xref: /netbsd-src/external/gpl3/gcc.old/dist/NEWS (revision 53d1339bf7f9c7367b35a9e1ebe693f9b047a47b)
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-9/index.html
9                              GCC 9 Release Series
10
11   Mar 12, 2020
12
13   The [1]GNU project and the GCC developers are pleased to announce the
14   release of GCC 9.3.
15
16   This release is a bug-fix release, containing fixes for regressions in
17   GCC 9.2 relative to previous releases of GCC.
18
19Release History
20
21   GCC 9.3
22          Mar 12, 2020 ([2]changes, [3]documentation)
23
24   GCC 9.2
25          Aug 12, 2019 ([4]changes, [5]documentation)
26
27   GCC 9.1
28          May 3, 2019 ([6]changes, [7]documentation)
29
30References and Acknowledgements
31
32   GCC used to stand for the GNU C Compiler, but since the compiler
33   supports several other languages aside from C, it now stands for the
34   GNU Compiler Collection.
35
36   A list of [8]successful builds is updated as new information becomes
37   available.
38
39   The GCC developers would like to thank the numerous people that have
40   contributed new features, improvements, bug fixes, and other changes as
41   well as test results to GCC. This [9]amazing group of volunteers is
42   what makes GCC successful.
43
44   For additional information about GCC please refer to the [10]GCC
45   project web site or contact the [11]GCC development mailing list.
46
47   To obtain GCC please use [12]our mirror sites or [13]our version
48   control system.
49
50
51    For questions related to the use of GCC, please consult these web
52    pages and the [14]GCC manuals. If that fails, the
53    [15]gcc-help@gcc.gnu.org mailing list might help. Comments on these
54    web pages and the development of GCC are welcome on our developer
55    list at [16]gcc@gcc.gnu.org. All of [17]our lists have public
56    archives.
57
58   Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and
59   distribution of this entire article is permitted in any medium,
60   provided this notice is preserved.
61
62   These pages are [19]maintained by the GCC team. Last modified
63   2020-03-12[20].
64
65References
66
67   1. http://www.gnu.org/
68   2. http://gcc.gnu.org/gcc-9/changes.html
69   3. http://gcc.gnu.org/onlinedocs/9.3.0/
70   4. http://gcc.gnu.org/gcc-9/changes.html
71   5. http://gcc.gnu.org/onlinedocs/9.2.0/
72   6. http://gcc.gnu.org/gcc-9/changes.html
73   7. http://gcc.gnu.org/onlinedocs/9.1.0/
74   8. http://gcc.gnu.org/gcc-9/buildstat.html
75   9. http://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Contributors.html
76  10. http://gcc.gnu.org/index.html
77  11. mailto:gcc@gcc.gnu.org
78  12. http://gcc.gnu.org/mirrors.html
79  13. http://gcc.gnu.org/git.html
80  14. https://gcc.gnu.org/onlinedocs/
81  15. mailto:gcc-help@gcc.gnu.org
82  16. mailto:gcc@gcc.gnu.org
83  17. https://gcc.gnu.org/lists.html
84  18. https://www.fsf.org/
85  19. https://gcc.gnu.org/about.html
86  20. http://validator.w3.org/check/referer
87======================================================================
88http://gcc.gnu.org/gcc-9/changes.html
89                              GCC 9 Release Series
90                        Changes, New Features, and Fixes
91
92   This page is a "brief" summary of some of the huge number of
93   improvements in GCC 9. You may also want to check out our [1]Porting to
94   GCC 9 page and the [2]full GCC documentation.
95
96Caveats
97
98     * On Arm targets (arm*-*-*), [3]a bug in the implementation of the
99       procedure call standard (AAPCS) in the GCC 6, 7 and 8 releases has
100       been fixed: a structure containing a bit-field based on a 64-bit
101       integral type and where no other element in a structure required
102       64-bit alignment could be passed incorrectly to functions. This is
103       an ABI change. If the option -Wpsabi is enabled (on by default) the
104       compiler will emit a diagnostic note for code that might be
105       affected.
106     * Support for a number of older systems and recently unmaintained or
107       untested target ports of GCC has been declared obsolete in GCC 9.
108       Unless there is activity to revive them, the next release of GCC
109       will have their sources permanently removed.
110       The following ports for individual systems on particular
111       architectures have been obsoleted:
112          + Solaris 10 (*-*-solaris2.10). Details can be found in the
113            [4]announcement.
114          + Cell Broadband Engine SPU (spu*-*-*). Details can be found in
115            the [5]announcement.
116     * A change to the C++ std::rotate algorithm in GCC 9.1.0 can cause
117       ABI incompatibilities with object files compiled with other
118       versions of GCC. If the std::rotate algorithm is called with an
119       empty range then it might cause a divide-by-zero error (as a SIGFPE
120       signal) and crash. The change has been reverted for GCC 9.2.0 and
121       future releases. For more details see [6]Bug 90920. The problem can
122       be avoided by recompiling any objects that might call std::rotate
123       with an empty range, so that the GCC 9.1.0 definition of
124       std::rotate is not used.
125     * The automatic template instantiation at link time ([7]-frepo) has
126       been deprecated and will be removed in a future release.
127     * The --with-default-libstdcxx-abi=gcc4-compatible configure option
128       is broken in the 9.1 and 9.2 releases, producing a shared library
129       with missing symbols (see [8]Bug 90361). As a workaround, configure
130       without that option and build GCC as normal, then edit the
131       installed <bits/c++config.h> headers to define the
132       _GLIBCXX_USE_CXX11_ABI macro to 0.
133
134General Improvements
135
136   The following GCC command line options have been introduced or
137   improved.
138     * All command line options that take a byte-size argument accept
139       64-bit integers as well as standard SI and IEC suffixes such as kb
140       and KiB, MB and MiB, or GB and GiB denoting the corresponding
141       multiples of bytes. See [9]Invoking GCC for more.
142     * A new option,
143       [10]-flive-patching=[inline-only-static|inline-clone], has been
144       introduced to provide a safe compilation for live-patching. At the
145       same time, provides multiple-level control on the enabled IPA
146       optimizations. See the user guide for more details about the
147       option.
148     * A new option, --completion, has been added to provide more fine
149       option completion in a shell. It is intended to be used by
150       Bash-completion.
151     * GCC's diagnostics now print source code with a left margin showing
152       line numbers, configurable with
153       [11]-fno-diagnostics-show-line-numbers.
154       GCC's diagnostics can also now label regions of the source code to
155       show pertinent information, such as the types within an expression.
156$ g++ t.cc
157t.cc: In function 'int test(const shape&, const shape&)':
158t.cc:15:4: error: no match for 'operator+' (operand types are 'boxed_value<doubl
159e>' and 'boxed_value<double>')
160   14 |   return (width(s1) * height(s1)
161      |           ~~~~~~~~~~~~~~~~~~~~~~
162      |                     |
163      |                     boxed_value<[...]>
164   15 |    + width(s2) * height(s2));
165      |    ^ ~~~~~~~~~~~~~~~~~~~~~~
166      |                |
167      |                boxed_value<[...]>
168
169       These labels can be disabled via [12]-fno-diagnostics-show-labels.
170     * A new option [13]-fdiagnostics-format=json has been introduced for
171       emitting diagnostics in a machine-readable format.
172     * The alignment-related options [14]-falign-functions,
173       [15]-falign-labels, [16]-falign-loops, and [17]-falign-jumps
174       received support for a secondary alignment (e.g.
175       -falign-loops=n:m:n2:m2).
176     * New pair of profiling options ([18]-fprofile-filter-files and
177       [19]-fprofile-exclude-files) has been added. The options help to
178       filter which source files are instrumented.
179     * AddressSanitizer generates more compact redzones for automatic
180       variables. That helps to reduce memory footprint of a sanitized
181       binary.
182     * Numerous improvements have been made to the output of
183       [20]-fopt-info.
184       Messages are now prefixed with optimized, missed, or note, rather
185       than the old behavior of all being prefixed with note.
186       The output from -fopt-info can now contain information on inlining
187       decisions:
188$ g++ -c inline.cc -O2 -fopt-info-inline-all
189inline.cc:24:11: note: Considering inline candidate void foreach(T, T, void (*)(
190E)) [with T = char**; E = char*]/2.
191inline.cc:24:11: optimized:  Inlining void foreach(T, T, void (*)(E)) [with T =
192char**; E = char*]/2 into int main(int, char**)/1.
193inline.cc:19:12: missed:   not inlinable: void inline_me(char*)/0 -> int std::pu
194ts(const char*)/3, function body not available
195inline.cc:13:8: optimized:  Inlined void inline_me(char*)/4 into int main(int, c
196har**)/1 which now has time 127.363637 and size 11, net change of +0.
197Unit growth for small function inlining: 16->16 (0%)
198
199Inlined 2 calls, eliminated 1 functions
200
201
202       The output from the vectorizer has been rationalized so that failed
203       attempts to vectorize a loop are displayed in the form
204    [LOOP-LOCATION]: couldn't vectorize this loop
205    [PROBLEM-LOCATION]: because of [REASON]
206
207       rather than an exhaustive log of all decisions made by the
208       vectorizer. For example:
209$ gcc -c v.c -O3 -fopt-info-all-vec
210v.c:7:3: missed: couldn't vectorize loop
211v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" :  :  : "me
212mory");
213v.c:3:6: note: vectorized 0 loops in function.
214v.c:10:7: missed: statement clobbers memory: __asm__ __volatile__("" :  :  : "me
215mory");
216
217       The old behavior can be obtained via a new -internals suboption of
218       -fopt-info.
219     * A new option, [21]-fsave-optimization-record has been added, which
220       writes a SRCFILE.opt-record.json.gz file describing the
221       optimization decisions made by GCC. This is similar to the output
222       of -fopt-info, but with additional metadata such as the inlining
223       chain, and profile information (if available).
224     * Inter-procedural propagation of stack alignment can now be
225       controlled by [22]-fipa-stack-alignment.
226     * Propagation of addressability, readonly and writeonly flags on
227       static variables can now be controlled by
228       [23]-fipa-reference-addressable.
229
230   The following built-in functions have been introduced.
231     * [24]__builtin_expect_with_probability to provide branch prediction
232       probability hints to the optimizer.
233     * [25]__builtin_has_attribute determines whether a function, type, or
234       variable has been declared with some attribute.
235     * [26]__builtin_speculation_safe_value can be used to help mitigate
236       against unsafe speculative execution.
237
238   The following attributes have been introduced.
239     * The [27]copy function attribute has been added. The attribute can
240       also be applied to type definitions and to variable declarations.
241
242   A large number of improvements to code generation have been made,
243   including but not limited to the following.
244     * Switch expansion has been improved by using a different strategy
245       (jump table, bit test, decision tree) for a subset of switch cases.
246     * A linear function expression defined as a switch statement can be
247       transformed by [28]-ftree-switch-conversion. For example:
248
249int
250foo (int how)
251{
252  switch (how) {
253    case 2: how = 205; break;
254    case 3: how = 305; break;
255    case 4: how = 405; break;
256    case 5: how = 505; break;
257    case 6: how = 605; break;
258  }
259  return how;
260}
261
262       can be transformed into 100 * how + 5 (for values defined in the
263       switch statement).
264     * Inter-procedural optimization improvements:
265          + Inliner defaults was tuned to better suits modern C++
266            codebases especially when built with link time optimizations.
267            New parameters max-inline-insns-small, max-inline-insns-size,
268            uninlined-function-insns, uninlined-function-time,
269            uninlined-thunk-insns, and uninlined-thunk-time were added.
270          + Hot/cold partitioning is now more precise and aggressive.
271          + Improved scalability for very large translation units
272            (especially when link-time optimizing large programs).
273     * Profile driven optimization improvements:
274          + [29]-fprofile-use now enables [30]-fversion-loops-for-strides,
275            [31]-floop-interchange, [32]-floop-unroll-and-jam,
276            [33]-ftree-loop-distribution.
277          + Streaming of counter histograms was removed. This reduces the
278            size of profile files. Histogram is computed on the fly with
279            link-time optimization. Parameter hot-bb-count-ws-permille was
280            reduced from 999 to 990 to account for more precise
281            histograms.
282     * Link-time optimization improvements:
283          + Types are now simplified prior streaming resulting in
284            significant reductions of the LTO object files, link-time
285            memory use, and improvements of link-time parallelism.
286          + Default number of partitions (--param lto-partitions) was
287            increased from 32 to 128 enabling effective use of CPUs with
288            more than 32 hyperthreads. --param
289            lto-max-streaming-parallelism can now be used to control
290            number of streaming processes.
291          + Warnings on C++ One Decl Rule violations (-Wodr) are now more
292            informative and produce fewer redundant results.
293       Overall compile time of Firefox 66 and LibreOffice 6.2.3 on an
294       8-core machine was reduced by about 5% compared to GCC 8.3, and the
295       size of LTO object files by 7%. LTO link-time improves by 11% on an
296       8-core machine and scales significantly better for more parallel
297       build environments. The serial stage of the link-time optimization
298       is 28% faster consuming 20% less memory. The parallel stage now
299       scales to up to 128 partitions rather than 32 and reduces memory
300       use for every worker by 30%.
301
302   The following improvements to the gcov command-line utility have been
303   made.
304     * The gcov tool received a new option [34]--use-hotness-colors (-q)
305       that can provide perf-like coloring of hot functions.
306     * The gcov tool has changed its intermediate format to a new JSON
307       format.
308
309New Languages and Language specific improvements
310
311   [35]OpenACC support in C, C++, and Fortran continues to be maintained
312   and improved. Most of the OpenACC 2.5 specification is implemented. See
313   the [36]implementation status section on the OpenACC wiki page for
314   further information.
315
316  C family
317
318     * Version 5.0 of the [37]OpenMP specification is now partially
319       supported in the C and C++ compilers. For details which features of
320       OpenMP 5.0 are and which are not supported in the GCC 9 release see
321       [38]this mail.
322     * New extensions:
323          + [39]__builtin_convertvector built-in for vector conversions
324            has been added.
325     * New warnings:
326          + [40]-Waddress-of-packed-member, enabled by default, warns
327            about an unaligned pointer value from the address of a packed
328            member of a struct or union.
329     * Enhancements to existing warnings:
330          + [41]-Warray-bounds detects more instances of out-of-bounds
331            indices.
332          + [42]-Wattribute-alias also detects attribute mismatches
333            between alias declarations and their targets, in addition to
334            mismatches between their types.
335          + [43]-Wformat-overflow and [44]-Wformat-truncation have been
336            extended to all formatted input/output functions (where
337            applicable) and enhanced to detect a subset of instances of
338            reading past the end of unterminated constant character arrays
339            in %s directives.
340          + [45]-Wmissing-attributes detects instances of missing function
341            attributes on declarations of aliases and weak references.
342          + [46]-Wstringop-truncation also detects a subset of instances
343            of reading past the end of unterminated constant character
344            arrays,
345     * If a macro is used with the wrong argument count, the C and C++
346       front ends now show the definition of that macro via a note.
347     * The spelling corrector now considers transposed letters, and the
348       threshold for similarity has been tightened, to avoid nonsensical
349       suggestions.
350
351  C
352
353     * There is now experimental support for -std=c2x, to select support
354       for the upcoming C2X revision of the ISO C standard. This standard
355       is in the early stages of development and the only feature
356       supported in GCC 9 is _Static_assert with a single argument
357       (support for _Static_assert with two arguments was added in C11 and
358       GCC 4.6). There are also new options -std=gnu2x, for C2X with GNU
359       extensions, and -Wc11-c2x-compat, to warn for uses of features
360       added in C2X (such warnings are also enabled by use of -Wpedantic
361       if not using -std=c2x or -std=gnu2x).
362     * New warnings:
363          + [47]-Wabsolute-value warns for calls to standard functions
364            that compute the absolute value of an argument when a more
365            appropriate standard function is available. For example,
366            calling abs(3.14) triggers the warning because the appropriate
367            function to call to compute the absolute value of a double
368            argument is fabs. The option also triggers warnings when the
369            argument in a call to such a function has an unsigned type.
370            This warning can be suppressed with an explicit type cast and
371            it is also enabled by -Wextra.
372
373  C++
374
375     * New warnings:
376          + [48]-Wdeprecated-copy, implied by -Wextra, warns about the
377            C++11 deprecation of implicitly declared copy constructor and
378            assignment operator if one of them is user-provided.
379            -Wdeprecated-copy-dtor also warns if the destructor is
380            user-provided, as specified in C++11.
381          + [49]-Winit-list-lifetime, on by default, warns about uses of
382            std::initializer_list that are likely to result in a dangling
383            pointer, such as returning or assigning from a temporary list.
384          + [50]-Wredundant-move, implied by -Wextra, warns about
385            redundant calls to std::move.
386          + [51]-Wpessimizing-move, implied by -Wall, warns when a call to
387            std::move prevents copy elision.
388          + [52]-Wclass-conversion, on by default, warns when a conversion
389            function will never be called due to the type it converts to.
390     * The C++ front end has experimental support for some of the upcoming
391       C++2a draft features with the -std=c++2a or -std=gnu++2a flags,
392       including range-based for statements with initializer, default
393       constructible and assignable stateless lambdas, lambdas in
394       unevaluated contexts, language support for empty data members,
395       allowing pack expansion in lambda init-capture, likely and unlikely
396       attributes, class types in non-type template parameters, allowing
397       virtual function calls in constant expressions, explicit(bool),
398       std::is_constant_evaluated, nested inline namespaces, etc. For a
399       full list of new features, see [53]the C++ status page.
400     * The C++ front end now preserves source locations for literals,
401       id-expression, and mem-initializer for longer. For example it is
402       now able to pin-point the pertinent locations for bad
403       initializations such as these
404$ g++ -c bad-inits.cc
405bad-inits.cc:10:14: error: cannot convert 'json' to 'int' in initialization
406   10 |   { 3, json::object },
407      |        ~~~~~~^~~~~~
408      |              |
409      |              json
410bad-inits.cc:14:31: error: initializer-string for array of chars is too long [-f
411permissive]
412   14 | char buffers[3][5] = { "red", "green", "blue" };
413      |                               ^~~~~~~
414bad-inits.cc: In constructor 'X::X()':
415bad-inits.cc:17:13: error: invalid conversion from 'int' to 'void*' [-fpermissiv
416e]
417   17 |   X() : one(42), two(42), three(42)
418      |             ^~
419      |             |
420      |             int
421
422       rather than emitting the error at the final closing parenthesis or
423       brace.
424     * Error-reporting of overload resolution has been special-cased to
425       make the case of a single failed candidate easier to read. For
426       example:
427$ g++ param-type-mismatch.cc
428param-type-mismatch.cc: In function 'int test(int, const char*, float)':
429param-type-mismatch.cc:8:32: error: cannot convert 'const char*' to 'const char*
430*'
431    8 |   return foo::member_1 (first, second, third);
432      |                                ^~~~~~
433      |                                |
434      |                                const char*
435param-type-mismatch.cc:3:46: note:   initializing argument 2 of 'static int foo:
436:member_1(int, const char**, float)'
437    3 |   static int member_1 (int one, const char **two, float three);
438      |                                 ~~~~~~~~~~~~~^~~
439
440       highlights both the problematic argument, and the parameter that it
441       can't be converted to.
442     * Diagnostics involving binary operators now use color to distinguish
443       the two operands, and label them separately (as per the example of
444       source labelling above).
445     * Diagnostics involving function calls now highlight the pertinent
446       parameter of the declaration in more places.
447$ g++ bad-conversion.cc
448bad-conversion.cc: In function 'void caller()':
449bad-conversion.cc:9:14: error: cannot convert 'bool' to 'void*'
450    9 |   callee (0, false, 2);
451      |              ^~~~~
452      |              |
453      |              bool
454bad-conversion.cc:3:19: note:   initializing argument 2 of 'void callee(int, voi
455d*, int)'
456    3 | void callee (int, void *, int)
457      |                   ^~~~~~
458
459     * The C++ front end's implementation of [54]-Wformat now shows
460       precise locations within string literals, and underlines the
461       pertinent arguments at bogus call sites (the C front end has been
462       doing this since GCC 7). For example:
463$ g++ -c bad-printf.cc -Wall
464bad-printf.cc: In function 'void print_field(const char*, float, long int, long
465int)':
466bad-printf.cc:6:17: warning: field width specifier '*' expects argument of type
467'int', but argument 3 has type 'long int' [-Wformat=]
468    6 |   printf ("%s: %*ld ", fieldname, column - width, value);
469      |                ~^~~               ~~~~~~~~~~~~~~
470      |                 |                        |
471      |                 int                      long int
472bad-printf.cc:6:19: warning: format '%ld' expects argument of type 'long int', b
473ut argument 4 has type 'double' [-Wformat=]
474    6 |   printf ("%s: %*ld ", fieldname, column - width, value);
475      |                ~~~^                               ~~~~~
476      |                   |                               |
477      |                   long int                        double
478      |                %*f
479
480     * The C++ front end has gained new fix-it hints for forgetting the
481       return *this; needed by various C++ operators:
482$ g++ -c operator.cc
483operator.cc: In member function 'boxed_ptr& boxed_ptr::operator=(const boxed_ptr
484&)':
485operator.cc:7:3: warning: no return statement in function returning non-void [-W
486return-type]
487    6 |     m_ptr = other.m_ptr;
488  +++ |+    return *this;
489    7 |   }
490      |   ^
491
492       for when the compiler needs a typename:
493$ g++ -c template.cc
494template.cc:3:3: error: need 'typename' before 'Traits::type' because 'Traits' i
495s a dependent scope
496    3 |   Traits::type type;
497      |   ^~~~~~
498      |   typename
499
500       when trying to use an accessor member as if it were a data member:
501$ g++ -c fncall.cc
502fncall.cc: In function 'void hangman(const mystring&)':
503fncall.cc:12:11: error: invalid use of member function 'int mystring::get_length
504() const' (did you forget the '()' ?)
505   12 |   if (str.get_length > 0)
506      |       ~~~~^~~~~~~~~~
507      |                     ()
508
509       for C++11's scoped enums:
510$ g++ -c enums.cc
511enums.cc: In function 'void json::test(const json::value&)':
512enums.cc:12:26: error: 'STRING' was not declared in this scope; did you mean 'js
513on::kind::STRING'?
514   12 |     if (v.get_kind () == STRING)
515      |                          ^~~~~~
516      |                          json::kind::STRING
517enums.cc:3:44: note: 'json::kind::STRING' declared here
518    3 |   enum class kind { OBJECT, ARRAY, NUMBER, STRING, TRUE, FALSE, NULL_ };
519      |                                            ^~~~~~
520
521       and a tweak to integrate the suggestions about misspelled members
522       with that for accessors:
523$ g++ -c accessor-fixit.cc
524accessor-fixit.cc: In function 'int test(t*)':
525accessor-fixit.cc:17:15: error: 'class t' has no member named 'ratio'; did you m
526ean 'int t::m_ratio'? (accessible via 'int t::get_ratio() const')
527   17 |   return ptr->ratio;
528      |               ^~~~~
529      |               get_ratio()
530
531       In addition, various diagnostics in the C++ front-end have been
532       streamlined by consolidating the suggestion into the initial error,
533       rather than emitting a follow-up note:
534$ g++ typo.cc
535typo.cc:5:13: error: 'BUFSIZE' was not declared in this scope; did you mean 'BUF
536_SIZE'?
537    5 | uint8_t buf[BUFSIZE];
538      |             ^~~~~~~
539      |             BUF_SIZE
540
541    Runtime Library (libstdc++)
542
543     * Improved support for C++17, including:
544          + The C++17 implementation is no longer experimental.
545          + Parallel algorithms and <execution> (requires [55]Thread
546            Building Blocks 2018 or newer).
547          + <memory_resource>.
548          + Using the types and functions in <filesystem> does not require
549            linking with -lstdc++fs now.
550     * Improved experimental support for C++2a, including:
551          + Type traits std::remove_cvref, std::unwrap_reference,
552            std::unwrap_decay_ref, std::is_nothrow_convertible, and
553            std::type_identity.
554          + Headers <bit> and <version>.
555          + Uniform container erasure (std::erase_if).
556          + contains member of maps and sets.
557          + String prefix and suffix checking (starts_with, ends_with).
558          + Functions std::midpoint and std::lerp for interpolation.
559          + std::bind_front.
560          + std::visit<R>.
561          + std::assume_aligned.
562          + Uses-allocator construction utilities.
563          + std::pmr::polymorphic_allocator<std::byte>.
564          + Library support for char8_t type.
565          + Destroying delete.
566          + std::is_constant_evaluated() function.
567     * Support for opening file streams with wide character paths on
568       Windows
569     * Incomplete support for the C++17 Filesystem library and the
570       Filesystem TS on Windows.
571     * Incomplete, experimental support for the Networking TS.
572
573  D
574
575     * Support for the D programming language has been added to GCC,
576       implementing version 2.076 of the language and run-time library.
577
578  Fortran
579
580     * Asynchronous I/O is now fully supported. The program needs to be
581       linked against the pthreads library to use it, otherwise the I/O is
582       done synchronously. For systems which do not support POSIX
583       condition variables, such as AIX, all I/O is still done
584       synchronously.
585     * The BACK argument for MINLOC and MAXLOC has been implemented.
586     * The FINDLOC intrinsic function has been implemented.
587     * The IS_CONTIGUOUS intrinsic function has been implemented.
588     * Direct access to the real and imaginary parts of a complex variable
589       via c%re and c%im has been implemented.
590     * Type parameter inquiry via str%len and a%kind has been implemented.
591     * C descriptors and the ISO_Fortran_binding.h source file have been
592       implemented.
593     * The MAX and MIN intrinsics are no longer guaranteed to return any
594       particular value in case one of the arguments is a NaN. Note that
595       this conforms to the Fortran standard and to what other Fortran
596       compilers do. If there is a need to handle that case in some
597       specific way, one needs to explicitly check for NaN's before
598       calling MAX or MIN, e.g. by using the IEEE_IS_NAN function from the
599       intrinsic module IEEE_ARITHMETIC.
600     * A new command-line option [56]-fdec-include, set also by the
601       [57]-fdec option, has been added to increase compatibility with
602       legacy code. With this option, an INCLUDE directive is also parsed
603       as a statement, which allows the directive to be spread across
604       multiple source lines with line continuations.
605     * A new [58]BUILTIN directive, has been added. The purpose of the
606       directive is to provide an API between the GCC compiler and the GNU
607       C Library which would define vector implementations of math
608       routines.
609
610  Go
611
612     * GCC 9 provides a complete implementation of the Go 1.12.2 user
613       packages.
614
615libgccjit
616
617     * The libgccjit API gained a new entry point:
618       [59]gcc_jit_context_add_driver_option.
619
620New Targets and Target Specific Improvements
621
622  AArch64 & Arm
623
624     * Support has been added for the following processors (GCC
625       identifiers in parentheses):
626          + Arm Cortex-A76 (cortex-a76).
627          + Arm Cortex-A55/Cortex-A76 DynamIQ big.LITTLE
628            (cortex-a76.cortex-a55).
629          + Arm Neoverse N1 (neoverse-n1).
630       The GCC identifiers can be used as arguments to the -mcpu or -mtune
631       options, for example: -mcpu=cortex-a76 or
632       -mtune=cortex-a76.cortex-a55 or as arguments to the equivalent
633       target attributes and pragmas.
634     * The Armv8.3-A complex number instructions are now supported via
635       intrinsics when the option -march=armv8.3-a or equivalent is
636       specified. For the half-precision floating-point variants of these
637       instructions use the architecture extension flag +fp16, e.g.
638       -march=armv8.3-a+fp16.
639       The intrinsics are defined by the ACLE specification.
640     * The Armv8.5-A architecture is now supported through the
641       -march=armv8.5-a option.
642     * The Armv8.5-A architecture also adds some security features that
643       are optional to all older architecture versions. These are now
644       supported and only affect the assembler.
645          + Speculation Barrier instruction through the -march=armv8-a+sb
646            option.
647          + Execution and Data Prediction Restriction instructions through
648            the -march=armv8-a+predres option.
649          + Speculative Store Bypass Safe instruction through the
650            -march=armv8-a+ssbs option. This does not require a compiler
651            option for Arm and thus -march=armv8-a+ssbs is an
652            AArch64-specific option.
653
654      AArch64 specific
655
656     * Support has been added for the Arm Neoverse E1 processor
657       (-mcpu=neoverse-e1).
658     * The AArch64 port now has support for stack clash protection using
659       the [60]-fstack-clash-protection option. The probing interval/guard
660       size can be set by using --param
661       stack-clash-protection-guard-size=12|16. The value of this
662       parameter must be in bytes represented as a power of two. The two
663       supported values for this parameter are 12 (for a 4KiB size, 2^12)
664       and 16 (for a 64KiB size, 2^16). The default value is 16 (64Kb) and
665       can be changed at configure time using the flag
666       --with-stack-clash-protection-guard-size=12|16.
667     * The option -msign-return-address= has been deprecated. This has
668       been replaced by the new -mbranch-protection= option. This new
669       option can now be used to enable the return address signing as well
670       as the new Branch Target Identification feature of Armv8.5-A
671       architecture. For more information on the arguments accepted by
672       this option, please refer to [61]AArch64-Options.
673     * The following optional extensions to Armv8.5-A architecture are now
674       supported and only affect the assembler.
675          + Random Number Generation instructions through the
676            -march=armv8.5-a+rng option.
677          + Memory Tagging Extension through the -march=armv8.5-a+memtag
678            option.
679
680      Arm specific
681
682     * Support for the deprecated Armv2 and Armv3 architectures and their
683       variants has been removed. Their corresponding -march values and
684       the -mcpu options that used these architectures have been removed.
685     * Support for the Armv5 and Armv5E architectures (which have no known
686       implementations) has been removed. Note that Armv5T, Armv5TE and
687       Armv5TEJ architectures remain supported.
688     * Corrected FPU configurations for Cortex-R7 and Cortex-R8 when using
689       their respective -mcpu options.
690
691  AMD GCN
692
693     * A new back end targeting AMD GCN GPUs has been contributed to GCC.
694       The implementation is currently limited to compiling
695       single-threaded, stand-alone programs. Future versions will add
696       support for offloading multi-threaded kernels via OpenMP and
697       OpenACC. The following devices are supported (GCC identifiers in
698       parentheses):
699          + Fiji (fiji).
700          + Vega 10 (gfx900).
701
702  ARC
703
704     * LRA is now on by default for the ARC target. This can be controlled
705       by -mlra.
706     * Add support for frame code-density and branch-and-index
707       instructions.
708
709  C-SKY
710
711     * A new back end targeting C-SKY V2 processors has been contributed
712       to GCC.
713
714  IA-32/x86-64
715
716     * Support of Intel MPX (Memory Protection Extensions) has been
717       removed.
718     * New ISA extension support for Intel PTWRITE was added to GCC.
719       PTWRITE intrinsics are available via the -mptwrite compiler switch.
720     * GCC now supports the Intel CPU named Cascade Lake with AVX512
721       extensions through -march=cascadelake. The switch enables the
722       following ISA extensions: AVX512F, AVX512VL, AVX512CD, AVX512BW,
723       AVX512DQ, AVX512VNNI.
724
725  OpenRISC
726
727     * A new back end targeting OpenRISC processors has been contributed
728       to GCC.
729
730  S/390, System z, IBM z Systems
731
732     * Support for the arch13 architecture has been added. When using the
733       -march=arch13 option, the compiler will generate code making use of
734       the new instructions introduced with the vector enhancement
735       facility 2 and the miscellaneous instruction extension facility 2.
736       The -mtune=arch13 option enables arch13 specific instruction
737       scheduling without making use of new instructions.
738     * Builtins for the new vector instructions have been added and can be
739       enabled using the -mzvector option.
740     * Support for ESA architecture machines g5 and g6 is deprecated since
741       GCC 6.1.0 and has been removed now.
742     * When compiling with -march=z14 or higher GCC emits alignments hints
743       on the vector load/store instructions (8 or 16 byte).
744     * Functions now have a default alignment of 16 bytes. This helps with
745       branch prediction effects.
746     * -mfentry is now supported. As well as the mcount mechanism the
747       __fentry__ is called before the function prologue. However, since
748       just a single instruction is required to call __fentry__ the call
749       sequence imposes a smaller overhead than mcount (4 instructions).
750       The produced code is compatible only with newer glibc versions,
751       which provide the __fentry__ symbol and do not clobber r0 when
752       resolving lazily bound functions. -mfentry is only supported when
753       generating 64 bit code and does not work with nested C functions.
754     * The -mnop-mcount option can be used to emit NOP instructions
755       instead of an mcount or fentry call stub.
756     * With the -mrecord-mcount option a __mcount_loc section is generated
757       containing pointers to each profiling call stub. This is useful for
758       automatically patching in and out calls.
759
760Operating Systems
761
762  Solaris
763
764     * g++ now unconditionally enables large file support when compiling
765       32-bit code.
766     * Support for the AddressSanitizer and UndefinedBehaviorSanitizer has
767       been merged from LLVM. For the moment, this only works for 32-bit
768       code on both SPARC and x86.
769     * An initial port of the D runtime library has been completed on
770       Solaris 11/x86. It requires the use of GNU as. Solaris 11/SPARC
771       support is still work-in-progress.
772
773  Windows
774
775     * A C++ Microsoft ABI bitfield layout bug, [62]PR87137 has been
776       fixed. A non-field declaration could cause the current bitfield
777       allocation unit to be completed, incorrectly placing a following
778       bitfield into a new allocation unit. The Microsoft ABI is selected
779       for:
780          + Mingw targets
781          + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields
782            option is specified, or __attribute__((ms_struct)) is used
783          + SuperH targets when the -mhitachi option is specified, or
784            __attribute__((renesas)) is used
785
786Improvements for plugin authors
787
788     * GCC's diagnostic subsystem now has a way to logically group
789       together related diagnostics, auto_diagnostic_group. Such
790       diagnostics will be nested by the output of
791       [63]-fdiagnostics-format=json.
792     * GCC now has a set of [64]user experience guidelines for GCC, with
793       information and advice on implementing new diagnostics.
794
795Other significant improvements
796
797     * GCC's internal "selftest" suite now runs for C++ as well as C (in
798       debug builds of the compiler).
799
800GCC 9.1
801
802   This is the [65]list of problem reports (PRs) from GCC's bug tracking
803   system that are known to be fixed in the 9.1 release. This list might
804   not be complete (that is, it is possible that some PRs that have been
805   fixed are not listed here).
806
807GCC 9.2
808
809   This is the [66]list of problem reports (PRs) from GCC's bug tracking
810   system that are known to be fixed in the 9.2 release. This list might
811   not be complete (that is, it is possible that some PRs that have been
812   fixed are not listed here).
813
814GCC 9.3
815
816   This is the [67]list of problem reports (PRs) from GCC's bug tracking
817   system that are known to be fixed in the 9.3 release. This list might
818   not be complete (that is, it is possible that some PRs that have been
819   fixed are not listed here).
820
821
822    For questions related to the use of GCC, please consult these web
823    pages and the [68]GCC manuals. If that fails, the
824    [69]gcc-help@gcc.gnu.org mailing list might help. Comments on these
825    web pages and the development of GCC are welcome on our developer
826    list at [70]gcc@gcc.gnu.org. All of [71]our lists have public
827    archives.
828
829   Copyright (C) [72]Free Software Foundation, Inc. Verbatim copying and
830   distribution of this entire article is permitted in any medium,
831   provided this notice is preserved.
832
833   These pages are [73]maintained by the GCC team. Last modified
834   2020-03-12[74].
835
836References
837
838   1. http://gcc.gnu.org/gcc-9/porting_to.html
839   2. http://gcc.gnu.org/onlinedocs/index.html#current
840   3. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=88469
841   4. https://gcc.gnu.org/ml/gcc/2018-10/msg00139.html
842   5. https://gcc.gnu.org/ml/gcc/2019-04/msg00023.html
843   6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90920
844   7. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-frepo
845   8. https://gcc.gnu.org/PR90361
846   9. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-GCC.html#Invoking-GCC
847  10. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-flive-patching
848  11. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-line-numbers
849  12. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fno-diagnostics-show-labels
850  13. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format
851  14. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-functions
852  15. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-labels
853  16. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-loops
854  17. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-falign-jumps
855  18. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-filter-files
856  19. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fprofile-exclude-files
857  20. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fopt-info
858  21. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Developer-Options.html#index-fsave-optimization-record
859  22. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-stack-alignment
860  23. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fipa-reference-addressable
861  24. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fexpect_005fwith_005fprobability
862  25. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fhas_005fattribute-1
863  26. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Other-Builtins.html#index-_005f_005fbuiltin_005fspeculation_005fsafe_005fvalue-1
864  27. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Common-Function-Attributes.html#index-copy-function-attribute
865  28. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-switch-conversion
866  29. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fprofile-use
867  30. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-fversion-loops-for-strides
868  31. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-interchange
869  32. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-floop-unroll-and-jam
870  33. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Optimize-Options.html#index-ftree-loop-distribution
871  34. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Invoking-Gcov.html#Invoking-Gcov
872  35. https://www.openacc.org/
873  36. https://gcc.gnu.org/wiki/OpenACC#status-9
874  37. https://www.openmp.org/specifications/
875  38. https://gcc.gnu.org/ml/gcc-patches/2018-11/msg00628.html
876  39. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Vector-Extensions.html#index-_005f_005fbuiltin_005fconvertvector
877  40. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Waddress-of-packed-member
878  41. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Warray-bounds
879  42. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wattribute-alias
880  43. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-overflow
881  44. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat-truncation
882  45. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wmissing-attributes
883  46. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wstringop-truncation
884  47. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wabsolute-value
885  48. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wdeprecated-copy
886  49. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Winit-list-lifetime
887  50. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wredundant-move
888  51. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wpessimizing-move
889  52. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-conversion
890  53. http://gcc.gnu.org/projects/cxx-status.html#cxx2a
891  54. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Warning-Options.html#index-Wformat
892  55. https://github.com/intel/tbb
893  56. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec-include
894  57. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/Fortran-Dialect-Options.html#index-fdec
895  58. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gfortran/BUILTIN-directive.html#BUILTIN-directive
896  59. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/jit/topics/contexts.html#gcc_jit_context_add_driver_option
897  60. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Instrumentation-Options.html#index-fstack-protector
898  61. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/AArch64-Options.html#AArch64-Options
899  62. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137
900  63. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gcc/Diagnostic-Message-Formatting-Options.html#index-fdiagnostics-format
901  64. https://gcc.gnu.org/onlinedocs/gcc-9.1.0/gccint/User-Experience-Guidelines.html
902  65. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.0
903  66. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.2
904  67. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=9.3
905  68. https://gcc.gnu.org/onlinedocs/
906  69. mailto:gcc-help@gcc.gnu.org
907  70. mailto:gcc@gcc.gnu.org
908  71. https://gcc.gnu.org/lists.html
909  72. https://www.fsf.org/
910  73. https://gcc.gnu.org/about.html
911  74. http://validator.w3.org/check/referer
912======================================================================
913http://gcc.gnu.org/gcc-8/index.html
914                              GCC 8 Release Series
915
916   Mar 4, 2020
917
918   The [1]GNU project and the GCC developers are pleased to announce the
919   release of GCC 8.4.
920
921   This release is a bug-fix release, containing fixes for regressions in
922   GCC 8.3 relative to previous releases of GCC.
923
924Release History
925
926   GCC 8.4
927          Mar 4, 2020 ([2]changes, [3]documentation)
928
929   GCC 8.3
930          Feb 22, 2019 ([4]changes, [5]documentation)
931
932   GCC 8.2
933          Jul 26, 2018 ([6]changes, [7]documentation)
934
935   GCC 8.1
936          May 2, 2018 ([8]changes, [9]documentation)
937
938References and Acknowledgements
939
940   GCC used to stand for the GNU C Compiler, but since the compiler
941   supports several other languages aside from C, it now stands for the
942   GNU Compiler Collection.
943
944   A list of [10]successful builds is updated as new information becomes
945   available.
946
947   The GCC developers would like to thank the numerous people that have
948   contributed new features, improvements, bug fixes, and other changes as
949   well as test results to GCC. This [11]amazing group of volunteers is
950   what makes GCC successful.
951
952   For additional information about GCC please refer to the [12]GCC
953   project web site or contact the [13]GCC development mailing list.
954
955   To obtain GCC please use [14]our mirror sites or [15]our version
956   control system.
957
958
959    For questions related to the use of GCC, please consult these web
960    pages and the [16]GCC manuals. If that fails, the
961    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
962    web pages and the development of GCC are welcome on our developer
963    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
964    archives.
965
966   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
967   distribution of this entire article is permitted in any medium,
968   provided this notice is preserved.
969
970   These pages are [21]maintained by the GCC team. Last modified
971   2020-03-04[22].
972
973References
974
975   1. http://www.gnu.org/
976   2. http://gcc.gnu.org/gcc-8/changes.html
977   3. http://gcc.gnu.org/onlinedocs/8.4.0/
978   4. http://gcc.gnu.org/gcc-8/changes.html
979   5. http://gcc.gnu.org/onlinedocs/8.3.0/
980   6. http://gcc.gnu.org/gcc-8/changes.html
981   7. http://gcc.gnu.org/onlinedocs/8.2.0/
982   8. http://gcc.gnu.org/gcc-8/changes.html
983   9. http://gcc.gnu.org/onlinedocs/8.1.0/
984  10. http://gcc.gnu.org/gcc-8/buildstat.html
985  11. http://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Contributors.html
986  12. http://gcc.gnu.org/index.html
987  13. mailto:gcc@gcc.gnu.org
988  14. http://gcc.gnu.org/mirrors.html
989  15. http://gcc.gnu.org/git.html
990  16. https://gcc.gnu.org/onlinedocs/
991  17. mailto:gcc-help@gcc.gnu.org
992  18. mailto:gcc@gcc.gnu.org
993  19. https://gcc.gnu.org/lists.html
994  20. https://www.fsf.org/
995  21. https://gcc.gnu.org/about.html
996  22. http://validator.w3.org/check/referer
997======================================================================
998http://gcc.gnu.org/gcc-8/changes.html
999                              GCC 8 Release Series
1000                        Changes, New Features, and Fixes
1001
1002   This page is a "brief" summary of some of the huge number of
1003   improvements in GCC 8. You may also want to check out our [1]Porting to
1004   GCC 8 page and the [2]full GCC documentation.
1005
1006Caveats
1007
1008     * Support for the obsolete SDB/coff debug info format has been
1009       removed. The option -gcoff no longer does anything.
1010     * The Cilk+ extensions to the C and C++ languages have been removed.
1011     * The MPX extensions to the C and C++ languages have been deprecated
1012       and will be removed in a future release.
1013     * The extension allowing arithmetic on std::atomic<void*> and types
1014       like std::atomic<R(*)()> has been deprecated.
1015     * The non-standard C++0x std::copy_exception function was removed.
1016       std::make_exception_ptr should be used instead.
1017     * Support for the powerpc*-*-*spe* target ports which have been
1018       recently unmaintained and untested in GCC has been declared
1019       obsolete in GCC 8 as [3]announced. Unless there is activity to
1020       revive them, the next release of GCC will have their sources
1021       permanently removed.
1022
1023General Improvements
1024
1025     * Inter-procedural optimization improvements:
1026          + Reworked run-time estimation metrics leading to more realistic
1027            guesses driving inliner and cloning heuristics.
1028          + The ipa-pure-const pass is extended to propagate the malloc
1029            attribute, and the corresponding warning option
1030            -Wsuggest-attribute=malloc emits a diagnostic for functions
1031            which can be annotated with the malloc attribute.
1032     * Profile driven optimization improvements:
1033          + New infrastructure for representing profiles (both statically
1034            guessed and profile feedback) which allows propagation of
1035            additional information about the reliability of the profile.
1036          + A number of improvements in the profile updating code solving
1037            problems found by new verification code.
1038          + Static detection of code which is not executed in a valid run
1039            of the program. This includes paths which trigger undefined
1040            behavior as well as calls to functions declared with the cold
1041            attribute. Newly the noreturn attribute does not imply all
1042            effects of cold to differentiate between exit (which is
1043            noreturn) and abort (which is in addition not executed in
1044            valid runs).
1045          + -freorder-blocks-and-partition, a pass splitting function
1046            bodies into hot and cold regions, is now enabled by default at
1047            -O2 and higher for x86 and x86-64.
1048     * Link-time optimization improvements:
1049          + We have significantly improved debug information on ELF
1050            targets using DWARF by properly preserving language-specific
1051            information. This allows for example the libstdc++
1052            pretty-printers to work with LTO optimized executables.
1053     * A new option -fcf-protection=[full|branch|return|none] is
1054       introduced to perform code instrumentation to increase program
1055       security by checking that target addresses of control-flow transfer
1056       instructions (such as indirect function call, function return,
1057       indirect jump) are valid. Currently the instrumentation is
1058       supported on x86 GNU/Linux targets only. See the user guide for
1059       further information about the option syntax and section "New
1060       Targets and Target Specific Improvements" for IA-32/x86-64 for more
1061       details.
1062     * The -gcolumn-info option is now enabled by default. It includes
1063       column information in addition to just filenames and line numbers
1064       in DWARF debugging information.
1065     * The polyhedral-based loop nest optimization pass
1066       -floop-nest-optimize has been overhauled. It's still considered
1067       experimental and may not result in any runtime improvements.
1068     * Two new classical loop nest optimization passes have been added.
1069       -floop-unroll-and-jam performs outer loop unrolling and fusing of
1070       the inner loop copies. -floop-interchange exchanges loops in a loop
1071       nest to improve data locality. Both passes are enabled by default
1072       at -O3 and above.
1073     * The classic loop nest optimization pass -ftree-loop-distribution
1074       has been improved and enabled by default at -O3 and above. It
1075       supports loop nest distribution in some restricted scenarios; it
1076       also supports cancellable innermost loop distribution with loop
1077       versioning under run-time alias checks.
1078     * The new option -fstack-clash-protection causes the compiler to
1079       insert probes whenever stack space is allocated statically or
1080       dynamically to reliably detect stack overflows and thus mitigate
1081       the attack vector that relies on jumping over a stack guard page as
1082       provided by the operating system.
1083     * A new pragma GCC unroll has been implemented in the C family of
1084       languages, as well as Fortran and Ada, so as to make it possible
1085       for the user to have a finer-grained control over the loop
1086       unrolling optimization.
1087     * GCC has been enhanced to detect more instances of meaningless or
1088       mutually exclusive attribute specifications and handle such
1089       conflicts more consistently. Mutually exclusive attribute
1090       specifications are ignored with a warning regardless of whether
1091       they appear on the same declaration or on distinct declarations of
1092       the same entity. For example, because the noreturn attribute on the
1093       second declaration below is mutually exclusive with the malloc
1094       attribute on the first, it is ignored and a warning is issued.
1095>
1096      void* __attribute__ ((malloc)) f (unsigned);
1097      void* __attribute__ ((noreturn)) f (unsigned);
1098
1099      warning: ignoring attribute 'noreturn' because it conflicts with attribute
1100 'malloc' [-Wattributes]
1101     * The gcov tool can distinguish functions that begin on a same line
1102       in a source file. This can be a different template instantiation or
1103       a class constructor:
1104
1105File 'ins.C'
1106Lines executed:100.00% of 8
1107Creating 'ins.C.gcov'
1108
1109        -:    0:Source:ins.C
1110        -:    0:Graph:ins.gcno
1111        -:    0:Data:ins.gcda
1112        -:    0:Runs:1
1113        -:    0:Programs:1
1114        -:    1:template<class T>
1115        -:    2:class Foo
1116        -:    3:{
1117        -:    4: public:
1118        2:    5:   Foo(): b (1000) {}
1119------------------
1120Foo<char>::Foo():
1121        1:    5:   Foo(): b (1000) {}
1122------------------
1123Foo<int>::Foo():
1124        1:    5:   Foo(): b (1000) {}
1125------------------
1126        2:    6:   void inc () { b++; }
1127------------------
1128Foo<char>::inc():
1129        1:    6:   void inc () { b++; }
1130------------------
1131Foo<int>::inc():
1132        1:    6:   void inc () { b++; }
1133------------------
1134        -:    7:
1135        -:    8:  private:
1136        -:    9:   int b;
1137        -:   10:};
1138        -:   11:
1139        1:   12:int main(int argc, char **argv)
1140        -:   13:{
1141        1:   14:  Foo<int> a;
1142        1:   15:  Foo<char> b;
1143        -:   16:
1144        1:   17:  a.inc ();
1145        1:   18:  b.inc ();
1146        1:   19:}
1147
1148     * The gcov tool has more accurate numbers for execution of lines in a
1149       source file.
1150     * The gcov tool can use TERM colors to provide more readable output.
1151     * AddressSanitizer gained a new pair of sanitization options,
1152       -fsanitize=pointer-compare and -fsanitize=pointer-subtract, which
1153       warn about subtraction (or comparison) of pointers that point to a
1154       different memory object:
1155
1156int
1157main ()
1158{
1159  /* Heap allocated memory.  */
1160  char *heap1 = (char *)__builtin_malloc (42);
1161  char *heap2 = (char *)__builtin_malloc (42);
1162  if (heap1 > heap2)
1163      return 1;
1164
1165  return 0;
1166}
1167
1168==17465==ERROR: AddressSanitizer: invalid-pointer-pair: 0x604000000010 0x6040000
116900050
1170    #0 0x40070f in main /tmp/pointer-compare.c:7
1171    #1 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86)
1172    #2 0x400629 in _start (/tmp/a.out+0x400629)
1173
11740x604000000010 is located 0 bytes inside of 42-byte region [0x604000000010,0x604
117500000003a)
1176allocated by thread T0 here:
1177    #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan
1178_malloc_linux.cc:86
1179    #1 0x4006ea in main /tmp/pointer-compare.c:5
1180    #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86)
1181
11820x604000000050 is located 0 bytes inside of 42-byte region [0x604000000050,0x604
118300000007a)
1184allocated by thread T0 here:
1185    #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan
1186_malloc_linux.cc:86
1187    #1 0x4006f8 in main /tmp/pointer-compare.c:6
1188    #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86)
1189
1190SUMMARY: AddressSanitizer: invalid-pointer-pair /tmp/pointer-compare.c:7 in main
1191
1192     * The store merging pass has been enhanced to handle bit-fields and
1193       not just constant stores, but also data copying from adjacent
1194       memory locations into other adjacent memory locations, including
1195       bitwise logical operations on the data. The pass can also handle
1196       byte swapping into memory locations.
1197     * The undefined behavior sanitizer gained two new options included in
1198       -fsanitize=undefined: -fsanitize=builtin which diagnoses at run
1199       time invalid arguments to __builtin_clz or __builtin_ctz prefixed
1200       builtins, and -fsanitize=pointer-overflow which performs cheap run
1201       time tests for pointer wrapping.
1202     * A new attribute no_sanitize can be applied to functions to instruct
1203       the compiler not to do sanitization of the options provided as
1204       arguments to the attribute. Acceptable values for no_sanitize match
1205       those acceptable by the -fsanitize command-line option.
1206
1207void __attribute__ ((no_sanitize ("alignment", "object-size")))
1208f () { /* Do something. */; }
1209
1210New Languages and Language specific improvements
1211
1212  Ada
1213
1214     * For its internal exception handling used on the host for error
1215       recovery in the front-end, the compiler now relies on the native
1216       exception handling mechanism of the host platform, which should be
1217       more efficient than the former mechanism.
1218
1219  BRIG (HSAIL)
1220
1221   In this release cycle, the focus for the BRIGFE was on stabilization
1222   and performance improvements. Also a couple of completely new features
1223   were added.
1224     * Improved support for function and module scope group segment
1225       variables. PRM specs define function and module scope group segment
1226       variables as an experimental feature. However, PRM test suite uses
1227       them. Now group segment is handled by separate book keeping of
1228       module scope and function (kernel) offsets. Each function has a
1229       "frame" in the group segment offset to which is given as an
1230       argument, similar to traditional call stack frame handling.
1231     * Reduce the number of type conversions due to the untyped HSAIL
1232       registers. Instead of always representing the HSAIL's untyped
1233       registers as unsigned int, the gccbrig now pre-analyzes the BRIG
1234       code and builds the register variables as a type used the most when
1235       storing or reading data to/from each register. This reduces the
1236       number of total casts which cannot be always optimized away.
1237     * Support for BRIG_KIND_NONE directives.
1238     * Made -O3 the default optimization level for BRIGFE.
1239     * Fixed illegal addresses generated from address expressions which
1240       refer only to offset 0.
1241     * Fixed a bug with reg+offset addressing on 32b segments. In 'large'
1242       mode, the offset is treated as 32bits unless it's in global,
1243       read-only or kernarg address space.
1244     * Fixed a crash caused sometimes by calls with more than 4 arguments.
1245     * Fixed a mis-execution issue with kernels that have both unexpanded
1246       ID functions and calls to subfunctions.
1247     * Treat HSAIL barrier builtins as setjmp/longjump style functions to
1248       avoid illegal optimizations.
1249     * Ensure per WI copies of private variables are aligned correctly.
1250     * libhsail-rt: Assume the host runtime allocates the work group
1251       memory.
1252
1253  C family
1254
1255     * New command-line options have been added for the C and C++
1256       compilers:
1257          + [4]-Wmultistatement-macros warns about unsafe macros expanding
1258            to multiple statements used as a body of a statement such as
1259            if, else, while, switch, or for.
1260          + [5]-Wstringop-truncation warns for calls to bounded string
1261            manipulation functions such as strncat, strncpy, and stpncpy
1262            that might either truncate the copied string or leave the
1263            destination unchanged. For example, the following call to
1264            strncat is diagnosed because it appends just three of the four
1265            characters from the source string.
1266void append (char *buf, size_t bufsize)
1267{
1268    strncat (buf, ".txt", 3);
1269}
1270warning: 'strncat' output truncated copying 3 bytes from a string of length 4 [-
1271Wstringop-truncation]
1272            Similarly, in the following example, the call to strncpy
1273            specifies the size of the destination buffer as the bound. If
1274            the length of the source string is equal to or greater than
1275            this size the result of the copy will not be NUL-terminated.
1276            Therefore, the call is also diagnosed. To avoid the warning,
1277            specify sizeof buf - 1 as the bound and set the last element
1278            of the buffer to NUL.
1279void copy (const char *s)
1280{
1281    char buf[80];
1282    strncpy (buf, s, sizeof buf);
12831284}
1285warning: 'strncpy' specified bound 80 equals destination size [-Wstringop-trunca
1286tion]
1287            The -Wstringop-truncation option is included in -Wall.
1288            Note that due to GCC bug [6]82944, defining strncat, strncpy,
1289            or stpncpy as a macro in a system header as some
1290            implementations do, suppresses the warning.
1291          + [7]-Wif-not-aligned controls warnings issued in response to
1292            invalid uses of objects declared with attribute
1293            [8]warn_if_not_aligned.
1294            The -Wif-not-aligned option is included in -Wall.
1295          + [9]-Wmissing-attributes warns when a declaration of a function
1296            is missing one or more attributes that a related function is
1297            declared with and whose absence may adversely affect the
1298            correctness or efficiency of generated code. For example, in
1299            C++, the warning is issued when an explicit specialization of
1300            a primary template declared with attribute alloc_align,
1301            alloc_size, assume_aligned, format, format_arg, malloc, or
1302            nonnull is declared without it. Attributes deprecated, error,
1303            and warning suppress the warning.
1304            The -Wmissing-attributes option is included in -Wall.
1305          + [10]-Wpacked-not-aligned warns when a struct or union declared
1306            with attribute packed defines a member with an explicitly
1307            specified alignment greater than 1. Such a member will wind up
1308            under-aligned. For example, a warning will be issued for the
1309            definition of struct A in the following:
1310struct __attribute__ ((aligned (8)))
1311S8 { char a[8]; };
1312
1313struct __attribute__ ((packed)) A
1314{
1315    struct S8 s8;
1316};
1317warning: alignment 1 of 'struct S' is less than 8 [-Wpacked-not-aligned]
1318            The -Wpacked-not-aligned option is included in -Wall.
1319          + -Wcast-function-type warns when a function pointer is cast to
1320            an incompatible function pointer. This warning is enabled by
1321            -Wextra.
1322          + -Wsizeof-pointer-div warns for suspicious divisions of the
1323            size of a pointer by the size of the elements it points to,
1324            which looks like the usual way to compute the array size but
1325            won't work out correctly with pointers. This warning is
1326            enabled by -Wall.
1327          + -Wcast-align=strict warns whenever a pointer is cast such that
1328            the required alignment of the target is increased. For
1329            example, warn if a char * is cast to an int * regardless of
1330            the target machine.
1331          + -fprofile-abs-path creates absolute path names in the .gcno
1332            files. This allows gcov to find the correct sources in
1333            projects where compilations occur with different working
1334            directories.
1335     * -fno-strict-overflow is now mapped to -fwrapv -fwrapv-pointer and
1336       signed integer overflow is now undefined by default at all
1337       optimization levels. Using -fsanitize=signed-integer-overflow is
1338       now the preferred way to audit code, -Wstrict-overflow is
1339       deprecated.
1340     * The [11]-Warray-bounds option has been improved to detect more
1341       instances of out-of-bounds array indices and pointer offsets. For
1342       example, negative or excessive indices into flexible array members
1343       and string literals are detected.
1344     * The [12]-Wrestrict option introduced in GCC 7 has been enhanced to
1345       detect many more instances of overlapping accesses to objects via
1346       restrict-qualified arguments to standard memory and string
1347       manipulation functions such as memcpy and strcpy. For example, the
1348       strcpy call in the function below attempts to truncate the string
1349       by replacing its initial characters with the last four. However,
1350       because the function writes the terminating NUL into a[4], the
1351       copies overlap and the call is diagnosed.
1352void f (void)
1353{
1354    char a[] = "abcd1234";
1355    strcpy (a, a + 4);
13561357}
1358warning: 'strcpy' accessing 5 bytes at offsets 0 and 4 overlaps 1 byte at offset
1359 4 [-Wrestrict]
1360       The -Wrestrict option is included in -Wall.
1361     * Several optimizer enhancements have enabled improvements to the
1362       [13]-Wformat-overflow and [14]-Wformat-truncation options. The
1363       warnings detect more instances of buffer overflow and truncation
1364       than in GCC 7 and are better at avoiding certain kinds of false
1365       positives.
1366     * When reporting mismatching argument types at a function call, the C
1367       and C++ compilers now underline both the argument and the pertinent
1368       parameter in the declaration.
1369$ gcc arg-type-mismatch.cc
1370arg-type-mismatch.cc: In function 'int caller(int, int, float)':
1371arg-type-mismatch.cc:5:24: error: invalid conversion from 'int' to 'const char*'
1372 [-fpermissive]
1373   return callee(first, second, third);
1374                        ^~~~~~
1375arg-type-mismatch.cc:1:40: note:   initializing argument 2 of 'int callee(int, c
1376onst char*, float)'
1377 extern int callee(int one, const char *two, float three);
1378                            ~~~~~~~~~~~~^~~
1379
1380     * When reporting on unrecognized identifiers, the C and C++ compilers
1381       will now emit fix-it hints suggesting #include directives for
1382       various headers in the C and C++ standard libraries.
1383$ gcc incomplete.c
1384incomplete.c: In function 'test':
1385incomplete.c:3:10: error: 'NULL' undeclared (first use in this function)
1386   return NULL;
1387          ^~~~
1388incomplete.c:3:10: note: 'NULL' is defined in header '<stddef.h>'; did you forge
1389t to '#include <stddef.h>'?
1390incomplete.c:1:1:
1391+#include <stddef.h>
1392 const char *test(void)
1393incomplete.c:3:10:
1394   return NULL;
1395          ^~~~
1396incomplete.c:3:10: note: each undeclared identifier is reported only once for ea
1397ch function it appears in
1398
1399$ gcc incomplete.cc
1400incomplete.cc:1:6: error: 'string' in namespace 'std' does not name a type
1401 std::string s("hello world");
1402      ^~~~~~
1403incomplete.cc:1:1: note: 'std::string' is defined in header '<string>'; did you
1404forget to '#include <string>'?
1405+#include <string>
1406 std::string s("hello world");
1407 ^~~
1408
1409     * The C and C++ compilers now use more intuitive locations when
1410       reporting on missing semicolons, and offer fix-it hints:
1411$ gcc t.c
1412t.c: In function 'test':
1413t.c:3:12: error: expected ';' before '}' token
1414   return 42
1415            ^
1416            ;
1417 }
1418 ~
1419
1420     * When reporting on missing '}' and ')' tokens, the C and C++
1421       compilers will now highlight the corresponding '{' and '(' token,
1422       issuing a 'note' if it's on a separate line:
1423$ gcc unclosed.c
1424unclosed.c: In function 'log_when_out_of_range':
1425unclosed.c:12:50: error: expected ')' before '{' token
1426       && (temperature < MIN || temperature > MAX) {
1427                                                  ^~
1428                                                  )
1429unclosed.c:11:6: note: to match this '('
1430   if (logging_enabled && check_range ()
1431      ^
1432
1433       or highlighting it directly if it's on the same line:
1434$ gcc unclosed-2.c
1435unclosed-2.c: In function 'test':
1436unclosed-2.c:8:45: error: expected ')' before '{' token
1437   if (temperature < MIN || temperature > MAX {
1438      ~                                      ^~
1439                                             )
1440
1441       They will also emit fix-it hints.
1442
1443  C++
1444
1445     * GCC 8 (-fabi-version=12) has a couple of corrections to the calling
1446       convention, which changes the ABI for some uncommon code:
1447          + Passing an empty class as an argument now takes up no space on
1448            x86_64, as required by the psABI.
1449          + Passing or returning a class with only deleted copy and move
1450            constructors now uses the same calling convention as a class
1451            with a non-trivial copy or move constructor. This only affects
1452            C++17 mode, as in earlier standards passing or returning such
1453            a class was impossible.
1454          + WARNING: In GCC 8.1 the second change mistakenly also affects
1455            classes with a deleted copy constructor and defaulted trivial
1456            move constructor (bug [15]c++/86094). This issue is fixed in
1457            GCC 8.2 (-fabi-version=13).
1458       You can test whether these changes affect your code with -Wabi=11
1459       (or -Wabi=12 in GCC 8.2 for the third issue); if these changes are
1460       problematic for your project, the GCC 7 ABI can be selected with
1461       -fabi-version=11.
1462     * The value of the C++11 alignof operator has been corrected to match
1463       C _Alignof (minimum alignment) rather than GNU __alignof__
1464       (preferred alignment); on ia32 targets this means that
1465       alignof(double) is now 4 rather than 8. Code that wants the
1466       preferred alignment should use __alignof__ instead.
1467     * New command-line options have been added for the C++ compiler to
1468       control warnings:
1469          + [16]-Wclass-memaccess warns when objects of non-trivial class
1470            types are manipulated in potentially unsafe ways by raw memory
1471            functions such as memcpy, or realloc. The warning helps detect
1472            calls that bypass user-defined constructors or copy-assignment
1473            operators, corrupt virtual table pointers, data members of
1474            const-qualified types or references, or member pointers. The
1475            warning also detects calls that would bypass access controls
1476            to data members. For example, a call such as:
1477        memcpy (&std::cout, &std::cerr, sizeof std::cout);
1478            results in
1479        warning: 'void* memcpy(void*, const void*, long unsigned int)' writing t
1480o an object of type 'std::ostream' {aka 'class std::basic_ostream<char>'} with n
1481o trivial copy-assignment [-Wclass-memaccess]
1482            The -Wclass-memaccess option is included in -Wall.
1483     * The C++ front end has experimental support for some of the upcoming
1484       C++2a draft features with the -std=c++2a or -std=gnu++2a flags,
1485       including designated initializers, default member initializers for
1486       bit-fields, __VA_OPT__ (except that #__VA_OPT__ is unsupported),
1487       lambda [=, this] captures, etc. For a full list of new features,
1488       see [17]the C++ status page.
1489     * When reporting on attempts to access private fields of a class or
1490       struct, the C++ compiler will now offer fix-it hints showing how to
1491       use an accessor function to get at the field in question, if one
1492       exists.
1493$ gcc accessor.cc
1494accessor.cc: In function 'void test(foo*)':
1495accessor.cc:12:12: error: 'double foo::m_ratio' is private within this context
1496   if (ptr->m_ratio >= 0.5)
1497            ^~~~~~~
1498accessor.cc:7:10: note: declared private here
1499   double m_ratio;
1500          ^~~~~~~
1501accessor.cc:12:12: note: field 'double foo::m_ratio' can be accessed via 'double
1502 foo::get_ratio() const'
1503   if (ptr->m_ratio >= 0.5)
1504            ^~~~~~~
1505            get_ratio()
1506
1507     * The C++ compiler can now give you a hint if you use a macro before
1508       it was defined (e.g. if you mess up the order of your #include
1509       directives):
1510$ gcc ordering.cc
1511ordering.cc:2:24: error: expected ';' at end of member declaration
1512   virtual void clone() const OVERRIDE { }
1513                        ^~~~~
1514                             ;
1515ordering.cc:2:30: error: 'OVERRIDE' does not name a type
1516   virtual void clone() const OVERRIDE { }
1517                              ^~~~~~~~
1518ordering.cc:2:30: note: the macro 'OVERRIDE' had not yet been defined
1519In file included from ordering.cc:5:
1520c++11-compat.h:2: note: it was later defined here
1521 #define OVERRIDE override
1522
1523
1524     * The -Wold-style-cast diagnostic can now emit fix-it hints telling
1525       you when you can use a static_cast, const_cast, or
1526       reinterpret_cast.
1527$ gcc -c old-style-cast-fixits.cc -Wold-style-cast
1528old-style-cast-fixits.cc: In function 'void test(void*)':
1529old-style-cast-fixits.cc:5:19: warning: use of old-style cast to 'struct foo*' [
1530-Wold-style-cast]
1531   foo *f = (foo *)ptr;
1532                   ^~~
1533            ----------
1534            static_cast<foo *> (ptr)
1535
1536     * When reporting on problems within extern "C" linkage
1537       specifications, the C++ compiler will now display the location of
1538       the start of the extern "C".
1539$ gcc -c extern-c.cc
1540extern-c.cc:3:1: error: template with C linkage
1541 template <typename T> void test (void);
1542 ^~~~~~~~
1543In file included from extern-c.cc:1:
1544unclosed.h:1:1: note: 'extern "C"' linkage started here
1545 extern "C" {
1546 ^~~~~~~~~~
1547extern-c.cc:3:39: error: expected '}' at end of input
1548 template <typename T> void test (void);
1549                                       ^
1550In file included from extern-c.cc:1:
1551unclosed.h:1:12: note: to match this '{'
1552 extern "C" {
1553            ^
1554
1555     * When reporting on mismatching template types, the C++ compiler will
1556       now use color to highlight the mismatching parts of the template,
1557       and will elide the parameters that are common between two
1558       mismatching templates, printing [...] instead:
1559$ gcc templates.cc
1560templates.cc: In function 'void test()':
1561templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl
1562e>' to 'vector<int>'
1563   fn_1(vector<double> ());
1564        ^~~~~~~~~~~~~~~~~
1565templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<[...]
1566,double>' to 'map<[...],int>'
1567   fn_2(map<int, double>());
1568        ^~~~~~~~~~~~~~~~~~
1569
1570       Those [...] elided parameters can be seen using -fno-elide-type:
1571$ gcc templates.cc -fno-elide-type
1572templates.cc: In function 'void test()':
1573templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl
1574e>' to 'vector<int>'
1575   fn_1(vector<double> ());
1576        ^~~~~~~~~~~~~~~~~
1577templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<int,d
1578ouble>' to 'map<int,int>'
1579   fn_2(map<int, double>());
1580        ^~~~~~~~~~~~~~~~~~
1581
1582       The C++ compiler has also gained an option
1583       -fdiagnostics-show-template-tree which visualizes such mismatching
1584       templates in a hierarchical form:
1585$ gcc templates-2.cc -fdiagnostics-show-template-tree
1586templates-2.cc: In function 'void test()':
1587templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou
1588ble>' to 'vector<int>'
1589  vector<
1590    [double != int]>
1591   fn_1(vector<double> ());
1592        ^~~~~~~~~~~~~~~~~
1593templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve
1594ctor<double> >()' from 'map<map<[...],vector<double>>,vector<double>>' to 'map<m
1595ap<[...],vector<float>>,vector<float>>'
1596  map<
1597    map<
1598      [...],
1599      vector<
1600        [double != float]>>,
1601    vector<
1602      [double != float]>>
1603   fn_2(map<map<int, vector<double>>, vector<double>> ());
1604        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1605
1606       which again works with -fno-elide-type:
1607$ gcc templates-2.cc -fdiagnostics-show-template-tree -fno-elide-type
1608templates-2.cc: In function 'void test()':
1609templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou
1610ble>' to 'vector<int>'
1611  vector<
1612    [double != int]>
1613   fn_1(vector<double> ());
1614        ^~~~~~~~~~~~~~~~~
1615templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve
1616ctor<double> >()' from 'map<map<int,vector<double>>,vector<double>>' to 'map<map
1617<int,vector<float>>,vector<float>>'
1618  map<
1619    map<
1620      int,
1621      vector<
1622        [double != float]>>,
1623    vector<
1624      [double != float]>>
1625   fn_2(map<map<int, vector<double>>, vector<double>> ());
1626        ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1627
1628     * Flowing off the end of a non-void function is considered
1629       unreachable and may be subject to optimization on that basis. As a
1630       result of this change, -Wreturn-type warnings are enabled by
1631       default for C++.
1632
1633    Runtime Library (libstdc++)
1634
1635     * Improved experimental support for C++17, including the following
1636       features:
1637          + Deduction guides to support class template argument deduction.
1638          + std::filesystem implementation.
1639          + std::char_traits<char> and std::char_traits<wchar_t> are
1640            usable in constant expressions.
1641          + std::to_chars and std::from_chars (for integers only, not for
1642            floating point types).
1643     * Experimental support for C++2a: std::to_address (thanks to Glen
1644       Fernandes) and std::endian.
1645     * On GNU/Linux, std::random_device::entropy() accesses the kernel's
1646       entropy count for the random device, if known (thanks to Xi
1647       Ruoyao).
1648     * Support for std::experimental::source_location.
1649     * AddressSanitizer integration for std::vector, detecting
1650       out-of-range accesses to the unused capacity of a vector.
1651     * Extensions __gnu_cxx::airy_ai and __gnu_cxx::airy_bi added to the
1652       Mathematical Special Functions.
1653
1654  Fortran
1655
1656     * The main version of libfortran has been changed to 5.
1657     * Parameterized derived types, a major feature of Fortran 2003, have
1658       been implemented.
1659     * Partial support is provided for Fortran 2018 teams, which are
1660       hierarchical subsets of images that execute independently of other
1661       image subsets.
1662     * The maximum rank for arrays has been increased to 15, conforming to
1663       the Fortran 2008 standard.
1664     * Transformational intrinsics are now fully supported in
1665       initialization expressions.
1666     * New flag -fc-prototypes to write C prototypes for BIND(C)
1667       procedures and variables.
1668     * If -fmax-stack-var-size is honored if given together with -Ofast,
1669       -fstack-arrays is no longer set in that case.
1670     * New options -fdefault-real-16 and -fdefault-real-10 to control the
1671       default kind of REAL variables.
1672     * A warning is now issued if an array subscript inside a DO loop
1673       could lead to an out-of-bounds-access. The new option
1674       -Wdo-subscript, enabled by -Wextra, warns about this even if the
1675       compiler can not prove that the code will be executed.
1676     * The Fortran front end now attempts to interchange loops if it is
1677       deemed profitable. So far, this is restricted to FORALL and DO
1678       CONCURRENT statements with multiple indices. This behavior be
1679       controlled with the new flag -ffrontend-loop-interchange, which is
1680       enabled with optimization by default. The
1681       -Wfrontend-loop-interchange option warns about such occurrences.
1682     * When an actual argument contains too few elements for a dummy
1683       argument, an error is now issued. The -std=legacy option can be
1684       used to still compile such code.
1685     * The RECL= argument to OPEN and INQUIRE statements now allows 64-bit
1686       integers, making records larger than 2GiB possible.
1687     * The GFORTRAN_DEFAULT_RECL environment variable no longer has any
1688       effect. The record length for preconnected units is now larger than
1689       any practical limit, same as for sequential access units opened
1690       without an explicit RECL= specifier.
1691     * Character variables longer than HUGE(0) elements are now possible
1692       on 64-bit targets. Note that this changes the procedure call ABI
1693       for all procedures with character arguments on 64-bit targets, as
1694       the type of the hidden character length argument has changed. The
1695       hidden character length argument is now of type INTEGER(C_SIZE_T).
1696     * Partial support is provided for Fortran 2018 teams, which are
1697       hierarchical subsets of images that execute independently of other
1698       image subsets.
1699
1700  Go
1701
1702     * GCC 8 provides a complete implementation of the Go 1.10.1 user
1703       packages.
1704     * The garbage collector is now fully concurrent. As before, values
1705       stored on the stack are scanned conservatively, but value stored in
1706       the heap are scanned precisely.
1707     * Escape analysis is fully implemented and enabled by default in the
1708       Go frontend. This significantly reduces the number of heap
1709       allocations by allocating values on the stack instead.
1710
1711libgccjit
1712
1713   The libgccjit API gained four new entry points:
1714     * [18]gcc_jit_type_get_vector and
1715     * [19]gcc_jit_context_new_rvalue_from_vector for working with
1716       vectors,
1717     * [20]gcc_jit_type_get_aligned
1718     * [21]gcc_jit_function_get_address
1719
1720   The C code generated by [22]gcc_jit_context_dump_reproducer_to_file is
1721   now easier-to-read.
1722
1723New Targets and Target Specific Improvements
1724
1725  AArch64
1726
1727     * The Armv8.4-A architecture is now supported. It can be used by
1728       specifying the -march=armv8.4-a option.
1729     * The Dot Product instructions are now supported as an optional
1730       extension to the Armv8.2-A architecture and newer and are mandatory
1731       on Armv8.4-A. The extension can be used by specifying the +dotprod
1732       architecture extension. E.g. -march=armv8.2-a+dotprod.
1733     * The Armv8-A +crypto extension has now been split into two
1734       extensions for finer grained control:
1735          + +aes which contains the Armv8-A AES crytographic instructions.
1736          + +sha2 which contains the Armv8-A SHA2 and SHA1 cryptographic
1737            instructions.
1738       Using +crypto will now enable these two extensions.
1739     * New Armv8.4-A FP16 Floating Point Multiplication Variant
1740       instructions have been added. These instructions are mandatory in
1741       Armv8.4-A but available as an optional extension to Armv8.2-A and
1742       Armv8.3-A. The new extension can be used by specifying the +fp16fml
1743       architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A
1744       the instructions can be enabled by specifying +fp16.
1745     * New cryptographic instructions have been added as optional
1746       extensions to Armv8.2-A and newer. These instructions can be
1747       enabled with:
1748          + +sha3 New SHA3 and SHA2 instructions from Armv8.4-A. This
1749            implies +sha2.
1750          + +sm4 New SM3 and SM4 instructions from Armv8.4-A.
1751     * The Scalable Vector Extension (SVE) is now supported as an optional
1752       extension to the Armv8.2-A architecture and newer. This support
1753       includes automatic vectorization with SVE instructions, but it does
1754       not yet include the SVE Arm C Language Extensions (ACLE). It can be
1755       enabled by specifying the +sve architecture extension (for example,
1756       -march=armv8.2-a+sve). By default, the generated code works with
1757       all vector lengths, but it can be made specific to N-bit vectors
1758       using -msve-vector-bits=N.
1759     * Support has been added for the following processors (GCC
1760       identifiers in parentheses):
1761          + Arm Cortex-A75 (cortex-a75).
1762          + Arm Cortex-A55 (cortex-a55).
1763          + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE
1764            (cortex-a75.cortex-a55).
1765       The GCC identifiers can be used as arguments to the -mcpu or -mtune
1766       options, for example: -mcpu=cortex-a75 or -mtune=cortex-a75 or as
1767       arguments to the equivalent target attributes and pragmas.
1768
1769  ARC
1770
1771     * Added support for:
1772          + Fast interrupts.
1773          + Naked functions.
1774          + aux variable attributes.
1775          + uncached type qualifier.
1776          + Secure functions via sjli instruction.
1777     * New exception handling implementation.
1778     * Revamped trampoline implementation.
1779     * Refactored small data feature implementation, controlled via -G
1780       command line option.
1781     * New support for reduced register set ARC architecture
1782       configurations, controlled via -mrf16 command line option.
1783     * Refurbished and improved support for zero overhead loops.
1784       Introduced -mlpc-width command line option to control the width of
1785       lp_count register.
1786
1787  ARM
1788
1789     * The -mfpu option now takes a new option setting of -mfpu=auto. When
1790       set to this the floating-point and SIMD settings are derived from
1791       the settings of the -mcpu or -march options. The internal CPU
1792       configurations have been updated with information about the
1793       permitted floating-point configurations supported. See the user
1794       guide for further information about the extended option syntax for
1795       controlling architectural extensions via the -march option.
1796       -mfpu=auto is now the default setting unless the compiler has been
1797       configured with an explicit --with-fpu option.
1798     * The -march and -mcpu options now accept optional extensions to the
1799       architecture or CPU option, allowing the user to enable or disable
1800       any such extensions supported by that architecture or CPU such as
1801       (but not limited to) floating-point and AdvancedSIMD. For example:
1802       the option -mcpu=cortex-a53+nofp will generate code for the
1803       Cortex-A53 processor with no floating-point support. This, in
1804       combination with the new -mfpu=auto option, provides a
1805       straightforward way of specifying a valid build target through a
1806       single -mcpu or -march option. The -mtune option accepts the same
1807       arguments as -mcpu but only the CPU name has an effect on tuning.
1808       The architecture extensions do not have any effect. For details of
1809       what extensions a particular architecture or CPU option supports
1810       please refer to the [23]documentation.
1811     * The -mstructure-size-boundary option has been deprecated and will
1812       be removed in a future release.
1813     * The default link behavior for Armv6 and Armv7-R targets has been
1814       changed to produce BE8 format when generating big-endian images. A
1815       new flag -mbe32 can be used to force the linker to produce legacy
1816       BE32 format images. There is no change of behavior for Armv6-M and
1817       other Armv7 or later targets: these already defaulted to BE8
1818       format. This change brings GCC into alignment with other compilers
1819       for the ARM architecture.
1820     * The Armv8-R architecture is now supported. It can be used by
1821       specifying the -march=armv8-r option.
1822     * The Armv8.3-A architecture is now supported. It can be used by
1823       specifying the -march=armv8.3-a option.
1824     * The Armv8.4-A architecture is now supported. It can be used by
1825       specifying the -march=armv8.4-a option.
1826     * The Dot Product instructions are now supported as an optional
1827       extension to the Armv8.2-A architecture and newer and are mandatory
1828       on Armv8.4-A. The extension can be used by specifying the +dotprod
1829       architecture extension. E.g. -march=armv8.2-a+dotprod.
1830     * Support for setting extensions and architectures using the GCC
1831       target pragma and attribute has been added. It can be used by
1832       specifying #pragma GCC target ("arch=..."), #pragma GCC target
1833       ("+extension"), __attribute__((target("arch=..."))) or
1834       __attribute__((target("+extension"))).
1835     * New Armv8.4-A FP16 Floating Point Multiplication Variant
1836       instructions have been added. These instructions are mandatory in
1837       Armv8.4-A but available as an optional extension to Armv8.2-A and
1838       Armv8.3-A. The new extension can be used by specifying the +fp16fml
1839       architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A
1840       the instructions can be enabled by specifying +fp16.
1841     * Support has been added for the following processors (GCC
1842       identifiers in parentheses):
1843          + Arm Cortex-A75 (cortex-a75).
1844          + Arm Cortex-A55 (cortex-a55).
1845          + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE
1846            (cortex-a75.cortex-a55).
1847          + Arm Cortex-R52 for Armv8-R (cortex-r52).
1848       The GCC identifiers can be used as arguments to the -mcpu or -mtune
1849       options, for example: -mcpu=cortex-a75 or -mtune=cortex-r52 or as
1850       arguments to the equivalent target attributes and pragmas.
1851
1852  AVR
1853
1854     * The AVR port now supports the following XMEGA-like devices:
1855
1856     ATtiny212, ATtiny214, ATtiny412, ATtiny414, ATtiny416, ATtiny417,
1857     ATtiny814, ATtiny816, ATtiny817, ATtiny1614, ATtiny1616, ATtiny1617,
1858     ATtiny3214, ATtiny3216, ATtiny3217
1859       The new devices are listed under [24]-mmcu=avrxmega3.
1860          + These devices see flash memory in the RAM address space, so
1861            that features like PROGMEM and __flash are not needed any more
1862            (as opposed to other AVR families for which read-only data
1863            will be located in RAM except special, non-standard features
1864            are used to locate and access such data). This requires that
1865            the compiler is used with Binutils 2.29 or newer so that
1866            [25]read-only data will be located in flash memory.
1867          + A new command-line option -mshort-calls is supported. This
1868            option is used internally for multilib selection of the
1869            avrxmega3 variants. It is not an optimization option. Do not
1870            set it by hand.
1871     * The compiler now generates [26]efficient interrupt service routine
1872       (ISR) prologues and epilogues. This is achieved by using the new
1873       [27]AVR pseudo instruction __gcc_isr which is supported and
1874       resolved by the GNU assembler.
1875          + As the __gcc_isr pseudo-instruction will be resolved by the
1876            assembler, inline assembly is transparent to the process. This
1877            means that when inline assembly uses an instruction like INC
1878            that clobbers the condition code, then the assembler will
1879            detect this and generate an appropriate ISR prologue /
1880            epilogue chunk to save / restore SREG as needed.
1881          + A new command-line option -mno-gas-isr-prologues disables the
1882            generation of the __gcc_isr pseudo instruction. Any non-naked
1883            ISR will save and restore SREG, tmp_reg and zero_reg, no
1884            matter whether the respective register is clobbered or used.
1885          + The feature is turned on per default for all optimization
1886            levels except for -O0 and -Og. It is explicitly enabled by
1887            means of option -mgas-isr-prologues.
1888          + Support has been added for a new [28]AVR function attribute
1889            no_gccisr. It can be used to disable __gcc_isr pseudo
1890            instruction generation for individual ISRs.
1891          + This optimization is only available if GCC is configured with
1892            GNU Binutils 2.29 or newer; or at least with a version of
1893            Binutils that implements feature [29]PR21683.
1894     * The compiler no more saves / restores registers in main; the effect
1895       is the same as if attribute OS_task was specified for main. This
1896       optimization can be switched off by the new command-line option
1897       -mno-main-is-OS_task.
1898
1899  IA-32/x86-64
1900
1901     * The x86 port now supports the naked function attribute.
1902     * Better tuning for znver1 and Intel Core based CPUs.
1903     * Vectorization cost metrics has been reworked leading to significant
1904       improvements on some benchmarks.
1905     * GCC now supports the Intel CPU named Cannonlake through
1906       -march=cannonlake. The switch enables the AVX512VBMI, AVX512IFMA
1907       and SHA ISA extensions.
1908     * GCC now supports the Intel CPU named Icelake through
1909       -march=icelake. The switch enables the AVX512VNNI, GFNI, VAES,
1910       AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID and AVX512VPOPCNTDQ
1911       ISA extensions.
1912     * GCC now supports the Intel Control-flow Enforcement Technology
1913       (CET) extension through -fcf-protection option.
1914
1915  NDS32
1916
1917     * New command-line options -mext-perf, -mext-perf2, and -mext-string
1918       have been added for performance extension instructions.
1919
1920  Nios II
1921
1922     * The Nios II back end has been improved to generate better-optimized
1923       code. Changes include switching to LRA, more accurate cost models,
1924       and more compact code for addressing static variables.
1925     * New command-line options -mgprel-sec= and -mr0rel-sec= have been
1926       added.
1927     * The stack-smashing protection options are now enabled on Nios II.
1928
1929  PA-RISC
1930
1931     * The default call ABI on 32-bit linux has been changed from callee
1932       copies to caller copies. This affects objects larger than eight
1933       bytes passed by value. The goal is to improve compatibility with
1934       x86 and resolve issues with OpenMP.
1935     * Other PA-RISC targets are unchanged.
1936
1937  PowerPC / PowerPC64 / RS6000
1938
1939     * The PowerPC SPE support is split off to a separate powerpcspe port.
1940       The separate port is deprecated and might be removed in a future
1941       release.
1942     * The Paired Single support (as used on some PPC750 CPUs, -mpaired,
1943       powerpc*-*-linux*paired*) is deprecated and will be removed in a
1944       future release.
1945     * The Xilinx floating point support (-mxilinx-fpu,
1946       powerpc-xilinx-eabi*) is deprecated and will be removed in a future
1947       release.
1948     * Support for using big-endian AltiVec intrinsics on a little-endian
1949       target (-maltivec=be) is deprecated and will be removed in a future
1950       release.
1951
1952  Tile
1953
1954     * The TILE-Gx port is deprecated and will be removed in a future
1955       release.
1956
1957Operating Systems
1958
1959  Windows
1960
1961     * GCC on Microsoft Windows can now be configured via
1962       --enable-mingw-wildcard or --disable-mingw-wildcard to force a
1963       specific behavior for GCC itself with regards to supporting the
1964       wildcard character. Prior versions of GCC would follow the
1965       configuration of the MinGW runtime. This behavior can still be
1966       obtained by not using the above options or by using
1967       --enable-mingw-wildcard=platform.
1968
1969Improvements for plugin authors
1970
1971     * Plugins can now register a callback hook for when comments are
1972       encountered by the C and C++ compilers, e.g. allowing for plugins
1973       to handle documentation markup in code comments.
1974     * The gdbinit support script for debugging GCC now has a
1975       break-on-diagnostic command, providing an easy way to trigger a
1976       breakpoint whenever a diagnostic is emitted.
1977     * The API for creating fix-it hints now supports newlines, and for
1978       emitting mutually incompatible fix-it hints for one diagnostic.
1979
1980GCC 8.1
1981
1982   This is the [30]list of problem reports (PRs) from GCC's bug tracking
1983   system that are known to be fixed in the 8.1 release. This list might
1984   not be complete (that is, it is possible that some PRs that have been
1985   fixed are not listed here).
1986
1987GCC 8.2
1988
1989   This is the [31]list of problem reports (PRs) from GCC's bug tracking
1990   system that are known to be fixed in the 8.2 release. This list might
1991   not be complete (that is, it is possible that some PRs that have been
1992   fixed are not listed here).
1993
1994  General Improvements
1995
1996     * Fixed LTO link-time performance problems caused by an overflow in
1997       the partitioning algorithm while building large binaries.
1998
1999  Language Specific Changes
2000
2001    C++
2002
2003   GCC 8.2 fixed a bug introduced in GCC 8.1 affecting passing or
2004   returning of classes with a deleted copy constructor and defaulted
2005   trivial move constructor (bug [32]c++/86094). GCC 8.2 introduces
2006   -fabi-version=13 and makes it the default, ABI incompatibilities
2007   between GCC 8.1 and 8.2 can be reported with -Wabi=12. See [33]C++
2008   changes for more details.
2009
2010  Target Specific Changes
2011
2012    IA-32/x86-64
2013
2014     * -mtune=native performance regression [34]PR84413 on Intel Skylake
2015       processors has been fixed.
2016
2017GCC 8.3
2018
2019   This is the [35]list of problem reports (PRs) from GCC's bug tracking
2020   system that are known to be fixed in the 8.3 release. This list might
2021   not be complete (that is, it is possible that some PRs that have been
2022   fixed are not listed here).
2023
2024  Windows
2025
2026     * A C++ Microsoft ABI bitfield layout bug, [36]PR87137 has been
2027       fixed. A non-field declaration could cause the current bitfield
2028       allocation unit to be completed, incorrectly placing a following
2029       bitfield into a new allocation unit. The Microsoft ABI is selected
2030       for:
2031          + Mingw targets
2032          + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields
2033            option is specified, or __attribute__((ms_struct)) is used
2034          + SuperH targets when the -mhitachi option is specified, or
2035            __attribute__((renesas)) is used
2036       GCC 8 introduced additional cases of this defect, but rather than
2037       resolve only those regressions, we decided to resolve all the cases
2038       of this defect in single change.
2039
2040GCC 8.4
2041
2042   This is the [37]list of problem reports (PRs) from GCC's bug tracking
2043   system that are known to be fixed in the 8.4 release. This list might
2044   not be complete (that is, it is possible that some PRs that have been
2045   fixed are not listed here).
2046
2047
2048    For questions related to the use of GCC, please consult these web
2049    pages and the [38]GCC manuals. If that fails, the
2050    [39]gcc-help@gcc.gnu.org mailing list might help. Comments on these
2051    web pages and the development of GCC are welcome on our developer
2052    list at [40]gcc@gcc.gnu.org. All of [41]our lists have public
2053    archives.
2054
2055   Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and
2056   distribution of this entire article is permitted in any medium,
2057   provided this notice is preserved.
2058
2059   These pages are [43]maintained by the GCC team. Last modified
2060   2020-03-04[44].
2061
2062References
2063
2064   1. http://gcc.gnu.org/gcc-8/porting_to.html
2065   2. http://gcc.gnu.org/onlinedocs/index.html#current
2066   3. https://gcc.gnu.org/ml/gcc/2018-04/msg00102.html
2067   4. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmultistatement-macros
2068   5. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wstringop-truncation
2069   6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82944
2070   7. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wif-not-aligned
2071   8. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Common-Variable-Attributes.html#index-warn_005fif_005fnot_005faligned-variable-attribute
2072   9. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmissing-attributes
2073  10. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wpacked-not-aligned
2074  11. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Warray-bounds
2075  12. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wrestrict
2076  13. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-overflow
2077  14. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-truncation
2078  15. https://gcc.gnu.org/PR86094
2079  16. https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess
2080  17. http://gcc.gnu.org/projects/cxx-status.html#cxx2a
2081  18. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_vector
2082  19. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/expressions.html#gcc_jit_context_new_rvalue_from_vector
2083  20. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_aligned
2084  21. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/function-pointers.html#gcc_jit_function_get_address
2085  22. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/contexts.html#gcc_jit_context_dump_reproducer_to_file
2086  23. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/ARM-Options.html#ARM-Options
2087  24. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Options.html
2088  25. https://sourceware.org/PR21472
2089  26. https://gcc.gnu.org/PR20296
2090  27. https://sourceware.org/binutils/docs-2.29/as/AVR-Pseudo-Instructions.html
2091  28. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Function-Attributes.html
2092  29. https://sourceware.org/PR21683
2093  30. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.0
2094  31. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.2
2095  32. https://gcc.gnu.org/PR86094
2096  33. http://gcc.gnu.org/gcc-8/changes.html#cxx
2097  34. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84413
2098  35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.3
2099  36. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137
2100  37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.4
2101  38. https://gcc.gnu.org/onlinedocs/
2102  39. mailto:gcc-help@gcc.gnu.org
2103  40. mailto:gcc@gcc.gnu.org
2104  41. https://gcc.gnu.org/lists.html
2105  42. https://www.fsf.org/
2106  43. https://gcc.gnu.org/about.html
2107  44. http://validator.w3.org/check/referer
2108======================================================================
2109http://gcc.gnu.org/gcc-7/index.html
2110                              GCC 7 Release Series
2111
2112   Nov 14, 2019
2113
2114   The [1]GNU project and the GCC developers are pleased to announce the
2115   release of GCC 7.5.
2116
2117   This release is a bug-fix release, containing fixes for regressions in
2118   GCC 7.4 relative to previous releases of GCC.
2119
2120Release History
2121
2122   GCC 7.5
2123          Nov 14, 2019 ([2]changes, [3]documentation)
2124
2125   GCC 7.4
2126          Dec 6, 2018 ([4]changes, [5]documentation)
2127
2128   GCC 7.3
2129          Jan 25, 2018 ([6]changes, [7]documentation)
2130
2131   GCC 7.2
2132          Aug 14, 2017 ([8]changes, [9]documentation)
2133
2134   GCC 7.1
2135          May 2, 2017 ([10]changes, [11]documentation)
2136
2137References and Acknowledgements
2138
2139   GCC used to stand for the GNU C Compiler, but since the compiler
2140   supports several other languages aside from C, it now stands for the
2141   GNU Compiler Collection.
2142
2143   A list of [12]successful builds is updated as new information becomes
2144   available.
2145
2146   The GCC developers would like to thank the numerous people that have
2147   contributed new features, improvements, bug fixes, and other changes as
2148   well as test results to GCC. This [13]amazing group of volunteers is
2149   what makes GCC successful.
2150
2151   For additional information about GCC please refer to the [14]GCC
2152   project web site or contact the [15]GCC development mailing list.
2153
2154   To obtain GCC please use [16]our mirror sites or [17]our version
2155   control system.
2156
2157
2158    For questions related to the use of GCC, please consult these web
2159    pages and the [18]GCC manuals. If that fails, the
2160    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
2161    web pages and the development of GCC are welcome on our developer
2162    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
2163    archives.
2164
2165   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
2166   distribution of this entire article is permitted in any medium,
2167   provided this notice is preserved.
2168
2169   These pages are [23]maintained by the GCC team. Last modified
2170   2020-01-14[24].
2171
2172References
2173
2174   1. http://www.gnu.org/
2175   2. http://gcc.gnu.org/gcc-7/changes.html
2176   3. http://gcc.gnu.org/onlinedocs/7.5.0/
2177   4. http://gcc.gnu.org/gcc-7/changes.html
2178   5. http://gcc.gnu.org/onlinedocs/7.4.0/
2179   6. http://gcc.gnu.org/gcc-7/changes.html
2180   7. http://gcc.gnu.org/onlinedocs/7.3.0/
2181   8. http://gcc.gnu.org/gcc-7/changes.html
2182   9. http://gcc.gnu.org/onlinedocs/7.2.0/
2183  10. http://gcc.gnu.org/gcc-7/changes.html
2184  11. http://gcc.gnu.org/onlinedocs/7.1.0/
2185  12. http://gcc.gnu.org/gcc-7/buildstat.html
2186  13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
2187  14. http://gcc.gnu.org/index.html
2188  15. mailto:gcc@gcc.gnu.org
2189  16. http://gcc.gnu.org/mirrors.html
2190  17. http://gcc.gnu.org/git.html
2191  18. https://gcc.gnu.org/onlinedocs/
2192  19. mailto:gcc-help@gcc.gnu.org
2193  20. mailto:gcc@gcc.gnu.org
2194  21. https://gcc.gnu.org/lists.html
2195  22. https://www.fsf.org/
2196  23. https://gcc.gnu.org/about.html
2197  24. http://validator.w3.org/check/referer
2198======================================================================
2199http://gcc.gnu.org/gcc-7/changes.html
2200                              GCC 7 Release Series
2201                        Changes, New Features, and Fixes
2202
2203   This page is a brief summary of some of the huge number of improvements
2204   in GCC 7. For more information, see the [1]Porting to GCC 7 page and
2205   the [2]full GCC documentation.
2206
2207Caveats
2208
2209     * GCC now uses [3]LRA (a new local register allocator) by default for
2210       new targets.
2211     * The non-standard C++0x type traits has_trivial_default_constructor,
2212       has_trivial_copy_constructor and has_trivial_copy_assign have been
2213       removed.
2214     * The libstdc++ [4]Profile Mode has been deprecated and will be
2215       removed in a future version.
2216     * The Cilk+ extensions to the C and C++ languages have been
2217       deprecated.
2218     * On ARM targets (arm*-*-*), [5]a bug introduced in GCC 5 that
2219       affects conformance to the procedure call standard (AAPCS) has been
2220       fixed. The bug affects some C++ code where class objects are passed
2221       by value to functions and could result in incorrect or inconsistent
2222       code being generated. This is an ABI change. If the option -Wpsabi
2223       is enabled (on by default) the compiler will emit a diagnostic note
2224       for code that might be affected.
2225
2226General Optimizer Improvements
2227
2228     * GCC 7 can determine the return value or range of return values of
2229       some calls to the sprintf family of functions and make it available
2230       to other optimization passes. Some calls to the snprintf function
2231       with a zero size argument can be folded into constants. This
2232       optimization is included in -O1 and can be selectively controlled
2233       by the -fprintf-return-value option.
2234     * A new store merging pass has been added. It merges constant stores
2235       to adjacent memory locations into fewer, wider, stores. It is
2236       enabled by the -fstore-merging option and at the -O2 optimization
2237       level or higher (and -Os).
2238     * A new code hoisting optimization has been added to the partial
2239       redundancy elimination pass. It attempts to move evaluation of
2240       expressions executed on all paths to the function exit as early as
2241       possible. This primarily helps improve code size, but can improve
2242       the speed of the generated code as well. It is enabled by the
2243       -fcode-hoisting option and at the -O2 optimization level or higher
2244       (and -Os).
2245     * A new interprocedural bitwise constant propagation optimization has
2246       been added, which propagates knowledge about which bits of
2247       variables are known to be zero (including pointer alignment
2248       information) across the call graph. It is enabled by the
2249       -fipa-bit-cp option if -fipa-cp is enabled as well, and is enabled
2250       at the -O2 optimization level and higher (and -Os). This
2251       optimization supersedes interprocedural alignment propagation of
2252       GCC 6, and therefore the option -fipa-cp-alignment is now
2253       deprecated and ignored.
2254     * A new interprocedural value range propagation optimization has been
2255       added, which propagates integral range information across the call
2256       graph when variable values can be proven to be within those ranges.
2257       It is enabled by the -fipa-vrp option and at the -O2 optimization
2258       level and higher (and -Os).
2259     * A new loop splitting optimization pass has been added. Certain
2260       loops which contain a condition that is always true on one side of
2261       the iteration space and always false on the other are split into
2262       two loops, such that each of the two new loops iterates on just one
2263       side of the iteration space and the condition does not need to be
2264       checked inside of the loop. It is enabled by the -fsplit-loops
2265       option and at the -O3 optimization level or higher.
2266     * The shrink-wrapping optimization can now separate portions of
2267       prologues and epilogues to improve performance if some of the work
2268       done traditionally by prologues and epilogues is not needed on
2269       certain paths. This is controlled by the -fshrink-wrap-separate
2270       option, enabled by default. It requires target support, which is
2271       currently only implemented in the PowerPC and AArch64 ports.
2272     * AddressSanitizer gained a new sanitization option,
2273       -fsanitize-address-use-after-scope, which enables sanitization of
2274       variables whose address is taken and used after a scope where the
2275       variable is defined:
2276
2277int
2278main (int argc, char **argv)
2279{
2280  char *ptr;
2281    {
2282      char my_char;
2283      ptr = &my_char;
2284    }
2285
2286  *ptr = 123;
2287  return *ptr;
2288}
2289
2290==28882==ERROR: AddressSanitizer: stack-use-after-scope on address 0x7fffb8dba99
22910 at pc 0x0000004006d5 bp 0x7fffb8dba960 sp 0x7fffb8dba958
2292WRITE of size 1 at 0x7fffb8dba990 thread T0
2293    #0 0x4006d4 in main /tmp/use-after-scope-1.c:10
2294    #1 0x7f9c71943290 in __libc_start_main (/lib64/libc.so.6+0x20290)
2295    #2 0x400739 in _start (/tmp/a.out+0x400739)
2296
2297Address 0x7fffb8dba990 is located in stack of thread T0 at offset 32 in frame
2298    #0 0x40067f in main /tmp/use-after-scope-1.c:3
2299
2300  This frame has 1 object(s):
2301    [32, 33) 'my_char' <== Memory access at offset 32 is inside this variable
2302
2303       The option is enabled by default with -fsanitize=address and
2304       disabled by default with -fsanitize=kernel-address. Compared to the
2305       LLVM compiler, where the option already exists, the implementation
2306       in the GCC compiler has some improvements and advantages:
2307          + Complex uses of gotos and case labels are properly handled and
2308            should not report any false positive or false negatives.
2309          + C++ temporaries are sanitized.
2310          + Sanitization can handle invalid memory stores that are
2311            optimized out by the LLVM compiler when optimization is
2312            enabled.
2313     * The -fsanitize=signed-integer-overflow suboption of the
2314       UndefinedBehavior Sanitizer now diagnoses arithmetic overflows even
2315       on arithmetic operations with generic vectors.
2316     * Version 5 of the [6]DWARF debugging information standard is
2317       supported through the -gdwarf-5 option. The DWARF version 4
2318       debugging information remains the default until consumers of
2319       debugging information are adjusted.
2320
2321New Languages and Language specific improvements
2322
2323   OpenACC support in C, C++, and Fortran continues to be maintained and
2324   improved. See the [7]OpenACC and [8]Offloading wiki pages for further
2325   information.
2326
2327  Ada
2328
2329     * On mainstream native platforms, Ada programs no longer require the
2330       stack to be made executable in order to run properly.
2331
2332  BRIG (HSAIL)
2333
2334   Support for processing BRIG 1.0 files was added in this release. BRIG
2335   is a binary format for HSAIL (Heterogeneous System Architecture
2336   Intermediate Language). The BRIG front end can be used for implementing
2337   HSAIL "finalizers" (compilation of HSAIL to a native ISA) for
2338   GCC-supported targets. An implementation of an HSAIL runtime library,
2339   libhsail-rt is also included.
2340
2341  C family
2342
2343     * New command-line options have been added for the C and C++
2344       compilers:
2345          + -Wimplicit-fallthrough warns when a switch case falls through.
2346            This warning has five different levels. The compiler is able
2347            to parse a wide range of fallthrough comments, depending on
2348            the level. It also handles control-flow statements, such as
2349            ifs. It's possible to suppress the warning by either adding a
2350            fallthrough comment, or by using a null statement:
2351            __attribute__ ((fallthrough)); (C, C++), or [[fallthrough]];
2352            (C++17), or [[gnu::fallthrough]]; (C++11/C++14). This warning
2353            is enabled by -Wextra.
2354          + -Wpointer-compare warns when a pointer is compared with a zero
2355            character constant. Such code is now invalid in C++11 and GCC
2356            rejects it. This warning is enabled by default.
2357          + -Wduplicated-branches warns when an if-else has identical
2358            branches.
2359          + -Wrestrict warns when an argument passed to a
2360            restrict-qualified parameter aliases with another argument.
2361          + -Wmemset-elt-size warns for memset calls, when the first
2362            argument references an array, and the third argument is a
2363            number equal to the number of elements of the array, but not
2364            the size of the array. This warning is enabled by -Wall.
2365          + -Wint-in-bool-context warns about suspicious uses of integer
2366            values where boolean values are expected. This warning is
2367            enabled by -Wall.
2368          + -Wswitch-unreachable warns when a switch statement has
2369            statements between the controlling expression and the first
2370            case label which will never be executed. This warning is
2371            enabled by default.
2372          + -Wexpansion-to-defined warns when defined is used outside #if.
2373            This warning is enabled by -Wextra or -Wpedantic.
2374          + -Wregister warns about uses of the register storage specifier.
2375            In C++17 this keyword has been removed and for C++17 this is a
2376            pedantic warning enabled by default. The warning is not
2377            emitted for the GNU Explicit Register Variables extension.
2378          + -Wvla-larger-than=N warns about unbounded uses of
2379            variable-length arrays, and about bounded uses of
2380            variable-length arrays whose bound can be larger than N bytes.
2381          + -Wduplicate-decl-specifier warns when a declaration has
2382            duplicate const, volatile, restrict or _Atomic specifier. This
2383            warning is enabled by -Wall.
2384     * GCC 6's C and C++ front ends were able to offer suggestions for
2385       misspelled field names:
2386
2387spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
2388you mean 'color'?
2389   return ptr->colour;
2390               ^~~~~~
2391
2392       GCC 7 greatly expands the scope of these suggestions. Firstly, it
2393       adds fix-it hints to such suggestions:
2394
2395spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
2396you mean 'color'?
2397   return ptr->colour;
2398               ^~~~~~
2399               color
2400
2401       The suggestions now cover many other things, such as misspelled
2402       function names:
2403
2404spellcheck-identifiers.c:11:3: warning: implicit declaration of function 'gtk_wi
2405dget_showall'; did you mean 'gtk_widget_show_all'? [-Wimplicit-function-declarat
2406ion]
2407   gtk_widget_showall (w);
2408   ^~~~~~~~~~~~~~~~~~
2409   gtk_widget_show_all
2410
2411       misspelled macro names and enum values:
2412
2413spellcheck-identifiers.cc:85:11: error: 'MAX_ITEM' undeclared here (not in a fun
2414ction); did you mean 'MAX_ITEMS'?
2415 int array[MAX_ITEM];
2416           ^~~~~~~~
2417           MAX_ITEMS
2418
2419       misspelled type names:
2420
2421spellcheck-typenames.c:7:14: error: unknown type name 'singed'; did you mean 'si
2422gned'?
2423 void test (singed char e);
2424            ^~~~~~
2425            signed
2426
2427       and, in the C front end, named initializers:
2428
2429test.c:7:20: error: 'struct s' has no member named 'colour'; did you mean 'color
2430'?
2431 struct s test = { .colour = 3 };
2432                    ^~~~~~
2433                    color
2434
2435     * The preprocessor can now offer suggestions for misspelled
2436       directives, e.g.:
2437
2438test.c:5:2: error:invalid preprocessing directive #endfi; did you mean #endif?
2439 #endfi
2440  ^~~~~
2441  endif
2442
2443     * Warnings about format strings now underline the pertinent part of
2444       the string, and can offer suggested fixes. In some cases, the
2445       pertinent argument is underlined.
2446
2447test.c:51:29: warning: format '%s' expects argument of type 'char *', but argume
2448nt 3 has type 'int' [-Wformat=]
2449   printf ("foo: %d  bar: %s baz: %d", 100, i + j, 102);
2450                          ~^                ~~~~~
2451                          %d
2452
2453     * The new -Wdangling-else command-line option has been split out of
2454       -Wparentheses and warns about dangling else.
2455     * The -Wshadow warning has been split into three variants.
2456       -Wshadow=global warns for any shadowing. This is the default when
2457       using -Wshadow without any argument. -Wshadow=local only warns for
2458       a local variable shadowing another local variable or parameter.
2459       -Wshadow=compatible-local only warns for a local variable shadowing
2460       another local variable or parameter whose type is compatible (in
2461       C++ compatible means that the type of the shadowing variable can be
2462       converted to that of the shadowed variable).
2463       The following example shows the different kinds of shadow warnings:
2464
2465enum operation { add, count };
2466struct container { int nr; };
2467
2468int
2469container_count (struct container c, int count)
2470{
2471  int r = 0;
2472  for (int count = 0; count > 0; count--)
2473    {
2474      struct container count = c;
2475      r += count.nr;
2476    }
2477  return r;
2478}
2479
2480       -Wshadow=compatible-local will warn for the parameter being
2481       shadowed with the same type:
2482
2483warn-test.c:8:12: warning: declaration of 'count' shadows a parameter [-Wshadow=
2484compatible-local]
2485   for (int count = 0; count > 0; count--)
2486            ^~~~~
2487warn-test.c:5:42: note: shadowed declaration is here
2488 container_count (struct container c, int count)
2489                                          ^~~~~
2490
2491       -Wshadow=local will warn for the above and for the shadowed
2492       declaration with incompatible type:
2493
2494warn-test.c:10:24: warning: declaration of 'count' shadows a previous local [-Ws
2495hadow=local]
2496       struct container count = c;
2497                        ^~~~~
2498warn-test.c:8:12: note: shadowed declaration is here
2499   for (int count = 0; count > 0; count--)
2500            ^~~~~
2501
2502       -Wshadow=global will warn for all of the above and the shadowing of
2503       the global declaration:
2504
2505warn-test.c:5:42: warning: declaration of 'count' shadows a global declaration [
2506-Wshadow]
2507 container_count (struct container c, int count)
2508                                          ^~~~~
2509warn-test.c:1:23: note: shadowed declaration is here
2510 enum operation { add, count };
2511                       ^~~~~
2512
2513     * GCC 7 contains a number of enhancements that help detect buffer
2514       overflow and other forms of invalid memory accesses.
