantics as the setjmp function. In
15658   particular, the compiler will ensure that all registers are dead before
15659   calling such a function and will emit a warning about the variables
15660   that may be clobbered after the second return from the function.
15661
15662GCC 4.0.4
15663
15664   This is the [15]list of problem reports (PRs) from GCC's bug tracking
15665   system that are known to be fixed in the 4.0.4 release. This list might
15666   not be complete (that is, it is possible that some PRs that have been
15667   fixed are not listed here).
15668
15669   The 4.0.4 release is provided for those that require a high degree of
15670   binary compatibility with previous 4.0.x releases. For most users, the
15671   GCC team recommends that version 4.1.1 or later be used instead."
15672
15673
15674    For questions related to the use of GCC, please consult these web
15675    pages and the [16]GCC manuals. If that fails, the
15676    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
15677    web pages and the development of GCC are welcome on our developer
15678    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
15679    archives.
15680
15681   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
15682   distribution of this entire article is permitted in any medium,
15683   provided this notice is preserved.
15684
15685   These pages are [21]maintained by the GCC team. Last modified
15686   2022-10-26.
15687
15688References
15689
15690   1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
15691   2. https://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html
15692   3. http://gcc.gnu.org/projects/tree-ssa/
15693   4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html
15694   5. http://gcc.gnu.org/news/sms.html
15695   6. https://www.akkadia.org/drepper/dsohowto.pdf
15696   7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility
15697   8. https://itanium-cxx-abi.github.io/cxx-abi/
15698   9. http://gcc.gnu.org/fortran/
15699  10. https://gcc.gnu.org/install/
15700  11. https://gcc.gnu.org/wiki/Visibility
15701  12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1
15702  13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2
15703  14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html
15704  15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4
15705  16. https://gcc.gnu.org/onlinedocs/
15706  17. mailto:gcc-help@gcc.gnu.org
15707  18. mailto:gcc@gcc.gnu.org
15708  19. https://gcc.gnu.org/lists.html
15709  20. https://www.fsf.org/
15710  21. https://gcc.gnu.org/about.html
15711======================================================================
15712http://gcc.gnu.org/gcc-3.4/index.html
15713
15714                             GCC 3.4 Release Series
15715
15716   (This release series is no longer supported.)
15717
15718   May 26, 2006
15719
15720   The [1]GNU project and the GCC developers are pleased to announce the
15721   release of GCC 3.4.6.
15722
15723   This release is a bug-fix release, containing fixes for regressions in
15724   GCC 3.4.4 relative to previous releases of GCC. This is the last of the
15725   3.4.x series.
15726
15727   The GCC 3.4 release series includes numerous [2]new features,
15728   improvements, bug fixes, and other changes, thanks to an [3]amazing
15729   group of volunteers.
15730
15731Release History
15732
15733   GCC 3.4.6
15734          March 6, 2006 ([4]changes)
15735
15736   GCC 3.4.5
15737          November 30, 2005 ([5]changes)
15738
15739   GCC 3.4.4
15740          May 18, 2005 ([6]changes)
15741
15742   GCC 3.4.3
15743          November 4, 2004 ([7]changes)
15744
15745   GCC 3.4.2
15746          September 6, 2004 ([8]changes)
15747
15748   GCC 3.4.1
15749          July 1, 2004 ([9]changes)
15750
15751   GCC 3.4.0
15752          April 18, 2004 ([10]changes)
15753
15754References and Acknowledgements
15755
15756   GCC used to stand for the GNU C Compiler, but since the compiler
15757   supports several other languages aside from C, it now stands for the
15758   GNU Compiler Collection.
15759
15760   A list of [11]successful builds is updated as new information becomes
15761   available.
15762
15763   The GCC developers would like to thank the numerous people that have
15764   contributed new features, improvements, bug fixes, and other changes as
15765   well as test results to GCC. This [12]amazing group of volunteers is
15766   what makes GCC successful.
15767
15768   For additional information about GCC please refer to the [13]GCC
15769   project web site or contact the [14]GCC development mailing list.
15770
15771   To obtain GCC please use [15]our mirror sites, or [16]our version
15772   control system.
15773
15774
15775    For questions related to the use of GCC, please consult these web
15776    pages and the [17]GCC manuals. If that fails, the
15777    [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
15778    web pages and the development of GCC are welcome on our developer
15779    list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
15780    archives.
15781
15782   Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
15783   distribution of this entire article is permitted in any medium,
15784   provided this notice is preserved.
15785
15786   These pages are [22]maintained by the GCC team. Last modified
15787   2022-10-26.
15788
15789References
15790
15791   1. http://www.gnu.org/
15792   2. http://gcc.gnu.org/gcc-3.4/changes.html
15793   3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
15794   4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
15795   5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5
15796   6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4
15797   7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3
15798   8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2
15799   9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1
15800  10. http://gcc.gnu.org/gcc-3.4/changes.html
15801  11. http://gcc.gnu.org/gcc-3.4/buildstat.html
15802  12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
15803  13. http://gcc.gnu.org/index.html
15804  14. mailto:gcc@gcc.gnu.org
15805  15. http://gcc.gnu.org/mirrors.html
15806  16. http://gcc.gnu.org/git.html
15807  17. https://gcc.gnu.org/onlinedocs/
15808  18. mailto:gcc-help@gcc.gnu.org
15809  19. mailto:gcc@gcc.gnu.org
15810  20. https://gcc.gnu.org/lists.html
15811  21. https://www.fsf.org/
15812  22. https://gcc.gnu.org/about.html
15813======================================================================
15814http://gcc.gnu.org/gcc-3.4/changes.html
15815
15816                             GCC 3.4 Release Series
15817                        Changes, New Features, and Fixes
15818
15819   The final release in the 3.4 release series is [1]GCC 3.4.6. The series
15820   is now closed.
15821
15822   GCC 3.4 has [2]many improvements in the C++ front end. Before reporting
15823   a bug, please make sure it's really GCC, and not your code, that is
15824   broken.
15825
15826Caveats
15827
15828     * GNU Make is now required to build GCC.
15829     * With -nostdinc the preprocessor used to ignore both standard
15830       include paths and include paths contained in environment variables.
15831       It was neither documented nor intended that environment variable
15832       paths be ignored, so this has been corrected.
15833     * GCC no longer accepts the options -fvolatile, -fvolatile-global and
15834       -fvolatile-static. It is unlikely that they worked correctly in any
15835       3.x release.
15836     * GCC no longer ships <varargs.h>. Use <stdarg.h> instead.
15837     * Support for all the systems [3]obsoleted in GCC 3.3 has been
15838       removed from GCC 3.4. See below for a [4]list of systems which are
15839       obsoleted in this release.
15840     * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C
15841       compilers will not work.
15842     * The implementation of the [5]MIPS ABIs has changed. As a result,
15843       the code generated for certain MIPS targets will not be binary
15844       compatible with earlier releases.
15845     * In previous releases, the MIPS port had a fake "hilo" register with
15846       the user-visible name accum. This register has been removed.
15847     * The implementation of the [6]SPARC ABIs has changed. As a result,
15848       the code generated will not be binary compatible with earlier
15849       releases in certain cases.
15850     * The configure option --enable-threads=pthreads has been removed;
15851       use --enable-threads=posix instead, which should have the same
15852       effect.
15853     * Code size estimates used by inlining heuristics for C, Objective-C,
15854       C++ and Java have been redesigned significantly. As a result the
15855       parameters of -finline-insns, --param max-inline-insns-single and
15856       --param max-inline-insns-auto need to be reconsidered.
15857     * --param max-inline-slope and --param min-inline-insns have been
15858       removed; they are not needed for the new bottom-up inlining
15859       heuristics.
15860     * The new unit-at-a-time compilation scheme has several compatibility
15861       issues:
15862          + The order in which functions, variables, and top-level asm
15863            statements are emitted may have changed. Code relying on some
15864            particular ordering needs to be updated. The majority of such
15865            top-level asm statements can be replaced by section
15866            attributes.
15867          + Unreferenced static variables and functions are removed. This
15868            may result in undefined references when an asm statement
15869            refers to the variable/function directly. In that case either
15870            the variable/function shall be listed in asm statement operand
15871            or in the case of top-level asm statements the attribute used
15872            shall be used to force function/variable to be always output
15873            and considered as a possibly used by unknown code.
15874            For variables the attribute is accepted only by GCC 3.4 and
15875            newer, while for earlier versions it is sufficient to use
15876            unused to silence warnings about the variables not being
15877            referenced. To keep code portable across different GCC
15878            versions, you can use appropriate preprocessor conditionals.
15879          + Static functions now can use non-standard passing conventions
15880            that may break asm statements calling functions directly.
15881            Again the attribute used shall be used to prevent this
15882            behavior.
15883       As a temporary workaround, -fno-unit-at-a-time can be used, but
15884       this scheme may not be supported by future releases of GCC.
15885     * GCC 3.4 automatically places zero-initialized variables in the .bss
15886       section on some operating systems. Versions of GNU Emacs up to (and
15887       including) 21.3 will not work correctly when using this
15888       optimization; you can use -fno-zero-initialized-in-bss to disable
15889       it.
15890     * If GCC 3.4 is configured with --enable-threads=posix (the default
15891       on most targets that support pthreads) then _REENTRANT will be
15892       defined unconditionally by some libstdc++ headers. C++ code which
15893       relies on that macro to detect whether multi-threaded code is being
15894       compiled might change in meaning, possibly resulting in linker
15895       errors for single-threaded programs. Affected users of [7]Boost
15896       should compile single-threaded code with -DBOOST_DISABLE_THREADS.
15897       See Bugzilla for [8]more information.
15898
15899General Optimizer Improvements
15900
15901     * Usability of the profile feedback and coverage testing has been
15902       improved.
15903          + Performance of profiled programs has been improved by faster
15904            profile merging code.
15905          + Better use of the profile feedback for optimization (loop
15906            unrolling and loop peeling).
15907          + File locking support allowing fork() calls and parallel runs
15908            of profiled programs.
15909          + Coverage file format has been redesigned.
15910          + gcov coverage tool has been improved.
15911          + make profiledbootstrap available to build a faster compiler.
15912            Experiments made on i386 hardware showed an 11% speedup on -O0
15913            and a 7.5% speedup on -O2 compilation of a [9]large C++
15914            testcase.
15915          + New value profiling pass enabled via -fprofile-values
15916          + New value profile transformations pass enabled via -fvpt aims
15917            to optimize some code sequences by exploiting knowledge about
15918            value ranges or other properties of the operands. At the
15919            moment a conversion of expensive divisions into cheaper
15920            operations has been implemented.
15921          + New -fprofile-generate and -fprofile-use command-line options
15922            to simplify the use of profile feedback.
15923     * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and
15924       Java which is enabled via -funit-at-a-time (and implied by -O2). In
15925       this scheme a whole file is parsed first and optimized later. The
15926       following basic inter-procedural optimizations are implemented:
15927          + Removal of unreachable functions and variables
15928          + Discovery of local functions (functions with static linkage
15929            whose address is never taken)
15930          + On i386, these local functions use register parameter passing
15931            conventions.
15932          + Reordering of functions in topological order of the call graph
15933            to enable better propagation of optimizing hints (such as the
15934            stack alignments needed by functions) in the back end.
15935          + Call graph based out-of-order inlining heuristics which allows
15936            to limit overall compilation unit growth (--param
15937            inline-unit-growth).
15938       Overall, the unit-at-a-time scheme produces a 1.3% improvement for
15939       the SPECint2000 benchmark on the i386 architecture (AMD Athlon
15940       CPU).
15941     * More realistic code size estimates used by inlining for C,
15942       Objective-C, C++ and Java. The growth of large functions can now be
15943       limited via --param large-function-insns and --param
15944       large-function-growth.
15945     * A new cfg-level loop optimizer pass replaces the old loop unrolling
15946       pass and adds two other loop transformations -- loop peeling and
15947       loop unswitching -- and also uses the profile feedback to limit
15948       code growth. (The three optimizations are enabled by
15949       -funroll-loops, -fpeel-loops and -funswitch-loops flags,
15950       respectively).
15951       The old loop unroller still can be enabled by -fold-unroll-loops
15952       and may produce better code in some cases, especially when the
15953       webizer optimization pass is not run.
15954     * A new web construction pass enabled via -fweb (and implied by -O3)
15955       improves the quality of register allocation, CSE, first scheduling
15956       pass and some other optimization passes by avoiding re-use of
15957       pseudo registers with non-overlapping live ranges. The pass almost
15958       always improves code quality but does make debugging difficult and
15959       thus is not enabled by default by -O2
15960       The pass is especially effective as cleanup after code duplication
15961       passes, such as the loop unroller or the tracer.
15962     * Experimental implementations of superblock or trace scheduling in
15963       the second scheduling pass can be enabled via
15964       -fsched2-use-superblocks and -fsched2-use-traces, respectively.
15965
15966New Languages and Language specific improvements
15967
15968  Ada
15969
15970     * The Ada front end has been updated to include numerous bug fixes
15971       and enhancements. These include:
15972          + Improved project file support
15973          + Additional set of warnings about potential wrong code
15974          + Improved error messages
15975          + Improved code generation
15976          + Improved cross reference information
15977          + Improved inlining
15978          + Better run-time check elimination
15979          + Better error recovery
15980          + More efficient implementation of unbounded strings
15981          + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools,
15982            ...
15983          + New GNAT.xxxx packages (e.g. GNAT.Strings,
15984            GNAT.Exception_Action)
15985          + New pragmas
15986          + New -gnatS switch replacing gnatpsta
15987          + Implementation of new Ada features (in particular limited
15988            with, limited aggregates)
15989
15990  C/Objective-C/C++
15991
15992     * Precompiled headers are now supported. Precompiled headers can
15993       dramatically speed up compilation of some projects. There are some
15994       known defects in the current precompiled header implementation that
15995       will result in compiler crashes in relatively rare situations.
15996       Therefore, precompiled headers should be considered a "technology
15997       preview" in this release. Read the manual for details about how to
15998       use precompiled headers.
15999     * File handling in the preprocessor has been rewritten. GCC no longer
16000       gets confused by symlinks and hardlinks, and now has a correct
16001       implementation of #import and #pragma once. These two directives
16002       have therefore been un-deprecated.
16003     * The undocumented extension that allowed C programs to have a label
16004       at the end of a compound statement, which has been deprecated since
16005       GCC 3.0, has been removed.
16006     * The cast-as-lvalue extension has been removed for C++ and
16007       deprecated for C and Objective-C. In particular, code like this:
16008        int i;
16009        (char) i = 5;
16010
16011       or this:
16012        char *p;
16013        ((int *) p)++;
16014
16015       is no longer accepted for C++ and will not be accepted for C and
16016       Objective-C in a future version.
16017     * The conditional-expression-as-lvalue extension has been deprecated
16018       for C and Objective-C. In particular, code like this:
16019        int a, b, c;
16020        (a ? b : c) = 2;
16021
16022       will not be accepted for C and Objective-C in a future version.
16023     * The compound-expression-as-lvalue extension has been deprecated for
16024       C and Objective-C. In particular, code like this:
16025        int a, b;
16026        (a, b) = 2;
16027
16028       will not be accepted for C and Objective-C in a future version. A
16029       possible non-intrusive workaround is the following:
16030        (*(a, &b)) = 2;
16031
16032     * Several [10]built-in functions such as __builtin_popcount for
16033       counting bits, finding the highest and lowest bit in a word, and
16034       parity have been added.
16035     * The -fwritable-strings option has been deprecated and will be
16036       removed.
16037     * Many C math library functions are now recognized as built-ins and
16038       optimized.
16039     * The C, C++, and Objective-C compilers can now handle source files
16040       written in any character encoding supported by the host C library.
16041       The default input character set is taken from the current locale,
16042       and may be overridden with the -finput-charset command line option.
16043       In the future we will add support for inline encoding markers.
16044
16045  C++
16046
16047     * G++ is now much closer to full conformance to the ISO/ANSI C++
16048       standard. This means, among other things, that a lot of invalid
16049       constructs which used to be accepted in previous versions will now
16050       be rejected. It is very likely that existing C++ code will need to
16051       be fixed. This document lists some of the most common issues.
16052     * A hand-written recursive-descent C++ parser has replaced the
16053       YACC-derived C++ parser from previous GCC releases. The new parser
16054       contains much improved infrastructure needed for better parsing of
16055       C++ source codes, handling of extensions, and clean separation
16056       (where possible) between proper semantics analysis and parsing. The
16057       new parser fixes many bugs that were found in the old parser.
16058     * You must now use the typename and template keywords to disambiguate
16059       dependent names, as required by the C++ standard.
16060        struct K {
16061          typedef int mytype_t;
16062        };
16063
16064        template <class T1> struct A {
16065          template <class T2> struct B {
16066              void callme(void);
16067            };
16068
16069          template <int N> void bar(void)
16070          {
16071            // Use 'typename' to tell the parser that T1::mytype_t names
16072            //  a type. This is needed because the name is dependent (in
16073            //  this case, on template parameter T1).
16074            typename T1::mytype_t x;
16075            x = 0;
16076          }
16077        };
16078
16079        template <class T> void template_func(void)
16080        {
16081          // Use 'template' to prefix member templates within
16082          //  dependent types (a has type A<T>, which depends on
16083          //  the template parameter T).
16084          A<T> a;
16085          a.template bar<0>();
16086
16087          // Use 'template' to tell the parser that B is a nested
16088          //  template class (dependent on template parameter T), and
16089          //  'typename' because the whole A<T>::B<int> is
16090          //  the name of a type (again, dependent).
16091          typename A<T>::template B<int> b;
16092          b.callme();
16093        }
16094
16095        void non_template_func(void)
16096        {
16097          // Outside of any template class or function, no names can be
16098          //  dependent, so the use of the keyword 'typename' and 'template'
16099          //  is not needed (and actually forbidden).
16100          A<K> a;
16101          a.bar<0>();
16102          A<K>::B<float> b;
16103          b.callme();
16104        }
16105     * In a template definition, unqualified names will no longer find
16106       members of a dependent base (as specified by [temp.dep]/3 in the
16107       C++ standard). For example,
16108        template <typename T> struct B {
16109          int m;
16110          int n;
16111          int f ();
16112          int g ();
16113        };
16114        int n;
16115        int g ();
16116        template <typename T> struct C : B<T> {
16117          void h ()
16118          {
16119            m = 0; // error
16120            f ();  // error
16121            n = 0; // ::n is modified
16122            g ();  // ::g is called
16123          }
16124        };
16125       You must make the names dependent, e.g. by prefixing them with
16126       this->. Here is the corrected definition of C<T>::h,
16127        template <typename T> void C<T>::h ()
16128        {
16129          this->m = 0;
16130          this->f ();
16131          this->n = 0
16132          this->g ();
16133        }
16134       As an alternative solution (unfortunately not backwards compatible
16135       with GCC 3.3), you may use using declarations instead of this->:
16136        template <typename T> struct C : B<T> {
16137          using B<T>::m;
16138          using B<T>::f;
16139          using B<T>::n;
16140          using B<T>::g;
16141          void h ()
16142          {
16143            m = 0;
16144            f ();
16145            n = 0;
16146            g ();
16147          }
16148        };
16149     * In templates, all non-dependent names are now looked up and bound
16150       at definition time (while parsing the code), instead of later when
16151       the template is instantiated. For instance:
16152        void foo(int);
16153
16154        template <int> struct A {
16155          static void bar(void){
16156            foo('a');
16157          }
16158        };
16159
16160        void foo(char);
16161
16162        int main()
16163        {
16164          A<0>::bar();    // Calls foo(int), used to call foo(char).
16165        }
16166
16167     * In an explicit instantiation of a class template, you must use
16168       class or struct before the template-id:
16169        template <int N>
16170        class A {};
16171
16172        template A<0>;         // error, not accepted anymore
16173        template class A<0>;   // OK
16174     * The "named return value" and "implicit typename" extensions have
16175       been removed.
16176     * Default arguments in function types have been deprecated and will
16177       be removed.
16178     * ARM-style name-injection of friend declarations has been deprecated
16179       and will be removed. For example: struct S { friend void f(); };
16180       void g() { f(); } will not be accepted by future versions of G++;
16181       instead a declaration of "f" will need to be present outside of the
16182       scope of "S".
16183     * Covariant returns are implemented for all but varadic functions
16184       that require an adjustment.
16185     * When -pedantic is used, G++ now issues errors about spurious
16186       semicolons. For example,
16187        namespace N {}; // Invalid semicolon.
16188        void f() {}; // Invalid semicolon.
16189     * G++ no longer accepts attributes for a declarator after the
16190       initializer associated with that declarator. For example,
16191        X x(1) __attribute__((...));
16192       is no longer accepted. Instead, use:
16193        X x __attribute__((...)) (1);
16194     * Inside the scope of a template class, the name of the class itself
16195       can be treated as either a class or a template. So GCC used to
16196       accept the class name as argument of type template, and template
16197       template parameter. However this is not C++ standard compliant. Now
16198       the name is not treated as a valid template template argument
16199       unless you qualify the name by its scope. For example, the code
16200       below no longer compiles.
16201        template <template <class> class TT> class X {};
16202        template <class T> class Y {
16203          X<Y> x; // Invalid, Y is always a type template parameter.
16204        };
16205       The valid code for the above example is
16206          X< ::Y> x; // Valid.
16207       (Notice the space between < and : to prevent GCC to interpret this
16208       as a digraph for [.)
16209     * Friend declarations that refer to template specializations are
16210       rejected if the template has not already been declared. For
16211       example,
16212        template <typename T>
16213        class C {
16214          friend void f<> (C&);
16215        };
16216       is rejected. You must first declare f as a template,
16217        template <typename T>
16218        void f(T);
16219     * In case of friend declarations, every name used in the friend
16220       declaration must be accessible at the point of that declaration.
16221       Previous versions of G++ used to be less strict about this and
16222       allowed friend declarations for private class members, for example.
16223       See the ISO C++ Standard Committee's [11]defect report #209 for
16224       details.
16225     * Declaration of member functions of class templates as friends are
16226       supported. For example,
16227        template <typename T> struct A {
16228          void f();
16229        };
16230        class C {
16231          template <typename T> friend void A<T>::f();
16232        };
16233     * You must use template <> to introduce template specializations, as
16234       required by the standard. For example,
16235        template <typename T>
16236        struct S;
16237
16238        struct S<int> { };
16239       is rejected. You must write,
16240        template <> struct S<int> {};
16241     * G++ used to accept code like this,
16242        struct S {
16243          int h();
16244          void f(int i = g());
16245          int g(int i = h());
16246        };
16247       This behavior is not mandated by the standard. Now G++ issues an
16248       error about this code. To avoid the error, you must move the
16249       declaration of g before the declaration of f. The default arguments
16250       for g must be visible at the point where it is called.
16251     * The C++ ABI Section 3.3.3 specifications for the array construction
16252       routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return
16253       NULL when the allocator argument returns NULL. These changes are
16254       incorporated into the libstdc++ runtime library.
16255     * Using a name introduced by a typedef in a friend declaration or in
16256       an explicit instantiation is now rejected, as specified by the ISO
16257       C++ standard.
16258        class A;
16259        typedef A B;
16260        class C {
16261          friend class B;      // error, no typedef name here
16262          friend B;            // error, friend always needs class/struct/enum
16263          friend class A;      // OK
16264        };
16265
16266        template <int> class Q {};
16267        typedef Q<0> R;
16268        template class R;      // error, no typedef name here
16269        template class Q<0>;   // OK
16270     * When allocating an array with a new expression, GCC used to allow
16271       parentheses around the type name. This is actually ill-formed and
16272       it is now rejected:
16273        int* a = new (int)[10];    // error, not accepted anymore
16274        int* a = new int[10];      // OK
16275     * When binding an rvalue of class type to a reference, the copy
16276       constructor of the class must be accessible. For instance, consider
16277       the following code:
16278        class A
16279        {
16280        public:
16281          A();
16282
16283        private:
16284          A(const A&);   // private copy ctor
16285        };
16286
16287        A makeA(void);
16288        void foo(const A&);
16289
16290        void bar(void)
16291        {
16292          foo(A());       // error, copy ctor is not accessible
16293          foo(makeA());   // error, copy ctor is not accessible
16294
16295          A a1;
16296          foo(a1);        // OK, a1 is a lvalue
16297        }
16298       This might be surprising at first sight, especially since most
16299       popular compilers do not correctly implement this rule ([12]further
16300       details).
16301     * When forming a pointer to member or a pointer to member function,
16302       access checks for class visibility (public, protected, private) are
16303       now performed using the qualifying scope of the name itself. This
16304       is better explained with an example:
16305        class A
16306        {
16307        public:
16308          void pub_func();
16309        protected:
16310          void prot_func();
16311        private:
16312          void priv_func();
16313        };
16314
16315        class B : public A
16316        {
16317        public:
16318          void foo()
16319          {
16320            &A::pub_func;   // OK, pub_func is accessible through A
16321            &A::prot_func;  // error, cannot access prot_func through A
16322            &A::priv_func;  // error, cannot access priv_func through A
16323
16324            &B::pub_func;   // OK, pub_func is accessible through B
16325            &B::prot_func;  // OK, can access prot_func through B (within B)
16326            &B::priv_func;  // error, cannot access priv_func through B
16327          }
16328        };
16329
16330    Runtime Library (libstdc++)
16331
16332     * Optimization work:
16333          + Streamlined streambuf, filebuf, separate synched with C
16334            Standard I/O streambuf.
16335          + All formatted I/O now uses cached locale information.
16336          + STL optimizations (memory/speed for list, red-black trees as
16337            used by sets and maps).
16338          + More use of GCC builtins.
16339          + String optimizations (avoid contention on
16340            increment/decrement-and-test of the reference count in the
16341            empty-string object, constructor from input_iterators
16342            speedup).
16343     * Static linkage size reductions.
16344     * Large File Support (files larger than 2 GB on 32-bit systems).
16345     * Wide character and variable encoding filebuf work (UTF-8, Unicode).
16346     * Generic character traits.
16347     * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x,
16348       Solaris 2.7 and above, AIX 5.x, Irix 6.5.
16349     * The allocator class is now standard-conformant, and two additional
16350       extension allocators have been added, mt_alloc and
16351       bitmap_allocator.
16352     * PCH support: -include bits/stdc++.h (2x compile speedup).
16353     * Rewrote __cxa_demangle with support for C++ style allocators.
16354     * New debug modes for STL containers and iterators.
16355     * Testsuite rewrite: five times as many tests, plus increasingly
16356       sophisticated tests, including I/O, MT, multi-locale, wide and
16357       narrow characters.
16358     * Use current versions of GNU "autotools" for build/configuration.
16359
16360  Objective-C
16361
16362     * The Objective-C front end has been updated to include the numerous
16363       bug fixes and enhancements previously available only in Apple's
16364       version of GCC. These include:
16365          + Structured exception (@try... @catch... @finally, @throw) and
16366            synchronization (@synchronized) support. These are accessible
16367            via the -fobjc-exceptions switch; as of this writing, they may
16368            only be used in conjunction with -fnext-runtime on Mac OS X
16369            10.3 and later. See [13]Options Controlling Objective-C
16370            Dialect for more information.
16371          + An overhaul of @encode logic. The C99 _Bool and C++ bool type
16372            may now be encoded as 'B'. In addition, the back-end/codegen
16373            dependencies have been removed.
16374          + An overhaul of message dispatch construction, ensuring that
16375            the various receiver types (and casts thereof) are handled
16376            properly, and that correct diagnostics are issued.
16377          + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue"
16378            (-freplace-objc-classes) debugging modes, currently available
16379            on Mac OS X 10.3 and later. See [14]Options Controlling
16380            Objective-C Dialect for more information.
16381          + Access to optimized runtime entry points (-fno-nil-receivers )
16382            on the assumption that message receivers are never nil. This
16383            is currently available on Mac OS X 10.3 and later. See
16384            [15]Options Controlling Objective-C Dialect for more
16385            information.
16386
16387  Java
16388
16389     * Compiling a .jar file will now cause non-.class entries to be
16390       automatically compiled as resources.
16391     * libgcj has been ported to Darwin.
16392     * Jeff Sturm has adapted Jan Hubicka's call graph optimization code
16393       to gcj.
16394     * libgcj has a new gcjlib URL type; this lets URLClassLoader load
16395       code from shared libraries.
16396     * libgcj has been much more completely merged with [16]GNU Classpath.
16397     * Class loading is now much more correct; in particular the caller's
16398       class loader is now used when that is required.
16399     * [17]Eclipse 2.x will run out of the box using gij.
16400     * Parts of java.nio have been implemented. Direct and indirect
16401       buffers work, as do fundamental file and socket operations.
16402     * java.awt has been improved, though it is still not ready for
16403       general use.
16404     * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST
16405       method.
16406     * The MinGW port has matured. Enhancements include socket timeout
16407       support, thread interruption, improved Runtime.exec() handling and
16408       support for accented characters in filenames.
16409
16410  Fortran
16411
16412     * Fortran improvements are listed in the [18]Fortran documentation.
16413
16414New Targets and Target Specific Improvements
16415
16416  Alpha
16417
16418     * Several [19]built-in functions have been added such as
16419       __builtin_alpha_zap to allow utilizing the more obscure
16420       instructions of the CPU.
16421     * Parameter passing of complex arguments has changed to match the
16422       ABI. This change is incompatible with previous GCC versions, but
16423       does fix compatibility with the Tru64 compiler and several corner
16424       cases where GCC was incompatible with itself.
16425
16426  ARM
16427
16428     * Nicolas Pitre has contributed his hand-coded floating-point support
16429       code for ARM. It is both significantly smaller and faster than the
16430       existing C-based implementation, even when building applications
16431       for Thumb. The arm-elf configuration has been converted to use the
16432       new code.
16433     * Support for the Intel's iWMMXt architecture, a second generation
16434       XScale processor, has been added. Enabled at run time with the
16435       -mcpu=iwmmxt command line switch.
16436     * A new ARM target has been added: arm-wince-pe. This is similar to
16437       the arm-pe target, but it defaults to using the APCS32 ABI.
16438     * The existing ARM pipeline description has been converted to the use
16439       the [20]DFA processor pipeline model. There is not much change in
16440       code performance, but the description is now [21]easier to
16441       understand.
16442     * Support for the Cirrus EP9312 Maverick floating point co-processor
16443       added. Enabled at run time with the -mcpu=ep9312 command line
16444       switch. Note however that the multilibs to support this chip are
16445       currently disabled in gcc/config/arm/t-arm-elf, so if you want to
16446       enable their production you will have to uncomment the entries in
16447       that file.
16448
16449  H8/300
16450
16451     * Support for long long has been added.
16452     * Support for saveall attribute has been added.
16453     * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code
16454       for H8/300H and H8S, which is much faster than the previous
16455       implementation.
16456     * A lot of small performance improvements.
16457
16458  IA-32/AMD64 (x86-64)
16459
16460     * Tuning for K8 (AMD Opteron/Athlon64) core is available via
16461       -march=k8 and -mcpu=k8.
16462     * Scalar SSE code generation carefully avoids reformatting penalties,
16463       hidden dependencies and minimizes the number of uops generated on
16464       both Intel and AMD CPUs.
16465     * Vector MMX and SSE operands are now passed in registers to improve
16466       performance and match the argument passing convention used by the
16467       Intel C++ Compiler. As a result it is not possible to call
16468       functions accepting vector arguments compiled by older GCC version.
16469     * Conditional jump elimination is now more aggressive on modern CPUs.
16470     * The Athlon ports has been converted to use the DFA processor
16471       pipeline description.
16472     * Optimization of indirect tail calls is now possible in a similar
16473       fashion as direct sibcall optimization.
16474     * Further small performance improvements.
16475     * -m128bit-long-double is now less buggy.
16476     * __float128 support in 64-bit compilation.
16477     * Support for data structures exceeding 2GB in 64-bit mode.
16478     * -mcpu has been renamed to -mtune.
16479
16480  IA-64
16481
16482     * Tuning code for the Itanium 2 processor has been added. The
16483       generation of code tuned for Itanium 2 (option -mtune=itanium2) is
16484       enabled by default now. To generate code tuned for Itanium 1 the
16485       option -mtune=itanium1 should be used.
16486     * [22]DFA processor pipeline descriptions for the IA-64 processors
16487       have been added. This resulted in about 3% improvement on the
16488       SPECInt2000 benchmark for Itanium 2.
16489     * Instruction bundling for the IA-64 processors has been rewritten
16490       using the DFA pipeline hazard recognizer. It resulted in about 60%
16491       compiler speedup on the SPECInt2000 C programs.
16492
16493  M32R
16494
16495     * Support for the M32R/2 processor has been added by Renesas.
16496     * Support for an M32R GNU/Linux target and PIC code generation has
16497       been added by Renesas.
16498
16499  M68000
16500
16501     * Bernardo Innocenti (Develer S.r.l.) has contributed the
16502       m68k-uclinux target, based on former work done by Paul Dale
16503       (SnapGear Inc.). Code generation for the ColdFire processors family
16504       has been enhanced and extended to support the MCF 53xx and MCF 54xx
16505       cores, integrating former work done by Peter Barada (Motorola).
16506
16507  MIPS
16508
16509    Processor-specific changes
16510
16511     * Support for the RM7000 and RM9000 processors has been added. It can
16512       be selected using the -march compiler option and should work with
16513       any MIPS I (mips-*) or MIPS III (mips64-*) configuration.
16514     * Support for revision 2 of the MIPS32 ISA has been added. It can be
16515       selected with the command-line option -march=mips32r2.
16516     * There is a new option, -mfix-sb1, to work around certain SB-1
16517       errata.
16518
16519    Configuration
16520
16521     * It is possible to customize GCC using the following configure-time
16522       options:
16523          + --with-arch, which specifies the default value of the -march
16524            option.
16525          + --with-tune, which specifies the default value of the -mtune
16526            option.
16527          + --with-abi, which specifies the default ABI.
16528          + --with-float=soft, which tells GCC to use software floating
16529            point by default.
16530          + --with-float=hard, which tells GCC to use hardware floating
16531            point by default.
16532     * A 64-bit GNU/Linux port has been added. The associated
16533       configurations are mips64-linux-gnu and mips64el-linux-gnu.
16534     * The 32-bit GNU/Linux port now supports Java.
16535     * The IRIX 6 configuration now supports the o32 ABI and will build
16536       o32 multilibs by default. This support is compatible with both
16537       binutils and the SGI tools, but note that several features,
16538       including debugging information and DWARF2 exception handling, are
16539       only available when using the GNU assembler. Use of the GNU
16540       assembler and linker (version 2.15 or above) is strongly
16541       recommended.
16542     * The IRIX 6 configuration now supports 128-bit long doubles.
16543     * There are two new RTEMS-specific configurations, mips-rtems and
16544       mipsel-rtems.
16545     * There are two new *-elf configurations, mipsisa32r2-elf and
16546       mipsisa32r2el-elf.
16547
16548    General
16549
16550     * Several [23]ABI bugs have been fixed. Unfortunately, these changes
16551       will break binary compatibility with earlier releases.
16552     * GCC can now use explicit relocation operators when generating
16553       -mabicalls code. This behavior is controlled by -mexplicit-relocs
16554       and can have several performance benefits. For example:
16555          + It allows for more optimization of GOT accesses, including
16556            better scheduling and redundancy elimination.
16557          + It allows sibling calls to be implemented as jumps.
16558          + n32 and n64 leaf functions can use a call-clobbered global
16559            pointer instead of $28.
16560          + The code to set up $gp can be removed from functions that
16561            don't need it.
16562     * A new option, -mxgot, allows the GOT to be bigger than 64k. This
16563       option is equivalent to the assembler's -xgot option and should be
16564       used instead of -Wa,-xgot.
16565     * Frame pointer elimination is now supported when generating 64-bit
16566       MIPS16 code.
16567     * Inline block moves have been optimized to take more account of
16568       alignment information.
16569     * Many internal changes have been made to the MIPS port, mostly aimed
16570       at reducing the reliance on assembler macros.
16571
16572  PowerPC
16573
16574     * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64
16575       [24]ABI incompatibilities regarding the way parameters are passed
16576       during functions calls. These changes may result in incompatibility
16577       between code compiled with GCC 3.3 and GCC 3.4.
16578
16579    PowerPC Darwin
16580
16581     * Support for shared/dylib gcc libraries has been added. It is
16582       enabled by default on powerpc-apple-darwin7.0.0 and up.
16583     * Libgcj is enabled by default. On systems older than
16584       powerpc-apple-darwin7.0.0 you need to install dlcompat.
16585     * 128-bit IBM extended precision format support added for long
16586       double.
16587
16588    PowerPC64 GNU/Linux
16589
16590     * By default, PowerPC64 GNU/Linux now uses natural alignment of
16591       structure elements. The old four byte alignment for double, with
16592       special rules for a struct starting with a double, can be chosen
16593       with -malign-power. This change may result in incompatibility
16594       between code compiled with GCC 3.3 and GCC 3.4.
16595     * -mabi=altivec is now the default rather than -mabi=no-altivec.
16596     * 128-bit IBM extended precision format support added for long
16597       double.
16598
16599  S/390 and zSeries
16600
16601     * New command-line options allow to specify the intended execution
16602       environment for generated code:
16603          + -mesa/-mzarch allows to specify whether to generate code
16604            running in ESA/390 mode or in z/Architecture mode (this is
16605            applicable to 31-bit code only).
16606          + -march allows to specify a minimum processor architecture
16607            level (g5, g6, z900, or z990).
16608          + -mtune allows to specify which processor to tune for.
16609     * It is possible to customize GCC using the following configure-time
16610       options:
16611          + --with-mode, which specifies whether to default to assuming
16612            ESA/390 or z/Architecture mode.
16613          + --with-arch, which specifies the default value of the -march
16614            option.
16615          + --with-tune, which specifies the default value of the -mtune
16616            option.
16617     * Support for the z990 processor has been added, and can be selected
16618       using -march=z990 or -mtune=z990. This includes instruction
16619       scheduling tuned for the superscalar instruction pipeline of the
16620       z990 processor as well as support for all new instructions provided
16621       by the long-displacement facility.
16622     * Support to generate 31-bit code optimized for zSeries processors
16623       (running in ESA/390 or in z/Architecture mode) has been added. This
16624       can be selected using -march=z900 and -mzarch respectively.
16625     * Instruction scheduling for the z900 and z990 processors now uses
16626       the DFA pipeline hazard recognizer.
16627     * GCC no longer generates code to maintain a stack backchain,
16628       previously used to generate stack backtraces for debugging
16629       purposes. As replacement that does not incur runtime overhead,
16630       DWARF-2 call frame information is provided by GCC; this is
16631       supported by GDB 6.1. The old behavior can be restored using the
16632       -mbackchain option.
