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 4.9.1 relative to previous releases of GCC.
22 October 30, 2014 ([2]changes, [3]documentation)
25 July 16, 2014 ([4]changes, [5]documentation)
28 April 22, 2014 ([6]changes, [7]documentation)
30 References and Acknowledgements
32 GCC used to stand for the GNU C Compiler, but since the compiler
33 supports several other languages aside from C, it now stands for the
34 GNU Compiler Collection.
36 A list of [8]successful builds is updated as new information becomes
39 The GCC developers would like to thank the numerous people that have
40 contributed new features, improvements, bug fixes, and other changes as
41 well as test results to GCC. This [9]amazing group of volunteers is
42 what makes GCC successful.
44 For additional information about GCC please refer to the [10]GCC
45 project web site or contact the [11]GCC development mailing list.
47 To obtain GCC please use [12]our mirror sites or [13]our SVN server.
50 For questions related to the use of GCC, please consult these web
51 pages and the [14]GCC manuals. If that fails, the
52 [15]gcc-help@gcc.gnu.org mailing list might help. Comments on these
53 web pages and the development of GCC are welcome on our developer
54 list at [16]gcc@gcc.gnu.org. All of [17]our lists have public
57 Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and
58 distribution of this entire article is permitted in any medium,
59 provided this notice is preserved.
61 These pages are [19]maintained by the GCC team. Last modified
66 1. http://www.gnu.org/
67 2. http://gcc.gnu.org/gcc-4.9/changes.html
68 3. http://gcc.gnu.org/onlinedocs/4.9.2/
69 4. http://gcc.gnu.org/gcc-4.9/changes.html
70 5. http://gcc.gnu.org/onlinedocs/4.9.1/
71 6. http://gcc.gnu.org/gcc-4.9/changes.html
72 7. https://gcc.gnu.org/onlinedocs/4.9.0/
73 8. http://gcc.gnu.org/gcc-4.9/buildstat.html
74 9. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
75 10. http://gcc.gnu.org/index.html
76 11. mailto:gcc@gcc.gnu.org
77 12. http://gcc.gnu.org/mirrors.html
78 13. http://gcc.gnu.org/svn.html
79 14. https://gcc.gnu.org/onlinedocs/
80 15. mailto:gcc-help@gcc.gnu.org
81 16. mailto:gcc@gcc.gnu.org
82 17. https://gcc.gnu.org/lists.html
83 18. http://www.fsf.org/
84 19. https://gcc.gnu.org/about.html
85 20. http://validator.w3.org/check/referer
86 ======================================================================
87 http://gcc.gnu.org/gcc-4.9/changes.html
88 GCC 4.9 Release Series
89 Changes, New Features, and Fixes
93 * The mudflap run time checker has been removed. The mudflap options
94 remain, but do nothing.
95 * Support for a number of older systems and recently unmaintained or
96 untested target ports of GCC has been declared obsolete in GCC 4.9.
97 Unless there is activity to revive them, the next release of GCC
98 will have their sources permanently removed.
99 The following ports for individual systems on particular
100 architectures have been obsoleted:
101 + Solaris 9 (*-*-solaris2.9). Details can be found in the
103 * On AArch64, the singleton vector types int64x1_t, uint64x1_t and
104 float64x1_t exported by arm_neon.h are defined to be the same as
105 their base types. This results in incorrect application of
106 parameter passing rules to arguments of types int64x1_t and
107 uint64x1_t, with respect to the AAPCS64 ABI specification. In
108 addition, names of C++ functions with parameters of these types
109 (including float64x1_t) are not mangled correctly. The current
110 typedef declarations also unintentionally allow implicit casting
111 between singleton vector types and their base types. These issues
112 will be resolved in a near future release. See [2]PR60825 for more
115 More information on porting to GCC 4.9 from previous versions of GCC
116 can be found in the [3]porting guide for this release.
118 General Optimizer Improvements
120 * AddressSanitizer, a fast memory error detector, is now available on
122 * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior
123 detector, has been added and can be enabled via
124 -fsanitize=undefined. Various computations will be instrumented to
125 detect undefined behavior at runtime. UndefinedBehaviorSanitizer is
126 currently available for the C and C++ languages.
127 * Link-time optimization (LTO) improvements:
128 + Type merging was rewritten. The new implementation is
129 significantly faster and uses less memory.
130 + Better partitioning algorithm resulting in less streaming
132 + Early removal of virtual methods reduces the size of object
133 files and improves link-time memory usage and compile time.
134 + Function bodies are now loaded on-demand and released early
135 improving overall memory usage at link time.
136 + C++ hidden keyed methods can now be optimized out.
137 + When using a linker plugin, compiling with the -flto option
138 now generates slim object files (.o) which only contain
139 intermediate language representation for LTO. Use
140 -ffat-lto-objects to create files which contain additionally
141 the object code. To generate static libraries suitable for LTO
142 processing, use gcc-ar and gcc-ranlib; to list symbols from a
143 slim object file use gcc-nm. (This requires that ar, ranlib
144 and nm have been compiled with plugin support.)
145 Memory usage building Firefox with debug enabled was reduced from
146 15GB to 3.5GB; link time from 1700 seconds to 350 seconds.
147 * Inter-procedural optimization improvements:
148 + New type inheritance analysis module improving
149 devirtualization. Devirtualization now takes into account
150 anonymous name-spaces and the C++11 final keyword.
151 + New speculative devirtualization pass (controlled by
152 -fdevirtualize-speculatively.
153 + Calls that were speculatively made direct are turned back to
154 indirect where direct call is not cheaper.
155 + Local aliases are introduced for symbols that are known to be
156 semantically equivalent across shared libraries improving
157 dynamic linking times.
158 * Feedback directed optimization improvements:
159 + Profiling of programs using C++ inline functions is now more
161 + New time profiling determines typical order in which functions
163 + A new function reordering pass (controlled by
164 -freorder-functions) significantly reduces startup time of
165 large applications. Until binutils support is completed, it is
166 effective only with link-time optimization.
167 + Feedback driven indirect call removal and devirtualization now
168 handle cross-module calls when link-time optimization is
171 New Languages and Language specific improvements
173 * Version 4.0 of the [4]OpenMP specification is now supported in the
174 C and C++ compilers and starting with the 4.9.1 release also in the
175 Fortran compiler. The new -fopenmp-simd option can be used to
176 enable OpenMP's SIMD directives, while ignoring other OpenMP
177 directives. The new [5]-fsimd-cost-model= option permits to tune
178 the vectorization cost model for loops annotated with OpenMP and
179 Cilk Plus simd directives; -Wopenmp-simd warns when the current
180 cost model overrides simd directives set by the user.
181 * The -Wdate-time option has been added for the C, C++ and Fortran
182 compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__
183 macros are used. Those macros might prevent bit-wise-identical
184 reproducible compilations.
188 * GNAT switched to Ada 2012 instead of Ada 2005 by default.
192 * Support for colorizing diagnostics emitted by GCC has been added.
193 The [6]-fdiagnostics-color=auto will enable it when outputting to
194 terminals, -fdiagnostics-color=always unconditionally. The
195 GCC_COLORS environment variable can be used to customize the colors
196 or disable coloring. If GCC_COLORS variable is present in the
197 environment, the default is -fdiagnostics-color=auto, otherwise
198 -fdiagnostics-color=never.
199 Sample diagnostics output:
200 $ g++ -fdiagnostics-color=always -S -Wall test.C
201 test.C: In function ‘int foo()’:
202 test.C:1:14: warning: no return statement in function returning non-void [-W
206 test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use
207 -ftemplate-depth= to increase the maximum) instantiating ‘struct X<100>’
208 template <int N> struct X { static const int value = X<N-1>::value; }; temp
211 test.C:2:46: recursively required from ‘const int X<999>::value’
212 test.C:2:46: required from ‘const int X<1000>::value’
213 test.C:2:88: required from here
215 test.C:2:46: error: incomplete type ‘X<100>’ used in nested name specifier
217 * With the new [7]#pragma GCC ivdep, the user can assert that there
218 are no loop-carried dependencies which would prevent concurrent
219 execution of consecutive iterations using SIMD (single instruction
220 multiple data) instructions.
221 * Support for [8]Cilk Plus has been added and can be enabled with the
222 -fcilkplus option. Cilk Plus is an extension to the C and C++
223 languages to support data and task parallelism. The present
224 implementation follows ABI version 1.2; all features but _Cilk_for
225 have been implemented.
229 * ISO C11 atomics (the _Atomic type specifier and qualifier and the
230 <stdatomic.h> header) are now supported.
231 * ISO C11 generic selections (_Generic keyword) are now supported.
232 * ISO C11 thread-local storage (_Thread_local, similar to GNU C
233 __thread) is now supported.
234 * ISO C11 support is now at a similar level of completeness to ISO
235 C99 support: substantially complete modulo bugs, extended
236 identifiers (supported except for corner cases when
237 -fextended-identifiers is used), floating-point issues (mainly but
238 not entirely relating to optional C99 features from Annexes F and
239 G) and the optional Annexes K (Bounds-checking interfaces) and L
241 * A new C extension __auto_type provides a subset of the
242 functionality of C++11 auto in GNU C.
246 * The G++ implementation of [9]C++1y return type deduction for normal
247 functions has been updated to conform to [10]N3638, the proposal
248 accepted into the working paper. Most notably, it adds
249 decltype(auto) for getting decltype semantics rather than the
250 template argument deduction semantics of plain auto:
253 auto i1 = f(); // int
254 decltype(auto) i2 = f(); // int&
256 * G++ supports [11]C++1y lambda capture initializers:
260 Actually, they have been accepted since GCC 4.5, but now the
261 compiler doesn't warn about them with -std=c++1y, and supports
262 parenthesized and brace-enclosed initializers as well.
263 * G++ supports [12]C++1y variable length arrays. G++ has supported
264 GNU/C99-style VLAs for a long time, but now additionally supports
265 initializers and lambda capture by reference. In C++1y mode G++
266 will complain about VLA uses that are not permitted by the draft
267 standard, such as forming a pointer to VLA type or applying sizeof
268 to a VLA variable. Note that it now appears that VLAs will not be
269 part of C++14, but will be part of a separate document and then
273 int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3
274 [&a]{ for (int i : a) { cout << i << endl; } }();
275 &a; // error, taking address of VLA
278 * G++ supports the [13]C++1y [[deprecated]] attribute modulo bugs in
279 the underlying [[gnu::deprecated]] attribute. Classes and functions
280 can be marked deprecated and a diagnostic message added:
284 #if __cplusplus > 201103
285 class [[deprecated("A is deprecated in C++14; Use B instead")]] A;
286 [[deprecated("bar is unsafe; use foo() instead")]]
292 A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead
293 int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo
296 * G++ supports [14]C++1y digit separators. Long numeric literals can
297 be subdivided with a single quote ' to enhance readability:
302 int m = 0'004'000'000;
303 int n = 0b0001'0000'0000'0000'0000'0000;
305 double x = 1.602'176'565e-19;
306 double y = 1.602'176'565e-1'9;
308 * G++ supports [15]C++1y generic (polymorphic) lambdas.
310 // a functional object that will increment any type
311 auto incr = [](auto x) { return x++; };
313 * As a GNU extension, G++ supports explicit template parameter syntax
314 for generic lambdas. This can be combined in the expected way with
315 the standard auto syntax.
317 // a functional object that will add two like-type objects
318 auto add = [] <typename T> (T a, T b) { return a + b; };
320 * G++ supports unconstrained generic functions as specified by §4.1.2
321 and §5.1.1 of [16]N3889: Concepts Lite Specification. Briefly, auto
322 may be used as a type-specifier in a parameter declaration of any
323 function declarator in order to introduce an implicit function
324 template parameter, akin to generic lambdas.
326 // the following two function declarations are equivalent
327 auto incr(auto x) { return x++; }
328 template <typename T>
329 auto incr(T x) { return x++; }
331 Runtime Library (libstdc++)
333 * [17]Improved support for C++11, including:
334 + support for <regex>;
335 + The associative containers in <map> and <set> and the
336 unordered associative containers in <unordered_map> and
337 <unordered_set> meet the allocator-aware container
339 * [18]Improved experimental support for the upcoming ISO C++
340 standard, C++14, including:
341 + fixing constexpr member functions without const;
342 + implementation of the std::exchange() utility function;
343 + addressing tuples by type;
344 + implemention of std::make_unique;
345 + implemention of std::shared_lock;
346 + making std::result_of SFINAE-friendly;
347 + adding operator() to std::integral_constant;
348 + adding user-defined literals for standard library types
349 std::basic_string, std::chrono::duration, and std::complex;
350 + adding two range overloads to non-modifying sequence oprations
351 std::equal and std::mismatch;
352 + adding IO manipulators for quoted strings;
353 + adding constexpr members to <utility>, <complex>, <chrono>,
355 + adding compile-time std::integer_sequence;
356 + adding cleaner transformation traits;
357 + making <functional>s operator functors easier to use and more
359 * An implementation of std::experimental::optional.
360 * An implementation of std::experimental::string_view.
361 * The non-standard function std::copy_exception has been deprecated
362 and will be removed in a future version. std::make_exception_ptr
363 should be used instead.
367 * Compatibility notice:
368 + Module files: The version of the module files (.mod) has been
369 incremented; additionally, module files are now compressed.
370 Fortran MODULEs compiled by earlier GCC versions have to be
371 recompiled, when they are USEd by files compiled with GCC 4.9.
372 GCC 4.9 is not able to read .mod files of earlier GCC
373 versions; attempting to do so gives an error message. Note:
374 The ABI of the produced assembler data itself has not changed:
375 object files and libraries are fully compatible with older
376 versions (except as stated below).
378 o The [19]argument passing ABI has changed for scalar dummy
379 arguments of type INTEGER, REAL, COMPLEX and LOGICAL,
380 which have both the VALUE and the OPTIONAL attributes.
381 o To support finalization the virtual table associated with
382 polymorphic variables has changed. Code containing CLASS
383 should be recompiled, including all files which define
384 derived types involved in the type definition used by
385 polymorphic variables. (Note: Due to the incremented
386 module version, trying to mix old code with new code will
387 usually give an error message.)
388 + GNU Fortran no longer deallocates allocatable variables or
389 allocatable components of variables declared in the main
390 program. Since Fortran 2008, the standard explicitly states
391 that variables declared in the Fortran main program
392 automatically have the SAVE attribute.
393 + When opening files, the close-on-exec flag is set if the
394 system supports such a feature. This is generally considered
395 good practice these days, but if there is a need to pass file
396 descriptors to child processes the parent process must now
397 remember to clear the close-on-exec flag by calling fcntl(),
398 e.g. via ISO_C_BINDING, before executing the child process.
399 * The deprecated command-line option -fno-whole-file has been
400 removed. (-fwhole-file is the default since GCC 4.6.)
401 -fwhole-file/-fno-whole-file continue to be accepted but do not
402 influence the code generation.
403 * The compiler no longer unconditionally warns about DO loops with
404 zero iterations. This warning is now controlled by the -Wzerotrips
405 option, which is implied by -Wall.
406 * The new NO_ARG_CHECK attribute of the [20]!GCC$ directive can be
407 used to disable the type-kind-rank (TKR) argument check for a dummy
408 argument. The feature is similar to ISO/IEC TS 29133:2012's
409 TYPE(*), except that it additionally also disables the rank check.
410 Variables with NO_ARG_CHECK have to be dummy arguments and may only
411 be used as argument to ISO_C_BINDING's C_LOC and as actual argument
412 to another NO_ARG_CHECK dummy argument; also the other constraints
413 of TYPE(*) apply. The dummy arguments should be declared as scalar
414 or assumed-size variable of type type(*) (recommended) – or of type
415 integer, real, complex or logical. With NO_ARG_CHECK, a pointer to
416 the data without further type or shape information is passed,
417 similar to C's void*. Note that also TS 29113's
418 type(*),dimension(..) accepts arguments of any type and rank;
419 contrary to NO_ARG_CHECK assumed-rank arguments pass an array
420 descriptor which contains the array shape and stride of the
423 + Finalization is now supported. It is currently only done for a
424 subset of those situations in which it should occur.
425 + Experimental support for scalar character components with
426 deferred length (i.e. allocatable string length) in derived
427 types has been added. (Deferred-length character variables are
428 supported since GCC 4.6.)
430 + When STOP or ERROR STOP are used to terminate the execution
431 and any exception (but inexact) is signaling, a warning is
432 printed to ERROR_UNIT, indicating which exceptions are
433 signaling. The [23]-ffpe-summary= command-line option can be
434 used to fine-tune for which exceptions the warning should be
436 + Rounding on input (READ) is now handled on systems where
437 strtod honours the rounding mode. (For output, rounding is
438 supported since GCC 4.5.) Note that for input, the compatible
439 rounding mode is handled as nearest (i.e., for a tie, rounding
440 to an even last significant [cf. IEC 60559:1989] – while
441 compatible rounds away from zero for a tie).
445 * GCC 4.9 provides a complete implementation of the Go 1.2.1 release.
447 New Targets and Target Specific Improvements
451 * The ARMv8-A crypto and CRC instructions are now supported through
452 intrinsics. These are enabled when the architecture supports these
453 and are available through the -march=armv8-a+crc and
454 -march=armv8-a+crypto options.
455 * Initial support for ILP32 has now been added to the compiler. This
456 is now available through the command-line option -mabi=ilp32.
457 Support for ILP32 is considered experimental as the ABI
458 specification is still beta.
459 * Coverage of more of the ISA including the SIMD extensions has been
460 added. The Advanced SIMD intrinsics have also been improved.
461 * The new local register allocator (LRA) is now on by default for the
463 * The REE (Redundant extension elimination) pass has now been enabled
464 by default for the AArch64 backend.
465 * Tuning for the Cortex-A53 and Cortex-A57 has been improved.
466 * Initial big.LITTLE tuning support for the combination of Cortex-A57
467 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53
469 * A number of structural changes have been made to both the ARM and
470 AArch64 backends to facilitate improved code-generation.
474 * A port for Synopsys Designware ARC has been contributed by Embecosm
479 * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been
480 disabled by default. This was found to generate better code in only
481 a small number of cases. It can be turned back on with the
482 -mneon-for-64bits option.
483 * Further support for the ARMv8-A architecture, notably implementing
484 the restriction around IT blocks in the Thumb32 instruction set has
485 been added. The -mrestrict-it option can be used with
486 -march=armv7-a or the -march=armv7ve options to make code
487 generation fully compatible with the deprecated instructions in
489 * Support has now been added for the ARMv7ve variant of the
490 architecture. This can be used by the -march=armv7ve option.
491 * The ARMv8-A crypto and CRC instructions are now supported through
492 intrinsics and are available through the -march=armv8-a+crc and
493 mfpu=crypto-neon-fp-armv8 options.
494 * LRA is now on by default for the ARM target. This can be turned off
495 using the -mno-lra option. This option is a purely transitionary
496 command-line option and will be removed in a future release. We are
497 interested in any bug reports regarding functional and performance
498 regressions with LRA.
499 * A new option -mslow-flash-data to improve performance of programs
500 fetching data on slow flash memory has now been introduced for the
501 ARMv7-M profile cores.
502 * A new option -mpic-data-is-text-relative for targets that allows
503 data segments to be relative to text segments has been added. This
504 is on by default for all targets except VxWorks RTP.
505 * A number of infrastructural changes have been made to both the ARM
506 and AArch64 backends to facilitate improved code-generation.
507 * GCC now supports Cortex-A12 and the Cortex-R7 through the
508 -mcpu=cortex-a12 and -mcpu=cortex-r7 options.
509 * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the
510 -mcpu=cortex-a57 and -mcpu=cortex-a53 options.
511 * Initial big.LITTLE tuning support for the combination of Cortex-A57
512 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53
513 option. Similar support was added for the combination of Cortex-A15
514 and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option.
515 * Further performance optimizations for the Cortex-A15 and the
516 Cortex-M4 have been added.
517 * A number of code generation improvements for Thumb2 to reduce code
518 size when compiling for the M-profile processors.
522 * -mfpmath=sse is now implied by -ffast-math on all targets where
524 * Intel AVX-512 support was added to GCC. That includes inline
525 assembly support, new registers and extending existing ones, new
526 intrinsics (covered by corresponding testsuite), and basic
527 autovectorization. AVX-512 instructions are available via the
528 following GCC switches: AVX-512 foundation instructions: -mavx512f,
529 AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and
530 reciprocal instructions: -mavx512er, AVX-512 conflict detection
531 instructions: -mavx512cd.
532 * It is now possible to call x86 intrinsics from select functions in
533 a file that are tagged with the corresponding target attribute
534 without having to compile the entire file with the -mxxx option.
535 This improves the usability of x86 intrinsics and is particularly
536 useful when doing [24]Function Multiversioning.
537 * GCC now supports the new Intel microarchitecture named Silvermont
538 through -march=silvermont.
539 * GCC now supports the new Intel microarchitecture named Broadwell
540 through -march=broadwell.
541 * Optimizing for other Intel microarchitectures have been renamed to
542 -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell.
543 * -march=generic has been retuned for better support of Intel core
544 and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel
545 Pentium-M, and Pentium4 based CPUs is no longer considered
546 important for generic.
547 * -mtune=intel can now be used to generate code running well on the
548 most current Intel processors, which are Haswell and Silvermont for
550 * Support to encode 32-bit assembly instructions in 16-bit format is
551 now available through the -m16 command-line option.
552 * Better inlining of memcpy and memset that is aware of value ranges
553 and produces shorter alignment prologues.
554 * -mno-accumulate-outgoing-args is now honored when unwind
555 information is output. Argument accumulation is also now turned off
556 for portions of programs optimized for size.
557 * Support for new AMD family 15h processors (Excavator core) is now
558 available through the -march=bdver4 and -mtune=bdver4 options.
562 * A new command-line option -mcpu= has been added to the MSP430
563 backend. This option is used to specify the ISA to be used.
564 Accepted values are msp430 (the default), msp430x and msp430xv2.
565 The ISA is no longer deduced from the -mmcu= option as there are
566 far too many different MCU names. The -mmcu= option is still
567 supported, and this is still used to select linker scripts and
568 generate a C preprocessor symbol that will be recognised by the
569 msp430.h header file.
573 * A new nds32 port supports the 32-bit architecture from Andes
574 Technology Corporation.
575 * The port provides initial support for the V2, V3, V3m instruction
580 * A port for the Altera Nios II has been contributed by Mentor
583 PowerPC / PowerPC64 / RS6000
585 * GCC now supports Power ISA 2.07, which includes support for
586 Hardware Transactional Memory (HTM), Quadword atomics and several
587 VMX and VSX additions, including Crypto, 64-bit integer, 128-bit
588 integer and decimal integer operations.
589 * Support for the POWER8 processor is now available through the
590 -mcpu=power8 and -mtune=power8 options.
591 * The libitm library has been modified to add a HTM fastpath that
592 automatically uses POWER's HTM hardware instructions when it is
593 executing on a HTM enabled processor.
594 * Support for the new powerpc64le-linux platform has been added. It
595 defaults to generating code that conforms to the ELFV2 ABI.
599 * Support for the Transactional Execution Facility included with the
600 IBM zEnterprise zEC12 processor has been added. A set of GCC style
601 builtins as well as XLC style builtins are provided. The builtins
602 are enabled by default when using the -march=zEC12 option but can
603 explicitly be disabled with -mno-htm. Using the GCC builtins also
604 libitm supports hardware transactions on S/390.
605 * The hotpatch features allows to prepare functions for hotpatching.
606 A certain amount of bytes is reserved before the function entry
607 label plus a NOP is inserted at its very beginning to implement a
608 backward jump when applying a patch. The feature can either be
609 enabled per compilation unit via the command-line option -mhotpatch
610 or per function using the hotpatch attribute.
611 * The shrink wrap optimization is now supported on S/390 and enabled
613 * A major rework of the routines to determine which registers need to
614 be saved and restored in function prologue/epilogue now allow to
615 use floating point registers as save slots. This will happen for
616 certain leaf function with -march=z10 or higher.
617 * The LRA rtl pass replaces reload by default on S/390.
621 * The port now allows to specify the RX100, RX200, and RX600
622 processors with the command-line options -mcpu=rx100, -mcpu=rx200
627 * Minor improvements to code generated for integer arithmetic and
628 code that involves the T bit.
629 * Added support for the SH2A clips and clipu instructions. The
630 compiler will now try to utilize them for min/max expressions such
631 as max (-128, min (127, x)).
632 * Added support for the cmp/str instruction through built-in
633 functions such as __builtin_strlen. When not optimizing for size,
634 the compiler will now expand calls to e.g. strlen as an inlined
635 sequences which utilize the cmp/str instruction.
636 * Improved code generated around volatile memory loads and stores.
637 * The option -mcbranchdi has been deprecated. Specifying it will
638 result in a warning and will not influence code generation.
639 * The option -mcmpeqdi has been deprecated. Specifying it will result
640 in a warning and will not influence code generation.
644 This is the [25]list of problem reports (PRs) from GCC's bug tracking
645 system that are known to be fixed in the 4.9.1 release. This list might
646 not be complete (that is, it is possible that some PRs that have been
647 fixed are not listed here).
649 Version 4.0 of the [26]OpenMP specification is supported even in
650 Fortran, not just C and C++.
654 This is the [27]list of problem reports (PRs) from GCC's bug tracking
655 system that are known to be fixed in the 4.9.2 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).
660 For questions related to the use of GCC, please consult these web
661 pages and the [28]GCC manuals. If that fails, the
662 [29]gcc-help@gcc.gnu.org mailing list might help. Comments on these
663 web pages and the development of GCC are welcome on our developer
664 list at [30]gcc@gcc.gnu.org. All of [31]our lists have public
667 Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and
668 distribution of this entire article is permitted in any medium,
669 provided this notice is preserved.
671 These pages are [33]maintained by the GCC team. Last modified
676 1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html
677 2. https://gcc.gnu.org/PR60825
678 3. https://gcc.gnu.org/gcc-4.9/porting_to.html
679 4. http://openmp.org/wp/openmp-specifications/
680 5. https://gcc.gnu.org/onlinedocs/gcc-4.9.0/gcc/Optimize-Options.html#index-fsimd-cost-model-908
681 6. https://gcc.gnu.org/onlinedocs/gcc-4.9.0/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252
682 7. https://gcc.gnu.org/onlinedocs/gcc-4.9.0/gcc/Loop-Specific-Pragmas.html
683 8. https://www.cilkplus.org/
684 9. http://gcc.gnu.org/projects/cxx1y.html
685 10. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html
686 11. http://gcc.gnu.org/projects/cxx1y.html
687 12. http://gcc.gnu.org/projects/cxx1y.html
688 13. http://gcc.gnu.org/projects/cxx1y.html
689 14. http://gcc.gnu.org/projects/cxx1y.html
690 15. http://gcc.gnu.org/projects/cxx1y.html
691 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf
692 17. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2011
693 18. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014
694 19. https://gcc.gnu.org/onlinedocs/gcc-4.9.0/gfortran/Argument-passing-conventions.html
695 20. https://gcc.gnu.org/onlinedocs/gcc-4.9.0/gfortran/GNU-Fortran-Compiler-Directives.html
696 21. https://gcc.gnu.org/wiki/Fortran2003Status
697 22. https://gcc.gnu.org/wiki/Fortran2008Status
698 23. https://gcc.gnu.org/onlinedocs/gcc-4.9.0/gfortran/Debugging-Options.html
699 24. https://gcc.gnu.org/onlinedocs/gcc-4.9.0/gcc/Function-Multiversioning.html
700 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1
701 26. http://openmp.org/wp/openmp-specifications/
702 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2
703 28. https://gcc.gnu.org/onlinedocs/
704 29. mailto:gcc-help@gcc.gnu.org
705 30. mailto:gcc@gcc.gnu.org
706 31. https://gcc.gnu.org/lists.html
707 32. http://www.fsf.org/
708 33. https://gcc.gnu.org/about.html
709 34. http://validator.w3.org/check/referer
710 ======================================================================
711 http://gcc.gnu.org/gcc-4.8/index.html
712 GCC 4.8 Release Series
716 The [1]GNU project and the GCC developers are pleased to announce the
717 release of GCC 4.8.3.
719 This release is a bug-fix release, containing fixes for regressions in
720 GCC 4.8.2 relative to previous releases of GCC.
725 May 22, 2014 ([2]changes, [3]documentation)
728 October 16, 2013 ([4]changes, [5]documentation)
731 May 31, 2013 ([6]changes, [7]documentation)
734 March 22, 2013 ([8]changes, [9]documentation)
736 References and Acknowledgements
738 GCC used to stand for the GNU C Compiler, but since the compiler
739 supports several other languages aside from C, it now stands for the
740 GNU Compiler Collection.
742 A list of [10]successful builds is updated as new information becomes
745 The GCC developers would like to thank the numerous people that have
746 contributed new features, improvements, bug fixes, and other changes as
747 well as test results to GCC. This [11]amazing group of volunteers is
748 what makes GCC successful.
750 For additional information about GCC please refer to the [12]GCC
751 project web site or contact the [13]GCC development mailing list.
753 To obtain GCC please use [14]our mirror sites or [15]our SVN server.
756 For questions related to the use of GCC, please consult these web
757 pages and the [16]GCC manuals. If that fails, the
758 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
759 web pages and the development of GCC are welcome on our developer
760 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
763 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
764 distribution of this entire article is permitted in any medium,
765 provided this notice is preserved.
767 These pages are [21]maintained by the GCC team. Last modified
772 1. http://www.gnu.org/
773 2. http://gcc.gnu.org/gcc-4.8/changes.html
774 3. https://gcc.gnu.org/onlinedocs/4.8.3/
775 4. http://gcc.gnu.org/gcc-4.8/changes.html
776 5. https://gcc.gnu.org/onlinedocs/4.8.2/
777 6. http://gcc.gnu.org/gcc-4.8/changes.html
778 7. https://gcc.gnu.org/onlinedocs/4.8.1/
779 8. http://gcc.gnu.org/gcc-4.8/changes.html
780 9. https://gcc.gnu.org/onlinedocs/4.8.0/
781 10. http://gcc.gnu.org/gcc-4.8/buildstat.html
782 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
783 12. http://gcc.gnu.org/index.html
784 13. mailto:gcc@gcc.gnu.org
785 14. http://gcc.gnu.org/mirrors.html
786 15. http://gcc.gnu.org/svn.html
787 16. https://gcc.gnu.org/onlinedocs/
788 17. mailto:gcc-help@gcc.gnu.org
789 18. mailto:gcc@gcc.gnu.org
790 19. https://gcc.gnu.org/lists.html
791 20. http://www.fsf.org/
792 21. https://gcc.gnu.org/about.html
793 22. http://validator.w3.org/check/referer
794 ======================================================================
795 http://gcc.gnu.org/gcc-4.8/changes.html
796 GCC 4.8 Release Series
797 Changes, New Features, and Fixes
801 GCC now uses C++ as its implementation language. This means that to
802 build GCC from sources, you will need a C++ compiler that understands
803 C++ 2003. For more details on the rationale and specific changes,
804 please refer to the [1]C++ conversion page.
806 To enable the Graphite framework for loop optimizations you now need
807 CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from
808 the [2]GCC infrastructure directory. The installation manual contains
809 more information about requirements to build GCC.
811 GCC now uses a more aggressive analysis to derive an upper bound for
812 the number of iterations of loops using constraints imposed by language
813 standards. This may cause non-conforming programs to no longer work as
814 expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new
815 option, -fno-aggressive-loop-optimizations, was added to disable this
816 aggressive analysis. In some loops that have known constant number of
817 iterations, but undefined behavior is known to occur in the loop before
818 reaching or during the last iteration, GCC will warn about the
819 undefined behavior in the loop instead of deriving lower upper bound of
820 the number of iterations for the loop. The warning can be disabled with
821 -Wno-aggressive-loop-optimizations.
823 On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules
824 for the layout of vectors that could lead to wrong code being
825 generated. Vectors larger than 8 bytes in size are now by default
826 aligned to an 8-byte boundary. This is an ABI change: code that makes
827 explicit use of vector types may be incompatible with binary objects
828 built with older versions of GCC. Auto-vectorized code is not affected
831 On AVR, support has been removed for the command-line option
832 -mshort-calls deprecated in GCC 4.7.
834 On AVR, the configure option --with-avrlibc supported since GCC 4.7.2
835 is turned on per default for all non-RTEMS configurations. This option
836 arranges for a better integration of [3]AVR Libc with avr-gcc. For
837 technical details, see [4]PR54461. To turn off the option in non-RTEMS
838 configurations, use --with-avrlibc=no. If the compiler is configured
839 for RTEMS, the option is always turned off.
841 More information on porting to GCC 4.8 from previous versions of GCC
842 can be found in the [5]porting guide for this release.
844 General Optimizer Improvements (and Changes)
846 * DWARF4 is now the default when generating DWARF debug information.
847 When -g is used on a platform that uses DWARF debugging
848 information, GCC will now default to -gdwarf-4
849 -fno-debug-types-section.
850 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information
851 consumers support DWARF4 by default. Before GCC 4.8 the default
852 version used was DWARF2. To make GCC 4.8 generate an older DWARF
853 version use -g together with -gdwarf-2 or -gdwarf-3. The default
854 for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf.
855 * A new general optimization level, -Og, has been introduced. It
856 addresses the need for fast compilation and a superior debugging
857 experience while providing a reasonable level of run-time
858 performance. Overall experience for development should be better
859 than the default optimization level -O0.
860 * A new option -ftree-partial-pre was added to control the partial
861 redundancy elimination (PRE) optimization. This option is enabled
862 by default at the -O3 optimization level, and it makes PRE more
864 * The option -fconserve-space has been removed; it was no longer
865 useful on most targets since GCC supports putting variables into
866 BSS without making them common.
867 * The struct reorg and matrix reorg optimizations (command-line
868 options -fipa-struct-reorg and -fipa-matrix-reorg) have been
869 removed. They did not always work correctly, nor did they work with
870 link-time optimization (LTO), hence were only applicable to
871 programs consisting of a single translation unit.
872 * Several scalability bottle-necks have been removed from GCC's
873 optimization passes. Compilation of extremely large functions, e.g.
874 due to the use of the flatten attribute in the "Eigen" C++ linear
875 algebra templates library, is significantly faster than previous
877 * Link-time optimization (LTO) improvements:
878 + LTO partitioning has been rewritten for better reliability and
879 maintanibility. Several important bugs leading to link
880 failures have been fixed.
881 * Interprocedural optimization improvements:
882 + A new symbol table has been implemented. It builds on existing
883 callgraph and varpool modules and provide a new API. Unusual
884 symbol visibilities and aliases are handled more consistently
885 leading to, for example, more aggressive unreachable code
887 + The inline heuristic can now bypass limits on the size of of
888 inlined functions when the inlining is particularly
889 profitable. This happens, for example, when loop bounds or
890 array strides get propagated.
891 + Values passed through aggregates (either by value or
892 reference) are now propagated at the inter-procedural level
893 leading to better inlining decisions (for example in the case
894 of Fortran array descriptors) and devirtualization.
895 * [6]AddressSanitizer , a fast memory error detector, has been added
896 and can be enabled via -fsanitize=address. Memory access
897 instructions will be instrumented to detect heap-, stack-, and
898 global-buffer overflow as well as use-after-free bugs. To get nicer
899 stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is
900 available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on
902 * [7]ThreadSanitizer has been added and can be enabled via
903 -fsanitize=thread. Instructions will be instrumented to detect data
904 races. The ThreadSanitizer is available on x86-64 GNU/Linux.
905 * A new local register allocator (LRA) has been implemented, which
906 replaces the 26 year old reload pass and improves generated code
907 quality. For now it is active on the IA-32 and x86-64 targets.
908 * Support for transactional memory has been implemented on the
909 following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and
912 New Languages and Language specific improvements
916 * Each diagnostic emitted now includes the original source line and a
917 caret '^' indicating the column. The option
918 -fno-diagnostics-show-caret suppresses this information.
919 * The option -ftrack-macro-expansion=2 is now enabled by default.
920 This allows the compiler to display the macro expansion stack in
921 diagnostics. Combined with the caret information, an example
922 diagnostic showing these two features is:
924 t.c:1:94: error: invalid operands to binary < (have ‘struct mystruct’ and ‘float
926 #define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _
927 _b = (B); __a < __b ? __b : __a; })
930 t.c:7:7: note: in expansion of macro 'MYMAX'
934 * A new -Wsizeof-pointer-memaccess warning has been added (also
935 enabled by -Wall) to warn about suspicious length parameters to
936 certain string and memory built-in functions if the argument uses
937 sizeof. This warning warns e.g. about memset (ptr, 0, sizeof
938 (ptr)); if ptr is not an array, but a pointer, and suggests a
939 possible fix, or about memcpy (&foo, ptr, sizeof (&foo));.
940 * The new option -Wpedantic is an alias for -pedantic, which is now
941 deprecated. The forms -Wno-pedantic, -Werror=pedantic, and
942 -Wno-error=pedantic work in the same way as for any other -W
943 option. One caveat is that -Werror=pedantic is not equivalent to
944 -pedantic-errors, since the latter makes into errors some warnings
945 that are not controlled by -Wpedantic, and the former only affects
946 diagnostics that are disabled when using -Wno-pedantic.
947 * The option -Wshadow no longer warns if a declaration shadows a
948 function declaration, unless the former declares a function or
949 pointer to function, because this is [8]a common and valid case in
954 * G++ now implements the [9]C++11 thread_local keyword; this differs
955 from the GNU __thread keyword primarily in that it allows dynamic
956 initialization and destruction semantics. Unfortunately, this
957 support requires a run-time penalty for references to
958 non-function-local thread_local variables defined in a different
959 translation unit even if they don't need dynamic initialization, so
960 users may want to continue to use __thread for TLS variables with
961 static initialization semantics.
962 If the programmer can be sure that no use of the variable in a
963 non-defining TU needs to trigger dynamic initialization (either
964 because the variable is statically initialized, or a use of the
965 variable in the defining TU will be executed before any uses in
966 another TU), they can avoid this overhead with the
967 -fno-extern-tls-init option.
968 OpenMP threadprivate variables now also support dynamic
969 initialization and destruction by the same mechanism.
970 * G++ now implements the [10]C++11 attribute syntax, e.g.
972 [[noreturn]] void f();
974 and also the alignment specifier, e.g.
976 alignas(double) int i;
978 * G++ now implements [11]C++11 inheriting constructors, e.g.
980 struct A { A(int); };
981 struct B: A { using A::A; }; // defines B::B(int)
984 * As of GCC 4.8.1, G++ implements the change to decltype semantics
988 decltype(f()) g(); // OK, return type of f() is not required to be complete.
990 * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g.
992 struct A { int f() &; };
993 int i = A().f(); // error, f() requires an lvalue object
995 * G++ now supports a -std=c++1y option for experimentation with
996 features proposed for the next revision of the standard, expected
997 around 2014. Currently the only difference from -std=c++11 is
998 support for return type deduction in normal functions, as proposed
999 in [14]N3386. Status of C++1y features in GCC 4.8 can be found
1001 * The G++ namespace association extension, __attribute ((strong)),
1002 has been deprecated. Inline namespaces should be used instead.
1003 * G++ now supports a -fext-numeric-literal option to control whether
1004 GNU numeric literal suffixes are accepted as extensions or
1005 processed as C++11 user-defined numeric literal suffixes. The flag
1006 is on (use suffixes for GNU literals) by default for -std=gnu++*,
1007 and -std=c++98. The flag is off (use suffixes for user-defined
1008 literals) by default for -std=c++11 and later.
1010 Runtime Library (libstdc++)
1012 * [16]Improved experimental support for the new ISO C++ standard,
1014 + forward_list meets the allocator-aware container requirements;
1015 + this_thread::sleep_for(), this_thread::sleep_until() and
1016 this_thread::yield() are defined without requiring the
1017 configure option --enable-libstdcxx-time;
1018 * Improvements to <random>:
1019 + SSE optimized normal_distribution.
1020 + Use of hardware RNG instruction for random_device on new x86
1021 processors (requires the assembler to support the
1024 + New random number engine simd_fast_mersenne_twister_engine
1025 with an optimized SSE implementation.
1026 + New random number distributions beta_distribution,
1027 normal_mv_distribution, rice_distribution,
1028 nakagami_distribution, pareto_distribution, k_distribution,
1029 arcsine_distribution, hoyt_distribution.
1030 * Added --disable-libstdcxx-verbose configure option to disable
1031 diagnostic messages issued when a process terminates abnormally.
1032 This may be useful for embedded systems to reduce the size of
1033 executables that link statically to the library.
1037 * Compatibility notice:
1038 + Module files: The version of module files (.mod) has been
1039 incremented. Fortran MODULEs compiled by earlier GCC versions
1040 have to be recompiled, when they are USEd by files compiled
1041 with GCC 4.8. GCC 4.8 is not able to read .mod files created
1042 by earlier versions; attempting to do so gives an error
1044 Note: The ABI of the produced assembler data itself has not
1045 changed; object files and libraries are fully compatible with
1046 older versions except as noted below.
1047 + ABI: Some internal names (used in the assembler/object file)
1048 have changed for symbols declared in the specification part of
1049 a module. If an affected module – or a file using it via use
1050 association – is recompiled, the module and all files which
1051 directly use such symbols have to be recompiled as well. This
1052 change only affects the following kind of module symbols:
1053 o Procedure pointers. Note: C-interoperable function
1054 pointers (type(c_funptr)) are not affected nor are
1055 procedure-pointer components.
1056 o Deferred-length character strings.
1057 * The [17]BACKTRACE intrinsic subroutine has been added. It shows a
1058 backtrace at an arbitrary place in user code; program execution
1059 continues normally afterwards.
1060 * The [18]-Wc-binding-type warning option has been added (disabled by
1061 default). It warns if the a variable might not be C interoperable;
1062 in particular, if the variable has been declared using an intrinsic
1063 type with default kind instead of using a kind parameter defined
1064 for C interoperability in the intrinsic ISO_C_Binding module.
1065 Before, this warning was always printed. The -Wc-binding-type
1066 option is enabled by -Wall.
1067 * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line
1068 options have been added, which diagnose when code to is inserted
1069 for automatic (re)allocation of a variable during assignment. This
1070 option can be used to decide whether it is safe to use
1071 [20]-fno-realloc-lhs. Additionally, it can be used to find
1072 automatic (re)allocation in hot loops. (For arrays, replacing
1073 "var=" by "var(:)=" disables the automatic reallocation.)
1074 * The [21]-Wcompare-reals command-line option has been added. When
1075 this is set, warnings are issued when comparing REAL or COMPLEX
1076 types for equality and inequality; consider replacing a == b by
1077 abs(a−b) < eps with a suitable eps. -Wcompare-reals is enabled by
1079 * The [22]-Wtarget-lifetime command-line option has been added
1080 (enabled with -Wall), which warns if the pointer in a pointer
1081 assignment might outlive its target.
1082 * Reading floating point numbers which use "q" for the exponential
1083 (such as 4.0q0) is now supported as vendor extension for better
1084 compatibility with old data files. It is strongly recommended to
1085 use for I/O the equivalent but standard conforming "e" (such as
1087 (For Fortran source code, consider replacing the "q" in
1088 floating-point literals by a kind parameter (e.g. 4.0e0_qp with a
1089 suitable qp). Note that – in Fortran source code – replacing "q" by
1090 a simple "e" is not equivalent.)
1091 * The GFORTRAN_TMPDIR environment variable for specifying a
1092 non-default directory for files opened with STATUS="SCRATCH", is
1093 not used anymore. Instead gfortran checks the POSIX/GNU standard
1094 TMPDIR environment variable. If TMPDIR is not defined, gfortran
1095 falls back to other methods to determine the directory for
1096 temporary files as documented in the [23]user manual.
1098 + Support for unlimited polymorphic variables (CLASS(*)) has
1099 been added. Nonconstant character lengths are not yet
1102 + Assumed types (TYPE(*)) are now supported.
1103 + Experimental support for assumed-rank arrays (dimension(..))
1104 has been added. Note that currently gfortran's own array
1105 descriptor is used, which is different from the one defined in
1106 TS29113, see [26]gfortran's header file or use the [27]Chasm
1107 Language Interoperability Tools.
1111 * GCC 4.8.2 provides a complete implementation of the Go 1.1.2
1113 * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1
1114 release. The library support is not quite complete.
1115 * Go has been tested on GNU/Linux and Solaris platforms for various
1116 processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may
1117 work on other platforms as well.
1119 New Targets and Target Specific Improvements
1123 * A new port has been added to support AArch64, the new 64-bit
1124 architecture from ARM. Note that this is a separate port from the
1125 existing 32-bit ARM port.
1126 * The port provides initial support for the Cortex-A53 and the
1127 Cortex-A57 processors with the command line options
1128 -mcpu=cortex-a53 and -mcpu=cortex-a57.
1132 * Initial support has been added for the AArch32 extensions defined
1133 in the ARMv8 architecture.
1134 * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs.
1135 * A new option, -mcpu=marvell-pj4, has been added to generate code
1136 for the Marvell PJ4 processor.
1137 * The compiler can now automatically generate the VFMA, VFMS, REVSH
1138 and REV16 instructions.
1139 * A new vectorizer cost model for Advanced SIMD configurations to
1140 improve the auto-vectorization strategies used.
1141 * The scheduler now takes into account the number of live registers
1142 to reduce the amount of spilling that can occur. This should
1143 improve code performance in large functions. The limit can be
1144 removed by using the option -fno-sched-pressure.
1145 * Improvements have been made to the Marvell iWMMX code generation
1146 and support for the iWMMX2 SIMD unit has been added. The option
1147 -mcpu=iwmmxt2 can be used to enable code generation for the latter.
1148 * A number of code generation improvements for Thumb2 to reduce code
1149 size when compiling for the M-profile processors.
1150 * The RTEMS (arm-rtems) port has been updated to use the EABI.
1151 * Code generation support for the old FPA and Maverick floating-point
1152 architectures has been removed. Ports that previously relied on
1153 these features have also been removed. This includes the targets:
1154 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
1155 + arm*-*-elf (use arm*-*-eabi)
1156 + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
1157 + arm*-*-ecos-elf (no alternative)
1158 + arm*-*-freebsd (no alternative)
1159 + arm*-wince-pe* (no alternative).
1163 * Support for the "Embedded C" fixed-point has been added. For
1164 details, see the [28]GCC wiki and the [29]user manual. The support
1166 * A new print modifier %r for register operands in inline assembler
1167 is supported. It will print the raw register number without the
1168 register prefix 'r':
1169 /* Return the most significant byte of 'val', a 64-bit value. */
1171 unsigned char msb (long long val)
1174 __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val));
1177 The inline assembler in this example will generate code like
1179 provided c is allocated to R24 and val is allocated to R8…R15. This
1180 works because the GNU assembler accepts plain register numbers
1181 without register prefix.
1182 * Static initializers with 3-byte symbols are supported now:
1183 extern const __memx char foo;
1184 const __memx void *pfoo = &foo;
1185 This requires at least Binutils 2.23.
1189 * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with
1190 SSE extensions disabled. Since the x86-64 ABI requires 16 byte
1191 stack alignment, this is ABI incompatible and intended to be used
1192 in controlled environments where stack space is an important
1193 limitation. This option will lead to wrong code when functions
1194 compiled with 16 byte stack alignment (such as functions from a
1195 standard library) are called with misaligned stack. In this case,
1196 SSE instructions may lead to misaligned memory access traps. In
1197 addition, variable arguments will be handled incorrectly for 16
1198 byte aligned objects (including x87 long double and __int128),
1199 leading to wrong results. You must build all modules with
1200 -mpreferred-stack-boundary=3, including any libraries. This
1201 includes the system libraries and startup modules.
1202 * Support for the new Intel processor codename Broadwell with RDSEED,
1203 ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw,
1204 -mrdseed command-line options.
1205 * Support for the Intel RTM and HLE intrinsics, built-in functions
1206 and code generation is available via -mrtm and -mhle.
1207 * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets.
1208 Intrinsics and built-in functions are available via -mfxsr, -mxsave
1209 and -mxsaveopt respectively.
1210 * New -maddress-mode=[short|long] options for x32.
1211 -maddress-mode=short overrides default 64-bit addresses to 32-bit
1212 by emitting the 0x67 address-size override prefix. This is the
1213 default address mode for x32.
1214 * New built-in functions to detect run-time CPU type and ISA:
1215 + A built-in function __builtin_cpu_is has been added to detect
1216 if the run-time CPU is of a particular type. It returns a
1217 positive integer on a match and zero otherwise. It accepts one
1218 string literal argument, the CPU name. For example,
1219 __builtin_cpu_is("westmere") returns a positive integer if the
1220 run-time CPU is an Intel Core i7 Westmere processor. Please
1221 refer to the [30]user manual for the list of valid CPU names
1223 + A built-in function __builtin_cpu_supports has been added to
1224 detect if the run-time CPU supports a particular ISA feature.
1225 It returns a positive integer on a match and zero otherwise.
1226 It accepts one string literal argument, the ISA feature. For
1227 example, __builtin_cpu_supports("ssse3") returns a positive
1228 integer if the run-time CPU supports SSSE3 instructions.
1229 Please refer to the [31]user manual for the list of valid ISA
1231 Caveat: If these built-in functions are called before any static
1232 constructors are invoked, like during IFUNC initialization, then
1233 the CPU detection initialization must be explicitly run using this
1234 newly provided built-in function, __builtin_cpu_init. The
1235 initialization needs to be done only once. For example, this is how
1236 the invocation would look like inside an IFUNC initializer:
1237 static void (*some_ifunc_resolver(void))(void)
1239 __builtin_cpu_init();
1240 if (__builtin_cpu_is("amdfam10h") ...
1241 if (__builtin_cpu_supports("popcnt") ...
1244 * Function Multiversioning Support with G++:
1245 It is now possible to create multiple function versions each
1246 targeting a specific processor and/or ISA. Function versions have
1247 the same signature but different target attributes. For example,
1248 here is a program with function versions:
1249 __attribute__ ((target ("default")))
1255 __attribute__ ((target ("sse4.2")))
1264 assert ((*p)() == foo());
1268 Please refer to this [32]wiki for more information.
1269 * The x86 back end has been improved to allow option -fschedule-insns
1270 to work reliably. This option can be used to schedule instructions
1271 better and leads to improved performace in certain cases.
1272 * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437
1273 from the Mingw-w64 trunk.
1274 * Support for new AMD family 15h processors (Steamroller core) is now
1275 available through the -march=bdver3 and -mtune=bdver3 options.
1276 * Support for new AMD family 16h processors (Jaguar core) is now
1277 available through the -march=btver2 and -mtune=btver2 options.
1281 * This target now supports the -fstack-usage command-line option.
1285 * GCC can now generate code specifically for the R4700, Broadcom XLP
1286 and MIPS 34kn processors. The associated -march options are
1287 -march=r4700, -march=xlp and -march=34kn respectively.
1288 * GCC now generates better DSP code for MIPS 74k cores thanks to
1289 further scheduling optimizations.
1290 * The MIPS port now supports the -fstack-check option.
1291 * GCC now passes the -mmcu and -mno-mcu options to the assembler.
1292 * Previous versions of GCC would silently accept -fpic and -fPIC for
1293 -mno-abicalls targets like mips*-elf. This combination was not
1294 intended or supported, and did not generate position-independent
1295 code. GCC 4.8 now reports an error when this combination is used.
1297 PowerPC / PowerPC64 / RS6000
1299 * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save,
1300 restore or update the VRSAVE register by default. The respective
1301 operating systems manage the VRSAVE register directly.
1302 * Large TOC support has been added for AIX through the command line
1303 option -mcmodel=large.
1304 * Native Thread-Local Storage support has been added for AIX.
1305 * VMX (Altivec) and VSX instruction sets now are enabled implicitly
1306 when targetting processors that support those hardware features on
1311 * This target will now issue a warning message whenever multiple fast
1312 interrupt handlers are found in the same compilation unit. This
1313 feature can be turned off by the new
1314 -mno-warn-multiple-fast-interrupts command-line option.
1318 * Support for the IBM zEnterprise zEC12 processor has been added.
1319 When using the -march=zEC12 option, the compiler will generate code
1320 making use of the following new instructions:
1321 + load and trap instructions
1322 + 2 new compare and trap instructions
1323 + rotate and insert selected bits - without CC clobber
1324 The -mtune=zEC12 option enables zEC12 specific instruction
1325 scheduling without making use of new instructions.
1326 * Register pressure sensitive instruction scheduling is enabled by
1328 * The ifunc function attribute is enabled by default.
1329 * memcpy and memcmp invokations on big memory chunks or with run time
1330 lengths are not generated inline anymore when tuning for z10 or
1331 higher. The purpose is to make use of the IFUNC optimized versions
1336 * The default alignment settings have been reduced to be less
1337 aggressive. This results in more compact code for optimization
1338 levels other than -Os.
1339 * Improved support for the __atomic built-in functions:
1340 + A new option -matomic-model=model selects the model for the
1341 generated atomic sequences. The following models are
1345 Software gUSA sequences (SH3* and SH4* only). On
1346 SH4A targets this will now also partially utilize
1347 the movco.l and movli.l instructions. This is the
1348 default when the target is sh3*-*-linux* or
1352 Hardware movco.l / movli.l sequences (SH4A only).
1355 Software thread control block sequences.
1358 Software interrupt flipping sequences (privileged
1359 mode only). This is the default when the target is
1360 sh1*-*-linux* or sh2*-*-linux*.
1363 Generates function calls to the respective __atomic
1364 built-in functions. This is the default for SH64
1365 targets or when the target is not sh*-*-linux*.
1367 + The option -msoft-atomic has been deprecated. It is now an
1368 alias for -matomic-model=soft-gusa.
1369 + A new option -mtas makes the compiler generate the tas.b
1370 instruction for the __atomic_test_and_set built-in function
1371 regardless of the selected atomic model.
1372 + The __sync functions in libgcc now reflect the selected atomic
1373 model when building the toolchain.
1374 * Added support for the mov.b and mov.w instructions with
1375 displacement addressing.
1376 * Added support for the SH2A instructions movu.b and movu.w.
1377 * Various improvements to code generated for integer arithmetic.
1378 * Improvements to conditional branches and code that involves the T
1379 bit. A new option -mzdcbranch tells the compiler to favor
1380 zero-displacement branches. This is enabled by default for SH4*
1382 * The pref instruction will now be emitted by the __builtin_prefetch
1383 built-in function for SH3* targets.
1384 * The fmac instruction will now be emitted by the fmaf standard
1385 function and the __builtin_fmaf built-in function.
1386 * The -mfused-madd option has been deprecated in favor of the
1387 machine-independent -ffp-contract option. Notice that the fmac
1388 instruction will now be generated by default for expressions like a
1389 * b + c. This is due to the compiler default setting
1391 * Added new options -mfsrra and -mfsca to allow the compiler using
1392 the fsrra and fsca instructions on targets other than SH4A (where
1393 they are already enabled by default).
1394 * Added support for the __builtin_bswap32 built-in function. It is
1395 now expanded as a sequence of swap.b and swap.w instructions
1396 instead of a library function call.
1397 * The behavior of the -mieee option has been fixed and the negative
1398 form -mno-ieee has been added to control the IEEE conformance of
1399 floating point comparisons. By default -mieee is now enabled and
1400 the option -ffinite-math-only implicitly sets -mno-ieee.
1401 * Added support for the built-in functions __builtin_thread_pointer
1402 and __builtin_set_thread_pointer. This assumes that GBR is used to
1403 hold the thread pointer of the current thread. Memory loads and
1404 stores relative to the address returned by __builtin_thread_pointer
1405 will now also utilize GBR based displacement address modes.
1406 * The -mdiv= option for targets other than SHmedia has been fixed and
1411 * Added optimized instruction scheduling for Niagara4.
1415 * Added support for the -mcmodel=MODEL command-line option. The
1416 models supported are small and large.
1420 * This target now supports the E3V5 architecture via the use of the
1421 new -mv850e3v5 command-line option. It also has experimental
1422 support for the e3v5 LOOP instruction which can be enabled via the
1423 new -mloop command-line option.
1427 * This target now supports the -fstack-usage command-line option.
1433 * Executables are now linked against shared libgcc by default. The
1434 previous default was to link statically, which can still be done by
1435 explicitly specifying -static or static-libgcc on the command line.
1436 However it is strongly advised against, as it will cause problems
1437 for any application that makes use of DLLs compiled by GCC. It
1438 should be alright for a monolithic stand-alone application that
1439 only links against the Windows DLLs, but offers little or no
1444 This is the [33]list of problem reports (PRs) from GCC's bug tracking
1445 system that are known to be fixed in the 4.8.1 release. This list might
1446 not be complete (that is, it is possible that some PRs that have been
1447 fixed are not listed here).
1449 The C++11 <chrono> std::chrono::system_clock and
1450 std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they
1451 both are now separate (never typedefs of each other), both use
1452 std::chrono::nanoseconds resolution, on most GNU/Linux configurations
1453 std::chrono::steady_clock is now finally monotonic, and both classes
1454 are mangled differently than in the previous GCC releases.
1455 std::chrono::system_clock::now() with std::chrono::microseconds resp.
1456 std::chrono::seconds resolution is still exported for backwards
1457 compatibility with default configured libstdc++. Note that libstdc++
1458 configured with --enable-libstdcxx-time= used to be ABI incompatible
1459 with default configured libstdc++ for those two classes and no ABI
1460 compatibility can be offered for those configurations, so any C++11
1461 code that uses those classes and has been compiled and linked against
1462 libstdc++ configured with the non-default --enable-libstdcxx-time=
1463 configuration option needs to be recompiled.
1467 This is the [34]list of problem reports (PRs) from GCC's bug tracking
1468 system that are known to be fixed in the 4.8.2 release. This list might
1469 not be complete (that is, it is possible that some PRs that have been
1470 fixed are not listed here).
1474 This is the [35]list of problem reports (PRs) from GCC's bug tracking
1475 system that are known to be fixed in the 4.8.3 release. This list might
1476 not be complete (that is, it is possible that some PRs that have been
1477 fixed are not listed here).
1479 Support for the new powerpc64le-linux platform has been added. It
1480 defaults to generating code that conforms to the ELFV2 ABI.
1483 For questions related to the use of GCC, please consult these web
1484 pages and the [36]GCC manuals. If that fails, the
1485 [37]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1486 web pages and the development of GCC are welcome on our developer
1487 list at [38]gcc@gcc.gnu.org. All of [39]our lists have public
1490 Copyright (C) [40]Free Software Foundation, Inc. Verbatim copying and
1491 distribution of this entire article is permitted in any medium,
1492 provided this notice is preserved.
1494 These pages are [41]maintained by the GCC team. Last modified
1499 1. https://gcc.gnu.org/wiki/cxx-conversion
1500 2. ftp://gcc.gnu.org/pub/gcc/infrastructure/
1501 3. http://www.nongnu.org/avr-libc/
1502 4. https://gcc.gnu.org/PR54461
1503 5. https://gcc.gnu.org/gcc-4.8/porting_to.html
1504 6. https://code.google.com/p/address-sanitizer/
1505 7. https://code.google.com/p/data-race-test/wiki/ThreadSanitizer
1506 8. https://lkml.org/lkml/2006/11/28/239
1507 9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
1508 10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
1509 11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
1510 12. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf
1511 13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html
1512 14. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html
1513 15. http://gcc.gnu.org/projects/cxx1y.html
1514 16. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2011
1515 17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html
1516 18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
1517 19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
1518 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
1519 21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
1520 22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html
1521 23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html
1522 24. https://gcc.gnu.org/wiki/Fortran2003Status
1523 25. https://gcc.gnu.org/wiki/TS29113Status
1524 26. https://gcc.gnu.org/viewcvs/trunk/libgfortran/libgfortran.h?content-type=text/plain&view=co
1525 27. http://chasm-interop.sourceforge.net/
1526 28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support
1527 29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html
1528 30. https://gcc.gnu.org/onlinedocs/gcc/X86-Built-in-Functions.html#X86-Built-in-Functions
1529 31. https://gcc.gnu.org/onlinedocs/gcc/X86-Built-in-Functions.html#X86-Built-in-Functions
1530 32. https://gcc.gnu.org/wiki/FunctionMultiVersioning
1531 33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1
1532 34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2
1533 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3
1534 36. https://gcc.gnu.org/onlinedocs/
1535 37. mailto:gcc-help@gcc.gnu.org
1536 38. mailto:gcc@gcc.gnu.org
1537 39. https://gcc.gnu.org/lists.html
1538 40. http://www.fsf.org/
1539 41. https://gcc.gnu.org/about.html
1540 42. http://validator.w3.org/check/referer
1541 ======================================================================
1542 http://gcc.gnu.org/gcc-4.7/index.html
1543 GCC 4.7 Release Series
1547 The [1]GNU project and the GCC developers are pleased to announce the
1548 release of GCC 4.7.4.
1550 This release is a bug-fix release, containing fixes for regressions in
1551 GCC 4.7.3 relative to previous releases of GCC.
1556 June 12, 2014 ([2]changes, [3]documentation)
1559 April 11, 2013 ([4]changes, [5]documentation)
1562 September 20, 2012 ([6]changes, [7]documentation)
1565 June 14, 2012 ([8]changes, [9]documentation)
1568 March 22, 2012 ([10]changes, [11]documentation)
1570 References and Acknowledgements
1572 GCC used to stand for the GNU C Compiler, but since the compiler
1573 supports several other languages aside from C, it now stands for the
1574 GNU Compiler Collection.
1576 A list of [12]successful builds is updated as new information becomes
1579 The GCC developers would like to thank the numerous people that have
1580 contributed new features, improvements, bug fixes, and other changes as
1581 well as test results to GCC. This [13]amazing group of volunteers is
1582 what makes GCC successful.
1584 For additional information about GCC please refer to the [14]GCC
1585 project web site or contact the [15]GCC development mailing list.
1587 To obtain GCC please use [16]our mirror sites or [17]our SVN server.
1590 For questions related to the use of GCC, please consult these web
1591 pages and the [18]GCC manuals. If that fails, the
1592 [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
1593 web pages and the development of GCC are welcome on our developer
1594 list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
1597 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
1598 distribution of this entire article is permitted in any medium,
1599 provided this notice is preserved.
1601 These pages are [23]maintained by the GCC team. Last modified
1606 1. http://www.gnu.org/
1607 2. http://gcc.gnu.org/gcc-4.7/changes.html
1608 3. https://gcc.gnu.org/onlinedocs/4.7.4/
1609 4. http://gcc.gnu.org/gcc-4.7/changes.html
1610 5. https://gcc.gnu.org/onlinedocs/4.7.3/
1611 6. http://gcc.gnu.org/gcc-4.7/changes.html
1612 7. https://gcc.gnu.org/onlinedocs/4.7.2/
1613 8. http://gcc.gnu.org/gcc-4.7/changes.html
1614 9. https://gcc.gnu.org/onlinedocs/4.7.1/
1615 10. http://gcc.gnu.org/gcc-4.7/changes.html
1616 11. https://gcc.gnu.org/onlinedocs/4.7.0/
1617 12. http://gcc.gnu.org/gcc-4.7/buildstat.html
1618 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
1619 14. http://gcc.gnu.org/index.html
1620 15. mailto:gcc@gcc.gnu.org
1621 16. http://gcc.gnu.org/mirrors.html
1622 17. http://gcc.gnu.org/svn.html
1623 18. https://gcc.gnu.org/onlinedocs/
1624 19. mailto:gcc-help@gcc.gnu.org
1625 20. mailto:gcc@gcc.gnu.org
1626 21. https://gcc.gnu.org/lists.html
1627 22. http://www.fsf.org/
1628 23. https://gcc.gnu.org/about.html
1629 24. http://validator.w3.org/check/referer
1630 ======================================================================
1631 http://gcc.gnu.org/gcc-4.7/changes.html
1632 GCC 4.7 Release Series
1633 Changes, New Features, and Fixes
1637 * The -fconserve-space flag has been deprecated. The flag had no
1638 effect for most targets: only targets without a global .bss section
1639 and without support for switchable sections. Furthermore, the flag
1640 only had an effect for G++, where it could result in wrong
1641 semantics (please refer to the GCC manual for further details). The
1642 flag will be removed in GCC 4.8
1643 * Support for a number of older systems and recently unmaintained or
1644 untested target ports of GCC has been declared obsolete in GCC 4.7.
1645 Unless there is activity to revive them, the next release of GCC
1646 will have their sources permanently removed.
1647 All GCC ports for the following processor architectures have been
1649 + picoChip (picochip-*)
1650 The following ports for individual systems on particular
1651 architectures have been obsoleted:
1652 + IRIX 6.5 (mips-sgi-irix6.5)
1653 + MIPS OpenBSD (mips*-*-openbsd*)
1654 + Solaris 8 (*-*-solaris2.8). Details can be found in the
1656 + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*)
1657 * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A,
1658 ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by
1659 default, which for some sources generates code that accesses memory
1660 on unaligned addresses. This requires the kernel of those systems
1661 to enable such accesses (controlled by CP15 register c1, refer to
1662 ARM documentation). Alternatively, or for compatibility with
1663 kernels where unaligned accesses are not supported, all code has to
1664 be compiled with -mno-unaligned-access. Upstream Linux kernel
1665 releases have automatically and unconditionally supported unaligned
1666 accesses as emitted by GCC due to this option being active since
1668 * Support on ARM for the legacy floating-point accelerator (FPA) and
1669 the mixed-endian floating-point format that it used has been
1670 obsoleted. The ports that still use this format have been obsoleted
1671 as well. Many legacy ARM ports already provide an alternative that
1672 uses the VFP floating-point format. The obsolete ports will be
1673 deleted in the next release.
1674 The obsolete ports with alternatives are:
1675 + arm*-*-rtems (use arm*-*-rtemseabi)
1676 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi)
1677 + arm*-*-elf (use arm*-*-eabi)
1678 + arm*-*-uclinux* (use arm*-*-uclinux*eabi)
1679 Note, however, that these alternatives are not binary compatible
1680 with their legacy counterparts (although some can support running
1681 legacy applications).
1682 The obsolete ports that currently lack a modern alternative are:
1686 New ports that support more recent versions of the architecture are
1688 * Support for the Maverick co-processor on ARM has been obsoleted.
1689 Code to support it will be deleted in the next release.
1690 * Support has been removed for Unix International threads on Solaris
1691 2, so the --enable-threads=solaris configure option and the
1692 -threads compiler option don't work any longer.
1693 * Support has been removed for the Solaris BSD Compatibility Package,
1694 which lives in /usr/ucbinclude and /usr/ucblib. It has been removed
1695 from Solaris 11, and was only intended as a migration aid from
1696 SunOS 4 to SunOS 5. The -compat-bsd compiler option is not
1697 recognized any longer.
1698 * The AVR port's libgcc has been improved and its multilib structure
1699 has been enhanced. As a result, all objects contributing to an
1700 application must either be compiled with GCC versions up to 4.6.x
1701 or with GCC versions 4.7.1 or later. If the compiler is used with
1702 AVR Libc, you need a version that supports the new layout, i.e.
1703 implements [2]#35407.
1704 * The AVR port's -mshort-calls command-line option has been
1705 deprecated. It will be removed in the GCC 4.8 release. See -mrelax
1707 * The AVR port only references startup code that clears .bss and the
1708 common section resp. initializes the .data and .rodata section
1709 provided respective sections (or subsections thereof) are not
1710 empty, see [3]PR18145. Applications that put all static storage
1711 objects into non-standard sections and / or define all static
1712 storage objects in assembler modules, must reference __do_clear_bss
1713 resp. __do_copy_data by hand or undefine the symbol(s) by means of
1714 -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data.
1715 * The ARM port's -mwords-little-endian option has been deprecated. It
1716 will be removed in a future release.
1717 * Support has been removed for the NetWare x86 configuration
1718 obsoleted in GCC 4.6.
1719 * It is no longer possible to use the "l" constraint in MIPS16 asm
1721 * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard
1722 library which affected the ABI in C++11 mode: a data member was
1723 added to std::list changing its size and altering the definitions
1724 of some member functions, and std::pair's move constructor was
1725 non-trivial which altered the calling convention for functions with
1726 std::pair arguments or return types. The ABI incompatibilities have
1727 been fixed for GCC version 4.7.2 but as a result C++11 code
1728 compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11
1729 code compiled with different GCC versions and with C++98/C++03 code
1730 compiled with any version.
1731 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
1732 rules for the layout of vectors that could lead to wrong code being
1733 generated. Vectors larger than 8 bytes in size are now by default
1734 aligned to an 8-byte boundary. This is an ABI change: code that
1735 makes explicit use of vector types may be incompatible with binary
1736 objects built with older versions of GCC. Auto-vectorized code is
1737 not affected by this change. (This change affects GCC versions
1739 * More information on porting to GCC 4.7 from previous versions of
1740 GCC can be found in the [4]porting guide for this release.
1742 General Optimizer Improvements
1744 * Support for a new parameter --param case-values-threshold=n was
1745 added to allow users to control the cutoff between doing switch
1746 statements as a series of if statements and using a jump table.
1747 * Link-time optimization (LTO) improvements:
1748 + Improved scalability and reduced memory usage. Link time
1749 optimization of Firefox now requires 3GB of RAM on a 64-bit
1750 system, while over 8GB was needed previously. Linking time has
1751 been improved, too. The serial stage of linking Firefox has
1752 been sped up by about a factor of 10.
1753 + Reduced size of object files and temporary storage used during
1755 + Streaming performance (both outbound and inbound) has been
1757 + ld -r is now supported with LTO.