2515          + The -Walloc-size-larger-than=size option detects calls to
2516            standard and user-defined memory allocation functions
2517            decorated with attribute alloc_size whose argument exceeds the
2518            specified size (PTRDIFF_MAX by default). The option also
2519            detects arithmetic overflow in the computation of the size in
2520            two-argument allocation functions like calloc where the total
2521            size is the product of the two arguments. Since calls with an
2522            excessive size cannot succeed they are typically the result of
2523            programming errors. Such bugs have been known to be the source
2524            of security vulnerabilities and a target of exploits.
2525            -Walloc-size-larger-than=PTRDIFF_MAX is included in -Wall.
2526            For example, the following call to malloc incorrectly tries to
2527            avoid passing a negative argument to the function and instead
2528            ends up unconditionally invoking it with an argument less than
2529            or equal to zero. Since after conversion to the type of the
2530            argument of the function (size_t) a negative argument results
2531            in a value in excess of the maximum PTRDIFF_MAX the call is
2532            diagnosed.
2533
2534void* f (int n)
2535{
2536  return malloc (n > 0 ? 0 : n);
2537}
2538
2539warning: argument 1 range [2147483648, 4294967295] exceeds maximum object size 2
2540147483647 [-Walloc-size-larger-than=]
2541
2542          + The -Walloc-zero option detects calls to standard and
2543            user-defined memory allocation functions decorated with
2544            attribute alloc_size with a zero argument. -Walloc-zero is not
2545            included in either -Wall or -Wextra and must be explicitly
2546            enabled.
2547          + The -Walloca option detects all calls to the alloca function
2548            in the program. -Walloca is not included in either -Wall or
2549            -Wextra and must be explicitly enabled.
2550          + The -Walloca-larger-than=size option detects calls to the
2551            alloca function whose argument either may exceed the specified
2552            size, or that is not known to be sufficiently constrained to
2553            avoid exceeding it. -Walloca-larger-than is not included in
2554            either -Wall or -Wextra and must be explicitly enabled.
2555            For example, compiling the following snippet with
2556            -Walloca-larger-than=1024 results in a warning because even
2557            though the code appears to call alloca only with sizes of 1kb
2558            and less, since n is signed, a negative value would result in
2559            a call to the function well in excess of the limit.
2560
2561void f (int n)
2562{
2563  char *d;
2564  if (n < 1025)
2565    d = alloca (n);
2566  else
2567    d = malloc (n);
25682569}
2570
2571warning: argument to 'alloca may be too large due to conversion from 'int' to 'l
2572ong unsigned int' [-Walloca-larger-than=]
2573
2574            In contrast, a call to alloca that isn't bounded at all such
2575            as in the following function will elicit the warning below
2576            regardless of the size argument to the option.
2577
2578void f (size_t n)
2579{
2580  char *d = alloca (n);
25812582}
2583
2584warning: unbounded use of 'alloca' [-Walloca-larger-than=]
2585
2586          + The -Wformat-overflow=level option detects certain and likely
2587            buffer overflow in calls to the sprintf family of formatted
2588            output functions. Although the option is enabled even without
2589            optimization it works best with -O2 and higher.
2590            For example, in the following snippet the call to sprintf is
2591            diagnosed because even though its output has been constrained
2592            using the modulo operation it could result in as many as three
2593            bytes if mday were negative. The solution is to either
2594            allocate a larger buffer or make sure the argument is not
2595            negative, for example by changing mday's type to unsigned or
2596            by making the type of the second operand of the modulo
2597            expression unsigned: 100U.
2598
2599void* f (int mday)
2600{
2601  char *buf = malloc (3);
2602  sprintf (buf, "%02i", mday % 100);
2603  return buf;
2604}
2605
2606warning: 'sprintf may write a terminating nul past the end of the destination [-
2607Wformat-overflow=]
2608note: 'sprintf' output between 3 and 4 bytes into a destination of size 3
2609
2610          + The -Wformat-truncation=level option detects certain and
2611            likely output truncation in calls to the snprintf family of
2612            formatted output functions. -Wformat-truncation=1 is included
2613            in -Wall and enabled without optimization but works best with
2614            -O2 and higher.
2615            For example, the following function attempts to format an
2616            integer between 0 and 255 in hexadecimal, including the 0x
2617            prefix, into a buffer of four characters. But since the
2618            function must always terminate output by the null character
2619            ('\0') such a buffer is only big enough to fit just one digit
2620            plus the prefix. Therefore the snprintf call is diagnosed. To
2621            avoid the warning either use a bigger buffer or handle the
2622            function's return value which indicates whether or not its
2623            output has been truncated.
2624
2625void f (unsigned x)
2626{
2627  char d[4];
2628  snprintf (d, sizeof d, "%#02x", x & 0xff);
26292630}
2631
2632warning: 'snprintf' output may be truncated before the last format character [-W
2633format-truncation=]
2634note: 'snprintf' output between 3 and 5 bytes into a destination of size 4
2635
2636          + The -Wnonnull option has been enhanced to detect a broader set
2637            of cases of passing null pointers to functions that expect a
2638            non-null argument (those decorated with attribute nonnull). By
2639            taking advantage of optimizations the option can detect many
2640            more cases of the problem than in prior GCC versions.
2641          + The -Wstringop-overflow=type option detects buffer overflow in
2642            calls to string handling functions like memcpy and strcpy. The
2643            option relies on [9]Object Size Checking and has an effect
2644            similar to defining the _FORTIFY_SOURCE macro.
2645            -Wstringop-overflow=2 is enabled by default.
2646            For example, in the following snippet, because the call to
2647            strncat specifies a maximum that allows the function to write
2648            past the end of the destination, it is diagnosed. To correct
2649            the problem and avoid the overflow the function should be
2650            called with a size of at most sizeof d - strlen(d) - 1.
2651
2652void f (const char *fname)
2653{
2654  char d[8];
2655  strncpy (d, "/tmp/", sizeof d);
2656  strncat (d, fname, sizeof d);
26572658}
2659
2660warning: specified bound 8 equals the size of the destination [-Wstringop-overfl
2661ow=]
2662
2663     * The <limits.h> header provided by GCC defines macros such as
2664       INT_WIDTH for the width in bits of integer types, if
2665       __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is
2666       included. The <stdint.h> header defines such macros as SIZE_WIDTH
2667       and INTMAX_WIDTH for the width of some standard typedef names for
2668       integer types, again if __STDC_WANT_IEC_60559_BFP_EXT__ is defined
2669       before the header is included; note that GCC's implementation of
2670       this header is only used for freestanding compilations, not hosted
2671       compilations, on most systems. These macros come from ISO/IEC TS
2672       18661-1:2014.
2673     * The <float.h> header provided by GCC defines the macro
2674       CR_DECIMAL_DIG, from ISO/IEC TS 18661-1:2014, if
2675       __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is
2676       included. This represents the number of decimal digits for which
2677       conversions between decimal character strings and binary formats,
2678       in both directions, are correctly rounded, and currently has the
2679       value of UINTMAX_MAX on all systems, reflecting that GCC's
2680       compile-time conversions are correctly rounded for any number of
2681       digits.
2682     * New __builtin_add_overflow_p, __builtin_sub_overflow_p,
2683       __builtin_mul_overflow_p built-in functions have been added. These
2684       work similarly to their siblings without the _p suffix, but do not
2685       actually store the result of the arithmetics anywhere, just return
2686       whether the operation would overflow. Calls to these built-ins with
2687       integer constant arguments evaluate to integer constants
2688       expressions.
2689       For example, in the following, c is assigned the result of a * b
2690       only if the multiplication does not overflow, otherwise it is
2691       assigned the value zero. The multiplication is performed at
2692       compile-time and without triggering a -Woverflow warning.
2693
2694enum {
2695  a = 12345678,
2696  b = 87654321,
2697  c = __builtin_mul_overflow_p (a, b, a) ? 0 : a * b
2698};
2699
2700  C
2701
2702     * The C front end now supports type names _FloatN for floating-point
2703       types with IEEE interchange formats and _FloatNx for floating-point
2704       types with IEEE extended formats. These type names come from
2705       ISO/IEC TS 18661-3:2015.
2706       The set of types supported depends on the target for which GCC is
2707       configured. Most targets support _Float32, _Float32x and _Float64.
2708       _Float128 is supported on targets where IEEE binary128 encoding was
2709       already supported as long double or __float128. _Float64x is
2710       supported on targets where a type with either binary128 or Intel
2711       extended precision format is available.
2712       Constants with these types are supported using suffixes fN, FN, fNx
2713       and FNx (e.g., 1.2f128 or 2.3F64x). Macros such as FLT128_MAX are
2714       defined in <float.h> if __STDC_WANT_IEC_60559_TYPES_EXT__ is
2715       defined before it is included.
2716       These new types are always distinct from each other and from float,
2717       double and long double, even if they have the same encoding.
2718       Complex types such as _Complex _Float128 are also supported.
2719       Type-generic built-in functions such as __builtin_isinf support the
2720       new types, and the following type-specific built-in functions have
2721       versions (suffixed fN or fNx) for the new types:
2722       __builtin_copysign, __builtin_fabs, __builtin_huge_val,
2723       __builtin_inf, __builtin_nan, __builtin_nans.
2724     * Compilation with -fopenmp is now compatible with the C11 _Atomic
2725       keyword.
2726
2727  C++
2728
2729     * The C++ front end has experimental support for all of the current
2730       C++17 draft with the -std=c++1z or -std=gnu++1z flags, including if
2731       constexpr, class template argument deduction, auto template
2732       parameters, and structured bindings. For a full list of new
2733       features, see [10]the C++ status page.
2734     * C++17 support for new of over-aligned types can be enabled in other
2735       modes with the -faligned-new flag.
2736     * The C++17 evaluation order requirements can be selected in other
2737       modes with the -fstrong-eval-order flag, or disabled in C++17 mode
2738       with -fno-strong-eval-order.
2739     * The default semantics of inherited constructors has changed in all
2740       modes, following [11]P0136. Essentially, overload resolution
2741       happens as if calling the inherited constructor directly, and the
2742       compiler fills in construction of the other bases and members as
2743       needed. Most uses should not need any changes. The old behavior can
2744       be restored with -fno-new-inheriting-ctors, or -fabi-version less
2745       than 11.
2746     * The resolution of DR 150 on matching of template template
2747       parameters, allowing default template arguments to make a template
2748       match a parameter, is currently enabled by default in C++17 mode
2749       only. The default can be overridden with -f{no-,}new-ttp-matching.
2750     * The C++ front end will now provide fix-it hints for some missing
2751       semicolons, allowing for automatic fixes by IDEs:
2752
2753test.cc:4:11: error: expected ';' after class definition
2754 class a {}
2755           ^
2756           ;
2757
2758     * -Waligned-new has been added to the C++ front end. It warns about
2759       new of type with extended alignment without -faligned-new.
2760
2761    Runtime Library (libstdc++)
2762
2763     * The type of exception thrown by iostreams, std::ios_base::failure,
2764       now uses the [12]cxx11 ABI.
2765     * Experimental support for C++17, including the following new
2766       features:
2767          + std::string_view;
2768          + std::any, std::optional, and std::variant;
2769          + std::invoke, std::is_invocable, std::is_nothrow_invocable, and
2770            invoke_result;
2771          + std::is_swappable, and std::is_nothrow_swappable;
2772          + std::apply, and std::make_from_tuple;
2773          + std::void_t, std::bool_constant, std::conjunction,
2774            std::disjunction, and std::negation;
2775          + Variable templates for type traits;
2776          + Mathematical Special Functions;
2777          + std::chrono::floor, std::chrono::ceil, std::chrono::round, and
2778            std::chrono::abs;
2779          + std::clamp, std::gcd, std::lcm, 3-dimensional std::hypot;
2780          + std::scoped_lock, std::shared_mutex,
2781            std::atomic<T>::is_always_lock_free;
2782          + std::sample, std::default_searcher, std::boyer_moore_searcher
2783            and std::boyer_moore_horspool_searcher;
2784          + Extraction and re-insertion of map and set nodes, try_emplace
2785            members for maps, and functions for accessing containers
2786            std::size, std::empty, and std::data;
2787          + std::shared_ptr support for arrays,
2788            std::shared_ptr<T>::weak_type,
2789            std::enable_shared_from_this<T>::weak_from_this(), and
2790            std::owner_less<void>;
2791          + std::byte;
2792          + std::as_const, std::not_fn,
2793            std::has_unique_object_representations, constexpr
2794            std::addressof.
2795       Thanks to Daniel Krügler, Tim Shen, Edward Smith-Rowland, and Ville
2796       Voutilainen for work on the C++17 support.
2797     * A new power-of-two rehashing policy for use with the _Hashtable
2798       internals, thanks to François Dumont.
2799
2800  Fortran
2801
2802     * Support for a number of extensions for compatibility with legacy
2803       code with new flags:
2804          + -fdec-structure Support for DEC STRUCTURE and UNION
2805          + -fdec-intrinsic-ints Support for new integer intrinsics with
2806            B/I/J/K prefixes such as BABS, JIAND...
2807          + -fdec-math Support for additional math intrinsics, including
2808            COTAN and degree-valued trigonometric functions such as TAND,
2809            ASIND...
2810          + -fdec Enable the -fdec-* family of extensions.
2811     * New flag -finit-derived to allow default initialization of
2812       derived-type variables.
2813     * Improved DO loops with step equal to 1 or -1, generates faster code
2814       without a loop preheader. A new warning, -Wundefined-do-loop, warns
2815       when a loop iterates either to HUGE(i) (with step equal to 1), or
2816       to -HUGE(i) (with step equal to -1). Invalid behavior can be caught
2817       at run time with -fcheck=do enabled:
2818
2819program test
2820  implicit none
2821  integer(1) :: i
2822  do i = -HUGE(i)+10, -HUGE(i)-1, -1
2823    print *, i
2824  end do
2825end program test
2826
2827At line 8 of file do_check_12.f90
2828Fortran runtime error: Loop iterates infinitely
2829
2830     * Version 4.5 of the [13]OpenMP specification is now partially
2831       supported in the Fortran compiler; the largest missing item is
2832       structure element mapping.
2833     * User-defined derived-type input/output (UDTIO) is added.
2834     * Derived type coarrays with allocatable and pointer components are
2835       partially supported.
2836     * Non-constant stop codes and error stop codes (Fortran 2015
2837       feature).
2838     * Derived types with allocatable components of recursive type.
2839     * Intrinsic assignment to polymorphic variables.
2840     * Improved submodule support.
2841     * Improved diagnostics (polymorphic results in pure functions).
2842     * Coarray: Support for failed images (Fortan 2015 feature).
2843
2844  Go
2845
2846     * GCC 7 provides a complete implementation of the Go 1.8.1 user
2847       packages.
2848     * Compared to the Go 1.8.1 toolchain, the garbage collector is more
2849       conservative and less concurrent.
2850     * Escape analysis is available for experimental use via the
2851       -fgo-optimize-allocs option. The -fgo-debug-escape prints
2852       information useful for debugging escape analysis choices.
2853
2854  Java (GCJ)
2855
2856   The GCC Java front end and associated libjava runtime library have been
2857   removed from GCC.
2858
2859libgccjit
2860
2861   The libgccjit API gained support for marking calls as requiring
2862   tail-call optimization via a new entry point:
2863   [14]gcc_jit_rvalue_set_bool_require_tail_call.
2864
2865   libgccjit performs numerous checks at the API boundary, but if these
2866   succeed, it previously ignored errors and other diagnostics emitted
2867   within the core of GCC, and treated the compile of a gcc_jit_context as
2868   having succeeded. As of GCC 7 it now ensures that if any diagnostics
2869   are emitted, they are visible from the libgccjit API, and that the the
2870   context is flagged as having failed.
2871
2872New Targets and Target Specific Improvements
2873
2874  AArch64
2875
2876     * GCC has been updated to the latest revision of the procedure call
2877       standard (AAPCS64) to provide support for parameter passing when
2878       data types have been over-aligned.
2879     * The ARMv8.3-A architecture is now supported. It can be used by
2880       specifying the -march=armv8.3-a option.
2881     * The option -msign-return-address= is supported to enable return
2882       address protection using ARMv8.3-A Pointer Authentication
2883       Extensions. For more information on the arguments accepted by this
2884       option, please refer to [15]AArch64-Options.
2885     * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point
2886       Extensions are now supported. They can be used by specifying the
2887       -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit
2888       Floating-Point Extensions introduce new half-precision data
2889       processing floating-point instructions.
2890     * Support has been added for the following processors (GCC
2891       identifiers in parentheses): ARM Cortex-A73 (cortex-a73), Broadcom
2892       Vulcan (vulcan), Cavium ThunderX CN81xx (thunderxt81), Cavium
2893       ThunderX CN83xx (thunderxt83), Cavium ThunderX CN88xx
2894       (thunderxt88), Cavium ThunderX CN88xx pass 1.x (thunderxt88p1),
2895       Cavium ThunderX 2 CN99xx (thunderx2t99), Qualcomm Falkor (falkor).
2896       The GCC identifiers can be used as arguments to the -mcpu or -mtune
2897       options, for example: -mcpu=cortex-a73 or -mtune=vulcan or as
2898       arguments to the equivalent target attributes and pragmas.
2899
2900  ARC
2901
2902     * Added support for ARC HS and ARC EM processors.
2903     * Added support for ARC EM variation found in Intel QuarkSE SoCs.
2904     * Added support for NPS400 ARC700 based CPUs.
2905     * Thread Local Storage is now supported by ARC CPUs.
2906     * Fixed errors for ARC600 when using 32x16 multiplier option.
2907     * Fixed PIE for ARC CPUs.
2908     * New CPU templates are supported via multilib.
2909
2910  ARM
2911
2912     * Support for the ARMv5 and ARMv5E architectures has been deprecated
2913       (which have no known implementations) and will be removed in a
2914       future GCC release. Note that ARMv5T, ARMv5TE and ARMv5TEJ
2915       architectures remain supported. The values armv5 and armv5e of
2916       -march are thus deprecated.
2917     * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point
2918       Extensions are now supported. They can be used by specifying the
2919       -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit
2920       Floating-Point Extensions introduce new half-precision data
2921       processing floating-point instructions.
2922     * The ARMv8-M architecture is now supported in its two architecture
2923       profiles: ARMv8-M Baseline and ARMv8-M Mainline with its DSP and
2924       Floating-Point Extensions. They can be used by specifying the
2925       -march=armv8-m.base, armv8-m.main or armv8-m.main+dsp options.
2926     * Support has been added for the following processors (GCC
2927       identifiers in parentheses): ARM Cortex-A73 (cortex-a73), ARM
2928       Cortex-M23 (cortex-m23) and ARM Cortex-M33 (cortex-m33). The GCC
2929       identifiers can be used as arguments to the -mcpu or -mtune
2930       options, for example: -mcpu=cortex-a73 or -mtune=cortex-m33.
2931     * A new command-line option -mpure-code has been added. It does not
2932       allow constant data to be placed in code sections. This option is
2933       only available when generating non-PIC code for ARMv7-M targets.
2934     * Support for the ACLE Coprocessor Intrinsics has been added. This
2935       enables the generation of coprocessor instructions through the use
2936       of intrinsics such as cdp, ldc, and others.
2937     * The configure option --with-multilib-list now accepts the value
2938       rmprofile to build multilib libraries for a range of embedded
2939       targets. See our [16]installation instructions for details.
2940
2941  AVR
2942
2943     * On the reduced Tiny cores, the progmem [17]variable attribute is
2944       now properly supported. Respective read-only variables are located
2945       in flash memory in section .progmem.data. No special code is needed
2946       to access such variables; the compiler automatically adds an offset
2947       of 0x4000 to all addresses, which is needed to access variables in
2948       flash memory. As opposed to ordinary cores where it is sufficient
2949       to specify the progmem attribute with definitions, on the reduced
2950       Tiny cores the attribute also has to be specified with (external)
2951       declarations:
2952
2953extern const int array[] __attribute__((__progmem__));
2954
2955int get_value2 (void)
2956{
2957  /* Access via addresses array + 0x4004 and array + 0x4005. */
2958  return array[2];
2959}
2960
2961const int* get_address (unsigned idx)
2962{
2963  /* Returns array + 0x4000 + 2 * idx. */
2964  return &array[idx];
2965}
2966
2967     * A new command-line option -Wmisspelled-isr has been added. It turns
2968       off — or turns into errors — warnings that are reported for
2969       interrupt service routines (ISRs) which don't follow AVR-LibC's
2970       naming convention of prefixing ISR names with __vector.
2971     * __builtin_avr_nops(n) is a new [18]built-in function that inserts n
2972       NOP instructions into the instruction stream. n must be a value
2973       known at compile time.
2974
2975  IA-32/x86-64
2976
2977     * Support for the AVX-512 Fused Multiply Accumulation Packed Single
2978       precision (4FMAPS), AVX-512 Vector Neural Network Instructions Word
2979       variable precision (4VNNIW), AVX-512 Vector Population Count
2980       (VPOPCNTDQ) and Software Guard Extensions (SGX) ISA extensions has
2981       been added.
2982
2983  NVPTX
2984
2985     * OpenMP target regions can now be offloaded to NVidia PTX GPGPUs.
2986       See the [19]Offloading Wiki on how to configure it.
2987
2988  PowerPC / PowerPC64 / RS6000
2989
2990     * The PowerPC port now uses LRA by default.
2991     * GCC now diagnoses inline assembly that clobbers register r2. This
2992       has always been invalid code, and is no longer quietly tolerated.
2993     * The PowerPC port's support for ISA 3.0 (-mcpu=power9) has been
2994       enhanced to generate more of the new instructions by default, and
2995       to provide more built-in functions to generate code for other new
2996       instructions.
2997     * The configuration option --enable-gnu-indirect-function is now
2998       enabled by default on PowerPC GNU/Linux builds.
2999     * The PowerPC port will now allow 64-bit and 32-bit integer types to
3000       be allocated to the VSX vector registers (ISA 2.06 and above). In
3001       addition, on ISA 3.0, 16-bit and 8-bit integer types can be
3002       allocated in the vector registers. Previously, only 64-bit integer
3003       types were allowed in the traditional floating point registers.
3004     * New options -mstack-protector-guard=global,
3005       -mstack-protector-guard=tls, -mstack-protector-guard-reg=, and
3006       -mstack-protector-guard-offset= change how the stack protector gets
3007       the value to use as canary.
3008
3009  S/390, System z, IBM z Systems, IBM Z
3010
3011     * Support for the IBM z14 processor has been added. When using the
3012       -march=z14 option, the compiler will generate code making use of
3013       the new instructions introduced with the vector enhancement
3014       facility and the miscellaneous instruction extension facility 2.
3015       The -mtune=z14 option enables z14 specific instruction scheduling
3016       without making use of new instructions.
3017     * Builtins for the new vector instructions have been added and can be
3018       enabled using the -mzvector option.
3019
3020  RISC-V
3021
3022     * Support for the RISC-V instruction set has been added.
3023
3024  RX
3025
3026   Basic support for atomic built-in function has been added. It is
3027   currently implemented by flipping interrupts off and on as needed.
3028
3029  SH
3030
3031     * Support for SH5/SH64 has been removed.
3032     * Improved utilization of delay slots on SH2A.
3033     * Improved utilization of zero-displacement conditional branches.
3034     * The following deprecated options have been removed
3035          + -mcbranchdi
3036          + -mcmpeqdi
3037          + -minvalid-symbols
3038          + -msoft-atomic
3039          + -mspace
3040          + -madjust-unroll
3041     * Support for the following SH2A instructions has been added
3042          + mov.b @-Rm,R0
3043          + mov.w @-Rm,R0
3044          + mov.l @-Rm,R0
3045          + mov.b R0,@Rn+
3046          + mov.w R0,@Rn+
3047          + mov.l R0,@Rn+
3048
3049  SPARC
3050
3051     * The SPARC port now uses LRA by default.
3052     * Support for the new Subtract-Extended-with-Carry instruction
3053       available in SPARC M7 (Niagara 7) has been added.
3054
3055Operating Systems
3056
3057  AIX
3058
3059     * Visibility support has been enabled for AIX 7.1 and above.
3060
3061  Fuchsia
3062
3063     * Support has been added for the [20]Fuchsia OS.
3064
3065  RTEMS
3066
3067     * The ABI changes on ARM so that no short enums are used by default.
3068
3069Other significant improvements
3070
3071     * -fverbose-asm previously emitted information on the meanings of
3072       assembly expressions. This has been extended so that it now also
3073       prints comments showing the source lines that correspond to the
3074       assembly, making it easier to read the generated assembly
3075       (especially with larger functions). For example, given this C
3076       source file:
3077
3078int test (int n)
3079{
3080  int i;
3081  int total = 0;
3082
3083  for (i = 0; i < n; i++)
3084    total += i * i;
3085  return total;
3086}
3087
3088       -fverbose-asm now gives output similar to this for the function
3089       body (when compiling for x86_64, with -Os):
3090
3091       .text
3092       .globl  test
3093       .type   test, @@function
3094test:
3095.LFB0:
3096       .cfi_startproc
3097# example.c:4:   int total = 0;
3098       xorl    %eax, %eax      # <retval>
3099# example.c:6:   for (i = 0; i < n; i++)
3100       xorl    %edx, %edx      # i
3101.L2:
3102# example.c:6:   for (i = 0; i < n; i++)
3103       cmpl    %edi, %edx      # n, i
3104       jge     .L5     #,
3105# example.c:7:     total += i * i;
3106       movl    %edx, %ecx      # i, tmp92
3107       imull   %edx, %ecx      # i, tmp92
3108# example.c:6:   for (i = 0; i < n; i++)
3109       incl    %edx    # i
3110# example.c:7:     total += i * i;
3111       addl    %ecx, %eax      # tmp92, <retval>
3112       jmp     .L2     #
3113.L5:
3114# example.c:10: }
3115       ret
3116       .cfi_endproc
3117
3118     * Two new options have been added for printing fix-it hints:
3119          + -fdiagnostics-parseable-fixits allows for fix-it hints to be
3120            emitted in a machine-readable form, suitable for consumption
3121            by IDEs. For example, given:
3122
3123spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
3124you mean 'color'?
3125   return ptr->colour;
3126               ^~~~~~
3127               color
3128
3129            it will emit:
3130
3131fix-it:"spellcheck-fields.cc":{52:13-52:19}:"color"
3132
3133          + -fdiagnostics-generate-patch will print a patch in "unified"
3134            format after any diagnostics are printed, showing the result
3135            of applying all fix-it hints. For the above example it would
3136            emit:
3137
3138--- spellcheck-fields.cc
3139+++ spellcheck-fields.cc
3140@@ -49,5 +49,5 @@
3141
3142 color get_color(struct s *ptr)
3143 {
3144-  return ptr->colour;
3145+  return ptr->color;
3146 }
3147
3148     * The gcc and g++ driver programs will now provide suggestions for
3149       misspelled arguments to command-line options.
3150
3151$ gcc -c test.c -ftls-model=global-dinamic
3152gcc: error: unknown TLS model 'global-dinamic'
3153gcc: note: valid arguments to '-ftls-model=' are: global-dynamic initial-exec lo
3154cal-dynamic local-exec; did you mean 'global-dynamic'?
3155
3156     * The compiler will now provide suggestions for misspelled
3157       parameters.
3158
3159$ gcc -c test.c --param max-early-inliner-iteration=3
3160cc1: error: invalid --param name 'max-early-inliner-iteration'; did you mean 'ma
3161x-early-inliner-iterations'?
3162
3163     * Profile-guided optimization (PGO) instrumentation, as well as test
3164       coverage (GCOV), can newly instrument constructors (functions marks
3165       with __attribute__((constructor))), destructors and C++
3166       constructors (and destructors) of classes that are used as the type
3167       of a global variable.
3168     * A new option -fprofile-update=atomic prevents creation of corrupted
3169       profiles created during an instrumentation run (-fprofile=generate)
3170       of an application. The downside of the option is a speed penalty.
3171       Providing -pthread on the command line selects atomic profile
3172       updating (when supported by the target).
3173     * GCC's already extensive testsuite has gained some new capabilities,
3174       to further improve the reliability of the compiler:
3175          + GCC now has an internal unit-testing API and a suite of tests
3176            for programmatic self-testing of subsystems.
3177          + GCC's C front end has been extended so that it can parse dumps
3178            of GCC's internal representations, allowing for DejaGnu tests
3179            that more directly exercise specific optimization passes. This
3180            covers both the [21]GIMPLE representation (for testing
3181            higher-level optimizations) and the [22]RTL representation,
3182            allowing for more direct testing of lower-level details, such
3183            as register allocation and instruction selection.
3184
3185GCC 7.1
3186
3187   This is the [23]list of problem reports (PRs) from GCC's bug tracking
3188   system that are known to be fixed in the 7.1 release. This list might
3189   not be complete (that is, it is possible that some PRs that have been
3190   fixed are not listed here).
3191
3192GCC 7.2
3193
3194   This is the [24]list of problem reports (PRs) from GCC's bug tracking
3195   system that are known to be fixed in the 7.2 release. This list might
3196   not be complete (that is, it is possible that some PRs that have been
3197   fixed are not listed here).
3198
3199  Target Specific Changes
3200
3201    SPARC
3202
3203     * Support for the SPARC M8 processor has been added.
3204     * The switches -mfix-ut700 and -mfix-gr712rc have been added to work
3205       around an erratum in LEON3FT processors.
3206     * Use of the Floating-point Multiply Single to Double (FsMULd)
3207       instruction can now be controlled by the -mfsmuld and -fno-fsmuld
3208       options.
3209
3210  Operating Systems
3211
3212    RTEMS
3213
3214     * The Ada run-time support uses now thread-local storage (TLS).
3215     * Support for RISC-V has been added.
3216     * Support for 64-bit PowerPC using the ELFv2 ABI with 64-bit long
3217       double has been added.
3218
3219GCC 7.3
3220
3221   This is the [25]list of problem reports (PRs) from GCC's bug tracking
3222   system that are known to be fixed in the 7.3 release. This list might
3223   not be complete (that is, it is possible that some PRs that have been
3224   fixed are not listed here).
3225
3226  Target Specific Changes
3227
3228    SPARC
3229
3230     * Workarounds for the four [26]LEON3FT errata GRLIB-TN-0010..0013
3231       have been added. Relevant errata are activated by the target
3232       specific -mfix-ut699, -mfix-ut700 and -mfix-gr712rc switches.
3233
3234  Operating Systems
3235
3236    RTEMS
3237
3238     * Support has been added for Epiphany target.
3239
3240GCC 7.4
3241
3242   This is the [27]list of problem reports (PRs) from GCC's bug tracking
3243   system that are known to be fixed in the 7.4 release. This list might
3244   not be complete (that is, it is possible that some PRs that have been
3245   fixed are not listed here).
3246
3247GCC 7.5
3248
3249   This is the [28]list of problem reports (PRs) from GCC's bug tracking
3250   system that are known to be fixed in the 7.5 release. This list might
3251   not be complete (that is, it is possible that some PRs that have been
3252   fixed are not listed here).
3253
3254
3255    For questions related to the use of GCC, please consult these web
3256    pages and the [29]GCC manuals. If that fails, the
3257    [30]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3258    web pages and the development of GCC are welcome on our developer
3259    list at [31]gcc@gcc.gnu.org. All of [32]our lists have public
3260    archives.
3261
3262   Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and
3263   distribution of this entire article is permitted in any medium,
3264   provided this notice is preserved.
3265
3266   These pages are [34]maintained by the GCC team. Last modified
3267   2019-11-28[35].
3268
3269References
3270
3271   1. http://gcc.gnu.org/gcc-7/porting_to.html
3272   2. http://gcc.gnu.org/onlinedocs/index.html#current
3273   3. https://gcc.gnu.org/wiki/LRAIsDefault
3274   4. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/manual/profile_mode.html
3275   5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77728
3276   6. http://www.dwarfstd.org/Download.php
3277   7. https://gcc.gnu.org/wiki/OpenACC
3278   8. https://gcc.gnu.org/wiki/Offloading
3279   9. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/Object-Size-Checking.html
3280  10. https://gcc.gnu.org/projects/cxx-status.html#cxx1z
3281  11. http://wg21.link/p0136
3282  12. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/using_dual_abi.html
3283  13. https://www.openmp.org/specifications/
3284  14. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/jit/topics/expressions.html#gcc_jit_rvalue_set_bool_require_tail_call
3285  15. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AArch64-Options.html#AArch64-Options
3286  16. https://gcc.gnu.org/install/configure.html
3287  17. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Variable-Attributes.html
3288  18. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Built-in-Functions.html
3289  19. https://gcc.gnu.org/wiki/Offloading
3290  20. https://fuchsia.googlesource.com/
3291  21. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/GIMPLE-Tests.html
3292  22. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/RTL-Tests.html
3293  23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.0
3294  24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.2
3295  25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.3
3296  26. https://www.gaisler.com/index.php/information/app-tech-notes
3297  27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.4
3298  28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.5
3299  29. https://gcc.gnu.org/onlinedocs/
3300  30. mailto:gcc-help@gcc.gnu.org
3301  31. mailto:gcc@gcc.gnu.org
3302  32. https://gcc.gnu.org/lists.html
3303  33. https://www.fsf.org/
3304  34. https://gcc.gnu.org/about.html
3305  35. http://validator.w3.org/check/referer
3306======================================================================
3307http://gcc.gnu.org/gcc-6/index.html
3308                              GCC 6 Release Series
3309
3310   (This release series is no longer supported.)
3311
3312   October 26, 2018
3313
3314   The [1]GNU project and the GCC developers are pleased to announce the
3315   release of GCC 6.5.
3316
3317   This release is a bug-fix release, containing fixes for regressions in
3318   GCC 6.4 relative to previous releases of GCC.
3319
3320Release History
3321
3322   GCC 6.5
3323          October 26, 2018 ([2]changes, [3]documentation)
3324
3325   GCC 6.4
3326          July 4, 2017 ([4]changes, [5]documentation)
3327
3328   GCC 6.3
3329          December 21, 2016 ([6]changes, [7]documentation)
3330
3331   GCC 6.2
3332          August 22, 2016 ([8]changes, [9]documentation)
3333
3334   GCC 6.1
3335          April 27, 2016 ([10]changes, [11]documentation)
3336
3337References and Acknowledgements
3338
3339   GCC used to stand for the GNU C Compiler, but since the compiler
3340   supports several other languages aside from C, it now stands for the
3341   GNU Compiler Collection.
3342
3343   A list of [12]successful builds is updated as new information becomes
3344   available.
3345
3346   The GCC developers would like to thank the numerous people that have
3347   contributed new features, improvements, bug fixes, and other changes as
3348   well as test results to GCC. This [13]amazing group of volunteers is
3349   what makes GCC successful.
3350
3351   For additional information about GCC please refer to the [14]GCC
3352   project web site or contact the [15]GCC development mailing list.
3353
3354   To obtain GCC please use [16]our mirror sites or [17]our version
3355   control system.
3356
3357
3358    For questions related to the use of GCC, please consult these web
3359    pages and the [18]GCC manuals. If that fails, the
3360    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3361    web pages and the development of GCC are welcome on our developer
3362    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
3363    archives.
3364
3365   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
3366   distribution of this entire article is permitted in any medium,
3367   provided this notice is preserved.
3368
3369   These pages are [23]maintained by the GCC team. Last modified
3370   2020-01-14[24].
3371
3372References
3373
3374   1. http://www.gnu.org/
3375   2. http://gcc.gnu.org/gcc-6/changes.html
3376   3. http://gcc.gnu.org/onlinedocs/6.5.0/
3377   4. http://gcc.gnu.org/gcc-6/changes.html
3378   5. http://gcc.gnu.org/onlinedocs/6.4.0/
3379   6. http://gcc.gnu.org/gcc-6/changes.html
3380   7. http://gcc.gnu.org/onlinedocs/6.3.0/
3381   8. http://gcc.gnu.org/gcc-6/changes.html
3382   9. http://gcc.gnu.org/onlinedocs/6.2.0/
3383  10. http://gcc.gnu.org/gcc-6/changes.html
3384  11. http://gcc.gnu.org/onlinedocs/6.1.0/
3385  12. http://gcc.gnu.org/gcc-6/buildstat.html
3386  13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
3387  14. http://gcc.gnu.org/index.html
3388  15. mailto:gcc@gcc.gnu.org
3389  16. http://gcc.gnu.org/mirrors.html
3390  17. http://gcc.gnu.org/git.html
3391  18. https://gcc.gnu.org/onlinedocs/
3392  19. mailto:gcc-help@gcc.gnu.org
3393  20. mailto:gcc@gcc.gnu.org
3394  21. https://gcc.gnu.org/lists.html
3395  22. https://www.fsf.org/
3396  23. https://gcc.gnu.org/about.html
3397  24. http://validator.w3.org/check/referer
3398======================================================================
3399http://gcc.gnu.org/gcc-6/changes.html
3400                              GCC 6 Release Series
3401                        Changes, New Features, and Fixes
3402
3403   This page is a brief summary of some of the huge number of improvements
3404   in GCC 6. For more information, see the [1]Porting to GCC 6 page and
3405   the [2]full GCC documentation.
3406
3407Caveats
3408
3409     * The default mode for C++ is now -std=gnu++14 instead of
3410       -std=gnu++98.
3411     * Support for a number of older systems and recently unmaintained or
3412       untested target ports of GCC has been declared obsolete in GCC 6.
3413       Unless there is activity to revive them, the next release of GCC
3414       will have their sources permanently removed.
3415       The following ports for individual systems on particular
3416       architectures have been obsoleted:
3417          + SH5 / SH64 (sh64-*-*) as announced [3]here.
3418     * The AVR port requires binutils version 2.26.1 or later for the fix
3419       for [4]PR71151 to work.
3420     * The GCC 6.5 release has an accidental ABI incompatibility for
3421       nested std::pair objects, for more details see [5]PR 87822. The bug
3422       causes a layout change for pairs where the first member is also a
3423       pair, e.g. std::pair<std::pair<X, Y>, Z>. The GCC 6 release series
3424       is closed so the bug in GCC 6.5 will not be fixed upstream, but
3425       there is a patch in the bug report to allow it to be fixed by
3426       anybody packaging GCC 6.5 or installing it themselves.
3427
3428General Optimizer Improvements
3429
3430     * UndefinedBehaviorSanitizer gained a new sanitization option,
3431       -fsanitize=bounds-strict, which enables strict checking of array
3432       bounds. In particular, it enables -fsanitize=bounds as well as
3433       instrumentation of flexible array member-like arrays.
3434     * Type-based alias analysis now disambiguates accesses to different
3435       pointers. This improves precision of the alias oracle by about
3436       20-30% on higher-level C++ programs. Programs doing invalid type
3437       punning of pointer types may now need -fno-strict-aliasing to work
3438       correctly.
3439     * Alias analysis now correctly supports the weakref and alias
3440       attributes. This allows accessing both a variable and its alias in
3441       one translation unit which is common with link-time optimization.
3442     * Value range propagation now assumes that the this pointer in C++
3443       member functions is non-null. This eliminates common null pointer
3444       checks but also breaks some non-conforming code-bases (such as
3445       Qt-5, Chromium, KDevelop). As a temporary work-around
3446       -fno-delete-null-pointer-checks can be used. Wrong code can be
3447       identified by using -fsanitize=undefined.
3448     * Link-time optimization improvements:
3449          + warning and error attributes are now correctly preserved by
3450            declaration linking and thus -D_FORTIFY_SOURCE=2 is now
3451            supported with -flto.
3452          + Type merging was fixed to handle C and Fortran
3453            interoperability rules as defined by the Fortran 2008 language
3454            standard.
3455            As an exception, CHARACTER(KIND=C_CHAR) is not inter-operable
3456            with char in all cases because it is an array while char is
3457            scalar. INTEGER(KIND=C_SIGNED_CHAR) should be used instead. In
3458            general, this inter-operability cannot be implemented, for
3459            example on targets where the argument passing convention for
3460            arrays differs from scalars.
3461          + More type information is now preserved at link time, reducing
3462            the loss of accuracy of the type-based alias analysis compared
3463            to builds without link-time optimization.
3464          + Invalid type punning on global variables and declarations is
3465            now reported with -Wodr-type-mismatch.
3466          + The size of LTO object files was reduced by about 11%
3467            (measured by compiling Firefox 46.0).
3468          + Link-time parallelization (enabled using -flto=n) was
3469            significantly improved by decreasing the size of streamed data
3470            when partitioning programs. The size of streamed IL while
3471            compiling Firefox 46.0 was reduced by 66%.
3472          + The linker plugin was extended to pass information about the
3473            type of binary produced to the GCC back end. (That can also be
3474            controlled manually by -flinker-output.) This makes it
3475            possible to properly configure the code generator and support
3476            incremental linking. Incremental linking of LTO objects by gcc
3477            -r is now supported for plugin-enabled setups.
3478            There are two ways to perform incremental linking:
3479              1. Linking by ld -r will result in an object file with all
3480                 sections from individual object files mechanically
3481                 merged. This delays the actual link-time optimization to
3482                 the final linking step and thus permits whole program
3483                 optimization. Linking the final binary with such object
3484                 files is however slower.
3485              2. Linking by gcc -r will lead to link-time optimization and
3486                 emit the final binary into the object file. Linking such
3487                 an object file is fast but avoids any benefits from whole
3488                 program optimization.
3489            GCC 7 will support incremental link-time optimization with gcc
3490            -r.
3491     * Inter-procedural optimization improvements:
3492          + Basic jump threading is now performed before profile
3493            construction and inline analysis, resulting in more realistic
3494            size and time estimates that drive the heuristics of the
3495            inliner and function cloning passes.
3496          + Function cloning now more aggressively eliminates unused
3497            function parameters.
3498
3499New Languages and Language specific improvements
3500
3501   Compared to GCC 5, the GCC 6 release series includes a much improved
3502   implementation of the [6]OpenACC 2.0a specification. Highlights are:
3503     * In addition to single-threaded host-fallback execution, offloading
3504       is supported for nvptx (Nvidia GPUs) on x86_64 and PowerPC 64-bit
3505       little-endian GNU/Linux host systems. For nvptx offloading, with
3506       the OpenACC parallel construct, the execution model allows for an
3507       arbitrary number of gangs, up to 32 workers, and 32 vectors.
3508     * Initial support for parallelized execution of OpenACC kernels
3509       constructs:
3510          + Parallelization of a kernels region is switched on by
3511            -fopenacc combined with -O2 or higher.
3512          + Code is offloaded onto multiple gangs, but executes with just
3513            one worker, and a vector length of 1.
3514          + Directives inside a kernels region are not supported.
3515          + Loops with reductions can be parallelized.
3516          + Only kernels regions with one loop nest are parallelized.
3517          + Only the outer-most loop of a loop nest can be parallelized.
3518          + Loop nests containing sibling loops are not parallelized.
3519       Typically, using the OpenACC parallel construct gives much better
3520       performance, compared to the initial support of the OpenACC kernels
3521       construct.
3522     * The device_type clause is not supported. The bind and nohost
3523       clauses are not supported. The host_data directive is not supported
3524       in Fortran.
3525     * Nested parallelism (cf. CUDA dynamic parallelism) is not supported.
3526     * Usage of OpenACC constructs inside multithreaded contexts (such as
3527       created by OpenMP, or pthread programming) is not supported.
3528     * If a call to the acc_on_device function has a compile-time constant
3529       argument, the function call evaluates to a compile-time constant
3530       value only for C and C++ but not for Fortran.
3531
3532   See the [7]OpenACC and [8]Offloading wiki pages for further
3533   information.
3534
3535  C family
3536
3537     * Version 4.5 of the [9]OpenMP specification is now supported in the
3538       C and C++ compilers.
3539     * The C and C++ compilers now support attributes on enumerators. For
3540       instance, it is now possible to mark enumerators as deprecated:
3541
3542enum {
3543  newval,
3544  oldval __attribute__ ((deprecated ("too old")))
3545};
3546
3547     * Source locations for the C and C++ compilers are now tracked as
3548       ranges, rather than just points, making it easier to identify the
3549       subexpression of interest within a complicated expression. For
3550       example:
3551
3552test.cc: In function 'int test(int, int, foo, int, int)':
3553test.cc:5:16: error: no match for 'operator*' (operand types are 'int' and 'foo'
3554)
3555   return p + q * r * s + t;
3556              ~~^~~
3557
3558       In addition, there is now initial support for precise diagnostic
3559       locations within strings:
3560
3561format-strings.c:3:14: warning: field width specifier '*' expects a matching 'in
3562t' argument [-Wformat=]
3563   printf("%*d");
3564            ^
3565
3566     * Diagnostics can now contain "fix-it hints", which are displayed in
3567       context underneath the relevant source code. For example:
3568
3569fixits.c: In function 'bad_deref':
3570fixits.c:11:13: error: 'ptr' is a pointer; did you mean to use '->'?
3571   return ptr.x;
3572             ^
3573             ->
3574
3575     * The C and C++ compilers now offer suggestions for misspelled field
3576       names:
3577
3578spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did
3579you mean 'color'?
3580   return ptr->colour;
3581               ^~~~~~
3582
3583     * New command-line options have been added for the C and C++
3584       compilers:
3585          + -Wshift-negative-value warns about left shifting a negative
3586            value.
3587          + -Wshift-overflow warns about left shift overflows. This
3588            warning is enabled by default. -Wshift-overflow=2 also warns
3589            about left-shifting 1 into the sign bit.
3590          + -Wtautological-compare warns if a self-comparison always
3591            evaluates to true or false. This warning is enabled by -Wall.
3592          + -Wnull-dereference warns if the compiler detects paths that
3593            trigger erroneous or undefined behavior due to dereferencing a
3594            null pointer. This option is only active when
3595            -fdelete-null-pointer-checks is active, which is enabled by
3596            optimizations in most targets. The precision of the warnings
3597            depends on the optimization options used.
3598          + -Wduplicated-cond warns about duplicated conditions in an
3599            if-else-if chain.
3600          + -Wmisleading-indentation warns about places where the
3601            indentation of the code gives a misleading idea of the block
3602            structure of the code to a human reader. For example, given
3603            [10]CVE-2014-1266:
3604
3605sslKeyExchange.c: In function 'SSLVerifySignedServerKeyExchange':
3606sslKeyExchange.c:629:3: warning: this 'if' clause does not guard... [-Wmisleadin
3607g-indentation]
3608    if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0)
3609    ^~
3610sslKeyExchange.c:631:5: note: ...this statement, but the latter is misleadingly
3611indented as if it is guarded by the 'if'
3612        goto fail;
3613        ^~~~
3614
3615            This warning is enabled by -Wall.
3616     * The C and C++ compilers now emit saner error messages if
3617       merge-conflict markers are present in a source file.
3618
3619test.c:3:1: error: version control conflict marker in file
3620 <<<<<<< HEAD
3621 ^~~~~~~
3622
3623  C
3624
3625     * It is possible to disable warnings when an initialized field of a
3626       structure or a union with side effects is being overridden when
3627       using designated initializers via a new warning option
3628       -Woverride-init-side-effects.
3629     * A new type attribute scalar_storage_order applying to structures
3630       and unions has been introduced. It specifies the storage order (aka
3631       endianness) in memory of scalar fields in structures or unions.
3632
3633  C++
3634
3635     * The default mode has been changed to -std=gnu++14.
3636     * [11]C++ Concepts are now supported when compiling with -fconcepts.
3637     * -flifetime-dse is more aggressive in dead-store elimination in
3638       situations where a memory store to a location precedes a
3639       constructor to that memory location.
3640     * G++ now supports [12]C++17 fold expressions, u8 character literals,
3641       extended static_assert, and nested namespace definitions.
3642     * G++ now allows constant evaluation for all non-type template
3643       arguments.
3644     * G++ now supports C++ Transactional Memory when compiling with
3645       -fgnu-tm.
3646
3647    Runtime Library (libstdc++)
3648
3649     * Extensions to the C++ Library to support mathematical special
3650       functions (ISO/IEC 29124:2010), thanks to Edward Smith-Rowland.
3651     * Experimental support for C++17, including the following new
3652       features:
3653          + std::uncaught_exceptions function (this is also available for
3654            -std=gnu++NN modes);
3655          + new member functions try_emplace and insert_or_assign for
3656            unique_key maps;
3657          + non-member functions std::size, std::empty, and std::data for
3658            accessing containers and arrays;
3659          + std::invoke;
3660          + std::shared_mutex;
3661          + std::void_t and std::bool_constant metaprogramming utilities.
3662       Thanks to Ville Voutilainen for contributing many of the C++17
3663       features.
3664     * An experimental implementation of the File System TS.
3665     * Experimental support for most features of the second version of the
3666       Library Fundamentals TS. This includes polymorphic memory resources
3667       and array support in shared_ptr, thanks to Fan You.
3668     * Some assertions checked by Debug Mode can now also be enabled by
3669       _GLIBCXX_ASSERTIONS. The subset of checks enabled by the new macro
3670       have less run-time overhead than the full _GLIBCXX_DEBUG checks and
3671       don't affect the library ABI, so can be enabled per-translation
3672       unit.
3673     * Timed mutex types are supported on more targets, including Darwin.
3674     * Improved std::locale support for DragonFly and FreeBSD, thanks to
3675       John Marino and Andreas Tobler.
3676
3677  Fortran
3678
3679     * Fortran 2008 SUBMODULE support.
3680     * Fortran 2015 EVENT_TYPE, EVENT_POST, EVENT_WAIT, and EVENT_QUERY
3681       support.
3682     * Improved support for Fortran 2003 deferred-length character
3683       variables.
3684     * Improved support for OpenMP and OpenACC.
3685     * The MATMUL intrinsic is now inlined for straightforward cases if
3686       front-end optimization is active. The maximum size for inlining can
3687       be set to n with the -finline-matmul-limit=n option and turned off
3688       with -finline-matmul-limit=0.
3689     * The -Wconversion-extra option will warn about REAL constants which
3690       have excess precision for their kind.
3691     * The -Winteger-division option has been added, which warns about
3692       divisions of integer constants which are truncated. This option is
3693       included in -Wall by default.
3694
3695libgccjit
3696
3697     * The driver code is now run in-process within libgccjit, providing a
3698       small speed-up of the compilation process.
3699     * The API has gained entrypoints for
3700          + [13]timing how long was spent in different parts of code,
3701          + [14]creating switch statements,
3702          + [15]allowing unreachable basic blocks in a function, and
3703          + [16]adding arbitrary command-line options to a compilation.
3704
3705New Targets and Target Specific Improvements
3706
3707  AArch64
3708
3709     * A number of AArch64-specific options have been added. The most
3710       important ones are summarised in this section; for more detailed
3711       information please refer to the documentation.
3712     * The command-line options -march=native, -mcpu=native and
3713       -mtune=native are now available on native AArch64 GNU/Linux
3714       systems. Specifying these options causes GCC to auto-detect the
3715       host CPU and choose the optimal setting for that system.
3716     * -fpic is now supported when generating code for the small code
3717       model (-mcmodel=small). The size of the global offset table (GOT)
3718       is limited to 28KiB under the LP64 SysV ABI, and 15KiB under the
3719       ILP32 SysV ABI.
3720     * The AArch64 port now supports target attributes and pragmas. Please
3721       refer to the [17]documentation for details of available attributes
3722       and pragmas as well as usage instructions.
3723     * Link-time optimization across translation units with different
3724       target-specific options is now supported.
3725     * The option -mtls-size= is now supported. It can be used to specify
3726       the bit size of TLS offsets, allowing GCC to generate better TLS
3727       instruction sequences.
3728     * The option -fno-plt is now fully functional.
3729     * The ARMv8.1-A architecture and the Large System Extensions are now
3730       supported. They can be used by specifying the -march=armv8.1-a
3731       option. Additionally, the +lse option extension can be used in a
3732       similar fashion to other option extensions. The Large System
3733       Extensions introduce new instructions that are used in the
3734       implementation of atomic operations.
3735     * The ACLE half-precision floating-point type __fp16 is now supported
3736       in the C and C++ languages.
3737     * The ARM Cortex-A35 processor is now supported via the
3738       -mcpu=cortex-a35 and -mtune=cortex-a35 options as well as the
3739       equivalent target attributes and pragmas.
3740     * The Qualcomm QDF24xx processor is now supported via the
3741       -mcpu=qdf24xx and -mtune=qdf24xx options as well as the equivalent
3742       target attributes and pragmas.
3743     * Code generation for the ARM Cortex-A57 processor is improved. Among
3744       general code generation improvements, a better algorithm is added
3745       for allocating registers to floating-point multiply-accumulate
3746       instructions offering increased performance when compiling with
3747       -mcpu=cortex-a57 or -mtune=cortex-a57.
3748     * Code generation for the ARM Cortex-A53 processor is improved. A
3749       more accurate instruction scheduling model for the processor is now
3750       used, and a number of compiler tuning parameters have been set to
3751       offer increased performance when compiling with -mcpu=cortex-a53 or
3752       -mtune=cortex-a53.
3753     * Code generation for the Samsung Exynos M1 processor is improved. A
3754       more accurate instruction scheduling model for the processor is now
3755       used, and a number of compiler tuning parameters have been set to
3756       offer increased performance when compiling with -mcpu=exynos-m1 or
3757       -mtune=exynos-m1.
3758     * Improvements in the generation of conditional branches and literal
3759       pools allow the compiler to compile functions of a large size.
3760       Constant pools are now placed into separate rodata sections. The
3761       new option -mpc-relative-literal-loads generates per-function
3762       literal pools, limiting the maximum size of functions to 1MiB.
3763     * Several correctness issues generating Advanced SIMD instructions
3764       for big-endian targets have been fixed resulting in improved code
3765       generation for ACLE intrinsics with -mbig-endian.
3766
3767  ARM
3768
3769     * Support for revisions of the ARM architecture prior to ARMv4t has
3770       been deprecated and will be removed in a future GCC release. The
3771       -mcpu and -mtune values that are deprecated are: arm2, arm250,
3772       arm3, arm6, arm60, arm600, arm610, arm620, arm7, arm7d, arm7di,
3773       arm70, arm700, arm700i, arm710, arm720, arm710c, arm7100, arm7500,
3774       arm7500fe, arm7m, arm7dm, arm7dmi, arm8, arm810, strongarm,
3775       strongarm110, strongarm1100, strongarm1110, fa526, fa626. The value
3776       arm7tdmi is still supported. The values of -march that are
3777       deprecated are: armv2,armv2a,armv3,armv3m,armv4.
3778     * The ARM port now supports target attributes and pragmas. Please
3779       refer to the [18]documentation for details of available attributes
3780       and pragmas as well as usage instructions.
3781     * Support has been added for the following processors (GCC
3782       identifiers in parentheses): ARM Cortex-A32 (cortex-a32), ARM
3783       Cortex-A35 (cortex-a35) and ARM Cortex-R8 (cortex-r8). The GCC
3784       identifiers can be used as arguments to the -mcpu or -mtune
3785       options, for example: -mcpu=cortex-a32 or -mtune=cortex-a35.
3786
3787  Heterogeneous Systems Architecture
3788
3789     * GCC can now generate HSAIL (Heterogeneous System Architecture
3790       Intermediate Language) for simple OpenMP device constructs if
3791       configured with --enable-offload-targets=hsa. A new libgomp plugin
3792       then runs the HSA GPU kernels implementing these constructs on HSA
3793       capable GPUs via a standard HSA run time.
3794       If the HSA compilation back end determines it cannot output HSAIL
3795       for a particular input, it gives a warning by default. These
3796       warnings can be suppressed with -Wno-hsa. To give a few examples,
3797       the HSA back end does not implement compilation of code using
3798       function pointers, automatic allocation of variable sized arrays,
3799       functions with variadic arguments as well as a number of other less
3800       common programming constructs.
3801       When compilation for HSA is enabled, the compiler attempts to
3802       compile composite OpenMP constructs
3803
3804#pragma omp target teams distribute parallel for
3805
3806       into parallel HSA GPU kernels.
3807
3808  IA-32/x86-64
3809
3810     * GCC now supports the Intel CPU named Skylake with AVX-512
3811       extensions through -march=skylake-avx512. The switch enables the
3812       following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW,
3813       AVX-512DQ.
3814     * Support for new AMD instructions monitorx and mwaitx has been
3815       added. This includes new intrinsic and built-in support. It is
3816       enabled through option -mmwaitx. The instructions monitorx and
3817       mwaitx implement the same functionality as the old monitor and
3818       mwait instructions. In addition mwaitx adds a configurable timer.
3819       The timer value is received as third argument and stored in
3820       register %ebx.
3821     * x86-64 targets now allow stack realignment from a word-aligned
3822       stack pointer using the command-line option -mstackrealign or
3823       __attribute__ ((force_align_arg_pointer)). This allows functions
3824       compiled with a vector-aligned stack to be invoked from objects
3825       that keep only word-alignment.
3826     * Support for address spaces __seg_fs, __seg_gs, and __seg_tls. These
3827       can be used to access data via the %fs and %gs segments without
3828       having to resort to inline assembly. Please refer to the
3829       [19]documentation for usage instructions.
3830     * Support for AMD Zen (family 17h) processors is now available
3831       through the -march=znver1 and -mtune=znver1 options.
3832
3833  MeP
3834
3835     * Support for the MeP (mep-elf) architecture has been deprecated and
3836       will be removed in a future GCC release.
3837
3838  MSP430
3839
3840     * The MSP430 compiler now has the ability to automatically distribute
3841       code and data between low memory (addresses below 64K) and high
3842       memory. This only applies to parts that actually have both memory
3843       regions and only if the linker script for the part has been
3844       specifically set up to support this feature.
3845       A new attribute of either can be applied to both functions and
3846       data, and this tells the compiler to place the object into low
3847       memory if there is room and into high memory otherwise. Two other
3848       new attributes - lower and upper - can be used to explicitly state
3849       that an object should be placed in the specified memory region. If
3850       there is not enough left in that region the compilation will fail.
3851       Two new command-line options - -mcode-region=[lower|upper|either]
3852       and -mdata-region=[lower|upper|either] - can be used to tell the
3853       compiler what to do with objects that do not have one of these new
3854       attributes.
3855
3856  PowerPC / PowerPC64 / RS6000
3857
3858     * PowerPC64 now supports IEEE 128-bit floating-point using the
3859       __float128 data type. In GCC 6, this is not enabled by default, but
3860       you can enable it with -mfloat128. The IEEE 128-bit floating-point
3861       support requires the use of the VSX instruction set. IEEE 128-bit
3862       floating-point values are passed and returned as a single vector
3863       value. The software emulator for IEEE 128-bit floating-point
3864       support is only built on PowerPC GNU/Linux systems where the
3865       default CPU is at least power7. On future ISA 3.0 systems (POWER 9
3866       and later), you will be able to use the -mfloat128-hardware option
3867       to use the ISA 3.0 instructions that support IEEE 128-bit
3868       floating-point. An additional type (__ibm128) has been added to
3869       refer to the IBM extended double type that normally implements long
3870       double. This will allow for a future transition to implementing
3871       long double with IEEE 128-bit floating-point.
3872     * Basic support has been added for POWER9 hardware that will use the
3873       recently published OpenPOWER ISA 3.0 instructions. The following
3874       new switches are available:
3875          + -mcpu=power9: Implement all of the ISA 3.0 instructions
3876            supported by the compiler.
3877          + -mtune=power9: In the future, apply tuning for POWER9 systems.
3878            Currently, POWER8 tunings are used.
3879          + -mmodulo: Generate code using the ISA 3.0 integer instructions
3880            (modulus, count trailing zeros, array index support, integer
3881            multiply/add).
3882          + -mpower9-fusion: Generate code to suitably fuse instruction
3883            sequences for a POWER9 system.
3884          + -mpower9-dform: Generate code to use the new D-form
3885            (register+offset) memory instructions for the vector
3886            registers.
3887          + -mpower9-vector: Generate code using the new ISA 3.0 vector
3888            (VSX or Altivec) instructions.
3889          + -mpower9-minmax: Reserved for future development.
3890          + -mtoc-fusion: Keep TOC entries together to provide more fusion
3891            opportunities.
3892     * New constraints have been added to support IEEE 128-bit
3893       floating-point and ISA 3.0 instructions:
3894          + wb: Altivec register if -mpower9-dform is enabled.
3895          + we: VSX register if -mpower9-vector is enabled for 64-bit code
3896            generation.
3897          + wo: VSX register if -mpower9-vector is enabled.
3898          + wp: Reserved for future use if long double is implemented with
3899            IEEE 128-bit floating-point instead of IBM extended double.
3900          + wq: VSX register if -mfloat128 is enabled.
3901          + wF: Memory operand suitable for POWER9 fusion load/store.
3902          + wG: Memory operand suitable for TOC fusion memory references.
3903          + wL: Integer constant identifying the element number mfvsrld
3904            accesses within a vector.
3905     * Support has been added for __builtin_cpu_is() and
3906       __builtin_cpu_supports(), allowing for very fast access to
3907       AT_PLATFORM, AT_HWCAP, and AT_HWCAP2 values. This requires use of
3908       glibc 2.23 or later.
3909     * All hardware transactional memory builtins now correctly behave as
3910       memory barriers. Programmers can use #ifdef __TM_FENCE__ to
3911       determine whether their "old" compiler treats the builtins as
3912       barriers.
3913     * Split-stack support has been added for gccgo on PowerPC64 for both
3914       big- and little-endian (but not for 32-bit). The gold linker from
3915       at least binutils 2.25.1 must be available in the PATH when
3916       configuring and building gccgo to enable split stack. (The
3917       requirement for binutils 2.25.1 applies to PowerPC64 only.) The
3918       split-stack feature allows a small initial stack size to be
3919       allocated for each goroutine, which increases as needed.
3920     * GCC on PowerPC now supports the standard lround function.
3921     * A new configuration option ---with-advance-toolchain=at was added
3922       for PowerPC 64-bit GNU/Linux systems to use the header files,
3923       library files, and the dynamic linker from a specific Advance
3924       Toolchain release instead of the default versions that are provided
3925       by the GNU/Linux distribution. In general, this option is intended
3926       for the developers of GCC, and it is not intended for general use.
3927     * The "q", "S", "T", and "t" asm-constraints have been removed.
3928     * The "b", "B", "m", "M", and "W" format modifiers have been removed.
3929
3930  S/390, System z, IBM z Systems
3931
3932     * Support for the IBM z13 processor has been added. When using the
3933       -march=z13 option, the compiler will generate code making use of
3934       the new instructions and registers introduced with the vector
3935       extension facility. The -mtune=z13 option enables z13 specific
3936       instruction scheduling without making use of new instructions.
3937       Compiling code with -march=z13 reduces the default alignment of
3938       vector types bigger than 8 bytes to 8. This is an ABI change and
3939       care must be taken when linking modules compiled with different
3940       arch levels which interchange variables containing vector type
3941       values. For newly compiled code the GNU linker will emit a warning.
3942     * The -mzvector option enables a C/C++ language extension. This
3943       extension provides a new keyword vector which can be used to define
3944       vector type variables. (Note: This is not available when enforcing
3945       strict standard compliance e.g. with -std=c99. Either enable GNU
3946       extensions with e.g. -std=gnu99 or use __vector instead of vector.)
3947       Additionally a set of overloaded builtins is provided which is
3948       partially compatible to the PowerPC Altivec builtins. In order to
3949       make use of these builtins the vecintrin.h header file needs to be
3950       included.
3951     * The new command-line options -march=native, and -mtune=native are
3952       now available on native IBM z Systems. Specifying these options
3953       causes GCC to auto-detect the host CPU and choose the optimal
3954       setting for that system. If GCC is unable to detect the host CPU
3955       these options have no effect.
3956     * The IBM z Systems port now supports target attributes and pragmas.
3957       Please refer to the [20]documentation for details of available
3958       attributes and pragmas as well as usage instructions.
3959     * -fsplit-stack is now supported as part of the IBM z Systems port.
3960       This feature requires a recent gold linker to be used.
3961     * Support for the g5 and g6 -march=/-mtune= CPU level switches has
3962       been deprecated and will be removed in a future GCC release. -m31
3963       from now on defaults to -march=z900 if not specified otherwise.
3964       -march=native on a g5/g6 machine will default to -march=z900.
3965
3966  SH
3967
3968     * Support for SH5 / SH64 has been declared obsolete and will be
3969       removed in future releases.
3970     * Support for the FDPIC ABI has been added. It can be enabled using
3971       the new -mfdpic target option and --enable-fdpic configure option.
3972
3973  SPARC
3974
3975     * An ABI bug has been fixed in 64-bit mode. Unfortunately, this
3976       change will break binary compatibility with earlier releases for
3977       code it affects, but this should be pretty rare in practice. The
3978       conditions are: a 16-byte structure containing a double or a 8-byte
3979       vector in the second half is passed to a subprogram in slot #15,
3980       for example as 16th parameter if the first 15 ones have at most 8
3981       bytes. The double or vector was wrongly passed in floating-point
3982       register %d32 in lieu of on the stack as per the SPARC calling
3983       conventions.
3984
3985Operating Systems
3986
3987  AIX
3988
3989     * DWARF debugging support for AIX 7.1 has been enabled as an optional
3990       debugging format. A more recent Technology Level (TL) and GCC built
3991       with that level are required for full exploitation of DWARF
3992       debugging capabilities.
3993
3994  Linux
3995
3996     * Support for the [21]musl C library was added for the AArch64, ARM,
3997       MicroBlaze, MIPS, MIPS64, PowerPC, PowerPC64, SH, i386, x32 and
3998       x86_64 targets. It can be selected using the new -mmusl option in
3999       case musl is not the default libc. GCC defaults to musl libc if it
4000       is built with a target triplet matching the *-linux-musl* pattern.
4001
4002  RTEMS
4003
4004     * The RTEMS thread model implementation changed. Mutexes now use
4005       self-contained objects defined in Newlib <sys/lock.h> instead of
4006       Classic API semaphores. The keys for thread specific data and the
4007       once function are directly defined via <pthread.h>. Self-contained
4008       condition variables are provided via Newlib <sys/lock.h>. The RTEMS
4009       thread model also supports C++11 threads.
4010     * OpenMP support now uses self-contained objects provided by Newlib
4011       <sys/lock.h> and offers a significantly better performance compared
4012       to the POSIX configuration of libgomp. It is possible to configure
4013       thread pools for each scheduler instance via the environment
4014       variable GOMP_RTEMS_THREAD_POOLS.
4015
4016  Solaris
4017
4018     * Solaris 12 is now fully supported. Minimal support had already been
4019       present in GCC 5.3.
4020     * Solaris 12 provides a full set of startup files (crt1.o, crti.o,
4021       crtn.o), which GCC now prefers over its own ones.
4022     * Position independent executables (PIE) are now supported on Solaris
4023       12.
4024     * Constructor priority is now supported on Solaris 12 with the system
4025       linker.
4026     * libvtv has been ported to Solaris 11 and up.
4027
4028  Windows
4029
4030     * The option -mstackrealign is now automatically activated in 32-bit
4031       mode whenever the use of SSE instructions is requested.
4032
4033Other significant improvements
4034
4035     * The gcc and g++ driver programs will now provide suggestions for
4036       misspelled command-line options.
4037
4038$ gcc -static-libfortran test.f95
4039gcc: error: unrecognized command line option '-static-libfortran'; did you mean
4040'-static-libgfortran'?
4041
4042     * The --enable-default-pie configure option enables generation of PIE
4043       by default.
4044
4045                                    GCC 6.2
4046
4047   This is the [22]list of problem reports (PRs) from GCC's bug tracking
4048   system that are known to be fixed in the 6.2 release. This list might
4049   not be complete (that is, it is possible that some PRs that have been
4050   fixed are not listed here).
4051
4052Target Specific Changes
4053
4054  SPARC
4055
4056     * Support for --with-cpu-32 and --with-cpu-64 configure options has
4057       been added on bi-architecture platforms.
4058     * Support for the SPARC M7 (Niagara 7) processor has been added.
4059     * Support for the VIS 4.0 instruction set has been added.
4060
4061                                    GCC 6.3
4062
4063   This is the [23]list of problem reports (PRs) from GCC's bug tracking
4064   system that are known to be fixed in the 6.3 release. This list might
4065   not be complete (that is, it is possible that some PRs that have been
4066   fixed are not listed here).
4067
4068Target Specific Changes
4069
4070  IA-32/x86-64
4071
4072     * Support for the [24]deprecated pcommit instruction has been
4073       removed.
4074
4075                                    GCC 6.4
4076
4077   This is the [25]list of problem reports (PRs) from GCC's bug tracking
4078   system that are known to be fixed in the 6.4 release. This list might
4079   not be complete (that is, it is possible that some PRs that have been
4080   fixed are not listed here).
4081
4082Operating Systems
4083
4084  RTEMS
4085
4086     * The ABI changes on ARM so that no short enums are used by default.
4087
4088                                    GCC 6.5
4089
4090   This is the [26]list of problem reports (PRs) from GCC's bug tracking
4091   system that are known to be fixed in the 6.5 release. This list might
4092   not be complete (that is, it is possible that some PRs that have been
4093   fixed are not listed here).
4094
4095
4096    For questions related to the use of GCC, please consult these web
4097    pages and the [27]GCC manuals. If that fails, the
4098    [28]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4099    web pages and the development of GCC are welcome on our developer
4100    list at [29]gcc@gcc.gnu.org. All of [30]our lists have public
4101    archives.
4102
4103   Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and
4104   distribution of this entire article is permitted in any medium,
4105   provided this notice is preserved.
4106
4107   These pages are [32]maintained by the GCC team. Last modified
4108   2019-11-28[33].
4109
4110References
4111
4112   1. http://gcc.gnu.org/gcc-6/porting_to.html
4113   2. http://gcc.gnu.org/onlinedocs/index.html#current
4114   3. https://gcc.gnu.org/ml/gcc/2015-08/msg00101.html
4115   4. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71151
4116   5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87822
4117   6. https://www.openacc.org/
4118   7. https://gcc.gnu.org/wiki/OpenACC
4119   8. https://gcc.gnu.org/wiki/Offloading
4120   9. https://www.openmp.org/specifications/
4121  10. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2014-1266
4122  11. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4377.pdf
4123  12. https://gcc.gnu.org/projects/cxx-status.html#cxx1z
4124  13. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/performance.html
4125  14. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/functions.html#gcc_jit_block_end_with_switch
4126  15. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_set_bool_allow_unreachable_blocks
4127  16. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_add_command_line_option
4128  17. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/AArch64-Function-Attributes.html#AArch64-Function-Attributes
4129  18. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/ARM-Function-Attributes.html#ARM-Function-Attributes
4130  19. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/Named-Address-Spaces.html#Named-Address-Spaces
4131  20. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/S_002f390-Function-Attributes.html#S_002f390-Function-Attributes
4132  21. http://www.musl-libc.org/
4133  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.2
4134  23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.3
4135  24. https://software.intel.com/en-us/blogs/2016/09/12/deprecate-pcommit-instruction
4136  25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.4
4137  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.5
4138  27. https://gcc.gnu.org/onlinedocs/
4139  28. mailto:gcc-help@gcc.gnu.org
4140  29. mailto:gcc@gcc.gnu.org
4141  30. https://gcc.gnu.org/lists.html
4142  31. https://www.fsf.org/
4143  32. https://gcc.gnu.org/about.html
4144  33. http://validator.w3.org/check/referer
4145======================================================================
4146http://gcc.gnu.org/gcc-5/index.html
4147                              GCC 5 Release Series
4148
4149   (This release series is no longer supported.)
4150
4151   October 10, 2017
4152
4153   The [1]GNU project and the GCC developers are pleased to announce the
4154   release of GCC 5.5.
4155
4156   This release is a bug-fix release, containing fixes for regressions in
4157   GCC 5.4 relative to previous releases of GCC.
4158
4159Release History
4160
4161   GCC 5.5
4162          October 10, 2017 ([2]changes, [3]documentation)
4163
4164   GCC 5.4
4165          June 3, 2016 ([4]changes, [5]documentation)
4166
4167   GCC 5.3
4168          December 4, 2015 ([6]changes, [7]documentation)
4169
4170   GCC 5.2
4171          July 16, 2015 ([8]changes, [9]documentation)
4172
4173   GCC 5.1
4174          April 22, 2015 ([10]changes, [11]documentation)
4175
4176References and Acknowledgements
4177
4178   GCC used to stand for the GNU C Compiler, but since the compiler
4179   supports several other languages aside from C, it now stands for the
4180   GNU Compiler Collection.
4181
4182   A list of [12]successful builds is updated as new information becomes
4183   available.
4184
4185   The GCC developers would like to thank the numerous people that have
4186   contributed new features, improvements, bug fixes, and other changes as
4187   well as test results to GCC. This [13]amazing group of volunteers is
4188   what makes GCC successful.
4189
4190   For additional information about GCC please refer to the [14]GCC
4191   project web site or contact the [15]GCC development mailing list.
4192
4193   To obtain GCC please use [16]our mirror sites or [17]our version
4194   control system.
4195
4196
4197    For questions related to the use of GCC, please consult these web
4198    pages and the [18]GCC manuals. If that fails, the
4199    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4200    web pages and the development of GCC are welcome on our developer
4201    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
4202    archives.
4203
4204   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
4205   distribution of this entire article is permitted in any medium,
4206   provided this notice is preserved.
4207
4208   These pages are [23]maintained by the GCC team. Last modified
4209   2020-01-14[24].
4210
4211References
4212
4213   1. http://www.gnu.org/
4214   2. http://gcc.gnu.org/gcc-5/changes.html
4215   3. http://gcc.gnu.org/onlinedocs/5.5.0/
4216   4. http://gcc.gnu.org/gcc-5/changes.html
4217   5. http://gcc.gnu.org/onlinedocs/5.4.0/
4218   6. http://gcc.gnu.org/gcc-5/changes.html
4219   7. http://gcc.gnu.org/onlinedocs/5.3.0/
4220   8. http://gcc.gnu.org/gcc-5/changes.html
4221   9. http://gcc.gnu.org/onlinedocs/5.2.0/
4222  10. http://gcc.gnu.org/gcc-5/changes.html
4223  11. http://gcc.gnu.org/onlinedocs/5.1.0/
4224  12. http://gcc.gnu.org/gcc-5/buildstat.html
4225  13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html
4226  14. http://gcc.gnu.org/index.html
4227  15. mailto:gcc@gcc.gnu.org
4228  16. http://gcc.gnu.org/mirrors.html
4229  17. http://gcc.gnu.org/git.html
4230  18. https://gcc.gnu.org/onlinedocs/
4231  19. mailto:gcc-help@gcc.gnu.org
4232  20. mailto:gcc@gcc.gnu.org
4233  21. https://gcc.gnu.org/lists.html
4234  22. https://www.fsf.org/
4235  23. https://gcc.gnu.org/about.html
4236  24. http://validator.w3.org/check/referer
4237======================================================================
4238http://gcc.gnu.org/gcc-5/changes.html
4239                              GCC 5 Release Series
4240                        Changes, New Features, and Fixes
4241
4242Caveats
4243
4244     * The default mode for C is now -std=gnu11 instead of -std=gnu89.
4245     * The C++ runtime library (libstdc++) uses a new ABI by default (see
4246       [1]below).
4247     * The Graphite framework for loop optimizations no longer requires
4248       the CLooG library, only ISL version 0.14 (recommended) or 0.12.2.
4249       The installation manual contains more information about
4250       requirements to build GCC.
4251     * The non-standard C++0x type traits has_trivial_default_constructor,
4252       has_trivial_copy_constructor and has_trivial_copy_assign have been
4253       deprecated and will be removed in a future version. The standard
4254       C++11 traits is_trivially_default_constructible,
4255       is_trivially_copy_constructible and is_trivially_copy_assignable
4256       should be used instead.
4257     * On AVR, support has been added for the devices
4258       ATtiny4/5/9/10/20/40. This requires Binutils 2.25 or newer.
4259     * The AVR port uses a new scheme to describe supported devices: For
4260       each supported device the compiler provides a device-specific
4261       [2]spec file. If the compiler is used together with AVR-LibC, this
4262       requires at least GCC 5.2 and a version of AVR-LibC which
4263       implements [3]feature #44574.
4264
4265General Optimizer Improvements
4266
4267     * Inter-procedural optimization improvements:
4268          + An Identical Code Folding (ICF) pass (controlled via
4269            -fipa-icf) has been added. Compared to the identical code
4270            folding performed by the Gold linker this pass does not
4271            require function sections. It also performs merging before
4272            inlining, so inter-procedural optimizations are aware of the
4273            code re-use. On the other hand not all unifications performed
4274            by a linker are doable by GCC which must honor aliasing
4275            information. During link-time optimization of Firefox, this
4276            pass unifies about 31000 functions, that is 14% overall.
4277          + The devirtualization pass was significantly improved by adding
4278            better support for speculative devirtualization and dynamic
4279            type detection. About 50% of virtual calls in Firefox are now
4280            speculatively devirtualized during link-time optimization.
4281          + A new comdat localization pass allows the linker to eliminate
4282            more dead code in presence of C++ inline functions.
4283          + Virtual tables are now optimized. Local aliases are used to
4284            reduce dynamic linking time of C++ virtual tables on ELF
4285            targets and data alignment has been reduced to limit data
4286            segment bloat.
4287          + A new -fno-semantic-interposition option can be used to
4288            improve code quality of shared libraries where interposition
4289            of exported symbols is not allowed.
4290          + Write-only variables are now detected and optimized out.
4291          + With profile feedback the function inliner can now bypass
4292            --param inline-insns-auto and --param inline-insns-single
4293            limits for hot calls.
4294          + The IPA reference pass was significantly sped up making it
4295            feasible to enable -fipa-reference with -fprofile-generate.
4296            This also solves a bottleneck seen when building Chromium with
4297            link-time optimization.
4298          + The symbol table and call-graph API was reworked to C++ and
4299            simplified.
4300          + The interprocedural propagation of constants now also
4301            propagates alignments of pointer parameters. This for example
4302            means that the vectorizer often does not need to generate loop
4303            prologues and epilogues to make up for potential
4304            misalignments.
4305     * Link-time optimization improvements:
4306          + One Definition Rule based merging of C++ types has been
4307            implemented. Type merging enables better devirtualization and
4308            alias analysis. Streaming extra information needed to merge
4309            types adds about 2-6% of memory size and object size increase.
4310            This can be controlled by -flto-odr-type-merging.
4311          + Command-line optimization and target options are now streamed
4312            on a per-function basis and honored by the link-time
4313            optimizer. This change makes link-time optimization a more
4314            transparent replacement of per-file optimizations. It is now
4315            possible to build projects that require different optimization
4316            settings for different translation units (such as -ffast-math,
4317            -mavx, or -finline). Contrary to earlier GCC releases, the
4318            optimization and target options passed on the link command
4319            line are ignored.
4320            Note that this applies only to those command-line options that
4321            can be passed to optimize and target attributes. Command-line
4322            options affecting global code generation (such as -fpic),
4323            warnings (such as -Wodr), optimizations affecting the way
4324            static variables are optimized (such as -fcommon), debug
4325            output (such as -g), and --param parameters can be applied
4326            only to the whole link-time optimization unit. In these cases,
4327            it is recommended to consistently use the same options at both
4328            compile time and link time.
4329          + GCC bootstrap now uses slim LTO object files.
4330          + Memory usage and link times were improved. Tree merging was
4331            sped up, memory usage of GIMPLE declarations and types was
4332            reduced, and, support for on-demand streaming of variable
4333            constructors was added.
4334     * Feedback directed optimization improvements:
4335          + A new auto-FDO mode uses profiles collected by low overhead
4336            profiling tools (perf) instead of more expensive program
4337            instrumentation (via -fprofile-generate). SPEC2006 benchmarks
4338            on x86-64 improve by 4.7% with auto-FDO and by 7.3% with
4339            traditional feedback directed optimization.
4340          + Profile precision was improved in presence of C++ inline and
4341            extern inline functions.
4342          + The new gcov-tool utility allows manipulating profiles.
4343          + Profiles are now more tolerant to source file changes (this
4344            can be controlled by --param profile-func-internal-id).
4345     * Register allocation improvements:
4346          + A new local register allocator (LRA) sub-pass, controlled by
4347            -flra-remat, implements control-flow sensitive global register
4348            rematerialization. Instead of spilling and restoring a
4349            register value, it is recalculated if it is profitable. The
4350            sub-pass improved SPEC2000 generated code by 1% and 0.5%
4351            correspondingly on ARM and x86-64.
4352          + Reuse of the PIC hard register, instead of using a fixed
4353            register, was implemented on x86/x86-64 targets. This improves
4354            generated PIC code performance as more hard registers can be
4355            used. Shared libraries can significantly benefit from this
4356            optimization. Currently it is switched on only for x86/x86-64
4357            targets. As RA infrastructure is already implemented for PIC
4358            register reuse, other targets might follow this in the future.
4359          + A simple form of inter-procedural RA was implemented. When it
4360            is known that a called function does not use caller-saved
4361            registers, save/restore code is not generated around the call
4362            for such registers. This optimization can be controlled by
4363            -fipa-ra
4364          + LRA is now much more effective at generating spills of general
4365            registers into vector registers instead of memory on
4366            architectures (e.g., modern Intel processors) where this is
4367            profitable.
4368     * UndefinedBehaviorSanitizer gained a few new sanitization options:
4369          + -fsanitize=float-divide-by-zero: detect floating-point
4370            division by zero;
4371          + -fsanitize=float-cast-overflow: check that the result of
4372            floating-point type to integer conversions do not overflow;
4373          + -fsanitize=bounds: enable instrumentation of array bounds and
4374            detect out-of-bounds accesses;
4375          + -fsanitize=alignment: enable alignment checking, detect
4376            various misaligned objects;
4377          + -fsanitize=object-size: enable object size checking, detect
4378            various out-of-bounds accesses.
4379          + -fsanitize=vptr: enable checking of C++ member function calls,
4380            member accesses and some conversions between pointers to base
4381            and derived classes, detect if the referenced object does not
4382            have the correct dynamic type.
4383     * Pointer Bounds Checker, a bounds violation detector, has been added
4384       and can be enabled via -fcheck-pointer-bounds. Memory accesses are
4385       instrumented with run-time checks of used pointers against their
4386       bounds to detect pointer bounds violations (overflows). The Pointer
4387       Bounds Checker is available on x86/x86-64 GNU/Linux targets with a
4388       new ISA extension Intel MPX support. See the Pointer Bounds Checker
4389       [4]Wiki page for more details.
4390
4391New Languages and Language specific improvements
4392
4393     * [5]OpenMP 4.0 specification offloading features are now supported
4394       by the C, C++, and Fortran compilers. Generic changes:
4395          + Infrastructure (suitable for any vendor).
4396          + Testsuite which covers offloading from the [6]OpenMP 4.0
4397            Examples document.
4398       Specific for upcoming Intel Xeon Phi products:
4399          + Run-time library.
4400          + Card emulator.
4401     * GCC 5 includes a preliminary implementation of the OpenACC 2.0a
4402       specification. OpenACC is intended for programming accelerator
4403       devices such as GPUs. See [7]the OpenACC wiki page for more
4404       information.
4405
4406  C family
4407
4408     * The default setting of the -fdiagnostics-color= command-line option
4409       is now [8]configurable when building GCC using configuration option
4410       --with-diagnostics-color=. The possible values are: never, always,
4411       auto and auto-if-env. The new default auto uses color only when the
4412       standard error is a terminal. The default in GCC 4.9 was
4413       auto-if-env, which is equivalent to auto if there is a non-empty
4414       GCC_COLORS environment variable, and never otherwise. As in GCC
4415       4.9, an empty GCC_COLORS variable in the environment will always
4416       disable colors, no matter what the default is or what command-line
4417       options are used.
4418     * A new command-line option -Wswitch-bool has been added for the C
4419       and C++ compilers, which warns whenever a switch statement has an
4420       index of boolean type.
4421     * A new command-line option -Wlogical-not-parentheses has been added
4422       for the C and C++ compilers, which warns about "logical not" used
4423       on the left hand side operand of a comparison.
4424     * A new command-line option -Wsizeof-array-argument has been added
4425       for the C and C++ compilers, which warns when the sizeof operator
4426       is applied to a parameter that has been declared as an array in a
4427       function definition.
4428     * A new command-line option -Wbool-compare has been added for the C
4429       and C++ compilers, which warns about boolean expressions compared
4430       with an integer value different from true/false.
4431     * Full support for [9]Cilk Plus has been added to the GCC compiler.
4432       Cilk Plus is an extension to the C and C++ languages to support
4433       data and task parallelism.
4434     * A new attribute no_reorder prevents reordering of selected symbols
4435       against other such symbols or inline assembler. This enables to
4436       link-time optimize the Linux kernel without having to resort to
4437       -fno-toplevel-reorder that disables several optimizations.
4438     * New preprocessor constructs, __has_include and __has_include_next,
4439       to test the availability of headers have been added.
4440       This demonstrates a way to include the header <optional> only if it
4441       is available:
4442
4443#ifdef __has_include
4444#  if __has_include(<optional>)
4445#    include <optional>
4446#    define have_optional 1
4447#  elif __has_include(<experimental/optional>)
4448#    include <experimental/optional>
4449#    define have_optional 1
4450#    define experimental_optional
4451#  else
4452#    define have_optional 0
4453#  endif
4454#endif
4455
4456       The header search paths for __has_include and __has_include_next
4457       are equivalent to those of the standard directive #include and the
4458       extension #include_next respectively.
4459     * A new built-in function-like macro to determine the existence of an
4460       attribute, __has_attribute, has been added. The equivalent built-in
4461       macro __has_cpp_attribute was added to C++ to support
4462       [10]Feature-testing recommendations for C++. The macro
4463       __has_attribute is added to all C-like languages as an extension:
4464
4465int
4466#ifdef __has_attribute
4467#  if __has_attribute(__noinline__)
4468  __attribute__((__noinline__))
4469#  endif
4470#endif
4471foo(int x);
4472
4473       If an attribute exists, a nonzero constant integer is returned. For
4474       standardized C++ attributes a date is returned, otherwise the
4475       constant returned is 1. Both __has_attribute and
4476       __has_cpp_attribute will add underscores to an attribute name if
4477       necessary to resolve the name. For C++11 and onwards the attribute
4478       may be scoped.
4479     * A new set of built-in functions for arithmetics with overflow
4480       checking has been added: __builtin_add_overflow,
4481       __builtin_sub_overflow and __builtin_mul_overflow and for
4482       compatibility with clang also other variants. These builtins have
4483       two integral arguments (which don't need to have the same type),
4484       the arguments are extended to infinite precision signed type, +, -
4485       or * is performed on those, and the result is stored in an integer
4486       variable pointed to by the last argument. If the stored value is
4487       equal to the infinite precision result, the built-in functions
4488       return false, otherwise true. The type of the integer variable that
4489       will hold the result can be different from the types of the first
4490       two arguments. The following snippet demonstrates how this can be
4491       used in computing the size for the calloc function:
4492
4493void *
4494calloc (size_t x, size_t y)
4495{
4496  size_t sz;
4497  if (__builtin_mul_overflow (x, y, &sz))
4498    return NULL;
4499  void *ret = malloc (sz);
4500  if (ret) memset (res, 0, sz);
4501  return ret;
4502}
4503
4504       On e.g. i?86 or x86-64 the above will result in a mul instruction
4505       followed by a jump on overflow.
4506     * The option -fextended-identifiers is now enabled by default for
4507       C++, and for C99 and later C versions. Various bugs in the
4508       implementation of extended identifiers have been fixed.
4509
4510  C
4511
4512     * The default mode has been changed to -std=gnu11.
4513     * A new command-line option -Wc90-c99-compat has been added to warn
4514       about features not present in ISO C90, but present in ISO C99.
4515     * A new command-line option -Wc99-c11-compat has been added to warn
4516       about features not present in ISO C99, but present in ISO C11.
4517     * It is possible to disable warnings about conversions between
4518       pointers that have incompatible types via a new warning option
4519       -Wno-incompatible-pointer-types; warnings about implicit
4520       incompatible integer to pointer and pointer to integer conversions
4521       via a new warning option -Wno-int-conversion; and warnings about
4522       qualifiers on pointers being discarded via a new warning option
4523       -Wno-discarded-qualifiers.
4524     * To allow proper use of const qualifiers with multidimensional
4525       arrays, GCC will not warn about incompatible pointer types anymore
4526       for conversions between pointers to arrays with and without const
4527       qualifier (except when using -pedantic). Instead, a new warning is
4528       emitted only if the const qualifier is lost. This can be controlled
4529       with a new warning option -Wno-discarded-array-qualifiers.
4530     * The C front end now generates more precise caret diagnostics.
4531     * The -pg command-line option now only affects the current file in an
4532       LTO build.
4533
4534  C++
4535
4536     * G++ now supports [11]C++14 variable templates.
4537     * -Wnon-virtual-dtor doesn't warn anymore for final classes.
4538     * Excessive template instantiation depth is now a fatal error. This
4539       prevents excessive diagnostics that usually do not help to identify
4540       the problem.
4541     * G++ and libstdc++ now implement the feature-testing macros from
4542       [12]Feature-testing recommendations for C++.
4543     * G++ now allows typename in a template template parameter.
4544
4545template<template<typename> typename X> struct D; // OK
4546
4547     * G++ now supports [13]C++14 aggregates with non-static data member
4548       initializers.
4549
4550struct A { int i, j = i; };
4551A a = { 42 }; // a.j is also 42
4552
4553     * G++ now supports [14]C++14 extended constexpr.
4554
4555constexpr int f (int i)
4556{
4557  int j = 0;
4558  for (; i > 0; --i)
4559    ++j;
4560  return j;
4561}
4562
4563constexpr int i = f(42); // i is 42
4564
4565     * G++ now supports the [15]C++14 sized deallocation functions.
4566
4567void operator delete (void *, std::size_t) noexcept;
4568void operator delete[] (void *, std::size_t) noexcept;
4569
4570     * A new One Definition Rule violation warning (controlled by -Wodr)
4571       detects mismatches in type definitions and virtual table contents
4572       during link-time optimization.
4573     * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help
4574       developers to annotate programs with final specifiers (or anonymous
4575       namespaces) to improve code generation. These warnings can be used
4576       at compile time, but they are more useful in combination with
4577       link-time optimization.
4578     * G++ no longer supports [16]N3639 variable length arrays, as they
4579       were removed from the C++14 working paper prior to ratification.
4580       GNU VLAs are still supported, so VLA support is now the same in
4581       C++14 mode as in C++98 and C++11 modes.
4582     * G++ now allows passing a non-trivially-copyable class via C
4583       varargs, which is conditionally-supported with
4584       implementation-defined semantics in the standard. This uses the
4585       same calling convention as a normal value parameter.
4586     * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So
4587       various mangling bugs are fixed, but G++ will still emit aliases
4588       with the old, wrong mangling where feasible. -Wabi=2 will warn
4589       about differences between ABI version 2 and the current setting.
4590     * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely
4591       to be unaffected, but -Wabi=8 will warn about a non-static data
4592       member with type std::nullptr_t which changes position due to this
4593       change.
4594
4595    Runtime Library (libstdc++)
4596
4597     * A [17]Dual ABI is provided by the library. A new ABI is enabled by
4598       default. The old ABI is still supported and can be used by defining
4599       the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++
4600       standard library headers.
4601     * A new implementation of std::string is enabled by default, using
4602       the small string optimization instead of copy-on-write reference
4603       counting.
4604     * A new implementation of std::list is enabled by default, with an
4605       O(1) size() function;
4606     * [18]Full support for C++11, including the following new features:
4607          + std::deque and std::vector<bool> meet the allocator-aware
4608            container requirements;
4609          + movable and swappable iostream classes;
4610          + support for std::align and std::aligned_union;
4611          + type traits std::is_trivially_copyable,
4612            std::is_trivially_constructible, std::is_trivially_assignable
4613            etc.;
4614          + I/O manipulators std::put_time, std::get_time, std::hexfloat
4615            and std::defaultfloat;
4616          + generic locale-aware std::isblank;
4617          + locale facets for Unicode conversion;
4618          + atomic operations for std::shared_ptr;
4619          + std::notify_all_at_thread_exit() and functions for making
4620            futures ready at thread exit.
4621     * Support for the C++11 hexfloat manipulator changes how the num_put
4622       facet formats floating point types when
4623       ios_base::fixed|ios_base::scientific is set in a stream's fmtflags.
4624       This change affects all language modes, even though the C++98
4625       standard gave no special meaning to that combination of flags. To
4626       prevent the use of hexadecimal notation for floating point types
4627       use str.unsetf(std::ios_base::floatfield) to clear the relevant
4628       bits in str.flags().
4629     * [19]Full experimental support for C++14, including the following
4630       new features:
4631          + std::is_final type trait;
4632          + heterogeneous comparison lookup in associative containers.
4633          + global functions cbegin, cend, rbegin, rend, crbegin, and
4634            crend for range access to containers, arrays and initializer
4635            lists.
4636     * [20]Improved experimental support for the Library Fundamentals TS,
4637       including:
4638          + class std::experimental::any;
4639          + function template std::experimental::apply;
4640          + function template std::experimental::sample;
4641          + function template std::experimental::search and related
4642            searcher types;
4643          + variable templates for type traits;
4644          + function template std::experimental::not_fn.
4645     * New random number distributions logistic_distribution and
4646       uniform_on_sphere_distribution as extensions.
4647     * [21]GDB Xmethods for containers and std::unique_ptr.
4648
4649  Fortran
4650
4651     * Compatibility notice:
4652          + The version of the module files (.mod) has been incremented.
4653          + For free-form source files [22]-Werror=line-truncation is now
4654            enabled by default. Note that comments exceeding the line
4655            length are not diagnosed. (For fixed-form source code, the
4656            same warning is available but turned off by default, such that
4657            excess characters are ignored. -ffree-line-length-n and
4658            -ffixed-line-length-n can be used to modify the default line
4659            lengths of 132 and 72 columns, respectively.)
4660          + The -Wtabs option is now more sensible: with -Wtabs the
4661            compiler warns if it encounters tabs and with -Wno-tabs this
4662            warning is turned off. Before, -Wno-tabs warned and -Wtabs
4663            disabled the warning. As before, this warning is also enabled
4664            by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts
4665            options of -std=.
4666     * Incomplete support for colorizing diagnostics emitted by gfortran
4667       has been added. The option [23]-fdiagnostics-color controls when
4668       color is used in diagnostics. The default value of this option can
4669       be [24]configured when building GCC. The GCC_COLORS environment
4670       variable can be used to customize the colors or disable coloring
4671       completely. Sample diagnostics output:
4672      $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90
4673      test.f90:6:1:
4674
4675       0 continue
4676       1
4677      Error: Zero is not a valid statement label at (1)
4678      test.f90:9:6:
4679
4680         USE foo
4681            1
4682      Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only]
4683
4684     * The -Wuse-without-only option has been added to warn when a USE
4685       statement has no ONLY qualifier and thus implicitly imports all
4686       public entities of the used module.
4687     * Formatted READ and WRITE statements now work correctly in
4688       locale-aware programs. For more information and potential caveats,
4689       see [25]Section 5.3 Thread-safety of the runtime library in the
4690       manual.
4691     * [26]Fortran 2003:
4692          + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and
4693            IEEE_ARITHMETIC) are now supported.
4694     * [27]Fortran 2008:
4695          + [28]Coarrays: Full experimental support of Fortran 2008's
4696            coarrays with -fcoarray=lib except for allocatable/pointer
4697            components of derived-type coarrays. GCC currently only ships
4698            with a single-image library (libcaf_single), but multi-image
4699            support based on MPI and GASNet is provided by the libraries
4700            of the [29]OpenCoarrays project.
4701     * TS18508 Additional Parallel Features in Fortran:
4702          + Support for the collective intrinsic subroutines CO_MAX,
4703            CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added,
4704            including -fcoarray=lib support.
4705          + Support for the new atomic intrinsics has been added,
4706            including -fcoarray=lib support.
4707     * Fortran 2015:
4708          + Support for IMPLICIT NONE (external, type).
4709          + ERROR STOP is now permitted in pure procedures.
4710
4711  Go
4712
4713     * GCC 5 provides a complete implementation of the Go 1.4.2 release.
4714     * Building GCC 5 with Go enabled will install two new programs:
4715       [30]go and [31]gofmt.
4716
4717libgccjit
4718
4719   New in GCC 5 is the ability to build GCC as a shared library for
4720   embedding in other processes (such as interpreters), suitable for
4721   Just-In-Time compilation to machine code.
4722
4723   The shared library has a [32]C API and a [33]C++ wrapper API providing
4724   some "syntactic sugar". There are also bindings available from 3rd
4725   parties for [34]Python and for [35]D.
4726
4727   For example, this library can be used by interpreters for [36]compiling
4728   functions from bytecode to machine code.
4729
4730   The library can also be used for ahead-of-time compilation, enabling
4731   GCC to be plugged into a pre-existing front end. An example of using
4732   this to build a compiler for an esoteric language we'll refer to as
4733   "brainf" can be seen [37]here.
4734
4735   libgccjit is licensed under the GPLv3 (or at your option, any later
4736   version)
4737
4738   It should be regarded as experimental at this time.
4739
4740New Targets and Target Specific Improvements
4741
4742  Reporting stack usage
4743
4744     * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the
4745       -fstack-usage option.
4746
4747  AArch64
4748
4749     * Code generation for the ARM Cortex-A57 processor has been improved.
4750       A more accurate instruction scheduling model for the processor is
4751       now used, and a number of compiler tuning parameters have been set
4752       to offer increased performance when compiling with -mcpu=cortex-a57
4753       or -mtune=cortex-a57.
4754     * A workaround for the ARM Cortex-A53 erratum 835769 has been added
4755       and can be enabled by giving the -mfix-cortex-a53-835769 option.
4756       Alternatively it can be enabled by default by configuring GCC with
4757       the --enable-fix-cortex-a53-835769 option.
4758     * The optional cryptographic extensions to the ARMv8-A architecture
4759       are no longer enabled by default when specifying the
4760       -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53
4761       options. To enable these extensions add +crypto to the value of
4762       -mcpu or -march e.g. -mcpu=cortex-a53+crypto.
4763     * Support has been added for the following processors (GCC
4764       identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and
4765       initial support for its big.LITTLE combination with the ARM
4766       Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx),
4767       Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1).
4768       The GCC identifiers can be used as arguments to the -mcpu or -mtune
4769       options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53.
4770       Using -mcpu=cortex-a72 requires a version of GNU binutils that has
4771       support for the Cortex-A72.
4772     * The transitional options -mlra and -mno-lra have been removed. The
4773       AArch64 backend now uses the local register allocator (LRA) only.
4774
4775  ARM
4776
4777     * Thumb-1 assembly code is now generated in unified syntax. The new
4778       option -masm-syntax-unified specifies whether inline assembly code
4779       is using unified syntax. By default the option is off which means
4780       non-unified syntax is used. However this is subject to change in
4781       future releases. Eventually the non-unified syntax will be
4782       deprecated.
4783     * It is now a configure-time error to use the --with-cpu configure
4784       option with either of --with-tune or --with-arch.
4785     * Code generation for the ARM Cortex-A57 processor has been improved.
4786       A more accurate instruction scheduling model for the processor is
4787       now used, and a number of compiler tuning parameters have been set
4788       to offer increased performance when compiling with -mcpu=cortex-a57
4789       or -mtune=cortex-a57.
4790     * Support has been added for the following processors (GCC
4791       identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and
4792       initial support for its big.LITTLE combination with the ARM
4793       Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and
4794       initial support for its big.LITTLE combination with the ARM
4795       Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7),
4796       Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1).
4797       The GCC identifiers can be used as arguments to the -mcpu or -mtune
4798       options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53.
4799       Using -mcpu=cortex-a72 requires a version of GNU binutils that has
4800       support for the Cortex-A72.
4801     * The deprecated option -mwords-little-endian has been removed.
4802     * The options -mapcs, -mapcs-frame, -mtpcs-frame and
4803       -mtpcs-leaf-frame which are only applicable to the old ABI have
4804       been deprecated.
4805     * The transitional options -mlra and -mno-lra have been removed. The
4806       ARM backend now uses the local register allocator (LRA) only.
4807
4808  AVR
4809
4810     * The compiler no more supports individual devices like ATmega8.
4811       Specifying, say, -mmcu=atmega8 triggers the usage of the
4812       device-specific [38]spec file specs-atmega8 which is part of the
4813       installation and describes options for the sub-processes like
4814       compiler proper, assembler and linker. You can add support for a
4815       new device -mmcu=mydevice as follows:
4816         1. In an empty directory /someplace, create a new directory
4817            device-specs.
4818         2. Copy a device spec file from the installed device-specs
4819            folder, follow the comments in that file and then save it as
4820            /someplace/device-specs/specs-mydevice.
4821         3. Add -B /someplace -mmcu=mydevice to the compiler's
4822            command-line options. Notice that /someplace must specify an
4823            absolute path and that mydevice must not start with "avr".
4824         4. Provided you have a device-specific library libmydevice.a
4825            available, you can put it at /someplace, dito for a
4826            device-specific startup file crtmydevice.o.
4827       The contents of the device spec files depend on the compiler's
4828       configuration, in particular on --with-avrlibc=no and whether or
4829       not it is configured for RTEMS.
4830     * A new command-line option -nodevicelib has been added. It prevents
4831       the compiler from linking against AVR-LibC's device-specific
4832       library libdevice.a.
4833     * The following three command-line options have been added:
4834
4835        -mrmw
4836                Set if the device supports the read-modify-write
4837                instructions LAC, LAS, LAT and XCH.
4838
4839        -mn-flash=size
4840                Specify the flash size of the device in units of 64 KiB,
4841                rounded up to the next integer as needed. This option
4842                affects the availability of the [39]AVR address-spaces.
4843
4844        -mskip-bug
4845                Set if the device is affected by the respective silicon
4846                bug.
4847
4848       In general, you don't need to set these options by hand. The new
4849       device-specific spec file will set them as needed.
4850
4851  IA-32/x86-64
4852
4853     * New [40]ISA extensions support AVX-512{BW,DQ,VL,IFMA,VBMI} of
4854       Intel's CPU codenamed Skylake Server was added to GCC. That
4855       includes inline assembly support, new intrinsics, and basic
4856       autovectorization. These new AVX-512 extensions are available via
4857       the following GCC switches: AVX-512 Vector Length EVEX feature:
4858       -mavx512vl, AVX-512 Byte and Word instructions: -mavx512bw, AVX-512
4859       Dword and Qword instructions: -mavx512dq, AVX-512 FMA-52
4860       instructions: -mavx512ifma and for AVX-512 Vector Bit Manipulation
4861       Instructions: -mavx512vbmi.
4862     * New ISA extensions support Intel MPX was added to GCC. This new
4863       extension is available via the -mmpx compiler switch. Intel MPX is
4864       a set of processor features which, with compiler, run-time library
4865       and OS support, brings increased robustness to software by run-time
4866       checking pointer references against their bounds. In GCC Intel MPX
4867       is supported by Pointer Bounds Checker and libmpx run-time
4868       libraries.
4869     * The new -mrecord-mcount option for -pg generates a Linux kernel
4870       style table of pointers to mcount or __fentry__ calls at the
4871       beginning of functions. The new -mnop-mcount option in addition
4872       also generates nops in place of the __fentry__ or mcount call, so
4873       that a call per function can be later patched in. This can be used
4874       for low overhead tracing or hot code patching.
4875     * The new -malign-data option controls how GCC aligns variables.
4876       -malign-data=compat uses increased alignment compatible with GCC
4877       4.8 and earlier, -malign-data=abi uses alignment as specified by
4878       the psABI, and -malign-data=cacheline uses increased alignment to
4879       match the cache line size. -malign-data=compat is the default.
4880     * The new -mskip-rax-setup option skips setting up the RAX register
4881       when SSE is disabled and there are no variable arguments passed in
4882       vector registers. This can be used to optimize the Linux kernel.
4883
4884  MIPS
4885
4886     * MIPS Releases 3 and 5 are now directly supported. Use the
4887       command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5
4888       to enable code-generation for these processors.
4889     * The Imagination P5600 processor is now supported using the
4890       -march=p5600 command-line option.
4891     * The Cavium Octeon3 processor is now supported using the
4892       -march=octeon3 command-line option.
4893     * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6
4894       command-line options.
4895     * The o32 ABI has been modified and extended. The o32 64-bit
4896       floating-point register support is now obsolete and has been
4897       removed. It has been replaced by three ABI extensions FPXX, FP64A,
4898       and FP64. The meaning of the -mfp64 command-line option has
4899       changed. It is now used to enable the FP64A and FP64 ABI
4900       extensions.
4901          + The FPXX extension requires that code generated to access
4902            double-precision values use even-numbered registers. Code that
4903            adheres to this extension is link-compatible with all other
4904            o32 double-precision ABI variants and will execute correctly
4905            in all hardware FPU modes. The command-line options -mabi=32
4906            -mfpxx can be used to enable this extension. MIPS II is the
4907            minimum processor required.
4908          + The o32 FP64A extension requires that floating-point registers
4909            be 64-bit and odd-numbered single-precision registers are not
4910            allowed. Code that adheres to the o32 FP64A variant is
4911            link-compatible with all other o32 double-precision ABI
4912            variants. The command-line options -mabi=32 -mfp64
4913            -mno-odd-spreg can be used to enable this extension. MIPS32R2
4914            is the minimum processor required.
4915          + The o32 FP64 extension also requires that floating-point
4916            registers be 64-bit, but permits the use of single-precision
4917            registers. Code that adheres to the o32 FP64 variant is
4918            link-compatible with o32 FPXX and o32 FP64A variants only,
4919            i.e. it is not compatible with the original o32
4920            double-precision ABI. The command-line options -mabi=32 -mfp64
4921            -modd-spreg can be used to enable this extension. MIPS32R2 is
4922            the minimum processor required.
4923       The new ABI variants can be enabled by default using the configure
4924       time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32.
4925       It is strongly recommended that all vendors begin to set o32 FPXX
4926       as the default ABI. This will be required to run the generated code
4927       on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and
4928       MIPSR6 cores.
4929     * GCC will now pass all floating-point options to the assembler if
4930       GNU binutils 2.25 is used. As a result, any inline assembly code
4931       that uses hard-float instructions should be amended to include a
4932       .set directive to override the global assembler options when
4933       compiling for soft-float targets.
4934
4935  NDS32
4936
4937     * The variadic function ABI implementation is now compatible with
4938       past Andes toolchains where the caller uses registers to pass
4939       arguments and the callee is in charge of pushing them on stack.
4940     * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have
4941       been removed since they are not yet available in the nds32 port of
4942       GNU binutils.
4943     * A new option -mcmodel=[small|medium|large] supports varied code
4944       models on code generation. The -mgp-direct option became
4945       meaningless and can be discarded.