16633     * The stack frame size of functions may now exceed 2 GB in 64-bit
16634       code.
16635     * A port for the 64-bit IBM TPF operating system has been added; the
16636       configuration is s390x-ibm-tpf. This configuration is supported as
16637       cross-compilation target only.
16638     * Various changes to improve the generated code have been
16639       implemented, including:
16640          + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT
16641            instructions to significantly speed up many floating-point
16642            applications.
16643          + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL
16644            WITH BORROW instructions to speed up long long arithmetic.
16645          + GCC now uses the SEARCH STRING instruction to implement
16646            strlen().
16647          + In many cases, function call overhead for 31-bit code has been
16648            reduced by placing the literal pool after the function code
16649            instead of after the function prolog.
16650          + Register 14 is no longer reserved in 64-bit code.
16651          + Handling of global register variables has been improved.
16652
16653  SPARC
16654
16655     * The option -mflat is deprecated.
16656     * Support for large (> 2GB) frames has been added to the 64-bit port.
16657     * Several [25]ABI bugs have been fixed. Unfortunately, these changes
16658       will break binary compatibility with earlier releases.
16659     * The default debugging format has been switched from STABS to
16660       DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already
16661       the default debugging format for 64-bit code on Solaris.
16662
16663  SuperH
16664
16665     * Support for the SH2E processor has been added. Enabled at run time
16666       with the -m2e command line switch, or at configure time by
16667       specifying sh2e as the machine part of the target triple.
16668
16669  V850
16670
16671     * Support for the Mitsubishi V850E1 processor has been added. This is
16672       a variant of the V850E processor with some additional debugging
16673       instructions.
16674
16675  Xtensa
16676
16677     * Several ABI bugs have been fixed. Unfortunately, these changes
16678       break binary compatibility with earlier releases.
16679          + For big-endian processors, the padding of aggregate return
16680            values larger than a word has changed. If the size of an
16681            aggregate return value is not a multiple of 32 bits, previous
16682            versions of GCC inserted padding in the most-significant bytes
16683            of the first return value register. Aggregates larger than a
16684            word are now padded in the least-significant bytes of the last
16685            return value register used. Aggregates smaller than a word are
16686            still padded in the most-significant bytes. The return value
16687            padding has not changed for little-endian processors.
16688          + Function arguments with 16-byte alignment are now properly
16689            aligned.
16690          + The implementation of the va_list type has changed. A va_list
16691            value created by va_start from a previous release cannot be
16692            used with va_arg from this release, or vice versa.
16693     * More processor configuration options for Xtensa processors are
16694       supported:
16695          + the ABS instruction is now optional;
16696          + the ADDX* and SUBX* instructions are now optional;
16697          + an experimental CONST16 instruction can be used to synthesize
16698            constants instead of loading them from constant pools.
16699       These and other Xtensa processor configuration options can no
16700       longer be enabled or disabled by command-line options; the
16701       processor configuration must be specified by the xtensa-config.h
16702       header file when building GCC. Additionally, the
16703       -mno-serialize-volatile option is no longer supported.
16704
16705Obsolete Systems
16706
16707   Support for a number of older systems has been declared obsolete in GCC
16708   3.4. Unless there is activity to revive them, the next release of GCC
16709   will have their sources permanently removed.
16710
16711   All configurations of the following processor architectures have been
16712   declared obsolete:
16713     * Mitsubishi D30V, d30v-*
16714     * AT&T DSP1600 and DSP1610, dsp16xx-*
16715     * Intel 80960, i960
16716
16717   Also, some individual systems have been obsoleted:
16718     * ARM Family
16719          + Support for generating code for operation in APCS/26 mode
16720            (-mapcs-26).
16721     * IBM ESA/390
16722          + "Bigfoot" port, i370-*. (The other port, s390-*, is actively
16723            maintained and supported.)
16724     * Intel 386 family
16725          + MOSS, i?86-moss-msdos and i?86-*-moss*
16726          + NCR 3000 running System V r.4, i?86-ncr-sysv4*
16727          + FreeBSD with a.out object format, i?86-*-freebsd*aout* and
16728            i?86-*-freebsd2*
16729          + GNU/Linux with a.out object format, i?86-linux*aout*
16730          + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1*
16731          + Interix versions before Interix 3, i?86-*-interix
16732          + Mach microkernel, i?86-mach*
16733          + SCO UnixWare with UDK, i?86-*-udk*
16734          + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]*
16735          + VSTa microkernel, i386-*-vsta
16736     * Motorola M68000 family
16737          + HPUX, m68k-hp-hpux* and m68000-hp-hpux*
16738          + NetBSD with a.out object format (before NetBSD 1.4),
16739            m68k-*-*-netbsd* except m68k-*-*-netbsdelf*
16740          + Generic System V r.4, m68k-*-sysv4*
16741     * VAX
16742          + Generic VAX, vax-*-* (This is generic VAX only; we have not
16743            obsoleted any VAX triples for specific operating systems.)
16744
16745Documentation improvements
16746
16747Other significant improvements
16748
16749     * The build system has undergone several significant cleanups.
16750       Subdirectories will only be configured if they are being built, and
16751       all subdirectory configures are run from the make command. The top
16752       level has been autoconfiscated.
16753     * Building GCC no longer writes to its source directory. This should
16754       help those wishing to share a read-only source directory over NFS
16755       or build from a CD. The exceptions to this feature are if you
16756       configure with either --enable-maintainer-mode or
16757       --enable-generated-files-in-srcdir.
16758     * The -W warning option has been renamed to -Wextra, which is more
16759       easily understood. The older spelling will be retained for
16760       backwards compatibility.
16761     * Substantial improvements in compile time have been made,
16762       particularly for non-optimizing compilations.
16763     __________________________________________________________________
16764
16765GCC 3.4.0
16766
16767  Bug Fixes
16768
16769   A vast number of bugs have been fixed in 3.4.0, too many to publish a
16770   complete list here. [26]Follow this link to query the Bugzilla database
16771   for the list of over 900 bugs fixed in 3.4.0. This is the list of all
16772   bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4
16773   regressions.
16774     __________________________________________________________________
16775
16776GCC 3.4.1
16777
16778  Bug Fixes
16779
16780   This section lists the problem reports (PRs) from GCC's bug tracking
16781   system that are known to be fixed in the 3.4.1 release. This list might
16782   not be complete (that is, it is possible that some PRs that have been
16783   fixed are not listed here).
16784
16785    Bootstrap failures
16786
16787     * [27]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler
16788       emitted - PIC related
16789     * [28]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf
16790     * [29]14760 A bug in configure.in prevents using both
16791       --program-suffix and --program-prefix
16792     * [30]14671 [hppa64] bootstrap fails: ICE in
16793       save_call_clobbered_regs, in caller_save.c
16794     * [31]15093 [alpha][Java] make bootstrap fails to configure libffi on
16795       Alpha
16796     * [32]15178 Solaris 9/x86 fails linking after stage 3
16797
16798    Multi-platform internal compiler errors (ICEs)
16799
16800     * [33]12753 (preprocessor) Memory corruption in preprocessor on bad
16801       input
16802     * [34]13985 ICE in gcc.c-torture/compile/930621-1.c
16803     * [35]14810 (c++) tree check failures with invalid code involving
16804       templates
16805     * [36]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in
16806       cp/parser.c
16807     * [37]15044 (c++) ICE on syntax error, template header
16808     * [38]15057 (c++) Compiling of conditional value throw constructs
16809       cause a segmentation violation
16810     * [39]15064 (c++) typeid of template parameter gives ICE
16811     * [40]15142 (c++) ICE when passing a string where a char* is expected
16812       in a throw statement
16813     * [41]15159 ICE in rtl_verify_flow_info_1
16814     * [42]15165 (c++) ICE in instantiate_template
16815     * [43]15193 Unary minus using pointer to V4SF vector causes
16816       -fforce-mem to exhaust all memory
16817     * [44]15209 (c++) Runs out of memory with packed structs
16818     * [45]15227 (c++) Trouble with invalid function definition
16819     * [46]15285 (c++) instantiate_type ICE when forming pointer to
16820       template function
16821     * [47]15299 (c++) ICE in resolve_overloaded_unification
16822     * [48]15329 (c++) ICE on constructor of member template
16823     * [49]15550 ICE in extract_insn, in recog.c
16824     * [50]15554 (c++) ICE in tsubst_copy, in cp/pt.c
16825     * [51]15640 (c++) ICE on invalid code in arg_assoc, in
16826       cp/name-lookup.c
16827     * [52]15666 [unit-at-a-time] Gcc abort on valid code
16828     * [53]15696 (c++) ICE with bad pointer-to-member code
16829     * [54]15701 (c++) ICE with friends and template template parameter
16830     * [55]15761 ICE in do_SUBST, in combine.c
16831     * [56]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops
16832
16833    Ada
16834
16835     * [57]14538 All RTEMS targets broken for gnat
16836
16837    C front end
16838
16839     * [58]12391 missing warning about assigning to an incomplete type
16840     * [59]14649 atan(1.0) should not be a constant expression
16841     * [60]15004 [unit-at-a-time] no warning for unused paramater in
16842       static function
16843     * [61]15749 --pedantic-errors behaves differently from --pedantic
16844       with C-compiler on GNU/Linux
16845
16846    C++ compiler and library
16847
16848     * [62]10646 non-const reference is incorrectly matched in a "const T"
16849       partial specialization
16850     * [63]12077 wcin.rdbuf()->in_avail() return value too high
16851     * [64]13598 enc_filebuf doesn't work
16852     * [65]14211 const_cast returns lvalue but should be rvalue
16853     * [66]14220 num_put::do_put() undesired float/double behavior
16854     * [67]14245 problem with user-defined allocators in std::basic_string
16855     * [68]14340 libstdc++ Debug mode: failure to convert iterator to
16856       const_iterator
16857     * [69]14600 __gnu_cxx::stdio_sync_filebuf should expose internal
16858       FILE*
16859     * [70]14668 no warning anymore for reevaluation of declaration
16860     * [71]14775 LFS (large file support) tests missing
16861     * [72]14821 Duplicate namespace alias declaration should not conflict
16862     * [73]14930 Friend declaration ignored
16863     * [74]14932 cannot use offsetof to get offsets of array elements in
16864       g++ 3.4.0
16865     * [75]14950 [non unit-at-a-time] always_inline does not mix with
16866       templates and -O0
16867     * [76]14962 g++ ignores #pragma redefine_extname
16868     * [77]14975 Segfault on low-level write error during imbue
16869     * [78]15002 Linewise stream input is unusably slow (std::string slow)
16870     * [79]15025 compiler accepts redeclaration of template as
16871       non-template
16872     * [80]15046 [arm] Math functions misdetected by cross configuration
16873     * [81]15069 a bit test on a variable of enum type is miscompiled
16874     * [82]15074 g++ -lsupc++ still links against libstdc++
16875     * [83]15083 spurious "statement has no effect" warning
16876     * [84]15096 parse error with templates and pointer to const member
16877     * [85]15287 combination of operator[] and operator .* fails in
16878       templates
16879     * [86]15317 __attribute__ unused in first parameter of constructor
16880       gives error
16881     * [87]15337 sizeof on incomplete type diagnostic
16882     * [88]15361 bitset<>::_Find_next fails
16883     * [89]15412 _GLIBCXX_ symbols symbols defined and used in different
16884       namespaces
16885     * [90]15427 valid code results in incomplete type error
16886     * [91]15471 Incorrect member pointer offsets in anonymous
16887       structs/unions
16888     * [92]15503 nested template problem
16889     * [93]15507 compiler hangs while laying out union
16890     * [94]15542 operator & and template definitions
16891     * [95]15565 SLES9: leading + sign for unsigned int with showpos
16892     * [96]15625 friend defined inside a template fails to find static
16893       function
16894     * [97]15629 Function templates, overloads, and friend name injection
16895     * [98]15742 'noreturn' attribute ignored in method of template
16896       functions.
16897     * [99]15775 Allocator::pointer consistently ignored
16898     * [100]15821 Duplicate namespace alias within namespace rejected
16899     * [101]15862 'enum yn' fails (confict with undeclared builtin)
16900     * [102]15875 rejects pointer to member in template
16901     * [103]15877 valid code using templates and anonymous enums is
16902       rejected
16903     * [104]15947 Puzzling error message for wrong destructor declaration
16904       in template class
16905     * [105]16020 cannot copy __gnu_debug::bitset
16906     * [106]16154 input iterator concept too restrictive
16907     * [107]16174 deducing top-level consts
16908
16909    Java
16910
16911     * [108]14315 Java compiler is not parallel make safe
16912
16913    Fortran
16914
16915     * [109]15151 [g77] incorrect logical i/o in 64-bit mode
16916
16917    Objective-C
16918
16919     * [110]7993 private variables cannot be shadowed in subclasses
16920
16921    Optimization bugs
16922
16923     * [111]15228 useless copies of floating point operands
16924     * [112]15345 [non-unit-at-a-time] unreferenced nested inline
16925       functions not optimized away
16926     * [113]15945 Incorrect floating point optimization
16927     * [114]15526 ftrapv aborts on 0 * (-1)
16928     * [115]14690 Miscompiled POOMA tests
16929     * [116]15112 GCC generates code to write to unchanging memory
16930
16931    Preprocessor
16932
16933     * [117]15067 Minor glitch in the source of cpp
16934
16935    Main driver program bugs
16936
16937     * [118]1963 collect2 interprets -oldstyle_liblookup as -o
16938       ldstyle_liblookup
16939
16940    x86-specific (Intel/AMD)
16941
16942     * [119]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND*
16943       section}
16944
16945    HPPA-specific
16946
16947     * [120]14782 GCC produces an unaligned data access at -O2
16948     * [121]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2
16949     * [122]15202 ICE in reload_cse_simplify_operands, in postreload.c
16950
16951    IA64-specific
16952
16953     * [123]14610 __float80 constants incorrectly emitted
16954     * [124]14813 init_array sections are initialized in the wrong order
16955     * [125]14857 GCC segfault on duplicated asm statement
16956     * [126]15598 Gcc 3.4 ICE on valid code
16957     * [127]15653 Gcc 3.4 ICE on valid code
16958
16959    MIPS-specific
16960
16961     * [128]15189 wrong filling of delay slot with -march=mips1 -G0
16962       -mno-split-addresses -mno-explicit-relocs
16963     * [129]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as
16964       2.14.91
16965     * [130]16144 Bogus reference to __divdf3 when -O1
16966     * [131]16176 Miscompilation of unaligned data in MIPS backend
16967
16968    PowerPC-specific
16969
16970     * [132]11591 ICE in gcc.dg/altivec-5.c
16971     * [133]12028 powerpc-eabispe produces bad sCOND operation
16972     * [134]14478 rs6000 geu/ltu patterns generate incorrect code
16973     * [135]14567 long double and va_arg complex args
16974     * [136]14715 Altivec stack layout may overlap gpr save with stack
16975       temps
16976     * [137]14902 (libstdc++) Stream checking functions fail when -pthread
16977       option is used.
16978     * [138]14924 Compiler ICE on valid code
16979     * [139]14960 -maltivec affects vector return with -mabi=no-altivec
16980     * [140]15106 vector varargs failure passing from altivec to
16981       non-altivec code for -m32
16982     * [141]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 &
16983       half-word operation
16984     * [142]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx
16985       and stvx
16986     * [143]15662 Segmentation fault when an exception is thrown - even if
16987       try and catch are specified
16988
16989    s390-specific
16990
16991     * [144]15054 Bad code due to overlapping stack temporaries
16992
16993    SPARC-specific
16994
16995     * [145]15783 ICE with union assignment in 64-bit mode
16996     * [146]15626 GCC 3.4 emits "ld: warning: relocation error:
16997       R_SPARC_UA32"
16998
16999    x86-64-specific
17000
17001     * [147]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64
17002     * [148]14723 Backported -march=nocona from mainline
17003     * [149]15290 __float128 failed to pass to function properly
17004
17005    Cygwin/Mingw32-specific
17006
17007     * [150]15250 Option -mms-bitfields support on GCC 3.4 is not
17008       conformant to MS layout
17009     * [151]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe
17010       worker on windows32 targets
17011
17012    Bugs specific to embedded processors
17013
17014     * [152]8309 [m68k] -m5200 produces erroneous SImode set of short
17015       varaible on stack
17016     * [153]13250 [SH] Gcc code for rotation clobbers the register, but
17017       gcc continues to use the register as if it was not clobbered
17018     * [154]13803 [coldfire] movqi operand constraints too restrictivefor
17019       TARGET_COLDFIRE
17020     * [155]14093 [SH] ICE for code when using -mhitachi option in SH
17021     * [156]14457 [m6811hc] ICE with simple c++ source
17022     * [157]14542 [m6811hc] ICE on simple source
17023     * [158]15100 [SH] cc1plus got hang-up on
17024       libstdc++-v3/testsuite/abi_check.cc
17025     * [159]15296 [CRIS] Delayed branch scheduling causing invalid code on
17026       cris-*
17027     * [160]15396 [SH] ICE with -O2 -fPIC
17028     * [161]15782 [coldfire] m68k_output_mi_thunk emits wrong code for
17029       ColdFire
17030
17031    Testsuite problems (compiler not affected)
17032
17033     * [162]11610 libstdc++ testcases 27_io/* don't work properly remotely
17034     * [163]15488 (libstdc++) possibly insufficient file permissions for
17035       executing test suite
17036     * [164]15489 (libstdc++) testsuite_files determined incorrectly
17037
17038    Documentation bugs
17039
17040     * [165]13928 (libstdc++) no whatis info in some man pages generated
17041       by doxygen
17042     * [166]14150 Ada documentation out of date
17043     * [167]14949 (c++) Need to document method visibility changes
17044     * [168]15123 libstdc++-doc: Allocators.3 manpage is empty
17045     __________________________________________________________________
17046
17047GCC 3.4.2
17048
17049  Bug Fixes
17050
17051   This section lists the problem reports (PRs) from GCC's bug tracking
17052   system that are known to be fixed in the 3.4.2 release. This list might
17053   not be complete (that is, it is possible that some PRs that have been
17054   fixed are not listed here).
17055
17056    Bootstrap failures and issues
17057
17058     * [169]16469 [mips-sgi-irix5.3] bootstrap fails in
17059       libstdc++-v3/testsuite
17060     * [170]16344 [hppa-linux-gnu] libstdc++'s PCH built by
17061       profiledbootstrap does not work with the built compiler
17062     * [171]16842 [Solaris/x86] mkheaders can not find mkheaders.conf
17063
17064    Multi-platform internal compiler errors (ICEs)
17065
17066     * [172]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in
17067       cp_parser_class_specifier, in cp/parser.c
17068     * [173]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
17069     * [174]15461 (c++) ICE due to NRV and inlining
17070     * [175]15890 (c++) ICE in c_expand_expr, in c-common.c
17071     * [176]16180 ICE: segmentation fault in RTL optimization
17072     * [177]16224 (c++) ICE in write_unscoped_name (template/namespace)
17073     * [178]16408 ICE: in delete_insn, in cfgrtl.c
17074     * [179]16529 (c++) ICE for: namespace-alias shall not be declared as
17075       the name of any other entity
17076     * [180]16698 (c++) ICE with exceptions and declaration of __cxa_throw
17077     * [181]16706 (c++) ICE in finish_member_declaration, in
17078       cp/semantics.c
17079     * [182]16810 (c++) Legal C++ program with cast gives ICE in
17080       build_ptrmemfunc
17081     * [183]16851 (c++) ICE when throwing a comma expression
17082     * [184]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c
17083     * [185]16904 (c++) ICE in finish_class_member_access_expr, in
17084       cp/typeck.c
17085     * [186]16905 (c++) ICE (segfault) with exceptions
17086     * [187]16964 (c++) ICE in cp_parser_class_specifier due to
17087       redefinition
17088     * [188]17068 (c++) ICE: tree check: expected class 'd', have 'x'
17089       (identifier_node) in dependent_template_p, in cp/pt.c
17090
17091    Preprocessor bugs
17092
17093     * [189]16366 Preprocessor option -remap causes memory corruption
17094
17095    Optimization
17096
17097     * [190]15345 unreferenced nested inline functions not optimized away
17098     * [191]16590 Incorrect execution when compiling with -O2
17099     * [192]16693 Bitwise AND is lost when used within a cast to an enum
17100       of the same precision
17101     * [193]17078 Jump into if(0) substatement fails
17102
17103    Problems in generated debug information
17104
17105     * [194]13956 incorrect stabs for nested local variables
17106
17107    C front end bugs
17108
17109     * [195]16684 GCC should not warn about redundant redeclarations of
17110       built-ins
17111
17112    C++ compiler and library
17113
17114     * [196]12658 Thread safety problems in locale::global() and
17115       locale::locale()
17116     * [197]13092 g++ accepts invalid pointer-to-member conversion
17117     * [198]15320 Excessive memory consumption
17118     * [199]16246 Incorrect template argument deduction
17119     * [200]16273 Memory exhausted when using nested classes and virtual
17120       functions
17121     * [201]16401 ostringstream in gcc 3.4.x very slow for big data
17122     * [202]16411 undefined reference to
17123       __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char>
17124       >::file()
17125     * [203]16489 G++ incorrectly rejects use of a null constant integral
17126       expression as a null constant pointer
17127     * [204]16618 offsetof fails with constant member
17128     * [205]16637 syntax error reported for valid input code
17129     * [206]16717 __attribute__((constructor)) broken in C++
17130     * [207]16813 compiler error in DEBUG version of range insertion
17131       std::map::insert
17132     * [208]16853 pointer-to-member initialization from incompatible one
17133       accepted
17134     * [209]16889 ambiguity is not detected
17135     * [210]16959 Segmentation fault in ios_base::sync_with_stdio
17136
17137    Java compiler and library
17138
17139     * [211]7587 direct threaded interpreter not thread-safe
17140     * [212]16473 ServerSocket accept() leaks file descriptors
17141     * [213]16478 Hash synchronization deadlock with finalizers
17142
17143    Alpha-specific
17144
17145     * [214]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c
17146     * [215]16974 could not split insn (ice in final_scan_insn, in
17147       final.c)
17148
17149    x86-specific
17150
17151     * [216]16298 ICE in output_operand
17152     * [217]17113 ICE with SSE2 intrinsics
17153
17154    x86-64 specific
17155
17156     * [218]14697 libstdc++ couldn't find 32bit libgcc_s
17157
17158    MIPS-specific
17159
17160     * [219]15869 [mips64] No NOP after LW (with -mips1 -O0)
17161     * [220]16325 [mips64] value profiling clobbers gp on mips
17162     * [221]16357 [mipsisa64-elf] ICE copying 7 bytes between extern
17163       char[]s
17164     * [222]16380 [mips64] Use of uninitialised register after dbra
17165       conversion
17166     * [223]16407 [mips64] Unaligned access to local variables
17167     * [224]16643 [mips64] verify_local_live_at_start ICE after
17168       crossjumping & cfgcleanup
17169
17170    ARM-specific
17171
17172     * [225]15927 THUMB -O2: strength-reduced iteration variable ends up
17173       off by 1
17174     * [226]15948 THUMB: ICE with non-commutative cbranch
17175     * [227]17019 THUMB: bad switch statement in md code for
17176       addsi3_cbranch_scratch
17177
17178    IA64-specific
17179
17180     * [228]16130 ICE on valid code: in bundling, in config/ia64/ia64.c
17181       (-mtune=merced)
17182     * [229]16142 ICE on valid code: in bundling, in config/ia64/ia64.c
17183       (-mtune=itanium)
17184     * [230]16278 Gcc failed to build Linux kernel with -mtune=merced
17185     * [231]16414 ICE on valid code: typo in comparison of asm_noperands
17186       result
17187     * [232]16445 ICE on valid code: don't count ignored insns
17188     * [233]16490 ICE (segfault) while compiling with -fprofile-use
17189     * [234]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS
17190
17191    PowerPC-specific
17192
17193     * [235]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x
17194     * [236]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions
17195       issue)
17196
17197    SPARC-specific
17198
17199     * [237]16199 ICE while compiling apache 2.0.49
17200     * [238]16416 -m64 doesn't imply -mcpu=v9 anymore
17201     * [239]16430 ICE when returning non-C aggregates larger than 16 bytes
17202
17203    Bugs specific to embedded processors
17204
17205     * [240]16379 [m32r] can't output large model function call of memcpy
17206     * [241]17093 [m32r] ICE with -msdata=use -O0
17207     * [242]17119 [m32r] ICE at switch case 0x8000
17208
17209    DJGPP-specific
17210
17211     * [243]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp
17212
17213    Alpha Tru64-specific
17214
17215     * [244]16210 libstdc++ gratuitously omits "long long" I/O
17216
17217    Testsuite, documentation issues (compiler is not affected):
17218
17219     * [245]15488 (libstdc++) possibly insufficient file permissions for
17220       executing test suite
17221     * [246]16250 ada/doctools runs makeinfo even in release tarball
17222     __________________________________________________________________
17223
17224GCC 3.4.3
17225
17226   This is the [247]list of problem reports (PRs) from GCC's bug tracking
17227   system that are known to be fixed in the 3.4.3 release. This list might
17228   not be complete (that is, it is possible that some PRs that have been
17229   fixed are not listed here).
17230
17231    Bootstrap failures
17232
17233     * [248]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1
17234     * [249]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold
17235       when undeclared
17236
17237    Internal compiler errors (ICEs) affecting multiple platforms
17238
17239     * [250]13948 (java) GCJ segmentation fault while compiling GL4Java
17240       .class files
17241     * [251]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
17242     * [252]16301 (c++) ICE when "strong" attribute is attached to a using
17243       directive
17244     * [253]16566 ICE with flexible arrays
17245     * [254]17023 ICE with nested functions in parameter declaration
17246     * [255]17027 ICE with noreturn function in loop at -O2
17247     * [256]17524 ICE in grokdeclarator, in cp/decl.c
17248     * [257]17826 (c++) ICE in cp_tree_equal
17249
17250    C and optimization bugs
17251
17252     * [258]15526 -ftrapv aborts on 0 * (-1)
17253     * [259]16999 #ident stopped working
17254     * [260]17503 quadratic behaviour in invalid_mode_change_p
17255     * [261]17581 Long long arithmetic fails inside a switch/case
17256       statement when compiled with -O2
17257     * [262]18129 -fwritable-strings doesn't work
17258
17259    C++ compiler and library bugs
17260
17261     * [263]10975 incorrect initial ostringstream::tellp()
17262     * [264]11722 Unbuffered filebuf::sgetn is slow
17263     * [265]14534 Unrecognizing static function as a template parameter
17264       when its return value is also templated
17265     * [266]15172 Copy constructor optimization in aggregate
17266       initialization
17267     * [267]15786 Bad error message for frequently occuring error.
17268     * [268]16162 Rejects valid member-template-definition
17269     * [269]16612 empty basic_strings can't live in shared memory
17270     * [270]16715 std::basic_iostream is instantiated when used, even
17271       though instantiations are already contained in libstdc++
17272     * [271]16848 code in /ext/demangle.h appears broken
17273     * [272]17132 GCC fails to eliminate function template specialization
17274       when argument deduction fails
17275     * [273]17259 One more _S_leaf incorrectly qualified with _RopeRep::
17276       in ropeimpl.h
17277     * [274]17327 use of `enumeral_type' in template type unification
17278     * [275]17393 "unused variable '._0'" warning with -Wall
17279     * [276]17501 Confusion with member templates
17280     * [277]17537 g++ not passing -lstdc++ to linker when all command line
17281       arguments are libraries
17282     * [278]17585 usage of unqualified name of static member from within
17283       class not allowed
17284     * [279]17821 Poor diagnostic for using "." instead of "->"
17285     * [280]17829 wrong error: call of overloaded function is ambiguous
17286     * [281]17851 Misleading diagnostic for invalid function declarations
17287       with undeclared types
17288     * [282]17976 Destructor is called twice
17289     * [283]18020 rejects valid definition of enum value in template
17290     * [284]18093 bogus conflict in namespace aliasing
17291     * [285]18140 C++ parser bug when using >> in templates
17292
17293    Fortran
17294
17295     * [286]17541 data statements with double precision constants fail
17296
17297    x86-specific
17298
17299     * [287]17853 -O2 ICE for MMX testcase
17300
17301    SPARC-specific
17302
17303     * [288]17245 ICE compiling gsl-1.5 statistics/lag1.c
17304
17305    Darwin-specific
17306
17307     * [289]17167 FATAL:Symbol L_foo$stub already defined.
17308
17309    AIX-specific
17310
17311     * [290]17277 could not catch an exception when specified -maix64
17312
17313    Solaris-specific
17314
17315     * [291]17505 <cmath> calls acosf(), ceilf(), and other functions
17316       missing from system libraries
17317
17318    HP/UX specific:
17319
17320     * [292]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl
17321
17322    ARM-specific
17323
17324     * [293]17384 ICE with mode attribute on structures
17325
17326    MIPS-specific
17327
17328     * [294]17770 No NOP after LWL with -mips1
17329
17330    Other embedded target specific
17331
17332     * [295]11476 [arc-elf] gcc ICE on newlib's vfprintf.c
17333     * [296]14064 [avr-elf] -fdata-sections triggers ICE
17334     * [297]14678 [m68hc11-elf] gcc ICE
17335     * [298]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__
17336     * [299]15790 [i686-coff] Alignment error building gcc with i686-coff
17337       target
17338     * [300]15886 [SH] Miscompilation with -O2 -fPIC
17339     * [301]16884 [avr-elf] [fweb related] bug while initializing
17340       variables
17341
17342    Bugs relating to debugger support
17343
17344     * [302]13841 missing debug info for _Complex function arguments
17345     * [303]15860 [big-endian targets] No DW_AT_location debug info is
17346       emitted for formal arguments to a function that uses "register"
17347       qualifiers
17348
17349    Testsuite issues (compiler not affected)
17350
17351     * [304]17465 Testsuite in libffi overrides LD_LIBRARY_PATH
17352     * [305]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH
17353     * [306]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit
17354       testsuite
17355
17356    Documentation
17357
17358     * [307]15498 typo in gcc manual: non-existing locale example en_UK,
17359       should be en_GB
17360     * [308]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap:
17361       document broken shell
17362     * [309]16406 USE_LD_AS_NEEDED undocumented
17363     __________________________________________________________________
17364
17365GCC 3.4.4
17366
17367   This is the [310]list of problem reports (PRs) from GCC's bug tracking
17368   system that are known to be fixed in the 3.4.4 release. This list might
17369   not be complete (that is, it is possible that some PRs that have been
17370   fixed are not listed here).
17371     __________________________________________________________________
17372
17373GCC 3.4.5
17374
17375   This is the [311]list of problem reports (PRs) from GCC's bug tracking
17376   system that are known to be fixed in the 3.4.5 release. This list might
17377   not be complete (that is, it is possible that some PRs that have been
17378   fixed are not listed here).
17379
17380    Bootstrap issues
17381
17382     * [312]24688 sco_math fixincl breaks math.h
17383
17384    C compiler bugs
17385
17386     * [313]17188 struct Foo { } redefinition
17387     * [314]20187 wrong code for ((unsigned char)(unsigned long
17388       long)((a?a:1)&(a*b)))?0:1)
17389     * [315]21873 infinite warning loop on bad array initializer
17390     * [316]21899 enum definition accepts values to be overriden
17391     * [317]22061 ICE in find_function_data, in function.c
17392     * [318]22308 Failure to diagnose violation of constraint 6.516p2
17393     * [319]22458 ICE on missing brace
17394     * [320]22589 ICE casting to long long
17395     * [321]24101 Segfault with preprocessed source
17396
17397    C++ compiler and library bugs
17398
17399     * [322]10611 operations on vector mode not recognized in C++
17400     * [323]13377 unexpected behavior of namespace usage directive
17401     * [324]16002 Strange error message with new parser
17402     * [325]17413 local classes as template argument
17403     * [326]17609 spurious error message after using keyword
17404     * [327]17618 ICE in cp_convert_to_pointer, in cp/cvt.c
17405     * [328]18124 ICE with invalid template template parameter
17406     * [329]18155 typedef in template declaration not rejected
17407     * [330]18177 ICE with const_cast for undeclared variable
17408     * [331]18368 C++ error message regression
17409     * [332]16378 ICE when returning a copy of a packed member
17410     * [333]18466 int ::i; accepted
17411     * [334]18512 ICE on invalid usage of template base class
17412     * [335]18454 ICE when returning undefined type
17413     * [336]18738 typename not allowed with non-dependent qualified name
17414     * [337]18803 rejects access to operator() in template
17415     * [338]19004 ICE in uses_template_parms, in cp/pt.c
17416     * [339]19208 Spurious error about variably modified type
17417     * [340]18253 bad error message / ICE for invalid template parameter
17418     * [341]19608 ICE after friend function definition in local class
17419     * [342]19884 ICE on explicit instantiation of a non-template
17420       constructor
17421     * [343]20153 ICE when C++ template function contains anonymous union
17422     * [344]20563 Infinite loop in diagnostic (and ice after error
17423       message)
17424     * [345]20789 ICE with incomplete type in template
17425     * [346]21336 Internal compiler error when using custom new operators
17426     * [347]21768 ICE in error message due to violation of coding
17427       conventions
17428     * [348]21853 constness of pointer to data member ignored
17429     * [349]21903 Default argument of template function causes a
17430       compile-time error
17431     * [350]21983 multiple diagnostics
17432     * [351]21987 New testsuite failure
17433       g++.dg/warn/conversion-function-1.C
17434     * [352]22153 ICE on invalid template specialization
17435     * [353]22172 Internal compiler error, seg fault.
17436     * [354]21286 filebuf::xsgetn vs pipes
17437     * [355]22233 ICE with wrong number of template parameters
17438     * [356]22508 ICE after invalid operator new
17439     * [357]22545 ICE with pointer to class member & user defined
17440       conversion operator
17441     * [358]23528 Wrong default allocator in ext/hash_map
17442     * [359]23550 char_traits requirements/1.cc test bad math
17443     * [360]23586 Bad diagnostic for invalid namespace-name
17444     * [361]23624 ICE in invert_truthvalue, in fold-const.c
17445     * [362]23639 Bad error message: not a member of '<declaration error>'
17446     * [363]23797 ICE on typename outside template
17447     * [364]23965 Bogus error message: no matching function for call to
17448       'foo(<type error>)'
17449     * [365]24052 &#`label_decl' not supported by dump_expr#<expression
17450       error>
17451     * [366]24580 virtual base class cause exception not to be caught
17452
17453    Problems in generated debug information
17454
17455     * [367]24267 Bad DWARF for altivec vectors
17456
17457    Optimizations issues
17458
17459     * [368]17810 ICE in verify_local_live_at_start
17460     * [369]17860 Wrong generated code for loop with varying bound
17461     * [370]21709 ICE on compile-time complex NaN
17462     * [371]21964 broken tail call at -O2 or more
17463     * [372]22167 Strange optimization bug when using -Os
17464     * [373]22619 Compilation failure for real_const_1.f and
17465       real_const_2.f90
17466     * [374]23241 Invalid code generated for comparison of uchar to 255
17467     * [375]23478 Miscompilation due to reloading of a var that is also
17468       used in EH pad
17469     * [376]24470 segmentation fault in cc1plus when compiling with -O
17470     * [377]24950 ICE in operand_subword_force
17471
17472    Precompiled headers problems
17473
17474     * [378]14400 Cannot compile qt-x11-free-3.3.0
17475     * [379]14940 PCH largefile test fails on various platforms
17476
17477    Preprocessor bugs
17478
17479     * [380]20239 ICE on empty preprocessed input
17480     * [381]15220 "gcc -E -MM -MG" reports missing system headers in
17481       source directory
17482
17483    Testsuite issues
17484
17485     * [382]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on
17486       i686-pc-linux-gnu
17487
17488    Alpha specific
17489
17490     * [383]21888 bootstrap failure with linker relaxation enabled
17491
17492    ARM specific
17493
17494     * [384]15342 [arm-linux]: ICE in verify_local_live_at_start
17495     * [385]23985 Memory aliasing information incorrect in inlined memcpy
17496
17497    ColdFile specific
17498
17499     * [386]16719 Illegal move of byte into address register causes
17500       compiler to ICE
17501
17502    HPPA specific
17503
17504     * [387]21723 ICE while building libgfortran
17505     * [388]21841 -mhp-ld/-mgnu-ld documentation
17506
17507    IA-64 specific
17508
17509     * [389]23644 IA-64 hardware models and configuration options
17510       documentation error
17511     * [390]24718 Shared libgcc not used for linking by default
17512
17513    M68000 specific
17514
17515     * [391]18421 ICE in reload_cse_simplify_operands, in postreload.c
17516
17517    MIPS specific
17518
17519     * [392]20621 ICE in change_address_1, in emit-rtl.c
17520
17521    PowerPC and PowerPC64 specific
17522
17523     * [393]18583 error on valid code: const
17524       __attribute__((altivec(vector__))) doesn't work in arrays
17525     * [394]20191 ICE in reload_cse_simplify_operands
17526     * [395]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined
17527     * [396]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set
17528     * [397]23404 gij trashes args of functions with more than 8 fp args
17529     * [398]23539 C & C++ compiler generating misaligned references
17530       regardless of compiler flags
17531     * [399]24102 floatdisf2_internal2 broken
17532     * [400]24465 -mminimal-toc miscompilation of __thread vars
17533
17534    Solaris specific
17535
17536     * [401]19933 Problem with define of HUGE_VAL in math_c99
17537     * [402]21889 Native Solaris assembler cannot grok DTP-relative debug
17538       symbols
17539
17540    SPARC specific
17541
17542     * [403]19300 PCH failures on sparc-linux
17543     * [404]20301 Assembler labels have a leading "-"
17544     * [405]20673 C PCH testsuite assembly comparison failure
17545
17546    x86 and x86_64 specific
17547
17548     * [406]18582 ICE with arrays of type V2DF
17549     * [407]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2
17550       -fsched2-use-traces
17551     * [408]21716 ICE in reg-stack.c's swap_rtx_condition
17552     * [409]24315 amd64 fails -fpeephole2
17553     __________________________________________________________________
17554
17555GCC 3.4.6
17556
17557   This is the [410]list of problem reports (PRs) from GCC's bug tracking
17558   system that are known to be fixed in the 3.4.6 release. This list might
17559   not be complete (that is, it is possible that some PRs that have been
17560   fixed are not listed here).
