1 This file contains information about GCC releases which has been generated
2 automatically 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
4 that led to GCC 3. For information on GCC 2.8.1 and older releases of GCC 2,
7 ======================================================================
8 http://gcc.gnu.org/gcc-4.9/index.html
13 The [1]GNU project and the GCC developers are pleased to announce the
16 This release is a bug-fix release, containing fixes for regressions in
17 GCC 4.9.2 relative to previous releases of GCC.
22 June 26, 2015 ([2]changes, [3]documentation)
25 October 30, 2014 ([4]changes, [5]documentation)
28 July 16, 2014 ([6]changes, [7]documentation)
31 April 22, 2014 ([8]changes, [9]documentation)
33 References and Acknowledgements
35 GCC used to stand for the GNU C Compiler, but since the compiler
36 supports several other languages aside from C, it now stands for the
37 GNU Compiler Collection.
39 A list of [10]successful builds is updated as new information becomes
42 The GCC developers would like to thank the numerous people that have
43 contributed new features, improvements, bug fixes, and other changes as
44 well as test results to GCC. This [11]amazing group of volunteers is
45 what makes GCC successful.
47 For additional information about GCC please refer to the [12]GCC
48 project web site or contact the [13]GCC development mailing list.
50 To obtain GCC please use [14]our mirror sites or [15]our SVN server.
53 For questions related to the use of GCC, please consult these web
54 pages and the [16]GCC manuals. If that fails, the
55 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
56 web pages and the development of GCC are welcome on our developer
57 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
60 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
61 distribution of this entire article is permitted in any medium,
62 provided this notice is preserved.
64 These pages are [21]maintained by the GCC team. Last modified
69 1. http://www.gnu.org/
70 2. http://gcc.gnu.org/gcc-4.9/changes.html
71 3. http://gcc.gnu.org/onlinedocs/4.9.3/
72 4. http://gcc.gnu.org/gcc-4.9/changes.html
73 5. http://gcc.gnu.org/onlinedocs/4.9.2/
74 6. http://gcc.gnu.org/gcc-4.9/changes.html
75 7. http://gcc.gnu.org/onlinedocs/4.9.1/
76 8. http://gcc.gnu.org/gcc-4.9/changes.html
77 9. https://gcc.gnu.org/onlinedocs/4.9.0/
78 10. http://gcc.gnu.org/gcc-4.9/buildstat.html
79 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
80 12. http://gcc.gnu.org/index.html
81 13. mailto:gcc@gcc.gnu.org
82 14. http://gcc.gnu.org/mirrors.html
83 15. http://gcc.gnu.org/svn.html
84 16. https://gcc.gnu.org/onlinedocs/
85 17. mailto:gcc-help@gcc.gnu.org
86 18. mailto:gcc@gcc.gnu.org
87 19. https://gcc.gnu.org/lists.html
88 20. http://www.fsf.org/
89 21. https://gcc.gnu.org/about.html
90 22. http://validator.w3.org/check/referer
91 ======================================================================
92 http://gcc.gnu.org/gcc-4.9/changes.html
93 GCC 4.9 Release Series
94 Changes, New Features, and Fixes
98 * The mudflap run time checker has been removed. The mudflap options
99 remain, but do nothing.
100 * Support for a number of older systems and recently unmaintained or
101 untested target ports of GCC has been declared obsolete in GCC 4.9.
102 Unless there is activity to revive them, the next release of GCC
103 will have their sources permanently removed.
104 The following ports for individual systems on particular
105 architectures have been obsoleted:
106 + Solaris 9 (*-*-solaris2.9). Details can be found in the
108 * On AArch64, the singleton vector types int64x1_t, uint64x1_t and
109 float64x1_t exported by arm_neon.h are defined to be the same as
110 their base types. This results in incorrect application of
111 parameter passing rules to arguments of types int64x1_t and
112 uint64x1_t, with respect to the AAPCS64 ABI specification. In
113 addition, names of C++ functions with parameters of these types
114 (including float64x1_t) are not mangled correctly. The current
115 typedef declarations also unintentionally allow implicit casting
116 between singleton vector types and their base types. These issues
117 will be resolved in a near future release. See [2]PR60825 for more
120 More information on porting to GCC 4.9 from previous versions of GCC
121 can be found in the [3]porting guide for this release.
123 General Optimizer Improvements
125 * AddressSanitizer, a fast memory error detector, is now available on
127 * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior
128 detector, has been added and can be enabled via
129 -fsanitize=undefined. Various computations will be instrumented to
130 detect undefined behavior at runtime. UndefinedBehaviorSanitizer is
131 currently available for the C and C++ languages.
132 * Link-time optimization (LTO) improvements:
133 + Type merging was rewritten. The new implementation is
134 significantly faster and uses less memory.
135 + Better partitioning algorithm resulting in less streaming
137 + Early removal of virtual methods reduces the size of object
138 files and improves link-time memory usage and compile time.
139 + Function bodies are now loaded on-demand and released early
140 improving overall memory usage at link time.
141 + C++ hidden keyed methods can now be optimized out.
142 + When using a linker plugin, compiling with the -flto option
143 now generates slim object files (.o) which only contain
144 intermediate language representation for LTO. Use
145 -ffat-lto-objects to create files which contain additionally
146 the object code. To generate static libraries suitable for LTO
147 processing, use gcc-ar and gcc-ranlib; to list symbols from a
148 slim object file use gcc-nm. (This requires that ar, ranlib
149 and nm have been compiled with plugin support.)
150 Memory usage building Firefox with debug enabled was reduced from
151 15GB to 3.5GB; link time from 1700 seconds to 350 seconds.
152 * Inter-procedural optimization improvements:
153 + New type inheritance analysis module improving
154 devirtualization. Devirtualization now takes into account
155 anonymous name-spaces and the C++11 final keyword.
156 + New speculative devirtualization pass (controlled by
157 -fdevirtualize-speculatively.
158 + Calls that were speculatively made direct are turned back to
159 indirect where direct call is not cheaper.
160 + Local aliases are introduced for symbols that are known to be
161 semantically equivalent across shared libraries improving
162 dynamic linking times.
163 * Feedback directed optimization improvements:
164 + Profiling of programs using C++ inline functions is now more
166 + New time profiling determines typical order in which functions
168 + A new function reordering pass (controlled by
169 -freorder-functions) significantly reduces startup time of
170 large applications. Until binutils support is completed, it is
171 effective only with link-time optimization.
172 + Feedback driven indirect call removal and devirtualization now
173 handle cross-module calls when link-time optimization is
176 New Languages and Language specific improvements
178 * Version 4.0 of the [4]OpenMP specification is now supported in the
179 C and C++ compilers and starting with the 4.9.1 release also in the
180 Fortran compiler. The new -fopenmp-simd option can be used to
181 enable OpenMP's SIMD directives, while ignoring other OpenMP
182 directives. The new [5]-fsimd-cost-model= option permits to tune
183 the vectorization cost model for loops annotated with OpenMP and
184 Cilk Plus simd directives; -Wopenmp-simd warns when the current
185 cost model overrides simd directives set by the user.
186 * The -Wdate-time option has been added for the C, C++ and Fortran
187 compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__
188 macros are used. Those macros might prevent bit-wise-identical
189 reproducible compilations.
193 * GNAT switched to Ada 2012 instead of Ada 2005 by default.
197 * Support for colorizing diagnostics emitted by GCC has been added.
198 The [6]-fdiagnostics-color=auto will enable it when outputting to
199 terminals, -fdiagnostics-color=always unconditionally. The
200 GCC_COLORS environment variable can be used to customize the colors
201 or disable coloring. If GCC_COLORS variable is present in the
202 environment, the default is -fdiagnostics-color=auto, otherwise
203 -fdiagnostics-color=never.
204 Sample diagnostics output:
205 $ g++ -fdiagnostics-color=always -S -Wall test.C
206 test.C: In function ‘int foo()’:
207 test.C:1:14: warning: no return statement in function returning non-void [-W
211 test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use
212 -ftemplate-depth= to increase the maximum) instantiating ‘struct X<100>’
213 template <int N> struct X { static const int value = X<N-1>::value; }; temp
216 test.C:2:46: recursively required from ‘const int X<999>::value’
217 test.C:2:46: required from ‘const int X<1000>::value’
218 test.C:2:88: required from here
220 test.C:2:46: error: incomplete type ‘X<100>’ used in nested name specifier
222 * With the new [7]#pragma GCC ivdep, the user can assert that there
223 are no loop-carried dependencies which would prevent concurrent
224 execution of consecutive iterations using SIMD (single instruction
225 multiple data) instructions.
226 * Support for [8]Cilk Plus has been added and can be enabled with the
227 -fcilkplus option. Cilk Plus is an extension to the C and C++
228 languages to support data and task parallelism. The present
229 implementation follows ABI version 1.2; all features but _Cilk_for
230 have been implemented.
234 * ISO C11 atomics (the _Atomic type specifier and qualifier and the
235 <stdatomic.h> header) are now supported.
236 * ISO C11 generic selections (_Generic keyword) are now supported.
237 * ISO C11 thread-local storage (_Thread_local, similar to GNU C
238 __thread) is now supported.
239 * ISO C11 support is now at a similar level of completeness to ISO
240 C99 support: substantially complete modulo bugs, extended
241 identifiers (supported except for corner cases when
242 -fextended-identifiers is used), floating-point issues (mainly but
243 not entirely relating to optional C99 features from Annexes F and
244 G) and the optional Annexes K (Bounds-checking interfaces) and L
246 * A new C extension __auto_type provides a subset of the
247 functionality of C++11 auto in GNU C.
251 * The G++ implementation of [9]C++1y return type deduction for normal
252 functions has been updated to conform to [10]N3638, the proposal
253 accepted into the working paper. Most notably, it adds
254 decltype(auto) for getting decltype semantics rather than the
255 template argument deduction semantics of plain auto:
258 auto i1 = f(); // int
259 decltype(auto) i2 = f(); // int&
261 * G++ supports [11]C++1y lambda capture initializers:
265 Actually, they have been accepted since GCC 4.5, but now the
266 compiler doesn't warn about them with -std=c++1y, and supports
267 parenthesized and brace-enclosed initializers as well.
268 * G++ supports [12]C++1y variable length arrays. G++ has supported
269 GNU/C99-style VLAs for a long time, but now additionally supports
270 initializers and lambda capture by reference. In C++1y mode G++
271 will complain about VLA uses that are not permitted by the draft
272 standard, such as forming a pointer to VLA type or applying sizeof
273 to a VLA variable. Note that it now appears that VLAs will not be
274 part of C++14, but will be part of a separate document and then
278 int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3
279 [&a]{ for (int i : a) { cout << i << endl; } }();
280 &a; // error, taking address of VLA
283 * G++ supports the [13]C++1y [[deprecated]] attribute modulo bugs in
284 the underlying [[gnu::deprecated]] attribute. Classes and functions
285 can be marked deprecated and a diagnostic message added:
289 #if __cplusplus > 201103
290 class [[deprecated("A is deprecated in C++14; Use B instead")]] A;
291 [[deprecated("bar is unsafe; use foo() instead")]]
297 A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead
298 int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo
301 * G++ supports [14]C++1y digit separators. Long numeric literals can
302 be subdivided with a single quote ' to enhance readability:
307 int m = 0'004'000'000;
308 int n = 0b0001'0000'0000'0000'0000'0000;
310 double x = 1.602'176'565e-19;
311 double y = 1.602'176'565e-1'9;
313 * G++ supports [15]C++1y generic (polymorphic) lambdas.
315 // a functional object that will increment any type
316 auto incr = [](auto x) { return x++; };
318 * As a GNU extension, G++ supports explicit template parameter syntax
319 for generic lambdas. This can be combined in the expected way with
320 the standard auto syntax.
322 // a functional object that will add two like-type objects
323 auto add = [] <typename T> (T a, T b) { return a + b; };
325 * G++ supports unconstrained generic functions as specified by §4.1.2
326 and §5.1.1 of [16]N3889: Concepts Lite Specification. Briefly, auto
327 may be used as a type-specifier in a parameter declaration of any
328 function declarator in order to introduce an implicit function
329 template parameter, akin to generic lambdas.
331 // the following two function declarations are equivalent
332 auto incr(auto x) { return x++; }
333 template <typename T>
334 auto incr(T x) { return x++; }
336 Runtime Library (libstdc++)
338 * [17]Improved support for C++11, including:
339 + support for <regex>;
340 + The associative containers in <map> and <set> and the
341 unordered associative containers in <unordered_map> and
342 <unordered_set> meet the allocator-aware container
344 * [18]Improved experimental support for the upcoming ISO C++
345 standard, C++14, including:
346 + fixing constexpr member functions without const;
347 + implementation of the std::exchange() utility function;
348 + addressing tuples by type;
349 + implemention of std::make_unique;
350 + implemention of std::shared_lock;
351 + making std::result_of SFINAE-friendly;
352 + adding operator() to std::integral_constant;
353 + adding user-defined literals for standard library types
354 std::basic_string, std::chrono::duration, and std::complex;
355 + adding two range overloads to non-modifying sequence oprations
356 std::equal and std::mismatch;
357 + adding IO manipulators for quoted strings;
358 + adding constexpr members to <utility>, <complex>, <chrono>,
360 + adding compile-time std::integer_sequence;
361 + adding cleaner transformation traits;
362 + making <functional>s operator functors easier to use and more
364 * An implementation of std::experimental::optional.
365 * An implementation of std::experimental::string_view.
366 * The non-standard function std::copy_exception has been deprecated
367 and will be removed in a future version. std::make_exception_ptr
368 should be used instead.
372 * Compatibility notice:
373 + Module files: The version of the module files (.mod) has been
374 incremented; additionally, module files are now compressed.
375 Fortran MODULEs compiled by earlier GCC versions have to be
376 recompiled, when they are USEd by files compiled with GCC 4.9.
377 GCC 4.9 is not able to read .mod files of earlier GCC
378 versions; attempting to do so gives an error message. Note:
379 The ABI of the produced assembler data itself has not changed:
380 object files and libraries are fully compatible with older
381 versions (except as stated below).
383 o The [19]argument passing ABI has changed for scalar dummy
384 arguments of type INTEGER, REAL, COMPLEX and LOGICAL,
385 which have both the VALUE and the OPTIONAL attributes.
386 o To support finalization the virtual table associated with
387 polymorphic variables has changed. Code containing CLASS
388 should be recompiled, including all files which define
389 derived types involved in the type definition used by
390 polymorphic variables. (Note: Due to the incremented
391 module version, trying to mix old code with new code will
392 usually give an error message.)
393 + GNU Fortran no longer deallocates allocatable variables or
394 allocatable components of variables declared in the main
395 program. Since Fortran 2008, the standard explicitly states
396 that variables declared in the Fortran main program
397 automatically have the SAVE attribute.
398 + When opening files, the close-on-exec flag is set if the
399 system supports such a feature. This is generally considered
400 good practice these days, but if there is a need to pass file
401 descriptors to child processes the parent process must now
402 remember to clear the close-on-exec flag by calling fcntl(),
403 e.g. via ISO_C_BINDING, before executing the child process.
404 * The deprecated command-line option -fno-whole-file has been
405 removed. (-fwhole-file is the default since GCC 4.6.)
406 -fwhole-file/-fno-whole-file continue to be accepted but do not
407 influence the code generation.
408 * The compiler no longer unconditionally warns about DO loops with
409 zero iterations. This warning is now controlled by the -Wzerotrip
410 option, which is implied by -Wall.
411 * The new NO_ARG_CHECK attribute of the [20]!GCC$ directive can be
412 used to disable the type-kind-rank (TKR) argument check for a dummy
413 argument. The feature is similar to ISO/IEC TS 29133:2012's
414 TYPE(*), except that it additionally also disables the rank check.
415 Variables with NO_ARG_CHECK have to be dummy arguments and may only
416 be used as argument to ISO_C_BINDING's C_LOC and as actual argument
417 to another NO_ARG_CHECK dummy argument; also the other constraints
418 of TYPE(*) apply. The dummy arguments should be declared as scalar
419 or assumed-size variable of type type(*) (recommended) – or of type
420 integer, real, complex or logical. With NO_ARG_CHECK, a pointer to
421 the data without further type or shape information is passed,
422 similar to C's void*. Note that also TS 29113's
423 type(*),dimension(..) accepts arguments of any type and rank;
424 contrary to NO_ARG_CHECK assumed-rank arguments pass an array
425 descriptor which contains the array shape and stride of the
428 + Finalization is now supported. It is currently only done for a
429 subset of those situations in which it should occur.
430 + Experimental support for scalar character components with
431 deferred length (i.e. allocatable string length) in derived
432 types has been added. (Deferred-length character variables are
433 supported since GCC 4.6.)
435 + When STOP or ERROR STOP are used to terminate the execution
436 and any exception (but inexact) is signaling, a warning is
437 printed to ERROR_UNIT, indicating which exceptions are
438 signaling. The [23]-ffpe-summary= command-line option can be
439 used to fine-tune for which exceptions the warning should be
441 + Rounding on input (READ) is now handled on systems where
442 strtod honours the rounding mode. (For output, rounding is
443 supported since GCC 4.5.) Note that for input, the compatible
444 rounding mode is handled as nearest (i.e., for a tie, rounding
445 to an even last significant [cf. IEC 60559:1989] – while
446 compatible rounds away from zero for a tie).
450 * GCC 4.9 provides a complete implementation of the Go 1.2.1 release.
452 New Targets and Target Specific Improvements
456 * The ARMv8-A crypto and CRC instructions are now supported through
457 intrinsics. These are enabled when the architecture supports these
458 and are available through the -march=armv8-a+crc and
459 -march=armv8-a+crypto options.
460 * Initial support for ILP32 has now been added to the compiler. This
461 is now available through the command-line option -mabi=ilp32.
462 Support for ILP32 is considered experimental as the ABI
463 specification is still beta.
464 * Coverage of more of the ISA including the SIMD extensions has been
465 added. The Advanced SIMD intrinsics have also been improved.
466 * The new local register allocator (LRA) is now on by default for the
468 * The REE (Redundant extension elimination) pass has now been enabled
469 by default for the AArch64 backend.
470 * Tuning for the Cortex-A53 and Cortex-A57 has been improved.
471 * Initial big.LITTLE tuning support for the combination of Cortex-A57
472 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53
474 * A number of structural changes have been made to both the ARM and
475 AArch64 backends to facilitate improved code-generation.
476 * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769
477 has been added and can be enabled by giving the
478 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by
479 default by configuring GCC with the --enable-fix-cortex-a53-835769
484 * A port for Synopsys Designware ARC has been contributed by Embecosm
489 * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been
490 disabled by default. This was found to generate better code in only
491 a small number of cases. It can be turned back on with the
492 -mneon-for-64bits option.
493 * Further support for the ARMv8-A architecture, notably implementing
494 the restriction around IT blocks in the Thumb32 instruction set has
495 been added. The -mrestrict-it option can be used with
496 -march=armv7-a or the -march=armv7ve options to make code
497 generation fully compatible with the deprecated instructions in
499 * Support has now been added for the ARMv7ve variant of the
500 architecture. This can be used by the -march=armv7ve option.
501 * The ARMv8-A crypto and CRC instructions are now supported through
502 intrinsics and are available through the -march=armv8-a+crc and
503 mfpu=crypto-neon-fp-armv8 options.
504 * LRA is now on by default for the ARM target. This can be turned off
505 using the -mno-lra option. This option is a purely transitionary
506 command-line option and will be removed in a future release. We are
507 interested in any bug reports regarding functional and performance
508 regressions with LRA.
509 * A new option -mslow-flash-data to improve performance of programs
510 fetching data on slow flash memory has now been introduced for the
511 ARMv7-M profile cores.
512 * A new option -mpic-data-is-text-relative for targets that allows
513 data segments to be relative to text segments has been added. This
514 is on by default for all targets except VxWorks RTP.
515 * A number of infrastructural changes have been made to both the ARM
516 and AArch64 backends to facilitate improved code-generation.
517 * GCC now supports Cortex-A12 and the Cortex-R7 through the
518 -mcpu=cortex-a12 and -mcpu=cortex-r7 options.
519 * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the
520 -mcpu=cortex-a57 and -mcpu=cortex-a53 options.
521 * Initial big.LITTLE tuning support for the combination of Cortex-A57
522 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53
523 option. Similar support was added for the combination of Cortex-A15
524 and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option.
525 * Further performance optimizations for the Cortex-A15 and the
526 Cortex-M4 have been added.
527 * A number of code generation improvements for Thumb2 to reduce code
528 size when compiling for the M-profile processors.
532 * -mfpmath=sse is now implied by -ffast-math on all targets where
534 * Intel AVX-512 support was added to GCC. That includes inline
535 assembly support, new registers and extending existing ones, new
536 intrinsics (covered by corresponding testsuite), and basic
537 autovectorization. AVX-512 instructions are available via the
538 following GCC switches: AVX-512 foundation instructions: -mavx512f,
539 AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and
540 reciprocal instructions: -mavx512er, AVX-512 conflict detection
541 instructions: -mavx512cd.
542 * It is now possible to call x86 intrinsics from select functions in
543 a file that are tagged with the corresponding target attribute
544 without having to compile the entire file with the -mxxx option.
545 This improves the usability of x86 intrinsics and is particularly
546 useful when doing [24]Function Multiversioning.
547 * GCC now supports the new Intel microarchitecture named Silvermont
548 through -march=silvermont.
549 * GCC now supports the new Intel microarchitecture named Broadwell
550 through -march=broadwell.
551 * Optimizing for other Intel microarchitectures have been renamed to
552 -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell.
553 * -march=generic has been retuned for better support of Intel core
554 and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel
555 Pentium-M, and Pentium4 based CPUs is no longer considered
556 important for generic.
557 * -mtune=intel can now be used to generate code running well on the
558 most current Intel processors, which are Haswell and Silvermont for
560 * Support to encode 32-bit assembly instructions in 16-bit format is
561 now available through the -m16 command-line option.
562 * Better inlining of memcpy and memset that is aware of value ranges
563 and produces shorter alignment prologues.
564 * -mno-accumulate-outgoing-args is now honored when unwind
565 information is output. Argument accumulation is also now turned off
566 for portions of programs optimized for size.
567 * Support for new AMD family 15h processors (Excavator core) is now
568 available through the -march=bdver4 and -mtune=bdver4 options.
572 * A new command-line option -mcpu= has been added to the MSP430
573 backend. This option is used to specify the ISA to be used.
574 Accepted values are msp430 (the default), msp430x and msp430xv2.
575 The ISA is no longer deduced from the -mmcu= option as there are
576 far too many different MCU names. The -mmcu= option is still
577 supported, and this is still used to select linker scripts and
578 generate a C preprocessor symbol that will be recognised by the
579 msp430.h header file.
583 * A new nds32 port supports the 32-bit architecture from Andes
584 Technology Corporation.
585 * The port provides initial support for the V2, V3, V3m instruction
590 * A port for the Altera Nios II has been contributed by Mentor
593 PowerPC / PowerPC64 / RS6000
595 * GCC now supports Power ISA 2.07, which includes support for
596 Hardware Transactional Memory (HTM), Quadword atomics and several
597 VMX and VSX additions, including Crypto, 64-bit integer, 128-bit
598 integer and decimal integer operations.
599 * Support for the POWER8 processor is now available through the
600 -mcpu=power8 and -mtune=power8 options.
601 * The libitm library has been modified to add a HTM fastpath that
602 automatically uses POWER's HTM hardware instructions when it is
603 executing on a HTM enabled processor.
604 * Support for the new powerpc64le-linux platform has been added. It
605 defaults to generating code that conforms to the ELFV2 ABI.
609 * Support for the Transactional Execution Facility included with the
610 IBM zEnterprise zEC12 processor has been added. A set of GCC style
611 builtins as well as XLC style builtins are provided. The builtins
612 are enabled by default when using the -march=zEC12 option but can
613 explicitly be disabled with -mno-htm. Using the GCC builtins also
614 libitm supports hardware transactions on S/390.
615 * The hotpatch features allows to prepare functions for hotpatching.
616 A certain amount of bytes is reserved before the function entry
617 label plus a NOP is inserted at its very beginning to implement a
618 backward jump when applying a patch. The feature can either be
619 enabled per compilation unit via the command-line option -mhotpatch
620 or per function using the hotpatch attribute.
621 * The shrink wrap optimization is now supported on S/390 and enabled
623 * A major rework of the routines to determine which registers need to
624 be saved and restored in function prologue/epilogue now allow to
625 use floating point registers as save slots. This will happen for
626 certain leaf function with -march=z10 or higher.
627 * The LRA rtl pass replaces reload by default on S/390.
631 * The port now allows to specify the RX100, RX200, and RX600
632 processors with the command-line options -mcpu=rx100, -mcpu=rx200
637 * Minor improvements to code generated for integer arithmetic and
638 code that involves the T bit.
639 * Added support for the SH2A clips and clipu instructions. The
640 compiler will now try to utilize them for min/max expressions such
641 as max (-128, min (127, x)).
642 * Added support for the cmp/str instruction through built-in
643 functions such as __builtin_strlen. When not optimizing for size,
644 the compiler will now expand calls to e.g. strlen as an inlined
645 sequences which utilize the cmp/str instruction.
646 * Improved code generated around volatile memory loads and stores.
647 * The option -mcbranchdi has been deprecated. Specifying it will
648 result in a warning and will not influence code generation.
649 * The option -mcmpeqdi has been deprecated. Specifying it will result
650 in a warning and will not influence code generation.
654 This is the [25]list of problem reports (PRs) from GCC's bug tracking
655 system that are known to be fixed in the 4.9.1 release. This list might
656 not be complete (that is, it is possible that some PRs that have been
657 fixed are not listed here).
659 Version 4.0 of the [26]OpenMP specification is supported even in
660 Fortran, not just C and C++.
664 This is the [27]list of problem reports (PRs) from GCC's bug tracking
665 system that are known to be fixed in the 4.9.2 release. This list might
666 not be complete (that is, it is possible that some PRs that have been
667 fixed are not listed here).
671 This is the [28]list of problem reports (PRs) from GCC's bug tracking
672 system that are known to be fixed in the 4.9.3 release. This list might
673 not be complete (that is, it is possible that some PRs that have been
674 fixed are not listed here).
677 For questions related to the use of GCC, please consult these web
678 pages and the [29]GCC manuals. If that fails, the
679 [30]gcc-help@gcc.gnu.org mailing list might help. Comments on these
680 web pages and the development of GCC are welcome on our developer
681 list at [31]gcc@gcc.gnu.org. All of [32]our lists have public
684 Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and
685 distribution of this entire article is permitted in any medium,
686 provided this notice is preserved.
688 These pages are [34]maintained by the GCC team. Last modified
693 1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html
694 2. https://gcc.gnu.org/PR60825
695 3. https://gcc.gnu.org/gcc-4.9/porting_to.html
696 4. http://openmp.org/wp/openmp-specifications/
697 5. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908
698 6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252
699 7. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html
700 8. https://www.cilkplus.org/
701 9. http://gcc.gnu.org/projects/cxx1y.html
702 10. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html
703 11. http://gcc.gnu.org/projects/cxx1y.html
704 12. http://gcc.gnu.org/projects/cxx1y.html
705 13. http://gcc.gnu.org/projects/cxx1y.html
706 14. http://gcc.gnu.org/projects/cxx1y.html
707 15. http://gcc.gnu.org/projects/cxx1y.html
708 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf
709 17. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011
710 18. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014
711 19. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html
712 20. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html
713 21. https://gcc.gnu.org/wiki/Fortran2003Status
714 22. https://gcc.gnu.org/wiki/Fortran2008Status
715 23. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html
716 24. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html
717 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1
718 26. http://openmp.org/wp/openmp-specifications/
719 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2
720 28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3
721 29. https://gcc.gnu.org/onlinedocs/
722 30. mailto:gcc-help@gcc.gnu.org
723 31. mailto:gcc@gcc.gnu.org
724 32. https://gcc.gnu.org/lists.html
725 33. http://www.fsf.org/
726 34. https://gcc.gnu.org/about.html
727 35. http://validator.w3.org/check/referer
728 ======================================================================
729 http://gcc.gnu.org/gcc-4.8/index.html
730 GCC 4.8 Release Series
734 The [1]GNU project and the GCC developers are pleased to announce the
735 release of GCC 4.8.5.
737 This release is a bug-fix release, containing fixes for regressions in
738 GCC 4.8.4 relative to previous releases of GCC.
743 June 23, 2015 ([2]changes, [3]documentation)
746 December 19, 2014 ([4]changes, [5]documentation)
749 May 22, 2014 ([6]changes, [7]documentation)
752 October 16, 2013 ([8]changes, [9]documentation)
755 May 31, 2013 ([10]changes, [11]documentation)
758 March 22, 2013 ([12]changes, [13]documentation)
760 References and Acknowledgements
762 GCC used to stand for the GNU C Compiler, but since the compiler
763 supports several other languages aside from C, it now stands for the
764 GNU Compiler Collection.
766 A list of [14]successful builds is updated as new information becomes
769 The GCC developers would like to thank the numerous people that have
770 contributed new features, improvements, bug fixes, and other changes as
771 well as test results to GCC. This [15]amazing group of volunteers is
772 what makes GCC successful.
774 For additional information about GCC please refer to the [16]GCC
775 project web site or contact the [17]GCC development mailing list.
777 To obtain GCC please use [18]our mirror sites or [19]our SVN server.
780 For questions related to the use of GCC, please consult these web
781 pages and the [20]GCC manuals. If that fails, the
782 [21]gcc-help@gcc.gnu.org mailing list might help. Comments on these
783 web pages and the development of GCC are welcome on our developer
784 list at [22]gcc@gcc.gnu.org. All of [23]our lists have public
787 Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and
788 distribution of this entire article is permitted in any medium,
789 provided this notice is preserved.
791 These pages are [25]maintained by the GCC team. Last modified
796 1. http://www.gnu.org/
797 2. http://gcc.gnu.org/gcc-4.8/changes.html
798 3. https://gcc.gnu.org/onlinedocs/4.8.5/
799 4. http://gcc.gnu.org/gcc-4.8/changes.html
800 5. https://gcc.gnu.org/onlinedocs/4.8.4/
801 6. http://gcc.gnu.org/gcc-4.8/changes.html
802 7. https://gcc.gnu.org/onlinedocs/4.8.3/
803 8. http://gcc.gnu.org/gcc-4.8/changes.html
804 9. https://gcc.gnu.org/onlinedocs/4.8.2/
805 10. http://gcc.gnu.org/gcc-4.8/changes.html
806 11. https://gcc.gnu.org/onlinedocs/4.8.1/
807 12. http://gcc.gnu.org/gcc-4.8/changes.html
808 13. https://gcc.gnu.org/onlinedocs/4.8.0/
809 14. http://gcc.gnu.org/gcc-4.8/buildstat.html
810 15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
811 16. http://gcc.gnu.org/index.html
812 17. mailto:gcc@gcc.gnu.org
813 18. http://gcc.gnu.org/mirrors.html
814 19. http://gcc.gnu.org/svn.html
815 20. https://gcc.gnu.org/onlinedocs/
816 21. mailto:gcc-help@gcc.gnu.org
817 22. mailto:gcc@gcc.gnu.org
818 23. https://gcc.gnu.org/lists.html
819 24. http://www.fsf.org/
820 25. https://gcc.gnu.org/about.html
821 26. http://validator.w3.org/check/referer
822 ======================================================================
823 http://gcc.gnu.org/gcc-4.8/changes.html
824 GCC 4.8 Release Series
825 Changes, New Features, and Fixes
829 GCC now uses C++ as its implementation language. This means that to
830 build GCC from sources, you will need a C++ compiler that understands
831 C++ 2003. For more details on the rationale and specific changes,
832 please refer to the [1]C++ conversion page.
834 To enable the Graphite framework for loop optimizations you now need
835 CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from
836 the [2]GCC infrastructure directory. The installation manual contains
837 more information about requirements to build GCC.
839 GCC now uses a more aggressive analysis to derive an upper bound for
840 the number of iterations of loops using constraints imposed by language
841 standards. This may cause non-conforming programs to no longer work as
842 expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new
843 option, -fno-aggressive-loop-optimizations, was added to disable this
844 aggressive analysis. In some loops that have known constant number of
845 iterations, but undefined behavior is known to occur in the loop before
846 reaching or during the last iteration, GCC will warn about the
847 undefined behavior in the loop instead of deriving lower upper bound of
848 the number of iterations for the loop. The warning can be disabled with
849 -Wno-aggressive-loop-optimizations.
851 On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules
852 for the layout of vectors that could lead to wrong code being
853 generated. Vectors larger than 8 bytes in size are now by default
854 aligned to an 8-byte boundary. This is an ABI change: code that makes
855 explicit use of vector types may be incompatible with binary objects
856 built with older versions of GCC. Auto-vectorized code is not affected
859 On AVR, support has been removed for the command-line option
860 -mshort-calls deprecated in GCC 4.7.
862 On AVR, the configure option --with-avrlibc supported since GCC 4.7.2
863 is turned on per default for all non-RTEMS configurations. This option
864 arranges for a better integration of [3]AVR Libc with avr-gcc. For
865 technical details, see [4]PR54461. To turn off the option in non-RTEMS
866 configurations, use --with-avrlibc=no. If the compiler is configured
867 for RTEMS, the option is always turned off.
869 More information on porting to GCC 4.8 from previous versions of GCC
870 can be found in the [5]porting guide for this release.
872 General Optimizer Improvements (and Changes)
874 * DWARF4 is now the default when generating DWARF debug information.
875 When -g is used on a platform that uses DWARF debugging
876 information, GCC will now default to -gdwarf-4
877 -fno-debug-types-section.
878 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information
879 consumers support DWARF4 by default. Before GCC 4.8 the default
880 version used was DWARF2. To make GCC 4.8 generate an older DWARF
881 version use -g together with -gdwarf-2 or -gdwarf-3. The default
882 for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf.
883 * A new general optimization level, -Og, has been introduced. It
884 addresses the need for fast compilation and a superior debugging
885 experience while providing a reasonable level of run-time
886 performance. Overall experience for development should be better
887 than the default optimization level -O0.
888 * A new option -ftree-partial-pre was added to control the partial
889 redundancy elimination (PRE) optimization. This option is enabled
890 by default at the -O3 optimization level, and it makes PRE more
892 * The option -fconserve-space has been removed; it was no longer
893 useful on most targets since GCC supports putting variables into
894 BSS without making them common.
895 * The struct reorg and matrix reorg optimizations (command-line
896 options -fipa-struct-reorg and -fipa-matrix-reorg) have been
897 removed. They did not always work correctly, nor did they work with
898 link-time optimization (LTO), hence were only applicable to
899 programs consisting of a single translation unit.
900 * Several scalability bottle-necks have been removed from GCC's
901 optimization passes. Compilation of extremely large functions, e.g.
902 due to the use of the flatten attribute in the "Eigen" C++ linear
903 algebra templates library, is significantly faster than previous
905 * Link-time optimization (LTO) improvements:
906 + LTO partitioning has been rewritten for better reliability and
907 maintanibility. Several important bugs leading to link
908 failures have been fixed.
909 * Interprocedural optimization improvements:
910 + A new symbol table has been implemented. It builds on existing
911 callgraph and varpool modules and provide a new API. Unusual
912 symbol visibilities and aliases are handled more consistently
913 leading to, for example, more aggressive unreachable code
915 + The inline heuristic can now bypass limits on the size of of
916 inlined functions when the inlining is particularly
917 profitable. This happens, for example, when loop bounds or
918 array strides get propagated.
919 + Values passed through aggregates (either by value or
920 reference) are now propagated at the inter-procedural level
921 leading to better inlining decisions (for example in the case
922 of Fortran array descriptors) and devirtualization.
923 * [6]AddressSanitizer , a fast memory error detector, has been added
924 and can be enabled via -fsanitize=address. Memory access
925 instructions will be instrumented to detect heap-, stack-, and
926 global-buffer overflow as well as use-after-free bugs. To get nicer
927 stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is
928 available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on
930 * [7]ThreadSanitizer has been added and can be enabled via
931 -fsanitize=thread. Instructions will be instrumented to detect data
932 races. The ThreadSanitizer is available on x86-64 GNU/Linux.
933 * A new local register allocator (LRA) has been implemented, which
934 replaces the 26 year old reload pass and improves generated code
935 quality. For now it is active on the IA-32 and x86-64 targets.
936 * Support for transactional memory has been implemented on the
937 following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and
940 New Languages and Language specific improvements
944 * Each diagnostic emitted now includes the original source line and a
945 caret '^' indicating the column. The option
946 -fno-diagnostics-show-caret suppresses this information.
947 * The option -ftrack-macro-expansion=2 is now enabled by default.
948 This allows the compiler to display the macro expansion stack in
949 diagnostics. Combined with the caret information, an example
950 diagnostic showing these two features is:
952 t.c:1:94: error: invalid operands to binary < (have ‘struct mystruct’ and ‘float
954 #define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _
955 _b = (B); __a < __b ? __b : __a; })
958 t.c:7:7: note: in expansion of macro 'MYMAX'
962 * A new -Wsizeof-pointer-memaccess warning has been added (also
963 enabled by -Wall) to warn about suspicious length parameters to
964 certain string and memory built-in functions if the argument uses
965 sizeof. This warning warns e.g. about memset (ptr, 0, sizeof
966 (ptr)); if ptr is not an array, but a pointer, and suggests a
967 possible fix, or about memcpy (&foo, ptr, sizeof (&foo));.
968 * The new option -Wpedantic is an alias for -pedantic, which is now
969 deprecated. The forms -Wno-pedantic, -Werror=pedantic, and
970 -Wno-error=pedantic work in the same way as for any other -W
971 option. One caveat is that -Werror=pedantic is not equivalent to
972 -pedantic-errors, since the latter makes into errors some warnings
973 that are not controlled by -Wpedantic, and the former only affects
974 diagnostics that are disabled when using -Wno-pedantic.
975 * The option -Wshadow no longer warns if a declaration shadows a
976 function declaration, unless the former declares a function or
977 pointer to function, because this is [8]a common and valid case in
982 * G++ now implements the [9]C++11 thread_local keyword; this differs
983 from the GNU __thread keyword primarily in that it allows dynamic
984 initialization and destruction semantics. Unfortunately, this
985 support requires a run-time penalty for references to
986 non-function-local thread_local variables defined in a different
987 translation unit even if they don't need dynamic initialization, so
988 users may want to continue to use __thread for TLS variables with
989 static initialization semantics.
990 If the programmer can be sure that no use of the variable in a
991 non-defining TU needs to trigger dynamic initialization (either
992 because the variable is statically initialized, or a use of the
993 variable in the defining TU will be executed before any uses in
994 another TU), they can avoid this overhead with the
995 -fno-extern-tls-init option.
996 OpenMP threadprivate variables now also support dynamic
997 initialization and destruction by the same mechanism.
998 * G++ now implements the [10]C++11 attribute syntax, e.g.
1000 [[noreturn]] void f();
1002 and also the alignment specifier, e.g.
1004 alignas(double) int i;
1006 * G++ now implements [11]C++11 inheriting constructors, e.g.
1008 struct A { A(int); };
1009 struct B: A { using A::A; }; // defines B::B(int)
1012 * As of GCC 4.8.1, G++ implements the change to decltype semantics
1016 decltype(f()) g(); // OK, return type of f() is not required to be complete.
1018 * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g.
1020 struct A { int f() &; };
1021 int i = A().f(); // error, f() requires an lvalue object
1023 * G++ now supports a -std=c++1y option for experimentation with
1024 features proposed for the next revision of the standard, expected
1025 around 2014. Currently the only difference from -std=c++11 is
1026 support for return type deduction in normal functions, as proposed
1027 in [14]N3386. Status of C++1y features in GCC 4.8 can be found
1029 * The G++ namespace association extension, __attribute ((strong)),
1030 has been deprecated. Inline namespaces should be used instead.
1031 * G++ now supports a -fext-numeric-literal option to control whether
1032 GNU numeric literal suffixes are accepted as extensions or
1033 processed as C++11 user-defined numeric literal suffixes. The flag
1034 is on (use suffixes for GNU literals) by default for -std=gnu++*,
1035 and -std=c++98. The flag is off (use suffixes for user-defined
1036 literals) by default for -std=c++11 and later.
1038 Runtime Library (libstdc++)
1040 * [16]Improved experimental support for the new ISO C++ standard,
1042 + forward_list meets the allocator-aware container requirements;
1043 + this_thread::sleep_for(), this_thread::sleep_until() and
1044 this_thread::yield() are defined without requiring the
1045 configure option --enable-libstdcxx-time;
1046 * Improvements to <random>:
1047 + SSE optimized normal_distribution.
1048 + Use of hardware RNG instruction for random_device on new x86
1049 processors (requires the assembler to support the
1052 + New random number engine simd_fast_mersenne_twister_engine
1053 with an optimized SSE implementation.
1054 + New random number distributions beta_distribution,
1055 normal_mv_distribution, rice_distribution,
1056 nakagami_distribution, pareto_distribution, k_distribution,
1057 arcsine_distribution, hoyt_distribution.
1058 * Added --disable-libstdcxx-verbose configure option to disable
1059 diagnostic messages issued when a process terminates abnormally.
1060 This may be useful for embedded systems to reduce the size of
1061 executables that link statically to the library.
1065 * Compatibility notice:
1066 + Module files: The version of module files (.mod) has been
1067 incremented. Fortran MODULEs compiled by earlier GCC versions
1068 have to be recompiled, when they are USEd by files compiled
1069 with GCC 4.8. GCC 4.8 is not able to read .mod files created
1070 by earlier versions; attempting to do so gives an error
1072 Note: The ABI of the produced assembler data itself has not
1073 changed; object files and libraries are fully compatible with
1074 older versions except as noted below.
1075 + ABI: Some internal names (used in the assembler/object file)
1076 have changed for symbols declared in the specification part of
1077 a module. If an affected module – or a file using it via use
1078 association – is recompiled, the module and all files which
1079 directly use such symbols have to be recompiled as well. This
1080 change only affects the following kind of module symbols:
1081 o Procedure pointers. Note: C-interoperable function
1082 pointers (type(c_funptr)) are not affected nor are
1083 procedure-pointer components.
1084 o Deferred-length character strings.
1085 * The [17]BACKTRACE intrinsic subroutine has been added. It shows a
1086 backtrace at an arbitrary place in user code; program execution
1087 continues normally afterwards.
1088 * The [18]-Wc-binding-type warning option has been added (disabled by
1089 default). It warns if the a variable might not be C interoperable;
1090 in particular, if the variable has been declared using an intrinsic
1091 type with default kind instead of using a kind parameter defined
1092 for C interoperability in the intrinsic ISO_C_Binding module.
1093 Before, this warning was always printed. The -Wc-binding-type
1094 option is enabled by -Wall.
1095 * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line
1096 options have been added, which diagnose when code is inserted for
1097 automatic (re)allocation of a variable during assignment. This
1098 option can be used to decide whether it is safe to use
1099 [20]-fno-realloc-lhs. Additionally, it can be used to find
1100 automatic (re)allocation in hot loops. (For arrays, replacing
1101 "var=" by "var(:)=" disables the automatic reallocation.)
1102 * The [21]-Wcompare-reals command-line option has been added. When
1103 this is set, warnings are issued when comparing REAL or COMPLEX
1104 types for equality and inequality; consider replacing a == b by
1105 abs(a−b) < eps with a suitable eps. -Wcompare-reals is enabled by
1107 * The [22]-Wtarget-lifetime command-line option has been added
1108 (enabled with -Wall), which warns if the pointer in a pointer
1109 assignment might outlive its target.
1110 * Reading floating point numbers which use "q" for the exponential
1111 (such as 4.0q0) is now supported as vendor extension for better
1112 compatibility with old data files. It is strongly recommended to
1113 use for I/O the equivalent but standard conforming "e" (such as
1115 (For Fortran source code, consider replacing the "q" in
1116 floating-point literals by a kind parameter (e.g. 4.0e0_qp with a
1117 suitable qp). Note that – in Fortran source code – replacing "q" by
1118 a simple "e" is not equivalent.)
1119 * The GFORTRAN_TMPDIR environment variable for specifying a
1120 non-default directory for files opened with STATUS="SCRATCH", is
1121 not used anymore. Instead gfortran checks the POSIX/GNU standard
1122 TMPDIR environment variable. If TMPDIR is not defined, gfortran
1123 falls back to other methods to determine the directory for
1124 temporary files as documented in the [23]user manual.
1126 + Support for unlimited polymorphic variables (CLASS(*)) has
1127 been added. Nonconstant character lengths are not yet
1130 + Assumed types (TYPE(*)) are now supported.
1131 + Experimental support for assumed-rank arrays (dimension(..))
1132 has been added. Note that currently gfortran's own array
1133 descriptor is used, which is different from the one defined in
1134 TS29113, see [26]gfortran's header file or use the [27]Chasm
1135 Language Interoperability Tools.
1139 * GCC 4.8.2 provides a complete implementation of the Go 1.1.2
1141 * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1
1142 release. The library support is not quite complete.
1143 * Go has been tested on GNU/Linux and Solaris platforms for various
1144 processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may
1145 work on other platforms as well.
1147 New Targets and Target Specific Improvements
1151 * A new port has been added to support AArch64, the new 64-bit
1152 architecture from ARM. Note that this is a separate port from the
1153 existing 32-bit ARM port.
1154 * The port provides initial support for the Cortex-A53 and the
1155 Cortex-A57 processors with the command line options
1156 -mcpu=cortex-a53 and -mcpu=cortex-a57.
1157 * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769
1158 has been added and can be enabled by giving the
1159 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by
1160 default by configuring GCC with the --enable-fix-cortex-a53-835769
1165 * Initial support has been added for the AArch32 extensions defined
1166 in the ARMv8 architecture.
1167 * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs.
1168 * A new option, -mcpu=marvell-pj4, has been added to generate code
1169 for the Marvell PJ4 processor.
1170 * The compiler can now automatically generate the VFMA, VFMS, REVSH
1171 and REV16 instructions.
1172 * A new vectorizer cost model for Advanced SIMD configurations to
1173 improve the auto-vectorization strategies used.
1174 * The scheduler now takes into account the number of live registers
1175 to reduce the amount of spilling that can occur. This should
1176 improve code performance in large functions. The limit can be
1177 removed by using the option -fno-sched-pressure.
1178 * Improvements have been made to the Marvell iWMMX code generation
1179 and support for the iWMMX2 SIMD unit has been added. The option
1180 -mcpu=iwmmxt2 can be used to enable code generation for the latter.
1181 * A number of code generation improvements for Thumb2 to reduce code
1182 size when compiling for the M-profile processors.
1183 * The RTEMS (arm-rtems) port has been updated to use the EABI.
1184 * Code generation support for the old FPA and Maverick floating-point
1185 architectures has been removed. Ports that previously relied on
1186 these features have also been removed. This includes the targets:
1187 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
1188 + arm*-*-elf (use arm*-*-eabi)
1189 + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
1190 + arm*-*-ecos-elf (no alternative)
1191 + arm*-*-freebsd (no alternative)
1192 + arm*-wince-pe* (no alternative).
1196 * Support for the "Embedded C" fixed-point has been added. For
1197 details, see the [28]GCC wiki and the [29]user manual. The support
1199 * A new print modifier %r for register operands in inline assembler
1200 is supported. It will print the raw register number without the
1201 register prefix 'r':
1202 /* Return the most significant byte of 'val', a 64-bit value. */
1204 unsigned char msb (long long val)
1207 __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val));
1210 The inline assembler in this example will generate code like
1212 provided c is allocated to R24 and val is allocated to R8…R15. This
1213 works because the GNU assembler accepts plain register numbers
1214 without register prefix.
1215 * Static initializers with 3-byte symbols are supported now:
1216 extern const __memx char foo;
1217 const __memx void *pfoo = &foo;
1218 This requires at least Binutils 2.23.
1222 * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with
1223 SSE extensions disabled. Since the x86-64 ABI requires 16 byte
1224 stack alignment, this is ABI incompatible and intended to be used
1225 in controlled environments where stack space is an important
1226 limitation. This option will lead to wrong code when functions
1227 compiled with 16 byte stack alignment (such as functions from a
1228 standard library) are called with misaligned stack. In this case,
1229 SSE instructions may lead to misaligned memory access traps. In
1230 addition, variable arguments will be handled incorrectly for 16
1231 byte aligned objects (including x87 long double and __int128),
1232 leading to wrong results. You must build all modules with
1233 -mpreferred-stack-boundary=3, including any libraries. This
1234 includes the system libraries and startup modules.
1235 * Support for the new Intel processor codename Broadwell with RDSEED,
1236 ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw,
1237 -mrdseed command-line options.
1238 * Support for the Intel RTM and HLE intrinsics, built-in functions
1239 and code generation is available via -mrtm and -mhle.
1240 * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets.
1241 Intrinsics and built-in functions are available via -mfxsr, -mxsave
1242 and -mxsaveopt respectively.
1243 * New -maddress-mode=[short|long] options for x32.
1244 -maddress-mode=short overrides default 64-bit addresses to 32-bit
1245 by emitting the 0x67 address-size override prefix. This is the
1246 default address mode for x32.
1247 * New built-in functions to detect run-time CPU type and ISA:
1248 + A built-in function __builtin_cpu_is has been added to detect
1249 if the run-time CPU is of a particular type. It returns a
1250 positive integer on a match and zero otherwise. It accepts one
1251 string literal argument, the CPU name. For example,
1252 __builtin_cpu_is("westmere") returns a positive integer if the
1253 run-time CPU is an Intel Core i7 Westmere processor. Please
1254 refer to the [30]user manual for the list of valid CPU names
1256 + A built-in function __builtin_cpu_supports has been added to
1257 detect if the run-time CPU supports a particular ISA feature.
1258 It returns a positive integer on a match and zero otherwise.
1259 It accepts one string literal argument, the ISA feature. For
1260 example, __builtin_cpu_supports("ssse3") returns a positive
1261 integer if the run-time CPU supports SSSE3 instructions.
1262 Please refer to the [31]user manual for the list of valid ISA
1264 Caveat: If these built-in functions are called before any static
1265 constructors are invoked, like during IFUNC initialization, then
1266 the CPU detection initialization must be explicitly run using this
1267 newly provided built-in function, __builtin_cpu_init. The
1268 initialization needs to be done only once. For example, this is how
1269 the invocation would look like inside an IFUNC initializer:
1270 static void (*some_ifunc_resolver(void))(void)
1272 __builtin_cpu_init();
1273 if (__builtin_cpu_is("amdfam10h") ...
1274 if (__builtin_cpu_supports("popcnt") ...
1277 * Function Multiversioning Support with G++:
1278 It is now possible to create multiple function versions each
1279 targeting a specific processor and/or ISA. Function versions have
1280 the same signature but different target attributes. For example,
1281 here is a program with function versions:
1282 __attribute__ ((target ("default")))
1288 __attribute__ ((target ("sse4.2")))
1297 assert ((*p)() == foo());
1301 Please refer to this [32]wiki for more information.
1302 * The x86 back end has been improved to allow option -fschedule-insns
1303 to work reliably. This option can be used to schedule instructions
1304 better and leads to improved performace in certain cases.
1305 * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437
1306 from the Mingw-w64 trunk.
1307 * Support for new AMD family 15h processors (Steamroller core) is now
1308 available through the -march=bdver3 and -mtune=bdver3 options.
1309 * Support for new AMD family 16h processors (Jaguar core) is now
1310 available through the -march=btver2 and -mtune=btver2 options.
1314 * This target now supports the -fstack-usage command-line option.
1318 * GCC can now generate code specifically for the R4700, Broadcom XLP
1319 and MIPS 34kn processors. The associated -march options are
1320 -march=r4700, -march=xlp and -march=34kn respectively.
1321 * GCC now generates better DSP code for MIPS 74k cores thanks to
1322 further scheduling optimizations.
1323 * The MIPS port now supports the -fstack-check option.
1324 * GCC now passes the -mmcu and -mno-mcu options to the assembler.
1325 * Previous versions of GCC would silently accept -fpic and -fPIC for
1326 -mno-abicalls targets like mips*-elf. This combination was not
1327 intended or supported, and did not generate position-independent
1328 code. GCC 4.8 now reports an error when this combination is used.
1330 PowerPC / PowerPC64 / RS6000
1332 * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save,
1333 restore or update the VRSAVE register by default. The respective
1334 operating systems manage the VRSAVE register directly.
1335 * Large TOC support has been added for AIX through the command line
1336 option -mcmodel=large.
1337 * Native Thread-Local Storage support has been added for AIX.
1338 * VMX (Altivec) and VSX instruction sets now are enabled implicitly
1339 when targetting processors that support those hardware features on
1344 * This target will now issue a warning message whenever multiple fast
1345 interrupt handlers are found in the same compilation unit. This
1346 feature can be turned off by the new
1347 -mno-warn-multiple-fast-interrupts command-line option.
1351 * Support for the IBM zEnterprise zEC12 processor has been added.
1352 When using the -march=zEC12 option, the compiler will generate code
1353 making use of the following new instructions:
1354 + load and trap instructions
1355 + 2 new compare and trap instructions
1356 + rotate and insert selected bits - without CC clobber
1357 The -mtune=zEC12 option enables zEC12 specific instruction
1358 scheduling without making use of new instructions.
1359 * Register pressure sensitive instruction scheduling is enabled by
1361 * The ifunc function attribute is enabled by default.
1362 * memcpy and memcmp invokations on big memory chunks or with run time
1363 lengths are not generated inline anymore when tuning for z10 or
1364 higher. The purpose is to make use of the IFUNC optimized versions
1369 * The default alignment settings have been reduced to be less
1370 aggressive. This results in more compact code for optimization
1371 levels other than -Os.
1372 * Improved support for the __atomic built-in functions:
1373 + A new option -matomic-model=model selects the model for the
1374 generated atomic sequences. The following models are
1378 Software gUSA sequences (SH3* and SH4* only). On
1379 SH4A targets this will now also partially utilize
1380 the movco.l and movli.l instructions. This is the
1381 default when the target is sh3*-*-linux* or
1385 Hardware movco.l / movli.l sequences (SH4A only).
1388 Software thread control block sequences.
1391 Software interrupt flipping sequences (privileged
1392 mode only). This is the default when the target is
1393 sh1*-*-linux* or sh2*-*-linux*.
1396 Generates function calls to the respective __atomic
1397 built-in functions. This is the default for SH64
1398 targets or when the target is not sh*-*-linux*.
1400 + The option -msoft-atomic has been deprecated. It is now an
1401 alias for -matomic-model=soft-gusa.
1402 + A new option -mtas makes the compiler generate the tas.b
1403 instruction for the __atomic_test_and_set built-in function
1404 regardless of the selected atomic model.
1405 + The __sync functions in libgcc now reflect the selected atomic
1406 model when building the toolchain.
1407 * Added support for the mov.b and mov.w instructions with
1408 displacement addressing.
1409 * Added support for the SH2A instructions movu.b and movu.w.
1410 * Various improvements to code generated for integer arithmetic.
1411 * Improvements to conditional branches and code that involves the T
1412 bit. A new option -mzdcbranch tells the compiler to favor
1413 zero-displacement branches. This is enabled by default for SH4*
1415 * The pref instruction will now be emitted by the __builtin_prefetch
1416 built-in function for SH3* targets.
1417 * The fmac instruction will now be emitted by the fmaf standard
1418 function and the __builtin_fmaf built-in function.
1419 * The -mfused-madd option has been deprecated in favor of the
1420 machine-independent -ffp-contract option. Notice that the fmac
1421 instruction will now be generated by default for expressions like a
1422 * b + c. This is due to the compiler default setting
1424 * Added new options -mfsrra and -mfsca to allow the compiler using
1425 the fsrra and fsca instructions on targets other than SH4A (where
1426 they are already enabled by default).
1427 * Added support for the __builtin_bswap32 built-in function. It is
1428 now expanded as a sequence of swap.b and swap.w instructions
1429 instead of a library function call.
1430 * The behavior of the -mieee option has been fixed and the negative
1431 form -mno-ieee has been added to control the IEEE conformance of
1432 floating point comparisons. By default -mieee is now enabled and
1433 the option -ffinite-math-only implicitly sets -mno-ieee.
1434 * Added support for the built-in functions __builtin_thread_pointer
1435 and __builtin_set_thread_pointer. This assumes that GBR is used to
1436 hold the thread pointer of the current thread. Memory loads and
1437 stores relative to the address returned by __builtin_thread_pointer
1438 will now also utilize GBR based displacement address modes.
1439 * The -mdiv= option for targets other than SHmedia has been fixed and
1444 * Added optimized instruction scheduling for Niagara4.
1448 * Added support for the -mcmodel=MODEL command-line option. The
1449 models supported are small and large.
1453 * This target now supports the E3V5 architecture via the use of the
1454 new -mv850e3v5 command-line option. It also has experimental
1455 support for the e3v5 LOOP instruction which can be enabled via the
1456 new -mloop command-line option.
1460 * This target now supports the -fstack-usage command-line option.
1466 * Executables are now linked against shared libgcc by default. The
1467 previous default was to link statically, which can still be done by
1468 explicitly specifying -static or static-libgcc on the command line.
1469 However it is strongly advised against, as it will cause problems
1470 for any application that makes use of DLLs compiled by GCC. It
1471 should be alright for a monolithic stand-alone application that
1472 only links against the Windows DLLs, but offers little or no
1477 This is the [33]list of problem reports (PRs) from GCC's bug tracking
1478 system that are known to be fixed in the 4.8.1 release. This list might
1479 not be complete (that is, it is possible that some PRs that have been
1480 fixed are not listed here).
1482 The C++11 <chrono> std::chrono::system_clock and
1483 std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they
1484 both are now separate (never typedefs of each other), both use
1485 std::chrono::nanoseconds resolution, on most GNU/Linux configurations
1486 std::chrono::steady_clock is now finally monotonic, and both classes
1487 are mangled differently than in the previous GCC releases.
1488 std::chrono::system_clock::now() with std::chrono::microseconds resp.
1489 std::chrono::seconds resolution is still exported for backwards
1490 compatibility with default configured libstdc++. Note that libstdc++
1491 configured with --enable-libstdcxx-time= used to be ABI incompatible
1492 with default configured libstdc++ for those two classes and no ABI
1493 compatibility can be offered for those configurations, so any C++11
1494 code that uses those classes and has been compiled and linked against
1495 libstdc++ configured with the non-default --enable-libstdcxx-time=
1496 configuration option needs to be recompiled.
1500 This is the [34]list of problem reports (PRs) from GCC's bug tracking
1501 system that are known to be fixed in the 4.8.2 release. This list might
1502 not be complete (that is, it is possible that some PRs that have been
1503 fixed are not listed here).
1507 This is the [35]list of problem reports (PRs) from GCC's bug tracking
1508 system that are known to be fixed in the 4.8.3 release. This list might
1509 not be complete (that is, it is possible that some PRs that have been
1510 fixed are not listed here).
1512 Support for the new powerpc64le-linux platform has been added. It
1513 defaults to generating code that conforms to the ELFV2 ABI.
1517 This is the [36]list of problem reports (PRs) from GCC's bug tracking
1518 system that are known to be fixed in the 4.8.4 release. This list might
1519 not be complete (that is, it is possible that some PRs that have been
1520 fixed are not listed here).
1524 This is the [37]list of problem reports (PRs) from GCC's bug tracking
1525 system that are known to be fixed in the 4.8.5 release. This list might
1526 not be complete (that is, it is possible that some PRs that have been
1527 fixed are not listed here).
1530 For questions related to the use of GCC, please consult these web
1531 pages and the [38]GCC manuals. If that fails, the
1532 [39]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1533 web pages and the development of GCC are welcome on our developer
1534 list at [40]gcc@gcc.gnu.org. All of [41]our lists have public
1537 Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and
1538 distribution of this entire article is permitted in any medium,
1539 provided this notice is preserved.
1541 These pages are [43]maintained by the GCC team. Last modified
1546 1. https://gcc.gnu.org/wiki/cxx-conversion
1547 2. ftp://gcc.gnu.org/pub/gcc/infrastructure/
1548 3. http://www.nongnu.org/avr-libc/
1549 4. https://gcc.gnu.org/PR54461
1550 5. https://gcc.gnu.org/gcc-4.8/porting_to.html
1551 6. https://code.google.com/p/address-sanitizer/
1552 7. https://code.google.com/p/data-race-test/wiki/ThreadSanitizer
1553 8. https://lkml.org/lkml/2006/11/28/239
1554 9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
1555 10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
1556 11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
1557 12. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf
1558 13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
1559 14. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html
1560 15. http://gcc.gnu.org/projects/cxx1y.html
1561 16. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011
1562 17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html
1563 18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
1564 19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
1565 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
1566 21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
1567 22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
1568 23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html
1569 24. https://gcc.gnu.org/wiki/Fortran2003Status
1570 25. https://gcc.gnu.org/wiki/TS29113Status
1571 26. https://gcc.gnu.org/viewcvs/trunk/libgfortran/libgfortran.h?content-type=text/plain&view=co
1572 27. http://chasm-interop.sourceforge.net/
1573 28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support
1574 29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html
1575 30. https://gcc.gnu.org/onlinedocs/gcc/X86-Built-in-Functions.html#X86-Built-in-Functions
1576 31. https://gcc.gnu.org/onlinedocs/gcc/X86-Built-in-Functions.html#X86-Built-in-Functions
1577 32. https://gcc.gnu.org/wiki/FunctionMultiVersioning
1578 33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1
1579 34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2
1580 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3
1581 36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4
1582 37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5
1583 38. https://gcc.gnu.org/onlinedocs/
1584 39. mailto:gcc-help@gcc.gnu.org
1585 40. mailto:gcc@gcc.gnu.org
1586 41. https://gcc.gnu.org/lists.html
1587 42. http://www.fsf.org/
1588 43. https://gcc.gnu.org/about.html
1589 44. http://validator.w3.org/check/referer
1590 ======================================================================
1591 http://gcc.gnu.org/gcc-4.7/index.html
1592 GCC 4.7 Release Series
1596 The [1]GNU project and the GCC developers are pleased to announce the
1597 release of GCC 4.7.4.
1599 This release is a bug-fix release, containing fixes for regressions in
1600 GCC 4.7.3 relative to previous releases of GCC.
1605 June 12, 2014 ([2]changes, [3]documentation)
1608 April 11, 2013 ([4]changes, [5]documentation)
1611 September 20, 2012 ([6]changes, [7]documentation)
1614 June 14, 2012 ([8]changes, [9]documentation)
1617 March 22, 2012 ([10]changes, [11]documentation)
1619 References and Acknowledgements
1621 GCC used to stand for the GNU C Compiler, but since the compiler
1622 supports several other languages aside from C, it now stands for the
1623 GNU Compiler Collection.
1625 A list of [12]successful builds is updated as new information becomes
1628 The GCC developers would like to thank the numerous people that have
1629 contributed new features, improvements, bug fixes, and other changes as
1630 well as test results to GCC. This [13]amazing group of volunteers is
1631 what makes GCC successful.
1633 For additional information about GCC please refer to the [14]GCC
1634 project web site or contact the [15]GCC development mailing list.
1636 To obtain GCC please use [16]our mirror sites or [17]our SVN server.
1639 For questions related to the use of GCC, please consult these web
1640 pages and the [18]GCC manuals. If that fails, the
1641 [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1642 web pages and the development of GCC are welcome on our developer
1643 list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
1646 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
1647 distribution of this entire article is permitted in any medium,
1648 provided this notice is preserved.
1650 These pages are [23]maintained by the GCC team. Last modified
1655 1. http://www.gnu.org/
1656 2. http://gcc.gnu.org/gcc-4.7/changes.html
1657 3. https://gcc.gnu.org/onlinedocs/4.7.4/
1658 4. http://gcc.gnu.org/gcc-4.7/changes.html
1659 5. https://gcc.gnu.org/onlinedocs/4.7.3/
1660 6. http://gcc.gnu.org/gcc-4.7/changes.html
1661 7. https://gcc.gnu.org/onlinedocs/4.7.2/
1662 8. http://gcc.gnu.org/gcc-4.7/changes.html
1663 9. https://gcc.gnu.org/onlinedocs/4.7.1/
1664 10. http://gcc.gnu.org/gcc-4.7/changes.html
1665 11. https://gcc.gnu.org/onlinedocs/4.7.0/
1666 12. http://gcc.gnu.org/gcc-4.7/buildstat.html
1667 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
1668 14. http://gcc.gnu.org/index.html
1669 15. mailto:gcc@gcc.gnu.org
1670 16. http://gcc.gnu.org/mirrors.html
1671 17. http://gcc.gnu.org/svn.html
1672 18. https://gcc.gnu.org/onlinedocs/
1673 19. mailto:gcc-help@gcc.gnu.org
1674 20. mailto:gcc@gcc.gnu.org
1675 21. https://gcc.gnu.org/lists.html
1676 22. http://www.fsf.org/
1677 23. https://gcc.gnu.org/about.html
1678 24. http://validator.w3.org/check/referer
1679 ======================================================================
1680 http://gcc.gnu.org/gcc-4.7/changes.html
1681 GCC 4.7 Release Series
1682 Changes, New Features, and Fixes
1686 * The -fconserve-space flag has been deprecated. The flag had no
1687 effect for most targets: only targets without a global .bss section
1688 and without support for switchable sections. Furthermore, the flag
1689 only had an effect for G++, where it could result in wrong
1690 semantics (please refer to the GCC manual for further details). The
1691 flag will be removed in GCC 4.8
1692 * Support for a number of older systems and recently unmaintained or
1693 untested target ports of GCC has been declared obsolete in GCC 4.7.
1694 Unless there is activity to revive them, the next release of GCC
1695 will have their sources permanently removed.
1696 All GCC ports for the following processor architectures have been
1698 + picoChip (picochip-*)
1699 The following ports for individual systems on particular
1700 architectures have been obsoleted:
1701 + IRIX 6.5 (mips-sgi-irix6.5)
1702 + MIPS OpenBSD (mips*-*-openbsd*)
1703 + Solaris 8 (*-*-solaris2.8). Details can be found in the
1705 + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*)
1706 * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A,
1707 ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by
1708 default, which for some sources generates code that accesses memory
1709 on unaligned addresses. This requires the kernel of those systems
1710 to enable such accesses (controlled by CP15 register c1, refer to
1711 ARM documentation). Alternatively, or for compatibility with
1712 kernels where unaligned accesses are not supported, all code has to
1713 be compiled with -mno-unaligned-access. Upstream Linux kernel
1714 releases have automatically and unconditionally supported unaligned
1715 accesses as emitted by GCC due to this option being active since
1717 * Support on ARM for the legacy floating-point accelerator (FPA) and
1718 the mixed-endian floating-point format that it used has been
1719 obsoleted. The ports that still use this format have been obsoleted
1720 as well. Many legacy ARM ports already provide an alternative that
1721 uses the VFP floating-point format. The obsolete ports will be
1722 deleted in the next release.
1723 The obsolete ports with alternatives are:
1724 + arm*-*-rtems (use arm*-*-rtemseabi)
1725 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
1726 + arm*-*-elf (use arm*-*-eabi)
1727 + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
1728 Note, however, that these alternatives are not binary compatible
1729 with their legacy counterparts (although some can support running
1730 legacy applications).
1731 The obsolete ports that currently lack a modern alternative are:
1735 New ports that support more recent versions of the architecture are
1737 * Support for the Maverick co-processor on ARM has been obsoleted.
1738 Code to support it will be deleted in the next release.
1739 * Support has been removed for Unix International threads on Solaris
1740 2, so the --enable-threads=solaris configure option and the
1741 -threads compiler option don't work any longer.
1742 * Support has been removed for the Solaris BSD Compatibility Package,
1743 which lives in /usr/ucbinclude and /usr/ucblib. It has been removed
1744 from Solaris 11, and was only intended as a migration aid from
1745 SunOS 4 to SunOS 5. The -compat-bsd compiler option is not
1746 recognized any longer.
1747 * The AVR port's libgcc has been improved and its multilib structure
1748 has been enhanced. As a result, all objects contributing to an
1749 application must either be compiled with GCC versions up to 4.6.x
1750 or with GCC versions 4.7.1 or later. If the compiler is used with
1751 AVR Libc, you need a version that supports the new layout, i.e.
1752 implements [2]#35407.
1753 * The AVR port's -mshort-calls command-line option has been
1754 deprecated. It will be removed in the GCC 4.8 release. See -mrelax
1756 * The AVR port only references startup code that clears .bss and the
1757 common section resp. initializes the .data and .rodata section
1758 provided respective sections (or subsections thereof) are not
1759 empty, see [3]PR18145. Applications that put all static storage
1760 objects into non-standard sections and / or define all static
1761 storage objects in assembler modules, must reference __do_clear_bss
1762 resp. __do_copy_data by hand or undefine the symbol(s) by means of
1763 -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data.
1764 * The ARM port's -mwords-little-endian option has been deprecated. It
1765 will be removed in a future release.
1766 * Support has been removed for the NetWare x86 configuration
1767 obsoleted in GCC 4.6.
1768 * It is no longer possible to use the "l" constraint in MIPS16 asm
1770 * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard
1771 library which affected the ABI in C++11 mode: a data member was
1772 added to std::list changing its size and altering the definitions
1773 of some member functions, and std::pair's move constructor was
1774 non-trivial which altered the calling convention for functions with
1775 std::pair arguments or return types. The ABI incompatibilities have
1776 been fixed for GCC version 4.7.2 but as a result C++11 code
1777 compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11
1778 code compiled with different GCC versions and with C++98/C++03 code
1779 compiled with any version.
1780 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
1781 rules for the layout of vectors that could lead to wrong code being
1782 generated. Vectors larger than 8 bytes in size are now by default
1783 aligned to an 8-byte boundary. This is an ABI change: code that
1784 makes explicit use of vector types may be incompatible with binary
1785 objects built with older versions of GCC. Auto-vectorized code is
1786 not affected by this change. (This change affects GCC versions
1788 * More information on porting to GCC 4.7 from previous versions of
1789 GCC can be found in the [4]porting guide for this release.
1791 General Optimizer Improvements
1793 * Support for a new parameter --param case-values-threshold=n was
1794 added to allow users to control the cutoff between doing switch
1795 statements as a series of if statements and using a jump table.
1796 * Link-time optimization (LTO) improvements:
1797 + Improved scalability and reduced memory usage. Link time
1798 optimization of Firefox now requires 3GB of RAM on a 64-bit
1799 system, while over 8GB was needed previously. Linking time has
1800 been improved, too. The serial stage of linking Firefox has
1801 been sped up by about a factor of 10.
1802 + Reduced size of object files and temporary storage used during
1804 + Streaming performance (both outbound and inbound) has been
1806 + ld -r is now supported with LTO.
1807 + Several bug fixes, especially in symbol table handling and
1809 * Interprocedural optimization improvements:
1810 + Heuristics now take into account that after inlining code will
1811 be optimized out because of known values (or properties) of
1812 function parameters. For example:
1823 The call of foo will be inlined into bar even when optimizing
1824 for code size. Constructs based on __builtin_constant_p are
1825 now understood by the inliner and code size estimates are
1826 evaluated a lot more realistically.
1827 + The representation of C++ virtual thunks and aliases (both
1828 implicit and defined via the alias attribute) has been
1829 re-engineered. Aliases no longer pose optimization barriers
1830 and calls to an alias can be inlined and otherwise optimized.
1831 + The inter-procedural constant propagation pass has been
1832 rewritten. It now performs generic function specialization.
1833 For example when compiling the following:
1837 ... do something ...
1839 ... do something else ...
1851 GCC will now produce two copies of foo. One with flag being
1852 true, while other with flag being false. This leads to
1853 performance improvements previously possible only by inlining
1854 all calls. Cloning causes a lot less code size growth.
1855 * A string length optimization pass has been added. It attempts to
1856 track string lengths and optimize various standard C string
1857 functions like strlen, strchr, strcpy, strcat, stpcpy and their
1858 _FORTIFY_SOURCE counterparts into faster alternatives. This pass is
1859 enabled by default at -O2 or above, unless optimizing for size, and
1860 can be disabled by the -fno-optimize-strlen option. The pass can
1862 char *bar (const char *a)
1864 size_t l = strlen (a) + 2;
1865 char *p = malloc (l); if (p == NULL) return p;
1866 strcpy (p, a); strcat (p, "/"); return p;
1870 char *bar (const char *a)
1872 size_t tmp = strlen (a);
1873 char *p = malloc (tmp + 2); if (p == NULL) return p;
1874 memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p;
1877 or for hosted compilations where stpcpy is available in the runtime
1878 and headers provide its prototype, e.g.