4946
4947  RX
4948
4949     * A new command line option -mno-allow-string-insns can be used to
4950       disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL,
4951       SWHILE and RMPA instructions. An erratum released by Renesas shows
4952       that it is unsafe to use these instructions on addresses within the
4953       I/O space of the processor. The new option can be used when the
4954       programmer is concerned that the I/O space might be accessed. The
4955       default is still to enable these instructions.
4956
4957  SH
4958
4959     * The compiler will now pass the appropriate --isa= option to the
4960       assembler.
4961     * The default handling for the GBR has been changed from call
4962       clobbered to call preserved. The old behavior can be reinstated by
4963       specifying the option -fcall-used-gbr.
4964     * Support for the SH4A fpchg instruction has been added which will be
4965       utilized when switching between single and double precision FPU
4966       modes.
4967     * The compiler no longer uses the __fpscr_values array for switching
4968       between single and double FPU precision modes on non-SH4A targets.
4969       Instead mode switching will now be performed by storing, modifying
4970       and reloading the FPSCR, so that other FPSCR bits are preserved
4971       across mode switches. The __fpscr_values array that is defined in
4972       libgcc is still present for backwards compatibility, but it will
4973       not be referenced by compiler generated code anymore.
4974     * New builtin functions __builtin_sh_get_fpscr and
4975       __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr
4976       function will mask the specified bits in such a way that the SZ, PR
4977       and FR mode bits will be preserved, while changing the other bits.
4978       These new functions do not reference the __fpscr_values array. The
4979       old functions __set_fpscr and __get_fpscr in libgcc which access
4980       the __fpscr_values array are still present for backwards
4981       compatibility, but their usage is highly discouraged.
4982     * Some improvements to code generated for __atomic built-in
4983       functions.
4984     * When compiling for SH2E the compiler will no longer force the usage
4985       of delay slots for conditional branch instructions bt and bf. The
4986       old behavior can be reinstated (e.g. to work around a hardware bug
4987       in the original SH7055) by specifying the new option
4988       -mcbranch-force-delay-slot.
4989
4990Operating Systems
4991
4992  AIX
4993
4994     * GCC now supports stabs debugging continuation lines to allow long
4995       stabs debug information without overflow that generates AIX linker
4996       errors.
4997
4998  DragonFly BSD
4999
5000     * GCC now supports the DragonFly BSD operating system.
5001
5002  FreeBSD
5003
5004     * GCC now supports the FreeBSD operating system for the arm port
5005       through the arm*-*-freebsd* target triplets.
5006
5007  VxWorks MILS
5008
5009     * GCC now supports the MILS (Multiple Independent Levels of Security)
5010       variant of WindRiver's VxWorks operating system for PowerPC
5011       targets.
5012
5013Other significant improvements
5014
5015     * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option
5016       to set the compiler to use.
5017
5018     * When the new command-line option -freport-bug is used, GCC
5019       automatically generates a developer-friendly reproducer whenever an
5020       internal compiler error is encountered.
5021
5022                                    GCC 5.2
5023
5024   This is the [41]list of problem reports (PRs) from GCC's bug tracking
5025   system that are known to be fixed in the 5.2 release. This list might
5026   not be complete (that is, it is possible that some PRs that have been
5027   fixed are not listed here).
5028
5029Target Specific Changes
5030
5031  IA-32/x86-64
5032
5033     * Support for new AMD instructions monitorx and mwaitx has been
5034       added. This includes new intrinsic and built-in support. It is
5035       enabled through option -mmwaitx. The instructions monitorx and
5036       mwaitx implement the same functionality as the old monitor and
5037       mwait instructions. In addition mwaitx adds a configurable timer.
5038       The timer value is received as third argument and stored in
5039       register %ebx.
5040
5041  S/390, System z, IBM z Systems
5042
5043     * Support for the IBM z13 processor has been added. When using the
5044       -march=z13 option, the compiler will generate code making use of
5045       the new instructions and registers introduced with the vector
5046       extension facility. The -mtune=z13 option enables z13 specific
5047       instruction scheduling without making use of new instructions.
5048       Compiling code with -march=z13 reduces the default alignment of
5049       vector types bigger than 8 bytes to 8. This is an ABI change and
5050       care must be taken when linking modules compiled with different
5051       arch levels which interchange variables containing vector type
5052       values. For newly compiled code the GNU linker will emit a warning.
5053     * The -mzvector option enables a C/C++ language extension. This
5054       extension provides a new keyword vector which can be used to define
5055       vector type variables. (Note: This is not available when enforcing
5056       strict standard compliance e.g. with -std=c99. Either enable GNU
5057       extensions with e.g. -std=gnu99 or use __vector instead of vector.)
5058       Additionally a set of overloaded builtins is provided which is
5059       partially compatible to the PowerPC Altivec builtins. In order to
5060       make use of these builtins the vecintrin.h header file needs to be
5061       included.
5062
5063                                    GCC 5.3
5064
5065   This is the [42]list of problem reports (PRs) from GCC's bug tracking
5066   system that are known to be fixed in the 5.3 release. This list might
5067   not be complete (that is, it is possible that some PRs that have been
5068   fixed are not listed here).
5069
5070Target Specific Changes
5071
5072  IA-32/x86-64
5073
5074     * GCC now supports the Intel CPU named Skylake with AVX-512
5075       extensions through -march=skylake-avx512. The switch enables the
5076       following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW,
5077       AVX-512DQ.
5078
5079  S/390, System z, IBM z Systems
5080
5081     * With this version of GCC IBM z Systems support has been added to
5082       the GO runtime environment. GCC 5.3 has proven to be able to
5083       compile larger GO applications on IBM z Systems.
5084
5085                                    GCC 5.4
5086
5087   This is the [43]list of problem reports (PRs) from GCC's bug tracking
5088   system that are known to be fixed in the 5.4 release. This list might
5089   not be complete (that is, it is possible that some PRs that have been
5090   fixed are not listed here).
5091
5092                                    GCC 5.5
5093
5094   This is the [44]list of problem reports (PRs) from GCC's bug tracking
5095   system that are known to be fixed in the 5.5 release. This list might
5096   not be complete (that is, it is possible that some PRs that have been
5097   fixed are not listed here).
5098
5099Target Specific Changes
5100
5101  IA-32/x86-64
5102
5103     * Support for the [45]deprecated pcommit instruction has been
5104       removed.
5105
5106
5107    For questions related to the use of GCC, please consult these web
5108    pages and the [46]GCC manuals. If that fails, the
5109    [47]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5110    web pages and the development of GCC are welcome on our developer
5111    list at [48]gcc@gcc.gnu.org. All of [49]our lists have public
5112    archives.
5113
5114   Copyright (C) [50]Free Software Foundation, Inc. Verbatim copying and
5115   distribution of this entire article is permitted in any medium,
5116   provided this notice is preserved.
5117
5118   These pages are [51]maintained by the GCC team. Last modified
5119   2019-11-28[52].
5120
5121References
5122
5123   1. http://gcc.gnu.org/gcc-5/changes.html#libstdcxx
5124   2. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html
5125   3. https://savannah.nongnu.org/bugs/?44574
5126   4. https://gcc.gnu.org/wiki/Intel MPX support in the GCC compiler
5127   5. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.pdf
5128   6. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.Examples.pdf
5129   7. https://gcc.gnu.org/wiki/OpenACC
5130   8. https://gcc.gnu.org/install/configure.html
5131   9. https://www.cilkplus.org/
5132  10. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations
5133  11. http://gcc.gnu.org/projects/cxx1y.html
5134  12. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations
5135  13. http://gcc.gnu.org/projects/cxx1y.html
5136  14. http://gcc.gnu.org/projects/cxx1y.html
5137  15. http://gcc.gnu.org/projects/cxx1y.html
5138  16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html
5139  17. http://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html
5140  18. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011
5141  19. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014
5142  20. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014
5143  21. https://sourceware.org/gdb/current/onlinedocs/gdb/Xmethods-In-Python.html
5144  22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html
5145  23. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html
5146  24. https://gcc.gnu.org/install/configure.html
5147  25. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html
5148  26. https://gcc.gnu.org/wiki/Fortran2003Status
5149  27. https://gcc.gnu.org/wiki/Fortran2008Status
5150  28. https://gcc.gnu.org/wiki/Coarray
5151  29. http://www.opencoarrays.org/
5152  30. https://golang.org/cmd/go/
5153  31. https://golang.org/cmd/gofmt/
5154  32. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html
5155  33. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html
5156  34. https://github.com/davidmalcolm/pygccjit
5157  35. https://github.com/ibuclaw/gccjitd
5158  36. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html
5159  37. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html
5160  38. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html
5161  39. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html
5162  40. https://software.intel.com/sites/default/files/managed/39/c5/325462-sdm-vol-1-2abcd-3abcd.pdf
5163  41. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2
5164  42. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3
5165  43. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.4
5166  44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.5
5167  45. https://software.intel.com/en-us/blogs/2016/09/12/deprecate-pcommit-instruction
5168  46. https://gcc.gnu.org/onlinedocs/
5169  47. mailto:gcc-help@gcc.gnu.org
5170  48. mailto:gcc@gcc.gnu.org
5171  49. https://gcc.gnu.org/lists.html
5172  50. https://www.fsf.org/
5173  51. https://gcc.gnu.org/about.html
5174  52. http://validator.w3.org/check/referer
5175======================================================================
5176http://gcc.gnu.org/gcc-4.9/index.html
5177                             GCC 4.9 Release Series
5178
5179   (This release series is no longer supported.)
5180
5181   Aug 3, 2016
5182
5183   The [1]GNU project and the GCC developers are pleased to announce the
5184   release of GCC 4.9.4.
5185
5186   This release is a bug-fix release, containing fixes for regressions in
5187   GCC 4.9.3 relative to previous releases of GCC.
5188
5189Release History
5190
5191   GCC 4.9.4
5192          Aug 3, 2016 ([2]changes, [3]documentation)
5193
5194   GCC 4.9.3
5195          June 26, 2015 ([4]changes, [5]documentation)
5196
5197   GCC 4.9.2
5198          October 30, 2014 ([6]changes, [7]documentation)
5199
5200   GCC 4.9.1
5201          July 16, 2014 ([8]changes, [9]documentation)
5202
5203   GCC 4.9.0
5204          April 22, 2014 ([10]changes, [11]documentation)
5205
5206References and Acknowledgements
5207
5208   GCC used to stand for the GNU C Compiler, but since the compiler
5209   supports several other languages aside from C, it now stands for the
5210   GNU Compiler Collection.
5211
5212   A list of [12]successful builds is updated as new information becomes
5213   available.
5214
5215   The GCC developers would like to thank the numerous people that have
5216   contributed new features, improvements, bug fixes, and other changes as
5217   well as test results to GCC. This [13]amazing group of volunteers is
5218   what makes GCC successful.
5219
5220   For additional information about GCC please refer to the [14]GCC
5221   project web site or contact the [15]GCC development mailing list.
5222
5223   To obtain GCC please use [16]our mirror sites or [17]our version
5224   control system.
5225
5226
5227    For questions related to the use of GCC, please consult these web
5228    pages and the [18]GCC manuals. If that fails, the
5229    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5230    web pages and the development of GCC are welcome on our developer
5231    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
5232    archives.
5233
5234   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
5235   distribution of this entire article is permitted in any medium,
5236   provided this notice is preserved.
5237
5238   These pages are [23]maintained by the GCC team. Last modified
5239   2020-01-14[24].
5240
5241References
5242
5243   1. http://www.gnu.org/
5244   2. http://gcc.gnu.org/gcc-4.9/changes.html
5245   3. http://gcc.gnu.org/onlinedocs/4.9.4/
5246   4. http://gcc.gnu.org/gcc-4.9/changes.html
5247   5. http://gcc.gnu.org/onlinedocs/4.9.3/
5248   6. http://gcc.gnu.org/gcc-4.9/changes.html
5249   7. http://gcc.gnu.org/onlinedocs/4.9.2/
5250   8. http://gcc.gnu.org/gcc-4.9/changes.html
5251   9. http://gcc.gnu.org/onlinedocs/4.9.1/
5252  10. http://gcc.gnu.org/gcc-4.9/changes.html
5253  11. https://gcc.gnu.org/onlinedocs/4.9.0/
5254  12. http://gcc.gnu.org/gcc-4.9/buildstat.html
5255  13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
5256  14. http://gcc.gnu.org/index.html
5257  15. mailto:gcc@gcc.gnu.org
5258  16. http://gcc.gnu.org/mirrors.html
5259  17. http://gcc.gnu.org/git.html
5260  18. https://gcc.gnu.org/onlinedocs/
5261  19. mailto:gcc-help@gcc.gnu.org
5262  20. mailto:gcc@gcc.gnu.org
5263  21. https://gcc.gnu.org/lists.html
5264  22. https://www.fsf.org/
5265  23. https://gcc.gnu.org/about.html
5266  24. http://validator.w3.org/check/referer
5267======================================================================
5268http://gcc.gnu.org/gcc-4.9/changes.html
5269                             GCC 4.9 Release Series
5270                        Changes, New Features, and Fixes
5271
5272Caveats
5273
5274     * The mudflap run time checker has been removed. The mudflap options
5275       remain, but do nothing.
5276     * Support for a number of older systems and recently unmaintained or
5277       untested target ports of GCC has been declared obsolete in GCC 4.9.
5278       Unless there is activity to revive them, the next release of GCC
5279       will have their sources permanently removed.
5280       The following ports for individual systems on particular
5281       architectures have been obsoleted:
5282          + Solaris 9 (*-*-solaris2.9). Details can be found in the
5283            [1]announcement.
5284     * On AArch64, the singleton vector types int64x1_t, uint64x1_t and
5285       float64x1_t exported by arm_neon.h are defined to be the same as
5286       their base types. This results in incorrect application of
5287       parameter passing rules to arguments of types int64x1_t and
5288       uint64x1_t, with respect to the AAPCS64 ABI specification. In
5289       addition, names of C++ functions with parameters of these types
5290       (including float64x1_t) are not mangled correctly. The current
5291       typedef declarations also unintentionally allow implicit casting
5292       between singleton vector types and their base types. These issues
5293       will be resolved in a near future release. See [2]PR60825 for more
5294       information.
5295
5296   More information on porting to GCC 4.9 from previous versions of GCC
5297   can be found in the [3]porting guide for this release.
5298
5299General Optimizer Improvements
5300
5301     * AddressSanitizer, a fast memory error detector, is now available on
5302       ARM.
5303     * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior
5304       detector, has been added and can be enabled via
5305       -fsanitize=undefined. Various computations will be instrumented to
5306       detect undefined behavior at runtime. UndefinedBehaviorSanitizer is
5307       currently available for the C and C++ languages.
5308     * Link-time optimization (LTO) improvements:
5309          + Type merging was rewritten. The new implementation is
5310            significantly faster and uses less memory.
5311          + Better partitioning algorithm resulting in less streaming
5312            during link time.
5313          + Early removal of virtual methods reduces the size of object
5314            files and improves link-time memory usage and compile time.
5315          + Function bodies are now loaded on-demand and released early
5316            improving overall memory usage at link time.
5317          + C++ hidden keyed methods can now be optimized out.
5318          + When using a linker plugin, compiling with the -flto option
5319            now generates slim object files (.o) which only contain
5320            intermediate language representation for LTO. Use
5321            -ffat-lto-objects to create files which contain additionally
5322            the object code. To generate static libraries suitable for LTO
5323            processing, use gcc-ar and gcc-ranlib; to list symbols from a
5324            slim object file use gcc-nm. (This requires that ar, ranlib
5325            and nm have been compiled with plugin support.)
5326       Memory usage building Firefox with debug enabled was reduced from
5327       15GB to 3.5GB; link time from 1700 seconds to 350 seconds.
5328     * Inter-procedural optimization improvements:
5329          + New type inheritance analysis module improving
5330            devirtualization. Devirtualization now takes into account
5331            anonymous name-spaces and the C++11 final keyword.
5332          + New speculative devirtualization pass (controlled by
5333            -fdevirtualize-speculatively.
5334          + Calls that were speculatively made direct are turned back to
5335            indirect where direct call is not cheaper.
5336          + Local aliases are introduced for symbols that are known to be
5337            semantically equivalent across shared libraries improving
5338            dynamic linking times.
5339     * Feedback directed optimization improvements:
5340          + Profiling of programs using C++ inline functions is now more
5341            reliable.
5342          + New time profiling determines typical order in which functions
5343            are executed.
5344          + A new function reordering pass (controlled by
5345            -freorder-functions) significantly reduces startup time of
5346            large applications. Until binutils support is completed, it is
5347            effective only with link-time optimization.
5348          + Feedback driven indirect call removal and devirtualization now
5349            handle cross-module calls when link-time optimization is
5350            enabled.
5351
5352New Languages and Language specific improvements
5353
5354     * Version 4.0 of the [4]OpenMP specification is now supported in the
5355       C and C++ compilers and starting with the 4.9.1 release also in the
5356       Fortran compiler. The new -fopenmp-simd option can be used to
5357       enable OpenMP's SIMD directives while ignoring other OpenMP
5358       directives. The new [5]-fsimd-cost-model= option permits to tune
5359       the vectorization cost model for loops annotated with OpenMP and
5360       Cilk Plus simd directives. -Wopenmp-simd warns when the current
5361       cost model overrides simd directives set by the user.
5362     * The -Wdate-time option has been added for the C, C++ and Fortran
5363       compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__
5364       macros are used. Those macros might prevent bit-wise-identical
5365       reproducible compilations.
5366
5367  Ada
5368
5369     * GNAT switched to Ada 2012 instead of Ada 2005 by default.
5370
5371  C family
5372
5373     * Support for colorizing diagnostics emitted by GCC has been added.
5374       The [6]-fdiagnostics-color=auto will enable it when outputting to
5375       terminals, -fdiagnostics-color=always unconditionally. The
5376       GCC_COLORS environment variable can be used to customize the colors
5377       or disable coloring. If GCC_COLORS variable is present in the
5378       environment, the default is -fdiagnostics-color=auto, otherwise
5379       -fdiagnostics-color=never.
5380       Sample diagnostics output:
5381    $ g++ -fdiagnostics-color=always -S -Wall test.C
5382    test.C: In function ‘int foo()’:
5383    test.C:1:14: warning: no return statement in function returning non-void [-W
5384return-type]
5385     int foo () { }
5386                  ^
5387    test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use
5388 -ftemplate-depth= to increase the maximum) instantiating ‘struct X<100>’
5389     template <int N> struct X { static const int value = X<N-1>::value; }; temp
5390late struct X<1000>;
5391                                                  ^
5392    test.C:2:46:   recursively required from ‘const int X<999>::value’
5393    test.C:2:46:   required from ‘const int X<1000>::value’
5394    test.C:2:88:   required from here
5395
5396    test.C:2:46: error: incomplete type ‘X<100>’ used in nested name specifier
5397
5398     * With the new [7]#pragma GCC ivdep, the user can assert that there
5399       are no loop-carried dependencies which would prevent concurrent
5400       execution of consecutive iterations using SIMD (single instruction
5401       multiple data) instructions.
5402     * Support for [8]Cilk Plus has been added and can be enabled with the
5403       -fcilkplus option. Cilk Plus is an extension to the C and C++
5404       languages to support data and task parallelism. The present
5405       implementation follows ABI version 1.2; all features but _Cilk_for
5406       have been implemented.
5407
5408  C
5409
5410     * ISO C11 atomics (the _Atomic type specifier and qualifier and the
5411       <stdatomic.h> header) are now supported.
5412     * ISO C11 generic selections (_Generic keyword) are now supported.
5413     * ISO C11 thread-local storage (_Thread_local, similar to GNU C
5414       __thread) is now supported.
5415     * ISO C11 support is now at a similar level of completeness to ISO
5416       C99 support: substantially complete modulo bugs, extended
5417       identifiers (supported except for corner cases when
5418       -fextended-identifiers is used), floating-point issues (mainly but
5419       not entirely relating to optional C99 features from Annexes F and
5420       G) and the optional Annexes K (Bounds-checking interfaces) and L
5421       (Analyzability).
5422     * A new C extension __auto_type provides a subset of the
5423       functionality of C++11 auto in GNU C.
5424
5425  C++
5426
5427     * The G++ implementation of [9]C++1y return type deduction for normal
5428       functions has been updated to conform to [10]N3638, the proposal
5429       accepted into the working paper. Most notably, it adds
5430       decltype(auto) for getting decltype semantics rather than the
5431       template argument deduction semantics of plain auto:
5432
5433int& f();
5434         auto  i1 = f(); // int
5435decltype(auto) i2 = f(); // int&
5436
5437     * G++ supports [11]C++1y lambda capture initializers:
5438
5439[x = 42]{ ... };
5440
5441       Actually, they have been accepted since GCC 4.5, but now the
5442       compiler doesn't warn about them with -std=c++1y, and supports
5443       parenthesized and brace-enclosed initializers as well.
5444     * G++ supports [12]C++1y variable length arrays. G++ has supported
5445       GNU/C99-style VLAs for a long time, but now additionally supports
5446       initializers and lambda capture by reference. In C++1y mode G++
5447       will complain about VLA uses that are not permitted by the draft
5448       standard, such as forming a pointer to VLA type or applying sizeof
5449       to a VLA variable. Note that it now appears that VLAs will not be
5450       part of C++14, but will be part of a separate document and then
5451       perhaps C++17.
5452
5453void f(int n) {
5454  int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3
5455  [&a]{ for (int i : a) { cout << i << endl; } }();
5456  &a; // error, taking address of VLA
5457}
5458
5459     * G++ supports the [13]C++1y [[deprecated]] attribute modulo bugs in
5460       the underlying [[gnu::deprecated]] attribute. Classes and functions
5461       can be marked deprecated and a diagnostic message added:
5462
5463class A;
5464int bar(int n);
5465#if __cplusplus > 201103
5466class [[deprecated("A is deprecated in C++14; Use B instead")]] A;
5467[[deprecated("bar is unsafe; use foo() instead")]]
5468int bar(int n);
5469
5470int foo(int n);
5471class B;
5472#endif
5473A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead
5474int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo
5475o() instead
5476
5477     * G++ supports [14]C++1y digit separators. Long numeric literals can
5478       be subdivided with a single quote ' to enhance readability:
5479
5480int i = 1048576;
5481int j = 1'048'576;
5482int k = 0x10'0000;
5483int m = 0'004'000'000;
5484int n = 0b0001'0000'0000'0000'0000'0000;
5485
5486double x = 1.602'176'565e-19;
5487double y = 1.602'176'565e-1'9;
5488
5489     * G++ supports [15]C++1y generic (polymorphic) lambdas.
5490
5491// a functional object that will increment any type
5492auto incr = [](auto x) { return x++; };
5493
5494     * As a GNU extension, G++ supports explicit template parameter syntax
5495       for generic lambdas. This can be combined in the expected way with
5496       the standard auto syntax.
5497
5498// a functional object that will add two like-type objects
5499auto add = [] <typename T> (T a, T b) { return a + b; };
5500
5501     * G++ supports unconstrained generic functions as specified by §4.1.2
5502       and §5.1.1 of [16]N3889: Concepts Lite Specification. Briefly, auto
5503       may be used as a type-specifier in a parameter declaration of any
5504       function declarator in order to introduce an implicit function
5505       template parameter, akin to generic lambdas.
5506
5507// the following two function declarations are equivalent
5508auto incr(auto x) { return x++; }
5509template <typename T>
5510auto incr(T x) { return x++; }
5511
5512    Runtime Library (libstdc++)
5513
5514     * [17]Improved support for C++11, including:
5515          + support for <regex>;
5516          + The associative containers in <map> and <set> and the
5517            unordered associative containers in <unordered_map> and
5518            <unordered_set> meet the allocator-aware container
5519            requirements;
5520     * [18]Improved experimental support for the upcoming ISO C++
5521       standard, C++14, including:
5522          + fixing constexpr member functions without const;
5523          + implementation of the std::exchange() utility function;
5524          + addressing tuples by type;
5525          + implemention of std::make_unique;
5526          + implemention of std::shared_lock;
5527          + making std::result_of SFINAE-friendly;
5528          + adding operator() to std::integral_constant;
5529          + adding user-defined literals for standard library types
5530            std::basic_string, std::chrono::duration, and std::complex;
5531          + adding two range overloads to non-modifying sequence oprations
5532            std::equal and std::mismatch;
5533          + adding IO manipulators for quoted strings;
5534          + adding constexpr members to <utility>, <complex>, <chrono>,
5535            and some containers;
5536          + adding compile-time std::integer_sequence;
5537          + adding cleaner transformation traits;
5538          + making <functional>s operator functors easier to use and more
5539            generic;
5540     * An implementation of std::experimental::optional.
5541     * An implementation of std::experimental::string_view.
5542     * The non-standard function std::copy_exception has been deprecated
5543       and will be removed in a future version. std::make_exception_ptr
5544       should be used instead.
5545
5546  Fortran
5547
5548     * Compatibility notice:
5549          + Module files: The version of the module files (.mod) has been
5550            incremented; additionally, module files are now compressed.
5551            Fortran MODULEs compiled by earlier GCC versions have to be
5552            recompiled, when they are USEd by files compiled with GCC 4.9.
5553            GCC 4.9 is not able to read .mod files of earlier GCC
5554            versions; attempting to do so gives an error message. Note:
5555            The ABI of the produced assembler data itself has not changed:
5556            object files and libraries are fully compatible with older
5557            versions (except as stated below).
5558          + ABI changes:
5559               o The [19]argument passing ABI has changed for scalar dummy
5560                 arguments of type INTEGER, REAL, COMPLEX and LOGICAL,
5561                 which have both the VALUE and the OPTIONAL attributes.
5562               o To support finalization the virtual table associated with
5563                 polymorphic variables has changed. Code containing CLASS
5564                 should be recompiled, including all files which define
5565                 derived types involved in the type definition used by
5566                 polymorphic variables. (Note: Due to the incremented
5567                 module version, trying to mix old code with new code will
5568                 usually give an error message.)
5569          + GNU Fortran no longer deallocates allocatable variables or
5570            allocatable components of variables declared in the main
5571            program. Since Fortran 2008, the standard explicitly states
5572            that variables declared in the Fortran main program
5573            automatically have the SAVE attribute.
5574          + When opening files, the close-on-exec flag is set if the
5575            system supports such a feature. This is generally considered
5576            good practice these days, but if there is a need to pass file
5577            descriptors to child processes the parent process must now
5578            remember to clear the close-on-exec flag by calling fcntl(),
5579            e.g. via ISO_C_BINDING, before executing the child process.
5580     * The deprecated command-line option -fno-whole-file has been
5581       removed. (-fwhole-file is the default since GCC 4.6.)
5582       -fwhole-file/-fno-whole-file continue to be accepted but do not
5583       influence the code generation.
5584     * The compiler no longer unconditionally warns about DO loops with
5585       zero iterations. This warning is now controlled by the -Wzerotrip
5586       option, which is implied by -Wall.
5587     * The new NO_ARG_CHECK attribute of the [20]!GCC$ directive can be
5588       used to disable the type-kind-rank (TKR) argument check for a dummy
5589       argument. The feature is similar to ISO/IEC TS 29133:2012's
5590       TYPE(*), except that it additionally also disables the rank check.
5591       Variables with NO_ARG_CHECK have to be dummy arguments and may only
5592       be used as argument to ISO_C_BINDING's C_LOC and as actual argument
5593       to another NO_ARG_CHECK dummy argument; also the other constraints
5594       of TYPE(*) apply. The dummy arguments should be declared as scalar
5595       or assumed-size variable of type type(*) (recommended) – or of type
5596       integer, real, complex or logical. With NO_ARG_CHECK, a pointer to
5597       the data without further type or shape information is passed,
5598       similar to C's void*. Note that also TS 29113's
5599       type(*),dimension(..) accepts arguments of any type and rank;
5600       contrary to NO_ARG_CHECK assumed-rank arguments pass an array
5601       descriptor which contains the array shape and stride of the
5602       argument.
5603     * [21]Fortran 2003:
5604          + Finalization is now supported. It is currently only done for a
5605            subset of those situations in which it should occur.
5606          + Experimental support for scalar character components with
5607            deferred length (i.e. allocatable string length) in derived
5608            types has been added. (Deferred-length character variables are
5609            supported since GCC 4.6.)
5610     * [22]Fortran 2008:
5611          + When STOP or ERROR STOP are used to terminate the execution
5612            and any exception (but inexact) is signaling, a warning is
5613            printed to ERROR_UNIT, indicating which exceptions are
5614            signaling. The [23]-ffpe-summary= command-line option can be
5615            used to fine-tune for which exceptions the warning should be
5616            shown.
5617          + Rounding on input (READ) is now handled on systems where
5618            strtod honours the rounding mode. (For output, rounding is
5619            supported since GCC 4.5.) Note that for input, the compatible
5620            rounding mode is handled as nearest (i.e., rounding to an even
5621            least significant [cf. IEC 60559:1989] for a tie, while
5622            compatible rounds away from zero in that case).
5623
5624  Go
5625
5626     * GCC 4.9 provides a complete implementation of the Go 1.2.1 release.
5627
5628New Targets and Target Specific Improvements
5629
5630  AArch64
5631
5632     * The ARMv8-A crypto and CRC instructions are now supported through
5633       intrinsics. These are enabled when the architecture supports these
5634       and are available through the -march=armv8-a+crc and
5635       -march=armv8-a+crypto options.
5636     * Initial support for ILP32 has now been added to the compiler. This
5637       is now available through the command-line option -mabi=ilp32.
5638       Support for ILP32 is considered experimental as the ABI
5639       specification is still beta.
5640     * Coverage of more of the ISA including the SIMD extensions has been
5641       added. The Advanced SIMD intrinsics have also been improved.
5642     * The new local register allocator (LRA) is now on by default for the
5643       AArch64 backend.
5644     * The REE (Redundant extension elimination) pass has now been enabled
5645       by default for the AArch64 backend.
5646     * Tuning for the Cortex-A53 and Cortex-A57 has been improved.
5647     * Initial big.LITTLE tuning support for the combination of Cortex-A57
5648       and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53
5649       option.
5650     * A number of structural changes have been made to both the ARM and
5651       AArch64 backends to facilitate improved code-generation.
5652     * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769
5653       has been added and can be enabled by giving the
5654       -mfix-cortex-a53-835769 option. Alternatively it can be enabled by
5655       default by configuring GCC with the --enable-fix-cortex-a53-835769
5656       option.
5657
5658  ARC
5659
5660     * A port for Synopsys Designware ARC has been contributed by Embecosm
5661       and Synopsys Inc.
5662
5663  ARM
5664
5665     * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been
5666       disabled by default. This was found to generate better code in only
5667       a small number of cases. It can be turned back on with the
5668       -mneon-for-64bits option.
5669     * Further support for the ARMv8-A architecture, notably implementing
5670       the restriction around IT blocks in the Thumb32 instruction set has
5671       been added. The -mrestrict-it option can be used with
5672       -march=armv7-a or the -march=armv7ve options to make code
5673       generation fully compatible with the deprecated instructions in
5674       ARMv8-A.
5675     * Support has now been added for the ARMv7ve variant of the
5676       architecture. This can be used by the -march=armv7ve option.
5677     * The ARMv8-A crypto and CRC instructions are now supported through
5678       intrinsics and are available through the -march=armv8-a+crc and
5679       mfpu=crypto-neon-fp-armv8 options.
5680     * LRA is now on by default for the ARM target. This can be turned off
5681       using the -mno-lra option. This option is a purely transitionary
5682       command-line option and will be removed in a future release. We are
5683       interested in any bug reports regarding functional and performance
5684       regressions with LRA.
5685     * A new option -mslow-flash-data to improve performance of programs
5686       fetching data on slow flash memory has now been introduced for the
5687       ARMv7-M profile cores.
5688     * A new option -mpic-data-is-text-relative for targets that allows
5689       data segments to be relative to text segments has been added. This
5690       is on by default for all targets except VxWorks RTP.
5691     * A number of infrastructural changes have been made to both the ARM
5692       and AArch64 backends to facilitate improved code-generation.
5693     * GCC now supports Cortex-A12 and the Cortex-R7 through the
5694       -mcpu=cortex-a12 and -mcpu=cortex-r7 options.
5695     * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the
5696       -mcpu=cortex-a57 and -mcpu=cortex-a53 options.
5697     * Initial big.LITTLE tuning support for the combination of Cortex-A57
5698       and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53
5699       option. Similar support was added for the combination of Cortex-A15
5700       and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option.
5701     * Further performance optimizations for the Cortex-A15 and the
5702       Cortex-M4 have been added.
5703     * A number of code generation improvements for Thumb2 to reduce code
5704       size when compiling for the M-profile processors.
5705
5706  AVR
5707
5708     * A new command-line option -mfract-convert-truncate has been added.
5709       It allows compiler to use truncation instead of rounding towards
5710       zero for fractional fixed-point types.
5711
5712  IA-32/x86-64
5713
5714     * -mfpmath=sse is now implied by -ffast-math on all targets where
5715       SSE2 is supported.
5716     * Intel AVX-512 support was added to GCC. That includes inline
5717       assembly support, new registers and extending existing ones, new
5718       intrinsics (covered by corresponding testsuite), and basic
5719       autovectorization. AVX-512 instructions are available via the
5720       following GCC switches: AVX-512 foundation instructions: -mavx512f,
5721       AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and
5722       reciprocal instructions: -mavx512er, AVX-512 conflict detection
5723       instructions: -mavx512cd.
5724     * It is now possible to call x86 intrinsics from select functions in
5725       a file that are tagged with the corresponding target attribute
5726       without having to compile the entire file with the -mxxx option.
5727       This improves the usability of x86 intrinsics and is particularly
5728       useful when doing [24]Function Multiversioning.
5729     * GCC now supports the new Intel microarchitecture named Silvermont
5730       through -march=silvermont.
5731     * GCC now supports the new Intel microarchitecture named Broadwell
5732       through -march=broadwell.
5733     * Optimizing for other Intel microarchitectures have been renamed to
5734       -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell.
5735     * -march=generic has been retuned for better support of Intel core
5736       and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel
5737       Pentium-M, and Pentium4 based CPUs is no longer considered
5738       important for generic.
5739     * -mtune=intel can now be used to generate code running well on the
5740       most current Intel processors, which are Haswell and Silvermont for
5741       GCC 4.9.
5742     * Support to encode 32-bit assembly instructions in 16-bit format is
5743       now available through the -m16 command-line option.
5744     * Better inlining of memcpy and memset that is aware of value ranges
5745       and produces shorter alignment prologues.
5746     * -mno-accumulate-outgoing-args is now honored when unwind
5747       information is output. Argument accumulation is also now turned off
5748       for portions of programs optimized for size.
5749     * Support for new AMD family 15h processors (Excavator core) is now
5750       available through the -march=bdver4 and -mtune=bdver4 options.
5751
5752  MSP430
5753
5754     * A new command-line option -mcpu= has been added to the MSP430
5755       backend. This option is used to specify the ISA to be used.
5756       Accepted values are msp430 (the default), msp430x and msp430xv2.
5757       The ISA is no longer deduced from the -mmcu= option as there are
5758       far too many different MCU names. The -mmcu= option is still
5759       supported, and this is still used to select linker scripts and
5760       generate a C preprocessor symbol that will be recognised by the
5761       msp430.h header file.
5762
5763  NDS32
5764
5765     * A new nds32 port supports the 32-bit architecture from Andes
5766       Technology Corporation.
5767     * The port provides initial support for the V2, V3, V3m instruction
5768       set architectures.
5769
5770  Nios II
5771
5772     * A port for the Altera Nios II has been contributed by Mentor
5773       Graphics.
5774
5775  PowerPC / PowerPC64 / RS6000
5776
5777     * GCC now supports Power ISA 2.07, which includes support for
5778       Hardware Transactional Memory (HTM), Quadword atomics and several
5779       VMX and VSX additions, including Crypto, 64-bit integer, 128-bit
5780       integer and decimal integer operations.
5781     * Support for the POWER8 processor is now available through the
5782       -mcpu=power8 and -mtune=power8 options.
5783     * The libitm library has been modified to add a HTM fastpath that
5784       automatically uses POWER's HTM hardware instructions when it is
5785       executing on a HTM enabled processor.
5786     * Support for the new powerpc64le-linux platform has been added. It
5787       defaults to generating code that conforms to the ELFV2 ABI.
5788
5789  S/390, System z
5790
5791     * Support for the Transactional Execution Facility included with the
5792       IBM zEnterprise zEC12 processor has been added. A set of GCC style
5793       builtins as well as XLC style builtins are provided. The builtins
5794       are enabled by default when using the -march=zEC12 option but can
5795       explicitly be disabled with -mno-htm. Using the GCC builtins also
5796       libitm supports hardware transactions on S/390.
5797     * The hotpatch features allows to prepare functions for hotpatching.
5798       A certain amount of bytes is reserved before the function entry
5799       label plus a NOP is inserted at its very beginning to implement a
5800       backward jump when applying a patch. The feature can either be
5801       enabled per compilation unit via the command-line option -mhotpatch
5802       or per function using the hotpatch attribute.
5803     * The shrink wrap optimization is now supported on S/390 and enabled
5804       by default.
5805     * A major rework of the routines to determine which registers need to
5806       be saved and restored in function prologue/epilogue now allow to
5807       use floating point registers as save slots. This will happen for
5808       certain leaf function with -march=z10 or higher.
5809     * The LRA rtl pass replaces reload by default on S/390.
5810
5811  RX
5812
5813     * The port now allows to specify the RX100, RX200, and RX600
5814       processors with the command-line options -mcpu=rx100, -mcpu=rx200
5815       and -mcpu=rx600.
5816
5817  SH
5818
5819     * Minor improvements to code generated for integer arithmetic and
5820       code that involves the T bit.
5821     * Added support for the SH2A clips and clipu instructions. The
5822       compiler will now try to utilize them for min/max expressions such
5823       as max (-128, min (127, x)).
5824     * Added support for the cmp/str instruction through built-in
5825       functions such as __builtin_strlen. When not optimizing for size,
5826       the compiler will now expand calls to e.g. strlen as an inlined
5827       sequences which utilize the cmp/str instruction.
5828     * Improved code generated around volatile memory loads and stores.
5829     * The option -mcbranchdi has been deprecated. Specifying it will
5830       result in a warning and will not influence code generation.
5831     * The option -mcmpeqdi has been deprecated. Specifying it will result
5832       in a warning and will not influence code generation.
5833
5834GCC 4.9.1
5835
5836   This is the [25]list of problem reports (PRs) from GCC's bug tracking
5837   system that are known to be fixed in the 4.9.1 release. This list might
5838   not be complete (that is, it is possible that some PRs that have been
5839   fixed are not listed here).
5840
5841   Version 4.0 of the OpenMP specification is supported even in Fortran,
5842   not just C and C++.
5843
5844GCC 4.9.2
5845
5846   This is the [26]list of problem reports (PRs) from GCC's bug tracking
5847   system that are known to be fixed in the 4.9.2 release. This list might
5848   not be complete (that is, it is possible that some PRs that have been
5849   fixed are not listed here).
5850
5851GCC 4.9.3
5852
5853   This is the [27]list of problem reports (PRs) from GCC's bug tracking
5854   system that are known to be fixed in the 4.9.3 release. This list might
5855   not be complete (that is, it is possible that some PRs that have been
5856   fixed are not listed here).
5857
5858GCC 4.9.4
5859
5860   This is the [28]list of problem reports (PRs) from GCC's bug tracking
5861   system that are known to be fixed in the 4.9.4 release. This list might
5862   not be complete (that is, it is possible that some PRs that have been
5863   fixed are not listed here).
5864
5865
5866    For questions related to the use of GCC, please consult these web
5867    pages and the [29]GCC manuals. If that fails, the
5868    [30]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5869    web pages and the development of GCC are welcome on our developer
5870    list at [31]gcc@gcc.gnu.org. All of [32]our lists have public
5871    archives.
5872
5873   Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and
5874   distribution of this entire article is permitted in any medium,
5875   provided this notice is preserved.
5876
5877   These pages are [34]maintained by the GCC team. Last modified
5878   2019-11-28[35].
5879
5880References
5881
5882   1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html
5883   2. https://gcc.gnu.org/PR60825
5884   3. https://gcc.gnu.org/gcc-4.9/porting_to.html
5885   4. https://www.openmp.org/specifications/
5886   5. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908
5887   6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252
5888   7. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html
5889   8. https://www.cilkplus.org/
5890   9. http://gcc.gnu.org/projects/cxx1y.html
5891  10. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html
5892  11. http://gcc.gnu.org/projects/cxx1y.html
5893  12. http://gcc.gnu.org/projects/cxx1y.html
5894  13. http://gcc.gnu.org/projects/cxx1y.html
5895  14. http://gcc.gnu.org/projects/cxx1y.html
5896  15. http://gcc.gnu.org/projects/cxx1y.html
5897  16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf
5898  17. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011
5899  18. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014
5900  19. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html
5901  20. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html
5902  21. https://gcc.gnu.org/wiki/Fortran2003Status
5903  22. https://gcc.gnu.org/wiki/Fortran2008Status
5904  23. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html
5905  24. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html
5906  25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1
5907  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2
5908  27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3
5909  28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.4
5910  29. https://gcc.gnu.org/onlinedocs/
5911  30. mailto:gcc-help@gcc.gnu.org
5912  31. mailto:gcc@gcc.gnu.org
5913  32. https://gcc.gnu.org/lists.html
5914  33. https://www.fsf.org/
5915  34. https://gcc.gnu.org/about.html
5916  35. http://validator.w3.org/check/referer
5917======================================================================
5918http://gcc.gnu.org/gcc-4.8/index.html
5919                             GCC 4.8 Release Series
5920
5921   (This release series is no longer supported.)
5922
5923   June 23, 2015
5924
5925   The [1]GNU project and the GCC developers are pleased to announce the
5926   release of GCC 4.8.5.
5927
5928   This release is a bug-fix release, containing fixes for regressions in
5929   GCC 4.8.4 relative to previous releases of GCC.
5930
5931Release History
5932
5933   GCC 4.8.5
5934          June 23, 2015 ([2]changes, [3]documentation)
5935
5936   GCC 4.8.4
5937          December 19, 2014 ([4]changes, [5]documentation)
5938
5939   GCC 4.8.3
5940          May 22, 2014 ([6]changes, [7]documentation)
5941
5942   GCC 4.8.2
5943          October 16, 2013 ([8]changes, [9]documentation)
5944
5945   GCC 4.8.1
5946          May 31, 2013 ([10]changes, [11]documentation)
5947
5948   GCC 4.8.0
5949          March 22, 2013 ([12]changes, [13]documentation)
5950
5951References and Acknowledgements
5952
5953   GCC used to stand for the GNU C Compiler, but since the compiler
5954   supports several other languages aside from C, it now stands for the
5955   GNU Compiler Collection.
5956
5957   A list of [14]successful builds is updated as new information becomes
5958   available.
5959
5960   The GCC developers would like to thank the numerous people that have
5961   contributed new features, improvements, bug fixes, and other changes as
5962   well as test results to GCC. This [15]amazing group of volunteers is
5963   what makes GCC successful.
5964
5965   For additional information about GCC please refer to the [16]GCC
5966   project web site or contact the [17]GCC development mailing list.
5967
5968   To obtain GCC please use [18]our mirror sites or [19]our version
5969   control system.
5970
5971
5972    For questions related to the use of GCC, please consult these web
5973    pages and the [20]GCC manuals. If that fails, the
5974    [21]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5975    web pages and the development of GCC are welcome on our developer
5976    list at [22]gcc@gcc.gnu.org. All of [23]our lists have public
5977    archives.
5978
5979   Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and
5980   distribution of this entire article is permitted in any medium,
5981   provided this notice is preserved.
5982
5983   These pages are [25]maintained by the GCC team. Last modified
5984   2020-01-14[26].
5985
5986References
5987
5988   1. http://www.gnu.org/
5989   2. http://gcc.gnu.org/gcc-4.8/changes.html
5990   3. https://gcc.gnu.org/onlinedocs/4.8.5/
5991   4. http://gcc.gnu.org/gcc-4.8/changes.html
5992   5. https://gcc.gnu.org/onlinedocs/4.8.4/
5993   6. http://gcc.gnu.org/gcc-4.8/changes.html
5994   7. https://gcc.gnu.org/onlinedocs/4.8.3/
5995   8. http://gcc.gnu.org/gcc-4.8/changes.html
5996   9. https://gcc.gnu.org/onlinedocs/4.8.2/
5997  10. http://gcc.gnu.org/gcc-4.8/changes.html
5998  11. https://gcc.gnu.org/onlinedocs/4.8.1/
5999  12. http://gcc.gnu.org/gcc-4.8/changes.html
6000  13. https://gcc.gnu.org/onlinedocs/4.8.0/
6001  14. http://gcc.gnu.org/gcc-4.8/buildstat.html
6002  15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6003  16. http://gcc.gnu.org/index.html
6004  17. mailto:gcc@gcc.gnu.org
6005  18. http://gcc.gnu.org/mirrors.html
6006  19. http://gcc.gnu.org/git.html
6007  20. https://gcc.gnu.org/onlinedocs/
6008  21. mailto:gcc-help@gcc.gnu.org
6009  22. mailto:gcc@gcc.gnu.org
6010  23. https://gcc.gnu.org/lists.html
6011  24. https://www.fsf.org/
6012  25. https://gcc.gnu.org/about.html
6013  26. http://validator.w3.org/check/referer
6014======================================================================
6015http://gcc.gnu.org/gcc-4.8/changes.html
6016                             GCC 4.8 Release Series
6017                        Changes, New Features, and Fixes
6018
6019Caveats
6020
6021   GCC now uses C++ as its implementation language. This means that to
6022   build GCC from sources, you will need a C++ compiler that understands
6023   C++ 2003. For more details on the rationale and specific changes,
6024   please refer to the [1]C++ conversion page.
6025
6026   To enable the Graphite framework for loop optimizations you now need
6027   CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from
6028   the [2]GCC infrastructure directory. The installation manual contains
6029   more information about requirements to build GCC.
6030
6031   GCC now uses a more aggressive analysis to derive an upper bound for
6032   the number of iterations of loops using constraints imposed by language
6033   standards. This may cause non-conforming programs to no longer work as
6034   expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new
6035   option, -fno-aggressive-loop-optimizations, was added to disable this
6036   aggressive analysis. In some loops that have known constant number of
6037   iterations, but undefined behavior is known to occur in the loop before
6038   reaching or during the last iteration, GCC will warn about the
6039   undefined behavior in the loop instead of deriving lower upper bound of
6040   the number of iterations for the loop. The warning can be disabled with
6041   -Wno-aggressive-loop-optimizations.
6042
6043   On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules
6044   for the layout of vectors that could lead to wrong code being
6045   generated. Vectors larger than 8 bytes in size are now by default
6046   aligned to an 8-byte boundary. This is an ABI change: code that makes
6047   explicit use of vector types may be incompatible with binary objects
6048   built with older versions of GCC. Auto-vectorized code is not affected
6049   by this change.
6050
6051   On AVR, support has been removed for the command-line option
6052   -mshort-calls deprecated in GCC 4.7.
6053
6054   On AVR, the configure option --with-avrlibc supported since GCC 4.7.2
6055   is turned on per default for all non-RTEMS configurations. This option
6056   arranges for a better integration of [3]AVR Libc with avr-gcc. For
6057   technical details, see [4]PR54461. To turn off the option in non-RTEMS
6058   configurations, use --with-avrlibc=no. If the compiler is configured
6059   for RTEMS, the option is always turned off.
6060
6061   More information on porting to GCC 4.8 from previous versions of GCC
6062   can be found in the [5]porting guide for this release.
6063
6064General Optimizer Improvements (and Changes)
6065
6066     * DWARF4 is now the default when generating DWARF debug information.
6067       When -g is used on a platform that uses DWARF debugging
6068       information, GCC will now default to -gdwarf-4
6069       -fno-debug-types-section.
6070       GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information
6071       consumers support DWARF4 by default. Before GCC 4.8 the default
6072       version used was DWARF2. To make GCC 4.8 generate an older DWARF
6073       version use -g together with -gdwarf-2 or -gdwarf-3. The default
6074       for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf.
6075     * A new general optimization level, -Og, has been introduced. It
6076       addresses the need for fast compilation and a superior debugging
6077       experience while providing a reasonable level of run-time
6078       performance. Overall experience for development should be better
6079       than the default optimization level -O0.
6080     * A new option -ftree-partial-pre was added to control the partial
6081       redundancy elimination (PRE) optimization. This option is enabled
6082       by default at the -O3 optimization level, and it makes PRE more
6083       aggressive.
6084     * The option -fconserve-space has been removed; it was no longer
6085       useful on most targets since GCC supports putting variables into
6086       BSS without making them common.
6087     * The struct reorg and matrix reorg optimizations (command-line
6088       options -fipa-struct-reorg and -fipa-matrix-reorg) have been
6089       removed. They did not always work correctly, nor did they work with
6090       link-time optimization (LTO), hence were only applicable to
6091       programs consisting of a single translation unit.
6092     * Several scalability bottle-necks have been removed from GCC's
6093       optimization passes. Compilation of extremely large functions, e.g.
6094       due to the use of the flatten attribute in the "Eigen" C++ linear
6095       algebra templates library, is significantly faster than previous
6096       releases of GCC.
6097     * Link-time optimization (LTO) improvements:
6098          + LTO partitioning has been rewritten for better reliability and
6099            maintanibility. Several important bugs leading to link
6100            failures have been fixed.
6101     * Interprocedural optimization improvements:
6102          + A new symbol table has been implemented. It builds on existing
6103            callgraph and varpool modules and provide a new API. Unusual
6104            symbol visibilities and aliases are handled more consistently
6105            leading to, for example, more aggressive unreachable code
6106            removal with LTO.
6107          + The inline heuristic can now bypass limits on the size of of
6108            inlined functions when the inlining is particularly
6109            profitable. This happens, for example, when loop bounds or
6110            array strides get propagated.
6111          + Values passed through aggregates (either by value or
6112            reference) are now propagated at the inter-procedural level
6113            leading to better inlining decisions (for example in the case
6114            of Fortran array descriptors) and devirtualization.
6115     * [6]AddressSanitizer , a fast memory error detector, has been added
6116       and can be enabled via -fsanitize=address. Memory access
6117       instructions will be instrumented to detect heap-, stack-, and
6118       global-buffer overflow as well as use-after-free bugs. To get nicer
6119       stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is
6120       available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on
6121       x86-64 Darwin.
6122     * [7]ThreadSanitizer has been added and can be enabled via
6123       -fsanitize=thread. Instructions will be instrumented to detect data
6124       races. The ThreadSanitizer is available on x86-64 GNU/Linux.
6125     * A new local register allocator (LRA) has been implemented, which
6126       replaces the 26 year old reload pass and improves generated code
6127       quality. For now it is active on the IA-32 and x86-64 targets.
6128     * Support for transactional memory has been implemented on the
6129       following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and
6130       Alpha.
6131
6132New Languages and Language specific improvements
6133
6134  C family
6135
6136     * Each diagnostic emitted now includes the original source line and a
6137       caret '^' indicating the column. The option
6138       -fno-diagnostics-show-caret suppresses this information.
6139     * The option -ftrack-macro-expansion=2 is now enabled by default.
6140       This allows the compiler to display the macro expansion stack in
6141       diagnostics. Combined with the caret information, an example
6142       diagnostic showing these two features is:
6143
6144t.c:1:94: error: invalid operands to binary < (have ‘struct mystruct’ and ‘float
6145’)
6146 #define MYMAX(A,B)    __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _
6147_b = (B); __a < __b ? __b : __a; })
6148
6149              ^
6150t.c:7:7: note: in expansion of macro 'MYMAX'
6151   X = MYMAX(P, F);
6152       ^
6153
6154     * A new -Wsizeof-pointer-memaccess warning has been added (also
6155       enabled by -Wall) to warn about suspicious length parameters to
6156       certain string and memory built-in functions if the argument uses
6157       sizeof. This warning warns e.g. about memset (ptr, 0, sizeof
6158       (ptr)); if ptr is not an array, but a pointer, and suggests a
6159       possible fix, or about memcpy (&foo, ptr, sizeof (&foo));.
6160     * The new option -Wpedantic is an alias for -pedantic, which is now
6161       deprecated. The forms -Wno-pedantic, -Werror=pedantic, and
6162       -Wno-error=pedantic work in the same way as for any other -W
6163       option. One caveat is that -Werror=pedantic is not equivalent to
6164       -pedantic-errors, since the latter makes into errors some warnings
6165       that are not controlled by -Wpedantic, and the former only affects
6166       diagnostics that are disabled when using -Wno-pedantic.
6167     * The option -Wshadow no longer warns if a declaration shadows a
6168       function declaration, unless the former declares a function or
6169       pointer to function, because this is [8]a common and valid case in
6170       real-world code.
6171
6172  C++
6173
6174     * G++ now implements the [9]C++11 thread_local keyword; this differs
6175       from the GNU __thread keyword primarily in that it allows dynamic
6176       initialization and destruction semantics. Unfortunately, this
6177       support requires a run-time penalty for references to
6178       non-function-local thread_local variables defined in a different
6179       translation unit even if they don't need dynamic initialization, so
6180       users may want to continue to use __thread for TLS variables with
6181       static initialization semantics.
6182       If the programmer can be sure that no use of the variable in a
6183       non-defining TU needs to trigger dynamic initialization (either
6184       because the variable is statically initialized, or a use of the
6185       variable in the defining TU will be executed before any uses in
6186       another TU), they can avoid this overhead with the
6187       -fno-extern-tls-init option.
6188       OpenMP threadprivate variables now also support dynamic
6189       initialization and destruction by the same mechanism.
6190     * G++ now implements the [10]C++11 attribute syntax, e.g.
6191
6192[[noreturn]] void f();
6193
6194       and also the alignment specifier, e.g.
6195
6196alignas(double) int i;
6197
6198     * G++ now implements [11]C++11 inheriting constructors, e.g.
6199
6200struct A { A(int); };
6201struct B: A { using A::A; }; // defines B::B(int)
6202B b(42); // OK
6203
6204     * As of GCC 4.8.1, G++ implements the change to decltype semantics
6205       from [12]N3276.
6206
6207struct A f();
6208decltype(f()) g();    // OK, return type of f() is not required to be complete.
6209
6210     * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g.
6211
6212struct A { int f() &; };
6213int i = A().f();  // error, f() requires an lvalue object
6214
6215     * G++ now supports a -std=c++1y option for experimentation with
6216       features proposed for the next revision of the standard, expected
6217       around 2014. Currently the only difference from -std=c++11 is
6218       support for return type deduction in normal functions, as proposed
6219       in [14]N3386. Status of C++1y features in GCC 4.8 can be found
6220       [15]here.
6221     * The G++ namespace association extension, __attribute ((strong)),
6222       has been deprecated. Inline namespaces should be used instead.
6223     * G++ now supports a -fext-numeric-literal option to control whether
6224       GNU numeric literal suffixes are accepted as extensions or
6225       processed as C++11 user-defined numeric literal suffixes. The flag
6226       is on (use suffixes for GNU literals) by default for -std=gnu++*,
6227       and -std=c++98. The flag is off (use suffixes for user-defined
6228       literals) by default for -std=c++11 and later.
6229
6230    Runtime Library (libstdc++)
6231
6232     * [16]Improved experimental support for the new ISO C++ standard,
6233       C++11, including:
6234          + forward_list meets the allocator-aware container requirements;
6235          + this_thread::sleep_for(), this_thread::sleep_until() and
6236            this_thread::yield() are defined without requiring the
6237            configure option --enable-libstdcxx-time;
6238     * Improvements to <random>:
6239          + SSE optimized normal_distribution.
6240          + Use of hardware RNG instruction for random_device on new x86
6241            processors (requires the assembler to support the
6242            instruction.)
6243       and <ext/random>:
6244          + New random number engine simd_fast_mersenne_twister_engine
6245            with an optimized SSE implementation.
6246          + New random number distributions beta_distribution,
6247            normal_mv_distribution, rice_distribution,
6248            nakagami_distribution, pareto_distribution, k_distribution,
6249            arcsine_distribution, hoyt_distribution.
6250     * Added --disable-libstdcxx-verbose configure option to disable
6251       diagnostic messages issued when a process terminates abnormally.
6252       This may be useful for embedded systems to reduce the size of
6253       executables that link statically to the library.
6254
6255  Fortran
6256
6257     * Compatibility notice:
6258          + Module files: The version of module files (.mod) has been
6259            incremented. Fortran MODULEs compiled by earlier GCC versions
6260            have to be recompiled, when they are USEd by files compiled
6261            with GCC 4.8. GCC 4.8 is not able to read .mod files created
6262            by earlier versions; attempting to do so gives an error
6263            message.
6264            Note: The ABI of the produced assembler data itself has not
6265            changed; object files and libraries are fully compatible with
6266            older versions except as noted below.
6267          + ABI: Some internal names (used in the assembler/object file)
6268            have changed for symbols declared in the specification part of
6269            a module. If an affected module – or a file using it via use
6270            association – is recompiled, the module and all files which
6271            directly use such symbols have to be recompiled as well. This
6272            change only affects the following kind of module symbols:
6273               o Procedure pointers. Note: C-interoperable function
6274                 pointers (type(c_funptr)) are not affected nor are
6275                 procedure-pointer components.
6276               o Deferred-length character strings.
6277     * The [17]BACKTRACE intrinsic subroutine has been added. It shows a
6278       backtrace at an arbitrary place in user code; program execution
6279       continues normally afterwards.
6280     * The [18]-Wc-binding-type warning option has been added (disabled by
6281       default). It warns if the a variable might not be C interoperable;
6282       in particular, if the variable has been declared using an intrinsic
6283       type with default kind instead of using a kind parameter defined
6284       for C interoperability in the intrinsic ISO_C_Binding module.
6285       Before, this warning was always printed. The -Wc-binding-type
6286       option is enabled by -Wall.
6287     * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line
6288       options have been added, which diagnose when code is inserted for
6289       automatic (re)allocation of a variable during assignment. This
6290       option can be used to decide whether it is safe to use
6291       [20]-fno-realloc-lhs. Additionally, it can be used to find
6292       automatic (re)allocation in hot loops. (For arrays, replacing
6293       "var=" by "var(:)=" disables the automatic reallocation.)
6294     * The [21]-Wcompare-reals command-line option has been added. When
6295       this is set, warnings are issued when comparing REAL or COMPLEX
6296       types for equality and inequality; consider replacing a == b by
6297       abs(a−b) < eps with a suitable eps. -Wcompare-reals is enabled by
6298       -Wextra.
6299     * The [22]-Wtarget-lifetime command-line option has been added
6300       (enabled with -Wall), which warns if the pointer in a pointer
6301       assignment might outlive its target.
6302     * Reading floating point numbers which use "q" for the exponential
6303       (such as 4.0q0) is now supported as vendor extension for better
6304       compatibility with old data files. It is strongly recommended to
6305       use for I/O the equivalent but standard conforming "e" (such as
6306       4.0e0).
6307       (For Fortran source code, consider replacing the "q" in
6308       floating-point literals by a kind parameter (e.g. 4.0e0_qp with a
6309       suitable qp). Note that – in Fortran source code – replacing "q" by
6310       a simple "e" is not equivalent.)
6311     * The GFORTRAN_TMPDIR environment variable for specifying a
6312       non-default directory for files opened with STATUS="SCRATCH", is
6313       not used anymore. Instead gfortran checks the POSIX/GNU standard
6314       TMPDIR environment variable. If TMPDIR is not defined, gfortran
6315       falls back to other methods to determine the directory for
6316       temporary files as documented in the [23]user manual.
6317     * [24]Fortran 2003:
6318          + Support for unlimited polymorphic variables (CLASS(*)) has
6319            been added. Nonconstant character lengths are not yet
6320            supported.
6321     * [25]TS 29113:
6322          + Assumed types (TYPE(*)) are now supported.
6323          + Experimental support for assumed-rank arrays (dimension(..))
6324            has been added. Note that currently gfortran's own array
6325            descriptor is used, which is different from the one defined in
6326            TS29113, see [26]gfortran's header file or use the [27]Chasm
6327            Language Interoperability Tools.
6328
6329  Go
6330
6331     * GCC 4.8.2 provides a complete implementation of the Go 1.1.2
6332       release.
6333     * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1
6334       release. The library support is not quite complete.
6335     * Go has been tested on GNU/Linux and Solaris platforms for various
6336       processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may
6337       work on other platforms as well.
6338
6339New Targets and Target Specific Improvements
6340
6341  AArch64
6342
6343     * A new port has been added to support AArch64, the new 64-bit
6344       architecture from ARM. Note that this is a separate port from the
6345       existing 32-bit ARM port.
6346     * The port provides initial support for the Cortex-A53 and the
6347       Cortex-A57 processors with the command line options
6348       -mcpu=cortex-a53 and -mcpu=cortex-a57.
6349     * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769
6350       has been added and can be enabled by giving the
6351       -mfix-cortex-a53-835769 option. Alternatively it can be enabled by
6352       default by configuring GCC with the --enable-fix-cortex-a53-835769
6353       option.
6354
6355  ARM
6356
6357     * Initial support has been added for the AArch32 extensions defined
6358       in the ARMv8 architecture.
6359     * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs.
6360     * A new option, -mcpu=marvell-pj4, has been added to generate code
6361       for the Marvell PJ4 processor.
6362     * The compiler can now automatically generate the VFMA, VFMS, REVSH
6363       and REV16 instructions.
6364     * A new vectorizer cost model for Advanced SIMD configurations to
6365       improve the auto-vectorization strategies used.
6366     * The scheduler now takes into account the number of live registers
6367       to reduce the amount of spilling that can occur. This should
6368       improve code performance in large functions. The limit can be
6369       removed by using the option -fno-sched-pressure.
6370     * Improvements have been made to the Marvell iWMMX code generation
6371       and support for the iWMMX2 SIMD unit has been added. The option
6372       -mcpu=iwmmxt2 can be used to enable code generation for the latter.
6373     * A number of code generation improvements for Thumb2 to reduce code
6374       size when compiling for the M-profile processors.
6375     * The RTEMS (arm-rtems) port has been updated to use the EABI.
6376     * Code generation support for the old FPA and Maverick floating-point
6377       architectures has been removed. Ports that previously relied on
6378       these features have also been removed. This includes the targets:
6379          + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
6380          + arm*-*-elf (use arm*-*-eabi)
6381          + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
6382          + arm*-*-ecos-elf (no alternative)
6383          + arm*-*-freebsd (no alternative)
6384          + arm*-wince-pe* (no alternative).
6385
6386  AVR
6387
6388     * Support for the "Embedded C" fixed-point has been added. For
6389       details, see the [28]GCC wiki and the [29]user manual. The support
6390       is not complete.
6391     * A new print modifier %r for register operands in inline assembler
6392       is supported. It will print the raw register number without the
6393       register prefix 'r':
6394    /* Return the most significant byte of 'val', a 64-bit value.  */
6395
6396    unsigned char msb (long long val)
6397    {
6398      unsigned char c;
6399      __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val));
6400      return c;
6401    }
6402       The inline assembler in this example will generate code like
6403    mov r24, 8+7
6404       provided c is allocated to R24 and val is allocated to R8…R15. This
6405       works because the GNU assembler accepts plain register numbers
6406       without register prefix.
6407     * Static initializers with 3-byte symbols are supported now:
6408    extern const __memx char foo;
6409    const __memx void *pfoo = &foo;
6410       This requires at least Binutils 2.23.
6411
6412  IA-32/x86-64
6413
6414     * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with
6415       SSE extensions disabled. Since the x86-64 ABI requires 16 byte
6416       stack alignment, this is ABI incompatible and intended to be used
6417       in controlled environments where stack space is an important
6418       limitation. This option will lead to wrong code when functions
6419       compiled with 16 byte stack alignment (such as functions from a
6420       standard library) are called with misaligned stack. In this case,
6421       SSE instructions may lead to misaligned memory access traps. In
6422       addition, variable arguments will be handled incorrectly for 16
6423       byte aligned objects (including x87 long double and __int128),
6424       leading to wrong results. You must build all modules with
6425       -mpreferred-stack-boundary=3, including any libraries. This
6426       includes the system libraries and startup modules.
6427     * Support for the new Intel processor codename Broadwell with RDSEED,
6428       ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw,
6429       -mrdseed command-line options.
6430     * Support for the Intel RTM and HLE intrinsics, built-in functions
6431       and code generation is available via -mrtm and -mhle.
6432     * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets.
6433       Intrinsics and built-in functions are available via -mfxsr, -mxsave
6434       and -mxsaveopt respectively.
6435     * New -maddress-mode=[short|long] options for x32.
6436       -maddress-mode=short overrides default 64-bit addresses to 32-bit
6437       by emitting the 0x67 address-size override prefix. This is the
6438       default address mode for x32.
6439     * New built-in functions to detect run-time CPU type and ISA:
6440          + A built-in function __builtin_cpu_is has been added to detect
6441            if the run-time CPU is of a particular type. It returns a
6442            positive integer on a match and zero otherwise. It accepts one
6443            string literal argument, the CPU name. For example,
6444            __builtin_cpu_is("westmere") returns a positive integer if the
6445            run-time CPU is an Intel Core i7 Westmere processor. Please
6446            refer to the [30]user manual for the list of valid CPU names
6447            recognized.
6448          + A built-in function __builtin_cpu_supports has been added to
6449            detect if the run-time CPU supports a particular ISA feature.
6450            It returns a positive integer on a match and zero otherwise.
6451            It accepts one string literal argument, the ISA feature. For
6452            example, __builtin_cpu_supports("ssse3") returns a positive
6453            integer if the run-time CPU supports SSSE3 instructions.
6454            Please refer to the [31]user manual for the list of valid ISA
6455            names recognized.
6456       Caveat: If these built-in functions are called before any static
6457       constructors are invoked, like during IFUNC initialization, then
6458       the CPU detection initialization must be explicitly run using this
6459       newly provided built-in function, __builtin_cpu_init. The
6460       initialization needs to be done only once. For example, this is how
6461       the invocation would look like inside an IFUNC initializer:
6462    static void (*some_ifunc_resolver(void))(void)
6463    {
6464      __builtin_cpu_init();
6465      if (__builtin_cpu_is("amdfam10h") ...
6466      if (__builtin_cpu_supports("popcnt") ...
6467    }
6468
6469     * Function Multiversioning Support with G++:
6470       It is now possible to create multiple function versions each
6471       targeting a specific processor and/or ISA. Function versions have
6472       the same signature but different target attributes. For example,
6473       here is a program with function versions:
6474    __attribute__ ((target ("default")))
6475    int foo(void)
6476    {
6477      return 1;
6478    }
6479
6480    __attribute__ ((target ("sse4.2")))
6481    int foo(void)
6482    {
6483      return 2;
6484    }
6485
6486    int main (void)
6487    {
6488      int (*p) = &foo;
6489      assert ((*p)() == foo());
6490      return 0;
6491    }
6492
6493       Please refer to this [32]wiki for more information.
6494     * The x86 back end has been improved to allow option -fschedule-insns
6495       to work reliably. This option can be used to schedule instructions
6496       better and leads to improved performace in certain cases.
6497     * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437
6498       from the Mingw-w64 trunk.
6499     * Support for new AMD family 15h processors (Steamroller core) is now
6500       available through the -march=bdver3 and -mtune=bdver3 options.
6501     * Support for new AMD family 16h processors (Jaguar core) is now
6502       available through the -march=btver2 and -mtune=btver2 options.
6503
6504  FRV
6505
6506     * This target now supports the -fstack-usage command-line option.
6507
6508  MIPS
6509
6510     * GCC can now generate code specifically for the R4700, Broadcom XLP
6511       and MIPS 34kn processors. The associated -march options are
6512       -march=r4700, -march=xlp and -march=34kn respectively.
6513     * GCC now generates better DSP code for MIPS 74k cores thanks to
6514       further scheduling optimizations.
6515     * The MIPS port now supports the -fstack-check option.
6516     * GCC now passes the -mmcu and -mno-mcu options to the assembler.
6517     * Previous versions of GCC would silently accept -fpic and -fPIC for
6518       -mno-abicalls targets like mips*-elf. This combination was not
6519       intended or supported, and did not generate position-independent
6520       code. GCC 4.8 now reports an error when this combination is used.
6521
6522  PowerPC / PowerPC64 / RS6000
6523
6524     * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save,
6525       restore or update the VRSAVE register by default. The respective
6526       operating systems manage the VRSAVE register directly.
6527     * Large TOC support has been added for AIX through the command line
6528       option -mcmodel=large.
6529     * Native Thread-Local Storage support has been added for AIX.
6530     * VMX (Altivec) and VSX instruction sets now are enabled implicitly
6531       when targetting processors that support those hardware features on
6532       AIX 6.1 and above.
6533
6534  RX
6535
6536     * This target will now issue a warning message whenever multiple fast
6537       interrupt handlers are found in the same compilation unit. This
6538       feature can be turned off by the new
6539       -mno-warn-multiple-fast-interrupts command-line option.
6540
6541  S/390, System z
6542
6543     * Support for the IBM zEnterprise zEC12 processor has been added.
6544       When using the -march=zEC12 option, the compiler will generate code
6545       making use of the following new instructions:
6546          + load and trap instructions
6547          + 2 new compare and trap instructions
6548          + rotate and insert selected bits - without CC clobber
6549       The -mtune=zEC12 option enables zEC12 specific instruction
6550       scheduling without making use of new instructions.
6551     * Register pressure sensitive instruction scheduling is enabled by
6552       default.
6553     * The ifunc function attribute is enabled by default.
6554     * memcpy and memcmp invokations on big memory chunks or with run time
6555       lengths are not generated inline anymore when tuning for z10 or
6556       higher. The purpose is to make use of the IFUNC optimized versions
6557       in Glibc.
6558
6559  SH
6560
6561     * The default alignment settings have been reduced to be less
6562       aggressive. This results in more compact code for optimization
6563       levels other than -Os.
6564     * Improved support for the __atomic built-in functions:
6565          + A new option -matomic-model=model selects the model for the
6566            generated atomic sequences. The following models are
6567            supported:
6568
6569              soft-gusa
6570                      Software gUSA sequences (SH3* and SH4* only). On
6571                      SH4A targets this will now also partially utilize
6572                      the movco.l and movli.l instructions. This is the
6573                      default when the target is sh3*-*-linux* or
6574                      sh4*-*-linux*.
6575
6576              hard-llcs
6577                      Hardware movco.l / movli.l sequences (SH4A only).
6578
6579              soft-tcb
6580                      Software thread control block sequences.
6581
6582              soft-imask
6583                      Software interrupt flipping sequences (privileged
6584                      mode only). This is the default when the target is
6585                      sh1*-*-linux* or sh2*-*-linux*.
6586
6587              none
6588                      Generates function calls to the respective __atomic
6589                      built-in functions. This is the default for SH64
6590                      targets or when the target is not sh*-*-linux*.
6591
6592          + The option -msoft-atomic has been deprecated. It is now an
6593            alias for -matomic-model=soft-gusa.
6594          + A new option -mtas makes the compiler generate the tas.b
6595            instruction for the __atomic_test_and_set built-in function
6596            regardless of the selected atomic model.
6597          + The __sync functions in libgcc now reflect the selected atomic
6598            model when building the toolchain.
6599     * Added support for the mov.b and mov.w instructions with
6600       displacement addressing.
6601     * Added support for the SH2A instructions movu.b and movu.w.
6602     * Various improvements to code generated for integer arithmetic.
6603     * Improvements to conditional branches and code that involves the T
6604       bit. A new option -mzdcbranch tells the compiler to favor
6605       zero-displacement branches. This is enabled by default for SH4*
6606       targets.
6607     * The pref instruction will now be emitted by the __builtin_prefetch
6608       built-in function for SH3* targets.
6609     * The fmac instruction will now be emitted by the fmaf standard
6610       function and the __builtin_fmaf built-in function.
6611     * The -mfused-madd option has been deprecated in favor of the
6612       machine-independent -ffp-contract option. Notice that the fmac
6613       instruction will now be generated by default for expressions like a
6614       * b + c. This is due to the compiler default setting
6615       -ffp-contract=fast.
6616     * Added new options -mfsrra and -mfsca to allow the compiler using
6617       the fsrra and fsca instructions on targets other than SH4A (where
6618       they are already enabled by default).
6619     * Added support for the __builtin_bswap32 built-in function. It is
6620       now expanded as a sequence of swap.b and swap.w instructions
6621       instead of a library function call.
6622     * The behavior of the -mieee option has been fixed and the negative
6623       form -mno-ieee has been added to control the IEEE conformance of
6624       floating point comparisons. By default -mieee is now enabled and
6625       the option -ffinite-math-only implicitly sets -mno-ieee.
6626     * Added support for the built-in functions __builtin_thread_pointer
6627       and __builtin_set_thread_pointer. This assumes that GBR is used to
6628       hold the thread pointer of the current thread. Memory loads and
6629       stores relative to the address returned by __builtin_thread_pointer
6630       will now also utilize GBR based displacement address modes.
6631     * The -mdiv= option for targets other than SHmedia has been fixed and
6632       documented.
6633
6634  SPARC
6635
6636     * Added optimized instruction scheduling for Niagara4.
6637
6638  TILE-Gx
6639
6640     * Added support for the -mcmodel=MODEL command-line option. The
6641       models supported are small and large.
6642
6643  V850
6644
6645     * This target now supports the E3V5 architecture via the use of the
6646       new -mv850e3v5 command-line option. It also has experimental
6647       support for the e3v5 LOOP instruction which can be enabled via the
6648       new -mloop command-line option.
6649
6650  XStormy16
6651
6652     * This target now supports the -fstack-usage command-line option.
6653
6654Operating Systems
6655
6656  OpenBSD
6657
6658     * Support for OpenBSD/amd64 (x86_64-*-openbsd*) has been added and
6659       support for OpenBSD/i386 (i386-*-openbsd*) has been rejuvenated.
6660
6661  Windows (Cygwin)
6662
6663     * Executables are now linked against shared libgcc by default. The
6664       previous default was to link statically, which can still be done by
6665       explicitly specifying -static or static-libgcc on the command line.
6666       However it is strongly advised against, as it will cause problems
6667       for any application that makes use of DLLs compiled by GCC. It
6668       should be alright for a monolithic stand-alone application that
6669       only links against the Windows DLLs, but offers little or no
6670       benefit.
6671
6672GCC 4.8.1
6673
6674   This is the [33]list of problem reports (PRs) from GCC's bug tracking
6675   system that are known to be fixed in the 4.8.1 release. This list might
6676   not be complete (that is, it is possible that some PRs that have been
6677   fixed are not listed here).
6678
6679   The C++11 <chrono> std::chrono::system_clock and
6680   std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they
6681   both are now separate (never typedefs of each other), both use
6682   std::chrono::nanoseconds resolution, on most GNU/Linux configurations
6683   std::chrono::steady_clock is now finally monotonic, and both classes
6684   are mangled differently than in the previous GCC releases.
6685   std::chrono::system_clock::now() with std::chrono::microseconds resp.
6686   std::chrono::seconds resolution is still exported for backwards
6687   compatibility with default configured libstdc++. Note that libstdc++
6688   configured with --enable-libstdcxx-time= used to be ABI incompatible
6689   with default configured libstdc++ for those two classes and no ABI
6690   compatibility can be offered for those configurations, so any C++11
6691   code that uses those classes and has been compiled and linked against
6692   libstdc++ configured with the non-default --enable-libstdcxx-time=
6693   configuration option needs to be recompiled.
6694
6695GCC 4.8.2
6696
6697   This is the [34]list of problem reports (PRs) from GCC's bug tracking
6698   system that are known to be fixed in the 4.8.2 release. This list might
6699   not be complete (that is, it is possible that some PRs that have been
6700   fixed are not listed here).
6701
6702GCC 4.8.3
6703
6704   This is the [35]list of problem reports (PRs) from GCC's bug tracking
6705   system that are known to be fixed in the 4.8.3 release. This list might
6706   not be complete (that is, it is possible that some PRs that have been
6707   fixed are not listed here).
6708
6709   Support for the new powerpc64le-linux platform has been added. It
6710   defaults to generating code that conforms to the ELFV2 ABI.
6711
6712GCC 4.8.4
6713
6714   This is the [36]list of problem reports (PRs) from GCC's bug tracking
6715   system that are known to be fixed in the 4.8.4 release. This list might
6716   not be complete (that is, it is possible that some PRs that have been
6717   fixed are not listed here).
6718
6719GCC 4.8.5
6720
6721   This is the [37]list of problem reports (PRs) from GCC's bug tracking
6722   system that are known to be fixed in the 4.8.5 release. This list might
6723   not be complete (that is, it is possible that some PRs that have been
6724   fixed are not listed here).
6725
6726
6727    For questions related to the use of GCC, please consult these web
6728    pages and the [38]GCC manuals. If that fails, the
6729    [39]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6730    web pages and the development of GCC are welcome on our developer
6731    list at [40]gcc@gcc.gnu.org. All of [41]our lists have public
6732    archives.
6733
6734   Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and
6735   distribution of this entire article is permitted in any medium,
6736   provided this notice is preserved.
6737
6738   These pages are [43]maintained by the GCC team. Last modified
6739   2019-11-28[44].
6740
6741References
6742
6743   1. https://gcc.gnu.org/wiki/cxx-conversion
6744   2. ftp://gcc.gnu.org/pub/gcc/infrastructure/
6745   3. http://www.nongnu.org/avr-libc/
6746   4. https://gcc.gnu.org/PR54461
6747   5. https://gcc.gnu.org/gcc-4.8/porting_to.html
6748   6. https://github.com/google/sanitizers
6749   7. https://code.google.com/archive/p/data-race-test/wikis/ThreadSanitizer.wiki
6750   8. https://lkml.org/lkml/2006/11/28/239
6751   9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
6752  10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
6753  11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
6754  12. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf
6755  13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
6756  14. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html
6757  15. http://gcc.gnu.org/projects/cxx1y.html
6758  16. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011
6759  17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html
6760  18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
6761  19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
6762  20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
6763  21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
6764  22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
6765  23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html
6766  24. https://gcc.gnu.org/wiki/Fortran2003Status
6767  25. https://gcc.gnu.org/wiki/TS29113Status
6768  26. https://gcc.gnu.org/viewcvs/trunk/libgfortran/libgfortran.h?content-type=text/plain&view=co
6769  27. http://chasm-interop.sourceforge.net/
6770  28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support
6771  29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html
6772  30. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html
6773  31. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html
6774  32. https://gcc.gnu.org/wiki/FunctionMultiVersioning
6775  33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1
6776  34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2
6777  35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3
6778  36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4
6779  37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5
6780  38. https://gcc.gnu.org/onlinedocs/
6781  39. mailto:gcc-help@gcc.gnu.org
6782  40. mailto:gcc@gcc.gnu.org
6783  41. https://gcc.gnu.org/lists.html
6784  42. https://www.fsf.org/
6785  43. https://gcc.gnu.org/about.html
6786  44. http://validator.w3.org/check/referer
6787======================================================================
6788http://gcc.gnu.org/gcc-4.7/index.html
6789                             GCC 4.7 Release Series
6790
6791   (This release series is no longer supported.)
6792
6793   June 12, 2014
6794
6795   The [1]GNU project and the GCC developers are pleased to announce the
6796   release of GCC 4.7.4.
6797
6798   This release is a bug-fix release, containing fixes for regressions in
6799   GCC 4.7.3 relative to previous releases of GCC.
6800
6801Release History
6802
6803   GCC 4.7.4
6804          June 12, 2014 ([2]changes, [3]documentation)
6805
6806   GCC 4.7.3
6807          April 11, 2013 ([4]changes, [5]documentation)
6808
6809   GCC 4.7.2
6810          September 20, 2012 ([6]changes, [7]documentation)
6811
6812   GCC 4.7.1
6813          June 14, 2012 ([8]changes, [9]documentation)
6814
6815   GCC 4.7.0
6816          March 22, 2012 ([10]changes, [11]documentation)
6817
6818References and Acknowledgements
6819
6820   GCC used to stand for the GNU C Compiler, but since the compiler
6821   supports several other languages aside from C, it now stands for the
6822   GNU Compiler Collection.
6823
6824   A list of [12]successful builds is updated as new information becomes
6825   available.
6826
6827   The GCC developers would like to thank the numerous people that have
6828   contributed new features, improvements, bug fixes, and other changes as
6829   well as test results to GCC. This [13]amazing group of volunteers is
6830   what makes GCC successful.
6831
6832   For additional information about GCC please refer to the [14]GCC
6833   project web site or contact the [15]GCC development mailing list.
6834
6835   To obtain GCC please use [16]our mirror sites or [17]our version
6836   control system.
6837
6838
6839    For questions related to the use of GCC, please consult these web
6840    pages and the [18]GCC manuals. If that fails, the
6841    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6842    web pages and the development of GCC are welcome on our developer
6843    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
6844    archives.
6845
6846   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
6847   distribution of this entire article is permitted in any medium,
6848   provided this notice is preserved.
6849
6850   These pages are [23]maintained by the GCC team. Last modified
6851   2020-01-14[24].
6852
6853References
6854
6855   1. http://www.gnu.org/
6856   2. http://gcc.gnu.org/gcc-4.7/changes.html
6857   3. https://gcc.gnu.org/onlinedocs/4.7.4/
6858   4. http://gcc.gnu.org/gcc-4.7/changes.html
6859   5. https://gcc.gnu.org/onlinedocs/4.7.3/
6860   6. http://gcc.gnu.org/gcc-4.7/changes.html
6861   7. https://gcc.gnu.org/onlinedocs/4.7.2/
6862   8. http://gcc.gnu.org/gcc-4.7/changes.html
6863   9. https://gcc.gnu.org/onlinedocs/4.7.1/
6864  10. http://gcc.gnu.org/gcc-4.7/changes.html
6865  11. https://gcc.gnu.org/onlinedocs/4.7.0/
6866  12. http://gcc.gnu.org/gcc-4.7/buildstat.html
6867  13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6868  14. http://gcc.gnu.org/index.html
6869  15. mailto:gcc@gcc.gnu.org
6870  16. http://gcc.gnu.org/mirrors.html
6871  17. http://gcc.gnu.org/git.html
6872  18. https://gcc.gnu.org/onlinedocs/
6873  19. mailto:gcc-help@gcc.gnu.org
6874  20. mailto:gcc@gcc.gnu.org
6875  21. https://gcc.gnu.org/lists.html
6876  22. https://www.fsf.org/
6877  23. https://gcc.gnu.org/about.html
6878  24. http://validator.w3.org/check/referer
6879======================================================================
6880http://gcc.gnu.org/gcc-4.7/changes.html
6881                             GCC 4.7 Release Series
6882                        Changes, New Features, and Fixes
6883
6884Caveats
6885
6886     * The -fconserve-space flag has been deprecated. The flag had no
6887       effect for most targets: only targets without a global .bss section
6888       and without support for switchable sections. Furthermore, the flag
6889       only had an effect for G++, where it could result in wrong
6890       semantics (please refer to the GCC manual for further details). The
6891       flag will be removed in GCC 4.8
6892     * Support for a number of older systems and recently unmaintained or
6893       untested target ports of GCC has been declared obsolete in GCC 4.7.
6894       Unless there is activity to revive them, the next release of GCC
6895       will have their sources permanently removed.
6896       All GCC ports for the following processor architectures have been
6897       declared obsolete:
6898          + picoChip (picochip-*)
6899       The following ports for individual systems on particular
6900       architectures have been obsoleted:
6901          + IRIX 6.5 (mips-sgi-irix6.5)
6902          + MIPS OpenBSD (mips*-*-openbsd*)
6903          + Solaris 8 (*-*-solaris2.8). Details can be found in the
6904            [1]announcement.
6905          + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*)
6906     * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A,
6907       ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by
6908       default, which for some sources generates code that accesses memory
6909       on unaligned addresses. This requires the kernel of those systems
6910       to enable such accesses (controlled by CP15 register c1, refer to
6911       ARM documentation). Alternatively, or for compatibility with
6912       kernels where unaligned accesses are not supported, all code has to
6913       be compiled with -mno-unaligned-access. Upstream Linux kernel
6914       releases have automatically and unconditionally supported unaligned
6915       accesses as emitted by GCC due to this option being active since
6916       version 2.6.28.
6917     * Support on ARM for the legacy floating-point accelerator (FPA) and
6918       the mixed-endian floating-point format that it used has been
6919       obsoleted. The ports that still use this format have been obsoleted
6920       as well. Many legacy ARM ports already provide an alternative that
6921       uses the VFP floating-point format. The obsolete ports will be
6922       deleted in the next release.
6923       The obsolete ports with alternatives are:
6924          + arm*-*-rtems (use arm*-*-rtemseabi)
6925          + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
6926          + arm*-*-elf (use arm*-*-eabi)
6927          + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
6928       Note, however, that these alternatives are not binary compatible
6929       with their legacy counterparts (although some can support running
6930       legacy applications).
6931       The obsolete ports that currently lack a modern alternative are:
6932          + arm*-*-ecos-elf
6933          + arm*-*-freebsd
6934          + arm*-wince-pe*
6935       New ports that support more recent versions of the architecture are
6936       welcome.
6937     * Support for the Maverick co-processor on ARM has been obsoleted.
6938       Code to support it will be deleted in the next release.
6939     * Support has been removed for Unix International threads on Solaris
6940       2, so the --enable-threads=solaris configure option and the
6941       -threads compiler option don't work any longer.
6942     * Support has been removed for the Solaris BSD Compatibility Package,
6943       which lives in /usr/ucbinclude and /usr/ucblib. It has been removed
6944       from Solaris 11, and was only intended as a migration aid from
6945       SunOS 4 to SunOS 5. The -compat-bsd compiler option is not
6946       recognized any longer.
6947     * The AVR port's libgcc has been improved and its multilib structure
6948       has been enhanced. As a result, all objects contributing to an
6949       application must either be compiled with GCC versions up to 4.6.x
6950       or with GCC versions 4.7.1 or later. If the compiler is used with
6951       AVR Libc, you need a version that supports the new layout, i.e.
6952       implements [2]#35407.
6953     * The AVR port's -mshort-calls command-line option has been
6954       deprecated. It will be removed in the GCC 4.8 release. See -mrelax
6955       for a replacement.
6956     * The AVR port only references startup code that clears .bss and the
6957       common section resp. initializes the .data and .rodata section
6958       provided respective sections (or subsections thereof) are not
6959       empty, see [3]PR18145. Applications that put all static storage
6960       objects into non-standard sections and / or define all static
6961       storage objects in assembler modules, must reference __do_clear_bss
6962       resp. __do_copy_data by hand or undefine the symbol(s) by means of
6963       -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data.
6964     * The ARM port's -mwords-little-endian option has been deprecated. It
6965       will be removed in a future release.
6966     * Support has been removed for the NetWare x86 configuration
6967       obsoleted in GCC 4.6.
6968     * It is no longer possible to use the "l" constraint in MIPS16 asm
6969       statements.
6970     * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard
6971       library which affected the ABI in C++11 mode: a data member was
6972       added to std::list changing its size and altering the definitions
6973       of some member functions, and std::pair's move constructor was
6974       non-trivial which altered the calling convention for functions with
6975       std::pair arguments or return types. The ABI incompatibilities have
6976       been fixed for GCC version 4.7.2 but as a result C++11 code
6977       compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11
6978       code compiled with different GCC versions and with C++98/C++03 code
6979       compiled with any version.
6980     * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
6981       rules for the layout of vectors that could lead to wrong code being
6982       generated. Vectors larger than 8 bytes in size are now by default
6983       aligned to an 8-byte boundary. This is an ABI change: code that
6984       makes explicit use of vector types may be incompatible with binary
6985       objects built with older versions of GCC. Auto-vectorized code is
6986       not affected by this change. (This change affects GCC versions
6987       4.7.2 and later.)
6988     * More information on porting to GCC 4.7 from previous versions of
6989       GCC can be found in the [4]porting guide for this release.
6990
6991General Optimizer Improvements
6992
6993     * Support for a new parameter --param case-values-threshold=n was
6994       added to allow users to control the cutoff between doing switch
6995       statements as a series of if statements and using a jump table.
6996     * Link-time optimization (LTO) improvements:
6997          + Improved scalability and reduced memory usage. Link time
6998            optimization of Firefox now requires 3GB of RAM on a 64-bit
6999            system, while over 8GB was needed previously. Linking time has
7000            been improved, too. The serial stage of linking Firefox has
7001            been sped up by about a factor of 10.
7002          + Reduced size of object files and temporary storage used during
7003            linking.
7004          + Streaming performance (both outbound and inbound) has been
7005            improved.
7006          + ld -r is now supported with LTO.
7007          + Several bug fixes, especially in symbol table handling and
7008            merging.
7009     * Interprocedural optimization improvements:
7010          + Heuristics now take into account that after inlining code will
7011            be optimized out because of known values (or properties) of
7012            function parameters. For example:
7013void foo(int a)
7014{
7015  if (a > 10)
7016    ... huge code ...
7017}
7018void bar (void)
7019{
7020  foo (0);
7021}
7022
7023            The call of foo will be inlined into bar even when optimizing
7024            for code size. Constructs based on __builtin_constant_p are
7025            now understood by the inliner and code size estimates are
7026            evaluated a lot more realistically.
7027          + The representation of C++ virtual thunks and aliases (both
7028            implicit and defined via the alias attribute) has been
7029            re-engineered. Aliases no longer pose optimization barriers
7030            and calls to an alias can be inlined and otherwise optimized.
7031          + The inter-procedural constant propagation pass has been
7032            rewritten. It now performs generic function specialization.
7033            For example when compiling the following:
7034void foo(bool flag)
7035{
7036  if (flag)
7037    ... do something ...
7038  else
7039    ... do something else ...
7040}
7041void bar (void)
7042{
7043  foo (false);
7044  foo (true);
7045  foo (false);
7046  foo (true);
7047  foo (false);
7048  foo (true);
7049}
7050
7051            GCC will now produce two copies of foo. One with flag being
7052            true, while other with flag being false. This leads to
7053            performance improvements previously possible only by inlining
7054            all calls. Cloning causes a lot less code size growth.
7055     * A string length optimization pass has been added. It attempts to
7056       track string lengths and optimize various standard C string
7057       functions like strlen, strchr, strcpy, strcat, stpcpy and their
7058       _FORTIFY_SOURCE counterparts into faster alternatives. This pass is
7059       enabled by default at -O2 or above, unless optimizing for size, and
7060       can be disabled by the -fno-optimize-strlen option. The pass can
7061       e.g. optimize
7062char *bar (const char *a)
7063{
7064  size_t l = strlen (a) + 2;
7065  char *p = malloc (l); if (p == NULL) return p;
7066  strcpy (p, a); strcat (p, "/"); return p;
7067}
7068
7069       into:
7070char *bar (const char *a)
7071{
7072  size_t tmp = strlen (a);
7073  char *p = malloc (tmp + 2); if (p == NULL) return p;
7074  memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p;
7075}
7076
7077       or for hosted compilations where stpcpy is available in the runtime
7078       and headers provide its prototype, e.g.
7079void foo (char *a, const char *b, const char *c, const char *d)
7080{
7081  strcpy (a, b); strcat (a, c); strcat (a, d);
7082}
7083
7084       can be optimized into:
7085void foo (char *a, const char *b, const char *c, const char *d)
7086{
7087  strcpy (stpcpy (stpcpy (a, b), c), d);
7088}
7089
7090New Languages and Language specific improvements
7091
7092     * Version 3.1 of the OpenMP specification is now supported for the C,
7093       C++, and Fortran compilers.
7094
7095  Ada
7096
7097     * The command-line option -feliminate-unused-debug-types has been
7098       re-enabled by default, as it is for the other languages, leading to
7099       a reduction in debug info size of 12.5% and more for relevant
7100       cases, as well as to a small compilation speedup.
7101
7102  C family
7103
7104     * A new built-in, __builtin_assume_aligned, has been added, through
7105       which the compiler can be hinted about pointer alignment and can
7106       use it to improve generated code.
7107     * A new warning option -Wunused-local-typedefs was added for C, C++,
7108       Objective-C and Objective-C++. This warning diagnoses typedefs
7109       locally defined in a function, and otherwise not used.
7110     * A new experimental command-line option -ftrack-macro-expansion was
7111       added for C, C++, Objective-C, Objective-C++ and Fortran. It allows
7112       the compiler to emit diagnostic about the current macro expansion
7113       stack when a compilation error occurs in a macro expansion.
7114     * Experimental support for transactional memory has been added. It
7115       includes support in the compiler, as well as a supporting runtime
7116       library called libitm. To compile code with transactional memory
7117       constructs, use the -fgnu-tm option.
7118       Support is currently available for Alpha, ARM, PowerPC, SH, SPARC,
7119       and 32-bit/64-bit x86 platforms.
7120       For more details on transactional memory see [5]the GCC WiKi.
7121     * Support for atomic operations specifying the C++11/C11 memory model
7122       has been added. These new __atomic routines replace the existing
7123       __sync built-in routines.
7124       Atomic support is also available for memory blocks. Lock-free
7125       instructions will be used if a memory block is the same size and
7126       alignment as a supported integer type. Atomic operations which do
7127       not have lock-free support are left as function calls. A set of
7128       library functions is available on the GCC atomic wiki in the
7129       "External Atomics Library" section.
7130       For more details on the memory models and features, see the
7131       [6]atomic wiki.
7132     * When a binary operation is performed on vector types and one of the
7133       operands is a uniform vector, it is possible to replace the vector
7134       with the generating element. For example:
7135typedef int v4si __attribute__ ((vector_size (16)));
7136v4si res, a = {1,2,3,4};
7137int x;
7138
7139res = 2 + a;  /* means {2,2,2,2} + a  */
7140res = a - x;  /* means a - {x,x,x,x}  */
7141
7142  C
7143
7144     * There is support for some more features from the C11 revision of
7145       the ISO C standard. GCC now accepts the options -std=c11 and
7146       -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x.
7147          + Unicode strings (previously supported only with options such
7148            as -std=gnu11, now supported with -std=c11), and the
7149            predefined macros __STDC_UTF_16__ and __STDC_UTF_32__.
7150          + Nonreturning functions (_Noreturn and <stdnoreturn.h>).
7151          + Alignment support (_Alignas, _Alignof, max_align_t,
7152            <stdalign.h>).
7153          + A built-in function __builtin_complex is provided to support C
7154            library implementation of the CMPLX family of macros.
7155
7156  C++
7157
7158     * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat
7159       options, which are equivalent to -std=c++0x, -std=gnu++0x, and
7160       -Wc++0x-compat, respectively.
7161     * G++ now implements [7]C++11 extended friend syntax:
7162
7163template<class W>
7164class Q
7165{
7166  static const int I = 2;
7167public:
7168  friend W;
7169};
7170
7171struct B
7172{
7173  int ar[Q<B>::I];
7174};
7175
7176     * Thanks to Ville Voutilainen, G++ now implements [8]C++11 explicit
7177       override control.
7178
7179struct B {
7180  virtual void f() const final;
7181  virtual void f(int);
7182};
7183
7184struct D : B {
7185  void f() const;            // error: D::f attempts to override final B::f
7186  void f(long) override;     // error: doesn't override anything
7187  void f(int) override;      // ok
7188};
7189
7190struct E final { };
7191struct F: E { }; // error: deriving from final class
7192
7193     * G++ now implements [9]C++11 non-static data member initializers.
7194
7195struct A {
7196  int i = 42;
7197} a; // initializes a.i to 42
7198
7199     * Thanks to Ed Smith-Rowland, G++ now implements [10]C++11
7200       user-defined literals.
7201
7202// Not actually a good approximation.  :)
7203constexpr long double operator"" _degrees (long double d) { return d * 0.0175; }
7204long double pi = 180.0_degrees;
7205
7206     * G++ now implements [11]C++11 alias-declarations.
7207
7208template <class T> using Ptr = T*;
7209Ptr<int> ip;  // decltype(ip) is int*
7210
7211     * Thanks to Ville Voutilainen and Pedro Lamarão, G++ now implements
7212       [12]C++11 delegating constructors.
7213
7214struct A {
7215  A(int);
7216  A(): A(42) { } // delegate to the A(int) constructor
7217};
7218
7219     * G++ now fully implements C++11 atomic classes rather than just
7220       integer derived classes.
7221
7222class POD {
7223  int a;
7224  int b;
7225};
7226std::atomic<POD> my_atomic_POD;
7227
7228     * G++ now sets the predefined macro __cplusplus to the correct value,
7229       199711L for C++98/03, and 201103L for C++11.
7230     * G++ now correctly implements the two-phase lookup rules such that
7231       an unqualified name used in a template must have an appropriate
7232       declaration found either in scope at the point of definition of the
7233       template or by argument-dependent lookup at the point of
7234       instantiation. As a result, code that relies on a second
7235       unqualified lookup at the point of instantiation to find functions
7236       declared after the template or in dependent bases will be rejected.
7237       The compiler will suggest ways to fix affected code, and using the
7238       -fpermissive compiler flag will allow the code to compile with a
7239       warning.
7240
7241template <class T>
7242void f() { g(T()); } // error, g(int) not found by argument-dependent lookup
7243void g(int) { } // fix by moving this declaration before the declaration of f
7244
7245template <class T>
7246struct A: T {
7247  // error, B::g(B) not found by argument-dependent lookup
7248  void f() { g(T()); } // fix by using this->g or A::g
7249};
7250
7251struct B { void g(B); };
7252
7253int main()
7254{
7255  f<int>();
7256  A<B>().f();
7257}
7258
7259     * G++ now properly re-uses stack space allocated for temporary
7260       objects when their lifetime ends, which can significantly lower
7261       stack consumption for some C++ functions. As a result of this, some
7262       code with undefined behavior will now break:
7263
7264const int &f(const int &i) { return i; }
7265....
7266const int &x = f(1);
7267const int &y = f(2);
7268
7269       Here, x refers to the temporary allocated to hold the 1 argument,
7270       which only lives until the end of the initialization; it
7271       immediately becomes a dangling reference. So the next statement
7272       re-uses the stack slot to hold the 2 argument, and users of x get
7273       that value instead.
7274       Note that this should not cause any change of behavior for
7275       temporaries of types with non-trivial destructors, as they are
7276       already destroyed at end of full-expression; the change is that now
7277       the storage is released as well.
7278     * A new command-line option -Wdelete-non-virtual-dtor has been added
7279       to warn when delete is used to destroy an instance of a class which
7280       has virtual functions and non-virtual destructor. It is unsafe to
7281       delete an instance of a derived class through a pointer to a base
7282       class if the base class does not have a virtual destructor. This
7283       warning is enabled by -Wall.
7284     * A new command-line option -Wzero-as-null-pointer-constant has been
7285       added to warn when a literal '0' is used as null pointer constant.
7286       It can be useful to facilitate the conversion to nullptr in C++11.
7287     * As per C++98, access-declarations are now deprecated by G++.
7288       Using-declarations are to be used instead. Furthermore, some
7289       efforts have been made to improve the support of class scope
7290       using-declarations. In particular, using-declarations referring to
7291       a dependent type now work as expected ([13]bug c++/14258).
7292     * The ELF symbol visibility of a template instantiation is now
7293       properly constrained by the visibility of its template arguments
7294       ([14]bug c++/35688).
7295
7296    Runtime Library (libstdc++)
7297
7298     * [15]Improved experimental support for the new ISO C++ standard,
7299       C++11, including:
7300          + using noexcept in most of the library;
7301          + implementations of pointer_traits, allocator_traits and
7302            scoped_allocator_adaptor;
7303          + uses-allocator construction for tuple;
7304          + vector meets the allocator-aware container requirements;
7305          + replacing monotonic_clock with steady_clock;
7306          + enabling the thread support library on most POSIX targets;
7307          + many small improvements to conform to the FDIS.
7308     * Added --enable-clocale=newlib configure option.
7309     * Debug Mode iterators for unordered associative containers.
7310     * Avoid polluting the global namespace and do not include <unistd.h>.
7311
7312  Fortran
7313
7314     * The compile flag [16]-fstack-arrays has been added, which causes
7315       all local arrays to be put on stack memory. For some programs this
7316       will improve the performance significantly. If your program uses
7317       very large local arrays, it is possible that you will have to
7318       extend your runtime limits for stack memory.
7319     * The [17]-Ofast flag now also implies [18]-fno-protect-parens and
7320       [19]-fstack-arrays.
7321     * Front-end optimizations can now be selected by the
7322       [20]-ffrontend-optimize option and deselected by the
7323       -fno-frontend-optimize option.
7324     * When front-end optimization removes a function call,
7325       [21]-Wfunction-elimination warns about that.
7326     * When performing front-end-optimization, the
7327       [22]-faggressive-function-elimination option allows the removal of
7328       duplicate function calls even for impure functions.
7329     * The flag [23]-Wreal-q-constant has been added, which warns if
7330       floating-point literals have been specified using q (such as
7331       1.0q0); the q marker is now supported as a vendor extension to
7332       denote quad precision (REAL(16) or, if not available, REAL(10)).
7333       Consider using a kind parameter (such as in 1.0_qp) instead, which
7334       can be obtained via [24]SELECTED_REAL_KIND.
7335     * The GFORTRAN_USE_STDERR environment variable has been removed. GNU
7336       Fortran now always prints error messages to standard error. If you
7337       wish to redirect standard error, please consult the manual for your
7338       OS, shell, batch environment etc. as appropriate.
7339     * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment
7340       variable have been removed. When encountering a serious error,
7341       gfortran will now always abort the program. Whether a core dump is
7342       generated depends on the user environment settings; see the ulimit
7343       -c setting for POSIX shells, limit coredumpsize for C shells, and
7344       the [25]WER user-mode dumps settings on Windows.
7345     * The [26]-fbacktrace option is now enabled by default. When
7346       encountering a fatal error, gfortran will attempt to print a
7347       backtrace to standard error before aborting. It can be disabled
7348       with -fno-backtrace. Note: On POSIX targets with the addr2line
7349       utility from GNU binutils, GNU Fortran can print a backtrace with
7350       function name, file name, line number information in addition to
7351       the addresses; otherwise only the addresses are printed.
7352     * [27]Fortran 2003:
7353          + Generic interface names which have the same name as derived
7354            types are now supported, which allows to write constructor
7355            functions. Note that Fortran does not support static
7356            constructor functions; only default initialization or an
7357            explicit structure-constructor initialization are available.
7358          + [28]Polymorphic (class) arrays are now supported.
7359     * [29]Fortran 2008:
7360          + Support for the DO CONCURRENT construct has been added, which
7361            allows the user to specify that individual loop iterations
7362            have no interdependencies.
7363          + [30]Coarrays: Full single-image support except for polymorphic
7364            coarrays. Additionally, preliminary support for multiple
7365            images via an MPI-based [31]coarray communication library has
7366            been added. Note: The library version is not yet usable as
7367            remote coarray access is not yet possible.
7368     * [32]TS 29113:
7369          + New flag [33]-std=f2008ts permits programs that are expected
7370            to conform to the Fortran 2008 standard and the draft
7371            Technical Specification (TS) 29113 on Further Interoperability
7372            of Fortran with C.
7373          + The OPTIONAL attribute is now allowed for dummy arguments of
7374            BIND(C) procedures.
7375          + The RANK intrinsic has been added.
7376          + The implementation of the ASYNCHRONOUS attribute in GCC is
7377            compatible with the candidate draft of TS 29113 (since GCC
7378            4.6).
7379
7380  Go
7381
7382     * GCC 4.7 implements the [34]Go 1 language standard. The library
7383       support in 4.7.0 is not quite complete, due to release timing.
7384       Release 4.7.1 includes complete support for Go 1. The Go library is
7385       from the Go 1.0.1 release.
7386     * Go has been tested on GNU/Linux and Solaris platforms. It may work
7387       on other platforms as well.
7388
7389New Targets and Target Specific Improvements
7390
7391  ARM
7392
7393     * GCC now supports the Cortex-A7 processor implementing the v7-a
7394       version of the architecture using the option -mcpu=cortex-a7.
7395     * The default vector size in auto-vectorization for NEON is now 128
7396       bits. If vectorization fails thusly, the vectorizer tries again
7397       with 64-bit vectors.
7398     * A new option -mvectorize-with-neon-double was added to allow users
7399       to change the vector size to 64 bits.
7400
7401  AVR
7402
7403     * GCC now supports the XMEGA architecture. This requires GNU binutils
7404       2.22 or later.
7405     * Support for the [35]named address spaces __flash, __flash1, …,
7406       __flash5 and __memx has been added. These address spaces locate
7407       read-only data in flash memory and allow reading from flash memory
7408       by means of ordinary C code, i.e. without the need of (inline)
7409       assembler code:
7410
7411const __flash int values[] = { 42, 31 };
7412
7413int add_values (const __flash int *p, int i)
7414{
7415    return values[i] + *p;
7416}
7417
7418     * Support has been added for the AVR-specific configure option
7419       --with-avrlibc=yes in order to arrange for better integration of
7420       [36]AVR-Libc. This configure option is supported in avr-gcc 4.7.2
7421       and newer and will only take effect in non-RTEMS configurations. If
7422       avr-gcc is configured for RTEMS, the option will be ignored which
7423       is the same as specifying --with-avrlibc=no. See [37]PR54461 for
7424       more technical details.
7425     * Support for AVR-specific [38]built-in functions has been added.
7426     * Support has been added for the signed and unsigned 24-bit scalar
7427       integer types __int24 and __uint24.
7428     * New command-line options -maccumulate-args, -mbranch-cost=cost and
7429       -mstrict-X were added to allow better fine-tuning of code
7430       optimization.
7431     * The command-line option -fdata-sections now also takes affect on
7432       the section names of variables with the progmem attribute.
7433     * A new inline assembler print modifier %i to print a RAM address as
7434       I/O address has been added:
7435
7436#include <avr/io.h> /* Port Definitions from AVR-LibC */
7437
7438void set_portb (uint8_t value)
7439{
7440    asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory");
7441}
7442
7443       The offset between an I/O address and the RAM address for that I/O
7444       location is device-specific. This offset is taken into account when
7445       printing a RAM address with the %i modifier so that the address is
7446       suitable to be used as operand in an I/O command. The address must
7447       be a constant integer known at compile time.
7448     * The inline assembler constraint "R" to represent integers in the
7449       range −6 … 5 has been removed without replacement.
7450     * Many optimizations to:
7451          + 64-bit integer arithmetic
7452          + Widening multiplication
7453          + Integer division by a constant
7454          + Avoid constant reloading in multi-byte instructions.
7455          + Micro-optimizations for special instruction sequences.
7456          + Generic built-in functions like __builtin_ffs*,
7457            __builtin_clz*, etc.
7458          + If-else decision trees generated by switch instructions
7459          + Merging of data located in flash memory
7460          + New libgcc variants for devices with 8-bit wide stack pointer
7461          + …
7462     * Better documentation:
7463          + Handling of EIND and indirect jumps on devices with more than
7464            128 KiB of program memory.
7465          + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function
7466            registers.
7467          + Function attributes OS_main and OS_task.
7468          + AVR-specific built-in macros.
7469
7470  C6X
7471
7472     * Support has been added for the Texas Instruments C6X family of
7473       processors.
