xref: /netbsd-src/external/gpl3/gcc/dist/NEWS (revision 867d70fc718005c0918b8b8b2f9d7f2d52d0a0db)
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-10/index.html
9                             GCC 10 Release Series
10
11   June 28, 2022
12
13   The [1]GNU project and the GCC developers are pleased to announce the
14   release of GCC 10.4.
15
16   This release is a bug-fix release, containing fixes for regressions in
17   GCC 10.3 relative to previous releases of GCC.
18
19Release History
20
21   GCC 10.4
22          June 28, 2022 ([2]changes, [3]documentation)
23
24   GCC 10.3
25          April 8, 2021 ([4]changes, [5]documentation)
26
27   GCC 10.2
28          July 23, 2020 ([6]changes, [7]documentation)
29
30   GCC 10.1
31          May 7, 2020 ([8]changes, [9]documentation)
32
33References and Acknowledgements
34
35   GCC used to stand for the GNU C Compiler, but since the compiler
36   supports several other languages aside from C, it now stands for the
37   GNU Compiler Collection.
38
39   A list of [10]successful builds is updated as new information becomes
40   available.
41
42   The GCC developers would like to thank the numerous people that have
43   contributed new features, improvements, bug fixes, and other changes as
44   well as test results to GCC. This [11]amazing group of volunteers is
45   what makes GCC successful.
46
47   For additional information about GCC please refer to the [12]GCC
48   project web site or contact the [13]GCC development mailing list.
49
50   To obtain GCC please use [14]our mirror sites or [15]our version
51   control system.
52
53
54    For questions related to the use of GCC, please consult these web
55    pages and the [16]GCC manuals. If that fails, the
56    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
57    web pages and the development of GCC are welcome on our developer
58    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
59    archives.
60
61   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
62   distribution of this entire article is permitted in any medium,
63   provided this notice is preserved.
64
65   These pages are [21]maintained by the GCC team. Last modified
66   2022-06-28[22].
67
68References
69
70   1. http://www.gnu.org/
71   2. http://gcc.gnu.org/gcc-10/changes.html
72   3. http://gcc.gnu.org/onlinedocs/10.4.0/
73   4. http://gcc.gnu.org/gcc-10/changes.html
74   5. http://gcc.gnu.org/onlinedocs/10.3.0/
75   6. http://gcc.gnu.org/gcc-10/changes.html
76   7. http://gcc.gnu.org/onlinedocs/10.2.0/
77   8. http://gcc.gnu.org/gcc-10/changes.html
78   9. http://gcc.gnu.org/onlinedocs/10.1.0/
79  10. http://gcc.gnu.org/gcc-10/buildstat.html
80  11. http://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Contributors.html
81  12. http://gcc.gnu.org/index.html
82  13. mailto:gcc@gcc.gnu.org
83  14. http://gcc.gnu.org/mirrors.html
84  15. http://gcc.gnu.org/git.html
85  16. https://gcc.gnu.org/onlinedocs/
86  17. mailto:gcc-help@gcc.gnu.org
87  18. mailto:gcc@gcc.gnu.org
88  19. https://gcc.gnu.org/lists.html
89  20. https://www.fsf.org/
90  21. https://gcc.gnu.org/about.html
91  22. http://validator.w3.org/check/referer
92======================================================================
93http://gcc.gnu.org/gcc-10/changes.html
94                             GCC 10 Release Series
95                        Changes, New Features, and Fixes
96
97   This page is a "brief" summary of some of the huge number of
98   improvements in GCC 10. You may also want to check out our [1]Porting
99   to GCC 10 page and the [2]full GCC documentation.
100
101Caveats
102
103     * An ABI incompatibility between C++14 and C++17 has been fixed. On
104       some targets a class with a zero-sized subobject would be passed
105       incorrectly when compiled as C++17 or C++20. See the [3]C++ notes
106       below for more details.
107     * The deprecated Profile Mode and array_allocator extensions have
108       been removed from libstdc++.
109     * The non-standard std::__is_nullptr_t type trait is deprecated and
110       will be removed from libstdc++ in a future release. The standard
111       trait std::is_null_pointer should be instead.
112     * The minimum version of the [4]MPFR library required for building
113       GCC has been increased to version 3.1.0 (released 2011-10-03).
114     * The automatic template instantiation at link time (-frepo) has been
115       removed.
116     * The --param allow-store-data-races internal parameter has been
117       removed in favor of a new official option -fallow-store-data-races.
118       While default behavior is unchanged and the new option allows to
119       correctly maintain a per compilation unit setting across link-time
120       optimization, alteration of the default via --param
121       allow-store-data-races will now be diagnosed and build systems have
122       to be adjusted accordingly.
123     * Offloading to Heterogeneous System Architecture Intermediate
124       Language (HSAIL) has been deprecated and will likely be removed in
125       a future release.
126     * The type of the std::iterator base class of
127       std::istreambuf_iterator was changed in C++98 mode to be consistent
128       with C++11 and later standards. See the [5]libstdc++ notes below
129       for more details.
130
131General Improvements
132
133     * New built-in functions:
134          + The [6]__has_builtin built-in preprocessor operator can be
135            used to query support for built-in functions provided by GCC
136            and other compilers that support it.
137          + __builtin_roundeven for the corresponding function from
138            ISO/IEC TS 18661.
139     * New command-line options:
140          + [7]-fallocation-dce removes unneeded pairs of new and delete
141            operators.
142          + [8]-fprofile-partial-training can now be used to inform the
143            compiler that code paths not covered by the training run
144            should not be optimized for size.
145          + [9]-fprofile-reproducible controls level of reproducibility of
146            profile gathered by [10]-fprofile-generate. This makes it
147            possible to rebuild program with same outcome which is useful,
148            for example, for distribution packages.
149          + [11]-fprofile-prefix-path can be used in combination with
150            -fprofile-generate=profile_dir and -fprofile-use=profile_dir
151            to inform GCC where the base directory of build source tree is
152            in case it differs between instrumentation and optimized
153            builds.
154          + [12]-fanalyzer enables a new static analysis pass and
155            associated warnings. This pass performs a time-consuming
156            exploration of paths through the code in the hope of detecting
157            various common errors, such as double-free bugs. This option
158            should be regarded as experimental in this release. In
159            particular, analysis of non-C code is unlikely to work.
160     * Inter-procedural optimization improvements:
161          + The inter-procedural scalar replacement of aggregates
162            (IPA-SRA) pass was re-implemented to work at link-time and can
163            now also remove computing and returning unused return values.
164          + [13]-finline-functions is now enabled at -O2 and was retuned
165            for better code size versus runtime performance trade-offs.
166            Inliner heuristics was also significantly sped up to avoid
167            negative impact to -flto -O2 compile times.
168          + Inliner heuristics and function cloning can now use
169            value-range information to predict effectivity of individual
170            transformations.
171          + During link-time optimization the C++ One Definition Rule is
172            used to increase precision of type based alias analysis.
173     * Link-time optimization improvements:
174          + A new binary [14]lto-dump has been added. It dumps various
175            information about LTO bytecode object files.
176          + The parallel phase of the LTO can automatically detect a
177            running make's jobserver or fall back to number of available
178            cores.
179          + The LTO bytecode can be compressed with the [15]zstd
180            algorithm. The configure script automatically detects zstd
181            support.
182          + Most --param values can now be specified at translation unit
183            granularity. This includes all parameters controlling the
184            inliner and other inter-procedural optimizations. Unlike
185            earlier releases, GCC 10 will ignore parameters controlling
186            optimizations specified at link-time and apply parameters
187            specified at compile-time in the same manner as done for
188            optimization flags.
189     * Profile driven optimization improvements:
190          + Profile maintenance during compilation and hot/cold code
191            partitioning have been improved.
192          + Using [16]-fprofile-values, an instrumented binary can track
193            multiple values (up to 4) for e.g. indirect calls and provide
194            more precise profile information.
195
196New Languages and Language-Specific Improvements
197
198     * Version 2.6 of the [17]OpenACC specification is now supported in
199       the C, C++ and Fortran compilers. See the [18]implementation status
200       section on the OpenACC wiki page and the [19]run-time library
201       documentation for further information.
202     * GCC 10 adds a number of newly implemented [20]OpenMP 5.0 features
203       on top of the GCC 9 release such as conditional lastprivate clause,
204       scan and loop directives, order(concurrent) and use_device_addr
205       clauses support, if clause on simd construct or partial support for
206       the declare variant directive, getting closer to full support of
207       the OpenMP 5.0 standard.
208     * OpenMP and OpenACC now support [21]offloading to AMD Radeon (GCN)
209       GPUs; supported are the third-generation Fiji (fiji) and the
210       fifth-generation VEGA 10/VEGA 20 (gfx900 or gfx906).
211
212  C family
213
214     * New attributes:
215          + The access function and type attribute has been added to
216            describe how a function accesses objects passed to it by
217            pointer or reference, and to associate such arguments with
218            integer arguments denoting the objects' sizes. The attribute
219            is used to enable the detection of invalid accesses by
220            user-defined functions, such as those diagnosed by
221            -Wstringop-overflow.
222          + The symver attribute can be used to bind symbols to specific
223            version nodes on ELF platforms. This is preferred to using
224            inline assembly with GNU as symver directive because the
225            latter is not compatible with link-time optimizations.
226     * New warnings:
227          + [22]-Wstring-compare, enabled by -Wextra, warns about equality
228            and inequality expressions between zero and the result of a
229            call to either strcmp and strncmp that evaluate to a constant
230            as a result of the length of one argument being greater than
231            the size of the array pointed to by the other.
232          + [23]-Wzero-length-bounds, enabled by -Warray-bounds, warns
233            about accesses to elements of zero-length arrays that might
234            overlap other members of the same object.
235     * Enhancements to existing warnings:
236          + [24]-Warray-bounds detects more out-of-bounds accesses to
237            member arrays as well as accesses to elements of zero-length
238            arrays.
239          + [25]-Wformat-overflow makes full use of string length
240            information computed by the strlen optimization pass.
241          + [26]-Wrestrict detects overlapping accesses to dynamically
242            allocated objects.
243          + [27]-Wreturn-local-addr diagnoses more instances of return
244            statements returning addresses of automatic variables.
245          + [28]-Wstringop-overflow detects more out-of-bounds stores to
246            member arrays including zero-length arrays, dynamically
247            allocated objects and variable length arrays, as well as more
248            instances of reads of unterminated character arrays by string
249            built-in functions. The warning also detects out-of-bounds
250            accesses by calls to user-defined functions declared with the
251            new attribute access.
252          + [29]-Warith-conversion re-enables warnings from -Wconversion,
253            -Wfloat-conversion, and -Wsign-conversion that are now off by
254            default for an expression where the result of an arithmetic
255            operation will not fit in the target type due to promotion,
256            but the operands of the expression do fit in the target type.
257     * Extended characters in identifiers may now be specified directly in
258       the input encoding (UTF-8, by default), in addition to the UCN
259       syntax (\uNNNN or \UNNNNNNNN) that is already supported:
260
261static const int π = 3;
262int get_naïve_pi() {
263  return π;
264}
265
266  C
267
268     * Several new features from the upcoming C2X revision of the ISO C
269       standard are supported with -std=c2x and -std=gnu2x. Some of these
270       features are also supported as extensions when compiling for older
271       language versions. In addition to the features listed, some
272       features previously supported as extensions and now added to the C
273       standard are enabled by default in C2X mode and not diagnosed with
274       -std=c2x -Wpedantic.
275          + The [[]] attribute syntax is supported, as in C++. Existing
276            attributes can be used with this syntax in forms such as
277            [[gnu::const]]. The standard attributes [[deprecated]],
278            [[fallthrough]] and [[maybe_unused]] are supported.
279          + UTF-8 character constants using the u8'' syntax are supported.
280          + <float.h> defines macros FLT_NORM_MAX, DBL_NORM_MAX and
281            LDBL_NORM_MAX.
282          + When decimal floating-point arithmetic is supported, <float.h>
283            defines macros DEC32_TRUE_MIN, DEC64_TRUE_MIN and
284            DEC128_TRUE_MIN, in addition to the macros that were
285            previously only defined if __STDC_WANT_DEC_FP__ was defined
286            before including <float.h>.
287          + In C2X mode, empty parentheses in a function definition give
288            that function a type with a prototype for subsequent calls;
289            other old-style function definitions are diagnosed by default
290            in C2X mode.
291          + The strftime format checking supports the %OB and %Ob formats.
292          + In C2X mode, -fno-fp-int-builtin-inexact is enabled by
293            default.
294     * GCC now defaults to -fno-common. As a result, global variable
295       accesses are more efficient on various targets. In C, global
296       variables with multiple tentative definitions now result in linker
297       errors. With -fcommon such definitions are silently merged during
298       linking.
299
300  C++
301
302     * Several C++20 features have been implemented:
303          + Concepts, including P0734R0, P0857R0, P1084R2, P1141R2,
304            P0848R3, P1616R1, P1452R2
305          + P1668R1, Permitting Unevaluated inline-assembly in constexpr
306            Functions
307          + P1161R3, Deprecate a[b,c]
308          + P0848R3, Conditionally Trivial Special Member Functions
309          + P1091R3, Extending structured bindings
310          + P1143R2, Adding the constinit keyword
311          + P1152R4, Deprecating volatile
312          + P0388R4, Permit conversions to arrays of unknown bound
313          + P0784R7, constexpr new
314          + P1301R4, [[nodiscard("with reason")]]
315          + P1814R0, class template argument deduction for alias templates
316          + P1816R0, class template argument deduction for aggregates
317          + P0960R3, Parenthesized initialization of aggregates
318          + P1331R2, Allow trivial default initialization in constexpr
319            contexts
320          + P1327R1, Allowing dynamic_cast and polymorphic typeid in
321            constexpr contexts
322          + P0912R5, Coroutines (requires -fcoroutines)
323     * Several C++ Defect Reports have been resolved, e.g.:
324          + DR 1560, lvalue-to-rvalue conversion in ?:
325          + DR 1813, __is_standard_layout for a class with repeated bases
326          + DR 2094, volatile scalars are trivially copyable,
327          + DR 2096, constraints on literal unions
328          + DR 2413, typename in conversion-function-ids
329          + DR 2352, Similar types and reference binding
330          + DR 1601, Promotion of enumeration with fixed underlying type
331          + DR 330, Qualification conversions and pointers to arrays of
332            pointers
333          + DR 1307, Overload resolution based on size of array
334            initializer-list
335          + DR 1710, Missing template keyword in class-or-decltype
336     * New warnings:
337          + [30]-Wmismatched-tags, disabled by default, warns about
338            declarations of structs, classes, and class templates and
339            their specializations with a class-key that does not match
340            either the definition or the first declaration if no
341            definition is provided. The option is provided to ease
342            portability to Windows-based compilers.
343          + [31]-Wredundant-tags, disabled by default, warns about
344            redundant class-key and enum-key in contexts where the key can
345            be eliminated without causing an syntactic ambiguity.
346     * G++ can now detect modifying constant objects in constexpr
347       evaluation (which is undefined behavior).
348     * G++ no longer emits bogus -Wsign-conversion warnings with explicit
349       casts.
350     * Narrowing is now detected in more contexts (e.g., case values).
351     * Memory consumption of the compiler has been reduced in constexpr
352       evaluation.
353     * The noexcept-specifier is now properly treated as a complete-class
354       context as per [class.mem].
355     * The attribute deprecated can now be used on namespaces too.
356     * The ABI of passing and returning certain C++ classes by value
357       changed on several targets in GCC 10, including [32]AArch64,
358       [33]ARM, [34]PowerPC ELFv2, [35]S/390 and [36]Itanium. These
359       changes affect classes with a zero-sized subobject (an empty base
360       class, or data member with the [[no_unique_address]] attribute)
361       where all other non-static data members have the same type (this is
362       called a "homogeneous aggregate" in some ABI specifications, or if
363       there is only one such member, a "single element"). In -std=c++17
364       and -std=c++20 modes, classes with an empty base class were not
365       considered to have a single element or to be a homogeneous
366       aggregate, and so could be passed differently (in the wrong
367       registers or at the wrong stack address). This could make code
368       compiled with -std=c++17 and -std=c++14 ABI incompatible. This has
369       been corrected and the empty bases are ignored in those ABI
370       decisions, so functions compiled with -std=c++14 and -std=c++17 are
371       now ABI compatible again. Example: struct empty {}; struct S :
372       empty { float f; }; void f(S);. Similarly, in classes containing
373       non-static data members with empty class types using the C++20
374       [[no_unique_address]] attribute, those members weren't ignored in
375       the ABI argument passing decisions as they should be. Both of these
376       ABI changes are now diagnosed with -Wpsabi.
377
378    Runtime Library (libstdc++)
379
380     * Improved experimental C++2a support, including:
381          + Library concepts in <concepts> and <iterator>.
382          + Constrained algorithms in <ranges>, <algorithm>, and <memory>
383            (thanks to Patrick Palka).
384          + New algorithms shift_left and shift_right (thanks to Patrick
385            Palka).
386          + std::span (thanks to JeanHeyd Meneide).
387          + Three-way comparisons in <compare> and throughout the library.
388          + Constexpr support in <algorithm> and elsewhere (thanks to
389            Edward Smith-Rowland).
390          + <stop_token> and std::jthread (thanks to Thomas Rodgers).
391          + std::atomic_ref and std::atomic<floating point>.
392          + Integer comparison functions (cmp_equal, cmp_less etc.).
393          + std::ssize, std::to_array.
394          + std::construct_at, std::destroy, constexpr std::allocator.
395          + Mathematical constants in <numbers>.
396     * Support for RDSEED in std::random_device.
397     * Reduced header dependencies, leading to faster compilation for some
398       code.
399     * The std::iterator base class of std::istreambuf_iterator was
400       changed in C++98 mode to be consistent with C++11 and later
401       standards. This is expected to have no noticeable effect except in
402       the unlikely case of a class which has potentially overlapping
403       subobjects of type std::istreambuf_iterator<C> and another iterator
404       type with a std::iterator<input_iterator_tag, C, ...> base class.
405       The layout of such a type might change when compiled as C++98.
406       [37]Bug 92285 has more details and concrete examples.
407
408  D
409
410     * Support for static foreach has been implemented.
411     * Aliases can now be created directly from any __traits that return
412       symbols or tuples. Previously, an AliasSeq was necessary in order
413       to alias their return.
414     * It is now possible to detect the language ABI specified for a
415       struct, class, or interface using __traits(getLinkage, ...).
416     * Support for core.math.toPrec intrinsics has been added. These
417       intrinsics guarantee the rounding to specific floating-point
418       precisions at specified points in the code.
419     * Support for pragma(inline) has been implemented. Previously the
420       pragma was recognized, but had no effect on the compilation.
421     * Optional parentheses in asm operands are deprecated and will be
422       removed in a future release.
423     * All content imported files are now included in the make dependency
424       list when compiling with -M.
425     * Compiler recognized attributes provided by the gcc.attribute module
426       will now take effect when applied to function prototypes as well as
427       when applied to full function declarations.
428     * Added a --enable-libphobos-checking configure option to control
429       whether run-time checks are compiled into the D runtime library.
430     * Added a --with-libphobos-druntime-only configure option to indicate
431       whether to build only the core D runtime library, or both the core
432       and standard libraries into libphobos.
433
434  Fortran
435
436     * use_device_addr of version 5.0 of the [38]OpenMP specification is
437       now supported. Note that otherwise OpenMP 4.5 is partially
438       supported in the Fortran compiler; the largest missing item is
439       structure element mapping.
440     * The default buffer size for I/O using unformatted files has been
441       increased to 1048576. The buffer size for can now be set at runtime
442       via the environment variables GFORTRAN_FORMATTED_BUFFER_SIZE and
443       GFORTRAN_UNFORMATTED_BUFFER_SIZE for formatted and unformatted
444       files, respectively.
445     * Mismatches between actual and dummy argument lists in a single file
446       are now rejected with an error. Use the new option
447       -fallow-argument-mismatch to turn these errors into warnings; this
448       option is implied with -std=legacy. -Wargument-mismatch has been
449       removed.
450     * The handling of a BOZ literal constant has been reworked to provide
451       better conformance to the Fortran 2008 and 2018 standards. In these
452       Fortran standards, a BOZ literal constant is a typeless and
453       kindless entity. As a part of the rework, documented and
454       undocumented extensions to the Fortran standard now emit errors
455       during compilation. Some of these extensions are permitted with the
456       -fallow-invalid-boz option, which degrades the error to a warning
457       and the code is compiled as with older gfortran.
458     * At any optimization level except-Os, gfortran now uses inline
459       packing for arguments instead of calling a library routine. If the
460       source contains a large number of arguments that need to be
461       repacked, code size or time for compilation can become excessive.
462       If that is the case, -fno-inline-arg-packing can be used to disable
463       inline argument packing.
464     * Legacy extensions:
465          + For formatted input/output, if the explicit widths after the
466            data-edit descriptors I, F and G have been omitted, default
467            widths are used.
468          + A blank format item at the end of a format specification, i.e.
469            nothing following the final comma, is allowed. Use the option
470            -fdec-blank-format-item; this option is implied with -fdec.
471          + The existing support for AUTOMATIC and STATIC attributes has
472            been extended to allow variables with the AUTOMATIC attribute
473            to be used in EQUIVALENCE statements. Use -fdec-static; this
474            option is implied by -fdec.
475          + Allow character literals in assignments and DATA statements
476            for numeric (INTEGER, REAL, or COMPLEX) or LOGICAL variables.
477            Use the option -fdec-char-conversions; this option is implied
478            with -fdec.
479          + DEC comparisons, i.e. allow Hollerith constants to be used in
480            comparisons with INTEGER, REAL, COMPLEX and CHARACTER
481            expressions. Use the option -fdec.
482     * Character type names in errors and warnings now include len in
483       addition to kind; * is used for assumed length. The kind is omitted
484       if it is the default kind. Examples: CHARACTER(12), CHARACTER(6,4).
485     * CO_BROADCAST now supports derived type variables including objects
486       with allocatable components. In this case, the optional arguments
487       STAT= and ERRMSG= are currently ignored.
488     * The handling of module and submodule names has been reworked to
489       allow the full 63-character length mandated by the standard.
490       Previously symbol names were truncated if the combined length of
491       module, submodule, and function name exceeded 126 characters. This
492       change therefore breaks the ABI, but only for cases where this 126
493       character limit was exceeded.
494
495  Go
496
497     * GCC 10 provides a complete implementation of the Go 1.14.6 user
498       packages.
499
500libgccjit
501
502     * The libgccjit API gained four new entry points:
503          + [39]gcc_jit_version_major, [40]gcc_jit_version_minor, and
504            [41]gcc_jit_version_patchlevel for programmatically checking
505            the libgccjit version from client code, and
506          + [42]gcc_jit_context_new_bitfield
507
508New Targets and Target Specific Improvements
509
510  AArch64 & arm
511
512     * The AArch64 and arm ports now support condition flag output
513       constraints in inline assembly, as indicated by the
514       __GCC_ASM_FLAG_OUTPUTS__. On arm this feature is only available for
515       A32 and T32 targets. Please refer to the documentation for more
516       details.
517
518  AArch64
519
520     * There have been several improvements related to the Scalable Vector
521       Extension (SVE):
522          + The SVE ACLE types and intrinsics are now supported. They can
523            be accessed using the header file arm_sve.h.
524          + It is now possible to create fixed-length SVE types using the
525            arm_sve_vector_bits attribute. For example:
526#if __ARM_FEATURE_SVE_BITS==512
527typedef svint32_t vec512 __attribute__((arm_sve_vector_bits(512)));
528typedef svbool_t pred512 __attribute__((arm_sve_vector_bits(512)));
529#endif
530          + -mlow-precision-div, -mlow-precision-sqrt and
531            -mlow-precision-recip-sqrt now work for SVE.
532          + -msve-vector-bits=128 now generates vector-length-specific
533            code for little-endian targets. It continues to generate
534            vector-length-agnostic code for big-endian targets, just as
535            previous releases did for all targets.
536          + The vectorizer is now able to use extending loads and
537            truncating stores, including gather loads and scatter stores.
538          + The vectorizer now compares the cost of vectorizing with SVE
539            and vectorizing with Advanced SIMD and tries to pick the best
540            one. Previously it would always use SVE if possible.
541          + If a vector loop uses Advanced SIMD rather than SVE, the
542            vectorizer now considers using SVE to vectorize the left-over
543            elements (the “scalar tail” or “epilog”).
544          + Besides these specific points, there have been many general
545            improvements to the way that the vectorizer uses SVE.
546     * The -mbranch-protection=pac-ret option now accepts the optional
547       argument +b-key extension to perform return address signing with
548       the B-key instead of the A-key.
549     * The option -moutline-atomics has been added to aid deployment of
550       the Large System Extensions (LSE) on GNU/Linux systems built with a
551       baseline architecture targeting Armv8-A. When the option is
552       specified code is emitted to detect the presence of LSE
553       instructions at runtime and use them for standard atomic
554       operations. For more information please refer to the documentation.
555     * The Transactional Memory Extension is now supported through ACLE
556       intrinsics. It can be enabled through the +tme option extension
557       (for example, -march=armv8.5-a+tme).
558     * A number of features from Armv8.5-A are now supported through ACLE
559       intrinsics. These include:
560          + The random number instructions that can be enabled through the
561            (already present in GCC 9.1) +rng option extension.
562          + Floating-point intrinsics to round to integer instructions
563            from Armv8.5-A when targeting -march=armv8.5-a or later.
564          + Memory Tagging Extension intrinsics enabled through the
565            +memtag option extension.
566     * Similarly, the following Armv8.6-A features are now supported
567       through ACLE intrinsics:
568          + The bfloat16 extension. This extension is enabled
569            automatically when Armv8.6-A is selected (such as by
570            -march=armv8.6-a). It can also be enabled for Armv8.2-A and
571            later using the +bf16 option extension.
572          + The Matrix Multiply extension. This extension is split into
573            three parts, one for each supported data type:
574               o Support for 8-bit integer matrix multiply instructions.
575                 This extension is enabled automatically when Armv8.6-A is
576                 selected. It can also be enabled for Armv8.2-A and later
577                 using the +i8mm option extension.
578               o Support for 32-bit floating-point matrix multiply
579                 instructions. This extension can be enabled using the
580                 +f32mm option extension, which also has the effect of
581                 enabling SVE.
582               o Support for 64-bit floating-point matrix multiply
583                 instructions. This extension can be enabled using the
584                 +f64mm option extension, which likewise has the effect of
585                 enabling SVE.
586     * SVE2 is now supported through ACLE intrinsics and (to a limited
587       extent) through autovectorization. It can be enabled through the
588       +sve2 option extension (for example, -march=armv8.5-a+sve2).
589       Additional extensions can be enabled through +sve2-sm4, +sve2-aes,
590       +sve2-sha3 and +sve2-bitperm.
591     * Support has been added for the following processors (GCC
592       identifiers in parentheses):
593          + Arm Cortex-A77 (cortex-a77).
594          + Arm Cortex-A76AE (cortex-a76ae).
595          + Arm Cortex-A65 (cortex-a65).
596          + Arm Cortex-A65AE (cortex-a65ae).
597          + Arm Cortex-A34 (cortex-a34).
598          + Marvell ThunderX3 (thunderx3t110).
599       The GCC identifiers can be used as arguments to the -mcpu or -mtune
600       options, for example: -mcpu=cortex-a77 or -mtune=cortex-a65ae or as
601       arguments to the equivalent target attributes and pragmas.
602
603  arm
604
605     * Support for the FDPIC ABI has been added. It uses 64-bit function
606       descriptors to represent pointers to functions, and enables code
607       sharing on MMU-less systems. The corresponding target triple is
608       arm-uclinuxfdpiceabi, and the C library is uclibc-ng.
609     * Support has been added for the Arm EABI on NetBSD through the
610       arm*-*-netbsdelf-*eabi* triplet.
611     * The handling of 64-bit integer operations has been significantly
612       reworked and improved leading to improved performance and reduced
613       stack usage when using 64-bit integral data types. The option
614       -mneon-for-64bits is now deprecated and will be removed in a future
615       release.
616     * Support has been added for the following processors (GCC
617       identifiers in parentheses):
618          + Arm Cortex-A77 (cortex-a77).
619          + Arm Cortex-A76AE (cortex-a76ae).
620          + Arm Cortex-M35P (cortex-m35p).
621          + Arm Cortex-M55 (cortex-m55).
622       The GCC identifiers can be used as arguments to the -mcpu or -mtune
623       options, for example: -mcpu=cortex-a77 or -mtune=cortex-m35p.
624     * Support has been extended for the ACLE [43]data-processing
625       intrinsics to include 32-bit SIMD, saturating arithmetic, 16-bit
626       multiplication and other related intrinsics aimed at DSP algorithm
627       optimization.
628     * Support for -mpure-code in Thumb-1 (v6m) has been added: this
629       M-profile feature is no longer restricted to targets with MOVT. For
630       example, -mcpu=cortex-m0 now supports this option.
631     * Support for the [44]Armv8.1-M Mainline Architecture has been added.
632          + Armv8.1-M Mainline can be enabled by using the
633            -march=armv8.1-m.main command-line option.
634     * Support for the [45]MVE beta ACLE intrinsics has been added. These
635       intrinsics can be enabled by including the arm_mve.h header file
636       and passing the +mve or +mve.fp option extensions (for example:
637       -march=armv8.1-m.main+mve).
638     * Support for the Custom Datapath Extension beta ACLE [46]intrinsics
639       has been added.
640     * Support for Armv8.1-M Mainline Security Extensions architecture has
641       been added. The -mcmse option, when used in combination with an
642       Armv8.1-M Mainline architecture (for example: -march=armv8.1-m.main
643       -mcmse), now leads to the generation of improved code sequences
644       when changing security states.
645
646  AMD Radeon (GCN)
647
648     * The code generation and in particular the vectorization support has
649       been much improved.
650
651  ARC
652
653     * The interrupt service routine functions save all used registers,
654       including extension registers and auxiliary registers used by Zero
655       Overhead Loops.
656     * Improve code size by using multiple short instructions instead of a
657       single long mov or ior instruction when its long immediate constant
658       is known.
659     * Fix usage of the accumulator register for ARC600.
660     * Fix issues with uncached attribute.
661     * Remove -mq-class option.
662     * Improve 64-bit integer addition and subtraction operations.
663
664  AVR
665
666     * Support for the XMEGA-like devices
667
668     ATtiny202, ATtiny204, ATtiny402, ATtiny404, ATtiny406, ATtiny804,
669     ATtiny806, ATtiny807, ATtiny1604, ATtiny1606, ATtiny1607, ATmega808,
670     ATmega809, ATmega1608, ATmega1609, ATmega3208, ATmega3209,
671     ATmega4808, ATmega4809
672       has been added.
673     * A new command-line option -nodevicespecs has been added. It allows
674       to provide a custom device-specs file by means of
675
676     avr-gcc -nodevicespecs -specs=my-spec-file <options>
677       and without the need to provide options -B and -mmcu=. See [47]AVR
678       command-line options for details. This feature is also available in
679       GCC 9.3+ and GCC 8.4+.
680     * New command-line options -mdouble=[32,64] and -mlong-double=[32,64]
681       have been added. They allow to choose the size (in bits) of the
682       double and long double types, respectively. Whether or not the
683       mentioned layouts are available, whether the options act as a
684       multilib option, and the default for either option are controlled
685       by the new [48]AVR configure options --with-double= and
686       --with-long-double=.
687     * A new configure option --with-libf7= has been added. It controls to
688       which level avr-libgcc provides 64-bit floating point support by
689       means of [49]Libf7.
690     * A new configure option --with-double-comparison= has been added.
691       It's unlikely you need to set this option by hand.
692
693  IA-32/x86-64
694
695     * Support to expand __builtin_roundeven into the appropriate SSE 4.1
696       instruction has been added.
697     * New ISA extension support for Intel ENQCMD was added to GCC. ENQCMD
698       intrinsics are available via the -menqcmd compiler switch.
699     * GCC now supports the Intel CPU named Cooperlake through
700       -march=cooperlake. The switch enables the AVX512BF16 ISA
701       extensions.
702     * GCC now supports the Intel CPU named Tigerlake through
703       -march=tigerlake. The switch enables the MOVDIRI MOVDIR64B
704       AVX512VP2INTERSECT ISA extensions.
705
706  MIPS
707
708     * The mips*-*-linux* targets now mark object files with appropriate
709       GNU-stack note, facilitating use of non-executable stack hardening
710       on GNU/Linux. The soft-float targets have this feature enabled by
711       default, while for hard-float targets it is required for GCC to be
712       configured with --with-glibc-version=2.31 against glibc 2.31 or
713       later.
714
715  PowerPC / PowerPC64 / RS6000
716
717     * Many vector builtins have been listed as deprecated in the
718       [50]64-Bit ELF V2 ABI Specification for quite a number of years.
719       The vector builtins listed in Tables A.8 through A.10 are now
720       deprecated for GCC 10, and will likely be removed from support in
721       GCC 11. Note that this does not result in any loss of function.
722       These deprecated builtins generally provide somewhat nonsensical
723       argument lists (for example, mixing signed, unsigned, and bool
724       vector arguments arbitrarily), or are duplicate builtins that are
725       inconsistent with the expected naming scheme. We expect that this
726       will be unlikely to affect much if any code, and any required code
727       changes will be trivial.
728
729  PRU
730
731     * A new back end targeting TI PRU I/O processors has been contributed
732       to GCC.
733
734  RISC-V
735
736     * The riscv*-*-* targets now require GNU binutils version 2.30 or
737       later, to support new assembly instructions produced by GCC.
738
739  V850
740
741     * The ABI for V850 nested functions has been changed. Previously the
742       V850 port used %r20 for the static chain pointer, now the port uses
743       %r19. This corrects a long standing latent bug in the v850 port
744       where a call to a nested function would unexpectedly change the
745       value in %r20.
746
747Operating Systems
748
749Improvements for plugin authors
750
751     * GCC diagnostics can now have a chain of events associated with
752       them, describing a path through the code that triggers the problem.
753       These can be printed by the diagnostics subsystem in various ways,
754       controlled by the [51]-fdiagnostics-path-format option, or captured
755       in JSON form via [52]-fdiagnostics-format=json.
756     * GCC diagnostics can now be associated with [53]CWE weakness
757       identifiers, which will appear on the standard error stream, and in
758       the JSON output from [54]-fdiagnostics-format=json.
759
760Other significant improvements
761
762     * To allow inline expansion of both memcpy and memmove, the existing
763       movmem instruction patterns used for non-overlapping memory copies
764       have been renamed to cpymem. The movmem name is now used for
765       overlapping memory moves, consistent with the library functions
766       memcpy and memmove.
767     * For many releases, when GCC emits a warning it prints the option
768       controlling that warning. As of GCC 10, that option text is now a
769       clickable hyperlink for the documentation of that option (assuming
770       a [55]sufficiently capable terminal). This behavior can be
771       controlled via a new [56]-fdiagnostics-urls option (along with
772       various environment variables and heuristics documented with that
773       option).
774
775GCC 10.1
776
777   This is the [57]list of problem reports (PRs) from GCC's bug tracking
778   system that are known to be fixed in the 10.1 release. This list might
779   not be complete (that is, it is possible that some PRs that have been
780   fixed are not listed here).
781
782GCC 10.2
783
784   This is the [58]list of problem reports (PRs) from GCC's bug tracking
785   system that are known to be fixed in the 10.2 release. This list might
786   not be complete (that is, it is possible that some PRs that have been
787   fixed are not listed here).
788
789GCC 10.3
790
791   This is the [59]list of problem reports (PRs) from GCC's bug tracking
792   system that are known to be fixed in the 10.3 release. This list might
793   not be complete (that is, it is possible that some PRs that have been
794   fixed are not listed here).
795
796  Target Specific Changes
797
798    AArch64
799
800     * A bug with the Random Number intrinsics in the arm_acle.h header
801       that resulted in an incorrect status result being returned has been
802       fixed.
803     * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune
804       options are -mcpu=a64fx and -mtune=a64fx respectively. In
805       particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and
806       tunes the code for the A64FX. This includes tuning the SVE code,
807       although by default the code is still length-agnostic and so works
808       for all SVE implementations. Adding -msve-vector-bits=512 makes the
809       code specific to 512-bit SVE.
810
811  x86-64
812
813     * GCC 10.3 supports AMD CPUs based on the znver3 core via
814       -march=znver3.
815
816GCC 10.4
817
818   This is the [60]list of problem reports (PRs) from GCC's bug tracking
819   system that are known to be fixed in the 10.4 release. This list might
820   not be complete (that is, it is possible that some PRs that have been
821   fixed are not listed here).
822
823  Target Specific Changes
824
825    x86-64
826
827     * The x86-64 ABI of passing and returning structures with a 64-bit
828       integer vector changed in GCC 10.1 when MMX is disabled. Disabling
829       MMX no longer changes how they are passed nor returned. This ABI
830       change is now diagnosed with -Wpsabi.
831
832
833    For questions related to the use of GCC, please consult these web
834    pages and the [61]GCC manuals. If that fails, the
835    [62]gcc-help@gcc.gnu.org mailing list might help. Comments on these
836    web pages and the development of GCC are welcome on our developer
837    list at [63]gcc@gcc.gnu.org. All of [64]our lists have public
838    archives.
839
840   Copyright (C) [65]Free Software Foundation, Inc. Verbatim copying and
841   distribution of this entire article is permitted in any medium,
842   provided this notice is preserved.
843
844   These pages are [66]maintained by the GCC team. Last modified
845   2022-06-28[67].
846
847References
848
849   1. http://gcc.gnu.org/gcc-10/porting_to.html
850   2. http://gcc.gnu.org/onlinedocs/index.html#current
851   3. http://gcc.gnu.org/gcc-10/changes.html#empty_base
852   4. https://www.mpfr.org/
853   5. http://gcc.gnu.org/gcc-10/changes.html#iterator_base
854   6. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/cpp/_005f_005fhas_005fbuiltin.html#g_t_005f_005fhas_005fbuiltin
855   7. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fno-allocation-dce
856   8. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-partial-training
857   9. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-reproducible
858  10. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-generate
859  11. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Instrumentation-Options.html#index-fprofile-prefix-path
860  12. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Static-Analyzer-Options.html
861  13. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-finline-functions
862  14. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/lto-dump.html
863  15. https://facebook.github.io/zstd/
864  16. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Optimize-Options.html#index-fprofile-values
865  17. https://www.openacc.org/
866  18. https://gcc.gnu.org/wiki/OpenACC/Implementation Status#status-10
867  19. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/libgomp/#toc-Enabling-OpenACC-1
868  20. https://www.openmp.org/specifications/
869  21. https://gcc.gnu.org/wiki/Offloading
870  22. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstring-compare
871  23. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wzero-length-bounds
872  24. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warray-bounds
873  25. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wformat-overflow
874  26. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wrestrict
875  27. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wreturn-local-addr
876  28. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Wstringop-overflow
877  29. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Warning-Options.html#index-Warith-conversion
878  30. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wmismatched-tags
879  31. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-tags
880  32. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94383
881  33. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94711
882  34. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94707
883  35. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94704
884  36. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94706
885  37. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92285
886  38. https://www.openmp.org/specifications/
887  39. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_major
888  40. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_minor
889  41. https://gcc.gnu.org/onlinedocs/jit/topics/compatibility.html#c.gcc_jit_version_patchlevel
890  42. https://gcc.gnu.org/onlinedocs/jit/topics/types.html#c.gcc_jit_context_new_bitfield
891  43. https://developer.arm.com/documentation/101028/0009/Data-processing-intrinsics
892  44. https://developer.arm.com/architectures/cpu-architecture/m-profile
893  45. https://developer.arm.com/architectures/instruction-sets/simd-isas/helium/helium-intrinsics
894  46. https://developer.arm.com/documentation/101028/0010/Custom-Datapath-Extension
895  47. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/AVR-Options.html#index-nodevicespecs
896  48. https://gcc.gnu.org/install/configure.html#avr
897  49. https://gcc.gnu.org/wiki/avr-gcc#Libf7
898  50. https://openpowerfoundation.org/?resource_lib=64-bit-elf-v2-abi-specification-power-architecture
899  51. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-path-format
900  52. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format
901  53. https://cwe.mitre.org/
902  54. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format
903  55. https://gist.github.com/egmontkob/eb114294efbcd5adb1944c9f3cb5feda
904  56. https://gcc.gnu.org/onlinedocs/gcc-10.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-urls
905  57. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.0
906  58. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.2
907  59. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.3
908  60. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=10.4
909  61. https://gcc.gnu.org/onlinedocs/
910  62. mailto:gcc-help@gcc.gnu.org
911  63. mailto:gcc@gcc.gnu.org
912  64. https://gcc.gnu.org/lists.html
913  65. https://www.fsf.org/
914  66. https://gcc.gnu.org/about.html
915  67. http://validator.w3.org/check/referer
916======================================================================
917http://gcc.gnu.org/gcc-9/index.html
918                              GCC 9 Release Series
919
920   May 27, 2022
921
922   The [1]GNU project and the GCC developers are pleased to announce the
923   release of GCC 9.5.
924
925   This release is a bug-fix release, containing fixes for regressions in
926   GCC 9.4 relative to previous releases of GCC.
927
928Release History
929
930   GCC 9.5
931          May 27, 2022 ([2]changes, [3]documentation)
932
933   GCC 9.4
934          June 1, 2021 ([4]changes, [5]documentation)
935
936   GCC 9.3
937          Mar 12, 2020 ([6]changes, [7]documentation)
938
939   GCC 9.2
940          Aug 12, 2019 ([8]changes, [9]documentation)
941
942   GCC 9.1
943          May 3, 2019 ([10]changes, [11]documentation)
944
945References and Acknowledgements
946
947   GCC used to stand for the GNU C Compiler, but since the compiler
948   supports several other languages aside from C, it now stands for the
949   GNU Compiler Collection.
950
951   A list of [12]successful builds is updated as new information becomes
952   available.
953
954   The GCC developers would like to thank the numerous people that have
955   contributed new features, improvements, bug fixes, and other changes as
956   well as test results to GCC. This [13]amazing group of volunteers is
957   what makes GCC successful.
958
959   For additional information about GCC please refer to the [14]GCC
960   project web site or contact the [15]GCC development mailing list.
961
962   To obtain GCC please use [16]our mirror sites or [17]our version
963   control system.
964
965
966    For questions related to the use of GCC, please consult these web
967    pages and the [18]GCC manuals. If that fails, the
968    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
969    web pages and the development of GCC are welcome on our developer
970    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
971    archives.
972
973   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
974   distribution of this entire article is permitted in any medium,
975   provided this notice is preserved.
976
977   These pages are [23]maintained by the GCC team. Last modified
978   2022-05-27[24].
979
980References
981
982   1. http://www.gnu.org/
983   2. http://gcc.gnu.org/gcc-9/changes.html
984   3. http://gcc.gnu.org/onlinedocs/9.5.0/
985   4. http://gcc.gnu.org/gcc-9/changes.html
986   5. http://gcc.gnu.org/onlinedocs/9.4.0/
987   6. http://gcc.gnu.org/gcc-9/changes.html
988   7. http://gcc.gnu.org/onlinedocs/9.3.0/
989   8. http://gcc.gnu.org/gcc-9/changes.html
990   9. http://gcc.gnu.org/onlinedocs/9.2.0/
991  10. http://gcc.gnu.org/gcc-9/changes.html
992  11. http://gcc.gnu.org/onlinedocs/9.1.0/
993  12. http://gcc.gnu.org/gcc-9/buildstat.html
994  13. http://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Contributors.html
995  14. http://gcc.gnu.org/index.html
996  15. mailto:gcc@gcc.gnu.org
997  16. http://gcc.gnu.org/mirrors.html
998  17. http://gcc.gnu.org/git.html
999  18. https://gcc.gnu.org/onlinedocs/
1000  19. mailto:gcc-help@gcc.gnu.org
1001  20. mailto:gcc@gcc.gnu.org
1002  21. https://gcc.gnu.org/lists.html
1003  22. https://www.fsf.org/
1004  23. https://gcc.gnu.org/about.html
1005  24. http://validator.w3.org/check/referer
1006======================================================================
1007http://gcc.gnu.org/gcc-9/changes.html
1008                              GCC 9 Release Series
1009                        Changes, New Features, and Fixes
1010
1011   This page is a "brief" summary of some of the huge number of
1012   improvements in GCC 9. You may also want to check out our [1]Porting to
1013   GCC 9 page and the [2]full GCC documentation.
1014
1015Caveats
1016
1017     * On Arm targets (arm*-*-*), [3]a bug in the implementation of the
1018       procedure call standard (AAPCS) in the GCC 6, 7 and 8 releases has
1019       been fixed: a structure containing a bit-field based on a 64-bit
1020       integral type and where no other element in a structure required
1021       64-bit alignment could be passed incorrectly to functions. This is
1022       an ABI change. If the option -Wpsabi is enabled (on by default) the
1023       compiler will emit a diagnostic note for code that might be
1024       affected.
1025     * Support for a number of older systems and recently unmaintained or
1026       untested target ports of GCC has been declared obsolete in GCC 9.
1027       Unless there is activity to revive them, the next release of GCC
1028       will have their sources permanently removed.
1029       The following ports for individual systems on particular
1030       architectures have been obsoleted:
1031          + Solaris 10 (*-*-solaris2.10). Details can be found in the
1032            [4]announcement.
1033          + Cell Broadband Engine SPU (spu*-*-*). Details can be found in
1034            the [5]announcement.
1035     * A change to the C++ std::rotate algorithm in GCC 9.1.0 can cause
1036       ABI incompatibilities with object files compiled with other
1037       versions of GCC. If the std::rotate algorithm is called with an
1038       empty range then it might cause a divide-by-zero error (as a SIGFPE
1039       signal) and crash. The change has been reverted for GCC 9.2.0 and
1040       future releases. For more details see [6]Bug 90920. The problem can
1041       be avoided by recompiling any objects that might call std::rotate
1042       with an empty range, so that the GCC 9.1.0 definition of
1043       std::rotate is not used.
1044     * The automatic template instantiation at link time ([7]-frepo) has
1045       been deprecated and will be removed in a future release.
1046     * The --with-default-libstdcxx-abi=gcc4-compatible configure option
1047       is broken in the 9.1 and 9.2 releases, producing a shared library
1048       with missing symbols (see [8]Bug 90361). As a workaround, configure
1049       without that option and build GCC as normal, then edit the
1050       installed <bits/c++config.h> headers to define the
1051       _GLIBCXX_USE_CXX11_ABI macro to 0.
1052
1053General Improvements
1054
1055   The following GCC command line options have been introduced or
1056   improved.
1057     * All command line options that take a byte-size argument accept
1058       64-bit integers as well as standard SI and IEC suffixes such as kb
1059       and KiB, MB and MiB, or GB and GiB denoting the corresponding
1060       multiples of bytes. See [9]Invoking GCC for more.
1061     * A new option,
1062       [10]-flive-patching=[inline-only-static|inline-clone], has been
1063       introduced to provide a safe compilation for live-patching. At the
1064       same time, provides multiple-level control on the enabled IPA
1065       optimizations. See the user guide for more details about the
1066       option.
1067     * A new option, --completion, has been added to provide more fine
1068       option completion in a shell. It is intended to be used by
1069       Bash-completion.
1070     * GCC's diagnostics now print source code with a left margin showing
1071       line numbers, configurable with
1072       [11]-fno-diagnostics-show-line-numbers.
1073       GCC's diagnostics can also now label regions of the source code to
1074       show pertinent information, such as the types within an expression.
1075$ g++ t.cc
1076t.cc: In function 'int test(const shape&, const shape&)':
1077t.cc:15:4: error: no match for 'operator+' (operand types are 'boxed_value<doubl
1078e>' and 'boxed_value<double>')
1079   14 |   return (width(s1) * height(s1)
1080      |           ~~~~~~~~~~~~~~~~~~~~~~
1081      |                     |
1082      |                     boxed_value<[...]>
1083   15 |    + width(s2) * height(s2));
1084      |    ^ ~~~~~~~~~~~~~~~~~~~~~~
1085      |                |
1086      |                boxed_value<[...]>
1087
1088       These labels can be disabled via [12]-fno-diagnostics-show-labels.
1089     * A new option [13]-fdiagnostics-format=json has been introduced for
1090       emitting diagnostics in a machine-readable format.
1091     * The alignment-related options [14]-falign-functions,
1092       [15]-falign-labels, [16]-falign-loops, and [17]-falign-jumps
1093       received support for a secondary alignment (e.g.
1094       -falign-loops=n:m:n2:m2).
1095     * New pair of profiling options ([18]-fprofile-filter-files and
1096       [19]-fprofile-exclude-files) has been added. The options help to
1097       filter which source files are instrumented.
1098     * AddressSanitizer generates more compact redzones for automatic
1099       variables. That helps to reduce memory footprint of a sanitized
1100       binary.
1101     * Numerous improvements have been made to the output of
1102       [20]-fopt-info.
1103       Messages are now prefixed with optimized, missed, or note, rather
1104       than the old behavior of all being prefixed with note.
1105       The output from -fopt-info can now contain information on inlining
1106       decisions:
1107$ g++ -c inline.cc -O2 -fopt-info-inline-all
1108inline.cc:24:11: note: Considering inline candidate void foreach(T, T, void (*)(
1109E)) [with T = char**; E = char*]/2.
1110inline.cc:24:11: optimized:  Inlining void foreach(T, T, void (*)(E)) [with T =
1111char**; E = char*]/2 into int main(int, char**)/1.
1112inline.cc:19:12: missed:   not inlinable: void inline_me(char*)/0 -> int std::pu
1113ts(const char*)/3, function body not available
1114inline.cc:13:8: optimized:  Inlined void inline_me(char*)/4 into int main(int, c
1115har**)/1 which now has time 127.363637 and size 11, net change of +0.
1116Unit growth for small function inlining: 16->16 (0%)
1117
1118Inlined 2 calls, eliminated 1 functions
1119
1120
1121       The output from the vectorizer has been rationalized so that failed
1122       attempts to vectorize a loop are displayed in the form
1123    [LOOP-LOCATION]: couldn't vectorize this loop
1124    [PROBLEM-LOCATION]: because of [REASON]
1125
1126       rather than an exhaustive log of all decisions made by the
1127       vectorizer. For example:
1128$ gcc -c v.c -O3 -fopt-info-all-vec
1129v.c:7:3: missed: couldn't vectorize loop
1130v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" :  :  : "me
1131mory");
1132v.c:3:6: note: vectorized 0 loops in function.
1133v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" :  :  : "me
1134mory");
1135
1136       The old behavior can be obtained via a new -internals suboption of
1137       -fopt-info.
1138     * A new option, [21]-fsave-optimization-record has been added, which
1139       writes a SRCFILE.opt-record.json.gz file describing the
1140       optimization decisions made by GCC. This is similar to the output
1141       of -fopt-info, but with additional metadata such as the inlining
1142       chain, and profile information (if available).
1143     * Inter-procedural propagation of stack alignment can now be
1144       controlled by [22]-fipa-stack-alignment.
1145     * Propagation of addressability, readonly and writeonly flags on
1146       static variables can now be controlled by
1147       [23]-fipa-reference-addressable.
1148
1149   The following built-in functions have been introduced.
1150     * [24]__builtin_expect_with_probability to provide branch prediction
1151       probability hints to the optimizer.
1152     * [25]__builtin_has_attribute determines whether a function, type, or
1153       variable has been declared with some attribute.
1154     * [26]__builtin_speculation_safe_value can be used to help mitigate
1155       against unsafe speculative execution.
1156
1157   The following attributes have been introduced.
1158     * The [27]copy function attribute has been added. The attribute can
1159       also be applied to type definitions and to variable declarations.
1160
1161   A large number of improvements to code generation have been made,
1162   including but not limited to the following.
1163     * Switch expansion has been improved by using a different strategy
1164       (jump table, bit test, decision tree) for a subset of switch cases.
1165     * A linear function expression defined as a switch statement can be
1166       transformed by [28]-ftree-switch-conversion. For example:
1167
1168int
1169foo (int how)
1170{
1171  switch (how) {
1172    case 2: how = 205; break;
1173    case 3: how = 305; break;
1174    case 4: how = 405; break;
1175    case 5: how = 505; break;
1176    case 6: how = 605; break;
1177  }
1178  return how;
1179}
1180
1181       can be transformed into 100 * how + 5 (for values defined in the
1182       switch statement).
1183     * Inter-procedural optimization improvements:
1184          + Inliner defaults was tuned to better suits modern C++
1185            codebases especially when built with link time optimizations.
1186            New parameters max-inline-insns-small, max-inline-insns-size,
1187            uninlined-function-insns, uninlined-function-time,
1188            uninlined-thunk-insns, and uninlined-thunk-time were added.
1189          + Hot/cold partitioning is now more precise and aggressive.
1190          + Improved scalability for very large translation units
1191            (especially when link-time optimizing large programs).
1192     * Profile driven optimization improvements:
1193          + [29]-fprofile-use now enables [30]-fversion-loops-for-strides,
1194            [31]-floop-interchange, [32]-floop-unroll-and-jam,
1195            [33]-ftree-loop-distribution.
1196          + Streaming of counter histograms was removed. This reduces the
1197            size of profile files. Histogram is computed on the fly with
1198            link-time optimization. Parameter hot-bb-count-ws-permille was
1199            reduced from 999 to 990 to account for more precise
1200            histograms.
1201     * Link-time optimization improvements:
1202          + Types are now simplified prior streaming resulting in
1203            significant reductions of the LTO object files, link-time
1204            memory use, and improvements of link-time parallelism.
1205          + Default number of partitions (--param lto-partitions) was
1206            increased from 32 to 128 enabling effective use of CPUs with
1207            more than 32 hyperthreads. --param
1208            lto-max-streaming-parallelism can now be used to control
1209            number of streaming processes.
1210          + Warnings on C++ One Decl Rule violations (-Wodr) are now more
1211            informative and produce fewer redundant results.
1212       Overall compile time of Firefox 66 and LibreOffice 6.2.3 on an
1213       8-core machine was reduced by about 5% compared to GCC 8.3, and the
1214       size of LTO object files by 7%. LTO link-time improves by 11% on an
1215       8-core machine and scales significantly better for more parallel
1216       build environments. The serial stage of the link-time optimization
1217       is 28% faster consuming 20% less memory. The parallel stage now
1218       scales to up to 128 partitions rather than 32 and reduces memory
1219       use for every worker by 30%.
1220
1221   The following improvements to the gcov command-line utility have been
1222   made.
1223     * The gcov tool received a new option [34]--use-hotness-colors (-q)
1224       that can provide perf-like coloring of hot functions.
1225     * The gcov tool has changed its intermediate format to a new JSON
1226       format.
1227
1228New Languages and Language specific improvements
1229
1230   [35]OpenACC support in C, C++, and Fortran continues to be maintained
1231   and improved. Most of the OpenACC 2.5 specification is implemented. See
1232   the [36]implementation status section on the OpenACC wiki page for
1233   further information.
1234
1235  C family
1236
1237     * Version 5.0 of the [37]OpenMP specification is now partially
1238       supported in the C and C++ compilers. For details which features of
1239       OpenMP 5.0 are and which are not supported in the GCC 9 release see
1240       [38]this mail.
1241     * New extensions:
1242          + [39]__builtin_convertvector built-in for vector conversions
1243            has been added.
1244     * New warnings:
1245          + [40]-Waddress-of-packed-member, enabled by default, warns
1246            about an unaligned pointer value from the address of a packed
1247            member of a struct or union.
1248     * Enhancements to existing warnings:
1249          + [41]-Warray-bounds detects more instances of out-of-bounds
1250            indices.
1251          + [42]-Wattribute-alias also detects attribute mismatches
1252            between alias declarations and their targets, in addition to
1253            mismatches between their types.
1254          + [43]-Wformat-overflow and [44]-Wformat-truncation have been
1255            extended to all formatted input/output functions (where
1256            applicable) and enhanced to detect a subset of instances of
1257            reading past the end of unterminated constant character arrays
1258            in %s directives.
1259          + [45]-Wmissing-attributes detects instances of missing function
1260            attributes on declarations of aliases and weak references.
1261          + [46]-Wstringop-truncation also detects a subset of instances
1262            of reading past the end of unterminated constant character
1263            arrays,
1264     * If a macro is used with the wrong argument count, the C and C++
1265       front ends now show the definition of that macro via a note.
1266     * The spelling corrector now considers transposed letters, and the
1267       threshold for similarity has been tightened, to avoid nonsensical
1268       suggestions.
1269
1270  C
1271
1272     * There is now experimental support for -std=c2x, to select support
1273       for the upcoming C2X revision of the ISO C standard. This standard
1274       is in the early stages of development and the only feature
1275       supported in GCC 9 is _Static_assert with a single argument
1276       (support for _Static_assert with two arguments was added in C11 and
1277       GCC 4.6). There are also new options -std=gnu2x, for C2X with GNU
1278       extensions, and -Wc11-c2x-compat, to warn for uses of features
1279       added in C2X (such warnings are also enabled by use of -Wpedantic
1280       if not using -std=c2x or -std=gnu2x).
1281     * New warnings:
1282          + [47]-Wabsolute-value warns for calls to standard functions
1283            that compute the absolute value of an argument when a more
1284            appropriate standard function is available. For example,
1285            calling abs(3.14) triggers the warning because the appropriate
1286            function to call to compute the absolute value of a double
1287            argument is fabs. The option also triggers warnings when the
1288            argument in a call to such a function has an unsigned type.
1289            This warning can be suppressed with an explicit type cast and
1290            it is also enabled by -Wextra.
1291
1292  C++
1293
1294     * New warnings:
1295          + [48]-Wdeprecated-copy, implied by -Wextra, warns about the
1296            C++11 deprecation of implicitly declared copy constructor and
1297            assignment operator if one of them is user-provided.
1298            -Wdeprecated-copy-dtor also warns if the destructor is
1299            user-provided, as specified in C++11.
1300          + [49]-Winit-list-lifetime, on by default, warns about uses of
1301            std::initializer_list that are likely to result in a dangling
1302            pointer, such as returning or assigning from a temporary list.
1303          + [50]-Wredundant-move, implied by -Wextra, warns about
1304            redundant calls to std::move.
1305          + [51]-Wpessimizing-move, implied by -Wall, warns when a call to
1306            std::move prevents copy elision.
1307          + [52]-Wclass-conversion, on by default, warns when a conversion
1308            function will never be called due to the type it converts to.
1309     * The C++ front end has experimental support for some of the upcoming
1310       C++2a draft features with the -std=c++2a or -std=gnu++2a flags,
1311       including range-based for statements with initializer, default
1312       constructible and assignable stateless lambdas, lambdas in
1313       unevaluated contexts, language support for empty data members,
1314       allowing pack expansion in lambda init-capture, likely and unlikely
1315       attributes, class types in non-type template parameters, allowing
1316       virtual function calls in constant expressions, explicit(bool),
1317       std::is_constant_evaluated, nested inline namespaces, etc. For a
1318       full list of new features, see [53]the C++ status page.
1319     * The C++ front end now preserves source locations for literals,
1320       id-expression, and mem-initializer for longer. For example it is
1321       now able to pin-point the pertinent locations for bad
1322       initializations such as these
1323$ g++ -c bad-inits.cc
1324bad-inits.cc:10:14: error: cannot convert 'json' to 'int' in initialization
1325   10 |   { 3, json::object },
1326      |        ~~~~~~^~~~~~
1327      |              |
1328      |              json
1329bad-inits.cc:14:31: error: initializer-string for array of chars is too long [-f
1330permissive]
1331   14 | char buffers[3][5] = { "red", "green", "blue" };
1332      |                               ^~~~~~~
1333bad-inits.cc: In constructor 'X::X()':
1334bad-inits.cc:17:13: error: invalid conversion from 'int' to 'void*' [-fpermissiv
1335e]
1336   17 |   X() : one(42), two(42), three(42)
1337      |             ^~
1338      |             |
1339      |             int
1340
1341       rather than emitting the error at the final closing parenthesis or
1342       brace.
1343     * Error-reporting of overload resolution has been special-cased to
1344       make the case of a single failed candidate easier to read. For
1345       example:
1346$ g++ param-type-mismatch.cc
1347param-type-mismatch.cc: In function 'int test(int, const char*, float)':
1348param-type-mismatch.cc:8:32: error: cannot convert 'const char*' to 'const char*
1349*'
1350    8 |   return foo::member_1 (first, second, third);
1351      |                                ^~~~~~
1352      |                                |
1353      |                                const char*
1354param-type-mismatch.cc:3:46: note:   initializing argument 2 of 'static int foo:
1355:member_1(int, const char**, float)'
1356    3 |   static int member_1 (int one, const char **two, float three);
1357      |                                 ~~~~~~~~~~~~~^~~
1358
1359       highlights both the problematic argument, and the parameter that it
1360       can't be converted to.
1361     * Diagnostics involving binary operators now use color to distinguish
1362       the two operands, and label them separately (as per the example of
1363       source labelling above).
1364     * Diagnostics involving function calls now highlight the pertinent
1365       parameter of the declaration in more places.
1366$ g++ bad-conversion.cc
1367bad-conversion.cc: In function 'void caller()':
1368bad-conversion.cc:9:14: error: cannot convert 'bool' to 'void*'
1369    9 |   callee (0, false, 2);
1370      |              ^~~~~
1371      |              |
1372      |              bool
1373bad-conversion.cc:3:19: note:   initializing argument 2 of 'void callee(int, voi
1374d*, int)'
1375    3 | void callee (int, void *, int)
1376      |                   ^~~~~~
1377
1378     * The C++ front end's implementation of [54]-Wformat now shows
1379       precise locations within string literals, and underlines the
1380       pertinent arguments at bogus call sites (the C front end has been
1381       doing this since GCC 7). For example:
1382$ g++ -c bad-printf.cc -Wall
1383bad-printf.cc: In function 'void print_field(const char*, float, long int, long
1384int)':
1385bad-printf.cc:6:17: warning: field width specifier '*' expects argument of type
1386'int', but argument 3 has type 'long int' [-Wformat=]
1387    6 |   printf ("%s: %*ld ", fieldname, column - width, value);
1388      |                ~^~~               ~~~~~~~~~~~~~~
1389      |                 |                        |
1390      |                 int                      long int
1391bad-printf.cc:6:19: warning: format '%ld' expects argument of type 'long int', b
1392ut argument 4 has type 'double' [-Wformat=]
1393    6 |   printf ("%s: %*ld ", fieldname, column - width, value);
1394      |                ~~~^                               ~~~~~
1395      |                   |                               |
1396      |                   long int                        double
1397      |                %*f
1398
1399     * The C++ front end has gained new fix-it hints for forgetting the
1400       return *this; needed by various C++ operators:
1401$ g++ -c operator.cc
1402operator.cc: In member function 'boxed_ptr& boxed_ptr::operator=(const boxed_ptr
1403&)':
1404operator.cc:7:3: warning: no return statement in function returning non-void [-W
1405return-type]
1406    6 |     m_ptr = other.m_ptr;
1407  +++ |+    return *this;
1408    7 |   }
1409      |   ^
1410
1411       for when the compiler needs a typename:
1412$ g++ -c template.cc
1413template.cc:3:3: error: need 'typename' before 'Traits::type' because 'Traits' i
1414s a dependent scope
1415    3 |   Traits::type type;
1416      |   ^~~~~~
1417      |   typename
1418
1419       when trying to use an accessor member as if it were a data member:
1420$ g++ -c fncall.cc
1421fncall.cc: In function 'void hangman(const mystring&)':
1422fncall.cc:12:11: error: invalid use of member function 'int mystring::get_length
1423() const' (did you forget the '()' ?)
1424   12 |   if (str.get_length > 0)
1425      |       ~~~~^~~~~~~~~~
1426      |                     ()
1427
1428       for C++11's scoped enums:
1429$ g++ -c enums.cc
1430enums.cc: In function 'void json::test(const json::value&)':
1431enums.cc:12:26: error: 'STRING' was not declared in this scope; did you mean 'js
1432on::kind::STRING'?
1433   12 |     if (v.get_kind () == STRING)
1434      |                          ^~~~~~
1435      |                          json::kind::STRING
1436enums.cc:3:44: note: 'json::kind::STRING' declared here
1437    3 |   enum class kind { OBJECT, ARRAY, NUMBER, STRING, TRUE, FALSE, NULL_ };
1438      |                                            ^~~~~~
1439
1440       and a tweak to integrate the suggestions about misspelled members
1441       with that for accessors:
1442$ g++ -c accessor-fixit.cc
1443accessor-fixit.cc: In function 'int test(t*)':
1444accessor-fixit.cc:17:15: error: 'class t' has no member named 'ratio'; did you m
1445ean 'int t::m_ratio'? (accessible via 'int t::get_ratio() const')
1446   17 |   return ptr->ratio;
1447      |               ^~~~~
1448      |               get_ratio()
1449
1450       In addition, various diagnostics in the C++ front-end have been
1451       streamlined by consolidating the suggestion into the initial error,
1452       rather than emitting a follow-up note:
1453$ g++ typo.cc
1454typo.cc:5:13: error: 'BUFSIZE' was not declared in this scope; did you mean 'BUF
1455_SIZE'?
1456    5 | uint8_t buf[BUFSIZE];
1457      |             ^~~~~~~
1458      |             BUF_SIZE
1459
1460    Runtime Library (libstdc++)
1461
1462     * Improved support for C++17, including:
1463          + The C++17 implementation is no longer experimental.
1464          + Parallel algorithms and <execution> (requires [55]Thread
1465            Building Blocks 2018 or newer).
1466          + <memory_resource>.
1467          + Using the types and functions in <filesystem> does not require
1468            linking with -lstdc++fs now.
1469     * Improved experimental support for C++2a, including:
1470          + Type traits std::remove_cvref, std::unwrap_reference,
1471            std::unwrap_decay_ref, std::is_nothrow_convertible, and
1472            std::type_identity.
1473          + Headers <bit> and <version>.
1474          + Uniform container erasure (std::erase_if).
1475          + contains member of maps and sets.
1476          + String prefix and suffix checking (starts_with, ends_with).
1477          + Functions std::midpoint and std::lerp for interpolation.
1478          + std::bind_front.
1479          + std::visit<R>.
1480          + std::assume_aligned.
1481          + Uses-allocator construction utilities.
1482          + std::pmr::polymorphic_allocator<std::byte>.
1483          + Library support for char8_t type.
1484          + Destroying delete.
1485          + std::is_constant_evaluated() function.
1486     * Support for opening file streams with wide character paths on
1487       Windows
1488     * Incomplete support for the C++17 Filesystem library and the
1489       Filesystem TS on Windows.
1490     * Incomplete, experimental support for the Networking TS.
1491
1492  D
1493
1494     * Support for the D programming language has been added to GCC,
1495       implementing version 2.076 of the language and run-time library.
1496
1497  Fortran
1498
1499     * Asynchronous I/O is now fully supported. The program needs to be
1500       linked against the pthreads library to use it, otherwise the I/O is
1501       done synchronously. For systems which do not support POSIX
1502       condition variables, such as AIX, all I/O is still done
1503       synchronously.
1504     * The BACK argument for MINLOC and MAXLOC has been implemented.
1505     * The FINDLOC intrinsic function has been implemented.
1506     * The IS_CONTIGUOUS intrinsic function has been implemented.
1507     * Direct access to the real and imaginary parts of a complex variable
1508       via c%re and c%im has been implemented.
1509     * Type parameter inquiry via str%len and a%kind has been implemented.
1510     * C descriptors and the ISO_Fortran_binding.h source file have been
1511       implemented.
1512     * The MAX and MIN intrinsics are no longer guaranteed to return any
1513       particular value in case one of the arguments is a NaN. Note that
1514       this conforms to the Fortran standard and to what other Fortran
1515       compilers do. If there is a need to handle that case in some
1516       specific way, one needs to explicitly check for NaN's before
1517       calling MAX or MIN, e.g. by using the IEEE_IS_NAN function from the
1518       intrinsic module IEEE_ARITHMETIC.
1519     * A new command-line option [56]-fdec-include, set also by the
1520       [57]-fdec option, has been added to increase compatibility with
1521       legacy code. With this option, an INCLUDE directive is also parsed
1522       as a statement, which allows the directive to be spread across
1523       multiple source lines with line continuations.
1524     * A new [58]BUILTIN directive, has been added. The purpose of the
1525       directive is to provide an API between the GCC compiler and the GNU
1526       C Library which would define vector implementations of math
1527       routines.
1528
1529  Go
1530
1531     * GCC 9 provides a complete implementation of the Go 1.12.2 user
1532       packages.
1533
1534libgccjit
1535
1536     * The libgccjit API gained a new entry point:
1537       [59]gcc_jit_context_add_driver_option.
1538
1539New Targets and Target Specific Improvements
1540
1541  AArch64 & Arm
1542
1543     * Support has been added for the following processors (GCC
1544       identifiers in parentheses):
1545          + Arm Cortex-A76 (cortex-a76).
1546          + Arm Cortex-A55/Cortex-A76 DynamIQ big.LITTLE
1547            (cortex-a76.cortex-a55).
1548          + Arm Neoverse N1 (neoverse-n1).
1549       The GCC identifiers can be used as arguments to the -mcpu or -mtune
1550       options, for example: -mcpu=cortex-a76 or
1551       -mtune=cortex-a76.cortex-a55 or as arguments to the equivalent
1552       target attributes and pragmas.
1553     * The Armv8.3-A complex number instructions are now supported via
1554       intrinsics when the option -march=armv8.3-a or equivalent is
1555       specified. For the half-precision floating-point variants of these
1556       instructions use the architecture extension flag +fp16, e.g.
1557       -march=armv8.3-a+fp16.
1558       The intrinsics are defined by the ACLE specification.
1559     * The Armv8.5-A architecture is now supported through the
1560       -march=armv8.5-a option.
1561     * The Armv8.5-A architecture also adds some security features that
1562       are optional to all older architecture versions. These are now
1563       supported and only affect the assembler.
1564          + Speculation Barrier instruction through the -march=armv8-a+sb
1565            option.
1566          + Execution and Data Prediction Restriction instructions through
1567            the -march=armv8-a+predres option.
1568          + Speculative Store Bypass Safe instruction through the
1569            -march=armv8-a+ssbs option. This does not require a compiler
1570            option for Arm and thus -march=armv8-a+ssbs is an
1571            AArch64-specific option.
1572
1573      AArch64 specific
1574
1575     * Support has been added for the Arm Neoverse E1 processor
1576       (-mcpu=neoverse-e1).
1577     * The AArch64 port now has support for stack clash protection using
1578       the [60]-fstack-clash-protection option. The probing interval/guard
1579       size can be set by using --param
1580       stack-clash-protection-guard-size=12|16. The value of this
1581       parameter must be in bytes represented as a power of two. The two
1582       supported values for this parameter are 12 (for a 4KiB size, 2^12)
1583       and 16 (for a 64KiB size, 2^16). The default value is 16 (64Kb) and
1584       can be changed at configure time using the flag
1585       --with-stack-clash-protection-guard-size=12|16.
1586     * The option -msign-return-address= has been deprecated. This has
1587       been replaced by the new -mbranch-protection= option. This new
1588       option can now be used to enable the return address signing as well
1589       as the new Branch Target Identification feature of Armv8.5-A
1590       architecture. For more information on the arguments accepted by
1591       this option, please refer to [61]AArch64-Options.
1592     * The following optional extensions to Armv8.5-A architecture are now
1593       supported and only affect the assembler.
1594          + Random Number Generation instructions through the
1595            -march=armv8.5-a+rng option.
1596          + Memory Tagging Extension through the -march=armv8.5-a+memtag
1597            option.
1598
1599      Arm specific
1600
1601     * Support for the deprecated Armv2 and Armv3 architectures and their
1602       variants has been removed. Their corresponding -march values and
1603       the -mcpu options that used these architectures have been removed.
1604     * Support for the Armv5 and Armv5E architectures (which have no known
1605       implementations) has been removed. Note that Armv5T, Armv5TE and
1606       Armv5TEJ architectures remain supported.
1607     * Corrected FPU configurations for Cortex-R7 and Cortex-R8 when using
1608       their respective -mcpu options.
1609
1610  AMD GCN
1611
1612     * A new back end targeting AMD GCN GPUs has been contributed to GCC.
1613       The implementation is currently limited to compiling
1614       single-threaded, stand-alone programs. Future versions will add
1615       support for offloading multi-threaded kernels via OpenMP and
1616       OpenACC. The following devices are supported (GCC identifiers in
1617       parentheses):
1618          + Fiji (fiji).
1619          + Vega 10 (gfx900).
1620
1621  ARC
1622
1623     * LRA is now on by default for the ARC target. This can be controlled
1624       by -mlra.
1625     * Add support for frame code-density and branch-and-index
1626       instructions.
1627
1628  C-SKY
1629
1630     * A new back end targeting C-SKY V2 processors has been contributed
1631       to GCC.
1632
1633  IA-32/x86-64
1634
1635     * Support of Intel MPX (Memory Protection Extensions) has been
1636       removed.
1637     * New ISA extension support for Intel PTWRITE was added to GCC.
1638       PTWRITE intrinsics are available via the -mptwrite compiler switch.
1639     * GCC now supports the Intel CPU named Cascade Lake with AVX512
1640       extensions through -march=cascadelake. The switch enables the
1641       following ISA extensions: AVX512F, AVX512VL, AVX512CD, AVX512BW,
1642       AVX512DQ, AVX512VNNI.
1643
1644  OpenRISC
1645
1646     * A new back end targeting OpenRISC processors has been contributed
1647       to GCC.
1648
1649  S/390, System z, IBM z Systems
1650
1651     * Support for the arch13 architecture has been added. When using the
1652       -march=arch13 option, the compiler will generate code making use of
1653       the new instructions introduced with the vector enhancement
1654       facility 2 and the miscellaneous instruction extension facility 2.
1655       The -mtune=arch13 option enables arch13 specific instruction
1656       scheduling without making use of new instructions.
1657     * Builtins for the new vector instructions have been added and can be
1658       enabled using the -mzvector option.
1659     * Support for ESA architecture machines g5 and g6 is deprecated since
1660       GCC 6.1.0 and has been removed now.
1661     * When compiling with -march=z14 or higher GCC emits alignments hints
1662       on the vector load/store instructions (8 or 16 byte).
1663     * Functions now have a default alignment of 16 bytes. This helps with
1664       branch prediction effects.
1665     * -mfentry is now supported. As well as the mcount mechanism the
1666       __fentry__ is called before the function prologue. However, since
1667       just a single instruction is required to call __fentry__ the call
1668       sequence imposes a smaller overhead than mcount (4 instructions).
1669       The produced code is compatible only with newer glibc versions,
1670       which provide the __fentry__ symbol and do not clobber r0 when
1671       resolving lazily bound functions. -mfentry is only supported when
1672       generating 64 bit code and does not work with nested C functions.
1673     * The -mnop-mcount option can be used to emit NOP instructions
1674       instead of an mcount or fentry call stub.
1675     * With the -mrecord-mcount option a __mcount_loc section is generated
1676       containing pointers to each profiling call stub. This is useful for
1677       automatically patching in and out calls.
1678
1679Operating Systems
1680
1681  Solaris
1682
1683     * g++ now unconditionally enables large file support when compiling
1684       32-bit code.
1685     * Support for the AddressSanitizer and UndefinedBehaviorSanitizer has
1686       been merged from LLVM. For the moment, this only works for 32-bit
1687       code on both SPARC and x86.
1688     * An initial port of the D runtime library has been completed on
1689       Solaris 11/x86. It requires the use of GNU as. Solaris 11/SPARC
1690       support is still work-in-progress.
1691
1692  Windows
1693
1694     * A C++ Microsoft ABI bitfield layout bug, [62]PR87137 has been
1695       fixed. A non-field declaration could cause the current bitfield
1696       allocation unit to be completed, incorrectly placing a following
1697       bitfield into a new allocation unit. The Microsoft ABI is selected
1698       for:
1699          + Mingw targets
1700          + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields
1701            option is specified, or __attribute__((ms_struct)) is used
1702          + SuperH targets when the -mhitachi option is specified, or
1703            __attribute__((renesas)) is used
1704
1705Improvements for plugin authors
1706
1707     * GCC's diagnostic subsystem now has a way to logically group
1708       together related diagnostics, auto_diagnostic_group. Such
1709       diagnostics will be nested by the output of
1710       [63]-fdiagnostics-format=json.
1711     * GCC now has a set of [64]user experience guidelines for GCC, with
1712       information and advice on implementing new diagnostics.
1713
1714Other significant improvements
1715
1716     * GCC's internal "selftest" suite now runs for C++ as well as C (in
1717       debug builds of the compiler).
1718
1719GCC 9.1
1720
1721   This is the [65]list of problem reports (PRs) from GCC's bug tracking
1722   system that are known to be fixed in the 9.1 release. This list might
1723   not be complete (that is, it is possible that some PRs that have been
1724   fixed are not listed here).
1725
1726GCC 9.2
1727
1728   This is the [66]list of problem reports (PRs) from GCC's bug tracking
1729   system that are known to be fixed in the 9.2 release. This list might
1730   not be complete (that is, it is possible that some PRs that have been
1731   fixed are not listed here).
1732
1733GCC 9.3
1734
1735   This is the [67]list of problem reports (PRs) from GCC's bug tracking
1736   system that are known to be fixed in the 9.3 release. This list might
1737   not be complete (that is, it is possible that some PRs that have been
1738   fixed are not listed here).
1739
1740GCC 9.4
1741
1742   This is the [68]list of problem reports (PRs) from GCC's bug tracking
1743   system that are known to be fixed in the 9.4 release. This list might
1744   not be complete (that is, it is possible that some PRs that have been
1745   fixed are not listed here).
1746
1747  Target Specific Changes
1748
1749    AArch64
1750
1751     * The option -moutline-atomics has been added to aid deployment of
1752       the Large System Extensions (LSE) on GNU/Linux systems built with a
1753       baseline architecture targeting Armv8-A. When the option is
1754       specified code is emitted to detect the presence of LSE
1755       instructions at run time and use them for standard atomic
1756       operations. For more information please refer to the documentation.
1757     * GCC now supports the Fujitsu A64FX. The associated -mcpu and -mtune
1758       options are -mcpu=a64fx and -mtune=a64fx respectively. In
1759       particular, -mcpu=a64fx generates code for Armv8.2-A with SVE and
1760       tunes the code for the A64FX. This includes tuning the SVE code,
1761       although by default the code is still length-agnostic and so works
1762       for all SVE implementations. Adding -msve-vector-bits=512 makes the
1763       code specific to 512-bit SVE.
1764
1765GCC 9.5
1766
1767   This is the [69]list of problem reports (PRs) from GCC's bug tracking
1768   system that are known to be fixed in the 9.5 release. This list might
1769   not be complete (that is, it is possible that some PRs that have been
1770   fixed are not listed here).
1771
1772
1773    For questions related to the use of GCC, please consult these web
1774    pages and the [70]GCC manuals. If that fails, the
1775    [71]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1776    web pages and the development of GCC are welcome on our developer
1777    list at [72]gcc@gcc.gnu.org. All of [73]our lists have public
1778    archives.
1779
1780   Copyright (C) [74]Free Software Foundation, Inc. Verbatim copying and
1781   distribution of this entire article is permitted in any medium,
1782   provided this notice is preserved.
1783
1784   These pages are [75]maintained by the GCC team. Last modified
1785   2022-05-27[76].
1786
1787References
1788
1789   1. http://gcc.gnu.org/gcc-9/porting_to.html
1790   2. http://gcc.gnu.org/onlinedocs/index.html#current
1791   3. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88469
1792   4. https://gcc.gnu.org/ml/gcc/2018-10/msg00139.html
1793   5. https://gcc.gnu.org/ml/gcc/2019-04/msg00023.html
1794   6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90920
1795   7. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-frepo
1796   8. https://gcc.gnu.org/PR90361
1797   9. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-GCC.html#Invoking-GCC
1798  10. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-flive-patching
1799  11. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-line-numbers
1800  12. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-labels
1801  13. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format
1802  14. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-functions
1803  15. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-labels
1804  16. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-loops
1805  17. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-jumps
1806  18. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-filter-files
1807  19. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-exclude-files
1808  20. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fopt-info
1809  21. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fsave-optimization-record
1810  22. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-stack-alignment
1811  23. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-reference-addressable
1812  24. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fexpect_005fwith_005fprobability
1813  25. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fhas_005fattribute-1
1814  26. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fspeculation_005fsafe_005fvalue-1
1815  27. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Common-Function-Attributes.html#index-copy-function-attribute
1816  28. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-switch-conversion
1817  29. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fprofile-use
1818  30. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fversion-loops-for-strides
1819  31. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-interchange
1820  32. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-unroll-and-jam
1821  33. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-loop-distribution
1822  34. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-Gcov.html#Invoking-Gcov
1823  35. https://www.openacc.org/
1824  36. https://gcc.gnu.org/wiki/OpenACC/Implementation Status#status-9
1825  37. https://www.openmp.org/specifications/
1826  38. https://gcc.gnu.org/ml/gcc-patches/2018-11/msg00628.html
1827  39. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Vector-Extensions.html#index-_005f_005fbuiltin_005fconvertvector
1828  40. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Waddress-of-packed-member
1829  41. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Warray-bounds
1830  42. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wattribute-alias
1831  43. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-overflow
1832  44. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-truncation
1833  45. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wmissing-attributes
1834  46. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wstringop-truncation
1835  47. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wabsolute-value
1836  48. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-copy
1837  49. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Winit-list-lifetime
1838  50. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-move
1839  51. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wpessimizing-move
1840  52. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-conversion
1841  53. http://gcc.gnu.org/projects/cxx-status.html#cxx2a
1842  54. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat
1843  55. https://github.com/oneapi-src/oneTBB
1844  56. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec-include
1845  57. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec
1846  58. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/BUILTIN-directive.html#BUILTIN-directive
1847  59. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/jit/topics/contexts.html#gcc_jit_context_add_driver_option
1848  60. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fstack-protector
1849  61. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/AArch64-Options.html#AArch64-Options
1850  62. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137
1851  63. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format
1852  64. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gccint/User-Experience-Guidelines.html
1853  65. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.0
1854  66. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.2
1855  67. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.3
1856  68. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.4
1857  69. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.5
1858  70. https://gcc.gnu.org/onlinedocs/
1859  71. mailto:gcc-help@gcc.gnu.org
1860  72. mailto:gcc@gcc.gnu.org
1861  73. https://gcc.gnu.org/lists.html
1862  74. https://www.fsf.org/
1863  75. https://gcc.gnu.org/about.html
1864  76. http://validator.w3.org/check/referer
1865======================================================================
1866http://gcc.gnu.org/gcc-8/index.html
1867                              GCC 8 Release Series
1868
1869   (This release series is no longer supported.)
1870
1871   May 14, 2021
1872
1873   The [1]GNU project and the GCC developers are pleased to announce the
1874   release of GCC 8.5.
1875
1876   This release is a bug-fix release, containing fixes for regressions in
1877   GCC 8.4 relative to previous releases of GCC.
1878
1879Release History
1880
1881   GCC 8.5
1882          May 14, 2021 ([2]changes, [3]documentation)
1883
1884   GCC 8.4
1885          Mar 4, 2020 ([4]changes, [5]documentation)
1886
1887   GCC 8.3
1888          Feb 22, 2019 ([6]changes, [7]documentation)
1889
1890   GCC 8.2
1891          Jul 26, 2018 ([8]changes, [9]documentation)
1892
1893   GCC 8.1
1894          May 2, 2018 ([10]changes, [11]documentation)
1895
1896References and Acknowledgements
1897
1898   GCC used to stand for the GNU C Compiler, but since the compiler
1899   supports several other languages aside from C, it now stands for the
1900   GNU Compiler Collection.
1901
1902   A list of [12]successful builds is updated as new information becomes
1903   available.
1904
1905   The GCC developers would like to thank the numerous people that have
1906   contributed new features, improvements, bug fixes, and other changes as
1907   well as test results to GCC. This [13]amazing group of volunteers is
1908   what makes GCC successful.
1909
1910   For additional information about GCC please refer to the [14]GCC
1911   project web site or contact the [15]GCC development mailing list.
1912
1913   To obtain GCC please use [16]our mirror sites or [17]our version
1914   control system.
1915
1916
1917    For questions related to the use of GCC, please consult these web
1918    pages and the [18]GCC manuals. If that fails, the
1919    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1920    web pages and the development of GCC are welcome on our developer
1921    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
1922    archives.
1923
1924   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
1925   distribution of this entire article is permitted in any medium,
1926   provided this notice is preserved.
1927
1928   These pages are [23]maintained by the GCC team. Last modified
1929   2022-05-06[24].
1930
1931References
1932
1933   1. http://www.gnu.org/
1934   2. http://gcc.gnu.org/gcc-8/changes.html
1935   3. http://gcc.gnu.org/onlinedocs/8.5.0/
1936   4. http://gcc.gnu.org/gcc-8/changes.html
1937   5. http://gcc.gnu.org/onlinedocs/8.4.0/
1938   6. http://gcc.gnu.org/gcc-8/changes.html
1939   7. http://gcc.gnu.org/onlinedocs/8.3.0/
1940   8. http://gcc.gnu.org/gcc-8/changes.html
1941   9. http://gcc.gnu.org/onlinedocs/8.2.0/
1942  10. http://gcc.gnu.org/gcc-8/changes.html
1943  11. http://gcc.gnu.org/onlinedocs/8.1.0/
1944  12. http://gcc.gnu.org/gcc-8/buildstat.html
1945  13. http://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Contributors.html
1946  14. http://gcc.gnu.org/index.html
1947  15. mailto:gcc@gcc.gnu.org
1948  16. http://gcc.gnu.org/mirrors.html
1949  17. http://gcc.gnu.org/git.html
1950  18. https://gcc.gnu.org/onlinedocs/
1951  19. mailto:gcc-help@gcc.gnu.org
1952  20. mailto:gcc@gcc.gnu.org
1953  21. https://gcc.gnu.org/lists.html
1954  22. https://www.fsf.org/
1955  23. https://gcc.gnu.org/about.html
1956  24. http://validator.w3.org/check/referer
1957======================================================================
1958http://gcc.gnu.org/gcc-8/changes.html
1959                              GCC 8 Release Series
1960                        Changes, New Features, and Fixes
1961
1962   This page is a "brief" summary of some of the huge number of
1963   improvements in GCC 8. You may also want to check out our [1]Porting to
1964   GCC 8 page and the [2]full GCC documentation.
1965
1966Caveats
1967
1968     * Support for the obsolete SDB/coff debug info format has been
1969       removed. The option -gcoff no longer does anything.
1970     * The Cilk+ extensions to the C and C++ languages have been removed.
1971     * The MPX extensions to the C and C++ languages have been deprecated
1972       and will be removed in a future release.
1973     * The extension allowing arithmetic on std::atomic<void*> and types
1974       like std::atomic<R(*)()> has been deprecated.
1975     * The non-standard C++0x std::copy_exception function was removed.
1976       std::make_exception_ptr should be used instead.
1977     * Support for the powerpc*-*-*spe* target ports which have been
1978       recently unmaintained and untested in GCC has been declared
1979       obsolete in GCC 8 as [3]announced. Unless there is activity to
1980       revive them, the next release of GCC will have their sources
1981       permanently removed.
1982
1983General Improvements
1984
1985     * Inter-procedural optimization improvements:
1986          + Reworked run-time estimation metrics leading to more realistic
1987            guesses driving inliner and cloning heuristics.
1988          + The ipa-pure-const pass is extended to propagate the malloc
1989            attribute, and the corresponding warning option
1990            -Wsuggest-attribute=malloc emits a diagnostic for functions
1991            which can be annotated with the malloc attribute.
1992     * Profile driven optimization improvements:
1993          + New infrastructure for representing profiles (both statically
1994            guessed and profile feedback) which allows propagation of
1995            additional information about the reliability of the profile.
1996          + A number of improvements in the profile updating code solving
1997            problems found by new verification code.
1998          + Static detection of code which is not executed in a valid run
1999            of the program. This includes paths which trigger undefined
2000            behavior as well as calls to functions declared with the cold
2001            attribute. Newly the noreturn attribute does not imply all
2002            effects of cold to differentiate between exit (which is
2003            noreturn) and abort (which is in addition not executed in
2004            valid runs).
2005          + -freorder-blocks-and-partition, a pass splitting function
2006            bodies into hot and cold regions, is now enabled by default at
2007            -O2 and higher for x86 and x86-64.
2008     * Link-time optimization improvements:
2009          + We have significantly improved debug information on ELF
2010            targets using DWARF by properly preserving language-specific
2011            information. This allows for example the libstdc++
2012            pretty-printers to work with LTO optimized executables.
2013     * A new option -fcf-protection=[full|branch|return|none] is
2014       introduced to perform code instrumentation to increase program
2015       security by checking that target addresses of control-flow transfer
2016       instructions (such as indirect function call, function return,
2017       indirect jump) are valid. Currently the instrumentation is
2018       supported on x86 GNU/Linux targets only. See the user guide for
2019       further information about the option syntax and section "New
2020       Targets and Target Specific Improvements" for IA-32/x86-64 for more
2021       details.
2022     * The -gcolumn-info option is now enabled by default. It includes
2023       column information in addition to just filenames and line numbers
2024       in DWARF debugging information.
2025     * The polyhedral-based loop nest optimization pass
2026       -floop-nest-optimize has been overhauled. It's still considered
2027       experimental and may not result in any runtime improvements.
2028     * Two new classical loop nest optimization passes have been added.
2029       -floop-unroll-and-jam performs outer loop unrolling and fusing of
2030       the inner loop copies. -floop-interchange exchanges loops in a loop
2031       nest to improve data locality. Both passes are enabled by default
2032       at -O3 and above.
2033     * The classic loop nest optimization pass -ftree-loop-distribution
2034       has been improved and enabled by default at -O3 and above. It
2035       supports loop nest distribution in some restricted scenarios; it
2036       also supports cancellable innermost loop distribution with loop
2037       versioning under run-time alias checks.
2038     * The new option -fstack-clash-protection causes the compiler to
2039       insert probes whenever stack space is allocated statically or
2040       dynamically to reliably detect stack overflows and thus mitigate
2041       the attack vector that relies on jumping over a stack guard page as
2042       provided by the operating system.
2043     * A new pragma GCC unroll has been implemented in the C family of
2044       languages, as well as Fortran and Ada, so as to make it possible
2045       for the user to have a finer-grained control over the loop
2046       unrolling optimization.
2047     * GCC has been enhanced to detect more instances of meaningless or
2048       mutually exclusive attribute specifications and handle such
2049       conflicts more consistently. Mutually exclusive attribute
2050       specifications are ignored with a warning regardless of whether
2051       they appear on the same declaration or on distinct declarations of
2052       the same entity. For example, because the noreturn attribute on the
2053       second declaration below is mutually exclusive with the malloc
2054       attribute on the first, it is ignored and a warning is issued.
2055>
2056      void* __attribute__ ((malloc)) f (unsigned);
2057      void* __attribute__ ((noreturn)) f (unsigned);
2058
2059      warning: ignoring attribute 'noreturn' because it conflicts with attribute
2060 'malloc' [-Wattributes]
2061     * The gcov tool can distinguish functions that begin on a same line
2062       in a source file. This can be a different template instantiation or
2063       a class constructor:
2064
2065File 'ins.C'
2066Lines executed:100.00% of 8
2067Creating 'ins.C.gcov'
2068
2069        -:    0:Source:ins.C
2070        -:    0:Graph:ins.gcno
2071        -:    0:Data:ins.gcda
2072        -:    0:Runs:1
2073        -:    0:Programs:1
2074        -:    1:template<class T>
2075        -:    2:class Foo
2076        -:    3:{
2077        -:    4: public:
2078        2:    5:   Foo(): b (1000) {}
2079------------------
2080Foo<char>::Foo():
2081        1:    5:   Foo(): b (1000) {}
2082------------------
2083Foo<int>::Foo():
2084        1:    5:   Foo(): b (1000) {}
2085------------------
2086        2:    6:   void inc () { b++; }
2087------------------
2088Foo<char>::inc():
2089        1:    6:   void inc () { b++; }
2090------------------
2091Foo<int>::inc():
2092        1:    6:   void inc () { b++; }
2093------------------
2094        -:    7:
2095        -:    8:  private:
2096        -:    9:   int b;
2097        -:   10:};
2098        -:   11:
2099        1:   12:int main(int argc, char **argv)
2100        -:   13:{
2101        1:   14:  Foo<int> a;
2102        1:   15:  Foo<char> b;
2103        -:   16:
2104        1:   17:  a.inc ();
2105        1:   18:  b.inc ();
2106        1:   19:}
2107
2108     * The gcov tool has more accurate numbers for execution of lines in a
2109       source file.
2110     * The gcov tool can use TERM colors to provide more readable output.
2111     * AddressSanitizer gained a new pair of sanitization options,
2112       -fsanitize=pointer-compare and -fsanitize=pointer-subtract, which
2113       warn about subtraction (or comparison) of pointers that point to a
2114       different memory object:
2115
2116int
2117main ()
2118{
2119  /* Heap allocated memory.  */
2120  char *heap1 = (char *)__builtin_malloc (42);
2121  char *heap2 = (char *)__builtin_malloc (42);
2122  if (heap1 > heap2)
2123      return 1;
2124
2125  return 0;
2126}
2127
2128==17465==ERROR: AddressSanitizer: invalid-pointer-pair: 0x604000000010 0x6040000
212900050
2130    #0 0x40070f in main /tmp/pointer-compare.c:7
2131    #1 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86)
2132    #2 0x400629 in _start (/tmp/a.out+0x400629)
2133
21340x604000000010 is located 0 bytes inside of 42-byte region [0x604000000010,0x604
213500000003a)
2136allocated by thread T0 here:
2137    #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan
2138_malloc_linux.cc:86
2139    #1 0x4006ea in main /tmp/pointer-compare.c:5
2140    #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86)
2141
21420x604000000050 is located 0 bytes inside of 42-byte region [0x604000000050,0x604
214300000007a)
2144allocated by thread T0 here:
2145    #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan
2146_malloc_linux.cc:86
2147    #1 0x4006f8 in main /tmp/pointer-compare.c:6
2148    #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86)
2149
2150SUMMARY: AddressSanitizer: invalid-pointer-pair /tmp/pointer-compare.c:7 in main
2151
2152     * The store merging pass has been enhanced to handle bit-fields and
2153       not just constant stores, but also data copying from adjacent
2154       memory locations into other adjacent memory locations, including
2155       bitwise logical operations on the data. The pass can also handle
2156       byte swapping into memory locations.
2157     * The undefined behavior sanitizer gained two new options included in
2158       -fsanitize=undefined: -fsanitize=builtin which diagnoses at run
2159       time invalid arguments to __builtin_clz or __builtin_ctz prefixed
2160       builtins, and -fsanitize=pointer-overflow which performs cheap run
2161       time tests for pointer wrapping.
2162     * A new attribute no_sanitize can be applied to functions to instruct
2163       the compiler not to do sanitization of the options provided as
2164       arguments to the attribute. Acceptable values for no_sanitize match
2165       those acceptable by the -fsanitize command-line option.
2166
2167void __attribute__ ((no_sanitize ("alignment", "object-size")))
2168f () { /* Do something. */; }
2169
2170New Languages and Language specific improvements
2171
2172  Ada
2173
2174     * For its internal exception handling used on the host for error
2175       recovery in the front-end, the compiler now relies on the native
2176       exception handling mechanism of the host platform, which should be
2177       more efficient than the former mechanism.
2178
2179  BRIG (HSAIL)
2180
2181   In this release cycle, the focus for the BRIGFE was on stabilization
2182   and performance improvements. Also a couple of completely new features
2183   were added.
2184     * Improved support for function and module scope group segment
2185       variables. PRM specs define function and module scope group segment
2186       variables as an experimental feature. However, PRM test suite uses
2187       them. Now group segment is handled by separate book keeping of
2188       module scope and function (kernel) offsets. Each function has a
2189       "frame" in the group segment offset to which is given as an
2190       argument, similar to traditional call stack frame handling.
2191     * Reduce the number of type conversions due to the untyped HSAIL
2192       registers. Instead of always representing the HSAIL's untyped
2193       registers as unsigned int, the gccbrig now pre-analyzes the BRIG
2194       code and builds the register variables as a type used the most when
2195       storing or reading data to/from each register. This reduces the
2196       number of total casts which cannot be always optimized away.
2197     * Support for BRIG_KIND_NONE directives.
2198     * Made -O3 the default optimization level for BRIGFE.
2199     * Fixed illegal addresses generated from address expressions which
2200       refer only to offset 0.
2201     * Fixed a bug with reg+offset addressing on 32b segments. In 'large'
2202       mode, the offset is treated as 32-bit unless it's in global,
2203       read-only or kernarg address space.
2204     * Fixed a crash caused sometimes by calls with more than 4 arguments.
2205     * Fixed a mis-execution issue with kernels that have both unexpanded
2206       ID functions and calls to subfunctions.
2207     * Treat HSAIL barrier builtins as setjmp/longjump style functions to
2208       avoid illegal optimizations.
2209     * Ensure per WI copies of private variables are aligned correctly.
2210     * libhsail-rt: Assume the host runtime allocates the work group
2211       memory.
2212
2213  C family
2214
2215     * New command-line options have been added for the C and C++
2216       compilers:
2217          + [4]-Wmultistatement-macros warns about unsafe macros expanding
2218            to multiple statements used as a body of a statement such as
2219            if, else, while, switch, or for.
2220          + [5]-Wstringop-truncation warns for calls to bounded string
2221            manipulation functions such as strncat, strncpy, and stpncpy
2222            that might either truncate the copied string or leave the
2223            destination unchanged. For example, the following call to
2224            strncat is diagnosed because it appends just three of the four
2225            characters from the source string.
2226void append (char *buf, size_t bufsize)
2227{
2228    strncat (buf, ".txt", 3);
2229}
2230warning: 'strncat' output truncated copying 3 bytes from a string of length 4 [-
2231Wstringop-truncation]
2232            Similarly, in the following example, the call to strncpy
2233            specifies the size of the destination buffer as the bound. If
2234            the length of the source string is equal to or greater than
2235            this size the result of the copy will not be NUL-terminated.
2236            Therefore, the call is also diagnosed. To avoid the warning,
2237            specify sizeof buf - 1 as the bound and set the last element
2238            of the buffer to NUL.
2239void copy (const char *s)
2240{
2241    char buf[80];
2242    strncpy (buf, s, sizeof buf);
22432244}
2245warning: 'strncpy' specified bound 80 equals destination size [-Wstringop-trunca
2246tion]
2247            The -Wstringop-truncation option is included in -Wall.
2248            Note that due to GCC bug [6]82944, defining strncat, strncpy,
2249            or stpncpy as a macro in a system header as some
2250            implementations do, suppresses the warning.
2251          + [7]-Wif-not-aligned controls warnings issued in response to
2252            invalid uses of objects declared with attribute
2253            [8]warn_if_not_aligned.
2254            The -Wif-not-aligned option is included in -Wall.
2255          + [9]-Wmissing-attributes warns when a declaration of a function
2256            is missing one or more attributes that a related function is
2257            declared with and whose absence may adversely affect the
2258            correctness or efficiency of generated code. For example, in
2259            C++, the warning is issued when an explicit specialization of
2260            a primary template declared with attribute alloc_align,
2261            alloc_size, assume_aligned, format, format_arg, malloc, or
2262            nonnull is declared without it. Attributes deprecated, error,
2263            and warning suppress the warning.
2264            The -Wmissing-attributes option is included in -Wall.
2265          + [10]-Wpacked-not-aligned warns when a struct or union declared
2266            with attribute packed defines a member with an explicitly
2267            specified alignment greater than 1. Such a member will wind up
2268            under-aligned. For example, a warning will be issued for the
2269            definition of struct A in the following:
2270struct __attribute__ ((aligned (8)))
2271S8 { char a[8]; };
2272
2273struct __attribute__ ((packed)) A
2274{
2275    struct S8 s8;
2276};
2277warning: alignment 1 of 'struct S' is less than 8 [-Wpacked-not-aligned]
2278            The -Wpacked-not-aligned option is included in -Wall.
2279          + -Wcast-function-type warns when a function pointer is cast to
2280            an incompatible function pointer. This warning is enabled by
2281            -Wextra.
2282          + -Wsizeof-pointer-div warns for suspicious divisions of the
2283            size of a pointer by the size of the elements it points to,
2284            which looks like the usual way to compute the array size but
2285            won't work out correctly with pointers. This warning is
2286            enabled by -Wall.
2287          + -Wcast-align=strict warns whenever a pointer is cast such that
2288            the required alignment of the target is increased. For
2289            example, warn if a char * is cast to an int * regardless of
2290            the target machine.
2291          + -fprofile-abs-path creates absolute path names in the .gcno
2292            files. This allows gcov to find the correct sources in
2293            projects where compilations occur with different working
2294            directories.
2295     * -fno-strict-overflow is now mapped to -fwrapv -fwrapv-pointer and
2296       signed integer overflow is now undefined by default at all
2297       optimization levels. Using -fsanitize=signed-integer-overflow is
2298       now the preferred way to audit code, -Wstrict-overflow is
2299       deprecated.
2300     * The [11]-Warray-bounds option has been improved to detect more
2301       instances of out-of-bounds array indices and pointer offsets. For
2302       example, negative or excessive indices into flexible array members
2303       and string literals are detected.
2304     * The [12]-Wrestrict option introduced in GCC 7 has been enhanced to
2305       detect many more instances of overlapping accesses to objects via
2306       restrict-qualified arguments to standard memory and string
2307       manipulation functions such as memcpy and strcpy. For example, the
2308       strcpy call in the function below attempts to truncate the string
2309       by replacing its initial characters with the last four. However,
2310       because the function writes the terminating NUL into a[4], the
2311       copies overlap and the call is diagnosed.
2312void f (void)
2313{
2314    char a[] = "abcd1234";
2315    strcpy (a, a + 4);
23162317}
2318warning: 'strcpy' accessing 5 bytes at offsets 0 and 4 overlaps 1 byte at offset
2319 4 [-Wrestrict]
2320       The -Wrestrict option is included in -Wall.
2321     * Several optimizer enhancements have enabled improvements to the
2322       [13]-Wformat-overflow and [14]-Wformat-truncation options. The
2323       warnings detect more instances of buffer overflow and truncation
2324       than in GCC 7 and are better at avoiding certain kinds of false
2325       positives.
2326     * When reporting mismatching argument types at a function call, the C
2327       and C++ compilers now underline both the argument and the pertinent
2328       parameter in the declaration.
2329$ gcc arg-type-mismatch.cc
2330arg-type-mismatch.cc: In function 'int caller(int, int, float)':
2331arg-type-mismatch.cc:5:24: error: invalid conversion from 'int' to 'const char*'
2332 [-fpermissive]
2333   return callee(first, second, third);
2334                        ^~~~~~
2335arg-type-mismatch.cc:1:40: note:   initializing argument 2 of 'int callee(int, c
2336onst char*, float)'
2337 extern int callee(int one, const char *two, float three);
2338                            ~~~~~~~~~~~~^~~
2339
2340     * When reporting on unrecognized identifiers, the C and C++ compilers
2341       will now emit fix-it hints suggesting #include directives for
2342       various headers in the C and C++ standard libraries.
2343$ gcc incomplete.c
2344incomplete.c: In function 'test':
2345incomplete.c:3:10: error: 'NULL' undeclared (first use in this function)
2346   return NULL;
2347          ^~~~
2348incomplete.c:3:10: note: 'NULL' is defined in header '<stddef.h>'; did you forge
2349t to '#include <stddef.h>'?
2350incomplete.c:1:1:
2351+#include <stddef.h>
2352 const char *test(void)
2353incomplete.c:3:10:
2354   return NULL;
2355          ^~~~
2356incomplete.c:3:10: note: each undeclared identifier is reported only once for ea
2357ch function it appears in
2358
2359$ gcc incomplete.cc
2360incomplete.cc:1:6: error: 'string' in namespace 'std' does not name a type
2361 std::string s("hello world");
2362      ^~~~~~
2363incomplete.cc:1:1: note: 'std::string' is defined in header '<string>'; did you
2364forget to '#include <string>'?
2365+#include <string>
2366 std::string s("hello world");
2367 ^~~
2368
2369     * The C and C++ compilers now use more intuitive locations when
2370       reporting on missing semicolons, and offer fix-it hints:
2371$ gcc t.c
2372t.c: In function 'test':
2373t.c:3:12: error: expected ';' before '}' token
2374   return 42
2375            ^
2376            ;
2377 }
2378 ~
2379
2380     * When reporting on missing '}' and ')' tokens, the C and C++
2381       compilers will now highlight the corresponding '{' and '(' token,
2382       issuing a 'note' if it's on a separate line:
2383$ gcc unclosed.c
2384unclosed.c: In function 'log_when_out_of_range':
2385unclosed.c:12:50: error: expected ')' before '{' token
2386       && (temperature < MIN || temperature > MAX) {
2387                                                  ^~
2388                                                  )
2389unclosed.c:11:6: note: to match this '('
2390   if (logging_enabled && check_range ()
2391      ^
2392
2393       or highlighting it directly if it's on the same line:
2394$ gcc unclosed-2.c
2395unclosed-2.c: In function 'test':
2396unclosed-2.c:8:45: error: expected ')' before '{' token
2397   if (temperature < MIN || temperature > MAX {
2398      ~                                      ^~
2399                                             )
2400
2401       They will also emit fix-it hints.
2402
2403  C++
2404
2405     * GCC 8 (-fabi-version=12) has a couple of corrections to the calling
2406       convention, which changes the ABI for some uncommon code:
2407          + Passing an empty class as an argument now takes up no space on
2408            x86_64, as required by the psABI.
2409          + Passing or returning a class with only deleted copy and move
2410            constructors now uses the same calling convention as a class
2411            with a non-trivial copy or move constructor. This only affects
2412            C++17 mode, as in earlier standards passing or returning such
2413            a class was impossible.
2414          + WARNING: In GCC 8.1 the second change mistakenly also affects
2415            classes with a deleted copy constructor and defaulted trivial
2416            move constructor (bug [15]c++/86094). This issue is fixed in
2417            GCC 8.2 (-fabi-version=13).
2418       You can test whether these changes affect your code with -Wabi=11
2419       (or -Wabi=12 in GCC 8.2 for the third issue); if these changes are
2420       problematic for your project, the GCC 7 ABI can be selected with
2421       -fabi-version=11.
2422     * The value of the C++11 alignof operator has been corrected to match
2423       C _Alignof (minimum alignment) rather than GNU __alignof__
2424       (preferred alignment); on ia32 targets this means that
2425       alignof(double) is now 4 rather than 8. Code that wants the
2426       preferred alignment should use __alignof__ instead.
2427     * New command-line options have been added for the C++ compiler to
2428       control warnings:
2429          + [16]-Wclass-memaccess warns when objects of non-trivial class
2430            types are manipulated in potentially unsafe ways by raw memory
2431            functions such as memcpy, or realloc. The warning helps detect
2432            calls that bypass user-defined constructors or copy-assignment
2433            operators, corrupt virtual table pointers, data members of
2434            const-qualified types or references, or member pointers. The
2435            warning also detects calls that would bypass access controls
2436            to data members. For example, a call such as:
2437        memcpy (&std::cout, &std::cerr, sizeof std::cout);
2438            results in
2439        warning: 'void* memcpy(void*, const void*, long unsigned int)' writing t
2440o an object of type 'std::ostream' {aka 'class std::basic_ostream<char>'} with n
2441o trivial copy-assignment [-Wclass-memaccess]
2442            The -Wclass-memaccess option is included in -Wall.
2443     * The C++ front end has experimental support for some of the upcoming
2444       C++2a draft features with the -std=c++2a or -std=gnu++2a flags,
2445       including designated initializers, default member initializers for
2446       bit-fields, __VA_OPT__ (except that #__VA_OPT__ is unsupported),
2447       lambda [=, this] captures, etc. For a full list of new features,
2448       see [17]the C++ status page.
2449     * When reporting on attempts to access private fields of a class or
2450       struct, the C++ compiler will now offer fix-it hints showing how to
2451       use an accessor function to get at the field in question, if one
2452       exists.
2453$ gcc accessor.cc
2454accessor.cc: In function 'void test(foo*)':
2455accessor.cc:12:12: error: 'double foo::m_ratio' is private within this context
2456   if (ptr->m_ratio >= 0.5)
2457            ^~~~~~~
2458accessor.cc:7:10: note: declared private here
2459   double m_ratio;
2460          ^~~~~~~
2461accessor.cc:12:12: note: field 'double foo::m_ratio' can be accessed via 'double
2462 foo::get_ratio() const'
2463   if (ptr->m_ratio >= 0.5)
2464            ^~~~~~~
2465            get_ratio()
2466
2467     * The C++ compiler can now give you a hint if you use a macro before
2468       it was defined (e.g. if you mess up the order of your #include
2469       directives):
2470$ gcc ordering.cc
2471ordering.cc:2:24: error: expected ';' at end of member declaration
2472   virtual void clone() const OVERRIDE { }
2473                        ^~~~~
2474                             ;
2475ordering.cc:2:30: error: 'OVERRIDE' does not name a type
2476   virtual void clone() const OVERRIDE { }
2477                              ^~~~~~~~
2478ordering.cc:2:30: note: the macro 'OVERRIDE' had not yet been defined
2479In file included from ordering.cc:5:
2480c++11-compat.h:2: note: it was later defined here
2481 #define OVERRIDE override
2482
2483
2484     * The -Wold-style-cast diagnostic can now emit fix-it hints telling
2485       you when you can use a static_cast, const_cast, or
2486       reinterpret_cast.
2487$ gcc -c old-style-cast-fixits.cc -Wold-style-cast
2488old-style-cast-fixits.cc: In function 'void test(void*)':
2489old-style-cast-fixits.cc:5:19: warning: use of old-style cast to 'struct foo*' [
2490-Wold-style-cast]
2491   foo *f = (foo *)ptr;
2492                   ^~~
2493            ----------
2494            static_cast<foo *> (ptr)
2495
2496     * When reporting on problems within extern "C" linkage
2497       specifications, the C++ compiler will now display the location of
2498       the start of the extern "C".
2499$ gcc -c extern-c.cc
2500extern-c.cc:3:1: error: template with C linkage
2501 template <typename T> void test (void);
2502 ^~~~~~~~
2503In file included from extern-c.cc:1:
2504unclosed.h:1:1: note: 'extern "C"' linkage started here
2505 extern "C" {
2506 ^~~~~~~~~~
2507extern-c.cc:3:39: error: expected '}' at end of input
2508 template <typename T> void test (void);
2509                                       ^
2510In file included from extern-c.cc:1:
2511unclosed.h:1:12: note: to match this '{'
2512 extern "C" {
2513            ^
2514
2515     * When reporting on mismatching template types, the C++ compiler will
2516       now use color to highlight the mismatching parts of the template,
2517       and will elide the parameters that are common between two
2518       mismatching templates, printing [...] instead:
2519$ gcc templates.cc
2520templates.cc: In function 'void test()':
2521templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl
2522e>' to 'vector<int>'
2523   fn_1(vector<double> ());
2524        ^~~~~~~~~~~~~~~~~
2525templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<[...]
2526,double>' to 'map<[...],int>'
2527   fn_2(map<int, double>());
2528        ^~~~~~~~~~~~~~~~~~
2529
2530       Those [...] elided parameters can be seen using -fno-elide-type:
2531$ gcc templates.cc -fno-elide-type
2532templates.cc: In function 'void test()':
2533templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl
2534e>' to 'vector<int>'
2535   fn_1(vector<double> ());
2536        ^~~~~~~~~~~~~~~~~
2537templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<int,d
2538ouble>' to 'map<int,int>'
2539   fn_2(map<int, double>());
2540        ^~~~~~~~~~~~~~~~~~
2541
2542       The C++ compiler has also gained an option
2543       -fdiagnostics-show-template-tree which visualizes such mismatching
2544       templates in a hierarchical form:
2545$ gcc templates-2.cc -fdiagnostics-show-template-tree
2546templates-2.cc: In function 'void test()':
2547templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou
2548ble>' to 'vector<int>'
2549  vector<
2550    [double != int]>
2551   fn_1(vector<double> ());
2552        ^~~~~~~~~~~~~~~~~
2553templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve
2554ctor<double> >()' from 'map<map<[...],vector<double>>,vector<double>>' to 'map<m
2555ap<[...],vector<float>>,vector<float>>'
2556  map<
2557    map<
2558      [...],
2559      vector<
2560        [double != float]>>,
2561    vector<
2562      [double != float]>>
2563   fn_2(map<map<int, vector<double>>, vector<double>> ());
2564        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2565
2566       which again works with -fno-elide-type:
2567$ gcc templates-2.cc -fdiagnostics-show-template-tree -fno-elide-type
2568templates-2.cc: In function 'void test()':
2569templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou
2570ble>' to 'vector<int>'
2571  vector<
2572    [double != int]>
2573   fn_1(vector<double> ());
2574        ^~~~~~~~~~~~~~~~~
2575templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve
2576ctor<double> >()' from 'map<map<int,vector<double>>,vector<double>>' to 'map<map
2577<int,vector<float>>,vector<float>>'
2578  map<
2579    map<
2580      int,
2581      vector<
2582        [double != float]>>,
2583    vector<
2584      [double != float]>>
2585   fn_2(map<map<int, vector<double>>, vector<double>> ());
2586        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2587
2588     * Flowing off the end of a non-void function is considered
2589       unreachable and may be subject to optimization on that basis. As a
2590       result of this change, -Wreturn-type warnings are enabled by
2591       default for C++.