17561
17562
17563    For questions related to the use of GCC, please consult these web
17564    pages and the [411]GCC manuals. If that fails, the
17565    [412]gcc-help@gcc.gnu.org mailing list might help. Comments on these
17566    web pages and the development of GCC are welcome on our developer
17567    list at [413]gcc@gcc.gnu.org. All of [414]our lists have public
17568    archives.
17569
17570   Copyright (C) [415]Free Software Foundation, Inc. Verbatim copying and
17571   distribution of this entire article is permitted in any medium,
17572   provided this notice is preserved.
17573
17574   These pages are [416]maintained by the GCC team. Last modified
17575   2023-01-27.
17576
17577References
17578
17579   1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
17580   2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus
17581   3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
17582   4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems
17583   5. http://gcc.gnu.org/gcc-3.4/mips-abi.html
17584   6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
17585   7. https://www.boost.org/
17586   8. https://gcc.gnu.org/PR11953
17587   9. https://gcc.gnu.org/PR8361
17588  10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other%20Builtins
17589  11. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209
17590  12. http://gcc.gnu.org/bugs/#cxx_rvalbind
17591  13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
17592  14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
17593  15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
17594  16. http://www.gnu.org/software/classpath/
17595  17. https://www.eclipse.org/
17596  18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html
17597  19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html
17598  20. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
17599  21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html
17600  22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
17601  23. http://gcc.gnu.org/gcc-3.4/mips-abi.html
17602  24. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html
17603  25. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
17604  26. https://gcc.gnu.org/bugzilla/buglist.cgi?short_desc_type=notregexp&short_desc=%5C%5B3%5C.4.*%5BRr%5Degression&target_milestone=3.4.0&bug_status=RESOLVED&resolution=FIXED
17605  27. https://gcc.gnu.org/PR10129
17606  28. https://gcc.gnu.org/PR14576
17607  29. https://gcc.gnu.org/PR14760
17608  30. https://gcc.gnu.org/PR14671
17609  31. https://gcc.gnu.org/PR15093
17610  32. https://gcc.gnu.org/PR15178
17611  33. https://gcc.gnu.org/PR12753
17612  34. https://gcc.gnu.org/PR13985
17613  35. https://gcc.gnu.org/PR14810
17614  36. https://gcc.gnu.org/PR14883
17615  37. https://gcc.gnu.org/PR15044
17616  38. https://gcc.gnu.org/PR15057
17617  39. https://gcc.gnu.org/PR15064
17618  40. https://gcc.gnu.org/PR15142
17619  41. https://gcc.gnu.org/PR15159
17620  42. https://gcc.gnu.org/PR15165
17621  43. https://gcc.gnu.org/PR15193
17622  44. https://gcc.gnu.org/PR15209
17623  45. https://gcc.gnu.org/PR15227
17624  46. https://gcc.gnu.org/PR15285
17625  47. https://gcc.gnu.org/PR15299
17626  48. https://gcc.gnu.org/PR15329
17627  49. https://gcc.gnu.org/PR15550
17628  50. https://gcc.gnu.org/PR15554
17629  51. https://gcc.gnu.org/PR15640
17630  52. https://gcc.gnu.org/PR15666
17631  53. https://gcc.gnu.org/PR15696
17632  54. https://gcc.gnu.org/PR15701
17633  55. https://gcc.gnu.org/PR15761
17634  56. https://gcc.gnu.org/PR15829
17635  57. https://gcc.gnu.org/PR14538
17636  58. https://gcc.gnu.org/PR12391
17637  59. https://gcc.gnu.org/PR14649
17638  60. https://gcc.gnu.org/PR15004
17639  61. https://gcc.gnu.org/PR15749
17640  62. https://gcc.gnu.org/PR10646
17641  63. https://gcc.gnu.org/PR12077
17642  64. https://gcc.gnu.org/PR13598
17643  65. https://gcc.gnu.org/PR14211
17644  66. https://gcc.gnu.org/PR14220
17645  67. https://gcc.gnu.org/PR14245
17646  68. https://gcc.gnu.org/PR14340
17647  69. https://gcc.gnu.org/PR14600
17648  70. https://gcc.gnu.org/PR14668
17649  71. https://gcc.gnu.org/PR14775
17650  72. https://gcc.gnu.org/PR14821
17651  73. https://gcc.gnu.org/PR14930
17652  74. https://gcc.gnu.org/PR14932
17653  75. https://gcc.gnu.org/PR14950
17654  76. https://gcc.gnu.org/PR14962
17655  77. https://gcc.gnu.org/PR14975
17656  78. https://gcc.gnu.org/PR15002
17657  79. https://gcc.gnu.org/PR15025
17658  80. https://gcc.gnu.org/PR15046
17659  81. https://gcc.gnu.org/PR15069
17660  82. https://gcc.gnu.org/PR15074
17661  83. https://gcc.gnu.org/PR15083
17662  84. https://gcc.gnu.org/PR15096
17663  85. https://gcc.gnu.org/PR15287
17664  86. https://gcc.gnu.org/PR15317
17665  87. https://gcc.gnu.org/PR15337
17666  88. https://gcc.gnu.org/PR15361
17667  89. https://gcc.gnu.org/PR15412
17668  90. https://gcc.gnu.org/PR15427
17669  91. https://gcc.gnu.org/PR15471
17670  92. https://gcc.gnu.org/PR15503
17671  93. https://gcc.gnu.org/PR15507
17672  94. https://gcc.gnu.org/PR15542
17673  95. https://gcc.gnu.org/PR15565
17674  96. https://gcc.gnu.org/PR15625
17675  97. https://gcc.gnu.org/PR15629
17676  98. https://gcc.gnu.org/PR15742
17677  99. https://gcc.gnu.org/PR15775
17678 100. https://gcc.gnu.org/PR15821
17679 101. https://gcc.gnu.org/PR15862
17680 102. https://gcc.gnu.org/PR15875
17681 103. https://gcc.gnu.org/PR15877
17682 104. https://gcc.gnu.org/PR15947
17683 105. https://gcc.gnu.org/PR16020
17684 106. https://gcc.gnu.org/PR16154
17685 107. https://gcc.gnu.org/PR16174
17686 108. https://gcc.gnu.org/PR14315
17687 109. https://gcc.gnu.org/PR15151
17688 110. https://gcc.gnu.org/PR7993
17689 111. https://gcc.gnu.org/PR15228
17690 112. https://gcc.gnu.org/PR15345
17691 113. https://gcc.gnu.org/PR15945
17692 114. https://gcc.gnu.org/PR15526
17693 115. https://gcc.gnu.org/PR14690
17694 116. https://gcc.gnu.org/PR15112
17695 117. https://gcc.gnu.org/PR15067
17696 118. https://gcc.gnu.org/PR1963
17697 119. https://gcc.gnu.org/PR15717
17698 120. https://gcc.gnu.org/PR14782
17699 121. https://gcc.gnu.org/PR14828
17700 122. https://gcc.gnu.org/PR15202
17701 123. https://gcc.gnu.org/PR14610
17702 124. https://gcc.gnu.org/PR14813
17703 125. https://gcc.gnu.org/PR14857
17704 126. https://gcc.gnu.org/PR15598
17705 127. https://gcc.gnu.org/PR15653
17706 128. https://gcc.gnu.org/PR15189
17707 129. https://gcc.gnu.org/PR15331
17708 130. https://gcc.gnu.org/PR16144
17709 131. https://gcc.gnu.org/PR16176
17710 132. https://gcc.gnu.org/PR11591
17711 133. https://gcc.gnu.org/PR12028
17712 134. https://gcc.gnu.org/PR14478
17713 135. https://gcc.gnu.org/PR14567
17714 136. https://gcc.gnu.org/PR14715
17715 137. https://gcc.gnu.org/PR14902
17716 138. https://gcc.gnu.org/PR14924
17717 139. https://gcc.gnu.org/PR14960
17718 140. https://gcc.gnu.org/PR15106
17719 141. https://gcc.gnu.org/PR16026
17720 142. https://gcc.gnu.org/PR15191
17721 143. https://gcc.gnu.org/PR15662
17722 144. https://gcc.gnu.org/PR15054
17723 145. https://gcc.gnu.org/PR15783
17724 146. https://gcc.gnu.org/PR15626
17725 147. https://gcc.gnu.org/PR14326
17726 148. https://gcc.gnu.org/PR14723
17727 149. https://gcc.gnu.org/PR15290
17728 150. https://gcc.gnu.org/PR15250
17729 151. https://gcc.gnu.org/PR15551
17730 152. https://gcc.gnu.org/PR8309
17731 153. https://gcc.gnu.org/PR13250
17732 154. https://gcc.gnu.org/PR13803
17733 155. https://gcc.gnu.org/PR14093
17734 156. https://gcc.gnu.org/PR14457
17735 157. https://gcc.gnu.org/PR14542
17736 158. https://gcc.gnu.org/PR15100
17737 159. https://gcc.gnu.org/PR15296
17738 160. https://gcc.gnu.org/PR15396
17739 161. https://gcc.gnu.org/PR15782
17740 162. https://gcc.gnu.org/PR11610
17741 163. https://gcc.gnu.org/PR15488
17742 164. https://gcc.gnu.org/PR15489
17743 165. https://gcc.gnu.org/PR13928
17744 166. https://gcc.gnu.org/PR14150
17745 167. https://gcc.gnu.org/PR14949
17746 168. https://gcc.gnu.org/PR15123
17747 169. https://gcc.gnu.org/PR16469
17748 170. https://gcc.gnu.org/PR16344
17749 171. https://gcc.gnu.org/PR16842
17750 172. https://gcc.gnu.org/PR12608
17751 173. https://gcc.gnu.org/PR14492
17752 174. https://gcc.gnu.org/PR15461
17753 175. https://gcc.gnu.org/PR15890
17754 176. https://gcc.gnu.org/PR16180
17755 177. https://gcc.gnu.org/PR16224
17756 178. https://gcc.gnu.org/PR16408
17757 179. https://gcc.gnu.org/PR16529
17758 180. https://gcc.gnu.org/PR16698
17759 181. https://gcc.gnu.org/PR16706
17760 182. https://gcc.gnu.org/PR16810
17761 183. https://gcc.gnu.org/PR16851
17762 184. https://gcc.gnu.org/PR16870
17763 185. https://gcc.gnu.org/PR16904
17764 186. https://gcc.gnu.org/PR16905
17765 187. https://gcc.gnu.org/PR16964
17766 188. https://gcc.gnu.org/PR17068
17767 189. https://gcc.gnu.org/PR16366
17768 190. https://gcc.gnu.org/PR15345
17769 191. https://gcc.gnu.org/PR16590
17770 192. https://gcc.gnu.org/PR16693
17771 193. https://gcc.gnu.org/PR17078
17772 194. https://gcc.gnu.org/PR13956
17773 195. https://gcc.gnu.org/PR16684
17774 196. https://gcc.gnu.org/PR12658
17775 197. https://gcc.gnu.org/PR13092
17776 198. https://gcc.gnu.org/PR15320
17777 199. https://gcc.gnu.org/PR16246
17778 200. https://gcc.gnu.org/PR16273
17779 201. https://gcc.gnu.org/PR16401
17780 202. https://gcc.gnu.org/PR16411
17781 203. https://gcc.gnu.org/PR16489
17782 204. https://gcc.gnu.org/PR16618
17783 205. https://gcc.gnu.org/PR16637
17784 206. https://gcc.gnu.org/PR16717
17785 207. https://gcc.gnu.org/PR16813
17786 208. https://gcc.gnu.org/PR16853
17787 209. https://gcc.gnu.org/PR16889
17788 210. https://gcc.gnu.org/PR16959
17789 211. https://gcc.gnu.org/PR7587
17790 212. https://gcc.gnu.org/PR16473
17791 213. https://gcc.gnu.org/PR16478
17792 214. https://gcc.gnu.org/PR10695
17793 215. https://gcc.gnu.org/PR16974
17794 216. https://gcc.gnu.org/PR16298
17795 217. https://gcc.gnu.org/PR17113
17796 218. https://gcc.gnu.org/PR14697
17797 219. https://gcc.gnu.org/PR15869
17798 220. https://gcc.gnu.org/PR16325
17799 221. https://gcc.gnu.org/PR16357
17800 222. https://gcc.gnu.org/PR16380
17801 223. https://gcc.gnu.org/PR16407
17802 224. https://gcc.gnu.org/PR16643
17803 225. https://gcc.gnu.org/PR15927
17804 226. https://gcc.gnu.org/PR15948
17805 227. https://gcc.gnu.org/PR17019
17806 228. https://gcc.gnu.org/PR16130
17807 229. https://gcc.gnu.org/PR16142
17808 230. https://gcc.gnu.org/PR16278
17809 231. https://gcc.gnu.org/PR16414
17810 232. https://gcc.gnu.org/PR16445
17811 233. https://gcc.gnu.org/PR16490
17812 234. https://gcc.gnu.org/PR16683
17813 235. https://gcc.gnu.org/PR16195
17814 236. https://gcc.gnu.org/PR16239
17815 237. https://gcc.gnu.org/PR16199
17816 238. https://gcc.gnu.org/PR16416
17817 239. https://gcc.gnu.org/PR16430
17818 240. https://gcc.gnu.org/PR16379
17819 241. https://gcc.gnu.org/PR17093
17820 242. https://gcc.gnu.org/PR17119
17821 243. https://gcc.gnu.org/PR15928
17822 244. https://gcc.gnu.org/PR16210
17823 245. https://gcc.gnu.org/PR15488
17824 246. https://gcc.gnu.org/PR16250
17825 247. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3
17826 248. https://gcc.gnu.org/PR17369
17827 249. https://gcc.gnu.org/PR17850
17828 250. https://gcc.gnu.org/PR13948
17829 251. https://gcc.gnu.org/PR14492
17830 252. https://gcc.gnu.org/PR16301
17831 253. https://gcc.gnu.org/PR16566
17832 254. https://gcc.gnu.org/PR17023
17833 255. https://gcc.gnu.org/PR17027
17834 256. https://gcc.gnu.org/PR17524
17835 257. https://gcc.gnu.org/PR17826
17836 258. https://gcc.gnu.org/PR15526
17837 259. https://gcc.gnu.org/PR16999
17838 260. https://gcc.gnu.org/PR17503
17839 261. https://gcc.gnu.org/PR17581
17840 262. https://gcc.gnu.org/PR18129
17841 263. https://gcc.gnu.org/PR10975
17842 264. https://gcc.gnu.org/PR11722
17843 265. https://gcc.gnu.org/PR14534
17844 266. https://gcc.gnu.org/PR15172
17845 267. https://gcc.gnu.org/PR15786
17846 268. https://gcc.gnu.org/PR16162
17847 269. https://gcc.gnu.org/PR16612
17848 270. https://gcc.gnu.org/PR16715
17849 271. https://gcc.gnu.org/PR16848
17850 272. https://gcc.gnu.org/PR17132
17851 273. https://gcc.gnu.org/PR17259
17852 274. https://gcc.gnu.org/PR17327
17853 275. https://gcc.gnu.org/PR17393
17854 276. https://gcc.gnu.org/PR17501
17855 277. https://gcc.gnu.org/PR17537
17856 278. https://gcc.gnu.org/PR17585
17857 279. https://gcc.gnu.org/PR17821
17858 280. https://gcc.gnu.org/PR17829
17859 281. https://gcc.gnu.org/PR17851
17860 282. https://gcc.gnu.org/PR17976
17861 283. https://gcc.gnu.org/PR18020
17862 284. https://gcc.gnu.org/PR18093
17863 285. https://gcc.gnu.org/PR18140
17864 286. https://gcc.gnu.org/PR17541
17865 287. https://gcc.gnu.org/PR17853
17866 288. https://gcc.gnu.org/PR17245
17867 289. https://gcc.gnu.org/PR17167
17868 290. https://gcc.gnu.org/PR17277
17869 291. https://gcc.gnu.org/PR17505
17870 292. https://gcc.gnu.org/PR17684
17871 293. https://gcc.gnu.org/PR17384
17872 294. https://gcc.gnu.org/PR17770
17873 295. https://gcc.gnu.org/PR11476
17874 296. https://gcc.gnu.org/PR14064
17875 297. https://gcc.gnu.org/PR14678
17876 298. https://gcc.gnu.org/PR15583
17877 299. https://gcc.gnu.org/PR15790
17878 300. https://gcc.gnu.org/PR15886
17879 301. https://gcc.gnu.org/PR16884
17880 302. https://gcc.gnu.org/PR13841
17881 303. https://gcc.gnu.org/PR15860
17882 304. https://gcc.gnu.org/PR17465
17883 305. https://gcc.gnu.org/PR17469
17884 306. https://gcc.gnu.org/PR18138
17885 307. https://gcc.gnu.org/PR15498
17886 308. https://gcc.gnu.org/PR15747
17887 309. https://gcc.gnu.org/PR16406
17888 310. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4
17889 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5
17890 312. https://gcc.gnu.org/PR24688
17891 313. https://gcc.gnu.org/PR17188
17892 314. https://gcc.gnu.org/PR20187
17893 315. https://gcc.gnu.org/PR21873
17894 316. https://gcc.gnu.org/PR21899
17895 317. https://gcc.gnu.org/PR22061
17896 318. https://gcc.gnu.org/PR22208
17897 319. https://gcc.gnu.org/PR22458
17898 320. https://gcc.gnu.org/PR22589
17899 321. https://gcc.gnu.org/PR24101
17900 322. https://gcc.gnu.org/PR10611
17901 323. https://gcc.gnu.org/PR13377
17902 324. https://gcc.gnu.org/PR16002
17903 325. https://gcc.gnu.org/PR17413
17904 326. https://gcc.gnu.org/PR17609
17905 327. https://gcc.gnu.org/PR17618
17906 328. https://gcc.gnu.org/PR18124
17907 329. https://gcc.gnu.org/PR18155
17908 330. https://gcc.gnu.org/PR18177
17909 331. https://gcc.gnu.org/PR18368
17910 332. https://gcc.gnu.org/PR18378
17911 333. https://gcc.gnu.org/PR18466
17912 334. https://gcc.gnu.org/PR18512
17913 335. https://gcc.gnu.org/PR18545
17914 336. https://gcc.gnu.org/PR18738
17915 337. https://gcc.gnu.org/PR18803
17916 338. https://gcc.gnu.org/PR19004
17917 339. https://gcc.gnu.org/PR19208
17918 340. https://gcc.gnu.org/PR19253
17919 341. https://gcc.gnu.org/PR19608
17920 342. https://gcc.gnu.org/PR19884
17921 343. https://gcc.gnu.org/PR20153
17922 344. https://gcc.gnu.org/PR20563
17923 345. https://gcc.gnu.org/PR20789
17924 346. https://gcc.gnu.org/PR21336
17925 347. https://gcc.gnu.org/PR21768
17926 348. https://gcc.gnu.org/PR21853
17927 349. https://gcc.gnu.org/PR21903
17928 350. https://gcc.gnu.org/PR21983
17929 351. https://gcc.gnu.org/PR21987
17930 352. https://gcc.gnu.org/PR22153
17931 353. https://gcc.gnu.org/PR22172
17932 354. https://gcc.gnu.org/PR21286
17933 355. https://gcc.gnu.org/PR22233
17934 356. https://gcc.gnu.org/PR22508
17935 357. https://gcc.gnu.org/PR22545
17936 358. https://gcc.gnu.org/PR23528
17937 359. https://gcc.gnu.org/PR23550
17938 360. https://gcc.gnu.org/PR23586
17939 361. https://gcc.gnu.org/PR23624
17940 362. https://gcc.gnu.org/PR23639
17941 363. https://gcc.gnu.org/PR23797
17942 364. https://gcc.gnu.org/PR23965
17943 365. https://gcc.gnu.org/PR24052
17944 366. https://gcc.gnu.org/PR24580
17945 367. https://gcc.gnu.org/PR24267
17946 368. https://gcc.gnu.org/PR17810
17947 369. https://gcc.gnu.org/PR17860
17948 370. https://gcc.gnu.org/PR21709
17949 371. https://gcc.gnu.org/PR21964
17950 372. https://gcc.gnu.org/PR22167
17951 373. https://gcc.gnu.org/PR22619
17952 374. https://gcc.gnu.org/PR23241
17953 375. https://gcc.gnu.org/PR23478
17954 376. https://gcc.gnu.org/PR24470
17955 377. https://gcc.gnu.org/PR24950
17956 378. https://gcc.gnu.org/PR14400
17957 379. https://gcc.gnu.org/PR14940
17958 380. https://gcc.gnu.org/PR20239
17959 381. https://gcc.gnu.org/PR15220
17960 382. https://gcc.gnu.org/PR19275
17961 383. https://gcc.gnu.org/PR21888
17962 384. https://gcc.gnu.org/PR15342
17963 385. https://gcc.gnu.org/PR23985
17964 386. https://gcc.gnu.org/PR16719
17965 387. https://gcc.gnu.org/PR21723
17966 388. https://gcc.gnu.org/PR21841
17967 389. https://gcc.gnu.org/PR23644
17968 390. https://gcc.gnu.org/PR24718
17969 391. https://gcc.gnu.org/PR18421
17970 392. https://gcc.gnu.org/PR20621
17971 393. https://gcc.gnu.org/PR18583
17972 394. https://gcc.gnu.org/PR20191
17973 395. https://gcc.gnu.org/PR22083
17974 396. https://gcc.gnu.org/PR23070
17975 397. https://gcc.gnu.org/PR23404
17976 398. https://gcc.gnu.org/PR23539
17977 399. https://gcc.gnu.org/PR24102
17978 400. https://gcc.gnu.org/PR24465
17979 401. https://gcc.gnu.org/PR19933
17980 402. https://gcc.gnu.org/PR21889
17981 403. https://gcc.gnu.org/PR19300
17982 404. https://gcc.gnu.org/PR20301
17983 405. https://gcc.gnu.org/PR20673
17984 406. https://gcc.gnu.org/PR18582
17985 407. https://gcc.gnu.org/PR19340
17986 408. https://gcc.gnu.org/PR21716
17987 409. https://gcc.gnu.org/PR24315
17988 410. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6
17989 411. https://gcc.gnu.org/onlinedocs/
17990 412. mailto:gcc-help@gcc.gnu.org
17991 413. mailto:gcc@gcc.gnu.org
17992 414. https://gcc.gnu.org/lists.html
17993 415. https://www.fsf.org/
17994 416. https://gcc.gnu.org/about.html
17995======================================================================
17996http://gcc.gnu.org/gcc-3.3/index.html
17997
17998                             GCC 3.3 Release Series
17999
18000   (This release series is no longer supported.)
18001
18002   May 03, 2005
18003
18004   The [1]GNU project and the GCC developers are pleased to announce the
18005   release of GCC 3.3.6.
18006
18007   This release is a bug-fix release, containing fixes for regressions in
18008   GCC 3.3.5 relative to previous releases of GCC.
18009
18010   This release is the last of the series 3.3.x.
18011
18012   The GCC 3.3 release series includes numerous [2]new features,
18013   improvements, bug fixes, and other changes, thanks to an [3]amazing
18014   group of volunteers.
18015
18016Release History
18017
18018   GCC 3.3.6
18019          May 3, 2005 ([4]changes)
18020
18021   GCC 3.3.5
18022          September 30, 2004 ([5]changes)
18023
18024   GCC 3.3.4
18025          May 31, 2004 ([6]changes)
18026
18027   GCC 3.3.3
18028          February 14, 2004 ([7]changes)
18029
18030   GCC 3.3.2
18031          October 16, 2003 ([8]changes)
18032
18033   GCC 3.3.1
18034          August 8, 2003 ([9]changes)
18035
18036   GCC 3.3
18037          May 14, 2003 ([10]changes)
18038
18039References and Acknowledgements
18040
18041   GCC used to stand for the GNU C Compiler, but since the compiler
18042   supports several other languages aside from C, it now stands for the
18043   GNU Compiler Collection.
18044
18045   A list of [11]successful builds is updated as new information becomes
18046   available.
18047
18048   The GCC developers would like to thank the numerous people that have
18049   contributed new features, improvements, bug fixes, and other changes as
18050   well as test results to GCC. This [12]amazing group of volunteers is
18051   what makes GCC successful.
18052
18053   For additional information about GCC please refer to the [13]GCC
18054   project web site or contact the [14]GCC development mailing list.
18055
18056   To obtain GCC please use [15]our mirror sites, or our CVS server.
18057
18058
18059    For questions related to the use of GCC, please consult these web
18060    pages and the [16]GCC manuals. If that fails, the
18061    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
18062    web pages and the development of GCC are welcome on our developer
18063    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
18064    archives.
18065
18066   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
18067   distribution of this entire article is permitted in any medium,
18068   provided this notice is preserved.
18069
18070   These pages are [21]maintained by the GCC team. Last modified
18071   2022-10-26.
18072
18073References
18074
18075   1. http://www.gnu.org/
18076   2. http://gcc.gnu.org/gcc-3.3/changes.html
18077   3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
18078   4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
18079   5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5
18080   6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4
18081   7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3
18082   8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2
18083   9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1
18084  10. http://gcc.gnu.org/gcc-3.3/changes.html
18085  11. http://gcc.gnu.org/gcc-3.3/buildstat.html
18086  12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
18087  13. http://gcc.gnu.org/index.html
18088  14. mailto:gcc@gcc.gnu.org
18089  15. http://gcc.gnu.org/mirrors.html
18090  16. https://gcc.gnu.org/onlinedocs/
18091  17. mailto:gcc-help@gcc.gnu.org
18092  18. mailto:gcc@gcc.gnu.org
18093  19. https://gcc.gnu.org/lists.html
18094  20. https://www.fsf.org/
18095  21. https://gcc.gnu.org/about.html
18096======================================================================
18097http://gcc.gnu.org/gcc-3.3/changes.html
18098
18099                             GCC 3.3 Release Series
18100                        Changes, New Features, and Fixes
18101
18102   The latest release in the 3.3 release series is [1]GCC 3.3.6.
18103
18104Caveats
18105
18106     * The preprocessor no longer accepts multi-line string literals. They
18107       were deprecated in 3.0, 3.1, and 3.2.
18108     * The preprocessor no longer supports the -A- switch when appearing
18109       alone. -A- followed by an assertion is still supported.
18110     * Support for all the systems [2]obsoleted in GCC 3.1 has been
18111       removed from GCC 3.3. See below for a [3]list of systems which are
18112       obsoleted in this release.
18113     * Checking for null format arguments has been decoupled from the rest
18114       of the format checking mechanism. Programs which use the format
18115       attribute may regain this functionality by using the new [4]nonnull
18116       function attribute. Note that all functions for which GCC has a
18117       built-in format attribute, an appropriate built-in nonnull
18118       attribute is also applied.
18119     * The DWARF (version 1) debugging format has been deprecated and will
18120       be removed in a future version of GCC. Version 2 of the DWARF
18121       debugging format will continue to be supported for the foreseeable
18122       future.
18123     * The C and Objective-C compilers no longer accept the "Naming Types"
18124       extension (typedef foo = bar); it was already unavailable in C++.
18125       Code which uses it will need to be changed to use the "typeof"
18126       extension instead: typedef typeof(bar) foo. (We have removed this
18127       extension without a period of deprecation because it has caused the
18128       compiler to crash since version 3.0 and no one noticed until very
18129       recently. Thus we conclude it is not in widespread use.)
18130     * The -traditional C compiler option has been removed. It was
18131       deprecated in 3.1 and 3.2. (Traditional preprocessing remains
18132       available.) The <varargs.h> header, used for writing variadic
18133       functions in traditional C, still exists but will produce an error
18134       message if used.
18135     * GCC 3.3.1 automatically places zero-initialized variables in the
18136       .bss section on some operating systems. Versions of GNU Emacs up to
18137       (and including) 21.3 will not work correctly when using this
18138       optimization; you can use -fno-zero-initialized-in-bss to disable
18139       it.
18140
18141General Optimizer Improvements
18142
18143     * A new scheme for accurately describing processor pipelines, the
18144       [5]DFA scheduler, has been added.
18145     * Pavel Nejedly, Charles University Prague, has contributed new file
18146       format used by the edge coverage profiler (-fprofile-arcs).
18147       The new format is robust and diagnoses common mistakes where
18148       profiles from different versions (or compilations) of the program
18149       are combined resulting in nonsensical profiles and slow code to
18150       produced with profile feedback. Additionally this format allows
18151       extra data to be gathered. Currently, overall statistics are
18152       produced helping optimizers to identify hot spots of a program
18153       globally replacing the old intra-procedural scheme and resulting in
18154       better code. Note that the gcov tool from older GCC versions will
18155       not be able to parse the profiles generated by GCC 3.3 and vice
18156       versa.
18157     * Jan Hubicka, SuSE Labs, has contributed a new superblock formation
18158       pass enabled using -ftracer. This pass simplifies the control flow
18159       of functions allowing other optimizations to do better job.
18160       He also contributed the function reordering pass
18161       (-freorder-functions) to optimize function placement using profile
18162       feedback.
18163
18164New Languages and Language specific improvements
18165
18166  C/ObjC/C++
18167
18168     * The preprocessor now accepts directives within macro arguments. It
18169       processes them just as if they had not been within macro arguments.
18170     * The separate ISO and traditional preprocessors have been completely
18171       removed. The front end handles either type of preprocessed output
18172       if necessary.
18173     * In C99 mode preprocessor arithmetic is done in the precision of the
18174       target's intmax_t, as required by that standard.
18175     * The preprocessor can now copy comments inside macros to the output
18176       file when the macro is expanded. This feature, enabled using the
18177       -CC option, is intended for use by applications which place
18178       metadata or directives inside comments, such as lint.
18179     * The method of constructing the list of directories to be searched
18180       for header files has been revised. If a directory named by a -I
18181       option is a standard system include directory, the option is
18182       ignored to ensure that the default search order for system
18183       directories and the special treatment of system header files are
18184       not defeated.
18185     * A few more [6]ISO C99 features now work correctly.
18186     * A new function attribute, nonnull, has been added which allows
18187       pointer arguments to functions to be specified as requiring a
18188       non-null value. The compiler currently uses this information to
18189       issue a warning when it detects a null value passed in such an
18190       argument slot.
18191     * A new type attribute, may_alias, has been added. Accesses to
18192       objects with types with this attribute are not subjected to
18193       type-based alias analysis, but are instead assumed to be able to
18194       alias any other type of objects, just like the char type.
18195
18196  C++
18197
18198     * Type based alias analysis has been implemented for C++ aggregate
18199       types.
18200
18201  Objective-C
18202
18203     * Generate an error if Objective-C objects are passed by value in
18204       function and method calls.
18205     * When -Wselector is used, check the whole list of selectors at the
18206       end of compilation, and emit a warning if a @selector() is not
18207       known.
18208     * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime.
18209     * No longer need to include objc/objc-class.h to compile self calls
18210       in class methods (NeXT runtime only).
18211     * New -Wundeclared-selector option.
18212     * Removed selector bloating which was causing object files to be 10%
18213       bigger on average (GNU runtime only).
18214     * Using at run time @protocol() objects has been fixed in certain
18215       situations (GNU runtime only).
18216     * Type checking has been fixed and improved in many situations
18217       involving protocols.
18218
18219  Java
18220
18221     * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK
18222       1.4) API.
18223     * The JDK 1.4 assert facility has been implemented.
18224     * The bytecode interpreter is now direct threaded and thus faster.
18225
18226  Fortran
18227
18228     * Fortran improvements are listed in [7]the Fortran documentation.
18229
18230  Ada
18231
18232     * Ada tasking now works with glibc 2.3.x threading libraries.
18233
18234New Targets and Target Specific Improvements
18235
18236     * The following changes have been made to the HP-PA port:
18237          + The port now defaults to scheduling for the PA8000 series of
18238            processors.
18239          + Scheduling support for the PA7300 processor has been added.
18240          + The 32-bit port now supports weak symbols under HP-UX 11.
18241          + The handling of initializers and finalizers has been improved
18242            under HP-UX 11. The 64-bit port no longer uses collect2.
18243          + Dwarf2 EH support has been added to the 32-bit GNU/Linux port.
18244          + ABI fixes to correct the passing of small structures by value.
18245     * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to
18246       use the DFA processor pipeline description.
18247     * The following NetBSD configurations for the SuperH processor family
18248       have been added:
18249          + SH3, big-endian, sh-*-netbsdelf*
18250          + SH3, little-endian, shle-*-netbsdelf*
18251          + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd*
18252          + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd*
18253          + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd*
18254          + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd*
18255     * The following changes have been made to the IA-32/x86-64 port:
18256          + SSE2 and 3dNOW! intrinsics are now supported.
18257          + Support for thread local storage has been added to the IA-32
18258            and x86-64 ports.
18259          + The x86-64 port has been significantly improved.
18260     * The following changes have been made to the MIPS port:
18261          + All configurations now accept the -mabi switch. Note that you
18262            will need appropriate multilibs for this option to work
18263            properly.
18264          + ELF configurations will always pass an ABI flag to the
18265            assembler, except when the MIPS EABI is selected.
18266          + -mabi=64 no longer selects MIPS IV code.
18267          + The -mcpu option, which was deprecated in 3.1 and 3.2, has
18268            been removed from this release.
18269          + -march now changes the core ISA level. In previous releases,
18270            it would change the use of processor-specific extensions, but
18271            would leave the core ISA unchanged. For example, mips64-elf
18272            -march=r8000 will now generate MIPS IV code.
18273          + Under most configurations, -mipsN now acts as a synonym for
18274            -march.
18275          + There are some new preprocessor macros to describe the -march
18276            and -mtune settings. See the documentation of those options
18277            for details.
18278          + Support for the NEC VR-Series processors has been added. This
18279            includes the 54xx, 5500, and 41xx series.
18280          + Support for the Sandcraft sr71k processor has been added.
18281     * The following changes have been made to the S/390 port:
18282          + Support to build the Java runtime libraries has been added.
18283            Java is now enabled by default on s390-*-linux* and
18284            s390x-*-linux* targets.
18285          + Multilib support for the s390x-*-linux* target has been added;
18286            this allows to build 31-bit binaries using the -m31 option.
18287          + Support for thread local storage has been added.
18288          + Inline assembler code may now use the 'Q' constraint to
18289            specify memory operands without index register.
18290          + Various platform-specific performance improvements have been
18291            implemented; in particular, the compiler now uses the BRANCH
18292            ON COUNT family of instructions and makes more frequent use of
18293            the TEST UNDER MASK family of instructions.
18294     * The following changes have been made to the PowerPC port:
18295          + Support for IBM Power4 processor added.
18296          + Support for Motorola e500 SPE added.
18297          + Support for AIX 5.2 added.
18298          + Function and Data sections now supported on AIX.
18299          + Sibcall optimizations added.
18300     * The support for H8 Tiny is added to the H8/300 port with -mn.
18301
18302Obsolete Systems
18303
18304   Support for a number of older systems has been declared obsolete in GCC
18305   3.3. Unless there is activity to revive them, the next release of GCC
18306   will have their sources permanently removed.
18307
18308   All configurations of the following processor architectures have been
18309   declared obsolete:
18310     * Matsushita MN10200, mn10200-*-*
18311     * Motorola 88000, m88k-*-*
18312     * IBM ROMP, romp-*-*
18313
18314   Also, some individual systems have been obsoleted:
18315     * Alpha
18316          + Interix, alpha*-*-interix*
18317          + Linux libc1, alpha*-*-linux*libc1*
18318          + Linux ECOFF, alpha*-*-linux*ecoff*
18319     * ARM
18320          + Generic a.out, arm*-*-aout*
18321          + Conix, arm*-*-conix*
18322          + "Old ABI," arm*-*-oabi
18323          + StrongARM/COFF, strongarm-*-coff*
18324     * HPPA (PA-RISC)
18325          + Generic OSF, hppa1.0-*-osf*
18326          + Generic BSD, hppa1.0-*-bsd*
18327          + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]*
18328          + HiUX, hppa*-*-hiux*
18329          + Mach Lites, hppa*-*-lites*
18330     * Intel 386 family
18331          + Windows NT 3.x, i?86-*-win32
18332     * MC68000 family
18333          + HP systems, m68000-hp-bsd* and m68k-hp-bsd*
18334          + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and
18335            m68k-sun-mach*
18336          + AT&T systems, m68000-att-sysv*
18337          + Atari systems, m68k-atari-sysv*
18338          + Motorola systems, m68k-motorola-sysv*
18339          + NCR systems, m68k-ncr-sysv*
18340          + Plexus systems, m68k-plexus-sysv*
18341          + Commodore systems, m68k-cbm-sysv*
18342          + Citicorp TTI, m68k-tti-*
18343          + Unos, m68k-crds-unos*
18344          + Concurrent RTU, m68k-ccur-rtu*
18345          + Linux a.out, m68k-*-linux*aout*
18346          + Linux libc1, m68k-*-linux*libc1*
18347          + pSOS, m68k-*-psos*
18348     * MIPS
18349          + Generic ECOFF, mips*-*-ecoff*
18350          + SINIX, mips-sni-sysv4
18351          + Orion RTEMS, mips64orion-*-rtems*
18352     * National Semiconductor 32000
18353          + OpenBSD, ns32k-*-openbsd*
18354     * POWER (aka RS/6000) and PowerPC
18355          + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]*
18356          + Bull BOSX, rs6000-bull-bosx
18357          + Generic Mach, rs6000-*-mach*
18358          + Generic SysV, powerpc*-*-sysv*
18359          + Linux libc1, powerpc*-*-linux*libc1*
18360     * Sun SPARC
18361          + Generic a.out, sparc-*-aout*, sparclet-*-aout*,
18362            sparclite-*-aout*, and sparc86x-*-aout*
18363          + NetBSD a.out, sparc-*-netbsd*aout*
18364          + Generic BSD, sparc-*-bsd*
18365          + ChorusOS, sparc-*-chorusos*
18366          + Linux a.out, sparc-*-linux*aout*
18367          + Linux libc1, sparc-*-linux*libc1*
18368          + LynxOS, sparc-*-lynxos*
18369          + Solaris on HAL hardware, sparc-hal-solaris2*
18370          + SunOS versions 3 and 4, sparc-*-sunos[34]*
18371     * NEC V850
18372          + RTEMS, v850-*-rtems*
18373     * VAX
18374          + VMS, vax-*-vms*
18375
18376Documentation improvements
18377
18378Other significant improvements
18379
18380     * Almost all front-end dependencies in the compiler have been
18381       separated out into a set of language hooks. This should make adding
18382       a new front end clearer and easier.
18383     * One effect of removing the separate preprocessor is a small
18384       increase in the robustness of the compiler in general, and the
18385       maintainability of target descriptions. Previously target-specific
18386       built-in macros and others, such as __FAST_MATH__, had to be
18387       handled with so-called specs that were hard to maintain. Often they
18388       would fail to behave properly when conflicting options were
18389       supplied on the command line, and define macros in the user's
18390       namespace even when strict ISO compliance was requested.