1879 void foo (char *a, const char *b, const char *c, const char *d)
1881 strcpy (a, b); strcat (a, c); strcat (a, d);
1884 can be optimized into:
1885 void foo (char *a, const char *b, const char *c, const char *d)
1887 strcpy (stpcpy (stpcpy (a, b), c), d);
1890 New Languages and Language specific improvements
1892 * Version 3.1 of the [5]OpenMP specification is now supported for the
1893 C, C++, and Fortran compilers.
1897 * The command-line option -feliminate-unused-debug-types has been
1898 re-enabled by default, as it is for the other languages, leading to
1899 a reduction in debug info size of 12.5% and more for relevant
1900 cases, as well as to a small compilation speedup.
1904 * A new built-in, __builtin_assume_aligned, has been added, through
1905 which the compiler can be hinted about pointer alignment and can
1906 use it to improve generated code.
1907 * A new warning option -Wunused-local-typedefs was added for C, C++,
1908 Objective-C and Objective-C++. This warning diagnoses typedefs
1909 locally defined in a function, and otherwise not used.
1910 * A new experimental command-line option -ftrack-macro-expansion was
1911 added for C, C++, Objective-C, Objective-C++ and Fortran. It allows
1912 the compiler to emit diagnostic about the current macro expansion
1913 stack when a compilation error occurs in a macro expansion.
1914 * Experimental support for transactional memory has been added. It
1915 includes support in the compiler, as well as a supporting runtime
1916 library called libitm. To compile code with transactional memory
1917 constructs, use the -fgnu-tm option.
1918 Support is currently available for Alpha, ARM, PowerPC, SH, SPARC,
1919 and 32-bit/64-bit x86 platforms.
1920 For more details on transactional memory see [6]the GCC WiKi.
1921 * Support for atomic operations specifying the C++11/C11 memory model
1922 has been added. These new __atomic routines replace the existing
1923 __sync built-in routines.
1924 Atomic support is also available for memory blocks. Lock-free
1925 instructions will be used if a memory block is the same size and
1926 alignment as a supported integer type. Atomic operations which do
1927 not have lock-free support are left as function calls. A set of
1928 library functions is available on the GCC atomic wiki in the
1929 "External Atomics Library" section.
1930 For more details on the memory models and features, see the
1932 * When a binary operation is performed on vector types and one of the
1933 operands is a uniform vector, it is possible to replace the vector
1934 with the generating element. For example:
1935 typedef int v4si __attribute__ ((vector_size (16)));
1936 v4si res, a = {1,2,3,4};
1939 res = 2 + a; /* means {2,2,2,2} + a */
1940 res = a - x; /* means a - {x,x,x,x} */
1944 * There is support for some more features from the C11 revision of
1945 the ISO C standard. GCC now accepts the options -std=c11 and
1946 -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x.
1947 + Unicode strings (previously supported only with options such
1948 as -std=gnu11, now supported with -std=c11), and the
1949 predefined macros __STDC_UTF_16__ and __STDC_UTF_32__.
1950 + Nonreturning functions (_Noreturn and <stdnoreturn.h>).
1951 + Alignment support (_Alignas, _Alignof, max_align_t,
1953 + A built-in function __builtin_complex is provided to support C
1954 library implementation of the CMPLX family of macros.
1958 * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat
1959 options, which are equivalent to -std=c++0x, -std=gnu++0x, and
1960 -Wc++0x-compat, respectively.
1961 * G++ now implements [8]C++11 extended friend syntax:
1966 static const int I = 2;
1976 * Thanks to Ville Voutilainen, G++ now implements [9]C++11 explicit
1980 virtual void f() const final;
1981 virtual void f(int);
1985 void f() const; // error: D::f attempts to override final B::f
1986 void f(long) override; // error: doesn't override anything
1987 void f(int) override; // ok
1991 struct F: E { }; // error: deriving from final class
1993 * G++ now implements [10]C++11 non-static data member initializers.
1997 } a; // initializes a.i to 42
1999 * Thanks to Ed Smith-Rowland, G++ now implements [11]C++11
2000 user-defined literals.
2002 // Not actually a good approximation. :)
2003 constexpr long double operator"" _degrees (long double d) { return d * 0.0175; }
2004 long double pi = 180.0_degrees;
2006 * G++ now implements [12]C++11 alias-declarations.
2008 template <class T> using Ptr = T*;
2009 Ptr<int> ip; // decltype(ip) is int*
2011 * Thanks to Ville Voutilainen and Pedro Lamarão, G++ now implements
2012 [13]C++11 delegating constructors.
2016 A(): A(42) { } // delegate to the A(int) constructor
2019 * G++ now fully implements C++11 atomic classes rather than just
2020 integer derived classes.
2026 std::atomic<POD> my_atomic_POD;
2028 * G++ now sets the predefined macro __cplusplus to the correct value,
2029 199711L for C++98/03, and 201103L for C++11.
2030 * G++ now correctly implements the two-phase lookup rules such that
2031 an unqualified name used in a template must have an appropriate
2032 declaration found either in scope at the point of definition of the
2033 template or by argument-dependent lookup at the point of
2034 instantiation. As a result, code that relies on a second
2035 unqualified lookup at the point of instantiation to find functions
2036 declared after the template or in dependent bases will be rejected.
2037 The compiler will suggest ways to fix affected code, and using the
2038 -fpermissive compiler flag will allow the code to compile with a
2042 void f() { g(T()); } // error, g(int) not found by argument-dependent lookup
2043 void g(int) { } // fix by moving this declaration before the declaration of f
2047 // error, B::g(B) not found by argument-dependent lookup
2048 void f() { g(T()); } // fix by using this->g or A::g
2051 struct B { void g(B); };
2059 * G++ now properly re-uses stack space allocated for temporary
2060 objects when their lifetime ends, which can significantly lower
2061 stack consumption for some C++ functions. As a result of this, some
2062 code with undefined behavior will now break:
2064 const int &f(const int &i) { return i; }
2066 const int &x = f(1);
2067 const int &y = f(2);
2069 Here, x refers to the temporary allocated to hold the 1 argument,
2070 which only lives until the end of the initialization; it
2071 immediately becomes a dangling reference. So the next statement
2072 re-uses the stack slot to hold the 2 argument, and users of x get
2074 Note that this should not cause any change of behavior for
2075 temporaries of types with non-trivial destructors, as they are
2076 already destroyed at end of full-expression; the change is that now
2077 the storage is released as well.
2078 * A new command-line option -Wdelete-non-virtual-dtor has been added
2079 to warn when delete is used to destroy an instance of a class which
2080 has virtual functions and non-virtual destructor. It is unsafe to
2081 delete an instance of a derived class through a pointer to a base
2082 class if the base class does not have a virtual destructor. This
2083 warning is enabled by -Wall.
2084 * A new command-line option -Wzero-as-null-pointer-constant has been
2085 added to warn when a literal '0' is used as null pointer constant.
2086 It can be useful to facilitate the conversion to nullptr in C++11.
2087 * As per C++98, access-declarations are now deprecated by G++.
2088 Using-declarations are to be used instead. Furthermore, some
2089 efforts have been made to improve the support of class scope
2090 using-declarations. In particular, using-declarations referring to
2091 a dependent type now work as expected ([14]bug c++/14258).
2092 * The ELF symbol visibility of a template instantiation is now
2093 properly constrained by the visibility of its template arguments
2094 ([15]bug c++/35688).
2096 Runtime Library (libstdc++)
2098 * [16]Improved experimental support for the new ISO C++ standard,
2100 + using noexcept in most of the library;
2101 + implementations of pointer_traits, allocator_traits and
2102 scoped_allocator_adaptor;
2103 + uses-allocator construction for tuple;
2104 + vector meets the allocator-aware container requirements;
2105 + replacing monotonic_clock with steady_clock;
2106 + enabling the thread support library on most POSIX targets;
2107 + many small improvements to conform to the FDIS.
2108 * Added --enable-clocale=newlib configure option.
2109 * Debug Mode iterators for unordered associative containers.
2110 * Avoid polluting the global namespace and do not include <unistd.h>.
2114 * The compile flag [17]-fstack-arrays has been added, which causes
2115 all local arrays to be put on stack memory. For some programs this
2116 will improve the performance significantly. If your program uses
2117 very large local arrays, it is possible that you will have to
2118 extend your runtime limits for stack memory.
2119 * The [18]-Ofast flag now also implies [19]-fno-protect-parens and
2121 * Front-end optimizations can now be selected by the
2122 [21]-ffrontend-optimize option and deselected by the
2123 -fno-frontend-optimize option.
2124 * When front-end optimization removes a function call,
2125 [22]-Wfunction-elimination warns about that.
2126 * When performing front-end-optimization, the
2127 [23]-faggressive-function-elimination option allows the removal of
2128 duplicate function calls even for impure functions.
2129 * The flag [24]-Wreal-q-constant has been added, which warns if
2130 floating-point literals have been specified using q (such as
2131 1.0q0); the q marker is now supported as a vendor extension to
2132 denote quad precision (REAL(16) or, if not available, REAL(10)).
2133 Consider using a kind parameter (such as in 1.0_qp) instead, which
2134 can be obtained via [25]SELECTED_REAL_KIND.
2135 * The GFORTRAN_USE_STDERR environment variable has been removed. GNU
2136 Fortran now always prints error messages to standard error. If you
2137 wish to redirect standard error, please consult the manual for your
2138 OS, shell, batch environment etc. as appropriate.
2139 * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment
2140 variable have been removed. When encountering a serious error,
2141 gfortran will now always abort the program. Whether a core dump is
2142 generated depends on the user environment settings; see the ulimit
2143 -c setting for POSIX shells, limit coredumpsize for C shells, and
2144 the [26]WER user-mode dumps settings on Windows.
2145 * The [27]-fbacktrace option is now enabled by default. When
2146 encountering a fatal error, gfortran will attempt to print a
2147 backtrace to standard error before aborting. It can be disabled
2148 with -fno-backtrace. Note: On POSIX targets with the addr2line
2149 utility from GNU binutils, GNU Fortran can print a backtrace with
2150 function name, file name, line number information in addition to
2151 the addresses; otherwise only the addresses are printed.
2153 + Generic interface names which have the same name as derived
2154 types are now supported, which allows to write constructor
2155 functions. Note that Fortran does not support static
2156 constructor functions; only default initialization or an
2157 explicit structure-constructor initialization are available.
2158 + [29]Polymorphic (class) arrays are now supported.
2160 + Support for the DO CONCURRENT construct has been added, which
2161 allows the user to specify that individual loop iterations
2162 have no interdependencies.
2163 + [31]Coarrays: Full single-image support except for polymorphic
2164 coarrays. Additionally, preliminary support for multiple
2165 images via an MPI-based [32]coarray communication library has
2166 been added. Note: The library version is not yet usable as
2167 remote coarray access is not yet possible.
2169 + New flag [34]-std=f2008ts permits programs that are expected
2170 to conform to the Fortran 2008 standard and the draft
2171 Technical Specification (TS) 29113 on Further Interoperability
2173 + The OPTIONAL attribute is now allowed for dummy arguments of
2175 + The RANK intrinsic has been added.
2176 + The implementation of the ASYNCHRONOUS attribute in GCC is
2177 compatible with the candidate draft of TS 29113 (since GCC
2182 * GCC 4.7 implements the [35]Go 1 language standard. The library
2183 support in 4.7.0 is not quite complete, due to release timing.
2184 Release 4.7.1 includes complete support for Go 1. The Go library is
2185 from the Go 1.0.1 release.
2186 * Go has been tested on GNU/Linux and Solaris platforms. It may work
2187 on other platforms as well.
2189 New Targets and Target Specific Improvements
2193 * GCC now supports the Cortex-A7 processor implementing the v7-a
2194 version of the architecture using the option -mcpu=cortex-a7.
2195 * The default vector size in auto-vectorization for NEON is now 128
2196 bits. If vectorization fails thusly, the vectorizer tries again
2197 with 64-bit vectors.
2198 * A new option -mvectorize-with-neon-double was added to allow users
2199 to change the vector size to 64 bits.
2203 * GCC now supports the XMEGA architecture. This requires GNU binutils
2205 * Support for the [36]named address spaces __flash, __flash1, …,
2206 __flash5 and __memx has been added. These address spaces locate
2207 read-only data in flash memory and allow reading from flash memory
2208 by means of ordinary C code, i.e. without the need of (inline)
2211 const __flash int values[] = { 42, 31 };
2213 int add_values (const __flash int *p, int i)
2215 return values[i] + *p;
2218 * Support has been added for the AVR-specific configure option
2219 --with-avrlibc=yes in order to arrange for better integration of
2220 [37]AVR-Libc. This configure option is supported in avr-gcc 4.7.2
2221 and newer and will only take effect in non-RTEMS configurations. If
2222 avr-gcc is configured for RTEMS, the option will be ignored which
2223 is the same as specifying --with-avrlibc=no. See [38]PR54461 for
2224 more technical details.
2225 * Support for AVR-specific [39]built-in functions has been added.
2226 * Support has been added for the signed and unsigned 24-bit scalar
2227 integer types __int24 and __uint24.
2228 * New command-line options -maccumulate-args, -mbranch-cost=cost and
2229 -mstrict-X were added to allow better fine-tuning of code
2231 * The command option -fdata-sections now also takes affect on the
2232 section names of variables with the progmem attribute.
2233 * A new inline assembler print modifier %i to print a RAM address as
2234 I/O address has been added:
2236 #include <avr/io.h> /* Port Definitions from AVR-LibC */
2238 void set_portb (uint8_t value)
2240 asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory");
2243 The offset between an I/O address and the RAM address for that I/O
2244 location is device-specific. This offset is taken into account when
2245 printing a RAM address with the %i modifier so that the address is
2246 suitable to be used as operand in an I/O command. The address must
2247 be a constant integer known at compile time.
2248 * The inline assembler constraint "R" to represent integers in the
2249 range −6 … 5 has been removed without replacement.
2250 * Many optimizations to:
2251 + 64-bit integer arithmetic
2252 + Widening multiplication
2253 + Integer division by a constant
2254 + Avoid constant reloading in multi-byte instructions.
2255 + Micro-optimizations for special instruction sequences.
2256 + Generic built-in functions like __builtin_ffs*,
2257 __builtin_clz*, etc.
2258 + If-else decision trees generated by switch instructions
2259 + Merging of data located in flash memory
2260 + New libgcc variants for devices with 8-bit wide stack pointer
2262 * Better documentation:
2263 + Handling of EIND and indirect jumps on devices with more than
2264 128 KiB of program memory.
2265 + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function
2267 + Function attributes OS_main and OS_task.
2268 + AVR-specific built-in macros.
2272 * Support has been added for the Texas Instruments C6X family of
2277 * Support has been added for National Semiconductor's CR16
2282 * Support has been added for Adapteva's Epiphany architecture.
2286 * Support for Intel AVX2 intrinsics, built-in functions and code
2287 generation is available via -mavx2.
2288 * Support for Intel BMI2 intrinsics, built-in functions and code
2289 generation is available via -mbmi2.
2290 * Implementation and automatic generation of __builtin_clz* using the
2291 lzcnt instruction is available via -mlzcnt.
2292 * Support for Intel FMA3 intrinsics and code generation is available
2294 * A new -mfsgsbase command-line option is available that makes GCC
2295 generate new segment register read/write instructions through
2296 dedicated built-ins.
2297 * Support for the new Intel rdrnd instruction is available via
2299 * Two additional AVX vector conversion instructions are available via
2301 * Support for new Intel processor codename IvyBridge with RDRND,
2302 FSGSBASE and F16C is available through -march=core-avx-i.
2303 * Support for the new Intel processor codename Haswell with AVX2,
2304 FMA, BMI, BMI2, LZCNT is available through -march=core-avx2.
2305 * Support for new AMD family 15h processors (Piledriver core) is now
2306 available through -march=bdver2 and -mtune=bdver2 options.
2307 * Support for [40]the x32 psABI is now available through the -mx32
2309 * Windows mingw targets are using the -mms-bitfields option by
2311 * Windows x86 targets are using the __thiscall calling convention for
2312 C++ class-member functions.
2313 * Support for the configure option --with-threads=posix for Windows
2318 * GCC now supports thread-local storage (TLS) for MIPS16. This
2319 requires GNU binutils 2.22 or later.
2320 * GCC can now generate code specifically for the Cavium Octeon+ and
2321 Octeon2 processors. The associated command-line options are
2322 -march=octeon+ and -march=octeon2 respectively. Both options
2323 require GNU binutils 2.22 or later.
2324 * GCC can now work around certain 24k errata, under the control of
2325 the command-line option -mfix-24k. These workarounds require GNU
2326 binutils 2.20 or later.
2327 * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build
2328 n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux
2329 toolchain that generates 32-bit code by default. Use the
2330 configure-time option --enable-targets=all to select these extra
2332 * Passing -fno-delayed-branch now also stops the assembler from
2333 automatically filling delay slots.
2337 * Vectors of type vector long long or vector long are passed and
2338 returned using the same method as other vectors with the VSX
2339 instruction set. Previously GCC did not adhere to the ABI for
2340 128-bit vectors with 64-bit integer base types (PR 48857). This
2341 will also be fixed in the GCC 4.6.1 and 4.5.4 releases.
2342 * A new option -mno-pointers-to-nested-functions was added to allow
2343 AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify
2344 that the compiler should not load up the chain register (r11)
2345 before calling a function through a pointer. If you use this
2346 option, you cannot call nested functions through a pointer, or call
2347 other languages that might use the static chain.
2348 * A new option msave-toc-indirect was added to allow AIX
2349 32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we
2350 save the TOC in the prologue for indirect calls or generate the
2351 save inline. This can speed up some programs that call through a
2352 function pointer a lot, but it can slow down other functions that
2353 only call through a function pointer in exceptional cases.
2354 * The PowerPC port will now enable machine-specific built-in
2355 functions when the user switches the target machine using the
2356 #pragma GCC target or __attribute__ ((__target__ ("target"))) code
2357 sequences. In addition, the target macros are updated. However, due
2358 to the way the -save-temps switch is implemented, you won't see the
2359 effect of these additional macros being defined in preprocessor
2364 * A new option -msoft-atomic has been added. When it is specified,
2365 GCC will generate GNU/Linux-compatible gUSA atomic sequences for
2366 the new __atomic routines.
2367 * Since it is neither supported by GAS nor officially documented,
2368 code generation for little endian SH2A has been disabled.
2369 Specifying -ml with -m2a* will now result in a compiler error.
2370 * The defunct -mbranch-cost option has been fixed.
2371 * Some improvements to the generated code of:
2372 + Utilization of the tst #imm,R0 instruction.
2373 + Dynamic shift instructions on SH2A.
2374 + Integer absolute value calculations.
2375 * The -mdiv= option for targets other than SHmedia has been fixed and
2380 * The option -mflat has been reinstated. When it is specified, the
2381 compiler will generate code for a single register window model.
2382 This is essentially a new implementation and the corresponding
2383 debugger support has been added to GDB 7.4.
2384 * Support for the options -mtune=native and -mcpu=native has been
2385 added on selected native platforms (GNU/Linux and Solaris).
2386 * Support for the SPARC T3 (Niagara 3) processor has been added.
2388 + An intrinsics header visintrin.h has been added.
2389 + Builtin intrinsics for the VIS 1.0 edge handling and pixel
2390 compare instructions have been added.
2391 + The little-endian version of alignaddr is now supported.
2392 + When possible, VIS builtins are marked const, which should
2393 increase the compiler's ability to optimize VIS operations.
2394 + The compiler now properly tracks the %gsr register and how it
2395 behaves as an input for various VIS instructions.
2396 + Akin to fzero, the compiler can now generate fone instructions
2397 in order to set all of the bits of a floating-point register
2399 + The documentation for the VIS intrinsics in the GCC manual has
2400 been brought up to date and many inaccuracies were fixed.
2401 + Intrinsics for the VIS 2.0 bmask, bshuffle, and
2402 non-condition-code setting edge instructions have been added.
2403 Their availability is controlled by the new -mvis2 and
2404 -mno-vis2 options. They are enabled by default on
2405 UltraSPARC-III and later CPUs.
2406 * Support for UltraSPARC Fused Multiply-Add floating-point extensions
2407 has been added. These instructions are enabled by default on SPARC
2408 T3 (Niagara 3) and later CPUs.
2412 * Support has been added for the Tilera TILE-Gx and TILEPro families
2415 Other significant improvements
2417 * A new option (-grecord-gcc-switches) was added that appends
2418 compiler command-line options that might affect code generation to
2419 the DW_AT_producer attribute string in the DWARF debugging
2421 * GCC now supports various new GNU extensions to the DWARF debugging
2422 information format, like [41]entry value and [42]call site
2423 information, [43]typed DWARF stack or [44]a more compact macro
2424 representation. Support for these extensions has been added to GDB
2425 7.4. They can be disabled through the -gstrict-dwarf command-line
2430 This is the [45]list of problem reports (PRs) from GCC's bug tracking
2431 system that are known to be fixed in the 4.7.1 release. This list might
2432 not be complete (that is, it is possible that some PRs that have been
2433 fixed are not listed here).
2435 The Go frontend in the 4.7.1 release fully supports the [46]Go 1
2440 This is the [47]list of problem reports (PRs) from GCC's bug tracking
2441 system that are known to be fixed in the 4.7.2 release. This list might
2442 not be complete (that is, it is possible that some PRs that have been
2443 fixed are not listed here).
2447 This is the [48]list of problem reports (PRs) from GCC's bug tracking
2448 system that are known to be fixed in the 4.7.3 release. This list might
2449 not be complete (that is, it is possible that some PRs that have been
2450 fixed are not listed here).
2454 This is the [49]list of problem reports (PRs) from GCC's bug tracking
2455 system that are known to be fixed in the 4.7.4 release. This list might
2456 not be complete (that is, it is possible that some PRs that have been
2457 fixed are not listed here).
2460 For questions related to the use of GCC, please consult these web
2461 pages and the [50]GCC manuals. If that fails, the
2462 [51]gcc-help@gcc.gnu.org mailing list might help. Comments on these
2463 web pages and the development of GCC are welcome on our developer
2464 list at [52]gcc@gcc.gnu.org. All of [53]our lists have public
2467 Copyright (C) [54]Free Software Foundation, Inc. Verbatim copying and
2468 distribution of this entire article is permitted in any medium,
2469 provided this notice is preserved.
2471 These pages are [55]maintained by the GCC team. Last modified
2476 1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html
2477 2. http://savannah.nongnu.org/bugs/?35407
2478 3. https://gcc.gnu.org/PR18145
2479 4. https://gcc.gnu.org/gcc-4.7/porting_to.html
2480 5. http://openmp.org/wp/openmp-specifications/
2481 6. https://gcc.gnu.org/wiki/TransactionalMemory
2482 7. https://gcc.gnu.org/wiki/Atomic/GCCMM
2483 8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
2484 9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
2485 10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
2486 11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
2487 12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
2488 13. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
2489 14. https://gcc.gnu.org/PR14258
2490 15. https://gcc.gnu.org/PR35688
2491 16. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011
2492 17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
2493 18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689
2494 19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270
2495 20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
2496 21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275
2497 22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170
2498 23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270
2499 24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWreal-q-constant_007d-149
2500 25. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html
2501 26. http://msdn.microsoft.com/en-us/library/bb787181(v=vs.85).aspx
2502 27. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183
2503 28. https://gcc.gnu.org/wiki/Fortran2003Status
2504 29. https://gcc.gnu.org/wiki/OOP
2505 30. https://gcc.gnu.org/wiki/Fortran2008Status
2506 31. https://gcc.gnu.org/wiki/Coarray
2507 32. https://gcc.gnu.org/wiki/CoarrayLib
2508 33. https://gcc.gnu.org/wiki/TS29113Status
2509 34. 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
2510 35. http://weekly.golang.org/doc/go1.html
2511 36. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html
2512 37. http://nongnu.org/avr-libc/
2513 38. https://gcc.gnu.org/PR54461
2514 39. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built_002din-Functions.html
2515 40. https://sites.google.com/site/x32abi/
2516 41. http://www.dwarfstd.org/ShowIssue.php?issue=100909.1
2517 42. http://www.dwarfstd.org/ShowIssue.php?issue=100909.2
2518 43. http://www.dwarfstd.org/doc/040408.1.html
2519 44. http://www.dwarfstd.org/ShowIssue.php?issue=110722.1
2520 45. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1
2521 46. http://weekly.golang.org/doc/go1.html
2522 47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2
2523 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3
2524 49. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4
2525 50. https://gcc.gnu.org/onlinedocs/
2526 51. mailto:gcc-help@gcc.gnu.org
2527 52. mailto:gcc@gcc.gnu.org
2528 53. https://gcc.gnu.org/lists.html
2529 54. http://www.fsf.org/
2530 55. https://gcc.gnu.org/about.html
2531 56. http://validator.w3.org/check/referer
2532 ======================================================================
2533 http://gcc.gnu.org/gcc-4.6/index.html
2534 GCC 4.6 Release Series
2538 The [1]GNU project and the GCC developers are pleased to announce the
2539 release of GCC 4.6.4.
2541 This release is a bug-fix release, containing fixes for regressions in
2542 GCC 4.6.3 relative to previous releases of GCC.
2547 April 12, 2013 ([2]changes, [3]documentation)
2550 March 1, 2012 ([4]changes, [5]documentation)
2553 October 26, 2011 ([6]changes, [7]documentation)
2556 June 27, 2011 ([8]changes, [9]documentation)
2559 March 25, 2011 ([10]changes, [11]documentation)
2561 References and Acknowledgements
2563 GCC used to stand for the GNU C Compiler, but since the compiler
2564 supports several other languages aside from C, it now stands for the
2565 GNU Compiler Collection.
2567 A list of [12]successful builds is updated as new information becomes
2570 The GCC developers would like to thank the numerous people that have
2571 contributed new features, improvements, bug fixes, and other changes as
2572 well as test results to GCC. This [13]amazing group of volunteers is
2573 what makes GCC successful.
2575 For additional information about GCC please refer to the [14]GCC
2576 project web site or contact the [15]GCC development mailing list.
2578 To obtain GCC please use [16]our mirror sites or [17]our SVN server.
2581 For questions related to the use of GCC, please consult these web
2582 pages and the [18]GCC manuals. If that fails, the
2583 [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
2584 web pages and the development of GCC are welcome on our developer
2585 list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
2588 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
2589 distribution of this entire article is permitted in any medium,
2590 provided this notice is preserved.
2592 These pages are [23]maintained by the GCC team. Last modified
2597 1. http://www.gnu.org/
2598 2. http://gcc.gnu.org/gcc-4.6/changes.html
2599 3. https://gcc.gnu.org/onlinedocs/4.6.4/
2600 4. http://gcc.gnu.org/gcc-4.6/changes.html
2601 5. https://gcc.gnu.org/onlinedocs/4.6.3/
2602 6. http://gcc.gnu.org/gcc-4.6/changes.html
2603 7. https://gcc.gnu.org/onlinedocs/4.6.2/
2604 8. http://gcc.gnu.org/gcc-4.6/changes.html
2605 9. https://gcc.gnu.org/onlinedocs/4.6.1/
2606 10. http://gcc.gnu.org/gcc-4.6/changes.html
2607 11. https://gcc.gnu.org/onlinedocs/4.6.0/
2608 12. http://gcc.gnu.org/gcc-4.6/buildstat.html
2609 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
2610 14. http://gcc.gnu.org/index.html
2611 15. mailto:gcc@gcc.gnu.org
2612 16. http://gcc.gnu.org/mirrors.html
2613 17. http://gcc.gnu.org/svn.html
2614 18. https://gcc.gnu.org/onlinedocs/
2615 19. mailto:gcc-help@gcc.gnu.org
2616 20. mailto:gcc@gcc.gnu.org
2617 21. https://gcc.gnu.org/lists.html
2618 22. http://www.fsf.org/
2619 23. https://gcc.gnu.org/about.html
2620 24. http://validator.w3.org/check/referer
2621 ======================================================================
2622 http://gcc.gnu.org/gcc-4.6/changes.html
2623 GCC 4.6 Release Series
2624 Changes, New Features, and Fixes
2628 * The options -b <machine> and -V <version> have been removed because
2629 they were unreliable. Instead, users should directly run
2630 <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to
2631 run a different version of gcc.
2632 * GCC now has stricter checks for invalid command-line options. In
2633 particular, when gcc was called to link object files rather than
2634 compile source code, it would previously accept and ignore all
2635 options starting with --, including linker options such as
2636 --as-needed and --export-dynamic, although such options would
2637 result in errors if any source code was compiled. Such options, if
2638 unknown to the compiler, are now rejected in all cases; if the
2639 intent was to pass them to the linker, options such as
2640 -Wl,--as-needed should be used.
2641 * Versions of the GNU C library up to and including 2.11.1 included
2642 an [1]incorrect implementation of the cproj function. GCC optimizes
2643 its builtin cproj according to the behavior specified and allowed
2644 by the ISO C99 standard. If you want to avoid discrepancies between
2645 the C library and GCC's builtin transformations when using cproj in
2646 your code, use GLIBC 2.12 or later. If you are using an older GLIBC
2647 and actually rely on the incorrect behavior of cproj, then you can
2648 disable GCC's transformations using -fno-builtin-cproj.
2649 * The C-only intermodule optimization framework (IMA, enabled by
2650 -combine) has been removed in favor of the new generic link-time
2651 optimization framework (LTO) introduced in [2]GCC 4.5.0.
2652 * GCC now ships with the LGPL-licensed libquadmath library, which
2653 provides quad-precision mathematical functions for targets with a
2654 __float128 datatype. __float128 is available for targets on 32-bit
2655 x86, x86-64 and Itanium architectures. The libquadmath library is
2656 automatically built on such targets when building the Fortran
2658 * New -Wunused-but-set-variable and -Wunused-but-set-parameter
2659 warnings were added for C, C++, Objective-C and Objective-C++.
2660 These warnings diagnose variables respective parameters which are
2661 only set in the code and never otherwise used. Usually such
2662 variables are useless and often even the value assigned to them is
2663 computed needlessly, sometimes expensively. The
2664 -Wunused-but-set-variable warning is enabled by default by -Wall
2665 flag and -Wunused-but-set-parameter by -Wall -Wextra flags.
2666 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
2667 rules for the layout of vectors that could lead to wrong code being
2668 generated. Vectors larger than 8 bytes in size are now by default
2669 aligned to an 8-byte boundary. This is an ABI change: code that
2670 makes explicit use of vector types may be incompatible with binary
2671 objects built with older versions of GCC. Auto-vectorized code is
2672 not affected by this change. (This change affects GCC versions
2673 4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.)
2674 * On AVR, variables with the progmem attribute to locate data in
2675 flash memory must be qualified as const.
2676 * Support for a number of older systems and recently unmaintained or
2677 untested target ports of GCC has been declared obsolete in GCC 4.6.
2678 Unless there is activity to revive them, the next release of GCC
2679 will have their sources permanently removed.
2680 All GCC ports for the following processor architectures have been
2682 + Argonaut ARC (arc-*)
2683 + National Semiconductor CRX (crx-*)
2684 + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*,
2685 m68hc12-*-*, m6812-*-*)
2686 + Sunplus S+core (score-*)
2687 The following ports for individual systems on particular
2688 architectures have been obsoleted:
2689 + Interix (i[34567]86-*-interix3*)
2690 + NetWare x86 (i[3456x]86-*-netware*)
2691 + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*)
2692 + MCore PE (mcore-*-pe*)
2693 + SH SymbianOS (sh*-*-symbianelf*)
2694 + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*)
2695 + M68K uClinux old ABI (m68k-*-uclinuxoldabi*)
2696 + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*,
2697 vax-*-netbsd*, but not *-*-netbsdelf*)
2698 The i[34567]86-*-pe alias for Cygwin targets has also been
2699 obsoleted; users should configure for i[34567]86-*-cygwin* instead.
2700 Certain configure options to control the set of libraries built
2701 with GCC on some targets have been obsoleted. On ARM targets, the
2702 options --disable-fpu, --disable-26bit, --disable-underscore,
2703 --disable-interwork, --disable-biendian and --disable-nofmult have
2704 been obsoleted. On MIPS targets, the options
2705 --disable-single-float, --disable-biendian and --disable-softfloat
2706 have been obsoleted.
2707 * Support has been removed for all the [3]configurations obsoleted in
2709 * More information on porting to GCC 4.6 from previous versions of
2710 GCC can be found in the [4]porting guide for this release.
2712 General Optimizer Improvements
2714 * A new general optimization level, -Ofast, has been introduced. It
2715 combines the existing optimization level -O3 with options that can
2716 affect standards compliance but result in better optimized code.
2717 For example, -Ofast enables -ffast-math.
2718 * Link-time optimization improvements:
2719 + The [5]Scalable Whole Program Optimizer (WHOPR) project has
2720 stabilized to the point of being usable. It has become the
2721 default mode when using the LTO optimization model. Link time
2722 optimization can now split itself into multiple parallel
2723 compilations. Parallelism is controlled with -flto=n (where n
2724 specifies the number of compilations to execute in parallel).
2725 GCC can also cooperate with a GNU make job server by
2726 specifying the -flto=jobserver option and adding + to the
2727 beginning of the Makefile rule executing the linker.
2728 Classical LTO mode can be enforced by -flto-partition=none.
2729 This may result in small code quality improvements.
2730 + A large number of bugs were fixed. GCC itself, Mozilla Firefox
2731 and other large applications can be built with LTO enabled.
2732 + The linker plugin support improvements
2733 o Linker plugin is now enabled by default when the linker
2734 is detected to have plugin support. This is the case for
2735 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and
2736 the Gold linker on ELF targets. Plugin support of the
2737 Apple linker on Darwin is not compatible with GCC. The
2738 linker plugin can also be controlled by the
2739 -fuse-linker-plugin command line option.
2740 o Resolution information from the linker plugin is used to
2741 drive whole program assumptions. Use of the linker plugin
2742 results in more aggressive optimization on binaries and
2743 on shared libraries that use the hidden visibility
2744 attribute. Consequently the use of -fwhole-program is not
2745 necessary in addition to LTO.
2746 + Hidden symbols used from non-LTO objects now have to be
2747 explicitly annotated with externally_visible when the linker
2749 + C++ inline functions and virtual tables are now privatized
2750 more aggressively, leading to better inter-procedural
2751 optimization and faster dynamic linking.
2752 + Memory usage and intermediate language streaming performance
2754 + Static constructors and destructors from individual units are
2755 inlined into a single function. This can significantly improve
2756 startup times of large C++ applications where static
2757 constructors are very common. For example, static constructors
2758 are used when including the iostream header.
2759 + Support for the Ada language has been added.
2760 * Interprocedural optimization improvements
2761 + The interprocedural framework was re-tuned for link time
2762 optimization. Several scalability issues were resolved.
2763 + Improved auto-detection of const and pure functions. Newly,
2764 noreturn functions are auto-detected.
2765 The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is
2766 available that informs users when adding attributes to headers
2767 might improve code generation.
2768 + A number of inlining heuristic improvements. In particular:
2769 o Partial inlining is now supported and enabled by default
2770 at -O2 and greater. The feature can be controlled via
2772 Partial inlining splits functions with short hot path to
2773 return. This allows more aggressive inlining of the hot
2774 path leading to better performance and often to code size
2775 reductions (because cold parts of functions are not
2777 o Scalability for large compilation units was improved
2779 o Inlining of callbacks is now more aggressive.
2780 o Virtual methods are considered for inlining when the
2781 caller is inlined and devirtualization is then possible.
2782 o Inlining when optimizing for size (either in cold regions
2783 of a program or when compiling with -Os) was improved to
2784 better handle C++ programs with larger abstraction
2785 penalty, leading to smaller and faster code.
2786 + The IPA reference optimization pass detecting global variables
2787 used or modified by functions was strengthened and sped up.
2788 + Functions whose address was taken are now optimized out when
2789 all references to them are dead.
2790 + A new inter-procedural static profile estimation pass detects
2791 functions that are executed once or unlikely to be executed.
2792 Unlikely executed functions are optimized for size. Functions
2793 executed once are optimized for size except for the inner
2795 + On most targets with named section support, functions used
2796 only at startup (static constructors and main), functions used
2797 only at exit and functions detected to be cold are placed into
2798 separate text segment subsections. This extends the
2799 -freorder-functions feature and is controlled by the same
2800 switch. The goal is to improve the startup time of large C++
2802 Proper function placement requires linker support. GNU ld
2803 2.21.51 on ELF targets was updated to place those functions
2804 together within the text section leading to better code
2805 locality and faster startup times of large C++ programs. The
2806 feature is also supported in the Apple linker. Support in the
2807 gold linker is planned.
2808 * A new switch -fstack-usage has been added. It makes the compiler
2809 output stack usage information for the program, on a per-function
2810 basis, in an auxiliary file.
2811 * A new switch -fcombine-stack-adjustments has been added. It can be
2812 used to enable or disable the compiler's stack-slot combining pass
2813 which before was enabled automatically at -O1 and above, but could
2814 not be controlled on its own.
2815 * A new switch -fstrict-volatile-bitfields has been added. Using it
2816 indicates that accesses to volatile bitfields should use a single
2817 access of the width of the field's type. This option can be useful
2818 for precisely defining and accessing memory-mapped peripheral
2819 registers from C or C++.
2821 Compile time and memory usage improvements
2823 * Datastructures used by the dataflow framework in GCC were
2824 reorganized for better memory usage and more cache locality.
2825 Compile time is improved especially on units with large functions
2826 (possibly resulting from a lot of inlining) not fitting into the
2827 processor cache. The compile time of the GCC C compiler binary with
2828 link-time optimization went down by over 10% (benchmarked on x86-64
2831 New Languages and Language specific improvements
2835 * Stack checking has been improved on selected architectures (Alpha,
2836 IA-32/x86-64, RS/6000 and SPARC): it now will detect stack
2837 overflows in all cases on these architectures.
2838 * Initial support for Ada 2012 has been added.
2842 * A new warning, enabled by -Wdouble-promotion, has been added that
2843 warns about cases where a value of type float is implicitly
2844 promoted to double. This is especially helpful for CPUs that handle
2845 the former in hardware, but emulate the latter in software.
2846 * A new function attribute leaf was introduced. This attribute allows
2847 better inter-procedural optimization across calls to functions that
2848 return to the current unit only via returning or exception
2849 handling. This is the case for most library functions that have no
2851 * Support for a new data type __int128 for targets having wide enough
2852 machine-mode support.
2853 * The new function attribute callee_pop_aggregate allows to specify
2854 if the caller or callee is responsible for popping the aggregate
2855 return pointer value from the stack.
2856 * Support for selectively enabling and disabling warnings via #pragma
2857 GCC diagnostic has been added. For instance:
2858 #pragma GCC diagnostic error "-Wuninitialized"
2859 foo(a); /* error is given for this one */
2860 #pragma GCC diagnostic push
2861 #pragma GCC diagnostic ignored "-Wuninitialized"
2862 foo(b); /* no diagnostic for this one */
2863 #pragma GCC diagnostic pop
2864 foo(c); /* error is given for this one */
2865 #pragma GCC diagnostic pop
2866 foo(d); /* depends on command line options */
2868 * The -fmax-errors=N option is now supported. Using this option
2869 causes the compiler to exit after N errors have been issued.
2873 * There is now experimental support for some features from the
2874 upcoming C1X revision of the ISO C standard. This support may be
2875 selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions.
2876 Note that this support is experimental and may change incompatibly
2877 in future releases for consistency with changes to the C1X standard
2878 draft. The following features are newly supported as described in
2879 the N1539 draft of C1X (with changes agreed at the March 2011 WG14
2880 meeting); some other features were already supported with no
2881 compiler changes being needed, or have some support but not in full
2882 accord with N1539 (as amended).
2883 + Static assertions (_Static_assert keyword)
2884 + Typedef redefinition
2885 + New macros in <float.h>
2886 + Anonymous structures and unions
2887 * The new -fplan9-extensions option directs the compiler to support
2888 some extensions for anonymous struct fields which are implemented
2889 by the Plan 9 compiler. A pointer to a struct may be automatically
2890 converted to a pointer to an anonymous field when calling a
2891 function, in order to make the types match. An anonymous struct
2892 field whose type is a typedef name may be referred to using the
2897 * Improved [7]experimental support for the upcoming C++0x ISO C++
2898 standard, including support for constexpr (thanks to Gabriel Dos
2899 Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide),
2900 noexcept, unrestricted unions, range-based for loops (thanks to
2901 Rodrigo Rivas Costa), opaque enum declarations (thanks also to
2902 Rodrigo), implicitly deleted functions and implicit move
2904 * When an extern declaration within a function does not match a
2905 declaration in the enclosing context, G++ now properly declares the
2906 name within the namespace of the function rather than the namespace
2907 which was open just before the function definition ([8]c++/43145).
2908 * GCC now warns by default when casting integers to larger pointer
2909 types. These warnings can be disabled with the option
2910 -Wno-int-to-pointer-cast, which is now also available in C++.
2911 * G++ no longer optimizes using the assumption that a value of
2912 enumeration type will fall within the range specified by the
2913 standard, since that assumption is easily violated with a
2914 conversion from integer type ([9]c++/43680). The old behavior can
2915 be restored with -fstrict-enums.
2916 * The new -fnothrow-opt flag changes the semantics of a throw()
2917 exception specification to match the proposed semantics of the
2918 noexcept specification: just call terminate if an exception tries
2919 to propagate out of a function with such an exception
2920 specification. This dramatically reduces or eliminates the code
2921 size overhead from adding the exception specification.
2922 * The new -Wnoexcept flag will suggest adding a noexcept qualifier to
2923 a function that the compiler can tell doesn't throw if it would
2924 change the value of a noexcept expression.
2925 * The -Wshadow option now warns if a local variable or type
2926 declaration shadows another type in C++. Note that the compiler
2927 will not warn if a local variable shadows a struct/class/enum, but
2928 will warn if it shadows an explicit typedef.
2929 * When an identifier is not found in the current scope, G++ now
2930 offers suggestions about which identifier might have been intended.
2931 * G++ now issues clearer diagnostics for missing semicolons after
2932 class, struct, and union definitions.
2933 * G++ now issues clearer diagnostics for missing semicolons after
2934 class member declarations.
2935 * G++ now issues clearer diagnostics when a colon is used in a place
2936 where a double-colon was intended.
2937 * G++ no longer accepts mutable on reference members ([10]c++/33558).
2938 Use -fpermissive to allow the old, non-conforming behaviour.
2939 * A few mangling fixes have been made, to attribute const/volatile on
2940 function pointer types, decltype of a plain decl, and use of a
2941 function parameter in the declaration of another parameter. By
2942 default the compiler still uses the old mangling, but emits aliases
2943 with the new mangling on targets that support strong aliases. Users
2944 can switch over entirely to the new mangling with -fabi-version=5
2945 or -fabi-version=0. -Wabi will now warn about code that uses the
2947 * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified
2948 type to be default initialized unless the type has a user-declared
2949 default constructor. In 4.6.2 G++ implements the proposed
2950 resolution of [11]DR 253, so default initialization is allowed if
2951 it initializes all subobjects. Code that fails to compile can be
2952 fixed by providing an initializer e.g.
2954 struct B : A { int i; };
2956 Use -fpermissive to allow the old, non-conforming behaviour.
2958 Runtime Library (libstdc++)
2960 * [12]Improved experimental support for the upcoming ISO C++
2961 standard, C++0x, including using constexpr and nullptr.
2962 * Performance improvements to the [13]Debug Mode, thanks to François
2964 * Atomic operations used for reference-counting are annotated so that
2965 they can be understood by race detectors such as Helgrind, see
2966 [14]Data Race Hunting.
2967 * Most libstdc++ standard headers have been changed to no longer
2968 include the cstddef header as an implementation detail. Code that
2969 relied on that header being included as side-effect of including
2970 other standard headers will need to include cstddef explicitly.
2974 * On systems supporting the libquadmath library, GNU Fortran now also
2975 supports a quad-precision, kind=16 floating-point data type
2976 (REAL(16), COMPLEX(16)). As the data type is not fully supported in
2977 hardware, calculations might be one to two orders of magnitude
2978 slower than with the 4, 8 or 10 bytes floating-point data types.
2979 This change does not affect systems which support REAL(16) in
2980 hardware nor those which do not support libquadmath.
2981 * Much improved compile time for large array constructors.
2982 * In order to reduce execution time and memory consumption, use of
2983 temporary arrays in assignment expressions is avoided for many
2984 cases. The compiler now reverses loops in order to avoid generating
2985 a temporary array where possible.
2986 * Improved diagnostics, especially with -fwhole-file.
2987 * The -fwhole-file flag is now enabled by default. This improves code
2988 generation and diagnostics. It can be disabled using the deprecated
2989 -fno-whole-file flag.
2990 * Support the generation of Makefile dependencies via the [15]-M...
2991 flags of GCC; you may need to specify the -cpp option in addition.
2992 The dependencies take modules, Fortran's include, and CPP's
2993 #include into account. Note: Using -M for the module path is no
2994 longer supported, use -J instead.
2995 * The flag -Wconversion has been modified to only issue warnings
2996 where a conversion leads to information loss. This drastically
2997 reduces the number of warnings; -Wconversion is thus now enabled
2998 with -Wall. The flag -Wconversion-extra has been added and also
2999 warns about other conversions; -Wconversion-extra typically issues
3000 a huge number of warnings, most of which can be ignored.
3001 * A new command-line option -Wunused-dummy-argument warns about
3002 unused dummy arguments and is included in -Wall. Before,
3003 -Wunused-variable also warned about unused dummy arguments.
3004 * Fortran 2003 support has been extended:
3005 + Improved support for polymorphism between libraries and
3006 programs and for complicated inheritance patterns (cf.
3007 [16]object-oriented programming).
3008 + Experimental support of the ASSOCIATE construct.
3009 + In pointer assignments it is now possible to specify the lower
3010 bounds of the pointer and, for a rank-1 or a simply contiguous
3011 data-target, to remap the bounds.
3012 + Automatic (re)allocation: In intrinsic assignments to
3013 allocatable variables the left-hand side will be automatically
3014 allocated (if unallocated) or reallocated (if the shape or
3015 type parameter is different). To avoid the small performance
3016 penalty, you can use a(:) = ... instead of a = ... for arrays
3017 and character strings – or disable the feature using -std=f95
3018 or -fno-realloc-lhs.
3019 + Deferred type parameter: For scalar allocatable and pointer
3020 variables the character length can be deferred.
3021 + Namelist variables with allocatable and pointer attribute and
3022 nonconstant length type parameter are supported.
3023 * Fortran 2008 support has been extended:
3024 + Experimental [17]coarray support (for one image only, i.e.
3025 num_images() == 1); use the [18]-fcoarray=single flag to
3027 + The STOP and the new ERROR STOP statements now support all
3028 constant expressions.
3029 + Support for the CONTIGUOUS attribute.
3030 + Support for ALLOCATE with MOLD.
3031 + Support for the STORAGE_SIZE intrinsic inquiry function.
3032 + Support of the NORM2 and PARITY intrinsic functions.
3033 + The following bit intrinsics were added: POPCNT and POPPAR for
3034 counting the number of 1 bits and returning the parity; BGE,
3035 BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR
3036 for combined left and right shifts, MASKL and MASKR for simple
3037 left and right justified masks, MERGE_BITS for a bitwise merge
3038 using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations,
3039 and the transformational bit intrinsics IALL, IANY and
3041 + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine.
3042 + Support for the IMPURE attribute for procedures, which allows
3043 for ELEMENTAL procedures without the restrictions of PURE.
3044 + Null pointers (including NULL()) and not allocated variables
3045 can be used as actual argument to optional non-pointer,
3046 non-allocatable dummy arguments, denoting an absent argument.
3047 + Non-pointer variables with TARGET attribute can be used as
3048 actual argument to POINTER dummies with INTENT(IN)
3049 + Pointers including procedure pointers and those in a derived
3050 type (pointer components) can now be initialized by a target
3051 instead of only by NULL.
3052 + The EXIT statement (with construct-name) can now be used to
3053 leave not only the DO but also the ASSOCIATE, BLOCK, IF,
3054 SELECT CASE and SELECT TYPE constructs.
3055 + Internal procedures can now be used as actual argument.
3056 + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS
3057 and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV
3058 have been added; these arrays contain the supported kind
3059 values for the respective types.
3060 + The module procedures C_SIZEOF of the intrinsic module
3061 ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of
3062 ISO_FORTRAN_ENV have been implemented.
3063 + Minor changes: obsolescence diagnostics for ENTRY was added
3064 for -std=f2008; a line may start with a semicolon; for
3065 internal and module procedures END can be used instead of END
3066 SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes
3067 a RADIX argument; intrinsic types are supported for
3068 TYPE(intrinsic-type-spec); multiple type-bound procedures can
3069 be declared in a single PROCEDURE statement; implied-shape
3070 arrays are supported for named constants (PARAMETER). The
3071 transformational, three argument versions of BESSEL_JN and
3072 BESSEL_YN were added – the elemental, two-argument version had
3073 been added in GCC 4.4; note that the transformational
3074 functions use a recurrence algorithm.
3078 Support for the [19]Go programming language has been added to GCC. It
3079 is not enabled by default when you build GCC; use the
3080 --enable-languages configure option to build it. The driver program for
3081 compiling Go code is gccgo.
3083 Go is currently known to work on GNU/Linux and RTEMS. Solaris support
3084 is in progress. It may or may not work on other platforms.
3086 Objective-C and Objective-C++
3088 * The -fobjc-exceptions flag is now required to enable Objective-C
3089 exception and synchronization syntax (introduced by the keywords
3090 @try, @catch, @finally and @synchronized).
3091 * A number of Objective-C 2.0 features and extensions are now
3092 supported by GCC. These features are enabled by default; you can
3093 disable them by using the new -fobjc-std=objc1 command-line option.
3094 * The Objective-C 2.0 dot-syntax is now supported. It is an
3095 alternative syntax for using getters and setters; object.count is
3096 automatically converted into [object count] or [object setCount:
3097 ...] depending on context; for example if (object.count > 0) is
3098 automatically compiled into the equivalent of if ([object count] >
3099 0) while object.count = 0; is automatically compiled into the
3100 equivalent ot [object setCount: 0];. The dot-syntax can be used
3101 with instance and class objects and with any setters or getters, no
3102 matter if they are part of a declared property or not.
3103 * Objective-C 2.0 declared properties are now supported. They are
3104 declared using the new @property keyword, and are most commonly
3105 used in conjunction with the new Objective-C 2.0 dot-syntax. The
3106 nonatomic, readonly, readwrite, assign, retain, copy, setter and
3107 getter attributes are all supported. Marking declared properties
3108 with __attribute__ ((deprecated)) is supported too.
3109 * The Objective-C 2.0 @synthesize and @dynamic keywords are
3110 supported. @synthesize causes the compiler to automatically
3111 synthesize a declared property, while @dynamic is used to disable
3112 all warnings for a declared property for which no implementation is
3113 provided at compile time. Synthesizing declared properties requires
3114 runtime support in most useful cases; to be able to use it with the
3115 GNU runtime, appropriate helper functions have been added to the
3116 GNU Objective-C runtime ABI, and are implemented by the GNU
3117 Objective-C runtime library shipped with GCC.
3118 * The Objective-C 2.0 fast enumeration syntax is supported in
3119 Objective-C. This is currently not yet available in Objective-C++.
3120 Fast enumeration requires support in the runtime, and such support
3121 has been added to the GNU Objective-C runtime library (shipped with
3123 * The Objective-C 2.0 @optional keyword is supported. It allows you
3124 to mark methods or properties in a protocol as optional as opposed
3126 * The Objective-C 2.0 @package keyword is supported. It has currently
3127 the same effect as the @public keyword.
3128 * Objective-C 2.0 method attributes are supported. Currently the
3129 supported attributes are deprecated, sentinel, noreturn and format.
3130 * Objective-C 2.0 method argument attributes are supported. The most
3131 widely used attribute is unused, to mark an argument as unused in
3133 * Objective-C 2.0 class and protocol attributes are supported.
3134 Currently the only supported attribute is deprecated.
3135 * Objective-C 2.0 class extensions are supported. A class extension
3136 has the same syntax as a category declaration with no category
3137 name, and the methods and properties declared in it are added
3138 directly to the main class. It is mostly used as an alternative to
3139 a category to add methods to a class without advertising them in
3140 the public headers, with the advantage that for class extensions
3141 the compiler checks that all the privately declared methods are
3142 actually implemented.
3143 * As a result of these enhancements, GCC can now be used to build
3144 Objective-C and Objective-C++ software that uses Foundation and
3145 other important system frameworks with the NeXT runtime on Darwin 9
3146 and Darwin 10 (OSX 10.5 and 10.6).
3147 * Many bugs in the compiler have been fixed in this release; in
3148 particular, LTO can now be used when compiling Objective-C and
3149 Objective-C++ and the parser is much more robust in dealing with
3152 Runtime Library (libobjc)
3154 * The GNU Objective-C runtime library now defines the macro
3155 __GNU_LIBOBJC__ (with a value that is increased at every release
3156 where there is any change to the API) in objc/objc.h, making it
3157 easy to determine if the GNU Objective-C runtime library is being
3158 used, and if so, which version. Previous versions of the GNU
3159 Objective-C runtime library (and other Objective-C runtime
3160 libraries such as the Apple one) do not define this macro.
3161 * A new Objective-C 2.0 API, almost identical to the one implemented
3162 by the Apple Objective-C runtime, has been implemented in the GNU
3163 Objective-C runtime library. The new API hides the internals of
3164 most runtime structures but provides a more extensive set of
3165 functions to operate on them. It is much easier, for example, to
3166 create or modify classes at runtime. The new API also makes it
3167 easier to port software from Apple to GNU as almost no changes
3168 should be required. The old API is still supported for backwards
3169 compatibility; including the old objc/objc-api.h header file
3170 automatically selects the old API, while including the new
3171 objc/runtime.h header file automatically selects the new API.
3172 Support for the old API is being phased out and upgrading the
3173 software to use the new API is strongly recommended. To check for
3174 the availability of the new API, the __GNU_LIBOBJC__ macro can be
3175 used as older versions of the GNU Objective-C runtime library,
3176 which do not support the new API, do not define such a macro.
3177 * Runtime support for @synchronized has been added.
3178 * Runtime support for Objective-C 2.0 synthesized property accessors
3180 * Runtime support for Objective-C 2.0 fast enumeration has been
3183 New Targets and Target Specific Improvements
3187 * GCC now supports the Cortex-M4 processor implementing the v7-em
3188 version of the architecture using the option -mcpu=cortex-m4.
3189 * Scheduling descriptions for the Cortex-M4, the Neon and the
3190 floating point units of the Cortex-A9 and a pipeline description
3191 for the Cortex-A5 have been added.
3192 * Synchronization primitives such as __sync_fetch_and_add and friends
3193 are now inlined for supported architectures rather than calling
3194 into a kernel helper function.
3195 * SSA loop prefetching is enabled by default for the Cortex-A9 at
3197 * Several improvements were committed to improve code generation for
3198 the ARM architecture including a rewritten implementation for load
3199 and store multiples.
3200 * Several enhancements were committed to improve SIMD code generation
3201 for NEON by adding support for widening instructions, misaligned
3202 loads and stores, vector conditionals and support for 64 bit
3204 * Support was added for the Faraday cores fa526, fa606te, fa626te,
3205 fmp626te, fmp626 and fa726te and can be used with the respective
3206 names as parameters to the -mcpu= option.
3207 * Basic support was added for Cortex-A15 and is available through
3209 * GCC for AAPCS configurations now more closely adheres to the AAPCS
3210 specification by enabling -fstrict-volatile-bitfields by default.
3214 * The new -fsplit-stack option permits programs to use a
3215 discontiguous stack. This is useful for threaded programs, in that
3216 it is no longer necessary to specify the maximum stack size when
3217 creating a thread. This feature is currently only implemented for
3218 32-bit and 64-bit x86 GNU/Linux targets.
3219 * Support for emitting profiler counter calls before function
3220 prologues. This is enabled via a new command-line option -mfentry.
3221 * Optimization for the Intel Core 2 processors is now available
3222 through the -march=core2 and -mtune=core2 options.
3223 * Support for Intel Core i3/i5/i7 processors is now available through
3224 the -march=corei7 and -mtune=corei7 options.
3225 * Support for Intel Core i3/i5/i7 processors with AVX is now
3226 available through the -march=corei7-avx and -mtune=corei7-avx
3228 * Support for AMD Bobcat (family 14) processors is now available
3229 through the -march=btver1 and -mtune=btver1 options.
3230 * Support for AMD Bulldozer (family 15) processors is now available
3231 through the -march=bdver1 and -mtune=bdver1 options.
3232 * The default setting (when not optimizing for size) for 32-bit
3233 GNU/Linux and Darwin x86 targets has been changed to
3234 -fomit-frame-pointer. The default can be reverted to
3235 -fno-omit-frame-pointer by configuring GCC with the
3236 --enable-frame-pointer configure option.
3237 * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support
3238 __float128 on 32-bit and 64-bit x86 targets.
3239 * AVX floating-point arithmetic can now be enabled by default at
3240 configure time with the new --with-fpmath=avx option.
3241 * The SSA loop prefetching pass is enabled when using -O3 when
3242 optimizing for CPUs where prefetching is beneficial (AMD CPUs newer
3244 * Support for TBM (Trailing Bit Manipulation) built-in functions and
3245 code generation is available via -mtbm.
3246 * Support for AMD's BMI (Bit Manipulation) built-in functions and
3247 code generation is available via -mbmi.
3251 * Support has been added for the Xilinx MicroBlaze softcore processor
3252 (microblaze-elf) embedded target. This configurable processor is
3253 supported on several Xilinx Spartan and Virtex FPGAs.
3257 * GCC now supports the Loongson 3A processor. Its canonical -march=
3258 and -mtune= name is loongson3a.
3262 * The inline assembly register constraint "A" has been renamed "c".
3263 This constraint is used to select a floating-point register that
3264 can be used as the destination of a multiply-accumulate
3266 * New inline assembly register constraints "A" and "D" have been
3267 added. These constraint letters resolve to all general registers
3268 when compiling for AM33, and resolve to address registers only or
3269 data registers only when compiling for MN10300.
3270 * The MDR register is represented in the compiler. One can access the
3271 register via the "z" constraint in inline assembly. It can be
3272 marked as clobbered or used as a local register variable via the
3273 "mdr" name. The compiler uses the RETF instruction if the function
3274 does not modify the MDR register, so it is important that inline
3275 assembly properly annotate any usage of the register.
3279 * GCC now supports the Applied Micro Titan processor with
3281 * The -mrecip option has been added, which indicates whether the
3282 reciprocal and reciprocal square root instructions should be used.
3283 * The -mveclibabi=mass option can be used to enable the compiler to
3284 autovectorize mathematical functions using the Mathematical
3285 Acceleration Subsystem library.
3286 * The -msingle-pic-base option has been added, which instructs the
3287 compiler to avoid loading the PIC base register in function
3288 prologues. The PIC base register must be initialized by the runtime
3290 * The -mblock-move-inline-limit option has been added, which enables
3291 the user to control the maximum size of inlined memcpy calls and
3293 * PowerPC64 GNU/Linux support for applications requiring a large TOC
3294 section has been improved. A new command-line option,
3295 -mcmodel=MODEL, controls this feature; valid values for MODEL are
3296 small, medium, or large.
3297 * The Altivec builtin functions vec_ld and vec_st have been modified
3298 to generate the Altivec memory instructions LVX and STVX, even if
3299 the -mvsx option is used. In the initial GCC 4.5 release, these
3300 builtin functions were changed to generate VSX memory reference
3301 instructions instead of Altivec memory instructions, but there are
3302 differences between the two instructions. If the VSX instruction
3303 set is available, you can now use the new builtin functions
3304 vec_vsx_ld and vec_vsx_st which always generates the VSX memory
3306 * The GCC compiler on AIX now defaults to a process layout with a
3307 larger data space allowing larger programs to be compiled.
3308 * The GCC long double type on AIX 6.1 and above has reverted to 64
3309 bit double precision, matching the AIX XL compiler default, because
3310 of missing C99 symbols required by the GCC runtime.
3311 * The default processor scheduling model and tuning for PowerPC64
3312 GNU/Linux and for AIX 6.1 and above now is POWER7.
3313 * Starting with GCC 4.6.1, vectors of type vector long long or vector
3314 long are passed and returned in the same method as other vectors
3315 with the VSX instruction set. Previously the GCC compiler did not
3316 adhere to the ABI for 128-bit vectors with 64-bit integer base
3317 types (PR 48857). This is also fixed in the GCC 4.5.4 release.
3319 S/390, zSeries and System z9/z10, IBM zEnterprise z196
3321 * Support for the zEnterprise z196 processor has been added. When
3322 using the -march=z196 option, the compiler will generate code
3323 making use of the following instruction facilities:
3324 + Conditional load/store
3326 + Floating-point-extension
3327 + Interlocked-access
3329 The -mtune=z196 option avoids the compare and branch instructions
3330 as well as the load address instruction with an index register as
3331 much as possible and performs instruction scheduling appropriate
3332 for the new out-of-order pipeline architecture.
3333 * When using the -m31 -mzarch options the generated code still
3334 conforms to the 32-bit ABI but uses the general purpose registers
3335 as 64-bit registers internally. This requires a Linux kernel saving
3336 the whole 64-bit registers when doing a context switch. Kernels
3337 providing that feature indicate that by the 'highgprs' string in
3339 * The SSA loop prefetching pass is enabled when using -O3.
3343 * GCC now supports the LEON series of SPARC V8 processors. The code
3344 generated by the compiler can either be tuned to it by means of the
3345 --with-tune=leon configure option and -mtune=leon compilation
3346 option, or the compiler can be built for the sparc-leon-{elf,linux}
3347 and sparc-leon3-{elf,linux} targets directly.
3348 * GCC has stopped sign/zero-extending parameter registers in the
3349 callee for functions taking parameters with sub-word size in 32-bit
3350 mode, since this is redundant with the specification of the ABI.
3351 GCC has never done so in 64-bit mode since this is also redundant.
3352 * The command line option -mfix-at697f has been added to enable the
3353 documented workaround for the single erratum of the Atmel AT697F
3360 * GCC now supports the Bionic C library and provides a convenient way
3361 of building native libraries and applications for the Android
3362 platform. Refer to the documentation of the -mandroid and -mbionic
3363 options for details on building native code. At the moment, Android
3364 support is enabled only for ARM.
3369 + Initial support for CFString types has been added.
3370 This allows GCC to build projects including the system Core
3371 Foundation frameworks. The GCC Objective-C family supports
3372 CFString "toll-free bridged" as per the Mac OS X system tools.
3373 CFString is also recognized in the context of format
3374 attributes and arguments (see the documentation for format
3375 attributes for limitations). At present, 8-bit character types
3377 + Object file size reduction.
3378 The Darwin zeroed memory allocators have been re-written to
3379 make more use of .zerofill sections. For non-debug code, this
3380 can reduce object file size significantly.
3381 + Objective-C family 64-bit support (NeXT ABI 2).
3382 Initial support has been added to support 64-bit Objective-C
3383 code using the Darwin/OS X native (NeXT) runtime. ABI version
3384 2 will be selected automatically when 64-bit code is built.
3385 + Objective-C family 32-bit ABI 1.
3386 For 32-bit code ABI 1 is also now also allowed. At present it
3387 must be selected manually using -fobjc-abi-version=1 where
3388 applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6).
3390 + The -mdynamic-no-pic option has been enabled.
3391 Code supporting -mdynamic-no-pic optimization has been added
3392 and is applicable to -m32 builds. The compiler bootstrap uses
3393 the option where appropriate.
3394 + The default value for -mtune= has been changed.
3395 Since Darwin systems are primarily Xeon, Core-2 or similar the
3396 default tuning has been changed to -mtune=core2.
3397 + Enable 128-bit long double (__float128) support on Darwin.
3400 Several significant bugs have been fixed, such that GCC now
3401 produces code compatible with the Darwin64 PowerPC ABI.
3402 + libffi and boehm-gc.
3403 The Darwin ports of the libffi and boehm-gc libraries have
3404 been upgraded to include a Darwin64 implementation. This means
3405 that powerpc*-*-darwin9 platforms may now, for example, build
3406 Java applications with -m64 enabled.
3407 + Plug-in support has been enabled.
3408 + The -fsection-anchors option is now available although,
3409 presently, not heavily tested.
3415 * Support symbol versioning with the Sun linker.
3416 * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+.
3417 * Support thread-local storage (TLS) with the Sun assembler on
3419 * Support TLS on Solaris 8/9 if prerequisites are met.
3420 * Support COMDAT group with the GNU assembler and recent Sun linker.
3421 * Support the Sun assembler visibility syntax.
3422 * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp.
3423 -march=pentiumpro (Solaris 8/9).
3424 * Don't use SSE on Solaris 8/9 x86 by default.
3425 * Enable 128-bit long double (__float128) support on Solaris 2/x86.
3429 * Change the ABI for returning 8-byte vectors like __m64 in MMX
3430 registers on Solaris 10+/x86 to match the Sun Studio 12.1+
3431 compilers. This is an incompatible change. If you use such types,
3432 you must either recompile all your code with the new compiler or
3433 use the new -mvect8-ret-in-mem option to remain compatible with
3434 previous versions of GCC and Sun Studio.
3438 * Initial support for decimal floating point.
3439 * Support for the __thiscall calling-convention.
3440 * Support for hot-patchable function prologues via the
3441 ms_hook_prologue attribute for x86_64 in addition to 32-bit x86.
3442 * Improvements of stack-probing and stack-allocation mechanisms.
3443 * Support of push/pop-macro pragma as preprocessor command.
3444 With #pragma push_macro("macro-name") the current definition of
3445 macro-name is saved and can be restored with #pragma
3446 pop_macro("macro-name") to its saved definition.
3447 * Enable 128-bit long double (__float128) support on MinGW and
3450 Other significant improvements
3452 Installation changes
3454 * An install-strip make target is provided that installs stripped
3455 executables, and may install libraries with unneeded or debugging
3457 * On Power7 systems, there is a potential problem if you build the
3458 GCC compiler with a host compiler using options that enable the VSX
3459 instruction set generation. If the host compiler has been patched
3460 so that the vec_ld and vec_st builtin functions generate Altivec
3461 memory instructions instead of VSX memory instructions, then you
3462 should be able to build the compiler with VSX instruction
3465 Changes for GCC Developers
3467 Note: these changes concern developers that develop GCC itself or
3468 software that integrates with GCC, such as plugins, and not the general
3470 * The gengtype utility, which previously was internal to the GCC
3471 build process, has been enchanced to provide GC root information
3472 for plugins as necessary.
3473 * The old GC allocation interface of ggc_alloc and friends was
3474 replaced with a type-safe alternative.
3478 This is the [20]list of problem reports (PRs) from GCC's bug tracking
3479 system that are known to be fixed in the 4.6.1 release. This list might
3480 not be complete (that is, it is possible that some PRs that have been
3481 fixed are not listed here).
3485 This is the [21]list of problem reports (PRs) from GCC's bug tracking
3486 system that are known to be fixed in the 4.6.2 release. This list might
3487 not be complete (that is, it is possible that some PRs that have been
3488 fixed are not listed here).
3492 This is the [22]list of problem reports (PRs) from GCC's bug tracking
3493 system that are known to be fixed in the 4.6.3 release. This list might
3494 not be complete (that is, it is possible that some PRs that have been
3495 fixed are not listed here).
3499 This is the [23]list of problem reports (PRs) from GCC's bug tracking
3500 system that are known to be fixed in the 4.6.4 release. This list might
3501 not be complete (that is, it is possible that some PRs that have been
3502 fixed are not listed here).
3505 For questions related to the use of GCC, please consult these web
3506 pages and the [24]GCC manuals. If that fails, the
3507 [25]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3508 web pages and the development of GCC are welcome on our developer
3509 list at [26]gcc@gcc.gnu.org. All of [27]our lists have public
3512 Copyright (C) [28]Free Software Foundation, Inc. Verbatim copying and
3513 distribution of this entire article is permitted in any medium,
3514 provided this notice is preserved.
3516 These pages are [29]maintained by the GCC team. Last modified
3521 1. http://sourceware.org/bugzilla/show_bug.cgi?id=10401
3522 2. http://gcc.gnu.org/gcc-4.5/changes.html
3523 3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted
3524 4. http://gcc.gnu.org/gcc-4.6/porting_to.html
3525 5. http://gcc.gnu.org/projects/lto/whopr.pdf
3526 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
3527 7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html
3528 8. https://gcc.gnu.org/PR43145
3529 9. https://gcc.gnu.org/PR43680
3530 10. https://gcc.gnu.org/PR33558
3531 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253
3532 12. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x
3533 13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html
3534 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races
3535 15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html
3536 16. https://gcc.gnu.org/wiki/OOP
3537 17. https://gcc.gnu.org/wiki/Coarray
3538 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233
3539 19. http://golang.org/
3540 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1
3541 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2
3542 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3
3543 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4
3544 24. https://gcc.gnu.org/onlinedocs/
3545 25. mailto:gcc-help@gcc.gnu.org
3546 26. mailto:gcc@gcc.gnu.org
3547 27. https://gcc.gnu.org/lists.html
3548 28. http://www.fsf.org/
3549 29. https://gcc.gnu.org/about.html
3550 30. http://validator.w3.org/check/referer
3551 ======================================================================
3552 http://gcc.gnu.org/gcc-4.5/index.html
3553 GCC 4.5 Release Series
3557 The [1]GNU project and the GCC developers are pleased to announce the
3558 release of GCC 4.5.4.
3560 This release is a bug-fix release, containing fixes for regressions in
3561 GCC 4.5.3 relative to previous releases of GCC.
3566 Jul 2, 2012 ([2]changes)
3569 Apr 28, 2011 ([3]changes)
3572 Dec 16, 2010 ([4]changes)
3575 Jul 31, 2010 ([5]changes)
3578 April 14, 2010 ([6]changes)
3580 References and Acknowledgements
3582 GCC used to stand for the GNU C Compiler, but since the compiler
3583 supports several other languages aside from C, it now stands for the
3584 GNU Compiler Collection.
3586 A list of [7]successful builds is updated as new information becomes
3589 The GCC developers would like to thank the numerous people that have
3590 contributed new features, improvements, bug fixes, and other changes as
3591 well as test results to GCC. This [8]amazing group of volunteers is
3592 what makes GCC successful.
3594 For additional information about GCC please refer to the [9]GCC project
3595 web site or contact the [10]GCC development mailing list.
3597 To obtain GCC please use [11]our mirror sites or [12]our SVN server.
3600 For questions related to the use of GCC, please consult these web
3601 pages and the [13]GCC manuals. If that fails, the
3602 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3603 web pages and the development of GCC are welcome on our developer
3604 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
3607 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
3608 distribution of this entire article is permitted in any medium,
3609 provided this notice is preserved.
3611 These pages are [18]maintained by the GCC team. Last modified
3616 1. http://www.gnu.org/
3617 2. http://gcc.gnu.org/gcc-4.5/changes.html
3618 3. http://gcc.gnu.org/gcc-4.5/changes.html
3619 4. http://gcc.gnu.org/gcc-4.5/changes.html
3620 5. http://gcc.gnu.org/gcc-4.5/changes.html
3621 6. http://gcc.gnu.org/gcc-4.5/changes.html
3622 7. http://gcc.gnu.org/gcc-4.5/buildstat.html
3623 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
3624 9. http://gcc.gnu.org/index.html
3625 10. mailto:gcc@gcc.gnu.org
3626 11. http://gcc.gnu.org/mirrors.html
3627 12. http://gcc.gnu.org/svn.html
3628 13. https://gcc.gnu.org/onlinedocs/
3629 14. mailto:gcc-help@gcc.gnu.org
3630 15. mailto:gcc@gcc.gnu.org
3631 16. https://gcc.gnu.org/lists.html
3632 17. http://www.fsf.org/
3633 18. https://gcc.gnu.org/about.html
3634 19. http://validator.w3.org/check/referer
3635 ======================================================================
3636 http://gcc.gnu.org/gcc-4.5/changes.html
3637 GCC 4.5 Release Series
3638 Changes, New Features, and Fixes
3642 * GCC now requires the [1]MPC library in order to build. See the
3643 [2]prerequisites page for version requirements.