7474
7475  CR16
7476
7477     * Support has been added for National Semiconductor's CR16
7478       architecture.
7479
7480  Epiphany
7481
7482     * Support has been added for Adapteva's Epiphany architecture.
7483
7484  IA-32/x86-64
7485
7486     * Support for Intel AVX2 intrinsics, built-in functions and code
7487       generation is available via -mavx2.
7488     * Support for Intel BMI2 intrinsics, built-in functions and code
7489       generation is available via -mbmi2.
7490     * Implementation and automatic generation of __builtin_clz* using the
7491       lzcnt instruction is available via -mlzcnt.
7492     * Support for Intel FMA3 intrinsics and code generation is available
7493       via -mfma.
7494     * A new -mfsgsbase command-line option is available that makes GCC
7495       generate new segment register read/write instructions through
7496       dedicated built-ins.
7497     * Support for the new Intel rdrnd instruction is available via
7498       -mrdrnd.
7499     * Two additional AVX vector conversion instructions are available via
7500       -mf16c.
7501     * Support for new Intel processor codename IvyBridge with RDRND,
7502       FSGSBASE and F16C is available through -march=core-avx-i.
7503     * Support for the new Intel processor codename Haswell with AVX2,
7504       FMA, BMI, BMI2, LZCNT is available through -march=core-avx2.
7505     * Support for new AMD family 15h processors (Piledriver core) is now
7506       available through -march=bdver2 and -mtune=bdver2 options.
7507     * Support for [39]the x32 psABI is now available through the -mx32
7508       option.
7509     * Windows mingw targets are using the -mms-bitfields option by
7510       default.
7511     * Windows x86 targets are using the __thiscall calling convention for
7512       C++ class-member functions.
7513     * Support for the configure option --with-threads=posix for Windows
7514       mingw targets.
7515
7516  MIPS
7517
7518     * GCC now supports thread-local storage (TLS) for MIPS16. This
7519       requires GNU binutils 2.22 or later.
7520     * GCC can now generate code specifically for the Cavium Octeon+ and
7521       Octeon2 processors. The associated command-line options are
7522       -march=octeon+ and -march=octeon2 respectively. Both options
7523       require GNU binutils 2.22 or later.
7524     * GCC can now work around certain 24k errata, under the control of
7525       the command-line option -mfix-24k. These workarounds require GNU
7526       binutils 2.20 or later.
7527     * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build
7528       n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux
7529       toolchain that generates 32-bit code by default. Use the
7530       configure-time option --enable-targets=all to select these extra
7531       multilibs.
7532     * Passing -fno-delayed-branch now also stops the assembler from
7533       automatically filling delay slots.
7534
7535  PowerPC/PowerPC64
7536
7537     * Vectors of type vector long long or vector long are passed and
7538       returned using the same method as other vectors with the VSX
7539       instruction set. Previously GCC did not adhere to the ABI for
7540       128-bit vectors with 64-bit integer base types (PR 48857). This
7541       will also be fixed in the GCC 4.6.1 and 4.5.4 releases.
7542     * A new option -mno-pointers-to-nested-functions was added to allow
7543       AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify
7544       that the compiler should not load up the chain register (r11)
7545       before calling a function through a pointer. If you use this
7546       option, you cannot call nested functions through a pointer, or call
7547       other languages that might use the static chain.
7548     * A new option msave-toc-indirect was added to allow AIX
7549       32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we
7550       save the TOC in the prologue for indirect calls or generate the
7551       save inline. This can speed up some programs that call through a
7552       function pointer a lot, but it can slow down other functions that
7553       only call through a function pointer in exceptional cases.
7554     * The PowerPC port will now enable machine-specific built-in
7555       functions when the user switches the target machine using the
7556       #pragma GCC target or __attribute__ ((__target__ ("target"))) code
7557       sequences. In addition, the target macros are updated. However, due
7558       to the way the -save-temps switch is implemented, you won't see the
7559       effect of these additional macros being defined in preprocessor
7560       output.
7561
7562  SH
7563
7564     * A new option -msoft-atomic has been added. When it is specified,
7565       GCC will generate GNU/Linux-compatible gUSA atomic sequences for
7566       the new __atomic routines.
7567     * Since it is neither supported by GAS nor officially documented,
7568       code generation for little endian SH2A has been disabled.
7569       Specifying -ml with -m2a* will now result in a compiler error.
7570     * The defunct -mbranch-cost option has been fixed.
7571     * Some improvements to the generated code of:
7572          + Utilization of the tst #imm,R0 instruction.
7573          + Dynamic shift instructions on SH2A.
7574          + Integer absolute value calculations.
7575     * The -mdiv= option for targets other than SHmedia has been fixed and
7576       documented.
7577
7578  SPARC
7579
7580     * The option -mflat has been reinstated. When it is specified, the
7581       compiler will generate code for a single register window model.
7582       This is essentially a new implementation and the corresponding
7583       debugger support has been added to GDB 7.4.
7584     * Support for the options -mtune=native and -mcpu=native has been
7585       added on selected native platforms (GNU/Linux and Solaris).
7586     * Support for the SPARC T3 (Niagara 3) processor has been added.
7587     * VIS:
7588          + An intrinsics header visintrin.h has been added.
7589          + Builtin intrinsics for the VIS 1.0 edge handling and pixel
7590            compare instructions have been added.
7591          + The little-endian version of alignaddr is now supported.
7592          + When possible, VIS builtins are marked const, which should
7593            increase the compiler's ability to optimize VIS operations.
7594          + The compiler now properly tracks the %gsr register and how it
7595            behaves as an input for various VIS instructions.
7596          + Akin to fzero, the compiler can now generate fone instructions
7597            in order to set all of the bits of a floating-point register
7598            to 1.
7599          + The documentation for the VIS intrinsics in the GCC manual has
7600            been brought up to date and many inaccuracies were fixed.
7601          + Intrinsics for the VIS 2.0 bmask, bshuffle, and
7602            non-condition-code setting edge instructions have been added.
7603            Their availability is controlled by the new -mvis2 and
7604            -mno-vis2 options. They are enabled by default on
7605            UltraSPARC-III and later CPUs.
7606     * Support for UltraSPARC Fused Multiply-Add floating-point extensions
7607       has been added. These instructions are enabled by default on SPARC
7608       T3 (Niagara 3) and later CPUs.
7609
7610  TILE-Gx/TILEPro
7611
7612     * Support has been added for the Tilera TILE-Gx and TILEPro families
7613       of processors.
7614
7615Other significant improvements
7616
7617     * A new option (-grecord-gcc-switches) was added that appends
7618       compiler command-line options that might affect code generation to
7619       the DW_AT_producer attribute string in the DWARF debugging
7620       information.
7621     * GCC now supports various new GNU extensions to the DWARF debugging
7622       information format, like [40]entry value and [41]call site
7623       information, [42]typed DWARF stack or [43]a more compact macro
7624       representation. Support for these extensions has been added to GDB
7625       7.4. They can be disabled through the -gstrict-dwarf command-line
7626       option.
7627
7628GCC 4.7.1
7629
7630   This is the [44]list of problem reports (PRs) from GCC's bug tracking
7631   system that are known to be fixed in the 4.7.1 release. This list might
7632   not be complete (that is, it is possible that some PRs that have been
7633   fixed are not listed here).
7634
7635   The Go front end in the 4.7.1 release fully supports the [45]Go 1
7636   language standard.
7637
7638GCC 4.7.2
7639
7640   This is the [46]list of problem reports (PRs) from GCC's bug tracking
7641   system that are known to be fixed in the 4.7.2 release. This list might
7642   not be complete (that is, it is possible that some PRs that have been
7643   fixed are not listed here).
7644
7645GCC 4.7.3
7646
7647   This is the [47]list of problem reports (PRs) from GCC's bug tracking
7648   system that are known to be fixed in the 4.7.3 release. This list might
7649   not be complete (that is, it is possible that some PRs that have been
7650   fixed are not listed here).
7651
7652GCC 4.7.4
7653
7654   This is the [48]list of problem reports (PRs) from GCC's bug tracking
7655   system that are known to be fixed in the 4.7.4 release. This list might
7656   not be complete (that is, it is possible that some PRs that have been
7657   fixed are not listed here).
7658
7659
7660    For questions related to the use of GCC, please consult these web
7661    pages and the [49]GCC manuals. If that fails, the
7662    [50]gcc-help@gcc.gnu.org mailing list might help. Comments on these
7663    web pages and the development of GCC are welcome on our developer
7664    list at [51]gcc@gcc.gnu.org. All of [52]our lists have public
7665    archives.
7666
7667   Copyright (C) [53]Free Software Foundation, Inc. Verbatim copying and
7668   distribution of this entire article is permitted in any medium,
7669   provided this notice is preserved.
7670
7671   These pages are [54]maintained by the GCC team. Last modified
7672   2019-11-28[55].
7673
7674References
7675
7676   1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html
7677   2. http://savannah.nongnu.org/bugs/?35407
7678   3. https://gcc.gnu.org/PR18145
7679   4. https://gcc.gnu.org/gcc-4.7/porting_to.html
7680   5. https://gcc.gnu.org/wiki/TransactionalMemory
7681   6. https://gcc.gnu.org/wiki/Atomic/GCCMM
7682   7. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
7683   8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
7684   9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
7685  10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
7686  11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
7687  12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
7688  13. https://gcc.gnu.org/PR14258
7689  14. https://gcc.gnu.org/PR35688
7690  15. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011
7691  16. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
7692  17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689
7693  18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270
7694  19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
7695  20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275
7696  21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170
7697  22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270
7698  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
7699  24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html
7700  25. https://docs.microsoft.com/en-us/windows/win32/wer/collecting-user-mode-dumps
7701  26. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183
7702  27. https://gcc.gnu.org/wiki/Fortran2003Status
7703  28. https://gcc.gnu.org/wiki/OOP
7704  29. https://gcc.gnu.org/wiki/Fortran2008Status
7705  30. https://gcc.gnu.org/wiki/Coarray
7706  31. https://gcc.gnu.org/wiki/CoarrayLib
7707  32. https://gcc.gnu.org/wiki/TS29113Status
7708  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
7709  34. https://golang.org/doc/go1
7710  35. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html
7711  36. http://nongnu.org/avr-libc/
7712  37. https://gcc.gnu.org/PR54461
7713  38. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built_002din-Functions.html
7714  39. https://sites.google.com/site/x32abi/
7715  40. http://www.dwarfstd.org/ShowIssue.php?issue=100909.1
7716  41. http://www.dwarfstd.org/ShowIssue.php?issue=100909.2
7717  42. http://www.dwarfstd.org/ShowIssue.php?issue=140425.1
7718  43. http://www.dwarfstd.org/ShowIssue.php?issue=110722.1
7719  44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1
7720  45. https://golang.org/doc/go1
7721  46. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2
7722  47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3
7723  48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4
7724  49. https://gcc.gnu.org/onlinedocs/
7725  50. mailto:gcc-help@gcc.gnu.org
7726  51. mailto:gcc@gcc.gnu.org
7727  52. https://gcc.gnu.org/lists.html
7728  53. https://www.fsf.org/
7729  54. https://gcc.gnu.org/about.html
7730  55. http://validator.w3.org/check/referer
7731======================================================================
7732http://gcc.gnu.org/gcc-4.6/index.html
7733                             GCC 4.6 Release Series
7734
7735   (This release series is no longer supported.)
7736
7737   April 12, 2013
7738
7739   The [1]GNU project and the GCC developers are pleased to announce the
7740   release of GCC 4.6.4.
7741
7742   This release is a bug-fix release, containing fixes for regressions in
7743   GCC 4.6.3 relative to previous releases of GCC.
7744
7745Release History
7746
7747   GCC 4.6.4
7748          April 12, 2013 ([2]changes, [3]documentation)
7749
7750   GCC 4.6.3
7751          March 1, 2012 ([4]changes, [5]documentation)
7752
7753   GCC 4.6.2
7754          October 26, 2011 ([6]changes, [7]documentation)
7755
7756   GCC 4.6.1
7757          June 27, 2011 ([8]changes, [9]documentation)
7758
7759   GCC 4.6.0
7760          March 25, 2011 ([10]changes, [11]documentation)
7761
7762References and Acknowledgements
7763
7764   GCC used to stand for the GNU C Compiler, but since the compiler
7765   supports several other languages aside from C, it now stands for the
7766   GNU Compiler Collection.
7767
7768   A list of [12]successful builds is updated as new information becomes
7769   available.
7770
7771   The GCC developers would like to thank the numerous people that have
7772   contributed new features, improvements, bug fixes, and other changes as
7773   well as test results to GCC. This [13]amazing group of volunteers is
7774   what makes GCC successful.
7775
7776   For additional information about GCC please refer to the [14]GCC
7777   project web site or contact the [15]GCC development mailing list.
7778
7779   To obtain GCC please use [16]our mirror sites or [17]our version
7780   control system.
7781
7782
7783    For questions related to the use of GCC, please consult these web
7784    pages and the [18]GCC manuals. If that fails, the
7785    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
7786    web pages and the development of GCC are welcome on our developer
7787    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
7788    archives.
7789
7790   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
7791   distribution of this entire article is permitted in any medium,
7792   provided this notice is preserved.
7793
7794   These pages are [23]maintained by the GCC team. Last modified
7795   2020-01-14[24].
7796
7797References
7798
7799   1. http://www.gnu.org/
7800   2. http://gcc.gnu.org/gcc-4.6/changes.html
7801   3. https://gcc.gnu.org/onlinedocs/4.6.4/
7802   4. http://gcc.gnu.org/gcc-4.6/changes.html
7803   5. https://gcc.gnu.org/onlinedocs/4.6.3/
7804   6. http://gcc.gnu.org/gcc-4.6/changes.html
7805   7. https://gcc.gnu.org/onlinedocs/4.6.2/
7806   8. http://gcc.gnu.org/gcc-4.6/changes.html
7807   9. https://gcc.gnu.org/onlinedocs/4.6.1/
7808  10. http://gcc.gnu.org/gcc-4.6/changes.html
7809  11. https://gcc.gnu.org/onlinedocs/4.6.0/
7810  12. http://gcc.gnu.org/gcc-4.6/buildstat.html
7811  13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
7812  14. http://gcc.gnu.org/index.html
7813  15. mailto:gcc@gcc.gnu.org
7814  16. http://gcc.gnu.org/mirrors.html
7815  17. http://gcc.gnu.org/git.html
7816  18. https://gcc.gnu.org/onlinedocs/
7817  19. mailto:gcc-help@gcc.gnu.org
7818  20. mailto:gcc@gcc.gnu.org
7819  21. https://gcc.gnu.org/lists.html
7820  22. https://www.fsf.org/
7821  23. https://gcc.gnu.org/about.html
7822  24. http://validator.w3.org/check/referer
7823======================================================================
7824http://gcc.gnu.org/gcc-4.6/changes.html
7825                             GCC 4.6 Release Series
7826                        Changes, New Features, and Fixes
7827
7828Caveats
7829
7830     * The options -b <machine> and -V <version> have been removed because
7831       they were unreliable. Instead, users should directly run
7832       <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to
7833       run a different version of gcc.
7834     * GCC now has stricter checks for invalid command-line options. In
7835       particular, when gcc was called to link object files rather than
7836       compile source code, it would previously accept and ignore all
7837       options starting with --, including linker options such as
7838       --as-needed and --export-dynamic, although such options would
7839       result in errors if any source code was compiled. Such options, if
7840       unknown to the compiler, are now rejected in all cases; if the
7841       intent was to pass them to the linker, options such as
7842       -Wl,--as-needed should be used.
7843     * Versions of the GNU C library up to and including 2.11.1 included
7844       an [1]incorrect implementation of the cproj function. GCC optimizes
7845       its builtin cproj according to the behavior specified and allowed
7846       by the ISO C99 standard. If you want to avoid discrepancies between
7847       the C library and GCC's builtin transformations when using cproj in
7848       your code, use GLIBC 2.12 or later. If you are using an older GLIBC
7849       and actually rely on the incorrect behavior of cproj, then you can
7850       disable GCC's transformations using -fno-builtin-cproj.
7851     * The C-only intermodule optimization framework (IMA, enabled by
7852       -combine) has been removed in favor of the new generic link-time
7853       optimization framework (LTO) introduced in [2]GCC 4.5.0.
7854     * GCC now ships with the LGPL-licensed libquadmath library, which
7855       provides quad-precision mathematical functions for targets with a
7856       __float128 datatype. __float128 is available for targets on 32-bit
7857       x86, x86-64 and Itanium architectures. The libquadmath library is
7858       automatically built on such targets when building the Fortran
7859       compiler.
7860     * New -Wunused-but-set-variable and -Wunused-but-set-parameter
7861       warnings were added for C, C++, Objective-C and Objective-C++.
7862       These warnings diagnose variables respective parameters which are
7863       only set in the code and never otherwise used. Usually such
7864       variables are useless and often even the value assigned to them is
7865       computed needlessly, sometimes expensively. The
7866       -Wunused-but-set-variable warning is enabled by default by -Wall
7867       flag and -Wunused-but-set-parameter by -Wall -Wextra flags.
7868     * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
7869       rules for the layout of vectors that could lead to wrong code being
7870       generated. Vectors larger than 8 bytes in size are now by default
7871       aligned to an 8-byte boundary. This is an ABI change: code that
7872       makes explicit use of vector types may be incompatible with binary
7873       objects built with older versions of GCC. Auto-vectorized code is
7874       not affected by this change. (This change affects GCC versions
7875       4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.)
7876     * On AVR, variables with the progmem attribute to locate data in
7877       flash memory must be qualified as const.
7878     * Support for a number of older systems and recently unmaintained or
7879       untested target ports of GCC has been declared obsolete in GCC 4.6.
7880       Unless there is activity to revive them, the next release of GCC
7881       will have their sources permanently removed.
7882       All GCC ports for the following processor architectures have been
7883       declared obsolete:
7884          + Argonaut ARC (arc-*)
7885          + National Semiconductor CRX (crx-*)
7886          + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*,
7887            m68hc12-*-*, m6812-*-*)
7888          + Sunplus S+core (score-*)
7889       The following ports for individual systems on particular
7890       architectures have been obsoleted:
7891          + Interix (i[34567]86-*-interix3*)
7892          + NetWare x86 (i[3456x]86-*-netware*)
7893          + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*)
7894          + MCore PE (mcore-*-pe*)
7895          + SH SymbianOS (sh*-*-symbianelf*)
7896          + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*)
7897          + M68K uClinux old ABI (m68k-*-uclinuxoldabi*)
7898          + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*,
7899            vax-*-netbsd*, but not *-*-netbsdelf*)
7900       The i[34567]86-*-pe alias for Cygwin targets has also been
7901       obsoleted; users should configure for i[34567]86-*-cygwin* instead.
7902       Certain configure options to control the set of libraries built
7903       with GCC on some targets have been obsoleted. On ARM targets, the
7904       options --disable-fpu, --disable-26bit, --disable-underscore,
7905       --disable-interwork, --disable-biendian and --disable-nofmult have
7906       been obsoleted. On MIPS targets, the options
7907       --disable-single-float, --disable-biendian and --disable-softfloat
7908       have been obsoleted.
7909     * Support has been removed for all the [3]configurations obsoleted in
7910       GCC 4.5.
7911     * More information on porting to GCC 4.6 from previous versions of
7912       GCC can be found in the [4]porting guide for this release.
7913
7914General Optimizer Improvements
7915
7916     * A new general optimization level, -Ofast, has been introduced. It
7917       combines the existing optimization level -O3 with options that can
7918       affect standards compliance but result in better optimized code.
7919       For example, -Ofast enables -ffast-math.
7920     * Link-time optimization improvements:
7921          + The [5]Scalable Whole Program Optimizer (WHOPR) project has
7922            stabilized to the point of being usable. It has become the
7923            default mode when using the LTO optimization model. Link time
7924            optimization can now split itself into multiple parallel
7925            compilations. Parallelism is controlled with -flto=n (where n
7926            specifies the number of compilations to execute in parallel).
7927            GCC can also cooperate with a GNU make job server by
7928            specifying the -flto=jobserver option and adding + to the
7929            beginning of the Makefile rule executing the linker.
7930            Classical LTO mode can be enforced by -flto-partition=none.
7931            This may result in small code quality improvements.
7932          + A large number of bugs were fixed. GCC itself, Mozilla Firefox
7933            and other large applications can be built with LTO enabled.
7934          + The linker plugin support improvements
7935               o Linker plugin is now enabled by default when the linker
7936                 is detected to have plugin support. This is the case for
7937                 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and
7938                 the Gold linker on ELF targets. Plugin support of the
7939                 Apple linker on Darwin is not compatible with GCC. The
7940                 linker plugin can also be controlled by the
7941                 -fuse-linker-plugin command-line option.
7942               o Resolution information from the linker plugin is used to
7943                 drive whole program assumptions. Use of the linker plugin
7944                 results in more aggressive optimization on binaries and
7945                 on shared libraries that use the hidden visibility
7946                 attribute. Consequently the use of -fwhole-program is not
7947                 necessary in addition to LTO.
7948          + Hidden symbols used from non-LTO objects now have to be
7949            explicitly annotated with externally_visible when the linker
7950            plugin is not used.
7951          + C++ inline functions and virtual tables are now privatized
7952            more aggressively, leading to better inter-procedural
7953            optimization and faster dynamic linking.
7954          + Memory usage and intermediate language streaming performance
7955            have been improved.
7956          + Static constructors and destructors from individual units are
7957            inlined into a single function. This can significantly improve
7958            startup times of large C++ applications where static
7959            constructors are very common. For example, static constructors
7960            are used when including the iostream header.
7961          + Support for the Ada language has been added.
7962     * Interprocedural optimization improvements
7963          + The interprocedural framework was re-tuned for link time
7964            optimization. Several scalability issues were resolved.
7965          + Improved auto-detection of const and pure functions. Newly,
7966            noreturn functions are auto-detected.
7967            The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is
7968            available that informs users when adding attributes to headers
7969            might improve code generation.
7970          + A number of inlining heuristic improvements. In particular:
7971               o Partial inlining is now supported and enabled by default
7972                 at -O2 and greater. The feature can be controlled via
7973                 -fpartial-inlining.
7974                 Partial inlining splits functions with short hot path to
7975                 return. This allows more aggressive inlining of the hot
7976                 path leading to better performance and often to code size
7977                 reductions (because cold parts of functions are not
7978                 duplicated).
7979               o Scalability for large compilation units was improved
7980                 significantly.
7981               o Inlining of callbacks is now more aggressive.
7982               o Virtual methods are considered for inlining when the
7983                 caller is inlined and devirtualization is then possible.
7984               o Inlining when optimizing for size (either in cold regions
7985                 of a program or when compiling with -Os) was improved to
7986                 better handle C++ programs with larger abstraction
7987                 penalty, leading to smaller and faster code.
7988          + The IPA reference optimization pass detecting global variables
7989            used or modified by functions was strengthened and sped up.
7990          + Functions whose address was taken are now optimized out when
7991            all references to them are dead.
7992          + A new inter-procedural static profile estimation pass detects
7993            functions that are executed once or unlikely to be executed.
7994            Unlikely executed functions are optimized for size. Functions
7995            executed once are optimized for size except for the inner
7996            loops.
7997          + On most targets with named section support, functions used
7998            only at startup (static constructors and main), functions used
7999            only at exit and functions detected to be cold are placed into
8000            separate text segment subsections. This extends the
8001            -freorder-functions feature and is controlled by the same
8002            switch. The goal is to improve the startup time of large C++
8003            programs.
8004            Proper function placement requires linker support. GNU ld
8005            2.21.51 on ELF targets was updated to place those functions
8006            together within the text section leading to better code
8007            locality and faster startup times of large C++ programs. The
8008            feature is also supported in the Apple linker. Support in the
8009            gold linker is planned.
8010     * A new switch -fstack-usage has been added. It makes the compiler
8011       output stack usage information for the program, on a per-function
8012       basis, in an auxiliary file.
8013     * A new switch -fcombine-stack-adjustments has been added. It can be
8014       used to enable or disable the compiler's stack-slot combining pass
8015       which before was enabled automatically at -O1 and above, but could
8016       not be controlled on its own.
8017     * A new switch -fstrict-volatile-bitfields has been added. Using it
8018       indicates that accesses to volatile bitfields should use a single
8019       access of the width of the field's type. This option can be useful
8020       for precisely defining and accessing memory-mapped peripheral
8021       registers from C or C++.
8022
8023Compile time and memory usage improvements
8024
8025     * Datastructures used by the dataflow framework in GCC were
8026       reorganized for better memory usage and more cache locality.
8027       Compile time is improved especially on units with large functions
8028       (possibly resulting from a lot of inlining) not fitting into the
8029       processor cache. The compile time of the GCC C compiler binary with
8030       link-time optimization went down by over 10% (benchmarked on x86-64
8031       target).
8032
8033New Languages and Language specific improvements
8034
8035  Ada
8036
8037     * Stack checking has been improved on selected architectures (Alpha,
8038       IA-32/x86-64, RS/6000 and SPARC): it now will detect stack
8039       overflows in all cases on these architectures.
8040     * Initial support for Ada 2012 has been added.
8041
8042  C family
8043
8044     * A new warning, enabled by -Wdouble-promotion, has been added that
8045       warns about cases where a value of type float is implicitly
8046       promoted to double. This is especially helpful for CPUs that handle
8047       the former in hardware, but emulate the latter in software.
8048     * A new function attribute leaf was introduced. This attribute allows
8049       better inter-procedural optimization across calls to functions that
8050       return to the current unit only via returning or exception
8051       handling. This is the case for most library functions that have no
8052       callbacks.
8053     * Support for a new data type __int128 for targets having wide enough
8054       machine-mode support.
8055     * The new function attribute callee_pop_aggregate allows to specify
8056       if the caller or callee is responsible for popping the aggregate
8057       return pointer value from the stack.
8058     * Support for selectively enabling and disabling warnings via #pragma
8059       GCC diagnostic has been added. For instance:
8060#pragma GCC diagnostic error "-Wuninitialized"
8061  foo(a);                       /* error is given for this one */
8062#pragma GCC diagnostic push
8063#pragma GCC diagnostic ignored "-Wuninitialized"
8064  foo(b);                       /* no diagnostic for this one */
8065#pragma GCC diagnostic pop
8066  foo(c);                       /* error is given for this one */
8067#pragma GCC diagnostic pop
8068  foo(d);                       /* depends on command-line options */
8069
8070     * The -fmax-errors=N option is now supported. Using this option
8071       causes the compiler to exit after N errors have been issued.
8072
8073  C
8074
8075     * There is now experimental support for some features from the
8076       upcoming C1X revision of the ISO C standard. This support may be
8077       selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions.
8078       Note that this support is experimental and may change incompatibly
8079       in future releases for consistency with changes to the C1X standard
8080       draft. The following features are newly supported as described in
8081       the N1539 draft of C1X (with changes agreed at the March 2011 WG14
8082       meeting); some other features were already supported with no
8083       compiler changes being needed, or have some support but not in full
8084       accord with N1539 (as amended).
8085          + Static assertions (_Static_assert keyword)
8086          + Typedef redefinition
8087          + New macros in <float.h>
8088          + Anonymous structures and unions
8089     * The new -fplan9-extensions option directs the compiler to support
8090       some extensions for anonymous struct fields which are implemented
8091       by the Plan 9 compiler. A pointer to a struct may be automatically
8092       converted to a pointer to an anonymous field when calling a
8093       function, in order to make the types match. An anonymous struct
8094       field whose type is a typedef name may be referred to using the
8095       typedef name.
8096
8097  C++
8098
8099     * Improved [7]experimental support for the upcoming C++0x ISO C++
8100       standard, including support for constexpr (thanks to Gabriel Dos
8101       Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide),
8102       noexcept, unrestricted unions, range-based for loops (thanks to
8103       Rodrigo Rivas Costa), opaque enum declarations (thanks also to
8104       Rodrigo), implicitly deleted functions and implicit move
8105       constructors.
8106     * When an extern declaration within a function does not match a
8107       declaration in the enclosing context, G++ now properly declares the
8108       name within the namespace of the function rather than the namespace
8109       which was open just before the function definition ([8]c++/43145).
8110     * GCC now warns by default when casting integers to larger pointer
8111       types. These warnings can be disabled with the option
8112       -Wno-int-to-pointer-cast, which is now also available in C++.
8113     * G++ no longer optimizes using the assumption that a value of
8114       enumeration type will fall within the range specified by the
8115       standard, since that assumption is easily violated with a
8116       conversion from integer type ([9]c++/43680). The old behavior can
8117       be restored with -fstrict-enums.
8118     * The new -fnothrow-opt flag changes the semantics of a throw()
8119       exception specification to match the proposed semantics of the
8120       noexcept specification: just call terminate if an exception tries
8121       to propagate out of a function with such an exception
8122       specification. This dramatically reduces or eliminates the code
8123       size overhead from adding the exception specification.
8124     * The new -Wnoexcept flag will suggest adding a noexcept qualifier to
8125       a function that the compiler can tell doesn't throw if it would
8126       change the value of a noexcept expression.
8127     * The -Wshadow option now warns if a local variable or type
8128       declaration shadows another type in C++. Note that the compiler
8129       will not warn if a local variable shadows a struct/class/enum, but
8130       will warn if it shadows an explicit typedef.
8131     * When an identifier is not found in the current scope, G++ now
8132       offers suggestions about which identifier might have been intended.
8133     * G++ now issues clearer diagnostics for missing semicolons after
8134       class, struct, and union definitions.
8135     * G++ now issues clearer diagnostics for missing semicolons after
8136       class member declarations.
8137     * G++ now issues clearer diagnostics when a colon is used in a place
8138       where a double-colon was intended.
8139     * G++ no longer accepts mutable on reference members ([10]c++/33558).
8140       Use -fpermissive to allow the old, non-conforming behaviour.
8141     * A few mangling fixes have been made, to attribute const/volatile on
8142       function pointer types, decltype of a plain decl, and use of a
8143       function parameter in the declaration of another parameter. By
8144       default the compiler still uses the old mangling, but emits aliases
8145       with the new mangling on targets that support strong aliases. Users
8146       can switch over entirely to the new mangling with -fabi-version=5
8147       or -fabi-version=0. -Wabi will now warn about code that uses the
8148       old mangling.
8149     * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified
8150       type to be default initialized unless the type has a user-declared
8151       default constructor. In 4.6.2 G++ implements the proposed
8152       resolution of [11]DR 253, so default initialization is allowed if
8153       it initializes all subobjects. Code that fails to compile can be
8154       fixed by providing an initializer e.g.
8155    struct A { A(); };
8156    struct B : A { int i; };
8157    const B b = B();
8158       Use -fpermissive to allow the old, non-conforming behaviour.
8159
8160    Runtime Library (libstdc++)
8161
8162     * [12]Improved experimental support for the upcoming ISO C++
8163       standard, C++0x, including using constexpr and nullptr.
8164     * Performance improvements to the [13]Debug Mode, thanks to François
8165       Dumont.
8166     * Atomic operations used for reference-counting are annotated so that
8167       they can be understood by race detectors such as Helgrind, see
8168       [14]Data Race Hunting.
8169     * Most libstdc++ standard headers have been changed to no longer
8170       include the cstddef header as an implementation detail. Code that
8171       relied on that header being included as side-effect of including
8172       other standard headers will need to include cstddef explicitly.
8173
8174  Fortran
8175
8176     * On systems supporting the libquadmath library, GNU Fortran now also
8177       supports a quad-precision, kind=16 floating-point data type
8178       (REAL(16), COMPLEX(16)). As the data type is not fully supported in
8179       hardware, calculations might be one to two orders of magnitude
8180       slower than with the 4, 8 or 10 bytes floating-point data types.
8181       This change does not affect systems which support REAL(16) in
8182       hardware nor those which do not support libquadmath.
8183     * Much improved compile time for large array constructors.
8184     * In order to reduce execution time and memory consumption, use of
8185       temporary arrays in assignment expressions is avoided for many
8186       cases. The compiler now reverses loops in order to avoid generating
8187       a temporary array where possible.
8188     * Improved diagnostics, especially with -fwhole-file.
8189     * The -fwhole-file flag is now enabled by default. This improves code
8190       generation and diagnostics. It can be disabled using the deprecated
8191       -fno-whole-file flag.
8192     * Support the generation of Makefile dependencies via the [15]-M...
8193       flags of GCC; you may need to specify the -cpp option in addition.
8194       The dependencies take modules, Fortran's include, and CPP's
8195       #include into account. Note: Using -M for the module path is no
8196       longer supported, use -J instead.
8197     * The flag -Wconversion has been modified to only issue warnings
8198       where a conversion leads to information loss. This drastically
8199       reduces the number of warnings; -Wconversion is thus now enabled
8200       with -Wall. The flag -Wconversion-extra has been added and also
8201       warns about other conversions; -Wconversion-extra typically issues
8202       a huge number of warnings, most of which can be ignored.
8203     * A new command-line option -Wunused-dummy-argument warns about
8204       unused dummy arguments and is included in -Wall. Before,
8205       -Wunused-variable also warned about unused dummy arguments.
8206     * Fortran 2003 support has been extended:
8207          + Improved support for polymorphism between libraries and
8208            programs and for complicated inheritance patterns (cf.
8209            [16]object-oriented programming).
8210          + Experimental support of the ASSOCIATE construct.
8211          + In pointer assignments it is now possible to specify the lower
8212            bounds of the pointer and, for a rank-1 or a simply contiguous
8213            data-target, to remap the bounds.
8214          + Automatic (re)allocation: In intrinsic assignments to
8215            allocatable variables the left-hand side will be automatically
8216            allocated (if unallocated) or reallocated (if the shape or
8217            type parameter is different). To avoid the small performance
8218            penalty, you can use a(:) = ... instead of a = ... for arrays
8219            and character strings – or disable the feature using -std=f95
8220            or -fno-realloc-lhs.
8221          + Deferred type parameter: For scalar allocatable and pointer
8222            variables the character length can be deferred.
8223          + Namelist variables with allocatable and pointer attribute and
8224            nonconstant length type parameter are supported.
8225     * Fortran 2008 support has been extended:
8226          + Experimental [17]coarray support (for one image only, i.e.
8227            num_images() == 1); use the [18]-fcoarray=single flag to
8228            enable it.
8229          + The STOP and the new ERROR STOP statements now support all
8230            constant expressions.
8231          + Support for the CONTIGUOUS attribute.
8232          + Support for ALLOCATE with MOLD.
8233          + Support for the STORAGE_SIZE intrinsic inquiry function.
8234          + Support of the NORM2 and PARITY intrinsic functions.
8235          + The following bit intrinsics were added: POPCNT and POPPAR for
8236            counting the number of 1 bits and returning the parity; BGE,
8237            BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR
8238            for combined left and right shifts, MASKL and MASKR for simple
8239            left and right justified masks, MERGE_BITS for a bitwise merge
8240            using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations,
8241            and the transformational bit intrinsics IALL, IANY and
8242            IPARITY.
8243          + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine.
8244          + Support for the IMPURE attribute for procedures, which allows
8245            for ELEMENTAL procedures without the restrictions of PURE.
8246          + Null pointers (including NULL()) and not allocated variables
8247            can be used as actual argument to optional non-pointer,
8248            non-allocatable dummy arguments, denoting an absent argument.
8249          + Non-pointer variables with TARGET attribute can be used as
8250            actual argument to POINTER dummies with INTENT(IN)
8251          + Pointers including procedure pointers and those in a derived
8252            type (pointer components) can now be initialized by a target
8253            instead of only by NULL.
8254          + The EXIT statement (with construct-name) can now be used to
8255            leave not only the DO but also the ASSOCIATE, BLOCK, IF,
8256            SELECT CASE and SELECT TYPE constructs.
8257          + Internal procedures can now be used as actual argument.
8258          + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS
8259            and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV
8260            have been added; these arrays contain the supported kind
8261            values for the respective types.
8262          + The module procedures C_SIZEOF of the intrinsic module
8263            ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of
8264            ISO_FORTRAN_ENV have been implemented.
8265          + Minor changes: obsolescence diagnostics for ENTRY was added
8266            for -std=f2008; a line may start with a semicolon; for
8267            internal and module procedures END can be used instead of END
8268            SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes
8269            a RADIX argument; intrinsic types are supported for
8270            TYPE(intrinsic-type-spec); multiple type-bound procedures can
8271            be declared in a single PROCEDURE statement; implied-shape
8272            arrays are supported for named constants (PARAMETER). The
8273            transformational, three argument versions of BESSEL_JN and
8274            BESSEL_YN were added – the elemental, two-argument version had
8275            been added in GCC 4.4; note that the transformational
8276            functions use a recurrence algorithm.
8277
8278  Go
8279
8280   Support for the [19]Go programming language has been added to GCC. It
8281   is not enabled by default when you build GCC; use the
8282   --enable-languages configure option to build it. The driver program for
8283   compiling Go code is gccgo.
8284
8285   Go is currently known to work on GNU/Linux and RTEMS. Solaris support
8286   is in progress. It may or may not work on other platforms.
8287
8288  Objective-C and Objective-C++
8289
8290     * The -fobjc-exceptions flag is now required to enable Objective-C
8291       exception and synchronization syntax (introduced by the keywords
8292       @try, @catch, @finally and @synchronized).
8293     * A number of Objective-C 2.0 features and extensions are now
8294       supported by GCC. These features are enabled by default; you can
8295       disable them by using the new -fobjc-std=objc1 command-line option.
8296     * The Objective-C 2.0 dot-syntax is now supported. It is an
8297       alternative syntax for using getters and setters; object.count is
8298       automatically converted into [object count] or [object setCount:
8299       ...] depending on context; for example if (object.count > 0) is
8300       automatically compiled into the equivalent of if ([object count] >
8301       0) while object.count = 0; is automatically compiled into the
8302       equivalent ot [object setCount: 0];. The dot-syntax can be used
8303       with instance and class objects and with any setters or getters, no
8304       matter if they are part of a declared property or not.
8305     * Objective-C 2.0 declared properties are now supported. They are
8306       declared using the new @property keyword, and are most commonly
8307       used in conjunction with the new Objective-C 2.0 dot-syntax. The
8308       nonatomic, readonly, readwrite, assign, retain, copy, setter and
8309       getter attributes are all supported. Marking declared properties
8310       with __attribute__ ((deprecated)) is supported too.
8311     * The Objective-C 2.0 @synthesize and @dynamic keywords are
8312       supported. @synthesize causes the compiler to automatically
8313       synthesize a declared property, while @dynamic is used to disable
8314       all warnings for a declared property for which no implementation is
8315       provided at compile time. Synthesizing declared properties requires
8316       runtime support in most useful cases; to be able to use it with the
8317       GNU runtime, appropriate helper functions have been added to the
8318       GNU Objective-C runtime ABI, and are implemented by the GNU
8319       Objective-C runtime library shipped with GCC.
8320     * The Objective-C 2.0 fast enumeration syntax is supported in
8321       Objective-C. This is currently not yet available in Objective-C++.
8322       Fast enumeration requires support in the runtime, and such support
8323       has been added to the GNU Objective-C runtime library (shipped with
8324       GCC).
8325     * The Objective-C 2.0 @optional keyword is supported. It allows you
8326       to mark methods or properties in a protocol as optional as opposed
8327       to required.
8328     * The Objective-C 2.0 @package keyword is supported. It has currently
8329       the same effect as the @public keyword.
8330     * Objective-C 2.0 method attributes are supported. Currently the
8331       supported attributes are deprecated, sentinel, noreturn and format.
8332     * Objective-C 2.0 method argument attributes are supported. The most
8333       widely used attribute is unused, to mark an argument as unused in
8334       the implementation.
8335     * Objective-C 2.0 class and protocol attributes are supported.
8336       Currently the only supported attribute is deprecated.
8337     * Objective-C 2.0 class extensions are supported. A class extension
8338       has the same syntax as a category declaration with no category
8339       name, and the methods and properties declared in it are added
8340       directly to the main class. It is mostly used as an alternative to
8341       a category to add methods to a class without advertising them in
8342       the public headers, with the advantage that for class extensions
8343       the compiler checks that all the privately declared methods are
8344       actually implemented.
8345     * As a result of these enhancements, GCC can now be used to build
8346       Objective-C and Objective-C++ software that uses Foundation and
8347       other important system frameworks with the NeXT runtime on Darwin 9
8348       and Darwin 10 (OSX 10.5 and 10.6).
8349     * Many bugs in the compiler have been fixed in this release; in
8350       particular, LTO can now be used when compiling Objective-C and
8351       Objective-C++ and the parser is much more robust in dealing with
8352       invalid code.
8353
8354    Runtime Library (libobjc)
8355
8356     * The GNU Objective-C runtime library now defines the macro
8357       __GNU_LIBOBJC__ (with a value that is increased at every release
8358       where there is any change to the API) in objc/objc.h, making it
8359       easy to determine if the GNU Objective-C runtime library is being
8360       used, and if so, which version. Previous versions of the GNU
8361       Objective-C runtime library (and other Objective-C runtime
8362       libraries such as the Apple one) do not define this macro.
8363     * A new Objective-C 2.0 API, almost identical to the one implemented
8364       by the Apple Objective-C runtime, has been implemented in the GNU
8365       Objective-C runtime library. The new API hides the internals of
8366       most runtime structures but provides a more extensive set of
8367       functions to operate on them. It is much easier, for example, to
8368       create or modify classes at runtime. The new API also makes it
8369       easier to port software from Apple to GNU as almost no changes
8370       should be required. The old API is still supported for backwards
8371       compatibility; including the old objc/objc-api.h header file
8372       automatically selects the old API, while including the new
8373       objc/runtime.h header file automatically selects the new API.
8374       Support for the old API is being phased out and upgrading the
8375       software to use the new API is strongly recommended. To check for
8376       the availability of the new API, the __GNU_LIBOBJC__ macro can be
8377       used as older versions of the GNU Objective-C runtime library,
8378       which do not support the new API, do not define such a macro.
8379     * Runtime support for @synchronized has been added.
8380     * Runtime support for Objective-C 2.0 synthesized property accessors
8381       has been added.
8382     * Runtime support for Objective-C 2.0 fast enumeration has been
8383       added.
8384
8385New Targets and Target Specific Improvements
8386
8387  ARM
8388
8389     * GCC now supports the Cortex-M4 processor implementing the v7-em
8390       version of the architecture using the option -mcpu=cortex-m4.
8391     * Scheduling descriptions for the Cortex-M4, the Neon and the
8392       floating point units of the Cortex-A9 and a pipeline description
8393       for the Cortex-A5 have been added.
8394     * Synchronization primitives such as __sync_fetch_and_add and friends
8395       are now inlined for supported architectures rather than calling
8396       into a kernel helper function.
8397     * SSA loop prefetching is enabled by default for the Cortex-A9 at
8398       -O3.
8399     * Several improvements were committed to improve code generation for
8400       the ARM architecture including a rewritten implementation for load
8401       and store multiples.
8402     * Several enhancements were committed to improve SIMD code generation
8403       for NEON by adding support for widening instructions, misaligned
8404       loads and stores, vector conditionals and support for 64 bit
8405       arithmetic.
8406     * Support was added for the Faraday cores fa526, fa606te, fa626te,
8407       fmp626te, fmp626 and fa726te and can be used with the respective
8408       names as parameters to the -mcpu= option.
8409     * Basic support was added for Cortex-A15 and is available through
8410       -mcpu=cortex-a15.
8411     * GCC for AAPCS configurations now more closely adheres to the AAPCS
8412       specification by enabling -fstrict-volatile-bitfields by default.
8413
8414  IA-32/x86-64
8415
8416     * The new -fsplit-stack option permits programs to use a
8417       discontiguous stack. This is useful for threaded programs, in that
8418       it is no longer necessary to specify the maximum stack size when
8419       creating a thread. This feature is currently only implemented for
8420       32-bit and 64-bit x86 GNU/Linux targets.
8421     * Support for emitting profiler counter calls before function
8422       prologues. This is enabled via a new command-line option -mfentry.
8423     * Optimization for the Intel Core 2 processors is now available
8424       through the -march=core2 and -mtune=core2 options.
8425     * Support for Intel Core i3/i5/i7 processors is now available through
8426       the -march=corei7 and -mtune=corei7 options.
8427     * Support for Intel Core i3/i5/i7 processors with AVX is now
8428       available through the -march=corei7-avx and -mtune=corei7-avx
8429       options.
8430     * Support for AMD Bobcat (family 14) processors is now available
8431       through the -march=btver1 and -mtune=btver1 options.
8432     * Support for AMD Bulldozer (family 15) processors is now available
8433       through the -march=bdver1 and -mtune=bdver1 options.
8434     * The default setting (when not optimizing for size) for 32-bit
8435       GNU/Linux and Darwin x86 targets has been changed to
8436       -fomit-frame-pointer. The default can be reverted to
8437       -fno-omit-frame-pointer by configuring GCC with the
8438       --enable-frame-pointer configure option.
8439     * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support
8440       __float128 on 32-bit and 64-bit x86 targets.
8441     * AVX floating-point arithmetic can now be enabled by default at
8442       configure time with the new --with-fpmath=avx option.
8443     * The SSA loop prefetching pass is enabled when using -O3 when
8444       optimizing for CPUs where prefetching is beneficial (AMD CPUs newer
8445       than K6).
8446     * Support for TBM (Trailing Bit Manipulation) built-in functions and
8447       code generation is available via -mtbm.
8448     * Support for AMD's BMI (Bit Manipulation) built-in functions and
8449       code generation is available via -mbmi.
8450
8451  MicroBlaze
8452
8453     * Support has been added for the Xilinx MicroBlaze softcore processor
8454       (microblaze-elf) embedded target. This configurable processor is
8455       supported on several Xilinx Spartan and Virtex FPGAs.
8456
8457  MIPS
8458
8459     * GCC now supports the Loongson 3A processor. Its canonical -march=
8460       and -mtune= name is loongson3a.
8461
8462  MN10300 / AM33
8463
8464     * The inline assembly register constraint "A" has been renamed "c".
8465       This constraint is used to select a floating-point register that
8466       can be used as the destination of a multiply-accumulate
8467       instruction.
8468     * New inline assembly register constraints "A" and "D" have been
8469       added. These constraint letters resolve to all general registers
8470       when compiling for AM33, and resolve to address registers only or
8471       data registers only when compiling for MN10300.
8472     * The MDR register is represented in the compiler. One can access the
8473       register via the "z" constraint in inline assembly. It can be
8474       marked as clobbered or used as a local register variable via the
8475       "mdr" name. The compiler uses the RETF instruction if the function
8476       does not modify the MDR register, so it is important that inline
8477       assembly properly annotate any usage of the register.
8478
8479  PowerPC/PowerPC64
8480
8481     * GCC now supports the Applied Micro Titan processor with
8482       -mcpu=titan.
8483     * The -mrecip option has been added, which indicates whether the
8484       reciprocal and reciprocal square root instructions should be used.
8485     * The -mveclibabi=mass option can be used to enable the compiler to
8486       autovectorize mathematical functions using the Mathematical
8487       Acceleration Subsystem library.
8488     * The -msingle-pic-base option has been added, which instructs the
8489       compiler to avoid loading the PIC base register in function
8490       prologues. The PIC base register must be initialized by the runtime
8491       system.
8492     * The -mblock-move-inline-limit option has been added, which enables
8493       the user to control the maximum size of inlined memcpy calls and
8494       similar.
8495     * PowerPC64 GNU/Linux support for applications requiring a large TOC
8496       section has been improved. A new command-line option,
8497       -mcmodel=MODEL, controls this feature; valid values for MODEL are
8498       small, medium, or large.
8499     * The Altivec builtin functions vec_ld and vec_st have been modified
8500       to generate the Altivec memory instructions LVX and STVX, even if
8501       the -mvsx option is used. In the initial GCC 4.5 release, these
8502       builtin functions were changed to generate VSX memory reference
8503       instructions instead of Altivec memory instructions, but there are
8504       differences between the two instructions. If the VSX instruction
8505       set is available, you can now use the new builtin functions
8506       vec_vsx_ld and vec_vsx_st which always generates the VSX memory
8507       instructions.
8508     * The GCC compiler on AIX now defaults to a process layout with a
8509       larger data space allowing larger programs to be compiled.
8510     * The GCC long double type on AIX 6.1 and above has reverted to 64
8511       bit double precision, matching the AIX XL compiler default, because
8512       of missing C99 symbols required by the GCC runtime.
8513     * The default processor scheduling model and tuning for PowerPC64
8514       GNU/Linux and for AIX 6.1 and above now is POWER7.
8515     * Starting with GCC 4.6.1, vectors of type vector long long or vector
8516       long are passed and returned in the same method as other vectors
8517       with the VSX instruction set. Previously the GCC compiler did not
8518       adhere to the ABI for 128-bit vectors with 64-bit integer base
8519       types (PR 48857). This is also fixed in the GCC 4.5.4 release.
8520
8521  S/390, zSeries and System z9/z10, IBM zEnterprise z196
8522
8523     * Support for the zEnterprise z196 processor has been added. When
8524       using the -march=z196 option, the compiler will generate code
8525       making use of the following instruction facilities:
8526          + Conditional load/store
8527          + Distinct-operands
8528          + Floating-point-extension
8529          + Interlocked-access
8530          + Population-count
8531       The -mtune=z196 option avoids the compare and branch instructions
8532       as well as the load address instruction with an index register as
8533       much as possible and performs instruction scheduling appropriate
8534       for the new out-of-order pipeline architecture.
8535     * When using the -m31 -mzarch options the generated code still
8536       conforms to the 32-bit ABI but uses the general purpose registers
8537       as 64-bit registers internally. This requires a Linux kernel saving
8538       the whole 64-bit registers when doing a context switch. Kernels
8539       providing that feature indicate that by the 'highgprs' string in
8540       /proc/cpuinfo.
8541     * The SSA loop prefetching pass is enabled when using -O3.
8542
8543  SPARC
8544
8545     * GCC now supports the LEON series of SPARC V8 processors. The code
8546       generated by the compiler can either be tuned to it by means of the
8547       --with-tune=leon configure option and -mtune=leon compilation
8548       option, or the compiler can be built for the sparc-leon-{elf,linux}
8549       and sparc-leon3-{elf,linux} targets directly.
8550     * GCC has stopped sign/zero-extending parameter registers in the
8551       callee for functions taking parameters with sub-word size in 32-bit
8552       mode, since this is redundant with the specification of the ABI.
8553       GCC has never done so in 64-bit mode since this is also redundant.
8554     * The command-line option -mfix-at697f has been added to enable the
8555       documented workaround for the single erratum of the Atmel AT697F
8556       processor.
8557
8558Operating Systems
8559
8560  Android
8561
8562     * GCC now supports the Bionic C library and provides a convenient way
8563       of building native libraries and applications for the Android
8564       platform. Refer to the documentation of the -mandroid and -mbionic
8565       options for details on building native code. At the moment, Android
8566       support is enabled only for ARM.
8567
8568  Darwin/Mac OS X
8569
8570     * General
8571          + Initial support for CFString types has been added.
8572            This allows GCC to build projects including the system Core
8573            Foundation frameworks. The GCC Objective-C family supports
8574            CFString "toll-free bridged" as per the Mac OS X system tools.
8575            CFString is also recognized in the context of format
8576            attributes and arguments (see the documentation for format
8577            attributes for limitations). At present, 8-bit character types
8578            are supported.
8579          + Object file size reduction.
8580            The Darwin zeroed memory allocators have been re-written to
8581            make more use of .zerofill sections. For non-debug code, this
8582            can reduce object file size significantly.
8583          + Objective-C family 64-bit support (NeXT ABI 2).
8584            Initial support has been added to support 64-bit Objective-C
8585            code using the Darwin/OS X native (NeXT) runtime. ABI version
8586            2 will be selected automatically when 64-bit code is built.
8587          + Objective-C family 32-bit ABI 1.
8588            For 32-bit code ABI 1 is also now also allowed. At present it
8589            must be selected manually using -fobjc-abi-version=1 where
8590            applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6).
8591     * x86 Architecture
8592          + The -mdynamic-no-pic option has been enabled.
8593            Code supporting -mdynamic-no-pic optimization has been added
8594            and is applicable to -m32 builds. The compiler bootstrap uses
8595            the option where appropriate.
8596          + The default value for -mtune= has been changed.
8597            Since Darwin systems are primarily Xeon, Core-2 or similar the
8598            default tuning has been changed to -mtune=core2.
8599          + Enable 128-bit long double (__float128) support on Darwin.
8600     * PPC Architecture
8601          + Darwin64 ABI.
8602            Several significant bugs have been fixed, such that GCC now
8603            produces code compatible with the Darwin64 PowerPC ABI.
8604          + libffi and boehm-gc.
8605            The Darwin ports of the libffi and boehm-gc libraries have
8606            been upgraded to include a Darwin64 implementation. This means
8607            that powerpc*-*-darwin9 platforms may now, for example, build
8608            Java applications with -m64 enabled.
8609          + Plug-in support has been enabled.
8610          + The -fsection-anchors option is now available although,
8611            presently, not heavily tested.
8612
8613  Solaris 2
8614
8615    New Features
8616
8617     * Support symbol versioning with the Sun linker.
8618     * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+.
8619     * Support thread-local storage (TLS) with the Sun assembler on
8620       Solaris 2/x86.
8621     * Support TLS on Solaris 8/9 if prerequisites are met.
8622     * Support COMDAT group with the GNU assembler and recent Sun linker.
8623     * Support the Sun assembler visibility syntax.
8624     * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp.
8625       -march=pentiumpro (Solaris 8/9).
8626     * Don't use SSE on Solaris 8/9 x86 by default.
8627     * Enable 128-bit long double (__float128) support on Solaris 2/x86.
8628
8629    ABI Change
8630
8631     * Change the ABI for returning 8-byte vectors like __m64 in MMX
8632       registers on Solaris 10+/x86 to match the Sun Studio 12.1+
8633       compilers. This is an incompatible change. If you use such types,
8634       you must either recompile all your code with the new compiler or
8635       use the new -mvect8-ret-in-mem option to remain compatible with
8636       previous versions of GCC and Sun Studio.
8637
8638  Windows x86/x86_64
8639
8640     * Initial support for decimal floating point.
8641     * Support for the __thiscall calling-convention.
8642     * Support for hot-patchable function prologues via the
8643       ms_hook_prologue attribute for x86_64 in addition to 32-bit x86.
8644     * Improvements of stack-probing and stack-allocation mechanisms.
8645     * Support of push/pop-macro pragma as preprocessor command.
8646       With #pragma push_macro("macro-name") the current definition of
8647       macro-name is saved and can be restored with #pragma
8648       pop_macro("macro-name") to its saved definition.
8649     * Enable 128-bit long double (__float128) support on MinGW and
8650       Cygwin.
8651
8652Other significant improvements
8653
8654  Installation changes
8655
8656     * An install-strip make target is provided that installs stripped
8657       executables, and may install libraries with unneeded or debugging
8658       sections stripped.
8659     * On Power7 systems, there is a potential problem if you build the
8660       GCC compiler with a host compiler using options that enable the VSX
8661       instruction set generation. If the host compiler has been patched
8662       so that the vec_ld and vec_st builtin functions generate Altivec
8663       memory instructions instead of VSX memory instructions, then you
8664       should be able to build the compiler with VSX instruction
8665       generation.
8666
8667Changes for GCC Developers
8668
8669   Note: these changes concern developers that develop GCC itself or
8670   software that integrates with GCC, such as plugins, and not the general
8671   GCC users.
8672     * The gengtype utility, which previously was internal to the GCC
8673       build process, has been enchanced to provide GC root information
8674       for plugins as necessary.
8675     * The old GC allocation interface of ggc_alloc and friends was
8676       replaced with a type-safe alternative.
8677
8678GCC 4.6.1
8679
8680   This is the [20]list of problem reports (PRs) from GCC's bug tracking
8681   system that are known to be fixed in the 4.6.1 release. This list might
8682   not be complete (that is, it is possible that some PRs that have been
8683   fixed are not listed here).
8684
8685GCC 4.6.2
8686
8687   This is the [21]list of problem reports (PRs) from GCC's bug tracking
8688   system that are known to be fixed in the 4.6.2 release. This list might
8689   not be complete (that is, it is possible that some PRs that have been
8690   fixed are not listed here).
8691
8692GCC 4.6.3
8693
8694   This is the [22]list of problem reports (PRs) from GCC's bug tracking
8695   system that are known to be fixed in the 4.6.3 release. This list might
8696   not be complete (that is, it is possible that some PRs that have been
8697   fixed are not listed here).
8698
8699GCC 4.6.4
8700
8701   This is the [23]list of problem reports (PRs) from GCC's bug tracking
8702   system that are known to be fixed in the 4.6.4 release. This list might
8703   not be complete (that is, it is possible that some PRs that have been
8704   fixed are not listed here).
8705
8706
8707    For questions related to the use of GCC, please consult these web
8708    pages and the [24]GCC manuals. If that fails, the
8709    [25]gcc-help@gcc.gnu.org mailing list might help. Comments on these
8710    web pages and the development of GCC are welcome on our developer
8711    list at [26]gcc@gcc.gnu.org. All of [27]our lists have public
8712    archives.
8713
8714   Copyright (C) [28]Free Software Foundation, Inc. Verbatim copying and
8715   distribution of this entire article is permitted in any medium,
8716   provided this notice is preserved.
8717
8718   These pages are [29]maintained by the GCC team. Last modified
8719   2019-11-28[30].
8720
8721References
8722
8723   1. https://sourceware.org/bugzilla/show_bug.cgi?id=10401
8724   2. http://gcc.gnu.org/gcc-4.5/changes.html
8725   3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted
8726   4. http://gcc.gnu.org/gcc-4.6/porting_to.html
8727   5. http://gcc.gnu.org/projects/lto/whopr.pdf
8728   6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
8729   7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html
8730   8. https://gcc.gnu.org/PR43145
8731   9. https://gcc.gnu.org/PR43680
8732  10. https://gcc.gnu.org/PR33558
8733  11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253
8734  12. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x
8735  13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html
8736  14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races
8737  15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html
8738  16. https://gcc.gnu.org/wiki/OOP
8739  17. https://gcc.gnu.org/wiki/Coarray
8740  18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233
8741  19. https://golang.org/
8742  20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1
8743  21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2
8744  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3
8745  23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4
8746  24. https://gcc.gnu.org/onlinedocs/
8747  25. mailto:gcc-help@gcc.gnu.org
8748  26. mailto:gcc@gcc.gnu.org
8749  27. https://gcc.gnu.org/lists.html
8750  28. https://www.fsf.org/
8751  29. https://gcc.gnu.org/about.html
8752  30. http://validator.w3.org/check/referer
8753======================================================================
8754http://gcc.gnu.org/gcc-4.5/index.html
8755                             GCC 4.5 Release Series
8756
8757   (This release series is no longer supported.)
8758
8759   Jul 2, 2012
8760
8761   The [1]GNU project and the GCC developers are pleased to announce the
8762   release of GCC 4.5.4.
8763
8764   This release is a bug-fix release, containing fixes for regressions in
8765   GCC 4.5.3 relative to previous releases of GCC.
8766
8767Release History
8768
8769   GCC 4.5.4
8770          Jul 2, 2012 ([2]changes)
8771
8772   GCC 4.5.3
8773          Apr 28, 2011 ([3]changes)
8774
8775   GCC 4.5.2
8776          Dec 16, 2010 ([4]changes)
8777
8778   GCC 4.5.1
8779          Jul 31, 2010 ([5]changes)
8780
8781   GCC 4.5.0
8782          April 14, 2010 ([6]changes)
8783
8784References and Acknowledgements
8785
8786   GCC used to stand for the GNU C Compiler, but since the compiler
8787   supports several other languages aside from C, it now stands for the
8788   GNU Compiler Collection.
8789
8790   A list of [7]successful builds is updated as new information becomes
8791   available.
8792
8793   The GCC developers would like to thank the numerous people that have
8794   contributed new features, improvements, bug fixes, and other changes as
8795   well as test results to GCC. This [8]amazing group of volunteers is
8796   what makes GCC successful.
8797
8798   For additional information about GCC please refer to the [9]GCC project
8799   web site or contact the [10]GCC development mailing list.
8800
8801   To obtain GCC please use [11]our mirror sites or [12]our version
8802   control system.
8803
8804
8805    For questions related to the use of GCC, please consult these web
8806    pages and the [13]GCC manuals. If that fails, the
8807    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
8808    web pages and the development of GCC are welcome on our developer
8809    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
8810    archives.
8811
8812   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
8813   distribution of this entire article is permitted in any medium,
8814   provided this notice is preserved.
8815
8816   These pages are [18]maintained by the GCC team. Last modified
8817   2020-01-14[19].
8818
8819References
8820
8821   1. http://www.gnu.org/
8822   2. http://gcc.gnu.org/gcc-4.5/changes.html
8823   3. http://gcc.gnu.org/gcc-4.5/changes.html
8824   4. http://gcc.gnu.org/gcc-4.5/changes.html
8825   5. http://gcc.gnu.org/gcc-4.5/changes.html
8826   6. http://gcc.gnu.org/gcc-4.5/changes.html
8827   7. http://gcc.gnu.org/gcc-4.5/buildstat.html
8828   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
8829   9. http://gcc.gnu.org/index.html
8830  10. mailto:gcc@gcc.gnu.org
8831  11. http://gcc.gnu.org/mirrors.html
8832  12. http://gcc.gnu.org/git.html
8833  13. https://gcc.gnu.org/onlinedocs/
8834  14. mailto:gcc-help@gcc.gnu.org
8835  15. mailto:gcc@gcc.gnu.org
8836  16. https://gcc.gnu.org/lists.html
8837  17. https://www.fsf.org/
8838  18. https://gcc.gnu.org/about.html
8839  19. http://validator.w3.org/check/referer
8840======================================================================
8841http://gcc.gnu.org/gcc-4.5/changes.html
8842                             GCC 4.5 Release Series
8843                        Changes, New Features, and Fixes
8844
8845Caveats
8846
8847     * GCC now requires the [1]MPC library in order to build. See the
8848       [2]prerequisites page for version requirements.
8849     * Support for a number of older systems and recently unmaintained or
8850       untested target ports of GCC has been declared obsolete in GCC 4.5.
8851       Unless there is activity to revive them, the next release of GCC
8852       will have their sources permanently removed.
8853       The following ports for individual systems on particular
8854       architectures have been obsoleted:
8855          + IRIX releases before 6.5 (mips-sgi-irix5*,
8856            mips-sgi-irix6.[0-4])
8857          + Solaris 7 (*-*-solaris2.7)
8858          + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*,
8859            alpha-dec-osf5.0*)
8860          + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions
8861            can be found in the [3]announcement.
8862       Support for the classic POWER architecture implemented in the
8863       original RIOS and RIOS2 processors of the old IBM RS/6000 product
8864       line has been obsoleted in the rs6000 port. This does not affect
8865       the new generation Power and PowerPC architectures.
8866     * Support has been removed for all the [4]configurations obsoleted in
8867       GCC 4.4.
8868     * Support has been removed for the protoize and unprotoize utilities,
8869       obsoleted in GCC 4.4.
8870     * Support has been removed for tuning for Itanium1 (Merced) variants.
8871       Note that code tuned for Itanium2 should also run correctly on
8872       Itanium1.
8873     * GCC now generates unwind info also for epilogues. DWARF debuginfo
8874       generated by GCC now uses more features of DWARF3 than before, and
8875       also some DWARF4 features. GDB older than 7.0 is not able to handle
8876       either of these, so to debug GCC 4.5 generated binaries or
8877       libraries GDB 7.0 or later is needed. You can disable use of DWARF4
8878       features with the -gdwarf-3 -gstrict-dwarf options, or use
8879       -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but
8880       epilogue unwind info is emitted unconditionally whenever unwind
8881       info is emitted.
8882     * On x86 targets, code containing floating-point calculations may run
8883       significantly more slowly when compiled with GCC 4.5 in strict C99
8884       conformance mode than they did with earlier GCC versions. This is
8885       due to stricter standard conformance of the compiler and can be
8886       avoided by using the option -fexcess-precision=fast; also see
8887       [5]below.
8888     * The function attribute noinline no longer prevents GCC from cloning
8889       the function. A new attribute noclone has been introduced for this
8890       purpose. Cloning a function means that it is duplicated and the new
8891       copy is specialized for certain contexts (for example when a
8892       parameter is a known constant).
8893
8894General Optimizer Improvements
8895
8896     * The -save-temps now takes an optional argument. The -save-temps and
8897       -save-temps=cwd switches write the temporary files in the current
8898       working directory based on the original source file. The
8899       -save-temps=obj switch will write files into the directory
8900       specified with the -o option, and the intermediate filenames are
8901       based on the output file. This will allow the user to get the
8902       compiler intermediate files when doing parallel builds without two
8903       builds of the same filename located in different directories from
8904       interfering with each other.
8905     * Debugging dumps are now created in the same directory as the object
8906       file rather than in the current working directory. This allows the
8907       user to get debugging dumps when doing parallel builds without two
8908       builds of the same filename interfering with each other.
8909     * GCC has been integrated with the MPC library. This allows GCC to
8910       evaluate complex arithmetic at compile time [6]more accurately. It
8911       also allows GCC to evaluate calls to complex built-in math
8912       functions having constant arguments and replace them at compile
8913       time with their mathematically equivalent results. In doing so, GCC
8914       can generate correct results regardless of the math library
8915       implementation or floating point precision of the host platform.
8916       This also allows GCC to generate identical results regardless of
8917       whether one compiles in native or cross-compile configurations to a
8918       particular target. The following built-in functions take advantage
8919       of this new capability: cacos, cacosh, casin, casinh, catan,
8920       catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan,
8921       and ctanh. The float and long double variants of these functions
8922       (e.g. csinf and csinl) are also handled.
8923     * A new link-time optimizer has been added ([7]-flto). When this
8924       option is used, GCC generates a bytecode representation of each
8925       input file and writes it to specially-named sections in each object
8926       file. When the object files are linked together, all the function
8927       bodies are read from these named sections and instantiated as if
8928       they had been part of the same translation unit. This enables
8929       interprocedural optimizations to work across different files (and
8930       even different languages), potentially improving the performance of
8931       the generated code. To use the link-timer optimizer, -flto needs to
8932       be specified at compile time and during the final link. If the
8933       program does not require any symbols to be exported, it is possible
8934       to combine -flto and the experimental [8]-fwhopr with
8935       [9]-fwhole-program to allow the interprocedural optimizers to use
8936       more aggressive assumptions.
8937     * The automatic parallelization pass was enhanced to support
8938       parallelization of outer loops.
8939     * Automatic parallelization can be enabled as part of Graphite. In
8940       addition to -ftree-parallelize-loops=, specify
8941       -floop-parallelize-all to enable the Graphite-based optimization.
8942     * The infrastructure for optimizing based on [10]restrict qualified
8943       pointers has been rewritten and should result in code generation
8944       improvements. Optimizations based on restrict qualified pointers
8945       are now also available when using -fno-strict-aliasing.
8946     * There is a new optimization pass that attempts to change prototype
8947       of functions to avoid unused parameters, pass only relevant parts
8948       of structures and turn arguments passed by reference to arguments
8949       passed by value when possible. It is enabled by -O2 and above as
8950       well as -Os and can be manually invoked using the new command-line
8951       switch -fipa-sra.
8952     * GCC now optimize exception handling code. In particular cleanup
8953       regions that are proved to not have any effect are optimized out.
8954
8955New Languages and Language specific improvements
8956
8957  All languages
8958
8959     * The -fshow-column option is now on by default. This means error
8960       messages now have a column associated with them.
8961
8962  Ada
8963
8964     * Compilation of programs heavily using discriminated record types
8965       with variant parts has been sped up and generates more compact
8966       code.
8967     * Stack checking now works reasonably well on most plaforms. In some
8968       specific cases, stack overflows may still fail to be detected, but
8969       a compile-time warning will be issued for these cases.
8970
8971  C family
8972
8973     * If a header named in a #include directive is not found, the
8974       compiler exits immediately. This avoids a cascade of errors arising
8975       from declarations expected to be found in that header being
8976       missing.
8977     * A new built-in function __builtin_unreachable() has been added that
8978       tells the compiler that control will never reach that point. It may
8979       be used after asm statements that terminate by transferring control
8980       elsewhere, and in other places that are known to be unreachable.
8981     * The -Wlogical-op option now warns for logical expressions such as
8982       (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be
8983       mistakes. This option is disabled by default.
8984     * An asm goto feature has been added to allow asm statements that
8985       jump to C labels.
8986     * C++0x raw strings are supported for C++ and for C with -std=gnu99.
8987     * The deprecated attribute now takes an optional string argument, for
8988       example, __attribute__((deprecated("text string"))), that will be
8989       printed together with the deprecation warning.
8990
8991  C
8992
8993     * The -Wenum-compare option, which warns when comparing values of
8994       different enum types, now works for C. It formerly only worked for
8995       C++. This warning is enabled by -Wall. It may be avoided by using a
8996       type cast.
8997     * The -Wcast-qual option now warns about casts which are unsafe in
8998       that they permit const-correctness to be violated without further
8999       warnings. Specifically, it warns about cases where a qualifier is
9000       added when all the lower types are not const. For example, it warns
9001       about a cast from char ** to const char **.
9002     * The -Wc++-compat option is significantly improved. It issues new
9003       warnings for:
9004          + Using C++ reserved operator names as identifiers.
9005          + Conversions to enum types without explicit casts.
9006          + Using va_arg with an enum type.
9007          + Using different enum types in the two branches of ?:.
9008          + Using ++ or -- on a variable of enum type.
9009          + Using the same name as both a struct, union or enum tag and a
9010            typedef, unless the typedef refers to the tagged type itself.
9011          + Using a struct, union, or enum which is defined within another
9012            struct or union.
9013          + A struct field defined using a typedef if there is a field in
9014            the struct, or an enclosing struct, whose name is the typedef
9015            name.
9016          + Duplicate definitions at file scope.
9017          + Uninitialized const variables.
9018          + A global variable with an anonymous struct, union, or enum
9019            type.
9020          + Using a string constant to initialize a char array whose size
9021            is the length of the string.
9022     * The new -Wjump-misses-init option warns about cases where a goto or
9023       switch skips the initialization of a variable. This sort of branch
9024       is an error in C++ but not in C. This warning is enabled by
9025       -Wc++-compat.
9026     * GCC now ensures that a C99-conforming <stdint.h> is present on most
9027       targets, and uses information about the types in this header to
9028       implement the Fortran bindings to those types. GCC does not ensure
9029       the presence of such a header, and does not implement the Fortran
9030       bindings, on the following targets: NetBSD, VxWorks, VMS,
9031       SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF.
9032     * GCC now implements C90- and C99-conforming rules for constant
9033       expressions. This may cause warnings or errors for some code using
9034       expressions that can be folded to a constant but are not constant
9035       expressions as defined by ISO C.
9036     * All known target-independent C90 and C90 Amendment 1 conformance
9037       bugs, and all known target-independent C99 conformance bugs not
9038       related to floating point or extended identifiers, have been fixed.
9039     * The C decimal floating point support now includes support for the
9040       FLOAT_CONST_DECIMAL64 pragma.
9041     * The named address space feature from ISO/IEC TR 18037 is now
9042       supported. This is currently only implemented for the SPU
9043       processor.
9044
9045  C++
9046
9047     * Improved [11]experimental support for the upcoming C++0x ISO C++
9048       standard, including support for raw strings, lambda expressions and
9049       explicit type conversion operators.
9050     * When printing the name of a class template specialization, G++ will
9051       now omit any template arguments which come from default template
9052       arguments. This behavior (and the pretty-printing of function
9053       template specializations as template signature and arguments) can
9054       be disabled with the -fno-pretty-templates option.
9055     * Access control is now applied to typedef names used in a template,
9056       which may cause G++ to reject some ill-formed code that was
9057       accepted by earlier releases. The -fno-access-control option can be
9058       used as a temporary workaround until the code is corrected.
9059     * Compilation time for code that uses templates should now scale
9060       linearly with the number of instantiations rather than
9061       quadratically, as template instantiations are now looked up using
9062       hash tables.
9063     * Declarations of functions that look like builtin declarations of
9064       library functions are only considered to be redeclarations if they
9065       are declared with extern "C". This may cause problems with code
9066       that omits extern "C" on hand-written declarations of C library
9067       functions such as abort or memcpy. Such code is ill-formed, but was
9068       accepted by earlier releases.
9069     * Diagnostics that used to complain about passing non-POD types to
9070       ... or jumping past the declaration of a non-POD variable now check
9071       for triviality rather than PODness, as per C++0x.
9072     * In C++0x mode local and anonymous classes are now allowed as
9073       template arguments, and in declarations of variables and functions
9074       with linkage, so long as any such declaration that is used is also
9075       defined ([12]DR 757).
9076     * Labels may now have attributes, as has been permitted for a while
9077       in C. This is only permitted when the label definition and the
9078       attribute specifier is followed by a semicolon—i.e., the label
9079       applies to an empty statement. The only useful attribute for a
9080       label is unused.
9081     * G++ now implements [13]DR 176. Previously G++ did not support using
9082       the injected-class-name of a template base class as a type name,
9083       and lookup of the name found the declaration of the template in the
9084       enclosing scope. Now lookup of the name finds the
9085       injected-class-name, which can be used either as a type or as a
9086       template, depending on whether or not the name is followed by a
9087       template argument list. As a result of this change, some code that
9088       was previously accepted may be ill-formed because
9089         1. The injected-class-name is not accessible because it's from a
9090            private base, or
9091         2. The injected-class-name cannot be used as an argument for a
9092            template template parameter.
9093       In either of these cases, the code can be fixed by adding a
9094       nested-name-specifier to explicitly name the template. The first
9095       can be worked around with -fno-access-control; the second is only
9096       rejected with -pedantic.
9097     * A new standard mangling for SIMD vector types has been added, to
9098       avoid name clashes on systems with vectors of varying length. By
9099       default the compiler still uses the old mangling, but emits aliases
9100       with the new mangling on targets that support strong aliases. Users
9101       can switch over entirely to the new mangling with -fabi-version=4
9102       or -fabi-version=0. -Wabi will now warn about code that uses the
9103       old mangling.
9104     * The command-line option -ftemplate-depth-N is now written as
9105       -ftemplate-depth=N and the old form is deprecated.
9106     * Conversions between NULL and non-pointer types are now warned by
9107       default. The new option -Wno-conversion-null disables these
9108       warnings. Previously these warnings were only available when using
9109       -Wconversion explicitly.
9110
9111    Runtime Library (libstdc++)
9112
9113     * Improved experimental support for the upcoming ISO C++ standard,
9114       C++0x, including:
9115          + Support for <future>, <functional>, and <random>.
9116          + Existing facilities now exploit explicit operators and the
9117            newly implemented core C++0x features.
9118          + The header <cstdatomic> has been renamed to <atomic>.
9119     * An experimental [14]profile mode has been added. This is an
9120       implementation of many C++ standard library constructs with an
9121       additional analysis layer that gives performance improvement advice
9122       based on recognition of suboptimal usage patterns. For example,
9123#include <vector>
9124int main()
9125{
9126  std::vector<int> v;
9127  for (int k = 0; k < 1024; ++k)
9128    v.insert(v.begin(), k);
9129}
9130
9131       When instrumented via the profile mode, can return suggestions
9132       about the initial size and choice of the container used as follows:
9133vector-to-list: improvement = 5: call stack = 0x804842c ...
9134    : advice = change std::vector to std::list
9135vector-size: improvement = 3: call stack = 0x804842c ...
9136    : advice = change initial container size from 0 to 1024
9137
9138       These constructs can be substituted for the normal libstdc++
9139       constructs on a piecemeal basis, or all existing components can be
9140       transformed via the -D_GLIBCXX_PROFILE macro.
9141     * [15]Support for decimal floating-point arithmetic (aka ISO C++ TR
9142       24733) has been added. This support is in header file
9143       <decimal/decimal>, uses namespace std::decimal, and includes
9144       classes decimal32, decimal64, and decimal128.
9145     * Sources have been audited for application of function attributes
9146       nothrow, const, pure, and noreturn.
9147     * Python pretty-printers have been added for many standard library
9148       components that simplify the internal representation and present a
9149       more intuitive view of components when used with
9150       appropriately-advanced versions of GDB. For more information,
9151       please consult the more [16]detailed description.
9152     * The default behavior for comparing typeinfo names has changed, so
9153       in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero.
9154     * The new -static-libstdc++ option directs g++ to link the C++
9155       library statically, even if the default would normally be to link
9156       it dynamically.
9157
9158  Fortran
9159
9160     * The COMMON default padding has been changed – instead of adding the
9161       padding before a variable it is now added afterwards, which
9162       increases the compatibility with other vendors and helps to obtain
9163       the correct output in some cases. Cf. also the -falign-commons
9164       option ([17]added in 4.4).
9165     * The -finit-real= option now also supports the value snan for
9166       signaling not-a-number; to be effective, one additionally needs to
9167       enable trapping (e.g. via -ffpe-trap=). Note: Compile-time
9168       optimizations can turn a signaling NaN into a quiet one.
9169     * The new option -fcheck= has been added with the options bounds,
9170       array-temps, do, pointer, and recursive. The bounds and array-temps
9171       options are equivalent to -fbounds-check and
9172       -fcheck-array-temporaries. The do option checks for invalid
9173       modification of loop iteration variables, and the recursive option
9174       tests for recursive calls to subroutines/functions which are not
9175       marked as recursive. With pointer pointer association checks in
9176       calls are performed; however, neither undefined pointers nor
9177       pointers in expressions are handled. Using -fcheck=all enables all
9178       these run-time checks.
9179     * The run-time checking -fcheck=bounds now warns about invalid string
9180       lengths of character dummy arguments. Additionally, more
9181       compile-time checks have been added.
9182     * The new option [18]-fno-protect-parens has been added; if set, the
9183       compiler may reorder REAL and COMPLEX expressions without regard to
9184       parentheses.
9185     * GNU Fortran no longer links against libgfortranbegin. As before,
9186       MAIN__ (assembler symbol name) is the actual Fortran main program,
9187       which is invoked by the main function. However, main is now
9188       generated and put in the same object file as MAIN__. For the time
9189       being, libgfortranbegin still exists for backward compatibility.
9190       For details see the new [19]Mixed-Language Programming chapter in
9191       the manual.
9192     * The I/O library was restructured for performance and cleaner code.
9193     * Array assignments and WHERE are now run in parallel when OpenMP's
9194       WORKSHARE is used.
9195     * The experimental option -fwhole-file was added. The option allows
9196       whole-file checking of procedure arguments and allows for better
9197       optimizations. It can also be used with -fwhole-program, which is
9198       now also supported in gfortran.
9199     * More Fortran 2003 and Fortran 2008 mathematical functions can now
9200       be used as initialization expressions.
9201     * Some extended attributes such as STDCALL are now supported via the
9202       [20]GCC$ compiler directive.
9203     * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN
9204       intrinsic behaves now as if zero were always positive.
9205     * For legacy compatibiliy: On Cygwin and MinGW, the special files
9206       CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now
9207       supported.
9208     * Fortran 2003 support has been extended:
9209          + Procedure-pointer function results and procedure-pointer
9210            components (including PASS),
9211          + allocatable scalars (experimental),
9212          + DEFERRED type-bound procedures,
9213          + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements
9214            have been implemented.
9215          + The ALLOCATE statement supports type-specs and the SOURCE=
9216            argument.
9217          + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC
9218            type-bound procedure (i.e. as type-bound operators).
9219          + Rounding (ROUND=, RZ, ...) for output is now supported.
9220          + The INT_FAST{8,16,32,64,128}_T kind type parameters of the
9221            intrinsic module ISO_C_BINDING are now supported, except for
9222            the targets listed above as ones where GCC does not have
9223            <stdint.h> type information.
9224          + Extensible derived types with type-bound procedure or
9225            procedure pointer with PASS attribute now have to use CLASS in
9226            line with the Fortran 2003 standard; the workaround to use
9227            TYPE is no longer supported.
9228          + [21]Experimental, incomplete support for polymorphism,
9229            including CLASS, SELECT TYPE and dynamic dispatch of
9230            type-bound procedure calls. Some features do not work yet such
9231            as unlimited polymorphism (CLASS(*)).
9232     * Fortran 2008 support has been extended:
9233          + The OPEN statement now supports the NEWUNIT= option, which
9234            returns a unique file unit, thus preventing inadvertent use of
9235            the same unit in different parts of the program.
9236          + Support for unlimited format items has been added.
9237          + The INT{8,16,32} and REAL{32,64,128} kind type parameters of
9238            the intrinsic module ISO_FORTRAN_ENV are now supported.
9239          + Using complex arguments with TAN, SINH, COSH, TANH, ASIN,
9240            ACOS, and ATAN is now possible; the functions ASINH, ACOSH,
9241            and ATANH have been added (for real and complex arguments) and
9242            ATAN(Y,X) is now an alias for ATAN2(Y,X).
9243          + The BLOCK construct has been implemented.
9244
9245New Targets and Target Specific Improvements
9246
9247  AIX
9248
9249     * Full cross-toolchain support now available with GNU Binutils
9250
9251  ARM
9252
9253     * GCC now supports the Cortex-M0 and Cortex-A5 processors.
9254     * GCC now supports the ARM v7E-M architecture.
9255     * GCC now supports VFPv4-based FPUs and FPUs with
9256       single-precision-only VFP.
9257     * GCC has many improvements to optimization for other ARM processors,
9258       including scheduling support for the integer pipeline on Cortex-A9.
9259     * GCC now supports the IEEE 754-2008 half-precision floating-point
9260       type, and a variant ARM-specific half-precision type. This type is
9261       specified using __fp16, with the layout determined by
9262       -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and
9263       VFPv4 half-precision instructions will be used.
9264     * GCC now supports the variant of AAPCS that uses VFP registers for
9265       parameter passing and return values.
9266
9267  AVR
9268
9269     * The -mno-tablejump option has been removed because it has the same
9270       effect as the -fno-jump-tables option.
9271     * Added support for these new AVR devices:
9272          + ATmega8U2
9273          + ATmega16U2
9274          + ATmega32U2
9275
9276  IA-32/x86-64
9277
9278     * GCC now will set the default for -march= based on the configure
9279       target.
9280     * GCC now supports handling floating-point excess precision arising
9281       from use of the x87 floating-point unit in a way that conforms to
9282       ISO C99. This is enabled with -fexcess-precision=standard and with
9283       standards conformance options such as -std=c99, and may be disabled
9284       using -fexcess-precision=fast.
9285     * Support for the Intel Atom processor is now available through the
9286       -march=atom and -mtune=atom options.
9287     * A new -mcrc32 option is now available to enable crc32 intrinsics.
9288     * A new -mmovbe option is now available to enable GCC to use the
9289       movbe instruction to implement __builtin_bswap32 and
9290       __builtin_bswap64.
9291     * SSE math now can be enabled by default at configure time with the
9292       new --with-fpmath=sse option.
9293     * There is a new intrinsic header file, <x86intrin.h>. It should be
9294       included before using any IA-32/x86-64 intrinsics.
9295     * Support for the XOP, FMA4, and LWP instruction sets for the AMD
9296       Orochi processors are now available with the -mxop, -mfma4, and
9297       -mlwp options.
9298     * The -mabm option enables GCC to use the popcnt and lzcnt
9299       instructions on AMD processors.
9300     * The -mpopcnt option enables GCC to use the popcnt instructions on
9301       both AMD and Intel processors.
9302
9303  M68K/ColdFire
9304
9305     * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x
9306       and 5441x devices.
9307     * GCC now supports thread-local storage (TLS) on M68K and ColdFire
9308       processors.
9309
9310  MeP
9311
9312   Support has been added for the Toshiba Media embedded Processor (MeP,
9313   or mep-elf) embedded target.
9314
9315  MIPS
9316
9317     * GCC now supports MIPS 1004K processors.
9318     * GCC can now be configured with options --with-arch-32,
9319       --with-arch-64, --with-tune-32 and --with-tune-64 to control the
9320       default optimization separately for 32-bit and 64-bit modes.
9321     * MIPS targets now support an alternative _mcount interface, in which
9322       register $12 points to the function's save slot for register $31.
9323       This interface is selected by the -mcount-ra-address option; see
9324       the documentation for more details.
9325     * GNU/Linux targets can now generate read-only .eh_frame sections.
9326       This optimization requires GNU binutils 2.20 or above, and is only
9327       available if GCC is configured with a suitable version of binutils.
9328     * GNU/Linux targets can now attach special relocations to indirect
9329       calls, so that the linker can turn them into direct jumps or
9330       branches. This optimization requires GNU binutils 2.20 or later,
9331       and is automatically selected if GCC is configured with an
9332       appropriate version of binutils. It can be explicitly enabled or
9333       disabled using the -mrelax-pic-calls command-line option.
9334     * GCC now generates more heavily-optimized atomic operations on
9335       Octeon processors.
9336     * MIPS targets now support the -fstack-protector option.
9337     * GCC now supports an -msynci option, which specifies that synci is
9338       enough to flush the instruction cache, without help from the
9339       operating system. GCC uses this information to optimize
9340       automatically-generated cache flush operations, such as those used
9341       for nested functions in C. There is also a --with-synci
9342       configure-time option, which makes -msynci the default.
9343     * GCC supports four new function attributes for interrupt handlers:
9344       interrupt, use_shadow_register_set, keep_interrupts_masked and
9345       use_debug_exception_return. See the documentation for more details
9346       about these attributes.
9347
9348  RS/6000 (POWER/PowerPC)
9349
9350     * GCC now supports the Power ISA 2.06, which includes the VSX
9351       instructions that add vector 64-bit floating point support, new
9352       population count instructions, and conversions between floating
9353       point and unsigned types.
9354     * Support for the power7 processor is now available through the
9355       -mcpu=power7 and -mtune=power7.
9356     * GCC will now vectorize loops that contain simple math functions
9357       like copysign when generating code for altivec or VSX targets.
9358     * Support for the A2 processor is now available through the -mcpu=a2
9359       and -mtune=a2 options.
9360     * Support for the 476 processor is now available through the
9361       -mcpu={476,476fp} and -mtune={476,476fp} options.
9362     * Support for the e500mc64 processor is now available through the
9363       -mcpu=e500mc64 and -mtune=e500mc64 options.
9364     * GCC can now be configured with options --with-cpu-32,
9365       --with-cpu-64, --with-tune-32 and --with-tune-64 to control the
9366       default optimization separately for 32-bit and 64-bit modes.
9367     * Starting with GCC 4.5.4, vectors of type vector long long or vector
9368       long are passed and returned in the same method as other vectors
9369       with the VSX instruction set. Previously the GCC compiler did not
9370       adhere to the ABI for 128-bit vectors with 64-bit integer base
9371       types (PR 48857). This is also fixed in the GCC 4.6.1 release.
9372
9373  RX
9374
9375   Support has been added for the Renesas RX Processor (rx-elf) target.
9376
9377Operating Systems
9378
9379  Windows (Cygwin and MinGW)
9380
9381     * GCC now installs all the major language runtime libraries as DLLs
9382       when configured with the --enable-shared option.
9383     * GCC now makes use of the new support for aligned common variables
9384       in versions of binutils >= 2.20 to fix bugs in the support for SSE
9385       data types.
9386     * Improvements to the libffi support library increase the reliability
9387       of code generated by GCJ on all Windows platforms. Libgcj is
9388       enabled by default for the first time.
9389     * Libtool improvements simplify installation by placing the generated
9390       DLLs in the correct binaries directory.
9391     * Numerous other minor bugfixes and improvements, and substantial
9392       enhancements to the Fortran language support library.
9393
9394   >
9395
9396Other significant improvements
9397
9398  Plugins
9399
9400     * It is now possible to extend the compiler without having to modify
9401       its source code. A new option -fplugin=file.so tells GCC to load
9402       the shared object file.so and execute it as part of the compiler.
9403       The internal documentation describes the details on how plugins can
9404       interact with the compiler.
9405
9406  Installation changes
9407
9408     * The move to newer autotools changed default installation
9409       directories and switches to control them: The --with-datarootdir,
9410       --with-docdir, --with-pdfdir, and --with-htmldir switches are not
9411       used any more. Instead, you can now use --datarootdir, --docdir,
9412       --htmldir, and --pdfdir. The default installation directories have
9413       changed as follows according to the GNU Coding Standards:
9414
9415       datarootdir read-only architecture-independent data root [PREFIX/share]
9416       localedir   locale-specific message catalogs [DATAROOTDIR/locale]
9417       docdir      documentation root [DATAROOTDIR/doc/PACKAGE]
9418       htmldir     html documentation [DOCDIR]
9419       dvidir      dvi documentation [DOCDIR]
9420       pdfdir      pdf documentation [DOCDIR]
9421       psdir       ps documentation [DOCDIR]
9422       The following variables have new default values:
9423
9424       datadir read-only architecture-independent data [DATAROOTDIR]
9425       infodir info documentation [DATAROOTDIR/info]
9426       mandir  man documentation [DATAROOTDIR/man]
9427
9428GCC 4.5.1
9429
9430   This is the [22]list of problem reports (PRs) from GCC's bug tracking
9431   system that are known to be fixed in the 4.5.1 release. This list might
9432   not be complete (that is, it is possible that some PRs that have been
9433   fixed are not listed here).
9434
9435  All languages
9436
9437     * GCC's new link-time optimizer ([23]-flto) now also works on a few
9438       non-ELF targets:
9439          + Cygwin (*-cygwin*)
9440          + MinGW (*-mingw*)
9441          + Darwin on x86-64 (x86_64-apple-darwin*)
9442       LTO is not enabled by default for these targets. To enable LTO, you
9443       should configure with the --enable-lto option.
9444
9445GCC 4.5.2
9446
9447   This is the [24]list of problem reports (PRs) from GCC's bug tracking
9448   system that are known to be fixed in the 4.5.2 release. This list might
9449   not be complete (that is, it is possible that some PRs that have been
9450   fixed are not listed here).
9451
9452GCC 4.5.3
9453
9454   This is the [25]list of problem reports (PRs) from GCC's bug tracking
9455   system that are known to be fixed in the 4.5.3 release. This list might
9456   not be complete (that is, it is possible that some PRs that have been
9457   fixed are not listed here).
9458
9459   On the PowerPC compiler, the Altivec builtin functions vec_ld and
9460   vec_st have been modified to generate the Altivec memory instructions
9461   LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5
9462   release, these builtin functions were changed to generate VSX memory
9463   reference instructions instead of Altivec memory instructions, but
9464   there are differences between the two instructions. If the VSX
9465   instruction set is available, you can now use the new builtin functions
9466   vec_vsx_ld and vec_vsx_st which always generates the VSX memory
9467   instructions.
9468
9469GCC 4.5.4
9470
9471   This is the [26]list of problem reports (PRs) from GCC's bug tracking
9472   system that are known to be fixed in the 4.5.4 release. This list might
9473   not be complete (that is, it is possible that some PRs that have been
9474   fixed are not listed here).
9475
9476
9477    For questions related to the use of GCC, please consult these web
9478    pages and the [27]GCC manuals. If that fails, the
9479    [28]gcc-help@gcc.gnu.org mailing list might help. Comments on these
9480    web pages and the development of GCC are welcome on our developer
9481    list at [29]gcc@gcc.gnu.org. All of [30]our lists have public
9482    archives.
9483
9484   Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and
9485   distribution of this entire article is permitted in any medium,
9486   provided this notice is preserved.
9487
9488   These pages are [32]maintained by the GCC team. Last modified
9489   2019-11-28[33].
9490
9491References
9492
9493   1. http://www.multiprecision.org/mpc/
9494   2. https://gcc.gnu.org/install/prerequisites.html
9495   3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html
9496   4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted
9497   5. http://gcc.gnu.org/gcc-4.5/changes.html#x86
9498   6. https://gcc.gnu.org/PR30789
9499   7. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
9500   8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802
9501   9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800
9502  10. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html
9503  11. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html
9504  12. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757
9505  13. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176
9506  14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html
9507  15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733
9508  16. https://sourceware.org/gdb/wiki/STLSupport
9509  17. http://gcc.gnu.org/gcc-4.4/changes.html
9510  18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
9511  19. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html
9512  20. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html
9513  21. https://gcc.gnu.org/wiki/OOP
9514  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1
9515  23. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
9516  24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2
9517  25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3
9518  26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4
9519  27. https://gcc.gnu.org/onlinedocs/
9520  28. mailto:gcc-help@gcc.gnu.org
9521  29. mailto:gcc@gcc.gnu.org
9522  30. https://gcc.gnu.org/lists.html
9523  31. https://www.fsf.org/
9524  32. https://gcc.gnu.org/about.html
9525  33. http://validator.w3.org/check/referer
9526======================================================================
9527http://gcc.gnu.org/gcc-4.4/index.html
9528                             GCC 4.4 Release Series
9529
9530   This release series is no longer maintained.
9531
9532   March 13, 2012
9533
9534   The [1]GNU project and the GCC developers are pleased to announce the
9535   release of GCC 4.4.7.
9536
9537   This release is a bug-fix release, containing fixes for regressions in
9538   GCC 4.4.6 relative to previous releases of GCC.
9539
9540Release History
9541
9542   GCC 4.4.7
9543          March 13, 2012 ([2]changes)
9544
9545   GCC 4.4.6
9546          April 16, 2011 ([3]changes)
9547
9548   GCC 4.4.5
9549          October 1, 2010 ([4]changes)
9550
9551   GCC 4.4.4
9552          April 29, 2010 ([5]changes)
9553
9554   GCC 4.4.3
9555          January 21, 2010 ([6]changes)
9556
9557   GCC 4.4.2
9558          October 15, 2009 ([7]changes)
9559
9560   GCC 4.4.1
9561          July 22, 2009 ([8]changes)
9562
9563   GCC 4.4.0
9564          April 21, 2009 ([9]changes)
9565
9566References and Acknowledgements
9567
9568   GCC used to stand for the GNU C Compiler, but since the compiler
9569   supports several other languages aside from C, it now stands for the
9570   GNU Compiler Collection.
9571
9572   A list of [10]successful builds is updated as new information becomes
9573   available.
9574
9575   The GCC developers would like to thank the numerous people that have
9576   contributed new features, improvements, bug fixes, and other changes as
9577   well as test results to GCC. This [11]amazing group of volunteers is
9578   what makes GCC successful.
9579
9580   For additional information about GCC please refer to the [12]GCC
9581   project web site or contact the [13]GCC development mailing list.
9582
9583   To obtain GCC please use [14]our mirror sites or [15]our version
9584   control system.
9585
9586
9587    For questions related to the use of GCC, please consult these web
9588    pages and the [16]GCC manuals. If that fails, the
9589    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
9590    web pages and the development of GCC are welcome on our developer
9591    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
9592    archives.
9593
9594   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
9595   distribution of this entire article is permitted in any medium,
9596   provided this notice is preserved.
9597
9598   These pages are [21]maintained by the GCC team. Last modified
9599   2020-01-14[22].
9600
9601References
9602
9603   1. http://www.gnu.org/
9604   2. http://gcc.gnu.org/gcc-4.4/changes.html
9605   3. http://gcc.gnu.org/gcc-4.4/changes.html
9606   4. http://gcc.gnu.org/gcc-4.4/changes.html
9607   5. http://gcc.gnu.org/gcc-4.4/changes.html
9608   6. http://gcc.gnu.org/gcc-4.4/changes.html
9609   7. http://gcc.gnu.org/gcc-4.4/changes.html
9610   8. http://gcc.gnu.org/gcc-4.4/changes.html
9611   9. http://gcc.gnu.org/gcc-4.4/changes.html
9612  10. http://gcc.gnu.org/gcc-4.4/buildstat.html
9613  11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
9614  12. http://gcc.gnu.org/index.html
9615  13. mailto:gcc@gcc.gnu.org
9616  14. http://gcc.gnu.org/mirrors.html
9617  15. http://gcc.gnu.org/git.html
9618  16. https://gcc.gnu.org/onlinedocs/
9619  17. mailto:gcc-help@gcc.gnu.org
9620  18. mailto:gcc@gcc.gnu.org
9621  19. https://gcc.gnu.org/lists.html
9622  20. https://www.fsf.org/
9623  21. https://gcc.gnu.org/about.html
9624  22. http://validator.w3.org/check/referer
9625======================================================================
9626http://gcc.gnu.org/gcc-4.4/changes.html
9627                             GCC 4.4 Release Series
9628                        Changes, New Features, and Fixes
9629
9630   The latest release in the 4.4 release series is [1]GCC 4.4.7.
9631
9632Caveats
9633
9634     * __builtin_stdarg_start has been completely removed from GCC.
9635       Support for <varargs.h> had been deprecated since GCC 4.0. Use
9636       __builtin_va_start as a replacement.
9637     * Some of the errors issued by the C++ front end that could be
9638       downgraded to warnings in previous releases by using -fpermissive
9639       are now warnings by default. They can be converted into errors by
9640       using -pedantic-errors.
9641     * Use of the cpp assertion extension will now emit a warning when
9642       -Wdeprecated or -pedantic is used. This extension has been
9643       deprecated for many years, but never warned about.
9644     * Packed bit-fields of type char were not properly bit-packed on many
9645       targets prior to GCC 4.4. On these targets, the fix in GCC 4.4
9646       causes an ABI change. For example there is no longer a 4-bit
9647       padding between field a and b in this structure:
9648    struct foo
9649    {
9650      char a:4;
9651      char b:8;
9652    } __attribute__ ((packed));
9653       There is a new warning to help identify fields that are affected:
9654    foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4
9655       The warning can be disabled with -Wno-packed-bitfield-compat.
9656     * On ARM EABI targets, the C++ mangling of the va_list type has been
9657       changed to conform to the current revision of the EABI. This does
9658       not affect the libstdc++ library included with GCC.
9659     * The SCOUNT and POS bits of the MIPS DSP control register are now
9660       treated as global. Previous versions of GCC treated these fields as
9661       call-clobbered instead.
9662     * The MIPS port no longer recognizes the h asm constraint. It was
9663       necessary to remove this constraint in order to avoid generating
9664       unpredictable code sequences.
9665       One of the main uses of the h constraint was to extract the high
9666       part of a multiplication on 64-bit targets. For example:
9667    asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y));
9668       You can now achieve the same effect using 128-bit types:
9669    typedef unsigned int uint128_t __attribute__((mode(TI)));
9670    result = ((uint128_t) x * y) >> 64;
9671       The second sequence is better in many ways. For example, if x and y
9672       are constants, the compiler can perform the multiplication at
9673       compile time. If x and y are not constants, the compiler can
9674       schedule the runtime multiplication better than it can schedule an
9675       asm statement.
9676     * Support for a number of older systems and recently unmaintained or
9677       untested target ports of GCC has been declared obsolete in GCC 4.4.
9678       Unless there is activity to revive them, the next release of GCC
9679       will have their sources permanently removed.
9680       The following ports for individual systems on particular
9681       architectures have been obsoleted:
9682          + Generic a.out on IA32 and m68k (i[34567]86-*-aout*,
9683            m68k-*-aout*)
9684          + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*,
9685            armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*,
9686            sh-*-*). This does not affect other more specific targets
9687            using the COFF object format on those architectures, or the
9688            more specific H8300 and SH targets (h8300-*-rtems*,
9689            h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*,
9690            sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks).
9691          + 2BSD on PDP-11 (pdp11-*-bsd)
9692          + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*,
9693            powerpc-ibm-aix4.[12]*)
9694          + Tuning support for Itanium1 (Merced) variants. Note that code
9695            tuned for Itanium2 should also run correctly on Itanium1.
9696     * The protoize and unprotoize utilities have been obsoleted and will
9697       be removed in GCC 4.5. These utilities have not been installed by
9698       default since GCC 3.0.
9699     * Support has been removed for all the [2]configurations obsoleted in
9700       GCC 4.3.
9701     * Unknown -Wno-* options are now silently ignored by GCC if no other
9702       diagnostics are issued. If other diagnostics are issued, then GCC
9703       warns about the unknown options.
9704     * More information on porting to GCC 4.4 from previous versions of
9705       GCC can be found in the [3]porting guide for this release.
9706
9707General Optimizer Improvements
9708
9709     * A new command-line switch -findirect-inlining has been added. When
9710       turned on it allows the inliner to also inline indirect calls that
9711       are discovered to have known targets at compile time thanks to
9712       previous inlining.
9713     * A new command-line switch -ftree-switch-conversion has been added.
9714       This new pass turns simple initializations of scalar variables in
9715       switch statements into initializations from a static array, given
9716       that all the values are known at compile time and the ratio between
9717       the new array size and the original switch branches does not exceed
9718       the parameter --param switch-conversion-max-branch-ratio (default
9719       is eight).
9720     * A new command-line switch -ftree-builtin-call-dce has been added.
9721       This optimization eliminates unnecessary calls to certain builtin
9722       functions when the return value is not used, in cases where the
9723       calls can not be eliminated entirely because the function may set
9724       errno. This optimization is on by default at -O2 and above.
9725     * A new command-line switch -fconserve-stack directs the compiler to
9726       minimize stack usage even if it makes the generated code slower.
9727       This affects inlining decisions.
9728     * When the assembler supports it, the compiler will now emit unwind
9729       information using assembler .cfi directives. This makes it possible
9730       to use such directives in inline assembler code. The new option
9731       -fno-dwarf2-cfi-asm directs the compiler to not use .cfi
9732       directives.
9733     * The [4]Graphite branch has been merged. This merge has brought in a
9734       new framework for loop optimizations based on a polyhedral
9735       intermediate representation. These optimizations apply to all the
9736       languages supported by GCC. The following new code transformations
9737       are available in GCC 4.4:
9738          + -floop-interchange performs loop interchange transformations
9739            on loops. Interchanging two nested loops switches the inner
9740            and outer loops. For example, given a loop like:
9741          DO J = 1, M
9742            DO I = 1, N
9743              A(J, I) = A(J, I) * C
9744            ENDDO
9745          ENDDO
9746
9747            loop interchange will transform the loop as if the user had
9748            written:
9749          DO I = 1, N
9750            DO J = 1, M
9751              A(J, I) = A(J, I) * C
9752            ENDDO
9753          ENDDO
9754
9755            which can be beneficial when N is larger than the caches,
9756            because in Fortran, the elements of an array are stored in
9757            memory contiguously by column, and the original loop iterates
9758            over rows, potentially creating at each access a cache miss.
9759          + -floop-strip-mine performs loop strip mining transformations
9760            on loops. Strip mining splits a loop into two nested loops.
9761            The outer loop has strides equal to the strip size and the
9762            inner loop has strides of the original loop within a strip.
9763            For example, given a loop like:
9764          DO I = 1, N
9765            A(I) = A(I) + C
9766          ENDDO
9767
9768            loop strip mining will transform the loop as if the user had
9769            written:
9770          DO II = 1, N, 4
9771            DO I = II, min (II + 3, N)
9772              A(I) = A(I) + C
9773            ENDDO
9774          ENDDO
9775
9776          + -floop-block performs loop blocking transformations on loops.
9777            Blocking strip mines each loop in the loop nest such that the
9778            memory accesses of the element loops fit inside caches. For
9779            example, given a loop like:
9780          DO I = 1, N
9781            DO J = 1, M
9782              A(J, I) = B(I) + C(J)
9783            ENDDO
9784          ENDDO
9785
9786            loop blocking will transform the loop as if the user had
9787            written:
9788          DO II = 1, N, 64
9789            DO JJ = 1, M, 64
9790              DO I = II, min (II + 63, N)
9791                DO J = JJ, min (JJ + 63, M)
9792                  A(J, I) = B(I) + C(J)
9793                ENDDO
9794              ENDDO
9795            ENDDO
9796          ENDDO
9797
9798            which can be beneficial when M is larger than the caches,
9799            because the innermost loop will iterate over a smaller amount
9800            of data that can be kept in the caches.
9801     * A new register allocator has replaced the old one. It is called
9802       integrated register allocator (IRA) because coalescing, register
9803       live range splitting, and hard register preferencing are done
9804       on-the-fly during coloring. It also has better integration with the
9805       reload pass. IRA is a regional register allocator which uses modern
9806       Chaitin-Briggs coloring instead of Chow's priority coloring used in
9807       the old register allocator. More info about IRA internals and
9808       options can be found in the GCC manuals.
9809     * A new instruction scheduler and software pipeliner, based on the
9810       selective scheduling approach, has been added. The new pass
9811       performs instruction unification, register renaming, substitution
9812       through register copies, and speculation during scheduling. The
9813       software pipeliner is able to pipeline non-countable loops. The new
9814       pass is targeted at scheduling-eager in-order platforms. In GCC 4.4
9815       it is available for the Intel Itanium platform working by default
9816       as the second scheduling pass (after register allocation) at the
9817       -O3 optimization level.
9818     * When using -fprofile-generate with a multi-threaded program, the
9819       profile counts may be slightly wrong due to race conditions. The
9820       new -fprofile-correction option directs the compiler to apply
9821       heuristics to smooth out the inconsistencies. By default the
9822       compiler will give an error message when it finds an inconsistent
9823       profile.
9824     * The new -fprofile-dir=PATH option permits setting the directory
9825       where profile data files are stored when using -fprofile-generate
9826       and friends, and the directory used when reading profile data files
9827       using -fprofile-use and friends.
9828
9829New warning options
9830
9831     * The new -Wframe-larger-than=NUMBER option directs GCC to emit a
9832       warning if any stack frame is larger than NUMBER bytes. This may be
9833       used to help ensure that code fits within a limited amount of stack
9834       space.
9835     * The command-line option -Wlarger-than-N is now written as
9836       -Wlarger-than=N and the old form is deprecated.
9837     * The new -Wno-mudflap option disables warnings about constructs
9838       which can not be instrumented when using -fmudflap.
9839
9840New Languages and Language specific improvements
9841
9842     * Version 3.0 of the OpenMP specification is now supported for the C,
9843       C++, and Fortran compilers.
9844     * New character data types, per [5]TR 19769: New character types in
9845       C, are now supported for the C compiler in -std=gnu99 mode, as
9846       __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in
9847       -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too.
9848
9849  C family
9850
9851     * A new optimize attribute was added to allow programmers to change
9852       the optimization level and particular optimization options for an
9853       individual function. You can also change the optimization options
9854       via the GCC optimize pragma for functions defined after the pragma.
9855       The GCC push_options pragma and the GCC pop_options pragma allow
9856       you temporarily save and restore the options used. The GCC
9857       reset_options pragma restores the options to what was specified on
9858       the command line.
9859     * Uninitialized warnings do not require enabling optimization
9860       anymore, that is, -Wuninitialized can be used together with -O0.
9861       Nonetheless, the warnings given by -Wuninitialized will probably be
9862       more accurate if optimization is enabled.
9863     * -Wparentheses now warns about expressions such as (!x | y) and (!x
9864       & y). Using explicit parentheses, such as in ((!x) | y), silences
9865       this warning.
9866     * -Wsequence-point now warns within if, while,do while and for
9867       conditions, and within for begin/end expressions.
9868     * A new option -dU is available to dump definitions of preprocessor
9869       macros that are tested or expanded.
9870
9871  C++
9872
9873     * [6]Improved experimental support for the upcoming ISO C++ standard,
9874       C++0x. Including support for auto, inline namespaces, generalized
9875       initializer lists, defaulted and deleted functions, new character
9876       types, and scoped enums.