18391       Integrating the preprocessor has cleanly solved these issues.
18392     * The Makefile suite now supports redirection of make install by
18393       means of the variable DESTDIR.
18394     __________________________________________________________________
18395
18396GCC 3.3
18397
18398   Detailed release notes for the GCC 3.3 release follow.
18399
18400  Bug Fixes
18401
18402    bootstrap failures
18403
18404     * [8]10140 cross compiler build failures: missing __mempcpy (DUP:
18405       [9]10198,[10]10338)
18406
18407    Internal compiler errors (multi-platform)
18408
18409     * [11]3581 large string causes segmentation fault in cc1
18410     * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler
18411     * [13]5533 (c++) ICE when processing std::accumulate(begin, end,
18412       init, invalid_op)
18413     * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out
18414     * [15]6412 (c++) ICE in retrieve_specialization
18415     * [16]6620 (c++) partial template specialization causes an ICE
18416       (segmentation fault)
18417     * [17]6663 (c++) ICE with attribute aligned
18418     * [18]7068 ICE with incomplete types
18419     * [19]7083 (c++) ICE using -gstabs with dodgy class derivation
18420     * [20]7647 (c++) ICE when data member has the name of the enclosing
18421       class
18422     * [21]7675 ICE in fixup_var_refs_1
18423     * [22]7718 'complex' template instantiation causes ICE
18424     * [23]8116 (c++) ICE in member template function
18425     * [24]8358 (ada) Ada compiler accesses freed memory, crashes
18426     * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation
18427       fault
18428     * [26]8564 (c++) ICE in find_function_data, in function.c
18429     * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c
18430     * [28]8766 (c++) ICE after failed initialization of static template
18431       variable
18432     * [29]8803 ICE in instantiate_virtual_regs_1, in function.c
18433     * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set
18434     * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class
18435       definition
18436     * [32]9216 (c++) ICE on missing template parameter
18437     * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c
18438     * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO
18439       loop
18440     * [35]9429 (c++) ICE in template instantiation with a pointered new
18441       operator
18442     * [36]9516 Internal error when using a big array
18443     * [37]9600 (c++) ICE with typedefs in template class
18444     * [38]9629 (c++) virtual inheritance segfault
18445     * [39]9672 (c++) ICE: Error reporting routines re-entered
18446     * [40]9749 (c++) ICE in write_expression on invalid function
18447       prototype
18448     * [41]9794 (fortran) ICE: floating point exception during constant
18449       folding
18450     * [42]9829 (c++) Missing colon in nested namespace usage causes ICE
18451     * [43]9916 (c++) ICE with noreturn function in ?: statement
18452     * [44]9936 ICE with local function and variable-length 2d array
18453     * [45]10262 (c++) cc1plus crashes with large generated code
18454     * [46]10278 (c++) ICE in parser for invalid code
18455     * [47]10446 (c++) ICE on definition of nonexistent member function of
18456       nested class in a class template
18457     * [48]10451 (c++) ICE in grokdeclarator on spurious mutable
18458       declaration
18459     * [49]10506 (c++) ICE in build_new at cp/init.c with
18460       -fkeep-inline-functions and multiple inheritance
18461     * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the
18462       precision of the declared type
18463
18464    Optimization bugs
18465
18466     * [51]2001 Inordinately long compile times in reload CSE regs
18467     * [52]2391 Exponential compilation time explosion in combine
18468     * [53]2960 Duplicate loop conditions even with -Os
18469     * [54]4046 redundant conditional branch
18470     * [55]6405 Loop-unrolling related performance regressions
18471     * [56]6798 very long compile time with large case-statement
18472     * [57]6871 const objects shouldn't be moved to .bss
18473     * [58]6909 problem w/ -Os on modified loop-2c.c test case
18474     * [59]7189 gcc -O2 -Wall does not print ``control reaches end of
18475       non-void function'' warning
18476     * [60]7642 optimization problem with signbit()
18477     * [61]8634 incorrect code for inlining of memcpy under -O2
18478     * [62]8750 Cygwin prolog generation erroneously emitting __alloca as
18479       regular function call
18480
18481    C front end
18482
18483     * [63]2161 long if-else cascade overflows parser stack
18484     * [64]4319 short accepted on typedef'd char
18485     * [65]8602 incorrect line numbers in warning messages when using
18486       inline functions
18487     * [66]9177 -fdump-translation-unit: C front end deletes function_decl
18488       AST nodes and breaks debugging dumps
18489     * [67]9853 miscompilation of non-constant structure initializer
18490
18491    c++ compiler and library
18492
18493     * [68]45 legal template specialization code is rejected (DUP:
18494       [69]3784)
18495     * [70]764 lookup failure: friend operator and dereferencing a pointer
18496       and templates (DUP: [71]5116)
18497     * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP:
18498       2863)
18499     * [73]3663 G++ doesn't check access control during template
18500       instantiation
18501     * [74]3797 gcc fails to emit explicit specialization of a template
18502       member
18503     * [75]3948 Two destructors are called when no copy destructor is
18504       defined (ABI change)
18505     * [76]4137 Conversion operator within template is not accepted
18506     * [77]4361 bogus ambiguity taking the address of a member template
18507     * [78]4802 g++ accepts illegal template code (access to private
18508       member; DUP: [79]5837)
18509     * [80]4803 inline function is used but never defined, and g++ does
18510       not object
18511     * [81]5094 Partial specialization cannot be friend?
18512     * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66
18513     * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run
18514       time
18515     * [84]7015 certain __asm__ constructs rejected
18516     * [85]7086 compile time regression (quadratic behavior in
18517       fixup_var_refs)
18518     * [86]7099 G++ doesn't set the noreturn attribute on std::exit and
18519       std::abort
18520     * [87]7247 copy constructor missing when inlining enabled (invalid
18521       optimization?)
18522     * [88]7441 string array initialization compilation time regression
18523       from seconds to minutes
18524     * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong
18525     * [90]7804 bad printing of floating point constant in warning message
18526     * [91]8099 Friend classes and template specializations
18527     * [92]8117 member function pointers and multiple inheritance
18528     * [93]8205 using declaration and multiple inheritance
18529     * [94]8645 unnecessary non-zero checks in stl_tree.h
18530     * [95]8724 explicit destructor call for incomplete class allowed
18531     * [96]8805 compile time regression with many member variables
18532     * [97]8691 -O3 and -fno-implicit-templates are incompatible
18533     * [98]8700 unhelpful error message for binding temp to reference
18534     * [99]8724 explicit destructor call for incomplete class allowed
18535     * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems
18536     * [101]9016 Failure to consistently constant fold "constant" C++
18537       objects
18538     * [102]9053 g++ confused about ambiguity of overloaded function
18539       templates
18540     * [103]9152 undefined virtual thunks
18541     * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out
18542     * [105]9297 data corruption due to codegen bug (when copying.)
18543     * [106]9318 i/ostream::operator>>/<<(streambuf*) broken
18544     * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf
18545     * [108]9400 bogus -Wshadow warning: shadowed declaration of this in
18546       local classes
18547     * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters
18548     * [110]9425 filebuf::pbackfail broken (DUP: [111]9439)
18549     * [112]9474 GCC freezes in compiling a weird code mixing <iostream>
18550       and <iostream.h>
18551     * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1)
18552       [114][DR 231]
18553     * [115]9555 ostream inserters fail to set badbit on exception
18554     * [116]9561 ostream inserters rethrow exception of wrong type
18555     * [117]9563 ostream::sentry returns true after a failed preparation
18556     * [118]9582 one-definition rule violation in std::allocator
18557     * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors
18558     * [120]9683 bug in initialization chains for static const variables
18559       from template classes
18560     * [121]9791 -Woverloaded-virtual reports hiding of destructor
18561     * [122]9817 collate::compare doesn't handle nul characters
18562     * [123]9825 filebuf::sputbackc breaks sbumpc
18563     * [124]9826 operator>>(basic_istream, basic_string) fails to compile
18564       with custom traits
18565     * [125]9924 Multiple using statements for builtin functions not
18566       allowed
18567     * [126]9946 destructor is not called for temporary object
18568     * [127]9964 filebuf::close() sometimes fails to close file
18569     * [128]9988 filebuf::overflow writes EOF to file
18570     * [129]10033 optimization breaks polymorphic references w/ typeid
18571       operator
18572     * [130]10097 filebuf::underflow drops characters
18573     * [131]10132 filebuf destructor can throw exceptions
18574     * [132]10180 gcc fails to warn about non-inlined function
18575     * [133]10199 method parametrized by template does not work everywhere
18576     * [134]10300 use of array-new (nothrow) in segfaults on NULL return
18577     * [135]10427 Stack corruption with variable-length automatic arrays
18578       and virtual destructors
18579     * [136]10503 Compilation never stops in fixed_type_or_null
18580
18581    Objective-C
18582
18583     * [137]5956 selectors aren't matched properly when added to the
18584       selector table
18585
18586    Fortran compiler and library
18587
18588     * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't
18589       detect
18590     * [139]3924 g77 generates code that is rejected by GAS if COFF debug
18591       info requested
18592     * [140]5634 doc: explain that configure --prefix=~/... does not work
18593     * [141]6367 multiple repeat counts confuse namelist read into array
18594     * [142]6491 Logical operations error on logicals when using
18595       -fugly-logint
18596     * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C"
18597     * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os
18598       on irix6.5
18599     * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should
18600       assume a direct access file
18601     * [146]7278 g77 "bug"; the executable misbehaves (with -O2
18602       -fno-automatic)
18603     * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows
18604     * [148]7388 Incorrect output with 0-based array of characters
18605     * [149]8587 Double complex zero ** double precision number -> NaN
18606       instead of zero
18607     * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning:
18608       unknown register name line-length-none
18609     * [151]10197 Direct access files not unformatted by default
18610
18611    Java compiler and library
18612
18613     * [152]6005 gcj fails to build rhug on alpha
18614     * [153]6389 System.getProperty("") should always throw an
18615       IllegalArgumentException
18616     * [154]6576 java.util.ResourceBundle.getResource ignores locale
18617     * [155]6652 new java.io.File("").getCanonicalFile() throws exception
18618     * [156]7060 getMethod() doesn't search super interface
18619     * [157]7073 bytecode interpreter gives wrong answer for interface
18620       getSuperclass()
18621     * [158]7180 possible bug in
18622       javax.naming.spi.NamingManager.getPlusPath()
18623     * [159]7416 java.security startup refs "GNU libgcj.security"
18624     * [160]7570 Runtime.exec with null envp: child doesn't inherit parent
18625       env (DUP: [161]7578)
18626     * [162]7611 Internal error while compiling libjava with -O
18627     * [163]7709 NullPointerException in _Jv_ResolvePoolEntry
18628     * [164]7766 ZipInputStream.available returns 0 immediately after
18629       construction
18630     * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public
18631     * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented
18632     * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens'
18633     * [168]8234 ZipInputStream chokes when InputStream.read() returns
18634       small chunks
18635     * [169]8415 reflection bug: exception info for Method
18636     * [170]8481 java.Random.nextInt(int) may return negative
18637     * [171]8593 Error reading GZIPped files with BufferedReader
18638     * [172]8759 java.beans.Introspector has no flushCaches() or
18639       flushFromCaches() methods
18640     * [173]8997 spin() calls Thread.sleep
18641     * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd
18642       instead of the root content of C:
18643     * [175]9254 java::lang::Object::wait(), threads-win32.cc returns
18644       wrong return codes
18645     * [176]9271 Severe bias in java.security.SecureRandom
18646
18647    Ada compiler and library
18648
18649     * [177]6767 make gnatlib-shared fails on -laddr2line
18650     * [178]9911 gnatmake fails to link when GCC configured with
18651       --with-sjlj-exceptions=yes
18652     * [179]10020 Can't bootstrap gcc on AIX with Ada enabled
18653     * [180]10546 Ada tasking not working on Red Hat 9
18654
18655    preprocessor
18656
18657     * [181]7029 preprocessor should ignore #warning with -M
18658
18659    ARM-specific
18660
18661     * [182]2903 [arm] Optimization bug with long long arithmetic
18662     * [183]7873 arm-linux-gcc fails when assigning address to a bit field
18663
18664    FreeBSD-specific
18665
18666     * [184]7680 float functions undefined in math.h/cmath with #define
18667       _XOPEN_SOURCE
18668
18669    HP-UX or HP-PA-specific
18670
18671     * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c
18672     * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to
18673       fputc_unlocked
18674     * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen
18675
18676    m68hc11-specific
18677
18678     * [188]6744 Bad assembler code generated: reference to pseudo
18679       register z
18680     * [189]7361 Internal compiler error in reload_cse_simplify_operands,
18681       in reload1.c
18682
18683    MIPS-specific
18684
18685     * [190]9496 [mips-linux] bug in optimizer?
18686
18687    PowerPC-specific
18688
18689     * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of
18690       space
18691     * [192]8480 reload ICEs for LAPACK code on powerpc64-linux
18692     * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg
18693     * [194]10315 [powerpc] ICE: in extract_insn, in recog.c
18694
18695    SPARC-specific
18696
18697     * [195]10267 (documentation) Wrong build instructions for
18698       *-*-solaris2*
18699
18700    x86-specific (Intel/AMD)
18701
18702     * [196]7916 ICE in instantiate_virtual_register_1
18703     * [197]7926 (c++) i486 instructions in header files make c++ programs
18704       crash on i386
18705     * [198]8555 ICE in gen_split_1231
18706     * [199]8994 ICE with -O -march=pentium4
18707     * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs
18708     * [201]9806 ICE in inline assembly with -fPIC flag
18709     * [202]10077 gcc -msse2 generates movd to move dwords between xmm
18710       regs
18711     * [203]10233 64-bit comparison only comparing bottom 32-bits
18712     * [204]10286 type-punning doesn't work with __m64 and -O
18713     * [205]10308 [x86] ICE with -O -fgcse or -O2
18714     __________________________________________________________________
18715
18716GCC 3.3.1
18717
18718  Bug Fixes
18719
18720   This section lists the problem reports (PRs) from GCC's bug tracking
18721   system that are known to be fixed in the 3.3.1 release. This list might
18722   not be complete (that is, it is possible that some PRs that have been
18723   fixed are not listed here).
18724
18725    Bootstrap failures
18726
18727     * [206]11272 [Solaris] make bootstrap fails while building libstdc++
18728
18729    Internal compiler errors (multi-platform)
18730
18731     * [207]5754 ICE on invalid nested template class
18732     * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64
18733       and --enable-checking
18734     * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c
18735     * [210]7053 (c++) ICE when declaring a function already defined as a
18736       friend method of a template class
18737     * [211]8164 (c++) ICE when using different const expressions as
18738       template parameter
18739     * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c
18740     * [213]9559 (c++) ICE with invalid initialization of a static const
18741     * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c
18742       when redeclaring a static member variable
18743     * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in
18744       dwarfout.c with -g -O -finline-functions
18745     * [216]10432 (c++) ICE in poplevel, in cp/decl.c
18746     * [217]10475 ICE in subreg_highpart_offset for code with long long
18747     * [218]10635 (c++) ICE when dereferencing an incomplete type casted
18748       from a void pointer
18749     * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while
18750       instantiating static member variables
18751     * [220]10700 ICE in copy_to_mode_reg on 64-bit targets
18752     * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c
18753     * [222]10796 (c++) ICE when defining an enum with two values: -1 and
18754       MAX_INT_64BIT
18755     * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x
18756       sched.c
18757     * [224]10939 (c++) ICE with template code
18758     * [225]10956 (c++) ICE when specializing a template member function
18759       of a template class, in tsubst, in cp/pt.c
18760     * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*()
18761       defined)
18762     * [227]11059 (c++) ICE with empty union
18763     * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with
18764       -O2 -fnon-call-exceptions
18765     * [229]11105 (c++) ICE in mangle_conv_op_name_for_type
18766     * [230]11149 (c++) ICE on error when instantiation with call function
18767       of a base type
18768     * [231]11228 (c++) ICE on new-expression using array operator new and
18769       default-initialization
18770     * [232]11282 (c++) Infinite memory usage after syntax error
18771     * [233]11301 (fortran) ICE with -fno-globals
18772     * [234]11308 (c++) ICE when using an enum type name as if it were a
18773       class or namespace
18774     * [235]11473 (c++) ICE with -gstabs when empty struct inherits from
18775       an empty struct
18776     * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR
18777     * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c:
18778       template member functions
18779
18780    Optimization bugs
18781
18782     * [238]11198 -O2 -frename-registers generates wrong code (aliasing
18783       problem)
18784     * [239]11304 Wrong code production with -fomit-frame-pointer
18785     * [240]11381 volatile memory access optimized away
18786     * [241]11536 [strength-reduce] -O2 optimization produces wrong code
18787     * [242]11557 constant folding bug generates wrong code
18788
18789    C front end
18790
18791     * [243]5897 No warning for statement after return
18792     * [244]11279 DWARF-2 output mishandles large enums
18793
18794    Preprocessor bugs
18795
18796     * [245]11022 no warning for non-compatible macro redefinition
18797
18798    C++ compiler and library
18799
18800     * [246]2330 static_cast<>() to a private base is allowed
18801     * [247]5388 Incorrect message "operands to ?: have different types"
18802     * [248]5390 Libiberty fails to demangle multi-digit template
18803       parameters
18804     * [249]7877 Incorrect parameter passing to specializations of member
18805       function templates
18806     * [250]9393 Anonymous namespaces and compiling the same file twice
18807     * [251]10032 -pedantic converts some errors to warnings
18808     * [252]10468 const typeof(x) is non-const, but only in templates
18809     * [253]10527 confused error message with "new int()" parameter
18810       initializer
18811     * [254]10679 parameter MIN_INLINE_INSNS is not honored
18812     * [255]10682 gcc chokes on a typedef for an enum inside a class
18813       template
18814     * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of
18815       0.
18816     * [257]10845 template member function (with nested template as
18817       parameter) cannot be called anymore if another unrelated template
18818       member function is defined
18819     * [258]10849 Cannot define an out-of-class specialization of a
18820       private nested template class
18821     * [259]10888 Suppress -Winline warnings for system headers
18822     * [260]10929 -Winline warns about functions for which no definition
18823       is visible
18824     * [261]10931 valid conversion static_cast<const unsigned
18825       int&>(lvalue-of-type-int) is rejected
18826     * [262]10940 Bad code with explicit specialization
18827     * [263]10968 If member function implicitly instantiated, explicit
18828       instantiation of class fails to instantiate it
18829     * [264]10990 Cannot convert with dynamic_cast<> to a private base
18830       class from within a member function
18831     * [265]11039 Bad interaction between implicit typename deprecation
18832       and friendship
18833     * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say
18834       "__unused__" instead
18835     * [267]11095 C++ iostream manipulator causes segfault when called
18836       with negative argument
18837     * [268]11098 g++ doesn't emit complete debugging information for
18838       local variables in destructors
18839     * [269]11137 GNU/Linux shared library constructors not called unless
18840       there's one global object
18841     * [270]11154 spurious ambiguity report for template class
18842       specialization
18843     * [271]11329 Compiler cannot find user defined implicit typecast
18844     * [272]11332 Spurious error with casts in ?: expression
18845     * [273]11431 static_cast behavior with subclasses when default
18846       constructor available
18847     * [274]11528 money_get facet does not accept "$.00" as valid
18848     * [275]11546 Type lookup problems in out-of-line definition of a
18849       class doubly nested from a template class
18850     * [276]11567 C++ code containing templated member function with same
18851       name as pure virtual member function results in linking failure
18852     * [277]11645 Failure to deal with using and private inheritance
18853
18854    Java compiler and library
18855
18856     * [278]5179 Qualified static field access doesn't initialize its
18857       class
18858     * [279]8204 gcj -O2 to native reorders certain instructions
18859       improperly
18860     * [280]10838 java.io.ObjectInputStream syntax error
18861     * [281]10886 The RMI registry that comes with GCJ does not work
18862       correctly
18863     * [282]11349 JNDI URL context factories not located correctly
18864
18865    x86-specific (Intel/AMD)
18866
18867     * [283]4823 ICE on inline assembly code
18868     * [284]8878 miscompilation with -O and SSE
18869     * [285]9815 (c++ library) atomicity.h - fails to compile with -O3
18870       -masm=intel
18871     * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads,
18872       in reload1.c
18873     * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2
18874     * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source
18875     * [289]11044 [x86] out of range loop instructions for FP code on K6
18876     * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE
18877       built-ins
18878     * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC"
18879       is used
18880
18881    SPARC- or Solaris- specific
18882
18883     * [292]9362 solaris 'as' dies when fed .s and "-gstabs"
18884     * [293]10142 [SPARC64] gcc produces wrong code when passing
18885       structures by value
18886     * [294]10663 New configure check aborts with Sun tools.
18887     * [295]10835 combinatorial explosion in scheduler on HyperSPARC
18888     * [296]10876 ICE in calculate_giv_inc when building KDE
18889     * [297]10955 wrong code at -O3 for structure argument in context of
18890       structure return
18891     * [298]11018 -mcpu=ultrasparc busts tar-1.13.25
18892     * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x
18893       Linux kernel
18894
18895    ia64 specific
18896
18897     * [300]10907 gcc violates the ia64 ABI (GP must be preserved)
18898     * [301]11320 scheduler bug (in machine depended reorganization pass)
18899     * [302]11599 bug with conditional and __builtin_prefetch
18900
18901    PowerPC specific
18902
18903     * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem
18904       during loop)
18905     * [304]10871 error in rs6000_stack_info save_size computation
18906     * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse
18907       cures it
18908
18909    m68k-specific
18910
18911     * [306]7594 [m68k] ICE on legal code associated with simplify-rtx
18912     * [307]10557 [m68k] ICE in subreg_offset_representable_p
18913     * [308]11054 [m68k] ICE in reg_overlap_mentioned_p
18914
18915    ARM-specific
18916
18917     * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for
18918       functions with __attribute__ ((interrupt ("IRQ")))
18919     * [310]10842 [arm] Clobbered link register is copied to pc under
18920       certain circumstances
18921     * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes
18922     * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno
18923       (3.4)
18924
18925    MIPS-specific
18926
18927     * [313]11084 ICE in propagate_one_insn, in flow.c
18928
18929    SH-specific
18930
18931     * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf
18932     * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c
18933     * [316]11096 i686-linux to sh-linux cross compiler fails to compile
18934       C++ files
18935
18936    GNU/Linux (or Hurd?) specific
18937
18938     * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3
18939
18940    UnixWare specific
18941
18942     * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare
18943       7.1.1
18944
18945    Cygwin (or mingw) specific
18946
18947     * [319]5287 ICE with dllimport attribute
18948     * [320]10148 [MingW/CygWin] Compiler dumps core
18949
18950    DJGPP specific
18951
18952     * [321]8787 GCC fails to emit .intel_syntax when invoked with
18953       -masm=intel on DJGPP
18954
18955    Darwin (and MacOS X) specific
18956
18957     * [322]10900 trampolines crash
18958
18959    Documentation
18960
18961     * [323]1607 (c++) Format attributes on methods undocumented
18962     * [324]4252 Invalid option `-fdump-translation-unit'
18963     * [325]4490 Clarify restrictions on -m96bit-long-double,
18964       -m128bit-long-double
18965     * [326]10355 document an issue with regparm attribute on some systems
18966       (e.g. Solaris)
18967     * [327]10726 (fortran) Documentation for function "IDate Intrinsic
18968       (Unix)" is wrong
18969     * [328]10805 document bug in old version of Sun assembler
18970     * [329]10815 warn against GNU binutils on AIX
18971     * [330]10877 document need for newer binutils on i?86-*-linux-gnu
18972     * [331]11280 Manual incorrect with respect to -freorder-blocks
18973     * [332]11466 Document -mlittle-endian and its restrictions for the
18974       sparc64 port
18975
18976    Testsuite bugs (compiler itself is not affected)
18977
18978     * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly
18979       report failure
18980     * [334]10810 gcc-3.3 fails make check: buffer overrun in
18981       test_demangle.c
18982     __________________________________________________________________
18983
18984GCC 3.3.2
18985
18986  Bug Fixes
18987
18988   This section lists the problem reports (PRs) from GCC's bug tracker
18989   that are known to be fixed in the 3.3.2 release. This list might not be
18990   complete (that is, it is possible that some PRs that have been fixed
18991   are not listed here).
18992
18993    Bootstrap failures and problems
18994
18995     * [335]8336 [SCO5] bootstrap config still tries to use COFF options
18996     * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with
18997       --enable-threads=posix
18998     * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap
18999     * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare
19000       7.1.1)
19001     * [339]11687 xstormy16-elf build fails in libf2c
19002     * [340]12263 [SGI IRIX] bootstrap fails during compile of
19003       libf2c/libI77/backspace.c
19004     * [341]12490 buffer overflow in scan-decls.c (during Solaris 9
19005       fix-header processing)
19006
19007    Internal compiler errors (multi-platform)
19008
19009     * [342]7277 Casting integers to vector types causes ICE
19010     * [343]7939 (c++) ICE on invalid function template specialization
19011     * [344]11063 (c++) ICE on parsing initialization list of const array
19012       member
19013     * [345]11207 ICE with negative index in array element designator
19014     * [346]11522 (fortran) g77 dwarf-2 ICE in
19015       add_abstract_origin_attribute
19016     * [347]11595 (c++) ICE on duplicate label definition
19017     * [348]11646 (c++) ICE in commit_one_edge_insertion with
19018       -fnon-call-exceptions -fgcse -O
19019     * [349]11665 ICE in struct initializer when taking address
19020     * [350]11852 (c++) ICE with bad struct initializer.
19021     * [351]11878 (c++) ICE in cp_expr_size
19022     * [352]11883 ICE with any -O on mercury-generated C code
19023     * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in
19024       cp/typeck2.c when applying typeid operator to template template
19025       parameter
19026     * [354]12146 ICE in lookup_template_function, in cp/pt.c
19027     * [355]12215 ICE in make_label_edge with -fnon-call-exceptions
19028       -fno-gcse -O2
19029     * [356]12369 (c++) ICE with templates and friends
19030     * [357]12446 ICE in emit_move_insn on complicated array reference
19031     * [358]12510 ICE in final_scan_insn
19032     * [359]12544 ICE with large parameters used in nested functions
19033
19034    C and optimization bugs
19035
19036     * [360]9862 spurious warnings with -W -finline-functions
19037     * [361]10962 lookup_field is a linear search on a linked list (can be
19038       slow if large struct)
19039     * [362]11370 -Wunreachable-code gives false complaints
19040     * [363]11637 invalid assembly with -fnon-call-exceptions
19041     * [364]11885 Problem with bitfields in packed structs
19042     * [365]12082 Inappropriate unreachable code warnings
19043     * [366]12180 Inline optimization fails for variadic function
19044     * [367]12340 loop unroller + gcse produces wrong code
19045
19046    C++ compiler and library
19047
19048     * [368]3907 nested template parameter collides with member name
19049     * [369]5293 confusing message when binding a temporary to a reference
19050     * [370]5296 [DR115] Pointers to functions and to template functions
19051       behave differently in deduction
19052     * [371]7939 ICE on function template specialization
19053     * [372]8656 Unable to assign function with __attribute__ and pointer
19054       return type to an appropriate variable
19055     * [373]10147 Confusing error message for invalid template function
19056       argument
19057     * [374]11400 std::search_n() makes assumptions about Size parameter
19058     * [375]11409 issues with using declarations, overloading, and
19059       built-in functions
19060     * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle
19061       multiple bits in mask
19062     * [377]11786 operator() call on variable in other namespace not
19063       recognized
19064     * [378]11867 static_cast ignores ambiguity
19065     * [379]11928 bug with conversion operators that are typedefs
19066     * [380]12114 Uninitialized memory accessed in dtor
19067     * [381]12163 static_cast + explicit constructor regression
19068     * [382]12181 Wrong code with comma operator and c++
19069     * [383]12236 regparm and fastcall messes up parameters
19070     * [384]12266 incorrect instantiation of unneeded template during
19071       overload resolution
19072     * [385]12296 istream::peek() doesn't set eofbit
19073     * [386]12298 [sjlj exceptions] Stack unwind destroys
19074       not-yet-constructed object
19075     * [387]12369 ICE with templates and friends
19076     * [388]12337 apparently infinite loop in g++
19077     * [389]12344 stdcall attribute ignored if function returns a pointer
19078     * [390]12451 missing(late) class forward declaration in cxxabi.h
19079     * [391]12486 g++ accepts invalid use of a qualified name
19080
19081    x86 specific (Intel/AMD)
19082
19083     * [392]8869 [x86 MMX] ICE with const variable optimization and MMX
19084       builtins
19085     * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions
19086       -O2
19087     * [394]11689 g++3.3 emits un-assembleable code for k6 architecture
19088     * [395]12116 [k6] Invalid assembly output values with X-MAME code
19089     * [396]12070 ICE converting between double and long double with
19090       -msoft-float
19091
19092    ia64-specific
19093
19094     * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc
19095     * [398]11535 __builtin_return_address may not work on ia64
19096     * [399]11693 [ia64] ICE in gen_nop_type
19097     * [400]12224 [ia64] Thread-local storage doesn't work
19098
19099    PowerPC-specific
19100
19101     * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux
19102       kernel
19103     * [402]11319 loop miscompiled on ppc32
19104     * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code
19105
19106    SPARC-specific
19107
19108     * [404]11662 wrong code for expr. with cast to long long and
19109       exclusive or
19110     * [405]11965 invalid assembler code for a shift < 32 operation
19111     * [406]12301 (c++) stack corruption when a returned expression throws
19112       an exception
19113
19114    Alpha-specific
19115
19116     * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of
19117       kernel 2.4.22-pre8
19118
19119    HPUX-specific
19120
19121     * [408]11313 problem with #pragma weak and static inline functions
19122     * [409]11712 __STDC_EXT__ not defined for C++ by default anymore?
19123
19124    Solaris specific
19125
19126     * [410]12166 Profiled programs crash if PROFDIR is set
19127
19128    Solaris-x86 specific
19129
19130     * [411]12101 i386 Solaris no longer works with GNU as?
19131
19132    Miscellaneous embedded target-specific bugs
19133
19134     * [412]10988 [m32r-elf] wrong blockmove code with -O3
19135     * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with
19136       -O2
19137     * [414]11902 [sh4] spec file improperly inserts rpath even when none
19138       needed
19139     * [415]11903 [sh4] -pthread fails to link due to error in spec file
19140       on sh4
19141     __________________________________________________________________
19142
19143GCC 3.3.3
19144
19145  Minor features
19146
19147   In addition to the bug fixes documented below, this release contains
19148   few minor features such as:
19149     * Support for --with-sysroot
19150     * Support for automatic detection of executable stacks
19151     * Support for SSE3 instructions
19152     * Support for thread local storage debugging under GDB on S390
19153
19154  Bug Fixes
19155
19156   This section lists the problem reports (PRs) from GCC's bug tracker
19157   that are known to be fixed in the 3.3.3 release. This list might not be
19158   complete (that is, it is possible that some PRs that have been fixed
19159   are not listed here).
19160
19161    Bootstrap failures and issues
19162
19163     * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails
19164     * [417]12399 boehm-gc fails (when building a cross compiler): libtool
19165       unable to infer tagged configuration
19166     * [418]13068 mklibgcc.in doesn't handle multi-level multilib
19167       subdirectories properly
19168
19169    Internal compiler errors (multi-platform)
19170
19171     * [419]10060 ICE (stack overflow) on huge file (300k lines) due to
19172       recursive behaviour of copy_rtx_if_shared, in emit_rtl.c
19173     * [420]10555 (c++) ICE on undefined template argument
19174     * [421]10706 (c++) ICE in mangle_class_name_for_template
19175     * [422]11496 (fortran) error in flow_loops_find when -funroll-loops
19176       active
19177     * [423]11741 ICE in pre_insert_copy_insn, in gcse.c
19178     * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0
19179     * [425]12632 (fortran) -fbounds-check ICE
19180     * [426]12712 (c++) ICE on short legit C++ code fragment with gcc
19181       3.3.2
19182     * [427]12726 (c++) ICE (segfault) on trivial code
19183     * [428]12890 (c++) ICE on compilation of class with throwing method
19184     * [429]12900 (c++) ICE in rtl_verify_flow_info_1
19185     * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on
19186       correct code with -O2 -fno-force-mem
19187     * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive
19188       template
19189     * [432]13318 ICE: floating point exception in the loop optimizer
19190     * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in
19191       except.c
19192     * [434]13574 (c++) invalid array default initializer in class lets
19193       gcc consume all memory and die
19194     * [435]13475 ICE on SIMD variables with partial value initialization
19195     * [436]13797 (c++) ICE on invalid template parameter
19196     * [437]13824 (java) gcj SEGV with simple .java program
19197
19198    C and optimization bugs
19199
19200     * [438]8776 loop invariants are not removed (most likely)
19201     * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing
19202       strncmp by memcmp
19203     * [440]11350 undefined labels with -Os -fPIC
19204     * [441]12826 Optimizer removes reference through volatile pointer
19205     * [442]12500 stabs debug info: void no longer a predefined / builtin
19206       type
19207     * [443]12941 builtin-bitops-1.c miscompilation (latent bug)
19208     * [444]12953 tree inliner bug (in inline_forbidden_p) and fix
19209     * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled
19210     * [446]13507 spurious printf format warning
19211     * [447]13382 Type information for const pointer disappears during
19212       optimization.
19213     * [448]13394 noreturn attribute ignored on recursive invokation
19214     * [449]13400 Compiled code crashes storing to read-only location
19215     * [450]13521 Endless loop in calculate_global_regs_live
19216
19217    C++ compiler and library
19218
19219   Some of the bug fixes in this list were made to implement decisions
19220   that the ISO C++ standards committee has made concerning several defect
19221   reports (DRs). Links in the list below point to detailed discussion of
19222   the relevant defect report.
19223     * [451]2094 unimplemented: use of `ptrmem_cst' in template type
19224       unification
19225     * [452]2294 using declaration confusion
19226     * [453]5050 template instantiation depth exceeds limit: recursion
19227       problem?
19228     * [454]9371 Bad exception handling in
19229       i/ostream::operator>>/<<(streambuf*)
19230     * [455]9546 bad exception handling in ostream members
19231     * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the
19232       face of unknown locales
19233     * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work
19234     * [459]10095 istream::operator>>(int&) sets ios::badbit when
19235       ios::failbit is set.
19236     * [460]11554 Warning about reordering of initializers doesn't mention
19237       location of constructor
19238     * [461]12297 istream::sentry::sentry() handles eof() incorrectly.
19239     * [462]12352 Exception safety problems in src/localename.cc
19240     * [463]12438 Memory leak in locale::combine()
19241     * [464]12540 Memory leak in locale::locale(const char*)
19242     * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented
19243     * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented
19244     * [470]12696 memory eating infinite loop in diagnostics (error
19245       recovery problem)
19246     * [471]12815 Code compiled with optimization behaves unexpectedly
19247     * [472]12862 Conflicts between typedefs/enums and namespace member
19248       declarations
19249     * [473]12926 Wrong value after assignment in initialize list using
19250       bit-fields
19251     * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented
19252     * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented
19253     * [478]13007 basic_streambuf::pubimbue, imbue wrong
19254     * [479]13009 Implicitly-defined assignment operator writes to wrong
19255       memory
19256     * [480]13057 regparm attribute not applied to destructor
19257     * [481]13070 -Wformat option ignored in g++
19258     * [482]13081 forward template declarations in <complex> let inlining
19259       fail
19260     * [483]13239 Assertion does not seem to work correctly anymore
19261     * [484]13262 "xxx is private within this context" when initializing a
19262       self-contained template class
19263     * [485]13290 simple typo in concept checking for std::generate_n
19264     * [486]13323 Template code does not compile in presence of typedef
19265     * [487]13369 __verify_grouping (and __add_grouping?) not correct
19266     * [488]13371 infinite loop with packed struct and inlining
19267     * [489]13445 Template argument replacement "dereferences" a typedef
19268     * [490]13461 Fails to access protected-ctor from public constant
19269     * [491]13462 Non-standard-conforming type set::pointer
19270     * [492]13478 gcc uses wrong constructor to initialize a const
19271       reference
19272     * [493]13544 "conflicting types" for enums in different scopes
19273     * [494]13650 string::compare should not (always) use
19274       traits_type::length()
19275     * [495]13683 bogus warning about passing non-PODs through ellipsis
19276     * [496]13688 Derived class is denied access to protected base class
19277       member class
19278     * [497]13774 Member variable cleared in virtual multiple inheritance
19279       class
19280     * [498]13884 Protect sstream.tcc from extern template use
19281
19282    Java compiler and library
19283
19284     * [499]10746 [win32] garbage collection crash in GCJ
19285
19286    Objective-C compiler and library
19287
19288     * [500]11433 Crash due to dereferencing null pointer when querying
19289       protocol
19290
19291    Fortran compiler and library
19292
19293     * [501]12633 logical expression gives incorrect result with
19294       -fugly-logint option
19295     * [502]13037 [gcse-lm] g77 generates incorrect code
19296     * [503]13213 Hex constant problem when compiling with -fugly-logint
19297       and -ftypeless-boz
19298
19299    x86-specific (Intel/AMD)
19300
19301     * [504]4490 ICE with -m128bit-long-double
19302     * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have
19303       `reg' in make_field_assignment, in combine.c
19304     * [506]12441 ICE: can't find a register to spill
19305     * [507]12943 array static-init failure under -fpic, -fPIC
19306     * [508]13608 Incorrect code with -O3 -ffast-math
19307
19308    PowerPC-specific
19309
19310     * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of
19311       __attribute__((aligned(16)))
19312     * [510]11793 ICE in extract_insn, in recog.c (const_vector's)
19313     * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in
19314       altivec.md)
19315     * [512]12537 g++ generates writeable text sections
19316
19317    SPARC-specific
19318
19319     * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0
19320       -m64
19321     * [514]12865 mprotect call to make trampoline executable may fail
19322     * [515]13354 ICE in sparc_emit_set_const32
19323
19324    ARM-specific
19325
19326     * [516]10467 [arm] ICE in pre_insert_copy_insn,
19327
19328    ia64-specific
19329
19330     * [517]11226 ICE passing struct arg with two floats
19331     * [518]11227 ICE for _Complex float, _Complex long double args
19332     * [519]12644 GCC 3.3.2 fails to compile glibc on ia64
19333     * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn
19334     * Various fixes for libunwind
19335
19336    Alpha-specific
19337
19338     * [521]12654 Incorrect comparison code generated for Alpha
19339     * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2
19340     * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2
19341
19342    HPPA-specific
19343
19344     * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c
19345     * [525]12158 [hppa] compilation does not terminate at -O1
19346
19347    S390-specific
19348
19349     * [526]11992 Wrong built-in code for memcmp with length 1<<24: only
19350       (1<<24)-1 possible for CLCL-Instruction
19351
19352    SH-specific
19353
19354     * [527]9365 segfault in gen_far_branch (config/sh/sh.c)
19355     * [528]10392 optimizer generates faulty array indexing
19356     * [529]11322 SH profiler outputs multiple definitions of symbol
19357     * [530]13069 gcc/config/sh/rtems.h broken
19358     * [531]13302 Putting a va_list in a struct causes seg fault
19359     * [532]13585 Incorrect optimization of call to sfunc
19360     * Fix inappropriately exported libgcc functions from the shared
19361       library
19362
19363    Other embedded target specific
19364
19365     * [533]8916 [mcore] unsigned char assign gets hosed.