3644 * Support for a number of older systems and recently unmaintained or
3645 untested target ports of GCC has been declared obsolete in GCC 4.5.
3646 Unless there is activity to revive them, the next release of GCC
3647 will have their sources permanently removed.
3648 The following ports for individual systems on particular
3649 architectures have been obsoleted:
3650 + IRIX releases before 6.5 (mips-sgi-irix5*,
3651 mips-sgi-irix6.[0-4])
3652 + Solaris 7 (*-*-solaris2.7)
3653 + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*,
3655 + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions
3656 can be found in the [3]announcement.
3657 Support for the classic POWER architecture implemented in the
3658 original RIOS and RIOS2 processors of the old IBM RS/6000 product
3659 line has been obsoleted in the rs6000 port. This does not affect
3660 the new generation Power and PowerPC architectures.
3661 * Support has been removed for all the [4]configurations obsoleted in
3663 * Support has been removed for the protoize and unprotoize utilities,
3664 obsoleted in GCC 4.4.
3665 * Support has been removed for tuning for Itanium1 (Merced) variants.
3666 Note that code tuned for Itanium2 should also run correctly on
3668 * GCC now generates unwind info also for epilogues. DWARF debuginfo
3669 generated by GCC now uses more features of DWARF3 than before, and
3670 also some DWARF4 features. GDB older than 7.0 is not able to handle
3671 either of these, so to debug GCC 4.5 generated binaries or
3672 libraries GDB 7.0 or later is needed. You can disable use of DWARF4
3673 features with the -gdwarf-3 -gstrict-dwarf options, or use
3674 -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but
3675 epilogue unwind info is emitted unconditionally whenever unwind
3677 * On x86 targets, code containing floating-point calculations may run
3678 significantly slower when compiled with GCC 4.5 in strict C99
3679 conformance mode than they did with earlier GCC versions. This is
3680 due to stricter standard conformance of the compiler and can be
3681 avoided by using the option -fexcess-precision=fast; also see
3683 * The function attribute noinline no longer prevents GCC from cloning
3684 the function. A new attribute noclone has been introduced for this
3685 purpose. Cloning a function means that it is duplicated and the new
3686 copy is specialized for certain contexts (for example when a
3687 parameter is a known constant).
3689 General Optimizer Improvements
3691 * The -save-temps now takes an optional argument. The -save-temps and
3692 -save-temps=cwd switches write the temporary files in the current
3693 working directory based on the original source file. The
3694 -save-temps=obj switch will write files into the directory
3695 specified with the -o option, and the intermediate filenames are
3696 based on the output file. This will allow the user to get the
3697 compiler intermediate files when doing parallel builds without two
3698 builds of the same filename located in different directories from
3699 interfering with each other.
3700 * Debugging dumps are now created in the same directory as the object
3701 file rather than in the current working directory. This allows the
3702 user to get debugging dumps when doing parallel builds without two
3703 builds of the same filename interfering with each other.
3704 * GCC has been integrated with the [6]MPC library. This allows GCC to
3705 evaluate complex arithmetic at compile time [7]more accurately. It
3706 also allows GCC to evaluate calls to complex built-in math
3707 functions having constant arguments and replace them at compile
3708 time with their mathematically equivalent results. In doing so, GCC
3709 can generate correct results regardless of the math library
3710 implementation or floating point precision of the host platform.
3711 This also allows GCC to generate identical results regardless of
3712 whether one compiles in native or cross-compile configurations to a
3713 particular target. The following built-in functions take advantage
3714 of this new capability: cacos, cacosh, casin, casinh, catan,
3715 catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan,
3716 and ctanh. The float and long double variants of these functions
3717 (e.g. csinf and csinl) are also handled.
3718 * A new link-time optimizer has been added ([8]-flto). When this
3719 option is used, GCC generates a bytecode representation of each
3720 input file and writes it to specially-named sections in each object
3721 file. When the object files are linked together, all the function
3722 bodies are read from these named sections and instantiated as if
3723 they had been part of the same translation unit. This enables
3724 interprocedural optimizations to work across different files (and
3725 even different languages), potentially improving the performance of
3726 the generated code. To use the link-timer optimizer, -flto needs to
3727 be specified at compile time and during the final link. If the
3728 program does not require any symbols to be exported, it is possible
3729 to combine -flto and the experimental [9]-fwhopr with
3730 [10]-fwhole-program to allow the interprocedural optimizers to use
3731 more aggressive assumptions.
3732 * The automatic parallelization pass was enhanced to support
3733 parallelization of outer loops.
3734 * Automatic parallelization can be enabled as part of Graphite. In
3735 addition to -ftree-parallelize-loops=, specify
3736 -floop-parallelize-all to enable the Graphite-based optimization.
3737 * The infrastructure for optimizing based on [11]restrict qualified
3738 pointers has been rewritten and should result in code generation
3739 improvements. Optimizations based on restrict qualified pointers
3740 are now also available when using -fno-strict-aliasing.
3741 * There is a new optimization pass that attempts to change prototype
3742 of functions to avoid unused parameters, pass only relevant parts
3743 of structures and turn arguments passed by reference to arguments
3744 passed by value when possible. It is enabled by -O2 and above as
3745 well as -Os and can be manually invoked using the new command-line
3747 * GCC now optimize exception handling code. In particular cleanup
3748 regions that are proved to not have any effect are optimized out.
3750 New Languages and Language specific improvements
3754 * The -fshow-column option is now on by default. This means error
3755 messages now have a column associated with them.
3759 * Compilation of programs heavily using discriminated record types
3760 with variant parts has been sped up and generates more compact
3762 * Stack checking now works reasonably well on most plaforms. In some
3763 specific cases, stack overflows may still fail to be detected, but
3764 a compile-time warning will be issued for these cases.
3768 * If a header named in a #include directive is not found, the
3769 compiler exits immediately. This avoids a cascade of errors arising
3770 from declarations expected to be found in that header being
3772 * A new built-in function __builtin_unreachable() has been added that
3773 tells the compiler that control will never reach that point. It may
3774 be used after asm statements that terminate by transferring control
3775 elsewhere, and in other places that are known to be unreachable.
3776 * The -Wlogical-op option now warns for logical expressions such as
3777 (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be
3778 mistakes. This option is disabled by default.
3779 * An asm goto feature has been added to allow asm statements that
3781 * C++0x raw strings are supported for C++ and for C with -std=gnu99.
3782 * The deprecated attribute now takes an optional string argument, for
3783 example, __attribute__((deprecated("text string"))), that will be
3784 printed together with the deprecation warning.
3788 * The -Wenum-compare option, which warns when comparing values of
3789 different enum types, now works for C. It formerly only worked for
3790 C++. This warning is enabled by -Wall. It may be avoided by using a
3792 * The -Wcast-qual option now warns about casts which are unsafe in
3793 that they permit const-correctness to be violated without further
3794 warnings. Specifically, it warns about cases where a qualifier is
3795 added when all the lower types are not const. For example, it warns
3796 about a cast from char ** to const char **.
3797 * The -Wc++-compat option is significantly improved. It issues new
3799 + Using C++ reserved operator names as identifiers.
3800 + Conversions to enum types without explicit casts.
3801 + Using va_arg with an enum type.
3802 + Using different enum types in the two branches of ?:.
3803 + Using ++ or -- on a variable of enum type.
3804 + Using the same name as both a struct, union or enum tag and a
3805 typedef, unless the typedef refers to the tagged type itself.
3806 + Using a struct, union, or enum which is defined within another
3808 + A struct field defined using a typedef if there is a field in
3809 the struct, or an enclosing struct, whose name is the typedef
3811 + Duplicate definitions at file scope.
3812 + Uninitialized const variables.
3813 + A global variable with an anonymous struct, union, or enum
3815 + Using a string constant to initialize a char array whose size
3816 is the length of the string.
3817 * The new -Wjump-misses-init option warns about cases where a goto or
3818 switch skips the initialization of a variable. This sort of branch
3819 is an error in C++ but not in C. This warning is enabled by
3821 * GCC now ensures that a C99-conforming <stdint.h> is present on most
3822 targets, and uses information about the types in this header to
3823 implement the Fortran bindings to those types. GCC does not ensure
3824 the presence of such a header, and does not implement the Fortran
3825 bindings, on the following targets: NetBSD, VxWorks, VMS,
3826 SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF.
3827 * GCC now implements C90- and C99-conforming rules for constant
3828 expressions. This may cause warnings or errors for some code using
3829 expressions that can be folded to a constant but are not constant
3830 expressions as defined by ISO C.
3831 * All known target-independent C90 and C90 Amendment 1 conformance
3832 bugs, and all known target-independent C99 conformance bugs not
3833 related to floating point or extended identifiers, have been fixed.
3834 * The C decimal floating point support now includes support for the
3835 FLOAT_CONST_DECIMAL64 pragma.
3836 * The named address space feature from ISO/IEC TR 18037 is now
3837 supported. This is currently only implemented for the SPU
3842 * Improved [12]experimental support for the upcoming C++0x ISO C++
3843 standard, including support for raw strings, lambda expressions and
3844 explicit type conversion operators.
3845 * When printing the name of a class template specialization, G++ will
3846 now omit any template arguments which come from default template
3847 arguments. This behavior (and the pretty-printing of function
3848 template specializations as template signature and arguments) can
3849 be disabled with the -fno-pretty-templates option.
3850 * Access control is now applied to typedef names used in a template,
3851 which may cause G++ to reject some ill-formed code that was
3852 accepted by earlier releases. The -fno-access-control option can be
3853 used as a temporary workaround until the code is corrected.
3854 * Compilation time for code that uses templates should now scale
3855 linearly with the number of instantiations rather than
3856 quadratically, as template instantiations are now looked up using
3858 * Declarations of functions that look like builtin declarations of
3859 library functions are only considered to be redeclarations if they
3860 are declared with extern "C". This may cause problems with code
3861 that omits extern "C" on hand-written declarations of C library
3862 functions such as abort or memcpy. Such code is ill-formed, but was
3863 accepted by earlier releases.
3864 * Diagnostics that used to complain about passing non-POD types to
3865 ... or jumping past the declaration of a non-POD variable now check
3866 for triviality rather than PODness, as per C++0x.
3867 * In C++0x mode local and anonymous classes are now allowed as
3868 template arguments, and in declarations of variables and functions
3869 with linkage, so long as any such declaration that is used is also
3870 defined ([13]DR 757).
3871 * Labels may now have attributes, as has been permitted for a while
3872 in C. This is only permitted when the label definition and the
3873 attribute specifier is followed by a semicolon—i.e., the label
3874 applies to an empty statement. The only useful attribute for a
3876 * G++ now implements [14]DR 176. Previously G++ did not support using
3877 the injected-class-name of a template base class as a type name,
3878 and lookup of the name found the declaration of the template in the
3879 enclosing scope. Now lookup of the name finds the
3880 injected-class-name, which can be used either as a type or as a
3881 template, depending on whether or not the name is followed by a
3882 template argument list. As a result of this change, some code that
3883 was previously accepted may be ill-formed because
3884 1. The injected-class-name is not accessible because it's from a
3886 2. The injected-class-name cannot be used as an argument for a
3887 template template parameter.
3888 In either of these cases, the code can be fixed by adding a
3889 nested-name-specifier to explicitly name the template. The first
3890 can be worked around with -fno-access-control; the second is only
3891 rejected with -pedantic.
3892 * A new standard mangling for SIMD vector types has been added, to
3893 avoid name clashes on systems with vectors of varying length. By
3894 default the compiler still uses the old mangling, but emits aliases
3895 with the new mangling on targets that support strong aliases. Users
3896 can switch over entirely to the new mangling with -fabi-version=4
3897 or -fabi-version=0. -Wabi will now warn about code that uses the
3899 * The command-line option -ftemplate-depth-N is now written as
3900 -ftemplate-depth=N and the old form is deprecated.
3901 * Conversions between NULL and non-pointer types are now warned by
3902 default. The new option -Wno-conversion-null disables these
3903 warnings. Previously these warnings were only available when using
3904 -Wconversion explicitly.
3906 Runtime Library (libstdc++)
3908 * Improved experimental support for the upcoming ISO C++ standard,
3910 + Support for <future>, <functional>, and <random>.
3911 + Existing facilities now exploit explicit operators and the
3912 newly implemented core C++0x features.
3913 + The header <cstdatomic> has been renamed to <atomic>.
3914 * An experimental [15]profile mode has been added. This is an
3915 implementation of many C++ standard library constructs with an
3916 additional analysis layer that gives performance improvement advice
3917 based on recognition of suboptimal usage patterns. For example,
3922 for (int k = 0; k < 1024; ++k)
3923 v.insert(v.begin(), k);
3926 When instrumented via the profile mode, can return suggestions
3927 about the initial size and choice of the container used as follows:
3928 vector-to-list: improvement = 5: call stack = 0x804842c ...
3929 : advice = change std::vector to std::list
3930 vector-size: improvement = 3: call stack = 0x804842c ...
3931 : advice = change initial container size from 0 to 1024
3933 These constructs can be substituted for the normal libstdc++
3934 constructs on a piecemeal basis, or all existing components can be
3935 transformed via the -D_GLIBCXX_PROFILE macro.
3936 * [16]Support for decimal floating-point arithmetic (aka ISO C++ TR
3937 24733) has been added. This support is in header file
3938 <decimal/decimal>, uses namespace std::decimal, and includes
3939 classes decimal32, decimal64, and decimal128.
3940 * Sources have been audited for application of function attributes
3941 nothrow, const, pure, and noreturn.
3942 * Python pretty-printers have been added for many standard library
3943 components that simplify the internal representation and present a
3944 more intuitive view of components when used with
3945 appropriately-advanced versions of GDB. For more information,
3946 please consult the more [17]detailed description.
3947 * The default behavior for comparing typeinfo names has changed, so
3948 in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero.
3949 * The new -static-libstdc++ option directs g++ to link the C++
3950 library statically, even if the default would normally be to link
3955 * The COMMON default padding has been changed – instead of adding the
3956 padding before a variable it is now added afterwards, which
3957 increases the compatibility with other vendors and helps to obtain
3958 the correct output in some cases. Cf. also the -falign-commons
3959 option ([18]added in 4.4).
3960 * The -finit-real= option now also supports the value snan for
3961 signalling not-a-number; to be effective, one additionally needs to
3962 enable trapping (e.g. via -ffpe-trap=). Note: Compile-time
3963 optimizations can turn a signalling NaN into a quiet one.
3964 * The new option -fcheck= has been added with the options bounds,
3965 array-temps, do, pointer, and recursive. The bounds and array-temps
3966 options are equivalent to -fbounds-check and
3967 -fcheck-array-temporaries. The do option checks for invalid
3968 modification of loop iteration variables, and the recursive option
3969 tests for recursive calls to subroutines/functions which are not
3970 marked as recursive. With pointer pointer association checks in
3971 calls are performed; however, neither undefined pointers nor
3972 pointers in expressions are handled. Using -fcheck=all enables all
3973 these run-time checks.
3974 * The run-time checking -fcheck=bounds now warns about invalid string
3975 lengths of character dummy arguments. Additionally, more
3976 compile-time checks have been added.
3977 * The new option [19]-fno-protect-parens has been added; if set, the
3978 compiler may reorder REAL and COMPLEX expressions without regard to
3980 * GNU Fortran no longer links against libgfortranbegin. As before,
3981 MAIN__ (assembler symbol name) is the actual Fortran main program,
3982 which is invoked by the main function. However, main is now
3983 generated and put in the same object file as MAIN__. For the time
3984 being, libgfortranbegin still exists for backward compatibility.
3985 For details see the new [20]Mixed-Language Programming chapter in
3987 * The I/O library was restructured for performance and cleaner code.
3988 * Array assignments and WHERE are now run in parallel when OpenMP's
3990 * The experimental option -fwhole-file was added. The option allows
3991 whole-file checking of procedure arguments and allows for better
3992 optimizations. It can also be used with -fwhole-program, which is
3993 now also supported in gfortran.
3994 * More Fortran 2003 and Fortran 2008 mathematical functions can now
3995 be used as initialization expressions.
3996 * Some extended attributes such as STDCALL are now supported via the
3997 [21]GCC$ compiler directive.
3998 * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN
3999 intrinsic behaves now as if zero were always positive.
4000 * For legacy compatibiliy: On Cygwin and MinGW, the special files
4001 CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now
4003 * Fortran 2003 support has been extended:
4004 + Procedure-pointer function results and procedure-pointer
4005 components (including PASS),
4006 + allocatable scalars (experimental),
4007 + DEFERRED type-bound procedures,
4008 + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements
4009 have been implemented.
4010 + The ALLOCATE statement supports type-specs and the SOURCE=
4012 + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC
4013 type-bound procedure (i.e. as type-bound operators).
4014 + Rounding (ROUND=, RZ, ...) for output is now supported.
4015 + The INT_FAST{8,16,32,64,128}_T kind type parameters of the
4016 intrinsic module ISO_C_BINDING are now supported, except for
4017 the targets listed above as ones where GCC does not have
4018 <stdint.h> type information.
4019 + Extensible derived types with type-bound procedure or
4020 procedure pointer with PASS attribute now have to use CLASS in
4021 line with the Fortran 2003 standard; the workaround to use
4022 TYPE is no longer supported.
4023 + [22]Experimental, incomplete support for polymorphism,
4024 including CLASS, SELECT TYPE and dynamic dispatch of
4025 type-bound procedure calls. Some features do not work yet such
4026 as unlimited polymorphism (CLASS(*)).
4027 * Fortran 2008 support has been extended:
4028 + The OPEN statement now supports the NEWUNIT= option, which
4029 returns a unique file unit, thus preventing inadvertent use of
4030 the same unit in different parts of the program.
4031 + Support for unlimited format items has been added.
4032 + The INT{8,16,32} and REAL{32,64,128} kind type parameters of
4033 the intrinsic module ISO_FORTRAN_ENV are now supported.
4034 + Using complex arguments with TAN, SINH, COSH, TANH, ASIN,
4035 ACOS, and ATAN is now possible; the functions ASINH, ACOSH,
4036 and ATANH have been added (for real and complex arguments) and
4037 ATAN(Y,X) is now an alias for ATAN2(Y,X).
4038 + The BLOCK construct has been implemented.
4040 New Targets and Target Specific Improvements
4044 * Full cross-toolchain support now available with GNU Binutils
4048 * GCC now supports the Cortex-M0 and Cortex-A5 processors.
4049 * GCC now supports the ARM v7E-M architecture.
4050 * GCC now supports VFPv4-based FPUs and FPUs with
4051 single-precision-only VFP.
4052 * GCC has many improvements to optimization for other ARM processors,
4053 including scheduling support for the integer pipeline on Cortex-A9.
4054 * GCC now supports the IEEE 754-2008 half-precision floating-point
4055 type, and a variant ARM-specific half-precision type. This type is
4056 specified using __fp16, with the layout determined by
4057 -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and
4058 VFPv4 half-precision instructions will be used.
4059 * GCC now supports the variant of AAPCS that uses VFP registers for
4060 parameter passing and return values.
4064 * The -mno-tablejump option has been removed because it has the same
4065 effect as the -fno-jump-tables option.
4066 * Added support for these new AVR devices:
4073 * GCC now will set the default for -march= based on the configure
4075 * GCC now supports handling floating-point excess precision arising
4076 from use of the x87 floating-point unit in a way that conforms to
4077 ISO C99. This is enabled with -fexcess-precision=standard and with
4078 standards conformance options such as -std=c99, and may be disabled
4079 using -fexcess-precision=fast.
4080 * Support for the Intel Atom processor is now available through the
4081 -march=atom and -mtune=atom options.
4082 * A new -mcrc32 option is now available to enable crc32 intrinsics.
4083 * A new -mmovbe option is now available to enable GCC to use the
4084 movbe instruction to implement __builtin_bswap32 and
4086 * SSE math now can be enabled by default at configure time with the
4087 new --with-fpmath=sse option.
4088 * There is a new intrinsic header file, <x86intrin.h>. It should be
4089 included before using any IA-32/x86-64 intrinsics.
4090 * Support for the XOP, FMA4, and LWP instruction sets for the AMD
4091 Orochi processors are now available with the -mxop, -mfma4, and
4093 * The -mabm option enables GCC to use the popcnt and lzcnt
4094 instructions on AMD processors.
4095 * The -mpopcnt option enables GCC to use the popcnt instructions on
4096 both AMD and Intel processors.
4100 * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x
4102 * GCC now supports thread-local storage (TLS) on M68K and ColdFire
4107 Support has been added for the Toshiba Media embedded Processor (MeP,
4108 or mep-elf) embedded target.
4112 * GCC now supports MIPS 1004K processors.
4113 * GCC can now be configured with options --with-arch-32,
4114 --with-arch-64, --with-tune-32 and --with-tune-64 to control the
4115 default optimization separately for 32-bit and 64-bit modes.
4116 * MIPS targets now support an alternative _mcount interface, in which
4117 register $12 points to the function's save slot for register $31.
4118 This interface is selected by the -mcount-ra-address option; see
4119 the documentation for more details.
4120 * GNU/Linux targets can now generate read-only .eh_frame sections.
4121 This optimization requires GNU binutils 2.20 or above, and is only
4122 available if GCC is configured with a suitable version of binutils.
4123 * GNU/Linux targets can now attach special relocations to indirect
4124 calls, so that the linker can turn them into direct jumps or
4125 branches. This optimization requires GNU binutils 2.20 or later,
4126 and is automatically selected if GCC is configured with an
4127 appropriate version of binutils. It can be explicitly enabled or
4128 disabled using the -mrelax-pic-calls command-line option.
4129 * GCC now generates more heavily-optimized atomic operations on
4131 * MIPS targets now support the -fstack-protector option.
4132 * GCC now supports an -msynci option, which specifies that synci is
4133 enough to flush the instruction cache, without help from the
4134 operating system. GCC uses this information to optimize
4135 automatically-generated cache flush operations, such as those used
4136 for nested functions in C. There is also a --with-synci
4137 configure-time option, which makes -msynci the default.
4138 * GCC supports four new function attributes for interrupt handlers:
4139 interrupt, use_shadow_register_set, keep_interrupts_masked and
4140 use_debug_exception_return. See the documentation for more details
4141 about these attributes.
4143 RS/6000 (POWER/PowerPC)
4145 * GCC now supports the Power ISA 2.06, which includes the VSX
4146 instructions that add vector 64-bit floating point support, new
4147 population count instructions, and conversions between floating
4148 point and unsigned types.
4149 * Support for the power7 processor is now available through the
4150 -mcpu=power7 and -mtune=power7.
4151 * GCC will now vectorize loops that contain simple math functions
4152 like copysign when generating code for altivec or VSX targets.
4153 * Support for the A2 processor is now available through the -mcpu=a2
4154 and -mtune=a2 options.
4155 * Support for the 476 processor is now available through the
4156 -mcpu={476,476fp} and -mtune={476,476fp} options.
4157 * Support for the e500mc64 processor is now available through the
4158 -mcpu=e500mc64 and -mtune=e500mc64 options.
4159 * GCC can now be configured with options --with-cpu-32,
4160 --with-cpu-64, --with-tune-32 and --with-tune-64 to control the
4161 default optimization separately for 32-bit and 64-bit modes.
4162 * Starting with GCC 4.5.4, vectors of type vector long long or vector
4163 long are passed and returned in the same method as other vectors
4164 with the VSX instruction set. Previously the GCC compiler did not
4165 adhere to the ABI for 128-bit vectors with 64-bit integer base
4166 types (PR 48857). This is also fixed in the GCC 4.6.1 release.
4170 Support has been added for the Renesas RX Processor (rx-elf) target.
4174 Windows (Cygwin and MinGW)
4176 * GCC now installs all the major language runtime libraries as DLLs
4177 when configured with the --enable-shared option.
4178 * GCC now makes use of the new support for aligned common variables
4179 in versions of binutils >= 2.20 to fix bugs in the support for SSE
4181 * Improvements to the libffi support library increase the reliability
4182 of code generated by GCJ on all Windows platforms. Libgcj is
4183 enabled by default for the first time.
4184 * Libtool improvements simplify installation by placing the generated
4185 DLLs in the correct binaries directory.
4186 * Numerous other minor bugfixes and improvements, and substantial
4187 enhancements to the Fortran language support library.
4191 Other significant improvements
4195 * It is now possible to extend the compiler without having to modify
4196 its source code. A new option -fplugin=file.so tells GCC to load
4197 the shared object file.so and execute it as part of the compiler.
4198 The internal documentation describes the details on how plugins can
4199 interact with the compiler.
4201 Installation changes
4203 * The move to newer autotools changed default installation
4204 directories and switches to control them: The --with-datarootdir,
4205 --with-docdir, --with-pdfdir, and --with-htmldir switches are not
4206 used any more. Instead, you can now use --datarootdir, --docdir,
4207 --htmldir, and --pdfdir. The default installation directories have
4208 changed as follows according to the GNU Coding Standards:
4210 datarootdir read-only architecture-independent data root [PREFIX/share]
4211 localedir locale-specific message catalogs [DATAROOTDIR/locale]
4212 docdir documentation root [DATAROOTDIR/doc/PACKAGE]
4213 htmldir html documentation [DOCDIR]
4214 dvidir dvi documentation [DOCDIR]
4215 pdfdir pdf documentation [DOCDIR]
4216 psdir ps documentation [DOCDIR]
4217 The following variables have new default values:
4219 datadir read-only architecture-independent data [DATAROOTDIR]
4220 infodir info documentation [DATAROOTDIR/info]
4221 mandir man documentation [DATAROOTDIR/man]
4225 This is the [23]list of problem reports (PRs) from GCC's bug tracking
4226 system that are known to be fixed in the 4.5.1 release. This list might
4227 not be complete (that is, it is possible that some PRs that have been
4228 fixed are not listed here).
4232 * GCC's new link-time optimizer ([24]-flto) now also works on a few
4234 + Cygwin (*-cygwin*)
4236 + Darwin on x86-64 (x86_64-apple-darwin*)
4237 LTO is not enabled by default for these targets. To enable LTO, you
4238 should configure with the --enable-lto option.
4242 This is the [25]list of problem reports (PRs) from GCC's bug tracking
4243 system that are known to be fixed in the 4.5.2 release. This list might
4244 not be complete (that is, it is possible that some PRs that have been
4245 fixed are not listed here).
4249 This is the [26]list of problem reports (PRs) from GCC's bug tracking
4250 system that are known to be fixed in the 4.5.3 release. This list might
4251 not be complete (that is, it is possible that some PRs that have been
4252 fixed are not listed here).
4254 On the PowerPC compiler, the Altivec builtin functions vec_ld and
4255 vec_st have been modified to generate the Altivec memory instructions
4256 LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5
4257 release, these builtin functions were changed to generate VSX memory
4258 reference instructions instead of Altivec memory instructions, but
4259 there are differences between the two instructions. If the VSX
4260 instruction set is available, you can now use the new builtin functions
4261 vec_vsx_ld and vec_vsx_st which always generates the VSX memory
4266 This is the [27]list of problem reports (PRs) from GCC's bug tracking
4267 system that are known to be fixed in the 4.5.4 release. This list might
4268 not be complete (that is, it is possible that some PRs that have been
4269 fixed are not listed here).
4272 For questions related to the use of GCC, please consult these web
4273 pages and the [28]GCC manuals. If that fails, the
4274 [29]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4275 web pages and the development of GCC are welcome on our developer
4276 list at [30]gcc@gcc.gnu.org. All of [31]our lists have public
4279 Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and
4280 distribution of this entire article is permitted in any medium,
4281 provided this notice is preserved.
4283 These pages are [33]maintained by the GCC team. Last modified
4288 1. http://www.multiprecision.org/
4289 2. https://gcc.gnu.org/install/prerequisites.html
4290 3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html
4291 4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted
4292 5. http://gcc.gnu.org/gcc-4.5/changes.html#x86
4293 6. http://www.multiprecision.org/
4294 7. https://gcc.gnu.org/PR30789
4295 8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
4296 9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802
4297 10. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800
4298 11. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html
4299 12. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html
4300 13. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757
4301 14. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176
4302 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html
4303 16. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733
4304 17. http://sourceware.org/gdb/wiki/STLSupport
4305 18. http://gcc.gnu.org/gcc-4.4/changes.html
4306 19. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
4307 20. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html
4308 21. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html
4309 22. https://gcc.gnu.org/wiki/OOP
4310 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1
4311 24. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
4312 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2
4313 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3
4314 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4
4315 28. https://gcc.gnu.org/onlinedocs/
4316 29. mailto:gcc-help@gcc.gnu.org
4317 30. mailto:gcc@gcc.gnu.org
4318 31. https://gcc.gnu.org/lists.html
4319 32. http://www.fsf.org/
4320 33. https://gcc.gnu.org/about.html
4321 34. http://validator.w3.org/check/referer
4322 ======================================================================
4323 http://gcc.gnu.org/gcc-4.4/index.html
4324 GCC 4.4 Release Series
4328 The [1]GNU project and the GCC developers are pleased to announce the
4329 release of GCC 4.4.7.
4331 This release is a bug-fix release, containing fixes for regressions in
4332 GCC 4.4.6 relative to previous releases of GCC.
4337 March 13, 2012 ([2]changes)
4340 April 16, 2011 ([3]changes)
4343 October 1, 2010 ([4]changes)
4346 April 29, 2010 ([5]changes)
4349 January 21, 2010 ([6]changes)
4352 October 15, 2009 ([7]changes)
4355 July 22, 2009 ([8]changes)
4358 April 21, 2009 ([9]changes)
4360 References and Acknowledgements
4362 GCC used to stand for the GNU C Compiler, but since the compiler
4363 supports several other languages aside from C, it now stands for the
4364 GNU Compiler Collection.
4366 A list of [10]successful builds is updated as new information becomes
4369 The GCC developers would like to thank the numerous people that have
4370 contributed new features, improvements, bug fixes, and other changes as
4371 well as test results to GCC. This [11]amazing group of volunteers is
4372 what makes GCC successful.
4374 For additional information about GCC please refer to the [12]GCC
4375 project web site or contact the [13]GCC development mailing list.
4377 To obtain GCC please use [14]our mirror sites or [15]our SVN server.
4380 For questions related to the use of GCC, please consult these web
4381 pages and the [16]GCC manuals. If that fails, the
4382 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4383 web pages and the development of GCC are welcome on our developer
4384 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
4387 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
4388 distribution of this entire article is permitted in any medium,
4389 provided this notice is preserved.
4391 These pages are [21]maintained by the GCC team. Last modified
4396 1. http://www.gnu.org/
4397 2. http://gcc.gnu.org/gcc-4.4/changes.html
4398 3. http://gcc.gnu.org/gcc-4.4/changes.html
4399 4. http://gcc.gnu.org/gcc-4.4/changes.html
4400 5. http://gcc.gnu.org/gcc-4.4/changes.html
4401 6. http://gcc.gnu.org/gcc-4.4/changes.html
4402 7. http://gcc.gnu.org/gcc-4.4/changes.html
4403 8. http://gcc.gnu.org/gcc-4.4/changes.html
4404 9. http://gcc.gnu.org/gcc-4.4/changes.html
4405 10. http://gcc.gnu.org/gcc-4.4/buildstat.html
4406 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
4407 12. http://gcc.gnu.org/index.html
4408 13. mailto:gcc@gcc.gnu.org
4409 14. http://gcc.gnu.org/mirrors.html
4410 15. http://gcc.gnu.org/svn.html
4411 16. https://gcc.gnu.org/onlinedocs/
4412 17. mailto:gcc-help@gcc.gnu.org
4413 18. mailto:gcc@gcc.gnu.org
4414 19. https://gcc.gnu.org/lists.html
4415 20. http://www.fsf.org/
4416 21. https://gcc.gnu.org/about.html
4417 22. http://validator.w3.org/check/referer
4418 ======================================================================
4419 http://gcc.gnu.org/gcc-4.4/changes.html
4420 GCC 4.4 Release Series
4421 Changes, New Features, and Fixes
4423 The latest release in the 4.4 release series is [1]GCC 4.4.7.
4427 * __builtin_stdarg_start has been completely removed from GCC.
4428 Support for <varargs.h> had been deprecated since GCC 4.0. Use
4429 __builtin_va_start as a replacement.
4430 * Some of the errors issued by the C++ front end that could be
4431 downgraded to warnings in previous releases by using -fpermissive
4432 are now warnings by default. They can be converted into errors by
4433 using -pedantic-errors.
4434 * Use of the cpp assertion extension will now emit a warning when
4435 -Wdeprecated or -pedantic is used. This extension has been
4436 deprecated for many years, but never warned about.
4437 * Packed bit-fields of type char were not properly bit-packed on many
4438 targets prior to GCC 4.4. On these targets, the fix in GCC 4.4
4439 causes an ABI change. For example there is no longer a 4-bit
4440 padding between field a and b in this structure:
4445 } __attribute__ ((packed));
4446 There is a new warning to help identify fields that are affected:
4447 foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4
4448 The warning can be disabled with -Wno-packed-bitfield-compat.
4449 * On ARM EABI targets, the C++ mangling of the va_list type has been
4450 changed to conform to the current revision of the EABI. This does
4451 not affect the libstdc++ library included with GCC.
4452 * The SCOUNT and POS bits of the MIPS DSP control register are now
4453 treated as global. Previous versions of GCC treated these fields as
4454 call-clobbered instead.
4455 * The MIPS port no longer recognizes the h asm constraint. It was
4456 necessary to remove this constraint in order to avoid generating
4457 unpredictable code sequences.
4458 One of the main uses of the h constraint was to extract the high
4459 part of a multiplication on 64-bit targets. For example:
4460 asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y));
4461 You can now achieve the same effect using 128-bit types:
4462 typedef unsigned int uint128_t __attribute__((mode(TI)));
4463 result = ((uint128_t) x * y) >> 64;
4464 The second sequence is better in many ways. For example, if x and y
4465 are constants, the compiler can perform the multiplication at
4466 compile time. If x and y are not constants, the compiler can
4467 schedule the runtime multiplication better than it can schedule an
4469 * Support for a number of older systems and recently unmaintained or
4470 untested target ports of GCC has been declared obsolete in GCC 4.4.
4471 Unless there is activity to revive them, the next release of GCC
4472 will have their sources permanently removed.
4473 The following ports for individual systems on particular
4474 architectures have been obsoleted:
4475 + Generic a.out on IA32 and m68k (i[34567]86-*-aout*,
4477 + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*,
4478 armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*,
4479 sh-*-*). This does not affect other more specific targets
4480 using the COFF object format on those architectures, or the
4481 more specific H8300 and SH targets (h8300-*-rtems*,
4482 h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*,
4483 sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks).
4484 + 2BSD on PDP-11 (pdp11-*-bsd)
4485 + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*,
4486 powerpc-ibm-aix4.[12]*)
4487 + Tuning support for Itanium1 (Merced) variants. Note that code
4488 tuned for Itanium2 should also run correctly on Itanium1.
4489 * The protoize and unprotoize utilities have been obsoleted and will
4490 be removed in GCC 4.5. These utilities have not been installed by
4491 default since GCC 3.0.
4492 * Support has been removed for all the [2]configurations obsoleted in
4494 * Unknown -Wno-* options are now silently ignored by GCC if no other
4495 diagnostics are issued. If other diagnostics are issued, then GCC
4496 warns about the unknown options.
4497 * More information on porting to GCC 4.4 from previous versions of
4498 GCC can be found in the [3]porting guide for this release.
4500 General Optimizer Improvements
4502 * A new command-line switch -findirect-inlining has been added. When
4503 turned on it allows the inliner to also inline indirect calls that
4504 are discovered to have known targets at compile time thanks to
4506 * A new command-line switch -ftree-switch-conversion has been added.
4507 This new pass turns simple initializations of scalar variables in
4508 switch statements into initializations from a static array, given
4509 that all the values are known at compile time and the ratio between
4510 the new array size and the original switch branches does not exceed
4511 the parameter --param switch-conversion-max-branch-ratio (default
4513 * A new command-line switch -ftree-builtin-call-dce has been added.
4514 This optimization eliminates unnecessary calls to certain builtin
4515 functions when the return value is not used, in cases where the
4516 calls can not be eliminated entirely because the function may set
4517 errno. This optimization is on by default at -O2 and above.
4518 * A new command-line switch -fconserve-stack directs the compiler to
4519 minimize stack usage even if it makes the generated code slower.
4520 This affects inlining decisions.
4521 * When the assembler supports it, the compiler will now emit unwind
4522 information using assembler .cfi directives. This makes it possible
4523 to use such directives in inline assembler code. The new option
4524 -fno-dwarf2-cfi-asm directs the compiler to not use .cfi
4526 * The [4]Graphite branch has been merged. This merge has brought in a
4527 new framework for loop optimizations based on a polyhedral
4528 intermediate representation. These optimizations apply to all the
4529 languages supported by GCC. The following new code transformations
4530 are available in GCC 4.4:
4531 + -floop-interchange performs loop interchange transformations
4532 on loops. Interchanging two nested loops switches the inner
4533 and outer loops. For example, given a loop like:
4536 A(J, I) = A(J, I) * C
4540 loop interchange will transform the loop as if the user had
4544 A(J, I) = A(J, I) * C
4548 which can be beneficial when N is larger than the caches,
4549 because in Fortran, the elements of an array are stored in
4550 memory contiguously by column, and the original loop iterates
4551 over rows, potentially creating at each access a cache miss.
4552 + -floop-strip-mine performs loop strip mining transformations
4553 on loops. Strip mining splits a loop into two nested loops.
4554 The outer loop has strides equal to the strip size and the
4555 inner loop has strides of the original loop within a strip.
4556 For example, given a loop like:
4561 loop strip mining will transform the loop as if the user had
4564 DO I = II, min (II + 3, N)
4569 + -floop-block performs loop blocking transformations on loops.
4570 Blocking strip mines each loop in the loop nest such that the
4571 memory accesses of the element loops fit inside caches. For
4572 example, given a loop like:
4575 A(J, I) = B(I) + C(J)
4579 loop blocking will transform the loop as if the user had
4583 DO I = II, min (II + 63, N)
4584 DO J = JJ, min (JJ + 63, M)
4585 A(J, I) = B(I) + C(J)
4591 which can be beneficial when M is larger than the caches,
4592 because the innermost loop will iterate over a smaller amount
4593 of data that can be kept in the caches.
4594 * A new register allocator has replaced the old one. It is called
4595 integrated register allocator (IRA) because coalescing, register
4596 live range splitting, and hard register preferencing are done
4597 on-the-fly during coloring. It also has better integration with the
4598 reload pass. IRA is a regional register allocator which uses modern
4599 Chaitin-Briggs coloring instead of Chow's priority coloring used in
4600 the old register allocator. More info about IRA internals and
4601 options can be found in the GCC manuals.
4602 * A new instruction scheduler and software pipeliner, based on the
4603 selective scheduling approach, has been added. The new pass
4604 performs instruction unification, register renaming, substitution
4605 through register copies, and speculation during scheduling. The
4606 software pipeliner is able to pipeline non-countable loops. The new
4607 pass is targeted at scheduling-eager in-order platforms. In GCC 4.4
4608 it is available for the Intel Itanium platform working by default
4609 as the second scheduling pass (after register allocation) at the
4610 -O3 optimization level.
4611 * When using -fprofile-generate with a multi-threaded program, the
4612 profile counts may be slightly wrong due to race conditions. The
4613 new -fprofile-correction option directs the compiler to apply
4614 heuristics to smooth out the inconsistencies. By default the
4615 compiler will give an error message when it finds an inconsistent
4617 * The new -fprofile-dir=PATH option permits setting the directory
4618 where profile data files are stored when using -fprofile-generate
4619 and friends, and the directory used when reading profile data files
4620 using -fprofile-use and friends.
4624 * The new -Wframe-larger-than=NUMBER option directs GCC to emit a
4625 warning if any stack frame is larger than NUMBER bytes. This may be
4626 used to help ensure that code fits within a limited amount of stack
4628 * The command-line option -Wlarger-than-N is now written as
4629 -Wlarger-than=N and the old form is deprecated.
4630 * The new -Wno-mudflap option disables warnings about constructs
4631 which can not be instrumented when using -fmudflap.
4633 New Languages and Language specific improvements
4635 * Version 3.0 of the [5]OpenMP specification is now supported for the
4636 C, C++, and Fortran compilers.
4637 * New character data types, per [6]TR 19769: New character types in
4638 C, are now supported for the C compiler in -std=gnu99 mode, as
4639 __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in
4640 -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too.
4644 * A new optimize attribute was added to allow programmers to change
4645 the optimization level and particular optimization options for an
4646 individual function. You can also change the optimization options
4647 via the GCC optimize pragma for functions defined after the pragma.
4648 The GCC push_options pragma and the GCC pop_options pragma allow
4649 you temporarily save and restore the options used. The GCC
4650 reset_options pragma restores the options to what was specified on
4652 * Uninitialized warnings do not require enabling optimization
4653 anymore, that is, -Wuninitialized can be used together with -O0.
4654 Nonetheless, the warnings given by -Wuninitialized will probably be
4655 more accurate if optimization is enabled.
4656 * -Wparentheses now warns about expressions such as (!x | y) and (!x
4657 & y). Using explicit parentheses, such as in ((!x) | y), silences
4659 * -Wsequence-point now warns within if, while,do while and for
4660 conditions, and within for begin/end expressions.
4661 * A new option -dU is available to dump definitions of preprocessor
4662 macros that are tested or expanded.
4666 * [7]Improved experimental support for the upcoming ISO C++ standard,
4667 C++0x. Including support for auto, inline namespaces, generalized
4668 initializer lists, defaulted and deleted functions, new character
4669 types, and scoped enums.
4670 * Those errors that may be downgraded to warnings to build legacy
4671 code now mention -fpermissive when -fdiagnostics-show-option is
4673 * -Wconversion now warns if the result of a static_cast to enumeral
4674 type is unspecified because the value is outside the range of the
4676 * -Wuninitialized now warns if a non-static reference or non-static
4677 const member appears in a class without constructors.
4678 * G++ now properly implements value-initialization, so objects with
4679 an initializer of () and an implicitly defined default constructor
4680 will be zero-initialized before the default constructor is called.
4682 Runtime Library (libstdc++)
4684 * Improved experimental support for the upcoming ISO C++ standard,
4686 + Support for <chrono>, <condition_variable>, <cstdatomic>,
4687 <forward_list>, <initializer_list>, <mutex>, <ratio>,
4688 <system_error>, and <thread>.
4689 + unique_ptr, <algorithm> additions, exception propagation, and
4690 support for the new character types in <string> and <limits>.
4691 + Existing facilities now exploit initializer lists, defaulted
4692 and deleted functions, and the newly implemented core C++0x
4694 + Some standard containers are more efficient together with
4695 stateful allocators, i.e., no allocator is constructed on the
4696 fly at element construction time.
4697 * Experimental support for non-standard pointer types in containers.
4698 * The long standing libstdc++/30928 has been fixed for targets
4699 running glibc 2.10 or later.
4700 * As usual, many small and larger bug fixes, in particular quite a
4701 few corner cases in <locale>.
4705 * GNU Fortran now employs libcpp directly instead of using cc1 as an
4706 external preprocessor. The [8]-cpp option was added to allow manual
4707 invocation of the preprocessor without relying on filename
4709 * The [9]-Warray-temporaries option warns about array temporaries
4710 generated by the compiler, as an aid to optimization.
4711 * The [10]-fcheck-array-temporaries option has been added, printing a
4712 notification at run time, when an array temporary had to be created
4713 for an function argument. Contrary to -Warray-temporaries the
4714 warning is only printed if the array is noncontiguous.
4715 * Improved generation of DWARF debugging symbols
4716 * If using an intrinsic not part of the selected standard (via -std=
4717 and -fall-intrinsics) gfortran will now treat it as if this
4718 procedure were declared EXTERNAL and try to link to a user-supplied
4719 procedure. -Wintrinsics-std will warn whenever this happens. The
4720 now-useless option -Wnonstd-intrinsic was removed.
4721 * The flag -falign-commons has been added to control the alignment of
4722 variables in COMMON blocks, which is enabled by default in line
4723 with previous GCC version. Using -fno-align-commons one can force
4724 commons to be contiguous in memory as required by the Fortran
4725 standard, however, this slows down the memory access. The option
4726 -Walign-commons, which is enabled by default, warns when padding
4727 bytes were added for alignment. The proper solution is to sort the
4728 common objects by decreasing storage size, which avoids the
4730 * Fortran 2003 support has been extended:
4731 + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is
4732 now supported (except internal reads from/writes to wide
4733 strings). [11]-fbackslash now supports also \unnnn and
4734 \Unnnnnnnn to enter Unicode characters.
4735 + Asynchronous I/O (implemented as synchronous I/O) and the
4736 decimal=, size=, sign=, pad=, blank=, and delim= specifiers
4737 are now supported in I/O statements.
4738 + Support for Fortran 2003 structure constructors and for array
4739 constructor with typespec has been added.
4740 + Procedure Pointers (but not yet as component in derived types
4741 and as function results) are now supported.
4742 + Abstract types, type extension, and type-bound procedures
4743 (both PROCEDURE and GENERIC but not as operators). Note: As
4744 CLASS/polymorphyic types are not implemented, type-bound
4745 procedures with PASS accept as non-standard extension TYPE
4747 * Fortran 2008 support has been added:
4748 + The -std=f2008 option and support for the file extensions
4749 .f2008 and .F2008 has been added.
4750 + The g0 format descriptor is now supported.
4751 + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH,
4752 ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED
4753 are now available (some of them existed as GNU extension
4754 before). Note: The hyperbolic functions are not yet supporting
4755 complex arguments and the three- argument version of BESSEL_*N
4757 + The bit intrinsics LEADZ and TRAILZ have been added.
4763 * The Ada runtime now supports multilibs on many platforms including
4764 x86_64, SPARC and PowerPC. Their build is enabled by default.
4766 New Targets and Target Specific Improvements
4770 * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and
4771 Cortex-R4F processors and has many other improvements to
4772 optimization for ARM processors.
4773 * GCC now supports the VFPv3 variant with 16 double-precision
4774 registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been
4775 renamed to -mfpu=vfpv3.
4776 * GCC now supports the -mfix-cortex-m3-ldrd option to work around an
4777 erratum on Cortex-M3 processors.
4778 * GCC now supports the __sync_* atomic operations for ARM EABI
4780 * The section anchors optimization is now enabled by default when
4782 * GCC now uses a new EABI-compatible profiling interface for EABI
4783 targets. This requires a function __gnu_mcount_nc, which is
4784 provided by GNU libc versions 2.8 and later.
4788 * The -mno-tablejump option has been deprecated because it has the
4789 same effect as the -fno-jump-tables option.
4790 * Added support for these new AVR devices:
4819 * Support for Intel AES built-in functions and code generation is
4820 available via -maes.
4821 * Support for Intel PCLMUL built-in function and code generation is
4822 available via -mpclmul.
4823 * Support for Intel AVX built-in functions and code generation is
4824 available via -mavx.
4825 * Automatically align the stack for local variables with alignment
4827 * GCC can now utilize the SVML library for vectorizing calls to a set
4828 of C99 functions if -mveclibabi=svml is specified and you link to
4829 an SVML ABI compatible library.
4830 * On x86-64, the ABI has been changed in the following cases to
4831 conform to the x86-64 ABI:
4832 + Passing/returning structures with flexible array member:
4838 + Passing/returning structures with complex float member:
4842 __complex__ float f;
4844 + Passing/returning unions with long double member:
4850 Code built with previous versions of GCC that uses any of these is
4851 not compatible with code built with GCC 4.4.0 or later.
4852 * A new target attribute was added to allow programmers to change the
4853 target options like -msse2 or -march=k8 for an individual function.
4854 You can also change the target options via the GCC target pragma
4855 for functions defined after the pragma.
4856 * GCC can now be configured with options --with-arch-32,
4857 --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and
4858 --with-tune-64 to control the default optimization separately for
4859 32-bit and 64-bit modes.
4863 * Support for __float128 (TFmode) IEEE quad type and corresponding
4864 TCmode IEEE complex quad type is available via the soft-fp library
4865 on IA-32/IA64 targets. This includes basic arithmetic operations
4866 (addition, subtraction, negation, multiplication and division) on
4867 __float128 real and TCmode complex values, the full set of IEEE
4868 comparisons between __float128 values, conversions to and from
4869 float, double and long double floating point types, as well as
4870 conversions to and from signed or unsigned integer, signed or
4871 unsigned long integer and signed or unsigned quad (TImode, IA64
4872 only) integer types. Additionally, all operations generate the full
4873 set of IEEE exceptions and support the full set of IEEE rounding
4878 * GCC now supports instruction scheduling for ColdFire V1, V3 and V4
4879 processors. (Scheduling support for ColdFire V2 processors was
4881 * GCC now supports the -mxgot option to support programs requiring
4882 many GOT entries on ColdFire.
4883 * The m68k-*-linux-gnu target now builds multilibs by default.
4887 * MIPS Technologies have extended the original MIPS SVR4 ABI to
4888 include support for procedure linkage tables (PLTs) and copy
4889 relocations. These extensions allow GNU/Linux executables to use a
4890 significantly more efficient code model than the one defined by the
4892 GCC support for this code model is available via a new command-line
4893 option, -mplt. There is also a new configure-time option,
4894 --with-mips-plt, to make -mplt the default.
4895 The new code model requires support from the assembler, the linker,
4896 and the runtime C library. This support is available in binutils
4898 * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables
4899 and 32-bit GNU/Linux shared libraries. This feature requires GNU
4900 binutils 2.19 or above.
4901 * Support for RMI's XLR processor is now available through the
4902 -march=xlr and -mtune=xlr options.
4903 * 64-bit targets can now perform 128-bit multiplications inline,
4904 instead of relying on a libgcc function.
4905 * Native GNU/Linux toolchains now support -march=native and
4906 -mtune=native, which select the host processor.
4907 * GCC now supports the R10K, R12K, R14K and R16K processors. The
4908 canonical -march= and -mtune= names for these processors are
4909 r10000, r12000, r14000 and r16000 respectively.
4910 * GCC can now work around the side effects of speculative execution
4911 on R10K processors. Please see the documentation of the
4912 -mr10k-cache-barrier option for details.
4913 * Support for the MIPS64 Release 2 instruction set has been added.
4914 The option -march=mips64r2 enables generation of these
4916 * GCC now supports Cavium Networks' Octeon processor. This support is
4917 available through the -march=octeon and -mtune=octeon options.
4918 * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The
4919 canonical -march= and -mtune= names for these processors are
4920 loongson2e and loongson2f.
4924 Picochip is a 16-bit processor. A typical picoChip contains over 250
4925 small cores, each with small amounts of memory. There are three
4926 processor variants (STAN, MEM and CTRL) with different instruction sets
4927 and memory configurations and they can be chosen using the -mae option.
4929 This port is intended to be a "C" only port.
4931 Power Architecture and PowerPC
4933 * GCC now supports the e300c2, e300c3 and e500mc processors.
4934 * GCC now supports Xilinx processors with a single-precision FPU.
4935 * Decimal floating point is now supported for e500 processors.
4937 S/390, zSeries and System z9/z10
4939 * Support for the IBM System z10 EC/BC processor has been added. When
4940 using the -march=z10 option, the compiler will generate code making
4941 use of instructions provided by the General-Instruction-Extension
4942 Facility and the Execute-Extension Facility.
4946 * GCC now supports the thread-local storage mechanism used on
4951 * GCC now supports thread-local storage (TLS) for Xtensa processor
4952 configurations that include the Thread Pointer option. TLS also
4953 requires support from the assembler and linker; this support is
4954 provided in the GNU binutils beginning with version 2.19.
4956 Documentation improvements
4958 Other significant improvements
4962 This is the [12]list of problem reports (PRs) from GCC's bug tracking
4963 system that are known to be fixed in the 4.4.1 release. This list might
4964 not be complete (that is, it is possible that some PRs that have been
4965 fixed are not listed here).
4969 This is the [13]list of problem reports (PRs) from GCC's bug tracking
4970 system that are known to be fixed in the 4.4.2 release. This list might
4971 not be complete (that is, it is possible that some PRs that have been
4972 fixed are not listed here).
4976 This is the [14]list of problem reports (PRs) from GCC's bug tracking
4977 system that are known to be fixed in the 4.4.3 release. This list might
4978 not be complete (that is, it is possible that some PRs that have been
4979 fixed are not listed here).
4983 This is the [15]list of problem reports (PRs) from GCC's bug tracking
4984 system that are known to be fixed in the 4.4.4 release. This list might
4985 not be complete (that is, it is possible that some PRs that have been
4986 fixed are not listed here).
4990 This is the [16]list of problem reports (PRs) from GCC's bug tracking
4991 system that are known to be fixed in the 4.4.5 release. This list might
4992 not be complete (that is, it is possible that some PRs that have been
4993 fixed are not listed here).
4997 This is the [17]list of problem reports (PRs) from GCC's bug tracking
4998 system that are known to be fixed in the 4.4.6 release. This list might
4999 not be complete (that is, it is possible that some PRs that have been
5000 fixed are not listed here).
5004 This is the [18]list of problem reports (PRs) from GCC's bug tracking
5005 system that are known to be fixed in the 4.4.7 release. This list might
5006 not be complete (that is, it is possible that some PRs that have been
5007 fixed are not listed here).
5010 For questions related to the use of GCC, please consult these web
5011 pages and the [19]GCC manuals. If that fails, the
5012 [20]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5013 web pages and the development of GCC are welcome on our developer
5014 list at [21]gcc@gcc.gnu.org. All of [22]our lists have public
5017 Copyright (C) [23]Free Software Foundation, Inc. Verbatim copying and
5018 distribution of this entire article is permitted in any medium,
5019 provided this notice is preserved.
5021 These pages are [24]maintained by the GCC team. Last modified
5026 1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7
5027 2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted
5028 3. http://gcc.gnu.org/gcc-4.4/porting_to.html
5029 4. https://gcc.gnu.org/wiki/Graphite
5030 5. http://openmp.org/wp/openmp-specifications/
5031 6. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf
5032 7. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html
5033 8. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html
5034 9. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125
5035 10. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221
5036 11. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34
5037 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1
5038 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2
5039 14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3
5040 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4
5041 16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5
5042 17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6
5043 18. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7
5044 19. https://gcc.gnu.org/onlinedocs/
5045 20. mailto:gcc-help@gcc.gnu.org
5046 21. mailto:gcc@gcc.gnu.org
5047 22. https://gcc.gnu.org/lists.html
5048 23. http://www.fsf.org/
5049 24. https://gcc.gnu.org/about.html
5050 25. http://validator.w3.org/check/referer
5051 ======================================================================
5052 http://gcc.gnu.org/gcc-4.3/index.html
5053 GCC 4.3 Release Series
5057 The [1]GNU project and the GCC developers are pleased to announce the
5058 release of GCC 4.3.6.
5060 This release is a bug-fix release, containing fixes for regressions in
5061 GCC 4.3.5 relative to previous releases of GCC.
5066 Jun 27, 2011 ([2]changes)
5069 May 22, 2010 ([3]changes)
5072 August 4, 2009 ([4]changes)
5075 January 24, 2009 ([5]changes)
5078 August 27, 2008 ([6]changes)
5081 June 6, 2008 ([7]changes)
5084 March 5, 2008 ([8]changes)
5086 References and Acknowledgements
5088 GCC used to stand for the GNU C Compiler, but since the compiler
5089 supports several other languages aside from C, it now stands for the
5090 GNU Compiler Collection.
5092 A list of [9]successful builds is updated as new information becomes
5095 The GCC developers would like to thank the numerous people that have
5096 contributed new features, improvements, bug fixes, and other changes as
5097 well as test results to GCC. This [10]amazing group of volunteers is
5098 what makes GCC successful.
5100 For additional information about GCC please refer to the [11]GCC
5101 project web site or contact the [12]GCC development mailing list.
5103 To obtain GCC please use [13]our mirror sites or [14]our SVN server.
5106 For questions related to the use of GCC, please consult these web
5107 pages and the [15]GCC manuals. If that fails, the
5108 [16]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5109 web pages and the development of GCC are welcome on our developer
5110 list at [17]gcc@gcc.gnu.org. All of [18]our lists have public
5113 Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and
5114 distribution of this entire article is permitted in any medium,
5115 provided this notice is preserved.
5117 These pages are [20]maintained by the GCC team. Last modified
5122 1. http://www.gnu.org/
5123 2. http://gcc.gnu.org/gcc-4.3/changes.html
5124 3. http://gcc.gnu.org/gcc-4.3/changes.html
5125 4. http://gcc.gnu.org/gcc-4.3/changes.html
5126 5. http://gcc.gnu.org/gcc-4.3/changes.html
5127 6. http://gcc.gnu.org/gcc-4.3/changes.html
5128 7. http://gcc.gnu.org/gcc-4.3/changes.html
5129 8. http://gcc.gnu.org/gcc-4.3/changes.html
5130 9. http://gcc.gnu.org/gcc-4.3/buildstat.html
5131 10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
5132 11. http://gcc.gnu.org/index.html
5133 12. mailto:gcc@gcc.gnu.org
5134 13. http://gcc.gnu.org/mirrors.html
5135 14. http://gcc.gnu.org/svn.html
5136 15. https://gcc.gnu.org/onlinedocs/
5137 16. mailto:gcc-help@gcc.gnu.org
5138 17. mailto:gcc@gcc.gnu.org
5139 18. https://gcc.gnu.org/lists.html
5140 19. http://www.fsf.org/
5141 20. https://gcc.gnu.org/about.html
5142 21. http://validator.w3.org/check/referer
5143 ======================================================================
5144 http://gcc.gnu.org/gcc-4.3/changes.html
5145 GCC 4.3 Release Series
5146 Changes, New Features, and Fixes
5148 The latest release in the 4.3 release series is [1]GCC 4.3.5.
5152 * GCC requires the [2]GMP and [3]MPFR libraries for building all the
5153 various front-end languages it supports. See the [4]prerequisites
5154 page for version requirements.
5155 * ColdFire targets now treat long double as having the same format as
5156 double. In earlier versions of GCC, they used the 68881 long double
5158 * The m68k-uclinux target now uses the same calling conventions as
5159 m68k-linux-gnu. You can select the original calling conventions by
5160 configuring for m68k-uclinuxoldabi instead. Note that
5161 m68k-uclinuxoldabi also retains the original 80-bit long double on
5163 * The -fforce-mem option has been removed because it has had no
5164 effect in the last few GCC releases.
5165 * The i386 -msvr3-shlib option has been removed since it is no longer
5167 * Fastcall for i386 has been changed not to pass aggregate arguments
5168 in registers, following Microsoft compilers.
5169 * Support for the AOF assembler has been removed from the ARM back
5170 end; this affects only the targets arm-semi-aof and armel-semi-aof,
5171 which are no longer recognized. We removed these targets without a
5172 deprecation period because we discovered that they have been
5173 unusable since GCC 4.0.0.
5174 * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*)
5175 has been removed. This support had been deprecated since GCC 4.0.0.
5176 * Support for a number of older systems and recently unmaintained or
5177 untested target ports of GCC has been declared obsolete in GCC 4.3.
5178 Unless there is activity to revive them, the next release of GCC
5179 will have their sources permanently removed.
5180 All GCC ports for the following processor architectures have been
5183 The following aliases for processor architectures have been
5184 declared obsolete. Users should use the indicated generic target
5185 names instead, with compile-time options such as -mcpu or
5186 configure-time options such as --with-cpu to control the
5187 configuration more precisely.
5188 + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-*
5190 + parisc*-*-* (use hppa*-*-* instead).
5191 + m680[012]0-*-* (use m68k-*-* instead).
5192 All GCC ports for the following operating systems have been
5196 + GNU/Linux using the a.out object format (*-*-linux*aout*)
5197 + GNU/Linux using version 1 of the GNU C Library
5199 + Solaris versions before Solaris 7 (*-*-solaris2.[0-6],
5200 *-*-solaris2.[0-6].*)
5201 + Miscellaneous System V (*-*-sysv*)
5202 + WindISS (*-*-windiss*)
5203 Also, those for some individual systems on particular architectures
5204 have been obsoleted:
5205 + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*)
5206 + CRIS with a.out object format (cris-*-aout)
5207 + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*)
5208 + OSF/1 on PA-RISC (hppa1.1-*-osf*)
5209 + PRO on PA-RISC (hppa1.1-*-pro*)
5210 + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*,
5211 i[34567]86-sequent-sysv4*)
5212 + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*)
5213 + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host
5214 was previously [5]removed in 2001, leaving only the support
5215 for UWIN as a target now being deprecated)
5216 + ChorusOS on PowerPC (powerpc-*-chorusos*)
5217 + All VAX configurations apart from NetBSD and OpenBSD
5218 (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*)
5219 * The [6]-Wconversion option has been modified. Its purpose now is to
5220 warn for implicit conversions that may alter a value. This new
5221 behavior is available for both C and C++. Warnings about
5222 conversions between signed and unsigned integers can be disabled by
5223 using -Wno-sign-conversion. In C++, they are disabled by default
5224 unless -Wsign-conversion is explicitly requested. The old behavior
5225 of -Wconversion, that is, warn for prototypes causing a type
5226 conversion that is different from what would happen to the same
5227 argument in the absence of a prototype, has been moved to a new
5228 option -Wtraditional-conversion, which is only available for C.
5229 * The -m386, -m486, -mpentium and -mpentiumpro tuning options have
5230 been removed because they were deprecated for more than 3 GCC major
5231 releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or
5232 -mtune=pentiumpro as a replacement.
5233 * The -funsafe-math-optimizations option now automatically turns on
5234 -fno-trapping-math in addition to -fno-signed-zeros, as it enables
5235 reassociation and thus may introduce or remove traps.
5236 * The -ftree-vectorize option is now on by default under -O3. In
5237 order to generate code for a SIMD extension, it has to be enabled
5238 as well: use -maltivec for PowerPC platforms and -msse/-msse2 for
5240 * More information on porting to GCC 4.3 from previous versions of
5241 GCC can be found in the [7]porting guide for this release.
5243 General Optimizer Improvements
5245 * The GCC middle-end has been integrated with the [8]MPFR library.
5246 This allows GCC to evaluate and replace at compile-time calls to
5247 built-in math functions having constant arguments with their
5248 mathematically equivalent results. In making use of [9]MPFR, GCC
5249 can generate correct results regardless of the math library
5250 implementation or floating point precision of the host platform.
5251 This also allows GCC to generate identical results regardless of
5252 whether one compiles in native or cross-compile configurations to a
5253 particular target. The following built-in functions take advantage
5254 of this new capability: acos, acosh, asin, asinh, atan2, atan,
5255 atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1,
5256 fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10,
5257 log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh,
5258 tan, tanh, tgamma, y0, y1 and yn. The float and long double
5259 variants of these functions (e.g. sinf and sinl) are also handled.
5260 The sqrt and cabs functions with constant arguments were already
5261 optimized in prior GCC releases. Now they also use [10]MPFR.
5262 * A new forward propagation pass on RTL was added. The new pass
5263 replaces several slower transformations, resulting in compile-time
5264 improvements as well as better code generation in some cases.
5265 * A new command-line switch -frecord-gcc-switches has been added to
5266 GCC, although it is only enabled for some targets. The switch
5267 causes the command line that was used to invoke the compiler to be
5268 recorded into the object file that is being created. The exact
5269 format of this recording is target and binary file format
5270 dependent, but it usually takes the form of a note section
5271 containing ASCII text. The switch is related to the -fverbose-asm
5272 switch, but that one only records the information in the assembler
5273 output file as comments, so the information never reaches the
5275 * The inliner heuristic is now aware of stack frame consumption. New
5276 command-line parameters --param large-stack-frame and --param
5277 large-stack-frame-growth can be used to limit stack frame size
5278 growth caused by inlining.
5279 * During feedback directed optimizations, the expected block size the
5280 memcpy, memset and bzero functions operate on is discovered and for
5281 cases of commonly used small sizes, specialized inline code is
5283 * __builtin_expect no longer requires its argument to be a compile
5285 * Interprocedural optimization was reorganized to work on functions
5286 in SSA form. This enables more precise and cheaper dataflow
5287 analysis and makes writing interprocedural optimizations easier.
5288 The following improvements have been implemented on top of this
5290 + Pre-inline optimization: Selected local optimization passes
5291 are run before the inliner (and other interprocedural passes)
5292 are executed. This significantly improves the accuracy of code
5293 growth estimates used by the inliner and reduces the overall
5294 memory footprint for large compilation units.
5295 + Early inlining (a simple bottom-up inliner pass inlining only
5296 functions whose body is smaller than the expected call
5297 overhead) is now executed with the early optimization passes,
5298 thus inlining already optimized function bodies into an
5299 unoptimized function that is subsequently optimized by early
5300 optimizers. This enables the compiler to quickly eliminate
5301 abstraction penalty in C++ programs.
5302 + Interprocedural constant propagation now operate on SSA form
5303 increasing accuracy of the analysis.
5304 * A new internal representation for GIMPLE statements has been
5305 contributed, resulting in compile-time memory savings.
5306 * The vectorizer was enhanced to support vectorization of outer
5307 loops, intra-iteration parallelism (loop-aware SLP), vectorization
5308 of strided accesses and loops with multiple data-types. Run-time
5309 dependency testing using loop versioning was added. The cost model,
5310 turned on by -fvect-cost-model, was developed.
5312 New Languages and Language specific improvements
5314 * We have added new command-line options
5315 -finstrument-functions-exclude-function-list and
5316 -finstrument-functions-exclude-file-list. They provide more control
5317 over which functions are annotated by the -finstrument-functions
5322 * Implicit conversions between generic vector types are now only
5323 permitted when the two vectors in question have the same number of
5324 elements and compatible element types. (Note that the restriction
5325 involves compatible element types, not implicitly-convertible
5326 element types: thus, a vector type with element type int may not be
5327 implicitly converted to a vector type with element type unsigned
5328 int.) This restriction, which is in line with specifications for
5329 SIMD architectures such as AltiVec, may be relaxed using the flag
5330 -flax-vector-conversions. This flag is intended only as a
5331 compatibility measure and should not be used for new code.
5332 * -Warray-bounds has been added and is now enabled by default for
5333 -Wall . It produces warnings for array subscripts that can be
5334 determined at compile time to be always out of bounds.
5335 -Wno-array-bounds will disable the warning.
5336 * The constructor and destructor function attributes now accept
5337 optional priority arguments which control the order in which the
5338 constructor and destructor functions are run.
5339 * New [11]command-line options -Wtype-limits,
5340 -Wold-style-declaration, -Wmissing-parameter-type, -Wempty-body,
5341 -Wclobbered and -Wignored-qualifiers have been added for finer
5342 control of the diverse warnings enabled by -Wextra.
5343 * A new function attribute alloc_size has been added to mark up
5344 malloc style functions. For constant sized allocations this can be
5345 used to find out the size of the returned pointer using the
5346 __builtin_object_size() function for buffer overflow checking and
5347 similar. This supplements the already built-in malloc and calloc
5348 constant size handling.
5349 * Integer constants written in binary are now supported as a GCC
5350 extension. They consist of a prefix 0b or 0B, followed by a
5351 sequence of 0 and 1 digits.
5352 * A new predefined macro __COUNTER__ has been added. It expands to
5353 sequential integral values starting from 0. In conjunction with the
5354 ## operator, this provides a convenient means to generate unique
5356 * A new command-line option -fdirectives-only has been added. It
5357 enables a special preprocessing mode which improves the performance
5358 of applications like distcc and ccache.
5359 * Fixed-point data types and operators have been added. They are
5360 based on Chapter 4 of the Embedded-C specification (n1169.pdf).
5361 Currently, only MIPS targets are supported.
5362 * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732,
5363 N1241, is now supported as a GCC extension to C for targets
5364 i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu,
5365 and x86_64-*-linux-gnu. The feature introduces new data types
5366 _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF,
5371 * [12]Experimental support for the upcoming ISO C++ standard, C++0x.
5372 * -Wc++0x-compat has been added and is now enabled by default for
5373 -Wall. It produces warnings for constructs whose meaning differs
5374 between ISO C++ 1998 and C++0x.
5375 * The -Wparentheses option now works for C++ as it does for C. It
5376 warns if parentheses are omitted when operators with confusing
5377 precedence are nested. It also warns about ambiguous else
5378 statements. Since -Wparentheses is enabled by -Wall, this may cause
5379 additional warnings with existing C++ code which uses -Wall. These
5380 new warnings may be disabled by using -Wall -Wno-parentheses.
5381 * The -Wmissing-declarations now works for C++ as it does for C.
5382 * The -fvisibility-ms-compat flag was added, to make it easier to
5383 port larger projects using shared libraries from Microsoft's Visual
5384 Studio to ELF and Mach-O systems.
5385 * C++ attribute handling has been overhauled for template arguments
5386 (ie dependent types). In particular, __attribute__((aligned(T)));
5387 works for C++ types.
5389 Runtime Library (libstdc++)
5391 * [13]Experimental support for the upcoming ISO C++ standard, C++0x.
5392 * Support for TR1 mathematical special functions and regular
5393 expressions. ([14]Implementation status of TR1)
5394 * Default what implementations give more elaborate exception strings
5395 for bad_cast, bad_typeid, bad_exception, and bad_alloc.
5396 * Header dependencies have been streamlined, reducing unnecessary
5397 includes and pre-processed bloat.
5398 * Variadic template implementations of items in <tuple> and
5400 * An experimental [15]parallel mode has been added. This is a
5401 parallel implementation of many C++ Standard library algorithms,
5402 like std::accumulate, std::for_each, std::transform, or std::sort,
5403 to give but four examples. These algorithms can be substituted for
5404 the normal (sequential) libstdc++ algorithms on a piecemeal basis,
5405 or all existing algorithms can be transformed via the
5406 -D_GLIBCXX_PARALLEL macro.
5407 * Debug mode versions of classes in <unordered_set> and
5409 * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are
5410 now <backward/hash_set> and <backward/hash_map>. This code:
5411 #include <ext/hash_set>
5412 __gnu_cxx::hash_set<int> s;
5414 Can be transformed (in order of preference) to:
5415 #include <tr1/unordered_set>
5416 std::tr1::unordered_set<int> s;
5419 #include <backward/hash_set>
5420 __gnu_cxx::hash_set<int> s;
5422 Similar transformations apply to __gnu_cxx::hash_map,
5423 __gnu_cxx::hash_multimap, __gnu_cxx::hash_set,
5424 __gnu_cxx::hash_multiset.
5428 * Due to the fact that the [16]GMP and [17]MPFR libraries are
5429 required for all languages, Fortran is no longer special in this
5430 regard and is available by default.
5431 * The [18]-fexternal-blas option has been added, which generates
5432 calls to BLAS routines for intrinsic matrix operations such as
5433 matmul rather than using the built-in algorithms.
5434 * Support to give a backtrace (compiler flag -fbacktrace or
5435 environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems
5436 only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a
5437 run-time error occured.
5438 * GNU Fortran now defines __GFORTRAN__ when it runs the C
5440 * The [19]-finit-local-zero, -finit-real, -finit-integer,
5441 -finit-character, and -finit-logical options have been added, which
5442 can be used to initialize local variables.
5443 * The intrinsic procedures [20]GAMMA and [21]LGAMMA have been added,
5444 which calculate the Gamma function and its logarithm. Use EXTERNAL
5445 gamma if you want to use your own gamma function.
5446 * GNU Fortran now regards the backslash character as literal (as
5447 required by the Fortran 2003 standard); using [22]-fbackslash GNU
5448 Fortran interprets backslashes as C-style escape characters.
5449 * The [23]interpretation of binary, octal and hexadecimal (BOZ)
5450 literal constants has been changed. Before they were always
5451 interpreted as integer; now they are bit-wise transferred as
5452 argument of INT, REAL, DBLE and CMPLX as required by the Fortran
5453 2003 standard, and for real and complex variables in DATA
5454 statements or when directly assigned to real and complex variables.