2592
2593    Runtime Library (libstdc++)
2594
2595     * Improved experimental support for C++17, including the following
2596       features:
2597          + Deduction guides to support class template argument deduction.
2598          + std::filesystem implementation.
2599          + std::char_traits<char> and std::char_traits<wchar_t> are
2600            usable in constant expressions.
2601          + std::to_chars and std::from_chars (for integers only, not for
2602            floating point types).
2603     * Experimental support for C++2a: std::to_address (thanks to Glen
2604       Fernandes) and std::endian.
2605     * On GNU/Linux, std::random_device::entropy() accesses the kernel's
2606       entropy count for the random device, if known (thanks to Xi
2607       Ruoyao).
2608     * Support for std::experimental::source_location.
2609     * AddressSanitizer integration for std::vector, detecting
2610       out-of-range accesses to the unused capacity of a vector.
2611     * Extensions __gnu_cxx::airy_ai and __gnu_cxx::airy_bi added to the
2612       Mathematical Special Functions.
2613
2614  Fortran
2615
2616     * The main version of libfortran has been changed to 5.
2617     * Parameterized derived types, a major feature of Fortran 2003, have
2618       been implemented.
2619     * Partial support is provided for Fortran 2018 teams, which are
2620       hierarchical subsets of images that execute independently of other
2621       image subsets.
2622     * The maximum rank for arrays has been increased to 15, conforming to
2623       the Fortran 2008 standard.
2624     * Transformational intrinsics are now fully supported in
2625       initialization expressions.
2626     * New flag -fc-prototypes to write C prototypes for BIND(C)
2627       procedures and variables.
2628     * If -fmax-stack-var-size is honored if given together with -Ofast,
2629       -fstack-arrays is no longer set in that case.
2630     * New options -fdefault-real-16 and -fdefault-real-10 to control the
2631       default kind of REAL variables.
2632     * A warning is now issued if an array subscript inside a DO loop
2633       could lead to an out-of-bounds-access. The new option
2634       -Wdo-subscript, enabled by -Wextra, warns about this even if the
2635       compiler can not prove that the code will be executed.
2636     * The Fortran front end now attempts to interchange loops if it is
2637       deemed profitable. So far, this is restricted to FORALL and DO
2638       CONCURRENT statements with multiple indices. This behavior be
2639       controlled with the new flag -ffrontend-loop-interchange, which is
2640       enabled with optimization by default. The
2641       -Wfrontend-loop-interchange option warns about such occurrences.
2642     * When an actual argument contains too few elements for a dummy
2643       argument, an error is now issued. The -std=legacy option can be
2644       used to still compile such code.
2645     * The RECL= argument to OPEN and INQUIRE statements now allows 64-bit
2646       integers, making records larger than 2GiB possible.
2647     * The GFORTRAN_DEFAULT_RECL environment variable no longer has any
2648       effect. The record length for preconnected units is now larger than
2649       any practical limit, same as for sequential access units opened
2650       without an explicit RECL= specifier.
2651     * Character variables longer than HUGE(0) elements are now possible
2652       on 64-bit targets. Note that this changes the procedure call ABI
2653       for all procedures with character arguments on 64-bit targets, as
2654       the type of the hidden character length argument has changed. The
2655       hidden character length argument is now of type INTEGER(C_SIZE_T).
2656     * Partial support is provided for Fortran 2018 teams, which are
2657       hierarchical subsets of images that execute independently of other
2658       image subsets.
2659
2660  Go
2661
2662     * GCC 8 provides a complete implementation of the Go 1.10.1 user
2663       packages.
2664     * The garbage collector is now fully concurrent. As before, values
2665       stored on the stack are scanned conservatively, but value stored in
2666       the heap are scanned precisely.
2667     * Escape analysis is fully implemented and enabled by default in the
2668       Go front end. This significantly reduces the number of heap
2669       allocations by allocating values on the stack instead.
2670
2671libgccjit
2672
2673   The libgccjit API gained four new entry points:
2674     * [18]gcc_jit_type_get_vector and
2675     * [19]gcc_jit_context_new_rvalue_from_vector for working with
2676       vectors,
2677     * [20]gcc_jit_type_get_aligned
2678     * [21]gcc_jit_function_get_address
2679
2680   The C code generated by [22]gcc_jit_context_dump_reproducer_to_file is
2681   now easier-to-read.
2682
2683New Targets and Target Specific Improvements
2684
2685  AArch64
2686
2687     * The Armv8.4-A architecture is now supported. It can be used by
2688       specifying the -march=armv8.4-a option.
2689     * The Dot Product instructions are now supported as an optional
2690       extension to the Armv8.2-A architecture and newer and are mandatory
2691       on Armv8.4-A. The extension can be used by specifying the +dotprod
2692       architecture extension. E.g. -march=armv8.2-a+dotprod.
2693     * The Armv8-A +crypto extension has now been split into two
2694       extensions for finer grained control:
2695          + +aes which contains the Armv8-A AES crytographic instructions.
2696          + +sha2 which contains the Armv8-A SHA2 and SHA1 cryptographic
2697            instructions.
2698       Using +crypto will now enable these two extensions.
2699     * New Armv8.4-A FP16 Floating Point Multiplication Variant
2700       instructions have been added. These instructions are mandatory in
2701       Armv8.4-A but available as an optional extension to Armv8.2-A and
2702       Armv8.3-A. The new extension can be used by specifying the +fp16fml
2703       architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A
2704       the instructions can be enabled by specifying +fp16.
2705     * New cryptographic instructions have been added as optional
2706       extensions to Armv8.2-A and newer. These instructions can be
2707       enabled with:
2708          + +sha3 New SHA3 and SHA2 instructions from Armv8.4-A. This
2709            implies +sha2.
2710          + +sm4 New SM3 and SM4 instructions from Armv8.4-A.
2711     * The Scalable Vector Extension (SVE) is now supported as an optional
2712       extension to the Armv8.2-A architecture and newer. This support
2713       includes automatic vectorization with SVE instructions, but it does
2714       not yet include the SVE Arm C Language Extensions (ACLE). It can be
2715       enabled by specifying the +sve architecture extension (for example,
2716       -march=armv8.2-a+sve). By default, the generated code works with
2717       all vector lengths, but it can be made specific to N-bit vectors
2718       using -msve-vector-bits=N.
2719     * Support has been added for the following processors (GCC
2720       identifiers in parentheses):
2721          + Arm Cortex-A75 (cortex-a75).
2722          + Arm Cortex-A55 (cortex-a55).
2723          + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE
2724            (cortex-a75.cortex-a55).
2725       The GCC identifiers can be used as arguments to the -mcpu or -mtune
2726       options, for example: -mcpu=cortex-a75 or -mtune=cortex-a75 or as
2727       arguments to the equivalent target attributes and pragmas.
2728
2729  ARC
2730
2731     * Added support for:
2732          + Fast interrupts.
2733          + Naked functions.
2734          + aux variable attributes.
2735          + uncached type qualifier.
2736          + Secure functions via sjli instruction.
2737     * New exception handling implementation.
2738     * Revamped trampoline implementation.
2739     * Refactored small data feature implementation, controlled via -G
2740       command line option.
2741     * New support for reduced register set ARC architecture
2742       configurations, controlled via -mrf16 command line option.
2743     * Refurbished and improved support for zero overhead loops.
2744       Introduced -mlpc-width command line option to control the width of
2745       lp_count register.
2746
2747  ARM
2748
2749     * The -mfpu option now takes a new option setting of -mfpu=auto. When
2750       set to this the floating-point and SIMD settings are derived from
2751       the settings of the -mcpu or -march options. The internal CPU
2752       configurations have been updated with information about the
2753       permitted floating-point configurations supported. See the user
2754       guide for further information about the extended option syntax for
2755       controlling architectural extensions via the -march option.
2756       -mfpu=auto is now the default setting unless the compiler has been
2757       configured with an explicit --with-fpu option.
2758     * The -march and -mcpu options now accept optional extensions to the
2759       architecture or CPU option, allowing the user to enable or disable
2760       any such extensions supported by that architecture or CPU such as
2761       (but not limited to) floating-point and AdvancedSIMD. For example:
2762       the option -mcpu=cortex-a53+nofp will generate code for the
2763       Cortex-A53 processor with no floating-point support. This, in
2764       combination with the new -mfpu=auto option, provides a
2765       straightforward way of specifying a valid build target through a
2766       single -mcpu or -march option. The -mtune option accepts the same
2767       arguments as -mcpu but only the CPU name has an effect on tuning.
2768       The architecture extensions do not have any effect. For details of
2769       what extensions a particular architecture or CPU option supports
2770       please refer to the [23]documentation.
2771     * The -mstructure-size-boundary option has been deprecated and will
2772       be removed in a future release.
2773     * The default link behavior for Armv6 and Armv7-R targets has been
2774       changed to produce BE8 format when generating big-endian images. A
2775       new flag -mbe32 can be used to force the linker to produce legacy
2776       BE32 format images. There is no change of behavior for Armv6-M and
2777       other Armv7 or later targets: these already defaulted to BE8
2778       format. This change brings GCC into alignment with other compilers
2779       for the ARM architecture.
2780     * The Armv8-R architecture is now supported. It can be used by
2781       specifying the -march=armv8-r option.
2782     * The Armv8.3-A architecture is now supported. It can be used by
2783       specifying the -march=armv8.3-a option.
2784     * The Armv8.4-A architecture is now supported. It can be used by
2785       specifying the -march=armv8.4-a option.
2786     * The Dot Product instructions are now supported as an optional
2787       extension to the Armv8.2-A architecture and newer and are mandatory
2788       on Armv8.4-A. The extension can be used by specifying the +dotprod
2789       architecture extension. E.g. -march=armv8.2-a+dotprod.
2790     * Support for setting extensions and architectures using the GCC
2791       target pragma and attribute has been added. It can be used by
2792       specifying #pragma GCC target ("arch=..."), #pragma GCC target
2793       ("+extension"), __attribute__((target("arch=..."))) or
2794       __attribute__((target("+extension"))).
2795     * New Armv8.4-A FP16 Floating Point Multiplication Variant
2796       instructions have been added. These instructions are mandatory in
2797       Armv8.4-A but available as an optional extension to Armv8.2-A and
2798       Armv8.3-A. The new extension can be used by specifying the +fp16fml
2799       architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A
2800       the instructions can be enabled by specifying +fp16.
2801     * Support has been added for the following processors (GCC
2802       identifiers in parentheses):
2803          + Arm Cortex-A75 (cortex-a75).
2804          + Arm Cortex-A55 (cortex-a55).
2805          + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE
2806            (cortex-a75.cortex-a55).
2807          + Arm Cortex-R52 for Armv8-R (cortex-r52).
2808       The GCC identifiers can be used as arguments to the -mcpu or -mtune
2809       options, for example: -mcpu=cortex-a75 or -mtune=cortex-r52 or as
2810       arguments to the equivalent target attributes and pragmas.
2811
2812  AVR
2813
2814     * The AVR port now supports the following XMEGA-like devices:
2815
2816     ATtiny212, ATtiny214, ATtiny412, ATtiny414, ATtiny416, ATtiny417,
2817     ATtiny814, ATtiny816, ATtiny817, ATtiny1614, ATtiny1616, ATtiny1617,
2818     ATtiny3214, ATtiny3216, ATtiny3217
2819       The new devices are listed under [24]-mmcu=avrxmega3.
2820          + These devices see flash memory in the RAM address space, so
2821            that features like PROGMEM and __flash are not needed any more
2822            (as opposed to other AVR families for which read-only data
2823            will be located in RAM except special, non-standard features
2824            are used to locate and access such data). This requires that
2825            the compiler is used with Binutils 2.29 or newer so that
2826            [25]read-only data will be located in flash memory.
2827          + A new command-line option -mshort-calls is supported. This
2828            option is used internally for multilib selection of the
2829            avrxmega3 variants. It is not an optimization option. Do not
2830            set it by hand.
2831     * The compiler now generates [26]efficient interrupt service routine
2832       (ISR) prologues and epilogues. This is achieved by using the new
2833       [27]AVR pseudo instruction __gcc_isr which is supported and
2834       resolved by the GNU assembler.
2835          + As the __gcc_isr pseudo-instruction will be resolved by the
2836            assembler, inline assembly is transparent to the process. This
2837            means that when inline assembly uses an instruction like INC
2838            that clobbers the condition code, then the assembler will
2839            detect this and generate an appropriate ISR prologue /
2840            epilogue chunk to save / restore SREG as needed.
2841          + A new command-line option -mno-gas-isr-prologues disables the
2842            generation of the __gcc_isr pseudo instruction. Any non-naked
2843            ISR will save and restore SREG, tmp_reg and zero_reg, no
2844            matter whether the respective register is clobbered or used.
2845          + The feature is turned on per default for all optimization
2846            levels except for -O0 and -Og. It is explicitly enabled by
2847            means of option -mgas-isr-prologues.
2848          + Support has been added for a new [28]AVR function attribute
2849            no_gccisr. It can be used to disable __gcc_isr pseudo
2850            instruction generation for individual ISRs.
2851          + This optimization is only available if GCC is configured with
2852            GNU Binutils 2.29 or newer; or at least with a version of
2853            Binutils that implements feature [29]PR21683.
2854     * The compiler no more saves / restores registers in main; the effect
2855       is the same as if attribute OS_task was specified for main. This
2856       optimization can be switched off by the new command-line option
2857       -mno-main-is-OS_task.
2858
2859  IA-32/x86-64
2860
2861     * The x86 port now supports the naked function attribute.
2862     * Better tuning for znver1 and Intel Core based CPUs.
2863     * Vectorization cost metrics has been reworked leading to significant
2864       improvements on some benchmarks.
2865     * GCC now supports the Intel CPU named Cannonlake through
2866       -march=cannonlake. The switch enables the AVX512VBMI, AVX512IFMA
2867       and SHA ISA extensions.
2868     * GCC now supports the Intel CPU named Icelake through
2869       -march=icelake. The switch enables the AVX512VNNI, GFNI, VAES,
2870       AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID and AVX512VPOPCNTDQ
2871       ISA extensions.
2872     * GCC now supports the Intel Control-flow Enforcement Technology
2873       (CET) extension through -fcf-protection option.
2874
2875  NDS32
2876
2877     * New command-line options -mext-perf, -mext-perf2, and -mext-string
2878       have been added for performance extension instructions.
2879
2880  Nios II
2881
2882     * The Nios II back end has been improved to generate better-optimized
2883       code. Changes include switching to LRA, more accurate cost models,
2884       and more compact code for addressing static variables.
2885     * New command-line options -mgprel-sec= and -mr0rel-sec= have been
2886       added.
2887     * The stack-smashing protection options are now enabled on Nios II.
2888
2889  PA-RISC
2890
2891     * The default call ABI on 32-bit linux has been changed from callee
2892       copies to caller copies. This affects objects larger than eight
2893       bytes passed by value. The goal is to improve compatibility with
2894       x86 and resolve issues with OpenMP.
2895     * Other PA-RISC targets are unchanged.
2896
2897  PowerPC / PowerPC64 / RS6000
2898
2899     * The PowerPC SPE support is split off to a separate powerpcspe port.
2900       The separate port is deprecated and might be removed in a future
2901       release.
2902     * The Paired Single support (as used on some PPC750 CPUs, -mpaired,
2903       powerpc*-*-linux*paired*) is deprecated and will be removed in a
2904       future release.
2905     * The Xilinx floating point support (-mxilinx-fpu,
2906       powerpc-xilinx-eabi*) is deprecated and will be removed in a future
2907       release.
2908     * Support for using big-endian AltiVec intrinsics on a little-endian
2909       target (-maltivec=be) is deprecated and will be removed in a future
2910       release.
2911
2912  Tile
2913
2914     * The TILE-Gx port is deprecated and will be removed in a future
2915       release.
2916
2917Operating Systems
2918
2919  Windows
2920
2921     * GCC on Microsoft Windows can now be configured via
2922       --enable-mingw-wildcard or --disable-mingw-wildcard to force a
2923       specific behavior for GCC itself with regards to supporting the
2924       wildcard character. Prior versions of GCC would follow the
2925       configuration of the MinGW runtime. This behavior can still be
2926       obtained by not using the above options or by using
2927       --enable-mingw-wildcard=platform.
2928
2929Improvements for plugin authors
2930
2931     * Plugins can now register a callback hook for when comments are
2932       encountered by the C and C++ compilers, e.g. allowing for plugins
2933       to handle documentation markup in code comments.
2934     * The gdbinit support script for debugging GCC now has a
2935       break-on-diagnostic command, providing an easy way to trigger a
2936       breakpoint whenever a diagnostic is emitted.
2937     * The API for creating fix-it hints now supports newlines, and for
2938       emitting mutually incompatible fix-it hints for one diagnostic.
2939
2940GCC 8.1
2941
2942   This is the [30]list of problem reports (PRs) from GCC's bug tracking
2943   system that are known to be fixed in the 8.1 release. This list might
2944   not be complete (that is, it is possible that some PRs that have been
2945   fixed are not listed here).
2946
2947GCC 8.2
2948
2949   This is the [31]list of problem reports (PRs) from GCC's bug tracking
2950   system that are known to be fixed in the 8.2 release. This list might
2951   not be complete (that is, it is possible that some PRs that have been
2952   fixed are not listed here).
2953
2954  General Improvements
2955
2956     * Fixed LTO link-time performance problems caused by an overflow in
2957       the partitioning algorithm while building large binaries.
2958
2959  Language Specific Changes
2960
2961    C++
2962
2963   GCC 8.2 fixed a bug introduced in GCC 8.1 affecting passing or
2964   returning of classes with a deleted copy constructor and defaulted
2965   trivial move constructor (bug [32]c++/86094). GCC 8.2 introduces
2966   -fabi-version=13 and makes it the default, ABI incompatibilities
2967   between GCC 8.1 and 8.2 can be reported with -Wabi=12. See [33]C++
2968   changes for more details.
2969
2970  Target Specific Changes
2971
2972    IA-32/x86-64
2973
2974     * -mtune=native performance regression [34]PR84413 on Intel Skylake
2975       processors has been fixed.
2976
2977GCC 8.3
2978
2979   This is the [35]list of problem reports (PRs) from GCC's bug tracking
2980   system that are known to be fixed in the 8.3 release. This list might
2981   not be complete (that is, it is possible that some PRs that have been
2982   fixed are not listed here).
2983
2984  Windows
2985
2986     * A C++ Microsoft ABI bitfield layout bug, [36]PR87137 has been
2987       fixed. A non-field declaration could cause the current bitfield
2988       allocation unit to be completed, incorrectly placing a following
2989       bitfield into a new allocation unit. The Microsoft ABI is selected
2990       for:
2991          + Mingw targets
2992          + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields
2993            option is specified, or __attribute__((ms_struct)) is used
2994          + SuperH targets when the -mhitachi option is specified, or
2995            __attribute__((renesas)) is used
2996       GCC 8 introduced additional cases of this defect, but rather than
2997       resolve only those regressions, we decided to resolve all the cases
2998       of this defect in single change.
2999
3000GCC 8.4
3001
3002   This is the [37]list of problem reports (PRs) from GCC's bug tracking
3003   system that are known to be fixed in the 8.4 release. This list might
3004   not be complete (that is, it is possible that some PRs that have been
3005   fixed are not listed here).
3006
3007GCC 8.5
3008
3009   This is the [38]list of problem reports (PRs) from GCC's bug tracking
3010   system that are known to be fixed in the 8.5 release. This list might
3011   not be complete (that is, it is possible that some PRs that have been
3012   fixed are not listed here).
3013
3014  Target Specific Changes
3015
3016    AArch64
3017
3018     * The option -moutline-atomics has been added to aid deployment of
3019       the Large System Extensions (LSE) on GNU/Linux systems built with a
3020       baseline architecture targeting Armv8-A. When the option is
3021       specified code is emitted to detect the presence of LSE
3022       instructions at run time and use them for standard atomic
3023       operations. For more information please refer to the documentation.
3024
3025
3026    For questions related to the use of GCC, please consult these web
3027    pages and the [39]GCC manuals. If that fails, the
3028    [40]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3029    web pages and the development of GCC are welcome on our developer
3030    list at [41]gcc@gcc.gnu.org. All of [42]our lists have public
3031    archives.
3032
3033   Copyright (C) [43]Free Software Foundation, Inc. Verbatim copying and
3034   distribution of this entire article is permitted in any medium,
3035   provided this notice is preserved.
3036
3037   These pages are [44]maintained by the GCC team. Last modified
3038   2021-10-01[45].
3039
3040References
3041
3042   1. http://gcc.gnu.org/gcc-8/porting_to.html
3043   2. http://gcc.gnu.org/onlinedocs/index.html#current
3044   3. https://gcc.gnu.org/ml/gcc/2018-04/msg00102.html
3045   4. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmultistatement-macros
3046   5. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wstringop-truncation
3047   6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82944
3048   7. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wif-not-aligned
3049   8. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Common-Variable-Attributes.html#index-warn_005fif_005fnot_005faligned-variable-attribute
3050   9. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmissing-attributes
3051  10. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wpacked-not-aligned
3052  11. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Warray-bounds
3053  12. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wrestrict
3054  13. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-overflow
3055  14. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-truncation
3056  15. https://gcc.gnu.org/PR86094
3057  16. https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess
3058  17. http://gcc.gnu.org/projects/cxx-status.html#cxx2a
3059  18. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_vector
3060  19. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/expressions.html#gcc_jit_context_new_rvalue_from_vector
3061  20. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_aligned
3062  21. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/function-pointers.html#gcc_jit_function_get_address
3063  22. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/contexts.html#gcc_jit_context_dump_reproducer_to_file
3064  23. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/ARM-Options.html#ARM-Options
3065  24. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Options.html
3066  25. https://sourceware.org/PR21472
3067  26. https://gcc.gnu.org/PR20296
3068  27. https://sourceware.org/binutils/docs-2.29/as/AVR-Pseudo-Instructions.html
3069  28. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Function-Attributes.html
3070  29. https://sourceware.org/PR21683
3071  30. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.0
3072  31. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.2
3073  32. https://gcc.gnu.org/PR86094
3074  33. http://gcc.gnu.org/gcc-8/changes.html#cxx
3075  34. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84413
3076  35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.3
3077  36. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137
3078  37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.4
3079  38. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.5
3080  39. https://gcc.gnu.org/onlinedocs/
3081  40. mailto:gcc-help@gcc.gnu.org
3082  41. mailto:gcc@gcc.gnu.org
3083  42. https://gcc.gnu.org/lists.html
3084  43. https://www.fsf.org/
3085  44. https://gcc.gnu.org/about.html
3086  45. http://validator.w3.org/check/referer
3087======================================================================
3088http://gcc.gnu.org/gcc-7/index.html
3089                              GCC 7 Release Series
3090
3091   (This release series is no longer supported.)
3092
3093   Nov 14, 2019
3094
3095   The [1]GNU project and the GCC developers are pleased to announce the
3096   release of GCC 7.5.
3097
3098   This release is a bug-fix release, containing fixes for regressions in
3099   GCC 7.4 relative to previous releases of GCC.
3100
3101Release History
3102
3103   GCC 7.5
3104          Nov 14, 2019 ([2]changes, [3]documentation)
3105
3106   GCC 7.4
3107          Dec 6, 2018 ([4]changes, [5]documentation)
3108
3109   GCC 7.3
3110          Jan 25, 2018 ([6]changes, [7]documentation)
3111
3112   GCC 7.2
3113          Aug 14, 2017 ([8]changes, [9]documentation)
3114
3115   GCC 7.1
3116          May 2, 2017 ([10]changes, [11]documentation)
3117
3118References and Acknowledgements
3119
3120   GCC used to stand for the GNU C Compiler, but since the compiler
3121   supports several other languages aside from C, it now stands for the
3122   GNU Compiler Collection.
3123
3124   A list of [12]successful builds is updated as new information becomes
3125   available.
3126
3127   The GCC developers would like to thank the numerous people that have
3128   contributed new features, improvements, bug fixes, and other changes as
3129   well as test results to GCC. This [13]amazing group of volunteers is
3130   what makes GCC successful.
3131
3132   For additional information about GCC please refer to the [14]GCC
3133   project web site or contact the [15]GCC development mailing list.
3134
3135   To obtain GCC please use [16]our mirror sites or [17]our version
3136   control system.
3137
3138
3139    For questions related to the use of GCC, please consult these web
3140    pages and the [18]GCC manuals. If that fails, the
3141    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3142    web pages and the development of GCC are welcome on our developer
3143    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
3144    archives.
3145
3146   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
3147   distribution of this entire article is permitted in any medium,
3148   provided this notice is preserved.
3149
3150   These pages are [23]maintained by the GCC team. Last modified
3151   2022-05-06[24].
3152
3153References
3154
3155   1. http://www.gnu.org/
3156   2. http://gcc.gnu.org/gcc-7/changes.html
3157   3. http://gcc.gnu.org/onlinedocs/7.5.0/
3158   4. http://gcc.gnu.org/gcc-7/changes.html
3159   5. http://gcc.gnu.org/onlinedocs/7.4.0/
3160   6. http://gcc.gnu.org/gcc-7/changes.html
3161   7. http://gcc.gnu.org/onlinedocs/7.3.0/
3162   8. http://gcc.gnu.org/gcc-7/changes.html
3163   9. http://gcc.gnu.org/onlinedocs/7.2.0/
3164  10. http://gcc.gnu.org/gcc-7/changes.html
3165  11. http://gcc.gnu.org/onlinedocs/7.1.0/
3166  12. http://gcc.gnu.org/gcc-7/buildstat.html
3167  13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
3168  14. http://gcc.gnu.org/index.html
3169  15. mailto:gcc@gcc.gnu.org
3170  16. http://gcc.gnu.org/mirrors.html
3171  17. http://gcc.gnu.org/git.html
3172  18. https://gcc.gnu.org/onlinedocs/
3173  19. mailto:gcc-help@gcc.gnu.org
3174  20. mailto:gcc@gcc.gnu.org
3175  21. https://gcc.gnu.org/lists.html
3176  22. https://www.fsf.org/
3177  23. https://gcc.gnu.org/about.html
3178  24. http://validator.w3.org/check/referer
3179======================================================================
3180http://gcc.gnu.org/gcc-7/changes.html
3181                              GCC 7 Release Series
3182                        Changes, New Features, and Fixes
3183
3184   This page is a brief summary of some of the huge number of improvements
3185   in GCC 7. For more information, see the [1]Porting to GCC 7 page and
3186   the [2]full GCC documentation.
3187
3188Caveats
3189
3190     * GCC now uses [3]LRA (a new local register allocator) by default for
3191       new targets.
3192     * The non-standard C++0x type traits has_trivial_default_constructor,
3193       has_trivial_copy_constructor and has_trivial_copy_assign have been
3194       removed.
3195     * The libstdc++ [4]Profile Mode has been deprecated and will be
3196       removed in a future version.
3197     * The Cilk+ extensions to the C and C++ languages have been
3198       deprecated.
3199     * On ARM targets (arm*-*-*), [5]a bug introduced in GCC 5 that
3200       affects conformance to the procedure call standard (AAPCS) has been
3201       fixed. The bug affects some C++ code where class objects are passed
3202       by value to functions and could result in incorrect or inconsistent
3203       code being generated. This is an ABI change. If the option -Wpsabi
3204       is enabled (on by default) the compiler will emit a diagnostic note
3205       for code that might be affected.
3206
3207General Optimizer Improvements
3208
3209     * GCC 7 can determine the return value or range of return values of
3210       some calls to the sprintf family of functions and make it available
3211       to other optimization passes. Some calls to the snprintf function
3212       with a zero size argument can be folded into constants. This
3213       optimization is included in -O1 and can be selectively controlled
3214       by the -fprintf-return-value option.
3215     * A new store merging pass has been added. It merges constant stores
3216       to adjacent memory locations into fewer, wider, stores. It is
3217       enabled by the -fstore-merging option and at the -O2 optimization
3218       level or higher (and -Os).
3219     * A new code hoisting optimization has been added to the partial
3220       redundancy elimination pass. It attempts to move evaluation of
3221       expressions executed on all paths to the function exit as early as
3222       possible. This primarily helps improve code size, but can improve
3223       the speed of the generated code as well. It is enabled by the
3224       -fcode-hoisting option and at the -O2 optimization level or higher
3225       (and -Os).
3226     * A new interprocedural bitwise constant propagation optimization has
3227       been added, which propagates knowledge about which bits of
3228       variables are known to be zero (including pointer alignment
3229       information) across the call graph. It is enabled by the
3230       -fipa-bit-cp option if -fipa-cp is enabled as well, and is enabled
3231       at the -O2 optimization level and higher (and -Os). This
3232       optimization supersedes interprocedural alignment propagation of
3233       GCC 6, and therefore the option -fipa-cp-alignment is now
3234       deprecated and ignored.
3235     * A new interprocedural value range propagation optimization has been
3236       added, which propagates integral range information across the call
3237       graph when variable values can be proven to be within those ranges.
3238       It is enabled by the -fipa-vrp option and at the -O2 optimization
3239       level and higher (and -Os).
3240     * A new loop splitting optimization pass has been added. Certain
3241       loops which contain a condition that is always true on one side of
3242       the iteration space and always false on the other are split into
3243       two loops, such that each of the two new loops iterates on just one
3244       side of the iteration space and the condition does not need to be
3245       checked inside of the loop. It is enabled by the -fsplit-loops
3246       option and at the -O3 optimization level or higher.
3247     * The shrink-wrapping optimization can now separate portions of
3248       prologues and epilogues to improve performance if some of the work
3249       done traditionally by prologues and epilogues is not needed on
3250       certain paths. This is controlled by the -fshrink-wrap-separate
3251       option, enabled by default. It requires target support, which is
3252       currently only implemented in the PowerPC and AArch64 ports.
3253     * AddressSanitizer gained a new sanitization option,
3254       -fsanitize-address-use-after-scope, which enables sanitization of
3255       variables whose address is taken and used after a scope where the
3256       variable is defined:
3257
3258int
3259main (int argc, char **argv)
3260{
3261  char *ptr;
3262    {
3263      char my_char;
3264      ptr = &my_char;
3265    }
3266
3267  *ptr = 123;
3268  return *ptr;
3269}
3270
3271==28882==ERROR: AddressSanitizer: stack-use-after-scope on address 0x7fffb8dba99
32720 at pc 0x0000004006d5 bp 0x7fffb8dba960 sp 0x7fffb8dba958
3273WRITE of size 1 at 0x7fffb8dba990 thread T0
3274    #0 0x4006d4 in main /tmp/use-after-scope-1.c:10
3275    #1 0x7f9c71943290 in __libc_start_main (/lib64/libc.so.6+0x20290)
3276    #2 0x400739 in _start (/tmp/a.out+0x400739)
3277
3278Address 0x7fffb8dba990 is located in stack of thread T0 at offset 32 in frame
3279    #0 0x40067f in main /tmp/use-after-scope-1.c:3
3280
3281  This frame has 1 object(s):
3282    [32, 33) 'my_char' <== Memory access at offset 32 is inside this variable
3283
3284       The option is enabled by default with -fsanitize=address and
3285       disabled by default with -fsanitize=kernel-address. Compared to the
3286       LLVM compiler, where the option already exists, the implementation
3287       in the GCC compiler has some improvements and advantages:
3288          + Complex uses of gotos and case labels are properly handled and
3289            should not report any false positive or false negatives.
3290          + C++ temporaries are sanitized.
3291          + Sanitization can handle invalid memory stores that are
3292            optimized out by the LLVM compiler when optimization is
3293            enabled.
3294     * The -fsanitize=signed-integer-overflow suboption of the
3295       UndefinedBehavior Sanitizer now diagnoses arithmetic overflows even
3296       on arithmetic operations with generic vectors.
3297     * Version 5 of the DWARF debugging information standard is supported
3298       through the -gdwarf-5 option. The DWARF version 4 debugging
3299       information remains the default until consumers of debugging
3300       information are adjusted.
3301
3302New Languages and Language specific improvements
3303
3304   OpenACC support in C, C++, and Fortran continues to be maintained and
3305   improved. See the [6]OpenACC and [7]Offloading wiki pages for further
3306   information.
3307
3308  Ada
3309
3310     * On mainstream native platforms, Ada programs no longer require the
3311       stack to be made executable in order to run properly.
3312
3313  BRIG (HSAIL)
3314
3315   Support for processing BRIG 1.0 files was added in this release. BRIG
3316   is a binary format for HSAIL (Heterogeneous System Architecture
3317   Intermediate Language). The BRIG front end can be used for implementing
3318   HSAIL "finalizers" (compilation of HSAIL to a native ISA) for
3319   GCC-supported targets. An implementation of an HSAIL runtime library,
3320   libhsail-rt is also included.
3321
3322  C family
3323
3324     * New command-line options have been added for the C and C++
3325       compilers:
3326          + -Wimplicit-fallthrough warns when a switch case falls through.
3327            This warning has five different levels. The compiler is able
3328            to parse a wide range of fallthrough comments, depending on
3329            the level. It also handles control-flow statements, such as
3330            ifs. It's possible to suppress the warning by either adding a
3331            fallthrough comment, or by using a null statement:
3332            __attribute__ ((fallthrough)); (C, C++), or [[fallthrough]];
3333            (C++17), or [[gnu::fallthrough]]; (C++11/C++14). This warning
3334            is enabled by -Wextra.
3335          + -Wpointer-compare warns when a pointer is compared with a zero
3336            character constant. Such code is now invalid in C++11 and GCC
3337            rejects it. This warning is enabled by default.
3338          + -Wduplicated-branches warns when an if-else has identical
3339            branches.
3340          + -Wrestrict warns when an argument passed to a
3341            restrict-qualified parameter aliases with another argument.
3342          + -Wmemset-elt-size warns for memset calls, when the first
3343            argument references an array, and the third argument is a
3344            number equal to the number of elements of the array, but not
3345            the size of the array. This warning is enabled by -Wall.
3346          + -Wint-in-bool-context warns about suspicious uses of integer
3347            values where boolean values are expected. This warning is
3348            enabled by -Wall.
3349          + -Wswitch-unreachable warns when a switch statement has
3350            statements between the controlling expression and the first
3351            case label which will never be executed. This warning is
3352            enabled by default.
3353          + -Wexpansion-to-defined warns when defined is used outside #if.
3354            This warning is enabled by -Wextra or -Wpedantic.
3355          + -Wregister warns about uses of the register storage specifier.
3356            In C++17 this keyword has been removed and for C++17 this is a
3357            pedantic warning enabled by default. The warning is not
3358            emitted for the GNU Explicit Register Variables extension.
3359          + -Wvla-larger-than=N warns about unbounded uses of
3360            variable-length arrays, and about bounded uses of
3361            variable-length arrays whose bound can be larger than N bytes.
3362          + -Wduplicate-decl-specifier warns when a declaration has
3363            duplicate const, volatile, restrict or _Atomic specifier. This
3364            warning is enabled by -Wall.
3365     * GCC 6's C and C++ front ends were able to offer suggestions for
3366       misspelled field names:
3367
3368spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
3369you mean 'color'?
3370   return ptr->colour;
3371               ^~~~~~
3372
3373       GCC 7 greatly expands the scope of these suggestions. Firstly, it
3374       adds fix-it hints to such suggestions:
3375
3376spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
3377you mean 'color'?
3378   return ptr->colour;
3379               ^~~~~~
3380               color
3381
3382       The suggestions now cover many other things, such as misspelled
3383       function names:
3384
3385spellcheck-identifiers.c:11:3: warning: implicit declaration of function 'gtk_wi
3386dget_showall'; did you mean 'gtk_widget_show_all'? [-Wimplicit-function-declarat
3387ion]
3388   gtk_widget_showall (w);
3389   ^~~~~~~~~~~~~~~~~~
3390   gtk_widget_show_all
3391
3392       misspelled macro names and enum values:
3393
3394spellcheck-identifiers.cc:85:11: error: 'MAX_ITEM' undeclared here (not in a fun
3395ction); did you mean 'MAX_ITEMS'?
3396 int array[MAX_ITEM];
3397           ^~~~~~~~
3398           MAX_ITEMS
3399
3400       misspelled type names:
3401
3402spellcheck-typenames.c:7:14: error: unknown type name 'singed'; did you mean 'si
3403gned'?
3404 void test (singed char e);
3405            ^~~~~~
3406            signed
3407
3408       and, in the C front end, named initializers:
3409
3410test.c:7:20: error: 'struct s' has no member named 'colour'; did you mean 'color
3411'?
3412 struct s test = { .colour = 3 };
3413                    ^~~~~~
3414                    color
3415
3416     * The preprocessor can now offer suggestions for misspelled
3417       directives, e.g.:
3418
3419test.c:5:2: error:invalid preprocessing directive #endfi; did you mean #endif?
3420 #endfi
3421  ^~~~~
3422  endif
3423
3424     * Warnings about format strings now underline the pertinent part of
3425       the string, and can offer suggested fixes. In some cases, the
3426       pertinent argument is underlined.
3427
3428test.c:51:29: warning: format '%s' expects argument of type 'char *', but argume
3429nt 3 has type 'int' [-Wformat=]
3430   printf ("foo: %d  bar: %s baz: %d", 100, i + j, 102);
3431                          ~^                ~~~~~
3432                          %d
3433
3434     * The new -Wdangling-else command-line option has been split out of
3435       -Wparentheses and warns about dangling else.
3436     * The -Wshadow warning has been split into three variants.
3437       -Wshadow=global warns for any shadowing. This is the default when
3438       using -Wshadow without any argument. -Wshadow=local only warns for
3439       a local variable shadowing another local variable or parameter.
3440       -Wshadow=compatible-local only warns for a local variable shadowing
3441       another local variable or parameter whose type is compatible (in
3442       C++ compatible means that the type of the shadowing variable can be
3443       converted to that of the shadowed variable).
3444       The following example shows the different kinds of shadow warnings:
3445
3446enum operation { add, count };
3447struct container { int nr; };
3448
3449int
3450container_count (struct container c, int count)
3451{
3452  int r = 0;
3453  for (int count = 0; count > 0; count--)
3454    {
3455      struct container count = c;
3456      r += count.nr;
3457    }
3458  return r;
3459}
3460
3461       -Wshadow=compatible-local will warn for the parameter being
3462       shadowed with the same type:
3463
3464warn-test.c:8:12: warning: declaration of 'count' shadows a parameter [-Wshadow=
3465compatible-local]
3466   for (int count = 0; count > 0; count--)
3467            ^~~~~
3468warn-test.c:5:42: note: shadowed declaration is here
3469 container_count (struct container c, int count)
3470                                          ^~~~~
3471
3472       -Wshadow=local will warn for the above and for the shadowed
3473       declaration with incompatible type:
3474
3475warn-test.c:10:24: warning: declaration of 'count' shadows a previous local [-Ws
3476hadow=local]
3477       struct container count = c;
3478                        ^~~~~
3479warn-test.c:8:12: note: shadowed declaration is here
3480   for (int count = 0; count > 0; count--)
3481            ^~~~~
3482
3483       -Wshadow=global will warn for all of the above and the shadowing of
3484       the global declaration:
3485
3486warn-test.c:5:42: warning: declaration of 'count' shadows a global declaration [
3487-Wshadow]
3488 container_count (struct container c, int count)
3489                                          ^~~~~
3490warn-test.c:1:23: note: shadowed declaration is here
3491 enum operation { add, count };
3492                       ^~~~~
3493
3494     * GCC 7 contains a number of enhancements that help detect buffer
3495       overflow and other forms of invalid memory accesses.
3496          + The -Walloc-size-larger-than=size option detects calls to
3497            standard and user-defined memory allocation functions
3498            decorated with attribute alloc_size whose argument exceeds the
3499            specified size (PTRDIFF_MAX by default). The option also
3500            detects arithmetic overflow in the computation of the size in
3501            two-argument allocation functions like calloc where the total
3502            size is the product of the two arguments. Since calls with an
3503            excessive size cannot succeed they are typically the result of
3504            programming errors. Such bugs have been known to be the source
3505            of security vulnerabilities and a target of exploits.
3506            -Walloc-size-larger-than=PTRDIFF_MAX is included in -Wall.
3507            For example, the following call to malloc incorrectly tries to
3508            avoid passing a negative argument to the function and instead
3509            ends up unconditionally invoking it with an argument less than
3510            or equal to zero. Since after conversion to the type of the
3511            argument of the function (size_t) a negative argument results
3512            in a value in excess of the maximum PTRDIFF_MAX the call is
3513            diagnosed.
3514
3515void* f (int n)
3516{
3517  return malloc (n > 0 ? 0 : n);
3518}
3519
3520warning: argument 1 range [2147483648, 4294967295] exceeds maximum object size 2
3521147483647 [-Walloc-size-larger-than=]
3522
3523          + The -Walloc-zero option detects calls to standard and
3524            user-defined memory allocation functions decorated with
3525            attribute alloc_size with a zero argument. -Walloc-zero is not
3526            included in either -Wall or -Wextra and must be explicitly
3527            enabled.
3528          + The -Walloca option detects all calls to the alloca function
3529            in the program. -Walloca is not included in either -Wall or
3530            -Wextra and must be explicitly enabled.
3531          + The -Walloca-larger-than=size option detects calls to the
3532            alloca function whose argument either may exceed the specified
3533            size, or that is not known to be sufficiently constrained to
3534            avoid exceeding it. -Walloca-larger-than is not included in
3535            either -Wall or -Wextra and must be explicitly enabled.
3536            For example, compiling the following snippet with
3537            -Walloca-larger-than=1024 results in a warning because even
3538            though the code appears to call alloca only with sizes of 1kb
3539            and less, since n is signed, a negative value would result in
3540            a call to the function well in excess of the limit.
3541
3542void f (int n)
3543{
3544  char *d;
3545  if (n < 1025)
3546    d = alloca (n);
3547  else
3548    d = malloc (n);
35493550}
3551
3552warning: argument to 'alloca may be too large due to conversion from 'int' to 'l
3553ong unsigned int' [-Walloca-larger-than=]
3554
3555            In contrast, a call to alloca that isn't bounded at all such
3556            as in the following function will elicit the warning below
3557            regardless of the size argument to the option.
3558
3559void f (size_t n)
3560{
3561  char *d = alloca (n);
35623563}
3564
3565warning: unbounded use of 'alloca' [-Walloca-larger-than=]
3566
3567          + The -Wformat-overflow=level option detects certain and likely
3568            buffer overflow in calls to the sprintf family of formatted
3569            output functions. Although the option is enabled even without
3570            optimization it works best with -O2 and higher.
3571            For example, in the following snippet the call to sprintf is
3572            diagnosed because even though its output has been constrained
3573            using the modulo operation it could result in as many as three
3574            bytes if mday were negative. The solution is to either
3575            allocate a larger buffer or make sure the argument is not
3576            negative, for example by changing mday's type to unsigned or
3577            by making the type of the second operand of the modulo
3578            expression unsigned: 100U.
3579
3580void* f (int mday)
3581{
3582  char *buf = malloc (3);
3583  sprintf (buf, "%02i", mday % 100);
3584  return buf;
3585}
3586
3587warning: 'sprintf may write a terminating nul past the end of the destination [-
3588Wformat-overflow=]
3589note: 'sprintf' output between 3 and 4 bytes into a destination of size 3
3590
3591          + The -Wformat-truncation=level option detects certain and
3592            likely output truncation in calls to the snprintf family of
3593            formatted output functions. -Wformat-truncation=1 is included
3594            in -Wall and enabled without optimization but works best with
3595            -O2 and higher.
3596            For example, the following function attempts to format an
3597            integer between 0 and 255 in hexadecimal, including the 0x
3598            prefix, into a buffer of four characters. But since the
3599            function must always terminate output by the null character
3600            ('\0') such a buffer is only big enough to fit just one digit
3601            plus the prefix. Therefore the snprintf call is diagnosed. To
3602            avoid the warning either use a bigger buffer or handle the
3603            function's return value which indicates whether or not its
3604            output has been truncated.
3605
3606void f (unsigned x)
3607{
3608  char d[4];
3609  snprintf (d, sizeof d, "%#02x", x & 0xff);
36103611}
3612
3613warning: 'snprintf' output may be truncated before the last format character [-W
3614format-truncation=]
3615note: 'snprintf' output between 3 and 5 bytes into a destination of size 4
3616
3617          + The -Wnonnull option has been enhanced to detect a broader set
3618            of cases of passing null pointers to functions that expect a
3619            non-null argument (those decorated with attribute nonnull). By
3620            taking advantage of optimizations the option can detect many
3621            more cases of the problem than in prior GCC versions.
3622          + The -Wstringop-overflow=type option detects buffer overflow in
3623            calls to string handling functions like memcpy and strcpy. The
3624            option relies on [8]Object Size Checking and has an effect
3625            similar to defining the _FORTIFY_SOURCE macro.
3626            -Wstringop-overflow=2 is enabled by default.
3627            For example, in the following snippet, because the call to
3628            strncat specifies a maximum that allows the function to write
3629            past the end of the destination, it is diagnosed. To correct
3630            the problem and avoid the overflow the function should be
3631            called with a size of at most sizeof d - strlen(d) - 1.
3632
3633void f (const char *fname)
3634{
3635  char d[8];
3636  strncpy (d, "/tmp/", sizeof d);
3637  strncat (d, fname, sizeof d);
36383639}
3640
3641warning: specified bound 8 equals the size of the destination [-Wstringop-overfl
3642ow=]
3643
3644     * The <limits.h> header provided by GCC defines macros such as
3645       INT_WIDTH for the width in bits of integer types, if
3646       __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is
3647       included. The <stdint.h> header defines such macros as SIZE_WIDTH
3648       and INTMAX_WIDTH for the width of some standard typedef names for
3649       integer types, again if __STDC_WANT_IEC_60559_BFP_EXT__ is defined
3650       before the header is included; note that GCC's implementation of
3651       this header is only used for freestanding compilations, not hosted
3652       compilations, on most systems. These macros come from ISO/IEC TS
3653       18661-1:2014.
3654     * The <float.h> header provided by GCC defines the macro
3655       CR_DECIMAL_DIG, from ISO/IEC TS 18661-1:2014, if
3656       __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is
3657       included. This represents the number of decimal digits for which
3658       conversions between decimal character strings and binary formats,
3659       in both directions, are correctly rounded, and currently has the
3660       value of UINTMAX_MAX on all systems, reflecting that GCC's
3661       compile-time conversions are correctly rounded for any number of
3662       digits.
3663     * New __builtin_add_overflow_p, __builtin_sub_overflow_p,
3664       __builtin_mul_overflow_p built-in functions have been added. These
3665       work similarly to their siblings without the _p suffix, but do not
3666       actually store the result of the arithmetics anywhere, just return
3667       whether the operation would overflow. Calls to these built-ins with
3668       integer constant arguments evaluate to integer constants
3669       expressions.
3670       For example, in the following, c is assigned the result of a * b
3671       only if the multiplication does not overflow, otherwise it is
3672       assigned the value zero. The multiplication is performed at
3673       compile-time and without triggering a -Woverflow warning.
3674
3675enum {
3676  a = 12345678,
3677  b = 87654321,
3678  c = __builtin_mul_overflow_p (a, b, a) ? 0 : a * b
3679};
3680
3681  C
3682
3683     * The C front end now supports type names _FloatN for floating-point
3684       types with IEEE interchange formats and _FloatNx for floating-point
3685       types with IEEE extended formats. These type names come from
3686       ISO/IEC TS 18661-3:2015.
3687       The set of types supported depends on the target for which GCC is
3688       configured. Most targets support _Float32, _Float32x and _Float64.
3689       _Float128 is supported on targets where IEEE binary128 encoding was
3690       already supported as long double or __float128. _Float64x is
3691       supported on targets where a type with either binary128 or Intel
3692       extended precision format is available.
3693       Constants with these types are supported using suffixes fN, FN, fNx
3694       and FNx (e.g., 1.2f128 or 2.3F64x). Macros such as FLT128_MAX are
3695       defined in <float.h> if __STDC_WANT_IEC_60559_TYPES_EXT__ is
3696       defined before it is included.
3697       These new types are always distinct from each other and from float,
3698       double and long double, even if they have the same encoding.
3699       Complex types such as _Complex _Float128 are also supported.
3700       Type-generic built-in functions such as __builtin_isinf support the
3701       new types, and the following type-specific built-in functions have
3702       versions (suffixed fN or fNx) for the new types:
3703       __builtin_copysign, __builtin_fabs, __builtin_huge_val,
3704       __builtin_inf, __builtin_nan, __builtin_nans.
3705     * Compilation with -fopenmp is now compatible with the C11 _Atomic
3706       keyword.
3707
3708  C++
3709
3710     * The C++ front end has experimental support for all of the current
3711       C++17 draft with the -std=c++1z or -std=gnu++1z flags, including if
3712       constexpr, class template argument deduction, auto template
3713       parameters, and structured bindings. For a full list of new
3714       features, see [9]the C++ status page.
3715     * C++17 support for new of over-aligned types can be enabled in other
3716       modes with the -faligned-new flag.
3717     * The C++17 evaluation order requirements can be selected in other
3718       modes with the -fstrong-eval-order flag, or disabled in C++17 mode
3719       with -fno-strong-eval-order.
3720     * The default semantics of inherited constructors has changed in all
3721       modes, following [10]P0136. Essentially, overload resolution
3722       happens as if calling the inherited constructor directly, and the
3723       compiler fills in construction of the other bases and members as
3724       needed. Most uses should not need any changes. The old behavior can
3725       be restored with -fno-new-inheriting-ctors, or -fabi-version less
3726       than 11.
3727     * The resolution of DR 150 on matching of template template
3728       parameters, allowing default template arguments to make a template
3729       match a parameter, is currently enabled by default in C++17 mode
3730       only. The default can be overridden with -f{no-,}new-ttp-matching.
3731     * The C++ front end will now provide fix-it hints for some missing
3732       semicolons, allowing for automatic fixes by IDEs:
3733
3734test.cc:4:11: error: expected ';' after class definition
3735 class a {}
3736           ^
3737           ;
3738
3739     * -Waligned-new has been added to the C++ front end. It warns about
3740       new of type with extended alignment without -faligned-new.
3741
3742    Runtime Library (libstdc++)
3743
3744     * The type of exception thrown by iostreams, std::ios_base::failure,
3745       now uses the [11]cxx11 ABI.
3746     * Experimental support for C++17, including the following new
3747       features:
3748          + std::string_view;
3749          + std::any, std::optional, and std::variant;
3750          + std::invoke, std::is_invocable, std::is_nothrow_invocable, and
3751            invoke_result;
3752          + std::is_swappable, and std::is_nothrow_swappable;
3753          + std::apply, and std::make_from_tuple;
3754          + std::void_t, std::bool_constant, std::conjunction,
3755            std::disjunction, and std::negation;
3756          + Variable templates for type traits;
3757          + Mathematical Special Functions;
3758          + std::chrono::floor, std::chrono::ceil, std::chrono::round, and
3759            std::chrono::abs;
3760          + std::clamp, std::gcd, std::lcm, 3-dimensional std::hypot;
3761          + std::scoped_lock, std::shared_mutex,
3762            std::atomic<T>::is_always_lock_free;
3763          + std::sample, std::default_searcher, std::boyer_moore_searcher
3764            and std::boyer_moore_horspool_searcher;
3765          + Extraction and re-insertion of map and set nodes, try_emplace
3766            members for maps, and functions for accessing containers
3767            std::size, std::empty, and std::data;
3768          + std::shared_ptr support for arrays,
3769            std::shared_ptr<T>::weak_type,
3770            std::enable_shared_from_this<T>::weak_from_this(), and
3771            std::owner_less<void>;
3772          + std::byte;
3773          + std::as_const, std::not_fn,
3774            std::has_unique_object_representations, constexpr
3775            std::addressof.
3776       Thanks to Daniel Krügler, Tim Shen, Edward Smith-Rowland, and Ville
3777       Voutilainen for work on the C++17 support.
3778     * A new power-of-two rehashing policy for use with the _Hashtable
3779       internals, thanks to François Dumont.
3780
3781  Fortran
3782
3783     * Support for a number of extensions for compatibility with legacy
3784       code with new flags:
3785          + -fdec-structure Support for DEC STRUCTURE and UNION
3786          + -fdec-intrinsic-ints Support for new integer intrinsics with
3787            B/I/J/K prefixes such as BABS, JIAND...
3788          + -fdec-math Support for additional math intrinsics, including
3789            COTAN and degree-valued trigonometric functions such as TAND,
3790            ASIND...
3791          + -fdec Enable the -fdec-* family of extensions.
3792     * New flag -finit-derived to allow default initialization of
3793       derived-type variables.
3794     * Improved DO loops with step equal to 1 or -1, generates faster code
3795       without a loop preheader. A new warning, -Wundefined-do-loop, warns
3796       when a loop iterates either to HUGE(i) (with step equal to 1), or
3797       to -HUGE(i) (with step equal to -1). Invalid behavior can be caught
3798       at run time with -fcheck=do enabled:
3799
3800program test
3801  implicit none
3802  integer(1) :: i
3803  do i = -HUGE(i)+10, -HUGE(i)-1, -1
3804    print *, i
3805  end do
3806end program test
3807
3808At line 8 of file do_check_12.f90
3809Fortran runtime error: Loop iterates infinitely
3810
3811     * Version 4.5 of the [12]OpenMP specification is now partially
3812       supported in the Fortran compiler; the largest missing item is
3813       structure element mapping.
3814     * User-defined derived-type input/output (UDTIO) is added.
3815     * Derived type coarrays with allocatable and pointer components are
3816       partially supported.
3817     * Non-constant stop codes and error stop codes (Fortran 2015
3818       feature).
3819     * Derived types with allocatable components of recursive type.
3820     * Intrinsic assignment to polymorphic variables.
3821     * Improved submodule support.
3822     * Improved diagnostics (polymorphic results in pure functions).
3823     * Coarray: Support for failed images (Fortan 2015 feature).
3824
3825  Go
3826
3827     * GCC 7 provides a complete implementation of the Go 1.8.1 user
3828       packages.
3829     * Compared to the Go 1.8.1 toolchain, the garbage collector is more
3830       conservative and less concurrent.
3831     * Escape analysis is available for experimental use via the
3832       -fgo-optimize-allocs option. The -fgo-debug-escape prints
3833       information useful for debugging escape analysis choices.
3834
3835  Java (GCJ)
3836
3837   The GCC Java front end and associated libjava runtime library have been
3838   removed from GCC.
3839
3840libgccjit
3841
3842   The libgccjit API gained support for marking calls as requiring
3843   tail-call optimization via a new entry point:
3844   [13]gcc_jit_rvalue_set_bool_require_tail_call.
3845
3846   libgccjit performs numerous checks at the API boundary, but if these
3847   succeed, it previously ignored errors and other diagnostics emitted
3848   within the core of GCC, and treated the compile of a gcc_jit_context as
3849   having succeeded. As of GCC 7 it now ensures that if any diagnostics
3850   are emitted, they are visible from the libgccjit API, and that the the
3851   context is flagged as having failed.
3852
3853New Targets and Target Specific Improvements
3854
3855  AArch64
3856
3857     * GCC has been updated to the latest revision of the procedure call
3858       standard (AAPCS64) to provide support for parameter passing when
3859       data types have been over-aligned.
3860     * The ARMv8.3-A architecture is now supported. It can be used by
3861       specifying the -march=armv8.3-a option.
3862     * The option -msign-return-address= is supported to enable return
3863       address protection using ARMv8.3-A Pointer Authentication
3864       Extensions. For more information on the arguments accepted by this
3865       option, please refer to [14]AArch64-Options.
3866     * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point
3867       Extensions are now supported. They can be used by specifying the
3868       -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit
3869       Floating-Point Extensions introduce new half-precision data
3870       processing floating-point instructions.
3871     * Support has been added for the following processors (GCC
3872       identifiers in parentheses): ARM Cortex-A73 (cortex-a73), Broadcom
3873       Vulcan (vulcan), Cavium ThunderX CN81xx (thunderxt81), Cavium
3874       ThunderX CN83xx (thunderxt83), Cavium ThunderX CN88xx
3875       (thunderxt88), Cavium ThunderX CN88xx pass 1.x (thunderxt88p1),
3876       Cavium ThunderX 2 CN99xx (thunderx2t99), Qualcomm Falkor (falkor).
3877       The GCC identifiers can be used as arguments to the -mcpu or -mtune
3878       options, for example: -mcpu=cortex-a73 or -mtune=vulcan or as
3879       arguments to the equivalent target attributes and pragmas.
3880
3881  ARC
3882
3883     * Added support for ARC HS and ARC EM processors.
3884     * Added support for ARC EM variation found in Intel QuarkSE SoCs.
3885     * Added support for NPS400 ARC700 based CPUs.
3886     * Thread Local Storage is now supported by ARC CPUs.
3887     * Fixed errors for ARC600 when using 32x16 multiplier option.
3888     * Fixed PIE for ARC CPUs.
3889     * New CPU templates are supported via multilib.
3890
3891  ARM
3892
3893     * Support for the ARMv5 and ARMv5E architectures has been deprecated
3894       (which have no known implementations) and will be removed in a
3895       future GCC release. Note that ARMv5T, ARMv5TE and ARMv5TEJ
3896       architectures remain supported. The values armv5 and armv5e of
3897       -march are thus deprecated.
3898     * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point
3899       Extensions are now supported. They can be used by specifying the
3900       -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit
3901       Floating-Point Extensions introduce new half-precision data
3902       processing floating-point instructions.
3903     * The ARMv8-M architecture is now supported in its two architecture
3904       profiles: ARMv8-M Baseline and ARMv8-M Mainline with its DSP and
3905       Floating-Point Extensions. They can be used by specifying the
3906       -march=armv8-m.base, armv8-m.main or armv8-m.main+dsp options.
3907     * Support has been added for the following processors (GCC
3908       identifiers in parentheses): ARM Cortex-A73 (cortex-a73), ARM
3909       Cortex-M23 (cortex-m23) and ARM Cortex-M33 (cortex-m33). The GCC
3910       identifiers can be used as arguments to the -mcpu or -mtune
3911       options, for example: -mcpu=cortex-a73 or -mtune=cortex-m33.
3912     * A new command-line option -mpure-code has been added. It does not
3913       allow constant data to be placed in code sections. This option is
3914       only available when generating non-PIC code for ARMv7-M targets.
3915     * Support for the ACLE Coprocessor Intrinsics has been added. This
3916       enables the generation of coprocessor instructions through the use
3917       of intrinsics such as cdp, ldc, and others.
3918     * The configure option --with-multilib-list now accepts the value
3919       rmprofile to build multilib libraries for a range of embedded
3920       targets. See our [15]installation instructions for details.
3921
3922  AVR
3923
3924     * On the reduced Tiny cores, the progmem [16]variable attribute is
3925       now properly supported. Respective read-only variables are located
3926       in flash memory in section .progmem.data. No special code is needed
3927       to access such variables; the compiler automatically adds an offset
3928       of 0x4000 to all addresses, which is needed to access variables in
3929       flash memory. As opposed to ordinary cores where it is sufficient
3930       to specify the progmem attribute with definitions, on the reduced
3931       Tiny cores the attribute also has to be specified with (external)
3932       declarations:
3933
3934extern const int array[] __attribute__((__progmem__));
3935
3936int get_value2 (void)
3937{
3938  /* Access via addresses array + 0x4004 and array + 0x4005. */
3939  return array[2];
3940}
3941
3942const int* get_address (unsigned idx)
3943{
3944  /* Returns array + 0x4000 + 2 * idx. */
3945  return &array[idx];
3946}
3947
3948     * A new command-line option -Wmisspelled-isr has been added. It turns
3949       off — or turns into errors — warnings that are reported for
3950       interrupt service routines (ISRs) which don't follow AVR-LibC's
3951       naming convention of prefixing ISR names with __vector.
3952     * __builtin_avr_nops(n) is a new [17]built-in function that inserts n
3953       NOP instructions into the instruction stream. n must be a value
3954       known at compile time.
3955
3956  IA-32/x86-64
3957
3958     * Support for the AVX-512 Fused Multiply Accumulation Packed Single
3959       precision (4FMAPS), AVX-512 Vector Neural Network Instructions Word
3960       variable precision (4VNNIW), AVX-512 Vector Population Count
3961       (VPOPCNTDQ) and Software Guard Extensions (SGX) ISA extensions has
3962       been added.
3963
3964  NVPTX
3965
3966     * OpenMP target regions can now be offloaded to NVidia PTX GPGPUs.
3967       See the [18]Offloading Wiki on how to configure it.
3968
3969  PowerPC / PowerPC64 / RS6000
3970
3971     * The PowerPC port now uses LRA by default.
3972     * GCC now diagnoses inline assembly that clobbers register r2. This
3973       has always been invalid code, and is no longer quietly tolerated.
3974     * The PowerPC port's support for ISA 3.0 (-mcpu=power9) has been
3975       enhanced to generate more of the new instructions by default, and
3976       to provide more built-in functions to generate code for other new
3977       instructions.
3978     * The configuration option --enable-gnu-indirect-function is now
3979       enabled by default on PowerPC GNU/Linux builds.
3980     * The PowerPC port will now allow 64-bit and 32-bit integer types to
3981       be allocated to the VSX vector registers (ISA 2.06 and above). In
3982       addition, on ISA 3.0, 16-bit and 8-bit integer types can be
3983       allocated in the vector registers. Previously, only 64-bit integer
3984       types were allowed in the traditional floating point registers.
3985     * New options -mstack-protector-guard=global,
3986       -mstack-protector-guard=tls, -mstack-protector-guard-reg=, and
3987       -mstack-protector-guard-offset= change how the stack protector gets
3988       the value to use as canary.
3989
3990  S/390, System z, IBM z Systems, IBM Z
3991
3992     * Support for the IBM z14 processor has been added. When using the
3993       -march=z14 option, the compiler will generate code making use of
3994       the new instructions introduced with the vector enhancement
3995       facility and the miscellaneous instruction extension facility 2.
3996       The -mtune=z14 option enables z14 specific instruction scheduling
3997       without making use of new instructions.
3998     * Builtins for the new vector instructions have been added and can be
3999       enabled using the -mzvector option.
4000
4001  RISC-V
4002
4003     * Support for the RISC-V instruction set has been added.
4004
4005  RX
4006
4007   Basic support for atomic built-in function has been added. It is
4008   currently implemented by flipping interrupts off and on as needed.
4009
4010  SH
4011
4012     * Support for SH5/SH64 has been removed.
4013     * Improved utilization of delay slots on SH2A.
4014     * Improved utilization of zero-displacement conditional branches.
4015     * The following deprecated options have been removed
4016          + -mcbranchdi
4017          + -mcmpeqdi
4018          + -minvalid-symbols
4019          + -msoft-atomic
4020          + -mspace
4021          + -madjust-unroll
4022     * Support for the following SH2A instructions has been added
4023          + mov.b @-Rm,R0
4024          + mov.w @-Rm,R0
4025          + mov.l @-Rm,R0
4026          + mov.b R0,@Rn+
4027          + mov.w R0,@Rn+
4028          + mov.l R0,@Rn+
4029
4030  SPARC
4031
4032     * The SPARC port now uses LRA by default.
4033     * Support for the new Subtract-Extended-with-Carry instruction
4034       available in SPARC M7 (Niagara 7) has been added.
4035
4036Operating Systems
4037
4038  AIX
4039
4040     * Visibility support has been enabled for AIX 7.1 and above.
4041
4042  Fuchsia
4043
4044     * Support has been added for the [19]Fuchsia OS.
4045
4046  RTEMS
4047
4048     * The ABI changes on ARM so that no short enums are used by default.
4049
4050Other significant improvements
4051
4052     * -fverbose-asm previously emitted information on the meanings of
4053       assembly expressions. This has been extended so that it now also
4054       prints comments showing the source lines that correspond to the
4055       assembly, making it easier to read the generated assembly
4056       (especially with larger functions). For example, given this C
4057       source file:
4058
4059int test (int n)
4060{
4061  int i;
4062  int total = 0;
4063
4064  for (i = 0; i < n; i++)
4065    total += i * i;
4066  return total;
4067}
4068
4069       -fverbose-asm now gives output similar to this for the function
4070       body (when compiling for x86_64, with -Os):
4071
4072       .text
4073       .globl  test
4074       .type   test, @@function
4075test:
4076.LFB0:
4077       .cfi_startproc
4078# example.c:4:   int total = 0;
4079       xorl    %eax, %eax      # <retval>
4080# example.c:6:   for (i = 0; i < n; i++)
4081       xorl    %edx, %edx      # i
4082.L2:
4083# example.c:6:   for (i = 0; i < n; i++)
4084       cmpl    %edi, %edx      # n, i
4085       jge     .L5     #,
4086# example.c:7:     total += i * i;
4087       movl    %edx, %ecx      # i, tmp92
4088       imull   %edx, %ecx      # i, tmp92
4089# example.c:6:   for (i = 0; i < n; i++)
4090       incl    %edx    # i
4091# example.c:7:     total += i * i;
4092       addl    %ecx, %eax      # tmp92, <retval>
4093       jmp     .L2     #
4094.L5:
4095# example.c:10: }
4096       ret
4097       .cfi_endproc
4098
4099     * Two new options have been added for printing fix-it hints:
4100          + -fdiagnostics-parseable-fixits allows for fix-it hints to be
4101            emitted in a machine-readable form, suitable for consumption
4102            by IDEs. For example, given:
4103
4104spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
4105you mean 'color'?
4106   return ptr->colour;
4107               ^~~~~~
4108               color
4109
4110            it will emit:
4111
4112fix-it:"spellcheck-fields.cc":{52:13-52:19}:"color"
4113
4114          + -fdiagnostics-generate-patch will print a patch in "unified"
4115            format after any diagnostics are printed, showing the result
4116            of applying all fix-it hints. For the above example it would
4117            emit:
4118
4119--- spellcheck-fields.cc
4120+++ spellcheck-fields.cc
4121@@ -49,5 +49,5 @@
4122
4123 color get_color(struct s *ptr)
4124 {
4125-  return ptr->colour;
4126+  return ptr->color;
4127 }
4128
4129     * The gcc and g++ driver programs will now provide suggestions for
4130       misspelled arguments to command-line options.
4131
4132$ gcc -c test.c -ftls-model=global-dinamic
4133gcc: error: unknown TLS model 'global-dinamic'
4134gcc: note: valid arguments to '-ftls-model=' are: global-dynamic initial-exec lo
4135cal-dynamic local-exec; did you mean 'global-dynamic'?
4136
4137     * The compiler will now provide suggestions for misspelled
4138       parameters.
4139
4140$ gcc -c test.c --param max-early-inliner-iteration=3
4141cc1: error: invalid --param name 'max-early-inliner-iteration'; did you mean 'ma
4142x-early-inliner-iterations'?
4143
4144     * Profile-guided optimization (PGO) instrumentation, as well as test
4145       coverage (GCOV), can newly instrument constructors (functions marks
4146       with __attribute__((constructor))), destructors and C++
4147       constructors (and destructors) of classes that are used as the type
4148       of a global variable.
4149     * A new option -fprofile-update=atomic prevents creation of corrupted
4150       profiles created during an instrumentation run (-fprofile=generate)
4151       of an application. The downside of the option is a speed penalty.
4152       Providing -pthread on the command line selects atomic profile
4153       updating (when supported by the target).
4154     * GCC's already extensive testsuite has gained some new capabilities,
4155       to further improve the reliability of the compiler:
4156          + GCC now has an internal unit-testing API and a suite of tests
4157            for programmatic self-testing of subsystems.
4158          + GCC's C front end has been extended so that it can parse dumps
4159            of GCC's internal representations, allowing for DejaGnu tests
4160            that more directly exercise specific optimization passes. This
4161            covers both the [20]GIMPLE representation (for testing
4162            higher-level optimizations) and the [21]RTL representation,
4163            allowing for more direct testing of lower-level details, such
4164            as register allocation and instruction selection.
4165
4166GCC 7.1
4167
4168   This is the [22]list of problem reports (PRs) from GCC's bug tracking
4169   system that are known to be fixed in the 7.1 release. This list might
4170   not be complete (that is, it is possible that some PRs that have been
4171   fixed are not listed here).
4172
4173GCC 7.2
4174
4175   This is the [23]list of problem reports (PRs) from GCC's bug tracking
4176   system that are known to be fixed in the 7.2 release. This list might
4177   not be complete (that is, it is possible that some PRs that have been
4178   fixed are not listed here).
4179
4180  Target Specific Changes
4181
4182    SPARC
4183
4184     * Support for the SPARC M8 processor has been added.
4185     * The switches -mfix-ut700 and -mfix-gr712rc have been added to work
4186       around an erratum in LEON3FT processors.
4187     * Use of the Floating-point Multiply Single to Double (FsMULd)
4188       instruction can now be controlled by the -mfsmuld and -fno-fsmuld
4189       options.
4190
4191  Operating Systems
4192
4193    RTEMS
4194
4195     * The Ada run-time support uses now thread-local storage (TLS).
4196     * Support for RISC-V has been added.
4197     * Support for 64-bit PowerPC using the ELFv2 ABI with 64-bit long
4198       double has been added.
4199
4200GCC 7.3
4201
4202   This is the [24]list of problem reports (PRs) from GCC's bug tracking
4203   system that are known to be fixed in the 7.3 release. This list might
4204   not be complete (that is, it is possible that some PRs that have been
4205   fixed are not listed here).
4206
4207  Target Specific Changes
4208
4209    SPARC
4210
4211     * Workarounds for the four [25]LEON3FT errata GRLIB-TN-0010..0013
4212       have been added. Relevant errata are activated by the target
4213       specific -mfix-ut699, -mfix-ut700 and -mfix-gr712rc switches.
4214
4215  Operating Systems
4216
4217    RTEMS
4218
4219     * Support has been added for Epiphany target.
4220
4221GCC 7.4
4222
4223   This is the [26]list of problem reports (PRs) from GCC's bug tracking
4224   system that are known to be fixed in the 7.4 release. This list might
4225   not be complete (that is, it is possible that some PRs that have been
4226   fixed are not listed here).
4227
4228GCC 7.5
4229
4230   This is the [27]list of problem reports (PRs) from GCC's bug tracking
4231   system that are known to be fixed in the 7.5 release. This list might
4232   not be complete (that is, it is possible that some PRs that have been
4233   fixed are not listed here).
4234
4235
4236    For questions related to the use of GCC, please consult these web
4237    pages and the [28]GCC manuals. If that fails, the
4238    [29]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4239    web pages and the development of GCC are welcome on our developer
4240    list at [30]gcc@gcc.gnu.org. All of [31]our lists have public
4241    archives.
4242
4243   Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and
4244   distribution of this entire article is permitted in any medium,
4245   provided this notice is preserved.
4246
4247   These pages are [33]maintained by the GCC team. Last modified
4248   2021-10-17[34].
4249
4250References
4251
4252   1. http://gcc.gnu.org/gcc-7/porting_to.html
4253   2. http://gcc.gnu.org/onlinedocs/index.html#current
4254   3. https://gcc.gnu.org/wiki/LRAIsDefault
4255   4. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/manual/profile_mode.html
4256   5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77728
4257   6. https://gcc.gnu.org/wiki/OpenACC
4258   7. https://gcc.gnu.org/wiki/Offloading
4259   8. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/Object-Size-Checking.html
4260   9. https://gcc.gnu.org/projects/cxx-status.html#cxx1z
4261  10. https://wg21.link/p0136
4262  11. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/using_dual_abi.html
4263  12. https://www.openmp.org/specifications/
4264  13. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/jit/topics/expressions.html#gcc_jit_rvalue_set_bool_require_tail_call
4265  14. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AArch64-Options.html#AArch64-Options
4266  15. https://gcc.gnu.org/install/configure.html
4267  16. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Variable-Attributes.html
4268  17. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Built-in-Functions.html
4269  18. https://gcc.gnu.org/wiki/Offloading
4270  19. https://fuchsia.googlesource.com/
4271  20. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/GIMPLE-Tests.html
4272  21. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/RTL-Tests.html
4273  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.0
4274  23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.2
4275  24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.3
4276  25. https://www.gaisler.com/index.php/information/app-tech-notes
4277  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.4
4278  27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.5
4279  28. https://gcc.gnu.org/onlinedocs/
4280  29. mailto:gcc-help@gcc.gnu.org
4281  30. mailto:gcc@gcc.gnu.org
4282  31. https://gcc.gnu.org/lists.html
4283  32. https://www.fsf.org/
4284  33. https://gcc.gnu.org/about.html
4285  34. http://validator.w3.org/check/referer
4286======================================================================
4287http://gcc.gnu.org/gcc-6/index.html
4288                              GCC 6 Release Series
4289
4290   (This release series is no longer supported.)
4291
4292   October 26, 2018
4293
4294   The [1]GNU project and the GCC developers are pleased to announce the
4295   release of GCC 6.5.
4296
4297   This release is a bug-fix release, containing fixes for regressions in
4298   GCC 6.4 relative to previous releases of GCC.
4299
4300Release History
4301
4302   GCC 6.5
4303          October 26, 2018 ([2]changes, [3]documentation)
4304
4305   GCC 6.4
4306          July 4, 2017 ([4]changes, [5]documentation)
4307
4308   GCC 6.3
4309          December 21, 2016 ([6]changes, [7]documentation)
4310
4311   GCC 6.2
4312          August 22, 2016 ([8]changes, [9]documentation)
4313
4314   GCC 6.1
4315          April 27, 2016 ([10]changes, [11]documentation)
4316
4317References and Acknowledgements
4318
4319   GCC used to stand for the GNU C Compiler, but since the compiler
4320   supports several other languages aside from C, it now stands for the
4321   GNU Compiler Collection.
4322
4323   A list of [12]successful builds is updated as new information becomes
4324   available.
4325
4326   The GCC developers would like to thank the numerous people that have
4327   contributed new features, improvements, bug fixes, and other changes as
4328   well as test results to GCC. This [13]amazing group of volunteers is
4329   what makes GCC successful.
4330
4331   For additional information about GCC please refer to the [14]GCC
4332   project web site or contact the [15]GCC development mailing list.
4333
4334   To obtain GCC please use [16]our mirror sites or [17]our version
4335   control system.
4336
4337
4338    For questions related to the use of GCC, please consult these web
4339    pages and the [18]GCC manuals. If that fails, the
4340    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4341    web pages and the development of GCC are welcome on our developer
4342    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
4343    archives.
4344
4345   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
4346   distribution of this entire article is permitted in any medium,
4347   provided this notice is preserved.
4348
4349   These pages are [23]maintained by the GCC team. Last modified
4350   2021-07-28[24].
4351
4352References
4353
4354   1. http://www.gnu.org/
4355   2. http://gcc.gnu.org/gcc-6/changes.html
4356   3. http://gcc.gnu.org/onlinedocs/6.5.0/
4357   4. http://gcc.gnu.org/gcc-6/changes.html
4358   5. http://gcc.gnu.org/onlinedocs/6.4.0/
4359   6. http://gcc.gnu.org/gcc-6/changes.html
4360   7. http://gcc.gnu.org/onlinedocs/6.3.0/
4361   8. http://gcc.gnu.org/gcc-6/changes.html
4362   9. http://gcc.gnu.org/onlinedocs/6.2.0/
4363  10. http://gcc.gnu.org/gcc-6/changes.html
4364  11. http://gcc.gnu.org/onlinedocs/6.1.0/
4365  12. http://gcc.gnu.org/gcc-6/buildstat.html
4366  13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
4367  14. http://gcc.gnu.org/index.html
4368  15. mailto:gcc@gcc.gnu.org
4369  16. http://gcc.gnu.org/mirrors.html
4370  17. http://gcc.gnu.org/git.html
4371  18. https://gcc.gnu.org/onlinedocs/
4372  19. mailto:gcc-help@gcc.gnu.org
4373  20. mailto:gcc@gcc.gnu.org
4374  21. https://gcc.gnu.org/lists.html
4375  22. https://www.fsf.org/
4376  23. https://gcc.gnu.org/about.html
4377  24. http://validator.w3.org/check/referer
4378======================================================================
4379http://gcc.gnu.org/gcc-6/changes.html
4380                              GCC 6 Release Series
4381                        Changes, New Features, and Fixes
4382
4383   This page is a brief summary of some of the huge number of improvements
4384   in GCC 6. For more information, see the [1]Porting to GCC 6 page and
4385   the [2]full GCC documentation.
4386
4387Caveats
4388
4389     * The default mode for C++ is now -std=gnu++14 instead of
4390       -std=gnu++98.
4391     * Support for a number of older systems and recently unmaintained or
4392       untested target ports of GCC has been declared obsolete in GCC 6.
4393       Unless there is activity to revive them, the next release of GCC
4394       will have their sources permanently removed.
4395       The following ports for individual systems on particular
4396       architectures have been obsoleted:
4397          + SH5 / SH64 (sh64-*-*) as announced [3]here.
4398     * The AVR port requires binutils version 2.26.1 or later for the fix
4399       for [4]PR71151 to work.
4400     * The GCC 6.5 release has an accidental ABI incompatibility for
4401       nested std::pair objects, for more details see [5]PR 87822. The bug
4402       causes a layout change for pairs where the first member is also a
4403       pair, e.g. std::pair<std::pair<X, Y>, Z>. The GCC 6 release series
4404       is closed so the bug in GCC 6.5 will not be fixed upstream, but
4405       there is a patch in the bug report to allow it to be fixed by
4406       anybody packaging GCC 6.5 or installing it themselves.
4407
4408General Optimizer Improvements
4409
4410     * UndefinedBehaviorSanitizer gained a new sanitization option,
4411       -fsanitize=bounds-strict, which enables strict checking of array
4412       bounds. In particular, it enables -fsanitize=bounds as well as
4413       instrumentation of flexible array member-like arrays.
4414     * Type-based alias analysis now disambiguates accesses to different
4415       pointers. This improves precision of the alias oracle by about
4416       20-30% on higher-level C++ programs. Programs doing invalid type
4417       punning of pointer types may now need -fno-strict-aliasing to work
4418       correctly.
4419     * Alias analysis now correctly supports the weakref and alias
4420       attributes. This allows accessing both a variable and its alias in
4421       one translation unit which is common with link-time optimization.
4422     * Value range propagation now assumes that the this pointer in C++
4423       member functions is non-null. This eliminates common null pointer
4424       checks but also breaks some non-conforming code-bases (such as
4425       Qt-5, Chromium, KDevelop). As a temporary work-around
4426       -fno-delete-null-pointer-checks can be used. Wrong code can be
4427       identified by using -fsanitize=undefined.
4428     * Link-time optimization improvements:
4429          + warning and error attributes are now correctly preserved by
4430            declaration linking and thus -D_FORTIFY_SOURCE=2 is now
4431            supported with -flto.
4432          + Type merging was fixed to handle C and Fortran
4433            interoperability rules as defined by the Fortran 2008 language
4434            standard.
4435            As an exception, CHARACTER(KIND=C_CHAR) is not inter-operable
4436            with char in all cases because it is an array while char is
4437            scalar. INTEGER(KIND=C_SIGNED_CHAR) should be used instead. In
4438            general, this inter-operability cannot be implemented, for
4439            example on targets where the argument passing convention for
4440            arrays differs from scalars.
4441          + More type information is now preserved at link time, reducing
4442            the loss of accuracy of the type-based alias analysis compared
4443            to builds without link-time optimization.
4444          + Invalid type punning on global variables and declarations is
4445            now reported with -Wodr-type-mismatch.
4446          + The size of LTO object files was reduced by about 11%
4447            (measured by compiling Firefox 46.0).
4448          + Link-time parallelization (enabled using -flto=n) was
4449            significantly improved by decreasing the size of streamed data
4450            when partitioning programs. The size of streamed IL while
4451            compiling Firefox 46.0 was reduced by 66%.
4452          + The linker plugin was extended to pass information about the
4453            type of binary produced to the GCC back end. (That can also be
4454            controlled manually by -flinker-output.) This makes it
4455            possible to properly configure the code generator and support
4456            incremental linking. Incremental linking of LTO objects by gcc
4457            -r is now supported for plugin-enabled setups.
4458            There are two ways to perform incremental linking:
4459              1. Linking by ld -r will result in an object file with all
4460                 sections from individual object files mechanically
4461                 merged. This delays the actual link-time optimization to
4462                 the final linking step and thus permits whole program
4463                 optimization. Linking the final binary with such object
4464                 files is however slower.
4465              2. Linking by gcc -r will lead to link-time optimization and
4466                 emit the final binary into the object file. Linking such
4467                 an object file is fast but avoids any benefits from whole
4468                 program optimization.
4469            GCC 7 will support incremental link-time optimization with gcc
4470            -r.
4471     * Inter-procedural optimization improvements:
4472          + Basic jump threading is now performed before profile
4473            construction and inline analysis, resulting in more realistic
4474            size and time estimates that drive the heuristics of the
4475            inliner and function cloning passes.
4476          + Function cloning now more aggressively eliminates unused
4477            function parameters.
4478
4479New Languages and Language specific improvements
4480
4481   Compared to GCC 5, the GCC 6 release series includes a much improved
4482   implementation of the [6]OpenACC 2.0a specification. Highlights are:
4483     * In addition to single-threaded host-fallback execution, offloading
4484       is supported for nvptx (Nvidia GPUs) on x86_64 and PowerPC 64-bit
4485       little-endian GNU/Linux host systems. For nvptx offloading, with
4486       the OpenACC parallel construct, the execution model allows for an
4487       arbitrary number of gangs, up to 32 workers, and 32 vectors.
4488     * Initial support for parallelized execution of OpenACC kernels
4489       constructs:
4490          + Parallelization of a kernels region is switched on by
4491            -fopenacc combined with -O2 or higher.
4492          + Code is offloaded onto multiple gangs, but executes with just
4493            one worker, and a vector length of 1.
4494          + Directives inside a kernels region are not supported.
4495          + Loops with reductions can be parallelized.
4496          + Only kernels regions with one loop nest are parallelized.
4497          + Only the outer-most loop of a loop nest can be parallelized.
4498          + Loop nests containing sibling loops are not parallelized.
4499       Typically, using the OpenACC parallel construct gives much better
4500       performance, compared to the initial support of the OpenACC kernels
4501       construct.
4502     * The device_type clause is not supported. The bind and nohost
4503       clauses are not supported. The host_data directive is not supported
4504       in Fortran.
4505     * Nested parallelism (cf. CUDA dynamic parallelism) is not supported.
4506     * Usage of OpenACC constructs inside multithreaded contexts (such as
4507       created by OpenMP, or pthread programming) is not supported.
4508     * If a call to the acc_on_device function has a compile-time constant
4509       argument, the function call evaluates to a compile-time constant
4510       value only for C and C++ but not for Fortran.
4511
4512   See the [7]OpenACC and [8]Offloading wiki pages for further
4513   information.
4514
4515  C family
4516
4517     * Version 4.5 of the [9]OpenMP specification is now supported in the
4518       C and C++ compilers.
4519     * The C and C++ compilers now support attributes on enumerators. For
4520       instance, it is now possible to mark enumerators as deprecated:
4521
4522enum {
4523  newval,
4524  oldval __attribute__ ((deprecated ("too old")))
4525};
4526
4527     * Source locations for the C and C++ compilers are now tracked as
4528       ranges, rather than just points, making it easier to identify the
4529       subexpression of interest within a complicated expression. For
4530       example:
4531
4532test.cc: In function 'int test(int, int, foo, int, int)':
4533test.cc:5:16: error: no match for 'operator*' (operand types are 'int' and 'foo'
4534)
4535   return p + q * r * s + t;
4536              ~~^~~
4537
4538       In addition, there is now initial support for precise diagnostic
4539       locations within strings:
4540
4541format-strings.c:3:14: warning: field width specifier '*' expects a matching 'in
4542t' argument [-Wformat=]
4543   printf("%*d");
4544            ^
4545
4546     * Diagnostics can now contain "fix-it hints", which are displayed in
4547       context underneath the relevant source code. For example:
4548
4549fixits.c: In function 'bad_deref':
4550fixits.c:11:13: error: 'ptr' is a pointer; did you mean to use '->'?
4551   return ptr.x;
4552             ^
4553             ->
4554
4555     * The C and C++ compilers now offer suggestions for misspelled field
4556       names:
4557
4558spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
4559you mean 'color'?
4560   return ptr->colour;
4561               ^~~~~~
4562
4563     * New command-line options have been added for the C and C++
4564       compilers:
4565          + -Wshift-negative-value warns about left shifting a negative
4566            value.
4567          + -Wshift-overflow warns about left shift overflows. This
4568            warning is enabled by default. -Wshift-overflow=2 also warns
4569            about left-shifting 1 into the sign bit.
4570          + -Wtautological-compare warns if a self-comparison always
4571            evaluates to true or false. This warning is enabled by -Wall.
4572          + -Wnull-dereference warns if the compiler detects paths that
4573            trigger erroneous or undefined behavior due to dereferencing a
4574            null pointer. This option is only active when
4575            -fdelete-null-pointer-checks is active, which is enabled by
4576            optimizations in most targets. The precision of the warnings
4577            depends on the optimization options used.
4578          + -Wduplicated-cond warns about duplicated conditions in an
4579            if-else-if chain.
4580          + -Wmisleading-indentation warns about places where the
4581            indentation of the code gives a misleading idea of the block
4582            structure of the code to a human reader. For example, given
4583            [10]CVE-2014-1266:
4584
4585sslKeyExchange.c: In function 'SSLVerifySignedServerKeyExchange':
4586sslKeyExchange.c:629:3: warning: this 'if' clause does not guard... [-Wmisleadin
4587g-indentation]
4588    if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0)
4589    ^~
4590sslKeyExchange.c:631:5: note: ...this statement, but the latter is misleadingly
4591indented as if it is guarded by the 'if'
4592        goto fail;
4593        ^~~~
4594
4595            This warning is enabled by -Wall.
4596     * The C and C++ compilers now emit saner error messages if
4597       merge-conflict markers are present in a source file.
4598
4599test.c:3:1: error: version control conflict marker in file
4600 <<<<<<< HEAD
4601 ^~~~~~~
4602
4603  C
4604
4605     * It is possible to disable warnings when an initialized field of a
4606       structure or a union with side effects is being overridden when
4607       using designated initializers via a new warning option
4608       -Woverride-init-side-effects.
4609     * A new type attribute scalar_storage_order applying to structures
4610       and unions has been introduced. It specifies the storage order (aka
4611       endianness) in memory of scalar fields in structures or unions.
4612
4613  C++
4614
4615     * The default mode has been changed to -std=gnu++14.
4616     * [11]C++ Concepts are now supported when compiling with -fconcepts.
4617     * -flifetime-dse is more aggressive in dead-store elimination in
4618       situations where a memory store to a location precedes a
4619       constructor to that memory location.
4620     * G++ now supports [12]C++17 fold expressions, u8 character literals,
4621       extended static_assert, and nested namespace definitions.
4622     * G++ now allows constant evaluation for all non-type template
4623       arguments.
4624     * G++ now supports C++ Transactional Memory when compiling with
4625       -fgnu-tm.
4626
4627    Runtime Library (libstdc++)
4628
4629     * Extensions to the C++ Library to support mathematical special
4630       functions (ISO/IEC 29124:2010), thanks to Edward Smith-Rowland.
4631     * Experimental support for C++17, including the following new
4632       features:
4633          + std::uncaught_exceptions function (this is also available for
4634            -std=gnu++NN modes);
4635          + new member functions try_emplace and insert_or_assign for
4636            unique_key maps;
4637          + non-member functions std::size, std::empty, and std::data for
4638            accessing containers and arrays;
4639          + std::invoke;
4640          + std::shared_mutex;
4641          + std::void_t and std::bool_constant metaprogramming utilities.
4642       Thanks to Ville Voutilainen for contributing many of the C++17
4643       features.
4644     * An experimental implementation of the File System TS.
4645     * Experimental support for most features of the second version of the
4646       Library Fundamentals TS. This includes polymorphic memory resources
4647       and array support in shared_ptr, thanks to Fan You.
4648     * Some assertions checked by Debug Mode can now also be enabled by
4649       _GLIBCXX_ASSERTIONS. The subset of checks enabled by the new macro
4650       have less run-time overhead than the full _GLIBCXX_DEBUG checks and
4651       don't affect the library ABI, so can be enabled per-translation
4652       unit.
4653     * Timed mutex types are supported on more targets, including Darwin.
4654     * Improved std::locale support for DragonFly and FreeBSD, thanks to
4655       John Marino and Andreas Tobler.
4656
4657  Fortran
4658
4659     * Fortran 2008 SUBMODULE support.
4660     * Fortran 2015 EVENT_TYPE, EVENT_POST, EVENT_WAIT, and EVENT_QUERY
4661       support.
4662     * Improved support for Fortran 2003 deferred-length character
4663       variables.
4664     * Improved support for OpenMP and OpenACC.
4665     * The MATMUL intrinsic is now inlined for straightforward cases if
4666       front-end optimization is active. The maximum size for inlining can
4667       be set to n with the -finline-matmul-limit=n option and turned off
4668       with -finline-matmul-limit=0.
4669     * The -Wconversion-extra option will warn about REAL constants which
4670       have excess precision for their kind.
4671     * The -Winteger-division option has been added, which warns about
4672       divisions of integer constants which are truncated. This option is
4673       included in -Wall by default.
4674
4675libgccjit
4676
4677     * The driver code is now run in-process within libgccjit, providing a
4678       small speed-up of the compilation process.
4679     * The API has gained entrypoints for
4680          + [13]timing how long was spent in different parts of code,
4681          + [14]creating switch statements,
4682          + [15]allowing unreachable basic blocks in a function, and
4683          + [16]adding arbitrary command-line options to a compilation.
4684
4685New Targets and Target Specific Improvements
4686
4687  AArch64
4688
4689     * A number of AArch64-specific options have been added. The most
4690       important ones are summarised in this section; for more detailed
4691       information please refer to the documentation.
4692     * The command-line options -march=native, -mcpu=native and
4693       -mtune=native are now available on native AArch64 GNU/Linux
4694       systems. Specifying these options causes GCC to auto-detect the
4695       host CPU and choose the optimal setting for that system.
4696     * -fpic is now supported when generating code for the small code
4697       model (-mcmodel=small). The size of the global offset table (GOT)
4698       is limited to 28KiB under the LP64 SysV ABI, and 15KiB under the
4699       ILP32 SysV ABI.
4700     * The AArch64 port now supports target attributes and pragmas. Please
4701       refer to the [17]documentation for details of available attributes
4702       and pragmas as well as usage instructions.
4703     * Link-time optimization across translation units with different
4704       target-specific options is now supported.
4705     * The option -mtls-size= is now supported. It can be used to specify
4706       the bit size of TLS offsets, allowing GCC to generate better TLS
4707       instruction sequences.
4708     * The option -fno-plt is now fully functional.
4709     * The ARMv8.1-A architecture and the Large System Extensions are now
4710       supported. They can be used by specifying the -march=armv8.1-a
4711       option. Additionally, the +lse option extension can be used in a
4712       similar fashion to other option extensions. The Large System
4713       Extensions introduce new instructions that are used in the
4714       implementation of atomic operations.
4715     * The ACLE half-precision floating-point type __fp16 is now supported
4716       in the C and C++ languages.
4717     * The ARM Cortex-A35 processor is now supported via the
4718       -mcpu=cortex-a35 and -mtune=cortex-a35 options as well as the
4719       equivalent target attributes and pragmas.
4720     * The Qualcomm QDF24xx processor is now supported via the
4721       -mcpu=qdf24xx and -mtune=qdf24xx options as well as the equivalent
4722       target attributes and pragmas.
4723     * Code generation for the ARM Cortex-A57 processor is improved. Among
4724       general code generation improvements, a better algorithm is added
4725       for allocating registers to floating-point multiply-accumulate
4726       instructions offering increased performance when compiling with
4727       -mcpu=cortex-a57 or -mtune=cortex-a57.
4728     * Code generation for the ARM Cortex-A53 processor is improved. A
4729       more accurate instruction scheduling model for the processor is now
4730       used, and a number of compiler tuning parameters have been set to
4731       offer increased performance when compiling with -mcpu=cortex-a53 or
4732       -mtune=cortex-a53.
4733     * Code generation for the Samsung Exynos M1 processor is improved. A
4734       more accurate instruction scheduling model for the processor is now
4735       used, and a number of compiler tuning parameters have been set to
4736       offer increased performance when compiling with -mcpu=exynos-m1 or
4737       -mtune=exynos-m1.
4738     * Improvements in the generation of conditional branches and literal
4739       pools allow the compiler to compile functions of a large size.
4740       Constant pools are now placed into separate rodata sections. The
4741       new option -mpc-relative-literal-loads generates per-function
4742       literal pools, limiting the maximum size of functions to 1MiB.
4743     * Several correctness issues generating Advanced SIMD instructions
4744       for big-endian targets have been fixed resulting in improved code
4745       generation for ACLE intrinsics with -mbig-endian.
4746
4747  ARM
4748
4749     * Support for revisions of the ARM architecture prior to ARMv4t has
4750       been deprecated and will be removed in a future GCC release. The
4751       -mcpu and -mtune values that are deprecated are: arm2, arm250,
4752       arm3, arm6, arm60, arm600, arm610, arm620, arm7, arm7d, arm7di,
4753       arm70, arm700, arm700i, arm710, arm720, arm710c, arm7100, arm7500,
4754       arm7500fe, arm7m, arm7dm, arm7dmi, arm8, arm810, strongarm,
4755       strongarm110, strongarm1100, strongarm1110, fa526, fa626. The value
4756       arm7tdmi is still supported. The values of -march that are
4757       deprecated are: armv2,armv2a,armv3,armv3m,armv4.
4758     * The ARM port now supports target attributes and pragmas. Please
4759       refer to the [18]documentation for details of available attributes
4760       and pragmas as well as usage instructions.
4761     * Support has been added for the following processors (GCC
4762       identifiers in parentheses): ARM Cortex-A32 (cortex-a32), ARM
4763       Cortex-A35 (cortex-a35) and ARM Cortex-R8 (cortex-r8). The GCC
4764       identifiers can be used as arguments to the -mcpu or -mtune
4765       options, for example: -mcpu=cortex-a32 or -mtune=cortex-a35.
4766
4767  Heterogeneous Systems Architecture
4768
4769     * GCC can now generate HSAIL (Heterogeneous System Architecture
4770       Intermediate Language) for simple OpenMP device constructs if
4771       configured with --enable-offload-targets=hsa. A new libgomp plugin
4772       then runs the HSA GPU kernels implementing these constructs on HSA
4773       capable GPUs via a standard HSA run time.
4774       If the HSA compilation back end determines it cannot output HSAIL
4775       for a particular input, it gives a warning by default. These
4776       warnings can be suppressed with -Wno-hsa. To give a few examples,
4777       the HSA back end does not implement compilation of code using
4778       function pointers, automatic allocation of variable sized arrays,
4779       functions with variadic arguments as well as a number of other less
4780       common programming constructs.
4781       When compilation for HSA is enabled, the compiler attempts to
4782       compile composite OpenMP constructs
4783
4784#pragma omp target teams distribute parallel for
4785
4786       into parallel HSA GPU kernels.
4787
4788  IA-32/x86-64
4789
4790     * GCC now supports the Intel CPU named Skylake with AVX-512
4791       extensions through -march=skylake-avx512. The switch enables the
4792       following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW,
4793       AVX-512DQ.
4794     * Support for new AMD instructions monitorx and mwaitx has been
4795       added. This includes new intrinsic and built-in support. It is
4796       enabled through option -mmwaitx. The instructions monitorx and
4797       mwaitx implement the same functionality as the old monitor and
4798       mwait instructions. In addition mwaitx adds a configurable timer.
4799       The timer value is received as third argument and stored in
4800       register %ebx.
4801     * x86-64 targets now allow stack realignment from a word-aligned
4802       stack pointer using the command-line option -mstackrealign or
4803       __attribute__ ((force_align_arg_pointer)). This allows functions
4804       compiled with a vector-aligned stack to be invoked from objects
4805       that keep only word-alignment.
4806     * Support for address spaces __seg_fs, __seg_gs, and __seg_tls. These
4807       can be used to access data via the %fs and %gs segments without
4808       having to resort to inline assembly. Please refer to the
4809       [19]documentation for usage instructions.
4810     * Support for AMD Zen (family 17h) processors is now available
4811       through the -march=znver1 and -mtune=znver1 options.
4812
4813  MeP
4814
4815     * Support for the MeP (mep-elf) architecture has been deprecated and
4816       will be removed in a future GCC release.
4817
4818  MSP430
4819
4820     * The MSP430 compiler now has the ability to automatically distribute
4821       code and data between low memory (addresses below 64K) and high
4822       memory. This only applies to parts that actually have both memory
4823       regions and only if the linker script for the part has been
4824       specifically set up to support this feature.
4825       A new attribute of either can be applied to both functions and
4826       data, and this tells the compiler to place the object into low
4827       memory if there is room and into high memory otherwise. Two other
4828       new attributes - lower and upper - can be used to explicitly state
4829       that an object should be placed in the specified memory region. If
4830       there is not enough left in that region the compilation will fail.
4831       Two new command-line options - -mcode-region=[lower|upper|either]
4832       and -mdata-region=[lower|upper|either] - can be used to tell the
4833       compiler what to do with objects that do not have one of these new
4834       attributes.
4835
4836  PowerPC / PowerPC64 / RS6000
4837
4838     * PowerPC64 now supports IEEE 128-bit floating-point using the
4839       __float128 data type. In GCC 6, this is not enabled by default, but
4840       you can enable it with -mfloat128. The IEEE 128-bit floating-point
4841       support requires the use of the VSX instruction set. IEEE 128-bit
4842       floating-point values are passed and returned as a single vector
4843       value. The software emulator for IEEE 128-bit floating-point
4844       support is only built on PowerPC GNU/Linux systems where the
4845       default CPU is at least power7. On future ISA 3.0 systems (POWER 9
4846       and later), you will be able to use the -mfloat128-hardware option
4847       to use the ISA 3.0 instructions that support IEEE 128-bit
4848       floating-point. An additional type (__ibm128) has been added to
4849       refer to the IBM extended double type that normally implements long
4850       double. This will allow for a future transition to implementing
4851       long double with IEEE 128-bit floating-point.
4852     * Basic support has been added for POWER9 hardware that will use the
4853       recently published OpenPOWER ISA 3.0 instructions. The following
4854       new switches are available:
4855          + -mcpu=power9: Implement all of the ISA 3.0 instructions
4856            supported by the compiler.
4857          + -mtune=power9: In the future, apply tuning for POWER9 systems.
4858            Currently, POWER8 tunings are used.
4859          + -mmodulo: Generate code using the ISA 3.0 integer instructions
4860            (modulus, count trailing zeros, array index support, integer
4861            multiply/add).
4862          + -mpower9-fusion: Generate code to suitably fuse instruction
4863            sequences for a POWER9 system.
4864          + -mpower9-dform: Generate code to use the new D-form
4865            (register+offset) memory instructions for the vector
4866            registers.
4867          + -mpower9-vector: Generate code using the new ISA 3.0 vector
4868            (VSX or Altivec) instructions.
4869          + -mpower9-minmax: Reserved for future development.
4870          + -mtoc-fusion: Keep TOC entries together to provide more fusion
4871            opportunities.
4872     * New constraints have been added to support IEEE 128-bit
4873       floating-point and ISA 3.0 instructions:
4874          + wb: Altivec register if -mpower9-dform is enabled.
4875          + we: VSX register if -mpower9-vector is enabled for 64-bit code
4876            generation.
4877          + wo: VSX register if -mpower9-vector is enabled.
4878          + wp: Reserved for future use if long double is implemented with
4879            IEEE 128-bit floating-point instead of IBM extended double.
4880          + wq: VSX register if -mfloat128 is enabled.
4881          + wF: Memory operand suitable for POWER9 fusion load/store.
4882          + wG: Memory operand suitable for TOC fusion memory references.
4883          + wL: Integer constant identifying the element number mfvsrld
4884            accesses within a vector.
4885     * Support has been added for __builtin_cpu_is() and
4886       __builtin_cpu_supports(), allowing for very fast access to
4887       AT_PLATFORM, AT_HWCAP, and AT_HWCAP2 values. This requires use of
4888       glibc 2.23 or later.
4889     * All hardware transactional memory builtins now correctly behave as
4890       memory barriers. Programmers can use #ifdef __TM_FENCE__ to
4891       determine whether their "old" compiler treats the builtins as
4892       barriers.
4893     * Split-stack support has been added for gccgo on PowerPC64 for both
4894       big- and little-endian (but not for 32-bit). The gold linker from
4895       at least binutils 2.25.1 must be available in the PATH when
4896       configuring and building gccgo to enable split stack. (The
4897       requirement for binutils 2.25.1 applies to PowerPC64 only.) The
4898       split-stack feature allows a small initial stack size to be
4899       allocated for each goroutine, which increases as needed.
4900     * GCC on PowerPC now supports the standard lround function.
4901     * A new configuration option ---with-advance-toolchain=at was added
4902       for PowerPC 64-bit GNU/Linux systems to use the header files,
4903       library files, and the dynamic linker from a specific Advance
4904       Toolchain release instead of the default versions that are provided
4905       by the GNU/Linux distribution. In general, this option is intended
4906       for the developers of GCC, and it is not intended for general use.
4907     * The "q", "S", "T", and "t" asm-constraints have been removed.
4908     * The "b", "B", "m", "M", and "W" format modifiers have been removed.
4909
4910  S/390, System z, IBM z Systems
4911
4912     * Support for the IBM z13 processor has been added. When using the
4913       -march=z13 option, the compiler will generate code making use of
4914       the new instructions and registers introduced with the vector
4915       extension facility. The -mtune=z13 option enables z13 specific
4916       instruction scheduling without making use of new instructions.
4917       Compiling code with -march=z13 reduces the default alignment of
4918       vector types bigger than 8 bytes to 8. This is an ABI change and
4919       care must be taken when linking modules compiled with different
4920       arch levels which interchange variables containing vector type
4921       values. For newly compiled code the GNU linker will emit a warning.
4922     * The -mzvector option enables a C/C++ language extension. This
4923       extension provides a new keyword vector which can be used to define
4924       vector type variables. (Note: This is not available when enforcing
4925       strict standard compliance e.g. with -std=c99. Either enable GNU
4926       extensions with e.g. -std=gnu99 or use __vector instead of vector.)
4927       Additionally a set of overloaded builtins is provided which is
4928       partially compatible to the PowerPC Altivec builtins. In order to
4929       make use of these builtins the vecintrin.h header file needs to be
4930       included.
4931     * The new command-line options -march=native, and -mtune=native are
4932       now available on native IBM z Systems. Specifying these options
4933       causes GCC to auto-detect the host CPU and choose the optimal
4934       setting for that system. If GCC is unable to detect the host CPU
4935       these options have no effect.
4936     * The IBM z Systems port now supports target attributes and pragmas.
4937       Please refer to the [20]documentation for details of available
4938       attributes and pragmas as well as usage instructions.
4939     * -fsplit-stack is now supported as part of the IBM z Systems port.
4940       This feature requires a recent gold linker to be used.
4941     * Support for the g5 and g6 -march=/-mtune= CPU level switches has
4942       been deprecated and will be removed in a future GCC release. -m31
4943       from now on defaults to -march=z900 if not specified otherwise.
4944       -march=native on a g5/g6 machine will default to -march=z900.
4945
4946  SH
4947
4948     * Support for SH5 / SH64 has been declared obsolete and will be
4949       removed in future releases.
4950     * Support for the FDPIC ABI has been added. It can be enabled using
4951       the new -mfdpic target option and --enable-fdpic configure option.
4952
4953  SPARC
4954
4955     * An ABI bug has been fixed in 64-bit mode. Unfortunately, this
4956       change will break binary compatibility with earlier releases for
4957       code it affects, but this should be pretty rare in practice. The
4958       conditions are: a 16-byte structure containing a double or a 8-byte
4959       vector in the second half is passed to a subprogram in slot #15,
4960       for example as 16th parameter if the first 15 ones have at most 8
4961       bytes. The double or vector was wrongly passed in floating-point
4962       register %d32 in lieu of on the stack as per the SPARC calling
4963       conventions.
4964
4965Operating Systems
4966
4967  AIX
4968
4969     * DWARF debugging support for AIX 7.1 has been enabled as an optional
4970       debugging format. A more recent Technology Level (TL) and GCC built
4971       with that level are required for full exploitation of DWARF
4972       debugging capabilities.
4973
4974  Linux
4975
4976     * Support for the [21]musl C library was added for the AArch64, ARM,
4977       MicroBlaze, MIPS, MIPS64, PowerPC, PowerPC64, SH, i386, x32 and
4978       x86_64 targets. It can be selected using the new -mmusl option in
4979       case musl is not the default libc. GCC defaults to musl libc if it
4980       is built with a target triplet matching the *-linux-musl* pattern.
4981
4982  RTEMS
4983
4984     * The RTEMS thread model implementation changed. Mutexes now use
4985       self-contained objects defined in Newlib <sys/lock.h> instead of
4986       Classic API semaphores. The keys for thread specific data and the
4987       once function are directly defined via <pthread.h>. Self-contained
4988       condition variables are provided via Newlib <sys/lock.h>. The RTEMS
4989       thread model also supports C++11 threads.
4990     * OpenMP support now uses self-contained objects provided by Newlib
4991       <sys/lock.h> and offers a significantly better performance compared
4992       to the POSIX configuration of libgomp. It is possible to configure
4993       thread pools for each scheduler instance via the environment
4994       variable GOMP_RTEMS_THREAD_POOLS.
4995
4996  Solaris
4997
4998     * Solaris 12 is now fully supported. Minimal support had already been
4999       present in GCC 5.3.
5000     * Solaris 12 provides a full set of startup files (crt1.o, crti.o,
5001       crtn.o), which GCC now prefers over its own ones.
5002     * Position independent executables (PIE) are now supported on Solaris
5003       12.
5004     * Constructor priority is now supported on Solaris 12 with the system
5005       linker.
5006     * libvtv has been ported to Solaris 11 and up.
5007
5008  Windows
5009
5010     * The option -mstackrealign is now automatically activated in 32-bit
5011       mode whenever the use of SSE instructions is requested.
5012
5013Other significant improvements
5014
5015     * The gcc and g++ driver programs will now provide suggestions for
5016       misspelled command-line options.
5017
5018$ gcc -static-libfortran test.f95
5019gcc: error: unrecognized command line option '-static-libfortran'; did you mean
5020'-static-libgfortran'?
5021
5022     * The --enable-default-pie configure option enables generation of PIE
5023       by default.
5024
5025                                    GCC 6.2
5026
5027   This is the [22]list of problem reports (PRs) from GCC's bug tracking
5028   system that are known to be fixed in the 6.2 release. This list might
5029   not be complete (that is, it is possible that some PRs that have been
5030   fixed are not listed here).
5031
5032Target Specific Changes
5033
5034  SPARC
5035
5036     * Support for --with-cpu-32 and --with-cpu-64 configure options has
5037       been added on bi-architecture platforms.
5038     * Support for the SPARC M7 (Niagara 7) processor has been added.
5039     * Support for the VIS 4.0 instruction set has been added.
5040
5041                                    GCC 6.3
5042
5043   This is the [23]list of problem reports (PRs) from GCC's bug tracking
5044   system that are known to be fixed in the 6.3 release. This list might
5045   not be complete (that is, it is possible that some PRs that have been
5046   fixed are not listed here).
5047
5048Target Specific Changes
5049
5050  IA-32/x86-64
5051
5052     * Support for the [24]deprecated pcommit instruction has been
5053       removed.
5054
5055                                    GCC 6.4
5056
5057   This is the [25]list of problem reports (PRs) from GCC's bug tracking
5058   system that are known to be fixed in the 6.4 release. This list might
5059   not be complete (that is, it is possible that some PRs that have been
5060   fixed are not listed here).
5061
5062Operating Systems
5063
5064  RTEMS
5065
5066     * The ABI changes on ARM so that no short enums are used by default.
5067
5068                                    GCC 6.5
5069
5070   This is the [26]list of problem reports (PRs) from GCC's bug tracking
5071   system that are known to be fixed in the 6.5 release. This list might
5072   not be complete (that is, it is possible that some PRs that have been
5073   fixed are not listed here).
5074
5075
5076    For questions related to the use of GCC, please consult these web
5077    pages and the [27]GCC manuals. If that fails, the
5078    [28]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5079    web pages and the development of GCC are welcome on our developer
5080    list at [29]gcc@gcc.gnu.org. All of [30]our lists have public
5081    archives.
5082
5083   Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and
5084   distribution of this entire article is permitted in any medium,
5085   provided this notice is preserved.
5086
5087   These pages are [32]maintained by the GCC team. Last modified
5088   2021-10-31[33].
5089
5090References
5091
5092   1. http://gcc.gnu.org/gcc-6/porting_to.html
5093   2. http://gcc.gnu.org/onlinedocs/index.html#current
5094   3. https://gcc.gnu.org/ml/gcc/2015-08/msg00101.html
5095   4. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71151
5096   5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87822
5097   6. https://www.openacc.org/
5098   7. https://gcc.gnu.org/wiki/OpenACC
5099   8. https://gcc.gnu.org/wiki/Offloading
5100   9. https://www.openmp.org/specifications/
5101  10. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2014-1266
5102  11. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4377.pdf
5103  12. https://gcc.gnu.org/projects/cxx-status.html#cxx1z
5104  13. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/performance.html
5105  14. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/functions.html#gcc_jit_block_end_with_switch
5106  15. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_set_bool_allow_unreachable_blocks
5107  16. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_add_command_line_option
5108  17. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/AArch64-Function-Attributes.html#AArch64-Function-Attributes
5109  18. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/ARM-Function-Attributes.html#ARM-Function-Attributes
5110  19. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/Named-Address-Spaces.html#Named-Address-Spaces
5111  20. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/S_002f390-Function-Attributes.html#S_002f390-Function-Attributes
5112  21. http://www.musl-libc.org/
5113  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.2
5114  23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.3
5115  24. https://www.intel.com/content/www/us/en/developer/articles/technical/deprecate-pcommit-instruction.html
5116  25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.4
5117  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.5
5118  27. https://gcc.gnu.org/onlinedocs/
5119  28. mailto:gcc-help@gcc.gnu.org
5120  29. mailto:gcc@gcc.gnu.org
5121  30. https://gcc.gnu.org/lists.html
5122  31. https://www.fsf.org/
5123  32. https://gcc.gnu.org/about.html
5124  33. http://validator.w3.org/check/referer
5125======================================================================
5126http://gcc.gnu.org/gcc-5/index.html
5127                              GCC 5 Release Series
5128
5129   (This release series is no longer supported.)
5130
5131   October 10, 2017
5132
5133   The [1]GNU project and the GCC developers are pleased to announce the
5134   release of GCC 5.5.
5135
5136   This release is a bug-fix release, containing fixes for regressions in
5137   GCC 5.4 relative to previous releases of GCC.
5138
5139Release History
5140
5141   GCC 5.5
5142          October 10, 2017 ([2]changes, [3]documentation)
5143
5144   GCC 5.4
5145          June 3, 2016 ([4]changes, [5]documentation)
5146
5147   GCC 5.3
5148          December 4, 2015 ([6]changes, [7]documentation)
5149
5150   GCC 5.2
5151          July 16, 2015 ([8]changes, [9]documentation)
5152
5153   GCC 5.1
5154          April 22, 2015 ([10]changes, [11]documentation)
5155
5156References and Acknowledgements
5157
5158   GCC used to stand for the GNU C Compiler, but since the compiler
5159   supports several other languages aside from C, it now stands for the
5160   GNU Compiler Collection.
5161
5162   A list of [12]successful builds is updated as new information becomes
5163   available.
5164
5165   The GCC developers would like to thank the numerous people that have
5166   contributed new features, improvements, bug fixes, and other changes as
5167   well as test results to GCC. This [13]amazing group of volunteers is
5168   what makes GCC successful.
5169
5170   For additional information about GCC please refer to the [14]GCC
5171   project web site or contact the [15]GCC development mailing list.