9877     * Those errors that may be downgraded to warnings to build legacy
9878       code now mention -fpermissive when -fdiagnostics-show-option is
9879       enabled.
9880     * -Wconversion now warns if the result of a static_cast to enumeral
9881       type is unspecified because the value is outside the range of the
9882       enumeral type.
9883     * -Wuninitialized now warns if a non-static reference or non-static
9884       const member appears in a class without constructors.
9885     * G++ now properly implements value-initialization, so objects with
9886       an initializer of () and an implicitly defined default constructor
9887       will be zero-initialized before the default constructor is called.
9888
9889    Runtime Library (libstdc++)
9890
9891     * Improved experimental support for the upcoming ISO C++ standard,
9892       C++0x, including:
9893          + Support for <chrono>, <condition_variable>, <cstdatomic>,
9894            <forward_list>, <initializer_list>, <mutex>, <ratio>,
9895            <system_error>, and <thread>.
9896          + unique_ptr, <algorithm> additions, exception propagation, and
9897            support for the new character types in <string> and <limits>.
9898          + Existing facilities now exploit initializer lists, defaulted
9899            and deleted functions, and the newly implemented core C++0x
9900            features.
9901          + Some standard containers are more efficient together with
9902            stateful allocators, i.e., no allocator is constructed on the
9903            fly at element construction time.
9904     * Experimental support for non-standard pointer types in containers.
9905     * The long standing libstdc++/30928 has been fixed for targets
9906       running glibc 2.10 or later.
9907     * As usual, many small and larger bug fixes, in particular quite a
9908       few corner cases in <locale>.
9909
9910  Fortran
9911
9912     * GNU Fortran now employs libcpp directly instead of using cc1 as an
9913       external preprocessor. The [7]-cpp option was added to allow manual
9914       invocation of the preprocessor without relying on filename
9915       extensions.
9916     * The [8]-Warray-temporaries option warns about array temporaries
9917       generated by the compiler, as an aid to optimization.
9918     * The [9]-fcheck-array-temporaries option has been added, printing a
9919       notification at run time, when an array temporary had to be created
9920       for an function argument. Contrary to -Warray-temporaries the
9921       warning is only printed if the array is noncontiguous.
9922     * Improved generation of DWARF debugging symbols
9923     * If using an intrinsic not part of the selected standard (via -std=
9924       and -fall-intrinsics) gfortran will now treat it as if this
9925       procedure were declared EXTERNAL and try to link to a user-supplied
9926       procedure. -Wintrinsics-std will warn whenever this happens. The
9927       now-useless option -Wnonstd-intrinsic was removed.
9928     * The flag -falign-commons has been added to control the alignment of
9929       variables in COMMON blocks, which is enabled by default in line
9930       with previous GCC version. Using -fno-align-commons one can force
9931       commons to be contiguous in memory as required by the Fortran
9932       standard, however, this slows down the memory access. The option
9933       -Walign-commons, which is enabled by default, warns when padding
9934       bytes were added for alignment. The proper solution is to sort the
9935       common objects by decreasing storage size, which avoids the
9936       alignment problems.
9937     * Fortran 2003 support has been extended:
9938          + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is
9939            now supported (except internal reads from/writes to wide
9940            strings). [10]-fbackslash now supports also \unnnn and
9941            \Unnnnnnnn to enter Unicode characters.
9942          + Asynchronous I/O (implemented as synchronous I/O) and the
9943            decimal=, size=, sign=, pad=, blank=, and delim= specifiers
9944            are now supported in I/O statements.
9945          + Support for Fortran 2003 structure constructors and for array
9946            constructor with typespec has been added.
9947          + Procedure Pointers (but not yet as component in derived types
9948            and as function results) are now supported.
9949          + Abstract types, type extension, and type-bound procedures
9950            (both PROCEDURE and GENERIC but not as operators). Note: As
9951            CLASS/polymorphyic types are not implemented, type-bound
9952            procedures with PASS accept as non-standard extension TYPE
9953            arguments.
9954     * Fortran 2008 support has been added:
9955          + The -std=f2008 option and support for the file extensions
9956            .f2008 and .F2008 has been added.
9957          + The g0 format descriptor is now supported.
9958          + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH,
9959            ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED
9960            are now available (some of them existed as GNU extension
9961            before). Note: The hyperbolic functions are not yet supporting
9962            complex arguments and the three- argument version of BESSEL_*N
9963            is not available.
9964          + The bit intrinsics LEADZ and TRAILZ have been added.
9965
9966  Java (GCJ)
9967
9968  Ada
9969
9970     * The Ada runtime now supports multilibs on many platforms including
9971       x86_64, SPARC and PowerPC. Their build is enabled by default.
9972
9973New Targets and Target Specific Improvements
9974
9975  ARM
9976
9977     * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and
9978       Cortex-R4F processors and has many other improvements to
9979       optimization for ARM processors.
9980     * GCC now supports the VFPv3 variant with 16 double-precision
9981       registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been
9982       renamed to -mfpu=vfpv3.
9983     * GCC now supports the -mfix-cortex-m3-ldrd option to work around an
9984       erratum on Cortex-M3 processors.
9985     * GCC now supports the __sync_* atomic operations for ARM EABI
9986       GNU/Linux.
9987     * The section anchors optimization is now enabled by default when
9988       optimizing for ARM.
9989     * GCC now uses a new EABI-compatible profiling interface for EABI
9990       targets. This requires a function __gnu_mcount_nc, which is
9991       provided by GNU libc versions 2.8 and later.
9992
9993  AVR
9994
9995     * The -mno-tablejump option has been deprecated because it has the
9996       same effect as the -fno-jump-tables option.
9997     * Added support for these new AVR devices:
9998          + ATA6289
9999          + ATtiny13A
10000          + ATtiny87
10001          + ATtiny167
10002          + ATtiny327
10003          + ATmega8C1
10004          + ATmega16C1
10005          + ATmega32C1
10006          + ATmega8M1
10007          + ATmega16M1
10008          + ATmega32M1
10009          + ATmega32U4
10010          + ATmega16HVB
10011          + ATmega4HVD
10012          + ATmega8HVD
10013          + ATmega64C1
10014          + ATmega64M1
10015          + ATmega16U4
10016          + ATmega32U6
10017          + ATmega128RFA1
10018          + AT90PWM81
10019          + AT90SCR100
10020          + M3000F
10021          + M3000S
10022          + M3001B
10023
10024  IA-32/x86-64
10025
10026     * Support for Intel AES built-in functions and code generation is
10027       available via -maes.
10028     * Support for Intel PCLMUL built-in function and code generation is
10029       available via -mpclmul.
10030     * Support for Intel AVX built-in functions and code generation is
10031       available via -mavx.
10032     * Automatically align the stack for local variables with alignment
10033       requirement.
10034     * GCC can now utilize the SVML library for vectorizing calls to a set
10035       of C99 functions if -mveclibabi=svml is specified and you link to
10036       an SVML ABI compatible library.
10037     * On x86-64, the ABI has been changed in the following cases to
10038       conform to the x86-64 ABI:
10039          + Passing/returning structures with flexible array member:
10040  struct foo
10041    {
10042      int i;
10043      int flex[];
10044    };
10045          + Passing/returning structures with complex float member:
10046  struct foo
10047    {
10048      int i;
10049      __complex__ float f;
10050    };
10051          + Passing/returning unions with long double member:
10052  union foo
10053    {
10054      int x;
10055      long double ld;
10056    };
10057       Code built with previous versions of GCC that uses any of these is
10058       not compatible with code built with GCC 4.4.0 or later.
10059     * A new target attribute was added to allow programmers to change the
10060       target options like -msse2 or -march=k8 for an individual function.
10061       You can also change the target options via the GCC target pragma
10062       for functions defined after the pragma.
10063     * GCC can now be configured with options --with-arch-32,
10064       --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and
10065       --with-tune-64 to control the default optimization separately for
10066       32-bit and 64-bit modes.
10067
10068  IA-32/IA64
10069
10070     * Support for __float128 (TFmode) IEEE quad type and corresponding
10071       TCmode IEEE complex quad type is available via the soft-fp library
10072       on IA-32/IA64 targets. This includes basic arithmetic operations
10073       (addition, subtraction, negation, multiplication and division) on
10074       __float128 real and TCmode complex values, the full set of IEEE
10075       comparisons between __float128 values, conversions to and from
10076       float, double and long double floating point types, as well as
10077       conversions to and from signed or unsigned integer, signed or
10078       unsigned long integer and signed or unsigned quad (TImode, IA64
10079       only) integer types. Additionally, all operations generate the full
10080       set of IEEE exceptions and support the full set of IEEE rounding
10081       modes.
10082
10083  M68K/ColdFire
10084
10085     * GCC now supports instruction scheduling for ColdFire V1, V3 and V4
10086       processors. (Scheduling support for ColdFire V2 processors was
10087       added in GCC 4.3.)
10088     * GCC now supports the -mxgot option to support programs requiring
10089       many GOT entries on ColdFire.
10090     * The m68k-*-linux-gnu target now builds multilibs by default.
10091
10092  MIPS
10093
10094     * MIPS Technologies have extended the original MIPS SVR4 ABI to
10095       include support for procedure linkage tables (PLTs) and copy
10096       relocations. These extensions allow GNU/Linux executables to use a
10097       significantly more efficient code model than the one defined by the
10098       original ABI.
10099       GCC support for this code model is available via a new command-line
10100       option, -mplt. There is also a new configure-time option,
10101       --with-mips-plt, to make -mplt the default.
10102       The new code model requires support from the assembler, the linker,
10103       and the runtime C library. This support is available in binutils
10104       2.19 and GLIBC 2.9.
10105     * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables
10106       and 32-bit GNU/Linux shared libraries. This feature requires GNU
10107       binutils 2.19 or above.
10108     * Support for RMI's XLR processor is now available through the
10109       -march=xlr and -mtune=xlr options.
10110     * 64-bit targets can now perform 128-bit multiplications inline,
10111       instead of relying on a libgcc function.
10112     * Native GNU/Linux toolchains now support -march=native and
10113       -mtune=native, which select the host processor.
10114     * GCC now supports the R10K, R12K, R14K and R16K processors. The
10115       canonical -march= and -mtune= names for these processors are
10116       r10000, r12000, r14000 and r16000 respectively.
10117     * GCC can now work around the side effects of speculative execution
10118       on R10K processors. Please see the documentation of the
10119       -mr10k-cache-barrier option for details.
10120     * Support for the MIPS64 Release 2 instruction set has been added.
10121       The option -march=mips64r2 enables generation of these
10122       instructions.
10123     * GCC now supports Cavium Networks' Octeon processor. This support is
10124       available through the -march=octeon and -mtune=octeon options.
10125     * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The
10126       canonical -march= and -mtune= names for these processors are
10127       loongson2e and loongson2f.
10128
10129  picochip
10130
10131   Picochip is a 16-bit processor. A typical picoChip contains over 250
10132   small cores, each with small amounts of memory. There are three
10133   processor variants (STAN, MEM and CTRL) with different instruction sets
10134   and memory configurations and they can be chosen using the -mae option.
10135
10136   This port is intended to be a "C" only port.
10137
10138  Power Architecture and PowerPC
10139
10140     * GCC now supports the e300c2, e300c3 and e500mc processors.
10141     * GCC now supports Xilinx processors with a single-precision FPU.
10142     * Decimal floating point is now supported for e500 processors.
10143
10144  S/390, zSeries and System z9/z10
10145
10146     * Support for the IBM System z10 EC/BC processor has been added. When
10147       using the -march=z10 option, the compiler will generate code making
10148       use of instructions provided by the General-Instruction-Extension
10149       Facility and the Execute-Extension Facility.
10150
10151  VxWorks
10152
10153     * GCC now supports the thread-local storage mechanism used on
10154       VxWorks.
10155
10156  Xtensa
10157
10158     * GCC now supports thread-local storage (TLS) for Xtensa processor
10159       configurations that include the Thread Pointer option. TLS also
10160       requires support from the assembler and linker; this support is
10161       provided in the GNU binutils beginning with version 2.19.
10162
10163Documentation improvements
10164
10165Other significant improvements
10166
10167GCC 4.4.1
10168
10169   This is the [11]list of problem reports (PRs) from GCC's bug tracking
10170   system that are known to be fixed in the 4.4.1 release. This list might
10171   not be complete (that is, it is possible that some PRs that have been
10172   fixed are not listed here).
10173
10174GCC 4.4.2
10175
10176   This is the [12]list of problem reports (PRs) from GCC's bug tracking
10177   system that are known to be fixed in the 4.4.2 release. This list might
10178   not be complete (that is, it is possible that some PRs that have been
10179   fixed are not listed here).
10180
10181GCC 4.4.3
10182
10183   This is the [13]list of problem reports (PRs) from GCC's bug tracking
10184   system that are known to be fixed in the 4.4.3 release. This list might
10185   not be complete (that is, it is possible that some PRs that have been
10186   fixed are not listed here).
10187
10188GCC 4.4.4
10189
10190   This is the [14]list of problem reports (PRs) from GCC's bug tracking
10191   system that are known to be fixed in the 4.4.4 release. This list might
10192   not be complete (that is, it is possible that some PRs that have been
10193   fixed are not listed here).
10194
10195GCC 4.4.5
10196
10197   This is the [15]list of problem reports (PRs) from GCC's bug tracking
10198   system that are known to be fixed in the 4.4.5 release. This list might
10199   not be complete (that is, it is possible that some PRs that have been
10200   fixed are not listed here).
10201
10202GCC 4.4.6
10203
10204   This is the [16]list of problem reports (PRs) from GCC's bug tracking
10205   system that are known to be fixed in the 4.4.6 release. This list might
10206   not be complete (that is, it is possible that some PRs that have been
10207   fixed are not listed here).
10208
10209GCC 4.4.7
10210
10211   This is the [17]list of problem reports (PRs) from GCC's bug tracking
10212   system that are known to be fixed in the 4.4.7 release. This list might
10213   not be complete (that is, it is possible that some PRs that have been
10214   fixed are not listed here).
10215
10216
10217    For questions related to the use of GCC, please consult these web
10218    pages and the [18]GCC manuals. If that fails, the
10219    [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
10220    web pages and the development of GCC are welcome on our developer
10221    list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
10222    archives.
10223
10224   Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
10225   distribution of this entire article is permitted in any medium,
10226   provided this notice is preserved.
10227
10228   These pages are [23]maintained by the GCC team. Last modified
10229   2019-11-28[24].
10230
10231References
10232
10233   1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7
10234   2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted
10235   3. http://gcc.gnu.org/gcc-4.4/porting_to.html
10236   4. https://gcc.gnu.org/wiki/Graphite
10237   5. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf
10238   6. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html
10239   7. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html
10240   8. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125
10241   9. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221
10242  10. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34
10243  11. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1
10244  12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2
10245  13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3
10246  14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4
10247  15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5
10248  16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6
10249  17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7
10250  18. https://gcc.gnu.org/onlinedocs/
10251  19. mailto:gcc-help@gcc.gnu.org
10252  20. mailto:gcc@gcc.gnu.org
10253  21. https://gcc.gnu.org/lists.html
10254  22. https://www.fsf.org/
10255  23. https://gcc.gnu.org/about.html
10256  24. http://validator.w3.org/check/referer
10257======================================================================
10258http://gcc.gnu.org/gcc-4.3/index.html
10259                             GCC 4.3 Release Series
10260
10261   (This release series is no longer supported.)
10262
10263   Jun 27, 2011
10264
10265   The [1]GNU project and the GCC developers are pleased to announce the
10266   release of GCC 4.3.6.
10267
10268   This release is a bug-fix release, containing fixes for regressions in
10269   GCC 4.3.5 relative to previous releases of GCC.
10270
10271Release History
10272
10273   GCC 4.3.6
10274          Jun 27, 2011 ([2]changes)
10275
10276   GCC 4.3.5
10277          May 22, 2010 ([3]changes)
10278
10279   GCC 4.3.4
10280          August 4, 2009 ([4]changes)
10281
10282   GCC 4.3.3
10283          January 24, 2009 ([5]changes)
10284
10285   GCC 4.3.2
10286          August 27, 2008 ([6]changes)
10287
10288   GCC 4.3.1
10289          June 6, 2008 ([7]changes)
10290
10291   GCC 4.3.0
10292          March 5, 2008 ([8]changes)
10293
10294References and Acknowledgements
10295
10296   GCC used to stand for the GNU C Compiler, but since the compiler
10297   supports several other languages aside from C, it now stands for the
10298   GNU Compiler Collection.
10299
10300   A list of [9]successful builds is updated as new information becomes
10301   available.
10302
10303   The GCC developers would like to thank the numerous people that have
10304   contributed new features, improvements, bug fixes, and other changes as
10305   well as test results to GCC. This [10]amazing group of volunteers is
10306   what makes GCC successful.
10307
10308   For additional information about GCC please refer to the [11]GCC
10309   project web site or contact the [12]GCC development mailing list.
10310
10311   To obtain GCC please use [13]our mirror sites or [14]our version
10312   control system.
10313
10314
10315    For questions related to the use of GCC, please consult these web
10316    pages and the [15]GCC manuals. If that fails, the
10317    [16]gcc-help@gcc.gnu.org mailing list might help. Comments on these
10318    web pages and the development of GCC are welcome on our developer
10319    list at [17]gcc@gcc.gnu.org. All of [18]our lists have public
10320    archives.
10321
10322   Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and
10323   distribution of this entire article is permitted in any medium,
10324   provided this notice is preserved.
10325
10326   These pages are [20]maintained by the GCC team. Last modified
10327   2020-01-14[21].
10328
10329References
10330
10331   1. http://www.gnu.org/
10332   2. http://gcc.gnu.org/gcc-4.3/changes.html
10333   3. http://gcc.gnu.org/gcc-4.3/changes.html
10334   4. http://gcc.gnu.org/gcc-4.3/changes.html
10335   5. http://gcc.gnu.org/gcc-4.3/changes.html
10336   6. http://gcc.gnu.org/gcc-4.3/changes.html
10337   7. http://gcc.gnu.org/gcc-4.3/changes.html
10338   8. http://gcc.gnu.org/gcc-4.3/changes.html
10339   9. http://gcc.gnu.org/gcc-4.3/buildstat.html
10340  10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
10341  11. http://gcc.gnu.org/index.html
10342  12. mailto:gcc@gcc.gnu.org
10343  13. http://gcc.gnu.org/mirrors.html
10344  14. http://gcc.gnu.org/git.html
10345  15. https://gcc.gnu.org/onlinedocs/
10346  16. mailto:gcc-help@gcc.gnu.org
10347  17. mailto:gcc@gcc.gnu.org
10348  18. https://gcc.gnu.org/lists.html
10349  19. https://www.fsf.org/
10350  20. https://gcc.gnu.org/about.html
10351  21. http://validator.w3.org/check/referer
10352======================================================================
10353http://gcc.gnu.org/gcc-4.3/changes.html
10354                             GCC 4.3 Release Series
10355                        Changes, New Features, and Fixes
10356
10357   The latest release in the 4.3 release series is [1]GCC 4.3.5.
10358
10359Caveats
10360
10361     * GCC requires the [2]GMP and [3]MPFR libraries for building all the
10362       various front-end languages it supports. See the [4]prerequisites
10363       page for version requirements.
10364     * ColdFire targets now treat long double as having the same format as
10365       double. In earlier versions of GCC, they used the 68881 long double
10366       format instead.
10367     * The m68k-uclinux target now uses the same calling conventions as
10368       m68k-linux-gnu. You can select the original calling conventions by
10369       configuring for m68k-uclinuxoldabi instead. Note that
10370       m68k-uclinuxoldabi also retains the original 80-bit long double on
10371       ColdFire targets.
10372     * The -fforce-mem option has been removed because it has had no
10373       effect in the last few GCC releases.
10374     * The i386 -msvr3-shlib option has been removed since it is no longer
10375       used.
10376     * Fastcall for i386 has been changed not to pass aggregate arguments
10377       in registers, following Microsoft compilers.
10378     * Support for the AOF assembler has been removed from the ARM back
10379       end; this affects only the targets arm-semi-aof and armel-semi-aof,
10380       which are no longer recognized. We removed these targets without a
10381       deprecation period because we discovered that they have been
10382       unusable since GCC 4.0.0.
10383     * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*)
10384       has been removed. This support had been deprecated since GCC 4.0.0.
10385     * Support for a number of older systems and recently unmaintained or
10386       untested target ports of GCC has been declared obsolete in GCC 4.3.
10387       Unless there is activity to revive them, the next release of GCC
10388       will have their sources permanently removed.
10389       All GCC ports for the following processor architectures have been
10390       declared obsolete:
10391          + Morpho MT (mt-*)
10392       The following aliases for processor architectures have been
10393       declared obsolete. Users should use the indicated generic target
10394       names instead, with compile-time options such as -mcpu or
10395       configure-time options such as --with-cpu to control the
10396       configuration more precisely.
10397          + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-*
10398            instead).
10399          + parisc*-*-* (use hppa*-*-* instead).
10400          + m680[012]0-*-* (use m68k-*-* instead).
10401       All GCC ports for the following operating systems have been
10402       declared obsolete:
10403          + BeOS (*-*-beos*)
10404          + kaOS (*-*-kaos*)
10405          + GNU/Linux using the a.out object format (*-*-linux*aout*)
10406          + GNU/Linux using version 1 of the GNU C Library
10407            (*-*-linux*libc1*)
10408          + Solaris versions before Solaris 7 (*-*-solaris2.[0-6],
10409            *-*-solaris2.[0-6].*)
10410          + Miscellaneous System V (*-*-sysv*)
10411          + WindISS (*-*-windiss*)
10412       Also, those for some individual systems on particular architectures
10413       have been obsoleted:
10414          + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*)
10415          + CRIS with a.out object format (cris-*-aout)
10416          + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*)
10417          + OSF/1 on PA-RISC (hppa1.1-*-osf*)
10418          + PRO on PA-RISC (hppa1.1-*-pro*)
10419          + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*,
10420            i[34567]86-sequent-sysv4*)
10421          + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*)
10422          + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host
10423            was previously [5]removed in 2001, leaving only the support
10424            for UWIN as a target now being deprecated)
10425          + ChorusOS on PowerPC (powerpc-*-chorusos*)
10426          + All VAX configurations apart from NetBSD and OpenBSD
10427            (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*)
10428     * The [6]-Wconversion option has been modified. Its purpose now is to
10429       warn for implicit conversions that may alter a value. This new
10430       behavior is available for both C and C++. Warnings about
10431       conversions between signed and unsigned integers can be disabled by
10432       using -Wno-sign-conversion. In C++, they are disabled by default
10433       unless -Wsign-conversion is explicitly requested. The old behavior
10434       of -Wconversion, that is, warn for prototypes causing a type
10435       conversion that is different from what would happen to the same
10436       argument in the absence of a prototype, has been moved to a new
10437       option -Wtraditional-conversion, which is only available for C.
10438     * The -m386, -m486, -mpentium and -mpentiumpro tuning options have
10439       been removed because they were deprecated for more than 3 GCC major
10440       releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or
10441       -mtune=pentiumpro as a replacement.
10442     * The -funsafe-math-optimizations option now automatically turns on
10443       -fno-trapping-math in addition to -fno-signed-zeros, as it enables
10444       reassociation and thus may introduce or remove traps.
10445     * The -ftree-vectorize option is now on by default under -O3. In
10446       order to generate code for a SIMD extension, it has to be enabled
10447       as well: use -maltivec for PowerPC platforms and -msse/-msse2 for
10448       i?86 and x86_64.
10449     * More information on porting to GCC 4.3 from previous versions of
10450       GCC can be found in the [7]porting guide for this release.
10451
10452General Optimizer Improvements
10453
10454     * The GCC middle-end has been integrated with the MPFR library. This
10455       allows GCC to evaluate and replace at compile-time calls to
10456       built-in math functions having constant arguments with their
10457       mathematically equivalent results. In making use of MPFR, GCC can
10458       generate correct results regardless of the math library
10459       implementation or floating point precision of the host platform.
10460       This also allows GCC to generate identical results regardless of
10461       whether one compiles in native or cross-compile configurations to a
10462       particular target. The following built-in functions take advantage
10463       of this new capability: acos, acosh, asin, asinh, atan2, atan,
10464       atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1,
10465       fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10,
10466       log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh,
10467       tan, tanh, tgamma, y0, y1 and yn. The float and long double
10468       variants of these functions (e.g. sinf and sinl) are also handled.
10469       The sqrt and cabs functions with constant arguments were already
10470       optimized in prior GCC releases. Now they also use MPFR.
10471     * A new forward propagation pass on RTL was added. The new pass
10472       replaces several slower transformations, resulting in compile-time
10473       improvements as well as better code generation in some cases.
10474     * A new command-line switch -frecord-gcc-switches has been added to
10475       GCC, although it is only enabled for some targets. The switch
10476       causes the command line that was used to invoke the compiler to be
10477       recorded into the object file that is being created. The exact
10478       format of this recording is target and binary file format
10479       dependent, but it usually takes the form of a note section
10480       containing ASCII text. The switch is related to the -fverbose-asm
10481       switch, but that one only records the information in the assembler
10482       output file as comments, so the information never reaches the
10483       object file.
10484     * The inliner heuristic is now aware of stack frame consumption. New
10485       command-line parameters --param large-stack-frame and --param
10486       large-stack-frame-growth can be used to limit stack frame size
10487       growth caused by inlining.
10488     * During feedback directed optimizations, the expected block size the
10489       memcpy, memset and bzero functions operate on is discovered and for
10490       cases of commonly used small sizes, specialized inline code is
10491       generated.
10492     * __builtin_expect no longer requires its argument to be a compile
10493       time constant.
10494     * Interprocedural optimization was reorganized to work on functions
10495       in SSA form. This enables more precise and cheaper dataflow
10496       analysis and makes writing interprocedural optimizations easier.
10497       The following improvements have been implemented on top of this
10498       framework:
10499          + Pre-inline optimization: Selected local optimization passes
10500            are run before the inliner (and other interprocedural passes)
10501            are executed. This significantly improves the accuracy of code
10502            growth estimates used by the inliner and reduces the overall
10503            memory footprint for large compilation units.
10504          + Early inlining (a simple bottom-up inliner pass inlining only
10505            functions whose body is smaller than the expected call
10506            overhead) is now executed with the early optimization passes,
10507            thus inlining already optimized function bodies into an
10508            unoptimized function that is subsequently optimized by early
10509            optimizers. This enables the compiler to quickly eliminate
10510            abstraction penalty in C++ programs.
10511          + Interprocedural constant propagation now operate on SSA form
10512            increasing accuracy of the analysis.
10513     * A new internal representation for GIMPLE statements has been
10514       contributed, resulting in compile-time memory savings.
10515     * The vectorizer was enhanced to support vectorization of outer
10516       loops, intra-iteration parallelism (loop-aware SLP), vectorization
10517       of strided accesses and loops with multiple data-types. Run-time
10518       dependency testing using loop versioning was added. The cost model,
10519       turned on by -fvect-cost-model, was developed.
10520
10521New Languages and Language specific improvements
10522
10523     * We have added new command-line options
10524       -finstrument-functions-exclude-function-list and
10525       -finstrument-functions-exclude-file-list. They provide more control
10526       over which functions are annotated by the -finstrument-functions
10527       option.
10528
10529  C family
10530
10531     * Implicit conversions between generic vector types are now only
10532       permitted when the two vectors in question have the same number of
10533       elements and compatible element types. (Note that the restriction
10534       involves compatible element types, not implicitly-convertible
10535       element types: thus, a vector type with element type int may not be
10536       implicitly converted to a vector type with element type unsigned
10537       int.) This restriction, which is in line with specifications for
10538       SIMD architectures such as AltiVec, may be relaxed using the flag
10539       -flax-vector-conversions. This flag is intended only as a
10540       compatibility measure and should not be used for new code.
10541     * -Warray-bounds has been added and is now enabled by default for
10542       -Wall . It produces warnings for array subscripts that can be
10543       determined at compile time to be always out of bounds.
10544       -Wno-array-bounds will disable the warning.
10545     * The constructor and destructor function attributes now accept
10546       optional priority arguments which control the order in which the
10547       constructor and destructor functions are run.
10548     * New [8]command-line options -Wtype-limits, -Wold-style-declaration,
10549       -Wmissing-parameter-type, -Wempty-body, -Wclobbered and
10550       -Wignored-qualifiers have been added for finer control of the
10551       diverse warnings enabled by -Wextra.
10552     * A new function attribute alloc_size has been added to mark up
10553       malloc style functions. For constant sized allocations this can be
10554       used to find out the size of the returned pointer using the
10555       __builtin_object_size() function for buffer overflow checking and
10556       similar. This supplements the already built-in malloc and calloc
10557       constant size handling.
10558     * Integer constants written in binary are now supported as a GCC
10559       extension. They consist of a prefix 0b or 0B, followed by a
10560       sequence of 0 and 1 digits.
10561     * A new predefined macro __COUNTER__ has been added. It expands to
10562       sequential integral values starting from 0. In conjunction with the
10563       ## operator, this provides a convenient means to generate unique
10564       identifiers.
10565     * A new command-line option -fdirectives-only has been added. It
10566       enables a special preprocessing mode which improves the performance
10567       of applications like distcc and ccache.
10568     * Fixed-point data types and operators have been added. They are
10569       based on Chapter 4 of the Embedded-C specification (n1169.pdf).
10570       Currently, only MIPS targets are supported.
10571     * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732,
10572       N1241, is now supported as a GCC extension to C for targets
10573       i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu,
10574       and x86_64-*-linux-gnu. The feature introduces new data types
10575       _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF,
10576       DD, and DL.
10577
10578  C++
10579
10580     * [9]Experimental support for the upcoming ISO C++ standard, C++0x.
10581     * -Wc++0x-compat has been added and is now enabled by default for
10582       -Wall. It produces warnings for constructs whose meaning differs
10583       between ISO C++ 1998 and C++0x.
10584     * The -Wparentheses option now works for C++ as it does for C. It
10585       warns if parentheses are omitted when operators with confusing
10586       precedence are nested. It also warns about ambiguous else
10587       statements. Since -Wparentheses is enabled by -Wall, this may cause
10588       additional warnings with existing C++ code which uses -Wall. These
10589       new warnings may be disabled by using -Wall -Wno-parentheses.
10590     * The -Wmissing-declarations now works for C++ as it does for C.
10591     * The -fvisibility-ms-compat flag was added, to make it easier to
10592       port larger projects using shared libraries from Microsoft's Visual
10593       Studio to ELF and Mach-O systems.
10594     * C++ attribute handling has been overhauled for template arguments
10595       (ie dependent types). In particular, __attribute__((aligned(T)));
10596       works for C++ types.
10597
10598    Runtime Library (libstdc++)
10599
10600     * [10]Experimental support for the upcoming ISO C++ standard, C++0x.
10601     * Support for TR1 mathematical special functions and regular
10602       expressions.
10603     * Default what implementations give more elaborate exception strings
10604       for bad_cast, bad_typeid, bad_exception, and bad_alloc.
10605     * Header dependencies have been streamlined, reducing unnecessary
10606       includes and pre-processed bloat.
10607     * Variadic template implementations of items in <tuple> and
10608       <functional>.
10609     * An experimental [11]parallel mode has been added. This is a
10610       parallel implementation of many C++ Standard library algorithms,
10611       like std::accumulate, std::for_each, std::transform, or std::sort,
10612       to give but four examples. These algorithms can be substituted for
10613       the normal (sequential) libstdc++ algorithms on a piecemeal basis,
10614       or all existing algorithms can be transformed via the
10615       -D_GLIBCXX_PARALLEL macro.
10616     * Debug mode versions of classes in <unordered_set> and
10617       <unordered_map>.
10618     * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are
10619       now <backward/hash_set> and <backward/hash_map>. This code:
10620    #include <ext/hash_set>
10621    __gnu_cxx::hash_set<int> s;
10622
10623       Can be transformed (in order of preference) to:
10624    #include <tr1/unordered_set>
10625    std::tr1::unordered_set<int> s;
10626
10627       or
10628    #include <backward/hash_set>
10629    __gnu_cxx::hash_set<int> s;
10630
10631       Similar transformations apply to __gnu_cxx::hash_map,
10632       __gnu_cxx::hash_multimap, __gnu_cxx::hash_set,
10633       __gnu_cxx::hash_multiset.
10634
10635  Fortran
10636
10637     * Due to the fact that the GMP and MPFR libraries are required for
10638       all languages, Fortran is no longer special in this regard and is
10639       available by default.
10640     * The [12]-fexternal-blas option has been added, which generates
10641       calls to BLAS routines for intrinsic matrix operations such as
10642       matmul rather than using the built-in algorithms.
10643     * Support to give a backtrace (compiler flag -fbacktrace or
10644       environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems
10645       only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a
10646       run-time error occured.
10647     * GNU Fortran now defines __GFORTRAN__ when it runs the C
10648       preprocessor (CPP).
10649     * The [13]-finit-local-zero, -finit-real, -finit-integer,
10650       -finit-character, and -finit-logical options have been added, which
10651       can be used to initialize local variables.
10652     * The intrinsic procedures [14]GAMMA and [15]LGAMMA have been added,
10653       which calculate the Gamma function and its logarithm. Use EXTERNAL
10654       gamma if you want to use your own gamma function.
10655     * GNU Fortran now regards the backslash character as literal (as
10656       required by the Fortran 2003 standard); using [16]-fbackslash GNU
10657       Fortran interprets backslashes as C-style escape characters.
10658     * The [17]interpretation of binary, octal and hexadecimal (BOZ)
10659       literal constants has been changed. Before they were always
10660       interpreted as integer; now they are bit-wise transferred as
10661       argument of INT, REAL, DBLE and CMPLX as required by the Fortran
10662       2003 standard, and for real and complex variables in DATA
10663       statements or when directly assigned to real and complex variables.
10664       Everywhere else and especially in expressions they are still
10665       regarded as integer constants.
10666     * Fortran 2003 support has been extended:
10667          + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE
10668          + Pointer intent
10669          + Intrinsic module ISO_ENV_FORTRAN
10670          + Interoperability with C (ISO C Bindings)
10671          + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER
10672            attribute)
10673          + Fortran 2003 BOZ
10674
10675  Java (GCJ)
10676
10677     * GCJ now uses the Eclipse Java compiler for its Java parsing needs.
10678       This enables the use of all 1.5 language features, and fixes most
10679       existing front end bugs.
10680     * libgcj now supports all 1.5 language features which require runtime
10681       support: foreach, enum, annotations, generics, and auto-boxing.
10682     * We've made many changes to the tools shipped with gcj.
10683          + The old jv-scan tool has been removed. This tool never really
10684            worked properly. There is no replacement.
10685          + gcjh has been rewritten. Some of its more obscure options no
10686            longer work, but are still recognized in an attempt at
10687            compatibility. gjavah is a new program with similar
10688            functionality but different command-line options.
10689          + grmic and grmiregistry have been rewritten. grmid has been
10690            added.
10691          + gjar replaces the old fastjar.
10692          + gjarsigner (used for signing jars), gkeytool (used for key
10693            management), gorbd (for CORBA), gserialver (computes
10694            serialization UIDs), and gtnameserv (also for CORBA) are now
10695            installed.
10696     * The ability to dump the contents of the java run time heap to a
10697       file for off-line analysis has been added. The heap dumps may be
10698       analyzed with the new gc-analyze tool. They may be generated on
10699       out-of-memory conditions or on demand and are controlled by the new
10700       run time class gnu.gcj.util.GCInfo.
10701     * java.util.TimeZone can now read files from /usr/share/zoneinfo to
10702       provide correct, updated, timezone information. This means that
10703       packagers no longer have to update libgcj when a time zone change
10704       is published.
10705
10706New Targets and Target Specific Improvements
10707
10708  IA-32/x86-64
10709
10710     * Tuning for Intel Core 2 processors is available via -mtune=core2
10711       and -march=core2.
10712     * Tuning for AMD Geode processors is available via -mtune=geode and
10713       -march=geode.
10714     * Code generation of block move (memcpy) and block set (memset) was
10715       rewritten. GCC can now pick the best algorithm (loop, unrolled
10716       loop, instruction with rep prefix or a library call) based on the
10717       size of the block being copied and the CPU being optimized for. A
10718       new option -minline-stringops-dynamically has been added. With this
10719       option string operations of unknown size are expanded such that
10720       small blocks are copied by in-line code, while for large blocks a
10721       library call is used. This results in faster code than
10722       -minline-all-stringops when the library implementation is capable
10723       of using cache hierarchy hints. The heuristic choosing the
10724       particular algorithm can be overwritten via -mstringop-strategy.
10725       Newly also memset of values different from 0 is inlined.
10726     * GCC no longer places the cld instruction before string operations.
10727       Both i386 and x86-64 ABI documents mandate the direction flag to be
10728       clear at the entry of a function. It is now invalid to set the flag
10729       in asm statement without reseting it afterward.
10730     * Support for SSSE3 built-in functions and code generation are
10731       available via -mssse3.
10732     * Support for SSE4.1 built-in functions and code generation are
10733       available via -msse4.1.
10734     * Support for SSE4.2 built-in functions and code generation are
10735       available via -msse4.2.
10736     * Both SSE4.1 and SSE4.2 support can be enabled via -msse4.
10737     * A new set of options -mpc32, -mpc64 and -mpc80 have been added to
10738       allow explicit control of x87 floating point precision.
10739     * Support for __float128 (TFmode) IEEE quad type and corresponding
10740       TCmode IEEE complex quad type is available via the soft-fp library
10741       on x86_64 targets. This includes basic arithmetic operations
10742       (addition, subtraction, negation, multiplication and division) on
10743       __float128 real and TCmode complex values, the full set of IEEE
10744       comparisons between __float128 values, conversions to and from
10745       float, double and long double floating point types, as well as
10746       conversions to and from signed or unsigned integer, signed or
10747       unsigned long integer and signed or unsigned quad (TImode) integer
10748       types. Additionally, all operations generate the full set of IEEE
10749       exceptions and support the full set of IEEE rounding modes.
10750     * GCC can now utilize the ACML library for vectorizing calls to a set
10751       of C99 functions on x86_64 if -mveclibabi=acml is specified and you
10752       link to an ACML ABI compatible library.
10753
10754  ARM
10755
10756     * Compiler and Library support for Thumb-2 and the ARMv7 architecture
10757       has been added.
10758
10759  CRIS
10760
10761    New features
10762
10763     * Compiler and Library support for the CRIS v32 architecture, as
10764       found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been
10765       added.
10766
10767    Configuration changes
10768
10769     * The cris-*-elf target now includes support for CRIS v32, including
10770       libraries, through the -march=v32 option.
10771     * A new crisv32-*-elf target defaults to generate code for CRIS v32.
10772     * A new crisv32-*-linux* target defaults to generate code for CRIS
10773       v32.
10774     * The cris-*-aout target has been obsoleted.
10775
10776    Improved support for built-in functions
10777
10778     * GCC can now use the lz and swapwbr instructions to implement the
10779       __builtin_clz, __builtin_ctz and __builtin_ffs family of functions.
10780     * __builtin_bswap32 is now implemented using the swapwb instruction,
10781       when available.
10782
10783  m68k and ColdFire
10784
10785    New features
10786
10787     * Support for several new ColdFire processors has been added. You can
10788       generate code for them using the new -mcpu option.
10789     * All targets now support ColdFire processors.
10790     * m68k-uclinux targets have improved support for C++ constructors and
10791       destructors, and for shared libraries.
10792     * It is now possible to set breakpoints on the first or last line of
10793       a function, even if there are no statements on that line.
10794
10795    Optimizations
10796
10797     * Support for sibling calls has been added.
10798     * More use is now made of the ColdFire mov3q instruction.
10799     * __builtin_clz is now implemented using the ff1 ColdFire
10800       instruction, when available.
10801     * GCC now honors the -m68010 option. 68010 code now uses clr rather
10802       than move to zero volatile memory.
10803     * 68020 targets and above can now use symbol(index.size*scale)
10804       addresses for indexed array accesses. Earlier compilers would
10805       always load the symbol into a base register first.
10806
10807    Configuration changes
10808
10809     * All m68k and ColdFire targets now allow the default processor to be
10810       set at configure time using --with-cpu.
10811     * A --with-arch configuration option has been added. This option
10812       allows you to restrict a target to ColdFire or non-ColdFire
10813       processors.
10814
10815    Preprocessor macros
10816
10817     * An __mcfv*__ macro is now defined for all ColdFire targets.
10818       (Earlier versions of GCC only defined __mcfv4e__.)
10819     * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added.
10820     * All targets now define __mc68010 and __mc68010__ when generating
10821       68010 code.
10822
10823    Command-line changes
10824
10825     * New command-line options -march, -mcpu, -mtune and -mhard-float
10826       have been added. These options apply to both m68k and ColdFire
10827       targets.
10828     * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative
10829       versions of -mshort, etc.
10830     * -fforce-addr has been removed. It is now ignored by the compiler.
10831
10832    Other improvements
10833
10834     * ColdFire targets now try to maintain a 4-byte-aligned stack where
10835       possible.
10836     * m68k-uclinux targets now try to avoid situations that lead to the
10837       load-time error: BINFMT_FLAT: reloc outside program.
10838
10839  MIPS
10840
10841    Changes to existing configurations
10842
10843     * libffi and libjava now support all three GNU/Linux ABIs: o32, n32
10844       and n64. Every GNU/Linux configuration now builds these libraries
10845       by default.
10846     * GNU/Linux configurations now generate -mno-shared code unless
10847       overridden by -fpic, -fPIC, -fpie or -fPIE.
10848     * mipsisa32*-linux-gnu configurations now generate hard-float code by
10849       default, just like other mipsisa32* and mips*-linux-gnu
10850       configurations. You can build a soft-float version of any
10851       mips*-linux-gnu configuration by passing --with-float=soft to
10852       configure.
10853     * mips-wrs-vxworks now supports run-time processes (RTPs).
10854
10855    Changes to existing command-line options
10856
10857     * The -march and -mtune options no longer accept 24k as a processor
10858       name. Please use 24kc, 24kf2_1 or 24kf1_1 instead.
10859     * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and
10860       34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The
10861       options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for
10862       24kx, 24kex and 34kx.
10863
10864    New configurations
10865
10866   GCC now supports the following configurations:
10867     * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by
10868       default. Earlier releases also recognized this configuration, but
10869       they treated it in the same way as mipsisa32*-linux-gnu*. Note that
10870       you can customize any mips*-linux-gnu* configuration to a
10871       particular ISA or processor by passing an appropriate --with-arch
10872       option to configure.
10873     * mipsisa*-sde-elf*, which provides compatibility with MIPS
10874       Technologies' SDE toolchains. The configuration uses the SDE
10875       libraries by default, but you can use it like other newlib-based
10876       ELF configurations by passing --with-newlib to configure. It is the
10877       only configuration besides mips64vr*-elf* to build MIPS16 as well
10878       as non-MIPS16 libraries.
10879     * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf*
10880       configuration, but uses the o32 and o64 ABIs instead of the 32-bit
10881       and 64-bit forms of the EABI.
10882
10883    New processors and application-specific extensions
10884
10885     * Support for the SmartMIPS ASE is available through the new
10886       -msmartmips option.
10887     * Support for revision 2 of the DSP ASE is available through the new
10888       -mdspr2 option. A new preprocessor macro called __mips_dsp_rev
10889       indicates the revision of the ASE in use.
10890     * Support for the 4KS and 74K families of processors is available
10891       through the -march and -mtune options.
10892
10893    Improved support for built-in functions
10894
10895     * GCC can now use load-linked, store-conditional and sync
10896       instructions to implement atomic built-in functions such as
10897       __sync_fetch_and_add. The memory reference must be 4 bytes wide for
10898       32-bit targets and either 4 or 8 bytes wide for 64-bit targets.
10899     * GCC can now use the clz and dclz instructions to implement the
10900       __builtin_ctz and __builtin_ffs families of functions.
10901     * There is a new __builtin___clear_cache function for flushing the
10902       instruction cache. GCC expands this function inline on MIPS32
10903       revision 2 targets, otherwise it calls the function specified by
10904       -mcache-flush-func.
10905
10906    MIPS16 improvements
10907
10908     * GCC can now compile objects that contain a mixture of MIPS16 and
10909       non-MIPS16 code. There are two new attributes, mips16 and nomips16,
10910       for specifying which mode a function should use.
10911     * A new option called -minterlink-mips16 makes non-MIPS16 code
10912       link-compatible with MIPS16 code.
10913     * After many bug fixes, the long-standing MIPS16 -mhard-float support
10914       should now work fairly reliably.
10915     * GCC can now use the MIPS16e save and restore instructions.
10916     * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled
10917       with -G0 -fsection-anchors is often smaller than code compiled with
10918       -G8. However, please note that you must usually compile all objects
10919       in your application with the same -G option; see the documentation
10920       of -G for details.
10921     * A new option called-mcode-readable specifies which instructions are
10922       allowed to load from the code segment. -mcode-readable=yes is the
10923       default and says that any instruction may load from the code
10924       segment. The other alternatives are -mcode-readable=pcrel, which
10925       says that only PC-relative MIPS16 instructions may load from the
10926       code segment, and -mcode-readable=no, which says that no
10927       instruction may do so. Please see the documentation for more
10928       details, including example uses.
10929
10930    Small-data improvements
10931
10932   There are three new options for controlling small data:
10933     * -mno-extern-sdata, which disables small-data accesses for
10934       externally-defined variables. Code compiled with -Gn
10935       -mno-extern-sdata will be link-compatible with any -G setting
10936       between -G0 and -Gn inclusive.
10937     * -mno-local-sdata, which disables the use of small-data sections for
10938       data that is not externally visible. This option can be a useful
10939       way of reducing small-data usage in less performance-critical parts
10940       of an application.
10941     * -mno-gpopt, which disables the use of the $gp register while still
10942       honoring the -G limit when placing externally-visible data. This
10943       option implies -mno-extern-sdata and -mno-local-sdata and it can be
10944       useful in situations where $gp does not necessarily hold the
10945       expected value.
10946
10947    Miscellaneous improvements
10948
10949     * There is a new option called -mbranch-cost for tweaking the
10950       perceived cost of branches.
10951     * If GCC is configured to use a version of GAS that supports the
10952       .gnu_attribute directive, it will use that directive to record
10953       certain properties of the output code. .gnu_attribute is new to GAS
10954       2.18.
10955     * There are two new function attributes, near and far, for overriding
10956       the command-line setting of -mlong-calls on a function-by-function
10957       basis.
10958     * -mfp64, which previously required a 64-bit target, now works with
10959       MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and
10960       mipsisa*-sde-elf* configurations provide suitable library support.
10961     * GCC now recognizes the -mdmx and -mmt options and passes them down
10962       to the assembler. It does nothing else with the options at present.
10963
10964  SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture
10965  (BEA)
10966
10967     * Support has been added for this new architecture.
10968
10969  RS6000 (POWER/PowerPC)
10970
10971     * Support for the PowerPC 750CL paired-single instructions has been
10972       added with a new powerpc-*-linux*paired* target configuration. It
10973       is enabled by an associated -mpaired option and can be accessed
10974       using new built-in functions.
10975     * Support for auto-detecting architecture and system configuration to
10976       auto-select processor optimization tuning.
10977     * Support for VMX on AIX 5.3 has been added.
10978     * Support for AIX Version 6.1 has been added.
10979
10980  S/390, zSeries and System z9
10981
10982     * Support for the IBM System z9 EC/BC processor (z9 GA3) has been
10983       added. When using the -march=z9-ec option, the compiler will
10984       generate code making use of instructions provided by the decimal
10985       floating point facility and the floating point conversion facility
10986       (pfpo). Besides the instructions used to implement decimal floating
10987       point operations these facilities also contain instructions to move
10988       between general purpose and floating point registers and to modify
10989       and copy the sign-bit of floating point values.
10990     * When the -march=z9-ec option is used the new
10991       -mhard-dfp/-mno-hard-dfp options can be used to specify whether the
10992       decimal floating point hardware instructions will be used or not.
10993       If none of them is given the hardware support is enabled by
10994       default.
10995     * The -mstack-guard option can now be omitted when using stack
10996       checking via -mstack-size in order to let GCC choose a sensible
10997       stack guard value according to the frame size of each function.
10998     * Various changes to improve performance of generated code have been
10999       implemented, including:
11000          + The condition code set by an add logical with carry
11001            instruction is now available for overflow checks like: a + b +
11002            carry < b.
11003          + The test data class instruction is now used to implement
11004            sign-bit and infinity checks of binary and decimal floating
11005            point numbers.
11006
11007  SPARC
11008
11009     * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been
11010       added.
11011
11012  Xtensa
11013
11014     * Stack unwinding for exception handling now uses by default a
11015       specialized version of DWARF unwinding. This is not
11016       binary-compatible with the setjmp/longjmp (sjlj) unwinding used for
11017       Xtensa with previous versions of GCC.
11018     * For Xtensa processors that include the Conditional Store option,
11019       the built-in functions for atomic memory access are now implemented
11020       using S32C1I instructions.
11021     * If the Xtensa NSA option is available, GCC will use it to implement
11022       the __builtin_ctz and __builtin_clz functions.
11023
11024Documentation improvements
11025
11026     * Existing libstdc++ documentation has been edited and restructured
11027       into a single DocBook XML manual. The results can be viewed online
11028       [18]here.
11029
11030Other significant improvements
11031
11032     * The compiler's --help command-line option has been extended so that
11033       it now takes an optional set of arguments. These arguments restrict
11034       the information displayed to specific classes of command-line
11035       options, and possibly only a subset of those options. It is also
11036       now possible to replace the descriptive text associated with each
11037       displayed option with an indication of its current value, or for
11038       binary options, whether it has been enabled or disabled.
11039       Here are some examples. The following will display all the options
11040       controlling warning messages:
11041      --help=warnings
11042
11043       Whereas this will display all the undocumented, target specific
11044       options:
11045      --help=target,undocumented
11046
11047       This sequence of commands will display the binary optimizations
11048       that are enabled by -O3:
11049      gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
11050      gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
11051      diff /tmp/O2-opts /tmp/O3-opts | grep enabled
11052
11053     * The configure options --with-pkgversion and --with-bugurl have been
11054       added. These allow distributors of GCC to include a
11055       distributor-specific string in manuals and --version output and to
11056       specify the URL for reporting bugs in their versions of GCC.
11057
11058GCC 4.3.1
11059
11060   This is the [19]list of problem reports (PRs) from GCC's bug tracking
11061   system that are known to be fixed in the 4.3.1 release. This list might
11062   not be complete (that is, it is possible that some PRs that have been
11063   fixed are not listed here).
11064
11065Target Specific Changes
11066
11067  IA-32/x86-64
11068
11069    ABI changes
11070
11071     * Starting with GCC 4.3.1, decimal floating point variables are
11072       aligned to their natural boundaries when they are passed on the
11073       stack for i386.
11074
11075    Command-line changes
11076
11077     * Starting with GCC 4.3.1, the -mcld option has been added to
11078       automatically generate a cld instruction in the prologue of
11079       functions that use string instructions. This option is used for
11080       backward compatibility on some operating systems and can be enabled
11081       by default for 32-bit x86 targets by configuring GCC with the
11082       --enable-cld configure option.
11083
11084GCC 4.3.2
11085
11086   This is the [20]list of problem reports (PRs) from GCC's bug tracking
11087   system that are known to be fixed in the 4.3.2 release. This list might
11088   not be complete (that is, it is possible that some PRs that have been
11089   fixed are not listed here).
11090
11091GCC 4.3.3
11092
11093   This is the [21]list of problem reports (PRs) from GCC's bug tracking
11094   system that are known to be fixed in the 4.3.3 release. This list might
11095   not be complete (that is, it is possible that some PRs that have been
11096   fixed are not listed here).
11097
11098GCC 4.3.4
11099
11100   This is the [22]list of problem reports (PRs) from GCC's bug tracking
11101   system that are known to be fixed in the 4.3.4 release. This list might
11102   not be complete (that is, it is possible that some PRs that have been
11103   fixed are not listed here).
11104
11105GCC 4.3.5
11106
11107   This is the [23]list of problem reports (PRs) from GCC's bug tracking
11108   system that are known to be fixed in the 4.3.5 release. This list might
11109   not be complete (that is, it is possible that some PRs that have been
11110   fixed are not listed here).
11111
11112GCC 4.3.6
11113
11114   This is the [24]list of problem reports (PRs) from GCC's bug tracking
11115   system that are known to be fixed in the 4.3.6 release. This list might
11116   not be complete (that is, it is possible that some PRs that have been
11117   fixed are not listed here).
11118
11119
11120    For questions related to the use of GCC, please consult these web
11121    pages and the [25]GCC manuals. If that fails, the
11122    [26]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11123    web pages and the development of GCC are welcome on our developer
11124    list at [27]gcc@gcc.gnu.org. All of [28]our lists have public
11125    archives.
11126
11127   Copyright (C) [29]Free Software Foundation, Inc. Verbatim copying and
11128   distribution of this entire article is permitted in any medium,
11129   provided this notice is preserved.
11130
11131   These pages are [30]maintained by the GCC team. Last modified
11132   2019-11-28[31].
11133
11134References
11135
11136   1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5
11137   2. https://gmplib.org/
11138   3. https://www.mpfr.org/
11139   4. https://gcc.gnu.org/install/prerequisites.html
11140   5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html
11141   6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
11142   7. http://gcc.gnu.org/gcc-4.3/porting_to.html
11143   8. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
11144   9. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
11145  10. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
11146  11. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html
11147  12. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options
11148  13. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167
11149  14. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html
11150  15. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html
11151  16. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html
11152  17. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html
11153  18. https://gcc.gnu.org/onlinedocs/libstdc++/
11154  19. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1
11155  20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2
11156  21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3
11157  22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4
11158  23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5
11159  24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6
11160  25. https://gcc.gnu.org/onlinedocs/
11161  26. mailto:gcc-help@gcc.gnu.org
11162  27. mailto:gcc@gcc.gnu.org
11163  28. https://gcc.gnu.org/lists.html
11164  29. https://www.fsf.org/
11165  30. https://gcc.gnu.org/about.html
11166  31. http://validator.w3.org/check/referer
11167======================================================================
11168http://gcc.gnu.org/gcc-4.2/index.html
11169                             GCC 4.2 Release Series
11170
11171   (This release series is no longer supported.)
11172
11173   May 19, 2008
11174
11175   The [1]GNU project and the GCC developers are pleased to announce the
11176   release of GCC 4.2.4.
11177
11178   This release is a bug-fix release, containing fixes for regressions in
11179   GCC 4.2.3 relative to previous releases of GCC.
11180
11181Release History
11182
11183   GCC 4.2.4
11184          May 19, 2008 ([2]changes)
11185
11186   GCC 4.2.3
11187          February 1, 2008 ([3]changes)
11188
11189   GCC 4.2.2
11190          October 7, 2007 ([4]changes)
11191
11192   GCC 4.2.1
11193          July 18, 2007 ([5]changes)
11194
11195   GCC 4.2.0
11196          May 13, 2007 ([6]changes)
11197
11198References and Acknowledgements
11199
11200   GCC used to stand for the GNU C Compiler, but since the compiler
11201   supports several other languages aside from C, it now stands for the
11202   GNU Compiler Collection.
11203
11204   A list of [7]successful builds is updated as new information becomes
11205   available.
11206
11207   The GCC developers would like to thank the numerous people that have
11208   contributed new features, improvements, bug fixes, and other changes as
11209   well as test results to GCC. This [8]amazing group of volunteers is
11210   what makes GCC successful.
11211
11212   For additional information about GCC please refer to the [9]GCC project
11213   web site or contact the [10]GCC development mailing list.
11214
11215   To obtain GCC please use [11]our mirror sites or [12]our version
11216   control system.
11217
11218
11219    For questions related to the use of GCC, please consult these web
11220    pages and the [13]GCC manuals. If that fails, the
11221    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11222    web pages and the development of GCC are welcome on our developer
11223    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
11224    archives.
11225
11226   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
11227   distribution of this entire article is permitted in any medium,
11228   provided this notice is preserved.
11229
11230   These pages are [18]maintained by the GCC team. Last modified
11231   2020-01-14[19].
11232
11233References
11234
11235   1. http://www.gnu.org/
11236   2. http://gcc.gnu.org/gcc-4.2/changes.html
11237   3. http://gcc.gnu.org/gcc-4.2/changes.html
11238   4. http://gcc.gnu.org/gcc-4.2/changes.html
11239   5. http://gcc.gnu.org/gcc-4.2/changes.html
11240   6. http://gcc.gnu.org/gcc-4.2/changes.html
11241   7. http://gcc.gnu.org/gcc-4.2/buildstat.html
11242   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
11243   9. http://gcc.gnu.org/index.html
11244  10. mailto:gcc@gcc.gnu.org
11245  11. http://gcc.gnu.org/mirrors.html
11246  12. http://gcc.gnu.org/git.html
11247  13. https://gcc.gnu.org/onlinedocs/
11248  14. mailto:gcc-help@gcc.gnu.org
11249  15. mailto:gcc@gcc.gnu.org
11250  16. https://gcc.gnu.org/lists.html
11251  17. https://www.fsf.org/
11252  18. https://gcc.gnu.org/about.html
11253  19. http://validator.w3.org/check/referer
11254======================================================================
11255http://gcc.gnu.org/gcc-4.2/changes.html
11256                             GCC 4.2 Release Series
11257                        Changes, New Features, and Fixes
11258
11259Caveats
11260
11261     * GCC no longer accepts the -fshared-data option. This option has had
11262       no effect in any GCC 4 release; the targets to which the option
11263       used to apply had been removed before GCC 4.0.
11264
11265General Optimizer Improvements
11266
11267     * New command-line options specify the possible relationships among
11268       parameters and between parameters and global data. For example,
11269       -fargument-noalias-anything specifies that arguments do not alias
11270       any other storage.
11271       Each language will automatically use whatever option is required by
11272       the language standard. You should not need to use these options
11273       yourself.
11274
11275New Languages and Language specific improvements
11276
11277     * [1]OpenMP is now supported for the C, C++ and Fortran compilers.
11278     * New command-line options -fstrict-overflow and -Wstrict-overflow
11279       have been added. -fstrict-overflow tells the compiler that it may
11280       assume that the program follows the strict signed overflow
11281       semantics permitted for the language: for C and C++ this means that
11282       the compiler may assume that signed overflow does not occur. For
11283       example, a loop like
11284      for (i = 1; i > 0; i *= 2)
11285
11286       is presumably intended to continue looping until i overflows. With
11287       -fstrict-overflow, the compiler may assume that signed overflow
11288       will not occur, and transform this into an infinite loop.
11289       -fstrict-overflow is turned on by default at -O2, and may be
11290       disabled via -fno-strict-overflow. The -Wstrict-overflow option may
11291       be used to warn about cases where the compiler assumes that signed
11292       overflow will not occur. It takes five different levels:
11293       -Wstrict-overflow=1 to 5. See the [2]documentation for details.
11294       -Wstrict-overflow=1 is enabled by -Wall.
11295     * The new command-line option -fno-toplevel-reorder directs GCC to
11296       emit top-level functions, variables, and asm statements in the same
11297       order that they appear in the input file. This is intended to
11298       support existing code which relies on a particular ordering (for
11299       example, code which uses top-level asm statements to switch
11300       sections). For new code, it is generally better to use function and
11301       variable attributes. The -fno-toplevel-reorder option may be used
11302       for most cases which currently use -fno-unit-at-a-time. The
11303       -fno-unit-at-a-time option will be removed in some future version
11304       of GCC. If you know of a case which requires -fno-unit-at-a-time
11305       which is not fixed by -fno-toplevel-reorder, please open a bug
11306       report.
11307
11308  C family
11309
11310     * The pragma redefine_extname will now macro expand its tokens for
11311       compatibility with SunPRO.
11312     * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct
11313       GCC to handle inline functions as specified in the C99 standard. In
11314       preparation for this, GCC 4.2 will warn about any use of non-static
11315       inline functions in gnu99 or c99 mode. This new warning may be
11316       disabled with the new gnu_inline function attribute or the new
11317       -fgnu89-inline command-line option. Also, GCC 4.2 and later will
11318       define one of the preprocessor macros __GNUC_GNU_INLINE__ or
11319       __GNUC_STDC_INLINE__ to indicate the semantics of inline functions
11320       in the current compilation.
11321     * A new command-line option -Waddress has been added to warn about
11322       suspicious uses of memory addresses as, for example, using the
11323       address of a function in a conditional expression, and comparisons
11324       against the memory address of a string literal. This warning is
11325       enabled by -Wall.
11326
11327  C++
11328
11329     * C++ visibility handling has been overhauled.
11330       Restricted visiblity is propagated from classes to members, from
11331       functions to local statics, and from templates and template
11332       arguments to instantiations, unless the latter has explicitly
11333       declared visibility.
11334       The visibility attribute for a class must come between the
11335       class-key and the name, not after the closing brace.
11336       Attributes are now allowed for enums and elaborated-type-specifiers
11337       that only declare a type.
11338       Members of the anonymous namespace are now local to a particular
11339       translation unit, along with any other declarations which use them,
11340       though they are still treated as having external linkage for
11341       language semantics.
11342     * The (undocumented) extension which permitted templates with default
11343       arguments to be bound to template template parameters with fewer
11344       parameters has been removed. For example:
11345        template <template <typename> class C>
11346        void f(C<double>) {}
11347
11348        template <typename T, typename U = int>
11349        struct S {};
11350
11351        template void f(S<double>);
11352
11353       is no longer accepted by G++. The reason this code is not accepted
11354       is that S is a template with two parameters; therefore, it cannot
11355       be bound to C which has only one parameter.
11356     * The <?, >?, <?=, and >?= operators, deprecated in previous GCC
11357       releases, have been removed.
11358     * The command-line option -fconst-strings, deprecated in previous GCC
11359       releases, has been removed.
11360     * The configure variable enable-__cxa_atexit is now enabled by
11361       default for more targets. Enabling this variable is necessary in
11362       order for static destructors to be executed in the correct order,
11363       but it depends upon the presence of a non-standard C library in the
11364       target library in order to work. The variable is now enabled for
11365       more targets which are known to have suitable C libraries.
11366     * -Wextra will produce warnings for if statements with a semicolon as
11367       the only body, to catch code like:
11368         if (a);
11369            return 1;
11370         return 0;
11371
11372       To suppress the warning in valid cases, use { } instead.
11373     * The C++ front end now also produces strict aliasing warnings when
11374       -fstrict-aliasing -Wstrict-aliasing is in effect.
11375
11376    Runtime Library (libstdc++)
11377
11378     * Added support for TR1 <random>, <complex>, and C compatibility
11379       headers. In addition, a lock-free version of shared_ptr was
11380       contributed as part of Phillip Jordan's Google Summer of Code
11381       project on lock-free containers.
11382     * In association with the Summer of Code work on lock-free
11383       containers, the interface for atomic builtins was adjusted,
11384       creating simpler alternatives for non-threaded code paths. Also,
11385       usage was consolidated and all elements were moved from namespace
11386       std to namespace__gnu_cxx. Affected interfaces are the functions
11387       __exchange_and_add, __atomic_add, and the objects __mutex,
11388       __recursive_mutex, and __scoped_lock.
11389     * Support for versioning weak symbol names via namespace association
11390       was added. However, as this changes the names of exported symbols,
11391       this is turned off by default in the current ABI. Intrepid users
11392       can enable this feature by using
11393       --enable-symvers=gnu-versioned-namespace during configuration.
11394     * Revised, simplified, and expanded policy-based associative
11395       containers, including data types for tree and trie forms
11396       (basic_tree, tree, trie), lists (list_update), and both
11397       collision-chaining and probing hash-based containers
11398       (basic_hash_table, cc_hash_table, gp_hash_table). More details per
11399       the [3]documentation.
11400     * The implementation of the debug mode was modified, whereby the
11401       debug namespaces were nested inside of namespace std and namespace
11402       __gnu_cxx in order to resolve some long standing corner cases
11403       involving name lookup. Debug functionality from the policy-based
11404       data structures was consolidated and enabled with the single macro,
11405       _GLIBCXX_DEBUG. See PR 26142 for more information.
11406     * Added extensions for type traits: __conditional_type,
11407       __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if.
11408     * Added a typelist implementation for compile-time meta-programming.
11409       Elements for typelist construction and operation can be found
11410       within namespace __gnu_cxx::typelist.
11411     * Added a new allocator, __gnu_cxx::throw_allocator, for testing
11412       exception-safety.
11413     * Enabled library-wide visibility control, allowing -fvisibility to
11414       be used.
11415     * Consolidated all nested namespaces and the conversion of
11416       __gnu_internal implementation-private details to anonymous
11417       namespaces whenever possible.
11418     * Implemented LWG resolutions DR 431 and DR 538.
11419
11420  Fortran
11421
11422     * Support for allocatable components has been added (TR 15581 and
11423       Fortran 2003).
11424     * Support for the Fortran 2003 streaming IO extension has been added.
11425     * The GNU Fortran compiler now uses 4-byte record markers by default
11426       for unformatted files to be compatible with g77 and most other
11427       compilers. The implementation allows for records greater than 2 GB
11428       and is compatible with several other compilers. Older versions of
11429       gfortran used 8-byte record markers by default (on most systems).
11430       In order to change the length of the record markers, e.g. to read
11431       unformatted files created by older gfortran versions, the
11432       [4]-frecord-marker=8 option can be used.
11433
11434  Java (GCJ)
11435
11436     * A new command-line option -static-libgcj has been added for targets
11437       that use a linker compatible with GNU Binutils. As its name
11438       implies, this causes libgcj to be linked statically. In some cases
11439       this causes the resulting executable to start faster and use less
11440       memory than if the shared version of libgcj were used. However
11441       caution should be used as it can also cause essential parts of the
11442       library to be omitted. Some of these issues are discussed in:
11443       [5]https://gcc.gnu.org/wiki/Statically_linking_libgcj
11444     * fastjar is no longer bundled with GCC. To build libgcj, you will
11445       need either InfoZIP (both zip and unzip) or an external jar
11446       program. In the former case, the GCC build will install a jar shell
11447       script that is based on InfoZIP and provides the same functionality
11448       as fastjar.
11449
11450New Targets and Target Specific Improvements
11451
11452  IA-32/x86-64
11453
11454     * -mtune=generic can now be used to generate code running well on
11455       common x86 chips. This includes AMD Athlon, AMD Opteron, Intel
11456       Pentium-M, Intel Pentium 4 and Intel Core 2.
11457     * -mtune=native and -march=native will produce code optimized for the
11458       host architecture as detected using the cpuid instruction.
11459     * Added a new command-line option -fstackrealign and and
11460       __attribute__ ((force_align_arg_pointer)) to realign the stack at
11461       runtime. This allows functions compiled with a vector-aligned stack
11462       to be invoked from legacy objects that keep only word-alignment.
11463
11464  SPARC
11465
11466     * The default CPU setting has been changed from V7 to V9 in 32-bit
11467       mode on Solaris 7 and above. This is already the case in 64-bit
11468       mode. It can be overridden by specifying --with-cpu at configure
11469       time.
11470     * Back-end support of built-in functions for atomic memory access has
11471       been implemented.
11472     * Support for the Sun UltraSPARC T1 (Niagara) processor has been
11473       added.
11474
11475  M32C
11476
11477     * Various bug fixes have made some functions (notably, functions
11478       returning structures) incompatible with previous releases.
11479       Recompiling all libraries is recommended. Note that code quality
11480       has considerably improved since 4.1, making a recompile even more
11481       beneficial.
11482
11483  MIPS
11484
11485     * Added support for the Broadcom SB-1A core.
11486
11487  IA-64
11488
11489     * Added support for IA-64 data and control speculation. By default
11490       speculation is enabled only during second scheduler pass. A number
11491       of machine flags was introduced to control the usage of speculation
11492       for both scheduler passes.
11493
11494  HPPA
11495
11496     * Added Java language support (libffi and libjava) for 32-bit HP-UX
11497       11 target.
11498
11499Obsolete Systems
11500
11501Documentation improvements
11502
11503  PDF Documentation
11504
11505     * A make pdf target has been added to the top-level makefile,
11506       enabling automated production of PDF documentation files.
11507       (Front-ends external to GCC should modify their Make-lang.in file
11508       to add a lang.pdf: target.)
11509
11510Other significant improvements
11511
11512  Build system improvements
11513
11514     * All the components of the compiler are now bootstrapped by default.
11515       This improves the resilience to bugs in the system compiler or
11516       binary compatibility problems, as well as providing better testing
11517       of GCC 4.2 itself. In addition, if you build the compiler from a
11518       combined tree, the assembler, linker, etc. will also be
11519       bootstrapped (i.e. built with themselves).
11520       You can disable this behavior, and go back to the pre-GCC 4.2 set
11521       up, by configuring GCC with --disable-bootstrap.
11522     * The rules that configure follows to find target tools resemble more
11523       closely the locations that the built compiler will search. In
11524       addition, you can use the new configure option --with-target-tools
11525       to specify where to find the target tools used during the build,
11526       without affecting what the built compiler will use.
11527       This can be especially useful when building packages of GCC. For
11528       example, you may want to build GCC with GNU as or ld, even if the
11529       resulting compiler to work with the native assembler and linker. To
11530       do so, you can use --with-target-tools to point to the native
11531       tools.
11532
11533  Incompatible changes to the build system
11534
11535     * Front-ends external to GCC should modify their Make-lang.in file to
11536       replace double-colon rules (e.g. dvi::) with normal rules (like
11537       lang.dvi:). Front-end makefile hooks do not use double-colon rules
11538       anymore.
11539     * Up to GCC 4.1, a popular way to specify the target tools used
11540       during the build was to create directories named gas, binutils,
11541       etc. in the build tree, and create links to the tools from there.
11542       This does not work any more when the compiler is bootstrapped. The
11543       new configure option --with-target-tools provides a better way to
11544       achieve the same effect, and works for all native and cross
11545       settings.
11546
11547
11548    For questions related to the use of GCC, please consult these web
11549    pages and the [6]GCC manuals. If that fails, the
11550    [7]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11551    web pages and the development of GCC are welcome on our developer
11552    list at [8]gcc@gcc.gnu.org. All of [9]our lists have public archives.
11553
11554   Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and
11555   distribution of this entire article is permitted in any medium,
11556   provided this notice is preserved.
11557
11558   These pages are [11]maintained by the GCC team. Last modified
11559   2019-11-28[12].
11560
11561References
11562
11563   1. http://gcc.gnu.org/projects/gomp/
11564   2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
11565   3. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html
11566   4. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html
11567   5. https://gcc.gnu.org/wiki/Statically_linking_libgcj
11568   6. https://gcc.gnu.org/onlinedocs/
11569   7. mailto:gcc-help@gcc.gnu.org
11570   8. mailto:gcc@gcc.gnu.org
11571   9. https://gcc.gnu.org/lists.html
11572  10. https://www.fsf.org/
11573  11. https://gcc.gnu.org/about.html
11574  12. http://validator.w3.org/check/referer
11575======================================================================
11576http://gcc.gnu.org/gcc-4.1/index.html
11577                             GCC 4.1 Release Series
11578
11579   (This release series is no longer supported.)
11580
11581   February 13, 2007
11582
11583   The [1]GNU project and the GCC developers are pleased to announce the
11584   release of GCC 4.1.2.
11585
11586   This release is a bug-fix release, containing fixes for regressions in
11587   GCC 4.1.1 relative to previous releases of GCC.
11588
11589Release History
11590
11591   GCC 4.1.2
11592          February 13, 2007 ([2]changes)
11593
11594   GCC 4.1.1
11595          May 24, 2006 ([3]changes)
11596
11597   GCC 4.1.0
11598          February 28, 2006 ([4]changes)
11599
11600References and Acknowledgements
11601
11602   GCC used to stand for the GNU C Compiler, but since the compiler
11603   supports several other languages aside from C, it now stands for the
11604   GNU Compiler Collection.
11605
11606   A list of [5]successful builds is updated as new information becomes
11607   available.
11608
11609   The GCC developers would like to thank the numerous people that have
11610   contributed new features, improvements, bug fixes, and other changes as
11611   well as test results to GCC. This [6]amazing group of volunteers is
11612   what makes GCC successful.
11613
11614   For additional information about GCC please refer to the [7]GCC project
11615   web site or contact the [8]GCC development mailing list.
11616
11617   To obtain GCC please use [9]our mirror sites or [10]our version control
11618   system.
11619
11620
11621    For questions related to the use of GCC, please consult these web
11622    pages and the [11]GCC manuals. If that fails, the
11623    [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11624    web pages and the development of GCC are welcome on our developer
11625    list at [13]gcc@gcc.gnu.org. All of [14]our lists have public
11626    archives.
11627
11628   Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and
11629   distribution of this entire article is permitted in any medium,
11630   provided this notice is preserved.
11631
11632   These pages are [16]maintained by the GCC team. Last modified
11633   2020-01-14[17].
11634
11635References
11636
11637   1. http://www.gnu.org/
11638   2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
11639   3. http://gcc.gnu.org/gcc-4.1/changes.html
11640   4. http://gcc.gnu.org/gcc-4.1/changes.html
11641   5. http://gcc.gnu.org/gcc-4.1/buildstat.html
11642   6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
11643   7. http://gcc.gnu.org/index.html
11644   8. mailto:gcc@gcc.gnu.org
11645   9. http://gcc.gnu.org/mirrors.html
11646  10. http://gcc.gnu.org/git.html
11647  11. https://gcc.gnu.org/onlinedocs/
11648  12. mailto:gcc-help@gcc.gnu.org
11649  13. mailto:gcc@gcc.gnu.org
11650  14. https://gcc.gnu.org/lists.html
11651  15. https://www.fsf.org/
11652  16. https://gcc.gnu.org/about.html
11653  17. http://validator.w3.org/check/referer
11654======================================================================
11655http://gcc.gnu.org/gcc-4.1/changes.html
11656                             GCC 4.1 Release Series
11657                        Changes, New Features, and Fixes
11658
11659   The latest release in the 4.1 release series is [1]GCC 4.1.2.
11660
11661Caveats
11662
11663General Optimizer Improvements
11664
11665     * GCC now has infrastructure for inter-procedural optimizations and
11666       the following inter-procedural optimizations are implemented:
11667          + Profile guided inlining. When doing profile feedback guided
11668            optimization, GCC can now use the profile to make better
11669            informed decisions on whether inlining of a function is
11670            profitable or not. This means that GCC will no longer inline
11671            functions at call sites that are not executed very often, and
11672            that functions at hot call sites are more likely to be
11673            inlined.
11674            A new parameter min-inline-recursive-probability is also now
11675            available to throttle recursive inlining of functions with
11676            small average recursive depths.
11677          + Discovery of pure and const functions, a form of side-effects
11678            analysis. While older GCC releases could also discover such
11679            special functions, the new IPA-based pass runs earlier so that
11680            the results are available to more optimizers. The pass is also
11681            simply more powerful than the old one.
11682          + Analysis of references to static variables and type escape
11683            analysis, also forms of side-effects analysis. The results of
11684            these passes allow the compiler to be less conservative about
11685            call-clobbered variables and references. This results in more
11686            redundant loads being eliminated and in making static
11687            variables candidates for register promotion.
11688          + Improvement of RTL-based alias analysis. The results of type
11689            escape analysis are fed to the RTL type-based alias analyzer,
11690            allowing it to disambiguate more memory references.
11691          + Interprocedural constant propagation and function versioning.
11692            This pass looks for functions that are always called with the
11693            same constant value for one or more of the function arguments,
11694            and propagates those constants into those functions.
11695          + GCC will now eliminate static variables whose usage was
11696            optimized out.
11697          + -fwhole-program --combine can now be used to make all
11698            functions in program static allowing whole program
11699            optimization. As an exception, the main function and all
11700            functions marked with the new externally_visible attribute are
11701            kept global so that programs can link with runtime libraries.
11702     * GCC can now do a form of partial dead code elimination (PDCE) that
11703       allows code motion of expressions to the paths where the result of
11704       the expression is actually needed. This is not always a win, so the
11705       pass has been limited to only consider profitable cases. Here is an
11706       example:
11707    int foo (int *, int *);
11708    int
11709    bar (int d)
11710    {
11711      int a, b, c;
11712      b = d + 1;
11713      c = d + 2;
11714      a = b + c;
11715      if (d)
11716        {
11717          foo (&b, &c);
11718          a = b + c;
11719        }
11720      printf ("%d\n", a);
11721    }
11722
11723       The a = b + c can be sunk to right before the printf. Normal code
11724       sinking will not do this, it will sink the first one above into the
11725       else-branch of the conditional jump, which still gives you two
11726       copies of the code.
11727     * GCC now has a value range propagation pass. This allows the
11728       compiler to eliminate bounds checks and branches. The results of
11729       the pass can also be used to accurately compute branch
11730       probabilities.
11731     * The pass to convert PHI nodes to straight-line code (a form of
11732       if-conversion for GIMPLE) has been improved significantly. The two
11733       most significant improvements are an improved algorithm to
11734       determine the order in which the PHI nodes are considered, and an
11735       improvement that allow the pass to consider if-conversions of basic
11736       blocks with more than two predecessors.
11737     * Alias analysis improvements. GCC can now differentiate between
11738       different fields of structures in Tree-SSA's virtual operands form.
11739       This lets stores/loads from non-overlapping structure fields not
11740       conflict. A new algorithm to compute points-to sets was contributed
11741       that can allows GCC to see now that p->a and p->b, where p is a
11742       pointer to a structure, can never point to the same field.
11743     * Various enhancements to auto-vectorization:
11744          + Incrementally preserve SSA form when vectorizing.
11745          + Incrementally preserve loop-closed form when vectorizing.
11746          + Improvements to peeling for alignment: generate better code
11747            when the misalignment of an access is known at compile time,
11748            or when different accesses are known to have the same
11749            misalignment, even if the misalignment amount itself is
11750            unknown.
11751          + Consider dependence distance in the vectorizer.
11752          + Externalize generic parts of data reference analysis to make
11753            this analysis available to other passes.
11754          + Vectorization of conditional code.
11755          + Reduction support.
11756     * GCC can now partition functions in sections of hot and cold code.
11757       This can significantly improve performance due to better
11758       instruction cache locality. This feature works best together with
11759       profile feedback driven optimization.
11760     * A new pass to avoid saving of unneeded arguments to the stack in
11761       vararg functions if the compiler can prove that they will not be
11762       needed.
11763     * Transition of basic block profiling to tree level implementation
11764       has been completed. The new implementation should be considerably
11765       more reliable (hopefully avoiding profile mismatch errors when
11766       using -fprofile-use or -fbranch-probabilities) and can be used to
11767       drive higher level optimizations, such as inlining.
11768       The -ftree-based-profiling command-line option was removed and
11769       -fprofile-use now implies disabling old RTL level loop optimizer
11770       (-fno-loop-optimize). Speculative prefetching optimization
11771       (originally enabled by -fspeculative-prefetching) was removed.
11772
11773New Languages and Language specific improvements
11774
11775  C and Objective-C
11776
11777     * The old Bison-based C and Objective-C parser has been replaced by a
11778       new, faster hand-written recursive-descent parser.
11779
11780  Ada
11781
11782     * The build infrastructure for the Ada runtime library and tools has
11783       been changed to be better integrated with the rest of the build
11784       infrastructure of GCC. This should make doing cross builds of Ada a
11785       bit easier.
11786
11787  C++
11788
11789     * ARM-style name-injection of friend declarations is no longer the
11790       default. For example:
11791          struct S {
11792            friend void f();
11793          };
11794
11795          void g() { f(); }
11796       will not be accepted; instead a declaration of f will need to be
11797       present outside of the scope of S. The new -ffriend-injection
11798       option will enable the old behavior.
11799     * The (undocumented) extension which permitted templates with default
11800       arguments to be bound to template template parameters with fewer
11801       parameters has been deprecated, and will be removed in the next
11802       major release of G++. For example:
11803       template <template <typename> class C>
11804       void f(C<double>) {}
11805
11806       template <typename T, typename U = int>
11807       struct S {};
11808
11809       template void f(S<double>);
11810
11811       makes use of the deprecated extension. The reason this code is not
11812       valid ISO C++ is that S is a template with two parameters;
11813       therefore, it cannot be bound to C which has only one parameter.
11814
11815    Runtime Library (libstdc++)
11816
11817     * Optimization work:
11818          + A new implementation of std::search_n is provided, better
11819            performing in case of random access iterators.
11820          + Added further efficient specializations of istream functions,
11821            i.e., character array and string extractors.
11822          + Other smaller improvements throughout.
11823     * Policy-based associative containers, designed for high-performance,
11824       flexibility and semantic safety are delivered in ext/pb_assoc.
11825     * A versatile string class, __gnu_cxx::__versa_string, providing
11826       facilities conforming to the standard requirements for
11827       basic_string, is delivered in <ext/vstring.h>. In particular:
11828          + Two base classes are provided: the default one avoids
11829            reference counting and is optimized for short strings; the
11830            alternate one, still uses it while improving in a few low
11831            level areas (e.g., alignment). See vstring_fwd.h for some
11832            useful typedefs.
11833          + Various algorithms have been rewritten (e.g., replace), the
11834            code streamlined and simple optimizations added.
11835          + Option 3 of DR 431 is implemented for both available bases,
11836            thus improving the support for stateful allocators.
11837     * As usual, many bugs have been fixed (e.g., libstdc++/13583,
11838       libstdc++/23953) and LWG resolutions put into effect for the first
11839       time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1
11840       Issue 6.19). The implementation status of TR1 is now tracked in the
11841       docs in tr1.html.
11842
11843  Objective-C++
11844
11845     * A new language front end for Objective-C++ has been added. This
11846       language allows users to mix the object oriented features of
11847       Objective-C with those of C++.
11848
11849  Java (GCJ)
11850
11851     * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19
11852       features (plus some 0.20 bug-fixes)
11853          + Networking
11854               o The java.net.HttpURLConnection implementation no longer
11855                 buffers the entire response body in memory. This means
11856                 that response bodies larger than available memory can now
11857                 be handled.
11858          + (N)IO
11859               o NIO FileChannel.map implementation, fast bulk put
11860                 implementation for DirectByteBuffer (speeds up this
11861                 method 10x).
11862               o FileChannel.lock() and FileChannel.force() implemented.
11863          + XML
11864               o gnu.xml fix for nodes created outside a namespace
11865                 context.
11866               o Add support for output indenting and
11867                 cdata-section-elements output instruction in
11868                 xml.transform.
11869               o xml.xpath corrections for cases where elements/attributes
11870                 might have been created in non-namespace-aware mode.
11871                 Corrections to handling of XSL variables and minor
11872                 conformance updates.
11873          + AWT
11874               o GNU JAWT implementation, the AWT Native Interface, which
11875                 allows direct access to native screen resources from
11876                 within a Canvas's paint method. GNU Classpath Examples
11877                 comes with a Demo, see libjava/classpath/examples/README.
11878               o awt.datatransfer updated to 1.5 with support for
11879                 FlavorEvents. The gtk+ awt peers now allow copy/paste of
11880                 text, images, URIs/files and serialized objects with
11881                 other applications and tracking clipboard change events
11882                 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized
11883                 objects are supported). A GNU Classpath Examples
11884                 datatransfer Demo was added to show the new
11885                 functionality.
11886               o Split gtk+ awt peers event handling in two threads and
11887                 improve gdk lock handling (solves several awt lock ups).
11888               o Speed up awt Image loading.
11889               o Better gtk+ scrollbar peer implementation when using gtk+
11890                 >= 2.6.
11891               o Handle image loading errors correctly for gdkpixbuf and
11892                 MediaTracker.
11893               o Better handle GDK lock. Properly prefix gtkpeer native
11894                 functions (cp_gtk).
11895               o GdkGraphics2D has been updated to use Cairo 0.5.x or
11896                 higher.
11897               o BufferedImage and GtkImage rewrites. All image drawing
11898                 operations should now work correctly (flipping requires
11899                 gtk+ >= 2.6)
11900               o Future Graphics2D, image and text work is documented at:
11901                 [2]http://developer.classpath.org/mediation/ClasspathGrap
11902                 hicsImagesText
11903               o When gtk+ 2.6 or higher is installed the default log
11904                 handler will produce stack traces whenever a WARNING,
11905                 CRITICAL or ERROR message is produced.
11906          + Free Swing
11907               o The RepaintManager has been reworked for more efficient
11908                 painting, especially for large GUIs.
11909               o The layout manager OverlayLayout has been implemented,
11910                 the BoxLayout has been rewritten to make use of the
11911                 SizeRequirements utility class and caching for more
11912                 efficient layout.
11913               o Improved accessibility support.
11914               o Significant progress has been made in the implementation
11915                 of the javax.swing.plaf.metal package, with most UI
11916                 delegates in a working state now. Please test this with
11917                 your own applications and provide feedback that will help
11918                 us to improve this package.
11919               o The GUI demo (gnu.classpath.examples.swing.Demo) has been
11920                 extended to highlight various features in our Free Swing
11921                 implementation. And it includes a look and feel switcher
11922                 for Metal (default), Ocean and GNU themes.
11923               o The javax.swing.plaf.multi package is now implemented.
11924               o Editing and several key actions for JTree and JTable were
11925                 implemented.
11926               o Lots of icons and look and feel improvements for Free
11927                 Swing basic and metal themes were added. Try running the
11928                 GNU Classpath Swing Demo in examples
11929                 (gnu.classpath.examples.swing.Demo) with:
11930                 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee
11931                 l or
11932                 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee
11933                 l
11934               o Start of styled text capabilites for java.swing.text.
11935               o DefaultMutableTreeNode pre-order, post-order, depth-first
11936                 and breadth-first traversal enumerations implemented.
11937               o JInternalFrame colors and titlebar draw properly.
11938               o JTree is working up to par (icons, selection and keyboard
11939                 traversal).
11940               o JMenus were made more compatible in visual and
11941                 programmatic behavior.
11942               o JTable changeSelection and multiple selections
11943                 implemented.
11944               o JButton and JToggleButton change states work properly
11945                 now.
11946               o JFileChooser fixes.
11947               o revalidate() and repaint() fixes which make Free Swing
11948                 much more responsive.
11949               o MetalIconFactory implemented.
11950               o Free Swing Top-Level Compatibility. JFrame, JDialog,
11951                 JApplet, JInternalFrame, and JWindow are now 1.5
11952                 compatible in the sense that you can call add() and
11953                 setLayout() directly on them, which will have the same
11954                 effect as calling getContentPane().add() and
11955                 getContentPane().setLayout().
11956               o The JTree interface has been completed. JTrees now
11957                 recognizes mouse clicks and selections work.
11958               o BoxLayout works properly now.
11959               o Fixed GrayFilter to actually work.
11960               o Metal SplitPane implemented.
11961               o Lots of Free Swing text and editor stuff work now.
11962          + Free RMI and Corba
11963               o Andrew Watson, Vice President and Technical Director of
11964                 the Object Management Group, has officially assigned us
11965                 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will
11966                 mark remote classpath-specific system exceptions.
11967                 Obtaining the VMCID means that GNU Classpath now is a
11968                 recogniseable type of node in a highly interoperable
11969                 CORBA world.
11970               o GNU Classpath now includes the first working draft to
11971                 support the RMI over IIOP protocol. The current
11972                 implementation is capable of remote invocations,
11973                 transferring various Serializables and Externalizables
11974                 via RMI-IIOP protocol. It can flatten graphs and, at
11975                 least for the simple cases, is interoperable with 1.5
11976                 JDKs.
11977               o org.omg.PortableInterceptor and related functionality in
11978                 other packages is now implemented:
11979                    # The sever and client interceptors work as required
11980                      since 1.4.
11981                    # The IOR interceptor works as needed for 1.5.
11982               o The org.omg.DynamicAny package is completed and passes
11983                 the prepared tests.
11984               o The Portable Object Adapter should now support the output
11985                 of the recent IDL to java compilers. These compilers now
11986                 generate servants and not CORBA objects as before, making
11987                 the output depend on the existing POA implementation.
11988                 Completing POA means that such code can already be tried
11989                 to run on Classpath. Our POA is tested for the following
11990                 usager scenarios:
11991                    # POA converts servant to the CORBA object.
11992                    # Servant provides to the CORBA object.
11993                    # POA activates new CORBA object with the given Object
11994                      Id (byte array) that is later accessible for the
11995                      servant.
11996                    # During the first call, the ServantActivator provides
11997                      servant for this and all subsequent calls on the
11998                      current object.
11999                    # During each call, the ServantLocator provides
12000                      servant for this call only.
12001                    # ServantLocator or ServantActivator forwards call to
12002                      another server.
12003                    # POA has a single servant, responsible for all
12004                      objects.
12005                    # POA has a default servant, but some objects are
12006                      explicitly connected to they specific servants.
12007                 The POA is verified using tests from the former
12008                 cost.omg.org.
12009               o The CORBA implementation is now a working prototype that
12010                 should support features up to 1.3 inclusive. We invite
12011                 groups writing CORBA dependent applications to try
12012                 Classpath implementation, reporting any possible bugs.
12013                 The CORBA prototype is interoperable with Sun's
12014                 implementation v 1.4, transferring object references,
12015                 primitive types, narrow and wide strings, arrays,
12016                 structures, trees, abstract interfaces and value types
12017                 (feature of CORBA 2.3) between these two platforms.
12018                 Remote exceptions are transferred and handled correctly.
12019                 The stringified object references (IORs) from various
12020                 sources are parsed as required. The transient (for
12021                 current session) and permanent (till jre restart)
12022                 redirections work. Both Little and Big Endian encoded
12023                 messages are accepted. The implementation is verified
12024                 using tests from the former cost.omg.org. The current
12025                 release includes working examples (see the examples
12026                 directory), demonstrating the client-server
12027                 communication, using either CORBA Request or IDL-based
12028                 stub (usually generated by a IDL to java compiler). These
12029                 examples also show how to use the Classpath CORBA naming
12030                 service. The IDL to java compiler is not yet written, but
12031                 as our library must be compatible, it naturally accepts
12032                 the output of other idlj implementations.
12033          + Misc
12034               o Updated TimeZone data against Olson tzdata2005l.
12035               o Make zip and jar packages UTF-8 clean.
12036               o "native" code builds and compiles (warning free) on
12037                 Darwin and Solaris.
12038               o java.util.logging.FileHandler now rotates files.
12039               o Start of a generic JDWP framework in gnu/classpath/jdwp.
12040                 This is unfinished, but feedback (at classpath@gnu.org)
12041                 from runtime hackers is greatly appreciated. Although
12042                 most of the work is currently being done around gcj/gij
12043                 we want this framework to be as VM neutral as possible.
12044                 Early design is described in:
12045                 [3]https://gcc.gnu.org/ml/java/2005-05/msg00260.html
12046               o QT4 AWT peers, enable by giving configure
12047                 --enable-qt-peer. Included, but not ready for production
12048                 yet. They are explicitly disabled and not supported. But
12049                 if you want to help with the development of these new
12050                 features we are interested in feedback. You will have to
12051                 explicitly enable them to try them out (and they will
12052                 most likely contain bugs).
12053               o Documentation fixes all over the place. See
12054                 [4]http://developer.classpath.org/doc/
12055
12056New Targets and Target Specific Improvements
12057
12058  IA-32/x86-64
12059
12060     * The x86-64 medium model (that allows building applications whose
12061       data segment exceeds 4GB) was redesigned to match latest ABI draft.
12062       New implementation split large datastructures into separate segment
12063       improving performance of accesses to small datastructures and also
12064       allows linking of small model libraries into medium model programs
12065       as long as the libraries are not accessing the large datastructures
12066       directly. Medium model is also supported in position independent
12067       code now.
12068       The ABI change results in partial incompatibility among medium
12069       model objects. Linking medium model libraries (or objects) compiled
12070       with new compiler into medium model program compiled with older
12071       will likely result in exceeding ranges of relocations.
12072       Binutils 2.16.91 or newer are required for compiling medium model
12073       now.
12074
12075  RS6000 (POWER/PowerPC)
12076
12077     * The AltiVec vector primitives in <altivec.h> are now implemented in
12078       a way that puts a smaller burden on the preprocessor, instead
12079       processing the "overloading" in the front ends. This should benefit
12080       compilation speed on AltiVec vector code.
12081     * AltiVec initializers now are generated more efficiently.
12082     * The popcountb instruction available on POWER5 now is generated.
12083     * The floating point round to integer instructions available on
12084       POWER5+ now is generated.
12085     * Floating point divides can be synthesized using the floating point
12086       reciprocal estimate instructions.
12087     * Double precision floating point constants are initialized as single
12088       precision values if they can be represented exactly.
12089
12090  S/390, zSeries and System z9
12091
12092     * Support for the IBM System z9 109 processor has been added. When
12093       using the -march=z9-109 option, the compiler will generate code
12094       making use of instructions provided by the extended immediate
12095       facility.
12096     * Support for 128-bit IEEE floating point has been added. When using
12097       the -mlong-double-128 option, the compiler will map the long double
12098       data type to 128-bit IEEE floating point. Using this option
12099       constitutes an ABI change, and requires glibc support.
12100     * Various changes to improve performance of generated code have been
12101       implemented, including:
12102          + In functions that do not require a literal pool, register %r13
12103            (which is traditionally reserved as literal pool pointer), can
12104            now be freely used for other purposes by the compiler.
12105          + More precise tracking of register use allows the compiler to
12106            generate more efficient function prolog and epilog code in
12107            certain cases.
12108          + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING
12109            instructions are now used to implement C string functions.
12110          + The MOVE CHARACTER instruction with single byte overlap is now
12111            used to implement the memset function with non-zero fill byte.
12112          + The LOAD ZERO instructions are now used where appropriate.
12113          + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK,
12114            and INSERT IMMEDIATE instructions are now used more frequently
12115            to optimize bitfield operations.
12116          + The BRANCH ON COUNT instruction is now used more frequently.
12117            In particular, the fact that a loop contains a subroutine call
12118            no longer prevents the compiler from using this instruction.
12119          + The compiler is now aware that all shift and rotate
12120            instructions implicitly truncate the shift count to six bits.
12121     * Back-end support for the following generic features has been
12122       implemented:
12123          + The full set of [5]built-in functions for atomic memory
12124            access.
12125          + The -fstack-protector feature.
12126          + The optimization pass avoiding unnecessary stores of incoming
12127            argument registers in functions with variable argument list.
12128
12129  SPARC
12130
12131     * The default code model in 64-bit mode has been changed from
12132       Medium/Anywhere to Medium/Middle on Solaris.
12133     * TLS support is disabled by default on Solaris prior to release 10.
12134       It can be enabled on TLS-capable Solaris 9 versions (4/04 release
12135       and later) by specifying --enable-tls at configure time.
12136
12137  MorphoSys
12138
12139     * Support has been added for this new architecture.
12140
12141Obsolete Systems
12142
12143Documentation improvements
12144
12145Other significant improvements
12146
12147     * GCC can now emit code for protecting applications from
12148       stack-smashing attacks. The protection is realized by buffer
12149       overflow detection and reordering of stack variables to avoid
12150       pointer corruption.
12151     * Some built-in functions have been fortified to protect them against
12152       various buffer overflow (and format string) vulnerabilities.
12153       Compared to the mudflap bounds checking feature, the safe builtins
12154       have far smaller overhead. This means that programs built using
12155       safe builtins should not experience any measurable slowdown.
12156
12157GCC 4.1.2
12158
12159   This is the [6]list of problem reports (PRs) from GCC's bug tracking
12160   system that are known to be fixed in the 4.1.2 release. This list might
12161   not be complete (that is, it is possible that some PRs that have been
12162   fixed are not listed here).
12163
12164   When generating code for a shared library, GCC now recognizes that
12165   global functions may be replaced when the program runs. Therefore, it
12166   is now more conservative in deducing information from the bodies of
12167   functions. For example, in this example:
12168    void f() {}
12169    void g() {
12170     try { f(); }
12171     catch (...) {
12172       cout << "Exception";
12173     }
12174    }
12175
12176   G++ would previously have optimized away the catch clause, since it
12177   would have concluded that f cannot throw exceptions. Because users may
12178   replace f with another function in the main body of the program, this
12179   optimization is unsafe, and is no longer performed. If you wish G++ to
12180   continue to optimize as before, you must add a throw() clause to the
12181   declaration of f to make clear that it does not throw exceptions.
12182
12183
12184    For questions related to the use of GCC, please consult these web
12185    pages and the [7]GCC manuals. If that fails, the
12186    [8]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12187    web pages and the development of GCC are welcome on our developer
12188    list at [9]gcc@gcc.gnu.org. All of [10]our lists have public
12189    archives.
12190
12191   Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and
12192   distribution of this entire article is permitted in any medium,
12193   provided this notice is preserved.
12194
12195   These pages are [12]maintained by the GCC team. Last modified
12196   2019-11-28[13].
12197
12198References
12199
12200   1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
12201   2. http://developer.classpath.org/mediation/ClasspathGraphicsImagesText
12202   3. https://gcc.gnu.org/ml/java/2005-05/msg00260.html
12203   4. http://developer.classpath.org/doc/
12204   5. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html
12205   6. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2
12206   7. https://gcc.gnu.org/onlinedocs/
12207   8. mailto:gcc-help@gcc.gnu.org
12208   9. mailto:gcc@gcc.gnu.org
12209  10. https://gcc.gnu.org/lists.html
12210  11. https://www.fsf.org/
12211  12. https://gcc.gnu.org/about.html
12212  13. http://validator.w3.org/check/referer
12213======================================================================
12214http://gcc.gnu.org/gcc-4.0/index.html
12215                             GCC 4.0 Release Series
12216
12217   (This release series is no longer supported.)
12218
12219   January 31, 2007
12220
12221   The [1]GNU project and the GCC developers are pleased to announce the
12222   release of GCC 4.0.4.
12223
12224   This release is a bug-fix release, containing fixes for regressions in
12225   GCC 4.0.3 relative to previous releases of GCC.
12226
12227Release History
12228
12229   GCC 4.0.4
12230          January 31, 2007 ([2]changes)
12231
12232   GCC 4.0.3
12233          March 10, 2006 ([3]changes)
12234
12235   GCC 4.0.2
12236          September 28, 2005 ([4]changes)
12237
12238   GCC 4.0.1
12239          July 7, 2005 ([5]changes)
12240
12241   GCC 4.0.0
12242          April 20, 2005 ([6]changes)
12243
12244References and Acknowledgements
12245
12246   GCC used to stand for the GNU C Compiler, but since the compiler
12247   supports several other languages aside from C, it now stands for the
12248   GNU Compiler Collection.
12249
12250   A list of [7]successful builds is updated as new information becomes
12251   available.
12252
12253   The GCC developers would like to thank the numerous people that have
12254   contributed new features, improvements, bug fixes, and other changes as
12255   well as test results to GCC. This [8]amazing group of volunteers is
12256   what makes GCC successful.
12257
12258   For additional information about GCC please refer to the [9]GCC project
12259   web site or contact the [10]GCC development mailing list.
12260
12261   To obtain GCC please use [11]our mirror sites, or [12]our version
12262   control system.
12263
12264
12265    For questions related to the use of GCC, please consult these web
12266    pages and the [13]GCC manuals. If that fails, the
12267    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12268    web pages and the development of GCC are welcome on our developer
12269    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
12270    archives.
12271
12272   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
12273   distribution of this entire article is permitted in any medium,
12274   provided this notice is preserved.
12275
12276   These pages are [18]maintained by the GCC team. Last modified
12277   2020-01-14[19].
12278
12279References
12280
12281   1. http://www.gnu.org/
12282   2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
12283   3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3
12284   4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2
12285   5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1
12286   6. http://gcc.gnu.org/gcc-4.0/changes.html
12287   7. http://gcc.gnu.org/gcc-4.0/buildstat.html
12288   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12289   9. http://gcc.gnu.org/index.html
12290  10. mailto:gcc@gcc.gnu.org
12291  11. http://gcc.gnu.org/mirrors.html
12292  12. http://gcc.gnu.org/git.html
12293  13. https://gcc.gnu.org/onlinedocs/
12294  14. mailto:gcc-help@gcc.gnu.org
12295  15. mailto:gcc@gcc.gnu.org
12296  16. https://gcc.gnu.org/lists.html
12297  17. https://www.fsf.org/
12298  18. https://gcc.gnu.org/about.html
12299  19. http://validator.w3.org/check/referer
12300======================================================================
12301http://gcc.gnu.org/gcc-4.0/changes.html
12302                             GCC 4.0 Release Series
12303                        Changes, New Features, and Fixes
12304
12305   The latest release in the 4.0 release series is [1]GCC 4.0.4.
12306
12307Caveats
12308
12309     * GCC now generates location lists by default when compiling with
12310       debug info and optimization.
12311          + GDB 6.0 and older crashes when it sees location lists. GDB 6.1
12312            or later is needed to debug binaries containing location
12313            lists.
12314          + When you are trying to view a value of a variable in a part of
12315            a function where it has no location (for example when the
12316            variable is no longer used and thus its location was used for
12317            something else) GDB will say that it is not available.
12318       You can disable generating location lists by -fno-var-tracking.
12319     * GCC no longer accepts the -fwritable-strings option. Use named
12320       character arrays when you need a writable string.
12321     * The options -freduce-all-givs and -fmove-all-movables have been
12322       discontinued. They were used to circumvent a shortcoming in the
12323       heuristics of the old loop optimization code with respect to common
12324       Fortran constructs. The new (tree) loop optimizer works differently
12325       and doesn't need those work-arounds.
12326     * The graph-coloring register allocator, formerly enabled by the
12327       option -fnew-ra, has been discontinued.
12328     * -I- has been deprecated. -iquote is meant to replace the need for
12329       this option.
12330     * The MIPS -membedded-pic and -mrnames options have been removed.
12331     * All MIPS targets now require the GNU assembler. In particular, IRIX
12332       configurations can no longer use the MIPSpro assemblers, although
12333       they do still support the MIPSpro linkers.
12334     * The SPARC option -mflat has been removed.
12335     * English-language diagnostic messages will now use Unicode quotation
12336       marks in UTF-8 locales. (Non-English messages already used the
12337       quotes appropriate for the language in previous releases.) If your
12338       terminal does not support UTF-8 but you are using a UTF-8 locale
12339       (such locales are the default on many GNU/Linux systems) then you
12340       should set LC_CTYPE=C in the environment to disable that locale.
12341       Programs that parse diagnostics and expect plain ASCII
12342       English-language messages should set LC_ALL=C. See [2]Markus Kuhn's
12343       explanation of Unicode quotation marks for more information.
12344     * The specs file is no longer installed on most platforms. Most users
12345       will be totally unaffected. However, if you are accustomed to
12346       editing the specs file yourself, you will now have to use the
12347       -dumpspecs option to generate the specs file, and then edit the
12348       resulting file.
12349
12350General Optimizer Improvements
12351
12352     * The [3]tree ssa branch has been merged. This merge has brought in a
12353       completely new optimization framework based on a higher level
12354       intermediate representation than the existing RTL representation.
12355       Numerous new code transformations based on the new framework are
12356       available in GCC 4.0, including:
12357          + Scalar replacement of aggregates
12358          + Constant propagation
12359          + Value range propagation
12360          + Partial redundancy elimination
12361          + Load and store motion
12362          + Strength reduction
12363          + Dead store elimination
12364          + Dead and unreachable code elimination
12365          + [4]Autovectorization
12366          + Loop interchange
12367          + Tail recursion by accumulation
12368       Many of these passes outperform their counterparts from previous
12369       GCC releases.
12370     * [5]Swing Modulo Scheduling (SMS). An RTL level instruction
12371       scheduling optimization intended for loops that perform heavy
12372       computations.
12373
12374New Languages and Language specific improvements
12375
12376  C family
12377
12378     * The sentinel attribute has been added to GCC. This function
12379       attribute allows GCC to warn when variadic functions such as execl
12380       are not NULL terminated. See the GCC manual for a complete
12381       description of its behavior.
12382     * Given __attribute__((alias("target"))) it is now an error if target
12383       is not a symbol, defined in the same translation unit. This also
12384       applies to aliases created by #pragma weak alias=target. This is
12385       because it's meaningless to define an alias to an undefined symbol.
12386       On Solaris, the native assembler would have caught this error, but
12387       GNU as does not.
12388
12389  C and Objective-C
12390
12391     * The -Wstrict-aliasing=2 option has been added. This warning catches
12392       all unsafe cases, but it may also give a warning for some cases
12393       that are safe.
12394     * The cast-as-lvalue, conditional-expression-as-lvalue and
12395       compound-expression-as-lvalue extensions, which were deprecated in
12396       3.3.4 and 3.4, have been removed.
12397     * The -fwritable-strings option, which was deprecated in 3.4, has
12398       been removed.
12399     * #pragma pack() semantics have been brought closer to those used by
12400       other compilers. This also applies to C++.
12401     * Taking the address of a variable with register storage is invalid
12402       in C. GCC now issues an error instead of a warning.
12403     * Arrays of incomplete element type are invalid in C. GCC now issues
12404       an error for such arrays. Declarations such as extern struct s x[];
12405       (where struct s has not been defined) can be moved after the
12406       definition of struct s. Function parameters declared as arrays of
12407       incomplete type can instead be declared as pointers.
12408
12409  C++
12410
12411     * When compiling without optimizations (-O0), the C++ front end is
12412       much faster than in any previous versions of GCC. Independent
12413       testers have measured speed-ups up to 25% in real-world production
12414       code, compared to the 3.4 family (which was already the fastest
12415       version to date). Upgrading from older versions might show even
12416       bigger improvements.
12417     * ELF visibility attributes can now be applied to a class type, so
12418       that it affects every member function of a class at once, without
12419       having to specify each individually:
12420class __attribute__ ((visibility("hidden"))) Foo
12421{
12422   int foo1();
12423   void foo2();
12424};
12425       The syntax is deliberately similar to the __declspec() system used
12426       by Microsoft Windows based compilers, allowing cross-platform
12427       projects to easily reuse their existing macro system for denoting
12428       exports and imports. By explicitly marking internal classes never
12429       used outside a binary as hidden, one can completely avoid PLT
12430       indirection overheads during their usage by the compiler. You can
12431       find out more about the advantages of this at
12432       [6]https://www.akkadia.org/drepper/dsohowto.pdf
12433     * The -fvisibility-inlines-hidden option has been added which marks
12434       all inlineable functions as having hidden ELF visibility, thus
12435       removing their symbol and typeinfo from the exported symbol table
12436       of the output ELF binary. Using this option can reduce the exported
12437       symbol count of template-heavy code by up to 40% with no code
12438       change at all, thus notably improving link and load times for the
12439       binary as well as a reduction in size of up to 10%. Also, check the
12440       new [7]-fvisibility option.
12441     * The compiler now uses the library interface specified by the [8]C++
12442       ABI for thread-safe initialization of function-scope static
12443       variables. Most users should leave this alone, but embedded
12444       programmers may want to disable this by specifying
12445       -fno-threadsafe-statics for a small savings in code size.
12446     * Taking the address of an explicit register variable is no longer
12447       supported. Note that C++ allows taking the address of variables
12448       with register storage so this will continue to compile with a
12449       warning. For example, assuming that r0 is a machine register:
12450register int foo asm ("r0");
12451register int bar;
12452&foo; // error, no longer accepted
12453&bar; // OK, with a warning
12454     * G++ has an undocumented extension to virtual function covariancy
12455       rules that allowed the overrider to return a type that was
12456       implicitly convertable to the overridden function's return type.