5455 Everywhere else and especially in expressions they are still
5456 regarded as integer constants.
5457 * Fortran 2003 support has been extended:
5458 + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE
5460 + Intrinsic module ISO_ENV_FORTRAN
5461 + Interoperability with C (ISO C Bindings)
5462 + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER
5468 * GCJ now uses the Eclipse Java compiler for its Java parsing needs.
5469 This enables the use of all 1.5 language features, and fixes most
5470 existing front end bugs.
5471 * libgcj now supports all 1.5 language features which require runtime
5472 support: foreach, enum, annotations, generics, and auto-boxing.
5473 * We've made many changes to the tools shipped with gcj.
5474 + The old jv-scan tool has been removed. This tool never really
5475 worked properly. There is no replacement.
5476 + gcjh has been rewritten. Some of its more obscure options no
5477 longer work, but are still recognized in an attempt at
5478 compatibility. gjavah is a new program with similar
5479 functionality but different command-line options.
5480 + grmic and grmiregistry have been rewritten. grmid has been
5482 + gjar replaces the old fastjar.
5483 + gjarsigner (used for signing jars), gkeytool (used for key
5484 management), gorbd (for CORBA), gserialver (computes
5485 serialization UIDs), and gtnameserv (also for CORBA) are now
5487 * The ability to dump the contents of the java run time heap to a
5488 file for off-line analysis has been added. The heap dumps may be
5489 analyzed with the new gc-analyze tool. They may be generated on
5490 out-of-memory conditions or on demand and are controlled by the new
5491 run time class gnu.gcj.util.GCInfo.
5492 * java.util.TimeZone can now read files from /usr/share/zoneinfo to
5493 provide correct, updated, timezone information. This means that
5494 packagers no longer have to update libgcj when a time zone change
5497 New Targets and Target Specific Improvements
5501 * Tuning for Intel Core 2 processors is available via -mtune=core2
5503 * Tuning for AMD Geode processors is available via -mtune=geode and
5505 * Code generation of block move (memcpy) and block set (memset) was
5506 rewritten. GCC can now pick the best algorithm (loop, unrolled
5507 loop, instruction with rep prefix or a library call) based on the
5508 size of the block being copied and the CPU being optimized for. A
5509 new option -minline-stringops-dynamically has been added. With this
5510 option string operations of unknown size are expanded such that
5511 small blocks are copied by in-line code, while for large blocks a
5512 library call is used. This results in faster code than
5513 -minline-all-stringops when the library implementation is capable
5514 of using cache hierarchy hints. The heuristic choosing the
5515 particular algorithm can be overwritten via -mstringop-strategy.
5516 Newly also memset of values different from 0 is inlined.
5517 * GCC no longer places the cld instruction before string operations.
5518 Both i386 and x86-64 ABI documents mandate the direction flag to be
5519 clear at the entry of a function. It is now invalid to set the flag
5520 in asm statement without reseting it afterward.
5521 * Support for SSSE3 built-in functions and code generation are
5522 available via -mssse3.
5523 * Support for SSE4.1 built-in functions and code generation are
5524 available via -msse4.1.
5525 * Support for SSE4.2 built-in functions and code generation are
5526 available via -msse4.2.
5527 * Both SSE4.1 and SSE4.2 support can be enabled via -msse4.
5528 * A new set of options -mpc32, -mpc64 and -mpc80 have been added to
5529 allow explicit control of x87 floating point precision.
5530 * Support for __float128 (TFmode) IEEE quad type and corresponding
5531 TCmode IEEE complex quad type is available via the soft-fp library
5532 on x86_64 targets. This includes basic arithmetic operations
5533 (addition, subtraction, negation, multiplication and division) on
5534 __float128 real and TCmode complex values, the full set of IEEE
5535 comparisons between __float128 values, conversions to and from
5536 float, double and long double floating point types, as well as
5537 conversions to and from signed or unsigned integer, signed or
5538 unsigned long integer and signed or unsigned quad (TImode) integer
5539 types. Additionally, all operations generate the full set of IEEE
5540 exceptions and support the full set of IEEE rounding modes.
5541 * GCC can now utilize the ACML library for vectorizing calls to a set
5542 of C99 functions on x86_64 if -mveclibabi=acml is specified and you
5543 link to an ACML ABI compatible library.
5547 * Compiler and Library support for Thumb-2 and the ARMv7 architecture
5554 * Compiler and Library support for the CRIS v32 architecture, as
5555 found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been
5558 Configuration changes
5560 * The cris-*-elf target now includes support for CRIS v32, including
5561 libraries, through the -march=v32 option.
5562 * A new crisv32-*-elf target defaults to generate code for CRIS v32.
5563 * A new crisv32-*-linux* target defaults to generate code for CRIS
5565 * The cris-*-aout target has been obsoleted.
5567 Improved support for built-in functions
5569 * GCC can now use the lz and swapwbr instructions to implement the
5570 __builtin_clz, __builtin_ctz and __builtin_ffs family of functions.
5571 * __builtin_bswap32 is now implemented using the swapwb instruction,
5578 * Support for several new ColdFire processors has been added. You can
5579 generate code for them using the new -mcpu option.
5580 * All targets now support ColdFire processors.
5581 * m68k-uclinux targets have improved support for C++ constructors and
5582 destructors, and for shared libraries.
5583 * It is now possible to set breakpoints on the first or last line of
5584 a function, even if there are no statements on that line.
5588 * Support for sibling calls has been added.
5589 * More use is now made of the ColdFire mov3q instruction.
5590 * __builtin_clz is now implemented using the ff1 ColdFire
5591 instruction, when available.
5592 * GCC now honors the -m68010 option. 68010 code now uses clr rather
5593 than move to zero volatile memory.
5594 * 68020 targets and above can now use symbol(index.size*scale)
5595 addresses for indexed array accesses. Earlier compilers would
5596 always load the symbol into a base register first.
5598 Configuration changes
5600 * All m68k and ColdFire targets now allow the default processor to be
5601 set at configure time using --with-cpu.
5602 * A --with-arch configuration option has been added. This option
5603 allows you to restrict a target to ColdFire or non-ColdFire
5608 * An __mcfv*__ macro is now defined for all ColdFire targets.
5609 (Earlier versions of GCC only defined __mcfv4e__.)
5610 * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added.
5611 * All targets now define __mc68010 and __mc68010__ when generating
5614 Command-line changes
5616 * New command-line options -march, -mcpu, -mtune and -mhard-float
5617 have been added. These options apply to both m68k and ColdFire
5619 * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative
5620 versions of -mshort, etc.
5621 * -fforce-addr has been removed. It is now ignored by the compiler.
5625 * ColdFire targets now try to maintain a 4-byte-aligned stack where
5627 * m68k-uclinux targets now try to avoid situations that lead to the
5628 load-time error: BINFMT_FLAT: reloc outside program.
5632 Changes to existing configurations
5634 * libffi and libjava now support all three GNU/Linux ABIs: o32, n32
5635 and n64. Every GNU/Linux configuration now builds these libraries
5637 * GNU/Linux configurations now generate -mno-shared code unless
5638 overridden by -fpic, -fPIC, -fpie or -fPIE.
5639 * mipsisa32*-linux-gnu configurations now generate hard-float code by
5640 default, just like other mipsisa32* and mips*-linux-gnu
5641 configurations. You can build a soft-float version of any
5642 mips*-linux-gnu configuration by passing --with-float=soft to
5644 * mips-wrs-vxworks now supports run-time processes (RTPs).
5646 Changes to existing command-line options
5648 * The -march and -mtune options no longer accept 24k as a processor
5649 name. Please use 24kc, 24kf2_1 or 24kf1_1 instead.
5650 * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and
5651 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The
5652 options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for
5653 24kx, 24kex and 34kx.
5657 GCC now supports the following configurations:
5658 * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by
5659 default. Earlier releases also recognized this configuration, but
5660 they treated it in the same way as mipsisa32*-linux-gnu*. Note that
5661 you can customize any mips*-linux-gnu* configuration to a
5662 particular ISA or processor by passing an appropriate --with-arch
5663 option to configure.
5664 * mipsisa*-sde-elf*, which provides compatibility with MIPS
5665 Technologies' SDE toolchains. The configuration uses the SDE
5666 libraries by default, but you can use it like other newlib-based
5667 ELF configurations by passing --with-newlib to configure. It is the
5668 only configuration besides mips64vr*-elf* to build MIPS16 as well
5669 as non-MIPS16 libraries.
5670 * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf*
5671 configuration, but uses the o32 and o64 ABIs instead of the 32-bit
5672 and 64-bit forms of the EABI.
5674 New processors and application-specific extensions
5676 * Support for the SmartMIPS ASE is available through the new
5678 * Support for revision 2 of the DSP ASE is available through the new
5679 -mdspr2 option. A new preprocessor macro called __mips_dsp_rev
5680 indicates the revision of the ASE in use.
5681 * Support for the 4KS and 74K families of processors is available
5682 through the -march and -mtune options.
5684 Improved support for built-in functions
5686 * GCC can now use load-linked, store-conditional and sync
5687 instructions to implement atomic built-in functions such as
5688 __sync_fetch_and_add. The memory reference must be 4 bytes wide for
5689 32-bit targets and either 4 or 8 bytes wide for 64-bit targets.
5690 * GCC can now use the clz and dclz instructions to implement the
5691 __builtin_ctz and __builtin_ffs families of functions.
5692 * There is a new __builtin___clear_cache function for flushing the
5693 instruction cache. GCC expands this function inline on MIPS32
5694 revision 2 targets, otherwise it calls the function specified by
5699 * GCC can now compile objects that contain a mixture of MIPS16 and
5700 non-MIPS16 code. There are two new attributes, mips16 and nomips16,
5701 for specifying which mode a function should use.
5702 * A new option called -minterlink-mips16 makes non-MIPS16 code
5703 link-compatible with MIPS16 code.
5704 * After many bug fixes, the long-standing MIPS16 -mhard-float support
5705 should now work fairly reliably.
5706 * GCC can now use the MIPS16e save and restore instructions.
5707 * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled
5708 with -G0 -fsection-anchors is often smaller than code compiled with
5709 -G8. However, please note that you must usually compile all objects
5710 in your application with the same -G option; see the documentation
5712 * A new option called-mcode-readable specifies which instructions are
5713 allowed to load from the code segment. -mcode-readable=yes is the
5714 default and says that any instruction may load from the code
5715 segment. The other alternatives are -mcode-readable=pcrel, which
5716 says that only PC-relative MIPS16 instructions may load from the
5717 code segment, and -mcode-readable=no, which says that no
5718 instruction may do so. Please see the documentation for more
5719 details, including example uses.
5721 Small-data improvements
5723 There are three new options for controlling small data:
5724 * -mno-extern-sdata, which disables small-data accesses for
5725 externally-defined variables. Code compiled with -Gn
5726 -mno-extern-sdata will be link-compatible with any -G setting
5727 between -G0 and -Gn inclusive.
5728 * -mno-local-sdata, which disables the use of small-data sections for
5729 data that is not externally visible. This option can be a useful
5730 way of reducing small-data usage in less performance-critical parts
5732 * -mno-gpopt, which disables the use of the $gp register while still
5733 honoring the -G limit when placing externally-visible data. This
5734 option implies -mno-extern-sdata and -mno-local-sdata and it can be
5735 useful in situations where $gp does not necessarily hold the
5738 Miscellaneous improvements
5740 * There is a new option called -mbranch-cost for tweaking the
5741 perceived cost of branches.
5742 * If GCC is configured to use a version of GAS that supports the
5743 .gnu_attribute directive, it will use that directive to record
5744 certain properties of the output code. .gnu_attribute is new to GAS
5746 * There are two new function attributes, near and far, for overriding
5747 the command-line setting of -mlong-calls on a function-by-function
5749 * -mfp64, which previously required a 64-bit target, now works with
5750 MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and
5751 mipsisa*-sde-elf* configurations provide suitable library support.
5752 * GCC now recognizes the -mdmx and -mmt options and passes them down
5753 to the assembler. It does nothing else with the options at present.
5755 SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture
5758 * Support has been added for this new architecture.
5760 RS6000 (POWER/PowerPC)
5762 * Support for the PowerPC 750CL paired-single instructions has been
5763 added with a new powerpc-*-linux*paired* target configuration. It
5764 is enabled by an associated -mpaired option and can be accessed
5765 using new built-in functions.
5766 * Support for auto-detecting architecture and system configuration to
5767 auto-select processor optimization tuning.
5768 * Support for VMX on AIX 5.3 has been added.
5769 * Support for AIX Version 6.1 has been added.
5771 S/390, zSeries and System z9
5773 * Support for the IBM System z9 EC/BC processor (z9 GA3) has been
5774 added. When using the -march=z9-ec option, the compiler will
5775 generate code making use of instructions provided by the decimal
5776 floating point facility and the floating point conversion facility
5777 (pfpo). Besides the instructions used to implement decimal floating
5778 point operations these facilities also contain instructions to move
5779 between general purpose and floating point registers and to modify
5780 and copy the sign-bit of floating point values.
5781 * When the -march=z9-ec option is used the new
5782 -mhard-dfp/-mno-hard-dfp options can be used to specify whether the
5783 decimal floating point hardware instructions will be used or not.
5784 If none of them is given the hardware support is enabled by
5786 * The -mstack-guard option can now be omitted when using stack
5787 checking via -mstack-size in order to let GCC choose a sensible
5788 stack guard value according to the frame size of each function.
5789 * Various changes to improve performance of generated code have been
5790 implemented, including:
5791 + The condition code set by an add logical with carry
5792 instruction is now available for overflow checks like: a + b +
5794 + The test data class instruction is now used to implement
5795 sign-bit and infinity checks of binary and decimal floating
5800 * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been
5805 * Stack unwinding for exception handling now uses by default a
5806 specialized version of DWARF unwinding. This is not
5807 binary-compatible with the setjmp/longjmp (sjlj) unwinding used for
5808 Xtensa with previous versions of GCC.
5809 * For Xtensa processors that include the Conditional Store option,
5810 the built-in functions for atomic memory access are now implemented
5811 using S32C1I instructions.
5812 * If the Xtensa NSA option is available, GCC will use it to implement
5813 the __builtin_ctz and __builtin_clz functions.
5815 Documentation improvements
5817 * Existing libstdc++ documentation has been edited and restructured
5818 into a single DocBook XML manual. The results can be viewed online
5821 Other significant improvements
5823 * The compiler's --help command-line option has been extended so that
5824 it now takes an optional set of arguments. These arguments restrict
5825 the information displayed to specific classes of command-line
5826 options, and possibly only a subset of those options. It is also
5827 now possible to replace the descriptive text associated with each
5828 displayed option with an indication of its current value, or for
5829 binary options, whether it has been enabled or disabled.
5830 Here are some examples. The following will display all the options
5831 controlling warning messages:
5834 Whereas this will display all the undocumented, target specific
5836 --help=target,undocumented
5838 This sequence of commands will display the binary optimizations
5839 that are enabled by -O3:
5840 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
5841 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
5842 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
5844 * The configure options --with-pkgversion and --with-bugurl have been
5845 added. These allow distributors of GCC to include a
5846 distributor-specific string in manuals and --version output and to
5847 specify the URL for reporting bugs in their versions of GCC.
5851 This is the [25]list of problem reports (PRs) from GCC's bug tracking
5852 system that are known to be fixed in the 4.3.1 release. This list might
5853 not be complete (that is, it is possible that some PRs that have been
5854 fixed are not listed here).
5856 Target Specific Changes
5862 * Starting with GCC 4.3.1, decimal floating point variables are
5863 aligned to their natural boundaries when they are passed on the
5866 Command-line changes
5868 * Starting with GCC 4.3.1, the -mcld option has been added to
5869 automatically generate a cld instruction in the prologue of
5870 functions that use string instructions. This option is used for
5871 backward compatibility on some operating systems and can be enabled
5872 by default for 32-bit x86 targets by configuring GCC with the
5873 --enable-cld configure option.
5877 This is the [26]list of problem reports (PRs) from GCC's bug tracking
5878 system that are known to be fixed in the 4.3.2 release. This list might
5879 not be complete (that is, it is possible that some PRs that have been
5880 fixed are not listed here).
5884 This is the [27]list of problem reports (PRs) from GCC's bug tracking
5885 system that are known to be fixed in the 4.3.3 release. This list might
5886 not be complete (that is, it is possible that some PRs that have been
5887 fixed are not listed here).
5891 This is the [28]list of problem reports (PRs) from GCC's bug tracking
5892 system that are known to be fixed in the 4.3.4 release. This list might
5893 not be complete (that is, it is possible that some PRs that have been
5894 fixed are not listed here).
5898 This is the [29]list of problem reports (PRs) from GCC's bug tracking
5899 system that are known to be fixed in the 4.3.5 release. This list might
5900 not be complete (that is, it is possible that some PRs that have been
5901 fixed are not listed here).
5905 This is the [30]list of problem reports (PRs) from GCC's bug tracking
5906 system that are known to be fixed in the 4.3.6 release. This list might
5907 not be complete (that is, it is possible that some PRs that have been
5908 fixed are not listed here).
5911 For questions related to the use of GCC, please consult these web
5912 pages and the [31]GCC manuals. If that fails, the
5913 [32]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5914 web pages and the development of GCC are welcome on our developer
5915 list at [33]gcc@gcc.gnu.org. All of [34]our lists have public
5918 Copyright (C) [35]Free Software Foundation, Inc. Verbatim copying and
5919 distribution of this entire article is permitted in any medium,
5920 provided this notice is preserved.
5922 These pages are [36]maintained by the GCC team. Last modified
5927 1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5
5928 2. http://gmplib.org/
5929 3. http://www.mpfr.org/
5930 4. https://gcc.gnu.org/install/prerequisites.html
5931 5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html
5932 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
5933 7. http://gcc.gnu.org/gcc-4.3/porting_to.html
5934 8. http://www.mpfr.org/
5935 9. http://www.mpfr.org/
5936 10. http://www.mpfr.org/
5937 11. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
5938 12. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
5939 13. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
5940 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/bk01pt01ch01.html#m anual.intro.status.standard.tr1
5941 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html
5942 16. http://gmplib.org/
5943 17. http://www.mpfr.org/
5944 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options
5945 19. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167
5946 20. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html
5947 21. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html
5948 22. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html
5949 23. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html
5950 24. https://gcc.gnu.org/onlinedocs/libstdc++/
5951 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1
5952 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2
5953 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3
5954 28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4
5955 29. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5
5956 30. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6
5957 31. https://gcc.gnu.org/onlinedocs/
5958 32. mailto:gcc-help@gcc.gnu.org
5959 33. mailto:gcc@gcc.gnu.org
5960 34. https://gcc.gnu.org/lists.html
5961 35. http://www.fsf.org/
5962 36. https://gcc.gnu.org/about.html
5963 37. http://validator.w3.org/check/referer
5964 ======================================================================
5965 http://gcc.gnu.org/gcc-4.2/index.html
5966 GCC 4.2 Release Series
5970 The [1]GNU project and the GCC developers are pleased to announce the
5971 release of GCC 4.2.4.
5973 This release is a bug-fix release, containing fixes for regressions in
5974 GCC 4.2.3 relative to previous releases of GCC.
5979 May 19, 2008 ([2]changes)
5982 February 1, 2008 ([3]changes)
5985 October 7, 2007 ([4]changes)
5988 July 18, 2007 ([5]changes)
5991 May 13, 2007 ([6]changes)
5993 References and Acknowledgements
5995 GCC used to stand for the GNU C Compiler, but since the compiler
5996 supports several other languages aside from C, it now stands for the
5997 GNU Compiler Collection.
5999 A list of [7]successful builds is updated as new information becomes
6002 The GCC developers would like to thank the numerous people that have
6003 contributed new features, improvements, bug fixes, and other changes as
6004 well as test results to GCC. This [8]amazing group of volunteers is
6005 what makes GCC successful.
6007 For additional information about GCC please refer to the [9]GCC project
6008 web site or contact the [10]GCC development mailing list.
6010 To obtain GCC please use [11]our mirror sites or [12]our SVN server.
6013 For questions related to the use of GCC, please consult these web
6014 pages and the [13]GCC manuals. If that fails, the
6015 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6016 web pages and the development of GCC are welcome on our developer
6017 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
6020 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
6021 distribution of this entire article is permitted in any medium,
6022 provided this notice is preserved.
6024 These pages are [18]maintained by the GCC team. Last modified
6029 1. http://www.gnu.org/
6030 2. http://gcc.gnu.org/gcc-4.2/changes.html
6031 3. http://gcc.gnu.org/gcc-4.2/changes.html
6032 4. http://gcc.gnu.org/gcc-4.2/changes.html
6033 5. http://gcc.gnu.org/gcc-4.2/changes.html
6034 6. http://gcc.gnu.org/gcc-4.2/changes.html
6035 7. http://gcc.gnu.org/gcc-4.2/buildstat.html
6036 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6037 9. http://gcc.gnu.org/index.html
6038 10. mailto:gcc@gcc.gnu.org
6039 11. http://gcc.gnu.org/mirrors.html
6040 12. http://gcc.gnu.org/svn.html
6041 13. https://gcc.gnu.org/onlinedocs/
6042 14. mailto:gcc-help@gcc.gnu.org
6043 15. mailto:gcc@gcc.gnu.org
6044 16. https://gcc.gnu.org/lists.html
6045 17. http://www.fsf.org/
6046 18. https://gcc.gnu.org/about.html
6047 19. http://validator.w3.org/check/referer
6048 ======================================================================
6049 http://gcc.gnu.org/gcc-4.2/changes.html
6050 GCC 4.2 Release Series
6051 Changes, New Features, and Fixes
6055 * GCC no longer accepts the -fshared-data option. This option has had
6056 no effect in any GCC 4 release; the targets to which the option
6057 used to apply had been removed before GCC 4.0.
6059 General Optimizer Improvements
6061 * New command-line options specify the possible relationships among
6062 parameters and between parameters and global data. For example,
6063 -fargument-noalias-anything specifies that arguments do not alias
6065 Each language will automatically use whatever option is required by
6066 the language standard. You should not need to use these options
6069 New Languages and Language specific improvements
6071 * [1]OpenMP is now supported for the C, C++ and Fortran compilers.
6072 * New command-line options -fstrict-overflow and -Wstrict-overflow
6073 have been added. -fstrict-overflow tells the compiler that it may
6074 assume that the program follows the strict signed overflow
6075 semantics permitted for the language: for C and C++ this means that
6076 the compiler may assume that signed overflow does not occur. For
6077 example, a loop like
6078 for (i = 1; i > 0; i *= 2)
6080 is presumably intended to continue looping until i overflows. With
6081 -fstrict-overflow, the compiler may assume that signed overflow
6082 will not occur, and transform this into an infinite loop.
6083 -fstrict-overflow is turned on by default at -O2, and may be
6084 disabled via -fno-strict-overflow. The -Wstrict-overflow option may
6085 be used to warn about cases where the compiler assumes that signed
6086 overflow will not occur. It takes five different levels:
6087 -Wstrict-overflow=1 to 5. See the [2]documentation for details.
6088 -Wstrict-overflow=1 is enabled by -Wall.
6089 * The new command-line option -fno-toplevel-reorder directs GCC to
6090 emit top-level functions, variables, and asm statements in the same
6091 order that they appear in the input file. This is intended to
6092 support existing code which relies on a particular ordering (for
6093 example, code which uses top-level asm statements to switch
6094 sections). For new code, it is generally better to use function and
6095 variable attributes. The -fno-toplevel-reorder option may be used
6096 for most cases which currently use -fno-unit-at-a-time. The
6097 -fno-unit-at-a-time option will be removed in some future version
6098 of GCC. If you know of a case which requires -fno-unit-at-a-time
6099 which is not fixed by -fno-toplevel-reorder, please open a bug
6104 * The pragma redefine_extname will now macro expand its tokens for
6105 compatibility with SunPRO.
6106 * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct
6107 GCC to handle inline functions as specified in the C99 standard. In
6108 preparation for this, GCC 4.2 will warn about any use of non-static
6109 inline functions in gnu99 or c99 mode. This new warning may be
6110 disabled with the new gnu_inline function attribute or the new
6111 -fgnu89-inline command-line option. Also, GCC 4.2 and later will
6112 define one of the preprocessor macros __GNUC_GNU_INLINE__ or
6113 __GNUC_STDC_INLINE__ to indicate the semantics of inline functions
6114 in the current compilation.
6115 * A new command-line option -Waddress has been added to warn about
6116 suspicious uses of memory addresses as, for example, using the
6117 address of a function in a conditional expression, and comparisons
6118 against the memory address of a string literal. This warning is
6123 * C++ visibility handling has been overhauled.
6124 Restricted visiblity is propagated from classes to members, from
6125 functions to local statics, and from templates and template
6126 arguments to instantiations, unless the latter has explicitly
6127 declared visibility.
6128 The visibility attribute for a class must come between the
6129 class-key and the name, not after the closing brace.
6130 Attributes are now allowed for enums and elaborated-type-specifiers
6131 that only declare a type.
6132 Members of the anonymous namespace are now local to a particular
6133 translation unit, along with any other declarations which use them,
6134 though they are still treated as having external linkage for
6136 * The (undocumented) extension which permitted templates with default
6137 arguments to be bound to template template parameters with fewer
6138 parameters has been removed. For example:
6139 template <template <typename> class C>
6140 void f(C<double>) {}
6142 template <typename T, typename U = int>
6145 template void f(S<double>);
6147 is no longer accepted by G++. The reason this code is not accepted
6148 is that S is a template with two parameters; therefore, it cannot
6149 be bound to C which has only one parameter.
6150 * The <?, >?, <?=, and >?= operators, deprecated in previous GCC
6151 releases, have been removed.
6152 * The command-line option -fconst-strings, deprecated in previous GCC
6153 releases, has been removed.
6154 * The configure variable enable-__cxa_atexit is now enabled by
6155 default for more targets. Enabling this variable is necessary in
6156 order for static destructors to be executed in the correct order,
6157 but it depends upon the presence of a non-standard C library in the
6158 target library in order to work. The variable is now enabled for
6159 more targets which are known to have suitable C libraries.
6160 * -Wextra will produce warnings for if statements with a semicolon as
6161 the only body, to catch code like:
6166 To suppress the warning in valid cases, use { } instead.
6167 * The C++ frontend now also produces strict aliasing warnings when
6168 -fstrict-aliasing -Wstrict-aliasing is in effect.
6170 Runtime Library (libstdc++)
6172 * Added support for TR1 <random>, <complex>, and C compatibility
6173 headers. In addition, a lock-free version of shared_ptr was
6174 contributed as part of Phillip Jordan's Google Summer of Code
6175 project on lock-free containers. ([3]Implementation status of TR1)
6176 * In association with the Summer of Code work on lock-free
6177 containers, the interface for atomic builtins was adjusted,
6178 creating simpler alternatives for non-threaded code paths. Also,
6179 usage was consolidated and all elements were moved from namespace
6180 std to namespace__gnu_cxx. Affected interfaces are the functions
6181 __exchange_and_add, __atomic_add, and the objects __mutex,
6182 __recursive_mutex, and __scoped_lock.
6183 * Support for versioning weak symbol names via namespace association
6184 was added. However, as this changes the names of exported symbols,
6185 this is turned off by default in the current ABI. Intrepid users
6186 can enable this feature by using
6187 --enable-symvers=gnu-versioned-namespace during configuration.
6188 * Revised, simplified, and expanded policy-based associative
6189 containers, including data types for tree and trie forms
6190 (basic_tree, tree, trie), lists (list_update), and both
6191 collision-chaining and probing hash-based containers
6192 (basic_hash_table, cc_hash_table, gp_hash_table). More details per
6193 the [4]documentation.
6194 * The implementation of the debug mode was modified, whereby the
6195 debug namespaces were nested inside of namespace std and namespace
6196 __gnu_cxx in order to resolve some long standing corner cases
6197 involving name lookup. Debug functionality from the policy-based
6198 data structures was consolidated and enabled with the single macro,
6199 _GLIBCXX_DEBUG. See PR 26142 for more information.
6200 * Added extensions for type traits: __conditional_type,
6201 __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if.
6202 * Added a typelist implementation for compile-time meta-programming.
6203 Elements for typelist construction and operation can be found
6204 within namespace __gnu_cxx::typelist.
6205 * Added a new allocator, __gnu_cxx::throw_allocator, for testing
6207 * Enabled library-wide visibility control, allowing -fvisibility to
6209 * Consolidated all nested namespaces and the conversion of
6210 __gnu_internal implementation-private details to anonymous
6211 namespaces whenever possible.
6212 * Implemented LWG resolutions DR 431 and DR 538.
6216 * Support for allocatable components has been added (TR 15581 and
6218 * Support for the Fortran 2003 streaming IO extension has been added.
6219 * The GNU Fortran compiler now uses 4-byte record markers by default
6220 for unformatted files to be compatible with g77 and most other
6221 compilers. The implementation allows for records greater than 2 GB
6222 and is compatible with several other compilers. Older versions of
6223 gfortran used 8-byte record markers by default (on most systems).
6224 In order to change the length of the record markers, e.g. to read
6225 unformatted files created by older gfortran versions, the
6226 [5]-frecord-marker=8 option can be used.
6230 * A new command-line option -static-libgcj has been added for targets
6231 that use a linker compatible with GNU Binutils. As its name
6232 implies, this causes libgcj to be linked statically. In some cases
6233 this causes the resulting executable to start faster and use less
6234 memory than if the shared version of libgcj were used. However
6235 caution should be used as it can also cause essential parts of the
6236 library to be omitted. Some of these issues are discussed in:
6237 [6]https://gcc.gnu.org/wiki/Statically_linking_libgcj
6238 * fastjar is no longer bundled with GCC. To build libgcj, you will
6239 need either InfoZIP (both zip and unzip) or an external jar
6240 program. In the former case, the GCC build will install a jar shell
6241 script that is based on InfoZIP and provides the same functionality
6244 New Targets and Target Specific Improvements
6248 * -mtune=generic can now be used to generate code running well on
6249 common x86 chips. This includes AMD Athlon, AMD Opteron, Intel
6250 Pentium-M, Intel Pentium 4 and Intel Core 2.
6251 * -mtune=native and -march=native will produce code optimized for the
6252 host architecture as detected using the cpuid instruction.
6253 * Added a new command-line option -fstackrealign and and
6254 __attribute__ ((force_align_arg_pointer)) to realign the stack at
6255 runtime. This allows functions compiled with a vector-aligned stack
6256 to be invoked from legacy objects that keep only word-alignment.
6260 * The default CPU setting has been changed from V7 to V9 in 32-bit
6261 mode on Solaris 7 and above. This is already the case in 64-bit
6262 mode. It can be overridden by specifying --with-cpu at configure
6264 * Back-end support of built-in functions for atomic memory access has
6266 * Support for the Sun UltraSPARC T1 (Niagara) processor has been
6271 * Various bug fixes have made some functions (notably, functions
6272 returning structures) incompatible with previous releases.
6273 Recompiling all libraries is recommended. Note that code quality
6274 has considerably improved since 4.1, making a recompile even more
6279 * Added support for the Broadcom SB-1A core.
6283 * Added support for IA-64 data and control speculation. By default
6284 speculation is enabled only during second scheduler pass. A number
6285 of machine flags was introduced to control the usage of speculation
6286 for both scheduler passes.
6290 * Added Java language support (libffi and libjava) for 32-bit HP-UX
6295 Documentation improvements
6299 * A make pdf target has been added to the top-level makefile,
6300 enabling automated production of PDF documentation files.
6301 (Front-ends external to GCC should modify their Make-lang.in file
6302 to add a lang.pdf: target.)
6304 Other significant improvements
6306 Build system improvements
6308 * All the components of the compiler are now bootstrapped by default.
6309 This improves the resilience to bugs in the system compiler or
6310 binary compatibility problems, as well as providing better testing
6311 of GCC 4.2 itself. In addition, if you build the compiler from a
6312 combined tree, the assembler, linker, etc. will also be
6313 bootstrapped (i.e. built with themselves).
6314 You can disable this behavior, and go back to the pre-GCC 4.2 set
6315 up, by configuring GCC with --disable-bootstrap.
6316 * The rules that configure follows to find target tools resemble more
6317 closely the locations that the built compiler will search. In
6318 addition, you can use the new configure option --with-target-tools
6319 to specify where to find the target tools used during the build,
6320 without affecting what the built compiler will use.
6321 This can be especially useful when building packages of GCC. For
6322 example, you may want to build GCC with GNU as or ld, even if the
6323 resulting compiler to work with the native assembler and linker. To
6324 do so, you can use --with-target-tools to point to the native
6327 Incompatible changes to the build system
6329 * Front-ends external to GCC should modify their Make-lang.in file to
6330 replace double-colon rules (e.g. dvi::) with normal rules (like
6331 lang.dvi:). Front-end makefile hooks do not use double-colon rules
6333 * Up to GCC 4.1, a popular way to specify the target tools used
6334 during the build was to create directories named gas, binutils,
6335 etc. in the build tree, and create links to the tools from there.
6336 This does not work any more when the compiler is bootstrapped. The
6337 new configure option --with-target-tools provides a better way to
6338 achieve the same effect, and works for all native and cross
6342 For questions related to the use of GCC, please consult these web
6343 pages and the [7]GCC manuals. If that fails, the
6344 [8]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6345 web pages and the development of GCC are welcome on our developer
6346 list at [9]gcc@gcc.gnu.org. All of [10]our lists have public
6349 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and
6350 distribution of this entire article is permitted in any medium,
6351 provided this notice is preserved.
6353 These pages are [12]maintained by the GCC team. Last modified
6358 1. http://gcc.gnu.org/projects/gomp/
6359 2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
6360 3. https://gcc.gnu.org/onlinedocs/libstdc++/manual/bk01pt01ch01.html#manual.intro.status.standard.tr1
6361 4. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html
6362 5. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html
6363 6. https://gcc.gnu.org/wiki/Statically_linking_libgcj
6364 7. https://gcc.gnu.org/onlinedocs/
6365 8. mailto:gcc-help@gcc.gnu.org
6366 9. mailto:gcc@gcc.gnu.org
6367 10. https://gcc.gnu.org/lists.html
6368 11. http://www.fsf.org/
6369 12. https://gcc.gnu.org/about.html
6370 13. http://validator.w3.org/check/referer
6371 ======================================================================
6372 http://gcc.gnu.org/gcc-4.1/index.html
6373 GCC 4.1 Release Series
6377 The [1]GNU project and the GCC developers are pleased to announce the
6378 release of GCC 4.1.2.
6380 This release is a bug-fix release, containing fixes for regressions in
6381 GCC 4.1.1 relative to previous releases of GCC.
6386 February 13, 2007 ([2]changes)
6389 May 24, 2006 ([3]changes)
6392 February 28, 2006 ([4]changes)
6394 References and Acknowledgements
6396 GCC used to stand for the GNU C Compiler, but since the compiler
6397 supports several other languages aside from C, it now stands for the
6398 GNU Compiler Collection.
6400 A list of [5]successful builds is updated as new information becomes
6403 The GCC developers would like to thank the numerous people that have
6404 contributed new features, improvements, bug fixes, and other changes as
6405 well as test results to GCC. This [6]amazing group of volunteers is
6406 what makes GCC successful.
6408 For additional information about GCC please refer to the [7]GCC project
6409 web site or contact the [8]GCC development mailing list.
6411 To obtain GCC please use [9]our mirror sites or [10]our SVN server.
6414 For questions related to the use of GCC, please consult these web
6415 pages and the [11]GCC manuals. If that fails, the
6416 [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6417 web pages and the development of GCC are welcome on our developer
6418 list at [13]gcc@gcc.gnu.org. All of [14]our lists have public
6421 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and
6422 distribution of this entire article is permitted in any medium,
6423 provided this notice is preserved.
6425 These pages are [16]maintained by the GCC team. Last modified
6430 1. http://www.gnu.org/
6431 2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
6432 3. http://gcc.gnu.org/gcc-4.1/changes.html
6433 4. http://gcc.gnu.org/gcc-4.1/changes.html
6434 5. http://gcc.gnu.org/gcc-4.1/buildstat.html
6435 6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6436 7. http://gcc.gnu.org/index.html
6437 8. mailto:gcc@gcc.gnu.org
6438 9. http://gcc.gnu.org/mirrors.html
6439 10. http://gcc.gnu.org/svn.html
6440 11. https://gcc.gnu.org/onlinedocs/
6441 12. mailto:gcc-help@gcc.gnu.org
6442 13. mailto:gcc@gcc.gnu.org
6443 14. https://gcc.gnu.org/lists.html
6444 15. http://www.fsf.org/
6445 16. https://gcc.gnu.org/about.html
6446 17. http://validator.w3.org/check/referer
6447 ======================================================================
6448 http://gcc.gnu.org/gcc-4.1/changes.html
6449 GCC 4.1 Release Series
6450 Changes, New Features, and Fixes
6452 The latest release in the 4.1 release series is [1]GCC 4.1.2.
6456 General Optimizer Improvements
6458 * GCC now has infrastructure for inter-procedural optimizations and
6459 the following inter-procedural optimizations are implemented:
6460 + Profile guided inlining. When doing profile feedback guided
6461 optimization, GCC can now use the profile to make better
6462 informed decisions on whether inlining of a function is
6463 profitable or not. This means that GCC will no longer inline
6464 functions at call sites that are not executed very often, and
6465 that functions at hot call sites are more likely to be
6467 A new parameter min-inline-recursive-probability is also now
6468 available to throttle recursive inlining of functions with
6469 small average recursive depths.
6470 + Discovery of pure and const functions, a form of side-effects
6471 analysis. While older GCC releases could also discover such
6472 special functions, the new IPA-based pass runs earlier so that
6473 the results are available to more optimizers. The pass is also
6474 simply more powerful than the old one.
6475 + Analysis of references to static variables and type escape
6476 analysis, also forms of side-effects analysis. The results of
6477 these passes allow the compiler to be less conservative about
6478 call-clobbered variables and references. This results in more
6479 redundant loads being eliminated and in making static
6480 variables candidates for register promotion.
6481 + Improvement of RTL-based alias analysis. The results of type
6482 escape analysis are fed to the RTL type-based alias analyzer,
6483 allowing it to disambiguate more memory references.
6484 + Interprocedural constant propagation and function versioning.
6485 This pass looks for functions that are always called with the
6486 same constant value for one or more of the function arguments,
6487 and propagates those constants into those functions.
6488 + GCC will now eliminate static variables whose usage was
6490 + -fwhole-program --combine can now be used to make all
6491 functions in program static allowing whole program
6492 optimization. As an exception, the main function and all
6493 functions marked with the new externally_visible attribute are
6494 kept global so that programs can link with runtime libraries.
6495 * GCC can now do a form of partial dead code elimination (PDCE) that
6496 allows code motion of expressions to the paths where the result of
6497 the expression is actually needed. This is not always a win, so the
6498 pass has been limited to only consider profitable cases. Here is an
6500 int foo (int *, int *);
6516 The a = b + c can be sunk to right before the printf. Normal code
6517 sinking will not do this, it will sink the first one above into the
6518 else-branch of the conditional jump, which still gives you two
6520 * GCC now has a value range propagation pass. This allows the
6521 compiler to eliminate bounds checks and branches. The results of
6522 the pass can also be used to accurately compute branch
6524 * The pass to convert PHI nodes to straight-line code (a form of
6525 if-conversion for GIMPLE) has been improved significantly. The two
6526 most significant improvements are an improved algorithm to
6527 determine the order in which the PHI nodes are considered, and an
6528 improvement that allow the pass to consider if-conversions of basic
6529 blocks with more than two predecessors.
6530 * Alias analysis improvements. GCC can now differentiate between
6531 different fields of structures in Tree-SSA's virtual operands form.
6532 This lets stores/loads from non-overlapping structure fields not
6533 conflict. A new algorithm to compute points-to sets was contributed
6534 that can allows GCC to see now that p->a and p->b, where p is a
6535 pointer to a structure, can never point to the same field.
6536 * Various enhancements to auto-vectorization:
6537 + Incrementally preserve SSA form when vectorizing.
6538 + Incrementally preserve loop-closed form when vectorizing.
6539 + Improvements to peeling for alignment: generate better code
6540 when the misalignment of an access is known at compile time,
6541 or when different accesses are known to have the same
6542 misalignment, even if the misalignment amount itself is
6544 + Consider dependence distance in the vectorizer.
6545 + Externalize generic parts of data reference analysis to make
6546 this analysis available to other passes.
6547 + Vectorization of conditional code.
6548 + Reduction support.
6549 * GCC can now partition functions in sections of hot and cold code.
6550 This can significantly improve performance due to better
6551 instruction cache locality. This feature works best together with
6552 profile feedback driven optimization.
6553 * A new pass to avoid saving of unneeded arguments to the stack in
6554 vararg functions if the compiler can prove that they will not be
6556 * Transition of basic block profiling to tree level implementation
6557 has been completed. The new implementation should be considerably
6558 more reliable (hopefully avoiding profile mismatch errors when
6559 using -fprofile-use or -fbranch-probabilities) and can be used to
6560 drive higher level optimizations, such as inlining.
6561 The -ftree-based-profiling command-line option was removed and
6562 -fprofile-use now implies disabling old RTL level loop optimizer
6563 (-fno-loop-optimize). Speculative prefetching optimization
6564 (originally enabled by -fspeculative-prefetching) was removed.
6566 New Languages and Language specific improvements
6570 * The old Bison-based C and Objective-C parser has been replaced by a
6571 new, faster hand-written recursive-descent parser.
6575 * The build infrastructure for the Ada runtime library and tools has
6576 been changed to be better integrated with the rest of the build
6577 infrastructure of GCC. This should make doing cross builds of Ada a
6582 * ARM-style name-injection of friend declarations is no longer the
6583 default. For example:
6589 will not be accepted; instead a declaration of f will need to be
6590 present outside of the scope of S. The new -ffriend-injection
6591 option will enable the old behavior.
6592 * The (undocumented) extension which permitted templates with default
6593 arguments to be bound to template template parameters with fewer
6594 parameters has been deprecated, and will be removed in the next
6595 major release of G++. For example:
6596 template <template <typename> class C>
6597 void f(C<double>) {}
6599 template <typename T, typename U = int>
6602 template void f(S<double>);
6604 makes use of the deprecated extension. The reason this code is not
6605 valid ISO C++ is that S is a template with two parameters;
6606 therefore, it cannot be bound to C which has only one parameter.
6608 Runtime Library (libstdc++)
6610 * Optimization work:
6611 + A new implementation of std::search_n is provided, better
6612 performing in case of random access iterators.
6613 + Added further efficient specializations of istream functions,
6614 i.e., character array and string extractors.
6615 + Other smaller improvements throughout.
6616 * Policy-based associative containers, designed for high-performance,
6617 flexibility and semantic safety are delivered in ext/pb_assoc.
6618 * A versatile string class, __gnu_cxx::__versa_string, providing
6619 facilities conforming to the standard requirements for
6620 basic_string, is delivered in <ext/vstring.h>. In particular:
6621 + Two base classes are provided: the default one avoids
6622 reference counting and is optimized for short strings; the
6623 alternate one, still uses it while improving in a few low
6624 level areas (e.g., alignment). See vstring_fwd.h for some
6626 + Various algorithms have been rewritten (e.g., replace), the
6627 code streamlined and simple optimizations added.
6628 + Option 3 of DR 431 is implemented for both available bases,
6629 thus improving the support for stateful allocators.
6630 * As usual, many bugs have been fixed (e.g., libstdc++/13583,
6631 libstdc++/23953) and LWG resolutions put into effect for the first
6632 time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1
6633 Issue 6.19). The implementation status of TR1 is now tracked in the
6638 * A new language front end for Objective-C++ has been added. This
6639 language allows users to mix the object oriented features of
6640 Objective-C with those of C++.
6644 * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19
6645 features (plus some 0.20 bug-fixes)
6647 o The java.net.HttpURLConnection implementation no longer
6648 buffers the entire response body in memory. This means
6649 that response bodies larger than available memory can now
6652 o NIO FileChannel.map implementation, fast bulk put
6653 implementation for DirectByteBuffer (speeds up this
6655 o FileChannel.lock() and FileChannel.force() implemented.
6657 o gnu.xml fix for nodes created outside a namespace
6659 o Add support for output indenting and
6660 cdata-section-elements output instruction in
6662 o xml.xpath corrections for cases where elements/attributes
6663 might have been created in non-namespace-aware mode.
6664 Corrections to handling of XSL variables and minor
6665 conformance updates.
6667 o GNU JAWT implementation, the AWT Native Interface, which
6668 allows direct access to native screen resources from
6669 within a Canvas's paint method. GNU Classpath Examples
6670 comes with a Demo, see libjava/classpath/examples/README.
6671 o awt.datatransfer updated to 1.5 with support for
6672 FlavorEvents. The gtk+ awt peers now allow copy/paste of
6673 text, images, URIs/files and serialized objects with
6674 other applications and tracking clipboard change events
6675 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized
6676 objects are supported). A GNU Classpath Examples
6677 datatransfer Demo was added to show the new
6679 o Split gtk+ awt peers event handling in two threads and
6680 improve gdk lock handling (solves several awt lock ups).
6681 o Speed up awt Image loading.
6682 o Better gtk+ scrollbar peer implementation when using gtk+
6684 o Handle image loading errors correctly for gdkpixbuf and
6686 o Better handle GDK lock. Properly prefix gtkpeer native
6688 o GdkGraphics2D has been updated to use Cairo 0.5.x or
6690 o BufferedImage and GtkImage rewrites. All image drawing
6691 operations should now work correctly (flipping requires
6693 o Future Graphics2D, image and text work is documented at:
6694 [2]http://developer.classpath.org/mediation/ClasspathGrap
6696 o When gtk+ 2.6 or higher is installed the default log
6697 handler will produce stack traces whenever a WARNING,
6698 CRITICAL or ERROR message is produced.
6700 o The RepaintManager has been reworked for more efficient
6701 painting, especially for large GUIs.
6702 o The layout manager OverlayLayout has been implemented,
6703 the BoxLayout has been rewritten to make use of the
6704 SizeRequirements utility class and caching for more
6706 o Improved accessibility support.
6707 o Significant progress has been made in the implementation
6708 of the javax.swing.plaf.metal package, with most UI
6709 delegates in a working state now. Please test this with
6710 your own applications and provide feedback that will help
6711 us to improve this package.
6712 o The GUI demo (gnu.classpath.examples.swing.Demo) has been
6713 extended to highlight various features in our Free Swing
6714 implementation. And it includes a look and feel switcher
6715 for Metal (default), Ocean and GNU themes.
6716 o The javax.swing.plaf.multi package is now implemented.
6717 o Editing and several key actions for JTree and JTable were
6719 o Lots of icons and look and feel improvements for Free
6720 Swing basic and metal themes were added. Try running the
6721 GNU Classpath Swing Demo in examples
6722 (gnu.classpath.examples.swing.Demo) with:
6723 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee
6725 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee
6727 o Start of styled text capabilites for java.swing.text.
6728 o DefaultMutableTreeNode pre-order, post-order, depth-first
6729 and breadth-first traversal enumerations implemented.
6730 o JInternalFrame colors and titlebar draw properly.
6731 o JTree is working up to par (icons, selection and keyboard
6733 o JMenus were made more compatible in visual and
6734 programmatic behavior.
6735 o JTable changeSelection and multiple selections
6737 o JButton and JToggleButton change states work properly
6739 o JFileChooser fixes.
6740 o revalidate() and repaint() fixes which make Free Swing
6741 much more responsive.
6742 o MetalIconFactory implemented.
6743 o Free Swing Top-Level Compatibility. JFrame, JDialog,
6744 JApplet, JInternalFrame, and JWindow are now 1.5
6745 compatible in the sense that you can call add() and
6746 setLayout() directly on them, which will have the same
6747 effect as calling getContentPane().add() and
6748 getContentPane().setLayout().
6749 o The JTree interface has been completed. JTrees now
6750 recognizes mouse clicks and selections work.
6751 o BoxLayout works properly now.
6752 o Fixed GrayFilter to actually work.
6753 o Metal SplitPane implemented.
6754 o Lots of Free Swing text and editor stuff work now.
6755 + Free RMI and Corba
6756 o Andrew Watson, Vice President and Technical Director of
6757 the Object Management Group, has officially assigned us
6758 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will
6759 mark remote classpath-specific system exceptions.
6760 Obtaining the VMCID means that GNU Classpath now is a
6761 recogniseable type of node in a highly interoperable
6763 o GNU Classpath now includes the first working draft to
6764 support the RMI over IIOP protocol. The current
6765 implementation is capable of remote invocations,
6766 transferring various Serializables and Externalizables
6767 via RMI-IIOP protocol. It can flatten graphs and, at
6768 least for the simple cases, is interoperable with 1.5
6770 o org.omg.PortableInterceptor and related functionality in
6771 other packages is now implemented:
6772 # The sever and client interceptors work as required
6774 # The IOR interceptor works as needed for 1.5.
6775 o The org.omg.DynamicAny package is completed and passes
6777 o The Portable Object Adapter should now support the output
6778 of the recent IDL to java compilers. These compilers now
6779 generate servants and not CORBA objects as before, making
6780 the output depend on the existing POA implementation.
6781 Completing POA means that such code can already be tried
6782 to run on Classpath. Our POA is tested for the following
6784 # POA converts servant to the CORBA object.
6785 # Servant provides to the CORBA object.
6786 # POA activates new CORBA object with the given Object
6787 Id (byte array) that is later accessible for the
6789 # During the first call, the ServantActivator provides
6790 servant for this and all subsequent calls on the
6792 # During each call, the ServantLocator provides
6793 servant for this call only.
6794 # ServantLocator or ServantActivator forwards call to
6796 # POA has a single servant, responsible for all
6798 # POA has a default servant, but some objects are
6799 explicitly connected to they specific servants.
6800 The POA is verified using tests from the former
6802 o The CORBA implementation is now a working prototype that
6803 should support features up to 1.3 inclusive. We invite
6804 groups writing CORBA dependent applications to try
6805 Classpath implementation, reporting any possible bugs.
6806 The CORBA prototype is interoperable with Sun's
6807 implementation v 1.4, transferring object references,
6808 primitive types, narrow and wide strings, arrays,
6809 structures, trees, abstract interfaces and value types
6810 (feature of CORBA 2.3) between these two platforms.
6811 Remote exceptions are transferred and handled correctly.
6812 The stringified object references (IORs) from various
6813 sources are parsed as required. The transient (for
6814 current session) and permanent (till jre restart)
6815 redirections work. Both Little and Big Endian encoded
6816 messages are accepted. The implementation is verified
6817 using tests from the former cost.omg.org. The current
6818 release includes working examples (see the examples
6819 directory), demonstrating the client-server
6820 communication, using either CORBA Request or IDL-based
6821 stub (usually generated by a IDL to java compiler). These
6822 examples also show how to use the Classpath CORBA naming
6823 service. The IDL to java compiler is not yet written, but
6824 as our library must be compatible, it naturally accepts
6825 the output of other idlj implementations.
6827 o Updated TimeZone data against Olson tzdata2005l.
6828 o Make zip and jar packages UTF-8 clean.
6829 o "native" code builds and compiles (warning free) on
6831 o java.util.logging.FileHandler now rotates files.
6832 o Start of a generic JDWP framework in gnu/classpath/jdwp.
6833 This is unfinished, but feedback (at classpath@gnu.org)
6834 from runtime hackers is greatly appreciated. Although
6835 most of the work is currently being done around gcj/gij
6836 we want this framework to be as VM neutral as possible.
6837 Early design is described in:
6838 [3]https://gcc.gnu.org/ml/java/2005-05/msg00260.html
6839 o QT4 AWT peers, enable by giving configure
6840 --enable-qt-peer. Included, but not ready for production
6841 yet. They are explicitly disabled and not supported. But
6842 if you want to help with the development of these new
6843 features we are interested in feedback. You will have to
6844 explicitly enable them to try them out (and they will
6845 most likely contain bugs).
6846 o Documentation fixes all over the place. See
6847 [4]http://developer.classpath.org/doc/
6849 New Targets and Target Specific Improvements
6853 * The x86-64 medium model (that allows building applications whose
6854 data segment exceeds 4GB) was redesigned to match latest ABI draft.
6855 New implementation split large datastructures into separate segment
6856 improving performance of accesses to small datastructures and also
6857 allows linking of small model libraries into medium model programs
6858 as long as the libraries are not accessing the large datastructures
6859 directly. Medium model is also supported in position independent
6861 The ABI change results in partial incompatibility among medium
6862 model objects. Linking medium model libraries (or objects) compiled
6863 with new compiler into medium model program compiled with older
6864 will likely result in exceeding ranges of relocations.
6865 Binutils 2.16.91 or newer are required for compiling medium model
6868 RS6000 (POWER/PowerPC)
6870 * The AltiVec vector primitives in <altivec.h> are now implemented in
6871 a way that puts a smaller burden on the preprocessor, instead
6872 processing the "overloading" in the front ends. This should benefit
6873 compilation speed on AltiVec vector code.
6874 * AltiVec initializers now are generated more efficiently.
6875 * The popcountb instruction available on POWER5 now is generated.
6876 * The floating point round to integer instructions available on
6877 POWER5+ now is generated.
6878 * Floating point divides can be synthesized using the floating point
6879 reciprocal estimate instructions.
6880 * Double precision floating point constants are initialized as single
6881 precision values if they can be represented exactly.
6883 S/390, zSeries and System z9
6885 * Support for the IBM System z9 109 processor has been added. When
6886 using the -march=z9-109 option, the compiler will generate code
6887 making use of instructions provided by the extended immediate
6889 * Support for 128-bit IEEE floating point has been added. When using
6890 the -mlong-double-128 option, the compiler will map the long double
6891 data type to 128-bit IEEE floating point. Using this option
6892 constitutes an ABI change, and requires glibc support.
6893 * Various changes to improve performance of generated code have been
6894 implemented, including:
6895 + In functions that do not require a literal pool, register %r13
6896 (which is traditionally reserved as literal pool pointer), can
6897 now be freely used for other purposes by the compiler.
6898 + More precise tracking of register use allows the compiler to
6899 generate more efficient function prolog and epilog code in
6901 + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING
6902 instructions are now used to implement C string functions.
6903 + The MOVE CHARACTER instruction with single byte overlap is now
6904 used to implement the memset function with non-zero fill byte.
6905 + The LOAD ZERO instructions are now used where appropriate.
6906 + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK,
6907 and INSERT IMMEDIATE instructions are now used more frequently
6908 to optimize bitfield operations.
6909 + The BRANCH ON COUNT instruction is now used more frequently.
6910 In particular, the fact that a loop contains a subroutine call
6911 no longer prevents the compiler from using this instruction.
6912 + The compiler is now aware that all shift and rotate
6913 instructions implicitly truncate the shift count to six bits.
6914 * Back-end support for the following generic features has been
6916 + The full set of [5]built-in functions for atomic memory
6918 + The -fstack-protector feature.
6919 + The optimization pass avoiding unnecessary stores of incoming
6920 argument registers in functions with variable argument list.
6924 * The default code model in 64-bit mode has been changed from
6925 Medium/Anywhere to Medium/Middle on Solaris.
6926 * TLS support is disabled by default on Solaris prior to release 10.
6927 It can be enabled on TLS-capable Solaris 9 versions (4/04 release
6928 and later) by specifying --enable-tls at configure time.
6932 * Support has been added for this new architecture.
6936 Documentation improvements
6938 Other significant improvements
6940 * GCC can now emit code for protecting applications from
6941 stack-smashing attacks. The protection is realized by buffer
6942 overflow detection and reordering of stack variables to avoid
6944 * Some built-in functions have been fortified to protect them against
6945 various buffer overflow (and format string) vulnerabilities.
6946 Compared to the mudflap bounds checking feature, the safe builtins
6947 have far smaller overhead. This means that programs built using
6948 safe builtins should not experience any measurable slowdown.
6952 This is the [6]list of problem reports (PRs) from GCC's bug tracking
6953 system that are known to be fixed in the 4.1.2 release. This list might
6954 not be complete (that is, it is possible that some PRs that have been
6955 fixed are not listed here).
6957 When generating code for a shared library, GCC now recognizes that
6958 global functions may be replaced when the program runs. Therefore, it
6959 is now more conservative in deducing information from the bodies of
6960 functions. For example, in this example:
6965 cout << "Exception";
6969 G++ would previously have optimized away the catch clause, since it
6970 would have concluded that f cannot throw exceptions. Because users may
6971 replace f with another function in the main body of the program, this
6972 optimization is unsafe, and is no longer performed. If you wish G++ to
6973 continue to optimize as before, you must add a throw() clause to the
6974 declaration of f to make clear that it does not throw exceptions.
6977 For questions related to the use of GCC, please consult these web
6978 pages and the [7]GCC manuals. If that fails, the
6979 [8]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6980 web pages and the development of GCC are welcome on our developer
6981 list at [9]gcc@gcc.gnu.org. All of [10]our lists have public
6984 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and
6985 distribution of this entire article is permitted in any medium,
6986 provided this notice is preserved.
6988 These pages are [12]maintained by the GCC team. Last modified
6993 1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
6994 2. http://developer.classpath.org/mediation/ClasspathGraphicsImagesText
6995 3. https://gcc.gnu.org/ml/java/2005-05/msg00260.html
6996 4. http://developer.classpath.org/doc/
6997 5. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html
6998 6. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2
6999 7. https://gcc.gnu.org/onlinedocs/
7000 8. mailto:gcc-help@gcc.gnu.org
7001 9. mailto:gcc@gcc.gnu.org
7002 10. https://gcc.gnu.org/lists.html
7003 11. http://www.fsf.org/
7004 12. https://gcc.gnu.org/about.html
7005 13. http://validator.w3.org/check/referer
7006 ======================================================================
7007 http://gcc.gnu.org/gcc-4.0/index.html
7008 GCC 4.0 Release Series
7012 The [1]GNU project and the GCC developers are pleased to announce the
7013 release of GCC 4.0.4.
7015 This release is a bug-fix release, containing fixes for regressions in
7016 GCC 4.0.3 relative to previous releases of GCC.
7021 January 31, 2007 ([2]changes)
7024 March 10, 2006 ([3]changes)
7027 September 28, 2005 ([4]changes)
7030 July 7, 2005 ([5]changes)
7033 April 20, 2005 ([6]changes)
7035 References and Acknowledgements
7037 GCC used to stand for the GNU C Compiler, but since the compiler
7038 supports several other languages aside from C, it now stands for the
7039 GNU Compiler Collection.
7041 A list of [7]successful builds is updated as new information becomes
7044 The GCC developers would like to thank the numerous people that have
7045 contributed new features, improvements, bug fixes, and other changes as
7046 well as test results to GCC. This [8]amazing group of volunteers is
7047 what makes GCC successful.
7049 For additional information about GCC please refer to the [9]GCC project
7050 web site or contact the [10]GCC development mailing list.
7052 To obtain GCC please use [11]our mirror sites, or [12]our SVN server.
7055 For questions related to the use of GCC, please consult these web
7056 pages and the [13]GCC manuals. If that fails, the
7057 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
7058 web pages and the development of GCC are welcome on our developer
7059 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
7062 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
7063 distribution of this entire article is permitted in any medium,
7064 provided this notice is preserved.
7066 These pages are [18]maintained by the GCC team. Last modified
7071 1. http://www.gnu.org/
7072 2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
7073 3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3
7074 4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2
7075 5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1
7076 6. http://gcc.gnu.org/gcc-4.0/changes.html
7077 7. http://gcc.gnu.org/gcc-4.0/buildstat.html
7078 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
7079 9. http://gcc.gnu.org/index.html
7080 10. mailto:gcc@gcc.gnu.org
7081 11. http://gcc.gnu.org/mirrors.html
7082 12. http://gcc.gnu.org/svn.html
7083 13. https://gcc.gnu.org/onlinedocs/
7084 14. mailto:gcc-help@gcc.gnu.org
7085 15. mailto:gcc@gcc.gnu.org
7086 16. https://gcc.gnu.org/lists.html
7087 17. http://www.fsf.org/
7088 18. https://gcc.gnu.org/about.html
7089 19. http://validator.w3.org/check/referer
7090 ======================================================================
7091 http://gcc.gnu.org/gcc-4.0/changes.html
7092 GCC 4.0 Release Series
7093 Changes, New Features, and Fixes
7095 The latest release in the 4.0 release series is [1]GCC 4.0.4.
7099 * GCC now generates location lists by default when compiling with
7100 debug info and optimization.
7101 + GDB 6.0 and older crashes when it sees location lists. GDB 6.1
7102 or later is needed to debug binaries containing location
7104 + When you are trying to view a value of a variable in a part of
7105 a function where it has no location (for example when the
7106 variable is no longer used and thus its location was used for
7107 something else) GDB will say that it is not available.
7108 You can disable generating location lists by -fno-var-tracking.
7109 * GCC no longer accepts the -fwritable-strings option. Use named
7110 character arrays when you need a writable string.
7111 * The options -freduce-all-givs and -fmove-all-movables have been
7112 discontinued. They were used to circumvent a shortcoming in the
7113 heuristics of the old loop optimization code with respect to common
7114 Fortran constructs. The new (tree) loop optimizer works differently
7115 and doesn't need those work-arounds.
7116 * The graph-coloring register allocator, formerly enabled by the
7117 option -fnew-ra, has been discontinued.
7118 * -I- has been deprecated. -iquote is meant to replace the need for
7120 * The MIPS -membedded-pic and -mrnames options have been removed.
7121 * All MIPS targets now require the GNU assembler. In particular, IRIX
7122 configurations can no longer use the MIPSpro assemblers, although
7123 they do still support the MIPSpro linkers.
7124 * The SPARC option -mflat has been removed.
7125 * English-language diagnostic messages will now use Unicode quotation
7126 marks in UTF-8 locales. (Non-English messages already used the
7127 quotes appropriate for the language in previous releases.) If your
7128 terminal does not support UTF-8 but you are using a UTF-8 locale
7129 (such locales are the default on many GNU/Linux systems) then you
7130 should set LC_CTYPE=C in the environment to disable that locale.
7131 Programs that parse diagnostics and expect plain ASCII
7132 English-language messages should set LC_ALL=C. See [2]Markus Kuhn's
7133 explanation of Unicode quotation marks for more information.
7134 * The specs file is no longer installed on most platforms. Most users
7135 will be totally unaffected. However, if you are accustomed to
7136 editing the specs file yourself, you will now have to use the
7137 -dumpspecs option to generate the specs file, and then edit the
7140 General Optimizer Improvements
7142 * The [3]tree ssa branch has been merged. This merge has brought in a
7143 completely new optimization framework based on a higher level
7144 intermediate representation than the existing RTL representation.
7145 Numerous new code transformations based on the new framework are
7146 available in GCC 4.0, including:
7147 + Scalar replacement of aggregates
7148 + Constant propagation
7149 + Value range propagation
7150 + Partial redundancy elimination
7151 + Load and store motion
7152 + Strength reduction
7153 + Dead store elimination
7154 + Dead and unreachable code elimination
7155 + [4]Autovectorization
7157 + Tail recursion by accumulation
7158 Many of these passes outperform their counterparts from previous
7160 * [5]Swing Modulo Scheduling (SMS). An RTL level instruction
7161 scheduling optimization intended for loops that perform heavy
7164 New Languages and Language specific improvements
7168 * The sentinel attribute has been added to GCC. This function
7169 attribute allows GCC to warn when variadic functions such as execl
7170 are not NULL terminated. See the GCC manual for a complete
7171 description of its behavior.
7172 * Given __attribute__((alias("target"))) it is now an error if target
7173 is not a symbol, defined in the same translation unit. This also
7174 applies to aliases created by #pragma weak alias=target. This is
7175 because it's meaningless to define an alias to an undefined symbol.
7176 On Solaris, the native assembler would have caught this error, but
7181 * The -Wstrict-aliasing=2 option has been added. This warning catches
7182 all unsafe cases, but it may also give a warning for some cases
7184 * The cast-as-lvalue, conditional-expression-as-lvalue and
7185 compound-expression-as-lvalue extensions, which were deprecated in
7186 3.3.4 and 3.4, have been removed.
7187 * The -fwritable-strings option, which was deprecated in 3.4, has
7189 * #pragma pack() semantics have been brought closer to those used by
7190 other compilers. This also applies to C++.
7191 * Taking the address of a variable with register storage is invalid
7192 in C. GCC now issues an error instead of a warning.
7193 * Arrays of incomplete element type are invalid in C. GCC now issues
7194 an error for such arrays. Declarations such as extern struct s x[];
7195 (where struct s has not been defined) can be moved after the
7196 definition of struct s. Function parameters declared as arrays of
7197 incomplete type can instead be declared as pointers.
7201 * When compiling without optimizations (-O0), the C++ frontend is
7202 much faster than in any previous versions of GCC. Independent
7203 testers have measured speed-ups up to 25% in real-world production
7204 code, compared to the 3.4 family (which was already the fastest
7205 version to date). Upgrading from older versions might show even
7206 bigger improvements.
7207 * ELF visibility attributes can now be applied to a class type, so
7208 that it affects every member function of a class at once, without
7209 having to specify each individually:
7210 class __attribute__ ((visibility("hidden"))) Foo
7215 The syntax is deliberately similar to the __declspec() system used
7216 by Microsoft Windows based compilers, allowing cross-platform
7217 projects to easily reuse their existing macro system for denoting
7218 exports and imports. By explicitly marking internal classes never
7219 used outside a binary as hidden, one can completely avoid PLT
7220 indirection overheads during their usage by the compiler. You can
7221 find out more about the advantages of this at
7222 [6]http://www.akkadia.org/drepper/dsohowto.pdf
7223 * The -fvisibility-inlines-hidden option has been added which marks
7224 all inlineable functions as having hidden ELF visibility, thus
7225 removing their symbol and typeinfo from the exported symbol table
7226 of the output ELF binary. Using this option can reduce the exported
7227 symbol count of template-heavy code by up to 40% with no code
7228 change at all, thus notably improving link and load times for the
7229 binary as well as a reduction in size of up to 10%. Also, check the
7230 new [7]-fvisibility option.
7231 * The compiler now uses the library interface specified by the [8]C++
7232 ABI for thread-safe initialization of function-scope static
7233 variables. Most users should leave this alone, but embedded
7234 programmers may want to disable this by specifying
7235 -fno-threadsafe-statics for a small savings in code size.
7236 * Taking the address of an explicit register variable is no longer
7237 supported. Note that C++ allows taking the address of variables
7238 with register storage so this will continue to compile with a
7239 warning. For example, assuming that r0 is a machine register:
7240 register int foo asm ("r0");
7242 &foo; // error, no longer accepted
7243 &bar; // OK, with a warning
7244 * G++ has an undocumented extension to virtual function covariancy
7245 rules that allowed the overrider to return a type that was
7246 implicitly convertable to the overridden function's return type.
7247 For instance a function returning void * could be overridden by a
7248 function returning T *. This is now deprecated and will be removed
7249 in a future release.
7250 * The G++ minimum and maximum operators (<? and >?) and their
7251 compound forms (<?=) and >?=) have been deprecated and will be
7252 removed in a future version. Code using these operators should be
7253 modified to use std::min and std::max instead.
7254 * Declaration of nested classes of class templates as friends are
7256 template <typename T> struct A {
7260 template <typename T> friend class A<T>::B;
7262 This complements the feature member functions of class templates as
7263 friends introduced in GCC 3.4.0.
7264 * When declaring a friend class using an unqualified name, classes
7265 outside the innermost non-class scope are not searched:
7269 friend class A; // Refer to N::A which has not been declared yet
7270 // because name outside namespace N are not searched
7271 friend class ::A; // Refer to ::A
7274 Hiding the friend name until declaration is still not implemented.
7275 * Friends of classes defined outside their namespace are correctly
7281 friend class B; // Refer to N::B in GCC 4.0.0
7282 // but ::B in earlier versions of GCC
7285 Runtime Library (libstdc++)
7287 * Optimization work:
7288 + Added efficient specializations of istream functions for char
7290 + Further performance tuning of strings, in particular wrt
7291 single-char append and getline.
7292 + iter_swap - and therefore most of the mutating algorithms -
7293 now makes an unqualified call to swap when the value_type of
7294 the two iterators is the same.
7295 * A large subset of the features in Technical Report 1 (TR1 for
7296 short) is experimentally delivered (i.e., no guarantees about the
7297 implementation are provided. In particular it is not promised that
7298 the library will remain link-compatible when code using TR1 is
7300 + General utilities such as reference_wrapper and shared_ptr.
7301 + Function objects, i.e., result_of, mem_fn, bind, function.
7302 + Support for metaprogramming.
7303 + New containers such as tuple, array, unordered_set,
7304 unordered_map, unordered_multiset, unordered_multimap.
7305 * As usual, many bugs have been fixed and LWG resolutions implemented
7306 for the first time (e.g., DR 409).
7310 * In order to prevent naming conflicts with other implementations of
7311 these tools, some GCJ binaries have been renamed:
7312 + rmic is now grmic,
7313 + rmiregistry is now grmiregistry, and
7314 + jar is now fastjar.
7315 In particular, these names were problematic for the jpackage.org
7316 packaging conventions which install symlinks in /usr/bin that point
7317 to the preferred versions of these tools.
7318 * The -findirect-dispatch argument to the compiler now works and
7319 generates code following a new "binary compatibility" ABI. Code
7320 compiled this way follows the binary compatibility rules of the
7321 Java Language Specification.
7322 * libgcj now has support for using GCJ as a JIT, using the
7323 gnu.gcj.jit family of system properties.
7324 * libgcj can now find a shared library corresponding to the bytecode
7325 representation of a class. See the documentation for the new
7326 gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system
7328 * There have been many improvements to the class library. Here are
7330 + Much more of AWT and Swing exist.
7331 + Many new packages and classes were added, including
7332 java.util.regex, java.net.URI, javax.crypto,
7333 javax.crypto.interfaces, javax.crypto.spec, javax.net,
7334 javax.net.ssl, javax.security.auth,
7335 javax.security.auth.callback, javax.security.auth.login,
7336 javax.security.auth.x500, javax.security.sasl, org.ietf.jgss,
7337 javax.imageio, javax.imageio.event, javax.imageio.spi,
7338 javax.print, javax.print.attribute,
7339 javax.print.attribute.standard, javax.print.event, and
7341 + Updated SAX and DOM, and imported GNU JAXP
7345 * A new [9]Fortran front end has replaced the aging GNU Fortran 77
7346 front end. The new front end supports Fortran 90 and Fortran 95. It
7347 may not yet be as stable as the old Fortran front end.
7351 * Ada (with tasking and Zero Cost Exceptions) is now available on
7352 many more targets, including but not limited to: alpha-linux,
7353 hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux,
7354 s390x-linux, sparc-linux.
7355 * Some of the new Ada 2005 features are now implemented like
7356 Wide_Wide_Character and Ada.Containers.
7357 * Many bugs have been fixed, tools and documentation improved.
7358 * To compile Ada from the sources, install an older working Ada
7359 compiler and then use --enable-languages=ada at configuration time,
7360 since the Ada frontend is not currently activated by default. See
7361 the [10]Installing GCC for details.
7363 New Targets and Target Specific Improvements
7367 * The frame layout has changed. In the new layout, the prologue of a
7368 function first saves registers and then allocate space for locals,
7369 resulting in an 1% improvement on code size.
7371 IA-32/x86-64 (AMD64)
7373 * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10,
7374 log1p, log2, logb and tan mathematical builtins (and their float
7375 and long double variants) are now implemented as inline x87
7376 intrinsics when using -ffast-math.
7377 * The ceil, floor, nearbyint, rint and trunc mathematical builtins
7378 (and their float and long double variants) are now implemented as
7379 inline x87 intrinsics when using -ffast-math.
7380 * The x87's fsincos instruction is now used automatically with
7381 -ffast-math when calculating both the sin and cos of the same
7383 * Instruction selection for multiplication and division by constants
7388 * Floating point division, integer division and sqrt are now inlined,
7389 resulting in significant performance improvements on some codes.
7393 * Division by zero checks now use conditional traps if the target
7394 processor supports them. This decreases code size by one word per
7395 division operation. The old behavior (branch and break) can be
7396 obtained either at configure time by passing --with-divide=breaks
7397 to configure or at runtime by passing -mdivide-breaks to GCC.