5172
5173   To obtain GCC please use [16]our mirror sites or [17]our version
5174   control system.
5175
5176
5177    For questions related to the use of GCC, please consult these web
5178    pages and the [18]GCC manuals. If that fails, the
5179    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5180    web pages and the development of GCC are welcome on our developer
5181    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
5182    archives.
5183
5184   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
5185   distribution of this entire article is permitted in any medium,
5186   provided this notice is preserved.
5187
5188   These pages are [23]maintained by the GCC team. Last modified
5189   2021-07-28[24].
5190
5191References
5192
5193   1. http://www.gnu.org/
5194   2. http://gcc.gnu.org/gcc-5/changes.html
5195   3. http://gcc.gnu.org/onlinedocs/5.5.0/
5196   4. http://gcc.gnu.org/gcc-5/changes.html
5197   5. http://gcc.gnu.org/onlinedocs/5.4.0/
5198   6. http://gcc.gnu.org/gcc-5/changes.html
5199   7. http://gcc.gnu.org/onlinedocs/5.3.0/
5200   8. http://gcc.gnu.org/gcc-5/changes.html
5201   9. http://gcc.gnu.org/onlinedocs/5.2.0/
5202  10. http://gcc.gnu.org/gcc-5/changes.html
5203  11. http://gcc.gnu.org/onlinedocs/5.1.0/
5204  12. http://gcc.gnu.org/gcc-5/buildstat.html
5205  13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
5206  14. http://gcc.gnu.org/index.html
5207  15. mailto:gcc@gcc.gnu.org
5208  16. http://gcc.gnu.org/mirrors.html
5209  17. http://gcc.gnu.org/git.html
5210  18. https://gcc.gnu.org/onlinedocs/
5211  19. mailto:gcc-help@gcc.gnu.org
5212  20. mailto:gcc@gcc.gnu.org
5213  21. https://gcc.gnu.org/lists.html
5214  22. https://www.fsf.org/
5215  23. https://gcc.gnu.org/about.html
5216  24. http://validator.w3.org/check/referer
5217======================================================================
5218http://gcc.gnu.org/gcc-5/changes.html
5219                              GCC 5 Release Series
5220                        Changes, New Features, and Fixes
5221
5222Caveats
5223
5224     * The default mode for C is now -std=gnu11 instead of -std=gnu89.
5225     * The C++ runtime library (libstdc++) uses a new ABI by default (see
5226       [1]below).
5227     * The Graphite framework for loop optimizations no longer requires
5228       the CLooG library, only ISL version 0.14 (recommended) or 0.12.2.
5229       The installation manual contains more information about
5230       requirements to build GCC.
5231     * The non-standard C++0x type traits has_trivial_default_constructor,
5232       has_trivial_copy_constructor and has_trivial_copy_assign have been
5233       deprecated and will be removed in a future version. The standard
5234       C++11 traits is_trivially_default_constructible,
5235       is_trivially_copy_constructible and is_trivially_copy_assignable
5236       should be used instead.
5237     * On AVR, support has been added for the devices
5238       ATtiny4/5/9/10/20/40. This requires Binutils 2.25 or newer.
5239     * The AVR port uses a new scheme to describe supported devices: For
5240       each supported device the compiler provides a device-specific
5241       [2]spec file. If the compiler is used together with AVR-LibC, this
5242       requires at least GCC 5.2 and a version of AVR-LibC which
5243       implements [3]feature #44574.
5244
5245General Optimizer Improvements
5246
5247     * Inter-procedural optimization improvements:
5248          + An Identical Code Folding (ICF) pass (controlled via
5249            -fipa-icf) has been added. Compared to the identical code
5250            folding performed by the Gold linker this pass does not
5251            require function sections. It also performs merging before
5252            inlining, so inter-procedural optimizations are aware of the
5253            code re-use. On the other hand not all unifications performed
5254            by a linker are doable by GCC which must honor aliasing
5255            information. During link-time optimization of Firefox, this
5256            pass unifies about 31000 functions, that is 14% overall.
5257          + The devirtualization pass was significantly improved by adding
5258            better support for speculative devirtualization and dynamic
5259            type detection. About 50% of virtual calls in Firefox are now
5260            speculatively devirtualized during link-time optimization.
5261          + A new comdat localization pass allows the linker to eliminate
5262            more dead code in presence of C++ inline functions.
5263          + Virtual tables are now optimized. Local aliases are used to
5264            reduce dynamic linking time of C++ virtual tables on ELF
5265            targets and data alignment has been reduced to limit data
5266            segment bloat.
5267          + A new -fno-semantic-interposition option can be used to
5268            improve code quality of shared libraries where interposition
5269            of exported symbols is not allowed.
5270          + Write-only variables are now detected and optimized out.
5271          + With profile feedback the function inliner can now bypass
5272            --param inline-insns-auto and --param inline-insns-single
5273            limits for hot calls.
5274          + The IPA reference pass was significantly sped up making it
5275            feasible to enable -fipa-reference with -fprofile-generate.
5276            This also solves a bottleneck seen when building Chromium with
5277            link-time optimization.
5278          + The symbol table and call-graph API was reworked to C++ and
5279            simplified.
5280          + The interprocedural propagation of constants now also
5281            propagates alignments of pointer parameters. This for example
5282            means that the vectorizer often does not need to generate loop
5283            prologues and epilogues to make up for potential
5284            misalignments.
5285     * Link-time optimization improvements:
5286          + One Definition Rule based merging of C++ types has been
5287            implemented. Type merging enables better devirtualization and
5288            alias analysis. Streaming extra information needed to merge
5289            types adds about 2-6% of memory size and object size increase.
5290            This can be controlled by -flto-odr-type-merging.
5291          + Command-line optimization and target options are now streamed
5292            on a per-function basis and honored by the link-time
5293            optimizer. This change makes link-time optimization a more
5294            transparent replacement of per-file optimizations. It is now
5295            possible to build projects that require different optimization
5296            settings for different translation units (such as -ffast-math,
5297            -mavx, or -finline). Contrary to earlier GCC releases, the
5298            optimization and target options passed on the link command
5299            line are ignored.
5300            Note that this applies only to those command-line options that
5301            can be passed to optimize and target attributes. Command-line
5302            options affecting global code generation (such as -fpic),
5303            warnings (such as -Wodr), optimizations affecting the way
5304            static variables are optimized (such as -fcommon), debug
5305            output (such as -g), and --param parameters can be applied
5306            only to the whole link-time optimization unit. In these cases,
5307            it is recommended to consistently use the same options at both
5308            compile time and link time.
5309          + GCC bootstrap now uses slim LTO object files.
5310          + Memory usage and link times were improved. Tree merging was
5311            sped up, memory usage of GIMPLE declarations and types was
5312            reduced, and, support for on-demand streaming of variable
5313            constructors was added.
5314     * Feedback directed optimization improvements:
5315          + A new auto-FDO mode uses profiles collected by low overhead
5316            profiling tools (perf) instead of more expensive program
5317            instrumentation (via -fprofile-generate). SPEC2006 benchmarks
5318            on x86-64 improve by 4.7% with auto-FDO and by 7.3% with
5319            traditional feedback directed optimization.
5320          + Profile precision was improved in presence of C++ inline and
5321            extern inline functions.
5322          + The new gcov-tool utility allows manipulating profiles.
5323          + Profiles are now more tolerant to source file changes (this
5324            can be controlled by --param profile-func-internal-id).
5325     * Register allocation improvements:
5326          + A new local register allocator (LRA) sub-pass, controlled by
5327            -flra-remat, implements control-flow sensitive global register
5328            rematerialization. Instead of spilling and restoring a
5329            register value, it is recalculated if it is profitable. The
5330            sub-pass improved SPEC2000 generated code by 1% and 0.5%
5331            correspondingly on ARM and x86-64.
5332          + Reuse of the PIC hard register, instead of using a fixed
5333            register, was implemented on x86/x86-64 targets. This improves
5334            generated PIC code performance as more hard registers can be
5335            used. Shared libraries can significantly benefit from this
5336            optimization. Currently it is switched on only for x86/x86-64
5337            targets. As RA infrastructure is already implemented for PIC
5338            register reuse, other targets might follow this in the future.
5339          + A simple form of inter-procedural RA was implemented. When it
5340            is known that a called function does not use caller-saved
5341            registers, save/restore code is not generated around the call
5342            for such registers. This optimization can be controlled by
5343            -fipa-ra
5344          + LRA is now much more effective at generating spills of general
5345            registers into vector registers instead of memory on
5346            architectures (e.g., modern Intel processors) where this is
5347            profitable.
5348     * UndefinedBehaviorSanitizer gained a few new sanitization options:
5349          + -fsanitize=float-divide-by-zero: detect floating-point
5350            division by zero;
5351          + -fsanitize=float-cast-overflow: check that the result of
5352            floating-point type to integer conversions do not overflow;
5353          + -fsanitize=bounds: enable instrumentation of array bounds and
5354            detect out-of-bounds accesses;
5355          + -fsanitize=alignment: enable alignment checking, detect
5356            various misaligned objects;
5357          + -fsanitize=object-size: enable object size checking, detect
5358            various out-of-bounds accesses.
5359          + -fsanitize=vptr: enable checking of C++ member function calls,
5360            member accesses and some conversions between pointers to base
5361            and derived classes, detect if the referenced object does not
5362            have the correct dynamic type.
5363     * Pointer Bounds Checker, a bounds violation detector, has been added
5364       and can be enabled via -fcheck-pointer-bounds. Memory accesses are
5365       instrumented with run-time checks of used pointers against their
5366       bounds to detect pointer bounds violations (overflows). The Pointer
5367       Bounds Checker is available on x86/x86-64 GNU/Linux targets with a
5368       new ISA extension Intel MPX support. See the Pointer Bounds Checker
5369       [4]Wiki page for more details.
5370
5371New Languages and Language specific improvements
5372
5373     * [5]OpenMP 4.0 specification offloading features are now supported
5374       by the C, C++, and Fortran compilers. Generic changes:
5375          + Infrastructure (suitable for any vendor).
5376          + Testsuite which covers offloading from the [6]OpenMP 4.0
5377            Examples document.
5378       Specific for upcoming Intel Xeon Phi products:
5379          + Run-time library.
5380          + Card emulator.
5381     * GCC 5 includes a preliminary implementation of the OpenACC 2.0a
5382       specification. OpenACC is intended for programming accelerator
5383       devices such as GPUs. See [7]the OpenACC wiki page for more
5384       information.
5385
5386  C family
5387
5388     * The default setting of the -fdiagnostics-color= command-line option
5389       is now [8]configurable when building GCC using configuration option
5390       --with-diagnostics-color=. The possible values are: never, always,
5391       auto and auto-if-env. The new default auto uses color only when the
5392       standard error is a terminal. The default in GCC 4.9 was
5393       auto-if-env, which is equivalent to auto if there is a non-empty
5394       GCC_COLORS environment variable, and never otherwise. As in GCC
5395       4.9, an empty GCC_COLORS variable in the environment will always
5396       disable colors, no matter what the default is or what command-line
5397       options are used.
5398     * A new command-line option -Wswitch-bool has been added for the C
5399       and C++ compilers, which warns whenever a switch statement has an
5400       index of boolean type.
5401     * A new command-line option -Wlogical-not-parentheses has been added
5402       for the C and C++ compilers, which warns about "logical not" used
5403       on the left hand side operand of a comparison.
5404     * A new command-line option -Wsizeof-array-argument has been added
5405       for the C and C++ compilers, which warns when the sizeof operator
5406       is applied to a parameter that has been declared as an array in a
5407       function definition.
5408     * A new command-line option -Wbool-compare has been added for the C
5409       and C++ compilers, which warns about boolean expressions compared
5410       with an integer value different from true/false.
5411     * Full support for Cilk Plus has been added to the GCC compiler. Cilk
5412       Plus is an extension to the C and C++ languages to support data and
5413       task parallelism.
5414     * A new attribute no_reorder prevents reordering of selected symbols
5415       against other such symbols or inline assembler. This enables to
5416       link-time optimize the Linux kernel without having to resort to
5417       -fno-toplevel-reorder that disables several optimizations.
5418     * New preprocessor constructs, __has_include and __has_include_next,
5419       to test the availability of headers have been added.
5420       This demonstrates a way to include the header <optional> only if it
5421       is available:
5422
5423#ifdef __has_include
5424#  if __has_include(<optional>)
5425#    include <optional>
5426#    define have_optional 1
5427#  elif __has_include(<experimental/optional>)
5428#    include <experimental/optional>
5429#    define have_optional 1
5430#    define experimental_optional
5431#  else
5432#    define have_optional 0
5433#  endif
5434#endif
5435
5436       The header search paths for __has_include and __has_include_next
5437       are equivalent to those of the standard directive #include and the
5438       extension #include_next respectively.
5439     * A new built-in function-like macro to determine the existence of an
5440       attribute, __has_attribute, has been added. The equivalent built-in
5441       macro __has_cpp_attribute was added to C++ to support
5442       [9]Feature-testing recommendations for C++. The macro
5443       __has_attribute is added to all C-like languages as an extension:
5444
5445int
5446#ifdef __has_attribute
5447#  if __has_attribute(__noinline__)
5448  __attribute__((__noinline__))
5449#  endif
5450#endif
5451foo(int x);
5452
5453       If an attribute exists, a nonzero constant integer is returned. For
5454       standardized C++ attributes a date is returned, otherwise the
5455       constant returned is 1. Both __has_attribute and
5456       __has_cpp_attribute will add underscores to an attribute name if
5457       necessary to resolve the name. For C++11 and onwards the attribute
5458       may be scoped.
5459     * A new set of built-in functions for arithmetics with overflow
5460       checking has been added: __builtin_add_overflow,
5461       __builtin_sub_overflow and __builtin_mul_overflow and for
5462       compatibility with clang also other variants. These builtins have
5463       two integral arguments (which don't need to have the same type),
5464       the arguments are extended to infinite precision signed type, +, -
5465       or * is performed on those, and the result is stored in an integer
5466       variable pointed to by the last argument. If the stored value is
5467       equal to the infinite precision result, the built-in functions
5468       return false, otherwise true. The type of the integer variable that
5469       will hold the result can be different from the types of the first
5470       two arguments. The following snippet demonstrates how this can be
5471       used in computing the size for the calloc function:
5472
5473void *
5474calloc (size_t x, size_t y)
5475{
5476  size_t sz;
5477  if (__builtin_mul_overflow (x, y, &sz))
5478    return NULL;
5479  void *ret = malloc (sz);
5480  if (ret) memset (res, 0, sz);
5481  return ret;
5482}
5483
5484       On e.g. i?86 or x86-64 the above will result in a mul instruction
5485       followed by a jump on overflow.
5486     * The option -fextended-identifiers is now enabled by default for
5487       C++, and for C99 and later C versions. Various bugs in the
5488       implementation of extended identifiers have been fixed.
5489
5490  C
5491
5492     * The default mode has been changed to -std=gnu11.
5493     * A new command-line option -Wc90-c99-compat has been added to warn
5494       about features not present in ISO C90, but present in ISO C99.
5495     * A new command-line option -Wc99-c11-compat has been added to warn
5496       about features not present in ISO C99, but present in ISO C11.
5497     * It is possible to disable warnings about conversions between
5498       pointers that have incompatible types via a new warning option
5499       -Wno-incompatible-pointer-types; warnings about implicit
5500       incompatible integer to pointer and pointer to integer conversions
5501       via a new warning option -Wno-int-conversion; and warnings about
5502       qualifiers on pointers being discarded via a new warning option
5503       -Wno-discarded-qualifiers.
5504     * To allow proper use of const qualifiers with multidimensional
5505       arrays, GCC will not warn about incompatible pointer types anymore
5506       for conversions between pointers to arrays with and without const
5507       qualifier (except when using -pedantic). Instead, a new warning is
5508       emitted only if the const qualifier is lost. This can be controlled
5509       with a new warning option -Wno-discarded-array-qualifiers.
5510     * The C front end now generates more precise caret diagnostics.
5511     * The -pg command-line option now only affects the current file in an
5512       LTO build.
5513
5514  C++
5515
5516     * G++ now supports [10]C++14 variable templates.
5517     * -Wnon-virtual-dtor doesn't warn anymore for final classes.
5518     * Excessive template instantiation depth is now a fatal error. This
5519       prevents excessive diagnostics that usually do not help to identify
5520       the problem.
5521     * G++ and libstdc++ now implement the feature-testing macros from
5522       [11]Feature-testing recommendations for C++.
5523     * G++ now allows typename in a template template parameter.
5524
5525template<template<typename> typename X> struct D; // OK
5526
5527     * G++ now supports [12]C++14 aggregates with non-static data member
5528       initializers.
5529
5530struct A { int i, j = i; };
5531A a = { 42 }; // a.j is also 42
5532
5533     * G++ now supports [13]C++14 extended constexpr.
5534
5535constexpr int f (int i)
5536{
5537  int j = 0;
5538  for (; i > 0; --i)
5539    ++j;
5540  return j;
5541}
5542
5543constexpr int i = f(42); // i is 42
5544
5545     * G++ now supports the [14]C++14 sized deallocation functions.
5546
5547void operator delete (void *, std::size_t) noexcept;
5548void operator delete[] (void *, std::size_t) noexcept;
5549
5550     * A new One Definition Rule violation warning (controlled by -Wodr)
5551       detects mismatches in type definitions and virtual table contents
5552       during link-time optimization.
5553     * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help
5554       developers to annotate programs with final specifiers (or anonymous
5555       namespaces) to improve code generation. These warnings can be used
5556       at compile time, but they are more useful in combination with
5557       link-time optimization.
5558     * G++ no longer supports [15]N3639 variable length arrays, as they
5559       were removed from the C++14 working paper prior to ratification.
5560       GNU VLAs are still supported, so VLA support is now the same in
5561       C++14 mode as in C++98 and C++11 modes.
5562     * G++ now allows passing a non-trivially-copyable class via C
5563       varargs, which is conditionally-supported with
5564       implementation-defined semantics in the standard. This uses the
5565       same calling convention as a normal value parameter.
5566     * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So
5567       various mangling bugs are fixed, but G++ will still emit aliases
5568       with the old, wrong mangling where feasible. -Wabi=2 will warn
5569       about differences between ABI version 2 and the current setting.
5570     * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely
5571       to be unaffected, but -Wabi=8 will warn about a non-static data
5572       member with type std::nullptr_t which changes position due to this
5573       change.
5574
5575    Runtime Library (libstdc++)
5576
5577     * A [16]Dual ABI is provided by the library. A new ABI is enabled by
5578       default. The old ABI is still supported and can be used by defining
5579       the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++
5580       standard library headers.
5581     * A new implementation of std::string is enabled by default, using
5582       the small string optimization instead of copy-on-write reference
5583       counting.
5584     * A new implementation of std::list is enabled by default, with an
5585       O(1) size() function;
5586     * [17]Full support for C++11, including the following new features:
5587          + std::deque and std::vector<bool> meet the allocator-aware
5588            container requirements;
5589          + movable and swappable iostream classes;
5590          + support for std::align and std::aligned_union;
5591          + type traits std::is_trivially_copyable,
5592            std::is_trivially_constructible, std::is_trivially_assignable
5593            etc.;
5594          + I/O manipulators std::put_time, std::get_time, std::hexfloat
5595            and std::defaultfloat;
5596          + generic locale-aware std::isblank;
5597          + locale facets for Unicode conversion;
5598          + atomic operations for std::shared_ptr;
5599          + std::notify_all_at_thread_exit() and functions for making
5600            futures ready at thread exit.
5601     * Support for the C++11 hexfloat manipulator changes how the num_put
5602       facet formats floating point types when
5603       ios_base::fixed|ios_base::scientific is set in a stream's fmtflags.
5604       This change affects all language modes, even though the C++98
5605       standard gave no special meaning to that combination of flags. To
5606       prevent the use of hexadecimal notation for floating point types
5607       use str.unsetf(std::ios_base::floatfield) to clear the relevant
5608       bits in str.flags().
5609     * [18]Full experimental support for C++14, including the following
5610       new features:
5611          + std::is_final type trait;
5612          + heterogeneous comparison lookup in associative containers.
5613          + global functions cbegin, cend, rbegin, rend, crbegin, and
5614            crend for range access to containers, arrays and initializer
5615            lists.
5616     * [19]Improved experimental support for the Library Fundamentals TS,
5617       including:
5618          + class std::experimental::any;
5619          + function template std::experimental::apply;
5620          + function template std::experimental::sample;
5621          + function template std::experimental::search and related
5622            searcher types;
5623          + variable templates for type traits;
5624          + function template std::experimental::not_fn.
5625     * New random number distributions logistic_distribution and
5626       uniform_on_sphere_distribution as extensions.
5627     * [20]GDB Xmethods for containers and std::unique_ptr.
5628
5629  Fortran
5630
5631     * Compatibility notice:
5632          + The version of the module files (.mod) has been incremented.
5633          + For free-form source files [21]-Werror=line-truncation is now
5634            enabled by default. Note that comments exceeding the line
5635            length are not diagnosed. (For fixed-form source code, the
5636            same warning is available but turned off by default, such that
5637            excess characters are ignored. -ffree-line-length-n and
5638            -ffixed-line-length-n can be used to modify the default line
5639            lengths of 132 and 72 columns, respectively.)
5640          + The -Wtabs option is now more sensible: with -Wtabs the
5641            compiler warns if it encounters tabs and with -Wno-tabs this
5642            warning is turned off. Before, -Wno-tabs warned and -Wtabs
5643            disabled the warning. As before, this warning is also enabled
5644            by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts
5645            options of -std=.
5646     * Incomplete support for colorizing diagnostics emitted by gfortran
5647       has been added. The option [22]-fdiagnostics-color controls when
5648       color is used in diagnostics. The default value of this option can
5649       be [23]configured when building GCC. The GCC_COLORS environment
5650       variable can be used to customize the colors or disable coloring
5651       completely. Sample diagnostics output:
5652      $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90
5653      test.f90:6:1:
5654
5655       0 continue
5656       1
5657      Error: Zero is not a valid statement label at (1)
5658      test.f90:9:6:
5659
5660         USE foo
5661            1
5662      Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only]
5663
5664     * The -Wuse-without-only option has been added to warn when a USE
5665       statement has no ONLY qualifier and thus implicitly imports all
5666       public entities of the used module.
5667     * Formatted READ and WRITE statements now work correctly in
5668       locale-aware programs. For more information and potential caveats,
5669       see [24]Section 5.3 Thread-safety of the runtime library in the
5670       manual.
5671     * [25]Fortran 2003:
5672          + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and
5673            IEEE_ARITHMETIC) are now supported.
5674     * [26]Fortran 2008:
5675          + [27]Coarrays: Full experimental support of Fortran 2008's
5676            coarrays with -fcoarray=lib except for allocatable/pointer
5677            components of derived-type coarrays. GCC currently only ships
5678            with a single-image library (libcaf_single), but multi-image
5679            support based on MPI and GASNet is provided by the libraries
5680            of the [28]OpenCoarrays project.
5681     * TS18508 Additional Parallel Features in Fortran:
5682          + Support for the collective intrinsic subroutines CO_MAX,
5683            CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added,
5684            including -fcoarray=lib support.
5685          + Support for the new atomic intrinsics has been added,
5686            including -fcoarray=lib support.
5687     * Fortran 2015:
5688          + Support for IMPLICIT NONE (external, type).
5689          + ERROR STOP is now permitted in pure procedures.
5690
5691  Go
5692
5693     * GCC 5 provides a complete implementation of the Go 1.4.2 release.
5694     * Building GCC 5 with Go enabled will install two new programs:
5695       [29]go and [30]gofmt.
5696
5697libgccjit
5698
5699   New in GCC 5 is the ability to build GCC as a shared library for
5700   embedding in other processes (such as interpreters), suitable for
5701   Just-In-Time compilation to machine code.
5702
5703   The shared library has a [31]C API and a [32]C++ wrapper API providing
5704   some "syntactic sugar". There are also bindings available from 3rd
5705   parties for [33]Python and for [34]D.
5706
5707   For example, this library can be used by interpreters for [35]compiling
5708   functions from bytecode to machine code.
5709
5710   The library can also be used for ahead-of-time compilation, enabling
5711   GCC to be plugged into a pre-existing front end. An example of using
5712   this to build a compiler for an esoteric language we'll refer to as
5713   "brainf" can be seen [36]here.
5714
5715   libgccjit is licensed under the GPLv3 (or at your option, any later
5716   version)
5717
5718   It should be regarded as experimental at this time.
5719
5720New Targets and Target Specific Improvements
5721
5722  Reporting stack usage
5723
5724     * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the
5725       -fstack-usage option.
5726
5727  AArch64
5728
5729     * Code generation for the ARM Cortex-A57 processor has been improved.
5730       A more accurate instruction scheduling model for the processor is
5731       now used, and a number of compiler tuning parameters have been set
5732       to offer increased performance when compiling with -mcpu=cortex-a57
5733       or -mtune=cortex-a57.
5734     * A workaround for the ARM Cortex-A53 erratum 835769 has been added
5735       and can be enabled by giving the -mfix-cortex-a53-835769 option.
5736       Alternatively it can be enabled by default by configuring GCC with
5737       the --enable-fix-cortex-a53-835769 option.
5738     * The optional cryptographic extensions to the ARMv8-A architecture
5739       are no longer enabled by default when specifying the
5740       -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53
5741       options. To enable these extensions add +crypto to the value of
5742       -mcpu or -march e.g. -mcpu=cortex-a53+crypto.
5743     * Support has been added for the following processors (GCC
5744       identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and
5745       initial support for its big.LITTLE combination with the ARM
5746       Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx),
5747       Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1).
5748       The GCC identifiers can be used as arguments to the -mcpu or -mtune
5749       options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53.
5750       Using -mcpu=cortex-a72 requires a version of GNU binutils that has
5751       support for the Cortex-A72.
5752     * The transitional options -mlra and -mno-lra have been removed. The
5753       AArch64 backend now uses the local register allocator (LRA) only.
5754
5755  ARM
5756
5757     * Thumb-1 assembly code is now generated in unified syntax. The new
5758       option -masm-syntax-unified specifies whether inline assembly code
5759       is using unified syntax. By default the option is off which means
5760       non-unified syntax is used. However this is subject to change in
5761       future releases. Eventually the non-unified syntax will be
5762       deprecated.
5763     * It is now a configure-time error to use the --with-cpu configure
5764       option with either of --with-tune or --with-arch.
5765     * Code generation for the ARM Cortex-A57 processor has been improved.
5766       A more accurate instruction scheduling model for the processor is
5767       now used, and a number of compiler tuning parameters have been set
5768       to offer increased performance when compiling with -mcpu=cortex-a57
5769       or -mtune=cortex-a57.
5770     * Support has been added for the following processors (GCC
5771       identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and
5772       initial support for its big.LITTLE combination with the ARM
5773       Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and
5774       initial support for its big.LITTLE combination with the ARM
5775       Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7),
5776       Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1).
5777       The GCC identifiers can be used as arguments to the -mcpu or -mtune
5778       options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53.
5779       Using -mcpu=cortex-a72 requires a version of GNU binutils that has
5780       support for the Cortex-A72.
5781     * The deprecated option -mwords-little-endian has been removed.
5782     * The options -mapcs, -mapcs-frame, -mtpcs-frame and
5783       -mtpcs-leaf-frame which are only applicable to the old ABI have
5784       been deprecated.
5785     * The transitional options -mlra and -mno-lra have been removed. The
5786       ARM backend now uses the local register allocator (LRA) only.
5787
5788  AVR
5789
5790     * The compiler no more supports individual devices like ATmega8.
5791       Specifying, say, -mmcu=atmega8 triggers the usage of the
5792       device-specific [37]spec file specs-atmega8 which is part of the
5793       installation and describes options for the sub-processes like
5794       compiler proper, assembler and linker. You can add support for a
5795       new device -mmcu=mydevice as follows:
5796         1. In an empty directory /someplace, create a new directory
5797            device-specs.
5798         2. Copy a device spec file from the installed device-specs
5799            folder, follow the comments in that file and then save it as
5800            /someplace/device-specs/specs-mydevice.
5801         3. Add -B /someplace -mmcu=mydevice to the compiler's
5802            command-line options. Notice that /someplace must specify an
5803            absolute path and that mydevice must not start with "avr".
5804         4. Provided you have a device-specific library libmydevice.a
5805            available, you can put it at /someplace, dito for a
5806            device-specific startup file crtmydevice.o.
5807       The contents of the device spec files depend on the compiler's
5808       configuration, in particular on --with-avrlibc=no and whether or
5809       not it is configured for RTEMS.
5810     * A new command-line option -nodevicelib has been added. It prevents
5811       the compiler from linking against AVR-LibC's device-specific
5812       library libdevice.a.
5813     * The following three command-line options have been added:
5814
5815        -mrmw
5816                Set if the device supports the read-modify-write
5817                instructions LAC, LAS, LAT and XCH.
5818
5819        -mn-flash=size
5820                Specify the flash size of the device in units of 64 KiB,
5821                rounded up to the next integer as needed. This option
5822                affects the availability of the [38]AVR address-spaces.
5823
5824        -mskip-bug
5825                Set if the device is affected by the respective silicon
5826                bug.
5827
5828       In general, you don't need to set these options by hand. The new
5829       device-specific spec file will set them as needed.
5830
5831  IA-32/x86-64
5832
5833     * New ISA extensions support AVX-512{BW,DQ,VL,IFMA,VBMI} of Intel's
5834       CPU codenamed Skylake Server was added to GCC. That includes inline
5835       assembly support, new intrinsics, and basic autovectorization.
5836       These new AVX-512 extensions are available via the following GCC
5837       switches: AVX-512 Vector Length EVEX feature: -mavx512vl, AVX-512
5838       Byte and Word instructions: -mavx512bw, AVX-512 Dword and Qword
5839       instructions: -mavx512dq, AVX-512 FMA-52 instructions: -mavx512ifma
5840       and for AVX-512 Vector Bit Manipulation Instructions: -mavx512vbmi.
5841     * New ISA extensions support Intel MPX was added to GCC. This new
5842       extension is available via the -mmpx compiler switch. Intel MPX is
5843       a set of processor features which, with compiler, run-time library
5844       and OS support, brings increased robustness to software by run-time
5845       checking pointer references against their bounds. In GCC Intel MPX
5846       is supported by Pointer Bounds Checker and libmpx run-time
5847       libraries.
5848     * The new -mrecord-mcount option for -pg generates a Linux kernel
5849       style table of pointers to mcount or __fentry__ calls at the
5850       beginning of functions. The new -mnop-mcount option in addition
5851       also generates nops in place of the __fentry__ or mcount call, so
5852       that a call per function can be later patched in. This can be used
5853       for low overhead tracing or hot code patching.
5854     * The new -malign-data option controls how GCC aligns variables.
5855       -malign-data=compat uses increased alignment compatible with GCC
5856       4.8 and earlier, -malign-data=abi uses alignment as specified by
5857       the psABI, and -malign-data=cacheline uses increased alignment to
5858       match the cache line size. -malign-data=compat is the default.
5859     * The new -mskip-rax-setup option skips setting up the RAX register
5860       when SSE is disabled and there are no variable arguments passed in
5861       vector registers. This can be used to optimize the Linux kernel.
5862
5863  MIPS
5864
5865     * MIPS Releases 3 and 5 are now directly supported. Use the
5866       command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5
5867       to enable code-generation for these processors.
5868     * The Imagination P5600 processor is now supported using the
5869       -march=p5600 command-line option.
5870     * The Cavium Octeon3 processor is now supported using the
5871       -march=octeon3 command-line option.
5872     * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6
5873       command-line options.
5874     * The o32 ABI has been modified and extended. The o32 64-bit
5875       floating-point register support is now obsolete and has been
5876       removed. It has been replaced by three ABI extensions FPXX, FP64A,
5877       and FP64. The meaning of the -mfp64 command-line option has
5878       changed. It is now used to enable the FP64A and FP64 ABI
5879       extensions.
5880          + The FPXX extension requires that code generated to access
5881            double-precision values use even-numbered registers. Code that
5882            adheres to this extension is link-compatible with all other
5883            o32 double-precision ABI variants and will execute correctly
5884            in all hardware FPU modes. The command-line options -mabi=32
5885            -mfpxx can be used to enable this extension. MIPS II is the
5886            minimum processor required.
5887          + The o32 FP64A extension requires that floating-point registers
5888            be 64-bit and odd-numbered single-precision registers are not
5889            allowed. Code that adheres to the o32 FP64A variant is
5890            link-compatible with all other o32 double-precision ABI
5891            variants. The command-line options -mabi=32 -mfp64
5892            -mno-odd-spreg can be used to enable this extension. MIPS32R2
5893            is the minimum processor required.
5894          + The o32 FP64 extension also requires that floating-point
5895            registers be 64-bit, but permits the use of single-precision
5896            registers. Code that adheres to the o32 FP64 variant is
5897            link-compatible with o32 FPXX and o32 FP64A variants only,
5898            i.e. it is not compatible with the original o32
5899            double-precision ABI. The command-line options -mabi=32 -mfp64
5900            -modd-spreg can be used to enable this extension. MIPS32R2 is
5901            the minimum processor required.
5902       The new ABI variants can be enabled by default using the configure
5903       time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32.
5904       It is strongly recommended that all vendors begin to set o32 FPXX
5905       as the default ABI. This will be required to run the generated code
5906       on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and
5907       MIPSR6 cores.
5908     * GCC will now pass all floating-point options to the assembler if
5909       GNU binutils 2.25 is used. As a result, any inline assembly code
5910       that uses hard-float instructions should be amended to include a
5911       .set directive to override the global assembler options when
5912       compiling for soft-float targets.
5913
5914  NDS32
5915
5916     * The variadic function ABI implementation is now compatible with
5917       past Andes toolchains where the caller uses registers to pass
5918       arguments and the callee is in charge of pushing them on stack.
5919     * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have
5920       been removed since they are not yet available in the nds32 port of
5921       GNU binutils.
5922     * A new option -mcmodel=[small|medium|large] supports varied code
5923       models on code generation. The -mgp-direct option became
5924       meaningless and can be discarded.
5925
5926  RX
5927
5928     * A new command line option -mno-allow-string-insns can be used to
5929       disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL,
5930       SWHILE and RMPA instructions. An erratum released by Renesas shows
5931       that it is unsafe to use these instructions on addresses within the
5932       I/O space of the processor. The new option can be used when the
5933       programmer is concerned that the I/O space might be accessed. The
5934       default is still to enable these instructions.
5935
5936  SH
5937
5938     * The compiler will now pass the appropriate --isa= option to the
5939       assembler.
5940     * The default handling for the GBR has been changed from call
5941       clobbered to call preserved. The old behavior can be reinstated by
5942       specifying the option -fcall-used-gbr.
5943     * Support for the SH4A fpchg instruction has been added which will be
5944       utilized when switching between single and double precision FPU
5945       modes.
5946     * The compiler no longer uses the __fpscr_values array for switching
5947       between single and double FPU precision modes on non-SH4A targets.
5948       Instead mode switching will now be performed by storing, modifying
5949       and reloading the FPSCR, so that other FPSCR bits are preserved
5950       across mode switches. The __fpscr_values array that is defined in
5951       libgcc is still present for backwards compatibility, but it will
5952       not be referenced by compiler generated code anymore.
5953     * New builtin functions __builtin_sh_get_fpscr and
5954       __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr
5955       function will mask the specified bits in such a way that the SZ, PR
5956       and FR mode bits will be preserved, while changing the other bits.
5957       These new functions do not reference the __fpscr_values array. The
5958       old functions __set_fpscr and __get_fpscr in libgcc which access
5959       the __fpscr_values array are still present for backwards
5960       compatibility, but their usage is highly discouraged.
5961     * Some improvements to code generated for __atomic built-in
5962       functions.
5963     * When compiling for SH2E the compiler will no longer force the usage
5964       of delay slots for conditional branch instructions bt and bf. The
5965       old behavior can be reinstated (e.g. to work around a hardware bug
5966       in the original SH7055) by specifying the new option
5967       -mcbranch-force-delay-slot.
5968
5969Operating Systems
5970
5971  AIX
5972
5973     * GCC now supports stabs debugging continuation lines to allow long
5974       stabs debug information without overflow that generates AIX linker
5975       errors.
5976
5977  DragonFly BSD
5978
5979     * GCC now supports the DragonFly BSD operating system.
5980
5981  FreeBSD
5982
5983     * GCC now supports the FreeBSD operating system for the arm port
5984       through the arm*-*-freebsd* target triplets.
5985
5986  VxWorks MILS
5987
5988     * GCC now supports the MILS (Multiple Independent Levels of Security)
5989       variant of WindRiver's VxWorks operating system for PowerPC
5990       targets.
5991
5992Other significant improvements
5993
5994     * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option
5995       to set the compiler to use.
5996
5997     * When the new command-line option -freport-bug is used, GCC
5998       automatically generates a developer-friendly reproducer whenever an
5999       internal compiler error is encountered.
6000
6001                                    GCC 5.2
6002
6003   This is the [39]list of problem reports (PRs) from GCC's bug tracking
6004   system that are known to be fixed in the 5.2 release. This list might
6005   not be complete (that is, it is possible that some PRs that have been
6006   fixed are not listed here).
6007
6008Target Specific Changes
6009
6010  IA-32/x86-64
6011
6012     * Support for new AMD instructions monitorx and mwaitx has been
6013       added. This includes new intrinsic and built-in support. It is
6014       enabled through option -mmwaitx. The instructions monitorx and
6015       mwaitx implement the same functionality as the old monitor and
6016       mwait instructions. In addition mwaitx adds a configurable timer.
6017       The timer value is received as third argument and stored in
6018       register %ebx.
6019
6020  S/390, System z, IBM z Systems
6021
6022     * Support for the IBM z13 processor has been added. When using the
6023       -march=z13 option, the compiler will generate code making use of
6024       the new instructions and registers introduced with the vector
6025       extension facility. The -mtune=z13 option enables z13 specific
6026       instruction scheduling without making use of new instructions.
6027       Compiling code with -march=z13 reduces the default alignment of
6028       vector types bigger than 8 bytes to 8. This is an ABI change and
6029       care must be taken when linking modules compiled with different
6030       arch levels which interchange variables containing vector type
6031       values. For newly compiled code the GNU linker will emit a warning.
6032     * The -mzvector option enables a C/C++ language extension. This
6033       extension provides a new keyword vector which can be used to define
6034       vector type variables. (Note: This is not available when enforcing
6035       strict standard compliance e.g. with -std=c99. Either enable GNU
6036       extensions with e.g. -std=gnu99 or use __vector instead of vector.)
6037       Additionally a set of overloaded builtins is provided which is
6038       partially compatible to the PowerPC Altivec builtins. In order to
6039       make use of these builtins the vecintrin.h header file needs to be
6040       included.
6041
6042                                    GCC 5.3
6043
6044   This is the [40]list of problem reports (PRs) from GCC's bug tracking
6045   system that are known to be fixed in the 5.3 release. This list might
6046   not be complete (that is, it is possible that some PRs that have been
6047   fixed are not listed here).
6048
6049Target Specific Changes
6050
6051  IA-32/x86-64
6052
6053     * GCC now supports the Intel CPU named Skylake with AVX-512
6054       extensions through -march=skylake-avx512. The switch enables the
6055       following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW,
6056       AVX-512DQ.
6057
6058  S/390, System z, IBM z Systems
6059
6060     * With this version of GCC IBM z Systems support has been added to
6061       the GO runtime environment. GCC 5.3 has proven to be able to
6062       compile larger GO applications on IBM z Systems.
6063
6064                                    GCC 5.4
6065
6066   This is the [41]list of problem reports (PRs) from GCC's bug tracking
6067   system that are known to be fixed in the 5.4 release. This list might
6068   not be complete (that is, it is possible that some PRs that have been
6069   fixed are not listed here).
6070
6071                                    GCC 5.5
6072
6073   This is the [42]list of problem reports (PRs) from GCC's bug tracking
6074   system that are known to be fixed in the 5.5 release. This list might
6075   not be complete (that is, it is possible that some PRs that have been
6076   fixed are not listed here).
6077
6078Target Specific Changes
6079
6080  IA-32/x86-64
6081
6082     * Support for the [43]deprecated pcommit instruction has been
6083       removed.
6084
6085
6086    For questions related to the use of GCC, please consult these web
6087    pages and the [44]GCC manuals. If that fails, the
6088    [45]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6089    web pages and the development of GCC are welcome on our developer
6090    list at [46]gcc@gcc.gnu.org. All of [47]our lists have public
6091    archives.
6092
6093   Copyright (C) [48]Free Software Foundation, Inc. Verbatim copying and
6094   distribution of this entire article is permitted in any medium,
6095   provided this notice is preserved.
6096
6097   These pages are [49]maintained by the GCC team. Last modified
6098   2022-03-11[50].
6099
6100References
6101
6102   1. http://gcc.gnu.org/gcc-5/changes.html#libstdcxx
6103   2. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html
6104   3. https://savannah.nongnu.org/bugs/?44574
6105   4. https://gcc.gnu.org/wiki/Intel MPX support in the GCC compiler
6106   5. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.pdf
6107   6. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.Examples.pdf
6108   7. https://gcc.gnu.org/wiki/OpenACC
6109   8. https://gcc.gnu.org/install/configure.html
6110   9. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations
6111  10. http://gcc.gnu.org/projects/cxx1y.html
6112  11. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations
6113  12. http://gcc.gnu.org/projects/cxx1y.html
6114  13. http://gcc.gnu.org/projects/cxx1y.html
6115  14. http://gcc.gnu.org/projects/cxx1y.html
6116  15. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html
6117  16. http://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html
6118  17. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011
6119  18. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014
6120  19. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014
6121  20. https://sourceware.org/gdb/current/onlinedocs/gdb/Xmethods-In-Python.html
6122  21. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html
6123  22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html
6124  23. https://gcc.gnu.org/install/configure.html
6125  24. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html
6126  25. https://gcc.gnu.org/wiki/Fortran2003Status
6127  26. https://gcc.gnu.org/wiki/Fortran2008Status
6128  27. https://gcc.gnu.org/wiki/Coarray
6129  28. http://www.opencoarrays.org/
6130  29. https://pkg.go.dev/cmd/go
6131  30. https://pkg.go.dev/cmd/gofmt
6132  31. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html
6133  32. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html
6134  33. https://github.com/davidmalcolm/pygccjit
6135  34. https://github.com/ibuclaw/gccjitd
6136  35. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html
6137  36. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html
6138  37. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html
6139  38. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html
6140  39. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2
6141  40. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3
6142  41. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.4
6143  42. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.5
6144  43. https://www.intel.com/content/www/us/en/developer/articles/technical/deprecate-pcommit-instruction.html
6145  44. https://gcc.gnu.org/onlinedocs/
6146  45. mailto:gcc-help@gcc.gnu.org
6147  46. mailto:gcc@gcc.gnu.org
6148  47. https://gcc.gnu.org/lists.html
6149  48. https://www.fsf.org/
6150  49. https://gcc.gnu.org/about.html
6151  50. http://validator.w3.org/check/referer
6152======================================================================
6153http://gcc.gnu.org/gcc-4.9/index.html
6154                             GCC 4.9 Release Series
6155
6156   (This release series is no longer supported.)
6157
6158   Aug 3, 2016
6159
6160   The [1]GNU project and the GCC developers are pleased to announce the
6161   release of GCC 4.9.4.
6162
6163   This release is a bug-fix release, containing fixes for regressions in
6164   GCC 4.9.3 relative to previous releases of GCC.
6165
6166Release History
6167
6168   GCC 4.9.4
6169          Aug 3, 2016 ([2]changes, [3]documentation)
6170
6171   GCC 4.9.3
6172          June 26, 2015 ([4]changes, [5]documentation)
6173
6174   GCC 4.9.2
6175          October 30, 2014 ([6]changes, [7]documentation)
6176
6177   GCC 4.9.1
6178          July 16, 2014 ([8]changes, [9]documentation)
6179
6180   GCC 4.9.0
6181          April 22, 2014 ([10]changes, [11]documentation)
6182
6183References and Acknowledgements
6184
6185   GCC used to stand for the GNU C Compiler, but since the compiler
6186   supports several other languages aside from C, it now stands for the
6187   GNU Compiler Collection.
6188
6189   A list of [12]successful builds is updated as new information becomes
6190   available.
6191
6192   The GCC developers would like to thank the numerous people that have
6193   contributed new features, improvements, bug fixes, and other changes as
6194   well as test results to GCC. This [13]amazing group of volunteers is
6195   what makes GCC successful.
6196
6197   For additional information about GCC please refer to the [14]GCC
6198   project web site or contact the [15]GCC development mailing list.
6199
6200   To obtain GCC please use [16]our mirror sites or [17]our version
6201   control system.
6202
6203
6204    For questions related to the use of GCC, please consult these web
6205    pages and the [18]GCC manuals. If that fails, the
6206    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6207    web pages and the development of GCC are welcome on our developer
6208    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
6209    archives.
6210
6211   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
6212   distribution of this entire article is permitted in any medium,
6213   provided this notice is preserved.
6214
6215   These pages are [23]maintained by the GCC team. Last modified
6216   2021-07-28[24].
6217
6218References
6219
6220   1. http://www.gnu.org/
6221   2. http://gcc.gnu.org/gcc-4.9/changes.html
6222   3. http://gcc.gnu.org/onlinedocs/4.9.4/
6223   4. http://gcc.gnu.org/gcc-4.9/changes.html
6224   5. http://gcc.gnu.org/onlinedocs/4.9.3/
6225   6. http://gcc.gnu.org/gcc-4.9/changes.html
6226   7. http://gcc.gnu.org/onlinedocs/4.9.2/
6227   8. http://gcc.gnu.org/gcc-4.9/changes.html
6228   9. http://gcc.gnu.org/onlinedocs/4.9.1/
6229  10. http://gcc.gnu.org/gcc-4.9/changes.html
6230  11. https://gcc.gnu.org/onlinedocs/4.9.0/
6231  12. http://gcc.gnu.org/gcc-4.9/buildstat.html
6232  13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6233  14. http://gcc.gnu.org/index.html
6234  15. mailto:gcc@gcc.gnu.org
6235  16. http://gcc.gnu.org/mirrors.html
6236  17. http://gcc.gnu.org/git.html
6237  18. https://gcc.gnu.org/onlinedocs/
6238  19. mailto:gcc-help@gcc.gnu.org
6239  20. mailto:gcc@gcc.gnu.org
6240  21. https://gcc.gnu.org/lists.html
6241  22. https://www.fsf.org/
6242  23. https://gcc.gnu.org/about.html
6243  24. http://validator.w3.org/check/referer
6244======================================================================
6245http://gcc.gnu.org/gcc-4.9/changes.html
6246                             GCC 4.9 Release Series
6247                        Changes, New Features, and Fixes
6248
6249Caveats
6250
6251     * The mudflap run time checker has been removed. The mudflap options
6252       remain, but do nothing.
6253     * Support for a number of older systems and recently unmaintained or
6254       untested target ports of GCC has been declared obsolete in GCC 4.9.
6255       Unless there is activity to revive them, the next release of GCC
6256       will have their sources permanently removed.
6257       The following ports for individual systems on particular
6258       architectures have been obsoleted:
6259          + Solaris 9 (*-*-solaris2.9). Details can be found in the
6260            [1]announcement.
6261     * On AArch64, the singleton vector types int64x1_t, uint64x1_t and
6262       float64x1_t exported by arm_neon.h are defined to be the same as
6263       their base types. This results in incorrect application of
6264       parameter passing rules to arguments of types int64x1_t and
6265       uint64x1_t, with respect to the AAPCS64 ABI specification. In
6266       addition, names of C++ functions with parameters of these types
6267       (including float64x1_t) are not mangled correctly. The current
6268       typedef declarations also unintentionally allow implicit casting
6269       between singleton vector types and their base types. These issues
6270       will be resolved in a near future release. See [2]PR60825 for more
6271       information.
6272
6273   More information on porting to GCC 4.9 from previous versions of GCC
6274   can be found in the [3]porting guide for this release.
6275
6276General Optimizer Improvements
6277
6278     * AddressSanitizer, a fast memory error detector, is now available on
6279       ARM.
6280     * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior
6281       detector, has been added and can be enabled via
6282       -fsanitize=undefined. Various computations will be instrumented to
6283       detect undefined behavior at runtime. UndefinedBehaviorSanitizer is
6284       currently available for the C and C++ languages.
6285     * Link-time optimization (LTO) improvements:
6286          + Type merging was rewritten. The new implementation is
6287            significantly faster and uses less memory.
6288          + Better partitioning algorithm resulting in less streaming
6289            during link time.
6290          + Early removal of virtual methods reduces the size of object
6291            files and improves link-time memory usage and compile time.
6292          + Function bodies are now loaded on-demand and released early
6293            improving overall memory usage at link time.
6294          + C++ hidden keyed methods can now be optimized out.
6295          + When using a linker plugin, compiling with the -flto option
6296            now generates slim object files (.o) which only contain
6297            intermediate language representation for LTO. Use
6298            -ffat-lto-objects to create files which contain additionally
6299            the object code. To generate static libraries suitable for LTO
6300            processing, use gcc-ar and gcc-ranlib; to list symbols from a
6301            slim object file use gcc-nm. (This requires that ar, ranlib
6302            and nm have been compiled with plugin support.)
6303       Memory usage building Firefox with debug enabled was reduced from
6304       15GB to 3.5GB; link time from 1700 seconds to 350 seconds.
6305     * Inter-procedural optimization improvements:
6306          + New type inheritance analysis module improving
6307            devirtualization. Devirtualization now takes into account
6308            anonymous name-spaces and the C++11 final keyword.
6309          + New speculative devirtualization pass (controlled by
6310            -fdevirtualize-speculatively.
6311          + Calls that were speculatively made direct are turned back to
6312            indirect where direct call is not cheaper.
6313          + Local aliases are introduced for symbols that are known to be
6314            semantically equivalent across shared libraries improving
6315            dynamic linking times.
6316     * Feedback directed optimization improvements:
6317          + Profiling of programs using C++ inline functions is now more
6318            reliable.
6319          + New time profiling determines typical order in which functions
6320            are executed.
6321          + A new function reordering pass (controlled by
6322            -freorder-functions) significantly reduces startup time of
6323            large applications. Until binutils support is completed, it is
6324            effective only with link-time optimization.
6325          + Feedback driven indirect call removal and devirtualization now
6326            handle cross-module calls when link-time optimization is
6327            enabled.
6328
6329New Languages and Language specific improvements
6330
6331     * Version 4.0 of the [4]OpenMP specification is now supported in the
6332       C and C++ compilers and starting with the 4.9.1 release also in the
6333       Fortran compiler. The new -fopenmp-simd option can be used to
6334       enable OpenMP's SIMD directives while ignoring other OpenMP
6335       directives. The new [5]-fsimd-cost-model= option permits to tune
6336       the vectorization cost model for loops annotated with OpenMP and
6337       Cilk Plus simd directives. -Wopenmp-simd warns when the current
6338       cost model overrides simd directives set by the user.
6339     * The -Wdate-time option has been added for the C, C++ and Fortran
6340       compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__
6341       macros are used. Those macros might prevent bit-wise-identical
6342       reproducible compilations.
6343
6344  Ada
6345
6346     * GNAT switched to Ada 2012 instead of Ada 2005 by default.
6347
6348  C family
6349
6350     * Support for colorizing diagnostics emitted by GCC has been added.
6351       The [6]-fdiagnostics-color=auto will enable it when outputting to
6352       terminals, -fdiagnostics-color=always unconditionally. The
6353       GCC_COLORS environment variable can be used to customize the colors
6354       or disable coloring. If GCC_COLORS variable is present in the
6355       environment, the default is -fdiagnostics-color=auto, otherwise
6356       -fdiagnostics-color=never.
6357       Sample diagnostics output:
6358    $ g++ -fdiagnostics-color=always -S -Wall test.C
6359    test.C: In function ‘int foo()’:
6360    test.C:1:14: warning: no return statement in function returning non-void [-W
6361return-type]
6362     int foo () { }
6363                  ^
6364    test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use
6365 -ftemplate-depth= to increase the maximum) instantiating ‘struct X<100>’
6366     template <int N> struct X { static const int value = X<N-1>::value; }; temp
6367late struct X<1000>;
6368                                                  ^
6369    test.C:2:46:   recursively required from ‘const int X<999>::value’
6370    test.C:2:46:   required from ‘const int X<1000>::value’
6371    test.C:2:88:   required from here
6372
6373    test.C:2:46: error: incomplete type ‘X<100>’ used in nested name specifier
6374
6375     * With the new [7]#pragma GCC ivdep, the user can assert that there
6376       are no loop-carried dependencies which would prevent concurrent
6377       execution of consecutive iterations using SIMD (single instruction
6378       multiple data) instructions.
6379     * Support for Cilk Plus has been added and can be enabled with the
6380       -fcilkplus option. Cilk Plus is an extension to the C and C++
6381       languages to support data and task parallelism. The present
6382       implementation follows ABI version 1.2; all features but _Cilk_for
6383       have been implemented.
6384
6385  C
6386
6387     * ISO C11 atomics (the _Atomic type specifier and qualifier and the
6388       <stdatomic.h> header) are now supported.
6389     * ISO C11 generic selections (_Generic keyword) are now supported.
6390     * ISO C11 thread-local storage (_Thread_local, similar to GNU C
6391       __thread) is now supported.
6392     * ISO C11 support is now at a similar level of completeness to ISO
6393       C99 support: substantially complete modulo bugs, extended
6394       identifiers (supported except for corner cases when
6395       -fextended-identifiers is used), floating-point issues (mainly but
6396       not entirely relating to optional C99 features from Annexes F and
6397       G) and the optional Annexes K (Bounds-checking interfaces) and L
6398       (Analyzability).
6399     * A new C extension __auto_type provides a subset of the
6400       functionality of C++11 auto in GNU C.
6401
6402  C++
6403
6404     * The G++ implementation of [8]C++1y return type deduction for normal
6405       functions has been updated to conform to [9]N3638, the proposal
6406       accepted into the working paper. Most notably, it adds
6407       decltype(auto) for getting decltype semantics rather than the
6408       template argument deduction semantics of plain auto:
6409
6410int& f();
6411         auto  i1 = f(); // int
6412decltype(auto) i2 = f(); // int&
6413
6414     * G++ supports [10]C++1y lambda capture initializers:
6415
6416[x = 42]{ ... };
6417
6418       Actually, they have been accepted since GCC 4.5, but now the
6419       compiler doesn't warn about them with -std=c++1y, and supports
6420       parenthesized and brace-enclosed initializers as well.
6421     * G++ supports [11]C++1y variable length arrays. G++ has supported
6422       GNU/C99-style VLAs for a long time, but now additionally supports
6423       initializers and lambda capture by reference. In C++1y mode G++
6424       will complain about VLA uses that are not permitted by the draft
6425       standard, such as forming a pointer to VLA type or applying sizeof
6426       to a VLA variable. Note that it now appears that VLAs will not be
6427       part of C++14, but will be part of a separate document and then
6428       perhaps C++17.
6429
6430void f(int n) {
6431  int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3
6432  [&a]{ for (int i : a) { cout << i << endl; } }();
6433  &a; // error, taking address of VLA
6434}
6435
6436     * G++ supports the [12]C++1y [[deprecated]] attribute modulo bugs in
6437       the underlying [[gnu::deprecated]] attribute. Classes and functions
6438       can be marked deprecated and a diagnostic message added:
6439
6440class A;
6441int bar(int n);
6442#if __cplusplus > 201103
6443class [[deprecated("A is deprecated in C++14; Use B instead")]] A;
6444[[deprecated("bar is unsafe; use foo() instead")]]
6445int bar(int n);
6446
6447int foo(int n);
6448class B;
6449#endif
6450A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead
6451int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo
6452o() instead
6453
6454     * G++ supports [13]C++1y digit separators. Long numeric literals can
6455       be subdivided with a single quote ' to enhance readability:
6456
6457int i = 1048576;
6458int j = 1'048'576;
6459int k = 0x10'0000;
6460int m = 0'004'000'000;
6461int n = 0b0001'0000'0000'0000'0000'0000;
6462
6463double x = 1.602'176'565e-19;
6464double y = 1.602'176'565e-1'9;
6465
6466     * G++ supports [14]C++1y generic (polymorphic) lambdas.
6467
6468// a functional object that will increment any type
6469auto incr = [](auto x) { return x++; };
6470
6471     * As a GNU extension, G++ supports explicit template parameter syntax
6472       for generic lambdas. This can be combined in the expected way with
6473       the standard auto syntax.
6474
6475// a functional object that will add two like-type objects
6476auto add = [] <typename T> (T a, T b) { return a + b; };
6477
6478     * G++ supports unconstrained generic functions as specified by §4.1.2
6479       and §5.1.1 of [15]N3889: Concepts Lite Specification. Briefly, auto
6480       may be used as a type-specifier in a parameter declaration of any
6481       function declarator in order to introduce an implicit function
6482       template parameter, akin to generic lambdas.
6483
6484// the following two function declarations are equivalent
6485auto incr(auto x) { return x++; }
6486template <typename T>
6487auto incr(T x) { return x++; }
6488
6489    Runtime Library (libstdc++)
6490
6491     * [16]Improved support for C++11, including:
6492          + support for <regex>;
6493          + The associative containers in <map> and <set> and the
6494            unordered associative containers in <unordered_map> and
6495            <unordered_set> meet the allocator-aware container
6496            requirements;
6497     * [17]Improved experimental support for the upcoming ISO C++
6498       standard, C++14, including:
6499          + fixing constexpr member functions without const;
6500          + implementation of the std::exchange() utility function;
6501          + addressing tuples by type;
6502          + implemention of std::make_unique;
6503          + implemention of std::shared_lock;
6504          + making std::result_of SFINAE-friendly;
6505          + adding operator() to std::integral_constant;
6506          + adding user-defined literals for standard library types
6507            std::basic_string, std::chrono::duration, and std::complex;
6508          + adding two range overloads to non-modifying sequence oprations
6509            std::equal and std::mismatch;
6510          + adding IO manipulators for quoted strings;
6511          + adding constexpr members to <utility>, <complex>, <chrono>,
6512            and some containers;
6513          + adding compile-time std::integer_sequence;
6514          + adding cleaner transformation traits;
6515          + making <functional>s operator functors easier to use and more
6516            generic;
6517     * An implementation of std::experimental::optional.
6518     * An implementation of std::experimental::string_view.
6519     * The non-standard function std::copy_exception has been deprecated
6520       and will be removed in a future version. std::make_exception_ptr
6521       should be used instead.
6522
6523  Fortran
6524
6525     * Compatibility notice:
6526          + Module files: The version of the module files (.mod) has been
6527            incremented; additionally, module files are now compressed.
6528            Fortran MODULEs compiled by earlier GCC versions have to be
6529            recompiled, when they are USEd by files compiled with GCC 4.9.
6530            GCC 4.9 is not able to read .mod files of earlier GCC
6531            versions; attempting to do so gives an error message. Note:
6532            The ABI of the produced assembler data itself has not changed:
6533            object files and libraries are fully compatible with older
6534            versions (except as stated below).
6535          + ABI changes:
6536               o The [18]argument passing ABI has changed for scalar dummy
6537                 arguments of type INTEGER, REAL, COMPLEX and LOGICAL,
6538                 which have both the VALUE and the OPTIONAL attributes.
6539               o To support finalization the virtual table associated with
6540                 polymorphic variables has changed. Code containing CLASS
6541                 should be recompiled, including all files which define
6542                 derived types involved in the type definition used by
6543                 polymorphic variables. (Note: Due to the incremented
6544                 module version, trying to mix old code with new code will
6545                 usually give an error message.)
6546          + GNU Fortran no longer deallocates allocatable variables or
6547            allocatable components of variables declared in the main
6548            program. Since Fortran 2008, the standard explicitly states
6549            that variables declared in the Fortran main program
6550            automatically have the SAVE attribute.
6551          + When opening files, the close-on-exec flag is set if the
6552            system supports such a feature. This is generally considered
6553            good practice these days, but if there is a need to pass file
6554            descriptors to child processes the parent process must now
6555            remember to clear the close-on-exec flag by calling fcntl(),
6556            e.g. via ISO_C_BINDING, before executing the child process.
6557     * The deprecated command-line option -fno-whole-file has been
6558       removed. (-fwhole-file is the default since GCC 4.6.)
6559       -fwhole-file/-fno-whole-file continue to be accepted but do not
6560       influence the code generation.
6561     * The compiler no longer unconditionally warns about DO loops with
6562       zero iterations. This warning is now controlled by the -Wzerotrip
6563       option, which is implied by -Wall.
6564     * The new NO_ARG_CHECK attribute of the [19]!GCC$ directive can be
6565       used to disable the type-kind-rank (TKR) argument check for a dummy
6566       argument. The feature is similar to ISO/IEC TS 29133:2012's
6567       TYPE(*), except that it additionally also disables the rank check.
6568       Variables with NO_ARG_CHECK have to be dummy arguments and may only
6569       be used as argument to ISO_C_BINDING's C_LOC and as actual argument
6570       to another NO_ARG_CHECK dummy argument; also the other constraints
6571       of TYPE(*) apply. The dummy arguments should be declared as scalar
6572       or assumed-size variable of type type(*) (recommended) – or of type
6573       integer, real, complex or logical. With NO_ARG_CHECK, a pointer to
6574       the data without further type or shape information is passed,
6575       similar to C's void*. Note that also TS 29113's
6576       type(*),dimension(..) accepts arguments of any type and rank;
6577       contrary to NO_ARG_CHECK assumed-rank arguments pass an array
6578       descriptor which contains the array shape and stride of the
6579       argument.
6580     * [20]Fortran 2003:
6581          + Finalization is now supported. It is currently only done for a
6582            subset of those situations in which it should occur.
6583          + Experimental support for scalar character components with
6584            deferred length (i.e. allocatable string length) in derived
6585            types has been added. (Deferred-length character variables are
6586            supported since GCC 4.6.)
6587     * [21]Fortran 2008:
6588          + When STOP or ERROR STOP are used to terminate the execution
6589            and any exception (but inexact) is signaling, a warning is
6590            printed to ERROR_UNIT, indicating which exceptions are
6591            signaling. The [22]-ffpe-summary= command-line option can be
6592            used to fine-tune for which exceptions the warning should be
6593            shown.
6594          + Rounding on input (READ) is now handled on systems where
6595            strtod honours the rounding mode. (For output, rounding is
6596            supported since GCC 4.5.) Note that for input, the compatible
6597            rounding mode is handled as nearest (i.e., rounding to an even
6598            least significant [cf. IEC 60559:1989] for a tie, while
6599            compatible rounds away from zero in that case).
6600
6601  Go
6602
6603     * GCC 4.9 provides a complete implementation of the Go 1.2.1 release.
6604
6605New Targets and Target Specific Improvements
6606
6607  AArch64
6608
6609     * The ARMv8-A crypto and CRC instructions are now supported through
6610       intrinsics. These are enabled when the architecture supports these
6611       and are available through the -march=armv8-a+crc and
6612       -march=armv8-a+crypto options.
6613     * Initial support for ILP32 has now been added to the compiler. This
6614       is now available through the command-line option -mabi=ilp32.
6615       Support for ILP32 is considered experimental as the ABI
6616       specification is still beta.
6617     * Coverage of more of the ISA including the SIMD extensions has been
6618       added. The Advanced SIMD intrinsics have also been improved.
6619     * The new local register allocator (LRA) is now on by default for the
6620       AArch64 backend.
6621     * The REE (Redundant extension elimination) pass has now been enabled
6622       by default for the AArch64 backend.
6623     * Tuning for the Cortex-A53 and Cortex-A57 has been improved.
6624     * Initial big.LITTLE tuning support for the combination of Cortex-A57
6625       and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53
6626       option.
6627     * A number of structural changes have been made to both the ARM and
6628       AArch64 backends to facilitate improved code-generation.
6629     * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769
6630       has been added and can be enabled by giving the
6631       -mfix-cortex-a53-835769 option. Alternatively it can be enabled by
6632       default by configuring GCC with the --enable-fix-cortex-a53-835769
6633       option.
6634
6635  ARC
6636
6637     * A port for Synopsys Designware ARC has been contributed by Embecosm
6638       and Synopsys Inc.
6639
6640  ARM
6641
6642     * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been
6643       disabled by default. This was found to generate better code in only
6644       a small number of cases. It can be turned back on with the
6645       -mneon-for-64bits option.
6646     * Further support for the ARMv8-A architecture, notably implementing
6647       the restriction around IT blocks in the Thumb32 instruction set has
6648       been added. The -mrestrict-it option can be used with
6649       -march=armv7-a or the -march=armv7ve options to make code
6650       generation fully compatible with the deprecated instructions in
6651       ARMv8-A.
6652     * Support has now been added for the ARMv7ve variant of the
6653       architecture. This can be used by the -march=armv7ve option.
6654     * The ARMv8-A crypto and CRC instructions are now supported through
6655       intrinsics and are available through the -march=armv8-a+crc and
6656       mfpu=crypto-neon-fp-armv8 options.
6657     * LRA is now on by default for the ARM target. This can be turned off
6658       using the -mno-lra option. This option is a purely transitionary
6659       command-line option and will be removed in a future release. We are
6660       interested in any bug reports regarding functional and performance
6661       regressions with LRA.
6662     * A new option -mslow-flash-data to improve performance of programs
6663       fetching data on slow flash memory has now been introduced for the
6664       ARMv7-M profile cores.
6665     * A new option -mpic-data-is-text-relative for targets that allows
6666       data segments to be relative to text segments has been added. This
6667       is on by default for all targets except VxWorks RTP.
6668     * A number of infrastructural changes have been made to both the ARM
6669       and AArch64 backends to facilitate improved code-generation.
6670     * GCC now supports Cortex-A12 and the Cortex-R7 through the
6671       -mcpu=cortex-a12 and -mcpu=cortex-r7 options.
6672     * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the
6673       -mcpu=cortex-a57 and -mcpu=cortex-a53 options.
6674     * Initial big.LITTLE tuning support for the combination of Cortex-A57
6675       and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53
6676       option. Similar support was added for the combination of Cortex-A15
6677       and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option.
6678     * Further performance optimizations for the Cortex-A15 and the
6679       Cortex-M4 have been added.
6680     * A number of code generation improvements for Thumb2 to reduce code
6681       size when compiling for the M-profile processors.
6682
6683  AVR
6684
6685     * A new command-line option -mfract-convert-truncate has been added.
6686       It allows compiler to use truncation instead of rounding towards
6687       zero for fractional fixed-point types.
6688
6689  IA-32/x86-64
6690
6691     * -mfpmath=sse is now implied by -ffast-math on all targets where
6692       SSE2 is supported.
6693     * Intel AVX-512 support was added to GCC. That includes inline
6694       assembly support, new registers and extending existing ones, new
6695       intrinsics (covered by corresponding testsuite), and basic
6696       autovectorization. AVX-512 instructions are available via the
6697       following GCC switches: AVX-512 foundation instructions: -mavx512f,
6698       AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and
6699       reciprocal instructions: -mavx512er, AVX-512 conflict detection
6700       instructions: -mavx512cd.
6701     * It is now possible to call x86 intrinsics from select functions in
6702       a file that are tagged with the corresponding target attribute
6703       without having to compile the entire file with the -mxxx option.
6704       This improves the usability of x86 intrinsics and is particularly
6705       useful when doing [23]Function Multiversioning.
6706     * GCC now supports the new Intel microarchitecture named Silvermont
6707       through -march=silvermont.
6708     * GCC now supports the new Intel microarchitecture named Broadwell
6709       through -march=broadwell.
6710     * Optimizing for other Intel microarchitectures have been renamed to
6711       -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell.
6712     * -march=generic has been retuned for better support of Intel core
6713       and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel
6714       Pentium-M, and Pentium4 based CPUs is no longer considered
6715       important for generic.
6716     * -mtune=intel can now be used to generate code running well on the
6717       most current Intel processors, which are Haswell and Silvermont for
6718       GCC 4.9.
6719     * Support to encode 32-bit assembly instructions in 16-bit format is
6720       now available through the -m16 command-line option.
6721     * Better inlining of memcpy and memset that is aware of value ranges
6722       and produces shorter alignment prologues.
6723     * -mno-accumulate-outgoing-args is now honored when unwind
6724       information is output. Argument accumulation is also now turned off
6725       for portions of programs optimized for size.
6726     * Support for new AMD family 15h processors (Excavator core) is now
6727       available through the -march=bdver4 and -mtune=bdver4 options.
6728
6729  MSP430
6730
6731     * A new command-line option -mcpu= has been added to the MSP430
6732       backend. This option is used to specify the ISA to be used.
6733       Accepted values are msp430 (the default), msp430x and msp430xv2.
6734       The ISA is no longer deduced from the -mmcu= option as there are
6735       far too many different MCU names. The -mmcu= option is still
6736       supported, and this is still used to select linker scripts and
6737       generate a C preprocessor symbol that will be recognised by the
6738       msp430.h header file.
6739
6740  NDS32
6741
6742     * A new nds32 port supports the 32-bit architecture from Andes
6743       Technology Corporation.
6744     * The port provides initial support for the V2, V3, V3m instruction
6745       set architectures.
6746
6747  Nios II
6748
6749     * A port for the Altera Nios II has been contributed by Mentor
6750       Graphics.
6751
6752  PowerPC / PowerPC64 / RS6000
6753
6754     * GCC now supports Power ISA 2.07, which includes support for
6755       Hardware Transactional Memory (HTM), Quadword atomics and several
6756       VMX and VSX additions, including Crypto, 64-bit integer, 128-bit
6757       integer and decimal integer operations.
6758     * Support for the POWER8 processor is now available through the
6759       -mcpu=power8 and -mtune=power8 options.
6760     * The libitm library has been modified to add a HTM fastpath that
6761       automatically uses POWER's HTM hardware instructions when it is
6762       executing on a HTM enabled processor.
6763     * Support for the new powerpc64le-linux platform has been added. It
6764       defaults to generating code that conforms to the ELFV2 ABI.
6765
6766  S/390, System z
6767
6768     * Support for the Transactional Execution Facility included with the
6769       IBM zEnterprise zEC12 processor has been added. A set of GCC style
6770       builtins as well as XLC style builtins are provided. The builtins
6771       are enabled by default when using the -march=zEC12 option but can
6772       explicitly be disabled with -mno-htm. Using the GCC builtins also
6773       libitm supports hardware transactions on S/390.
6774     * The hotpatch features allows to prepare functions for hotpatching.
6775       A certain amount of bytes is reserved before the function entry
6776       label plus a NOP is inserted at its very beginning to implement a
6777       backward jump when applying a patch. The feature can either be
6778       enabled per compilation unit via the command-line option -mhotpatch
6779       or per function using the hotpatch attribute.
6780     * The shrink wrap optimization is now supported on S/390 and enabled
6781       by default.
6782     * A major rework of the routines to determine which registers need to
6783       be saved and restored in function prologue/epilogue now allow to
6784       use floating point registers as save slots. This will happen for
6785       certain leaf function with -march=z10 or higher.
6786     * The LRA rtl pass replaces reload by default on S/390.
6787
6788  RX
6789
6790     * The port now allows to specify the RX100, RX200, and RX600
6791       processors with the command-line options -mcpu=rx100, -mcpu=rx200
6792       and -mcpu=rx600.
6793
6794  SH
6795
6796     * Minor improvements to code generated for integer arithmetic and
6797       code that involves the T bit.
6798     * Added support for the SH2A clips and clipu instructions. The
6799       compiler will now try to utilize them for min/max expressions such
6800       as max (-128, min (127, x)).
6801     * Added support for the cmp/str instruction through built-in
6802       functions such as __builtin_strlen. When not optimizing for size,
6803       the compiler will now expand calls to e.g. strlen as an inlined
6804       sequences which utilize the cmp/str instruction.
6805     * Improved code generated around volatile memory loads and stores.
6806     * The option -mcbranchdi has been deprecated. Specifying it will
6807       result in a warning and will not influence code generation.
6808     * The option -mcmpeqdi has been deprecated. Specifying it will result
6809       in a warning and will not influence code generation.
6810
6811GCC 4.9.1
6812
6813   This is the [24]list of problem reports (PRs) from GCC's bug tracking
6814   system that are known to be fixed in the 4.9.1 release. This list might
6815   not be complete (that is, it is possible that some PRs that have been
6816   fixed are not listed here).
6817
6818   Version 4.0 of the OpenMP specification is supported even in Fortran,
6819   not just C and C++.
6820
6821GCC 4.9.2
6822
6823   This is the [25]list of problem reports (PRs) from GCC's bug tracking
6824   system that are known to be fixed in the 4.9.2 release. This list might
6825   not be complete (that is, it is possible that some PRs that have been
6826   fixed are not listed here).
6827
6828GCC 4.9.3
6829
6830   This is the [26]list of problem reports (PRs) from GCC's bug tracking
6831   system that are known to be fixed in the 4.9.3 release. This list might
6832   not be complete (that is, it is possible that some PRs that have been
6833   fixed are not listed here).
6834
6835GCC 4.9.4
6836
6837   This is the [27]list of problem reports (PRs) from GCC's bug tracking
6838   system that are known to be fixed in the 4.9.4 release. This list might
6839   not be complete (that is, it is possible that some PRs that have been
6840   fixed are not listed here).
6841
6842
6843    For questions related to the use of GCC, please consult these web
6844    pages and the [28]GCC manuals. If that fails, the
6845    [29]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6846    web pages and the development of GCC are welcome on our developer
6847    list at [30]gcc@gcc.gnu.org. All of [31]our lists have public
6848    archives.
6849
6850   Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and
6851   distribution of this entire article is permitted in any medium,
6852   provided this notice is preserved.
6853
6854   These pages are [33]maintained by the GCC team. Last modified
6855   2021-07-28[34].
6856
6857References
6858
6859   1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html
6860   2. https://gcc.gnu.org/PR60825
6861   3. https://gcc.gnu.org/gcc-4.9/porting_to.html
6862   4. https://www.openmp.org/specifications/
6863   5. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908
6864   6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252
6865   7. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html
6866   8. http://gcc.gnu.org/projects/cxx1y.html
6867   9. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html
6868  10. http://gcc.gnu.org/projects/cxx1y.html
6869  11. http://gcc.gnu.org/projects/cxx1y.html
6870  12. http://gcc.gnu.org/projects/cxx1y.html
6871  13. http://gcc.gnu.org/projects/cxx1y.html
6872  14. http://gcc.gnu.org/projects/cxx1y.html
6873  15. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf
6874  16. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011
6875  17. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014
6876  18. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html
6877  19. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html
6878  20. https://gcc.gnu.org/wiki/Fortran2003Status
6879  21. https://gcc.gnu.org/wiki/Fortran2008Status
6880  22. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html
6881  23. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html
6882  24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1
6883  25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2
6884  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3
6885  27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.4
6886  28. https://gcc.gnu.org/onlinedocs/
6887  29. mailto:gcc-help@gcc.gnu.org
6888  30. mailto:gcc@gcc.gnu.org
6889  31. https://gcc.gnu.org/lists.html
6890  32. https://www.fsf.org/
6891  33. https://gcc.gnu.org/about.html
6892  34. http://validator.w3.org/check/referer
6893======================================================================
6894http://gcc.gnu.org/gcc-4.8/index.html
6895                             GCC 4.8 Release Series
6896
6897   (This release series is no longer supported.)
6898
6899   June 23, 2015
6900
6901   The [1]GNU project and the GCC developers are pleased to announce the
6902   release of GCC 4.8.5.
6903
6904   This release is a bug-fix release, containing fixes for regressions in
6905   GCC 4.8.4 relative to previous releases of GCC.
6906
6907Release History
6908
6909   GCC 4.8.5
6910          June 23, 2015 ([2]changes, [3]documentation)
6911
6912   GCC 4.8.4
6913          December 19, 2014 ([4]changes, [5]documentation)
6914
6915   GCC 4.8.3
6916          May 22, 2014 ([6]changes, [7]documentation)
6917
6918   GCC 4.8.2
6919          October 16, 2013 ([8]changes, [9]documentation)
6920
6921   GCC 4.8.1
6922          May 31, 2013 ([10]changes, [11]documentation)
6923
6924   GCC 4.8.0
6925          March 22, 2013 ([12]changes, [13]documentation)
6926
6927References and Acknowledgements
6928
6929   GCC used to stand for the GNU C Compiler, but since the compiler
6930   supports several other languages aside from C, it now stands for the
6931   GNU Compiler Collection.
6932
6933   A list of [14]successful builds is updated as new information becomes
6934   available.
6935
6936   The GCC developers would like to thank the numerous people that have
6937   contributed new features, improvements, bug fixes, and other changes as
6938   well as test results to GCC. This [15]amazing group of volunteers is
6939   what makes GCC successful.
6940
6941   For additional information about GCC please refer to the [16]GCC
6942   project web site or contact the [17]GCC development mailing list.
6943
6944   To obtain GCC please use [18]our mirror sites or [19]our version
6945   control system.
6946
6947
6948    For questions related to the use of GCC, please consult these web
6949    pages and the [20]GCC manuals. If that fails, the
6950    [21]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6951    web pages and the development of GCC are welcome on our developer
6952    list at [22]gcc@gcc.gnu.org. All of [23]our lists have public
6953    archives.
6954
6955   Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and
6956   distribution of this entire article is permitted in any medium,
6957   provided this notice is preserved.
6958
6959   These pages are [25]maintained by the GCC team. Last modified
6960   2021-07-28[26].
6961
6962References
6963
6964   1. http://www.gnu.org/
6965   2. http://gcc.gnu.org/gcc-4.8/changes.html
6966   3. https://gcc.gnu.org/onlinedocs/4.8.5/
6967   4. http://gcc.gnu.org/gcc-4.8/changes.html
6968   5. https://gcc.gnu.org/onlinedocs/4.8.4/
6969   6. http://gcc.gnu.org/gcc-4.8/changes.html
6970   7. https://gcc.gnu.org/onlinedocs/4.8.3/
6971   8. http://gcc.gnu.org/gcc-4.8/changes.html
6972   9. https://gcc.gnu.org/onlinedocs/4.8.2/
6973  10. http://gcc.gnu.org/gcc-4.8/changes.html
6974  11. https://gcc.gnu.org/onlinedocs/4.8.1/
6975  12. http://gcc.gnu.org/gcc-4.8/changes.html
6976  13. https://gcc.gnu.org/onlinedocs/4.8.0/
6977  14. http://gcc.gnu.org/gcc-4.8/buildstat.html
6978  15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6979  16. http://gcc.gnu.org/index.html
6980  17. mailto:gcc@gcc.gnu.org
6981  18. http://gcc.gnu.org/mirrors.html
6982  19. http://gcc.gnu.org/git.html
6983  20. https://gcc.gnu.org/onlinedocs/
6984  21. mailto:gcc-help@gcc.gnu.org
6985  22. mailto:gcc@gcc.gnu.org
6986  23. https://gcc.gnu.org/lists.html
6987  24. https://www.fsf.org/
6988  25. https://gcc.gnu.org/about.html
6989  26. http://validator.w3.org/check/referer
6990======================================================================
6991http://gcc.gnu.org/gcc-4.8/changes.html
6992                             GCC 4.8 Release Series
6993                        Changes, New Features, and Fixes
6994
6995Caveats
6996
6997   GCC now uses C++ as its implementation language. This means that to
6998   build GCC from sources, you will need a C++ compiler that understands
6999   C++ 2003. For more details on the rationale and specific changes,
7000   please refer to the [1]C++ conversion page.
7001
7002   To enable the Graphite framework for loop optimizations you now need
7003   CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from
7004   the [2]GCC infrastructure directory. The installation manual contains
7005   more information about requirements to build GCC.
7006
7007   GCC now uses a more aggressive analysis to derive an upper bound for
7008   the number of iterations of loops using constraints imposed by language
7009   standards. This may cause non-conforming programs to no longer work as
7010   expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new
7011   option, -fno-aggressive-loop-optimizations, was added to disable this
7012   aggressive analysis. In some loops that have known constant number of
7013   iterations, but undefined behavior is known to occur in the loop before
7014   reaching or during the last iteration, GCC will warn about the
7015   undefined behavior in the loop instead of deriving lower upper bound of
7016   the number of iterations for the loop. The warning can be disabled with
7017   -Wno-aggressive-loop-optimizations.
7018
7019   On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules
7020   for the layout of vectors that could lead to wrong code being
7021   generated. Vectors larger than 8 bytes in size are now by default
7022   aligned to an 8-byte boundary. This is an ABI change: code that makes
7023   explicit use of vector types may be incompatible with binary objects
7024   built with older versions of GCC. Auto-vectorized code is not affected
7025   by this change.
7026
7027   On AVR, support has been removed for the command-line option
7028   -mshort-calls deprecated in GCC 4.7.
7029
7030   On AVR, the configure option --with-avrlibc supported since GCC 4.7.2
7031   is turned on per default for all non-RTEMS configurations. This option
7032   arranges for a better integration of [3]AVR Libc with avr-gcc. For
7033   technical details, see [4]PR54461. To turn off the option in non-RTEMS
7034   configurations, use --with-avrlibc=no. If the compiler is configured
7035   for RTEMS, the option is always turned off.
7036
7037   More information on porting to GCC 4.8 from previous versions of GCC
7038   can be found in the [5]porting guide for this release.
7039
7040General Optimizer Improvements (and Changes)
7041
7042     * DWARF4 is now the default when generating DWARF debug information.
7043       When -g is used on a platform that uses DWARF debugging
7044       information, GCC will now default to -gdwarf-4
7045       -fno-debug-types-section.
7046       GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information
7047       consumers support DWARF4 by default. Before GCC 4.8 the default
7048       version used was DWARF2. To make GCC 4.8 generate an older DWARF
7049       version use -g together with -gdwarf-2 or -gdwarf-3. The default
7050       for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf.
7051     * A new general optimization level, -Og, has been introduced. It
7052       addresses the need for fast compilation and a superior debugging
7053       experience while providing a reasonable level of run-time
7054       performance. Overall experience for development should be better
7055       than the default optimization level -O0.
7056     * A new option -ftree-partial-pre was added to control the partial
7057       redundancy elimination (PRE) optimization. This option is enabled
7058       by default at the -O3 optimization level, and it makes PRE more
7059       aggressive.
7060     * The option -fconserve-space has been removed; it was no longer
7061       useful on most targets since GCC supports putting variables into
7062       BSS without making them common.
7063     * The struct reorg and matrix reorg optimizations (command-line
7064       options -fipa-struct-reorg and -fipa-matrix-reorg) have been
7065       removed. They did not always work correctly, nor did they work with
7066       link-time optimization (LTO), hence were only applicable to
7067       programs consisting of a single translation unit.
7068     * Several scalability bottle-necks have been removed from GCC's
7069       optimization passes. Compilation of extremely large functions, e.g.
7070       due to the use of the flatten attribute in the "Eigen" C++ linear
7071       algebra templates library, is significantly faster than previous
7072       releases of GCC.
7073     * Link-time optimization (LTO) improvements:
7074          + LTO partitioning has been rewritten for better reliability and
7075            maintanibility. Several important bugs leading to link
7076            failures have been fixed.
7077     * Interprocedural optimization improvements:
7078          + A new symbol table has been implemented. It builds on existing
7079            callgraph and varpool modules and provide a new API. Unusual
7080            symbol visibilities and aliases are handled more consistently
7081            leading to, for example, more aggressive unreachable code
7082            removal with LTO.
7083          + The inline heuristic can now bypass limits on the size of of
7084            inlined functions when the inlining is particularly
7085            profitable. This happens, for example, when loop bounds or
7086            array strides get propagated.
7087          + Values passed through aggregates (either by value or
7088            reference) are now propagated at the inter-procedural level
7089            leading to better inlining decisions (for example in the case
7090            of Fortran array descriptors) and devirtualization.
7091     * [6]AddressSanitizer , a fast memory error detector, has been added
7092       and can be enabled via -fsanitize=address. Memory access
7093       instructions will be instrumented to detect heap-, stack-, and
7094       global-buffer overflow as well as use-after-free bugs. To get nicer
7095       stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is
7096       available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on
7097       x86-64 Darwin.
7098     * [7]ThreadSanitizer has been added and can be enabled via
7099       -fsanitize=thread. Instructions will be instrumented to detect data
7100       races. The ThreadSanitizer is available on x86-64 GNU/Linux.
7101     * A new local register allocator (LRA) has been implemented, which
7102       replaces the 26 year old reload pass and improves generated code
7103       quality. For now it is active on the IA-32 and x86-64 targets.
7104     * Support for transactional memory has been implemented on the
7105       following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and
7106       Alpha.
7107
7108New Languages and Language specific improvements
7109
7110  C family
7111
7112     * Each diagnostic emitted now includes the original source line and a
7113       caret '^' indicating the column. The option
7114       -fno-diagnostics-show-caret suppresses this information.
7115     * The option -ftrack-macro-expansion=2 is now enabled by default.
7116       This allows the compiler to display the macro expansion stack in
7117       diagnostics. Combined with the caret information, an example
7118       diagnostic showing these two features is:
7119
7120t.c:1:94: error: invalid operands to binary < (have ‘struct mystruct’ and ‘float
7121’)
7122 #define MYMAX(A,B)    __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _
7123_b = (B); __a < __b ? __b : __a; })
7124
7125              ^
7126t.c:7:7: note: in expansion of macro 'MYMAX'
7127   X = MYMAX(P, F);
7128       ^
7129
7130     * A new -Wsizeof-pointer-memaccess warning has been added (also
7131       enabled by -Wall) to warn about suspicious length parameters to
7132       certain string and memory built-in functions if the argument uses
7133       sizeof. This warning warns e.g. about memset (ptr, 0, sizeof
7134       (ptr)); if ptr is not an array, but a pointer, and suggests a
7135       possible fix, or about memcpy (&foo, ptr, sizeof (&foo));.
7136     * The new option -Wpedantic is an alias for -pedantic, which is now
7137       deprecated. The forms -Wno-pedantic, -Werror=pedantic, and
7138       -Wno-error=pedantic work in the same way as for any other -W
7139       option. One caveat is that -Werror=pedantic is not equivalent to
7140       -pedantic-errors, since the latter makes into errors some warnings
7141       that are not controlled by -Wpedantic, and the former only affects
7142       diagnostics that are disabled when using -Wno-pedantic.
7143     * The option -Wshadow no longer warns if a declaration shadows a
7144       function declaration, unless the former declares a function or
7145       pointer to function, because this is [8]a common and valid case in
7146       real-world code.
7147
7148  C++
7149
7150     * G++ now implements the [9]C++11 thread_local keyword; this differs
7151       from the GNU __thread keyword primarily in that it allows dynamic
7152       initialization and destruction semantics. Unfortunately, this
7153       support requires a run-time penalty for references to
7154       non-function-local thread_local variables defined in a different
7155       translation unit even if they don't need dynamic initialization, so
7156       users may want to continue to use __thread for TLS variables with
7157       static initialization semantics.
7158       If the programmer can be sure that no use of the variable in a
7159       non-defining TU needs to trigger dynamic initialization (either
7160       because the variable is statically initialized, or a use of the
7161       variable in the defining TU will be executed before any uses in
7162       another TU), they can avoid this overhead with the
7163       -fno-extern-tls-init option.
7164       OpenMP threadprivate variables now also support dynamic
7165       initialization and destruction by the same mechanism.
7166     * G++ now implements the [10]C++11 attribute syntax, e.g.
7167
7168[[noreturn]] void f();
7169
7170       and also the alignment specifier, e.g.
7171
7172alignas(double) int i;
7173
7174     * G++ now implements [11]C++11 inheriting constructors, e.g.
7175
7176struct A { A(int); };
7177struct B: A { using A::A; }; // defines B::B(int)
7178B b(42); // OK
7179
7180     * As of GCC 4.8.1, G++ implements the change to decltype semantics
7181       from [12]N3276.
7182
7183struct A f();
7184decltype(f()) g();    // OK, return type of f() is not required to be complete.
7185
7186     * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g.
7187
7188struct A { int f() &; };
7189int i = A().f();  // error, f() requires an lvalue object
7190
7191     * G++ now supports a -std=c++1y option for experimentation with
7192       features proposed for the next revision of the standard, expected
7193       around 2014. Currently the only difference from -std=c++11 is
7194       support for return type deduction in normal functions, as proposed
7195       in [14]N3386. Status of C++1y features in GCC 4.8 can be found
7196       [15]here.
7197     * The G++ namespace association extension, __attribute ((strong)),
7198       has been deprecated. Inline namespaces should be used instead.
7199     * G++ now supports a -fext-numeric-literal option to control whether
7200       GNU numeric literal suffixes are accepted as extensions or
7201       processed as C++11 user-defined numeric literal suffixes. The flag
7202       is on (use suffixes for GNU literals) by default for -std=gnu++*,
7203       and -std=c++98. The flag is off (use suffixes for user-defined
7204       literals) by default for -std=c++11 and later.
7205
7206    Runtime Library (libstdc++)
7207
7208     * [16]Improved experimental support for the new ISO C++ standard,
7209       C++11, including:
7210          + forward_list meets the allocator-aware container requirements;
7211          + this_thread::sleep_for(), this_thread::sleep_until() and
7212            this_thread::yield() are defined without requiring the
7213            configure option --enable-libstdcxx-time;
7214     * Improvements to <random>:
7215          + SSE optimized normal_distribution.
7216          + Use of hardware RNG instruction for random_device on new x86
7217            processors (requires the assembler to support the
7218            instruction.)
7219       and <ext/random>:
7220          + New random number engine simd_fast_mersenne_twister_engine
7221            with an optimized SSE implementation.
7222          + New random number distributions beta_distribution,
7223            normal_mv_distribution, rice_distribution,
7224            nakagami_distribution, pareto_distribution, k_distribution,
7225            arcsine_distribution, hoyt_distribution.
7226     * Added --disable-libstdcxx-verbose configure option to disable
7227       diagnostic messages issued when a process terminates abnormally.
7228       This may be useful for embedded systems to reduce the size of
7229       executables that link statically to the library.
7230
7231  Fortran
7232
7233     * Compatibility notice:
7234          + Module files: The version of module files (.mod) has been
7235            incremented. Fortran MODULEs compiled by earlier GCC versions
7236            have to be recompiled, when they are USEd by files compiled
7237            with GCC 4.8. GCC 4.8 is not able to read .mod files created
7238            by earlier versions; attempting to do so gives an error
7239            message.
7240            Note: The ABI of the produced assembler data itself has not
7241            changed; object files and libraries are fully compatible with
7242            older versions except as noted below.
7243          + ABI: Some internal names (used in the assembler/object file)
7244            have changed for symbols declared in the specification part of
7245            a module. If an affected module – or a file using it via use
7246            association – is recompiled, the module and all files which
7247            directly use such symbols have to be recompiled as well. This
7248            change only affects the following kind of module symbols:
7249               o Procedure pointers. Note: C-interoperable function
7250                 pointers (type(c_funptr)) are not affected nor are
7251                 procedure-pointer components.
7252               o Deferred-length character strings.
7253     * The [17]BACKTRACE intrinsic subroutine has been added. It shows a
7254       backtrace at an arbitrary place in user code; program execution
7255       continues normally afterwards.
7256     * The [18]-Wc-binding-type warning option has been added (disabled by
7257       default). It warns if the a variable might not be C interoperable;
7258       in particular, if the variable has been declared using an intrinsic
7259       type with default kind instead of using a kind parameter defined
7260       for C interoperability in the intrinsic ISO_C_Binding module.
7261       Before, this warning was always printed. The -Wc-binding-type
7262       option is enabled by -Wall.
7263     * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line
7264       options have been added, which diagnose when code is inserted for
7265       automatic (re)allocation of a variable during assignment. This
7266       option can be used to decide whether it is safe to use
7267       [20]-fno-realloc-lhs. Additionally, it can be used to find
7268       automatic (re)allocation in hot loops. (For arrays, replacing
7269       "var=" by "var(:)=" disables the automatic reallocation.)
7270     * The [21]-Wcompare-reals command-line option has been added. When
7271       this is set, warnings are issued when comparing REAL or COMPLEX
7272       types for equality and inequality; consider replacing a == b by
7273       abs(a−b) < eps with a suitable eps. -Wcompare-reals is enabled by
7274       -Wextra.
7275     * The [22]-Wtarget-lifetime command-line option has been added
7276       (enabled with -Wall), which warns if the pointer in a pointer
7277       assignment might outlive its target.
7278     * Reading floating point numbers which use "q" for the exponential
7279       (such as 4.0q0) is now supported as vendor extension for better
7280       compatibility with old data files. It is strongly recommended to
7281       use for I/O the equivalent but standard conforming "e" (such as
7282       4.0e0).
7283       (For Fortran source code, consider replacing the "q" in
7284       floating-point literals by a kind parameter (e.g. 4.0e0_qp with a
7285       suitable qp). Note that – in Fortran source code – replacing "q" by
7286       a simple "e" is not equivalent.)
7287     * The GFORTRAN_TMPDIR environment variable for specifying a
7288       non-default directory for files opened with STATUS="SCRATCH", is
7289       not used anymore. Instead gfortran checks the POSIX/GNU standard
7290       TMPDIR environment variable. If TMPDIR is not defined, gfortran
7291       falls back to other methods to determine the directory for
7292       temporary files as documented in the [23]user manual.
7293     * [24]Fortran 2003:
7294          + Support for unlimited polymorphic variables (CLASS(*)) has
7295            been added. Nonconstant character lengths are not yet
7296            supported.
7297     * [25]TS 29113:
7298          + Assumed types (TYPE(*)) are now supported.
7299          + Experimental support for assumed-rank arrays (dimension(..))
7300            has been added. Note that currently gfortran's own array
7301            descriptor is used, which is different from the one defined in
7302            TS29113, see [26]gfortran's header file or use the [27]Chasm
7303            Language Interoperability Tools.
7304
7305  Go
7306
7307     * GCC 4.8.2 provides a complete implementation of the Go 1.1.2
7308       release.
7309     * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1
7310       release. The library support is not quite complete.
7311     * Go has been tested on GNU/Linux and Solaris platforms for various
7312       processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may
7313       work on other platforms as well.
7314
7315New Targets and Target Specific Improvements
7316
7317  AArch64
7318
7319     * A new port has been added to support AArch64, the new 64-bit
7320       architecture from ARM. Note that this is a separate port from the
7321       existing 32-bit ARM port.
7322     * The port provides initial support for the Cortex-A53 and the
7323       Cortex-A57 processors with the command line options
7324       -mcpu=cortex-a53 and -mcpu=cortex-a57.
7325     * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769
7326       has been added and can be enabled by giving the
7327       -mfix-cortex-a53-835769 option. Alternatively it can be enabled by
7328       default by configuring GCC with the --enable-fix-cortex-a53-835769
7329       option.
7330
7331  ARM
7332
7333     * Initial support has been added for the AArch32 extensions defined
7334       in the ARMv8 architecture.
7335     * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs.
7336     * A new option, -mcpu=marvell-pj4, has been added to generate code
7337       for the Marvell PJ4 processor.
7338     * The compiler can now automatically generate the VFMA, VFMS, REVSH
7339       and REV16 instructions.
7340     * A new vectorizer cost model for Advanced SIMD configurations to
7341       improve the auto-vectorization strategies used.
7342     * The scheduler now takes into account the number of live registers
7343       to reduce the amount of spilling that can occur. This should
7344       improve code performance in large functions. The limit can be
7345       removed by using the option -fno-sched-pressure.
7346     * Improvements have been made to the Marvell iWMMX code generation
7347       and support for the iWMMX2 SIMD unit has been added. The option
7348       -mcpu=iwmmxt2 can be used to enable code generation for the latter.
7349     * A number of code generation improvements for Thumb2 to reduce code
7350       size when compiling for the M-profile processors.
7351     * The RTEMS (arm-rtems) port has been updated to use the EABI.
7352     * Code generation support for the old FPA and Maverick floating-point
7353       architectures has been removed. Ports that previously relied on
7354       these features have also been removed. This includes the targets:
7355          + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
7356          + arm*-*-elf (use arm*-*-eabi)
7357          + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
7358          + arm*-*-ecos-elf (no alternative)
7359          + arm*-*-freebsd (no alternative)
7360          + arm*-wince-pe* (no alternative).
7361
7362  AVR
7363
7364     * Support for the "Embedded C" fixed-point has been added. For
7365       details, see the [28]GCC wiki and the [29]user manual. The support
7366       is not complete.
7367     * A new print modifier %r for register operands in inline assembler
7368       is supported. It will print the raw register number without the
7369       register prefix 'r':
7370    /* Return the most significant byte of 'val', a 64-bit value.  */
7371
7372    unsigned char msb (long long val)
7373    {
7374      unsigned char c;
7375      __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val));
7376      return c;
7377    }
7378       The inline assembler in this example will generate code like
7379    mov r24, 8+7
7380       provided c is allocated to R24 and val is allocated to R8…R15. This
7381       works because the GNU assembler accepts plain register numbers
7382       without register prefix.
7383     * Static initializers with 3-byte symbols are supported now:
7384    extern const __memx char foo;
7385    const __memx void *pfoo = &foo;
7386       This requires at least Binutils 2.23.
7387
7388  IA-32/x86-64
7389
7390     * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with
7391       SSE extensions disabled. Since the x86-64 ABI requires 16 byte
7392       stack alignment, this is ABI incompatible and intended to be used
7393       in controlled environments where stack space is an important
7394       limitation. This option will lead to wrong code when functions
7395       compiled with 16 byte stack alignment (such as functions from a
7396       standard library) are called with misaligned stack. In this case,
7397       SSE instructions may lead to misaligned memory access traps. In
7398       addition, variable arguments will be handled incorrectly for 16
7399       byte aligned objects (including x87 long double and __int128),
7400       leading to wrong results. You must build all modules with
7401       -mpreferred-stack-boundary=3, including any libraries. This
7402       includes the system libraries and startup modules.
7403     * Support for the new Intel processor codename Broadwell with RDSEED,
7404       ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw,
7405       -mrdseed command-line options.
7406     * Support for the Intel RTM and HLE intrinsics, built-in functions
7407       and code generation is available via -mrtm and -mhle.
7408     * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets.
7409       Intrinsics and built-in functions are available via -mfxsr, -mxsave
7410       and -mxsaveopt respectively.
7411     * New -maddress-mode=[short|long] options for x32.
7412       -maddress-mode=short overrides default 64-bit addresses to 32-bit
7413       by emitting the 0x67 address-size override prefix. This is the
7414       default address mode for x32.
7415     * New built-in functions to detect run-time CPU type and ISA:
7416          + A built-in function __builtin_cpu_is has been added to detect
7417            if the run-time CPU is of a particular type. It returns a
7418            positive integer on a match and zero otherwise. It accepts one
7419            string literal argument, the CPU name. For example,
7420            __builtin_cpu_is("westmere") returns a positive integer if the
7421            run-time CPU is an Intel Core i7 Westmere processor. Please
7422            refer to the [30]user manual for the list of valid CPU names
7423            recognized.
7424          + A built-in function __builtin_cpu_supports has been added to
7425            detect if the run-time CPU supports a particular ISA feature.
7426            It returns a positive integer on a match and zero otherwise.
7427            It accepts one string literal argument, the ISA feature. For
7428            example, __builtin_cpu_supports("ssse3") returns a positive
7429            integer if the run-time CPU supports SSSE3 instructions.
7430            Please refer to the [31]user manual for the list of valid ISA
7431            names recognized.
7432       Caveat: If these built-in functions are called before any static
7433       constructors are invoked, like during IFUNC initialization, then
7434       the CPU detection initialization must be explicitly run using this
7435       newly provided built-in function, __builtin_cpu_init. The
7436       initialization needs to be done only once. For example, this is how
7437       the invocation would look like inside an IFUNC initializer:
7438    static void (*some_ifunc_resolver(void))(void)
7439    {
7440      __builtin_cpu_init();
7441      if (__builtin_cpu_is("amdfam10h") ...
7442      if (__builtin_cpu_supports("popcnt") ...
7443    }
7444
7445     * Function Multiversioning Support with G++:
7446       It is now possible to create multiple function versions each
7447       targeting a specific processor and/or ISA. Function versions have
7448       the same signature but different target attributes. For example,
7449       here is a program with function versions:
7450    __attribute__ ((target ("default")))
7451    int foo(void)
7452    {
7453      return 1;
7454    }
7455
7456    __attribute__ ((target ("sse4.2")))
7457    int foo(void)
7458    {
7459      return 2;
7460    }
7461
7462    int main (void)
7463    {
7464      int (*p) = &foo;
7465      assert ((*p)() == foo());
7466      return 0;
7467    }
7468
7469       Please refer to this [32]wiki for more information.
7470     * The x86 back end has been improved to allow option -fschedule-insns
7471       to work reliably. This option can be used to schedule instructions
7472       better and leads to improved performace in certain cases.
7473     * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437
7474       from the Mingw-w64 trunk.
7475     * Support for new AMD family 15h processors (Steamroller core) is now
7476       available through the -march=bdver3 and -mtune=bdver3 options.
7477     * Support for new AMD family 16h processors (Jaguar core) is now
7478       available through the -march=btver2 and -mtune=btver2 options.
7479
7480  FRV
7481
7482     * This target now supports the -fstack-usage command-line option.
7483
7484  MIPS
7485
7486     * GCC can now generate code specifically for the R4700, Broadcom XLP
7487       and MIPS 34kn processors. The associated -march options are
7488       -march=r4700, -march=xlp and -march=34kn respectively.
7489     * GCC now generates better DSP code for MIPS 74k cores thanks to
7490       further scheduling optimizations.
7491     * The MIPS port now supports the -fstack-check option.
7492     * GCC now passes the -mmcu and -mno-mcu options to the assembler.
7493     * Previous versions of GCC would silently accept -fpic and -fPIC for
7494       -mno-abicalls targets like mips*-elf. This combination was not
7495       intended or supported, and did not generate position-independent
7496       code. GCC 4.8 now reports an error when this combination is used.
7497
7498  PowerPC / PowerPC64 / RS6000
7499
7500     * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save,
7501       restore or update the VRSAVE register by default. The respective
7502       operating systems manage the VRSAVE register directly.
7503     * Large TOC support has been added for AIX through the command line
7504       option -mcmodel=large.
7505     * Native Thread-Local Storage support has been added for AIX.
7506     * VMX (Altivec) and VSX instruction sets now are enabled implicitly
7507       when targetting processors that support those hardware features on
7508       AIX 6.1 and above.
7509
7510  RX
7511
7512     * This target will now issue a warning message whenever multiple fast
7513       interrupt handlers are found in the same compilation unit. This
7514       feature can be turned off by the new
7515       -mno-warn-multiple-fast-interrupts command-line option.
7516
7517  S/390, System z
7518
7519     * Support for the IBM zEnterprise zEC12 processor has been added.
7520       When using the -march=zEC12 option, the compiler will generate code
7521       making use of the following new instructions:
7522          + load and trap instructions
7523          + 2 new compare and trap instructions
7524          + rotate and insert selected bits - without CC clobber
7525       The -mtune=zEC12 option enables zEC12 specific instruction
7526       scheduling without making use of new instructions.
7527     * Register pressure sensitive instruction scheduling is enabled by
7528       default.
7529     * The ifunc function attribute is enabled by default.
7530     * memcpy and memcmp invokations on big memory chunks or with run time
7531       lengths are not generated inline anymore when tuning for z10 or
7532       higher. The purpose is to make use of the IFUNC optimized versions
7533       in Glibc.
7534
7535  SH
7536
7537     * The default alignment settings have been reduced to be less
7538       aggressive. This results in more compact code for optimization
7539       levels other than -Os.
7540     * Improved support for the __atomic built-in functions:
7541          + A new option -matomic-model=model selects the model for the
7542            generated atomic sequences. The following models are
7543            supported:
7544
7545              soft-gusa
7546                      Software gUSA sequences (SH3* and SH4* only). On
7547                      SH4A targets this will now also partially utilize
7548                      the movco.l and movli.l instructions. This is the
7549                      default when the target is sh3*-*-linux* or
7550                      sh4*-*-linux*.
7551
7552              hard-llcs
7553                      Hardware movco.l / movli.l sequences (SH4A only).
7554
7555              soft-tcb
7556                      Software thread control block sequences.
7557
7558              soft-imask
7559                      Software interrupt flipping sequences (privileged
7560                      mode only). This is the default when the target is
7561                      sh1*-*-linux* or sh2*-*-linux*.
7562
7563              none
7564                      Generates function calls to the respective __atomic
7565                      built-in functions. This is the default for SH64
7566                      targets or when the target is not sh*-*-linux*.
7567
7568          + The option -msoft-atomic has been deprecated. It is now an
7569            alias for -matomic-model=soft-gusa.
7570          + A new option -mtas makes the compiler generate the tas.b
7571            instruction for the __atomic_test_and_set built-in function
7572            regardless of the selected atomic model.
7573          + The __sync functions in libgcc now reflect the selected atomic
7574            model when building the toolchain.
7575     * Added support for the mov.b and mov.w instructions with
7576       displacement addressing.
7577     * Added support for the SH2A instructions movu.b and movu.w.
7578     * Various improvements to code generated for integer arithmetic.
7579     * Improvements to conditional branches and code that involves the T
7580       bit. A new option -mzdcbranch tells the compiler to favor
7581       zero-displacement branches. This is enabled by default for SH4*
7582       targets.
7583     * The pref instruction will now be emitted by the __builtin_prefetch
7584       built-in function for SH3* targets.
7585     * The fmac instruction will now be emitted by the fmaf standard
7586       function and the __builtin_fmaf built-in function.
7587     * The -mfused-madd option has been deprecated in favor of the
7588       machine-independent -ffp-contract option. Notice that the fmac
7589       instruction will now be generated by default for expressions like a
7590       * b + c. This is due to the compiler default setting
7591       -ffp-contract=fast.
7592     * Added new options -mfsrra and -mfsca to allow the compiler using
7593       the fsrra and fsca instructions on targets other than SH4A (where
7594       they are already enabled by default).
7595     * Added support for the __builtin_bswap32 built-in function. It is
7596       now expanded as a sequence of swap.b and swap.w instructions
7597       instead of a library function call.
7598     * The behavior of the -mieee option has been fixed and the negative
7599       form -mno-ieee has been added to control the IEEE conformance of
7600       floating point comparisons. By default -mieee is now enabled and
7601       the option -ffinite-math-only implicitly sets -mno-ieee.
7602     * Added support for the built-in functions __builtin_thread_pointer
7603       and __builtin_set_thread_pointer. This assumes that GBR is used to
7604       hold the thread pointer of the current thread. Memory loads and
7605       stores relative to the address returned by __builtin_thread_pointer
7606       will now also utilize GBR based displacement address modes.
7607     * The -mdiv= option for targets other than SHmedia has been fixed and
7608       documented.
7609
7610  SPARC
7611
7612     * Added optimized instruction scheduling for Niagara4.
7613
7614  TILE-Gx
7615
7616     * Added support for the -mcmodel=MODEL command-line option. The
7617       models supported are small and large.
7618
7619  V850
7620
7621     * This target now supports the E3V5 architecture via the use of the
7622       new -mv850e3v5 command-line option. It also has experimental
7623       support for the e3v5 LOOP instruction which can be enabled via the
7624       new -mloop command-line option.
7625
7626  XStormy16
7627
7628     * This target now supports the -fstack-usage command-line option.
7629
7630Operating Systems
7631
7632  OpenBSD
7633
7634     * Support for OpenBSD/amd64 (x86_64-*-openbsd*) has been added and
7635       support for OpenBSD/i386 (i386-*-openbsd*) has been rejuvenated.
7636
7637  Windows (Cygwin)
7638
7639     * Executables are now linked against shared libgcc by default. The
7640       previous default was to link statically, which can still be done by
7641       explicitly specifying -static or static-libgcc on the command line.
7642       However it is strongly advised against, as it will cause problems
7643       for any application that makes use of DLLs compiled by GCC. It
7644       should be alright for a monolithic stand-alone application that
7645       only links against the Windows DLLs, but offers little or no
7646       benefit.
7647
7648GCC 4.8.1
7649
7650   This is the [33]list of problem reports (PRs) from GCC's bug tracking
7651   system that are known to be fixed in the 4.8.1 release. This list might
7652   not be complete (that is, it is possible that some PRs that have been
7653   fixed are not listed here).
7654
7655   The C++11 <chrono> std::chrono::system_clock and
7656   std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they
7657   both are now separate (never typedefs of each other), both use
7658   std::chrono::nanoseconds resolution, on most GNU/Linux configurations
7659   std::chrono::steady_clock is now finally monotonic, and both classes
7660   are mangled differently than in the previous GCC releases.
7661   std::chrono::system_clock::now() with std::chrono::microseconds resp.
7662   std::chrono::seconds resolution is still exported for backwards
7663   compatibility with default configured libstdc++. Note that libstdc++
7664   configured with --enable-libstdcxx-time= used to be ABI incompatible
7665   with default configured libstdc++ for those two classes and no ABI
7666   compatibility can be offered for those configurations, so any C++11
7667   code that uses those classes and has been compiled and linked against
7668   libstdc++ configured with the non-default --enable-libstdcxx-time=
7669   configuration option needs to be recompiled.
7670
7671GCC 4.8.2
7672
7673   This is the [34]list of problem reports (PRs) from GCC's bug tracking
7674   system that are known to be fixed in the 4.8.2 release. This list might
7675   not be complete (that is, it is possible that some PRs that have been
7676   fixed are not listed here).
7677
7678GCC 4.8.3
7679
7680   This is the [35]list of problem reports (PRs) from GCC's bug tracking
7681   system that are known to be fixed in the 4.8.3 release. This list might
7682   not be complete (that is, it is possible that some PRs that have been
7683   fixed are not listed here).
7684
7685   Support for the new powerpc64le-linux platform has been added. It
7686   defaults to generating code that conforms to the ELFV2 ABI.
7687
7688GCC 4.8.4
7689
7690   This is the [36]list of problem reports (PRs) from GCC's bug tracking
7691   system that are known to be fixed in the 4.8.4 release. This list might
7692   not be complete (that is, it is possible that some PRs that have been
7693   fixed are not listed here).
7694
7695GCC 4.8.5
7696
7697   This is the [37]list of problem reports (PRs) from GCC's bug tracking
7698   system that are known to be fixed in the 4.8.5 release. This list might
7699   not be complete (that is, it is possible that some PRs that have been
7700   fixed are not listed here).
7701
7702
7703    For questions related to the use of GCC, please consult these web
7704    pages and the [38]GCC manuals. If that fails, the
7705    [39]gcc-help@gcc.gnu.org mailing list might help. Comments on these
7706    web pages and the development of GCC are welcome on our developer
7707    list at [40]gcc@gcc.gnu.org. All of [41]our lists have public
7708    archives.
7709
7710   Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and
7711   distribution of this entire article is permitted in any medium,
7712   provided this notice is preserved.
7713
7714   These pages are [43]maintained by the GCC team. Last modified
7715   2021-07-28[44].
7716
7717References
7718
7719   1. https://gcc.gnu.org/wiki/cxx-conversion
7720   2. ftp://gcc.gnu.org/pub/gcc/infrastructure/
7721   3. http://www.nongnu.org/avr-libc/
7722   4. https://gcc.gnu.org/PR54461
7723   5. https://gcc.gnu.org/gcc-4.8/porting_to.html
7724   6. https://github.com/google/sanitizers
7725   7. https://code.google.com/archive/p/data-race-test/wikis/ThreadSanitizer.wiki
7726   8. https://lkml.org/lkml/2006/11/28/239
7727   9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
7728  10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
7729  11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
7730  12. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf
7731  13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
7732  14. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html
7733  15. http://gcc.gnu.org/projects/cxx1y.html
7734  16. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011
7735  17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html
7736  18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
7737  19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
7738  20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
7739  21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
7740  22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
7741  23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html
7742  24. https://gcc.gnu.org/wiki/Fortran2003Status
7743  25. https://gcc.gnu.org/wiki/TS29113Status
7744  26. https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=libgfortran/libgfortran.h
7745  27. http://chasm-interop.sourceforge.net/
7746  28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support
7747  29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html
7748  30. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html
7749  31. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html
7750  32. https://gcc.gnu.org/wiki/FunctionMultiVersioning
7751  33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1
7752  34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2
7753  35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3
7754  36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4
7755  37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5
7756  38. https://gcc.gnu.org/onlinedocs/
7757  39. mailto:gcc-help@gcc.gnu.org
7758  40. mailto:gcc@gcc.gnu.org
7759  41. https://gcc.gnu.org/lists.html
7760  42. https://www.fsf.org/
7761  43. https://gcc.gnu.org/about.html
7762  44. http://validator.w3.org/check/referer
7763======================================================================
7764http://gcc.gnu.org/gcc-4.7/index.html
7765                             GCC 4.7 Release Series
7766
7767   (This release series is no longer supported.)