19366     * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c
19367     * [535]13122 [h8300] local variable gets corrupted by function call
19368       when -fomit-frame-pointer is given
19369     * [536]13256 [cris] strict_low_part mistreated in delay slots
19370     * [537]13373 [mcore] optimization with -frerun-cse-after-loop
19371       -fexpensive-optimizations produces wrong code on mcore
19372
19373    GNU HURD-specific
19374
19375     * [538]12561 gcc/config/t-gnu needs updating to work with
19376       --with-sysroot
19377
19378    Tru64 Unix specific
19379
19380     * [539]6243 testsuite fails almost all tests due to no libintl in
19381       LD_LIBRARY_PATH during test.
19382     * [540]11397 weak aliases broken on Tru64 UNIX
19383
19384    AIX-specific
19385
19386     * [541]12505 build failure due to defines of uchar in cpphash.h and
19387       sys/types.h
19388     * [542]13150 WEAK symbols not exported by collect2
19389
19390    IRIX-specific
19391
19392     * [543]12666 fixincludes problem on IRIX 6.5.19m
19393
19394    Solaris-specific
19395
19396     * [544]12969 Including sys/byteorder.h breaks configure checks
19397
19398    Testsuite problems (compiler is not affected)
19399
19400     * [545]10819 testsuite creates CR+LF on compiler version lines in
19401       test summary files
19402     * [546]11612 abi_check not finding correct libgcc_s.so.1
19403
19404    Miscellaneous
19405
19406     * [547]13211 using -###, incorrect warnings about unused linker file
19407       are produced
19408     __________________________________________________________________
19409
19410GCC 3.3.4
19411
19412   This is the [548]list of problem reports (PRs) from GCC's bug tracking
19413   system that are known to be fixed in the 3.3.4 release. This list might
19414   not be complete (that is, it is possible that some PRs that have been
19415   fixed are not listed here).
19416     __________________________________________________________________
19417
19418GCC 3.3.5
19419
19420   This is the [549]list of problem reports (PRs) from GCC's bug tracking
19421   system that are known to be fixed in the 3.3.5 release. This list might
19422   not be complete (that is, it is possible that some PRs that have been
19423   fixed are not listed here).
19424     __________________________________________________________________
19425
19426GCC 3.3.6
19427
19428   This is the [550]list of problem reports (PRs) from GCC's bug tracking
19429   system that are known to be fixed in the 3.3.6 release. This list might
19430   not be complete (that is, it is possible that some PRs that have been
19431   fixed are not listed here).
19432
19433
19434    For questions related to the use of GCC, please consult these web
19435    pages and the [551]GCC manuals. If that fails, the
19436    [552]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19437    web pages and the development of GCC are welcome on our developer
19438    list at [553]gcc@gcc.gnu.org. All of [554]our lists have public
19439    archives.
19440
19441   Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and
19442   distribution of this entire article is permitted in any medium,
19443   provided this notice is preserved.
19444
19445   These pages are [556]maintained by the GCC team. Last modified
19446   2023-01-19.
19447
19448References
19449
19450   1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
19451   2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems
19452   3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
19453   4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute
19454   5. http://gcc.gnu.org/news/dfa.html
19455   6. http://gcc.gnu.org/c99status.html
19456   7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html
19457   8. https://gcc.gnu.org/PR10140
19458   9. https://gcc.gnu.org/PR10198
19459  10. https://gcc.gnu.org/PR10338
19460  11. https://gcc.gnu.org/PR3581
19461  12. https://gcc.gnu.org/PR4382
19462  13. https://gcc.gnu.org/PR5533
19463  14. https://gcc.gnu.org/PR6387
19464  15. https://gcc.gnu.org/PR6412
19465  16. https://gcc.gnu.org/PR6620
19466  17. https://gcc.gnu.org/PR6663
19467  18. https://gcc.gnu.org/PR7068
19468  19. https://gcc.gnu.org/PR7083
19469  20. https://gcc.gnu.org/PR7647
19470  21. https://gcc.gnu.org/PR7675
19471  22. https://gcc.gnu.org/PR7718
19472  23. https://gcc.gnu.org/PR8116
19473  24. https://gcc.gnu.org/PR8358
19474  25. https://gcc.gnu.org/PR8511
19475  26. https://gcc.gnu.org/PR8564
19476  27. https://gcc.gnu.org/PR8660
19477  28. https://gcc.gnu.org/PR8766
19478  29. https://gcc.gnu.org/PR8803
19479  30. https://gcc.gnu.org/PR8846
19480  31. https://gcc.gnu.org/PR8906
19481  32. https://gcc.gnu.org/PR9216
19482  33. https://gcc.gnu.org/PR9261
19483  34. https://gcc.gnu.org/PR9263
19484  35. https://gcc.gnu.org/PR9429
19485  36. https://gcc.gnu.org/PR9516
19486  37. https://gcc.gnu.org/PR9600
19487  38. https://gcc.gnu.org/PR9629
19488  39. https://gcc.gnu.org/PR9672
19489  40. https://gcc.gnu.org/PR9749
19490  41. https://gcc.gnu.org/PR9794
19491  42. https://gcc.gnu.org/PR9829
19492  43. https://gcc.gnu.org/PR9916
19493  44. https://gcc.gnu.org/PR9936
19494  45. https://gcc.gnu.org/PR10262
19495  46. https://gcc.gnu.org/PR10278
19496  47. https://gcc.gnu.org/PR10446
19497  48. https://gcc.gnu.org/PR10451
19498  49. https://gcc.gnu.org/PR10506
19499  50. https://gcc.gnu.org/PR10549
19500  51. https://gcc.gnu.org/PR2001
19501  52. https://gcc.gnu.org/PR2391
19502  53. https://gcc.gnu.org/PR2960
19503  54. https://gcc.gnu.org/PR4046
19504  55. https://gcc.gnu.org/PR6405
19505  56. https://gcc.gnu.org/PR6798
19506  57. https://gcc.gnu.org/PR6871
19507  58. https://gcc.gnu.org/PR6909
19508  59. https://gcc.gnu.org/PR7189
19509  60. https://gcc.gnu.org/PR7642
19510  61. https://gcc.gnu.org/PR8634
19511  62. https://gcc.gnu.org/PR8750
19512  63. https://gcc.gnu.org/PR2161
19513  64. https://gcc.gnu.org/PR4319
19514  65. https://gcc.gnu.org/PR8602
19515  66. https://gcc.gnu.org/PR9177
19516  67. https://gcc.gnu.org/PR9853
19517  68. https://gcc.gnu.org/PR45
19518  69. https://gcc.gnu.org/PR3784
19519  70. https://gcc.gnu.org/PR764
19520  71. https://gcc.gnu.org/PR5116
19521  72. https://gcc.gnu.org/PR2862
19522  73. https://gcc.gnu.org/PR3663
19523  74. https://gcc.gnu.org/PR3797
19524  75. https://gcc.gnu.org/PR3948
19525  76. https://gcc.gnu.org/PR4137
19526  77. https://gcc.gnu.org/PR4361
19527  78. https://gcc.gnu.org/PR4802
19528  79. https://gcc.gnu.org/PR5837
19529  80. https://gcc.gnu.org/PR4803
19530  81. https://gcc.gnu.org/PR5094
19531  82. https://gcc.gnu.org/PR5730
19532  83. https://gcc.gnu.org/PR6713
19533  84. https://gcc.gnu.org/PR7015
19534  85. https://gcc.gnu.org/PR7086
19535  86. https://gcc.gnu.org/PR7099
19536  87. https://gcc.gnu.org/PR7247
19537  88. https://gcc.gnu.org/PR7441
19538  89. https://gcc.gnu.org/PR7768
19539  90. https://gcc.gnu.org/PR7804
19540  91. https://gcc.gnu.org/PR8099
19541  92. https://gcc.gnu.org/PR8117
19542  93. https://gcc.gnu.org/PR8205
19543  94. https://gcc.gnu.org/PR8645
19544  95. https://gcc.gnu.org/PR8724
19545  96. https://gcc.gnu.org/PR8805
19546  97. https://gcc.gnu.org/PR8691
19547  98. https://gcc.gnu.org/PR8700
19548  99. https://gcc.gnu.org/PR8724
19549 100. https://gcc.gnu.org/PR8949
19550 101. https://gcc.gnu.org/PR9016
19551 102. https://gcc.gnu.org/PR9053
19552 103. https://gcc.gnu.org/PR9152
19553 104. https://gcc.gnu.org/PR9182
19554 105. https://gcc.gnu.org/PR9297
19555 106. https://gcc.gnu.org/PR9318
19556 107. https://gcc.gnu.org/PR9320
19557 108. https://gcc.gnu.org/PR9400
19558 109. https://gcc.gnu.org/PR9424
19559 110. https://gcc.gnu.org/PR9425
19560 111. https://gcc.gnu.org/PR9439
19561 112. https://gcc.gnu.org/PR9474
19562 113. https://gcc.gnu.org/PR9548
19563 114. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231
19564 115. https://gcc.gnu.org/PR9555
19565 116. https://gcc.gnu.org/PR9561
19566 117. https://gcc.gnu.org/PR9563
19567 118. https://gcc.gnu.org/PR9582
19568 119. https://gcc.gnu.org/PR9622
19569 120. https://gcc.gnu.org/PR9683
19570 121. https://gcc.gnu.org/PR9791
19571 122. https://gcc.gnu.org/PR9817
19572 123. https://gcc.gnu.org/PR9825
19573 124. https://gcc.gnu.org/PR9826
19574 125. https://gcc.gnu.org/PR9924
19575 126. https://gcc.gnu.org/PR9946
19576 127. https://gcc.gnu.org/PR9964
19577 128. https://gcc.gnu.org/PR9988
19578 129. https://gcc.gnu.org/PR10033
19579 130. https://gcc.gnu.org/PR10097
19580 131. https://gcc.gnu.org/PR10132
19581 132. https://gcc.gnu.org/PR10180
19582 133. https://gcc.gnu.org/PR10199
19583 134. https://gcc.gnu.org/PR10300
19584 135. https://gcc.gnu.org/PR10427
19585 136. https://gcc.gnu.org/PR10503
19586 137. https://gcc.gnu.org/PR5956
19587 138. https://gcc.gnu.org/PR1832
19588 139. https://gcc.gnu.org/PR3924
19589 140. https://gcc.gnu.org/PR5634
19590 141. https://gcc.gnu.org/PR6367
19591 142. https://gcc.gnu.org/PR6491
19592 143. https://gcc.gnu.org/PR6742
19593 144. https://gcc.gnu.org/PR7113
19594 145. https://gcc.gnu.org/PR7236
19595 146. https://gcc.gnu.org/PR7278
19596 147. https://gcc.gnu.org/PR7384
19597 148. https://gcc.gnu.org/PR7388
19598 149. https://gcc.gnu.org/PR8587
19599 150. https://gcc.gnu.org/PR9038
19600 151. https://gcc.gnu.org/PR10197
19601 152. https://gcc.gnu.org/PR6005
19602 153. https://gcc.gnu.org/PR6389
19603 154. https://gcc.gnu.org/PR6576
19604 155. https://gcc.gnu.org/PR6652
19605 156. https://gcc.gnu.org/PR7060
19606 157. https://gcc.gnu.org/PR7073
19607 158. https://gcc.gnu.org/PR7180
19608 159. https://gcc.gnu.org/PR7416
19609 160. https://gcc.gnu.org/PR7570
19610 161. https://gcc.gnu.org/PR7578
19611 162. https://gcc.gnu.org/PR7611
19612 163. https://gcc.gnu.org/PR7709
19613 164. https://gcc.gnu.org/PR7766
19614 165. https://gcc.gnu.org/PR7785
19615 166. https://gcc.gnu.org/PR7786
19616 167. https://gcc.gnu.org/PR8142
19617 168. https://gcc.gnu.org/PR8234
19618 169. https://gcc.gnu.org/PR8415
19619 170. https://gcc.gnu.org/PR8481
19620 171. https://gcc.gnu.org/PR8593
19621 172. https://gcc.gnu.org/PR8759
19622 173. https://gcc.gnu.org/PR8997
19623 174. https://gcc.gnu.org/PR9253
19624 175. https://gcc.gnu.org/PR9254
19625 176. https://gcc.gnu.org/PR9271
19626 177. https://gcc.gnu.org/PR6767
19627 178. https://gcc.gnu.org/PR9911
19628 179. https://gcc.gnu.org/PR10020
19629 180. https://gcc.gnu.org/PR10546
19630 181. https://gcc.gnu.org/PR7029
19631 182. https://gcc.gnu.org/PR2903
19632 183. https://gcc.gnu.org/PR7873
19633 184. https://gcc.gnu.org/PR7680
19634 185. https://gcc.gnu.org/PR8705
19635 186. https://gcc.gnu.org/PR9986
19636 187. https://gcc.gnu.org/PR10056
19637 188. https://gcc.gnu.org/PR6744
19638 189. https://gcc.gnu.org/PR7361
19639 190. https://gcc.gnu.org/PR9496
19640 191. https://gcc.gnu.org/PR7067
19641 192. https://gcc.gnu.org/PR8480
19642 193. https://gcc.gnu.org/PR8784
19643 194. https://gcc.gnu.org/PR10315
19644 195. https://gcc.gnu.org/PR10267
19645 196. https://gcc.gnu.org/PR7916
19646 197. https://gcc.gnu.org/PR7926
19647 198. https://gcc.gnu.org/PR8555
19648 199. https://gcc.gnu.org/PR8994
19649 200. https://gcc.gnu.org/PR9426
19650 201. https://gcc.gnu.org/PR9806
19651 202. https://gcc.gnu.org/PR10077
19652 203. https://gcc.gnu.org/PR10233
19653 204. https://gcc.gnu.org/PR10286
19654 205. https://gcc.gnu.org/PR10308
19655 206. https://gcc.gnu.org/PR11272
19656 207. https://gcc.gnu.org/PR5754
19657 208. https://gcc.gnu.org/PR6597
19658 209. https://gcc.gnu.org/PR6949
19659 210. https://gcc.gnu.org/PR7053
19660 211. https://gcc.gnu.org/PR8164
19661 212. https://gcc.gnu.org/PR8384
19662 213. https://gcc.gnu.org/PR9559
19663 214. https://gcc.gnu.org/PR9649
19664 215. https://gcc.gnu.org/PR9864
19665 216. https://gcc.gnu.org/PR10432
19666 217. https://gcc.gnu.org/PR10475
19667 218. https://gcc.gnu.org/PR10635
19668 219. https://gcc.gnu.org/PR10661
19669 220. https://gcc.gnu.org/PR10700
19670 221. https://gcc.gnu.org/PR10712
19671 222. https://gcc.gnu.org/PR10796
19672 223. https://gcc.gnu.org/PR10890
19673 224. https://gcc.gnu.org/PR10939
19674 225. https://gcc.gnu.org/PR10956
19675 226. https://gcc.gnu.org/PR11041
19676 227. https://gcc.gnu.org/PR11059
19677 228. https://gcc.gnu.org/PR11083
19678 229. https://gcc.gnu.org/PR11105
19679 230. https://gcc.gnu.org/PR11149
19680 231. https://gcc.gnu.org/PR11228
19681 232. https://gcc.gnu.org/PR11282
19682 233. https://gcc.gnu.org/PR11301
19683 234. https://gcc.gnu.org/PR11308
19684 235. https://gcc.gnu.org/PR11473
19685 236. https://gcc.gnu.org/PR11503
19686 237. https://gcc.gnu.org/PR11513
19687 238. https://gcc.gnu.org/PR11198
19688 239. https://gcc.gnu.org/PR11304
19689 240. https://gcc.gnu.org/PR11381
19690 241. https://gcc.gnu.org/PR11536
19691 242. https://gcc.gnu.org/PR11557
19692 243. https://gcc.gnu.org/PR5897
19693 244. https://gcc.gnu.org/PR11279
19694 245. https://gcc.gnu.org/PR11022
19695 246. https://gcc.gnu.org/PR2330
19696 247. https://gcc.gnu.org/PR5388
19697 248. https://gcc.gnu.org/PR5390
19698 249. https://gcc.gnu.org/PR7877
19699 250. https://gcc.gnu.org/PR9393
19700 251. https://gcc.gnu.org/PR10032
19701 252. https://gcc.gnu.org/PR10468
19702 253. https://gcc.gnu.org/PR10527
19703 254. https://gcc.gnu.org/PR10679
19704 255. https://gcc.gnu.org/PR10682
19705 256. https://gcc.gnu.org/PR10689
19706 257. https://gcc.gnu.org/PR10845
19707 258. https://gcc.gnu.org/PR10849
19708 259. https://gcc.gnu.org/PR10888
19709 260. https://gcc.gnu.org/PR10929
19710 261. https://gcc.gnu.org/PR10931
19711 262. https://gcc.gnu.org/PR10940
19712 263. https://gcc.gnu.org/PR10968
19713 264. https://gcc.gnu.org/PR10990
19714 265. https://gcc.gnu.org/PR11039
19715 266. https://gcc.gnu.org/PR11062
19716 267. https://gcc.gnu.org/PR11095
19717 268. https://gcc.gnu.org/PR11098
19718 269. https://gcc.gnu.org/PR11137
19719 270. https://gcc.gnu.org/PR11154
19720 271. https://gcc.gnu.org/PR11329
19721 272. https://gcc.gnu.org/PR11332
19722 273. https://gcc.gnu.org/PR11431
19723 274. https://gcc.gnu.org/PR11528
19724 275. https://gcc.gnu.org/PR11546
19725 276. https://gcc.gnu.org/PR11567
19726 277. https://gcc.gnu.org/PR11645
19727 278. https://gcc.gnu.org/PR5179
19728 279. https://gcc.gnu.org/PR8204
19729 280. https://gcc.gnu.org/PR10838
19730 281. https://gcc.gnu.org/PR10886
19731 282. https://gcc.gnu.org/PR11349
19732 283. https://gcc.gnu.org/PR4823
19733 284. https://gcc.gnu.org/PR8878
19734 285. https://gcc.gnu.org/PR9815
19735 286. https://gcc.gnu.org/PR10402
19736 287. https://gcc.gnu.org/PR10504
19737 288. https://gcc.gnu.org/PR10673
19738 289. https://gcc.gnu.org/PR11044
19739 290. https://gcc.gnu.org/PR11089
19740 291. https://gcc.gnu.org/PR11420
19741 292. https://gcc.gnu.org/PR9362
19742 293. https://gcc.gnu.org/PR10142
19743 294. https://gcc.gnu.org/PR10663
19744 295. https://gcc.gnu.org/PR10835
19745 296. https://gcc.gnu.org/PR10876
19746 297. https://gcc.gnu.org/PR10955
19747 298. https://gcc.gnu.org/PR11018
19748 299. https://gcc.gnu.org/PR11556
19749 300. https://gcc.gnu.org/PR10907
19750 301. https://gcc.gnu.org/PR11320
19751 302. https://gcc.gnu.org/PR11599
19752 303. https://gcc.gnu.org/PR9745
19753 304. https://gcc.gnu.org/PR10871
19754 305. https://gcc.gnu.org/PR11440
19755 306. https://gcc.gnu.org/PR7594
19756 307. https://gcc.gnu.org/PR10557
19757 308. https://gcc.gnu.org/PR11054
19758 309. https://gcc.gnu.org/PR10834
19759 310. https://gcc.gnu.org/PR10842
19760 311. https://gcc.gnu.org/PR11052
19761 312. https://gcc.gnu.org/PR11183
19762 313. https://gcc.gnu.org/PR11084
19763 314. https://gcc.gnu.org/PR10331
19764 315. https://gcc.gnu.org/PR10413
19765 316. https://gcc.gnu.org/PR11096
19766 317. https://gcc.gnu.org/PR2873
19767 318. https://gcc.gnu.org/PR3163
19768 319. https://gcc.gnu.org/PR5287
19769 320. https://gcc.gnu.org/PR10148
19770 321. https://gcc.gnu.org/PR8787
19771 322. https://gcc.gnu.org/PR10900
19772 323. https://gcc.gnu.org/PR1607
19773 324. https://gcc.gnu.org/PR4252
19774 325. https://gcc.gnu.org/PR4490
19775 326. https://gcc.gnu.org/PR10355
19776 327. https://gcc.gnu.org/PR10726
19777 328. https://gcc.gnu.org/PR10805
19778 329. https://gcc.gnu.org/PR10815
19779 330. https://gcc.gnu.org/PR10877
19780 331. https://gcc.gnu.org/PR11280
19781 332. https://gcc.gnu.org/PR11466
19782 333. https://gcc.gnu.org/PR10737
19783 334. https://gcc.gnu.org/PR10810
19784 335. https://gcc.gnu.org/PR8336
19785 336. https://gcc.gnu.org/PR9330
19786 337. https://gcc.gnu.org/PR9631
19787 338. https://gcc.gnu.org/PR9877
19788 339. https://gcc.gnu.org/PR11687
19789 340. https://gcc.gnu.org/PR12263
19790 341. https://gcc.gnu.org/PR12490
19791 342. https://gcc.gnu.org/PR7277
19792 343. https://gcc.gnu.org/PR7939
19793 344. https://gcc.gnu.org/PR11063
19794 345. https://gcc.gnu.org/PR11207
19795 346. https://gcc.gnu.org/PR11522
19796 347. https://gcc.gnu.org/PR11595
19797 348. https://gcc.gnu.org/PR11646
19798 349. https://gcc.gnu.org/PR11665
19799 350. https://gcc.gnu.org/PR11852
19800 351. https://gcc.gnu.org/PR11878
19801 352. https://gcc.gnu.org/PR11883
19802 353. https://gcc.gnu.org/PR11991
19803 354. https://gcc.gnu.org/PR12146
19804 355. https://gcc.gnu.org/PR12215
19805 356. https://gcc.gnu.org/PR12369
19806 357. https://gcc.gnu.org/PR12446
19807 358. https://gcc.gnu.org/PR12510
19808 359. https://gcc.gnu.org/PR12544
19809 360. https://gcc.gnu.org/PR9862
19810 361. https://gcc.gnu.org/PR10962
19811 362. https://gcc.gnu.org/PR11370
19812 363. https://gcc.gnu.org/PR11637
19813 364. https://gcc.gnu.org/PR11885
19814 365. https://gcc.gnu.org/PR12082
19815 366. https://gcc.gnu.org/PR12180
19816 367. https://gcc.gnu.org/PR12340
19817 368. https://gcc.gnu.org/PR3907
19818 369. https://gcc.gnu.org/PR5293
19819 370. https://gcc.gnu.org/PR5296
19820 371. https://gcc.gnu.org/PR7939
19821 372. https://gcc.gnu.org/PR8656
19822 373. https://gcc.gnu.org/PR10147
19823 374. https://gcc.gnu.org/PR11400
19824 375. https://gcc.gnu.org/PR11409
19825 376. https://gcc.gnu.org/PR11740
19826 377. https://gcc.gnu.org/PR11786
19827 378. https://gcc.gnu.org/PR11867
19828 379. https://gcc.gnu.org/PR11928
19829 380. https://gcc.gnu.org/PR12114
19830 381. https://gcc.gnu.org/PR12163
19831 382. https://gcc.gnu.org/PR12181
19832 383. https://gcc.gnu.org/PR12236
19833 384. https://gcc.gnu.org/PR12266
19834 385. https://gcc.gnu.org/PR12296
19835 386. https://gcc.gnu.org/PR12298
19836 387. https://gcc.gnu.org/PR12369
19837 388. https://gcc.gnu.org/PR12337
19838 389. https://gcc.gnu.org/PR12344
19839 390. https://gcc.gnu.org/PR12451
19840 391. https://gcc.gnu.org/PR12486
19841 392. https://gcc.gnu.org/PR8869
19842 393. https://gcc.gnu.org/PR9786
19843 394. https://gcc.gnu.org/PR11689
19844 395. https://gcc.gnu.org/PR12116
19845 396. https://gcc.gnu.org/PR12070
19846 397. https://gcc.gnu.org/PR11184
19847 398. https://gcc.gnu.org/PR11535
19848 399. https://gcc.gnu.org/PR11693
19849 400. https://gcc.gnu.org/PR12224
19850 401. https://gcc.gnu.org/PR11087
19851 402. https://gcc.gnu.org/PR11319
19852 403. https://gcc.gnu.org/PR11949
19853 404. https://gcc.gnu.org/PR11662
19854 405. https://gcc.gnu.org/PR11965
19855 406. https://gcc.gnu.org/PR12301
19856 407. https://gcc.gnu.org/PR11717
19857 408. https://gcc.gnu.org/PR11313
19858 409. https://gcc.gnu.org/PR11712
19859 410. https://gcc.gnu.org/PR12166
19860 411. https://gcc.gnu.org/PR12101
19861 412. https://gcc.gnu.org/PR10988
19862 413. https://gcc.gnu.org/PR11805
19863 414. https://gcc.gnu.org/PR11902
19864 415. https://gcc.gnu.org/PR11903
19865 416. https://gcc.gnu.org/PR11890
19866 417. https://gcc.gnu.org/PR12399
19867 418. https://gcc.gnu.org/PR13068
19868 419. https://gcc.gnu.org/PR10060
19869 420. https://gcc.gnu.org/PR10555
19870 421. https://gcc.gnu.org/PR10706
19871 422. https://gcc.gnu.org/PR11496
19872 423. https://gcc.gnu.org/PR11741
19873 424. https://gcc.gnu.org/PR12440
19874 425. https://gcc.gnu.org/PR12632
19875 426. https://gcc.gnu.org/PR12712
19876 427. https://gcc.gnu.org/PR12726
19877 428. https://gcc.gnu.org/PR12890
19878 429. https://gcc.gnu.org/PR12900
19879 430. https://gcc.gnu.org/PR13060
19880 431. https://gcc.gnu.org/PR13289
19881 432. https://gcc.gnu.org/PR13318
19882 433. https://gcc.gnu.org/PR13392
19883 434. https://gcc.gnu.org/PR13574
19884 435. https://gcc.gnu.org/PR13475
19885 436. https://gcc.gnu.org/PR13797
19886 437. https://gcc.gnu.org/PR13824
19887 438. https://gcc.gnu.org/PR8776
19888 439. https://gcc.gnu.org/PR10339
19889 440. https://gcc.gnu.org/PR11350
19890 441. https://gcc.gnu.org/PR12826
19891 442. https://gcc.gnu.org/PR12500
19892 443. https://gcc.gnu.org/PR12941
19893 444. https://gcc.gnu.org/PR12953
19894 445. https://gcc.gnu.org/PR13041
19895 446. https://gcc.gnu.org/PR13507
19896 447. https://gcc.gnu.org/PR13382
19897 448. https://gcc.gnu.org/PR13394
19898 449. https://gcc.gnu.org/PR13400
19899 450. https://gcc.gnu.org/PR13521
19900 451. https://gcc.gnu.org/PR2094
19901 452. https://gcc.gnu.org/PR2294
19902 453. https://gcc.gnu.org/PR5050
19903 454. https://gcc.gnu.org/PR9371
19904 455. https://gcc.gnu.org/PR9546
19905 456. https://gcc.gnu.org/PR10081
19906 457. https://gcc.gnu.org/PR10093
19907 458. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61
19908 459. https://gcc.gnu.org/PR10095
19909 460. https://gcc.gnu.org/PR11554
19910 461. https://gcc.gnu.org/PR12297
19911 462. https://gcc.gnu.org/PR12352
19912 463. https://gcc.gnu.org/PR12438
19913 464. https://gcc.gnu.org/PR12540
19914 465. https://gcc.gnu.org/PR12594
19915 466. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60
19916 467. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63
19917 468. https://gcc.gnu.org/PR12657
19918 469. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292
19919 470. https://gcc.gnu.org/PR12696
19920 471. https://gcc.gnu.org/PR12815
19921 472. https://gcc.gnu.org/PR12862
19922 473. https://gcc.gnu.org/PR12926
19923 474. https://gcc.gnu.org/PR12967
19924 475. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html
19925 476. https://gcc.gnu.org/PR12971
19926 477. https://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328
19927 478. https://gcc.gnu.org/PR13007
19928 479. https://gcc.gnu.org/PR13009
19929 480. https://gcc.gnu.org/PR13057
19930 481. https://gcc.gnu.org/PR13070
19931 482. https://gcc.gnu.org/PR13081
19932 483. https://gcc.gnu.org/PR13239
19933 484. https://gcc.gnu.org/PR13262
19934 485. https://gcc.gnu.org/PR13290
19935 486. https://gcc.gnu.org/PR13323
19936 487. https://gcc.gnu.org/PR13369
19937 488. https://gcc.gnu.org/PR13371
19938 489. https://gcc.gnu.org/PR13445
19939 490. https://gcc.gnu.org/PR13461
19940 491. https://gcc.gnu.org/PR13462
19941 492. https://gcc.gnu.org/PR13478
19942 493. https://gcc.gnu.org/PR13544
19943 494. https://gcc.gnu.org/PR13650
19944 495. https://gcc.gnu.org/PR13683
19945 496. https://gcc.gnu.org/PR13688
19946 497. https://gcc.gnu.org/PR13774
19947 498. https://gcc.gnu.org/PR13884
19948 499. https://gcc.gnu.org/PR10746
19949 500. https://gcc.gnu.org/PR11433
19950 501. https://gcc.gnu.org/PR12633
19951 502. https://gcc.gnu.org/PR13037
19952 503. https://gcc.gnu.org/PR13213
19953 504. https://gcc.gnu.org/PR4490
19954 505. https://gcc.gnu.org/PR12292
19955 506. https://gcc.gnu.org/PR12441
19956 507. https://gcc.gnu.org/PR12943
19957 508. https://gcc.gnu.org/PR13608
19958 509. https://gcc.gnu.org/PR11598
19959 510. https://gcc.gnu.org/PR11793
19960 511. https://gcc.gnu.org/PR12467
19961 512. https://gcc.gnu.org/PR12537
19962 513. https://gcc.gnu.org/PR12496
19963 514. https://gcc.gnu.org/PR12865
19964 515. https://gcc.gnu.org/PR13354
19965 516. https://gcc.gnu.org/PR10467
19966 517. https://gcc.gnu.org/PR11226
19967 518. https://gcc.gnu.org/PR11227
19968 519. https://gcc.gnu.org/PR12644
19969 520. https://gcc.gnu.org/PR13149
19970 521. https://gcc.gnu.org/PR12654
19971 522. https://gcc.gnu.org/PR12965
19972 523. https://gcc.gnu.org/PR13031
19973 524. https://gcc.gnu.org/PR11634
19974 525. https://gcc.gnu.org/PR12158
19975 526. https://gcc.gnu.org/PR11992
19976 527. https://gcc.gnu.org/PR9365
19977 528. https://gcc.gnu.org/PR10392
19978 529. https://gcc.gnu.org/PR11322
19979 530. https://gcc.gnu.org/PR13069
19980 531. https://gcc.gnu.org/PR13302
19981 532. https://gcc.gnu.org/PR13585
19982 533. https://gcc.gnu.org/PR8916
19983 534. https://gcc.gnu.org/PR11576
19984 535. https://gcc.gnu.org/PR13122
19985 536. https://gcc.gnu.org/PR13256
19986 537. https://gcc.gnu.org/PR13373
19987 538. https://gcc.gnu.org/PR12561
19988 539. https://gcc.gnu.org/PR6243
19989 540. https://gcc.gnu.org/PR11397
19990 541. https://gcc.gnu.org/PR12505
19991 542. https://gcc.gnu.org/PR13150
19992 543. https://gcc.gnu.org/PR12666
19993 544. https://gcc.gnu.org/PR12969
19994 545. https://gcc.gnu.org/PR10819
19995 546. https://gcc.gnu.org/PR11612
19996 547. https://gcc.gnu.org/PR13211
19997 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4
19998 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5
19999 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6
20000 551. https://gcc.gnu.org/onlinedocs/
20001 552. mailto:gcc-help@gcc.gnu.org
20002 553. mailto:gcc@gcc.gnu.org
20003 554. https://gcc.gnu.org/lists.html
20004 555. https://www.fsf.org/
20005 556. https://gcc.gnu.org/about.html
20006======================================================================
20007http://gcc.gnu.org/gcc-3.2/index.html
20008
20009                             GCC 3.2 Release Series
20010
20011   (This release series is no longer supported.)
20012
20013   April 25, 2003
20014
20015   The [1]GNU project and the GCC developers are pleased to announce the
20016   release of GCC 3.2.3.
20017
20018   The purpose of the GCC 3.2 release series is to provide a stable
20019   platform for OS distributors to use building their next releases. A
20020   primary objective was to stabilize the C++ ABI; we believe that the
20021   interface to the compiler and the C++ standard library are now
20022   relatively stable.
20023
20024   Be aware that C++ code compiled by GCC 3.2.x will (in general) not
20025   interoperate with code compiled by GCC 3.1.1 or earlier.
20026
20027   Please refer to our [2]detailed list of news, caveats, and bug-fixes
20028   for further information.
20029
20030Release History
20031
20032   GCC 3.2.3
20033          April 25, 2003 ([3]changes)
20034
20035   GCC 3.2.2
20036          February 5, 2003 ([4]changes)
20037
20038   GCC 3.2.1
20039          November 19, 2002 ([5]changes)
20040
20041   GCC 3.2
20042          August 14, 2002 ([6]changes)
20043
20044References and Acknowledgements
20045
20046   GCC used to stand for the GNU C Compiler, but since the compiler
20047   supports several other languages aside from C, it now stands for the
20048   GNU Compiler Collection.
20049
20050   A list of [7]successful builds is updated as new information becomes
20051   available.
20052
20053   The GCC developers would like to thank the numerous people that have
20054   contributed new features, improvements, bug fixes, and other changes as
20055   well as test results to GCC. This [8]amazing group of volunteers is
20056   what makes GCC successful.
20057
20058   For additional information about GCC please refer to the [9]GCC project
20059   web site or contact the [10]GCC development mailing list.
20060
20061   To obtain GCC please use [11]our mirror sites, or our CVS server.
20062
20063
20064    For questions related to the use of GCC, please consult these web
20065    pages and the [12]GCC manuals. If that fails, the
20066    [13]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20067    web pages and the development of GCC are welcome on our developer
20068    list at [14]gcc@gcc.gnu.org. All of [15]our lists have public
20069    archives.
20070
20071   Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and
20072   distribution of this entire article is permitted in any medium,
20073   provided this notice is preserved.
20074
20075   These pages are [17]maintained by the GCC team. Last modified
20076   2022-10-26.
20077
20078References
20079
20080   1. http://www.gnu.org/
20081   2. http://gcc.gnu.org/gcc-3.2/changes.html
20082   3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
20083   4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2
20084   5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1
20085   6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2
20086   7. http://gcc.gnu.org/gcc-3.2/buildstat.html
20087   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
20088   9. http://gcc.gnu.org/index.html
20089  10. mailto:gcc@gcc.gnu.org
20090  11. http://gcc.gnu.org/mirrors.html
20091  12. https://gcc.gnu.org/onlinedocs/
20092  13. mailto:gcc-help@gcc.gnu.org
20093  14. mailto:gcc@gcc.gnu.org
20094  15. https://gcc.gnu.org/lists.html
20095  16. https://www.fsf.org/
20096  17. https://gcc.gnu.org/about.html
20097======================================================================
20098http://gcc.gnu.org/gcc-3.2/changes.html
20099
20100                             GCC 3.2 Release Series
20101                        Changes, New Features, and Fixes
20102
20103   The latest release in the 3.2 release series is [1]GCC 3.2.3.
20104
20105Caveats and New Features
20106
20107  Caveats
20108
20109     * The C++ compiler does not correctly zero-initialize
20110       pointers-to-data members. You must explicitly initialize them. For
20111       example: int S::*m(0); will work, but depending on
20112       default-initialization to zero will not work. This bug cannot be
20113       fixed in GCC 3.2 without inducing unacceptable risks. It will be
20114       fixed in GCC 3.3.
20115     * This GCC release is based on the GCC 3.1 sourcebase, and thus has
20116       all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has
20117       a number of C++ ABI fixes which make its C++ compiler generate
20118       binary code which is incompatible with the C++ compilers found in
20119       earlier GCC releases, including GCC 3.1 and GCC 3.1.1.
20120
20121  Frontend Enhancements
20122
20123    C/C++/Objective-C
20124
20125     * The method of constructing the list of directories to be searched
20126       for header files has been revised. If a directory named by a -I
20127       option is a standard system include directory, the option is
20128       ignored to ensure that the default search order for system
20129       directories and the special treatment of system header files are
20130       not defeated.
20131     * The C and Objective-C compilers no longer accept the "Naming Types"
20132       extension (typedef foo = bar); it was already unavailable in C++.
20133       Code which uses it will need to be changed to use the "typeof"
20134       extension instead: typedef typeof(bar) foo. (We have removed this
20135       extension without a period of deprecation because it has caused the
20136       compiler to crash since version 3.0 and no one noticed until very
20137       recently. Thus we conclude it is not in widespread use.)
20138
20139    C++
20140
20141     * GCC 3.2 fixed serveral differences between the C++ ABI implemented
20142       in GCC and the multi-vendor standard, but more have been found
20143       since the release. 3.2.1 adds a new warning, -Wabi, to warn about
20144       code which is affected by these bugs. We will fix these bugs in
20145       some future release, once we are confident that all have been
20146       found; until then, it is our intention to make changes to the ABI
20147       only if they are necessary for correct compilation of C++, as
20148       opposed to conformance to the ABI documents.
20149     * For details on how to build an ABI compliant compiler for GNU/Linux
20150       systems, check the [3]common C++ ABI page.
20151
20152  New Targets and Target Specific Improvements
20153
20154    IA-32
20155
20156     * Fixed a number of bugs in SSE and MMX intrinsics.