1758 + Several bug fixes, especially in symbol table handling and
1760 * Interprocedural optimization improvements:
1761 + Heuristics now take into account that after inlining code will
1762 be optimized out because of known values (or properties) of
1763 function parameters. For example:
1774 The call of foo will be inlined into bar even when optimizing
1775 for code size. Constructs based on __builtin_constant_p are
1776 now understood by the inliner and code size estimates are
1777 evaluated a lot more realistically.
1778 + The representation of C++ virtual thunks and aliases (both
1779 implicit and defined via the alias attribute) has been
1780 re-engineered. Aliases no longer pose optimization barriers
1781 and calls to an alias can be inlined and otherwise optimized.
1782 + The inter-procedural constant propagation pass has been
1783 rewritten. It now performs generic function specialization.
1784 For example when compiling the following:
1788 ... do something ...
1790 ... do something else ...
1802 GCC will now produce two copies of foo. One with flag being
1803 true, while other with flag being false. This leads to
1804 performance improvements previously possible only by inlining
1805 all calls. Cloning causes a lot less code size growth.
1806 * A string length optimization pass has been added. It attempts to
1807 track string lengths and optimize various standard C string
1808 functions like strlen, strchr, strcpy, strcat, stpcpy and their
1809 _FORTIFY_SOURCE counterparts into faster alternatives. This pass is
1810 enabled by default at -O2 or above, unless optimizing for size, and
1811 can be disabled by the -fno-optimize-strlen option. The pass can
1813 char *bar (const char *a)
1815 size_t l = strlen (a) + 2;
1816 char *p = malloc (l); if (p == NULL) return p;
1817 strcpy (p, a); strcat (p, "/"); return p;
1821 char *bar (const char *a)
1823 size_t tmp = strlen (a);
1824 char *p = malloc (tmp + 2); if (p == NULL) return p;
1825 memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p;
1828 or for hosted compilations where stpcpy is available in the runtime
1829 and headers provide its prototype, e.g.
1830 void foo (char *a, const char *b, const char *c, const char *d)
1832 strcpy (a, b); strcat (a, c); strcat (a, d);
1835 can be optimized into:
1836 void foo (char *a, const char *b, const char *c, const char *d)
1838 strcpy (stpcpy (stpcpy (a, b), c), d);
1841 New Languages and Language specific improvements
1843 * Version 3.1 of the [5]OpenMP specification is now supported for the
1844 C, C++, and Fortran compilers.
1848 * The command-line option -feliminate-unused-debug-types has been
1849 re-enabled by default, as it is for the other languages, leading to
1850 a reduction in debug info size of 12.5% and more for relevant
1851 cases, as well as to a small compilation speedup.
1855 * A new built-in, __builtin_assume_aligned, has been added, through
1856 which the compiler can be hinted about pointer alignment and can
1857 use it to improve generated code.
1858 * A new warning option -Wunused-local-typedefs was added for C, C++,
1859 Objective-C and Objective-C++. This warning diagnoses typedefs
1860 locally defined in a function, and otherwise not used.
1861 * A new experimental command-line option -ftrack-macro-expansion was
1862 added for C, C++, Objective-C, Objective-C++ and Fortran. It allows
1863 the compiler to emit diagnostic about the current macro expansion
1864 stack when a compilation error occurs in a macro expansion.
1865 * Experimental support for transactional memory has been added. It
1866 includes support in the compiler, as well as a supporting runtime
1867 library called libitm. To compile code with transactional memory
1868 constructs, use the -fgnu-tm option.
1869 Support is currently available for Alpha, ARM, PowerPC, SH, SPARC,
1870 and 32-bit/64-bit x86 platforms.
1871 For more details on transactional memory see [6]the GCC WiKi.
1872 * Support for atomic operations specifying the C++11/C11 memory model
1873 has been added. These new __atomic routines replace the existing
1874 __sync built-in routines.
1875 Atomic support is also available for memory blocks. Lock-free
1876 instructions will be used if a memory block is the same size and
1877 alignment as a supported integer type. Atomic operations which do
1878 not have lock-free support are left as function calls. A set of
1879 library functions is available on the GCC atomic wiki in the
1880 "External Atomics Library" section.
1881 For more details on the memory models and features, see the
1883 * When a binary operation is performed on vector types and one of the
1884 operands is a uniform vector, it is possible to replace the vector
1885 with the generating element. For example:
1886 typedef int v4si __attribute__ ((vector_size (16)));
1887 v4si res, a = {1,2,3,4};
1890 res = 2 + a; /* means {2,2,2,2} + a */
1891 res = a - x; /* means a - {x,x,x,x} */
1895 * There is support for some more features from the C11 revision of
1896 the ISO C standard. GCC now accepts the options -std=c11 and
1897 -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x.
1898 + Unicode strings (previously supported only with options such
1899 as -std=gnu11, now supported with -std=c11), and the
1900 predefined macros __STDC_UTF_16__ and __STDC_UTF_32__.
1901 + Nonreturning functions (_Noreturn and <stdnoreturn.h>).
1902 + Alignment support (_Alignas, _Alignof, max_align_t,
1904 + A built-in function __builtin_complex is provided to support C
1905 library implementation of the CMPLX family of macros.
1909 * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat
1910 options, which are equivalent to -std=c++0x, -std=gnu++0x, and
1911 -Wc++0x-compat, respectively.
1912 * G++ now implements [8]C++11 extended friend syntax:
1917 static const int I = 2;
1927 * Thanks to Ville Voutilainen, G++ now implements [9]C++11 explicit
1931 virtual void f() const final;
1932 virtual void f(int);
1936 void f() const; // error: D::f attempts to override final B::f
1937 void f(long) override; // error: doesn't override anything
1938 void f(int) override; // ok
1942 struct F: E { }; // error: deriving from final class
1944 * G++ now implements [10]C++11 non-static data member initializers.
1948 } a; // initializes a.i to 42
1950 * Thanks to Ed Smith-Rowland, G++ now implements [11]C++11
1951 user-defined literals.
1953 // Not actually a good approximation. :)
1954 constexpr long double operator"" _degrees (long double d) { return d * 0.0175; }
1955 long double pi = 180.0_degrees;
1957 * G++ now implements [12]C++11 alias-declarations.
1959 template <class T> using Ptr = T*;
1960 Ptr<int> ip; // decltype(ip) is int*
1962 * Thanks to Ville Voutilainen and Pedro Lamarão, G++ now implements
1963 [13]C++11 delegating constructors.
1967 A(): A(42) { } // delegate to the A(int) constructor
1970 * G++ now fully implements C++11 atomic classes rather than just
1971 integer derived classes.
1977 std::atomic<POD> my_atomic_POD;
1979 * G++ now sets the predefined macro __cplusplus to the correct value,
1980 199711L for C++98/03, and 201103L for C++11.
1981 * G++ now correctly implements the two-phase lookup rules such that
1982 an unqualified name used in a template must have an appropriate
1983 declaration found either in scope at the point of definition of the
1984 template or by argument-dependent lookup at the point of
1985 instantiation. As a result, code that relies on a second
1986 unqualified lookup at the point of instantiation to find functions
1987 declared after the template or in dependent bases will be rejected.
1988 The compiler will suggest ways to fix affected code, and using the
1989 -fpermissive compiler flag will allow the code to compile with a
1993 void f() { g(T()); } // error, g(int) not found by argument-dependent lookup
1994 void g(int) { } // fix by moving this declaration before the declaration of f
1998 // error, B::g(B) not found by argument-dependent lookup
1999 void f() { g(T()); } // fix by using this->g or A::g
2002 struct B { void g(B); };
2010 * G++ now properly re-uses stack space allocated for temporary
2011 objects when their lifetime ends, which can significantly lower
2012 stack consumption for some C++ functions. As a result of this, some
2013 code with undefined behavior will now break:
2015 const int &f(const int &i) { return i; }
2017 const int &x = f(1);
2018 const int &y = f(2);
2020 Here, x refers to the temporary allocated to hold the 1 argument,
2021 which only lives until the end of the initialization; it
2022 immediately becomes a dangling reference. So the next statement
2023 re-uses the stack slot to hold the 2 argument, and users of x get
2025 Note that this should not cause any change of behavior for
2026 temporaries of types with non-trivial destructors, as they are
2027 already destroyed at end of full-expression; the change is that now
2028 the storage is released as well.
2029 * A new command-line option -Wdelete-non-virtual-dtor has been added
2030 to warn when delete is used to destroy an instance of a class which
2031 has virtual functions and non-virtual destructor. It is unsafe to
2032 delete an instance of a derived class through a pointer to a base
2033 class if the base class does not have a virtual destructor. This
2034 warning is enabled by -Wall.
2035 * A new command-line option -Wzero-as-null-pointer-constant has been
2036 added to warn when a literal '0' is used as null pointer constant.
2037 It can be useful to facilitate the conversion to nullptr in C++11.
2038 * As per C++98, access-declarations are now deprecated by G++.
2039 Using-declarations are to be used instead. Furthermore, some
2040 efforts have been made to improve the support of class scope
2041 using-declarations. In particular, using-declarations referring to
2042 a dependent type now work as expected ([14]bug c++/14258).
2043 * The ELF symbol visibility of a template instantiation is now
2044 properly constrained by the visibility of its template arguments
2045 ([15]bug c++/35688).
2047 Runtime Library (libstdc++)
2049 * [16]Improved experimental support for the new ISO C++ standard,
2051 + using noexcept in most of the library;
2052 + implementations of pointer_traits, allocator_traits and
2053 scoped_allocator_adaptor;
2054 + uses-allocator construction for tuple;
2055 + vector meets the allocator-aware container requirements;
2056 + replacing monotonic_clock with steady_clock;
2057 + enabling the thread support library on most POSIX targets;
2058 + many small improvements to conform to the FDIS.
2059 * Added --enable-clocale=newlib configure option.
2060 * Debug Mode iterators for unordered associative containers.
2061 * Avoid polluting the global namespace and do not include <unistd.h>.
2065 * The compile flag [17]-fstack-arrays has been added, which causes
2066 all local arrays to be put on stack memory. For some programs this
2067 will improve the performance significantly. If your program uses
2068 very large local arrays, it is possible that you will have to
2069 extend your runtime limits for stack memory.
2070 * The [18]-Ofast flag now also implies [19]-fno-protect-parens and
2072 * Front-end optimizations can now be selected by the
2073 [21]-ffrontend-optimize option and deselected by the
2074 -fno-frontend-optimize option.
2075 * When front-end optimization removes a function call,
2076 [22]-Wfunction-elimination warns about that.
2077 * When performing front-end-optimization, the
2078 [23]-faggressive-function-elimination option allows the removal of
2079 duplicate function calls even for impure functions.
2080 * The flag [24]-Wreal-q-constant has been added, which warns if
2081 floating-point literals have been specified using q (such as
2082 1.0q0); the q marker is now supported as a vendor extension to
2083 denote quad precision (REAL(16) or, if not available, REAL(10)).
2084 Consider using a kind parameter (such as in 1.0_qp) instead, which
2085 can be obtained via [25]SELECTED_REAL_KIND.
2086 * The GFORTRAN_USE_STDERR environment variable has been removed. GNU
2087 Fortran now always prints error messages to standard error. If you
2088 wish to redirect standard error, please consult the manual for your
2089 OS, shell, batch environment etc. as appropriate.
2090 * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment
2091 variable have been removed. When encountering a serious error,
2092 gfortran will now always abort the program. Whether a core dump is
2093 generated depends on the user environment settings; see the ulimit
2094 -c setting for POSIX shells, limit coredumpsize for C shells, and
2095 the [26]WER user-mode dumps settings on Windows.
2096 * The [27]-fbacktrace option is now enabled by default. When
2097 encountering a fatal error, gfortran will attempt to print a
2098 backtrace to standard error before aborting. It can be disabled
2099 with -fno-backtrace. Note: On POSIX targets with the addr2line
2100 utility from GNU binutils, GNU Fortran can print a backtrace with
2101 function name, file name, line number information in addition to
2102 the addresses; otherwise only the addresses are printed.
2104 + Generic interface names which have the same name as derived
2105 types are now supported, which allows to write constructor
2106 functions. Note that Fortran does not support static
2107 constructor functions; only default initialization or an
2108 explicit structure-constructor initialization are available.
2109 + [29]Polymorphic (class) arrays are now supported.
2111 + Support for the DO CONCURRENT construct has been added, which
2112 allows the user to specify that individual loop iterations
2113 have no interdependencies.
2114 + [31]Coarrays: Full single-image support except for polymorphic
2115 coarrays. Additionally, preliminary support for multiple
2116 images via an MPI-based [32]coarray communication library has
2117 been added. Note: The library version is not yet usable as
2118 remote coarray access is not yet possible.
2120 + New flag [34]-std=f2008ts permits programs that are expected
2121 to conform to the Fortran 2008 standard and the draft
2122 Technical Specification (TS) 29113 on Further Interoperability
2124 + The OPTIONAL attribute is now allowed for dummy arguments of
2126 + The RANK intrinsic has been added.
2127 + The implementation of the ASYNCHRONOUS attribute in GCC is
2128 compatible with the candidate draft of TS 29113 (since GCC
2133 * GCC 4.7 implements the [35]Go 1 language standard. The library
2134 support in 4.7.0 is not quite complete, due to release timing.
2135 Release 4.7.1 includes complete support for Go 1. The Go library is
2136 from the Go 1.0.1 release.
2137 * Go has been tested on GNU/Linux and Solaris platforms. It may work
2138 on other platforms as well.
2140 New Targets and Target Specific Improvements
2144 * GCC now supports the Cortex-A7 processor implementing the v7-a
2145 version of the architecture using the option -mcpu=cortex-a7.
2146 * The default vector size in auto-vectorization for NEON is now 128
2147 bits. If vectorization fails thusly, the vectorizer tries again
2148 with 64-bit vectors.
2149 * A new option -mvectorize-with-neon-double was added to allow users
2150 to change the vector size to 64 bits.
2154 * GCC now supports the XMEGA architecture. This requires GNU binutils
2156 * Support for the [36]named address spaces __flash, __flash1, …,
2157 __flash5 and __memx has been added. These address spaces locate
2158 read-only data in flash memory and allow reading from flash memory
2159 by means of ordinary C code, i.e. without the need of (inline)
2162 const __flash int values[] = { 42, 31 };
2164 int add_values (const __flash int *p, int i)
2166 return values[i] + *p;
2169 * Support has been added for the AVR-specific configure option
2170 --with-avrlibc=yes in order to arrange for better integration of
2171 [37]AVR-Libc. This configure option is supported in avr-gcc 4.7.2
2172 and newer and will only take effect in non-RTEMS configurations. If
2173 avr-gcc is configured for RTEMS, the option will be ignored which
2174 is the same as specifying --with-avrlibc=no. See [38]PR54461 for
2175 more technical details.
2176 * Support for AVR-specific [39]built-in functions has been added.
2177 * Support has been added for the signed and unsigned 24-bit scalar
2178 integer types __int24 and __uint24.
2179 * New command-line options -maccumulate-args, -mbranch-cost=cost and
2180 -mstrict-X were added to allow better fine-tuning of code
2182 * The command option -fdata-sections now also takes affect on the
2183 section names of variables with the progmem attribute.
2184 * A new inline assembler print modifier %i to print a RAM address as
2185 I/O address has been added:
2187 #include <avr/io.h> /* Port Definitions from AVR-LibC */
2189 void set_portb (uint8_t value)
2191 asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory");
2194 The offset between an I/O address and the RAM address for that I/O
2195 location is device-specific. This offset is taken into account when
2196 printing a RAM address with the %i modifier so that the address is
2197 suitable to be used as operand in an I/O command. The address must
2198 be a constant integer known at compile time.
2199 * The inline assembler constraint "R" to represent integers in the
2200 range −6 … 5 has been removed without replacement.
2201 * Many optimizations to:
2202 + 64-bit integer arithmetic
2203 + Widening multiplication
2204 + Integer division by a constant
2205 + Avoid constant reloading in multi-byte instructions.
2206 + Micro-optimizations for special instruction sequences.
2207 + Generic built-in functions like __builtin_ffs*,
2208 __builtin_clz*, etc.
2209 + If-else decision trees generated by switch instructions
2210 + Merging of data located in flash memory
2211 + New libgcc variants for devices with 8-bit wide stack pointer
2213 * Better documentation:
2214 + Handling of EIND and indirect jumps on devices with more than
2215 128 KiB of program memory.
2216 + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function
2218 + Function attributes OS_main and OS_task.
2219 + AVR-specific built-in macros.
2223 * Support has been added for the Texas Instruments C6X family of
2228 * Support has been added for National Semiconductor's CR16
2233 * Support has been added for Adapteva's Epiphany architecture.
2237 * Support for Intel AVX2 intrinsics, built-in functions and code
2238 generation is available via -mavx2.
2239 * Support for Intel BMI2 intrinsics, built-in functions and code
2240 generation is available via -mbmi2.
2241 * Implementation and automatic generation of __builtin_clz* using the
2242 lzcnt instruction is available via -mlzcnt.
2243 * Support for Intel FMA3 intrinsics and code generation is available
2245 * A new -mfsgsbase command-line option is available that makes GCC
2246 generate new segment register read/write instructions through
2247 dedicated built-ins.
2248 * Support for the new Intel rdrnd instruction is available via
2250 * Two additional AVX vector conversion instructions are available via
2252 * Support for new Intel processor codename IvyBridge with RDRND,
2253 FSGSBASE and F16C is available through -march=core-avx-i.
2254 * Support for the new Intel processor codename Haswell with AVX2,
2255 FMA, BMI, BMI2, LZCNT is available through -march=core-avx2.
2256 * Support for new AMD family 15h processors (Piledriver core) is now
2257 available through -march=bdver2 and -mtune=bdver2 options.
2258 * Support for [40]the x32 psABI is now available through the -mx32
2260 * Windows mingw targets are using the -mms-bitfields option by
2262 * Windows x86 targets are using the __thiscall calling convention for
2263 C++ class-member functions.
2264 * Support for the configure option --with-threads=posix for Windows
2269 * GCC now supports thread-local storage (TLS) for MIPS16. This
2270 requires GNU binutils 2.22 or later.
2271 * GCC can now generate code specifically for the Cavium Octeon+ and
2272 Octeon2 processors. The associated command-line options are
2273 -march=octeon+ and -march=octeon2 respectively. Both options
2274 require GNU binutils 2.22 or later.
2275 * GCC can now work around certain 24k errata, under the control of
2276 the command-line option -mfix-24k. These workarounds require GNU
2277 binutils 2.20 or later.
2278 * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build
2279 n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux
2280 toolchain that generates 32-bit code by default. Use the
2281 configure-time option --enable-targets=all to select these extra
2283 * Passing -fno-delayed-branch now also stops the assembler from
2284 automatically filling delay slots.
2288 * Vectors of type vector long long or vector long are passed and
2289 returned using the same method as other vectors with the VSX
2290 instruction set. Previously GCC did not adhere to the ABI for
2291 128-bit vectors with 64-bit integer base types (PR 48857). This
2292 will also be fixed in the GCC 4.6.1 and 4.5.4 releases.
2293 * A new option -mno-pointers-to-nested-functions was added to allow
2294 AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify
2295 that the compiler should not load up the chain register (r11)
2296 before calling a function through a pointer. If you use this
2297 option, you cannot call nested functions through a pointer, or call
2298 other languages that might use the static chain.
2299 * A new option msave-toc-indirect was added to allow AIX
2300 32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we
2301 save the TOC in the prologue for indirect calls or generate the
2302 save inline. This can speed up some programs that call through a
2303 function pointer a lot, but it can slow down other functions that
2304 only call through a function pointer in exceptional cases.
2305 * The PowerPC port will now enable machine-specific built-in
2306 functions when the user switches the target machine using the
2307 #pragma GCC target or __attribute__ ((__target__ ("target"))) code
2308 sequences. In addition, the target macros are updated. However, due
2309 to the way the -save-temps switch is implemented, you won't see the
2310 effect of these additional macros being defined in preprocessor
2315 * A new option -msoft-atomic has been added. When it is specified,
2316 GCC will generate GNU/Linux-compatible gUSA atomic sequences for
2317 the new __atomic routines.
2318 * Since it is neither supported by GAS nor officially documented,
2319 code generation for little endian SH2A has been disabled.
2320 Specifying -ml with -m2a* will now result in a compiler error.
2321 * The defunct -mbranch-cost option has been fixed.
2322 * Some improvements to the generated code of:
2323 + Utilization of the tst #imm,R0 instruction.
2324 + Dynamic shift instructions on SH2A.
2325 + Integer absolute value calculations.
2326 * The -mdiv= option for targets other than SHmedia has been fixed and
2331 * The option -mflat has been reinstated. When it is specified, the
2332 compiler will generate code for a single register window model.
2333 This is essentially a new implementation and the corresponding
2334 debugger support has been added to GDB 7.4.
2335 * Support for the options -mtune=native and -mcpu=native has been
2336 added on selected native platforms (GNU/Linux and Solaris).
2337 * Support for the SPARC T3 (Niagara 3) processor has been added.
2339 + An intrinsics header visintrin.h has been added.
2340 + Builtin intrinsics for the VIS 1.0 edge handling and pixel
2341 compare instructions have been added.
2342 + The little-endian version of alignaddr is now supported.
2343 + When possible, VIS builtins are marked const, which should
2344 increase the compiler's ability to optimize VIS operations.
2345 + The compiler now properly tracks the %gsr register and how it
2346 behaves as an input for various VIS instructions.
2347 + Akin to fzero, the compiler can now generate fone instructions
2348 in order to set all of the bits of a floating-point register
2350 + The documentation for the VIS intrinsics in the GCC manual has
2351 been brought up to date and many inaccuracies were fixed.
2352 + Intrinsics for the VIS 2.0 bmask, bshuffle, and
2353 non-condition-code setting edge instructions have been added.
2354 Their availability is controlled by the new -mvis2 and
2355 -mno-vis2 options. They are enabled by default on
2356 UltraSPARC-III and later CPUs.
2357 * Support for UltraSPARC Fused Multiply-Add floating-point extensions
2358 has been added. These instructions are enabled by default on SPARC
2359 T3 (Niagara 3) and later CPUs.
2363 * Support has been added for the Tilera TILE-Gx and TILEPro families
2366 Other significant improvements
2368 * A new option (-grecord-gcc-switches) was added that appends
2369 compiler command-line options that might affect code generation to
2370 the DW_AT_producer attribute string in the DWARF debugging
2372 * GCC now supports various new GNU extensions to the DWARF debugging
2373 information format, like [41]entry value and [42]call site
2374 information, [43]typed DWARF stack or [44]a more compact macro
2375 representation. Support for these extensions has been added to GDB
2376 7.4. They can be disabled through the -gstrict-dwarf command-line
2381 This is the [45]list of problem reports (PRs) from GCC's bug tracking
2382 system that are known to be fixed in the 4.7.1 release. This list might
2383 not be complete (that is, it is possible that some PRs that have been
2384 fixed are not listed here).
2386 The Go frontend in the 4.7.1 release fully supports the [46]Go 1
2391 This is the [47]list of problem reports (PRs) from GCC's bug tracking
2392 system that are known to be fixed in the 4.7.2 release. This list might
2393 not be complete (that is, it is possible that some PRs that have been
2394 fixed are not listed here).
2398 This is the [48]list of problem reports (PRs) from GCC's bug tracking
2399 system that are known to be fixed in the 4.7.3 release. This list might
2400 not be complete (that is, it is possible that some PRs that have been
2401 fixed are not listed here).
2405 This is the [49]list of problem reports (PRs) from GCC's bug tracking
2406 system that are known to be fixed in the 4.7.4 release. This list might
2407 not be complete (that is, it is possible that some PRs that have been
2408 fixed are not listed here).
2411 For questions related to the use of GCC, please consult these web
2412 pages and the [50]GCC manuals. If that fails, the
2413 [51]gcc-help@gcc.gnu.org mailing list might help. Comments on these
2414 web pages and the development of GCC are welcome on our developer
2415 list at [52]gcc@gcc.gnu.org. All of [53]our lists have public
2418 Copyright (C) [54]Free Software Foundation, Inc. Verbatim copying and
2419 distribution of this entire article is permitted in any medium,
2420 provided this notice is preserved.
2422 These pages are [55]maintained by the GCC team. Last modified
2427 1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html
2428 2. http://savannah.nongnu.org/bugs/?35407
2429 3. https://gcc.gnu.org/PR18145
2430 4. https://gcc.gnu.org/gcc-4.7/porting_to.html
2431 5. http://openmp.org/wp/openmp-specifications/
2432 6. https://gcc.gnu.org/wiki/TransactionalMemory
2433 7. https://gcc.gnu.org/wiki/Atomic/GCCMM
2434 8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
2435 9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
2436 10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
2437 11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
2438 12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
2439 13. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html
2440 14. https://gcc.gnu.org/PR14258
2441 15. https://gcc.gnu.org/PR35688
2442 16. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/libstdc++/manual/manual/status.html#status.iso.2011
2443 17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
2444 18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689
2445 19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270
2446 20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254
2447 21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275
2448 22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170
2449 23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270
2450 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
2451 25. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html
2452 26. http://msdn.microsoft.com/en-us/library/bb787181(v=vs.85).aspx
2453 27. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183
2454 28. https://gcc.gnu.org/wiki/Fortran2003Status
2455 29. https://gcc.gnu.org/wiki/OOP
2456 30. https://gcc.gnu.org/wiki/Fortran2008Status
2457 31. https://gcc.gnu.org/wiki/Coarray
2458 32. https://gcc.gnu.org/wiki/CoarrayLib
2459 33. https://gcc.gnu.org/wiki/TS29113Status
2460 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
2461 35. http://weekly.golang.org/doc/go1.html
2462 36. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html
2463 37. http://nongnu.org/avr-libc/
2464 38. https://gcc.gnu.org/PR54461
2465 39. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built_002din-Functions.html
2466 40. https://sites.google.com/site/x32abi/
2467 41. http://www.dwarfstd.org/ShowIssue.php?issue=100909.1
2468 42. http://www.dwarfstd.org/ShowIssue.php?issue=100909.2
2469 43. http://www.dwarfstd.org/doc/040408.1.html
2470 44. http://www.dwarfstd.org/ShowIssue.php?issue=110722.1
2471 45. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1
2472 46. http://weekly.golang.org/doc/go1.html
2473 47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2
2474 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3
2475 49. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4
2476 50. https://gcc.gnu.org/onlinedocs/
2477 51. mailto:gcc-help@gcc.gnu.org
2478 52. mailto:gcc@gcc.gnu.org
2479 53. https://gcc.gnu.org/lists.html
2480 54. http://www.fsf.org/
2481 55. https://gcc.gnu.org/about.html
2482 56. http://validator.w3.org/check/referer
2483 ======================================================================
2484 http://gcc.gnu.org/gcc-4.6/index.html
2485 GCC 4.6 Release Series
2489 The [1]GNU project and the GCC developers are pleased to announce the
2490 release of GCC 4.6.4.
2492 This release is a bug-fix release, containing fixes for regressions in
2493 GCC 4.6.3 relative to previous releases of GCC.
2498 April 12, 2013 ([2]changes, [3]documentation)
2501 March 1, 2012 ([4]changes, [5]documentation)
2504 October 26, 2011 ([6]changes, [7]documentation)
2507 June 27, 2011 ([8]changes, [9]documentation)
2510 March 25, 2011 ([10]changes, [11]documentation)
2512 References and Acknowledgements
2514 GCC used to stand for the GNU C Compiler, but since the compiler
2515 supports several other languages aside from C, it now stands for the
2516 GNU Compiler Collection.
2518 A list of [12]successful builds is updated as new information becomes
2521 The GCC developers would like to thank the numerous people that have
2522 contributed new features, improvements, bug fixes, and other changes as
2523 well as test results to GCC. This [13]amazing group of volunteers is
2524 what makes GCC successful.
2526 For additional information about GCC please refer to the [14]GCC
2527 project web site or contact the [15]GCC development mailing list.
2529 To obtain GCC please use [16]our mirror sites or [17]our SVN server.
2532 For questions related to the use of GCC, please consult these web
2533 pages and the [18]GCC manuals. If that fails, the
2534 [19]gcc-help@gcc.gnu.org mailing list might help. Comments on these
2535 web pages and the development of GCC are welcome on our developer
2536 list at [20]gcc@gcc.gnu.org. All of [21]our lists have public
2539 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and
2540 distribution of this entire article is permitted in any medium,
2541 provided this notice is preserved.
2543 These pages are [23]maintained by the GCC team. Last modified
2548 1. http://www.gnu.org/
2549 2. http://gcc.gnu.org/gcc-4.6/changes.html
2550 3. https://gcc.gnu.org/onlinedocs/4.6.4/
2551 4. http://gcc.gnu.org/gcc-4.6/changes.html
2552 5. https://gcc.gnu.org/onlinedocs/4.6.3/
2553 6. http://gcc.gnu.org/gcc-4.6/changes.html
2554 7. https://gcc.gnu.org/onlinedocs/4.6.2/
2555 8. http://gcc.gnu.org/gcc-4.6/changes.html
2556 9. https://gcc.gnu.org/onlinedocs/4.6.1/
2557 10. http://gcc.gnu.org/gcc-4.6/changes.html
2558 11. https://gcc.gnu.org/onlinedocs/4.6.0/
2559 12. http://gcc.gnu.org/gcc-4.6/buildstat.html
2560 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
2561 14. http://gcc.gnu.org/index.html
2562 15. mailto:gcc@gcc.gnu.org
2563 16. http://gcc.gnu.org/mirrors.html
2564 17. http://gcc.gnu.org/svn.html
2565 18. https://gcc.gnu.org/onlinedocs/
2566 19. mailto:gcc-help@gcc.gnu.org
2567 20. mailto:gcc@gcc.gnu.org
2568 21. https://gcc.gnu.org/lists.html
2569 22. http://www.fsf.org/
2570 23. https://gcc.gnu.org/about.html
2571 24. http://validator.w3.org/check/referer
2572 ======================================================================
2573 http://gcc.gnu.org/gcc-4.6/changes.html
2574 GCC 4.6 Release Series
2575 Changes, New Features, and Fixes
2579 * The options -b <machine> and -V <version> have been removed because
2580 they were unreliable. Instead, users should directly run
2581 <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to
2582 run a different version of gcc.
2583 * GCC now has stricter checks for invalid command-line options. In
2584 particular, when gcc was called to link object files rather than
2585 compile source code, it would previously accept and ignore all
2586 options starting with --, including linker options such as
2587 --as-needed and --export-dynamic, although such options would
2588 result in errors if any source code was compiled. Such options, if
2589 unknown to the compiler, are now rejected in all cases; if the
2590 intent was to pass them to the linker, options such as
2591 -Wl,--as-needed should be used.
2592 * Versions of the GNU C library up to and including 2.11.1 included
2593 an [1]incorrect implementation of the cproj function. GCC optimizes
2594 its builtin cproj according to the behavior specified and allowed
2595 by the ISO C99 standard. If you want to avoid discrepancies between
2596 the C library and GCC's builtin transformations when using cproj in
2597 your code, use GLIBC 2.12 or later. If you are using an older GLIBC
2598 and actually rely on the incorrect behavior of cproj, then you can
2599 disable GCC's transformations using -fno-builtin-cproj.
2600 * The C-only intermodule optimization framework (IMA, enabled by
2601 -combine) has been removed in favor of the new generic link-time
2602 optimization framework (LTO) introduced in [2]GCC 4.5.0.
2603 * GCC now ships with the LGPL-licensed libquadmath library, which
2604 provides quad-precision mathematical functions for targets with a
2605 __float128 datatype. __float128 is available for targets on 32-bit
2606 x86, x86-64 and Itanium architectures. The libquadmath library is
2607 automatically built on such targets when building the Fortran
2609 * New -Wunused-but-set-variable and -Wunused-but-set-parameter
2610 warnings were added for C, C++, Objective-C and Objective-C++.
2611 These warnings diagnose variables respective parameters which are
2612 only set in the code and never otherwise used. Usually such
2613 variables are useless and often even the value assigned to them is
2614 computed needlessly, sometimes expensively. The
2615 -Wunused-but-set-variable warning is enabled by default by -Wall
2616 flag and -Wunused-but-set-parameter by -Wall -Wextra flags.
2617 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS
2618 rules for the layout of vectors that could lead to wrong code being
2619 generated. Vectors larger than 8 bytes in size are now by default
2620 aligned to an 8-byte boundary. This is an ABI change: code that
2621 makes explicit use of vector types may be incompatible with binary
2622 objects built with older versions of GCC. Auto-vectorized code is
2623 not affected by this change. (This change affects GCC versions
2624 4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.)
2625 * On AVR, variables with the progmem attribute to locate data in
2626 flash memory must be qualified as const.
2627 * Support for a number of older systems and recently unmaintained or
2628 untested target ports of GCC has been declared obsolete in GCC 4.6.
2629 Unless there is activity to revive them, the next release of GCC
2630 will have their sources permanently removed.
2631 All GCC ports for the following processor architectures have been
2633 + Argonaut ARC (arc-*)
2634 + National Semiconductor CRX (crx-*)
2635 + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*,
2636 m68hc12-*-*, m6812-*-*)
2637 + Sunplus S+core (score-*)
2638 The following ports for individual systems on particular
2639 architectures have been obsoleted:
2640 + Interix (i[34567]86-*-interix3*)
2641 + NetWare x86 (i[3456x]86-*-netware*)
2642 + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*)
2643 + MCore PE (mcore-*-pe*)
2644 + SH SymbianOS (sh*-*-symbianelf*)
2645 + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*)
2646 + M68K uClinux old ABI (m68k-*-uclinuxoldabi*)
2647 + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*,
2648 vax-*-netbsd*, but not *-*-netbsdelf*)
2649 The i[34567]86-*-pe alias for Cygwin targets has also been
2650 obsoleted; users should configure for i[34567]86-*-cygwin* instead.
2651 Certain configure options to control the set of libraries built
2652 with GCC on some targets have been obsoleted. On ARM targets, the
2653 options --disable-fpu, --disable-26bit, --disable-underscore,
2654 --disable-interwork, --disable-biendian and --disable-nofmult have
2655 been obsoleted. On MIPS targets, the options
2656 --disable-single-float, --disable-biendian and --disable-softfloat
2657 have been obsoleted.
2658 * Support has been removed for all the [3]configurations obsoleted in
2660 * More information on porting to GCC 4.6 from previous versions of
2661 GCC can be found in the [4]porting guide for this release.
2663 General Optimizer Improvements
2665 * A new general optimization level, -Ofast, has been introduced. It
2666 combines the existing optimization level -O3 with options that can
2667 affect standards compliance but result in better optimized code.
2668 For example, -Ofast enables -ffast-math.
2669 * Link-time optimization improvements:
2670 + The [5]Scalable Whole Program Optimizer (WHOPR) project has
2671 stabilized to the point of being usable. It has become the
2672 default mode when using the LTO optimization model. Link time
2673 optimization can now split itself into multiple parallel
2674 compilations. Parallelism is controlled with -flto=n (where n
2675 specifies the number of compilations to execute in parallel).
2676 GCC can also cooperate with a GNU make job server by
2677 specifying the -flto=jobserver option and adding + to the
2678 beginning of the Makefile rule executing the linker.
2679 Classical LTO mode can be enforced by -flto-partition=none.
2680 This may result in small code quality improvements.
2681 + A large number of bugs were fixed. GCC itself, Mozilla Firefox
2682 and other large applications can be built with LTO enabled.
2683 + The linker plugin support improvements
2684 o Linker plugin is now enabled by default when the linker
2685 is detected to have plugin support. This is the case for
2686 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and
2687 the Gold linker on ELF targets. Plugin support of the
2688 Apple linker on Darwin is not compatible with GCC. The
2689 linker plugin can also be controlled by the
2690 -fuse-linker-plugin command line option.
2691 o Resolution information from the linker plugin is used to
2692 drive whole program assumptions. Use of the linker plugin
2693 results in more aggressive optimization on binaries and
2694 on shared libraries that use the hidden visibility
2695 attribute. Consequently the use of -fwhole-program is not
2696 necessary in addition to LTO.
2697 + Hidden symbols used from non-LTO objects now have to be
2698 explicitly annotated with externally_visible when the linker
2700 + C++ inline functions and virtual tables are now privatized
2701 more aggressively, leading to better inter-procedural
2702 optimization and faster dynamic linking.
2703 + Memory usage and intermediate language streaming performance
2705 + Static constructors and destructors from individual units are
2706 inlined into a single function. This can significantly improve
2707 startup times of large C++ applications where static
2708 constructors are very common. For example, static constructors
2709 are used when including the iostream header.
2710 + Support for the Ada language has been added.
2711 * Interprocedural optimization improvements
2712 + The interprocedural framework was re-tuned for link time
2713 optimization. Several scalability issues were resolved.
2714 + Improved auto-detection of const and pure functions. Newly,
2715 noreturn functions are auto-detected.
2716 The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is
2717 available that informs users when adding attributes to headers
2718 might improve code generation.
2719 + A number of inlining heuristic improvements. In particular:
2720 o Partial inlining is now supported and enabled by default
2721 at -O2 and greater. The feature can be controlled via
2723 Partial inlining splits functions with short hot path to
2724 return. This allows more aggressive inlining of the hot
2725 path leading to better performance and often to code size
2726 reductions (because cold parts of functions are not
2728 o Scalability for large compilation units was improved
2730 o Inlining of callbacks is now more aggressive.
2731 o Virtual methods are considered for inlining when the
2732 caller is inlined and devirtualization is then possible.
2733 o Inlining when optimizing for size (either in cold regions
2734 of a program or when compiling with -Os) was improved to
2735 better handle C++ programs with larger abstraction
2736 penalty, leading to smaller and faster code.
2737 + The IPA reference optimization pass detecting global variables
2738 used or modified by functions was strengthened and sped up.
2739 + Functions whose address was taken are now optimized out when
2740 all references to them are dead.
2741 + A new inter-procedural static profile estimation pass detects
2742 functions that are executed once or unlikely to be executed.
2743 Unlikely executed functions are optimized for size. Functions
2744 executed once are optimized for size except for the inner
2746 + On most targets with named section support, functions used
2747 only at startup (static constructors and main), functions used
2748 only at exit and functions detected to be cold are placed into
2749 separate text segment subsections. This extends the
2750 -freorder-functions feature and is controlled by the same
2751 switch. The goal is to improve the startup time of large C++
2753 Proper function placement requires linker support. GNU ld
2754 2.21.51 on ELF targets was updated to place those functions
2755 together within the text section leading to better code
2756 locality and faster startup times of large C++ programs. The
2757 feature is also supported in the Apple linker. Support in the
2758 gold linker is planned.
2759 * A new switch -fstack-usage has been added. It makes the compiler
2760 output stack usage information for the program, on a per-function
2761 basis, in an auxiliary file.
2762 * A new switch -fcombine-stack-adjustments has been added. It can be
2763 used to enable or disable the compiler's stack-slot combining pass
2764 which before was enabled automatically at -O1 and above, but could
2765 not be controlled on its own.
2766 * A new switch -fstrict-volatile-bitfields has been added. Using it
2767 indicates that accesses to volatile bitfields should use a single
2768 access of the width of the field's type. This option can be useful
2769 for precisely defining and accessing memory-mapped peripheral
2770 registers from C or C++.
2772 Compile time and memory usage improvements
2774 * Datastructures used by the dataflow framework in GCC were
2775 reorganized for better memory usage and more cache locality.
2776 Compile time is improved especially on units with large functions
2777 (possibly resulting from a lot of inlining) not fitting into the
2778 processor cache. The compile time of the GCC C compiler binary with
2779 link-time optimization went down by over 10% (benchmarked on x86-64
2782 New Languages and Language specific improvements
2786 * Stack checking has been improved on selected architectures (Alpha,
2787 IA-32/x86-64, RS/6000 and SPARC): it now will detect stack
2788 overflows in all cases on these architectures.
2789 * Initial support for Ada 2012 has been added.
2793 * A new warning, enabled by -Wdouble-promotion, has been added that
2794 warns about cases where a value of type float is implicitly
2795 promoted to double. This is especially helpful for CPUs that handle
2796 the former in hardware, but emulate the latter in software.
2797 * A new function attribute leaf was introduced. This attribute allows
2798 better inter-procedural optimization across calls to functions that
2799 return to the current unit only via returning or exception
2800 handling. This is the case for most library functions that have no
2802 * Support for a new data type __int128 for targets having wide enough
2803 machine-mode support.
2804 * The new function attribute callee_pop_aggregate allows to specify
2805 if the caller or callee is responsible for popping the aggregate
2806 return pointer value from the stack.
2807 * Support for selectively enabling and disabling warnings via #pragma
2808 GCC diagnostic has been added. For instance:
2809 #pragma GCC diagnostic error "-Wuninitialized"
2810 foo(a); /* error is given for this one */
2811 #pragma GCC diagnostic push
2812 #pragma GCC diagnostic ignored "-Wuninitialized"
2813 foo(b); /* no diagnostic for this one */
2814 #pragma GCC diagnostic pop
2815 foo(c); /* error is given for this one */
2816 #pragma GCC diagnostic pop
2817 foo(d); /* depends on command line options */
2819 * The -fmax-errors=N option is now supported. Using this option
2820 causes the compiler to exit after N errors have been issued.
2824 * There is now experimental support for some features from the
2825 upcoming C1X revision of the ISO C standard. This support may be
2826 selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions.
2827 Note that this support is experimental and may change incompatibly
2828 in future releases for consistency with changes to the C1X standard
2829 draft. The following features are newly supported as described in
2830 the N1539 draft of C1X (with changes agreed at the March 2011 WG14
2831 meeting); some other features were already supported with no
2832 compiler changes being needed, or have some support but not in full
2833 accord with N1539 (as amended).
2834 + Static assertions (_Static_assert keyword)
2835 + Typedef redefinition
2836 + New macros in <float.h>
2837 + Anonymous structures and unions
2838 * The new -fplan9-extensions option directs the compiler to support
2839 some extensions for anonymous struct fields which are implemented
2840 by the Plan 9 compiler. A pointer to a struct may be automatically
2841 converted to a pointer to an anonymous field when calling a
2842 function, in order to make the types match. An anonymous struct
2843 field whose type is a typedef name may be referred to using the
2848 * Improved [7]experimental support for the upcoming C++0x ISO C++
2849 standard, including support for constexpr (thanks to Gabriel Dos
2850 Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide),
2851 noexcept, unrestricted unions, range-based for loops (thanks to
2852 Rodrigo Rivas Costa), opaque enum declarations (thanks also to
2853 Rodrigo), implicitly deleted functions and implicit move
2855 * When an extern declaration within a function does not match a
2856 declaration in the enclosing context, G++ now properly declares the
2857 name within the namespace of the function rather than the namespace
2858 which was open just before the function definition ([8]c++/43145).
2859 * GCC now warns by default when casting integers to larger pointer
2860 types. These warnings can be disabled with the option
2861 -Wno-int-to-pointer-cast, which is now also available in C++.
2862 * G++ no longer optimizes using the assumption that a value of
2863 enumeration type will fall within the range specified by the
2864 standard, since that assumption is easily violated with a
2865 conversion from integer type ([9]c++/43680). The old behavior can
2866 be restored with -fstrict-enums.
2867 * The new -fnothrow-opt flag changes the semantics of a throw()
2868 exception specification to match the proposed semantics of the
2869 noexcept specification: just call terminate if an exception tries
2870 to propagate out of a function with such an exception
2871 specification. This dramatically reduces or eliminates the code
2872 size overhead from adding the exception specification.
2873 * The new -Wnoexcept flag will suggest adding a noexcept qualifier to
2874 a function that the compiler can tell doesn't throw if it would
2875 change the value of a noexcept expression.
2876 * The -Wshadow option now warns if a local variable or type
2877 declaration shadows another type in C++. Note that the compiler
2878 will not warn if a local variable shadows a struct/class/enum, but
2879 will warn if it shadows an explicit typedef.
2880 * When an identifier is not found in the current scope, G++ now
2881 offers suggestions about which identifier might have been intended.
2882 * G++ now issues clearer diagnostics for missing semicolons after
2883 class, struct, and union definitions.
2884 * G++ now issues clearer diagnostics for missing semicolons after
2885 class member declarations.
2886 * G++ now issues clearer diagnostics when a colon is used in a place
2887 where a double-colon was intended.
2888 * G++ no longer accepts mutable on reference members ([10]c++/33558).
2889 Use -fpermissive to allow the old, non-conforming behaviour.
2890 * A few mangling fixes have been made, to attribute const/volatile on
2891 function pointer types, decltype of a plain decl, and use of a
2892 function parameter in the declaration of another parameter. By
2893 default the compiler still uses the old mangling, but emits aliases
2894 with the new mangling on targets that support strong aliases. Users
2895 can switch over entirely to the new mangling with -fabi-version=5
2896 or -fabi-version=0. -Wabi will now warn about code that uses the
2898 * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified
2899 type to be default initialized unless the type has a user-declared
2900 default constructor. In 4.6.2 G++ implements the proposed
2901 resolution of [11]DR 253, so default initialization is allowed if
2902 it initializes all subobjects. Code that fails to compile can be
2903 fixed by providing an initializer e.g.
2905 struct B : A { int i; };
2907 Use -fpermissive to allow the old, non-conforming behaviour.
2909 Runtime Library (libstdc++)
2911 * [12]Improved experimental support for the upcoming ISO C++
2912 standard, C++0x, including using constexpr and nullptr.
2913 * Performance improvements to the [13]Debug Mode, thanks to François
2915 * Atomic operations used for reference-counting are annotated so that
2916 they can be understood by race detectors such as Helgrind, see
2917 [14]Data Race Hunting.
2918 * Most libstdc++ standard headers have been changed to no longer
2919 include the cstddef header as an implementation detail. Code that
2920 relied on that header being included as side-effect of including
2921 other standard headers will need to include cstddef explicitly.
2925 * On systems supporting the libquadmath library, GNU Fortran now also
2926 supports a quad-precision, kind=16 floating-point data type
2927 (REAL(16), COMPLEX(16)). As the data type is not fully supported in
2928 hardware, calculations might be one to two orders of magnitude
2929 slower than with the 4, 8 or 10 bytes floating-point data types.
2930 This change does not affect systems which support REAL(16) in
2931 hardware nor those which do not support libquadmath.
2932 * Much improved compile time for large array constructors.
2933 * In order to reduce execution time and memory consumption, use of
2934 temporary arrays in assignment expressions is avoided for many
2935 cases. The compiler now reverses loops in order to avoid generating
2936 a temporary array where possible.
2937 * Improved diagnostics, especially with -fwhole-file.
2938 * The -fwhole-file flag is now enabled by default. This improves code
2939 generation and diagnostics. It can be disabled using the deprecated
2940 -fno-whole-file flag.
2941 * Support the generation of Makefile dependencies via the [15]-M...
2942 flags of GCC; you may need to specify the -cpp option in addition.
2943 The dependencies take modules, Fortran's include, and CPP's
2944 #include into account. Note: Using -M for the module path is no
2945 longer supported, use -J instead.
2946 * The flag -Wconversion has been modified to only issue warnings
2947 where a conversion leads to information loss. This drastically
2948 reduces the number of warnings; -Wconversion is thus now enabled
2949 with -Wall. The flag -Wconversion-extra has been added and also
2950 warns about other conversions; -Wconversion-extra typically issues
2951 a huge number of warnings, most of which can be ignored.
2952 * A new command-line option -Wunused-dummy-argument warns about
2953 unused dummy arguments and is included in -Wall. Before,
2954 -Wunused-variable also warned about unused dummy arguments.
2955 * Fortran 2003 support has been extended:
2956 + Improved support for polymorphism between libraries and
2957 programs and for complicated inheritance patterns (cf.
2958 [16]object-oriented programming).
2959 + Experimental support of the ASSOCIATE construct.
2960 + In pointer assignments it is now possible to specify the lower
2961 bounds of the pointer and, for a rank-1 or a simply contiguous
2962 data-target, to remap the bounds.
2963 + Automatic (re)allocation: In intrinsic assignments to
2964 allocatable variables the left-hand side will be automatically
2965 allocated (if unallocated) or reallocated (if the shape or
2966 type parameter is different). To avoid the small performance
2967 penalty, you can use a(:) = ... instead of a = ... for arrays
2968 and character strings – or disable the feature using -std=f95
2969 or -fno-realloc-lhs.
2970 + Deferred type parameter: For scalar allocatable and pointer
2971 variables the character length can be deferred.
2972 + Namelist variables with allocatable and pointer attribute and
2973 nonconstant length type parameter are supported.
2974 * Fortran 2008 support has been extended:
2975 + Experimental [17]coarray support (for one image only, i.e.
2976 num_images() == 1); use the [18]-fcoarray=single flag to
2978 + The STOP and the new ERROR STOP statements now support all
2979 constant expressions.
2980 + Support for the CONTIGUOUS attribute.
2981 + Support for ALLOCATE with MOLD.
2982 + Support for the STORAGE_SIZE intrinsic inquiry function.
2983 + Support of the NORM2 and PARITY intrinsic functions.
2984 + The following bit intrinsics were added: POPCNT and POPPAR for
2985 counting the number of 1 bits and returning the parity; BGE,
2986 BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR
2987 for combined left and right shifts, MASKL and MASKR for simple
2988 left and right justified masks, MERGE_BITS for a bitwise merge
2989 using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations,
2990 and the transformational bit intrinsics IALL, IANY and
2992 + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine.
2993 + Support for the IMPURE attribute for procedures, which allows
2994 for ELEMENTAL procedures without the restrictions of PURE.
2995 + Null pointers (including NULL()) and not allocated variables
2996 can be used as actual argument to optional non-pointer,
2997 non-allocatable dummy arguments, denoting an absent argument.
2998 + Non-pointer variables with TARGET attribute can be used as
2999 actual argument to POINTER dummies with INTENT(IN)
3000 + Pointers including procedure pointers and those in a derived
3001 type (pointer components) can now be initialized by a target
3002 instead of only by NULL.
3003 + The EXIT statement (with construct-name) can now be used to
3004 leave not only the DO but also the ASSOCIATE, BLOCK, IF,
3005 SELECT CASE and SELECT TYPE constructs.
3006 + Internal procedures can now be used as actual argument.
3007 + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS
3008 and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV
3009 have been added; these arrays contain the supported kind
3010 values for the respective types.
3011 + The module procedures C_SIZEOF of the intrinsic module
3012 ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of
3013 ISO_FORTRAN_ENV have been implemented.
3014 + Minor changes: obsolescence diagnostics for ENTRY was added
3015 for -std=f2008; a line may start with a semicolon; for
3016 internal and module procedures END can be used instead of END
3017 SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes
3018 a RADIX argument; intrinsic types are supported for
3019 TYPE(intrinsic-type-spec); multiple type-bound procedures can
3020 be declared in a single PROCEDURE statement; implied-shape
3021 arrays are supported for named constants (PARAMETER). The
3022 transformational, three argument versions of BESSEL_JN and
3023 BESSEL_YN were added – the elemental, two-argument version had
3024 been added in GCC 4.4; note that the transformational
3025 functions use a recurrence algorithm.
3029 Support for the [19]Go programming language has been added to GCC. It
3030 is not enabled by default when you build GCC; use the
3031 --enable-languages configure option to build it. The driver program for
3032 compiling Go code is gccgo.
3034 Go is currently known to work on GNU/Linux and RTEMS. Solaris support
3035 is in progress. It may or may not work on other platforms.
3037 Objective-C and Objective-C++
3039 * The -fobjc-exceptions flag is now required to enable Objective-C
3040 exception and synchronization syntax (introduced by the keywords
3041 @try, @catch, @finally and @synchronized).
3042 * A number of Objective-C 2.0 features and extensions are now
3043 supported by GCC. These features are enabled by default; you can
3044 disable them by using the new -fobjc-std=objc1 command-line option.
3045 * The Objective-C 2.0 dot-syntax is now supported. It is an
3046 alternative syntax for using getters and setters; object.count is
3047 automatically converted into [object count] or [object setCount:
3048 ...] depending on context; for example if (object.count > 0) is
3049 automatically compiled into the equivalent of if ([object count] >
3050 0) while object.count = 0; is automatically compiled into the
3051 equivalent ot [object setCount: 0];. The dot-syntax can be used
3052 with instance and class objects and with any setters or getters, no
3053 matter if they are part of a declared property or not.
3054 * Objective-C 2.0 declared properties are now supported. They are
3055 declared using the new @property keyword, and are most commonly
3056 used in conjunction with the new Objective-C 2.0 dot-syntax. The
3057 nonatomic, readonly, readwrite, assign, retain, copy, setter and
3058 getter attributes are all supported. Marking declared properties
3059 with __attribute__ ((deprecated)) is supported too.
3060 * The Objective-C 2.0 @synthesize and @dynamic keywords are
3061 supported. @synthesize causes the compiler to automatically
3062 synthesize a declared property, while @dynamic is used to disable
3063 all warnings for a declared property for which no implementation is
3064 provided at compile time. Synthesizing declared properties requires
3065 runtime support in most useful cases; to be able to use it with the
3066 GNU runtime, appropriate helper functions have been added to the
3067 GNU Objective-C runtime ABI, and are implemented by the GNU
3068 Objective-C runtime library shipped with GCC.
3069 * The Objective-C 2.0 fast enumeration syntax is supported in
3070 Objective-C. This is currently not yet available in Objective-C++.
3071 Fast enumeration requires support in the runtime, and such support
3072 has been added to the GNU Objective-C runtime library (shipped with
3074 * The Objective-C 2.0 @optional keyword is supported. It allows you
3075 to mark methods or properties in a protocol as optional as opposed
3077 * The Objective-C 2.0 @package keyword is supported. It has currently
3078 the same effect as the @public keyword.
3079 * Objective-C 2.0 method attributes are supported. Currently the
3080 supported attributes are deprecated, sentinel, noreturn and format.
3081 * Objective-C 2.0 method argument attributes are supported. The most
3082 widely used attribute is unused, to mark an argument as unused in
3084 * Objective-C 2.0 class and protocol attributes are supported.
3085 Currently the only supported attribute is deprecated.
3086 * Objective-C 2.0 class extensions are supported. A class extension
3087 has the same syntax as a category declaration with no category
3088 name, and the methods and properties declared in it are added
3089 directly to the main class. It is mostly used as an alternative to
3090 a category to add methods to a class without advertising them in
3091 the public headers, with the advantage that for class extensions
3092 the compiler checks that all the privately declared methods are
3093 actually implemented.
3094 * As a result of these enhancements, GCC can now be used to build
3095 Objective-C and Objective-C++ software that uses Foundation and
3096 other important system frameworks with the NeXT runtime on Darwin 9
3097 and Darwin 10 (OSX 10.5 and 10.6).
3098 * Many bugs in the compiler have been fixed in this release; in
3099 particular, LTO can now be used when compiling Objective-C and
3100 Objective-C++ and the parser is much more robust in dealing with
3103 Runtime Library (libobjc)
3105 * The GNU Objective-C runtime library now defines the macro
3106 __GNU_LIBOBJC__ (with a value that is increased at every release
3107 where there is any change to the API) in objc/objc.h, making it
3108 easy to determine if the GNU Objective-C runtime library is being
3109 used, and if so, which version. Previous versions of the GNU
3110 Objective-C runtime library (and other Objective-C runtime
3111 libraries such as the Apple one) do not define this macro.
3112 * A new Objective-C 2.0 API, almost identical to the one implemented
3113 by the Apple Objective-C runtime, has been implemented in the GNU
3114 Objective-C runtime library. The new API hides the internals of
3115 most runtime structures but provides a more extensive set of
3116 functions to operate on them. It is much easier, for example, to
3117 create or modify classes at runtime. The new API also makes it
3118 easier to port software from Apple to GNU as almost no changes
3119 should be required. The old API is still supported for backwards
3120 compatibility; including the old objc/objc-api.h header file
3121 automatically selects the old API, while including the new
3122 objc/runtime.h header file automatically selects the new API.
3123 Support for the old API is being phased out and upgrading the
3124 software to use the new API is strongly recommended. To check for
3125 the availability of the new API, the __GNU_LIBOBJC__ macro can be
3126 used as older versions of the GNU Objective-C runtime library,
3127 which do not support the new API, do not define such a macro.
3128 * Runtime support for @synchronized has been added.
3129 * Runtime support for Objective-C 2.0 synthesized property accessors
3131 * Runtime support for Objective-C 2.0 fast enumeration has been
3134 New Targets and Target Specific Improvements
3138 * GCC now supports the Cortex-M4 processor implementing the v7-em
3139 version of the architecture using the option -mcpu=cortex-m4.
3140 * Scheduling descriptions for the Cortex-M4, the Neon and the
3141 floating point units of the Cortex-A9 and a pipeline description
3142 for the Cortex-A5 have been added.
3143 * Synchronization primitives such as __sync_fetch_and_add and friends
3144 are now inlined for supported architectures rather than calling
3145 into a kernel helper function.
3146 * SSA loop prefetching is enabled by default for the Cortex-A9 at
3148 * Several improvements were committed to improve code generation for
3149 the ARM architecture including a rewritten implementation for load
3150 and store multiples.
3151 * Several enhancements were committed to improve SIMD code generation
3152 for NEON by adding support for widening instructions, misaligned
3153 loads and stores, vector conditionals and support for 64 bit
3155 * Support was added for the Faraday cores fa526, fa606te, fa626te,
3156 fmp626te, fmp626 and fa726te and can be used with the respective
3157 names as parameters to the -mcpu= option.
3158 * Basic support was added for Cortex-A15 and is available through
3160 * GCC for AAPCS configurations now more closely adheres to the AAPCS
3161 specification by enabling -fstrict-volatile-bitfields by default.
3165 * The new -fsplit-stack option permits programs to use a
3166 discontiguous stack. This is useful for threaded programs, in that
3167 it is no longer necessary to specify the maximum stack size when
3168 creating a thread. This feature is currently only implemented for
3169 32-bit and 64-bit x86 GNU/Linux targets.
3170 * Support for emitting profiler counter calls before function
3171 prologues. This is enabled via a new command-line option -mfentry.
3172 * Optimization for the Intel Core 2 processors is now available
3173 through the -march=core2 and -mtune=core2 options.
3174 * Support for Intel Core i3/i5/i7 processors is now available through
3175 the -march=corei7 and -mtune=corei7 options.
3176 * Support for Intel Core i3/i5/i7 processors with AVX is now
3177 available through the -march=corei7-avx and -mtune=corei7-avx
3179 * Support for AMD Bobcat (family 14) processors is now available
3180 through the -march=btver1 and -mtune=btver1 options.
3181 * Support for AMD Bulldozer (family 15) processors is now available
3182 through the -march=bdver1 and -mtune=bdver1 options.
3183 * The default setting (when not optimizing for size) for 32-bit
3184 GNU/Linux and Darwin x86 targets has been changed to
3185 -fomit-frame-pointer. The default can be reverted to
3186 -fno-omit-frame-pointer by configuring GCC with the
3187 --enable-frame-pointer configure option.
3188 * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support
3189 __float128 on 32-bit and 64-bit x86 targets.
3190 * AVX floating-point arithmetic can now be enabled by default at
3191 configure time with the new --with-fpmath=avx option.
3192 * The SSA loop prefetching pass is enabled when using -O3 when
3193 optimizing for CPUs where prefetching is beneficial (AMD CPUs newer
3195 * Support for TBM (Trailing Bit Manipulation) built-in functions and
3196 code generation is available via -mtbm.
3197 * Support for AMD's BMI (Bit Manipulation) built-in functions and
3198 code generation is available via -mbmi.
3202 * Support has been added for the Xilinx MicroBlaze softcore processor
3203 (microblaze-elf) embedded target. This configurable processor is
3204 supported on several Xilinx Spartan and Virtex FPGAs.
3208 * GCC now supports the Loongson 3A processor. Its canonical -march=
3209 and -mtune= name is loongson3a.
3213 * The inline assembly register constraint "A" has been renamed "c".
3214 This constraint is used to select a floating-point register that
3215 can be used as the destination of a multiply-accumulate
3217 * New inline assembly register constraints "A" and "D" have been
3218 added. These constraint letters resolve to all general registers
3219 when compiling for AM33, and resolve to address registers only or
3220 data registers only when compiling for MN10300.
3221 * The MDR register is represented in the compiler. One can access the
3222 register via the "z" constraint in inline assembly. It can be
3223 marked as clobbered or used as a local register variable via the
3224 "mdr" name. The compiler uses the RETF instruction if the function
3225 does not modify the MDR register, so it is important that inline
3226 assembly properly annotate any usage of the register.
3230 * GCC now supports the Applied Micro Titan processor with
3232 * The -mrecip option has been added, which indicates whether the
3233 reciprocal and reciprocal square root instructions should be used.
3234 * The -mveclibabi=mass option can be used to enable the compiler to
3235 autovectorize mathematical functions using the Mathematical
3236 Acceleration Subsystem library.
3237 * The -msingle-pic-base option has been added, which instructs the
3238 compiler to avoid loading the PIC base register in function
3239 prologues. The PIC base register must be initialized by the runtime
3241 * The -mblock-move-inline-limit option has been added, which enables
3242 the user to control the maximum size of inlined memcpy calls and
3244 * PowerPC64 GNU/Linux support for applications requiring a large TOC
3245 section has been improved. A new command-line option,
3246 -mcmodel=MODEL, controls this feature; valid values for MODEL are
3247 small, medium, or large.
3248 * The Altivec builtin functions vec_ld and vec_st have been modified
3249 to generate the Altivec memory instructions LVX and STVX, even if
3250 the -mvsx option is used. In the initial GCC 4.5 release, these
3251 builtin functions were changed to generate VSX memory reference
3252 instructions instead of Altivec memory instructions, but there are
3253 differences between the two instructions. If the VSX instruction
3254 set is available, you can now use the new builtin functions
3255 vec_vsx_ld and vec_vsx_st which always generates the VSX memory
3257 * The GCC compiler on AIX now defaults to a process layout with a
3258 larger data space allowing larger programs to be compiled.
3259 * The GCC long double type on AIX 6.1 and above has reverted to 64
3260 bit double precision, matching the AIX XL compiler default, because
3261 of missing C99 symbols required by the GCC runtime.
3262 * The default processor scheduling model and tuning for PowerPC64
3263 GNU/Linux and for AIX 6.1 and above now is POWER7.
3264 * Starting with GCC 4.6.1, vectors of type vector long long or vector
3265 long are passed and returned in the same method as other vectors
3266 with the VSX instruction set. Previously the GCC compiler did not
3267 adhere to the ABI for 128-bit vectors with 64-bit integer base
3268 types (PR 48857). This is also fixed in the GCC 4.5.4 release.
3270 S/390, zSeries and System z9/z10, IBM zEnterprise z196
3272 * Support for the zEnterprise z196 processor has been added. When
3273 using the -march=z196 option, the compiler will generate code
3274 making use of the following instruction facilities:
3275 + Conditional load/store
3277 + Floating-point-extension
3278 + Interlocked-access
3280 The -mtune=z196 option avoids the compare and branch instructions
3281 as well as the load address instruction with an index register as
3282 much as possible and performs instruction scheduling appropriate
3283 for the new out-of-order pipeline architecture.
3284 * When using the -m31 -mzarch options the generated code still
3285 conforms to the 32-bit ABI but uses the general purpose registers
3286 as 64-bit registers internally. This requires a Linux kernel saving
3287 the whole 64-bit registers when doing a context switch. Kernels
3288 providing that feature indicate that by the 'highgprs' string in
3290 * The SSA loop prefetching pass is enabled when using -O3.
3294 * GCC now supports the LEON series of SPARC V8 processors. The code
3295 generated by the compiler can either be tuned to it by means of the
3296 --with-tune=leon configure option and -mtune=leon compilation
3297 option, or the compiler can be built for the sparc-leon-{elf,linux}
3298 and sparc-leon3-{elf,linux} targets directly.
3299 * GCC has stopped sign/zero-extending parameter registers in the
3300 callee for functions taking parameters with sub-word size in 32-bit
3301 mode, since this is redundant with the specification of the ABI.
3302 GCC has never done so in 64-bit mode since this is also redundant.
3303 * The command line option -mfix-at697f has been added to enable the
3304 documented workaround for the single erratum of the Atmel AT697F
3311 * GCC now supports the Bionic C library and provides a convenient way
3312 of building native libraries and applications for the Android
3313 platform. Refer to the documentation of the -mandroid and -mbionic
3314 options for details on building native code. At the moment, Android
3315 support is enabled only for ARM.
3320 + Initial support for CFString types has been added.
3321 This allows GCC to build projects including the system Core
3322 Foundation frameworks. The GCC Objective-C family supports
3323 CFString "toll-free bridged" as per the Mac OS X system tools.
3324 CFString is also recognized in the context of format
3325 attributes and arguments (see the documentation for format
3326 attributes for limitations). At present, 8-bit character types
3328 + Object file size reduction.
3329 The Darwin zeroed memory allocators have been re-written to
3330 make more use of .zerofill sections. For non-debug code, this
3331 can reduce object file size significantly.
3332 + Objective-C family 64-bit support (NeXT ABI 2).
3333 Initial support has been added to support 64-bit Objective-C
3334 code using the Darwin/OS X native (NeXT) runtime. ABI version
3335 2 will be selected automatically when 64-bit code is built.
3336 + Objective-C family 32-bit ABI 1.
3337 For 32-bit code ABI 1 is also now also allowed. At present it
3338 must be selected manually using -fobjc-abi-version=1 where
3339 applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6).
3341 + The -mdynamic-no-pic option has been enabled.
3342 Code supporting -mdynamic-no-pic optimization has been added
3343 and is applicable to -m32 builds. The compiler bootstrap uses
3344 the option where appropriate.
3345 + The default value for -mtune= has been changed.
3346 Since Darwin systems are primarily Xeon, Core-2 or similar the
3347 default tuning has been changed to -mtune=core2.
3348 + Enable 128-bit long double (__float128) support on Darwin.
3351 Several significant bugs have been fixed, such that GCC now
3352 produces code compatible with the Darwin64 PowerPC ABI.
3353 + libffi and boehm-gc.
3354 The Darwin ports of the libffi and boehm-gc libraries have
3355 been upgraded to include a Darwin64 implementation. This means
3356 that powerpc*-*-darwin9 platforms may now, for example, build
3357 Java applications with -m64 enabled.
3358 + Plug-in support has been enabled.
3359 + The -fsection-anchors option is now available although,
3360 presently, not heavily tested.
3366 * Support symbol versioning with the Sun linker.
3367 * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+.
3368 * Support thread-local storage (TLS) with the Sun assembler on
3370 * Support TLS on Solaris 8/9 if prerequisites are met.
3371 * Support COMDAT group with the GNU assembler and recent Sun linker.
3372 * Support the Sun assembler visibility syntax.
3373 * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp.
3374 -march=pentiumpro (Solaris 8/9).
3375 * Don't use SSE on Solaris 8/9 x86 by default.
3376 * Enable 128-bit long double (__float128) support on Solaris 2/x86.
3380 * Change the ABI for returning 8-byte vectors like __m64 in MMX
3381 registers on Solaris 10+/x86 to match the Sun Studio 12.1+
3382 compilers. This is an incompatible change. If you use such types,
3383 you must either recompile all your code with the new compiler or
3384 use the new -mvect8-ret-in-mem option to remain compatible with
3385 previous versions of GCC and Sun Studio.
3389 * Initial support for decimal floating point.
3390 * Support for the __thiscall calling-convention.
3391 * Support for hot-patchable function prologues via the
3392 ms_hook_prologue attribute for x86_64 in addition to 32-bit x86.
3393 * Improvements of stack-probing and stack-allocation mechanisms.
3394 * Support of push/pop-macro pragma as preprocessor command.
3395 With #pragma push_macro("macro-name") the current definition of
3396 macro-name is saved and can be restored with #pragma
3397 pop_macro("macro-name") to its saved definition.
3398 * Enable 128-bit long double (__float128) support on MinGW and
3401 Other significant improvements
3403 Installation changes
3405 * An install-strip make target is provided that installs stripped
3406 executables, and may install libraries with unneeded or debugging
3408 * On Power7 systems, there is a potential problem if you build the
3409 GCC compiler with a host compiler using options that enable the VSX
3410 instruction set generation. If the host compiler has been patched
3411 so that the vec_ld and vec_st builtin functions generate Altivec
3412 memory instructions instead of VSX memory instructions, then you
3413 should be able to build the compiler with VSX instruction
3416 Changes for GCC Developers
3418 Note: these changes concern developers that develop GCC itself or
3419 software that integrates with GCC, such as plugins, and not the general
3421 * The gengtype utility, which previously was internal to the GCC
3422 build process, has been enchanced to provide GC root information
3423 for plugins as necessary.
3424 * The old GC allocation interface of ggc_alloc and friends was
3425 replaced with a type-safe alternative.
3429 This is the [20]list of problem reports (PRs) from GCC's bug tracking
3430 system that are known to be fixed in the 4.6.1 release. This list might
3431 not be complete (that is, it is possible that some PRs that have been
3432 fixed are not listed here).
3436 This is the [21]list of problem reports (PRs) from GCC's bug tracking
3437 system that are known to be fixed in the 4.6.2 release. This list might
3438 not be complete (that is, it is possible that some PRs that have been
3439 fixed are not listed here).
3443 This is the [22]list of problem reports (PRs) from GCC's bug tracking
3444 system that are known to be fixed in the 4.6.3 release. This list might
3445 not be complete (that is, it is possible that some PRs that have been
3446 fixed are not listed here).
3450 This is the [23]list of problem reports (PRs) from GCC's bug tracking
3451 system that are known to be fixed in the 4.6.4 release. This list might
3452 not be complete (that is, it is possible that some PRs that have been
3453 fixed are not listed here).