7768
7769   June 12, 2014
7770
7771   The [1]GNU project and the GCC developers are pleased to announce the
7772   release of GCC 4.7.4.
7773
7774   This release is a bug-fix release, containing fixes for regressions in
7775   GCC 4.7.3 relative to previous releases of GCC.
7776
7777Release History
7778
7779   GCC 4.7.4
7780          June 12, 2014 ([2]changes, [3]documentation)
7781
7782   GCC 4.7.3
7783          April 11, 2013 ([4]changes, [5]documentation)
7784
7785   GCC 4.7.2
7786          September 20, 2012 ([6]changes, [7]documentation)
7787
7788   GCC 4.7.1
7789          June 14, 2012 ([8]changes, [9]documentation)
7790
7791   GCC 4.7.0
7792          March 22, 2012 ([10]changes, [11]documentation)
7793
7794References and Acknowledgements
7795
7796   GCC used to stand for the GNU C Compiler, but since the compiler
7797   supports several other languages aside from C, it now stands for the
7798   GNU Compiler Collection.
7799
7800   A list of [12]successful builds is updated as new information becomes
7801   available.
7802
7803   The GCC developers would like to thank the numerous people that have
7804   contributed new features, improvements, bug fixes, and other changes as
7805   well as test results to GCC. This [13]amazing group of volunteers is
7806   what makes GCC successful.
7807
7808   For additional information about GCC please refer to the [14]GCC
7809   project web site or contact the [15]GCC development mailing list.
7810
7811   To obtain GCC please use [16]our mirror sites or [17]our version
7812   control system.
7813
7814
7815    For questions related to the use of GCC, please consult these web
7816    pages and the [18]GCC manuals. If that fails, the
7817    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
7818    web pages and the development of GCC are welcome on our developer
7819    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
7820    archives.
7821
7822   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
7823   distribution of this entire article is permitted in any medium,
7824   provided this notice is preserved.
7825
7826   These pages are [23]maintained by the GCC team. Last modified
7827   2021-07-28[24].
7828
7829References
7830
7831   1. http://www.gnu.org/
7832   2. http://gcc.gnu.org/gcc-4.7/changes.html
7833   3. https://gcc.gnu.org/onlinedocs/4.7.4/
7834   4. http://gcc.gnu.org/gcc-4.7/changes.html
7835   5. https://gcc.gnu.org/onlinedocs/4.7.3/
7836   6. http://gcc.gnu.org/gcc-4.7/changes.html
7837   7. https://gcc.gnu.org/onlinedocs/4.7.2/
7838   8. http://gcc.gnu.org/gcc-4.7/changes.html
7839   9. https://gcc.gnu.org/onlinedocs/4.7.1/
7840  10. http://gcc.gnu.org/gcc-4.7/changes.html
7841  11. https://gcc.gnu.org/onlinedocs/4.7.0/
7842  12. http://gcc.gnu.org/gcc-4.7/buildstat.html
7843  13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
7844  14. http://gcc.gnu.org/index.html
7845  15. mailto:gcc@gcc.gnu.org
7846  16. http://gcc.gnu.org/mirrors.html
7847  17. http://gcc.gnu.org/git.html
7848  18. https://gcc.gnu.org/onlinedocs/
7849  19. mailto:gcc-help@gcc.gnu.org
7850  20. mailto:gcc@gcc.gnu.org
7851  21. https://gcc.gnu.org/lists.html
7852  22. https://www.fsf.org/
7853  23. https://gcc.gnu.org/about.html
7854  24. http://validator.w3.org/check/referer
7855======================================================================
7856http://gcc.gnu.org/gcc-4.7/changes.html
7857                             GCC 4.7 Release Series
7858                        Changes, New Features, and Fixes
7859
7860Caveats
7861
7862     * The -fconserve-space flag has been deprecated. The flag had no
7863       effect for most targets: only targets without a global .bss section
7864       and without support for switchable sections. Furthermore, the flag
7865       only had an effect for G++, where it could result in wrong
7866       semantics (please refer to the GCC manual for further details). The
7867       flag will be removed in GCC 4.8
7868     * Support for a number of older systems and recently unmaintained or
7869       untested target ports of GCC has been declared obsolete in GCC 4.7.
7870       Unless there is activity to revive them, the next release of GCC
7871       will have their sources permanently removed.
7872       All GCC ports for the following processor architectures have been
7873       declared obsolete:
7874          + picoChip (picochip-*)
7875       The following ports for individual systems on particular
7876       architectures have been obsoleted:
7877          + IRIX 6.5 (mips-sgi-irix6.5)
7878          + MIPS OpenBSD (mips*-*-openbsd*)
7879          + Solaris 8 (*-*-solaris2.8). Details can be found in the
7880            [1]announcement.
7881          + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*)
7882     * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A,
7883       ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by
7884       default, which for some sources generates code that accesses memory
7885       on unaligned addresses. This requires the kernel of those systems
7886       to enable such accesses (controlled by CP15 register c1, refer to
7887       ARM documentation). Alternatively, or for compatibility with
7888       kernels where unaligned accesses are not supported, all code has to
7889       be compiled with -mno-unaligned-access. Upstream Linux kernel
7890       releases have automatically and unconditionally supported unaligned
7891       accesses as emitted by GCC due to this option being active since
7892       version 2.6.28.
7893     * Support on ARM for the legacy floating-point accelerator (FPA) and
7894       the mixed-endian floating-point format that it used has been
7895       obsoleted. The ports that still use this format have been obsoleted
7896       as well. Many legacy ARM ports already provide an alternative that
7897       uses the VFP floating-point format. The obsolete ports will be
7898       deleted in the next release.
7899       The obsolete ports with alternatives are:
7900          + arm*-*-rtems (use arm*-*-rtemseabi)
7901          + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
7902          + arm*-*-elf (use arm*-*-eabi)
7903          + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
7904       Note, however, that these alternatives are not binary compatible
7905       with their legacy counterparts (although some can support running
7906       legacy applications).
7907       The obsolete ports that currently lack a modern alternative are:
7908          + arm*-*-ecos-elf
7909          + arm*-*-freebsd
7910          + arm*-wince-pe*
7911       New ports that support more recent versions of the architecture are
7912       welcome.
7913     * Support for the Maverick co-processor on ARM has been obsoleted.
7914       Code to support it will be deleted in the next release.
7915     * Support has been removed for Unix International threads on Solaris
7916       2, so the --enable-threads=solaris configure option and the
7917       -threads compiler option don't work any longer.
7918     * Support has been removed for the Solaris BSD Compatibility Package,
7919       which lives in /usr/ucbinclude and /usr/ucblib. It has been removed
7920       from Solaris 11, and was only intended as a migration aid from
7921       SunOS 4 to SunOS 5. The -compat-bsd compiler option is not
7922       recognized any longer.
7923     * The AVR port's libgcc has been improved and its multilib structure
7924       has been enhanced. As a result, all objects contributing to an
7925       application must either be compiled with GCC versions up to 4.6.x
7926       or with GCC versions 4.7.1 or later. If the compiler is used with
7927       AVR Libc, you need a version that supports the new layout, i.e.
7928       implements [2]#35407.
7929     * The AVR port's -mshort-calls command-line option has been
7930       deprecated. It will be removed in the GCC 4.8 release. See -mrelax
7931       for a replacement.
7932     * The AVR port only references startup code that clears .bss and the
7933       common section resp. initializes the .data and .rodata section
7934       provided respective sections (or subsections thereof) are not
7935       empty, see [3]PR18145. Applications that put all static storage
7936       objects into non-standard sections and / or define all static
7937       storage objects in assembler modules, must reference __do_clear_bss
7938       resp. __do_copy_data by hand or undefine the symbol(s) by means of
7939       -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data.
7940     * The ARM port's -mwords-little-endian option has been deprecated. It
7941       will be removed in a future release.
7942     * Support has been removed for the NetWare x86 configuration
7943       obsoleted in GCC 4.6.
7944     * It is no longer possible to use the "l" constraint in MIPS16 asm
7945       statements.
7946     * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard
7947       library which affected the ABI in C++11 mode: a data member was
7948       added to std::list changing its size and altering the definitions
7949       of some member functions, and std::pair's move constructor was
7950       non-trivial which altered the calling convention for functions with
7951       std::pair arguments or return types. The ABI incompatibilities have
7952       been fixed for GCC version 4.7.2 but as a result C++11 code
7953       compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11
7954       code compiled with different GCC versions and with C++98/C++03 code
7955       compiled with any version.
7956     * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
7957       rules for the layout of vectors that could lead to wrong code being
7958       generated. Vectors larger than 8 bytes in size are now by default
7959       aligned to an 8-byte boundary. This is an ABI change: code that
7960       makes explicit use of vector types may be incompatible with binary
7961       objects built with older versions of GCC. Auto-vectorized code is
7962       not affected by this change. (This change affects GCC versions
7963       4.7.2 and later.)
7964     * More information on porting to GCC 4.7 from previous versions of
7965       GCC can be found in the [4]porting guide for this release.
7966
7967General Optimizer Improvements
7968
7969     * Support for a new parameter --param case-values-threshold=n was
7970       added to allow users to control the cutoff between doing switch
7971       statements as a series of if statements and using a jump table.
7972     * Link-time optimization (LTO) improvements:
7973          + Improved scalability and reduced memory usage. Link time
7974            optimization of Firefox now requires 3GB of RAM on a 64-bit
7975            system, while over 8GB was needed previously. Linking time has
7976            been improved, too. The serial stage of linking Firefox has
7977            been sped up by about a factor of 10.
7978          + Reduced size of object files and temporary storage used during
7979            linking.
7980          + Streaming performance (both outbound and inbound) has been
7981            improved.
7982          + ld -r is now supported with LTO.
7983          + Several bug fixes, especially in symbol table handling and
7984            merging.
7985     * Interprocedural optimization improvements:
7986          + Heuristics now take into account that after inlining code will
7987            be optimized out because of known values (or properties) of
7988            function parameters. For example:
7989void foo(int a)
7990{
7991  if (a > 10)
7992    ... huge code ...
7993}
7994void bar (void)
7995{
7996  foo (0);
7997}
7998
7999            The call of foo will be inlined into bar even when optimizing
8000            for code size. Constructs based on __builtin_constant_p are
8001            now understood by the inliner and code size estimates are
8002            evaluated a lot more realistically.
8003          + The representation of C++ virtual thunks and aliases (both
8004            implicit and defined via the alias attribute) has been
8005            re-engineered. Aliases no longer pose optimization barriers
8006            and calls to an alias can be inlined and otherwise optimized.
8007          + The inter-procedural constant propagation pass has been
8008            rewritten. It now performs generic function specialization.
8009            For example when compiling the following:
8010void foo(bool flag)
8011{
8012  if (flag)
8013    ... do something ...
8014  else
8015    ... do something else ...
8016}
8017void bar (void)
8018{
8019  foo (false);
8020  foo (true);
8021  foo (false);
8022  foo (true);
8023  foo (false);
8024  foo (true);
8025}
8026
8027            GCC will now produce two copies of foo. One with flag being
8028            true, while other with flag being false. This leads to
8029            performance improvements previously possible only by inlining
8030            all calls. Cloning causes a lot less code size growth.
8031     * A string length optimization pass has been added. It attempts to
8032       track string lengths and optimize various standard C string
8033       functions like strlen, strchr, strcpy, strcat, stpcpy and their
8034       _FORTIFY_SOURCE counterparts into faster alternatives. This pass is
8035       enabled by default at -O2 or above, unless optimizing for size, and
8036       can be disabled by the -fno-optimize-strlen option. The pass can
8037       e.g. optimize
8038char *bar (const char *a)
8039{
8040  size_t l = strlen (a) + 2;
8041  char *p = malloc (l); if (p == NULL) return p;
8042  strcpy (p, a); strcat (p, "/"); return p;
8043}
8044
8045       into:
8046char *bar (const char *a)
8047{
8048  size_t tmp = strlen (a);
8049  char *p = malloc (tmp + 2); if (p == NULL) return p;
8050  memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p;
8051}
8052
8053       or for hosted compilations where stpcpy is available in the runtime
8054       and headers provide its prototype, e.g.
8055void foo (char *a, const char *b, const char *c, const char *d)
8056{
8057  strcpy (a, b); strcat (a, c); strcat (a, d);
8058}
8059
8060       can be optimized into:
8061void foo (char *a, const char *b, const char *c, const char *d)
8062{
8063  strcpy (stpcpy (stpcpy (a, b), c), d);
8064}
8065
8066New Languages and Language specific improvements
8067
8068     * Version 3.1 of the OpenMP specification is now supported for the C,
8069       C++, and Fortran compilers.
8070
8071  Ada
8072
8073     * The command-line option -feliminate-unused-debug-types has been
8074       re-enabled by default, as it is for the other languages, leading to
8075       a reduction in debug info size of 12.5% and more for relevant
8076       cases, as well as to a small compilation speedup.
8077
8078  C family
8079
8080     * A new built-in, __builtin_assume_aligned, has been added, through
8081       which the compiler can be hinted about pointer alignment and can
8082       use it to improve generated code.
8083     * A new warning option -Wunused-local-typedefs was added for C, C++,
8084       Objective-C and Objective-C++. This warning diagnoses typedefs
8085       locally defined in a function, and otherwise not used.
8086     * A new experimental command-line option -ftrack-macro-expansion was
8087       added for C, C++, Objective-C, Objective-C++ and Fortran. It allows
8088       the compiler to emit diagnostic about the current macro expansion
8089       stack when a compilation error occurs in a macro expansion.
8090     * Experimental support for transactional memory has been added. It
8091       includes support in the compiler, as well as a supporting runtime
8092       library called libitm. To compile code with transactional memory
8093       constructs, use the -fgnu-tm option.
8094       Support is currently available for Alpha, ARM, PowerPC, SH, SPARC,
8095       and 32-bit/64-bit x86 platforms.
8096       For more details on transactional memory see [5]the GCC WiKi.
8097     * Support for atomic operations specifying the C++11/C11 memory model
8098       has been added. These new __atomic routines replace the existing
8099       __sync built-in routines.
8100       Atomic support is also available for memory blocks. Lock-free
8101       instructions will be used if a memory block is the same size and
8102       alignment as a supported integer type. Atomic operations which do
8103       not have lock-free support are left as function calls. A set of
8104       library functions is available on the GCC atomic wiki in the
8105       "External Atomics Library" section.
8106       For more details on the memory models and features, see the
8107       [6]atomic wiki.
8108     * When a binary operation is performed on vector types and one of the
8109       operands is a uniform vector, it is possible to replace the vector
8110       with the generating element. For example:
8111typedef int v4si __attribute__ ((vector_size (16)));
8112v4si res, a = {1,2,3,4};
8113int x;
8114
8115res = 2 + a;  /* means {2,2,2,2} + a  */
8116res = a - x;  /* means a - {x,x,x,x}  */
8117
8118  C
8119
8120     * There is support for some more features from the C11 revision of
8121       the ISO C standard. GCC now accepts the options -std=c11 and
8122       -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x.
8123          + Unicode strings (previously supported only with options such
8124            as -std=gnu11, now supported with -std=c11), and the
8125            predefined macros __STDC_UTF_16__ and __STDC_UTF_32__.
8126          + Nonreturning functions (_Noreturn and <stdnoreturn.h>).
8127          + Alignment support (_Alignas, _Alignof, max_align_t,
8128            <stdalign.h>).
8129          + A built-in function __builtin_complex is provided to support C
8130            library implementation of the CMPLX family of macros.
8131
8132  C++
8133
8134     * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat
8135       options, which are equivalent to -std=c++0x, -std=gnu++0x, and
8136       -Wc++0x-compat, respectively.
8137     * G++ now implements [7]C++11 extended friend syntax:
8138
8139template<class W>
8140class Q
8141{
8142  static const int I = 2;
8143public:
8144  friend W;
8145};
8146
8147struct B
8148{
8149  int ar[Q<B>::I];
8150};
8151
8152     * Thanks to Ville Voutilainen, G++ now implements [8]C++11 explicit
8153       override control.
8154
8155struct B {
8156  virtual void f() const final;
8157  virtual void f(int);
8158};
8159
8160struct D : B {
8161  void f() const;            // error: D::f attempts to override final B::f
8162  void f(long) override;     // error: doesn't override anything
8163  void f(int) override;      // ok
8164};
8165
8166struct E final { };
8167struct F: E { }; // error: deriving from final class
8168
8169     * G++ now implements [9]C++11 non-static data member initializers.
8170
8171struct A {
8172  int i = 42;
8173} a; // initializes a.i to 42
8174
8175     * Thanks to Ed Smith-Rowland, G++ now implements [10]C++11
8176       user-defined literals.
8177
8178// Not actually a good approximation.  :)
8179constexpr long double operator"" _degrees (long double d) { return d * 0.0175; }
8180long double pi = 180.0_degrees;
8181
8182     * G++ now implements [11]C++11 alias-declarations.
8183
8184template <class T> using Ptr = T*;
8185Ptr<int> ip;  // decltype(ip) is int*
8186
8187     * Thanks to Ville Voutilainen and Pedro Lamarão, G++ now implements
8188       [12]C++11 delegating constructors.
8189
8190struct A {
8191  A(int);
8192  A(): A(42) { } // delegate to the A(int) constructor
8193};
8194
8195     * G++ now fully implements C++11 atomic classes rather than just
8196       integer derived classes.
8197
8198class POD {
8199  int a;
8200  int b;
8201};
8202std::atomic<POD> my_atomic_POD;
8203
8204     * G++ now sets the predefined macro __cplusplus to the correct value,
8205       199711L for C++98/03, and 201103L for C++11.
8206     * G++ now correctly implements the two-phase lookup rules such that
8207       an unqualified name used in a template must have an appropriate
8208       declaration found either in scope at the point of definition of the
8209       template or by argument-dependent lookup at the point of
8210       instantiation. As a result, code that relies on a second
8211       unqualified lookup at the point of instantiation to find functions
8212       declared after the template or in dependent bases will be rejected.
8213       The compiler will suggest ways to fix affected code, and using the
8214       -fpermissive compiler flag will allow the code to compile with a
8215       warning.
8216
8217template <class T>
8218void f() { g(T()); } // error, g(int) not found by argument-dependent lookup
8219void g(int) { } // fix by moving this declaration before the declaration of f
8220
8221template <class T>
8222struct A: T {
8223  // error, B::g(B) not found by argument-dependent lookup
8224  void f() { g(T()); } // fix by using this->g or A::g
8225};
8226
8227struct B { void g(B); };
8228
8229int main()
8230{
8231  f<int>();
8232  A<B>().f();
8233}
8234
8235     * G++ now properly re-uses stack space allocated for temporary
8236       objects when their lifetime ends, which can significantly lower
8237       stack consumption for some C++ functions. As a result of this, some
8238       code with undefined behavior will now break:
8239
8240const int &f(const int &i) { return i; }
8241....
8242const int &x = f(1);
8243const int &y = f(2);
8244
8245       Here, x refers to the temporary allocated to hold the 1 argument,
8246       which only lives until the end of the initialization; it
8247       immediately becomes a dangling reference. So the next statement
8248       re-uses the stack slot to hold the 2 argument, and users of x get
8249       that value instead.
8250       Note that this should not cause any change of behavior for
8251       temporaries of types with non-trivial destructors, as they are
8252       already destroyed at end of full-expression; the change is that now
8253       the storage is released as well.
8254     * A new command-line option -Wdelete-non-virtual-dtor has been added
8255       to warn when delete is used to destroy an instance of a class which
8256       has virtual functions and non-virtual destructor. It is unsafe to
8257       delete an instance of a derived class through a pointer to a base
8258       class if the base class does not have a virtual destructor. This
8259       warning is enabled by -Wall.
8260     * A new command-line option -Wzero-as-null-pointer-constant has been
8261       added to warn when a literal '0' is used as null pointer constant.
8262       It can be useful to facilitate the conversion to nullptr in C++11.
8263     * As per C++98, access-declarations are now deprecated by G++.
8264       Using-declarations are to be used instead. Furthermore, some
8265       efforts have been made to improve the support of class scope
8266       using-declarations. In particular, using-declarations referring to
8267       a dependent type now work as expected ([13]bug c++/14258).
8268     * The ELF symbol visibility of a template instantiation is now
8269       properly constrained by the visibility of its template arguments
8270       ([14]bug c++/35688).
8271
8272    Runtime Library (libstdc++)
8273
8274     * [15]Improved experimental support for the new ISO C++ standard,
8275       C++11, including:
8276          + using noexcept in most of the library;
8277          + implementations of pointer_traits, allocator_traits and
8278            scoped_allocator_adaptor;
8279          + uses-allocator construction for tuple;
8280          + vector meets the allocator-aware container requirements;
8281          + replacing monotonic_clock with steady_clock;
8282          + enabling the thread support library on most POSIX targets;
8283          + many small improvements to conform to the FDIS.
8284     * Added --enable-clocale=newlib configure option.
8285     * Debug Mode iterators for unordered associative containers.
8286     * Avoid polluting the global namespace and do not include <unistd.h>.
8287
8288  Fortran
8289
8290     * The compile flag [16]-fstack-arrays has been added, which causes
8291       all local arrays to be put on stack memory. For some programs this
8292       will improve the performance significantly. If your program uses
8293       very large local arrays, it is possible that you will have to
8294       extend your runtime limits for stack memory.
8295     * The [17]-Ofast flag now also implies [18]-fno-protect-parens and
8296       [19]-fstack-arrays.
8297     * Front-end optimizations can now be selected by the
8298       [20]-ffrontend-optimize option and deselected by the
8299       -fno-frontend-optimize option.
8300     * When front-end optimization removes a function call,
8301       [21]-Wfunction-elimination warns about that.
8302     * When performing front-end-optimization, the
8303       [22]-faggressive-function-elimination option allows the removal of
8304       duplicate function calls even for impure functions.
8305     * The flag [23]-Wreal-q-constant has been added, which warns if
8306       floating-point literals have been specified using q (such as
8307       1.0q0); the q marker is now supported as a vendor extension to
8308       denote quad precision (REAL(16) or, if not available, REAL(10)).
8309       Consider using a kind parameter (such as in 1.0_qp) instead, which
8310       can be obtained via [24]SELECTED_REAL_KIND.
8311     * The GFORTRAN_USE_STDERR environment variable has been removed. GNU
8312       Fortran now always prints error messages to standard error. If you
8313       wish to redirect standard error, please consult the manual for your
8314       OS, shell, batch environment etc. as appropriate.
8315     * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment
8316       variable have been removed. When encountering a serious error,
8317       gfortran will now always abort the program. Whether a core dump is
8318       generated depends on the user environment settings; see the ulimit
8319       -c setting for POSIX shells, limit coredumpsize for C shells, and
8320       the [25]WER user-mode dumps settings on Windows.
8321     * The [26]-fbacktrace option is now enabled by default. When
8322       encountering a fatal error, gfortran will attempt to print a
8323       backtrace to standard error before aborting. It can be disabled
8324       with -fno-backtrace. Note: On POSIX targets with the addr2line
8325       utility from GNU binutils, GNU Fortran can print a backtrace with
8326       function name, file name, line number information in addition to
8327       the addresses; otherwise only the addresses are printed.
8328     * [27]Fortran 2003:
8329          + Generic interface names which have the same name as derived
8330            types are now supported, which allows to write constructor
8331            functions. Note that Fortran does not support static
8332            constructor functions; only default initialization or an
8333            explicit structure-constructor initialization are available.
8334          + [28]Polymorphic (class) arrays are now supported.
8335     * [29]Fortran 2008:
8336          + Support for the DO CONCURRENT construct has been added, which
8337            allows the user to specify that individual loop iterations
8338            have no interdependencies.
8339          + [30]Coarrays: Full single-image support except for polymorphic
8340            coarrays. Additionally, preliminary support for multiple
8341            images via an MPI-based [31]coarray communication library has
8342            been added. Note: The library version is not yet usable as
8343            remote coarray access is not yet possible.
8344     * [32]TS 29113:
8345          + New flag [33]-std=f2008ts permits programs that are expected
8346            to conform to the Fortran 2008 standard and the draft
8347            Technical Specification (TS) 29113 on Further Interoperability
8348            of Fortran with C.
8349          + The OPTIONAL attribute is now allowed for dummy arguments of
8350            BIND(C) procedures.
8351          + The RANK intrinsic has been added.
8352          + The implementation of the ASYNCHRONOUS attribute in GCC is
8353            compatible with the candidate draft of TS 29113 (since GCC
8354            4.6).
8355
8356  Go
8357
8358     * GCC 4.7 implements the [34]Go 1 language standard. The library
8359       support in 4.7.0 is not quite complete, due to release timing.
8360       Release 4.7.1 includes complete support for Go 1. The Go library is
8361       from the Go 1.0.1 release.
8362     * Go has been tested on GNU/Linux and Solaris platforms. It may work
8363       on other platforms as well.
8364
8365New Targets and Target Specific Improvements
8366
8367  ARM
8368
8369     * GCC now supports the Cortex-A7 processor implementing the v7-a
8370       version of the architecture using the option -mcpu=cortex-a7.
8371     * The default vector size in auto-vectorization for NEON is now 128
8372       bits. If vectorization fails thusly, the vectorizer tries again
8373       with 64-bit vectors.
8374     * A new option -mvectorize-with-neon-double was added to allow users
8375       to change the vector size to 64 bits.
8376
8377  AVR
8378
8379     * GCC now supports the XMEGA architecture. This requires GNU binutils
8380       2.22 or later.
8381     * Support for the [35]named address spaces __flash, __flash1, …,
8382       __flash5 and __memx has been added. These address spaces locate
8383       read-only data in flash memory and allow reading from flash memory
8384       by means of ordinary C code, i.e. without the need of (inline)
8385       assembler code:
8386
8387const __flash int values[] = { 42, 31 };
8388
8389int add_values (const __flash int *p, int i)
8390{
8391    return values[i] + *p;
8392}
8393
8394     * Support has been added for the AVR-specific configure option
8395       --with-avrlibc=yes in order to arrange for better integration of
8396       [36]AVR-Libc. This configure option is supported in avr-gcc 4.7.2
8397       and newer and will only take effect in non-RTEMS configurations. If
8398       avr-gcc is configured for RTEMS, the option will be ignored which
8399       is the same as specifying --with-avrlibc=no. See [37]PR54461 for
8400       more technical details.
8401     * Support for AVR-specific [38]built-in functions has been added.
8402     * Support has been added for the signed and unsigned 24-bit scalar
8403       integer types __int24 and __uint24.
8404     * New command-line options -maccumulate-args, -mbranch-cost=cost and
8405       -mstrict-X were added to allow better fine-tuning of code
8406       optimization.
8407     * The command-line option -fdata-sections now also takes affect on
8408       the section names of variables with the progmem attribute.
8409     * A new inline assembler print modifier %i to print a RAM address as
8410       I/O address has been added:
8411
8412#include <avr/io.h> /* Port Definitions from AVR-LibC */
8413
8414void set_portb (uint8_t value)
8415{
8416    asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory");
8417}
8418
8419       The offset between an I/O address and the RAM address for that I/O
8420       location is device-specific. This offset is taken into account when
8421       printing a RAM address with the %i modifier so that the address is
8422       suitable to be used as operand in an I/O command. The address must
8423       be a constant integer known at compile time.
8424     * The inline assembler constraint "R" to represent integers in the
8425       range −6 … 5 has been removed without replacement.
8426     * Many optimizations to:
8427          + 64-bit integer arithmetic
8428          + Widening multiplication
8429          + Integer division by a constant
8430          + Avoid constant reloading in multi-byte instructions.
8431          + Micro-optimizations for special instruction sequences.
8432          + Generic built-in functions like __builtin_ffs*,
8433            __builtin_clz*, etc.
8434          + If-else decision trees generated by switch instructions
8435          + Merging of data located in flash memory
8436          + New libgcc variants for devices with 8-bit wide stack pointer
8437          + …
8438     * Better documentation:
8439          + Handling of EIND and indirect jumps on devices with more than
8440            128 KiB of program memory.
8441          + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function
8442            registers.
8443          + Function attributes OS_main and OS_task.
8444          + AVR-specific built-in macros.
8445
8446  C6X
8447
8448     * Support has been added for the Texas Instruments C6X family of
8449       processors.
8450
8451  CR16
8452
8453     * Support has been added for National Semiconductor's CR16
8454       architecture.
8455
8456  Epiphany
8457
8458     * Support has been added for Adapteva's Epiphany architecture.
8459
8460  IA-32/x86-64
8461
8462     * Support for Intel AVX2 intrinsics, built-in functions and code
8463       generation is available via -mavx2.
8464     * Support for Intel BMI2 intrinsics, built-in functions and code
8465       generation is available via -mbmi2.
8466     * Implementation and automatic generation of __builtin_clz* using the
8467       lzcnt instruction is available via -mlzcnt.
8468     * Support for Intel FMA3 intrinsics and code generation is available
8469       via -mfma.
8470     * A new -mfsgsbase command-line option is available that makes GCC
8471       generate new segment register read/write instructions through
8472       dedicated built-ins.
8473     * Support for the new Intel rdrnd instruction is available via
8474       -mrdrnd.
8475     * Two additional AVX vector conversion instructions are available via
8476       -mf16c.
8477     * Support for new Intel processor codename IvyBridge with RDRND,
8478       FSGSBASE and F16C is available through -march=core-avx-i.
8479     * Support for the new Intel processor codename Haswell with AVX2,
8480       FMA, BMI, BMI2, LZCNT is available through -march=core-avx2.
8481     * Support for new AMD family 15h processors (Piledriver core) is now
8482       available through -march=bdver2 and -mtune=bdver2 options.
8483     * Support for [39]the x32 psABI is now available through the -mx32
8484       option.
8485     * Windows mingw targets are using the -mms-bitfields option by
8486       default.
8487     * Windows x86 targets are using the __thiscall calling convention for
8488       C++ class-member functions.
8489     * Support for the configure option --with-threads=posix for Windows
8490       mingw targets.
8491
8492  MIPS
8493
8494     * GCC now supports thread-local storage (TLS) for MIPS16. This
8495       requires GNU binutils 2.22 or later.
8496     * GCC can now generate code specifically for the Cavium Octeon+ and
8497       Octeon2 processors. The associated command-line options are
8498       -march=octeon+ and -march=octeon2 respectively. Both options
8499       require GNU binutils 2.22 or later.
8500     * GCC can now work around certain 24k errata, under the control of
8501       the command-line option -mfix-24k. These workarounds require GNU
8502       binutils 2.20 or later.
8503     * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build
8504       n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux
8505       toolchain that generates 32-bit code by default. Use the
8506       configure-time option --enable-targets=all to select these extra
8507       multilibs.
8508     * Passing -fno-delayed-branch now also stops the assembler from
8509       automatically filling delay slots.
8510
8511  PowerPC/PowerPC64
8512
8513     * Vectors of type vector long long or vector long are passed and
8514       returned using the same method as other vectors with the VSX
8515       instruction set. Previously GCC did not adhere to the ABI for
8516       128-bit vectors with 64-bit integer base types (PR 48857). This
8517       will also be fixed in the GCC 4.6.1 and 4.5.4 releases.
8518     * A new option -mno-pointers-to-nested-functions was added to allow
8519       AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify
8520       that the compiler should not load up the chain register (r11)
8521       before calling a function through a pointer. If you use this
8522       option, you cannot call nested functions through a pointer, or call
8523       other languages that might use the static chain.
8524     * A new option msave-toc-indirect was added to allow AIX
8525       32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we
8526       save the TOC in the prologue for indirect calls or generate the
8527       save inline. This can speed up some programs that call through a
8528       function pointer a lot, but it can slow down other functions that
8529       only call through a function pointer in exceptional cases.
8530     * The PowerPC port will now enable machine-specific built-in
8531       functions when the user switches the target machine using the
8532       #pragma GCC target or __attribute__ ((__target__ ("target"))) code
8533       sequences. In addition, the target macros are updated. However, due
8534       to the way the -save-temps switch is implemented, you won't see the
8535       effect of these additional macros being defined in preprocessor
8536       output.
8537
8538  SH
8539
8540     * A new option -msoft-atomic has been added. When it is specified,
8541       GCC will generate GNU/Linux-compatible gUSA atomic sequences for
8542       the new __atomic routines.
8543     * Since it is neither supported by GAS nor officially documented,
8544       code generation for little endian SH2A has been disabled.
8545       Specifying -ml with -m2a* will now result in a compiler error.
8546     * The defunct -mbranch-cost option has been fixed.
8547     * Some improvements to the generated code of:
8548          + Utilization of the tst #imm,R0 instruction.
8549          + Dynamic shift instructions on SH2A.
8550          + Integer absolute value calculations.
8551     * The -mdiv= option for targets other than SHmedia has been fixed and
8552       documented.
8553
8554  SPARC
8555
8556     * The option -mflat has been reinstated. When it is specified, the
8557       compiler will generate code for a single register window model.
8558       This is essentially a new implementation and the corresponding
8559       debugger support has been added to GDB 7.4.
8560     * Support for the options -mtune=native and -mcpu=native has been
8561       added on selected native platforms (GNU/Linux and Solaris).
8562     * Support for the SPARC T3 (Niagara 3) processor has been added.
8563     * VIS:
8564          + An intrinsics header visintrin.h has been added.
8565          + Builtin intrinsics for the VIS 1.0 edge handling and pixel
8566            compare instructions have been added.
8567          + The little-endian version of alignaddr is now supported.
8568          + When possible, VIS builtins are marked const, which should
8569            increase the compiler's ability to optimize VIS operations.
8570          + The compiler now properly tracks the %gsr register and how it
8571            behaves as an input for various VIS instructions.
8572          + Akin to fzero, the compiler can now generate fone instructions
8573            in order to set all of the bits of a floating-point register
8574            to 1.
8575          + The documentation for the VIS intrinsics in the GCC manual has
8576            been brought up to date and many inaccuracies were fixed.
8577          + Intrinsics for the VIS 2.0 bmask, bshuffle, and
8578            non-condition-code setting edge instructions have been added.
8579            Their availability is controlled by the new -mvis2 and
8580            -mno-vis2 options. They are enabled by default on
8581            UltraSPARC-III and later CPUs.
8582     * Support for UltraSPARC Fused Multiply-Add floating-point extensions
8583       has been added. These instructions are enabled by default on SPARC
8584       T3 (Niagara 3) and later CPUs.
8585
8586  TILE-Gx/TILEPro
8587
8588     * Support has been added for the Tilera TILE-Gx and TILEPro families
8589       of processors.
8590
8591Other significant improvements
8592
8593     * A new option (-grecord-gcc-switches) was added that appends
8594       compiler command-line options that might affect code generation to
8595       the DW_AT_producer attribute string in the DWARF debugging
8596       information.
8597     * GCC now supports various new GNU extensions to the DWARF debugging
8598       information format, like [40]entry value and [41]call site
8599       information, [42]typed DWARF stack or [43]a more compact macro
8600       representation. Support for these extensions has been added to GDB
8601       7.4. They can be disabled through the -gstrict-dwarf command-line
8602       option.
8603
8604GCC 4.7.1
8605
8606   This is the [44]list of problem reports (PRs) from GCC's bug tracking
8607   system that are known to be fixed in the 4.7.1 release. This list might
8608   not be complete (that is, it is possible that some PRs that have been
8609   fixed are not listed here).
8610
8611   The Go front end in the 4.7.1 release fully supports the [45]Go 1
8612   language standard.
8613
8614GCC 4.7.2
8615
8616   This is the [46]list of problem reports (PRs) from GCC's bug tracking
8617   system that are known to be fixed in the 4.7.2 release. This list might
8618   not be complete (that is, it is possible that some PRs that have been
8619   fixed are not listed here).
8620
8621GCC 4.7.3
8622
8623   This is the [47]list of problem reports (PRs) from GCC's bug tracking
8624   system that are known to be fixed in the 4.7.3 release. This list might
8625   not be complete (that is, it is possible that some PRs that have been
8626   fixed are not listed here).
8627
8628GCC 4.7.4
8629
8630   This is the [48]list of problem reports (PRs) from GCC's bug tracking
8631   system that are known to be fixed in the 4.7.4 release. This list might
8632   not be complete (that is, it is possible that some PRs that have been
8633   fixed are not listed here).
8634
8635
8636    For questions related to the use of GCC, please consult these web
8637    pages and the [49]GCC manuals. If that fails, the
8638    [50]gcc-help@gcc.gnu.org mailing list might help. Comments on these
8639    web pages and the development of GCC are welcome on our developer
8640    list at [51]gcc@gcc.gnu.org. All of [52]our lists have public
8641    archives.
8642
8643   Copyright (C) [53]Free Software Foundation, Inc. Verbatim copying and
8644   distribution of this entire article is permitted in any medium,
8645   provided this notice is preserved.
8646
8647   These pages are [54]maintained by the GCC team. Last modified
8648   2022-02-10[55].
8649
8650References
8651
8652   1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html
8653   2. http://savannah.nongnu.org/bugs/?35407
8654   3. https://gcc.gnu.org/PR18145
8655   4. https://gcc.gnu.org/gcc-4.7/porting_to.html
8656   5. https://gcc.gnu.org/wiki/TransactionalMemory
8657   6. https://gcc.gnu.org/wiki/Atomic/GCCMM
8658   7. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
8659   8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
8660   9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
8661  10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
8662  11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
8663  12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
8664  13. https://gcc.gnu.org/PR14258
8665  14. https://gcc.gnu.org/PR35688
8666  15. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011
8667  16. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
8668  17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689
8669  18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270
8670  19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
8671  20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275
8672  21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170
8673  22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270
8674  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
8675  24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html
8676  25. https://docs.microsoft.com/en-us/windows/win32/wer/collecting-user-mode-dumps
8677  26. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183
8678  27. https://gcc.gnu.org/wiki/Fortran2003Status
8679  28. https://gcc.gnu.org/wiki/OOP
8680  29. https://gcc.gnu.org/wiki/Fortran2008Status
8681  30. https://gcc.gnu.org/wiki/Coarray
8682  31. https://gcc.gnu.org/wiki/CoarrayLib
8683  32. https://gcc.gnu.org/wiki/TS29113Status
8684  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
8685  34. https://go.dev/doc/go1
8686  35. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html
8687  36. http://www.nongnu.org/avr-libc/
8688  37. https://gcc.gnu.org/PR54461
8689  38. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built_002din-Functions.html
8690  39. https://sites.google.com/site/x32abi/
8691  40. https://www.dwarfstd.org/ShowIssue.php?issue=100909.1
8692  41. https://www.dwarfstd.org/ShowIssue.php?issue=100909.2
8693  42. https://www.dwarfstd.org/ShowIssue.php?issue=140425.1
8694  43. https://www.dwarfstd.org/ShowIssue.php?issue=110722.1
8695  44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1
8696  45. https://go.dev/doc/go1
8697  46. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2
8698  47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3
8699  48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4
8700  49. https://gcc.gnu.org/onlinedocs/
8701  50. mailto:gcc-help@gcc.gnu.org
8702  51. mailto:gcc@gcc.gnu.org
8703  52. https://gcc.gnu.org/lists.html
8704  53. https://www.fsf.org/
8705  54. https://gcc.gnu.org/about.html
8706  55. http://validator.w3.org/check/referer
8707======================================================================
8708http://gcc.gnu.org/gcc-4.6/index.html
8709                             GCC 4.6 Release Series
8710
8711   (This release series is no longer supported.)
8712
8713   April 12, 2013
8714
8715   The [1]GNU project and the GCC developers are pleased to announce the
8716   release of GCC 4.6.4.
8717
8718   This release is a bug-fix release, containing fixes for regressions in
8719   GCC 4.6.3 relative to previous releases of GCC.
8720
8721Release History
8722
8723   GCC 4.6.4
8724          April 12, 2013 ([2]changes, [3]documentation)
8725
8726   GCC 4.6.3
8727          March 1, 2012 ([4]changes, [5]documentation)
8728
8729   GCC 4.6.2
8730          October 26, 2011 ([6]changes, [7]documentation)
8731
8732   GCC 4.6.1
8733          June 27, 2011 ([8]changes, [9]documentation)
8734
8735   GCC 4.6.0
8736          March 25, 2011 ([10]changes, [11]documentation)
8737
8738References and Acknowledgements
8739
8740   GCC used to stand for the GNU C Compiler, but since the compiler
8741   supports several other languages aside from C, it now stands for the
8742   GNU Compiler Collection.
8743
8744   A list of [12]successful builds is updated as new information becomes
8745   available.
8746
8747   The GCC developers would like to thank the numerous people that have
8748   contributed new features, improvements, bug fixes, and other changes as
8749   well as test results to GCC. This [13]amazing group of volunteers is
8750   what makes GCC successful.
8751
8752   For additional information about GCC please refer to the [14]GCC
8753   project web site or contact the [15]GCC development mailing list.
8754
8755   To obtain GCC please use [16]our mirror sites or [17]our version
8756   control system.
8757
8758
8759    For questions related to the use of GCC, please consult these web
8760    pages and the [18]GCC manuals. If that fails, the
8761    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
8762    web pages and the development of GCC are welcome on our developer
8763    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
8764    archives.
8765
8766   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
8767   distribution of this entire article is permitted in any medium,
8768   provided this notice is preserved.
8769
8770   These pages are [23]maintained by the GCC team. Last modified
8771   2021-07-28[24].
8772
8773References
8774
8775   1. http://www.gnu.org/
8776   2. http://gcc.gnu.org/gcc-4.6/changes.html
8777   3. https://gcc.gnu.org/onlinedocs/4.6.4/
8778   4. http://gcc.gnu.org/gcc-4.6/changes.html
8779   5. https://gcc.gnu.org/onlinedocs/4.6.3/
8780   6. http://gcc.gnu.org/gcc-4.6/changes.html
8781   7. https://gcc.gnu.org/onlinedocs/4.6.2/
8782   8. http://gcc.gnu.org/gcc-4.6/changes.html
8783   9. https://gcc.gnu.org/onlinedocs/4.6.1/
8784  10. http://gcc.gnu.org/gcc-4.6/changes.html
8785  11. https://gcc.gnu.org/onlinedocs/4.6.0/
8786  12. http://gcc.gnu.org/gcc-4.6/buildstat.html
8787  13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
8788  14. http://gcc.gnu.org/index.html
8789  15. mailto:gcc@gcc.gnu.org
8790  16. http://gcc.gnu.org/mirrors.html
8791  17. http://gcc.gnu.org/git.html
8792  18. https://gcc.gnu.org/onlinedocs/
8793  19. mailto:gcc-help@gcc.gnu.org
8794  20. mailto:gcc@gcc.gnu.org
8795  21. https://gcc.gnu.org/lists.html
8796  22. https://www.fsf.org/
8797  23. https://gcc.gnu.org/about.html
8798  24. http://validator.w3.org/check/referer
8799======================================================================
8800http://gcc.gnu.org/gcc-4.6/changes.html
8801                             GCC 4.6 Release Series
8802                        Changes, New Features, and Fixes
8803
8804Caveats
8805
8806     * The options -b <machine> and -V <version> have been removed because
8807       they were unreliable. Instead, users should directly run
8808       <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to
8809       run a different version of gcc.
8810     * GCC now has stricter checks for invalid command-line options. In
8811       particular, when gcc was called to link object files rather than
8812       compile source code, it would previously accept and ignore all
8813       options starting with --, including linker options such as
8814       --as-needed and --export-dynamic, although such options would
8815       result in errors if any source code was compiled. Such options, if
8816       unknown to the compiler, are now rejected in all cases; if the
8817       intent was to pass them to the linker, options such as
8818       -Wl,--as-needed should be used.
8819     * Versions of the GNU C library up to and including 2.11.1 included
8820       an [1]incorrect implementation of the cproj function. GCC optimizes
8821       its builtin cproj according to the behavior specified and allowed
8822       by the ISO C99 standard. If you want to avoid discrepancies between
8823       the C library and GCC's builtin transformations when using cproj in
8824       your code, use GLIBC 2.12 or later. If you are using an older GLIBC
8825       and actually rely on the incorrect behavior of cproj, then you can
8826       disable GCC's transformations using -fno-builtin-cproj.
8827     * The C-only intermodule optimization framework (IMA, enabled by
8828       -combine) has been removed in favor of the new generic link-time
8829       optimization framework (LTO) introduced in [2]GCC 4.5.0.
8830     * GCC now ships with the LGPL-licensed libquadmath library, which
8831       provides quad-precision mathematical functions for targets with a
8832       __float128 datatype. __float128 is available for targets on 32-bit
8833       x86, x86-64 and Itanium architectures. The libquadmath library is
8834       automatically built on such targets when building the Fortran
8835       compiler.
8836     * New -Wunused-but-set-variable and -Wunused-but-set-parameter
8837       warnings were added for C, C++, Objective-C and Objective-C++.
8838       These warnings diagnose variables respective parameters which are
8839       only set in the code and never otherwise used. Usually such
8840       variables are useless and often even the value assigned to them is
8841       computed needlessly, sometimes expensively. The
8842       -Wunused-but-set-variable warning is enabled by default by -Wall
8843       flag and -Wunused-but-set-parameter by -Wall -Wextra flags.
8844     * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
8845       rules for the layout of vectors that could lead to wrong code being
8846       generated. Vectors larger than 8 bytes in size are now by default
8847       aligned to an 8-byte boundary. This is an ABI change: code that
8848       makes explicit use of vector types may be incompatible with binary
8849       objects built with older versions of GCC. Auto-vectorized code is
8850       not affected by this change. (This change affects GCC versions
8851       4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.)
8852     * On AVR, variables with the progmem attribute to locate data in
8853       flash memory must be qualified as const.
8854     * Support for a number of older systems and recently unmaintained or
8855       untested target ports of GCC has been declared obsolete in GCC 4.6.
8856       Unless there is activity to revive them, the next release of GCC
8857       will have their sources permanently removed.
8858       All GCC ports for the following processor architectures have been
8859       declared obsolete:
8860          + Argonaut ARC (arc-*)
8861          + National Semiconductor CRX (crx-*)
8862          + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*,
8863            m68hc12-*-*, m6812-*-*)
8864          + Sunplus S+core (score-*)
8865       The following ports for individual systems on particular
8866       architectures have been obsoleted:
8867          + Interix (i[34567]86-*-interix3*)
8868          + NetWare x86 (i[3456x]86-*-netware*)
8869          + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*)
8870          + MCore PE (mcore-*-pe*)
8871          + SH SymbianOS (sh*-*-symbianelf*)
8872          + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*)
8873          + M68K uClinux old ABI (m68k-*-uclinuxoldabi*)
8874          + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*,
8875            vax-*-netbsd*, but not *-*-netbsdelf*)
8876       The i[34567]86-*-pe alias for Cygwin targets has also been
8877       obsoleted; users should configure for i[34567]86-*-cygwin* instead.
8878       Certain configure options to control the set of libraries built
8879       with GCC on some targets have been obsoleted. On ARM targets, the
8880       options --disable-fpu, --disable-26bit, --disable-underscore,
8881       --disable-interwork, --disable-biendian and --disable-nofmult have
8882       been obsoleted. On MIPS targets, the options
8883       --disable-single-float, --disable-biendian and --disable-softfloat
8884       have been obsoleted.
8885     * Support has been removed for all the [3]configurations obsoleted in
8886       GCC 4.5.
8887     * More information on porting to GCC 4.6 from previous versions of
8888       GCC can be found in the [4]porting guide for this release.
8889
8890General Optimizer Improvements
8891
8892     * A new general optimization level, -Ofast, has been introduced. It
8893       combines the existing optimization level -O3 with options that can
8894       affect standards compliance but result in better optimized code.
8895       For example, -Ofast enables -ffast-math.
8896     * Link-time optimization improvements:
8897          + The [5]Scalable Whole Program Optimizer (WHOPR) project has
8898            stabilized to the point of being usable. It has become the
8899            default mode when using the LTO optimization model. Link time
8900            optimization can now split itself into multiple parallel
8901            compilations. Parallelism is controlled with -flto=n (where n
8902            specifies the number of compilations to execute in parallel).
8903            GCC can also cooperate with a GNU make job server by
8904            specifying the -flto=jobserver option and adding + to the
8905            beginning of the Makefile rule executing the linker.
8906            Classical LTO mode can be enforced by -flto-partition=none.
8907            This may result in small code quality improvements.
8908          + A large number of bugs were fixed. GCC itself, Mozilla Firefox
8909            and other large applications can be built with LTO enabled.
8910          + The linker plugin support improvements
8911               o Linker plugin is now enabled by default when the linker
8912                 is detected to have plugin support. This is the case for
8913                 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and
8914                 the Gold linker on ELF targets. Plugin support of the
8915                 Apple linker on Darwin is not compatible with GCC. The
8916                 linker plugin can also be controlled by the
8917                 -fuse-linker-plugin command-line option.
8918               o Resolution information from the linker plugin is used to
8919                 drive whole program assumptions. Use of the linker plugin
8920                 results in more aggressive optimization on binaries and
8921                 on shared libraries that use the hidden visibility
8922                 attribute. Consequently the use of -fwhole-program is not
8923                 necessary in addition to LTO.
8924          + Hidden symbols used from non-LTO objects now have to be
8925            explicitly annotated with externally_visible when the linker
8926            plugin is not used.
8927          + C++ inline functions and virtual tables are now privatized
8928            more aggressively, leading to better inter-procedural
8929            optimization and faster dynamic linking.
8930          + Memory usage and intermediate language streaming performance
8931            have been improved.
8932          + Static constructors and destructors from individual units are
8933            inlined into a single function. This can significantly improve
8934            startup times of large C++ applications where static
8935            constructors are very common. For example, static constructors
8936            are used when including the iostream header.
8937          + Support for the Ada language has been added.
8938     * Interprocedural optimization improvements
8939          + The interprocedural framework was re-tuned for link time
8940            optimization. Several scalability issues were resolved.
8941          + Improved auto-detection of const and pure functions. Newly,
8942            noreturn functions are auto-detected.
8943            The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is
8944            available that informs users when adding attributes to headers
8945            might improve code generation.
8946          + A number of inlining heuristic improvements. In particular:
8947               o Partial inlining is now supported and enabled by default
8948                 at -O2 and greater. The feature can be controlled via
8949                 -fpartial-inlining.
8950                 Partial inlining splits functions with short hot path to
8951                 return. This allows more aggressive inlining of the hot
8952                 path leading to better performance and often to code size
8953                 reductions (because cold parts of functions are not
8954                 duplicated).
8955               o Scalability for large compilation units was improved
8956                 significantly.
8957               o Inlining of callbacks is now more aggressive.
8958               o Virtual methods are considered for inlining when the
8959                 caller is inlined and devirtualization is then possible.
8960               o Inlining when optimizing for size (either in cold regions
8961                 of a program or when compiling with -Os) was improved to
8962                 better handle C++ programs with larger abstraction
8963                 penalty, leading to smaller and faster code.
8964          + The IPA reference optimization pass detecting global variables
8965            used or modified by functions was strengthened and sped up.
8966          + Functions whose address was taken are now optimized out when
8967            all references to them are dead.
8968          + A new inter-procedural static profile estimation pass detects
8969            functions that are executed once or unlikely to be executed.
8970            Unlikely executed functions are optimized for size. Functions
8971            executed once are optimized for size except for the inner
8972            loops.
8973          + On most targets with named section support, functions used
8974            only at startup (static constructors and main), functions used
8975            only at exit and functions detected to be cold are placed into
8976            separate text segment subsections. This extends the
8977            -freorder-functions feature and is controlled by the same
8978            switch. The goal is to improve the startup time of large C++
8979            programs.
8980            Proper function placement requires linker support. GNU ld
8981            2.21.51 on ELF targets was updated to place those functions
8982            together within the text section leading to better code
8983            locality and faster startup times of large C++ programs. The
8984            feature is also supported in the Apple linker. Support in the
8985            gold linker is planned.
8986     * A new switch -fstack-usage has been added. It makes the compiler
8987       output stack usage information for the program, on a per-function
8988       basis, in an auxiliary file.
8989     * A new switch -fcombine-stack-adjustments has been added. It can be
8990       used to enable or disable the compiler's stack-slot combining pass
8991       which before was enabled automatically at -O1 and above, but could
8992       not be controlled on its own.
8993     * A new switch -fstrict-volatile-bitfields has been added. Using it
8994       indicates that accesses to volatile bitfields should use a single
8995       access of the width of the field's type. This option can be useful
8996       for precisely defining and accessing memory-mapped peripheral
8997       registers from C or C++.
8998
8999Compile time and memory usage improvements
9000
9001     * Datastructures used by the dataflow framework in GCC were
9002       reorganized for better memory usage and more cache locality.
9003       Compile time is improved especially on units with large functions
9004       (possibly resulting from a lot of inlining) not fitting into the
9005       processor cache. The compile time of the GCC C compiler binary with
9006       link-time optimization went down by over 10% (benchmarked on x86-64
9007       target).
9008
9009New Languages and Language specific improvements
9010
9011  Ada
9012
9013     * Stack checking has been improved on selected architectures (Alpha,
9014       IA-32/x86-64, RS/6000 and SPARC): it now will detect stack
9015       overflows in all cases on these architectures.
9016     * Initial support for Ada 2012 has been added.
9017
9018  C family
9019
9020     * A new warning, enabled by -Wdouble-promotion, has been added that
9021       warns about cases where a value of type float is implicitly
9022       promoted to double. This is especially helpful for CPUs that handle
9023       the former in hardware, but emulate the latter in software.
9024     * A new function attribute leaf was introduced. This attribute allows
9025       better inter-procedural optimization across calls to functions that
9026       return to the current unit only via returning or exception
9027       handling. This is the case for most library functions that have no
9028       callbacks.
9029     * Support for a new data type __int128 for targets having wide enough
9030       machine-mode support.
9031     * The new function attribute callee_pop_aggregate allows to specify
9032       if the caller or callee is responsible for popping the aggregate
9033       return pointer value from the stack.
9034     * Support for selectively enabling and disabling warnings via #pragma
9035       GCC diagnostic has been added. For instance:
9036#pragma GCC diagnostic error "-Wuninitialized"
9037  foo(a);                       /* error is given for this one */
9038#pragma GCC diagnostic push
9039#pragma GCC diagnostic ignored "-Wuninitialized"
9040  foo(b);                       /* no diagnostic for this one */
9041#pragma GCC diagnostic pop
9042  foo(c);                       /* error is given for this one */
9043#pragma GCC diagnostic pop
9044  foo(d);                       /* depends on command-line options */
9045
9046     * The -fmax-errors=N option is now supported. Using this option
9047       causes the compiler to exit after N errors have been issued.
9048
9049  C
9050
9051     * There is now experimental support for some features from the
9052       upcoming C1X revision of the ISO C standard. This support may be
9053       selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions.
9054       Note that this support is experimental and may change incompatibly
9055       in future releases for consistency with changes to the C1X standard
9056       draft. The following features are newly supported as described in
9057       the N1539 draft of C1X (with changes agreed at the March 2011 WG14
9058       meeting); some other features were already supported with no
9059       compiler changes being needed, or have some support but not in full
9060       accord with N1539 (as amended).
9061          + Static assertions (_Static_assert keyword)
9062          + Typedef redefinition
9063          + New macros in <float.h>
9064          + Anonymous structures and unions
9065     * The new -fplan9-extensions option directs the compiler to support
9066       some extensions for anonymous struct fields which are implemented
9067       by the Plan 9 compiler. A pointer to a struct may be automatically
9068       converted to a pointer to an anonymous field when calling a
9069       function, in order to make the types match. An anonymous struct
9070       field whose type is a typedef name may be referred to using the
9071       typedef name.
9072
9073  C++
9074
9075     * Improved [7]experimental support for the upcoming C++0x ISO C++
9076       standard, including support for constexpr (thanks to Gabriel Dos
9077       Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide),
9078       noexcept, unrestricted unions, range-based for loops (thanks to
9079       Rodrigo Rivas Costa), opaque enum declarations (thanks also to
9080       Rodrigo), implicitly deleted functions and implicit move
9081       constructors.
9082     * When an extern declaration within a function does not match a
9083       declaration in the enclosing context, G++ now properly declares the
9084       name within the namespace of the function rather than the namespace
9085       which was open just before the function definition ([8]c++/43145).
9086     * GCC now warns by default when casting integers to larger pointer
9087       types. These warnings can be disabled with the option
9088       -Wno-int-to-pointer-cast, which is now also available in C++.
9089     * G++ no longer optimizes using the assumption that a value of
9090       enumeration type will fall within the range specified by the
9091       standard, since that assumption is easily violated with a
9092       conversion from integer type ([9]c++/43680). The old behavior can
9093       be restored with -fstrict-enums.
9094     * The new -fnothrow-opt flag changes the semantics of a throw()
9095       exception specification to match the proposed semantics of the
9096       noexcept specification: just call terminate if an exception tries
9097       to propagate out of a function with such an exception
9098       specification. This dramatically reduces or eliminates the code
9099       size overhead from adding the exception specification.
9100     * The new -Wnoexcept flag will suggest adding a noexcept qualifier to
9101       a function that the compiler can tell doesn't throw if it would
9102       change the value of a noexcept expression.
9103     * The -Wshadow option now warns if a local variable or type
9104       declaration shadows another type in C++. Note that the compiler
9105       will not warn if a local variable shadows a struct/class/enum, but
9106       will warn if it shadows an explicit typedef.
9107     * When an identifier is not found in the current scope, G++ now
9108       offers suggestions about which identifier might have been intended.
9109     * G++ now issues clearer diagnostics for missing semicolons after
9110       class, struct, and union definitions.
9111     * G++ now issues clearer diagnostics for missing semicolons after
9112       class member declarations.
9113     * G++ now issues clearer diagnostics when a colon is used in a place
9114       where a double-colon was intended.
9115     * G++ no longer accepts mutable on reference members ([10]c++/33558).
9116       Use -fpermissive to allow the old, non-conforming behaviour.
9117     * A few mangling fixes have been made, to attribute const/volatile on
9118       function pointer types, decltype of a plain decl, and use of a
9119       function parameter in the declaration of another parameter. By
9120       default the compiler still uses the old mangling, but emits aliases
9121       with the new mangling on targets that support strong aliases. Users
9122       can switch over entirely to the new mangling with -fabi-version=5
9123       or -fabi-version=0. -Wabi will now warn about code that uses the
9124       old mangling.
9125     * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified
9126       type to be default initialized unless the type has a user-declared
9127       default constructor. In 4.6.2 G++ implements the proposed
9128       resolution of [11]DR 253, so default initialization is allowed if
9129       it initializes all subobjects. Code that fails to compile can be
9130       fixed by providing an initializer e.g.
9131    struct A { A(); };
9132    struct B : A { int i; };
9133    const B b = B();
9134       Use -fpermissive to allow the old, non-conforming behaviour.
9135
9136    Runtime Library (libstdc++)
9137
9138     * [12]Improved experimental support for the upcoming ISO C++
9139       standard, C++0x, including using constexpr and nullptr.
9140     * Performance improvements to the [13]Debug Mode, thanks to François
9141       Dumont.
9142     * Atomic operations used for reference-counting are annotated so that
9143       they can be understood by race detectors such as Helgrind, see
9144       [14]Data Race Hunting.
9145     * Most libstdc++ standard headers have been changed to no longer
9146       include the cstddef header as an implementation detail. Code that
9147       relied on that header being included as side-effect of including
9148       other standard headers will need to include cstddef explicitly.
9149
9150  Fortran
9151
9152     * On systems supporting the libquadmath library, GNU Fortran now also
9153       supports a quad-precision, kind=16 floating-point data type
9154       (REAL(16), COMPLEX(16)). As the data type is not fully supported in
9155       hardware, calculations might be one to two orders of magnitude
9156       slower than with the 4, 8 or 10 bytes floating-point data types.
9157       This change does not affect systems which support REAL(16) in
9158       hardware nor those which do not support libquadmath.
9159     * Much improved compile time for large array constructors.
9160     * In order to reduce execution time and memory consumption, use of
9161       temporary arrays in assignment expressions is avoided for many
9162       cases. The compiler now reverses loops in order to avoid generating
9163       a temporary array where possible.
9164     * Improved diagnostics, especially with -fwhole-file.
9165     * The -fwhole-file flag is now enabled by default. This improves code
9166       generation and diagnostics. It can be disabled using the deprecated
9167       -fno-whole-file flag.
9168     * Support the generation of Makefile dependencies via the [15]-M...
9169       flags of GCC; you may need to specify the -cpp option in addition.
9170       The dependencies take modules, Fortran's include, and CPP's
9171       #include into account. Note: Using -M for the module path is no
9172       longer supported, use -J instead.
9173     * The flag -Wconversion has been modified to only issue warnings
9174       where a conversion leads to information loss. This drastically
9175       reduces the number of warnings; -Wconversion is thus now enabled
9176       with -Wall. The flag -Wconversion-extra has been added and also
9177       warns about other conversions; -Wconversion-extra typically issues
9178       a huge number of warnings, most of which can be ignored.
9179     * A new command-line option -Wunused-dummy-argument warns about
9180       unused dummy arguments and is included in -Wall. Before,
9181       -Wunused-variable also warned about unused dummy arguments.
9182     * Fortran 2003 support has been extended:
9183          + Improved support for polymorphism between libraries and
9184            programs and for complicated inheritance patterns (cf.
9185            [16]object-oriented programming).
9186          + Experimental support of the ASSOCIATE construct.
9187          + In pointer assignments it is now possible to specify the lower
9188            bounds of the pointer and, for a rank-1 or a simply contiguous
9189            data-target, to remap the bounds.
9190          + Automatic (re)allocation: In intrinsic assignments to
9191            allocatable variables the left-hand side will be automatically
9192            allocated (if unallocated) or reallocated (if the shape or
9193            type parameter is different). To avoid the small performance
9194            penalty, you can use a(:) = ... instead of a = ... for arrays
9195            and character strings – or disable the feature using -std=f95
9196            or -fno-realloc-lhs.
9197          + Deferred type parameter: For scalar allocatable and pointer
9198            variables the character length can be deferred.
9199          + Namelist variables with allocatable and pointer attribute and
9200            nonconstant length type parameter are supported.
9201     * Fortran 2008 support has been extended:
9202          + Experimental [17]coarray support (for one image only, i.e.
9203            num_images() == 1); use the [18]-fcoarray=single flag to
9204            enable it.
9205          + The STOP and the new ERROR STOP statements now support all
9206            constant expressions.
9207          + Support for the CONTIGUOUS attribute.
9208          + Support for ALLOCATE with MOLD.
9209          + Support for the STORAGE_SIZE intrinsic inquiry function.
9210          + Support of the NORM2 and PARITY intrinsic functions.
9211          + The following bit intrinsics were added: POPCNT and POPPAR for
9212            counting the number of 1 bits and returning the parity; BGE,
9213            BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR
9214            for combined left and right shifts, MASKL and MASKR for simple
9215            left and right justified masks, MERGE_BITS for a bitwise merge
9216            using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations,
9217            and the transformational bit intrinsics IALL, IANY and
9218            IPARITY.
9219          + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine.
9220          + Support for the IMPURE attribute for procedures, which allows
9221            for ELEMENTAL procedures without the restrictions of PURE.
9222          + Null pointers (including NULL()) and not allocated variables
9223            can be used as actual argument to optional non-pointer,
9224            non-allocatable dummy arguments, denoting an absent argument.
9225          + Non-pointer variables with TARGET attribute can be used as
9226            actual argument to POINTER dummies with INTENT(IN)
9227          + Pointers including procedure pointers and those in a derived
9228            type (pointer components) can now be initialized by a target
9229            instead of only by NULL.
9230          + The EXIT statement (with construct-name) can now be used to
9231            leave not only the DO but also the ASSOCIATE, BLOCK, IF,
9232            SELECT CASE and SELECT TYPE constructs.
9233          + Internal procedures can now be used as actual argument.
9234          + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS
9235            and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV
9236            have been added; these arrays contain the supported kind
9237            values for the respective types.
9238          + The module procedures C_SIZEOF of the intrinsic module
9239            ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of
9240            ISO_FORTRAN_ENV have been implemented.
9241          + Minor changes: obsolescence diagnostics for ENTRY was added
9242            for -std=f2008; a line may start with a semicolon; for
9243            internal and module procedures END can be used instead of END
9244            SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes
9245            a RADIX argument; intrinsic types are supported for
9246            TYPE(intrinsic-type-spec); multiple type-bound procedures can
9247            be declared in a single PROCEDURE statement; implied-shape
9248            arrays are supported for named constants (PARAMETER). The
9249            transformational, three argument versions of BESSEL_JN and
9250            BESSEL_YN were added – the elemental, two-argument version had
9251            been added in GCC 4.4; note that the transformational
9252            functions use a recurrence algorithm.
9253
9254  Go
9255
9256   Support for the Go programming language has been added to GCC. It is
9257   not enabled by default when you build GCC; use the --enable-languages
9258   configure option to build it. The driver program for compiling Go code
9259   is gccgo.
9260
9261   Go is currently known to work on GNU/Linux and RTEMS. Solaris support
9262   is in progress. It may or may not work on other platforms.
9263
9264  Objective-C and Objective-C++
9265
9266     * The -fobjc-exceptions flag is now required to enable Objective-C
9267       exception and synchronization syntax (introduced by the keywords
9268       @try, @catch, @finally and @synchronized).
9269     * A number of Objective-C 2.0 features and extensions are now
9270       supported by GCC. These features are enabled by default; you can
9271       disable them by using the new -fobjc-std=objc1 command-line option.
9272     * The Objective-C 2.0 dot-syntax is now supported. It is an
9273       alternative syntax for using getters and setters; object.count is
9274       automatically converted into [object count] or [object setCount:
9275       ...] depending on context; for example if (object.count > 0) is
9276       automatically compiled into the equivalent of if ([object count] >
9277       0) while object.count = 0; is automatically compiled into the
9278       equivalent ot [object setCount: 0];. The dot-syntax can be used
9279       with instance and class objects and with any setters or getters, no
9280       matter if they are part of a declared property or not.
9281     * Objective-C 2.0 declared properties are now supported. They are
9282       declared using the new @property keyword, and are most commonly
9283       used in conjunction with the new Objective-C 2.0 dot-syntax. The
9284       nonatomic, readonly, readwrite, assign, retain, copy, setter and
9285       getter attributes are all supported. Marking declared properties
9286       with __attribute__ ((deprecated)) is supported too.
9287     * The Objective-C 2.0 @synthesize and @dynamic keywords are
9288       supported. @synthesize causes the compiler to automatically
9289       synthesize a declared property, while @dynamic is used to disable
9290       all warnings for a declared property for which no implementation is
9291       provided at compile time. Synthesizing declared properties requires
9292       runtime support in most useful cases; to be able to use it with the
9293       GNU runtime, appropriate helper functions have been added to the
9294       GNU Objective-C runtime ABI, and are implemented by the GNU
9295       Objective-C runtime library shipped with GCC.
9296     * The Objective-C 2.0 fast enumeration syntax is supported in
9297       Objective-C. This is currently not yet available in Objective-C++.
9298       Fast enumeration requires support in the runtime, and such support
9299       has been added to the GNU Objective-C runtime library (shipped with
9300       GCC).
9301     * The Objective-C 2.0 @optional keyword is supported. It allows you
9302       to mark methods or properties in a protocol as optional as opposed
9303       to required.
9304     * The Objective-C 2.0 @package keyword is supported. It has currently
9305       the same effect as the @public keyword.
9306     * Objective-C 2.0 method attributes are supported. Currently the
9307       supported attributes are deprecated, sentinel, noreturn and format.
9308     * Objective-C 2.0 method argument attributes are supported. The most
9309       widely used attribute is unused, to mark an argument as unused in
9310       the implementation.
9311     * Objective-C 2.0 class and protocol attributes are supported.
9312       Currently the only supported attribute is deprecated.
9313     * Objective-C 2.0 class extensions are supported. A class extension
9314       has the same syntax as a category declaration with no category
9315       name, and the methods and properties declared in it are added
9316       directly to the main class. It is mostly used as an alternative to
9317       a category to add methods to a class without advertising them in
9318       the public headers, with the advantage that for class extensions
9319       the compiler checks that all the privately declared methods are
9320       actually implemented.
9321     * As a result of these enhancements, GCC can now be used to build
9322       Objective-C and Objective-C++ software that uses Foundation and
9323       other important system frameworks with the NeXT runtime on Darwin 9
9324       and Darwin 10 (OSX 10.5 and 10.6).
9325     * Many bugs in the compiler have been fixed in this release; in
9326       particular, LTO can now be used when compiling Objective-C and
9327       Objective-C++ and the parser is much more robust in dealing with
9328       invalid code.
9329
9330    Runtime Library (libobjc)
9331
9332     * The GNU Objective-C runtime library now defines the macro
9333       __GNU_LIBOBJC__ (with a value that is increased at every release
9334       where there is any change to the API) in objc/objc.h, making it
9335       easy to determine if the GNU Objective-C runtime library is being
9336       used, and if so, which version. Previous versions of the GNU
9337       Objective-C runtime library (and other Objective-C runtime
9338       libraries such as the Apple one) do not define this macro.
9339     * A new Objective-C 2.0 API, almost identical to the one implemented
9340       by the Apple Objective-C runtime, has been implemented in the GNU
9341       Objective-C runtime library. The new API hides the internals of
9342       most runtime structures but provides a more extensive set of
9343       functions to operate on them. It is much easier, for example, to
9344       create or modify classes at runtime. The new API also makes it
9345       easier to port software from Apple to GNU as almost no changes
9346       should be required. The old API is still supported for backwards
9347       compatibility; including the old objc/objc-api.h header file
9348       automatically selects the old API, while including the new
9349       objc/runtime.h header file automatically selects the new API.
9350       Support for the old API is being phased out and upgrading the
9351       software to use the new API is strongly recommended. To check for
9352       the availability of the new API, the __GNU_LIBOBJC__ macro can be
9353       used as older versions of the GNU Objective-C runtime library,
9354       which do not support the new API, do not define such a macro.
9355     * Runtime support for @synchronized has been added.
9356     * Runtime support for Objective-C 2.0 synthesized property accessors
9357       has been added.
9358     * Runtime support for Objective-C 2.0 fast enumeration has been
9359       added.
9360
9361New Targets and Target Specific Improvements
9362
9363  ARM
9364
9365     * GCC now supports the Cortex-M4 processor implementing the v7-em
9366       version of the architecture using the option -mcpu=cortex-m4.
9367     * Scheduling descriptions for the Cortex-M4, the Neon and the
9368       floating point units of the Cortex-A9 and a pipeline description
9369       for the Cortex-A5 have been added.
9370     * Synchronization primitives such as __sync_fetch_and_add and friends
9371       are now inlined for supported architectures rather than calling
9372       into a kernel helper function.
9373     * SSA loop prefetching is enabled by default for the Cortex-A9 at
9374       -O3.
9375     * Several improvements were committed to improve code generation for
9376       the ARM architecture including a rewritten implementation for load
9377       and store multiples.
9378     * Several enhancements were committed to improve SIMD code generation
9379       for NEON by adding support for widening instructions, misaligned
9380       loads and stores, vector conditionals and support for 64 bit
9381       arithmetic.
9382     * Support was added for the Faraday cores fa526, fa606te, fa626te,
9383       fmp626te, fmp626 and fa726te and can be used with the respective
9384       names as parameters to the -mcpu= option.
9385     * Basic support was added for Cortex-A15 and is available through
9386       -mcpu=cortex-a15.
9387     * GCC for AAPCS configurations now more closely adheres to the AAPCS
9388       specification by enabling -fstrict-volatile-bitfields by default.
9389
9390  IA-32/x86-64
9391
9392     * The new -fsplit-stack option permits programs to use a
9393       discontiguous stack. This is useful for threaded programs, in that
9394       it is no longer necessary to specify the maximum stack size when
9395       creating a thread. This feature is currently only implemented for
9396       32-bit and 64-bit x86 GNU/Linux targets.
9397     * Support for emitting profiler counter calls before function
9398       prologues. This is enabled via a new command-line option -mfentry.
9399     * Optimization for the Intel Core 2 processors is now available
9400       through the -march=core2 and -mtune=core2 options.
9401     * Support for Intel Core i3/i5/i7 processors is now available through
9402       the -march=corei7 and -mtune=corei7 options.
9403     * Support for Intel Core i3/i5/i7 processors with AVX is now
9404       available through the -march=corei7-avx and -mtune=corei7-avx
9405       options.
9406     * Support for AMD Bobcat (family 14) processors is now available
9407       through the -march=btver1 and -mtune=btver1 options.
9408     * Support for AMD Bulldozer (family 15) processors is now available
9409       through the -march=bdver1 and -mtune=bdver1 options.
9410     * The default setting (when not optimizing for size) for 32-bit
9411       GNU/Linux and Darwin x86 targets has been changed to
9412       -fomit-frame-pointer. The default can be reverted to
9413       -fno-omit-frame-pointer by configuring GCC with the
9414       --enable-frame-pointer configure option.
9415     * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support
9416       __float128 on 32-bit and 64-bit x86 targets.
9417     * AVX floating-point arithmetic can now be enabled by default at
9418       configure time with the new --with-fpmath=avx option.
9419     * The SSA loop prefetching pass is enabled when using -O3 when
9420       optimizing for CPUs where prefetching is beneficial (AMD CPUs newer
9421       than K6).
9422     * Support for TBM (Trailing Bit Manipulation) built-in functions and
9423       code generation is available via -mtbm.
9424     * Support for AMD's BMI (Bit Manipulation) built-in functions and
9425       code generation is available via -mbmi.
9426
9427  MicroBlaze
9428
9429     * Support has been added for the Xilinx MicroBlaze softcore processor
9430       (microblaze-elf) embedded target. This configurable processor is
9431       supported on several Xilinx Spartan and Virtex FPGAs.
9432
9433  MIPS
9434
9435     * GCC now supports the Loongson 3A processor. Its canonical -march=
9436       and -mtune= name is loongson3a.
9437
9438  MN10300 / AM33
9439
9440     * The inline assembly register constraint "A" has been renamed "c".
9441       This constraint is used to select a floating-point register that
9442       can be used as the destination of a multiply-accumulate
9443       instruction.
9444     * New inline assembly register constraints "A" and "D" have been
9445       added. These constraint letters resolve to all general registers
9446       when compiling for AM33, and resolve to address registers only or
9447       data registers only when compiling for MN10300.
9448     * The MDR register is represented in the compiler. One can access the
9449       register via the "z" constraint in inline assembly. It can be
9450       marked as clobbered or used as a local register variable via the
9451       "mdr" name. The compiler uses the RETF instruction if the function
9452       does not modify the MDR register, so it is important that inline
9453       assembly properly annotate any usage of the register.
9454
9455  PowerPC/PowerPC64
9456
9457     * GCC now supports the Applied Micro Titan processor with
9458       -mcpu=titan.
9459     * The -mrecip option has been added, which indicates whether the
9460       reciprocal and reciprocal square root instructions should be used.
9461     * The -mveclibabi=mass option can be used to enable the compiler to
9462       autovectorize mathematical functions using the Mathematical
9463       Acceleration Subsystem library.
9464     * The -msingle-pic-base option has been added, which instructs the
9465       compiler to avoid loading the PIC base register in function
9466       prologues. The PIC base register must be initialized by the runtime
9467       system.
9468     * The -mblock-move-inline-limit option has been added, which enables
9469       the user to control the maximum size of inlined memcpy calls and
9470       similar.
9471     * PowerPC64 GNU/Linux support for applications requiring a large TOC
9472       section has been improved. A new command-line option,
9473       -mcmodel=MODEL, controls this feature; valid values for MODEL are
9474       small, medium, or large.
9475     * The Altivec builtin functions vec_ld and vec_st have been modified
9476       to generate the Altivec memory instructions LVX and STVX, even if
9477       the -mvsx option is used. In the initial GCC 4.5 release, these
9478       builtin functions were changed to generate VSX memory reference
9479       instructions instead of Altivec memory instructions, but there are
9480       differences between the two instructions. If the VSX instruction
9481       set is available, you can now use the new builtin functions
9482       vec_vsx_ld and vec_vsx_st which always generates the VSX memory
9483       instructions.
9484     * The GCC compiler on AIX now defaults to a process layout with a
9485       larger data space allowing larger programs to be compiled.
9486     * The GCC long double type on AIX 6.1 and above has reverted to 64
9487       bit double precision, matching the AIX XL compiler default, because
9488       of missing C99 symbols required by the GCC runtime.
9489     * The default processor scheduling model and tuning for PowerPC64
9490       GNU/Linux and for AIX 6.1 and above now is POWER7.
9491     * Starting with GCC 4.6.1, vectors of type vector long long or vector
9492       long are passed and returned in the same method as other vectors
9493       with the VSX instruction set. Previously the GCC compiler did not
9494       adhere to the ABI for 128-bit vectors with 64-bit integer base
9495       types (PR 48857). This is also fixed in the GCC 4.5.4 release.
9496
9497  S/390, zSeries and System z9/z10, IBM zEnterprise z196
9498
9499     * Support for the zEnterprise z196 processor has been added. When
9500       using the -march=z196 option, the compiler will generate code
9501       making use of the following instruction facilities:
9502          + Conditional load/store
9503          + Distinct-operands
9504          + Floating-point-extension
9505          + Interlocked-access
9506          + Population-count
9507       The -mtune=z196 option avoids the compare and branch instructions
9508       as well as the load address instruction with an index register as
9509       much as possible and performs instruction scheduling appropriate
9510       for the new out-of-order pipeline architecture.
9511     * When using the -m31 -mzarch options the generated code still
9512       conforms to the 32-bit ABI but uses the general purpose registers
9513       as 64-bit registers internally. This requires a Linux kernel saving
9514       the whole 64-bit registers when doing a context switch. Kernels
9515       providing that feature indicate that by the 'highgprs' string in
9516       /proc/cpuinfo.
9517     * The SSA loop prefetching pass is enabled when using -O3.
9518
9519  SPARC
9520
9521     * GCC now supports the LEON series of SPARC V8 processors. The code
9522       generated by the compiler can either be tuned to it by means of the
9523       --with-tune=leon configure option and -mtune=leon compilation
9524       option, or the compiler can be built for the sparc-leon-{elf,linux}
9525       and sparc-leon3-{elf,linux} targets directly.
9526     * GCC has stopped sign/zero-extending parameter registers in the
9527       callee for functions taking parameters with sub-word size in 32-bit
9528       mode, since this is redundant with the specification of the ABI.
9529       GCC has never done so in 64-bit mode since this is also redundant.
9530     * The command-line option -mfix-at697f has been added to enable the
9531       documented workaround for the single erratum of the Atmel AT697F
9532       processor.
9533
9534Operating Systems
9535
9536  Android
9537
9538     * GCC now supports the Bionic C library and provides a convenient way
9539       of building native libraries and applications for the Android
9540       platform. Refer to the documentation of the -mandroid and -mbionic
9541       options for details on building native code. At the moment, Android
9542       support is enabled only for ARM.
9543
9544  Darwin/Mac OS X
9545
9546     * General
9547          + Initial support for CFString types has been added.
9548            This allows GCC to build projects including the system Core
9549            Foundation frameworks. The GCC Objective-C family supports
9550            CFString "toll-free bridged" as per the Mac OS X system tools.
9551            CFString is also recognized in the context of format
9552            attributes and arguments (see the documentation for format
9553            attributes for limitations). At present, 8-bit character types
9554            are supported.
9555          + Object file size reduction.
9556            The Darwin zeroed memory allocators have been re-written to
9557            make more use of .zerofill sections. For non-debug code, this
9558            can reduce object file size significantly.
9559          + Objective-C family 64-bit support (NeXT ABI 2).
9560            Initial support has been added to support 64-bit Objective-C
9561            code using the Darwin/OS X native (NeXT) runtime. ABI version
9562            2 will be selected automatically when 64-bit code is built.
9563          + Objective-C family 32-bit ABI 1.
9564            For 32-bit code ABI 1 is also now also allowed. At present it
9565            must be selected manually using -fobjc-abi-version=1 where
9566            applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6).
9567     * x86 Architecture
9568          + The -mdynamic-no-pic option has been enabled.
9569            Code supporting -mdynamic-no-pic optimization has been added
9570            and is applicable to -m32 builds. The compiler bootstrap uses
9571            the option where appropriate.
9572          + The default value for -mtune= has been changed.
9573            Since Darwin systems are primarily Xeon, Core-2 or similar the
9574            default tuning has been changed to -mtune=core2.
9575          + Enable 128-bit long double (__float128) support on Darwin.
9576     * PPC Architecture
9577          + Darwin64 ABI.
9578            Several significant bugs have been fixed, such that GCC now
9579            produces code compatible with the Darwin64 PowerPC ABI.
9580          + libffi and boehm-gc.
9581            The Darwin ports of the libffi and boehm-gc libraries have
9582            been upgraded to include a Darwin64 implementation. This means
9583            that powerpc*-*-darwin9 platforms may now, for example, build
9584            Java applications with -m64 enabled.
9585          + Plug-in support has been enabled.
9586          + The -fsection-anchors option is now available although,
9587            presently, not heavily tested.
9588
9589  Solaris 2
9590
9591    New Features
9592
9593     * Support symbol versioning with the Sun linker.
9594     * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+.
9595     * Support thread-local storage (TLS) with the Sun assembler on
9596       Solaris 2/x86.
9597     * Support TLS on Solaris 8/9 if prerequisites are met.
9598     * Support COMDAT group with the GNU assembler and recent Sun linker.
9599     * Support the Sun assembler visibility syntax.
9600     * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp.
9601       -march=pentiumpro (Solaris 8/9).
9602     * Don't use SSE on Solaris 8/9 x86 by default.
9603     * Enable 128-bit long double (__float128) support on Solaris 2/x86.
9604
9605    ABI Change
9606
9607     * Change the ABI for returning 8-byte vectors like __m64 in MMX
9608       registers on Solaris 10+/x86 to match the Sun Studio 12.1+
9609       compilers. This is an incompatible change. If you use such types,
9610       you must either recompile all your code with the new compiler or
9611       use the new -mvect8-ret-in-mem option to remain compatible with
9612       previous versions of GCC and Sun Studio.
9613
9614  Windows x86/x86_64
9615
9616     * Initial support for decimal floating point.
9617     * Support for the __thiscall calling-convention.
9618     * Support for hot-patchable function prologues via the
9619       ms_hook_prologue attribute for x86_64 in addition to 32-bit x86.
9620     * Improvements of stack-probing and stack-allocation mechanisms.
9621     * Support of push/pop-macro pragma as preprocessor command.
9622       With #pragma push_macro("macro-name") the current definition of
9623       macro-name is saved and can be restored with #pragma
9624       pop_macro("macro-name") to its saved definition.
9625     * Enable 128-bit long double (__float128) support on MinGW and
9626       Cygwin.
9627
9628Other significant improvements
9629
9630  Installation changes
9631
9632     * An install-strip make target is provided that installs stripped
9633       executables, and may install libraries with unneeded or debugging
9634       sections stripped.
9635     * On Power7 systems, there is a potential problem if you build the
9636       GCC compiler with a host compiler using options that enable the VSX
9637       instruction set generation. If the host compiler has been patched
9638       so that the vec_ld and vec_st builtin functions generate Altivec
9639       memory instructions instead of VSX memory instructions, then you
9640       should be able to build the compiler with VSX instruction
9641       generation.
9642
9643Changes for GCC Developers
9644
9645   Note: these changes concern developers that develop GCC itself or
9646   software that integrates with GCC, such as plugins, and not the general
9647   GCC users.
9648     * The gengtype utility, which previously was internal to the GCC
9649       build process, has been enchanced to provide GC root information
9650       for plugins as necessary.
9651     * The old GC allocation interface of ggc_alloc and friends was
9652       replaced with a type-safe alternative.
9653
9654GCC 4.6.1
9655
9656   This is the [19]list of problem reports (PRs) from GCC's bug tracking
9657   system that are known to be fixed in the 4.6.1 release. This list might
9658   not be complete (that is, it is possible that some PRs that have been
9659   fixed are not listed here).
9660
9661GCC 4.6.2
9662
9663   This is the [20]list of problem reports (PRs) from GCC's bug tracking
9664   system that are known to be fixed in the 4.6.2 release. This list might
9665   not be complete (that is, it is possible that some PRs that have been
9666   fixed are not listed here).
9667
9668GCC 4.6.3
9669
9670   This is the [21]list of problem reports (PRs) from GCC's bug tracking
9671   system that are known to be fixed in the 4.6.3 release. This list might
9672   not be complete (that is, it is possible that some PRs that have been
9673   fixed are not listed here).
9674
9675GCC 4.6.4
9676
9677   This is the [22]list of problem reports (PRs) from GCC's bug tracking
9678   system that are known to be fixed in the 4.6.4 release. This list might
9679   not be complete (that is, it is possible that some PRs that have been
9680   fixed are not listed here).
9681
9682
9683    For questions related to the use of GCC, please consult these web
9684    pages and the [23]GCC manuals. If that fails, the
9685    [24]gcc-help@gcc.gnu.org mailing list might help. Comments on these
9686    web pages and the development of GCC are welcome on our developer
9687    list at [25]gcc@gcc.gnu.org. All of [26]our lists have public
9688    archives.
9689
9690   Copyright (C) [27]Free Software Foundation, Inc. Verbatim copying and
9691   distribution of this entire article is permitted in any medium,
9692   provided this notice is preserved.
9693
9694   These pages are [28]maintained by the GCC team. Last modified
9695   2021-12-05[29].
9696
9697References
9698
9699   1. https://sourceware.org/bugzilla/show_bug.cgi?id=10401
9700   2. http://gcc.gnu.org/gcc-4.5/changes.html
9701   3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted
9702   4. http://gcc.gnu.org/gcc-4.6/porting_to.html
9703   5. http://gcc.gnu.org/projects/lto/whopr.pdf
9704   6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
9705   7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html
9706   8. https://gcc.gnu.org/PR43145
9707   9. https://gcc.gnu.org/PR43680
9708  10. https://gcc.gnu.org/PR33558
9709  11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253
9710  12. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x
9711  13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html
9712  14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races
9713  15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html
9714  16. https://gcc.gnu.org/wiki/OOP
9715  17. https://gcc.gnu.org/wiki/Coarray
9716  18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233
9717  19. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1
9718  20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2
9719  21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3
9720  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4
9721  23. https://gcc.gnu.org/onlinedocs/
9722  24. mailto:gcc-help@gcc.gnu.org
9723  25. mailto:gcc@gcc.gnu.org
9724  26. https://gcc.gnu.org/lists.html
9725  27. https://www.fsf.org/
9726  28. https://gcc.gnu.org/about.html
9727  29. http://validator.w3.org/check/referer
9728======================================================================
9729http://gcc.gnu.org/gcc-4.5/index.html
9730                             GCC 4.5 Release Series
9731
9732   (This release series is no longer supported.)
9733
9734   Jul 2, 2012
9735
9736   The [1]GNU project and the GCC developers are pleased to announce the
9737   release of GCC 4.5.4.
9738
9739   This release is a bug-fix release, containing fixes for regressions in
9740   GCC 4.5.3 relative to previous releases of GCC.
9741
9742Release History
9743
9744   GCC 4.5.4
9745          Jul 2, 2012 ([2]changes)
9746
9747   GCC 4.5.3
9748          Apr 28, 2011 ([3]changes)
9749
9750   GCC 4.5.2
9751          Dec 16, 2010 ([4]changes)
9752
9753   GCC 4.5.1
9754          Jul 31, 2010 ([5]changes)
9755
9756   GCC 4.5.0
9757          April 14, 2010 ([6]changes)
9758
9759References and Acknowledgements
9760
9761   GCC used to stand for the GNU C Compiler, but since the compiler
9762   supports several other languages aside from C, it now stands for the
9763   GNU Compiler Collection.
9764
9765   A list of [7]successful builds is updated as new information becomes
9766   available.
9767
9768   The GCC developers would like to thank the numerous people that have
9769   contributed new features, improvements, bug fixes, and other changes as
9770   well as test results to GCC. This [8]amazing group of volunteers is
9771   what makes GCC successful.
9772
9773   For additional information about GCC please refer to the [9]GCC project
9774   web site or contact the [10]GCC development mailing list.
9775
9776   To obtain GCC please use [11]our mirror sites or [12]our version
9777   control system.
9778
9779
9780    For questions related to the use of GCC, please consult these web
9781    pages and the [13]GCC manuals. If that fails, the
9782    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
9783    web pages and the development of GCC are welcome on our developer
9784    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
9785    archives.
9786
9787   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
9788   distribution of this entire article is permitted in any medium,
9789   provided this notice is preserved.
9790
9791   These pages are [18]maintained by the GCC team. Last modified
9792   2021-07-28[19].
9793
9794References
9795
9796   1. http://www.gnu.org/
9797   2. http://gcc.gnu.org/gcc-4.5/changes.html
9798   3. http://gcc.gnu.org/gcc-4.5/changes.html
9799   4. http://gcc.gnu.org/gcc-4.5/changes.html
9800   5. http://gcc.gnu.org/gcc-4.5/changes.html
9801   6. http://gcc.gnu.org/gcc-4.5/changes.html
9802   7. http://gcc.gnu.org/gcc-4.5/buildstat.html
9803   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
9804   9. http://gcc.gnu.org/index.html
9805  10. mailto:gcc@gcc.gnu.org
9806  11. http://gcc.gnu.org/mirrors.html
9807  12. http://gcc.gnu.org/git.html
9808  13. https://gcc.gnu.org/onlinedocs/
9809  14. mailto:gcc-help@gcc.gnu.org
9810  15. mailto:gcc@gcc.gnu.org
9811  16. https://gcc.gnu.org/lists.html
9812  17. https://www.fsf.org/
9813  18. https://gcc.gnu.org/about.html
9814  19. http://validator.w3.org/check/referer
9815======================================================================
9816http://gcc.gnu.org/gcc-4.5/changes.html
9817                             GCC 4.5 Release Series
9818                        Changes, New Features, and Fixes
9819
9820Caveats
9821
9822     * GCC now requires the [1]MPC library in order to build. See the
9823       [2]prerequisites page for version requirements.
9824     * Support for a number of older systems and recently unmaintained or
9825       untested target ports of GCC has been declared obsolete in GCC 4.5.
9826       Unless there is activity to revive them, the next release of GCC
9827       will have their sources permanently removed.
9828       The following ports for individual systems on particular
9829       architectures have been obsoleted:
9830          + IRIX releases before 6.5 (mips-sgi-irix5*,
9831            mips-sgi-irix6.[0-4])
9832          + Solaris 7 (*-*-solaris2.7)
9833          + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*,
9834            alpha-dec-osf5.0*)
9835          + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions
9836            can be found in the [3]announcement.
9837       Support for the classic POWER architecture implemented in the
9838       original RIOS and RIOS2 processors of the old IBM RS/6000 product
9839       line has been obsoleted in the rs6000 port. This does not affect
9840       the new generation Power and PowerPC architectures.
9841     * Support has been removed for all the [4]configurations obsoleted in
9842       GCC 4.4.
9843     * Support has been removed for the protoize and unprotoize utilities,
9844       obsoleted in GCC 4.4.
9845     * Support has been removed for tuning for Itanium1 (Merced) variants.
9846       Note that code tuned for Itanium2 should also run correctly on
9847       Itanium1.
9848     * GCC now generates unwind info also for epilogues. DWARF debuginfo
9849       generated by GCC now uses more features of DWARF3 than before, and
9850       also some DWARF4 features. GDB older than 7.0 is not able to handle
9851       either of these, so to debug GCC 4.5 generated binaries or
9852       libraries GDB 7.0 or later is needed. You can disable use of DWARF4
9853       features with the -gdwarf-3 -gstrict-dwarf options, or use
9854       -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but
9855       epilogue unwind info is emitted unconditionally whenever unwind
9856       info is emitted.
9857     * On x86 targets, code containing floating-point calculations may run
9858       significantly more slowly when compiled with GCC 4.5 in strict C99
9859       conformance mode than they did with earlier GCC versions. This is
9860       due to stricter standard conformance of the compiler and can be
9861       avoided by using the option -fexcess-precision=fast; also see
9862       [5]below.
9863     * The function attribute noinline no longer prevents GCC from cloning
9864       the function. A new attribute noclone has been introduced for this
9865       purpose. Cloning a function means that it is duplicated and the new
9866       copy is specialized for certain contexts (for example when a
9867       parameter is a known constant).
9868
9869General Optimizer Improvements
9870
9871     * The -save-temps now takes an optional argument. The -save-temps and
9872       -save-temps=cwd switches write the temporary files in the current
9873       working directory based on the original source file. The
9874       -save-temps=obj switch will write files into the directory
9875       specified with the -o option, and the intermediate filenames are
9876       based on the output file. This will allow the user to get the
9877       compiler intermediate files when doing parallel builds without two
9878       builds of the same filename located in different directories from
9879       interfering with each other.
9880     * Debugging dumps are now created in the same directory as the object
9881       file rather than in the current working directory. This allows the
9882       user to get debugging dumps when doing parallel builds without two
9883       builds of the same filename interfering with each other.
9884     * GCC has been integrated with the MPC library. This allows GCC to
9885       evaluate complex arithmetic at compile time [6]more accurately. It
9886       also allows GCC to evaluate calls to complex built-in math
9887       functions having constant arguments and replace them at compile
9888       time with their mathematically equivalent results. In doing so, GCC
9889       can generate correct results regardless of the math library
9890       implementation or floating point precision of the host platform.
9891       This also allows GCC to generate identical results regardless of
9892       whether one compiles in native or cross-compile configurations to a
9893       particular target. The following built-in functions take advantage
9894       of this new capability: cacos, cacosh, casin, casinh, catan,
9895       catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan,
9896       and ctanh. The float and long double variants of these functions
9897       (e.g. csinf and csinl) are also handled.
9898     * A new link-time optimizer has been added ([7]-flto). When this
9899       option is used, GCC generates a bytecode representation of each
9900       input file and writes it to specially-named sections in each object
9901       file. When the object files are linked together, all the function
9902       bodies are read from these named sections and instantiated as if
9903       they had been part of the same translation unit. This enables
9904       interprocedural optimizations to work across different files (and
9905       even different languages), potentially improving the performance of
9906       the generated code. To use the link-timer optimizer, -flto needs to
9907       be specified at compile time and during the final link. If the
9908       program does not require any symbols to be exported, it is possible
9909       to combine -flto and the experimental [8]-fwhopr with
9910       [9]-fwhole-program to allow the interprocedural optimizers to use
9911       more aggressive assumptions.
9912     * The automatic parallelization pass was enhanced to support
9913       parallelization of outer loops.
9914     * Automatic parallelization can be enabled as part of Graphite. In
9915       addition to -ftree-parallelize-loops=, specify
9916       -floop-parallelize-all to enable the Graphite-based optimization.
9917     * The infrastructure for optimizing based on [10]restrict qualified
9918       pointers has been rewritten and should result in code generation
9919       improvements. Optimizations based on restrict qualified pointers
9920       are now also available when using -fno-strict-aliasing.
9921     * There is a new optimization pass that attempts to change prototype
9922       of functions to avoid unused parameters, pass only relevant parts
9923       of structures and turn arguments passed by reference to arguments
9924       passed by value when possible. It is enabled by -O2 and above as
9925       well as -Os and can be manually invoked using the new command-line
9926       switch -fipa-sra.
9927     * GCC now optimize exception handling code. In particular cleanup
9928       regions that are proved to not have any effect are optimized out.
9929
9930New Languages and Language specific improvements
9931
9932  All languages
9933
9934     * The -fshow-column option is now on by default. This means error
9935       messages now have a column associated with them.
9936
9937  Ada
9938
9939     * Compilation of programs heavily using discriminated record types
9940       with variant parts has been sped up and generates more compact
9941       code.
9942     * Stack checking now works reasonably well on most plaforms. In some
9943       specific cases, stack overflows may still fail to be detected, but
9944       a compile-time warning will be issued for these cases.
9945
9946  C family
9947
9948     * If a header named in a #include directive is not found, the
9949       compiler exits immediately. This avoids a cascade of errors arising
9950       from declarations expected to be found in that header being
9951       missing.
9952     * A new built-in function __builtin_unreachable() has been added that
9953       tells the compiler that control will never reach that point. It may
9954       be used after asm statements that terminate by transferring control
9955       elsewhere, and in other places that are known to be unreachable.
9956     * The -Wlogical-op option now warns for logical expressions such as
9957       (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be
9958       mistakes. This option is disabled by default.
9959     * An asm goto feature has been added to allow asm statements that
9960       jump to C labels.
9961     * C++0x raw strings are supported for C++ and for C with -std=gnu99.
9962     * The deprecated attribute now takes an optional string argument, for
9963       example, __attribute__((deprecated("text string"))), that will be
9964       printed together with the deprecation warning.
9965
9966  C
9967
9968     * The -Wenum-compare option, which warns when comparing values of
9969       different enum types, now works for C. It formerly only worked for
9970       C++. This warning is enabled by -Wall. It may be avoided by using a
9971       type cast.
9972     * The -Wcast-qual option now warns about casts which are unsafe in
9973       that they permit const-correctness to be violated without further
9974       warnings. Specifically, it warns about cases where a qualifier is
9975       added when all the lower types are not const. For example, it warns
9976       about a cast from char ** to const char **.
9977     * The -Wc++-compat option is significantly improved. It issues new
9978       warnings for:
9979          + Using C++ reserved operator names as identifiers.
9980          + Conversions to enum types without explicit casts.
9981          + Using va_arg with an enum type.
9982          + Using different enum types in the two branches of ?:.
9983          + Using ++ or -- on a variable of enum type.
9984          + Using the same name as both a struct, union or enum tag and a
9985            typedef, unless the typedef refers to the tagged type itself.
9986          + Using a struct, union, or enum which is defined within another
9987            struct or union.
9988          + A struct field defined using a typedef if there is a field in
9989            the struct, or an enclosing struct, whose name is the typedef
9990            name.
9991          + Duplicate definitions at file scope.
9992          + Uninitialized const variables.
9993          + A global variable with an anonymous struct, union, or enum
9994            type.
9995          + Using a string constant to initialize a char array whose size
9996            is the length of the string.
9997     * The new -Wjump-misses-init option warns about cases where a goto or
9998       switch skips the initialization of a variable. This sort of branch
9999       is an error in C++ but not in C. This warning is enabled by
10000       -Wc++-compat.
10001     * GCC now ensures that a C99-conforming <stdint.h> is present on most
10002       targets, and uses information about the types in this header to
10003       implement the Fortran bindings to those types. GCC does not ensure
10004       the presence of such a header, and does not implement the Fortran
10005       bindings, on the following targets: NetBSD, VxWorks, VMS,
10006       SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF.
10007     * GCC now implements C90- and C99-conforming rules for constant
10008       expressions. This may cause warnings or errors for some code using
10009       expressions that can be folded to a constant but are not constant
10010       expressions as defined by ISO C.
10011     * All known target-independent C90 and C90 Amendment 1 conformance
10012       bugs, and all known target-independent C99 conformance bugs not
10013       related to floating point or extended identifiers, have been fixed.
10014     * The C decimal floating point support now includes support for the
10015       FLOAT_CONST_DECIMAL64 pragma.
10016     * The named address space feature from ISO/IEC TR 18037 is now
10017       supported. This is currently only implemented for the SPU
10018       processor.
10019
10020  C++
10021
10022     * Improved [11]experimental support for the upcoming C++0x ISO C++
10023       standard, including support for raw strings, lambda expressions and
10024       explicit type conversion operators.
10025     * When printing the name of a class template specialization, G++ will
10026       now omit any template arguments which come from default template
10027       arguments. This behavior (and the pretty-printing of function
10028       template specializations as template signature and arguments) can
10029       be disabled with the -fno-pretty-templates option.
10030     * Access control is now applied to typedef names used in a template,
10031       which may cause G++ to reject some ill-formed code that was
10032       accepted by earlier releases. The -fno-access-control option can be
10033       used as a temporary workaround until the code is corrected.
10034     * Compilation time for code that uses templates should now scale
10035       linearly with the number of instantiations rather than
10036       quadratically, as template instantiations are now looked up using
10037       hash tables.
10038     * Declarations of functions that look like builtin declarations of
10039       library functions are only considered to be redeclarations if they
10040       are declared with extern "C". This may cause problems with code
10041       that omits extern "C" on hand-written declarations of C library
10042       functions such as abort or memcpy. Such code is ill-formed, but was
10043       accepted by earlier releases.
10044     * Diagnostics that used to complain about passing non-POD types to
10045       ... or jumping past the declaration of a non-POD variable now check
10046       for triviality rather than PODness, as per C++0x.
10047     * In C++0x mode local and anonymous classes are now allowed as
10048       template arguments, and in declarations of variables and functions
10049       with linkage, so long as any such declaration that is used is also
10050       defined ([12]DR 757).
10051     * Labels may now have attributes, as has been permitted for a while
10052       in C. This is only permitted when the label definition and the
10053       attribute specifier is followed by a semicolon—i.e., the label
10054       applies to an empty statement. The only useful attribute for a
10055       label is unused.
10056     * G++ now implements [13]DR 176. Previously G++ did not support using
10057       the injected-class-name of a template base class as a type name,
10058       and lookup of the name found the declaration of the template in the
10059       enclosing scope. Now lookup of the name finds the
10060       injected-class-name, which can be used either as a type or as a
10061       template, depending on whether or not the name is followed by a
10062       template argument list. As a result of this change, some code that
10063       was previously accepted may be ill-formed because
10064         1. The injected-class-name is not accessible because it's from a
10065            private base, or
10066         2. The injected-class-name cannot be used as an argument for a
10067            template template parameter.
10068       In either of these cases, the code can be fixed by adding a
10069       nested-name-specifier to explicitly name the template. The first
10070       can be worked around with -fno-access-control; the second is only
10071       rejected with -pedantic.
10072     * A new standard mangling for SIMD vector types has been added, to
10073       avoid name clashes on systems with vectors of varying length. By
10074       default the compiler still uses the old mangling, but emits aliases
10075       with the new mangling on targets that support strong aliases. Users
10076       can switch over entirely to the new mangling with -fabi-version=4
10077       or -fabi-version=0. -Wabi will now warn about code that uses the
10078       old mangling.
10079     * The command-line option -ftemplate-depth-N is now written as
10080       -ftemplate-depth=N and the old form is deprecated.
10081     * Conversions between NULL and non-pointer types are now warned by
10082       default. The new option -Wno-conversion-null disables these
10083       warnings. Previously these warnings were only available when using
10084       -Wconversion explicitly.
10085
10086    Runtime Library (libstdc++)
10087
10088     * Improved experimental support for the upcoming ISO C++ standard,
10089       C++0x, including:
10090          + Support for <future>, <functional>, and <random>.
10091          + Existing facilities now exploit explicit operators and the
10092            newly implemented core C++0x features.
10093          + The header <cstdatomic> has been renamed to <atomic>.
10094     * An experimental [14]profile mode has been added. This is an
10095       implementation of many C++ standard library constructs with an
10096       additional analysis layer that gives performance improvement advice
10097       based on recognition of suboptimal usage patterns. For example,
10098#include <vector>
10099int main()
10100{
10101  std::vector<int> v;
10102  for (int k = 0; k < 1024; ++k)
10103    v.insert(v.begin(), k);
10104}
10105
10106       When instrumented via the profile mode, can return suggestions
10107       about the initial size and choice of the container used as follows:
10108vector-to-list: improvement = 5: call stack = 0x804842c ...
10109    : advice = change std::vector to std::list
10110vector-size: improvement = 3: call stack = 0x804842c ...
10111    : advice = change initial container size from 0 to 1024
10112
10113       These constructs can be substituted for the normal libstdc++
10114       constructs on a piecemeal basis, or all existing components can be
10115       transformed via the -D_GLIBCXX_PROFILE macro.
10116     * [15]Support for decimal floating-point arithmetic (aka ISO C++ TR
10117       24733) has been added. This support is in header file
10118       <decimal/decimal>, uses namespace std::decimal, and includes
10119       classes decimal32, decimal64, and decimal128.
10120     * Sources have been audited for application of function attributes
10121       nothrow, const, pure, and noreturn.
10122     * Python pretty-printers have been added for many standard library
10123       components that simplify the internal representation and present a
10124       more intuitive view of components when used with
10125       appropriately-advanced versions of GDB. For more information,
10126       please consult the more [16]detailed description.
10127     * The default behavior for comparing typeinfo names has changed, so
10128       in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero.
10129     * The new -static-libstdc++ option directs g++ to link the C++
10130       library statically, even if the default would normally be to link
10131       it dynamically.
10132
10133  Fortran
10134
10135     * The COMMON default padding has been changed – instead of adding the
10136       padding before a variable it is now added afterwards, which
10137       increases the compatibility with other vendors and helps to obtain
10138       the correct output in some cases. Cf. also the -falign-commons
10139       option ([17]added in 4.4).
10140     * The -finit-real= option now also supports the value snan for
10141       signaling not-a-number; to be effective, one additionally needs to
10142       enable trapping (e.g. via -ffpe-trap=). Note: Compile-time
10143       optimizations can turn a signaling NaN into a quiet one.
10144     * The new option -fcheck= has been added with the options bounds,
10145       array-temps, do, pointer, and recursive. The bounds and array-temps
10146       options are equivalent to -fbounds-check and
10147       -fcheck-array-temporaries. The do option checks for invalid
10148       modification of loop iteration variables, and the recursive option
10149       tests for recursive calls to subroutines/functions which are not
10150       marked as recursive. With pointer pointer association checks in
10151       calls are performed; however, neither undefined pointers nor
10152       pointers in expressions are handled. Using -fcheck=all enables all
10153       these run-time checks.
10154     * The run-time checking -fcheck=bounds now warns about invalid string
10155       lengths of character dummy arguments. Additionally, more
10156       compile-time checks have been added.
10157     * The new option [18]-fno-protect-parens has been added; if set, the
10158       compiler may reorder REAL and COMPLEX expressions without regard to
10159       parentheses.
10160     * GNU Fortran no longer links against libgfortranbegin. As before,
10161       MAIN__ (assembler symbol name) is the actual Fortran main program,
10162       which is invoked by the main function. However, main is now
10163       generated and put in the same object file as MAIN__. For the time
10164       being, libgfortranbegin still exists for backward compatibility.
10165       For details see the new [19]Mixed-Language Programming chapter in
10166       the manual.
10167     * The I/O library was restructured for performance and cleaner code.
10168     * Array assignments and WHERE are now run in parallel when OpenMP's
10169       WORKSHARE is used.
10170     * The experimental option -fwhole-file was added. The option allows
10171       whole-file checking of procedure arguments and allows for better
10172       optimizations. It can also be used with -fwhole-program, which is
10173       now also supported in gfortran.
10174     * More Fortran 2003 and Fortran 2008 mathematical functions can now
10175       be used as initialization expressions.
10176     * Some extended attributes such as STDCALL are now supported via the
10177       [20]GCC$ compiler directive.
10178     * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN
10179       intrinsic behaves now as if zero were always positive.
10180     * For legacy compatibiliy: On Cygwin and MinGW, the special files
10181       CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now
10182       supported.
10183     * Fortran 2003 support has been extended:
10184          + Procedure-pointer function results and procedure-pointer
10185            components (including PASS),
10186          + allocatable scalars (experimental),
10187          + DEFERRED type-bound procedures,
10188          + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements
10189            have been implemented.
10190          + The ALLOCATE statement supports type-specs and the SOURCE=
10191            argument.
10192          + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC
10193            type-bound procedure (i.e. as type-bound operators).
10194          + Rounding (ROUND=, RZ, ...) for output is now supported.
10195          + The INT_FAST{8,16,32,64,128}_T kind type parameters of the
10196            intrinsic module ISO_C_BINDING are now supported, except for
10197            the targets listed above as ones where GCC does not have
10198            <stdint.h> type information.
10199          + Extensible derived types with type-bound procedure or
10200            procedure pointer with PASS attribute now have to use CLASS in
10201            line with the Fortran 2003 standard; the workaround to use
10202            TYPE is no longer supported.
10203          + [21]Experimental, incomplete support for polymorphism,
10204            including CLASS, SELECT TYPE and dynamic dispatch of
10205            type-bound procedure calls. Some features do not work yet such
10206            as unlimited polymorphism (CLASS(*)).
10207     * Fortran 2008 support has been extended:
10208          + The OPEN statement now supports the NEWUNIT= option, which
10209            returns a unique file unit, thus preventing inadvertent use of
10210            the same unit in different parts of the program.
10211          + Support for unlimited format items has been added.
10212          + The INT{8,16,32} and REAL{32,64,128} kind type parameters of
10213            the intrinsic module ISO_FORTRAN_ENV are now supported.
10214          + Using complex arguments with TAN, SINH, COSH, TANH, ASIN,
10215            ACOS, and ATAN is now possible; the functions ASINH, ACOSH,
10216            and ATANH have been added (for real and complex arguments) and
10217            ATAN(Y,X) is now an alias for ATAN2(Y,X).
10218          + The BLOCK construct has been implemented.
10219
10220New Targets and Target Specific Improvements
10221
10222  AIX
10223
10224     * Full cross-toolchain support now available with GNU Binutils
10225
10226  ARM
10227
10228     * GCC now supports the Cortex-M0 and Cortex-A5 processors.
10229     * GCC now supports the ARM v7E-M architecture.
10230     * GCC now supports VFPv4-based FPUs and FPUs with
10231       single-precision-only VFP.
10232     * GCC has many improvements to optimization for other ARM processors,
10233       including scheduling support for the integer pipeline on Cortex-A9.
10234     * GCC now supports the IEEE 754-2008 half-precision floating-point
10235       type, and a variant ARM-specific half-precision type. This type is
10236       specified using __fp16, with the layout determined by
10237       -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and
10238       VFPv4 half-precision instructions will be used.
10239     * GCC now supports the variant of AAPCS that uses VFP registers for
10240       parameter passing and return values.
10241
10242  AVR
10243
10244     * The -mno-tablejump option has been removed because it has the same
10245       effect as the -fno-jump-tables option.
10246     * Added support for these new AVR devices:
10247          + ATmega8U2
10248          + ATmega16U2
10249          + ATmega32U2
10250
10251  IA-32/x86-64
10252
10253     * GCC now will set the default for -march= based on the configure
10254       target.
10255     * GCC now supports handling floating-point excess precision arising
10256       from use of the x87 floating-point unit in a way that conforms to
10257       ISO C99. This is enabled with -fexcess-precision=standard and with
10258       standards conformance options such as -std=c99, and may be disabled
10259       using -fexcess-precision=fast.
10260     * Support for the Intel Atom processor is now available through the
10261       -march=atom and -mtune=atom options.
10262     * A new -mcrc32 option is now available to enable crc32 intrinsics.
10263     * A new -mmovbe option is now available to enable GCC to use the
10264       movbe instruction to implement __builtin_bswap32 and
10265       __builtin_bswap64.
10266     * SSE math now can be enabled by default at configure time with the
10267       new --with-fpmath=sse option.
10268     * There is a new intrinsic header file, <x86intrin.h>. It should be
10269       included before using any IA-32/x86-64 intrinsics.
10270     * Support for the XOP, FMA4, and LWP instruction sets for the AMD
10271       Orochi processors are now available with the -mxop, -mfma4, and
10272       -mlwp options.
10273     * The -mabm option enables GCC to use the popcnt and lzcnt
10274       instructions on AMD processors.
10275     * The -mpopcnt option enables GCC to use the popcnt instructions on
10276       both AMD and Intel processors.
10277
10278  M68K/ColdFire
10279
10280     * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x
10281       and 5441x devices.
10282     * GCC now supports thread-local storage (TLS) on M68K and ColdFire
10283       processors.
10284
10285  MeP
10286
10287   Support has been added for the Toshiba Media embedded Processor (MeP,
10288   or mep-elf) embedded target.
10289
10290  MIPS
10291
10292     * GCC now supports MIPS 1004K processors.
10293     * GCC can now be configured with options --with-arch-32,
10294       --with-arch-64, --with-tune-32 and --with-tune-64 to control the
10295       default optimization separately for 32-bit and 64-bit modes.
10296     * MIPS targets now support an alternative _mcount interface, in which
10297       register $12 points to the function's save slot for register $31.
10298       This interface is selected by the -mcount-ra-address option; see
10299       the documentation for more details.
10300     * GNU/Linux targets can now generate read-only .eh_frame sections.
10301       This optimization requires GNU binutils 2.20 or above, and is only
10302       available if GCC is configured with a suitable version of binutils.
10303     * GNU/Linux targets can now attach special relocations to indirect
10304       calls, so that the linker can turn them into direct jumps or
10305       branches. This optimization requires GNU binutils 2.20 or later,
10306       and is automatically selected if GCC is configured with an
10307       appropriate version of binutils. It can be explicitly enabled or
10308       disabled using the -mrelax-pic-calls command-line option.
10309     * GCC now generates more heavily-optimized atomic operations on
10310       Octeon processors.
10311     * MIPS targets now support the -fstack-protector option.