20157     * Fixed common compiler crashes with SSE instruction set enabled
20158       (implied by -march=pentium3, pentium4, athlon-xp)
20159     * __m128 and __m128i is not 128bit aligned when used in structures.
20160
20161    x86-64
20162
20163     * A bug whereby the compiler could generate bad code for bzero has
20164       been fixed.
20165     * ABI fixes (implying ABI incompatibilities with previous version in
20166       some corner cases)
20167     * Fixed prefetch code generation
20168     __________________________________________________________________
20169
20170GCC 3.2.3
20171
20172   3.2.3 is a bug fix release only; there are no new features that were
20173   not present in GCC 3.2.2.
20174
20175  Bug Fixes
20176
20177   This section lists the problem reports (PRs) from GCC's bug tracking
20178   system that are known to be fixed in the 3.2.3 release. This list might
20179   not be complete (that is, it is possible that some PRs that have been
20180   fixed are not listed here), and some of the titles have been changed to
20181   make them more clear.
20182
20183    Internal Compiler Errors (multi-platform)
20184
20185     * [4]3782: (c++) -quiet -fstats produces a segmentation fault in
20186       cc1plus
20187     * [5]6440: (c++) template specializations cause ICE
20188     * [6]7050: (c++) ICE on: (i ? get_string() : throw)
20189     * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c)
20190     * [8]7982: (c++) ICE due to infinite recursion (using STL set)
20191     * [9]8068: exceedingly high (infinite) memory usage
20192     * [10]8178: ICE with __builtin_ffs
20193     * [11]8396: ICE in copy_to_mode_reg, in explow.c
20194     * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c
20195     * [13]9768: ICE when optimizing inline code at -O2
20196     * [14]9798: (c++) Infinite recursion (segfault) in
20197       cp/decl.c:push_using_directive with recursive using directives
20198     * [15]9799: mismatching structure initializer with nested flexible
20199       array member: ICE
20200     * [16]9928: ICE on duplicate enum declaration
20201     * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects
20202       sparc, alpha)
20203     * [18]10352: ICE in find_reloads_toplev
20204     * [19]10336: ICE with -Wunreachable-code
20205
20206    C/optimizer bugs:
20207
20208     * [20]8224: Incorrect joining of signed and unsigned division
20209     * [21]8613: -O2 produces wrong code with builtin strlen and
20210       postincrements
20211     * [22]8828: gcc reports some code is unreachable when it is not
20212     * [23]9226: GCSE breaking argument passing
20213     * [24]9853: miscompilation of non-constant structure initializer
20214     * [25]9797: C99-style struct initializers are miscompiled
20215     * [26]9967: Some standard C function calls should not be replaced
20216       when optimizing for size
20217     * [27]10116: ce2: invalid merge of join_bb in the context of switch
20218       statements
20219     * [28]10171: wrong code for inlined function
20220     * [29]10175: -Wunreachable-code doesn't work for single lines
20221
20222    C++ compiler and library:
20223
20224     * [30]8316: Confusing diagnostic for code that misuses conversion
20225       operators
20226     * [31]9169: filebuf output fails if codecvt<>::out returns noconv
20227     * [32]9420: incomplete type incorrectly reported
20228     * [33]9459: typeof in return type specification of template not
20229       supported
20230     * [34]9507: filebuf::open handles ios_base::ate incorrectly
20231     * [35]9538: Out-of-bounds memory access in streambuf::sputbackc
20232     * [36]9602: Total confusion about template/friend/virtual/abstract
20233     * [37]9993: destructor not called for local object created within and
20234       returned from infinite loop
20235     * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2
20236       system
20237
20238    Java compiler and library:
20239
20240     * [39]9652: libgcj build fails on irix6.5.1[78]
20241     * [40]10144: gas on solaris complains about bad .stabs lines for
20242       java, native as unaffected
20243
20244    x86-specific (Intel/AMD):
20245
20246     * [41]8746: gcc miscompiles Linux kernel ppa driver on x86
20247     * [42]9888: -mcpu=k6 -Os produces out of range loop instructions
20248     * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu
20249       failed
20250     * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib)
20251       failed
20252
20253    SPARC-specific:
20254
20255     * [45]7784: [Sparc] ICE in extract_insn, in recog.c
20256     * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in
20257       unroll.c
20258     * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc
20259     * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in
20260       execute/loop-2d.c
20261     * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc
20262     * [50]9414: Scheduling bug on Ultrasparc
20263     * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64
20264
20265    m68k-specific:
20266
20267     * [52]7248: broken "inclusive or" code
20268     * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1
20269
20270    PowerPC-specific:
20271
20272     * [54]9732: Wrong code with -O2 -fPIC
20273     * [55]10073: ICE: powerpc cannot split insn
20274
20275    Alpha-specific:
20276
20277     * [56]7702: optimization problem on a DEC alpha under OSF1
20278     * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system
20279
20280    HP-specific:
20281
20282     * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275)
20283     * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10
20284       (missing symbol)
20285     * [60]10271: Floating point args don't get reloaded across function
20286       calls with -O2
20287
20288    MIPS specific:
20289
20290     * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in
20291       compile/920501-4.c
20292
20293    CRIS specific:
20294
20295     * [62]10377: gcc-3.2.2 creates bad assembler code for cris
20296
20297    Miscellaneous and minor bugs:
20298
20299     * [63]6955: collect2 says "core dumped" when there is no core
20300     __________________________________________________________________
20301
20302GCC 3.2.2
20303
20304   Beginning with 3.2.2, GCC's Makefile suite supports redirection of make
20305   install by means of the DESTDIR variable. Parts of the GCC tree have
20306   featured that support long before, but now it is available even from
20307   the top level.
20308
20309   Other than that, GCC 3.2.2 is a bug fix release only; there are no new
20310   features that were not present in GCC 3.2.1.
20311
20312  Bug Fixes
20313
20314   On the following i386-based systems GCC 3.2.1 broke the C ABI wrt.
20315   functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped
20316   with FreeBSD 5.0 does not have this problem), Interix, a.out-based
20317   GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI
20318   change, and thus restores ABI-compatibility with previous releases
20319   (except GCC 3.2.1) on these platforms.
20320
20321   This section lists the problem reports (PRs) from GCC's bug tracking
20322   system that are known to be fixed in the 3.2.2 release. This list might
20323   not be complete (that is, it is possible that some PRs that have been
20324   fixed are not listed here) and some of the titles have been changed to
20325   make them more clear.
20326
20327    Internal Compiler Errors (multi-platform)
20328
20329     * [64]5919: (c++) ICE when passing variable array to template
20330       function
20331     * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=)
20332     * [66]7507: ICE with -O2 when address of called function is a
20333       complicated expression
20334     * [67]7622: ICE with nested inline functions if function's address is
20335       taken
20336     * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR
20337       [69]9258)
20338     * [70]8031: (c++) ICE in code comparing typeids and casting from
20339       virtual base
20340     * [71]8275: ICE in simplify_subreg
20341     * [72]8332: (c++) builtin strlen/template interaction causes ICE
20342     * [73]8372: (c++) ICE on explicit call of destructor
20343     * [74]8439: (c, not c++) empty struct causes ICE
20344     * [75]8442: (c++) ICE with nested template classes
20345     * [76]8518: ICE when compiling mplayer ("extern inline" issue)
20346     * [77]8615: (c++) ICE with out-of-range character constant template
20347       argument
20348     * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307
20349     * [79]8799: (c++) ICE: error reporting routines re-entered
20350     * [80]9328: (c++) ICE with typeof(X) for overloaded X
20351     * [81]9465: (preprocessor) cpp -traditional ICE on null bytes
20352
20353    C++ (compiler and library) bugs
20354
20355     * [82]47: scoping in nested classes is broken
20356     * [83]6745: problems with iostream rdbuf() member function
20357     * [84]8214: conversion from const char* const to char* sometimes
20358       accepted illegally
20359     * [85]8493: builtin strlen and overload resolution (same bug as
20360       [86]8332)
20361     * [87]8503: strange behaviour of function types
20362     * [88]8727: compiler confused by inheritance from an anonymous struct
20363     * [89]7445: poor performance of std::locale::classic() in
20364       multi-threaded applications
20365     * [90]8230: mishandling of overflow in vector<T>::resize
20366     * [91]8399: sync_with_stdio(false) breaks unformatted input
20367     * [92]8662: illegal access of private member of unnamed class is
20368       accepted
20369     * [93]8707: "make distclean" fails in libstdc++-v3 directory
20370     * [94]8708: __USE_MALLOC doesn't work
20371     * [95]8790: Use of non-thread-safe strtok in src/localename.cc
20372     * [96]8887: Bug in date formats with --enable-clocale=generic
20373     * [97]9076: Call Frame Instructions are not handled correctly during
20374       unwind operation
20375     * [98]9151: std::setprecision limited to 16 digits when outputting a
20376       double to a stream
20377     * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers
20378     * [100]9269: libstdc++ headers: explicit specialization of function
20379       must precede its first use
20380     * [101]9322: return value of basic_streambuf<>::getloc affected by
20381       locale::global
20382     * [102]9433: segfault in runtime support for dynamic_cast
20383
20384    C and optimizer bugs
20385
20386     * [103]8032: GCC incorrectly initializes static structs that have
20387       flexible arrays
20388     * [104]8639: simple arithmetic expression broken
20389     * [105]8794: optimization improperly eliminates certain expressions
20390     * [106]8832: traditional "asm volatile" code is illegally optimized
20391     * [107]8988: loop optimizer bug: with -O2, code is generated that
20392       segfaults (found on i386, bug present for all platforms)
20393     * [108]9492: structure copy clobbers subsequent stores to structure
20394
20395    Objective-C bugs
20396
20397     * [109]9267: Objective-C parser won't build with newer bison versions
20398       (e.g. 1.875)
20399
20400    Ada bugs
20401
20402     * [110]8344: Ada build problem due to conflict between gcc/final.o,
20403       gcc/ada/final.o
20404
20405    Preprocessor bugs
20406
20407     * [111]8524: _Pragma within macros is improperly expanded
20408     * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with
20409       -fshort-wchar
20410
20411    ARM-specific
20412
20413     * [113]9090: arm ICE with >= -O2; regression from gcc-2.95
20414
20415    x86-specific (Intel/AMD)
20416
20417     * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction)
20418     * [115]8599: loop unroll bug with -march=k6-3
20419     * [116]9506: ABI breakage in structure return (affects BSD and
20420       Cygwin, but not GNU/Linux)
20421
20422    FreeBSD 5.0 specific
20423
20424     * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0
20425
20426    RTEMS-specific
20427
20428     * [118]9292: hppa1.1-rtems configurery problems
20429     * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug
20430     * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue
20431     * [121]9296: gthr-rtems regression
20432     * [122]9316: powerpc-rtems: extending multilibs
20433
20434    HP-PA specific
20435
20436     * [123]9493: ICE with -O2 when building a simple function
20437
20438    Documentation
20439
20440     * [124]7341: hyperlink to gcov in GCC documentation doesn't work
20441     * [125]8947: Please add a warning about "-malign-double" in docs
20442     * [126]7448, [127]8882: typo cleanups
20443     __________________________________________________________________
20444
20445GCC 3.2.1
20446
20447   3.2.1 adds a new warning, -Wabi. This option warns when GNU C++
20448   generates code that is known not to be binary-compatible with the
20449   vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included
20450   in the distribution, for details.
20451
20452   This release also removes an old GCC extension, "naming types", and the
20453   documentation now directs users to use a different GCC extension,
20454   __typeof__, instead. The feature had evidently been broken for a while.
20455
20456   Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and
20457   the new warning there are no new features that were not present in GCC
20458   3.2.
20459
20460   In addition, the previous fix for [128]PR 7445 (poor performance of
20461   std::locale::classic() in multi-threaded applications) was reverted
20462   ("unfixed"), because the "fix" was not thread-safe.
20463
20464  Bug Fixes
20465
20466   This section lists the problem reports (PRs) from GCC's bug tracking
20467   system that are known to be fixed in the 3.2.1 release. This list might
20468   not be complete (that is, it is possible that some PRs that have been
20469   fixed are not listed here). As you can see, the number of bug fixes is
20470   quite large, so it is strongly recommended that users of earlier GCC
20471   3.x releases upgrade to GCC 3.2.1.
20472
20473    Internal Compiler Errors (multi-platform)
20474
20475     * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c
20476     * [130]5661: (c++) ICE instantiating template on array of unknown
20477       size (bad code)
20478     * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on
20479       64-bit platforms
20480     * [132]6994: (c++) ICE in find_function_data
20481     * [133]7150: preprocessor: GCC -dM -E gives an ICE
20482     * [134]7160: ICE when optimizing branches without a return value
20483     * [135]7228: (c++) ICE when using member template and template
20484       function
20485     * [136]7266: (c++) ICE with -pedantic on missing typename
20486     * [137]7353: ICE from use of "Naming Types" extension, see above
20487     * [138]7411: ICE in instantiate_virtual_regs_1, in function.c
20488     * [139]7478: (c++) ICE on static_cast inside template
20489     * [140]7526: preprocessor core dump when _Pragma implies #pragma
20490       dependency
20491     * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803
20492       is a duplicate)
20493     * [143]7754: (c++) ICE on union with template parameter
20494     * [144]7788: (c++) redeclaring a definition as an incomplete class
20495       causes ICE
20496     * [145]8031: (c++) ICE in comptypes, in cp/typeck.c
20497     * [146]8055: preprocessor dies with SIG11 when building FreeBSD
20498       kernel
20499     * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related
20500       variables
20501     * [148]8134: (c++) ICE in force_store_init_value on legal code
20502     * [149]8149: (c++) ICE on incomplete type
20503     * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array
20504       initialization
20505
20506    C++ (compiler and library) bugs
20507
20508     * [151]5607: No pointer adjustment in covariant return types
20509     * [152]6579: Infinite loop with statement expressions in member
20510       initialization
20511     * [153]6803: Default copy constructor bug in GCC 3.1
20512     * [154]7176: g++ confused by friend and static member with same name
20513     * [155]7188: Segfault with template class and recursive (incorrect)
20514       initializer list
20515     * [156]7306: Regression: GCC 3.x fails to compile code with virtual
20516       inheritance if a method has a variable number of arguments
20517     * [157]7461: ctype<char>::classic_table() returns offset array on
20518       Cygwin
20519     * [158]7524: f(const float arg[3]) fails
20520     * [159]7584: Erroneous ambiguous base error on using declaration
20521     * [160]7676: Member template overloading problem
20522     * [161]7679: infinite loop when a right parenthesis is missing
20523     * [162]7811: default locale not taken from environment
20524     * [163]7961: compare( char *) implemented incorrectly in
20525       basic_string<>
20526     * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if
20527       streambuf::underflow() leaves gptr() NULL (dups: [165]8127,
20528       [166]6745)
20529     * [167]8096: deque::at() throws std::range_error instead of
20530       std::out_of_range
20531     * [168]8127: cout << cin.rdbuf() infinite loop
20532     * [169]8218: Excessively large memory consumed for classes with large
20533       array members
20534     * [170]8287: GCC 3.2: Destructor called for non-constructed local
20535       object
20536     * [171]8347: empty vector range used in string construction causes
20537       core dump
20538     * [172]8348: fail() flag is set in istringstream when eof() flag is
20539       set
20540     * [173]8391: regression: infinite loop in cp/decl2.c(finish_file)
20541
20542    C and optimizer bugs
20543
20544     * [174]6627: -fno-align-functions doesn't seem to disable function
20545       alignment
20546     * [175]6631: life_analysis misoptimizes code to initialize fields of
20547       a structure
20548     * [176]7102: unsigned char division results in floating exception
20549     * [177]7120: Run once loop should *always* be unrolled
20550       (pessimization)
20551     * [178]7209: Bug involving array referencing and ?: operator
20552     * [179]7515: invalid inlining of global function with -O3
20553     * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test
20554     * [181]8467: bug in sibling call optimization
20555
20556    Preprocessor bugs
20557
20558     * [182]4890: incorrect line markers from the traditional preprocessor
20559     * [183]7357: -M option omits system headers files (making it the same
20560       as -MM)
20561     * [184]7358: Changes to Sun's make Dependencies
20562     * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as
20563       C headers
20564     * [186]7862: Interrupting GCC -MD removes .d file but not .o
20565     * [187]8190: Failed compilation deletes -MD dependency file
20566     * [188]8524: _Pragma within macro is improperly expanded
20567
20568    x86 specific (Intel/AMD)
20569
20570     * [189]5351: (i686-only) function pass-by-value structure copy
20571       corrupts stack ([190]7591 is a duplicate)
20572     * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with
20573       -march=pentium3/pentium2/athlon (these are all the same underlying
20574       bug, in MMX register use)
20575     * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same
20576       as above?)
20577     * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken
20578     * [199]6981: wrong code in 64-bit manipulation on x86
20579     * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__
20580       macro
20581     * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE
20582       intrinsics are broken
20583     * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with
20584       -march=pentium4
20585     * [203]7693: Typo in i386 mmintrin.h header
20586     * [204]7723: ICE - Pentium3 sse - GCC 3.2
20587     * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse
20588     * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3
20589
20590    PowerPC specific
20591
20592     * [207]5967: GCC bug when profiling nested functions on powerpc
20593     * [208]6984: wrong code generated with -O2, -O3, -Os for do-while
20594       loop on PowerPC
20595     * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5
20596     * [210]7130: miscompiled code for GCC-3.1 on
20597       powerpc-unknown-linux-gnu with -funroll-all-loops
20598     * [211]7133: PowerPC ICE: unrecognizable insn
20599     * [212]7380: ICE in extract_insn, at recog.c:2148
20600     * [213]8252: ICE on Altivec code with optimization turned on
20601     * [214]8451: Altivec ICE in GCC 3.2
20602
20603    HP/PA specific
20604
20605     * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa
20606
20607    SPARC specific
20608
20609     * [216]6668: when using --disable-multilib, libgcc_s.so is installed
20610       in the wrong place on sparc-solaris
20611     * [217]7151: ICE when compiling for UltraSPARC
20612     * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long
20613       double and -O1
20614     * [219]7842: [REGRESSION] SPARC code gen bug
20615
20616    ARM specific
20617
20618     * [220]7856: [arm] invalid offset in constant pool reference
20619     * [221]7967: optimization produces wrong code (ARM)
20620
20621    Alpha specific
20622
20623     * [222]7374: __builtin_fabsl broken on alpha
20624
20625    IBM s390 specific
20626
20627     * [223]7370: ICE in fixup_var_refs_1 on s390x
20628     * [224]7409: loop optimization bug on s390x-linux-gnu
20629     * [225]8232: s390x: ICE when using bcmp with int length argument
20630
20631    SCO specific
20632
20633     * [226]7623: SCO OpenServer build fails with machmode.def: undefined
20634       symbol: BITS_PER_UNIT
20635
20636    m68k/Coldfire specific
20637
20638     * [227]8314: crtbegin, crtend need to be multilib'ed for this
20639       platform
20640
20641    Documentation
20642
20643     * [228]761: Document some undocumented options
20644     * [229]5610: Fix documentation about invoking SSE instructions
20645       (-mfpmath=sse)
20646     * [230]7484: List -Wmissing-declarations as C-only option
20647     * [231]7531: -mcmodel not documented for x86-64
20648     * [232]8120: Update documentation of bad use of ##
20649     __________________________________________________________________
20650
20651GCC 3.2
20652
20653   3.2 is a small bug fix release, but there is a change to the
20654   application binary interface (ABI), hence the change to the second part
20655   of the version number.
20656
20657   The main purpose of the 3.2 release is to correct a couple of problems
20658   in the C++ ABI, with the intention of providing a stable interface
20659   going forward.  Accordingly, 3.2 is only a small change to 3.1.1.
20660
20661  Bug Fixes
20662
20663    C++
20664
20665     * [233]7320: g++ 3.2 relocation problem
20666     * [234]7470: vtable: virtual function pointers not in declaration
20667       order
20668
20669    libstdc++
20670
20671     * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t
20672     * [236]6503, [237]6642, [238]7186: Problems with comparing or
20673       subtracting various types of const and non-const iterators
20674     * [239]7216: ambiguity with basic_iostream::traits_type
20675     * [240]7220: problem with basic_istream::ignore(0,delimiter)
20676     * [241]7222: locale::operator==() doesn't work on std::locale("")
20677     * [242]7286: placement operator delete issue
20678     * [243]7442: cxxabi.h does not match the C++ ABI
20679     * [244]7445: poor performance of std::locale::classic() in
20680       multi-threaded applications
20681
20682    x86-64 specific
20683
20684     * [245]7291: off-by-one in generated inline bzero code for x86-64
20685
20686
20687    For questions related to the use of GCC, please consult these web
20688    pages and the [246]GCC manuals. If that fails, the
20689    [247]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20690    web pages and the development of GCC are welcome on our developer
20691    list at [248]gcc@gcc.gnu.org. All of [249]our lists have public
20692    archives.
20693
20694   Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and
20695   distribution of this entire article is permitted in any medium,
20696   provided this notice is preserved.
20697
20698   These pages are [251]maintained by the GCC team. Last modified
20699   2022-10-26.
20700
20701References
20702
20703   1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
20704   2. http://gcc.gnu.org/gcc-3.1/changes.html
20705   3. http://gcc.gnu.org/gcc-3.2/c++-abi.html
20706   4. https://gcc.gnu.org/PR3782
20707   5. https://gcc.gnu.org/PR6440
20708   6. https://gcc.gnu.org/PR7050
20709   7. https://gcc.gnu.org/PR7741
20710   8. https://gcc.gnu.org/PR7982
20711   9. https://gcc.gnu.org/PR8068
20712  10. https://gcc.gnu.org/PR8178
20713  11. https://gcc.gnu.org/PR8396
20714  12. https://gcc.gnu.org/PR8674
20715  13. https://gcc.gnu.org/PR9768
20716  14. https://gcc.gnu.org/PR9798
20717  15. https://gcc.gnu.org/PR9799
20718  16. https://gcc.gnu.org/PR9928
20719  17. https://gcc.gnu.org/PR10114
20720  18. https://gcc.gnu.org/PR10352
20721  19. https://gcc.gnu.org/PR10336
20722  20. https://gcc.gnu.org/PR8224
20723  21. https://gcc.gnu.org/PR8613
20724  22. https://gcc.gnu.org/PR8828
20725  23. https://gcc.gnu.org/PR9226
20726  24. https://gcc.gnu.org/PR9853
20727  25. https://gcc.gnu.org/PR9797
20728  26. https://gcc.gnu.org/PR9967
20729  27. https://gcc.gnu.org/PR10116
20730  28. https://gcc.gnu.org/PR10171
20731  29. https://gcc.gnu.org/PR10175
20732  30. https://gcc.gnu.org/PR8316
20733  31. https://gcc.gnu.org/PR9169
20734  32. https://gcc.gnu.org/PR9420
20735  33. https://gcc.gnu.org/PR9459
20736  34. https://gcc.gnu.org/PR9507
20737  35. https://gcc.gnu.org/PR9538
20738  36. https://gcc.gnu.org/PR9602
20739  37. https://gcc.gnu.org/PR9993
20740  38. https://gcc.gnu.org/PR10167
20741  39. https://gcc.gnu.org/PR9652
20742  40. https://gcc.gnu.org/PR10144
20743  41. https://gcc.gnu.org/PR8746
20744  42. https://gcc.gnu.org/PR9888
20745  43. https://gcc.gnu.org/PR9638
20746  44. https://gcc.gnu.org/PR9954
20747  45. https://gcc.gnu.org/PR7784
20748  46. https://gcc.gnu.org/PR7796
20749  47. https://gcc.gnu.org/PR8281
20750  48. https://gcc.gnu.org/PR8366
20751  49. https://gcc.gnu.org/PR8726
20752  50. https://gcc.gnu.org/PR9414
20753  51. https://gcc.gnu.org/PR10067
20754  52. https://gcc.gnu.org/PR7248
20755  53. https://gcc.gnu.org/PR8343
20756  54. https://gcc.gnu.org/PR9732
20757  55. https://gcc.gnu.org/PR10073
20758  56. https://gcc.gnu.org/PR7702
20759  57. https://gcc.gnu.org/PR9671
20760  58. https://gcc.gnu.org/PR8694
20761  59. https://gcc.gnu.org/PR9953
20762  60. https://gcc.gnu.org/PR10271
20763  61. https://gcc.gnu.org/PR6362
20764  62. https://gcc.gnu.org/PR10377
20765  63. https://gcc.gnu.org/PR6955
20766  64. https://gcc.gnu.org/PR5919
20767  65. https://gcc.gnu.org/PR7129
20768  66. https://gcc.gnu.org/PR7507
20769  67. https://gcc.gnu.org/PR7622
20770  68. https://gcc.gnu.org/PR7681
20771  69. https://gcc.gnu.org/PR9528
20772  70. https://gcc.gnu.org/PR8031
20773  71. https://gcc.gnu.org/PR8275
20774  72. https://gcc.gnu.org/PR8332
20775  73. https://gcc.gnu.org/PR8372
20776  74. https://gcc.gnu.org/PR8439
20777  75. https://gcc.gnu.org/PR8442
20778  76. https://gcc.gnu.org/PR8518
20779  77. https://gcc.gnu.org/PR8615
20780  78. https://gcc.gnu.org/PR8663
20781  79. https://gcc.gnu.org/PR8799
20782  80. https://gcc.gnu.org/PR9328
20783  81. https://gcc.gnu.org/PR9465
20784  82. https://gcc.gnu.org/PR47
20785  83. https://gcc.gnu.org/PR6745
20786  84. https://gcc.gnu.org/PR8214
20787  85. https://gcc.gnu.org/PR8493
20788  86. https://gcc.gnu.org/PR8332
20789  87. https://gcc.gnu.org/PR8503
20790  88. https://gcc.gnu.org/PR8727
20791  89. https://gcc.gnu.org/PR7445
20792  90. https://gcc.gnu.org/PR8230
20793  91. https://gcc.gnu.org/PR8399
20794  92. https://gcc.gnu.org/PR8662
20795  93. https://gcc.gnu.org/PR8707
20796  94. https://gcc.gnu.org/PR8708
20797  95. https://gcc.gnu.org/PR8790
20798  96. https://gcc.gnu.org/PR8887
20799  97. https://gcc.gnu.org/PR9076
20800  98. https://gcc.gnu.org/PR9151
20801  99. https://gcc.gnu.org/PR9168
20802 100. https://gcc.gnu.org/PR9269
20803 101. https://gcc.gnu.org/PR9322
20804 102. https://gcc.gnu.org/PR9433
20805 103. https://gcc.gnu.org/PR8032
20806 104. https://gcc.gnu.org/PR8639
20807 105. https://gcc.gnu.org/PR8794
20808 106. https://gcc.gnu.org/PR8832
20809 107. https://gcc.gnu.org/PR8988
20810 108. https://gcc.gnu.org/PR9492
20811 109. https://gcc.gnu.org/PR9267
20812 110. https://gcc.gnu.org/PR8344
20813 111. https://gcc.gnu.org/PR8524
20814 112. https://gcc.gnu.org/PR8880
20815 113. https://gcc.gnu.org/PR9090
20816 114. https://gcc.gnu.org/PR8588
20817 115. https://gcc.gnu.org/PR8599
20818 116. https://gcc.gnu.org/PR9506
20819 117. https://gcc.gnu.org/PR9484
20820 118. https://gcc.gnu.org/PR9292
20821 119. https://gcc.gnu.org/PR9293
20822 120. https://gcc.gnu.org/PR9295
20823 121. https://gcc.gnu.org/PR9296
20824 122. https://gcc.gnu.org/PR9316
20825 123. https://gcc.gnu.org/PR9493
20826 124. https://gcc.gnu.org/PR7341
20827 125. https://gcc.gnu.org/PR8947
20828 126. https://gcc.gnu.org/PR7448
20829 127. https://gcc.gnu.org/PR8882
20830 128. https://gcc.gnu.org/PR7445
20831 129. https://gcc.gnu.org/PR2521
20832 130. https://gcc.gnu.org/PR5661
20833 131. https://gcc.gnu.org/PR6419
20834 132. https://gcc.gnu.org/PR6994
20835 133. https://gcc.gnu.org/PR7150
20836 134. https://gcc.gnu.org/PR7160
20837 135. https://gcc.gnu.org/PR7228
20838 136. https://gcc.gnu.org/PR7266
20839 137. https://gcc.gnu.org/PR7353
20840 138. https://gcc.gnu.org/PR7411
20841 139. https://gcc.gnu.org/PR7478
20842 140. https://gcc.gnu.org/PR7526
20843 141. https://gcc.gnu.org/PR7721
20844 142. https://gcc.gnu.org/PR7803
20845 143. https://gcc.gnu.org/PR7754
20846 144. https://gcc.gnu.org/PR7788
20847 145. https://gcc.gnu.org/PR8031
20848 146. https://gcc.gnu.org/PR8055
20849 147. https://gcc.gnu.org/PR8067
20850 148. https://gcc.gnu.org/PR8134
20851 149. https://gcc.gnu.org/PR8149
20852 150. https://gcc.gnu.org/PR8160
20853 151. https://gcc.gnu.org/PR5607
20854 152. https://gcc.gnu.org/PR6579
20855 153. https://gcc.gnu.org/PR6803
20856 154. https://gcc.gnu.org/PR7176
20857 155. https://gcc.gnu.org/PR7188
20858 156. https://gcc.gnu.org/PR7306
20859 157. https://gcc.gnu.org/PR7461
20860 158. https://gcc.gnu.org/PR7524
20861 159. https://gcc.gnu.org/PR7584
20862 160. https://gcc.gnu.org/PR7676
20863 161. https://gcc.gnu.org/PR7679
20864 162. https://gcc.gnu.org/PR7811
20865 163. https://gcc.gnu.org/PR7961
20866 164. https://gcc.gnu.org/PR8071
20867 165. https://gcc.gnu.org/PR8127
20868 166. https://gcc.gnu.org/PR6745
20869 167. https://gcc.gnu.org/PR8096
20870 168. https://gcc.gnu.org/PR8127
20871 169. https://gcc.gnu.org/PR8218
20872 170. https://gcc.gnu.org/PR8287
20873 171. https://gcc.gnu.org/PR8347
20874 172. https://gcc.gnu.org/PR8348
20875 173. https://gcc.gnu.org/PR8391
20876 174. https://gcc.gnu.org/PR6627
20877 175. https://gcc.gnu.org/PR6631
20878 176. https://gcc.gnu.org/PR7102
20879 177. https://gcc.gnu.org/PR7120
20880 178. https://gcc.gnu.org/PR7209
20881 179. https://gcc.gnu.org/PR7515
20882 180. https://gcc.gnu.org/PR7814
20883 181. https://gcc.gnu.org/PR8467
20884 182. https://gcc.gnu.org/PR4890
20885 183. https://gcc.gnu.org/PR7357
20886 184. https://gcc.gnu.org/PR7358
20887 185. https://gcc.gnu.org/PR7602
20888 186. https://gcc.gnu.org/PR7862
20889 187. https://gcc.gnu.org/PR8190
20890 188. https://gcc.gnu.org/PR8524
20891 189. https://gcc.gnu.org/PR5351
20892 190. https://gcc.gnu.org/PR7591
20893 191. https://gcc.gnu.org/PR6845
20894 192. https://gcc.gnu.org/PR7034
20895 193. https://gcc.gnu.org/PR7124
20896 194. https://gcc.gnu.org/PR7174
20897 195. https://gcc.gnu.org/PR7134
20898 196. https://gcc.gnu.org/PR7375
20899 197. https://gcc.gnu.org/PR7390
20900 198. https://gcc.gnu.org/PR6890
20901 199. https://gcc.gnu.org/PR6981
20902 200. https://gcc.gnu.org/PR7242
20903 201. https://gcc.gnu.org/PR7396
20904 202. https://gcc.gnu.org/PR7630
20905 203. https://gcc.gnu.org/PR7693
20906 204. https://gcc.gnu.org/PR7723
20907 205. https://gcc.gnu.org/PR7951
20908 206. https://gcc.gnu.org/PR8146
20909 207. https://gcc.gnu.org/PR5967
20910 208. https://gcc.gnu.org/PR6984
20911 209. https://gcc.gnu.org/PR7114
20912 210. https://gcc.gnu.org/PR7130
20913 211. https://gcc.gnu.org/PR7133
20914 212. https://gcc.gnu.org/PR7380
20915 213. https://gcc.gnu.org/PR8252
20916 214. https://gcc.gnu.org/PR8451
20917 215. https://gcc.gnu.org/PR7250
20918 216. https://gcc.gnu.org/PR6668
20919 217. https://gcc.gnu.org/PR7151
20920 218. https://gcc.gnu.org/PR7335
20921 219. https://gcc.gnu.org/PR7842
20922 220. https://gcc.gnu.org/PR7856
20923 221. https://gcc.gnu.org/PR7967
20924 222. https://gcc.gnu.org/PR7374
20925 223. https://gcc.gnu.org/PR7370
20926 224. https://gcc.gnu.org/PR7409
20927 225. https://gcc.gnu.org/PR8232
20928 226. https://gcc.gnu.org/PR7623
20929 227. https://gcc.gnu.org/PR8314
20930 228. https://gcc.gnu.org/PR761
20931 229. https://gcc.gnu.org/PR5610
20932 230. https://gcc.gnu.org/PR7484
20933 231. https://gcc.gnu.org/PR7531
20934 232. https://gcc.gnu.org/PR8120
20935 233. https://gcc.gnu.org/PR7320
20936 234. https://gcc.gnu.org/PR7470
20937 235. https://gcc.gnu.org/PR6410
20938 236. https://gcc.gnu.org/PR6503
20939 237. https://gcc.gnu.org/PR6642
20940 238. https://gcc.gnu.org/PR7186
20941 239. https://gcc.gnu.org/PR7216
20942 240. https://gcc.gnu.org/PR7220
20943 241. https://gcc.gnu.org/PR7222
20944 242. https://gcc.gnu.org/PR7286
20945 243. https://gcc.gnu.org/PR7442
20946 244. https://gcc.gnu.org/PR7445
20947 245. https://gcc.gnu.org/PR7291
20948 246. https://gcc.gnu.org/onlinedocs/
20949 247. mailto:gcc-help@gcc.gnu.org
20950 248. mailto:gcc@gcc.gnu.org
20951 249. https://gcc.gnu.org/lists.html
20952 250. https://www.fsf.org/
20953 251. https://gcc.gnu.org/about.html
20954======================================================================
20955http://gcc.gnu.org/gcc-3.1/index.html
20956
20957                                    GCC 3.1
20958
20959   (This release series is no longer supported.)
20960
20961   July 27, 2002
20962
20963   The [1]GNU project and the GCC developers are pleased to announce the
20964   release of GCC 3.1.1.
20965
20966   The links below still apply to GCC 3.1.1.
20967
20968   May 15, 2002
20969
20970   The [2]GNU project and the GCC developers are pleased to announce the
20971   release of GCC 3.1.
20972
20973   GCC used to stand for the GNU C Compiler, but since the compiler
20974   supports several other languages aside from C, it now stands for the
20975   GNU Compiler Collection.
20976
20977   A list of [3]successful builds is updated as new information becomes
20978   available.
20979
20980   The GCC developers would like to thank the numerous people that have
20981   contributed [4]new features, improvements, bug fixes, and other changes
20982   as well as test results to GCC. This [5]amazing group of volunteers is
20983   what makes GCC successful.
20984
20985   For additional information about GCC please refer to the [6]GCC project
20986   web site or contact the [7]GCC development mailing list.
20987
20988   To obtain GCC please use [8]our mirror sites, or our CVS server.
20989     __________________________________________________________________
20990
20991
20992    For questions related to the use of GCC, please consult these web
20993    pages and the [9]GCC manuals. If that fails, the
20994    [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20995    web pages and the development of GCC are welcome on our developer
20996    list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
20997    archives.
20998
20999   Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
21000   distribution of this entire article is permitted in any medium,
21001   provided this notice is preserved.
21002
21003   These pages are [14]maintained by the GCC team. Last modified
21004   2022-10-26.
21005
21006References
21007
21008   1. http://www.gnu.org/
21009   2. http://www.gnu.org/
21010   3. http://gcc.gnu.org/gcc-3.1/buildstat.html
21011   4. http://gcc.gnu.org/gcc-3.1/changes.html
21012   5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
21013   6. http://gcc.gnu.org/index.html
21014   7. mailto:gcc@gcc.gnu.org
21015   8. http://gcc.gnu.org/mirrors.html
21016   9. https://gcc.gnu.org/onlinedocs/
21017  10. mailto:gcc-help@gcc.gnu.org
21018  11. mailto:gcc@gcc.gnu.org
21019  12. https://gcc.gnu.org/lists.html
21020  13. https://www.fsf.org/
21021  14. https://gcc.gnu.org/about.html
21022======================================================================
21023http://gcc.gnu.org/gcc-3.1/changes.html
21024
21025                             GCC 3.1 Release Series
21026                        Changes, New Features, and Fixes
21027
21028Additional changes in GCC 3.1.1
21029
21030     * A bug related to how structures and unions are returned has been
21031       fixed for powerpc-*-netbsd*.
21032     * An important bug in the implementation of -fprefetch-loop-arrays
21033       has been fixed. Previously the optimization prefetched random
21034       blocks of memory for most targets except for i386.
21035     * The Java compiler now compiles Java programs much faster and also
21036       works with parallel make.
21037     * Nested functions have been fixed for mips*-*-netbsd*.
21038     * Some missing floating point support routines have beed added for
21039       mips*-*-netbsd*.
21040     * This [1]message gives additional information about the bugs fixed
21041       in this release.
21042
21043Caveats
21044
21045     * The -traditional C compiler option has been deprecated and will be
21046       removed in GCC 3.3. (It remains possible to preprocess non-C code
21047       with the traditional preprocessor.)
21048     * The default debugging format for most ELF platforms (including
21049       GNU/Linux and FreeBSD; notable exception is Solaris) has changed
21050       from stabs to DWARF2. This requires GDB 5.1.1 or later.
21051
21052General Optimizer Improvements
21053
21054     * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat,
21055       and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure
21056       for profile driven optimizations.
21057       Options -fprofile-arcs and -fbranch-probabilities can now be used
21058       to improve speed of the generated code by profiling the actual
21059       program behaviour on typical runs. In the absence of profile info
21060       the compiler attempts to guess the profile statically.
21061     * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to
21062       monitor performance of the generated code.