3456 For questions related to the use of GCC, please consult these web
3457 pages and the [24]GCC manuals. If that fails, the
3458 [25]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3459 web pages and the development of GCC are welcome on our developer
3460 list at [26]gcc@gcc.gnu.org. All of [27]our lists have public
3463 Copyright (C) [28]Free Software Foundation, Inc. Verbatim copying and
3464 distribution of this entire article is permitted in any medium,
3465 provided this notice is preserved.
3467 These pages are [29]maintained by the GCC team. Last modified
3472 1. http://sourceware.org/bugzilla/show_bug.cgi?id=10401
3473 2. http://gcc.gnu.org/gcc-4.5/changes.html
3474 3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted
3475 4. http://gcc.gnu.org/gcc-4.6/porting_to.html
3476 5. http://gcc.gnu.org/projects/lto/whopr.pdf
3477 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
3478 7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html
3479 8. https://gcc.gnu.org/PR43145
3480 9. https://gcc.gnu.org/PR43680
3481 10. https://gcc.gnu.org/PR33558
3482 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253
3483 12. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.200x
3484 13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html
3485 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races
3486 15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html
3487 16. https://gcc.gnu.org/wiki/OOP
3488 17. https://gcc.gnu.org/wiki/Coarray
3489 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233
3490 19. http://golang.org/
3491 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1
3492 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2
3493 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3
3494 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4
3495 24. https://gcc.gnu.org/onlinedocs/
3496 25. mailto:gcc-help@gcc.gnu.org
3497 26. mailto:gcc@gcc.gnu.org
3498 27. https://gcc.gnu.org/lists.html
3499 28. http://www.fsf.org/
3500 29. https://gcc.gnu.org/about.html
3501 30. http://validator.w3.org/check/referer
3502 ======================================================================
3503 http://gcc.gnu.org/gcc-4.5/index.html
3504 GCC 4.5 Release Series
3508 The [1]GNU project and the GCC developers are pleased to announce the
3509 release of GCC 4.5.4.
3511 This release is a bug-fix release, containing fixes for regressions in
3512 GCC 4.5.3 relative to previous releases of GCC.
3517 Jul 2, 2012 ([2]changes)
3520 Apr 28, 2011 ([3]changes)
3523 Dec 16, 2010 ([4]changes)
3526 Jul 31, 2010 ([5]changes)
3529 April 14, 2010 ([6]changes)
3531 References and Acknowledgements
3533 GCC used to stand for the GNU C Compiler, but since the compiler
3534 supports several other languages aside from C, it now stands for the
3535 GNU Compiler Collection.
3537 A list of [7]successful builds is updated as new information becomes
3540 The GCC developers would like to thank the numerous people that have
3541 contributed new features, improvements, bug fixes, and other changes as
3542 well as test results to GCC. This [8]amazing group of volunteers is
3543 what makes GCC successful.
3545 For additional information about GCC please refer to the [9]GCC project
3546 web site or contact the [10]GCC development mailing list.
3548 To obtain GCC please use [11]our mirror sites or [12]our SVN server.
3551 For questions related to the use of GCC, please consult these web
3552 pages and the [13]GCC manuals. If that fails, the
3553 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
3554 web pages and the development of GCC are welcome on our developer
3555 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
3558 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
3559 distribution of this entire article is permitted in any medium,
3560 provided this notice is preserved.
3562 These pages are [18]maintained by the GCC team. Last modified
3567 1. http://www.gnu.org/
3568 2. http://gcc.gnu.org/gcc-4.5/changes.html
3569 3. http://gcc.gnu.org/gcc-4.5/changes.html
3570 4. http://gcc.gnu.org/gcc-4.5/changes.html
3571 5. http://gcc.gnu.org/gcc-4.5/changes.html
3572 6. http://gcc.gnu.org/gcc-4.5/changes.html
3573 7. http://gcc.gnu.org/gcc-4.5/buildstat.html
3574 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
3575 9. http://gcc.gnu.org/index.html
3576 10. mailto:gcc@gcc.gnu.org
3577 11. http://gcc.gnu.org/mirrors.html
3578 12. http://gcc.gnu.org/svn.html
3579 13. https://gcc.gnu.org/onlinedocs/
3580 14. mailto:gcc-help@gcc.gnu.org
3581 15. mailto:gcc@gcc.gnu.org
3582 16. https://gcc.gnu.org/lists.html
3583 17. http://www.fsf.org/
3584 18. https://gcc.gnu.org/about.html
3585 19. http://validator.w3.org/check/referer
3586 ======================================================================
3587 http://gcc.gnu.org/gcc-4.5/changes.html
3588 GCC 4.5 Release Series
3589 Changes, New Features, and Fixes
3593 * GCC now requires the [1]MPC library in order to build. See the
3594 [2]prerequisites page for version requirements.
3595 * Support for a number of older systems and recently unmaintained or
3596 untested target ports of GCC has been declared obsolete in GCC 4.5.
3597 Unless there is activity to revive them, the next release of GCC
3598 will have their sources permanently removed.
3599 The following ports for individual systems on particular
3600 architectures have been obsoleted:
3601 + IRIX releases before 6.5 (mips-sgi-irix5*,
3602 mips-sgi-irix6.[0-4])
3603 + Solaris 7 (*-*-solaris2.7)
3604 + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*,
3606 + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions
3607 can be found in the [3]announcement.
3608 Support for the classic POWER architecture implemented in the
3609 original RIOS and RIOS2 processors of the old IBM RS/6000 product
3610 line has been obsoleted in the rs6000 port. This does not affect
3611 the new generation Power and PowerPC architectures.
3612 * Support has been removed for all the [4]configurations obsoleted in
3614 * Support has been removed for the protoize and unprotoize utilities,
3615 obsoleted in GCC 4.4.
3616 * Support has been removed for tuning for Itanium1 (Merced) variants.
3617 Note that code tuned for Itanium2 should also run correctly on
3619 * GCC now generates unwind info also for epilogues. DWARF debuginfo
3620 generated by GCC now uses more features of DWARF3 than before, and
3621 also some DWARF4 features. GDB older than 7.0 is not able to handle
3622 either of these, so to debug GCC 4.5 generated binaries or
3623 libraries GDB 7.0 or later is needed. You can disable use of DWARF4
3624 features with the -gdwarf-3 -gstrict-dwarf options, or use
3625 -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but
3626 epilogue unwind info is emitted unconditionally whenever unwind
3628 * On x86 targets, code containing floating-point calculations may run
3629 significantly slower when compiled with GCC 4.5 in strict C99
3630 conformance mode than they did with earlier GCC versions. This is
3631 due to stricter standard conformance of the compiler and can be
3632 avoided by using the option -fexcess-precision=fast; also see
3634 * The function attribute noinline no longer prevents GCC from cloning
3635 the function. A new attribute noclone has been introduced for this
3636 purpose. Cloning a function means that it is duplicated and the new
3637 copy is specialized for certain contexts (for example when a
3638 parameter is a known constant).
3640 General Optimizer Improvements
3642 * The -save-temps now takes an optional argument. The -save-temps and
3643 -save-temps=cwd switches write the temporary files in the current
3644 working directory based on the original source file. The
3645 -save-temps=obj switch will write files into the directory
3646 specified with the -o option, and the intermediate filenames are
3647 based on the output file. This will allow the user to get the
3648 compiler intermediate files when doing parallel builds without two
3649 builds of the same filename located in different directories from
3650 interfering with each other.
3651 * Debugging dumps are now created in the same directory as the object
3652 file rather than in the current working directory. This allows the
3653 user to get debugging dumps when doing parallel builds without two
3654 builds of the same filename interfering with each other.
3655 * GCC has been integrated with the [6]MPC library. This allows GCC to
3656 evaluate complex arithmetic at compile time [7]more accurately. It
3657 also allows GCC to evaluate calls to complex built-in math
3658 functions having constant arguments and replace them at compile
3659 time with their mathematically equivalent results. In doing so, GCC
3660 can generate correct results regardless of the math library
3661 implementation or floating point precision of the host platform.
3662 This also allows GCC to generate identical results regardless of
3663 whether one compiles in native or cross-compile configurations to a
3664 particular target. The following built-in functions take advantage
3665 of this new capability: cacos, cacosh, casin, casinh, catan,
3666 catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan,
3667 and ctanh. The float and long double variants of these functions
3668 (e.g. csinf and csinl) are also handled.
3669 * A new link-time optimizer has been added ([8]-flto). When this
3670 option is used, GCC generates a bytecode representation of each
3671 input file and writes it to specially-named sections in each object
3672 file. When the object files are linked together, all the function
3673 bodies are read from these named sections and instantiated as if
3674 they had been part of the same translation unit. This enables
3675 interprocedural optimizations to work across different files (and
3676 even different languages), potentially improving the performance of
3677 the generated code. To use the link-timer optimizer, -flto needs to
3678 be specified at compile time and during the final link. If the
3679 program does not require any symbols to be exported, it is possible
3680 to combine -flto and the experimental [9]-fwhopr with
3681 [10]-fwhole-program to allow the interprocedural optimizers to use
3682 more aggressive assumptions.
3683 * The automatic parallelization pass was enhanced to support
3684 parallelization of outer loops.
3685 * Automatic parallelization can be enabled as part of Graphite. In
3686 addition to -ftree-parallelize-loops=, specify
3687 -floop-parallelize-all to enable the Graphite-based optimization.
3688 * The infrastructure for optimizing based on [11]restrict qualified
3689 pointers has been rewritten and should result in code generation
3690 improvements. Optimizations based on restrict qualified pointers
3691 are now also available when using -fno-strict-aliasing.
3692 * There is a new optimization pass that attempts to change prototype
3693 of functions to avoid unused parameters, pass only relevant parts
3694 of structures and turn arguments passed by reference to arguments
3695 passed by value when possible. It is enabled by -O2 and above as
3696 well as -Os and can be manually invoked using the new command-line
3698 * GCC now optimize exception handling code. In particular cleanup
3699 regions that are proved to not have any effect are optimized out.
3701 New Languages and Language specific improvements
3705 * The -fshow-column option is now on by default. This means error
3706 messages now have a column associated with them.
3710 * Compilation of programs heavily using discriminated record types
3711 with variant parts has been sped up and generates more compact
3713 * Stack checking now works reasonably well on most plaforms. In some
3714 specific cases, stack overflows may still fail to be detected, but
3715 a compile-time warning will be issued for these cases.
3719 * If a header named in a #include directive is not found, the
3720 compiler exits immediately. This avoids a cascade of errors arising
3721 from declarations expected to be found in that header being
3723 * A new built-in function __builtin_unreachable() has been added that
3724 tells the compiler that control will never reach that point. It may
3725 be used after asm statements that terminate by transferring control
3726 elsewhere, and in other places that are known to be unreachable.
3727 * The -Wlogical-op option now warns for logical expressions such as
3728 (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be
3729 mistakes. This option is disabled by default.
3730 * An asm goto feature has been added to allow asm statements that
3732 * C++0x raw strings are supported for C++ and for C with -std=gnu99.
3733 * The deprecated attribute now takes an optional string argument, for
3734 example, __attribute__((deprecated("text string"))), that will be
3735 printed together with the deprecation warning.
3739 * The -Wenum-compare option, which warns when comparing values of
3740 different enum types, now works for C. It formerly only worked for
3741 C++. This warning is enabled by -Wall. It may be avoided by using a
3743 * The -Wcast-qual option now warns about casts which are unsafe in
3744 that they permit const-correctness to be violated without further
3745 warnings. Specifically, it warns about cases where a qualifier is
3746 added when all the lower types are not const. For example, it warns
3747 about a cast from char ** to const char **.
3748 * The -Wc++-compat option is significantly improved. It issues new
3750 + Using C++ reserved operator names as identifiers.
3751 + Conversions to enum types without explicit casts.
3752 + Using va_arg with an enum type.
3753 + Using different enum types in the two branches of ?:.
3754 + Using ++ or -- on a variable of enum type.
3755 + Using the same name as both a struct, union or enum tag and a
3756 typedef, unless the typedef refers to the tagged type itself.
3757 + Using a struct, union, or enum which is defined within another
3759 + A struct field defined using a typedef if there is a field in
3760 the struct, or an enclosing struct, whose name is the typedef
3762 + Duplicate definitions at file scope.
3763 + Uninitialized const variables.
3764 + A global variable with an anonymous struct, union, or enum
3766 + Using a string constant to initialize a char array whose size
3767 is the length of the string.
3768 * The new -Wjump-misses-init option warns about cases where a goto or
3769 switch skips the initialization of a variable. This sort of branch
3770 is an error in C++ but not in C. This warning is enabled by
3772 * GCC now ensures that a C99-conforming <stdint.h> is present on most
3773 targets, and uses information about the types in this header to
3774 implement the Fortran bindings to those types. GCC does not ensure
3775 the presence of such a header, and does not implement the Fortran
3776 bindings, on the following targets: NetBSD, VxWorks, VMS,
3777 SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF.
3778 * GCC now implements C90- and C99-conforming rules for constant
3779 expressions. This may cause warnings or errors for some code using
3780 expressions that can be folded to a constant but are not constant
3781 expressions as defined by ISO C.
3782 * All known target-independent C90 and C90 Amendment 1 conformance
3783 bugs, and all known target-independent C99 conformance bugs not
3784 related to floating point or extended identifiers, have been fixed.
3785 * The C decimal floating point support now includes support for the
3786 FLOAT_CONST_DECIMAL64 pragma.
3787 * The named address space feature from ISO/IEC TR 18037 is now
3788 supported. This is currently only implemented for the SPU
3793 * Improved [12]experimental support for the upcoming C++0x ISO C++
3794 standard, including support for raw strings, lambda expressions and
3795 explicit type conversion operators.
3796 * When printing the name of a class template specialization, G++ will
3797 now omit any template arguments which come from default template
3798 arguments. This behavior (and the pretty-printing of function
3799 template specializations as template signature and arguments) can
3800 be disabled with the -fno-pretty-templates option.
3801 * Access control is now applied to typedef names used in a template,
3802 which may cause G++ to reject some ill-formed code that was
3803 accepted by earlier releases. The -fno-access-control option can be
3804 used as a temporary workaround until the code is corrected.
3805 * Compilation time for code that uses templates should now scale
3806 linearly with the number of instantiations rather than
3807 quadratically, as template instantiations are now looked up using
3809 * Declarations of functions that look like builtin declarations of
3810 library functions are only considered to be redeclarations if they
3811 are declared with extern "C". This may cause problems with code
3812 that omits extern "C" on hand-written declarations of C library
3813 functions such as abort or memcpy. Such code is ill-formed, but was
3814 accepted by earlier releases.
3815 * Diagnostics that used to complain about passing non-POD types to
3816 ... or jumping past the declaration of a non-POD variable now check
3817 for triviality rather than PODness, as per C++0x.
3818 * In C++0x mode local and anonymous classes are now allowed as
3819 template arguments, and in declarations of variables and functions
3820 with linkage, so long as any such declaration that is used is also
3821 defined ([13]DR 757).
3822 * Labels may now have attributes, as has been permitted for a while
3823 in C. This is only permitted when the label definition and the
3824 attribute specifier is followed by a semicolon—i.e., the label
3825 applies to an empty statement. The only useful attribute for a
3827 * G++ now implements [14]DR 176. Previously G++ did not support using
3828 the injected-class-name of a template base class as a type name,
3829 and lookup of the name found the declaration of the template in the
3830 enclosing scope. Now lookup of the name finds the
3831 injected-class-name, which can be used either as a type or as a
3832 template, depending on whether or not the name is followed by a
3833 template argument list. As a result of this change, some code that
3834 was previously accepted may be ill-formed because
3835 1. The injected-class-name is not accessible because it's from a
3837 2. The injected-class-name cannot be used as an argument for a
3838 template template parameter.
3839 In either of these cases, the code can be fixed by adding a
3840 nested-name-specifier to explicitly name the template. The first
3841 can be worked around with -fno-access-control; the second is only
3842 rejected with -pedantic.
3843 * A new standard mangling for SIMD vector types has been added, to
3844 avoid name clashes on systems with vectors of varying length. By
3845 default the compiler still uses the old mangling, but emits aliases
3846 with the new mangling on targets that support strong aliases. Users
3847 can switch over entirely to the new mangling with -fabi-version=4
3848 or -fabi-version=0. -Wabi will now warn about code that uses the
3850 * The command-line option -ftemplate-depth-N is now written as
3851 -ftemplate-depth=N and the old form is deprecated.
3852 * Conversions between NULL and non-pointer types are now warned by
3853 default. The new option -Wno-conversion-null disables these
3854 warnings. Previously these warnings were only available when using
3855 -Wconversion explicitly.
3857 Runtime Library (libstdc++)
3859 * [15]Improved experimental support for the upcoming ISO C++
3860 standard, C++0x, including:
3861 + Support for <future>, <functional>, and <random>.
3862 + Existing facilities now exploit explicit operators and the
3863 newly implemented core C++0x features.
3864 * An experimental [16]profile mode has been added. This is an
3865 implementation of many C++ standard library constructs with an
3866 additional analysis layer that gives performance improvement advice
3867 based on recognition of suboptimal usage patterns. For example,
3872 for (int k = 0; k < 1024; ++k)
3873 v.insert(v.begin(), k);
3876 When instrumented via the profile mode, can return suggestions
3877 about the initial size and choice of the container used as follows:
3878 vector-to-list: improvement = 5: call stack = 0x804842c ...
3879 : advice = change std::vector to std::list
3880 vector-size: improvement = 3: call stack = 0x804842c ...
3881 : advice = change initial container size from 0 to 1024
3883 These constructs can be substituted for the normal libstdc++
3884 constructs on a piecemeal basis, or all existing components can be
3885 transformed via the -D_GLIBCXX_PROFILE macro.
3886 * [17]Support for decimal floating-point arithmetic (aka ISO C++ TR
3887 24733) has been added. This support is in header file
3888 <decimal/decimal>, uses namespace std::decimal, and includes
3889 classes decimal32, decimal64, and decimal128.
3890 * Sources have been audited for application of function attributes
3891 nothrow, const, pure, and noreturn.
3892 * Python pretty-printers have been added for many standard library
3893 components that simplify the internal representation and present a
3894 more intuitive view of components when used with
3895 appropriately-advanced versions of GDB. For more information,
3896 please consult the more [18]detailed description.
3897 * The default behavior for comparing typeinfo names has changed, so
3898 in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero.
3899 * The new -static-libstdc++ option directs g++ to link the C++
3900 library statically, even if the default would normally be to link
3905 * The COMMON default padding has been changed – instead of adding the
3906 padding before a variable it is now added afterwards, which
3907 increases the compatibility with other vendors and helps to obtain
3908 the correct output in some cases. Cf. also the -falign-commons
3909 option ([19]added in 4.4).
3910 * The -finit-real= option now also supports the value snan for
3911 signalling not-a-number; to be effective, one additionally needs to
3912 enable trapping (e.g. via -ffpe-trap=). Note: Compile-time
3913 optimizations can turn a signalling NaN into a quiet one.
3914 * The new option -fcheck= has been added with the options bounds,
3915 array-temps, do, pointer, and recursive. The bounds and array-temps
3916 options are equivalent to -fbounds-check and
3917 -fcheck-array-temporaries. The do option checks for invalid
3918 modification of loop iteration variables, and the recursive option
3919 tests for recursive calls to subroutines/functions which are not
3920 marked as recursive. With pointer pointer association checks in
3921 calls are performed; however, neither undefined pointers nor
3922 pointers in expressions are handled. Using -fcheck=all enables all
3923 these run-time checks.
3924 * The run-time checking -fcheck=bounds now warns about invalid string
3925 lengths of character dummy arguments. Additionally, more
3926 compile-time checks have been added.
3927 * The new option [20]-fno-protect-parens has been added; if set, the
3928 compiler may reorder REAL and COMPLEX expressions without regard to
3930 * GNU Fortran no longer links against libgfortranbegin. As before,
3931 MAIN__ (assembler symbol name) is the actual Fortran main program,
3932 which is invoked by the main function. However, main is now
3933 generated and put in the same object file as MAIN__. For the time
3934 being, libgfortranbegin still exists for backward compatibility.
3935 For details see the new [21]Mixed-Language Programming chapter in
3937 * The I/O library was restructured for performance and cleaner code.
3938 * Array assignments and WHERE are now run in parallel when OpenMP's
3940 * The experimental option -fwhole-file was added. The option allows
3941 whole-file checking of procedure arguments and allows for better
3942 optimizations. It can also be used with -fwhole-program, which is
3943 now also supported in gfortran.
3944 * More Fortran 2003 and Fortran 2008 mathematical functions can now
3945 be used as initialization expressions.
3946 * Some extended attributes such as STDCALL are now supported via the
3947 [22]GCC$ compiler directive.
3948 * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN
3949 intrinsic behaves now as if zero were always positive.
3950 * For legacy compatibiliy: On Cygwin and MinGW, the special files
3951 CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now
3953 * Fortran 2003 support has been extended:
3954 + Procedure-pointer function results and procedure-pointer
3955 components (including PASS),
3956 + allocatable scalars (experimental),
3957 + DEFERRED type-bound procedures,
3958 + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements
3959 have been implemented.
3960 + The ALLOCATE statement supports type-specs and the SOURCE=
3962 + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC
3963 type-bound procedure (i.e. as type-bound operators).
3964 + Rounding (ROUND=, RZ, ...) for output is now supported.
3965 + The INT_FAST{8,16,32,64,128}_T kind type parameters of the
3966 intrinsic module ISO_C_BINDING are now supported, except for
3967 the targets listed above as ones where GCC does not have
3968 <stdint.h> type information.
3969 + Extensible derived types with type-bound procedure or
3970 procedure pointer with PASS attribute now have to use CLASS in
3971 line with the Fortran 2003 standard; the workaround to use
3972 TYPE is no longer supported.
3973 + [23]Experimental, incomplete support for polymorphism,
3974 including CLASS, SELECT TYPE and dynamic dispatch of
3975 type-bound procedure calls. Some features do not work yet such
3976 as unlimited polymorphism (CLASS(*)).
3977 * Fortran 2008 support has been extended:
3978 + The OPEN statement now supports the NEWUNIT= option, which
3979 returns a unique file unit, thus preventing inadvertent use of
3980 the same unit in different parts of the program.
3981 + Support for unlimited format items has been added.
3982 + The INT{8,16,32} and REAL{32,64,128} kind type parameters of
3983 the intrinsic module ISO_FORTRAN_ENV are now supported.
3984 + Using complex arguments with TAN, SINH, COSH, TANH, ASIN,
3985 ACOS, and ATAN is now possible; the functions ASINH, ACOSH,
3986 and ATANH have been added (for real and complex arguments) and
3987 ATAN(Y,X) is now an alias for ATAN2(Y,X).
3988 + The BLOCK construct has been implemented.
3990 New Targets and Target Specific Improvements
3994 * Full cross-toolchain support now available with GNU Binutils
3998 * GCC now supports the Cortex-M0 and Cortex-A5 processors.
3999 * GCC now supports the ARM v7E-M architecture.
4000 * GCC now supports VFPv4-based FPUs and FPUs with
4001 single-precision-only VFP.
4002 * GCC has many improvements to optimization for other ARM processors,
4003 including scheduling support for the integer pipeline on Cortex-A9.
4004 * GCC now supports the IEEE 754-2008 half-precision floating-point
4005 type, and a variant ARM-specific half-precision type. This type is
4006 specified using __fp16, with the layout determined by
4007 -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and
4008 VFPv4 half-precision instructions will be used.
4009 * GCC now supports the variant of AAPCS that uses VFP registers for
4010 parameter passing and return values.
4014 * The -mno-tablejump option has been removed because it has the same
4015 effect as the -fno-jump-tables option.
4016 * Added support for these new AVR devices:
4023 * GCC now will set the default for -march= based on the configure
4025 * GCC now supports handling floating-point excess precision arising
4026 from use of the x87 floating-point unit in a way that conforms to
4027 ISO C99. This is enabled with -fexcess-precision=standard and with
4028 standards conformance options such as -std=c99, and may be disabled
4029 using -fexcess-precision=fast.
4030 * Support for the Intel Atom processor is now available through the
4031 -march=atom and -mtune=atom options.
4032 * A new -mcrc32 option is now available to enable crc32 intrinsics.
4033 * A new -mmovbe option is now available to enable GCC to use the
4034 movbe instruction to implement __builtin_bswap32 and
4036 * SSE math now can be enabled by default at configure time with the
4037 new --with-fpmath=sse option.
4038 * There is a new intrinsic header file, <x86intrin.h>. It should be
4039 included before using any IA-32/x86-64 intrinsics.
4040 * Support for the XOP, FMA4, and LWP instruction sets for the AMD
4041 Orochi processors are now available with the -mxop, -mfma4, and
4043 * The -mabm option enables GCC to use the popcnt and lzcnt
4044 instructions on AMD processors.
4045 * The -mpopcnt option enables GCC to use the popcnt instructions on
4046 both AMD and Intel processors.
4050 * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x
4052 * GCC now supports thread-local storage (TLS) on M68K and ColdFire
4057 Support has been added for the Toshiba Media embedded Processor (MeP,
4058 or mep-elf) embedded target.
4062 * GCC now supports MIPS 1004K processors.
4063 * GCC can now be configured with options --with-arch-32,
4064 --with-arch-64, --with-tune-32 and --with-tune-64 to control the
4065 default optimization separately for 32-bit and 64-bit modes.
4066 * MIPS targets now support an alternative _mcount interface, in which
4067 register $12 points to the function's save slot for register $31.
4068 This interface is selected by the -mcount-ra-address option; see
4069 the documentation for more details.
4070 * GNU/Linux targets can now generate read-only .eh_frame sections.
4071 This optimization requires GNU binutils 2.20 or above, and is only
4072 available if GCC is configured with a suitable version of binutils.
4073 * GNU/Linux targets can now attach special relocations to indirect
4074 calls, so that the linker can turn them into direct jumps or
4075 branches. This optimization requires GNU binutils 2.20 or later,
4076 and is automatically selected if GCC is configured with an
4077 appropriate version of binutils. It can be explicitly enabled or
4078 disabled using the -mrelax-pic-calls command-line option.
4079 * GCC now generates more heavily-optimized atomic operations on
4081 * MIPS targets now support the -fstack-protector option.
4082 * GCC now supports an -msynci option, which specifies that synci is
4083 enough to flush the instruction cache, without help from the
4084 operating system. GCC uses this information to optimize
4085 automatically-generated cache flush operations, such as those used
4086 for nested functions in C. There is also a --with-synci
4087 configure-time option, which makes -msynci the default.
4088 * GCC supports four new function attributes for interrupt handlers:
4089 interrupt, use_shadow_register_set, keep_interrupts_masked and
4090 use_debug_exception_return. See the documentation for more details
4091 about these attributes.
4093 RS/6000 (POWER/PowerPC)
4095 * GCC now supports the Power ISA 2.06, which includes the VSX
4096 instructions that add vector 64-bit floating point support, new
4097 population count instructions, and conversions between floating
4098 point and unsigned types.
4099 * Support for the power7 processor is now available through the
4100 -mcpu=power7 and -mtune=power7.
4101 * GCC will now vectorize loops that contain simple math functions
4102 like copysign when generating code for altivec or VSX targets.
4103 * Support for the A2 processor is now available through the -mcpu=a2
4104 and -mtune=a2 options.
4105 * Support for the 476 processor is now available through the
4106 -mcpu={476,476fp} and -mtune={476,476fp} options.
4107 * Support for the e500mc64 processor is now available through the
4108 -mcpu=e500mc64 and -mtune=e500mc64 options.
4109 * GCC can now be configured with options --with-cpu-32,
4110 --with-cpu-64, --with-tune-32 and --with-tune-64 to control the
4111 default optimization separately for 32-bit and 64-bit modes.
4112 * Starting with GCC 4.5.4, vectors of type vector long long or vector
4113 long are passed and returned in the same method as other vectors
4114 with the VSX instruction set. Previously the GCC compiler did not
4115 adhere to the ABI for 128-bit vectors with 64-bit integer base
4116 types (PR 48857). This is also fixed in the GCC 4.6.1 release.
4120 Support has been added for the Renesas RX Processor (rx-elf) target.
4124 Windows (Cygwin and MinGW)
4126 * GCC now installs all the major language runtime libraries as DLLs
4127 when configured with the --enable-shared option.
4128 * GCC now makes use of the new support for aligned common variables
4129 in versions of binutils >= 2.20 to fix bugs in the support for SSE
4131 * Improvements to the libffi support library increase the reliability
4132 of code generated by GCJ on all Windows platforms. Libgcj is
4133 enabled by default for the first time.
4134 * Libtool improvements simplify installation by placing the generated
4135 DLLs in the correct binaries directory.
4136 * Numerous other minor bugfixes and improvements, and substantial
4137 enhancements to the Fortran language support library.
4141 Other significant improvements
4145 * It is now possible to extend the compiler without having to modify
4146 its source code. A new option -fplugin=file.so tells GCC to load
4147 the shared object file.so and execute it as part of the compiler.
4148 The internal documentation describes the details on how plugins can
4149 interact with the compiler.
4151 Installation changes
4153 * The move to newer autotools changed default installation
4154 directories and switches to control them: The --with-datarootdir,
4155 --with-docdir, --with-pdfdir, and --with-htmldir switches are not
4156 used any more. Instead, you can now use --datarootdir, --docdir,
4157 --htmldir, and --pdfdir. The default installation directories have
4158 changed as follows according to the GNU Coding Standards:
4160 datarootdir read-only architecture-independent data root [PREFIX/share]
4161 localedir locale-specific message catalogs [DATAROOTDIR/locale]
4162 docdir documentation root [DATAROOTDIR/doc/PACKAGE]
4163 htmldir html documentation [DOCDIR]
4164 dvidir dvi documentation [DOCDIR]
4165 pdfdir pdf documentation [DOCDIR]
4166 psdir ps documentation [DOCDIR]
4167 The following variables have new default values:
4169 datadir read-only architecture-independent data [DATAROOTDIR]
4170 infodir info documentation [DATAROOTDIR/info]
4171 mandir man documentation [DATAROOTDIR/man]
4175 This is the [24]list of problem reports (PRs) from GCC's bug tracking
4176 system that are known to be fixed in the 4.5.1 release. This list might
4177 not be complete (that is, it is possible that some PRs that have been
4178 fixed are not listed here).
4182 * GCC's new link-time optimizer ([25]-flto) now also works on a few
4184 + Cygwin (*-cygwin*)
4186 + Darwin on x86-64 (x86_64-apple-darwin*)
4187 LTO is not enabled by default for these targets. To enable LTO, you
4188 should configure with the --enable-lto option.
4192 This is the [26]list of problem reports (PRs) from GCC's bug tracking
4193 system that are known to be fixed in the 4.5.2 release. This list might
4194 not be complete (that is, it is possible that some PRs that have been
4195 fixed are not listed here).
4199 This is the [27]list of problem reports (PRs) from GCC's bug tracking
4200 system that are known to be fixed in the 4.5.3 release. This list might
4201 not be complete (that is, it is possible that some PRs that have been
4202 fixed are not listed here).
4204 On the PowerPC compiler, the Altivec builtin functions vec_ld and
4205 vec_st have been modified to generate the Altivec memory instructions
4206 LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5
4207 release, these builtin functions were changed to generate VSX memory
4208 reference instructions instead of Altivec memory instructions, but
4209 there are differences between the two instructions. If the VSX
4210 instruction set is available, you can now use the new builtin functions
4211 vec_vsx_ld and vec_vsx_st which always generates the VSX memory
4216 This is the [28]list of problem reports (PRs) from GCC's bug tracking
4217 system that are known to be fixed in the 4.5.4 release. This list might
4218 not be complete (that is, it is possible that some PRs that have been
4219 fixed are not listed here).
4222 For questions related to the use of GCC, please consult these web
4223 pages and the [29]GCC manuals. If that fails, the
4224 [30]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4225 web pages and the development of GCC are welcome on our developer
4226 list at [31]gcc@gcc.gnu.org. All of [32]our lists have public
4229 Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and
4230 distribution of this entire article is permitted in any medium,
4231 provided this notice is preserved.
4233 These pages are [34]maintained by the GCC team. Last modified
4238 1. http://www.multiprecision.org/
4239 2. https://gcc.gnu.org/install/prerequisites.html
4240 3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html
4241 4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted
4242 5. http://gcc.gnu.org/gcc-4.5/changes.html#x86
4243 6. http://www.multiprecision.org/
4244 7. https://gcc.gnu.org/PR30789
4245 8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
4246 9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802
4247 10. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800
4248 11. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html
4249 12. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html
4250 13. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757
4251 14. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176
4252 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.200x
4253 16. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html
4254 17. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733
4255 18. http://sourceware.org/gdb/wiki/STLSupport
4256 19. http://gcc.gnu.org/gcc-4.4/changes.html
4257 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html
4258 21. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html
4259 22. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html
4260 23. https://gcc.gnu.org/wiki/OOP
4261 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1
4262 25. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801
4263 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2
4264 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3
4265 28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4
4266 29. https://gcc.gnu.org/onlinedocs/
4267 30. mailto:gcc-help@gcc.gnu.org
4268 31. mailto:gcc@gcc.gnu.org
4269 32. https://gcc.gnu.org/lists.html
4270 33. http://www.fsf.org/
4271 34. https://gcc.gnu.org/about.html
4272 35. http://validator.w3.org/check/referer
4273 ======================================================================
4274 http://gcc.gnu.org/gcc-4.4/index.html
4275 GCC 4.4 Release Series
4279 The [1]GNU project and the GCC developers are pleased to announce the
4280 release of GCC 4.4.7.
4282 This release is a bug-fix release, containing fixes for regressions in
4283 GCC 4.4.6 relative to previous releases of GCC.
4288 March 13, 2012 ([2]changes)
4291 April 16, 2011 ([3]changes)
4294 October 1, 2010 ([4]changes)
4297 April 29, 2010 ([5]changes)
4300 January 21, 2010 ([6]changes)
4303 October 15, 2009 ([7]changes)
4306 July 22, 2009 ([8]changes)
4309 April 21, 2009 ([9]changes)
4311 References and Acknowledgements
4313 GCC used to stand for the GNU C Compiler, but since the compiler
4314 supports several other languages aside from C, it now stands for the
4315 GNU Compiler Collection.
4317 A list of [10]successful builds is updated as new information becomes
4320 The GCC developers would like to thank the numerous people that have
4321 contributed new features, improvements, bug fixes, and other changes as
4322 well as test results to GCC. This [11]amazing group of volunteers is
4323 what makes GCC successful.
4325 For additional information about GCC please refer to the [12]GCC
4326 project web site or contact the [13]GCC development mailing list.
4328 To obtain GCC please use [14]our mirror sites or [15]our SVN server.
4331 For questions related to the use of GCC, please consult these web
4332 pages and the [16]GCC manuals. If that fails, the
4333 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4334 web pages and the development of GCC are welcome on our developer
4335 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
4338 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
4339 distribution of this entire article is permitted in any medium,
4340 provided this notice is preserved.
4342 These pages are [21]maintained by the GCC team. Last modified
4347 1. http://www.gnu.org/
4348 2. http://gcc.gnu.org/gcc-4.4/changes.html
4349 3. http://gcc.gnu.org/gcc-4.4/changes.html
4350 4. http://gcc.gnu.org/gcc-4.4/changes.html
4351 5. http://gcc.gnu.org/gcc-4.4/changes.html
4352 6. http://gcc.gnu.org/gcc-4.4/changes.html
4353 7. http://gcc.gnu.org/gcc-4.4/changes.html
4354 8. http://gcc.gnu.org/gcc-4.4/changes.html
4355 9. http://gcc.gnu.org/gcc-4.4/changes.html
4356 10. http://gcc.gnu.org/gcc-4.4/buildstat.html
4357 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
4358 12. http://gcc.gnu.org/index.html
4359 13. mailto:gcc@gcc.gnu.org
4360 14. http://gcc.gnu.org/mirrors.html
4361 15. http://gcc.gnu.org/svn.html
4362 16. https://gcc.gnu.org/onlinedocs/
4363 17. mailto:gcc-help@gcc.gnu.org
4364 18. mailto:gcc@gcc.gnu.org
4365 19. https://gcc.gnu.org/lists.html
4366 20. http://www.fsf.org/
4367 21. https://gcc.gnu.org/about.html
4368 22. http://validator.w3.org/check/referer
4369 ======================================================================
4370 http://gcc.gnu.org/gcc-4.4/changes.html
4371 GCC 4.4 Release Series
4372 Changes, New Features, and Fixes
4374 The latest release in the 4.4 release series is [1]GCC 4.4.7.
4378 * __builtin_stdarg_start has been completely removed from GCC.
4379 Support for <varargs.h> had been deprecated since GCC 4.0. Use
4380 __builtin_va_start as a replacement.
4381 * Some of the errors issued by the C++ front end that could be
4382 downgraded to warnings in previous releases by using -fpermissive
4383 are now warnings by default. They can be converted into errors by
4384 using -pedantic-errors.
4385 * Use of the cpp assertion extension will now emit a warning when
4386 -Wdeprecated or -pedantic is used. This extension has been
4387 deprecated for many years, but never warned about.
4388 * Packed bit-fields of type char were not properly bit-packed on many
4389 targets prior to GCC 4.4. On these targets, the fix in GCC 4.4
4390 causes an ABI change. For example there is no longer a 4-bit
4391 padding between field a and b in this structure:
4396 } __attribute__ ((packed));
4397 There is a new warning to help identify fields that are affected:
4398 foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4
4399 The warning can be disabled with -Wno-packed-bitfield-compat.
4400 * On ARM EABI targets, the C++ mangling of the va_list type has been
4401 changed to conform to the current revision of the EABI. This does
4402 not affect the libstdc++ library included with GCC.
4403 * The SCOUNT and POS bits of the MIPS DSP control register are now
4404 treated as global. Previous versions of GCC treated these fields as
4405 call-clobbered instead.
4406 * The MIPS port no longer recognizes the h asm constraint. It was
4407 necessary to remove this constraint in order to avoid generating
4408 unpredictable code sequences.
4409 One of the main uses of the h constraint was to extract the high
4410 part of a multiplication on 64-bit targets. For example:
4411 asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y));
4412 You can now achieve the same effect using 128-bit types:
4413 typedef unsigned int uint128_t __attribute__((mode(TI)));
4414 result = ((uint128_t) x * y) >> 64;
4415 The second sequence is better in many ways. For example, if x and y
4416 are constants, the compiler can perform the multiplication at
4417 compile time. If x and y are not constants, the compiler can
4418 schedule the runtime multiplication better than it can schedule an
4420 * Support for a number of older systems and recently unmaintained or
4421 untested target ports of GCC has been declared obsolete in GCC 4.4.
4422 Unless there is activity to revive them, the next release of GCC
4423 will have their sources permanently removed.
4424 The following ports for individual systems on particular
4425 architectures have been obsoleted:
4426 + Generic a.out on IA32 and m68k (i[34567]86-*-aout*,
4428 + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*,
4429 armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*,
4430 sh-*-*). This does not affect other more specific targets
4431 using the COFF object format on those architectures, or the
4432 more specific H8300 and SH targets (h8300-*-rtems*,
4433 h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*,
4434 sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks).
4435 + 2BSD on PDP-11 (pdp11-*-bsd)
4436 + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*,
4437 powerpc-ibm-aix4.[12]*)
4438 + Tuning support for Itanium1 (Merced) variants. Note that code
4439 tuned for Itanium2 should also run correctly on Itanium1.
4440 * The protoize and unprotoize utilities have been obsoleted and will
4441 be removed in GCC 4.5. These utilities have not been installed by
4442 default since GCC 3.0.
4443 * Support has been removed for all the [2]configurations obsoleted in
4445 * Unknown -Wno-* options are now silently ignored by GCC if no other
4446 diagnostics are issued. If other diagnostics are issued, then GCC
4447 warns about the unknown options.
4448 * More information on porting to GCC 4.4 from previous versions of
4449 GCC can be found in the [3]porting guide for this release.
4451 General Optimizer Improvements
4453 * A new command-line switch -findirect-inlining has been added. When
4454 turned on it allows the inliner to also inline indirect calls that
4455 are discovered to have known targets at compile time thanks to
4457 * A new command-line switch -ftree-switch-conversion has been added.
4458 This new pass turns simple initializations of scalar variables in
4459 switch statements into initializations from a static array, given
4460 that all the values are known at compile time and the ratio between
4461 the new array size and the original switch branches does not exceed
4462 the parameter --param switch-conversion-max-branch-ratio (default
4464 * A new command-line switch -ftree-builtin-call-dce has been added.
4465 This optimization eliminates unnecessary calls to certain builtin
4466 functions when the return value is not used, in cases where the
4467 calls can not be eliminated entirely because the function may set
4468 errno. This optimization is on by default at -O2 and above.
4469 * A new command-line switch -fconserve-stack directs the compiler to
4470 minimize stack usage even if it makes the generated code slower.
4471 This affects inlining decisions.
4472 * When the assembler supports it, the compiler will now emit unwind
4473 information using assembler .cfi directives. This makes it possible
4474 to use such directives in inline assembler code. The new option
4475 -fno-dwarf2-cfi-asm directs the compiler to not use .cfi
4477 * The [4]Graphite branch has been merged. This merge has brought in a
4478 new framework for loop optimizations based on a polyhedral
4479 intermediate representation. These optimizations apply to all the
4480 languages supported by GCC. The following new code transformations
4481 are available in GCC 4.4:
4482 + -floop-interchange performs loop interchange transformations
4483 on loops. Interchanging two nested loops switches the inner
4484 and outer loops. For example, given a loop like:
4487 A(J, I) = A(J, I) * C
4491 loop interchange will transform the loop as if the user had
4495 A(J, I) = A(J, I) * C
4499 which can be beneficial when N is larger than the caches,
4500 because in Fortran, the elements of an array are stored in
4501 memory contiguously by column, and the original loop iterates
4502 over rows, potentially creating at each access a cache miss.
4503 + -floop-strip-mine performs loop strip mining transformations
4504 on loops. Strip mining splits a loop into two nested loops.
4505 The outer loop has strides equal to the strip size and the
4506 inner loop has strides of the original loop within a strip.
4507 For example, given a loop like:
4512 loop strip mining will transform the loop as if the user had
4515 DO I = II, min (II + 3, N)
4520 + -floop-block performs loop blocking transformations on loops.
4521 Blocking strip mines each loop in the loop nest such that the
4522 memory accesses of the element loops fit inside caches. For
4523 example, given a loop like:
4526 A(J, I) = B(I) + C(J)
4530 loop blocking will transform the loop as if the user had
4534 DO I = II, min (II + 63, N)
4535 DO J = JJ, min (JJ + 63, M)
4536 A(J, I) = B(I) + C(J)
4542 which can be beneficial when M is larger than the caches,
4543 because the innermost loop will iterate over a smaller amount
4544 of data that can be kept in the caches.
4545 * A new register allocator has replaced the old one. It is called
4546 integrated register allocator (IRA) because coalescing, register
4547 live range splitting, and hard register preferencing are done
4548 on-the-fly during coloring. It also has better integration with the
4549 reload pass. IRA is a regional register allocator which uses modern
4550 Chaitin-Briggs coloring instead of Chow's priority coloring used in
4551 the old register allocator. More info about IRA internals and
4552 options can be found in the GCC manuals.
4553 * A new instruction scheduler and software pipeliner, based on the
4554 selective scheduling approach, has been added. The new pass
4555 performs instruction unification, register renaming, substitution
4556 through register copies, and speculation during scheduling. The
4557 software pipeliner is able to pipeline non-countable loops. The new
4558 pass is targeted at scheduling-eager in-order platforms. In GCC 4.4
4559 it is available for the Intel Itanium platform working by default
4560 as the second scheduling pass (after register allocation) at the
4561 -O3 optimization level.
4562 * When using -fprofile-generate with a multi-threaded program, the
4563 profile counts may be slightly wrong due to race conditions. The
4564 new -fprofile-correction option directs the compiler to apply
4565 heuristics to smooth out the inconsistencies. By default the
4566 compiler will give an error message when it finds an inconsistent
4568 * The new -fprofile-dir=PATH option permits setting the directory
4569 where profile data files are stored when using -fprofile-generate
4570 and friends, and the directory used when reading profile data files
4571 using -fprofile-use and friends.
4575 * The new -Wframe-larger-than=NUMBER option directs GCC to emit a
4576 warning if any stack frame is larger than NUMBER bytes. This may be
4577 used to help ensure that code fits within a limited amount of stack
4579 * The command-line option -Wlarger-than-N is now written as
4580 -Wlarger-than=N and the old form is deprecated.
4581 * The new -Wno-mudflap option disables warnings about constructs
4582 which can not be instrumented when using -fmudflap.
4584 New Languages and Language specific improvements
4586 * Version 3.0 of the [5]OpenMP specification is now supported for the
4587 C, C++, and Fortran compilers.
4588 * New character data types, per [6]TR 19769: New character types in
4589 C, are now supported for the C compiler in -std=gnu99 mode, as
4590 __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in
4591 -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too.
4595 * A new optimize attribute was added to allow programmers to change
4596 the optimization level and particular optimization options for an
4597 individual function. You can also change the optimization options
4598 via the GCC optimize pragma for functions defined after the pragma.
4599 The GCC push_options pragma and the GCC pop_options pragma allow
4600 you temporarily save and restore the options used. The GCC
4601 reset_options pragma restores the options to what was specified on
4603 * Uninitialized warnings do not require enabling optimization
4604 anymore, that is, -Wuninitialized can be used together with -O0.
4605 Nonetheless, the warnings given by -Wuninitialized will probably be
4606 more accurate if optimization is enabled.
4607 * -Wparentheses now warns about expressions such as (!x | y) and (!x
4608 & y). Using explicit parentheses, such as in ((!x) | y), silences
4610 * -Wsequence-point now warns within if, while,do while and for
4611 conditions, and within for begin/end expressions.
4612 * A new option -dU is available to dump definitions of preprocessor
4613 macros that are tested or expanded.
4617 * [7]Improved experimental support for the upcoming ISO C++ standard,
4618 C++0x. Including support for auto, inline namespaces, generalized
4619 initializer lists, defaulted and deleted functions, new character
4620 types, and scoped enums.
4621 * Those errors that may be downgraded to warnings to build legacy
4622 code now mention -fpermissive when -fdiagnostics-show-option is
4624 * -Wconversion now warns if the result of a static_cast to enumeral
4625 type is unspecified because the value is outside the range of the
4627 * -Wuninitialized now warns if a non-static reference or non-static
4628 const member appears in a class without constructors.
4629 * G++ now properly implements value-initialization, so objects with
4630 an initializer of () and an implicitly defined default constructor
4631 will be zero-initialized before the default constructor is called.
4633 Runtime Library (libstdc++)
4635 * [8]Improved experimental support for the upcoming ISO C++ standard,
4637 + Support for <chrono>, <condition_variable>, <cstdatomic>,
4638 <forward_list>, <initializer_list>, <mutex>, <ratio>,
4639 <system_error>, and <thread>.
4640 + unique_ptr, <algorithm> additions, exception propagation, and
4641 support for the new character types in <string> and <limits>.
4642 + Existing facilities now exploit initializer lists, defaulted
4643 and deleted functions, and the newly implemented core C++0x
4645 + Some standard containers are more efficient together with
4646 stateful allocators, i.e., no allocator is constructed on the
4647 fly at element construction time.
4648 * Experimental support for non-standard pointer types in containers.
4649 * The long standing libstdc++/30928 has been fixed for targets
4650 running glibc 2.10 or later.
4651 * As usual, many small and larger bug fixes, in particular quite a
4652 few corner cases in <locale>.
4656 * GNU Fortran now employs libcpp directly instead of using cc1 as an
4657 external preprocessor. The [9]-cpp option was added to allow manual
4658 invocation of the preprocessor without relying on filename
4660 * The [10]-Warray-temporaries option warns about array temporaries
4661 generated by the compiler, as an aid to optimization.
4662 * The [11]-fcheck-array-temporaries option has been added, printing a
4663 notification at run time, when an array temporary had to be created
4664 for an function argument. Contrary to -Warray-temporaries the
4665 warning is only printed if the array is noncontiguous.
4666 * Improved generation of DWARF debugging symbols
4667 * If using an intrinsic not part of the selected standard (via -std=
4668 and -fall-intrinsics) gfortran will now treat it as if this
4669 procedure were declared EXTERNAL and try to link to a user-supplied
4670 procedure. -Wintrinsics-std will warn whenever this happens. The
4671 now-useless option -Wnonstd-intrinsic was removed.
4672 * The flag -falign-commons has been added to control the alignment of
4673 variables in COMMON blocks, which is enabled by default in line
4674 with previous GCC version. Using -fno-align-commons one can force
4675 commons to be contiguous in memory as required by the Fortran
4676 standard, however, this slows down the memory access. The option
4677 -Walign-commons, which is enabled by default, warns when padding
4678 bytes were added for alignment. The proper solution is to sort the
4679 common objects by decreasing storage size, which avoids the
4681 * Fortran 2003 support has been extended:
4682 + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is
4683 now supported (except internal reads from/writes to wide
4684 strings). [12]-fbackslash now supports also \unnnn and
4685 \Unnnnnnnn to enter Unicode characters.
4686 + Asynchronous I/O (implemented as synchronous I/O) and the
4687 decimal=, size=, sign=, pad=, blank=, and delim= specifiers
4688 are now supported in I/O statements.
4689 + Support for Fortran 2003 structure constructors and for array
4690 constructor with typespec has been added.
4691 + Procedure Pointers (but not yet as component in derived types
4692 and as function results) are now supported.
4693 + Abstract types, type extension, and type-bound procedures
4694 (both PROCEDURE and GENERIC but not as operators). Note: As
4695 CLASS/polymorphyic types are not implemented, type-bound
4696 procedures with PASS accept as non-standard extension TYPE
4698 * Fortran 2008 support has been added:
4699 + The -std=f2008 option and support for the file extensions
4700 .f2008 and .F2008 has been added.
4701 + The g0 format descriptor is now supported.
4702 + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH,
4703 ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED
4704 are now available (some of them existed as GNU extension
4705 before). Note: The hyperbolic functions are not yet supporting
4706 complex arguments and the three- argument version of BESSEL_*N
4708 + The bit intrinsics LEADZ and TRAILZ have been added.
4714 * The Ada runtime now supports multilibs on many platforms including
4715 x86_64, SPARC and PowerPC. Their build is enabled by default.
4717 New Targets and Target Specific Improvements
4721 * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and
4722 Cortex-R4F processors and has many other improvements to
4723 optimization for ARM processors.
4724 * GCC now supports the VFPv3 variant with 16 double-precision
4725 registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been
4726 renamed to -mfpu=vfpv3.
4727 * GCC now supports the -mfix-cortex-m3-ldrd option to work around an
4728 erratum on Cortex-M3 processors.
4729 * GCC now supports the __sync_* atomic operations for ARM EABI
4731 * The section anchors optimization is now enabled by default when
4733 * GCC now uses a new EABI-compatible profiling interface for EABI
4734 targets. This requires a function __gnu_mcount_nc, which is
4735 provided by GNU libc versions 2.8 and later.
4739 * The -mno-tablejump option has been deprecated because it has the
4740 same effect as the -fno-jump-tables option.
4741 * Added support for these new AVR devices:
4770 * Support for Intel AES built-in functions and code generation is
4771 available via -maes.
4772 * Support for Intel PCLMUL built-in function and code generation is
4773 available via -mpclmul.
4774 * Support for Intel AVX built-in functions and code generation is
4775 available via -mavx.
4776 * Automatically align the stack for local variables with alignment
4778 * GCC can now utilize the SVML library for vectorizing calls to a set
4779 of C99 functions if -mveclibabi=svml is specified and you link to
4780 an SVML ABI compatible library.
4781 * On x86-64, the ABI has been changed in the following cases to
4782 conform to the x86-64 ABI:
4783 + Passing/returning structures with flexible array member:
4789 + Passing/returning structures with complex float member:
4793 __complex__ float f;
4795 + Passing/returning unions with long double member:
4801 Code built with previous versions of GCC that uses any of these is
4802 not compatible with code built with GCC 4.4.0 or later.
4803 * A new target attribute was added to allow programmers to change the
4804 target options like -msse2 or -march=k8 for an individual function.
4805 You can also change the target options via the GCC target pragma
4806 for functions defined after the pragma.
4807 * GCC can now be configured with options --with-arch-32,
4808 --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and
4809 --with-tune-64 to control the default optimization separately for
4810 32-bit and 64-bit modes.
4814 * Support for __float128 (TFmode) IEEE quad type and corresponding
4815 TCmode IEEE complex quad type is available via the soft-fp library
4816 on IA-32/IA64 targets. This includes basic arithmetic operations
4817 (addition, subtraction, negation, multiplication and division) on
4818 __float128 real and TCmode complex values, the full set of IEEE
4819 comparisons between __float128 values, conversions to and from
4820 float, double and long double floating point types, as well as
4821 conversions to and from signed or unsigned integer, signed or
4822 unsigned long integer and signed or unsigned quad (TImode, IA64
4823 only) integer types. Additionally, all operations generate the full
4824 set of IEEE exceptions and support the full set of IEEE rounding
4829 * GCC now supports instruction scheduling for ColdFire V1, V3 and V4
4830 processors. (Scheduling support for ColdFire V2 processors was
4832 * GCC now supports the -mxgot option to support programs requiring
4833 many GOT entries on ColdFire.
4834 * The m68k-*-linux-gnu target now builds multilibs by default.
4838 * MIPS Technologies have extended the original MIPS SVR4 ABI to
4839 include support for procedure linkage tables (PLTs) and copy
4840 relocations. These extensions allow GNU/Linux executables to use a
4841 significantly more efficient code model than the one defined by the
4843 GCC support for this code model is available via a new command-line
4844 option, -mplt. There is also a new configure-time option,
4845 --with-mips-plt, to make -mplt the default.
4846 The new code model requires support from the assembler, the linker,
4847 and the runtime C library. This support is available in binutils
4849 * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables
4850 and 32-bit GNU/Linux shared libraries. This feature requires GNU
4851 binutils 2.19 or above.
4852 * Support for RMI's XLR processor is now available through the
4853 -march=xlr and -mtune=xlr options.
4854 * 64-bit targets can now perform 128-bit multiplications inline,
4855 instead of relying on a libgcc function.
4856 * Native GNU/Linux toolchains now support -march=native and
4857 -mtune=native, which select the host processor.
4858 * GCC now supports the R10K, R12K, R14K and R16K processors. The
4859 canonical -march= and -mtune= names for these processors are
4860 r10000, r12000, r14000 and r16000 respectively.
4861 * GCC can now work around the side effects of speculative execution
4862 on R10K processors. Please see the documentation of the
4863 -mr10k-cache-barrier option for details.
4864 * Support for the MIPS64 Release 2 instruction set has been added.
4865 The option -march=mips64r2 enables generation of these
4867 * GCC now supports Cavium Networks' Octeon processor. This support is
4868 available through the -march=octeon and -mtune=octeon options.
4869 * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The
4870 canonical -march= and -mtune= names for these processors are
4871 loongson2e and loongson2f.
4875 Picochip is a 16-bit processor. A typical picoChip contains over 250
4876 small cores, each with small amounts of memory. There are three
4877 processor variants (STAN, MEM and CTRL) with different instruction sets
4878 and memory configurations and they can be chosen using the -mae option.
4880 This port is intended to be a "C" only port.
4882 Power Architecture and PowerPC
4884 * GCC now supports the e300c2, e300c3 and e500mc processors.
4885 * GCC now supports Xilinx processors with a single-precision FPU.
4886 * Decimal floating point is now supported for e500 processors.
4888 S/390, zSeries and System z9/z10
4890 * Support for the IBM System z10 EC/BC processor has been added. When
4891 using the -march=z10 option, the compiler will generate code making
4892 use of instructions provided by the General-Instruction-Extension
4893 Facility and the Execute-Extension Facility.
4897 * GCC now supports the thread-local storage mechanism used on
4902 * GCC now supports thread-local storage (TLS) for Xtensa processor
4903 configurations that include the Thread Pointer option. TLS also
4904 requires support from the assembler and linker; this support is
4905 provided in the GNU binutils beginning with version 2.19.
4907 Documentation improvements
4909 Other significant improvements
4913 This is the [13]list of problem reports (PRs) from GCC's bug tracking
4914 system that are known to be fixed in the 4.4.1 release. This list might
4915 not be complete (that is, it is possible that some PRs that have been
4916 fixed are not listed here).
4920 This is the [14]list of problem reports (PRs) from GCC's bug tracking
4921 system that are known to be fixed in the 4.4.2 release. This list might
4922 not be complete (that is, it is possible that some PRs that have been
4923 fixed are not listed here).
4927 This is the [15]list of problem reports (PRs) from GCC's bug tracking
4928 system that are known to be fixed in the 4.4.3 release. This list might
4929 not be complete (that is, it is possible that some PRs that have been
4930 fixed are not listed here).
4934 This is the [16]list of problem reports (PRs) from GCC's bug tracking
4935 system that are known to be fixed in the 4.4.4 release. This list might
4936 not be complete (that is, it is possible that some PRs that have been
4937 fixed are not listed here).
4941 This is the [17]list of problem reports (PRs) from GCC's bug tracking
4942 system that are known to be fixed in the 4.4.5 release. This list might
4943 not be complete (that is, it is possible that some PRs that have been
4944 fixed are not listed here).
4948 This is the [18]list of problem reports (PRs) from GCC's bug tracking
4949 system that are known to be fixed in the 4.4.6 release. This list might
4950 not be complete (that is, it is possible that some PRs that have been
4951 fixed are not listed here).
4955 This is the [19]list of problem reports (PRs) from GCC's bug tracking
4956 system that are known to be fixed in the 4.4.7 release. This list might
4957 not be complete (that is, it is possible that some PRs that have been
4958 fixed are not listed here).
4961 For questions related to the use of GCC, please consult these web
4962 pages and the [20]GCC manuals. If that fails, the
4963 [21]gcc-help@gcc.gnu.org mailing list might help. Comments on these
4964 web pages and the development of GCC are welcome on our developer
4965 list at [22]gcc@gcc.gnu.org. All of [23]our lists have public
4968 Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and
4969 distribution of this entire article is permitted in any medium,
4970 provided this notice is preserved.
4972 These pages are [25]maintained by the GCC team. Last modified
4977 1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7
4978 2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted
4979 3. http://gcc.gnu.org/gcc-4.4/porting_to.html
4980 4. https://gcc.gnu.org/wiki/Graphite
4981 5. http://openmp.org/wp/openmp-specifications/
4982 6. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf
4983 7. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html
4984 8. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#id476343
4985 9. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html
4986 10. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125
4987 11. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221
4988 12. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34
4989 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1
4990 14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2
4991 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3
4992 16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4
4993 17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5
4994 18. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6
4995 19. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7
4996 20. https://gcc.gnu.org/onlinedocs/
4997 21. mailto:gcc-help@gcc.gnu.org
4998 22. mailto:gcc@gcc.gnu.org
4999 23. https://gcc.gnu.org/lists.html
5000 24. http://www.fsf.org/
5001 25. https://gcc.gnu.org/about.html
5002 26. http://validator.w3.org/check/referer
5003 ======================================================================
5004 http://gcc.gnu.org/gcc-4.3/index.html
5005 GCC 4.3 Release Series
5009 The [1]GNU project and the GCC developers are pleased to announce the
5010 release of GCC 4.3.6.
5012 This release is a bug-fix release, containing fixes for regressions in
5013 GCC 4.3.5 relative to previous releases of GCC.
5018 Jun 27, 2011 ([2]changes)
5021 May 22, 2010 ([3]changes)
5024 August 4, 2009 ([4]changes)
5027 January 24, 2009 ([5]changes)
5030 August 27, 2008 ([6]changes)
5033 June 6, 2008 ([7]changes)
5036 March 5, 2008 ([8]changes)
5038 References and Acknowledgements
5040 GCC used to stand for the GNU C Compiler, but since the compiler
5041 supports several other languages aside from C, it now stands for the
5042 GNU Compiler Collection.
5044 A list of [9]successful builds is updated as new information becomes
5047 The GCC developers would like to thank the numerous people that have
5048 contributed new features, improvements, bug fixes, and other changes as
5049 well as test results to GCC. This [10]amazing group of volunteers is
5050 what makes GCC successful.
5052 For additional information about GCC please refer to the [11]GCC
5053 project web site or contact the [12]GCC development mailing list.
5055 To obtain GCC please use [13]our mirror sites or [14]our SVN server.
5058 For questions related to the use of GCC, please consult these web
5059 pages and the [15]GCC manuals. If that fails, the
5060 [16]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5061 web pages and the development of GCC are welcome on our developer
5062 list at [17]gcc@gcc.gnu.org. All of [18]our lists have public
5065 Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and
5066 distribution of this entire article is permitted in any medium,
5067 provided this notice is preserved.
5069 These pages are [20]maintained by the GCC team. Last modified
5074 1. http://www.gnu.org/
5075 2. http://gcc.gnu.org/gcc-4.3/changes.html
5076 3. http://gcc.gnu.org/gcc-4.3/changes.html
5077 4. http://gcc.gnu.org/gcc-4.3/changes.html
5078 5. http://gcc.gnu.org/gcc-4.3/changes.html
5079 6. http://gcc.gnu.org/gcc-4.3/changes.html
5080 7. http://gcc.gnu.org/gcc-4.3/changes.html
5081 8. http://gcc.gnu.org/gcc-4.3/changes.html
5082 9. http://gcc.gnu.org/gcc-4.3/buildstat.html
5083 10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
5084 11. http://gcc.gnu.org/index.html
5085 12. mailto:gcc@gcc.gnu.org
5086 13. http://gcc.gnu.org/mirrors.html
5087 14. http://gcc.gnu.org/svn.html
5088 15. https://gcc.gnu.org/onlinedocs/
5089 16. mailto:gcc-help@gcc.gnu.org
5090 17. mailto:gcc@gcc.gnu.org
5091 18. https://gcc.gnu.org/lists.html
5092 19. http://www.fsf.org/
5093 20. https://gcc.gnu.org/about.html
5094 21. http://validator.w3.org/check/referer
5095 ======================================================================
5096 http://gcc.gnu.org/gcc-4.3/changes.html
5097 GCC 4.3 Release Series
5098 Changes, New Features, and Fixes
5100 The latest release in the 4.3 release series is [1]GCC 4.3.5.
5104 * GCC requires the [2]GMP and [3]MPFR libraries for building all the
5105 various front-end languages it supports. See the [4]prerequisites
5106 page for version requirements.
5107 * ColdFire targets now treat long double as having the same format as
5108 double. In earlier versions of GCC, they used the 68881 long double
5110 * The m68k-uclinux target now uses the same calling conventions as
5111 m68k-linux-gnu. You can select the original calling conventions by
5112 configuring for m68k-uclinuxoldabi instead. Note that
5113 m68k-uclinuxoldabi also retains the original 80-bit long double on
5115 * The -fforce-mem option has been removed because it has had no
5116 effect in the last few GCC releases.
5117 * The i386 -msvr3-shlib option has been removed since it is no longer
5119 * Fastcall for i386 has been changed not to pass aggregate arguments
5120 in registers, following Microsoft compilers.
5121 * Support for the AOF assembler has been removed from the ARM back
5122 end; this affects only the targets arm-semi-aof and armel-semi-aof,
5123 which are no longer recognized. We removed these targets without a
5124 deprecation period because we discovered that they have been
5125 unusable since GCC 4.0.0.
5126 * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*)
5127 has been removed. This support had been deprecated since GCC 4.0.0.
5128 * Support for a number of older systems and recently unmaintained or
5129 untested target ports of GCC has been declared obsolete in GCC 4.3.
5130 Unless there is activity to revive them, the next release of GCC
5131 will have their sources permanently removed.
5132 All GCC ports for the following processor architectures have been
5135 The following aliases for processor architectures have been
5136 declared obsolete. Users should use the indicated generic target
5137 names instead, with compile-time options such as -mcpu or
5138 configure-time options such as --with-cpu to control the
5139 configuration more precisely.
5140 + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-*
5142 + parisc*-*-* (use hppa*-*-* instead).
5143 + m680[012]0-*-* (use m68k-*-* instead).
5144 All GCC ports for the following operating systems have been
5148 + GNU/Linux using the a.out object format (*-*-linux*aout*)
5149 + GNU/Linux using version 1 of the GNU C Library
5151 + Solaris versions before Solaris 7 (*-*-solaris2.[0-6],
5152 *-*-solaris2.[0-6].*)
5153 + Miscellaneous System V (*-*-sysv*)
5154 + WindISS (*-*-windiss*)
5155 Also, those for some individual systems on particular architectures
5156 have been obsoleted:
5157 + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*)
5158 + CRIS with a.out object format (cris-*-aout)
5159 + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*)
5160 + OSF/1 on PA-RISC (hppa1.1-*-osf*)
5161 + PRO on PA-RISC (hppa1.1-*-pro*)
5162 + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*,
5163 i[34567]86-sequent-sysv4*)
5164 + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*)
5165 + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host
5166 was previously [5]removed in 2001, leaving only the support
5167 for UWIN as a target now being deprecated)
5168 + ChorusOS on PowerPC (powerpc-*-chorusos*)
5169 + All VAX configurations apart from NetBSD and OpenBSD
5170 (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*)
5171 * The [6]-Wconversion option has been modified. Its purpose now is to
5172 warn for implicit conversions that may alter a value. This new
5173 behavior is available for both C and C++. Warnings about
5174 conversions between signed and unsigned integers can be disabled by
5175 using -Wno-sign-conversion. In C++, they are disabled by default
5176 unless -Wsign-conversion is explicitly requested. The old behavior
5177 of -Wconversion, that is, warn for prototypes causing a type
5178 conversion that is different from what would happen to the same
5179 argument in the absence of a prototype, has been moved to a new
5180 option -Wtraditional-conversion, which is only available for C.
5181 * The -m386, -m486, -mpentium and -mpentiumpro tuning options have
5182 been removed because they were deprecated for more than 3 GCC major
5183 releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or
5184 -mtune=pentiumpro as a replacement.
5185 * The -funsafe-math-optimizations option now automatically turns on
5186 -fno-trapping-math in addition to -fno-signed-zeros, as it enables
5187 reassociation and thus may introduce or remove traps.
5188 * The -ftree-vectorize option is now on by default under -O3. In
5189 order to generate code for a SIMD extension, it has to be enabled
5190 as well: use -maltivec for PowerPC platforms and -msse/-msse2 for
5192 * More information on porting to GCC 4.3 from previous versions of
5193 GCC can be found in the [7]porting guide for this release.
5195 General Optimizer Improvements
5197 * The GCC middle-end has been integrated with the [8]MPFR library.
5198 This allows GCC to evaluate and replace at compile-time calls to
5199 built-in math functions having constant arguments with their
5200 mathematically equivalent results. In making use of [9]MPFR, GCC
5201 can generate correct results regardless of the math library
5202 implementation or floating point precision of the host platform.
5203 This also allows GCC to generate identical results regardless of
5204 whether one compiles in native or cross-compile configurations to a
5205 particular target. The following built-in functions take advantage
5206 of this new capability: acos, acosh, asin, asinh, atan2, atan,
5207 atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1,
5208 fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10,
5209 log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh,
5210 tan, tanh, tgamma, y0, y1 and yn. The float and long double
5211 variants of these functions (e.g. sinf and sinl) are also handled.
5212 The sqrt and cabs functions with constant arguments were already
5213 optimized in prior GCC releases. Now they also use [10]MPFR.
5214 * A new forward propagation pass on RTL was added. The new pass
5215 replaces several slower transformations, resulting in compile-time
5216 improvements as well as better code generation in some cases.
5217 * A new command-line switch -frecord-gcc-switches has been added to
5218 GCC, although it is only enabled for some targets. The switch
5219 causes the command line that was used to invoke the compiler to be
5220 recorded into the object file that is being created. The exact
5221 format of this recording is target and binary file format
5222 dependent, but it usually takes the form of a note section
5223 containing ASCII text. The switch is related to the -fverbose-asm
5224 switch, but that one only records the information in the assembler
5225 output file as comments, so the information never reaches the
5227 * The inliner heuristic is now aware of stack frame consumption. New
5228 command-line parameters --param large-stack-frame and --param
5229 large-stack-frame-growth can be used to limit stack frame size
5230 growth caused by inlining.
5231 * During feedback directed optimizations, the expected block size the
5232 memcpy, memset and bzero functions operate on is discovered and for
5233 cases of commonly used small sizes, specialized inline code is
5235 * __builtin_expect no longer requires its argument to be a compile
5237 * Interprocedural optimization was reorganized to work on functions
5238 in SSA form. This enables more precise and cheaper dataflow
5239 analysis and makes writing interprocedural optimizations easier.
5240 The following improvements have been implemented on top of this
5242 + Pre-inline optimization: Selected local optimization passes
5243 are run before the inliner (and other interprocedural passes)
5244 are executed. This significantly improves the accuracy of code
5245 growth estimates used by the inliner and reduces the overall
5246 memory footprint for large compilation units.
5247 + Early inlining (a simple bottom-up inliner pass inlining only
5248 functions whose body is smaller than the expected call
5249 overhead) is now executed with the early optimization passes,
5250 thus inlining already optimized function bodies into an
5251 unoptimized function that is subsequently optimized by early
5252 optimizers. This enables the compiler to quickly eliminate
5253 abstraction penalty in C++ programs.
5254 + Interprocedural constant propagation now operate on SSA form
5255 increasing accuracy of the analysis.
5256 * A new internal representation for GIMPLE statements has been
5257 contributed, resulting in compile-time memory savings.