7398 * Support for MIPS64 paired-single instructions has been added. It is
7399 enabled by -mpaired-single and can be accessed using both the
7400 target-independent vector extensions and new MIPS-specific built-in
7402 * Support for the MIPS-3D ASE has been added. It is enabled by
7403 -mips3d and provides new MIPS-3D-specific built-in functions.
7404 * The -mexplicit-relocs option now supports static n64 code (as is
7405 used, for example, in 64-bit linux kernels). -mexplicit-relocs
7406 should now be feature-complete and is enabled by default when GCC
7407 is configured to use a compatible assembler.
7408 * Support for the NEC VR4130 series has been added. This support
7409 includes the use of VR-specific instructions and a new VR4130
7410 scheduler. Full VR4130 support can be selected with -march=vr4130
7411 while code for any ISA can be tuned for the VR4130 using
7412 -mtune=vr4130. There is also a new -mvr4130-align option that
7413 produces better schedules at the cost of increased code size.
7414 * Support for the Broadcom SB-1 has been extended. There is now an
7415 SB-1 scheduler as well as support for the SB-1-specific
7416 paired-single instructions. Full SB-1 support can be selected with
7417 -march=sb1 while code for any ISA can be optimized for the SB-1
7419 * The compiler can now work around errata in R4000, R4400, VR4120 and
7420 VR4130 processors. These workarounds are enabled by -mfix-r4000,
7421 -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120
7422 and VR4130 workarounds need binutils 2.16 or above.
7423 * IRIX shared libraries are now installed into the standard library
7424 directories: o32 libraries go into lib/, n32 libraries go into
7425 lib32/ and n64 libraries go into lib64/.
7426 * The compiler supports a new -msym32 option. It can be used to
7427 optimize n64 code in which all symbols are known to have 32-bit
7432 * New command-line options help to generate code intended to run in
7433 an environment where stack space is restricted, e.g. Linux kernel
7435 + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time
7436 warnings for single functions that require large or dynamic
7438 + -mstack-size and -mstack-guard generate code that checks for
7439 stack overflow at run time.
7440 + -mpacked-stack generates code that reduces the stack frame
7441 size of many functions by reusing unneeded parts of the stack
7443 * The -msoft-float option now ensures that generated code never
7444 accesses floating point registers.
7445 * The s390x-ibm-tpf target now fully supports C++, including
7446 exceptions and threads.
7447 * Various changes to improve performance of the generated code have
7448 been implemented, including:
7449 + GCC now uses sibling calls where possible.
7450 + Condition code handling has been optimized, allowing GCC to
7451 omit redundant comparisons in certain cases.
7452 + The cost function guiding many optimizations has been refined
7453 to more accurately represent the z900 and z990 processors.
7454 + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW
7455 instructions are now used to avoid conditional branches in
7457 + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to
7458 optimize address arithmetic required to access large stack
7460 + GCC now makes more efficient use of memory-to-memory type
7461 instructions (MVC, CLC, ...).
7462 + More precise tracking of special register use allows better
7463 instruction scheduling, in particular of the function prologue
7464 and epilogue sequences.
7465 + The Java front end now generates inline code to implement
7466 integer division, instead of calling library routines.
7470 * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and
7471 -mf934 have been removed. They have been replaced with -mcpu=xxx.
7472 * The internal model used to estimate the relative cost of each
7473 instruction has been updated. It is expected to give better results
7474 on recent UltraSPARC processors.
7475 * Code generation for function prologues and epilogues has been
7476 improved, resulting in better scheduling and allowing multiple exit
7477 points in functions.
7478 * Support for Sun's Visual Instruction Set (VIS) has been enhanced.
7479 It is enabled by -mvis and provides new built-in functions for VIS
7480 instructions on UltraSPARC processors.
7481 * The option -mapp-regs has been turned on by default on Solaris too.
7485 * Novell NetWare (on ix86, no other hardware platform was ever really
7486 supported by this OS) has been re-enabled and the ABI supported by
7487 GCC has been brought into sync with that of MetroWerks CodeWarrior
7488 (the ABI previously supported was that of some Unix systems, which
7489 NetWare never tried to support).
7493 Support for a number of older systems has been declared obsolete in GCC
7494 4.0. Unless there is activity to revive them, the next release of GCC
7495 will have their sources permanently removed.
7497 All GCC ports for the following processor architectures have been
7501 * National Semiconductor NS32K
7502 * Texas Instruments TMS320C[34]x
7504 Also, those for some individual systems have been obsoleted:
7506 + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf,
7508 + OpenBSD 32-bit (sparc-*-openbsd*)
7510 Documentation improvements
7512 Other significant improvements
7514 * Location lists are now generated by default when compiling with
7515 debug info and optimization. Location lists provide more accurate
7516 debug info about locations of variables and they allow debugging
7517 code compiled with -fomit-frame-pointer.
7518 * The -fvisibility option has been added which allows the default ELF
7519 visibility of all symbols to be set per compilation and the new
7520 #pragma GCC visibility preprocessor command allows the setting of
7521 default ELF visibility for a region of code. Using
7522 -fvisibility=hidden especially in combination with the new
7523 -fvisibility-inlines-hidden can yield substantial improvements in
7524 output binary quality including avoiding PLT indirection overheads,
7525 reduction of the exported symbol count by up to 60% (with resultant
7526 improvements to link and load times), better scope for the
7527 optimizer to improve code and up to a 20% reduction in binary size.
7528 Using these options correctly yields a binary with a similar symbol
7529 count to a Windows DLL.
7530 Perhaps more importantly, this new feature finally allows (with
7531 careful planning) complete avoidance of symbol clashes when
7532 manually loading shared objects with RTLD_GLOBAL, thus finally
7533 solving problems many projects such as python were forced to use
7534 RTLD_LOCAL for (with its resulting issues for C++ correctness). You
7535 can find more information about using these options at
7536 [11]https://gcc.gnu.org/wiki/Visibility.
7537 __________________________________________________________________
7541 This is the [12]list of problem reports (PRs) from GCC's bug tracking
7542 system that are known to be fixed in the 4.0.1 release. This list might
7543 not be complete (that is, it is possible that some PRs that have been
7544 fixed are not listed here).
7548 This is the [13]list of problem reports (PRs) from GCC's bug tracking
7549 system that are known to be fixed in the 4.0.2 release. This list might
7550 not be complete (that is, it is possible that some PRs that have been
7551 fixed are not listed here).
7553 Unfortunately, due to a release engineering failure, this release has a
7554 regression on Solaris that will affect some C++ programs. We suggest
7555 that Solaris users apply a [14]patch that corrects the problem. Users
7556 who do not wish to apply the patch should explicitly link C++ programs
7557 with the -pthreads option, even if they do not use threads. This
7558 problem has been corrected in the current 4.0 branch sources and will
7559 not be present in GCC 4.0.3.
7563 Starting with this release, the function getcontext is recognized by
7564 the compiler as having the same semantics as the setjmp function. In
7565 particular, the compiler will ensure that all registers are dead before
7566 calling such a function and will emit a warning about the variables
7567 that may be clobbered after the second return from the function.
7571 This is the [15]list of problem reports (PRs) from GCC's bug tracking
7572 system that are known to be fixed in the 4.0.4 release. This list might
7573 not be complete (that is, it is possible that some PRs that have been
7574 fixed are not listed here).
7576 The 4.0.4 release is provided for those that require a high degree of
7577 binary compatibility with previous 4.0.x releases. For most users, the
7578 GCC team recommends that version 4.1.1 or later be used instead."
7581 For questions related to the use of GCC, please consult these web
7582 pages and the [16]GCC manuals. If that fails, the
7583 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
7584 web pages and the development of GCC are welcome on our developer
7585 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
7588 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
7589 distribution of this entire article is permitted in any medium,
7590 provided this notice is preserved.
7592 These pages are [21]maintained by the GCC team. Last modified
7597 1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
7598 2. http://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html
7599 3. http://gcc.gnu.org/projects/tree-ssa/
7600 4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html
7601 5. http://gcc.gnu.org/news/sms.html
7602 6. http://www.akkadia.org/drepper/dsohowto.pdf
7603 7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility
7604 8. http://mentorembedded.github.com/cxx-abi/
7605 9. http://gcc.gnu.org/fortran/
7606 10. https://gcc.gnu.org/install/
7607 11. https://gcc.gnu.org/wiki/Visibility
7608 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1
7609 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2
7610 14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html
7611 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4
7612 16. https://gcc.gnu.org/onlinedocs/
7613 17. mailto:gcc-help@gcc.gnu.org
7614 18. mailto:gcc@gcc.gnu.org
7615 19. https://gcc.gnu.org/lists.html
7616 20. http://www.fsf.org/
7617 21. https://gcc.gnu.org/about.html
7618 22. http://validator.w3.org/check/referer
7619 ======================================================================
7620 http://gcc.gnu.org/gcc-3.4/index.html
7621 GCC 3.4 Release Series
7625 The [1]GNU project and the GCC developers are pleased to announce the
7626 release of GCC 3.4.6.
7628 This release is a bug-fix release, containing fixes for regressions in
7629 GCC 3.4.4 relative to previous releases of GCC. This is the last of the
7632 The GCC 3.4 release series includes numerous [2]new features,
7633 improvements, bug fixes, and other changes, thanks to an [3]amazing
7634 group of volunteers.
7639 March 6, 2006 ([4]changes)
7642 November 30, 2005 ([5]changes)
7645 May 18, 2005 ([6]changes)
7648 November 4, 2004 ([7]changes)
7651 September 6, 2004 ([8]changes)
7654 July 1, 2004 ([9]changes)
7657 April 18, 2004 ([10]changes)
7659 References and Acknowledgements
7661 GCC used to stand for the GNU C Compiler, but since the compiler
7662 supports several other languages aside from C, it now stands for the
7663 GNU Compiler Collection.
7665 A list of [11]successful builds is updated as new information becomes
7668 The GCC developers would like to thank the numerous people that have
7669 contributed new features, improvements, bug fixes, and other changes as
7670 well as test results to GCC. This [12]amazing group of volunteers is
7671 what makes GCC successful.
7673 For additional information about GCC please refer to the [13]GCC
7674 project web site or contact the [14]GCC development mailing list.
7676 To obtain GCC please use [15]our mirror sites, or [16]our SVN server.
7679 For questions related to the use of GCC, please consult these web
7680 pages and the [17]GCC manuals. If that fails, the
7681 [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
7682 web pages and the development of GCC are welcome on our developer
7683 list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
7686 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
7687 distribution of this entire article is permitted in any medium,
7688 provided this notice is preserved.
7690 These pages are [22]maintained by the GCC team. Last modified
7695 1. http://www.gnu.org/
7696 2. http://gcc.gnu.org/gcc-3.4/changes.html
7697 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
7698 4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
7699 5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5
7700 6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4
7701 7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3
7702 8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2
7703 9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1
7704 10. http://gcc.gnu.org/gcc-3.4/changes.html
7705 11. http://gcc.gnu.org/gcc-3.4/buildstat.html
7706 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
7707 13. http://gcc.gnu.org/index.html
7708 14. mailto:gcc@gcc.gnu.org
7709 15. http://gcc.gnu.org/mirrors.html
7710 16. http://gcc.gnu.org/svn.html
7711 17. https://gcc.gnu.org/onlinedocs/
7712 18. mailto:gcc-help@gcc.gnu.org
7713 19. mailto:gcc@gcc.gnu.org
7714 20. https://gcc.gnu.org/lists.html
7715 21. http://www.fsf.org/
7716 22. https://gcc.gnu.org/about.html
7717 23. http://validator.w3.org/check/referer
7718 ======================================================================
7719 http://gcc.gnu.org/gcc-3.4/changes.html
7720 GCC 3.4 Release Series
7721 Changes, New Features, and Fixes
7723 The final release in the 3.4 release series is [1]GCC 3.4.6. The series
7726 GCC 3.4 has [2]many improvements in the C++ frontend. Before reporting
7727 a bug, please make sure it's really GCC, and not your code, that is
7732 * GNU Make is now required to build GCC.
7733 * With -nostdinc the preprocessor used to ignore both standard
7734 include paths and include paths contained in environment variables.
7735 It was neither documented nor intended that environment variable
7736 paths be ignored, so this has been corrected.
7737 * GCC no longer accepts the options -fvolatile, -fvolatile-global and
7738 -fvolatile-static. It is unlikely that they worked correctly in any
7740 * GCC no longer ships <varargs.h>. Use <stdarg.h> instead.
7741 * Support for all the systems [3]obsoleted in GCC 3.3 has been
7742 removed from GCC 3.4. See below for a [4]list of systems which are
7743 obsoleted in this release.
7744 * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C
7745 compilers will not work.
7746 * The implementation of the [5]MIPS ABIs has changed. As a result,
7747 the code generated for certain MIPS targets will not be binary
7748 compatible with earlier releases.
7749 * In previous releases, the MIPS port had a fake "hilo" register with
7750 the user-visible name accum. This register has been removed.
7751 * The implementation of the [6]SPARC ABIs has changed. As a result,
7752 the code generated will not be binary compatible with earlier
7753 releases in certain cases.
7754 * The configure option --enable-threads=pthreads has been removed;
7755 use --enable-threads=posix instead, which should have the same
7757 * Code size estimates used by inlining heuristics for C, Objective-C,
7758 C++ and Java have been redesigned significantly. As a result the
7759 parameters of -finline-insns, --param max-inline-insns-single and
7760 --param max-inline-insns-auto need to be reconsidered.
7761 * --param max-inline-slope and --param min-inline-insns have been
7762 removed; they are not needed for the new bottom-up inlining
7764 * The new unit-at-a-time compilation scheme has several compatibility
7766 + The order in which functions, variables, and top-level asm
7767 statements are emitted may have changed. Code relying on some
7768 particular ordering needs to be updated. The majority of such
7769 top-level asm statements can be replaced by section
7771 + Unreferenced static variables and functions are removed. This
7772 may result in undefined references when an asm statement
7773 refers to the variable/function directly. In that case either
7774 the variable/function shall be listed in asm statement operand
7775 or in the case of top-level asm statements the attribute used
7776 shall be used to force function/variable to be always output
7777 and considered as a possibly used by unknown code.
7778 For variables the attribute is accepted only by GCC 3.4 and
7779 newer, while for earlier versions it is sufficient to use
7780 unused to silence warnings about the variables not being
7781 referenced. To keep code portable across different GCC
7782 versions, you can use appropriate preprocessor conditionals.
7783 + Static functions now can use non-standard passing conventions
7784 that may break asm statements calling functions directly.
7785 Again the attribute used shall be used to prevent this
7787 As a temporary workaround, -fno-unit-at-a-time can be used, but
7788 this scheme may not be supported by future releases of GCC.
7789 * GCC 3.4 automatically places zero-initialized variables in the .bss
7790 section on some operating systems. Versions of GNU Emacs up to (and
7791 including) 21.3 will not work correctly when using this
7792 optimization; you can use -fno-zero-initialized-in-bss to disable
7794 * If GCC 3.4 is configured with --enable-threads=posix (the default
7795 on most targets that support pthreads) then _REENTRANT will be
7796 defined unconditionally by some libstdc++ headers. C++ code which
7797 relies on that macro to detect whether multi-threaded code is being
7798 compiled might change in meaning, possibly resulting in linker
7799 errors for single-threaded programs. Affected users of [7]Boost
7800 should compile single-threaded code with -DBOOST_DISABLE_THREADS.
7801 See Bugzilla for [8]more information.
7803 General Optimizer Improvements
7805 * Usability of the profile feedback and coverage testing has been
7807 + Performance of profiled programs has been improved by faster
7808 profile merging code.
7809 + Better use of the profile feedback for optimization (loop
7810 unrolling and loop peeling).
7811 + File locking support allowing fork() calls and parallel runs
7812 of profiled programs.
7813 + Coverage file format has been redesigned.
7814 + gcov coverage tool has been improved.
7815 + make profiledbootstrap available to build a faster compiler.
7816 Experiments made on i386 hardware showed an 11% speedup on -O0
7817 and a 7.5% speedup on -O2 compilation of a [9]large C++
7819 + New value profiling pass enabled via -fprofile-values
7820 + New value profile transformations pass enabled via -fvpt aims
7821 to optimize some code sequences by exploiting knowledge about
7822 value ranges or other properties of the operands. At the
7823 moment a conversion of expensive divisions into cheaper
7824 operations has been implemented.
7825 + New -fprofile-generate and -fprofile-use command-line options
7826 to simplify the use of profile feedback.
7827 * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and
7828 Java which is enabled via -funit-at-a-time (and implied by -O2). In
7829 this scheme a whole file is parsed first and optimized later. The
7830 following basic inter-procedural optimizations are implemented:
7831 + Removal of unreachable functions and variables
7832 + Discovery of local functions (functions with static linkage
7833 whose address is never taken)
7834 + On i386, these local functions use register parameter passing
7836 + Reordering of functions in topological order of the call graph
7837 to enable better propagation of optimizing hints (such as the
7838 stack alignments needed by functions) in the back end.
7839 + Call graph based out-of-order inlining heuristics which allows
7840 to limit overall compilation unit growth (--param
7841 inline-unit-growth).
7842 Overall, the unit-at-a-time scheme produces a 1.3% improvement for
7843 the SPECint2000 benchmark on the i386 architecture (AMD Athlon
7845 * More realistic code size estimates used by inlining for C,
7846 Objective-C, C++ and Java. The growth of large functions can now be
7847 limited via --param large-function-insns and --param
7848 large-function-growth.
7849 * A new cfg-level loop optimizer pass replaces the old loop unrolling
7850 pass and adds two other loop transformations -- loop peeling and
7851 loop unswitching -- and also uses the profile feedback to limit
7852 code growth. (The three optimizations are enabled by
7853 -funroll-loops, -fpeel-loops and -funswitch-loops flags,
7855 The old loop unroller still can be enabled by -fold-unroll-loops
7856 and may produce better code in some cases, especially when the
7857 webizer optimization pass is not run.
7858 * A new web construction pass enabled via -fweb (and implied by -O3)
7859 improves the quality of register allocation, CSE, first scheduling
7860 pass and some other optimization passes by avoiding re-use of
7861 pseudo registers with non-overlapping live ranges. The pass almost
7862 always improves code quality but does make debugging difficult and
7863 thus is not enabled by default by -O2
7864 The pass is especially effective as cleanup after code duplication
7865 passes, such as the loop unroller or the tracer.
7866 * Experimental implementations of superblock or trace scheduling in
7867 the second scheduling pass can be enabled via
7868 -fsched2-use-superblocks and -fsched2-use-traces, respectively.
7870 New Languages and Language specific improvements
7874 * The Ada front end has been updated to include numerous bug fixes
7875 and enhancements. These include:
7876 + Improved project file support
7877 + Additional set of warnings about potential wrong code
7878 + Improved error messages
7879 + Improved code generation
7880 + Improved cross reference information
7882 + Better run-time check elimination
7883 + Better error recovery
7884 + More efficient implementation of unbounded strings
7885 + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools,
7887 + New GNAT.xxxx packages (e.g. GNAT.Strings,
7888 GNAT.Exception_Action)
7890 + New -gnatS switch replacing gnatpsta
7891 + Implementation of new Ada features (in particular limited
7892 with, limited aggregates)
7896 * Precompiled headers are now supported. Precompiled headers can
7897 dramatically speed up compilation of some projects. There are some
7898 known defects in the current precompiled header implementation that
7899 will result in compiler crashes in relatively rare situations.
7900 Therefore, precompiled headers should be considered a "technology
7901 preview" in this release. Read the manual for details about how to
7902 use precompiled headers.
7903 * File handling in the preprocessor has been rewritten. GCC no longer
7904 gets confused by symlinks and hardlinks, and now has a correct
7905 implementation of #import and #pragma once. These two directives
7906 have therefore been un-deprecated.
7907 * The undocumented extension that allowed C programs to have a label
7908 at the end of a compound statement, which has been deprecated since
7909 GCC 3.0, has been removed.
7910 * The cast-as-lvalue extension has been removed for C++ and
7911 deprecated for C and Objective-C. In particular, code like this:
7919 is no longer accepted for C++ and will not be accepted for C and
7920 Objective-C in a future version.
7921 * The conditional-expression-as-lvalue extension has been deprecated
7922 for C and Objective-C. In particular, code like this:
7926 will not be accepted for C and Objective-C in a future version.
7927 * The compound-expression-as-lvalue extension has been deprecated for
7928 C and Objective-C. In particular, code like this:
7932 will not be accepted for C and Objective-C in a future version. A
7933 possible non-intrusive workaround is the following:
7936 * Several [10]built-in functions such as __builtin_popcount for
7937 counting bits, finding the highest and lowest bit in a word, and
7938 parity have been added.
7939 * The -fwritable-strings option has been deprecated and will be
7941 * Many C math library functions are now recognized as built-ins and
7943 * The C, C++, and Objective-C compilers can now handle source files
7944 written in any character encoding supported by the host C library.
7945 The default input character set is taken from the current locale,
7946 and may be overridden with the -finput-charset command line option.
7947 In the future we will add support for inline encoding markers.
7951 * G++ is now much closer to full conformance to the ISO/ANSI C++
7952 standard. This means, among other things, that a lot of invalid
7953 constructs which used to be accepted in previous versions will now
7954 be rejected. It is very likely that existing C++ code will need to
7955 be fixed. This document lists some of the most common issues.
7956 * A hand-written recursive-descent C++ parser has replaced the
7957 YACC-derived C++ parser from previous GCC releases. The new parser
7958 contains much improved infrastructure needed for better parsing of
7959 C++ source codes, handling of extensions, and clean separation
7960 (where possible) between proper semantics analysis and parsing. The
7961 new parser fixes many bugs that were found in the old parser.
7962 * You must now use the typename and template keywords to disambiguate
7963 dependent names, as required by the C++ standard.
7965 typedef int mytype_t;
7968 template <class T1> struct A {
7969 template <class T2> struct B {
7973 template <int N> void bar(void)
7975 // Use 'typename' to tell the parser that T1::mytype_t names
7976 // a type. This is needed because the name is dependent (in
7977 // this case, on template parameter T1).
7978 typename T1::mytype_t x;
7983 template <class T> void template_func(void)
7985 // Use 'template' to prefix member templates within
7986 // dependent types (a has type A<T>, which depends on
7987 // the template parameter T).
7989 a.template bar<0>();
7991 // Use 'template' to tell the parser that B is a nested
7992 // template class (dependent on template parameter T), and
7993 // 'typename' because the whole A<T>::B<int> is
7994 // the name of a type (again, dependent).
7995 typename A<T>::template B<int> b;
7999 void non_template_func(void)
8001 // Outside of any template class or function, no names can be
8002 // dependent, so the use of the keyword 'typename' and 'template'
8003 // is not needed (and actually forbidden).
8009 * In a template definition, unqualified names will no longer find
8010 members of a dependent base (as specified by [temp.dep]/3 in the
8011 C++ standard). For example,
8012 template <typename T> struct B {
8020 template <typename T> struct C : B<T> {
8025 n = 0; // ::n is modified
8026 g (); // ::g is called
8029 You must make the names dependent, e.g. by prefixing them with
8030 this->. Here is the corrected definition of C<T>::h,
8031 template <typename T> void C<T>::h ()
8038 As an alternative solution (unfortunately not backwards compatible
8039 with GCC 3.3), you may use using declarations instead of this->:
8040 template <typename T> struct C : B<T> {
8053 * In templates, all non-dependent names are now looked up and bound
8054 at definition time (while parsing the code), instead of later when
8055 the template is instantiated. For instance:
8058 template <int> struct A {
8059 static void bar(void){
8068 A<0>::bar(); // Calls foo(int), used to call foo(char).
8071 * In an explicit instantiation of a class template, you must use
8072 class or struct before the template-id:
8076 template A<0>; // error, not accepted anymore
8077 template class A<0>; // OK
8078 * The "named return value" and "implicit typename" extensions have
8080 * Default arguments in function types have been deprecated and will
8082 * ARM-style name-injection of friend declarations has been deprecated
8083 and will be removed. For example: struct S { friend void f(); };
8084 void g() { f(); } will not be accepted by future versions of G++;
8085 instead a declaration of "f" will need to be present outside of the
8087 * Covariant returns are implemented for all but varadic functions
8088 that require an adjustment.
8089 * When -pedantic is used, G++ now issues errors about spurious
8090 semicolons. For example,
8091 namespace N {}; // Invalid semicolon.
8092 void f() {}; // Invalid semicolon.
8093 * G++ no longer accepts attributes for a declarator after the
8094 initializer associated with that declarator. For example,
8095 X x(1) __attribute__((...));
8096 is no longer accepted. Instead, use:
8097 X x __attribute__((...)) (1);
8098 * Inside the scope of a template class, the name of the class itself
8099 can be treated as either a class or a template. So GCC used to
8100 accept the class name as argument of type template, and template
8101 template parameter. However this is not C++ standard compliant. Now
8102 the name is not treated as a valid template template argument
8103 unless you qualify the name by its scope. For example, the code
8104 below no longer compiles.
8105 template <template <class> class TT> class X {};
8106 template <class T> class Y {
8107 X<Y> x; // Invalid, Y is always a type template parameter.
8109 The valid code for the above example is
8110 X< ::Y> x; // Valid.
8111 (Notice the space between < and : to prevent GCC to interpret this
8112 as a digraph for [.)
8113 * Friend declarations that refer to template specializations are
8114 rejected if the template has not already been declared. For
8116 template <typename T>
8118 friend void f<> (C&);
8120 is rejected. You must first declare f as a template,
8121 template <typename T>
8123 * In case of friend declarations, every name used in the friend
8124 declaration must be accessible at the point of that declaration.
8125 Previous versions of G++ used to be less strict about this and
8126 allowed friend declarations for private class members, for example.
8127 See the ISO C++ Standard Committee's [11]defect report #209 for
8129 * Declaration of member functions of class templates as friends are
8130 supported. For example,
8131 template <typename T> struct A {
8135 template <typename T> friend void A<T>::f();
8137 * You must use template <> to introduce template specializations, as
8138 required by the standard. For example,
8139 template <typename T>
8143 is rejected. You must write,
8144 template <> struct S<int> {};
8145 * G++ used to accept code like this,
8148 void f(int i = g());
8151 This behavior is not mandated by the standard. Now G++ issues an
8152 error about this code. To avoid the error, you must move the
8153 declaration of g before the declaration of f. The default arguments
8154 for g must be visible at the point where it is called.
8155 * The C++ ABI Section 3.3.3 specifications for the array construction
8156 routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return
8157 NULL when the allocator argument returns NULL. These changes are
8158 incorporated into the libstdc++ runtime library.
8159 * Using a name introduced by a typedef in a friend declaration or in
8160 an explicit instantiation is now rejected, as specified by the ISO
8165 friend class B; // error, no typedef name here
8166 friend B; // error, friend always needs class/struct/enum
8167 friend class A; // OK
8170 template <int> class Q {};
8172 template class R; // error, no typedef name here
8173 template class Q<0>; // OK
8174 * When allocating an array with a new expression, GCC used to allow
8175 parentheses around the type name. This is actually ill-formed and
8177 int* a = new (int)[10]; // error, not accepted anymore
8178 int* a = new int[10]; // OK
8179 * When binding an rvalue of class type to a reference, the copy
8180 constructor of the class must be accessible. For instance, consider
8188 A(const A&); // private copy ctor
8196 foo(A()); // error, copy ctor is not accessible
8197 foo(makeA()); // error, copy ctor is not accessible
8200 foo(a1); // OK, a1 is a lvalue
8202 This might be surprising at first sight, especially since most
8203 popular compilers do not correctly implement this rule ([12]further
8205 * When forming a pointer to member or a pointer to member function,
8206 access checks for class visibility (public, protected, private) are
8207 now performed using the qualifying scope of the name itself. This
8208 is better explained with an example:
8224 &A::pub_func; // OK, pub_func is accessible through A
8225 &A::prot_func; // error, cannot access prot_func through A
8226 &A::priv_func; // error, cannot access priv_func through A
8228 &B::pub_func; // OK, pub_func is accessible through B
8229 &B::prot_func; // OK, can access prot_func through B (within B)
8230 &B::priv_func; // error, cannot access priv_func through B
8234 Runtime Library (libstdc++)
8236 * Optimization work:
8237 + Streamlined streambuf, filebuf, separate synched with C
8238 Standard I/O streambuf.
8239 + All formatted I/O now uses cached locale information.
8240 + STL optimizations (memory/speed for list, red-black trees as
8241 used by sets and maps).
8242 + More use of GCC builtins.
8243 + String optimizations (avoid contention on
8244 increment/decrement-and-test of the reference count in the
8245 empty-string object, constructor from input_iterators
8247 * Static linkage size reductions.
8248 * Large File Support (files larger than 2 GB on 32-bit systems).
8249 * Wide character and variable encoding filebuf work (UTF-8, Unicode).
8250 * Generic character traits.
8251 * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x,
8252 Solaris 2.7 and above, AIX 5.x, Irix 6.5.
8253 * The allocator class is now standard-conformant, and two additional
8254 extension allocators have been added, mt_alloc and
8256 * PCH support: -include bits/stdc++.h (2x compile speedup).
8257 * Rewrote __cxa_demangle with support for C++ style allocators.
8258 * New debug modes for STL containers and iterators.
8259 * Testsuite rewrite: five times as many tests, plus increasingly
8260 sophisticated tests, including I/O, MT, multi-locale, wide and
8262 * Use current versions of GNU "autotools" for build/configuration.
8266 * The Objective-C front end has been updated to include the numerous
8267 bug fixes and enhancements previously available only in Apple's
8268 version of GCC. These include:
8269 + Structured exception (@try... @catch... @finally, @throw) and
8270 synchronization (@synchronized) support. These are accessible
8271 via the -fobjc-exceptions switch; as of this writing, they may
8272 only be used in conjunction with -fnext-runtime on Mac OS X
8273 10.3 and later. See [13]Options Controlling Objective-C
8274 Dialect for more information.
8275 + An overhaul of @encode logic. The C99 _Bool and C++ bool type
8276 may now be encoded as 'B'. In addition, the back-end/codegen
8277 dependencies have been removed.
8278 + An overhaul of message dispatch construction, ensuring that
8279 the various receiver types (and casts thereof) are handled
8280 properly, and that correct diagnostics are issued.
8281 + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue"
8282 (-freplace-objc-classes) debugging modes, currently available
8283 on Mac OS X 10.3 and later. See [14]Options Controlling
8284 Objective-C Dialect for more information.
8285 + Access to optimized runtime entry points (-fno-nil-receivers )
8286 on the assumption that message receivers are never nil. This
8287 is currently available on Mac OS X 10.3 and later. See
8288 [15]Options Controlling Objective-C Dialect for more
8293 * Compiling a .jar file will now cause non-.class entries to be
8294 automatically compiled as resources.
8295 * libgcj has been ported to Darwin.
8296 * Jeff Sturm has adapted Jan Hubicka's call graph optimization code
8298 * libgcj has a new gcjlib URL type; this lets URLClassLoader load
8299 code from shared libraries.
8300 * libgcj has been much more completely merged with [16]GNU Classpath.
8301 * Class loading is now much more correct; in particular the caller's
8302 class loader is now used when that is required.
8303 * [17]Eclipse 2.x will run out of the box using gij.
8304 * Parts of java.nio have been implemented. Direct and indirect
8305 buffers work, as do fundamental file and socket operations.
8306 * java.awt has been improved, though it is still not ready for
8308 * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST
8310 * The MinGW port has matured. Enhancements include socket timeout
8311 support, thread interruption, improved Runtime.exec() handling and
8312 support for accented characters in filenames.
8316 * Fortran improvements are listed in the [18]Fortran documentation.
8318 New Targets and Target Specific Improvements
8322 * Several [19]built-in functions have been added such as
8323 __builtin_alpha_zap to allow utilizing the more obscure
8324 instructions of the CPU.
8325 * Parameter passing of complex arguments has changed to match the
8326 [20]ABI. This change is incompatible with previous GCC versions,
8327 but does fix compatibility with the Tru64 compiler and several
8328 corner cases where GCC was incompatible with itself.
8332 * Nicolas Pitre has contributed his hand-coded floating-point support
8333 code for ARM. It is both significantly smaller and faster than the
8334 existing C-based implementation, even when building applications
8335 for Thumb. The arm-elf configuration has been converted to use the
8337 * Support for the Intel's iWMMXt architecture, a second generation
8338 XScale processor, has been added. Enabled at run time with the
8339 -mcpu=iwmmxt command line switch.
8340 * A new ARM target has been added: arm-wince-pe. This is similar to
8341 the arm-pe target, but it defaults to using the APCS32 ABI.
8342 * The existing ARM pipeline description has been converted to the use
8343 the [21]DFA processor pipeline model. There is not much change in
8344 code performance, but the description is now [22]easier to
8346 * Support for the Cirrus EP9312 Maverick floating point co-processor
8347 added. Enabled at run time with the -mcpu=ep9312 command line
8348 switch. Note however that the multilibs to support this chip are
8349 currently disabled in gcc/config/arm/t-arm-elf, so if you want to
8350 enable their production you will have to uncomment the entries in
8355 * Support for long long has been added.
8356 * Support for saveall attribute has been added.
8357 * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code
8358 for H8/300H and H8S, which is much faster than the previous
8360 * A lot of small performance improvements.
8362 IA-32/AMD64 (x86-64)
8364 * Tuning for K8 (AMD Opteron/Athlon64) core is available via
8365 -march=k8 and -mcpu=k8.
8366 * Scalar SSE code generation carefully avoids reformatting penalties,
8367 hidden dependencies and minimizes the number of uops generated on
8368 both Intel and AMD CPUs.
8369 * Vector MMX and SSE operands are now passed in registers to improve
8370 performance and match the argument passing convention used by the
8371 Intel C++ Compiler. As a result it is not possible to call
8372 functions accepting vector arguments compiled by older GCC version.
8373 * Conditional jump elimination is now more aggressive on modern CPUs.
8374 * The Athlon ports has been converted to use the DFA processor
8375 pipeline description.
8376 * Optimization of indirect tail calls is now possible in a similar
8377 fashion as direct sibcall optimization.
8378 * Further small performance improvements.
8379 * -m128bit-long-double is now less buggy.
8380 * __float128 support in 64-bit compilation.
8381 * Support for data structures exceeding 2GB in 64-bit mode.
8382 * -mcpu has been renamed to -mtune.
8386 * Tuning code for the Itanium 2 processor has been added. The
8387 generation of code tuned for Itanium 2 (option -mtune=itanium2) is
8388 enabled by default now. To generate code tuned for Itanium 1 the
8389 option -mtune=itanium1 should be used.
8390 * [23]DFA processor pipeline descriptions for the IA-64 processors
8391 have been added. This resulted in about 3% improvement on the
8392 SPECInt2000 benchmark for Itanium 2.
8393 * Instruction bundling for the IA-64 processors has been rewritten
8394 using the DFA pipeline hazard recognizer. It resulted in about 60%
8395 compiler speedup on the SPECInt2000 C programs.
8399 * Support for the M32R/2 processor has been added by Renesas.
8400 * Support for an M32R GNU/Linux target and PIC code generation has
8401 been added by Renesas.
8405 * Bernardo Innocenti (Develer S.r.l.) has contributed the
8406 m68k-uclinux target, based on former work done by Paul Dale
8407 (SnapGear Inc.). Code generation for the ColdFire processors family
8408 has been enhanced and extended to support the MCF 53xx and MCF 54xx
8409 cores, integrating former work done by Peter Barada (Motorola).
8413 Processor-specific changes
8415 * Support for the RM7000 and RM9000 processors has been added. It can
8416 be selected using the -march compiler option and should work with
8417 any MIPS I (mips-*) or MIPS III (mips64-*) configuration.
8418 * Support for revision 2 of the MIPS32 ISA has been added. It can be
8419 selected with the command-line option -march=mips32r2.
8420 * There is a new option, -mfix-sb1, to work around certain SB-1
8425 * It is possible to customize GCC using the following configure-time
8427 + --with-arch, which specifies the default value of the -march
8429 + --with-tune, which specifies the default value of the -mtune
8431 + --with-abi, which specifies the default ABI.
8432 + --with-float=soft, which tells GCC to use software floating
8434 + --with-float=hard, which tells GCC to use hardware floating
8436 * A 64-bit GNU/Linux port has been added. The associated
8437 configurations are mips64-linux-gnu and mips64el-linux-gnu.
8438 * The 32-bit GNU/Linux port now supports Java.
8439 * The IRIX 6 configuration now supports the o32 ABI and will build
8440 o32 multilibs by default. This support is compatible with both
8441 binutils and the SGI tools, but note that several features,
8442 including debugging information and DWARF2 exception handling, are
8443 only available when using the GNU assembler. Use of the GNU
8444 assembler and linker (version 2.15 or above) is strongly
8446 * The IRIX 6 configuration now supports 128-bit long doubles.
8447 * There are two new RTEMS-specific configurations, mips-rtems and
8449 * There are two new *-elf configurations, mipsisa32r2-elf and
8454 * Several [24]ABI bugs have been fixed. Unfortunately, these changes
8455 will break binary compatibility with earlier releases.
8456 * GCC can now use explicit relocation operators when generating
8457 -mabicalls code. This behavior is controlled by -mexplicit-relocs
8458 and can have several performance benefits. For example:
8459 + It allows for more optimization of GOT accesses, including
8460 better scheduling and redundancy elimination.
8461 + It allows sibling calls to be implemented as jumps.
8462 + n32 and n64 leaf functions can use a call-clobbered global
8463 pointer instead of $28.
8464 + The code to set up $gp can be removed from functions that
8466 * A new option, -mxgot, allows the GOT to be bigger than 64k. This
8467 option is equivalent to the assembler's -xgot option and should be
8468 used instead of -Wa,-xgot.
8469 * Frame pointer elimination is now supported when generating 64-bit
8471 * Inline block moves have been optimized to take more account of
8472 alignment information.
8473 * Many internal changes have been made to the MIPS port, mostly aimed
8474 at reducing the reliance on assembler macros.
8478 * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64
8479 [25]ABI incompatibilities regarding the way parameters are passed
8480 during functions calls. These changes may result in incompatibility
8481 between code compiled with GCC 3.3 and GCC 3.4.
8485 * Support for shared/dylib gcc libraries has been added. It is
8486 enabled by default on powerpc-apple-darwin7.0.0 and up.
8487 * Libgcj is enabled by default. On systems older than
8488 powerpc-apple-darwin7.0.0 you need to install dlcompat.
8489 * 128-bit IBM extended precision format support added for long
8494 * By default, PowerPC64 GNU/Linux now uses natural alignment of
8495 structure elements. The old four byte alignment for double, with
8496 special rules for a struct starting with a double, can be chosen
8497 with -malign-power. This change may result in incompatibility
8498 between code compiled with GCC 3.3 and GCC 3.4.
8499 * -mabi=altivec is now the default rather than -mabi=no-altivec.
8500 * 128-bit IBM extended precision format support added for long
8505 * New command-line options allow to specify the intended execution
8506 environment for generated code:
8507 + -mesa/-mzarch allows to specify whether to generate code
8508 running in ESA/390 mode or in z/Architecture mode (this is
8509 applicable to 31-bit code only).
8510 + -march allows to specify a minimum processor architecture
8511 level (g5, g6, z900, or z990).
8512 + -mtune allows to specify which processor to tune for.
8513 * It is possible to customize GCC using the following configure-time
8515 + --with-mode, which specifies whether to default to assuming
8516 ESA/390 or z/Architecture mode.
8517 + --with-arch, which specifies the default value of the -march
8519 + --with-tune, which specifies the default value of the -mtune
8521 * Support for the z990 processor has been added, and can be selected
8522 using -march=z990 or -mtune=z990. This includes instruction
8523 scheduling tuned for the superscalar instruction pipeline of the
8524 z990 processor as well as support for all new instructions provided
8525 by the long-displacement facility.
8526 * Support to generate 31-bit code optimized for zSeries processors
8527 (running in ESA/390 or in z/Architecture mode) has been added. This
8528 can be selected using -march=z900 and -mzarch respectively.
8529 * Instruction scheduling for the z900 and z990 processors now uses
8530 the DFA pipeline hazard recognizer.
8531 * GCC no longer generates code to maintain a stack backchain,
8532 previously used to generate stack backtraces for debugging
8533 purposes. As replacement that does not incur runtime overhead,
8534 DWARF-2 call frame information is provided by GCC; this is
8535 supported by GDB 6.1. The old behavior can be restored using the
8537 * The stack frame size of functions may now exceed 2 GB in 64-bit
8539 * A port for the 64-bit IBM TPF operating system has been added; the
8540 configuration is s390x-ibm-tpf. This configuration is supported as
8541 cross-compilation target only.
8542 * Various changes to improve the generated code have been
8543 implemented, including:
8544 + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT
8545 instructions to significantly speed up many floating-point
8547 + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL
8548 WITH BORROW instructions to speed up long long arithmetic.
8549 + GCC now uses the SEARCH STRING instruction to implement
8551 + In many cases, function call overhead for 31-bit code has been
8552 reduced by placing the literal pool after the function code
8553 instead of after the function prolog.
8554 + Register 14 is no longer reserved in 64-bit code.
8555 + Handling of global register variables has been improved.
8559 * The option -mflat is deprecated.
8560 * Support for large (> 2GB) frames has been added to the 64-bit port.
8561 * Several [26]ABI bugs have been fixed. Unfortunately, these changes
8562 will break binary compatibility with earlier releases.
8563 * The default debugging format has been switched from STABS to
8564 DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already
8565 the default debugging format for 64-bit code on Solaris.
8569 * Support for the SH2E processor has been added. Enabled at run time
8570 with the -m2e command line switch, or at configure time by
8571 specifying sh2e as the machine part of the target triple.
8575 * Support for the Mitsubishi V850E1 processor has been added. This is
8576 a variant of the V850E processor with some additional debugging
8581 * Several ABI bugs have been fixed. Unfortunately, these changes
8582 break binary compatibility with earlier releases.
8583 + For big-endian processors, the padding of aggregate return
8584 values larger than a word has changed. If the size of an
8585 aggregate return value is not a multiple of 32 bits, previous
8586 versions of GCC inserted padding in the most-significant bytes
8587 of the first return value register. Aggregates larger than a
8588 word are now padded in the least-significant bytes of the last
8589 return value register used. Aggregates smaller than a word are
8590 still padded in the most-significant bytes. The return value
8591 padding has not changed for little-endian processors.
8592 + Function arguments with 16-byte alignment are now properly
8594 + The implementation of the va_list type has changed. A va_list
8595 value created by va_start from a previous release cannot be
8596 used with va_arg from this release, or vice versa.
8597 * More processor configuration options for Xtensa processors are
8599 + the ABS instruction is now optional;
8600 + the ADDX* and SUBX* instructions are now optional;
8601 + an experimental CONST16 instruction can be used to synthesize
8602 constants instead of loading them from constant pools.
8603 These and other Xtensa processor configuration options can no
8604 longer be enabled or disabled by command-line options; the
8605 processor configuration must be specified by the xtensa-config.h
8606 header file when building GCC. Additionally, the
8607 -mno-serialize-volatile option is no longer supported.
8611 Support for a number of older systems has been declared obsolete in GCC
8612 3.4. Unless there is activity to revive them, the next release of GCC
8613 will have their sources permanently removed.
8615 All configurations of the following processor architectures have been
8617 * Mitsubishi D30V, d30v-*
8618 * AT&T DSP1600 and DSP1610, dsp16xx-*
8621 Also, some individual systems have been obsoleted:
8623 + Support for generating code for operation in APCS/26 mode
8626 + "Bigfoot" port, i370-*. (The other port, s390-*, is actively
8627 maintained and supported.)
8629 + MOSS, i?86-moss-msdos and i?86-*-moss*
8630 + NCR 3000 running System V r.4, i?86-ncr-sysv4*
8631 + FreeBSD with a.out object format, i?86-*-freebsd*aout* and
8633 + GNU/Linux with a.out object format, i?86-linux*aout*
8634 + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1*
8635 + Interix versions before Interix 3, i?86-*-interix
8636 + Mach microkernel, i?86-mach*
8637 + SCO UnixWare with UDK, i?86-*-udk*
8638 + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]*
8639 + VSTa microkernel, i386-*-vsta
8640 * Motorola M68000 family
8641 + HPUX, m68k-hp-hpux* and m68000-hp-hpux*
8642 + NetBSD with a.out object format (before NetBSD 1.4),
8643 m68k-*-*-netbsd* except m68k-*-*-netbsdelf*
8644 + Generic System V r.4, m68k-*-sysv4*
8646 + Generic VAX, vax-*-* (This is generic VAX only; we have not
8647 obsoleted any VAX triples for specific operating systems.)
8649 Documentation improvements
8651 Other significant improvements
8653 * The build system has undergone several significant cleanups.
8654 Subdirectories will only be configured if they are being built, and
8655 all subdirectory configures are run from the make command. The top
8656 level has been autoconfiscated.
8657 * Building GCC no longer writes to its source directory. This should
8658 help those wishing to share a read-only source directory over NFS
8659 or build from a CD. The exceptions to this feature are if you
8660 configure with either --enable-maintainer-mode or
8661 --enable-generated-files-in-srcdir.
8662 * The -W warning option has been renamed to -Wextra, which is more
8663 easily understood. The older spelling will be retained for
8664 backwards compatibility.
8665 * Substantial improvements in compile time have been made,
8666 particularly for non-optimizing compilations.
8667 __________________________________________________________________
8673 A vast number of bugs have been fixed in 3.4.0, too many to publish a
8674 complete list here. [27]Follow this link to query the Bugzilla database
8675 for the list of over 900 bugs fixed in 3.4.0. This is the list of all
8676 bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4
8678 __________________________________________________________________
8684 This section lists the problem reports (PRs) from GCC's bug tracking
8685 system that are known to be fixed in the 3.4.1 release. This list might
8686 not be complete (that is, it is possible that some PRs that have been
8687 fixed are not listed here).
8691 * [28]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler
8692 emitted - PIC related
8693 * [29]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf
8694 * [30]14760 A bug in configure.in prevents using both
8695 --program-suffix and --program-prefix
8696 * [31]14671 [hppa64] bootstrap fails: ICE in
8697 save_call_clobbered_regs, in caller_save.c
8698 * [32]15093 [alpha][Java] make bootstrap fails to configure libffi on
8700 * [33]15178 Solaris 9/x86 fails linking after stage 3
8702 Multi-platform internal compiler errors (ICEs)
8704 * [34]12753 (preprocessor) Memory corruption in preprocessor on bad
8706 * [35]13985 ICE in gcc.c-torture/compile/930621-1.c
8707 * [36]14810 (c++) tree check failures with invalid code involving
8709 * [37]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in
8711 * [38]15044 (c++) ICE on syntax error, template header
8712 * [39]15057 (c++) Compiling of conditional value throw constructs
8713 cause a segmentation violation
8714 * [40]15064 (c++) typeid of template parameter gives ICE
8715 * [41]15142 (c++) ICE when passing a string where a char* is expected
8716 in a throw statement
8717 * [42]15159 ICE in rtl_verify_flow_info_1
8718 * [43]15165 (c++) ICE in instantiate_template
8719 * [44]15193 Unary minus using pointer to V4SF vector causes
8720 -fforce-mem to exhaust all memory
8721 * [45]15209 (c++) Runs out of memory with packed structs
8722 * [46]15227 (c++) Trouble with invalid function definition
8723 * [47]15285 (c++) instantiate_type ICE when forming pointer to
8725 * [48]15299 (c++) ICE in resolve_overloaded_unification
8726 * [49]15329 (c++) ICE on constructor of member template
8727 * [50]15550 ICE in extract_insn, in recog.c
8728 * [51]15554 (c++) ICE in tsubst_copy, in cp/pt.c
8729 * [52]15640 (c++) ICE on invalid code in arg_assoc, in
8731 * [53]15666 [unit-at-a-time] Gcc abort on valid code
8732 * [54]15696 (c++) ICE with bad pointer-to-member code
8733 * [55]15701 (c++) ICE with friends and template template parameter
8734 * [56]15761 ICE in do_SUBST, in combine.c
8735 * [57]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops
8739 * [58]14538 All RTEMS targets broken for gnat
8743 * [59]12391 missing warning about assigning to an incomplete type
8744 * [60]14649 atan(1.0) should not be a constant expression
8745 * [61]15004 [unit-at-a-time] no warning for unused paramater in
8747 * [62]15749 --pedantic-errors behaves differently from --pedantic
8748 with C-compiler on GNU/Linux
8750 C++ compiler and library
8752 * [63]10646 non-const reference is incorrectly matched in a "const T"
8753 partial specialization
8754 * [64]12077 wcin.rdbuf()->in_avail() return value too high
8755 * [65]13598 enc_filebuf doesn't work
8756 * [66]14211 const_cast returns lvalue but should be rvalue
8757 * [67]14220 num_put::do_put() undesired float/double behavior
8758 * [68]14245 problem with user-defined allocators in std::basic_string
8759 * [69]14340 libstdc++ Debug mode: failure to convert iterator to
8761 * [70]14600 __gnu_cxx::stdio_sync_filebuf should expose internal
8763 * [71]14668 no warning anymore for reevaluation of declaration
8764 * [72]14775 LFS (large file support) tests missing
8765 * [73]14821 Duplicate namespace alias declaration should not conflict
8766 * [74]14930 Friend declaration ignored
8767 * [75]14932 cannot use offsetof to get offsets of array elements in
8769 * [76]14950 [non unit-at-a-time] always_inline does not mix with
8771 * [77]14962 g++ ignores #pragma redefine_extname
8772 * [78]14975 Segfault on low-level write error during imbue
8773 * [79]15002 Linewise stream input is unusably slow (std::string slow)
8774 * [80]15025 compiler accepts redeclaration of template as
8776 * [81]15046 [arm] Math functions misdetected by cross configuration
8777 * [82]15069 a bit test on a variable of enum type is miscompiled
8778 * [83]15074 g++ -lsupc++ still links against libstdc++
8779 * [84]15083 spurious "statement has no effect" warning
8780 * [85]15096 parse error with templates and pointer to const member
8781 * [86]15287 combination of operator[] and operator .* fails in
8783 * [87]15317 __attribute__ unused in first parameter of constructor
8785 * [88]15337 sizeof on incomplete type diagnostic
8786 * [89]15361 bitset<>::_Find_next fails
8787 * [90]15412 _GLIBCXX_ symbols symbols defined and used in different
8789 * [91]15427 valid code results in incomplete type error
8790 * [92]15471 Incorrect member pointer offsets in anonymous
8792 * [93]15503 nested template problem
8793 * [94]15507 compiler hangs while laying out union
8794 * [95]15542 operator & and template definitions
8795 * [96]15565 SLES9: leading + sign for unsigned int with showpos
8796 * [97]15625 friend defined inside a template fails to find static
8798 * [98]15629 Function templates, overloads, and friend name injection
8799 * [99]15742 'noreturn' attribute ignored in method of template
8801 * [100]15775 Allocator::pointer consistently ignored
8802 * [101]15821 Duplicate namespace alias within namespace rejected
8803 * [102]15862 'enum yn' fails (confict with undeclared builtin)
8804 * [103]15875 rejects pointer to member in template
8805 * [104]15877 valid code using templates and anonymous enums is
8807 * [105]15947 Puzzling error message for wrong destructor declaration
8809 * [106]16020 cannot copy __gnu_debug::bitset
8810 * [107]16154 input iterator concept too restrictive
8811 * [108]16174 deducing top-level consts
8815 * [109]14315 Java compiler is not parallel make safe
8819 * [110]15151 [g77] incorrect logical i/o in 64-bit mode
8823 * [111]7993 private variables cannot be shadowed in subclasses
8827 * [112]15228 useless copies of floating point operands
8828 * [113]15345 [non-unit-at-a-time] unreferenced nested inline
8829 functions not optimized away
8830 * [114]15945 Incorrect floating point optimization
8831 * [115]15526 ftrapv aborts on 0 * (-1)
8832 * [116]14690 Miscompiled POOMA tests
8833 * [117]15112 GCC generates code to write to unchanging memory
8837 * [118]15067 Minor glitch in the source of cpp
8839 Main driver program bugs
8841 * [119]1963 collect2 interprets -oldstyle_liblookup as -o
8844 x86-specific (Intel/AMD)
8846 * [120]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND*
8851 * [121]14782 GCC produces an unaligned data access at -O2
8852 * [122]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2
8853 * [123]15202 ICE in reload_cse_simplify_operands, in postreload.c
8857 * [124]14610 __float80 constants incorrectly emitted
8858 * [125]14813 init_array sections are initialized in the wrong order
8859 * [126]14857 GCC segfault on duplicated asm statement
8860 * [127]15598 Gcc 3.4 ICE on valid code
8861 * [128]15653 Gcc 3.4 ICE on valid code
8865 * [129]15189 wrong filling of delay slot with -march=mips1 -G0
8866 -mno-split-addresses -mno-explicit-relocs
8867 * [130]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as
8869 * [131]16144 Bogus reference to __divdf3 when -O1
8870 * [132]16176 Miscompilation of unaligned data in MIPS backend
8874 * [133]11591 ICE in gcc.dg/altivec-5.c
8875 * [134]12028 powerpc-eabispe produces bad sCOND operation
8876 * [135]14478 rs6000 geu/ltu patterns generate incorrect code
8877 * [136]14567 long double and va_arg complex args
8878 * [137]14715 Altivec stack layout may overlap gpr save with stack
8880 * [138]14902 (libstdc++) Stream checking functions fail when -pthread
8882 * [139]14924 Compiler ICE on valid code
8883 * [140]14960 -maltivec affects vector return with -mabi=no-altivec
8884 * [141]15106 vector varargs failure passing from altivec to
8885 non-altivec code for -m32
8886 * [142]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 &
8888 * [143]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx
8890 * [144]15662 Segmentation fault when an exception is thrown - even if
8891 try and catch are specified
8895 * [145]15054 Bad code due to overlapping stack temporaries
8899 * [146]15783 ICE with union assignment in 64-bit mode
8900 * [147]15626 GCC 3.4 emits "ld: warning: relocation error:
8905 * [148]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64
8906 * [149]14723 Backported -march=nocona from mainline
8907 * [150]15290 __float128 failed to pass to function properly
8909 Cygwin/Mingw32-specific
8911 * [151]15250 Option -mms-bitfields support on GCC 3.4 is not
8912 conformant to MS layout
8913 * [152]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe
8914 worker on windows32 targets
8916 Bugs specific to embedded processors
8918 * [153]8309 [m68k] -m5200 produces erroneous SImode set of short
8920 * [154]13250 [SH] Gcc code for rotation clobbers the register, but
8921 gcc continues to use the register as if it was not clobbered
8922 * [155]13803 [coldfire] movqi operand constraints too restrictivefor
8924 * [156]14093 [SH] ICE for code when using -mhitachi option in SH
8925 * [157]14457 [m6811hc] ICE with simple c++ source
8926 * [158]14542 [m6811hc] ICE on simple source
8927 * [159]15100 [SH] cc1plus got hang-up on
8928 libstdc++-v3/testsuite/abi_check.cc
8929 * [160]15296 [CRIS] Delayed branch scheduling causing invalid code on
8931 * [161]15396 [SH] ICE with -O2 -fPIC
8932 * [162]15782 [coldfire] m68k_output_mi_thunk emits wrong code for
8935 Testsuite problems (compiler not affected)
8937 * [163]11610 libstdc++ testcases 27_io/* don't work properly remotely
8938 * [164]15488 (libstdc++) possibly insufficient file permissions for
8939 executing test suite
8940 * [165]15489 (libstdc++) testsuite_files determined incorrectly
8944 * [166]13928 (libstdc++) no whatis info in some man pages generated
8946 * [167]14150 Ada documentation out of date
8947 * [168]14949 (c++) Need to document method visibility changes
8948 * [169]15123 libstdc++-doc: Allocators.3 manpage is empty
8949 __________________________________________________________________
8955 This section lists the problem reports (PRs) from GCC's bug tracking
8956 system that are known to be fixed in the 3.4.2 release. This list might
8957 not be complete (that is, it is possible that some PRs that have been
8958 fixed are not listed here).
8960 Bootstrap failures and issues
8962 * [170]16469 [mips-sgi-irix5.3] bootstrap fails in
8963 libstdc++-v3/testsuite
8964 * [171]16344 [hppa-linux-gnu] libstdc++'s PCH built by
8965 profiledbootstrap does not work with the built compiler
8966 * [172]16842 [Solaris/x86] mkheaders can not find mkheaders.conf
8968 Multi-platform internal compiler errors (ICEs)
8970 * [173]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in
8971 cp_parser_class_specifier, in cp/parser.c
8972 * [174]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
8973 * [175]15461 (c++) ICE due to NRV and inlining
8974 * [176]15890 (c++) ICE in c_expand_expr, in c-common.c
8975 * [177]16180 ICE: segmentation fault in RTL optimization
8976 * [178]16224 (c++) ICE in write_unscoped_name (template/namespace)
8977 * [179]16408 ICE: in delete_insn, in cfgrtl.c
8978 * [180]16529 (c++) ICE for: namespace-alias shall not be declared as
8979 the name of any other entity
8980 * [181]16698 (c++) ICE with exceptions and declaration of __cxa_throw
8981 * [182]16706 (c++) ICE in finish_member_declaration, in
8983 * [183]16810 (c++) Legal C++ program with cast gives ICE in
8985 * [184]16851 (c++) ICE when throwing a comma expression
8986 * [185]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c
8987 * [186]16904 (c++) ICE in finish_class_member_access_expr, in
8989 * [187]16905 (c++) ICE (segfault) with exceptions
8990 * [188]16964 (c++) ICE in cp_parser_class_specifier due to
8992 * [189]17068 (c++) ICE: tree check: expected class 'd', have 'x'
8993 (identifier_node) in dependent_template_p, in cp/pt.c
8997 * [190]16366 Preprocessor option -remap causes memory corruption
9001 * [191]15345 unreferenced nested inline functions not optimized away
9002 * [192]16590 Incorrect execution when compiling with -O2
9003 * [193]16693 Bitwise AND is lost when used within a cast to an enum
9004 of the same precision
9005 * [194]17078 Jump into if(0) substatement fails
9007 Problems in generated debug information
9009 * [195]13956 incorrect stabs for nested local variables
9013 * [196]16684 GCC should not warn about redundant redeclarations of
9016 C++ compiler and library
9018 * [197]12658 Thread safety problems in locale::global() and
9020 * [198]13092 g++ accepts invalid pointer-to-member conversion
9021 * [199]15320 Excessive memory consumption
9022 * [200]16246 Incorrect template argument deduction
9023 * [201]16273 Memory exhausted when using nested classes and virtual
9025 * [202]16401 ostringstream in gcc 3.4.x very slow for big data
9026 * [203]16411 undefined reference to
9027 __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char>
9029 * [204]16489 G++ incorrectly rejects use of a null constant integral
9030 expression as a null constant pointer
9031 * [205]16618 offsetof fails with constant member
9032 * [206]16637 syntax error reported for valid input code
9033 * [207]16717 __attribute__((constructor)) broken in C++
9034 * [208]16813 compiler error in DEBUG version of range insertion
9036 * [209]16853 pointer-to-member initialization from incompatible one
9038 * [210]16889 ambiguity is not detected
9039 * [211]16959 Segmentation fault in ios_base::sync_with_stdio
9041 Java compiler and library
9043 * [212]7587 direct threaded interpreter not thread-safe
9044 * [213]16473 ServerSocket accept() leaks file descriptors
9045 * [214]16478 Hash synchronization deadlock with finalizers
9049 * [215]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c
9050 * [216]16974 could not split insn (ice in final_scan_insn, in
9055 * [217]16298 ICE in output_operand
9056 * [218]17113 ICE with SSE2 intrinsics
9060 * [219]14697 libstdc++ couldn't find 32bit libgcc_s
9064 * [220]15869 [mips64] No NOP after LW (with -mips1 -O0)
9065 * [221]16325 [mips64] value profiling clobbers gp on mips
9066 * [222]16357 [mipsisa64-elf] ICE copying 7 bytes between extern
9068 * [223]16380 [mips64] Use of uninitialised register after dbra
9070 * [224]16407 [mips64] Unaligned access to local variables
9071 * [225]16643 [mips64] verify_local_live_at_start ICE after
9072 crossjumping & cfgcleanup
9076 * [226]15927 THUMB -O2: strength-reduced iteration variable ends up
9078 * [227]15948 THUMB: ICE with non-commutative cbranch
9079 * [228]17019 THUMB: bad switch statement in md code for
9080 addsi3_cbranch_scratch
9084 * [229]16130 ICE on valid code: in bundling, in config/ia64/ia64.c
9086 * [230]16142 ICE on valid code: in bundling, in config/ia64/ia64.c
9088 * [231]16278 Gcc failed to build Linux kernel with -mtune=merced
9089 * [232]16414 ICE on valid code: typo in comparison of asm_noperands
9091 * [233]16445 ICE on valid code: don't count ignored insns
9092 * [234]16490 ICE (segfault) while compiling with -fprofile-use
9093 * [235]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS
9097 * [236]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x
9098 * [237]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions
9103 * [238]16199 ICE while compiling apache 2.0.49
9104 * [239]16416 -m64 doesn't imply -mcpu=v9 anymore
9105 * [240]16430 ICE when returning non-C aggregates larger than 16 bytes
9107 Bugs specific to embedded processors
9109 * [241]16379 [m32r] can't output large model function call of memcpy
9110 * [242]17093 [m32r] ICE with -msdata=use -O0
9111 * [243]17119 [m32r] ICE at switch case 0x8000
9115 * [244]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp
9117 Alpha Tru64-specific
9119 * [245]16210 libstdc++ gratuitously omits "long long" I/O
9121 Testsuite, documentation issues (compiler is not affected):
9123 * [246]15488 (libstdc++) possibly insufficient file permissions for
9124 executing test suite
9125 * [247]16250 ada/doctools runs makeinfo even in release tarball
9126 __________________________________________________________________
9130 This is the [248]list of problem reports (PRs) from GCC's bug tracking
9131 system that are known to be fixed in the 3.4.3 release. This list might
9132 not be complete (that is, it is possible that some PRs that have been
9133 fixed are not listed here).
9137 * [249]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1
9138 * [250]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold
9141 Internal compiler errors (ICEs) affecting multiple platforms
9143 * [251]13948 (java) GCJ segmentation fault while compiling GL4Java
9145 * [252]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
9146 * [253]16301 (c++) ICE when "strong" attribute is attached to a using
9148 * [254]16566 ICE with flexible arrays
9149 * [255]17023 ICE with nested functions in parameter declaration
9150 * [256]17027 ICE with noreturn function in loop at -O2
9151 * [257]17524 ICE in grokdeclarator, in cp/decl.c
9152 * [258]17826 (c++) ICE in cp_tree_equal
9154 C and optimization bugs
9156 * [259]15526 -ftrapv aborts on 0 * (-1)
9157 * [260]16999 #ident stopped working
9158 * [261]17503 quadratic behaviour in invalid_mode_change_p
9159 * [262]17581 Long long arithmetic fails inside a switch/case
9160 statement when compiled with -O2
9161 * [263]18129 -fwritable-strings doesn't work
9163 C++ compiler and library bugs
9165 * [264]10975 incorrect initial ostringstream::tellp()
9166 * [265]11722 Unbuffered filebuf::sgetn is slow
9167 * [266]14534 Unrecognizing static function as a template parameter
9168 when its return value is also templated
9169 * [267]15172 Copy constructor optimization in aggregate
9171 * [268]15786 Bad error message for frequently occuring error.
9172 * [269]16162 Rejects valid member-template-definition
9173 * [270]16612 empty basic_strings can't live in shared memory
9174 * [271]16715 std::basic_iostream is instantiated when used, even
9175 though instantiations are already contained in libstdc++
9176 * [272]16848 code in /ext/demangle.h appears broken
9177 * [273]17132 GCC fails to eliminate function template specialization
9178 when argument deduction fails
9179 * [274]17259 One more _S_leaf incorrectly qualified with _RopeRep::
9181 * [275]17327 use of `enumeral_type' in template type unification
9182 * [276]17393 "unused variable '._0'" warning with -Wall
9183 * [277]17501 Confusion with member templates
9184 * [278]17537 g++ not passing -lstdc++ to linker when all command line
9185 arguments are libraries
9186 * [279]17585 usage of unqualified name of static member from within
9188 * [280]17821 Poor diagnostic for using "." instead of "->"
9189 * [281]17829 wrong error: call of overloaded function is ambiguous
9190 * [282]17851 Misleading diagnostic for invalid function declarations
9191 with undeclared types
9192 * [283]17976 Destructor is called twice
9193 * [284]18020 rejects valid definition of enum value in template
9194 * [285]18093 bogus conflict in namespace aliasing
9195 * [286]18140 C++ parser bug when using >> in templates
9199 * [287]17541 data statements with double precision constants fail
9203 * [288]17853 -O2 ICE for MMX testcase
9207 * [289]17245 ICE compiling gsl-1.5 statistics/lag1.c
9211 * [290]17167 FATAL:Symbol L_foo$stub already defined.
9215 * [291]17277 could not catch an exception when specified -maix64
9219 * [292]17505 <cmath> calls acosf(), ceilf(), and other functions
9220 missing from system libraries
9224 * [293]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl
9228 * [294]17384 ICE with mode attribute on structures
9232 * [295]17770 No NOP after LWL with -mips1
9234 Other embedded target specific
9236 * [296]11476 [arc-elf] gcc ICE on newlib's vfprintf.c
9237 * [297]14064 [avr-elf] -fdata-sections triggers ICE
9238 * [298]14678 [m68hc11-elf] gcc ICE
9239 * [299]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__
9240 * [300]15790 [i686-coff] Alignment error building gcc with i686-coff
9242 * [301]15886 [SH] Miscompilation with -O2 -fPIC
9243 * [302]16884 [avr-elf] [fweb related] bug while initializing
9246 Bugs relating to debugger support
9248 * [303]13841 missing debug info for _Complex function arguments
9249 * [304]15860 [big-endian targets] No DW_AT_location debug info is
9250 emitted for formal arguments to a function that uses "register"
9253 Testsuite issues (compiler not affected)
9255 * [305]17465 Testsuite in libffi overrides LD_LIBRARY_PATH
9256 * [306]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH
9257 * [307]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit
9262 * [308]15498 typo in gcc manual: non-existing locale example en_UK,
9264 * [309]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap:
9265 document broken shell
9266 * [310]16406 USE_LD_AS_NEEDED undocumented
9267 __________________________________________________________________
9271 This is the [311]list of problem reports (PRs) from GCC's bug tracking
9272 system that are known to be fixed in the 3.4.4 release. This list might
9273 not be complete (that is, it is possible that some PRs that have been
9274 fixed are not listed here).
9275 __________________________________________________________________
9279 This is the [312]list of problem reports (PRs) from GCC's bug tracking
9280 system that are known to be fixed in the 3.4.5 release. This list might
9281 not be complete (that is, it is possible that some PRs that have been
9282 fixed are not listed here).
9286 * [313]24688 sco_math fixincl breaks math.h
9290 * [314]17188 struct Foo { } redefinition
9291 * [315]20187 wrong code for ((unsigned char)(unsigned long
9292 long)((a?a:1)&(a*b)))?0:1)
9293 * [316]21873 infinite warning loop on bad array initializer
9294 * [317]21899 enum definition accepts values to be overriden
9295 * [318]22061 ICE in find_function_data, in function.c
9296 * [319]22308 Failure to diagnose violation of constraint 6.516p2
9297 * [320]22458 ICE on missing brace
9298 * [321]22589 ICE casting to long long
9299 * [322]24101 Segfault with preprocessed source
9301 C++ compiler and library bugs
9303 * [323]10611 operations on vector mode not recognized in C++
9304 * [324]13377 unexpected behavior of namespace usage directive
9305 * [325]16002 Strange error message with new parser
9306 * [326]17413 local classes as template argument
9307 * [327]17609 spurious error message after using keyword
9308 * [328]17618 ICE in cp_convert_to_pointer, in cp/cvt.c
9309 * [329]18124 ICE with invalid template template parameter
9310 * [330]18155 typedef in template declaration not rejected
9311 * [331]18177 ICE with const_cast for undeclared variable
9312 * [332]18368 C++ error message regression
9313 * [333]16378 ICE when returning a copy of a packed member
9314 * [334]18466 int ::i; accepted
9315 * [335]18512 ICE on invalid usage of template base class
9316 * [336]18454 ICE when returning undefined type
9317 * [337]18738 typename not allowed with non-dependent qualified name
9318 * [338]18803 rejects access to operator() in template
9319 * [339]19004 ICE in uses_template_parms, in cp/pt.c
9320 * [340]19208 Spurious error about variably modified type
9321 * [341]18253 bad error message / ICE for invalid template parameter
9322 * [342]19608 ICE after friend function definition in local class
9323 * [343]19884 ICE on explicit instantiation of a non-template
9325 * [344]20153 ICE when C++ template function contains anonymous union
9326 * [345]20563 Infinite loop in diagnostic (and ice after error
9328 * [346]20789 ICE with incomplete type in template
9329 * [347]21336 Internal compiler error when using custom new operators
9330 * [348]21768 ICE in error message due to violation of coding
9332 * [349]21853 constness of pointer to data member ignored
9333 * [350]21903 Default argument of template function causes a
9335 * [351]21983 multiple diagnostics
9336 * [352]21987 New testsuite failure
9337 g++.dg/warn/conversion-function-1.C
9338 * [353]22153 ICE on invalid template specialization
9339 * [354]22172 Internal compiler error, seg fault.
9340 * [355]21286 filebuf::xsgetn vs pipes
9341 * [356]22233 ICE with wrong number of template parameters
9342 * [357]22508 ICE after invalid operator new
9343 * [358]22545 ICE with pointer to class member & user defined
9345 * [359]23528 Wrong default allocator in ext/hash_map
9346 * [360]23550 char_traits requirements/1.cc test bad math
9347 * [361]23586 Bad diagnostic for invalid namespace-name
9348 * [362]23624 ICE in invert_truthvalue, in fold-const.c
9349 * [363]23639 Bad error message: not a member of '<declaration error>'
9350 * [364]23797 ICE on typename outside template
9351 * [365]23965 Bogus error message: no matching function for call to
9353 * [366]24052 &#`label_decl' not supported by dump_expr#<expression
9355 * [367]24580 virtual base class cause exception not to be caught
9357 Problems in generated debug information
9359 * [368]24267 Bad DWARF for altivec vectors
9361 Optimizations issues
9363 * [369]17810 ICE in verify_local_live_at_start
9364 * [370]17860 Wrong generated code for loop with varying bound
9365 * [371]21709 ICE on compile-time complex NaN
9366 * [372]21964 broken tail call at -O2 or more
9367 * [373]22167 Strange optimization bug when using -Os
9368 * [374]22619 Compilation failure for real_const_1.f and
9370 * [375]23241 Invalid code generated for comparison of uchar to 255
9371 * [376]23478 Miscompilation due to reloading of a var that is also
9373 * [377]24470 segmentation fault in cc1plus when compiling with -O
9374 * [378]24950 ICE in operand_subword_force
9376 Precompiled headers problems
9378 * [379]14400 Cannot compile qt-x11-free-3.3.0
9379 * [380]14940 PCH largefile test fails on various platforms
9383 * [381]20239 ICE on empty preprocessed input
9384 * [382]15220 "gcc -E -MM -MG" reports missing system headers in
9389 * [383]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on
9394 * [384]21888 bootstrap failure with linker relaxation enabled
9398 * [385]15342 [arm-linux]: ICE in verify_local_live_at_start
9399 * [386]23985 Memory aliasing information incorrect in inlined memcpy
9403 * [387]16719 Illegal move of byte into address register causes
9408 * [388]21723 ICE while building libgfortran
9409 * [389]21841 -mhp-ld/-mgnu-ld documentation
9413 * [390]23644 IA-64 hardware models and configuration options
9415 * [391]24718 Shared libgcc not used for linking by default
9419 * [392]18421 ICE in reload_cse_simplify_operands, in postreload.c
9423 * [393]20621 ICE in change_address_1, in emit-rtl.c
9425 PowerPC and PowerPC64 specific
9427 * [394]18583 error on valid code: const
9428 __attribute__((altivec(vector__))) doesn't work in arrays
9429 * [395]20191 ICE in reload_cse_simplify_operands
9430 * [396]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined
9431 * [397]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set
9432 * [398]23404 gij trashes args of functions with more than 8 fp args
9433 * [399]23539 C & C++ compiler generating misaligned references
9434 regardless of compiler flags
9435 * [400]24102 floatdisf2_internal2 broken
9436 * [401]24465 -mminimal-toc miscompilation of __thread vars
9440 * [402]19933 Problem with define of HUGE_VAL in math_c99
9441 * [403]21889 Native Solaris assembler cannot grok DTP-relative debug
9446 * [404]19300 PCH failures on sparc-linux
9447 * [405]20301 Assembler labels have a leading "-"
9448 * [406]20673 C PCH testsuite assembly comparison failure
9450 x86 and x86_64 specific
9452 * [407]18582 ICE with arrays of type V2DF
9453 * [408]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2
9455 * [409]21716 ICE in reg-stack.c's swap_rtx_condition
9456 * [410]24315 amd64 fails -fpeephole2
9457 __________________________________________________________________
9461 This is the [411]list of problem reports (PRs) from GCC's bug tracking
9462 system that are known to be fixed in the 3.4.6 release. This list might
9463 not be complete (that is, it is possible that some PRs that have been
9464 fixed are not listed here).