21063       According to the SPECInt2000 results on an AMD Athlon CPU, the code
21064       generated by GCC 3.1 is 6% faster on the average (8.2% faster with
21065       profile feedback) compared to GCC 3.0. The code produced by GCC 3.0
21066       is about 2.1% faster compared to 2.95.3. Tests were done using the
21067       -O2 -march=athlon command-line options.
21068     * Alexandre Oliva, of Red Hat, has generalized the tree inlining
21069       infrastructure developed by CodeSourcery, LLC for the C++ front
21070       end, so that it is now used in the C front end too. Inlining
21071       functions as trees exposes them earlier to the compiler, giving it
21072       more opportunities for optimization.
21073     * Support for data prefetching instructions has been added to the GCC
21074       back end and several targets. A new __builtin_prefetch intrinsic is
21075       available to explicitly insert prefetch instructions and
21076       experimental support for loop array prefetching has been added (see
21077       -fprefetch-loop-array documentation).
21078     * Support for emitting debugging information for macros has been
21079       added for DWARF2. It is activated using -g3.
21080
21081New Languages and Language specific improvements
21082
21083  C/C++
21084
21085     * A few more [4]ISO C99 features.
21086     * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0.
21087     * The preprocessor's symbol table has been merged with the symbol
21088       table of the C, C++ and Objective-C front ends.
21089     * The preprocessor consumes less memory than the preprocessor in GCC
21090       3.0, often significantly so. On normal input files, it typically
21091       consumes less memory than pre-3.0 cccp-based GCC, too.
21092
21093  C++
21094
21095     * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std
21096       was a workaround to allow std compliant code to work with the
21097       non-std compliant libstdc++-v2. libstdc++-v3 is std compliant.
21098     * The C++ ABI has been fixed so that void (A::*)() const is mangled
21099       as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only
21100       affects pointer to cv-qualified member function types.
21101     * The C++ ABI has been changed to correctly handle this code:
21102    struct A {
21103      void operator delete[] (void *, size_t);
21104    };
21105
21106    struct B : public A {
21107    };
21108
21109    new B[10];
21110
21111       The amount of storage allocated for the array will be greater than
21112       it was in 3.0, in order to store the number of elements in the
21113       array, so that the correct size can be passed to operator delete[]
21114       when the array is deleted. Previously, the value passed to operator
21115       delete[] was unpredictable.
21116       This change will only affect code that declares a two-argument
21117       operator delete[] with a second parameter of type size_t in a base
21118       class, and does not override that definition in a derived class.
21119     * The C++ ABI has been changed so that:
21120    struct A {
21121      void operator delete[] (void *, size_t);
21122      void operator delete[] (void *);
21123    };
21124
21125       does not cause unnecessary storage to be allocated when an array of
21126       A objects is allocated.
21127       This change will only affect code that declares both of these forms
21128       of operator delete[], and declared the two-argument form before the
21129       one-argument form.
21130     * The C++ ABI has been changed so that when a parameter is passed by
21131       value, any cleanup for that parameter is performed in the caller,
21132       as specified by the ia64 C++ ABI, rather than the called function
21133       as before. As a result, classes with a non-trivial destructor but a
21134       trivial copy constructor will be passed and returned by invisible
21135       reference, rather than by bitwise copy as before.
21136     * G++ now supports the "named return value optimization": for code
21137       like
21138    A f () {
21139      A a;
21140      ...
21141      return a;
21142    }
21143
21144       G++ will allocate a in the return value slot, so that the return
21145       becomes a no-op. For this to work, all return statements in the
21146       function must return the same variable.
21147     * Improvements to the C++ library are listed in [5]the libstdc++-v3
21148       FAQ.
21149
21150  Objective-C
21151
21152     * Annoying linker warnings (due to incorrect code being generated)
21153       have been fixed.
21154     * If a class method cannot be found, the compiler no longer issues a
21155       warning if a corresponding instance method exists in the root
21156       class.
21157     * Forward @protocol declarations have been fixed.
21158     * Loading of categories has been fixed in certain situations (GNU run
21159       time only).
21160     * The class lookup in the run-time library has been rewritten so that
21161       class method dispatch is more than twice as fast as it used to be
21162       (GNU run time only).
21163
21164  Java
21165
21166     * libgcj now includes RMI, java.lang.ref.*, javax.naming, and
21167       javax.transaction.
21168     * Property files and other system resources can be compiled into
21169       executables which use libgcj using the new gcj --resource feature.
21170     * libgcj has been ported to more platforms. In particular there is
21171       now a mostly-functional mingw32 (Windows) target port.
21172     * JNI and CNI invocation interfaces were implemented, so gcj-compiled
21173       Java code can now be called from a C/C++ application.
21174     * gcj can now use builtin functions for certain known methods, for
21175       instance Math.cos.
21176     * gcj can now automatically remove redundant array-store checks in
21177       some common cases.
21178     * The --no-store-checks optimization option was added. This can be
21179       used to omit runtime store checks for code which is known not to
21180       throw ArrayStoreException
21181     * The following third party interface standards were added to libgcj:
21182       org.w3c.dom and org.xml.sax.
21183     * java.security has been merged with GNU Classpath. The new package
21184       is now JDK 1.2 compliant, and much more complete.
21185     * A bytecode verifier was added to the libgcj interpreter.
21186     * java.lang.Character was rewritten to comply with the Unicode 3.0
21187       standard, and improve performance.
21188     * Partial support for many more locales was added to libgcj.
21189     * Socket timeouts have been implemented.
21190     * libgcj has been merged into a single shared library. There are no
21191       longer separate shared libraries for the garbage collector and
21192       zlib.
21193     * Several performance improvements were made to gcj and libgcj:
21194          + Hash synchronization (thin locks)
21195          + A special allocation path for finalizer-free objects
21196          + Thread-local allocation
21197          + Parallel GC, and other GC tweaks
21198
21199  Fortran
21200
21201   Fortran improvements are listed in [6]the Fortran documentation.
21202
21203  Ada
21204
21205   [7]AdaCore, has contributed its GNAT Ada 95 front end and associated
21206   tools. The GNAT compiler fully implements the Ada language as defined
21207   by the ISO/IEC 8652 standard.
21208
21209   Please note that the integration of the Ada front end is still work in
21210   progress.
21211
21212New Targets and Target Specific Improvements
21213
21214     * Hans-Peter Nilsson has contributed a port to MMIX, the CPU
21215       architecture used in new editions of Donald E. Knuth's The Art of
21216       Computer Programming.
21217     * Axis Communications has contributed its port to the CRIS CPU
21218       architecture, used in the ETRAX system-on-a-chip series.
21219     * Alexandre Oliva, of Red Hat, has contributed a port to the SuperH
21220       SH5 64-bit RISC microprocessor architecture, extending the existing
21221       SH port.
21222     * UltraSPARC is fully supported in 64-bit mode. The option -m64
21223       enables it.
21224     * For compatibility with the Sun compiler #pragma redefine_extname
21225       has been implemented on Solaris.
21226     * The x86 back end has had some noticeable work done to it.
21227          + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas
21228            Jaeger have contributed a port to the AMD x86-64 architecture.
21229            For more information on x86-64 see http://www.x86-64.org.
21230          + The compiler now supports MMX, 3DNow!, SSE, and SSE2
21231            instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will
21232            enable the respective instruction sets. Intel C++ compatible
21233            MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics
21234            will be added in next major release.
21235          + Following those improvements, targets for Pentium MMX, K6-2,
21236            K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were
21237            added. Refer to the documentation on -march= and -mcpu=
21238            options for details.
21239          + For those targets that support it, -mfpmath=sse will cause the
21240            compiler to generate SSE/SSE2 instructions for floating point
21241            math instead of x87 instructions. Usually, this will lead to
21242            quicker code -- especially on the Pentium 4. Note that only
21243            scalar floating point instructions are used and GCC does not
21244            exploit SIMD features yet.
21245          + Prefetch support has been added to the Pentium III, Pentium 4,
21246            K6-2, K6-3, and Athlon series.
21247          + Code generated for floating point to integer conversions has
21248            been improved leading to better performance of many 3D
21249            applications.
21250     * The PowerPC back end has added 64-bit PowerPC GNU/Linux support.
21251     * C++ support for AIX has been improved.
21252     * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the
21253       PowerPC port supporting the AltiVec programming model (SIMD). The
21254       support, though presently useful, is experimental and is expected
21255       to stabilize for 3.2. The support is written to conform to
21256       Motorola's AltiVec specs. See -maltivec.
21257
21258Obsolete Systems
21259
21260   Support for a number of older systems has been declared obsolete in GCC
21261   3.1. Unless there is activity to revive them, the next release of GCC
21262   will have their sources permanently removed.
21263
21264   All configurations of the following processor architectures have been
21265   declared obsolete:
21266     * MIL-STD-1750A, 1750a-*-*
21267     * AMD A29k, a29k-*-*
21268     * Convex, c*-convex-*
21269     * Clipper, clipper-*-*
21270     * Elxsi, elxsi-*-*
21271     * Intel i860, i860-*-*
21272     * Sun picoJava, pj-*-* and pjl-*-*
21273     * Western Electric 32000, we32k-*-*
21274
21275   Most configurations of the following processor architectures have been
21276   declared obsolete, but we are preserving a few systems which may have
21277   active developers. It is unlikely that the remaining systems will
21278   survive much longer unless we see definite signs of port activity.
21279     * Motorola 88000 except
21280          + Generic a.out, m88k-*-aout*
21281          + Generic SVR4, m88k-*-sysv4
21282          + OpenBSD, m88k-*-openbsd*
21283     * NS32k except
21284          + NetBSD, ns32k-*-netbsd*
21285          + OpenBSD, ns32k-*-openbsd*.
21286     * ROMP except
21287          + OpenBSD, romp-*-openbsd*.
21288
21289   Finally, only some configurations of these processor architectures are
21290   being obsoleted.
21291     * Alpha:
21292          + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka
21293            alpha*-*-osf[45], are still supported.)
21294     * ARM:
21295          + RISCiX, arm-*-riscix*.
21296     * i386:
21297          + 386BSD, i?86-*-bsd*
21298          + Chorus, i?86-*-chorusos*
21299          + DG/UX, i?86-*-dgux*
21300          + FreeBSD 1.x, i?86-*-freebsd1.*
21301          + IBM AIX, i?86-*-aix*
21302          + ISC UNIX, i?86-*-isc*
21303          + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld*
21304          + NEXTstep, i?86-next-*
21305          + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose*
21306          + RTEMS/coff, i?86-*-rtemscoff*
21307          + RTEMS/go32, i?86-go32-rtems*
21308          + Sequent/BSD, i?86-sequent-bsd*
21309          + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and
21310            i?86-sequent-sysv3*
21311          + SunOS, i?86-*-sunos*
21312     * Motorola 68000:
21313          + Altos, m68[k0]*-altos-*
21314          + Apollo, m68[k0]*-apollo-*
21315          + Apple A/UX, m68[k0]*-apple-*
21316          + Bull, m68[k0]*-bull-*
21317          + Convergent, m68[k0]*-convergent-*
21318          + Generic SVR3, m68[k0]*-*-sysv3*
21319          + ISI, m68[k0]*-isi-*
21320          + LynxOS, m68[k0]*-*-lynxos*
21321          + NEXT, m68[k0]*-next-*
21322          + RTEMS/coff, m68[k0]*-*-rtemscoff*
21323          + Sony, m68[k0]*-sony-*
21324     * MIPS:
21325          + DEC Ultrix, mips-*-ultrix* and mips-dec-*
21326          + Generic BSD, mips-*-bsd*
21327          + Generic System V, mips-*-sysv*
21328          + IRIX before version 5, mips-sgi-irix[1234]*
21329          + RiscOS, mips-*-riscos*
21330          + Sony, mips-sony-*
21331          + Tandem, mips-tandem-*
21332     * SPARC:
21333          + RTEMS/a.out, sparc-*-rtemsaout*.
21334
21335Documentation improvements
21336
21337     * The old manual ("Using and Porting the GNU Compiler Collection")
21338       has been replaced by a users manual ("Using the GNU Compiler
21339       Collection") and a separate internals reference manual ("GNU
21340       Compiler Collection Internals").
21341     * More complete and much improved documentation about GCC's internal
21342       representation used by the C and C++ front ends.
21343     * Many cleanups and improvements in general.
21344
21345
21346    For questions related to the use of GCC, please consult these web
21347    pages and the [8]GCC manuals. If that fails, the
21348    [9]gcc-help@gcc.gnu.org mailing list might help. Comments on these
21349    web pages and the development of GCC are welcome on our developer
21350    list at [10]gcc@gcc.gnu.org. All of [11]our lists have public
21351    archives.
21352
21353   Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and
21354   distribution of this entire article is permitted in any medium,
21355   provided this notice is preserved.
21356
21357   These pages are [13]maintained by the GCC team. Last modified
21358   2022-10-26.
21359
21360References
21361
21362   1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html
21363   2. http://gcc.gnu.org/news/profiledriven.html
21364   3. http://gcc.gnu.org/benchmarks/
21365   4. http://gcc.gnu.org/c99status.html
21366   5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html
21367   6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html
21368   7. https://www.adacore.com/
21369   8. https://gcc.gnu.org/onlinedocs/
21370   9. mailto:gcc-help@gcc.gnu.org
21371  10. mailto:gcc@gcc.gnu.org
21372  11. https://gcc.gnu.org/lists.html
21373  12. https://www.fsf.org/
21374  13. https://gcc.gnu.org/about.html
21375======================================================================
21376http://gcc.gnu.org/gcc-3.0/index.html
21377
21378                                   GCC 3.0.4
21379
21380   (This release series is no longer supported.)
21381
21382   February 20, 2002
21383
21384   The [1]GNU project and the GCC developers are pleased to announce the
21385   release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0
21386   series.
21387
21388   GCC used to stand for the GNU C Compiler, but since the compiler
21389   supports several other languages aside from C, it now stands for the
21390   GNU Compiler Collection.
21391
21392   GCC 3.0.x has several new optimizations, new targets, new languages and
21393   many other new features, relative to GCC 2.95.x. See the [2]new
21394   features page for a more complete list.
21395
21396   A list of [3]successful builds is updated as new information becomes
21397   available.
21398
21399   The GCC developers would like to thank the numerous people that have
21400   contributed new features, test results, bug fixes, etc to GCC. This
21401   [4]amazing group of volunteers is what makes GCC successful.
21402
21403   And finally, we can't in good conscience fail to mention some
21404   [5]caveats to using GCC 3.0.x.
21405
21406   For additional information about GCC please refer to the [6]GCC project
21407   web site or contact the [7]GCC development mailing list.
21408
21409   To obtain GCC please use [8]our mirror sites, or our CVS server.
21410     __________________________________________________________________
21411
21412Previous 3.0.x Releases
21413
21414   December 20, 2001: GCC 3.0.3 has been released.
21415   October 25, 2001: GCC 3.0.2 has been released.
21416   August 20, 2001: GCC 3.0.1 has been released.
21417   June 18, 2001: GCC 3.0 has been released.
21418
21419
21420    For questions related to the use of GCC, please consult these web
21421    pages and the [9]GCC manuals. If that fails, the
21422    [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
21423    web pages and the development of GCC are welcome on our developer
21424    list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
21425    archives.
21426
21427   Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
21428   distribution of this entire article is permitted in any medium,
21429   provided this notice is preserved.
21430
21431   These pages are [14]maintained by the GCC team. Last modified
21432   2022-10-26.
21433
21434References
21435
21436   1. http://www.gnu.org/
21437   2. http://gcc.gnu.org/gcc-3.0/features.html
21438   3. http://gcc.gnu.org/gcc-3.0/buildstat.html
21439   4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
21440   5. http://gcc.gnu.org/gcc-3.0/caveats.html
21441   6. http://gcc.gnu.org/index.html
21442   7. mailto:gcc@gcc.gnu.org
21443   8. http://gcc.gnu.org/mirrors.html
21444   9. https://gcc.gnu.org/onlinedocs/
21445  10. mailto:gcc-help@gcc.gnu.org
21446  11. mailto:gcc@gcc.gnu.org
21447  12. https://gcc.gnu.org/lists.html
21448  13. https://www.fsf.org/
21449  14. https://gcc.gnu.org/about.html
21450======================================================================
21451http://gcc.gnu.org/gcc-3.0/features.html
21452
21453                              GCC 3.0 New Features
21454
21455Additional changes in GCC 3.0.4
21456
21457     * GCC 3.0 now supports newer versions of the [1]NetBSD operating
21458       system, which use the ELF object file format, on x86 processors.
21459     * Correct debugging information is generated from functions that have
21460       lines from multiple files (e.g. yacc output).
21461     * A fix for whitespace handling in the -traditional preprocessor,
21462       which can affect Fortran.
21463     * Fixes to the exception handling runtime.
21464     * More fixes for bad code generation in C++.
21465     * A fix for shared library generation under AIX 4.3.
21466     * Documentation updates.
21467     * Port of GCC to Tensilica's Xtensa processor contributed.
21468     * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link).
21469
21470Additional changes in GCC 3.0.3
21471
21472     * A fix to correct an accidental change to the PowerPC ABI.
21473     * Fixes for bad code generation on a variety of architectures.
21474     * Improvements to the debugging information generated for C++
21475       classes.
21476     * Fixes for bad code generation in C++.
21477     * A fix to avoid crashes in the C++ demangler.
21478     * A fix to the C++ standard library to avoid buffer overflows.
21479     * Miscellaneous improvements for a variety of architectures.
21480
21481Additional changes in GCC 3.0.2
21482
21483     * Fixes for bad code generation during loop unrolling.
21484     * Fixes for bad code generation by the sibling call optimization.
21485     * Minor improvements to x86 code generation.
21486     * Implementation of function descriptors in C++ vtables for IA64.
21487     * Numerous minor bug-fixes.
21488
21489Additional changes in GCC 3.0.1
21490
21491     * C++ fixes for incorrect code-generation.
21492     * Improved cross-compiling support for the C++ standard library.
21493     * Fixes for some embedded targets that worked in GCC 2.95.3, but not
21494       in GCC 3.0.
21495     * Fixes for various exception-handling bugs.
21496     * A port to the S/390 architecture.
21497
21498General Optimizer Improvements
21499
21500     * [2]Basic block reordering pass.
21501     * New if-conversion pass with support for conditional (predicated)
21502       execution.
21503     * New tail call and sibling call elimination optimizations.
21504     * New register renaming pass.
21505     * New (experimental) [3]static single assignment (SSA) representation
21506       support.
21507     * New dead-code elimination pass implemented using the SSA
21508       representation.
21509     * [4]Global null pointer test elimination.
21510     * [5]Global code hoisting/unification.
21511     * More builtins and optimizations for stdio.h, string.h and old BSD
21512       functions, as well as for ISO C99 functions.
21513     * New builtin __builtin_expect for giving hints to the branch
21514       predictor.
21515
21516New Languages and Language specific improvements
21517
21518     * The GNU Compiler for the Java(TM) language (GCJ) is now integrated
21519       and supported, including the run-time library containing most
21520       common non-GUI Java classes, a bytecode interpreter, and the Boehm
21521       conservative garbage collector. Many bugs have been fixed. GCJ can
21522       compile Java source or Java bytecodes to either native code or Java
21523       class files, and supports native methods written in either the
21524       standard JNI or the more efficient and convenient CNI.
21525     * Here is a [6]partial list of C++ improvements, both new features
21526       and those no longer supported.
21527     * New C++ ABI. On the IA-64 platform GCC is capable of
21528       inter-operating with other IA-64 compilers.
21529     * The new ABI also significantly reduces the size of symbol and debug
21530       information.
21531     * New C++ support library and many C++ bug fixes, vastly improving
21532       our conformance to the ISO C++ standard.
21533     * New [7]inliner for C++.
21534     * Rewritten C preprocessor, integrated into the C, C++ and Objective
21535       C compilers, with very many improvements including ISO C99 support
21536       and [8]improvements to dependency generation.
21537     * Support for more [9]ISO C99 features.
21538     * Many improvements to support for checking calls to format functions
21539       such as printf and scanf, including support for ISO C99 format
21540       features, extensions from the Single Unix Specification and GNU
21541       libc 2.2, checking of strfmon formats and features to assist in
21542       auditing for format string security bugs.
21543     * New warnings for C code that may have undefined semantics because
21544       of violations of sequence point rules in the C standard (such as a
21545       = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall.
21546     * Additional warning option -Wfloat-equal.
21547     * Improvements to -Wtraditional.
21548     * Fortran improvements are listed in [10]the Fortran documentation.
21549
21550New Targets and Target Specific Improvements
21551
21552     * New x86 back end, generating much improved code.
21553     * Support for a generic i386-elf target contributed.
21554     * New option to emit x86 assembly code using Intel style syntax
21555       (-mintel-syntax).
21556     * HPUX 11 support contributed.
21557     * Improved PowerPC code generation, including scheduled prologue and
21558       epilogue.
21559     * Port of GCC to Intel's IA-64 processor contributed.
21560     * Port of GCC to Motorola's MCore 210 and 340 contributed.
21561     * New unified back-end for Arm, Thumb and StrongArm contributed.
21562     * Port of GCC to Intel's XScale processor contributed.
21563     * Port of GCC to Atmel's AVR microcontrollers contributed.
21564     * Port of GCC to Mitsubishi's D30V processor contributed.
21565     * Port of GCC to Matsushita's AM33 processor (a member of the MN10300
21566       processor family) contributed.
21567     * Port of GCC to Fujitsu's FR30 processor contributed.
21568     * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed.
21569     * Port of GCC to Sun's picoJava processor core contributed.
21570
21571Documentation improvements
21572
21573     * Substantially rewritten and improved C preprocessor manual.
21574     * Many improvements to other documentation.
21575     * Manpages for gcc, cpp and gcov are now generated automatically from
21576       the master Texinfo manual, eliminating the problem of manpages
21577       being out of date. (The generated manpages are only extracts from
21578       the full manual, which is provided in Texinfo form, from which
21579       info, HTML, other formats and a printed manual can be generated.)
21580     * Generated info files are included in the release tarballs alongside
21581       their Texinfo sources, avoiding problems on some platforms with
21582       building makeinfo as part of the GCC distribution.
21583
21584Other significant improvements
21585
21586     * Garbage collection used internally by the compiler for most memory
21587       allocation instead of obstacks.
21588     * Lengauer and Tarjan algorithm used for computing dominators in the
21589       CFG. This algorithm can be significantly faster and more space
21590       efficient than our older algorithm.
21591     * gccbug script provided to assist in submitting bug reports to our
21592       bug tracking system. (Bug reports previously submitted directly to
21593       our mailing lists, for which you received no bug tracking number,
21594       should be submitted again using gccbug if you can reproduce the
21595       problem with GCC 3.0.)
21596     * The internal libgcc library is [11]built as a shared library on
21597       systems that support it.
21598     * Extensive testsuite included with GCC, with many new tests. In
21599       addition to tests for GCC bugs that have been fixed, many tests
21600       have been added for language features, compiler warnings and
21601       builtin functions.
21602     * Additional language-independent warning options -Wpacked, -Wpadded,
21603       -Wunreachable-code and -Wdisabled-optimization.
21604     * Target-independent options -falign-functions, -falign-loops and
21605       -falign-jumps.
21606
21607   Plus a great many bug fixes and almost all the [12]features found in
21608   GCC 2.95.
21609
21610
21611    For questions related to the use of GCC, please consult these web
21612    pages and the [13]GCC manuals. If that fails, the
21613    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
21614    web pages and the development of GCC are welcome on our developer
21615    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
21616    archives.
21617
21618   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
21619   distribution of this entire article is permitted in any medium,
21620   provided this notice is preserved.
21621
21622   These pages are [18]maintained by the GCC team. Last modified
21623   2022-10-26.
21624
21625References
21626
21627   1. http://www.netbsd.org/
21628   2. http://gcc.gnu.org/news/reorder.html
21629   3. http://gcc.gnu.org/news/ssa.html
21630   4. http://gcc.gnu.org/news/null.html
21631   5. http://gcc.gnu.org/news/unify.html
21632   6. http://gcc.gnu.org/gcc-3.0/c++features.html
21633   7. http://gcc.gnu.org/news/inlining.html
21634   8. http://gcc.gnu.org/news/dependencies.html
21635   9. http://gcc.gnu.org/c99status.html
21636  10. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
21637  11. http://gcc.gnu.org/gcc-3.0/libgcc.html
21638  12. http://gcc.gnu.org/gcc-2.95/features.html
21639  13. https://gcc.gnu.org/onlinedocs/
21640  14. mailto:gcc-help@gcc.gnu.org
21641  15. mailto:gcc@gcc.gnu.org
21642  16. https://gcc.gnu.org/lists.html
21643  17. https://www.fsf.org/
21644  18. https://gcc.gnu.org/about.html
21645======================================================================
21646http://gcc.gnu.org/gcc-3.0/caveats.html
21647
21648                                GCC 3.0 Caveats
21649
21650     * -fstrict-aliasing is now part of -O2 and higher optimization
21651       levels. This allows the compiler to assume the strictest aliasing
21652       rules applicable to the language being compiled. For C and C++,
21653       this activates optimizations based on the type of expressions. This
21654       optimization may thus break old, non-compliant code.
21655     * Enumerations are now properly promoted to int in function
21656       parameters and function returns. Normally this change is not
21657       visible, but when using -fshort-enums this is an ABI change.
21658     * The undocumented extension that allowed C programs to have a label
21659       at the end of a compound statement has been deprecated and may be
21660       removed in a future version. Programs that now generate a warning
21661       about this may be fixed by adding a null statement (a single
21662       semicolon) after the label.
21663     * The poorly documented extension that allowed string constants in C,
21664       C++ and Objective C to contain unescaped newlines has been
21665       deprecated and may be removed in a future version. Programs using
21666       this extension may be fixed in several ways: the bare newline may
21667       be replaced by \n, or preceded by \n\, or string concatenation may
21668       be used with the bare newline preceded by \n" and " placed at the
21669       start of the next line.
21670     * The Chill compiler is not included in GCC 3.0, because of the lack
21671       of a volunteer to convert it to use garbage collection.
21672     * Certain non-standard iostream methods from earlier versions of
21673       libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach,
21674       ostream::form, and istream::gets.
21675     * The new C++ ABI is not yet fully supported by current (as of
21676       2001-07-01) releases and development versions of GDB, or any
21677       earlier versions. There is a problem setting breakpoints by line
21678       number, and other related issues that have been fixed in GCC 3.0
21679       but not yet handled in GDB:
21680       [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
21681
21682
21683    For questions related to the use of GCC, please consult these web
21684    pages and the [2]GCC manuals. If that fails, the
21685    [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
21686    web pages and the development of GCC are welcome on our developer
21687    list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
21688
21689   Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
21690   distribution of this entire article is permitted in any medium,
21691   provided this notice is preserved.
21692
21693   These pages are [7]maintained by the GCC team. Last modified
21694   2022-10-26.
21695
21696References
21697
21698   1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
21699   2. https://gcc.gnu.org/onlinedocs/
21700   3. mailto:gcc-help@gcc.gnu.org
21701   4. mailto:gcc@gcc.gnu.org
21702   5. https://gcc.gnu.org/lists.html
21703   6. https://www.fsf.org/
21704   7. https://gcc.gnu.org/about.html
21705======================================================================
21706http://gcc.gnu.org/gcc-2.95/index.html
21707
21708                                    GCC 2.95
21709
21710   (This release series is no longer supported.)
21711
21712   March 16, 2001: The GNU project and the GCC developers are pleased to
21713   announce the release of GCC version 2.95.3.
21714
21715Release History
21716
21717   GCC 2.95.3
21718          March 16, 2001
21719
21720   GCC 2.95.2
21721          October 27, 1999
21722
21723   GCC 2.95.1
21724          August 19, 1999
21725
21726   GCC 2.95
21727          July 31, 1999. This is the first release of GCC since the April
21728          1999 GCC/EGCS reunification and includes nearly a year's worth
21729          of new development and bugfixes.
21730
21731References and Acknowledgements
21732
21733   GCC used to stand for the GNU C Compiler, but since the compiler
21734   supports several other languages aside from C, it now stands for the
21735   GNU Compiler Collection.
21736
21737   The whole suite has been extensively [1]regression tested and
21738   [2]package tested. It should be reliable and suitable for widespread
21739   use.
21740
21741   The compiler has several new optimizations, new targets, new languages
21742   and other new features. See the [3]new features page for a more
21743   complete list of new features found in the GCC 2.95 releases.
21744
21745   The sources include installation instructions in both HTML and
21746   plaintext forms in the install directory in the distribution. However,
21747   the most up to date installation instructions and [4]build/test status
21748   are on the web pages. We will update those pages as new information
21749   becomes available.
21750
21751   The GCC developers would like to thank the numerous people that have
21752   contributed new features, test results, bugfixes, etc to GCC. This
21753   [5]amazing group of volunteers is what makes GCC successful.
21754
21755   And finally, we can't in good conscience fail to mention some
21756   [6]caveats to using GCC 2.95.
21757
21758   Download GCC 2.95 from one of our many [7]mirror sites.
21759
21760   For additional information about GCC please see the [8]GCC project web
21761   server or contact the [9]GCC development mailing list.
21762
21763
21764    For questions related to the use of GCC, please consult these web
21765    pages and the [10]GCC manuals. If that fails, the
21766    [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these
21767    web pages and the development of GCC are welcome on our developer
21768    list at [12]gcc@gcc.gnu.org. All of [13]our lists have public
21769    archives.
21770
21771   Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and
21772   distribution of this entire article is permitted in any medium,
21773   provided this notice is preserved.
21774
21775   These pages are [15]maintained by the GCC team. Last modified
21776   2022-10-26.
21777
21778References
21779
21780   1. http://gcc.gnu.org/gcc-2.95/regress.html
21781   2. http://gcc.gnu.org/gcc-2.95/othertest.html
21782   3. http://gcc.gnu.org/gcc-2.95/features.html
21783   4. http://gcc.gnu.org/gcc-2.95/buildstat.html
21784   5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
21785   6. http://gcc.gnu.org/gcc-2.95/caveats.html
21786   7. http://gcc.gnu.org/mirrors.html
21787   8. http://gcc.gnu.org/index.html
21788   9. mailto:gcc@gcc.gnu.org
21789  10. https://gcc.gnu.org/onlinedocs/
21790  11. mailto:gcc-help@gcc.gnu.org
21791  12. mailto:gcc@gcc.gnu.org
21792  13. https://gcc.gnu.org/lists.html
21793  14. https://www.fsf.org/
21794  15. https://gcc.gnu.org/about.html
21795======================================================================
21796http://gcc.gnu.org/gcc-2.95/features.html
21797
21798                             GCC 2.95 New Features
21799
21800     * General Optimizer Improvements:
21801          + [1]Localized register spilling to improve speed and code
21802            density especially on small register class machines.
21803          + [2]Global CSE using lazy code motion algorithms.
21804          + [3]Improved global constant/copy propagation.
21805          + [4]Improved control flow graph analysis and manipulation.
21806          + [5]Local dead store elimination.
21807          + [6]Memory Load hoisting/store sinking in loops.
21808          + [7]Type based alias analysis is enabled by default. Note this
21809            feature will expose bugs in the Linux kernel. Please refer to
21810            the FAQ (as shipped with GCC 2.95) for additional information
21811            on this issue.
21812          + Major revamp of GIV detection, combination and simplification
21813            to improve loop performance.
21814          + Major improvements to register allocation and reloading.
21815     * New Languages and Language specific improvements
21816          + [8]Many C++ improvements.
21817          + [9]Many Fortran improvements.
21818          + [10]Java front-end has been integrated. A [11]runtime library
21819            is available separately.
21820          + [12]ISO C99 support
21821          + [13]Chill front-end and runtime has been integrated.
21822          + Boehm garbage collector support in libobjc.
21823          + More support for various pragmas which appear in vendor
21824            include files
21825     * New Targets and Target Specific Improvements
21826          + [14]SPARC backend rewrite.
21827          + -mschedule=8000 will optimize code for PA8000 class
21828            processors; -mpa-risc-2-0 will generate code for PA2.0
21829            processors
21830          + Various micro-optimizations for the ia32 port. K6
21831            optimizations
21832          + Compiler will attempt to align doubles in the stack on the
21833            ia32 port
21834          + Alpha EV6 support
21835          + PowerPC 750
21836          + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403.
21837            -mcpu=e603e was added to do -mcpu=603e and -msoft-float.
21838          + c3x, c4x
21839          + HyperSPARC
21840          + SparcLite86x
21841          + sh4
21842          + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix,
21843            arm-linux)
21844          + vxWorks targets include support for vxWorks threads
21845          + StrongARM 110 and ARM9 support added. ARM Scheduling
21846            parameters rewritten.
21847          + Various changes to the MIPS port to avoid assembler macros,
21848            which in turn improves performance
21849          + Various performance improvements to the i960 port.
21850          + Major rewrite of ns32k port
21851     * Other significant improvements
21852          + [15]Ability to dump cfg information and display it using vcg.
21853          + The new faster scheme for fixing vendor header files is
21854            enabled by default.
21855          + Experimental internationalization support.
21856          + multibyte character support
21857          + Some compile-time speedups for pathological problems
21858          + Better support for complex types
21859     * Plus the usual mountain of bugfixes
21860     * Core compiler is based on the gcc2 development tree from Sept 30,
21861       1998, so we have all of the [16]features found in GCC 2.8.
21862
21863Additional Changes in GCC 2.95.1
21864
21865     * Generic bugfixes and improvements
21866          + Various documentation fixes related to the GCC/EGCS merger.
21867          + Fix memory management bug which could lead to spurious aborts,
21868            core dumps or random parsing errors in the compiler.
21869          + Fix a couple bugs in the dwarf1 and dwarf2 debug record
21870            support.
21871          + Fix infinite loop in the CSE optimizer.
21872          + Avoid undefined behavior in compiler FP emulation code
21873          + Fix install problem when prefix is overridden on the make
21874            install command.
21875          + Fix problem with unwanted installation of assert.h on some
21876            systems.
21877          + Fix problem with finding the wrong assembler in a single tree
21878            build.
21879          + Avoid increasing the known alignment of a register that is
21880            already known to be a pointer.
21881     * Platform specific bugfixes and improvements
21882          + Codegen bugfix for prologue/epilogue for cpu32 target.
21883          + Fix long long code generation bug for the Coldfire target.
21884          + Fix various aborts in the SH compiler.
21885          + Fix bugs in libgcc support library for the SH.
21886          + Fix alpha ev6 code generation bug.
21887          + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on
21888            AIX platforms.
21889          + Fix -fpic code generation bug for rs6000/ppc svr4 targets.
21890          + Fix varargs/stdarg code generation bug for rs6000/ppc svr4
21891            targets.
21892          + Fix weak symbol handling for rs6000/ppc svr4 targets.
21893          + Fix various problems with 64bit code generation for the
21894            rs6000/ppc port.
21895          + Fix codegen bug which caused tetex to be mis-compiled on the
21896            x86.
21897          + Fix compiler abort in new cfg code exposed by x86 port.
21898          + Fix out of range array reference in code convert flat
21899            registers to the x87 stacked FP register file.
21900          + Fix minor vxworks configuration bug.
21901          + Fix return type of bsearch for SunOS 4.x.
21902     * Language & Runtime specific fixes.
21903          + The G++ signature extension has been deprecated. It will be
21904            removed in the next major release of G++. Use of signatures
21905            will result in a warning from the compiler.
21906          + Several bugs relating to templates and namespaces were fixed.
21907          + A bug that caused crashes when combining templates with -g on
21908            DWARF1 platforms was fixed.
21909          + Pointers-to-members, virtual functions, and multiple
21910            inheritance should now work together correctly.
21911          + Some code-generation bugs relating to function try blocks were
21912            fixed.
21913          + G++ is a little bit more lenient with certain archaic
21914            constructs than in GCC 2.95.
21915          + Fix to prevent shared library version #s from bring truncated
21916            to 1 digit
21917          + Fix missing std:: in the libstdc++ library.
21918          + Fix stream locking problems in libio.
21919          + Fix problem in java compiler driver.
21920
21921Additional Changes in GCC 2.95.2
21922
21923   The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While
21924   the optimizations performed by -fstrict-aliasing are valid according to
21925   the C and C++ standards, the optimization have caused some problems,
21926   particularly with old non-conforming code.
21927
21928   The GCC developers are experimenting with ways to warn users about code
21929   which violates the C/C++ standards, but those warnings are not ready
21930   for widespread use at this time. Rather than wait for those warnings
21931   the GCC developers have chosen to disable -fstrict-aliasing by default
21932   for the GCC 2.95.2 release.
21933
21934   We strongly encourage developers to find and fix code which violates
21935   the C/C++ standards as -fstrict-aliasing may be enabled by default in
21936   future releases. Use the option -fstrict-aliasing to re-enable these
21937   optimizations.
21938     * Generic bugfixes and improvements
21939          + Fix incorrectly optimized memory reference in global common
21940            subexpression elimination (GCSE) optimization pass.
21941          + Fix code generation bug in regmove.c in which it could
21942            incorrectly change a "const" value.
21943          + Fix bug in optimization of conditionals involving volatile
21944            memory references.
21945          + Avoid over-allocation of stack space for some procedures.
21946          + Fixed bug in the compiler which caused incorrect optimization
21947            of an obscure series of bit manipulations, shifts and
21948            arithmetic.
21949          + Fixed register allocator bug which caused teTeX to be
21950            mis-compiled on SPARC targets.
21951          + Avoid incorrect optimization of degenerate case statements for
21952            certain targets such as the ARM.
21953          + Fix out of range memory reference in the jump optimizer.
21954          + Avoid dereferencing null pointer in fix-header.
21955          + Fix test for GCC specific features so that it is possible to
21956            bootstrap with gcc-2.6.2 and older versions of GCC.
21957          + Fix typo in scheduler which could potentially cause out of
21958            range memory accesses.
21959          + Avoid incorrect loop reversal which caused incorrect code for
21960            certain loops on PowerPC targets.
21961          + Avoid incorrect optimization of switch statements on certain
21962            targets (for example the ARM).
21963     * Platform specific bugfixes and improvements
21964          + Work around bug in Sun V5.0 compilers which caused bootstrap
21965            comparison failures on SPARC targets.
21966          + Fix SPARC backend bug which caused aborts in final.c.
21967          + Fix sparc-hal-solaris2* configuration fragments.
21968          + Fix bug in sparc block profiling.
21969          + Fix obscure code generation bug for the PARISC targets.
21970          + Define __STDC_EXT__ for HPUX configurations.
21971          + Various POWERPC64 code generation bugfixes.
21972          + Fix abort for PPC targets using ELF (ex GNU/Linux).
21973          + Fix collect2 problems for AIX targets.