5258 * The vectorizer was enhanced to support vectorization of outer
5259 loops, intra-iteration parallelism (loop-aware SLP), vectorization
5260 of strided accesses and loops with multiple data-types. Run-time
5261 dependency testing using loop versioning was added. The cost model,
5262 turned on by -fvect-cost-model, was developed.
5264 New Languages and Language specific improvements
5266 * We have added new command-line options
5267 -finstrument-functions-exclude-function-list and
5268 -finstrument-functions-exclude-file-list. They provide more control
5269 over which functions are annotated by the -finstrument-functions
5274 * Implicit conversions between generic vector types are now only
5275 permitted when the two vectors in question have the same number of
5276 elements and compatible element types. (Note that the restriction
5277 involves compatible element types, not implicitly-convertible
5278 element types: thus, a vector type with element type int may not be
5279 implicitly converted to a vector type with element type unsigned
5280 int.) This restriction, which is in line with specifications for
5281 SIMD architectures such as AltiVec, may be relaxed using the flag
5282 -flax-vector-conversions. This flag is intended only as a
5283 compatibility measure and should not be used for new code.
5284 * -Warray-bounds has been added and is now enabled by default for
5285 -Wall . It produces warnings for array subscripts that can be
5286 determined at compile time to be always out of bounds.
5287 -Wno-array-bounds will disable the warning.
5288 * The constructor and destructor function attributes now accept
5289 optional priority arguments which control the order in which the
5290 constructor and destructor functions are run.
5291 * New [11]command-line options -Wtype-limits,
5292 -Wold-style-declaration, -Wmissing-parameter-type, -Wempty-body,
5293 -Wclobbered and -Wignored-qualifiers have been added for finer
5294 control of the diverse warnings enabled by -Wextra.
5295 * A new function attribute alloc_size has been added to mark up
5296 malloc style functions. For constant sized allocations this can be
5297 used to find out the size of the returned pointer using the
5298 __builtin_object_size() function for buffer overflow checking and
5299 similar. This supplements the already built-in malloc and calloc
5300 constant size handling.
5301 * Integer constants written in binary are now supported as a GCC
5302 extension. They consist of a prefix 0b or 0B, followed by a
5303 sequence of 0 and 1 digits.
5304 * A new predefined macro __COUNTER__ has been added. It expands to
5305 sequential integral values starting from 0. In conjunction with the
5306 ## operator, this provides a convenient means to generate unique
5308 * A new command-line option -fdirectives-only has been added. It
5309 enables a special preprocessing mode which improves the performance
5310 of applications like distcc and ccache.
5311 * Fixed-point data types and operators have been added. They are
5312 based on Chapter 4 of the Embedded-C specification (n1169.pdf).
5313 Currently, only MIPS targets are supported.
5314 * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732,
5315 N1241, is now supported as a GCC extension to C for targets
5316 i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu,
5317 and x86_64-*-linux-gnu. The feature introduces new data types
5318 _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF,
5323 * [12]Experimental support for the upcoming ISO C++ standard, C++0x.
5324 * -Wc++0x-compat has been added and is now enabled by default for
5325 -Wall. It produces warnings for constructs whose meaning differs
5326 between ISO C++ 1998 and C++0x.
5327 * The -Wparentheses option now works for C++ as it does for C. It
5328 warns if parentheses are omitted when operators with confusing
5329 precedence are nested. It also warns about ambiguous else
5330 statements. Since -Wparentheses is enabled by -Wall, this may cause
5331 additional warnings with existing C++ code which uses -Wall. These
5332 new warnings may be disabled by using -Wall -Wno-parentheses.
5333 * The -Wmissing-declarations now works for C++ as it does for C.
5334 * The -fvisibility-ms-compat flag was added, to make it easier to
5335 port larger projects using shared libraries from Microsoft's Visual
5336 Studio to ELF and Mach-O systems.
5337 * C++ attribute handling has been overhauled for template arguments
5338 (ie dependent types). In particular, __attribute__((aligned(T)));
5339 works for C++ types.
5341 Runtime Library (libstdc++)
5343 * [13]Experimental support for the upcoming ISO C++ standard, C++0x.
5344 * Support for TR1 mathematical special functions and regular
5345 expressions. ([14]Implementation status of TR1)
5346 * Default what implementations give more elaborate exception strings
5347 for bad_cast, bad_typeid, bad_exception, and bad_alloc.
5348 * Header dependencies have been streamlined, reducing unnecessary
5349 includes and pre-processed bloat.
5350 * Variadic template implementations of items in <tuple> and
5352 * An experimental [15]parallel mode has been added. This is a
5353 parallel implementation of many C++ Standard library algorithms,
5354 like std::accumulate, std::for_each, std::transform, or std::sort,
5355 to give but four examples. These algorithms can be substituted for
5356 the normal (sequential) libstdc++ algorithms on a piecemeal basis,
5357 or all existing algorithms can be transformed via the
5358 -D_GLIBCXX_PARALLEL macro.
5359 * Debug mode versions of classes in <unordered_set> and
5361 * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are
5362 now <backward/hash_set> and <backward/hash_map>. This code:
5363 #include <ext/hash_set>
5364 __gnu_cxx::hash_set<int> s;
5366 Can be transformed (in order of preference) to:
5367 #include <tr1/unordered_set>
5368 std::tr1::unordered_set<int> s;
5371 #include <backward/hash_set>
5372 __gnu_cxx::hash_set<int> s;
5374 Similar transformations apply to __gnu_cxx::hash_map,
5375 __gnu_cxx::hash_multimap, __gnu_cxx::hash_set,
5376 __gnu_cxx::hash_multiset.
5380 * Due to the fact that the [16]GMP and [17]MPFR libraries are
5381 required for all languages, Fortran is no longer special in this
5382 regard and is available by default.
5383 * The [18]-fexternal-blas option has been added, which generates
5384 calls to BLAS routines for intrinsic matrix operations such as
5385 matmul rather than using the built-in algorithms.
5386 * Support to give a backtrace (compiler flag -fbacktrace or
5387 environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems
5388 only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a
5389 run-time error occured.
5390 * GNU Fortran now defines __GFORTRAN__ when it runs the C
5392 * The [19]-finit-local-zero, -finit-real, -finit-integer,
5393 -finit-character, and -finit-logical options have been added, which
5394 can be used to initialize local variables.
5395 * The intrinsic procedures [20]GAMMA and [21]LGAMMA have been added,
5396 which calculate the Gamma function and its logarithm. Use EXTERNAL
5397 gamma if you want to use your own gamma function.
5398 * GNU Fortran now regards the backslash character as literal (as
5399 required by the Fortran 2003 standard); using [22]-fbackslash GNU
5400 Fortran interprets backslashes as C-style escape characters.
5401 * The [23]interpretation of binary, octal and hexadecimal (BOZ)
5402 literal constants has been changed. Before they were always
5403 interpreted as integer; now they are bit-wise transferred as
5404 argument of INT, REAL, DBLE and CMPLX as required by the Fortran
5405 2003 standard, and for real and complex variables in DATA
5406 statements or when directly assigned to real and complex variables.
5407 Everywhere else and especially in expressions they are still
5408 regarded as integer constants.
5409 * Fortran 2003 support has been extended:
5410 + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE
5412 + Intrinsic module ISO_ENV_FORTRAN
5413 + Interoperability with C (ISO C Bindings)
5414 + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER
5420 * GCJ now uses the Eclipse Java compiler for its Java parsing needs.
5421 This enables the use of all 1.5 language features, and fixes most
5422 existing front end bugs.
5423 * libgcj now supports all 1.5 language features which require runtime
5424 support: foreach, enum, annotations, generics, and auto-boxing.
5425 * We've made many changes to the tools shipped with gcj.
5426 + The old jv-scan tool has been removed. This tool never really
5427 worked properly. There is no replacement.
5428 + gcjh has been rewritten. Some of its more obscure options no
5429 longer work, but are still recognized in an attempt at
5430 compatibility. gjavah is a new program with similar
5431 functionality but different command-line options.
5432 + grmic and grmiregistry have been rewritten. grmid has been
5434 + gjar replaces the old fastjar.
5435 + gjarsigner (used for signing jars), gkeytool (used for key
5436 management), gorbd (for CORBA), gserialver (computes
5437 serialization UIDs), and gtnameserv (also for CORBA) are now
5439 * The ability to dump the contents of the java run time heap to a
5440 file for off-line analysis has been added. The heap dumps may be
5441 analyzed with the new gc-analyze tool. They may be generated on
5442 out-of-memory conditions or on demand and are controlled by the new
5443 run time class gnu.gcj.util.GCInfo.
5444 * java.util.TimeZone can now read files from /usr/share/zoneinfo to
5445 provide correct, updated, timezone information. This means that
5446 packagers no longer have to update libgcj when a time zone change
5449 New Targets and Target Specific Improvements
5453 * Tuning for Intel Core 2 processors is available via -mtune=core2
5455 * Tuning for AMD Geode processors is available via -mtune=geode and
5457 * Code generation of block move (memcpy) and block set (memset) was
5458 rewritten. GCC can now pick the best algorithm (loop, unrolled
5459 loop, instruction with rep prefix or a library call) based on the
5460 size of the block being copied and the CPU being optimized for. A
5461 new option -minline-stringops-dynamically has been added. With this
5462 option string operations of unknown size are expanded such that
5463 small blocks are copied by in-line code, while for large blocks a
5464 library call is used. This results in faster code than
5465 -minline-all-stringops when the library implementation is capable
5466 of using cache hierarchy hints. The heuristic choosing the
5467 particular algorithm can be overwritten via -mstringop-strategy.
5468 Newly also memset of values different from 0 is inlined.
5469 * GCC no longer places the cld instruction before string operations.
5470 Both i386 and x86-64 ABI documents mandate the direction flag to be
5471 clear at the entry of a function. It is now invalid to set the flag
5472 in asm statement without reseting it afterward.
5473 * Support for SSSE3 built-in functions and code generation are
5474 available via -mssse3.
5475 * Support for SSE4.1 built-in functions and code generation are
5476 available via -msse4.1.
5477 * Support for SSE4.2 built-in functions and code generation are
5478 available via -msse4.2.
5479 * Both SSE4.1 and SSE4.2 support can be enabled via -msse4.
5480 * A new set of options -mpc32, -mpc64 and -mpc80 have been added to
5481 allow explicit control of x87 floating point precision.
5482 * Support for __float128 (TFmode) IEEE quad type and corresponding
5483 TCmode IEEE complex quad type is available via the soft-fp library
5484 on x86_64 targets. This includes basic arithmetic operations
5485 (addition, subtraction, negation, multiplication and division) on
5486 __float128 real and TCmode complex values, the full set of IEEE
5487 comparisons between __float128 values, conversions to and from
5488 float, double and long double floating point types, as well as
5489 conversions to and from signed or unsigned integer, signed or
5490 unsigned long integer and signed or unsigned quad (TImode) integer
5491 types. Additionally, all operations generate the full set of IEEE
5492 exceptions and support the full set of IEEE rounding modes.
5493 * GCC can now utilize the ACML library for vectorizing calls to a set
5494 of C99 functions on x86_64 if -mveclibabi=acml is specified and you
5495 link to an ACML ABI compatible library.
5499 * Compiler and Library support for Thumb-2 and the ARMv7 architecture
5506 * Compiler and Library support for the CRIS v32 architecture, as
5507 found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been
5510 Configuration changes
5512 * The cris-*-elf target now includes support for CRIS v32, including
5513 libraries, through the -march=v32 option.
5514 * A new crisv32-*-elf target defaults to generate code for CRIS v32.
5515 * A new crisv32-*-linux* target defaults to generate code for CRIS
5517 * The cris-*-aout target has been obsoleted.
5519 Improved support for built-in functions
5521 * GCC can now use the lz and swapwbr instructions to implement the
5522 __builtin_clz, __builtin_ctz and __builtin_ffs family of functions.
5523 * __builtin_bswap32 is now implemented using the swapwb instruction,
5530 * Support for several new ColdFire processors has been added. You can
5531 generate code for them using the new -mcpu option.
5532 * All targets now support ColdFire processors.
5533 * m68k-uclinux targets have improved support for C++ constructors and
5534 destructors, and for shared libraries.
5535 * It is now possible to set breakpoints on the first or last line of
5536 a function, even if there are no statements on that line.
5540 * Support for sibling calls has been added.
5541 * More use is now made of the ColdFire mov3q instruction.
5542 * __builtin_clz is now implemented using the ff1 ColdFire
5543 instruction, when available.
5544 * GCC now honors the -m68010 option. 68010 code now uses clr rather
5545 than move to zero volatile memory.
5546 * 68020 targets and above can now use symbol(index.size*scale)
5547 addresses for indexed array accesses. Earlier compilers would
5548 always load the symbol into a base register first.
5550 Configuration changes
5552 * All m68k and ColdFire targets now allow the default processor to be
5553 set at configure time using --with-cpu.
5554 * A --with-arch configuration option has been added. This option
5555 allows you to restrict a target to ColdFire or non-ColdFire
5560 * An __mcfv*__ macro is now defined for all ColdFire targets.
5561 (Earlier versions of GCC only defined __mcfv4e__.)
5562 * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added.
5563 * All targets now define __mc68010 and __mc68010__ when generating
5566 Command-line changes
5568 * New command-line options -march, -mcpu, -mtune and -mhard-float
5569 have been added. These options apply to both m68k and ColdFire
5571 * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative
5572 versions of -mshort, etc.
5573 * -fforce-addr has been removed. It is now ignored by the compiler.
5577 * ColdFire targets now try to maintain a 4-byte-aligned stack where
5579 * m68k-uclinux targets now try to avoid situations that lead to the
5580 load-time error: BINFMT_FLAT: reloc outside program.
5584 Changes to existing configurations
5586 * libffi and libjava now support all three GNU/Linux ABIs: o32, n32
5587 and n64. Every GNU/Linux configuration now builds these libraries
5589 * GNU/Linux configurations now generate -mno-shared code unless
5590 overridden by -fpic, -fPIC, -fpie or -fPIE.
5591 * mipsisa32*-linux-gnu configurations now generate hard-float code by
5592 default, just like other mipsisa32* and mips*-linux-gnu
5593 configurations. You can build a soft-float version of any
5594 mips*-linux-gnu configuration by passing --with-float=soft to
5596 * mips-wrs-vxworks now supports run-time processes (RTPs).
5598 Changes to existing command-line options
5600 * The -march and -mtune options no longer accept 24k as a processor
5601 name. Please use 24kc, 24kf2_1 or 24kf1_1 instead.
5602 * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and
5603 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The
5604 options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for
5605 24kx, 24kex and 34kx.
5609 GCC now supports the following configurations:
5610 * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by
5611 default. Earlier releases also recognized this configuration, but
5612 they treated it in the same way as mipsisa32*-linux-gnu*. Note that
5613 you can customize any mips*-linux-gnu* configuration to a
5614 particular ISA or processor by passing an appropriate --with-arch
5615 option to configure.
5616 * mipsisa*-sde-elf*, which provides compatibility with MIPS
5617 Technologies' SDE toolchains. The configuration uses the SDE
5618 libraries by default, but you can use it like other newlib-based
5619 ELF configurations by passing --with-newlib to configure. It is the
5620 only configuration besides mips64vr*-elf* to build MIPS16 as well
5621 as non-MIPS16 libraries.
5622 * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf*
5623 configuration, but uses the o32 and o64 ABIs instead of the 32-bit
5624 and 64-bit forms of the EABI.
5626 New processors and application-specific extensions
5628 * Support for the SmartMIPS ASE is available through the new
5630 * Support for revision 2 of the DSP ASE is available through the new
5631 -mdspr2 option. A new preprocessor macro called __mips_dsp_rev
5632 indicates the revision of the ASE in use.
5633 * Support for the 4KS and 74K families of processors is available
5634 through the -march and -mtune options.
5636 Improved support for built-in functions
5638 * GCC can now use load-linked, store-conditional and sync
5639 instructions to implement atomic built-in functions such as
5640 __sync_fetch_and_add. The memory reference must be 4 bytes wide for
5641 32-bit targets and either 4 or 8 bytes wide for 64-bit targets.
5642 * GCC can now use the clz and dclz instructions to implement the
5643 __builtin_ctz and __builtin_ffs families of functions.
5644 * There is a new __builtin___clear_cache function for flushing the
5645 instruction cache. GCC expands this function inline on MIPS32
5646 revision 2 targets, otherwise it calls the function specified by
5651 * GCC can now compile objects that contain a mixture of MIPS16 and
5652 non-MIPS16 code. There are two new attributes, mips16 and nomips16,
5653 for specifying which mode a function should use.
5654 * A new option called -minterlink-mips16 makes non-MIPS16 code
5655 link-compatible with MIPS16 code.
5656 * After many bug fixes, the long-standing MIPS16 -mhard-float support
5657 should now work fairly reliably.
5658 * GCC can now use the MIPS16e save and restore instructions.
5659 * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled
5660 with -G0 -fsection-anchors is often smaller than code compiled with
5661 -G8. However, please note that you must usually compile all objects
5662 in your application with the same -G option; see the documentation
5664 * A new option called-mcode-readable specifies which instructions are
5665 allowed to load from the code segment. -mcode-readable=yes is the
5666 default and says that any instruction may load from the code
5667 segment. The other alternatives are -mcode-readable=pcrel, which
5668 says that only PC-relative MIPS16 instructions may load from the
5669 code segment, and -mcode-readable=no, which says that no
5670 instruction may do so. Please see the documentation for more
5671 details, including example uses.
5673 Small-data improvements
5675 There are three new options for controlling small data:
5676 * -mno-extern-sdata, which disables small-data accesses for
5677 externally-defined variables. Code compiled with -Gn
5678 -mno-extern-sdata will be link-compatible with any -G setting
5679 between -G0 and -Gn inclusive.
5680 * -mno-local-sdata, which disables the use of small-data sections for
5681 data that is not externally visible. This option can be a useful
5682 way of reducing small-data usage in less performance-critical parts
5684 * -mno-gpopt, which disables the use of the $gp register while still
5685 honoring the -G limit when placing externally-visible data. This
5686 option implies -mno-extern-sdata and -mno-local-sdata and it can be
5687 useful in situations where $gp does not necessarily hold the
5690 Miscellaneous improvements
5692 * There is a new option called -mbranch-cost for tweaking the
5693 perceived cost of branches.
5694 * If GCC is configured to use a version of GAS that supports the
5695 .gnu_attribute directive, it will use that directive to record
5696 certain properties of the output code. .gnu_attribute is new to GAS
5698 * There are two new function attributes, near and far, for overriding
5699 the command-line setting of -mlong-calls on a function-by-function
5701 * -mfp64, which previously required a 64-bit target, now works with
5702 MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and
5703 mipsisa*-sde-elf* configurations provide suitable library support.
5704 * GCC now recognizes the -mdmx and -mmt options and passes them down
5705 to the assembler. It does nothing else with the options at present.
5707 SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture
5710 * Support has been added for this new architecture.
5712 RS6000 (POWER/PowerPC)
5714 * Support for the PowerPC 750CL paired-single instructions has been
5715 added with a new powerpc-*-linux*paired* target configuration. It
5716 is enabled by an associated -mpaired option and can be accessed
5717 using new built-in functions.
5718 * Support for auto-detecting architecture and system configuration to
5719 auto-select processor optimization tuning.
5720 * Support for VMX on AIX 5.3 has been added.
5721 * Support for AIX Version 6.1 has been added.
5723 S/390, zSeries and System z9
5725 * Support for the IBM System z9 EC/BC processor (z9 GA3) has been
5726 added. When using the -march=z9-ec option, the compiler will
5727 generate code making use of instructions provided by the decimal
5728 floating point facility and the floating point conversion facility
5729 (pfpo). Besides the instructions used to implement decimal floating
5730 point operations these facilities also contain instructions to move
5731 between general purpose and floating point registers and to modify
5732 and copy the sign-bit of floating point values.
5733 * When the -march=z9-ec option is used the new
5734 -mhard-dfp/-mno-hard-dfp options can be used to specify whether the
5735 decimal floating point hardware instructions will be used or not.
5736 If none of them is given the hardware support is enabled by
5738 * The -mstack-guard option can now be omitted when using stack
5739 checking via -mstack-size in order to let GCC choose a sensible
5740 stack guard value according to the frame size of each function.
5741 * Various changes to improve performance of generated code have been
5742 implemented, including:
5743 + The condition code set by an add logical with carry
5744 instruction is now available for overflow checks like: a + b +
5746 + The test data class instruction is now used to implement
5747 sign-bit and infinity checks of binary and decimal floating
5752 * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been
5757 * Stack unwinding for exception handling now uses by default a
5758 specialized version of DWARF unwinding. This is not
5759 binary-compatible with the setjmp/longjmp (sjlj) unwinding used for
5760 Xtensa with previous versions of GCC.
5761 * For Xtensa processors that include the Conditional Store option,
5762 the built-in functions for atomic memory access are now implemented
5763 using S32C1I instructions.
5764 * If the Xtensa NSA option is available, GCC will use it to implement
5765 the __builtin_ctz and __builtin_clz functions.
5767 Documentation improvements
5769 * Existing libstdc++ documentation has been edited and restructured
5770 into a single DocBook XML manual. The results can be viewed online
5773 Other significant improvements
5775 * The compiler's --help command-line option has been extended so that
5776 it now takes an optional set of arguments. These arguments restrict
5777 the information displayed to specific classes of command-line
5778 options, and possibly only a subset of those options. It is also
5779 now possible to replace the descriptive text associated with each
5780 displayed option with an indication of its current value, or for
5781 binary options, whether it has been enabled or disabled.
5782 Here are some examples. The following will display all the options
5783 controlling warning messages:
5786 Whereas this will display all the undocumented, target specific
5788 --help=target,undocumented
5790 This sequence of commands will display the binary optimizations
5791 that are enabled by -O3:
5792 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
5793 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
5794 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
5796 * The configure options --with-pkgversion and --with-bugurl have been
5797 added. These allow distributors of GCC to include a
5798 distributor-specific string in manuals and --version output and to
5799 specify the URL for reporting bugs in their versions of GCC.
5803 This is the [25]list of problem reports (PRs) from GCC's bug tracking
5804 system that are known to be fixed in the 4.3.1 release. This list might
5805 not be complete (that is, it is possible that some PRs that have been
5806 fixed are not listed here).
5808 Target Specific Changes
5814 * Starting with GCC 4.3.1, decimal floating point variables are
5815 aligned to their natural boundaries when they are passed on the
5818 Command-line changes
5820 * Starting with GCC 4.3.1, the -mcld option has been added to
5821 automatically generate a cld instruction in the prologue of
5822 functions that use string instructions. This option is used for
5823 backward compatibility on some operating systems and can be enabled
5824 by default for 32-bit x86 targets by configuring GCC with the
5825 --enable-cld configure option.
5829 This is the [26]list of problem reports (PRs) from GCC's bug tracking
5830 system that are known to be fixed in the 4.3.2 release. This list might
5831 not be complete (that is, it is possible that some PRs that have been
5832 fixed are not listed here).
5836 This is the [27]list of problem reports (PRs) from GCC's bug tracking
5837 system that are known to be fixed in the 4.3.3 release. This list might
5838 not be complete (that is, it is possible that some PRs that have been
5839 fixed are not listed here).
5843 This is the [28]list of problem reports (PRs) from GCC's bug tracking
5844 system that are known to be fixed in the 4.3.4 release. This list might
5845 not be complete (that is, it is possible that some PRs that have been
5846 fixed are not listed here).
5850 This is the [29]list of problem reports (PRs) from GCC's bug tracking
5851 system that are known to be fixed in the 4.3.5 release. This list might
5852 not be complete (that is, it is possible that some PRs that have been
5853 fixed are not listed here).
5857 This is the [30]list of problem reports (PRs) from GCC's bug tracking
5858 system that are known to be fixed in the 4.3.6 release. This list might
5859 not be complete (that is, it is possible that some PRs that have been
5860 fixed are not listed here).
5863 For questions related to the use of GCC, please consult these web
5864 pages and the [31]GCC manuals. If that fails, the
5865 [32]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5866 web pages and the development of GCC are welcome on our developer
5867 list at [33]gcc@gcc.gnu.org. All of [34]our lists have public
5870 Copyright (C) [35]Free Software Foundation, Inc. Verbatim copying and
5871 distribution of this entire article is permitted in any medium,
5872 provided this notice is preserved.
5874 These pages are [36]maintained by the GCC team. Last modified
5879 1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5
5880 2. http://gmplib.org/
5881 3. http://www.mpfr.org/
5882 4. https://gcc.gnu.org/install/prerequisites.html
5883 5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html
5884 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options
5885 7. http://gcc.gnu.org/gcc-4.3/porting_to.html
5886 8. http://www.mpfr.org/
5887 9. http://www.mpfr.org/
5888 10. http://www.mpfr.org/
5889 11. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
5890 12. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
5891 13. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html
5892 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/bk01pt01ch01.html#m anual.intro.status.standard.tr1
5893 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html
5894 16. http://gmplib.org/
5895 17. http://www.mpfr.org/
5896 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options
5897 19. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167
5898 20. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html
5899 21. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html
5900 22. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html
5901 23. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html
5902 24. https://gcc.gnu.org/onlinedocs/libstdc++/
5903 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1
5904 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2
5905 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3
5906 28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4
5907 29. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5
5908 30. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6
5909 31. https://gcc.gnu.org/onlinedocs/
5910 32. mailto:gcc-help@gcc.gnu.org
5911 33. mailto:gcc@gcc.gnu.org
5912 34. https://gcc.gnu.org/lists.html
5913 35. http://www.fsf.org/
5914 36. https://gcc.gnu.org/about.html
5915 37. http://validator.w3.org/check/referer
5916 ======================================================================
5917 http://gcc.gnu.org/gcc-4.2/index.html
5918 GCC 4.2 Release Series
5922 The [1]GNU project and the GCC developers are pleased to announce the
5923 release of GCC 4.2.4.
5925 This release is a bug-fix release, containing fixes for regressions in
5926 GCC 4.2.3 relative to previous releases of GCC.
5931 May 19, 2008 ([2]changes)
5934 February 1, 2008 ([3]changes)
5937 October 7, 2007 ([4]changes)
5940 July 18, 2007 ([5]changes)
5943 May 13, 2007 ([6]changes)
5945 References and Acknowledgements
5947 GCC used to stand for the GNU C Compiler, but since the compiler
5948 supports several other languages aside from C, it now stands for the
5949 GNU Compiler Collection.
5951 A list of [7]successful builds is updated as new information becomes
5954 The GCC developers would like to thank the numerous people that have
5955 contributed new features, improvements, bug fixes, and other changes as
5956 well as test results to GCC. This [8]amazing group of volunteers is
5957 what makes GCC successful.
5959 For additional information about GCC please refer to the [9]GCC project
5960 web site or contact the [10]GCC development mailing list.
5962 To obtain GCC please use [11]our mirror sites or [12]our SVN server.
5965 For questions related to the use of GCC, please consult these web
5966 pages and the [13]GCC manuals. If that fails, the
5967 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
5968 web pages and the development of GCC are welcome on our developer
5969 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
5972 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
5973 distribution of this entire article is permitted in any medium,
5974 provided this notice is preserved.
5976 These pages are [18]maintained by the GCC team. Last modified
5981 1. http://www.gnu.org/
5982 2. http://gcc.gnu.org/gcc-4.2/changes.html
5983 3. http://gcc.gnu.org/gcc-4.2/changes.html
5984 4. http://gcc.gnu.org/gcc-4.2/changes.html
5985 5. http://gcc.gnu.org/gcc-4.2/changes.html
5986 6. http://gcc.gnu.org/gcc-4.2/changes.html
5987 7. http://gcc.gnu.org/gcc-4.2/buildstat.html
5988 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
5989 9. http://gcc.gnu.org/index.html
5990 10. mailto:gcc@gcc.gnu.org
5991 11. http://gcc.gnu.org/mirrors.html
5992 12. http://gcc.gnu.org/svn.html
5993 13. https://gcc.gnu.org/onlinedocs/
5994 14. mailto:gcc-help@gcc.gnu.org
5995 15. mailto:gcc@gcc.gnu.org
5996 16. https://gcc.gnu.org/lists.html
5997 17. http://www.fsf.org/
5998 18. https://gcc.gnu.org/about.html
5999 19. http://validator.w3.org/check/referer
6000 ======================================================================
6001 http://gcc.gnu.org/gcc-4.2/changes.html
6002 GCC 4.2 Release Series
6003 Changes, New Features, and Fixes
6007 * GCC no longer accepts the -fshared-data option. This option has had
6008 no effect in any GCC 4 release; the targets to which the option
6009 used to apply had been removed before GCC 4.0.
6011 General Optimizer Improvements
6013 * New command-line options specify the possible relationships among
6014 parameters and between parameters and global data. For example,
6015 -fargument-noalias-anything specifies that arguments do not alias
6017 Each language will automatically use whatever option is required by
6018 the language standard. You should not need to use these options
6021 New Languages and Language specific improvements
6023 * [1]OpenMP is now supported for the C, C++ and Fortran compilers.
6024 * New command-line options -fstrict-overflow and -Wstrict-overflow
6025 have been added. -fstrict-overflow tells the compiler that it may
6026 assume that the program follows the strict signed overflow
6027 semantics permitted for the language: for C and C++ this means that
6028 the compiler may assume that signed overflow does not occur. For
6029 example, a loop like
6030 for (i = 1; i > 0; i *= 2)
6032 is presumably intended to continue looping until i overflows. With
6033 -fstrict-overflow, the compiler may assume that signed overflow
6034 will not occur, and transform this into an infinite loop.
6035 -fstrict-overflow is turned on by default at -O2, and may be
6036 disabled via -fno-strict-overflow. The -Wstrict-overflow option may
6037 be used to warn about cases where the compiler assumes that signed
6038 overflow will not occur. It takes five different levels:
6039 -Wstrict-overflow=1 to 5. See the [2]documentation for details.
6040 -Wstrict-overflow=1 is enabled by -Wall.
6041 * The new command-line option -fno-toplevel-reorder directs GCC to
6042 emit top-level functions, variables, and asm statements in the same
6043 order that they appear in the input file. This is intended to
6044 support existing code which relies on a particular ordering (for
6045 example, code which uses top-level asm statements to switch
6046 sections). For new code, it is generally better to use function and
6047 variable attributes. The -fno-toplevel-reorder option may be used
6048 for most cases which currently use -fno-unit-at-a-time. The
6049 -fno-unit-at-a-time option will be removed in some future version
6050 of GCC. If you know of a case which requires -fno-unit-at-a-time
6051 which is not fixed by -fno-toplevel-reorder, please open a bug
6056 * The pragma redefine_extname will now macro expand its tokens for
6057 compatibility with SunPRO.
6058 * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct
6059 GCC to handle inline functions as specified in the C99 standard. In
6060 preparation for this, GCC 4.2 will warn about any use of non-static
6061 inline functions in gnu99 or c99 mode. This new warning may be
6062 disabled with the new gnu_inline function attribute or the new
6063 -fgnu89-inline command-line option. Also, GCC 4.2 and later will
6064 define one of the preprocessor macros __GNUC_GNU_INLINE__ or
6065 __GNUC_STDC_INLINE__ to indicate the semantics of inline functions
6066 in the current compilation.
6067 * A new command-line option -Waddress has been added to warn about
6068 suspicious uses of memory addresses as, for example, using the
6069 address of a function in a conditional expression, and comparisons
6070 against the memory address of a string literal. This warning is
6075 * C++ visibility handling has been overhauled.
6076 Restricted visiblity is propagated from classes to members, from
6077 functions to local statics, and from templates and template
6078 arguments to instantiations, unless the latter has explicitly
6079 declared visibility.
6080 The visibility attribute for a class must come between the
6081 class-key and the name, not after the closing brace.
6082 Attributes are now allowed for enums and elaborated-type-specifiers
6083 that only declare a type.
6084 Members of the anonymous namespace are now local to a particular
6085 translation unit, along with any other declarations which use them,
6086 though they are still treated as having external linkage for
6088 * The (undocumented) extension which permitted templates with default
6089 arguments to be bound to template template parameters with fewer
6090 parameters has been removed. For example:
6091 template <template <typename> class C>
6092 void f(C<double>) {}
6094 template <typename T, typename U = int>
6097 template void f(S<double>);
6099 is no longer accepted by G++. The reason this code is not accepted
6100 is that S is a template with two parameters; therefore, it cannot
6101 be bound to C which has only one parameter.
6102 * The <?, >?, <?=, and >?= operators, deprecated in previous GCC
6103 releases, have been removed.
6104 * The command-line option -fconst-strings, deprecated in previous GCC
6105 releases, has been removed.
6106 * The configure variable enable-__cxa_atexit is now enabled by
6107 default for more targets. Enabling this variable is necessary in
6108 order for static destructors to be executed in the correct order,
6109 but it depends upon the presence of a non-standard C library in the
6110 target library in order to work. The variable is now enabled for
6111 more targets which are known to have suitable C libraries.
6112 * -Wextra will produce warnings for if statements with a semicolon as
6113 the only body, to catch code like:
6118 To suppress the warning in valid cases, use { } instead.
6119 * The C++ frontend now also produces strict aliasing warnings when
6120 -fstrict-aliasing -Wstrict-aliasing is in effect.
6122 Runtime Library (libstdc++)
6124 * Added support for TR1 <random>, <complex>, and C compatibility
6125 headers. In addition, a lock-free version of shared_ptr was
6126 contributed as part of Phillip Jordan's Google Summer of Code
6127 project on lock-free containers. ([3]Implementation status of TR1)
6128 * In association with the Summer of Code work on lock-free
6129 containers, the interface for atomic builtins was adjusted,
6130 creating simpler alternatives for non-threaded code paths. Also,
6131 usage was consolidated and all elements were moved from namespace
6132 std to namespace__gnu_cxx. Affected interfaces are the functions
6133 __exchange_and_add, __atomic_add, and the objects __mutex,
6134 __recursive_mutex, and __scoped_lock.
6135 * Support for versioning weak symbol names via namespace association
6136 was added. However, as this changes the names of exported symbols,
6137 this is turned off by default in the current ABI. Intrepid users
6138 can enable this feature by using
6139 --enable-symvers=gnu-versioned-namespace during configuration.
6140 * Revised, simplified, and expanded policy-based associative
6141 containers, including data types for tree and trie forms
6142 (basic_tree, tree, trie), lists (list_update), and both
6143 collision-chaining and probing hash-based containers
6144 (basic_hash_table, cc_hash_table, gp_hash_table). More details per
6145 the [4]documentation.
6146 * The implementation of the debug mode was modified, whereby the
6147 debug namespaces were nested inside of namespace std and namespace
6148 __gnu_cxx in order to resolve some long standing corner cases
6149 involving name lookup. Debug functionality from the policy-based
6150 data structures was consolidated and enabled with the single macro,
6151 _GLIBCXX_DEBUG. See PR 26142 for more information.
6152 * Added extensions for type traits: __conditional_type,
6153 __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if.
6154 * Added a typelist implementation for compile-time meta-programming.
6155 Elements for typelist construction and operation can be found
6156 within namespace __gnu_cxx::typelist.
6157 * Added a new allocator, __gnu_cxx::throw_allocator, for testing
6159 * Enabled library-wide visibility control, allowing -fvisibility to
6161 * Consolidated all nested namespaces and the conversion of
6162 __gnu_internal implementation-private details to anonymous
6163 namespaces whenever possible.
6164 * Implemented LWG resolutions DR 431 and DR 538.
6168 * Support for allocatable components has been added (TR 15581 and
6170 * Support for the Fortran 2003 streaming IO extension has been added.
6171 * The GNU Fortran compiler now uses 4-byte record markers by default
6172 for unformatted files to be compatible with g77 and most other
6173 compilers. The implementation allows for records greater than 2 GB
6174 and is compatible with several other compilers. Older versions of
6175 gfortran used 8-byte record markers by default (on most systems).
6176 In order to change the length of the record markers, e.g. to read
6177 unformatted files created by older gfortran versions, the
6178 [5]-frecord-marker=8 option can be used.
6182 * A new command-line option -static-libgcj has been added for targets
6183 that use a linker compatible with GNU Binutils. As its name
6184 implies, this causes libgcj to be linked statically. In some cases
6185 this causes the resulting executable to start faster and use less
6186 memory than if the shared version of libgcj were used. However
6187 caution should be used as it can also cause essential parts of the
6188 library to be omitted. Some of these issues are discussed in:
6189 [6]https://gcc.gnu.org/wiki/Statically_linking_libgcj
6190 * fastjar is no longer bundled with GCC. To build libgcj, you will
6191 need either InfoZIP (both zip and unzip) or an external jar
6192 program. In the former case, the GCC build will install a jar shell
6193 script that is based on InfoZIP and provides the same functionality
6196 New Targets and Target Specific Improvements
6200 * -mtune=generic can now be used to generate code running well on
6201 common x86 chips. This includes AMD Athlon, AMD Opteron, Intel
6202 Pentium-M, Intel Pentium 4 and Intel Core 2.
6203 * -mtune=native and -march=native will produce code optimized for the
6204 host architecture as detected using the cpuid instruction.
6205 * Added a new command-line option -fstackrealign and and
6206 __attribute__ ((force_align_arg_pointer)) to realign the stack at
6207 runtime. This allows functions compiled with a vector-aligned stack
6208 to be invoked from legacy objects that keep only word-alignment.
6212 * The default CPU setting has been changed from V7 to V9 in 32-bit
6213 mode on Solaris 7 and above. This is already the case in 64-bit
6214 mode. It can be overridden by specifying --with-cpu at configure
6216 * Back-end support of built-in functions for atomic memory access has
6218 * Support for the Sun UltraSPARC T1 (Niagara) processor has been
6223 * Various bug fixes have made some functions (notably, functions
6224 returning structures) incompatible with previous releases.
6225 Recompiling all libraries is recommended. Note that code quality
6226 has considerably improved since 4.1, making a recompile even more
6231 * Added support for the Broadcom SB-1A core.
6235 * Added support for IA-64 data and control speculation. By default
6236 speculation is enabled only during second scheduler pass. A number
6237 of machine flags was introduced to control the usage of speculation
6238 for both scheduler passes.
6242 * Added Java language support (libffi and libjava) for 32-bit HP-UX
6247 Documentation improvements
6251 * A make pdf target has been added to the top-level makefile,
6252 enabling automated production of PDF documentation files.
6253 (Front-ends external to GCC should modify their Make-lang.in file
6254 to add a lang.pdf: target.)
6256 Other significant improvements
6258 Build system improvements
6260 * All the components of the compiler are now bootstrapped by default.
6261 This improves the resilience to bugs in the system compiler or
6262 binary compatibility problems, as well as providing better testing
6263 of GCC 4.2 itself. In addition, if you build the compiler from a
6264 combined tree, the assembler, linker, etc. will also be
6265 bootstrapped (i.e. built with themselves).
6266 You can disable this behavior, and go back to the pre-GCC 4.2 set
6267 up, by configuring GCC with --disable-bootstrap.
6268 * The rules that configure follows to find target tools resemble more
6269 closely the locations that the built compiler will search. In
6270 addition, you can use the new configure option --with-target-tools
6271 to specify where to find the target tools used during the build,
6272 without affecting what the built compiler will use.
6273 This can be especially useful when building packages of GCC. For
6274 example, you may want to build GCC with GNU as or ld, even if the
6275 resulting compiler to work with the native assembler and linker. To
6276 do so, you can use --with-target-tools to point to the native
6279 Incompatible changes to the build system
6281 * Front-ends external to GCC should modify their Make-lang.in file to
6282 replace double-colon rules (e.g. dvi::) with normal rules (like
6283 lang.dvi:). Front-end makefile hooks do not use double-colon rules
6285 * Up to GCC 4.1, a popular way to specify the target tools used
6286 during the build was to create directories named gas, binutils,
6287 etc. in the build tree, and create links to the tools from there.
6288 This does not work any more when the compiler is bootstrapped. The
6289 new configure option --with-target-tools provides a better way to
6290 achieve the same effect, and works for all native and cross
6294 For questions related to the use of GCC, please consult these web
6295 pages and the [7]GCC manuals. If that fails, the
6296 [8]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6297 web pages and the development of GCC are welcome on our developer
6298 list at [9]gcc@gcc.gnu.org. All of [10]our lists have public
6301 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and
6302 distribution of this entire article is permitted in any medium,
6303 provided this notice is preserved.
6305 These pages are [12]maintained by the GCC team. Last modified
6310 1. http://gcc.gnu.org/projects/gomp/
6311 2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html
6312 3. https://gcc.gnu.org/onlinedocs/libstdc++/manual/bk01pt01ch01.html#manual.intro.status.standard.tr1
6313 4. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html
6314 5. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html
6315 6. https://gcc.gnu.org/wiki/Statically_linking_libgcj
6316 7. https://gcc.gnu.org/onlinedocs/
6317 8. mailto:gcc-help@gcc.gnu.org
6318 9. mailto:gcc@gcc.gnu.org
6319 10. https://gcc.gnu.org/lists.html
6320 11. http://www.fsf.org/
6321 12. https://gcc.gnu.org/about.html
6322 13. http://validator.w3.org/check/referer
6323 ======================================================================
6324 http://gcc.gnu.org/gcc-4.1/index.html
6325 GCC 4.1 Release Series
6329 The [1]GNU project and the GCC developers are pleased to announce the
6330 release of GCC 4.1.2.
6332 This release is a bug-fix release, containing fixes for regressions in
6333 GCC 4.1.1 relative to previous releases of GCC.
6338 February 13, 2007 ([2]changes)
6341 May 24, 2006 ([3]changes)
6344 February 28, 2006 ([4]changes)
6346 References and Acknowledgements
6348 GCC used to stand for the GNU C Compiler, but since the compiler
6349 supports several other languages aside from C, it now stands for the
6350 GNU Compiler Collection.
6352 A list of [5]successful builds is updated as new information becomes
6355 The GCC developers would like to thank the numerous people that have
6356 contributed new features, improvements, bug fixes, and other changes as
6357 well as test results to GCC. This [6]amazing group of volunteers is
6358 what makes GCC successful.
6360 For additional information about GCC please refer to the [7]GCC project
6361 web site or contact the [8]GCC development mailing list.
6363 To obtain GCC please use [9]our mirror sites or [10]our SVN server.
6366 For questions related to the use of GCC, please consult these web
6367 pages and the [11]GCC manuals. If that fails, the
6368 [12]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6369 web pages and the development of GCC are welcome on our developer
6370 list at [13]gcc@gcc.gnu.org. All of [14]our lists have public
6373 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and
6374 distribution of this entire article is permitted in any medium,
6375 provided this notice is preserved.
6377 These pages are [16]maintained by the GCC team. Last modified
6382 1. http://www.gnu.org/
6383 2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
6384 3. http://gcc.gnu.org/gcc-4.1/changes.html
6385 4. http://gcc.gnu.org/gcc-4.1/changes.html
6386 5. http://gcc.gnu.org/gcc-4.1/buildstat.html
6387 6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
6388 7. http://gcc.gnu.org/index.html
6389 8. mailto:gcc@gcc.gnu.org
6390 9. http://gcc.gnu.org/mirrors.html
6391 10. http://gcc.gnu.org/svn.html
6392 11. https://gcc.gnu.org/onlinedocs/
6393 12. mailto:gcc-help@gcc.gnu.org
6394 13. mailto:gcc@gcc.gnu.org
6395 14. https://gcc.gnu.org/lists.html
6396 15. http://www.fsf.org/
6397 16. https://gcc.gnu.org/about.html
6398 17. http://validator.w3.org/check/referer
6399 ======================================================================
6400 http://gcc.gnu.org/gcc-4.1/changes.html
6401 GCC 4.1 Release Series
6402 Changes, New Features, and Fixes
6404 The latest release in the 4.1 release series is [1]GCC 4.1.2.
6408 General Optimizer Improvements
6410 * GCC now has infrastructure for inter-procedural optimizations and
6411 the following inter-procedural optimizations are implemented:
6412 + Profile guided inlining. When doing profile feedback guided
6413 optimization, GCC can now use the profile to make better
6414 informed decisions on whether inlining of a function is
6415 profitable or not. This means that GCC will no longer inline
6416 functions at call sites that are not executed very often, and
6417 that functions at hot call sites are more likely to be
6419 A new parameter min-inline-recursive-probability is also now
6420 available to throttle recursive inlining of functions with
6421 small average recursive depths.
6422 + Discovery of pure and const functions, a form of side-effects
6423 analysis. While older GCC releases could also discover such
6424 special functions, the new IPA-based pass runs earlier so that
6425 the results are available to more optimizers. The pass is also
6426 simply more powerful than the old one.
6427 + Analysis of references to static variables and type escape
6428 analysis, also forms of side-effects analysis. The results of
6429 these passes allow the compiler to be less conservative about
6430 call-clobbered variables and references. This results in more
6431 redundant loads being eliminated and in making static
6432 variables candidates for register promotion.
6433 + Improvement of RTL-based alias analysis. The results of type
6434 escape analysis are fed to the RTL type-based alias analyzer,
6435 allowing it to disambiguate more memory references.
6436 + Interprocedural constant propagation and function versioning.
6437 This pass looks for functions that are always called with the
6438 same constant value for one or more of the function arguments,
6439 and propagates those constants into those functions.
6440 + GCC will now eliminate static variables whose usage was
6442 + -fwhole-program --combine can now be used to make all
6443 functions in program static allowing whole program
6444 optimization. As an exception, the main function and all
6445 functions marked with the new externally_visible attribute are
6446 kept global so that programs can link with runtime libraries.
6447 * GCC can now do a form of partial dead code elimination (PDCE) that
6448 allows code motion of expressions to the paths where the result of
6449 the expression is actually needed. This is not always a win, so the
6450 pass has been limited to only consider profitable cases. Here is an
6452 int foo (int *, int *);
6468 The a = b + c can be sunk to right before the printf. Normal code
6469 sinking will not do this, it will sink the first one above into the
6470 else-branch of the conditional jump, which still gives you two
6472 * GCC now has a value range propagation pass. This allows the
6473 compiler to eliminate bounds checks and branches. The results of
6474 the pass can also be used to accurately compute branch
6476 * The pass to convert PHI nodes to straight-line code (a form of
6477 if-conversion for GIMPLE) has been improved significantly. The two
6478 most significant improvements are an improved algorithm to
6479 determine the order in which the PHI nodes are considered, and an
6480 improvement that allow the pass to consider if-conversions of basic
6481 blocks with more than two predecessors.
6482 * Alias analysis improvements. GCC can now differentiate between
6483 different fields of structures in Tree-SSA's virtual operands form.
6484 This lets stores/loads from non-overlapping structure fields not
6485 conflict. A new algorithm to compute points-to sets was contributed
6486 that can allows GCC to see now that p->a and p->b, where p is a
6487 pointer to a structure, can never point to the same field.
6488 * Various enhancements to auto-vectorization:
6489 + Incrementally preserve SSA form when vectorizing.
6490 + Incrementally preserve loop-closed form when vectorizing.
6491 + Improvements to peeling for alignment: generate better code
6492 when the misalignment of an access is known at compile time,
6493 or when different accesses are known to have the same
6494 misalignment, even if the misalignment amount itself is
6496 + Consider dependence distance in the vectorizer.
6497 + Externalize generic parts of data reference analysis to make
6498 this analysis available to other passes.
6499 + Vectorization of conditional code.
6500 + Reduction support.
6501 * GCC can now partition functions in sections of hot and cold code.
6502 This can significantly improve performance due to better
6503 instruction cache locality. This feature works best together with
6504 profile feedback driven optimization.
6505 * A new pass to avoid saving of unneeded arguments to the stack in
6506 vararg functions if the compiler can prove that they will not be
6508 * Transition of basic block profiling to tree level implementation
6509 has been completed. The new implementation should be considerably
6510 more reliable (hopefully avoiding profile mismatch errors when
6511 using -fprofile-use or -fbranch-probabilities) and can be used to
6512 drive higher level optimizations, such as inlining.
6513 The -ftree-based-profiling command-line option was removed and
6514 -fprofile-use now implies disabling old RTL level loop optimizer
6515 (-fno-loop-optimize). Speculative prefetching optimization
6516 (originally enabled by -fspeculative-prefetching) was removed.
6518 New Languages and Language specific improvements
6522 * The old Bison-based C and Objective-C parser has been replaced by a
6523 new, faster hand-written recursive-descent parser.
6527 * The build infrastructure for the Ada runtime library and tools has
6528 been changed to be better integrated with the rest of the build
6529 infrastructure of GCC. This should make doing cross builds of Ada a
6534 * ARM-style name-injection of friend declarations is no longer the
6535 default. For example:
6541 will not be accepted; instead a declaration of f will need to be
6542 present outside of the scope of S. The new -ffriend-injection
6543 option will enable the old behavior.
6544 * The (undocumented) extension which permitted templates with default
6545 arguments to be bound to template template parameters with fewer
6546 parameters has been deprecated, and will be removed in the next
6547 major release of G++. For example:
6548 template <template <typename> class C>
6549 void f(C<double>) {}
6551 template <typename T, typename U = int>
6554 template void f(S<double>);
6556 makes use of the deprecated extension. The reason this code is not
6557 valid ISO C++ is that S is a template with two parameters;
6558 therefore, it cannot be bound to C which has only one parameter.
6560 Runtime Library (libstdc++)
6562 * Optimization work:
6563 + A new implementation of std::search_n is provided, better
6564 performing in case of random access iterators.
6565 + Added further efficient specializations of istream functions,
6566 i.e., character array and string extractors.
6567 + Other smaller improvements throughout.
6568 * Policy-based associative containers, designed for high-performance,
6569 flexibility and semantic safety are delivered in ext/pb_assoc.
6570 * A versatile string class, __gnu_cxx::__versa_string, providing
6571 facilities conforming to the standard requirements for
6572 basic_string, is delivered in <ext/vstring.h>. In particular:
6573 + Two base classes are provided: the default one avoids
6574 reference counting and is optimized for short strings; the
6575 alternate one, still uses it while improving in a few low
6576 level areas (e.g., alignment). See vstring_fwd.h for some
6578 + Various algorithms have been rewritten (e.g., replace), the
6579 code streamlined and simple optimizations added.
6580 + Option 3 of DR 431 is implemented for both available bases,
6581 thus improving the support for stateful allocators.
6582 * As usual, many bugs have been fixed (e.g., libstdc++/13583,
6583 libstdc++/23953) and LWG resolutions put into effect for the first
6584 time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1
6585 Issue 6.19). The implementation status of TR1 is now tracked in the
6590 * A new language front end for Objective-C++ has been added. This
6591 language allows users to mix the object oriented features of
6592 Objective-C with those of C++.
6596 * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19
6597 features (plus some 0.20 bug-fixes)
6599 o The java.net.HttpURLConnection implementation no longer
6600 buffers the entire response body in memory. This means
6601 that response bodies larger than available memory can now
6604 o NIO FileChannel.map implementation, fast bulk put
6605 implementation for DirectByteBuffer (speeds up this
6607 o FileChannel.lock() and FileChannel.force() implemented.
6609 o gnu.xml fix for nodes created outside a namespace
6611 o Add support for output indenting and
6612 cdata-section-elements output instruction in
6614 o xml.xpath corrections for cases where elements/attributes
6615 might have been created in non-namespace-aware mode.
6616 Corrections to handling of XSL variables and minor
6617 conformance updates.
6619 o GNU JAWT implementation, the AWT Native Interface, which
6620 allows direct access to native screen resources from
6621 within a Canvas's paint method. GNU Classpath Examples
6622 comes with a Demo, see libjava/classpath/examples/README.
6623 o awt.datatransfer updated to 1.5 with support for
6624 FlavorEvents. The gtk+ awt peers now allow copy/paste of
6625 text, images, URIs/files and serialized objects with
6626 other applications and tracking clipboard change events
6627 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized
6628 objects are supported). A GNU Classpath Examples
6629 datatransfer Demo was added to show the new
6631 o Split gtk+ awt peers event handling in two threads and
6632 improve gdk lock handling (solves several awt lock ups).
6633 o Speed up awt Image loading.
6634 o Better gtk+ scrollbar peer implementation when using gtk+
6636 o Handle image loading errors correctly for gdkpixbuf and
6638 o Better handle GDK lock. Properly prefix gtkpeer native
6640 o GdkGraphics2D has been updated to use Cairo 0.5.x or
6642 o BufferedImage and GtkImage rewrites. All image drawing
6643 operations should now work correctly (flipping requires
6645 o Future Graphics2D, image and text work is documented at:
6646 [2]http://developer.classpath.org/mediation/ClasspathGrap
6648 o When gtk+ 2.6 or higher is installed the default log
6649 handler will produce stack traces whenever a WARNING,
6650 CRITICAL or ERROR message is produced.
6652 o The RepaintManager has been reworked for more efficient
6653 painting, especially for large GUIs.
6654 o The layout manager OverlayLayout has been implemented,
6655 the BoxLayout has been rewritten to make use of the
6656 SizeRequirements utility class and caching for more
6658 o Improved accessibility support.
6659 o Significant progress has been made in the implementation
6660 of the javax.swing.plaf.metal package, with most UI
6661 delegates in a working state now. Please test this with
6662 your own applications and provide feedback that will help
6663 us to improve this package.
6664 o The GUI demo (gnu.classpath.examples.swing.Demo) has been
6665 extended to highlight various features in our Free Swing
6666 implementation. And it includes a look and feel switcher
6667 for Metal (default), Ocean and GNU themes.
6668 o The javax.swing.plaf.multi package is now implemented.
6669 o Editing and several key actions for JTree and JTable were
6671 o Lots of icons and look and feel improvements for Free
6672 Swing basic and metal themes were added. Try running the
6673 GNU Classpath Swing Demo in examples
6674 (gnu.classpath.examples.swing.Demo) with:
6675 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee
6677 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee
6679 o Start of styled text capabilites for java.swing.text.
6680 o DefaultMutableTreeNode pre-order, post-order, depth-first
6681 and breadth-first traversal enumerations implemented.
6682 o JInternalFrame colors and titlebar draw properly.
6683 o JTree is working up to par (icons, selection and keyboard
6685 o JMenus were made more compatible in visual and
6686 programmatic behavior.
6687 o JTable changeSelection and multiple selections
6689 o JButton and JToggleButton change states work properly
6691 o JFileChooser fixes.
6692 o revalidate() and repaint() fixes which make Free Swing
6693 much more responsive.
6694 o MetalIconFactory implemented.
6695 o Free Swing Top-Level Compatibility. JFrame, JDialog,
6696 JApplet, JInternalFrame, and JWindow are now 1.5
6697 compatible in the sense that you can call add() and
6698 setLayout() directly on them, which will have the same
6699 effect as calling getContentPane().add() and
6700 getContentPane().setLayout().
6701 o The JTree interface has been completed. JTrees now
6702 recognizes mouse clicks and selections work.
6703 o BoxLayout works properly now.
6704 o Fixed GrayFilter to actually work.
6705 o Metal SplitPane implemented.
6706 o Lots of Free Swing text and editor stuff work now.
6707 + Free RMI and Corba
6708 o Andrew Watson, Vice President and Technical Director of
6709 the Object Management Group, has officially assigned us
6710 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will
6711 mark remote classpath-specific system exceptions.
6712 Obtaining the VMCID means that GNU Classpath now is a
6713 recogniseable type of node in a highly interoperable
6715 o GNU Classpath now includes the first working draft to
6716 support the RMI over IIOP protocol. The current
6717 implementation is capable of remote invocations,
6718 transferring various Serializables and Externalizables
6719 via RMI-IIOP protocol. It can flatten graphs and, at
6720 least for the simple cases, is interoperable with 1.5
6722 o org.omg.PortableInterceptor and related functionality in
6723 other packages is now implemented:
6724 # The sever and client interceptors work as required
6726 # The IOR interceptor works as needed for 1.5.
6727 o The org.omg.DynamicAny package is completed and passes
6729 o The Portable Object Adapter should now support the output
6730 of the recent IDL to java compilers. These compilers now
6731 generate servants and not CORBA objects as before, making
6732 the output depend on the existing POA implementation.
6733 Completing POA means that such code can already be tried
6734 to run on Classpath. Our POA is tested for the following
6736 # POA converts servant to the CORBA object.
6737 # Servant provides to the CORBA object.
6738 # POA activates new CORBA object with the given Object
6739 Id (byte array) that is later accessible for the
6741 # During the first call, the ServantActivator provides
6742 servant for this and all subsequent calls on the
6744 # During each call, the ServantLocator provides
6745 servant for this call only.
6746 # ServantLocator or ServantActivator forwards call to
6748 # POA has a single servant, responsible for all
6750 # POA has a default servant, but some objects are
6751 explicitly connected to they specific servants.
6752 The POA is verified using tests from the former
6754 o The CORBA implementation is now a working prototype that
6755 should support features up to 1.3 inclusive. We invite
6756 groups writing CORBA dependent applications to try
6757 Classpath implementation, reporting any possible bugs.
6758 The CORBA prototype is interoperable with Sun's
6759 implementation v 1.4, transferring object references,
6760 primitive types, narrow and wide strings, arrays,
6761 structures, trees, abstract interfaces and value types
6762 (feature of CORBA 2.3) between these two platforms.
6763 Remote exceptions are transferred and handled correctly.
6764 The stringified object references (IORs) from various
6765 sources are parsed as required. The transient (for
6766 current session) and permanent (till jre restart)
6767 redirections work. Both Little and Big Endian encoded
6768 messages are accepted. The implementation is verified
6769 using tests from the former cost.omg.org. The current
6770 release includes working examples (see the examples
6771 directory), demonstrating the client-server
6772 communication, using either CORBA Request or IDL-based
6773 stub (usually generated by a IDL to java compiler). These
6774 examples also show how to use the Classpath CORBA naming
6775 service. The IDL to java compiler is not yet written, but
6776 as our library must be compatible, it naturally accepts
6777 the output of other idlj implementations.
6779 o Updated TimeZone data against Olson tzdata2005l.
6780 o Make zip and jar packages UTF-8 clean.
6781 o "native" code builds and compiles (warning free) on
6783 o java.util.logging.FileHandler now rotates files.
6784 o Start of a generic JDWP framework in gnu/classpath/jdwp.
6785 This is unfinished, but feedback (at classpath@gnu.org)
6786 from runtime hackers is greatly appreciated. Although
6787 most of the work is currently being done around gcj/gij
6788 we want this framework to be as VM neutral as possible.
6789 Early design is described in:
6790 [3]https://gcc.gnu.org/ml/java/2005-05/msg00260.html
6791 o QT4 AWT peers, enable by giving configure
6792 --enable-qt-peer. Included, but not ready for production
6793 yet. They are explicitly disabled and not supported. But
6794 if you want to help with the development of these new
6795 features we are interested in feedback. You will have to
6796 explicitly enable them to try them out (and they will
6797 most likely contain bugs).
6798 o Documentation fixes all over the place. See
6799 [4]http://developer.classpath.org/doc/
6801 New Targets and Target Specific Improvements
6805 * The x86-64 medium model (that allows building applications whose
6806 data segment exceeds 4GB) was redesigned to match latest ABI draft.
6807 New implementation split large datastructures into separate segment
6808 improving performance of accesses to small datastructures and also
6809 allows linking of small model libraries into medium model programs
6810 as long as the libraries are not accessing the large datastructures
6811 directly. Medium model is also supported in position independent
6813 The ABI change results in partial incompatibility among medium
6814 model objects. Linking medium model libraries (or objects) compiled
6815 with new compiler into medium model program compiled with older
6816 will likely result in exceeding ranges of relocations.
6817 Binutils 2.16.91 or newer are required for compiling medium model
6820 RS6000 (POWER/PowerPC)
6822 * The AltiVec vector primitives in <altivec.h> are now implemented in
6823 a way that puts a smaller burden on the preprocessor, instead
6824 processing the "overloading" in the front ends. This should benefit
6825 compilation speed on AltiVec vector code.
6826 * AltiVec initializers now are generated more efficiently.
6827 * The popcountb instruction available on POWER5 now is generated.
6828 * The floating point round to integer instructions available on
6829 POWER5+ now is generated.
6830 * Floating point divides can be synthesized using the floating point
6831 reciprocal estimate instructions.
6832 * Double precision floating point constants are initialized as single
6833 precision values if they can be represented exactly.
6835 S/390, zSeries and System z9
6837 * Support for the IBM System z9 109 processor has been added. When
6838 using the -march=z9-109 option, the compiler will generate code
6839 making use of instructions provided by the extended immediate
6841 * Support for 128-bit IEEE floating point has been added. When using
6842 the -mlong-double-128 option, the compiler will map the long double
6843 data type to 128-bit IEEE floating point. Using this option
6844 constitutes an ABI change, and requires glibc support.
6845 * Various changes to improve performance of generated code have been
6846 implemented, including:
6847 + In functions that do not require a literal pool, register %r13
6848 (which is traditionally reserved as literal pool pointer), can
6849 now be freely used for other purposes by the compiler.
6850 + More precise tracking of register use allows the compiler to
6851 generate more efficient function prolog and epilog code in
6853 + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING
6854 instructions are now used to implement C string functions.
6855 + The MOVE CHARACTER instruction with single byte overlap is now
6856 used to implement the memset function with non-zero fill byte.
6857 + The LOAD ZERO instructions are now used where appropriate.
6858 + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK,
6859 and INSERT IMMEDIATE instructions are now used more frequently
6860 to optimize bitfield operations.
6861 + The BRANCH ON COUNT instruction is now used more frequently.
6862 In particular, the fact that a loop contains a subroutine call
6863 no longer prevents the compiler from using this instruction.
6864 + The compiler is now aware that all shift and rotate
6865 instructions implicitly truncate the shift count to six bits.
6866 * Back-end support for the following generic features has been
6868 + The full set of [5]built-in functions for atomic memory
6870 + The -fstack-protector feature.
6871 + The optimization pass avoiding unnecessary stores of incoming
6872 argument registers in functions with variable argument list.
6876 * The default code model in 64-bit mode has been changed from
6877 Medium/Anywhere to Medium/Middle on Solaris.
6878 * TLS support is disabled by default on Solaris prior to release 10.
6879 It can be enabled on TLS-capable Solaris 9 versions (4/04 release
6880 and later) by specifying --enable-tls at configure time.
6884 * Support has been added for this new architecture.
6888 Documentation improvements
6890 Other significant improvements
6892 * GCC can now emit code for protecting applications from
6893 stack-smashing attacks. The protection is realized by buffer
6894 overflow detection and reordering of stack variables to avoid
6896 * Some built-in functions have been fortified to protect them against
6897 various buffer overflow (and format string) vulnerabilities.
6898 Compared to the mudflap bounds checking feature, the safe builtins
6899 have far smaller overhead. This means that programs built using
6900 safe builtins should not experience any measurable slowdown.
6904 This is the [6]list of problem reports (PRs) from GCC's bug tracking
6905 system that are known to be fixed in the 4.1.2 release. This list might
6906 not be complete (that is, it is possible that some PRs that have been
6907 fixed are not listed here).
6909 When generating code for a shared library, GCC now recognizes that
6910 global functions may be replaced when the program runs. Therefore, it
6911 is now more conservative in deducing information from the bodies of
6912 functions. For example, in this example:
6917 cout << "Exception";
6921 G++ would previously have optimized away the catch clause, since it
6922 would have concluded that f cannot throw exceptions. Because users may
6923 replace f with another function in the main body of the program, this
6924 optimization is unsafe, and is no longer performed. If you wish G++ to
6925 continue to optimize as before, you must add a throw() clause to the
6926 declaration of f to make clear that it does not throw exceptions.
6929 For questions related to the use of GCC, please consult these web
6930 pages and the [7]GCC manuals. If that fails, the
6931 [8]gcc-help@gcc.gnu.org mailing list might help. Comments on these
6932 web pages and the development of GCC are welcome on our developer
6933 list at [9]gcc@gcc.gnu.org. All of [10]our lists have public
6936 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and
6937 distribution of this entire article is permitted in any medium,
6938 provided this notice is preserved.
6940 These pages are [12]maintained by the GCC team. Last modified
6945 1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2
6946 2. http://developer.classpath.org/mediation/ClasspathGraphicsImagesText
6947 3. https://gcc.gnu.org/ml/java/2005-05/msg00260.html
6948 4. http://developer.classpath.org/doc/
6949 5. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html
6950 6. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2
6951 7. https://gcc.gnu.org/onlinedocs/
6952 8. mailto:gcc-help@gcc.gnu.org
6953 9. mailto:gcc@gcc.gnu.org
6954 10. https://gcc.gnu.org/lists.html
6955 11. http://www.fsf.org/
6956 12. https://gcc.gnu.org/about.html
6957 13. http://validator.w3.org/check/referer
6958 ======================================================================
6959 http://gcc.gnu.org/gcc-4.0/index.html
6960 GCC 4.0 Release Series
6964 The [1]GNU project and the GCC developers are pleased to announce the
6965 release of GCC 4.0.4.
6967 This release is a bug-fix release, containing fixes for regressions in
6968 GCC 4.0.3 relative to previous releases of GCC.
6973 January 31, 2007 ([2]changes)
6976 March 10, 2006 ([3]changes)
6979 September 28, 2005 ([4]changes)
6982 July 7, 2005 ([5]changes)
6985 April 20, 2005 ([6]changes)
6987 References and Acknowledgements
6989 GCC used to stand for the GNU C Compiler, but since the compiler
6990 supports several other languages aside from C, it now stands for the
6991 GNU Compiler Collection.
6993 A list of [7]successful builds is updated as new information becomes
6996 The GCC developers would like to thank the numerous people that have
6997 contributed new features, improvements, bug fixes, and other changes as
6998 well as test results to GCC. This [8]amazing group of volunteers is
6999 what makes GCC successful.
7001 For additional information about GCC please refer to the [9]GCC project
7002 web site or contact the [10]GCC development mailing list.
7004 To obtain GCC please use [11]our mirror sites, or [12]our SVN server.
7007 For questions related to the use of GCC, please consult these web
7008 pages and the [13]GCC manuals. If that fails, the
7009 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
7010 web pages and the development of GCC are welcome on our developer
7011 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
7014 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
7015 distribution of this entire article is permitted in any medium,
7016 provided this notice is preserved.
7018 These pages are [18]maintained by the GCC team. Last modified
7023 1. http://www.gnu.org/
7024 2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
7025 3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3
7026 4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2
7027 5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1
7028 6. http://gcc.gnu.org/gcc-4.0/changes.html
7029 7. http://gcc.gnu.org/gcc-4.0/buildstat.html
7030 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
7031 9. http://gcc.gnu.org/index.html
7032 10. mailto:gcc@gcc.gnu.org
7033 11. http://gcc.gnu.org/mirrors.html
7034 12. http://gcc.gnu.org/svn.html
7035 13. https://gcc.gnu.org/onlinedocs/
7036 14. mailto:gcc-help@gcc.gnu.org
7037 15. mailto:gcc@gcc.gnu.org
7038 16. https://gcc.gnu.org/lists.html
7039 17. http://www.fsf.org/
7040 18. https://gcc.gnu.org/about.html
7041 19. http://validator.w3.org/check/referer
7042 ======================================================================
7043 http://gcc.gnu.org/gcc-4.0/changes.html
7044 GCC 4.0 Release Series
7045 Changes, New Features, and Fixes
7047 The latest release in the 4.0 release series is [1]GCC 4.0.4.
7051 * GCC now generates location lists by default when compiling with
7052 debug info and optimization.
7053 + GDB 6.0 and older crashes when it sees location lists. GDB 6.1
7054 or later is needed to debug binaries containing location
7056 + When you are trying to view a value of a variable in a part of
7057 a function where it has no location (for example when the
7058 variable is no longer used and thus its location was used for
7059 something else) GDB will say that it is not available.
7060 You can disable generating location lists by -fno-var-tracking.
7061 * GCC no longer accepts the -fwritable-strings option. Use named
7062 character arrays when you need a writable string.
7063 * The options -freduce-all-givs and -fmove-all-movables have been
7064 discontinued. They were used to circumvent a shortcoming in the
7065 heuristics of the old loop optimization code with respect to common
7066 Fortran constructs. The new (tree) loop optimizer works differently
7067 and doesn't need those work-arounds.
7068 * The graph-coloring register allocator, formerly enabled by the
7069 option -fnew-ra, has been discontinued.
7070 * -I- has been deprecated. -iquote is meant to replace the need for
7072 * The MIPS -membedded-pic and -mrnames options have been removed.
7073 * All MIPS targets now require the GNU assembler. In particular, IRIX
7074 configurations can no longer use the MIPSpro assemblers, although
7075 they do still support the MIPSpro linkers.
7076 * The SPARC option -mflat has been removed.
7077 * English-language diagnostic messages will now use Unicode quotation
7078 marks in UTF-8 locales. (Non-English messages already used the
7079 quotes appropriate for the language in previous releases.) If your
7080 terminal does not support UTF-8 but you are using a UTF-8 locale
7081 (such locales are the default on many GNU/Linux systems) then you
7082 should set LC_CTYPE=C in the environment to disable that locale.
7083 Programs that parse diagnostics and expect plain ASCII
7084 English-language messages should set LC_ALL=C. See [2]Markus Kuhn's
7085 explanation of Unicode quotation marks for more information.
7086 * The specs file is no longer installed on most platforms. Most users
7087 will be totally unaffected. However, if you are accustomed to
7088 editing the specs file yourself, you will now have to use the
7089 -dumpspecs option to generate the specs file, and then edit the
7092 General Optimizer Improvements
7094 * The [3]tree ssa branch has been merged. This merge has brought in a
7095 completely new optimization framework based on a higher level
7096 intermediate representation than the existing RTL representation.