12457       For instance a function returning void * could be overridden by a
12458       function returning T *. This is now deprecated and will be removed
12459       in a future release.
12460     * The G++ minimum and maximum operators (<? and >?) and their
12461       compound forms (<?=) and >?=) have been deprecated and will be
12462       removed in a future version. Code using these operators should be
12463       modified to use std::min and std::max instead.
12464     * Declaration of nested classes of class templates as friends are
12465       supported:
12466template <typename T> struct A {
12467  class B {};
12468};
12469class C {
12470  template <typename T> friend class A<T>::B;
12471};
12472       This complements the feature member functions of class templates as
12473       friends introduced in GCC 3.4.0.
12474     * When declaring a friend class using an unqualified name, classes
12475       outside the innermost non-class scope are not searched:
12476class A;
12477namespace N {
12478  class B {
12479    friend class A;   // Refer to N::A which has not been declared yet
12480                      // because name outside namespace N are not searched
12481    friend class ::A; // Refer to ::A
12482  };
12483}
12484       Hiding the friend name until declaration is still not implemented.
12485     * Friends of classes defined outside their namespace are correctly
12486       handled:
12487namespace N {
12488  class A;
12489}
12490class N::A {
12491  friend class B; // Refer to N::B in GCC 4.0.0
12492                  // but ::B in earlier versions of GCC
12493};
12494
12495    Runtime Library (libstdc++)
12496
12497     * Optimization work:
12498          + Added efficient specializations of istream functions for char
12499            and wchar_t.
12500          + Further performance tuning of strings, in particular wrt
12501            single-char append and getline.
12502          + iter_swap - and therefore most of the mutating algorithms -
12503            now makes an unqualified call to swap when the value_type of
12504            the two iterators is the same.
12505     * A large subset of the features in Technical Report 1 (TR1 for
12506       short) is experimentally delivered (i.e., no guarantees about the
12507       implementation are provided. In particular it is not promised that
12508       the library will remain link-compatible when code using TR1 is
12509       used):
12510          + General utilities such as reference_wrapper and shared_ptr.
12511          + Function objects, i.e., result_of, mem_fn, bind, function.
12512          + Support for metaprogramming.
12513          + New containers such as tuple, array, unordered_set,
12514            unordered_map, unordered_multiset, unordered_multimap.
12515     * As usual, many bugs have been fixed and LWG resolutions implemented
12516       for the first time (e.g., DR 409).
12517
12518  Java
12519
12520     * In order to prevent naming conflicts with other implementations of
12521       these tools, some GCJ binaries have been renamed:
12522          + rmic is now grmic,
12523          + rmiregistry is now grmiregistry, and
12524          + jar is now fastjar.
12525       In particular, these names were problematic for the jpackage.org
12526       packaging conventions which install symlinks in /usr/bin that point
12527       to the preferred versions of these tools.
12528     * The -findirect-dispatch argument to the compiler now works and
12529       generates code following a new "binary compatibility" ABI. Code
12530       compiled this way follows the binary compatibility rules of the
12531       Java Language Specification.
12532     * libgcj now has support for using GCJ as a JIT, using the
12533       gnu.gcj.jit family of system properties.
12534     * libgcj can now find a shared library corresponding to the bytecode
12535       representation of a class. See the documentation for the new
12536       gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system
12537       property.
12538     * There have been many improvements to the class library. Here are
12539       some highlights:
12540          + Much more of AWT and Swing exist.
12541          + Many new packages and classes were added, including
12542            java.util.regex, java.net.URI, javax.crypto,
12543            javax.crypto.interfaces, javax.crypto.spec, javax.net,
12544            javax.net.ssl, javax.security.auth,
12545            javax.security.auth.callback, javax.security.auth.login,
12546            javax.security.auth.x500, javax.security.sasl, org.ietf.jgss,
12547            javax.imageio, javax.imageio.event, javax.imageio.spi,
12548            javax.print, javax.print.attribute,
12549            javax.print.attribute.standard, javax.print.event, and
12550            javax.xml
12551          + Updated SAX and DOM, and imported GNU JAXP
12552
12553  Fortran
12554
12555     * A new [9]Fortran front end has replaced the aging GNU Fortran 77
12556       front end. The new front end supports Fortran 90 and Fortran 95. It
12557       may not yet be as stable as the old Fortran front end.
12558
12559  Ada
12560
12561     * Ada (with tasking and Zero Cost Exceptions) is now available on
12562       many more targets, including but not limited to: alpha-linux,
12563       hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux,
12564       s390x-linux, sparc-linux.
12565     * Some of the new Ada 2005 features are now implemented like
12566       Wide_Wide_Character and Ada.Containers.
12567     * Many bugs have been fixed, tools and documentation improved.
12568     * To compile Ada from the sources, install an older working Ada
12569       compiler and then use --enable-languages=ada at configuration time,
12570       since the Ada front end is not currently activated by default. See
12571       the [10]Installing GCC for details.
12572
12573New Targets and Target Specific Improvements
12574
12575  H8/300
12576
12577     * The frame layout has changed. In the new layout, the prologue of a
12578       function first saves registers and then allocate space for locals,
12579       resulting in an 1% improvement on code size.
12580
12581  IA-32/x86-64 (AMD64)
12582
12583     * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10,
12584       log1p, log2, logb and tan mathematical builtins (and their float
12585       and long double variants) are now implemented as inline x87
12586       intrinsics when using -ffast-math.
12587     * The ceil, floor, nearbyint, rint and trunc mathematical builtins
12588       (and their float and long double variants) are now implemented as
12589       inline x87 intrinsics when using -ffast-math.
12590     * The x87's fsincos instruction is now used automatically with
12591       -ffast-math when calculating both the sin and cos of the same
12592       argument.
12593     * Instruction selection for multiplication and division by constants
12594       has been improved.
12595
12596  IA-64
12597
12598     * Floating point division, integer division and sqrt are now inlined,
12599       resulting in significant performance improvements on some codes.
12600
12601  MIPS
12602
12603     * Division by zero checks now use conditional traps if the target
12604       processor supports them. This decreases code size by one word per
12605       division operation. The old behavior (branch and break) can be
12606       obtained either at configure time by passing --with-divide=breaks
12607       to configure or at runtime by passing -mdivide-breaks to GCC.
12608     * Support for MIPS64 paired-single instructions has been added. It is
12609       enabled by -mpaired-single and can be accessed using both the
12610       target-independent vector extensions and new MIPS-specific built-in
12611       functions.
12612     * Support for the MIPS-3D ASE has been added. It is enabled by
12613       -mips3d and provides new MIPS-3D-specific built-in functions.
12614     * The -mexplicit-relocs option now supports static n64 code (as is
12615       used, for example, in 64-bit linux kernels). -mexplicit-relocs
12616       should now be feature-complete and is enabled by default when GCC
12617       is configured to use a compatible assembler.
12618     * Support for the NEC VR4130 series has been added. This support
12619       includes the use of VR-specific instructions and a new VR4130
12620       scheduler. Full VR4130 support can be selected with -march=vr4130
12621       while code for any ISA can be tuned for the VR4130 using
12622       -mtune=vr4130. There is also a new -mvr4130-align option that
12623       produces better schedules at the cost of increased code size.
12624     * Support for the Broadcom SB-1 has been extended. There is now an
12625       SB-1 scheduler as well as support for the SB-1-specific
12626       paired-single instructions. Full SB-1 support can be selected with
12627       -march=sb1 while code for any ISA can be optimized for the SB-1
12628       using -mtune=sb1.
12629     * The compiler can now work around errata in R4000, R4400, VR4120 and
12630       VR4130 processors. These workarounds are enabled by -mfix-r4000,
12631       -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120
12632       and VR4130 workarounds need binutils 2.16 or above.
12633     * IRIX shared libraries are now installed into the standard library
12634       directories: o32 libraries go into lib/, n32 libraries go into
12635       lib32/ and n64 libraries go into lib64/.
12636     * The compiler supports a new -msym32 option. It can be used to
12637       optimize n64 code in which all symbols are known to have 32-bit
12638       values.
12639
12640  S/390 and zSeries
12641
12642     * New command-line options help to generate code intended to run in
12643       an environment where stack space is restricted, e.g. Linux kernel
12644       code:
12645          + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time
12646            warnings for single functions that require large or dynamic
12647            stack frames.
12648          + -mstack-size and -mstack-guard generate code that checks for
12649            stack overflow at run time.
12650          + -mpacked-stack generates code that reduces the stack frame
12651            size of many functions by reusing unneeded parts of the stack
12652            bias area.
12653     * The -msoft-float option now ensures that generated code never
12654       accesses floating point registers.
12655     * The s390x-ibm-tpf target now fully supports C++, including
12656       exceptions and threads.
12657     * Various changes to improve performance of the generated code have
12658       been implemented, including:
12659          + GCC now uses sibling calls where possible.
12660          + Condition code handling has been optimized, allowing GCC to
12661            omit redundant comparisons in certain cases.
12662          + The cost function guiding many optimizations has been refined
12663            to more accurately represent the z900 and z990 processors.
12664          + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW
12665            instructions are now used to avoid conditional branches in
12666            certain cases.
12667          + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to
12668            optimize address arithmetic required to access large stack
12669            frames.
12670          + GCC now makes more efficient use of memory-to-memory type
12671            instructions (MVC, CLC, ...).
12672          + More precise tracking of special register use allows better
12673            instruction scheduling, in particular of the function prologue
12674            and epilogue sequences.
12675          + The Java front end now generates inline code to implement
12676            integer division, instead of calling library routines.
12677
12678  SPARC
12679
12680     * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and
12681       -mf934 have been removed. They have been replaced with -mcpu=xxx.
12682     * The internal model used to estimate the relative cost of each
12683       instruction has been updated. It is expected to give better results
12684       on recent UltraSPARC processors.
12685     * Code generation for function prologues and epilogues has been
12686       improved, resulting in better scheduling and allowing multiple exit
12687       points in functions.
12688     * Support for Sun's Visual Instruction Set (VIS) has been enhanced.
12689       It is enabled by -mvis and provides new built-in functions for VIS
12690       instructions on UltraSPARC processors.
12691     * The option -mapp-regs has been turned on by default on Solaris too.
12692
12693  NetWare
12694
12695     * Novell NetWare (on ix86, no other hardware platform was ever really
12696       supported by this OS) has been re-enabled and the ABI supported by
12697       GCC has been brought into sync with that of MetroWerks CodeWarrior
12698       (the ABI previously supported was that of some Unix systems, which
12699       NetWare never tried to support).
12700
12701Obsolete Systems
12702
12703   Support for a number of older systems has been declared obsolete in GCC
12704   4.0. Unless there is activity to revive them, the next release of GCC
12705   will have their sources permanently removed.
12706
12707   All GCC ports for the following processor architectures have been
12708   declared obsolete:
12709     * Intel i860
12710     * Ubicom IP2022
12711     * National Semiconductor NS32K (ns32k)
12712     * Texas Instruments TMS320C[34]x
12713
12714   Also, those for some individual systems have been obsoleted:
12715     * SPARC family
12716          + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf,
12717            sparc86x-*-elf)
12718          + OpenBSD 32-bit (sparc-*-openbsd*)
12719
12720Documentation improvements
12721
12722Other significant improvements
12723
12724     * Location lists are now generated by default when compiling with
12725       debug info and optimization. Location lists provide more accurate
12726       debug info about locations of variables and they allow debugging
12727       code compiled with -fomit-frame-pointer.
12728     * The -fvisibility option has been added which allows the default ELF
12729       visibility of all symbols to be set per compilation and the new
12730       #pragma GCC visibility preprocessor command allows the setting of
12731       default ELF visibility for a region of code. Using
12732       -fvisibility=hidden especially in combination with the new
12733       -fvisibility-inlines-hidden can yield substantial improvements in
12734       output binary quality including avoiding PLT indirection overheads,
12735       reduction of the exported symbol count by up to 60% (with resultant
12736       improvements to link and load times), better scope for the
12737       optimizer to improve code and up to a 20% reduction in binary size.
12738       Using these options correctly yields a binary with a similar symbol
12739       count to a Windows DLL.
12740       Perhaps more importantly, this new feature finally allows (with
12741       careful planning) complete avoidance of symbol clashes when
12742       manually loading shared objects with RTLD_GLOBAL, thus finally
12743       solving problems many projects such as python were forced to use
12744       RTLD_LOCAL for (with its resulting issues for C++ correctness). You
12745       can find more information about using these options at
12746       [11]https://gcc.gnu.org/wiki/Visibility.
12747     __________________________________________________________________
12748
12749GCC 4.0.1
12750
12751   This is the [12]list of problem reports (PRs) from GCC's bug tracking
12752   system that are known to be fixed in the 4.0.1 release. This list might
12753   not be complete (that is, it is possible that some PRs that have been
12754   fixed are not listed here).
12755
12756GCC 4.0.2
12757
12758   This is the [13]list of problem reports (PRs) from GCC's bug tracking
12759   system that are known to be fixed in the 4.0.2 release. This list might
12760   not be complete (that is, it is possible that some PRs that have been
12761   fixed are not listed here).
12762
12763   Unfortunately, due to a release engineering failure, this release has a
12764   regression on Solaris that will affect some C++ programs. We suggest
12765   that Solaris users apply a [14]patch that corrects the problem. Users
12766   who do not wish to apply the patch should explicitly link C++ programs
12767   with the -pthreads option, even if they do not use threads. This
12768   problem has been corrected in the current 4.0 branch sources and will
12769   not be present in GCC 4.0.3.
12770
12771GCC 4.0.3
12772
12773   Starting with this release, the function getcontext is recognized by
12774   the compiler as having the same semantics as the setjmp function. In
12775   particular, the compiler will ensure that all registers are dead before
12776   calling such a function and will emit a warning about the variables
12777   that may be clobbered after the second return from the function.
12778
12779GCC 4.0.4
12780
12781   This is the [15]list of problem reports (PRs) from GCC's bug tracking
12782   system that are known to be fixed in the 4.0.4 release. This list might
12783   not be complete (that is, it is possible that some PRs that have been
12784   fixed are not listed here).
12785
12786   The 4.0.4 release is provided for those that require a high degree of
12787   binary compatibility with previous 4.0.x releases. For most users, the
12788   GCC team recommends that version 4.1.1 or later be used instead."
12789
12790
12791    For questions related to the use of GCC, please consult these web
12792    pages and the [16]GCC manuals. If that fails, the
12793    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12794    web pages and the development of GCC are welcome on our developer
12795    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
12796    archives.
12797
12798   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
12799   distribution of this entire article is permitted in any medium,
12800   provided this notice is preserved.
12801
12802   These pages are [21]maintained by the GCC team. Last modified
12803   2019-11-28[22].
12804
12805References
12806
12807   1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
12808   2. https://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html
12809   3. http://gcc.gnu.org/projects/tree-ssa/
12810   4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html
12811   5. http://gcc.gnu.org/news/sms.html
12812   6. https://www.akkadia.org/drepper/dsohowto.pdf
12813   7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility
12814   8. https://itanium-cxx-abi.github.io/cxx-abi/
12815   9. http://gcc.gnu.org/fortran/
12816  10. https://gcc.gnu.org/install/
12817  11. https://gcc.gnu.org/wiki/Visibility
12818  12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1
12819  13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2
12820  14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html
12821  15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4
12822  16. https://gcc.gnu.org/onlinedocs/
12823  17. mailto:gcc-help@gcc.gnu.org
12824  18. mailto:gcc@gcc.gnu.org
12825  19. https://gcc.gnu.org/lists.html
12826  20. https://www.fsf.org/
12827  21. https://gcc.gnu.org/about.html
12828  22. http://validator.w3.org/check/referer
12829======================================================================
12830http://gcc.gnu.org/gcc-3.4/index.html
12831                             GCC 3.4 Release Series
12832
12833   (This release series is no longer supported.)
12834
12835   May 26, 2006
12836
12837   The [1]GNU project and the GCC developers are pleased to announce the
12838   release of GCC 3.4.6.
12839
12840   This release is a bug-fix release, containing fixes for regressions in
12841   GCC 3.4.4 relative to previous releases of GCC. This is the last of the
12842   3.4.x series.
12843
12844   The GCC 3.4 release series includes numerous [2]new features,
12845   improvements, bug fixes, and other changes, thanks to an [3]amazing
12846   group of volunteers.
12847
12848Release History
12849
12850   GCC 3.4.6
12851          March 6, 2006 ([4]changes)
12852
12853   GCC 3.4.5
12854          November 30, 2005 ([5]changes)
12855
12856   GCC 3.4.4
12857          May 18, 2005 ([6]changes)
12858
12859   GCC 3.4.3
12860          November 4, 2004 ([7]changes)
12861
12862   GCC 3.4.2
12863          September 6, 2004 ([8]changes)
12864
12865   GCC 3.4.1
12866          July 1, 2004 ([9]changes)
12867
12868   GCC 3.4.0
12869          April 18, 2004 ([10]changes)
12870
12871References and Acknowledgements
12872
12873   GCC used to stand for the GNU C Compiler, but since the compiler
12874   supports several other languages aside from C, it now stands for the
12875   GNU Compiler Collection.
12876
12877   A list of [11]successful builds is updated as new information becomes
12878   available.
12879
12880   The GCC developers would like to thank the numerous people that have
12881   contributed new features, improvements, bug fixes, and other changes as
12882   well as test results to GCC. This [12]amazing group of volunteers is
12883   what makes GCC successful.
12884
12885   For additional information about GCC please refer to the [13]GCC
12886   project web site or contact the [14]GCC development mailing list.
12887
12888   To obtain GCC please use [15]our mirror sites, or [16]our version
12889   control system.
12890
12891
12892    For questions related to the use of GCC, please consult these web
12893    pages and the [17]GCC manuals. If that fails, the
12894    [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12895    web pages and the development of GCC are welcome on our developer
12896    list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
12897    archives.
12898
12899   Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
12900   distribution of this entire article is permitted in any medium,
12901   provided this notice is preserved.
12902
12903   These pages are [22]maintained by the GCC team. Last modified
12904   2020-01-14[23].
12905
12906References
12907
12908   1. http://www.gnu.org/
12909   2. http://gcc.gnu.org/gcc-3.4/changes.html
12910   3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12911   4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
12912   5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5
12913   6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4
12914   7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3
12915   8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2
12916   9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1
12917  10. http://gcc.gnu.org/gcc-3.4/changes.html
12918  11. http://gcc.gnu.org/gcc-3.4/buildstat.html
12919  12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12920  13. http://gcc.gnu.org/index.html
12921  14. mailto:gcc@gcc.gnu.org
12922  15. http://gcc.gnu.org/mirrors.html
12923  16. http://gcc.gnu.org/git.html
12924  17. https://gcc.gnu.org/onlinedocs/
12925  18. mailto:gcc-help@gcc.gnu.org
12926  19. mailto:gcc@gcc.gnu.org
12927  20. https://gcc.gnu.org/lists.html
12928  21. https://www.fsf.org/
12929  22. https://gcc.gnu.org/about.html
12930  23. http://validator.w3.org/check/referer
12931======================================================================
12932http://gcc.gnu.org/gcc-3.4/changes.html
12933                             GCC 3.4 Release Series
12934                        Changes, New Features, and Fixes
12935
12936   The final release in the 3.4 release series is [1]GCC 3.4.6. The series
12937   is now closed.
12938
12939   GCC 3.4 has [2]many improvements in the C++ front end. Before reporting
12940   a bug, please make sure it's really GCC, and not your code, that is
12941   broken.
12942
12943Caveats
12944
12945     * GNU Make is now required to build GCC.
12946     * With -nostdinc the preprocessor used to ignore both standard
12947       include paths and include paths contained in environment variables.
12948       It was neither documented nor intended that environment variable
12949       paths be ignored, so this has been corrected.
12950     * GCC no longer accepts the options -fvolatile, -fvolatile-global and
12951       -fvolatile-static. It is unlikely that they worked correctly in any
12952       3.x release.
12953     * GCC no longer ships <varargs.h>. Use <stdarg.h> instead.
12954     * Support for all the systems [3]obsoleted in GCC 3.3 has been
12955       removed from GCC 3.4. See below for a [4]list of systems which are
12956       obsoleted in this release.
12957     * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C
12958       compilers will not work.
12959     * The implementation of the [5]MIPS ABIs has changed. As a result,
12960       the code generated for certain MIPS targets will not be binary
12961       compatible with earlier releases.
12962     * In previous releases, the MIPS port had a fake "hilo" register with
12963       the user-visible name accum. This register has been removed.
12964     * The implementation of the [6]SPARC ABIs has changed. As a result,
12965       the code generated will not be binary compatible with earlier
12966       releases in certain cases.
12967     * The configure option --enable-threads=pthreads has been removed;
12968       use --enable-threads=posix instead, which should have the same
12969       effect.
12970     * Code size estimates used by inlining heuristics for C, Objective-C,
12971       C++ and Java have been redesigned significantly. As a result the
12972       parameters of -finline-insns, --param max-inline-insns-single and
12973       --param max-inline-insns-auto need to be reconsidered.
12974     * --param max-inline-slope and --param min-inline-insns have been
12975       removed; they are not needed for the new bottom-up inlining
12976       heuristics.
12977     * The new unit-at-a-time compilation scheme has several compatibility
12978       issues:
12979          + The order in which functions, variables, and top-level asm
12980            statements are emitted may have changed. Code relying on some
12981            particular ordering needs to be updated. The majority of such
12982            top-level asm statements can be replaced by section
12983            attributes.
12984          + Unreferenced static variables and functions are removed. This
12985            may result in undefined references when an asm statement
12986            refers to the variable/function directly. In that case either
12987            the variable/function shall be listed in asm statement operand
12988            or in the case of top-level asm statements the attribute used
12989            shall be used to force function/variable to be always output
12990            and considered as a possibly used by unknown code.
12991            For variables the attribute is accepted only by GCC 3.4 and
12992            newer, while for earlier versions it is sufficient to use
12993            unused to silence warnings about the variables not being
12994            referenced. To keep code portable across different GCC
12995            versions, you can use appropriate preprocessor conditionals.
12996          + Static functions now can use non-standard passing conventions
12997            that may break asm statements calling functions directly.
12998            Again the attribute used shall be used to prevent this
12999            behavior.
13000       As a temporary workaround, -fno-unit-at-a-time can be used, but
13001       this scheme may not be supported by future releases of GCC.
13002     * GCC 3.4 automatically places zero-initialized variables in the .bss
13003       section on some operating systems. Versions of GNU Emacs up to (and
13004       including) 21.3 will not work correctly when using this
13005       optimization; you can use -fno-zero-initialized-in-bss to disable
13006       it.
13007     * If GCC 3.4 is configured with --enable-threads=posix (the default
13008       on most targets that support pthreads) then _REENTRANT will be
13009       defined unconditionally by some libstdc++ headers. C++ code which
13010       relies on that macro to detect whether multi-threaded code is being
13011       compiled might change in meaning, possibly resulting in linker
13012       errors for single-threaded programs. Affected users of [7]Boost
13013       should compile single-threaded code with -DBOOST_DISABLE_THREADS.
13014       See Bugzilla for [8]more information.
13015
13016General Optimizer Improvements
13017
13018     * Usability of the profile feedback and coverage testing has been
13019       improved.
13020          + Performance of profiled programs has been improved by faster
13021            profile merging code.
13022          + Better use of the profile feedback for optimization (loop
13023            unrolling and loop peeling).
13024          + File locking support allowing fork() calls and parallel runs
13025            of profiled programs.
13026          + Coverage file format has been redesigned.
13027          + gcov coverage tool has been improved.
13028          + make profiledbootstrap available to build a faster compiler.
13029            Experiments made on i386 hardware showed an 11% speedup on -O0
13030            and a 7.5% speedup on -O2 compilation of a [9]large C++
13031            testcase.
13032          + New value profiling pass enabled via -fprofile-values
13033          + New value profile transformations pass enabled via -fvpt aims
13034            to optimize some code sequences by exploiting knowledge about
13035            value ranges or other properties of the operands. At the
13036            moment a conversion of expensive divisions into cheaper
13037            operations has been implemented.
13038          + New -fprofile-generate and -fprofile-use command-line options
13039            to simplify the use of profile feedback.
13040     * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and
13041       Java which is enabled via -funit-at-a-time (and implied by -O2). In
13042       this scheme a whole file is parsed first and optimized later. The
13043       following basic inter-procedural optimizations are implemented:
13044          + Removal of unreachable functions and variables
13045          + Discovery of local functions (functions with static linkage
13046            whose address is never taken)
13047          + On i386, these local functions use register parameter passing
13048            conventions.
13049          + Reordering of functions in topological order of the call graph
13050            to enable better propagation of optimizing hints (such as the
13051            stack alignments needed by functions) in the back end.
13052          + Call graph based out-of-order inlining heuristics which allows
13053            to limit overall compilation unit growth (--param
13054            inline-unit-growth).
13055       Overall, the unit-at-a-time scheme produces a 1.3% improvement for
13056       the SPECint2000 benchmark on the i386 architecture (AMD Athlon
13057       CPU).
13058     * More realistic code size estimates used by inlining for C,
13059       Objective-C, C++ and Java. The growth of large functions can now be
13060       limited via --param large-function-insns and --param
13061       large-function-growth.
13062     * A new cfg-level loop optimizer pass replaces the old loop unrolling
13063       pass and adds two other loop transformations -- loop peeling and
13064       loop unswitching -- and also uses the profile feedback to limit
13065       code growth. (The three optimizations are enabled by
13066       -funroll-loops, -fpeel-loops and -funswitch-loops flags,
13067       respectively).
13068       The old loop unroller still can be enabled by -fold-unroll-loops
13069       and may produce better code in some cases, especially when the
13070       webizer optimization pass is not run.
13071     * A new web construction pass enabled via -fweb (and implied by -O3)
13072       improves the quality of register allocation, CSE, first scheduling
13073       pass and some other optimization passes by avoiding re-use of
13074       pseudo registers with non-overlapping live ranges. The pass almost
13075       always improves code quality but does make debugging difficult and
13076       thus is not enabled by default by -O2
13077       The pass is especially effective as cleanup after code duplication
13078       passes, such as the loop unroller or the tracer.
13079     * Experimental implementations of superblock or trace scheduling in
13080       the second scheduling pass can be enabled via
13081       -fsched2-use-superblocks and -fsched2-use-traces, respectively.
13082
13083New Languages and Language specific improvements
13084
13085  Ada
13086
13087     * The Ada front end has been updated to include numerous bug fixes
13088       and enhancements. These include:
13089          + Improved project file support
13090          + Additional set of warnings about potential wrong code
13091          + Improved error messages
13092          + Improved code generation
13093          + Improved cross reference information
13094          + Improved inlining
13095          + Better run-time check elimination
13096          + Better error recovery
13097          + More efficient implementation of unbounded strings
13098          + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools,
13099            ...
13100          + New GNAT.xxxx packages (e.g. GNAT.Strings,
13101            GNAT.Exception_Action)
13102          + New pragmas
13103          + New -gnatS switch replacing gnatpsta
13104          + Implementation of new Ada features (in particular limited
13105            with, limited aggregates)
13106
13107  C/Objective-C/C++
13108
13109     * Precompiled headers are now supported. Precompiled headers can
13110       dramatically speed up compilation of some projects. There are some
13111       known defects in the current precompiled header implementation that
13112       will result in compiler crashes in relatively rare situations.
13113       Therefore, precompiled headers should be considered a "technology
13114       preview" in this release. Read the manual for details about how to
13115       use precompiled headers.
13116     * File handling in the preprocessor has been rewritten. GCC no longer
13117       gets confused by symlinks and hardlinks, and now has a correct
13118       implementation of #import and #pragma once. These two directives
13119       have therefore been un-deprecated.
13120     * The undocumented extension that allowed C programs to have a label
13121       at the end of a compound statement, which has been deprecated since
13122       GCC 3.0, has been removed.
13123     * The cast-as-lvalue extension has been removed for C++ and
13124       deprecated for C and Objective-C. In particular, code like this:
13125        int i;
13126        (char) i = 5;
13127
13128       or this:
13129        char *p;
13130        ((int *) p)++;
13131
13132       is no longer accepted for C++ and will not be accepted for C and
13133       Objective-C in a future version.
13134     * The conditional-expression-as-lvalue extension has been deprecated
13135       for C and Objective-C. In particular, code like this:
13136        int a, b, c;
13137        (a ? b : c) = 2;
13138
13139       will not be accepted for C and Objective-C in a future version.
13140     * The compound-expression-as-lvalue extension has been deprecated for
13141       C and Objective-C. In particular, code like this:
13142        int a, b;
13143        (a, b) = 2;
13144
13145       will not be accepted for C and Objective-C in a future version. A
13146       possible non-intrusive workaround is the following:
13147        (*(a, &b)) = 2;
13148
13149     * Several [10]built-in functions such as __builtin_popcount for
13150       counting bits, finding the highest and lowest bit in a word, and
13151       parity have been added.
13152     * The -fwritable-strings option has been deprecated and will be
13153       removed.
13154     * Many C math library functions are now recognized as built-ins and
13155       optimized.
13156     * The C, C++, and Objective-C compilers can now handle source files
13157       written in any character encoding supported by the host C library.
13158       The default input character set is taken from the current locale,
13159       and may be overridden with the -finput-charset command line option.
13160       In the future we will add support for inline encoding markers.
13161
13162  C++
13163
13164     * G++ is now much closer to full conformance to the ISO/ANSI C++
13165       standard. This means, among other things, that a lot of invalid
13166       constructs which used to be accepted in previous versions will now
13167       be rejected. It is very likely that existing C++ code will need to
13168       be fixed. This document lists some of the most common issues.
13169     * A hand-written recursive-descent C++ parser has replaced the
13170       YACC-derived C++ parser from previous GCC releases. The new parser
13171       contains much improved infrastructure needed for better parsing of
13172       C++ source codes, handling of extensions, and clean separation
13173       (where possible) between proper semantics analysis and parsing. The
13174       new parser fixes many bugs that were found in the old parser.
13175     * You must now use the typename and template keywords to disambiguate
13176       dependent names, as required by the C++ standard.
13177        struct K {
13178          typedef int mytype_t;
13179        };
13180
13181        template <class T1> struct A {
13182          template <class T2> struct B {
13183              void callme(void);
13184            };
13185
13186          template <int N> void bar(void)
13187          {
13188            // Use 'typename' to tell the parser that T1::mytype_t names
13189            //  a type. This is needed because the name is dependent (in
13190            //  this case, on template parameter T1).
13191            typename T1::mytype_t x;
13192            x = 0;
13193          }
13194        };
13195
13196        template <class T> void template_func(void)
13197        {
13198          // Use 'template' to prefix member templates within
13199          //  dependent types (a has type A<T>, which depends on
13200          //  the template parameter T).
13201          A<T> a;
13202          a.template bar<0>();
13203
13204          // Use 'template' to tell the parser that B is a nested
13205          //  template class (dependent on template parameter T), and
13206          //  'typename' because the whole A<T>::B<int> is
13207          //  the name of a type (again, dependent).
13208          typename A<T>::template B<int> b;
13209          b.callme();
13210        }
13211
13212        void non_template_func(void)
13213        {
13214          // Outside of any template class or function, no names can be
13215          //  dependent, so the use of the keyword 'typename' and 'template'
13216          //  is not needed (and actually forbidden).
13217          A<K> a;
13218          a.bar<0>();
13219          A<K>::B<float> b;
13220          b.callme();
13221        }
13222     * In a template definition, unqualified names will no longer find
13223       members of a dependent base (as specified by [temp.dep]/3 in the
13224       C++ standard). For example,
13225        template <typename T> struct B {
13226          int m;
13227          int n;
13228          int f ();
13229          int g ();
13230        };
13231        int n;
13232        int g ();
13233        template <typename T> struct C : B<T> {
13234          void h ()
13235          {
13236            m = 0; // error
13237            f ();  // error
13238            n = 0; // ::n is modified
13239            g ();  // ::g is called
13240          }
13241        };
13242       You must make the names dependent, e.g. by prefixing them with
13243       this->. Here is the corrected definition of C<T>::h,
13244        template <typename T> void C<T>::h ()
13245        {
13246          this->m = 0;
13247          this->f ();
13248          this->n = 0
13249          this->g ();
13250        }
13251       As an alternative solution (unfortunately not backwards compatible
13252       with GCC 3.3), you may use using declarations instead of this->:
13253        template <typename T> struct C : B<T> {
13254          using B<T>::m;
13255          using B<T>::f;
13256          using B<T>::n;
13257          using B<T>::g;
13258          void h ()
13259          {
13260            m = 0;
13261            f ();
13262            n = 0;
13263            g ();
13264          }
13265        };
13266     * In templates, all non-dependent names are now looked up and bound
13267       at definition time (while parsing the code), instead of later when
13268       the template is instantiated. For instance:
13269        void foo(int);
13270
13271        template <int> struct A {
13272          static void bar(void){
13273            foo('a');
13274          }
13275        };
13276
13277        void foo(char);
13278
13279        int main()
13280        {
13281          A<0>::bar();    // Calls foo(int), used to call foo(char).
13282        }
13283
13284     * In an explicit instantiation of a class template, you must use
13285       class or struct before the template-id:
13286        template <int N>
13287        class A {};
13288
13289        template A<0>;         // error, not accepted anymore
13290        template class A<0>;   // OK
13291     * The "named return value" and "implicit typename" extensions have
13292       been removed.
13293     * Default arguments in function types have been deprecated and will
13294       be removed.
13295     * ARM-style name-injection of friend declarations has been deprecated
13296       and will be removed. For example: struct S { friend void f(); };
13297       void g() { f(); } will not be accepted by future versions of G++;
13298       instead a declaration of "f" will need to be present outside of the
13299       scope of "S".
13300     * Covariant returns are implemented for all but varadic functions
13301       that require an adjustment.
13302     * When -pedantic is used, G++ now issues errors about spurious
13303       semicolons. For example,
13304        namespace N {}; // Invalid semicolon.
13305        void f() {}; // Invalid semicolon.
13306     * G++ no longer accepts attributes for a declarator after the
13307       initializer associated with that declarator. For example,
13308        X x(1) __attribute__((...));
13309       is no longer accepted. Instead, use:
13310        X x __attribute__((...)) (1);
13311     * Inside the scope of a template class, the name of the class itself
13312       can be treated as either a class or a template. So GCC used to
13313       accept the class name as argument of type template, and template
13314       template parameter. However this is not C++ standard compliant. Now
13315       the name is not treated as a valid template template argument
13316       unless you qualify the name by its scope. For example, the code
13317       below no longer compiles.
13318        template <template <class> class TT> class X {};
13319        template <class T> class Y {
13320          X<Y> x; // Invalid, Y is always a type template parameter.
13321        };
13322       The valid code for the above example is
13323          X< ::Y> x; // Valid.
13324       (Notice the space between < and : to prevent GCC to interpret this
13325       as a digraph for [.)
13326     * Friend declarations that refer to template specializations are
13327       rejected if the template has not already been declared. For
13328       example,
13329        template <typename T>
13330        class C {
13331          friend void f<> (C&);
13332        };
13333       is rejected. You must first declare f as a template,
13334        template <typename T>
13335        void f(T);
13336     * In case of friend declarations, every name used in the friend
13337       declaration must be accessible at the point of that declaration.
13338       Previous versions of G++ used to be less strict about this and
13339       allowed friend declarations for private class members, for example.
13340       See the ISO C++ Standard Committee's [11]defect report #209 for
13341       details.
13342     * Declaration of member functions of class templates as friends are
13343       supported. For example,
13344        template <typename T> struct A {
13345          void f();
13346        };
13347        class C {
13348          template <typename T> friend void A<T>::f();
13349        };
13350     * You must use template <> to introduce template specializations, as
13351       required by the standard. For example,
13352        template <typename T>
13353        struct S;
13354
13355        struct S<int> { };
13356       is rejected. You must write,
13357        template <> struct S<int> {};
13358     * G++ used to accept code like this,
13359        struct S {
13360          int h();
13361          void f(int i = g());
13362          int g(int i = h());
13363        };
13364       This behavior is not mandated by the standard. Now G++ issues an
13365       error about this code. To avoid the error, you must move the
13366       declaration of g before the declaration of f. The default arguments
13367       for g must be visible at the point where it is called.
13368     * The C++ ABI Section 3.3.3 specifications for the array construction
13369       routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return
13370       NULL when the allocator argument returns NULL. These changes are
13371       incorporated into the libstdc++ runtime library.
13372     * Using a name introduced by a typedef in a friend declaration or in
13373       an explicit instantiation is now rejected, as specified by the ISO
13374       C++ standard.
13375        class A;
13376        typedef A B;
13377        class C {
13378          friend class B;      // error, no typedef name here
13379          friend B;            // error, friend always needs class/struct/enum
13380          friend class A;      // OK
13381        };
13382
13383        template <int> class Q {};
13384        typedef Q<0> R;
13385        template class R;      // error, no typedef name here
13386        template class Q<0>;   // OK
13387     * When allocating an array with a new expression, GCC used to allow
13388       parentheses around the type name. This is actually ill-formed and
13389       it is now rejected:
13390        int* a = new (int)[10];    // error, not accepted anymore
13391        int* a = new int[10];      // OK
13392     * When binding an rvalue of class type to a reference, the copy
13393       constructor of the class must be accessible. For instance, consider
13394       the following code:
13395        class A
13396        {
13397        public:
13398          A();
13399
13400        private:
13401          A(const A&);   // private copy ctor
13402        };
13403
13404        A makeA(void);
13405        void foo(const A&);
13406
13407        void bar(void)
13408        {
13409          foo(A());       // error, copy ctor is not accessible
13410          foo(makeA());   // error, copy ctor is not accessible
13411
13412          A a1;
13413          foo(a1);        // OK, a1 is a lvalue
13414        }
13415       This might be surprising at first sight, especially since most
13416       popular compilers do not correctly implement this rule ([12]further
13417       details).
13418     * When forming a pointer to member or a pointer to member function,
13419       access checks for class visibility (public, protected, private) are
13420       now performed using the qualifying scope of the name itself. This
13421       is better explained with an example:
13422        class A
13423        {
13424        public:
13425          void pub_func();
13426        protected:
13427          void prot_func();
13428        private:
13429          void priv_func();
13430        };
13431
13432        class B : public A
13433        {
13434        public:
13435          void foo()
13436          {
13437            &A::pub_func;   // OK, pub_func is accessible through A
13438            &A::prot_func;  // error, cannot access prot_func through A
13439            &A::priv_func;  // error, cannot access priv_func through A
13440
13441            &B::pub_func;   // OK, pub_func is accessible through B
13442            &B::prot_func;  // OK, can access prot_func through B (within B)
13443            &B::priv_func;  // error, cannot access priv_func through B
13444          }
13445        };
13446
13447    Runtime Library (libstdc++)
13448
13449     * Optimization work:
13450          + Streamlined streambuf, filebuf, separate synched with C
13451            Standard I/O streambuf.
13452          + All formatted I/O now uses cached locale information.
13453          + STL optimizations (memory/speed for list, red-black trees as
13454            used by sets and maps).
13455          + More use of GCC builtins.
13456          + String optimizations (avoid contention on
13457            increment/decrement-and-test of the reference count in the
13458            empty-string object, constructor from input_iterators
13459            speedup).
13460     * Static linkage size reductions.
13461     * Large File Support (files larger than 2 GB on 32-bit systems).
13462     * Wide character and variable encoding filebuf work (UTF-8, Unicode).
13463     * Generic character traits.
13464     * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x,
13465       Solaris 2.7 and above, AIX 5.x, Irix 6.5.
13466     * The allocator class is now standard-conformant, and two additional
13467       extension allocators have been added, mt_alloc and
13468       bitmap_allocator.
13469     * PCH support: -include bits/stdc++.h (2x compile speedup).
13470     * Rewrote __cxa_demangle with support for C++ style allocators.
13471     * New debug modes for STL containers and iterators.
13472     * Testsuite rewrite: five times as many tests, plus increasingly
13473       sophisticated tests, including I/O, MT, multi-locale, wide and
13474       narrow characters.
13475     * Use current versions of GNU "autotools" for build/configuration.
13476
13477  Objective-C
13478
13479     * The Objective-C front end has been updated to include the numerous
13480       bug fixes and enhancements previously available only in Apple's
13481       version of GCC. These include:
13482          + Structured exception (@try... @catch... @finally, @throw) and
13483            synchronization (@synchronized) support. These are accessible
13484            via the -fobjc-exceptions switch; as of this writing, they may
13485            only be used in conjunction with -fnext-runtime on Mac OS X
13486            10.3 and later. See [13]Options Controlling Objective-C
13487            Dialect for more information.
13488          + An overhaul of @encode logic. The C99 _Bool and C++ bool type
13489            may now be encoded as 'B'. In addition, the back-end/codegen
13490            dependencies have been removed.
13491          + An overhaul of message dispatch construction, ensuring that
13492            the various receiver types (and casts thereof) are handled
13493            properly, and that correct diagnostics are issued.
13494          + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue"
13495            (-freplace-objc-classes) debugging modes, currently available
13496            on Mac OS X 10.3 and later. See [14]Options Controlling
13497            Objective-C Dialect for more information.
13498          + Access to optimized runtime entry points (-fno-nil-receivers )
13499            on the assumption that message receivers are never nil. This
13500            is currently available on Mac OS X 10.3 and later. See
13501            [15]Options Controlling Objective-C Dialect for more
13502            information.
13503
13504  Java
13505
13506     * Compiling a .jar file will now cause non-.class entries to be
13507       automatically compiled as resources.
13508     * libgcj has been ported to Darwin.
13509     * Jeff Sturm has adapted Jan Hubicka's call graph optimization code
13510       to gcj.
13511     * libgcj has a new gcjlib URL type; this lets URLClassLoader load
13512       code from shared libraries.
13513     * libgcj has been much more completely merged with [16]GNU Classpath.
13514     * Class loading is now much more correct; in particular the caller's
13515       class loader is now used when that is required.
13516     * [17]Eclipse 2.x will run out of the box using gij.
13517     * Parts of java.nio have been implemented. Direct and indirect
13518       buffers work, as do fundamental file and socket operations.
13519     * java.awt has been improved, though it is still not ready for
13520       general use.
13521     * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST
13522       method.
13523     * The MinGW port has matured. Enhancements include socket timeout
13524       support, thread interruption, improved Runtime.exec() handling and
13525       support for accented characters in filenames.
13526
13527  Fortran
13528
13529     * Fortran improvements are listed in the [18]Fortran documentation.
13530
13531New Targets and Target Specific Improvements
13532
13533  Alpha
13534
13535     * Several [19]built-in functions have been added such as
13536       __builtin_alpha_zap to allow utilizing the more obscure
13537       instructions of the CPU.
13538     * Parameter passing of complex arguments has changed to match the
13539       ABI. This change is incompatible with previous GCC versions, but
13540       does fix compatibility with the Tru64 compiler and several corner
13541       cases where GCC was incompatible with itself.
13542
13543  ARM
13544
13545     * Nicolas Pitre has contributed his hand-coded floating-point support
13546       code for ARM. It is both significantly smaller and faster than the
13547       existing C-based implementation, even when building applications
13548       for Thumb. The arm-elf configuration has been converted to use the
13549       new code.
13550     * Support for the Intel's iWMMXt architecture, a second generation
13551       XScale processor, has been added. Enabled at run time with the
13552       -mcpu=iwmmxt command line switch.
13553     * A new ARM target has been added: arm-wince-pe. This is similar to
13554       the arm-pe target, but it defaults to using the APCS32 ABI.
13555     * The existing ARM pipeline description has been converted to the use
13556       the [20]DFA processor pipeline model. There is not much change in
13557       code performance, but the description is now [21]easier to
13558       understand.
13559     * Support for the Cirrus EP9312 Maverick floating point co-processor
13560       added. Enabled at run time with the -mcpu=ep9312 command line
13561       switch. Note however that the multilibs to support this chip are
13562       currently disabled in gcc/config/arm/t-arm-elf, so if you want to
13563       enable their production you will have to uncomment the entries in
13564       that file.
13565
13566  H8/300
13567
13568     * Support for long long has been added.
13569     * Support for saveall attribute has been added.
13570     * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code
13571       for H8/300H and H8S, which is much faster than the previous
13572       implementation.
13573     * A lot of small performance improvements.
13574
13575  IA-32/AMD64 (x86-64)
13576
13577     * Tuning for K8 (AMD Opteron/Athlon64) core is available via
13578       -march=k8 and -mcpu=k8.
13579     * Scalar SSE code generation carefully avoids reformatting penalties,
13580       hidden dependencies and minimizes the number of uops generated on
13581       both Intel and AMD CPUs.
13582     * Vector MMX and SSE operands are now passed in registers to improve
13583       performance and match the argument passing convention used by the
13584       Intel C++ Compiler. As a result it is not possible to call
13585       functions accepting vector arguments compiled by older GCC version.
13586     * Conditional jump elimination is now more aggressive on modern CPUs.
13587     * The Athlon ports has been converted to use the DFA processor
13588       pipeline description.
13589     * Optimization of indirect tail calls is now possible in a similar
13590       fashion as direct sibcall optimization.
13591     * Further small performance improvements.
13592     * -m128bit-long-double is now less buggy.
13593     * __float128 support in 64-bit compilation.
13594     * Support for data structures exceeding 2GB in 64-bit mode.
13595     * -mcpu has been renamed to -mtune.
13596
13597  IA-64
13598
13599     * Tuning code for the Itanium 2 processor has been added. The
13600       generation of code tuned for Itanium 2 (option -mtune=itanium2) is
13601       enabled by default now. To generate code tuned for Itanium 1 the
13602       option -mtune=itanium1 should be used.
13603     * [22]DFA processor pipeline descriptions for the IA-64 processors
13604       have been added. This resulted in about 3% improvement on the
13605       SPECInt2000 benchmark for Itanium 2.
13606     * Instruction bundling for the IA-64 processors has been rewritten
13607       using the DFA pipeline hazard recognizer. It resulted in about 60%
13608       compiler speedup on the SPECInt2000 C programs.
13609
13610  M32R
13611
13612     * Support for the M32R/2 processor has been added by Renesas.
13613     * Support for an M32R GNU/Linux target and PIC code generation has
13614       been added by Renesas.
13615
13616  M68000
13617
13618     * Bernardo Innocenti (Develer S.r.l.) has contributed the
13619       m68k-uclinux target, based on former work done by Paul Dale
13620       (SnapGear Inc.). Code generation for the ColdFire processors family
13621       has been enhanced and extended to support the MCF 53xx and MCF 54xx
13622       cores, integrating former work done by Peter Barada (Motorola).
13623
13624  MIPS
13625
13626    Processor-specific changes
13627
13628     * Support for the RM7000 and RM9000 processors has been added. It can
13629       be selected using the -march compiler option and should work with
13630       any MIPS I (mips-*) or MIPS III (mips64-*) configuration.
13631     * Support for revision 2 of the MIPS32 ISA has been added. It can be
13632       selected with the command-line option -march=mips32r2.
13633     * There is a new option, -mfix-sb1, to work around certain SB-1
13634       errata.
13635
13636    Configuration
13637
13638     * It is possible to customize GCC using the following configure-time
13639       options:
13640          + --with-arch, which specifies the default value of the -march
13641            option.
13642          + --with-tune, which specifies the default value of the -mtune
13643            option.
13644          + --with-abi, which specifies the default ABI.
13645          + --with-float=soft, which tells GCC to use software floating
13646            point by default.
13647          + --with-float=hard, which tells GCC to use hardware floating
13648            point by default.
13649     * A 64-bit GNU/Linux port has been added. The associated
13650       configurations are mips64-linux-gnu and mips64el-linux-gnu.
13651     * The 32-bit GNU/Linux port now supports Java.
13652     * The IRIX 6 configuration now supports the o32 ABI and will build
13653       o32 multilibs by default. This support is compatible with both
13654       binutils and the SGI tools, but note that several features,
13655       including debugging information and DWARF2 exception handling, are
13656       only available when using the GNU assembler. Use of the GNU
13657       assembler and linker (version 2.15 or above) is strongly
13658       recommended.
13659     * The IRIX 6 configuration now supports 128-bit long doubles.
13660     * There are two new RTEMS-specific configurations, mips-rtems and
13661       mipsel-rtems.
13662     * There are two new *-elf configurations, mipsisa32r2-elf and
13663       mipsisa32r2el-elf.
13664
13665    General
13666
13667     * Several [23]ABI bugs have been fixed. Unfortunately, these changes
13668       will break binary compatibility with earlier releases.
13669     * GCC can now use explicit relocation operators when generating
13670       -mabicalls code. This behavior is controlled by -mexplicit-relocs
13671       and can have several performance benefits. For example:
13672          + It allows for more optimization of GOT accesses, including
13673            better scheduling and redundancy elimination.
13674          + It allows sibling calls to be implemented as jumps.
13675          + n32 and n64 leaf functions can use a call-clobbered global
13676            pointer instead of $28.
13677          + The code to set up $gp can be removed from functions that
13678            don't need it.
13679     * A new option, -mxgot, allows the GOT to be bigger than 64k. This
13680       option is equivalent to the assembler's -xgot option and should be
13681       used instead of -Wa,-xgot.
13682     * Frame pointer elimination is now supported when generating 64-bit
13683       MIPS16 code.
13684     * Inline block moves have been optimized to take more account of
13685       alignment information.
13686     * Many internal changes have been made to the MIPS port, mostly aimed
13687       at reducing the reliance on assembler macros.
13688
13689  PowerPC
13690
13691     * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64
13692       [24]ABI incompatibilities regarding the way parameters are passed
13693       during functions calls. These changes may result in incompatibility
13694       between code compiled with GCC 3.3 and GCC 3.4.
13695
13696    PowerPC Darwin
13697
13698     * Support for shared/dylib gcc libraries has been added. It is
13699       enabled by default on powerpc-apple-darwin7.0.0 and up.
13700     * Libgcj is enabled by default. On systems older than
13701       powerpc-apple-darwin7.0.0 you need to install dlcompat.
13702     * 128-bit IBM extended precision format support added for long
13703       double.
13704
13705    PowerPC64 GNU/Linux
13706
13707     * By default, PowerPC64 GNU/Linux now uses natural alignment of
13708       structure elements. The old four byte alignment for double, with
13709       special rules for a struct starting with a double, can be chosen
13710       with -malign-power. This change may result in incompatibility
13711       between code compiled with GCC 3.3 and GCC 3.4.
13712     * -mabi=altivec is now the default rather than -mabi=no-altivec.
13713     * 128-bit IBM extended precision format support added for long
13714       double.
13715
13716  S/390 and zSeries
13717
13718     * New command-line options allow to specify the intended execution
13719       environment for generated code:
13720          + -mesa/-mzarch allows to specify whether to generate code
13721            running in ESA/390 mode or in z/Architecture mode (this is
13722            applicable to 31-bit code only).
13723          + -march allows to specify a minimum processor architecture
13724            level (g5, g6, z900, or z990).
13725          + -mtune allows to specify which processor to tune for.
13726     * It is possible to customize GCC using the following configure-time
13727       options:
13728          + --with-mode, which specifies whether to default to assuming
13729            ESA/390 or z/Architecture mode.
13730          + --with-arch, which specifies the default value of the -march
13731            option.
13732          + --with-tune, which specifies the default value of the -mtune
13733            option.
13734     * Support for the z990 processor has been added, and can be selected
13735       using -march=z990 or -mtune=z990. This includes instruction
13736       scheduling tuned for the superscalar instruction pipeline of the
13737       z990 processor as well as support for all new instructions provided
13738       by the long-displacement facility.
13739     * Support to generate 31-bit code optimized for zSeries processors
13740       (running in ESA/390 or in z/Architecture mode) has been added. This
13741       can be selected using -march=z900 and -mzarch respectively.
13742     * Instruction scheduling for the z900 and z990 processors now uses
13743       the DFA pipeline hazard recognizer.
13744     * GCC no longer generates code to maintain a stack backchain,
13745       previously used to generate stack backtraces for debugging
13746       purposes. As replacement that does not incur runtime overhead,
13747       DWARF-2 call frame information is provided by GCC; this is
13748       supported by GDB 6.1. The old behavior can be restored using the
13749       -mbackchain option.
13750     * The stack frame size of functions may now exceed 2 GB in 64-bit
13751       code.
13752     * A port for the 64-bit IBM TPF operating system has been added; the
13753       configuration is s390x-ibm-tpf. This configuration is supported as
13754       cross-compilation target only.
13755     * Various changes to improve the generated code have been
13756       implemented, including:
13757          + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT
13758            instructions to significantly speed up many floating-point
13759            applications.
13760          + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL
13761            WITH BORROW instructions to speed up long long arithmetic.
13762          + GCC now uses the SEARCH STRING instruction to implement
13763            strlen().
13764          + In many cases, function call overhead for 31-bit code has been
13765            reduced by placing the literal pool after the function code
13766            instead of after the function prolog.
13767          + Register 14 is no longer reserved in 64-bit code.
13768          + Handling of global register variables has been improved.
13769
13770  SPARC
13771
13772     * The option -mflat is deprecated.
13773     * Support for large (> 2GB) frames has been added to the 64-bit port.
13774     * Several [25]ABI bugs have been fixed. Unfortunately, these changes
13775       will break binary compatibility with earlier releases.
13776     * The default debugging format has been switched from STABS to
13777       DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already
13778       the default debugging format for 64-bit code on Solaris.
13779
13780  SuperH
13781
13782     * Support for the SH2E processor has been added. Enabled at run time
13783       with the -m2e command line switch, or at configure time by
13784       specifying sh2e as the machine part of the target triple.
13785
13786  V850
13787
13788     * Support for the Mitsubishi V850E1 processor has been added. This is
13789       a variant of the V850E processor with some additional debugging
13790       instructions.
13791
13792  Xtensa
13793
13794     * Several ABI bugs have been fixed. Unfortunately, these changes
13795       break binary compatibility with earlier releases.
13796          + For big-endian processors, the padding of aggregate return
13797            values larger than a word has changed. If the size of an
13798            aggregate return value is not a multiple of 32 bits, previous
13799            versions of GCC inserted padding in the most-significant bytes
13800            of the first return value register. Aggregates larger than a
13801            word are now padded in the least-significant bytes of the last
13802            return value register used. Aggregates smaller than a word are
13803            still padded in the most-significant bytes. The return value
13804            padding has not changed for little-endian processors.
13805          + Function arguments with 16-byte alignment are now properly
13806            aligned.
13807          + The implementation of the va_list type has changed. A va_list
13808            value created by va_start from a previous release cannot be
13809            used with va_arg from this release, or vice versa.
13810     * More processor configuration options for Xtensa processors are
13811       supported:
13812          + the ABS instruction is now optional;
13813          + the ADDX* and SUBX* instructions are now optional;
13814          + an experimental CONST16 instruction can be used to synthesize
13815            constants instead of loading them from constant pools.
13816       These and other Xtensa processor configuration options can no
13817       longer be enabled or disabled by command-line options; the
13818       processor configuration must be specified by the xtensa-config.h
13819       header file when building GCC. Additionally, the
13820       -mno-serialize-volatile option is no longer supported.
13821
13822Obsolete Systems
13823
13824   Support for a number of older systems has been declared obsolete in GCC
13825   3.4. Unless there is activity to revive them, the next release of GCC
13826   will have their sources permanently removed.
13827
13828   All configurations of the following processor architectures have been
13829   declared obsolete:
13830     * Mitsubishi D30V, d30v-*
13831     * AT&T DSP1600 and DSP1610, dsp16xx-*
13832     * Intel 80960, i960
13833
13834   Also, some individual systems have been obsoleted:
13835     * ARM Family
13836          + Support for generating code for operation in APCS/26 mode
13837            (-mapcs-26).
13838     * IBM ESA/390
13839          + "Bigfoot" port, i370-*. (The other port, s390-*, is actively
13840            maintained and supported.)
13841     * Intel 386 family
13842          + MOSS, i?86-moss-msdos and i?86-*-moss*
13843          + NCR 3000 running System V r.4, i?86-ncr-sysv4*
13844          + FreeBSD with a.out object format, i?86-*-freebsd*aout* and
13845            i?86-*-freebsd2*
13846          + GNU/Linux with a.out object format, i?86-linux*aout*
13847          + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1*
13848          + Interix versions before Interix 3, i?86-*-interix
13849          + Mach microkernel, i?86-mach*
13850          + SCO UnixWare with UDK, i?86-*-udk*
13851          + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]*
13852          + VSTa microkernel, i386-*-vsta
13853     * Motorola M68000 family
13854          + HPUX, m68k-hp-hpux* and m68000-hp-hpux*
13855          + NetBSD with a.out object format (before NetBSD 1.4),
13856            m68k-*-*-netbsd* except m68k-*-*-netbsdelf*
13857          + Generic System V r.4, m68k-*-sysv4*
13858     * VAX
13859          + Generic VAX, vax-*-* (This is generic VAX only; we have not
13860            obsoleted any VAX triples for specific operating systems.)
13861
13862Documentation improvements
13863
13864Other significant improvements
13865
13866     * The build system has undergone several significant cleanups.
13867       Subdirectories will only be configured if they are being built, and
13868       all subdirectory configures are run from the make command. The top
13869       level has been autoconfiscated.
13870     * Building GCC no longer writes to its source directory. This should
13871       help those wishing to share a read-only source directory over NFS
13872       or build from a CD. The exceptions to this feature are if you
13873       configure with either --enable-maintainer-mode or
13874       --enable-generated-files-in-srcdir.
13875     * The -W warning option has been renamed to -Wextra, which is more
13876       easily understood. The older spelling will be retained for
13877       backwards compatibility.
13878     * Substantial improvements in compile time have been made,
13879       particularly for non-optimizing compilations.
13880     __________________________________________________________________
13881
13882GCC 3.4.0
13883
13884  Bug Fixes
13885
13886   A vast number of bugs have been fixed in 3.4.0, too many to publish a
13887   complete list here. [26]Follow this link to query the Bugzilla database
13888   for the list of over 900 bugs fixed in 3.4.0. This is the list of all
13889   bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4
13890   regressions.
13891     __________________________________________________________________
13892
13893GCC 3.4.1
13894
13895  Bug Fixes
13896
13897   This section lists the problem reports (PRs) from GCC's bug tracking
13898   system that are known to be fixed in the 3.4.1 release. This list might
13899   not be complete (that is, it is possible that some PRs that have been
13900   fixed are not listed here).
13901
13902    Bootstrap failures
13903
13904     * [27]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler
13905       emitted - PIC related
13906     * [28]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf
13907     * [29]14760 A bug in configure.in prevents using both
13908       --program-suffix and --program-prefix
13909     * [30]14671 [hppa64] bootstrap fails: ICE in
13910       save_call_clobbered_regs, in caller_save.c
13911     * [31]15093 [alpha][Java] make bootstrap fails to configure libffi on
13912       Alpha
13913     * [32]15178 Solaris 9/x86 fails linking after stage 3
13914
13915    Multi-platform internal compiler errors (ICEs)
13916
13917     * [33]12753 (preprocessor) Memory corruption in preprocessor on bad
13918       input
13919     * [34]13985 ICE in gcc.c-torture/compile/930621-1.c
13920     * [35]14810 (c++) tree check failures with invalid code involving
13921       templates
13922     * [36]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in
13923       cp/parser.c
13924     * [37]15044 (c++) ICE on syntax error, template header
13925     * [38]15057 (c++) Compiling of conditional value throw constructs
13926       cause a segmentation violation
13927     * [39]15064 (c++) typeid of template parameter gives ICE
13928     * [40]15142 (c++) ICE when passing a string where a char* is expected
13929       in a throw statement
13930     * [41]15159 ICE in rtl_verify_flow_info_1
13931     * [42]15165 (c++) ICE in instantiate_template
13932     * [43]15193 Unary minus using pointer to V4SF vector causes
13933       -fforce-mem to exhaust all memory
13934     * [44]15209 (c++) Runs out of memory with packed structs
13935     * [45]15227 (c++) Trouble with invalid function definition
13936     * [46]15285 (c++) instantiate_type ICE when forming pointer to
13937       template function
13938     * [47]15299 (c++) ICE in resolve_overloaded_unification
13939     * [48]15329 (c++) ICE on constructor of member template
13940     * [49]15550 ICE in extract_insn, in recog.c
13941     * [50]15554 (c++) ICE in tsubst_copy, in cp/pt.c
13942     * [51]15640 (c++) ICE on invalid code in arg_assoc, in
13943       cp/name-lookup.c
13944     * [52]15666 [unit-at-a-time] Gcc abort on valid code
13945     * [53]15696 (c++) ICE with bad pointer-to-member code
13946     * [54]15701 (c++) ICE with friends and template template parameter
13947     * [55]15761 ICE in do_SUBST, in combine.c
13948     * [56]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops
13949
13950    Ada
13951
13952     * [57]14538 All RTEMS targets broken for gnat
13953
13954    C front end
13955
13956     * [58]12391 missing warning about assigning to an incomplete type
13957     * [59]14649 atan(1.0) should not be a constant expression
13958     * [60]15004 [unit-at-a-time] no warning for unused paramater in
13959       static function
13960     * [61]15749 --pedantic-errors behaves differently from --pedantic
13961       with C-compiler on GNU/Linux
13962
13963    C++ compiler and library
13964
13965     * [62]10646 non-const reference is incorrectly matched in a "const T"
13966       partial specialization
13967     * [63]12077 wcin.rdbuf()->in_avail() return value too high
13968     * [64]13598 enc_filebuf doesn't work
13969     * [65]14211 const_cast returns lvalue but should be rvalue
13970     * [66]14220 num_put::do_put() undesired float/double behavior
13971     * [67]14245 problem with user-defined allocators in std::basic_string
13972     * [68]14340 libstdc++ Debug mode: failure to convert iterator to
13973       const_iterator
13974     * [69]14600 __gnu_cxx::stdio_sync_filebuf should expose internal
13975       FILE*
13976     * [70]14668 no warning anymore for reevaluation of declaration
13977     * [71]14775 LFS (large file support) tests missing
13978     * [72]14821 Duplicate namespace alias declaration should not conflict
13979     * [73]14930 Friend declaration ignored
13980     * [74]14932 cannot use offsetof to get offsets of array elements in
13981       g++ 3.4.0
13982     * [75]14950 [non unit-at-a-time] always_inline does not mix with
13983       templates and -O0
13984     * [76]14962 g++ ignores #pragma redefine_extname
13985     * [77]14975 Segfault on low-level write error during imbue
13986     * [78]15002 Linewise stream input is unusably slow (std::string slow)
13987     * [79]15025 compiler accepts redeclaration of template as
13988       non-template
13989     * [80]15046 [arm] Math functions misdetected by cross configuration
13990     * [81]15069 a bit test on a variable of enum type is miscompiled
13991     * [82]15074 g++ -lsupc++ still links against libstdc++
13992     * [83]15083 spurious "statement has no effect" warning
13993     * [84]15096 parse error with templates and pointer to const member
13994     * [85]15287 combination of operator[] and operator .* fails in
13995       templates
13996     * [86]15317 __attribute__ unused in first parameter of constructor
13997       gives error
13998     * [87]15337 sizeof on incomplete type diagnostic
13999     * [88]15361 bitset<>::_Find_next fails
14000     * [89]15412 _GLIBCXX_ symbols symbols defined and used in different
14001       namespaces
14002     * [90]15427 valid code results in incomplete type error
14003     * [91]15471 Incorrect member pointer offsets in anonymous
14004       structs/unions
14005     * [92]15503 nested template problem
14006     * [93]15507 compiler hangs while laying out union
14007     * [94]15542 operator & and template definitions
14008     * [95]15565 SLES9: leading + sign for unsigned int with showpos
14009     * [96]15625 friend defined inside a template fails to find static
14010       function
14011     * [97]15629 Function templates, overloads, and friend name injection
14012     * [98]15742 'noreturn' attribute ignored in method of template
14013       functions.
14014     * [99]15775 Allocator::pointer consistently ignored
14015     * [100]15821 Duplicate namespace alias within namespace rejected
14016     * [101]15862 'enum yn' fails (confict with undeclared builtin)
14017     * [102]15875 rejects pointer to member in template
14018     * [103]15877 valid code using templates and anonymous enums is
14019       rejected
14020     * [104]15947 Puzzling error message for wrong destructor declaration
14021       in template class
14022     * [105]16020 cannot copy __gnu_debug::bitset
14023     * [106]16154 input iterator concept too restrictive
14024     * [107]16174 deducing top-level consts
14025
14026    Java
14027
14028     * [108]14315 Java compiler is not parallel make safe
14029
14030    Fortran
14031
14032     * [109]15151 [g77] incorrect logical i/o in 64-bit mode
14033
14034    Objective-C
14035
14036     * [110]7993 private variables cannot be shadowed in subclasses
14037
14038    Optimization bugs
14039
14040     * [111]15228 useless copies of floating point operands
14041     * [112]15345 [non-unit-at-a-time] unreferenced nested inline
14042       functions not optimized away
14043     * [113]15945 Incorrect floating point optimization
14044     * [114]15526 ftrapv aborts on 0 * (-1)
14045     * [115]14690 Miscompiled POOMA tests
14046     * [116]15112 GCC generates code to write to unchanging memory
14047
14048    Preprocessor
14049
14050     * [117]15067 Minor glitch in the source of cpp
14051
14052    Main driver program bugs
14053
14054     * [118]1963 collect2 interprets -oldstyle_liblookup as -o
14055       ldstyle_liblookup
14056
14057    x86-specific (Intel/AMD)
14058
14059     * [119]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND*
14060       section}
14061
14062    HPPA-specific
14063
14064     * [120]14782 GCC produces an unaligned data access at -O2
14065     * [121]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2
14066     * [122]15202 ICE in reload_cse_simplify_operands, in postreload.c
14067
14068    IA64-specific
14069
14070     * [123]14610 __float80 constants incorrectly emitted
14071     * [124]14813 init_array sections are initialized in the wrong order
14072     * [125]14857 GCC segfault on duplicated asm statement
14073     * [126]15598 Gcc 3.4 ICE on valid code
14074     * [127]15653 Gcc 3.4 ICE on valid code
14075
14076    MIPS-specific
14077
14078     * [128]15189 wrong filling of delay slot with -march=mips1 -G0
14079       -mno-split-addresses -mno-explicit-relocs
14080     * [129]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as
14081       2.14.91
14082     * [130]16144 Bogus reference to __divdf3 when -O1
14083     * [131]16176 Miscompilation of unaligned data in MIPS backend
14084
14085    PowerPC-specific
14086
14087     * [132]11591 ICE in gcc.dg/altivec-5.c
14088     * [133]12028 powerpc-eabispe produces bad sCOND operation
14089     * [134]14478 rs6000 geu/ltu patterns generate incorrect code
14090     * [135]14567 long double and va_arg complex args
14091     * [136]14715 Altivec stack layout may overlap gpr save with stack
14092       temps
14093     * [137]14902 (libstdc++) Stream checking functions fail when -pthread
14094       option is used.
14095     * [138]14924 Compiler ICE on valid code
14096     * [139]14960 -maltivec affects vector return with -mabi=no-altivec
14097     * [140]15106 vector varargs failure passing from altivec to
14098       non-altivec code for -m32
14099     * [141]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 &
14100       half-word operation
14101     * [142]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx
14102       and stvx
14103     * [143]15662 Segmentation fault when an exception is thrown - even if
14104       try and catch are specified
14105
14106    s390-specific
14107
14108     * [144]15054 Bad code due to overlapping stack temporaries
14109
14110    SPARC-specific
14111
14112     * [145]15783 ICE with union assignment in 64-bit mode
14113     * [146]15626 GCC 3.4 emits "ld: warning: relocation error:
14114       R_SPARC_UA32"
14115
14116    x86-64-specific
14117
14118     * [147]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64
14119     * [148]14723 Backported -march=nocona from mainline
14120     * [149]15290 __float128 failed to pass to function properly
14121
14122    Cygwin/Mingw32-specific
14123
14124     * [150]15250 Option -mms-bitfields support on GCC 3.4 is not
14125       conformant to MS layout
14126     * [151]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe
14127       worker on windows32 targets
14128
14129    Bugs specific to embedded processors
14130
14131     * [152]8309 [m68k] -m5200 produces erroneous SImode set of short
14132       varaible on stack
14133     * [153]13250 [SH] Gcc code for rotation clobbers the register, but
14134       gcc continues to use the register as if it was not clobbered
14135     * [154]13803 [coldfire] movqi operand constraints too restrictivefor
14136       TARGET_COLDFIRE
14137     * [155]14093 [SH] ICE for code when using -mhitachi option in SH
14138     * [156]14457 [m6811hc] ICE with simple c++ source
14139     * [157]14542 [m6811hc] ICE on simple source
14140     * [158]15100 [SH] cc1plus got hang-up on
14141       libstdc++-v3/testsuite/abi_check.cc
14142     * [159]15296 [CRIS] Delayed branch scheduling causing invalid code on
14143       cris-*
14144     * [160]15396 [SH] ICE with -O2 -fPIC
14145     * [161]15782 [coldfire] m68k_output_mi_thunk emits wrong code for
14146       ColdFire
14147
14148    Testsuite problems (compiler not affected)
14149
14150     * [162]11610 libstdc++ testcases 27_io/* don't work properly remotely
14151     * [163]15488 (libstdc++) possibly insufficient file permissions for
14152       executing test suite
14153     * [164]15489 (libstdc++) testsuite_files determined incorrectly
14154
14155    Documentation bugs
14156
14157     * [165]13928 (libstdc++) no whatis info in some man pages generated
14158       by doxygen
14159     * [166]14150 Ada documentation out of date
14160     * [167]14949 (c++) Need to document method visibility changes
14161     * [168]15123 libstdc++-doc: Allocators.3 manpage is empty
14162     __________________________________________________________________
14163
14164GCC 3.4.2
14165
14166  Bug Fixes
14167
14168   This section lists the problem reports (PRs) from GCC's bug tracking
14169   system that are known to be fixed in the 3.4.2 release. This list might
14170   not be complete (that is, it is possible that some PRs that have been
14171   fixed are not listed here).
14172
14173    Bootstrap failures and issues
14174
14175     * [169]16469 [mips-sgi-irix5.3] bootstrap fails in
14176       libstdc++-v3/testsuite
14177     * [170]16344 [hppa-linux-gnu] libstdc++'s PCH built by
14178       profiledbootstrap does not work with the built compiler
14179     * [171]16842 [Solaris/x86] mkheaders can not find mkheaders.conf
14180
14181    Multi-platform internal compiler errors (ICEs)
14182
14183     * [172]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in
14184       cp_parser_class_specifier, in cp/parser.c
14185     * [173]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
14186     * [174]15461 (c++) ICE due to NRV and inlining
14187     * [175]15890 (c++) ICE in c_expand_expr, in c-common.c
14188     * [176]16180 ICE: segmentation fault in RTL optimization
14189     * [177]16224 (c++) ICE in write_unscoped_name (template/namespace)
14190     * [178]16408 ICE: in delete_insn, in cfgrtl.c
14191     * [179]16529 (c++) ICE for: namespace-alias shall not be declared as
14192       the name of any other entity
14193     * [180]16698 (c++) ICE with exceptions and declaration of __cxa_throw
14194     * [181]16706 (c++) ICE in finish_member_declaration, in
14195       cp/semantics.c
14196     * [182]16810 (c++) Legal C++ program with cast gives ICE in
14197       build_ptrmemfunc
14198     * [183]16851 (c++) ICE when throwing a comma expression
14199     * [184]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c
14200     * [185]16904 (c++) ICE in finish_class_member_access_expr, in
14201       cp/typeck.c
14202     * [186]16905 (c++) ICE (segfault) with exceptions
14203     * [187]16964 (c++) ICE in cp_parser_class_specifier due to
14204       redefinition
14205     * [188]17068 (c++) ICE: tree check: expected class 'd', have 'x'
14206       (identifier_node) in dependent_template_p, in cp/pt.c
14207
14208    Preprocessor bugs
14209
14210     * [189]16366 Preprocessor option -remap causes memory corruption
14211
14212    Optimization
14213
14214     * [190]15345 unreferenced nested inline functions not optimized away
14215     * [191]16590 Incorrect execution when compiling with -O2
14216     * [192]16693 Bitwise AND is lost when used within a cast to an enum
14217       of the same precision
14218     * [193]17078 Jump into if(0) substatement fails
14219
14220    Problems in generated debug information
14221
14222     * [194]13956 incorrect stabs for nested local variables
14223
14224    C front end bugs
14225
14226     * [195]16684 GCC should not warn about redundant redeclarations of
14227       built-ins
14228
14229    C++ compiler and library
14230
14231     * [196]12658 Thread safety problems in locale::global() and
14232       locale::locale()
14233     * [197]13092 g++ accepts invalid pointer-to-member conversion
14234     * [198]15320 Excessive memory consumption
14235     * [199]16246 Incorrect template argument deduction
14236     * [200]16273 Memory exhausted when using nested classes and virtual
14237       functions
14238     * [201]16401 ostringstream in gcc 3.4.x very slow for big data
14239     * [202]16411 undefined reference to
14240       __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char>
14241       >::file()
14242     * [203]16489 G++ incorrectly rejects use of a null constant integral
14243       expression as a null constant pointer
14244     * [204]16618 offsetof fails with constant member
14245     * [205]16637 syntax error reported for valid input code
14246     * [206]16717 __attribute__((constructor)) broken in C++
14247     * [207]16813 compiler error in DEBUG version of range insertion
14248       std::map::insert
14249     * [208]16853 pointer-to-member initialization from incompatible one
14250       accepted
14251     * [209]16889 ambiguity is not detected
14252     * [210]16959 Segmentation fault in ios_base::sync_with_stdio
14253
14254    Java compiler and library
14255
14256     * [211]7587 direct threaded interpreter not thread-safe
14257     * [212]16473 ServerSocket accept() leaks file descriptors
14258     * [213]16478 Hash synchronization deadlock with finalizers
14259
14260    Alpha-specific
14261
14262     * [214]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c
14263     * [215]16974 could not split insn (ice in final_scan_insn, in
14264       final.c)
14265
14266    x86-specific
14267
14268     * [216]16298 ICE in output_operand
14269     * [217]17113 ICE with SSE2 intrinsics
14270
14271    x86-64 specific
14272
14273     * [218]14697 libstdc++ couldn't find 32bit libgcc_s
14274
14275    MIPS-specific
14276
14277     * [219]15869 [mips64] No NOP after LW (with -mips1 -O0)
14278     * [220]16325 [mips64] value profiling clobbers gp on mips
14279     * [221]16357 [mipsisa64-elf] ICE copying 7 bytes between extern
14280       char[]s
14281     * [222]16380 [mips64] Use of uninitialised register after dbra
14282       conversion
14283     * [223]16407 [mips64] Unaligned access to local variables
14284     * [224]16643 [mips64] verify_local_live_at_start ICE after
14285       crossjumping & cfgcleanup
14286
14287    ARM-specific
14288
14289     * [225]15927 THUMB -O2: strength-reduced iteration variable ends up
14290       off by 1
14291     * [226]15948 THUMB: ICE with non-commutative cbranch
14292     * [227]17019 THUMB: bad switch statement in md code for
14293       addsi3_cbranch_scratch
14294
14295    IA64-specific
14296
14297     * [228]16130 ICE on valid code: in bundling, in config/ia64/ia64.c
14298       (-mtune=merced)
14299     * [229]16142 ICE on valid code: in bundling, in config/ia64/ia64.c
14300       (-mtune=itanium)
14301     * [230]16278 Gcc failed to build Linux kernel with -mtune=merced
14302     * [231]16414 ICE on valid code: typo in comparison of asm_noperands
14303       result
14304     * [232]16445 ICE on valid code: don't count ignored insns
14305     * [233]16490 ICE (segfault) while compiling with -fprofile-use
14306     * [234]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS
14307
14308    PowerPC-specific
14309
14310     * [235]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x
14311     * [236]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions
14312       issue)
14313
14314    SPARC-specific
14315
14316     * [237]16199 ICE while compiling apache 2.0.49
14317     * [238]16416 -m64 doesn't imply -mcpu=v9 anymore
14318     * [239]16430 ICE when returning non-C aggregates larger than 16 bytes
14319
14320    Bugs specific to embedded processors
14321
14322     * [240]16379 [m32r] can't output large model function call of memcpy
14323     * [241]17093 [m32r] ICE with -msdata=use -O0
14324     * [242]17119 [m32r] ICE at switch case 0x8000
14325
14326    DJGPP-specific
14327
14328     * [243]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp
14329
14330    Alpha Tru64-specific
14331
14332     * [244]16210 libstdc++ gratuitously omits "long long" I/O
14333
14334    Testsuite, documentation issues (compiler is not affected):
14335
14336     * [245]15488 (libstdc++) possibly insufficient file permissions for
14337       executing test suite
14338     * [246]16250 ada/doctools runs makeinfo even in release tarball
14339     __________________________________________________________________
14340
14341GCC 3.4.3
14342
14343   This is the [247]list of problem reports (PRs) from GCC's bug tracking
14344   system that are known to be fixed in the 3.4.3 release. This list might
14345   not be complete (that is, it is possible that some PRs that have been
14346   fixed are not listed here).
14347
14348    Bootstrap failures
14349
14350     * [248]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1
14351     * [249]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold
14352       when undeclared
14353
14354    Internal compiler errors (ICEs) affecting multiple platforms
14355
14356     * [250]13948 (java) GCJ segmentation fault while compiling GL4Java
14357       .class files
14358     * [251]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
14359     * [252]16301 (c++) ICE when "strong" attribute is attached to a using
14360       directive
14361     * [253]16566 ICE with flexible arrays
14362     * [254]17023 ICE with nested functions in parameter declaration
14363     * [255]17027 ICE with noreturn function in loop at -O2
14364     * [256]17524 ICE in grokdeclarator, in cp/decl.c
14365     * [257]17826 (c++) ICE in cp_tree_equal
14366
14367    C and optimization bugs
14368
14369     * [258]15526 -ftrapv aborts on 0 * (-1)
14370     * [259]16999 #ident stopped working
14371     * [260]17503 quadratic behaviour in invalid_mode_change_p
14372     * [261]17581 Long long arithmetic fails inside a switch/case
14373       statement when compiled with -O2
14374     * [262]18129 -fwritable-strings doesn't work
14375
14376    C++ compiler and library bugs
14377
14378     * [263]10975 incorrect initial ostringstream::tellp()
14379     * [264]11722 Unbuffered filebuf::sgetn is slow
14380     * [265]14534 Unrecognizing static function as a template parameter
14381       when its return value is also templated
14382     * [266]15172 Copy constructor optimization in aggregate
14383       initialization
14384     * [267]15786 Bad error message for frequently occuring error.
14385     * [268]16162 Rejects valid member-template-definition
14386     * [269]16612 empty basic_strings can't live in shared memory
14387     * [270]16715 std::basic_iostream is instantiated when used, even
14388       though instantiations are already contained in libstdc++
14389     * [271]16848 code in /ext/demangle.h appears broken
14390     * [272]17132 GCC fails to eliminate function template specialization
14391       when argument deduction fails
14392     * [273]17259 One more _S_leaf incorrectly qualified with _RopeRep::
14393       in ropeimpl.h
14394     * [274]17327 use of `enumeral_type' in template type unification
14395     * [275]17393 "unused variable '._0'" warning with -Wall
14396     * [276]17501 Confusion with member templates
14397     * [277]17537 g++ not passing -lstdc++ to linker when all command line
14398       arguments are libraries
14399     * [278]17585 usage of unqualified name of static member from within
14400       class not allowed
14401     * [279]17821 Poor diagnostic for using "." instead of "->"
14402     * [280]17829 wrong error: call of overloaded function is ambiguous
14403     * [281]17851 Misleading diagnostic for invalid function declarations
14404       with undeclared types
14405     * [282]17976 Destructor is called twice
14406     * [283]18020 rejects valid definition of enum value in template
14407     * [284]18093 bogus conflict in namespace aliasing
14408     * [285]18140 C++ parser bug when using >> in templates
14409
14410    Fortran
14411
14412     * [286]17541 data statements with double precision constants fail
14413
14414    x86-specific
14415
14416     * [287]17853 -O2 ICE for MMX testcase
14417
14418    SPARC-specific
14419
14420     * [288]17245 ICE compiling gsl-1.5 statistics/lag1.c
14421
14422    Darwin-specific
14423
14424     * [289]17167 FATAL:Symbol L_foo$stub already defined.
14425
14426    AIX-specific
14427
14428     * [290]17277 could not catch an exception when specified -maix64
14429
14430    Solaris-specific
14431
14432     * [291]17505 <cmath> calls acosf(), ceilf(), and other functions
14433       missing from system libraries
14434
14435    HP/UX specific:
14436
14437     * [292]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl
14438
14439    ARM-specific
14440
14441     * [293]17384 ICE with mode attribute on structures
14442
14443    MIPS-specific
14444
14445     * [294]17770 No NOP after LWL with -mips1
14446
14447    Other embedded target specific
14448
14449     * [295]11476 [arc-elf] gcc ICE on newlib's vfprintf.c
14450     * [296]14064 [avr-elf] -fdata-sections triggers ICE
14451     * [297]14678 [m68hc11-elf] gcc ICE
14452     * [298]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__
14453     * [299]15790 [i686-coff] Alignment error building gcc with i686-coff
14454       target
14455     * [300]15886 [SH] Miscompilation with -O2 -fPIC
14456     * [301]16884 [avr-elf] [fweb related] bug while initializing
14457       variables
14458
14459    Bugs relating to debugger support
14460
14461     * [302]13841 missing debug info for _Complex function arguments
14462     * [303]15860 [big-endian targets] No DW_AT_location debug info is
14463       emitted for formal arguments to a function that uses "register"
14464       qualifiers
14465
14466    Testsuite issues (compiler not affected)
14467
14468     * [304]17465 Testsuite in libffi overrides LD_LIBRARY_PATH
14469     * [305]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH
14470     * [306]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit
14471       testsuite
14472
14473    Documentation
14474
14475     * [307]15498 typo in gcc manual: non-existing locale example en_UK,
14476       should be en_GB
14477     * [308]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap:
14478       document broken shell
14479     * [309]16406 USE_LD_AS_NEEDED undocumented
14480     __________________________________________________________________
14481
14482GCC 3.4.4
14483
14484   This is the [310]list of problem reports (PRs) from GCC's bug tracking
14485   system that are known to be fixed in the 3.4.4 release. This list might
14486   not be complete (that is, it is possible that some PRs that have been
14487   fixed are not listed here).
14488     __________________________________________________________________
14489
14490GCC 3.4.5
14491
14492   This is the [311]list of problem reports (PRs) from GCC's bug tracking
14493   system that are known to be fixed in the 3.4.5 release. This list might
14494   not be complete (that is, it is possible that some PRs that have been
14495   fixed are not listed here).
14496
14497    Bootstrap issues
14498
14499     * [312]24688 sco_math fixincl breaks math.h
14500
14501    C compiler bugs
14502
14503     * [313]17188 struct Foo { } redefinition
14504     * [314]20187 wrong code for ((unsigned char)(unsigned long
14505       long)((a?a:1)&(a*b)))?0:1)
14506     * [315]21873 infinite warning loop on bad array initializer
14507     * [316]21899 enum definition accepts values to be overriden
14508     * [317]22061 ICE in find_function_data, in function.c
14509     * [318]22308 Failure to diagnose violation of constraint 6.516p2
14510     * [319]22458 ICE on missing brace
14511     * [320]22589 ICE casting to long long
14512     * [321]24101 Segfault with preprocessed source
14513
14514    C++ compiler and library bugs
14515
14516     * [322]10611 operations on vector mode not recognized in C++
14517     * [323]13377 unexpected behavior of namespace usage directive
14518     * [324]16002 Strange error message with new parser
14519     * [325]17413 local classes as template argument
14520     * [326]17609 spurious error message after using keyword
14521     * [327]17618 ICE in cp_convert_to_pointer, in cp/cvt.c
14522     * [328]18124 ICE with invalid template template parameter
14523     * [329]18155 typedef in template declaration not rejected
14524     * [330]18177 ICE with const_cast for undeclared variable
14525     * [331]18368 C++ error message regression
14526     * [332]16378 ICE when returning a copy of a packed member
14527     * [333]18466 int ::i; accepted
14528     * [334]18512 ICE on invalid usage of template base class
14529     * [335]18454 ICE when returning undefined type
14530     * [336]18738 typename not allowed with non-dependent qualified name
14531     * [337]18803 rejects access to operator() in template
14532     * [338]19004 ICE in uses_template_parms, in cp/pt.c
14533     * [339]19208 Spurious error about variably modified type
14534     * [340]18253 bad error message / ICE for invalid template parameter
14535     * [341]19608 ICE after friend function definition in local class
14536     * [342]19884 ICE on explicit instantiation of a non-template
14537       constructor
14538     * [343]20153 ICE when C++ template function contains anonymous union
14539     * [344]20563 Infinite loop in diagnostic (and ice after error
14540       message)
14541     * [345]20789 ICE with incomplete type in template
14542     * [346]21336 Internal compiler error when using custom new operators
14543     * [347]21768 ICE in error message due to violation of coding
14544       conventions
14545     * [348]21853 constness of pointer to data member ignored
14546     * [349]21903 Default argument of template function causes a
14547       compile-time error
14548     * [350]21983 multiple diagnostics
14549     * [351]21987 New testsuite failure
14550       g++.dg/warn/conversion-function-1.C
14551     * [352]22153 ICE on invalid template specialization
14552     * [353]22172 Internal compiler error, seg fault.
14553     * [354]21286 filebuf::xsgetn vs pipes
14554     * [355]22233 ICE with wrong number of template parameters
14555     * [356]22508 ICE after invalid operator new
14556     * [357]22545 ICE with pointer to class member & user defined
14557       conversion operator
14558     * [358]23528 Wrong default allocator in ext/hash_map
14559     * [359]23550 char_traits requirements/1.cc test bad math
14560     * [360]23586 Bad diagnostic for invalid namespace-name
14561     * [361]23624 ICE in invert_truthvalue, in fold-const.c
14562     * [362]23639 Bad error message: not a member of '<declaration error>'
14563     * [363]23797 ICE on typename outside template
14564     * [364]23965 Bogus error message: no matching function for call to
14565       'foo(<type error>)'
14566     * [365]24052 &#`label_decl' not supported by dump_expr#<expression
14567       error>
14568     * [366]24580 virtual base class cause exception not to be caught
14569
14570    Problems in generated debug information
14571
14572     * [367]24267 Bad DWARF for altivec vectors
14573
14574    Optimizations issues
14575
14576     * [368]17810 ICE in verify_local_live_at_start
14577     * [369]17860 Wrong generated code for loop with varying bound
14578     * [370]21709 ICE on compile-time complex NaN
14579     * [371]21964 broken tail call at -O2 or more
14580     * [372]22167 Strange optimization bug when using -Os
14581     * [373]22619 Compilation failure for real_const_1.f and
14582       real_const_2.f90
14583     * [374]23241 Invalid code generated for comparison of uchar to 255
14584     * [375]23478 Miscompilation due to reloading of a var that is also
14585       used in EH pad
14586     * [376]24470 segmentation fault in cc1plus when compiling with -O
14587     * [377]24950 ICE in operand_subword_force
14588
14589    Precompiled headers problems
14590
14591     * [378]14400 Cannot compile qt-x11-free-3.3.0
14592     * [379]14940 PCH largefile test fails on various platforms
14593
14594    Preprocessor bugs
14595
14596     * [380]20239 ICE on empty preprocessed input
14597     * [381]15220 "gcc -E -MM -MG" reports missing system headers in
14598       source directory
14599
14600    Testsuite issues
14601
14602     * [382]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on
14603       i686-pc-linux-gnu
14604
14605    Alpha specific
14606
14607     * [383]21888 bootstrap failure with linker relaxation enabled
14608
14609    ARM specific
14610
14611     * [384]15342 [arm-linux]: ICE in verify_local_live_at_start
14612     * [385]23985 Memory aliasing information incorrect in inlined memcpy
14613
14614    ColdFile specific
14615
14616     * [386]16719 Illegal move of byte into address register causes
14617       compiler to ICE
14618
14619    HPPA specific
14620
14621     * [387]21723 ICE while building libgfortran
14622     * [388]21841 -mhp-ld/-mgnu-ld documentation
14623
14624    IA-64 specific
14625
14626     * [389]23644 IA-64 hardware models and configuration options
14627       documentation error
14628     * [390]24718 Shared libgcc not used for linking by default
14629
14630    M68000 specific
14631
14632     * [391]18421 ICE in reload_cse_simplify_operands, in postreload.c
14633
14634    MIPS specific
14635
14636     * [392]20621 ICE in change_address_1, in emit-rtl.c
14637
14638    PowerPC and PowerPC64 specific
14639
14640     * [393]18583 error on valid code: const
14641       __attribute__((altivec(vector__))) doesn't work in arrays
14642     * [394]20191 ICE in reload_cse_simplify_operands
14643     * [395]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined
14644     * [396]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set
14645     * [397]23404 gij trashes args of functions with more than 8 fp args
14646     * [398]23539 C & C++ compiler generating misaligned references
14647       regardless of compiler flags
14648     * [399]24102 floatdisf2_internal2 broken
14649     * [400]24465 -mminimal-toc miscompilation of __thread vars
14650
14651    Solaris specific
14652
14653     * [401]19933 Problem with define of HUGE_VAL in math_c99
14654     * [402]21889 Native Solaris assembler cannot grok DTP-relative debug
14655       symbols
14656
14657    SPARC specific
14658
14659     * [403]19300 PCH failures on sparc-linux
14660     * [404]20301 Assembler labels have a leading "-"
14661     * [405]20673 C PCH testsuite assembly comparison failure
14662
14663    x86 and x86_64 specific
14664
14665     * [406]18582 ICE with arrays of type V2DF
14666     * [407]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2
14667       -fsched2-use-traces
14668     * [408]21716 ICE in reg-stack.c's swap_rtx_condition
14669     * [409]24315 amd64 fails -fpeephole2
14670     __________________________________________________________________
14671
14672GCC 3.4.6
14673
14674   This is the [410]list of problem reports (PRs) from GCC's bug tracking
14675   system that are known to be fixed in the 3.4.6 release. This list might
14676   not be complete (that is, it is possible that some PRs that have been
14677   fixed are not listed here).
14678
14679
14680    For questions related to the use of GCC, please consult these web
14681    pages and the [411]GCC manuals. If that fails, the
14682    [412]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14683    web pages and the development of GCC are welcome on our developer
14684    list at [413]gcc@gcc.gnu.org. All of [414]our lists have public
14685    archives.
14686
14687   Copyright (C) [415]Free Software Foundation, Inc. Verbatim copying and
14688   distribution of this entire article is permitted in any medium,
14689   provided this notice is preserved.
14690
14691   These pages are [416]maintained by the GCC team. Last modified
14692   2019-11-28[417].
14693
14694References
14695
14696   1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
14697   2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus
14698   3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
14699   4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems
14700   5. http://gcc.gnu.org/gcc-3.4/mips-abi.html
14701   6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
14702   7. https://www.boost.org/
14703   8. https://gcc.gnu.org/PR11953
14704   9. https://gcc.gnu.org/PR8361
14705  10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other Builtins
14706  11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209
14707  12. http://gcc.gnu.org/bugs/#cxx_rvalbind
14708  13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
14709  14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
14710  15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
14711  16. http://www.gnu.org/software/classpath/
14712  17. http://www.eclipse.org/
14713  18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html
14714  19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html
14715  20. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
14716  21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html
14717  22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
14718  23. http://gcc.gnu.org/gcc-3.4/mips-abi.html
14719  24. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html
14720  25. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
14721  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
14722  27. https://gcc.gnu.org/PR10129
14723  28. https://gcc.gnu.org/PR14576
14724  29. https://gcc.gnu.org/PR14760
14725  30. https://gcc.gnu.org/PR14671
14726  31. https://gcc.gnu.org/PR15093
14727  32. https://gcc.gnu.org/PR15178
14728  33. https://gcc.gnu.org/PR12753
14729  34. https://gcc.gnu.org/PR13985
14730  35. https://gcc.gnu.org/PR14810
14731  36. https://gcc.gnu.org/PR14883
14732  37. https://gcc.gnu.org/PR15044
14733  38. https://gcc.gnu.org/PR15057
14734  39. https://gcc.gnu.org/PR15064
14735  40. https://gcc.gnu.org/PR15142
14736  41. https://gcc.gnu.org/PR15159
14737  42. https://gcc.gnu.org/PR15165
14738  43. https://gcc.gnu.org/PR15193
14739  44. https://gcc.gnu.org/PR15209
14740  45. https://gcc.gnu.org/PR15227
14741  46. https://gcc.gnu.org/PR15285
14742  47. https://gcc.gnu.org/PR15299
14743  48. https://gcc.gnu.org/PR15329
14744  49. https://gcc.gnu.org/PR15550
14745  50. https://gcc.gnu.org/PR15554
14746  51. https://gcc.gnu.org/PR15640
14747  52. https://gcc.gnu.org/PR15666
14748  53. https://gcc.gnu.org/PR15696
14749  54. https://gcc.gnu.org/PR15701
14750  55. https://gcc.gnu.org/PR15761
14751  56. https://gcc.gnu.org/PR15829
14752  57. https://gcc.gnu.org/PR14538
14753  58. https://gcc.gnu.org/PR12391
14754  59. https://gcc.gnu.org/PR14649
14755  60. https://gcc.gnu.org/PR15004
14756  61. https://gcc.gnu.org/PR15749
14757  62. https://gcc.gnu.org/PR10646
14758  63. https://gcc.gnu.org/PR12077
14759  64. https://gcc.gnu.org/PR13598
14760  65. https://gcc.gnu.org/PR14211
14761  66. https://gcc.gnu.org/PR14220
14762  67. https://gcc.gnu.org/PR14245
14763  68. https://gcc.gnu.org/PR14340
14764  69. https://gcc.gnu.org/PR14600
14765  70. https://gcc.gnu.org/PR14668
14766  71. https://gcc.gnu.org/PR14775
14767  72. https://gcc.gnu.org/PR14821
14768  73. https://gcc.gnu.org/PR14930
14769  74. https://gcc.gnu.org/PR14932
14770  75. https://gcc.gnu.org/PR14950
14771  76. https://gcc.gnu.org/PR14962
14772  77. https://gcc.gnu.org/PR14975
14773  78. https://gcc.gnu.org/PR15002
14774  79. https://gcc.gnu.org/PR15025
14775  80. https://gcc.gnu.org/PR15046
14776  81. https://gcc.gnu.org/PR15069
14777  82. https://gcc.gnu.org/PR15074
14778  83. https://gcc.gnu.org/PR15083
14779  84. https://gcc.gnu.org/PR15096
14780  85. https://gcc.gnu.org/PR15287
14781  86. https://gcc.gnu.org/PR15317
14782  87. https://gcc.gnu.org/PR15337
14783  88. https://gcc.gnu.org/PR15361
14784  89. https://gcc.gnu.org/PR15412
14785  90. https://gcc.gnu.org/PR15427
14786  91. https://gcc.gnu.org/PR15471
14787  92. https://gcc.gnu.org/PR15503
14788  93. https://gcc.gnu.org/PR15507
14789  94. https://gcc.gnu.org/PR15542
14790  95. https://gcc.gnu.org/PR15565
14791  96. https://gcc.gnu.org/PR15625
14792  97. https://gcc.gnu.org/PR15629
14793  98. https://gcc.gnu.org/PR15742
14794  99. https://gcc.gnu.org/PR15775
14795 100. https://gcc.gnu.org/PR15821
14796 101. https://gcc.gnu.org/PR15862
14797 102. https://gcc.gnu.org/PR15875
14798 103. https://gcc.gnu.org/PR15877
14799 104. https://gcc.gnu.org/PR15947
14800 105. https://gcc.gnu.org/PR16020
14801 106. https://gcc.gnu.org/PR16154
14802 107. https://gcc.gnu.org/PR16174
14803 108. https://gcc.gnu.org/PR14315
14804 109. https://gcc.gnu.org/PR15151
14805 110. https://gcc.gnu.org/PR7993
14806 111. https://gcc.gnu.org/PR15228
14807 112. https://gcc.gnu.org/PR15345
14808 113. https://gcc.gnu.org/PR15945
14809 114. https://gcc.gnu.org/PR15526
14810 115. https://gcc.gnu.org/PR14690
14811 116. https://gcc.gnu.org/PR15112
14812 117. https://gcc.gnu.org/PR15067
14813 118. https://gcc.gnu.org/PR1963
14814 119. https://gcc.gnu.org/PR15717
14815 120. https://gcc.gnu.org/PR14782
14816 121. https://gcc.gnu.org/PR14828
14817 122. https://gcc.gnu.org/PR15202
14818 123. https://gcc.gnu.org/PR14610
14819 124. https://gcc.gnu.org/PR14813
14820 125. https://gcc.gnu.org/PR14857
14821 126. https://gcc.gnu.org/PR15598
14822 127. https://gcc.gnu.org/PR15653
14823 128. https://gcc.gnu.org/PR15189
14824 129. https://gcc.gnu.org/PR15331
14825 130. https://gcc.gnu.org/PR16144
14826 131. https://gcc.gnu.org/PR16176
14827 132. https://gcc.gnu.org/PR11591
14828 133. https://gcc.gnu.org/PR12028
14829 134. https://gcc.gnu.org/PR14478
14830 135. https://gcc.gnu.org/PR14567
14831 136. https://gcc.gnu.org/PR14715
14832 137. https://gcc.gnu.org/PR14902
14833 138. https://gcc.gnu.org/PR14924
14834 139. https://gcc.gnu.org/PR14960
14835 140. https://gcc.gnu.org/PR15106
14836 141. https://gcc.gnu.org/PR16026
14837 142. https://gcc.gnu.org/PR15191
14838 143. https://gcc.gnu.org/PR15662
14839 144. https://gcc.gnu.org/PR15054
14840 145. https://gcc.gnu.org/PR15783
14841 146. https://gcc.gnu.org/PR15626
14842 147. https://gcc.gnu.org/PR14326
14843 148. https://gcc.gnu.org/PR14723
14844 149. https://gcc.gnu.org/PR15290
14845 150. https://gcc.gnu.org/PR15250
14846 151. https://gcc.gnu.org/PR15551
14847 152. https://gcc.gnu.org/PR8309
14848 153. https://gcc.gnu.org/PR13250
14849 154. https://gcc.gnu.org/PR13803
14850 155. https://gcc.gnu.org/PR14093
14851 156. https://gcc.gnu.org/PR14457
14852 157. https://gcc.gnu.org/PR14542
14853 158. https://gcc.gnu.org/PR15100
14854 159. https://gcc.gnu.org/PR15296
14855 160. https://gcc.gnu.org/PR15396
14856 161. https://gcc.gnu.org/PR15782
14857 162. https://gcc.gnu.org/PR11610
14858 163. https://gcc.gnu.org/PR15488
14859 164. https://gcc.gnu.org/PR15489
14860 165. https://gcc.gnu.org/PR13928
14861 166. https://gcc.gnu.org/PR14150
14862 167. https://gcc.gnu.org/PR14949
14863 168. https://gcc.gnu.org/PR15123
14864 169. https://gcc.gnu.org/PR16469
14865 170. https://gcc.gnu.org/PR16344
14866 171. https://gcc.gnu.org/PR16842
14867 172. https://gcc.gnu.org/PR12608
14868 173. https://gcc.gnu.org/PR14492
14869 174. https://gcc.gnu.org/PR15461
14870 175. https://gcc.gnu.org/PR15890
14871 176. https://gcc.gnu.org/PR16180
14872 177. https://gcc.gnu.org/PR16224
14873 178. https://gcc.gnu.org/PR16408
14874 179. https://gcc.gnu.org/PR16529
14875 180. https://gcc.gnu.org/PR16698
14876 181. https://gcc.gnu.org/PR16706
14877 182. https://gcc.gnu.org/PR16810
14878 183. https://gcc.gnu.org/PR16851
14879 184. https://gcc.gnu.org/PR16870
14880 185. https://gcc.gnu.org/PR16904
14881 186. https://gcc.gnu.org/PR16905
14882 187. https://gcc.gnu.org/PR16964
14883 188. https://gcc.gnu.org/PR17068
14884 189. https://gcc.gnu.org/PR16366
14885 190. https://gcc.gnu.org/PR15345
14886 191. https://gcc.gnu.org/PR16590
14887 192. https://gcc.gnu.org/PR16693
14888 193. https://gcc.gnu.org/PR17078
14889 194. https://gcc.gnu.org/PR13956
14890 195. https://gcc.gnu.org/PR16684
14891 196. https://gcc.gnu.org/PR12658
14892 197. https://gcc.gnu.org/PR13092
14893 198. https://gcc.gnu.org/PR15320
14894 199. https://gcc.gnu.org/PR16246
14895 200. https://gcc.gnu.org/PR16273
14896 201. https://gcc.gnu.org/PR16401
14897 202. https://gcc.gnu.org/PR16411
14898 203. https://gcc.gnu.org/PR16489
14899 204. https://gcc.gnu.org/PR16618
14900 205. https://gcc.gnu.org/PR16637
14901 206. https://gcc.gnu.org/PR16717
14902 207. https://gcc.gnu.org/PR16813
14903 208. https://gcc.gnu.org/PR16853
14904 209. https://gcc.gnu.org/PR16889
14905 210. https://gcc.gnu.org/PR16959
14906 211. https://gcc.gnu.org/PR7587
14907 212. https://gcc.gnu.org/PR16473
14908 213. https://gcc.gnu.org/PR16478
14909 214. https://gcc.gnu.org/PR10695
14910 215. https://gcc.gnu.org/PR16974
14911 216. https://gcc.gnu.org/PR16298
14912 217. https://gcc.gnu.org/PR17113
14913 218. https://gcc.gnu.org/PR14697
14914 219. https://gcc.gnu.org/PR15869
14915 220. https://gcc.gnu.org/PR16325
14916 221. https://gcc.gnu.org/PR16357
14917 222. https://gcc.gnu.org/PR16380
14918 223. https://gcc.gnu.org/PR16407
14919 224. https://gcc.gnu.org/PR16643
14920 225. https://gcc.gnu.org/PR15927
14921 226. https://gcc.gnu.org/PR15948
14922 227. https://gcc.gnu.org/PR17019
14923 228. https://gcc.gnu.org/PR16130
14924 229. https://gcc.gnu.org/PR16142
14925 230. https://gcc.gnu.org/PR16278
14926 231. https://gcc.gnu.org/PR16414
14927 232. https://gcc.gnu.org/PR16445
14928 233. https://gcc.gnu.org/PR16490
14929 234. https://gcc.gnu.org/PR16683
14930 235. https://gcc.gnu.org/PR16195
14931 236. https://gcc.gnu.org/PR16239
14932 237. https://gcc.gnu.org/PR16199
14933 238. https://gcc.gnu.org/PR16416
14934 239. https://gcc.gnu.org/PR16430
14935 240. https://gcc.gnu.org/PR16379
14936 241. https://gcc.gnu.org/PR17093
14937 242. https://gcc.gnu.org/PR17119
14938 243. https://gcc.gnu.org/PR15928
14939 244. https://gcc.gnu.org/PR16210
14940 245. https://gcc.gnu.org/PR15488
14941 246. https://gcc.gnu.org/PR16250
14942 247. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3
14943 248. https://gcc.gnu.org/PR17369
14944 249. https://gcc.gnu.org/PR17850
14945 250. https://gcc.gnu.org/PR13948
14946 251. https://gcc.gnu.org/PR14492
14947 252. https://gcc.gnu.org/PR16301
14948 253. https://gcc.gnu.org/PR16566
14949 254. https://gcc.gnu.org/PR17023
14950 255. https://gcc.gnu.org/PR17027
14951 256. https://gcc.gnu.org/PR17524
14952 257. https://gcc.gnu.org/PR17826
14953 258. https://gcc.gnu.org/PR15526
14954 259. https://gcc.gnu.org/PR16999
14955 260. https://gcc.gnu.org/PR17503
14956 261. https://gcc.gnu.org/PR17581
14957 262. https://gcc.gnu.org/PR18129
14958 263. https://gcc.gnu.org/PR10975
14959 264. https://gcc.gnu.org/PR11722
14960 265. https://gcc.gnu.org/PR14534
14961 266. https://gcc.gnu.org/PR15172
14962 267. https://gcc.gnu.org/PR15786
14963 268. https://gcc.gnu.org/PR16162
14964 269. https://gcc.gnu.org/PR16612
14965 270. https://gcc.gnu.org/PR16715
14966 271. https://gcc.gnu.org/PR16848
14967 272. https://gcc.gnu.org/PR17132
14968 273. https://gcc.gnu.org/PR17259
14969 274. https://gcc.gnu.org/PR17327
14970 275. https://gcc.gnu.org/PR17393
14971 276. https://gcc.gnu.org/PR17501
14972 277. https://gcc.gnu.org/PR17537
14973 278. https://gcc.gnu.org/PR17585
14974 279. https://gcc.gnu.org/PR17821
14975 280. https://gcc.gnu.org/PR17829
14976 281. https://gcc.gnu.org/PR17851
14977 282. https://gcc.gnu.org/PR17976
14978 283. https://gcc.gnu.org/PR18020
14979 284. https://gcc.gnu.org/PR18093
14980 285. https://gcc.gnu.org/PR18140
14981 286. https://gcc.gnu.org/PR17541
14982 287. https://gcc.gnu.org/PR17853
14983 288. https://gcc.gnu.org/PR17245
14984 289. https://gcc.gnu.org/PR17167
14985 290. https://gcc.gnu.org/PR17277
14986 291. https://gcc.gnu.org/PR17505
14987 292. https://gcc.gnu.org/PR17684
14988 293. https://gcc.gnu.org/PR17384
14989 294. https://gcc.gnu.org/PR17770
14990 295. https://gcc.gnu.org/PR11476
14991 296. https://gcc.gnu.org/PR14064
14992 297. https://gcc.gnu.org/PR14678
14993 298. https://gcc.gnu.org/PR15583
14994 299. https://gcc.gnu.org/PR15790
14995 300. https://gcc.gnu.org/PR15886
14996 301. https://gcc.gnu.org/PR16884
14997 302. https://gcc.gnu.org/PR13841
14998 303. https://gcc.gnu.org/PR15860
14999 304. https://gcc.gnu.org/PR17465
15000 305. https://gcc.gnu.org/PR17469
15001 306. https://gcc.gnu.org/PR18138
15002 307. https://gcc.gnu.org/PR15498
15003 308. https://gcc.gnu.org/PR15747
15004 309. https://gcc.gnu.org/PR16406
15005 310. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4
15006 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5
15007 312. https://gcc.gnu.org/PR24688
15008 313. https://gcc.gnu.org/PR17188
15009 314. https://gcc.gnu.org/PR20187
15010 315. https://gcc.gnu.org/PR21873
15011 316. https://gcc.gnu.org/PR21899
15012 317. https://gcc.gnu.org/PR22061
15013 318. https://gcc.gnu.org/PR22208
15014 319. https://gcc.gnu.org/PR22458
15015 320. https://gcc.gnu.org/PR22589
15016 321. https://gcc.gnu.org/PR24101
15017 322. https://gcc.gnu.org/PR10611
15018 323. https://gcc.gnu.org/PR13377
15019 324. https://gcc.gnu.org/PR16002
15020 325. https://gcc.gnu.org/PR17413
15021 326. https://gcc.gnu.org/PR17609
15022 327. https://gcc.gnu.org/PR17618
15023 328. https://gcc.gnu.org/PR18124
15024 329. https://gcc.gnu.org/PR18155
15025 330. https://gcc.gnu.org/PR18177
15026 331. https://gcc.gnu.org/PR18368
15027 332. https://gcc.gnu.org/PR18378
15028 333. https://gcc.gnu.org/PR18466
15029 334. https://gcc.gnu.org/PR18512
15030 335. https://gcc.gnu.org/PR18545
15031 336. https://gcc.gnu.org/PR18738
15032 337. https://gcc.gnu.org/PR18803
15033 338. https://gcc.gnu.org/PR19004
15034 339. https://gcc.gnu.org/PR19208
15035 340. https://gcc.gnu.org/PR19253
15036 341. https://gcc.gnu.org/PR19608
15037 342. https://gcc.gnu.org/PR19884
15038 343. https://gcc.gnu.org/PR20153
15039 344. https://gcc.gnu.org/PR20563
15040 345. https://gcc.gnu.org/PR20789
15041 346. https://gcc.gnu.org/PR21336
15042 347. https://gcc.gnu.org/PR21768
15043 348. https://gcc.gnu.org/PR21853
15044 349. https://gcc.gnu.org/PR21903
15045 350. https://gcc.gnu.org/PR21983
15046 351. https://gcc.gnu.org/PR21987
15047 352. https://gcc.gnu.org/PR22153
15048 353. https://gcc.gnu.org/PR22172
15049 354. https://gcc.gnu.org/PR21286
15050 355. https://gcc.gnu.org/PR22233
15051 356. https://gcc.gnu.org/PR22508
15052 357. https://gcc.gnu.org/PR22545
15053 358. https://gcc.gnu.org/PR23528
15054 359. https://gcc.gnu.org/PR23550
15055 360. https://gcc.gnu.org/PR23586
15056 361. https://gcc.gnu.org/PR23624
15057 362. https://gcc.gnu.org/PR23639
15058 363. https://gcc.gnu.org/PR23797
15059 364. https://gcc.gnu.org/PR23965
15060 365. https://gcc.gnu.org/PR24052
15061 366. https://gcc.gnu.org/PR24580
15062 367. https://gcc.gnu.org/PR24267
15063 368. https://gcc.gnu.org/PR17810
15064 369. https://gcc.gnu.org/PR17860
15065 370. https://gcc.gnu.org/PR21709
15066 371. https://gcc.gnu.org/PR21964
15067 372. https://gcc.gnu.org/PR22167
15068 373. https://gcc.gnu.org/PR22619
15069 374. https://gcc.gnu.org/PR23241
15070 375. https://gcc.gnu.org/PR23478
15071 376. https://gcc.gnu.org/PR24470
15072 377. https://gcc.gnu.org/PR24950
15073 378. https://gcc.gnu.org/PR14400
15074 379. https://gcc.gnu.org/PR14940
15075 380. https://gcc.gnu.org/PR20239
15076 381. https://gcc.gnu.org/PR15220
15077 382. https://gcc.gnu.org/PR19275
15078 383. https://gcc.gnu.org/PR21888
15079 384. https://gcc.gnu.org/PR15342
15080 385. https://gcc.gnu.org/PR23985
15081 386. https://gcc.gnu.org/PR16719
15082 387. https://gcc.gnu.org/PR21723
15083 388. https://gcc.gnu.org/PR21841
15084 389. https://gcc.gnu.org/PR23644
15085 390. https://gcc.gnu.org/PR24718
15086 391. https://gcc.gnu.org/PR18421
15087 392. https://gcc.gnu.org/PR20621
15088 393. https://gcc.gnu.org/PR18583
15089 394. https://gcc.gnu.org/PR20191
15090 395. https://gcc.gnu.org/PR22083
15091 396. https://gcc.gnu.org/PR23070
15092 397. https://gcc.gnu.org/PR23404
15093 398. https://gcc.gnu.org/PR23539
15094 399. https://gcc.gnu.org/PR24102
15095 400. https://gcc.gnu.org/PR24465
15096 401. https://gcc.gnu.org/PR19933
15097 402. https://gcc.gnu.org/PR21889
15098 403. https://gcc.gnu.org/PR19300
15099 404. https://gcc.gnu.org/PR20301
15100 405. https://gcc.gnu.org/PR20673
15101 406. https://gcc.gnu.org/PR18582
15102 407. https://gcc.gnu.org/PR19340
15103 408. https://gcc.gnu.org/PR21716
15104 409. https://gcc.gnu.org/PR24315
15105 410. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6
15106 411. https://gcc.gnu.org/onlinedocs/
15107 412. mailto:gcc-help@gcc.gnu.org
15108 413. mailto:gcc@gcc.gnu.org
15109 414. https://gcc.gnu.org/lists.html
15110 415. https://www.fsf.org/
15111 416. https://gcc.gnu.org/about.html
15112 417. http://validator.w3.org/check/referer
15113======================================================================
15114http://gcc.gnu.org/gcc-3.3/index.html
15115                             GCC 3.3 Release Series
15116
15117   (This release series is no longer supported.)