21974          + Correct handling of .file directive for PPC targets.
21975          + Fix bug in fix_trunc x86 patterns.
21976          + Fix x86 port to correctly pop the FP stack for functions that
21977            return structures in memory.
21978          + Fix minor bug in strlen x86 pattern.
21979          + Use stabs debugging instead of dwarf1 for x86-solaris targets.
21980          + Fix template repository code to handle leading underscore in
21981            mangled names.
21982          + Fix weak/weak alias support for OpenBSD.
21983          + GNU/Linux for the ARM has C++ compatible include files.
21984     * Language & Runtime specific fixes.
21985          + Fix handling of constructor attribute in the C front-end which
21986            caused problems building the Chill runtime library on some
21987            targets.
21988          + Fix minor problem merging type qualifiers in the C front-end.
21989          + Fix aliasing bug for pointers and references (C/C++).
21990          + Fix incorrect "non-constant initializer bug" when -traditional
21991            or -fwritable-strings is enabled.
21992          + Fix build error for Chill front-end on SunOS.
21993          + Do not complain about duplicate instantiations when using
21994            -frepo (C++).
21995          + Fix array bounds handling in C++ front-end which caused
21996            problems with dwarf debugging information in some
21997            circumstances.
21998          + Fix minor namespace problem.
21999          + Fix problem linking java programs.
22000
22001Additional Changes in GCC 2.95.3
22002
22003     * Generic bugfixes and improvements
22004          + Fix numerous problems that caused incorrect optimization in
22005            the register reloading code.
22006          + Fix numerous problems that caused incorrect optimization in
22007            the loop optimizer.
22008          + Fix aborts in the functions build_insn_chain and scan_loops
22009            under some circumstances.
22010          + Fix an alias analysis bug.
22011          + Fix an infinite compilation bug in the combiner.
22012          + A few problems with complex number support have been fixed.
22013          + It is no longer possible for gcc to act as a fork bomb when
22014            installed incorrectly.
22015          + The -fpack-struct option should be recognized now.
22016          + Fixed a bug that caused incorrect code to be generated due to
22017            a lost stack adjustment.
22018     * Platform specific bugfixes and improvements
22019          + Support building ARM toolchains hosted on Windows.
22020          + Fix attribute calculations in ARM toolchains.
22021          + arm-linux support has been improved.
22022          + Fix a PIC failure on sparc targets.
22023          + On ix86 targets, the regparm attribute should now work
22024            reliably.
22025          + Several updates for the h8300 port.
22026          + Fix problem building libio with glibc 2.2.
22027
22028
22029    For questions related to the use of GCC, please consult these web
22030    pages and the [17]GCC manuals. If that fails, the
22031    [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
22032    web pages and the development of GCC are welcome on our developer
22033    list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
22034    archives.
22035
22036   Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
22037   distribution of this entire article is permitted in any medium,
22038   provided this notice is preserved.
22039
22040   These pages are [22]maintained by the GCC team. Last modified
22041   2022-10-26.
22042
22043References
22044
22045   1. http://gcc.gnu.org/news/spill.html
22046   2. http://gcc.gnu.org/news/lcm.html
22047   3. http://gcc.gnu.org/news/cprop.html
22048   4. http://gcc.gnu.org/news/cfg.html
22049   5. http://gcc.gnu.org/news/dse.html
22050   6. http://gcc.gnu.org/news/hoist.html
22051   7. http://gcc.gnu.org/news/alias.html
22052   8. http://gcc.gnu.org/gcc-2.95/c++features.html
22053   9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
22054  10. http://gcc.gnu.org/news/gcj-announce.txt
22055  11. http://gcc.gnu.org/news/javaannounce.html
22056  12. http://gcc.gnu.org/c99status.html
22057  13. http://gcc.gnu.org/news/chill.html
22058  14. http://gcc.gnu.org/news/sparc.html
22059  15. http://gcc.gnu.org/news/egcs-vcg.html
22060  16. http://gcc.gnu.org/egcs-1.0/features-2.8.html
22061  17. https://gcc.gnu.org/onlinedocs/
22062  18. mailto:gcc-help@gcc.gnu.org
22063  19. mailto:gcc@gcc.gnu.org
22064  20. https://gcc.gnu.org/lists.html
22065  21. https://www.fsf.org/
22066  22. https://gcc.gnu.org/about.html
22067======================================================================
22068http://gcc.gnu.org/gcc-2.95/caveats.html
22069
22070                                GCC 2.95 Caveats
22071
22072     * GCC 2.95 will issue an error for invalid asm statements that had
22073       been silently accepted by earlier versions of the compiler. This is
22074       particularly noticeable when compiling older versions of the Linux
22075       kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95)
22076       for more information on this issue.
22077     * GCC 2.95 implements type based alias analysis to disambiguate
22078       memory references. Some programs, particularly the Linux kernel
22079       violate ANSI/ISO aliasing rules and therefore may not operate
22080       correctly when compiled with GCC 2.95. Please refer to the FAQ (as
22081       shipped with GCC 2.95) for more information on this issue.
22082     * GCC 2.95 has a known bug in its handling of complex variables for
22083       64bit targets. Instead of silently generating incorrect code, GCC
22084       2.95 will issue a fatal error for situations it can not handle.
22085       This primarily affects the Fortran community as Fortran makes more
22086       use of complex variables than C or C++.
22087     * GCC 2.95 has an integrated libstdc++, but does not have an
22088       integrated libg++. Furthermore old libg++ releases will not work
22089       with GCC 2.95. You can retrieve a recent copy of libg++ from the
22090       [1]GCC ftp server.
22091       Note most C++ programs only need libstdc++.
22092     * Exception handling may not work with shared libraries, particularly
22093       on alphas, hppas, rs6000/powerpc and mips based platforms.
22094       Exception handling is known to work on x86 GNU/Linux platforms with
22095       shared libraries.
22096     * In general, GCC 2.95 is more rigorous about rejecting invalid C++
22097       code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0,
22098       or EGCS 1.1. As a result it may be necessary to fix C++ code before
22099       it will compile with GCC 2.95.
22100     * G++ is also converting toward the ISO C++ standard; as a result
22101       code which was previously valid (and thus accepted by other
22102       compilers and older versions of g++) may no longer be accepted. The
22103       flag -fpermissive may allow some non-conforming code to compile
22104       with GCC 2.95.
22105     * GCC 2.95 compiled C++ code is not binary compatible with EGCS
22106       1.1.x, EGCS 1.0.x or GCC 2.8.x.
22107     * GCC 2.95 does not have changes from the GCC 2.8 tree that were made
22108       between Sept 30, 1998 and April 30, 1999 (the official end of the
22109       GCC 2.8 project). Future GCC releases will include all the changes
22110       from the defunct GCC 2.8 sources.
22111
22112
22113    For questions related to the use of GCC, please consult these web
22114    pages and the [2]GCC manuals. If that fails, the
22115    [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
22116    web pages and the development of GCC are welcome on our developer
22117    list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
22118
22119   Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
22120   distribution of this entire article is permitted in any medium,
22121   provided this notice is preserved.
22122
22123   These pages are [7]maintained by the GCC team. Last modified
22124   2022-10-26.
22125
22126References
22127
22128   1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz
22129   2. https://gcc.gnu.org/onlinedocs/
22130   3. mailto:gcc-help@gcc.gnu.org
22131   4. mailto:gcc@gcc.gnu.org
22132   5. https://gcc.gnu.org/lists.html
22133   6. https://www.fsf.org/
22134   7. https://gcc.gnu.org/about.html
22135======================================================================
22136http://gcc.gnu.org/egcs-1.1/index.html
22137
22138                                    EGCS 1.1
22139
22140   September 3, 1998: We are pleased to announce the release of EGCS 1.1.
22141   December 1, 1998: We are pleased to announce the release of EGCS 1.1.1.
22142   March 15, 1999: We are pleased to announce the release of EGCS 1.1.2.
22143
22144   EGCS is a free software project to further the development of the GNU
22145   compilers using an open development environment.
22146
22147   EGCS 1.1 is a major new release of the EGCS compiler system. It has
22148   been [1]extensively tested and is believed to be stable and suitable
22149   for widespread use.
22150
22151   EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8
22152   development sources; it contains all of the new features found in GCC
22153   2.8.1 as well as all new development from GCC up to June 6, 1998.
22154
22155   EGCS 1.1 also contains many improvements and features not found in GCC
22156   or in older versions of EGCS:
22157     * Global common subexpression elimination and global constant/copy
22158       propagation (aka [2]gcse)
22159     * Ongoing improvements to the [3]alias analysis support to allow for
22160       better optimizations throughout the compiler.
22161     * Vastly improved [4]C++ compiler and integrated C++ runtime
22162       libraries.
22163     * Fixes for the /tmp symlink race security problems.
22164     * New targets including mips16, arm-thumb and 64 bit PowerPC.
22165     * Improvements to GNU Fortran (g77) compiler and runtime library made
22166       since g77 version 0.5.23.
22167
22168   See the [5]new features page for a more complete list of new features
22169   found in EGCS 1.1 releases.
22170
22171   EGCS 1.1.1 is a minor update to fix several serious problems in EGCS
22172   1.1:
22173     * General improvements and fixes
22174          + Avoid some stack overflows when compiling large functions.
22175          + Avoid incorrect loop invariant code motions.
22176          + Fix some core dumps on Linux kernel code.
22177          + Bring back the imake -Di386 and friends fix from EGCS 1.0.2.
22178          + Fix code generation problem in gcse.
22179          + Various documentation related fixes.
22180     * g++/libstdc++ improvements and fixes
22181          + MT safe EH fix for setjmp/longjmp based exception handling.
22182          + Fix a few bad interactions between optimization and exception
22183            handling.
22184          + Fixes for demangling of template names starting with "__".
22185          + Fix a bug that would fail to run destructors in some cases
22186            with -O2.
22187          + Fix 'new' of classes with virtual bases.
22188          + Fix crash building Qt on the Alpha.
22189          + Fix failure compiling WIFEXITED macro on GNU/Linux.
22190          + Fix some -frepo failures.
22191     * g77 and libf2c improvements and fixes
22192          + Various documentation fixes.
22193          + Avoid compiler crash on RAND intrinsic.
22194          + Fix minor bugs in makefiles exposed by BSD make programs.
22195          + Define _XOPEN_SOURCE for libI77 build to avoid potential
22196            problems on some 64-bit systems.
22197          + Fix problem with implicit endfile on rewind.
22198          + Fix spurious recursive I/O errors.
22199     * platform specific improvements and fixes
22200          + Match all versions of UnixWare7.
22201          + Do not assume x86 SVR4 or UnixWare targets can handle stabs.
22202          + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion
22203            from unsigned ints to double precision floats.
22204          + Fix ARM ABI issue with NetBSD.
22205          + Fix a few arm code generation bugs.
22206          + Fixincludes will fix additional broken SCO OpenServer header
22207            files.
22208          + Fix a m68k backend bug which caused invalid offsets in reg+d
22209            addresses.
22210          + Fix problems with 64bit AIX 4.3 support.
22211          + Fix handling of long longs for varargs/stdarg functions on the
22212            ppc.
22213          + Minor fixes to CPP predefines for Windows.
22214          + Fix code generation problems with gpr<->fpr copies for 64bit
22215            ppc.
22216          + Fix a few coldfire code generation bugs.
22217          + Fix some more header file problems on SunOS 4.x.
22218          + Fix assert.h handling for RTEMS.
22219          + Fix Windows handling of TREE_SYMBOL_REFERENCED.
22220          + Fix x86 compiler abort in reg-stack pass.
22221          + Fix cygwin/windows problem with section attributes.
22222          + Fix Alpha code generation problem exposed by SMP Linux
22223            kernels.
22224          + Fix typo in m68k 32->64bit integer conversion.
22225          + Make sure target libraries build with -fPIC for PPC & Alpha
22226            targets.
22227
22228   EGCS 1.1.2 is a minor update to fix several serious problems in EGCS
22229   1.1.1:
22230     * General improvements and fixes
22231          + Fix bug in loop optimizer which caused the SPARC (and
22232            potentially other) ports to segfault.
22233          + Fix infinite recursion in alias analysis and combiner code.
22234          + Fix bug in regclass preferencing.
22235          + Fix incorrect loop reversal which caused incorrect code to be
22236            generated for several targets.
22237          + Fix return value for builtin memcpy.
22238          + Reduce compile time for certain loops which exposed quadratic
22239            behavior in the loop optimizer.
22240          + Fix bug which caused volatile memory to be written multiple
22241            times when only one write was needed/desired.
22242          + Fix compiler abort in caller-save.c
22243          + Fix combiner bug which caused incorrect code generation for
22244            certain division by constant operations.
22245          + Fix incorrect code generation due to a bug in range check
22246            optimizations.
22247          + Fix incorrect code generation due to mis-handling of clobbered
22248            values in CSE.
22249          + Fix compiler abort/segfault due to incorrect register
22250            splitting when unrolling loops.
22251          + Fix code generation involving autoincremented addresses with
22252            ternary operators.
22253          + Work around bug in the scheduler which caused qt to be
22254            mis-compiled on some platforms.
22255          + Fix code generation problems with -fshort-enums.
22256          + Tighten security for temporary files.
22257          + Improve compile time for codes which make heavy use of
22258            overloaded functions.
22259          + Fix multiply defined constructor/destructor symbol problems.
22260          + Avoid setting bogus RPATH environment variable during
22261            bootstrap.
22262          + Avoid GNU-make dependencies in the texinfo subdir.
22263          + Install CPP wrapper script in $(prefix)/bin if --enable-cpp.
22264            --enable-cpp=<dirname> can be used to specify an additional
22265            install directory for the cpp wrapper script.
22266          + Fix CSE bug which caused incorrect label-label refs to appear
22267            on some platforms.
22268          + Avoid linking in EH routines from libgcc if they are not
22269            needed.
22270          + Avoid obscure bug in aliasing code.
22271          + Fix bug in weak symbol handling.
22272     * Platform-specific improvements and fixes
22273          + Fix detection of PPro/PII on Unixware 7.
22274          + Fix compiler segfault when building spec99 and other programs
22275            for SPARC targets.
22276          + Fix code-generation bugs for integer and floating point
22277            conditional move instructions on the PPro/PII.
22278          + Use fixincludes to fix byteorder problems on i?86-*-sysv.
22279          + Fix build failure for the arc port.
22280          + Fix floating point format configuration for i?86-gnu port.
22281          + Fix problems with hppa1.0-hp-hpux10.20 configuration when
22282            threads are enabled.
22283          + Fix coldfire code generation bugs.
22284          + Fix "unrecognized insn" problems for Alpha and PPC ports.
22285          + Fix h8/300 code generation problem with floating point values
22286            in memory.
22287          + Fix unrecognized insn problems for the m68k port.
22288          + Fix namespace-pollution problem for the x86 port.
22289          + Fix problems with old assembler on x86 NeXT systems.
22290          + Fix PIC code-generation problems for the SPARC port.
22291          + Fix minor bug with LONG_CALLS in PowerPC SVR4 support.
22292          + Fix minor ISO namespace violation in Alpha varargs/stdarg
22293            support.
22294          + Fix incorrect "braf" instruction usage for the SH port.
22295          + Fix minor bug in va-sh which prevented its use with -ansi.
22296          + Fix problems recognizing and supporting FreeBSD.
22297          + Handle OpenBSD systems correctly.
22298          + Minor fixincludes fix for Digital UNIX 4.0B.
22299          + Fix problems with ctors/dtors in SCO shared libraries.
22300          + Abort instead of generating incorrect code for PPro/PII
22301            floating point conditional moves.
22302          + Avoid multiply defined symbols on GNU/Linux systems using
22303            libc-5.4.xx.
22304          + Fix abort in alpha compiler.
22305     * Fortran-specific fixes
22306          + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year
22307            is in the documented, non-Y2K-compliant range of 0-99, instead
22308            of being returned as 100 in the year 2000.
22309          + Fix the `Date_and_Time' intrinsic (in libg2c) to return the
22310            milliseconds value properly in Values(8).
22311          + Fix the `LStat' intrinsic (in libg2c) to return device-ID
22312            information properly in SArray(7).
22313
22314   Each release includes installation instructions in both HTML and
22315   plaintext forms (see the INSTALL directory in the toplevel directory of
22316   the distribution). However, we also keep the most up to date
22317   installation instructions and [6]build/test status on our web page. We
22318   will update those pages as new information becomes available.
22319
22320   The EGCS project would like to thank the numerous people that have
22321   contributed new features, test results, bugfixes, etc. This [7]amazing
22322   group of volunteers is what makes EGCS successful.
22323
22324   And finally, we can't in good conscience fail to mention some
22325   [8]caveats to using EGCS 1.1.
22326
22327   Download EGCS from egcs.cygnus.com (USA California).
22328
22329   The EGCS 1.1 release is also available on many mirror sites.
22330   [9]Goto mirror list to find a closer site.
22331
22332
22333    For questions related to the use of GCC, please consult these web
22334    pages and the [10]GCC manuals. If that fails, the
22335    [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these
22336    web pages and the development of GCC are welcome on our developer
22337    list at [12]gcc@gcc.gnu.org. All of [13]our lists have public
22338    archives.
22339
22340   Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and
22341   distribution of this entire article is permitted in any medium,
22342   provided this notice is preserved.
22343
22344   These pages are [15]maintained by the GCC team. Last modified
22345   2023-01-21.
22346
22347References
22348
22349   1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html
22350   2. http://gcc.gnu.org/news/gcse.html
22351   3. http://gcc.gnu.org/news/alias.html
22352   4. http://gcc.gnu.org/egcs-1.1/c++features.html
22353   5. http://gcc.gnu.org/egcs-1.1/features.html
22354   6. http://gcc.gnu.org/egcs-1.1/buildstat.html
22355   7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
22356   8. http://gcc.gnu.org/egcs-1.1/caveats.html
22357   9. http://gcc.gnu.org/mirrors.html
22358  10. https://gcc.gnu.org/onlinedocs/
22359  11. mailto:gcc-help@gcc.gnu.org
22360  12. mailto:gcc@gcc.gnu.org
22361  13. https://gcc.gnu.org/lists.html
22362  14. https://www.fsf.org/
22363  15. https://gcc.gnu.org/about.html
22364======================================================================
22365http://gcc.gnu.org/egcs-1.1/features.html
22366
22367                             EGCS 1.1 new features
22368
22369     * Integrated GNU Fortran (g77) compiler and runtime library with
22370       improvements, based on g77 version 0.5.23.
22371     * Vast improvements in the C++ compiler; so many they have [1]page of
22372       their own!
22373     * Compiler implements [2]global common subexpression elimination and
22374       global copy/constant propagation.
22375     * More major improvements in the [3]alias analysis code.
22376     * More major improvements in the exception handling code to improve
22377       performance, lower static overhead and provide the infrastructure
22378       for future improvements.
22379     * The infamous /tmp symlink race security problems have been fixed.
22380     * The regmove optimization pass has been nearly completely rewritten
22381       to improve performance of generated code.
22382     * The compiler now recomputes register usage information before local
22383       register allocation. By providing more accurate information to the
22384       priority based allocator, we get better register allocation.
22385     * The register reloading phase of the compiler optimizes spill code
22386       much better than in previous releases.
22387     * Some bad interactions between the register allocator and
22388       instruction scheduler have been fixed, resulting in much better
22389       code for certain programs. Additionally, we have tuned the
22390       scheduler in various ways to improve performance of generated code
22391       for some architectures.
22392     * The compiler's branch shortening algorithms have been significantly
22393       improved to work better on targets which align jump targets.
22394     * The compiler now supports -Os to prefer optimizing for code space
22395       over optimizing for code speed.
22396     * The compiler will now totally eliminate library calls which compute
22397       constant values. This primarily helps targets with no integer
22398       div/mul support and targets without floating point support.
22399     * The compiler now supports an extensive "--help" option.
22400     * cpplib has been greatly improved and may be suitable for limited
22401       use.
22402     * Memory footprint for the compiler has been significantly reduced
22403       for some pathological cases.
22404     * The time to build EGCS has been improved for certain targets
22405       (particularly the alpha and mips platforms).
22406     * Many infrastructure improvements throughout the compiler, plus the
22407       usual mountain of bugfixes and minor improvements.
22408     * Target dependent improvements:
22409          + SPARC port now includes V8 plus and V9 support as well as
22410            performance tuning for Ultra class machines. The SPARC port
22411            now uses the Haifa scheduler.
22412          + Alpha port has been tuned for the EV6 processor and has an
22413            optimized expansion of memcpy/bzero. The Alpha port now uses
22414            the Haifa scheduler.
22415          + RS6000/PowerPC: support for the Power64 architecture and AIX
22416            4.3. The RS6000/PowerPC port now uses the Haifa scheduler.
22417          + x86: Alignment of static store data and jump targets is per
22418            Intel recommendations now. Various improvements throughout the
22419            x86 port to improve performance on Pentium processors
22420            (including improved epilogue sequences for Pentium chips and
22421            backend improvements which should help register allocation on
22422            all x86 variants. Conditional move support has been fixed and
22423            enabled for PPro processors. The x86 port also better supports
22424            64bit operations now. Unixware 7, a System V Release 5 target,
22425            is now supported and SCO OpenServer targets can support GAS.
22426          + MIPS has improved multiply/multiply-add support and now
22427            includes mips16 ISA support.
22428          + M68k has many micro-optimizations and Coldfire fixes.
22429     * Core compiler is based on the GCC development tree from June 9,
22430       1998, so we have all of the [4]features found in GCC 2.8.
22431
22432
22433    For questions related to the use of GCC, please consult these web
22434    pages and the [5]GCC manuals. If that fails, the
22435    [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
22436    web pages and the development of GCC are welcome on our developer
22437    list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
22438
22439   Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
22440   distribution of this entire article is permitted in any medium,
22441   provided this notice is preserved.
22442
22443   These pages are [10]maintained by the GCC team. Last modified
22444   2022-10-26.
22445
22446References
22447
22448   1. http://gcc.gnu.org/egcs-1.1/c++features.html
22449   2. http://gcc.gnu.org/news/gcse.html
22450   3. http://gcc.gnu.org/news/alias.html
22451   4. http://gcc.gnu.org/egcs-1.0/features-2.8.html
22452   5. https://gcc.gnu.org/onlinedocs/
22453   6. mailto:gcc-help@gcc.gnu.org
22454   7. mailto:gcc@gcc.gnu.org
22455   8. https://gcc.gnu.org/lists.html
22456   9. https://www.fsf.org/
22457  10. https://gcc.gnu.org/about.html
22458======================================================================
22459http://gcc.gnu.org/egcs-1.1/caveats.html
22460
22461                                EGCS 1.1 Caveats
22462
22463     * EGCS has an integrated libstdc++, but does not have an integrated
22464       libg++. Furthermore old libg++ releases will not work with EGCS; HJ
22465       Lu has made a libg++-2.8.1.2 snapshot available which may work with
22466       EGCS.
22467       Note most C++ programs only need libstdc++.
22468     * Exception handling may not work with shared libraries, particularly
22469       on alphas, hppas, rs6000/powerpc and mips based platforms.
22470       Exception handling is known to work on x86-linux platforms with
22471       shared libraries.
22472     * Some versions of the Linux kernel have bugs which prevent them from
22473       being compiled or from running when compiled by EGCS. See the FAQ
22474       (as shipped with EGCS 1.1) for additional information.
22475     * In general, EGCS is more rigorous about rejecting invalid C++ code
22476       or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As
22477       a result it may be necessary to fix C++ code before it will compile
22478       with EGCS.
22479     * G++ is also converting toward the ISO C++ standard; as a result
22480       code which was previously valid (and thus accepted by other
22481       compilers and older versions of g++) may no longer be accepted.
22482     * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x
22483       or GCC 2.8.x due to changes necessary to support thread safe
22484       exception handling.
22485
22486
22487    For questions related to the use of GCC, please consult these web
22488    pages and the [1]GCC manuals. If that fails, the
22489    [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
22490    web pages and the development of GCC are welcome on our developer
22491    list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
22492
22493   Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
22494   distribution of this entire article is permitted in any medium,
22495   provided this notice is preserved.
22496
22497   These pages are [6]maintained by the GCC team. Last modified
22498   2022-10-26.
22499
22500References
22501
22502   1. https://gcc.gnu.org/onlinedocs/
22503   2. mailto:gcc-help@gcc.gnu.org
22504   3. mailto:gcc@gcc.gnu.org
22505   4. https://gcc.gnu.org/lists.html
22506   5. https://www.fsf.org/
22507   6. https://gcc.gnu.org/about.html
22508======================================================================
22509http://gcc.gnu.org/egcs-1.0/index.html
22510
22511                                    EGCS 1.0
22512
22513   December 3, 1997: We are pleased to announce the release of EGCS 1.0.
22514   January 6, 1998: We are pleased to announce the release of EGCS 1.0.1.
22515   March 16, 1998: We are pleased to announce the release of EGCS 1.0.2.
22516   May 15, 1998 We are pleased to announce the release of EGCS 1.0.3.
22517
22518   EGCS is a collaborative effort involving several groups of hackers
22519   using an open development model to accelerate development and testing
22520   of GNU compilers and runtime libraries.
22521
22522   An important goal of EGCS is to allow wide scale testing of
22523   experimental features and optimizations; therefore, EGCS contains some
22524   features and optimizations which are still under development. However,
22525   EGCS has been carefully tested and should be comparable in quality to
22526   most GCC releases.
22527
22528   EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8
22529   development sources; it contains nearly all of the new features found
22530   in GCC 2.8.
22531
22532   EGCS 1.0 also contains many improvements and features not found in GCC
22533   2.7 and even the GCC 2.8 series (which was released after the original
22534   EGCS 1.0 release).
22535     * Integrated C++ runtime libraries, including support for most major
22536       GNU/Linux systems!
22537     * The integrated libstdc++ library includes a verbatim copy of SGI's
22538       STL release.
22539     * Integrated GNU Fortran compiler.
22540     * New instruction scheduler.
22541     * New alias analysis code.
22542
22543   See the [1]new features page for a more complete list of new features.
22544
22545   EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few
22546   critical bugs and add support for Red Hat 5.0 Linux. Changes since the
22547   EGCS 1.0 release:
22548     * Add support for Red Hat 5.0 Linux and better support for Linux
22549       systems using glibc2.
22550       Many programs failed to link when compiled with EGCS 1.0 on Red Hat
22551       5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should
22552       fix these problems.
22553     * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception
22554       handling interfaces.
22555       To avoid future compatibility problems, we strongly urge anyone who
22556       is planning on distributing shared libraries that contain C++ code
22557       to upgrade to EGCS 1.0.1 first.
22558       Soon after EGCS 1.0 was released, the GCC developers made some
22559       incompatible changes in libgcc's exception handling interfaces.
22560       These changes were needed to solve problems on some platforms. This
22561       means that GCC 2.8.0, when released, will not be seamlessly
22562       compatible with shared libraries built by EGCS 1.0. The reason is
22563       that the libgcc.a in GCC 2.8.0 will not contain a function needed
22564       by the old interface.
22565       The result of this is that there may be compatibility problems with
22566       shared libraries built by EGCS 1.0 when used with GCC 2.8.0.
22567       With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface,
22568       and libgcc.a has the support routines for both the old and the new
22569       interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed,
22570       and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed).
22571       The maintainers of GCC 2.x have decided against including seamless
22572       support for the old interface in 2.8.0, since it was never
22573       "official", so to avoid future compatibility problems we recommend
22574       against distributing any shared libraries built by EGCS 1.0 that
22575       contain C++ code (upgrade to 1.0.1 and use that).
22576     * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends.
22577       The x86 changes fix code generation errors exposed when building
22578       glibc2 and the usual GNU/Linux dynamic linker (ld.so).
22579       The hppa change fixes a compiler abort when configured for use with
22580       RTEMS.
22581       The MIPS changes fix problems with the definition of LONG_MAX on
22582       newer systems, allow for command line selection of the target ABI,
22583       and fix one code generation problem.
22584       The rs6000/ppc change fixes some problems with passing structures
22585       to varargs/stdarg functions.
22586     * A few machine independent bugfixes, mostly to fix code generation
22587       errors when building Linux kernels or glibc.
22588     * Fix a few critical exception handling and template bugs in the C++
22589       compiler.
22590     * Fix Fortran namelist bug on alphas.
22591     * Fix build problems on x86-solaris systems.
22592
22593   EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several
22594   serious problems in EGCS 1.0.1.
22595     * General improvements and fixes
22596          + Memory consumption significantly reduced, especially for
22597            templates and inline functions.
22598          + Fix various problems with glibc2.1.
22599          + Fix loop optimization bug exposed by rs6000/ppc port.
22600          + Fix to avoid potential code generation problems in jump.c.
22601          + Fix some undefined symbol problems in dwarf1 debug support.
22602     * g++/libstdc++ improvements and fixes
22603          + libstdc++ in the EGCS release has been updated and should be
22604            link compatible with libstdc++-2.8.
22605          + Various fixes in libio/libstdc++ to work better on GNU/Linux
22606            systems.
22607          + Fix problems with duplicate symbols on systems that do not
22608            support weak symbols.
22609          + Memory corruption bug and undefined symbols in bastring have
22610            been fixed.
22611          + Various exception handling fixes.
22612          + Fix compiler abort for very long thunk names.
22613     * g77 improvements and fixes
22614          + Fix compiler crash for omitted bound in Fortran CASE
22615            statement.
22616          + Add missing entries to g77 lang-options.
22617          + Fix problem with -fpedantic in the g77 compiler.
22618          + Fix "backspace" problem with g77 on alphas.
22619          + Fix x86 backend problem with Fortran literals and -fpic.
22620          + Fix some of the problems with negative subscripts for g77 on
22621            alphas.
22622          + Fixes for Fortran builds on cygwin32/mingw32.
22623     * platform specific improvements and fixes
22624          + Fix long double problems on x86 (exposed by glibc).
22625          + x86 ports define i386 again to keep imake happy.
22626          + Fix exception handling support on NetBSD ports.
22627          + Several changes to collect2 to fix many problems with AIX.
22628          + Define __ELF__ for GNU/Linux on rs6000.
22629          + Fix -mcall-linux problem on GNU/Linux on rs6000.
22630          + Fix stdarg/vararg problem for GNU/Linux on rs6000.
22631          + Allow autoconf to select a proper install problem on AIX 3.1.
22632          + m68k port support includes -mcpu32 option as well as cpu32
22633            multilibs.
22634          + Fix stdarg bug for irix6.
22635          + Allow EGCS to build on irix5 without the gnu assembler.
22636          + Fix problem with static linking on sco5.
22637          + Fix bootstrap on sco5 with native compiler.
22638          + Fix for abort building newlib on H8 target.
22639          + Fix fixincludes handling of math.h on SunOS.
22640          + Minor fix for Motorola 3300 m68k systems.
22641
22642   EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few
22643   problems reported by Red Hat for builds of Red Hat 5.1.
22644     * Generic bugfixes:
22645          + Fix a typo in the libio library which resulted in incorrect
22646            behavior of istream::get.
22647          + Fix the Fortran negative array index problem.
22648          + Fix a major problem with the ObjC runtime thread support
22649            exposed by glibc2.
22650          + Reduce memory consumption of the Haifa scheduler.
22651     * Target specific bugfixes:
22652          + Fix one x86 floating point code generation bug exposed by
22653            glibc2 builds.
22654          + Fix one x86 internal compiler error exposed by glibc2 builds.
22655          + Fix profiling bugs on the Alpha.
22656          + Fix ImageMagick & emacs 20.2 build problems on the Alpha.
22657          + Fix rs6000/ppc bug when converting values from integer types
22658            to floating point types.
22659
22660   The EGCS 1.0 releases include installation instructions in both HTML
22661   and plaintext forms (see the INSTALL directory in the toplevel
22662   directory of the distribution). However, we also keep the most up to
22663   date installation instructions and [2]build/test status on our web
22664   page. We will update those pages as new information becomes available.
22665
22666   And, we can't in good conscience fail to mention some [3]caveats to
22667   using EGCS.
22668
22669   Update: Big thanks to Stanford for providing a high speed link for
22670   downloading EGCS (go.cygnus.com)!
22671
22672   Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com
22673   (USA California -- High speed link provided by Stanford).
22674
22675   The EGCS 1.0 release is also available many mirror sites.
22676   [4]Goto mirror list to find a closer site
22677
22678   We'd like to thank the numerous people that have contributed new
22679   features, test results, bugfixes, etc. Unfortunately, they're far too
22680   numerous to mention by name.
22681
22682
22683    For questions related to the use of GCC, please consult these web
22684    pages and the [5]GCC manuals. If that fails, the
22685    [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
22686    web pages and the development of GCC are welcome on our developer
22687    list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
22688
22689   Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
22690   distribution of this entire article is permitted in any medium,
22691   provided this notice is preserved.
22692
22693   These pages are [10]maintained by the GCC team. Last modified
22694   2023-01-21.
22695
22696References
22697
22698   1. http://gcc.gnu.org/egcs-1.0/features.html
22699   2. http://gcc.gnu.org/egcs-1.0/buildstat.html
22700   3. http://gcc.gnu.org/egcs-1.0/caveats.html
22701   4. http://gcc.gnu.org/mirrors.html
22702   5. https://gcc.gnu.org/onlinedocs/
22703   6. mailto:gcc-help@gcc.gnu.org
22704   7. mailto:gcc@gcc.gnu.org
22705   8. https://gcc.gnu.org/lists.html
22706   9. https://www.fsf.org/
22707  10. https://gcc.gnu.org/about.html
22708======================================================================
22709http://gcc.gnu.org/egcs-1.0/features.html
22710
22711                               EGCS 1.0 features
22712
22713     * Core compiler is based on the gcc2 development tree from Aug 2,
22714       1997, so we have most of the [1]features found in GCC 2.8.
22715     * Integrated GNU Fortran compiler based on g77-0.5.22-19970929.
22716     * Vast improvements in the C++ compiler; so many they have [2]page of
22717       their own!
22718     * Integrated C++ runtime libraries, including support for most major
22719       GNU/Linux systems!
22720     * New instruction scheduler from IBM Haifa which includes support for
22721       function wide instruction scheduling as well as superscalar
22722       scheduling.
22723     * Significantly improved alias analysis code.
22724     * Improved register allocation for two address machines.
22725     * Significant code generation improvements for Fortran code on
22726       Alphas.
22727     * Various optimizations from the g77 project as well as improved loop
22728       optimizations.
22729     * Dwarf2 debug format support for some targets.
22730     * egcs libstdc++ includes the SGI STL implementation without changes.
22731     * As a result of these and other changes, egcs libstc++ is not binary
22732       compatible with previous releases of libstdc++.
22733     * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO
22734       Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and
22735       1.1), Support for RTEMS on several embedded targets, Support for
22736       arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and
22737       MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc.
22738     * Integrated testsuites for gcc, g++, g77, libstdc++ and libio.
22739     * RS6000/PowerPC ports generate code which can run on all
22740       RS6000/PowerPC variants by default.
22741     * -mcpu= and -march= switches for the x86 port to allow better
22742       control over how the x86 port generates code.
22743     * Includes the template repository patch (aka repo patch); note the
22744       new template code makes repo obsolete for ELF systems using gnu-ld
22745       such as GNU/Linux.
22746     * Plus the usual assortment of bugfixes and improvements.
22747
22748
22749    For questions related to the use of GCC, please consult these web
22750    pages and the [3]GCC manuals. If that fails, the
22751    [4]gcc-help@gcc.gnu.org mailing list might help. Comments on these
22752    web pages and the development of GCC are welcome on our developer
22753    list at [5]gcc@gcc.gnu.org. All of [6]our lists have public archives.
22754
22755   Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and
22756   distribution of this entire article is permitted in any medium,
22757   provided this notice is preserved.
22758
22759   These pages are [8]maintained by the GCC team. Last modified
22760   2022-10-26.
22761
22762References
22763
22764   1. http://gcc.gnu.org/egcs-1.0/features-2.8.html
22765   2. http://gcc.gnu.org/egcs-1.0/c++features.html
22766   3. https://gcc.gnu.org/onlinedocs/
22767   4. mailto:gcc-help@gcc.gnu.org
22768   5. mailto:gcc@gcc.gnu.org
22769   6. https://gcc.gnu.org/lists.html
22770   7. https://www.fsf.org/
22771   8. https://gcc.gnu.org/about.html
22772======================================================================
22773http://gcc.gnu.org/egcs-1.0/caveats.html
22774
22775                                EGCS 1.0 Caveats
22776
22777     * EGCS has an integrated libstdc++, but does not have an integrated
22778       libg++. Furthermore old libg++ releases will not work with egc; HJ
22779       Lu has made a libg++-2.8.1.2 available which may work with EGCS.
22780       Note most C++ programs only need libstdc++.
22781     * Note that using -pedantic or -Wreturn-type can cause an explosion
22782       in the amount of memory needed for template-heavy C++ code, such as
22783       code that uses STL. Also note that -Wall includes -Wreturn-type, so
22784       if you use -Wall you will need to specify -Wno-return-type to turn
22785       it off.
22786     * Exception handling may not work with shared libraries, particularly
22787       on alphas, hppas, and mips based platforms. Exception handling is
22788       known to work on x86-linux platforms with shared libraries.
22789     * Some versions of the Linux kernel have bugs which prevent them from
22790       being compiled or from running when compiled by EGCS. See the FAQ
22791       (as shipped with EGCS 1.0) for additional information.
22792     * In general, EGCS is more rigorous about rejecting invalid C++ code
22793       or deprecated C++ constructs than G++ 2.7. As a result it may be
22794       necessary to fix C++ code before it will compile with EGCS.
22795     * G++ is also aggressively tracking the C++ standard; as a result
22796       code which was previously valid (and thus accepted by other
22797       compilers and older versions of G++) may no longer be accepted.
22798     * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS
22799       1.0.x and later releases should work with Red Hat Linux 5.0.
22800
22801
22802    For questions related to the use of GCC, please consult these web
22803    pages and the [1]GCC manuals. If that fails, the
22804    [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
22805    web pages and the development of GCC are welcome on our developer
22806    list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
22807
22808   Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
22809   distribution of this entire article is permitted in any medium,
22810   provided this notice is preserved.
22811
22812   These pages are [6]maintained by the GCC team. Last modified
22813   2022-10-26.
22814
22815References
22816
22817   1. https://gcc.gnu.org/onlinedocs/
22818   2. mailto:gcc-help@gcc.gnu.org
22819   3. mailto:gcc@gcc.gnu.org
22820   4. https://gcc.gnu.org/lists.html
22821   5. https://www.fsf.org/
22822   6. https://gcc.gnu.org/about.html
22823======================================================================
22824