7097 Numerous new code transformations based on the new framework are
7098 available in GCC 4.0, including:
7099 + Scalar replacement of aggregates
7100 + Constant propagation
7101 + Value range propagation
7102 + Partial redundancy elimination
7103 + Load and store motion
7104 + Strength reduction
7105 + Dead store elimination
7106 + Dead and unreachable code elimination
7107 + [4]Autovectorization
7109 + Tail recursion by accumulation
7110 Many of these passes outperform their counterparts from previous
7112 * [5]Swing Modulo Scheduling (SMS). An RTL level instruction
7113 scheduling optimization intended for loops that perform heavy
7116 New Languages and Language specific improvements
7120 * The sentinel attribute has been added to GCC. This function
7121 attribute allows GCC to warn when variadic functions such as execl
7122 are not NULL terminated. See the GCC manual for a complete
7123 description of its behavior.
7124 * Given __attribute__((alias("target"))) it is now an error if target
7125 is not a symbol, defined in the same translation unit. This also
7126 applies to aliases created by #pragma weak alias=target. This is
7127 because it's meaningless to define an alias to an undefined symbol.
7128 On Solaris, the native assembler would have caught this error, but
7133 * The -Wstrict-aliasing=2 option has been added. This warning catches
7134 all unsafe cases, but it may also give a warning for some cases
7136 * The cast-as-lvalue, conditional-expression-as-lvalue and
7137 compound-expression-as-lvalue extensions, which were deprecated in
7138 3.3.4 and 3.4, have been removed.
7139 * The -fwritable-strings option, which was deprecated in 3.4, has
7141 * #pragma pack() semantics have been brought closer to those used by
7142 other compilers. This also applies to C++.
7143 * Taking the address of a variable with register storage is invalid
7144 in C. GCC now issues an error instead of a warning.
7145 * Arrays of incomplete element type are invalid in C. GCC now issues
7146 an error for such arrays. Declarations such as extern struct s x[];
7147 (where struct s has not been defined) can be moved after the
7148 definition of struct s. Function parameters declared as arrays of
7149 incomplete type can instead be declared as pointers.
7153 * When compiling without optimizations (-O0), the C++ frontend is
7154 much faster than in any previous versions of GCC. Independent
7155 testers have measured speed-ups up to 25% in real-world production
7156 code, compared to the 3.4 family (which was already the fastest
7157 version to date). Upgrading from older versions might show even
7158 bigger improvements.
7159 * ELF visibility attributes can now be applied to a class type, so
7160 that it affects every member function of a class at once, without
7161 having to specify each individually:
7162 class __attribute__ ((visibility("hidden"))) Foo
7167 The syntax is deliberately similar to the __declspec() system used
7168 by Microsoft Windows based compilers, allowing cross-platform
7169 projects to easily reuse their existing macro system for denoting
7170 exports and imports. By explicitly marking internal classes never
7171 used outside a binary as hidden, one can completely avoid PLT
7172 indirection overheads during their usage by the compiler. You can
7173 find out more about the advantages of this at
7174 [6]http://www.akkadia.org/drepper/dsohowto.pdf
7175 * The -fvisibility-inlines-hidden option has been added which marks
7176 all inlineable functions as having hidden ELF visibility, thus
7177 removing their symbol and typeinfo from the exported symbol table
7178 of the output ELF binary. Using this option can reduce the exported
7179 symbol count of template-heavy code by up to 40% with no code
7180 change at all, thus notably improving link and load times for the
7181 binary as well as a reduction in size of up to 10%. Also, check the
7182 new [7]-fvisibility option.
7183 * The compiler now uses the library interface specified by the [8]C++
7184 ABI for thread-safe initialization of function-scope static
7185 variables. Most users should leave this alone, but embedded
7186 programmers may want to disable this by specifying
7187 -fno-threadsafe-statics for a small savings in code size.
7188 * Taking the address of an explicit register variable is no longer
7189 supported. Note that C++ allows taking the address of variables
7190 with register storage so this will continue to compile with a
7191 warning. For example, assuming that r0 is a machine register:
7192 register int foo asm ("r0");
7194 &foo; // error, no longer accepted
7195 &bar; // OK, with a warning
7196 * G++ has an undocumented extension to virtual function covariancy
7197 rules that allowed the overrider to return a type that was
7198 implicitly convertable to the overridden function's return type.
7199 For instance a function returning void * could be overridden by a
7200 function returning T *. This is now deprecated and will be removed
7201 in a future release.
7202 * The G++ minimum and maximum operators (<? and >?) and their
7203 compound forms (<?=) and >?=) have been deprecated and will be
7204 removed in a future version. Code using these operators should be
7205 modified to use std::min and std::max instead.
7206 * Declaration of nested classes of class templates as friends are
7208 template <typename T> struct A {
7212 template <typename T> friend class A<T>::B;
7214 This complements the feature member functions of class templates as
7215 friends introduced in GCC 3.4.0.
7216 * When declaring a friend class using an unqualified name, classes
7217 outside the innermost non-class scope are not searched:
7221 friend class A; // Refer to N::A which has not been declared yet
7222 // because name outside namespace N are not searched
7223 friend class ::A; // Refer to ::A
7226 Hiding the friend name until declaration is still not implemented.
7227 * Friends of classes defined outside their namespace are correctly
7233 friend class B; // Refer to N::B in GCC 4.0.0
7234 // but ::B in earlier versions of GCC
7237 Runtime Library (libstdc++)
7239 * Optimization work:
7240 + Added efficient specializations of istream functions for char
7242 + Further performance tuning of strings, in particular wrt
7243 single-char append and getline.
7244 + iter_swap - and therefore most of the mutating algorithms -
7245 now makes an unqualified call to swap when the value_type of
7246 the two iterators is the same.
7247 * A large subset of the features in Technical Report 1 (TR1 for
7248 short) is experimentally delivered (i.e., no guarantees about the
7249 implementation are provided. In particular it is not promised that
7250 the library will remain link-compatible when code using TR1 is
7252 + General utilities such as reference_wrapper and shared_ptr.
7253 + Function objects, i.e., result_of, mem_fn, bind, function.
7254 + Support for metaprogramming.
7255 + New containers such as tuple, array, unordered_set,
7256 unordered_map, unordered_multiset, unordered_multimap.
7257 * As usual, many bugs have been fixed and LWG resolutions implemented
7258 for the first time (e.g., DR 409).
7262 * In order to prevent naming conflicts with other implementations of
7263 these tools, some GCJ binaries have been renamed:
7264 + rmic is now grmic,
7265 + rmiregistry is now grmiregistry, and
7266 + jar is now fastjar.
7267 In particular, these names were problematic for the jpackage.org
7268 packaging conventions which install symlinks in /usr/bin that point
7269 to the preferred versions of these tools.
7270 * The -findirect-dispatch argument to the compiler now works and
7271 generates code following a new "binary compatibility" ABI. Code
7272 compiled this way follows the binary compatibility rules of the
7273 Java Language Specification.
7274 * libgcj now has support for using GCJ as a JIT, using the
7275 gnu.gcj.jit family of system properties.
7276 * libgcj can now find a shared library corresponding to the bytecode
7277 representation of a class. See the documentation for the new
7278 gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system
7280 * There have been many improvements to the class library. Here are
7282 + Much more of AWT and Swing exist.
7283 + Many new packages and classes were added, including
7284 java.util.regex, java.net.URI, javax.crypto,
7285 javax.crypto.interfaces, javax.crypto.spec, javax.net,
7286 javax.net.ssl, javax.security.auth,
7287 javax.security.auth.callback, javax.security.auth.login,
7288 javax.security.auth.x500, javax.security.sasl, org.ietf.jgss,
7289 javax.imageio, javax.imageio.event, javax.imageio.spi,
7290 javax.print, javax.print.attribute,
7291 javax.print.attribute.standard, javax.print.event, and
7293 + Updated SAX and DOM, and imported GNU JAXP
7297 * A new [9]Fortran front end has replaced the aging GNU Fortran 77
7298 front end. The new front end supports Fortran 90 and Fortran 95. It
7299 may not yet be as stable as the old Fortran front end.
7303 * Ada (with tasking and Zero Cost Exceptions) is now available on
7304 many more targets, including but not limited to: alpha-linux,
7305 hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux,
7306 s390x-linux, sparc-linux.
7307 * Some of the new Ada 2005 features are now implemented like
7308 Wide_Wide_Character and Ada.Containers.
7309 * Many bugs have been fixed, tools and documentation improved.
7310 * To compile Ada from the sources, install an older working Ada
7311 compiler and then use --enable-languages=ada at configuration time,
7312 since the Ada frontend is not currently activated by default. See
7313 the [10]Installing GCC for details.
7315 New Targets and Target Specific Improvements
7319 * The frame layout has changed. In the new layout, the prologue of a
7320 function first saves registers and then allocate space for locals,
7321 resulting in an 1% improvement on code size.
7323 IA-32/x86-64 (AMD64)
7325 * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10,
7326 log1p, log2, logb and tan mathematical builtins (and their float
7327 and long double variants) are now implemented as inline x87
7328 intrinsics when using -ffast-math.
7329 * The ceil, floor, nearbyint, rint and trunc mathematical builtins
7330 (and their float and long double variants) are now implemented as
7331 inline x87 intrinsics when using -ffast-math.
7332 * The x87's fsincos instruction is now used automatically with
7333 -ffast-math when calculating both the sin and cos of the same
7335 * Instruction selection for multiplication and division by constants
7340 * Floating point division, integer division and sqrt are now inlined,
7341 resulting in significant performance improvements on some codes.
7345 * Division by zero checks now use conditional traps if the target
7346 processor supports them. This decreases code size by one word per
7347 division operation. The old behavior (branch and break) can be
7348 obtained either at configure time by passing --with-divide=breaks
7349 to configure or at runtime by passing -mdivide-breaks to GCC.
7350 * Support for MIPS64 paired-single instructions has been added. It is
7351 enabled by -mpaired-single and can be accessed using both the
7352 target-independent vector extensions and new MIPS-specific built-in
7354 * Support for the MIPS-3D ASE has been added. It is enabled by
7355 -mips3d and provides new MIPS-3D-specific built-in functions.
7356 * The -mexplicit-relocs option now supports static n64 code (as is
7357 used, for example, in 64-bit linux kernels). -mexplicit-relocs
7358 should now be feature-complete and is enabled by default when GCC
7359 is configured to use a compatible assembler.
7360 * Support for the NEC VR4130 series has been added. This support
7361 includes the use of VR-specific instructions and a new VR4130
7362 scheduler. Full VR4130 support can be selected with -march=vr4130
7363 while code for any ISA can be tuned for the VR4130 using
7364 -mtune=vr4130. There is also a new -mvr4130-align option that
7365 produces better schedules at the cost of increased code size.
7366 * Support for the Broadcom SB-1 has been extended. There is now an
7367 SB-1 scheduler as well as support for the SB-1-specific
7368 paired-single instructions. Full SB-1 support can be selected with
7369 -march=sb1 while code for any ISA can be optimized for the SB-1
7371 * The compiler can now work around errata in R4000, R4400, VR4120 and
7372 VR4130 processors. These workarounds are enabled by -mfix-r4000,
7373 -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120
7374 and VR4130 workarounds need binutils 2.16 or above.
7375 * IRIX shared libraries are now installed into the standard library
7376 directories: o32 libraries go into lib/, n32 libraries go into
7377 lib32/ and n64 libraries go into lib64/.
7378 * The compiler supports a new -msym32 option. It can be used to
7379 optimize n64 code in which all symbols are known to have 32-bit
7384 * New command-line options help to generate code intended to run in
7385 an environment where stack space is restricted, e.g. Linux kernel
7387 + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time
7388 warnings for single functions that require large or dynamic
7390 + -mstack-size and -mstack-guard generate code that checks for
7391 stack overflow at run time.
7392 + -mpacked-stack generates code that reduces the stack frame
7393 size of many functions by reusing unneeded parts of the stack
7395 * The -msoft-float option now ensures that generated code never
7396 accesses floating point registers.
7397 * The s390x-ibm-tpf target now fully supports C++, including
7398 exceptions and threads.
7399 * Various changes to improve performance of the generated code have
7400 been implemented, including:
7401 + GCC now uses sibling calls where possible.
7402 + Condition code handling has been optimized, allowing GCC to
7403 omit redundant comparisons in certain cases.
7404 + The cost function guiding many optimizations has been refined
7405 to more accurately represent the z900 and z990 processors.
7406 + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW
7407 instructions are now used to avoid conditional branches in
7409 + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to
7410 optimize address arithmetic required to access large stack
7412 + GCC now makes more efficient use of memory-to-memory type
7413 instructions (MVC, CLC, ...).
7414 + More precise tracking of special register use allows better
7415 instruction scheduling, in particular of the function prologue
7416 and epilogue sequences.
7417 + The Java front end now generates inline code to implement
7418 integer division, instead of calling library routines.
7422 * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and
7423 -mf934 have been removed. They have been replaced with -mcpu=xxx.
7424 * The internal model used to estimate the relative cost of each
7425 instruction has been updated. It is expected to give better results
7426 on recent UltraSPARC processors.
7427 * Code generation for function prologues and epilogues has been
7428 improved, resulting in better scheduling and allowing multiple exit
7429 points in functions.
7430 * Support for Sun's Visual Instruction Set (VIS) has been enhanced.
7431 It is enabled by -mvis and provides new built-in functions for VIS
7432 instructions on UltraSPARC processors.
7433 * The option -mapp-regs has been turned on by default on Solaris too.
7437 * Novell NetWare (on ix86, no other hardware platform was ever really
7438 supported by this OS) has been re-enabled and the ABI supported by
7439 GCC has been brought into sync with that of MetroWerks CodeWarrior
7440 (the ABI previously supported was that of some Unix systems, which
7441 NetWare never tried to support).
7445 Support for a number of older systems has been declared obsolete in GCC
7446 4.0. Unless there is activity to revive them, the next release of GCC
7447 will have their sources permanently removed.
7449 All GCC ports for the following processor architectures have been
7453 * National Semiconductor NS32K
7454 * Texas Instruments TMS320C[34]x
7456 Also, those for some individual systems have been obsoleted:
7458 + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf,
7460 + OpenBSD 32-bit (sparc-*-openbsd*)
7462 Documentation improvements
7464 Other significant improvements
7466 * Location lists are now generated by default when compiling with
7467 debug info and optimization. Location lists provide more accurate
7468 debug info about locations of variables and they allow debugging
7469 code compiled with -fomit-frame-pointer.
7470 * The -fvisibility option has been added which allows the default ELF
7471 visibility of all symbols to be set per compilation and the new
7472 #pragma GCC visibility preprocessor command allows the setting of
7473 default ELF visibility for a region of code. Using
7474 -fvisibility=hidden especially in combination with the new
7475 -fvisibility-inlines-hidden can yield substantial improvements in
7476 output binary quality including avoiding PLT indirection overheads,
7477 reduction of the exported symbol count by up to 60% (with resultant
7478 improvements to link and load times), better scope for the
7479 optimizer to improve code and up to a 20% reduction in binary size.
7480 Using these options correctly yields a binary with a similar symbol
7481 count to a Windows DLL.
7482 Perhaps more importantly, this new feature finally allows (with
7483 careful planning) complete avoidance of symbol clashes when
7484 manually loading shared objects with RTLD_GLOBAL, thus finally
7485 solving problems many projects such as python were forced to use
7486 RTLD_LOCAL for (with its resulting issues for C++ correctness). You
7487 can find more information about using these options at
7488 [11]https://gcc.gnu.org/wiki/Visibility.
7489 __________________________________________________________________
7493 This is the [12]list of problem reports (PRs) from GCC's bug tracking
7494 system that are known to be fixed in the 4.0.1 release. This list might
7495 not be complete (that is, it is possible that some PRs that have been
7496 fixed are not listed here).
7500 This is the [13]list of problem reports (PRs) from GCC's bug tracking
7501 system that are known to be fixed in the 4.0.2 release. This list might
7502 not be complete (that is, it is possible that some PRs that have been
7503 fixed are not listed here).
7505 Unfortunately, due to a release engineering failure, this release has a
7506 regression on Solaris that will affect some C++ programs. We suggest
7507 that Solaris users apply a [14]patch that corrects the problem. Users
7508 who do not wish to apply the patch should explicitly link C++ programs
7509 with the -pthreads option, even if they do not use threads. This
7510 problem has been corrected in the current 4.0 branch sources and will
7511 not be present in GCC 4.0.3.
7515 Starting with this release, the function getcontext is recognized by
7516 the compiler as having the same semantics as the setjmp function. In
7517 particular, the compiler will ensure that all registers are dead before
7518 calling such a function and will emit a warning about the variables
7519 that may be clobbered after the second return from the function.
7523 This is the [15]list of problem reports (PRs) from GCC's bug tracking
7524 system that are known to be fixed in the 4.0.4 release. This list might
7525 not be complete (that is, it is possible that some PRs that have been
7526 fixed are not listed here).
7528 The 4.0.4 release is provided for those that require a high degree of
7529 binary compatibility with previous 4.0.x releases. For most users, the
7530 GCC team recommends that version 4.1.1 or later be used instead."
7533 For questions related to the use of GCC, please consult these web
7534 pages and the [16]GCC manuals. If that fails, the
7535 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
7536 web pages and the development of GCC are welcome on our developer
7537 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
7540 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
7541 distribution of this entire article is permitted in any medium,
7542 provided this notice is preserved.
7544 These pages are [21]maintained by the GCC team. Last modified
7549 1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4
7550 2. http://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html
7551 3. http://gcc.gnu.org/projects/tree-ssa/
7552 4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html
7553 5. http://gcc.gnu.org/news/sms.html
7554 6. http://www.akkadia.org/drepper/dsohowto.pdf
7555 7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility
7556 8. http://mentorembedded.github.com/cxx-abi/
7557 9. http://gcc.gnu.org/fortran/
7558 10. https://gcc.gnu.org/install/
7559 11. https://gcc.gnu.org/wiki/Visibility
7560 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1
7561 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2
7562 14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html
7563 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4
7564 16. https://gcc.gnu.org/onlinedocs/
7565 17. mailto:gcc-help@gcc.gnu.org
7566 18. mailto:gcc@gcc.gnu.org
7567 19. https://gcc.gnu.org/lists.html
7568 20. http://www.fsf.org/
7569 21. https://gcc.gnu.org/about.html
7570 22. http://validator.w3.org/check/referer
7571 ======================================================================
7572 http://gcc.gnu.org/gcc-3.4/index.html
7573 GCC 3.4 Release Series
7577 The [1]GNU project and the GCC developers are pleased to announce the
7578 release of GCC 3.4.6.
7580 This release is a bug-fix release, containing fixes for regressions in
7581 GCC 3.4.4 relative to previous releases of GCC. This is the last of the
7584 The GCC 3.4 release series includes numerous [2]new features,
7585 improvements, bug fixes, and other changes, thanks to an [3]amazing
7586 group of volunteers.
7591 March 6, 2006 ([4]changes)
7594 November 30, 2005 ([5]changes)
7597 May 18, 2005 ([6]changes)
7600 November 4, 2004 ([7]changes)
7603 September 6, 2004 ([8]changes)
7606 July 1, 2004 ([9]changes)
7609 April 18, 2004 ([10]changes)
7611 References and Acknowledgements
7613 GCC used to stand for the GNU C Compiler, but since the compiler
7614 supports several other languages aside from C, it now stands for the
7615 GNU Compiler Collection.
7617 A list of [11]successful builds is updated as new information becomes
7620 The GCC developers would like to thank the numerous people that have
7621 contributed new features, improvements, bug fixes, and other changes as
7622 well as test results to GCC. This [12]amazing group of volunteers is
7623 what makes GCC successful.
7625 For additional information about GCC please refer to the [13]GCC
7626 project web site or contact the [14]GCC development mailing list.
7628 To obtain GCC please use [15]our mirror sites, or [16]our SVN server.
7631 For questions related to the use of GCC, please consult these web
7632 pages and the [17]GCC manuals. If that fails, the
7633 [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
7634 web pages and the development of GCC are welcome on our developer
7635 list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
7638 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
7639 distribution of this entire article is permitted in any medium,
7640 provided this notice is preserved.
7642 These pages are [22]maintained by the GCC team. Last modified
7647 1. http://www.gnu.org/
7648 2. http://gcc.gnu.org/gcc-3.4/changes.html
7649 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
7650 4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
7651 5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5
7652 6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4
7653 7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3
7654 8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2
7655 9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1
7656 10. http://gcc.gnu.org/gcc-3.4/changes.html
7657 11. http://gcc.gnu.org/gcc-3.4/buildstat.html
7658 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
7659 13. http://gcc.gnu.org/index.html
7660 14. mailto:gcc@gcc.gnu.org
7661 15. http://gcc.gnu.org/mirrors.html
7662 16. http://gcc.gnu.org/svn.html
7663 17. https://gcc.gnu.org/onlinedocs/
7664 18. mailto:gcc-help@gcc.gnu.org
7665 19. mailto:gcc@gcc.gnu.org
7666 20. https://gcc.gnu.org/lists.html
7667 21. http://www.fsf.org/
7668 22. https://gcc.gnu.org/about.html
7669 23. http://validator.w3.org/check/referer
7670 ======================================================================
7671 http://gcc.gnu.org/gcc-3.4/changes.html
7672 GCC 3.4 Release Series
7673 Changes, New Features, and Fixes
7675 The final release in the 3.4 release series is [1]GCC 3.4.6. The series
7678 GCC 3.4 has [2]many improvements in the C++ frontend. Before reporting
7679 a bug, please make sure it's really GCC, and not your code, that is
7684 * GNU Make is now required to build GCC.
7685 * With -nostdinc the preprocessor used to ignore both standard
7686 include paths and include paths contained in environment variables.
7687 It was neither documented nor intended that environment variable
7688 paths be ignored, so this has been corrected.
7689 * GCC no longer accepts the options -fvolatile, -fvolatile-global and
7690 -fvolatile-static. It is unlikely that they worked correctly in any
7692 * GCC no longer ships <varargs.h>. Use <stdarg.h> instead.
7693 * Support for all the systems [3]obsoleted in GCC 3.3 has been
7694 removed from GCC 3.4. See below for a [4]list of systems which are
7695 obsoleted in this release.
7696 * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C
7697 compilers will not work.
7698 * The implementation of the [5]MIPS ABIs has changed. As a result,
7699 the code generated for certain MIPS targets will not be binary
7700 compatible with earlier releases.
7701 * In previous releases, the MIPS port had a fake "hilo" register with
7702 the user-visible name accum. This register has been removed.
7703 * The implementation of the [6]SPARC ABIs has changed. As a result,
7704 the code generated will not be binary compatible with earlier
7705 releases in certain cases.
7706 * The configure option --enable-threads=pthreads has been removed;
7707 use --enable-threads=posix instead, which should have the same
7709 * Code size estimates used by inlining heuristics for C, Objective-C,
7710 C++ and Java have been redesigned significantly. As a result the
7711 parameters of -finline-insns, --param max-inline-insns-single and
7712 --param max-inline-insns-auto need to be reconsidered.
7713 * --param max-inline-slope and --param min-inline-insns have been
7714 removed; they are not needed for the new bottom-up inlining
7716 * The new unit-at-a-time compilation scheme has several compatibility
7718 + The order in which functions, variables, and top-level asm
7719 statements are emitted may have changed. Code relying on some
7720 particular ordering needs to be updated. The majority of such
7721 top-level asm statements can be replaced by section
7723 + Unreferenced static variables and functions are removed. This
7724 may result in undefined references when an asm statement
7725 refers to the variable/function directly. In that case either
7726 the variable/function shall be listed in asm statement operand
7727 or in the case of top-level asm statements the attribute used
7728 shall be used to force function/variable to be always output
7729 and considered as a possibly used by unknown code.
7730 For variables the attribute is accepted only by GCC 3.4 and
7731 newer, while for earlier versions it is sufficient to use
7732 unused to silence warnings about the variables not being
7733 referenced. To keep code portable across different GCC
7734 versions, you can use appropriate preprocessor conditionals.
7735 + Static functions now can use non-standard passing conventions
7736 that may break asm statements calling functions directly.
7737 Again the attribute used shall be used to prevent this
7739 As a temporary workaround, -fno-unit-at-a-time can be used, but
7740 this scheme may not be supported by future releases of GCC.
7741 * GCC 3.4 automatically places zero-initialized variables in the .bss
7742 section on some operating systems. Versions of GNU Emacs up to (and
7743 including) 21.3 will not work correctly when using this
7744 optimization; you can use -fno-zero-initialized-in-bss to disable
7746 * If GCC 3.4 is configured with --enable-threads=posix (the default
7747 on most targets that support pthreads) then _REENTRANT will be
7748 defined unconditionally by some libstdc++ headers. C++ code which
7749 relies on that macro to detect whether multi-threaded code is being
7750 compiled might change in meaning, possibly resulting in linker
7751 errors for single-threaded programs. Affected users of [7]Boost
7752 should compile single-threaded code with -DBOOST_DISABLE_THREADS.
7753 See Bugzilla for [8]more information.
7755 General Optimizer Improvements
7757 * Usability of the profile feedback and coverage testing has been
7759 + Performance of profiled programs has been improved by faster
7760 profile merging code.
7761 + Better use of the profile feedback for optimization (loop
7762 unrolling and loop peeling).
7763 + File locking support allowing fork() calls and parallel runs
7764 of profiled programs.
7765 + Coverage file format has been redesigned.
7766 + gcov coverage tool has been improved.
7767 + make profiledbootstrap available to build a faster compiler.
7768 Experiments made on i386 hardware showed an 11% speedup on -O0
7769 and a 7.5% speedup on -O2 compilation of a [9]large C++
7771 + New value profiling pass enabled via -fprofile-values
7772 + New value profile transformations pass enabled via -fvpt aims
7773 to optimize some code sequences by exploiting knowledge about
7774 value ranges or other properties of the operands. At the
7775 moment a conversion of expensive divisions into cheaper
7776 operations has been implemented.
7777 + New -fprofile-generate and -fprofile-use command-line options
7778 to simplify the use of profile feedback.
7779 * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and
7780 Java which is enabled via -funit-at-a-time (and implied by -O2). In
7781 this scheme a whole file is parsed first and optimized later. The
7782 following basic inter-procedural optimizations are implemented:
7783 + Removal of unreachable functions and variables
7784 + Discovery of local functions (functions with static linkage
7785 whose address is never taken)
7786 + On i386, these local functions use register parameter passing
7788 + Reordering of functions in topological order of the call graph
7789 to enable better propagation of optimizing hints (such as the
7790 stack alignments needed by functions) in the back end.
7791 + Call graph based out-of-order inlining heuristics which allows
7792 to limit overall compilation unit growth (--param
7793 inline-unit-growth).
7794 Overall, the unit-at-a-time scheme produces a 1.3% improvement for
7795 the SPECint2000 benchmark on the i386 architecture (AMD Athlon
7797 * More realistic code size estimates used by inlining for C,
7798 Objective-C, C++ and Java. The growth of large functions can now be
7799 limited via --param large-function-insns and --param
7800 large-function-growth.
7801 * A new cfg-level loop optimizer pass replaces the old loop unrolling
7802 pass and adds two other loop transformations -- loop peeling and
7803 loop unswitching -- and also uses the profile feedback to limit
7804 code growth. (The three optimizations are enabled by
7805 -funroll-loops, -fpeel-loops and -funswitch-loops flags,
7807 The old loop unroller still can be enabled by -fold-unroll-loops
7808 and may produce better code in some cases, especially when the
7809 webizer optimization pass is not run.
7810 * A new web construction pass enabled via -fweb (and implied by -O3)
7811 improves the quality of register allocation, CSE, first scheduling
7812 pass and some other optimization passes by avoiding re-use of
7813 pseudo registers with non-overlapping live ranges. The pass almost
7814 always improves code quality but does make debugging difficult and
7815 thus is not enabled by default by -O2
7816 The pass is especially effective as cleanup after code duplication
7817 passes, such as the loop unroller or the tracer.
7818 * Experimental implementations of superblock or trace scheduling in
7819 the second scheduling pass can be enabled via
7820 -fsched2-use-superblocks and -fsched2-use-traces, respectively.
7822 New Languages and Language specific improvements
7826 * The Ada front end has been updated to include numerous bug fixes
7827 and enhancements. These include:
7828 + Improved project file support
7829 + Additional set of warnings about potential wrong code
7830 + Improved error messages
7831 + Improved code generation
7832 + Improved cross reference information
7834 + Better run-time check elimination
7835 + Better error recovery
7836 + More efficient implementation of unbounded strings
7837 + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools,
7839 + New GNAT.xxxx packages (e.g. GNAT.Strings,
7840 GNAT.Exception_Action)
7842 + New -gnatS switch replacing gnatpsta
7843 + Implementation of new Ada features (in particular limited
7844 with, limited aggregates)
7848 * Precompiled headers are now supported. Precompiled headers can
7849 dramatically speed up compilation of some projects. There are some
7850 known defects in the current precompiled header implementation that
7851 will result in compiler crashes in relatively rare situations.
7852 Therefore, precompiled headers should be considered a "technology
7853 preview" in this release. Read the manual for details about how to
7854 use precompiled headers.
7855 * File handling in the preprocessor has been rewritten. GCC no longer
7856 gets confused by symlinks and hardlinks, and now has a correct
7857 implementation of #import and #pragma once. These two directives
7858 have therefore been un-deprecated.
7859 * The undocumented extension that allowed C programs to have a label
7860 at the end of a compound statement, which has been deprecated since
7861 GCC 3.0, has been removed.
7862 * The cast-as-lvalue extension has been removed for C++ and
7863 deprecated for C and Objective-C. In particular, code like this:
7871 is no longer accepted for C++ and will not be accepted for C and
7872 Objective-C in a future version.
7873 * The conditional-expression-as-lvalue extension has been deprecated
7874 for C and Objective-C. In particular, code like this:
7878 will not be accepted for C and Objective-C in a future version.
7879 * The compound-expression-as-lvalue extension has been deprecated for
7880 C and Objective-C. In particular, code like this:
7884 will not be accepted for C and Objective-C in a future version. A
7885 possible non-intrusive workaround is the following:
7888 * Several [10]built-in functions such as __builtin_popcount for
7889 counting bits, finding the highest and lowest bit in a word, and
7890 parity have been added.
7891 * The -fwritable-strings option has been deprecated and will be
7893 * Many C math library functions are now recognized as built-ins and
7895 * The C, C++, and Objective-C compilers can now handle source files
7896 written in any character encoding supported by the host C library.
7897 The default input character set is taken from the current locale,
7898 and may be overridden with the -finput-charset command line option.
7899 In the future we will add support for inline encoding markers.
7903 * G++ is now much closer to full conformance to the ISO/ANSI C++
7904 standard. This means, among other things, that a lot of invalid
7905 constructs which used to be accepted in previous versions will now
7906 be rejected. It is very likely that existing C++ code will need to
7907 be fixed. This document lists some of the most common issues.
7908 * A hand-written recursive-descent C++ parser has replaced the
7909 YACC-derived C++ parser from previous GCC releases. The new parser
7910 contains much improved infrastructure needed for better parsing of
7911 C++ source codes, handling of extensions, and clean separation
7912 (where possible) between proper semantics analysis and parsing. The
7913 new parser fixes many bugs that were found in the old parser.
7914 * You must now use the typename and template keywords to disambiguate
7915 dependent names, as required by the C++ standard.
7917 typedef int mytype_t;
7920 template <class T1> struct A {
7921 template <class T2> struct B {
7925 template <int N> void bar(void)
7927 // Use 'typename' to tell the parser that T1::mytype_t names
7928 // a type. This is needed because the name is dependent (in
7929 // this case, on template parameter T1).
7930 typename T1::mytype_t x;
7935 template <class T> void template_func(void)
7937 // Use 'template' to prefix member templates within
7938 // dependent types (a has type A<T>, which depends on
7939 // the template parameter T).
7941 a.template bar<0>();
7943 // Use 'template' to tell the parser that B is a nested
7944 // template class (dependent on template parameter T), and
7945 // 'typename' because the whole A<T>::B<int> is
7946 // the name of a type (again, dependent).
7947 typename A<T>::template B<int> b;
7951 void non_template_func(void)
7953 // Outside of any template class or function, no names can be
7954 // dependent, so the use of the keyword 'typename' and 'template'
7955 // is not needed (and actually forbidden).
7961 * In a template definition, unqualified names will no longer find
7962 members of a dependent base (as specified by [temp.dep]/3 in the
7963 C++ standard). For example,
7964 template <typename T> struct B {
7972 template <typename T> struct C : B<T> {
7977 n = 0; // ::n is modified
7978 g (); // ::g is called
7981 You must make the names dependent, e.g. by prefixing them with
7982 this->. Here is the corrected definition of C<T>::h,
7983 template <typename T> void C<T>::h ()
7990 As an alternative solution (unfortunately not backwards compatible
7991 with GCC 3.3), you may use using declarations instead of this->:
7992 template <typename T> struct C : B<T> {
8005 * In templates, all non-dependent names are now looked up and bound
8006 at definition time (while parsing the code), instead of later when
8007 the template is instantiated. For instance:
8010 template <int> struct A {
8011 static void bar(void){
8020 A<0>::bar(); // Calls foo(int), used to call foo(char).
8023 * In an explicit instantiation of a class template, you must use
8024 class or struct before the template-id:
8028 template A<0>; // error, not accepted anymore
8029 template class A<0>; // OK
8030 * The "named return value" and "implicit typename" extensions have
8032 * Default arguments in function types have been deprecated and will
8034 * ARM-style name-injection of friend declarations has been deprecated
8035 and will be removed. For example: struct S { friend void f(); };
8036 void g() { f(); } will not be accepted by future versions of G++;
8037 instead a declaration of "f" will need to be present outside of the
8039 * Covariant returns are implemented for all but varadic functions
8040 that require an adjustment.
8041 * When -pedantic is used, G++ now issues errors about spurious
8042 semicolons. For example,
8043 namespace N {}; // Invalid semicolon.
8044 void f() {}; // Invalid semicolon.
8045 * G++ no longer accepts attributes for a declarator after the
8046 initializer associated with that declarator. For example,
8047 X x(1) __attribute__((...));
8048 is no longer accepted. Instead, use:
8049 X x __attribute__((...)) (1);
8050 * Inside the scope of a template class, the name of the class itself
8051 can be treated as either a class or a template. So GCC used to
8052 accept the class name as argument of type template, and template
8053 template parameter. However this is not C++ standard compliant. Now
8054 the name is not treated as a valid template template argument
8055 unless you qualify the name by its scope. For example, the code
8056 below no longer compiles.
8057 template <template <class> class TT> class X {};
8058 template <class T> class Y {
8059 X<Y> x; // Invalid, Y is always a type template parameter.
8061 The valid code for the above example is
8062 X< ::Y> x; // Valid.
8063 (Notice the space between < and : to prevent GCC to interpret this
8064 as a digraph for [.)
8065 * Friend declarations that refer to template specializations are
8066 rejected if the template has not already been declared. For
8068 template <typename T>
8070 friend void f<> (C&);
8072 is rejected. You must first declare f as a template,
8073 template <typename T>
8075 * In case of friend declarations, every name used in the friend
8076 declaration must be accessible at the point of that declaration.
8077 Previous versions of G++ used to be less strict about this and
8078 allowed friend declarations for private class members, for example.
8079 See the ISO C++ Standard Committee's [11]defect report #209 for
8081 * Declaration of member functions of class templates as friends are
8082 supported. For example,
8083 template <typename T> struct A {
8087 template <typename T> friend void A<T>::f();
8089 * You must use template <> to introduce template specializations, as
8090 required by the standard. For example,
8091 template <typename T>
8095 is rejected. You must write,
8096 template <> struct S<int> {};
8097 * G++ used to accept code like this,
8100 void f(int i = g());
8103 This behavior is not mandated by the standard. Now G++ issues an
8104 error about this code. To avoid the error, you must move the
8105 declaration of g before the declaration of f. The default arguments
8106 for g must be visible at the point where it is called.
8107 * The C++ ABI Section 3.3.3 specifications for the array construction
8108 routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return
8109 NULL when the allocator argument returns NULL. These changes are
8110 incorporated into the libstdc++ runtime library.
8111 * Using a name introduced by a typedef in a friend declaration or in
8112 an explicit instantiation is now rejected, as specified by the ISO
8117 friend class B; // error, no typedef name here
8118 friend B; // error, friend always needs class/struct/enum
8119 friend class A; // OK
8122 template <int> class Q {};
8124 template class R; // error, no typedef name here
8125 template class Q<0>; // OK
8126 * When allocating an array with a new expression, GCC used to allow
8127 parentheses around the type name. This is actually ill-formed and
8129 int* a = new (int)[10]; // error, not accepted anymore
8130 int* a = new int[10]; // OK
8131 * When binding an rvalue of class type to a reference, the copy
8132 constructor of the class must be accessible. For instance, consider
8140 A(const A&); // private copy ctor
8148 foo(A()); // error, copy ctor is not accessible
8149 foo(makeA()); // error, copy ctor is not accessible
8152 foo(a1); // OK, a1 is a lvalue
8154 This might be surprising at first sight, especially since most
8155 popular compilers do not correctly implement this rule ([12]further
8157 * When forming a pointer to member or a pointer to member function,
8158 access checks for class visibility (public, protected, private) are
8159 now performed using the qualifying scope of the name itself. This
8160 is better explained with an example:
8176 &A::pub_func; // OK, pub_func is accessible through A
8177 &A::prot_func; // error, cannot access prot_func through A
8178 &A::priv_func; // error, cannot access priv_func through A
8180 &B::pub_func; // OK, pub_func is accessible through B
8181 &B::prot_func; // OK, can access prot_func through B (within B)
8182 &B::priv_func; // error, cannot access priv_func through B
8186 Runtime Library (libstdc++)
8188 * Optimization work:
8189 + Streamlined streambuf, filebuf, separate synched with C
8190 Standard I/O streambuf.
8191 + All formatted I/O now uses cached locale information.
8192 + STL optimizations (memory/speed for list, red-black trees as
8193 used by sets and maps).
8194 + More use of GCC builtins.
8195 + String optimizations (avoid contention on
8196 increment/decrement-and-test of the reference count in the
8197 empty-string object, constructor from input_iterators
8199 * Static linkage size reductions.
8200 * Large File Support (files larger than 2 GB on 32-bit systems).
8201 * Wide character and variable encoding filebuf work (UTF-8, Unicode).
8202 * Generic character traits.
8203 * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x,
8204 Solaris 2.7 and above, AIX 5.x, Irix 6.5.
8205 * The allocator class is now standard-conformant, and two additional
8206 extension allocators have been added, mt_alloc and
8208 * PCH support: -include bits/stdc++.h (2x compile speedup).
8209 * Rewrote __cxa_demangle with support for C++ style allocators.
8210 * New debug modes for STL containers and iterators.
8211 * Testsuite rewrite: five times as many tests, plus increasingly
8212 sophisticated tests, including I/O, MT, multi-locale, wide and
8214 * Use current versions of GNU "autotools" for build/configuration.
8218 * The Objective-C front end has been updated to include the numerous
8219 bug fixes and enhancements previously available only in Apple's
8220 version of GCC. These include:
8221 + Structured exception (@try... @catch... @finally, @throw) and
8222 synchronization (@synchronized) support. These are accessible
8223 via the -fobjc-exceptions switch; as of this writing, they may
8224 only be used in conjunction with -fnext-runtime on Mac OS X
8225 10.3 and later. See [13]Options Controlling Objective-C
8226 Dialect for more information.
8227 + An overhaul of @encode logic. The C99 _Bool and C++ bool type
8228 may now be encoded as 'B'. In addition, the back-end/codegen
8229 dependencies have been removed.
8230 + An overhaul of message dispatch construction, ensuring that
8231 the various receiver types (and casts thereof) are handled
8232 properly, and that correct diagnostics are issued.
8233 + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue"
8234 (-freplace-objc-classes) debugging modes, currently available
8235 on Mac OS X 10.3 and later. See [14]Options Controlling
8236 Objective-C Dialect for more information.
8237 + Access to optimized runtime entry points (-fno-nil-receivers )
8238 on the assumption that message receivers are never nil. This
8239 is currently available on Mac OS X 10.3 and later. See
8240 [15]Options Controlling Objective-C Dialect for more
8245 * Compiling a .jar file will now cause non-.class entries to be
8246 automatically compiled as resources.
8247 * libgcj has been ported to Darwin.
8248 * Jeff Sturm has adapted Jan Hubicka's call graph optimization code
8250 * libgcj has a new gcjlib URL type; this lets URLClassLoader load
8251 code from shared libraries.
8252 * libgcj has been much more completely merged with [16]GNU Classpath.
8253 * Class loading is now much more correct; in particular the caller's
8254 class loader is now used when that is required.
8255 * [17]Eclipse 2.x will run out of the box using gij.
8256 * Parts of java.nio have been implemented. Direct and indirect
8257 buffers work, as do fundamental file and socket operations.
8258 * java.awt has been improved, though it is still not ready for
8260 * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST
8262 * The MinGW port has matured. Enhancements include socket timeout
8263 support, thread interruption, improved Runtime.exec() handling and
8264 support for accented characters in filenames.
8268 * Fortran improvements are listed in the [18]Fortran documentation.
8270 New Targets and Target Specific Improvements
8274 * Several [19]built-in functions have been added such as
8275 __builtin_alpha_zap to allow utilizing the more obscure
8276 instructions of the CPU.
8277 * Parameter passing of complex arguments has changed to match the
8278 [20]ABI. This change is incompatible with previous GCC versions,
8279 but does fix compatibility with the Tru64 compiler and several
8280 corner cases where GCC was incompatible with itself.
8284 * Nicolas Pitre has contributed his hand-coded floating-point support
8285 code for ARM. It is both significantly smaller and faster than the
8286 existing C-based implementation, even when building applications
8287 for Thumb. The arm-elf configuration has been converted to use the
8289 * Support for the Intel's iWMMXt architecture, a second generation
8290 XScale processor, has been added. Enabled at run time with the
8291 -mcpu=iwmmxt command line switch.
8292 * A new ARM target has been added: arm-wince-pe. This is similar to
8293 the arm-pe target, but it defaults to using the APCS32 ABI.
8294 * The existing ARM pipeline description has been converted to the use
8295 the [21]DFA processor pipeline model. There is not much change in
8296 code performance, but the description is now [22]easier to
8298 * Support for the Cirrus EP9312 Maverick floating point co-processor
8299 added. Enabled at run time with the -mcpu=ep9312 command line
8300 switch. Note however that the multilibs to support this chip are
8301 currently disabled in gcc/config/arm/t-arm-elf, so if you want to
8302 enable their production you will have to uncomment the entries in
8307 * Support for long long has been added.
8308 * Support for saveall attribute has been added.
8309 * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code
8310 for H8/300H and H8S, which is much faster than the previous
8312 * A lot of small performance improvements.
8314 IA-32/AMD64 (x86-64)
8316 * Tuning for K8 (AMD Opteron/Athlon64) core is available via
8317 -march=k8 and -mcpu=k8.
8318 * Scalar SSE code generation carefully avoids reformatting penalties,
8319 hidden dependencies and minimizes the number of uops generated on
8320 both Intel and AMD CPUs.
8321 * Vector MMX and SSE operands are now passed in registers to improve
8322 performance and match the argument passing convention used by the
8323 Intel C++ Compiler. As a result it is not possible to call
8324 functions accepting vector arguments compiled by older GCC version.
8325 * Conditional jump elimination is now more aggressive on modern CPUs.
8326 * The Athlon ports has been converted to use the DFA processor
8327 pipeline description.
8328 * Optimization of indirect tail calls is now possible in a similar
8329 fashion as direct sibcall optimization.
8330 * Further small performance improvements.
8331 * -m128bit-long-double is now less buggy.
8332 * __float128 support in 64-bit compilation.
8333 * Support for data structures exceeding 2GB in 64-bit mode.
8334 * -mcpu has been renamed to -mtune.
8338 * Tuning code for the Itanium 2 processor has been added. The
8339 generation of code tuned for Itanium 2 (option -mtune=itanium2) is
8340 enabled by default now. To generate code tuned for Itanium 1 the
8341 option -mtune=itanium1 should be used.
8342 * [23]DFA processor pipeline descriptions for the IA-64 processors
8343 have been added. This resulted in about 3% improvement on the
8344 SPECInt2000 benchmark for Itanium 2.
8345 * Instruction bundling for the IA-64 processors has been rewritten
8346 using the DFA pipeline hazard recognizer. It resulted in about 60%
8347 compiler speedup on the SPECInt2000 C programs.
8351 * Support for the M32R/2 processor has been added by Renesas.
8352 * Support for an M32R GNU/Linux target and PIC code generation has
8353 been added by Renesas.
8357 * Bernardo Innocenti (Develer S.r.l.) has contributed the
8358 m68k-uclinux target, based on former work done by Paul Dale
8359 (SnapGear Inc.). Code generation for the ColdFire processors family
8360 has been enhanced and extended to support the MCF 53xx and MCF 54xx
8361 cores, integrating former work done by Peter Barada (Motorola).
8365 Processor-specific changes
8367 * Support for the RM7000 and RM9000 processors has been added. It can
8368 be selected using the -march compiler option and should work with
8369 any MIPS I (mips-*) or MIPS III (mips64-*) configuration.
8370 * Support for revision 2 of the MIPS32 ISA has been added. It can be
8371 selected with the command-line option -march=mips32r2.
8372 * There is a new option, -mfix-sb1, to work around certain SB-1
8377 * It is possible to customize GCC using the following configure-time
8379 + --with-arch, which specifies the default value of the -march
8381 + --with-tune, which specifies the default value of the -mtune
8383 + --with-abi, which specifies the default ABI.
8384 + --with-float=soft, which tells GCC to use software floating
8386 + --with-float=hard, which tells GCC to use hardware floating
8388 * A 64-bit GNU/Linux port has been added. The associated
8389 configurations are mips64-linux-gnu and mips64el-linux-gnu.
8390 * The 32-bit GNU/Linux port now supports Java.
8391 * The IRIX 6 configuration now supports the o32 ABI and will build
8392 o32 multilibs by default. This support is compatible with both
8393 binutils and the SGI tools, but note that several features,
8394 including debugging information and DWARF2 exception handling, are
8395 only available when using the GNU assembler. Use of the GNU
8396 assembler and linker (version 2.15 or above) is strongly
8398 * The IRIX 6 configuration now supports 128-bit long doubles.
8399 * There are two new RTEMS-specific configurations, mips-rtems and
8401 * There are two new *-elf configurations, mipsisa32r2-elf and
8406 * Several [24]ABI bugs have been fixed. Unfortunately, these changes
8407 will break binary compatibility with earlier releases.
8408 * GCC can now use explicit relocation operators when generating
8409 -mabicalls code. This behavior is controlled by -mexplicit-relocs
8410 and can have several performance benefits. For example:
8411 + It allows for more optimization of GOT accesses, including
8412 better scheduling and redundancy elimination.
8413 + It allows sibling calls to be implemented as jumps.
8414 + n32 and n64 leaf functions can use a call-clobbered global
8415 pointer instead of $28.
8416 + The code to set up $gp can be removed from functions that
8418 * A new option, -mxgot, allows the GOT to be bigger than 64k. This
8419 option is equivalent to the assembler's -xgot option and should be
8420 used instead of -Wa,-xgot.
8421 * Frame pointer elimination is now supported when generating 64-bit
8423 * Inline block moves have been optimized to take more account of
8424 alignment information.
8425 * Many internal changes have been made to the MIPS port, mostly aimed
8426 at reducing the reliance on assembler macros.
8430 * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64
8431 [25]ABI incompatibilities regarding the way parameters are passed
8432 during functions calls. These changes may result in incompatibility
8433 between code compiled with GCC 3.3 and GCC 3.4.
8437 * Support for shared/dylib gcc libraries has been added. It is
8438 enabled by default on powerpc-apple-darwin7.0.0 and up.
8439 * Libgcj is enabled by default. On systems older than
8440 powerpc-apple-darwin7.0.0 you need to install dlcompat.
8441 * 128-bit IBM extended precision format support added for long
8446 * By default, PowerPC64 GNU/Linux now uses natural alignment of
8447 structure elements. The old four byte alignment for double, with
8448 special rules for a struct starting with a double, can be chosen
8449 with -malign-power. This change may result in incompatibility
8450 between code compiled with GCC 3.3 and GCC 3.4.
8451 * -mabi=altivec is now the default rather than -mabi=no-altivec.
8452 * 128-bit IBM extended precision format support added for long
8457 * New command-line options allow to specify the intended execution
8458 environment for generated code:
8459 + -mesa/-mzarch allows to specify whether to generate code
8460 running in ESA/390 mode or in z/Architecture mode (this is
8461 applicable to 31-bit code only).
8462 + -march allows to specify a minimum processor architecture
8463 level (g5, g6, z900, or z990).
8464 + -mtune allows to specify which processor to tune for.
8465 * It is possible to customize GCC using the following configure-time
8467 + --with-mode, which specifies whether to default to assuming
8468 ESA/390 or z/Architecture mode.
8469 + --with-arch, which specifies the default value of the -march
8471 + --with-tune, which specifies the default value of the -mtune
8473 * Support for the z990 processor has been added, and can be selected
8474 using -march=z990 or -mtune=z990. This includes instruction
8475 scheduling tuned for the superscalar instruction pipeline of the
8476 z990 processor as well as support for all new instructions provided
8477 by the long-displacement facility.
8478 * Support to generate 31-bit code optimized for zSeries processors
8479 (running in ESA/390 or in z/Architecture mode) has been added. This
8480 can be selected using -march=z900 and -mzarch respectively.
8481 * Instruction scheduling for the z900 and z990 processors now uses
8482 the DFA pipeline hazard recognizer.
8483 * GCC no longer generates code to maintain a stack backchain,
8484 previously used to generate stack backtraces for debugging
8485 purposes. As replacement that does not incur runtime overhead,
8486 DWARF-2 call frame information is provided by GCC; this is
8487 supported by GDB 6.1. The old behavior can be restored using the
8489 * The stack frame size of functions may now exceed 2 GB in 64-bit
8491 * A port for the 64-bit IBM TPF operating system has been added; the
8492 configuration is s390x-ibm-tpf. This configuration is supported as
8493 cross-compilation target only.
8494 * Various changes to improve the generated code have been
8495 implemented, including:
8496 + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT
8497 instructions to significantly speed up many floating-point
8499 + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL
8500 WITH BORROW instructions to speed up long long arithmetic.
8501 + GCC now uses the SEARCH STRING instruction to implement
8503 + In many cases, function call overhead for 31-bit code has been
8504 reduced by placing the literal pool after the function code
8505 instead of after the function prolog.
8506 + Register 14 is no longer reserved in 64-bit code.
8507 + Handling of global register variables has been improved.
8511 * The option -mflat is deprecated.
8512 * Support for large (> 2GB) frames has been added to the 64-bit port.
8513 * Several [26]ABI bugs have been fixed. Unfortunately, these changes
8514 will break binary compatibility with earlier releases.
8515 * The default debugging format has been switched from STABS to
8516 DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already
8517 the default debugging format for 64-bit code on Solaris.
8521 * Support for the SH2E processor has been added. Enabled at run time
8522 with the -m2e command line switch, or at configure time by
8523 specifying sh2e as the machine part of the target triple.
8527 * Support for the Mitsubishi V850E1 processor has been added. This is
8528 a variant of the V850E processor with some additional debugging
8533 * Several ABI bugs have been fixed. Unfortunately, these changes
8534 break binary compatibility with earlier releases.
8535 + For big-endian processors, the padding of aggregate return
8536 values larger than a word has changed. If the size of an
8537 aggregate return value is not a multiple of 32 bits, previous
8538 versions of GCC inserted padding in the most-significant bytes
8539 of the first return value register. Aggregates larger than a
8540 word are now padded in the least-significant bytes of the last
8541 return value register used. Aggregates smaller than a word are
8542 still padded in the most-significant bytes. The return value
8543 padding has not changed for little-endian processors.
8544 + Function arguments with 16-byte alignment are now properly
8546 + The implementation of the va_list type has changed. A va_list
8547 value created by va_start from a previous release cannot be
8548 used with va_arg from this release, or vice versa.
8549 * More processor configuration options for Xtensa processors are
8551 + the ABS instruction is now optional;
8552 + the ADDX* and SUBX* instructions are now optional;
8553 + an experimental CONST16 instruction can be used to synthesize
8554 constants instead of loading them from constant pools.
8555 These and other Xtensa processor configuration options can no
8556 longer be enabled or disabled by command-line options; the
8557 processor configuration must be specified by the xtensa-config.h
8558 header file when building GCC. Additionally, the
8559 -mno-serialize-volatile option is no longer supported.
8563 Support for a number of older systems has been declared obsolete in GCC
8564 3.4. Unless there is activity to revive them, the next release of GCC
8565 will have their sources permanently removed.
8567 All configurations of the following processor architectures have been
8569 * Mitsubishi D30V, d30v-*
8570 * AT&T DSP1600 and DSP1610, dsp16xx-*
8573 Also, some individual systems have been obsoleted:
8575 + Support for generating code for operation in APCS/26 mode
8578 + "Bigfoot" port, i370-*. (The other port, s390-*, is actively
8579 maintained and supported.)
8581 + MOSS, i?86-moss-msdos and i?86-*-moss*
8582 + NCR 3000 running System V r.4, i?86-ncr-sysv4*
8583 + FreeBSD with a.out object format, i?86-*-freebsd*aout* and
8585 + GNU/Linux with a.out object format, i?86-linux*aout*
8586 + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1*
8587 + Interix versions before Interix 3, i?86-*-interix
8588 + Mach microkernel, i?86-mach*
8589 + SCO UnixWare with UDK, i?86-*-udk*
8590 + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]*
8591 + VSTa microkernel, i386-*-vsta
8592 * Motorola M68000 family
8593 + HPUX, m68k-hp-hpux* and m68000-hp-hpux*
8594 + NetBSD with a.out object format (before NetBSD 1.4),
8595 m68k-*-*-netbsd* except m68k-*-*-netbsdelf*
8596 + Generic System V r.4, m68k-*-sysv4*
8598 + Generic VAX, vax-*-* (This is generic VAX only; we have not
8599 obsoleted any VAX triples for specific operating systems.)
8601 Documentation improvements
8603 Other significant improvements
8605 * The build system has undergone several significant cleanups.
8606 Subdirectories will only be configured if they are being built, and
8607 all subdirectory configures are run from the make command. The top
8608 level has been autoconfiscated.
8609 * Building GCC no longer writes to its source directory. This should
8610 help those wishing to share a read-only source directory over NFS
8611 or build from a CD. The exceptions to this feature are if you
8612 configure with either --enable-maintainer-mode or
8613 --enable-generated-files-in-srcdir.
8614 * The -W warning option has been renamed to -Wextra, which is more
8615 easily understood. The older spelling will be retained for
8616 backwards compatibility.
8617 * Substantial improvements in compile time have been made,
8618 particularly for non-optimizing compilations.
8619 __________________________________________________________________
8625 A vast number of bugs have been fixed in 3.4.0, too many to publish a
8626 complete list here. [27]Follow this link to query the Bugzilla database
8627 for the list of over 900 bugs fixed in 3.4.0. This is the list of all
8628 bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4
8630 __________________________________________________________________
8636 This section lists the problem reports (PRs) from GCC's bug tracking
8637 system that are known to be fixed in the 3.4.1 release. This list might
8638 not be complete (that is, it is possible that some PRs that have been
8639 fixed are not listed here).
8643 * [28]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler
8644 emitted - PIC related
8645 * [29]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf
8646 * [30]14760 A bug in configure.in prevents using both
8647 --program-suffix and --program-prefix
8648 * [31]14671 [hppa64] bootstrap fails: ICE in
8649 save_call_clobbered_regs, in caller_save.c
8650 * [32]15093 [alpha][Java] make bootstrap fails to configure libffi on
8652 * [33]15178 Solaris 9/x86 fails linking after stage 3
8654 Multi-platform internal compiler errors (ICEs)
8656 * [34]12753 (preprocessor) Memory corruption in preprocessor on bad
8658 * [35]13985 ICE in gcc.c-torture/compile/930621-1.c
8659 * [36]14810 (c++) tree check failures with invalid code involving
8661 * [37]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in
8663 * [38]15044 (c++) ICE on syntax error, template header
8664 * [39]15057 (c++) Compiling of conditional value throw constructs
8665 cause a segmentation violation
8666 * [40]15064 (c++) typeid of template parameter gives ICE
8667 * [41]15142 (c++) ICE when passing a string where a char* is expected
8668 in a throw statement
8669 * [42]15159 ICE in rtl_verify_flow_info_1
8670 * [43]15165 (c++) ICE in instantiate_template
8671 * [44]15193 Unary minus using pointer to V4SF vector causes
8672 -fforce-mem to exhaust all memory
8673 * [45]15209 (c++) Runs out of memory with packed structs
8674 * [46]15227 (c++) Trouble with invalid function definition
8675 * [47]15285 (c++) instantiate_type ICE when forming pointer to
8677 * [48]15299 (c++) ICE in resolve_overloaded_unification
8678 * [49]15329 (c++) ICE on constructor of member template
8679 * [50]15550 ICE in extract_insn, in recog.c
8680 * [51]15554 (c++) ICE in tsubst_copy, in cp/pt.c
8681 * [52]15640 (c++) ICE on invalid code in arg_assoc, in
8683 * [53]15666 [unit-at-a-time] Gcc abort on valid code
8684 * [54]15696 (c++) ICE with bad pointer-to-member code
8685 * [55]15701 (c++) ICE with friends and template template parameter
8686 * [56]15761 ICE in do_SUBST, in combine.c
8687 * [57]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops
8691 * [58]14538 All RTEMS targets broken for gnat
8695 * [59]12391 missing warning about assigning to an incomplete type
8696 * [60]14649 atan(1.0) should not be a constant expression
8697 * [61]15004 [unit-at-a-time] no warning for unused paramater in
8699 * [62]15749 --pedantic-errors behaves differently from --pedantic
8700 with C-compiler on GNU/Linux
8702 C++ compiler and library
8704 * [63]10646 non-const reference is incorrectly matched in a "const T"
8705 partial specialization
8706 * [64]12077 wcin.rdbuf()->in_avail() return value too high
8707 * [65]13598 enc_filebuf doesn't work
8708 * [66]14211 const_cast returns lvalue but should be rvalue
8709 * [67]14220 num_put::do_put() undesired float/double behavior
8710 * [68]14245 problem with user-defined allocators in std::basic_string
8711 * [69]14340 libstdc++ Debug mode: failure to convert iterator to
8713 * [70]14600 __gnu_cxx::stdio_sync_filebuf should expose internal
8715 * [71]14668 no warning anymore for reevaluation of declaration
8716 * [72]14775 LFS (large file support) tests missing
8717 * [73]14821 Duplicate namespace alias declaration should not conflict
8718 * [74]14930 Friend declaration ignored
8719 * [75]14932 cannot use offsetof to get offsets of array elements in
8721 * [76]14950 [non unit-at-a-time] always_inline does not mix with
8723 * [77]14962 g++ ignores #pragma redefine_extname
8724 * [78]14975 Segfault on low-level write error during imbue
8725 * [79]15002 Linewise stream input is unusably slow (std::string slow)
8726 * [80]15025 compiler accepts redeclaration of template as
8728 * [81]15046 [arm] Math functions misdetected by cross configuration
8729 * [82]15069 a bit test on a variable of enum type is miscompiled
8730 * [83]15074 g++ -lsupc++ still links against libstdc++
8731 * [84]15083 spurious "statement has no effect" warning
8732 * [85]15096 parse error with templates and pointer to const member
8733 * [86]15287 combination of operator[] and operator .* fails in
8735 * [87]15317 __attribute__ unused in first parameter of constructor
8737 * [88]15337 sizeof on incomplete type diagnostic
8738 * [89]15361 bitset<>::_Find_next fails
8739 * [90]15412 _GLIBCXX_ symbols symbols defined and used in different
8741 * [91]15427 valid code results in incomplete type error
8742 * [92]15471 Incorrect member pointer offsets in anonymous
8744 * [93]15503 nested template problem
8745 * [94]15507 compiler hangs while laying out union
8746 * [95]15542 operator & and template definitions
8747 * [96]15565 SLES9: leading + sign for unsigned int with showpos
8748 * [97]15625 friend defined inside a template fails to find static
8750 * [98]15629 Function templates, overloads, and friend name injection
8751 * [99]15742 'noreturn' attribute ignored in method of template
8753 * [100]15775 Allocator::pointer consistently ignored
8754 * [101]15821 Duplicate namespace alias within namespace rejected
8755 * [102]15862 'enum yn' fails (confict with undeclared builtin)
8756 * [103]15875 rejects pointer to member in template
8757 * [104]15877 valid code using templates and anonymous enums is
8759 * [105]15947 Puzzling error message for wrong destructor declaration
8761 * [106]16020 cannot copy __gnu_debug::bitset
8762 * [107]16154 input iterator concept too restrictive
8763 * [108]16174 deducing top-level consts
8767 * [109]14315 Java compiler is not parallel make safe
8771 * [110]15151 [g77] incorrect logical i/o in 64-bit mode
8775 * [111]7993 private variables cannot be shadowed in subclasses
8779 * [112]15228 useless copies of floating point operands
8780 * [113]15345 [non-unit-at-a-time] unreferenced nested inline
8781 functions not optimized away
8782 * [114]15945 Incorrect floating point optimization
8783 * [115]15526 ftrapv aborts on 0 * (-1)
8784 * [116]14690 Miscompiled POOMA tests
8785 * [117]15112 GCC generates code to write to unchanging memory
8789 * [118]15067 Minor glitch in the source of cpp
8791 Main driver program bugs
8793 * [119]1963 collect2 interprets -oldstyle_liblookup as -o
8796 x86-specific (Intel/AMD)
8798 * [120]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND*
8803 * [121]14782 GCC produces an unaligned data access at -O2
8804 * [122]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2
8805 * [123]15202 ICE in reload_cse_simplify_operands, in postreload.c
8809 * [124]14610 __float80 constants incorrectly emitted
8810 * [125]14813 init_array sections are initialized in the wrong order
8811 * [126]14857 GCC segfault on duplicated asm statement
8812 * [127]15598 Gcc 3.4 ICE on valid code
8813 * [128]15653 Gcc 3.4 ICE on valid code
8817 * [129]15189 wrong filling of delay slot with -march=mips1 -G0
8818 -mno-split-addresses -mno-explicit-relocs
8819 * [130]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as
8821 * [131]16144 Bogus reference to __divdf3 when -O1
8822 * [132]16176 Miscompilation of unaligned data in MIPS backend
8826 * [133]11591 ICE in gcc.dg/altivec-5.c
8827 * [134]12028 powerpc-eabispe produces bad sCOND operation
8828 * [135]14478 rs6000 geu/ltu patterns generate incorrect code
8829 * [136]14567 long double and va_arg complex args
8830 * [137]14715 Altivec stack layout may overlap gpr save with stack
8832 * [138]14902 (libstdc++) Stream checking functions fail when -pthread
8834 * [139]14924 Compiler ICE on valid code
8835 * [140]14960 -maltivec affects vector return with -mabi=no-altivec
8836 * [141]15106 vector varargs failure passing from altivec to
8837 non-altivec code for -m32
8838 * [142]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 &
8840 * [143]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx
8842 * [144]15662 Segmentation fault when an exception is thrown - even if
8843 try and catch are specified
8847 * [145]15054 Bad code due to overlapping stack temporaries
8851 * [146]15783 ICE with union assignment in 64-bit mode
8852 * [147]15626 GCC 3.4 emits "ld: warning: relocation error:
8857 * [148]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64
8858 * [149]14723 Backported -march=nocona from mainline
8859 * [150]15290 __float128 failed to pass to function properly
8861 Cygwin/Mingw32-specific
8863 * [151]15250 Option -mms-bitfields support on GCC 3.4 is not
8864 conformant to MS layout
8865 * [152]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe
8866 worker on windows32 targets
8868 Bugs specific to embedded processors
8870 * [153]8309 [m68k] -m5200 produces erroneous SImode set of short
8872 * [154]13250 [SH] Gcc code for rotation clobbers the register, but
8873 gcc continues to use the register as if it was not clobbered
8874 * [155]13803 [coldfire] movqi operand constraints too restrictivefor
8876 * [156]14093 [SH] ICE for code when using -mhitachi option in SH
8877 * [157]14457 [m6811hc] ICE with simple c++ source
8878 * [158]14542 [m6811hc] ICE on simple source
8879 * [159]15100 [SH] cc1plus got hang-up on
8880 libstdc++-v3/testsuite/abi_check.cc
8881 * [160]15296 [CRIS] Delayed branch scheduling causing invalid code on
8883 * [161]15396 [SH] ICE with -O2 -fPIC
8884 * [162]15782 [coldfire] m68k_output_mi_thunk emits wrong code for
8887 Testsuite problems (compiler not affected)
8889 * [163]11610 libstdc++ testcases 27_io/* don't work properly remotely
8890 * [164]15488 (libstdc++) possibly insufficient file permissions for
8891 executing test suite
8892 * [165]15489 (libstdc++) testsuite_files determined incorrectly
8896 * [166]13928 (libstdc++) no whatis info in some man pages generated
8898 * [167]14150 Ada documentation out of date
8899 * [168]14949 (c++) Need to document method visibility changes
8900 * [169]15123 libstdc++-doc: Allocators.3 manpage is empty
8901 __________________________________________________________________
8907 This section lists the problem reports (PRs) from GCC's bug tracking
8908 system that are known to be fixed in the 3.4.2 release. This list might
8909 not be complete (that is, it is possible that some PRs that have been
8910 fixed are not listed here).
8912 Bootstrap failures and issues
8914 * [170]16469 [mips-sgi-irix5.3] bootstrap fails in
8915 libstdc++-v3/testsuite
8916 * [171]16344 [hppa-linux-gnu] libstdc++'s PCH built by
8917 profiledbootstrap does not work with the built compiler
8918 * [172]16842 [Solaris/x86] mkheaders can not find mkheaders.conf
8920 Multi-platform internal compiler errors (ICEs)
8922 * [173]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in
8923 cp_parser_class_specifier, in cp/parser.c
8924 * [174]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
8925 * [175]15461 (c++) ICE due to NRV and inlining
8926 * [176]15890 (c++) ICE in c_expand_expr, in c-common.c
8927 * [177]16180 ICE: segmentation fault in RTL optimization
8928 * [178]16224 (c++) ICE in write_unscoped_name (template/namespace)
8929 * [179]16408 ICE: in delete_insn, in cfgrtl.c
8930 * [180]16529 (c++) ICE for: namespace-alias shall not be declared as
8931 the name of any other entity
8932 * [181]16698 (c++) ICE with exceptions and declaration of __cxa_throw
8933 * [182]16706 (c++) ICE in finish_member_declaration, in
8935 * [183]16810 (c++) Legal C++ program with cast gives ICE in
8937 * [184]16851 (c++) ICE when throwing a comma expression
8938 * [185]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c
8939 * [186]16904 (c++) ICE in finish_class_member_access_expr, in
8941 * [187]16905 (c++) ICE (segfault) with exceptions
8942 * [188]16964 (c++) ICE in cp_parser_class_specifier due to
8944 * [189]17068 (c++) ICE: tree check: expected class 'd', have 'x'
8945 (identifier_node) in dependent_template_p, in cp/pt.c
8949 * [190]16366 Preprocessor option -remap causes memory corruption
8953 * [191]15345 unreferenced nested inline functions not optimized away
8954 * [192]16590 Incorrect execution when compiling with -O2
8955 * [193]16693 Bitwise AND is lost when used within a cast to an enum
8956 of the same precision
8957 * [194]17078 Jump into if(0) substatement fails
8959 Problems in generated debug information
8961 * [195]13956 incorrect stabs for nested local variables
8965 * [196]16684 GCC should not warn about redundant redeclarations of
8968 C++ compiler and library
8970 * [197]12658 Thread safety problems in locale::global() and
8972 * [198]13092 g++ accepts invalid pointer-to-member conversion
8973 * [199]15320 Excessive memory consumption
8974 * [200]16246 Incorrect template argument deduction
8975 * [201]16273 Memory exhausted when using nested classes and virtual
8977 * [202]16401 ostringstream in gcc 3.4.x very slow for big data
8978 * [203]16411 undefined reference to
8979 __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char>
8981 * [204]16489 G++ incorrectly rejects use of a null constant integral
8982 expression as a null constant pointer
8983 * [205]16618 offsetof fails with constant member
8984 * [206]16637 syntax error reported for valid input code
8985 * [207]16717 __attribute__((constructor)) broken in C++
8986 * [208]16813 compiler error in DEBUG version of range insertion
8988 * [209]16853 pointer-to-member initialization from incompatible one
8990 * [210]16889 ambiguity is not detected
8991 * [211]16959 Segmentation fault in ios_base::sync_with_stdio
8993 Java compiler and library
8995 * [212]7587 direct threaded interpreter not thread-safe
8996 * [213]16473 ServerSocket accept() leaks file descriptors
8997 * [214]16478 Hash synchronization deadlock with finalizers
9001 * [215]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c
9002 * [216]16974 could not split insn (ice in final_scan_insn, in
9007 * [217]16298 ICE in output_operand
9008 * [218]17113 ICE with SSE2 intrinsics
9012 * [219]14697 libstdc++ couldn't find 32bit libgcc_s
9016 * [220]15869 [mips64] No NOP after LW (with -mips1 -O0)
9017 * [221]16325 [mips64] value profiling clobbers gp on mips
9018 * [222]16357 [mipsisa64-elf] ICE copying 7 bytes between extern
9020 * [223]16380 [mips64] Use of uninitialised register after dbra
9022 * [224]16407 [mips64] Unaligned access to local variables
9023 * [225]16643 [mips64] verify_local_live_at_start ICE after
9024 crossjumping & cfgcleanup
9028 * [226]15927 THUMB -O2: strength-reduced iteration variable ends up
9030 * [227]15948 THUMB: ICE with non-commutative cbranch
9031 * [228]17019 THUMB: bad switch statement in md code for
9032 addsi3_cbranch_scratch
9036 * [229]16130 ICE on valid code: in bundling, in config/ia64/ia64.c
9038 * [230]16142 ICE on valid code: in bundling, in config/ia64/ia64.c
9040 * [231]16278 Gcc failed to build Linux kernel with -mtune=merced
9041 * [232]16414 ICE on valid code: typo in comparison of asm_noperands
9043 * [233]16445 ICE on valid code: don't count ignored insns
9044 * [234]16490 ICE (segfault) while compiling with -fprofile-use
9045 * [235]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS
9049 * [236]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x
9050 * [237]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions
9055 * [238]16199 ICE while compiling apache 2.0.49
9056 * [239]16416 -m64 doesn't imply -mcpu=v9 anymore
9057 * [240]16430 ICE when returning non-C aggregates larger than 16 bytes
9059 Bugs specific to embedded processors
9061 * [241]16379 [m32r] can't output large model function call of memcpy
9062 * [242]17093 [m32r] ICE with -msdata=use -O0
9063 * [243]17119 [m32r] ICE at switch case 0x8000
9067 * [244]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp
9069 Alpha Tru64-specific
9071 * [245]16210 libstdc++ gratuitously omits "long long" I/O
9073 Testsuite, documentation issues (compiler is not affected):
9075 * [246]15488 (libstdc++) possibly insufficient file permissions for
9076 executing test suite
9077 * [247]16250 ada/doctools runs makeinfo even in release tarball
9078 __________________________________________________________________
9082 This is the [248]list of problem reports (PRs) from GCC's bug tracking
9083 system that are known to be fixed in the 3.4.3 release. This list might
9084 not be complete (that is, it is possible that some PRs that have been
9085 fixed are not listed here).