10312     * GCC now supports an -msynci option, which specifies that synci is
10313       enough to flush the instruction cache, without help from the
10314       operating system. GCC uses this information to optimize
10315       automatically-generated cache flush operations, such as those used
10316       for nested functions in C. There is also a --with-synci
10317       configure-time option, which makes -msynci the default.
10318     * GCC supports four new function attributes for interrupt handlers:
10319       interrupt, use_shadow_register_set, keep_interrupts_masked and
10320       use_debug_exception_return. See the documentation for more details
10321       about these attributes.
10322
10323  RS/6000 (POWER/PowerPC)
10324
10325     * GCC now supports the Power ISA 2.06, which includes the VSX
10326       instructions that add vector 64-bit floating point support, new
10327       population count instructions, and conversions between floating
10328       point and unsigned types.
10329     * Support for the power7 processor is now available through the
10330       -mcpu=power7 and -mtune=power7.
10331     * GCC will now vectorize loops that contain simple math functions
10332       like copysign when generating code for altivec or VSX targets.
10333     * Support for the A2 processor is now available through the -mcpu=a2
10334       and -mtune=a2 options.
10335     * Support for the 476 processor is now available through the
10336       -mcpu={476,476fp} and -mtune={476,476fp} options.
10337     * Support for the e500mc64 processor is now available through the
10338       -mcpu=e500mc64 and -mtune=e500mc64 options.
10339     * GCC can now be configured with options --with-cpu-32,
10340       --with-cpu-64, --with-tune-32 and --with-tune-64 to control the
10341       default optimization separately for 32-bit and 64-bit modes.
10342     * Starting with GCC 4.5.4, vectors of type vector long long or vector
10343       long are passed and returned in the same method as other vectors
10344       with the VSX instruction set. Previously the GCC compiler did not
10345       adhere to the ABI for 128-bit vectors with 64-bit integer base
10346       types (PR 48857). This is also fixed in the GCC 4.6.1 release.
10347
10348  RX
10349
10350   Support has been added for the Renesas RX Processor (rx-elf) target.
10351
10352Operating Systems
10353
10354  Windows (Cygwin and MinGW)
10355
10356     * GCC now installs all the major language runtime libraries as DLLs
10357       when configured with the --enable-shared option.
10358     * GCC now makes use of the new support for aligned common variables
10359       in versions of binutils >= 2.20 to fix bugs in the support for SSE
10360       data types.
10361     * Improvements to the libffi support library increase the reliability
10362       of code generated by GCJ on all Windows platforms. Libgcj is
10363       enabled by default for the first time.
10364     * Libtool improvements simplify installation by placing the generated
10365       DLLs in the correct binaries directory.
10366     * Numerous other minor bugfixes and improvements, and substantial
10367       enhancements to the Fortran language support library.
10368
10369   >
10370
10371Other significant improvements
10372
10373  Plugins
10374
10375     * It is now possible to extend the compiler without having to modify
10376       its source code. A new option -fplugin=file.so tells GCC to load
10377       the shared object file.so and execute it as part of the compiler.
10378       The internal documentation describes the details on how plugins can
10379       interact with the compiler.
10380
10381  Installation changes
10382
10383     * The move to newer autotools changed default installation
10384       directories and switches to control them: The --with-datarootdir,
10385       --with-docdir, --with-pdfdir, and --with-htmldir switches are not
10386       used any more. Instead, you can now use --datarootdir, --docdir,
10387       --htmldir, and --pdfdir. The default installation directories have
10388       changed as follows according to the GNU Coding Standards:
10389
10390       datarootdir read-only architecture-independent data root [PREFIX/share]
10391       localedir   locale-specific message catalogs [DATAROOTDIR/locale]
10392       docdir      documentation root [DATAROOTDIR/doc/PACKAGE]
10393       htmldir     html documentation [DOCDIR]
10394       dvidir      dvi documentation [DOCDIR]
10395       pdfdir      pdf documentation [DOCDIR]
10396       psdir       ps documentation [DOCDIR]
10397       The following variables have new default values:
10398
10399       datadir read-only architecture-independent data [DATAROOTDIR]
10400       infodir info documentation [DATAROOTDIR/info]
10401       mandir  man documentation [DATAROOTDIR/man]
10402
10403GCC 4.5.1
10404
10405   This is the [22]list of problem reports (PRs) from GCC's bug tracking
10406   system that are known to be fixed in the 4.5.1 release. This list might
10407   not be complete (that is, it is possible that some PRs that have been
10408   fixed are not listed here).
10409
10410  All languages
10411
10412     * GCC's new link-time optimizer ([23]-flto) now also works on a few
10413       non-ELF targets:
10414          + Cygwin (*-cygwin*)
10415          + MinGW (*-mingw*)
10416          + Darwin on x86-64 (x86_64-apple-darwin*)
10417       LTO is not enabled by default for these targets. To enable LTO, you
10418       should configure with the --enable-lto option.
10419
10420GCC 4.5.2
10421
10422   This is the [24]list of problem reports (PRs) from GCC's bug tracking
10423   system that are known to be fixed in the 4.5.2 release. This list might
10424   not be complete (that is, it is possible that some PRs that have been
10425   fixed are not listed here).
10426
10427GCC 4.5.3
10428
10429   This is the [25]list of problem reports (PRs) from GCC's bug tracking
10430   system that are known to be fixed in the 4.5.3 release. This list might
10431   not be complete (that is, it is possible that some PRs that have been
10432   fixed are not listed here).
10433
10434   On the PowerPC compiler, the Altivec builtin functions vec_ld and
10435   vec_st have been modified to generate the Altivec memory instructions
10436   LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5
10437   release, these builtin functions were changed to generate VSX memory
10438   reference instructions instead of Altivec memory instructions, but
10439   there are differences between the two instructions. If the VSX
10440   instruction set is available, you can now use the new builtin functions
10441   vec_vsx_ld and vec_vsx_st which always generates the VSX memory
10442   instructions.
10443
10444GCC 4.5.4
10445
10446   This is the [26]list of problem reports (PRs) from GCC's bug tracking
10447   system that are known to be fixed in the 4.5.4 release. This list might
10448   not be complete (that is, it is possible that some PRs that have been
10449   fixed are not listed here).
10450
10451
10452    For questions related to the use of GCC, please consult these web
10453    pages and the [27]GCC manuals. If that fails, the
10454    [28]gcc-help@gcc.gnu.org mailing list might help. Comments on these
10455    web pages and the development of GCC are welcome on our developer
10456    list at [29]gcc@gcc.gnu.org. All of [30]our lists have public
10457    archives.
10458
10459   Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and
10460   distribution of this entire article is permitted in any medium,
10461   provided this notice is preserved.
10462
10463   These pages are [32]maintained by the GCC team. Last modified
10464   2022-02-02[33].
10465
10466References
10467
10468   1. https://www.multiprecision.org/mpc/
10469   2. https://gcc.gnu.org/install/prerequisites.html
10470   3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html
10471   4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted
10472   5. http://gcc.gnu.org/gcc-4.5/changes.html#x86
10473   6. https://gcc.gnu.org/PR30789
10474   7. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
10475   8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802
10476   9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800
10477  10. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html
10478  11. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html
10479  12. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757
10480  13. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176
10481  14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html
10482  15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733
10483  16. https://sourceware.org/gdb/wiki/STLSupport
10484  17. http://gcc.gnu.org/gcc-4.4/changes.html
10485  18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
10486  19. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html
10487  20. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html
10488  21. https://gcc.gnu.org/wiki/OOP
10489  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1
10490  23. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
10491  24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2
10492  25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3
10493  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4
10494  27. https://gcc.gnu.org/onlinedocs/
10495  28. mailto:gcc-help@gcc.gnu.org
10496  29. mailto:gcc@gcc.gnu.org
10497  30. https://gcc.gnu.org/lists.html
10498  31. https://www.fsf.org/
10499  32. https://gcc.gnu.org/about.html
10500  33. http://validator.w3.org/check/referer
10501======================================================================
10502http://gcc.gnu.org/gcc-4.4/index.html
10503                             GCC 4.4 Release Series
10504
10505   This release series is no longer maintained.
10506
10507   March 13, 2012
10508
10509   The [1]GNU project and the GCC developers are pleased to announce the
10510   release of GCC 4.4.7.
10511
10512   This release is a bug-fix release, containing fixes for regressions in
10513   GCC 4.4.6 relative to previous releases of GCC.
10514
10515Release History
10516
10517   GCC 4.4.7
10518          March 13, 2012 ([2]changes)
10519
10520   GCC 4.4.6
10521          April 16, 2011 ([3]changes)
10522
10523   GCC 4.4.5
10524          October 1, 2010 ([4]changes)
10525
10526   GCC 4.4.4
10527          April 29, 2010 ([5]changes)
10528
10529   GCC 4.4.3
10530          January 21, 2010 ([6]changes)
10531
10532   GCC 4.4.2
10533          October 15, 2009 ([7]changes)
10534
10535   GCC 4.4.1
10536          July 22, 2009 ([8]changes)
10537
10538   GCC 4.4.0
10539          April 21, 2009 ([9]changes)
10540
10541References and Acknowledgements
10542
10543   GCC used to stand for the GNU C Compiler, but since the compiler
10544   supports several other languages aside from C, it now stands for the
10545   GNU Compiler Collection.
10546
10547   A list of [10]successful builds is updated as new information becomes
10548   available.
10549
10550   The GCC developers would like to thank the numerous people that have
10551   contributed new features, improvements, bug fixes, and other changes as
10552   well as test results to GCC. This [11]amazing group of volunteers is
10553   what makes GCC successful.
10554
10555   For additional information about GCC please refer to the [12]GCC
10556   project web site or contact the [13]GCC development mailing list.
10557
10558   To obtain GCC please use [14]our mirror sites or [15]our version
10559   control system.
10560
10561
10562    For questions related to the use of GCC, please consult these web
10563    pages and the [16]GCC manuals. If that fails, the
10564    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
10565    web pages and the development of GCC are welcome on our developer
10566    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
10567    archives.
10568
10569   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
10570   distribution of this entire article is permitted in any medium,
10571   provided this notice is preserved.
10572
10573   These pages are [21]maintained by the GCC team. Last modified
10574   2021-07-28[22].
10575
10576References
10577
10578   1. http://www.gnu.org/
10579   2. http://gcc.gnu.org/gcc-4.4/changes.html
10580   3. http://gcc.gnu.org/gcc-4.4/changes.html
10581   4. http://gcc.gnu.org/gcc-4.4/changes.html
10582   5. http://gcc.gnu.org/gcc-4.4/changes.html
10583   6. http://gcc.gnu.org/gcc-4.4/changes.html
10584   7. http://gcc.gnu.org/gcc-4.4/changes.html
10585   8. http://gcc.gnu.org/gcc-4.4/changes.html
10586   9. http://gcc.gnu.org/gcc-4.4/changes.html
10587  10. http://gcc.gnu.org/gcc-4.4/buildstat.html
10588  11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
10589  12. http://gcc.gnu.org/index.html
10590  13. mailto:gcc@gcc.gnu.org
10591  14. http://gcc.gnu.org/mirrors.html
10592  15. http://gcc.gnu.org/git.html
10593  16. https://gcc.gnu.org/onlinedocs/
10594  17. mailto:gcc-help@gcc.gnu.org
10595  18. mailto:gcc@gcc.gnu.org
10596  19. https://gcc.gnu.org/lists.html
10597  20. https://www.fsf.org/
10598  21. https://gcc.gnu.org/about.html
10599  22. http://validator.w3.org/check/referer
10600======================================================================
10601http://gcc.gnu.org/gcc-4.4/changes.html
10602                             GCC 4.4 Release Series
10603                        Changes, New Features, and Fixes
10604
10605   The latest release in the 4.4 release series is [1]GCC 4.4.7.
10606
10607Caveats
10608
10609     * __builtin_stdarg_start has been completely removed from GCC.
10610       Support for <varargs.h> had been deprecated since GCC 4.0. Use
10611       __builtin_va_start as a replacement.
10612     * Some of the errors issued by the C++ front end that could be
10613       downgraded to warnings in previous releases by using -fpermissive
10614       are now warnings by default. They can be converted into errors by
10615       using -pedantic-errors.
10616     * Use of the cpp assertion extension will now emit a warning when
10617       -Wdeprecated or -pedantic is used. This extension has been
10618       deprecated for many years, but never warned about.
10619     * Packed bit-fields of type char were not properly bit-packed on many
10620       targets prior to GCC 4.4. On these targets, the fix in GCC 4.4
10621       causes an ABI change. For example there is no longer a 4-bit
10622       padding between field a and b in this structure:
10623    struct foo
10624    {
10625      char a:4;
10626      char b:8;
10627    } __attribute__ ((packed));
10628       There is a new warning to help identify fields that are affected:
10629    foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4
10630       The warning can be disabled with -Wno-packed-bitfield-compat.
10631     * On ARM EABI targets, the C++ mangling of the va_list type has been
10632       changed to conform to the current revision of the EABI. This does
10633       not affect the libstdc++ library included with GCC.
10634     * The SCOUNT and POS bits of the MIPS DSP control register are now
10635       treated as global. Previous versions of GCC treated these fields as
10636       call-clobbered instead.
10637     * The MIPS port no longer recognizes the h asm constraint. It was
10638       necessary to remove this constraint in order to avoid generating
10639       unpredictable code sequences.
10640       One of the main uses of the h constraint was to extract the high
10641       part of a multiplication on 64-bit targets. For example:
10642    asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y));
10643       You can now achieve the same effect using 128-bit types:
10644    typedef unsigned int uint128_t __attribute__((mode(TI)));
10645    result = ((uint128_t) x * y) >> 64;
10646       The second sequence is better in many ways. For example, if x and y
10647       are constants, the compiler can perform the multiplication at
10648       compile time. If x and y are not constants, the compiler can
10649       schedule the runtime multiplication better than it can schedule an
10650       asm statement.
10651     * Support for a number of older systems and recently unmaintained or
10652       untested target ports of GCC has been declared obsolete in GCC 4.4.
10653       Unless there is activity to revive them, the next release of GCC
10654       will have their sources permanently removed.
10655       The following ports for individual systems on particular
10656       architectures have been obsoleted:
10657          + Generic a.out on IA32 and m68k (i[34567]86-*-aout*,
10658            m68k-*-aout*)
10659          + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*,
10660            armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*,
10661            sh-*-*). This does not affect other more specific targets
10662            using the COFF object format on those architectures, or the
10663            more specific H8300 and SH targets (h8300-*-rtems*,
10664            h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*,
10665            sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks).
10666          + 2BSD on PDP-11 (pdp11-*-bsd)
10667          + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*,
10668            powerpc-ibm-aix4.[12]*)
10669          + Tuning support for Itanium1 (Merced) variants. Note that code
10670            tuned for Itanium2 should also run correctly on Itanium1.
10671     * The protoize and unprotoize utilities have been obsoleted and will
10672       be removed in GCC 4.5. These utilities have not been installed by
10673       default since GCC 3.0.
10674     * Support has been removed for all the [2]configurations obsoleted in
10675       GCC 4.3.
10676     * Unknown -Wno-* options are now silently ignored by GCC if no other
10677       diagnostics are issued. If other diagnostics are issued, then GCC
10678       warns about the unknown options.
10679     * More information on porting to GCC 4.4 from previous versions of
10680       GCC can be found in the [3]porting guide for this release.
10681
10682General Optimizer Improvements
10683
10684     * A new command-line switch -findirect-inlining has been added. When
10685       turned on it allows the inliner to also inline indirect calls that
10686       are discovered to have known targets at compile time thanks to
10687       previous inlining.
10688     * A new command-line switch -ftree-switch-conversion has been added.
10689       This new pass turns simple initializations of scalar variables in
10690       switch statements into initializations from a static array, given
10691       that all the values are known at compile time and the ratio between
10692       the new array size and the original switch branches does not exceed
10693       the parameter --param switch-conversion-max-branch-ratio (default
10694       is eight).
10695     * A new command-line switch -ftree-builtin-call-dce has been added.
10696       This optimization eliminates unnecessary calls to certain builtin
10697       functions when the return value is not used, in cases where the
10698       calls can not be eliminated entirely because the function may set
10699       errno. This optimization is on by default at -O2 and above.
10700     * A new command-line switch -fconserve-stack directs the compiler to
10701       minimize stack usage even if it makes the generated code slower.
10702       This affects inlining decisions.
10703     * When the assembler supports it, the compiler will now emit unwind
10704       information using assembler .cfi directives. This makes it possible
10705       to use such directives in inline assembler code. The new option
10706       -fno-dwarf2-cfi-asm directs the compiler to not use .cfi
10707       directives.
10708     * The [4]Graphite branch has been merged. This merge has brought in a
10709       new framework for loop optimizations based on a polyhedral
10710       intermediate representation. These optimizations apply to all the
10711       languages supported by GCC. The following new code transformations
10712       are available in GCC 4.4:
10713          + -floop-interchange performs loop interchange transformations
10714            on loops. Interchanging two nested loops switches the inner
10715            and outer loops. For example, given a loop like:
10716          DO J = 1, M
10717            DO I = 1, N
10718              A(J, I) = A(J, I) * C
10719            ENDDO
10720          ENDDO
10721
10722            loop interchange will transform the loop as if the user had
10723            written:
10724          DO I = 1, N
10725            DO J = 1, M
10726              A(J, I) = A(J, I) * C
10727            ENDDO
10728          ENDDO
10729
10730            which can be beneficial when N is larger than the caches,
10731            because in Fortran, the elements of an array are stored in
10732            memory contiguously by column, and the original loop iterates
10733            over rows, potentially creating at each access a cache miss.
10734          + -floop-strip-mine performs loop strip mining transformations
10735            on loops. Strip mining splits a loop into two nested loops.
10736            The outer loop has strides equal to the strip size and the
10737            inner loop has strides of the original loop within a strip.
10738            For example, given a loop like:
10739          DO I = 1, N
10740            A(I) = A(I) + C
10741          ENDDO
10742
10743            loop strip mining will transform the loop as if the user had
10744            written:
10745          DO II = 1, N, 4
10746            DO I = II, min (II + 3, N)
10747              A(I) = A(I) + C
10748            ENDDO
10749          ENDDO
10750
10751          + -floop-block performs loop blocking transformations on loops.
10752            Blocking strip mines each loop in the loop nest such that the
10753            memory accesses of the element loops fit inside caches. For
10754            example, given a loop like:
10755          DO I = 1, N
10756            DO J = 1, M
10757              A(J, I) = B(I) + C(J)
10758            ENDDO
10759          ENDDO
10760
10761            loop blocking will transform the loop as if the user had
10762            written:
10763          DO II = 1, N, 64
10764            DO JJ = 1, M, 64
10765              DO I = II, min (II + 63, N)
10766                DO J = JJ, min (JJ + 63, M)
10767                  A(J, I) = B(I) + C(J)
10768                ENDDO
10769              ENDDO
10770            ENDDO
10771          ENDDO
10772
10773            which can be beneficial when M is larger than the caches,
10774            because the innermost loop will iterate over a smaller amount
10775            of data that can be kept in the caches.
10776     * A new register allocator has replaced the old one. It is called
10777       integrated register allocator (IRA) because coalescing, register
10778       live range splitting, and hard register preferencing are done
10779       on-the-fly during coloring. It also has better integration with the
10780       reload pass. IRA is a regional register allocator which uses modern
10781       Chaitin-Briggs coloring instead of Chow's priority coloring used in
10782       the old register allocator. More info about IRA internals and
10783       options can be found in the GCC manuals.
10784     * A new instruction scheduler and software pipeliner, based on the
10785       selective scheduling approach, has been added. The new pass
10786       performs instruction unification, register renaming, substitution
10787       through register copies, and speculation during scheduling. The
10788       software pipeliner is able to pipeline non-countable loops. The new
10789       pass is targeted at scheduling-eager in-order platforms. In GCC 4.4
10790       it is available for the Intel Itanium platform working by default
10791       as the second scheduling pass (after register allocation) at the
10792       -O3 optimization level.
10793     * When using -fprofile-generate with a multi-threaded program, the
10794       profile counts may be slightly wrong due to race conditions. The
10795       new -fprofile-correction option directs the compiler to apply
10796       heuristics to smooth out the inconsistencies. By default the
10797       compiler will give an error message when it finds an inconsistent
10798       profile.
10799     * The new -fprofile-dir=PATH option permits setting the directory
10800       where profile data files are stored when using -fprofile-generate
10801       and friends, and the directory used when reading profile data files
10802       using -fprofile-use and friends.
10803
10804New warning options
10805
10806     * The new -Wframe-larger-than=NUMBER option directs GCC to emit a
10807       warning if any stack frame is larger than NUMBER bytes. This may be
10808       used to help ensure that code fits within a limited amount of stack
10809       space.
10810     * The command-line option -Wlarger-than-N is now written as
10811       -Wlarger-than=N and the old form is deprecated.
10812     * The new -Wno-mudflap option disables warnings about constructs
10813       which can not be instrumented when using -fmudflap.
10814
10815New Languages and Language specific improvements
10816
10817     * Version 3.0 of the OpenMP specification is now supported for the C,
10818       C++, and Fortran compilers.
10819     * New character data types, per [5]TR 19769: New character types in
10820       C, are now supported for the C compiler in -std=gnu99 mode, as
10821       __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in
10822       -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too.
10823
10824  C family
10825
10826     * A new optimize attribute was added to allow programmers to change
10827       the optimization level and particular optimization options for an
10828       individual function. You can also change the optimization options
10829       via the GCC optimize pragma for functions defined after the pragma.
10830       The GCC push_options pragma and the GCC pop_options pragma allow
10831       you temporarily save and restore the options used. The GCC
10832       reset_options pragma restores the options to what was specified on
10833       the command line.
10834     * Uninitialized warnings do not require enabling optimization
10835       anymore, that is, -Wuninitialized can be used together with -O0.
10836       Nonetheless, the warnings given by -Wuninitialized will probably be
10837       more accurate if optimization is enabled.
10838     * -Wparentheses now warns about expressions such as (!x | y) and (!x
10839       & y). Using explicit parentheses, such as in ((!x) | y), silences
10840       this warning.
10841     * -Wsequence-point now warns within if, while,do while and for
10842       conditions, and within for begin/end expressions.
10843     * A new option -dU is available to dump definitions of preprocessor
10844       macros that are tested or expanded.
10845
10846  C++
10847
10848     * [6]Improved experimental support for the upcoming ISO C++ standard,
10849       C++0x. Including support for auto, inline namespaces, generalized
10850       initializer lists, defaulted and deleted functions, new character
10851       types, and scoped enums.
10852     * Those errors that may be downgraded to warnings to build legacy
10853       code now mention -fpermissive when -fdiagnostics-show-option is
10854       enabled.
10855     * -Wconversion now warns if the result of a static_cast to enumeral
10856       type is unspecified because the value is outside the range of the
10857       enumeral type.
10858     * -Wuninitialized now warns if a non-static reference or non-static
10859       const member appears in a class without constructors.
10860     * G++ now properly implements value-initialization, so objects with
10861       an initializer of () and an implicitly defined default constructor
10862       will be zero-initialized before the default constructor is called.
10863
10864    Runtime Library (libstdc++)
10865
10866     * Improved experimental support for the upcoming ISO C++ standard,
10867       C++0x, including:
10868          + Support for <chrono>, <condition_variable>, <cstdatomic>,
10869            <forward_list>, <initializer_list>, <mutex>, <ratio>,
10870            <system_error>, and <thread>.
10871          + unique_ptr, <algorithm> additions, exception propagation, and
10872            support for the new character types in <string> and <limits>.
10873          + Existing facilities now exploit initializer lists, defaulted
10874            and deleted functions, and the newly implemented core C++0x
10875            features.
10876          + Some standard containers are more efficient together with
10877            stateful allocators, i.e., no allocator is constructed on the
10878            fly at element construction time.
10879     * Experimental support for non-standard pointer types in containers.
10880     * The long standing libstdc++/30928 has been fixed for targets
10881       running glibc 2.10 or later.
10882     * As usual, many small and larger bug fixes, in particular quite a
10883       few corner cases in <locale>.
10884
10885  Fortran
10886
10887     * GNU Fortran now employs libcpp directly instead of using cc1 as an
10888       external preprocessor. The [7]-cpp option was added to allow manual
10889       invocation of the preprocessor without relying on filename
10890       extensions.
10891     * The [8]-Warray-temporaries option warns about array temporaries
10892       generated by the compiler, as an aid to optimization.
10893     * The [9]-fcheck-array-temporaries option has been added, printing a
10894       notification at run time, when an array temporary had to be created
10895       for an function argument. Contrary to -Warray-temporaries the
10896       warning is only printed if the array is noncontiguous.
10897     * Improved generation of DWARF debugging symbols
10898     * If using an intrinsic not part of the selected standard (via -std=
10899       and -fall-intrinsics) gfortran will now treat it as if this
10900       procedure were declared EXTERNAL and try to link to a user-supplied
10901       procedure. -Wintrinsics-std will warn whenever this happens. The
10902       now-useless option -Wnonstd-intrinsic was removed.
10903     * The flag -falign-commons has been added to control the alignment of
10904       variables in COMMON blocks, which is enabled by default in line
10905       with previous GCC version. Using -fno-align-commons one can force
10906       commons to be contiguous in memory as required by the Fortran
10907       standard, however, this slows down the memory access. The option
10908       -Walign-commons, which is enabled by default, warns when padding
10909       bytes were added for alignment. The proper solution is to sort the
10910       common objects by decreasing storage size, which avoids the
10911       alignment problems.
10912     * Fortran 2003 support has been extended:
10913          + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is
10914            now supported (except internal reads from/writes to wide
10915            strings). [10]-fbackslash now supports also \unnnn and
10916            \Unnnnnnnn to enter Unicode characters.
10917          + Asynchronous I/O (implemented as synchronous I/O) and the
10918            decimal=, size=, sign=, pad=, blank=, and delim= specifiers
10919            are now supported in I/O statements.
10920          + Support for Fortran 2003 structure constructors and for array
10921            constructor with typespec has been added.
10922          + Procedure Pointers (but not yet as component in derived types
10923            and as function results) are now supported.
10924          + Abstract types, type extension, and type-bound procedures
10925            (both PROCEDURE and GENERIC but not as operators). Note: As
10926            CLASS/polymorphyic types are not implemented, type-bound
10927            procedures with PASS accept as non-standard extension TYPE
10928            arguments.
10929     * Fortran 2008 support has been added:
10930          + The -std=f2008 option and support for the file extensions
10931            .f2008 and .F2008 has been added.
10932          + The g0 format descriptor is now supported.
10933          + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH,
10934            ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED
10935            are now available (some of them existed as GNU extension
10936            before). Note: The hyperbolic functions are not yet supporting
10937            complex arguments and the three- argument version of BESSEL_*N
10938            is not available.
10939          + The bit intrinsics LEADZ and TRAILZ have been added.
10940
10941  Java (GCJ)
10942
10943  Ada
10944
10945     * The Ada runtime now supports multilibs on many platforms including
10946       x86_64, SPARC and PowerPC. Their build is enabled by default.
10947
10948New Targets and Target Specific Improvements
10949
10950  ARM
10951
10952     * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and
10953       Cortex-R4F processors and has many other improvements to
10954       optimization for ARM processors.
10955     * GCC now supports the VFPv3 variant with 16 double-precision
10956       registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been
10957       renamed to -mfpu=vfpv3.
10958     * GCC now supports the -mfix-cortex-m3-ldrd option to work around an
10959       erratum on Cortex-M3 processors.
10960     * GCC now supports the __sync_* atomic operations for ARM EABI
10961       GNU/Linux.
10962     * The section anchors optimization is now enabled by default when
10963       optimizing for ARM.
10964     * GCC now uses a new EABI-compatible profiling interface for EABI
10965       targets. This requires a function __gnu_mcount_nc, which is
10966       provided by GNU libc versions 2.8 and later.
10967
10968  AVR
10969
10970     * The -mno-tablejump option has been deprecated because it has the
10971       same effect as the -fno-jump-tables option.
10972     * Added support for these new AVR devices:
10973          + ATA6289
10974          + ATtiny13A
10975          + ATtiny87
10976          + ATtiny167
10977          + ATtiny327
10978          + ATmega8C1
10979          + ATmega16C1
10980          + ATmega32C1
10981          + ATmega8M1
10982          + ATmega16M1
10983          + ATmega32M1
10984          + ATmega32U4
10985          + ATmega16HVB
10986          + ATmega4HVD
10987          + ATmega8HVD
10988          + ATmega64C1
10989          + ATmega64M1
10990          + ATmega16U4
10991          + ATmega32U6
10992          + ATmega128RFA1
10993          + AT90PWM81
10994          + AT90SCR100
10995          + M3000F
10996          + M3000S
10997          + M3001B
10998
10999  IA-32/x86-64
11000
11001     * Support for Intel AES built-in functions and code generation is
11002       available via -maes.
11003     * Support for Intel PCLMUL built-in function and code generation is
11004       available via -mpclmul.
11005     * Support for Intel AVX built-in functions and code generation is
11006       available via -mavx.
11007     * Automatically align the stack for local variables with alignment
11008       requirement.
11009     * GCC can now utilize the SVML library for vectorizing calls to a set
11010       of C99 functions if -mveclibabi=svml is specified and you link to
11011       an SVML ABI compatible library.
11012     * On x86-64, the ABI has been changed in the following cases to
11013       conform to the x86-64 ABI:
11014          + Passing/returning structures with flexible array member:
11015  struct foo
11016    {
11017      int i;
11018      int flex[];
11019    };
11020          + Passing/returning structures with complex float member:
11021  struct foo
11022    {
11023      int i;
11024      __complex__ float f;
11025    };
11026          + Passing/returning unions with long double member:
11027  union foo
11028    {
11029      int x;
11030      long double ld;
11031    };
11032       Code built with previous versions of GCC that uses any of these is
11033       not compatible with code built with GCC 4.4.0 or later.
11034     * A new target attribute was added to allow programmers to change the
11035       target options like -msse2 or -march=k8 for an individual function.
11036       You can also change the target options via the GCC target pragma
11037       for functions defined after the pragma.
11038     * GCC can now be configured with options --with-arch-32,
11039       --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and
11040       --with-tune-64 to control the default optimization separately for
11041       32-bit and 64-bit modes.
11042
11043  IA-32/IA64
11044
11045     * Support for __float128 (TFmode) IEEE quad type and corresponding
11046       TCmode IEEE complex quad type is available via the soft-fp library
11047       on IA-32/IA64 targets. This includes basic arithmetic operations
11048       (addition, subtraction, negation, multiplication and division) on
11049       __float128 real and TCmode complex values, the full set of IEEE
11050       comparisons between __float128 values, conversions to and from
11051       float, double and long double floating point types, as well as
11052       conversions to and from signed or unsigned integer, signed or
11053       unsigned long integer and signed or unsigned quad (TImode, IA64
11054       only) integer types. Additionally, all operations generate the full
11055       set of IEEE exceptions and support the full set of IEEE rounding
11056       modes.
11057
11058  M68K/ColdFire
11059
11060     * GCC now supports instruction scheduling for ColdFire V1, V3 and V4
11061       processors. (Scheduling support for ColdFire V2 processors was
11062       added in GCC 4.3.)
11063     * GCC now supports the -mxgot option to support programs requiring
11064       many GOT entries on ColdFire.
11065     * The m68k-*-linux-gnu target now builds multilibs by default.
11066
11067  MIPS
11068
11069     * MIPS Technologies have extended the original MIPS SVR4 ABI to
11070       include support for procedure linkage tables (PLTs) and copy
11071       relocations. These extensions allow GNU/Linux executables to use a
11072       significantly more efficient code model than the one defined by the
11073       original ABI.
11074       GCC support for this code model is available via a new command-line
11075       option, -mplt. There is also a new configure-time option,
11076       --with-mips-plt, to make -mplt the default.
11077       The new code model requires support from the assembler, the linker,
11078       and the runtime C library. This support is available in binutils
11079       2.19 and GLIBC 2.9.
11080     * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables
11081       and 32-bit GNU/Linux shared libraries. This feature requires GNU
11082       binutils 2.19 or above.
11083     * Support for RMI's XLR processor is now available through the
11084       -march=xlr and -mtune=xlr options.
11085     * 64-bit targets can now perform 128-bit multiplications inline,
11086       instead of relying on a libgcc function.
11087     * Native GNU/Linux toolchains now support -march=native and
11088       -mtune=native, which select the host processor.
11089     * GCC now supports the R10K, R12K, R14K and R16K processors. The
11090       canonical -march= and -mtune= names for these processors are
11091       r10000, r12000, r14000 and r16000 respectively.
11092     * GCC can now work around the side effects of speculative execution
11093       on R10K processors. Please see the documentation of the
11094       -mr10k-cache-barrier option for details.
11095     * Support for the MIPS64 Release 2 instruction set has been added.
11096       The option -march=mips64r2 enables generation of these
11097       instructions.
11098     * GCC now supports Cavium Networks' Octeon processor. This support is
11099       available through the -march=octeon and -mtune=octeon options.
11100     * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The
11101       canonical -march= and -mtune= names for these processors are
11102       loongson2e and loongson2f.
11103
11104  picochip
11105
11106   Picochip is a 16-bit processor. A typical picoChip contains over 250
11107   small cores, each with small amounts of memory. There are three
11108   processor variants (STAN, MEM and CTRL) with different instruction sets
11109   and memory configurations and they can be chosen using the -mae option.
11110
11111   This port is intended to be a "C" only port.
11112
11113  Power Architecture and PowerPC
11114
11115     * GCC now supports the e300c2, e300c3 and e500mc processors.
11116     * GCC now supports Xilinx processors with a single-precision FPU.
11117     * Decimal floating point is now supported for e500 processors.
11118
11119  S/390, zSeries and System z9/z10
11120
11121     * Support for the IBM System z10 EC/BC processor has been added. When
11122       using the -march=z10 option, the compiler will generate code making
11123       use of instructions provided by the General-Instruction-Extension
11124       Facility and the Execute-Extension Facility.
11125
11126  VxWorks
11127
11128     * GCC now supports the thread-local storage mechanism used on
11129       VxWorks.
11130
11131  Xtensa
11132
11133     * GCC now supports thread-local storage (TLS) for Xtensa processor
11134       configurations that include the Thread Pointer option. TLS also
11135       requires support from the assembler and linker; this support is
11136       provided in the GNU binutils beginning with version 2.19.
11137
11138Documentation improvements
11139
11140Other significant improvements
11141
11142GCC 4.4.1
11143
11144   This is the [11]list of problem reports (PRs) from GCC's bug tracking
11145   system that are known to be fixed in the 4.4.1 release. This list might
11146   not be complete (that is, it is possible that some PRs that have been
11147   fixed are not listed here).
11148
11149GCC 4.4.2
11150
11151   This is the [12]list of problem reports (PRs) from GCC's bug tracking
11152   system that are known to be fixed in the 4.4.2 release. This list might
11153   not be complete (that is, it is possible that some PRs that have been
11154   fixed are not listed here).
11155
11156GCC 4.4.3
11157
11158   This is the [13]list of problem reports (PRs) from GCC's bug tracking
11159   system that are known to be fixed in the 4.4.3 release. This list might
11160   not be complete (that is, it is possible that some PRs that have been
11161   fixed are not listed here).
11162
11163GCC 4.4.4
11164
11165   This is the [14]list of problem reports (PRs) from GCC's bug tracking
11166   system that are known to be fixed in the 4.4.4 release. This list might
11167   not be complete (that is, it is possible that some PRs that have been
11168   fixed are not listed here).
11169
11170GCC 4.4.5
11171
11172   This is the [15]list of problem reports (PRs) from GCC's bug tracking
11173   system that are known to be fixed in the 4.4.5 release. This list might
11174   not be complete (that is, it is possible that some PRs that have been
11175   fixed are not listed here).
11176
11177GCC 4.4.6
11178
11179   This is the [16]list of problem reports (PRs) from GCC's bug tracking
11180   system that are known to be fixed in the 4.4.6 release. This list might
11181   not be complete (that is, it is possible that some PRs that have been
11182   fixed are not listed here).
11183
11184GCC 4.4.7
11185
11186   This is the [17]list of problem reports (PRs) from GCC's bug tracking
11187   system that are known to be fixed in the 4.4.7 release. This list might
11188   not be complete (that is, it is possible that some PRs that have been
11189   fixed are not listed here).
11190
11191
11192    For questions related to the use of GCC, please consult these web
11193    pages and the [18]GCC manuals. If that fails, the
11194    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11195    web pages and the development of GCC are welcome on our developer
11196    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
11197    archives.
11198
11199   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
11200   distribution of this entire article is permitted in any medium,
11201   provided this notice is preserved.
11202
11203   These pages are [23]maintained by the GCC team. Last modified
11204   2021-07-28[24].
11205
11206References
11207
11208   1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7
11209   2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted
11210   3. http://gcc.gnu.org/gcc-4.4/porting_to.html
11211   4. https://gcc.gnu.org/wiki/Graphite
11212   5. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf
11213   6. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html
11214   7. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html
11215   8. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125
11216   9. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221
11217  10. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34
11218  11. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1
11219  12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2
11220  13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3
11221  14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4
11222  15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5
11223  16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6
11224  17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7
11225  18. https://gcc.gnu.org/onlinedocs/
11226  19. mailto:gcc-help@gcc.gnu.org
11227  20. mailto:gcc@gcc.gnu.org
11228  21. https://gcc.gnu.org/lists.html
11229  22. https://www.fsf.org/
11230  23. https://gcc.gnu.org/about.html
11231  24. http://validator.w3.org/check/referer
11232======================================================================
11233http://gcc.gnu.org/gcc-4.3/index.html
11234                             GCC 4.3 Release Series
11235
11236   (This release series is no longer supported.)
11237
11238   Jun 27, 2011
11239
11240   The [1]GNU project and the GCC developers are pleased to announce the
11241   release of GCC 4.3.6.
11242
11243   This release is a bug-fix release, containing fixes for regressions in
11244   GCC 4.3.5 relative to previous releases of GCC.
11245
11246Release History
11247
11248   GCC 4.3.6
11249          Jun 27, 2011 ([2]changes)
11250
11251   GCC 4.3.5
11252          May 22, 2010 ([3]changes)
11253
11254   GCC 4.3.4
11255          August 4, 2009 ([4]changes)
11256
11257   GCC 4.3.3
11258          January 24, 2009 ([5]changes)
11259
11260   GCC 4.3.2
11261          August 27, 2008 ([6]changes)
11262
11263   GCC 4.3.1
11264          June 6, 2008 ([7]changes)
11265
11266   GCC 4.3.0
11267          March 5, 2008 ([8]changes)
11268
11269References and Acknowledgements
11270
11271   GCC used to stand for the GNU C Compiler, but since the compiler
11272   supports several other languages aside from C, it now stands for the
11273   GNU Compiler Collection.
11274
11275   A list of [9]successful builds is updated as new information becomes
11276   available.
11277
11278   The GCC developers would like to thank the numerous people that have
11279   contributed new features, improvements, bug fixes, and other changes as
11280   well as test results to GCC. This [10]amazing group of volunteers is
11281   what makes GCC successful.
11282
11283   For additional information about GCC please refer to the [11]GCC
11284   project web site or contact the [12]GCC development mailing list.
11285
11286   To obtain GCC please use [13]our mirror sites or [14]our version
11287   control system.
11288
11289
11290    For questions related to the use of GCC, please consult these web
11291    pages and the [15]GCC manuals. If that fails, the
11292    [16]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11293    web pages and the development of GCC are welcome on our developer
11294    list at [17]gcc@gcc.gnu.org. All of [18]our lists have public
11295    archives.
11296
11297   Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and
11298   distribution of this entire article is permitted in any medium,
11299   provided this notice is preserved.
11300
11301   These pages are [20]maintained by the GCC team. Last modified
11302   2021-07-28[21].
11303
11304References
11305
11306   1. http://www.gnu.org/
11307   2. http://gcc.gnu.org/gcc-4.3/changes.html
11308   3. http://gcc.gnu.org/gcc-4.3/changes.html
11309   4. http://gcc.gnu.org/gcc-4.3/changes.html
11310   5. http://gcc.gnu.org/gcc-4.3/changes.html
11311   6. http://gcc.gnu.org/gcc-4.3/changes.html
11312   7. http://gcc.gnu.org/gcc-4.3/changes.html
11313   8. http://gcc.gnu.org/gcc-4.3/changes.html
11314   9. http://gcc.gnu.org/gcc-4.3/buildstat.html
11315  10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
11316  11. http://gcc.gnu.org/index.html
11317  12. mailto:gcc@gcc.gnu.org
11318  13. http://gcc.gnu.org/mirrors.html
11319  14. http://gcc.gnu.org/git.html
11320  15. https://gcc.gnu.org/onlinedocs/
11321  16. mailto:gcc-help@gcc.gnu.org
11322  17. mailto:gcc@gcc.gnu.org
11323  18. https://gcc.gnu.org/lists.html
11324  19. https://www.fsf.org/
11325  20. https://gcc.gnu.org/about.html
11326  21. http://validator.w3.org/check/referer
11327======================================================================
11328http://gcc.gnu.org/gcc-4.3/changes.html
11329                             GCC 4.3 Release Series
11330                        Changes, New Features, and Fixes
11331
11332   The latest release in the 4.3 release series is [1]GCC 4.3.5.
11333
11334Caveats
11335
11336     * GCC requires the [2]GMP and [3]MPFR libraries for building all the
11337       various front-end languages it supports. See the [4]prerequisites
11338       page for version requirements.
11339     * ColdFire targets now treat long double as having the same format as
11340       double. In earlier versions of GCC, they used the 68881 long double
11341       format instead.
11342     * The m68k-uclinux target now uses the same calling conventions as
11343       m68k-linux-gnu. You can select the original calling conventions by
11344       configuring for m68k-uclinuxoldabi instead. Note that
11345       m68k-uclinuxoldabi also retains the original 80-bit long double on
11346       ColdFire targets.
11347     * The -fforce-mem option has been removed because it has had no
11348       effect in the last few GCC releases.
11349     * The i386 -msvr3-shlib option has been removed since it is no longer
11350       used.
11351     * Fastcall for i386 has been changed not to pass aggregate arguments
11352       in registers, following Microsoft compilers.
11353     * Support for the AOF assembler has been removed from the ARM back
11354       end; this affects only the targets arm-semi-aof and armel-semi-aof,
11355       which are no longer recognized. We removed these targets without a
11356       deprecation period because we discovered that they have been
11357       unusable since GCC 4.0.0.
11358     * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*)
11359       has been removed. This support had been deprecated since GCC 4.0.0.
11360     * Support for a number of older systems and recently unmaintained or
11361       untested target ports of GCC has been declared obsolete in GCC 4.3.
11362       Unless there is activity to revive them, the next release of GCC
11363       will have their sources permanently removed.
11364       All GCC ports for the following processor architectures have been
11365       declared obsolete:
11366          + Morpho MT (mt-*)
11367       The following aliases for processor architectures have been
11368       declared obsolete. Users should use the indicated generic target
11369       names instead, with compile-time options such as -mcpu or
11370       configure-time options such as --with-cpu to control the
11371       configuration more precisely.
11372          + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-*
11373            instead).
11374          + parisc*-*-* (use hppa*-*-* instead).
11375          + m680[012]0-*-* (use m68k-*-* instead).
11376       All GCC ports for the following operating systems have been
11377       declared obsolete:
11378          + BeOS (*-*-beos*)
11379          + kaOS (*-*-kaos*)
11380          + GNU/Linux using the a.out object format (*-*-linux*aout*)
11381          + GNU/Linux using version 1 of the GNU C Library
11382            (*-*-linux*libc1*)
11383          + Solaris versions before Solaris 7 (*-*-solaris2.[0-6],
11384            *-*-solaris2.[0-6].*)
11385          + Miscellaneous System V (*-*-sysv*)
11386          + WindISS (*-*-windiss*)
11387       Also, those for some individual systems on particular architectures
11388       have been obsoleted:
11389          + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*)
11390          + CRIS with a.out object format (cris-*-aout)
11391          + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*)
11392          + OSF/1 on PA-RISC (hppa1.1-*-osf*)
11393          + PRO on PA-RISC (hppa1.1-*-pro*)
11394          + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*,
11395            i[34567]86-sequent-sysv4*)
11396          + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*)
11397          + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host
11398            was previously [5]removed in 2001, leaving only the support
11399            for UWIN as a target now being deprecated)
11400          + ChorusOS on PowerPC (powerpc-*-chorusos*)
11401          + All VAX configurations apart from NetBSD and OpenBSD
11402            (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*)
11403     * The [6]-Wconversion option has been modified. Its purpose now is to
11404       warn for implicit conversions that may alter a value. This new
11405       behavior is available for both C and C++. Warnings about
11406       conversions between signed and unsigned integers can be disabled by
11407       using -Wno-sign-conversion. In C++, they are disabled by default
11408       unless -Wsign-conversion is explicitly requested. The old behavior
11409       of -Wconversion, that is, warn for prototypes causing a type
11410       conversion that is different from what would happen to the same
11411       argument in the absence of a prototype, has been moved to a new
11412       option -Wtraditional-conversion, which is only available for C.
11413     * The -m386, -m486, -mpentium and -mpentiumpro tuning options have
11414       been removed because they were deprecated for more than 3 GCC major
11415       releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or
11416       -mtune=pentiumpro as a replacement.
11417     * The -funsafe-math-optimizations option now automatically turns on
11418       -fno-trapping-math in addition to -fno-signed-zeros, as it enables
11419       reassociation and thus may introduce or remove traps.
11420     * The -ftree-vectorize option is now on by default under -O3. In
11421       order to generate code for a SIMD extension, it has to be enabled
11422       as well: use -maltivec for PowerPC platforms and -msse/-msse2 for
11423       i?86 and x86_64.
11424     * More information on porting to GCC 4.3 from previous versions of
11425       GCC can be found in the [7]porting guide for this release.
11426
11427General Optimizer Improvements
11428
11429     * The GCC middle-end has been integrated with the MPFR library. This
11430       allows GCC to evaluate and replace at compile-time calls to
11431       built-in math functions having constant arguments with their
11432       mathematically equivalent results. In making use of MPFR, GCC can
11433       generate correct results regardless of the math library
11434       implementation or floating point precision of the host platform.
11435       This also allows GCC to generate identical results regardless of
11436       whether one compiles in native or cross-compile configurations to a
11437       particular target. The following built-in functions take advantage
11438       of this new capability: acos, acosh, asin, asinh, atan2, atan,
11439       atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1,
11440       fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10,
11441       log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh,
11442       tan, tanh, tgamma, y0, y1 and yn. The float and long double
11443       variants of these functions (e.g. sinf and sinl) are also handled.
11444       The sqrt and cabs functions with constant arguments were already
11445       optimized in prior GCC releases. Now they also use MPFR.
11446     * A new forward propagation pass on RTL was added. The new pass
11447       replaces several slower transformations, resulting in compile-time
11448       improvements as well as better code generation in some cases.
11449     * A new command-line switch -frecord-gcc-switches has been added to
11450       GCC, although it is only enabled for some targets. The switch
11451       causes the command line that was used to invoke the compiler to be
11452       recorded into the object file that is being created. The exact
11453       format of this recording is target and binary file format
11454       dependent, but it usually takes the form of a note section
11455       containing ASCII text. The switch is related to the -fverbose-asm
11456       switch, but that one only records the information in the assembler
11457       output file as comments, so the information never reaches the
11458       object file.
11459     * The inliner heuristic is now aware of stack frame consumption. New
11460       command-line parameters --param large-stack-frame and --param
11461       large-stack-frame-growth can be used to limit stack frame size
11462       growth caused by inlining.
11463     * During feedback directed optimizations, the expected block size the
11464       memcpy, memset and bzero functions operate on is discovered and for
11465       cases of commonly used small sizes, specialized inline code is
11466       generated.
11467     * __builtin_expect no longer requires its argument to be a compile
11468       time constant.
11469     * Interprocedural optimization was reorganized to work on functions
11470       in SSA form. This enables more precise and cheaper dataflow
11471       analysis and makes writing interprocedural optimizations easier.
11472       The following improvements have been implemented on top of this
11473       framework:
11474          + Pre-inline optimization: Selected local optimization passes
11475            are run before the inliner (and other interprocedural passes)
11476            are executed. This significantly improves the accuracy of code
11477            growth estimates used by the inliner and reduces the overall
11478            memory footprint for large compilation units.
11479          + Early inlining (a simple bottom-up inliner pass inlining only
11480            functions whose body is smaller than the expected call
11481            overhead) is now executed with the early optimization passes,
11482            thus inlining already optimized function bodies into an
11483            unoptimized function that is subsequently optimized by early
11484            optimizers. This enables the compiler to quickly eliminate
11485            abstraction penalty in C++ programs.
11486          + Interprocedural constant propagation now operate on SSA form
11487            increasing accuracy of the analysis.
11488     * A new internal representation for GIMPLE statements has been
11489       contributed, resulting in compile-time memory savings.
11490     * The vectorizer was enhanced to support vectorization of outer
11491       loops, intra-iteration parallelism (loop-aware SLP), vectorization
11492       of strided accesses and loops with multiple data-types. Run-time
11493       dependency testing using loop versioning was added. The cost model,
11494       turned on by -fvect-cost-model, was developed.
11495
11496New Languages and Language specific improvements
11497
11498     * We have added new command-line options
11499       -finstrument-functions-exclude-function-list and
11500       -finstrument-functions-exclude-file-list. They provide more control
11501       over which functions are annotated by the -finstrument-functions
11502       option.
11503
11504  C family
11505
11506     * Implicit conversions between generic vector types are now only
11507       permitted when the two vectors in question have the same number of
11508       elements and compatible element types. (Note that the restriction
11509       involves compatible element types, not implicitly-convertible
11510       element types: thus, a vector type with element type int may not be
11511       implicitly converted to a vector type with element type unsigned
11512       int.) This restriction, which is in line with specifications for
11513       SIMD architectures such as AltiVec, may be relaxed using the flag
11514       -flax-vector-conversions. This flag is intended only as a
11515       compatibility measure and should not be used for new code.
11516     * -Warray-bounds has been added and is now enabled by default for
11517       -Wall . It produces warnings for array subscripts that can be
11518       determined at compile time to be always out of bounds.
11519       -Wno-array-bounds will disable the warning.
11520     * The constructor and destructor function attributes now accept
11521       optional priority arguments which control the order in which the
11522       constructor and destructor functions are run.
11523     * New [8]command-line options -Wtype-limits, -Wold-style-declaration,
11524       -Wmissing-parameter-type, -Wempty-body, -Wclobbered and
11525       -Wignored-qualifiers have been added for finer control of the
11526       diverse warnings enabled by -Wextra.
11527     * A new function attribute alloc_size has been added to mark up
11528       malloc style functions. For constant sized allocations this can be
11529       used to find out the size of the returned pointer using the
11530       __builtin_object_size() function for buffer overflow checking and
11531       similar. This supplements the already built-in malloc and calloc
11532       constant size handling.
11533     * Integer constants written in binary are now supported as a GCC
11534       extension. They consist of a prefix 0b or 0B, followed by a
11535       sequence of 0 and 1 digits.
11536     * A new predefined macro __COUNTER__ has been added. It expands to
11537       sequential integral values starting from 0. In conjunction with the
11538       ## operator, this provides a convenient means to generate unique
11539       identifiers.
11540     * A new command-line option -fdirectives-only has been added. It
11541       enables a special preprocessing mode which improves the performance
11542       of applications like distcc and ccache.
11543     * Fixed-point data types and operators have been added. They are
11544       based on Chapter 4 of the Embedded-C specification (n1169.pdf).
11545       Currently, only MIPS targets are supported.
11546     * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732,
11547       N1241, is now supported as a GCC extension to C for targets
11548       i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu,
11549       and x86_64-*-linux-gnu. The feature introduces new data types
11550       _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF,
11551       DD, and DL.
11552
11553  C++
11554
11555     * [9]Experimental support for the upcoming ISO C++ standard, C++0x.
11556     * -Wc++0x-compat has been added and is now enabled by default for
11557       -Wall. It produces warnings for constructs whose meaning differs
11558       between ISO C++ 1998 and C++0x.
11559     * The -Wparentheses option now works for C++ as it does for C. It
11560       warns if parentheses are omitted when operators with confusing
11561       precedence are nested. It also warns about ambiguous else
11562       statements. Since -Wparentheses is enabled by -Wall, this may cause
11563       additional warnings with existing C++ code which uses -Wall. These
11564       new warnings may be disabled by using -Wall -Wno-parentheses.
11565     * The -Wmissing-declarations now works for C++ as it does for C.
11566     * The -fvisibility-ms-compat flag was added, to make it easier to
11567       port larger projects using shared libraries from Microsoft's Visual
11568       Studio to ELF and Mach-O systems.
11569     * C++ attribute handling has been overhauled for template arguments
11570       (ie dependent types). In particular, __attribute__((aligned(T)));
11571       works for C++ types.
11572
11573    Runtime Library (libstdc++)
11574
11575     * [10]Experimental support for the upcoming ISO C++ standard, C++0x.
11576     * Support for TR1 mathematical special functions and regular
11577       expressions.
11578     * Default what implementations give more elaborate exception strings
11579       for bad_cast, bad_typeid, bad_exception, and bad_alloc.
11580     * Header dependencies have been streamlined, reducing unnecessary
11581       includes and pre-processed bloat.
11582     * Variadic template implementations of items in <tuple> and
11583       <functional>.
11584     * An experimental [11]parallel mode has been added. This is a
11585       parallel implementation of many C++ Standard library algorithms,
11586       like std::accumulate, std::for_each, std::transform, or std::sort,
11587       to give but four examples. These algorithms can be substituted for
11588       the normal (sequential) libstdc++ algorithms on a piecemeal basis,
11589       or all existing algorithms can be transformed via the
11590       -D_GLIBCXX_PARALLEL macro.
11591     * Debug mode versions of classes in <unordered_set> and
11592       <unordered_map>.
11593     * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are
11594       now <backward/hash_set> and <backward/hash_map>. This code:
11595    #include <ext/hash_set>
11596    __gnu_cxx::hash_set<int> s;
11597
11598       Can be transformed (in order of preference) to:
11599    #include <tr1/unordered_set>
11600    std::tr1::unordered_set<int> s;
11601
11602       or
11603    #include <backward/hash_set>
11604    __gnu_cxx::hash_set<int> s;
11605
11606       Similar transformations apply to __gnu_cxx::hash_map,
11607       __gnu_cxx::hash_multimap, __gnu_cxx::hash_set,
11608       __gnu_cxx::hash_multiset.
11609
11610  Fortran
11611
11612     * Due to the fact that the GMP and MPFR libraries are required for
11613       all languages, Fortran is no longer special in this regard and is
11614       available by default.
11615     * The [12]-fexternal-blas option has been added, which generates
11616       calls to BLAS routines for intrinsic matrix operations such as
11617       matmul rather than using the built-in algorithms.
11618     * Support to give a backtrace (compiler flag -fbacktrace or
11619       environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems
11620       only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a
11621       run-time error occured.
11622     * GNU Fortran now defines __GFORTRAN__ when it runs the C
11623       preprocessor (CPP).
11624     * The [13]-finit-local-zero, -finit-real, -finit-integer,
11625       -finit-character, and -finit-logical options have been added, which
11626       can be used to initialize local variables.
11627     * The intrinsic procedures [14]GAMMA and [15]LGAMMA have been added,
11628       which calculate the Gamma function and its logarithm. Use EXTERNAL
11629       gamma if you want to use your own gamma function.
11630     * GNU Fortran now regards the backslash character as literal (as
11631       required by the Fortran 2003 standard); using [16]-fbackslash GNU
11632       Fortran interprets backslashes as C-style escape characters.
11633     * The [17]interpretation of binary, octal and hexadecimal (BOZ)
11634       literal constants has been changed. Before they were always
11635       interpreted as integer; now they are bit-wise transferred as
11636       argument of INT, REAL, DBLE and CMPLX as required by the Fortran
11637       2003 standard, and for real and complex variables in DATA
11638       statements or when directly assigned to real and complex variables.
11639       Everywhere else and especially in expressions they are still
11640       regarded as integer constants.
11641     * Fortran 2003 support has been extended:
11642          + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE
11643          + Pointer intent
11644          + Intrinsic module ISO_ENV_FORTRAN
11645          + Interoperability with C (ISO C Bindings)
11646          + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER
11647            attribute)
11648          + Fortran 2003 BOZ
11649
11650  Java (GCJ)
11651
11652     * GCJ now uses the Eclipse Java compiler for its Java parsing needs.
11653       This enables the use of all 1.5 language features, and fixes most
11654       existing front end bugs.
11655     * libgcj now supports all 1.5 language features which require runtime
11656       support: foreach, enum, annotations, generics, and auto-boxing.
11657     * We've made many changes to the tools shipped with gcj.
11658          + The old jv-scan tool has been removed. This tool never really
11659            worked properly. There is no replacement.
11660          + gcjh has been rewritten. Some of its more obscure options no
11661            longer work, but are still recognized in an attempt at
11662            compatibility. gjavah is a new program with similar
11663            functionality but different command-line options.
11664          + grmic and grmiregistry have been rewritten. grmid has been
11665            added.
11666          + gjar replaces the old fastjar.
11667          + gjarsigner (used for signing jars), gkeytool (used for key
11668            management), gorbd (for CORBA), gserialver (computes
11669            serialization UIDs), and gtnameserv (also for CORBA) are now
11670            installed.
11671     * The ability to dump the contents of the java run time heap to a
11672       file for off-line analysis has been added. The heap dumps may be
11673       analyzed with the new gc-analyze tool. They may be generated on
11674       out-of-memory conditions or on demand and are controlled by the new
11675       run time class gnu.gcj.util.GCInfo.
11676     * java.util.TimeZone can now read files from /usr/share/zoneinfo to
11677       provide correct, updated, timezone information. This means that
11678       packagers no longer have to update libgcj when a time zone change
11679       is published.
11680
11681New Targets and Target Specific Improvements
11682
11683  IA-32/x86-64
11684
11685     * Tuning for Intel Core 2 processors is available via -mtune=core2
11686       and -march=core2.
11687     * Tuning for AMD Geode processors is available via -mtune=geode and
11688       -march=geode.
11689     * Code generation of block move (memcpy) and block set (memset) was
11690       rewritten. GCC can now pick the best algorithm (loop, unrolled
11691       loop, instruction with rep prefix or a library call) based on the
11692       size of the block being copied and the CPU being optimized for. A
11693       new option -minline-stringops-dynamically has been added. With this
11694       option string operations of unknown size are expanded such that
11695       small blocks are copied by in-line code, while for large blocks a
11696       library call is used. This results in faster code than
11697       -minline-all-stringops when the library implementation is capable
11698       of using cache hierarchy hints. The heuristic choosing the
11699       particular algorithm can be overwritten via -mstringop-strategy.
11700       Newly also memset of values different from 0 is inlined.
11701     * GCC no longer places the cld instruction before string operations.
11702       Both i386 and x86-64 ABI documents mandate the direction flag to be
11703       clear at the entry of a function. It is now invalid to set the flag
11704       in asm statement without reseting it afterward.
11705     * Support for SSSE3 built-in functions and code generation are
11706       available via -mssse3.
11707     * Support for SSE4.1 built-in functions and code generation are
11708       available via -msse4.1.
11709     * Support for SSE4.2 built-in functions and code generation are
11710       available via -msse4.2.
11711     * Both SSE4.1 and SSE4.2 support can be enabled via -msse4.
11712     * A new set of options -mpc32, -mpc64 and -mpc80 have been added to
11713       allow explicit control of x87 floating point precision.
11714     * Support for __float128 (TFmode) IEEE quad type and corresponding
11715       TCmode IEEE complex quad type is available via the soft-fp library
11716       on x86_64 targets. This includes basic arithmetic operations
11717       (addition, subtraction, negation, multiplication and division) on
11718       __float128 real and TCmode complex values, the full set of IEEE
11719       comparisons between __float128 values, conversions to and from
11720       float, double and long double floating point types, as well as
11721       conversions to and from signed or unsigned integer, signed or
11722       unsigned long integer and signed or unsigned quad (TImode) integer
11723       types. Additionally, all operations generate the full set of IEEE
11724       exceptions and support the full set of IEEE rounding modes.
11725     * GCC can now utilize the ACML library for vectorizing calls to a set
11726       of C99 functions on x86_64 if -mveclibabi=acml is specified and you
11727       link to an ACML ABI compatible library.
11728
11729  ARM
11730
11731     * Compiler and Library support for Thumb-2 and the ARMv7 architecture
11732       has been added.
11733
11734  CRIS
11735
11736    New features
11737
11738     * Compiler and Library support for the CRIS v32 architecture, as
11739       found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been
11740       added.
11741
11742    Configuration changes
11743
11744     * The cris-*-elf target now includes support for CRIS v32, including
11745       libraries, through the -march=v32 option.
11746     * A new crisv32-*-elf target defaults to generate code for CRIS v32.
11747     * A new crisv32-*-linux* target defaults to generate code for CRIS
11748       v32.
11749     * The cris-*-aout target has been obsoleted.
11750
11751    Improved support for built-in functions
11752
11753     * GCC can now use the lz and swapwbr instructions to implement the
11754       __builtin_clz, __builtin_ctz and __builtin_ffs family of functions.
11755     * __builtin_bswap32 is now implemented using the swapwb instruction,
11756       when available.
11757
11758  m68k and ColdFire
11759
11760    New features
11761
11762     * Support for several new ColdFire processors has been added. You can
11763       generate code for them using the new -mcpu option.
11764     * All targets now support ColdFire processors.
11765     * m68k-uclinux targets have improved support for C++ constructors and
11766       destructors, and for shared libraries.
11767     * It is now possible to set breakpoints on the first or last line of
11768       a function, even if there are no statements on that line.
11769
11770    Optimizations
11771
11772     * Support for sibling calls has been added.
11773     * More use is now made of the ColdFire mov3q instruction.
11774     * __builtin_clz is now implemented using the ff1 ColdFire
11775       instruction, when available.
11776     * GCC now honors the -m68010 option. 68010 code now uses clr rather
11777       than move to zero volatile memory.
11778     * 68020 targets and above can now use symbol(index.size*scale)
11779       addresses for indexed array accesses. Earlier compilers would
11780       always load the symbol into a base register first.
11781
11782    Configuration changes
11783
11784     * All m68k and ColdFire targets now allow the default processor to be
11785       set at configure time using --with-cpu.
11786     * A --with-arch configuration option has been added. This option
11787       allows you to restrict a target to ColdFire or non-ColdFire
11788       processors.
11789
11790    Preprocessor macros
11791
11792     * An __mcfv*__ macro is now defined for all ColdFire targets.
11793       (Earlier versions of GCC only defined __mcfv4e__.)
11794     * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added.
11795     * All targets now define __mc68010 and __mc68010__ when generating
11796       68010 code.
11797
11798    Command-line changes
11799
11800     * New command-line options -march, -mcpu, -mtune and -mhard-float
11801       have been added. These options apply to both m68k and ColdFire
11802       targets.
11803     * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative
11804       versions of -mshort, etc.
11805     * -fforce-addr has been removed. It is now ignored by the compiler.
11806
11807    Other improvements
11808
11809     * ColdFire targets now try to maintain a 4-byte-aligned stack where
11810       possible.
11811     * m68k-uclinux targets now try to avoid situations that lead to the
11812       load-time error: BINFMT_FLAT: reloc outside program.
11813
11814  MIPS
11815
11816    Changes to existing configurations
11817
11818     * libffi and libjava now support all three GNU/Linux ABIs: o32, n32
11819       and n64. Every GNU/Linux configuration now builds these libraries
11820       by default.
11821     * GNU/Linux configurations now generate -mno-shared code unless
11822       overridden by -fpic, -fPIC, -fpie or -fPIE.
11823     * mipsisa32*-linux-gnu configurations now generate hard-float code by
11824       default, just like other mipsisa32* and mips*-linux-gnu
11825       configurations. You can build a soft-float version of any
11826       mips*-linux-gnu configuration by passing --with-float=soft to
11827       configure.
11828     * mips-wrs-vxworks now supports run-time processes (RTPs).
11829
11830    Changes to existing command-line options
11831
11832     * The -march and -mtune options no longer accept 24k as a processor
11833       name. Please use 24kc, 24kf2_1 or 24kf1_1 instead.
11834     * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and
11835       34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The
11836       options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for
11837       24kx, 24kex and 34kx.
11838
11839    New configurations
11840
11841   GCC now supports the following configurations:
11842     * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by
11843       default. Earlier releases also recognized this configuration, but
11844       they treated it in the same way as mipsisa32*-linux-gnu*. Note that
11845       you can customize any mips*-linux-gnu* configuration to a
11846       particular ISA or processor by passing an appropriate --with-arch
11847       option to configure.
11848     * mipsisa*-sde-elf*, which provides compatibility with MIPS
11849       Technologies' SDE toolchains. The configuration uses the SDE
11850       libraries by default, but you can use it like other newlib-based
11851       ELF configurations by passing --with-newlib to configure. It is the
11852       only configuration besides mips64vr*-elf* to build MIPS16 as well
11853       as non-MIPS16 libraries.
11854     * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf*
11855       configuration, but uses the o32 and o64 ABIs instead of the 32-bit
11856       and 64-bit forms of the EABI.
11857
11858    New processors and application-specific extensions
11859
11860     * Support for the SmartMIPS ASE is available through the new
11861       -msmartmips option.
11862     * Support for revision 2 of the DSP ASE is available through the new
11863       -mdspr2 option. A new preprocessor macro called __mips_dsp_rev
11864       indicates the revision of the ASE in use.
11865     * Support for the 4KS and 74K families of processors is available
11866       through the -march and -mtune options.
11867
11868    Improved support for built-in functions
11869
11870     * GCC can now use load-linked, store-conditional and sync
11871       instructions to implement atomic built-in functions such as
11872       __sync_fetch_and_add. The memory reference must be 4 bytes wide for
11873       32-bit targets and either 4 or 8 bytes wide for 64-bit targets.
11874     * GCC can now use the clz and dclz instructions to implement the
11875       __builtin_ctz and __builtin_ffs families of functions.
11876     * There is a new __builtin___clear_cache function for flushing the
11877       instruction cache. GCC expands this function inline on MIPS32
11878       revision 2 targets, otherwise it calls the function specified by
11879       -mcache-flush-func.
11880
11881    MIPS16 improvements
11882
11883     * GCC can now compile objects that contain a mixture of MIPS16 and
11884       non-MIPS16 code. There are two new attributes, mips16 and nomips16,
11885       for specifying which mode a function should use.
11886     * A new option called -minterlink-mips16 makes non-MIPS16 code
11887       link-compatible with MIPS16 code.
11888     * After many bug fixes, the long-standing MIPS16 -mhard-float support
11889       should now work fairly reliably.
11890     * GCC can now use the MIPS16e save and restore instructions.
11891     * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled
11892       with -G0 -fsection-anchors is often smaller than code compiled with
11893       -G8. However, please note that you must usually compile all objects
11894       in your application with the same -G option; see the documentation
11895       of -G for details.
11896     * A new option called-mcode-readable specifies which instructions are
11897       allowed to load from the code segment. -mcode-readable=yes is the
11898       default and says that any instruction may load from the code
11899       segment. The other alternatives are -mcode-readable=pcrel, which
11900       says that only PC-relative MIPS16 instructions may load from the
11901       code segment, and -mcode-readable=no, which says that no
11902       instruction may do so. Please see the documentation for more
11903       details, including example uses.
11904
11905    Small-data improvements
11906
11907   There are three new options for controlling small data:
11908     * -mno-extern-sdata, which disables small-data accesses for
11909       externally-defined variables. Code compiled with -Gn
11910       -mno-extern-sdata will be link-compatible with any -G setting
11911       between -G0 and -Gn inclusive.
11912     * -mno-local-sdata, which disables the use of small-data sections for
11913       data that is not externally visible. This option can be a useful
11914       way of reducing small-data usage in less performance-critical parts
11915       of an application.
11916     * -mno-gpopt, which disables the use of the $gp register while still
11917       honoring the -G limit when placing externally-visible data. This
11918       option implies -mno-extern-sdata and -mno-local-sdata and it can be
11919       useful in situations where $gp does not necessarily hold the
11920       expected value.
11921
11922    Miscellaneous improvements
11923
11924     * There is a new option called -mbranch-cost for tweaking the
11925       perceived cost of branches.
11926     * If GCC is configured to use a version of GAS that supports the
11927       .gnu_attribute directive, it will use that directive to record
11928       certain properties of the output code. .gnu_attribute is new to GAS
11929       2.18.
11930     * There are two new function attributes, near and far, for overriding
11931       the command-line setting of -mlong-calls on a function-by-function
11932       basis.
11933     * -mfp64, which previously required a 64-bit target, now works with
11934       MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and
11935       mipsisa*-sde-elf* configurations provide suitable library support.
11936     * GCC now recognizes the -mdmx and -mmt options and passes them down
11937       to the assembler. It does nothing else with the options at present.
11938
11939  SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture
11940  (BEA)
11941
11942     * Support has been added for this new architecture.
11943
11944  RS6000 (POWER/PowerPC)
11945
11946     * Support for the PowerPC 750CL paired-single instructions has been
11947       added with a new powerpc-*-linux*paired* target configuration. It
11948       is enabled by an associated -mpaired option and can be accessed
11949       using new built-in functions.
11950     * Support for auto-detecting architecture and system configuration to
11951       auto-select processor optimization tuning.
11952     * Support for VMX on AIX 5.3 has been added.
11953     * Support for AIX Version 6.1 has been added.
11954
11955  S/390, zSeries and System z9
11956
11957     * Support for the IBM System z9 EC/BC processor (z9 GA3) has been
11958       added. When using the -march=z9-ec option, the compiler will
11959       generate code making use of instructions provided by the decimal
11960       floating point facility and the floating point conversion facility
11961       (pfpo). Besides the instructions used to implement decimal floating
11962       point operations these facilities also contain instructions to move
11963       between general purpose and floating point registers and to modify
11964       and copy the sign-bit of floating point values.
11965     * When the -march=z9-ec option is used the new
11966       -mhard-dfp/-mno-hard-dfp options can be used to specify whether the
11967       decimal floating point hardware instructions will be used or not.
11968       If none of them is given the hardware support is enabled by
11969       default.
11970     * The -mstack-guard option can now be omitted when using stack
11971       checking via -mstack-size in order to let GCC choose a sensible
11972       stack guard value according to the frame size of each function.
11973     * Various changes to improve performance of generated code have been
11974       implemented, including:
11975          + The condition code set by an add logical with carry
11976            instruction is now available for overflow checks like: a + b +
11977            carry < b.
11978          + The test data class instruction is now used to implement
11979            sign-bit and infinity checks of binary and decimal floating
11980            point numbers.
11981
11982  SPARC
11983
11984     * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been
11985       added.
11986
11987  Xtensa
11988
11989     * Stack unwinding for exception handling now uses by default a
11990       specialized version of DWARF unwinding. This is not
11991       binary-compatible with the setjmp/longjmp (sjlj) unwinding used for
11992       Xtensa with previous versions of GCC.
11993     * For Xtensa processors that include the Conditional Store option,
11994       the built-in functions for atomic memory access are now implemented
11995       using S32C1I instructions.
11996     * If the Xtensa NSA option is available, GCC will use it to implement
11997       the __builtin_ctz and __builtin_clz functions.
11998
11999Documentation improvements
12000
12001     * Existing libstdc++ documentation has been edited and restructured
12002       into a single DocBook XML manual. The results can be viewed online
12003       [18]here.
12004
12005Other significant improvements
12006
12007     * The compiler's --help command-line option has been extended so that
12008       it now takes an optional set of arguments. These arguments restrict
12009       the information displayed to specific classes of command-line
12010       options, and possibly only a subset of those options. It is also
12011       now possible to replace the descriptive text associated with each
12012       displayed option with an indication of its current value, or for
12013       binary options, whether it has been enabled or disabled.
12014       Here are some examples. The following will display all the options
12015       controlling warning messages:
12016      --help=warnings
12017
12018       Whereas this will display all the undocumented, target specific
12019       options:
12020      --help=target,undocumented
12021
12022       This sequence of commands will display the binary optimizations
12023       that are enabled by -O3:
12024      gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
12025      gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
12026      diff /tmp/O2-opts /tmp/O3-opts | grep enabled
12027
12028     * The configure options --with-pkgversion and --with-bugurl have been
12029       added. These allow distributors of GCC to include a
12030       distributor-specific string in manuals and --version output and to
12031       specify the URL for reporting bugs in their versions of GCC.
12032
12033GCC 4.3.1
12034
12035   This is the [19]list of problem reports (PRs) from GCC's bug tracking
12036   system that are known to be fixed in the 4.3.1 release. This list might
12037   not be complete (that is, it is possible that some PRs that have been
12038   fixed are not listed here).
12039
12040Target Specific Changes
12041
12042  IA-32/x86-64
12043
12044    ABI changes
12045
12046     * Starting with GCC 4.3.1, decimal floating point variables are
12047       aligned to their natural boundaries when they are passed on the
12048       stack for i386.
12049
12050    Command-line changes
12051
12052     * Starting with GCC 4.3.1, the -mcld option has been added to
12053       automatically generate a cld instruction in the prologue of
12054       functions that use string instructions. This option is used for
12055       backward compatibility on some operating systems and can be enabled
12056       by default for 32-bit x86 targets by configuring GCC with the
12057       --enable-cld configure option.
12058
12059GCC 4.3.2
12060
12061   This is the [20]list of problem reports (PRs) from GCC's bug tracking
12062   system that are known to be fixed in the 4.3.2 release. This list might
12063   not be complete (that is, it is possible that some PRs that have been
12064   fixed are not listed here).
12065
12066GCC 4.3.3
12067
12068   This is the [21]list of problem reports (PRs) from GCC's bug tracking
12069   system that are known to be fixed in the 4.3.3 release. This list might
12070   not be complete (that is, it is possible that some PRs that have been
12071   fixed are not listed here).
12072
12073GCC 4.3.4
12074
12075   This is the [22]list of problem reports (PRs) from GCC's bug tracking
12076   system that are known to be fixed in the 4.3.4 release. This list might
12077   not be complete (that is, it is possible that some PRs that have been
12078   fixed are not listed here).
12079
12080GCC 4.3.5
12081
12082   This is the [23]list of problem reports (PRs) from GCC's bug tracking
12083   system that are known to be fixed in the 4.3.5 release. This list might
12084   not be complete (that is, it is possible that some PRs that have been
12085   fixed are not listed here).
12086
12087GCC 4.3.6
12088
12089   This is the [24]list of problem reports (PRs) from GCC's bug tracking
12090   system that are known to be fixed in the 4.3.6 release. This list might
12091   not be complete (that is, it is possible that some PRs that have been
12092   fixed are not listed here).
12093
12094
12095    For questions related to the use of GCC, please consult these web
12096    pages and the [25]GCC manuals. If that fails, the
12097    [26]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12098    web pages and the development of GCC are welcome on our developer
12099    list at [27]gcc@gcc.gnu.org. All of [28]our lists have public
12100    archives.
12101
12102   Copyright (C) [29]Free Software Foundation, Inc. Verbatim copying and
12103   distribution of this entire article is permitted in any medium,
12104   provided this notice is preserved.
12105
12106   These pages are [30]maintained by the GCC team. Last modified
12107   2021-07-28[31].
12108
12109References
12110
12111   1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5
12112   2. https://gmplib.org/
12113   3. https://www.mpfr.org/
12114   4. https://gcc.gnu.org/install/prerequisites.html
12115   5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html
12116   6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
12117   7. http://gcc.gnu.org/gcc-4.3/porting_to.html
12118   8. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
12119   9. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
12120  10. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
12121  11. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html
12122  12. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options
12123  13. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167
12124  14. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html
12125  15. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html
12126  16. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html
12127  17. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html
12128  18. https://gcc.gnu.org/onlinedocs/libstdc++/
12129  19. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1
12130  20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2
12131  21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3
12132  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4
12133  23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5
12134  24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6
12135  25. https://gcc.gnu.org/onlinedocs/
12136  26. mailto:gcc-help@gcc.gnu.org
12137  27. mailto:gcc@gcc.gnu.org
12138  28. https://gcc.gnu.org/lists.html
12139  29. https://www.fsf.org/
12140  30. https://gcc.gnu.org/about.html
12141  31. http://validator.w3.org/check/referer
12142======================================================================
12143http://gcc.gnu.org/gcc-4.2/index.html
12144                             GCC 4.2 Release Series
12145
12146   (This release series is no longer supported.)
12147
12148   May 19, 2008
12149
12150   The [1]GNU project and the GCC developers are pleased to announce the
12151   release of GCC 4.2.4.
12152
12153   This release is a bug-fix release, containing fixes for regressions in
12154   GCC 4.2.3 relative to previous releases of GCC.
12155
12156Release History
12157
12158   GCC 4.2.4
12159          May 19, 2008 ([2]changes)
12160
12161   GCC 4.2.3
12162          February 1, 2008 ([3]changes)
12163
12164   GCC 4.2.2
12165          October 7, 2007 ([4]changes)
12166
12167   GCC 4.2.1
12168          July 18, 2007 ([5]changes)
12169
12170   GCC 4.2.0
12171          May 13, 2007 ([6]changes)
12172
12173References and Acknowledgements
12174
12175   GCC used to stand for the GNU C Compiler, but since the compiler
12176   supports several other languages aside from C, it now stands for the
12177   GNU Compiler Collection.
12178
12179   A list of [7]successful builds is updated as new information becomes
12180   available.
12181
12182   The GCC developers would like to thank the numerous people that have
12183   contributed new features, improvements, bug fixes, and other changes as
12184   well as test results to GCC. This [8]amazing group of volunteers is
12185   what makes GCC successful.
12186
12187   For additional information about GCC please refer to the [9]GCC project
12188   web site or contact the [10]GCC development mailing list.
12189
12190   To obtain GCC please use [11]our mirror sites or [12]our version
12191   control system.
12192
12193
12194    For questions related to the use of GCC, please consult these web
12195    pages and the [13]GCC manuals. If that fails, the
12196    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12197    web pages and the development of GCC are welcome on our developer
12198    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
12199    archives.
12200
12201   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
12202   distribution of this entire article is permitted in any medium,
12203   provided this notice is preserved.
12204
12205   These pages are [18]maintained by the GCC team. Last modified
12206   2021-07-28[19].
12207
12208References
12209
12210   1. http://www.gnu.org/
12211   2. http://gcc.gnu.org/gcc-4.2/changes.html
12212   3. http://gcc.gnu.org/gcc-4.2/changes.html
12213   4. http://gcc.gnu.org/gcc-4.2/changes.html
12214   5. http://gcc.gnu.org/gcc-4.2/changes.html
12215   6. http://gcc.gnu.org/gcc-4.2/changes.html
12216   7. http://gcc.gnu.org/gcc-4.2/buildstat.html
12217   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12218   9. http://gcc.gnu.org/index.html
12219  10. mailto:gcc@gcc.gnu.org
12220  11. http://gcc.gnu.org/mirrors.html
12221  12. http://gcc.gnu.org/git.html
12222  13. https://gcc.gnu.org/onlinedocs/
12223  14. mailto:gcc-help@gcc.gnu.org
12224  15. mailto:gcc@gcc.gnu.org
12225  16. https://gcc.gnu.org/lists.html
12226  17. https://www.fsf.org/
12227  18. https://gcc.gnu.org/about.html
12228  19. http://validator.w3.org/check/referer
12229======================================================================
12230http://gcc.gnu.org/gcc-4.2/changes.html
12231                             GCC 4.2 Release Series
12232                        Changes, New Features, and Fixes
12233
12234Caveats
12235
12236     * GCC no longer accepts the -fshared-data option. This option has had
12237       no effect in any GCC 4 release; the targets to which the option
12238       used to apply had been removed before GCC 4.0.
12239
12240General Optimizer Improvements
12241
12242     * New command-line options specify the possible relationships among
12243       parameters and between parameters and global data. For example,
12244       -fargument-noalias-anything specifies that arguments do not alias
12245       any other storage.
12246       Each language will automatically use whatever option is required by
12247       the language standard. You should not need to use these options
12248       yourself.
12249
12250New Languages and Language specific improvements
12251
12252     * [1]OpenMP is now supported for the C, C++ and Fortran compilers.
12253     * New command-line options -fstrict-overflow and -Wstrict-overflow
12254       have been added. -fstrict-overflow tells the compiler that it may
12255       assume that the program follows the strict signed overflow
12256       semantics permitted for the language: for C and C++ this means that
12257       the compiler may assume that signed overflow does not occur. For
12258       example, a loop like
12259      for (i = 1; i > 0; i *= 2)
12260
12261       is presumably intended to continue looping until i overflows. With
12262       -fstrict-overflow, the compiler may assume that signed overflow
12263       will not occur, and transform this into an infinite loop.
12264       -fstrict-overflow is turned on by default at -O2, and may be
12265       disabled via -fno-strict-overflow. The -Wstrict-overflow option may
12266       be used to warn about cases where the compiler assumes that signed
12267       overflow will not occur. It takes five different levels:
12268       -Wstrict-overflow=1 to 5. See the [2]documentation for details.
12269       -Wstrict-overflow=1 is enabled by -Wall.
12270     * The new command-line option -fno-toplevel-reorder directs GCC to
12271       emit top-level functions, variables, and asm statements in the same
12272       order that they appear in the input file. This is intended to
12273       support existing code which relies on a particular ordering (for
12274       example, code which uses top-level asm statements to switch
12275       sections). For new code, it is generally better to use function and
12276       variable attributes. The -fno-toplevel-reorder option may be used
12277       for most cases which currently use -fno-unit-at-a-time. The
12278       -fno-unit-at-a-time option will be removed in some future version
12279       of GCC. If you know of a case which requires -fno-unit-at-a-time
12280       which is not fixed by -fno-toplevel-reorder, please open a bug
12281       report.
12282
12283  C family
12284
12285     * The pragma redefine_extname will now macro expand its tokens for
12286       compatibility with SunPRO.
12287     * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct
12288       GCC to handle inline functions as specified in the C99 standard. In
12289       preparation for this, GCC 4.2 will warn about any use of non-static
12290       inline functions in gnu99 or c99 mode. This new warning may be
12291       disabled with the new gnu_inline function attribute or the new
12292       -fgnu89-inline command-line option. Also, GCC 4.2 and later will
12293       define one of the preprocessor macros __GNUC_GNU_INLINE__ or
12294       __GNUC_STDC_INLINE__ to indicate the semantics of inline functions
12295       in the current compilation.
12296     * A new command-line option -Waddress has been added to warn about
12297       suspicious uses of memory addresses as, for example, using the
12298       address of a function in a conditional expression, and comparisons
12299       against the memory address of a string literal. This warning is
12300       enabled by -Wall.
12301
12302  C++
12303
12304     * C++ visibility handling has been overhauled.
12305       Restricted visiblity is propagated from classes to members, from
12306       functions to local statics, and from templates and template
12307       arguments to instantiations, unless the latter has explicitly
12308       declared visibility.
12309       The visibility attribute for a class must come between the
12310       class-key and the name, not after the closing brace.
12311       Attributes are now allowed for enums and elaborated-type-specifiers
12312       that only declare a type.
12313       Members of the anonymous namespace are now local to a particular
12314       translation unit, along with any other declarations which use them,
12315       though they are still treated as having external linkage for
12316       language semantics.
12317     * The (undocumented) extension which permitted templates with default
12318       arguments to be bound to template template parameters with fewer
12319       parameters has been removed. For example:
12320        template <template <typename> class C>
12321        void f(C<double>) {}
12322
12323        template <typename T, typename U = int>
12324        struct S {};
12325
12326        template void f(S<double>);
12327
12328       is no longer accepted by G++. The reason this code is not accepted
12329       is that S is a template with two parameters; therefore, it cannot
12330       be bound to C which has only one parameter.
12331     * The <?, >?, <?=, and >?= operators, deprecated in previous GCC
12332       releases, have been removed.
12333     * The command-line option -fconst-strings, deprecated in previous GCC
12334       releases, has been removed.
12335     * The configure variable enable-__cxa_atexit is now enabled by
12336       default for more targets. Enabling this variable is necessary in
12337       order for static destructors to be executed in the correct order,
12338       but it depends upon the presence of a non-standard C library in the
12339       target library in order to work. The variable is now enabled for
12340       more targets which are known to have suitable C libraries.
12341     * -Wextra will produce warnings for if statements with a semicolon as
12342       the only body, to catch code like:
12343         if (a);
12344            return 1;
12345         return 0;
12346
12347       To suppress the warning in valid cases, use { } instead.
12348     * The C++ front end now also produces strict aliasing warnings when
12349       -fstrict-aliasing -Wstrict-aliasing is in effect.
12350
12351    Runtime Library (libstdc++)
12352
12353     * Added support for TR1 <random>, <complex>, and C compatibility
12354       headers. In addition, a lock-free version of shared_ptr was
12355       contributed as part of Phillip Jordan's Google Summer of Code
12356       project on lock-free containers.
12357     * In association with the Summer of Code work on lock-free
12358       containers, the interface for atomic builtins was adjusted,
12359       creating simpler alternatives for non-threaded code paths. Also,
12360       usage was consolidated and all elements were moved from namespace
12361       std to namespace__gnu_cxx. Affected interfaces are the functions
12362       __exchange_and_add, __atomic_add, and the objects __mutex,
12363       __recursive_mutex, and __scoped_lock.
12364     * Support for versioning weak symbol names via namespace association
12365       was added. However, as this changes the names of exported symbols,
12366       this is turned off by default in the current ABI. Intrepid users
12367       can enable this feature by using
12368       --enable-symvers=gnu-versioned-namespace during configuration.
12369     * Revised, simplified, and expanded policy-based associative
12370       containers, including data types for tree and trie forms
12371       (basic_tree, tree, trie), lists (list_update), and both
12372       collision-chaining and probing hash-based containers
12373       (basic_hash_table, cc_hash_table, gp_hash_table). More details per
12374       the [3]documentation.
12375     * The implementation of the debug mode was modified, whereby the
12376       debug namespaces were nested inside of namespace std and namespace
12377       __gnu_cxx in order to resolve some long standing corner cases
12378       involving name lookup. Debug functionality from the policy-based
12379       data structures was consolidated and enabled with the single macro,
12380       _GLIBCXX_DEBUG. See PR 26142 for more information.
12381     * Added extensions for type traits: __conditional_type,
12382       __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if.
12383     * Added a typelist implementation for compile-time meta-programming.
12384       Elements for typelist construction and operation can be found
12385       within namespace __gnu_cxx::typelist.
12386     * Added a new allocator, __gnu_cxx::throw_allocator, for testing
12387       exception-safety.
12388     * Enabled library-wide visibility control, allowing -fvisibility to
12389       be used.
12390     * Consolidated all nested namespaces and the conversion of
12391       __gnu_internal implementation-private details to anonymous
12392       namespaces whenever possible.
12393     * Implemented LWG resolutions DR 431 and DR 538.
12394
12395  Fortran
12396
12397     * Support for allocatable components has been added (TR 15581 and
12398       Fortran 2003).
12399     * Support for the Fortran 2003 streaming IO extension has been added.
12400     * The GNU Fortran compiler now uses 4-byte record markers by default
12401       for unformatted files to be compatible with g77 and most other
12402       compilers. The implementation allows for records greater than 2 GB
12403       and is compatible with several other compilers. Older versions of
12404       gfortran used 8-byte record markers by default (on most systems).
12405       In order to change the length of the record markers, e.g. to read
12406       unformatted files created by older gfortran versions, the
12407       [4]-frecord-marker=8 option can be used.
12408
12409  Java (GCJ)
12410
12411     * A new command-line option -static-libgcj has been added for targets
12412       that use a linker compatible with GNU Binutils. As its name
12413       implies, this causes libgcj to be linked statically. In some cases
12414       this causes the resulting executable to start faster and use less
12415       memory than if the shared version of libgcj were used. However
12416       caution should be used as it can also cause essential parts of the
12417       library to be omitted. Some of these issues are discussed in:
12418       [5]https://gcc.gnu.org/wiki/Statically_linking_libgcj
12419     * fastjar is no longer bundled with GCC. To build libgcj, you will
12420       need either InfoZIP (both zip and unzip) or an external jar
12421       program. In the former case, the GCC build will install a jar shell
12422       script that is based on InfoZIP and provides the same functionality
12423       as fastjar.
12424
12425New Targets and Target Specific Improvements
12426
12427  IA-32/x86-64
12428
12429     * -mtune=generic can now be used to generate code running well on
12430       common x86 chips. This includes AMD Athlon, AMD Opteron, Intel
12431       Pentium-M, Intel Pentium 4 and Intel Core 2.
12432     * -mtune=native and -march=native will produce code optimized for the
12433       host architecture as detected using the cpuid instruction.
12434     * Added a new command-line option -fstackrealign and and
12435       __attribute__ ((force_align_arg_pointer)) to realign the stack at
12436       runtime. This allows functions compiled with a vector-aligned stack
12437       to be invoked from legacy objects that keep only word-alignment.
12438
12439  SPARC
12440
12441     * The default CPU setting has been changed from V7 to V9 in 32-bit
12442       mode on Solaris 7 and above. This is already the case in 64-bit
12443       mode. It can be overridden by specifying --with-cpu at configure
12444       time.
12445     * Back-end support of built-in functions for atomic memory access has
12446       been implemented.
12447     * Support for the Sun UltraSPARC T1 (Niagara) processor has been
12448       added.
12449
12450  M32C
12451
12452     * Various bug fixes have made some functions (notably, functions
12453       returning structures) incompatible with previous releases.
12454       Recompiling all libraries is recommended. Note that code quality
12455       has considerably improved since 4.1, making a recompile even more
12456       beneficial.
12457
12458  MIPS
12459
12460     * Added support for the Broadcom SB-1A core.
12461
12462  IA-64
12463
12464     * Added support for IA-64 data and control speculation. By default
12465       speculation is enabled only during second scheduler pass. A number
12466       of machine flags was introduced to control the usage of speculation
12467       for both scheduler passes.
12468
12469  HPPA
12470
12471     * Added Java language support (libffi and libjava) for 32-bit HP-UX
12472       11 target.
12473
12474Obsolete Systems
12475
12476Documentation improvements
12477
12478  PDF Documentation
12479
12480     * A make pdf target has been added to the top-level makefile,
12481       enabling automated production of PDF documentation files.
12482       (Front-ends external to GCC should modify their Make-lang.in file
12483       to add a lang.pdf: target.)
12484
12485Other significant improvements
12486
12487  Build system improvements
12488
12489     * All the components of the compiler are now bootstrapped by default.
12490       This improves the resilience to bugs in the system compiler or
12491       binary compatibility problems, as well as providing better testing
12492       of GCC 4.2 itself. In addition, if you build the compiler from a
12493       combined tree, the assembler, linker, etc. will also be
12494       bootstrapped (i.e. built with themselves).
12495       You can disable this behavior, and go back to the pre-GCC 4.2 set
12496       up, by configuring GCC with --disable-bootstrap.
12497     * The rules that configure follows to find target tools resemble more
12498       closely the locations that the built compiler will search. In
12499       addition, you can use the new configure option --with-target-tools
12500       to specify where to find the target tools used during the build,
12501       without affecting what the built compiler will use.
12502       This can be especially useful when building packages of GCC. For
12503       example, you may want to build GCC with GNU as or ld, even if the
12504       resulting compiler to work with the native assembler and linker. To
12505       do so, you can use --with-target-tools to point to the native
12506       tools.
12507
12508  Incompatible changes to the build system
12509
12510     * Front-ends external to GCC should modify their Make-lang.in file to
12511       replace double-colon rules (e.g. dvi::) with normal rules (like
12512       lang.dvi:). Front-end makefile hooks do not use double-colon rules
12513       anymore.
12514     * Up to GCC 4.1, a popular way to specify the target tools used
12515       during the build was to create directories named gas, binutils,
12516       etc. in the build tree, and create links to the tools from there.
12517       This does not work any more when the compiler is bootstrapped. The
12518       new configure option --with-target-tools provides a better way to
12519       achieve the same effect, and works for all native and cross
12520       settings.
12521
12522
12523    For questions related to the use of GCC, please consult these web
12524    pages and the [6]GCC manuals. If that fails, the
12525    [7]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12526    web pages and the development of GCC are welcome on our developer
12527    list at [8]gcc@gcc.gnu.org. All of [9]our lists have public archives.
12528
12529   Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and
12530   distribution of this entire article is permitted in any medium,
12531   provided this notice is preserved.
12532
12533   These pages are [11]maintained by the GCC team. Last modified
12534   2021-07-28[12].
12535
12536References
12537
12538   1. http://gcc.gnu.org/projects/gomp/
12539   2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
12540   3. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html
12541   4. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html
12542   5. https://gcc.gnu.org/wiki/Statically_linking_libgcj
12543   6. https://gcc.gnu.org/onlinedocs/
12544   7. mailto:gcc-help@gcc.gnu.org
12545   8. mailto:gcc@gcc.gnu.org
12546   9. https://gcc.gnu.org/lists.html
12547  10. https://www.fsf.org/
12548  11. https://gcc.gnu.org/about.html
12549  12. http://validator.w3.org/check/referer
12550======================================================================
12551http://gcc.gnu.org/gcc-4.1/index.html
12552                             GCC 4.1 Release Series
12553
12554   (This release series is no longer supported.)
12555
12556   February 13, 2007
12557
12558   The [1]GNU project and the GCC developers are pleased to announce the
12559   release of GCC 4.1.2.
12560
12561   This release is a bug-fix release, containing fixes for regressions in
12562   GCC 4.1.1 relative to previous releases of GCC.
12563
12564Release History
12565
12566   GCC 4.1.2
12567          February 13, 2007 ([2]changes)
12568
12569   GCC 4.1.1
12570          May 24, 2006 ([3]changes)
12571
12572   GCC 4.1.0
12573          February 28, 2006 ([4]changes)
12574
12575References and Acknowledgements
12576
12577   GCC used to stand for the GNU C Compiler, but since the compiler
12578   supports several other languages aside from C, it now stands for the
12579   GNU Compiler Collection.
12580
12581   A list of [5]successful builds is updated as new information becomes
12582   available.
12583
12584   The GCC developers would like to thank the numerous people that have
12585   contributed new features, improvements, bug fixes, and other changes as
12586   well as test results to GCC. This [6]amazing group of volunteers is
12587   what makes GCC successful.
12588
12589   For additional information about GCC please refer to the [7]GCC project
12590   web site or contact the [8]GCC development mailing list.
12591
12592   To obtain GCC please use [9]our mirror sites or [10]our version control
12593   system.
12594
12595
12596    For questions related to the use of GCC, please consult these web
12597    pages and the [11]GCC manuals. If that fails, the
12598    [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12599    web pages and the development of GCC are welcome on our developer
12600    list at [13]gcc@gcc.gnu.org. All of [14]our lists have public
12601    archives.
12602
12603   Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and
12604   distribution of this entire article is permitted in any medium,
12605   provided this notice is preserved.
12606
12607   These pages are [16]maintained by the GCC team. Last modified
12608   2021-07-28[17].
12609
12610References
12611
12612   1. http://www.gnu.org/
12613   2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
12614   3. http://gcc.gnu.org/gcc-4.1/changes.html
12615   4. http://gcc.gnu.org/gcc-4.1/changes.html
12616   5. http://gcc.gnu.org/gcc-4.1/buildstat.html
12617   6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12618   7. http://gcc.gnu.org/index.html
12619   8. mailto:gcc@gcc.gnu.org
12620   9. http://gcc.gnu.org/mirrors.html
12621  10. http://gcc.gnu.org/git.html
12622  11. https://gcc.gnu.org/onlinedocs/
12623  12. mailto:gcc-help@gcc.gnu.org
12624  13. mailto:gcc@gcc.gnu.org
12625  14. https://gcc.gnu.org/lists.html
12626  15. https://www.fsf.org/
12627  16. https://gcc.gnu.org/about.html
12628  17. http://validator.w3.org/check/referer
12629======================================================================
12630http://gcc.gnu.org/gcc-4.1/changes.html
12631                             GCC 4.1 Release Series
12632                        Changes, New Features, and Fixes
12633
12634   The latest release in the 4.1 release series is [1]GCC 4.1.2.
12635
12636Caveats
12637
12638General Optimizer Improvements
12639
12640     * GCC now has infrastructure for inter-procedural optimizations and
12641       the following inter-procedural optimizations are implemented:
12642          + Profile guided inlining. When doing profile feedback guided
12643            optimization, GCC can now use the profile to make better
12644            informed decisions on whether inlining of a function is
12645            profitable or not. This means that GCC will no longer inline
12646            functions at call sites that are not executed very often, and
12647            that functions at hot call sites are more likely to be
12648            inlined.
12649            A new parameter min-inline-recursive-probability is also now
12650            available to throttle recursive inlining of functions with
12651            small average recursive depths.
12652          + Discovery of pure and const functions, a form of side-effects
12653            analysis. While older GCC releases could also discover such
12654            special functions, the new IPA-based pass runs earlier so that
12655            the results are available to more optimizers. The pass is also
12656            simply more powerful than the old one.
12657          + Analysis of references to static variables and type escape
12658            analysis, also forms of side-effects analysis. The results of
12659            these passes allow the compiler to be less conservative about
12660            call-clobbered variables and references. This results in more
12661            redundant loads being eliminated and in making static
12662            variables candidates for register promotion.
12663          + Improvement of RTL-based alias analysis. The results of type
12664            escape analysis are fed to the RTL type-based alias analyzer,
12665            allowing it to disambiguate more memory references.
12666          + Interprocedural constant propagation and function versioning.
12667            This pass looks for functions that are always called with the
12668            same constant value for one or more of the function arguments,
12669            and propagates those constants into those functions.
12670          + GCC will now eliminate static variables whose usage was
12671            optimized out.
12672          + -fwhole-program --combine can now be used to make all
12673            functions in program static allowing whole program
12674            optimization. As an exception, the main function and all
12675            functions marked with the new externally_visible attribute are
12676            kept global so that programs can link with runtime libraries.
12677     * GCC can now do a form of partial dead code elimination (PDCE) that
12678       allows code motion of expressions to the paths where the result of
12679       the expression is actually needed. This is not always a win, so the
12680       pass has been limited to only consider profitable cases. Here is an
12681       example:
12682    int foo (int *, int *);
12683    int
12684    bar (int d)
12685    {
12686      int a, b, c;
12687      b = d + 1;
12688      c = d + 2;
12689      a = b + c;
12690      if (d)
12691        {
12692          foo (&b, &c);
12693          a = b + c;
12694        }
12695      printf ("%d\n", a);
12696    }
12697
12698       The a = b + c can be sunk to right before the printf. Normal code
12699       sinking will not do this, it will sink the first one above into the
12700       else-branch of the conditional jump, which still gives you two
12701       copies of the code.
12702     * GCC now has a value range propagation pass. This allows the
12703       compiler to eliminate bounds checks and branches. The results of
12704       the pass can also be used to accurately compute branch
12705       probabilities.
12706     * The pass to convert PHI nodes to straight-line code (a form of
12707       if-conversion for GIMPLE) has been improved significantly. The two
12708       most significant improvements are an improved algorithm to
12709       determine the order in which the PHI nodes are considered, and an
12710       improvement that allow the pass to consider if-conversions of basic
12711       blocks with more than two predecessors.
12712     * Alias analysis improvements. GCC can now differentiate between
12713       different fields of structures in Tree-SSA's virtual operands form.
12714       This lets stores/loads from non-overlapping structure fields not
12715       conflict. A new algorithm to compute points-to sets was contributed
12716       that can allows GCC to see now that p->a and p->b, where p is a
12717       pointer to a structure, can never point to the same field.
12718     * Various enhancements to auto-vectorization:
12719          + Incrementally preserve SSA form when vectorizing.
12720          + Incrementally preserve loop-closed form when vectorizing.
12721          + Improvements to peeling for alignment: generate better code
12722            when the misalignment of an access is known at compile time,
12723            or when different accesses are known to have the same
12724            misalignment, even if the misalignment amount itself is
12725            unknown.
12726          + Consider dependence distance in the vectorizer.
12727          + Externalize generic parts of data reference analysis to make
12728            this analysis available to other passes.
12729          + Vectorization of conditional code.
12730          + Reduction support.
12731     * GCC can now partition functions in sections of hot and cold code.
12732       This can significantly improve performance due to better
12733       instruction cache locality. This feature works best together with
12734       profile feedback driven optimization.
12735     * A new pass to avoid saving of unneeded arguments to the stack in
12736       vararg functions if the compiler can prove that they will not be
12737       needed.
12738     * Transition of basic block profiling to tree level implementation
12739       has been completed. The new implementation should be considerably
12740       more reliable (hopefully avoiding profile mismatch errors when
12741       using -fprofile-use or -fbranch-probabilities) and can be used to
12742       drive higher level optimizations, such as inlining.
12743       The -ftree-based-profiling command-line option was removed and
12744       -fprofile-use now implies disabling old RTL level loop optimizer
12745       (-fno-loop-optimize). Speculative prefetching optimization
12746       (originally enabled by -fspeculative-prefetching) was removed.
12747
12748New Languages and Language specific improvements
12749
12750  C and Objective-C
12751
12752     * The old Bison-based C and Objective-C parser has been replaced by a
12753       new, faster hand-written recursive-descent parser.
12754
12755  Ada
12756
12757     * The build infrastructure for the Ada runtime library and tools has
12758       been changed to be better integrated with the rest of the build
12759       infrastructure of GCC. This should make doing cross builds of Ada a
12760       bit easier.
12761
12762  C++
12763
12764     * ARM-style name-injection of friend declarations is no longer the
12765       default. For example:
12766          struct S {
12767            friend void f();
12768          };
12769
12770          void g() { f(); }
12771       will not be accepted; instead a declaration of f will need to be
12772       present outside of the scope of S. The new -ffriend-injection
12773       option will enable the old behavior.
12774     * The (undocumented) extension which permitted templates with default
12775       arguments to be bound to template template parameters with fewer
12776       parameters has been deprecated, and will be removed in the next
12777       major release of G++. For example:
12778       template <template <typename> class C>
12779       void f(C<double>) {}
12780
12781       template <typename T, typename U = int>
12782       struct S {};
12783
12784       template void f(S<double>);
12785
12786       makes use of the deprecated extension. The reason this code is not
12787       valid ISO C++ is that S is a template with two parameters;
12788       therefore, it cannot be bound to C which has only one parameter.
12789
12790    Runtime Library (libstdc++)
12791
12792     * Optimization work:
12793          + A new implementation of std::search_n is provided, better
12794            performing in case of random access iterators.
12795          + Added further efficient specializations of istream functions,
12796            i.e., character array and string extractors.
12797          + Other smaller improvements throughout.
12798     * Policy-based associative containers, designed for high-performance,
12799       flexibility and semantic safety are delivered in ext/pb_assoc.
12800     * A versatile string class, __gnu_cxx::__versa_string, providing
12801       facilities conforming to the standard requirements for
12802       basic_string, is delivered in <ext/vstring.h>. In particular:
12803          + Two base classes are provided: the default one avoids
12804            reference counting and is optimized for short strings; the
12805            alternate one, still uses it while improving in a few low
12806            level areas (e.g., alignment). See vstring_fwd.h for some
12807            useful typedefs.
12808          + Various algorithms have been rewritten (e.g., replace), the
12809            code streamlined and simple optimizations added.
12810          + Option 3 of DR 431 is implemented for both available bases,
12811            thus improving the support for stateful allocators.
12812     * As usual, many bugs have been fixed (e.g., libstdc++/13583,
12813       libstdc++/23953) and LWG resolutions put into effect for the first
12814       time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1
12815       Issue 6.19). The implementation status of TR1 is now tracked in the
12816       docs in tr1.html.
12817
12818  Objective-C++
12819
12820     * A new language front end for Objective-C++ has been added. This
12821       language allows users to mix the object oriented features of
12822       Objective-C with those of C++.
12823
12824  Java (GCJ)
12825
12826     * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19
12827       features (plus some 0.20 bug-fixes)
12828          + Networking
12829               o The java.net.HttpURLConnection implementation no longer
12830                 buffers the entire response body in memory. This means
12831                 that response bodies larger than available memory can now
12832                 be handled.
12833          + (N)IO
12834               o NIO FileChannel.map implementation, fast bulk put
12835                 implementation for DirectByteBuffer (speeds up this
12836                 method 10x).
12837               o FileChannel.lock() and FileChannel.force() implemented.
12838          + XML
12839               o gnu.xml fix for nodes created outside a namespace
12840                 context.
12841               o Add support for output indenting and
12842                 cdata-section-elements output instruction in
12843                 xml.transform.
12844               o xml.xpath corrections for cases where elements/attributes
12845                 might have been created in non-namespace-aware mode.
12846                 Corrections to handling of XSL variables and minor
12847                 conformance updates.
12848          + AWT
12849               o GNU JAWT implementation, the AWT Native Interface, which
12850                 allows direct access to native screen resources from
12851                 within a Canvas's paint method. GNU Classpath Examples
12852                 comes with a Demo, see libjava/classpath/examples/README.
12853               o awt.datatransfer updated to 1.5 with support for
12854                 FlavorEvents. The gtk+ awt peers now allow copy/paste of
12855                 text, images, URIs/files and serialized objects with
12856                 other applications and tracking clipboard change events
12857                 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized
12858                 objects are supported). A GNU Classpath Examples
12859                 datatransfer Demo was added to show the new
12860                 functionality.
12861               o Split gtk+ awt peers event handling in two threads and
12862                 improve gdk lock handling (solves several awt lock ups).
12863               o Speed up awt Image loading.
12864               o Better gtk+ scrollbar peer implementation when using gtk+
12865                 >= 2.6.
12866               o Handle image loading errors correctly for gdkpixbuf and
12867                 MediaTracker.
12868               o Better handle GDK lock. Properly prefix gtkpeer native
12869                 functions (cp_gtk).
12870               o GdkGraphics2D has been updated to use Cairo 0.5.x or
12871                 higher.
12872               o BufferedImage and GtkImage rewrites. All image drawing
12873                 operations should now work correctly (flipping requires
12874                 gtk+ >= 2.6)
12875               o When gtk+ 2.6 or higher is installed the default log
12876                 handler will produce stack traces whenever a WARNING,
12877                 CRITICAL or ERROR message is produced.
12878          + Free Swing
12879               o The RepaintManager has been reworked for more efficient
12880                 painting, especially for large GUIs.
12881               o The layout manager OverlayLayout has been implemented,
12882                 the BoxLayout has been rewritten to make use of the
12883                 SizeRequirements utility class and caching for more
12884                 efficient layout.
12885               o Improved accessibility support.
12886               o Significant progress has been made in the implementation
12887                 of the javax.swing.plaf.metal package, with most UI
12888                 delegates in a working state now. Please test this with
12889                 your own applications and provide feedback that will help
12890                 us to improve this package.
12891               o The GUI demo (gnu.classpath.examples.swing.Demo) has been
12892                 extended to highlight various features in our Free Swing
12893                 implementation. And it includes a look and feel switcher
12894                 for Metal (default), Ocean and GNU themes.
12895               o The javax.swing.plaf.multi package is now implemented.
12896               o Editing and several key actions for JTree and JTable were
12897                 implemented.
12898               o Lots of icons and look and feel improvements for Free
12899                 Swing basic and metal themes were added. Try running the
12900                 GNU Classpath Swing Demo in examples
12901                 (gnu.classpath.examples.swing.Demo) with:
12902                 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee
12903                 l or
12904                 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee
12905                 l
12906               o Start of styled text capabilites for java.swing.text.
12907               o DefaultMutableTreeNode pre-order, post-order, depth-first
12908                 and breadth-first traversal enumerations implemented.
12909               o JInternalFrame colors and titlebar draw properly.
12910               o JTree is working up to par (icons, selection and keyboard
12911                 traversal).
12912               o JMenus were made more compatible in visual and
12913                 programmatic behavior.
12914               o JTable changeSelection and multiple selections
12915                 implemented.
12916               o JButton and JToggleButton change states work properly
12917                 now.
12918               o JFileChooser fixes.
12919               o revalidate() and repaint() fixes which make Free Swing
12920                 much more responsive.
12921               o MetalIconFactory implemented.
12922               o Free Swing Top-Level Compatibility. JFrame, JDialog,
12923                 JApplet, JInternalFrame, and JWindow are now 1.5
12924                 compatible in the sense that you can call add() and
12925                 setLayout() directly on them, which will have the same
12926                 effect as calling getContentPane().add() and
12927                 getContentPane().setLayout().
12928               o The JTree interface has been completed. JTrees now
12929                 recognizes mouse clicks and selections work.
12930               o BoxLayout works properly now.
12931               o Fixed GrayFilter to actually work.
12932               o Metal SplitPane implemented.
12933               o Lots of Free Swing text and editor stuff work now.
12934          + Free RMI and Corba
12935               o Andrew Watson, Vice President and Technical Director of
12936                 the Object Management Group, has officially assigned us
12937                 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will
12938                 mark remote classpath-specific system exceptions.
12939                 Obtaining the VMCID means that GNU Classpath now is a
12940                 recogniseable type of node in a highly interoperable
12941                 CORBA world.
12942               o GNU Classpath now includes the first working draft to
12943                 support the RMI over IIOP protocol. The current
12944                 implementation is capable of remote invocations,
12945                 transferring various Serializables and Externalizables
12946                 via RMI-IIOP protocol. It can flatten graphs and, at
12947                 least for the simple cases, is interoperable with 1.5
12948                 JDKs.
12949               o org.omg.PortableInterceptor and related functionality in
12950                 other packages is now implemented:
12951                    # The sever and client interceptors work as required
12952                      since 1.4.
12953                    # The IOR interceptor works as needed for 1.5.
12954               o The org.omg.DynamicAny package is completed and passes
12955                 the prepared tests.
12956               o The Portable Object Adapter should now support the output
12957                 of the recent IDL to java compilers. These compilers now
12958                 generate servants and not CORBA objects as before, making
12959                 the output depend on the existing POA implementation.
12960                 Completing POA means that such code can already be tried
12961                 to run on Classpath. Our POA is tested for the following
12962                 usager scenarios:
12963                    # POA converts servant to the CORBA object.
12964                    # Servant provides to the CORBA object.
12965                    # POA activates new CORBA object with the given Object
12966                      Id (byte array) that is later accessible for the
12967                      servant.
12968                    # During the first call, the ServantActivator provides
12969                      servant for this and all subsequent calls on the
12970                      current object.
12971                    # During each call, the ServantLocator provides
12972                      servant for this call only.
12973                    # ServantLocator or ServantActivator forwards call to
12974                      another server.
12975                    # POA has a single servant, responsible for all
12976                      objects.
12977                    # POA has a default servant, but some objects are
12978                      explicitly connected to they specific servants.
12979                 The POA is verified using tests from the former
12980                 cost.omg.org.
12981               o The CORBA implementation is now a working prototype that
12982                 should support features up to 1.3 inclusive. We invite
12983                 groups writing CORBA dependent applications to try
12984                 Classpath implementation, reporting any possible bugs.
12985                 The CORBA prototype is interoperable with Sun's
12986                 implementation v 1.4, transferring object references,
12987                 primitive types, narrow and wide strings, arrays,
12988                 structures, trees, abstract interfaces and value types
12989                 (feature of CORBA 2.3) between these two platforms.
12990                 Remote exceptions are transferred and handled correctly.
12991                 The stringified object references (IORs) from various
12992                 sources are parsed as required. The transient (for
12993                 current session) and permanent (till jre restart)
12994                 redirections work. Both Little and Big Endian encoded
12995                 messages are accepted. The implementation is verified
12996                 using tests from the former cost.omg.org. The current
12997                 release includes working examples (see the examples
12998                 directory), demonstrating the client-server
12999                 communication, using either CORBA Request or IDL-based
13000                 stub (usually generated by a IDL to java compiler). These
13001                 examples also show how to use the Classpath CORBA naming
13002                 service. The IDL to java compiler is not yet written, but
13003                 as our library must be compatible, it naturally accepts
13004                 the output of other idlj implementations.
13005          + Misc
13006               o Updated TimeZone data against Olson tzdata2005l.
13007               o Make zip and jar packages UTF-8 clean.
13008               o "native" code builds and compiles (warning free) on
13009                 Darwin and Solaris.
13010               o java.util.logging.FileHandler now rotates files.
13011               o Start of a generic JDWP framework in gnu/classpath/jdwp.
13012                 This is unfinished, but feedback (at classpath@gnu.org)
13013                 from runtime hackers is greatly appreciated. Although
13014                 most of the work is currently being done around gcj/gij
13015                 we want this framework to be as VM neutral as possible.