9467 For questions related to the use of GCC, please consult these web
9468 pages and the [412]GCC manuals. If that fails, the
9469 [413]gcc-help@gcc.gnu.org mailing list might help. Comments on these
9470 web pages and the development of GCC are welcome on our developer
9471 list at [414]gcc@gcc.gnu.org. All of [415]our lists have public
9474 Copyright (C) [416]Free Software Foundation, Inc. Verbatim copying and
9475 distribution of this entire article is permitted in any medium,
9476 provided this notice is preserved.
9478 These pages are [417]maintained by the GCC team. Last modified
9483 1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
9484 2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus
9485 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
9486 4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems
9487 5. http://gcc.gnu.org/gcc-3.4/mips-abi.html
9488 6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
9489 7. http://www.boost.org/
9490 8. https://gcc.gnu.org/PR11953
9491 9. https://gcc.gnu.org/PR8361
9492 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other Builtins
9493 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209
9494 12. http://gcc.gnu.org/bugs/#cxx_rvalbind
9495 13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
9496 14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
9497 15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
9498 16. http://www.gnu.org/software/classpath/
9499 17. http://www.eclipse.org/
9500 18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html
9501 19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html
9502 20. http://h30097.www3.hp.com/docs/base_doc/DOCUMENTATION/V51A_HTML/ARH9MBTE/DTMNPLTN.HTM#normal-argument-list-structure
9503 21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
9504 22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html
9505 23. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
9506 24. http://gcc.gnu.org/gcc-3.4/mips-abi.html
9507 25. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html
9508 26. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
9509 27. 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
9510 28. https://gcc.gnu.org/PR10129
9511 29. https://gcc.gnu.org/PR14576
9512 30. https://gcc.gnu.org/PR14760
9513 31. https://gcc.gnu.org/PR14671
9514 32. https://gcc.gnu.org/PR15093
9515 33. https://gcc.gnu.org/PR15178
9516 34. https://gcc.gnu.org/PR12753
9517 35. https://gcc.gnu.org/PR13985
9518 36. https://gcc.gnu.org/PR14810
9519 37. https://gcc.gnu.org/PR14883
9520 38. https://gcc.gnu.org/PR15044
9521 39. https://gcc.gnu.org/PR15057
9522 40. https://gcc.gnu.org/PR15064
9523 41. https://gcc.gnu.org/PR15142
9524 42. https://gcc.gnu.org/PR15159
9525 43. https://gcc.gnu.org/PR15165
9526 44. https://gcc.gnu.org/PR15193
9527 45. https://gcc.gnu.org/PR15209
9528 46. https://gcc.gnu.org/PR15227
9529 47. https://gcc.gnu.org/PR15285
9530 48. https://gcc.gnu.org/PR15299
9531 49. https://gcc.gnu.org/PR15329
9532 50. https://gcc.gnu.org/PR15550
9533 51. https://gcc.gnu.org/PR15554
9534 52. https://gcc.gnu.org/PR15640
9535 53. https://gcc.gnu.org/PR15666
9536 54. https://gcc.gnu.org/PR15696
9537 55. https://gcc.gnu.org/PR15701
9538 56. https://gcc.gnu.org/PR15761
9539 57. https://gcc.gnu.org/PR15829
9540 58. https://gcc.gnu.org/PR14538
9541 59. https://gcc.gnu.org/PR12391
9542 60. https://gcc.gnu.org/PR14649
9543 61. https://gcc.gnu.org/PR15004
9544 62. https://gcc.gnu.org/PR15749
9545 63. https://gcc.gnu.org/PR10646
9546 64. https://gcc.gnu.org/PR12077
9547 65. https://gcc.gnu.org/PR13598
9548 66. https://gcc.gnu.org/PR14211
9549 67. https://gcc.gnu.org/PR14220
9550 68. https://gcc.gnu.org/PR14245
9551 69. https://gcc.gnu.org/PR14340
9552 70. https://gcc.gnu.org/PR14600
9553 71. https://gcc.gnu.org/PR14668
9554 72. https://gcc.gnu.org/PR14775
9555 73. https://gcc.gnu.org/PR14821
9556 74. https://gcc.gnu.org/PR14930
9557 75. https://gcc.gnu.org/PR14932
9558 76. https://gcc.gnu.org/PR14950
9559 77. https://gcc.gnu.org/PR14962
9560 78. https://gcc.gnu.org/PR14975
9561 79. https://gcc.gnu.org/PR15002
9562 80. https://gcc.gnu.org/PR15025
9563 81. https://gcc.gnu.org/PR15046
9564 82. https://gcc.gnu.org/PR15069
9565 83. https://gcc.gnu.org/PR15074
9566 84. https://gcc.gnu.org/PR15083
9567 85. https://gcc.gnu.org/PR15096
9568 86. https://gcc.gnu.org/PR15287
9569 87. https://gcc.gnu.org/PR15317
9570 88. https://gcc.gnu.org/PR15337
9571 89. https://gcc.gnu.org/PR15361
9572 90. https://gcc.gnu.org/PR15412
9573 91. https://gcc.gnu.org/PR15427
9574 92. https://gcc.gnu.org/PR15471
9575 93. https://gcc.gnu.org/PR15503
9576 94. https://gcc.gnu.org/PR15507
9577 95. https://gcc.gnu.org/PR15542
9578 96. https://gcc.gnu.org/PR15565
9579 97. https://gcc.gnu.org/PR15625
9580 98. https://gcc.gnu.org/PR15629
9581 99. https://gcc.gnu.org/PR15742
9582 100. https://gcc.gnu.org/PR15775
9583 101. https://gcc.gnu.org/PR15821
9584 102. https://gcc.gnu.org/PR15862
9585 103. https://gcc.gnu.org/PR15875
9586 104. https://gcc.gnu.org/PR15877
9587 105. https://gcc.gnu.org/PR15947
9588 106. https://gcc.gnu.org/PR16020
9589 107. https://gcc.gnu.org/PR16154
9590 108. https://gcc.gnu.org/PR16174
9591 109. https://gcc.gnu.org/PR14315
9592 110. https://gcc.gnu.org/PR15151
9593 111. https://gcc.gnu.org/PR7993
9594 112. https://gcc.gnu.org/PR15228
9595 113. https://gcc.gnu.org/PR15345
9596 114. https://gcc.gnu.org/PR15945
9597 115. https://gcc.gnu.org/PR15526
9598 116. https://gcc.gnu.org/PR14690
9599 117. https://gcc.gnu.org/PR15112
9600 118. https://gcc.gnu.org/PR15067
9601 119. https://gcc.gnu.org/PR1963
9602 120. https://gcc.gnu.org/PR15717
9603 121. https://gcc.gnu.org/PR14782
9604 122. https://gcc.gnu.org/PR14828
9605 123. https://gcc.gnu.org/PR15202
9606 124. https://gcc.gnu.org/PR14610
9607 125. https://gcc.gnu.org/PR14813
9608 126. https://gcc.gnu.org/PR14857
9609 127. https://gcc.gnu.org/PR15598
9610 128. https://gcc.gnu.org/PR15653
9611 129. https://gcc.gnu.org/PR15189
9612 130. https://gcc.gnu.org/PR15331
9613 131. https://gcc.gnu.org/PR16144
9614 132. https://gcc.gnu.org/PR16176
9615 133. https://gcc.gnu.org/PR11591
9616 134. https://gcc.gnu.org/PR12028
9617 135. https://gcc.gnu.org/PR14478
9618 136. https://gcc.gnu.org/PR14567
9619 137. https://gcc.gnu.org/PR14715
9620 138. https://gcc.gnu.org/PR14902
9621 139. https://gcc.gnu.org/PR14924
9622 140. https://gcc.gnu.org/PR14960
9623 141. https://gcc.gnu.org/PR15106
9624 142. https://gcc.gnu.org/PR16026
9625 143. https://gcc.gnu.org/PR15191
9626 144. https://gcc.gnu.org/PR15662
9627 145. https://gcc.gnu.org/PR15054
9628 146. https://gcc.gnu.org/PR15783
9629 147. https://gcc.gnu.org/PR15626
9630 148. https://gcc.gnu.org/PR14326
9631 149. https://gcc.gnu.org/PR14723
9632 150. https://gcc.gnu.org/PR15290
9633 151. https://gcc.gnu.org/PR15250
9634 152. https://gcc.gnu.org/PR15551
9635 153. https://gcc.gnu.org/PR8309
9636 154. https://gcc.gnu.org/PR13250
9637 155. https://gcc.gnu.org/PR13803
9638 156. https://gcc.gnu.org/PR14093
9639 157. https://gcc.gnu.org/PR14457
9640 158. https://gcc.gnu.org/PR14542
9641 159. https://gcc.gnu.org/PR15100
9642 160. https://gcc.gnu.org/PR15296
9643 161. https://gcc.gnu.org/PR15396
9644 162. https://gcc.gnu.org/PR15782
9645 163. https://gcc.gnu.org/PR11610
9646 164. https://gcc.gnu.org/PR15488
9647 165. https://gcc.gnu.org/PR15489
9648 166. https://gcc.gnu.org/PR13928
9649 167. https://gcc.gnu.org/PR14150
9650 168. https://gcc.gnu.org/PR14949
9651 169. https://gcc.gnu.org/PR15123
9652 170. https://gcc.gnu.org/PR16469
9653 171. https://gcc.gnu.org/PR16344
9654 172. https://gcc.gnu.org/PR16842
9655 173. https://gcc.gnu.org/PR12608
9656 174. https://gcc.gnu.org/PR14492
9657 175. https://gcc.gnu.org/PR15461
9658 176. https://gcc.gnu.org/PR15890
9659 177. https://gcc.gnu.org/PR16180
9660 178. https://gcc.gnu.org/PR16224
9661 179. https://gcc.gnu.org/PR16408
9662 180. https://gcc.gnu.org/PR16529
9663 181. https://gcc.gnu.org/PR16698
9664 182. https://gcc.gnu.org/PR16706
9665 183. https://gcc.gnu.org/PR16810
9666 184. https://gcc.gnu.org/PR16851
9667 185. https://gcc.gnu.org/PR16870
9668 186. https://gcc.gnu.org/PR16904
9669 187. https://gcc.gnu.org/PR16905
9670 188. https://gcc.gnu.org/PR16964
9671 189. https://gcc.gnu.org/PR17068
9672 190. https://gcc.gnu.org/PR16366
9673 191. https://gcc.gnu.org/PR15345
9674 192. https://gcc.gnu.org/PR16590
9675 193. https://gcc.gnu.org/PR16693
9676 194. https://gcc.gnu.org/PR17078
9677 195. https://gcc.gnu.org/PR13956
9678 196. https://gcc.gnu.org/PR16684
9679 197. https://gcc.gnu.org/PR12658
9680 198. https://gcc.gnu.org/PR13092
9681 199. https://gcc.gnu.org/PR15320
9682 200. https://gcc.gnu.org/PR16246
9683 201. https://gcc.gnu.org/PR16273
9684 202. https://gcc.gnu.org/PR16401
9685 203. https://gcc.gnu.org/PR16411
9686 204. https://gcc.gnu.org/PR16489
9687 205. https://gcc.gnu.org/PR16618
9688 206. https://gcc.gnu.org/PR16637
9689 207. https://gcc.gnu.org/PR16717
9690 208. https://gcc.gnu.org/PR16813
9691 209. https://gcc.gnu.org/PR16853
9692 210. https://gcc.gnu.org/PR16889
9693 211. https://gcc.gnu.org/PR16959
9694 212. https://gcc.gnu.org/PR7587
9695 213. https://gcc.gnu.org/PR16473
9696 214. https://gcc.gnu.org/PR16478
9697 215. https://gcc.gnu.org/PR10695
9698 216. https://gcc.gnu.org/PR16974
9699 217. https://gcc.gnu.org/PR16298
9700 218. https://gcc.gnu.org/PR17113
9701 219. https://gcc.gnu.org/PR14697
9702 220. https://gcc.gnu.org/PR15869
9703 221. https://gcc.gnu.org/PR16325
9704 222. https://gcc.gnu.org/PR16357
9705 223. https://gcc.gnu.org/PR16380
9706 224. https://gcc.gnu.org/PR16407
9707 225. https://gcc.gnu.org/PR16643
9708 226. https://gcc.gnu.org/PR15927
9709 227. https://gcc.gnu.org/PR15948
9710 228. https://gcc.gnu.org/PR17019
9711 229. https://gcc.gnu.org/PR16130
9712 230. https://gcc.gnu.org/PR16142
9713 231. https://gcc.gnu.org/PR16278
9714 232. https://gcc.gnu.org/PR16414
9715 233. https://gcc.gnu.org/PR16445
9716 234. https://gcc.gnu.org/PR16490
9717 235. https://gcc.gnu.org/PR16683
9718 236. https://gcc.gnu.org/PR16195
9719 237. https://gcc.gnu.org/PR16239
9720 238. https://gcc.gnu.org/PR16199
9721 239. https://gcc.gnu.org/PR16416
9722 240. https://gcc.gnu.org/PR16430
9723 241. https://gcc.gnu.org/PR16379
9724 242. https://gcc.gnu.org/PR17093
9725 243. https://gcc.gnu.org/PR17119
9726 244. https://gcc.gnu.org/PR15928
9727 245. https://gcc.gnu.org/PR16210
9728 246. https://gcc.gnu.org/PR15488
9729 247. https://gcc.gnu.org/PR16250
9730 248. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3
9731 249. https://gcc.gnu.org/PR17369
9732 250. https://gcc.gnu.org/PR17850
9733 251. https://gcc.gnu.org/PR13948
9734 252. https://gcc.gnu.org/PR14492
9735 253. https://gcc.gnu.org/PR16301
9736 254. https://gcc.gnu.org/PR16566
9737 255. https://gcc.gnu.org/PR17023
9738 256. https://gcc.gnu.org/PR17027
9739 257. https://gcc.gnu.org/PR17524
9740 258. https://gcc.gnu.org/PR17826
9741 259. https://gcc.gnu.org/PR15526
9742 260. https://gcc.gnu.org/PR16999
9743 261. https://gcc.gnu.org/PR17503
9744 262. https://gcc.gnu.org/PR17581
9745 263. https://gcc.gnu.org/PR18129
9746 264. https://gcc.gnu.org/PR10975
9747 265. https://gcc.gnu.org/PR11722
9748 266. https://gcc.gnu.org/PR14534
9749 267. https://gcc.gnu.org/PR15172
9750 268. https://gcc.gnu.org/PR15786
9751 269. https://gcc.gnu.org/PR16162
9752 270. https://gcc.gnu.org/PR16612
9753 271. https://gcc.gnu.org/PR16715
9754 272. https://gcc.gnu.org/PR16848
9755 273. https://gcc.gnu.org/PR17132
9756 274. https://gcc.gnu.org/PR17259
9757 275. https://gcc.gnu.org/PR17327
9758 276. https://gcc.gnu.org/PR17393
9759 277. https://gcc.gnu.org/PR17501
9760 278. https://gcc.gnu.org/PR17537
9761 279. https://gcc.gnu.org/PR17585
9762 280. https://gcc.gnu.org/PR17821
9763 281. https://gcc.gnu.org/PR17829
9764 282. https://gcc.gnu.org/PR17851
9765 283. https://gcc.gnu.org/PR17976
9766 284. https://gcc.gnu.org/PR18020
9767 285. https://gcc.gnu.org/PR18093
9768 286. https://gcc.gnu.org/PR18140
9769 287. https://gcc.gnu.org/PR17541
9770 288. https://gcc.gnu.org/PR17853
9771 289. https://gcc.gnu.org/PR17245
9772 290. https://gcc.gnu.org/PR17167
9773 291. https://gcc.gnu.org/PR17277
9774 292. https://gcc.gnu.org/PR17505
9775 293. https://gcc.gnu.org/PR17684
9776 294. https://gcc.gnu.org/PR17384
9777 295. https://gcc.gnu.org/PR17770
9778 296. https://gcc.gnu.org/PR11476
9779 297. https://gcc.gnu.org/PR14064
9780 298. https://gcc.gnu.org/PR14678
9781 299. https://gcc.gnu.org/PR15583
9782 300. https://gcc.gnu.org/PR15790
9783 301. https://gcc.gnu.org/PR15886
9784 302. https://gcc.gnu.org/PR16884
9785 303. https://gcc.gnu.org/PR13841
9786 304. https://gcc.gnu.org/PR15860
9787 305. https://gcc.gnu.org/PR17465
9788 306. https://gcc.gnu.org/PR17469
9789 307. https://gcc.gnu.org/PR18138
9790 308. https://gcc.gnu.org/PR15498
9791 309. https://gcc.gnu.org/PR15747
9792 310. https://gcc.gnu.org/PR16406
9793 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4
9794 312. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5
9795 313. https://gcc.gnu.org/PR24688
9796 314. https://gcc.gnu.org/PR17188
9797 315. https://gcc.gnu.org/PR20187
9798 316. https://gcc.gnu.org/PR21873
9799 317. https://gcc.gnu.org/PR21899
9800 318. https://gcc.gnu.org/PR22061
9801 319. https://gcc.gnu.org/PR22208
9802 320. https://gcc.gnu.org/PR22458
9803 321. https://gcc.gnu.org/PR22589
9804 322. https://gcc.gnu.org/PR24101
9805 323. https://gcc.gnu.org/PR10611
9806 324. https://gcc.gnu.org/PR13377
9807 325. https://gcc.gnu.org/PR16002
9808 326. https://gcc.gnu.org/PR17413
9809 327. https://gcc.gnu.org/PR17609
9810 328. https://gcc.gnu.org/PR17618
9811 329. https://gcc.gnu.org/PR18124
9812 330. https://gcc.gnu.org/PR18155
9813 331. https://gcc.gnu.org/PR18177
9814 332. https://gcc.gnu.org/PR18368
9815 333. https://gcc.gnu.org/PR18378
9816 334. https://gcc.gnu.org/PR18466
9817 335. https://gcc.gnu.org/PR18512
9818 336. https://gcc.gnu.org/PR18545
9819 337. https://gcc.gnu.org/PR18738
9820 338. https://gcc.gnu.org/PR18803
9821 339. https://gcc.gnu.org/PR19004
9822 340. https://gcc.gnu.org/PR19208
9823 341. https://gcc.gnu.org/PR19253
9824 342. https://gcc.gnu.org/PR19608
9825 343. https://gcc.gnu.org/PR19884
9826 344. https://gcc.gnu.org/PR20153
9827 345. https://gcc.gnu.org/PR20563
9828 346. https://gcc.gnu.org/PR20789
9829 347. https://gcc.gnu.org/PR21336
9830 348. https://gcc.gnu.org/PR21768
9831 349. https://gcc.gnu.org/PR21853
9832 350. https://gcc.gnu.org/PR21903
9833 351. https://gcc.gnu.org/PR21983
9834 352. https://gcc.gnu.org/PR21987
9835 353. https://gcc.gnu.org/PR22153
9836 354. https://gcc.gnu.org/PR22172
9837 355. https://gcc.gnu.org/PR21286
9838 356. https://gcc.gnu.org/PR22233
9839 357. https://gcc.gnu.org/PR22508
9840 358. https://gcc.gnu.org/PR22545
9841 359. https://gcc.gnu.org/PR23528
9842 360. https://gcc.gnu.org/PR23550
9843 361. https://gcc.gnu.org/PR23586
9844 362. https://gcc.gnu.org/PR23624
9845 363. https://gcc.gnu.org/PR23639
9846 364. https://gcc.gnu.org/PR23797
9847 365. https://gcc.gnu.org/PR23965
9848 366. https://gcc.gnu.org/PR24052
9849 367. https://gcc.gnu.org/PR24580
9850 368. https://gcc.gnu.org/PR24267
9851 369. https://gcc.gnu.org/PR17810
9852 370. https://gcc.gnu.org/PR17860
9853 371. https://gcc.gnu.org/PR21709
9854 372. https://gcc.gnu.org/PR21964
9855 373. https://gcc.gnu.org/PR22167
9856 374. https://gcc.gnu.org/PR22619
9857 375. https://gcc.gnu.org/PR23241
9858 376. https://gcc.gnu.org/PR23478
9859 377. https://gcc.gnu.org/PR24470
9860 378. https://gcc.gnu.org/PR24950
9861 379. https://gcc.gnu.org/PR14400
9862 380. https://gcc.gnu.org/PR14940
9863 381. https://gcc.gnu.org/PR20239
9864 382. https://gcc.gnu.org/PR15220
9865 383. https://gcc.gnu.org/PR19275
9866 384. https://gcc.gnu.org/PR21888
9867 385. https://gcc.gnu.org/PR15342
9868 386. https://gcc.gnu.org/PR23985
9869 387. https://gcc.gnu.org/PR16719
9870 388. https://gcc.gnu.org/PR21723
9871 389. https://gcc.gnu.org/PR21841
9872 390. https://gcc.gnu.org/PR23644
9873 391. https://gcc.gnu.org/PR24718
9874 392. https://gcc.gnu.org/PR18421
9875 393. https://gcc.gnu.org/PR20621
9876 394. https://gcc.gnu.org/PR18583
9877 395. https://gcc.gnu.org/PR20191
9878 396. https://gcc.gnu.org/PR22083
9879 397. https://gcc.gnu.org/PR23070
9880 398. https://gcc.gnu.org/PR23404
9881 399. https://gcc.gnu.org/PR23539
9882 400. https://gcc.gnu.org/PR24102
9883 401. https://gcc.gnu.org/PR24465
9884 402. https://gcc.gnu.org/PR19933
9885 403. https://gcc.gnu.org/PR21889
9886 404. https://gcc.gnu.org/PR19300
9887 405. https://gcc.gnu.org/PR20301
9888 406. https://gcc.gnu.org/PR20673
9889 407. https://gcc.gnu.org/PR18582
9890 408. https://gcc.gnu.org/PR19340
9891 409. https://gcc.gnu.org/PR21716
9892 410. https://gcc.gnu.org/PR24315
9893 411. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6
9894 412. https://gcc.gnu.org/onlinedocs/
9895 413. mailto:gcc-help@gcc.gnu.org
9896 414. mailto:gcc@gcc.gnu.org
9897 415. https://gcc.gnu.org/lists.html
9898 416. http://www.fsf.org/
9899 417. https://gcc.gnu.org/about.html
9900 418. http://validator.w3.org/check/referer
9901 ======================================================================
9902 http://gcc.gnu.org/gcc-3.3/index.html
9903 GCC 3.3 Release Series
9907 The [1]GNU project and the GCC developers are pleased to announce the
9908 release of GCC 3.3.6.
9910 This release is a bug-fix release, containing fixes for regressions in
9911 GCC 3.3.5 relative to previous releases of GCC.
9913 This release is the last of the series 3.3.x.
9915 The GCC 3.3 release series includes numerous [2]new features,
9916 improvements, bug fixes, and other changes, thanks to an [3]amazing
9917 group of volunteers.
9922 May 3, 2005 ([4]changes)
9925 September 30, 2004 ([5]changes)
9928 May 31, 2004 ([6]changes)
9931 February 14, 2004 ([7]changes)
9934 October 16, 2003 ([8]changes)
9937 August 8, 2003 ([9]changes)
9940 May 14, 2003 ([10]changes)
9942 References and Acknowledgements
9944 GCC used to stand for the GNU C Compiler, but since the compiler
9945 supports several other languages aside from C, it now stands for the
9946 GNU Compiler Collection.
9948 A list of [11]successful builds is updated as new information becomes
9951 The GCC developers would like to thank the numerous people that have
9952 contributed new features, improvements, bug fixes, and other changes as
9953 well as test results to GCC. This [12]amazing group of volunteers is
9954 what makes GCC successful.
9956 For additional information about GCC please refer to the [13]GCC
9957 project web site or contact the [14]GCC development mailing list.
9959 To obtain GCC please use [15]our mirror sites, or our CVS server.
9962 For questions related to the use of GCC, please consult these web
9963 pages and the [16]GCC manuals. If that fails, the
9964 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
9965 web pages and the development of GCC are welcome on our developer
9966 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
9969 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
9970 distribution of this entire article is permitted in any medium,
9971 provided this notice is preserved.
9973 These pages are [21]maintained by the GCC team. Last modified
9978 1. http://www.gnu.org/
9979 2. http://gcc.gnu.org/gcc-3.3/changes.html
9980 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
9981 4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
9982 5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5
9983 6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4
9984 7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3
9985 8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2
9986 9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1
9987 10. http://gcc.gnu.org/gcc-3.3/changes.html
9988 11. http://gcc.gnu.org/gcc-3.3/buildstat.html
9989 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
9990 13. http://gcc.gnu.org/index.html
9991 14. mailto:gcc@gcc.gnu.org
9992 15. http://gcc.gnu.org/mirrors.html
9993 16. https://gcc.gnu.org/onlinedocs/
9994 17. mailto:gcc-help@gcc.gnu.org
9995 18. mailto:gcc@gcc.gnu.org
9996 19. https://gcc.gnu.org/lists.html
9997 20. http://www.fsf.org/
9998 21. https://gcc.gnu.org/about.html
9999 22. http://validator.w3.org/check/referer
10000 ======================================================================
10001 http://gcc.gnu.org/gcc-3.3/changes.html
10002 GCC 3.3 Release Series
10003 Changes, New Features, and Fixes
10005 The latest release in the 3.3 release series is [1]GCC 3.3.6.
10009 * The preprocessor no longer accepts multi-line string literals. They
10010 were deprecated in 3.0, 3.1, and 3.2.
10011 * The preprocessor no longer supports the -A- switch when appearing
10012 alone. -A- followed by an assertion is still supported.
10013 * Support for all the systems [2]obsoleted in GCC 3.1 has been
10014 removed from GCC 3.3. See below for a [3]list of systems which are
10015 obsoleted in this release.
10016 * Checking for null format arguments has been decoupled from the rest
10017 of the format checking mechanism. Programs which use the format
10018 attribute may regain this functionality by using the new [4]nonnull
10019 function attribute. Note that all functions for which GCC has a
10020 built-in format attribute, an appropriate built-in nonnull
10021 attribute is also applied.
10022 * The DWARF (version 1) debugging format has been deprecated and will
10023 be removed in a future version of GCC. Version 2 of the DWARF
10024 debugging format will continue to be supported for the foreseeable
10026 * The C and Objective-C compilers no longer accept the "Naming Types"
10027 extension (typedef foo = bar); it was already unavailable in C++.
10028 Code which uses it will need to be changed to use the "typeof"
10029 extension instead: typedef typeof(bar) foo. (We have removed this
10030 extension without a period of deprecation because it has caused the
10031 compiler to crash since version 3.0 and no one noticed until very
10032 recently. Thus we conclude it is not in widespread use.)
10033 * The -traditional C compiler option has been removed. It was
10034 deprecated in 3.1 and 3.2. (Traditional preprocessing remains
10035 available.) The <varargs.h> header, used for writing variadic
10036 functions in traditional C, still exists but will produce an error
10038 * GCC 3.3.1 automatically places zero-initialized variables in the
10039 .bss section on some operating systems. Versions of GNU Emacs up to
10040 (and including) 21.3 will not work correctly when using this
10041 optimization; you can use -fno-zero-initialized-in-bss to disable
10044 General Optimizer Improvements
10046 * A new scheme for accurately describing processor pipelines, the
10047 [5]DFA scheduler, has been added.
10048 * Pavel Nejedly, Charles University Prague, has contributed new file
10049 format used by the edge coverage profiler (-fprofile-arcs).
10050 The new format is robust and diagnoses common mistakes where
10051 profiles from different versions (or compilations) of the program
10052 are combined resulting in nonsensical profiles and slow code to
10053 produced with profile feedback. Additionally this format allows
10054 extra data to be gathered. Currently, overall statistics are
10055 produced helping optimizers to identify hot spots of a program
10056 globally replacing the old intra-procedural scheme and resulting in
10057 better code. Note that the gcov tool from older GCC versions will
10058 not be able to parse the profiles generated by GCC 3.3 and vice
10060 * Jan Hubicka, SuSE Labs, has contributed a new superblock formation
10061 pass enabled using -ftracer. This pass simplifies the control flow
10062 of functions allowing other optimizations to do better job.
10063 He also contributed the function reordering pass
10064 (-freorder-functions) to optimize function placement using profile
10067 New Languages and Language specific improvements
10071 * The preprocessor now accepts directives within macro arguments. It
10072 processes them just as if they had not been within macro arguments.
10073 * The separate ISO and traditional preprocessors have been completely
10074 removed. The front end handles either type of preprocessed output
10076 * In C99 mode preprocessor arithmetic is done in the precision of the
10077 target's intmax_t, as required by that standard.
10078 * The preprocessor can now copy comments inside macros to the output
10079 file when the macro is expanded. This feature, enabled using the
10080 -CC option, is intended for use by applications which place
10081 metadata or directives inside comments, such as lint.
10082 * The method of constructing the list of directories to be searched
10083 for header files has been revised. If a directory named by a -I
10084 option is a standard system include directory, the option is
10085 ignored to ensure that the default search order for system
10086 directories and the special treatment of system header files are
10088 * A few more [6]ISO C99 features now work correctly.
10089 * A new function attribute, nonnull, has been added which allows
10090 pointer arguments to functions to be specified as requiring a
10091 non-null value. The compiler currently uses this information to
10092 issue a warning when it detects a null value passed in such an
10094 * A new type attribute, may_alias, has been added. Accesses to
10095 objects with types with this attribute are not subjected to
10096 type-based alias analysis, but are instead assumed to be able to
10097 alias any other type of objects, just like the char type.
10101 * Type based alias analysis has been implemented for C++ aggregate
10106 * Generate an error if Objective-C objects are passed by value in
10107 function and method calls.
10108 * When -Wselector is used, check the whole list of selectors at the
10109 end of compilation, and emit a warning if a @selector() is not
10111 * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime.
10112 * No longer need to include objc/objc-class.h to compile self calls
10113 in class methods (NeXT runtime only).
10114 * New -Wundeclared-selector option.
10115 * Removed selector bloating which was causing object files to be 10%
10116 bigger on average (GNU runtime only).
10117 * Using at run time @protocol() objects has been fixed in certain
10118 situations (GNU runtime only).
10119 * Type checking has been fixed and improved in many situations
10120 involving protocols.
10124 * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK
10126 * The JDK 1.4 assert facility has been implemented.
10127 * The bytecode interpreter is now direct threaded and thus faster.
10131 * Fortran improvements are listed in [7]the Fortran documentation.
10135 * Ada tasking now works with glibc 2.3.x threading libraries.
10137 New Targets and Target Specific Improvements
10139 * The following changes have been made to the HP-PA port:
10140 + The port now defaults to scheduling for the PA8000 series of
10142 + Scheduling support for the PA7300 processor has been added.
10143 + The 32-bit port now supports weak symbols under HP-UX 11.
10144 + The handling of initializers and finalizers has been improved
10145 under HP-UX 11. The 64-bit port no longer uses collect2.
10146 + Dwarf2 EH support has been added to the 32-bit GNU/Linux port.
10147 + ABI fixes to correct the passing of small structures by value.
10148 * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to
10149 use the DFA processor pipeline description.
10150 * The following NetBSD configurations for the SuperH processor family
10152 + SH3, big-endian, sh-*-netbsdelf*
10153 + SH3, little-endian, shle-*-netbsdelf*
10154 + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd*
10155 + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd*
10156 + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd*
10157 + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd*
10158 * The following changes have been made to the IA-32/x86-64 port:
10159 + SSE2 and 3dNOW! intrinsics are now supported.
10160 + Support for thread local storage has been added to the IA-32
10162 + The x86-64 port has been significantly improved.
10163 * The following changes have been made to the MIPS port:
10164 + All configurations now accept the -mabi switch. Note that you
10165 will need appropriate multilibs for this option to work
10167 + ELF configurations will always pass an ABI flag to the
10168 assembler, except when the MIPS EABI is selected.
10169 + -mabi=64 no longer selects MIPS IV code.
10170 + The -mcpu option, which was deprecated in 3.1 and 3.2, has
10171 been removed from this release.
10172 + -march now changes the core ISA level. In previous releases,
10173 it would change the use of processor-specific extensions, but
10174 would leave the core ISA unchanged. For example, mips64-elf
10175 -march=r8000 will now generate MIPS IV code.
10176 + Under most configurations, -mipsN now acts as a synonym for
10178 + There are some new preprocessor macros to describe the -march
10179 and -mtune settings. See the documentation of those options
10181 + Support for the NEC VR-Series processors has been added. This
10182 includes the 54xx, 5500, and 41xx series.
10183 + Support for the Sandcraft sr71k processor has been added.
10184 * The following changes have been made to the S/390 port:
10185 + Support to build the Java runtime libraries has been added.
10186 Java is now enabled by default on s390-*-linux* and
10187 s390x-*-linux* targets.
10188 + Multilib support for the s390x-*-linux* target has been added;
10189 this allows to build 31-bit binaries using the -m31 option.
10190 + Support for thread local storage has been added.
10191 + Inline assembler code may now use the 'Q' constraint to
10192 specify memory operands without index register.
10193 + Various platform-specific performance improvements have been
10194 implemented; in particular, the compiler now uses the BRANCH
10195 ON COUNT family of instructions and makes more frequent use of
10196 the TEST UNDER MASK family of instructions.
10197 * The following changes have been made to the PowerPC port:
10198 + Support for IBM Power4 processor added.
10199 + Support for Motorola e500 SPE added.
10200 + Support for AIX 5.2 added.
10201 + Function and Data sections now supported on AIX.
10202 + Sibcall optimizations added.
10203 * The support for H8 Tiny is added to the H8/300 port with -mn.
10207 Support for a number of older systems has been declared obsolete in GCC
10208 3.3. Unless there is activity to revive them, the next release of GCC
10209 will have their sources permanently removed.
10211 All configurations of the following processor architectures have been
10213 * Matsushita MN10200, mn10200-*-*
10214 * Motorola 88000, m88k-*-*
10215 * IBM ROMP, romp-*-*
10217 Also, some individual systems have been obsoleted:
10219 + Interix, alpha*-*-interix*
10220 + Linux libc1, alpha*-*-linux*libc1*
10221 + Linux ECOFF, alpha*-*-linux*ecoff*
10223 + Generic a.out, arm*-*-aout*
10224 + Conix, arm*-*-conix*
10225 + "Old ABI," arm*-*-oabi
10226 + StrongARM/COFF, strongarm-*-coff*
10228 + Generic OSF, hppa1.0-*-osf*
10229 + Generic BSD, hppa1.0-*-bsd*
10230 + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]*
10231 + HiUX, hppa*-*-hiux*
10232 + Mach Lites, hppa*-*-lites*
10234 + Windows NT 3.x, i?86-*-win32
10236 + HP systems, m68000-hp-bsd* and m68k-hp-bsd*
10237 + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and
10239 + AT&T systems, m68000-att-sysv*
10240 + Atari systems, m68k-atari-sysv*
10241 + Motorola systems, m68k-motorola-sysv*
10242 + NCR systems, m68k-ncr-sysv*
10243 + Plexus systems, m68k-plexus-sysv*
10244 + Commodore systems, m68k-cbm-sysv*
10245 + Citicorp TTI, m68k-tti-*
10246 + Unos, m68k-crds-unos*
10247 + Concurrent RTU, m68k-ccur-rtu*
10248 + Linux a.out, m68k-*-linux*aout*
10249 + Linux libc1, m68k-*-linux*libc1*
10250 + pSOS, m68k-*-psos*
10252 + Generic ECOFF, mips*-*-ecoff*
10253 + SINIX, mips-sni-sysv4
10254 + Orion RTEMS, mips64orion-*-rtems*
10255 * National Semiconductor 32000
10256 + OpenBSD, ns32k-*-openbsd*
10257 * POWER (aka RS/6000) and PowerPC
10258 + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]*
10259 + Bull BOSX, rs6000-bull-bosx
10260 + Generic Mach, rs6000-*-mach*
10261 + Generic SysV, powerpc*-*-sysv*
10262 + Linux libc1, powerpc*-*-linux*libc1*
10264 + Generic a.out, sparc-*-aout*, sparclet-*-aout*,
10265 sparclite-*-aout*, and sparc86x-*-aout*
10266 + NetBSD a.out, sparc-*-netbsd*aout*
10267 + Generic BSD, sparc-*-bsd*
10268 + ChorusOS, sparc-*-chorusos*
10269 + Linux a.out, sparc-*-linux*aout*
10270 + Linux libc1, sparc-*-linux*libc1*
10271 + LynxOS, sparc-*-lynxos*
10272 + Solaris on HAL hardware, sparc-hal-solaris2*
10273 + SunOS versions 3 and 4, sparc-*-sunos[34]*
10275 + RTEMS, v850-*-rtems*
10279 Documentation improvements
10281 Other significant improvements
10283 * Almost all front-end dependencies in the compiler have been
10284 separated out into a set of language hooks. This should make adding
10285 a new front end clearer and easier.
10286 * One effect of removing the separate preprocessor is a small
10287 increase in the robustness of the compiler in general, and the
10288 maintainability of target descriptions. Previously target-specific
10289 built-in macros and others, such as __FAST_MATH__, had to be
10290 handled with so-called specs that were hard to maintain. Often they
10291 would fail to behave properly when conflicting options were
10292 supplied on the command line, and define macros in the user's
10293 namespace even when strict ISO compliance was requested.
10294 Integrating the preprocessor has cleanly solved these issues.
10295 * The Makefile suite now supports redirection of make install by
10296 means of the variable DESTDIR.
10297 __________________________________________________________________
10301 Detailed release notes for the GCC 3.3 release follow.
10307 * [8]10140 cross compiler build failures: missing __mempcpy (DUP:
10308 [9]10198,[10]10338)
10310 Internal compiler errors (multi-platform)
10312 * [11]3581 large string causes segmentation fault in cc1
10313 * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler
10314 * [13]5533 (c++) ICE when processing std::accumulate(begin, end,
10316 * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out
10317 * [15]6412 (c++) ICE in retrieve_specialization
10318 * [16]6620 (c++) partial template specialization causes an ICE
10319 (segmentation fault)
10320 * [17]6663 (c++) ICE with attribute aligned
10321 * [18]7068 ICE with incomplete types
10322 * [19]7083 (c++) ICE using -gstabs with dodgy class derivation
10323 * [20]7647 (c++) ICE when data member has the name of the enclosing
10325 * [21]7675 ICE in fixup_var_refs_1
10326 * [22]7718 'complex' template instantiation causes ICE
10327 * [23]8116 (c++) ICE in member template function
10328 * [24]8358 (ada) Ada compiler accesses freed memory, crashes
10329 * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation
10331 * [26]8564 (c++) ICE in find_function_data, in function.c
10332 * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c
10333 * [28]8766 (c++) ICE after failed initialization of static template
10335 * [29]8803 ICE in instantiate_virtual_regs_1, in function.c
10336 * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set
10337 * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class
10339 * [32]9216 (c++) ICE on missing template parameter
10340 * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c
10341 * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO
10343 * [35]9429 (c++) ICE in template instantiation with a pointered new
10345 * [36]9516 Internal error when using a big array
10346 * [37]9600 (c++) ICE with typedefs in template class
10347 * [38]9629 (c++) virtual inheritance segfault
10348 * [39]9672 (c++) ICE: Error reporting routines re-entered
10349 * [40]9749 (c++) ICE in write_expression on invalid function
10351 * [41]9794 (fortran) ICE: floating point exception during constant
10353 * [42]9829 (c++) Missing colon in nested namespace usage causes ICE
10354 * [43]9916 (c++) ICE with noreturn function in ?: statement
10355 * [44]9936 ICE with local function and variable-length 2d array
10356 * [45]10262 (c++) cc1plus crashes with large generated code
10357 * [46]10278 (c++) ICE in parser for invalid code
10358 * [47]10446 (c++) ICE on definition of nonexistent member function of
10359 nested class in a class template
10360 * [48]10451 (c++) ICE in grokdeclarator on spurious mutable
10362 * [49]10506 (c++) ICE in build_new at cp/init.c with
10363 -fkeep-inline-functions and multiple inheritance
10364 * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the
10365 precision of the declared type
10369 * [51]2001 Inordinately long compile times in reload CSE regs
10370 * [52]2391 Exponential compilation time explosion in combine
10371 * [53]2960 Duplicate loop conditions even with -Os
10372 * [54]4046 redundant conditional branch
10373 * [55]6405 Loop-unrolling related performance regressions
10374 * [56]6798 very long compile time with large case-statement
10375 * [57]6871 const objects shouldn't be moved to .bss
10376 * [58]6909 problem w/ -Os on modified loop-2c.c test case
10377 * [59]7189 gcc -O2 -Wall does not print ``control reaches end of
10378 non-void function'' warning
10379 * [60]7642 optimization problem with signbit()
10380 * [61]8634 incorrect code for inlining of memcpy under -O2
10381 * [62]8750 Cygwin prolog generation erroneously emitting __alloca as
10382 regular function call
10386 * [63]2161 long if-else cascade overflows parser stack
10387 * [64]4319 short accepted on typedef'd char
10388 * [65]8602 incorrect line numbers in warning messages when using
10390 * [66]9177 -fdump-translation-unit: C front end deletes function_decl
10391 AST nodes and breaks debugging dumps
10392 * [67]9853 miscompilation of non-constant structure initializer
10394 c++ compiler and library
10396 * [68]45 legal template specialization code is rejected (DUP:
10398 * [70]764 lookup failure: friend operator and dereferencing a pointer
10399 and templates (DUP: [71]5116)
10400 * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP:
10402 * [73]3663 G++ doesn't check access control during template
10404 * [74]3797 gcc fails to emit explicit specialization of a template
10406 * [75]3948 Two destructors are called when no copy destructor is
10407 defined (ABI change)
10408 * [76]4137 Conversion operator within template is not accepted
10409 * [77]4361 bogus ambiguity taking the address of a member template
10410 * [78]4802 g++ accepts illegal template code (access to private
10411 member; DUP: [79]5837)
10412 * [80]4803 inline function is used but never defined, and g++ does
10414 * [81]5094 Partial specialization cannot be friend?
10415 * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66
10416 * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run
10418 * [84]7015 certain __asm__ constructs rejected
10419 * [85]7086 compile time regression (quadratic behavior in
10421 * [86]7099 G++ doesn't set the noreturn attribute on std::exit and
10423 * [87]7247 copy constructor missing when inlining enabled (invalid
10425 * [88]7441 string array initialization compilation time regression
10426 from seconds to minutes
10427 * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong
10428 * [90]7804 bad printing of floating point constant in warning message
10429 * [91]8099 Friend classes and template specializations
10430 * [92]8117 member function pointers and multiple inheritance
10431 * [93]8205 using declaration and multiple inheritance
10432 * [94]8645 unnecessary non-zero checks in stl_tree.h
10433 * [95]8724 explicit destructor call for incomplete class allowed
10434 * [96]8805 compile time regression with many member variables
10435 * [97]8691 -O3 and -fno-implicit-templates are incompatible
10436 * [98]8700 unhelpful error message for binding temp to reference
10437 * [99]8724 explicit destructor call for incomplete class allowed
10438 * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems
10439 * [101]9016 Failure to consistently constant fold "constant" C++
10441 * [102]9053 g++ confused about ambiguity of overloaded function
10443 * [103]9152 undefined virtual thunks
10444 * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out
10445 * [105]9297 data corruption due to codegen bug (when copying.)
10446 * [106]9318 i/ostream::operator>>/<<(streambuf*) broken
10447 * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf
10448 * [108]9400 bogus -Wshadow warning: shadowed declaration of this in
10450 * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters
10451 * [110]9425 filebuf::pbackfail broken (DUP: [111]9439)
10452 * [112]9474 GCC freezes in compiling a weird code mixing <iostream>
10454 * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1)
10456 * [115]9555 ostream inserters fail to set badbit on exception
10457 * [116]9561 ostream inserters rethrow exception of wrong type
10458 * [117]9563 ostream::sentry returns true after a failed preparation
10459 * [118]9582 one-definition rule violation in std::allocator
10460 * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors
10461 * [120]9683 bug in initialization chains for static const variables
10462 from template classes
10463 * [121]9791 -Woverloaded-virtual reports hiding of destructor
10464 * [122]9817 collate::compare doesn't handle nul characters
10465 * [123]9825 filebuf::sputbackc breaks sbumpc
10466 * [124]9826 operator>>(basic_istream, basic_string) fails to compile
10468 * [125]9924 Multiple using statements for builtin functions not
10470 * [126]9946 destructor is not called for temporary object
10471 * [127]9964 filebuf::close() sometimes fails to close file
10472 * [128]9988 filebuf::overflow writes EOF to file
10473 * [129]10033 optimization breaks polymorphic references w/ typeid
10475 * [130]10097 filebuf::underflow drops characters
10476 * [131]10132 filebuf destructor can throw exceptions
10477 * [132]10180 gcc fails to warn about non-inlined function
10478 * [133]10199 method parametrized by template does not work everywhere
10479 * [134]10300 use of array-new (nothrow) in segfaults on NULL return
10480 * [135]10427 Stack corruption with variable-length automatic arrays
10481 and virtual destructors
10482 * [136]10503 Compilation never stops in fixed_type_or_null
10486 * [137]5956 selectors aren't matched properly when added to the
10489 Fortran compiler and library
10491 * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't
10493 * [139]3924 g77 generates code that is rejected by GAS if COFF debug
10495 * [140]5634 doc: explain that configure --prefix=~/... does not work
10496 * [141]6367 multiple repeat counts confuse namelist read into array
10497 * [142]6491 Logical operations error on logicals when using
10499 * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C"
10500 * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os
10502 * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should
10503 assume a direct access file
10504 * [146]7278 g77 "bug"; the executable misbehaves (with -O2
10506 * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows
10507 * [148]7388 Incorrect output with 0-based array of characters
10508 * [149]8587 Double complex zero ** double precision number -> NaN
10510 * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning:
10511 unknown register name line-length-none
10512 * [151]10197 Direct access files not unformatted by default
10514 Java compiler and library
10516 * [152]6005 gcj fails to build rhug on alpha
10517 * [153]6389 System.getProperty("") should always throw an
10518 IllegalArgumentException
10519 * [154]6576 java.util.ResourceBundle.getResource ignores locale
10520 * [155]6652 new java.io.File("").getCanonicalFile() throws exception
10521 * [156]7060 getMethod() doesn't search super interface
10522 * [157]7073 bytecode interpreter gives wrong answer for interface
10524 * [158]7180 possible bug in
10525 javax.naming.spi.NamingManager.getPlusPath()
10526 * [159]7416 java.security startup refs "GNU libgcj.security"
10527 * [160]7570 Runtime.exec with null envp: child doesn't inherit parent
10528 env (DUP: [161]7578)
10529 * [162]7611 Internal error while compiling libjava with -O
10530 * [163]7709 NullPointerException in _Jv_ResolvePoolEntry
10531 * [164]7766 ZipInputStream.available returns 0 immediately after
10533 * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public
10534 * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented
10535 * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens'
10536 * [168]8234 ZipInputStream chokes when InputStream.read() returns
10538 * [169]8415 reflection bug: exception info for Method
10539 * [170]8481 java.Random.nextInt(int) may return negative
10540 * [171]8593 Error reading GZIPped files with BufferedReader
10541 * [172]8759 java.beans.Introspector has no flushCaches() or
10542 flushFromCaches() methods
10543 * [173]8997 spin() calls Thread.sleep
10544 * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd
10545 instead of the root content of C:
10546 * [175]9254 java::lang::Object::wait(), threads-win32.cc returns
10548 * [176]9271 Severe bias in java.security.SecureRandom
10550 Ada compiler and library
10552 * [177]6767 make gnatlib-shared fails on -laddr2line
10553 * [178]9911 gnatmake fails to link when GCC configured with
10554 --with-sjlj-exceptions=yes
10555 * [179]10020 Can't bootstrap gcc on AIX with Ada enabled
10556 * [180]10546 Ada tasking not working on Red Hat 9
10560 * [181]7029 preprocessor should ignore #warning with -M
10564 * [182]2903 [arm] Optimization bug with long long arithmetic
10565 * [183]7873 arm-linux-gcc fails when assigning address to a bit field
10569 * [184]7680 float functions undefined in math.h/cmath with #define
10572 HP-UX or HP-PA-specific
10574 * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c
10575 * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to
10577 * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen
10581 * [188]6744 Bad assembler code generated: reference to pseudo
10583 * [189]7361 Internal compiler error in reload_cse_simplify_operands,
10588 * [190]9496 [mips-linux] bug in optimizer?
10592 * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of
10594 * [192]8480 reload ICEs for LAPACK code on powerpc64-linux
10595 * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg
10596 * [194]10315 [powerpc] ICE: in extract_insn, in recog.c
10600 * [195]10267 (documentation) Wrong build instructions for
10603 x86-specific (Intel/AMD)
10605 * [196]7916 ICE in instantiate_virtual_register_1
10606 * [197]7926 (c++) i486 instructions in header files make c++ programs
10608 * [198]8555 ICE in gen_split_1231
10609 * [199]8994 ICE with -O -march=pentium4
10610 * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs
10611 * [201]9806 ICE in inline assembly with -fPIC flag
10612 * [202]10077 gcc -msse2 generates movd to move dwords between xmm
10614 * [203]10233 64-bit comparison only comparing bottom 32-bits
10615 * [204]10286 type-punning doesn't work with __m64 and -O
10616 * [205]10308 [x86] ICE with -O -fgcse or -O2
10617 __________________________________________________________________
10623 This section lists the problem reports (PRs) from GCC's bug tracking
10624 system that are known to be fixed in the 3.3.1 release. This list might
10625 not be complete (that is, it is possible that some PRs that have been
10626 fixed are not listed here).
10630 * [206]11272 [Solaris] make bootstrap fails while building libstdc++
10632 Internal compiler errors (multi-platform)
10634 * [207]5754 ICE on invalid nested template class
10635 * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64
10636 and --enable-checking
10637 * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c
10638 * [210]7053 (c++) ICE when declaring a function already defined as a
10639 friend method of a template class
10640 * [211]8164 (c++) ICE when using different const expressions as
10642 * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c
10643 * [213]9559 (c++) ICE with invalid initialization of a static const
10644 * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c
10645 when redeclaring a static member variable
10646 * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in
10647 dwarfout.c with -g -O -finline-functions
10648 * [216]10432 (c++) ICE in poplevel, in cp/decl.c
10649 * [217]10475 ICE in subreg_highpart_offset for code with long long
10650 * [218]10635 (c++) ICE when dereferencing an incomplete type casted
10651 from a void pointer
10652 * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while
10653 instantiating static member variables
10654 * [220]10700 ICE in copy_to_mode_reg on 64-bit targets
10655 * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c
10656 * [222]10796 (c++) ICE when defining an enum with two values: -1 and
10658 * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x
10660 * [224]10939 (c++) ICE with template code
10661 * [225]10956 (c++) ICE when specializing a template member function
10662 of a template class, in tsubst, in cp/pt.c
10663 * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*()
10665 * [227]11059 (c++) ICE with empty union
10666 * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with
10667 -O2 -fnon-call-exceptions
10668 * [229]11105 (c++) ICE in mangle_conv_op_name_for_type
10669 * [230]11149 (c++) ICE on error when instantiation with call function
10671 * [231]11228 (c++) ICE on new-expression using array operator new and
10672 default-initialization
10673 * [232]11282 (c++) Infinite memory usage after syntax error
10674 * [233]11301 (fortran) ICE with -fno-globals
10675 * [234]11308 (c++) ICE when using an enum type name as if it were a
10677 * [235]11473 (c++) ICE with -gstabs when empty struct inherits from
10679 * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR
10680 * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c:
10681 template member functions
10685 * [238]11198 -O2 -frename-registers generates wrong code (aliasing
10687 * [239]11304 Wrong code production with -fomit-frame-pointer
10688 * [240]11381 volatile memory access optimized away
10689 * [241]11536 [strength-reduce] -O2 optimization produces wrong code
10690 * [242]11557 constant folding bug generates wrong code
10694 * [243]5897 No warning for statement after return
10695 * [244]11279 DWARF-2 output mishandles large enums
10699 * [245]11022 no warning for non-compatible macro redefinition
10701 C++ compiler and library
10703 * [246]2330 static_cast<>() to a private base is allowed
10704 * [247]5388 Incorrect message "operands to ?: have different types"
10705 * [248]5390 Libiberty fails to demangle multi-digit template
10707 * [249]7877 Incorrect parameter passing to specializations of member
10709 * [250]9393 Anonymous namespaces and compiling the same file twice
10710 * [251]10032 -pedantic converts some errors to warnings
10711 * [252]10468 const typeof(x) is non-const, but only in templates
10712 * [253]10527 confused error message with "new int()" parameter
10714 * [254]10679 parameter MIN_INLINE_INSNS is not honored
10715 * [255]10682 gcc chokes on a typedef for an enum inside a class
10717 * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of
10719 * [257]10845 template member function (with nested template as
10720 parameter) cannot be called anymore if another unrelated template
10721 member function is defined
10722 * [258]10849 Cannot define an out-of-class specialization of a
10723 private nested template class
10724 * [259]10888 Suppress -Winline warnings for system headers
10725 * [260]10929 -Winline warns about functions for which no definition
10727 * [261]10931 valid conversion static_cast<const unsigned
10728 int&>(lvalue-of-type-int) is rejected
10729 * [262]10940 Bad code with explicit specialization
10730 * [263]10968 If member function implicitly instantiated, explicit
10731 instantiation of class fails to instantiate it
10732 * [264]10990 Cannot convert with dynamic_cast<> to a private base
10733 class from within a member function
10734 * [265]11039 Bad interaction between implicit typename deprecation
10736 * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say
10737 "__unused__" instead
10738 * [267]11095 C++ iostream manipulator causes segfault when called
10739 with negative argument
10740 * [268]11098 g++ doesn't emit complete debugging information for
10741 local variables in destructors
10742 * [269]11137 GNU/Linux shared library constructors not called unless
10743 there's one global object
10744 * [270]11154 spurious ambiguity report for template class
10746 * [271]11329 Compiler cannot find user defined implicit typecast
10747 * [272]11332 Spurious error with casts in ?: expression
10748 * [273]11431 static_cast behavior with subclasses when default
10749 constructor available
10750 * [274]11528 money_get facet does not accept "$.00" as valid
10751 * [275]11546 Type lookup problems in out-of-line definition of a
10752 class doubly nested from a template class
10753 * [276]11567 C++ code containing templated member function with same
10754 name as pure virtual member function results in linking failure
10755 * [277]11645 Failure to deal with using and private inheritance
10757 Java compiler and library
10759 * [278]5179 Qualified static field access doesn't initialize its
10761 * [279]8204 gcj -O2 to native reorders certain instructions
10763 * [280]10838 java.io.ObjectInputStream syntax error
10764 * [281]10886 The RMI registry that comes with GCJ does not work
10766 * [282]11349 JNDI URL context factories not located correctly
10768 x86-specific (Intel/AMD)
10770 * [283]4823 ICE on inline assembly code
10771 * [284]8878 miscompilation with -O and SSE
10772 * [285]9815 (c++ library) atomicity.h - fails to compile with -O3
10774 * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads,
10776 * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2
10777 * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source
10778 * [289]11044 [x86] out of range loop instructions for FP code on K6
10779 * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE
10781 * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC"
10784 SPARC- or Solaris- specific
10786 * [292]9362 solaris 'as' dies when fed .s and "-gstabs"
10787 * [293]10142 [SPARC64] gcc produces wrong code when passing
10788 structures by value
10789 * [294]10663 New configure check aborts with Sun tools.
10790 * [295]10835 combinatorial explosion in scheduler on HyperSPARC
10791 * [296]10876 ICE in calculate_giv_inc when building KDE
10792 * [297]10955 wrong code at -O3 for structure argument in context of
10794 * [298]11018 -mcpu=ultrasparc busts tar-1.13.25
10795 * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x
10800 * [300]10907 gcc violates the ia64 ABI (GP must be preserved)
10801 * [301]11320 scheduler bug (in machine depended reorganization pass)
10802 * [302]11599 bug with conditional and __builtin_prefetch
10806 * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem
10808 * [304]10871 error in rs6000_stack_info save_size computation
10809 * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse
10814 * [306]7594 [m68k] ICE on legal code associated with simplify-rtx
10815 * [307]10557 [m68k] ICE in subreg_offset_representable_p
10816 * [308]11054 [m68k] ICE in reg_overlap_mentioned_p
10820 * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for
10821 functions with __attribute__ ((interrupt ("IRQ")))
10822 * [310]10842 [arm] Clobbered link register is copied to pc under
10823 certain circumstances
10824 * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes
10825 * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno
10830 * [313]11084 ICE in propagate_one_insn, in flow.c
10834 * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf
10835 * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c
10836 * [316]11096 i686-linux to sh-linux cross compiler fails to compile
10839 GNU/Linux (or Hurd?) specific
10841 * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3
10845 * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare
10848 Cygwin (or mingw) specific
10850 * [319]5287 ICE with dllimport attribute
10851 * [320]10148 [MingW/CygWin] Compiler dumps core
10855 * [321]8787 GCC fails to emit .intel_syntax when invoked with
10856 -masm=intel on DJGPP
10858 Darwin (and MacOS X) specific
10860 * [322]10900 trampolines crash
10864 * [323]1607 (c++) Format attributes on methods undocumented
10865 * [324]4252 Invalid option `-fdump-translation-unit'
10866 * [325]4490 Clarify restrictions on -m96bit-long-double,
10867 -m128bit-long-double
10868 * [326]10355 document an issue with regparm attribute on some systems
10870 * [327]10726 (fortran) Documentation for function "IDate Intrinsic
10872 * [328]10805 document bug in old version of Sun assembler
10873 * [329]10815 warn against GNU binutils on AIX
10874 * [330]10877 document need for newer binutils on i?86-*-linux-gnu
10875 * [331]11280 Manual incorrect with respect to -freorder-blocks
10876 * [332]11466 Document -mlittle-endian and its restrictions for the
10879 Testsuite bugs (compiler itself is not affected)
10881 * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly
10883 * [334]10810 gcc-3.3 fails make check: buffer overrun in
10885 __________________________________________________________________
10891 This section lists the problem reports (PRs) from GCC's bug tracker
10892 that are known to be fixed in the 3.3.2 release. This list might not be
10893 complete (that is, it is possible that some PRs that have been fixed
10894 are not listed here).
10896 Bootstrap failures and problems
10898 * [335]8336 [SCO5] bootstrap config still tries to use COFF options
10899 * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with
10900 --enable-threads=posix
10901 * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap
10902 * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare
10904 * [339]11687 xstormy16-elf build fails in libf2c
10905 * [340]12263 [SGI IRIX] bootstrap fails during compile of
10906 libf2c/libI77/backspace.c
10907 * [341]12490 buffer overflow in scan-decls.c (during Solaris 9
10908 fix-header processing)
10910 Internal compiler errors (multi-platform)
10912 * [342]7277 Casting integers to vector types causes ICE
10913 * [343]7939 (c++) ICE on invalid function template specialization
10914 * [344]11063 (c++) ICE on parsing initialization list of const array
10916 * [345]11207 ICE with negative index in array element designator
10917 * [346]11522 (fortran) g77 dwarf-2 ICE in
10918 add_abstract_origin_attribute
10919 * [347]11595 (c++) ICE on duplicate label definition
10920 * [348]11646 (c++) ICE in commit_one_edge_insertion with
10921 -fnon-call-exceptions -fgcse -O
10922 * [349]11665 ICE in struct initializer when taking address
10923 * [350]11852 (c++) ICE with bad struct initializer.
10924 * [351]11878 (c++) ICE in cp_expr_size
10925 * [352]11883 ICE with any -O on mercury-generated C code
10926 * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in
10927 cp/typeck2.c when applying typeid operator to template template
10929 * [354]12146 ICE in lookup_template_function, in cp/pt.c
10930 * [355]12215 ICE in make_label_edge with -fnon-call-exceptions
10932 * [356]12369 (c++) ICE with templates and friends
10933 * [357]12446 ICE in emit_move_insn on complicated array reference
10934 * [358]12510 ICE in final_scan_insn
10935 * [359]12544 ICE with large parameters used in nested functions
10937 C and optimization bugs
10939 * [360]9862 spurious warnings with -W -finline-functions
10940 * [361]10962 lookup_field is a linear search on a linked list (can be
10941 slow if large struct)
10942 * [362]11370 -Wunreachable-code gives false complaints
10943 * [363]11637 invalid assembly with -fnon-call-exceptions
10944 * [364]11885 Problem with bitfields in packed structs
10945 * [365]12082 Inappropriate unreachable code warnings
10946 * [366]12180 Inline optimization fails for variadic function
10947 * [367]12340 loop unroller + gcse produces wrong code
10949 C++ compiler and library
10951 * [368]3907 nested template parameter collides with member name
10952 * [369]5293 confusing message when binding a temporary to a reference
10953 * [370]5296 [DR115] Pointers to functions and to template functions
10954 behave differently in deduction
10955 * [371]7939 ICE on function template specialization
10956 * [372]8656 Unable to assign function with __attribute__ and pointer
10957 return type to an appropriate variable
10958 * [373]10147 Confusing error message for invalid template function
10960 * [374]11400 std::search_n() makes assumptions about Size parameter
10961 * [375]11409 issues with using declarations, overloading, and
10963 * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle
10964 multiple bits in mask
10965 * [377]11786 operator() call on variable in other namespace not
10967 * [378]11867 static_cast ignores ambiguity
10968 * [379]11928 bug with conversion operators that are typedefs
10969 * [380]12114 Uninitialized memory accessed in dtor
10970 * [381]12163 static_cast + explicit constructor regression
10971 * [382]12181 Wrong code with comma operator and c++
10972 * [383]12236 regparm and fastcall messes up parameters
10973 * [384]12266 incorrect instantiation of unneeded template during
10974 overload resolution
10975 * [385]12296 istream::peek() doesn't set eofbit
10976 * [386]12298 [sjlj exceptions] Stack unwind destroys
10977 not-yet-constructed object
10978 * [387]12369 ICE with templates and friends
10979 * [388]12337 apparently infinite loop in g++
10980 * [389]12344 stdcall attribute ignored if function returns a pointer
10981 * [390]12451 missing(late) class forward declaration in cxxabi.h
10982 * [391]12486 g++ accepts invalid use of a qualified name
10984 x86 specific (Intel/AMD)
10986 * [392]8869 [x86 MMX] ICE with const variable optimization and MMX
10988 * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions
10990 * [394]11689 g++3.3 emits un-assembleable code for k6 architecture
10991 * [395]12116 [k6] Invalid assembly output values with X-MAME code
10992 * [396]12070 ICE converting between double and long double with
10997 * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc
10998 * [398]11535 __builtin_return_address may not work on ia64
10999 * [399]11693 [ia64] ICE in gen_nop_type
11000 * [400]12224 [ia64] Thread-local storage doesn't work
11004 * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux
11006 * [402]11319 loop miscompiled on ppc32
11007 * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code
11011 * [404]11662 wrong code for expr. with cast to long long and
11013 * [405]11965 invalid assembler code for a shift < 32 operation
11014 * [406]12301 (c++) stack corruption when a returned expression throws
11019 * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of
11024 * [408]11313 problem with #pragma weak and static inline functions
11025 * [409]11712 __STDC_EXT__ not defined for C++ by default anymore?
11029 * [410]12166 Profiled programs crash if PROFDIR is set
11031 Solaris-x86 specific
11033 * [411]12101 i386 Solaris no longer works with GNU as?
11035 Miscellaneous embedded target-specific bugs
11037 * [412]10988 [m32r-elf] wrong blockmove code with -O3
11038 * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with
11040 * [414]11902 [sh4] spec file improperly inserts rpath even when none
11042 * [415]11903 [sh4] -pthread fails to link due to error in spec file
11044 __________________________________________________________________
11050 In addition to the bug fixes documented below, this release contains
11051 few minor features such as:
11052 * Support for --with-sysroot
11053 * Support for automatic detection of executable stacks
11054 * Support for SSE3 instructions
11055 * Support for thread local storage debugging under GDB on S390
11059 This section lists the problem reports (PRs) from GCC's bug tracker
11060 that are known to be fixed in the 3.3.3 release. This list might not be
11061 complete (that is, it is possible that some PRs that have been fixed
11062 are not listed here).
11064 Bootstrap failures and issues
11066 * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails
11067 * [417]12399 boehm-gc fails (when building a cross compiler): libtool
11068 unable to infer tagged configuration
11069 * [418]13068 mklibgcc.in doesn't handle multi-level multilib
11070 subdirectories properly
11072 Internal compiler errors (multi-platform)
11074 * [419]10060 ICE (stack overflow) on huge file (300k lines) due to
11075 recursive behaviour of copy_rtx_if_shared, in emit_rtl.c
11076 * [420]10555 (c++) ICE on undefined template argument
11077 * [421]10706 (c++) ICE in mangle_class_name_for_template
11078 * [422]11496 (fortran) error in flow_loops_find when -funroll-loops
11080 * [423]11741 ICE in pre_insert_copy_insn, in gcse.c
11081 * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0
11082 * [425]12632 (fortran) -fbounds-check ICE
11083 * [426]12712 (c++) ICE on short legit C++ code fragment with gcc
11085 * [427]12726 (c++) ICE (segfault) on trivial code
11086 * [428]12890 (c++) ICE on compilation of class with throwing method
11087 * [429]12900 (c++) ICE in rtl_verify_flow_info_1
11088 * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on
11089 correct code with -O2 -fno-force-mem
11090 * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive
11092 * [432]13318 ICE: floating point exception in the loop optimizer
11093 * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in
11095 * [434]13574 (c++) invalid array default initializer in class lets
11096 gcc consume all memory and die
11097 * [435]13475 ICE on SIMD variables with partial value initialization
11098 * [436]13797 (c++) ICE on invalid template parameter
11099 * [437]13824 (java) gcj SEGV with simple .java program
11101 C and optimization bugs
11103 * [438]8776 loop invariants are not removed (most likely)
11104 * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing
11106 * [440]11350 undefined labels with -Os -fPIC
11107 * [441]12826 Optimizer removes reference through volatile pointer
11108 * [442]12500 stabs debug info: void no longer a predefined / builtin
11110 * [443]12941 builtin-bitops-1.c miscompilation (latent bug)
11111 * [444]12953 tree inliner bug (in inline_forbidden_p) and fix
11112 * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled
11113 * [446]13507 spurious printf format warning
11114 * [447]13382 Type information for const pointer disappears during
11116 * [448]13394 noreturn attribute ignored on recursive invokation
11117 * [449]13400 Compiled code crashes storing to read-only location
11118 * [450]13521 Endless loop in calculate_global_regs_live
11120 C++ compiler and library
11122 Some of the bug fixes in this list were made to implement decisions
11123 that the ISO C++ standards committee has made concerning several defect
11124 reports (DRs). Links in the list below point to detailed discussion of
11125 the relevant defect report.
11126 * [451]2094 unimplemented: use of `ptrmem_cst' in template type
11128 * [452]2294 using declaration confusion
11129 * [453]5050 template instantiation depth exceeds limit: recursion
11131 * [454]9371 Bad exception handling in
11132 i/ostream::operator>>/<<(streambuf*)
11133 * [455]9546 bad exception handling in ostream members
11134 * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the
11135 face of unknown locales
11136 * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work
11137 * [459]10095 istream::operator>>(int&) sets ios::badbit when
11138 ios::failbit is set.
11139 * [460]11554 Warning about reordering of initializers doesn't mention
11140 location of constructor
11141 * [461]12297 istream::sentry::sentry() handles eof() incorrectly.
11142 * [462]12352 Exception safety problems in src/localename.cc
11143 * [463]12438 Memory leak in locale::combine()
11144 * [464]12540 Memory leak in locale::locale(const char*)
11145 * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented
11146 * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented
11147 * [470]12696 memory eating infinite loop in diagnostics (error
11149 * [471]12815 Code compiled with optimization behaves unexpectedly
11150 * [472]12862 Conflicts between typedefs/enums and namespace member
11152 * [473]12926 Wrong value after assignment in initialize list using
11154 * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented
11155 * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented
11156 * [478]13007 basic_streambuf::pubimbue, imbue wrong
11157 * [479]13009 Implicitly-defined assignment operator writes to wrong
11159 * [480]13057 regparm attribute not applied to destructor
11160 * [481]13070 -Wformat option ignored in g++
11161 * [482]13081 forward template declarations in <complex> let inlining
11163 * [483]13239 Assertion does not seem to work correctly anymore
11164 * [484]13262 "xxx is private within this context" when initializing a
11165 self-contained template class
11166 * [485]13290 simple typo in concept checking for std::generate_n
11167 * [486]13323 Template code does not compile in presence of typedef
11168 * [487]13369 __verify_grouping (and __add_grouping?) not correct
11169 * [488]13371 infinite loop with packed struct and inlining
11170 * [489]13445 Template argument replacement "dereferences" a typedef
11171 * [490]13461 Fails to access protected-ctor from public constant
11172 * [491]13462 Non-standard-conforming type set::pointer
11173 * [492]13478 gcc uses wrong constructor to initialize a const
11175 * [493]13544 "conflicting types" for enums in different scopes
11176 * [494]13650 string::compare should not (always) use
11177 traits_type::length()
11178 * [495]13683 bogus warning about passing non-PODs through ellipsis
11179 * [496]13688 Derived class is denied access to protected base class
11181 * [497]13774 Member variable cleared in virtual multiple inheritance
11183 * [498]13884 Protect sstream.tcc from extern template use
11185 Java compiler and library
11187 * [499]10746 [win32] garbage collection crash in GCJ
11189 Objective-C compiler and library
11191 * [500]11433 Crash due to dereferencing null pointer when querying
11194 Fortran compiler and library
11196 * [501]12633 logical expression gives incorrect result with
11197 -fugly-logint option
11198 * [502]13037 [gcse-lm] g77 generates incorrect code
11199 * [503]13213 Hex constant problem when compiling with -fugly-logint
11202 x86-specific (Intel/AMD)
11204 * [504]4490 ICE with -m128bit-long-double
11205 * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have
11206 `reg' in make_field_assignment, in combine.c
11207 * [506]12441 ICE: can't find a register to spill
11208 * [507]12943 array static-init failure under -fpic, -fPIC
11209 * [508]13608 Incorrect code with -O3 -ffast-math
11213 * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of
11214 __attribute__((aligned(16)))
11215 * [510]11793 ICE in extract_insn, in recog.c (const_vector's)
11216 * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in
11218 * [512]12537 g++ generates writeable text sections
11222 * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0
11224 * [514]12865 mprotect call to make trampoline executable may fail
11225 * [515]13354 ICE in sparc_emit_set_const32
11229 * [516]10467 [arm] ICE in pre_insert_copy_insn,
11233 * [517]11226 ICE passing struct arg with two floats
11234 * [518]11227 ICE for _Complex float, _Complex long double args
11235 * [519]12644 GCC 3.3.2 fails to compile glibc on ia64
11236 * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn
11237 * Various fixes for libunwind
11241 * [521]12654 Incorrect comparison code generated for Alpha
11242 * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2
11243 * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2
11247 * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c
11248 * [525]12158 [hppa] compilation does not terminate at -O1
11252 * [526]11992 Wrong built-in code for memcmp with length 1<<24: only
11253 (1<<24)-1 possible for CLCL-Instruction
11257 * [527]9365 segfault in gen_far_branch (config/sh/sh.c)
11258 * [528]10392 optimizer generates faulty array indexing
11259 * [529]11322 SH profiler outputs multiple definitions of symbol
11260 * [530]13069 gcc/config/sh/rtems.h broken
11261 * [531]13302 Putting a va_list in a struct causes seg fault
11262 * [532]13585 Incorrect optimization of call to sfunc
11263 * Fix inappropriately exported libgcc functions from the shared
11266 Other embedded target specific
11268 * [533]8916 [mcore] unsigned char assign gets hosed.