9089 * [249]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1
9090 * [250]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold
9093 Internal compiler errors (ICEs) affecting multiple platforms
9095 * [251]13948 (java) GCJ segmentation fault while compiling GL4Java
9097 * [252]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c
9098 * [253]16301 (c++) ICE when "strong" attribute is attached to a using
9100 * [254]16566 ICE with flexible arrays
9101 * [255]17023 ICE with nested functions in parameter declaration
9102 * [256]17027 ICE with noreturn function in loop at -O2
9103 * [257]17524 ICE in grokdeclarator, in cp/decl.c
9104 * [258]17826 (c++) ICE in cp_tree_equal
9106 C and optimization bugs
9108 * [259]15526 -ftrapv aborts on 0 * (-1)
9109 * [260]16999 #ident stopped working
9110 * [261]17503 quadratic behaviour in invalid_mode_change_p
9111 * [262]17581 Long long arithmetic fails inside a switch/case
9112 statement when compiled with -O2
9113 * [263]18129 -fwritable-strings doesn't work
9115 C++ compiler and library bugs
9117 * [264]10975 incorrect initial ostringstream::tellp()
9118 * [265]11722 Unbuffered filebuf::sgetn is slow
9119 * [266]14534 Unrecognizing static function as a template parameter
9120 when its return value is also templated
9121 * [267]15172 Copy constructor optimization in aggregate
9123 * [268]15786 Bad error message for frequently occuring error.
9124 * [269]16162 Rejects valid member-template-definition
9125 * [270]16612 empty basic_strings can't live in shared memory
9126 * [271]16715 std::basic_iostream is instantiated when used, even
9127 though instantiations are already contained in libstdc++
9128 * [272]16848 code in /ext/demangle.h appears broken
9129 * [273]17132 GCC fails to eliminate function template specialization
9130 when argument deduction fails
9131 * [274]17259 One more _S_leaf incorrectly qualified with _RopeRep::
9133 * [275]17327 use of `enumeral_type' in template type unification
9134 * [276]17393 "unused variable '._0'" warning with -Wall
9135 * [277]17501 Confusion with member templates
9136 * [278]17537 g++ not passing -lstdc++ to linker when all command line
9137 arguments are libraries
9138 * [279]17585 usage of unqualified name of static member from within
9140 * [280]17821 Poor diagnostic for using "." instead of "->"
9141 * [281]17829 wrong error: call of overloaded function is ambiguous
9142 * [282]17851 Misleading diagnostic for invalid function declarations
9143 with undeclared types
9144 * [283]17976 Destructor is called twice
9145 * [284]18020 rejects valid definition of enum value in template
9146 * [285]18093 bogus conflict in namespace aliasing
9147 * [286]18140 C++ parser bug when using >> in templates
9151 * [287]17541 data statements with double precision constants fail
9155 * [288]17853 -O2 ICE for MMX testcase
9159 * [289]17245 ICE compiling gsl-1.5 statistics/lag1.c
9163 * [290]17167 FATAL:Symbol L_foo$stub already defined.
9167 * [291]17277 could not catch an exception when specified -maix64
9171 * [292]17505 <cmath> calls acosf(), ceilf(), and other functions
9172 missing from system libraries
9176 * [293]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl
9180 * [294]17384 ICE with mode attribute on structures
9184 * [295]17770 No NOP after LWL with -mips1
9186 Other embedded target specific
9188 * [296]11476 [arc-elf] gcc ICE on newlib's vfprintf.c
9189 * [297]14064 [avr-elf] -fdata-sections triggers ICE
9190 * [298]14678 [m68hc11-elf] gcc ICE
9191 * [299]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__
9192 * [300]15790 [i686-coff] Alignment error building gcc with i686-coff
9194 * [301]15886 [SH] Miscompilation with -O2 -fPIC
9195 * [302]16884 [avr-elf] [fweb related] bug while initializing
9198 Bugs relating to debugger support
9200 * [303]13841 missing debug info for _Complex function arguments
9201 * [304]15860 [big-endian targets] No DW_AT_location debug info is
9202 emitted for formal arguments to a function that uses "register"
9205 Testsuite issues (compiler not affected)
9207 * [305]17465 Testsuite in libffi overrides LD_LIBRARY_PATH
9208 * [306]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH
9209 * [307]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit
9214 * [308]15498 typo in gcc manual: non-existing locale example en_UK,
9216 * [309]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap:
9217 document broken shell
9218 * [310]16406 USE_LD_AS_NEEDED undocumented
9219 __________________________________________________________________
9223 This is the [311]list of problem reports (PRs) from GCC's bug tracking
9224 system that are known to be fixed in the 3.4.4 release. This list might
9225 not be complete (that is, it is possible that some PRs that have been
9226 fixed are not listed here).
9227 __________________________________________________________________
9231 This is the [312]list of problem reports (PRs) from GCC's bug tracking
9232 system that are known to be fixed in the 3.4.5 release. This list might
9233 not be complete (that is, it is possible that some PRs that have been
9234 fixed are not listed here).
9238 * [313]24688 sco_math fixincl breaks math.h
9242 * [314]17188 struct Foo { } redefinition
9243 * [315]20187 wrong code for ((unsigned char)(unsigned long
9244 long)((a?a:1)&(a*b)))?0:1)
9245 * [316]21873 infinite warning loop on bad array initializer
9246 * [317]21899 enum definition accepts values to be overriden
9247 * [318]22061 ICE in find_function_data, in function.c
9248 * [319]22308 Failure to diagnose violation of constraint 6.516p2
9249 * [320]22458 ICE on missing brace
9250 * [321]22589 ICE casting to long long
9251 * [322]24101 Segfault with preprocessed source
9253 C++ compiler and library bugs
9255 * [323]10611 operations on vector mode not recognized in C++
9256 * [324]13377 unexpected behavior of namespace usage directive
9257 * [325]16002 Strange error message with new parser
9258 * [326]17413 local classes as template argument
9259 * [327]17609 spurious error message after using keyword
9260 * [328]17618 ICE in cp_convert_to_pointer, in cp/cvt.c
9261 * [329]18124 ICE with invalid template template parameter
9262 * [330]18155 typedef in template declaration not rejected
9263 * [331]18177 ICE with const_cast for undeclared variable
9264 * [332]18368 C++ error message regression
9265 * [333]16378 ICE when returning a copy of a packed member
9266 * [334]18466 int ::i; accepted
9267 * [335]18512 ICE on invalid usage of template base class
9268 * [336]18454 ICE when returning undefined type
9269 * [337]18738 typename not allowed with non-dependent qualified name
9270 * [338]18803 rejects access to operator() in template
9271 * [339]19004 ICE in uses_template_parms, in cp/pt.c
9272 * [340]19208 Spurious error about variably modified type
9273 * [341]18253 bad error message / ICE for invalid template parameter
9274 * [342]19608 ICE after friend function definition in local class
9275 * [343]19884 ICE on explicit instantiation of a non-template
9277 * [344]20153 ICE when C++ template function contains anonymous union
9278 * [345]20563 Infinite loop in diagnostic (and ice after error
9280 * [346]20789 ICE with incomplete type in template
9281 * [347]21336 Internal compiler error when using custom new operators
9282 * [348]21768 ICE in error message due to violation of coding
9284 * [349]21853 constness of pointer to data member ignored
9285 * [350]21903 Default argument of template function causes a
9287 * [351]21983 multiple diagnostics
9288 * [352]21987 New testsuite failure
9289 g++.dg/warn/conversion-function-1.C
9290 * [353]22153 ICE on invalid template specialization
9291 * [354]22172 Internal compiler error, seg fault.
9292 * [355]21286 filebuf::xsgetn vs pipes
9293 * [356]22233 ICE with wrong number of template parameters
9294 * [357]22508 ICE after invalid operator new
9295 * [358]22545 ICE with pointer to class member & user defined
9297 * [359]23528 Wrong default allocator in ext/hash_map
9298 * [360]23550 char_traits requirements/1.cc test bad math
9299 * [361]23586 Bad diagnostic for invalid namespace-name
9300 * [362]23624 ICE in invert_truthvalue, in fold-const.c
9301 * [363]23639 Bad error message: not a member of '<declaration error>'
9302 * [364]23797 ICE on typename outside template
9303 * [365]23965 Bogus error message: no matching function for call to
9305 * [366]24052 &#`label_decl' not supported by dump_expr#<expression
9307 * [367]24580 virtual base class cause exception not to be caught
9309 Problems in generated debug information
9311 * [368]24267 Bad DWARF for altivec vectors
9313 Optimizations issues
9315 * [369]17810 ICE in verify_local_live_at_start
9316 * [370]17860 Wrong generated code for loop with varying bound
9317 * [371]21709 ICE on compile-time complex NaN
9318 * [372]21964 broken tail call at -O2 or more
9319 * [373]22167 Strange optimization bug when using -Os
9320 * [374]22619 Compilation failure for real_const_1.f and
9322 * [375]23241 Invalid code generated for comparison of uchar to 255
9323 * [376]23478 Miscompilation due to reloading of a var that is also
9325 * [377]24470 segmentation fault in cc1plus when compiling with -O
9326 * [378]24950 ICE in operand_subword_force
9328 Precompiled headers problems
9330 * [379]14400 Cannot compile qt-x11-free-3.3.0
9331 * [380]14940 PCH largefile test fails on various platforms
9335 * [381]20239 ICE on empty preprocessed input
9336 * [382]15220 "gcc -E -MM -MG" reports missing system headers in
9341 * [383]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on
9346 * [384]21888 bootstrap failure with linker relaxation enabled
9350 * [385]15342 [arm-linux]: ICE in verify_local_live_at_start
9351 * [386]23985 Memory aliasing information incorrect in inlined memcpy
9355 * [387]16719 Illegal move of byte into address register causes
9360 * [388]21723 ICE while building libgfortran
9361 * [389]21841 -mhp-ld/-mgnu-ld documentation
9365 * [390]23644 IA-64 hardware models and configuration options
9367 * [391]24718 Shared libgcc not used for linking by default
9371 * [392]18421 ICE in reload_cse_simplify_operands, in postreload.c
9375 * [393]20621 ICE in change_address_1, in emit-rtl.c
9377 PowerPC and PowerPC64 specific
9379 * [394]18583 error on valid code: const
9380 __attribute__((altivec(vector__))) doesn't work in arrays
9381 * [395]20191 ICE in reload_cse_simplify_operands
9382 * [396]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined
9383 * [397]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set
9384 * [398]23404 gij trashes args of functions with more than 8 fp args
9385 * [399]23539 C & C++ compiler generating misaligned references
9386 regardless of compiler flags
9387 * [400]24102 floatdisf2_internal2 broken
9388 * [401]24465 -mminimal-toc miscompilation of __thread vars
9392 * [402]19933 Problem with define of HUGE_VAL in math_c99
9393 * [403]21889 Native Solaris assembler cannot grok DTP-relative debug
9398 * [404]19300 PCH failures on sparc-linux
9399 * [405]20301 Assembler labels have a leading "-"
9400 * [406]20673 C PCH testsuite assembly comparison failure
9402 x86 and x86_64 specific
9404 * [407]18582 ICE with arrays of type V2DF
9405 * [408]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2
9407 * [409]21716 ICE in reg-stack.c's swap_rtx_condition
9408 * [410]24315 amd64 fails -fpeephole2
9409 __________________________________________________________________
9413 This is the [411]list of problem reports (PRs) from GCC's bug tracking
9414 system that are known to be fixed in the 3.4.6 release. This list might
9415 not be complete (that is, it is possible that some PRs that have been
9416 fixed are not listed here).
9419 For questions related to the use of GCC, please consult these web
9420 pages and the [412]GCC manuals. If that fails, the
9421 [413]gcc-help@gcc.gnu.org mailing list might help. Comments on these
9422 web pages and the development of GCC are welcome on our developer
9423 list at [414]gcc@gcc.gnu.org. All of [415]our lists have public
9426 Copyright (C) [416]Free Software Foundation, Inc. Verbatim copying and
9427 distribution of this entire article is permitted in any medium,
9428 provided this notice is preserved.
9430 These pages are [417]maintained by the GCC team. Last modified
9435 1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6
9436 2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus
9437 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
9438 4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems
9439 5. http://gcc.gnu.org/gcc-3.4/mips-abi.html
9440 6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
9441 7. http://www.boost.org/
9442 8. https://gcc.gnu.org/PR11953
9443 9. https://gcc.gnu.org/PR8361
9444 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other Builtins
9445 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209
9446 12. http://gcc.gnu.org/bugs/#cxx_rvalbind
9447 13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
9448 14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
9449 15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html
9450 16. http://www.gnu.org/software/classpath/
9451 17. http://www.eclipse.org/
9452 18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html
9453 19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html
9454 20. http://h30097.www3.hp.com/docs/base_doc/DOCUMENTATION/V51A_HTML/ARH9MBTE/DTMNPLTN.HTM#normal-argument-list-structure
9455 21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
9456 22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html
9457 23. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html
9458 24. http://gcc.gnu.org/gcc-3.4/mips-abi.html
9459 25. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html
9460 26. http://gcc.gnu.org/gcc-3.4/sparc-abi.html
9461 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
9462 28. https://gcc.gnu.org/PR10129
9463 29. https://gcc.gnu.org/PR14576
9464 30. https://gcc.gnu.org/PR14760
9465 31. https://gcc.gnu.org/PR14671
9466 32. https://gcc.gnu.org/PR15093
9467 33. https://gcc.gnu.org/PR15178
9468 34. https://gcc.gnu.org/PR12753
9469 35. https://gcc.gnu.org/PR13985
9470 36. https://gcc.gnu.org/PR14810
9471 37. https://gcc.gnu.org/PR14883
9472 38. https://gcc.gnu.org/PR15044
9473 39. https://gcc.gnu.org/PR15057
9474 40. https://gcc.gnu.org/PR15064
9475 41. https://gcc.gnu.org/PR15142
9476 42. https://gcc.gnu.org/PR15159
9477 43. https://gcc.gnu.org/PR15165
9478 44. https://gcc.gnu.org/PR15193
9479 45. https://gcc.gnu.org/PR15209
9480 46. https://gcc.gnu.org/PR15227
9481 47. https://gcc.gnu.org/PR15285
9482 48. https://gcc.gnu.org/PR15299
9483 49. https://gcc.gnu.org/PR15329
9484 50. https://gcc.gnu.org/PR15550
9485 51. https://gcc.gnu.org/PR15554
9486 52. https://gcc.gnu.org/PR15640
9487 53. https://gcc.gnu.org/PR15666
9488 54. https://gcc.gnu.org/PR15696
9489 55. https://gcc.gnu.org/PR15701
9490 56. https://gcc.gnu.org/PR15761
9491 57. https://gcc.gnu.org/PR15829
9492 58. https://gcc.gnu.org/PR14538
9493 59. https://gcc.gnu.org/PR12391
9494 60. https://gcc.gnu.org/PR14649
9495 61. https://gcc.gnu.org/PR15004
9496 62. https://gcc.gnu.org/PR15749
9497 63. https://gcc.gnu.org/PR10646
9498 64. https://gcc.gnu.org/PR12077
9499 65. https://gcc.gnu.org/PR13598
9500 66. https://gcc.gnu.org/PR14211
9501 67. https://gcc.gnu.org/PR14220
9502 68. https://gcc.gnu.org/PR14245
9503 69. https://gcc.gnu.org/PR14340
9504 70. https://gcc.gnu.org/PR14600
9505 71. https://gcc.gnu.org/PR14668
9506 72. https://gcc.gnu.org/PR14775
9507 73. https://gcc.gnu.org/PR14821
9508 74. https://gcc.gnu.org/PR14930
9509 75. https://gcc.gnu.org/PR14932
9510 76. https://gcc.gnu.org/PR14950
9511 77. https://gcc.gnu.org/PR14962
9512 78. https://gcc.gnu.org/PR14975
9513 79. https://gcc.gnu.org/PR15002
9514 80. https://gcc.gnu.org/PR15025
9515 81. https://gcc.gnu.org/PR15046
9516 82. https://gcc.gnu.org/PR15069
9517 83. https://gcc.gnu.org/PR15074
9518 84. https://gcc.gnu.org/PR15083
9519 85. https://gcc.gnu.org/PR15096
9520 86. https://gcc.gnu.org/PR15287
9521 87. https://gcc.gnu.org/PR15317
9522 88. https://gcc.gnu.org/PR15337
9523 89. https://gcc.gnu.org/PR15361
9524 90. https://gcc.gnu.org/PR15412
9525 91. https://gcc.gnu.org/PR15427
9526 92. https://gcc.gnu.org/PR15471
9527 93. https://gcc.gnu.org/PR15503
9528 94. https://gcc.gnu.org/PR15507
9529 95. https://gcc.gnu.org/PR15542
9530 96. https://gcc.gnu.org/PR15565
9531 97. https://gcc.gnu.org/PR15625
9532 98. https://gcc.gnu.org/PR15629
9533 99. https://gcc.gnu.org/PR15742
9534 100. https://gcc.gnu.org/PR15775
9535 101. https://gcc.gnu.org/PR15821
9536 102. https://gcc.gnu.org/PR15862
9537 103. https://gcc.gnu.org/PR15875
9538 104. https://gcc.gnu.org/PR15877
9539 105. https://gcc.gnu.org/PR15947
9540 106. https://gcc.gnu.org/PR16020
9541 107. https://gcc.gnu.org/PR16154
9542 108. https://gcc.gnu.org/PR16174
9543 109. https://gcc.gnu.org/PR14315
9544 110. https://gcc.gnu.org/PR15151
9545 111. https://gcc.gnu.org/PR7993
9546 112. https://gcc.gnu.org/PR15228
9547 113. https://gcc.gnu.org/PR15345
9548 114. https://gcc.gnu.org/PR15945
9549 115. https://gcc.gnu.org/PR15526
9550 116. https://gcc.gnu.org/PR14690
9551 117. https://gcc.gnu.org/PR15112
9552 118. https://gcc.gnu.org/PR15067
9553 119. https://gcc.gnu.org/PR1963
9554 120. https://gcc.gnu.org/PR15717
9555 121. https://gcc.gnu.org/PR14782
9556 122. https://gcc.gnu.org/PR14828
9557 123. https://gcc.gnu.org/PR15202
9558 124. https://gcc.gnu.org/PR14610
9559 125. https://gcc.gnu.org/PR14813
9560 126. https://gcc.gnu.org/PR14857
9561 127. https://gcc.gnu.org/PR15598
9562 128. https://gcc.gnu.org/PR15653
9563 129. https://gcc.gnu.org/PR15189
9564 130. https://gcc.gnu.org/PR15331
9565 131. https://gcc.gnu.org/PR16144
9566 132. https://gcc.gnu.org/PR16176
9567 133. https://gcc.gnu.org/PR11591
9568 134. https://gcc.gnu.org/PR12028
9569 135. https://gcc.gnu.org/PR14478
9570 136. https://gcc.gnu.org/PR14567
9571 137. https://gcc.gnu.org/PR14715
9572 138. https://gcc.gnu.org/PR14902
9573 139. https://gcc.gnu.org/PR14924
9574 140. https://gcc.gnu.org/PR14960
9575 141. https://gcc.gnu.org/PR15106
9576 142. https://gcc.gnu.org/PR16026
9577 143. https://gcc.gnu.org/PR15191
9578 144. https://gcc.gnu.org/PR15662
9579 145. https://gcc.gnu.org/PR15054
9580 146. https://gcc.gnu.org/PR15783
9581 147. https://gcc.gnu.org/PR15626
9582 148. https://gcc.gnu.org/PR14326
9583 149. https://gcc.gnu.org/PR14723
9584 150. https://gcc.gnu.org/PR15290
9585 151. https://gcc.gnu.org/PR15250
9586 152. https://gcc.gnu.org/PR15551
9587 153. https://gcc.gnu.org/PR8309
9588 154. https://gcc.gnu.org/PR13250
9589 155. https://gcc.gnu.org/PR13803
9590 156. https://gcc.gnu.org/PR14093
9591 157. https://gcc.gnu.org/PR14457
9592 158. https://gcc.gnu.org/PR14542
9593 159. https://gcc.gnu.org/PR15100
9594 160. https://gcc.gnu.org/PR15296
9595 161. https://gcc.gnu.org/PR15396
9596 162. https://gcc.gnu.org/PR15782
9597 163. https://gcc.gnu.org/PR11610
9598 164. https://gcc.gnu.org/PR15488
9599 165. https://gcc.gnu.org/PR15489
9600 166. https://gcc.gnu.org/PR13928
9601 167. https://gcc.gnu.org/PR14150
9602 168. https://gcc.gnu.org/PR14949
9603 169. https://gcc.gnu.org/PR15123
9604 170. https://gcc.gnu.org/PR16469
9605 171. https://gcc.gnu.org/PR16344
9606 172. https://gcc.gnu.org/PR16842
9607 173. https://gcc.gnu.org/PR12608
9608 174. https://gcc.gnu.org/PR14492
9609 175. https://gcc.gnu.org/PR15461
9610 176. https://gcc.gnu.org/PR15890
9611 177. https://gcc.gnu.org/PR16180
9612 178. https://gcc.gnu.org/PR16224
9613 179. https://gcc.gnu.org/PR16408
9614 180. https://gcc.gnu.org/PR16529
9615 181. https://gcc.gnu.org/PR16698
9616 182. https://gcc.gnu.org/PR16706
9617 183. https://gcc.gnu.org/PR16810
9618 184. https://gcc.gnu.org/PR16851
9619 185. https://gcc.gnu.org/PR16870
9620 186. https://gcc.gnu.org/PR16904
9621 187. https://gcc.gnu.org/PR16905
9622 188. https://gcc.gnu.org/PR16964
9623 189. https://gcc.gnu.org/PR17068
9624 190. https://gcc.gnu.org/PR16366
9625 191. https://gcc.gnu.org/PR15345
9626 192. https://gcc.gnu.org/PR16590
9627 193. https://gcc.gnu.org/PR16693
9628 194. https://gcc.gnu.org/PR17078
9629 195. https://gcc.gnu.org/PR13956
9630 196. https://gcc.gnu.org/PR16684
9631 197. https://gcc.gnu.org/PR12658
9632 198. https://gcc.gnu.org/PR13092
9633 199. https://gcc.gnu.org/PR15320
9634 200. https://gcc.gnu.org/PR16246
9635 201. https://gcc.gnu.org/PR16273
9636 202. https://gcc.gnu.org/PR16401
9637 203. https://gcc.gnu.org/PR16411
9638 204. https://gcc.gnu.org/PR16489
9639 205. https://gcc.gnu.org/PR16618
9640 206. https://gcc.gnu.org/PR16637
9641 207. https://gcc.gnu.org/PR16717
9642 208. https://gcc.gnu.org/PR16813
9643 209. https://gcc.gnu.org/PR16853
9644 210. https://gcc.gnu.org/PR16889
9645 211. https://gcc.gnu.org/PR16959
9646 212. https://gcc.gnu.org/PR7587
9647 213. https://gcc.gnu.org/PR16473
9648 214. https://gcc.gnu.org/PR16478
9649 215. https://gcc.gnu.org/PR10695
9650 216. https://gcc.gnu.org/PR16974
9651 217. https://gcc.gnu.org/PR16298
9652 218. https://gcc.gnu.org/PR17113
9653 219. https://gcc.gnu.org/PR14697
9654 220. https://gcc.gnu.org/PR15869
9655 221. https://gcc.gnu.org/PR16325
9656 222. https://gcc.gnu.org/PR16357
9657 223. https://gcc.gnu.org/PR16380
9658 224. https://gcc.gnu.org/PR16407
9659 225. https://gcc.gnu.org/PR16643
9660 226. https://gcc.gnu.org/PR15927
9661 227. https://gcc.gnu.org/PR15948
9662 228. https://gcc.gnu.org/PR17019
9663 229. https://gcc.gnu.org/PR16130
9664 230. https://gcc.gnu.org/PR16142
9665 231. https://gcc.gnu.org/PR16278
9666 232. https://gcc.gnu.org/PR16414
9667 233. https://gcc.gnu.org/PR16445
9668 234. https://gcc.gnu.org/PR16490
9669 235. https://gcc.gnu.org/PR16683
9670 236. https://gcc.gnu.org/PR16195
9671 237. https://gcc.gnu.org/PR16239
9672 238. https://gcc.gnu.org/PR16199
9673 239. https://gcc.gnu.org/PR16416
9674 240. https://gcc.gnu.org/PR16430
9675 241. https://gcc.gnu.org/PR16379
9676 242. https://gcc.gnu.org/PR17093
9677 243. https://gcc.gnu.org/PR17119
9678 244. https://gcc.gnu.org/PR15928
9679 245. https://gcc.gnu.org/PR16210
9680 246. https://gcc.gnu.org/PR15488
9681 247. https://gcc.gnu.org/PR16250
9682 248. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3
9683 249. https://gcc.gnu.org/PR17369
9684 250. https://gcc.gnu.org/PR17850
9685 251. https://gcc.gnu.org/PR13948
9686 252. https://gcc.gnu.org/PR14492
9687 253. https://gcc.gnu.org/PR16301
9688 254. https://gcc.gnu.org/PR16566
9689 255. https://gcc.gnu.org/PR17023
9690 256. https://gcc.gnu.org/PR17027
9691 257. https://gcc.gnu.org/PR17524
9692 258. https://gcc.gnu.org/PR17826
9693 259. https://gcc.gnu.org/PR15526
9694 260. https://gcc.gnu.org/PR16999
9695 261. https://gcc.gnu.org/PR17503
9696 262. https://gcc.gnu.org/PR17581
9697 263. https://gcc.gnu.org/PR18129
9698 264. https://gcc.gnu.org/PR10975
9699 265. https://gcc.gnu.org/PR11722
9700 266. https://gcc.gnu.org/PR14534
9701 267. https://gcc.gnu.org/PR15172
9702 268. https://gcc.gnu.org/PR15786
9703 269. https://gcc.gnu.org/PR16162
9704 270. https://gcc.gnu.org/PR16612
9705 271. https://gcc.gnu.org/PR16715
9706 272. https://gcc.gnu.org/PR16848
9707 273. https://gcc.gnu.org/PR17132
9708 274. https://gcc.gnu.org/PR17259
9709 275. https://gcc.gnu.org/PR17327
9710 276. https://gcc.gnu.org/PR17393
9711 277. https://gcc.gnu.org/PR17501
9712 278. https://gcc.gnu.org/PR17537
9713 279. https://gcc.gnu.org/PR17585
9714 280. https://gcc.gnu.org/PR17821
9715 281. https://gcc.gnu.org/PR17829
9716 282. https://gcc.gnu.org/PR17851
9717 283. https://gcc.gnu.org/PR17976
9718 284. https://gcc.gnu.org/PR18020
9719 285. https://gcc.gnu.org/PR18093
9720 286. https://gcc.gnu.org/PR18140
9721 287. https://gcc.gnu.org/PR17541
9722 288. https://gcc.gnu.org/PR17853
9723 289. https://gcc.gnu.org/PR17245
9724 290. https://gcc.gnu.org/PR17167
9725 291. https://gcc.gnu.org/PR17277
9726 292. https://gcc.gnu.org/PR17505
9727 293. https://gcc.gnu.org/PR17684
9728 294. https://gcc.gnu.org/PR17384
9729 295. https://gcc.gnu.org/PR17770
9730 296. https://gcc.gnu.org/PR11476
9731 297. https://gcc.gnu.org/PR14064
9732 298. https://gcc.gnu.org/PR14678
9733 299. https://gcc.gnu.org/PR15583
9734 300. https://gcc.gnu.org/PR15790
9735 301. https://gcc.gnu.org/PR15886
9736 302. https://gcc.gnu.org/PR16884
9737 303. https://gcc.gnu.org/PR13841
9738 304. https://gcc.gnu.org/PR15860
9739 305. https://gcc.gnu.org/PR17465
9740 306. https://gcc.gnu.org/PR17469
9741 307. https://gcc.gnu.org/PR18138
9742 308. https://gcc.gnu.org/PR15498
9743 309. https://gcc.gnu.org/PR15747
9744 310. https://gcc.gnu.org/PR16406
9745 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4
9746 312. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5
9747 313. https://gcc.gnu.org/PR24688
9748 314. https://gcc.gnu.org/PR17188
9749 315. https://gcc.gnu.org/PR20187
9750 316. https://gcc.gnu.org/PR21873
9751 317. https://gcc.gnu.org/PR21899
9752 318. https://gcc.gnu.org/PR22061
9753 319. https://gcc.gnu.org/PR22208
9754 320. https://gcc.gnu.org/PR22458
9755 321. https://gcc.gnu.org/PR22589
9756 322. https://gcc.gnu.org/PR24101
9757 323. https://gcc.gnu.org/PR10611
9758 324. https://gcc.gnu.org/PR13377
9759 325. https://gcc.gnu.org/PR16002
9760 326. https://gcc.gnu.org/PR17413
9761 327. https://gcc.gnu.org/PR17609
9762 328. https://gcc.gnu.org/PR17618
9763 329. https://gcc.gnu.org/PR18124
9764 330. https://gcc.gnu.org/PR18155
9765 331. https://gcc.gnu.org/PR18177
9766 332. https://gcc.gnu.org/PR18368
9767 333. https://gcc.gnu.org/PR18378
9768 334. https://gcc.gnu.org/PR18466
9769 335. https://gcc.gnu.org/PR18512
9770 336. https://gcc.gnu.org/PR18545
9771 337. https://gcc.gnu.org/PR18738
9772 338. https://gcc.gnu.org/PR18803
9773 339. https://gcc.gnu.org/PR19004
9774 340. https://gcc.gnu.org/PR19208
9775 341. https://gcc.gnu.org/PR19253
9776 342. https://gcc.gnu.org/PR19608
9777 343. https://gcc.gnu.org/PR19884
9778 344. https://gcc.gnu.org/PR20153
9779 345. https://gcc.gnu.org/PR20563
9780 346. https://gcc.gnu.org/PR20789
9781 347. https://gcc.gnu.org/PR21336
9782 348. https://gcc.gnu.org/PR21768
9783 349. https://gcc.gnu.org/PR21853
9784 350. https://gcc.gnu.org/PR21903
9785 351. https://gcc.gnu.org/PR21983
9786 352. https://gcc.gnu.org/PR21987
9787 353. https://gcc.gnu.org/PR22153
9788 354. https://gcc.gnu.org/PR22172
9789 355. https://gcc.gnu.org/PR21286
9790 356. https://gcc.gnu.org/PR22233
9791 357. https://gcc.gnu.org/PR22508
9792 358. https://gcc.gnu.org/PR22545
9793 359. https://gcc.gnu.org/PR23528
9794 360. https://gcc.gnu.org/PR23550
9795 361. https://gcc.gnu.org/PR23586
9796 362. https://gcc.gnu.org/PR23624
9797 363. https://gcc.gnu.org/PR23639
9798 364. https://gcc.gnu.org/PR23797
9799 365. https://gcc.gnu.org/PR23965
9800 366. https://gcc.gnu.org/PR24052
9801 367. https://gcc.gnu.org/PR24580
9802 368. https://gcc.gnu.org/PR24267
9803 369. https://gcc.gnu.org/PR17810
9804 370. https://gcc.gnu.org/PR17860
9805 371. https://gcc.gnu.org/PR21709
9806 372. https://gcc.gnu.org/PR21964
9807 373. https://gcc.gnu.org/PR22167
9808 374. https://gcc.gnu.org/PR22619
9809 375. https://gcc.gnu.org/PR23241
9810 376. https://gcc.gnu.org/PR23478
9811 377. https://gcc.gnu.org/PR24470
9812 378. https://gcc.gnu.org/PR24950
9813 379. https://gcc.gnu.org/PR14400
9814 380. https://gcc.gnu.org/PR14940
9815 381. https://gcc.gnu.org/PR20239
9816 382. https://gcc.gnu.org/PR15220
9817 383. https://gcc.gnu.org/PR19275
9818 384. https://gcc.gnu.org/PR21888
9819 385. https://gcc.gnu.org/PR15342
9820 386. https://gcc.gnu.org/PR23985
9821 387. https://gcc.gnu.org/PR16719
9822 388. https://gcc.gnu.org/PR21723
9823 389. https://gcc.gnu.org/PR21841
9824 390. https://gcc.gnu.org/PR23644
9825 391. https://gcc.gnu.org/PR24718
9826 392. https://gcc.gnu.org/PR18421
9827 393. https://gcc.gnu.org/PR20621
9828 394. https://gcc.gnu.org/PR18583
9829 395. https://gcc.gnu.org/PR20191
9830 396. https://gcc.gnu.org/PR22083
9831 397. https://gcc.gnu.org/PR23070
9832 398. https://gcc.gnu.org/PR23404
9833 399. https://gcc.gnu.org/PR23539
9834 400. https://gcc.gnu.org/PR24102
9835 401. https://gcc.gnu.org/PR24465
9836 402. https://gcc.gnu.org/PR19933
9837 403. https://gcc.gnu.org/PR21889
9838 404. https://gcc.gnu.org/PR19300
9839 405. https://gcc.gnu.org/PR20301
9840 406. https://gcc.gnu.org/PR20673
9841 407. https://gcc.gnu.org/PR18582
9842 408. https://gcc.gnu.org/PR19340
9843 409. https://gcc.gnu.org/PR21716
9844 410. https://gcc.gnu.org/PR24315
9845 411. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6
9846 412. https://gcc.gnu.org/onlinedocs/
9847 413. mailto:gcc-help@gcc.gnu.org
9848 414. mailto:gcc@gcc.gnu.org
9849 415. https://gcc.gnu.org/lists.html
9850 416. http://www.fsf.org/
9851 417. https://gcc.gnu.org/about.html
9852 418. http://validator.w3.org/check/referer
9853 ======================================================================
9854 http://gcc.gnu.org/gcc-3.3/index.html
9855 GCC 3.3 Release Series
9859 The [1]GNU project and the GCC developers are pleased to announce the
9860 release of GCC 3.3.6.
9862 This release is a bug-fix release, containing fixes for regressions in
9863 GCC 3.3.5 relative to previous releases of GCC.
9865 This release is the last of the series 3.3.x.
9867 The GCC 3.3 release series includes numerous [2]new features,
9868 improvements, bug fixes, and other changes, thanks to an [3]amazing
9869 group of volunteers.
9874 May 3, 2005 ([4]changes)
9877 September 30, 2004 ([5]changes)
9880 May 31, 2004 ([6]changes)
9883 February 14, 2004 ([7]changes)
9886 October 16, 2003 ([8]changes)
9889 August 8, 2003 ([9]changes)
9892 May 14, 2003 ([10]changes)
9894 References and Acknowledgements
9896 GCC used to stand for the GNU C Compiler, but since the compiler
9897 supports several other languages aside from C, it now stands for the
9898 GNU Compiler Collection.
9900 A list of [11]successful builds is updated as new information becomes
9903 The GCC developers would like to thank the numerous people that have
9904 contributed new features, improvements, bug fixes, and other changes as
9905 well as test results to GCC. This [12]amazing group of volunteers is
9906 what makes GCC successful.
9908 For additional information about GCC please refer to the [13]GCC
9909 project web site or contact the [14]GCC development mailing list.
9911 To obtain GCC please use [15]our mirror sites, or our CVS server.
9914 For questions related to the use of GCC, please consult these web
9915 pages and the [16]GCC manuals. If that fails, the
9916 [17]gcc-help@gcc.gnu.org mailing list might help. Comments on these
9917 web pages and the development of GCC are welcome on our developer
9918 list at [18]gcc@gcc.gnu.org. All of [19]our lists have public
9921 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and
9922 distribution of this entire article is permitted in any medium,
9923 provided this notice is preserved.
9925 These pages are [21]maintained by the GCC team. Last modified
9930 1. http://www.gnu.org/
9931 2. http://gcc.gnu.org/gcc-3.3/changes.html
9932 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
9933 4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
9934 5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5
9935 6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4
9936 7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3
9937 8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2
9938 9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1
9939 10. http://gcc.gnu.org/gcc-3.3/changes.html
9940 11. http://gcc.gnu.org/gcc-3.3/buildstat.html
9941 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
9942 13. http://gcc.gnu.org/index.html
9943 14. mailto:gcc@gcc.gnu.org
9944 15. http://gcc.gnu.org/mirrors.html
9945 16. https://gcc.gnu.org/onlinedocs/
9946 17. mailto:gcc-help@gcc.gnu.org
9947 18. mailto:gcc@gcc.gnu.org
9948 19. https://gcc.gnu.org/lists.html
9949 20. http://www.fsf.org/
9950 21. https://gcc.gnu.org/about.html
9951 22. http://validator.w3.org/check/referer
9952 ======================================================================
9953 http://gcc.gnu.org/gcc-3.3/changes.html
9954 GCC 3.3 Release Series
9955 Changes, New Features, and Fixes
9957 The latest release in the 3.3 release series is [1]GCC 3.3.6.
9961 * The preprocessor no longer accepts multi-line string literals. They
9962 were deprecated in 3.0, 3.1, and 3.2.
9963 * The preprocessor no longer supports the -A- switch when appearing
9964 alone. -A- followed by an assertion is still supported.
9965 * Support for all the systems [2]obsoleted in GCC 3.1 has been
9966 removed from GCC 3.3. See below for a [3]list of systems which are
9967 obsoleted in this release.
9968 * Checking for null format arguments has been decoupled from the rest
9969 of the format checking mechanism. Programs which use the format
9970 attribute may regain this functionality by using the new [4]nonnull
9971 function attribute. Note that all functions for which GCC has a
9972 built-in format attribute, an appropriate built-in nonnull
9973 attribute is also applied.
9974 * The DWARF (version 1) debugging format has been deprecated and will
9975 be removed in a future version of GCC. Version 2 of the DWARF
9976 debugging format will continue to be supported for the foreseeable
9978 * The C and Objective-C compilers no longer accept the "Naming Types"
9979 extension (typedef foo = bar); it was already unavailable in C++.
9980 Code which uses it will need to be changed to use the "typeof"
9981 extension instead: typedef typeof(bar) foo. (We have removed this
9982 extension without a period of deprecation because it has caused the
9983 compiler to crash since version 3.0 and no one noticed until very
9984 recently. Thus we conclude it is not in widespread use.)
9985 * The -traditional C compiler option has been removed. It was
9986 deprecated in 3.1 and 3.2. (Traditional preprocessing remains
9987 available.) The <varargs.h> header, used for writing variadic
9988 functions in traditional C, still exists but will produce an error
9990 * GCC 3.3.1 automatically places zero-initialized variables in the
9991 .bss section on some operating systems. Versions of GNU Emacs up to
9992 (and including) 21.3 will not work correctly when using this
9993 optimization; you can use -fno-zero-initialized-in-bss to disable
9996 General Optimizer Improvements
9998 * A new scheme for accurately describing processor pipelines, the
9999 [5]DFA scheduler, has been added.
10000 * Pavel Nejedly, Charles University Prague, has contributed new file
10001 format used by the edge coverage profiler (-fprofile-arcs).
10002 The new format is robust and diagnoses common mistakes where
10003 profiles from different versions (or compilations) of the program
10004 are combined resulting in nonsensical profiles and slow code to
10005 produced with profile feedback. Additionally this format allows
10006 extra data to be gathered. Currently, overall statistics are
10007 produced helping optimizers to identify hot spots of a program
10008 globally replacing the old intra-procedural scheme and resulting in
10009 better code. Note that the gcov tool from older GCC versions will
10010 not be able to parse the profiles generated by GCC 3.3 and vice
10012 * Jan Hubicka, SuSE Labs, has contributed a new superblock formation
10013 pass enabled using -ftracer. This pass simplifies the control flow
10014 of functions allowing other optimizations to do better job.
10015 He also contributed the function reordering pass
10016 (-freorder-functions) to optimize function placement using profile
10019 New Languages and Language specific improvements
10023 * The preprocessor now accepts directives within macro arguments. It
10024 processes them just as if they had not been within macro arguments.
10025 * The separate ISO and traditional preprocessors have been completely
10026 removed. The front end handles either type of preprocessed output
10028 * In C99 mode preprocessor arithmetic is done in the precision of the
10029 target's intmax_t, as required by that standard.
10030 * The preprocessor can now copy comments inside macros to the output
10031 file when the macro is expanded. This feature, enabled using the
10032 -CC option, is intended for use by applications which place
10033 metadata or directives inside comments, such as lint.
10034 * The method of constructing the list of directories to be searched
10035 for header files has been revised. If a directory named by a -I
10036 option is a standard system include directory, the option is
10037 ignored to ensure that the default search order for system
10038 directories and the special treatment of system header files are
10040 * A few more [6]ISO C99 features now work correctly.
10041 * A new function attribute, nonnull, has been added which allows
10042 pointer arguments to functions to be specified as requiring a
10043 non-null value. The compiler currently uses this information to
10044 issue a warning when it detects a null value passed in such an
10046 * A new type attribute, may_alias, has been added. Accesses to
10047 objects with types with this attribute are not subjected to
10048 type-based alias analysis, but are instead assumed to be able to
10049 alias any other type of objects, just like the char type.
10053 * Type based alias analysis has been implemented for C++ aggregate
10058 * Generate an error if Objective-C objects are passed by value in
10059 function and method calls.
10060 * When -Wselector is used, check the whole list of selectors at the
10061 end of compilation, and emit a warning if a @selector() is not
10063 * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime.
10064 * No longer need to include objc/objc-class.h to compile self calls
10065 in class methods (NeXT runtime only).
10066 * New -Wundeclared-selector option.
10067 * Removed selector bloating which was causing object files to be 10%
10068 bigger on average (GNU runtime only).
10069 * Using at run time @protocol() objects has been fixed in certain
10070 situations (GNU runtime only).
10071 * Type checking has been fixed and improved in many situations
10072 involving protocols.
10076 * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK
10078 * The JDK 1.4 assert facility has been implemented.
10079 * The bytecode interpreter is now direct threaded and thus faster.
10083 * Fortran improvements are listed in [7]the Fortran documentation.
10087 * Ada tasking now works with glibc 2.3.x threading libraries.
10089 New Targets and Target Specific Improvements
10091 * The following changes have been made to the HP-PA port:
10092 + The port now defaults to scheduling for the PA8000 series of
10094 + Scheduling support for the PA7300 processor has been added.
10095 + The 32-bit port now supports weak symbols under HP-UX 11.
10096 + The handling of initializers and finalizers has been improved
10097 under HP-UX 11. The 64-bit port no longer uses collect2.
10098 + Dwarf2 EH support has been added to the 32-bit GNU/Linux port.
10099 + ABI fixes to correct the passing of small structures by value.
10100 * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to
10101 use the DFA processor pipeline description.
10102 * The following NetBSD configurations for the SuperH processor family
10104 + SH3, big-endian, sh-*-netbsdelf*
10105 + SH3, little-endian, shle-*-netbsdelf*
10106 + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd*
10107 + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd*
10108 + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd*
10109 + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd*
10110 * The following changes have been made to the IA-32/x86-64 port:
10111 + SSE2 and 3dNOW! intrinsics are now supported.
10112 + Support for thread local storage has been added to the IA-32
10114 + The x86-64 port has been significantly improved.
10115 * The following changes have been made to the MIPS port:
10116 + All configurations now accept the -mabi switch. Note that you
10117 will need appropriate multilibs for this option to work
10119 + ELF configurations will always pass an ABI flag to the
10120 assembler, except when the MIPS EABI is selected.
10121 + -mabi=64 no longer selects MIPS IV code.
10122 + The -mcpu option, which was deprecated in 3.1 and 3.2, has
10123 been removed from this release.
10124 + -march now changes the core ISA level. In previous releases,
10125 it would change the use of processor-specific extensions, but
10126 would leave the core ISA unchanged. For example, mips64-elf
10127 -march=r8000 will now generate MIPS IV code.
10128 + Under most configurations, -mipsN now acts as a synonym for
10130 + There are some new preprocessor macros to describe the -march
10131 and -mtune settings. See the documentation of those options
10133 + Support for the NEC VR-Series processors has been added. This
10134 includes the 54xx, 5500, and 41xx series.
10135 + Support for the Sandcraft sr71k processor has been added.
10136 * The following changes have been made to the S/390 port:
10137 + Support to build the Java runtime libraries has been added.
10138 Java is now enabled by default on s390-*-linux* and
10139 s390x-*-linux* targets.
10140 + Multilib support for the s390x-*-linux* target has been added;
10141 this allows to build 31-bit binaries using the -m31 option.
10142 + Support for thread local storage has been added.
10143 + Inline assembler code may now use the 'Q' constraint to
10144 specify memory operands without index register.
10145 + Various platform-specific performance improvements have been
10146 implemented; in particular, the compiler now uses the BRANCH
10147 ON COUNT family of instructions and makes more frequent use of
10148 the TEST UNDER MASK family of instructions.
10149 * The following changes have been made to the PowerPC port:
10150 + Support for IBM Power4 processor added.
10151 + Support for Motorola e500 SPE added.
10152 + Support for AIX 5.2 added.
10153 + Function and Data sections now supported on AIX.
10154 + Sibcall optimizations added.
10155 * The support for H8 Tiny is added to the H8/300 port with -mn.
10159 Support for a number of older systems has been declared obsolete in GCC
10160 3.3. Unless there is activity to revive them, the next release of GCC
10161 will have their sources permanently removed.
10163 All configurations of the following processor architectures have been
10165 * Matsushita MN10200, mn10200-*-*
10166 * Motorola 88000, m88k-*-*
10167 * IBM ROMP, romp-*-*
10169 Also, some individual systems have been obsoleted:
10171 + Interix, alpha*-*-interix*
10172 + Linux libc1, alpha*-*-linux*libc1*
10173 + Linux ECOFF, alpha*-*-linux*ecoff*
10175 + Generic a.out, arm*-*-aout*
10176 + Conix, arm*-*-conix*
10177 + "Old ABI," arm*-*-oabi
10178 + StrongARM/COFF, strongarm-*-coff*
10180 + Generic OSF, hppa1.0-*-osf*
10181 + Generic BSD, hppa1.0-*-bsd*
10182 + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]*
10183 + HiUX, hppa*-*-hiux*
10184 + Mach Lites, hppa*-*-lites*
10186 + Windows NT 3.x, i?86-*-win32
10188 + HP systems, m68000-hp-bsd* and m68k-hp-bsd*
10189 + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and
10191 + AT&T systems, m68000-att-sysv*
10192 + Atari systems, m68k-atari-sysv*
10193 + Motorola systems, m68k-motorola-sysv*
10194 + NCR systems, m68k-ncr-sysv*
10195 + Plexus systems, m68k-plexus-sysv*
10196 + Commodore systems, m68k-cbm-sysv*
10197 + Citicorp TTI, m68k-tti-*
10198 + Unos, m68k-crds-unos*
10199 + Concurrent RTU, m68k-ccur-rtu*
10200 + Linux a.out, m68k-*-linux*aout*
10201 + Linux libc1, m68k-*-linux*libc1*
10202 + pSOS, m68k-*-psos*
10204 + Generic ECOFF, mips*-*-ecoff*
10205 + SINIX, mips-sni-sysv4
10206 + Orion RTEMS, mips64orion-*-rtems*
10207 * National Semiconductor 32000
10208 + OpenBSD, ns32k-*-openbsd*
10209 * POWER (aka RS/6000) and PowerPC
10210 + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]*
10211 + Bull BOSX, rs6000-bull-bosx
10212 + Generic Mach, rs6000-*-mach*
10213 + Generic SysV, powerpc*-*-sysv*
10214 + Linux libc1, powerpc*-*-linux*libc1*
10216 + Generic a.out, sparc-*-aout*, sparclet-*-aout*,
10217 sparclite-*-aout*, and sparc86x-*-aout*
10218 + NetBSD a.out, sparc-*-netbsd*aout*
10219 + Generic BSD, sparc-*-bsd*
10220 + ChorusOS, sparc-*-chorusos*
10221 + Linux a.out, sparc-*-linux*aout*
10222 + Linux libc1, sparc-*-linux*libc1*
10223 + LynxOS, sparc-*-lynxos*
10224 + Solaris on HAL hardware, sparc-hal-solaris2*
10225 + SunOS versions 3 and 4, sparc-*-sunos[34]*
10227 + RTEMS, v850-*-rtems*
10231 Documentation improvements
10233 Other significant improvements
10235 * Almost all front-end dependencies in the compiler have been
10236 separated out into a set of language hooks. This should make adding
10237 a new front end clearer and easier.
10238 * One effect of removing the separate preprocessor is a small
10239 increase in the robustness of the compiler in general, and the
10240 maintainability of target descriptions. Previously target-specific
10241 built-in macros and others, such as __FAST_MATH__, had to be
10242 handled with so-called specs that were hard to maintain. Often they
10243 would fail to behave properly when conflicting options were
10244 supplied on the command line, and define macros in the user's
10245 namespace even when strict ISO compliance was requested.
10246 Integrating the preprocessor has cleanly solved these issues.
10247 * The Makefile suite now supports redirection of make install by
10248 means of the variable DESTDIR.
10249 __________________________________________________________________
10253 Detailed release notes for the GCC 3.3 release follow.
10259 * [8]10140 cross compiler build failures: missing __mempcpy (DUP:
10260 [9]10198,[10]10338)
10262 Internal compiler errors (multi-platform)
10264 * [11]3581 large string causes segmentation fault in cc1
10265 * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler
10266 * [13]5533 (c++) ICE when processing std::accumulate(begin, end,
10268 * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out
10269 * [15]6412 (c++) ICE in retrieve_specialization
10270 * [16]6620 (c++) partial template specialization causes an ICE
10271 (segmentation fault)
10272 * [17]6663 (c++) ICE with attribute aligned
10273 * [18]7068 ICE with incomplete types
10274 * [19]7083 (c++) ICE using -gstabs with dodgy class derivation
10275 * [20]7647 (c++) ICE when data member has the name of the enclosing
10277 * [21]7675 ICE in fixup_var_refs_1
10278 * [22]7718 'complex' template instantiation causes ICE
10279 * [23]8116 (c++) ICE in member template function
10280 * [24]8358 (ada) Ada compiler accesses freed memory, crashes
10281 * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation
10283 * [26]8564 (c++) ICE in find_function_data, in function.c
10284 * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c
10285 * [28]8766 (c++) ICE after failed initialization of static template
10287 * [29]8803 ICE in instantiate_virtual_regs_1, in function.c
10288 * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set
10289 * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class
10291 * [32]9216 (c++) ICE on missing template parameter
10292 * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c
10293 * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO
10295 * [35]9429 (c++) ICE in template instantiation with a pointered new
10297 * [36]9516 Internal error when using a big array
10298 * [37]9600 (c++) ICE with typedefs in template class
10299 * [38]9629 (c++) virtual inheritance segfault
10300 * [39]9672 (c++) ICE: Error reporting routines re-entered
10301 * [40]9749 (c++) ICE in write_expression on invalid function
10303 * [41]9794 (fortran) ICE: floating point exception during constant
10305 * [42]9829 (c++) Missing colon in nested namespace usage causes ICE
10306 * [43]9916 (c++) ICE with noreturn function in ?: statement
10307 * [44]9936 ICE with local function and variable-length 2d array
10308 * [45]10262 (c++) cc1plus crashes with large generated code
10309 * [46]10278 (c++) ICE in parser for invalid code
10310 * [47]10446 (c++) ICE on definition of nonexistent member function of
10311 nested class in a class template
10312 * [48]10451 (c++) ICE in grokdeclarator on spurious mutable
10314 * [49]10506 (c++) ICE in build_new at cp/init.c with
10315 -fkeep-inline-functions and multiple inheritance
10316 * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the
10317 precision of the declared type
10321 * [51]2001 Inordinately long compile times in reload CSE regs
10322 * [52]2391 Exponential compilation time explosion in combine
10323 * [53]2960 Duplicate loop conditions even with -Os
10324 * [54]4046 redundant conditional branch
10325 * [55]6405 Loop-unrolling related performance regressions
10326 * [56]6798 very long compile time with large case-statement
10327 * [57]6871 const objects shouldn't be moved to .bss
10328 * [58]6909 problem w/ -Os on modified loop-2c.c test case
10329 * [59]7189 gcc -O2 -Wall does not print ``control reaches end of
10330 non-void function'' warning
10331 * [60]7642 optimization problem with signbit()
10332 * [61]8634 incorrect code for inlining of memcpy under -O2
10333 * [62]8750 Cygwin prolog generation erroneously emitting __alloca as
10334 regular function call
10338 * [63]2161 long if-else cascade overflows parser stack
10339 * [64]4319 short accepted on typedef'd char
10340 * [65]8602 incorrect line numbers in warning messages when using
10342 * [66]9177 -fdump-translation-unit: C front end deletes function_decl
10343 AST nodes and breaks debugging dumps
10344 * [67]9853 miscompilation of non-constant structure initializer
10346 c++ compiler and library
10348 * [68]45 legal template specialization code is rejected (DUP:
10350 * [70]764 lookup failure: friend operator and dereferencing a pointer
10351 and templates (DUP: [71]5116)
10352 * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP:
10354 * [73]3663 G++ doesn't check access control during template
10356 * [74]3797 gcc fails to emit explicit specialization of a template
10358 * [75]3948 Two destructors are called when no copy destructor is
10359 defined (ABI change)
10360 * [76]4137 Conversion operator within template is not accepted
10361 * [77]4361 bogus ambiguity taking the address of a member template
10362 * [78]4802 g++ accepts illegal template code (access to private
10363 member; DUP: [79]5837)
10364 * [80]4803 inline function is used but never defined, and g++ does
10366 * [81]5094 Partial specialization cannot be friend?
10367 * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66
10368 * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run
10370 * [84]7015 certain __asm__ constructs rejected
10371 * [85]7086 compile time regression (quadratic behavior in
10373 * [86]7099 G++ doesn't set the noreturn attribute on std::exit and
10375 * [87]7247 copy constructor missing when inlining enabled (invalid
10377 * [88]7441 string array initialization compilation time regression
10378 from seconds to minutes
10379 * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong
10380 * [90]7804 bad printing of floating point constant in warning message
10381 * [91]8099 Friend classes and template specializations
10382 * [92]8117 member function pointers and multiple inheritance
10383 * [93]8205 using declaration and multiple inheritance
10384 * [94]8645 unnecessary non-zero checks in stl_tree.h
10385 * [95]8724 explicit destructor call for incomplete class allowed
10386 * [96]8805 compile time regression with many member variables
10387 * [97]8691 -O3 and -fno-implicit-templates are incompatible
10388 * [98]8700 unhelpful error message for binding temp to reference
10389 * [99]8724 explicit destructor call for incomplete class allowed
10390 * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems
10391 * [101]9016 Failure to consistently constant fold "constant" C++
10393 * [102]9053 g++ confused about ambiguity of overloaded function
10395 * [103]9152 undefined virtual thunks
10396 * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out
10397 * [105]9297 data corruption due to codegen bug (when copying.)
10398 * [106]9318 i/ostream::operator>>/<<(streambuf*) broken
10399 * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf
10400 * [108]9400 bogus -Wshadow warning: shadowed declaration of this in
10402 * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters
10403 * [110]9425 filebuf::pbackfail broken (DUP: [111]9439)
10404 * [112]9474 GCC freezes in compiling a weird code mixing <iostream>
10406 * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1)
10408 * [115]9555 ostream inserters fail to set badbit on exception
10409 * [116]9561 ostream inserters rethrow exception of wrong type
10410 * [117]9563 ostream::sentry returns true after a failed preparation
10411 * [118]9582 one-definition rule violation in std::allocator
10412 * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors
10413 * [120]9683 bug in initialization chains for static const variables
10414 from template classes
10415 * [121]9791 -Woverloaded-virtual reports hiding of destructor
10416 * [122]9817 collate::compare doesn't handle nul characters
10417 * [123]9825 filebuf::sputbackc breaks sbumpc
10418 * [124]9826 operator>>(basic_istream, basic_string) fails to compile
10420 * [125]9924 Multiple using statements for builtin functions not
10422 * [126]9946 destructor is not called for temporary object
10423 * [127]9964 filebuf::close() sometimes fails to close file
10424 * [128]9988 filebuf::overflow writes EOF to file
10425 * [129]10033 optimization breaks polymorphic references w/ typeid
10427 * [130]10097 filebuf::underflow drops characters
10428 * [131]10132 filebuf destructor can throw exceptions
10429 * [132]10180 gcc fails to warn about non-inlined function
10430 * [133]10199 method parametrized by template does not work everywhere
10431 * [134]10300 use of array-new (nothrow) in segfaults on NULL return
10432 * [135]10427 Stack corruption with variable-length automatic arrays
10433 and virtual destructors
10434 * [136]10503 Compilation never stops in fixed_type_or_null
10438 * [137]5956 selectors aren't matched properly when added to the
10441 Fortran compiler and library
10443 * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't
10445 * [139]3924 g77 generates code that is rejected by GAS if COFF debug
10447 * [140]5634 doc: explain that configure --prefix=~/... does not work
10448 * [141]6367 multiple repeat counts confuse namelist read into array
10449 * [142]6491 Logical operations error on logicals when using
10451 * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C"
10452 * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os
10454 * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should
10455 assume a direct access file
10456 * [146]7278 g77 "bug"; the executable misbehaves (with -O2
10458 * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows
10459 * [148]7388 Incorrect output with 0-based array of characters
10460 * [149]8587 Double complex zero ** double precision number -> NaN
10462 * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning:
10463 unknown register name line-length-none
10464 * [151]10197 Direct access files not unformatted by default
10466 Java compiler and library
10468 * [152]6005 gcj fails to build rhug on alpha
10469 * [153]6389 System.getProperty("") should always throw an
10470 IllegalArgumentException
10471 * [154]6576 java.util.ResourceBundle.getResource ignores locale
10472 * [155]6652 new java.io.File("").getCanonicalFile() throws exception
10473 * [156]7060 getMethod() doesn't search super interface
10474 * [157]7073 bytecode interpreter gives wrong answer for interface
10476 * [158]7180 possible bug in
10477 javax.naming.spi.NamingManager.getPlusPath()
10478 * [159]7416 java.security startup refs "GNU libgcj.security"
10479 * [160]7570 Runtime.exec with null envp: child doesn't inherit parent
10480 env (DUP: [161]7578)
10481 * [162]7611 Internal error while compiling libjava with -O
10482 * [163]7709 NullPointerException in _Jv_ResolvePoolEntry
10483 * [164]7766 ZipInputStream.available returns 0 immediately after
10485 * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public
10486 * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented
10487 * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens'
10488 * [168]8234 ZipInputStream chokes when InputStream.read() returns
10490 * [169]8415 reflection bug: exception info for Method
10491 * [170]8481 java.Random.nextInt(int) may return negative
10492 * [171]8593 Error reading GZIPped files with BufferedReader
10493 * [172]8759 java.beans.Introspector has no flushCaches() or
10494 flushFromCaches() methods
10495 * [173]8997 spin() calls Thread.sleep
10496 * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd
10497 instead of the root content of C:
10498 * [175]9254 java::lang::Object::wait(), threads-win32.cc returns
10500 * [176]9271 Severe bias in java.security.SecureRandom
10502 Ada compiler and library
10504 * [177]6767 make gnatlib-shared fails on -laddr2line
10505 * [178]9911 gnatmake fails to link when GCC configured with
10506 --with-sjlj-exceptions=yes
10507 * [179]10020 Can't bootstrap gcc on AIX with Ada enabled
10508 * [180]10546 Ada tasking not working on Red Hat 9
10512 * [181]7029 preprocessor should ignore #warning with -M
10516 * [182]2903 [arm] Optimization bug with long long arithmetic
10517 * [183]7873 arm-linux-gcc fails when assigning address to a bit field
10521 * [184]7680 float functions undefined in math.h/cmath with #define
10524 HP-UX or HP-PA-specific
10526 * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c
10527 * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to
10529 * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen
10533 * [188]6744 Bad assembler code generated: reference to pseudo
10535 * [189]7361 Internal compiler error in reload_cse_simplify_operands,
10540 * [190]9496 [mips-linux] bug in optimizer?
10544 * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of
10546 * [192]8480 reload ICEs for LAPACK code on powerpc64-linux
10547 * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg
10548 * [194]10315 [powerpc] ICE: in extract_insn, in recog.c
10552 * [195]10267 (documentation) Wrong build instructions for
10555 x86-specific (Intel/AMD)
10557 * [196]7916 ICE in instantiate_virtual_register_1
10558 * [197]7926 (c++) i486 instructions in header files make c++ programs
10560 * [198]8555 ICE in gen_split_1231
10561 * [199]8994 ICE with -O -march=pentium4
10562 * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs
10563 * [201]9806 ICE in inline assembly with -fPIC flag
10564 * [202]10077 gcc -msse2 generates movd to move dwords between xmm
10566 * [203]10233 64-bit comparison only comparing bottom 32-bits
10567 * [204]10286 type-punning doesn't work with __m64 and -O
10568 * [205]10308 [x86] ICE with -O -fgcse or -O2
10569 __________________________________________________________________
10575 This section lists the problem reports (PRs) from GCC's bug tracking
10576 system that are known to be fixed in the 3.3.1 release. This list might
10577 not be complete (that is, it is possible that some PRs that have been
10578 fixed are not listed here).
10582 * [206]11272 [Solaris] make bootstrap fails while building libstdc++
10584 Internal compiler errors (multi-platform)
10586 * [207]5754 ICE on invalid nested template class
10587 * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64
10588 and --enable-checking
10589 * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c
10590 * [210]7053 (c++) ICE when declaring a function already defined as a
10591 friend method of a template class
10592 * [211]8164 (c++) ICE when using different const expressions as
10594 * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c
10595 * [213]9559 (c++) ICE with invalid initialization of a static const
10596 * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c
10597 when redeclaring a static member variable
10598 * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in
10599 dwarfout.c with -g -O -finline-functions
10600 * [216]10432 (c++) ICE in poplevel, in cp/decl.c
10601 * [217]10475 ICE in subreg_highpart_offset for code with long long
10602 * [218]10635 (c++) ICE when dereferencing an incomplete type casted
10603 from a void pointer
10604 * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while
10605 instantiating static member variables
10606 * [220]10700 ICE in copy_to_mode_reg on 64-bit targets
10607 * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c
10608 * [222]10796 (c++) ICE when defining an enum with two values: -1 and
10610 * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x
10612 * [224]10939 (c++) ICE with template code
10613 * [225]10956 (c++) ICE when specializing a template member function
10614 of a template class, in tsubst, in cp/pt.c
10615 * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*()
10617 * [227]11059 (c++) ICE with empty union
10618 * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with
10619 -O2 -fnon-call-exceptions
10620 * [229]11105 (c++) ICE in mangle_conv_op_name_for_type
10621 * [230]11149 (c++) ICE on error when instantiation with call function
10623 * [231]11228 (c++) ICE on new-expression using array operator new and
10624 default-initialization
10625 * [232]11282 (c++) Infinite memory usage after syntax error
10626 * [233]11301 (fortran) ICE with -fno-globals
10627 * [234]11308 (c++) ICE when using an enum type name as if it were a
10629 * [235]11473 (c++) ICE with -gstabs when empty struct inherits from
10631 * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR
10632 * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c:
10633 template member functions
10637 * [238]11198 -O2 -frename-registers generates wrong code (aliasing
10639 * [239]11304 Wrong code production with -fomit-frame-pointer
10640 * [240]11381 volatile memory access optimized away
10641 * [241]11536 [strength-reduce] -O2 optimization produces wrong code
10642 * [242]11557 constant folding bug generates wrong code
10646 * [243]5897 No warning for statement after return
10647 * [244]11279 DWARF-2 output mishandles large enums
10651 * [245]11022 no warning for non-compatible macro redefinition
10653 C++ compiler and library
10655 * [246]2330 static_cast<>() to a private base is allowed
10656 * [247]5388 Incorrect message "operands to ?: have different types"
10657 * [248]5390 Libiberty fails to demangle multi-digit template
10659 * [249]7877 Incorrect parameter passing to specializations of member
10661 * [250]9393 Anonymous namespaces and compiling the same file twice
10662 * [251]10032 -pedantic converts some errors to warnings
10663 * [252]10468 const typeof(x) is non-const, but only in templates
10664 * [253]10527 confused error message with "new int()" parameter
10666 * [254]10679 parameter MIN_INLINE_INSNS is not honored
10667 * [255]10682 gcc chokes on a typedef for an enum inside a class
10669 * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of
10671 * [257]10845 template member function (with nested template as
10672 parameter) cannot be called anymore if another unrelated template
10673 member function is defined
10674 * [258]10849 Cannot define an out-of-class specialization of a
10675 private nested template class
10676 * [259]10888 Suppress -Winline warnings for system headers
10677 * [260]10929 -Winline warns about functions for which no definition
10679 * [261]10931 valid conversion static_cast<const unsigned
10680 int&>(lvalue-of-type-int) is rejected
10681 * [262]10940 Bad code with explicit specialization
10682 * [263]10968 If member function implicitly instantiated, explicit
10683 instantiation of class fails to instantiate it
10684 * [264]10990 Cannot convert with dynamic_cast<> to a private base
10685 class from within a member function
10686 * [265]11039 Bad interaction between implicit typename deprecation
10688 * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say
10689 "__unused__" instead
10690 * [267]11095 C++ iostream manipulator causes segfault when called
10691 with negative argument
10692 * [268]11098 g++ doesn't emit complete debugging information for
10693 local variables in destructors
10694 * [269]11137 GNU/Linux shared library constructors not called unless
10695 there's one global object
10696 * [270]11154 spurious ambiguity report for template class
10698 * [271]11329 Compiler cannot find user defined implicit typecast
10699 * [272]11332 Spurious error with casts in ?: expression
10700 * [273]11431 static_cast behavior with subclasses when default
10701 constructor available
10702 * [274]11528 money_get facet does not accept "$.00" as valid
10703 * [275]11546 Type lookup problems in out-of-line definition of a
10704 class doubly nested from a template class
10705 * [276]11567 C++ code containing templated member function with same
10706 name as pure virtual member function results in linking failure
10707 * [277]11645 Failure to deal with using and private inheritance
10709 Java compiler and library
10711 * [278]5179 Qualified static field access doesn't initialize its
10713 * [279]8204 gcj -O2 to native reorders certain instructions
10715 * [280]10838 java.io.ObjectInputStream syntax error
10716 * [281]10886 The RMI registry that comes with GCJ does not work
10718 * [282]11349 JNDI URL context factories not located correctly
10720 x86-specific (Intel/AMD)
10722 * [283]4823 ICE on inline assembly code
10723 * [284]8878 miscompilation with -O and SSE
10724 * [285]9815 (c++ library) atomicity.h - fails to compile with -O3
10726 * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads,
10728 * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2
10729 * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source
10730 * [289]11044 [x86] out of range loop instructions for FP code on K6
10731 * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE
10733 * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC"
10736 SPARC- or Solaris- specific
10738 * [292]9362 solaris 'as' dies when fed .s and "-gstabs"
10739 * [293]10142 [SPARC64] gcc produces wrong code when passing
10740 structures by value
10741 * [294]10663 New configure check aborts with Sun tools.