13016                 Early design is described in:
13017                 [2]https://gcc.gnu.org/ml/java/2005-05/msg00260.html
13018               o QT4 AWT peers, enable by giving configure
13019                 --enable-qt-peer. Included, but not ready for production
13020                 yet. They are explicitly disabled and not supported. But
13021                 if you want to help with the development of these new
13022                 features we are interested in feedback. You will have to
13023                 explicitly enable them to try them out (and they will
13024                 most likely contain bugs).
13025               o Documentation fixes all over the place. See
13026                 [3]https://developer.classpath.org/doc/
13027
13028New Targets and Target Specific Improvements
13029
13030  IA-32/x86-64
13031
13032     * The x86-64 medium model (that allows building applications whose
13033       data segment exceeds 4GB) was redesigned to match latest ABI draft.
13034       New implementation split large datastructures into separate segment
13035       improving performance of accesses to small datastructures and also
13036       allows linking of small model libraries into medium model programs
13037       as long as the libraries are not accessing the large datastructures
13038       directly. Medium model is also supported in position independent
13039       code now.
13040       The ABI change results in partial incompatibility among medium
13041       model objects. Linking medium model libraries (or objects) compiled
13042       with new compiler into medium model program compiled with older
13043       will likely result in exceeding ranges of relocations.
13044       Binutils 2.16.91 or newer are required for compiling medium model
13045       now.
13046
13047  RS6000 (POWER/PowerPC)
13048
13049     * The AltiVec vector primitives in <altivec.h> are now implemented in
13050       a way that puts a smaller burden on the preprocessor, instead
13051       processing the "overloading" in the front ends. This should benefit
13052       compilation speed on AltiVec vector code.
13053     * AltiVec initializers now are generated more efficiently.
13054     * The popcountb instruction available on POWER5 now is generated.
13055     * The floating point round to integer instructions available on
13056       POWER5+ now is generated.
13057     * Floating point divides can be synthesized using the floating point
13058       reciprocal estimate instructions.
13059     * Double precision floating point constants are initialized as single
13060       precision values if they can be represented exactly.
13061
13062  S/390, zSeries and System z9
13063
13064     * Support for the IBM System z9 109 processor has been added. When
13065       using the -march=z9-109 option, the compiler will generate code
13066       making use of instructions provided by the extended immediate
13067       facility.
13068     * Support for 128-bit IEEE floating point has been added. When using
13069       the -mlong-double-128 option, the compiler will map the long double
13070       data type to 128-bit IEEE floating point. Using this option
13071       constitutes an ABI change, and requires glibc support.
13072     * Various changes to improve performance of generated code have been
13073       implemented, including:
13074          + In functions that do not require a literal pool, register %r13
13075            (which is traditionally reserved as literal pool pointer), can
13076            now be freely used for other purposes by the compiler.
13077          + More precise tracking of register use allows the compiler to
13078            generate more efficient function prolog and epilog code in
13079            certain cases.
13080          + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING
13081            instructions are now used to implement C string functions.
13082          + The MOVE CHARACTER instruction with single byte overlap is now
13083            used to implement the memset function with non-zero fill byte.
13084          + The LOAD ZERO instructions are now used where appropriate.
13085          + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK,
13086            and INSERT IMMEDIATE instructions are now used more frequently
13087            to optimize bitfield operations.
13088          + The BRANCH ON COUNT instruction is now used more frequently.
13089            In particular, the fact that a loop contains a subroutine call
13090            no longer prevents the compiler from using this instruction.
13091          + The compiler is now aware that all shift and rotate
13092            instructions implicitly truncate the shift count to six bits.
13093     * Back-end support for the following generic features has been
13094       implemented:
13095          + The full set of [4]built-in functions for atomic memory
13096            access.
13097          + The -fstack-protector feature.
13098          + The optimization pass avoiding unnecessary stores of incoming
13099            argument registers in functions with variable argument list.
13100
13101  SPARC
13102
13103     * The default code model in 64-bit mode has been changed from
13104       Medium/Anywhere to Medium/Middle on Solaris.
13105     * TLS support is disabled by default on Solaris prior to release 10.
13106       It can be enabled on TLS-capable Solaris 9 versions (4/04 release
13107       and later) by specifying --enable-tls at configure time.
13108
13109  MorphoSys
13110
13111     * Support has been added for this new architecture.
13112
13113Obsolete Systems
13114
13115Documentation improvements
13116
13117Other significant improvements
13118
13119     * GCC can now emit code for protecting applications from
13120       stack-smashing attacks. The protection is realized by buffer
13121       overflow detection and reordering of stack variables to avoid
13122       pointer corruption.
13123     * Some built-in functions have been fortified to protect them against
13124       various buffer overflow (and format string) vulnerabilities.
13125       Compared to the mudflap bounds checking feature, the safe builtins
13126       have far smaller overhead. This means that programs built using
13127       safe builtins should not experience any measurable slowdown.
13128
13129GCC 4.1.2
13130
13131   This is the [5]list of problem reports (PRs) from GCC's bug tracking
13132   system that are known to be fixed in the 4.1.2 release. This list might
13133   not be complete (that is, it is possible that some PRs that have been
13134   fixed are not listed here).
13135
13136   When generating code for a shared library, GCC now recognizes that
13137   global functions may be replaced when the program runs. Therefore, it
13138   is now more conservative in deducing information from the bodies of
13139   functions. For example, in this example:
13140    void f() {}
13141    void g() {
13142     try { f(); }
13143     catch (...) {
13144       cout << "Exception";
13145     }
13146    }
13147
13148   G++ would previously have optimized away the catch clause, since it
13149   would have concluded that f cannot throw exceptions. Because users may
13150   replace f with another function in the main body of the program, this
13151   optimization is unsafe, and is no longer performed. If you wish G++ to
13152   continue to optimize as before, you must add a throw() clause to the
13153   declaration of f to make clear that it does not throw exceptions.
13154
13155
13156    For questions related to the use of GCC, please consult these web
13157    pages and the [6]GCC manuals. If that fails, the
13158    [7]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13159    web pages and the development of GCC are welcome on our developer
13160    list at [8]gcc@gcc.gnu.org. All of [9]our lists have public archives.
13161
13162   Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and
13163   distribution of this entire article is permitted in any medium,
13164   provided this notice is preserved.
13165
13166   These pages are [11]maintained by the GCC team. Last modified
13167   2021-10-18[12].
13168
13169References
13170
13171   1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
13172   2. https://gcc.gnu.org/ml/java/2005-05/msg00260.html
13173   3. https://developer.classpath.org/doc/
13174   4. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html
13175   5. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2
13176   6. https://gcc.gnu.org/onlinedocs/
13177   7. mailto:gcc-help@gcc.gnu.org
13178   8. mailto:gcc@gcc.gnu.org
13179   9. https://gcc.gnu.org/lists.html
13180  10. https://www.fsf.org/
13181  11. https://gcc.gnu.org/about.html
13182  12. http://validator.w3.org/check/referer
13183======================================================================
13184http://gcc.gnu.org/gcc-4.0/index.html
13185                             GCC 4.0 Release Series
13186
13187   (This release series is no longer supported.)
13188
13189   January 31, 2007
13190
13191   The [1]GNU project and the GCC developers are pleased to announce the
13192   release of GCC 4.0.4.
13193
13194   This release is a bug-fix release, containing fixes for regressions in
13195   GCC 4.0.3 relative to previous releases of GCC.
13196
13197Release History
13198
13199   GCC 4.0.4
13200          January 31, 2007 ([2]changes)
13201
13202   GCC 4.0.3
13203          March 10, 2006 ([3]changes)
13204
13205   GCC 4.0.2
13206          September 28, 2005 ([4]changes)
13207
13208   GCC 4.0.1
13209          July 7, 2005 ([5]changes)
13210
13211   GCC 4.0.0
13212          April 20, 2005 ([6]changes)
13213
13214References and Acknowledgements
13215
13216   GCC used to stand for the GNU C Compiler, but since the compiler
13217   supports several other languages aside from C, it now stands for the
13218   GNU Compiler Collection.
13219
13220   A list of [7]successful builds is updated as new information becomes
13221   available.
13222
13223   The GCC developers would like to thank the numerous people that have
13224   contributed new features, improvements, bug fixes, and other changes as
13225   well as test results to GCC. This [8]amazing group of volunteers is
13226   what makes GCC successful.
13227
13228   For additional information about GCC please refer to the [9]GCC project
13229   web site or contact the [10]GCC development mailing list.
13230
13231   To obtain GCC please use [11]our mirror sites, or [12]our version
13232   control system.
13233
13234
13235    For questions related to the use of GCC, please consult these web
13236    pages and the [13]GCC manuals. If that fails, the
13237    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13238    web pages and the development of GCC are welcome on our developer
13239    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
13240    archives.
13241
13242   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
13243   distribution of this entire article is permitted in any medium,
13244   provided this notice is preserved.
13245
13246   These pages are [18]maintained by the GCC team. Last modified
13247   2021-07-28[19].
13248
13249References
13250
13251   1. http://www.gnu.org/
13252   2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
13253   3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3
13254   4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2
13255   5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1
13256   6. http://gcc.gnu.org/gcc-4.0/changes.html
13257   7. http://gcc.gnu.org/gcc-4.0/buildstat.html
13258   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
13259   9. http://gcc.gnu.org/index.html
13260  10. mailto:gcc@gcc.gnu.org
13261  11. http://gcc.gnu.org/mirrors.html
13262  12. http://gcc.gnu.org/git.html
13263  13. https://gcc.gnu.org/onlinedocs/
13264  14. mailto:gcc-help@gcc.gnu.org
13265  15. mailto:gcc@gcc.gnu.org
13266  16. https://gcc.gnu.org/lists.html
13267  17. https://www.fsf.org/
13268  18. https://gcc.gnu.org/about.html
13269  19. http://validator.w3.org/check/referer
13270======================================================================
13271http://gcc.gnu.org/gcc-4.0/changes.html
13272                             GCC 4.0 Release Series
13273                        Changes, New Features, and Fixes
13274
13275   The latest release in the 4.0 release series is [1]GCC 4.0.4.
13276
13277Caveats
13278
13279     * GCC now generates location lists by default when compiling with
13280       debug info and optimization.
13281          + GDB 6.0 and older crashes when it sees location lists. GDB 6.1
13282            or later is needed to debug binaries containing location
13283            lists.
13284          + When you are trying to view a value of a variable in a part of
13285            a function where it has no location (for example when the
13286            variable is no longer used and thus its location was used for
13287            something else) GDB will say that it is not available.
13288       You can disable generating location lists by -fno-var-tracking.
13289     * GCC no longer accepts the -fwritable-strings option. Use named
13290       character arrays when you need a writable string.
13291     * The options -freduce-all-givs and -fmove-all-movables have been
13292       discontinued. They were used to circumvent a shortcoming in the
13293       heuristics of the old loop optimization code with respect to common
13294       Fortran constructs. The new (tree) loop optimizer works differently
13295       and doesn't need those work-arounds.
13296     * The graph-coloring register allocator, formerly enabled by the
13297       option -fnew-ra, has been discontinued.
13298     * -I- has been deprecated. -iquote is meant to replace the need for
13299       this option.
13300     * The MIPS -membedded-pic and -mrnames options have been removed.
13301     * All MIPS targets now require the GNU assembler. In particular, IRIX
13302       configurations can no longer use the MIPSpro assemblers, although
13303       they do still support the MIPSpro linkers.
13304     * The SPARC option -mflat has been removed.
13305     * English-language diagnostic messages will now use Unicode quotation
13306       marks in UTF-8 locales. (Non-English messages already used the
13307       quotes appropriate for the language in previous releases.) If your
13308       terminal does not support UTF-8 but you are using a UTF-8 locale
13309       (such locales are the default on many GNU/Linux systems) then you
13310       should set LC_CTYPE=C in the environment to disable that locale.
13311       Programs that parse diagnostics and expect plain ASCII
13312       English-language messages should set LC_ALL=C. See [2]Markus Kuhn's
13313       explanation of Unicode quotation marks for more information.
13314     * The specs file is no longer installed on most platforms. Most users
13315       will be totally unaffected. However, if you are accustomed to
13316       editing the specs file yourself, you will now have to use the
13317       -dumpspecs option to generate the specs file, and then edit the
13318       resulting file.
13319
13320General Optimizer Improvements
13321
13322     * The [3]tree ssa branch has been merged. This merge has brought in a
13323       completely new optimization framework based on a higher level
13324       intermediate representation than the existing RTL representation.
13325       Numerous new code transformations based on the new framework are
13326       available in GCC 4.0, including:
13327          + Scalar replacement of aggregates
13328          + Constant propagation
13329          + Value range propagation
13330          + Partial redundancy elimination
13331          + Load and store motion
13332          + Strength reduction
13333          + Dead store elimination
13334          + Dead and unreachable code elimination
13335          + [4]Autovectorization
13336          + Loop interchange
13337          + Tail recursion by accumulation
13338       Many of these passes outperform their counterparts from previous
13339       GCC releases.
13340     * [5]Swing Modulo Scheduling (SMS). An RTL level instruction
13341       scheduling optimization intended for loops that perform heavy
13342       computations.
13343
13344New Languages and Language specific improvements
13345
13346  C family
13347
13348     * The sentinel attribute has been added to GCC. This function
13349       attribute allows GCC to warn when variadic functions such as execl
13350       are not NULL terminated. See the GCC manual for a complete
13351       description of its behavior.
13352     * Given __attribute__((alias("target"))) it is now an error if target
13353       is not a symbol, defined in the same translation unit. This also
13354       applies to aliases created by #pragma weak alias=target. This is
13355       because it's meaningless to define an alias to an undefined symbol.
13356       On Solaris, the native assembler would have caught this error, but
13357       GNU as does not.
13358
13359  C and Objective-C
13360
13361     * The -Wstrict-aliasing=2 option has been added. This warning catches
13362       all unsafe cases, but it may also give a warning for some cases
13363       that are safe.
13364     * The cast-as-lvalue, conditional-expression-as-lvalue and
13365       compound-expression-as-lvalue extensions, which were deprecated in
13366       3.3.4 and 3.4, have been removed.
13367     * The -fwritable-strings option, which was deprecated in 3.4, has
13368       been removed.
13369     * #pragma pack() semantics have been brought closer to those used by
13370       other compilers. This also applies to C++.
13371     * Taking the address of a variable with register storage is invalid
13372       in C. GCC now issues an error instead of a warning.
13373     * Arrays of incomplete element type are invalid in C. GCC now issues
13374       an error for such arrays. Declarations such as extern struct s x[];
13375       (where struct s has not been defined) can be moved after the
13376       definition of struct s. Function parameters declared as arrays of
13377       incomplete type can instead be declared as pointers.
13378
13379  C++
13380
13381     * When compiling without optimizations (-O0), the C++ front end is
13382       much faster than in any previous versions of GCC. Independent
13383       testers have measured speed-ups up to 25% in real-world production
13384       code, compared to the 3.4 family (which was already the fastest
13385       version to date). Upgrading from older versions might show even
13386       bigger improvements.
13387     * ELF visibility attributes can now be applied to a class type, so
13388       that it affects every member function of a class at once, without
13389       having to specify each individually:
13390class __attribute__ ((visibility("hidden"))) Foo
13391{
13392   int foo1();
13393   void foo2();
13394};
13395       The syntax is deliberately similar to the __declspec() system used
13396       by Microsoft Windows based compilers, allowing cross-platform
13397       projects to easily reuse their existing macro system for denoting
13398       exports and imports. By explicitly marking internal classes never
13399       used outside a binary as hidden, one can completely avoid PLT
13400       indirection overheads during their usage by the compiler. You can
13401       find out more about the advantages of this at
13402       [6]https://www.akkadia.org/drepper/dsohowto.pdf
13403     * The -fvisibility-inlines-hidden option has been added which marks
13404       all inlineable functions as having hidden ELF visibility, thus
13405       removing their symbol and typeinfo from the exported symbol table
13406       of the output ELF binary. Using this option can reduce the exported
13407       symbol count of template-heavy code by up to 40% with no code
13408       change at all, thus notably improving link and load times for the
13409       binary as well as a reduction in size of up to 10%. Also, check the
13410       new [7]-fvisibility option.
13411     * The compiler now uses the library interface specified by the [8]C++
13412       ABI for thread-safe initialization of function-scope static
13413       variables. Most users should leave this alone, but embedded
13414       programmers may want to disable this by specifying
13415       -fno-threadsafe-statics for a small savings in code size.
13416     * Taking the address of an explicit register variable is no longer
13417       supported. Note that C++ allows taking the address of variables
13418       with register storage so this will continue to compile with a
13419       warning. For example, assuming that r0 is a machine register:
13420register int foo asm ("r0");
13421register int bar;
13422&foo; // error, no longer accepted
13423&bar; // OK, with a warning
13424     * G++ has an undocumented extension to virtual function covariancy
13425       rules that allowed the overrider to return a type that was
13426       implicitly convertable to the overridden function's return type.
13427       For instance a function returning void * could be overridden by a
13428       function returning T *. This is now deprecated and will be removed
13429       in a future release.
13430     * The G++ minimum and maximum operators (<? and >?) and their
13431       compound forms (<?=) and >?=) have been deprecated and will be
13432       removed in a future version. Code using these operators should be
13433       modified to use std::min and std::max instead.
13434     * Declaration of nested classes of class templates as friends are
13435       supported:
13436template <typename T> struct A {
13437  class B {};
13438};
13439class C {
13440  template <typename T> friend class A<T>::B;
13441};
13442       This complements the feature member functions of class templates as
13443       friends introduced in GCC 3.4.0.
13444     * When declaring a friend class using an unqualified name, classes
13445       outside the innermost non-class scope are not searched:
13446class A;
13447namespace N {
13448  class B {
13449    friend class A;   // Refer to N::A which has not been declared yet
13450                      // because name outside namespace N are not searched
13451    friend class ::A; // Refer to ::A
13452  };
13453}
13454       Hiding the friend name until declaration is still not implemented.
13455     * Friends of classes defined outside their namespace are correctly
13456       handled:
13457namespace N {
13458  class A;
13459}
13460class N::A {
13461  friend class B; // Refer to N::B in GCC 4.0.0
13462                  // but ::B in earlier versions of GCC
13463};
13464
13465    Runtime Library (libstdc++)
13466
13467     * Optimization work:
13468          + Added efficient specializations of istream functions for char
13469            and wchar_t.
13470          + Further performance tuning of strings, in particular wrt
13471            single-char append and getline.
13472          + iter_swap - and therefore most of the mutating algorithms -
13473            now makes an unqualified call to swap when the value_type of
13474            the two iterators is the same.
13475     * A large subset of the features in Technical Report 1 (TR1 for
13476       short) is experimentally delivered (i.e., no guarantees about the
13477       implementation are provided. In particular it is not promised that
13478       the library will remain link-compatible when code using TR1 is
13479       used):
13480          + General utilities such as reference_wrapper and shared_ptr.
13481          + Function objects, i.e., result_of, mem_fn, bind, function.
13482          + Support for metaprogramming.
13483          + New containers such as tuple, array, unordered_set,
13484            unordered_map, unordered_multiset, unordered_multimap.
13485     * As usual, many bugs have been fixed and LWG resolutions implemented
13486       for the first time (e.g., DR 409).
13487
13488  Java
13489
13490     * In order to prevent naming conflicts with other implementations of
13491       these tools, some GCJ binaries have been renamed:
13492          + rmic is now grmic,
13493          + rmiregistry is now grmiregistry, and
13494          + jar is now fastjar.
13495       In particular, these names were problematic for the jpackage.org
13496       packaging conventions which install symlinks in /usr/bin that point
13497       to the preferred versions of these tools.
13498     * The -findirect-dispatch argument to the compiler now works and
13499       generates code following a new "binary compatibility" ABI. Code
13500       compiled this way follows the binary compatibility rules of the
13501       Java Language Specification.
13502     * libgcj now has support for using GCJ as a JIT, using the
13503       gnu.gcj.jit family of system properties.
13504     * libgcj can now find a shared library corresponding to the bytecode
13505       representation of a class. See the documentation for the new
13506       gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system
13507       property.
13508     * There have been many improvements to the class library. Here are
13509       some highlights:
13510          + Much more of AWT and Swing exist.
13511          + Many new packages and classes were added, including
13512            java.util.regex, java.net.URI, javax.crypto,
13513            javax.crypto.interfaces, javax.crypto.spec, javax.net,
13514            javax.net.ssl, javax.security.auth,
13515            javax.security.auth.callback, javax.security.auth.login,
13516            javax.security.auth.x500, javax.security.sasl, org.ietf.jgss,
13517            javax.imageio, javax.imageio.event, javax.imageio.spi,
13518            javax.print, javax.print.attribute,
13519            javax.print.attribute.standard, javax.print.event, and
13520            javax.xml
13521          + Updated SAX and DOM, and imported GNU JAXP
13522
13523  Fortran
13524
13525     * A new [9]Fortran front end has replaced the aging GNU Fortran 77
13526       front end. The new front end supports Fortran 90 and Fortran 95. It
13527       may not yet be as stable as the old Fortran front end.
13528
13529  Ada
13530
13531     * Ada (with tasking and Zero Cost Exceptions) is now available on
13532       many more targets, including but not limited to: alpha-linux,
13533       hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux,
13534       s390x-linux, sparc-linux.
13535     * Some of the new Ada 2005 features are now implemented like
13536       Wide_Wide_Character and Ada.Containers.
13537     * Many bugs have been fixed, tools and documentation improved.
13538     * To compile Ada from the sources, install an older working Ada
13539       compiler and then use --enable-languages=ada at configuration time,
13540       since the Ada front end is not currently activated by default. See
13541       the [10]Installing GCC for details.
13542
13543New Targets and Target Specific Improvements
13544
13545  H8/300
13546
13547     * The frame layout has changed. In the new layout, the prologue of a
13548       function first saves registers and then allocate space for locals,
13549       resulting in an 1% improvement on code size.
13550
13551  IA-32/x86-64 (AMD64)
13552
13553     * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10,
13554       log1p, log2, logb and tan mathematical builtins (and their float
13555       and long double variants) are now implemented as inline x87
13556       intrinsics when using -ffast-math.
13557     * The ceil, floor, nearbyint, rint and trunc mathematical builtins
13558       (and their float and long double variants) are now implemented as
13559       inline x87 intrinsics when using -ffast-math.
13560     * The x87's fsincos instruction is now used automatically with
13561       -ffast-math when calculating both the sin and cos of the same
13562       argument.
13563     * Instruction selection for multiplication and division by constants
13564       has been improved.
13565
13566  IA-64
13567
13568     * Floating point division, integer division and sqrt are now inlined,
13569       resulting in significant performance improvements on some codes.
13570
13571  MIPS
13572
13573     * Division by zero checks now use conditional traps if the target
13574       processor supports them. This decreases code size by one word per
13575       division operation. The old behavior (branch and break) can be
13576       obtained either at configure time by passing --with-divide=breaks
13577       to configure or at runtime by passing -mdivide-breaks to GCC.
13578     * Support for MIPS64 paired-single instructions has been added. It is
13579       enabled by -mpaired-single and can be accessed using both the
13580       target-independent vector extensions and new MIPS-specific built-in
13581       functions.
13582     * Support for the MIPS-3D ASE has been added. It is enabled by
13583       -mips3d and provides new MIPS-3D-specific built-in functions.
13584     * The -mexplicit-relocs option now supports static n64 code (as is
13585       used, for example, in 64-bit linux kernels). -mexplicit-relocs
13586       should now be feature-complete and is enabled by default when GCC
13587       is configured to use a compatible assembler.
13588     * Support for the NEC VR4130 series has been added. This support
13589       includes the use of VR-specific instructions and a new VR4130
13590       scheduler. Full VR4130 support can be selected with -march=vr4130
13591       while code for any ISA can be tuned for the VR4130 using
13592       -mtune=vr4130. There is also a new -mvr4130-align option that
13593       produces better schedules at the cost of increased code size.
13594     * Support for the Broadcom SB-1 has been extended. There is now an
13595       SB-1 scheduler as well as support for the SB-1-specific
13596       paired-single instructions. Full SB-1 support can be selected with
13597       -march=sb1 while code for any ISA can be optimized for the SB-1
13598       using -mtune=sb1.
13599     * The compiler can now work around errata in R4000, R4400, VR4120 and
13600       VR4130 processors. These workarounds are enabled by -mfix-r4000,
13601       -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120
13602       and VR4130 workarounds need binutils 2.16 or above.
13603     * IRIX shared libraries are now installed into the standard library
13604       directories: o32 libraries go into lib/, n32 libraries go into
13605       lib32/ and n64 libraries go into lib64/.
13606     * The compiler supports a new -msym32 option. It can be used to
13607       optimize n64 code in which all symbols are known to have 32-bit
13608       values.
13609
13610  S/390 and zSeries
13611
13612     * New command-line options help to generate code intended to run in
13613       an environment where stack space is restricted, e.g. Linux kernel
13614       code:
13615          + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time
13616            warnings for single functions that require large or dynamic
13617            stack frames.
13618          + -mstack-size and -mstack-guard generate code that checks for
13619            stack overflow at run time.
13620          + -mpacked-stack generates code that reduces the stack frame
13621            size of many functions by reusing unneeded parts of the stack
13622            bias area.
13623     * The -msoft-float option now ensures that generated code never
13624       accesses floating point registers.
13625     * The s390x-ibm-tpf target now fully supports C++, including
13626       exceptions and threads.
13627     * Various changes to improve performance of the generated code have
13628       been implemented, including:
13629          + GCC now uses sibling calls where possible.
13630          + Condition code handling has been optimized, allowing GCC to
13631            omit redundant comparisons in certain cases.
13632          + The cost function guiding many optimizations has been refined
13633            to more accurately represent the z900 and z990 processors.
13634          + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW
13635            instructions are now used to avoid conditional branches in
13636            certain cases.
13637          + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to
13638            optimize address arithmetic required to access large stack
13639            frames.
13640          + GCC now makes more efficient use of memory-to-memory type
13641            instructions (MVC, CLC, ...).
13642          + More precise tracking of special register use allows better
13643            instruction scheduling, in particular of the function prologue
13644            and epilogue sequences.
13645          + The Java front end now generates inline code to implement
13646            integer division, instead of calling library routines.
13647
13648  SPARC
13649
13650     * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and
13651       -mf934 have been removed. They have been replaced with -mcpu=xxx.
13652     * The internal model used to estimate the relative cost of each
13653       instruction has been updated. It is expected to give better results
13654       on recent UltraSPARC processors.
13655     * Code generation for function prologues and epilogues has been
13656       improved, resulting in better scheduling and allowing multiple exit
13657       points in functions.
13658     * Support for Sun's Visual Instruction Set (VIS) has been enhanced.
13659       It is enabled by -mvis and provides new built-in functions for VIS
13660       instructions on UltraSPARC processors.
13661     * The option -mapp-regs has been turned on by default on Solaris too.
13662
13663  NetWare
13664
13665     * Novell NetWare (on ix86, no other hardware platform was ever really
13666       supported by this OS) has been re-enabled and the ABI supported by
13667       GCC has been brought into sync with that of MetroWerks CodeWarrior
13668       (the ABI previously supported was that of some Unix systems, which
13669       NetWare never tried to support).
13670
13671Obsolete Systems
13672
13673   Support for a number of older systems has been declared obsolete in GCC
13674   4.0. Unless there is activity to revive them, the next release of GCC
13675   will have their sources permanently removed.
13676
13677   All GCC ports for the following processor architectures have been
13678   declared obsolete:
13679     * Intel i860
13680     * Ubicom IP2022
13681     * National Semiconductor NS32K (ns32k)
13682     * Texas Instruments TMS320C[34]x
13683
13684   Also, those for some individual systems have been obsoleted:
13685     * SPARC family
13686          + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf,
13687            sparc86x-*-elf)
13688          + OpenBSD 32-bit (sparc-*-openbsd*)
13689
13690Documentation improvements
13691
13692Other significant improvements
13693
13694     * Location lists are now generated by default when compiling with
13695       debug info and optimization. Location lists provide more accurate
13696       debug info about locations of variables and they allow debugging
13697       code compiled with -fomit-frame-pointer.
13698     * The -fvisibility option has been added which allows the default ELF
13699       visibility of all symbols to be set per compilation and the new
13700       #pragma GCC visibility preprocessor command allows the setting of
13701       default ELF visibility for a region of code. Using
13702       -fvisibility=hidden especially in combination with the new
13703       -fvisibility-inlines-hidden can yield substantial improvements in
13704       output binary quality including avoiding PLT indirection overheads,
13705       reduction of the exported symbol count by up to 60% (with resultant
13706       improvements to link and load times), better scope for the
13707       optimizer to improve code and up to a 20% reduction in binary size.
13708       Using these options correctly yields a binary with a similar symbol
13709       count to a Windows DLL.
13710       Perhaps more importantly, this new feature finally allows (with
13711       careful planning) complete avoidance of symbol clashes when
13712       manually loading shared objects with RTLD_GLOBAL, thus finally
13713       solving problems many projects such as python were forced to use
13714       RTLD_LOCAL for (with its resulting issues for C++ correctness). You
13715       can find more information about using these options at
13716       [11]https://gcc.gnu.org/wiki/Visibility.
13717     __________________________________________________________________
13718
13719GCC 4.0.1
13720
13721   This is the [12]list of problem reports (PRs) from GCC's bug tracking
13722   system that are known to be fixed in the 4.0.1 release. This list might
13723   not be complete (that is, it is possible that some PRs that have been
13724   fixed are not listed here).
13725
13726GCC 4.0.2
13727
13728   This is the [13]list of problem reports (PRs) from GCC's bug tracking
13729   system that are known to be fixed in the 4.0.2 release. This list might
13730   not be complete (that is, it is possible that some PRs that have been
13731   fixed are not listed here).
13732
13733   Unfortunately, due to a release engineering failure, this release has a
13734   regression on Solaris that will affect some C++ programs. We suggest
13735   that Solaris users apply a [14]patch that corrects the problem. Users
13736   who do not wish to apply the patch should explicitly link C++ programs
13737   with the -pthreads option, even if they do not use threads. This
13738   problem has been corrected in the current 4.0 branch sources and will
13739   not be present in GCC 4.0.3.
13740
13741GCC 4.0.3
13742
13743   Starting with this release, the function getcontext is recognized by
13744   the compiler as having the same semantics as the setjmp function. In
13745   particular, the compiler will ensure that all registers are dead before
13746   calling such a function and will emit a warning about the variables
13747   that may be clobbered after the second return from the function.
13748
13749GCC 4.0.4
13750
13751   This is the [15]list of problem reports (PRs) from GCC's bug tracking
13752   system that are known to be fixed in the 4.0.4 release. This list might
13753   not be complete (that is, it is possible that some PRs that have been
13754   fixed are not listed here).
13755
13756   The 4.0.4 release is provided for those that require a high degree of
13757   binary compatibility with previous 4.0.x releases. For most users, the
13758   GCC team recommends that version 4.1.1 or later be used instead."
13759
13760
13761    For questions related to the use of GCC, please consult these web
13762    pages and the [16]GCC manuals. If that fails, the
13763    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13764    web pages and the development of GCC are welcome on our developer
13765    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
13766    archives.
13767
13768   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
13769   distribution of this entire article is permitted in any medium,
13770   provided this notice is preserved.
13771
13772   These pages are [21]maintained by the GCC team. Last modified
13773   2021-07-28[22].
13774
13775References
13776
13777   1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
13778   2. https://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html
13779   3. http://gcc.gnu.org/projects/tree-ssa/
13780   4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html
13781   5. http://gcc.gnu.org/news/sms.html
13782   6. https://www.akkadia.org/drepper/dsohowto.pdf
13783   7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility
13784   8. https://itanium-cxx-abi.github.io/cxx-abi/
13785   9. http://gcc.gnu.org/fortran/
13786  10. https://gcc.gnu.org/install/
13787  11. https://gcc.gnu.org/wiki/Visibility
13788  12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1
13789  13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2
13790  14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html
13791  15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4
13792  16. https://gcc.gnu.org/onlinedocs/
13793  17. mailto:gcc-help@gcc.gnu.org
13794  18. mailto:gcc@gcc.gnu.org
13795  19. https://gcc.gnu.org/lists.html
13796  20. https://www.fsf.org/
13797  21. https://gcc.gnu.org/about.html
13798  22. http://validator.w3.org/check/referer
13799======================================================================
13800http://gcc.gnu.org/gcc-3.4/index.html
13801                             GCC 3.4 Release Series
13802
13803   (This release series is no longer supported.)
13804
13805   May 26, 2006
13806
13807   The [1]GNU project and the GCC developers are pleased to announce the
13808   release of GCC 3.4.6.
13809
13810   This release is a bug-fix release, containing fixes for regressions in
13811   GCC 3.4.4 relative to previous releases of GCC. This is the last of the
13812   3.4.x series.
13813
13814   The GCC 3.4 release series includes numerous [2]new features,
13815   improvements, bug fixes, and other changes, thanks to an [3]amazing
13816   group of volunteers.
13817
13818Release History
13819
13820   GCC 3.4.6
13821          March 6, 2006 ([4]changes)
13822
13823   GCC 3.4.5
13824          November 30, 2005 ([5]changes)
13825
13826   GCC 3.4.4
13827          May 18, 2005 ([6]changes)
13828
13829   GCC 3.4.3
13830          November 4, 2004 ([7]changes)
13831
13832   GCC 3.4.2
13833          September 6, 2004 ([8]changes)
13834
13835   GCC 3.4.1
13836          July 1, 2004 ([9]changes)
13837
13838   GCC 3.4.0
13839          April 18, 2004 ([10]changes)
13840
13841References and Acknowledgements
13842
13843   GCC used to stand for the GNU C Compiler, but since the compiler
13844   supports several other languages aside from C, it now stands for the
13845   GNU Compiler Collection.
13846
13847   A list of [11]successful builds is updated as new information becomes
13848   available.
13849
13850   The GCC developers would like to thank the numerous people that have
13851   contributed new features, improvements, bug fixes, and other changes as
13852   well as test results to GCC. This [12]amazing group of volunteers is
13853   what makes GCC successful.
13854
13855   For additional information about GCC please refer to the [13]GCC
13856   project web site or contact the [14]GCC development mailing list.
13857
13858   To obtain GCC please use [15]our mirror sites, or [16]our version
13859   control system.
13860
13861
13862    For questions related to the use of GCC, please consult these web
13863    pages and the [17]GCC manuals. If that fails, the
13864    [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13865    web pages and the development of GCC are welcome on our developer
13866    list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
13867    archives.
13868
13869   Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
13870   distribution of this entire article is permitted in any medium,
13871   provided this notice is preserved.
13872
13873   These pages are [22]maintained by the GCC team. Last modified
13874   2021-07-28[23].
13875
13876References
13877
13878   1. http://www.gnu.org/
13879   2. http://gcc.gnu.org/gcc-3.4/changes.html
13880   3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
13881   4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
13882   5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5
13883   6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4
13884   7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3
13885   8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2
13886   9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1
13887  10. http://gcc.gnu.org/gcc-3.4/changes.html
13888  11. http://gcc.gnu.org/gcc-3.4/buildstat.html
13889  12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
13890  13. http://gcc.gnu.org/index.html
13891  14. mailto:gcc@gcc.gnu.org
13892  15. http://gcc.gnu.org/mirrors.html
13893  16. http://gcc.gnu.org/git.html
13894  17. https://gcc.gnu.org/onlinedocs/
13895  18. mailto:gcc-help@gcc.gnu.org
13896  19. mailto:gcc@gcc.gnu.org
13897  20. https://gcc.gnu.org/lists.html
13898  21. https://www.fsf.org/
13899  22. https://gcc.gnu.org/about.html
13900  23. http://validator.w3.org/check/referer
13901======================================================================
13902http://gcc.gnu.org/gcc-3.4/changes.html
13903                             GCC 3.4 Release Series
13904                        Changes, New Features, and Fixes
13905
13906   The final release in the 3.4 release series is [1]GCC 3.4.6. The series
13907   is now closed.
13908
13909   GCC 3.4 has [2]many improvements in the C++ front end. Before reporting
13910   a bug, please make sure it's really GCC, and not your code, that is
13911   broken.
13912
13913Caveats
13914
13915     * GNU Make is now required to build GCC.
13916     * With -nostdinc the preprocessor used to ignore both standard
13917       include paths and include paths contained in environment variables.
13918       It was neither documented nor intended that environment variable
13919       paths be ignored, so this has been corrected.
13920     * GCC no longer accepts the options -fvolatile, -fvolatile-global and
13921       -fvolatile-static. It is unlikely that they worked correctly in any
13922       3.x release.
13923     * GCC no longer ships <varargs.h>. Use <stdarg.h> instead.
13924     * Support for all the systems [3]obsoleted in GCC 3.3 has been
13925       removed from GCC 3.4. See below for a [4]list of systems which are
13926       obsoleted in this release.
13927     * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C
13928       compilers will not work.
13929     * The implementation of the [5]MIPS ABIs has changed. As a result,
13930       the code generated for certain MIPS targets will not be binary
13931       compatible with earlier releases.
13932     * In previous releases, the MIPS port had a fake "hilo" register with
13933       the user-visible name accum. This register has been removed.
13934     * The implementation of the [6]SPARC ABIs has changed. As a result,
13935       the code generated will not be binary compatible with earlier
13936       releases in certain cases.
13937     * The configure option --enable-threads=pthreads has been removed;
13938       use --enable-threads=posix instead, which should have the same
13939       effect.
13940     * Code size estimates used by inlining heuristics for C, Objective-C,
13941       C++ and Java have been redesigned significantly. As a result the
13942       parameters of -finline-insns, --param max-inline-insns-single and
13943       --param max-inline-insns-auto need to be reconsidered.
13944     * --param max-inline-slope and --param min-inline-insns have been
13945       removed; they are not needed for the new bottom-up inlining
13946       heuristics.
13947     * The new unit-at-a-time compilation scheme has several compatibility
13948       issues:
13949          + The order in which functions, variables, and top-level asm
13950            statements are emitted may have changed. Code relying on some
13951            particular ordering needs to be updated. The majority of such
13952            top-level asm statements can be replaced by section
13953            attributes.
13954          + Unreferenced static variables and functions are removed. This
13955            may result in undefined references when an asm statement
13956            refers to the variable/function directly. In that case either
13957            the variable/function shall be listed in asm statement operand
13958            or in the case of top-level asm statements the attribute used
13959            shall be used to force function/variable to be always output
13960            and considered as a possibly used by unknown code.
13961            For variables the attribute is accepted only by GCC 3.4 and
13962            newer, while for earlier versions it is sufficient to use
13963            unused to silence warnings about the variables not being
13964            referenced. To keep code portable across different GCC
13965            versions, you can use appropriate preprocessor conditionals.
13966          + Static functions now can use non-standard passing conventions
13967            that may break asm statements calling functions directly.
13968            Again the attribute used shall be used to prevent this
13969            behavior.
13970       As a temporary workaround, -fno-unit-at-a-time can be used, but
13971       this scheme may not be supported by future releases of GCC.
13972     * GCC 3.4 automatically places zero-initialized variables in the .bss
13973       section on some operating systems. Versions of GNU Emacs up to (and
13974       including) 21.3 will not work correctly when using this
13975       optimization; you can use -fno-zero-initialized-in-bss to disable
13976       it.
13977     * If GCC 3.4 is configured with --enable-threads=posix (the default
13978       on most targets that support pthreads) then _REENTRANT will be
13979       defined unconditionally by some libstdc++ headers. C++ code which
13980       relies on that macro to detect whether multi-threaded code is being
13981       compiled might change in meaning, possibly resulting in linker
13982       errors for single-threaded programs. Affected users of [7]Boost
13983       should compile single-threaded code with -DBOOST_DISABLE_THREADS.
13984       See Bugzilla for [8]more information.
13985
13986General Optimizer Improvements
13987
13988     * Usability of the profile feedback and coverage testing has been
13989       improved.
13990          + Performance of profiled programs has been improved by faster
13991            profile merging code.
13992          + Better use of the profile feedback for optimization (loop
13993            unrolling and loop peeling).
13994          + File locking support allowing fork() calls and parallel runs
13995            of profiled programs.
13996          + Coverage file format has been redesigned.
13997          + gcov coverage tool has been improved.
13998          + make profiledbootstrap available to build a faster compiler.
13999            Experiments made on i386 hardware showed an 11% speedup on -O0
14000            and a 7.5% speedup on -O2 compilation of a [9]large C++
14001            testcase.
14002          + New value profiling pass enabled via -fprofile-values
14003          + New value profile transformations pass enabled via -fvpt aims
14004            to optimize some code sequences by exploiting knowledge about
14005            value ranges or other properties of the operands. At the
14006            moment a conversion of expensive divisions into cheaper
14007            operations has been implemented.
14008          + New -fprofile-generate and -fprofile-use command-line options
14009            to simplify the use of profile feedback.
14010     * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and
14011       Java which is enabled via -funit-at-a-time (and implied by -O2). In
14012       this scheme a whole file is parsed first and optimized later. The
14013       following basic inter-procedural optimizations are implemented:
14014          + Removal of unreachable functions and variables
14015          + Discovery of local functions (functions with static linkage
14016            whose address is never taken)
14017          + On i386, these local functions use register parameter passing
14018            conventions.
14019          + Reordering of functions in topological order of the call graph
14020            to enable better propagation of optimizing hints (such as the
14021            stack alignments needed by functions) in the back end.
14022          + Call graph based out-of-order inlining heuristics which allows
14023            to limit overall compilation unit growth (--param
14024            inline-unit-growth).
14025       Overall, the unit-at-a-time scheme produces a 1.3% improvement for
14026       the SPECint2000 benchmark on the i386 architecture (AMD Athlon
14027       CPU).
14028     * More realistic code size estimates used by inlining for C,
14029       Objective-C, C++ and Java. The growth of large functions can now be
14030       limited via --param large-function-insns and --param
14031       large-function-growth.
14032     * A new cfg-level loop optimizer pass replaces the old loop unrolling
14033       pass and adds two other loop transformations -- loop peeling and
14034       loop unswitching -- and also uses the profile feedback to limit
14035       code growth. (The three optimizations are enabled by
14036       -funroll-loops, -fpeel-loops and -funswitch-loops flags,
14037       respectively).
14038       The old loop unroller still can be enabled by -fold-unroll-loops
14039       and may produce better code in some cases, especially when the
14040       webizer optimization pass is not run.
14041     * A new web construction pass enabled via -fweb (and implied by -O3)
14042       improves the quality of register allocation, CSE, first scheduling
14043       pass and some other optimization passes by avoiding re-use of
14044       pseudo registers with non-overlapping live ranges. The pass almost
14045       always improves code quality but does make debugging difficult and
14046       thus is not enabled by default by -O2
14047       The pass is especially effective as cleanup after code duplication
14048       passes, such as the loop unroller or the tracer.
14049     * Experimental implementations of superblock or trace scheduling in
14050       the second scheduling pass can be enabled via
14051       -fsched2-use-superblocks and -fsched2-use-traces, respectively.
14052
14053New Languages and Language specific improvements
14054
14055  Ada
14056
14057     * The Ada front end has been updated to include numerous bug fixes
14058       and enhancements. These include:
14059          + Improved project file support
14060          + Additional set of warnings about potential wrong code
14061          + Improved error messages
14062          + Improved code generation
14063          + Improved cross reference information
14064          + Improved inlining
14065          + Better run-time check elimination
14066          + Better error recovery
14067          + More efficient implementation of unbounded strings
14068          + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools,
14069            ...
14070          + New GNAT.xxxx packages (e.g. GNAT.Strings,
14071            GNAT.Exception_Action)
14072          + New pragmas
14073          + New -gnatS switch replacing gnatpsta
14074          + Implementation of new Ada features (in particular limited
14075            with, limited aggregates)
14076
14077  C/Objective-C/C++
14078
14079     * Precompiled headers are now supported. Precompiled headers can
14080       dramatically speed up compilation of some projects. There are some
14081       known defects in the current precompiled header implementation that
14082       will result in compiler crashes in relatively rare situations.
14083       Therefore, precompiled headers should be considered a "technology
14084       preview" in this release. Read the manual for details about how to
14085       use precompiled headers.
14086     * File handling in the preprocessor has been rewritten. GCC no longer
14087       gets confused by symlinks and hardlinks, and now has a correct
14088       implementation of #import and #pragma once. These two directives
14089       have therefore been un-deprecated.
14090     * The undocumented extension that allowed C programs to have a label
14091       at the end of a compound statement, which has been deprecated since
14092       GCC 3.0, has been removed.
14093     * The cast-as-lvalue extension has been removed for C++ and
14094       deprecated for C and Objective-C. In particular, code like this:
14095        int i;
14096        (char) i = 5;
14097
14098       or this:
14099        char *p;
14100        ((int *) p)++;
14101
14102       is no longer accepted for C++ and will not be accepted for C and
14103       Objective-C in a future version.
14104     * The conditional-expression-as-lvalue extension has been deprecated
14105       for C and Objective-C. In particular, code like this:
14106        int a, b, c;
14107        (a ? b : c) = 2;
14108
14109       will not be accepted for C and Objective-C in a future version.
14110     * The compound-expression-as-lvalue extension has been deprecated for
14111       C and Objective-C. In particular, code like this:
14112        int a, b;
14113        (a, b) = 2;
14114
14115       will not be accepted for C and Objective-C in a future version. A
14116       possible non-intrusive workaround is the following:
14117        (*(a, &b)) = 2;
14118
14119     * Several [10]built-in functions such as __builtin_popcount for
14120       counting bits, finding the highest and lowest bit in a word, and
14121       parity have been added.
14122     * The -fwritable-strings option has been deprecated and will be
14123       removed.
14124     * Many C math library functions are now recognized as built-ins and
14125       optimized.
14126     * The C, C++, and Objective-C compilers can now handle source files
14127       written in any character encoding supported by the host C library.
14128       The default input character set is taken from the current locale,
14129       and may be overridden with the -finput-charset command line option.
14130       In the future we will add support for inline encoding markers.
14131
14132  C++
14133
14134     * G++ is now much closer to full conformance to the ISO/ANSI C++
14135       standard. This means, among other things, that a lot of invalid
14136       constructs which used to be accepted in previous versions will now
14137       be rejected. It is very likely that existing C++ code will need to
14138       be fixed. This document lists some of the most common issues.
14139     * A hand-written recursive-descent C++ parser has replaced the
14140       YACC-derived C++ parser from previous GCC releases. The new parser
14141       contains much improved infrastructure needed for better parsing of
14142       C++ source codes, handling of extensions, and clean separation
14143       (where possible) between proper semantics analysis and parsing. The
14144       new parser fixes many bugs that were found in the old parser.
14145     * You must now use the typename and template keywords to disambiguate
14146       dependent names, as required by the C++ standard.
14147        struct K {
14148          typedef int mytype_t;
14149        };
14150
14151        template <class T1> struct A {
14152          template <class T2> struct B {
14153              void callme(void);
14154            };
14155
14156          template <int N> void bar(void)
14157          {
14158            // Use 'typename' to tell the parser that T1::mytype_t names
14159            //  a type. This is needed because the name is dependent (in
14160            //  this case, on template parameter T1).
14161            typename T1::mytype_t x;
14162            x = 0;
14163          }
14164        };
14165
14166        template <class T> void template_func(void)
14167        {
14168          // Use 'template' to prefix member templates within
14169          //  dependent types (a has type A<T>, which depends on
14170          //  the template parameter T).
14171          A<T> a;
14172          a.template bar<0>();
14173
14174          // Use 'template' to tell the parser that B is a nested
14175          //  template class (dependent on template parameter T), and
14176          //  'typename' because the whole A<T>::B<int> is
14177          //  the name of a type (again, dependent).
14178          typename A<T>::template B<int> b;
14179          b.callme();
14180        }
14181
14182        void non_template_func(void)
14183        {
14184          // Outside of any template class or function, no names can be
14185          //  dependent, so the use of the keyword 'typename' and 'template'
14186          //  is not needed (and actually forbidden).
14187          A<K> a;
14188          a.bar<0>();
14189          A<K>::B<float> b;
14190          b.callme();
14191        }
14192     * In a template definition, unqualified names will no longer find
14193       members of a dependent base (as specified by [temp.dep]/3 in the
14194       C++ standard). For example,
14195        template <typename T> struct B {
14196          int m;
14197          int n;
14198          int f ();
14199          int g ();
14200        };
14201        int n;
14202        int g ();
14203        template <typename T> struct C : B<T> {
14204          void h ()
14205          {
14206            m = 0; // error
14207            f ();  // error
14208            n = 0; // ::n is modified
14209            g ();  // ::g is called
14210          }
14211        };
14212       You must make the names dependent, e.g. by prefixing them with
14213       this->. Here is the corrected definition of C<T>::h,
14214        template <typename T> void C<T>::h ()
14215        {
14216          this->m = 0;
14217          this->f ();
14218          this->n = 0
14219          this->g ();
14220        }
14221       As an alternative solution (unfortunately not backwards compatible
14222       with GCC 3.3), you may use using declarations instead of this->:
14223        template <typename T> struct C : B<T> {
14224          using B<T>::m;
14225          using B<T>::f;
14226          using B<T>::n;
14227          using B<T>::g;
14228          void h ()
14229          {
14230            m = 0;
14231            f ();
14232            n = 0;
14233            g ();
14234          }
14235        };
14236     * In templates, all non-dependent names are now looked up and bound
14237       at definition time (while parsing the code), instead of later when
14238       the template is instantiated. For instance:
14239        void foo(int);
14240
14241        template <int> struct A {
14242          static void bar(void){
14243            foo('a');
14244          }
14245        };
14246
14247        void foo(char);
14248
14249        int main()
14250        {
14251          A<0>::bar();    // Calls foo(int), used to call foo(char).
14252        }
14253
14254     * In an explicit instantiation of a class template, you must use
14255       class or struct before the template-id:
14256        template <int N>
14257        class A {};
14258
14259        template A<0>;         // error, not accepted anymore
14260        template class A<0>;   // OK
14261     * The "named return value" and "implicit typename" extensions have
14262       been removed.
14263     * Default arguments in function types have been deprecated and will
14264       be removed.
14265     * ARM-style name-injection of friend declarations has been deprecated
14266       and will be removed. For example: struct S { friend void f(); };
14267       void g() { f(); } will not be accepted by future versions of G++;
14268       instead a declaration of "f" will need to be present outside of the
14269       scope of "S".
14270     * Covariant returns are implemented for all but varadic functions
14271       that require an adjustment.
14272     * When -pedantic is used, G++ now issues errors about spurious
14273       semicolons. For example,
14274        namespace N {}; // Invalid semicolon.
14275        void f() {}; // Invalid semicolon.
14276     * G++ no longer accepts attributes for a declarator after the
14277       initializer associated with that declarator. For example,
14278        X x(1) __attribute__((...));
14279       is no longer accepted. Instead, use:
14280        X x __attribute__((...)) (1);
14281     * Inside the scope of a template class, the name of the class itself
14282       can be treated as either a class or a template. So GCC used to
14283       accept the class name as argument of type template, and template
14284       template parameter. However this is not C++ standard compliant. Now
14285       the name is not treated as a valid template template argument
14286       unless you qualify the name by its scope. For example, the code
14287       below no longer compiles.
14288        template <template <class> class TT> class X {};
14289        template <class T> class Y {
14290          X<Y> x; // Invalid, Y is always a type template parameter.
14291        };
14292       The valid code for the above example is
14293          X< ::Y> x; // Valid.
14294       (Notice the space between < and : to prevent GCC to interpret this
14295       as a digraph for [.)
14296     * Friend declarations that refer to template specializations are
14297       rejected if the template has not already been declared. For
14298       example,
14299        template <typename T>
14300        class C {
14301          friend void f<> (C&);
14302        };
14303       is rejected. You must first declare f as a template,
14304        template <typename T>
14305        void f(T);
14306     * In case of friend declarations, every name used in the friend
14307       declaration must be accessible at the point of that declaration.
14308       Previous versions of G++ used to be less strict about this and
14309       allowed friend declarations for private class members, for example.
14310       See the ISO C++ Standard Committee's [11]defect report #209 for
14311       details.
14312     * Declaration of member functions of class templates as friends are
14313       supported. For example,
14314        template <typename T> struct A {
14315          void f();
14316        };
14317        class C {
14318          template <typename T> friend void A<T>::f();
14319        };
14320     * You must use template <> to introduce template specializations, as
14321       required by the standard. For example,
14322        template <typename T>
14323        struct S;
14324
14325        struct S<int> { };
14326       is rejected. You must write,
14327        template <> struct S<int> {};
14328     * G++ used to accept code like this,
14329        struct S {
14330          int h();
14331          void f(int i = g());
14332          int g(int i = h());
14333        };
14334       This behavior is not mandated by the standard. Now G++ issues an
14335       error about this code. To avoid the error, you must move the
14336       declaration of g before the declaration of f. The default arguments
14337       for g must be visible at the point where it is called.
14338     * The C++ ABI Section 3.3.3 specifications for the array construction
14339       routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return
14340       NULL when the allocator argument returns NULL. These changes are
14341       incorporated into the libstdc++ runtime library.
14342     * Using a name introduced by a typedef in a friend declaration or in
14343       an explicit instantiation is now rejected, as specified by the ISO
14344       C++ standard.
14345        class A;
14346        typedef A B;
14347        class C {
14348          friend class B;      // error, no typedef name here
14349          friend B;            // error, friend always needs class/struct/enum
14350          friend class A;      // OK
14351        };
14352
14353        template <int> class Q {};
14354        typedef Q<0> R;
14355        template class R;      // error, no typedef name here
14356        template class Q<0>;   // OK
14357     * When allocating an array with a new expression, GCC used to allow
14358       parentheses around the type name. This is actually ill-formed and
14359       it is now rejected:
14360        int* a = new (int)[10];    // error, not accepted anymore
14361        int* a = new int[10];      // OK
14362     * When binding an rvalue of class type to a reference, the copy
14363       constructor of the class must be accessible. For instance, consider
14364       the following code:
14365        class A
14366        {
14367        public:
14368          A();
14369
14370        private:
14371          A(const A&);   // private copy ctor
14372        };
14373
14374        A makeA(void);
14375        void foo(const A&);
14376
14377        void bar(void)
14378        {
14379          foo(A());       // error, copy ctor is not accessible
14380          foo(makeA());   // error, copy ctor is not accessible
14381
14382          A a1;
14383          foo(a1);        // OK, a1 is a lvalue
14384        }
14385       This might be surprising at first sight, especially since most
14386       popular compilers do not correctly implement this rule ([12]further
14387       details).
14388     * When forming a pointer to member or a pointer to member function,
14389       access checks for class visibility (public, protected, private) are
14390       now performed using the qualifying scope of the name itself. This
14391       is better explained with an example:
14392        class A
14393        {
14394        public:
14395          void pub_func();
14396        protected:
14397          void prot_func();
14398        private:
14399          void priv_func();
14400        };
14401
14402        class B : public A
14403        {
14404        public:
14405          void foo()
14406          {
14407            &A::pub_func;   // OK, pub_func is accessible through A
14408            &A::prot_func;  // error, cannot access prot_func through A
14409            &A::priv_func;  // error, cannot access priv_func through A
14410
14411            &B::pub_func;   // OK, pub_func is accessible through B
14412            &B::prot_func;  // OK, can access prot_func through B (within B)
14413            &B::priv_func;  // error, cannot access priv_func through B
14414          }
14415        };
14416
14417    Runtime Library (libstdc++)
14418
14419     * Optimization work:
14420          + Streamlined streambuf, filebuf, separate synched with C
14421            Standard I/O streambuf.
14422          + All formatted I/O now uses cached locale information.
14423          + STL optimizations (memory/speed for list, red-black trees as
14424            used by sets and maps).
14425          + More use of GCC builtins.
14426          + String optimizations (avoid contention on
14427            increment/decrement-and-test of the reference count in the
14428            empty-string object, constructor from input_iterators
14429            speedup).
14430     * Static linkage size reductions.
14431     * Large File Support (files larger than 2 GB on 32-bit systems).
14432     * Wide character and variable encoding filebuf work (UTF-8, Unicode).
14433     * Generic character traits.
14434     * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x,
14435       Solaris 2.7 and above, AIX 5.x, Irix 6.5.
14436     * The allocator class is now standard-conformant, and two additional
14437       extension allocators have been added, mt_alloc and
14438       bitmap_allocator.
14439     * PCH support: -include bits/stdc++.h (2x compile speedup).
14440     * Rewrote __cxa_demangle with support for C++ style allocators.
14441     * New debug modes for STL containers and iterators.
14442     * Testsuite rewrite: five times as many tests, plus increasingly
14443       sophisticated tests, including I/O, MT, multi-locale, wide and
14444       narrow characters.
14445     * Use current versions of GNU "autotools" for build/configuration.
14446
14447  Objective-C
14448
14449     * The Objective-C front end has been updated to include the numerous
14450       bug fixes and enhancements previously available only in Apple's
14451       version of GCC. These include:
14452          + Structured exception (@try... @catch... @finally, @throw) and
14453            synchronization (@synchronized) support. These are accessible
14454            via the -fobjc-exceptions switch; as of this writing, they may
14455            only be used in conjunction with -fnext-runtime on Mac OS X
14456            10.3 and later. See [13]Options Controlling Objective-C
14457            Dialect for more information.
14458          + An overhaul of @encode logic. The C99 _Bool and C++ bool type
14459            may now be encoded as 'B'. In addition, the back-end/codegen
14460            dependencies have been removed.
14461          + An overhaul of message dispatch construction, ensuring that
14462            the various receiver types (and casts thereof) are handled
14463            properly, and that correct diagnostics are issued.
14464          + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue"
14465            (-freplace-objc-classes) debugging modes, currently available
14466            on Mac OS X 10.3 and later. See [14]Options Controlling
14467            Objective-C Dialect for more information.
14468          + Access to optimized runtime entry points (-fno-nil-receivers )
14469            on the assumption that message receivers are never nil. This
14470            is currently available on Mac OS X 10.3 and later. See
14471            [15]Options Controlling Objective-C Dialect for more
14472            information.
14473
14474  Java
14475
14476     * Compiling a .jar file will now cause non-.class entries to be
14477       automatically compiled as resources.
14478     * libgcj has been ported to Darwin.
14479     * Jeff Sturm has adapted Jan Hubicka's call graph optimization code
14480       to gcj.
14481     * libgcj has a new gcjlib URL type; this lets URLClassLoader load
14482       code from shared libraries.
14483     * libgcj has been much more completely merged with [16]GNU Classpath.
14484     * Class loading is now much more correct; in particular the caller's
14485       class loader is now used when that is required.
14486     * [17]Eclipse 2.x will run out of the box using gij.
14487     * Parts of java.nio have been implemented. Direct and indirect
14488       buffers work, as do fundamental file and socket operations.
14489     * java.awt has been improved, though it is still not ready for
14490       general use.
14491     * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST
14492       method.
14493     * The MinGW port has matured. Enhancements include socket timeout
14494       support, thread interruption, improved Runtime.exec() handling and
14495       support for accented characters in filenames.
14496
14497  Fortran
14498
14499     * Fortran improvements are listed in the [18]Fortran documentation.
14500
14501New Targets and Target Specific Improvements
14502
14503  Alpha
14504
14505     * Several [19]built-in functions have been added such as
14506       __builtin_alpha_zap to allow utilizing the more obscure
14507       instructions of the CPU.
14508     * Parameter passing of complex arguments has changed to match the
14509       ABI. This change is incompatible with previous GCC versions, but
14510       does fix compatibility with the Tru64 compiler and several corner
14511       cases where GCC was incompatible with itself.
14512
14513  ARM
14514
14515     * Nicolas Pitre has contributed his hand-coded floating-point support
14516       code for ARM. It is both significantly smaller and faster than the
14517       existing C-based implementation, even when building applications
14518       for Thumb. The arm-elf configuration has been converted to use the
14519       new code.
14520     * Support for the Intel's iWMMXt architecture, a second generation
14521       XScale processor, has been added. Enabled at run time with the
14522       -mcpu=iwmmxt command line switch.
14523     * A new ARM target has been added: arm-wince-pe. This is similar to
14524       the arm-pe target, but it defaults to using the APCS32 ABI.
14525     * The existing ARM pipeline description has been converted to the use
14526       the [20]DFA processor pipeline model. There is not much change in
14527       code performance, but the description is now [21]easier to
14528       understand.
14529     * Support for the Cirrus EP9312 Maverick floating point co-processor
14530       added. Enabled at run time with the -mcpu=ep9312 command line
14531       switch. Note however that the multilibs to support this chip are
14532       currently disabled in gcc/config/arm/t-arm-elf, so if you want to
14533       enable their production you will have to uncomment the entries in
14534       that file.
14535
14536  H8/300
14537
14538     * Support for long long has been added.
14539     * Support for saveall attribute has been added.
14540     * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code
14541       for H8/300H and H8S, which is much faster than the previous
14542       implementation.
14543     * A lot of small performance improvements.
14544
14545  IA-32/AMD64 (x86-64)
14546
14547     * Tuning for K8 (AMD Opteron/Athlon64) core is available via
14548       -march=k8 and -mcpu=k8.
14549     * Scalar SSE code generation carefully avoids reformatting penalties,
14550       hidden dependencies and minimizes the number of uops generated on
14551       both Intel and AMD CPUs.
14552     * Vector MMX and SSE operands are now passed in registers to improve
14553       performance and match the argument passing convention used by the
14554       Intel C++ Compiler. As a result it is not possible to call
14555       functions accepting vector arguments compiled by older GCC version.
14556     * Conditional jump elimination is now more aggressive on modern CPUs.
14557     * The Athlon ports has been converted to use the DFA processor
14558       pipeline description.
14559     * Optimization of indirect tail calls is now possible in a similar
14560       fashion as direct sibcall optimization.
14561     * Further small performance improvements.
14562     * -m128bit-long-double is now less buggy.
14563     * __float128 support in 64-bit compilation.
14564     * Support for data structures exceeding 2GB in 64-bit mode.
14565     * -mcpu has been renamed to -mtune.
14566
14567  IA-64
14568
14569     * Tuning code for the Itanium 2 processor has been added. The
14570       generation of code tuned for Itanium 2 (option -mtune=itanium2) is
14571       enabled by default now. To generate code tuned for Itanium 1 the
14572       option -mtune=itanium1 should be used.
14573     * [22]DFA processor pipeline descriptions for the IA-64 processors
14574       have been added. This resulted in about 3% improvement on the
14575       SPECInt2000 benchmark for Itanium 2.
14576     * Instruction bundling for the IA-64 processors has been rewritten
14577       using the DFA pipeline hazard recognizer. It resulted in about 60%
14578       compiler speedup on the SPECInt2000 C programs.
14579
14580  M32R
14581
14582     * Support for the M32R/2 processor has been added by Renesas.
14583     * Support for an M32R GNU/Linux target and PIC code generation has
14584       been added by Renesas.
14585
14586  M68000
14587
14588     * Bernardo Innocenti (Develer S.r.l.) has contributed the
14589       m68k-uclinux target, based on former work done by Paul Dale
14590       (SnapGear Inc.). Code generation for the ColdFire processors family
14591       has been enhanced and extended to support the MCF 53xx and MCF 54xx
14592       cores, integrating former work done by Peter Barada (Motorola).
14593
14594  MIPS
14595
14596    Processor-specific changes
14597
14598     * Support for the RM7000 and RM9000 processors has been added. It can
14599       be selected using the -march compiler option and should work with
14600       any MIPS I (mips-*) or MIPS III (mips64-*) configuration.
14601     * Support for revision 2 of the MIPS32 ISA has been added. It can be
14602       selected with the command-line option -march=mips32r2.
14603     * There is a new option, -mfix-sb1, to work around certain SB-1
14604       errata.
14605
14606    Configuration
14607
14608     * It is possible to customize GCC using the following configure-time
14609       options:
14610          + --with-arch, which specifies the default value of the -march
14611            option.
14612          + --with-tune, which specifies the default value of the -mtune
14613            option.
14614          + --with-abi, which specifies the default ABI.
14615          + --with-float=soft, which tells GCC to use software floating
14616            point by default.
14617          + --with-float=hard, which tells GCC to use hardware floating
14618            point by default.
14619     * A 64-bit GNU/Linux port has been added. The associated
14620       configurations are mips64-linux-gnu and mips64el-linux-gnu.
14621     * The 32-bit GNU/Linux port now supports Java.
14622     * The IRIX 6 configuration now supports the o32 ABI and will build
14623       o32 multilibs by default. This support is compatible with both
14624       binutils and the SGI tools, but note that several features,
14625       including debugging information and DWARF2 exception handling, are
14626       only available when using the GNU assembler. Use of the GNU
14627       assembler and linker (version 2.15 or above) is strongly
14628       recommended.
14629     * The IRIX 6 configuration now supports 128-bit long doubles.
14630     * There are two new RTEMS-specific configurations, mips-rtems and
14631       mipsel-rtems.
14632     * There are two new *-elf configurations, mipsisa32r2-elf and
14633       mipsisa32r2el-elf.
14634
14635    General
14636
14637     * Several [23]ABI bugs have been fixed. Unfortunately, these changes
14638       will break binary compatibility with earlier releases.
14639     * GCC can now use explicit relocation operators when generating
14640       -mabicalls code. This behavior is controlled by -mexplicit-relocs
14641       and can have several performance benefits. For example:
14642          + It allows for more optimization of GOT accesses, including
14643            better scheduling and redundancy elimination.
14644          + It allows sibling calls to be implemented as jumps.
14645          + n32 and n64 leaf functions can use a call-clobbered global
14646            pointer instead of $28.
14647          + The code to set up $gp can be removed from functions that
14648            don't need it.
14649     * A new option, -mxgot, allows the GOT to be bigger than 64k. This
14650       option is equivalent to the assembler's -xgot option and should be
14651       used instead of -Wa,-xgot.
14652     * Frame pointer elimination is now supported when generating 64-bit
14653       MIPS16 code.
14654     * Inline block moves have been optimized to take more account of
14655       alignment information.
14656     * Many internal changes have been made to the MIPS port, mostly aimed
14657       at reducing the reliance on assembler macros.
14658
14659  PowerPC
14660
14661     * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64
14662       [24]ABI incompatibilities regarding the way parameters are passed
14663       during functions calls. These changes may result in incompatibility
14664       between code compiled with GCC 3.3 and GCC 3.4.
14665
14666    PowerPC Darwin
14667
14668     * Support for shared/dylib gcc libraries has been added. It is
14669       enabled by default on powerpc-apple-darwin7.0.0 and up.
14670     * Libgcj is enabled by default. On systems older than
14671       powerpc-apple-darwin7.0.0 you need to install dlcompat.
14672     * 128-bit IBM extended precision format support added for long
14673       double.
14674
14675    PowerPC64 GNU/Linux
14676
14677     * By default, PowerPC64 GNU/Linux now uses natural alignment of
14678       structure elements. The old four byte alignment for double, with
14679       special rules for a struct starting with a double, can be chosen
14680       with -malign-power. This change may result in incompatibility
14681       between code compiled with GCC 3.3 and GCC 3.4.
14682     * -mabi=altivec is now the default rather than -mabi=no-altivec.
14683     * 128-bit IBM extended precision format support added for long
14684       double.
14685
14686  S/390 and zSeries
14687
14688     * New command-line options allow to specify the intended execution
14689       environment for generated code:
14690          + -mesa/-mzarch allows to specify whether to generate code
14691            running in ESA/390 mode or in z/Architecture mode (this is
14692            applicable to 31-bit code only).
14693          + -march allows to specify a minimum processor architecture
14694            level (g5, g6, z900, or z990).
14695          + -mtune allows to specify which processor to tune for.
14696     * It is possible to customize GCC using the following configure-time
14697       options:
14698          + --with-mode, which specifies whether to default to assuming
14699            ESA/390 or z/Architecture mode.
14700          + --with-arch, which specifies the default value of the -march
14701            option.
14702          + --with-tune, which specifies the default value of the -mtune
14703            option.
14704     * Support for the z990 processor has been added, and can be selected
14705       using -march=z990 or -mtune=z990. This includes instruction
14706       scheduling tuned for the superscalar instruction pipeline of the
14707       z990 processor as well as support for all new instructions provided
14708       by the long-displacement facility.
14709     * Support to generate 31-bit code optimized for zSeries processors
14710       (running in ESA/390 or in z/Architecture mode) has been added. This
14711       can be selected using -march=z900 and -mzarch respectively.
14712     * Instruction scheduling for the z900 and z990 processors now uses
14713       the DFA pipeline hazard recognizer.
14714     * GCC no longer generates code to maintain a stack backchain,
14715       previously used to generate stack backtraces for debugging
14716       purposes. As replacement that does not incur runtime overhead,
14717       DWARF-2 call frame information is provided by GCC; this is
14718       supported by GDB 6.1. The old behavior can be restored using the
14719       -mbackchain option.
14720     * The stack frame size of functions may now exceed 2 GB in 64-bit
14721       code.
14722     * A port for the 64-bit IBM TPF operating system has been added; the
14723       configuration is s390x-ibm-tpf. This configuration is supported as
14724       cross-compilation target only.
14725     * Various changes to improve the generated code have been
14726       implemented, including:
14727          + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT
14728            instructions to significantly speed up many floating-point
14729            applications.
14730          + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL
14731            WITH BORROW instructions to speed up long long arithmetic.
14732          + GCC now uses the SEARCH STRING instruction to implement
14733            strlen().
14734          + In many cases, function call overhead for 31-bit code has been
14735            reduced by placing the literal pool after the function code
14736            instead of after the function prolog.
14737          + Register 14 is no longer reserved in 64-bit code.
14738          + Handling of global register variables has been improved.
14739
14740  SPARC
14741
14742     * The option -mflat is deprecated.
14743     * Support for large (> 2GB) frames has been added to the 64-bit port.
14744     * Several [25]ABI bugs have been fixed. Unfortunately, these changes
14745       will break binary compatibility with earlier releases.
14746     * The default debugging format has been switched from STABS to
14747       DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already
14748       the default debugging format for 64-bit code on Solaris.
14749
14750  SuperH
14751
14752     * Support for the SH2E processor has been added. Enabled at run time
14753       with the -m2e command line switch, or at configure time by
14754       specifying sh2e as the machine part of the target triple.
14755
14756  V850
14757
14758     * Support for the Mitsubishi V850E1 processor has been added. This is
14759       a variant of the V850E processor with some additional debugging
14760       instructions.
14761
14762  Xtensa
14763
14764     * Several ABI bugs have been fixed. Unfortunately, these changes
14765       break binary compatibility with earlier releases.
14766          + For big-endian processors, the padding of aggregate return
14767            values larger than a word has changed. If the size of an
14768            aggregate return value is not a multiple of 32 bits, previous
14769            versions of GCC inserted padding in the most-significant bytes
14770            of the first return value register. Aggregates larger than a
14771            word are now padded in the least-significant bytes of the last
14772            return value register used. Aggregates smaller than a word are
14773            still padded in the most-significant bytes. The return value
14774            padding has not changed for little-endian processors.
14775          + Function arguments with 16-byte alignment are now properly
14776            aligned.
14777          + The implementation of the va_list type has changed. A va_list
14778            value created by va_start from a previous release cannot be
14779            used with va_arg from this release, or vice versa.
14780     * More processor configuration options for Xtensa processors are
14781       supported:
14782          + the ABS instruction is now optional;
14783          + the ADDX* and SUBX* instructions are now optional;
14784          + an experimental CONST16 instruction can be used to synthesize
14785            constants instead of loading them from constant pools.
14786       These and other Xtensa processor configuration options can no
14787       longer be enabled or disabled by command-line options; the
14788       processor configuration must be specified by the xtensa-config.h
14789       header file when building GCC. Additionally, the
14790       -mno-serialize-volatile option is no longer supported.
14791
14792Obsolete Systems
14793
14794   Support for a number of older systems has been declared obsolete in GCC
14795   3.4. Unless there is activity to revive them, the next release of GCC
14796   will have their sources permanently removed.
14797
14798   All configurations of the following processor architectures have been
14799   declared obsolete:
14800     * Mitsubishi D30V, d30v-*
14801     * AT&T DSP1600 and DSP1610, dsp16xx-*
14802     * Intel 80960, i960
14803
14804   Also, some individual systems have been obsoleted:
14805     * ARM Family
14806          + Support for generating code for operation in APCS/26 mode
14807            (-mapcs-26).
14808     * IBM ESA/390
14809          + "Bigfoot" port, i370-*. (The other port, s390-*, is actively
14810            maintained and supported.)
14811     * Intel 386 family
14812          + MOSS, i?86-moss-msdos and i?86-*-moss*
14813          + NCR 3000 running System V r.4, i?86-ncr-sysv4*
14814          + FreeBSD with a.out object format, i?86-*-freebsd*aout* and
14815            i?86-*-freebsd2*
14816          + GNU/Linux with a.out object format, i?86-linux*aout*
14817          + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1*
14818          + Interix versions before Interix 3, i?86-*-interix
14819          + Mach microkernel, i?86-mach*
14820          + SCO UnixWare with UDK, i?86-*-udk*
14821          + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]*
14822          + VSTa microkernel, i386-*-vsta
14823     * Motorola M68000 family
14824          + HPUX, m68k-hp-hpux* and m68000-hp-hpux*
14825          + NetBSD with a.out object format (before NetBSD 1.4),
14826            m68k-*-*-netbsd* except m68k-*-*-netbsdelf*
14827          + Generic System V r.4, m68k-*-sysv4*
14828     * VAX
14829          + Generic VAX, vax-*-* (This is generic VAX only; we have not
14830            obsoleted any VAX triples for specific operating systems.)
14831
14832Documentation improvements
14833
14834Other significant improvements
14835
14836     * The build system has undergone several significant cleanups.
14837       Subdirectories will only be configured if they are being built, and
14838       all subdirectory configures are run from the make command. The top
14839       level has been autoconfiscated.
14840     * Building GCC no longer writes to its source directory. This should
14841       help those wishing to share a read-only source directory over NFS
14842       or build from a CD. The exceptions to this feature are if you
14843       configure with either --enable-maintainer-mode or
14844       --enable-generated-files-in-srcdir.
14845     * The -W warning option has been renamed to -Wextra, which is more
14846       easily understood. The older spelling will be retained for
14847       backwards compatibility.
14848     * Substantial improvements in compile time have been made,
14849       particularly for non-optimizing compilations.
14850     __________________________________________________________________
14851
14852GCC 3.4.0
14853
14854  Bug Fixes
14855
14856   A vast number of bugs have been fixed in 3.4.0, too many to publish a
14857   complete list here. [26]Follow this link to query the Bugzilla database
14858   for the list of over 900 bugs fixed in 3.4.0. This is the list of all
14859   bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4
14860   regressions.
14861     __________________________________________________________________
14862
14863GCC 3.4.1
14864
14865  Bug Fixes
14866
14867   This section lists the problem reports (PRs) from GCC's bug tracking
14868   system that are known to be fixed in the 3.4.1 release. This list might
14869   not be complete (that is, it is possible that some PRs that have been
14870   fixed are not listed here).
14871
14872    Bootstrap failures
14873
14874     * [27]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler
14875       emitted - PIC related
14876     * [28]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf
14877     * [29]14760 A bug in configure.in prevents using both
14878       --program-suffix and --program-prefix
14879     * [30]14671 [hppa64] bootstrap fails: ICE in
14880       save_call_clobbered_regs, in caller_save.c
14881     * [31]15093 [alpha][Java] make bootstrap fails to configure libffi on
14882       Alpha
14883     * [32]15178 Solaris 9/x86 fails linking after stage 3
14884
14885    Multi-platform internal compiler errors (ICEs)
14886
14887     * [33]12753 (preprocessor) Memory corruption in preprocessor on bad
14888       input
14889     * [34]13985 ICE in gcc.c-torture/compile/930621-1.c
14890     * [35]14810 (c++) tree check failures with invalid code involving
14891       templates
14892     * [36]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in
14893       cp/parser.c
14894     * [37]15044 (c++) ICE on syntax error, template header
14895     * [38]15057 (c++) Compiling of conditional value throw constructs
14896       cause a segmentation violation
14897     * [39]15064 (c++) typeid of template parameter gives ICE
14898     * [40]15142 (c++) ICE when passing a string where a char* is expected
14899       in a throw statement
14900     * [41]15159 ICE in rtl_verify_flow_info_1
14901     * [42]15165 (c++) ICE in instantiate_template
14902     * [43]15193 Unary minus using pointer to V4SF vector causes
14903       -fforce-mem to exhaust all memory
14904     * [44]15209 (c++) Runs out of memory with packed structs
14905     * [45]15227 (c++) Trouble with invalid function definition
14906     * [46]15285 (c++) instantiate_type ICE when forming pointer to
14907       template function
14908     * [47]15299 (c++) ICE in resolve_overloaded_unification
14909     * [48]15329 (c++) ICE on constructor of member template
14910     * [49]15550 ICE in extract_insn, in recog.c
14911     * [50]15554 (c++) ICE in tsubst_copy, in cp/pt.c
14912     * [51]15640 (c++) ICE on invalid code in arg_assoc, in
14913       cp/name-lookup.c
14914     * [52]15666 [unit-at-a-time] Gcc abort on valid code
14915     * [53]15696 (c++) ICE with bad pointer-to-member code
14916     * [54]15701 (c++) ICE with friends and template template parameter
14917     * [55]15761 ICE in do_SUBST, in combine.c
14918     * [56]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops
14919
14920    Ada
14921
14922     * [57]14538 All RTEMS targets broken for gnat
14923
14924    C front end
14925
14926     * [58]12391 missing warning about assigning to an incomplete type
14927     * [59]14649 atan(1.0) should not be a constant expression
14928     * [60]15004 [unit-at-a-time] no warning for unused paramater in
14929       static function
14930     * [61]15749 --pedantic-errors behaves differently from --pedantic
14931       with C-compiler on GNU/Linux
14932
14933    C++ compiler and library
14934
14935     * [62]10646 non-const reference is incorrectly matched in a "const T"
14936       partial specialization
14937     * [63]12077 wcin.rdbuf()->in_avail() return value too high
14938     * [64]13598 enc_filebuf doesn't work
14939     * [65]14211 const_cast returns lvalue but should be rvalue
14940     * [66]14220 num_put::do_put() undesired float/double behavior
14941     * [67]14245 problem with user-defined allocators in std::basic_string
14942     * [68]14340 libstdc++ Debug mode: failure to convert iterator to
14943       const_iterator
14944     * [69]14600 __gnu_cxx::stdio_sync_filebuf should expose internal
14945       FILE*
14946     * [70]14668 no warning anymore for reevaluation of declaration
14947     * [71]14775 LFS (large file support) tests missing
14948     * [72]14821 Duplicate namespace alias declaration should not conflict
14949     * [73]14930 Friend declaration ignored
14950     * [74]14932 cannot use offsetof to get offsets of array elements in
14951       g++ 3.4.0
14952     * [75]14950 [non unit-at-a-time] always_inline does not mix with
14953       templates and -O0
14954     * [76]14962 g++ ignores #pragma redefine_extname
14955     * [77]14975 Segfault on low-level write error during imbue
14956     * [78]15002 Linewise stream input is unusably slow (std::string slow)
14957     * [79]15025 compiler accepts redeclaration of template as
14958       non-template
14959     * [80]15046 [arm] Math functions misdetected by cross configuration
14960     * [81]15069 a bit test on a variable of enum type is miscompiled
14961     * [82]15074 g++ -lsupc++ still links against libstdc++
14962     * [83]15083 spurious "statement has no effect" warning
14963     * [84]15096 parse error with templates and pointer to const member
14964     * [85]15287 combination of operator[] and operator .* fails in
14965       templates
14966     * [86]15317 __attribute__ unused in first parameter of constructor
14967       gives error
14968     * [87]15337 sizeof on incomplete type diagnostic
14969     * [88]15361 bitset<>::_Find_next fails
14970     * [89]15412 _GLIBCXX_ symbols symbols defined and used in different
14971       namespaces
14972     * [90]15427 valid code results in incomplete type error
14973     * [91]15471 Incorrect member pointer offsets in anonymous
14974       structs/unions
14975     * [92]15503 nested template problem
14976     * [93]15507 compiler hangs while laying out union
14977     * [94]15542 operator & and template definitions
14978     * [95]15565 SLES9: leading + sign for unsigned int with showpos
14979     * [96]15625 friend defined inside a template fails to find static
14980       function
14981     * [97]15629 Function templates, overloads, and friend name injection
14982     * [98]15742 'noreturn' attribute ignored in method of template
14983       functions.
14984     * [99]15775 Allocator::pointer consistently ignored
14985     * [100]15821 Duplicate namespace alias within namespace rejected
14986     * [101]15862 'enum yn' fails (confict with undeclared builtin)
14987     * [102]15875 rejects pointer to member in template
14988     * [103]15877 valid code using templates and anonymous enums is
14989       rejected
14990     * [104]15947 Puzzling error message for wrong destructor declaration
14991       in template class
14992     * [105]16020 cannot copy __gnu_debug::bitset
14993     * [106]16154 input iterator concept too restrictive
14994     * [107]16174 deducing top-level consts
14995
14996    Java
14997
14998     * [108]14315 Java compiler is not parallel make safe
14999
15000    Fortran
15001
15002     * [109]15151 [g77] incorrect logical i/o in 64-bit mode
15003
15004    Objective-C
15005
15006     * [110]7993 private variables cannot be shadowed in subclasses
15007
15008    Optimization bugs
15009
15010     * [111]15228 useless copies of floating point operands
15011     * [112]15345 [non-unit-at-a-time] unreferenced nested inline
15012       functions not optimized away
15013     * [113]15945 Incorrect floating point optimization
15014     * [114]15526 ftrapv aborts on 0 * (-1)
15015     * [115]14690 Miscompiled POOMA tests
15016     * [116]15112 GCC generates code to write to unchanging memory
15017
15018    Preprocessor
15019
15020     * [117]15067 Minor glitch in the source of cpp
15021
15022    Main driver program bugs
15023
15024     * [118]1963 collect2 interprets -oldstyle_liblookup as -o
15025       ldstyle_liblookup
15026
15027    x86-specific (Intel/AMD)
15028
15029     * [119]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND*
15030       section}
15031
15032    HPPA-specific
15033
15034     * [120]14782 GCC produces an unaligned data access at -O2
15035     * [121]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2
15036     * [122]15202 ICE in reload_cse_simplify_operands, in postreload.c
15037
15038    IA64-specific
15039
15040     * [123]14610 __float80 constants incorrectly emitted
15041     * [124]14813 init_array sections are initialized in the wrong order
15042     * [125]14857 GCC segfault on duplicated asm statement
15043     * [126]15598 Gcc 3.4 ICE on valid code
15044     * [127]15653 Gcc 3.4 ICE on valid code
15045
15046    MIPS-specific
15047
15048     * [128]15189 wrong filling of delay slot with -march=mips1 -G0
15049       -mno-split-addresses -mno-explicit-relocs
15050     * [129]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as
15051       2.14.91
15052     * [130]16144 Bogus reference to __divdf3 when -O1
15053     * [131]16176 Miscompilation of unaligned data in MIPS backend
15054
15055    PowerPC-specific
15056
15057     * [132]11591 ICE in gcc.dg/altivec-5.c
15058     * [133]12028 powerpc-eabispe produces bad sCOND operation
15059     * [134]14478 rs6000 geu/ltu patterns generate incorrect code
15060     * [135]14567 long double and va_arg complex args
15061     * [136]14715 Altivec stack layout may overlap gpr save with stack
15062       temps
15063     * [137]14902 (libstdc++) Stream checking functions fail when -pthread
15064       option is used.
15065     * [138]14924 Compiler ICE on valid code
15066     * [139]14960 -maltivec affects vector return with -mabi=no-altivec
15067     * [140]15106 vector varargs failure passing from altivec to
15068       non-altivec code for -m32
15069     * [141]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 &
15070       half-word operation
15071     * [142]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx
15072       and stvx
15073     * [143]15662 Segmentation fault when an exception is thrown - even if
15074       try and catch are specified
15075
15076    s390-specific
15077
15078     * [144]15054 Bad code due to overlapping stack temporaries
15079
15080    SPARC-specific
15081
15082     * [145]15783 ICE with union assignment in 64-bit mode
15083     * [146]15626 GCC 3.4 emits "ld: warning: relocation error:
15084       R_SPARC_UA32"
15085
15086    x86-64-specific
15087
15088     * [147]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64
15089     * [148]14723 Backported -march=nocona from mainline
15090     * [149]15290 __float128 failed to pass to function properly
15091
15092    Cygwin/Mingw32-specific
15093
15094     * [150]15250 Option -mms-bitfields support on GCC 3.4 is not
15095       conformant to MS layout
15096     * [151]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe
15097       worker on windows32 targets
15098
15099    Bugs specific to embedded processors
15100
15101     * [152]8309 [m68k] -m5200 produces erroneous SImode set of short
15102       varaible on stack
15103     * [153]13250 [SH] Gcc code for rotation clobbers the register, but
15104       gcc continues to use the register as if it was not clobbered
15105     * [154]13803 [coldfire] movqi operand constraints too restrictivefor
15106       TARGET_COLDFIRE
15107     * [155]14093 [SH] ICE for code when using -mhitachi option in SH
15108     * [156]14457 [m6811hc] ICE with simple c++ source
15109     * [157]14542 [m6811hc] ICE on simple source
15110     * [158]15100 [SH] cc1plus got hang-up on
15111       libstdc++-v3/testsuite/abi_check.cc
15112     * [159]15296 [CRIS] Delayed branch scheduling causing invalid code on
15113       cris-*
15114     * [160]15396 [SH] ICE with -O2 -fPIC
15115     * [161]15782 [coldfire] m68k_output_mi_thunk emits wrong code for
15116       ColdFire
15117
15118    Testsuite problems (compiler not affected)
15119
15120     * [162]11610 libstdc++ testcases 27_io/* don't work properly remotely
15121     * [163]15488 (libstdc++) possibly insufficient file permissions for
15122       executing test suite
15123     * [164]15489 (libstdc++) testsuite_files determined incorrectly
15124
15125    Documentation bugs
15126
15127     * [165]13928 (libstdc++) no whatis info in some man pages generated
15128       by doxygen
15129     * [166]14150 Ada documentation out of date
15130     * [167]14949 (c++) Need to document method visibility changes
15131     * [168]15123 libstdc++-doc: Allocators.3 manpage is empty
15132     __________________________________________________________________
15133
15134GCC 3.4.2
15135
15136  Bug Fixes
15137
15138   This section lists the problem reports (PRs) from GCC's bug tracking
15139   system that are known to be fixed in the 3.4.2 release. This list might
15140   not be complete (that is, it is possible that some PRs that have been
15141   fixed are not listed here).
15142
15143    Bootstrap failures and issues
15144
15145     * [169]16469 [mips-sgi-irix5.3] bootstrap fails in
15146       libstdc++-v3/testsuite
15147     * [170]16344 [hppa-linux-gnu] libstdc++'s PCH built by
15148       profiledbootstrap does not work with the built compiler
15149     * [171]16842 [Solaris/x86] mkheaders can not find mkheaders.conf
15150
15151    Multi-platform internal compiler errors (ICEs)
15152
15153     * [172]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in
15154       cp_parser_class_specifier, in cp/parser.c
15155     * [173]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
15156     * [174]15461 (c++) ICE due to NRV and inlining
15157     * [175]15890 (c++) ICE in c_expand_expr, in c-common.c
15158     * [176]16180 ICE: segmentation fault in RTL optimization
15159     * [177]16224 (c++) ICE in write_unscoped_name (template/namespace)
15160     * [178]16408 ICE: in delete_insn, in cfgrtl.c
15161     * [179]16529 (c++) ICE for: namespace-alias shall not be declared as
15162       the name of any other entity
15163     * [180]16698 (c++) ICE with exceptions and declaration of __cxa_throw
15164     * [181]16706 (c++) ICE in finish_member_declaration, in
15165       cp/semantics.c
15166     * [182]16810 (c++) Legal C++ program with cast gives ICE in
15167       build_ptrmemfunc
15168     * [183]16851 (c++) ICE when throwing a comma expression
15169     * [184]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c
15170     * [185]16904 (c++) ICE in finish_class_member_access_expr, in
15171       cp/typeck.c
15172     * [186]16905 (c++) ICE (segfault) with exceptions
15173     * [187]16964 (c++) ICE in cp_parser_class_specifier due to
15174       redefinition
15175     * [188]17068 (c++) ICE: tree check: expected class 'd', have 'x'
15176       (identifier_node) in dependent_template_p, in cp/pt.c
15177
15178    Preprocessor bugs
15179
15180     * [189]16366 Preprocessor option -remap causes memory corruption
15181
15182    Optimization
15183
15184     * [190]15345 unreferenced nested inline functions not optimized away
15185     * [191]16590 Incorrect execution when compiling with -O2
15186     * [192]16693 Bitwise AND is lost when used within a cast to an enum
15187       of the same precision
15188     * [193]17078 Jump into if(0) substatement fails
15189
15190    Problems in generated debug information
15191
15192     * [194]13956 incorrect stabs for nested local variables
15193
15194    C front end bugs
15195
15196     * [195]16684 GCC should not warn about redundant redeclarations of
15197       built-ins
15198
15199    C++ compiler and library
15200
15201     * [196]12658 Thread safety problems in locale::global() and
15202       locale::locale()
15203     * [197]13092 g++ accepts invalid pointer-to-member conversion
15204     * [198]15320 Excessive memory consumption
15205     * [199]16246 Incorrect template argument deduction
15206     * [200]16273 Memory exhausted when using nested classes and virtual
15207       functions
15208     * [201]16401 ostringstream in gcc 3.4.x very slow for big data
15209     * [202]16411 undefined reference to
15210       __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char>
15211       >::file()
15212     * [203]16489 G++ incorrectly rejects use of a null constant integral
15213       expression as a null constant pointer
15214     * [204]16618 offsetof fails with constant member
15215     * [205]16637 syntax error reported for valid input code
15216     * [206]16717 __attribute__((constructor)) broken in C++
15217     * [207]16813 compiler error in DEBUG version of range insertion
15218       std::map::insert
15219     * [208]16853 pointer-to-member initialization from incompatible one
15220       accepted
15221     * [209]16889 ambiguity is not detected
15222     * [210]16959 Segmentation fault in ios_base::sync_with_stdio
15223
15224    Java compiler and library
15225
15226     * [211]7587 direct threaded interpreter not thread-safe
15227     * [212]16473 ServerSocket accept() leaks file descriptors
15228     * [213]16478 Hash synchronization deadlock with finalizers
15229
15230    Alpha-specific
15231
15232     * [214]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c
15233     * [215]16974 could not split insn (ice in final_scan_insn, in
15234       final.c)
15235
15236    x86-specific
15237
15238     * [216]16298 ICE in output_operand
15239     * [217]17113 ICE with SSE2 intrinsics
15240
15241    x86-64 specific
15242
15243     * [218]14697 libstdc++ couldn't find 32bit libgcc_s
15244
15245    MIPS-specific
15246
15247     * [219]15869 [mips64] No NOP after LW (with -mips1 -O0)
15248     * [220]16325 [mips64] value profiling clobbers gp on mips
15249     * [221]16357 [mipsisa64-elf] ICE copying 7 bytes between extern
15250       char[]s
15251     * [222]16380 [mips64] Use of uninitialised register after dbra
15252       conversion
15253     * [223]16407 [mips64] Unaligned access to local variables
15254     * [224]16643 [mips64] verify_local_live_at_start ICE after
15255       crossjumping & cfgcleanup
15256
15257    ARM-specific
15258
15259     * [225]15927 THUMB -O2: strength-reduced iteration variable ends up
15260       off by 1
15261     * [226]15948 THUMB: ICE with non-commutative cbranch
15262     * [227]17019 THUMB: bad switch statement in md code for
15263       addsi3_cbranch_scratch
15264
15265    IA64-specific
15266
15267     * [228]16130 ICE on valid code: in bundling, in config/ia64/ia64.c
15268       (-mtune=merced)
15269     * [229]16142 ICE on valid code: in bundling, in config/ia64/ia64.c
15270       (-mtune=itanium)
15271     * [230]16278 Gcc failed to build Linux kernel with -mtune=merced
15272     * [231]16414 ICE on valid code: typo in comparison of asm_noperands
15273       result
15274     * [232]16445 ICE on valid code: don't count ignored insns
15275     * [233]16490 ICE (segfault) while compiling with -fprofile-use
15276     * [234]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS
15277
15278    PowerPC-specific
15279
15280     * [235]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x
15281     * [236]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions
15282       issue)
15283
15284    SPARC-specific
15285
15286     * [237]16199 ICE while compiling apache 2.0.49
15287     * [238]16416 -m64 doesn't imply -mcpu=v9 anymore
15288     * [239]16430 ICE when returning non-C aggregates larger than 16 bytes
15289
15290    Bugs specific to embedded processors
15291
15292     * [240]16379 [m32r] can't output large model function call of memcpy
15293     * [241]17093 [m32r] ICE with -msdata=use -O0
15294     * [242]17119 [m32r] ICE at switch case 0x8000
15295
15296    DJGPP-specific
15297
15298     * [243]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp
15299
15300    Alpha Tru64-specific
15301
15302     * [244]16210 libstdc++ gratuitously omits "long long" I/O
15303
15304    Testsuite, documentation issues (compiler is not affected):
15305
15306     * [245]15488 (libstdc++) possibly insufficient file permissions for
15307       executing test suite
15308     * [246]16250 ada/doctools runs makeinfo even in release tarball
15309     __________________________________________________________________
15310
15311GCC 3.4.3
15312
15313   This is the [247]list of problem reports (PRs) from GCC's bug tracking
15314   system that are known to be fixed in the 3.4.3 release. This list might
15315   not be complete (that is, it is possible that some PRs that have been
15316   fixed are not listed here).
15317
15318    Bootstrap failures
15319
15320     * [248]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1
15321     * [249]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold
15322       when undeclared
15323
15324    Internal compiler errors (ICEs) affecting multiple platforms
15325
15326     * [250]13948 (java) GCJ segmentation fault while compiling GL4Java
15327       .class files
15328     * [251]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
15329     * [252]16301 (c++) ICE when "strong" attribute is attached to a using
15330       directive
15331     * [253]16566 ICE with flexible arrays
15332     * [254]17023 ICE with nested functions in parameter declaration
15333     * [255]17027 ICE with noreturn function in loop at -O2
15334     * [256]17524 ICE in grokdeclarator, in cp/decl.c
15335     * [257]17826 (c++) ICE in cp_tree_equal
15336
15337    C and optimization bugs
15338
15339     * [258]15526 -ftrapv aborts on 0 * (-1)
15340     * [259]16999 #ident stopped working
15341     * [260]17503 quadratic behaviour in invalid_mode_change_p
15342     * [261]17581 Long long arithmetic fails inside a switch/case
15343       statement when compiled with -O2
15344     * [262]18129 -fwritable-strings doesn't work
15345
15346    C++ compiler and library bugs
15347
15348     * [263]10975 incorrect initial ostringstream::tellp()
15349     * [264]11722 Unbuffered filebuf::sgetn is slow
15350     * [265]14534 Unrecognizing static function as a template parameter
15351       when its return value is also templated
15352     * [266]15172 Copy constructor optimization in aggregate
15353       initialization
15354     * [267]15786 Bad error message for frequently occuring error.
15355     * [268]16162 Rejects valid member-template-definition
15356     * [269]16612 empty basic_strings can't live in shared memory
15357     * [270]16715 std::basic_iostream is instantiated when used, even
15358       though instantiations are already contained in libstdc++
15359     * [271]16848 code in /ext/demangle.h appears broken
15360     * [272]17132 GCC fails to eliminate function template specialization
15361       when argument deduction fails
15362     * [273]17259 One more _S_leaf incorrectly qualified with _RopeRep::
15363       in ropeimpl.h
15364     * [274]17327 use of `enumeral_type' in template type unification
15365     * [275]17393 "unused variable '._0'" warning with -Wall
15366     * [276]17501 Confusion with member templates
15367     * [277]17537 g++ not passing -lstdc++ to linker when all command line
15368       arguments are libraries
15369     * [278]17585 usage of unqualified name of static member from within
15370       class not allowed
15371     * [279]17821 Poor diagnostic for using "." instead of "->"
15372     * [280]17829 wrong error: call of overloaded function is ambiguous
15373     * [281]17851 Misleading diagnostic for invalid function declarations
15374       with undeclared types
15375     * [282]17976 Destructor is called twice
15376     * [283]18020 rejects valid definition of enum value in template
15377     * [284]18093 bogus conflict in namespace aliasing
15378     * [285]18140 C++ parser bug when using >> in templates
15379
15380    Fortran
15381
15382     * [286]17541 data statements with double precision constants fail
15383
15384    x86-specific
15385
15386     * [287]17853 -O2 ICE for MMX testcase
15387
15388    SPARC-specific
15389
15390     * [288]17245 ICE compiling gsl-1.5 statistics/lag1.c
15391
15392    Darwin-specific
15393
15394     * [289]17167 FATAL:Symbol L_foo$stub already defined.
15395
15396    AIX-specific
15397
15398     * [290]17277 could not catch an exception when specified -maix64
15399
15400    Solaris-specific
15401
15402     * [291]17505 <cmath> calls acosf(), ceilf(), and other functions
15403       missing from system libraries
15404
15405    HP/UX specific:
15406
15407     * [292]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl
15408
15409    ARM-specific
15410
15411     * [293]17384 ICE with mode attribute on structures
15412
15413    MIPS-specific
15414
15415     * [294]17770 No NOP after LWL with -mips1
15416
15417    Other embedded target specific
15418
15419     * [295]11476 [arc-elf] gcc ICE on newlib's vfprintf.c
15420     * [296]14064 [avr-elf] -fdata-sections triggers ICE
15421     * [297]14678 [m68hc11-elf] gcc ICE
15422     * [298]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__
15423     * [299]15790 [i686-coff] Alignment error building gcc with i686-coff
15424       target
15425     * [300]15886 [SH] Miscompilation with -O2 -fPIC
15426     * [301]16884 [avr-elf] [fweb related] bug while initializing
15427       variables
15428
15429    Bugs relating to debugger support
15430
15431     * [302]13841 missing debug info for _Complex function arguments
15432     * [303]15860 [big-endian targets] No DW_AT_location debug info is
15433       emitted for formal arguments to a function that uses "register"
15434       qualifiers
15435
15436    Testsuite issues (compiler not affected)
15437
15438     * [304]17465 Testsuite in libffi overrides LD_LIBRARY_PATH
15439     * [305]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH
15440     * [306]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit
15441       testsuite
15442
15443    Documentation
15444
15445     * [307]15498 typo in gcc manual: non-existing locale example en_UK,
15446       should be en_GB
15447     * [308]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap:
15448       document broken shell
15449     * [309]16406 USE_LD_AS_NEEDED undocumented
15450     __________________________________________________________________
15451
15452GCC 3.4.4
15453
15454   This is the [310]list of problem reports (PRs) from GCC's bug tracking
15455   system that are known to be fixed in the 3.4.4 release. This list might
15456   not be complete (that is, it is possible that some PRs that have been
15457   fixed are not listed here).
15458     __________________________________________________________________
15459
15460GCC 3.4.5
15461
15462   This is the [311]list of problem reports (PRs) from GCC's bug tracking
15463   system that are known to be fixed in the 3.4.5 release. This list might
15464   not be complete (that is, it is possible that some PRs that have been
15465   fixed are not listed here).
15466
15467    Bootstrap issues
15468
15469     * [312]24688 sco_math fixincl breaks math.h
15470
15471    C compiler bugs
15472
15473     * [313]17188 struct Foo { } redefinition
15474     * [314]20187 wrong code for ((unsigned char)(unsigned long
15475       long)((a?a:1)&(a*b)))?0:1)
15476     * [315]21873 infinite warning loop on bad array initializer
15477     * [316]21899 enum definition accepts values to be overriden
15478     * [317]22061 ICE in find_function_data, in function.c
15479     * [318]22308 Failure to diagnose violation of constraint 6.516p2
15480     * [319]22458 ICE on missing brace
15481     * [320]22589 ICE casting to long long
15482     * [321]24101 Segfault with preprocessed source
15483
15484    C++ compiler and library bugs
15485
15486     * [322]10611 operations on vector mode not recognized in C++
15487     * [323]13377 unexpected behavior of namespace usage directive
15488     * [324]16002 Strange error message with new parser
15489     * [325]17413 local classes as template argument
15490     * [326]17609 spurious error message after using keyword
15491     * [327]17618 ICE in cp_convert_to_pointer, in cp/cvt.c
15492     * [328]18124 ICE with invalid template template parameter
15493     * [329]18155 typedef in template declaration not rejected
15494     * [330]18177 ICE with const_cast for undeclared variable
15495     * [331]18368 C++ error message regression
15496     * [332]16378 ICE when returning a copy of a packed member
15497     * [333]18466 int ::i; accepted
15498     * [334]18512 ICE on invalid usage of template base class
15499     * [335]18454 ICE when returning undefined type
15500     * [336]18738 typename not allowed with non-dependent qualified name
15501     * [337]18803 rejects access to operator() in template
15502     * [338]19004 ICE in uses_template_parms, in cp/pt.c
15503     * [339]19208 Spurious error about variably modified type
15504     * [340]18253 bad error message / ICE for invalid template parameter
15505     * [341]19608 ICE after friend function definition in local class
15506     * [342]19884 ICE on explicit instantiation of a non-template
15507       constructor
15508     * [343]20153 ICE when C++ template function contains anonymous union
15509     * [344]20563 Infinite loop in diagnostic (and ice after error
15510       message)
15511     * [345]20789 ICE with incomplete type in template
15512     * [346]21336 Internal compiler error when using custom new operators
15513     * [347]21768 ICE in error message due to violation of coding
15514       conventions
15515     * [348]21853 constness of pointer to data member ignored
15516     * [349]21903 Default argument of template function causes a
15517       compile-time error
15518     * [350]21983 multiple diagnostics
15519     * [351]21987 New testsuite failure
15520       g++.dg/warn/conversion-function-1.C
15521     * [352]22153 ICE on invalid template specialization
15522     * [353]22172 Internal compiler error, seg fault.
15523     * [354]21286 filebuf::xsgetn vs pipes
15524     * [355]22233 ICE with wrong number of template parameters
15525     * [356]22508 ICE after invalid operator new
15526     * [357]22545 ICE with pointer to class member & user defined
15527       conversion operator
15528     * [358]23528 Wrong default allocator in ext/hash_map
15529     * [359]23550 char_traits requirements/1.cc test bad math
15530     * [360]23586 Bad diagnostic for invalid namespace-name
15531     * [361]23624 ICE in invert_truthvalue, in fold-const.c
15532     * [362]23639 Bad error message: not a member of '<declaration error>'
15533     * [363]23797 ICE on typename outside template
15534     * [364]23965 Bogus error message: no matching function for call to
15535       'foo(<type error>)'
15536     * [365]24052 &#`label_decl' not supported by dump_expr#<expression
15537       error>
15538     * [366]24580 virtual base class cause exception not to be caught
15539
15540    Problems in generated debug information
15541
15542     * [367]24267 Bad DWARF for altivec vectors
15543
15544    Optimizations issues
15545
15546     * [368]17810 ICE in verify_local_live_at_start
15547     * [369]17860 Wrong generated code for loop with varying bound
15548     * [370]21709 ICE on compile-time complex NaN
15549     * [371]21964 broken tail call at -O2 or more
15550     * [372]22167 Strange optimization bug when using -Os
15551     * [373]22619 Compilation failure for real_const_1.f and
15552       real_const_2.f90
15553     * [374]23241 Invalid code generated for comparison of uchar to 255
15554     * [375]23478 Miscompilation due to reloading of a var that is also
15555       used in EH pad
15556     * [376]24470 segmentation fault in cc1plus when compiling with -O
15557     * [377]24950 ICE in operand_subword_force
15558
15559    Precompiled headers problems
15560
15561     * [378]14400 Cannot compile qt-x11-free-3.3.0
15562     * [379]14940 PCH largefile test fails on various platforms
15563
15564    Preprocessor bugs
15565
15566     * [380]20239 ICE on empty preprocessed input
15567     * [381]15220 "gcc -E -MM -MG" reports missing system headers in
15568       source directory
15569
15570    Testsuite issues
15571
15572     * [382]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on
15573       i686-pc-linux-gnu
15574
15575    Alpha specific
15576
15577     * [383]21888 bootstrap failure with linker relaxation enabled
15578
15579    ARM specific
15580
15581     * [384]15342 [arm-linux]: ICE in verify_local_live_at_start
15582     * [385]23985 Memory aliasing information incorrect in inlined memcpy
15583
15584    ColdFile specific
15585
15586     * [386]16719 Illegal move of byte into address register causes
15587       compiler to ICE
15588
15589    HPPA specific
15590
15591     * [387]21723 ICE while building libgfortran
15592     * [388]21841 -mhp-ld/-mgnu-ld documentation
15593
15594    IA-64 specific
15595
15596     * [389]23644 IA-64 hardware models and configuration options
15597       documentation error
15598     * [390]24718 Shared libgcc not used for linking by default
15599
15600    M68000 specific
15601
15602     * [391]18421 ICE in reload_cse_simplify_operands, in postreload.c
15603
15604    MIPS specific
15605
15606     * [392]20621 ICE in change_address_1, in emit-rtl.c
15607
15608    PowerPC and PowerPC64 specific
15609
15610     * [393]18583 error on valid code: const
15611       __attribute__((altivec(vector__))) doesn't work in arrays
15612     * [394]20191 ICE in reload_cse_simplify_operands
15613     * [395]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined
15614     * [396]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set
15615     * [397]23404 gij trashes args of functions with more than 8 fp args
15616     * [398]23539 C & C++ compiler generating misaligned references
15617       regardless of compiler flags
15618     * [399]24102 floatdisf2_internal2 broken
15619     * [400]24465 -mminimal-toc miscompilation of __thread vars
15620
15621    Solaris specific
15622
15623     * [401]19933 Problem with define of HUGE_VAL in math_c99
15624     * [402]21889 Native Solaris assembler cannot grok DTP-relative debug
15625       symbols
15626
15627    SPARC specific
15628
15629     * [403]19300 PCH failures on sparc-linux
15630     * [404]20301 Assembler labels have a leading "-"
15631     * [405]20673 C PCH testsuite assembly comparison failure
15632
15633    x86 and x86_64 specific
15634
15635     * [406]18582 ICE with arrays of type V2DF
15636     * [407]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2
15637       -fsched2-use-traces
15638     * [408]21716 ICE in reg-stack.c's swap_rtx_condition
15639     * [409]24315 amd64 fails -fpeephole2
15640     __________________________________________________________________
15641
15642GCC 3.4.6
15643
15644   This is the [410]list of problem reports (PRs) from GCC's bug tracking
15645   system that are known to be fixed in the 3.4.6 release. This list might
15646   not be complete (that is, it is possible that some PRs that have been
15647   fixed are not listed here).
15648
15649
15650    For questions related to the use of GCC, please consult these web
15651    pages and the [411]GCC manuals. If that fails, the
15652    [412]gcc-help@gcc.gnu.org mailing list might help. Comments on these
15653    web pages and the development of GCC are welcome on our developer
15654    list at [413]gcc@gcc.gnu.org. All of [414]our lists have public
15655    archives.
15656
15657   Copyright (C) [415]Free Software Foundation, Inc. Verbatim copying and
15658   distribution of this entire article is permitted in any medium,
15659   provided this notice is preserved.
15660
15661   These pages are [416]maintained by the GCC team. Last modified
15662   2021-07-28[417].