11269 * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c
11270 * [535]13122 [h8300] local variable gets corrupted by function call
11271 when -fomit-frame-pointer is given
11272 * [536]13256 [cris] strict_low_part mistreated in delay slots
11273 * [537]13373 [mcore] optimization with -frerun-cse-after-loop
11274 -fexpensive-optimizations produces wrong code on mcore
11278 * [538]12561 gcc/config/t-gnu needs updating to work with
11281 Tru64 Unix specific
11283 * [539]6243 testsuite fails almost all tests due to no libintl in
11284 LD_LIBRARY_PATH during test.
11285 * [540]11397 weak aliases broken on Tru64 UNIX
11289 * [541]12505 build failure due to defines of uchar in cpphash.h and
11291 * [542]13150 WEAK symbols not exported by collect2
11295 * [543]12666 fixincludes problem on IRIX 6.5.19m
11299 * [544]12969 Including sys/byteorder.h breaks configure checks
11301 Testsuite problems (compiler is not affected)
11303 * [545]10819 testsuite creates CR+LF on compiler version lines in
11305 * [546]11612 abi_check not finding correct libgcc_s.so.1
11309 * [547]13211 using -###, incorrect warnings about unused linker file
11311 __________________________________________________________________
11315 This is the [548]list of problem reports (PRs) from GCC's bug tracking
11316 system that are known to be fixed in the 3.3.4 release. This list might
11317 not be complete (that is, it is possible that some PRs that have been
11318 fixed are not listed here).
11319 __________________________________________________________________
11323 This is the [549]list of problem reports (PRs) from GCC's bug tracking
11324 system that are known to be fixed in the 3.3.5 release. This list might
11325 not be complete (that is, it is possible that some PRs that have been
11326 fixed are not listed here).
11327 __________________________________________________________________
11331 This is the [550]list of problem reports (PRs) from GCC's bug tracking
11332 system that are known to be fixed in the 3.3.6 release. This list might
11333 not be complete (that is, it is possible that some PRs that have been
11334 fixed are not listed here).
11337 For questions related to the use of GCC, please consult these web
11338 pages and the [551]GCC manuals. If that fails, the
11339 [552]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11340 web pages and the development of GCC are welcome on our developer
11341 list at [553]gcc@gcc.gnu.org. All of [554]our lists have public
11344 Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and
11345 distribution of this entire article is permitted in any medium,
11346 provided this notice is preserved.
11348 These pages are [556]maintained by the GCC team. Last modified
11353 1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
11354 2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems
11355 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
11356 4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute
11357 5. http://gcc.gnu.org/news/dfa.html
11358 6. http://gcc.gnu.org/c99status.html
11359 7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html
11360 8. https://gcc.gnu.org/PR10140
11361 9. https://gcc.gnu.org/PR10198
11362 10. https://gcc.gnu.org/PR10338
11363 11. https://gcc.gnu.org/PR3581
11364 12. https://gcc.gnu.org/PR4382
11365 13. https://gcc.gnu.org/PR5533
11366 14. https://gcc.gnu.org/PR6387
11367 15. https://gcc.gnu.org/PR6412
11368 16. https://gcc.gnu.org/PR6620
11369 17. https://gcc.gnu.org/PR6663
11370 18. https://gcc.gnu.org/PR7068
11371 19. https://gcc.gnu.org/PR7083
11372 20. https://gcc.gnu.org/PR7647
11373 21. https://gcc.gnu.org/PR7675
11374 22. https://gcc.gnu.org/PR7718
11375 23. https://gcc.gnu.org/PR8116
11376 24. https://gcc.gnu.org/PR8358
11377 25. https://gcc.gnu.org/PR8511
11378 26. https://gcc.gnu.org/PR8564
11379 27. https://gcc.gnu.org/PR8660
11380 28. https://gcc.gnu.org/PR8766
11381 29. https://gcc.gnu.org/PR8803
11382 30. https://gcc.gnu.org/PR8846
11383 31. https://gcc.gnu.org/PR8906
11384 32. https://gcc.gnu.org/PR9216
11385 33. https://gcc.gnu.org/PR9261
11386 34. https://gcc.gnu.org/PR9263
11387 35. https://gcc.gnu.org/PR9429
11388 36. https://gcc.gnu.org/PR9516
11389 37. https://gcc.gnu.org/PR9600
11390 38. https://gcc.gnu.org/PR9629
11391 39. https://gcc.gnu.org/PR9672
11392 40. https://gcc.gnu.org/PR9749
11393 41. https://gcc.gnu.org/PR9794
11394 42. https://gcc.gnu.org/PR9829
11395 43. https://gcc.gnu.org/PR9916
11396 44. https://gcc.gnu.org/PR9936
11397 45. https://gcc.gnu.org/PR10262
11398 46. https://gcc.gnu.org/PR10278
11399 47. https://gcc.gnu.org/PR10446
11400 48. https://gcc.gnu.org/PR10451
11401 49. https://gcc.gnu.org/PR10506
11402 50. https://gcc.gnu.org/PR10549
11403 51. https://gcc.gnu.org/PR2001
11404 52. https://gcc.gnu.org/PR2391
11405 53. https://gcc.gnu.org/PR2960
11406 54. https://gcc.gnu.org/PR4046
11407 55. https://gcc.gnu.org/PR6405
11408 56. https://gcc.gnu.org/PR6798
11409 57. https://gcc.gnu.org/PR6871
11410 58. https://gcc.gnu.org/PR6909
11411 59. https://gcc.gnu.org/PR7189
11412 60. https://gcc.gnu.org/PR7642
11413 61. https://gcc.gnu.org/PR8634
11414 62. https://gcc.gnu.org/PR8750
11415 63. https://gcc.gnu.org/PR2161
11416 64. https://gcc.gnu.org/PR4319
11417 65. https://gcc.gnu.org/PR8602
11418 66. https://gcc.gnu.org/PR9177
11419 67. https://gcc.gnu.org/PR9853
11420 68. https://gcc.gnu.org/PR45
11421 69. https://gcc.gnu.org/PR3784
11422 70. https://gcc.gnu.org/PR764
11423 71. https://gcc.gnu.org/PR5116
11424 72. https://gcc.gnu.org/PR2862
11425 73. https://gcc.gnu.org/PR3663
11426 74. https://gcc.gnu.org/PR3797
11427 75. https://gcc.gnu.org/PR3948
11428 76. https://gcc.gnu.org/PR4137
11429 77. https://gcc.gnu.org/PR4361
11430 78. https://gcc.gnu.org/PR4802
11431 79. https://gcc.gnu.org/PR5837
11432 80. https://gcc.gnu.org/PR4803
11433 81. https://gcc.gnu.org/PR5094
11434 82. https://gcc.gnu.org/PR5730
11435 83. https://gcc.gnu.org/PR6713
11436 84. https://gcc.gnu.org/PR7015
11437 85. https://gcc.gnu.org/PR7086
11438 86. https://gcc.gnu.org/PR7099
11439 87. https://gcc.gnu.org/PR7247
11440 88. https://gcc.gnu.org/PR7441
11441 89. https://gcc.gnu.org/PR7768
11442 90. https://gcc.gnu.org/PR7804
11443 91. https://gcc.gnu.org/PR8099
11444 92. https://gcc.gnu.org/PR8117
11445 93. https://gcc.gnu.org/PR8205
11446 94. https://gcc.gnu.org/PR8645
11447 95. https://gcc.gnu.org/PR8724
11448 96. https://gcc.gnu.org/PR8805
11449 97. https://gcc.gnu.org/PR8691
11450 98. https://gcc.gnu.org/PR8700
11451 99. https://gcc.gnu.org/PR8724
11452 100. https://gcc.gnu.org/PR8949
11453 101. https://gcc.gnu.org/PR9016
11454 102. https://gcc.gnu.org/PR9053
11455 103. https://gcc.gnu.org/PR9152
11456 104. https://gcc.gnu.org/PR9182
11457 105. https://gcc.gnu.org/PR9297
11458 106. https://gcc.gnu.org/PR9318
11459 107. https://gcc.gnu.org/PR9320
11460 108. https://gcc.gnu.org/PR9400
11461 109. https://gcc.gnu.org/PR9424
11462 110. https://gcc.gnu.org/PR9425
11463 111. https://gcc.gnu.org/PR9439
11464 112. https://gcc.gnu.org/PR9474
11465 113. https://gcc.gnu.org/PR9548
11466 114. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231
11467 115. https://gcc.gnu.org/PR9555
11468 116. https://gcc.gnu.org/PR9561
11469 117. https://gcc.gnu.org/PR9563
11470 118. https://gcc.gnu.org/PR9582
11471 119. https://gcc.gnu.org/PR9622
11472 120. https://gcc.gnu.org/PR9683
11473 121. https://gcc.gnu.org/PR9791
11474 122. https://gcc.gnu.org/PR9817
11475 123. https://gcc.gnu.org/PR9825
11476 124. https://gcc.gnu.org/PR9826
11477 125. https://gcc.gnu.org/PR9924
11478 126. https://gcc.gnu.org/PR9946
11479 127. https://gcc.gnu.org/PR9964
11480 128. https://gcc.gnu.org/PR9988
11481 129. https://gcc.gnu.org/PR10033
11482 130. https://gcc.gnu.org/PR10097
11483 131. https://gcc.gnu.org/PR10132
11484 132. https://gcc.gnu.org/PR10180
11485 133. https://gcc.gnu.org/PR10199
11486 134. https://gcc.gnu.org/PR10300
11487 135. https://gcc.gnu.org/PR10427
11488 136. https://gcc.gnu.org/PR10503
11489 137. https://gcc.gnu.org/PR5956
11490 138. https://gcc.gnu.org/PR1832
11491 139. https://gcc.gnu.org/PR3924
11492 140. https://gcc.gnu.org/PR5634
11493 141. https://gcc.gnu.org/PR6367
11494 142. https://gcc.gnu.org/PR6491
11495 143. https://gcc.gnu.org/PR6742
11496 144. https://gcc.gnu.org/PR7113
11497 145. https://gcc.gnu.org/PR7236
11498 146. https://gcc.gnu.org/PR7278
11499 147. https://gcc.gnu.org/PR7384
11500 148. https://gcc.gnu.org/PR7388
11501 149. https://gcc.gnu.org/PR8587
11502 150. https://gcc.gnu.org/PR9038
11503 151. https://gcc.gnu.org/PR10197
11504 152. https://gcc.gnu.org/PR6005
11505 153. https://gcc.gnu.org/PR6389
11506 154. https://gcc.gnu.org/PR6576
11507 155. https://gcc.gnu.org/PR6652
11508 156. https://gcc.gnu.org/PR7060
11509 157. https://gcc.gnu.org/PR7073
11510 158. https://gcc.gnu.org/PR7180
11511 159. https://gcc.gnu.org/PR7416
11512 160. https://gcc.gnu.org/PR7570
11513 161. https://gcc.gnu.org/PR7578
11514 162. https://gcc.gnu.org/PR7611
11515 163. https://gcc.gnu.org/PR7709
11516 164. https://gcc.gnu.org/PR7766
11517 165. https://gcc.gnu.org/PR7785
11518 166. https://gcc.gnu.org/PR7786
11519 167. https://gcc.gnu.org/PR8142
11520 168. https://gcc.gnu.org/PR8234
11521 169. https://gcc.gnu.org/PR8415
11522 170. https://gcc.gnu.org/PR8481
11523 171. https://gcc.gnu.org/PR8593
11524 172. https://gcc.gnu.org/PR8759
11525 173. https://gcc.gnu.org/PR8997
11526 174. https://gcc.gnu.org/PR9253
11527 175. https://gcc.gnu.org/PR9254
11528 176. https://gcc.gnu.org/PR9271
11529 177. https://gcc.gnu.org/PR6767
11530 178. https://gcc.gnu.org/PR9911
11531 179. https://gcc.gnu.org/PR10020
11532 180. https://gcc.gnu.org/PR10546
11533 181. https://gcc.gnu.org/PR7029
11534 182. https://gcc.gnu.org/PR2903
11535 183. https://gcc.gnu.org/PR7873
11536 184. https://gcc.gnu.org/PR7680
11537 185. https://gcc.gnu.org/PR8705
11538 186. https://gcc.gnu.org/PR9986
11539 187. https://gcc.gnu.org/PR10056
11540 188. https://gcc.gnu.org/PR6744
11541 189. https://gcc.gnu.org/PR7361
11542 190. https://gcc.gnu.org/PR9496
11543 191. https://gcc.gnu.org/PR7067
11544 192. https://gcc.gnu.org/PR8480
11545 193. https://gcc.gnu.org/PR8784
11546 194. https://gcc.gnu.org/PR10315
11547 195. https://gcc.gnu.org/PR10267
11548 196. https://gcc.gnu.org/PR7916
11549 197. https://gcc.gnu.org/PR7926
11550 198. https://gcc.gnu.org/PR8555
11551 199. https://gcc.gnu.org/PR8994
11552 200. https://gcc.gnu.org/PR9426
11553 201. https://gcc.gnu.org/PR9806
11554 202. https://gcc.gnu.org/PR10077
11555 203. https://gcc.gnu.org/PR10233
11556 204. https://gcc.gnu.org/PR10286
11557 205. https://gcc.gnu.org/PR10308
11558 206. https://gcc.gnu.org/PR11272
11559 207. https://gcc.gnu.org/PR5754
11560 208. https://gcc.gnu.org/PR6597
11561 209. https://gcc.gnu.org/PR6949
11562 210. https://gcc.gnu.org/PR7053
11563 211. https://gcc.gnu.org/PR8164
11564 212. https://gcc.gnu.org/PR8384
11565 213. https://gcc.gnu.org/PR9559
11566 214. https://gcc.gnu.org/PR9649
11567 215. https://gcc.gnu.org/PR9864
11568 216. https://gcc.gnu.org/PR10432
11569 217. https://gcc.gnu.org/PR10475
11570 218. https://gcc.gnu.org/PR10635
11571 219. https://gcc.gnu.org/PR10661
11572 220. https://gcc.gnu.org/PR10700
11573 221. https://gcc.gnu.org/PR10712
11574 222. https://gcc.gnu.org/PR10796
11575 223. https://gcc.gnu.org/PR10890
11576 224. https://gcc.gnu.org/PR10939
11577 225. https://gcc.gnu.org/PR10956
11578 226. https://gcc.gnu.org/PR11041
11579 227. https://gcc.gnu.org/PR11059
11580 228. https://gcc.gnu.org/PR11083
11581 229. https://gcc.gnu.org/PR11105
11582 230. https://gcc.gnu.org/PR11149
11583 231. https://gcc.gnu.org/PR11228
11584 232. https://gcc.gnu.org/PR11282
11585 233. https://gcc.gnu.org/PR11301
11586 234. https://gcc.gnu.org/PR11308
11587 235. https://gcc.gnu.org/PR11473
11588 236. https://gcc.gnu.org/PR11503
11589 237. https://gcc.gnu.org/PR11513
11590 238. https://gcc.gnu.org/PR11198
11591 239. https://gcc.gnu.org/PR11304
11592 240. https://gcc.gnu.org/PR11381
11593 241. https://gcc.gnu.org/PR11536
11594 242. https://gcc.gnu.org/PR11557
11595 243. https://gcc.gnu.org/PR5897
11596 244. https://gcc.gnu.org/PR11279
11597 245. https://gcc.gnu.org/PR11022
11598 246. https://gcc.gnu.org/PR2330
11599 247. https://gcc.gnu.org/PR5388
11600 248. https://gcc.gnu.org/PR5390
11601 249. https://gcc.gnu.org/PR7877
11602 250. https://gcc.gnu.org/PR9393
11603 251. https://gcc.gnu.org/PR10032
11604 252. https://gcc.gnu.org/PR10468
11605 253. https://gcc.gnu.org/PR10527
11606 254. https://gcc.gnu.org/PR10679
11607 255. https://gcc.gnu.org/PR10682
11608 256. https://gcc.gnu.org/PR10689
11609 257. https://gcc.gnu.org/PR10845
11610 258. https://gcc.gnu.org/PR10849
11611 259. https://gcc.gnu.org/PR10888
11612 260. https://gcc.gnu.org/PR10929
11613 261. https://gcc.gnu.org/PR10931
11614 262. https://gcc.gnu.org/PR10940
11615 263. https://gcc.gnu.org/PR10968
11616 264. https://gcc.gnu.org/PR10990
11617 265. https://gcc.gnu.org/PR11039
11618 266. https://gcc.gnu.org/PR11062
11619 267. https://gcc.gnu.org/PR11095
11620 268. https://gcc.gnu.org/PR11098
11621 269. https://gcc.gnu.org/PR11137
11622 270. https://gcc.gnu.org/PR11154
11623 271. https://gcc.gnu.org/PR11329
11624 272. https://gcc.gnu.org/PR11332
11625 273. https://gcc.gnu.org/PR11431
11626 274. https://gcc.gnu.org/PR11528
11627 275. https://gcc.gnu.org/PR11546
11628 276. https://gcc.gnu.org/PR11567
11629 277. https://gcc.gnu.org/PR11645
11630 278. https://gcc.gnu.org/PR5179
11631 279. https://gcc.gnu.org/PR8204
11632 280. https://gcc.gnu.org/PR10838
11633 281. https://gcc.gnu.org/PR10886
11634 282. https://gcc.gnu.org/PR11349
11635 283. https://gcc.gnu.org/PR4823
11636 284. https://gcc.gnu.org/PR8878
11637 285. https://gcc.gnu.org/PR9815
11638 286. https://gcc.gnu.org/PR10402
11639 287. https://gcc.gnu.org/PR10504
11640 288. https://gcc.gnu.org/PR10673
11641 289. https://gcc.gnu.org/PR11044
11642 290. https://gcc.gnu.org/PR11089
11643 291. https://gcc.gnu.org/PR11420
11644 292. https://gcc.gnu.org/PR9362
11645 293. https://gcc.gnu.org/PR10142
11646 294. https://gcc.gnu.org/PR10663
11647 295. https://gcc.gnu.org/PR10835
11648 296. https://gcc.gnu.org/PR10876
11649 297. https://gcc.gnu.org/PR10955
11650 298. https://gcc.gnu.org/PR11018
11651 299. https://gcc.gnu.org/PR11556
11652 300. https://gcc.gnu.org/PR10907
11653 301. https://gcc.gnu.org/PR11320
11654 302. https://gcc.gnu.org/PR11599
11655 303. https://gcc.gnu.org/PR9745
11656 304. https://gcc.gnu.org/PR10871
11657 305. https://gcc.gnu.org/PR11440
11658 306. https://gcc.gnu.org/PR7594
11659 307. https://gcc.gnu.org/PR10557
11660 308. https://gcc.gnu.org/PR11054
11661 309. https://gcc.gnu.org/PR10834
11662 310. https://gcc.gnu.org/PR10842
11663 311. https://gcc.gnu.org/PR11052
11664 312. https://gcc.gnu.org/PR11183
11665 313. https://gcc.gnu.org/PR11084
11666 314. https://gcc.gnu.org/PR10331
11667 315. https://gcc.gnu.org/PR10413
11668 316. https://gcc.gnu.org/PR11096
11669 317. https://gcc.gnu.org/PR2873
11670 318. https://gcc.gnu.org/PR3163
11671 319. https://gcc.gnu.org/PR5287
11672 320. https://gcc.gnu.org/PR10148
11673 321. https://gcc.gnu.org/PR8787
11674 322. https://gcc.gnu.org/PR10900
11675 323. https://gcc.gnu.org/PR1607
11676 324. https://gcc.gnu.org/PR4252
11677 325. https://gcc.gnu.org/PR4490
11678 326. https://gcc.gnu.org/PR10355
11679 327. https://gcc.gnu.org/PR10726
11680 328. https://gcc.gnu.org/PR10805
11681 329. https://gcc.gnu.org/PR10815
11682 330. https://gcc.gnu.org/PR10877
11683 331. https://gcc.gnu.org/PR11280
11684 332. https://gcc.gnu.org/PR11466
11685 333. https://gcc.gnu.org/PR10737
11686 334. https://gcc.gnu.org/PR10810
11687 335. https://gcc.gnu.org/PR8336
11688 336. https://gcc.gnu.org/PR9330
11689 337. https://gcc.gnu.org/PR9631
11690 338. https://gcc.gnu.org/PR9877
11691 339. https://gcc.gnu.org/PR11687
11692 340. https://gcc.gnu.org/PR12263
11693 341. https://gcc.gnu.org/PR12490
11694 342. https://gcc.gnu.org/PR7277
11695 343. https://gcc.gnu.org/PR7939
11696 344. https://gcc.gnu.org/PR11063
11697 345. https://gcc.gnu.org/PR11207
11698 346. https://gcc.gnu.org/PR11522
11699 347. https://gcc.gnu.org/PR11595
11700 348. https://gcc.gnu.org/PR11646
11701 349. https://gcc.gnu.org/PR11665
11702 350. https://gcc.gnu.org/PR11852
11703 351. https://gcc.gnu.org/PR11878
11704 352. https://gcc.gnu.org/PR11883
11705 353. https://gcc.gnu.org/PR11991
11706 354. https://gcc.gnu.org/PR12146
11707 355. https://gcc.gnu.org/PR12215
11708 356. https://gcc.gnu.org/PR12369
11709 357. https://gcc.gnu.org/PR12446
11710 358. https://gcc.gnu.org/PR12510
11711 359. https://gcc.gnu.org/PR12544
11712 360. https://gcc.gnu.org/PR9862
11713 361. https://gcc.gnu.org/PR10962
11714 362. https://gcc.gnu.org/PR11370
11715 363. https://gcc.gnu.org/PR11637
11716 364. https://gcc.gnu.org/PR11885
11717 365. https://gcc.gnu.org/PR12082
11718 366. https://gcc.gnu.org/PR12180
11719 367. https://gcc.gnu.org/PR12340
11720 368. https://gcc.gnu.org/PR3907
11721 369. https://gcc.gnu.org/PR5293
11722 370. https://gcc.gnu.org/PR5296
11723 371. https://gcc.gnu.org/PR7939
11724 372. https://gcc.gnu.org/PR8656
11725 373. https://gcc.gnu.org/PR10147
11726 374. https://gcc.gnu.org/PR11400
11727 375. https://gcc.gnu.org/PR11409
11728 376. https://gcc.gnu.org/PR11740
11729 377. https://gcc.gnu.org/PR11786
11730 378. https://gcc.gnu.org/PR11867
11731 379. https://gcc.gnu.org/PR11928
11732 380. https://gcc.gnu.org/PR12114
11733 381. https://gcc.gnu.org/PR12163
11734 382. https://gcc.gnu.org/PR12181
11735 383. https://gcc.gnu.org/PR12236
11736 384. https://gcc.gnu.org/PR12266
11737 385. https://gcc.gnu.org/PR12296
11738 386. https://gcc.gnu.org/PR12298
11739 387. https://gcc.gnu.org/PR12369
11740 388. https://gcc.gnu.org/PR12337
11741 389. https://gcc.gnu.org/PR12344
11742 390. https://gcc.gnu.org/PR12451
11743 391. https://gcc.gnu.org/PR12486
11744 392. https://gcc.gnu.org/PR8869
11745 393. https://gcc.gnu.org/PR9786
11746 394. https://gcc.gnu.org/PR11689
11747 395. https://gcc.gnu.org/PR12116
11748 396. https://gcc.gnu.org/PR12070
11749 397. https://gcc.gnu.org/PR11184
11750 398. https://gcc.gnu.org/PR11535
11751 399. https://gcc.gnu.org/PR11693
11752 400. https://gcc.gnu.org/PR12224
11753 401. https://gcc.gnu.org/PR11087
11754 402. https://gcc.gnu.org/PR11319
11755 403. https://gcc.gnu.org/PR11949
11756 404. https://gcc.gnu.org/PR11662
11757 405. https://gcc.gnu.org/PR11965
11758 406. https://gcc.gnu.org/PR12301
11759 407. https://gcc.gnu.org/PR11717
11760 408. https://gcc.gnu.org/PR11313
11761 409. https://gcc.gnu.org/PR11712
11762 410. https://gcc.gnu.org/PR12166
11763 411. https://gcc.gnu.org/PR12101
11764 412. https://gcc.gnu.org/PR10988
11765 413. https://gcc.gnu.org/PR11805
11766 414. https://gcc.gnu.org/PR11902
11767 415. https://gcc.gnu.org/PR11903
11768 416. https://gcc.gnu.org/PR11890
11769 417. https://gcc.gnu.org/PR12399
11770 418. https://gcc.gnu.org/PR13068
11771 419. https://gcc.gnu.org/PR10060
11772 420. https://gcc.gnu.org/PR10555
11773 421. https://gcc.gnu.org/PR10706
11774 422. https://gcc.gnu.org/PR11496
11775 423. https://gcc.gnu.org/PR11741
11776 424. https://gcc.gnu.org/PR12440
11777 425. https://gcc.gnu.org/PR12632
11778 426. https://gcc.gnu.org/PR12712
11779 427. https://gcc.gnu.org/PR12726
11780 428. https://gcc.gnu.org/PR12890
11781 429. https://gcc.gnu.org/PR12900
11782 430. https://gcc.gnu.org/PR13060
11783 431. https://gcc.gnu.org/PR13289
11784 432. https://gcc.gnu.org/PR13318
11785 433. https://gcc.gnu.org/PR13392
11786 434. https://gcc.gnu.org/PR13574
11787 435. https://gcc.gnu.org/PR13475
11788 436. https://gcc.gnu.org/PR13797
11789 437. https://gcc.gnu.org/PR13824
11790 438. https://gcc.gnu.org/PR8776
11791 439. https://gcc.gnu.org/PR10339
11792 440. https://gcc.gnu.org/PR11350
11793 441. https://gcc.gnu.org/PR12826
11794 442. https://gcc.gnu.org/PR12500
11795 443. https://gcc.gnu.org/PR12941
11796 444. https://gcc.gnu.org/PR12953
11797 445. https://gcc.gnu.org/PR13041
11798 446. https://gcc.gnu.org/PR13507
11799 447. https://gcc.gnu.org/PR13382
11800 448. https://gcc.gnu.org/PR13394
11801 449. https://gcc.gnu.org/PR13400
11802 450. https://gcc.gnu.org/PR13521
11803 451. https://gcc.gnu.org/PR2094
11804 452. https://gcc.gnu.org/PR2294
11805 453. https://gcc.gnu.org/PR5050
11806 454. https://gcc.gnu.org/PR9371
11807 455. https://gcc.gnu.org/PR9546
11808 456. https://gcc.gnu.org/PR10081
11809 457. https://gcc.gnu.org/PR10093
11810 458. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61
11811 459. https://gcc.gnu.org/PR10095
11812 460. https://gcc.gnu.org/PR11554
11813 461. https://gcc.gnu.org/PR12297
11814 462. https://gcc.gnu.org/PR12352
11815 463. https://gcc.gnu.org/PR12438
11816 464. https://gcc.gnu.org/PR12540
11817 465. https://gcc.gnu.org/PR12594
11818 466. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60
11819 467. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63
11820 468. https://gcc.gnu.org/PR12657
11821 469. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292
11822 470. https://gcc.gnu.org/PR12696
11823 471. https://gcc.gnu.org/PR12815
11824 472. https://gcc.gnu.org/PR12862
11825 473. https://gcc.gnu.org/PR12926
11826 474. https://gcc.gnu.org/PR12967
11827 475. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html
11828 476. https://gcc.gnu.org/PR12971
11829 477. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328
11830 478. https://gcc.gnu.org/PR13007
11831 479. https://gcc.gnu.org/PR13009
11832 480. https://gcc.gnu.org/PR13057
11833 481. https://gcc.gnu.org/PR13070
11834 482. https://gcc.gnu.org/PR13081
11835 483. https://gcc.gnu.org/PR13239
11836 484. https://gcc.gnu.org/PR13262
11837 485. https://gcc.gnu.org/PR13290
11838 486. https://gcc.gnu.org/PR13323
11839 487. https://gcc.gnu.org/PR13369
11840 488. https://gcc.gnu.org/PR13371
11841 489. https://gcc.gnu.org/PR13445
11842 490. https://gcc.gnu.org/PR13461
11843 491. https://gcc.gnu.org/PR13462
11844 492. https://gcc.gnu.org/PR13478
11845 493. https://gcc.gnu.org/PR13544
11846 494. https://gcc.gnu.org/PR13650
11847 495. https://gcc.gnu.org/PR13683
11848 496. https://gcc.gnu.org/PR13688
11849 497. https://gcc.gnu.org/PR13774
11850 498. https://gcc.gnu.org/PR13884
11851 499. https://gcc.gnu.org/PR10746
11852 500. https://gcc.gnu.org/PR11433
11853 501. https://gcc.gnu.org/PR12633
11854 502. https://gcc.gnu.org/PR13037
11855 503. https://gcc.gnu.org/PR13213
11856 504. https://gcc.gnu.org/PR4490
11857 505. https://gcc.gnu.org/PR12292
11858 506. https://gcc.gnu.org/PR12441
11859 507. https://gcc.gnu.org/PR12943
11860 508. https://gcc.gnu.org/PR13608
11861 509. https://gcc.gnu.org/PR11598
11862 510. https://gcc.gnu.org/PR11793
11863 511. https://gcc.gnu.org/PR12467
11864 512. https://gcc.gnu.org/PR12537
11865 513. https://gcc.gnu.org/PR12496
11866 514. https://gcc.gnu.org/PR12865
11867 515. https://gcc.gnu.org/PR13354
11868 516. https://gcc.gnu.org/PR10467
11869 517. https://gcc.gnu.org/PR11226
11870 518. https://gcc.gnu.org/PR11227
11871 519. https://gcc.gnu.org/PR12644
11872 520. https://gcc.gnu.org/PR13149
11873 521. https://gcc.gnu.org/PR12654
11874 522. https://gcc.gnu.org/PR12965
11875 523. https://gcc.gnu.org/PR13031
11876 524. https://gcc.gnu.org/PR11634
11877 525. https://gcc.gnu.org/PR12158
11878 526. https://gcc.gnu.org/PR11992
11879 527. https://gcc.gnu.org/PR9365
11880 528. https://gcc.gnu.org/PR10392
11881 529. https://gcc.gnu.org/PR11322
11882 530. https://gcc.gnu.org/PR13069
11883 531. https://gcc.gnu.org/PR13302
11884 532. https://gcc.gnu.org/PR13585
11885 533. https://gcc.gnu.org/PR8916
11886 534. https://gcc.gnu.org/PR11576
11887 535. https://gcc.gnu.org/PR13122
11888 536. https://gcc.gnu.org/PR13256
11889 537. https://gcc.gnu.org/PR13373
11890 538. https://gcc.gnu.org/PR12561
11891 539. https://gcc.gnu.org/PR6243
11892 540. https://gcc.gnu.org/PR11397
11893 541. https://gcc.gnu.org/PR12505
11894 542. https://gcc.gnu.org/PR13150
11895 543. https://gcc.gnu.org/PR12666
11896 544. https://gcc.gnu.org/PR12969
11897 545. https://gcc.gnu.org/PR10819
11898 546. https://gcc.gnu.org/PR11612
11899 547. https://gcc.gnu.org/PR13211
11900 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4
11901 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5
11902 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6
11903 551. https://gcc.gnu.org/onlinedocs/
11904 552. mailto:gcc-help@gcc.gnu.org
11905 553. mailto:gcc@gcc.gnu.org
11906 554. https://gcc.gnu.org/lists.html
11907 555. http://www.fsf.org/
11908 556. https://gcc.gnu.org/about.html
11909 557. http://validator.w3.org/check/referer
11910 ======================================================================
11911 http://gcc.gnu.org/gcc-3.2/index.html
11912 GCC 3.2 Release Series
11916 The [1]GNU project and the GCC developers are pleased to announce the
11917 release of GCC 3.2.3.
11919 The purpose of the GCC 3.2 release series is to provide a stable
11920 platform for OS distributors to use building their next releases. A
11921 primary objective was to stabilize the C++ ABI; we believe that the
11922 interface to the compiler and the C++ standard library are now
11925 Be aware that C++ code compiled by GCC 3.2.x will (in general) not
11926 interoperate with code compiled by GCC 3.1.1 or earlier.
11928 Please refer to our [2]detailed list of news, caveats, and bug-fixes
11929 for further information.
11934 April 25, 2003 ([3]changes)
11937 February 5, 2003 ([4]changes)
11940 November 19, 2002 ([5]changes)
11943 August 14, 2002 ([6]changes)
11945 References and Acknowledgements
11947 GCC used to stand for the GNU C Compiler, but since the compiler
11948 supports several other languages aside from C, it now stands for the
11949 GNU Compiler Collection.
11951 A list of [7]successful builds is updated as new information becomes
11954 The GCC developers would like to thank the numerous people that have
11955 contributed new features, improvements, bug fixes, and other changes as
11956 well as test results to GCC. This [8]amazing group of volunteers is
11957 what makes GCC successful.
11959 For additional information about GCC please refer to the [9]GCC project
11960 web site or contact the [10]GCC development mailing list.
11962 To obtain GCC please use [11]our mirror sites, or our CVS server.
11965 For questions related to the use of GCC, please consult these web
11966 pages and the [12]GCC manuals. If that fails, the
11967 [13]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11968 web pages and the development of GCC are welcome on our developer
11969 list at [14]gcc@gcc.gnu.org. All of [15]our lists have public
11972 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and
11973 distribution of this entire article is permitted in any medium,
11974 provided this notice is preserved.
11976 These pages are [17]maintained by the GCC team. Last modified
11981 1. http://www.gnu.org/
11982 2. http://gcc.gnu.org/gcc-3.2/changes.html
11983 3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
11984 4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2
11985 5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1
11986 6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2
11987 7. http://gcc.gnu.org/gcc-3.2/buildstat.html
11988 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
11989 9. http://gcc.gnu.org/index.html
11990 10. mailto:gcc@gcc.gnu.org
11991 11. http://gcc.gnu.org/mirrors.html
11992 12. https://gcc.gnu.org/onlinedocs/
11993 13. mailto:gcc-help@gcc.gnu.org
11994 14. mailto:gcc@gcc.gnu.org
11995 15. https://gcc.gnu.org/lists.html
11996 16. http://www.fsf.org/
11997 17. https://gcc.gnu.org/about.html
11998 18. http://validator.w3.org/check/referer
11999 ======================================================================
12000 http://gcc.gnu.org/gcc-3.2/changes.html
12001 GCC 3.2 Release Series
12002 Changes, New Features, and Fixes
12004 The latest release in the 3.2 release series is [1]GCC 3.2.3.
12006 Caveats and New Features
12010 * The C++ compiler does not correctly zero-initialize
12011 pointers-to-data members. You must explicitly initialize them. For
12012 example: int S::*m(0); will work, but depending on
12013 default-initialization to zero will not work. This bug cannot be
12014 fixed in GCC 3.2 without inducing unacceptable risks. It will be
12016 * This GCC release is based on the GCC 3.1 sourcebase, and thus has
12017 all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has
12018 a number of C++ ABI fixes which make its C++ compiler generate
12019 binary code which is incompatible with the C++ compilers found in
12020 earlier GCC releases, including GCC 3.1 and GCC 3.1.1.
12022 Frontend Enhancements
12026 * The method of constructing the list of directories to be searched
12027 for header files has been revised. If a directory named by a -I
12028 option is a standard system include directory, the option is
12029 ignored to ensure that the default search order for system
12030 directories and the special treatment of system header files are
12032 * The C and Objective-C compilers no longer accept the "Naming Types"
12033 extension (typedef foo = bar); it was already unavailable in C++.
12034 Code which uses it will need to be changed to use the "typeof"
12035 extension instead: typedef typeof(bar) foo. (We have removed this
12036 extension without a period of deprecation because it has caused the
12037 compiler to crash since version 3.0 and no one noticed until very
12038 recently. Thus we conclude it is not in widespread use.)
12042 * GCC 3.2 fixed serveral differences between the C++ ABI implemented
12043 in GCC and the multi-vendor standard, but more have been found
12044 since the release. 3.2.1 adds a new warning, -Wabi, to warn about
12045 code which is affected by these bugs. We will fix these bugs in
12046 some future release, once we are confident that all have been
12047 found; until then, it is our intention to make changes to the ABI
12048 only if they are necessary for correct compilation of C++, as
12049 opposed to conformance to the ABI documents.
12050 * For details on how to build an ABI compliant compiler for GNU/Linux
12051 systems, check the [3]common C++ ABI page.
12053 New Targets and Target Specific Improvements
12057 * Fixed a number of bugs in SSE and MMX intrinsics.
12058 * Fixed common compiler crashes with SSE instruction set enabled
12059 (implied by -march=pentium3, pentium4, athlon-xp)
12060 * __m128 and __m128i is not 128bit aligned when used in structures.
12064 * A bug whereby the compiler could generate bad code for bzero has
12066 * ABI fixes (implying ABI incompatibilities with previous version in
12068 * Fixed prefetch code generation
12069 __________________________________________________________________
12073 3.2.3 is a bug fix release only; there are no new features that were
12074 not present in GCC 3.2.2.
12078 This section lists the problem reports (PRs) from GCC's bug tracking
12079 system that are known to be fixed in the 3.2.3 release. This list might
12080 not be complete (that is, it is possible that some PRs that have been
12081 fixed are not listed here), and some of the titles have been changed to
12082 make them more clear.
12084 Internal Compiler Errors (multi-platform)
12086 * [4]3782: (c++) -quiet -fstats produces a segmentation fault in
12088 * [5]6440: (c++) template specializations cause ICE
12089 * [6]7050: (c++) ICE on: (i ? get_string() : throw)
12090 * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c)
12091 * [8]7982: (c++) ICE due to infinite recursion (using STL set)
12092 * [9]8068: exceedingly high (infinite) memory usage
12093 * [10]8178: ICE with __builtin_ffs
12094 * [11]8396: ICE in copy_to_mode_reg, in explow.c
12095 * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c
12096 * [13]9768: ICE when optimizing inline code at -O2
12097 * [14]9798: (c++) Infinite recursion (segfault) in
12098 cp/decl.c:push_using_directive with recursive using directives
12099 * [15]9799: mismatching structure initializer with nested flexible
12101 * [16]9928: ICE on duplicate enum declaration
12102 * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects
12104 * [18]10352: ICE in find_reloads_toplev
12105 * [19]10336: ICE with -Wunreachable-code
12109 * [20]8224: Incorrect joining of signed and unsigned division
12110 * [21]8613: -O2 produces wrong code with builtin strlen and
12112 * [22]8828: gcc reports some code is unreachable when it is not
12113 * [23]9226: GCSE breaking argument passing
12114 * [24]9853: miscompilation of non-constant structure initializer
12115 * [25]9797: C99-style struct initializers are miscompiled
12116 * [26]9967: Some standard C function calls should not be replaced
12117 when optimizing for size
12118 * [27]10116: ce2: invalid merge of join_bb in the context of switch
12120 * [28]10171: wrong code for inlined function
12121 * [29]10175: -Wunreachable-code doesn't work for single lines
12123 C++ compiler and library:
12125 * [30]8316: Confusing diagnostic for code that misuses conversion
12127 * [31]9169: filebuf output fails if codecvt<>::out returns noconv
12128 * [32]9420: incomplete type incorrectly reported
12129 * [33]9459: typeof in return type specification of template not
12131 * [34]9507: filebuf::open handles ios_base::ate incorrectly
12132 * [35]9538: Out-of-bounds memory access in streambuf::sputbackc
12133 * [36]9602: Total confusion about template/friend/virtual/abstract
12134 * [37]9993: destructor not called for local object created within and
12135 returned from infinite loop
12136 * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2
12139 Java compiler and library:
12141 * [39]9652: libgcj build fails on irix6.5.1[78]
12142 * [40]10144: gas on solaris complains about bad .stabs lines for
12143 java, native as unaffected
12145 x86-specific (Intel/AMD):
12147 * [41]8746: gcc miscompiles Linux kernel ppa driver on x86
12148 * [42]9888: -mcpu=k6 -Os produces out of range loop instructions
12149 * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu
12151 * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib)
12156 * [45]7784: [Sparc] ICE in extract_insn, in recog.c
12157 * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in
12159 * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc
12160 * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in
12162 * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc
12163 * [50]9414: Scheduling bug on Ultrasparc
12164 * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64
12168 * [52]7248: broken "inclusive or" code
12169 * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1
12173 * [54]9732: Wrong code with -O2 -fPIC
12174 * [55]10073: ICE: powerpc cannot split insn
12178 * [56]7702: optimization problem on a DEC alpha under OSF1
12179 * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system
12183 * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275)
12184 * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10
12186 * [60]10271: Floating point args don't get reloaded across function
12191 * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in
12196 * [62]10377: gcc-3.2.2 creates bad assembler code for cris
12198 Miscellaneous and minor bugs:
12200 * [63]6955: collect2 says "core dumped" when there is no core
12201 __________________________________________________________________
12205 Beginning with 3.2.2, GCC's Makefile suite supports redirection of make
12206 install by means of the DESTDIR variable. Parts of the GCC tree have
12207 featured that support long before, but now it is available even from
12210 Other than that, GCC 3.2.2 is a bug fix release only; there are no new
12211 features that were not present in GCC 3.2.1.
12215 On the following i386-based systems GCC 3.2.1 broke the C ABI wrt.
12216 functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped
12217 with FreeBSD 5.0 does not have this problem), Interix, a.out-based
12218 GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI
12219 change, and thus restores ABI-compatibility with previous releases
12220 (except GCC 3.2.1) on these platforms.
12222 This section lists the problem reports (PRs) from GCC's bug tracking
12223 system that are known to be fixed in the 3.2.2 release. This list might
12224 not be complete (that is, it is possible that some PRs that have been
12225 fixed are not listed here) and some of the titles have been changed to
12226 make them more clear.
12228 Internal Compiler Errors (multi-platform)
12230 * [64]5919: (c++) ICE when passing variable array to template
12232 * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=)
12233 * [66]7507: ICE with -O2 when address of called function is a
12234 complicated expression
12235 * [67]7622: ICE with nested inline functions if function's address is
12237 * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR
12239 * [70]8031: (c++) ICE in code comparing typeids and casting from
12241 * [71]8275: ICE in simplify_subreg
12242 * [72]8332: (c++) builtin strlen/template interaction causes ICE
12243 * [73]8372: (c++) ICE on explicit call of destructor
12244 * [74]8439: (c, not c++) empty struct causes ICE
12245 * [75]8442: (c++) ICE with nested template classes
12246 * [76]8518: ICE when compiling mplayer ("extern inline" issue)
12247 * [77]8615: (c++) ICE with out-of-range character constant template
12249 * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307
12250 * [79]8799: (c++) ICE: error reporting routines re-entered
12251 * [80]9328: (c++) ICE with typeof(X) for overloaded X
12252 * [81]9465: (preprocessor) cpp -traditional ICE on null bytes
12254 C++ (compiler and library) bugs
12256 * [82]47: scoping in nested classes is broken
12257 * [83]6745: problems with iostream rdbuf() member function
12258 * [84]8214: conversion from const char* const to char* sometimes
12260 * [85]8493: builtin strlen and overload resolution (same bug as
12262 * [87]8503: strange behaviour of function types
12263 * [88]8727: compiler confused by inheritance from an anonymous struct
12264 * [89]7445: poor performance of std::locale::classic() in
12265 multi-threaded applications
12266 * [90]8230: mishandling of overflow in vector<T>::resize
12267 * [91]8399: sync_with_stdio(false) breaks unformatted input
12268 * [92]8662: illegal access of private member of unnamed class is
12270 * [93]8707: "make distclean" fails in libstdc++-v3 directory
12271 * [94]8708: __USE_MALLOC doesn't work
12272 * [95]8790: Use of non-thread-safe strtok in src/localename.cc
12273 * [96]8887: Bug in date formats with --enable-clocale=generic
12274 * [97]9076: Call Frame Instructions are not handled correctly during
12276 * [98]9151: std::setprecision limited to 16 digits when outputting a
12278 * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers
12279 * [100]9269: libstdc++ headers: explicit specialization of function
12280 must precede its first use
12281 * [101]9322: return value of basic_streambuf<>::getloc affected by
12283 * [102]9433: segfault in runtime support for dynamic_cast
12285 C and optimizer bugs
12287 * [103]8032: GCC incorrectly initializes static structs that have
12289 * [104]8639: simple arithmetic expression broken
12290 * [105]8794: optimization improperly eliminates certain expressions
12291 * [106]8832: traditional "asm volatile" code is illegally optimized
12292 * [107]8988: loop optimizer bug: with -O2, code is generated that
12293 segfaults (found on i386, bug present for all platforms)
12294 * [108]9492: structure copy clobbers subsequent stores to structure
12298 * [109]9267: Objective-C parser won't build with newer bison versions
12303 * [110]8344: Ada build problem due to conflict between gcc/final.o,
12308 * [111]8524: _Pragma within macros is improperly expanded
12309 * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with
12314 * [113]9090: arm ICE with >= -O2; regression from gcc-2.95
12316 x86-specific (Intel/AMD)
12318 * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction)
12319 * [115]8599: loop unroll bug with -march=k6-3
12320 * [116]9506: ABI breakage in structure return (affects BSD and
12321 Cygwin, but not GNU/Linux)
12323 FreeBSD 5.0 specific
12325 * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0
12329 * [118]9292: hppa1.1-rtems configurery problems
12330 * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug
12331 * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue
12332 * [121]9296: gthr-rtems regression
12333 * [122]9316: powerpc-rtems: extending multilibs
12337 * [123]9493: ICE with -O2 when building a simple function
12341 * [124]7341: hyperlink to gcov in GCC documentation doesn't work
12342 * [125]8947: Please add a warning about "-malign-double" in docs
12343 * [126]7448, [127]8882: typo cleanups
12344 __________________________________________________________________
12348 3.2.1 adds a new warning, -Wabi. This option warns when GNU C++
12349 generates code that is known not to be binary-compatible with the
12350 vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included
12351 in the distribution, for details.
12353 This release also removes an old GCC extension, "naming types", and the
12354 documentation now directs users to use a different GCC extension,
12355 __typeof__, instead. The feature had evidently been broken for a while.
12357 Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and
12358 the new warning there are no new features that were not present in GCC
12361 In addition, the previous fix for [128]PR 7445 (poor performance of
12362 std::locale::classic() in multi-threaded applications) was reverted
12363 ("unfixed"), because the "fix" was not thread-safe.
12367 This section lists the problem reports (PRs) from GCC's bug tracking
12368 system that are known to be fixed in the 3.2.1 release. This list might
12369 not be complete (that is, it is possible that some PRs that have been
12370 fixed are not listed here). As you can see, the number of bug fixes is
12371 quite large, so it is strongly recommended that users of earlier GCC
12372 3.x releases upgrade to GCC 3.2.1.
12374 Internal Compiler Errors (multi-platform)
12376 * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c
12377 * [130]5661: (c++) ICE instantiating template on array of unknown
12379 * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on
12381 * [132]6994: (c++) ICE in find_function_data
12382 * [133]7150: preprocessor: GCC -dM -E gives an ICE
12383 * [134]7160: ICE when optimizing branches without a return value
12384 * [135]7228: (c++) ICE when using member template and template
12386 * [136]7266: (c++) ICE with -pedantic on missing typename
12387 * [137]7353: ICE from use of "Naming Types" extension, see above
12388 * [138]7411: ICE in instantiate_virtual_regs_1, in function.c
12389 * [139]7478: (c++) ICE on static_cast inside template
12390 * [140]7526: preprocessor core dump when _Pragma implies #pragma
12392 * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803
12394 * [143]7754: (c++) ICE on union with template parameter
12395 * [144]7788: (c++) redeclaring a definition as an incomplete class
12397 * [145]8031: (c++) ICE in comptypes, in cp/typeck.c
12398 * [146]8055: preprocessor dies with SIG11 when building FreeBSD
12400 * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related
12402 * [148]8134: (c++) ICE in force_store_init_value on legal code
12403 * [149]8149: (c++) ICE on incomplete type
12404 * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array
12407 C++ (compiler and library) bugs
12409 * [151]5607: No pointer adjustment in covariant return types
12410 * [152]6579: Infinite loop with statement expressions in member
12412 * [153]6803: Default copy constructor bug in GCC 3.1
12413 * [154]7176: g++ confused by friend and static member with same name
12414 * [155]7188: Segfault with template class and recursive (incorrect)
12416 * [156]7306: Regression: GCC 3.x fails to compile code with virtual
12417 inheritance if a method has a variable number of arguments
12418 * [157]7461: ctype<char>::classic_table() returns offset array on
12420 * [158]7524: f(const float arg[3]) fails
12421 * [159]7584: Erroneous ambiguous base error on using declaration
12422 * [160]7676: Member template overloading problem
12423 * [161]7679: infinite loop when a right parenthesis is missing
12424 * [162]7811: default locale not taken from environment
12425 * [163]7961: compare( char *) implemented incorrectly in
12427 * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if
12428 streambuf::underflow() leaves gptr() NULL (dups: [165]8127,
12430 * [167]8096: deque::at() throws std::range_error instead of
12432 * [168]8127: cout << cin.rdbuf() infinite loop
12433 * [169]8218: Excessively large memory consumed for classes with large
12435 * [170]8287: GCC 3.2: Destructor called for non-constructed local
12437 * [171]8347: empty vector range used in string construction causes
12439 * [172]8348: fail() flag is set in istringstream when eof() flag is
12441 * [173]8391: regression: infinite loop in cp/decl2.c(finish_file)
12443 C and optimizer bugs
12445 * [174]6627: -fno-align-functions doesn't seem to disable function
12447 * [175]6631: life_analysis misoptimizes code to initialize fields of
12449 * [176]7102: unsigned char division results in floating exception
12450 * [177]7120: Run once loop should *always* be unrolled
12452 * [178]7209: Bug involving array referencing and ?: operator
12453 * [179]7515: invalid inlining of global function with -O3
12454 * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test
12455 * [181]8467: bug in sibling call optimization
12459 * [182]4890: incorrect line markers from the traditional preprocessor
12460 * [183]7357: -M option omits system headers files (making it the same
12462 * [184]7358: Changes to Sun's make Dependencies
12463 * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as
12465 * [186]7862: Interrupting GCC -MD removes .d file but not .o
12466 * [187]8190: Failed compilation deletes -MD dependency file
12467 * [188]8524: _Pragma within macro is improperly expanded
12469 x86 specific (Intel/AMD)
12471 * [189]5351: (i686-only) function pass-by-value structure copy
12472 corrupts stack ([190]7591 is a duplicate)
12473 * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with
12474 -march=pentium3/pentium2/athlon (these are all the same underlying
12475 bug, in MMX register use)
12476 * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same
12478 * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken
12479 * [199]6981: wrong code in 64-bit manipulation on x86
12480 * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__
12482 * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE
12483 intrinsics are broken
12484 * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with
12486 * [203]7693: Typo in i386 mmintrin.h header
12487 * [204]7723: ICE - Pentium3 sse - GCC 3.2
12488 * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse
12489 * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3
12493 * [207]5967: GCC bug when profiling nested functions on powerpc
12494 * [208]6984: wrong code generated with -O2, -O3, -Os for do-while
12496 * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5
12497 * [210]7130: miscompiled code for GCC-3.1 on
12498 powerpc-unknown-linux-gnu with -funroll-all-loops
12499 * [211]7133: PowerPC ICE: unrecognizable insn
12500 * [212]7380: ICE in extract_insn, at recog.c:2148
12501 * [213]8252: ICE on Altivec code with optimization turned on
12502 * [214]8451: Altivec ICE in GCC 3.2
12506 * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa
12510 * [216]6668: when using --disable-multilib, libgcc_s.so is installed
12511 in the wrong place on sparc-solaris
12512 * [217]7151: ICE when compiling for UltraSPARC
12513 * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long
12515 * [219]7842: [REGRESSION] SPARC code gen bug
12519 * [220]7856: [arm] invalid offset in constant pool reference
12520 * [221]7967: optimization produces wrong code (ARM)
12524 * [222]7374: __builtin_fabsl broken on alpha
12528 * [223]7370: ICE in fixup_var_refs_1 on s390x
12529 * [224]7409: loop optimization bug on s390x-linux-gnu
12530 * [225]8232: s390x: ICE when using bcmp with int length argument
12534 * [226]7623: SCO OpenServer build fails with machmode.def: undefined
12535 symbol: BITS_PER_UNIT
12537 m68k/Coldfire specific
12539 * [227]8314: crtbegin, crtend need to be multilib'ed for this
12544 * [228]761: Document some undocumented options
12545 * [229]5610: Fix documentation about invoking SSE instructions
12547 * [230]7484: List -Wmissing-declarations as C-only option
12548 * [231]7531: -mcmodel not documented for x86-64
12549 * [232]8120: Update documentation of bad use of ##
12550 __________________________________________________________________
12554 3.2 is a small bug fix release, but there is a change to the
12555 application binary interface (ABI), hence the change to the second part
12556 of the version number.
12558 The main purpose of the 3.2 release is to correct a couple of problems
12559 in the C++ ABI, with the intention of providing a stable interface
12560 going forward. Accordingly, 3.2 is only a small change to 3.1.1.
12566 * [233]7320: g++ 3.2 relocation problem
12567 * [234]7470: vtable: virtual function pointers not in declaration
12572 * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t
12573 * [236]6503, [237]6642, [238]7186: Problems with comparing or
12574 subtracting various types of const and non-const iterators
12575 * [239]7216: ambiguity with basic_iostream::traits_type
12576 * [240]7220: problem with basic_istream::ignore(0,delimiter)
12577 * [241]7222: locale::operator==() doesn't work on std::locale("")
12578 * [242]7286: placement operator delete issue
12579 * [243]7442: cxxabi.h does not match the C++ ABI
12580 * [244]7445: poor performance of std::locale::classic() in
12581 multi-threaded applications
12585 * [245]7291: off-by-one in generated inline bzero code for x86-64
12588 For questions related to the use of GCC, please consult these web
12589 pages and the [246]GCC manuals. If that fails, the
12590 [247]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12591 web pages and the development of GCC are welcome on our developer
12592 list at [248]gcc@gcc.gnu.org. All of [249]our lists have public
12595 Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and
12596 distribution of this entire article is permitted in any medium,
12597 provided this notice is preserved.
12599 These pages are [251]maintained by the GCC team. Last modified
12604 1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
12605 2. http://gcc.gnu.org/gcc-3.1/changes.html
12606 3. http://gcc.gnu.org/gcc-3.2/c++-abi.html
12607 4. https://gcc.gnu.org/PR3782
12608 5. https://gcc.gnu.org/PR6440
12609 6. https://gcc.gnu.org/PR7050
12610 7. https://gcc.gnu.org/PR7741
12611 8. https://gcc.gnu.org/PR7982
12612 9. https://gcc.gnu.org/PR8068
12613 10. https://gcc.gnu.org/PR8178
12614 11. https://gcc.gnu.org/PR8396
12615 12. https://gcc.gnu.org/PR8674
12616 13. https://gcc.gnu.org/PR9768
12617 14. https://gcc.gnu.org/PR9798
12618 15. https://gcc.gnu.org/PR9799
12619 16. https://gcc.gnu.org/PR9928
12620 17. https://gcc.gnu.org/PR10114
12621 18. https://gcc.gnu.org/PR10352
12622 19. https://gcc.gnu.org/PR10336
12623 20. https://gcc.gnu.org/PR8224
12624 21. https://gcc.gnu.org/PR8613
12625 22. https://gcc.gnu.org/PR8828
12626 23. https://gcc.gnu.org/PR9226
12627 24. https://gcc.gnu.org/PR9853
12628 25. https://gcc.gnu.org/PR9797
12629 26. https://gcc.gnu.org/PR9967
12630 27. https://gcc.gnu.org/PR10116
12631 28. https://gcc.gnu.org/PR10171
12632 29. https://gcc.gnu.org/PR10175
12633 30. https://gcc.gnu.org/PR8316
12634 31. https://gcc.gnu.org/PR9169
12635 32. https://gcc.gnu.org/PR9420
12636 33. https://gcc.gnu.org/PR9459
12637 34. https://gcc.gnu.org/PR9507
12638 35. https://gcc.gnu.org/PR9538
12639 36. https://gcc.gnu.org/PR9602
12640 37. https://gcc.gnu.org/PR9993
12641 38. https://gcc.gnu.org/PR10167
12642 39. https://gcc.gnu.org/PR9652
12643 40. https://gcc.gnu.org/PR10144
12644 41. https://gcc.gnu.org/PR8746
12645 42. https://gcc.gnu.org/PR9888
12646 43. https://gcc.gnu.org/PR9638
12647 44. https://gcc.gnu.org/PR9954
12648 45. https://gcc.gnu.org/PR7784
12649 46. https://gcc.gnu.org/PR7796
12650 47. https://gcc.gnu.org/PR8281
12651 48. https://gcc.gnu.org/PR8366
12652 49. https://gcc.gnu.org/PR8726
12653 50. https://gcc.gnu.org/PR9414
12654 51. https://gcc.gnu.org/PR10067
12655 52. https://gcc.gnu.org/PR7248
12656 53. https://gcc.gnu.org/PR8343
12657 54. https://gcc.gnu.org/PR9732
12658 55. https://gcc.gnu.org/PR10073
12659 56. https://gcc.gnu.org/PR7702
12660 57. https://gcc.gnu.org/PR9671
12661 58. https://gcc.gnu.org/PR8694
12662 59. https://gcc.gnu.org/PR9953
12663 60. https://gcc.gnu.org/PR10271
12664 61. https://gcc.gnu.org/PR6362
12665 62. https://gcc.gnu.org/PR10377
12666 63. https://gcc.gnu.org/PR6955
12667 64. https://gcc.gnu.org/PR5919
12668 65. https://gcc.gnu.org/PR7129
12669 66. https://gcc.gnu.org/PR7507
12670 67. https://gcc.gnu.org/PR7622
12671 68. https://gcc.gnu.org/PR7681
12672 69. https://gcc.gnu.org/PR9528
12673 70. https://gcc.gnu.org/PR8031
12674 71. https://gcc.gnu.org/PR8275
12675 72. https://gcc.gnu.org/PR8332
12676 73. https://gcc.gnu.org/PR8372
12677 74. https://gcc.gnu.org/PR8439
12678 75. https://gcc.gnu.org/PR8442
12679 76. https://gcc.gnu.org/PR8518
12680 77. https://gcc.gnu.org/PR8615
12681 78. https://gcc.gnu.org/PR8663
12682 79. https://gcc.gnu.org/PR8799
12683 80. https://gcc.gnu.org/PR9328
12684 81. https://gcc.gnu.org/PR9465
12685 82. https://gcc.gnu.org/PR47
12686 83. https://gcc.gnu.org/PR6745
12687 84. https://gcc.gnu.org/PR8214
12688 85. https://gcc.gnu.org/PR8493
12689 86. https://gcc.gnu.org/PR8332
12690 87. https://gcc.gnu.org/PR8503
12691 88. https://gcc.gnu.org/PR8727
12692 89. https://gcc.gnu.org/PR7445
12693 90. https://gcc.gnu.org/PR8230
12694 91. https://gcc.gnu.org/PR8399
12695 92. https://gcc.gnu.org/PR8662
12696 93. https://gcc.gnu.org/PR8707
12697 94. https://gcc.gnu.org/PR8708
12698 95. https://gcc.gnu.org/PR8790
12699 96. https://gcc.gnu.org/PR8887
12700 97. https://gcc.gnu.org/PR9076
12701 98. https://gcc.gnu.org/PR9151
12702 99. https://gcc.gnu.org/PR9168
12703 100. https://gcc.gnu.org/PR9269
12704 101. https://gcc.gnu.org/PR9322
12705 102. https://gcc.gnu.org/PR9433
12706 103. https://gcc.gnu.org/PR8032
12707 104. https://gcc.gnu.org/PR8639
12708 105. https://gcc.gnu.org/PR8794
12709 106. https://gcc.gnu.org/PR8832
12710 107. https://gcc.gnu.org/PR8988
12711 108. https://gcc.gnu.org/PR9492
12712 109. https://gcc.gnu.org/PR9267
12713 110. https://gcc.gnu.org/PR8344
12714 111. https://gcc.gnu.org/PR8524
12715 112. https://gcc.gnu.org/PR8880
12716 113. https://gcc.gnu.org/PR9090
12717 114. https://gcc.gnu.org/PR8588
12718 115. https://gcc.gnu.org/PR8599
12719 116. https://gcc.gnu.org/PR9506
12720 117. https://gcc.gnu.org/PR9484
12721 118. https://gcc.gnu.org/PR9292
12722 119. https://gcc.gnu.org/PR9293
12723 120. https://gcc.gnu.org/PR9295
12724 121. https://gcc.gnu.org/PR9296
12725 122. https://gcc.gnu.org/PR9316
12726 123. https://gcc.gnu.org/PR9493
12727 124. https://gcc.gnu.org/PR7341
12728 125. https://gcc.gnu.org/PR8947
12729 126. https://gcc.gnu.org/PR7448
12730 127. https://gcc.gnu.org/PR8882
12731 128. https://gcc.gnu.org/PR7445
12732 129. https://gcc.gnu.org/PR2521
12733 130. https://gcc.gnu.org/PR5661
12734 131. https://gcc.gnu.org/PR6419
12735 132. https://gcc.gnu.org/PR6994
12736 133. https://gcc.gnu.org/PR7150
12737 134. https://gcc.gnu.org/PR7160
12738 135. https://gcc.gnu.org/PR7228
12739 136. https://gcc.gnu.org/PR7266
12740 137. https://gcc.gnu.org/PR7353
12741 138. https://gcc.gnu.org/PR7411
12742 139. https://gcc.gnu.org/PR7478
12743 140. https://gcc.gnu.org/PR7526
12744 141. https://gcc.gnu.org/PR7721
12745 142. https://gcc.gnu.org/PR7803
12746 143. https://gcc.gnu.org/PR7754
12747 144. https://gcc.gnu.org/PR7788
12748 145. https://gcc.gnu.org/PR8031
12749 146. https://gcc.gnu.org/PR8055
12750 147. https://gcc.gnu.org/PR8067
12751 148. https://gcc.gnu.org/PR8134
12752 149. https://gcc.gnu.org/PR8149
12753 150. https://gcc.gnu.org/PR8160
12754 151. https://gcc.gnu.org/PR5607
12755 152. https://gcc.gnu.org/PR6579
12756 153. https://gcc.gnu.org/PR6803
12757 154. https://gcc.gnu.org/PR7176
12758 155. https://gcc.gnu.org/PR7188
12759 156. https://gcc.gnu.org/PR7306
12760 157. https://gcc.gnu.org/PR7461
12761 158. https://gcc.gnu.org/PR7524
12762 159. https://gcc.gnu.org/PR7584
12763 160. https://gcc.gnu.org/PR7676
12764 161. https://gcc.gnu.org/PR7679
12765 162. https://gcc.gnu.org/PR7811
12766 163. https://gcc.gnu.org/PR7961
12767 164. https://gcc.gnu.org/PR8071
12768 165. https://gcc.gnu.org/PR8127
12769 166. https://gcc.gnu.org/PR6745
12770 167. https://gcc.gnu.org/PR8096
12771 168. https://gcc.gnu.org/PR8127
12772 169. https://gcc.gnu.org/PR8218
12773 170. https://gcc.gnu.org/PR8287
12774 171. https://gcc.gnu.org/PR8347
12775 172. https://gcc.gnu.org/PR8348
12776 173. https://gcc.gnu.org/PR8391
12777 174. https://gcc.gnu.org/PR6627
12778 175. https://gcc.gnu.org/PR6631
12779 176. https://gcc.gnu.org/PR7102
12780 177. https://gcc.gnu.org/PR7120
12781 178. https://gcc.gnu.org/PR7209
12782 179. https://gcc.gnu.org/PR7515
12783 180. https://gcc.gnu.org/PR7814
12784 181. https://gcc.gnu.org/PR8467
12785 182. https://gcc.gnu.org/PR4890
12786 183. https://gcc.gnu.org/PR7357
12787 184. https://gcc.gnu.org/PR7358
12788 185. https://gcc.gnu.org/PR7602
12789 186. https://gcc.gnu.org/PR7862
12790 187. https://gcc.gnu.org/PR8190
12791 188. https://gcc.gnu.org/PR8524
12792 189. https://gcc.gnu.org/PR5351
12793 190. https://gcc.gnu.org/PR7591
12794 191. https://gcc.gnu.org/PR6845
12795 192. https://gcc.gnu.org/PR7034
12796 193. https://gcc.gnu.org/PR7124
12797 194. https://gcc.gnu.org/PR7174
12798 195. https://gcc.gnu.org/PR7134
12799 196. https://gcc.gnu.org/PR7375
12800 197. https://gcc.gnu.org/PR7390
12801 198. https://gcc.gnu.org/PR6890
12802 199. https://gcc.gnu.org/PR6981
12803 200. https://gcc.gnu.org/PR7242
12804 201. https://gcc.gnu.org/PR7396
12805 202. https://gcc.gnu.org/PR7630
12806 203. https://gcc.gnu.org/PR7693
12807 204. https://gcc.gnu.org/PR7723
12808 205. https://gcc.gnu.org/PR7951
12809 206. https://gcc.gnu.org/PR8146
12810 207. https://gcc.gnu.org/PR5967
12811 208. https://gcc.gnu.org/PR6984
12812 209. https://gcc.gnu.org/PR7114
12813 210. https://gcc.gnu.org/PR7130
12814 211. https://gcc.gnu.org/PR7133
12815 212. https://gcc.gnu.org/PR7380
12816 213. https://gcc.gnu.org/PR8252
12817 214. https://gcc.gnu.org/PR8451
12818 215. https://gcc.gnu.org/PR7250
12819 216. https://gcc.gnu.org/PR6668
12820 217. https://gcc.gnu.org/PR7151
12821 218. https://gcc.gnu.org/PR7335
12822 219. https://gcc.gnu.org/PR7842
12823 220. https://gcc.gnu.org/PR7856
12824 221. https://gcc.gnu.org/PR7967
12825 222. https://gcc.gnu.org/PR7374
12826 223. https://gcc.gnu.org/PR7370
12827 224. https://gcc.gnu.org/PR7409
12828 225. https://gcc.gnu.org/PR8232
12829 226. https://gcc.gnu.org/PR7623
12830 227. https://gcc.gnu.org/PR8314
12831 228. https://gcc.gnu.org/PR761
12832 229. https://gcc.gnu.org/PR5610
12833 230. https://gcc.gnu.org/PR7484
12834 231. https://gcc.gnu.org/PR7531
12835 232. https://gcc.gnu.org/PR8120
12836 233. https://gcc.gnu.org/PR7320
12837 234. https://gcc.gnu.org/PR7470
12838 235. https://gcc.gnu.org/PR6410
12839 236. https://gcc.gnu.org/PR6503
12840 237. https://gcc.gnu.org/PR6642
12841 238. https://gcc.gnu.org/PR7186
12842 239. https://gcc.gnu.org/PR7216
12843 240. https://gcc.gnu.org/PR7220
12844 241. https://gcc.gnu.org/PR7222
12845 242. https://gcc.gnu.org/PR7286
12846 243. https://gcc.gnu.org/PR7442
12847 244. https://gcc.gnu.org/PR7445
12848 245. https://gcc.gnu.org/PR7291
12849 246. https://gcc.gnu.org/onlinedocs/
12850 247. mailto:gcc-help@gcc.gnu.org
12851 248. mailto:gcc@gcc.gnu.org
12852 249. https://gcc.gnu.org/lists.html
12853 250. http://www.fsf.org/
12854 251. https://gcc.gnu.org/about.html
12855 252. http://validator.w3.org/check/referer
12856 ======================================================================
12857 http://gcc.gnu.org/gcc-3.1/index.html
12862 The [1]GNU project and the GCC developers are pleased to announce the
12863 release of GCC 3.1.1.
12865 The links below still apply to GCC 3.1.1.
12869 The [2]GNU project and the GCC developers are pleased to announce the
12870 release of GCC 3.1.
12872 GCC used to stand for the GNU C Compiler, but since the compiler
12873 supports several other languages aside from C, it now stands for the
12874 GNU Compiler Collection.
12876 A list of [3]successful builds is updated as new information becomes
12879 The GCC developers would like to thank the numerous people that have
12880 contributed [4]new features, improvements, bug fixes, and other changes
12881 as well as test results to GCC. This [5]amazing group of volunteers is
12882 what makes GCC successful.
12884 For additional information about GCC please refer to the [6]GCC project
12885 web site or contact the [7]GCC development mailing list.
12887 To obtain GCC please use [8]our mirror sites, or our CVS server.
12888 __________________________________________________________________
12891 For questions related to the use of GCC, please consult these web
12892 pages and the [9]GCC manuals. If that fails, the
12893 [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12894 web pages and the development of GCC are welcome on our developer
12895 list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
12898 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
12899 distribution of this entire article is permitted in any medium,
12900 provided this notice is preserved.
12902 These pages are [14]maintained by the GCC team. Last modified
12907 1. http://www.gnu.org/
12908 2. http://www.gnu.org/
12909 3. http://gcc.gnu.org/gcc-3.1/buildstat.html
12910 4. http://gcc.gnu.org/gcc-3.1/changes.html
12911 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12912 6. http://gcc.gnu.org/index.html
12913 7. mailto:gcc@gcc.gnu.org
12914 8. http://gcc.gnu.org/mirrors.html
12915 9. https://gcc.gnu.org/onlinedocs/
12916 10. mailto:gcc-help@gcc.gnu.org
12917 11. mailto:gcc@gcc.gnu.org
12918 12. https://gcc.gnu.org/lists.html
12919 13. http://www.fsf.org/
12920 14. https://gcc.gnu.org/about.html
12921 15. http://validator.w3.org/check/referer
12922 ======================================================================
12923 http://gcc.gnu.org/gcc-3.1/changes.html
12924 GCC 3.1 Release Series
12925 Changes, New Features, and Fixes
12927 Additional changes in GCC 3.1.1
12929 * A bug related to how structures and unions are returned has been
12930 fixed for powerpc-*-netbsd*.
12931 * An important bug in the implementation of -fprefetch-loop-arrays
12932 has been fixed. Previously the optimization prefetched random
12933 blocks of memory for most targets except for i386.
12934 * The Java compiler now compiles Java programs much faster and also
12935 works with parallel make.
12936 * Nested functions have been fixed for mips*-*-netbsd*.
12937 * Some missing floating point support routines have beed added for
12939 * This [1]message gives additional information about the bugs fixed
12944 * The -traditional C compiler option has been deprecated and will be
12945 removed in GCC 3.3. (It remains possible to preprocess non-C code
12946 with the traditional preprocessor.)
12947 * The default debugging format for most ELF platforms (including
12948 GNU/Linux and FreeBSD; notable exception is Solaris) has changed
12949 from stabs to DWARF2. This requires GDB 5.1.1 or later.
12951 General Optimizer Improvements
12953 * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat,
12954 and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure
12955 for profile driven optimizations.
12956 Options -fprofile-arcs and -fbranch-probabilities can now be used
12957 to improve speed of the generated code by profiling the actual
12958 program behaviour on typical runs. In the absence of profile info
12959 the compiler attempts to guess the profile statically.
12960 * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to
12961 monitor performance of the generated code.