15118
15119   May 03, 2005
15120
15121   The [1]GNU project and the GCC developers are pleased to announce the
15122   release of GCC 3.3.6.
15123
15124   This release is a bug-fix release, containing fixes for regressions in
15125   GCC 3.3.5 relative to previous releases of GCC.
15126
15127   This release is the last of the series 3.3.x.
15128
15129   The GCC 3.3 release series includes numerous [2]new features,
15130   improvements, bug fixes, and other changes, thanks to an [3]amazing
15131   group of volunteers.
15132
15133Release History
15134
15135   GCC 3.3.6
15136          May 3, 2005 ([4]changes)
15137
15138   GCC 3.3.5
15139          September 30, 2004 ([5]changes)
15140
15141   GCC 3.3.4
15142          May 31, 2004 ([6]changes)
15143
15144   GCC 3.3.3
15145          February 14, 2004 ([7]changes)
15146
15147   GCC 3.3.2
15148          October 16, 2003 ([8]changes)
15149
15150   GCC 3.3.1
15151          August 8, 2003 ([9]changes)
15152
15153   GCC 3.3
15154          May 14, 2003 ([10]changes)
15155
15156References and Acknowledgements
15157
15158   GCC used to stand for the GNU C Compiler, but since the compiler
15159   supports several other languages aside from C, it now stands for the
15160   GNU Compiler Collection.
15161
15162   A list of [11]successful builds is updated as new information becomes
15163   available.
15164
15165   The GCC developers would like to thank the numerous people that have
15166   contributed new features, improvements, bug fixes, and other changes as
15167   well as test results to GCC. This [12]amazing group of volunteers is
15168   what makes GCC successful.
15169
15170   For additional information about GCC please refer to the [13]GCC
15171   project web site or contact the [14]GCC development mailing list.
15172
15173   To obtain GCC please use [15]our mirror sites, or our CVS server.
15174
15175
15176    For questions related to the use of GCC, please consult these web
15177    pages and the [16]GCC manuals. If that fails, the
15178    [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
15179    web pages and the development of GCC are welcome on our developer
15180    list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
15181    archives.
15182
15183   Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
15184   distribution of this entire article is permitted in any medium,
15185   provided this notice is preserved.
15186
15187   These pages are [21]maintained by the GCC team. Last modified
15188   2019-11-28[22].
15189
15190References
15191
15192   1. http://www.gnu.org/
15193   2. http://gcc.gnu.org/gcc-3.3/changes.html
15194   3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
15195   4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
15196   5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5
15197   6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4
15198   7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3
15199   8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2
15200   9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1
15201  10. http://gcc.gnu.org/gcc-3.3/changes.html
15202  11. http://gcc.gnu.org/gcc-3.3/buildstat.html
15203  12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
15204  13. http://gcc.gnu.org/index.html
15205  14. mailto:gcc@gcc.gnu.org
15206  15. http://gcc.gnu.org/mirrors.html
15207  16. https://gcc.gnu.org/onlinedocs/
15208  17. mailto:gcc-help@gcc.gnu.org
15209  18. mailto:gcc@gcc.gnu.org
15210  19. https://gcc.gnu.org/lists.html
15211  20. https://www.fsf.org/
15212  21. https://gcc.gnu.org/about.html
15213  22. http://validator.w3.org/check/referer
15214======================================================================
15215http://gcc.gnu.org/gcc-3.3/changes.html
15216                             GCC 3.3 Release Series
15217                        Changes, New Features, and Fixes
15218
15219   The latest release in the 3.3 release series is [1]GCC 3.3.6.
15220
15221Caveats
15222
15223     * The preprocessor no longer accepts multi-line string literals. They
15224       were deprecated in 3.0, 3.1, and 3.2.
15225     * The preprocessor no longer supports the -A- switch when appearing
15226       alone. -A- followed by an assertion is still supported.
15227     * Support for all the systems [2]obsoleted in GCC 3.1 has been
15228       removed from GCC 3.3. See below for a [3]list of systems which are
15229       obsoleted in this release.
15230     * Checking for null format arguments has been decoupled from the rest
15231       of the format checking mechanism. Programs which use the format
15232       attribute may regain this functionality by using the new [4]nonnull
15233       function attribute. Note that all functions for which GCC has a
15234       built-in format attribute, an appropriate built-in nonnull
15235       attribute is also applied.
15236     * The DWARF (version 1) debugging format has been deprecated and will
15237       be removed in a future version of GCC. Version 2 of the DWARF
15238       debugging format will continue to be supported for the foreseeable
15239       future.
15240     * The C and Objective-C compilers no longer accept the "Naming Types"
15241       extension (typedef foo = bar); it was already unavailable in C++.
15242       Code which uses it will need to be changed to use the "typeof"
15243       extension instead: typedef typeof(bar) foo. (We have removed this
15244       extension without a period of deprecation because it has caused the
15245       compiler to crash since version 3.0 and no one noticed until very
15246       recently. Thus we conclude it is not in widespread use.)
15247     * The -traditional C compiler option has been removed. It was
15248       deprecated in 3.1 and 3.2. (Traditional preprocessing remains
15249       available.) The <varargs.h> header, used for writing variadic
15250       functions in traditional C, still exists but will produce an error
15251       message if used.
15252     * GCC 3.3.1 automatically places zero-initialized variables in the
15253       .bss section on some operating systems. Versions of GNU Emacs up to
15254       (and including) 21.3 will not work correctly when using this
15255       optimization; you can use -fno-zero-initialized-in-bss to disable
15256       it.
15257
15258General Optimizer Improvements
15259
15260     * A new scheme for accurately describing processor pipelines, the
15261       [5]DFA scheduler, has been added.
15262     * Pavel Nejedly, Charles University Prague, has contributed new file
15263       format used by the edge coverage profiler (-fprofile-arcs).
15264       The new format is robust and diagnoses common mistakes where
15265       profiles from different versions (or compilations) of the program
15266       are combined resulting in nonsensical profiles and slow code to
15267       produced with profile feedback. Additionally this format allows
15268       extra data to be gathered. Currently, overall statistics are
15269       produced helping optimizers to identify hot spots of a program
15270       globally replacing the old intra-procedural scheme and resulting in
15271       better code. Note that the gcov tool from older GCC versions will
15272       not be able to parse the profiles generated by GCC 3.3 and vice
15273       versa.
15274     * Jan Hubicka, SuSE Labs, has contributed a new superblock formation
15275       pass enabled using -ftracer. This pass simplifies the control flow
15276       of functions allowing other optimizations to do better job.
15277       He also contributed the function reordering pass
15278       (-freorder-functions) to optimize function placement using profile
15279       feedback.
15280
15281New Languages and Language specific improvements
15282
15283  C/ObjC/C++
15284
15285     * The preprocessor now accepts directives within macro arguments. It
15286       processes them just as if they had not been within macro arguments.
15287     * The separate ISO and traditional preprocessors have been completely
15288       removed. The front end handles either type of preprocessed output
15289       if necessary.
15290     * In C99 mode preprocessor arithmetic is done in the precision of the
15291       target's intmax_t, as required by that standard.
15292     * The preprocessor can now copy comments inside macros to the output
15293       file when the macro is expanded. This feature, enabled using the
15294       -CC option, is intended for use by applications which place
15295       metadata or directives inside comments, such as lint.
15296     * The method of constructing the list of directories to be searched
15297       for header files has been revised. If a directory named by a -I
15298       option is a standard system include directory, the option is
15299       ignored to ensure that the default search order for system
15300       directories and the special treatment of system header files are
15301       not defeated.
15302     * A few more [6]ISO C99 features now work correctly.
15303     * A new function attribute, nonnull, has been added which allows
15304       pointer arguments to functions to be specified as requiring a
15305       non-null value. The compiler currently uses this information to
15306       issue a warning when it detects a null value passed in such an
15307       argument slot.
15308     * A new type attribute, may_alias, has been added. Accesses to
15309       objects with types with this attribute are not subjected to
15310       type-based alias analysis, but are instead assumed to be able to
15311       alias any other type of objects, just like the char type.
15312
15313  C++
15314
15315     * Type based alias analysis has been implemented for C++ aggregate
15316       types.
15317
15318  Objective-C
15319
15320     * Generate an error if Objective-C objects are passed by value in
15321       function and method calls.
15322     * When -Wselector is used, check the whole list of selectors at the
15323       end of compilation, and emit a warning if a @selector() is not
15324       known.
15325     * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime.
15326     * No longer need to include objc/objc-class.h to compile self calls
15327       in class methods (NeXT runtime only).
15328     * New -Wundeclared-selector option.
15329     * Removed selector bloating which was causing object files to be 10%
15330       bigger on average (GNU runtime only).
15331     * Using at run time @protocol() objects has been fixed in certain
15332       situations (GNU runtime only).
15333     * Type checking has been fixed and improved in many situations
15334       involving protocols.
15335
15336  Java
15337
15338     * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK
15339       1.4) API.
15340     * The JDK 1.4 assert facility has been implemented.
15341     * The bytecode interpreter is now direct threaded and thus faster.
15342
15343  Fortran
15344
15345     * Fortran improvements are listed in [7]the Fortran documentation.
15346
15347  Ada
15348
15349     * Ada tasking now works with glibc 2.3.x threading libraries.
15350
15351New Targets and Target Specific Improvements
15352
15353     * The following changes have been made to the HP-PA port:
15354          + The port now defaults to scheduling for the PA8000 series of
15355            processors.
15356          + Scheduling support for the PA7300 processor has been added.
15357          + The 32-bit port now supports weak symbols under HP-UX 11.
15358          + The handling of initializers and finalizers has been improved
15359            under HP-UX 11. The 64-bit port no longer uses collect2.
15360          + Dwarf2 EH support has been added to the 32-bit GNU/Linux port.
15361          + ABI fixes to correct the passing of small structures by value.
15362     * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to
15363       use the DFA processor pipeline description.
15364     * The following NetBSD configurations for the SuperH processor family
15365       have been added:
15366          + SH3, big-endian, sh-*-netbsdelf*
15367          + SH3, little-endian, shle-*-netbsdelf*
15368          + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd*
15369          + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd*
15370          + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd*
15371          + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd*
15372     * The following changes have been made to the IA-32/x86-64 port:
15373          + SSE2 and 3dNOW! intrinsics are now supported.
15374          + Support for thread local storage has been added to the IA-32
15375            and x86-64 ports.
15376          + The x86-64 port has been significantly improved.
15377     * The following changes have been made to the MIPS port:
15378          + All configurations now accept the -mabi switch. Note that you
15379            will need appropriate multilibs for this option to work
15380            properly.
15381          + ELF configurations will always pass an ABI flag to the
15382            assembler, except when the MIPS EABI is selected.
15383          + -mabi=64 no longer selects MIPS IV code.
15384          + The -mcpu option, which was deprecated in 3.1 and 3.2, has
15385            been removed from this release.
15386          + -march now changes the core ISA level. In previous releases,
15387            it would change the use of processor-specific extensions, but
15388            would leave the core ISA unchanged. For example, mips64-elf
15389            -march=r8000 will now generate MIPS IV code.
15390          + Under most configurations, -mipsN now acts as a synonym for
15391            -march.
15392          + There are some new preprocessor macros to describe the -march
15393            and -mtune settings. See the documentation of those options
15394            for details.
15395          + Support for the NEC VR-Series processors has been added. This
15396            includes the 54xx, 5500, and 41xx series.
15397          + Support for the Sandcraft sr71k processor has been added.
15398     * The following changes have been made to the S/390 port:
15399          + Support to build the Java runtime libraries has been added.
15400            Java is now enabled by default on s390-*-linux* and
15401            s390x-*-linux* targets.
15402          + Multilib support for the s390x-*-linux* target has been added;
15403            this allows to build 31-bit binaries using the -m31 option.
15404          + Support for thread local storage has been added.
15405          + Inline assembler code may now use the 'Q' constraint to
15406            specify memory operands without index register.
15407          + Various platform-specific performance improvements have been
15408            implemented; in particular, the compiler now uses the BRANCH
15409            ON COUNT family of instructions and makes more frequent use of
15410            the TEST UNDER MASK family of instructions.
15411     * The following changes have been made to the PowerPC port:
15412          + Support for IBM Power4 processor added.
15413          + Support for Motorola e500 SPE added.
15414          + Support for AIX 5.2 added.
15415          + Function and Data sections now supported on AIX.
15416          + Sibcall optimizations added.
15417     * The support for H8 Tiny is added to the H8/300 port with -mn.
15418
15419Obsolete Systems
15420
15421   Support for a number of older systems has been declared obsolete in GCC
15422   3.3. Unless there is activity to revive them, the next release of GCC
15423   will have their sources permanently removed.
15424
15425   All configurations of the following processor architectures have been
15426   declared obsolete:
15427     * Matsushita MN10200, mn10200-*-*
15428     * Motorola 88000, m88k-*-*
15429     * IBM ROMP, romp-*-*
15430
15431   Also, some individual systems have been obsoleted:
15432     * Alpha
15433          + Interix, alpha*-*-interix*
15434          + Linux libc1, alpha*-*-linux*libc1*
15435          + Linux ECOFF, alpha*-*-linux*ecoff*
15436     * ARM
15437          + Generic a.out, arm*-*-aout*
15438          + Conix, arm*-*-conix*
15439          + "Old ABI," arm*-*-oabi
15440          + StrongARM/COFF, strongarm-*-coff*
15441     * HPPA (PA-RISC)
15442          + Generic OSF, hppa1.0-*-osf*
15443          + Generic BSD, hppa1.0-*-bsd*
15444          + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]*
15445          + HiUX, hppa*-*-hiux*
15446          + Mach Lites, hppa*-*-lites*
15447     * Intel 386 family
15448          + Windows NT 3.x, i?86-*-win32
15449     * MC68000 family
15450          + HP systems, m68000-hp-bsd* and m68k-hp-bsd*
15451          + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and
15452            m68k-sun-mach*
15453          + AT&T systems, m68000-att-sysv*
15454          + Atari systems, m68k-atari-sysv*
15455          + Motorola systems, m68k-motorola-sysv*
15456          + NCR systems, m68k-ncr-sysv*
15457          + Plexus systems, m68k-plexus-sysv*
15458          + Commodore systems, m68k-cbm-sysv*
15459          + Citicorp TTI, m68k-tti-*
15460          + Unos, m68k-crds-unos*
15461          + Concurrent RTU, m68k-ccur-rtu*
15462          + Linux a.out, m68k-*-linux*aout*
15463          + Linux libc1, m68k-*-linux*libc1*
15464          + pSOS, m68k-*-psos*
15465     * MIPS
15466          + Generic ECOFF, mips*-*-ecoff*
15467          + SINIX, mips-sni-sysv4
15468          + Orion RTEMS, mips64orion-*-rtems*
15469     * National Semiconductor 32000
15470          + OpenBSD, ns32k-*-openbsd*
15471     * POWER (aka RS/6000) and PowerPC
15472          + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]*
15473          + Bull BOSX, rs6000-bull-bosx
15474          + Generic Mach, rs6000-*-mach*
15475          + Generic SysV, powerpc*-*-sysv*
15476          + Linux libc1, powerpc*-*-linux*libc1*
15477     * Sun SPARC
15478          + Generic a.out, sparc-*-aout*, sparclet-*-aout*,
15479            sparclite-*-aout*, and sparc86x-*-aout*
15480          + NetBSD a.out, sparc-*-netbsd*aout*
15481          + Generic BSD, sparc-*-bsd*
15482          + ChorusOS, sparc-*-chorusos*
15483          + Linux a.out, sparc-*-linux*aout*
15484          + Linux libc1, sparc-*-linux*libc1*
15485          + LynxOS, sparc-*-lynxos*
15486          + Solaris on HAL hardware, sparc-hal-solaris2*
15487          + SunOS versions 3 and 4, sparc-*-sunos[34]*
15488     * NEC V850
15489          + RTEMS, v850-*-rtems*
15490     * VAX
15491          + VMS, vax-*-vms*
15492
15493Documentation improvements
15494
15495Other significant improvements
15496
15497     * Almost all front-end dependencies in the compiler have been
15498       separated out into a set of language hooks. This should make adding
15499       a new front end clearer and easier.
15500     * One effect of removing the separate preprocessor is a small
15501       increase in the robustness of the compiler in general, and the
15502       maintainability of target descriptions. Previously target-specific
15503       built-in macros and others, such as __FAST_MATH__, had to be
15504       handled with so-called specs that were hard to maintain. Often they
15505       would fail to behave properly when conflicting options were
15506       supplied on the command line, and define macros in the user's
15507       namespace even when strict ISO compliance was requested.
15508       Integrating the preprocessor has cleanly solved these issues.
15509     * The Makefile suite now supports redirection of make install by
15510       means of the variable DESTDIR.
15511     __________________________________________________________________
15512
15513GCC 3.3
15514
15515   Detailed release notes for the GCC 3.3 release follow.
15516
15517  Bug Fixes
15518
15519    bootstrap failures
15520
15521     * [8]10140 cross compiler build failures: missing __mempcpy (DUP:
15522       [9]10198,[10]10338)
15523
15524    Internal compiler errors (multi-platform)
15525
15526     * [11]3581 large string causes segmentation fault in cc1
15527     * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler
15528     * [13]5533 (c++) ICE when processing std::accumulate(begin, end,
15529       init, invalid_op)
15530     * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out
15531     * [15]6412 (c++) ICE in retrieve_specialization
15532     * [16]6620 (c++) partial template specialization causes an ICE
15533       (segmentation fault)
15534     * [17]6663 (c++) ICE with attribute aligned
15535     * [18]7068 ICE with incomplete types
15536     * [19]7083 (c++) ICE using -gstabs with dodgy class derivation
15537     * [20]7647 (c++) ICE when data member has the name of the enclosing
15538       class
15539     * [21]7675 ICE in fixup_var_refs_1
15540     * [22]7718 'complex' template instantiation causes ICE
15541     * [23]8116 (c++) ICE in member template function
15542     * [24]8358 (ada) Ada compiler accesses freed memory, crashes
15543     * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation
15544       fault
15545     * [26]8564 (c++) ICE in find_function_data, in function.c
15546     * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c
15547     * [28]8766 (c++) ICE after failed initialization of static template
15548       variable
15549     * [29]8803 ICE in instantiate_virtual_regs_1, in function.c
15550     * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set
15551     * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class
15552       definition
15553     * [32]9216 (c++) ICE on missing template parameter
15554     * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c
15555     * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO
15556       loop
15557     * [35]9429 (c++) ICE in template instantiation with a pointered new
15558       operator
15559     * [36]9516 Internal error when using a big array
15560     * [37]9600 (c++) ICE with typedefs in template class
15561     * [38]9629 (c++) virtual inheritance segfault
15562     * [39]9672 (c++) ICE: Error reporting routines re-entered
15563     * [40]9749 (c++) ICE in write_expression on invalid function
15564       prototype
15565     * [41]9794 (fortran) ICE: floating point exception during constant
15566       folding
15567     * [42]9829 (c++) Missing colon in nested namespace usage causes ICE
15568     * [43]9916 (c++) ICE with noreturn function in ?: statement
15569     * [44]9936 ICE with local function and variable-length 2d array
15570     * [45]10262 (c++) cc1plus crashes with large generated code
15571     * [46]10278 (c++) ICE in parser for invalid code
15572     * [47]10446 (c++) ICE on definition of nonexistent member function of
15573       nested class in a class template
15574     * [48]10451 (c++) ICE in grokdeclarator on spurious mutable
15575       declaration
15576     * [49]10506 (c++) ICE in build_new at cp/init.c with
15577       -fkeep-inline-functions and multiple inheritance
15578     * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the
15579       precision of the declared type
15580
15581    Optimization bugs
15582
15583     * [51]2001 Inordinately long compile times in reload CSE regs
15584     * [52]2391 Exponential compilation time explosion in combine
15585     * [53]2960 Duplicate loop conditions even with -Os
15586     * [54]4046 redundant conditional branch
15587     * [55]6405 Loop-unrolling related performance regressions
15588     * [56]6798 very long compile time with large case-statement
15589     * [57]6871 const objects shouldn't be moved to .bss
15590     * [58]6909 problem w/ -Os on modified loop-2c.c test case
15591     * [59]7189 gcc -O2 -Wall does not print ``control reaches end of
15592       non-void function'' warning
15593     * [60]7642 optimization problem with signbit()
15594     * [61]8634 incorrect code for inlining of memcpy under -O2
15595     * [62]8750 Cygwin prolog generation erroneously emitting __alloca as
15596       regular function call
15597
15598    C front end
15599
15600     * [63]2161 long if-else cascade overflows parser stack
15601     * [64]4319 short accepted on typedef'd char
15602     * [65]8602 incorrect line numbers in warning messages when using
15603       inline functions
15604     * [66]9177 -fdump-translation-unit: C front end deletes function_decl
15605       AST nodes and breaks debugging dumps
15606     * [67]9853 miscompilation of non-constant structure initializer
15607
15608    c++ compiler and library
15609
15610     * [68]45 legal template specialization code is rejected (DUP:
15611       [69]3784)
15612     * [70]764 lookup failure: friend operator and dereferencing a pointer
15613       and templates (DUP: [71]5116)
15614     * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP:
15615       2863)
15616     * [73]3663 G++ doesn't check access control during template
15617       instantiation
15618     * [74]3797 gcc fails to emit explicit specialization of a template
15619       member
15620     * [75]3948 Two destructors are called when no copy destructor is
15621       defined (ABI change)
15622     * [76]4137 Conversion operator within template is not accepted
15623     * [77]4361 bogus ambiguity taking the address of a member template
15624     * [78]4802 g++ accepts illegal template code (access to private
15625       member; DUP: [79]5837)
15626     * [80]4803 inline function is used but never defined, and g++ does
15627       not object
15628     * [81]5094 Partial specialization cannot be friend?
15629     * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66
15630     * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run
15631       time
15632     * [84]7015 certain __asm__ constructs rejected
15633     * [85]7086 compile time regression (quadratic behavior in
15634       fixup_var_refs)
15635     * [86]7099 G++ doesn't set the noreturn attribute on std::exit and
15636       std::abort
15637     * [87]7247 copy constructor missing when inlining enabled (invalid
15638       optimization?)
15639     * [88]7441 string array initialization compilation time regression
15640       from seconds to minutes
15641     * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong
15642     * [90]7804 bad printing of floating point constant in warning message
15643     * [91]8099 Friend classes and template specializations
15644     * [92]8117 member function pointers and multiple inheritance
15645     * [93]8205 using declaration and multiple inheritance
15646     * [94]8645 unnecessary non-zero checks in stl_tree.h
15647     * [95]8724 explicit destructor call for incomplete class allowed
15648     * [96]8805 compile time regression with many member variables
15649     * [97]8691 -O3 and -fno-implicit-templates are incompatible
15650     * [98]8700 unhelpful error message for binding temp to reference
15651     * [99]8724 explicit destructor call for incomplete class allowed
15652     * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems
15653     * [101]9016 Failure to consistently constant fold "constant" C++
15654       objects
15655     * [102]9053 g++ confused about ambiguity of overloaded function
15656       templates
15657     * [103]9152 undefined virtual thunks
15658     * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out
15659     * [105]9297 data corruption due to codegen bug (when copying.)
15660     * [106]9318 i/ostream::operator>>/<<(streambuf*) broken
15661     * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf
15662     * [108]9400 bogus -Wshadow warning: shadowed declaration of this in
15663       local classes
15664     * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters
15665     * [110]9425 filebuf::pbackfail broken (DUP: [111]9439)
15666     * [112]9474 GCC freezes in compiling a weird code mixing <iostream>
15667       and <iostream.h>
15668     * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1)
15669       [114][DR 231]
15670     * [115]9555 ostream inserters fail to set badbit on exception
15671     * [116]9561 ostream inserters rethrow exception of wrong type
15672     * [117]9563 ostream::sentry returns true after a failed preparation
15673     * [118]9582 one-definition rule violation in std::allocator
15674     * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors
15675     * [120]9683 bug in initialization chains for static const variables
15676       from template classes
15677     * [121]9791 -Woverloaded-virtual reports hiding of destructor
15678     * [122]9817 collate::compare doesn't handle nul characters
15679     * [123]9825 filebuf::sputbackc breaks sbumpc
15680     * [124]9826 operator>>(basic_istream, basic_string) fails to compile
15681       with custom traits
15682     * [125]9924 Multiple using statements for builtin functions not
15683       allowed
15684     * [126]9946 destructor is not called for temporary object
15685     * [127]9964 filebuf::close() sometimes fails to close file
15686     * [128]9988 filebuf::overflow writes EOF to file
15687     * [129]10033 optimization breaks polymorphic references w/ typeid
15688       operator
15689     * [130]10097 filebuf::underflow drops characters
15690     * [131]10132 filebuf destructor can throw exceptions
15691     * [132]10180 gcc fails to warn about non-inlined function
15692     * [133]10199 method parametrized by template does not work everywhere
15693     * [134]10300 use of array-new (nothrow) in segfaults on NULL return
15694     * [135]10427 Stack corruption with variable-length automatic arrays
15695       and virtual destructors
15696     * [136]10503 Compilation never stops in fixed_type_or_null
15697
15698    Objective-C
15699
15700     * [137]5956 selectors aren't matched properly when added to the
15701       selector table
15702
15703    Fortran compiler and library
15704
15705     * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't
15706       detect
15707     * [139]3924 g77 generates code that is rejected by GAS if COFF debug
15708       info requested
15709     * [140]5634 doc: explain that configure --prefix=~/... does not work
15710     * [141]6367 multiple repeat counts confuse namelist read into array
15711     * [142]6491 Logical operations error on logicals when using
15712       -fugly-logint
15713     * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C"
15714     * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os
15715       on irix6.5
15716     * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should
15717       assume a direct access file
15718     * [146]7278 g77 "bug"; the executable misbehaves (with -O2
15719       -fno-automatic)
15720     * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows
15721     * [148]7388 Incorrect output with 0-based array of characters
15722     * [149]8587 Double complex zero ** double precision number -> NaN
15723       instead of zero
15724     * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning:
15725       unknown register name line-length-none
15726     * [151]10197 Direct access files not unformatted by default
15727
15728    Java compiler and library
15729
15730     * [152]6005 gcj fails to build rhug on alpha
15731     * [153]6389 System.getProperty("") should always throw an
15732       IllegalArgumentException
15733     * [154]6576 java.util.ResourceBundle.getResource ignores locale
15734     * [155]6652 new java.io.File("").getCanonicalFile() throws exception
15735     * [156]7060 getMethod() doesn't search super interface
15736     * [157]7073 bytecode interpreter gives wrong answer for interface
15737       getSuperclass()
15738     * [158]7180 possible bug in
15739       javax.naming.spi.NamingManager.getPlusPath()
15740     * [159]7416 java.security startup refs "GNU libgcj.security"
15741     * [160]7570 Runtime.exec with null envp: child doesn't inherit parent
15742       env (DUP: [161]7578)
15743     * [162]7611 Internal error while compiling libjava with -O
15744     * [163]7709 NullPointerException in _Jv_ResolvePoolEntry
15745     * [164]7766 ZipInputStream.available returns 0 immediately after
15746       construction
15747     * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public
15748     * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented
15749     * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens'
15750     * [168]8234 ZipInputStream chokes when InputStream.read() returns
15751       small chunks
15752     * [169]8415 reflection bug: exception info for Method
15753     * [170]8481 java.Random.nextInt(int) may return negative
15754     * [171]8593 Error reading GZIPped files with BufferedReader
15755     * [172]8759 java.beans.Introspector has no flushCaches() or
15756       flushFromCaches() methods
15757     * [173]8997 spin() calls Thread.sleep
15758     * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd
15759       instead of the root content of C:
15760     * [175]9254 java::lang::Object::wait(), threads-win32.cc returns
15761       wrong return codes
15762     * [176]9271 Severe bias in java.security.SecureRandom
15763
15764    Ada compiler and library
15765
15766     * [177]6767 make gnatlib-shared fails on -laddr2line
15767     * [178]9911 gnatmake fails to link when GCC configured with
15768       --with-sjlj-exceptions=yes
15769     * [179]10020 Can't bootstrap gcc on AIX with Ada enabled
15770     * [180]10546 Ada tasking not working on Red Hat 9
15771
15772    preprocessor
15773
15774     * [181]7029 preprocessor should ignore #warning with -M
15775
15776    ARM-specific
15777
15778     * [182]2903 [arm] Optimization bug with long long arithmetic
15779     * [183]7873 arm-linux-gcc fails when assigning address to a bit field
15780
15781    FreeBSD-specific
15782
15783     * [184]7680 float functions undefined in math.h/cmath with #define
15784       _XOPEN_SOURCE
15785
15786    HP-UX or HP-PA-specific
15787
15788     * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c
15789     * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to
15790       fputc_unlocked
15791     * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen
15792
15793    m68hc11-specific
15794
15795     * [188]6744 Bad assembler code generated: reference to pseudo
15796       register z
15797     * [189]7361 Internal compiler error in reload_cse_simplify_operands,
15798       in reload1.c
15799
15800    MIPS-specific
15801
15802     * [190]9496 [mips-linux] bug in optimizer?
15803
15804    PowerPC-specific
15805
15806     * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of
15807       space
15808     * [192]8480 reload ICEs for LAPACK code on powerpc64-linux
15809     * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg
15810     * [194]10315 [powerpc] ICE: in extract_insn, in recog.c
15811
15812    SPARC-specific
15813
15814     * [195]10267 (documentation) Wrong build instructions for
15815       *-*-solaris2*
15816
15817    x86-specific (Intel/AMD)
15818
15819     * [196]7916 ICE in instantiate_virtual_register_1
15820     * [197]7926 (c++) i486 instructions in header files make c++ programs
15821       crash on i386
15822     * [198]8555 ICE in gen_split_1231
15823     * [199]8994 ICE with -O -march=pentium4
15824     * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs
15825     * [201]9806 ICE in inline assembly with -fPIC flag
15826     * [202]10077 gcc -msse2 generates movd to move dwords between xmm
15827       regs
15828     * [203]10233 64-bit comparison only comparing bottom 32-bits
15829     * [204]10286 type-punning doesn't work with __m64 and -O
15830     * [205]10308 [x86] ICE with -O -fgcse or -O2
15831     __________________________________________________________________
15832
15833GCC 3.3.1
15834
15835  Bug Fixes
15836
15837   This section lists the problem reports (PRs) from GCC's bug tracking
15838   system that are known to be fixed in the 3.3.1 release. This list might
15839   not be complete (that is, it is possible that some PRs that have been
15840   fixed are not listed here).
15841
15842    Bootstrap failures
15843
15844     * [206]11272 [Solaris] make bootstrap fails while building libstdc++
15845
15846    Internal compiler errors (multi-platform)
15847
15848     * [207]5754 ICE on invalid nested template class
15849     * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64
15850       and --enable-checking
15851     * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c
15852     * [210]7053 (c++) ICE when declaring a function already defined as a
15853       friend method of a template class
15854     * [211]8164 (c++) ICE when using different const expressions as
15855       template parameter
15856     * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c
15857     * [213]9559 (c++) ICE with invalid initialization of a static const
15858     * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c
15859       when redeclaring a static member variable
15860     * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in
15861       dwarfout.c with -g -O -finline-functions
15862     * [216]10432 (c++) ICE in poplevel, in cp/decl.c
15863     * [217]10475 ICE in subreg_highpart_offset for code with long long
15864     * [218]10635 (c++) ICE when dereferencing an incomplete type casted
15865       from a void pointer
15866     * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while
15867       instantiating static member variables
15868     * [220]10700 ICE in copy_to_mode_reg on 64-bit targets
15869     * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c
15870     * [222]10796 (c++) ICE when defining an enum with two values: -1 and
15871       MAX_INT_64BIT
15872     * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x
15873       sched.c
15874     * [224]10939 (c++) ICE with template code
15875     * [225]10956 (c++) ICE when specializing a template member function
15876       of a template class, in tsubst, in cp/pt.c
15877     * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*()
15878       defined)
15879     * [227]11059 (c++) ICE with empty union
15880     * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with
15881       -O2 -fnon-call-exceptions
15882     * [229]11105 (c++) ICE in mangle_conv_op_name_for_type
15883     * [230]11149 (c++) ICE on error when instantiation with call function
15884       of a base type
15885     * [231]11228 (c++) ICE on new-expression using array operator new and
15886       default-initialization
15887     * [232]11282 (c++) Infinite memory usage after syntax error
15888     * [233]11301 (fortran) ICE with -fno-globals
15889     * [234]11308 (c++) ICE when using an enum type name as if it were a
15890       class or namespace
15891     * [235]11473 (c++) ICE with -gstabs when empty struct inherits from
15892       an empty struct
15893     * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR
15894     * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c:
15895       template member functions
15896
15897    Optimization bugs
15898
15899     * [238]11198 -O2 -frename-registers generates wrong code (aliasing
15900       problem)
15901     * [239]11304 Wrong code production with -fomit-frame-pointer
15902     * [240]11381 volatile memory access optimized away
15903     * [241]11536 [strength-reduce] -O2 optimization produces wrong code
15904     * [242]11557 constant folding bug generates wrong code
15905
15906    C front end
15907
15908     * [243]5897 No warning for statement after return
15909     * [244]11279 DWARF-2 output mishandles large enums
15910
15911    Preprocessor bugs
15912
15913     * [245]11022 no warning for non-compatible macro redefinition
15914
15915    C++ compiler and library
15916
15917     * [246]2330 static_cast<>() to a private base is allowed
15918     * [247]5388 Incorrect message "operands to ?: have different types"
15919     * [248]5390 Libiberty fails to demangle multi-digit template
15920       parameters
15921     * [249]7877 Incorrect parameter passing to specializations of member
15922       function templates
15923     * [250]9393 Anonymous namespaces and compiling the same file twice
15924     * [251]10032 -pedantic converts some errors to warnings
15925     * [252]10468 const typeof(x) is non-const, but only in templates
15926     * [253]10527 confused error message with "new int()" parameter
15927       initializer
15928     * [254]10679 parameter MIN_INLINE_INSNS is not honored
15929     * [255]10682 gcc chokes on a typedef for an enum inside a class
15930       template
15931     * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of
15932       0.
15933     * [257]10845 template member function (with nested template as
15934       parameter) cannot be called anymore if another unrelated template
15935       member function is defined
15936     * [258]10849 Cannot define an out-of-class specialization of a
15937       private nested template class
15938     * [259]10888 Suppress -Winline warnings for system headers
15939     * [260]10929 -Winline warns about functions for which no definition
15940       is visible
15941     * [261]10931 valid conversion static_cast<const unsigned
15942       int&>(lvalue-of-type-int) is rejected
15943     * [262]10940 Bad code with explicit specialization
15944     * [263]10968 If member function implicitly instantiated, explicit
15945       instantiation of class fails to instantiate it
15946     * [264]10990 Cannot convert with dynamic_cast<> to a private base
15947       class from within a member function
15948     * [265]11039 Bad interaction between implicit typename deprecation
15949       and friendship
15950     * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say
15951       "__unused__" instead
15952     * [267]11095 C++ iostream manipulator causes segfault when called
15953       with negative argument
15954     * [268]11098 g++ doesn't emit complete debugging information for
15955       local variables in destructors
15956     * [269]11137 GNU/Linux shared library constructors not called unless
15957       there's one global object
15958     * [270]11154 spurious ambiguity report for template class
15959       specialization
15960     * [271]11329 Compiler cannot find user defined implicit typecast
15961     * [272]11332 Spurious error with casts in ?: expression
15962     * [273]11431 static_cast behavior with subclasses when default
15963       constructor available
15964     * [274]11528 money_get facet does not accept "$.00" as valid
15965     * [275]11546 Type lookup problems in out-of-line definition of a
15966       class doubly nested from a template class
15967     * [276]11567 C++ code containing templated member function with same
15968       name as pure virtual member function results in linking failure
15969     * [277]11645 Failure to deal with using and private inheritance
15970
15971    Java compiler and library
15972
15973     * [278]5179 Qualified static field access doesn't initialize its
15974       class
15975     * [279]8204 gcj -O2 to native reorders certain instructions
15976       improperly
15977     * [280]10838 java.io.ObjectInputStream syntax error
15978     * [281]10886 The RMI registry that comes with GCJ does not work
15979       correctly
15980     * [282]11349 JNDI URL context factories not located correctly
15981
15982    x86-specific (Intel/AMD)
15983
15984     * [283]4823 ICE on inline assembly code
15985     * [284]8878 miscompilation with -O and SSE
15986     * [285]9815 (c++ library) atomicity.h - fails to compile with -O3
15987       -masm=intel
15988     * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads,
15989       in reload1.c
15990     * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2
15991     * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source
15992     * [289]11044 [x86] out of range loop instructions for FP code on K6
15993     * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE
15994       built-ins
15995     * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC"
15996       is used
15997
15998    SPARC- or Solaris- specific
15999
16000     * [292]9362 solaris 'as' dies when fed .s and "-gstabs"
16001     * [293]10142 [SPARC64] gcc produces wrong code when passing
16002       structures by value
16003     * [294]10663 New configure check aborts with Sun tools.
16004     * [295]10835 combinatorial explosion in scheduler on HyperSPARC
16005     * [296]10876 ICE in calculate_giv_inc when building KDE
16006     * [297]10955 wrong code at -O3 for structure argument in context of
16007       structure return
16008     * [298]11018 -mcpu=ultrasparc busts tar-1.13.25
16009     * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x
16010       Linux kernel
16011
16012    ia64 specific
16013
16014     * [300]10907 gcc violates the ia64 ABI (GP must be preserved)
16015     * [301]11320 scheduler bug (in machine depended reorganization pass)
16016     * [302]11599 bug with conditional and __builtin_prefetch
16017
16018    PowerPC specific
16019
16020     * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem
16021       during loop)
16022     * [304]10871 error in rs6000_stack_info save_size computation
16023     * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse
16024       cures it
16025
16026    m68k-specific
16027
16028     * [306]7594 [m68k] ICE on legal code associated with simplify-rtx
16029     * [307]10557 [m68k] ICE in subreg_offset_representable_p
16030     * [308]11054 [m68k] ICE in reg_overlap_mentioned_p
16031
16032    ARM-specific
16033
16034     * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for
16035       functions with __attribute__ ((interrupt ("IRQ")))
16036     * [310]10842 [arm] Clobbered link register is copied to pc under
16037       certain circumstances
16038     * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes
16039     * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno
16040       (3.4)
16041
16042    MIPS-specific
16043
16044     * [313]11084 ICE in propagate_one_insn, in flow.c
16045
16046    SH-specific
16047
16048     * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf
16049     * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c
16050     * [316]11096 i686-linux to sh-linux cross compiler fails to compile
16051       C++ files
16052
16053    GNU/Linux (or Hurd?) specific
16054
16055     * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3
16056
16057    UnixWare specific
16058
16059     * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare
16060       7.1.1
16061
16062    Cygwin (or mingw) specific
16063
16064     * [319]5287 ICE with dllimport attribute
16065     * [320]10148 [MingW/CygWin] Compiler dumps core
16066
16067    DJGPP specific
16068
16069     * [321]8787 GCC fails to emit .intel_syntax when invoked with
16070       -masm=intel on DJGPP
16071
16072    Darwin (and MacOS X) specific
16073
16074     * [322]10900 trampolines crash
16075
16076    Documentation
16077
16078     * [323]1607 (c++) Format attributes on methods undocumented
16079     * [324]4252 Invalid option `-fdump-translation-unit'
16080     * [325]4490 Clarify restrictions on -m96bit-long-double,
16081       -m128bit-long-double
16082     * [326]10355 document an issue with regparm attribute on some systems
16083       (e.g. Solaris)
16084     * [327]10726 (fortran) Documentation for function "IDate Intrinsic
16085       (Unix)" is wrong
16086     * [328]10805 document bug in old version of Sun assembler
16087     * [329]10815 warn against GNU binutils on AIX
16088     * [330]10877 document need for newer binutils on i?86-*-linux-gnu
16089     * [331]11280 Manual incorrect with respect to -freorder-blocks
16090     * [332]11466 Document -mlittle-endian and its restrictions for the
16091       sparc64 port
16092
16093    Testsuite bugs (compiler itself is not affected)
16094
16095     * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly
16096       report failure
16097     * [334]10810 gcc-3.3 fails make check: buffer overrun in
16098       test_demangle.c
16099     __________________________________________________________________
16100
16101GCC 3.3.2
16102
16103  Bug Fixes
16104
16105   This section lists the problem reports (PRs) from GCC's bug tracker
16106   that are known to be fixed in the 3.3.2 release. This list might not be
16107   complete (that is, it is possible that some PRs that have been fixed
16108   are not listed here).
16109
16110    Bootstrap failures and problems
16111
16112     * [335]8336 [SCO5] bootstrap config still tries to use COFF options
16113     * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with
16114       --enable-threads=posix
16115     * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap
16116     * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare
16117       7.1.1)
16118     * [339]11687 xstormy16-elf build fails in libf2c
16119     * [340]12263 [SGI IRIX] bootstrap fails during compile of
16120       libf2c/libI77/backspace.c
16121     * [341]12490 buffer overflow in scan-decls.c (during Solaris 9
16122       fix-header processing)
16123
16124    Internal compiler errors (multi-platform)
16125
16126     * [342]7277 Casting integers to vector types causes ICE
16127     * [343]7939 (c++) ICE on invalid function template specialization
16128     * [344]11063 (c++) ICE on parsing initialization list of const array
16129       member
16130     * [345]11207 ICE with negative index in array element designator
16131     * [346]11522 (fortran) g77 dwarf-2 ICE in
16132       add_abstract_origin_attribute
16133     * [347]11595 (c++) ICE on duplicate label definition
16134     * [348]11646 (c++) ICE in commit_one_edge_insertion with
16135       -fnon-call-exceptions -fgcse -O
16136     * [349]11665 ICE in struct initializer when taking address
16137     * [350]11852 (c++) ICE with bad struct initializer.
16138     * [351]11878 (c++) ICE in cp_expr_size
16139     * [352]11883 ICE with any -O on mercury-generated C code
16140     * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in
16141       cp/typeck2.c when applying typeid operator to template template
16142       parameter
16143     * [354]12146 ICE in lookup_template_function, in cp/pt.c
16144     * [355]12215 ICE in make_label_edge with -fnon-call-exceptions
16145       -fno-gcse -O2
16146     * [356]12369 (c++) ICE with templates and friends
16147     * [357]12446 ICE in emit_move_insn on complicated array reference
16148     * [358]12510 ICE in final_scan_insn
16149     * [359]12544 ICE with large parameters used in nested functions
16150
16151    C and optimization bugs
16152
16153     * [360]9862 spurious warnings with -W -finline-functions
16154     * [361]10962 lookup_field is a linear search on a linked list (can be
16155       slow if large struct)
16156     * [362]11370 -Wunreachable-code gives false complaints
16157     * [363]11637 invalid assembly with -fnon-call-exceptions
16158     * [364]11885 Problem with bitfields in packed structs
16159     * [365]12082 Inappropriate unreachable code warnings
16160     * [366]12180 Inline optimization fails for variadic function
16161     * [367]12340 loop unroller + gcse produces wrong code
16162
16163    C++ compiler and library
16164
16165     * [368]3907 nested template parameter collides with member name
16166     * [369]5293 confusing message when binding a temporary to a reference
16167     * [370]5296 [DR115] Pointers to functions and to template functions
16168       behave differently in deduction
16169     * [371]7939 ICE on function template specialization
16170     * [372]8656 Unable to assign function with __attribute__ and pointer
16171       return type to an appropriate variable
16172     * [373]10147 Confusing error message for invalid template function
16173       argument
16174     * [374]11400 std::search_n() makes assumptions about Size parameter
16175     * [375]11409 issues with using declarations, overloading, and
16176       built-in functions
16177     * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle
16178       multiple bits in mask
16179     * [377]11786 operator() call on variable in other namespace not
16180       recognized
16181     * [378]11867 static_cast ignores ambiguity
16182     * [379]11928 bug with conversion operators that are typedefs
16183     * [380]12114 Uninitialized memory accessed in dtor
16184     * [381]12163 static_cast + explicit constructor regression
16185     * [382]12181 Wrong code with comma operator and c++
16186     * [383]12236 regparm and fastcall messes up parameters
16187     * [384]12266 incorrect instantiation of unneeded template during
16188       overload resolution
16189     * [385]12296 istream::peek() doesn't set eofbit
16190     * [386]12298 [sjlj exceptions] Stack unwind destroys
16191       not-yet-constructed object
16192     * [387]12369 ICE with templates and friends
16193     * [388]12337 apparently infinite loop in g++
16194     * [389]12344 stdcall attribute ignored if function returns a pointer
16195     * [390]12451 missing(late) class forward declaration in cxxabi.h
16196     * [391]12486 g++ accepts invalid use of a qualified name
16197
16198    x86 specific (Intel/AMD)
16199
16200     * [392]8869 [x86 MMX] ICE with const variable optimization and MMX
16201       builtins
16202     * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions
16203       -O2
16204     * [394]11689 g++3.3 emits un-assembleable code for k6 architecture
16205     * [395]12116 [k6] Invalid assembly output values with X-MAME code
16206     * [396]12070 ICE converting between double and long double with
16207       -msoft-float
16208
16209    ia64-specific
16210
16211     * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc
16212     * [398]11535 __builtin_return_address may not work on ia64
16213     * [399]11693 [ia64] ICE in gen_nop_type
16214     * [400]12224 [ia64] Thread-local storage doesn't work
16215
16216    PowerPC-specific
16217
16218     * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux
16219       kernel
16220     * [402]11319 loop miscompiled on ppc32
16221     * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code
16222
16223    SPARC-specific
16224
16225     * [404]11662 wrong code for expr. with cast to long long and
16226       exclusive or
16227     * [405]11965 invalid assembler code for a shift < 32 operation
16228     * [406]12301 (c++) stack corruption when a returned expression throws
16229       an exception
16230
16231    Alpha-specific
16232
16233     * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of
16234       kernel 2.4.22-pre8
16235
16236    HPUX-specific
16237
16238     * [408]11313 problem with #pragma weak and static inline functions
16239     * [409]11712 __STDC_EXT__ not defined for C++ by default anymore?
16240
16241    Solaris specific
16242
16243     * [410]12166 Profiled programs crash if PROFDIR is set
16244
16245    Solaris-x86 specific
16246
16247     * [411]12101 i386 Solaris no longer works with GNU as?
16248
16249    Miscellaneous embedded target-specific bugs
16250
16251     * [412]10988 [m32r-elf] wrong blockmove code with -O3
16252     * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with
16253       -O2
16254     * [414]11902 [sh4] spec file improperly inserts rpath even when none
16255       needed
16256     * [415]11903 [sh4] -pthread fails to link due to error in spec file
16257       on sh4
16258     __________________________________________________________________
16259
16260GCC 3.3.3
16261
16262  Minor features
16263
16264   In addition to the bug fixes documented below, this release contains
16265   few minor features such as:
16266     * Support for --with-sysroot
16267     * Support for automatic detection of executable stacks
16268     * Support for SSE3 instructions
16269     * Support for thread local storage debugging under GDB on S390
16270
16271  Bug Fixes
16272
16273   This section lists the problem reports (PRs) from GCC's bug tracker
16274   that are known to be fixed in the 3.3.3 release. This list might not be
16275   complete (that is, it is possible that some PRs that have been fixed
16276   are not listed here).
16277
16278    Bootstrap failures and issues
16279
16280     * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails
16281     * [417]12399 boehm-gc fails (when building a cross compiler): libtool
16282       unable to infer tagged configuration
16283     * [418]13068 mklibgcc.in doesn't handle multi-level multilib
16284       subdirectories properly
16285
16286    Internal compiler errors (multi-platform)
16287
16288     * [419]10060 ICE (stack overflow) on huge file (300k lines) due to
16289       recursive behaviour of copy_rtx_if_shared, in emit_rtl.c
16290     * [420]10555 (c++) ICE on undefined template argument
16291     * [421]10706 (c++) ICE in mangle_class_name_for_template
16292     * [422]11496 (fortran) error in flow_loops_find when -funroll-loops
16293       active
16294     * [423]11741 ICE in pre_insert_copy_insn, in gcse.c
16295     * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0
16296     * [425]12632 (fortran) -fbounds-check ICE
16297     * [426]12712 (c++) ICE on short legit C++ code fragment with gcc
16298       3.3.2
16299     * [427]12726 (c++) ICE (segfault) on trivial code
16300     * [428]12890 (c++) ICE on compilation of class with throwing method
16301     * [429]12900 (c++) ICE in rtl_verify_flow_info_1
16302     * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on
16303       correct code with -O2 -fno-force-mem
16304     * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive
16305       template
16306     * [432]13318 ICE: floating point exception in the loop optimizer
16307     * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in
16308       except.c
16309     * [434]13574 (c++) invalid array default initializer in class lets
16310       gcc consume all memory and die
16311     * [435]13475 ICE on SIMD variables with partial value initialization
16312     * [436]13797 (c++) ICE on invalid template parameter
16313     * [437]13824 (java) gcj SEGV with simple .java program
16314
16315    C and optimization bugs
16316
16317     * [438]8776 loop invariants are not removed (most likely)
16318     * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing
16319       strncmp by memcmp
16320     * [440]11350 undefined labels with -Os -fPIC
16321     * [441]12826 Optimizer removes reference through volatile pointer
16322     * [442]12500 stabs debug info: void no longer a predefined / builtin
16323       type
16324     * [443]12941 builtin-bitops-1.c miscompilation (latent bug)
16325     * [444]12953 tree inliner bug (in inline_forbidden_p) and fix
16326     * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled
16327     * [446]13507 spurious printf format warning
16328     * [447]13382 Type information for const pointer disappears during
16329       optimization.
16330     * [448]13394 noreturn attribute ignored on recursive invokation
16331     * [449]13400 Compiled code crashes storing to read-only location
16332     * [450]13521 Endless loop in calculate_global_regs_live
16333
16334    C++ compiler and library
16335
16336   Some of the bug fixes in this list were made to implement decisions
16337   that the ISO C++ standards committee has made concerning several defect
16338   reports (DRs). Links in the list below point to detailed discussion of
16339   the relevant defect report.
16340     * [451]2094 unimplemented: use of `ptrmem_cst' in template type
16341       unification
16342     * [452]2294 using declaration confusion
16343     * [453]5050 template instantiation depth exceeds limit: recursion
16344       problem?
16345     * [454]9371 Bad exception handling in
16346       i/ostream::operator>>/<<(streambuf*)
16347     * [455]9546 bad exception handling in ostream members
16348     * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the
16349       face of unknown locales
16350     * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work
16351     * [459]10095 istream::operator>>(int&) sets ios::badbit when
16352       ios::failbit is set.
16353     * [460]11554 Warning about reordering of initializers doesn't mention
16354       location of constructor
16355     * [461]12297 istream::sentry::sentry() handles eof() incorrectly.
16356     * [462]12352 Exception safety problems in src/localename.cc
16357     * [463]12438 Memory leak in locale::combine()
16358     * [464]12540 Memory leak in locale::locale(const char*)
16359     * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented
16360     * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented
16361     * [470]12696 memory eating infinite loop in diagnostics (error
16362       recovery problem)
16363     * [471]12815 Code compiled with optimization behaves unexpectedly
16364     * [472]12862 Conflicts between typedefs/enums and namespace member
16365       declarations
16366     * [473]12926 Wrong value after assignment in initialize list using
16367       bit-fields
16368     * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented
16369     * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented
16370     * [478]13007 basic_streambuf::pubimbue, imbue wrong
16371     * [479]13009 Implicitly-defined assignment operator writes to wrong
16372       memory
16373     * [480]13057 regparm attribute not applied to destructor
16374     * [481]13070 -Wformat option ignored in g++
16375     * [482]13081 forward template declarations in <complex> let inlining
16376       fail
16377     * [483]13239 Assertion does not seem to work correctly anymore
16378     * [484]13262 "xxx is private within this context" when initializing a
16379       self-contained template class
16380     * [485]13290 simple typo in concept checking for std::generate_n
16381     * [486]13323 Template code does not compile in presence of typedef
16382     * [487]13369 __verify_grouping (and __add_grouping?) not correct
16383     * [488]13371 infinite loop with packed struct and inlining
16384     * [489]13445 Template argument replacement "dereferences" a typedef
16385     * [490]13461 Fails to access protected-ctor from public constant
16386     * [491]13462 Non-standard-conforming type set::pointer
16387     * [492]13478 gcc uses wrong constructor to initialize a const
16388       reference
16389     * [493]13544 "conflicting types" for enums in different scopes
16390     * [494]13650 string::compare should not (always) use
16391       traits_type::length()
16392     * [495]13683 bogus warning about passing non-PODs through ellipsis
16393     * [496]13688 Derived class is denied access to protected base class
16394       member class
16395     * [497]13774 Member variable cleared in virtual multiple inheritance
16396       class
16397     * [498]13884 Protect sstream.tcc from extern template use
16398
16399    Java compiler and library
16400
16401     * [499]10746 [win32] garbage collection crash in GCJ
16402
16403    Objective-C compiler and library
16404
16405     * [500]11433 Crash due to dereferencing null pointer when querying
16406       protocol
16407
16408    Fortran compiler and library
16409
16410     * [501]12633 logical expression gives incorrect result with
16411       -fugly-logint option
16412     * [502]13037 [gcse-lm] g77 generates incorrect code
16413     * [503]13213 Hex constant problem when compiling with -fugly-logint
16414       and -ftypeless-boz
16415
16416    x86-specific (Intel/AMD)
16417
16418     * [504]4490 ICE with -m128bit-long-double
16419     * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have
16420       `reg' in make_field_assignment, in combine.c
16421     * [506]12441 ICE: can't find a register to spill
16422     * [507]12943 array static-init failure under -fpic, -fPIC
16423     * [508]13608 Incorrect code with -O3 -ffast-math
16424
16425    PowerPC-specific
16426
16427     * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of
16428       __attribute__((aligned(16)))
16429     * [510]11793 ICE in extract_insn, in recog.c (const_vector's)
16430     * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in
16431       altivec.md)
16432     * [512]12537 g++ generates writeable text sections
16433
16434    SPARC-specific
16435
16436     * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0
16437       -m64
16438     * [514]12865 mprotect call to make trampoline executable may fail
16439     * [515]13354 ICE in sparc_emit_set_const32
16440
16441    ARM-specific
16442
16443     * [516]10467 [arm] ICE in pre_insert_copy_insn,
16444
16445    ia64-specific
16446
16447     * [517]11226 ICE passing struct arg with two floats
16448     * [518]11227 ICE for _Complex float, _Complex long double args
16449     * [519]12644 GCC 3.3.2 fails to compile glibc on ia64
16450     * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn
16451     * Various fixes for libunwind
16452
16453    Alpha-specific
16454
16455     * [521]12654 Incorrect comparison code generated for Alpha
16456     * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2
16457     * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2
16458
16459    HPPA-specific
16460
16461     * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c
16462     * [525]12158 [hppa] compilation does not terminate at -O1
16463
16464    S390-specific
16465
16466     * [526]11992 Wrong built-in code for memcmp with length 1<<24: only
16467       (1<<24)-1 possible for CLCL-Instruction
16468
16469    SH-specific
16470
16471     * [527]9365 segfault in gen_far_branch (config/sh/sh.c)
16472     * [528]10392 optimizer generates faulty array indexing
16473     * [529]11322 SH profiler outputs multiple definitions of symbol
16474     * [530]13069 gcc/config/sh/rtems.h broken
16475     * [531]13302 Putting a va_list in a struct causes seg fault
16476     * [532]13585 Incorrect optimization of call to sfunc
16477     * Fix inappropriately exported libgcc functions from the shared
16478       library
16479
16480    Other embedded target specific
16481
16482     * [533]8916 [mcore] unsigned char assign gets hosed.
16483     * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c
16484     * [535]13122 [h8300] local variable gets corrupted by function call
16485       when -fomit-frame-pointer is given
16486     * [536]13256 [cris] strict_low_part mistreated in delay slots
16487     * [537]13373 [mcore] optimization with -frerun-cse-after-loop
16488       -fexpensive-optimizations produces wrong code on mcore
16489
16490    GNU HURD-specific
16491
16492     * [538]12561 gcc/config/t-gnu needs updating to work with
16493       --with-sysroot
16494
16495    Tru64 Unix specific
16496
16497     * [539]6243 testsuite fails almost all tests due to no libintl in
16498       LD_LIBRARY_PATH during test.
16499     * [540]11397 weak aliases broken on Tru64 UNIX
16500
16501    AIX-specific
16502
16503     * [541]12505 build failure due to defines of uchar in cpphash.h and
16504       sys/types.h
16505     * [542]13150 WEAK symbols not exported by collect2
16506
16507    IRIX-specific
16508
16509     * [543]12666 fixincludes problem on IRIX 6.5.19m
16510
16511    Solaris-specific
16512
16513     * [544]12969 Including sys/byteorder.h breaks configure checks
16514
16515    Testsuite problems (compiler is not affected)
16516
16517     * [545]10819 testsuite creates CR+LF on compiler version lines in
16518       test summary files
16519     * [546]11612 abi_check not finding correct libgcc_s.so.1
16520
16521    Miscellaneous
16522
16523     * [547]13211 using -###, incorrect warnings about unused linker file
16524       are produced
16525     __________________________________________________________________
16526
16527GCC 3.3.4
16528
16529   This is the [548]list of problem reports (PRs) from GCC's bug tracking
16530   system that are known to be fixed in the 3.3.4 release. This list might
16531   not be complete (that is, it is possible that some PRs that have been
16532   fixed are not listed here).
16533     __________________________________________________________________
16534
16535GCC 3.3.5
16536
16537   This is the [549]list of problem reports (PRs) from GCC's bug tracking
16538   system that are known to be fixed in the 3.3.5 release. This list might
16539   not be complete (that is, it is possible that some PRs that have been
16540   fixed are not listed here).
16541     __________________________________________________________________
16542
16543GCC 3.3.6
16544
16545   This is the [550]list of problem reports (PRs) from GCC's bug tracking
16546   system that are known to be fixed in the 3.3.6 release. This list might
16547   not be complete (that is, it is possible that some PRs that have been
16548   fixed are not listed here).
16549
16550
16551    For questions related to the use of GCC, please consult these web
16552    pages and the [551]GCC manuals. If that fails, the
16553    [552]gcc-help@gcc.gnu.org mailing list might help. Comments on these
16554    web pages and the development of GCC are welcome on our developer
16555    list at [553]gcc@gcc.gnu.org. All of [554]our lists have public
16556    archives.
16557
16558   Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and
16559   distribution of this entire article is permitted in any medium,
16560   provided this notice is preserved.
16561
16562   These pages are [556]maintained by the GCC team. Last modified
16563   2019-11-28[557].
16564
16565References
16566
16567   1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
16568   2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems
16569   3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
16570   4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute
16571   5. http://gcc.gnu.org/news/dfa.html
16572   6. http://gcc.gnu.org/c99status.html
16573   7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html
16574   8. https://gcc.gnu.org/PR10140
16575   9. https://gcc.gnu.org/PR10198
16576  10. https://gcc.gnu.org/PR10338
16577  11. https://gcc.gnu.org/PR3581
16578  12. https://gcc.gnu.org/PR4382
16579  13. https://gcc.gnu.org/PR5533
16580  14. https://gcc.gnu.org/PR6387
16581  15. https://gcc.gnu.org/PR6412
16582  16. https://gcc.gnu.org/PR6620
16583  17. https://gcc.gnu.org/PR6663
16584  18. https://gcc.gnu.org/PR7068
16585  19. https://gcc.gnu.org/PR7083
16586  20. https://gcc.gnu.org/PR7647
16587  21. https://gcc.gnu.org/PR7675
16588  22. https://gcc.gnu.org/PR7718
16589  23. https://gcc.gnu.org/PR8116
16590  24. https://gcc.gnu.org/PR8358
16591  25. https://gcc.gnu.org/PR8511
16592  26. https://gcc.gnu.org/PR8564
16593  27. https://gcc.gnu.org/PR8660
16594  28. https://gcc.gnu.org/PR8766
16595  29. https://gcc.gnu.org/PR8803
16596  30. https://gcc.gnu.org/PR8846
16597  31. https://gcc.gnu.org/PR8906
16598  32. https://gcc.gnu.org/PR9216
16599  33. https://gcc.gnu.org/PR9261
16600  34. https://gcc.gnu.org/PR9263
16601  35. https://gcc.gnu.org/PR9429
16602  36. https://gcc.gnu.org/PR9516
16603  37. https://gcc.gnu.org/PR9600
16604  38. https://gcc.gnu.org/PR9629
16605  39. https://gcc.gnu.org/PR9672
16606  40. https://gcc.gnu.org/PR9749
16607  41. https://gcc.gnu.org/PR9794
16608  42. https://gcc.gnu.org/PR9829
16609  43. https://gcc.gnu.org/PR9916
16610  44. https://gcc.gnu.org/PR9936
16611  45. https://gcc.gnu.org/PR10262
16612  46. https://gcc.gnu.org/PR10278
16613  47. https://gcc.gnu.org/PR10446
16614  48. https://gcc.gnu.org/PR10451
16615  49. https://gcc.gnu.org/PR10506
16616  50. https://gcc.gnu.org/PR10549
16617  51. https://gcc.gnu.org/PR2001
16618  52. https://gcc.gnu.org/PR2391
16619  53. https://gcc.gnu.org/PR2960
16620  54. https://gcc.gnu.org/PR4046
16621  55. https://gcc.gnu.org/PR6405
16622  56. https://gcc.gnu.org/PR6798
16623  57. https://gcc.gnu.org/PR6871
16624  58. https://gcc.gnu.org/PR6909
16625  59. https://gcc.gnu.org/PR7189
16626  60. https://gcc.gnu.org/PR7642
16627  61. https://gcc.gnu.org/PR8634
16628  62. https://gcc.gnu.org/PR8750
16629  63. https://gcc.gnu.org/PR2161
16630  64. https://gcc.gnu.org/PR4319
16631  65. https://gcc.gnu.org/PR8602
16632  66. https://gcc.gnu.org/PR9177
16633  67. https://gcc.gnu.org/PR9853
16634  68. https://gcc.gnu.org/PR45
16635  69. https://gcc.gnu.org/PR3784
16636  70. https://gcc.gnu.org/PR764
16637  71. https://gcc.gnu.org/PR5116
16638  72. https://gcc.gnu.org/PR2862
16639  73. https://gcc.gnu.org/PR3663
16640  74. https://gcc.gnu.org/PR3797
16641  75. https://gcc.gnu.org/PR3948
16642  76. https://gcc.gnu.org/PR4137
16643  77. https://gcc.gnu.org/PR4361
16644  78. https://gcc.gnu.org/PR4802
16645  79. https://gcc.gnu.org/PR5837
16646  80. https://gcc.gnu.org/PR4803
16647  81. https://gcc.gnu.org/PR5094
16648  82. https://gcc.gnu.org/PR5730
16649  83. https://gcc.gnu.org/PR6713
16650  84. https://gcc.gnu.org/PR7015
16651  85. https://gcc.gnu.org/PR7086
16652  86. https://gcc.gnu.org/PR7099
16653  87. https://gcc.gnu.org/PR7247
16654  88. https://gcc.gnu.org/PR7441
16655  89. https://gcc.gnu.org/PR7768
16656  90. https://gcc.gnu.org/PR7804
16657  91. https://gcc.gnu.org/PR8099
16658  92. https://gcc.gnu.org/PR8117
16659  93. https://gcc.gnu.org/PR8205
16660  94. https://gcc.gnu.org/PR8645
16661  95. https://gcc.gnu.org/PR8724
16662  96. https://gcc.gnu.org/PR8805
16663  97. https://gcc.gnu.org/PR8691
16664  98. https://gcc.gnu.org/PR8700
16665  99. https://gcc.gnu.org/PR8724
16666 100. https://gcc.gnu.org/PR8949
16667 101. https://gcc.gnu.org/PR9016
16668 102. https://gcc.gnu.org/PR9053
16669 103. https://gcc.gnu.org/PR9152
16670 104. https://gcc.gnu.org/PR9182
16671 105. https://gcc.gnu.org/PR9297
16672 106. https://gcc.gnu.org/PR9318
16673 107. https://gcc.gnu.org/PR9320
16674 108. https://gcc.gnu.org/PR9400
16675 109. https://gcc.gnu.org/PR9424
16676 110. https://gcc.gnu.org/PR9425
16677 111. https://gcc.gnu.org/PR9439
16678 112. https://gcc.gnu.org/PR9474
16679 113. https://gcc.gnu.org/PR9548
16680 114. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231
16681 115. https://gcc.gnu.org/PR9555
16682 116. https://gcc.gnu.org/PR9561
16683 117. https://gcc.gnu.org/PR9563
16684 118. https://gcc.gnu.org/PR9582
16685 119. https://gcc.gnu.org/PR9622
16686 120. https://gcc.gnu.org/PR9683
16687 121. https://gcc.gnu.org/PR9791
16688 122. https://gcc.gnu.org/PR9817
16689 123. https://gcc.gnu.org/PR9825
16690 124. https://gcc.gnu.org/PR9826
16691 125. https://gcc.gnu.org/PR9924
16692 126. https://gcc.gnu.org/PR9946
16693 127. https://gcc.gnu.org/PR9964
16694 128. https://gcc.gnu.org/PR9988
16695 129. https://gcc.gnu.org/PR10033
16696 130. https://gcc.gnu.org/PR10097
16697 131. https://gcc.gnu.org/PR10132
16698 132. https://gcc.gnu.org/PR10180
16699 133. https://gcc.gnu.org/PR10199
16700 134. https://gcc.gnu.org/PR10300
16701 135. https://gcc.gnu.org/PR10427
16702 136. https://gcc.gnu.org/PR10503
16703 137. https://gcc.gnu.org/PR5956
16704 138. https://gcc.gnu.org/PR1832
16705 139. https://gcc.gnu.org/PR3924
16706 140. https://gcc.gnu.org/PR5634
16707 141. https://gcc.gnu.org/PR6367
16708 142. https://gcc.gnu.org/PR6491
16709 143. https://gcc.gnu.org/PR6742
16710 144. https://gcc.gnu.org/PR7113
16711 145. https://gcc.gnu.org/PR7236
16712 146. https://gcc.gnu.org/PR7278
16713 147. https://gcc.gnu.org/PR7384
16714 148. https://gcc.gnu.org/PR7388
16715 149. https://gcc.gnu.org/PR8587
16716 150. https://gcc.gnu.org/PR9038
16717 151. https://gcc.gnu.org/PR10197
16718 152. https://gcc.gnu.org/PR6005
16719 153. https://gcc.gnu.org/PR6389
16720 154. https://gcc.gnu.org/PR6576
16721 155. https://gcc.gnu.org/PR6652
16722 156. https://gcc.gnu.org/PR7060
16723 157. https://gcc.gnu.org/PR7073
16724 158. https://gcc.gnu.org/PR7180
16725 159. https://gcc.gnu.org/PR7416
16726 160. https://gcc.gnu.org/PR7570
16727 161. https://gcc.gnu.org/PR7578
16728 162. https://gcc.gnu.org/PR7611
16729 163. https://gcc.gnu.org/PR7709
16730 164. https://gcc.gnu.org/PR7766
16731 165. https://gcc.gnu.org/PR7785
16732 166. https://gcc.gnu.org/PR7786
16733 167. https://gcc.gnu.org/PR8142
16734 168. https://gcc.gnu.org/PR8234
16735 169. https://gcc.gnu.org/PR8415
16736 170. https://gcc.gnu.org/PR8481
16737 171. https://gcc.gnu.org/PR8593
16738 172. https://gcc.gnu.org/PR8759
16739 173. https://gcc.gnu.org/PR8997
16740 174. https://gcc.gnu.org/PR9253
16741 175. https://gcc.gnu.org/PR9254
16742 176. https://gcc.gnu.org/PR9271
16743 177. https://gcc.gnu.org/PR6767
16744 178. https://gcc.gnu.org/PR9911
16745 179. https://gcc.gnu.org/PR10020
16746 180. https://gcc.gnu.org/PR10546
16747 181. https://gcc.gnu.org/PR7029
16748 182. https://gcc.gnu.org/PR2903
16749 183. https://gcc.gnu.org/PR7873
16750 184. https://gcc.gnu.org/PR7680
16751 185. https://gcc.gnu.org/PR8705
16752 186. https://gcc.gnu.org/PR9986
16753 187. https://gcc.gnu.org/PR10056
16754 188. https://gcc.gnu.org/PR6744
16755 189. https://gcc.gnu.org/PR7361
16756 190. https://gcc.gnu.org/PR9496
16757 191. https://gcc.gnu.org/PR7067
16758 192. https://gcc.gnu.org/PR8480
16759 193. https://gcc.gnu.org/PR8784
16760 194. https://gcc.gnu.org/PR10315
16761 195. https://gcc.gnu.org/PR10267
16762 196. https://gcc.gnu.org/PR7916
16763 197. https://gcc.gnu.org/PR7926
16764 198. https://gcc.gnu.org/PR8555
16765 199. https://gcc.gnu.org/PR8994
16766 200. https://gcc.gnu.org/PR9426
16767 201. https://gcc.gnu.org/PR9806
16768 202. https://gcc.gnu.org/PR10077
16769 203. https://gcc.gnu.org/PR10233
16770 204. https://gcc.gnu.org/PR10286
16771 205. https://gcc.gnu.org/PR10308
16772 206. https://gcc.gnu.org/PR11272
16773 207. https://gcc.gnu.org/PR5754
16774 208. https://gcc.gnu.org/PR6597
16775 209. https://gcc.gnu.org/PR6949
16776 210. https://gcc.gnu.org/PR7053
16777 211. https://gcc.gnu.org/PR8164
16778 212. https://gcc.gnu.org/PR8384
16779 213. https://gcc.gnu.org/PR9559
16780 214. https://gcc.gnu.org/PR9649
16781 215. https://gcc.gnu.org/PR9864
16782 216. https://gcc.gnu.org/PR10432
16783 217. https://gcc.gnu.org/PR10475
16784 218. https://gcc.gnu.org/PR10635
16785 219. https://gcc.gnu.org/PR10661
16786 220. https://gcc.gnu.org/PR10700
16787 221. https://gcc.gnu.org/PR10712
16788 222. https://gcc.gnu.org/PR10796
16789 223. https://gcc.gnu.org/PR10890
16790 224. https://gcc.gnu.org/PR10939
16791 225. https://gcc.gnu.org/PR10956
16792 226. https://gcc.gnu.org/PR11041
16793 227. https://gcc.gnu.org/PR11059
16794 228. https://gcc.gnu.org/PR11083
16795 229. https://gcc.gnu.org/PR11105
16796 230. https://gcc.gnu.org/PR11149
16797 231. https://gcc.gnu.org/PR11228
16798 232. https://gcc.gnu.org/PR11282
16799 233. https://gcc.gnu.org/PR11301
16800 234. https://gcc.gnu.org/PR11308
16801 235. https://gcc.gnu.org/PR11473
16802 236. https://gcc.gnu.org/PR11503
16803 237. https://gcc.gnu.org/PR11513
16804 238. https://gcc.gnu.org/PR11198
16805 239. https://gcc.gnu.org/PR11304
16806 240. https://gcc.gnu.org/PR11381
16807 241. https://gcc.gnu.org/PR11536
16808 242. https://gcc.gnu.org/PR11557
16809 243. https://gcc.gnu.org/PR5897
16810 244. https://gcc.gnu.org/PR11279
16811 245. https://gcc.gnu.org/PR11022
16812 246. https://gcc.gnu.org/PR2330
16813 247. https://gcc.gnu.org/PR5388
16814 248. https://gcc.gnu.org/PR5390
16815 249. https://gcc.gnu.org/PR7877
16816 250. https://gcc.gnu.org/PR9393
16817 251. https://gcc.gnu.org/PR10032
16818 252. https://gcc.gnu.org/PR10468
16819 253. https://gcc.gnu.org/PR10527
16820 254. https://gcc.gnu.org/PR10679
16821 255. https://gcc.gnu.org/PR10682
16822 256. https://gcc.gnu.org/PR10689
16823 257. https://gcc.gnu.org/PR10845
16824 258. https://gcc.gnu.org/PR10849
16825 259. https://gcc.gnu.org/PR10888
16826 260. https://gcc.gnu.org/PR10929
16827 261. https://gcc.gnu.org/PR10931
16828 262. https://gcc.gnu.org/PR10940
16829 263. https://gcc.gnu.org/PR10968
16830 264. https://gcc.gnu.org/PR10990
16831 265. https://gcc.gnu.org/PR11039
16832 266. https://gcc.gnu.org/PR11062
16833 267. https://gcc.gnu.org/PR11095
16834 268. https://gcc.gnu.org/PR11098
16835 269. https://gcc.gnu.org/PR11137
16836 270. https://gcc.gnu.org/PR11154
16837 271. https://gcc.gnu.org/PR11329
16838 272. https://gcc.gnu.org/PR11332
16839 273. https://gcc.gnu.org/PR11431
16840 274. https://gcc.gnu.org/PR11528
16841 275. https://gcc.gnu.org/PR11546
16842 276. https://gcc.gnu.org/PR11567
16843 277. https://gcc.gnu.org/PR11645
16844 278. https://gcc.gnu.org/PR5179
16845 279. https://gcc.gnu.org/PR8204
16846 280. https://gcc.gnu.org/PR10838
16847 281. https://gcc.gnu.org/PR10886
16848 282. https://gcc.gnu.org/PR11349
16849 283. https://gcc.gnu.org/PR4823
16850 284. https://gcc.gnu.org/PR8878
16851 285. https://gcc.gnu.org/PR9815
16852 286. https://gcc.gnu.org/PR10402
16853 287. https://gcc.gnu.org/PR10504
16854 288. https://gcc.gnu.org/PR10673
16855 289. https://gcc.gnu.org/PR11044
16856 290. https://gcc.gnu.org/PR11089
16857 291. https://gcc.gnu.org/PR11420
16858 292. https://gcc.gnu.org/PR9362
16859 293. https://gcc.gnu.org/PR10142
16860 294. https://gcc.gnu.org/PR10663
16861 295. https://gcc.gnu.org/PR10835
16862 296. https://gcc.gnu.org/PR10876
16863 297. https://gcc.gnu.org/PR10955
16864 298. https://gcc.gnu.org/PR11018
16865 299. https://gcc.gnu.org/PR11556
16866 300. https://gcc.gnu.org/PR10907
16867 301. https://gcc.gnu.org/PR11320
16868 302. https://gcc.gnu.org/PR11599
16869 303. https://gcc.gnu.org/PR9745
16870 304. https://gcc.gnu.org/PR10871
16871 305. https://gcc.gnu.org/PR11440
16872 306. https://gcc.gnu.org/PR7594
16873 307. https://gcc.gnu.org/PR10557
16874 308. https://gcc.gnu.org/PR11054
16875 309. https://gcc.gnu.org/PR10834
16876 310. https://gcc.gnu.org/PR10842
16877 311. https://gcc.gnu.org/PR11052
16878 312. https://gcc.gnu.org/PR11183
16879 313. https://gcc.gnu.org/PR11084
16880 314. https://gcc.gnu.org/PR10331
16881 315. https://gcc.gnu.org/PR10413
16882 316. https://gcc.gnu.org/PR11096
16883 317. https://gcc.gnu.org/PR2873
16884 318. https://gcc.gnu.org/PR3163
16885 319. https://gcc.gnu.org/PR5287
16886 320. https://gcc.gnu.org/PR10148
16887 321. https://gcc.gnu.org/PR8787
16888 322. https://gcc.gnu.org/PR10900
16889 323. https://gcc.gnu.org/PR1607
16890 324. https://gcc.gnu.org/PR4252
16891 325. https://gcc.gnu.org/PR4490
16892 326. https://gcc.gnu.org/PR10355
16893 327. https://gcc.gnu.org/PR10726
16894 328. https://gcc.gnu.org/PR10805
16895 329. https://gcc.gnu.org/PR10815
16896 330. https://gcc.gnu.org/PR10877
16897 331. https://gcc.gnu.org/PR11280
16898 332. https://gcc.gnu.org/PR11466
16899 333. https://gcc.gnu.org/PR10737
16900 334. https://gcc.gnu.org/PR10810
16901 335. https://gcc.gnu.org/PR8336
16902 336. https://gcc.gnu.org/PR9330
16903 337. https://gcc.gnu.org/PR9631
16904 338. https://gcc.gnu.org/PR9877
16905 339. https://gcc.gnu.org/PR11687
16906 340. https://gcc.gnu.org/PR12263
16907 341. https://gcc.gnu.org/PR12490
16908 342. https://gcc.gnu.org/PR7277
16909 343. https://gcc.gnu.org/PR7939
16910 344. https://gcc.gnu.org/PR11063
16911 345. https://gcc.gnu.org/PR11207
16912 346. https://gcc.gnu.org/PR11522
16913 347. https://gcc.gnu.org/PR11595
16914 348. https://gcc.gnu.org/PR11646
16915 349. https://gcc.gnu.org/PR11665
16916 350. https://gcc.gnu.org/PR11852
16917 351. https://gcc.gnu.org/PR11878
16918 352. https://gcc.gnu.org/PR11883
16919 353. https://gcc.gnu.org/PR11991
16920 354. https://gcc.gnu.org/PR12146
16921 355. https://gcc.gnu.org/PR12215
16922 356. https://gcc.gnu.org/PR12369
16923 357. https://gcc.gnu.org/PR12446
16924 358. https://gcc.gnu.org/PR12510
16925 359. https://gcc.gnu.org/PR12544
16926 360. https://gcc.gnu.org/PR9862
16927 361. https://gcc.gnu.org/PR10962
16928 362. https://gcc.gnu.org/PR11370
16929 363. https://gcc.gnu.org/PR11637
16930 364. https://gcc.gnu.org/PR11885
16931 365. https://gcc.gnu.org/PR12082
16932 366. https://gcc.gnu.org/PR12180
16933 367. https://gcc.gnu.org/PR12340
16934 368. https://gcc.gnu.org/PR3907
16935 369. https://gcc.gnu.org/PR5293
16936 370. https://gcc.gnu.org/PR5296
16937 371. https://gcc.gnu.org/PR7939
16938 372. https://gcc.gnu.org/PR8656
16939 373. https://gcc.gnu.org/PR10147
16940 374. https://gcc.gnu.org/PR11400
16941 375. https://gcc.gnu.org/PR11409
16942 376. https://gcc.gnu.org/PR11740
16943 377. https://gcc.gnu.org/PR11786
16944 378. https://gcc.gnu.org/PR11867
16945 379. https://gcc.gnu.org/PR11928
16946 380. https://gcc.gnu.org/PR12114
16947 381. https://gcc.gnu.org/PR12163
16948 382. https://gcc.gnu.org/PR12181
16949 383. https://gcc.gnu.org/PR12236
16950 384. https://gcc.gnu.org/PR12266
16951 385. https://gcc.gnu.org/PR12296
16952 386. https://gcc.gnu.org/PR12298
16953 387. https://gcc.gnu.org/PR12369
16954 388. https://gcc.gnu.org/PR12337
16955 389. https://gcc.gnu.org/PR12344
16956 390. https://gcc.gnu.org/PR12451
16957 391. https://gcc.gnu.org/PR12486
16958 392. https://gcc.gnu.org/PR8869
16959 393. https://gcc.gnu.org/PR9786
16960 394. https://gcc.gnu.org/PR11689
16961 395. https://gcc.gnu.org/PR12116
16962 396. https://gcc.gnu.org/PR12070
16963 397. https://gcc.gnu.org/PR11184
16964 398. https://gcc.gnu.org/PR11535
16965 399. https://gcc.gnu.org/PR11693
16966 400. https://gcc.gnu.org/PR12224
16967 401. https://gcc.gnu.org/PR11087
16968 402. https://gcc.gnu.org/PR11319
16969 403. https://gcc.gnu.org/PR11949
16970 404. https://gcc.gnu.org/PR11662
16971 405. https://gcc.gnu.org/PR11965
16972 406. https://gcc.gnu.org/PR12301
16973 407. https://gcc.gnu.org/PR11717
16974 408. https://gcc.gnu.org/PR11313
16975 409. https://gcc.gnu.org/PR11712
16976 410. https://gcc.gnu.org/PR12166
16977 411. https://gcc.gnu.org/PR12101
16978 412. https://gcc.gnu.org/PR10988
16979 413. https://gcc.gnu.org/PR11805
16980 414. https://gcc.gnu.org/PR11902
16981 415. https://gcc.gnu.org/PR11903
16982 416. https://gcc.gnu.org/PR11890
16983 417. https://gcc.gnu.org/PR12399
16984 418. https://gcc.gnu.org/PR13068
16985 419. https://gcc.gnu.org/PR10060
16986 420. https://gcc.gnu.org/PR10555
16987 421. https://gcc.gnu.org/PR10706
16988 422. https://gcc.gnu.org/PR11496
16989 423. https://gcc.gnu.org/PR11741
16990 424. https://gcc.gnu.org/PR12440
16991 425. https://gcc.gnu.org/PR12632
16992 426. https://gcc.gnu.org/PR12712
16993 427. https://gcc.gnu.org/PR12726
16994 428. https://gcc.gnu.org/PR12890
16995 429. https://gcc.gnu.org/PR12900
16996 430. https://gcc.gnu.org/PR13060
16997 431. https://gcc.gnu.org/PR13289
16998 432. https://gcc.gnu.org/PR13318
16999 433. https://gcc.gnu.org/PR13392
17000 434. https://gcc.gnu.org/PR13574
17001 435. https://gcc.gnu.org/PR13475
17002 436. https://gcc.gnu.org/PR13797
17003 437. https://gcc.gnu.org/PR13824
17004 438. https://gcc.gnu.org/PR8776
17005 439. https://gcc.gnu.org/PR10339
17006 440. https://gcc.gnu.org/PR11350
17007 441. https://gcc.gnu.org/PR12826
17008 442. https://gcc.gnu.org/PR12500
17009 443. https://gcc.gnu.org/PR12941
17010 444. https://gcc.gnu.org/PR12953
17011 445. https://gcc.gnu.org/PR13041
17012 446. https://gcc.gnu.org/PR13507
17013 447. https://gcc.gnu.org/PR13382
17014 448. https://gcc.gnu.org/PR13394
17015 449. https://gcc.gnu.org/PR13400
17016 450. https://gcc.gnu.org/PR13521
17017 451. https://gcc.gnu.org/PR2094
17018 452. https://gcc.gnu.org/PR2294
17019 453. https://gcc.gnu.org/PR5050
17020 454. https://gcc.gnu.org/PR9371
17021 455. https://gcc.gnu.org/PR9546
17022 456. https://gcc.gnu.org/PR10081
17023 457. https://gcc.gnu.org/PR10093
17024 458. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61
17025 459. https://gcc.gnu.org/PR10095
17026 460. https://gcc.gnu.org/PR11554
17027 461. https://gcc.gnu.org/PR12297
17028 462. https://gcc.gnu.org/PR12352
17029 463. https://gcc.gnu.org/PR12438
17030 464. https://gcc.gnu.org/PR12540
17031 465. https://gcc.gnu.org/PR12594
17032 466. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60
17033 467. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63
17034 468. https://gcc.gnu.org/PR12657
17035 469. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292
17036 470. https://gcc.gnu.org/PR12696
17037 471. https://gcc.gnu.org/PR12815
17038 472. https://gcc.gnu.org/PR12862
17039 473. https://gcc.gnu.org/PR12926
17040 474. https://gcc.gnu.org/PR12967
17041 475. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html
17042 476. https://gcc.gnu.org/PR12971
17043 477. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328
17044 478. https://gcc.gnu.org/PR13007
17045 479. https://gcc.gnu.org/PR13009
17046 480. https://gcc.gnu.org/PR13057
17047 481. https://gcc.gnu.org/PR13070
17048 482. https://gcc.gnu.org/PR13081
17049 483. https://gcc.gnu.org/PR13239
17050 484. https://gcc.gnu.org/PR13262
17051 485. https://gcc.gnu.org/PR13290
17052 486. https://gcc.gnu.org/PR13323
17053 487. https://gcc.gnu.org/PR13369
17054 488. https://gcc.gnu.org/PR13371
17055 489. https://gcc.gnu.org/PR13445
17056 490. https://gcc.gnu.org/PR13461
17057 491. https://gcc.gnu.org/PR13462
17058 492. https://gcc.gnu.org/PR13478
17059 493. https://gcc.gnu.org/PR13544
17060 494. https://gcc.gnu.org/PR13650
17061 495. https://gcc.gnu.org/PR13683
17062 496. https://gcc.gnu.org/PR13688
17063 497. https://gcc.gnu.org/PR13774
17064 498. https://gcc.gnu.org/PR13884
17065 499. https://gcc.gnu.org/PR10746
17066 500. https://gcc.gnu.org/PR11433
17067 501. https://gcc.gnu.org/PR12633
17068 502. https://gcc.gnu.org/PR13037
17069 503. https://gcc.gnu.org/PR13213
17070 504. https://gcc.gnu.org/PR4490
17071 505. https://gcc.gnu.org/PR12292
17072 506. https://gcc.gnu.org/PR12441
17073 507. https://gcc.gnu.org/PR12943
17074 508. https://gcc.gnu.org/PR13608
17075 509. https://gcc.gnu.org/PR11598
17076 510. https://gcc.gnu.org/PR11793
17077 511. https://gcc.gnu.org/PR12467
17078 512. https://gcc.gnu.org/PR12537
17079 513. https://gcc.gnu.org/PR12496
17080 514. https://gcc.gnu.org/PR12865
17081 515. https://gcc.gnu.org/PR13354
17082 516. https://gcc.gnu.org/PR10467
17083 517. https://gcc.gnu.org/PR11226
17084 518. https://gcc.gnu.org/PR11227
17085 519. https://gcc.gnu.org/PR12644
17086 520. https://gcc.gnu.org/PR13149
17087 521. https://gcc.gnu.org/PR12654
17088 522. https://gcc.gnu.org/PR12965
17089 523. https://gcc.gnu.org/PR13031
17090 524. https://gcc.gnu.org/PR11634
17091 525. https://gcc.gnu.org/PR12158
17092 526. https://gcc.gnu.org/PR11992
17093 527. https://gcc.gnu.org/PR9365
17094 528. https://gcc.gnu.org/PR10392
17095 529. https://gcc.gnu.org/PR11322
17096 530. https://gcc.gnu.org/PR13069
17097 531. https://gcc.gnu.org/PR13302
17098 532. https://gcc.gnu.org/PR13585
17099 533. https://gcc.gnu.org/PR8916
17100 534. https://gcc.gnu.org/PR11576
17101 535. https://gcc.gnu.org/PR13122
17102 536. https://gcc.gnu.org/PR13256
17103 537. https://gcc.gnu.org/PR13373
17104 538. https://gcc.gnu.org/PR12561
17105 539. https://gcc.gnu.org/PR6243
17106 540. https://gcc.gnu.org/PR11397
17107 541. https://gcc.gnu.org/PR12505
17108 542. https://gcc.gnu.org/PR13150
17109 543. https://gcc.gnu.org/PR12666
17110 544. https://gcc.gnu.org/PR12969
17111 545. https://gcc.gnu.org/PR10819
17112 546. https://gcc.gnu.org/PR11612
17113 547. https://gcc.gnu.org/PR13211
17114 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4
17115 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5
17116 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6
17117 551. https://gcc.gnu.org/onlinedocs/
17118 552. mailto:gcc-help@gcc.gnu.org
17119 553. mailto:gcc@gcc.gnu.org
17120 554. https://gcc.gnu.org/lists.html
17121 555. https://www.fsf.org/
17122 556. https://gcc.gnu.org/about.html
17123 557. http://validator.w3.org/check/referer
17124======================================================================
17125http://gcc.gnu.org/gcc-3.2/index.html
17126                             GCC 3.2 Release Series
17127
17128   (This release series is no longer supported.)
17129
17130   April 25, 2003
17131
17132   The [1]GNU project and the GCC developers are pleased to announce the
17133   release of GCC 3.2.3.
17134
17135   The purpose of the GCC 3.2 release series is to provide a stable
17136   platform for OS distributors to use building their next releases. A
17137   primary objective was to stabilize the C++ ABI; we believe that the
17138   interface to the compiler and the C++ standard library are now
17139   relatively stable.
17140
17141   Be aware that C++ code compiled by GCC 3.2.x will (in general) not
17142   interoperate with code compiled by GCC 3.1.1 or earlier.
17143
17144   Please refer to our [2]detailed list of news, caveats, and bug-fixes
17145   for further information.
17146
17147Release History
17148
17149   GCC 3.2.3
17150          April 25, 2003 ([3]changes)
17151
17152   GCC 3.2.2
17153          February 5, 2003 ([4]changes)
17154
17155   GCC 3.2.1
17156          November 19, 2002 ([5]changes)
17157
17158   GCC 3.2
17159          August 14, 2002 ([6]changes)
17160
17161References and Acknowledgements
17162
17163   GCC used to stand for the GNU C Compiler, but since the compiler
17164   supports several other languages aside from C, it now stands for the
17165   GNU Compiler Collection.
17166
17167   A list of [7]successful builds is updated as new information becomes
17168   available.
17169
17170   The GCC developers would like to thank the numerous people that have
17171   contributed new features, improvements, bug fixes, and other changes as
17172   well as test results to GCC. This [8]amazing group of volunteers is
17173   what makes GCC successful.
17174
17175   For additional information about GCC please refer to the [9]GCC project
17176   web site or contact the [10]GCC development mailing list.
17177
17178   To obtain GCC please use [11]our mirror sites, or our CVS server.
17179
17180
17181    For questions related to the use of GCC, please consult these web
17182    pages and the [12]GCC manuals. If that fails, the
17183    [13]gcc-help@gcc.gnu.org mailing list might help. Comments on these
17184    web pages and the development of GCC are welcome on our developer
17185    list at [14]gcc@gcc.gnu.org. All of [15]our lists have public
17186    archives.
17187
17188   Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and
17189   distribution of this entire article is permitted in any medium,
17190   provided this notice is preserved.
17191
17192   These pages are [17]maintained by the GCC team. Last modified
17193   2019-11-28[18].
17194
17195References
17196
17197   1. http://www.gnu.org/
17198   2. http://gcc.gnu.org/gcc-3.2/changes.html
17199   3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
17200   4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2
17201   5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1
17202   6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2
17203   7. http://gcc.gnu.org/gcc-3.2/buildstat.html
17204   8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
17205   9. http://gcc.gnu.org/index.html
17206  10. mailto:gcc@gcc.gnu.org
17207  11. http://gcc.gnu.org/mirrors.html
17208  12. https://gcc.gnu.org/onlinedocs/
17209  13. mailto:gcc-help@gcc.gnu.org
17210  14. mailto:gcc@gcc.gnu.org
17211  15. https://gcc.gnu.org/lists.html
17212  16. https://www.fsf.org/
17213  17. https://gcc.gnu.org/about.html
17214  18. http://validator.w3.org/check/referer
17215======================================================================
17216http://gcc.gnu.org/gcc-3.2/changes.html
17217                             GCC 3.2 Release Series
17218                        Changes, New Features, and Fixes
17219
17220   The latest release in the 3.2 release series is [1]GCC 3.2.3.
17221
17222Caveats and New Features
17223
17224  Caveats
17225
17226     * The C++ compiler does not correctly zero-initialize
17227       pointers-to-data members. You must explicitly initialize them. For
17228       example: int S::*m(0); will work, but depending on
17229       default-initialization to zero will not work. This bug cannot be
17230       fixed in GCC 3.2 without inducing unacceptable risks. It will be
17231       fixed in GCC 3.3.
17232     * This GCC release is based on the GCC 3.1 sourcebase, and thus has
17233       all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has
17234       a number of C++ ABI fixes which make its C++ compiler generate
17235       binary code which is incompatible with the C++ compilers found in
17236       earlier GCC releases, including GCC 3.1 and GCC 3.1.1.
17237
17238  Frontend Enhancements
17239
17240    C/C++/Objective-C
17241
17242     * The method of constructing the list of directories to be searched
17243       for header files has been revised. If a directory named by a -I
17244       option is a standard system include directory, the option is
17245       ignored to ensure that the default search order for system
17246       directories and the special treatment of system header files are
17247       not defeated.
17248     * The C and Objective-C compilers no longer accept the "Naming Types"
17249       extension (typedef foo = bar); it was already unavailable in C++.
17250       Code which uses it will need to be changed to use the "typeof"
17251       extension instead: typedef typeof(bar) foo. (We have removed this
17252       extension without a period of deprecation because it has caused the
17253       compiler to crash since version 3.0 and no one noticed until very
17254       recently. Thus we conclude it is not in widespread use.)
17255
17256    C++
17257
17258     * GCC 3.2 fixed serveral differences between the C++ ABI implemented
17259       in GCC and the multi-vendor standard, but more have been found
17260       since the release. 3.2.1 adds a new warning, -Wabi, to warn about
17261       code which is affected by these bugs. We will fix these bugs in
17262       some future release, once we are confident that all have been
17263       found; until then, it is our intention to make changes to the ABI
17264       only if they are necessary for correct compilation of C++, as
17265       opposed to conformance to the ABI documents.
17266     * For details on how to build an ABI compliant compiler for GNU/Linux
17267       systems, check the [3]common C++ ABI page.
17268
17269  New Targets and Target Specific Improvements
17270
17271    IA-32
17272
17273     * Fixed a number of bugs in SSE and MMX intrinsics.
17274     * Fixed common compiler crashes with SSE instruction set enabled
17275       (implied by -march=pentium3, pentium4, athlon-xp)
17276     * __m128 and __m128i is not 128bit aligned when used in structures.
17277
17278    x86-64
17279
17280     * A bug whereby the compiler could generate bad code for bzero has
17281       been fixed.
17282     * ABI fixes (implying ABI incompatibilities with previous version in
17283       some corner cases)
17284     * Fixed prefetch code generation
17285     __________________________________________________________________
17286
17287GCC 3.2.3
17288
17289   3.2.3 is a bug fix release only; there are no new features that were
17290   not present in GCC 3.2.2.
17291
17292  Bug Fixes
17293
17294   This section lists the problem reports (PRs) from GCC's bug tracking
17295   system that are known to be fixed in the 3.2.3 release. This list might
17296   not be complete (that is, it is possible that some PRs that have been
17297   fixed are not listed here), and some of the titles have been changed to
17298   make them more clear.
17299
17300    Internal Compiler Errors (multi-platform)
17301
17302     * [4]3782: (c++) -quiet -fstats produces a segmentation fault in
17303       cc1plus
17304     * [5]6440: (c++) template specializations cause ICE
17305     * [6]7050: (c++) ICE on: (i ? get_string() : throw)
17306     * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c)
17307     * [8]7982: (c++) ICE due to infinite recursion (using STL set)
17308     * [9]8068: exceedingly high (infinite) memory usage
17309     * [10]8178: ICE with __builtin_ffs
17310     * [11]8396: ICE in copy_to_mode_reg, in explow.c
17311     * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c
17312     * [13]9768: ICE when optimizing inline code at -O2
17313     * [14]9798: (c++) Infinite recursion (segfault) in
17314       cp/decl.c:push_using_directive with recursive using directives
17315     * [15]9799: mismatching structure initializer with nested flexible
17316       array member: ICE
17317     * [16]9928: ICE on duplicate enum declaration
17318     * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects
17319       sparc, alpha)
17320     * [18]10352: ICE in find_reloads_toplev
17321     * [19]10336: ICE with -Wunreachable-code
17322
17323    C/optimizer bugs:
17324
17325     * [20]8224: Incorrect joining of signed and unsigned division
17326     * [21]8613: -O2 produces wrong code with builtin strlen and
17327       postincrements
17328     * [22]8828: gcc reports some code is unreachable when it is not
17329     * [23]9226: GCSE breaking argument passing
17330     * [24]9853: miscompilation of non-constant structure initializer
17331     * [25]9797: C99-style struct initializers are miscompiled
17332     * [26]9967: Some standard C function calls should not be replaced
17333       when optimizing for size
17334     * [27]10116: ce2: invalid merge of join_bb in the context of switch
17335       statements
17336     * [28]10171: wrong code for inlined function
17337     * [29]10175: -Wunreachable-code doesn't work for single lines
17338
17339    C++ compiler and library:
17340
17341     * [30]8316: Confusing diagnostic for code that misuses conversion
17342       operators
17343     * [31]9169: filebuf output fails if codecvt<>::out returns noconv
17344     * [32]9420: incomplete type incorrectly reported
17345     * [33]9459: typeof in return type specification of template not
17346       supported
17347     * [34]9507: filebuf::open handles ios_base::ate incorrectly
17348     * [35]9538: Out-of-bounds memory access in streambuf::sputbackc
17349     * [36]9602: Total confusion about template/friend/virtual/abstract
17350     * [37]9993: destructor not called for local object created within and
17351       returned from infinite loop
17352     * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2
17353       system
17354
17355    Java compiler and library:
17356
17357     * [39]9652: libgcj build fails on irix6.5.1[78]
17358     * [40]10144: gas on solaris complains about bad .stabs lines for
17359       java, native as unaffected
17360
17361    x86-specific (Intel/AMD):
17362
17363     * [41]8746: gcc miscompiles Linux kernel ppa driver on x86
17364     * [42]9888: -mcpu=k6 -Os produces out of range loop instructions
17365     * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu
17366       failed
17367     * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib)
17368       failed
17369
17370    SPARC-specific:
17371
17372     * [45]7784: [Sparc] ICE in extract_insn, in recog.c
17373     * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in
17374       unroll.c
17375     * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc
17376     * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in
17377       execute/loop-2d.c
17378     * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc
17379     * [50]9414: Scheduling bug on Ultrasparc
17380     * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64
17381
17382    m68k-specific:
17383
17384     * [52]7248: broken "inclusive or" code
17385     * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1
17386
17387    PowerPC-specific:
17388
17389     * [54]9732: Wrong code with -O2 -fPIC
17390     * [55]10073: ICE: powerpc cannot split insn
17391
17392    Alpha-specific:
17393
17394     * [56]7702: optimization problem on a DEC alpha under OSF1
17395     * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system
17396
17397    HP-specific:
17398
17399     * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275)
17400     * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10
17401       (missing symbol)
17402     * [60]10271: Floating point args don't get reloaded across function
17403       calls with -O2
17404
17405    MIPS specific:
17406
17407     * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in
17408       compile/920501-4.c
17409
17410    CRIS specific:
17411
17412     * [62]10377: gcc-3.2.2 creates bad assembler code for cris
17413
17414    Miscellaneous and minor bugs:
17415
17416     * [63]6955: collect2 says "core dumped" when there is no core
17417     __________________________________________________________________
17418
17419GCC 3.2.2
17420
17421   Beginning with 3.2.2, GCC's Makefile suite supports redirection of make
17422   install by means of the DESTDIR variable. Parts of the GCC tree have
17423   featured that support long before, but now it is available even from
17424   the top level.
17425
17426   Other than that, GCC 3.2.2 is a bug fix release only; there are no new
17427   features that were not present in GCC 3.2.1.
17428
17429  Bug Fixes
17430
17431   On the following i386-based systems GCC 3.2.1 broke the C ABI wrt.
17432   functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped
17433   with FreeBSD 5.0 does not have this problem), Interix, a.out-based
17434   GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI
17435   change, and thus restores ABI-compatibility with previous releases
17436   (except GCC 3.2.1) on these platforms.
17437
17438   This section lists the problem reports (PRs) from GCC's bug tracking
17439   system that are known to be fixed in the 3.2.2 release. This list might
17440   not be complete (that is, it is possible that some PRs that have been
17441   fixed are not listed here) and some of the titles have been changed to
17442   make them more clear.