15663
15664References
15665
15666   1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
15667   2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus
15668   3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
15669   4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems
15670   5. http://gcc.gnu.org/gcc-3.4/mips-abi.html
15671   6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
15672   7. https://www.boost.org/
15673   8. https://gcc.gnu.org/PR11953
15674   9. https://gcc.gnu.org/PR8361
15675  10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other Builtins
15676  11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209
15677  12. http://gcc.gnu.org/bugs/#cxx_rvalbind
15678  13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
15679  14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
15680  15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
15681  16. http://www.gnu.org/software/classpath/
15682  17. http://www.eclipse.org/
15683  18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html
15684  19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html
15685  20. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
15686  21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html
15687  22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
15688  23. http://gcc.gnu.org/gcc-3.4/mips-abi.html
15689  24. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html
15690  25. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
15691  26. https://gcc.gnu.org/bugzilla/buglist.cgi?short_desc_type=notregexp&short_desc=\[3\.4.*[Rr]egression&target_milestone=3.4.0&bug_status=RESOLVED&resolution=FIXED
15692  27. https://gcc.gnu.org/PR10129
15693  28. https://gcc.gnu.org/PR14576
15694  29. https://gcc.gnu.org/PR14760
15695  30. https://gcc.gnu.org/PR14671
15696  31. https://gcc.gnu.org/PR15093
15697  32. https://gcc.gnu.org/PR15178
15698  33. https://gcc.gnu.org/PR12753
15699  34. https://gcc.gnu.org/PR13985
15700  35. https://gcc.gnu.org/PR14810
15701  36. https://gcc.gnu.org/PR14883
15702  37. https://gcc.gnu.org/PR15044
15703  38. https://gcc.gnu.org/PR15057
15704  39. https://gcc.gnu.org/PR15064
15705  40. https://gcc.gnu.org/PR15142
15706  41. https://gcc.gnu.org/PR15159
15707  42. https://gcc.gnu.org/PR15165
15708  43. https://gcc.gnu.org/PR15193
15709  44. https://gcc.gnu.org/PR15209
15710  45. https://gcc.gnu.org/PR15227
15711  46. https://gcc.gnu.org/PR15285
15712  47. https://gcc.gnu.org/PR15299
15713  48. https://gcc.gnu.org/PR15329
15714  49. https://gcc.gnu.org/PR15550
15715  50. https://gcc.gnu.org/PR15554
15716  51. https://gcc.gnu.org/PR15640
15717  52. https://gcc.gnu.org/PR15666
15718  53. https://gcc.gnu.org/PR15696
15719  54. https://gcc.gnu.org/PR15701
15720  55. https://gcc.gnu.org/PR15761
15721  56. https://gcc.gnu.org/PR15829
15722  57. https://gcc.gnu.org/PR14538
15723  58. https://gcc.gnu.org/PR12391
15724  59. https://gcc.gnu.org/PR14649
15725  60. https://gcc.gnu.org/PR15004
15726  61. https://gcc.gnu.org/PR15749
15727  62. https://gcc.gnu.org/PR10646
15728  63. https://gcc.gnu.org/PR12077
15729  64. https://gcc.gnu.org/PR13598
15730  65. https://gcc.gnu.org/PR14211
15731  66. https://gcc.gnu.org/PR14220
15732  67. https://gcc.gnu.org/PR14245
15733  68. https://gcc.gnu.org/PR14340
15734  69. https://gcc.gnu.org/PR14600
15735  70. https://gcc.gnu.org/PR14668
15736  71. https://gcc.gnu.org/PR14775
15737  72. https://gcc.gnu.org/PR14821
15738  73. https://gcc.gnu.org/PR14930
15739  74. https://gcc.gnu.org/PR14932
15740  75. https://gcc.gnu.org/PR14950
15741  76. https://gcc.gnu.org/PR14962
15742  77. https://gcc.gnu.org/PR14975
15743  78. https://gcc.gnu.org/PR15002
15744  79. https://gcc.gnu.org/PR15025
15745  80. https://gcc.gnu.org/PR15046
15746  81. https://gcc.gnu.org/PR15069
15747  82. https://gcc.gnu.org/PR15074
15748  83. https://gcc.gnu.org/PR15083
15749  84. https://gcc.gnu.org/PR15096
15750  85. https://gcc.gnu.org/PR15287
15751  86. https://gcc.gnu.org/PR15317
15752  87. https://gcc.gnu.org/PR15337
15753  88. https://gcc.gnu.org/PR15361
15754  89. https://gcc.gnu.org/PR15412
15755  90. https://gcc.gnu.org/PR15427
15756  91. https://gcc.gnu.org/PR15471
15757  92. https://gcc.gnu.org/PR15503
15758  93. https://gcc.gnu.org/PR15507
15759  94. https://gcc.gnu.org/PR15542
15760  95. https://gcc.gnu.org/PR15565
15761  96. https://gcc.gnu.org/PR15625
15762  97. https://gcc.gnu.org/PR15629
15763  98. https://gcc.gnu.org/PR15742
15764  99. https://gcc.gnu.org/PR15775
15765 100. https://gcc.gnu.org/PR15821
15766 101. https://gcc.gnu.org/PR15862
15767 102. https://gcc.gnu.org/PR15875
15768 103. https://gcc.gnu.org/PR15877
15769 104. https://gcc.gnu.org/PR15947
15770 105. https://gcc.gnu.org/PR16020
15771 106. https://gcc.gnu.org/PR16154
15772 107. https://gcc.gnu.org/PR16174
15773 108. https://gcc.gnu.org/PR14315
15774 109. https://gcc.gnu.org/PR15151
15775 110. https://gcc.gnu.org/PR7993
15776 111. https://gcc.gnu.org/PR15228
15777 112. https://gcc.gnu.org/PR15345
15778 113. https://gcc.gnu.org/PR15945
15779 114. https://gcc.gnu.org/PR15526
15780 115. https://gcc.gnu.org/PR14690
15781 116. https://gcc.gnu.org/PR15112
15782 117. https://gcc.gnu.org/PR15067
15783 118. https://gcc.gnu.org/PR1963
15784 119. https://gcc.gnu.org/PR15717
15785 120. https://gcc.gnu.org/PR14782
15786 121. https://gcc.gnu.org/PR14828
15787 122. https://gcc.gnu.org/PR15202
15788 123. https://gcc.gnu.org/PR14610
15789 124. https://gcc.gnu.org/PR14813
15790 125. https://gcc.gnu.org/PR14857
15791 126. https://gcc.gnu.org/PR15598
15792 127. https://gcc.gnu.org/PR15653
15793 128. https://gcc.gnu.org/PR15189
15794 129. https://gcc.gnu.org/PR15331
15795 130. https://gcc.gnu.org/PR16144
15796 131. https://gcc.gnu.org/PR16176
15797 132. https://gcc.gnu.org/PR11591
15798 133. https://gcc.gnu.org/PR12028
15799 134. https://gcc.gnu.org/PR14478
15800 135. https://gcc.gnu.org/PR14567
15801 136. https://gcc.gnu.org/PR14715
15802 137. https://gcc.gnu.org/PR14902
15803 138. https://gcc.gnu.org/PR14924
15804 139. https://gcc.gnu.org/PR14960
15805 140. https://gcc.gnu.org/PR15106
15806 141. https://gcc.gnu.org/PR16026
15807 142. https://gcc.gnu.org/PR15191
15808 143. https://gcc.gnu.org/PR15662
15809 144. https://gcc.gnu.org/PR15054
15810 145. https://gcc.gnu.org/PR15783
15811 146. https://gcc.gnu.org/PR15626
15812 147. https://gcc.gnu.org/PR14326
15813 148. https://gcc.gnu.org/PR14723
15814 149. https://gcc.gnu.org/PR15290
15815 150. https://gcc.gnu.org/PR15250
15816 151. https://gcc.gnu.org/PR15551
15817 152. https://gcc.gnu.org/PR8309
15818 153. https://gcc.gnu.org/PR13250
15819 154. https://gcc.gnu.org/PR13803
15820 155. https://gcc.gnu.org/PR14093
15821 156. https://gcc.gnu.org/PR14457
15822 157. https://gcc.gnu.org/PR14542
15823 158. https://gcc.gnu.org/PR15100
15824 159. https://gcc.gnu.org/PR15296
15825 160. https://gcc.gnu.org/PR15396
15826 161. https://gcc.gnu.org/PR15782
15827 162. https://gcc.gnu.org/PR11610
15828 163. https://gcc.gnu.org/PR15488
15829 164. https://gcc.gnu.org/PR15489
15830 165. https://gcc.gnu.org/PR13928
15831 166. https://gcc.gnu.org/PR14150
15832 167. https://gcc.gnu.org/PR14949
15833 168. https://gcc.gnu.org/PR15123
15834 169. https://gcc.gnu.org/PR16469
15835 170. https://gcc.gnu.org/PR16344
15836 171. https://gcc.gnu.org/PR16842
15837 172. https://gcc.gnu.org/PR12608
15838 173. https://gcc.gnu.org/PR14492
15839 174. https://gcc.gnu.org/PR15461
15840 175. https://gcc.gnu.org/PR15890
15841 176. https://gcc.gnu.org/PR16180
15842 177. https://gcc.gnu.org/PR16224
15843 178. https://gcc.gnu.org/PR16408
15844 179. https://gcc.gnu.org/PR16529
15845 180. https://gcc.gnu.org/PR16698
15846 181. https://gcc.gnu.org/PR16706
15847 182. https://gcc.gnu.org/PR16810
15848 183. https://gcc.gnu.org/PR16851
15849 184. https://gcc.gnu.org/PR16870
15850 185. https://gcc.gnu.org/PR16904
15851 186. https://gcc.gnu.org/PR16905
15852 187. https://gcc.gnu.org/PR16964
15853 188. https://gcc.gnu.org/PR17068
15854 189. https://gcc.gnu.org/PR16366
15855 190. https://gcc.gnu.org/PR15345
15856 191. https://gcc.gnu.org/PR16590
15857 192. https://gcc.gnu.org/PR16693
15858 193. https://gcc.gnu.org/PR17078
15859 194. https://gcc.gnu.org/PR13956
15860 195. https://gcc.gnu.org/PR16684
15861 196. https://gcc.gnu.org/PR12658
15862 197. https://gcc.gnu.org/PR13092
15863 198. https://gcc.gnu.org/PR15320
15864 199. https://gcc.gnu.org/PR16246
15865 200. https://gcc.gnu.org/PR16273
15866 201. https://gcc.gnu.org/PR16401
15867 202. https://gcc.gnu.org/PR16411
15868 203. https://gcc.gnu.org/PR16489
15869 204. https://gcc.gnu.org/PR16618
15870 205. https://gcc.gnu.org/PR16637
15871 206. https://gcc.gnu.org/PR16717
15872 207. https://gcc.gnu.org/PR16813
15873 208. https://gcc.gnu.org/PR16853
15874 209. https://gcc.gnu.org/PR16889
15875 210. https://gcc.gnu.org/PR16959
15876 211. https://gcc.gnu.org/PR7587
15877 212. https://gcc.gnu.org/PR16473
15878 213. https://gcc.gnu.org/PR16478
15879 214. https://gcc.gnu.org/PR10695
15880 215. https://gcc.gnu.org/PR16974
15881 216. https://gcc.gnu.org/PR16298
15882 217. https://gcc.gnu.org/PR17113
15883 218. https://gcc.gnu.org/PR14697
15884 219. https://gcc.gnu.org/PR15869
15885 220. https://gcc.gnu.org/PR16325
15886 221. https://gcc.gnu.org/PR16357
15887 222. https://gcc.gnu.org/PR16380
15888 223. https://gcc.gnu.org/PR16407
15889 224. https://gcc.gnu.org/PR16643
15890 225. https://gcc.gnu.org/PR15927
15891 226. https://gcc.gnu.org/PR15948
15892 227. https://gcc.gnu.org/PR17019
15893 228. https://gcc.gnu.org/PR16130
15894 229. https://gcc.gnu.org/PR16142
15895 230. https://gcc.gnu.org/PR16278
15896 231. https://gcc.gnu.org/PR16414
15897 232. https://gcc.gnu.org/PR16445
15898 233. https://gcc.gnu.org/PR16490
15899 234. https://gcc.gnu.org/PR16683
15900 235. https://gcc.gnu.org/PR16195
15901 236. https://gcc.gnu.org/PR16239
15902 237. https://gcc.gnu.org/PR16199
15903 238. https://gcc.gnu.org/PR16416
15904 239. https://gcc.gnu.org/PR16430
15905 240. https://gcc.gnu.org/PR16379
15906 241. https://gcc.gnu.org/PR17093
15907 242. https://gcc.gnu.org/PR17119
15908 243. https://gcc.gnu.org/PR15928
15909 244. https://gcc.gnu.org/PR16210
15910 245. https://gcc.gnu.org/PR15488
15911 246. https://gcc.gnu.org/PR16250
15912 247. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3
15913 248. https://gcc.gnu.org/PR17369
15914 249. https://gcc.gnu.org/PR17850
15915 250. https://gcc.gnu.org/PR13948
15916 251. https://gcc.gnu.org/PR14492
15917 252. https://gcc.gnu.org/PR16301
15918 253. https://gcc.gnu.org/PR16566
15919 254. https://gcc.gnu.org/PR17023
15920 255. https://gcc.gnu.org/PR17027
15921 256. https://gcc.gnu.org/PR17524
15922 257. https://gcc.gnu.org/PR17826
15923 258. https://gcc.gnu.org/PR15526
15924 259. https://gcc.gnu.org/PR16999
15925 260. https://gcc.gnu.org/PR17503
15926 261. https://gcc.gnu.org/PR17581
15927 262. https://gcc.gnu.org/PR18129
15928 263. https://gcc.gnu.org/PR10975
15929 264. https://gcc.gnu.org/PR11722
15930 265. https://gcc.gnu.org/PR14534
15931 266. https://gcc.gnu.org/PR15172
15932 267. https://gcc.gnu.org/PR15786
15933 268. https://gcc.gnu.org/PR16162
15934 269. https://gcc.gnu.org/PR16612
15935 270. https://gcc.gnu.org/PR16715
15936 271. https://gcc.gnu.org/PR16848
15937 272. https://gcc.gnu.org/PR17132
15938 273. https://gcc.gnu.org/PR17259
15939 274. https://gcc.gnu.org/PR17327
15940 275. https://gcc.gnu.org/PR17393
15941 276. https://gcc.gnu.org/PR17501
15942 277. https://gcc.gnu.org/PR17537
15943 278. https://gcc.gnu.org/PR17585
15944 279. https://gcc.gnu.org/PR17821
15945 280. https://gcc.gnu.org/PR17829
15946 281. https://gcc.gnu.org/PR17851
15947 282. https://gcc.gnu.org/PR17976
15948 283. https://gcc.gnu.org/PR18020
15949 284. https://gcc.gnu.org/PR18093
15950 285. https://gcc.gnu.org/PR18140
15951 286. https://gcc.gnu.org/PR17541
15952 287. https://gcc.gnu.org/PR17853
15953 288. https://gcc.gnu.org/PR17245
15954 289. https://gcc.gnu.org/PR17167
15955 290. https://gcc.gnu.org/PR17277
15956 291. https://gcc.gnu.org/PR17505
15957 292. https://gcc.gnu.org/PR17684
15958 293. https://gcc.gnu.org/PR17384
15959 294. https://gcc.gnu.org/PR17770
15960 295. https://gcc.gnu.org/PR11476
15961 296. https://gcc.gnu.org/PR14064
15962 297. https://gcc.gnu.org/PR14678
15963 298. https://gcc.gnu.org/PR15583
15964 299. https://gcc.gnu.org/PR15790
15965 300. https://gcc.gnu.org/PR15886
15966 301. https://gcc.gnu.org/PR16884
15967 302. https://gcc.gnu.org/PR13841
15968 303. https://gcc.gnu.org/PR15860
15969 304. https://gcc.gnu.org/PR17465
15970 305. https://gcc.gnu.org/PR17469
15971 306. https://gcc.gnu.org/PR18138
15972 307. https://gcc.gnu.org/PR15498
15973 308. https://gcc.gnu.org/PR15747
15974 309. https://gcc.gnu.org/PR16406
15975 310. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4
15976 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5
15977 312. https://gcc.gnu.org/PR24688
15978 313. https://gcc.gnu.org/PR17188
15979 314. https://gcc.gnu.org/PR20187
15980 315. https://gcc.gnu.org/PR21873
15981 316. https://gcc.gnu.org/PR21899
15982 317. https://gcc.gnu.org/PR22061
15983 318. https://gcc.gnu.org/PR22208
15984 319. https://gcc.gnu.org/PR22458
15985 320. https://gcc.gnu.org/PR22589
15986 321. https://gcc.gnu.org/PR24101
15987 322. https://gcc.gnu.org/PR10611
15988 323. https://gcc.gnu.org/PR13377
15989 324. https://gcc.gnu.org/PR16002
15990 325. https://gcc.gnu.org/PR17413
15991 326. https://gcc.gnu.org/PR17609
15992 327. https://gcc.gnu.org/PR17618
15993 328. https://gcc.gnu.org/PR18124
15994 329. https://gcc.gnu.org/PR18155
15995 330. https://gcc.gnu.org/PR18177
15996 331. https://gcc.gnu.org/PR18368
15997 332. https://gcc.gnu.org/PR18378
15998 333. https://gcc.gnu.org/PR18466
15999 334. https://gcc.gnu.org/PR18512
16000 335. https://gcc.gnu.org/PR18545
16001 336. https://gcc.gnu.org/PR18738
16002 337. https://gcc.gnu.org/PR18803
16003 338. https://gcc.gnu.org/PR19004
16004 339. https://gcc.gnu.org/PR19208
16005 340. https://gcc.gnu.org/PR19253
16006 341. https://gcc.gnu.org/PR19608
16007 342. https://gcc.gnu.org/PR19884
16008 343. https://gcc.gnu.org/PR20153
16009 344. https://gcc.gnu.org/PR20563
16010 345. https://gcc.gnu.org/PR20789
16011 346. https://gcc.gnu.org/PR21336
16012 347. https://gcc.gnu.org/PR21768
16013 348. https://gcc.gnu.org/PR21853
16014 349. https://gcc.gnu.org/PR21903
16015 350. https://gcc.gnu.org/PR21983
16016 351. https://gcc.gnu.org/PR21987
16017 352. https://gcc.gnu.org/PR22153
16018 353. https://gcc.gnu.org/PR22172
16019 354. https://gcc.gnu.org/PR21286
16020 355. https://gcc.gnu.org/PR22233
16021 356. https://gcc.gnu.org/PR22508
16022 357. https://gcc.gnu.org/PR22545
16023 358. https://gcc.gnu.org/PR23528
16024 359. https://gcc.gnu.org/PR23550
16025 360. https://gcc.gnu.org/PR23586
16026 361. https://gcc.gnu.org/PR23624
16027 362. https://gcc.gnu.org/PR23639
16028 363. https://gcc.gnu.org/PR23797
16029 364. https://gcc.gnu.org/PR23965
16030 365. https://gcc.gnu.org/PR24052
16031 366. https://gcc.gnu.org/PR24580
16032 367. https://gcc.gnu.org/PR24267
16033 368. https://gcc.gnu.org/PR17810
16034 369. https://gcc.gnu.org/PR17860
16035 370. https://gcc.gnu.org/PR21709
16036 371. https://gcc.gnu.org/PR21964
16037 372. https://gcc.gnu.org/PR22167
16038 373. https://gcc.gnu.org/PR22619
16039 374. https://gcc.gnu.org/PR23241
16040 375. https://gcc.gnu.org/PR23478
16041 376. https://gcc.gnu.org/PR24470
16042 377. https://gcc.gnu.org/PR24950
16043 378. https://gcc.gnu.org/PR14400
16044 379. https://gcc.gnu.org/PR14940
16045 380. https://gcc.gnu.org/PR20239
16046 381. https://gcc.gnu.org/PR15220
16047 382. https://gcc.gnu.org/PR19275
16048 383. https://gcc.gnu.org/PR21888
16049 384. https://gcc.gnu.org/PR15342
16050 385. https://gcc.gnu.org/PR23985
16051 386. https://gcc.gnu.org/PR16719
16052 387. https://gcc.gnu.org/PR21723
16053 388. https://gcc.gnu.org/PR21841
16054 389. https://gcc.gnu.org/PR23644
16055 390. https://gcc.gnu.org/PR24718
16056 391. https://gcc.gnu.org/PR18421
16057 392. https://gcc.gnu.org/PR20621
16058 393. https://gcc.gnu.org/PR18583
16059 394. https://gcc.gnu.org/PR20191
16060 395. https://gcc.gnu.org/PR22083
16061 396. https://gcc.gnu.org/PR23070
16062 397. https://gcc.gnu.org/PR23404
16063 398. https://gcc.gnu.org/PR23539
16064 399. https://gcc.gnu.org/PR24102
16065 400. https://gcc.gnu.org/PR24465
16066 401. https://gcc.gnu.org/PR19933
16067 402. https://gcc.gnu.org/PR21889
16068 403. https://gcc.gnu.org/PR19300
16069 404. https://gcc.gnu.org/PR20301
16070 405. https://gcc.gnu.org/PR20673
16071 406. https://gcc.gnu.org/PR18582
16072 407. https://gcc.gnu.org/PR19340
16073 408. https://gcc.gnu.org/PR21716
16074 409. https://gcc.gnu.org/PR24315
16075 410. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6
16076 411. https://gcc.gnu.org/onlinedocs/
16077 412. mailto:gcc-help@gcc.gnu.org
16078 413. mailto:gcc@gcc.gnu.org
16079 414. https://gcc.gnu.org/lists.html
16080 415. https://www.fsf.org/
16081 416. https://gcc.gnu.org/about.html
16082 417. http://validator.w3.org/check/referer
16083======================================================================
16084http://gcc.gnu.org/gcc-3.3/index.html
16085                             GCC 3.3 Release Series
16086
16087   (This release series is no longer supported.)
16088
16089   May 03, 2005
16090
16091   The [1]GNU project and the GCC developers are pleased to announce the
16092   release of GCC 3.3.6.
16093
16094   This release is a bug-fix release, containing fixes for regressions in
16095   GCC 3.3.5 relative to previous releases of GCC.
16096
16097   This release is the last of the series 3.3.x.
16098
16099   The GCC 3.3 release series includes numerous [2]new features,
16100   improvements, bug fixes, and other changes, thanks to an [3]amazing
16101   group of volunteers.
16102
16103Release History
16104
16105   GCC 3.3.6
16106          May 3, 2005 ([4]changes)
16107
16108   GCC 3.3.5
16109          September 30, 2004 ([5]changes)
16110
16111   GCC 3.3.4
16112          May 31, 2004 ([6]changes)
16113
16114   GCC 3.3.3
16115          February 14, 2004 ([7]changes)
16116
16117   GCC 3.3.2
16118          October 16, 2003 ([8]changes)
16119
16120   GCC 3.3.1
16121          August 8, 2003 ([9]changes)
16122
16123   GCC 3.3
16124          May 14, 2003 ([10]changes)
16125
16126References and Acknowledgements
16127
16128   GCC used to stand for the GNU C Compiler, but since the compiler
16129   supports several other languages aside from C, it now stands for the
16130   GNU Compiler Collection.
16131
16132   A list of [11]successful builds is updated as new information becomes
16133   available.
16134
16135   The GCC developers would like to thank the numerous people that have
16136   contributed new features, improvements, bug fixes, and other changes as
16137   well as test results to GCC. This [12]amazing group of volunteers is
16138   what makes GCC successful.
16139
16140   For additional information about GCC please refer to the [13]GCC
16141   project web site or contact the [14]GCC development mailing list.
16142
16143   To obtain GCC please use [15]our mirror sites, or our CVS server.
16144
16145
16146    For questions related to the use of GCC, please consult these web
16147    pages and the [16]GCC manuals. If that fails, the
16148    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
16149    web pages and the development of GCC are welcome on our developer
16150    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
16151    archives.
16152
16153   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
16154   distribution of this entire article is permitted in any medium,
16155   provided this notice is preserved.
16156
16157   These pages are [21]maintained by the GCC team. Last modified
16158   2021-07-28[22].
16159
16160References
16161
16162   1. http://www.gnu.org/
16163   2. http://gcc.gnu.org/gcc-3.3/changes.html
16164   3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
16165   4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
16166   5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5
16167   6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4
16168   7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3
16169   8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2
16170   9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1
16171  10. http://gcc.gnu.org/gcc-3.3/changes.html
16172  11. http://gcc.gnu.org/gcc-3.3/buildstat.html
16173  12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
16174  13. http://gcc.gnu.org/index.html
16175  14. mailto:gcc@gcc.gnu.org
16176  15. http://gcc.gnu.org/mirrors.html
16177  16. https://gcc.gnu.org/onlinedocs/
16178  17. mailto:gcc-help@gcc.gnu.org
16179  18. mailto:gcc@gcc.gnu.org
16180  19. https://gcc.gnu.org/lists.html
16181  20. https://www.fsf.org/
16182  21. https://gcc.gnu.org/about.html
16183  22. http://validator.w3.org/check/referer
16184======================================================================
16185http://gcc.gnu.org/gcc-3.3/changes.html
16186                             GCC 3.3 Release Series
16187                        Changes, New Features, and Fixes
16188
16189   The latest release in the 3.3 release series is [1]GCC 3.3.6.
16190
16191Caveats
16192
16193     * The preprocessor no longer accepts multi-line string literals. They
16194       were deprecated in 3.0, 3.1, and 3.2.
16195     * The preprocessor no longer supports the -A- switch when appearing
16196       alone. -A- followed by an assertion is still supported.
16197     * Support for all the systems [2]obsoleted in GCC 3.1 has been
16198       removed from GCC 3.3. See below for a [3]list of systems which are
16199       obsoleted in this release.
16200     * Checking for null format arguments has been decoupled from the rest
16201       of the format checking mechanism. Programs which use the format
16202       attribute may regain this functionality by using the new [4]nonnull
16203       function attribute. Note that all functions for which GCC has a
16204       built-in format attribute, an appropriate built-in nonnull
16205       attribute is also applied.
16206     * The DWARF (version 1) debugging format has been deprecated and will
16207       be removed in a future version of GCC. Version 2 of the DWARF
16208       debugging format will continue to be supported for the foreseeable
16209       future.
16210     * The C and Objective-C compilers no longer accept the "Naming Types"
16211       extension (typedef foo = bar); it was already unavailable in C++.
16212       Code which uses it will need to be changed to use the "typeof"
16213       extension instead: typedef typeof(bar) foo. (We have removed this
16214       extension without a period of deprecation because it has caused the
16215       compiler to crash since version 3.0 and no one noticed until very
16216       recently. Thus we conclude it is not in widespread use.)
16217     * The -traditional C compiler option has been removed. It was
16218       deprecated in 3.1 and 3.2. (Traditional preprocessing remains
16219       available.) The <varargs.h> header, used for writing variadic
16220       functions in traditional C, still exists but will produce an error
16221       message if used.
16222     * GCC 3.3.1 automatically places zero-initialized variables in the
16223       .bss section on some operating systems. Versions of GNU Emacs up to
16224       (and including) 21.3 will not work correctly when using this
16225       optimization; you can use -fno-zero-initialized-in-bss to disable
16226       it.
16227
16228General Optimizer Improvements
16229
16230     * A new scheme for accurately describing processor pipelines, the
16231       [5]DFA scheduler, has been added.
16232     * Pavel Nejedly, Charles University Prague, has contributed new file
16233       format used by the edge coverage profiler (-fprofile-arcs).
16234       The new format is robust and diagnoses common mistakes where
16235       profiles from different versions (or compilations) of the program
16236       are combined resulting in nonsensical profiles and slow code to
16237       produced with profile feedback. Additionally this format allows
16238       extra data to be gathered. Currently, overall statistics are
16239       produced helping optimizers to identify hot spots of a program
16240       globally replacing the old intra-procedural scheme and resulting in
16241       better code. Note that the gcov tool from older GCC versions will
16242       not be able to parse the profiles generated by GCC 3.3 and vice
16243       versa.
16244     * Jan Hubicka, SuSE Labs, has contributed a new superblock formation
16245       pass enabled using -ftracer. This pass simplifies the control flow
16246       of functions allowing other optimizations to do better job.
16247       He also contributed the function reordering pass
16248       (-freorder-functions) to optimize function placement using profile
16249       feedback.
16250
16251New Languages and Language specific improvements
16252
16253  C/ObjC/C++
16254
16255     * The preprocessor now accepts directives within macro arguments. It
16256       processes them just as if they had not been within macro arguments.
16257     * The separate ISO and traditional preprocessors have been completely
16258       removed. The front end handles either type of preprocessed output
16259       if necessary.
16260     * In C99 mode preprocessor arithmetic is done in the precision of the
16261       target's intmax_t, as required by that standard.
16262     * The preprocessor can now copy comments inside macros to the output
16263       file when the macro is expanded. This feature, enabled using the
16264       -CC option, is intended for use by applications which place
16265       metadata or directives inside comments, such as lint.
16266     * The method of constructing the list of directories to be searched
16267       for header files has been revised. If a directory named by a -I
16268       option is a standard system include directory, the option is
16269       ignored to ensure that the default search order for system
16270       directories and the special treatment of system header files are
16271       not defeated.
16272     * A few more [6]ISO C99 features now work correctly.
16273     * A new function attribute, nonnull, has been added which allows
16274       pointer arguments to functions to be specified as requiring a
16275       non-null value. The compiler currently uses this information to
16276       issue a warning when it detects a null value passed in such an
16277       argument slot.
16278     * A new type attribute, may_alias, has been added. Accesses to
16279       objects with types with this attribute are not subjected to
16280       type-based alias analysis, but are instead assumed to be able to
16281       alias any other type of objects, just like the char type.
16282
16283  C++
16284
16285     * Type based alias analysis has been implemented for C++ aggregate
16286       types.
16287
16288  Objective-C
16289
16290     * Generate an error if Objective-C objects are passed by value in
16291       function and method calls.
16292     * When -Wselector is used, check the whole list of selectors at the
16293       end of compilation, and emit a warning if a @selector() is not
16294       known.
16295     * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime.
16296     * No longer need to include objc/objc-class.h to compile self calls
16297       in class methods (NeXT runtime only).
16298     * New -Wundeclared-selector option.
16299     * Removed selector bloating which was causing object files to be 10%
16300       bigger on average (GNU runtime only).
16301     * Using at run time @protocol() objects has been fixed in certain
16302       situations (GNU runtime only).
16303     * Type checking has been fixed and improved in many situations
16304       involving protocols.
16305
16306  Java
16307
16308     * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK
16309       1.4) API.
16310     * The JDK 1.4 assert facility has been implemented.
16311     * The bytecode interpreter is now direct threaded and thus faster.
16312
16313  Fortran
16314
16315     * Fortran improvements are listed in [7]the Fortran documentation.
16316
16317  Ada
16318
16319     * Ada tasking now works with glibc 2.3.x threading libraries.
16320
16321New Targets and Target Specific Improvements
16322
16323     * The following changes have been made to the HP-PA port:
16324          + The port now defaults to scheduling for the PA8000 series of
16325            processors.
16326          + Scheduling support for the PA7300 processor has been added.
16327          + The 32-bit port now supports weak symbols under HP-UX 11.
16328          + The handling of initializers and finalizers has been improved
16329            under HP-UX 11. The 64-bit port no longer uses collect2.
16330          + Dwarf2 EH support has been added to the 32-bit GNU/Linux port.
16331          + ABI fixes to correct the passing of small structures by value.
16332     * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to
16333       use the DFA processor pipeline description.
16334     * The following NetBSD configurations for the SuperH processor family
16335       have been added:
16336          + SH3, big-endian, sh-*-netbsdelf*
16337          + SH3, little-endian, shle-*-netbsdelf*
16338          + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd*
16339          + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd*
16340          + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd*
16341          + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd*
16342     * The following changes have been made to the IA-32/x86-64 port:
16343          + SSE2 and 3dNOW! intrinsics are now supported.
16344          + Support for thread local storage has been added to the IA-32
16345            and x86-64 ports.
16346          + The x86-64 port has been significantly improved.
16347     * The following changes have been made to the MIPS port:
16348          + All configurations now accept the -mabi switch. Note that you
16349            will need appropriate multilibs for this option to work
16350            properly.
16351          + ELF configurations will always pass an ABI flag to the
16352            assembler, except when the MIPS EABI is selected.
16353          + -mabi=64 no longer selects MIPS IV code.
16354          + The -mcpu option, which was deprecated in 3.1 and 3.2, has
16355            been removed from this release.
16356          + -march now changes the core ISA level. In previous releases,
16357            it would change the use of processor-specific extensions, but
16358            would leave the core ISA unchanged. For example, mips64-elf
16359            -march=r8000 will now generate MIPS IV code.
16360          + Under most configurations, -mipsN now acts as a synonym for
16361            -march.
16362          + There are some new preprocessor macros to describe the -march
16363            and -mtune settings. See the documentation of those options
16364            for details.
16365          + Support for the NEC VR-Series processors has been added. This
16366            includes the 54xx, 5500, and 41xx series.
16367          + Support for the Sandcraft sr71k processor has been added.
16368     * The following changes have been made to the S/390 port:
16369          + Support to build the Java runtime libraries has been added.
16370            Java is now enabled by default on s390-*-linux* and
16371            s390x-*-linux* targets.
16372          + Multilib support for the s390x-*-linux* target has been added;
16373            this allows to build 31-bit binaries using the -m31 option.
16374          + Support for thread local storage has been added.
16375          + Inline assembler code may now use the 'Q' constraint to
16376            specify memory operands without index register.
16377          + Various platform-specific performance improvements have been
16378            implemented; in particular, the compiler now uses the BRANCH
16379            ON COUNT family of instructions and makes more frequent use of
16380            the TEST UNDER MASK family of instructions.
16381     * The following changes have been made to the PowerPC port:
16382          + Support for IBM Power4 processor added.
16383          + Support for Motorola e500 SPE added.
16384          + Support for AIX 5.2 added.
16385          + Function and Data sections now supported on AIX.
16386          + Sibcall optimizations added.
16387     * The support for H8 Tiny is added to the H8/300 port with -mn.
16388
16389Obsolete Systems
16390
16391   Support for a number of older systems has been declared obsolete in GCC
16392   3.3. Unless there is activity to revive them, the next release of GCC
16393   will have their sources permanently removed.
16394
16395   All configurations of the following processor architectures have been
16396   declared obsolete:
16397     * Matsushita MN10200, mn10200-*-*
16398     * Motorola 88000, m88k-*-*
16399     * IBM ROMP, romp-*-*
16400
16401   Also, some individual systems have been obsoleted:
16402     * Alpha
16403          + Interix, alpha*-*-interix*
16404          + Linux libc1, alpha*-*-linux*libc1*
16405          + Linux ECOFF, alpha*-*-linux*ecoff*
16406     * ARM
16407          + Generic a.out, arm*-*-aout*
16408          + Conix, arm*-*-conix*
16409          + "Old ABI," arm*-*-oabi
16410          + StrongARM/COFF, strongarm-*-coff*
16411     * HPPA (PA-RISC)
16412          + Generic OSF, hppa1.0-*-osf*
16413          + Generic BSD, hppa1.0-*-bsd*
16414          + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]*
16415          + HiUX, hppa*-*-hiux*
16416          + Mach Lites, hppa*-*-lites*
16417     * Intel 386 family
16418          + Windows NT 3.x, i?86-*-win32
16419     * MC68000 family
16420          + HP systems, m68000-hp-bsd* and m68k-hp-bsd*
16421          + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and
16422            m68k-sun-mach*
16423          + AT&T systems, m68000-att-sysv*
16424          + Atari systems, m68k-atari-sysv*
16425          + Motorola systems, m68k-motorola-sysv*
16426          + NCR systems, m68k-ncr-sysv*
16427          + Plexus systems, m68k-plexus-sysv*
16428          + Commodore systems, m68k-cbm-sysv*
16429          + Citicorp TTI, m68k-tti-*
16430          + Unos, m68k-crds-unos*
16431          + Concurrent RTU, m68k-ccur-rtu*
16432          + Linux a.out, m68k-*-linux*aout*
16433          + Linux libc1, m68k-*-linux*libc1*
16434          + pSOS, m68k-*-psos*
16435     * MIPS
16436          + Generic ECOFF, mips*-*-ecoff*
16437          + SINIX, mips-sni-sysv4
16438          + Orion RTEMS, mips64orion-*-rtems*
16439     * National Semiconductor 32000
16440          + OpenBSD, ns32k-*-openbsd*
16441     * POWER (aka RS/6000) and PowerPC
16442          + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]*
16443          + Bull BOSX, rs6000-bull-bosx
16444          + Generic Mach, rs6000-*-mach*
16445          + Generic SysV, powerpc*-*-sysv*
16446          + Linux libc1, powerpc*-*-linux*libc1*
16447     * Sun SPARC
16448          + Generic a.out, sparc-*-aout*, sparclet-*-aout*,
16449            sparclite-*-aout*, and sparc86x-*-aout*
16450          + NetBSD a.out, sparc-*-netbsd*aout*
16451          + Generic BSD, sparc-*-bsd*
16452          + ChorusOS, sparc-*-chorusos*
16453          + Linux a.out, sparc-*-linux*aout*
16454          + Linux libc1, sparc-*-linux*libc1*
16455          + LynxOS, sparc-*-lynxos*
16456          + Solaris on HAL hardware, sparc-hal-solaris2*
16457          + SunOS versions 3 and 4, sparc-*-sunos[34]*
16458     * NEC V850
16459          + RTEMS, v850-*-rtems*
16460     * VAX
16461          + VMS, vax-*-vms*
16462
16463Documentation improvements
16464
16465Other significant improvements
16466
16467     * Almost all front-end dependencies in the compiler have been
16468       separated out into a set of language hooks. This should make adding
16469       a new front end clearer and easier.
16470     * One effect of removing the separate preprocessor is a small
16471       increase in the robustness of the compiler in general, and the
16472       maintainability of target descriptions. Previously target-specific
16473       built-in macros and others, such as __FAST_MATH__, had to be
16474       handled with so-called specs that were hard to maintain. Often they
16475       would fail to behave properly when conflicting options were
16476       supplied on the command line, and define macros in the user's
16477       namespace even when strict ISO compliance was requested.
16478       Integrating the preprocessor has cleanly solved these issues.
16479     * The Makefile suite now supports redirection of make install by
16480       means of the variable DESTDIR.
16481     __________________________________________________________________
16482
16483GCC 3.3
16484
16485   Detailed release notes for the GCC 3.3 release follow.
16486
16487  Bug Fixes
16488
16489    bootstrap failures
16490
16491     * [8]10140 cross compiler build failures: missing __mempcpy (DUP:
16492       [9]10198,[10]10338)
16493
16494    Internal compiler errors (multi-platform)
16495
16496     * [11]3581 large string causes segmentation fault in cc1
16497     * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler
16498     * [13]5533 (c++) ICE when processing std::accumulate(begin, end,
16499       init, invalid_op)
16500     * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out
16501     * [15]6412 (c++) ICE in retrieve_specialization
16502     * [16]6620 (c++) partial template specialization causes an ICE
16503       (segmentation fault)
16504     * [17]6663 (c++) ICE with attribute aligned
16505     * [18]7068 ICE with incomplete types
16506     * [19]7083 (c++) ICE using -gstabs with dodgy class derivation
16507     * [20]7647 (c++) ICE when data member has the name of the enclosing
16508       class
16509     * [21]7675 ICE in fixup_var_refs_1
16510     * [22]7718 'complex' template instantiation causes ICE
16511     * [23]8116 (c++) ICE in member template function
16512     * [24]8358 (ada) Ada compiler accesses freed memory, crashes
16513     * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation
16514       fault
16515     * [26]8564 (c++) ICE in find_function_data, in function.c
16516     * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c
16517     * [28]8766 (c++) ICE after failed initialization of static template
16518       variable
16519     * [29]8803 ICE in instantiate_virtual_regs_1, in function.c
16520     * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set
16521     * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class
16522       definition
16523     * [32]9216 (c++) ICE on missing template parameter
16524     * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c
16525     * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO
16526       loop
16527     * [35]9429 (c++) ICE in template instantiation with a pointered new
16528       operator
16529     * [36]9516 Internal error when using a big array
16530     * [37]9600 (c++) ICE with typedefs in template class
16531     * [38]9629 (c++) virtual inheritance segfault
16532     * [39]9672 (c++) ICE: Error reporting routines re-entered
16533     * [40]9749 (c++) ICE in write_expression on invalid function
16534       prototype
16535     * [41]9794 (fortran) ICE: floating point exception during constant
16536       folding
16537     * [42]9829 (c++) Missing colon in nested namespace usage causes ICE
16538     * [43]9916 (c++) ICE with noreturn function in ?: statement
16539     * [44]9936 ICE with local function and variable-length 2d array
16540     * [45]10262 (c++) cc1plus crashes with large generated code
16541     * [46]10278 (c++) ICE in parser for invalid code
16542     * [47]10446 (c++) ICE on definition of nonexistent member function of
16543       nested class in a class template
16544     * [48]10451 (c++) ICE in grokdeclarator on spurious mutable
16545       declaration
16546     * [49]10506 (c++) ICE in build_new at cp/init.c with
16547       -fkeep-inline-functions and multiple inheritance
16548     * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the
16549       precision of the declared type
16550
16551    Optimization bugs
16552
16553     * [51]2001 Inordinately long compile times in reload CSE regs
16554     * [52]2391 Exponential compilation time explosion in combine
16555     * [53]2960 Duplicate loop conditions even with -Os
16556     * [54]4046 redundant conditional branch
16557     * [55]6405 Loop-unrolling related performance regressions
16558     * [56]6798 very long compile time with large case-statement
16559     * [57]6871 const objects shouldn't be moved to .bss
16560     * [58]6909 problem w/ -Os on modified loop-2c.c test case
16561     * [59]7189 gcc -O2 -Wall does not print ``control reaches end of
16562       non-void function'' warning
16563     * [60]7642 optimization problem with signbit()
16564     * [61]8634 incorrect code for inlining of memcpy under -O2
16565     * [62]8750 Cygwin prolog generation erroneously emitting __alloca as
16566       regular function call
16567
16568    C front end
16569
16570     * [63]2161 long if-else cascade overflows parser stack
16571     * [64]4319 short accepted on typedef'd char
16572     * [65]8602 incorrect line numbers in warning messages when using
16573       inline functions
16574     * [66]9177 -fdump-translation-unit: C front end deletes function_decl
16575       AST nodes and breaks debugging dumps
16576     * [67]9853 miscompilation of non-constant structure initializer
16577
16578    c++ compiler and library
16579
16580     * [68]45 legal template specialization code is rejected (DUP:
16581       [69]3784)
16582     * [70]764 lookup failure: friend operator and dereferencing a pointer
16583       and templates (DUP: [71]5116)
16584     * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP:
16585       2863)
16586     * [73]3663 G++ doesn't check access control during template
16587       instantiation
16588     * [74]3797 gcc fails to emit explicit specialization of a template
16589       member
16590     * [75]3948 Two destructors are called when no copy destructor is
16591       defined (ABI change)
16592     * [76]4137 Conversion operator within template is not accepted
16593     * [77]4361 bogus ambiguity taking the address of a member template
16594     * [78]4802 g++ accepts illegal template code (access to private
16595       member; DUP: [79]5837)
16596     * [80]4803 inline function is used but never defined, and g++ does
16597       not object
16598     * [81]5094 Partial specialization cannot be friend?
16599     * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66
16600     * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run
16601       time
16602     * [84]7015 certain __asm__ constructs rejected
16603     * [85]7086 compile time regression (quadratic behavior in
16604       fixup_var_refs)
16605     * [86]7099 G++ doesn't set the noreturn attribute on std::exit and
16606       std::abort
16607     * [87]7247 copy constructor missing when inlining enabled (invalid
16608       optimization?)
16609     * [88]7441 string array initialization compilation time regression
16610       from seconds to minutes
16611     * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong
16612     * [90]7804 bad printing of floating point constant in warning message
16613     * [91]8099 Friend classes and template specializations
16614     * [92]8117 member function pointers and multiple inheritance
16615     * [93]8205 using declaration and multiple inheritance
16616     * [94]8645 unnecessary non-zero checks in stl_tree.h
16617     * [95]8724 explicit destructor call for incomplete class allowed
16618     * [96]8805 compile time regression with many member variables
16619     * [97]8691 -O3 and -fno-implicit-templates are incompatible
16620     * [98]8700 unhelpful error message for binding temp to reference
16621     * [99]8724 explicit destructor call for incomplete class allowed
16622     * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems
16623     * [101]9016 Failure to consistently constant fold "constant" C++
16624       objects
16625     * [102]9053 g++ confused about ambiguity of overloaded function
16626       templates
16627     * [103]9152 undefined virtual thunks
16628     * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out
16629     * [105]9297 data corruption due to codegen bug (when copying.)
16630     * [106]9318 i/ostream::operator>>/<<(streambuf*) broken
16631     * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf
16632     * [108]9400 bogus -Wshadow warning: shadowed declaration of this in
16633       local classes
16634     * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters
16635     * [110]9425 filebuf::pbackfail broken (DUP: [111]9439)
16636     * [112]9474 GCC freezes in compiling a weird code mixing <iostream>
16637       and <iostream.h>
16638     * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1)
16639       [114][DR 231]
16640     * [115]9555 ostream inserters fail to set badbit on exception
16641     * [116]9561 ostream inserters rethrow exception of wrong type
16642     * [117]9563 ostream::sentry returns true after a failed preparation
16643     * [118]9582 one-definition rule violation in std::allocator
16644     * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors
16645     * [120]9683 bug in initialization chains for static const variables
16646       from template classes
16647     * [121]9791 -Woverloaded-virtual reports hiding of destructor
16648     * [122]9817 collate::compare doesn't handle nul characters
16649     * [123]9825 filebuf::sputbackc breaks sbumpc
16650     * [124]9826 operator>>(basic_istream, basic_string) fails to compile
16651       with custom traits
16652     * [125]9924 Multiple using statements for builtin functions not
16653       allowed
16654     * [126]9946 destructor is not called for temporary object
16655     * [127]9964 filebuf::close() sometimes fails to close file
16656     * [128]9988 filebuf::overflow writes EOF to file
16657     * [129]10033 optimization breaks polymorphic references w/ typeid
16658       operator
16659     * [130]10097 filebuf::underflow drops characters
16660     * [131]10132 filebuf destructor can throw exceptions
16661     * [132]10180 gcc fails to warn about non-inlined function
16662     * [133]10199 method parametrized by template does not work everywhere
16663     * [134]10300 use of array-new (nothrow) in segfaults on NULL return
16664     * [135]10427 Stack corruption with variable-length automatic arrays
16665       and virtual destructors
16666     * [136]10503 Compilation never stops in fixed_type_or_null
16667
16668    Objective-C
16669
16670     * [137]5956 selectors aren't matched properly when added to the
16671       selector table
16672
16673    Fortran compiler and library
16674
16675     * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't
16676       detect
16677     * [139]3924 g77 generates code that is rejected by GAS if COFF debug
16678       info requested
16679     * [140]5634 doc: explain that configure --prefix=~/... does not work
16680     * [141]6367 multiple repeat counts confuse namelist read into array
16681     * [142]6491 Logical operations error on logicals when using
16682       -fugly-logint
16683     * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C"
16684     * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os
16685       on irix6.5
16686     * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should
16687       assume a direct access file
16688     * [146]7278 g77 "bug"; the executable misbehaves (with -O2
16689       -fno-automatic)
16690     * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows
16691     * [148]7388 Incorrect output with 0-based array of characters
16692     * [149]8587 Double complex zero ** double precision number -> NaN
16693       instead of zero
16694     * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning:
16695       unknown register name line-length-none
16696     * [151]10197 Direct access files not unformatted by default
16697
16698    Java compiler and library
16699
16700     * [152]6005 gcj fails to build rhug on alpha
16701     * [153]6389 System.getProperty("") should always throw an
16702       IllegalArgumentException
16703     * [154]6576 java.util.ResourceBundle.getResource ignores locale
16704     * [155]6652 new java.io.File("").getCanonicalFile() throws exception
16705     * [156]7060 getMethod() doesn't search super interface
16706     * [157]7073 bytecode interpreter gives wrong answer for interface
16707       getSuperclass()
16708     * [158]7180 possible bug in
16709       javax.naming.spi.NamingManager.getPlusPath()
16710     * [159]7416 java.security startup refs "GNU libgcj.security"
16711     * [160]7570 Runtime.exec with null envp: child doesn't inherit parent
16712       env (DUP: [161]7578)
16713     * [162]7611 Internal error while compiling libjava with -O
16714     * [163]7709 NullPointerException in _Jv_ResolvePoolEntry
16715     * [164]7766 ZipInputStream.available returns 0 immediately after
16716       construction
16717     * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public
16718     * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented
16719     * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens'
16720     * [168]8234 ZipInputStream chokes when InputStream.read() returns
16721       small chunks
16722     * [169]8415 reflection bug: exception info for Method
16723     * [170]8481 java.Random.nextInt(int) may return negative
16724     * [171]8593 Error reading GZIPped files with BufferedReader
16725     * [172]8759 java.beans.Introspector has no flushCaches() or
16726       flushFromCaches() methods
16727     * [173]8997 spin() calls Thread.sleep
16728     * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd
16729       instead of the root content of C:
16730     * [175]9254 java::lang::Object::wait(), threads-win32.cc returns
16731       wrong return codes
16732     * [176]9271 Severe bias in java.security.SecureRandom
16733
16734    Ada compiler and library
16735
16736     * [177]6767 make gnatlib-shared fails on -laddr2line
16737     * [178]9911 gnatmake fails to link when GCC configured with
16738       --with-sjlj-exceptions=yes
16739     * [179]10020 Can't bootstrap gcc on AIX with Ada enabled
16740     * [180]10546 Ada tasking not working on Red Hat 9
16741
16742    preprocessor
16743
16744     * [181]7029 preprocessor should ignore #warning with -M
16745
16746    ARM-specific
16747
16748     * [182]2903 [arm] Optimization bug with long long arithmetic
16749     * [183]7873 arm-linux-gcc fails when assigning address to a bit field
16750
16751    FreeBSD-specific
16752
16753     * [184]7680 float functions undefined in math.h/cmath with #define
16754       _XOPEN_SOURCE
16755
16756    HP-UX or HP-PA-specific
16757
16758     * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c
16759     * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to
16760       fputc_unlocked
16761     * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen
16762
16763    m68hc11-specific
16764
16765     * [188]6744 Bad assembler code generated: reference to pseudo
16766       register z
16767     * [189]7361 Internal compiler error in reload_cse_simplify_operands,
16768       in reload1.c
16769
16770    MIPS-specific
16771
16772     * [190]9496 [mips-linux] bug in optimizer?
16773
16774    PowerPC-specific
16775
16776     * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of
16777       space
16778     * [192]8480 reload ICEs for LAPACK code on powerpc64-linux
16779     * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg
16780     * [194]10315 [powerpc] ICE: in extract_insn, in recog.c
16781
16782    SPARC-specific
16783
16784     * [195]10267 (documentation) Wrong build instructions for
16785       *-*-solaris2*
16786
16787    x86-specific (Intel/AMD)
16788
16789     * [196]7916 ICE in instantiate_virtual_register_1
16790     * [197]7926 (c++) i486 instructions in header files make c++ programs
16791       crash on i386
16792     * [198]8555 ICE in gen_split_1231
16793     * [199]8994 ICE with -O -march=pentium4
16794     * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs
16795     * [201]9806 ICE in inline assembly with -fPIC flag
16796     * [202]10077 gcc -msse2 generates movd to move dwords between xmm
16797       regs
16798     * [203]10233 64-bit comparison only comparing bottom 32-bits
16799     * [204]10286 type-punning doesn't work with __m64 and -O
16800     * [205]10308 [x86] ICE with -O -fgcse or -O2
16801     __________________________________________________________________
16802
16803GCC 3.3.1
16804
16805  Bug Fixes
16806
16807   This section lists the problem reports (PRs) from GCC's bug tracking
16808   system that are known to be fixed in the 3.3.1 release. This list might
16809   not be complete (that is, it is possible that some PRs that have been
16810   fixed are not listed here).
16811
16812    Bootstrap failures
16813
16814     * [206]11272 [Solaris] make bootstrap fails while building libstdc++
16815
16816    Internal compiler errors (multi-platform)
16817
16818     * [207]5754 ICE on invalid nested template class
16819     * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64
16820       and --enable-checking
16821     * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c
16822     * [210]7053 (c++) ICE when declaring a function already defined as a
16823       friend method of a template class
16824     * [211]8164 (c++) ICE when using different const expressions as
16825       template parameter
16826     * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c
16827     * [213]9559 (c++) ICE with invalid initialization of a static const
16828     * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c
16829       when redeclaring a static member variable
16830     * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in
16831       dwarfout.c with -g -O -finline-functions
16832     * [216]10432 (c++) ICE in poplevel, in cp/decl.c
16833     * [217]10475 ICE in subreg_highpart_offset for code with long long
16834     * [218]10635 (c++) ICE when dereferencing an incomplete type casted
16835       from a void pointer
16836     * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while
16837       instantiating static member variables
16838     * [220]10700 ICE in copy_to_mode_reg on 64-bit targets
16839     * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c
16840     * [222]10796 (c++) ICE when defining an enum with two values: -1 and
16841       MAX_INT_64BIT
16842     * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x
16843       sched.c
16844     * [224]10939 (c++) ICE with template code
16845     * [225]10956 (c++) ICE when specializing a template member function
16846       of a template class, in tsubst, in cp/pt.c
16847     * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*()
16848       defined)
16849     * [227]11059 (c++) ICE with empty union
16850     * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with
16851       -O2 -fnon-call-exceptions
16852     * [229]11105 (c++) ICE in mangle_conv_op_name_for_type
16853     * [230]11149 (c++) ICE on error when instantiation with call function
16854       of a base type
16855     * [231]11228 (c++) ICE on new-expression using array operator new and
16856       default-initialization
16857     * [232]11282 (c++) Infinite memory usage after syntax error
16858     * [233]11301 (fortran) ICE with -fno-globals
16859     * [234]11308 (c++) ICE when using an enum type name as if it were a
16860       class or namespace
16861     * [235]11473 (c++) ICE with -gstabs when empty struct inherits from
16862       an empty struct
16863     * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR
16864     * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c:
16865       template member functions
16866
16867    Optimization bugs
16868
16869     * [238]11198 -O2 -frename-registers generates wrong code (aliasing
16870       problem)
16871     * [239]11304 Wrong code production with -fomit-frame-pointer
16872     * [240]11381 volatile memory access optimized away
16873     * [241]11536 [strength-reduce] -O2 optimization produces wrong code
16874     * [242]11557 constant folding bug generates wrong code
16875
16876    C front end
16877
16878     * [243]5897 No warning for statement after return
16879     * [244]11279 DWARF-2 output mishandles large enums
16880
16881    Preprocessor bugs
16882
16883     * [245]11022 no warning for non-compatible macro redefinition
16884
16885    C++ compiler and library
16886
16887     * [246]2330 static_cast<>() to a private base is allowed
16888     * [247]5388 Incorrect message "operands to ?: have different types"
16889     * [248]5390 Libiberty fails to demangle multi-digit template
16890       parameters
16891     * [249]7877 Incorrect parameter passing to specializations of member
16892       function templates
16893     * [250]9393 Anonymous namespaces and compiling the same file twice
16894     * [251]10032 -pedantic converts some errors to warnings
16895     * [252]10468 const typeof(x) is non-const, but only in templates
16896     * [253]10527 confused error message with "new int()" parameter
16897       initializer
16898     * [254]10679 parameter MIN_INLINE_INSNS is not honored
16899     * [255]10682 gcc chokes on a typedef for an enum inside a class
16900       template
16901     * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of
16902       0.
16903     * [257]10845 template member function (with nested template as
16904       parameter) cannot be called anymore if another unrelated template
16905       member function is defined
16906     * [258]10849 Cannot define an out-of-class specialization of a
16907       private nested template class
16908     * [259]10888 Suppress -Winline warnings for system headers
16909     * [260]10929 -Winline warns about functions for which no definition
16910       is visible
16911     * [261]10931 valid conversion static_cast<const unsigned
16912       int&>(lvalue-of-type-int) is rejected
16913     * [262]10940 Bad code with explicit specialization
16914     * [263]10968 If member function implicitly instantiated, explicit
16915       instantiation of class fails to instantiate it
16916     * [264]10990 Cannot convert with dynamic_cast<> to a private base
16917       class from within a member function
16918     * [265]11039 Bad interaction between implicit typename deprecation
16919       and friendship
16920     * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say
16921       "__unused__" instead
16922     * [267]11095 C++ iostream manipulator causes segfault when called
16923       with negative argument
16924     * [268]11098 g++ doesn't emit complete debugging information for
16925       local variables in destructors
16926     * [269]11137 GNU/Linux shared library constructors not called unless
16927       there's one global object
16928     * [270]11154 spurious ambiguity report for template class
16929       specialization
16930     * [271]11329 Compiler cannot find user defined implicit typecast
16931     * [272]11332 Spurious error with casts in ?: expression
16932     * [273]11431 static_cast behavior with subclasses when default
16933       constructor available
16934     * [274]11528 money_get facet does not accept "$.00" as valid
16935     * [275]11546 Type lookup problems in out-of-line definition of a
16936       class doubly nested from a template class
16937     * [276]11567 C++ code containing templated member function with same
16938       name as pure virtual member function results in linking failure
16939     * [277]11645 Failure to deal with using and private inheritance
16940
16941    Java compiler and library
16942
16943     * [278]5179 Qualified static field access doesn't initialize its
16944       class
16945     * [279]8204 gcj -O2 to native reorders certain instructions
16946       improperly
16947     * [280]10838 java.io.ObjectInputStream syntax error
16948     * [281]10886 The RMI registry that comes with GCJ does not work
16949       correctly
16950     * [282]11349 JNDI URL context factories not located correctly
16951
16952    x86-specific (Intel/AMD)
16953
16954     * [283]4823 ICE on inline assembly code
16955     * [284]8878 miscompilation with -O and SSE
16956     * [285]9815 (c++ library) atomicity.h - fails to compile with -O3
16957       -masm=intel
16958     * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads,
16959       in reload1.c
16960     * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2
16961     * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source
16962     * [289]11044 [x86] out of range loop instructions for FP code on K6
16963     * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE
16964       built-ins
16965     * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC"
16966       is used
16967
16968    SPARC- or Solaris- specific
16969
16970     * [292]9362 solaris 'as' dies when fed .s and "-gstabs"
16971     * [293]10142 [SPARC64] gcc produces wrong code when passing
16972       structures by value
16973     * [294]10663 New configure check aborts with Sun tools.
16974     * [295]10835 combinatorial explosion in scheduler on HyperSPARC
16975     * [296]10876 ICE in calculate_giv_inc when building KDE
16976     * [297]10955 wrong code at -O3 for structure argument in context of
16977       structure return
16978     * [298]11018 -mcpu=ultrasparc busts tar-1.13.25
16979     * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x
16980       Linux kernel
16981
16982    ia64 specific
16983
16984     * [300]10907 gcc violates the ia64 ABI (GP must be preserved)
16985     * [301]11320 scheduler bug (in machine depended reorganization pass)
16986     * [302]11599 bug with conditional and __builtin_prefetch
16987
16988    PowerPC specific
16989
16990     * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem
16991       during loop)
16992     * [304]10871 error in rs6000_stack_info save_size computation
16993     * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse
16994       cures it
16995
16996    m68k-specific
16997
16998     * [306]7594 [m68k] ICE on legal code associated with simplify-rtx
16999     * [307]10557 [m68k] ICE in subreg_offset_representable_p
17000     * [308]11054 [m68k] ICE in reg_overlap_mentioned_p
17001
17002    ARM-specific
17003
17004     * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for
17005       functions with __attribute__ ((interrupt ("IRQ")))
17006     * [310]10842 [arm] Clobbered link register is copied to pc under
17007       certain circumstances
17008     * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes
17009     * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno
17010       (3.4)
17011
17012    MIPS-specific
17013
17014     * [313]11084 ICE in propagate_one_insn, in flow.c
17015
17016    SH-specific
17017
17018     * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf
17019     * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c
17020     * [316]11096 i686-linux to sh-linux cross compiler fails to compile
17021       C++ files
17022
17023    GNU/Linux (or Hurd?) specific
17024
17025     * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3
17026
17027    UnixWare specific
17028
17029     * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare
17030       7.1.1
17031
17032    Cygwin (or mingw) specific
17033
17034     * [319]5287 ICE with dllimport attribute
17035     * [320]10148 [MingW/CygWin] Compiler dumps core
17036
17037    DJGPP specific
17038
17039     * [321]8787 GCC fails to emit .intel_syntax when invoked with
17040       -masm=intel on DJGPP
17041
17042    Darwin (and MacOS X) specific
17043
17044     * [322]10900 trampolines crash
17045
17046    Documentation
17047
17048     * [323]1607 (c++) Format attributes on methods undocumented
17049     * [324]4252 Invalid option `-fdump-translation-unit'
17050     * [325]4490 Clarify restrictions on -m96bit-long-double,
17051       -m128bit-long-double
17052     * [326]10355 document an issue with regparm attribute on some systems
17053       (e.g. Solaris)
17054     * [327]10726 (fortran) Documentation for function "IDate Intrinsic
17055       (Unix)" is wrong
17056     * [328]10805 document bug in old version of Sun assembler
17057     * [329]10815 warn against GNU binutils on AIX
17058     * [330]10877 document need for newer binutils on i?86-*-linux-gnu
17059     * [331]11280 Manual incorrect with respect to -freorder-blocks
17060     * [332]11466 Document -mlittle-endian and its restrictions for the
17061       sparc64 port
17062
17063    Testsuite bugs (compiler itself is not affected)
17064
17065     * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly
17066       report failure
17067     * [334]10810 gcc-3.3 fails make check: buffer overrun in
17068       test_demangle.c
17069     __________________________________________________________________
17070
17071GCC 3.3.2
17072
17073  Bug Fixes
17074
17075   This section lists the problem reports (PRs) from GCC's bug tracker
17076   that are known to be fixed in the 3.3.2 release. This list might not be
17077   complete (that is, it is possible that some PRs that have been fixed
17078   are not listed here).
17079
17080    Bootstrap failures and problems
17081
17082     * [335]8336 [SCO5] bootstrap config still tries to use COFF options
17083     * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with
17084       --enable-threads=posix
17085     * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap
17086     * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare
17087       7.1.1)
17088     * [339]11687 xstormy16-elf build fails in libf2c
17089     * [340]12263 [SGI IRIX] bootstrap fails during compile of
17090       libf2c/libI77/backspace.c
17091     * [341]12490 buffer overflow in scan-decls.c (during Solaris 9
17092       fix-header processing)
17093
17094    Internal compiler errors (multi-platform)
17095
17096     * [342]7277 Casting integers to vector types causes ICE
17097     * [343]7939 (c++) ICE on invalid function template specialization
17098     * [344]11063 (c++) ICE on parsing initialization list of const array
17099       member
17100     * [345]11207 ICE with negative index in array element designator
17101     * [346]11522 (fortran) g77 dwarf-2 ICE in
17102       add_abstract_origin_attribute
17103     * [347]11595 (c++) ICE on duplicate label definition
17104     * [348]11646 (c++) ICE in commit_one_edge_insertion with
17105       -fnon-call-exceptions -fgcse -O
17106     * [349]11665 ICE in struct initializer when taking address
17107     * [350]11852 (c++) ICE with bad struct initializer.
17108     * [351]11878 (c++) ICE in cp_expr_size
17109     * [352]11883 ICE with any -O on mercury-generated C code
17110     * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in
17111       cp/typeck2.c when applying typeid operator to template template
17112       parameter
17113     * [354]12146 ICE in lookup_template_function, in cp/pt.c
17114     * [355]12215 ICE in make_label_edge with -fnon-call-exceptions
17115       -fno-gcse -O2
17116     * [356]12369 (c++) ICE with templates and friends
17117     * [357]12446 ICE in emit_move_insn on complicated array reference
17118     * [358]12510 ICE in final_scan_insn
17119     * [359]12544 ICE with large parameters used in nested functions
17120
17121    C and optimization bugs
17122
17123     * [360]9862 spurious warnings with -W -finline-functions
17124     * [361]10962 lookup_field is a linear search on a linked list (can be
17125       slow if large struct)
17126     * [362]11370 -Wunreachable-code gives false complaints
17127     * [363]11637 invalid assembly with -fnon-call-exceptions
17128     * [364]11885 Problem with bitfields in packed structs
17129     * [365]12082 Inappropriate unreachable code warnings
17130     * [366]12180 Inline optimization fails for variadic function
17131     * [367]12340 loop unroller + gcse produces wrong code
17132
17133    C++ compiler and library
17134
17135     * [368]3907 nested template parameter collides with member name
17136     * [369]5293 confusing message when binding a temporary to a reference
17137     * [370]5296 [DR115] Pointers to functions and to template functions
17138       behave differently in deduction
17139     * [371]7939 ICE on function template specialization
17140     * [372]8656 Unable to assign function with __attribute__ and pointer
17141       return type to an appropriate variable
17142     * [373]10147 Confusing error message for invalid template function
17143       argument
17144     * [374]11400 std::search_n() makes assumptions about Size parameter
17145     * [375]11409 issues with using declarations, overloading, and
17146       built-in functions
17147     * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle
17148       multiple bits in mask
17149     * [377]11786 operator() call on variable in other namespace not
17150       recognized
17151     * [378]11867 static_cast ignores ambiguity
17152     * [379]11928 bug with conversion operators that are typedefs
17153     * [380]12114 Uninitialized memory accessed in dtor
17154     * [381]12163 static_cast + explicit constructor regression
17155     * [382]12181 Wrong code with comma operator and c++
17156     * [383]12236 regparm and fastcall messes up parameters
17157     * [384]12266 incorrect instantiation of unneeded template during
17158       overload resolution
17159     * [385]12296 istream::peek() doesn't set eofbit
17160     * [386]12298 [sjlj exceptions] Stack unwind destroys
17161       not-yet-constructed object
17162     * [387]12369 ICE with templates and friends
17163     * [388]12337 apparently infinite loop in g++
17164     * [389]12344 stdcall attribute ignored if function returns a pointer
17165     * [390]12451 missing(late) class forward declaration in cxxabi.h
17166     * [391]12486 g++ accepts invalid use of a qualified name
17167
17168    x86 specific (Intel/AMD)
17169
17170     * [392]8869 [x86 MMX] ICE with const variable optimization and MMX
17171       builtins
17172     * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions
17173       -O2
17174     * [394]11689 g++3.3 emits un-assembleable code for k6 architecture
17175     * [395]12116 [k6] Invalid assembly output values with X-MAME code
17176     * [396]12070 ICE converting between double and long double with
17177       -msoft-float
17178
17179    ia64-specific
17180
17181     * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc
17182     * [398]11535 __builtin_return_address may not work on ia64
17183     * [399]11693 [ia64] ICE in gen_nop_type
17184     * [400]12224 [ia64] Thread-local storage doesn't work
17185
17186    PowerPC-specific
17187
17188     * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux
17189       kernel
17190     * [402]11319 loop miscompiled on ppc32
17191     * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code
17192
17193    SPARC-specific
17194
17195     * [404]11662 wrong code for expr. with cast to long long and
17196       exclusive or
17197     * [405]11965 invalid assembler code for a shift < 32 operation
17198     * [406]12301 (c++) stack corruption when a returned expression throws
17199       an exception
17200
17201    Alpha-specific
17202
17203     * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of
17204       kernel 2.4.22-pre8
17205
17206    HPUX-specific
17207
17208     * [408]11313 problem with #pragma weak and static inline functions
17209     * [409]11712 __STDC_EXT__ not defined for C++ by default anymore?
17210
17211    Solaris specific
17212
17213     * [410]12166 Profiled programs crash if PROFDIR is set
17214
17215    Solaris-x86 specific
17216
17217     * [411]12101 i386 Solaris no longer works with GNU as?
17218
17219    Miscellaneous embedded target-specific bugs
17220
17221     * [412]10988 [m32r-elf] wrong blockmove code with -O3
17222     * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with
17223       -O2
17224     * [414]11902 [sh4] spec file improperly inserts rpath even when none
17225       needed
17226     * [415]11903 [sh4] -pthread fails to link due to error in spec file
17227       on sh4
17228     __________________________________________________________________
17229
17230GCC 3.3.3
17231
17232  Minor features
17233
17234   In addition to the bug fixes documented below, this release contains
17235   few minor features such as:
17236     * Support for --with-sysroot
17237     * Support for automatic detection of executable stacks
17238     * Support for SSE3 instructions
17239     * Support for thread local storage debugging under GDB on S390
17240
17241  Bug Fixes
17242
17243   This section lists the problem reports (PRs) from GCC's bug tracker
17244   that are known to be fixed in the 3.3.3 release. This list might not be
17245   complete (that is, it is possible that some PRs that have been fixed
17246   are not listed here).
17247
17248    Bootstrap failures and issues
17249
17250     * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails
17251     * [417]12399 boehm-gc fails (when building a cross compiler): libtool
17252       unable to infer tagged configuration
17253     * [418]13068 mklibgcc.in doesn't handle multi-level multilib
17254       subdirectories properly
17255
17256    Internal compiler errors (multi-platform)
17257
17258     * [419]10060 ICE (stack overflow) on huge file (300k lines) due to
17259       recursive behaviour of copy_rtx_if_shared, in emit_rtl.c
17260     * [420]10555 (c++) ICE on undefined template argument
17261     * [421]10706 (c++) ICE in mangle_class_name_for_template
17262     * [422]11496 (fortran) error in flow_loops_find when -funroll-loops
17263       active
17264     * [423]11741 ICE in pre_insert_copy_insn, in gcse.c
17265     * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0
17266     * [425]12632 (fortran) -fbounds-check ICE
17267     * [426]12712 (c++) ICE on short legit C++ code fragment with gcc
17268       3.3.2
17269     * [427]12726 (c++) ICE (segfault) on trivial code
17270     * [428]12890 (c++) ICE on compilation of class with throwing method
17271     * [429]12900 (c++) ICE in rtl_verify_flow_info_1
17272     * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on
17273       correct code with -O2 -fno-force-mem
17274     * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive
17275       template
17276     * [432]13318 ICE: floating point exception in the loop optimizer
17277     * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in
17278       except.c
17279     * [434]13574 (c++) invalid array default initializer in class lets
17280       gcc consume all memory and die
17281     * [435]13475 ICE on SIMD variables with partial value initialization
17282     * [436]13797 (c++) ICE on invalid template parameter
17283     * [437]13824 (java) gcj SEGV with simple .java program
17284
17285    C and optimization bugs
17286
17287     * [438]8776 loop invariants are not removed (most likely)
17288     * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing
17289       strncmp by memcmp
17290     * [440]11350 undefined labels with -Os -fPIC
17291     * [441]12826 Optimizer removes reference through volatile pointer
17292     * [442]12500 stabs debug info: void no longer a predefined / builtin
17293       type
17294     * [443]12941 builtin-bitops-1.c miscompilation (latent bug)
17295     * [444]12953 tree inliner bug (in inline_forbidden_p) and fix
17296     * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled
17297     * [446]13507 spurious printf format warning
17298     * [447]13382 Type information for const pointer disappears during
17299       optimization.
17300     * [448]13394 noreturn attribute ignored on recursive invokation
17301     * [449]13400 Compiled code crashes storing to read-only location
17302     * [450]13521 Endless loop in calculate_global_regs_live
17303
17304    C++ compiler and library
17305
17306   Some of the bug fixes in this list were made to implement decisions
17307   that the ISO C++ standards committee has made concerning several defect
17308   reports (DRs). Links in the list below point to detailed discussion of
17309   the relevant defect report.
17310     * [451]2094 unimplemented: use of `ptrmem_cst' in template type
17311       unification
17312     * [452]2294 using declaration confusion
17313     * [453]5050 template instantiation depth exceeds limit: recursion
17314       problem?
17315     * [454]9371 Bad exception handling in
17316       i/ostream::operator>>/<<(streambuf*)
17317     * [455]9546 bad exception handling in ostream members
17318     * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the
17319       face of unknown locales
17320     * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work
17321     * [459]10095 istream::operator>>(int&) sets ios::badbit when
17322       ios::failbit is set.
17323     * [460]11554 Warning about reordering of initializers doesn't mention
17324       location of constructor
17325     * [461]12297 istream::sentry::sentry() handles eof() incorrectly.
17326     * [462]12352 Exception safety problems in src/localename.cc
17327     * [463]12438 Memory leak in locale::combine()
17328     * [464]12540 Memory leak in locale::locale(const char*)
17329     * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented
17330     * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented
17331     * [470]12696 memory eating infinite loop in diagnostics (error
17332       recovery problem)
17333     * [471]12815 Code compiled with optimization behaves unexpectedly
17334     * [472]12862 Conflicts between typedefs/enums and namespace member
17335       declarations
17336     * [473]12926 Wrong value after assignment in initialize list using
17337       bit-fields
17338     * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented
17339     * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented
17340     * [478]13007 basic_streambuf::pubimbue, imbue wrong
17341     * [479]13009 Implicitly-defined assignment operator writes to wrong
17342       memory
17343     * [480]13057 regparm attribute not applied to destructor
17344     * [481]13070 -Wformat option ignored in g++
17345     * [482]13081 forward template declarations in <complex> let inlining
17346       fail
17347     * [483]13239 Assertion does not seem to work correctly anymore
17348     * [484]13262 "xxx is private within this context" when initializing a
17349       self-contained template class
17350     * [485]13290 simple typo in concept checking for std::generate_n
17351     * [486]13323 Template code does not compile in presence of typedef
17352     * [487]13369 __verify_grouping (and __add_grouping?) not correct
17353     * [488]13371 infinite loop with packed struct and inlining
17354     * [489]13445 Template argument replacement "dereferences" a typedef
17355     * [490]13461 Fails to access protected-ctor from public constant
17356     * [491]13462 Non-standard-conforming type set::pointer
17357     * [492]13478 gcc uses wrong constructor to initialize a const
17358       reference
17359     * [493]13544 "conflicting types" for enums in different scopes
17360     * [494]13650 string::compare should not (always) use
17361       traits_type::length()
17362     * [495]13683 bogus warning about passing non-PODs through ellipsis
17363     * [496]13688 Derived class is denied access to protected base class
17364       member class
17365     * [497]13774 Member variable cleared in virtual multiple inheritance
17366       class
17367     * [498]13884 Protect sstream.tcc from extern template use
17368
17369    Java compiler and library
17370
17371     * [499]10746 [win32] garbage collection crash in GCJ
17372
17373    Objective-C compiler and library
17374
17375     * [500]11433 Crash due to dereferencing null pointer when querying
17376       protocol
17377
17378    Fortran compiler and library
17379
17380     * [501]12633 logical expression gives incorrect result with
17381       -fugly-logint option
17382     * [502]13037 [gcse-lm] g77 generates incorrect code
17383     * [503]13213 Hex constant problem when compiling with -fugly-logint
17384       and -ftypeless-boz
17385
17386    x86-specific (Intel/AMD)
17387
17388     * [504]4490 ICE with -m128bit-long-double
17389     * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have
17390       `reg' in make_field_assignment, in combine.c
17391     * [506]12441 ICE: can't find a register to spill
17392     * [507]12943 array static-init failure under -fpic, -fPIC
17393     * [508]13608 Incorrect code with -O3 -ffast-math
17394
17395    PowerPC-specific
17396
17397     * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of
17398       __attribute__((aligned(16)))
17399     * [510]11793 ICE in extract_insn, in recog.c (const_vector's)
17400     * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in
17401       altivec.md)
17402     * [512]12537 g++ generates writeable text sections
17403
17404    SPARC-specific
17405
17406     * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0
17407       -m64
17408     * [514]12865 mprotect call to make trampoline executable may fail
17409     * [515]13354 ICE in sparc_emit_set_const32
17410
17411    ARM-specific
17412
17413     * [516]10467 [arm] ICE in pre_insert_copy_insn,
17414
17415    ia64-specific
17416
17417     * [517]11226 ICE passing struct arg with two floats
17418     * [518]11227 ICE for _Complex float, _Complex long double args
17419     * [519]12644 GCC 3.3.2 fails to compile glibc on ia64
17420     * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn
17421     * Various fixes for libunwind
17422
17423    Alpha-specific
17424
17425     * [521]12654 Incorrect comparison code generated for Alpha
17426     * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2
17427     * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2
17428
17429    HPPA-specific
17430
17431     * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c
17432     * [525]12158 [hppa] compilation does not terminate at -O1
17433
17434    S390-specific
17435
17436     * [526]11992 Wrong built-in code for memcmp with length 1<<24: only
17437       (1<<24)-1 possible for CLCL-Instruction
17438
17439    SH-specific
17440
17441     * [527]9365 segfault in gen_far_branch (config/sh/sh.c)
17442     * [528]10392 optimizer generates faulty array indexing
17443     * [529]11322 SH profiler outputs multiple definitions of symbol
17444     * [530]13069 gcc/config/sh/rtems.h broken
17445     * [531]13302 Putting a va_list in a struct causes seg fault
17446     * [532]13585 Incorrect optimization of call to sfunc
17447     * Fix inappropriately exported libgcc functions from the shared
17448       library
17449
17450    Other embedded target specific
17451
17452     * [533]8916 [mcore] unsigned char assign gets hosed.
17453     * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c
17454     * [535]13122 [h8300] local variable gets corrupted by function call
17455       when -fomit-frame-pointer is given
17456     * [536]13256 [cris] strict_low_part mistreated in delay slots
17457     * [537]13373 [mcore] optimization with -frerun-cse-after-loop
17458       -fexpensive-optimizations produces wrong code on mcore
17459
17460    GNU HURD-specific
17461
17462     * [538]12561 gcc/config/t-gnu needs updating to work with
17463       --with-sysroot
17464
17465    Tru64 Unix specific
17466
17467     * [539]6243 testsuite fails almost all tests due to no libintl in
17468       LD_LIBRARY_PATH during test.
17469     * [540]11397 weak aliases broken on Tru64 UNIX
17470
17471    AIX-specific
17472
17473     * [541]12505 build failure due to defines of uchar in cpphash.h and
17474       sys/types.h
17475     * [542]13150 WEAK symbols not exported by collect2
17476
17477    IRIX-specific
17478
17479     * [543]12666 fixincludes problem on IRIX 6.5.19m
17480
17481    Solaris-specific
17482
17483     * [544]12969 Including sys/byteorder.h breaks configure checks
17484
17485    Testsuite problems (compiler is not affected)
17486
17487     * [545]10819 testsuite creates CR+LF on compiler version lines in
17488       test summary files
17489     * [546]11612 abi_check not finding correct libgcc_s.so.1
17490
17491    Miscellaneous
17492
17493     * [547]13211 using -###, incorrect warnings about unused linker file
17494       are produced
17495     __________________________________________________________________
17496
17497GCC 3.3.4
17498
17499   This is the [548]list of problem reports (PRs) from GCC's bug tracking
17500   system that are known to be fixed in the 3.3.4 release. This list might
17501   not be complete (that is, it is possible that some PRs that have been
17502   fixed are not listed here).
17503     __________________________________________________________________
17504
17505GCC 3.3.5
17506
17507   This is the [549]list of problem reports (PRs) from GCC's bug tracking
17508   system that are known to be fixed in the 3.3.5 release. This list might
17509   not be complete (that is, it is possible that some PRs that have been
17510   fixed are not listed here).
17511     __________________________________________________________________
17512
17513GCC 3.3.6
17514
17515   This is the [550]list of problem reports (PRs) from GCC's bug tracking
17516   system that are known to be fixed in the 3.3.6 release. This list might
17517   not be complete (that is, it is possible that some PRs that have been
17518   fixed are not listed here).
17519
17520
17521    For questions related to the use of GCC, please consult these web
17522    pages and the [551]GCC manuals. If that fails, the
17523    [552]gcc-help@gcc.gnu.org mailing list might help. Comments on these
17524    web pages and the development of GCC are welcome on our developer
17525    list at [553]gcc@gcc.gnu.org. All of [554]our lists have public
17526    archives.
17527
17528   Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and
17529   distribution of this entire article is permitted in any medium,
17530   provided this notice is preserved.
17531
17532   These pages are [556]maintained by the GCC team. Last modified
17533   2021-07-28[557].
17534
17535References
17536
17537   1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
17538   2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems
17539   3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
17540   4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute
17541   5. http://gcc.gnu.org/news/dfa.html
17542   6. http://gcc.gnu.org/c99status.html
17543   7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html
17544   8. https://gcc.gnu.org/PR10140
17545   9. https://gcc.gnu.org/PR10198
17546  10. https://gcc.gnu.org/PR10338
17547  11. https://gcc.gnu.org/PR3581
17548  12. https://gcc.gnu.org/PR4382
17549  13. https://gcc.gnu.org/PR5533
17550  14. https://gcc.gnu.org/PR6387
17551  15. https://gcc.gnu.org/PR6412
17552  16. https://gcc.gnu.org/PR6620
17553  17. https://gcc.gnu.org/PR6663
17554  18. https://gcc.gnu.org/PR7068
17555  19. https://gcc.gnu.org/PR7083
17556  20. https://gcc.gnu.org/PR7647
17557  21. https://gcc.gnu.org/PR7675
17558  22. https://gcc.gnu.org/PR7718
17559  23. https://gcc.gnu.org/PR8116
17560  24. https://gcc.gnu.org/PR8358
17561  25. https://gcc.gnu.org/PR8511
17562  26. https://gcc.gnu.org/PR8564
17563  27. https://gcc.gnu.org/PR8660
17564  28. https://gcc.gnu.org/PR8766
17565  29. https://gcc.gnu.org/PR8803
17566  30. https://gcc.gnu.org/PR8846
17567  31. https://gcc.gnu.org/PR8906
17568  32. https://gcc.gnu.org/PR9216
17569  33. https://gcc.gnu.org/PR9261
17570  34. https://gcc.gnu.org/PR9263
17571  35. https://gcc.gnu.org/PR9429
17572  36. https://gcc.gnu.org/PR9516
17573  37. https://gcc.gnu.org/PR9600
17574  38. https://gcc.gnu.org/PR9629
17575  39. https://gcc.gnu.org/PR9672
17576  40. https://gcc.gnu.org/PR9749
17577  41. https://gcc.gnu.org/PR9794
17578  42. https://gcc.gnu.org/PR9829
17579  43. https://gcc.gnu.org/PR9916
17580  44. https://gcc.gnu.org/PR9936
17581  45. https://gcc.gnu.org/PR10262
17582  46. https://gcc.gnu.org/PR10278
17583  47. https://gcc.gnu.org/PR10446
17584  48. https://gcc.gnu.org/PR10451
17585  49. https://gcc.gnu.org/PR10506
17586  50. https://gcc.gnu.org/PR10549
17587  51. https://gcc.gnu.org/PR2001
17588  52. https://gcc.gnu.org/PR2391
17589  53. https://gcc.gnu.org/PR2960
17590  54. https://gcc.gnu.org/PR4046
17591  55. https://gcc.gnu.org/PR6405
17592  56. https://gcc.gnu.org/PR6798
17593  57. https://gcc.gnu.org/PR6871
17594  58. https://gcc.gnu.org/PR6909
17595  59. https://gcc.gnu.org/PR7189
17596  60. https://gcc.gnu.org/PR7642
17597  61. https://gcc.gnu.org/PR8634
17598  62. https://gcc.gnu.org/PR8750
17599  63. https://gcc.gnu.org/PR2161
17600  64. https://gcc.gnu.org/PR4319
17601  65. https://gcc.gnu.org/PR8602
17602  66. https://gcc.gnu.org/PR9177
17603  67. https://gcc.gnu.org/PR9853
17604  68. https://gcc.gnu.org/PR45
17605  69. https://gcc.gnu.org/PR3784
17606  70. https://gcc.gnu.org/PR764
17607  71. https://gcc.gnu.org/PR5116
17608  72. https://gcc.gnu.org/PR2862
17609  73. https://gcc.gnu.org/PR3663
17610  74. https://gcc.gnu.org/PR3797
17611  75. https://gcc.gnu.org/PR3948
17612  76. https://gcc.gnu.org/PR4137
17613  77. https://gcc.gnu.org/PR4361
17614  78. https://gcc.gnu.org/PR4802
17615  79. https://gcc.gnu.org/PR5837
17616  80. https://gcc.gnu.org/PR4803
17617  81. https://gcc.gnu.org/PR5094
17618  82. https://gcc.gnu.org/PR5730
17619  83. https://gcc.gnu.org/PR6713
17620  84. https://gcc.gnu.org/PR7015
17621  85. https://gcc.gnu.org/PR7086
17622  86. https://gcc.gnu.org/PR7099
17623  87. https://gcc.gnu.org/PR7247
17624  88. https://gcc.gnu.org/PR7441
17625  89. https://gcc.gnu.org/PR7768
17626  90. https://gcc.gnu.org/PR7804
17627  91. https://gcc.gnu.org/PR8099
17628  92. https://gcc.gnu.org/PR8117
17629  93. https://gcc.gnu.org/PR8205
17630  94. https://gcc.gnu.org/PR8645
17631  95. https://gcc.gnu.org/PR8724
17632  96. https://gcc.gnu.org/PR8805
17633  97. https://gcc.gnu.org/PR8691
17634  98. https://gcc.gnu.org/PR8700
17635  99. https://gcc.gnu.org/PR8724
17636 100. https://gcc.gnu.org/PR8949
17637 101. https://gcc.gnu.org/PR9016
17638 102. https://gcc.gnu.org/PR9053
17639 103. https://gcc.gnu.org/PR9152
17640 104. https://gcc.gnu.org/PR9182
17641 105. https://gcc.gnu.org/PR9297
17642 106. https://gcc.gnu.org/PR9318
17643 107. https://gcc.gnu.org/PR9320
17644 108. https://gcc.gnu.org/PR9400
17645 109. https://gcc.gnu.org/PR9424
17646 110. https://gcc.gnu.org/PR9425
17647 111. https://gcc.gnu.org/PR9439
17648 112. https://gcc.gnu.org/PR9474
17649 113. https://gcc.gnu.org/PR9548
17650 114. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231
17651 115. https://gcc.gnu.org/PR9555
17652 116. https://gcc.gnu.org/PR9561
17653 117. https://gcc.gnu.org/PR9563
17654 118. https://gcc.gnu.org/PR9582
17655 119. https://gcc.gnu.org/PR9622
17656 120. https://gcc.gnu.org/PR9683
17657 121. https://gcc.gnu.org/PR9791
17658 122. https://gcc.gnu.org/PR9817
17659 123. https://gcc.gnu.org/PR9825
17660 124. https://gcc.gnu.org/PR9826
17661 125. https://gcc.gnu.org/PR9924
17662 126. https://gcc.gnu.org/PR9946
17663 127. https://gcc.gnu.org/PR9964
17664 128. https://gcc.gnu.org/PR9988
17665 129. https://gcc.gnu.org/PR10033
17666 130. https://gcc.gnu.org/PR10097
17667 131. https://gcc.gnu.org/PR10132
17668 132. https://gcc.gnu.org/PR10180
17669 133. https://gcc.gnu.org/PR10199
17670 134. https://gcc.gnu.org/PR10300
17671 135. https://gcc.gnu.org/PR10427
17672 136. https://gcc.gnu.org/PR10503
17673 137. https://gcc.gnu.org/PR5956
17674 138. https://gcc.gnu.org/PR1832
17675 139. https://gcc.gnu.org/PR3924
17676 140. https://gcc.gnu.org/PR5634
17677 141. https://gcc.gnu.org/PR6367
17678 142. https://gcc.gnu.org/PR6491
17679 143. https://gcc.gnu.org/PR6742
17680 144. https://gcc.gnu.org/PR7113
17681 145. https://gcc.gnu.org/PR7236
17682 146. https://gcc.gnu.org/PR7278
17683 147. https://gcc.gnu.org/PR7384
17684 148. https://gcc.gnu.org/PR7388
17685 149. https://gcc.gnu.org/PR8587
17686 150. https://gcc.gnu.org/PR9038
17687 151. https://gcc.gnu.org/PR10197
17688 152. https://gcc.gnu.org/PR6005
17689 153. https://gcc.gnu.org/PR6389
17690 154. https://gcc.gnu.org/PR6576
17691 155. https://gcc.gnu.org/PR6652
17692 156. https://gcc.gnu.org/PR7060
17693 157. https://gcc.gnu.org/PR7073
17694 158. https://gcc.gnu.org/PR7180
17695 159. https://gcc.gnu.org/PR7416
17696 160. https://gcc.gnu.org/PR7570
17697 161. https://gcc.gnu.org/PR7578
17698 162. https://gcc.gnu.org/PR7611
17699 163. https://gcc.gnu.org/PR7709
17700 164. https://gcc.gnu.org/PR7766
17701 165. https://gcc.gnu.org/PR7785
17702 166. https://gcc.gnu.org/PR7786
17703 167. https://gcc.gnu.org/PR8142
17704 168. https://gcc.gnu.org/PR8234
17705 169. https://gcc.gnu.org/PR8415
17706 170. https://gcc.gnu.org/PR8481
17707 171. https://gcc.gnu.org/PR8593
17708 172. https://gcc.gnu.org/PR8759
17709 173. https://gcc.gnu.org/PR8997
17710 174. https://gcc.gnu.org/PR9253
17711 175. https://gcc.gnu.org/PR9254
17712 176. https://gcc.gnu.org/PR9271
17713 177. https://gcc.gnu.org/PR6767
17714 178. https://gcc.gnu.org/PR9911
17715 179. https://gcc.gnu.org/PR10020
17716 180. https://gcc.gnu.org/PR10546
17717 181. https://gcc.gnu.org/PR7029
17718 182. https://gcc.gnu.org/PR2903
17719 183. https://gcc.gnu.org/PR7873
17720 184. https://gcc.gnu.org/PR7680
17721 185. https://gcc.gnu.org/PR8705
17722 186. https://gcc.gnu.org/PR9986
17723 187. https://gcc.gnu.org/PR10056
17724 188. https://gcc.gnu.org/PR6744
17725 189. https://gcc.gnu.org/PR7361
17726 190. https://gcc.gnu.org/PR9496
17727 191. https://gcc.gnu.org/PR7067
17728 192. https://gcc.gnu.org/PR8480
17729 193. https://gcc.gnu.org/PR8784
17730 194. https://gcc.gnu.org/PR10315
17731 195. https://gcc.gnu.org/PR10267
17732 196. https://gcc.gnu.org/PR7916
17733 197. https://gcc.gnu.org/PR7926
17734 198. https://gcc.gnu.org/PR8555
17735 199. https://gcc.gnu.org/PR8994
17736 200. https://gcc.gnu.org/PR9426
17737 201. https://gcc.gnu.org/PR9806
17738 202. https://gcc.gnu.org/PR10077
17739 203. https://gcc.gnu.org/PR10233
17740 204. https://gcc.gnu.org/PR10286
17741 205. https://gcc.gnu.org/PR10308
17742 206. https://gcc.gnu.org/PR11272
17743 207. https://gcc.gnu.org/PR5754
17744 208. https://gcc.gnu.org/PR6597
17745 209. https://gcc.gnu.org/PR6949
17746 210. https://gcc.gnu.org/PR7053
17747 211. https://gcc.gnu.org/PR8164
17748 212. https://gcc.gnu.org/PR8384
17749 213. https://gcc.gnu.org/PR9559
17750 214. https://gcc.gnu.org/PR9649
17751 215. https://gcc.gnu.org/PR9864
17752 216. https://gcc.gnu.org/PR10432
17753 217. https://gcc.gnu.org/PR10475
17754 218. https://gcc.gnu.org/PR10635
17755 219. https://gcc.gnu.org/PR10661
17756 220. https://gcc.gnu.org/PR10700
17757 221. https://gcc.gnu.org/PR10712
17758 222. https://gcc.gnu.org/PR10796
17759 223. https://gcc.gnu.org/PR10890
17760 224. https://gcc.gnu.org/PR10939
17761 225. https://gcc.gnu.org/PR10956
17762 226. https://gcc.gnu.org/PR11041
17763 227. https://gcc.gnu.org/PR11059
17764 228. https://gcc.gnu.org/PR11083
17765 229. https://gcc.gnu.org/PR11105
17766 230. https://gcc.gnu.org/PR11149
17767 231. https://gcc.gnu.org/PR11228
17768 232. https://gcc.gnu.org/PR11282
17769 233. https://gcc.gnu.org/PR11301
17770 234. https://gcc.gnu.org/PR11308
17771 235. https://gcc.gnu.org/PR11473
17772 236. https://gcc.gnu.org/PR11503
17773 237. https://gcc.gnu.org/PR11513
17774 238. https://gcc.gnu.org/PR11198
17775 239. https://gcc.gnu.org/PR11304
17776 240. https://gcc.gnu.org/PR11381
17777 241. https://gcc.gnu.org/PR11536
17778 242. https://gcc.gnu.org/PR11557
17779 243. https://gcc.gnu.org/PR5897
17780 244. https://gcc.gnu.org/PR11279
17781 245. https://gcc.gnu.org/PR11022
17782 246. https://gcc.gnu.org/PR2330
17783 247. https://gcc.gnu.org/PR5388
17784 248. https://gcc.gnu.org/PR5390
17785 249. https://gcc.gnu.org/PR7877
17786 250. https://gcc.gnu.org/PR9393
17787 251. https://gcc.gnu.org/PR10032
17788 252. https://gcc.gnu.org/PR10468
17789 253. https://gcc.gnu.org/PR10527
17790 254. https://gcc.gnu.org/PR10679
17791 255. https://gcc.gnu.org/PR10682
17792 256. https://gcc.gnu.org/PR10689
17793 257. https://gcc.gnu.org/PR10845
17794 258. https://gcc.gnu.org/PR10849
17795 259. https://gcc.gnu.org/PR10888
17796 260. https://gcc.gnu.org/PR10929
17797 261. https://gcc.gnu.org/PR10931
17798 262. https://gcc.gnu.org/PR10940
17799 263. https://gcc.gnu.org/PR10968
17800 264. https://gcc.gnu.org/PR10990
17801 265. https://gcc.gnu.org/PR11039
17802 266. https://gcc.gnu.org/PR11062
17803 267. https://gcc.gnu.org/PR11095
17804 268. https://gcc.gnu.org/PR11098
17805 269. https://gcc.gnu.org/PR11137
17806 270. https://gcc.gnu.org/PR11154
17807 271. https://gcc.gnu.org/PR11329
17808 272. https://gcc.gnu.org/PR11332
17809 273. https://gcc.gnu.org/PR11431
17810 274. https://gcc.gnu.org/PR11528
17811 275. https://gcc.gnu.org/PR11546
17812 276. https://gcc.gnu.org/PR11567
17813 277. https://gcc.gnu.org/PR11645
17814 278. https://gcc.gnu.org/PR5179
17815 279. https://gcc.gnu.org/PR8204
17816 280. https://gcc.gnu.org/PR10838
17817 281. https://gcc.gnu.org/PR10886
17818 282. https://gcc.gnu.org/PR11349
17819 283. https://gcc.gnu.org/PR4823
17820 284. https://gcc.gnu.org/PR8878
17821 285. https://gcc.gnu.org/PR9815
17822 286. https://gcc.gnu.org/PR10402
17823 287. https://gcc.gnu.org/PR10504
17824 288. https://gcc.gnu.org/PR10673
17825 289. https://gcc.gnu.org/PR11044
17826 290. https://gcc.gnu.org/PR11089
17827 291. https://gcc.gnu.org/PR11420
17828 292. https://gcc.gnu.org/PR9362
17829 293. https://gcc.gnu.org/PR10142
17830 294. https://gcc.gnu.org/PR10663
17831 295. https://gcc.gnu.org/PR10835
17832 296. https://gcc.gnu.org/PR10876
17833 297. https://gcc.gnu.org/PR10955
17834 298. https://gcc.gnu.org/PR11018
17835 299. https://gcc.gnu.org/PR11556
17836 300. https://gcc.gnu.org/PR10907
17837 301. https://gcc.gnu.org/PR11320
17838 302. https://gcc.gnu.org/PR11599
17839 303. https://gcc.gnu.org/PR9745
17840 304. https://gcc.gnu.org/PR10871
17841 305. https://gcc.gnu.org/PR11440
17842 306. https://gcc.gnu.org/PR7594
17843 307. https://gcc.gnu.org/PR10557
17844 308. https://gcc.gnu.org/PR11054
17845 309. https://gcc.gnu.org/PR10834
17846 310. https://gcc.gnu.org/PR10842
17847 311. https://gcc.gnu.org/PR11052
17848 312. https://gcc.gnu.org/PR11183
17849 313. https://gcc.gnu.org/PR11084
17850 314. https://gcc.gnu.org/PR10331
17851 315. https://gcc.gnu.org/PR10413
17852 316. https://gcc.gnu.org/PR11096
17853 317. https://gcc.gnu.org/PR2873
17854 318. https://gcc.gnu.org/PR3163
17855 319. https://gcc.gnu.org/PR5287
17856 320. https://gcc.gnu.org/PR10148
17857 321. https://gcc.gnu.org/PR8787
17858 322. https://gcc.gnu.org/PR10900
17859 323. https://gcc.gnu.org/PR1607
17860 324. https://gcc.gnu.org/PR4252
17861 325. https://gcc.gnu.org/PR4490
17862 326. https://gcc.gnu.org/PR10355
17863 327. https://gcc.gnu.org/PR10726
17864 328. https://gcc.gnu.org/PR10805
17865 329. https://gcc.gnu.org/PR10815
17866 330. https://gcc.gnu.org/PR10877
17867 331. https://gcc.gnu.org/PR11280
17868 332. https://gcc.gnu.org/PR11466
17869 333. https://gcc.gnu.org/PR10737
17870 334. https://gcc.gnu.org/PR10810
17871 335. https://gcc.gnu.org/PR8336
17872 336. https://gcc.gnu.org/PR9330
17873 337. https://gcc.gnu.org/PR9631
17874 338. https://gcc.gnu.org/PR9877
17875 339. https://gcc.gnu.org/PR11687
17876 340. https://gcc.gnu.org/PR12263
17877 341. https://gcc.gnu.org/PR12490
17878 342. https://gcc.gnu.org/PR7277
17879 343. https://gcc.gnu.org/PR7939
17880 344. https://gcc.gnu.org/PR11063
17881 345. https://gcc.gnu.org/PR11207
17882 346. https://gcc.gnu.org/PR11522
17883 347. https://gcc.gnu.org/PR11595
17884 348. https://gcc.gnu.org/PR11646
17885 349. https://gcc.gnu.org/PR11665
17886 350. https://gcc.gnu.org/PR11852
17887 351. https://gcc.gnu.org/PR11878
17888 352. https://gcc.gnu.org/PR11883
17889 353. https://gcc.gnu.org/PR11991
17890 354. https://gcc.gnu.org/PR12146
17891 355. https://gcc.gnu.org/PR12215
17892 356. https://gcc.gnu.org/PR12369
17893 357. https://gcc.gnu.org/PR12446
17894 358. https://gcc.gnu.org/PR12510
17895 359. https://gcc.gnu.org/PR12544
17896 360. https://gcc.gnu.org/PR9862
17897 361. https://gcc.gnu.org/PR10962
17898 362. https://gcc.gnu.org/PR11370
17899 363. https://gcc.gnu.org/PR11637
17900 364. https://gcc.gnu.org/PR11885
17901 365. https://gcc.gnu.org/PR12082
17902 366. https://gcc.gnu.org/PR12180
17903 367. https://gcc.gnu.org/PR12340
17904 368. https://gcc.gnu.org/PR3907
17905 369. https://gcc.gnu.org/PR5293
17906 370. https://gcc.gnu.org/PR5296
17907 371. https://gcc.gnu.org/PR7939
17908 372. https://gcc.gnu.org/PR8656
17909 373. https://gcc.gnu.org/PR10147
17910 374. https://gcc.gnu.org/PR11400
17911 375. https://gcc.gnu.org/PR11409
17912 376. https://gcc.gnu.org/PR11740
17913 377. https://gcc.gnu.org/PR11786
17914 378. https://gcc.gnu.org/PR11867
17915 379. https://gcc.gnu.org/PR11928
17916 380. https://gcc.gnu.org/PR12114
17917 381. https://gcc.gnu.org/PR12163
17918 382. https://gcc.gnu.org/PR12181
17919 383. https://gcc.gnu.org/PR12236
17920 384. https://gcc.gnu.org/PR12266
17921 385. https://gcc.gnu.org/PR12296
17922 386. https://gcc.gnu.org/PR12298
17923 387. https://gcc.gnu.org/PR12369
17924 388. https://gcc.gnu.org/PR12337
17925 389. https://gcc.gnu.org/PR12344
17926 390. https://gcc.gnu.org/PR12451
17927 391. https://gcc.gnu.org/PR12486
17928 392. https://gcc.gnu.org/PR8869
17929 393. https://gcc.gnu.org/PR9786
17930 394. https://gcc.gnu.org/PR11689
17931 395. https://gcc.gnu.org/PR12116
17932 396. https://gcc.gnu.org/PR12070
17933 397. https://gcc.gnu.org/PR11184
17934 398. https://gcc.gnu.org/PR11535
17935 399. https://gcc.gnu.org/PR11693
17936 400. https://gcc.gnu.org/PR12224
17937 401. https://gcc.gnu.org/PR11087
17938 402. https://gcc.gnu.org/PR11319
17939 403. https://gcc.gnu.org/PR11949
17940 404. https://gcc.gnu.org/PR11662
17941 405. https://gcc.gnu.org/PR11965
17942 406. https://gcc.gnu.org/PR12301
17943 407. https://gcc.gnu.org/PR11717
17944 408. https://gcc.gnu.org/PR11313
17945 409. https://gcc.gnu.org/PR11712
17946 410. https://gcc.gnu.org/PR12166
17947 411. https://gcc.gnu.org/PR12101
17948 412. https://gcc.gnu.org/PR10988
17949 413. https://gcc.gnu.org/PR11805
17950 414. https://gcc.gnu.org/PR11902
17951 415. https://gcc.gnu.org/PR11903
17952 416. https://gcc.gnu.org/PR11890
17953 417. https://gcc.gnu.org/PR12399
17954 418. https://gcc.gnu.org/PR13068
17955 419. https://gcc.gnu.org/PR10060
17956 420. https://gcc.gnu.org/PR10555
17957 421. https://gcc.gnu.org/PR10706
17958 422. https://gcc.gnu.org/PR11496
17959 423. https://gcc.gnu.org/PR11741
17960 424. https://gcc.gnu.org/PR12440
17961 425. https://gcc.gnu.org/PR12632
17962 426. https://gcc.gnu.org/PR12712
17963 427. https://gcc.gnu.org/PR12726
17964 428. https://gcc.gnu.org/PR12890
17965 429. https://gcc.gnu.org/PR12900
17966 430. https://gcc.gnu.org/PR13060
17967 431. https://gcc.gnu.org/PR13289
17968 432. https://gcc.gnu.org/PR13318
17969 433. https://gcc.gnu.org/PR13392
17970 434. https://gcc.gnu.org/PR13574
17971 435. https://gcc.gnu.org/PR13475
17972 436. https://gcc.gnu.org/PR13797
17973 437. https://gcc.gnu.org/PR13824
17974 438. https://gcc.gnu.org/PR8776
17975 439. https://gcc.gnu.org/PR10339
17976 440. https://gcc.gnu.org/PR11350
17977 441. https://gcc.gnu.org/PR12826
17978 442. https://gcc.gnu.org/PR12500
17979 443. https://gcc.gnu.org/PR12941
17980 444. https://gcc.gnu.org/PR12953
17981 445. https://gcc.gnu.org/PR13041
17982 446. https://gcc.gnu.org/PR13507
17983 447. https://gcc.gnu.org/PR13382
17984 448. https://gcc.gnu.org/PR13394
17985 449. https://gcc.gnu.org/PR13400
17986 450. https://gcc.gnu.org/PR13521
17987 451. https://gcc.gnu.org/PR2094
17988 452. https://gcc.gnu.org/PR2294
17989 453. https://gcc.gnu.org/PR5050
17990 454. https://gcc.gnu.org/PR9371
17991 455. https://gcc.gnu.org/PR9546
17992 456. https://gcc.gnu.org/PR10081
17993 457. https://gcc.gnu.org/PR10093
17994 458. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61
17995 459. https://gcc.gnu.org/PR10095
17996 460. https://gcc.gnu.org/PR11554
17997 461. https://gcc.gnu.org/PR12297
17998 462. https://gcc.gnu.org/PR12352
17999 463. https://gcc.gnu.org/PR12438
18000 464. https://gcc.gnu.org/PR12540
18001 465. https://gcc.gnu.org/PR12594
18002 466. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60
18003 467. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63
18004 468. https://gcc.gnu.org/PR12657
18005 469. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292
18006 470. https://gcc.gnu.org/PR12696
18007 471. https://gcc.gnu.org/PR12815
18008 472. https://gcc.gnu.org/PR12862
18009 473. https://gcc.gnu.org/PR12926
18010 474. https://gcc.gnu.org/PR12967
18011 475. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html
18012 476. https://gcc.gnu.org/PR12971
18013 477. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328
18014 478. https://gcc.gnu.org/PR13007
18015 479. https://gcc.gnu.org/PR13009
18016 480. https://gcc.gnu.org/PR13057
18017 481. https://gcc.gnu.org/PR13070
18018 482. https://gcc.gnu.org/PR13081
18019 483. https://gcc.gnu.org/PR13239
18020 484. https://gcc.gnu.org/PR13262
18021 485. https://gcc.gnu.org/PR13290
18022 486. https://gcc.gnu.org/PR13323
18023 487. https://gcc.gnu.org/PR13369
18024 488. https://gcc.gnu.org/PR13371
18025 489. https://gcc.gnu.org/PR13445
18026 490. https://gcc.gnu.org/PR13461
18027 491. https://gcc.gnu.org/PR13462
18028 492. https://gcc.gnu.org/PR13478
18029 493. https://gcc.gnu.org/PR13544
18030 494. https://gcc.gnu.org/PR13650
18031 495. https://gcc.gnu.org/PR13683
18032 496. https://gcc.gnu.org/PR13688
18033 497. https://gcc.gnu.org/PR13774
18034 498. https://gcc.gnu.org/PR13884
18035 499. https://gcc.gnu.org/PR10746
18036 500. https://gcc.gnu.org/PR11433
18037 501. https://gcc.gnu.org/PR12633
18038 502. https://gcc.gnu.org/PR13037
18039 503. https://gcc.gnu.org/PR13213
18040 504. https://gcc.gnu.org/PR4490
18041 505. https://gcc.gnu.org/PR12292
18042 506. https://gcc.gnu.org/PR12441
18043 507. https://gcc.gnu.org/PR12943
18044 508. https://gcc.gnu.org/PR13608
18045 509. https://gcc.gnu.org/PR11598
18046 510. https://gcc.gnu.org/PR11793
18047 511. https://gcc.gnu.org/PR12467
18048 512. https://gcc.gnu.org/PR12537
18049 513. https://gcc.gnu.org/PR12496
18050 514. https://gcc.gnu.org/PR12865
18051 515. https://gcc.gnu.org/PR13354
18052 516. https://gcc.gnu.org/PR10467
18053 517. https://gcc.gnu.org/PR11226
18054 518. https://gcc.gnu.org/PR11227
18055 519. https://gcc.gnu.org/PR12644
18056 520. https://gcc.gnu.org/PR13149
18057 521. https://gcc.gnu.org/PR12654
18058 522. https://gcc.gnu.org/PR12965
18059 523. https://gcc.gnu.org/PR13031
18060 524. https://gcc.gnu.org/PR11634
18061 525. https://gcc.gnu.org/PR12158
18062 526. https://gcc.gnu.org/PR11992
18063 527. https://gcc.gnu.org/PR9365
18064 528. https://gcc.gnu.org/PR10392
18065 529. https://gcc.gnu.org/PR11322
18066 530. https://gcc.gnu.org/PR13069
18067 531. https://gcc.gnu.org/PR13302
18068 532. https://gcc.gnu.org/PR13585
18069 533. https://gcc.gnu.org/PR8916
18070 534. https://gcc.gnu.org/PR11576
18071 535. https://gcc.gnu.org/PR13122
18072 536. https://gcc.gnu.org/PR13256
18073 537. https://gcc.gnu.org/PR13373
18074 538. https://gcc.gnu.org/PR12561
18075 539. https://gcc.gnu.org/PR6243
18076 540. https://gcc.gnu.org/PR11397
18077 541. https://gcc.gnu.org/PR12505
18078 542. https://gcc.gnu.org/PR13150
18079 543. https://gcc.gnu.org/PR12666
18080 544. https://gcc.gnu.org/PR12969
18081 545. https://gcc.gnu.org/PR10819
18082 546. https://gcc.gnu.org/PR11612
18083 547. https://gcc.gnu.org/PR13211
18084 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4
18085 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5
18086 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6
18087 551. https://gcc.gnu.org/onlinedocs/
18088 552. mailto:gcc-help@gcc.gnu.org
18089 553. mailto:gcc@gcc.gnu.org
18090 554. https://gcc.gnu.org/lists.html
18091 555. https://www.fsf.org/
18092 556. https://gcc.gnu.org/about.html
18093 557. http://validator.w3.org/check/referer
18094======================================================================
18095http://gcc.gnu.org/gcc-3.2/index.html
18096                             GCC 3.2 Release Series
18097
18098   (This release series is no longer supported.)
18099
18100   April 25, 2003
18101
18102   The [1]GNU project and the GCC developers are pleased to announce the
18103   release of GCC 3.2.3.
18104
18105   The purpose of the GCC 3.2 release series is to provide a stable
18106   platform for OS distributors to use building their next releases. A
18107   primary objective was to stabilize the C++ ABI; we believe that the
18108   interface to the compiler and the C++ standard library are now
18109   relatively stable.
18110
18111   Be aware that C++ code compiled by GCC 3.2.x will (in general) not
18112   interoperate with code compiled by GCC 3.1.1 or earlier.
18113
18114   Please refer to our [2]detailed list of news, caveats, and bug-fixes
18115   for further information.
18116
18117Release History
18118
18119   GCC 3.2.3
18120          April 25, 2003 ([3]changes)
18121
18122   GCC 3.2.2
18123          February 5, 2003 ([4]changes)
18124
18125   GCC 3.2.1
18126          November 19, 2002 ([5]changes)
18127
18128   GCC 3.2
18129          August 14, 2002 ([6]changes)
18130
18131References and Acknowledgements
18132
18133   GCC used to stand for the GNU C Compiler, but since the compiler
18134   supports several other languages aside from C, it now stands for the
18135   GNU Compiler Collection.
18136
18137   A list of [7]successful builds is updated as new information becomes
18138   available.
18139
18140   The GCC developers would like to thank the numerous people that have
18141   contributed new features, improvements, bug fixes, and other changes as
18142   well as test results to GCC. This [8]amazing group of volunteers is
18143   what makes GCC successful.
18144
18145   For additional information about GCC please refer to the [9]GCC project
18146   web site or contact the [10]GCC development mailing list.
18147
18148   To obtain GCC please use [11]our mirror sites, or our CVS server.
18149
18150
18151    For questions related to the use of GCC, please consult these web
18152    pages and the [12]GCC manuals. If that fails, the
18153    [13]gcc-help@gcc.gnu.org mailing list might help. Comments on these
18154    web pages and the development of GCC are welcome on our developer
18155    list at [14]gcc@gcc.gnu.org. All of [15]our lists have public
18156    archives.
18157
18158   Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and
18159   distribution of this entire article is permitted in any medium,
18160   provided this notice is preserved.
18161
18162   These pages are [17]maintained by the GCC team. Last modified
18163   2021-07-28[18].
18164
18165References
18166
18167   1. http://www.gnu.org/
18168   2. http://gcc.gnu.org/gcc-3.2/changes.html
18169   3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
18170   4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2
18171   5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1
18172   6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2
18173   7. http://gcc.gnu.org/gcc-3.2/buildstat.html
18174   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
18175   9. http://gcc.gnu.org/index.html
18176  10. mailto:gcc@gcc.gnu.org
18177  11. http://gcc.gnu.org/mirrors.html
18178  12. https://gcc.gnu.org/onlinedocs/
18179  13. mailto:gcc-help@gcc.gnu.org
18180  14. mailto:gcc@gcc.gnu.org
18181  15. https://gcc.gnu.org/lists.html
18182  16. https://www.fsf.org/
18183  17. https://gcc.gnu.org/about.html
18184  18. http://validator.w3.org/check/referer
18185======================================================================
18186http://gcc.gnu.org/gcc-3.2/changes.html
18187                             GCC 3.2 Release Series
18188                        Changes, New Features, and Fixes
18189
18190   The latest release in the 3.2 release series is [1]GCC 3.2.3.
18191
18192Caveats and New Features
18193
18194  Caveats
18195
18196     * The C++ compiler does not correctly zero-initialize
18197       pointers-to-data members. You must explicitly initialize them. For
18198       example: int S::*m(0); will work, but depending on
18199       default-initialization to zero will not work. This bug cannot be
18200       fixed in GCC 3.2 without inducing unacceptable risks. It will be
18201       fixed in GCC 3.3.
18202     * This GCC release is based on the GCC 3.1 sourcebase, and thus has
18203       all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has
18204       a number of C++ ABI fixes which make its C++ compiler generate
18205       binary code which is incompatible with the C++ compilers found in
18206       earlier GCC releases, including GCC 3.1 and GCC 3.1.1.
18207
18208  Frontend Enhancements
18209
18210    C/C++/Objective-C
18211
18212     * The method of constructing the list of directories to be searched
18213       for header files has been revised. If a directory named by a -I
18214       option is a standard system include directory, the option is
18215       ignored to ensure that the default search order for system
18216       directories and the special treatment of system header files are
18217       not defeated.
18218     * The C and Objective-C compilers no longer accept the "Naming Types"
18219       extension (typedef foo = bar); it was already unavailable in C++.
18220       Code which uses it will need to be changed to use the "typeof"
18221       extension instead: typedef typeof(bar) foo. (We have removed this
18222       extension without a period of deprecation because it has caused the
18223       compiler to crash since version 3.0 and no one noticed until very
18224       recently. Thus we conclude it is not in widespread use.)
18225
18226    C++
18227
18228     * GCC 3.2 fixed serveral differences between the C++ ABI implemented
18229       in GCC and the multi-vendor standard, but more have been found
18230       since the release. 3.2.1 adds a new warning, -Wabi, to warn about
18231       code which is affected by these bugs. We will fix these bugs in
18232       some future release, once we are confident that all have been
18233       found; until then, it is our intention to make changes to the ABI
18234       only if they are necessary for correct compilation of C++, as
18235       opposed to conformance to the ABI documents.
18236     * For details on how to build an ABI compliant compiler for GNU/Linux
18237       systems, check the [3]common C++ ABI page.
18238
18239  New Targets and Target Specific Improvements
18240
18241    IA-32
18242
18243     * Fixed a number of bugs in SSE and MMX intrinsics.
18244     * Fixed common compiler crashes with SSE instruction set enabled
18245       (implied by -march=pentium3, pentium4, athlon-xp)
18246     * __m128 and __m128i is not 128bit aligned when used in structures.
18247
18248    x86-64
18249
18250     * A bug whereby the compiler could generate bad code for bzero has
18251       been fixed.
18252     * ABI fixes (implying ABI incompatibilities with previous version in
18253       some corner cases)
18254     * Fixed prefetch code generation
18255     __________________________________________________________________
18256
18257GCC 3.2.3
18258
18259   3.2.3 is a bug fix release only; there are no new features that were
18260   not present in GCC 3.2.2.
18261
18262  Bug Fixes
18263
18264   This section lists the problem reports (PRs) from GCC's bug tracking
18265   system that are known to be fixed in the 3.2.3 release. This list might
18266   not be complete (that is, it is possible that some PRs that have been
18267   fixed are not listed here), and some of the titles have been changed to
18268   make them more clear.