12962 According to the SPECInt2000 results on an AMD Athlon CPU, the code
12963 generated by GCC 3.1 is 6% faster on the average (8.2% faster with
12964 profile feedback) compared to GCC 3.0. The code produced by GCC 3.0
12965 is about 2.1% faster compared to 2.95.3. Tests were done using the
12966 -O2 -march=athlon command-line options.
12967 * Alexandre Oliva, of Red Hat, has generalized the tree inlining
12968 infrastructure developed by CodeSourcery, LLC for the C++ front
12969 end, so that it is now used in the C front end too. Inlining
12970 functions as trees exposes them earlier to the compiler, giving it
12971 more opportunities for optimization.
12972 * Support for data prefetching instructions has been added to the GCC
12973 back end and several targets. A new __builtin_prefetch intrinsic is
12974 available to explicitly insert prefetch instructions and
12975 experimental support for loop array prefetching has been added (see
12976 -fprefetch-loop-array documentation).
12977 * Support for emitting debugging information for macros has been
12978 added for DWARF2. It is activated using -g3.
12980 New Languages and Language specific improvements
12984 * A few more [4]ISO C99 features.
12985 * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0.
12986 * The preprocessor's symbol table has been merged with the symbol
12987 table of the C, C++ and Objective-C front ends.
12988 * The preprocessor consumes less memory than the preprocessor in GCC
12989 3.0, often significantly so. On normal input files, it typically
12990 consumes less memory than pre-3.0 cccp-based GCC, too.
12994 * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std
12995 was a workaround to allow std compliant code to work with the
12996 non-std compliant libstdc++-v2. libstdc++-v3 is std compliant.
12997 * The C++ ABI has been fixed so that void (A::*)() const is mangled
12998 as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only
12999 affects pointer to cv-qualified member function types.
13000 * The C++ ABI has been changed to correctly handle this code:
13002 void operator delete[] (void *, size_t);
13005 struct B : public A {
13010 The amount of storage allocated for the array will be greater than
13011 it was in 3.0, in order to store the number of elements in the
13012 array, so that the correct size can be passed to operator delete[]
13013 when the array is deleted. Previously, the value passed to operator
13014 delete[] was unpredictable.
13015 This change will only affect code that declares a two-argument
13016 operator delete[] with a second parameter of type size_t in a base
13017 class, and does not override that definition in a derived class.
13018 * The C++ ABI has been changed so that:
13020 void operator delete[] (void *, size_t);
13021 void operator delete[] (void *);
13024 does not cause unnecessary storage to be allocated when an array of
13025 A objects is allocated.
13026 This change will only affect code that declares both of these forms
13027 of operator delete[], and declared the two-argument form before the
13029 * The C++ ABI has been changed so that when a parameter is passed by
13030 value, any cleanup for that parameter is performed in the caller,
13031 as specified by the ia64 C++ ABI, rather than the called function
13032 as before. As a result, classes with a non-trivial destructor but a
13033 trivial copy constructor will be passed and returned by invisible
13034 reference, rather than by bitwise copy as before.
13035 * G++ now supports the "named return value optimization": for code
13043 G++ will allocate a in the return value slot, so that the return
13044 becomes a no-op. For this to work, all return statements in the
13045 function must return the same variable.
13046 * Improvements to the C++ library are listed in [5]the libstdc++-v3
13051 * Annoying linker warnings (due to incorrect code being generated)
13053 * If a class method cannot be found, the compiler no longer issues a
13054 warning if a corresponding instance method exists in the root
13056 * Forward @protocol declarations have been fixed.
13057 * Loading of categories has been fixed in certain situations (GNU run
13059 * The class lookup in the run-time library has been rewritten so that
13060 class method dispatch is more than twice as fast as it used to be
13061 (GNU run time only).
13065 * libgcj now includes RMI, java.lang.ref.*, javax.naming, and
13067 * Property files and other system resources can be compiled into
13068 executables which use libgcj using the new gcj --resource feature.
13069 * libgcj has been ported to more platforms. In particular there is
13070 now a mostly-functional mingw32 (Windows) target port.
13071 * JNI and CNI invocation interfaces were implemented, so gcj-compiled
13072 Java code can now be called from a C/C++ application.
13073 * gcj can now use builtin functions for certain known methods, for
13075 * gcj can now automatically remove redundant array-store checks in
13077 * The --no-store-checks optimization option was added. This can be
13078 used to omit runtime store checks for code which is known not to
13079 throw ArrayStoreException
13080 * The following third party interface standards were added to libgcj:
13081 org.w3c.dom and org.xml.sax.
13082 * java.security has been merged with GNU Classpath. The new package
13083 is now JDK 1.2 compliant, and much more complete.
13084 * A bytecode verifier was added to the libgcj interpreter.
13085 * java.lang.Character was rewritten to comply with the Unicode 3.0
13086 standard, and improve performance.
13087 * Partial support for many more locales was added to libgcj.
13088 * Socket timeouts have been implemented.
13089 * libgcj has been merged into a single shared library. There are no
13090 longer separate shared libraries for the garbage collector and
13092 * Several performance improvements were made to gcj and libgcj:
13093 + Hash synchronization (thin locks)
13094 + A special allocation path for finalizer-free objects
13095 + Thread-local allocation
13096 + Parallel GC, and other GC tweaks
13100 Fortran improvements are listed in [6]the Fortran documentation.
13104 [7]Ada Core Technologies, Inc, has contributed its GNAT Ada 95 front
13105 end and associated tools. The GNAT compiler fully implements the Ada
13106 language as defined by the ISO/IEC 8652 standard.
13108 Please note that the integration of the Ada front end is still work in
13111 New Targets and Target Specific Improvements
13113 * Hans-Peter Nilsson has contributed a port to [8]MMIX, the CPU
13114 architecture used in new editions of Donald E. Knuth's The Art of
13115 Computer Programming.
13116 * [9]Axis Communications has contributed its port to the CRIS CPU
13117 architecture, used in the ETRAX system-on-a-chip series. See
13118 [10]Axis' developer site for technical information.
13119 * Alexandre Oliva, of Red Hat, has contributed a port to the
13120 [11]SuperH SH5 64-bit RISC microprocessor architecture, extending
13121 the existing SH port.
13122 * UltraSPARC is fully supported in 64-bit mode. The option -m64
13124 * For compatibility with the Sun compiler #pragma redefine_extname
13125 has been implemented on Solaris.
13126 * The x86 back end has had some noticeable work done to it.
13127 + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas
13128 Jaeger have contributed a port to the AMD x86-64 architecture.
13129 For more information on x86-64 see [12]http://www.x86-64.org.
13130 + The compiler now supports MMX, 3DNow!, SSE, and SSE2
13131 instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will
13132 enable the respective instruction sets. Intel C++ compatible
13133 MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics
13134 will be added in next major release.
13135 + Following those improvements, targets for Pentium MMX, K6-2,
13136 K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were
13137 added. Refer to the documentation on -march= and -mcpu=
13138 options for details.
13139 + For those targets that support it, -mfpmath=sse will cause the
13140 compiler to generate SSE/SSE2 instructions for floating point
13141 math instead of x87 instructions. Usually, this will lead to
13142 quicker code — especially on the Pentium 4. Note that only
13143 scalar floating point instructions are used and GCC does not
13144 exploit SIMD features yet.
13145 + Prefetch support has been added to the Pentium III, Pentium 4,
13146 K6-2, K6-3, and Athlon series.
13147 + Code generated for floating point to integer conversions has
13148 been improved leading to better performance of many 3D
13150 * The PowerPC back end has added 64-bit PowerPC GNU/Linux support.
13151 * C++ support for AIX has been improved.
13152 * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the
13153 PowerPC port supporting the AltiVec programming model (SIMD). The
13154 support, though presently useful, is experimental and is expected
13155 to stabilize for 3.2. The support is written to conform to
13156 Motorola's AltiVec specs. See -maltivec.
13160 Support for a number of older systems has been declared obsolete in GCC
13161 3.1. Unless there is activity to revive them, the next release of GCC
13162 will have their sources permanently removed.
13164 All configurations of the following processor architectures have been
13166 * MIL-STD-1750A, 1750a-*-*
13167 * AMD A29k, a29k-*-*
13168 * Convex, c*-convex-*
13169 * Clipper, clipper-*-*
13171 * Intel i860, i860-*-*
13172 * Sun picoJava, pj-*-* and pjl-*-*
13173 * Western Electric 32000, we32k-*-*
13175 Most configurations of the following processor architectures have been
13176 declared obsolete, but we are preserving a few systems which may have
13177 active developers. It is unlikely that the remaining systems will
13178 survive much longer unless we see definite signs of port activity.
13179 * Motorola 88000 except
13180 + Generic a.out, m88k-*-aout*
13181 + Generic SVR4, m88k-*-sysv4
13182 + OpenBSD, m88k-*-openbsd*
13184 + NetBSD, ns32k-*-netbsd*
13185 + OpenBSD, ns32k-*-openbsd*.
13187 + OpenBSD, romp-*-openbsd*.
13189 Finally, only some configurations of these processor architectures are
13192 + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka
13193 alpha*-*-osf[45], are still supported.)
13195 + RISCiX, arm-*-riscix*.
13197 + 386BSD, i?86-*-bsd*
13198 + Chorus, i?86-*-chorusos*
13199 + DG/UX, i?86-*-dgux*
13200 + FreeBSD 1.x, i?86-*-freebsd1.*
13201 + IBM AIX, i?86-*-aix*
13202 + ISC UNIX, i?86-*-isc*
13203 + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld*
13204 + NEXTstep, i?86-next-*
13205 + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose*
13206 + RTEMS/coff, i?86-*-rtemscoff*
13207 + RTEMS/go32, i?86-go32-rtems*
13208 + Sequent/BSD, i?86-sequent-bsd*
13209 + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and
13210 i?86-sequent-sysv3*
13211 + SunOS, i?86-*-sunos*
13213 + Altos, m68[k0]*-altos-*
13214 + Apollo, m68[k0]*-apollo-*
13215 + Apple A/UX, m68[k0]*-apple-*
13216 + Bull, m68[k0]*-bull-*
13217 + Convergent, m68[k0]*-convergent-*
13218 + Generic SVR3, m68[k0]*-*-sysv3*
13219 + ISI, m68[k0]*-isi-*
13220 + LynxOS, m68[k0]*-*-lynxos*
13221 + NEXT, m68[k0]*-next-*
13222 + RTEMS/coff, m68[k0]*-*-rtemscoff*
13223 + Sony, m68[k0]*-sony-*
13225 + DEC Ultrix, mips-*-ultrix* and mips-dec-*
13226 + Generic BSD, mips-*-bsd*
13227 + Generic System V, mips-*-sysv*
13228 + IRIX before version 5, mips-sgi-irix[1234]*
13229 + RiscOS, mips-*-riscos*
13230 + Sony, mips-sony-*
13231 + Tandem, mips-tandem-*
13233 + RTEMS/a.out, sparc-*-rtemsaout*.
13235 Documentation improvements
13237 * The old manual ("Using and Porting the GNU Compiler Collection")
13238 has been replaced by a users manual ("Using the GNU Compiler
13239 Collection") and a separate internals reference manual ("GNU
13240 Compiler Collection Internals").
13241 * More complete and much improved documentation about GCC's internal
13242 representation used by the C and C++ front ends.
13243 * Many cleanups and improvements in general.
13246 For questions related to the use of GCC, please consult these web
13247 pages and the [13]GCC manuals. If that fails, the
13248 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13249 web pages and the development of GCC are welcome on our developer
13250 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
13253 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
13254 distribution of this entire article is permitted in any medium,
13255 provided this notice is preserved.
13257 These pages are [18]maintained by the GCC team. Last modified
13262 1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html
13263 2. http://gcc.gnu.org/news/profiledriven.html
13264 3. http://gcc.gnu.org/benchmarks/
13265 4. http://gcc.gnu.org/c99status.html
13266 5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html
13267 6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html
13268 7. http://www.adacore.com/
13269 8. http://www-cs-faculty.stanford.edu/~knuth/mmix.html
13270 9. http://www.axis.com/
13271 10. http://developer.axis.com/
13272 11. http://www.superh.com/
13273 12. http://www.x86-64.org/
13274 13. https://gcc.gnu.org/onlinedocs/
13275 14. mailto:gcc-help@gcc.gnu.org
13276 15. mailto:gcc@gcc.gnu.org
13277 16. https://gcc.gnu.org/lists.html
13278 17. http://www.fsf.org/
13279 18. https://gcc.gnu.org/about.html
13280 19. http://validator.w3.org/check/referer
13281 ======================================================================
13282 http://gcc.gnu.org/gcc-3.0/index.html
13287 The [1]GNU project and the GCC developers are pleased to announce the
13288 release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0
13291 GCC used to stand for the GNU C Compiler, but since the compiler
13292 supports several other languages aside from C, it now stands for the
13293 GNU Compiler Collection.
13295 GCC 3.0.x has several new optimizations, new targets, new languages and
13296 many other new features, relative to GCC 2.95.x. See the [2]new
13297 features page for a more complete list.
13299 A list of [3]successful builds is updated as new information becomes
13302 The GCC developers would like to thank the numerous people that have
13303 contributed new features, test results, bug fixes, etc to GCC. This
13304 [4]amazing group of volunteers is what makes GCC successful.
13306 And finally, we can't in good conscience fail to mention some
13307 [5]caveats to using GCC 3.0.x.
13309 For additional information about GCC please refer to the [6]GCC project
13310 web site or contact the [7]GCC development mailing list.
13312 To obtain GCC please use [8]our mirror sites, or our CVS server.
13313 __________________________________________________________________
13315 Previous 3.0.x Releases
13317 December 20, 2001: GCC 3.0.3 has been released.
13318 October 25, 2001: GCC 3.0.2 has been released.
13319 August 20, 2001: GCC 3.0.1 has been released.
13320 June 18, 2001: GCC 3.0 has been released.
13323 For questions related to the use of GCC, please consult these web
13324 pages and the [9]GCC manuals. If that fails, the
13325 [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13326 web pages and the development of GCC are welcome on our developer
13327 list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
13330 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
13331 distribution of this entire article is permitted in any medium,
13332 provided this notice is preserved.
13334 These pages are [14]maintained by the GCC team. Last modified
13339 1. http://www.gnu.org/
13340 2. http://gcc.gnu.org/gcc-3.0/features.html
13341 3. http://gcc.gnu.org/gcc-3.0/buildstat.html
13342 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
13343 5. http://gcc.gnu.org/gcc-3.0/caveats.html
13344 6. http://gcc.gnu.org/index.html
13345 7. mailto:gcc@gcc.gnu.org
13346 8. http://gcc.gnu.org/mirrors.html
13347 9. https://gcc.gnu.org/onlinedocs/
13348 10. mailto:gcc-help@gcc.gnu.org
13349 11. mailto:gcc@gcc.gnu.org
13350 12. https://gcc.gnu.org/lists.html
13351 13. http://www.fsf.org/
13352 14. https://gcc.gnu.org/about.html
13353 15. http://validator.w3.org/check/referer
13354 ======================================================================
13355 http://gcc.gnu.org/gcc-3.0/features.html
13356 GCC 3.0 New Features
13358 Additional changes in GCC 3.0.4
13360 * GCC 3.0 now supports newer versions of the [1]NetBSD operating
13361 system, which use the ELF object file format, on x86 processors.
13362 * Correct debugging information is generated from functions that have
13363 lines from multiple files (e.g. yacc output).
13364 * A fix for whitespace handling in the -traditional preprocessor,
13365 which can affect Fortran.
13366 * Fixes to the exception handling runtime.
13367 * More fixes for bad code generation in C++.
13368 * A fix for shared library generation under AIX 4.3.
13369 * Documentation updates.
13370 * Port of GCC to Tensilica's Xtensa processor contributed.
13371 * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link).
13373 Additional changes in GCC 3.0.3
13375 * A fix to correct an accidental change to the PowerPC ABI.
13376 * Fixes for bad code generation on a variety of architectures.
13377 * Improvements to the debugging information generated for C++
13379 * Fixes for bad code generation in C++.
13380 * A fix to avoid crashes in the C++ demangler.
13381 * A fix to the C++ standard library to avoid buffer overflows.
13382 * Miscellaneous improvements for a variety of architectures.
13384 Additional changes in GCC 3.0.2
13386 * Fixes for bad code generation during loop unrolling.
13387 * Fixes for bad code generation by the sibling call optimization.
13388 * Minor improvements to x86 code generation.
13389 * Implementation of function descriptors in C++ vtables for IA64.
13390 * Numerous minor bug-fixes.
13392 Additional changes in GCC 3.0.1
13394 * C++ fixes for incorrect code-generation.
13395 * Improved cross-compiling support for the C++ standard library.
13396 * Fixes for some embedded targets that worked in GCC 2.95.3, but not
13398 * Fixes for various exception-handling bugs.
13399 * A port to the S/390 architecture.
13401 General Optimizer Improvements
13403 * [2]Basic block reordering pass.
13404 * New if-conversion pass with support for conditional (predicated)
13406 * New tail call and sibling call elimination optimizations.
13407 * New register renaming pass.
13408 * New (experimental) [3]static single assignment (SSA) representation
13410 * New dead-code elimination pass implemented using the SSA
13412 * [4]Global null pointer test elimination.
13413 * [5]Global code hoisting/unification.
13414 * More builtins and optimizations for stdio.h, string.h and old BSD
13415 functions, as well as for ISO C99 functions.
13416 * New builtin __builtin_expect for giving hints to the branch
13419 New Languages and Language specific improvements
13421 * The GNU Compiler for the Java(TM) language (GCJ) is now integrated
13422 and supported, including the run-time library containing most
13423 common non-GUI Java classes, a bytecode interpreter, and the Boehm
13424 conservative garbage collector. Many bugs have been fixed. GCJ can
13425 compile Java source or Java bytecodes to either native code or Java
13426 class files, and supports native methods written in either the
13427 standard JNI or the more efficient and convenient CNI.
13428 * Here is a [6]partial list of C++ improvements, both new features
13429 and those no longer supported.
13430 * New C++ ABI. On the IA-64 platform GCC is capable of
13431 inter-operating with other IA-64 compilers.
13432 * The new ABI also significantly reduces the size of symbol and debug
13434 * New [7]C++ support library and many C++ bug fixes, vastly improving
13435 our conformance to the ISO C++ standard.
13436 * New [8]inliner for C++.
13437 * Rewritten C preprocessor, integrated into the C, C++ and Objective
13438 C compilers, with very many improvements including ISO C99 support
13439 and [9]improvements to dependency generation.
13440 * Support for more [10]ISO C99 features.
13441 * Many improvements to support for checking calls to format functions
13442 such as printf and scanf, including support for ISO C99 format
13443 features, extensions from the Single Unix Specification and GNU
13444 libc 2.2, checking of strfmon formats and features to assist in
13445 auditing for format string security bugs.
13446 * New warnings for C code that may have undefined semantics because
13447 of violations of sequence point rules in the C standard (such as a
13448 = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall.
13449 * Additional warning option -Wfloat-equal.
13450 * Improvements to -Wtraditional.
13451 * Fortran improvements are listed in [11]the Fortran documentation.
13453 New Targets and Target Specific Improvements
13455 * New x86 back end, generating much improved code.
13456 * Support for a generic i386-elf target contributed.
13457 * New option to emit x86 assembly code using Intel style syntax
13459 * HPUX 11 support contributed.
13460 * Improved PowerPC code generation, including scheduled prologue and
13462 * Port of GCC to Intel's IA-64 processor contributed.
13463 * Port of GCC to Motorola's MCore 210 and 340 contributed.
13464 * New unified back-end for Arm, Thumb and StrongArm contributed.
13465 * Port of GCC to Intel's XScale processor contributed.
13466 * Port of GCC to Atmel's AVR microcontrollers contributed.
13467 * Port of GCC to Mitsubishi's D30V processor contributed.
13468 * Port of GCC to Matsushita's AM33 processor (a member of the MN10300
13469 processor family) contributed.
13470 * Port of GCC to Fujitsu's FR30 processor contributed.
13471 * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed.
13472 * Port of GCC to Sun's picoJava processor core contributed.
13474 Documentation improvements
13476 * Substantially rewritten and improved C preprocessor manual.
13477 * Many improvements to other documentation.
13478 * Manpages for gcc, cpp and gcov are now generated automatically from
13479 the master Texinfo manual, eliminating the problem of manpages
13480 being out of date. (The generated manpages are only extracts from
13481 the full manual, which is provided in Texinfo form, from which
13482 info, HTML, other formats and a printed manual can be generated.)
13483 * Generated info files are included in the release tarballs alongside
13484 their Texinfo sources, avoiding problems on some platforms with
13485 building makeinfo as part of the GCC distribution.
13487 Other significant improvements
13489 * Garbage collection used internally by the compiler for most memory
13490 allocation instead of obstacks.
13491 * Lengauer and Tarjan algorithm used for computing dominators in the
13492 CFG. This algorithm can be significantly faster and more space
13493 efficient than our older algorithm.
13494 * gccbug script provided to assist in submitting bug reports to our
13495 bug tracking system. (Bug reports previously submitted directly to
13496 our mailing lists, for which you received no bug tracking number,
13497 should be submitted again using gccbug if you can reproduce the
13498 problem with GCC 3.0.)
13499 * The internal libgcc library is [12]built as a shared library on
13500 systems that support it.
13501 * Extensive testsuite included with GCC, with many new tests. In
13502 addition to tests for GCC bugs that have been fixed, many tests
13503 have been added for language features, compiler warnings and
13505 * Additional language-independent warning options -Wpacked, -Wpadded,
13506 -Wunreachable-code and -Wdisabled-optimization.
13507 * Target-independent options -falign-functions, -falign-loops and
13510 Plus a great many bug fixes and almost all the [13]features found in
13514 For questions related to the use of GCC, please consult these web
13515 pages and the [14]GCC manuals. If that fails, the
13516 [15]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13517 web pages and the development of GCC are welcome on our developer
13518 list at [16]gcc@gcc.gnu.org. All of [17]our lists have public
13521 Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and
13522 distribution of this entire article is permitted in any medium,
13523 provided this notice is preserved.
13525 These pages are [19]maintained by the GCC team. Last modified
13530 1. http://www.netbsd.org/
13531 2. http://gcc.gnu.org/news/reorder.html
13532 3. http://gcc.gnu.org/news/ssa.html
13533 4. http://gcc.gnu.org/news/null.html
13534 5. http://gcc.gnu.org/news/unify.html
13535 6. http://gcc.gnu.org/gcc-3.0/c++features.html
13536 7. http://gcc.gnu.org/libstdc++/
13537 8. http://gcc.gnu.org/news/inlining.html
13538 9. http://gcc.gnu.org/news/dependencies.html
13539 10. http://gcc.gnu.org/c99status.html
13540 11. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
13541 12. http://gcc.gnu.org/gcc-3.0/libgcc.html
13542 13. http://gcc.gnu.org/gcc-2.95/features.html
13543 14. https://gcc.gnu.org/onlinedocs/
13544 15. mailto:gcc-help@gcc.gnu.org
13545 16. mailto:gcc@gcc.gnu.org
13546 17. https://gcc.gnu.org/lists.html
13547 18. http://www.fsf.org/
13548 19. https://gcc.gnu.org/about.html
13549 20. http://validator.w3.org/check/referer
13550 ======================================================================
13551 http://gcc.gnu.org/gcc-3.0/caveats.html
13554 * -fstrict-aliasing is now part of -O2 and higher optimization
13555 levels. This allows the compiler to assume the strictest aliasing
13556 rules applicable to the language being compiled. For C and C++,
13557 this activates optimizations based on the type of expressions. This
13558 optimization may thus break old, non-compliant code.
13559 * Enumerations are now properly promoted to int in function
13560 parameters and function returns. Normally this change is not
13561 visible, but when using -fshort-enums this is an ABI change.
13562 * The undocumented extension that allowed C programs to have a label
13563 at the end of a compound statement has been deprecated and may be
13564 removed in a future version. Programs that now generate a warning
13565 about this may be fixed by adding a null statement (a single
13566 semicolon) after the label.
13567 * The poorly documented extension that allowed string constants in C,
13568 C++ and Objective C to contain unescaped newlines has been
13569 deprecated and may be removed in a future version. Programs using
13570 this extension may be fixed in several ways: the bare newline may
13571 be replaced by \n, or preceded by \n\, or string concatenation may
13572 be used with the bare newline preceded by \n" and " placed at the
13573 start of the next line.
13574 * The Chill compiler is not included in GCC 3.0, because of the lack
13575 of a volunteer to convert it to use garbage collection.
13576 * Certain non-standard iostream methods from earlier versions of
13577 libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach,
13578 ostream::form, and istream::gets.
13579 * The new C++ ABI is not yet fully supported by current (as of
13580 2001-07-01) releases and development versions of GDB, or any
13581 earlier versions. There is a problem setting breakpoints by line
13582 number, and other related issues that have been fixed in GCC 3.0
13583 but not yet handled in GDB:
13584 [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
13587 For questions related to the use of GCC, please consult these web
13588 pages and the [2]GCC manuals. If that fails, the
13589 [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13590 web pages and the development of GCC are welcome on our developer
13591 list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
13593 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
13594 distribution of this entire article is permitted in any medium,
13595 provided this notice is preserved.
13597 These pages are [7]maintained by the GCC team. Last modified
13602 1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
13603 2. https://gcc.gnu.org/onlinedocs/
13604 3. mailto:gcc-help@gcc.gnu.org
13605 4. mailto:gcc@gcc.gnu.org
13606 5. https://gcc.gnu.org/lists.html
13607 6. http://www.fsf.org/
13608 7. https://gcc.gnu.org/about.html
13609 8. http://validator.w3.org/check/referer
13610 ======================================================================
13611 http://gcc.gnu.org/gcc-2.95/index.html
13614 March 16, 2001: The GNU project and the GCC developers are pleased to
13615 announce the release of GCC version 2.95.3.
13629 July 31, 1999. This is the first release of GCC since the April
13630 1999 GCC/EGCS reunification and includes nearly a year's worth
13631 of new development and bugfixes.
13633 References and Acknowledgements
13635 GCC used to stand for the GNU C Compiler, but since the compiler
13636 supports several other languages aside from C, it now stands for the
13637 GNU Compiler Collection.
13639 The whole suite has been extensively [1]regression tested and
13640 [2]package tested. It should be reliable and suitable for widespread
13643 The compiler has several new optimizations, new targets, new languages
13644 and other new features. See the [3]new features page for a more
13645 complete list of new features found in the GCC 2.95 releases.
13647 The sources include installation instructions in both HTML and
13648 plaintext forms in the install directory in the distribution. However,
13649 the most up to date installation instructions and [4]build/test status
13650 are on the web pages. We will update those pages as new information
13653 The GCC developers would like to thank the numerous people that have
13654 contributed new features, test results, bugfixes, etc to GCC. This
13655 [5]amazing group of volunteers is what makes GCC successful.
13657 And finally, we can't in good conscience fail to mention some
13658 [6]caveats to using GCC 2.95.
13660 Download GCC 2.95 from one of our many [7]mirror sites.
13662 For additional information about GCC please see the [8]GCC project web
13663 server or contact the [9]GCC development mailing list.
13666 For questions related to the use of GCC, please consult these web
13667 pages and the [10]GCC manuals. If that fails, the
13668 [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13669 web pages and the development of GCC are welcome on our developer
13670 list at [12]gcc@gcc.gnu.org. All of [13]our lists have public
13673 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and
13674 distribution of this entire article is permitted in any medium,
13675 provided this notice is preserved.
13677 These pages are [15]maintained by the GCC team. Last modified
13682 1. http://gcc.gnu.org/gcc-2.95/regress.html
13683 2. http://gcc.gnu.org/gcc-2.95/othertest.html
13684 3. http://gcc.gnu.org/gcc-2.95/features.html
13685 4. http://gcc.gnu.org/gcc-2.95/buildstat.html
13686 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
13687 6. http://gcc.gnu.org/gcc-2.95/caveats.html
13688 7. http://gcc.gnu.org/mirrors.html
13689 8. http://gcc.gnu.org/index.html
13690 9. mailto:gcc@gcc.gnu.org
13691 10. https://gcc.gnu.org/onlinedocs/
13692 11. mailto:gcc-help@gcc.gnu.org
13693 12. mailto:gcc@gcc.gnu.org
13694 13. https://gcc.gnu.org/lists.html
13695 14. http://www.fsf.org/
13696 15. https://gcc.gnu.org/about.html
13697 16. http://validator.w3.org/check/referer
13698 ======================================================================
13699 http://gcc.gnu.org/gcc-2.95/features.html
13700 GCC 2.95 New Features
13702 * General Optimizer Improvements:
13703 + [1]Localized register spilling to improve speed and code
13704 density especially on small register class machines.
13705 + [2]Global CSE using lazy code motion algorithms.
13706 + [3]Improved global constant/copy propagation.
13707 + [4]Improved control flow graph analysis and manipulation.
13708 + [5]Local dead store elimination.
13709 + [6]Memory Load hoisting/store sinking in loops.
13710 + [7]Type based alias analysis is enabled by default. Note this
13711 feature will expose bugs in the Linux kernel. Please refer to
13712 the FAQ (as shipped with GCC 2.95) for additional information
13714 + Major revamp of GIV detection, combination and simplification
13715 to improve loop performance.
13716 + Major improvements to register allocation and reloading.
13717 * New Languages and Language specific improvements
13718 + [8]Many C++ improvements.
13719 + [9]Many Fortran improvements.
13720 + [10]Java front-end has been integrated. [11]runtime library is
13721 available separately.
13722 + [12]ISO C99 support
13723 + [13]Chill front-end and runtime has been integrated.
13724 + Boehm garbage collector support in libobjc.
13725 + More support for various pragmas which appear in vendor
13727 * New Targets and Target Specific Improvements
13728 + [14]SPARC backend rewrite.
13729 + -mschedule=8000 will optimize code for PA8000 class
13730 processors; -mpa-risc-2-0 will generate code for PA2.0
13732 + Various micro-optimizations for the ia32 port. K6
13734 + Compiler will attempt to align doubles in the stack on the
13736 + Alpha EV6 support
13738 + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403.
13739 -mcpu=e603e was added to do -mcpu=603e and -msoft-float.
13744 + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix,
13746 + vxWorks targets include support for vxWorks threads
13747 + StrongARM 110 and ARM9 support added. ARM Scheduling
13748 parameters rewritten.
13749 + Various changes to the MIPS port to avoid assembler macros,
13750 which in turn improves performance
13751 + Various performance improvements to the i960 port.
13752 + Major rewrite of ns32k port
13753 * Other significant improvements
13754 + [15]Ability to dump cfg information and display it using vcg.
13755 + The new faster scheme for fixing vendor header files is
13756 enabled by default.
13757 + Experimental internationalization support.
13758 + multibyte character support
13759 + Some compile-time speedups for pathological problems
13760 + Better support for complex types
13761 * Plus the usual mountain of bugfixes
13762 * Core compiler is based on the gcc2 development tree from Sept 30,
13763 1998, so we have all of the [16]features found in GCC 2.8.
13765 Additional Changes in GCC 2.95.1
13767 * Generic bugfixes and improvements
13768 + Various documentation fixes related to the GCC/EGCS merger.
13769 + Fix memory management bug which could lead to spurious aborts,
13770 core dumps or random parsing errors in the compiler.
13771 + Fix a couple bugs in the dwarf1 and dwarf2 debug record
13773 + Fix infinite loop in the CSE optimizer.
13774 + Avoid undefined behavior in compiler FP emulation code
13775 + Fix install problem when prefix is overridden on the make
13777 + Fix problem with unwanted installation of assert.h on some
13779 + Fix problem with finding the wrong assembler in a single tree
13781 + Avoid increasing the known alignment of a register that is
13782 already known to be a pointer.
13783 * Platform specific bugfixes and improvements
13784 + Codegen bugfix for prologue/epilogue for cpu32 target.
13785 + Fix long long code generation bug for the Coldfire target.
13786 + Fix various aborts in the SH compiler.
13787 + Fix bugs in libgcc support library for the SH.
13788 + Fix alpha ev6 code generation bug.
13789 + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on
13791 + Fix -fpic code generation bug for rs6000/ppc svr4 targets.
13792 + Fix varargs/stdarg code generation bug for rs6000/ppc svr4
13794 + Fix weak symbol handling for rs6000/ppc svr4 targets.
13795 + Fix various problems with 64bit code generation for the
13797 + Fix codegen bug which caused tetex to be mis-compiled on the
13799 + Fix compiler abort in new cfg code exposed by x86 port.
13800 + Fix out of range array reference in code convert flat
13801 registers to the x87 stacked FP register file.
13802 + Fix minor vxworks configuration bug.
13803 + Fix return type of bsearch for SunOS 4.x.
13804 * Language & Runtime specific fixes.
13805 + The G++ signature extension has been deprecated. It will be
13806 removed in the next major release of G++. Use of signatures
13807 will result in a warning from the compiler.
13808 + Several bugs relating to templates and namespaces were fixed.
13809 + A bug that caused crashes when combining templates with -g on
13810 DWARF1 platforms was fixed.
13811 + Pointers-to-members, virtual functions, and multiple
13812 inheritance should now work together correctly.
13813 + Some code-generation bugs relating to function try blocks were
13815 + G++ is a little bit more lenient with certain archaic
13816 constructs than in GCC 2.95.
13817 + Fix to prevent shared library version #s from bring truncated
13819 + Fix missing std:: in the libstdc++ library.
13820 + Fix stream locking problems in libio.
13821 + Fix problem in java compiler driver.
13823 Additional Changes in GCC 2.95.2
13825 The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While
13826 the optimizations performed by -fstrict-aliasing are valid according to
13827 the C and C++ standards, the optimization have caused some problems,
13828 particularly with old non-conforming code.
13830 The GCC developers are experimenting with ways to warn users about code
13831 which violates the C/C++ standards, but those warnings are not ready
13832 for widespread use at this time. Rather than wait for those warnings
13833 the GCC developers have chosen to disable -fstrict-aliasing by default
13834 for the GCC 2.95.2 release.
13836 We strongly encourage developers to find and fix code which violates
13837 the C/C++ standards as -fstrict-aliasing may be enabled by default in
13838 future releases. Use the option -fstrict-aliasing to re-enable these
13840 * Generic bugfixes and improvements
13841 + Fix incorrectly optimized memory reference in global common
13842 subexpression elimination (GCSE) optimization pass.
13843 + Fix code generation bug in regmove.c in which it could
13844 incorrectly change a "const" value.
13845 + Fix bug in optimization of conditionals involving volatile
13847 + Avoid over-allocation of stack space for some procedures.
13848 + Fixed bug in the compiler which caused incorrect optimization
13849 of an obscure series of bit manipulations, shifts and
13851 + Fixed register allocator bug which caused teTeX to be
13852 mis-compiled on SPARC targets.
13853 + Avoid incorrect optimization of degenerate case statements for
13854 certain targets such as the ARM.
13855 + Fix out of range memory reference in the jump optimizer.
13856 + Avoid dereferencing null pointer in fix-header.
13857 + Fix test for GCC specific features so that it is possible to
13858 bootstrap with gcc-2.6.2 and older versions of GCC.
13859 + Fix typo in scheduler which could potentially cause out of
13860 range memory accesses.
13861 + Avoid incorrect loop reversal which caused incorrect code for
13862 certain loops on PowerPC targets.
13863 + Avoid incorrect optimization of switch statements on certain
13864 targets (for example the ARM).
13865 * Platform specific bugfixes and improvements
13866 + Work around bug in Sun V5.0 compilers which caused bootstrap
13867 comparison failures on SPARC targets.
13868 + Fix SPARC backend bug which caused aborts in final.c.
13869 + Fix sparc-hal-solaris2* configuration fragments.
13870 + Fix bug in sparc block profiling.
13871 + Fix obscure code generation bug for the PARISC targets.
13872 + Define __STDC_EXT__ for HPUX configurations.
13873 + Various POWERPC64 code generation bugfixes.
13874 + Fix abort for PPC targets using ELF (ex GNU/Linux).
13875 + Fix collect2 problems for AIX targets.
13876 + Correct handling of .file directive for PPC targets.
13877 + Fix bug in fix_trunc x86 patterns.
13878 + Fix x86 port to correctly pop the FP stack for functions that
13879 return structures in memory.
13880 + Fix minor bug in strlen x86 pattern.
13881 + Use stabs debugging instead of dwarf1 for x86-solaris targets.
13882 + Fix template repository code to handle leading underscore in
13884 + Fix weak/weak alias support for OpenBSD.
13885 + GNU/Linux for the ARM has C++ compatible include files.
13886 * Language & Runtime specific fixes.
13887 + Fix handling of constructor attribute in the C front-end which
13888 caused problems building the Chill runtime library on some
13890 + Fix minor problem merging type qualifiers in the C front-end.
13891 + Fix aliasing bug for pointers and references (C/C++).
13892 + Fix incorrect "non-constant initializer bug" when -traditional
13893 or -fwritable-strings is enabled.
13894 + Fix build error for Chill front-end on SunOS.
13895 + Do not complain about duplicate instantiations when using
13897 + Fix array bounds handling in C++ front-end which caused
13898 problems with dwarf debugging information in some
13900 + Fix minor namespace problem.
13901 + Fix problem linking java programs.
13903 Additional Changes in GCC 2.95.3
13905 * Generic bugfixes and improvements
13906 + Fix numerous problems that caused incorrect optimization in
13907 the register reloading code.
13908 + Fix numerous problems that caused incorrect optimization in
13909 the loop optimizer.
13910 + Fix aborts in the functions build_insn_chain and scan_loops
13911 under some circumstances.
13912 + Fix an alias analysis bug.
13913 + Fix an infinite compilation bug in the combiner.
13914 + A few problems with complex number support have been fixed.
13915 + It is no longer possible for gcc to act as a fork bomb when
13916 installed incorrectly.
13917 + The -fpack-struct option should be recognized now.
13918 + Fixed a bug that caused incorrect code to be generated due to
13919 a lost stack adjustment.
13920 * Platform specific bugfixes and improvements
13921 + Support building ARM toolchains hosted on Windows.
13922 + Fix attribute calculations in ARM toolchains.
13923 + arm-linux support has been improved.
13924 + Fix a PIC failure on sparc targets.
13925 + On ix86 targets, the regparm attribute should now work
13927 + Several updates for the h8300 port.
13928 + Fix problem building libio with glibc 2.2.
13931 For questions related to the use of GCC, please consult these web
13932 pages and the [17]GCC manuals. If that fails, the
13933 [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13934 web pages and the development of GCC are welcome on our developer
13935 list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
13938 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
13939 distribution of this entire article is permitted in any medium,
13940 provided this notice is preserved.
13942 These pages are [22]maintained by the GCC team. Last modified
13947 1. http://gcc.gnu.org/news/spill.html
13948 2. http://gcc.gnu.org/news/lcm.html
13949 3. http://gcc.gnu.org/news/cprop.html
13950 4. http://gcc.gnu.org/news/cfg.html
13951 5. http://gcc.gnu.org/news/dse.html
13952 6. http://gcc.gnu.org/news/hoist.html
13953 7. http://gcc.gnu.org/news/alias.html
13954 8. http://gcc.gnu.org/gcc-2.95/c++features.html
13955 9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
13956 10. http://gcc.gnu.org/java/gcj-announce.txt
13957 11. http://gcc.gnu.org/news/javaannounce.html
13958 12. http://gcc.gnu.org/c99status.html
13959 13. http://gcc.gnu.org/news/chill.html
13960 14. http://gcc.gnu.org/news/sparc.html
13961 15. http://gcc.gnu.org/news/egcs-vcg.html
13962 16. http://gcc.gnu.org/egcs-1.0/features-2.8.html
13963 17. https://gcc.gnu.org/onlinedocs/
13964 18. mailto:gcc-help@gcc.gnu.org
13965 19. mailto:gcc@gcc.gnu.org
13966 20. https://gcc.gnu.org/lists.html
13967 21. http://www.fsf.org/
13968 22. https://gcc.gnu.org/about.html
13969 23. http://validator.w3.org/check/referer
13970 ======================================================================
13971 http://gcc.gnu.org/gcc-2.95/caveats.html
13974 * GCC 2.95 will issue an error for invalid asm statements that had
13975 been silently accepted by earlier versions of the compiler. This is
13976 particularly noticeable when compiling older versions of the Linux
13977 kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95)
13978 for more information on this issue.
13979 * GCC 2.95 implements type based alias analysis to disambiguate
13980 memory references. Some programs, particularly the Linux kernel
13981 violate ANSI/ISO aliasing rules and therefore may not operate
13982 correctly when compiled with GCC 2.95. Please refer to the FAQ (as
13983 shipped with GCC 2.95) for more information on this issue.
13984 * GCC 2.95 has a known bug in its handling of complex variables for
13985 64bit targets. Instead of silently generating incorrect code, GCC
13986 2.95 will issue a fatal error for situations it can not handle.
13987 This primarily affects the Fortran community as Fortran makes more
13988 use of complex variables than C or C++.
13989 * GCC 2.95 has an integrated libstdc++, but does not have an
13990 integrated libg++. Furthermore old libg++ releases will not work
13991 with GCC 2.95. You can retrieve a recent copy of libg++ from the
13993 Note most C++ programs only need libstdc++.
13994 * Exception handling may not work with shared libraries, particularly
13995 on alphas, hppas, rs6000/powerpc and mips based platforms.
13996 Exception handling is known to work on x86 GNU/Linux platforms with
13998 * In general, GCC 2.95 is more rigorous about rejecting invalid C++
13999 code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0,
14000 or EGCS 1.1. As a result it may be necessary to fix C++ code before
14001 it will compile with GCC 2.95.
14002 * G++ is also converting toward the ISO C++ standard; as a result
14003 code which was previously valid (and thus accepted by other
14004 compilers and older versions of g++) may no longer be accepted. The
14005 flag -fpermissive may allow some non-conforming code to compile
14007 * GCC 2.95 compiled C++ code is not binary compatible with EGCS
14008 1.1.x, EGCS 1.0.x or GCC 2.8.x.
14009 * GCC 2.95 does not have changes from the GCC 2.8 tree that were made
14010 between Sept 30, 1998 and April 30, 1999 (the official end of the
14011 GCC 2.8 project). Future GCC releases will include all the changes
14012 from the defunct GCC 2.8 sources.
14015 For questions related to the use of GCC, please consult these web
14016 pages and the [2]GCC manuals. If that fails, the
14017 [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14018 web pages and the development of GCC are welcome on our developer
14019 list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
14021 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
14022 distribution of this entire article is permitted in any medium,
14023 provided this notice is preserved.
14025 These pages are [7]maintained by the GCC team. Last modified
14030 1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz
14031 2. https://gcc.gnu.org/onlinedocs/
14032 3. mailto:gcc-help@gcc.gnu.org
14033 4. mailto:gcc@gcc.gnu.org
14034 5. https://gcc.gnu.org/lists.html
14035 6. http://www.fsf.org/
14036 7. https://gcc.gnu.org/about.html
14037 8. http://validator.w3.org/check/referer
14038 ======================================================================
14039 http://gcc.gnu.org/egcs-1.1/index.html
14042 September 3, 1998: We are pleased to announce the release of EGCS 1.1.
14043 December 1, 1998: We are pleased to announce the release of EGCS 1.1.1.
14044 March 15, 1999: We are pleased to announce the release of EGCS 1.1.2.
14046 EGCS is a free software project to further the development of the GNU
14047 compilers using an open development environment.
14049 EGCS 1.1 is a major new release of the EGCS compiler system. It has
14050 been [1]extensively tested and is believed to be stable and suitable
14051 for widespread use.
14053 EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8
14054 development sources; it contains all of the new features found in GCC
14055 2.8.1 as well as all new development from GCC up to June 6, 1998.
14057 EGCS 1.1 also contains many improvements and features not found in GCC
14058 or in older versions of EGCS:
14059 * Global common subexpression elimination and global constant/copy
14060 propagation (aka [2]gcse)
14061 * Ongoing improvements to the [3]alias analysis support to allow for
14062 better optimizations throughout the compiler.
14063 * Vastly improved [4]C++ compiler and integrated C++ runtime
14065 * Fixes for the /tmp symlink race security problems.
14066 * New targets including mips16, arm-thumb and 64 bit PowerPC.
14067 * Improvements to GNU Fortran (g77) compiler and runtime library made
14068 since g77 version 0.5.23.
14070 See the [5]new features page for a more complete list of new features
14071 found in EGCS 1.1 releases.
14073 EGCS 1.1.1 is a minor update to fix several serious problems in EGCS
14075 * General improvements and fixes
14076 + Avoid some stack overflows when compiling large functions.
14077 + Avoid incorrect loop invariant code motions.
14078 + Fix some core dumps on Linux kernel code.
14079 + Bring back the imake -Di386 and friends fix from EGCS 1.0.2.
14080 + Fix code generation problem in gcse.
14081 + Various documentation related fixes.
14082 * g++/libstdc++ improvements and fixes
14083 + MT safe EH fix for setjmp/longjmp based exception handling.
14084 + Fix a few bad interactions between optimization and exception
14086 + Fixes for demangling of template names starting with "__".
14087 + Fix a bug that would fail to run destructors in some cases
14089 + Fix 'new' of classes with virtual bases.
14090 + Fix crash building Qt on the Alpha.
14091 + Fix failure compiling WIFEXITED macro on GNU/Linux.
14092 + Fix some -frepo failures.
14093 * g77 and libf2c improvements and fixes
14094 + Various documentation fixes.
14095 + Avoid compiler crash on RAND intrinsic.
14096 + Fix minor bugs in makefiles exposed by BSD make programs.
14097 + Define _XOPEN_SOURCE for libI77 build to avoid potential
14098 problems on some 64-bit systems.
14099 + Fix problem with implicit endfile on rewind.
14100 + Fix spurious recursive I/O errors.
14101 * platform specific improvements and fixes
14102 + Match all versions of UnixWare7.
14103 + Do not assume x86 SVR4 or UnixWare targets can handle stabs.
14104 + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion
14105 from unsigned ints to double precision floats.
14106 + Fix ARM ABI issue with NetBSD.
14107 + Fix a few arm code generation bugs.
14108 + Fixincludes will fix additional broken SCO OpenServer header
14110 + Fix a m68k backend bug which caused invalid offsets in reg+d
14112 + Fix problems with 64bit AIX 4.3 support.
14113 + Fix handling of long longs for varargs/stdarg functions on the
14115 + Minor fixes to CPP predefines for Windows.
14116 + Fix code generation problems with gpr<->fpr copies for 64bit
14118 + Fix a few coldfire code generation bugs.
14119 + Fix some more header file problems on SunOS 4.x.
14120 + Fix assert.h handling for RTEMS.
14121 + Fix Windows handling of TREE_SYMBOL_REFERENCED.
14122 + Fix x86 compiler abort in reg-stack pass.
14123 + Fix cygwin/windows problem with section attributes.
14124 + Fix Alpha code generation problem exposed by SMP Linux
14126 + Fix typo in m68k 32->64bit integer conversion.
14127 + Make sure target libraries build with -fPIC for PPC & Alpha
14130 EGCS 1.1.2 is a minor update to fix several serious problems in EGCS
14132 * General improvements and fixes
14133 + Fix bug in loop optimizer which caused the SPARC (and
14134 potentially other) ports to segfault.
14135 + Fix infinite recursion in alias analysis and combiner code.
14136 + Fix bug in regclass preferencing.
14137 + Fix incorrect loop reversal which caused incorrect code to be
14138 generated for several targets.
14139 + Fix return value for builtin memcpy.
14140 + Reduce compile time for certain loops which exposed quadratic
14141 behavior in the loop optimizer.
14142 + Fix bug which caused volatile memory to be written multiple
14143 times when only one write was needed/desired.
14144 + Fix compiler abort in caller-save.c
14145 + Fix combiner bug which caused incorrect code generation for
14146 certain division by constant operations.
14147 + Fix incorrect code generation due to a bug in range check
14149 + Fix incorrect code generation due to mis-handling of clobbered
14151 + Fix compiler abort/segfault due to incorrect register
14152 splitting when unrolling loops.
14153 + Fix code generation involving autoincremented addresses with
14155 + Work around bug in the scheduler which caused qt to be
14156 mis-compiled on some platforms.
14157 + Fix code generation problems with -fshort-enums.
14158 + Tighten security for temporary files.
14159 + Improve compile time for codes which make heavy use of
14160 overloaded functions.
14161 + Fix multiply defined constructor/destructor symbol problems.
14162 + Avoid setting bogus RPATH environment variable during
14164 + Avoid GNU-make dependencies in the texinfo subdir.
14165 + Install CPP wrapper script in $(prefix)/bin if --enable-cpp.
14166 --enable-cpp=<dirname> can be used to specify an additional
14167 install directory for the cpp wrapper script.
14168 + Fix CSE bug which caused incorrect label-label refs to appear
14170 + Avoid linking in EH routines from libgcc if they are not
14172 + Avoid obscure bug in aliasing code.
14173 + Fix bug in weak symbol handling.
14174 * Platform-specific improvements and fixes
14175 + Fix detection of PPro/PII on Unixware 7.
14176 + Fix compiler segfault when building spec99 and other programs
14178 + Fix code-generation bugs for integer and floating point
14179 conditional move instructions on the PPro/PII.
14180 + Use fixincludes to fix byteorder problems on i?86-*-sysv.
14181 + Fix build failure for the arc port.
14182 + Fix floating point format configuration for i?86-gnu port.
14183 + Fix problems with hppa1.0-hp-hpux10.20 configuration when
14184 threads are enabled.
14185 + Fix coldfire code generation bugs.
14186 + Fix "unrecognized insn" problems for Alpha and PPC ports.
14187 + Fix h8/300 code generation problem with floating point values
14189 + Fix unrecognized insn problems for the m68k port.
14190 + Fix namespace-pollution problem for the x86 port.
14191 + Fix problems with old assembler on x86 NeXT systems.
14192 + Fix PIC code-generation problems for the SPARC port.
14193 + Fix minor bug with LONG_CALLS in PowerPC SVR4 support.
14194 + Fix minor ISO namespace violation in Alpha varargs/stdarg
14196 + Fix incorrect "braf" instruction usage for the SH port.
14197 + Fix minor bug in va-sh which prevented its use with -ansi.
14198 + Fix problems recognizing and supporting FreeBSD.
14199 + Handle OpenBSD systems correctly.
14200 + Minor fixincludes fix for Digital UNIX 4.0B.
14201 + Fix problems with ctors/dtors in SCO shared libraries.
14202 + Abort instead of generating incorrect code for PPro/PII
14203 floating point conditional moves.
14204 + Avoid multiply defined symbols on GNU/Linux systems using
14206 + Fix abort in alpha compiler.
14207 * Fortran-specific fixes
14208 + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year
14209 is in the documented, non-Y2K-compliant range of 0-99, instead
14210 of being returned as 100 in the year 2000.
14211 + Fix the `Date_and_Time' intrinsic (in libg2c) to return the
14212 milliseconds value properly in Values(8).
14213 + Fix the `LStat' intrinsic (in libg2c) to return device-ID
14214 information properly in SArray(7).
14216 Each release includes installation instructions in both HTML and
14217 plaintext forms (see the INSTALL directory in the toplevel directory of
14218 the distribution). However, we also keep the most up to date
14219 installation instructions and [6]build/test status on our web page. We
14220 will update those pages as new information becomes available.
14222 The EGCS project would like to thank the numerous people that have
14223 contributed new features, test results, bugfixes, etc. This [7]amazing
14224 group of volunteers is what makes EGCS successful.
14226 And finally, we can't in good conscience fail to mention some
14227 [8]caveats to using EGCS 1.1.
14229 Download EGCS from egcs.cygnus.com (USA California).
14231 The EGCS 1.1 release is also available on many mirror sites.
14232 [9]Goto mirror list to find a closer site.
14235 For questions related to the use of GCC, please consult these web
14236 pages and the [10]GCC manuals. If that fails, the
14237 [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14238 web pages and the development of GCC are welcome on our developer
14239 list at [12]gcc@gcc.gnu.org. All of [13]our lists have public
14242 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and
14243 distribution of this entire article is permitted in any medium,
14244 provided this notice is preserved.
14246 These pages are [15]maintained by the GCC team. Last modified
14251 1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html
14252 2. http://gcc.gnu.org/news/gcse.html
14253 3. http://gcc.gnu.org/news/alias.html
14254 4. http://gcc.gnu.org/egcs-1.1/c++features.html
14255 5. http://gcc.gnu.org/egcs-1.1/features.html
14256 6. http://gcc.gnu.org/egcs-1.1/buildstat.html
14257 7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
14258 8. http://gcc.gnu.org/egcs-1.1/caveats.html
14259 9. http://gcc.gnu.org/mirrors.html
14260 10. https://gcc.gnu.org/onlinedocs/
14261 11. mailto:gcc-help@gcc.gnu.org
14262 12. mailto:gcc@gcc.gnu.org
14263 13. https://gcc.gnu.org/lists.html
14264 14. http://www.fsf.org/
14265 15. https://gcc.gnu.org/about.html
14266 16. http://validator.w3.org/check/referer
14267 ======================================================================
14268 http://gcc.gnu.org/egcs-1.1/features.html
14269 EGCS 1.1 new features
14271 * Integrated GNU Fortran (g77) compiler and runtime library with
14272 improvements, based on g77 version 0.5.23.
14273 * Vast improvements in the C++ compiler; so many they have [1]page of
14275 * Compiler implements [2]global common subexpression elimination and
14276 global copy/constant propagation.
14277 * More major improvements in the [3]alias analysis code.
14278 * More major improvements in the exception handling code to improve
14279 performance, lower static overhead and provide the infrastructure
14280 for future improvements.
14281 * The infamous /tmp symlink race security problems have been fixed.
14282 * The regmove optimization pass has been nearly completely rewritten
14283 to improve performance of generated code.
14284 * The compiler now recomputes register usage information before local
14285 register allocation. By providing more accurate information to the
14286 priority based allocator, we get better register allocation.
14287 * The register reloading phase of the compiler optimizes spill code
14288 much better than in previous releases.
14289 * Some bad interactions between the register allocator and
14290 instruction scheduler have been fixed, resulting in much better
14291 code for certain programs. Additionally, we have tuned the
14292 scheduler in various ways to improve performance of generated code
14293 for some architectures.
14294 * The compiler's branch shortening algorithms have been significantly
14295 improved to work better on targets which align jump targets.
14296 * The compiler now supports -Os to prefer optimizing for code space
14297 over optimizing for code speed.
14298 * The compiler will now totally eliminate library calls which compute
14299 constant values. This primarily helps targets with no integer
14300 div/mul support and targets without floating point support.
14301 * The compiler now supports an extensive "--help" option.
14302 * cpplib has been greatly improved and may be suitable for limited
14304 * Memory footprint for the compiler has been significantly reduced
14305 for some pathological cases.
14306 * The time to build EGCS has been improved for certain targets
14307 (particularly the alpha and mips platforms).
14308 * Many infrastructure improvements throughout the compiler, plus the
14309 usual mountain of bugfixes and minor improvements.
14310 * Target dependent improvements:
14311 + SPARC port now includes V8 plus and V9 support as well as
14312 performance tuning for Ultra class machines. The SPARC port
14313 now uses the Haifa scheduler.
14314 + Alpha port has been tuned for the EV6 processor and has an
14315 optimized expansion of memcpy/bzero. The Alpha port now uses
14316 the Haifa scheduler.
14317 + RS6000/PowerPC: support for the Power64 architecture and AIX
14318 4.3. The RS6000/PowerPC port now uses the Haifa scheduler.
14319 + x86: Alignment of static store data and jump targets is per
14320 Intel recommendations now. Various improvements throughout the
14321 x86 port to improve performance on Pentium processors
14322 (including improved epilogue sequences for Pentium chips and
14323 backend improvements which should help register allocation on
14324 all x86 variants. Conditional move support has been fixed and
14325 enabled for PPro processors. The x86 port also better supports
14326 64bit operations now. Unixware 7, a System V Release 5 target,
14327 is now supported and SCO OpenServer targets can support GAS.
14328 + MIPS has improved multiply/multiply-add support and now
14329 includes mips16 ISA support.
14330 + M68k has many micro-optimizations and Coldfire fixes.
14331 * Core compiler is based on the GCC development tree from June 9,
14332 1998, so we have all of the [4]features found in GCC 2.8.
14335 For questions related to the use of GCC, please consult these web
14336 pages and the [5]GCC manuals. If that fails, the
14337 [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14338 web pages and the development of GCC are welcome on our developer
14339 list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
14341 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
14342 distribution of this entire article is permitted in any medium,
14343 provided this notice is preserved.
14345 These pages are [10]maintained by the GCC team. Last modified
14350 1. http://gcc.gnu.org/egcs-1.1/c++features.html
14351 2. http://gcc.gnu.org/news/gcse.html
14352 3. http://gcc.gnu.org/news/alias.html
14353 4. http://gcc.gnu.org/egcs-1.0/features-2.8.html
14354 5. https://gcc.gnu.org/onlinedocs/
14355 6. mailto:gcc-help@gcc.gnu.org
14356 7. mailto:gcc@gcc.gnu.org
14357 8. https://gcc.gnu.org/lists.html
14358 9. http://www.fsf.org/
14359 10. https://gcc.gnu.org/about.html
14360 11. http://validator.w3.org/check/referer
14361 ======================================================================
14362 http://gcc.gnu.org/egcs-1.1/caveats.html
14365 * EGCS has an integrated libstdc++, but does not have an integrated
14366 libg++. Furthermore old libg++ releases will not work with EGCS; HJ
14367 Lu has made a libg++-2.8.1.2 snapshot available which may work with
14369 Note most C++ programs only need libstdc++.
14370 * Exception handling may not work with shared libraries, particularly
14371 on alphas, hppas, rs6000/powerpc and mips based platforms.
14372 Exception handling is known to work on x86-linux platforms with
14374 * Some versions of the Linux kernel have bugs which prevent them from
14375 being compiled or from running when compiled by EGCS. See the FAQ
14376 (as shipped with EGCS 1.1) for additional information.
14377 * In general, EGCS is more rigorous about rejecting invalid C++ code
14378 or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As
14379 a result it may be necessary to fix C++ code before it will compile
14381 * G++ is also converting toward the ISO C++ standard; as a result
14382 code which was previously valid (and thus accepted by other
14383 compilers and older versions of g++) may no longer be accepted.
14384 * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x
14385 or GCC 2.8.x due to changes necessary to support thread safe
14386 exception handling.
14389 For questions related to the use of GCC, please consult these web
14390 pages and the [1]GCC manuals. If that fails, the
14391 [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14392 web pages and the development of GCC are welcome on our developer
14393 list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
14395 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
14396 distribution of this entire article is permitted in any medium,
14397 provided this notice is preserved.
14399 These pages are [6]maintained by the GCC team. Last modified
14404 1. https://gcc.gnu.org/onlinedocs/
14405 2. mailto:gcc-help@gcc.gnu.org
14406 3. mailto:gcc@gcc.gnu.org
14407 4. https://gcc.gnu.org/lists.html
14408 5. http://www.fsf.org/
14409 6. https://gcc.gnu.org/about.html
14410 7. http://validator.w3.org/check/referer
14411 ======================================================================
14412 http://gcc.gnu.org/egcs-1.0/index.html
14415 December 3, 1997: We are pleased to announce the release of EGCS 1.0.
14416 January 6, 1998: We are pleased to announce the release of EGCS 1.0.1.
14417 March 16, 1998: We are pleased to announce the release of EGCS 1.0.2.
14418 May 15, 1998 We are pleased to announce the release of EGCS 1.0.3.
14420 EGCS is a collaborative effort involving several groups of hackers
14421 using an open development model to accelerate development and testing
14422 of GNU compilers and runtime libraries.
14424 An important goal of EGCS is to allow wide scale testing of
14425 experimental features and optimizations; therefore, EGCS contains some
14426 features and optimizations which are still under development. However,
14427 EGCS has been carefully tested and should be comparable in quality to
14430 EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8
14431 development sources; it contains nearly all of the new features found
14434 EGCS 1.0 also contains many improvements and features not found in GCC
14435 2.7 and even the GCC 2.8 series (which was released after the original
14437 * Integrated C++ runtime libraries, including support for most major
14439 * The integrated libstdc++ library includes a verbatim copy of SGI's
14441 * Integrated GNU Fortran compiler.
14442 * New instruction scheduler.
14443 * New alias analysis code.
14445 See the [1]new features page for a more complete list of new features.
14447 EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few
14448 critical bugs and add support for Red Hat 5.0 Linux. Changes since the
14450 * Add support for Red Hat 5.0 Linux and better support for Linux
14451 systems using glibc2.
14452 Many programs failed to link when compiled with EGCS 1.0 on Red Hat
14453 5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should
14454 fix these problems.
14455 * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception
14456 handling interfaces.
14457 To avoid future compatibility problems, we strongly urge anyone who
14458 is planning on distributing shared libraries that contain C++ code
14459 to upgrade to EGCS 1.0.1 first.
14460 Soon after EGCS 1.0 was released, the GCC developers made some
14461 incompatible changes in libgcc's exception handling interfaces.
14462 These changes were needed to solve problems on some platforms. This
14463 means that GCC 2.8.0, when released, will not be seamlessly
14464 compatible with shared libraries built by EGCS 1.0. The reason is
14465 that the libgcc.a in GCC 2.8.0 will not contain a function needed
14466 by the old interface.
14467 The result of this is that there may be compatibility problems with
14468 shared libraries built by EGCS 1.0 when used with GCC 2.8.0.
14469 With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface,
14470 and libgcc.a has the support routines for both the old and the new
14471 interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed,
14472 and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed).
14473 The maintainers of GCC 2.x have decided against including seamless
14474 support for the old interface in 2.8.0, since it was never
14475 "official", so to avoid future compatibility problems we recommend
14476 against distributing any shared libraries built by EGCS 1.0 that
14477 contain C++ code (upgrade to 1.0.1 and use that).
14478 * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends.
14479 The x86 changes fix code generation errors exposed when building
14480 glibc2 and the usual GNU/Linux dynamic linker (ld.so).
14481 The hppa change fixes a compiler abort when configured for use with
14483 The MIPS changes fix problems with the definition of LONG_MAX on
14484 newer systems, allow for command line selection of the target ABI,
14485 and fix one code generation problem.
14486 The rs6000/ppc change fixes some problems with passing structures
14487 to varargs/stdarg functions.
14488 * A few machine independent bugfixes, mostly to fix code generation
14489 errors when building Linux kernels or glibc.
14490 * Fix a few critical exception handling and template bugs in the C++
14492 * Fix Fortran namelist bug on alphas.
14493 * Fix build problems on x86-solaris systems.
14495 EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several
14496 serious problems in EGCS 1.0.1.
14497 * General improvements and fixes
14498 + Memory consumption significantly reduced, especially for
14499 templates and inline functions.
14500 + Fix various problems with glibc2.1.
14501 + Fix loop optimization bug exposed by rs6000/ppc port.
14502 + Fix to avoid potential code generation problems in jump.c.
14503 + Fix some undefined symbol problems in dwarf1 debug support.
14504 * g++/libstdc++ improvements and fixes
14505 + libstdc++ in the EGCS release has been updated and should be
14506 link compatible with libstdc++-2.8.
14507 + Various fixes in libio/libstdc++ to work better on GNU/Linux
14509 + Fix problems with duplicate symbols on systems that do not
14510 support weak symbols.
14511 + Memory corruption bug and undefined symbols in bastring have
14513 + Various exception handling fixes.
14514 + Fix compiler abort for very long thunk names.
14515 * g77 improvements and fixes
14516 + Fix compiler crash for omitted bound in Fortran CASE
14518 + Add missing entries to g77 lang-options.
14519 + Fix problem with -fpedantic in the g77 compiler.
14520 + Fix "backspace" problem with g77 on alphas.
14521 + Fix x86 backend problem with Fortran literals and -fpic.
14522 + Fix some of the problems with negative subscripts for g77 on
14524 + Fixes for Fortran builds on cygwin32/mingw32.
14525 * platform specific improvements and fixes
14526 + Fix long double problems on x86 (exposed by glibc).
14527 + x86 ports define i386 again to keep imake happy.
14528 + Fix exception handling support on NetBSD ports.
14529 + Several changes to collect2 to fix many problems with AIX.
14530 + Define __ELF__ for GNU/Linux on rs6000.
14531 + Fix -mcall-linux problem on GNU/Linux on rs6000.
14532 + Fix stdarg/vararg problem for GNU/Linux on rs6000.
14533 + Allow autoconf to select a proper install problem on AIX 3.1.
14534 + m68k port support includes -mcpu32 option as well as cpu32
14536 + Fix stdarg bug for irix6.
14537 + Allow EGCS to build on irix5 without the gnu assembler.
14538 + Fix problem with static linking on sco5.
14539 + Fix bootstrap on sco5 with native compiler.
14540 + Fix for abort building newlib on H8 target.
14541 + Fix fixincludes handling of math.h on SunOS.
14542 + Minor fix for Motorola 3300 m68k systems.
14544 EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few
14545 problems reported by Red Hat for builds of Red Hat 5.1.
14546 * Generic bugfixes:
14547 + Fix a typo in the libio library which resulted in incorrect
14548 behavior of istream::get.
14549 + Fix the Fortran negative array index problem.
14550 + Fix a major problem with the ObjC runtime thread support
14552 + Reduce memory consumption of the Haifa scheduler.
14553 * Target specific bugfixes:
14554 + Fix one x86 floating point code generation bug exposed by
14556 + Fix one x86 internal compiler error exposed by glibc2 builds.
14557 + Fix profiling bugs on the Alpha.
14558 + Fix ImageMagick & emacs 20.2 build problems on the Alpha.
14559 + Fix rs6000/ppc bug when converting values from integer types
14560 to floating point types.
14562 The EGCS 1.0 releases include installation instructions in both HTML
14563 and plaintext forms (see the INSTALL directory in the toplevel
14564 directory of the distribution). However, we also keep the most up to
14565 date installation instructions and [2]build/test status on our web
14566 page. We will update those pages as new information becomes available.
14568 And, we can't in good conscience fail to mention some [3]caveats to
14571 Update: Big thanks to Stanford for providing a high speed link for
14572 downloading EGCS (go.cygnus.com)!
14574 Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com
14575 (USA California -- High speed link provided by Stanford).
14577 The EGCS 1.0 release is also available many mirror sites.
14578 [4]Goto mirror list to find a closer site
14580 We'd like to thank the numerous people that have contributed new
14581 features, test results, bugfixes, etc. Unfortunately, they're far too
14582 numerous to mention by name.
14585 For questions related to the use of GCC, please consult these web
14586 pages and the [5]GCC manuals. If that fails, the
14587 [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14588 web pages and the development of GCC are welcome on our developer
14589 list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
14591 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
14592 distribution of this entire article is permitted in any medium,
14593 provided this notice is preserved.
14595 These pages are [10]maintained by the GCC team. Last modified
14600 1. http://gcc.gnu.org/egcs-1.0/features.html
14601 2. http://gcc.gnu.org/egcs-1.0/buildstat.html
14602 3. http://gcc.gnu.org/egcs-1.0/caveats.html
14603 4. http://gcc.gnu.org/mirrors.html
14604 5. https://gcc.gnu.org/onlinedocs/
14605 6. mailto:gcc-help@gcc.gnu.org
14606 7. mailto:gcc@gcc.gnu.org
14607 8. https://gcc.gnu.org/lists.html
14608 9. http://www.fsf.org/
14609 10. https://gcc.gnu.org/about.html
14610 11. http://validator.w3.org/check/referer
14611 ======================================================================
14612 http://gcc.gnu.org/egcs-1.0/features.html
14615 * Core compiler is based on the gcc2 development tree from Aug 2,
14616 1997, so we have most of the [1]features found in GCC 2.8.
14617 * Integrated GNU Fortran compiler based on g77-0.5.22-19970929.
14618 * Vast improvements in the C++ compiler; so many they have [2]page of
14620 * Integrated C++ runtime libraries, including support for most major
14622 * New instruction scheduler from IBM Haifa which includes support for
14623 function wide instruction scheduling as well as superscalar
14625 * Significantly improved alias analysis code.
14626 * Improved register allocation for two address machines.
14627 * Significant code generation improvements for Fortran code on
14629 * Various optimizations from the g77 project as well as improved loop
14631 * Dwarf2 debug format support for some targets.
14632 * egcs libstdc++ includes the SGI STL implementation without changes.
14633 * As a result of these and other changes, egcs libstc++ is not binary
14634 compatible with previous releases of libstdc++.
14635 * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO
14636 Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and
14637 1.1), Support for RTEMS on several embedded targets, Support for
14638 arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and
14639 MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc.
14640 * Integrated testsuites for gcc, g++, g77, libstdc++ and libio.
14641 * RS6000/PowerPC ports generate code which can run on all
14642 RS6000/PowerPC variants by default.
14643 * -mcpu= and -march= switches for the x86 port to allow better
14644 control over how the x86 port generates code.
14645 * Includes the template repository patch (aka repo patch); note the
14646 new template code makes repo obsolete for ELF systems using gnu-ld
14648 * Plus the usual assortment of bugfixes and improvements.
14651 For questions related to the use of GCC, please consult these web
14652 pages and the [3]GCC manuals. If that fails, the
14653 [4]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14654 web pages and the development of GCC are welcome on our developer
14655 list at [5]gcc@gcc.gnu.org. All of [6]our lists have public archives.
14657 Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and
14658 distribution of this entire article is permitted in any medium,
14659 provided this notice is preserved.
14661 These pages are [8]maintained by the GCC team. Last modified
14666 1. http://gcc.gnu.org/egcs-1.0/features-2.8.html
14667 2. http://gcc.gnu.org/egcs-1.0/c++features.html
14668 3. https://gcc.gnu.org/onlinedocs/
14669 4. mailto:gcc-help@gcc.gnu.org
14670 5. mailto:gcc@gcc.gnu.org
14671 6. https://gcc.gnu.org/lists.html
14672 7. http://www.fsf.org/
14673 8. https://gcc.gnu.org/about.html
14674 9. http://validator.w3.org/check/referer
14675 ======================================================================
14676 http://gcc.gnu.org/egcs-1.0/caveats.html
14679 * EGCS has an integrated libstdc++, but does not have an integrated
14680 libg++. Furthermore old libg++ releases will not work with egc; HJ
14681 Lu has made a libg++-2.8.1.2 available which may work with EGCS.
14682 Note most C++ programs only need libstdc++.
14683 * Note that using -pedantic or -Wreturn-type can cause an explosion
14684 in the amount of memory needed for template-heavy C++ code, such as
14685 code that uses STL. Also note that -Wall includes -Wreturn-type, so
14686 if you use -Wall you will need to specify -Wno-return-type to turn
14688 * Exception handling may not work with shared libraries, particularly
14689 on alphas, hppas, and mips based platforms. Exception handling is
14690 known to work on x86-linux platforms with shared libraries.
14691 * Some versions of the Linux kernel have bugs which prevent them from
14692 being compiled or from running when compiled by EGCS. See the FAQ
14693 (as shipped with EGCS 1.0) for additional information.
14694 * In general, EGCS is more rigorous about rejecting invalid C++ code
14695 or deprecated C++ constructs than G++ 2.7. As a result it may be
14696 necessary to fix C++ code before it will compile with EGCS.
14697 * G++ is also aggressively tracking the C++ standard; as a result
14698 code which was previously valid (and thus accepted by other
14699 compilers and older versions of G++) may no longer be accepted.
14700 * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS
14701 1.0.x and later releases should work with Red Hat Linux 5.0.
14704 For questions related to the use of GCC, please consult these web
14705 pages and the [1]GCC manuals. If that fails, the
14706 [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14707 web pages and the development of GCC are welcome on our developer
14708 list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
14710 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
14711 distribution of this entire article is permitted in any medium,
14712 provided this notice is preserved.
14714 These pages are [6]maintained by the GCC team. Last modified
14719 1. https://gcc.gnu.org/onlinedocs/
14720 2. mailto:gcc-help@gcc.gnu.org
14721 3. mailto:gcc@gcc.gnu.org
14722 4. https://gcc.gnu.org/lists.html
14723 5. http://www.fsf.org/
14724 6. https://gcc.gnu.org/about.html
14725 7. http://validator.w3.org/check/referer
14726 ======================================================================