17443
17444    Internal Compiler Errors (multi-platform)
17445
17446     * [64]5919: (c++) ICE when passing variable array to template
17447       function
17448     * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=)
17449     * [66]7507: ICE with -O2 when address of called function is a
17450       complicated expression
17451     * [67]7622: ICE with nested inline functions if function's address is
17452       taken
17453     * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR
17454       [69]9258)
17455     * [70]8031: (c++) ICE in code comparing typeids and casting from
17456       virtual base
17457     * [71]8275: ICE in simplify_subreg
17458     * [72]8332: (c++) builtin strlen/template interaction causes ICE
17459     * [73]8372: (c++) ICE on explicit call of destructor
17460     * [74]8439: (c, not c++) empty struct causes ICE
17461     * [75]8442: (c++) ICE with nested template classes
17462     * [76]8518: ICE when compiling mplayer ("extern inline" issue)
17463     * [77]8615: (c++) ICE with out-of-range character constant template
17464       argument
17465     * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307
17466     * [79]8799: (c++) ICE: error reporting routines re-entered
17467     * [80]9328: (c++) ICE with typeof(X) for overloaded X
17468     * [81]9465: (preprocessor) cpp -traditional ICE on null bytes
17469
17470    C++ (compiler and library) bugs
17471
17472     * [82]47: scoping in nested classes is broken
17473     * [83]6745: problems with iostream rdbuf() member function
17474     * [84]8214: conversion from const char* const to char* sometimes
17475       accepted illegally
17476     * [85]8493: builtin strlen and overload resolution (same bug as
17477       [86]8332)
17478     * [87]8503: strange behaviour of function types
17479     * [88]8727: compiler confused by inheritance from an anonymous struct
17480     * [89]7445: poor performance of std::locale::classic() in
17481       multi-threaded applications
17482     * [90]8230: mishandling of overflow in vector<T>::resize
17483     * [91]8399: sync_with_stdio(false) breaks unformatted input
17484     * [92]8662: illegal access of private member of unnamed class is
17485       accepted
17486     * [93]8707: "make distclean" fails in libstdc++-v3 directory
17487     * [94]8708: __USE_MALLOC doesn't work
17488     * [95]8790: Use of non-thread-safe strtok in src/localename.cc
17489     * [96]8887: Bug in date formats with --enable-clocale=generic
17490     * [97]9076: Call Frame Instructions are not handled correctly during
17491       unwind operation
17492     * [98]9151: std::setprecision limited to 16 digits when outputting a
17493       double to a stream
17494     * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers
17495     * [100]9269: libstdc++ headers: explicit specialization of function
17496       must precede its first use
17497     * [101]9322: return value of basic_streambuf<>::getloc affected by
17498       locale::global
17499     * [102]9433: segfault in runtime support for dynamic_cast
17500
17501    C and optimizer bugs
17502
17503     * [103]8032: GCC incorrectly initializes static structs that have
17504       flexible arrays
17505     * [104]8639: simple arithmetic expression broken
17506     * [105]8794: optimization improperly eliminates certain expressions
17507     * [106]8832: traditional "asm volatile" code is illegally optimized
17508     * [107]8988: loop optimizer bug: with -O2, code is generated that
17509       segfaults (found on i386, bug present for all platforms)
17510     * [108]9492: structure copy clobbers subsequent stores to structure
17511
17512    Objective-C bugs
17513
17514     * [109]9267: Objective-C parser won't build with newer bison versions
17515       (e.g. 1.875)
17516
17517    Ada bugs
17518
17519     * [110]8344: Ada build problem due to conflict between gcc/final.o,
17520       gcc/ada/final.o
17521
17522    Preprocessor bugs
17523
17524     * [111]8524: _Pragma within macros is improperly expanded
17525     * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with
17526       -fshort-wchar
17527
17528    ARM-specific
17529
17530     * [113]9090: arm ICE with >= -O2; regression from gcc-2.95
17531
17532    x86-specific (Intel/AMD)
17533
17534     * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction)
17535     * [115]8599: loop unroll bug with -march=k6-3
17536     * [116]9506: ABI breakage in structure return (affects BSD and
17537       Cygwin, but not GNU/Linux)
17538
17539    FreeBSD 5.0 specific
17540
17541     * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0
17542
17543    RTEMS-specific
17544
17545     * [118]9292: hppa1.1-rtems configurery problems
17546     * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug
17547     * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue
17548     * [121]9296: gthr-rtems regression
17549     * [122]9316: powerpc-rtems: extending multilibs
17550
17551    HP-PA specific
17552
17553     * [123]9493: ICE with -O2 when building a simple function
17554
17555    Documentation
17556
17557     * [124]7341: hyperlink to gcov in GCC documentation doesn't work
17558     * [125]8947: Please add a warning about "-malign-double" in docs
17559     * [126]7448, [127]8882: typo cleanups
17560     __________________________________________________________________
17561
17562GCC 3.2.1
17563
17564   3.2.1 adds a new warning, -Wabi. This option warns when GNU C++
17565   generates code that is known not to be binary-compatible with the
17566   vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included
17567   in the distribution, for details.
17568
17569   This release also removes an old GCC extension, "naming types", and the
17570   documentation now directs users to use a different GCC extension,
17571   __typeof__, instead. The feature had evidently been broken for a while.
17572
17573   Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and
17574   the new warning there are no new features that were not present in GCC
17575   3.2.
17576
17577   In addition, the previous fix for [128]PR 7445 (poor performance of
17578   std::locale::classic() in multi-threaded applications) was reverted
17579   ("unfixed"), because the "fix" was not thread-safe.
17580
17581  Bug Fixes
17582
17583   This section lists the problem reports (PRs) from GCC's bug tracking
17584   system that are known to be fixed in the 3.2.1 release. This list might
17585   not be complete (that is, it is possible that some PRs that have been
17586   fixed are not listed here). As you can see, the number of bug fixes is
17587   quite large, so it is strongly recommended that users of earlier GCC
17588   3.x releases upgrade to GCC 3.2.1.
17589
17590    Internal Compiler Errors (multi-platform)
17591
17592     * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c
17593     * [130]5661: (c++) ICE instantiating template on array of unknown
17594       size (bad code)
17595     * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on
17596       64-bit platforms
17597     * [132]6994: (c++) ICE in find_function_data
17598     * [133]7150: preprocessor: GCC -dM -E gives an ICE
17599     * [134]7160: ICE when optimizing branches without a return value
17600     * [135]7228: (c++) ICE when using member template and template
17601       function
17602     * [136]7266: (c++) ICE with -pedantic on missing typename
17603     * [137]7353: ICE from use of "Naming Types" extension, see above
17604     * [138]7411: ICE in instantiate_virtual_regs_1, in function.c
17605     * [139]7478: (c++) ICE on static_cast inside template
17606     * [140]7526: preprocessor core dump when _Pragma implies #pragma
17607       dependency
17608     * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803
17609       is a duplicate)
17610     * [143]7754: (c++) ICE on union with template parameter
17611     * [144]7788: (c++) redeclaring a definition as an incomplete class
17612       causes ICE
17613     * [145]8031: (c++) ICE in comptypes, in cp/typeck.c
17614     * [146]8055: preprocessor dies with SIG11 when building FreeBSD
17615       kernel
17616     * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related
17617       variables
17618     * [148]8134: (c++) ICE in force_store_init_value on legal code
17619     * [149]8149: (c++) ICE on incomplete type
17620     * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array
17621       initialization
17622
17623    C++ (compiler and library) bugs
17624
17625     * [151]5607: No pointer adjustment in covariant return types
17626     * [152]6579: Infinite loop with statement expressions in member
17627       initialization
17628     * [153]6803: Default copy constructor bug in GCC 3.1
17629     * [154]7176: g++ confused by friend and static member with same name
17630     * [155]7188: Segfault with template class and recursive (incorrect)
17631       initializer list
17632     * [156]7306: Regression: GCC 3.x fails to compile code with virtual
17633       inheritance if a method has a variable number of arguments
17634     * [157]7461: ctype<char>::classic_table() returns offset array on
17635       Cygwin
17636     * [158]7524: f(const float arg[3]) fails
17637     * [159]7584: Erroneous ambiguous base error on using declaration
17638     * [160]7676: Member template overloading problem
17639     * [161]7679: infinite loop when a right parenthesis is missing
17640     * [162]7811: default locale not taken from environment
17641     * [163]7961: compare( char *) implemented incorrectly in
17642       basic_string<>
17643     * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if
17644       streambuf::underflow() leaves gptr() NULL (dups: [165]8127,
17645       [166]6745)
17646     * [167]8096: deque::at() throws std::range_error instead of
17647       std::out_of_range
17648     * [168]8127: cout << cin.rdbuf() infinite loop
17649     * [169]8218: Excessively large memory consumed for classes with large
17650       array members
17651     * [170]8287: GCC 3.2: Destructor called for non-constructed local
17652       object
17653     * [171]8347: empty vector range used in string construction causes
17654       core dump
17655     * [172]8348: fail() flag is set in istringstream when eof() flag is
17656       set
17657     * [173]8391: regression: infinite loop in cp/decl2.c(finish_file)
17658
17659    C and optimizer bugs
17660
17661     * [174]6627: -fno-align-functions doesn't seem to disable function
17662       alignment
17663     * [175]6631: life_analysis misoptimizes code to initialize fields of
17664       a structure
17665     * [176]7102: unsigned char division results in floating exception
17666     * [177]7120: Run once loop should *always* be unrolled
17667       (pessimization)
17668     * [178]7209: Bug involving array referencing and ?: operator
17669     * [179]7515: invalid inlining of global function with -O3
17670     * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test
17671     * [181]8467: bug in sibling call optimization
17672
17673    Preprocessor bugs
17674
17675     * [182]4890: incorrect line markers from the traditional preprocessor
17676     * [183]7357: -M option omits system headers files (making it the same
17677       as -MM)
17678     * [184]7358: Changes to Sun's make Dependencies
17679     * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as
17680       C headers
17681     * [186]7862: Interrupting GCC -MD removes .d file but not .o
17682     * [187]8190: Failed compilation deletes -MD dependency file
17683     * [188]8524: _Pragma within macro is improperly expanded
17684
17685    x86 specific (Intel/AMD)
17686
17687     * [189]5351: (i686-only) function pass-by-value structure copy
17688       corrupts stack ([190]7591 is a duplicate)
17689     * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with
17690       -march=pentium3/pentium2/athlon (these are all the same underlying
17691       bug, in MMX register use)
17692     * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same
17693       as above?)
17694     * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken
17695     * [199]6981: wrong code in 64-bit manipulation on x86
17696     * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__
17697       macro
17698     * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE
17699       intrinsics are broken
17700     * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with
17701       -march=pentium4
17702     * [203]7693: Typo in i386 mmintrin.h header
17703     * [204]7723: ICE - Pentium3 sse - GCC 3.2
17704     * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse
17705     * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3
17706
17707    PowerPC specific
17708
17709     * [207]5967: GCC bug when profiling nested functions on powerpc
17710     * [208]6984: wrong code generated with -O2, -O3, -Os for do-while
17711       loop on PowerPC
17712     * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5
17713     * [210]7130: miscompiled code for GCC-3.1 on
17714       powerpc-unknown-linux-gnu with -funroll-all-loops
17715     * [211]7133: PowerPC ICE: unrecognizable insn
17716     * [212]7380: ICE in extract_insn, at recog.c:2148
17717     * [213]8252: ICE on Altivec code with optimization turned on
17718     * [214]8451: Altivec ICE in GCC 3.2
17719
17720    HP/PA specific
17721
17722     * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa
17723
17724    SPARC specific
17725
17726     * [216]6668: when using --disable-multilib, libgcc_s.so is installed
17727       in the wrong place on sparc-solaris
17728     * [217]7151: ICE when compiling for UltraSPARC
17729     * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long
17730       double and -O1
17731     * [219]7842: [REGRESSION] SPARC code gen bug
17732
17733    ARM specific
17734
17735     * [220]7856: [arm] invalid offset in constant pool reference
17736     * [221]7967: optimization produces wrong code (ARM)
17737
17738    Alpha specific
17739
17740     * [222]7374: __builtin_fabsl broken on alpha
17741
17742    IBM s390 specific
17743
17744     * [223]7370: ICE in fixup_var_refs_1 on s390x
17745     * [224]7409: loop optimization bug on s390x-linux-gnu
17746     * [225]8232: s390x: ICE when using bcmp with int length argument
17747
17748    SCO specific
17749
17750     * [226]7623: SCO OpenServer build fails with machmode.def: undefined
17751       symbol: BITS_PER_UNIT
17752
17753    m68k/Coldfire specific
17754
17755     * [227]8314: crtbegin, crtend need to be multilib'ed for this
17756       platform
17757
17758    Documentation
17759
17760     * [228]761: Document some undocumented options
17761     * [229]5610: Fix documentation about invoking SSE instructions
17762       (-mfpmath=sse)
17763     * [230]7484: List -Wmissing-declarations as C-only option
17764     * [231]7531: -mcmodel not documented for x86-64
17765     * [232]8120: Update documentation of bad use of ##
17766     __________________________________________________________________
17767
17768GCC 3.2
17769
17770   3.2 is a small bug fix release, but there is a change to the
17771   application binary interface (ABI), hence the change to the second part
17772   of the version number.
17773
17774   The main purpose of the 3.2 release is to correct a couple of problems
17775   in the C++ ABI, with the intention of providing a stable interface
17776   going forward.  Accordingly, 3.2 is only a small change to 3.1.1.
17777
17778  Bug Fixes
17779
17780    C++
17781
17782     * [233]7320: g++ 3.2 relocation problem
17783     * [234]7470: vtable: virtual function pointers not in declaration
17784       order
17785
17786    libstdc++
17787
17788     * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t
17789     * [236]6503, [237]6642, [238]7186: Problems with comparing or
17790       subtracting various types of const and non-const iterators
17791     * [239]7216: ambiguity with basic_iostream::traits_type
17792     * [240]7220: problem with basic_istream::ignore(0,delimiter)
17793     * [241]7222: locale::operator==() doesn't work on std::locale("")
17794     * [242]7286: placement operator delete issue
17795     * [243]7442: cxxabi.h does not match the C++ ABI
17796     * [244]7445: poor performance of std::locale::classic() in
17797       multi-threaded applications
17798
17799    x86-64 specific
17800
17801     * [245]7291: off-by-one in generated inline bzero code for x86-64
17802
17803
17804    For questions related to the use of GCC, please consult these web
17805    pages and the [246]GCC manuals. If that fails, the
17806    [247]gcc-help@gcc.gnu.org mailing list might help. Comments on these
17807    web pages and the development of GCC are welcome on our developer
17808    list at [248]gcc@gcc.gnu.org. All of [249]our lists have public
17809    archives.
17810
17811   Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and
17812   distribution of this entire article is permitted in any medium,
17813   provided this notice is preserved.
17814
17815   These pages are [251]maintained by the GCC team. Last modified
17816   2019-11-28[252].
17817
17818References
17819
17820   1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
17821   2. http://gcc.gnu.org/gcc-3.1/changes.html
17822   3. http://gcc.gnu.org/gcc-3.2/c++-abi.html
17823   4. https://gcc.gnu.org/PR3782
17824   5. https://gcc.gnu.org/PR6440
17825   6. https://gcc.gnu.org/PR7050
17826   7. https://gcc.gnu.org/PR7741
17827   8. https://gcc.gnu.org/PR7982
17828   9. https://gcc.gnu.org/PR8068
17829  10. https://gcc.gnu.org/PR8178
17830  11. https://gcc.gnu.org/PR8396
17831  12. https://gcc.gnu.org/PR8674
17832  13. https://gcc.gnu.org/PR9768
17833  14. https://gcc.gnu.org/PR9798
17834  15. https://gcc.gnu.org/PR9799
17835  16. https://gcc.gnu.org/PR9928
17836  17. https://gcc.gnu.org/PR10114
17837  18. https://gcc.gnu.org/PR10352
17838  19. https://gcc.gnu.org/PR10336
17839  20. https://gcc.gnu.org/PR8224
17840  21. https://gcc.gnu.org/PR8613
17841  22. https://gcc.gnu.org/PR8828
17842  23. https://gcc.gnu.org/PR9226
17843  24. https://gcc.gnu.org/PR9853
17844  25. https://gcc.gnu.org/PR9797
17845  26. https://gcc.gnu.org/PR9967
17846  27. https://gcc.gnu.org/PR10116
17847  28. https://gcc.gnu.org/PR10171
17848  29. https://gcc.gnu.org/PR10175
17849  30. https://gcc.gnu.org/PR8316
17850  31. https://gcc.gnu.org/PR9169
17851  32. https://gcc.gnu.org/PR9420
17852  33. https://gcc.gnu.org/PR9459
17853  34. https://gcc.gnu.org/PR9507
17854  35. https://gcc.gnu.org/PR9538
17855  36. https://gcc.gnu.org/PR9602
17856  37. https://gcc.gnu.org/PR9993
17857  38. https://gcc.gnu.org/PR10167
17858  39. https://gcc.gnu.org/PR9652
17859  40. https://gcc.gnu.org/PR10144
17860  41. https://gcc.gnu.org/PR8746
17861  42. https://gcc.gnu.org/PR9888
17862  43. https://gcc.gnu.org/PR9638
17863  44. https://gcc.gnu.org/PR9954
17864  45. https://gcc.gnu.org/PR7784
17865  46. https://gcc.gnu.org/PR7796
17866  47. https://gcc.gnu.org/PR8281
17867  48. https://gcc.gnu.org/PR8366
17868  49. https://gcc.gnu.org/PR8726
17869  50. https://gcc.gnu.org/PR9414
17870  51. https://gcc.gnu.org/PR10067
17871  52. https://gcc.gnu.org/PR7248
17872  53. https://gcc.gnu.org/PR8343
17873  54. https://gcc.gnu.org/PR9732
17874  55. https://gcc.gnu.org/PR10073
17875  56. https://gcc.gnu.org/PR7702
17876  57. https://gcc.gnu.org/PR9671
17877  58. https://gcc.gnu.org/PR8694
17878  59. https://gcc.gnu.org/PR9953
17879  60. https://gcc.gnu.org/PR10271
17880  61. https://gcc.gnu.org/PR6362
17881  62. https://gcc.gnu.org/PR10377
17882  63. https://gcc.gnu.org/PR6955
17883  64. https://gcc.gnu.org/PR5919
17884  65. https://gcc.gnu.org/PR7129
17885  66. https://gcc.gnu.org/PR7507
17886  67. https://gcc.gnu.org/PR7622
17887  68. https://gcc.gnu.org/PR7681
17888  69. https://gcc.gnu.org/PR9528
17889  70. https://gcc.gnu.org/PR8031
17890  71. https://gcc.gnu.org/PR8275
17891  72. https://gcc.gnu.org/PR8332
17892  73. https://gcc.gnu.org/PR8372
17893  74. https://gcc.gnu.org/PR8439
17894  75. https://gcc.gnu.org/PR8442
17895  76. https://gcc.gnu.org/PR8518
17896  77. https://gcc.gnu.org/PR8615
17897  78. https://gcc.gnu.org/PR8663
17898  79. https://gcc.gnu.org/PR8799
17899  80. https://gcc.gnu.org/PR9328
17900  81. https://gcc.gnu.org/PR9465
17901  82. https://gcc.gnu.org/PR47
17902  83. https://gcc.gnu.org/PR6745
17903  84. https://gcc.gnu.org/PR8214
17904  85. https://gcc.gnu.org/PR8493
17905  86. https://gcc.gnu.org/PR8332
17906  87. https://gcc.gnu.org/PR8503
17907  88. https://gcc.gnu.org/PR8727
17908  89. https://gcc.gnu.org/PR7445
17909  90. https://gcc.gnu.org/PR8230
17910  91. https://gcc.gnu.org/PR8399
17911  92. https://gcc.gnu.org/PR8662
17912  93. https://gcc.gnu.org/PR8707
17913  94. https://gcc.gnu.org/PR8708
17914  95. https://gcc.gnu.org/PR8790
17915  96. https://gcc.gnu.org/PR8887
17916  97. https://gcc.gnu.org/PR9076
17917  98. https://gcc.gnu.org/PR9151
17918  99. https://gcc.gnu.org/PR9168
17919 100. https://gcc.gnu.org/PR9269
17920 101. https://gcc.gnu.org/PR9322
17921 102. https://gcc.gnu.org/PR9433
17922 103. https://gcc.gnu.org/PR8032
17923 104. https://gcc.gnu.org/PR8639
17924 105. https://gcc.gnu.org/PR8794
17925 106. https://gcc.gnu.org/PR8832
17926 107. https://gcc.gnu.org/PR8988
17927 108. https://gcc.gnu.org/PR9492
17928 109. https://gcc.gnu.org/PR9267
17929 110. https://gcc.gnu.org/PR8344
17930 111. https://gcc.gnu.org/PR8524
17931 112. https://gcc.gnu.org/PR8880
17932 113. https://gcc.gnu.org/PR9090
17933 114. https://gcc.gnu.org/PR8588
17934 115. https://gcc.gnu.org/PR8599
17935 116. https://gcc.gnu.org/PR9506
17936 117. https://gcc.gnu.org/PR9484
17937 118. https://gcc.gnu.org/PR9292
17938 119. https://gcc.gnu.org/PR9293
17939 120. https://gcc.gnu.org/PR9295
17940 121. https://gcc.gnu.org/PR9296
17941 122. https://gcc.gnu.org/PR9316
17942 123. https://gcc.gnu.org/PR9493
17943 124. https://gcc.gnu.org/PR7341
17944 125. https://gcc.gnu.org/PR8947
17945 126. https://gcc.gnu.org/PR7448
17946 127. https://gcc.gnu.org/PR8882
17947 128. https://gcc.gnu.org/PR7445
17948 129. https://gcc.gnu.org/PR2521
17949 130. https://gcc.gnu.org/PR5661
17950 131. https://gcc.gnu.org/PR6419
17951 132. https://gcc.gnu.org/PR6994
17952 133. https://gcc.gnu.org/PR7150
17953 134. https://gcc.gnu.org/PR7160
17954 135. https://gcc.gnu.org/PR7228
17955 136. https://gcc.gnu.org/PR7266
17956 137. https://gcc.gnu.org/PR7353
17957 138. https://gcc.gnu.org/PR7411
17958 139. https://gcc.gnu.org/PR7478
17959 140. https://gcc.gnu.org/PR7526
17960 141. https://gcc.gnu.org/PR7721
17961 142. https://gcc.gnu.org/PR7803
17962 143. https://gcc.gnu.org/PR7754
17963 144. https://gcc.gnu.org/PR7788
17964 145. https://gcc.gnu.org/PR8031
17965 146. https://gcc.gnu.org/PR8055
17966 147. https://gcc.gnu.org/PR8067
17967 148. https://gcc.gnu.org/PR8134
17968 149. https://gcc.gnu.org/PR8149
17969 150. https://gcc.gnu.org/PR8160
17970 151. https://gcc.gnu.org/PR5607
17971 152. https://gcc.gnu.org/PR6579
17972 153. https://gcc.gnu.org/PR6803
17973 154. https://gcc.gnu.org/PR7176
17974 155. https://gcc.gnu.org/PR7188
17975 156. https://gcc.gnu.org/PR7306
17976 157. https://gcc.gnu.org/PR7461
17977 158. https://gcc.gnu.org/PR7524
17978 159. https://gcc.gnu.org/PR7584
17979 160. https://gcc.gnu.org/PR7676
17980 161. https://gcc.gnu.org/PR7679
17981 162. https://gcc.gnu.org/PR7811
17982 163. https://gcc.gnu.org/PR7961
17983 164. https://gcc.gnu.org/PR8071
17984 165. https://gcc.gnu.org/PR8127
17985 166. https://gcc.gnu.org/PR6745
17986 167. https://gcc.gnu.org/PR8096
17987 168. https://gcc.gnu.org/PR8127
17988 169. https://gcc.gnu.org/PR8218
17989 170. https://gcc.gnu.org/PR8287
17990 171. https://gcc.gnu.org/PR8347
17991 172. https://gcc.gnu.org/PR8348
17992 173. https://gcc.gnu.org/PR8391
17993 174. https://gcc.gnu.org/PR6627
17994 175. https://gcc.gnu.org/PR6631
17995 176. https://gcc.gnu.org/PR7102
17996 177. https://gcc.gnu.org/PR7120
17997 178. https://gcc.gnu.org/PR7209
17998 179. https://gcc.gnu.org/PR7515
17999 180. https://gcc.gnu.org/PR7814
18000 181. https://gcc.gnu.org/PR8467
18001 182. https://gcc.gnu.org/PR4890
18002 183. https://gcc.gnu.org/PR7357
18003 184. https://gcc.gnu.org/PR7358
18004 185. https://gcc.gnu.org/PR7602
18005 186. https://gcc.gnu.org/PR7862
18006 187. https://gcc.gnu.org/PR8190
18007 188. https://gcc.gnu.org/PR8524
18008 189. https://gcc.gnu.org/PR5351
18009 190. https://gcc.gnu.org/PR7591
18010 191. https://gcc.gnu.org/PR6845
18011 192. https://gcc.gnu.org/PR7034
18012 193. https://gcc.gnu.org/PR7124
18013 194. https://gcc.gnu.org/PR7174
18014 195. https://gcc.gnu.org/PR7134
18015 196. https://gcc.gnu.org/PR7375
18016 197. https://gcc.gnu.org/PR7390
18017 198. https://gcc.gnu.org/PR6890
18018 199. https://gcc.gnu.org/PR6981
18019 200. https://gcc.gnu.org/PR7242
18020 201. https://gcc.gnu.org/PR7396
18021 202. https://gcc.gnu.org/PR7630
18022 203. https://gcc.gnu.org/PR7693
18023 204. https://gcc.gnu.org/PR7723
18024 205. https://gcc.gnu.org/PR7951
18025 206. https://gcc.gnu.org/PR8146
18026 207. https://gcc.gnu.org/PR5967
18027 208. https://gcc.gnu.org/PR6984
18028 209. https://gcc.gnu.org/PR7114
18029 210. https://gcc.gnu.org/PR7130
18030 211. https://gcc.gnu.org/PR7133
18031 212. https://gcc.gnu.org/PR7380
18032 213. https://gcc.gnu.org/PR8252
18033 214. https://gcc.gnu.org/PR8451
18034 215. https://gcc.gnu.org/PR7250
18035 216. https://gcc.gnu.org/PR6668
18036 217. https://gcc.gnu.org/PR7151
18037 218. https://gcc.gnu.org/PR7335
18038 219. https://gcc.gnu.org/PR7842
18039 220. https://gcc.gnu.org/PR7856
18040 221. https://gcc.gnu.org/PR7967
18041 222. https://gcc.gnu.org/PR7374
18042 223. https://gcc.gnu.org/PR7370
18043 224. https://gcc.gnu.org/PR7409
18044 225. https://gcc.gnu.org/PR8232
18045 226. https://gcc.gnu.org/PR7623
18046 227. https://gcc.gnu.org/PR8314
18047 228. https://gcc.gnu.org/PR761
18048 229. https://gcc.gnu.org/PR5610
18049 230. https://gcc.gnu.org/PR7484
18050 231. https://gcc.gnu.org/PR7531
18051 232. https://gcc.gnu.org/PR8120
18052 233. https://gcc.gnu.org/PR7320
18053 234. https://gcc.gnu.org/PR7470
18054 235. https://gcc.gnu.org/PR6410
18055 236. https://gcc.gnu.org/PR6503
18056 237. https://gcc.gnu.org/PR6642
18057 238. https://gcc.gnu.org/PR7186
18058 239. https://gcc.gnu.org/PR7216
18059 240. https://gcc.gnu.org/PR7220
18060 241. https://gcc.gnu.org/PR7222
18061 242. https://gcc.gnu.org/PR7286
18062 243. https://gcc.gnu.org/PR7442
18063 244. https://gcc.gnu.org/PR7445
18064 245. https://gcc.gnu.org/PR7291
18065 246. https://gcc.gnu.org/onlinedocs/
18066 247. mailto:gcc-help@gcc.gnu.org
18067 248. mailto:gcc@gcc.gnu.org
18068 249. https://gcc.gnu.org/lists.html
18069 250. https://www.fsf.org/
18070 251. https://gcc.gnu.org/about.html
18071 252. http://validator.w3.org/check/referer
18072======================================================================
18073http://gcc.gnu.org/gcc-3.1/index.html
18074                                    GCC 3.1
18075
18076   (This release series is no longer supported.)
18077
18078   July 27, 2002
18079
18080   The [1]GNU project and the GCC developers are pleased to announce the
18081   release of GCC 3.1.1.
18082
18083   The links below still apply to GCC 3.1.1.
18084
18085   May 15, 2002
18086
18087   The [2]GNU project and the GCC developers are pleased to announce the
18088   release of GCC 3.1.
18089
18090   GCC used to stand for the GNU C Compiler, but since the compiler
18091   supports several other languages aside from C, it now stands for the
18092   GNU Compiler Collection.
18093
18094   A list of [3]successful builds is updated as new information becomes
18095   available.
18096
18097   The GCC developers would like to thank the numerous people that have
18098   contributed [4]new features, improvements, bug fixes, and other changes
18099   as well as test results to GCC. This [5]amazing group of volunteers is
18100   what makes GCC successful.
18101
18102   For additional information about GCC please refer to the [6]GCC project
18103   web site or contact the [7]GCC development mailing list.
18104
18105   To obtain GCC please use [8]our mirror sites, or our CVS server.
18106     __________________________________________________________________
18107
18108
18109    For questions related to the use of GCC, please consult these web
18110    pages and the [9]GCC manuals. If that fails, the
18111    [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
18112    web pages and the development of GCC are welcome on our developer
18113    list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
18114    archives.
18115
18116   Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
18117   distribution of this entire article is permitted in any medium,
18118   provided this notice is preserved.
18119
18120   These pages are [14]maintained by the GCC team. Last modified
18121   2019-11-28[15].
18122
18123References
18124
18125   1. http://www.gnu.org/
18126   2. http://www.gnu.org/
18127   3. http://gcc.gnu.org/gcc-3.1/buildstat.html
18128   4. http://gcc.gnu.org/gcc-3.1/changes.html
18129   5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
18130   6. http://gcc.gnu.org/index.html
18131   7. mailto:gcc@gcc.gnu.org
18132   8. http://gcc.gnu.org/mirrors.html
18133   9. https://gcc.gnu.org/onlinedocs/
18134  10. mailto:gcc-help@gcc.gnu.org
18135  11. mailto:gcc@gcc.gnu.org
18136  12. https://gcc.gnu.org/lists.html
18137  13. https://www.fsf.org/
18138  14. https://gcc.gnu.org/about.html
18139  15. http://validator.w3.org/check/referer
18140======================================================================
18141http://gcc.gnu.org/gcc-3.1/changes.html
18142                             GCC 3.1 Release Series
18143                        Changes, New Features, and Fixes
18144
18145Additional changes in GCC 3.1.1
18146
18147     * A bug related to how structures and unions are returned has been
18148       fixed for powerpc-*-netbsd*.
18149     * An important bug in the implementation of -fprefetch-loop-arrays
18150       has been fixed. Previously the optimization prefetched random
18151       blocks of memory for most targets except for i386.
18152     * The Java compiler now compiles Java programs much faster and also
18153       works with parallel make.
18154     * Nested functions have been fixed for mips*-*-netbsd*.
18155     * Some missing floating point support routines have beed added for
18156       mips*-*-netbsd*.
18157     * This [1]message gives additional information about the bugs fixed
18158       in this release.
18159
18160Caveats
18161
18162     * The -traditional C compiler option has been deprecated and will be
18163       removed in GCC 3.3. (It remains possible to preprocess non-C code
18164       with the traditional preprocessor.)
18165     * The default debugging format for most ELF platforms (including
18166       GNU/Linux and FreeBSD; notable exception is Solaris) has changed
18167       from stabs to DWARF2. This requires GDB 5.1.1 or later.
18168
18169General Optimizer Improvements
18170
18171     * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat,
18172       and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure
18173       for profile driven optimizations.
18174       Options -fprofile-arcs and -fbranch-probabilities can now be used
18175       to improve speed of the generated code by profiling the actual
18176       program behaviour on typical runs. In the absence of profile info
18177       the compiler attempts to guess the profile statically.
18178     * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to
18179       monitor performance of the generated code.
18180       According to the SPECInt2000 results on an AMD Athlon CPU, the code
18181       generated by GCC 3.1 is 6% faster on the average (8.2% faster with
18182       profile feedback) compared to GCC 3.0. The code produced by GCC 3.0
18183       is about 2.1% faster compared to 2.95.3. Tests were done using the
18184       -O2 -march=athlon command-line options.
18185     * Alexandre Oliva, of Red Hat, has generalized the tree inlining
18186       infrastructure developed by CodeSourcery, LLC for the C++ front
18187       end, so that it is now used in the C front end too. Inlining
18188       functions as trees exposes them earlier to the compiler, giving it
18189       more opportunities for optimization.
18190     * Support for data prefetching instructions has been added to the GCC
18191       back end and several targets. A new __builtin_prefetch intrinsic is
18192       available to explicitly insert prefetch instructions and
18193       experimental support for loop array prefetching has been added (see
18194       -fprefetch-loop-array documentation).
18195     * Support for emitting debugging information for macros has been
18196       added for DWARF2. It is activated using -g3.
18197
18198New Languages and Language specific improvements
18199
18200  C/C++
18201
18202     * A few more [4]ISO C99 features.
18203     * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0.
18204     * The preprocessor's symbol table has been merged with the symbol
18205       table of the C, C++ and Objective-C front ends.
18206     * The preprocessor consumes less memory than the preprocessor in GCC
18207       3.0, often significantly so. On normal input files, it typically
18208       consumes less memory than pre-3.0 cccp-based GCC, too.
18209
18210  C++
18211
18212     * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std
18213       was a workaround to allow std compliant code to work with the
18214       non-std compliant libstdc++-v2. libstdc++-v3 is std compliant.
18215     * The C++ ABI has been fixed so that void (A::*)() const is mangled
18216       as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only
18217       affects pointer to cv-qualified member function types.
18218     * The C++ ABI has been changed to correctly handle this code:
18219    struct A {
18220      void operator delete[] (void *, size_t);
18221    };
18222
18223    struct B : public A {
18224    };
18225
18226    new B[10];
18227
18228       The amount of storage allocated for the array will be greater than
18229       it was in 3.0, in order to store the number of elements in the
18230       array, so that the correct size can be passed to operator delete[]
18231       when the array is deleted. Previously, the value passed to operator
18232       delete[] was unpredictable.
18233       This change will only affect code that declares a two-argument
18234       operator delete[] with a second parameter of type size_t in a base
18235       class, and does not override that definition in a derived class.
18236     * The C++ ABI has been changed so that:
18237    struct A {
18238      void operator delete[] (void *, size_t);
18239      void operator delete[] (void *);
18240    };
18241
18242       does not cause unnecessary storage to be allocated when an array of
18243       A objects is allocated.
18244       This change will only affect code that declares both of these forms
18245       of operator delete[], and declared the two-argument form before the
18246       one-argument form.
18247     * The C++ ABI has been changed so that when a parameter is passed by
18248       value, any cleanup for that parameter is performed in the caller,
18249       as specified by the ia64 C++ ABI, rather than the called function
18250       as before. As a result, classes with a non-trivial destructor but a
18251       trivial copy constructor will be passed and returned by invisible
18252       reference, rather than by bitwise copy as before.
18253     * G++ now supports the "named return value optimization": for code
18254       like
18255    A f () {
18256      A a;
18257      ...
18258      return a;
18259    }
18260
18261       G++ will allocate a in the return value slot, so that the return
18262       becomes a no-op. For this to work, all return statements in the
18263       function must return the same variable.
18264     * Improvements to the C++ library are listed in [5]the libstdc++-v3
18265       FAQ.
18266
18267  Objective-C
18268
18269     * Annoying linker warnings (due to incorrect code being generated)
18270       have been fixed.
18271     * If a class method cannot be found, the compiler no longer issues a
18272       warning if a corresponding instance method exists in the root
18273       class.
18274     * Forward @protocol declarations have been fixed.
18275     * Loading of categories has been fixed in certain situations (GNU run
18276       time only).
18277     * The class lookup in the run-time library has been rewritten so that
18278       class method dispatch is more than twice as fast as it used to be
18279       (GNU run time only).
18280
18281  Java
18282
18283     * libgcj now includes RMI, java.lang.ref.*, javax.naming, and
18284       javax.transaction.
18285     * Property files and other system resources can be compiled into
18286       executables which use libgcj using the new gcj --resource feature.
18287     * libgcj has been ported to more platforms. In particular there is
18288       now a mostly-functional mingw32 (Windows) target port.
18289     * JNI and CNI invocation interfaces were implemented, so gcj-compiled
18290       Java code can now be called from a C/C++ application.
18291     * gcj can now use builtin functions for certain known methods, for
18292       instance Math.cos.
18293     * gcj can now automatically remove redundant array-store checks in
18294       some common cases.
18295     * The --no-store-checks optimization option was added. This can be
18296       used to omit runtime store checks for code which is known not to
18297       throw ArrayStoreException
18298     * The following third party interface standards were added to libgcj:
18299       org.w3c.dom and org.xml.sax.
18300     * java.security has been merged with GNU Classpath. The new package
18301       is now JDK 1.2 compliant, and much more complete.
18302     * A bytecode verifier was added to the libgcj interpreter.
18303     * java.lang.Character was rewritten to comply with the Unicode 3.0
18304       standard, and improve performance.
18305     * Partial support for many more locales was added to libgcj.
18306     * Socket timeouts have been implemented.
18307     * libgcj has been merged into a single shared library. There are no
18308       longer separate shared libraries for the garbage collector and
18309       zlib.
18310     * Several performance improvements were made to gcj and libgcj:
18311          + Hash synchronization (thin locks)
18312          + A special allocation path for finalizer-free objects
18313          + Thread-local allocation
18314          + Parallel GC, and other GC tweaks
18315
18316  Fortran
18317
18318   Fortran improvements are listed in [6]the Fortran documentation.
18319
18320  Ada
18321
18322   [7]AdaCore, has contributed its GNAT Ada 95 front end and associated
18323   tools. The GNAT compiler fully implements the Ada language as defined
18324   by the ISO/IEC 8652 standard.
18325
18326   Please note that the integration of the Ada front end is still work in
18327   progress.
18328
18329New Targets and Target Specific Improvements
18330
18331     * Hans-Peter Nilsson has contributed a port to MMIX, the CPU
18332       architecture used in new editions of Donald E. Knuth's The Art of
18333       Computer Programming.
18334     * Axis Communications has contributed its port to the CRIS CPU
18335       architecture, used in the ETRAX system-on-a-chip series.
18336     * Alexandre Oliva, of Red Hat, has contributed a port to the SuperH
18337       SH5 64-bit RISC microprocessor architecture, extending the existing
18338       SH port.
18339     * UltraSPARC is fully supported in 64-bit mode. The option -m64
18340       enables it.
18341     * For compatibility with the Sun compiler #pragma redefine_extname
18342       has been implemented on Solaris.
18343     * The x86 back end has had some noticeable work done to it.
18344          + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas
18345            Jaeger have contributed a port to the AMD x86-64 architecture.
18346            For more information on x86-64 see http://www.x86-64.org.
18347          + The compiler now supports MMX, 3DNow!, SSE, and SSE2
18348            instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will
18349            enable the respective instruction sets. Intel C++ compatible
18350            MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics
18351            will be added in next major release.
18352          + Following those improvements, targets for Pentium MMX, K6-2,
18353            K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were
18354            added. Refer to the documentation on -march= and -mcpu=
18355            options for details.
18356          + For those targets that support it, -mfpmath=sse will cause the
18357            compiler to generate SSE/SSE2 instructions for floating point
18358            math instead of x87 instructions. Usually, this will lead to
18359            quicker code — especially on the Pentium 4. Note that only
18360            scalar floating point instructions are used and GCC does not
18361            exploit SIMD features yet.
18362          + Prefetch support has been added to the Pentium III, Pentium 4,
18363            K6-2, K6-3, and Athlon series.
18364          + Code generated for floating point to integer conversions has
18365            been improved leading to better performance of many 3D
18366            applications.
18367     * The PowerPC back end has added 64-bit PowerPC GNU/Linux support.
18368     * C++ support for AIX has been improved.
18369     * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the
18370       PowerPC port supporting the AltiVec programming model (SIMD). The
18371       support, though presently useful, is experimental and is expected
18372       to stabilize for 3.2. The support is written to conform to
18373       Motorola's AltiVec specs. See -maltivec.
18374
18375Obsolete Systems
18376
18377   Support for a number of older systems has been declared obsolete in GCC
18378   3.1. Unless there is activity to revive them, the next release of GCC
18379   will have their sources permanently removed.
18380
18381   All configurations of the following processor architectures have been
18382   declared obsolete:
18383     * MIL-STD-1750A, 1750a-*-*
18384     * AMD A29k, a29k-*-*
18385     * Convex, c*-convex-*
18386     * Clipper, clipper-*-*
18387     * Elxsi, elxsi-*-*
18388     * Intel i860, i860-*-*
18389     * Sun picoJava, pj-*-* and pjl-*-*
18390     * Western Electric 32000, we32k-*-*
18391
18392   Most configurations of the following processor architectures have been
18393   declared obsolete, but we are preserving a few systems which may have
18394   active developers. It is unlikely that the remaining systems will
18395   survive much longer unless we see definite signs of port activity.
18396     * Motorola 88000 except
18397          + Generic a.out, m88k-*-aout*
18398          + Generic SVR4, m88k-*-sysv4
18399          + OpenBSD, m88k-*-openbsd*
18400     * NS32k except
18401          + NetBSD, ns32k-*-netbsd*
18402          + OpenBSD, ns32k-*-openbsd*.
18403     * ROMP except
18404          + OpenBSD, romp-*-openbsd*.
18405
18406   Finally, only some configurations of these processor architectures are
18407   being obsoleted.
18408     * Alpha:
18409          + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka
18410            alpha*-*-osf[45], are still supported.)
18411     * ARM:
18412          + RISCiX, arm-*-riscix*.
18413     * i386:
18414          + 386BSD, i?86-*-bsd*
18415          + Chorus, i?86-*-chorusos*
18416          + DG/UX, i?86-*-dgux*
18417          + FreeBSD 1.x, i?86-*-freebsd1.*
18418          + IBM AIX, i?86-*-aix*
18419          + ISC UNIX, i?86-*-isc*
18420          + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld*
18421          + NEXTstep, i?86-next-*
18422          + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose*
18423          + RTEMS/coff, i?86-*-rtemscoff*
18424          + RTEMS/go32, i?86-go32-rtems*
18425          + Sequent/BSD, i?86-sequent-bsd*
18426          + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and
18427            i?86-sequent-sysv3*
18428          + SunOS, i?86-*-sunos*
18429     * Motorola 68000:
18430          + Altos, m68[k0]*-altos-*
18431          + Apollo, m68[k0]*-apollo-*
18432          + Apple A/UX, m68[k0]*-apple-*
18433          + Bull, m68[k0]*-bull-*
18434          + Convergent, m68[k0]*-convergent-*
18435          + Generic SVR3, m68[k0]*-*-sysv3*
18436          + ISI, m68[k0]*-isi-*
18437          + LynxOS, m68[k0]*-*-lynxos*
18438          + NEXT, m68[k0]*-next-*
18439          + RTEMS/coff, m68[k0]*-*-rtemscoff*
18440          + Sony, m68[k0]*-sony-*
18441     * MIPS:
18442          + DEC Ultrix, mips-*-ultrix* and mips-dec-*
18443          + Generic BSD, mips-*-bsd*
18444          + Generic System V, mips-*-sysv*
18445          + IRIX before version 5, mips-sgi-irix[1234]*
18446          + RiscOS, mips-*-riscos*
18447          + Sony, mips-sony-*
18448          + Tandem, mips-tandem-*
18449     * SPARC:
18450          + RTEMS/a.out, sparc-*-rtemsaout*.
18451
18452Documentation improvements
18453
18454     * The old manual ("Using and Porting the GNU Compiler Collection")
18455       has been replaced by a users manual ("Using the GNU Compiler
18456       Collection") and a separate internals reference manual ("GNU
18457       Compiler Collection Internals").
18458     * More complete and much improved documentation about GCC's internal
18459       representation used by the C and C++ front ends.
18460     * Many cleanups and improvements in general.
18461
18462
18463    For questions related to the use of GCC, please consult these web
18464    pages and the [8]GCC manuals. If that fails, the
18465    [9]gcc-help@gcc.gnu.org mailing list might help. Comments on these
18466    web pages and the development of GCC are welcome on our developer
18467    list at [10]gcc@gcc.gnu.org. All of [11]our lists have public
18468    archives.
18469
18470   Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and
18471   distribution of this entire article is permitted in any medium,
18472   provided this notice is preserved.
18473
18474   These pages are [13]maintained by the GCC team. Last modified
18475   2019-11-28[14].
18476
18477References
18478
18479   1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html
18480   2. http://gcc.gnu.org/news/profiledriven.html
18481   3. http://gcc.gnu.org/benchmarks/
18482   4. http://gcc.gnu.org/c99status.html
18483   5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html
18484   6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html
18485   7. https://www.adacore.com/
18486   8. https://gcc.gnu.org/onlinedocs/
18487   9. mailto:gcc-help@gcc.gnu.org
18488  10. mailto:gcc@gcc.gnu.org
18489  11. https://gcc.gnu.org/lists.html
18490  12. https://www.fsf.org/
18491  13. https://gcc.gnu.org/about.html
18492  14. http://validator.w3.org/check/referer
18493======================================================================
18494http://gcc.gnu.org/gcc-3.0/index.html
18495                                   GCC 3.0.4
18496
18497   (This release series is no longer supported.)
18498
18499   February 20, 2002
18500
18501   The [1]GNU project and the GCC developers are pleased to announce the
18502   release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0
18503   series.
18504
18505   GCC used to stand for the GNU C Compiler, but since the compiler
18506   supports several other languages aside from C, it now stands for the
18507   GNU Compiler Collection.
18508
18509   GCC 3.0.x has several new optimizations, new targets, new languages and
18510   many other new features, relative to GCC 2.95.x. See the [2]new
18511   features page for a more complete list.
18512
18513   A list of [3]successful builds is updated as new information becomes
18514   available.
18515
18516   The GCC developers would like to thank the numerous people that have
18517   contributed new features, test results, bug fixes, etc to GCC. This
18518   [4]amazing group of volunteers is what makes GCC successful.
18519
18520   And finally, we can't in good conscience fail to mention some
18521   [5]caveats to using GCC 3.0.x.
18522
18523   For additional information about GCC please refer to the [6]GCC project
18524   web site or contact the [7]GCC development mailing list.
18525
18526   To obtain GCC please use [8]our mirror sites, or our CVS server.
18527     __________________________________________________________________
18528
18529Previous 3.0.x Releases
18530
18531   December 20, 2001: GCC 3.0.3 has been released.
18532   October 25, 2001: GCC 3.0.2 has been released.
18533   August 20, 2001: GCC 3.0.1 has been released.
18534   June 18, 2001: GCC 3.0 has been released.
18535
18536
18537    For questions related to the use of GCC, please consult these web
18538    pages and the [9]GCC manuals. If that fails, the
18539    [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
18540    web pages and the development of GCC are welcome on our developer
18541    list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
18542    archives.
18543
18544   Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
18545   distribution of this entire article is permitted in any medium,
18546   provided this notice is preserved.
18547
18548   These pages are [14]maintained by the GCC team. Last modified
18549   2019-11-28[15].
18550
18551References
18552
18553   1. http://www.gnu.org/
18554   2. http://gcc.gnu.org/gcc-3.0/features.html
18555   3. http://gcc.gnu.org/gcc-3.0/buildstat.html
18556   4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
18557   5. http://gcc.gnu.org/gcc-3.0/caveats.html
18558   6. http://gcc.gnu.org/index.html
18559   7. mailto:gcc@gcc.gnu.org
18560   8. http://gcc.gnu.org/mirrors.html
18561   9. https://gcc.gnu.org/onlinedocs/
18562  10. mailto:gcc-help@gcc.gnu.org
18563  11. mailto:gcc@gcc.gnu.org
18564  12. https://gcc.gnu.org/lists.html
18565  13. https://www.fsf.org/
18566  14. https://gcc.gnu.org/about.html
18567  15. http://validator.w3.org/check/referer
18568======================================================================
18569http://gcc.gnu.org/gcc-3.0/features.html
18570                              GCC 3.0 New Features
18571
18572Additional changes in GCC 3.0.4
18573
18574     * GCC 3.0 now supports newer versions of the [1]NetBSD operating
18575       system, which use the ELF object file format, on x86 processors.
18576     * Correct debugging information is generated from functions that have
18577       lines from multiple files (e.g. yacc output).
18578     * A fix for whitespace handling in the -traditional preprocessor,
18579       which can affect Fortran.
18580     * Fixes to the exception handling runtime.
18581     * More fixes for bad code generation in C++.
18582     * A fix for shared library generation under AIX 4.3.
18583     * Documentation updates.
18584     * Port of GCC to Tensilica's Xtensa processor contributed.
18585     * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link).
18586
18587Additional changes in GCC 3.0.3
18588
18589     * A fix to correct an accidental change to the PowerPC ABI.
18590     * Fixes for bad code generation on a variety of architectures.
18591     * Improvements to the debugging information generated for C++
18592       classes.
18593     * Fixes for bad code generation in C++.
18594     * A fix to avoid crashes in the C++ demangler.
18595     * A fix to the C++ standard library to avoid buffer overflows.
18596     * Miscellaneous improvements for a variety of architectures.
18597
18598Additional changes in GCC 3.0.2
18599
18600     * Fixes for bad code generation during loop unrolling.
18601     * Fixes for bad code generation by the sibling call optimization.
18602     * Minor improvements to x86 code generation.
18603     * Implementation of function descriptors in C++ vtables for IA64.
18604     * Numerous minor bug-fixes.
18605
18606Additional changes in GCC 3.0.1
18607
18608     * C++ fixes for incorrect code-generation.
18609     * Improved cross-compiling support for the C++ standard library.
18610     * Fixes for some embedded targets that worked in GCC 2.95.3, but not
18611       in GCC 3.0.
18612     * Fixes for various exception-handling bugs.
18613     * A port to the S/390 architecture.
18614
18615General Optimizer Improvements
18616
18617     * [2]Basic block reordering pass.
18618     * New if-conversion pass with support for conditional (predicated)
18619       execution.
18620     * New tail call and sibling call elimination optimizations.
18621     * New register renaming pass.
18622     * New (experimental) [3]static single assignment (SSA) representation
18623       support.
18624     * New dead-code elimination pass implemented using the SSA
18625       representation.
18626     * [4]Global null pointer test elimination.
18627     * [5]Global code hoisting/unification.
18628     * More builtins and optimizations for stdio.h, string.h and old BSD
18629       functions, as well as for ISO C99 functions.
18630     * New builtin __builtin_expect for giving hints to the branch
18631       predictor.
18632
18633New Languages and Language specific improvements
18634
18635     * The GNU Compiler for the Java(TM) language (GCJ) is now integrated
18636       and supported, including the run-time library containing most
18637       common non-GUI Java classes, a bytecode interpreter, and the Boehm
18638       conservative garbage collector. Many bugs have been fixed. GCJ can
18639       compile Java source or Java bytecodes to either native code or Java
18640       class files, and supports native methods written in either the
18641       standard JNI or the more efficient and convenient CNI.
18642     * Here is a [6]partial list of C++ improvements, both new features
18643       and those no longer supported.
18644     * New C++ ABI. On the IA-64 platform GCC is capable of
18645       inter-operating with other IA-64 compilers.
18646     * The new ABI also significantly reduces the size of symbol and debug
18647       information.
18648     * New C++ support library and many C++ bug fixes, vastly improving
18649       our conformance to the ISO C++ standard.
18650     * New [7]inliner for C++.
18651     * Rewritten C preprocessor, integrated into the C, C++ and Objective
18652       C compilers, with very many improvements including ISO C99 support
18653       and [8]improvements to dependency generation.
18654     * Support for more [9]ISO C99 features.
18655     * Many improvements to support for checking calls to format functions
18656       such as printf and scanf, including support for ISO C99 format
18657       features, extensions from the Single Unix Specification and GNU
18658       libc 2.2, checking of strfmon formats and features to assist in
18659       auditing for format string security bugs.
18660     * New warnings for C code that may have undefined semantics because
18661       of violations of sequence point rules in the C standard (such as a
18662       = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall.
18663     * Additional warning option -Wfloat-equal.
18664     * Improvements to -Wtraditional.
18665     * Fortran improvements are listed in [10]the Fortran documentation.
18666
18667New Targets and Target Specific Improvements
18668
18669     * New x86 back end, generating much improved code.
18670     * Support for a generic i386-elf target contributed.
18671     * New option to emit x86 assembly code using Intel style syntax
18672       (-mintel-syntax).
18673     * HPUX 11 support contributed.
18674     * Improved PowerPC code generation, including scheduled prologue and
18675       epilogue.
18676     * Port of GCC to Intel's IA-64 processor contributed.
18677     * Port of GCC to Motorola's MCore 210 and 340 contributed.
18678     * New unified back-end for Arm, Thumb and StrongArm contributed.
18679     * Port of GCC to Intel's XScale processor contributed.
18680     * Port of GCC to Atmel's AVR microcontrollers contributed.
18681     * Port of GCC to Mitsubishi's D30V processor contributed.
18682     * Port of GCC to Matsushita's AM33 processor (a member of the MN10300
18683       processor family) contributed.
18684     * Port of GCC to Fujitsu's FR30 processor contributed.
18685     * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed.
18686     * Port of GCC to Sun's picoJava processor core contributed.
18687
18688Documentation improvements
18689
18690     * Substantially rewritten and improved C preprocessor manual.
18691     * Many improvements to other documentation.
18692     * Manpages for gcc, cpp and gcov are now generated automatically from
18693       the master Texinfo manual, eliminating the problem of manpages
18694       being out of date. (The generated manpages are only extracts from
18695       the full manual, which is provided in Texinfo form, from which
18696       info, HTML, other formats and a printed manual can be generated.)
18697     * Generated info files are included in the release tarballs alongside
18698       their Texinfo sources, avoiding problems on some platforms with
18699       building makeinfo as part of the GCC distribution.
18700
18701Other significant improvements
18702
18703     * Garbage collection used internally by the compiler for most memory
18704       allocation instead of obstacks.
18705     * Lengauer and Tarjan algorithm used for computing dominators in the
18706       CFG. This algorithm can be significantly faster and more space
18707       efficient than our older algorithm.
18708     * gccbug script provided to assist in submitting bug reports to our
18709       bug tracking system. (Bug reports previously submitted directly to
18710       our mailing lists, for which you received no bug tracking number,
18711       should be submitted again using gccbug if you can reproduce the
18712       problem with GCC 3.0.)
18713     * The internal libgcc library is [11]built as a shared library on
18714       systems that support it.
18715     * Extensive testsuite included with GCC, with many new tests. In
18716       addition to tests for GCC bugs that have been fixed, many tests
18717       have been added for language features, compiler warnings and
18718       builtin functions.
18719     * Additional language-independent warning options -Wpacked, -Wpadded,
18720       -Wunreachable-code and -Wdisabled-optimization.
18721     * Target-independent options -falign-functions, -falign-loops and
18722       -falign-jumps.
18723
18724   Plus a great many bug fixes and almost all the [12]features found in
18725   GCC 2.95.
18726
18727
18728    For questions related to the use of GCC, please consult these web
18729    pages and the [13]GCC manuals. If that fails, the
18730    [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
18731    web pages and the development of GCC are welcome on our developer
18732    list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
18733    archives.
18734
18735   Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
18736   distribution of this entire article is permitted in any medium,
18737   provided this notice is preserved.
18738
18739   These pages are [18]maintained by the GCC team. Last modified
18740   2019-11-28[19].
18741
18742References
18743
18744   1. http://www.netbsd.org/
18745   2. http://gcc.gnu.org/news/reorder.html
18746   3. http://gcc.gnu.org/news/ssa.html
18747   4. http://gcc.gnu.org/news/null.html
18748   5. http://gcc.gnu.org/news/unify.html
18749   6. http://gcc.gnu.org/gcc-3.0/c++features.html
18750   7. http://gcc.gnu.org/news/inlining.html
18751   8. http://gcc.gnu.org/news/dependencies.html
18752   9. http://gcc.gnu.org/c99status.html
18753  10. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
18754  11. http://gcc.gnu.org/gcc-3.0/libgcc.html
18755  12. http://gcc.gnu.org/gcc-2.95/features.html
18756  13. https://gcc.gnu.org/onlinedocs/
18757  14. mailto:gcc-help@gcc.gnu.org
18758  15. mailto:gcc@gcc.gnu.org
18759  16. https://gcc.gnu.org/lists.html
18760  17. https://www.fsf.org/
18761  18. https://gcc.gnu.org/about.html
18762  19. http://validator.w3.org/check/referer
18763======================================================================
18764http://gcc.gnu.org/gcc-3.0/caveats.html
18765                                GCC 3.0 Caveats
18766
18767     * -fstrict-aliasing is now part of -O2 and higher optimization
18768       levels. This allows the compiler to assume the strictest aliasing
18769       rules applicable to the language being compiled. For C and C++,
18770       this activates optimizations based on the type of expressions. This
18771       optimization may thus break old, non-compliant code.
18772     * Enumerations are now properly promoted to int in function
18773       parameters and function returns. Normally this change is not
18774       visible, but when using -fshort-enums this is an ABI change.
18775     * The undocumented extension that allowed C programs to have a label
18776       at the end of a compound statement has been deprecated and may be
18777       removed in a future version. Programs that now generate a warning
18778       about this may be fixed by adding a null statement (a single
18779       semicolon) after the label.
18780     * The poorly documented extension that allowed string constants in C,
18781       C++ and Objective C to contain unescaped newlines has been
18782       deprecated and may be removed in a future version. Programs using
18783       this extension may be fixed in several ways: the bare newline may
18784       be replaced by \n, or preceded by \n\, or string concatenation may
18785       be used with the bare newline preceded by \n" and " placed at the
18786       start of the next line.
18787     * The Chill compiler is not included in GCC 3.0, because of the lack
18788       of a volunteer to convert it to use garbage collection.
18789     * Certain non-standard iostream methods from earlier versions of
18790       libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach,
18791       ostream::form, and istream::gets.
18792     * The new C++ ABI is not yet fully supported by current (as of
18793       2001-07-01) releases and development versions of GDB, or any
18794       earlier versions. There is a problem setting breakpoints by line
18795       number, and other related issues that have been fixed in GCC 3.0
18796       but not yet handled in GDB:
18797       [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
18798
18799
18800    For questions related to the use of GCC, please consult these web
18801    pages and the [2]GCC manuals. If that fails, the
18802    [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
18803    web pages and the development of GCC are welcome on our developer
18804    list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
18805
18806   Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
18807   distribution of this entire article is permitted in any medium,
18808   provided this notice is preserved.
18809
18810   These pages are [7]maintained by the GCC team. Last modified
18811   2019-11-28[8].
18812
18813References
18814
18815   1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
18816   2. https://gcc.gnu.org/onlinedocs/
18817   3. mailto:gcc-help@gcc.gnu.org
18818   4. mailto:gcc@gcc.gnu.org
18819   5. https://gcc.gnu.org/lists.html
18820   6. https://www.fsf.org/
18821   7. https://gcc.gnu.org/about.html
18822   8. http://validator.w3.org/check/referer
18823======================================================================
18824http://gcc.gnu.org/gcc-2.95/index.html
18825                                    GCC 2.95
18826
18827   (This release series is no longer supported.)
18828
18829   March 16, 2001: The GNU project and the GCC developers are pleased to
18830   announce the release of GCC version 2.95.3.
18831
18832Release History
18833
18834   GCC 2.95.3
18835          March 16, 2001
18836
18837   GCC 2.95.2
18838          October 27, 1999
18839
18840   GCC 2.95.1
18841          August 19, 1999
18842
18843   GCC 2.95
18844          July 31, 1999. This is the first release of GCC since the April
18845          1999 GCC/EGCS reunification and includes nearly a year's worth
18846          of new development and bugfixes.
18847
18848References and Acknowledgements
18849
18850   GCC used to stand for the GNU C Compiler, but since the compiler
18851   supports several other languages aside from C, it now stands for the
18852   GNU Compiler Collection.
18853
18854   The whole suite has been extensively [1]regression tested and
18855   [2]package tested. It should be reliable and suitable for widespread
18856   use.
18857
18858   The compiler has several new optimizations, new targets, new languages
18859   and other new features. See the [3]new features page for a more
18860   complete list of new features found in the GCC 2.95 releases.
18861
18862   The sources include installation instructions in both HTML and
18863   plaintext forms in the install directory in the distribution. However,
18864   the most up to date installation instructions and [4]build/test status
18865   are on the web pages. We will update those pages as new information
18866   becomes available.
18867
18868   The GCC developers would like to thank the numerous people that have
18869   contributed new features, test results, bugfixes, etc to GCC. This
18870   [5]amazing group of volunteers is what makes GCC successful.
18871
18872   And finally, we can't in good conscience fail to mention some
18873   [6]caveats to using GCC 2.95.
18874
18875   Download GCC 2.95 from one of our many [7]mirror sites.
18876
18877   For additional information about GCC please see the [8]GCC project web
18878   server or contact the [9]GCC development mailing list.
18879
18880
18881    For questions related to the use of GCC, please consult these web
18882    pages and the [10]GCC manuals. If that fails, the
18883    [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these
18884    web pages and the development of GCC are welcome on our developer
18885    list at [12]gcc@gcc.gnu.org. All of [13]our lists have public
18886    archives.
18887
18888   Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and
18889   distribution of this entire article is permitted in any medium,
18890   provided this notice is preserved.
18891
18892   These pages are [15]maintained by the GCC team. Last modified
18893   2019-11-28[16].
18894
18895References
18896
18897   1. http://gcc.gnu.org/gcc-2.95/regress.html
18898   2. http://gcc.gnu.org/gcc-2.95/othertest.html
18899   3. http://gcc.gnu.org/gcc-2.95/features.html
18900   4. http://gcc.gnu.org/gcc-2.95/buildstat.html
18901   5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
18902   6. http://gcc.gnu.org/gcc-2.95/caveats.html
18903   7. http://gcc.gnu.org/mirrors.html
18904   8. http://gcc.gnu.org/index.html
18905   9. mailto:gcc@gcc.gnu.org
18906  10. https://gcc.gnu.org/onlinedocs/
18907  11. mailto:gcc-help@gcc.gnu.org
18908  12. mailto:gcc@gcc.gnu.org
18909  13. https://gcc.gnu.org/lists.html
18910  14. https://www.fsf.org/
18911  15. https://gcc.gnu.org/about.html
18912  16. http://validator.w3.org/check/referer
18913======================================================================
18914http://gcc.gnu.org/gcc-2.95/features.html
18915                             GCC 2.95 New Features
18916
18917     * General Optimizer Improvements:
18918          + [1]Localized register spilling to improve speed and code
18919            density especially on small register class machines.
18920          + [2]Global CSE using lazy code motion algorithms.
18921          + [3]Improved global constant/copy propagation.
18922          + [4]Improved control flow graph analysis and manipulation.
18923          + [5]Local dead store elimination.
18924          + [6]Memory Load hoisting/store sinking in loops.
18925          + [7]Type based alias analysis is enabled by default. Note this
18926            feature will expose bugs in the Linux kernel. Please refer to
18927            the FAQ (as shipped with GCC 2.95) for additional information
18928            on this issue.
18929          + Major revamp of GIV detection, combination and simplification
18930            to improve loop performance.
18931          + Major improvements to register allocation and reloading.
18932     * New Languages and Language specific improvements
18933          + [8]Many C++ improvements.
18934          + [9]Many Fortran improvements.
18935          + [10]Java front-end has been integrated. A [11]runtime library
18936            is available separately.
18937          + [12]ISO C99 support
18938          + [13]Chill front-end and runtime has been integrated.
18939          + Boehm garbage collector support in libobjc.
18940          + More support for various pragmas which appear in vendor
18941            include files
18942     * New Targets and Target Specific Improvements
18943          + [14]SPARC backend rewrite.
18944          + -mschedule=8000 will optimize code for PA8000 class
18945            processors; -mpa-risc-2-0 will generate code for PA2.0
18946            processors
18947          + Various micro-optimizations for the ia32 port. K6
18948            optimizations
18949          + Compiler will attempt to align doubles in the stack on the
18950            ia32 port
18951          + Alpha EV6 support
18952          + PowerPC 750
18953          + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403.
18954            -mcpu=e603e was added to do -mcpu=603e and -msoft-float.
18955          + c3x, c4x
18956          + HyperSPARC
18957          + SparcLite86x
18958          + sh4
18959          + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix,
18960            arm-linux)
18961          + vxWorks targets include support for vxWorks threads
18962          + StrongARM 110 and ARM9 support added. ARM Scheduling
18963            parameters rewritten.
18964          + Various changes to the MIPS port to avoid assembler macros,
18965            which in turn improves performance
18966          + Various performance improvements to the i960 port.
18967          + Major rewrite of ns32k port
18968     * Other significant improvements
18969          + [15]Ability to dump cfg information and display it using vcg.
18970          + The new faster scheme for fixing vendor header files is
18971            enabled by default.
18972          + Experimental internationalization support.
18973          + multibyte character support
18974          + Some compile-time speedups for pathological problems
18975          + Better support for complex types
18976     * Plus the usual mountain of bugfixes
18977     * Core compiler is based on the gcc2 development tree from Sept 30,
18978       1998, so we have all of the [16]features found in GCC 2.8.
18979
18980Additional Changes in GCC 2.95.1
18981
18982     * Generic bugfixes and improvements
18983          + Various documentation fixes related to the GCC/EGCS merger.
18984          + Fix memory management bug which could lead to spurious aborts,
18985            core dumps or random parsing errors in the compiler.
18986          + Fix a couple bugs in the dwarf1 and dwarf2 debug record
18987            support.
18988          + Fix infinite loop in the CSE optimizer.
18989          + Avoid undefined behavior in compiler FP emulation code
18990          + Fix install problem when prefix is overridden on the make
18991            install command.
18992          + Fix problem with unwanted installation of assert.h on some
18993            systems.
18994          + Fix problem with finding the wrong assembler in a single tree
18995            build.
18996          + Avoid increasing the known alignment of a register that is
18997            already known to be a pointer.
18998     * Platform specific bugfixes and improvements
18999          + Codegen bugfix for prologue/epilogue for cpu32 target.
19000          + Fix long long code generation bug for the Coldfire target.
19001          + Fix various aborts in the SH compiler.
19002          + Fix bugs in libgcc support library for the SH.
19003          + Fix alpha ev6 code generation bug.
19004          + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on
19005            AIX platforms.
19006          + Fix -fpic code generation bug for rs6000/ppc svr4 targets.
19007          + Fix varargs/stdarg code generation bug for rs6000/ppc svr4
19008            targets.
19009          + Fix weak symbol handling for rs6000/ppc svr4 targets.
19010          + Fix various problems with 64bit code generation for the
19011            rs6000/ppc port.
19012          + Fix codegen bug which caused tetex to be mis-compiled on the
19013            x86.
19014          + Fix compiler abort in new cfg code exposed by x86 port.
19015          + Fix out of range array reference in code convert flat
19016            registers to the x87 stacked FP register file.
19017          + Fix minor vxworks configuration bug.
19018          + Fix return type of bsearch for SunOS 4.x.
19019     * Language & Runtime specific fixes.
19020          + The G++ signature extension has been deprecated. It will be
19021            removed in the next major release of G++. Use of signatures
19022            will result in a warning from the compiler.
19023          + Several bugs relating to templates and namespaces were fixed.
19024          + A bug that caused crashes when combining templates with -g on
19025            DWARF1 platforms was fixed.
19026          + Pointers-to-members, virtual functions, and multiple
19027            inheritance should now work together correctly.
19028          + Some code-generation bugs relating to function try blocks were
19029            fixed.
19030          + G++ is a little bit more lenient with certain archaic
19031            constructs than in GCC 2.95.
19032          + Fix to prevent shared library version #s from bring truncated
19033            to 1 digit
19034          + Fix missing std:: in the libstdc++ library.
19035          + Fix stream locking problems in libio.
19036          + Fix problem in java compiler driver.
19037
19038Additional Changes in GCC 2.95.2
19039
19040   The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While
19041   the optimizations performed by -fstrict-aliasing are valid according to
19042   the C and C++ standards, the optimization have caused some problems,
19043   particularly with old non-conforming code.
19044
19045   The GCC developers are experimenting with ways to warn users about code
19046   which violates the C/C++ standards, but those warnings are not ready
19047   for widespread use at this time. Rather than wait for those warnings
19048   the GCC developers have chosen to disable -fstrict-aliasing by default
19049   for the GCC 2.95.2 release.
19050
19051   We strongly encourage developers to find and fix code which violates
19052   the C/C++ standards as -fstrict-aliasing may be enabled by default in
19053   future releases. Use the option -fstrict-aliasing to re-enable these
19054   optimizations.
19055     * Generic bugfixes and improvements
19056          + Fix incorrectly optimized memory reference in global common
19057            subexpression elimination (GCSE) optimization pass.
19058          + Fix code generation bug in regmove.c in which it could
19059            incorrectly change a "const" value.
19060          + Fix bug in optimization of conditionals involving volatile
19061            memory references.
19062          + Avoid over-allocation of stack space for some procedures.
19063          + Fixed bug in the compiler which caused incorrect optimization
19064            of an obscure series of bit manipulations, shifts and
19065            arithmetic.
19066          + Fixed register allocator bug which caused teTeX to be
19067            mis-compiled on SPARC targets.
19068          + Avoid incorrect optimization of degenerate case statements for
19069            certain targets such as the ARM.
19070          + Fix out of range memory reference in the jump optimizer.
19071          + Avoid dereferencing null pointer in fix-header.
19072          + Fix test for GCC specific features so that it is possible to
19073            bootstrap with gcc-2.6.2 and older versions of GCC.
19074          + Fix typo in scheduler which could potentially cause out of
19075            range memory accesses.
19076          + Avoid incorrect loop reversal which caused incorrect code for
19077            certain loops on PowerPC targets.
19078          + Avoid incorrect optimization of switch statements on certain
19079            targets (for example the ARM).
19080     * Platform specific bugfixes and improvements
19081          + Work around bug in Sun V5.0 compilers which caused bootstrap
19082            comparison failures on SPARC targets.
19083          + Fix SPARC backend bug which caused aborts in final.c.
19084          + Fix sparc-hal-solaris2* configuration fragments.
19085          + Fix bug in sparc block profiling.
19086          + Fix obscure code generation bug for the PARISC targets.
19087          + Define __STDC_EXT__ for HPUX configurations.
19088          + Various POWERPC64 code generation bugfixes.
19089          + Fix abort for PPC targets using ELF (ex GNU/Linux).
19090          + Fix collect2 problems for AIX targets.
19091          + Correct handling of .file directive for PPC targets.
19092          + Fix bug in fix_trunc x86 patterns.
19093          + Fix x86 port to correctly pop the FP stack for functions that
19094            return structures in memory.
19095          + Fix minor bug in strlen x86 pattern.
19096          + Use stabs debugging instead of dwarf1 for x86-solaris targets.
19097          + Fix template repository code to handle leading underscore in
19098            mangled names.
19099          + Fix weak/weak alias support for OpenBSD.
19100          + GNU/Linux for the ARM has C++ compatible include files.
19101     * Language & Runtime specific fixes.
19102          + Fix handling of constructor attribute in the C front-end which
19103            caused problems building the Chill runtime library on some
19104            targets.
19105          + Fix minor problem merging type qualifiers in the C front-end.
19106          + Fix aliasing bug for pointers and references (C/C++).
19107          + Fix incorrect "non-constant initializer bug" when -traditional
19108            or -fwritable-strings is enabled.
19109          + Fix build error for Chill front-end on SunOS.
19110          + Do not complain about duplicate instantiations when using
19111            -frepo (C++).
19112          + Fix array bounds handling in C++ front-end which caused
19113            problems with dwarf debugging information in some
19114            circumstances.
19115          + Fix minor namespace problem.
19116          + Fix problem linking java programs.
19117
19118Additional Changes in GCC 2.95.3
19119
19120     * Generic bugfixes and improvements
19121          + Fix numerous problems that caused incorrect optimization in
19122            the register reloading code.
19123          + Fix numerous problems that caused incorrect optimization in
19124            the loop optimizer.
19125          + Fix aborts in the functions build_insn_chain and scan_loops
19126            under some circumstances.
19127          + Fix an alias analysis bug.
19128          + Fix an infinite compilation bug in the combiner.
19129          + A few problems with complex number support have been fixed.
19130          + It is no longer possible for gcc to act as a fork bomb when
19131            installed incorrectly.
19132          + The -fpack-struct option should be recognized now.
19133          + Fixed a bug that caused incorrect code to be generated due to
19134            a lost stack adjustment.
19135     * Platform specific bugfixes and improvements
19136          + Support building ARM toolchains hosted on Windows.
19137          + Fix attribute calculations in ARM toolchains.
19138          + arm-linux support has been improved.
19139          + Fix a PIC failure on sparc targets.
19140          + On ix86 targets, the regparm attribute should now work
19141            reliably.
19142          + Several updates for the h8300 port.
19143          + Fix problem building libio with glibc 2.2.
19144
19145
19146    For questions related to the use of GCC, please consult these web
19147    pages and the [17]GCC manuals. If that fails, the
19148    [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19149    web pages and the development of GCC are welcome on our developer
19150    list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
19151    archives.
19152
19153   Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
19154   distribution of this entire article is permitted in any medium,
19155   provided this notice is preserved.
19156
19157   These pages are [22]maintained by the GCC team. Last modified
19158   2019-11-28[23].
19159
19160References
19161
19162   1. http://gcc.gnu.org/news/spill.html
19163   2. http://gcc.gnu.org/news/lcm.html
19164   3. http://gcc.gnu.org/news/cprop.html
19165   4. http://gcc.gnu.org/news/cfg.html
19166   5. http://gcc.gnu.org/news/dse.html
19167   6. http://gcc.gnu.org/news/hoist.html
19168   7. http://gcc.gnu.org/news/alias.html
19169   8. http://gcc.gnu.org/gcc-2.95/c++features.html
19170   9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
19171  10. http://gcc.gnu.org/news/gcj-announce.txt
19172  11. http://gcc.gnu.org/news/javaannounce.html
19173  12. http://gcc.gnu.org/c99status.html
19174  13. http://gcc.gnu.org/news/chill.html
19175  14. http://gcc.gnu.org/news/sparc.html
19176  15. http://gcc.gnu.org/news/egcs-vcg.html
19177  16. http://gcc.gnu.org/egcs-1.0/features-2.8.html
19178  17. https://gcc.gnu.org/onlinedocs/
19179  18. mailto:gcc-help@gcc.gnu.org
19180  19. mailto:gcc@gcc.gnu.org
19181  20. https://gcc.gnu.org/lists.html
19182  21. https://www.fsf.org/
19183  22. https://gcc.gnu.org/about.html
19184  23. http://validator.w3.org/check/referer
19185======================================================================
19186http://gcc.gnu.org/gcc-2.95/caveats.html
19187                                GCC 2.95 Caveats
19188
19189     * GCC 2.95 will issue an error for invalid asm statements that had
19190       been silently accepted by earlier versions of the compiler. This is
19191       particularly noticeable when compiling older versions of the Linux
19192       kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95)
19193       for more information on this issue.
19194     * GCC 2.95 implements type based alias analysis to disambiguate
19195       memory references. Some programs, particularly the Linux kernel
19196       violate ANSI/ISO aliasing rules and therefore may not operate
19197       correctly when compiled with GCC 2.95. Please refer to the FAQ (as
19198       shipped with GCC 2.95) for more information on this issue.
19199     * GCC 2.95 has a known bug in its handling of complex variables for
19200       64bit targets. Instead of silently generating incorrect code, GCC
19201       2.95 will issue a fatal error for situations it can not handle.
19202       This primarily affects the Fortran community as Fortran makes more
19203       use of complex variables than C or C++.
19204     * GCC 2.95 has an integrated libstdc++, but does not have an
19205       integrated libg++. Furthermore old libg++ releases will not work
19206       with GCC 2.95. You can retrieve a recent copy of libg++ from the
19207       [1]GCC ftp server.
19208       Note most C++ programs only need libstdc++.
19209     * Exception handling may not work with shared libraries, particularly
19210       on alphas, hppas, rs6000/powerpc and mips based platforms.
19211       Exception handling is known to work on x86 GNU/Linux platforms with
19212       shared libraries.
19213     * In general, GCC 2.95 is more rigorous about rejecting invalid C++
19214       code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0,
19215       or EGCS 1.1. As a result it may be necessary to fix C++ code before
19216       it will compile with GCC 2.95.
19217     * G++ is also converting toward the ISO C++ standard; as a result
19218       code which was previously valid (and thus accepted by other
19219       compilers and older versions of g++) may no longer be accepted. The
19220       flag -fpermissive may allow some non-conforming code to compile
19221       with GCC 2.95.
19222     * GCC 2.95 compiled C++ code is not binary compatible with EGCS
19223       1.1.x, EGCS 1.0.x or GCC 2.8.x.
19224     * GCC 2.95 does not have changes from the GCC 2.8 tree that were made
19225       between Sept 30, 1998 and April 30, 1999 (the official end of the
19226       GCC 2.8 project). Future GCC releases will include all the changes
19227       from the defunct GCC 2.8 sources.
19228
19229
19230    For questions related to the use of GCC, please consult these web
19231    pages and the [2]GCC manuals. If that fails, the
19232    [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19233    web pages and the development of GCC are welcome on our developer
19234    list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
19235
19236   Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
19237   distribution of this entire article is permitted in any medium,
19238   provided this notice is preserved.
19239
19240   These pages are [7]maintained by the GCC team. Last modified
19241   2019-11-28[8].
19242
19243References
19244
19245   1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz
19246   2. https://gcc.gnu.org/onlinedocs/
19247   3. mailto:gcc-help@gcc.gnu.org
19248   4. mailto:gcc@gcc.gnu.org
19249   5. https://gcc.gnu.org/lists.html
19250   6. https://www.fsf.org/
19251   7. https://gcc.gnu.org/about.html
19252   8. http://validator.w3.org/check/referer
19253======================================================================
19254http://gcc.gnu.org/egcs-1.1/index.html
19255                                    EGCS 1.1
19256
19257   September 3, 1998: We are pleased to announce the release of EGCS 1.1.
19258   December 1, 1998: We are pleased to announce the release of EGCS 1.1.1.
19259   March 15, 1999: We are pleased to announce the release of EGCS 1.1.2.
19260
19261   EGCS is a free software project to further the development of the GNU
19262   compilers using an open development environment.
19263
19264   EGCS 1.1 is a major new release of the EGCS compiler system. It has
19265   been [1]extensively tested and is believed to be stable and suitable
19266   for widespread use.
19267
19268   EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8
19269   development sources; it contains all of the new features found in GCC
19270   2.8.1 as well as all new development from GCC up to June 6, 1998.
19271
19272   EGCS 1.1 also contains many improvements and features not found in GCC
19273   or in older versions of EGCS:
19274     * Global common subexpression elimination and global constant/copy
19275       propagation (aka [2]gcse)
19276     * Ongoing improvements to the [3]alias analysis support to allow for
19277       better optimizations throughout the compiler.
19278     * Vastly improved [4]C++ compiler and integrated C++ runtime
19279       libraries.
19280     * Fixes for the /tmp symlink race security problems.
19281     * New targets including mips16, arm-thumb and 64 bit PowerPC.
19282     * Improvements to GNU Fortran (g77) compiler and runtime library made
19283       since g77 version 0.5.23.
19284
19285   See the [5]new features page for a more complete list of new features
19286   found in EGCS 1.1 releases.
19287
19288   EGCS 1.1.1 is a minor update to fix several serious problems in EGCS
19289   1.1:
19290     * General improvements and fixes
19291          + Avoid some stack overflows when compiling large functions.
19292          + Avoid incorrect loop invariant code motions.
19293          + Fix some core dumps on Linux kernel code.
19294          + Bring back the imake -Di386 and friends fix from EGCS 1.0.2.
19295          + Fix code generation problem in gcse.
19296          + Various documentation related fixes.
19297     * g++/libstdc++ improvements and fixes
19298          + MT safe EH fix for setjmp/longjmp based exception handling.
19299          + Fix a few bad interactions between optimization and exception
19300            handling.
19301          + Fixes for demangling of template names starting with "__".
19302          + Fix a bug that would fail to run destructors in some cases
19303            with -O2.
19304          + Fix 'new' of classes with virtual bases.
19305          + Fix crash building Qt on the Alpha.
19306          + Fix failure compiling WIFEXITED macro on GNU/Linux.
19307          + Fix some -frepo failures.
19308     * g77 and libf2c improvements and fixes
19309          + Various documentation fixes.
19310          + Avoid compiler crash on RAND intrinsic.
19311          + Fix minor bugs in makefiles exposed by BSD make programs.
19312          + Define _XOPEN_SOURCE for libI77 build to avoid potential
19313            problems on some 64-bit systems.
19314          + Fix problem with implicit endfile on rewind.
19315          + Fix spurious recursive I/O errors.
19316     * platform specific improvements and fixes
19317          + Match all versions of UnixWare7.
19318          + Do not assume x86 SVR4 or UnixWare targets can handle stabs.
19319          + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion
19320            from unsigned ints to double precision floats.
19321          + Fix ARM ABI issue with NetBSD.
19322          + Fix a few arm code generation bugs.
19323          + Fixincludes will fix additional broken SCO OpenServer header
19324            files.
19325          + Fix a m68k backend bug which caused invalid offsets in reg+d
19326            addresses.
19327          + Fix problems with 64bit AIX 4.3 support.
19328          + Fix handling of long longs for varargs/stdarg functions on the
19329            ppc.
19330          + Minor fixes to CPP predefines for Windows.
19331          + Fix code generation problems with gpr<->fpr copies for 64bit
19332            ppc.
19333          + Fix a few coldfire code generation bugs.
19334          + Fix some more header file problems on SunOS 4.x.
19335          + Fix assert.h handling for RTEMS.
19336          + Fix Windows handling of TREE_SYMBOL_REFERENCED.
19337          + Fix x86 compiler abort in reg-stack pass.
19338          + Fix cygwin/windows problem with section attributes.
19339          + Fix Alpha code generation problem exposed by SMP Linux
19340            kernels.
19341          + Fix typo in m68k 32->64bit integer conversion.
19342          + Make sure target libraries build with -fPIC for PPC & Alpha
19343            targets.
19344
19345   EGCS 1.1.2 is a minor update to fix several serious problems in EGCS
19346   1.1.1:
19347     * General improvements and fixes
19348          + Fix bug in loop optimizer which caused the SPARC (and
19349            potentially other) ports to segfault.
19350          + Fix infinite recursion in alias analysis and combiner code.
19351          + Fix bug in regclass preferencing.
19352          + Fix incorrect loop reversal which caused incorrect code to be
19353            generated for several targets.
19354          + Fix return value for builtin memcpy.
19355          + Reduce compile time for certain loops which exposed quadratic
19356            behavior in the loop optimizer.
19357          + Fix bug which caused volatile memory to be written multiple
19358            times when only one write was needed/desired.
19359          + Fix compiler abort in caller-save.c
19360          + Fix combiner bug which caused incorrect code generation for
19361            certain division by constant operations.
19362          + Fix incorrect code generation due to a bug in range check
19363            optimizations.
19364          + Fix incorrect code generation due to mis-handling of clobbered
19365            values in CSE.
19366          + Fix compiler abort/segfault due to incorrect register
19367            splitting when unrolling loops.
19368          + Fix code generation involving autoincremented addresses with
19369            ternary operators.
19370          + Work around bug in the scheduler which caused qt to be
19371            mis-compiled on some platforms.
19372          + Fix code generation problems with -fshort-enums.
19373          + Tighten security for temporary files.
19374          + Improve compile time for codes which make heavy use of
19375            overloaded functions.
19376          + Fix multiply defined constructor/destructor symbol problems.
19377          + Avoid setting bogus RPATH environment variable during
19378            bootstrap.
19379          + Avoid GNU-make dependencies in the texinfo subdir.
19380          + Install CPP wrapper script in $(prefix)/bin if --enable-cpp.
19381            --enable-cpp=<dirname> can be used to specify an additional
19382            install directory for the cpp wrapper script.
19383          + Fix CSE bug which caused incorrect label-label refs to appear
19384            on some platforms.
19385          + Avoid linking in EH routines from libgcc if they are not
19386            needed.
19387          + Avoid obscure bug in aliasing code.
19388          + Fix bug in weak symbol handling.
19389     * Platform-specific improvements and fixes
19390          + Fix detection of PPro/PII on Unixware 7.
19391          + Fix compiler segfault when building spec99 and other programs
19392            for SPARC targets.
19393          + Fix code-generation bugs for integer and floating point
19394            conditional move instructions on the PPro/PII.
19395          + Use fixincludes to fix byteorder problems on i?86-*-sysv.
19396          + Fix build failure for the arc port.
19397          + Fix floating point format configuration for i?86-gnu port.
19398          + Fix problems with hppa1.0-hp-hpux10.20 configuration when
19399            threads are enabled.
19400          + Fix coldfire code generation bugs.
19401          + Fix "unrecognized insn" problems for Alpha and PPC ports.
19402          + Fix h8/300 code generation problem with floating point values
19403            in memory.
19404          + Fix unrecognized insn problems for the m68k port.
19405          + Fix namespace-pollution problem for the x86 port.
19406          + Fix problems with old assembler on x86 NeXT systems.
19407          + Fix PIC code-generation problems for the SPARC port.
19408          + Fix minor bug with LONG_CALLS in PowerPC SVR4 support.
19409          + Fix minor ISO namespace violation in Alpha varargs/stdarg
19410            support.
19411          + Fix incorrect "braf" instruction usage for the SH port.
19412          + Fix minor bug in va-sh which prevented its use with -ansi.
19413          + Fix problems recognizing and supporting FreeBSD.
19414          + Handle OpenBSD systems correctly.
19415          + Minor fixincludes fix for Digital UNIX 4.0B.
19416          + Fix problems with ctors/dtors in SCO shared libraries.
19417          + Abort instead of generating incorrect code for PPro/PII
19418            floating point conditional moves.
19419          + Avoid multiply defined symbols on GNU/Linux systems using
19420            libc-5.4.xx.
19421          + Fix abort in alpha compiler.
19422     * Fortran-specific fixes
19423          + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year
19424            is in the documented, non-Y2K-compliant range of 0-99, instead
19425            of being returned as 100 in the year 2000.
19426          + Fix the `Date_and_Time' intrinsic (in libg2c) to return the
19427            milliseconds value properly in Values(8).
19428          + Fix the `LStat' intrinsic (in libg2c) to return device-ID
19429            information properly in SArray(7).
19430
19431   Each release includes installation instructions in both HTML and
19432   plaintext forms (see the INSTALL directory in the toplevel directory of
19433   the distribution). However, we also keep the most up to date
19434   installation instructions and [6]build/test status on our web page. We
19435   will update those pages as new information becomes available.
19436
19437   The EGCS project would like to thank the numerous people that have
19438   contributed new features, test results, bugfixes, etc. This [7]amazing
19439   group of volunteers is what makes EGCS successful.
19440
19441   And finally, we can't in good conscience fail to mention some
19442   [8]caveats to using EGCS 1.1.
19443
19444   Download EGCS from egcs.cygnus.com (USA California).
19445
19446   The EGCS 1.1 release is also available on many mirror sites.
19447   [9]Goto mirror list to find a closer site.
19448
19449
19450    For questions related to the use of GCC, please consult these web
19451    pages and the [10]GCC manuals. If that fails, the
19452    [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19453    web pages and the development of GCC are welcome on our developer
19454    list at [12]gcc@gcc.gnu.org. All of [13]our lists have public
19455    archives.
19456
19457   Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and
19458   distribution of this entire article is permitted in any medium,
19459   provided this notice is preserved.
19460
19461   These pages are [15]maintained by the GCC team. Last modified
19462   2019-11-28[16].
19463
19464References
19465
19466   1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html
19467   2. http://gcc.gnu.org/news/gcse.html
19468   3. http://gcc.gnu.org/news/alias.html
19469   4. http://gcc.gnu.org/egcs-1.1/c++features.html
19470   5. http://gcc.gnu.org/egcs-1.1/features.html
19471   6. http://gcc.gnu.org/egcs-1.1/buildstat.html
19472   7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
19473   8. http://gcc.gnu.org/egcs-1.1/caveats.html
19474   9. http://gcc.gnu.org/mirrors.html
19475  10. https://gcc.gnu.org/onlinedocs/
19476  11. mailto:gcc-help@gcc.gnu.org
19477  12. mailto:gcc@gcc.gnu.org
19478  13. https://gcc.gnu.org/lists.html
19479  14. https://www.fsf.org/
19480  15. https://gcc.gnu.org/about.html
19481  16. http://validator.w3.org/check/referer
19482======================================================================
19483http://gcc.gnu.org/egcs-1.1/features.html
19484                             EGCS 1.1 new features
19485
19486     * Integrated GNU Fortran (g77) compiler and runtime library with
19487       improvements, based on g77 version 0.5.23.
19488     * Vast improvements in the C++ compiler; so many they have [1]page of
19489       their own!
19490     * Compiler implements [2]global common subexpression elimination and
19491       global copy/constant propagation.
19492     * More major improvements in the [3]alias analysis code.
19493     * More major improvements in the exception handling code to improve
19494       performance, lower static overhead and provide the infrastructure
19495       for future improvements.
19496     * The infamous /tmp symlink race security problems have been fixed.
19497     * The regmove optimization pass has been nearly completely rewritten
19498       to improve performance of generated code.
19499     * The compiler now recomputes register usage information before local
19500       register allocation. By providing more accurate information to the
19501       priority based allocator, we get better register allocation.
19502     * The register reloading phase of the compiler optimizes spill code
19503       much better than in previous releases.
19504     * Some bad interactions between the register allocator and
19505       instruction scheduler have been fixed, resulting in much better
19506       code for certain programs. Additionally, we have tuned the
19507       scheduler in various ways to improve performance of generated code
19508       for some architectures.
19509     * The compiler's branch shortening algorithms have been significantly
19510       improved to work better on targets which align jump targets.
19511     * The compiler now supports -Os to prefer optimizing for code space
19512       over optimizing for code speed.
19513     * The compiler will now totally eliminate library calls which compute
19514       constant values. This primarily helps targets with no integer
19515       div/mul support and targets without floating point support.
19516     * The compiler now supports an extensive "--help" option.
19517     * cpplib has been greatly improved and may be suitable for limited
19518       use.
19519     * Memory footprint for the compiler has been significantly reduced
19520       for some pathological cases.
19521     * The time to build EGCS has been improved for certain targets
19522       (particularly the alpha and mips platforms).
19523     * Many infrastructure improvements throughout the compiler, plus the
19524       usual mountain of bugfixes and minor improvements.
19525     * Target dependent improvements:
19526          + SPARC port now includes V8 plus and V9 support as well as
19527            performance tuning for Ultra class machines. The SPARC port
19528            now uses the Haifa scheduler.
19529          + Alpha port has been tuned for the EV6 processor and has an
19530            optimized expansion of memcpy/bzero. The Alpha port now uses
19531            the Haifa scheduler.
19532          + RS6000/PowerPC: support for the Power64 architecture and AIX
19533            4.3. The RS6000/PowerPC port now uses the Haifa scheduler.
19534          + x86: Alignment of static store data and jump targets is per
19535            Intel recommendations now. Various improvements throughout the
19536            x86 port to improve performance on Pentium processors
19537            (including improved epilogue sequences for Pentium chips and
19538            backend improvements which should help register allocation on
19539            all x86 variants. Conditional move support has been fixed and
19540            enabled for PPro processors. The x86 port also better supports
19541            64bit operations now. Unixware 7, a System V Release 5 target,
19542            is now supported and SCO OpenServer targets can support GAS.
19543          + MIPS has improved multiply/multiply-add support and now
19544            includes mips16 ISA support.
19545          + M68k has many micro-optimizations and Coldfire fixes.
19546     * Core compiler is based on the GCC development tree from June 9,
19547       1998, so we have all of the [4]features found in GCC 2.8.
19548
19549
19550    For questions related to the use of GCC, please consult these web
19551    pages and the [5]GCC manuals. If that fails, the
19552    [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19553    web pages and the development of GCC are welcome on our developer
19554    list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
19555
19556   Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
19557   distribution of this entire article is permitted in any medium,
19558   provided this notice is preserved.
19559
19560   These pages are [10]maintained by the GCC team. Last modified
19561   2019-11-28[11].
19562
19563References
19564
19565   1. http://gcc.gnu.org/egcs-1.1/c++features.html
19566   2. http://gcc.gnu.org/news/gcse.html
19567   3. http://gcc.gnu.org/news/alias.html
19568   4. http://gcc.gnu.org/egcs-1.0/features-2.8.html
19569   5. https://gcc.gnu.org/onlinedocs/
19570   6. mailto:gcc-help@gcc.gnu.org
19571   7. mailto:gcc@gcc.gnu.org
19572   8. https://gcc.gnu.org/lists.html
19573   9. https://www.fsf.org/
19574  10. https://gcc.gnu.org/about.html
19575  11. http://validator.w3.org/check/referer
19576======================================================================
19577http://gcc.gnu.org/egcs-1.1/caveats.html
19578                                EGCS 1.1 Caveats
19579
19580     * EGCS has an integrated libstdc++, but does not have an integrated
19581       libg++. Furthermore old libg++ releases will not work with EGCS; HJ
19582       Lu has made a libg++-2.8.1.2 snapshot available which may work with
19583       EGCS.
19584       Note most C++ programs only need libstdc++.
19585     * Exception handling may not work with shared libraries, particularly
19586       on alphas, hppas, rs6000/powerpc and mips based platforms.
19587       Exception handling is known to work on x86-linux platforms with
19588       shared libraries.
19589     * Some versions of the Linux kernel have bugs which prevent them from
19590       being compiled or from running when compiled by EGCS. See the FAQ
19591       (as shipped with EGCS 1.1) for additional information.
19592     * In general, EGCS is more rigorous about rejecting invalid C++ code
19593       or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As
19594       a result it may be necessary to fix C++ code before it will compile
19595       with EGCS.
19596     * G++ is also converting toward the ISO C++ standard; as a result
19597       code which was previously valid (and thus accepted by other
19598       compilers and older versions of g++) may no longer be accepted.
19599     * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x
19600       or GCC 2.8.x due to changes necessary to support thread safe
19601       exception handling.
19602
19603
19604    For questions related to the use of GCC, please consult these web
19605    pages and the [1]GCC manuals. If that fails, the
19606    [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19607    web pages and the development of GCC are welcome on our developer
19608    list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
19609
19610   Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
19611   distribution of this entire article is permitted in any medium,
19612   provided this notice is preserved.
19613
19614   These pages are [6]maintained by the GCC team. Last modified
19615   2019-11-28[7].
19616
19617References
19618
19619   1. https://gcc.gnu.org/onlinedocs/
19620   2. mailto:gcc-help@gcc.gnu.org
19621   3. mailto:gcc@gcc.gnu.org
19622   4. https://gcc.gnu.org/lists.html
19623   5. https://www.fsf.org/
19624   6. https://gcc.gnu.org/about.html
19625   7. http://validator.w3.org/check/referer
19626======================================================================
19627http://gcc.gnu.org/egcs-1.0/index.html
19628                                    EGCS 1.0
19629
19630   December 3, 1997: We are pleased to announce the release of EGCS 1.0.
19631   January 6, 1998: We are pleased to announce the release of EGCS 1.0.1.
19632   March 16, 1998: We are pleased to announce the release of EGCS 1.0.2.
19633   May 15, 1998 We are pleased to announce the release of EGCS 1.0.3.
19634
19635   EGCS is a collaborative effort involving several groups of hackers
19636   using an open development model to accelerate development and testing
19637   of GNU compilers and runtime libraries.
19638
19639   An important goal of EGCS is to allow wide scale testing of
19640   experimental features and optimizations; therefore, EGCS contains some
19641   features and optimizations which are still under development. However,
19642   EGCS has been carefully tested and should be comparable in quality to
19643   most GCC releases.
19644
19645   EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8
19646   development sources; it contains nearly all of the new features found
19647   in GCC 2.8.
19648
19649   EGCS 1.0 also contains many improvements and features not found in GCC
19650   2.7 and even the GCC 2.8 series (which was released after the original
19651   EGCS 1.0 release).
19652     * Integrated C++ runtime libraries, including support for most major
19653       GNU/Linux systems!
19654     * The integrated libstdc++ library includes a verbatim copy of SGI's
19655       STL release.
19656     * Integrated GNU Fortran compiler.
19657     * New instruction scheduler.
19658     * New alias analysis code.
19659
19660   See the [1]new features page for a more complete list of new features.
19661
19662   EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few
19663   critical bugs and add support for Red Hat 5.0 Linux. Changes since the
19664   EGCS 1.0 release:
19665     * Add support for Red Hat 5.0 Linux and better support for Linux
19666       systems using glibc2.
19667       Many programs failed to link when compiled with EGCS 1.0 on Red Hat
19668       5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should
19669       fix these problems.
19670     * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception
19671       handling interfaces.
19672       To avoid future compatibility problems, we strongly urge anyone who
19673       is planning on distributing shared libraries that contain C++ code
19674       to upgrade to EGCS 1.0.1 first.
19675       Soon after EGCS 1.0 was released, the GCC developers made some
19676       incompatible changes in libgcc's exception handling interfaces.
19677       These changes were needed to solve problems on some platforms. This
19678       means that GCC 2.8.0, when released, will not be seamlessly
19679       compatible with shared libraries built by EGCS 1.0. The reason is
19680       that the libgcc.a in GCC 2.8.0 will not contain a function needed
19681       by the old interface.
19682       The result of this is that there may be compatibility problems with
19683       shared libraries built by EGCS 1.0 when used with GCC 2.8.0.
19684       With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface,
19685       and libgcc.a has the support routines for both the old and the new
19686       interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed,
19687       and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed).
19688       The maintainers of GCC 2.x have decided against including seamless
19689       support for the old interface in 2.8.0, since it was never
19690       "official", so to avoid future compatibility problems we recommend
19691       against distributing any shared libraries built by EGCS 1.0 that
19692       contain C++ code (upgrade to 1.0.1 and use that).
19693     * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends.
19694       The x86 changes fix code generation errors exposed when building
19695       glibc2 and the usual GNU/Linux dynamic linker (ld.so).
19696       The hppa change fixes a compiler abort when configured for use with
19697       RTEMS.
19698       The MIPS changes fix problems with the definition of LONG_MAX on
19699       newer systems, allow for command line selection of the target ABI,
19700       and fix one code generation problem.
19701       The rs6000/ppc change fixes some problems with passing structures
19702       to varargs/stdarg functions.
19703     * A few machine independent bugfixes, mostly to fix code generation
19704       errors when building Linux kernels or glibc.
19705     * Fix a few critical exception handling and template bugs in the C++
19706       compiler.
19707     * Fix Fortran namelist bug on alphas.
19708     * Fix build problems on x86-solaris systems.
19709
19710   EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several
19711   serious problems in EGCS 1.0.1.
19712     * General improvements and fixes
19713          + Memory consumption significantly reduced, especially for
19714            templates and inline functions.
19715          + Fix various problems with glibc2.1.
19716          + Fix loop optimization bug exposed by rs6000/ppc port.
19717          + Fix to avoid potential code generation problems in jump.c.
19718          + Fix some undefined symbol problems in dwarf1 debug support.
19719     * g++/libstdc++ improvements and fixes
19720          + libstdc++ in the EGCS release has been updated and should be
19721            link compatible with libstdc++-2.8.
19722          + Various fixes in libio/libstdc++ to work better on GNU/Linux
19723            systems.
19724          + Fix problems with duplicate symbols on systems that do not
19725            support weak symbols.
19726          + Memory corruption bug and undefined symbols in bastring have
19727            been fixed.
19728          + Various exception handling fixes.
19729          + Fix compiler abort for very long thunk names.
19730     * g77 improvements and fixes
19731          + Fix compiler crash for omitted bound in Fortran CASE
19732            statement.
19733          + Add missing entries to g77 lang-options.
19734          + Fix problem with -fpedantic in the g77 compiler.
19735          + Fix "backspace" problem with g77 on alphas.
19736          + Fix x86 backend problem with Fortran literals and -fpic.
19737          + Fix some of the problems with negative subscripts for g77 on
19738            alphas.
19739          + Fixes for Fortran builds on cygwin32/mingw32.
19740     * platform specific improvements and fixes
19741          + Fix long double problems on x86 (exposed by glibc).
19742          + x86 ports define i386 again to keep imake happy.
19743          + Fix exception handling support on NetBSD ports.
19744          + Several changes to collect2 to fix many problems with AIX.
19745          + Define __ELF__ for GNU/Linux on rs6000.
19746          + Fix -mcall-linux problem on GNU/Linux on rs6000.
19747          + Fix stdarg/vararg problem for GNU/Linux on rs6000.
19748          + Allow autoconf to select a proper install problem on AIX 3.1.
19749          + m68k port support includes -mcpu32 option as well as cpu32
19750            multilibs.
19751          + Fix stdarg bug for irix6.
19752          + Allow EGCS to build on irix5 without the gnu assembler.
19753          + Fix problem with static linking on sco5.
19754          + Fix bootstrap on sco5 with native compiler.
19755          + Fix for abort building newlib on H8 target.
19756          + Fix fixincludes handling of math.h on SunOS.
19757          + Minor fix for Motorola 3300 m68k systems.
19758
19759   EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few
19760   problems reported by Red Hat for builds of Red Hat 5.1.
19761     * Generic bugfixes:
19762          + Fix a typo in the libio library which resulted in incorrect
19763            behavior of istream::get.
19764          + Fix the Fortran negative array index problem.
19765          + Fix a major problem with the ObjC runtime thread support
19766            exposed by glibc2.
19767          + Reduce memory consumption of the Haifa scheduler.
19768     * Target specific bugfixes:
19769          + Fix one x86 floating point code generation bug exposed by
19770            glibc2 builds.
19771          + Fix one x86 internal compiler error exposed by glibc2 builds.
19772          + Fix profiling bugs on the Alpha.
19773          + Fix ImageMagick & emacs 20.2 build problems on the Alpha.
19774          + Fix rs6000/ppc bug when converting values from integer types
19775            to floating point types.
19776
19777   The EGCS 1.0 releases include installation instructions in both HTML
19778   and plaintext forms (see the INSTALL directory in the toplevel
19779   directory of the distribution). However, we also keep the most up to
19780   date installation instructions and [2]build/test status on our web
19781   page. We will update those pages as new information becomes available.
19782
19783   And, we can't in good conscience fail to mention some [3]caveats to
19784   using EGCS.
19785
19786   Update: Big thanks to Stanford for providing a high speed link for
19787   downloading EGCS (go.cygnus.com)!
19788
19789   Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com
19790   (USA California -- High speed link provided by Stanford).
19791
19792   The EGCS 1.0 release is also available many mirror sites.
19793   [4]Goto mirror list to find a closer site
19794
19795   We'd like to thank the numerous people that have contributed new
19796   features, test results, bugfixes, etc. Unfortunately, they're far too
19797   numerous to mention by name.
19798
19799
19800    For questions related to the use of GCC, please consult these web
19801    pages and the [5]GCC manuals. If that fails, the
19802    [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19803    web pages and the development of GCC are welcome on our developer
19804    list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
19805
19806   Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
19807   distribution of this entire article is permitted in any medium,
19808   provided this notice is preserved.
19809
19810   These pages are [10]maintained by the GCC team. Last modified
19811   2019-11-28[11].
19812
19813References
19814
19815   1. http://gcc.gnu.org/egcs-1.0/features.html
19816   2. http://gcc.gnu.org/egcs-1.0/buildstat.html
19817   3. http://gcc.gnu.org/egcs-1.0/caveats.html
19818   4. http://gcc.gnu.org/mirrors.html
19819   5. https://gcc.gnu.org/onlinedocs/
19820   6. mailto:gcc-help@gcc.gnu.org
19821   7. mailto:gcc@gcc.gnu.org
19822   8. https://gcc.gnu.org/lists.html
19823   9. https://www.fsf.org/
19824  10. https://gcc.gnu.org/about.html
19825  11. http://validator.w3.org/check/referer
19826======================================================================
19827http://gcc.gnu.org/egcs-1.0/features.html
19828                               EGCS 1.0 features
19829
19830     * Core compiler is based on the gcc2 development tree from Aug 2,
19831       1997, so we have most of the [1]features found in GCC 2.8.
19832     * Integrated GNU Fortran compiler based on g77-0.5.22-19970929.
19833     * Vast improvements in the C++ compiler; so many they have [2]page of
19834       their own!
19835     * Integrated C++ runtime libraries, including support for most major
19836       GNU/Linux systems!
19837     * New instruction scheduler from IBM Haifa which includes support for
19838       function wide instruction scheduling as well as superscalar
19839       scheduling.
19840     * Significantly improved alias analysis code.
19841     * Improved register allocation for two address machines.
19842     * Significant code generation improvements for Fortran code on
19843       Alphas.
19844     * Various optimizations from the g77 project as well as improved loop
19845       optimizations.
19846     * Dwarf2 debug format support for some targets.
19847     * egcs libstdc++ includes the SGI STL implementation without changes.
19848     * As a result of these and other changes, egcs libstc++ is not binary
19849       compatible with previous releases of libstdc++.
19850     * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO
19851       Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and
19852       1.1), Support for RTEMS on several embedded targets, Support for
19853       arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and
19854       MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc.
19855     * Integrated testsuites for gcc, g++, g77, libstdc++ and libio.
19856     * RS6000/PowerPC ports generate code which can run on all
19857       RS6000/PowerPC variants by default.
19858     * -mcpu= and -march= switches for the x86 port to allow better
19859       control over how the x86 port generates code.
19860     * Includes the template repository patch (aka repo patch); note the
19861       new template code makes repo obsolete for ELF systems using gnu-ld
19862       such as GNU/Linux.
19863     * Plus the usual assortment of bugfixes and improvements.
19864
19865
19866    For questions related to the use of GCC, please consult these web
19867    pages and the [3]GCC manuals. If that fails, the
19868    [4]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19869    web pages and the development of GCC are welcome on our developer
19870    list at [5]gcc@gcc.gnu.org. All of [6]our lists have public archives.
19871
19872   Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and
19873   distribution of this entire article is permitted in any medium,
19874   provided this notice is preserved.
19875
19876   These pages are [8]maintained by the GCC team. Last modified
19877   2019-11-28[9].
19878
19879References
19880
19881   1. http://gcc.gnu.org/egcs-1.0/features-2.8.html
19882   2. http://gcc.gnu.org/egcs-1.0/c++features.html
19883   3. https://gcc.gnu.org/onlinedocs/
19884   4. mailto:gcc-help@gcc.gnu.org
19885   5. mailto:gcc@gcc.gnu.org
19886   6. https://gcc.gnu.org/lists.html
19887   7. https://www.fsf.org/
19888   8. https://gcc.gnu.org/about.html
19889   9. http://validator.w3.org/check/referer
19890======================================================================
19891http://gcc.gnu.org/egcs-1.0/caveats.html
19892                                EGCS 1.0 Caveats
19893
19894     * EGCS has an integrated libstdc++, but does not have an integrated
19895       libg++. Furthermore old libg++ releases will not work with egc; HJ
19896       Lu has made a libg++-2.8.1.2 available which may work with EGCS.
19897       Note most C++ programs only need libstdc++.
19898     * Note that using -pedantic or -Wreturn-type can cause an explosion
19899       in the amount of memory needed for template-heavy C++ code, such as
19900       code that uses STL. Also note that -Wall includes -Wreturn-type, so
19901       if you use -Wall you will need to specify -Wno-return-type to turn
19902       it off.
19903     * Exception handling may not work with shared libraries, particularly
19904       on alphas, hppas, and mips based platforms. Exception handling is
19905       known to work on x86-linux platforms with shared libraries.
19906     * Some versions of the Linux kernel have bugs which prevent them from
19907       being compiled or from running when compiled by EGCS. See the FAQ
19908       (as shipped with EGCS 1.0) for additional information.
19909     * In general, EGCS is more rigorous about rejecting invalid C++ code
19910       or deprecated C++ constructs than G++ 2.7. As a result it may be
19911       necessary to fix C++ code before it will compile with EGCS.
19912     * G++ is also aggressively tracking the C++ standard; as a result
19913       code which was previously valid (and thus accepted by other
19914       compilers and older versions of G++) may no longer be accepted.
19915     * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS
19916       1.0.x and later releases should work with Red Hat Linux 5.0.
19917
19918
19919    For questions related to the use of GCC, please consult these web
19920    pages and the [1]GCC manuals. If that fails, the
19921    [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
19922    web pages and the development of GCC are welcome on our developer
19923    list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
19924
19925   Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
19926   distribution of this entire article is permitted in any medium,
19927   provided this notice is preserved.
19928
19929   These pages are [6]maintained by the GCC team. Last modified
19930   2019-11-28[7].
19931
19932References
19933
19934   1. https://gcc.gnu.org/onlinedocs/
19935   2. mailto:gcc-help@gcc.gnu.org
19936   3. mailto:gcc@gcc.gnu.org
19937   4. https://gcc.gnu.org/lists.html
19938   5. https://www.fsf.org/
19939   6. https://gcc.gnu.org/about.html
19940   7. http://validator.w3.org/check/referer
19941======================================================================
19942