18269
18270    Internal Compiler Errors (multi-platform)
18271
18272     * [4]3782: (c++) -quiet -fstats produces a segmentation fault in
18273       cc1plus
18274     * [5]6440: (c++) template specializations cause ICE
18275     * [6]7050: (c++) ICE on: (i ? get_string() : throw)
18276     * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c)
18277     * [8]7982: (c++) ICE due to infinite recursion (using STL set)
18278     * [9]8068: exceedingly high (infinite) memory usage
18279     * [10]8178: ICE with __builtin_ffs
18280     * [11]8396: ICE in copy_to_mode_reg, in explow.c
18281     * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c
18282     * [13]9768: ICE when optimizing inline code at -O2
18283     * [14]9798: (c++) Infinite recursion (segfault) in
18284       cp/decl.c:push_using_directive with recursive using directives
18285     * [15]9799: mismatching structure initializer with nested flexible
18286       array member: ICE
18287     * [16]9928: ICE on duplicate enum declaration
18288     * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects
18289       sparc, alpha)
18290     * [18]10352: ICE in find_reloads_toplev
18291     * [19]10336: ICE with -Wunreachable-code
18292
18293    C/optimizer bugs:
18294
18295     * [20]8224: Incorrect joining of signed and unsigned division
18296     * [21]8613: -O2 produces wrong code with builtin strlen and
18297       postincrements
18298     * [22]8828: gcc reports some code is unreachable when it is not
18299     * [23]9226: GCSE breaking argument passing
18300     * [24]9853: miscompilation of non-constant structure initializer
18301     * [25]9797: C99-style struct initializers are miscompiled
18302     * [26]9967: Some standard C function calls should not be replaced
18303       when optimizing for size
18304     * [27]10116: ce2: invalid merge of join_bb in the context of switch
18305       statements
18306     * [28]10171: wrong code for inlined function
18307     * [29]10175: -Wunreachable-code doesn't work for single lines
18308
18309    C++ compiler and library:
18310
18311     * [30]8316: Confusing diagnostic for code that misuses conversion
18312       operators
18313     * [31]9169: filebuf output fails if codecvt<>::out returns noconv
18314     * [32]9420: incomplete type incorrectly reported
18315     * [33]9459: typeof in return type specification of template not
18316       supported
18317     * [34]9507: filebuf::open handles ios_base::ate incorrectly
18318     * [35]9538: Out-of-bounds memory access in streambuf::sputbackc
18319     * [36]9602: Total confusion about template/friend/virtual/abstract
18320     * [37]9993: destructor not called for local object created within and
18321       returned from infinite loop
18322     * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2
18323       system
18324
18325    Java compiler and library:
18326
18327     * [39]9652: libgcj build fails on irix6.5.1[78]
18328     * [40]10144: gas on solaris complains about bad .stabs lines for
18329       java, native as unaffected
18330
18331    x86-specific (Intel/AMD):
18332
18333     * [41]8746: gcc miscompiles Linux kernel ppa driver on x86
18334     * [42]9888: -mcpu=k6 -Os produces out of range loop instructions
18335     * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu
18336       failed
18337     * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib)
18338       failed
18339
18340    SPARC-specific:
18341
18342     * [45]7784: [Sparc] ICE in extract_insn, in recog.c
18343     * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in
18344       unroll.c
18345     * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc
18346     * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in
18347       execute/loop-2d.c
18348     * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc
18349     * [50]9414: Scheduling bug on Ultrasparc
18350     * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64
18351
18352    m68k-specific:
18353
18354     * [52]7248: broken "inclusive or" code
18355     * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1
18356
18357    PowerPC-specific:
18358
18359     * [54]9732: Wrong code with -O2 -fPIC
18360     * [55]10073: ICE: powerpc cannot split insn
18361
18362    Alpha-specific:
18363
18364     * [56]7702: optimization problem on a DEC alpha under OSF1
18365     * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system
18366
18367    HP-specific:
18368
18369     * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275)
18370     * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10
18371       (missing symbol)
18372     * [60]10271: Floating point args don't get reloaded across function
18373       calls with -O2
18374
18375    MIPS specific:
18376
18377     * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in
18378       compile/920501-4.c
18379
18380    CRIS specific:
18381
18382     * [62]10377: gcc-3.2.2 creates bad assembler code for cris
18383
18384    Miscellaneous and minor bugs:
18385
18386     * [63]6955: collect2 says "core dumped" when there is no core
18387     __________________________________________________________________
18388
18389GCC 3.2.2
18390
18391   Beginning with 3.2.2, GCC's Makefile suite supports redirection of make
18392   install by means of the DESTDIR variable. Parts of the GCC tree have
18393   featured that support long before, but now it is available even from
18394   the top level.
18395
18396   Other than that, GCC 3.2.2 is a bug fix release only; there are no new
18397   features that were not present in GCC 3.2.1.
18398
18399  Bug Fixes
18400
18401   On the following i386-based systems GCC 3.2.1 broke the C ABI wrt.
18402   functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped
18403   with FreeBSD 5.0 does not have this problem), Interix, a.out-based
18404   GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI
18405   change, and thus restores ABI-compatibility with previous releases
18406   (except GCC 3.2.1) on these platforms.
18407
18408   This section lists the problem reports (PRs) from GCC's bug tracking
18409   system that are known to be fixed in the 3.2.2 release. This list might
18410   not be complete (that is, it is possible that some PRs that have been
18411   fixed are not listed here) and some of the titles have been changed to
18412   make them more clear.
18413
18414    Internal Compiler Errors (multi-platform)
18415
18416     * [64]5919: (c++) ICE when passing variable array to template
18417       function
18418     * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=)
18419     * [66]7507: ICE with -O2 when address of called function is a
18420       complicated expression
18421     * [67]7622: ICE with nested inline functions if function's address is
18422       taken
18423     * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR
18424       [69]9258)
18425     * [70]8031: (c++) ICE in code comparing typeids and casting from
18426       virtual base
18427     * [71]8275: ICE in simplify_subreg
18428     * [72]8332: (c++) builtin strlen/template interaction causes ICE
18429     * [73]8372: (c++) ICE on explicit call of destructor
18430     * [74]8439: (c, not c++) empty struct causes ICE
18431     * [75]8442: (c++) ICE with nested template classes
18432     * [76]8518: ICE when compiling mplayer ("extern inline" issue)
18433     * [77]8615: (c++) ICE with out-of-range character constant template
18434       argument
18435     * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307
18436     * [79]8799: (c++) ICE: error reporting routines re-entered
18437     * [80]9328: (c++) ICE with typeof(X) for overloaded X
18438     * [81]9465: (preprocessor) cpp -traditional ICE on null bytes
18439
18440    C++ (compiler and library) bugs
18441
18442     * [82]47: scoping in nested classes is broken
18443     * [83]6745: problems with iostream rdbuf() member function
18444     * [84]8214: conversion from const char* const to char* sometimes
18445       accepted illegally
18446     * [85]8493: builtin strlen and overload resolution (same bug as
18447       [86]8332)
18448     * [87]8503: strange behaviour of function types
18449     * [88]8727: compiler confused by inheritance from an anonymous struct
18450     * [89]7445: poor performance of std::locale::classic() in
18451       multi-threaded applications
18452     * [90]8230: mishandling of overflow in vector<T>::resize
18453     * [91]8399: sync_with_stdio(false) breaks unformatted input
18454     * [92]8662: illegal access of private member of unnamed class is
18455       accepted
18456     * [93]8707: "make distclean" fails in libstdc++-v3 directory
18457     * [94]8708: __USE_MALLOC doesn't work
18458     * [95]8790: Use of non-thread-safe strtok in src/localename.cc
18459     * [96]8887: Bug in date formats with --enable-clocale=generic
18460     * [97]9076: Call Frame Instructions are not handled correctly during
18461       unwind operation
18462     * [98]9151: std::setprecision limited to 16 digits when outputting a
18463       double to a stream
18464     * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers
18465     * [100]9269: libstdc++ headers: explicit specialization of function
18466       must precede its first use
18467     * [101]9322: return value of basic_streambuf<>::getloc affected by
18468       locale::global
18469     * [102]9433: segfault in runtime support for dynamic_cast
18470
18471    C and optimizer bugs
18472
18473     * [103]8032: GCC incorrectly initializes static structs that have
18474       flexible arrays
18475     * [104]8639: simple arithmetic expression broken
18476     * [105]8794: optimization improperly eliminates certain expressions
18477     * [106]8832: traditional "asm volatile" code is illegally optimized
18478     * [107]8988: loop optimizer bug: with -O2, code is generated that
18479       segfaults (found on i386, bug present for all platforms)
18480     * [108]9492: structure copy clobbers subsequent stores to structure
18481
18482    Objective-C bugs
18483
18484     * [109]9267: Objective-C parser won't build with newer bison versions
18485       (e.g. 1.875)
18486
18487    Ada bugs
18488
18489     * [110]8344: Ada build problem due to conflict between gcc/final.o,
18490       gcc/ada/final.o
18491
18492    Preprocessor bugs
18493
18494     * [111]8524: _Pragma within macros is improperly expanded
18495     * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with
18496       -fshort-wchar
18497
18498    ARM-specific
18499
18500     * [113]9090: arm ICE with >= -O2; regression from gcc-2.95
18501
18502    x86-specific (Intel/AMD)
18503
18504     * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction)
18505     * [115]8599: loop unroll bug with -march=k6-3
18506     * [116]9506: ABI breakage in structure return (affects BSD and
18507       Cygwin, but not GNU/Linux)
18508
18509    FreeBSD 5.0 specific
18510
18511     * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0
18512
18513    RTEMS-specific
18514
18515     * [118]9292: hppa1.1-rtems configurery problems
18516     * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug
18517     * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue
18518     * [121]9296: gthr-rtems regression
18519     * [122]9316: powerpc-rtems: extending multilibs
18520
18521    HP-PA specific
18522
18523     * [123]9493: ICE with -O2 when building a simple function
18524
18525    Documentation
18526
18527     * [124]7341: hyperlink to gcov in GCC documentation doesn't work
18528     * [125]8947: Please add a warning about "-malign-double" in docs
18529     * [126]7448, [127]8882: typo cleanups
18530     __________________________________________________________________
18531
18532GCC 3.2.1
18533
18534   3.2.1 adds a new warning, -Wabi. This option warns when GNU C++
18535   generates code that is known not to be binary-compatible with the
18536   vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included
18537   in the distribution, for details.
18538
18539   This release also removes an old GCC extension, "naming types", and the
18540   documentation now directs users to use a different GCC extension,
18541   __typeof__, instead. The feature had evidently been broken for a while.
18542
18543   Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and
18544   the new warning there are no new features that were not present in GCC
18545   3.2.
18546
18547   In addition, the previous fix for [128]PR 7445 (poor performance of
18548   std::locale::classic() in multi-threaded applications) was reverted
18549   ("unfixed"), because the "fix" was not thread-safe.
18550
18551  Bug Fixes
18552
18553   This section lists the problem reports (PRs) from GCC's bug tracking
18554   system that are known to be fixed in the 3.2.1 release. This list might
18555   not be complete (that is, it is possible that some PRs that have been
18556   fixed are not listed here). As you can see, the number of bug fixes is
18557   quite large, so it is strongly recommended that users of earlier GCC
18558   3.x releases upgrade to GCC 3.2.1.
18559
18560    Internal Compiler Errors (multi-platform)
18561
18562     * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c
18563     * [130]5661: (c++) ICE instantiating template on array of unknown
18564       size (bad code)
18565     * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on
18566       64-bit platforms
18567     * [132]6994: (c++) ICE in find_function_data
18568     * [133]7150: preprocessor: GCC -dM -E gives an ICE
18569     * [134]7160: ICE when optimizing branches without a return value
18570     * [135]7228: (c++) ICE when using member template and template
18571       function
18572     * [136]7266: (c++) ICE with -pedantic on missing typename
18573     * [137]7353: ICE from use of "Naming Types" extension, see above
18574     * [138]7411: ICE in instantiate_virtual_regs_1, in function.c
18575     * [139]7478: (c++) ICE on static_cast inside template
18576     * [140]7526: preprocessor core dump when _Pragma implies #pragma
18577       dependency
18578     * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803
18579       is a duplicate)
18580     * [143]7754: (c++) ICE on union with template parameter
18581     * [144]7788: (c++) redeclaring a definition as an incomplete class
18582       causes ICE
18583     * [145]8031: (c++) ICE in comptypes, in cp/typeck.c
18584     * [146]8055: preprocessor dies with SIG11 when building FreeBSD
18585       kernel
18586     * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related
18587       variables
18588     * [148]8134: (c++) ICE in force_store_init_value on legal code
18589     * [149]8149: (c++) ICE on incomplete type
18590     * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array
18591       initialization
18592
18593    C++ (compiler and library) bugs
18594
18595     * [151]5607: No pointer adjustment in covariant return types
18596     * [152]6579: Infinite loop with statement expressions in member
18597       initialization
18598     * [153]6803: Default copy constructor bug in GCC 3.1
18599     * [154]7176: g++ confused by friend and static member with same name
18600     * [155]7188: Segfault with template class and recursive (incorrect)
18601       initializer list
18602     * [156]7306: Regression: GCC 3.x fails to compile code with virtual
18603       inheritance if a method has a variable number of arguments
18604     * [157]7461: ctype<char>::classic_table() returns offset array on
18605       Cygwin
18606     * [158]7524: f(const float arg[3]) fails
18607     * [159]7584: Erroneous ambiguous base error on using declaration
18608     * [160]7676: Member template overloading problem
18609     * [161]7679: infinite loop when a right parenthesis is missing
18610     * [162]7811: default locale not taken from environment
18611     * [163]7961: compare( char *) implemented incorrectly in
18612       basic_string<>
18613     * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if
18614       streambuf::underflow() leaves gptr() NULL (dups: [165]8127,
18615       [166]6745)
18616     * [167]8096: deque::at() throws std::range_error instead of
18617       std::out_of_range
18618     * [168]8127: cout << cin.rdbuf() infinite loop
18619     * [169]8218: Excessively large memory consumed for classes with large
18620       array members
18621     * [170]8287: GCC 3.2: Destructor called for non-constructed local
18622       object
18623     * [171]8347: empty vector range used in string construction causes
18624       core dump
18625     * [172]8348: fail() flag is set in istringstream when eof() flag is
18626       set
18627     * [173]8391: regression: infinite loop in cp/decl2.c(finish_file)
18628
18629    C and optimizer bugs
18630
18631     * [174]6627: -fno-align-functions doesn't seem to disable function
18632       alignment
18633     * [175]6631: life_analysis misoptimizes code to initialize fields of
18634       a structure
18635     * [176]7102: unsigned char division results in floating exception
18636     * [177]7120: Run once loop should *always* be unrolled
18637       (pessimization)
18638     * [178]7209: Bug involving array referencing and ?: operator
18639     * [179]7515: invalid inlining of global function with -O3
18640     * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test
18641     * [181]8467: bug in sibling call optimization
18642
18643    Preprocessor bugs
18644
18645     * [182]4890: incorrect line markers from the traditional preprocessor
18646     * [183]7357: -M option omits system headers files (making it the same
18647       as -MM)
18648     * [184]7358: Changes to Sun's make Dependencies
18649     * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as
18650       C headers
18651     * [186]7862: Interrupting GCC -MD removes .d file but not .o
18652     * [187]8190: Failed compilation deletes -MD dependency file
18653     * [188]8524: _Pragma within macro is improperly expanded
18654
18655    x86 specific (Intel/AMD)
18656
18657     * [189]5351: (i686-only) function pass-by-value structure copy
18658       corrupts stack ([190]7591 is a duplicate)
18659     * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with
18660       -march=pentium3/pentium2/athlon (these are all the same underlying
18661       bug, in MMX register use)
18662     * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same
18663       as above?)
18664     * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken
18665     * [199]6981: wrong code in 64-bit manipulation on x86
18666     * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__
18667       macro
18668     * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE
18669       intrinsics are broken
18670     * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with
18671       -march=pentium4
18672     * [203]7693: Typo in i386 mmintrin.h header
18673     * [204]7723: ICE - Pentium3 sse - GCC 3.2
18674     * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse
18675     * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3
18676
18677    PowerPC specific
18678
18679     * [207]5967: GCC bug when profiling nested functions on powerpc
18680     * [208]6984: wrong code generated with -O2, -O3, -Os for do-while
18681       loop on PowerPC
18682     * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5
18683     * [210]7130: miscompiled code for GCC-3.1 on
18684       powerpc-unknown-linux-gnu with -funroll-all-loops
18685     * [211]7133: PowerPC ICE: unrecognizable insn
18686     * [212]7380: ICE in extract_insn, at recog.c:2148
18687     * [213]8252: ICE on Altivec code with optimization turned on
18688     * [214]8451: Altivec ICE in GCC 3.2
18689
18690    HP/PA specific
18691
18692     * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa
18693
18694    SPARC specific
18695
18696     * [216]6668: when using --disable-multilib, libgcc_s.so is installed
18697       in the wrong place on sparc-solaris
18698     * [217]7151: ICE when compiling for UltraSPARC
18699     * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long
18700       double and -O1
18701     * [219]7842: [REGRESSION] SPARC code gen bug
18702
18703    ARM specific
18704
18705     * [220]7856: [arm] invalid offset in constant pool reference
18706     * [221]7967: optimization produces wrong code (ARM)
18707
18708    Alpha specific
18709
18710     * [222]7374: __builtin_fabsl broken on alpha
18711
18712    IBM s390 specific
18713
18714     * [223]7370: ICE in fixup_var_refs_1 on s390x
18715     * [224]7409: loop optimization bug on s390x-linux-gnu
18716     * [225]8232: s390x: ICE when using bcmp with int length argument
18717
18718    SCO specific
18719
18720     * [226]7623: SCO OpenServer build fails with machmode.def: undefined
18721       symbol: BITS_PER_UNIT
18722
18723    m68k/Coldfire specific
18724
18725     * [227]8314: crtbegin, crtend need to be multilib'ed for this
18726       platform
18727
18728    Documentation
18729
18730     * [228]761: Document some undocumented options
18731     * [229]5610: Fix documentation about invoking SSE instructions
18732       (-mfpmath=sse)
18733     * [230]7484: List -Wmissing-declarations as C-only option
18734     * [231]7531: -mcmodel not documented for x86-64
18735     * [232]8120: Update documentation of bad use of ##
18736     __________________________________________________________________
18737
18738GCC 3.2
18739
18740   3.2 is a small bug fix release, but there is a change to the
18741   application binary interface (ABI), hence the change to the second part
18742   of the version number.
18743
18744   The main purpose of the 3.2 release is to correct a couple of problems
18745   in the C++ ABI, with the intention of providing a stable interface
18746   going forward.  Accordingly, 3.2 is only a small change to 3.1.1.
18747
18748  Bug Fixes
18749
18750    C++
18751
18752     * [233]7320: g++ 3.2 relocation problem
18753     * [234]7470: vtable: virtual function pointers not in declaration
18754       order
18755
18756    libstdc++
18757
18758     * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t
18759     * [236]6503, [237]6642, [238]7186: Problems with comparing or
18760       subtracting various types of const and non-const iterators
18761     * [239]7216: ambiguity with basic_iostream::traits_type
18762     * [240]7220: problem with basic_istream::ignore(0,delimiter)
18763     * [241]7222: locale::operator==() doesn't work on std::locale("")
18764     * [242]7286: placement operator delete issue
18765     * [243]7442: cxxabi.h does not match the C++ ABI
18766     * [244]7445: poor performance of std::locale::classic() in
18767       multi-threaded applications
18768
18769    x86-64 specific
18770
18771     * [245]7291: off-by-one in generated inline bzero code for x86-64
18772
18773
18774    For questions related to the use of GCC, please consult these web
18775    pages and the [246]GCC manuals. If that fails, the
18776    [247]gcc-help@gcc.gnu.org mailing list might help. Comments on these
18777    web pages and the development of GCC are welcome on our developer
18778    list at [248]gcc@gcc.gnu.org. All of [249]our lists have public
18779    archives.
18780
18781   Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and
18782   distribution of this entire article is permitted in any medium,
18783   provided this notice is preserved.
18784
18785   These pages are [251]maintained by the GCC team. Last modified
18786   2021-07-28[252].
18787
18788References
18789
18790   1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
18791   2. http://gcc.gnu.org/gcc-3.1/changes.html
18792   3. http://gcc.gnu.org/gcc-3.2/c++-abi.html
18793   4. https://gcc.gnu.org/PR3782
18794   5. https://gcc.gnu.org/PR6440
18795   6. https://gcc.gnu.org/PR7050
18796   7. https://gcc.gnu.org/PR7741
18797   8. https://gcc.gnu.org/PR7982
18798   9. https://gcc.gnu.org/PR8068
18799  10. https://gcc.gnu.org/PR8178
18800  11. https://gcc.gnu.org/PR8396
18801  12. https://gcc.gnu.org/PR8674
18802  13. https://gcc.gnu.org/PR9768
18803  14. https://gcc.gnu.org/PR9798
18804  15. https://gcc.gnu.org/PR9799
18805  16. https://gcc.gnu.org/PR9928
18806  17. https://gcc.gnu.org/PR10114
18807  18. https://gcc.gnu.org/PR10352
18808  19. https://gcc.gnu.org/PR10336
18809  20. https://gcc.gnu.org/PR8224
18810  21. https://gcc.gnu.org/PR8613
18811  22. https://gcc.gnu.org/PR8828
18812  23. https://gcc.gnu.org/PR9226
18813  24. https://gcc.gnu.org/PR9853
18814  25. https://gcc.gnu.org/PR9797
18815  26. https://gcc.gnu.org/PR9967
18816  27. https://gcc.gnu.org/PR10116
18817  28. https://gcc.gnu.org/PR10171
18818  29. https://gcc.gnu.org/PR10175
18819  30. https://gcc.gnu.org/PR8316
18820  31. https://gcc.gnu.org/PR9169
18821  32. https://gcc.gnu.org/PR9420
18822  33. https://gcc.gnu.org/PR9459
18823  34. https://gcc.gnu.org/PR9507
18824  35. https://gcc.gnu.org/PR9538
18825  36. https://gcc.gnu.org/PR9602
18826  37. https://gcc.gnu.org/PR9993
18827  38. https://gcc.gnu.org/PR10167
18828  39. https://gcc.gnu.org/PR9652
18829  40. https://gcc.gnu.org/PR10144
18830  41. https://gcc.gnu.org/PR8746
18831  42. https://gcc.gnu.org/PR9888
18832  43. https://gcc.gnu.org/PR9638
18833  44. https://gcc.gnu.org/PR9954
18834  45. https://gcc.gnu.org/PR7784
18835  46. https://gcc.gnu.org/PR7796
18836  47. https://gcc.gnu.org/PR8281
18837  48. https://gcc.gnu.org/PR8366
18838  49. https://gcc.gnu.org/PR8726
18839  50. https://gcc.gnu.org/PR9414
18840  51. https://gcc.gnu.org/PR10067
18841  52. https://gcc.gnu.org/PR7248
18842  53. https://gcc.gnu.org/PR8343
18843  54. https://gcc.gnu.org/PR9732
18844  55. https://gcc.gnu.org/PR10073
18845  56. https://gcc.gnu.org/PR7702
18846  57. https://gcc.gnu.org/PR9671
18847  58. https://gcc.gnu.org/PR8694
18848  59. https://gcc.gnu.org/PR9953
18849  60. https://gcc.gnu.org/PR10271
18850  61. https://gcc.gnu.org/PR6362
18851  62. https://gcc.gnu.org/PR10377
18852  63. https://gcc.gnu.org/PR6955
18853  64. https://gcc.gnu.org/PR5919
18854  65. https://gcc.gnu.org/PR7129
18855  66. https://gcc.gnu.org/PR7507
18856  67. https://gcc.gnu.org/PR7622
18857  68. https://gcc.gnu.org/PR7681
18858  69. https://gcc.gnu.org/PR9528
18859  70. https://gcc.gnu.org/PR8031
18860  71. https://gcc.gnu.org/PR8275
18861  72. https://gcc.gnu.org/PR8332
18862  73. https://gcc.gnu.org/PR8372
18863  74. https://gcc.gnu.org/PR8439
18864  75. https://gcc.gnu.org/PR8442
18865  76. https://gcc.gnu.org/PR8518
18866  77. https://gcc.gnu.org/PR8615
18867  78. https://gcc.gnu.org/PR8663
18868  79. https://gcc.gnu.org/PR8799
18869  80. https://gcc.gnu.org/PR9328
18870  81. https://gcc.gnu.org/PR9465
18871  82. https://gcc.gnu.org/PR47
18872  83. https://gcc.gnu.org/PR6745
18873  84. https://gcc.gnu.org/PR8214
18874  85. https://gcc.gnu.org/PR8493
18875  86. https://gcc.gnu.org/PR8332
18876  87. https://gcc.gnu.org/PR8503
18877  88. https://gcc.gnu.org/PR8727
18878  89. https://gcc.gnu.org/PR7445
18879  90. https://gcc.gnu.org/PR8230
18880  91. https://gcc.gnu.org/PR8399
18881  92. https://gcc.gnu.org/PR8662
18882  93. https://gcc.gnu.org/PR8707
18883  94. https://gcc.gnu.org/PR8708
18884  95. https://gcc.gnu.org/PR8790
18885  96. https://gcc.gnu.org/PR8887
18886  97. https://gcc.gnu.org/PR9076
18887  98. https://gcc.gnu.org/PR9151
18888  99. https://gcc.gnu.org/PR9168
18889 100. https://gcc.gnu.org/PR9269
18890 101. https://gcc.gnu.org/PR9322
18891 102. https://gcc.gnu.org/PR9433
18892 103. https://gcc.gnu.org/PR8032
18893 104. https://gcc.gnu.org/PR8639
18894 105. https://gcc.gnu.org/PR8794
18895 106. https://gcc.gnu.org/PR8832
18896 107. https://gcc.gnu.org/PR8988
18897 108. https://gcc.gnu.org/PR9492
18898 109. https://gcc.gnu.org/PR9267
18899 110. https://gcc.gnu.org/PR8344
18900 111. https://gcc.gnu.org/PR8524
18901 112. https://gcc.gnu.org/PR8880
18902 113. https://gcc.gnu.org/PR9090
18903 114. https://gcc.gnu.org/PR8588
18904 115. https://gcc.gnu.org/PR8599
18905 116. https://gcc.gnu.org/PR9506
18906 117. https://gcc.gnu.org/PR9484
18907 118. https://gcc.gnu.org/PR9292
18908 119. https://gcc.gnu.org/PR9293
18909 120. https://gcc.gnu.org/PR9295
18910 121. https://gcc.gnu.org/PR9296
18911 122. https://gcc.gnu.org/PR9316
18912 123. https://gcc.gnu.org/PR9493
18913 124. https://gcc.gnu.org/PR7341
18914 125. https://gcc.gnu.org/PR8947
18915 126. https://gcc.gnu.org/PR7448
18916 127. https://gcc.gnu.org/PR8882
18917 128. https://gcc.gnu.org/PR7445
18918 129. https://gcc.gnu.org/PR2521
18919 130. https://gcc.gnu.org/PR5661
18920 131. https://gcc.gnu.org/PR6419
18921 132. https://gcc.gnu.org/PR6994
18922 133. https://gcc.gnu.org/PR7150
18923 134. https://gcc.gnu.org/PR7160
18924 135. https://gcc.gnu.org/PR7228
18925 136. https://gcc.gnu.org/PR7266
18926 137. https://gcc.gnu.org/PR7353
18927 138. https://gcc.gnu.org/PR7411
18928 139. https://gcc.gnu.org/PR7478
18929 140. https://gcc.gnu.org/PR7526
18930 141. https://gcc.gnu.org/PR7721
18931 142. https://gcc.gnu.org/PR7803
18932 143. https://gcc.gnu.org/PR7754
18933 144. https://gcc.gnu.org/PR7788
18934 145. https://gcc.gnu.org/PR8031
18935 146. https://gcc.gnu.org/PR8055
18936 147. https://gcc.gnu.org/PR8067
18937 148. https://gcc.gnu.org/PR8134
18938 149. https://gcc.gnu.org/PR8149
18939 150. https://gcc.gnu.org/PR8160
18940 151. https://gcc.gnu.org/PR5607
18941 152. https://gcc.gnu.org/PR6579
18942 153. https://gcc.gnu.org/PR6803
18943 154. https://gcc.gnu.org/PR7176
18944 155. https://gcc.gnu.org/PR7188
18945 156. https://gcc.gnu.org/PR7306
18946 157. https://gcc.gnu.org/PR7461
18947 158. https://gcc.gnu.org/PR7524
18948 159. https://gcc.gnu.org/PR7584
18949 160. https://gcc.gnu.org/PR7676
18950 161. https://gcc.gnu.org/PR7679
18951 162. https://gcc.gnu.org/PR7811
18952 163. https://gcc.gnu.org/PR7961
18953 164. https://gcc.gnu.org/PR8071
18954 165. https://gcc.gnu.org/PR8127
18955 166. https://gcc.gnu.org/PR6745
18956 167. https://gcc.gnu.org/PR8096
18957 168. https://gcc.gnu.org/PR8127
18958 169. https://gcc.gnu.org/PR8218
18959 170. https://gcc.gnu.org/PR8287
18960 171. https://gcc.gnu.org/PR8347
18961 172. https://gcc.gnu.org/PR8348
18962 173. https://gcc.gnu.org/PR8391
18963 174. https://gcc.gnu.org/PR6627
18964 175. https://gcc.gnu.org/PR6631
18965 176. https://gcc.gnu.org/PR7102
18966 177. https://gcc.gnu.org/PR7120
18967 178. https://gcc.gnu.org/PR7209
18968 179. https://gcc.gnu.org/PR7515
18969 180. https://gcc.gnu.org/PR7814
18970 181. https://gcc.gnu.org/PR8467
18971 182. https://gcc.gnu.org/PR4890
18972 183. https://gcc.gnu.org/PR7357
18973 184. https://gcc.gnu.org/PR7358
18974 185. https://gcc.gnu.org/PR7602
18975 186. https://gcc.gnu.org/PR7862
18976 187. https://gcc.gnu.org/PR8190
18977 188. https://gcc.gnu.org/PR8524
18978 189. https://gcc.gnu.org/PR5351
18979 190. https://gcc.gnu.org/PR7591
18980 191. https://gcc.gnu.org/PR6845
18981 192. https://gcc.gnu.org/PR7034
18982 193. https://gcc.gnu.org/PR7124
18983 194. https://gcc.gnu.org/PR7174
18984 195. https://gcc.gnu.org/PR7134
18985 196. https://gcc.gnu.org/PR7375
18986 197. https://gcc.gnu.org/PR7390
18987 198. https://gcc.gnu.org/PR6890
18988 199. https://gcc.gnu.org/PR6981
18989 200. https://gcc.gnu.org/PR7242
18990 201. https://gcc.gnu.org/PR7396
18991 202. https://gcc.gnu.org/PR7630
18992 203. https://gcc.gnu.org/PR7693
18993 204. https://gcc.gnu.org/PR7723
18994 205. https://gcc.gnu.org/PR7951
18995 206. https://gcc.gnu.org/PR8146
18996 207. https://gcc.gnu.org/PR5967
18997 208. https://gcc.gnu.org/PR6984
18998 209. https://gcc.gnu.org/PR7114
18999 210. https://gcc.gnu.org/PR7130
19000 211. https://gcc.gnu.org/PR7133
19001 212. https://gcc.gnu.org/PR7380
19002 213. https://gcc.gnu.org/PR8252
19003 214. https://gcc.gnu.org/PR8451
19004 215. https://gcc.gnu.org/PR7250
19005 216. https://gcc.gnu.org/PR6668
19006 217. https://gcc.gnu.org/PR7151
19007 218. https://gcc.gnu.org/PR7335
19008 219. https://gcc.gnu.org/PR7842
19009 220. https://gcc.gnu.org/PR7856
19010 221. https://gcc.gnu.org/PR7967
19011 222. https://gcc.gnu.org/PR7374
19012 223. https://gcc.gnu.org/PR7370
19013 224. https://gcc.gnu.org/PR7409
19014 225. https://gcc.gnu.org/PR8232
19015 226. https://gcc.gnu.org/PR7623
19016 227. https://gcc.gnu.org/PR8314
19017 228. https://gcc.gnu.org/PR761
19018 229. https://gcc.gnu.org/PR5610
19019 230. https://gcc.gnu.org/PR7484
19020 231. https://gcc.gnu.org/PR7531
19021 232. https://gcc.gnu.org/PR8120
19022 233. https://gcc.gnu.org/PR7320
19023 234. https://gcc.gnu.org/PR7470
19024 235. https://gcc.gnu.org/PR6410
19025 236. https://gcc.gnu.org/PR6503
19026 237. https://gcc.gnu.org/PR6642
19027 238. https://gcc.gnu.org/PR7186
19028 239. https://gcc.gnu.org/PR7216
19029 240. https://gcc.gnu.org/PR7220
19030 241. https://gcc.gnu.org/PR7222
19031 242. https://gcc.gnu.org/PR7286
19032 243. https://gcc.gnu.org/PR7442
19033 244. https://gcc.gnu.org/PR7445
19034 245. https://gcc.gnu.org/PR7291
19035 246. https://gcc.gnu.org/onlinedocs/
19036 247. mailto:gcc-help@gcc.gnu.org
19037 248. mailto:gcc@gcc.gnu.org
19038 249. https://gcc.gnu.org/lists.html
19039 250. https://www.fsf.org/
19040 251. https://gcc.gnu.org/about.html
19041 252. http://validator.w3.org/check/referer
19042======================================================================
19043http://gcc.gnu.org/gcc-3.1/index.html
19044                                    GCC 3.1
19045
19046   (This release series is no longer supported.)
19047
19048   July 27, 2002
19049
19050   The [1]GNU project and the GCC developers are pleased to announce the
19051   release of GCC 3.1.1.
19052
19053   The links below still apply to GCC 3.1.1.
19054
19055   May 15, 2002
19056
19057   The [2]GNU project and the GCC developers are pleased to announce the
19058   release of GCC 3.1.
19059
19060   GCC used to stand for the GNU C Compiler, but since the compiler
19061   supports several other languages aside from C, it now stands for the
19062   GNU Compiler Collection.
19063
19064   A list of [3]successful builds is updated as new information becomes
19065   available.
19066
19067   The GCC developers would like to thank the numerous people that have
19068   contributed [4]new features, improvements, bug fixes, and other changes
19069   as well as test results to GCC. This [5]amazing group of volunteers is
19070   what makes GCC successful.
19071
19072   For additional information about GCC please refer to the [6]GCC project
19073   web site or contact the [7]GCC development mailing list.
19074
19075   To obtain GCC please use [8]our mirror sites, or our CVS server.
19076     __________________________________________________________________
19077
19078
19079    For questions related to the use of GCC, please consult these web
19080    pages and the [9]GCC manuals. If that fails, the
19081    [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19082    web pages and the development of GCC are welcome on our developer
19083    list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
19084    archives.
19085
19086   Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
19087   distribution of this entire article is permitted in any medium,
19088   provided this notice is preserved.
19089
19090   These pages are [14]maintained by the GCC team. Last modified
19091   2021-07-28[15].
19092
19093References
19094
19095   1. http://www.gnu.org/
19096   2. http://www.gnu.org/
19097   3. http://gcc.gnu.org/gcc-3.1/buildstat.html
19098   4. http://gcc.gnu.org/gcc-3.1/changes.html
19099   5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
19100   6. http://gcc.gnu.org/index.html
19101   7. mailto:gcc@gcc.gnu.org
19102   8. http://gcc.gnu.org/mirrors.html
19103   9. https://gcc.gnu.org/onlinedocs/
19104  10. mailto:gcc-help@gcc.gnu.org
19105  11. mailto:gcc@gcc.gnu.org
19106  12. https://gcc.gnu.org/lists.html
19107  13. https://www.fsf.org/
19108  14. https://gcc.gnu.org/about.html
19109  15. http://validator.w3.org/check/referer
19110======================================================================
19111http://gcc.gnu.org/gcc-3.1/changes.html
19112                             GCC 3.1 Release Series
19113                        Changes, New Features, and Fixes
19114
19115Additional changes in GCC 3.1.1
19116
19117     * A bug related to how structures and unions are returned has been
19118       fixed for powerpc-*-netbsd*.
19119     * An important bug in the implementation of -fprefetch-loop-arrays
19120       has been fixed. Previously the optimization prefetched random
19121       blocks of memory for most targets except for i386.
19122     * The Java compiler now compiles Java programs much faster and also
19123       works with parallel make.
19124     * Nested functions have been fixed for mips*-*-netbsd*.
19125     * Some missing floating point support routines have beed added for
19126       mips*-*-netbsd*.
19127     * This [1]message gives additional information about the bugs fixed
19128       in this release.
19129
19130Caveats
19131
19132     * The -traditional C compiler option has been deprecated and will be
19133       removed in GCC 3.3. (It remains possible to preprocess non-C code
19134       with the traditional preprocessor.)
19135     * The default debugging format for most ELF platforms (including
19136       GNU/Linux and FreeBSD; notable exception is Solaris) has changed
19137       from stabs to DWARF2. This requires GDB 5.1.1 or later.
19138
19139General Optimizer Improvements
19140
19141     * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat,
19142       and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure
19143       for profile driven optimizations.
19144       Options -fprofile-arcs and -fbranch-probabilities can now be used
19145       to improve speed of the generated code by profiling the actual
19146       program behaviour on typical runs. In the absence of profile info
19147       the compiler attempts to guess the profile statically.
19148     * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to
19149       monitor performance of the generated code.
19150       According to the SPECInt2000 results on an AMD Athlon CPU, the code
19151       generated by GCC 3.1 is 6% faster on the average (8.2% faster with
19152       profile feedback) compared to GCC 3.0. The code produced by GCC 3.0
19153       is about 2.1% faster compared to 2.95.3. Tests were done using the
19154       -O2 -march=athlon command-line options.
19155     * Alexandre Oliva, of Red Hat, has generalized the tree inlining
19156       infrastructure developed by CodeSourcery, LLC for the C++ front
19157       end, so that it is now used in the C front end too. Inlining
19158       functions as trees exposes them earlier to the compiler, giving it
19159       more opportunities for optimization.
19160     * Support for data prefetching instructions has been added to the GCC
19161       back end and several targets. A new __builtin_prefetch intrinsic is
19162       available to explicitly insert prefetch instructions and
19163       experimental support for loop array prefetching has been added (see
19164       -fprefetch-loop-array documentation).
19165     * Support for emitting debugging information for macros has been
19166       added for DWARF2. It is activated using -g3.
19167
19168New Languages and Language specific improvements
19169
19170  C/C++
19171
19172     * A few more [4]ISO C99 features.
19173     * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0.
19174     * The preprocessor's symbol table has been merged with the symbol
19175       table of the C, C++ and Objective-C front ends.
19176     * The preprocessor consumes less memory than the preprocessor in GCC
19177       3.0, often significantly so. On normal input files, it typically
19178       consumes less memory than pre-3.0 cccp-based GCC, too.
19179
19180  C++
19181
19182     * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std
19183       was a workaround to allow std compliant code to work with the
19184       non-std compliant libstdc++-v2. libstdc++-v3 is std compliant.
19185     * The C++ ABI has been fixed so that void (A::*)() const is mangled
19186       as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only
19187       affects pointer to cv-qualified member function types.
19188     * The C++ ABI has been changed to correctly handle this code:
19189    struct A {
19190      void operator delete[] (void *, size_t);
19191    };
19192
19193    struct B : public A {
19194    };
19195
19196    new B[10];
19197
19198       The amount of storage allocated for the array will be greater than
19199       it was in 3.0, in order to store the number of elements in the
19200       array, so that the correct size can be passed to operator delete[]
19201       when the array is deleted. Previously, the value passed to operator
19202       delete[] was unpredictable.
19203       This change will only affect code that declares a two-argument
19204       operator delete[] with a second parameter of type size_t in a base
19205       class, and does not override that definition in a derived class.
19206     * The C++ ABI has been changed so that:
19207    struct A {
19208      void operator delete[] (void *, size_t);
19209      void operator delete[] (void *);
19210    };
19211
19212       does not cause unnecessary storage to be allocated when an array of
19213       A objects is allocated.
19214       This change will only affect code that declares both of these forms
19215       of operator delete[], and declared the two-argument form before the
19216       one-argument form.
19217     * The C++ ABI has been changed so that when a parameter is passed by
19218       value, any cleanup for that parameter is performed in the caller,
19219       as specified by the ia64 C++ ABI, rather than the called function
19220       as before. As a result, classes with a non-trivial destructor but a
19221       trivial copy constructor will be passed and returned by invisible
19222       reference, rather than by bitwise copy as before.
19223     * G++ now supports the "named return value optimization": for code
19224       like
19225    A f () {
19226      A a;
19227      ...
19228      return a;
19229    }
19230
19231       G++ will allocate a in the return value slot, so that the return
19232       becomes a no-op. For this to work, all return statements in the
19233       function must return the same variable.
19234     * Improvements to the C++ library are listed in [5]the libstdc++-v3
19235       FAQ.
19236
19237  Objective-C
19238
19239     * Annoying linker warnings (due to incorrect code being generated)
19240       have been fixed.
19241     * If a class method cannot be found, the compiler no longer issues a
19242       warning if a corresponding instance method exists in the root
19243       class.
19244     * Forward @protocol declarations have been fixed.
19245     * Loading of categories has been fixed in certain situations (GNU run
19246       time only).
19247     * The class lookup in the run-time library has been rewritten so that
19248       class method dispatch is more than twice as fast as it used to be
19249       (GNU run time only).
19250
19251  Java
19252
19253     * libgcj now includes RMI, java.lang.ref.*, javax.naming, and
19254       javax.transaction.
19255     * Property files and other system resources can be compiled into
19256       executables which use libgcj using the new gcj --resource feature.
19257     * libgcj has been ported to more platforms. In particular there is
19258       now a mostly-functional mingw32 (Windows) target port.
19259     * JNI and CNI invocation interfaces were implemented, so gcj-compiled
19260       Java code can now be called from a C/C++ application.
19261     * gcj can now use builtin functions for certain known methods, for
19262       instance Math.cos.
19263     * gcj can now automatically remove redundant array-store checks in
19264       some common cases.
19265     * The --no-store-checks optimization option was added. This can be
19266       used to omit runtime store checks for code which is known not to
19267       throw ArrayStoreException
19268     * The following third party interface standards were added to libgcj:
19269       org.w3c.dom and org.xml.sax.
19270     * java.security has been merged with GNU Classpath. The new package
19271       is now JDK 1.2 compliant, and much more complete.
19272     * A bytecode verifier was added to the libgcj interpreter.
19273     * java.lang.Character was rewritten to comply with the Unicode 3.0
19274       standard, and improve performance.
19275     * Partial support for many more locales was added to libgcj.
19276     * Socket timeouts have been implemented.
19277     * libgcj has been merged into a single shared library. There are no
19278       longer separate shared libraries for the garbage collector and
19279       zlib.
19280     * Several performance improvements were made to gcj and libgcj:
19281          + Hash synchronization (thin locks)
19282          + A special allocation path for finalizer-free objects
19283          + Thread-local allocation
19284          + Parallel GC, and other GC tweaks
19285
19286  Fortran
19287
19288   Fortran improvements are listed in [6]the Fortran documentation.
19289
19290  Ada
19291
19292   [7]AdaCore, has contributed its GNAT Ada 95 front end and associated
19293   tools. The GNAT compiler fully implements the Ada language as defined
19294   by the ISO/IEC 8652 standard.
19295
19296   Please note that the integration of the Ada front end is still work in
19297   progress.
19298
19299New Targets and Target Specific Improvements
19300
19301     * Hans-Peter Nilsson has contributed a port to MMIX, the CPU
19302       architecture used in new editions of Donald E. Knuth's The Art of
19303       Computer Programming.
19304     * Axis Communications has contributed its port to the CRIS CPU
19305       architecture, used in the ETRAX system-on-a-chip series.
19306     * Alexandre Oliva, of Red Hat, has contributed a port to the SuperH
19307       SH5 64-bit RISC microprocessor architecture, extending the existing
19308       SH port.
19309     * UltraSPARC is fully supported in 64-bit mode. The option -m64
19310       enables it.
19311     * For compatibility with the Sun compiler #pragma redefine_extname
19312       has been implemented on Solaris.
19313     * The x86 back end has had some noticeable work done to it.
19314          + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas
19315            Jaeger have contributed a port to the AMD x86-64 architecture.
19316            For more information on x86-64 see http://www.x86-64.org.
19317          + The compiler now supports MMX, 3DNow!, SSE, and SSE2
19318            instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will
19319            enable the respective instruction sets. Intel C++ compatible
19320            MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics
19321            will be added in next major release.
19322          + Following those improvements, targets for Pentium MMX, K6-2,
19323            K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were
19324            added. Refer to the documentation on -march= and -mcpu=
19325            options for details.
19326          + For those targets that support it, -mfpmath=sse will cause the
19327            compiler to generate SSE/SSE2 instructions for floating point
19328            math instead of x87 instructions. Usually, this will lead to
19329            quicker code — especially on the Pentium 4. Note that only
19330            scalar floating point instructions are used and GCC does not
19331            exploit SIMD features yet.
19332          + Prefetch support has been added to the Pentium III, Pentium 4,
19333            K6-2, K6-3, and Athlon series.
19334          + Code generated for floating point to integer conversions has
19335            been improved leading to better performance of many 3D
19336            applications.
19337     * The PowerPC back end has added 64-bit PowerPC GNU/Linux support.
19338     * C++ support for AIX has been improved.
19339     * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the
19340       PowerPC port supporting the AltiVec programming model (SIMD). The
19341       support, though presently useful, is experimental and is expected
19342       to stabilize for 3.2. The support is written to conform to
19343       Motorola's AltiVec specs. See -maltivec.
19344
19345Obsolete Systems
19346
19347   Support for a number of older systems has been declared obsolete in GCC
19348   3.1. Unless there is activity to revive them, the next release of GCC
19349   will have their sources permanently removed.
19350
19351   All configurations of the following processor architectures have been
19352   declared obsolete:
19353     * MIL-STD-1750A, 1750a-*-*
19354     * AMD A29k, a29k-*-*
19355     * Convex, c*-convex-*
19356     * Clipper, clipper-*-*
19357     * Elxsi, elxsi-*-*
19358     * Intel i860, i860-*-*
19359     * Sun picoJava, pj-*-* and pjl-*-*
19360     * Western Electric 32000, we32k-*-*
19361
19362   Most configurations of the following processor architectures have been
19363   declared obsolete, but we are preserving a few systems which may have
19364   active developers. It is unlikely that the remaining systems will
19365   survive much longer unless we see definite signs of port activity.
19366     * Motorola 88000 except
19367          + Generic a.out, m88k-*-aout*
19368          + Generic SVR4, m88k-*-sysv4
19369          + OpenBSD, m88k-*-openbsd*
19370     * NS32k except
19371          + NetBSD, ns32k-*-netbsd*
19372          + OpenBSD, ns32k-*-openbsd*.
19373     * ROMP except
19374          + OpenBSD, romp-*-openbsd*.
19375
19376   Finally, only some configurations of these processor architectures are
19377   being obsoleted.
19378     * Alpha:
19379          + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka
19380            alpha*-*-osf[45], are still supported.)
19381     * ARM:
19382          + RISCiX, arm-*-riscix*.
19383     * i386:
19384          + 386BSD, i?86-*-bsd*
19385          + Chorus, i?86-*-chorusos*
19386          + DG/UX, i?86-*-dgux*
19387          + FreeBSD 1.x, i?86-*-freebsd1.*
19388          + IBM AIX, i?86-*-aix*
19389          + ISC UNIX, i?86-*-isc*
19390          + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld*
19391          + NEXTstep, i?86-next-*
19392          + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose*
19393          + RTEMS/coff, i?86-*-rtemscoff*
19394          + RTEMS/go32, i?86-go32-rtems*
19395          + Sequent/BSD, i?86-sequent-bsd*
19396          + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and
19397            i?86-sequent-sysv3*
19398          + SunOS, i?86-*-sunos*
19399     * Motorola 68000:
19400          + Altos, m68[k0]*-altos-*
19401          + Apollo, m68[k0]*-apollo-*
19402          + Apple A/UX, m68[k0]*-apple-*
19403          + Bull, m68[k0]*-bull-*
19404          + Convergent, m68[k0]*-convergent-*
19405          + Generic SVR3, m68[k0]*-*-sysv3*
19406          + ISI, m68[k0]*-isi-*
19407          + LynxOS, m68[k0]*-*-lynxos*
19408          + NEXT, m68[k0]*-next-*
19409          + RTEMS/coff, m68[k0]*-*-rtemscoff*
19410          + Sony, m68[k0]*-sony-*
19411     * MIPS:
19412          + DEC Ultrix, mips-*-ultrix* and mips-dec-*
19413          + Generic BSD, mips-*-bsd*
19414          + Generic System V, mips-*-sysv*
19415          + IRIX before version 5, mips-sgi-irix[1234]*
19416          + RiscOS, mips-*-riscos*
19417          + Sony, mips-sony-*
19418          + Tandem, mips-tandem-*
19419     * SPARC:
19420          + RTEMS/a.out, sparc-*-rtemsaout*.
19421
19422Documentation improvements
19423
19424     * The old manual ("Using and Porting the GNU Compiler Collection")
19425       has been replaced by a users manual ("Using the GNU Compiler
19426       Collection") and a separate internals reference manual ("GNU
19427       Compiler Collection Internals").
19428     * More complete and much improved documentation about GCC's internal
19429       representation used by the C and C++ front ends.
19430     * Many cleanups and improvements in general.
19431
19432
19433    For questions related to the use of GCC, please consult these web
19434    pages and the [8]GCC manuals. If that fails, the
19435    [9]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19436    web pages and the development of GCC are welcome on our developer
19437    list at [10]gcc@gcc.gnu.org. All of [11]our lists have public
19438    archives.
19439
19440   Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and
19441   distribution of this entire article is permitted in any medium,
19442   provided this notice is preserved.
19443
19444   These pages are [13]maintained by the GCC team. Last modified
19445   2021-07-28[14].
19446
19447References
19448
19449   1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html
19450   2. http://gcc.gnu.org/news/profiledriven.html
19451   3. http://gcc.gnu.org/benchmarks/
19452   4. http://gcc.gnu.org/c99status.html
19453   5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html
19454   6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html
19455   7. https://www.adacore.com/
19456   8. https://gcc.gnu.org/onlinedocs/
19457   9. mailto:gcc-help@gcc.gnu.org
19458  10. mailto:gcc@gcc.gnu.org
19459  11. https://gcc.gnu.org/lists.html
19460  12. https://www.fsf.org/
19461  13. https://gcc.gnu.org/about.html
19462  14. http://validator.w3.org/check/referer
19463======================================================================
19464http://gcc.gnu.org/gcc-3.0/index.html
19465                                   GCC 3.0.4
19466
19467   (This release series is no longer supported.)
19468
19469   February 20, 2002
19470
19471   The [1]GNU project and the GCC developers are pleased to announce the
19472   release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0
19473   series.
19474
19475   GCC used to stand for the GNU C Compiler, but since the compiler
19476   supports several other languages aside from C, it now stands for the
19477   GNU Compiler Collection.
19478
19479   GCC 3.0.x has several new optimizations, new targets, new languages and
19480   many other new features, relative to GCC 2.95.x. See the [2]new
19481   features page for a more complete list.
19482
19483   A list of [3]successful builds is updated as new information becomes
19484   available.
19485
19486   The GCC developers would like to thank the numerous people that have
19487   contributed new features, test results, bug fixes, etc to GCC. This
19488   [4]amazing group of volunteers is what makes GCC successful.
19489
19490   And finally, we can't in good conscience fail to mention some
19491   [5]caveats to using GCC 3.0.x.
19492
19493   For additional information about GCC please refer to the [6]GCC project
19494   web site or contact the [7]GCC development mailing list.
19495
19496   To obtain GCC please use [8]our mirror sites, or our CVS server.
19497     __________________________________________________________________
19498
19499Previous 3.0.x Releases
19500
19501   December 20, 2001: GCC 3.0.3 has been released.
19502   October 25, 2001: GCC 3.0.2 has been released.
19503   August 20, 2001: GCC 3.0.1 has been released.
19504   June 18, 2001: GCC 3.0 has been released.
19505
19506
19507    For questions related to the use of GCC, please consult these web
19508    pages and the [9]GCC manuals. If that fails, the
19509    [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19510    web pages and the development of GCC are welcome on our developer
19511    list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
19512    archives.
19513
19514   Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
19515   distribution of this entire article is permitted in any medium,
19516   provided this notice is preserved.
19517
19518   These pages are [14]maintained by the GCC team. Last modified
19519   2021-07-28[15].
19520
19521References
19522
19523   1. http://www.gnu.org/
19524   2. http://gcc.gnu.org/gcc-3.0/features.html
19525   3. http://gcc.gnu.org/gcc-3.0/buildstat.html
19526   4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
19527   5. http://gcc.gnu.org/gcc-3.0/caveats.html
19528   6. http://gcc.gnu.org/index.html
19529   7. mailto:gcc@gcc.gnu.org
19530   8. http://gcc.gnu.org/mirrors.html
19531   9. https://gcc.gnu.org/onlinedocs/
19532  10. mailto:gcc-help@gcc.gnu.org
19533  11. mailto:gcc@gcc.gnu.org
19534  12. https://gcc.gnu.org/lists.html
19535  13. https://www.fsf.org/
19536  14. https://gcc.gnu.org/about.html
19537  15. http://validator.w3.org/check/referer
19538======================================================================
19539http://gcc.gnu.org/gcc-3.0/features.html
19540                              GCC 3.0 New Features
19541
19542Additional changes in GCC 3.0.4
19543
19544     * GCC 3.0 now supports newer versions of the [1]NetBSD operating
19545       system, which use the ELF object file format, on x86 processors.
19546     * Correct debugging information is generated from functions that have
19547       lines from multiple files (e.g. yacc output).
19548     * A fix for whitespace handling in the -traditional preprocessor,
19549       which can affect Fortran.
19550     * Fixes to the exception handling runtime.
19551     * More fixes for bad code generation in C++.
19552     * A fix for shared library generation under AIX 4.3.
19553     * Documentation updates.
19554     * Port of GCC to Tensilica's Xtensa processor contributed.
19555     * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link).
19556
19557Additional changes in GCC 3.0.3
19558
19559     * A fix to correct an accidental change to the PowerPC ABI.
19560     * Fixes for bad code generation on a variety of architectures.
19561     * Improvements to the debugging information generated for C++
19562       classes.
19563     * Fixes for bad code generation in C++.
19564     * A fix to avoid crashes in the C++ demangler.
19565     * A fix to the C++ standard library to avoid buffer overflows.
19566     * Miscellaneous improvements for a variety of architectures.
19567
19568Additional changes in GCC 3.0.2
19569
19570     * Fixes for bad code generation during loop unrolling.
19571     * Fixes for bad code generation by the sibling call optimization.
19572     * Minor improvements to x86 code generation.
19573     * Implementation of function descriptors in C++ vtables for IA64.
19574     * Numerous minor bug-fixes.
19575
19576Additional changes in GCC 3.0.1
19577
19578     * C++ fixes for incorrect code-generation.
19579     * Improved cross-compiling support for the C++ standard library.
19580     * Fixes for some embedded targets that worked in GCC 2.95.3, but not
19581       in GCC 3.0.
19582     * Fixes for various exception-handling bugs.
19583     * A port to the S/390 architecture.
19584
19585General Optimizer Improvements
19586
19587     * [2]Basic block reordering pass.
19588     * New if-conversion pass with support for conditional (predicated)
19589       execution.
19590     * New tail call and sibling call elimination optimizations.
19591     * New register renaming pass.
19592     * New (experimental) [3]static single assignment (SSA) representation
19593       support.
19594     * New dead-code elimination pass implemented using the SSA
19595       representation.
19596     * [4]Global null pointer test elimination.
19597     * [5]Global code hoisting/unification.
19598     * More builtins and optimizations for stdio.h, string.h and old BSD
19599       functions, as well as for ISO C99 functions.
19600     * New builtin __builtin_expect for giving hints to the branch
19601       predictor.
19602
19603New Languages and Language specific improvements
19604
19605     * The GNU Compiler for the Java(TM) language (GCJ) is now integrated
19606       and supported, including the run-time library containing most
19607       common non-GUI Java classes, a bytecode interpreter, and the Boehm
19608       conservative garbage collector. Many bugs have been fixed. GCJ can
19609       compile Java source or Java bytecodes to either native code or Java
19610       class files, and supports native methods written in either the
19611       standard JNI or the more efficient and convenient CNI.
19612     * Here is a [6]partial list of C++ improvements, both new features
19613       and those no longer supported.
19614     * New C++ ABI. On the IA-64 platform GCC is capable of
19615       inter-operating with other IA-64 compilers.
19616     * The new ABI also significantly reduces the size of symbol and debug
19617       information.
19618     * New C++ support library and many C++ bug fixes, vastly improving
19619       our conformance to the ISO C++ standard.
19620     * New [7]inliner for C++.
19621     * Rewritten C preprocessor, integrated into the C, C++ and Objective
19622       C compilers, with very many improvements including ISO C99 support
19623       and [8]improvements to dependency generation.
19624     * Support for more [9]ISO C99 features.
19625     * Many improvements to support for checking calls to format functions
19626       such as printf and scanf, including support for ISO C99 format
19627       features, extensions from the Single Unix Specification and GNU
19628       libc 2.2, checking of strfmon formats and features to assist in
19629       auditing for format string security bugs.
19630     * New warnings for C code that may have undefined semantics because
19631       of violations of sequence point rules in the C standard (such as a
19632       = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall.
19633     * Additional warning option -Wfloat-equal.
19634     * Improvements to -Wtraditional.
19635     * Fortran improvements are listed in [10]the Fortran documentation.
19636
19637New Targets and Target Specific Improvements
19638
19639     * New x86 back end, generating much improved code.
19640     * Support for a generic i386-elf target contributed.
19641     * New option to emit x86 assembly code using Intel style syntax
19642       (-mintel-syntax).
19643     * HPUX 11 support contributed.
19644     * Improved PowerPC code generation, including scheduled prologue and
19645       epilogue.
19646     * Port of GCC to Intel's IA-64 processor contributed.
19647     * Port of GCC to Motorola's MCore 210 and 340 contributed.
19648     * New unified back-end for Arm, Thumb and StrongArm contributed.
19649     * Port of GCC to Intel's XScale processor contributed.
19650     * Port of GCC to Atmel's AVR microcontrollers contributed.
19651     * Port of GCC to Mitsubishi's D30V processor contributed.
19652     * Port of GCC to Matsushita's AM33 processor (a member of the MN10300
19653       processor family) contributed.
19654     * Port of GCC to Fujitsu's FR30 processor contributed.
19655     * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed.
19656     * Port of GCC to Sun's picoJava processor core contributed.
19657
19658Documentation improvements
19659
19660     * Substantially rewritten and improved C preprocessor manual.
19661     * Many improvements to other documentation.
19662     * Manpages for gcc, cpp and gcov are now generated automatically from
19663       the master Texinfo manual, eliminating the problem of manpages
19664       being out of date. (The generated manpages are only extracts from
19665       the full manual, which is provided in Texinfo form, from which
19666       info, HTML, other formats and a printed manual can be generated.)
19667     * Generated info files are included in the release tarballs alongside
19668       their Texinfo sources, avoiding problems on some platforms with
19669       building makeinfo as part of the GCC distribution.
19670
19671Other significant improvements
19672
19673     * Garbage collection used internally by the compiler for most memory
19674       allocation instead of obstacks.
19675     * Lengauer and Tarjan algorithm used for computing dominators in the
19676       CFG. This algorithm can be significantly faster and more space
19677       efficient than our older algorithm.
19678     * gccbug script provided to assist in submitting bug reports to our
19679       bug tracking system. (Bug reports previously submitted directly to
19680       our mailing lists, for which you received no bug tracking number,
19681       should be submitted again using gccbug if you can reproduce the
19682       problem with GCC 3.0.)
19683     * The internal libgcc library is [11]built as a shared library on
19684       systems that support it.
19685     * Extensive testsuite included with GCC, with many new tests. In
19686       addition to tests for GCC bugs that have been fixed, many tests
19687       have been added for language features, compiler warnings and
19688       builtin functions.
19689     * Additional language-independent warning options -Wpacked, -Wpadded,
19690       -Wunreachable-code and -Wdisabled-optimization.
19691     * Target-independent options -falign-functions, -falign-loops and
19692       -falign-jumps.
19693
19694   Plus a great many bug fixes and almost all the [12]features found in
19695   GCC 2.95.
19696
19697
19698    For questions related to the use of GCC, please consult these web
19699    pages and the [13]GCC manuals. If that fails, the
19700    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19701    web pages and the development of GCC are welcome on our developer
19702    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
19703    archives.
19704
19705   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
19706   distribution of this entire article is permitted in any medium,
19707   provided this notice is preserved.
19708
19709   These pages are [18]maintained by the GCC team. Last modified
19710   2021-07-28[19].
19711
19712References
19713
19714   1. http://www.netbsd.org/
19715   2. http://gcc.gnu.org/news/reorder.html
19716   3. http://gcc.gnu.org/news/ssa.html
19717   4. http://gcc.gnu.org/news/null.html
19718   5. http://gcc.gnu.org/news/unify.html
19719   6. http://gcc.gnu.org/gcc-3.0/c++features.html
19720   7. http://gcc.gnu.org/news/inlining.html
19721   8. http://gcc.gnu.org/news/dependencies.html
19722   9. http://gcc.gnu.org/c99status.html
19723  10. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
19724  11. http://gcc.gnu.org/gcc-3.0/libgcc.html
19725  12. http://gcc.gnu.org/gcc-2.95/features.html
19726  13. https://gcc.gnu.org/onlinedocs/
19727  14. mailto:gcc-help@gcc.gnu.org
19728  15. mailto:gcc@gcc.gnu.org
19729  16. https://gcc.gnu.org/lists.html
19730  17. https://www.fsf.org/
19731  18. https://gcc.gnu.org/about.html
19732  19. http://validator.w3.org/check/referer
19733======================================================================
19734http://gcc.gnu.org/gcc-3.0/caveats.html
19735                                GCC 3.0 Caveats
19736
19737     * -fstrict-aliasing is now part of -O2 and higher optimization
19738       levels. This allows the compiler to assume the strictest aliasing
19739       rules applicable to the language being compiled. For C and C++,
19740       this activates optimizations based on the type of expressions. This
19741       optimization may thus break old, non-compliant code.
19742     * Enumerations are now properly promoted to int in function
19743       parameters and function returns. Normally this change is not
19744       visible, but when using -fshort-enums this is an ABI change.
19745     * The undocumented extension that allowed C programs to have a label
19746       at the end of a compound statement has been deprecated and may be
19747       removed in a future version. Programs that now generate a warning
19748       about this may be fixed by adding a null statement (a single
19749       semicolon) after the label.
19750     * The poorly documented extension that allowed string constants in C,
19751       C++ and Objective C to contain unescaped newlines has been
19752       deprecated and may be removed in a future version. Programs using
19753       this extension may be fixed in several ways: the bare newline may
19754       be replaced by \n, or preceded by \n\, or string concatenation may
19755       be used with the bare newline preceded by \n" and " placed at the
19756       start of the next line.
19757     * The Chill compiler is not included in GCC 3.0, because of the lack
19758       of a volunteer to convert it to use garbage collection.
19759     * Certain non-standard iostream methods from earlier versions of
19760       libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach,
19761       ostream::form, and istream::gets.
19762     * The new C++ ABI is not yet fully supported by current (as of
19763       2001-07-01) releases and development versions of GDB, or any
19764       earlier versions. There is a problem setting breakpoints by line
19765       number, and other related issues that have been fixed in GCC 3.0
19766       but not yet handled in GDB:
19767       [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
19768
19769
19770    For questions related to the use of GCC, please consult these web
19771    pages and the [2]GCC manuals. If that fails, the
19772    [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19773    web pages and the development of GCC are welcome on our developer
19774    list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
19775
19776   Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
19777   distribution of this entire article is permitted in any medium,
19778   provided this notice is preserved.
19779
19780   These pages are [7]maintained by the GCC team. Last modified
19781   2021-07-28[8].
19782
19783References
19784
19785   1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
19786   2. https://gcc.gnu.org/onlinedocs/
19787   3. mailto:gcc-help@gcc.gnu.org
19788   4. mailto:gcc@gcc.gnu.org
19789   5. https://gcc.gnu.org/lists.html
19790   6. https://www.fsf.org/
19791   7. https://gcc.gnu.org/about.html
19792   8. http://validator.w3.org/check/referer
19793======================================================================
19794http://gcc.gnu.org/gcc-2.95/index.html
19795                                    GCC 2.95
19796
19797   (This release series is no longer supported.)
19798
19799   March 16, 2001: The GNU project and the GCC developers are pleased to
19800   announce the release of GCC version 2.95.3.
19801
19802Release History
19803
19804   GCC 2.95.3
19805          March 16, 2001
19806
19807   GCC 2.95.2
19808          October 27, 1999
19809
19810   GCC 2.95.1
19811          August 19, 1999
19812
19813   GCC 2.95
19814          July 31, 1999. This is the first release of GCC since the April
19815          1999 GCC/EGCS reunification and includes nearly a year's worth
19816          of new development and bugfixes.
19817
19818References and Acknowledgements
19819
19820   GCC used to stand for the GNU C Compiler, but since the compiler
19821   supports several other languages aside from C, it now stands for the
19822   GNU Compiler Collection.
19823
19824   The whole suite has been extensively [1]regression tested and
19825   [2]package tested. It should be reliable and suitable for widespread
19826   use.
19827
19828   The compiler has several new optimizations, new targets, new languages
19829   and other new features. See the [3]new features page for a more
19830   complete list of new features found in the GCC 2.95 releases.
19831
19832   The sources include installation instructions in both HTML and
19833   plaintext forms in the install directory in the distribution. However,
19834   the most up to date installation instructions and [4]build/test status
19835   are on the web pages. We will update those pages as new information
19836   becomes available.
19837
19838   The GCC developers would like to thank the numerous people that have
19839   contributed new features, test results, bugfixes, etc to GCC. This
19840   [5]amazing group of volunteers is what makes GCC successful.
19841
19842   And finally, we can't in good conscience fail to mention some
19843   [6]caveats to using GCC 2.95.
19844
19845   Download GCC 2.95 from one of our many [7]mirror sites.
19846
19847   For additional information about GCC please see the [8]GCC project web
19848   server or contact the [9]GCC development mailing list.
19849
19850
19851    For questions related to the use of GCC, please consult these web
19852    pages and the [10]GCC manuals. If that fails, the
19853    [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19854    web pages and the development of GCC are welcome on our developer
19855    list at [12]gcc@gcc.gnu.org. All of [13]our lists have public
19856    archives.
19857
19858   Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and
19859   distribution of this entire article is permitted in any medium,
19860   provided this notice is preserved.
19861
19862   These pages are [15]maintained by the GCC team. Last modified
19863   2021-07-28[16].
19864
19865References
19866
19867   1. http://gcc.gnu.org/gcc-2.95/regress.html
19868   2. http://gcc.gnu.org/gcc-2.95/othertest.html
19869   3. http://gcc.gnu.org/gcc-2.95/features.html
19870   4. http://gcc.gnu.org/gcc-2.95/buildstat.html
19871   5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
19872   6. http://gcc.gnu.org/gcc-2.95/caveats.html
19873   7. http://gcc.gnu.org/mirrors.html
19874   8. http://gcc.gnu.org/index.html
19875   9. mailto:gcc@gcc.gnu.org
19876  10. https://gcc.gnu.org/onlinedocs/
19877  11. mailto:gcc-help@gcc.gnu.org
19878  12. mailto:gcc@gcc.gnu.org
19879  13. https://gcc.gnu.org/lists.html
19880  14. https://www.fsf.org/
19881  15. https://gcc.gnu.org/about.html
19882  16. http://validator.w3.org/check/referer
19883======================================================================
19884http://gcc.gnu.org/gcc-2.95/features.html
19885                             GCC 2.95 New Features
19886
19887     * General Optimizer Improvements:
19888          + [1]Localized register spilling to improve speed and code
19889            density especially on small register class machines.
19890          + [2]Global CSE using lazy code motion algorithms.
19891          + [3]Improved global constant/copy propagation.
19892          + [4]Improved control flow graph analysis and manipulation.
19893          + [5]Local dead store elimination.
19894          + [6]Memory Load hoisting/store sinking in loops.
19895          + [7]Type based alias analysis is enabled by default. Note this
19896            feature will expose bugs in the Linux kernel. Please refer to
19897            the FAQ (as shipped with GCC 2.95) for additional information
19898            on this issue.
19899          + Major revamp of GIV detection, combination and simplification
19900            to improve loop performance.
19901          + Major improvements to register allocation and reloading.
19902     * New Languages and Language specific improvements
19903          + [8]Many C++ improvements.
19904          + [9]Many Fortran improvements.
19905          + [10]Java front-end has been integrated. A [11]runtime library
19906            is available separately.
19907          + [12]ISO C99 support
19908          + [13]Chill front-end and runtime has been integrated.
19909          + Boehm garbage collector support in libobjc.
19910          + More support for various pragmas which appear in vendor
19911            include files
19912     * New Targets and Target Specific Improvements
19913          + [14]SPARC backend rewrite.
19914          + -mschedule=8000 will optimize code for PA8000 class
19915            processors; -mpa-risc-2-0 will generate code for PA2.0
19916            processors
19917          + Various micro-optimizations for the ia32 port. K6
19918            optimizations
19919          + Compiler will attempt to align doubles in the stack on the
19920            ia32 port
19921          + Alpha EV6 support
19922          + PowerPC 750
19923          + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403.
19924            -mcpu=e603e was added to do -mcpu=603e and -msoft-float.
19925          + c3x, c4x
19926          + HyperSPARC
19927          + SparcLite86x
19928          + sh4
19929          + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix,
19930            arm-linux)
19931          + vxWorks targets include support for vxWorks threads
19932          + StrongARM 110 and ARM9 support added. ARM Scheduling
19933            parameters rewritten.
19934          + Various changes to the MIPS port to avoid assembler macros,
19935            which in turn improves performance
19936          + Various performance improvements to the i960 port.
19937          + Major rewrite of ns32k port
19938     * Other significant improvements
19939          + [15]Ability to dump cfg information and display it using vcg.
19940          + The new faster scheme for fixing vendor header files is
19941            enabled by default.
19942          + Experimental internationalization support.
19943          + multibyte character support
19944          + Some compile-time speedups for pathological problems
19945          + Better support for complex types
19946     * Plus the usual mountain of bugfixes
19947     * Core compiler is based on the gcc2 development tree from Sept 30,
19948       1998, so we have all of the [16]features found in GCC 2.8.
19949
19950Additional Changes in GCC 2.95.1
19951
19952     * Generic bugfixes and improvements
19953          + Various documentation fixes related to the GCC/EGCS merger.
19954          + Fix memory management bug which could lead to spurious aborts,
19955            core dumps or random parsing errors in the compiler.
19956          + Fix a couple bugs in the dwarf1 and dwarf2 debug record
19957            support.
19958          + Fix infinite loop in the CSE optimizer.
19959          + Avoid undefined behavior in compiler FP emulation code
19960          + Fix install problem when prefix is overridden on the make
19961            install command.
19962          + Fix problem with unwanted installation of assert.h on some
19963            systems.
19964          + Fix problem with finding the wrong assembler in a single tree
19965            build.
19966          + Avoid increasing the known alignment of a register that is
19967            already known to be a pointer.
19968     * Platform specific bugfixes and improvements
19969          + Codegen bugfix for prologue/epilogue for cpu32 target.
19970          + Fix long long code generation bug for the Coldfire target.
19971          + Fix various aborts in the SH compiler.
19972          + Fix bugs in libgcc support library for the SH.
19973          + Fix alpha ev6 code generation bug.
19974          + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on
19975            AIX platforms.
19976          + Fix -fpic code generation bug for rs6000/ppc svr4 targets.
19977          + Fix varargs/stdarg code generation bug for rs6000/ppc svr4
19978            targets.
19979          + Fix weak symbol handling for rs6000/ppc svr4 targets.
19980          + Fix various problems with 64bit code generation for the
19981            rs6000/ppc port.
19982          + Fix codegen bug which caused tetex to be mis-compiled on the
19983            x86.
19984          + Fix compiler abort in new cfg code exposed by x86 port.
19985          + Fix out of range array reference in code convert flat
19986            registers to the x87 stacked FP register file.
19987          + Fix minor vxworks configuration bug.
19988          + Fix return type of bsearch for SunOS 4.x.
19989     * Language & Runtime specific fixes.
19990          + The G++ signature extension has been deprecated. It will be
19991            removed in the next major release of G++. Use of signatures
19992            will result in a warning from the compiler.
19993          + Several bugs relating to templates and namespaces were fixed.
19994          + A bug that caused crashes when combining templates with -g on
19995            DWARF1 platforms was fixed.
19996          + Pointers-to-members, virtual functions, and multiple
19997            inheritance should now work together correctly.
19998          + Some code-generation bugs relating to function try blocks were
19999            fixed.
20000          + G++ is a little bit more lenient with certain archaic
20001            constructs than in GCC 2.95.
20002          + Fix to prevent shared library version #s from bring truncated
20003            to 1 digit
20004          + Fix missing std:: in the libstdc++ library.
20005          + Fix stream locking problems in libio.
20006          + Fix problem in java compiler driver.
20007
20008Additional Changes in GCC 2.95.2
20009
20010   The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While
20011   the optimizations performed by -fstrict-aliasing are valid according to
20012   the C and C++ standards, the optimization have caused some problems,
20013   particularly with old non-conforming code.
20014
20015   The GCC developers are experimenting with ways to warn users about code
20016   which violates the C/C++ standards, but those warnings are not ready
20017   for widespread use at this time. Rather than wait for those warnings
20018   the GCC developers have chosen to disable -fstrict-aliasing by default
20019   for the GCC 2.95.2 release.
20020
20021   We strongly encourage developers to find and fix code which violates
20022   the C/C++ standards as -fstrict-aliasing may be enabled by default in
20023   future releases. Use the option -fstrict-aliasing to re-enable these
20024   optimizations.
20025     * Generic bugfixes and improvements
20026          + Fix incorrectly optimized memory reference in global common
20027            subexpression elimination (GCSE) optimization pass.
20028          + Fix code generation bug in regmove.c in which it could
20029            incorrectly change a "const" value.
20030          + Fix bug in optimization of conditionals involving volatile
20031            memory references.
20032          + Avoid over-allocation of stack space for some procedures.
20033          + Fixed bug in the compiler which caused incorrect optimization
20034            of an obscure series of bit manipulations, shifts and
20035            arithmetic.
20036          + Fixed register allocator bug which caused teTeX to be
20037            mis-compiled on SPARC targets.
20038          + Avoid incorrect optimization of degenerate case statements for
20039            certain targets such as the ARM.
20040          + Fix out of range memory reference in the jump optimizer.
20041          + Avoid dereferencing null pointer in fix-header.
20042          + Fix test for GCC specific features so that it is possible to
20043            bootstrap with gcc-2.6.2 and older versions of GCC.
20044          + Fix typo in scheduler which could potentially cause out of
20045            range memory accesses.
20046          + Avoid incorrect loop reversal which caused incorrect code for
20047            certain loops on PowerPC targets.
20048          + Avoid incorrect optimization of switch statements on certain
20049            targets (for example the ARM).
20050     * Platform specific bugfixes and improvements
20051          + Work around bug in Sun V5.0 compilers which caused bootstrap
20052            comparison failures on SPARC targets.
20053          + Fix SPARC backend bug which caused aborts in final.c.
20054          + Fix sparc-hal-solaris2* configuration fragments.
20055          + Fix bug in sparc block profiling.
20056          + Fix obscure code generation bug for the PARISC targets.
20057          + Define __STDC_EXT__ for HPUX configurations.
20058          + Various POWERPC64 code generation bugfixes.
20059          + Fix abort for PPC targets using ELF (ex GNU/Linux).
20060          + Fix collect2 problems for AIX targets.
20061          + Correct handling of .file directive for PPC targets.
20062          + Fix bug in fix_trunc x86 patterns.
20063          + Fix x86 port to correctly pop the FP stack for functions that
20064            return structures in memory.
20065          + Fix minor bug in strlen x86 pattern.
20066          + Use stabs debugging instead of dwarf1 for x86-solaris targets.
20067          + Fix template repository code to handle leading underscore in
20068            mangled names.
20069          + Fix weak/weak alias support for OpenBSD.
20070          + GNU/Linux for the ARM has C++ compatible include files.
20071     * Language & Runtime specific fixes.
20072          + Fix handling of constructor attribute in the C front-end which
20073            caused problems building the Chill runtime library on some
20074            targets.
20075          + Fix minor problem merging type qualifiers in the C front-end.
20076          + Fix aliasing bug for pointers and references (C/C++).
20077          + Fix incorrect "non-constant initializer bug" when -traditional
20078            or -fwritable-strings is enabled.
20079          + Fix build error for Chill front-end on SunOS.
20080          + Do not complain about duplicate instantiations when using
20081            -frepo (C++).
20082          + Fix array bounds handling in C++ front-end which caused
20083            problems with dwarf debugging information in some
20084            circumstances.
20085          + Fix minor namespace problem.
20086          + Fix problem linking java programs.
20087
20088Additional Changes in GCC 2.95.3
20089
20090     * Generic bugfixes and improvements
20091          + Fix numerous problems that caused incorrect optimization in
20092            the register reloading code.
20093          + Fix numerous problems that caused incorrect optimization in
20094            the loop optimizer.
20095          + Fix aborts in the functions build_insn_chain and scan_loops
20096            under some circumstances.
20097          + Fix an alias analysis bug.
20098          + Fix an infinite compilation bug in the combiner.
20099          + A few problems with complex number support have been fixed.
20100          + It is no longer possible for gcc to act as a fork bomb when
20101            installed incorrectly.
20102          + The -fpack-struct option should be recognized now.
20103          + Fixed a bug that caused incorrect code to be generated due to
20104            a lost stack adjustment.
20105     * Platform specific bugfixes and improvements
20106          + Support building ARM toolchains hosted on Windows.
20107          + Fix attribute calculations in ARM toolchains.
20108          + arm-linux support has been improved.
20109          + Fix a PIC failure on sparc targets.
20110          + On ix86 targets, the regparm attribute should now work
20111            reliably.
20112          + Several updates for the h8300 port.
20113          + Fix problem building libio with glibc 2.2.
20114
20115
20116    For questions related to the use of GCC, please consult these web
20117    pages and the [17]GCC manuals. If that fails, the
20118    [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20119    web pages and the development of GCC are welcome on our developer
20120    list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
20121    archives.
20122
20123   Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
20124   distribution of this entire article is permitted in any medium,
20125   provided this notice is preserved.
20126
20127   These pages are [22]maintained by the GCC team. Last modified
20128   2021-07-28[23].
20129
20130References
20131
20132   1. http://gcc.gnu.org/news/spill.html
20133   2. http://gcc.gnu.org/news/lcm.html
20134   3. http://gcc.gnu.org/news/cprop.html
20135   4. http://gcc.gnu.org/news/cfg.html
20136   5. http://gcc.gnu.org/news/dse.html
20137   6. http://gcc.gnu.org/news/hoist.html
20138   7. http://gcc.gnu.org/news/alias.html
20139   8. http://gcc.gnu.org/gcc-2.95/c++features.html
20140   9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
20141  10. http://gcc.gnu.org/news/gcj-announce.txt
20142  11. http://gcc.gnu.org/news/javaannounce.html
20143  12. http://gcc.gnu.org/c99status.html
20144  13. http://gcc.gnu.org/news/chill.html
20145  14. http://gcc.gnu.org/news/sparc.html
20146  15. http://gcc.gnu.org/news/egcs-vcg.html
20147  16. http://gcc.gnu.org/egcs-1.0/features-2.8.html
20148  17. https://gcc.gnu.org/onlinedocs/
20149  18. mailto:gcc-help@gcc.gnu.org
20150  19. mailto:gcc@gcc.gnu.org
20151  20. https://gcc.gnu.org/lists.html
20152  21. https://www.fsf.org/
20153  22. https://gcc.gnu.org/about.html
20154  23. http://validator.w3.org/check/referer
20155======================================================================
20156http://gcc.gnu.org/gcc-2.95/caveats.html
20157                                GCC 2.95 Caveats
20158
20159     * GCC 2.95 will issue an error for invalid asm statements that had
20160       been silently accepted by earlier versions of the compiler. This is
20161       particularly noticeable when compiling older versions of the Linux
20162       kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95)
20163       for more information on this issue.
20164     * GCC 2.95 implements type based alias analysis to disambiguate
20165       memory references. Some programs, particularly the Linux kernel
20166       violate ANSI/ISO aliasing rules and therefore may not operate
20167       correctly when compiled with GCC 2.95. Please refer to the FAQ (as
20168       shipped with GCC 2.95) for more information on this issue.
20169     * GCC 2.95 has a known bug in its handling of complex variables for
20170       64bit targets. Instead of silently generating incorrect code, GCC
20171       2.95 will issue a fatal error for situations it can not handle.
20172       This primarily affects the Fortran community as Fortran makes more
20173       use of complex variables than C or C++.
20174     * GCC 2.95 has an integrated libstdc++, but does not have an
20175       integrated libg++. Furthermore old libg++ releases will not work
20176       with GCC 2.95. You can retrieve a recent copy of libg++ from the
20177       [1]GCC ftp server.
20178       Note most C++ programs only need libstdc++.
20179     * Exception handling may not work with shared libraries, particularly
20180       on alphas, hppas, rs6000/powerpc and mips based platforms.
20181       Exception handling is known to work on x86 GNU/Linux platforms with
20182       shared libraries.
20183     * In general, GCC 2.95 is more rigorous about rejecting invalid C++
20184       code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0,
20185       or EGCS 1.1. As a result it may be necessary to fix C++ code before
20186       it will compile with GCC 2.95.
20187     * G++ is also converting toward the ISO C++ standard; as a result
20188       code which was previously valid (and thus accepted by other
20189       compilers and older versions of g++) may no longer be accepted. The
20190       flag -fpermissive may allow some non-conforming code to compile
20191       with GCC 2.95.
20192     * GCC 2.95 compiled C++ code is not binary compatible with EGCS
20193       1.1.x, EGCS 1.0.x or GCC 2.8.x.
20194     * GCC 2.95 does not have changes from the GCC 2.8 tree that were made
20195       between Sept 30, 1998 and April 30, 1999 (the official end of the
20196       GCC 2.8 project). Future GCC releases will include all the changes
20197       from the defunct GCC 2.8 sources.
20198
20199
20200    For questions related to the use of GCC, please consult these web
20201    pages and the [2]GCC manuals. If that fails, the
20202    [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20203    web pages and the development of GCC are welcome on our developer
20204    list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
20205
20206   Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
20207   distribution of this entire article is permitted in any medium,
20208   provided this notice is preserved.
20209
20210   These pages are [7]maintained by the GCC team. Last modified
20211   2021-07-28[8].
20212
20213References
20214
20215   1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz
20216   2. https://gcc.gnu.org/onlinedocs/
20217   3. mailto:gcc-help@gcc.gnu.org
20218   4. mailto:gcc@gcc.gnu.org
20219   5. https://gcc.gnu.org/lists.html
20220   6. https://www.fsf.org/
20221   7. https://gcc.gnu.org/about.html
20222   8. http://validator.w3.org/check/referer
20223======================================================================
20224http://gcc.gnu.org/egcs-1.1/index.html
20225                                    EGCS 1.1
20226
20227   September 3, 1998: We are pleased to announce the release of EGCS 1.1.
20228   December 1, 1998: We are pleased to announce the release of EGCS 1.1.1.
20229   March 15, 1999: We are pleased to announce the release of EGCS 1.1.2.
20230
20231   EGCS is a free software project to further the development of the GNU
20232   compilers using an open development environment.
20233
20234   EGCS 1.1 is a major new release of the EGCS compiler system. It has
20235   been [1]extensively tested and is believed to be stable and suitable
20236   for widespread use.
20237
20238   EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8
20239   development sources; it contains all of the new features found in GCC
20240   2.8.1 as well as all new development from GCC up to June 6, 1998.
20241
20242   EGCS 1.1 also contains many improvements and features not found in GCC
20243   or in older versions of EGCS:
20244     * Global common subexpression elimination and global constant/copy
20245       propagation (aka [2]gcse)
20246     * Ongoing improvements to the [3]alias analysis support to allow for
20247       better optimizations throughout the compiler.
20248     * Vastly improved [4]C++ compiler and integrated C++ runtime
20249       libraries.
20250     * Fixes for the /tmp symlink race security problems.
20251     * New targets including mips16, arm-thumb and 64 bit PowerPC.
20252     * Improvements to GNU Fortran (g77) compiler and runtime library made
20253       since g77 version 0.5.23.
20254
20255   See the [5]new features page for a more complete list of new features
20256   found in EGCS 1.1 releases.
20257
20258   EGCS 1.1.1 is a minor update to fix several serious problems in EGCS
20259   1.1:
20260     * General improvements and fixes
20261          + Avoid some stack overflows when compiling large functions.
20262          + Avoid incorrect loop invariant code motions.
20263          + Fix some core dumps on Linux kernel code.
20264          + Bring back the imake -Di386 and friends fix from EGCS 1.0.2.
20265          + Fix code generation problem in gcse.
20266          + Various documentation related fixes.
20267     * g++/libstdc++ improvements and fixes
20268          + MT safe EH fix for setjmp/longjmp based exception handling.
20269          + Fix a few bad interactions between optimization and exception
20270            handling.
20271          + Fixes for demangling of template names starting with "__".
20272          + Fix a bug that would fail to run destructors in some cases
20273            with -O2.
20274          + Fix 'new' of classes with virtual bases.
20275          + Fix crash building Qt on the Alpha.
20276          + Fix failure compiling WIFEXITED macro on GNU/Linux.
20277          + Fix some -frepo failures.
20278     * g77 and libf2c improvements and fixes
20279          + Various documentation fixes.
20280          + Avoid compiler crash on RAND intrinsic.
20281          + Fix minor bugs in makefiles exposed by BSD make programs.
20282          + Define _XOPEN_SOURCE for libI77 build to avoid potential
20283            problems on some 64-bit systems.
20284          + Fix problem with implicit endfile on rewind.
20285          + Fix spurious recursive I/O errors.
20286     * platform specific improvements and fixes
20287          + Match all versions of UnixWare7.
20288          + Do not assume x86 SVR4 or UnixWare targets can handle stabs.
20289          + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion
20290            from unsigned ints to double precision floats.
20291          + Fix ARM ABI issue with NetBSD.
20292          + Fix a few arm code generation bugs.
20293          + Fixincludes will fix additional broken SCO OpenServer header
20294            files.
20295          + Fix a m68k backend bug which caused invalid offsets in reg+d
20296            addresses.
20297          + Fix problems with 64bit AIX 4.3 support.
20298          + Fix handling of long longs for varargs/stdarg functions on the
20299            ppc.
20300          + Minor fixes to CPP predefines for Windows.
20301          + Fix code generation problems with gpr<->fpr copies for 64bit
20302            ppc.
20303          + Fix a few coldfire code generation bugs.
20304          + Fix some more header file problems on SunOS 4.x.
20305          + Fix assert.h handling for RTEMS.
20306          + Fix Windows handling of TREE_SYMBOL_REFERENCED.
20307          + Fix x86 compiler abort in reg-stack pass.
20308          + Fix cygwin/windows problem with section attributes.
20309          + Fix Alpha code generation problem exposed by SMP Linux
20310            kernels.
20311          + Fix typo in m68k 32->64bit integer conversion.
20312          + Make sure target libraries build with -fPIC for PPC & Alpha
20313            targets.
20314
20315   EGCS 1.1.2 is a minor update to fix several serious problems in EGCS
20316   1.1.1:
20317     * General improvements and fixes
20318          + Fix bug in loop optimizer which caused the SPARC (and
20319            potentially other) ports to segfault.
20320          + Fix infinite recursion in alias analysis and combiner code.
20321          + Fix bug in regclass preferencing.
20322          + Fix incorrect loop reversal which caused incorrect code to be
20323            generated for several targets.
20324          + Fix return value for builtin memcpy.
20325          + Reduce compile time for certain loops which exposed quadratic
20326            behavior in the loop optimizer.
20327          + Fix bug which caused volatile memory to be written multiple
20328            times when only one write was needed/desired.
20329          + Fix compiler abort in caller-save.c
20330          + Fix combiner bug which caused incorrect code generation for
20331            certain division by constant operations.
20332          + Fix incorrect code generation due to a bug in range check
20333            optimizations.
20334          + Fix incorrect code generation due to mis-handling of clobbered
20335            values in CSE.
20336          + Fix compiler abort/segfault due to incorrect register
20337            splitting when unrolling loops.
20338          + Fix code generation involving autoincremented addresses with
20339            ternary operators.
20340          + Work around bug in the scheduler which caused qt to be
20341            mis-compiled on some platforms.
20342          + Fix code generation problems with -fshort-enums.
20343          + Tighten security for temporary files.
20344          + Improve compile time for codes which make heavy use of
20345            overloaded functions.
20346          + Fix multiply defined constructor/destructor symbol problems.
20347          + Avoid setting bogus RPATH environment variable during
20348            bootstrap.
20349          + Avoid GNU-make dependencies in the texinfo subdir.
20350          + Install CPP wrapper script in $(prefix)/bin if --enable-cpp.
20351            --enable-cpp=<dirname> can be used to specify an additional
20352            install directory for the cpp wrapper script.
20353          + Fix CSE bug which caused incorrect label-label refs to appear
20354            on some platforms.
20355          + Avoid linking in EH routines from libgcc if they are not
20356            needed.
20357          + Avoid obscure bug in aliasing code.
20358          + Fix bug in weak symbol handling.
20359     * Platform-specific improvements and fixes
20360          + Fix detection of PPro/PII on Unixware 7.
20361          + Fix compiler segfault when building spec99 and other programs
20362            for SPARC targets.
20363          + Fix code-generation bugs for integer and floating point
20364            conditional move instructions on the PPro/PII.
20365          + Use fixincludes to fix byteorder problems on i?86-*-sysv.
20366          + Fix build failure for the arc port.
20367          + Fix floating point format configuration for i?86-gnu port.
20368          + Fix problems with hppa1.0-hp-hpux10.20 configuration when
20369            threads are enabled.
20370          + Fix coldfire code generation bugs.
20371          + Fix "unrecognized insn" problems for Alpha and PPC ports.
20372          + Fix h8/300 code generation problem with floating point values
20373            in memory.
20374          + Fix unrecognized insn problems for the m68k port.
20375          + Fix namespace-pollution problem for the x86 port.
20376          + Fix problems with old assembler on x86 NeXT systems.
20377          + Fix PIC code-generation problems for the SPARC port.
20378          + Fix minor bug with LONG_CALLS in PowerPC SVR4 support.
20379          + Fix minor ISO namespace violation in Alpha varargs/stdarg
20380            support.
20381          + Fix incorrect "braf" instruction usage for the SH port.
20382          + Fix minor bug in va-sh which prevented its use with -ansi.
20383          + Fix problems recognizing and supporting FreeBSD.
20384          + Handle OpenBSD systems correctly.
20385          + Minor fixincludes fix for Digital UNIX 4.0B.
20386          + Fix problems with ctors/dtors in SCO shared libraries.
20387          + Abort instead of generating incorrect code for PPro/PII
20388            floating point conditional moves.
20389          + Avoid multiply defined symbols on GNU/Linux systems using
20390            libc-5.4.xx.
20391          + Fix abort in alpha compiler.
20392     * Fortran-specific fixes
20393          + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year
20394            is in the documented, non-Y2K-compliant range of 0-99, instead
20395            of being returned as 100 in the year 2000.
20396          + Fix the `Date_and_Time' intrinsic (in libg2c) to return the
20397            milliseconds value properly in Values(8).
20398          + Fix the `LStat' intrinsic (in libg2c) to return device-ID
20399            information properly in SArray(7).
20400
20401   Each release includes installation instructions in both HTML and
20402   plaintext forms (see the INSTALL directory in the toplevel directory of
20403   the distribution). However, we also keep the most up to date
20404   installation instructions and [6]build/test status on our web page. We
20405   will update those pages as new information becomes available.
20406
20407   The EGCS project would like to thank the numerous people that have
20408   contributed new features, test results, bugfixes, etc. This [7]amazing
20409   group of volunteers is what makes EGCS successful.
20410
20411   And finally, we can't in good conscience fail to mention some
20412   [8]caveats to using EGCS 1.1.
20413
20414   Download EGCS from egcs.cygnus.com (USA California).
20415
20416   The EGCS 1.1 release is also available on many mirror sites.
20417   [9]Goto mirror list to find a closer site.
20418
20419
20420    For questions related to the use of GCC, please consult these web
20421    pages and the [10]GCC manuals. If that fails, the
20422    [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20423    web pages and the development of GCC are welcome on our developer
20424    list at [12]gcc@gcc.gnu.org. All of [13]our lists have public
20425    archives.
20426
20427   Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and
20428   distribution of this entire article is permitted in any medium,
20429   provided this notice is preserved.
20430
20431   These pages are [15]maintained by the GCC team. Last modified
20432   2021-07-28[16].
20433
20434References
20435
20436   1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html
20437   2. http://gcc.gnu.org/news/gcse.html
20438   3. http://gcc.gnu.org/news/alias.html
20439   4. http://gcc.gnu.org/egcs-1.1/c++features.html
20440   5. http://gcc.gnu.org/egcs-1.1/features.html
20441   6. http://gcc.gnu.org/egcs-1.1/buildstat.html
20442   7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
20443   8. http://gcc.gnu.org/egcs-1.1/caveats.html
20444   9. http://gcc.gnu.org/mirrors.html
20445  10. https://gcc.gnu.org/onlinedocs/
20446  11. mailto:gcc-help@gcc.gnu.org
20447  12. mailto:gcc@gcc.gnu.org
20448  13. https://gcc.gnu.org/lists.html
20449  14. https://www.fsf.org/
20450  15. https://gcc.gnu.org/about.html
20451  16. http://validator.w3.org/check/referer
20452======================================================================
20453http://gcc.gnu.org/egcs-1.1/features.html
20454                             EGCS 1.1 new features
20455
20456     * Integrated GNU Fortran (g77) compiler and runtime library with
20457       improvements, based on g77 version 0.5.23.
20458     * Vast improvements in the C++ compiler; so many they have [1]page of
20459       their own!
20460     * Compiler implements [2]global common subexpression elimination and
20461       global copy/constant propagation.
20462     * More major improvements in the [3]alias analysis code.
20463     * More major improvements in the exception handling code to improve
20464       performance, lower static overhead and provide the infrastructure
20465       for future improvements.
20466     * The infamous /tmp symlink race security problems have been fixed.
20467     * The regmove optimization pass has been nearly completely rewritten
20468       to improve performance of generated code.
20469     * The compiler now recomputes register usage information before local
20470       register allocation. By providing more accurate information to the
20471       priority based allocator, we get better register allocation.
20472     * The register reloading phase of the compiler optimizes spill code
20473       much better than in previous releases.
20474     * Some bad interactions between the register allocator and
20475       instruction scheduler have been fixed, resulting in much better
20476       code for certain programs. Additionally, we have tuned the
20477       scheduler in various ways to improve performance of generated code
20478       for some architectures.
20479     * The compiler's branch shortening algorithms have been significantly
20480       improved to work better on targets which align jump targets.
20481     * The compiler now supports -Os to prefer optimizing for code space
20482       over optimizing for code speed.
20483     * The compiler will now totally eliminate library calls which compute
20484       constant values. This primarily helps targets with no integer
20485       div/mul support and targets without floating point support.
20486     * The compiler now supports an extensive "--help" option.
20487     * cpplib has been greatly improved and may be suitable for limited
20488       use.
20489     * Memory footprint for the compiler has been significantly reduced
20490       for some pathological cases.
20491     * The time to build EGCS has been improved for certain targets
20492       (particularly the alpha and mips platforms).
20493     * Many infrastructure improvements throughout the compiler, plus the
20494       usual mountain of bugfixes and minor improvements.
20495     * Target dependent improvements:
20496          + SPARC port now includes V8 plus and V9 support as well as
20497            performance tuning for Ultra class machines. The SPARC port
20498            now uses the Haifa scheduler.
20499          + Alpha port has been tuned for the EV6 processor and has an
20500            optimized expansion of memcpy/bzero. The Alpha port now uses
20501            the Haifa scheduler.
20502          + RS6000/PowerPC: support for the Power64 architecture and AIX
20503            4.3. The RS6000/PowerPC port now uses the Haifa scheduler.
20504          + x86: Alignment of static store data and jump targets is per
20505            Intel recommendations now. Various improvements throughout the
20506            x86 port to improve performance on Pentium processors
20507            (including improved epilogue sequences for Pentium chips and
20508            backend improvements which should help register allocation on
20509            all x86 variants. Conditional move support has been fixed and
20510            enabled for PPro processors. The x86 port also better supports
20511            64bit operations now. Unixware 7, a System V Release 5 target,
20512            is now supported and SCO OpenServer targets can support GAS.
20513          + MIPS has improved multiply/multiply-add support and now
20514            includes mips16 ISA support.
20515          + M68k has many micro-optimizations and Coldfire fixes.
20516     * Core compiler is based on the GCC development tree from June 9,
20517       1998, so we have all of the [4]features found in GCC 2.8.
20518
20519
20520    For questions related to the use of GCC, please consult these web
20521    pages and the [5]GCC manuals. If that fails, the
20522    [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20523    web pages and the development of GCC are welcome on our developer
20524    list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
20525
20526   Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
20527   distribution of this entire article is permitted in any medium,
20528   provided this notice is preserved.
20529
20530   These pages are [10]maintained by the GCC team. Last modified
20531   2021-07-28[11].
20532
20533References
20534
20535   1. http://gcc.gnu.org/egcs-1.1/c++features.html
20536   2. http://gcc.gnu.org/news/gcse.html
20537   3. http://gcc.gnu.org/news/alias.html
20538   4. http://gcc.gnu.org/egcs-1.0/features-2.8.html
20539   5. https://gcc.gnu.org/onlinedocs/
20540   6. mailto:gcc-help@gcc.gnu.org
20541   7. mailto:gcc@gcc.gnu.org
20542   8. https://gcc.gnu.org/lists.html
20543   9. https://www.fsf.org/
20544  10. https://gcc.gnu.org/about.html
20545  11. http://validator.w3.org/check/referer
20546======================================================================
20547http://gcc.gnu.org/egcs-1.1/caveats.html
20548                                EGCS 1.1 Caveats
20549
20550     * EGCS has an integrated libstdc++, but does not have an integrated
20551       libg++. Furthermore old libg++ releases will not work with EGCS; HJ
20552       Lu has made a libg++-2.8.1.2 snapshot available which may work with
20553       EGCS.
20554       Note most C++ programs only need libstdc++.
20555     * Exception handling may not work with shared libraries, particularly
20556       on alphas, hppas, rs6000/powerpc and mips based platforms.
20557       Exception handling is known to work on x86-linux platforms with
20558       shared libraries.
20559     * Some versions of the Linux kernel have bugs which prevent them from
20560       being compiled or from running when compiled by EGCS. See the FAQ
20561       (as shipped with EGCS 1.1) for additional information.
20562     * In general, EGCS is more rigorous about rejecting invalid C++ code
20563       or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As
20564       a result it may be necessary to fix C++ code before it will compile
20565       with EGCS.
20566     * G++ is also converting toward the ISO C++ standard; as a result
20567       code which was previously valid (and thus accepted by other
20568       compilers and older versions of g++) may no longer be accepted.
20569     * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x
20570       or GCC 2.8.x due to changes necessary to support thread safe
20571       exception handling.
20572
20573
20574    For questions related to the use of GCC, please consult these web
20575    pages and the [1]GCC manuals. If that fails, the
20576    [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20577    web pages and the development of GCC are welcome on our developer
20578    list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
20579
20580   Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
20581   distribution of this entire article is permitted in any medium,
20582   provided this notice is preserved.
20583
20584   These pages are [6]maintained by the GCC team. Last modified
20585   2021-07-28[7].
20586
20587References
20588
20589   1. https://gcc.gnu.org/onlinedocs/
20590   2. mailto:gcc-help@gcc.gnu.org
20591   3. mailto:gcc@gcc.gnu.org
20592   4. https://gcc.gnu.org/lists.html
20593   5. https://www.fsf.org/
20594   6. https://gcc.gnu.org/about.html
20595   7. http://validator.w3.org/check/referer
20596======================================================================
20597http://gcc.gnu.org/egcs-1.0/index.html
20598                                    EGCS 1.0
20599
20600   December 3, 1997: We are pleased to announce the release of EGCS 1.0.
20601   January 6, 1998: We are pleased to announce the release of EGCS 1.0.1.
20602   March 16, 1998: We are pleased to announce the release of EGCS 1.0.2.
20603   May 15, 1998 We are pleased to announce the release of EGCS 1.0.3.
20604
20605   EGCS is a collaborative effort involving several groups of hackers
20606   using an open development model to accelerate development and testing
20607   of GNU compilers and runtime libraries.
20608
20609   An important goal of EGCS is to allow wide scale testing of
20610   experimental features and optimizations; therefore, EGCS contains some
20611   features and optimizations which are still under development. However,
20612   EGCS has been carefully tested and should be comparable in quality to
20613   most GCC releases.
20614
20615   EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8
20616   development sources; it contains nearly all of the new features found
20617   in GCC 2.8.
20618
20619   EGCS 1.0 also contains many improvements and features not found in GCC
20620   2.7 and even the GCC 2.8 series (which was released after the original
20621   EGCS 1.0 release).
20622     * Integrated C++ runtime libraries, including support for most major
20623       GNU/Linux systems!
20624     * The integrated libstdc++ library includes a verbatim copy of SGI's
20625       STL release.
20626     * Integrated GNU Fortran compiler.
20627     * New instruction scheduler.
20628     * New alias analysis code.
20629
20630   See the [1]new features page for a more complete list of new features.
20631
20632   EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few
20633   critical bugs and add support for Red Hat 5.0 Linux. Changes since the
20634   EGCS 1.0 release:
20635     * Add support for Red Hat 5.0 Linux and better support for Linux
20636       systems using glibc2.
20637       Many programs failed to link when compiled with EGCS 1.0 on Red Hat
20638       5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should
20639       fix these problems.
20640     * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception
20641       handling interfaces.
20642       To avoid future compatibility problems, we strongly urge anyone who
20643       is planning on distributing shared libraries that contain C++ code
20644       to upgrade to EGCS 1.0.1 first.
20645       Soon after EGCS 1.0 was released, the GCC developers made some
20646       incompatible changes in libgcc's exception handling interfaces.
20647       These changes were needed to solve problems on some platforms. This
20648       means that GCC 2.8.0, when released, will not be seamlessly
20649       compatible with shared libraries built by EGCS 1.0. The reason is
20650       that the libgcc.a in GCC 2.8.0 will not contain a function needed
20651       by the old interface.
20652       The result of this is that there may be compatibility problems with
20653       shared libraries built by EGCS 1.0 when used with GCC 2.8.0.
20654       With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface,
20655       and libgcc.a has the support routines for both the old and the new
20656       interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed,
20657       and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed).
20658       The maintainers of GCC 2.x have decided against including seamless
20659       support for the old interface in 2.8.0, since it was never
20660       "official", so to avoid future compatibility problems we recommend
20661       against distributing any shared libraries built by EGCS 1.0 that
20662       contain C++ code (upgrade to 1.0.1 and use that).
20663     * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends.
20664       The x86 changes fix code generation errors exposed when building
20665       glibc2 and the usual GNU/Linux dynamic linker (ld.so).
20666       The hppa change fixes a compiler abort when configured for use with
20667       RTEMS.
20668       The MIPS changes fix problems with the definition of LONG_MAX on
20669       newer systems, allow for command line selection of the target ABI,
20670       and fix one code generation problem.
20671       The rs6000/ppc change fixes some problems with passing structures
20672       to varargs/stdarg functions.
20673     * A few machine independent bugfixes, mostly to fix code generation
20674       errors when building Linux kernels or glibc.
20675     * Fix a few critical exception handling and template bugs in the C++
20676       compiler.
20677     * Fix Fortran namelist bug on alphas.
20678     * Fix build problems on x86-solaris systems.
20679
20680   EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several
20681   serious problems in EGCS 1.0.1.
20682     * General improvements and fixes
20683          + Memory consumption significantly reduced, especially for
20684            templates and inline functions.
20685          + Fix various problems with glibc2.1.
20686          + Fix loop optimization bug exposed by rs6000/ppc port.
20687          + Fix to avoid potential code generation problems in jump.c.
20688          + Fix some undefined symbol problems in dwarf1 debug support.
20689     * g++/libstdc++ improvements and fixes
20690          + libstdc++ in the EGCS release has been updated and should be
20691            link compatible with libstdc++-2.8.
20692          + Various fixes in libio/libstdc++ to work better on GNU/Linux
20693            systems.
20694          + Fix problems with duplicate symbols on systems that do not
20695            support weak symbols.
20696          + Memory corruption bug and undefined symbols in bastring have
20697            been fixed.
20698          + Various exception handling fixes.
20699          + Fix compiler abort for very long thunk names.
20700     * g77 improvements and fixes
20701          + Fix compiler crash for omitted bound in Fortran CASE
20702            statement.
20703          + Add missing entries to g77 lang-options.
20704          + Fix problem with -fpedantic in the g77 compiler.
20705          + Fix "backspace" problem with g77 on alphas.
20706          + Fix x86 backend problem with Fortran literals and -fpic.
20707          + Fix some of the problems with negative subscripts for g77 on
20708            alphas.
20709          + Fixes for Fortran builds on cygwin32/mingw32.
20710     * platform specific improvements and fixes
20711          + Fix long double problems on x86 (exposed by glibc).
20712          + x86 ports define i386 again to keep imake happy.
20713          + Fix exception handling support on NetBSD ports.
20714          + Several changes to collect2 to fix many problems with AIX.
20715          + Define __ELF__ for GNU/Linux on rs6000.
20716          + Fix -mcall-linux problem on GNU/Linux on rs6000.
20717          + Fix stdarg/vararg problem for GNU/Linux on rs6000.
20718          + Allow autoconf to select a proper install problem on AIX 3.1.
20719          + m68k port support includes -mcpu32 option as well as cpu32
20720            multilibs.
20721          + Fix stdarg bug for irix6.
20722          + Allow EGCS to build on irix5 without the gnu assembler.
20723          + Fix problem with static linking on sco5.
20724          + Fix bootstrap on sco5 with native compiler.
20725          + Fix for abort building newlib on H8 target.
20726          + Fix fixincludes handling of math.h on SunOS.
20727          + Minor fix for Motorola 3300 m68k systems.
20728
20729   EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few
20730   problems reported by Red Hat for builds of Red Hat 5.1.
20731     * Generic bugfixes:
20732          + Fix a typo in the libio library which resulted in incorrect
20733            behavior of istream::get.
20734          + Fix the Fortran negative array index problem.
20735          + Fix a major problem with the ObjC runtime thread support
20736            exposed by glibc2.
20737          + Reduce memory consumption of the Haifa scheduler.
20738     * Target specific bugfixes:
20739          + Fix one x86 floating point code generation bug exposed by
20740            glibc2 builds.
20741          + Fix one x86 internal compiler error exposed by glibc2 builds.
20742          + Fix profiling bugs on the Alpha.
20743          + Fix ImageMagick & emacs 20.2 build problems on the Alpha.
20744          + Fix rs6000/ppc bug when converting values from integer types
20745            to floating point types.
20746
20747   The EGCS 1.0 releases include installation instructions in both HTML
20748   and plaintext forms (see the INSTALL directory in the toplevel
20749   directory of the distribution). However, we also keep the most up to
20750   date installation instructions and [2]build/test status on our web
20751   page. We will update those pages as new information becomes available.
20752
20753   And, we can't in good conscience fail to mention some [3]caveats to
20754   using EGCS.
20755
20756   Update: Big thanks to Stanford for providing a high speed link for
20757   downloading EGCS (go.cygnus.com)!
20758
20759   Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com
20760   (USA California -- High speed link provided by Stanford).
20761
20762   The EGCS 1.0 release is also available many mirror sites.
20763   [4]Goto mirror list to find a closer site
20764
20765   We'd like to thank the numerous people that have contributed new
20766   features, test results, bugfixes, etc. Unfortunately, they're far too
20767   numerous to mention by name.
20768
20769
20770    For questions related to the use of GCC, please consult these web
20771    pages and the [5]GCC manuals. If that fails, the
20772    [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20773    web pages and the development of GCC are welcome on our developer
20774    list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
20775
20776   Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
20777   distribution of this entire article is permitted in any medium,
20778   provided this notice is preserved.
20779
20780   These pages are [10]maintained by the GCC team. Last modified
20781   2021-07-28[11].
20782
20783References
20784
20785   1. http://gcc.gnu.org/egcs-1.0/features.html
20786   2. http://gcc.gnu.org/egcs-1.0/buildstat.html
20787   3. http://gcc.gnu.org/egcs-1.0/caveats.html
20788   4. http://gcc.gnu.org/mirrors.html
20789   5. https://gcc.gnu.org/onlinedocs/
20790   6. mailto:gcc-help@gcc.gnu.org
20791   7. mailto:gcc@gcc.gnu.org
20792   8. https://gcc.gnu.org/lists.html
20793   9. https://www.fsf.org/
20794  10. https://gcc.gnu.org/about.html
20795  11. http://validator.w3.org/check/referer
20796======================================================================
20797http://gcc.gnu.org/egcs-1.0/features.html
20798                               EGCS 1.0 features
20799
20800     * Core compiler is based on the gcc2 development tree from Aug 2,
20801       1997, so we have most of the [1]features found in GCC 2.8.
20802     * Integrated GNU Fortran compiler based on g77-0.5.22-19970929.
20803     * Vast improvements in the C++ compiler; so many they have [2]page of
20804       their own!
20805     * Integrated C++ runtime libraries, including support for most major
20806       GNU/Linux systems!
20807     * New instruction scheduler from IBM Haifa which includes support for
20808       function wide instruction scheduling as well as superscalar
20809       scheduling.
20810     * Significantly improved alias analysis code.
20811     * Improved register allocation for two address machines.
20812     * Significant code generation improvements for Fortran code on
20813       Alphas.
20814     * Various optimizations from the g77 project as well as improved loop
20815       optimizations.
20816     * Dwarf2 debug format support for some targets.
20817     * egcs libstdc++ includes the SGI STL implementation without changes.
20818     * As a result of these and other changes, egcs libstc++ is not binary
20819       compatible with previous releases of libstdc++.
20820     * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO
20821       Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and
20822       1.1), Support for RTEMS on several embedded targets, Support for
20823       arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and
20824       MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc.
20825     * Integrated testsuites for gcc, g++, g77, libstdc++ and libio.
20826     * RS6000/PowerPC ports generate code which can run on all
20827       RS6000/PowerPC variants by default.
20828     * -mcpu= and -march= switches for the x86 port to allow better
20829       control over how the x86 port generates code.
20830     * Includes the template repository patch (aka repo patch); note the
20831       new template code makes repo obsolete for ELF systems using gnu-ld
20832       such as GNU/Linux.
20833     * Plus the usual assortment of bugfixes and improvements.
20834
20835
20836    For questions related to the use of GCC, please consult these web
20837    pages and the [3]GCC manuals. If that fails, the
20838    [4]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20839    web pages and the development of GCC are welcome on our developer
20840    list at [5]gcc@gcc.gnu.org. All of [6]our lists have public archives.
20841
20842   Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and
20843   distribution of this entire article is permitted in any medium,
20844   provided this notice is preserved.
20845
20846   These pages are [8]maintained by the GCC team. Last modified
20847   2021-07-28[9].
20848
20849References
20850
20851   1. http://gcc.gnu.org/egcs-1.0/features-2.8.html
20852   2. http://gcc.gnu.org/egcs-1.0/c++features.html
20853   3. https://gcc.gnu.org/onlinedocs/
20854   4. mailto:gcc-help@gcc.gnu.org
20855   5. mailto:gcc@gcc.gnu.org
20856   6. https://gcc.gnu.org/lists.html
20857   7. https://www.fsf.org/
20858   8. https://gcc.gnu.org/about.html
20859   9. http://validator.w3.org/check/referer
20860======================================================================
20861http://gcc.gnu.org/egcs-1.0/caveats.html
20862                                EGCS 1.0 Caveats
20863
20864     * EGCS has an integrated libstdc++, but does not have an integrated
20865       libg++. Furthermore old libg++ releases will not work with egc; HJ
20866       Lu has made a libg++-2.8.1.2 available which may work with EGCS.
20867       Note most C++ programs only need libstdc++.
20868     * Note that using -pedantic or -Wreturn-type can cause an explosion
20869       in the amount of memory needed for template-heavy C++ code, such as
20870       code that uses STL. Also note that -Wall includes -Wreturn-type, so
20871       if you use -Wall you will need to specify -Wno-return-type to turn
20872       it off.
20873     * Exception handling may not work with shared libraries, particularly
20874       on alphas, hppas, and mips based platforms. Exception handling is
20875       known to work on x86-linux platforms with shared libraries.
20876     * Some versions of the Linux kernel have bugs which prevent them from
20877       being compiled or from running when compiled by EGCS. See the FAQ
20878       (as shipped with EGCS 1.0) for additional information.
20879     * In general, EGCS is more rigorous about rejecting invalid C++ code
20880       or deprecated C++ constructs than G++ 2.7. As a result it may be
20881       necessary to fix C++ code before it will compile with EGCS.
20882     * G++ is also aggressively tracking the C++ standard; as a result
20883       code which was previously valid (and thus accepted by other
20884       compilers and older versions of G++) may no longer be accepted.
20885     * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS
20886       1.0.x and later releases should work with Red Hat Linux 5.0.
20887
20888
20889    For questions related to the use of GCC, please consult these web
20890    pages and the [1]GCC manuals. If that fails, the
20891    [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
20892    web pages and the development of GCC are welcome on our developer
20893    list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
20894
20895   Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
20896   distribution of this entire article is permitted in any medium,
20897   provided this notice is preserved.
20898
20899   These pages are [6]maintained by the GCC team. Last modified
20900   2021-07-28[7].
20901
20902References
20903
20904   1. https://gcc.gnu.org/onlinedocs/
20905   2. mailto:gcc-help@gcc.gnu.org
20906   3. mailto:gcc@gcc.gnu.org
20907   4. https://gcc.gnu.org/lists.html
20908   5. https://www.fsf.org/
20909   6. https://gcc.gnu.org/about.html
20910   7. http://validator.w3.org/check/referer
20911======================================================================
20912