10742 * [295]10835 combinatorial explosion in scheduler on HyperSPARC
10743 * [296]10876 ICE in calculate_giv_inc when building KDE
10744 * [297]10955 wrong code at -O3 for structure argument in context of
10746 * [298]11018 -mcpu=ultrasparc busts tar-1.13.25
10747 * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x
10752 * [300]10907 gcc violates the ia64 ABI (GP must be preserved)
10753 * [301]11320 scheduler bug (in machine depended reorganization pass)
10754 * [302]11599 bug with conditional and __builtin_prefetch
10758 * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem
10760 * [304]10871 error in rs6000_stack_info save_size computation
10761 * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse
10766 * [306]7594 [m68k] ICE on legal code associated with simplify-rtx
10767 * [307]10557 [m68k] ICE in subreg_offset_representable_p
10768 * [308]11054 [m68k] ICE in reg_overlap_mentioned_p
10772 * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for
10773 functions with __attribute__ ((interrupt ("IRQ")))
10774 * [310]10842 [arm] Clobbered link register is copied to pc under
10775 certain circumstances
10776 * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes
10777 * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno
10782 * [313]11084 ICE in propagate_one_insn, in flow.c
10786 * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf
10787 * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c
10788 * [316]11096 i686-linux to sh-linux cross compiler fails to compile
10791 GNU/Linux (or Hurd?) specific
10793 * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3
10797 * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare
10800 Cygwin (or mingw) specific
10802 * [319]5287 ICE with dllimport attribute
10803 * [320]10148 [MingW/CygWin] Compiler dumps core
10807 * [321]8787 GCC fails to emit .intel_syntax when invoked with
10808 -masm=intel on DJGPP
10810 Darwin (and MacOS X) specific
10812 * [322]10900 trampolines crash
10816 * [323]1607 (c++) Format attributes on methods undocumented
10817 * [324]4252 Invalid option `-fdump-translation-unit'
10818 * [325]4490 Clarify restrictions on -m96bit-long-double,
10819 -m128bit-long-double
10820 * [326]10355 document an issue with regparm attribute on some systems
10822 * [327]10726 (fortran) Documentation for function "IDate Intrinsic
10824 * [328]10805 document bug in old version of Sun assembler
10825 * [329]10815 warn against GNU binutils on AIX
10826 * [330]10877 document need for newer binutils on i?86-*-linux-gnu
10827 * [331]11280 Manual incorrect with respect to -freorder-blocks
10828 * [332]11466 Document -mlittle-endian and its restrictions for the
10831 Testsuite bugs (compiler itself is not affected)
10833 * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly
10835 * [334]10810 gcc-3.3 fails make check: buffer overrun in
10837 __________________________________________________________________
10843 This section lists the problem reports (PRs) from GCC's bug tracker
10844 that are known to be fixed in the 3.3.2 release. This list might not be
10845 complete (that is, it is possible that some PRs that have been fixed
10846 are not listed here).
10848 Bootstrap failures and problems
10850 * [335]8336 [SCO5] bootstrap config still tries to use COFF options
10851 * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with
10852 --enable-threads=posix
10853 * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap
10854 * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare
10856 * [339]11687 xstormy16-elf build fails in libf2c
10857 * [340]12263 [SGI IRIX] bootstrap fails during compile of
10858 libf2c/libI77/backspace.c
10859 * [341]12490 buffer overflow in scan-decls.c (during Solaris 9
10860 fix-header processing)
10862 Internal compiler errors (multi-platform)
10864 * [342]7277 Casting integers to vector types causes ICE
10865 * [343]7939 (c++) ICE on invalid function template specialization
10866 * [344]11063 (c++) ICE on parsing initialization list of const array
10868 * [345]11207 ICE with negative index in array element designator
10869 * [346]11522 (fortran) g77 dwarf-2 ICE in
10870 add_abstract_origin_attribute
10871 * [347]11595 (c++) ICE on duplicate label definition
10872 * [348]11646 (c++) ICE in commit_one_edge_insertion with
10873 -fnon-call-exceptions -fgcse -O
10874 * [349]11665 ICE in struct initializer when taking address
10875 * [350]11852 (c++) ICE with bad struct initializer.
10876 * [351]11878 (c++) ICE in cp_expr_size
10877 * [352]11883 ICE with any -O on mercury-generated C code
10878 * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in
10879 cp/typeck2.c when applying typeid operator to template template
10881 * [354]12146 ICE in lookup_template_function, in cp/pt.c
10882 * [355]12215 ICE in make_label_edge with -fnon-call-exceptions
10884 * [356]12369 (c++) ICE with templates and friends
10885 * [357]12446 ICE in emit_move_insn on complicated array reference
10886 * [358]12510 ICE in final_scan_insn
10887 * [359]12544 ICE with large parameters used in nested functions
10889 C and optimization bugs
10891 * [360]9862 spurious warnings with -W -finline-functions
10892 * [361]10962 lookup_field is a linear search on a linked list (can be
10893 slow if large struct)
10894 * [362]11370 -Wunreachable-code gives false complaints
10895 * [363]11637 invalid assembly with -fnon-call-exceptions
10896 * [364]11885 Problem with bitfields in packed structs
10897 * [365]12082 Inappropriate unreachable code warnings
10898 * [366]12180 Inline optimization fails for variadic function
10899 * [367]12340 loop unroller + gcse produces wrong code
10901 C++ compiler and library
10903 * [368]3907 nested template parameter collides with member name
10904 * [369]5293 confusing message when binding a temporary to a reference
10905 * [370]5296 [DR115] Pointers to functions and to template functions
10906 behave differently in deduction
10907 * [371]7939 ICE on function template specialization
10908 * [372]8656 Unable to assign function with __attribute__ and pointer
10909 return type to an appropriate variable
10910 * [373]10147 Confusing error message for invalid template function
10912 * [374]11400 std::search_n() makes assumptions about Size parameter
10913 * [375]11409 issues with using declarations, overloading, and
10915 * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle
10916 multiple bits in mask
10917 * [377]11786 operator() call on variable in other namespace not
10919 * [378]11867 static_cast ignores ambiguity
10920 * [379]11928 bug with conversion operators that are typedefs
10921 * [380]12114 Uninitialized memory accessed in dtor
10922 * [381]12163 static_cast + explicit constructor regression
10923 * [382]12181 Wrong code with comma operator and c++
10924 * [383]12236 regparm and fastcall messes up parameters
10925 * [384]12266 incorrect instantiation of unneeded template during
10926 overload resolution
10927 * [385]12296 istream::peek() doesn't set eofbit
10928 * [386]12298 [sjlj exceptions] Stack unwind destroys
10929 not-yet-constructed object
10930 * [387]12369 ICE with templates and friends
10931 * [388]12337 apparently infinite loop in g++
10932 * [389]12344 stdcall attribute ignored if function returns a pointer
10933 * [390]12451 missing(late) class forward declaration in cxxabi.h
10934 * [391]12486 g++ accepts invalid use of a qualified name
10936 x86 specific (Intel/AMD)
10938 * [392]8869 [x86 MMX] ICE with const variable optimization and MMX
10940 * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions
10942 * [394]11689 g++3.3 emits un-assembleable code for k6 architecture
10943 * [395]12116 [k6] Invalid assembly output values with X-MAME code
10944 * [396]12070 ICE converting between double and long double with
10949 * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc
10950 * [398]11535 __builtin_return_address may not work on ia64
10951 * [399]11693 [ia64] ICE in gen_nop_type
10952 * [400]12224 [ia64] Thread-local storage doesn't work
10956 * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux
10958 * [402]11319 loop miscompiled on ppc32
10959 * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code
10963 * [404]11662 wrong code for expr. with cast to long long and
10965 * [405]11965 invalid assembler code for a shift < 32 operation
10966 * [406]12301 (c++) stack corruption when a returned expression throws
10971 * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of
10976 * [408]11313 problem with #pragma weak and static inline functions
10977 * [409]11712 __STDC_EXT__ not defined for C++ by default anymore?
10981 * [410]12166 Profiled programs crash if PROFDIR is set
10983 Solaris-x86 specific
10985 * [411]12101 i386 Solaris no longer works with GNU as?
10987 Miscellaneous embedded target-specific bugs
10989 * [412]10988 [m32r-elf] wrong blockmove code with -O3
10990 * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with
10992 * [414]11902 [sh4] spec file improperly inserts rpath even when none
10994 * [415]11903 [sh4] -pthread fails to link due to error in spec file
10996 __________________________________________________________________
11002 In addition to the bug fixes documented below, this release contains
11003 few minor features such as:
11004 * Support for --with-sysroot
11005 * Support for automatic detection of executable stacks
11006 * Support for SSE3 instructions
11007 * Support for thread local storage debugging under GDB on S390
11011 This section lists the problem reports (PRs) from GCC's bug tracker
11012 that are known to be fixed in the 3.3.3 release. This list might not be
11013 complete (that is, it is possible that some PRs that have been fixed
11014 are not listed here).
11016 Bootstrap failures and issues
11018 * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails
11019 * [417]12399 boehm-gc fails (when building a cross compiler): libtool
11020 unable to infer tagged configuration
11021 * [418]13068 mklibgcc.in doesn't handle multi-level multilib
11022 subdirectories properly
11024 Internal compiler errors (multi-platform)
11026 * [419]10060 ICE (stack overflow) on huge file (300k lines) due to
11027 recursive behaviour of copy_rtx_if_shared, in emit_rtl.c
11028 * [420]10555 (c++) ICE on undefined template argument
11029 * [421]10706 (c++) ICE in mangle_class_name_for_template
11030 * [422]11496 (fortran) error in flow_loops_find when -funroll-loops
11032 * [423]11741 ICE in pre_insert_copy_insn, in gcse.c
11033 * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0
11034 * [425]12632 (fortran) -fbounds-check ICE
11035 * [426]12712 (c++) ICE on short legit C++ code fragment with gcc
11037 * [427]12726 (c++) ICE (segfault) on trivial code
11038 * [428]12890 (c++) ICE on compilation of class with throwing method
11039 * [429]12900 (c++) ICE in rtl_verify_flow_info_1
11040 * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on
11041 correct code with -O2 -fno-force-mem
11042 * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive
11044 * [432]13318 ICE: floating point exception in the loop optimizer
11045 * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in
11047 * [434]13574 (c++) invalid array default initializer in class lets
11048 gcc consume all memory and die
11049 * [435]13475 ICE on SIMD variables with partial value initialization
11050 * [436]13797 (c++) ICE on invalid template parameter
11051 * [437]13824 (java) gcj SEGV with simple .java program
11053 C and optimization bugs
11055 * [438]8776 loop invariants are not removed (most likely)
11056 * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing
11058 * [440]11350 undefined labels with -Os -fPIC
11059 * [441]12826 Optimizer removes reference through volatile pointer
11060 * [442]12500 stabs debug info: void no longer a predefined / builtin
11062 * [443]12941 builtin-bitops-1.c miscompilation (latent bug)
11063 * [444]12953 tree inliner bug (in inline_forbidden_p) and fix
11064 * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled
11065 * [446]13507 spurious printf format warning
11066 * [447]13382 Type information for const pointer disappears during
11068 * [448]13394 noreturn attribute ignored on recursive invokation
11069 * [449]13400 Compiled code crashes storing to read-only location
11070 * [450]13521 Endless loop in calculate_global_regs_live
11072 C++ compiler and library
11074 Some of the bug fixes in this list were made to implement decisions
11075 that the ISO C++ standards committee has made concerning several defect
11076 reports (DRs). Links in the list below point to detailed discussion of
11077 the relevant defect report.
11078 * [451]2094 unimplemented: use of `ptrmem_cst' in template type
11080 * [452]2294 using declaration confusion
11081 * [453]5050 template instantiation depth exceeds limit: recursion
11083 * [454]9371 Bad exception handling in
11084 i/ostream::operator>>/<<(streambuf*)
11085 * [455]9546 bad exception handling in ostream members
11086 * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the
11087 face of unknown locales
11088 * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work
11089 * [459]10095 istream::operator>>(int&) sets ios::badbit when
11090 ios::failbit is set.
11091 * [460]11554 Warning about reordering of initializers doesn't mention
11092 location of constructor
11093 * [461]12297 istream::sentry::sentry() handles eof() incorrectly.
11094 * [462]12352 Exception safety problems in src/localename.cc
11095 * [463]12438 Memory leak in locale::combine()
11096 * [464]12540 Memory leak in locale::locale(const char*)
11097 * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented
11098 * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented
11099 * [470]12696 memory eating infinite loop in diagnostics (error
11101 * [471]12815 Code compiled with optimization behaves unexpectedly
11102 * [472]12862 Conflicts between typedefs/enums and namespace member
11104 * [473]12926 Wrong value after assignment in initialize list using
11106 * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented
11107 * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented
11108 * [478]13007 basic_streambuf::pubimbue, imbue wrong
11109 * [479]13009 Implicitly-defined assignment operator writes to wrong
11111 * [480]13057 regparm attribute not applied to destructor
11112 * [481]13070 -Wformat option ignored in g++
11113 * [482]13081 forward template declarations in <complex> let inlining
11115 * [483]13239 Assertion does not seem to work correctly anymore
11116 * [484]13262 "xxx is private within this context" when initializing a
11117 self-contained template class
11118 * [485]13290 simple typo in concept checking for std::generate_n
11119 * [486]13323 Template code does not compile in presence of typedef
11120 * [487]13369 __verify_grouping (and __add_grouping?) not correct
11121 * [488]13371 infinite loop with packed struct and inlining
11122 * [489]13445 Template argument replacement "dereferences" a typedef
11123 * [490]13461 Fails to access protected-ctor from public constant
11124 * [491]13462 Non-standard-conforming type set::pointer
11125 * [492]13478 gcc uses wrong constructor to initialize a const
11127 * [493]13544 "conflicting types" for enums in different scopes
11128 * [494]13650 string::compare should not (always) use
11129 traits_type::length()
11130 * [495]13683 bogus warning about passing non-PODs through ellipsis
11131 * [496]13688 Derived class is denied access to protected base class
11133 * [497]13774 Member variable cleared in virtual multiple inheritance
11135 * [498]13884 Protect sstream.tcc from extern template use
11137 Java compiler and library
11139 * [499]10746 [win32] garbage collection crash in GCJ
11141 Objective-C compiler and library
11143 * [500]11433 Crash due to dereferencing null pointer when querying
11146 Fortran compiler and library
11148 * [501]12633 logical expression gives incorrect result with
11149 -fugly-logint option
11150 * [502]13037 [gcse-lm] g77 generates incorrect code
11151 * [503]13213 Hex constant problem when compiling with -fugly-logint
11154 x86-specific (Intel/AMD)
11156 * [504]4490 ICE with -m128bit-long-double
11157 * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have
11158 `reg' in make_field_assignment, in combine.c
11159 * [506]12441 ICE: can't find a register to spill
11160 * [507]12943 array static-init failure under -fpic, -fPIC
11161 * [508]13608 Incorrect code with -O3 -ffast-math
11165 * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of
11166 __attribute__((aligned(16)))
11167 * [510]11793 ICE in extract_insn, in recog.c (const_vector's)
11168 * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in
11170 * [512]12537 g++ generates writeable text sections
11174 * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0
11176 * [514]12865 mprotect call to make trampoline executable may fail
11177 * [515]13354 ICE in sparc_emit_set_const32
11181 * [516]10467 [arm] ICE in pre_insert_copy_insn,
11185 * [517]11226 ICE passing struct arg with two floats
11186 * [518]11227 ICE for _Complex float, _Complex long double args
11187 * [519]12644 GCC 3.3.2 fails to compile glibc on ia64
11188 * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn
11189 * Various fixes for libunwind
11193 * [521]12654 Incorrect comparison code generated for Alpha
11194 * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2
11195 * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2
11199 * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c
11200 * [525]12158 [hppa] compilation does not terminate at -O1
11204 * [526]11992 Wrong built-in code for memcmp with length 1<<24: only
11205 (1<<24)-1 possible for CLCL-Instruction
11209 * [527]9365 segfault in gen_far_branch (config/sh/sh.c)
11210 * [528]10392 optimizer generates faulty array indexing
11211 * [529]11322 SH profiler outputs multiple definitions of symbol
11212 * [530]13069 gcc/config/sh/rtems.h broken
11213 * [531]13302 Putting a va_list in a struct causes seg fault
11214 * [532]13585 Incorrect optimization of call to sfunc
11215 * Fix inappropriately exported libgcc functions from the shared
11218 Other embedded target specific
11220 * [533]8916 [mcore] unsigned char assign gets hosed.
11221 * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c
11222 * [535]13122 [h8300] local variable gets corrupted by function call
11223 when -fomit-frame-pointer is given
11224 * [536]13256 [cris] strict_low_part mistreated in delay slots
11225 * [537]13373 [mcore] optimization with -frerun-cse-after-loop
11226 -fexpensive-optimizations produces wrong code on mcore
11230 * [538]12561 gcc/config/t-gnu needs updating to work with
11233 Tru64 Unix specific
11235 * [539]6243 testsuite fails almost all tests due to no libintl in
11236 LD_LIBRARY_PATH during test.
11237 * [540]11397 weak aliases broken on Tru64 UNIX
11241 * [541]12505 build failure due to defines of uchar in cpphash.h and
11243 * [542]13150 WEAK symbols not exported by collect2
11247 * [543]12666 fixincludes problem on IRIX 6.5.19m
11251 * [544]12969 Including sys/byteorder.h breaks configure checks
11253 Testsuite problems (compiler is not affected)
11255 * [545]10819 testsuite creates CR+LF on compiler version lines in
11257 * [546]11612 abi_check not finding correct libgcc_s.so.1
11261 * [547]13211 using -###, incorrect warnings about unused linker file
11263 __________________________________________________________________
11267 This is the [548]list of problem reports (PRs) from GCC's bug tracking
11268 system that are known to be fixed in the 3.3.4 release. This list might
11269 not be complete (that is, it is possible that some PRs that have been
11270 fixed are not listed here).
11271 __________________________________________________________________
11275 This is the [549]list of problem reports (PRs) from GCC's bug tracking
11276 system that are known to be fixed in the 3.3.5 release. This list might
11277 not be complete (that is, it is possible that some PRs that have been
11278 fixed are not listed here).
11279 __________________________________________________________________
11283 This is the [550]list of problem reports (PRs) from GCC's bug tracking
11284 system that are known to be fixed in the 3.3.6 release. This list might
11285 not be complete (that is, it is possible that some PRs that have been
11286 fixed are not listed here).
11289 For questions related to the use of GCC, please consult these web
11290 pages and the [551]GCC manuals. If that fails, the
11291 [552]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11292 web pages and the development of GCC are welcome on our developer
11293 list at [553]gcc@gcc.gnu.org. All of [554]our lists have public
11296 Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and
11297 distribution of this entire article is permitted in any medium,
11298 provided this notice is preserved.
11300 These pages are [556]maintained by the GCC team. Last modified
11305 1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6
11306 2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems
11307 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems
11308 4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute
11309 5. http://gcc.gnu.org/news/dfa.html
11310 6. http://gcc.gnu.org/c99status.html
11311 7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html
11312 8. https://gcc.gnu.org/PR10140
11313 9. https://gcc.gnu.org/PR10198
11314 10. https://gcc.gnu.org/PR10338
11315 11. https://gcc.gnu.org/PR3581
11316 12. https://gcc.gnu.org/PR4382
11317 13. https://gcc.gnu.org/PR5533
11318 14. https://gcc.gnu.org/PR6387
11319 15. https://gcc.gnu.org/PR6412
11320 16. https://gcc.gnu.org/PR6620
11321 17. https://gcc.gnu.org/PR6663
11322 18. https://gcc.gnu.org/PR7068
11323 19. https://gcc.gnu.org/PR7083
11324 20. https://gcc.gnu.org/PR7647
11325 21. https://gcc.gnu.org/PR7675
11326 22. https://gcc.gnu.org/PR7718
11327 23. https://gcc.gnu.org/PR8116
11328 24. https://gcc.gnu.org/PR8358
11329 25. https://gcc.gnu.org/PR8511
11330 26. https://gcc.gnu.org/PR8564
11331 27. https://gcc.gnu.org/PR8660
11332 28. https://gcc.gnu.org/PR8766
11333 29. https://gcc.gnu.org/PR8803
11334 30. https://gcc.gnu.org/PR8846
11335 31. https://gcc.gnu.org/PR8906
11336 32. https://gcc.gnu.org/PR9216
11337 33. https://gcc.gnu.org/PR9261
11338 34. https://gcc.gnu.org/PR9263
11339 35. https://gcc.gnu.org/PR9429
11340 36. https://gcc.gnu.org/PR9516
11341 37. https://gcc.gnu.org/PR9600
11342 38. https://gcc.gnu.org/PR9629
11343 39. https://gcc.gnu.org/PR9672
11344 40. https://gcc.gnu.org/PR9749
11345 41. https://gcc.gnu.org/PR9794
11346 42. https://gcc.gnu.org/PR9829
11347 43. https://gcc.gnu.org/PR9916
11348 44. https://gcc.gnu.org/PR9936
11349 45. https://gcc.gnu.org/PR10262
11350 46. https://gcc.gnu.org/PR10278
11351 47. https://gcc.gnu.org/PR10446
11352 48. https://gcc.gnu.org/PR10451
11353 49. https://gcc.gnu.org/PR10506
11354 50. https://gcc.gnu.org/PR10549
11355 51. https://gcc.gnu.org/PR2001
11356 52. https://gcc.gnu.org/PR2391
11357 53. https://gcc.gnu.org/PR2960
11358 54. https://gcc.gnu.org/PR4046
11359 55. https://gcc.gnu.org/PR6405
11360 56. https://gcc.gnu.org/PR6798
11361 57. https://gcc.gnu.org/PR6871
11362 58. https://gcc.gnu.org/PR6909
11363 59. https://gcc.gnu.org/PR7189
11364 60. https://gcc.gnu.org/PR7642
11365 61. https://gcc.gnu.org/PR8634
11366 62. https://gcc.gnu.org/PR8750
11367 63. https://gcc.gnu.org/PR2161
11368 64. https://gcc.gnu.org/PR4319
11369 65. https://gcc.gnu.org/PR8602
11370 66. https://gcc.gnu.org/PR9177
11371 67. https://gcc.gnu.org/PR9853
11372 68. https://gcc.gnu.org/PR45
11373 69. https://gcc.gnu.org/PR3784
11374 70. https://gcc.gnu.org/PR764
11375 71. https://gcc.gnu.org/PR5116
11376 72. https://gcc.gnu.org/PR2862
11377 73. https://gcc.gnu.org/PR3663
11378 74. https://gcc.gnu.org/PR3797
11379 75. https://gcc.gnu.org/PR3948
11380 76. https://gcc.gnu.org/PR4137
11381 77. https://gcc.gnu.org/PR4361
11382 78. https://gcc.gnu.org/PR4802
11383 79. https://gcc.gnu.org/PR5837
11384 80. https://gcc.gnu.org/PR4803
11385 81. https://gcc.gnu.org/PR5094
11386 82. https://gcc.gnu.org/PR5730
11387 83. https://gcc.gnu.org/PR6713
11388 84. https://gcc.gnu.org/PR7015
11389 85. https://gcc.gnu.org/PR7086
11390 86. https://gcc.gnu.org/PR7099
11391 87. https://gcc.gnu.org/PR7247
11392 88. https://gcc.gnu.org/PR7441
11393 89. https://gcc.gnu.org/PR7768
11394 90. https://gcc.gnu.org/PR7804
11395 91. https://gcc.gnu.org/PR8099
11396 92. https://gcc.gnu.org/PR8117
11397 93. https://gcc.gnu.org/PR8205
11398 94. https://gcc.gnu.org/PR8645
11399 95. https://gcc.gnu.org/PR8724
11400 96. https://gcc.gnu.org/PR8805
11401 97. https://gcc.gnu.org/PR8691
11402 98. https://gcc.gnu.org/PR8700
11403 99. https://gcc.gnu.org/PR8724
11404 100. https://gcc.gnu.org/PR8949
11405 101. https://gcc.gnu.org/PR9016
11406 102. https://gcc.gnu.org/PR9053
11407 103. https://gcc.gnu.org/PR9152
11408 104. https://gcc.gnu.org/PR9182
11409 105. https://gcc.gnu.org/PR9297
11410 106. https://gcc.gnu.org/PR9318
11411 107. https://gcc.gnu.org/PR9320
11412 108. https://gcc.gnu.org/PR9400
11413 109. https://gcc.gnu.org/PR9424
11414 110. https://gcc.gnu.org/PR9425
11415 111. https://gcc.gnu.org/PR9439
11416 112. https://gcc.gnu.org/PR9474
11417 113. https://gcc.gnu.org/PR9548
11418 114. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231
11419 115. https://gcc.gnu.org/PR9555
11420 116. https://gcc.gnu.org/PR9561
11421 117. https://gcc.gnu.org/PR9563
11422 118. https://gcc.gnu.org/PR9582
11423 119. https://gcc.gnu.org/PR9622
11424 120. https://gcc.gnu.org/PR9683
11425 121. https://gcc.gnu.org/PR9791
11426 122. https://gcc.gnu.org/PR9817
11427 123. https://gcc.gnu.org/PR9825
11428 124. https://gcc.gnu.org/PR9826
11429 125. https://gcc.gnu.org/PR9924
11430 126. https://gcc.gnu.org/PR9946
11431 127. https://gcc.gnu.org/PR9964
11432 128. https://gcc.gnu.org/PR9988
11433 129. https://gcc.gnu.org/PR10033
11434 130. https://gcc.gnu.org/PR10097
11435 131. https://gcc.gnu.org/PR10132
11436 132. https://gcc.gnu.org/PR10180
11437 133. https://gcc.gnu.org/PR10199
11438 134. https://gcc.gnu.org/PR10300
11439 135. https://gcc.gnu.org/PR10427
11440 136. https://gcc.gnu.org/PR10503
11441 137. https://gcc.gnu.org/PR5956
11442 138. https://gcc.gnu.org/PR1832
11443 139. https://gcc.gnu.org/PR3924
11444 140. https://gcc.gnu.org/PR5634
11445 141. https://gcc.gnu.org/PR6367
11446 142. https://gcc.gnu.org/PR6491
11447 143. https://gcc.gnu.org/PR6742
11448 144. https://gcc.gnu.org/PR7113
11449 145. https://gcc.gnu.org/PR7236
11450 146. https://gcc.gnu.org/PR7278
11451 147. https://gcc.gnu.org/PR7384
11452 148. https://gcc.gnu.org/PR7388
11453 149. https://gcc.gnu.org/PR8587
11454 150. https://gcc.gnu.org/PR9038
11455 151. https://gcc.gnu.org/PR10197
11456 152. https://gcc.gnu.org/PR6005
11457 153. https://gcc.gnu.org/PR6389
11458 154. https://gcc.gnu.org/PR6576
11459 155. https://gcc.gnu.org/PR6652
11460 156. https://gcc.gnu.org/PR7060
11461 157. https://gcc.gnu.org/PR7073
11462 158. https://gcc.gnu.org/PR7180
11463 159. https://gcc.gnu.org/PR7416
11464 160. https://gcc.gnu.org/PR7570
11465 161. https://gcc.gnu.org/PR7578
11466 162. https://gcc.gnu.org/PR7611
11467 163. https://gcc.gnu.org/PR7709
11468 164. https://gcc.gnu.org/PR7766
11469 165. https://gcc.gnu.org/PR7785
11470 166. https://gcc.gnu.org/PR7786
11471 167. https://gcc.gnu.org/PR8142
11472 168. https://gcc.gnu.org/PR8234
11473 169. https://gcc.gnu.org/PR8415
11474 170. https://gcc.gnu.org/PR8481
11475 171. https://gcc.gnu.org/PR8593
11476 172. https://gcc.gnu.org/PR8759
11477 173. https://gcc.gnu.org/PR8997
11478 174. https://gcc.gnu.org/PR9253
11479 175. https://gcc.gnu.org/PR9254
11480 176. https://gcc.gnu.org/PR9271
11481 177. https://gcc.gnu.org/PR6767
11482 178. https://gcc.gnu.org/PR9911
11483 179. https://gcc.gnu.org/PR10020
11484 180. https://gcc.gnu.org/PR10546
11485 181. https://gcc.gnu.org/PR7029
11486 182. https://gcc.gnu.org/PR2903
11487 183. https://gcc.gnu.org/PR7873
11488 184. https://gcc.gnu.org/PR7680
11489 185. https://gcc.gnu.org/PR8705
11490 186. https://gcc.gnu.org/PR9986
11491 187. https://gcc.gnu.org/PR10056
11492 188. https://gcc.gnu.org/PR6744
11493 189. https://gcc.gnu.org/PR7361
11494 190. https://gcc.gnu.org/PR9496
11495 191. https://gcc.gnu.org/PR7067
11496 192. https://gcc.gnu.org/PR8480
11497 193. https://gcc.gnu.org/PR8784
11498 194. https://gcc.gnu.org/PR10315
11499 195. https://gcc.gnu.org/PR10267
11500 196. https://gcc.gnu.org/PR7916
11501 197. https://gcc.gnu.org/PR7926
11502 198. https://gcc.gnu.org/PR8555
11503 199. https://gcc.gnu.org/PR8994
11504 200. https://gcc.gnu.org/PR9426
11505 201. https://gcc.gnu.org/PR9806
11506 202. https://gcc.gnu.org/PR10077
11507 203. https://gcc.gnu.org/PR10233
11508 204. https://gcc.gnu.org/PR10286
11509 205. https://gcc.gnu.org/PR10308
11510 206. https://gcc.gnu.org/PR11272
11511 207. https://gcc.gnu.org/PR5754
11512 208. https://gcc.gnu.org/PR6597
11513 209. https://gcc.gnu.org/PR6949
11514 210. https://gcc.gnu.org/PR7053
11515 211. https://gcc.gnu.org/PR8164
11516 212. https://gcc.gnu.org/PR8384
11517 213. https://gcc.gnu.org/PR9559
11518 214. https://gcc.gnu.org/PR9649
11519 215. https://gcc.gnu.org/PR9864
11520 216. https://gcc.gnu.org/PR10432
11521 217. https://gcc.gnu.org/PR10475
11522 218. https://gcc.gnu.org/PR10635
11523 219. https://gcc.gnu.org/PR10661
11524 220. https://gcc.gnu.org/PR10700
11525 221. https://gcc.gnu.org/PR10712
11526 222. https://gcc.gnu.org/PR10796
11527 223. https://gcc.gnu.org/PR10890
11528 224. https://gcc.gnu.org/PR10939
11529 225. https://gcc.gnu.org/PR10956
11530 226. https://gcc.gnu.org/PR11041
11531 227. https://gcc.gnu.org/PR11059
11532 228. https://gcc.gnu.org/PR11083
11533 229. https://gcc.gnu.org/PR11105
11534 230. https://gcc.gnu.org/PR11149
11535 231. https://gcc.gnu.org/PR11228
11536 232. https://gcc.gnu.org/PR11282
11537 233. https://gcc.gnu.org/PR11301
11538 234. https://gcc.gnu.org/PR11308
11539 235. https://gcc.gnu.org/PR11473
11540 236. https://gcc.gnu.org/PR11503
11541 237. https://gcc.gnu.org/PR11513
11542 238. https://gcc.gnu.org/PR11198
11543 239. https://gcc.gnu.org/PR11304
11544 240. https://gcc.gnu.org/PR11381
11545 241. https://gcc.gnu.org/PR11536
11546 242. https://gcc.gnu.org/PR11557
11547 243. https://gcc.gnu.org/PR5897
11548 244. https://gcc.gnu.org/PR11279
11549 245. https://gcc.gnu.org/PR11022
11550 246. https://gcc.gnu.org/PR2330
11551 247. https://gcc.gnu.org/PR5388
11552 248. https://gcc.gnu.org/PR5390
11553 249. https://gcc.gnu.org/PR7877
11554 250. https://gcc.gnu.org/PR9393
11555 251. https://gcc.gnu.org/PR10032
11556 252. https://gcc.gnu.org/PR10468
11557 253. https://gcc.gnu.org/PR10527
11558 254. https://gcc.gnu.org/PR10679
11559 255. https://gcc.gnu.org/PR10682
11560 256. https://gcc.gnu.org/PR10689
11561 257. https://gcc.gnu.org/PR10845
11562 258. https://gcc.gnu.org/PR10849
11563 259. https://gcc.gnu.org/PR10888
11564 260. https://gcc.gnu.org/PR10929
11565 261. https://gcc.gnu.org/PR10931
11566 262. https://gcc.gnu.org/PR10940
11567 263. https://gcc.gnu.org/PR10968
11568 264. https://gcc.gnu.org/PR10990
11569 265. https://gcc.gnu.org/PR11039
11570 266. https://gcc.gnu.org/PR11062
11571 267. https://gcc.gnu.org/PR11095
11572 268. https://gcc.gnu.org/PR11098
11573 269. https://gcc.gnu.org/PR11137
11574 270. https://gcc.gnu.org/PR11154
11575 271. https://gcc.gnu.org/PR11329
11576 272. https://gcc.gnu.org/PR11332
11577 273. https://gcc.gnu.org/PR11431
11578 274. https://gcc.gnu.org/PR11528
11579 275. https://gcc.gnu.org/PR11546
11580 276. https://gcc.gnu.org/PR11567
11581 277. https://gcc.gnu.org/PR11645
11582 278. https://gcc.gnu.org/PR5179
11583 279. https://gcc.gnu.org/PR8204
11584 280. https://gcc.gnu.org/PR10838
11585 281. https://gcc.gnu.org/PR10886
11586 282. https://gcc.gnu.org/PR11349
11587 283. https://gcc.gnu.org/PR4823
11588 284. https://gcc.gnu.org/PR8878
11589 285. https://gcc.gnu.org/PR9815
11590 286. https://gcc.gnu.org/PR10402
11591 287. https://gcc.gnu.org/PR10504
11592 288. https://gcc.gnu.org/PR10673
11593 289. https://gcc.gnu.org/PR11044
11594 290. https://gcc.gnu.org/PR11089
11595 291. https://gcc.gnu.org/PR11420
11596 292. https://gcc.gnu.org/PR9362
11597 293. https://gcc.gnu.org/PR10142
11598 294. https://gcc.gnu.org/PR10663
11599 295. https://gcc.gnu.org/PR10835
11600 296. https://gcc.gnu.org/PR10876
11601 297. https://gcc.gnu.org/PR10955
11602 298. https://gcc.gnu.org/PR11018
11603 299. https://gcc.gnu.org/PR11556
11604 300. https://gcc.gnu.org/PR10907
11605 301. https://gcc.gnu.org/PR11320
11606 302. https://gcc.gnu.org/PR11599
11607 303. https://gcc.gnu.org/PR9745
11608 304. https://gcc.gnu.org/PR10871
11609 305. https://gcc.gnu.org/PR11440
11610 306. https://gcc.gnu.org/PR7594
11611 307. https://gcc.gnu.org/PR10557
11612 308. https://gcc.gnu.org/PR11054
11613 309. https://gcc.gnu.org/PR10834
11614 310. https://gcc.gnu.org/PR10842
11615 311. https://gcc.gnu.org/PR11052
11616 312. https://gcc.gnu.org/PR11183
11617 313. https://gcc.gnu.org/PR11084
11618 314. https://gcc.gnu.org/PR10331
11619 315. https://gcc.gnu.org/PR10413
11620 316. https://gcc.gnu.org/PR11096
11621 317. https://gcc.gnu.org/PR2873
11622 318. https://gcc.gnu.org/PR3163
11623 319. https://gcc.gnu.org/PR5287
11624 320. https://gcc.gnu.org/PR10148
11625 321. https://gcc.gnu.org/PR8787
11626 322. https://gcc.gnu.org/PR10900
11627 323. https://gcc.gnu.org/PR1607
11628 324. https://gcc.gnu.org/PR4252
11629 325. https://gcc.gnu.org/PR4490
11630 326. https://gcc.gnu.org/PR10355
11631 327. https://gcc.gnu.org/PR10726
11632 328. https://gcc.gnu.org/PR10805
11633 329. https://gcc.gnu.org/PR10815
11634 330. https://gcc.gnu.org/PR10877
11635 331. https://gcc.gnu.org/PR11280
11636 332. https://gcc.gnu.org/PR11466
11637 333. https://gcc.gnu.org/PR10737
11638 334. https://gcc.gnu.org/PR10810
11639 335. https://gcc.gnu.org/PR8336
11640 336. https://gcc.gnu.org/PR9330
11641 337. https://gcc.gnu.org/PR9631
11642 338. https://gcc.gnu.org/PR9877
11643 339. https://gcc.gnu.org/PR11687
11644 340. https://gcc.gnu.org/PR12263
11645 341. https://gcc.gnu.org/PR12490
11646 342. https://gcc.gnu.org/PR7277
11647 343. https://gcc.gnu.org/PR7939
11648 344. https://gcc.gnu.org/PR11063
11649 345. https://gcc.gnu.org/PR11207
11650 346. https://gcc.gnu.org/PR11522
11651 347. https://gcc.gnu.org/PR11595
11652 348. https://gcc.gnu.org/PR11646
11653 349. https://gcc.gnu.org/PR11665
11654 350. https://gcc.gnu.org/PR11852
11655 351. https://gcc.gnu.org/PR11878
11656 352. https://gcc.gnu.org/PR11883
11657 353. https://gcc.gnu.org/PR11991
11658 354. https://gcc.gnu.org/PR12146
11659 355. https://gcc.gnu.org/PR12215
11660 356. https://gcc.gnu.org/PR12369
11661 357. https://gcc.gnu.org/PR12446
11662 358. https://gcc.gnu.org/PR12510
11663 359. https://gcc.gnu.org/PR12544
11664 360. https://gcc.gnu.org/PR9862
11665 361. https://gcc.gnu.org/PR10962
11666 362. https://gcc.gnu.org/PR11370
11667 363. https://gcc.gnu.org/PR11637
11668 364. https://gcc.gnu.org/PR11885
11669 365. https://gcc.gnu.org/PR12082
11670 366. https://gcc.gnu.org/PR12180
11671 367. https://gcc.gnu.org/PR12340
11672 368. https://gcc.gnu.org/PR3907
11673 369. https://gcc.gnu.org/PR5293
11674 370. https://gcc.gnu.org/PR5296
11675 371. https://gcc.gnu.org/PR7939
11676 372. https://gcc.gnu.org/PR8656
11677 373. https://gcc.gnu.org/PR10147
11678 374. https://gcc.gnu.org/PR11400
11679 375. https://gcc.gnu.org/PR11409
11680 376. https://gcc.gnu.org/PR11740
11681 377. https://gcc.gnu.org/PR11786
11682 378. https://gcc.gnu.org/PR11867
11683 379. https://gcc.gnu.org/PR11928
11684 380. https://gcc.gnu.org/PR12114
11685 381. https://gcc.gnu.org/PR12163
11686 382. https://gcc.gnu.org/PR12181
11687 383. https://gcc.gnu.org/PR12236
11688 384. https://gcc.gnu.org/PR12266
11689 385. https://gcc.gnu.org/PR12296
11690 386. https://gcc.gnu.org/PR12298
11691 387. https://gcc.gnu.org/PR12369
11692 388. https://gcc.gnu.org/PR12337
11693 389. https://gcc.gnu.org/PR12344
11694 390. https://gcc.gnu.org/PR12451
11695 391. https://gcc.gnu.org/PR12486
11696 392. https://gcc.gnu.org/PR8869
11697 393. https://gcc.gnu.org/PR9786
11698 394. https://gcc.gnu.org/PR11689
11699 395. https://gcc.gnu.org/PR12116
11700 396. https://gcc.gnu.org/PR12070
11701 397. https://gcc.gnu.org/PR11184
11702 398. https://gcc.gnu.org/PR11535
11703 399. https://gcc.gnu.org/PR11693
11704 400. https://gcc.gnu.org/PR12224
11705 401. https://gcc.gnu.org/PR11087
11706 402. https://gcc.gnu.org/PR11319
11707 403. https://gcc.gnu.org/PR11949
11708 404. https://gcc.gnu.org/PR11662
11709 405. https://gcc.gnu.org/PR11965
11710 406. https://gcc.gnu.org/PR12301
11711 407. https://gcc.gnu.org/PR11717
11712 408. https://gcc.gnu.org/PR11313
11713 409. https://gcc.gnu.org/PR11712
11714 410. https://gcc.gnu.org/PR12166
11715 411. https://gcc.gnu.org/PR12101
11716 412. https://gcc.gnu.org/PR10988
11717 413. https://gcc.gnu.org/PR11805
11718 414. https://gcc.gnu.org/PR11902
11719 415. https://gcc.gnu.org/PR11903
11720 416. https://gcc.gnu.org/PR11890
11721 417. https://gcc.gnu.org/PR12399
11722 418. https://gcc.gnu.org/PR13068
11723 419. https://gcc.gnu.org/PR10060
11724 420. https://gcc.gnu.org/PR10555
11725 421. https://gcc.gnu.org/PR10706
11726 422. https://gcc.gnu.org/PR11496
11727 423. https://gcc.gnu.org/PR11741
11728 424. https://gcc.gnu.org/PR12440
11729 425. https://gcc.gnu.org/PR12632
11730 426. https://gcc.gnu.org/PR12712
11731 427. https://gcc.gnu.org/PR12726
11732 428. https://gcc.gnu.org/PR12890
11733 429. https://gcc.gnu.org/PR12900
11734 430. https://gcc.gnu.org/PR13060
11735 431. https://gcc.gnu.org/PR13289
11736 432. https://gcc.gnu.org/PR13318
11737 433. https://gcc.gnu.org/PR13392
11738 434. https://gcc.gnu.org/PR13574
11739 435. https://gcc.gnu.org/PR13475
11740 436. https://gcc.gnu.org/PR13797
11741 437. https://gcc.gnu.org/PR13824
11742 438. https://gcc.gnu.org/PR8776
11743 439. https://gcc.gnu.org/PR10339
11744 440. https://gcc.gnu.org/PR11350
11745 441. https://gcc.gnu.org/PR12826
11746 442. https://gcc.gnu.org/PR12500
11747 443. https://gcc.gnu.org/PR12941
11748 444. https://gcc.gnu.org/PR12953
11749 445. https://gcc.gnu.org/PR13041
11750 446. https://gcc.gnu.org/PR13507
11751 447. https://gcc.gnu.org/PR13382
11752 448. https://gcc.gnu.org/PR13394
11753 449. https://gcc.gnu.org/PR13400
11754 450. https://gcc.gnu.org/PR13521
11755 451. https://gcc.gnu.org/PR2094
11756 452. https://gcc.gnu.org/PR2294
11757 453. https://gcc.gnu.org/PR5050
11758 454. https://gcc.gnu.org/PR9371
11759 455. https://gcc.gnu.org/PR9546
11760 456. https://gcc.gnu.org/PR10081
11761 457. https://gcc.gnu.org/PR10093
11762 458. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61
11763 459. https://gcc.gnu.org/PR10095
11764 460. https://gcc.gnu.org/PR11554
11765 461. https://gcc.gnu.org/PR12297
11766 462. https://gcc.gnu.org/PR12352
11767 463. https://gcc.gnu.org/PR12438
11768 464. https://gcc.gnu.org/PR12540
11769 465. https://gcc.gnu.org/PR12594
11770 466. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60
11771 467. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63
11772 468. https://gcc.gnu.org/PR12657
11773 469. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292
11774 470. https://gcc.gnu.org/PR12696
11775 471. https://gcc.gnu.org/PR12815
11776 472. https://gcc.gnu.org/PR12862
11777 473. https://gcc.gnu.org/PR12926
11778 474. https://gcc.gnu.org/PR12967
11779 475. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html
11780 476. https://gcc.gnu.org/PR12971
11781 477. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328
11782 478. https://gcc.gnu.org/PR13007
11783 479. https://gcc.gnu.org/PR13009
11784 480. https://gcc.gnu.org/PR13057
11785 481. https://gcc.gnu.org/PR13070
11786 482. https://gcc.gnu.org/PR13081
11787 483. https://gcc.gnu.org/PR13239
11788 484. https://gcc.gnu.org/PR13262
11789 485. https://gcc.gnu.org/PR13290
11790 486. https://gcc.gnu.org/PR13323
11791 487. https://gcc.gnu.org/PR13369
11792 488. https://gcc.gnu.org/PR13371
11793 489. https://gcc.gnu.org/PR13445
11794 490. https://gcc.gnu.org/PR13461
11795 491. https://gcc.gnu.org/PR13462
11796 492. https://gcc.gnu.org/PR13478
11797 493. https://gcc.gnu.org/PR13544
11798 494. https://gcc.gnu.org/PR13650
11799 495. https://gcc.gnu.org/PR13683
11800 496. https://gcc.gnu.org/PR13688
11801 497. https://gcc.gnu.org/PR13774
11802 498. https://gcc.gnu.org/PR13884
11803 499. https://gcc.gnu.org/PR10746
11804 500. https://gcc.gnu.org/PR11433
11805 501. https://gcc.gnu.org/PR12633
11806 502. https://gcc.gnu.org/PR13037
11807 503. https://gcc.gnu.org/PR13213
11808 504. https://gcc.gnu.org/PR4490
11809 505. https://gcc.gnu.org/PR12292
11810 506. https://gcc.gnu.org/PR12441
11811 507. https://gcc.gnu.org/PR12943
11812 508. https://gcc.gnu.org/PR13608
11813 509. https://gcc.gnu.org/PR11598
11814 510. https://gcc.gnu.org/PR11793
11815 511. https://gcc.gnu.org/PR12467
11816 512. https://gcc.gnu.org/PR12537
11817 513. https://gcc.gnu.org/PR12496
11818 514. https://gcc.gnu.org/PR12865
11819 515. https://gcc.gnu.org/PR13354
11820 516. https://gcc.gnu.org/PR10467
11821 517. https://gcc.gnu.org/PR11226
11822 518. https://gcc.gnu.org/PR11227
11823 519. https://gcc.gnu.org/PR12644
11824 520. https://gcc.gnu.org/PR13149
11825 521. https://gcc.gnu.org/PR12654
11826 522. https://gcc.gnu.org/PR12965
11827 523. https://gcc.gnu.org/PR13031
11828 524. https://gcc.gnu.org/PR11634
11829 525. https://gcc.gnu.org/PR12158
11830 526. https://gcc.gnu.org/PR11992
11831 527. https://gcc.gnu.org/PR9365
11832 528. https://gcc.gnu.org/PR10392
11833 529. https://gcc.gnu.org/PR11322
11834 530. https://gcc.gnu.org/PR13069
11835 531. https://gcc.gnu.org/PR13302
11836 532. https://gcc.gnu.org/PR13585
11837 533. https://gcc.gnu.org/PR8916
11838 534. https://gcc.gnu.org/PR11576
11839 535. https://gcc.gnu.org/PR13122
11840 536. https://gcc.gnu.org/PR13256
11841 537. https://gcc.gnu.org/PR13373
11842 538. https://gcc.gnu.org/PR12561
11843 539. https://gcc.gnu.org/PR6243
11844 540. https://gcc.gnu.org/PR11397
11845 541. https://gcc.gnu.org/PR12505
11846 542. https://gcc.gnu.org/PR13150
11847 543. https://gcc.gnu.org/PR12666
11848 544. https://gcc.gnu.org/PR12969
11849 545. https://gcc.gnu.org/PR10819
11850 546. https://gcc.gnu.org/PR11612
11851 547. https://gcc.gnu.org/PR13211
11852 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4
11853 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5
11854 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6
11855 551. https://gcc.gnu.org/onlinedocs/
11856 552. mailto:gcc-help@gcc.gnu.org
11857 553. mailto:gcc@gcc.gnu.org
11858 554. https://gcc.gnu.org/lists.html
11859 555. http://www.fsf.org/
11860 556. https://gcc.gnu.org/about.html
11861 557. http://validator.w3.org/check/referer
11862 ======================================================================
11863 http://gcc.gnu.org/gcc-3.2/index.html
11864 GCC 3.2 Release Series
11868 The [1]GNU project and the GCC developers are pleased to announce the
11869 release of GCC 3.2.3.
11871 The purpose of the GCC 3.2 release series is to provide a stable
11872 platform for OS distributors to use building their next releases. A
11873 primary objective was to stabilize the C++ ABI; we believe that the
11874 interface to the compiler and the C++ standard library are now
11877 Be aware that C++ code compiled by GCC 3.2.x will (in general) not
11878 interoperate with code compiled by GCC 3.1.1 or earlier.
11880 Please refer to our [2]detailed list of news, caveats, and bug-fixes
11881 for further information.
11886 April 25, 2003 ([3]changes)
11889 February 5, 2003 ([4]changes)
11892 November 19, 2002 ([5]changes)
11895 August 14, 2002 ([6]changes)
11897 References and Acknowledgements
11899 GCC used to stand for the GNU C Compiler, but since the compiler
11900 supports several other languages aside from C, it now stands for the
11901 GNU Compiler Collection.
11903 A list of [7]successful builds is updated as new information becomes
11906 The GCC developers would like to thank the numerous people that have
11907 contributed new features, improvements, bug fixes, and other changes as
11908 well as test results to GCC. This [8]amazing group of volunteers is
11909 what makes GCC successful.
11911 For additional information about GCC please refer to the [9]GCC project
11912 web site or contact the [10]GCC development mailing list.
11914 To obtain GCC please use [11]our mirror sites, or our CVS server.
11917 For questions related to the use of GCC, please consult these web
11918 pages and the [12]GCC manuals. If that fails, the
11919 [13]gcc-help@gcc.gnu.org mailing list might help. Comments on these
11920 web pages and the development of GCC are welcome on our developer
11921 list at [14]gcc@gcc.gnu.org. All of [15]our lists have public
11924 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and
11925 distribution of this entire article is permitted in any medium,
11926 provided this notice is preserved.
11928 These pages are [17]maintained by the GCC team. Last modified
11933 1. http://www.gnu.org/
11934 2. http://gcc.gnu.org/gcc-3.2/changes.html
11935 3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
11936 4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2
11937 5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1
11938 6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2
11939 7. http://gcc.gnu.org/gcc-3.2/buildstat.html
11940 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
11941 9. http://gcc.gnu.org/index.html
11942 10. mailto:gcc@gcc.gnu.org
11943 11. http://gcc.gnu.org/mirrors.html
11944 12. https://gcc.gnu.org/onlinedocs/
11945 13. mailto:gcc-help@gcc.gnu.org
11946 14. mailto:gcc@gcc.gnu.org
11947 15. https://gcc.gnu.org/lists.html
11948 16. http://www.fsf.org/
11949 17. https://gcc.gnu.org/about.html
11950 18. http://validator.w3.org/check/referer
11951 ======================================================================
11952 http://gcc.gnu.org/gcc-3.2/changes.html
11953 GCC 3.2 Release Series
11954 Changes, New Features, and Fixes
11956 The latest release in the 3.2 release series is [1]GCC 3.2.3.
11958 Caveats and New Features
11962 * The C++ compiler does not correctly zero-initialize
11963 pointers-to-data members. You must explicitly initialize them. For
11964 example: int S::*m(0); will work, but depending on
11965 default-initialization to zero will not work. This bug cannot be
11966 fixed in GCC 3.2 without inducing unacceptable risks. It will be
11968 * This GCC release is based on the GCC 3.1 sourcebase, and thus has
11969 all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has
11970 a number of C++ ABI fixes which make its C++ compiler generate
11971 binary code which is incompatible with the C++ compilers found in
11972 earlier GCC releases, including GCC 3.1 and GCC 3.1.1.
11974 Frontend Enhancements
11978 * The method of constructing the list of directories to be searched
11979 for header files has been revised. If a directory named by a -I
11980 option is a standard system include directory, the option is
11981 ignored to ensure that the default search order for system
11982 directories and the special treatment of system header files are
11984 * The C and Objective-C compilers no longer accept the "Naming Types"
11985 extension (typedef foo = bar); it was already unavailable in C++.
11986 Code which uses it will need to be changed to use the "typeof"
11987 extension instead: typedef typeof(bar) foo. (We have removed this
11988 extension without a period of deprecation because it has caused the
11989 compiler to crash since version 3.0 and no one noticed until very
11990 recently. Thus we conclude it is not in widespread use.)
11994 * GCC 3.2 fixed serveral differences between the C++ ABI implemented
11995 in GCC and the multi-vendor standard, but more have been found
11996 since the release. 3.2.1 adds a new warning, -Wabi, to warn about
11997 code which is affected by these bugs. We will fix these bugs in
11998 some future release, once we are confident that all have been
11999 found; until then, it is our intention to make changes to the ABI
12000 only if they are necessary for correct compilation of C++, as
12001 opposed to conformance to the ABI documents.
12002 * For details on how to build an ABI compliant compiler for GNU/Linux
12003 systems, check the [3]common C++ ABI page.
12005 New Targets and Target Specific Improvements
12009 * Fixed a number of bugs in SSE and MMX intrinsics.
12010 * Fixed common compiler crashes with SSE instruction set enabled
12011 (implied by -march=pentium3, pentium4, athlon-xp)
12012 * __m128 and __m128i is not 128bit aligned when used in structures.
12016 * A bug whereby the compiler could generate bad code for bzero has
12018 * ABI fixes (implying ABI incompatibilities with previous version in
12020 * Fixed prefetch code generation
12021 __________________________________________________________________
12025 3.2.3 is a bug fix release only; there are no new features that were
12026 not present in GCC 3.2.2.
12030 This section lists the problem reports (PRs) from GCC's bug tracking
12031 system that are known to be fixed in the 3.2.3 release. This list might
12032 not be complete (that is, it is possible that some PRs that have been
12033 fixed are not listed here), and some of the titles have been changed to
12034 make them more clear.
12036 Internal Compiler Errors (multi-platform)
12038 * [4]3782: (c++) -quiet -fstats produces a segmentation fault in
12040 * [5]6440: (c++) template specializations cause ICE
12041 * [6]7050: (c++) ICE on: (i ? get_string() : throw)
12042 * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c)
12043 * [8]7982: (c++) ICE due to infinite recursion (using STL set)
12044 * [9]8068: exceedingly high (infinite) memory usage
12045 * [10]8178: ICE with __builtin_ffs
12046 * [11]8396: ICE in copy_to_mode_reg, in explow.c
12047 * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c
12048 * [13]9768: ICE when optimizing inline code at -O2
12049 * [14]9798: (c++) Infinite recursion (segfault) in
12050 cp/decl.c:push_using_directive with recursive using directives
12051 * [15]9799: mismatching structure initializer with nested flexible
12053 * [16]9928: ICE on duplicate enum declaration
12054 * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects
12056 * [18]10352: ICE in find_reloads_toplev
12057 * [19]10336: ICE with -Wunreachable-code
12061 * [20]8224: Incorrect joining of signed and unsigned division
12062 * [21]8613: -O2 produces wrong code with builtin strlen and
12064 * [22]8828: gcc reports some code is unreachable when it is not
12065 * [23]9226: GCSE breaking argument passing
12066 * [24]9853: miscompilation of non-constant structure initializer
12067 * [25]9797: C99-style struct initializers are miscompiled
12068 * [26]9967: Some standard C function calls should not be replaced
12069 when optimizing for size
12070 * [27]10116: ce2: invalid merge of join_bb in the context of switch
12072 * [28]10171: wrong code for inlined function
12073 * [29]10175: -Wunreachable-code doesn't work for single lines
12075 C++ compiler and library:
12077 * [30]8316: Confusing diagnostic for code that misuses conversion
12079 * [31]9169: filebuf output fails if codecvt<>::out returns noconv
12080 * [32]9420: incomplete type incorrectly reported
12081 * [33]9459: typeof in return type specification of template not
12083 * [34]9507: filebuf::open handles ios_base::ate incorrectly
12084 * [35]9538: Out-of-bounds memory access in streambuf::sputbackc
12085 * [36]9602: Total confusion about template/friend/virtual/abstract
12086 * [37]9993: destructor not called for local object created within and
12087 returned from infinite loop
12088 * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2
12091 Java compiler and library:
12093 * [39]9652: libgcj build fails on irix6.5.1[78]
12094 * [40]10144: gas on solaris complains about bad .stabs lines for
12095 java, native as unaffected
12097 x86-specific (Intel/AMD):
12099 * [41]8746: gcc miscompiles Linux kernel ppa driver on x86
12100 * [42]9888: -mcpu=k6 -Os produces out of range loop instructions
12101 * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu
12103 * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib)
12108 * [45]7784: [Sparc] ICE in extract_insn, in recog.c
12109 * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in
12111 * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc
12112 * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in
12114 * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc
12115 * [50]9414: Scheduling bug on Ultrasparc
12116 * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64
12120 * [52]7248: broken "inclusive or" code
12121 * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1
12125 * [54]9732: Wrong code with -O2 -fPIC
12126 * [55]10073: ICE: powerpc cannot split insn
12130 * [56]7702: optimization problem on a DEC alpha under OSF1
12131 * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system
12135 * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275)
12136 * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10
12138 * [60]10271: Floating point args don't get reloaded across function
12143 * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in
12148 * [62]10377: gcc-3.2.2 creates bad assembler code for cris
12150 Miscellaneous and minor bugs:
12152 * [63]6955: collect2 says "core dumped" when there is no core
12153 __________________________________________________________________
12157 Beginning with 3.2.2, GCC's Makefile suite supports redirection of make
12158 install by means of the DESTDIR variable. Parts of the GCC tree have
12159 featured that support long before, but now it is available even from
12162 Other than that, GCC 3.2.2 is a bug fix release only; there are no new
12163 features that were not present in GCC 3.2.1.
12167 On the following i386-based systems GCC 3.2.1 broke the C ABI wrt.
12168 functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped
12169 with FreeBSD 5.0 does not have this problem), Interix, a.out-based
12170 GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI
12171 change, and thus restores ABI-compatibility with previous releases
12172 (except GCC 3.2.1) on these platforms.
12174 This section lists the problem reports (PRs) from GCC's bug tracking
12175 system that are known to be fixed in the 3.2.2 release. This list might
12176 not be complete (that is, it is possible that some PRs that have been
12177 fixed are not listed here) and some of the titles have been changed to
12178 make them more clear.
12180 Internal Compiler Errors (multi-platform)
12182 * [64]5919: (c++) ICE when passing variable array to template
12184 * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=)
12185 * [66]7507: ICE with -O2 when address of called function is a
12186 complicated expression
12187 * [67]7622: ICE with nested inline functions if function's address is
12189 * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR
12191 * [70]8031: (c++) ICE in code comparing typeids and casting from
12193 * [71]8275: ICE in simplify_subreg
12194 * [72]8332: (c++) builtin strlen/template interaction causes ICE
12195 * [73]8372: (c++) ICE on explicit call of destructor
12196 * [74]8439: (c, not c++) empty struct causes ICE
12197 * [75]8442: (c++) ICE with nested template classes
12198 * [76]8518: ICE when compiling mplayer ("extern inline" issue)
12199 * [77]8615: (c++) ICE with out-of-range character constant template
12201 * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307
12202 * [79]8799: (c++) ICE: error reporting routines re-entered
12203 * [80]9328: (c++) ICE with typeof(X) for overloaded X
12204 * [81]9465: (preprocessor) cpp -traditional ICE on null bytes
12206 C++ (compiler and library) bugs
12208 * [82]47: scoping in nested classes is broken
12209 * [83]6745: problems with iostream rdbuf() member function
12210 * [84]8214: conversion from const char* const to char* sometimes
12212 * [85]8493: builtin strlen and overload resolution (same bug as
12214 * [87]8503: strange behaviour of function types
12215 * [88]8727: compiler confused by inheritance from an anonymous struct
12216 * [89]7445: poor performance of std::locale::classic() in
12217 multi-threaded applications
12218 * [90]8230: mishandling of overflow in vector<T>::resize
12219 * [91]8399: sync_with_stdio(false) breaks unformatted input
12220 * [92]8662: illegal access of private member of unnamed class is
12222 * [93]8707: "make distclean" fails in libstdc++-v3 directory
12223 * [94]8708: __USE_MALLOC doesn't work
12224 * [95]8790: Use of non-thread-safe strtok in src/localename.cc
12225 * [96]8887: Bug in date formats with --enable-clocale=generic
12226 * [97]9076: Call Frame Instructions are not handled correctly during
12228 * [98]9151: std::setprecision limited to 16 digits when outputting a
12230 * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers
12231 * [100]9269: libstdc++ headers: explicit specialization of function
12232 must precede its first use
12233 * [101]9322: return value of basic_streambuf<>::getloc affected by
12235 * [102]9433: segfault in runtime support for dynamic_cast
12237 C and optimizer bugs
12239 * [103]8032: GCC incorrectly initializes static structs that have
12241 * [104]8639: simple arithmetic expression broken
12242 * [105]8794: optimization improperly eliminates certain expressions
12243 * [106]8832: traditional "asm volatile" code is illegally optimized
12244 * [107]8988: loop optimizer bug: with -O2, code is generated that
12245 segfaults (found on i386, bug present for all platforms)
12246 * [108]9492: structure copy clobbers subsequent stores to structure
12250 * [109]9267: Objective-C parser won't build with newer bison versions
12255 * [110]8344: Ada build problem due to conflict between gcc/final.o,
12260 * [111]8524: _Pragma within macros is improperly expanded
12261 * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with
12266 * [113]9090: arm ICE with >= -O2; regression from gcc-2.95
12268 x86-specific (Intel/AMD)
12270 * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction)
12271 * [115]8599: loop unroll bug with -march=k6-3
12272 * [116]9506: ABI breakage in structure return (affects BSD and
12273 Cygwin, but not GNU/Linux)
12275 FreeBSD 5.0 specific
12277 * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0
12281 * [118]9292: hppa1.1-rtems configurery problems
12282 * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug
12283 * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue
12284 * [121]9296: gthr-rtems regression
12285 * [122]9316: powerpc-rtems: extending multilibs
12289 * [123]9493: ICE with -O2 when building a simple function
12293 * [124]7341: hyperlink to gcov in GCC documentation doesn't work
12294 * [125]8947: Please add a warning about "-malign-double" in docs
12295 * [126]7448, [127]8882: typo cleanups
12296 __________________________________________________________________
12300 3.2.1 adds a new warning, -Wabi. This option warns when GNU C++
12301 generates code that is known not to be binary-compatible with the
12302 vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included
12303 in the distribution, for details.
12305 This release also removes an old GCC extension, "naming types", and the
12306 documentation now directs users to use a different GCC extension,
12307 __typeof__, instead. The feature had evidently been broken for a while.
12309 Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and
12310 the new warning there are no new features that were not present in GCC
12313 In addition, the previous fix for [128]PR 7445 (poor performance of
12314 std::locale::classic() in multi-threaded applications) was reverted
12315 ("unfixed"), because the "fix" was not thread-safe.
12319 This section lists the problem reports (PRs) from GCC's bug tracking
12320 system that are known to be fixed in the 3.2.1 release. This list might
12321 not be complete (that is, it is possible that some PRs that have been
12322 fixed are not listed here). As you can see, the number of bug fixes is
12323 quite large, so it is strongly recommended that users of earlier GCC
12324 3.x releases upgrade to GCC 3.2.1.
12326 Internal Compiler Errors (multi-platform)
12328 * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c
12329 * [130]5661: (c++) ICE instantiating template on array of unknown
12331 * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on
12333 * [132]6994: (c++) ICE in find_function_data
12334 * [133]7150: preprocessor: GCC -dM -E gives an ICE
12335 * [134]7160: ICE when optimizing branches without a return value
12336 * [135]7228: (c++) ICE when using member template and template
12338 * [136]7266: (c++) ICE with -pedantic on missing typename
12339 * [137]7353: ICE from use of "Naming Types" extension, see above
12340 * [138]7411: ICE in instantiate_virtual_regs_1, in function.c
12341 * [139]7478: (c++) ICE on static_cast inside template
12342 * [140]7526: preprocessor core dump when _Pragma implies #pragma
12344 * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803
12346 * [143]7754: (c++) ICE on union with template parameter
12347 * [144]7788: (c++) redeclaring a definition as an incomplete class
12349 * [145]8031: (c++) ICE in comptypes, in cp/typeck.c
12350 * [146]8055: preprocessor dies with SIG11 when building FreeBSD
12352 * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related
12354 * [148]8134: (c++) ICE in force_store_init_value on legal code
12355 * [149]8149: (c++) ICE on incomplete type
12356 * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array
12359 C++ (compiler and library) bugs
12361 * [151]5607: No pointer adjustment in covariant return types
12362 * [152]6579: Infinite loop with statement expressions in member
12364 * [153]6803: Default copy constructor bug in GCC 3.1
12365 * [154]7176: g++ confused by friend and static member with same name
12366 * [155]7188: Segfault with template class and recursive (incorrect)
12368 * [156]7306: Regression: GCC 3.x fails to compile code with virtual
12369 inheritance if a method has a variable number of arguments
12370 * [157]7461: ctype<char>::classic_table() returns offset array on
12372 * [158]7524: f(const float arg[3]) fails
12373 * [159]7584: Erroneous ambiguous base error on using declaration
12374 * [160]7676: Member template overloading problem
12375 * [161]7679: infinite loop when a right parenthesis is missing
12376 * [162]7811: default locale not taken from environment
12377 * [163]7961: compare( char *) implemented incorrectly in
12379 * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if
12380 streambuf::underflow() leaves gptr() NULL (dups: [165]8127,
12382 * [167]8096: deque::at() throws std::range_error instead of
12384 * [168]8127: cout << cin.rdbuf() infinite loop
12385 * [169]8218: Excessively large memory consumed for classes with large
12387 * [170]8287: GCC 3.2: Destructor called for non-constructed local
12389 * [171]8347: empty vector range used in string construction causes
12391 * [172]8348: fail() flag is set in istringstream when eof() flag is
12393 * [173]8391: regression: infinite loop in cp/decl2.c(finish_file)
12395 C and optimizer bugs
12397 * [174]6627: -fno-align-functions doesn't seem to disable function
12399 * [175]6631: life_analysis misoptimizes code to initialize fields of
12401 * [176]7102: unsigned char division results in floating exception
12402 * [177]7120: Run once loop should *always* be unrolled
12404 * [178]7209: Bug involving array referencing and ?: operator
12405 * [179]7515: invalid inlining of global function with -O3
12406 * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test
12407 * [181]8467: bug in sibling call optimization
12411 * [182]4890: incorrect line markers from the traditional preprocessor
12412 * [183]7357: -M option omits system headers files (making it the same
12414 * [184]7358: Changes to Sun's make Dependencies
12415 * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as
12417 * [186]7862: Interrupting GCC -MD removes .d file but not .o
12418 * [187]8190: Failed compilation deletes -MD dependency file
12419 * [188]8524: _Pragma within macro is improperly expanded
12421 x86 specific (Intel/AMD)
12423 * [189]5351: (i686-only) function pass-by-value structure copy
12424 corrupts stack ([190]7591 is a duplicate)
12425 * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with
12426 -march=pentium3/pentium2/athlon (these are all the same underlying
12427 bug, in MMX register use)
12428 * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same
12430 * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken
12431 * [199]6981: wrong code in 64-bit manipulation on x86
12432 * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__
12434 * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE
12435 intrinsics are broken
12436 * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with
12438 * [203]7693: Typo in i386 mmintrin.h header
12439 * [204]7723: ICE - Pentium3 sse - GCC 3.2
12440 * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse
12441 * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3
12445 * [207]5967: GCC bug when profiling nested functions on powerpc
12446 * [208]6984: wrong code generated with -O2, -O3, -Os for do-while
12448 * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5
12449 * [210]7130: miscompiled code for GCC-3.1 on
12450 powerpc-unknown-linux-gnu with -funroll-all-loops
12451 * [211]7133: PowerPC ICE: unrecognizable insn
12452 * [212]7380: ICE in extract_insn, at recog.c:2148
12453 * [213]8252: ICE on Altivec code with optimization turned on
12454 * [214]8451: Altivec ICE in GCC 3.2
12458 * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa
12462 * [216]6668: when using --disable-multilib, libgcc_s.so is installed
12463 in the wrong place on sparc-solaris
12464 * [217]7151: ICE when compiling for UltraSPARC
12465 * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long
12467 * [219]7842: [REGRESSION] SPARC code gen bug
12471 * [220]7856: [arm] invalid offset in constant pool reference
12472 * [221]7967: optimization produces wrong code (ARM)
12476 * [222]7374: __builtin_fabsl broken on alpha
12480 * [223]7370: ICE in fixup_var_refs_1 on s390x
12481 * [224]7409: loop optimization bug on s390x-linux-gnu
12482 * [225]8232: s390x: ICE when using bcmp with int length argument
12486 * [226]7623: SCO OpenServer build fails with machmode.def: undefined
12487 symbol: BITS_PER_UNIT
12489 m68k/Coldfire specific
12491 * [227]8314: crtbegin, crtend need to be multilib'ed for this
12496 * [228]761: Document some undocumented options
12497 * [229]5610: Fix documentation about invoking SSE instructions
12499 * [230]7484: List -Wmissing-declarations as C-only option
12500 * [231]7531: -mcmodel not documented for x86-64
12501 * [232]8120: Update documentation of bad use of ##
12502 __________________________________________________________________
12506 3.2 is a small bug fix release, but there is a change to the
12507 application binary interface (ABI), hence the change to the second part
12508 of the version number.
12510 The main purpose of the 3.2 release is to correct a couple of problems
12511 in the C++ ABI, with the intention of providing a stable interface
12512 going forward. Accordingly, 3.2 is only a small change to 3.1.1.
12518 * [233]7320: g++ 3.2 relocation problem
12519 * [234]7470: vtable: virtual function pointers not in declaration
12524 * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t
12525 * [236]6503, [237]6642, [238]7186: Problems with comparing or
12526 subtracting various types of const and non-const iterators
12527 * [239]7216: ambiguity with basic_iostream::traits_type
12528 * [240]7220: problem with basic_istream::ignore(0,delimiter)
12529 * [241]7222: locale::operator==() doesn't work on std::locale("")
12530 * [242]7286: placement operator delete issue
12531 * [243]7442: cxxabi.h does not match the C++ ABI
12532 * [244]7445: poor performance of std::locale::classic() in
12533 multi-threaded applications
12537 * [245]7291: off-by-one in generated inline bzero code for x86-64
12540 For questions related to the use of GCC, please consult these web
12541 pages and the [246]GCC manuals. If that fails, the
12542 [247]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12543 web pages and the development of GCC are welcome on our developer
12544 list at [248]gcc@gcc.gnu.org. All of [249]our lists have public
12547 Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and
12548 distribution of this entire article is permitted in any medium,
12549 provided this notice is preserved.
12551 These pages are [251]maintained by the GCC team. Last modified
12556 1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3
12557 2. http://gcc.gnu.org/gcc-3.1/changes.html
12558 3. http://gcc.gnu.org/gcc-3.2/c++-abi.html
12559 4. https://gcc.gnu.org/PR3782
12560 5. https://gcc.gnu.org/PR6440
12561 6. https://gcc.gnu.org/PR7050
12562 7. https://gcc.gnu.org/PR7741
12563 8. https://gcc.gnu.org/PR7982
12564 9. https://gcc.gnu.org/PR8068
12565 10. https://gcc.gnu.org/PR8178
12566 11. https://gcc.gnu.org/PR8396
12567 12. https://gcc.gnu.org/PR8674
12568 13. https://gcc.gnu.org/PR9768
12569 14. https://gcc.gnu.org/PR9798
12570 15. https://gcc.gnu.org/PR9799
12571 16. https://gcc.gnu.org/PR9928
12572 17. https://gcc.gnu.org/PR10114
12573 18. https://gcc.gnu.org/PR10352
12574 19. https://gcc.gnu.org/PR10336
12575 20. https://gcc.gnu.org/PR8224
12576 21. https://gcc.gnu.org/PR8613
12577 22. https://gcc.gnu.org/PR8828
12578 23. https://gcc.gnu.org/PR9226
12579 24. https://gcc.gnu.org/PR9853
12580 25. https://gcc.gnu.org/PR9797
12581 26. https://gcc.gnu.org/PR9967
12582 27. https://gcc.gnu.org/PR10116
12583 28. https://gcc.gnu.org/PR10171
12584 29. https://gcc.gnu.org/PR10175
12585 30. https://gcc.gnu.org/PR8316
12586 31. https://gcc.gnu.org/PR9169
12587 32. https://gcc.gnu.org/PR9420
12588 33. https://gcc.gnu.org/PR9459
12589 34. https://gcc.gnu.org/PR9507
12590 35. https://gcc.gnu.org/PR9538
12591 36. https://gcc.gnu.org/PR9602
12592 37. https://gcc.gnu.org/PR9993
12593 38. https://gcc.gnu.org/PR10167
12594 39. https://gcc.gnu.org/PR9652
12595 40. https://gcc.gnu.org/PR10144
12596 41. https://gcc.gnu.org/PR8746
12597 42. https://gcc.gnu.org/PR9888
12598 43. https://gcc.gnu.org/PR9638
12599 44. https://gcc.gnu.org/PR9954
12600 45. https://gcc.gnu.org/PR7784
12601 46. https://gcc.gnu.org/PR7796
12602 47. https://gcc.gnu.org/PR8281
12603 48. https://gcc.gnu.org/PR8366
12604 49. https://gcc.gnu.org/PR8726
12605 50. https://gcc.gnu.org/PR9414
12606 51. https://gcc.gnu.org/PR10067
12607 52. https://gcc.gnu.org/PR7248
12608 53. https://gcc.gnu.org/PR8343
12609 54. https://gcc.gnu.org/PR9732
12610 55. https://gcc.gnu.org/PR10073
12611 56. https://gcc.gnu.org/PR7702
12612 57. https://gcc.gnu.org/PR9671
12613 58. https://gcc.gnu.org/PR8694
12614 59. https://gcc.gnu.org/PR9953
12615 60. https://gcc.gnu.org/PR10271
12616 61. https://gcc.gnu.org/PR6362
12617 62. https://gcc.gnu.org/PR10377
12618 63. https://gcc.gnu.org/PR6955
12619 64. https://gcc.gnu.org/PR5919
12620 65. https://gcc.gnu.org/PR7129
12621 66. https://gcc.gnu.org/PR7507
12622 67. https://gcc.gnu.org/PR7622
12623 68. https://gcc.gnu.org/PR7681
12624 69. https://gcc.gnu.org/PR9528
12625 70. https://gcc.gnu.org/PR8031
12626 71. https://gcc.gnu.org/PR8275
12627 72. https://gcc.gnu.org/PR8332
12628 73. https://gcc.gnu.org/PR8372
12629 74. https://gcc.gnu.org/PR8439
12630 75. https://gcc.gnu.org/PR8442
12631 76. https://gcc.gnu.org/PR8518
12632 77. https://gcc.gnu.org/PR8615
12633 78. https://gcc.gnu.org/PR8663
12634 79. https://gcc.gnu.org/PR8799
12635 80. https://gcc.gnu.org/PR9328
12636 81. https://gcc.gnu.org/PR9465
12637 82. https://gcc.gnu.org/PR47
12638 83. https://gcc.gnu.org/PR6745
12639 84. https://gcc.gnu.org/PR8214
12640 85. https://gcc.gnu.org/PR8493
12641 86. https://gcc.gnu.org/PR8332
12642 87. https://gcc.gnu.org/PR8503
12643 88. https://gcc.gnu.org/PR8727
12644 89. https://gcc.gnu.org/PR7445
12645 90. https://gcc.gnu.org/PR8230
12646 91. https://gcc.gnu.org/PR8399
12647 92. https://gcc.gnu.org/PR8662
12648 93. https://gcc.gnu.org/PR8707
12649 94. https://gcc.gnu.org/PR8708
12650 95. https://gcc.gnu.org/PR8790
12651 96. https://gcc.gnu.org/PR8887
12652 97. https://gcc.gnu.org/PR9076
12653 98. https://gcc.gnu.org/PR9151
12654 99. https://gcc.gnu.org/PR9168
12655 100. https://gcc.gnu.org/PR9269
12656 101. https://gcc.gnu.org/PR9322
12657 102. https://gcc.gnu.org/PR9433
12658 103. https://gcc.gnu.org/PR8032
12659 104. https://gcc.gnu.org/PR8639
12660 105. https://gcc.gnu.org/PR8794
12661 106. https://gcc.gnu.org/PR8832
12662 107. https://gcc.gnu.org/PR8988
12663 108. https://gcc.gnu.org/PR9492
12664 109. https://gcc.gnu.org/PR9267
12665 110. https://gcc.gnu.org/PR8344
12666 111. https://gcc.gnu.org/PR8524
12667 112. https://gcc.gnu.org/PR8880
12668 113. https://gcc.gnu.org/PR9090
12669 114. https://gcc.gnu.org/PR8588
12670 115. https://gcc.gnu.org/PR8599
12671 116. https://gcc.gnu.org/PR9506
12672 117. https://gcc.gnu.org/PR9484
12673 118. https://gcc.gnu.org/PR9292
12674 119. https://gcc.gnu.org/PR9293
12675 120. https://gcc.gnu.org/PR9295
12676 121. https://gcc.gnu.org/PR9296
12677 122. https://gcc.gnu.org/PR9316
12678 123. https://gcc.gnu.org/PR9493
12679 124. https://gcc.gnu.org/PR7341
12680 125. https://gcc.gnu.org/PR8947
12681 126. https://gcc.gnu.org/PR7448
12682 127. https://gcc.gnu.org/PR8882
12683 128. https://gcc.gnu.org/PR7445
12684 129. https://gcc.gnu.org/PR2521
12685 130. https://gcc.gnu.org/PR5661
12686 131. https://gcc.gnu.org/PR6419
12687 132. https://gcc.gnu.org/PR6994
12688 133. https://gcc.gnu.org/PR7150
12689 134. https://gcc.gnu.org/PR7160
12690 135. https://gcc.gnu.org/PR7228
12691 136. https://gcc.gnu.org/PR7266
12692 137. https://gcc.gnu.org/PR7353
12693 138. https://gcc.gnu.org/PR7411
12694 139. https://gcc.gnu.org/PR7478
12695 140. https://gcc.gnu.org/PR7526
12696 141. https://gcc.gnu.org/PR7721
12697 142. https://gcc.gnu.org/PR7803
12698 143. https://gcc.gnu.org/PR7754
12699 144. https://gcc.gnu.org/PR7788
12700 145. https://gcc.gnu.org/PR8031
12701 146. https://gcc.gnu.org/PR8055
12702 147. https://gcc.gnu.org/PR8067
12703 148. https://gcc.gnu.org/PR8134
12704 149. https://gcc.gnu.org/PR8149
12705 150. https://gcc.gnu.org/PR8160
12706 151. https://gcc.gnu.org/PR5607
12707 152. https://gcc.gnu.org/PR6579
12708 153. https://gcc.gnu.org/PR6803
12709 154. https://gcc.gnu.org/PR7176
12710 155. https://gcc.gnu.org/PR7188
12711 156. https://gcc.gnu.org/PR7306
12712 157. https://gcc.gnu.org/PR7461
12713 158. https://gcc.gnu.org/PR7524
12714 159. https://gcc.gnu.org/PR7584
12715 160. https://gcc.gnu.org/PR7676
12716 161. https://gcc.gnu.org/PR7679
12717 162. https://gcc.gnu.org/PR7811
12718 163. https://gcc.gnu.org/PR7961
12719 164. https://gcc.gnu.org/PR8071
12720 165. https://gcc.gnu.org/PR8127
12721 166. https://gcc.gnu.org/PR6745
12722 167. https://gcc.gnu.org/PR8096
12723 168. https://gcc.gnu.org/PR8127
12724 169. https://gcc.gnu.org/PR8218
12725 170. https://gcc.gnu.org/PR8287
12726 171. https://gcc.gnu.org/PR8347
12727 172. https://gcc.gnu.org/PR8348
12728 173. https://gcc.gnu.org/PR8391
12729 174. https://gcc.gnu.org/PR6627
12730 175. https://gcc.gnu.org/PR6631
12731 176. https://gcc.gnu.org/PR7102
12732 177. https://gcc.gnu.org/PR7120
12733 178. https://gcc.gnu.org/PR7209
12734 179. https://gcc.gnu.org/PR7515
12735 180. https://gcc.gnu.org/PR7814
12736 181. https://gcc.gnu.org/PR8467
12737 182. https://gcc.gnu.org/PR4890
12738 183. https://gcc.gnu.org/PR7357
12739 184. https://gcc.gnu.org/PR7358
12740 185. https://gcc.gnu.org/PR7602
12741 186. https://gcc.gnu.org/PR7862
12742 187. https://gcc.gnu.org/PR8190
12743 188. https://gcc.gnu.org/PR8524
12744 189. https://gcc.gnu.org/PR5351
12745 190. https://gcc.gnu.org/PR7591
12746 191. https://gcc.gnu.org/PR6845
12747 192. https://gcc.gnu.org/PR7034
12748 193. https://gcc.gnu.org/PR7124
12749 194. https://gcc.gnu.org/PR7174
12750 195. https://gcc.gnu.org/PR7134
12751 196. https://gcc.gnu.org/PR7375
12752 197. https://gcc.gnu.org/PR7390
12753 198. https://gcc.gnu.org/PR6890
12754 199. https://gcc.gnu.org/PR6981
12755 200. https://gcc.gnu.org/PR7242
12756 201. https://gcc.gnu.org/PR7396
12757 202. https://gcc.gnu.org/PR7630
12758 203. https://gcc.gnu.org/PR7693
12759 204. https://gcc.gnu.org/PR7723
12760 205. https://gcc.gnu.org/PR7951
12761 206. https://gcc.gnu.org/PR8146
12762 207. https://gcc.gnu.org/PR5967
12763 208. https://gcc.gnu.org/PR6984
12764 209. https://gcc.gnu.org/PR7114
12765 210. https://gcc.gnu.org/PR7130
12766 211. https://gcc.gnu.org/PR7133
12767 212. https://gcc.gnu.org/PR7380
12768 213. https://gcc.gnu.org/PR8252
12769 214. https://gcc.gnu.org/PR8451
12770 215. https://gcc.gnu.org/PR7250
12771 216. https://gcc.gnu.org/PR6668
12772 217. https://gcc.gnu.org/PR7151
12773 218. https://gcc.gnu.org/PR7335
12774 219. https://gcc.gnu.org/PR7842
12775 220. https://gcc.gnu.org/PR7856
12776 221. https://gcc.gnu.org/PR7967
12777 222. https://gcc.gnu.org/PR7374
12778 223. https://gcc.gnu.org/PR7370
12779 224. https://gcc.gnu.org/PR7409
12780 225. https://gcc.gnu.org/PR8232
12781 226. https://gcc.gnu.org/PR7623
12782 227. https://gcc.gnu.org/PR8314
12783 228. https://gcc.gnu.org/PR761
12784 229. https://gcc.gnu.org/PR5610
12785 230. https://gcc.gnu.org/PR7484
12786 231. https://gcc.gnu.org/PR7531
12787 232. https://gcc.gnu.org/PR8120
12788 233. https://gcc.gnu.org/PR7320
12789 234. https://gcc.gnu.org/PR7470
12790 235. https://gcc.gnu.org/PR6410
12791 236. https://gcc.gnu.org/PR6503
12792 237. https://gcc.gnu.org/PR6642
12793 238. https://gcc.gnu.org/PR7186
12794 239. https://gcc.gnu.org/PR7216
12795 240. https://gcc.gnu.org/PR7220
12796 241. https://gcc.gnu.org/PR7222
12797 242. https://gcc.gnu.org/PR7286
12798 243. https://gcc.gnu.org/PR7442
12799 244. https://gcc.gnu.org/PR7445
12800 245. https://gcc.gnu.org/PR7291
12801 246. https://gcc.gnu.org/onlinedocs/
12802 247. mailto:gcc-help@gcc.gnu.org
12803 248. mailto:gcc@gcc.gnu.org
12804 249. https://gcc.gnu.org/lists.html
12805 250. http://www.fsf.org/
12806 251. https://gcc.gnu.org/about.html
12807 252. http://validator.w3.org/check/referer
12808 ======================================================================
12809 http://gcc.gnu.org/gcc-3.1/index.html
12814 The [1]GNU project and the GCC developers are pleased to announce the
12815 release of GCC 3.1.1.
12817 The links below still apply to GCC 3.1.1.
12821 The [2]GNU project and the GCC developers are pleased to announce the
12822 release of GCC 3.1.
12824 GCC used to stand for the GNU C Compiler, but since the compiler
12825 supports several other languages aside from C, it now stands for the
12826 GNU Compiler Collection.
12828 A list of [3]successful builds is updated as new information becomes
12831 The GCC developers would like to thank the numerous people that have
12832 contributed [4]new features, improvements, bug fixes, and other changes
12833 as well as test results to GCC. This [5]amazing group of volunteers is
12834 what makes GCC successful.
12836 For additional information about GCC please refer to the [6]GCC project
12837 web site or contact the [7]GCC development mailing list.
12839 To obtain GCC please use [8]our mirror sites, or our CVS server.
12840 __________________________________________________________________
12843 For questions related to the use of GCC, please consult these web
12844 pages and the [9]GCC manuals. If that fails, the
12845 [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
12846 web pages and the development of GCC are welcome on our developer
12847 list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
12850 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
12851 distribution of this entire article is permitted in any medium,
12852 provided this notice is preserved.
12854 These pages are [14]maintained by the GCC team. Last modified
12859 1. http://www.gnu.org/
12860 2. http://www.gnu.org/
12861 3. http://gcc.gnu.org/gcc-3.1/buildstat.html
12862 4. http://gcc.gnu.org/gcc-3.1/changes.html
12863 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
12864 6. http://gcc.gnu.org/index.html
12865 7. mailto:gcc@gcc.gnu.org
12866 8. http://gcc.gnu.org/mirrors.html
12867 9. https://gcc.gnu.org/onlinedocs/
12868 10. mailto:gcc-help@gcc.gnu.org
12869 11. mailto:gcc@gcc.gnu.org
12870 12. https://gcc.gnu.org/lists.html
12871 13. http://www.fsf.org/
12872 14. https://gcc.gnu.org/about.html
12873 15. http://validator.w3.org/check/referer
12874 ======================================================================
12875 http://gcc.gnu.org/gcc-3.1/changes.html
12876 GCC 3.1 Release Series
12877 Changes, New Features, and Fixes
12879 Additional changes in GCC 3.1.1
12881 * A bug related to how structures and unions are returned has been
12882 fixed for powerpc-*-netbsd*.
12883 * An important bug in the implementation of -fprefetch-loop-arrays
12884 has been fixed. Previously the optimization prefetched random
12885 blocks of memory for most targets except for i386.
12886 * The Java compiler now compiles Java programs much faster and also
12887 works with parallel make.
12888 * Nested functions have been fixed for mips*-*-netbsd*.
12889 * Some missing floating point support routines have beed added for
12891 * This [1]message gives additional information about the bugs fixed
12896 * The -traditional C compiler option has been deprecated and will be
12897 removed in GCC 3.3. (It remains possible to preprocess non-C code
12898 with the traditional preprocessor.)
12899 * The default debugging format for most ELF platforms (including
12900 GNU/Linux and FreeBSD; notable exception is Solaris) has changed
12901 from stabs to DWARF2. This requires GDB 5.1.1 or later.
12903 General Optimizer Improvements
12905 * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat,
12906 and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure
12907 for profile driven optimizations.
12908 Options -fprofile-arcs and -fbranch-probabilities can now be used
12909 to improve speed of the generated code by profiling the actual
12910 program behaviour on typical runs. In the absence of profile info
12911 the compiler attempts to guess the profile statically.
12912 * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to
12913 monitor performance of the generated code.
12914 According to the SPECInt2000 results on an AMD Athlon CPU, the code
12915 generated by GCC 3.1 is 6% faster on the average (8.2% faster with
12916 profile feedback) compared to GCC 3.0. The code produced by GCC 3.0
12917 is about 2.1% faster compared to 2.95.3. Tests were done using the
12918 -O2 -march=athlon command-line options.
12919 * Alexandre Oliva, of Red Hat, has generalized the tree inlining
12920 infrastructure developed by CodeSourcery, LLC for the C++ front
12921 end, so that it is now used in the C front end too. Inlining
12922 functions as trees exposes them earlier to the compiler, giving it
12923 more opportunities for optimization.
12924 * Support for data prefetching instructions has been added to the GCC
12925 back end and several targets. A new __builtin_prefetch intrinsic is
12926 available to explicitly insert prefetch instructions and
12927 experimental support for loop array prefetching has been added (see
12928 -fprefetch-loop-array documentation).
12929 * Support for emitting debugging information for macros has been
12930 added for DWARF2. It is activated using -g3.
12932 New Languages and Language specific improvements
12936 * A few more [4]ISO C99 features.
12937 * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0.
12938 * The preprocessor's symbol table has been merged with the symbol
12939 table of the C, C++ and Objective-C front ends.
12940 * The preprocessor consumes less memory than the preprocessor in GCC
12941 3.0, often significantly so. On normal input files, it typically
12942 consumes less memory than pre-3.0 cccp-based GCC, too.
12946 * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std
12947 was a workaround to allow std compliant code to work with the
12948 non-std compliant libstdc++-v2. libstdc++-v3 is std compliant.
12949 * The C++ ABI has been fixed so that void (A::*)() const is mangled
12950 as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only
12951 affects pointer to cv-qualified member function types.
12952 * The C++ ABI has been changed to correctly handle this code:
12954 void operator delete[] (void *, size_t);
12957 struct B : public A {
12962 The amount of storage allocated for the array will be greater than
12963 it was in 3.0, in order to store the number of elements in the
12964 array, so that the correct size can be passed to operator delete[]
12965 when the array is deleted. Previously, the value passed to operator
12966 delete[] was unpredictable.
12967 This change will only affect code that declares a two-argument
12968 operator delete[] with a second parameter of type size_t in a base
12969 class, and does not override that definition in a derived class.
12970 * The C++ ABI has been changed so that:
12972 void operator delete[] (void *, size_t);
12973 void operator delete[] (void *);
12976 does not cause unnecessary storage to be allocated when an array of
12977 A objects is allocated.
12978 This change will only affect code that declares both of these forms
12979 of operator delete[], and declared the two-argument form before the
12981 * The C++ ABI has been changed so that when a parameter is passed by
12982 value, any cleanup for that parameter is performed in the caller,
12983 as specified by the ia64 C++ ABI, rather than the called function
12984 as before. As a result, classes with a non-trivial destructor but a
12985 trivial copy constructor will be passed and returned by invisible
12986 reference, rather than by bitwise copy as before.
12987 * G++ now supports the "named return value optimization": for code
12995 G++ will allocate a in the return value slot, so that the return
12996 becomes a no-op. For this to work, all return statements in the
12997 function must return the same variable.
12998 * Improvements to the C++ library are listed in [5]the libstdc++-v3
13003 * Annoying linker warnings (due to incorrect code being generated)
13005 * If a class method cannot be found, the compiler no longer issues a
13006 warning if a corresponding instance method exists in the root
13008 * Forward @protocol declarations have been fixed.
13009 * Loading of categories has been fixed in certain situations (GNU run
13011 * The class lookup in the run-time library has been rewritten so that
13012 class method dispatch is more than twice as fast as it used to be
13013 (GNU run time only).
13017 * libgcj now includes RMI, java.lang.ref.*, javax.naming, and
13019 * Property files and other system resources can be compiled into
13020 executables which use libgcj using the new gcj --resource feature.
13021 * libgcj has been ported to more platforms. In particular there is
13022 now a mostly-functional mingw32 (Windows) target port.
13023 * JNI and CNI invocation interfaces were implemented, so gcj-compiled
13024 Java code can now be called from a C/C++ application.
13025 * gcj can now use builtin functions for certain known methods, for
13027 * gcj can now automatically remove redundant array-store checks in
13029 * The --no-store-checks optimization option was added. This can be
13030 used to omit runtime store checks for code which is known not to
13031 throw ArrayStoreException
13032 * The following third party interface standards were added to libgcj:
13033 org.w3c.dom and org.xml.sax.
13034 * java.security has been merged with GNU Classpath. The new package
13035 is now JDK 1.2 compliant, and much more complete.
13036 * A bytecode verifier was added to the libgcj interpreter.
13037 * java.lang.Character was rewritten to comply with the Unicode 3.0
13038 standard, and improve performance.
13039 * Partial support for many more locales was added to libgcj.
13040 * Socket timeouts have been implemented.
13041 * libgcj has been merged into a single shared library. There are no
13042 longer separate shared libraries for the garbage collector and
13044 * Several performance improvements were made to gcj and libgcj:
13045 + Hash synchronization (thin locks)
13046 + A special allocation path for finalizer-free objects
13047 + Thread-local allocation
13048 + Parallel GC, and other GC tweaks
13052 Fortran improvements are listed in [6]the Fortran documentation.
13056 [7]Ada Core Technologies, Inc, has contributed its GNAT Ada 95 front
13057 end and associated tools. The GNAT compiler fully implements the Ada
13058 language as defined by the ISO/IEC 8652 standard.
13060 Please note that the integration of the Ada front end is still work in
13063 New Targets and Target Specific Improvements
13065 * Hans-Peter Nilsson has contributed a port to [8]MMIX, the CPU
13066 architecture used in new editions of Donald E. Knuth's The Art of
13067 Computer Programming.
13068 * [9]Axis Communications has contributed its port to the CRIS CPU
13069 architecture, used in the ETRAX system-on-a-chip series. See
13070 [10]Axis' developer site for technical information.
13071 * Alexandre Oliva, of Red Hat, has contributed a port to the
13072 [11]SuperH SH5 64-bit RISC microprocessor architecture, extending
13073 the existing SH port.
13074 * UltraSPARC is fully supported in 64-bit mode. The option -m64
13076 * For compatibility with the Sun compiler #pragma redefine_extname
13077 has been implemented on Solaris.
13078 * The x86 back end has had some noticeable work done to it.
13079 + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas
13080 Jaeger have contributed a port to the AMD x86-64 architecture.
13081 For more information on x86-64 see [12]http://www.x86-64.org.
13082 + The compiler now supports MMX, 3DNow!, SSE, and SSE2
13083 instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will
13084 enable the respective instruction sets. Intel C++ compatible
13085 MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics
13086 will be added in next major release.
13087 + Following those improvements, targets for Pentium MMX, K6-2,
13088 K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were
13089 added. Refer to the documentation on -march= and -mcpu=
13090 options for details.
13091 + For those targets that support it, -mfpmath=sse will cause the
13092 compiler to generate SSE/SSE2 instructions for floating point
13093 math instead of x87 instructions. Usually, this will lead to
13094 quicker code — especially on the Pentium 4. Note that only
13095 scalar floating point instructions are used and GCC does not
13096 exploit SIMD features yet.
13097 + Prefetch support has been added to the Pentium III, Pentium 4,
13098 K6-2, K6-3, and Athlon series.
13099 + Code generated for floating point to integer conversions has
13100 been improved leading to better performance of many 3D
13102 * The PowerPC back end has added 64-bit PowerPC GNU/Linux support.
13103 * C++ support for AIX has been improved.
13104 * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the
13105 PowerPC port supporting the AltiVec programming model (SIMD). The
13106 support, though presently useful, is experimental and is expected
13107 to stabilize for 3.2. The support is written to conform to
13108 Motorola's AltiVec specs. See -maltivec.
13112 Support for a number of older systems has been declared obsolete in GCC
13113 3.1. Unless there is activity to revive them, the next release of GCC
13114 will have their sources permanently removed.
13116 All configurations of the following processor architectures have been
13118 * MIL-STD-1750A, 1750a-*-*
13119 * AMD A29k, a29k-*-*
13120 * Convex, c*-convex-*
13121 * Clipper, clipper-*-*
13123 * Intel i860, i860-*-*
13124 * Sun picoJava, pj-*-* and pjl-*-*
13125 * Western Electric 32000, we32k-*-*
13127 Most configurations of the following processor architectures have been
13128 declared obsolete, but we are preserving a few systems which may have
13129 active developers. It is unlikely that the remaining systems will
13130 survive much longer unless we see definite signs of port activity.
13131 * Motorola 88000 except
13132 + Generic a.out, m88k-*-aout*
13133 + Generic SVR4, m88k-*-sysv4
13134 + OpenBSD, m88k-*-openbsd*
13136 + NetBSD, ns32k-*-netbsd*
13137 + OpenBSD, ns32k-*-openbsd*.
13139 + OpenBSD, romp-*-openbsd*.
13141 Finally, only some configurations of these processor architectures are
13144 + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka
13145 alpha*-*-osf[45], are still supported.)
13147 + RISCiX, arm-*-riscix*.
13149 + 386BSD, i?86-*-bsd*
13150 + Chorus, i?86-*-chorusos*
13151 + DG/UX, i?86-*-dgux*
13152 + FreeBSD 1.x, i?86-*-freebsd1.*
13153 + IBM AIX, i?86-*-aix*
13154 + ISC UNIX, i?86-*-isc*
13155 + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld*
13156 + NEXTstep, i?86-next-*
13157 + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose*
13158 + RTEMS/coff, i?86-*-rtemscoff*
13159 + RTEMS/go32, i?86-go32-rtems*
13160 + Sequent/BSD, i?86-sequent-bsd*
13161 + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and
13162 i?86-sequent-sysv3*
13163 + SunOS, i?86-*-sunos*
13165 + Altos, m68[k0]*-altos-*
13166 + Apollo, m68[k0]*-apollo-*
13167 + Apple A/UX, m68[k0]*-apple-*
13168 + Bull, m68[k0]*-bull-*
13169 + Convergent, m68[k0]*-convergent-*
13170 + Generic SVR3, m68[k0]*-*-sysv3*
13171 + ISI, m68[k0]*-isi-*
13172 + LynxOS, m68[k0]*-*-lynxos*
13173 + NEXT, m68[k0]*-next-*
13174 + RTEMS/coff, m68[k0]*-*-rtemscoff*
13175 + Sony, m68[k0]*-sony-*
13177 + DEC Ultrix, mips-*-ultrix* and mips-dec-*
13178 + Generic BSD, mips-*-bsd*
13179 + Generic System V, mips-*-sysv*
13180 + IRIX before version 5, mips-sgi-irix[1234]*
13181 + RiscOS, mips-*-riscos*
13182 + Sony, mips-sony-*
13183 + Tandem, mips-tandem-*
13185 + RTEMS/a.out, sparc-*-rtemsaout*.
13187 Documentation improvements
13189 * The old manual ("Using and Porting the GNU Compiler Collection")
13190 has been replaced by a users manual ("Using the GNU Compiler
13191 Collection") and a separate internals reference manual ("GNU
13192 Compiler Collection Internals").
13193 * More complete and much improved documentation about GCC's internal
13194 representation used by the C and C++ front ends.
13195 * Many cleanups and improvements in general.
13198 For questions related to the use of GCC, please consult these web
13199 pages and the [13]GCC manuals. If that fails, the
13200 [14]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13201 web pages and the development of GCC are welcome on our developer
13202 list at [15]gcc@gcc.gnu.org. All of [16]our lists have public
13205 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and
13206 distribution of this entire article is permitted in any medium,
13207 provided this notice is preserved.
13209 These pages are [18]maintained by the GCC team. Last modified
13214 1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html
13215 2. http://gcc.gnu.org/news/profiledriven.html
13216 3. http://gcc.gnu.org/benchmarks/
13217 4. http://gcc.gnu.org/c99status.html
13218 5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html
13219 6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html
13220 7. http://www.adacore.com/
13221 8. http://www-cs-faculty.stanford.edu/~knuth/mmix.html
13222 9. http://www.axis.com/
13223 10. http://developer.axis.com/
13224 11. http://www.superh.com/
13225 12. http://www.x86-64.org/
13226 13. https://gcc.gnu.org/onlinedocs/
13227 14. mailto:gcc-help@gcc.gnu.org
13228 15. mailto:gcc@gcc.gnu.org
13229 16. https://gcc.gnu.org/lists.html
13230 17. http://www.fsf.org/
13231 18. https://gcc.gnu.org/about.html
13232 19. http://validator.w3.org/check/referer
13233 ======================================================================
13234 http://gcc.gnu.org/gcc-3.0/index.html
13239 The [1]GNU project and the GCC developers are pleased to announce the
13240 release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0
13243 GCC used to stand for the GNU C Compiler, but since the compiler
13244 supports several other languages aside from C, it now stands for the
13245 GNU Compiler Collection.
13247 GCC 3.0.x has several new optimizations, new targets, new languages and
13248 many other new features, relative to GCC 2.95.x. See the [2]new
13249 features page for a more complete list.
13251 A list of [3]successful builds is updated as new information becomes
13254 The GCC developers would like to thank the numerous people that have
13255 contributed new features, test results, bug fixes, etc to GCC. This
13256 [4]amazing group of volunteers is what makes GCC successful.
13258 And finally, we can't in good conscience fail to mention some
13259 [5]caveats to using GCC 3.0.x.
13261 For additional information about GCC please refer to the [6]GCC project
13262 web site or contact the [7]GCC development mailing list.
13264 To obtain GCC please use [8]our mirror sites, or our CVS server.
13265 __________________________________________________________________
13267 Previous 3.0.x Releases
13269 December 20, 2001: GCC 3.0.3 has been released.
13270 October 25, 2001: GCC 3.0.2 has been released.
13271 August 20, 2001: GCC 3.0.1 has been released.
13272 June 18, 2001: GCC 3.0 has been released.
13275 For questions related to the use of GCC, please consult these web
13276 pages and the [9]GCC manuals. If that fails, the
13277 [10]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13278 web pages and the development of GCC are welcome on our developer
13279 list at [11]gcc@gcc.gnu.org. All of [12]our lists have public
13282 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and
13283 distribution of this entire article is permitted in any medium,
13284 provided this notice is preserved.
13286 These pages are [14]maintained by the GCC team. Last modified
13291 1. http://www.gnu.org/
13292 2. http://gcc.gnu.org/gcc-3.0/features.html
13293 3. http://gcc.gnu.org/gcc-3.0/buildstat.html
13294 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
13295 5. http://gcc.gnu.org/gcc-3.0/caveats.html
13296 6. http://gcc.gnu.org/index.html
13297 7. mailto:gcc@gcc.gnu.org
13298 8. http://gcc.gnu.org/mirrors.html
13299 9. https://gcc.gnu.org/onlinedocs/
13300 10. mailto:gcc-help@gcc.gnu.org
13301 11. mailto:gcc@gcc.gnu.org
13302 12. https://gcc.gnu.org/lists.html
13303 13. http://www.fsf.org/
13304 14. https://gcc.gnu.org/about.html
13305 15. http://validator.w3.org/check/referer
13306 ======================================================================
13307 http://gcc.gnu.org/gcc-3.0/features.html
13308 GCC 3.0 New Features
13310 Additional changes in GCC 3.0.4
13312 * GCC 3.0 now supports newer versions of the [1]NetBSD operating
13313 system, which use the ELF object file format, on x86 processors.
13314 * Correct debugging information is generated from functions that have
13315 lines from multiple files (e.g. yacc output).
13316 * A fix for whitespace handling in the -traditional preprocessor,
13317 which can affect Fortran.
13318 * Fixes to the exception handling runtime.
13319 * More fixes for bad code generation in C++.
13320 * A fix for shared library generation under AIX 4.3.
13321 * Documentation updates.
13322 * Port of GCC to Tensilica's Xtensa processor contributed.
13323 * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link).
13325 Additional changes in GCC 3.0.3
13327 * A fix to correct an accidental change to the PowerPC ABI.
13328 * Fixes for bad code generation on a variety of architectures.
13329 * Improvements to the debugging information generated for C++
13331 * Fixes for bad code generation in C++.
13332 * A fix to avoid crashes in the C++ demangler.
13333 * A fix to the C++ standard library to avoid buffer overflows.
13334 * Miscellaneous improvements for a variety of architectures.
13336 Additional changes in GCC 3.0.2
13338 * Fixes for bad code generation during loop unrolling.
13339 * Fixes for bad code generation by the sibling call optimization.
13340 * Minor improvements to x86 code generation.
13341 * Implementation of function descriptors in C++ vtables for IA64.
13342 * Numerous minor bug-fixes.
13344 Additional changes in GCC 3.0.1
13346 * C++ fixes for incorrect code-generation.
13347 * Improved cross-compiling support for the C++ standard library.
13348 * Fixes for some embedded targets that worked in GCC 2.95.3, but not
13350 * Fixes for various exception-handling bugs.
13351 * A port to the S/390 architecture.
13353 General Optimizer Improvements
13355 * [2]Basic block reordering pass.
13356 * New if-conversion pass with support for conditional (predicated)
13358 * New tail call and sibling call elimination optimizations.
13359 * New register renaming pass.
13360 * New (experimental) [3]static single assignment (SSA) representation
13362 * New dead-code elimination pass implemented using the SSA
13364 * [4]Global null pointer test elimination.
13365 * [5]Global code hoisting/unification.
13366 * More builtins and optimizations for stdio.h, string.h and old BSD
13367 functions, as well as for ISO C99 functions.
13368 * New builtin __builtin_expect for giving hints to the branch
13371 New Languages and Language specific improvements
13373 * The GNU Compiler for the Java(TM) language (GCJ) is now integrated
13374 and supported, including the run-time library containing most
13375 common non-GUI Java classes, a bytecode interpreter, and the Boehm
13376 conservative garbage collector. Many bugs have been fixed. GCJ can
13377 compile Java source or Java bytecodes to either native code or Java
13378 class files, and supports native methods written in either the
13379 standard JNI or the more efficient and convenient CNI.
13380 * Here is a [6]partial list of C++ improvements, both new features
13381 and those no longer supported.
13382 * New C++ ABI. On the IA-64 platform GCC is capable of
13383 inter-operating with other IA-64 compilers.
13384 * The new ABI also significantly reduces the size of symbol and debug
13386 * New [7]C++ support library and many C++ bug fixes, vastly improving
13387 our conformance to the ISO C++ standard.
13388 * New [8]inliner for C++.
13389 * Rewritten C preprocessor, integrated into the C, C++ and Objective
13390 C compilers, with very many improvements including ISO C99 support
13391 and [9]improvements to dependency generation.
13392 * Support for more [10]ISO C99 features.
13393 * Many improvements to support for checking calls to format functions
13394 such as printf and scanf, including support for ISO C99 format
13395 features, extensions from the Single Unix Specification and GNU
13396 libc 2.2, checking of strfmon formats and features to assist in
13397 auditing for format string security bugs.
13398 * New warnings for C code that may have undefined semantics because
13399 of violations of sequence point rules in the C standard (such as a
13400 = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall.
13401 * Additional warning option -Wfloat-equal.
13402 * Improvements to -Wtraditional.
13403 * Fortran improvements are listed in [11]the Fortran documentation.
13405 New Targets and Target Specific Improvements
13407 * New x86 back end, generating much improved code.
13408 * Support for a generic i386-elf target contributed.
13409 * New option to emit x86 assembly code using Intel style syntax
13411 * HPUX 11 support contributed.
13412 * Improved PowerPC code generation, including scheduled prologue and
13414 * Port of GCC to Intel's IA-64 processor contributed.
13415 * Port of GCC to Motorola's MCore 210 and 340 contributed.
13416 * New unified back-end for Arm, Thumb and StrongArm contributed.
13417 * Port of GCC to Intel's XScale processor contributed.
13418 * Port of GCC to Atmel's AVR microcontrollers contributed.
13419 * Port of GCC to Mitsubishi's D30V processor contributed.
13420 * Port of GCC to Matsushita's AM33 processor (a member of the MN10300
13421 processor family) contributed.
13422 * Port of GCC to Fujitsu's FR30 processor contributed.
13423 * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed.
13424 * Port of GCC to Sun's picoJava processor core contributed.
13426 Documentation improvements
13428 * Substantially rewritten and improved C preprocessor manual.
13429 * Many improvements to other documentation.
13430 * Manpages for gcc, cpp and gcov are now generated automatically from
13431 the master Texinfo manual, eliminating the problem of manpages
13432 being out of date. (The generated manpages are only extracts from
13433 the full manual, which is provided in Texinfo form, from which
13434 info, HTML, other formats and a printed manual can be generated.)
13435 * Generated info files are included in the release tarballs alongside
13436 their Texinfo sources, avoiding problems on some platforms with
13437 building makeinfo as part of the GCC distribution.
13439 Other significant improvements
13441 * Garbage collection used internally by the compiler for most memory
13442 allocation instead of obstacks.
13443 * Lengauer and Tarjan algorithm used for computing dominators in the
13444 CFG. This algorithm can be significantly faster and more space
13445 efficient than our older algorithm.
13446 * gccbug script provided to assist in submitting bug reports to our
13447 bug tracking system. (Bug reports previously submitted directly to
13448 our mailing lists, for which you received no bug tracking number,
13449 should be submitted again using gccbug if you can reproduce the
13450 problem with GCC 3.0.)
13451 * The internal libgcc library is [12]built as a shared library on
13452 systems that support it.
13453 * Extensive testsuite included with GCC, with many new tests. In
13454 addition to tests for GCC bugs that have been fixed, many tests
13455 have been added for language features, compiler warnings and
13457 * Additional language-independent warning options -Wpacked, -Wpadded,
13458 -Wunreachable-code and -Wdisabled-optimization.
13459 * Target-independent options -falign-functions, -falign-loops and
13462 Plus a great many bug fixes and almost all the [13]features found in
13466 For questions related to the use of GCC, please consult these web
13467 pages and the [14]GCC manuals. If that fails, the
13468 [15]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13469 web pages and the development of GCC are welcome on our developer
13470 list at [16]gcc@gcc.gnu.org. All of [17]our lists have public
13473 Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and
13474 distribution of this entire article is permitted in any medium,
13475 provided this notice is preserved.
13477 These pages are [19]maintained by the GCC team. Last modified
13482 1. http://www.netbsd.org/
13483 2. http://gcc.gnu.org/news/reorder.html
13484 3. http://gcc.gnu.org/news/ssa.html
13485 4. http://gcc.gnu.org/news/null.html
13486 5. http://gcc.gnu.org/news/unify.html
13487 6. http://gcc.gnu.org/gcc-3.0/c++features.html
13488 7. http://gcc.gnu.org/libstdc++/
13489 8. http://gcc.gnu.org/news/inlining.html
13490 9. http://gcc.gnu.org/news/dependencies.html
13491 10. http://gcc.gnu.org/c99status.html
13492 11. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
13493 12. http://gcc.gnu.org/gcc-3.0/libgcc.html
13494 13. http://gcc.gnu.org/gcc-2.95/features.html
13495 14. https://gcc.gnu.org/onlinedocs/
13496 15. mailto:gcc-help@gcc.gnu.org
13497 16. mailto:gcc@gcc.gnu.org
13498 17. https://gcc.gnu.org/lists.html
13499 18. http://www.fsf.org/
13500 19. https://gcc.gnu.org/about.html
13501 20. http://validator.w3.org/check/referer
13502 ======================================================================
13503 http://gcc.gnu.org/gcc-3.0/caveats.html
13506 * -fstrict-aliasing is now part of -O2 and higher optimization
13507 levels. This allows the compiler to assume the strictest aliasing
13508 rules applicable to the language being compiled. For C and C++,
13509 this activates optimizations based on the type of expressions. This
13510 optimization may thus break old, non-compliant code.
13511 * Enumerations are now properly promoted to int in function
13512 parameters and function returns. Normally this change is not
13513 visible, but when using -fshort-enums this is an ABI change.
13514 * The undocumented extension that allowed C programs to have a label
13515 at the end of a compound statement has been deprecated and may be
13516 removed in a future version. Programs that now generate a warning
13517 about this may be fixed by adding a null statement (a single
13518 semicolon) after the label.
13519 * The poorly documented extension that allowed string constants in C,
13520 C++ and Objective C to contain unescaped newlines has been
13521 deprecated and may be removed in a future version. Programs using
13522 this extension may be fixed in several ways: the bare newline may
13523 be replaced by \n, or preceded by \n\, or string concatenation may
13524 be used with the bare newline preceded by \n" and " placed at the
13525 start of the next line.
13526 * The Chill compiler is not included in GCC 3.0, because of the lack
13527 of a volunteer to convert it to use garbage collection.
13528 * Certain non-standard iostream methods from earlier versions of
13529 libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach,
13530 ostream::form, and istream::gets.
13531 * The new C++ ABI is not yet fully supported by current (as of
13532 2001-07-01) releases and development versions of GDB, or any
13533 earlier versions. There is a problem setting breakpoints by line
13534 number, and other related issues that have been fixed in GCC 3.0
13535 but not yet handled in GDB:
13536 [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
13539 For questions related to the use of GCC, please consult these web
13540 pages and the [2]GCC manuals. If that fails, the
13541 [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13542 web pages and the development of GCC are welcome on our developer
13543 list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
13545 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
13546 distribution of this entire article is permitted in any medium,
13547 provided this notice is preserved.
13549 These pages are [7]maintained by the GCC team. Last modified
13554 1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html
13555 2. https://gcc.gnu.org/onlinedocs/
13556 3. mailto:gcc-help@gcc.gnu.org
13557 4. mailto:gcc@gcc.gnu.org
13558 5. https://gcc.gnu.org/lists.html
13559 6. http://www.fsf.org/
13560 7. https://gcc.gnu.org/about.html
13561 8. http://validator.w3.org/check/referer
13562 ======================================================================
13563 http://gcc.gnu.org/gcc-2.95/index.html
13566 March 16, 2001: The GNU project and the GCC developers are pleased to
13567 announce the release of GCC version 2.95.3.
13581 July 31, 1999. This is the first release of GCC since the April
13582 1999 GCC/EGCS reunification and includes nearly a year's worth
13583 of new development and bugfixes.
13585 References and Acknowledgements
13587 GCC used to stand for the GNU C Compiler, but since the compiler
13588 supports several other languages aside from C, it now stands for the
13589 GNU Compiler Collection.
13591 The whole suite has been extensively [1]regression tested and
13592 [2]package tested. It should be reliable and suitable for widespread
13595 The compiler has several new optimizations, new targets, new languages
13596 and other new features. See the [3]new features page for a more
13597 complete list of new features found in the GCC 2.95 releases.
13599 The sources include installation instructions in both HTML and
13600 plaintext forms in the install directory in the distribution. However,
13601 the most up to date installation instructions and [4]build/test status
13602 are on the web pages. We will update those pages as new information
13605 The GCC developers would like to thank the numerous people that have
13606 contributed new features, test results, bugfixes, etc to GCC. This
13607 [5]amazing group of volunteers is what makes GCC successful.
13609 And finally, we can't in good conscience fail to mention some
13610 [6]caveats to using GCC 2.95.
13612 Download GCC 2.95 from one of our many [7]mirror sites.
13614 For additional information about GCC please see the [8]GCC project web
13615 server or contact the [9]GCC development mailing list.
13618 For questions related to the use of GCC, please consult these web
13619 pages and the [10]GCC manuals. If that fails, the
13620 [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13621 web pages and the development of GCC are welcome on our developer
13622 list at [12]gcc@gcc.gnu.org. All of [13]our lists have public
13625 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and
13626 distribution of this entire article is permitted in any medium,
13627 provided this notice is preserved.
13629 These pages are [15]maintained by the GCC team. Last modified
13634 1. http://gcc.gnu.org/gcc-2.95/regress.html
13635 2. http://gcc.gnu.org/gcc-2.95/othertest.html
13636 3. http://gcc.gnu.org/gcc-2.95/features.html
13637 4. http://gcc.gnu.org/gcc-2.95/buildstat.html
13638 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
13639 6. http://gcc.gnu.org/gcc-2.95/caveats.html
13640 7. http://gcc.gnu.org/mirrors.html
13641 8. http://gcc.gnu.org/index.html
13642 9. mailto:gcc@gcc.gnu.org
13643 10. https://gcc.gnu.org/onlinedocs/
13644 11. mailto:gcc-help@gcc.gnu.org
13645 12. mailto:gcc@gcc.gnu.org
13646 13. https://gcc.gnu.org/lists.html
13647 14. http://www.fsf.org/
13648 15. https://gcc.gnu.org/about.html
13649 16. http://validator.w3.org/check/referer
13650 ======================================================================
13651 http://gcc.gnu.org/gcc-2.95/features.html
13652 GCC 2.95 New Features
13654 * General Optimizer Improvements:
13655 + [1]Localized register spilling to improve speed and code
13656 density especially on small register class machines.
13657 + [2]Global CSE using lazy code motion algorithms.
13658 + [3]Improved global constant/copy propagation.
13659 + [4]Improved control flow graph analysis and manipulation.
13660 + [5]Local dead store elimination.
13661 + [6]Memory Load hoisting/store sinking in loops.
13662 + [7]Type based alias analysis is enabled by default. Note this
13663 feature will expose bugs in the Linux kernel. Please refer to
13664 the FAQ (as shipped with GCC 2.95) for additional information
13666 + Major revamp of GIV detection, combination and simplification
13667 to improve loop performance.
13668 + Major improvements to register allocation and reloading.
13669 * New Languages and Language specific improvements
13670 + [8]Many C++ improvements.
13671 + [9]Many Fortran improvements.
13672 + [10]Java front-end has been integrated. [11]runtime library is
13673 available separately.
13674 + [12]ISO C99 support
13675 + [13]Chill front-end and runtime has been integrated.
13676 + Boehm garbage collector support in libobjc.
13677 + More support for various pragmas which appear in vendor
13679 * New Targets and Target Specific Improvements
13680 + [14]SPARC backend rewrite.
13681 + -mschedule=8000 will optimize code for PA8000 class
13682 processors; -mpa-risc-2-0 will generate code for PA2.0
13684 + Various micro-optimizations for the ia32 port. K6
13686 + Compiler will attempt to align doubles in the stack on the
13688 + Alpha EV6 support
13690 + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403.
13691 -mcpu=e603e was added to do -mcpu=603e and -msoft-float.
13696 + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix,
13698 + vxWorks targets include support for vxWorks threads
13699 + StrongARM 110 and ARM9 support added. ARM Scheduling
13700 parameters rewritten.
13701 + Various changes to the MIPS port to avoid assembler macros,
13702 which in turn improves performance
13703 + Various performance improvements to the i960 port.
13704 + Major rewrite of ns32k port
13705 * Other significant improvements
13706 + [15]Ability to dump cfg information and display it using vcg.
13707 + The new faster scheme for fixing vendor header files is
13708 enabled by default.
13709 + Experimental internationalization support.
13710 + multibyte character support
13711 + Some compile-time speedups for pathological problems
13712 + Better support for complex types
13713 * Plus the usual mountain of bugfixes
13714 * Core compiler is based on the gcc2 development tree from Sept 30,
13715 1998, so we have all of the [16]features found in GCC 2.8.
13717 Additional Changes in GCC 2.95.1
13719 * Generic bugfixes and improvements
13720 + Various documentation fixes related to the GCC/EGCS merger.
13721 + Fix memory management bug which could lead to spurious aborts,
13722 core dumps or random parsing errors in the compiler.
13723 + Fix a couple bugs in the dwarf1 and dwarf2 debug record
13725 + Fix infinite loop in the CSE optimizer.
13726 + Avoid undefined behavior in compiler FP emulation code
13727 + Fix install problem when prefix is overridden on the make
13729 + Fix problem with unwanted installation of assert.h on some
13731 + Fix problem with finding the wrong assembler in a single tree
13733 + Avoid increasing the known alignment of a register that is
13734 already known to be a pointer.
13735 * Platform specific bugfixes and improvements
13736 + Codegen bugfix for prologue/epilogue for cpu32 target.
13737 + Fix long long code generation bug for the Coldfire target.
13738 + Fix various aborts in the SH compiler.
13739 + Fix bugs in libgcc support library for the SH.
13740 + Fix alpha ev6 code generation bug.
13741 + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on
13743 + Fix -fpic code generation bug for rs6000/ppc svr4 targets.
13744 + Fix varargs/stdarg code generation bug for rs6000/ppc svr4
13746 + Fix weak symbol handling for rs6000/ppc svr4 targets.
13747 + Fix various problems with 64bit code generation for the
13749 + Fix codegen bug which caused tetex to be mis-compiled on the
13751 + Fix compiler abort in new cfg code exposed by x86 port.
13752 + Fix out of range array reference in code convert flat
13753 registers to the x87 stacked FP register file.
13754 + Fix minor vxworks configuration bug.
13755 + Fix return type of bsearch for SunOS 4.x.
13756 * Language & Runtime specific fixes.
13757 + The G++ signature extension has been deprecated. It will be
13758 removed in the next major release of G++. Use of signatures
13759 will result in a warning from the compiler.
13760 + Several bugs relating to templates and namespaces were fixed.
13761 + A bug that caused crashes when combining templates with -g on
13762 DWARF1 platforms was fixed.
13763 + Pointers-to-members, virtual functions, and multiple
13764 inheritance should now work together correctly.
13765 + Some code-generation bugs relating to function try blocks were
13767 + G++ is a little bit more lenient with certain archaic
13768 constructs than in GCC 2.95.
13769 + Fix to prevent shared library version #s from bring truncated
13771 + Fix missing std:: in the libstdc++ library.
13772 + Fix stream locking problems in libio.
13773 + Fix problem in java compiler driver.
13775 Additional Changes in GCC 2.95.2
13777 The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While
13778 the optimizations performed by -fstrict-aliasing are valid according to
13779 the C and C++ standards, the optimization have caused some problems,
13780 particularly with old non-conforming code.
13782 The GCC developers are experimenting with ways to warn users about code
13783 which violates the C/C++ standards, but those warnings are not ready
13784 for widespread use at this time. Rather than wait for those warnings
13785 the GCC developers have chosen to disable -fstrict-aliasing by default
13786 for the GCC 2.95.2 release.
13788 We strongly encourage developers to find and fix code which violates
13789 the C/C++ standards as -fstrict-aliasing may be enabled by default in
13790 future releases. Use the option -fstrict-aliasing to re-enable these
13792 * Generic bugfixes and improvements
13793 + Fix incorrectly optimized memory reference in global common
13794 subexpression elimination (GCSE) optimization pass.
13795 + Fix code generation bug in regmove.c in which it could
13796 incorrectly change a "const" value.
13797 + Fix bug in optimization of conditionals involving volatile
13799 + Avoid over-allocation of stack space for some procedures.
13800 + Fixed bug in the compiler which caused incorrect optimization
13801 of an obscure series of bit manipulations, shifts and
13803 + Fixed register allocator bug which caused teTeX to be
13804 mis-compiled on SPARC targets.
13805 + Avoid incorrect optimization of degenerate case statements for
13806 certain targets such as the ARM.
13807 + Fix out of range memory reference in the jump optimizer.
13808 + Avoid dereferencing null pointer in fix-header.
13809 + Fix test for GCC specific features so that it is possible to
13810 bootstrap with gcc-2.6.2 and older versions of GCC.
13811 + Fix typo in scheduler which could potentially cause out of
13812 range memory accesses.
13813 + Avoid incorrect loop reversal which caused incorrect code for
13814 certain loops on PowerPC targets.
13815 + Avoid incorrect optimization of switch statements on certain
13816 targets (for example the ARM).
13817 * Platform specific bugfixes and improvements
13818 + Work around bug in Sun V5.0 compilers which caused bootstrap
13819 comparison failures on SPARC targets.
13820 + Fix SPARC backend bug which caused aborts in final.c.
13821 + Fix sparc-hal-solaris2* configuration fragments.
13822 + Fix bug in sparc block profiling.
13823 + Fix obscure code generation bug for the PARISC targets.
13824 + Define __STDC_EXT__ for HPUX configurations.
13825 + Various POWERPC64 code generation bugfixes.
13826 + Fix abort for PPC targets using ELF (ex GNU/Linux).
13827 + Fix collect2 problems for AIX targets.
13828 + Correct handling of .file directive for PPC targets.
13829 + Fix bug in fix_trunc x86 patterns.
13830 + Fix x86 port to correctly pop the FP stack for functions that
13831 return structures in memory.
13832 + Fix minor bug in strlen x86 pattern.
13833 + Use stabs debugging instead of dwarf1 for x86-solaris targets.
13834 + Fix template repository code to handle leading underscore in
13836 + Fix weak/weak alias support for OpenBSD.
13837 + GNU/Linux for the ARM has C++ compatible include files.
13838 * Language & Runtime specific fixes.
13839 + Fix handling of constructor attribute in the C front-end which
13840 caused problems building the Chill runtime library on some
13842 + Fix minor problem merging type qualifiers in the C front-end.
13843 + Fix aliasing bug for pointers and references (C/C++).
13844 + Fix incorrect "non-constant initializer bug" when -traditional
13845 or -fwritable-strings is enabled.
13846 + Fix build error for Chill front-end on SunOS.
13847 + Do not complain about duplicate instantiations when using
13849 + Fix array bounds handling in C++ front-end which caused
13850 problems with dwarf debugging information in some
13852 + Fix minor namespace problem.
13853 + Fix problem linking java programs.
13855 Additional Changes in GCC 2.95.3
13857 * Generic bugfixes and improvements
13858 + Fix numerous problems that caused incorrect optimization in
13859 the register reloading code.
13860 + Fix numerous problems that caused incorrect optimization in
13861 the loop optimizer.
13862 + Fix aborts in the functions build_insn_chain and scan_loops
13863 under some circumstances.
13864 + Fix an alias analysis bug.
13865 + Fix an infinite compilation bug in the combiner.
13866 + A few problems with complex number support have been fixed.
13867 + It is no longer possible for gcc to act as a fork bomb when
13868 installed incorrectly.
13869 + The -fpack-struct option should be recognized now.
13870 + Fixed a bug that caused incorrect code to be generated due to
13871 a lost stack adjustment.
13872 * Platform specific bugfixes and improvements
13873 + Support building ARM toolchains hosted on Windows.
13874 + Fix attribute calculations in ARM toolchains.
13875 + arm-linux support has been improved.
13876 + Fix a PIC failure on sparc targets.
13877 + On ix86 targets, the regparm attribute should now work
13879 + Several updates for the h8300 port.
13880 + Fix problem building libio with glibc 2.2.
13883 For questions related to the use of GCC, please consult these web
13884 pages and the [17]GCC manuals. If that fails, the
13885 [18]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13886 web pages and the development of GCC are welcome on our developer
13887 list at [19]gcc@gcc.gnu.org. All of [20]our lists have public
13890 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and
13891 distribution of this entire article is permitted in any medium,
13892 provided this notice is preserved.
13894 These pages are [22]maintained by the GCC team. Last modified
13899 1. http://gcc.gnu.org/news/spill.html
13900 2. http://gcc.gnu.org/news/lcm.html
13901 3. http://gcc.gnu.org/news/cprop.html
13902 4. http://gcc.gnu.org/news/cfg.html
13903 5. http://gcc.gnu.org/news/dse.html
13904 6. http://gcc.gnu.org/news/hoist.html
13905 7. http://gcc.gnu.org/news/alias.html
13906 8. http://gcc.gnu.org/gcc-2.95/c++features.html
13907 9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html
13908 10. http://gcc.gnu.org/java/gcj-announce.txt
13909 11. http://gcc.gnu.org/news/javaannounce.html
13910 12. http://gcc.gnu.org/c99status.html
13911 13. http://gcc.gnu.org/news/chill.html
13912 14. http://gcc.gnu.org/news/sparc.html
13913 15. http://gcc.gnu.org/news/egcs-vcg.html
13914 16. http://gcc.gnu.org/egcs-1.0/features-2.8.html
13915 17. https://gcc.gnu.org/onlinedocs/
13916 18. mailto:gcc-help@gcc.gnu.org
13917 19. mailto:gcc@gcc.gnu.org
13918 20. https://gcc.gnu.org/lists.html
13919 21. http://www.fsf.org/
13920 22. https://gcc.gnu.org/about.html
13921 23. http://validator.w3.org/check/referer
13922 ======================================================================
13923 http://gcc.gnu.org/gcc-2.95/caveats.html
13926 * GCC 2.95 will issue an error for invalid asm statements that had
13927 been silently accepted by earlier versions of the compiler. This is
13928 particularly noticeable when compiling older versions of the Linux
13929 kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95)
13930 for more information on this issue.
13931 * GCC 2.95 implements type based alias analysis to disambiguate
13932 memory references. Some programs, particularly the Linux kernel
13933 violate ANSI/ISO aliasing rules and therefore may not operate
13934 correctly when compiled with GCC 2.95. Please refer to the FAQ (as
13935 shipped with GCC 2.95) for more information on this issue.
13936 * GCC 2.95 has a known bug in its handling of complex variables for
13937 64bit targets. Instead of silently generating incorrect code, GCC
13938 2.95 will issue a fatal error for situations it can not handle.
13939 This primarily affects the Fortran community as Fortran makes more
13940 use of complex variables than C or C++.
13941 * GCC 2.95 has an integrated libstdc++, but does not have an
13942 integrated libg++. Furthermore old libg++ releases will not work
13943 with GCC 2.95. You can retrieve a recent copy of libg++ from the
13945 Note most C++ programs only need libstdc++.
13946 * Exception handling may not work with shared libraries, particularly
13947 on alphas, hppas, rs6000/powerpc and mips based platforms.
13948 Exception handling is known to work on x86 GNU/Linux platforms with
13950 * In general, GCC 2.95 is more rigorous about rejecting invalid C++
13951 code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0,
13952 or EGCS 1.1. As a result it may be necessary to fix C++ code before
13953 it will compile with GCC 2.95.
13954 * G++ is also converting toward the ISO C++ standard; as a result
13955 code which was previously valid (and thus accepted by other
13956 compilers and older versions of g++) may no longer be accepted. The
13957 flag -fpermissive may allow some non-conforming code to compile
13959 * GCC 2.95 compiled C++ code is not binary compatible with EGCS
13960 1.1.x, EGCS 1.0.x or GCC 2.8.x.
13961 * GCC 2.95 does not have changes from the GCC 2.8 tree that were made
13962 between Sept 30, 1998 and April 30, 1999 (the official end of the
13963 GCC 2.8 project). Future GCC releases will include all the changes
13964 from the defunct GCC 2.8 sources.
13967 For questions related to the use of GCC, please consult these web
13968 pages and the [2]GCC manuals. If that fails, the
13969 [3]gcc-help@gcc.gnu.org mailing list might help. Comments on these
13970 web pages and the development of GCC are welcome on our developer
13971 list at [4]gcc@gcc.gnu.org. All of [5]our lists have public archives.
13973 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and
13974 distribution of this entire article is permitted in any medium,
13975 provided this notice is preserved.
13977 These pages are [7]maintained by the GCC team. Last modified
13982 1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz
13983 2. https://gcc.gnu.org/onlinedocs/
13984 3. mailto:gcc-help@gcc.gnu.org
13985 4. mailto:gcc@gcc.gnu.org
13986 5. https://gcc.gnu.org/lists.html
13987 6. http://www.fsf.org/
13988 7. https://gcc.gnu.org/about.html
13989 8. http://validator.w3.org/check/referer
13990 ======================================================================
13991 http://gcc.gnu.org/egcs-1.1/index.html
13994 September 3, 1998: We are pleased to announce the release of EGCS 1.1.
13995 December 1, 1998: We are pleased to announce the release of EGCS 1.1.1.
13996 March 15, 1999: We are pleased to announce the release of EGCS 1.1.2.
13998 EGCS is a free software project to further the development of the GNU
13999 compilers using an open development environment.
14001 EGCS 1.1 is a major new release of the EGCS compiler system. It has
14002 been [1]extensively tested and is believed to be stable and suitable
14003 for widespread use.
14005 EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8
14006 development sources; it contains all of the new features found in GCC
14007 2.8.1 as well as all new development from GCC up to June 6, 1998.
14009 EGCS 1.1 also contains many improvements and features not found in GCC
14010 or in older versions of EGCS:
14011 * Global common subexpression elimination and global constant/copy
14012 propagation (aka [2]gcse)
14013 * Ongoing improvements to the [3]alias analysis support to allow for
14014 better optimizations throughout the compiler.
14015 * Vastly improved [4]C++ compiler and integrated C++ runtime
14017 * Fixes for the /tmp symlink race security problems.
14018 * New targets including mips16, arm-thumb and 64 bit PowerPC.
14019 * Improvements to GNU Fortran (g77) compiler and runtime library made
14020 since g77 version 0.5.23.
14022 See the [5]new features page for a more complete list of new features
14023 found in EGCS 1.1 releases.
14025 EGCS 1.1.1 is a minor update to fix several serious problems in EGCS
14027 * General improvements and fixes
14028 + Avoid some stack overflows when compiling large functions.
14029 + Avoid incorrect loop invariant code motions.
14030 + Fix some core dumps on Linux kernel code.
14031 + Bring back the imake -Di386 and friends fix from EGCS 1.0.2.
14032 + Fix code generation problem in gcse.
14033 + Various documentation related fixes.
14034 * g++/libstdc++ improvements and fixes
14035 + MT safe EH fix for setjmp/longjmp based exception handling.
14036 + Fix a few bad interactions between optimization and exception
14038 + Fixes for demangling of template names starting with "__".
14039 + Fix a bug that would fail to run destructors in some cases
14041 + Fix 'new' of classes with virtual bases.
14042 + Fix crash building Qt on the Alpha.
14043 + Fix failure compiling WIFEXITED macro on GNU/Linux.
14044 + Fix some -frepo failures.
14045 * g77 and libf2c improvements and fixes
14046 + Various documentation fixes.
14047 + Avoid compiler crash on RAND intrinsic.
14048 + Fix minor bugs in makefiles exposed by BSD make programs.
14049 + Define _XOPEN_SOURCE for libI77 build to avoid potential
14050 problems on some 64-bit systems.
14051 + Fix problem with implicit endfile on rewind.
14052 + Fix spurious recursive I/O errors.
14053 * platform specific improvements and fixes
14054 + Match all versions of UnixWare7.
14055 + Do not assume x86 SVR4 or UnixWare targets can handle stabs.
14056 + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion
14057 from unsigned ints to double precision floats.
14058 + Fix ARM ABI issue with NetBSD.
14059 + Fix a few arm code generation bugs.
14060 + Fixincludes will fix additional broken SCO OpenServer header
14062 + Fix a m68k backend bug which caused invalid offsets in reg+d
14064 + Fix problems with 64bit AIX 4.3 support.
14065 + Fix handling of long longs for varargs/stdarg functions on the
14067 + Minor fixes to CPP predefines for Windows.
14068 + Fix code generation problems with gpr<->fpr copies for 64bit
14070 + Fix a few coldfire code generation bugs.
14071 + Fix some more header file problems on SunOS 4.x.
14072 + Fix assert.h handling for RTEMS.
14073 + Fix Windows handling of TREE_SYMBOL_REFERENCED.
14074 + Fix x86 compiler abort in reg-stack pass.
14075 + Fix cygwin/windows problem with section attributes.
14076 + Fix Alpha code generation problem exposed by SMP Linux
14078 + Fix typo in m68k 32->64bit integer conversion.
14079 + Make sure target libraries build with -fPIC for PPC & Alpha
14082 EGCS 1.1.2 is a minor update to fix several serious problems in EGCS
14084 * General improvements and fixes
14085 + Fix bug in loop optimizer which caused the SPARC (and
14086 potentially other) ports to segfault.
14087 + Fix infinite recursion in alias analysis and combiner code.
14088 + Fix bug in regclass preferencing.
14089 + Fix incorrect loop reversal which caused incorrect code to be
14090 generated for several targets.
14091 + Fix return value for builtin memcpy.
14092 + Reduce compile time for certain loops which exposed quadratic
14093 behavior in the loop optimizer.
14094 + Fix bug which caused volatile memory to be written multiple
14095 times when only one write was needed/desired.
14096 + Fix compiler abort in caller-save.c
14097 + Fix combiner bug which caused incorrect code generation for
14098 certain division by constant operations.
14099 + Fix incorrect code generation due to a bug in range check
14101 + Fix incorrect code generation due to mis-handling of clobbered
14103 + Fix compiler abort/segfault due to incorrect register
14104 splitting when unrolling loops.
14105 + Fix code generation involving autoincremented addresses with
14107 + Work around bug in the scheduler which caused qt to be
14108 mis-compiled on some platforms.
14109 + Fix code generation problems with -fshort-enums.
14110 + Tighten security for temporary files.
14111 + Improve compile time for codes which make heavy use of
14112 overloaded functions.
14113 + Fix multiply defined constructor/destructor symbol problems.
14114 + Avoid setting bogus RPATH environment variable during
14116 + Avoid GNU-make dependencies in the texinfo subdir.
14117 + Install CPP wrapper script in $(prefix)/bin if --enable-cpp.
14118 --enable-cpp=<dirname> can be used to specify an additional
14119 install directory for the cpp wrapper script.
14120 + Fix CSE bug which caused incorrect label-label refs to appear
14122 + Avoid linking in EH routines from libgcc if they are not
14124 + Avoid obscure bug in aliasing code.
14125 + Fix bug in weak symbol handling.
14126 * Platform-specific improvements and fixes
14127 + Fix detection of PPro/PII on Unixware 7.
14128 + Fix compiler segfault when building spec99 and other programs
14130 + Fix code-generation bugs for integer and floating point
14131 conditional move instructions on the PPro/PII.
14132 + Use fixincludes to fix byteorder problems on i?86-*-sysv.
14133 + Fix build failure for the arc port.
14134 + Fix floating point format configuration for i?86-gnu port.
14135 + Fix problems with hppa1.0-hp-hpux10.20 configuration when
14136 threads are enabled.
14137 + Fix coldfire code generation bugs.
14138 + Fix "unrecognized insn" problems for Alpha and PPC ports.
14139 + Fix h8/300 code generation problem with floating point values
14141 + Fix unrecognized insn problems for the m68k port.
14142 + Fix namespace-pollution problem for the x86 port.
14143 + Fix problems with old assembler on x86 NeXT systems.
14144 + Fix PIC code-generation problems for the SPARC port.
14145 + Fix minor bug with LONG_CALLS in PowerPC SVR4 support.
14146 + Fix minor ISO namespace violation in Alpha varargs/stdarg
14148 + Fix incorrect "braf" instruction usage for the SH port.
14149 + Fix minor bug in va-sh which prevented its use with -ansi.
14150 + Fix problems recognizing and supporting FreeBSD.
14151 + Handle OpenBSD systems correctly.
14152 + Minor fixincludes fix for Digital UNIX 4.0B.
14153 + Fix problems with ctors/dtors in SCO shared libraries.
14154 + Abort instead of generating incorrect code for PPro/PII
14155 floating point conditional moves.
14156 + Avoid multiply defined symbols on GNU/Linux systems using
14158 + Fix abort in alpha compiler.
14159 * Fortran-specific fixes
14160 + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year
14161 is in the documented, non-Y2K-compliant range of 0-99, instead
14162 of being returned as 100 in the year 2000.
14163 + Fix the `Date_and_Time' intrinsic (in libg2c) to return the
14164 milliseconds value properly in Values(8).
14165 + Fix the `LStat' intrinsic (in libg2c) to return device-ID
14166 information properly in SArray(7).
14168 Each release includes installation instructions in both HTML and
14169 plaintext forms (see the INSTALL directory in the toplevel directory of
14170 the distribution). However, we also keep the most up to date
14171 installation instructions and [6]build/test status on our web page. We
14172 will update those pages as new information becomes available.
14174 The EGCS project would like to thank the numerous people that have
14175 contributed new features, test results, bugfixes, etc. This [7]amazing
14176 group of volunteers is what makes EGCS successful.
14178 And finally, we can't in good conscience fail to mention some
14179 [8]caveats to using EGCS 1.1.
14181 Download EGCS from egcs.cygnus.com (USA California).
14183 The EGCS 1.1 release is also available on many mirror sites.
14184 [9]Goto mirror list to find a closer site.
14187 For questions related to the use of GCC, please consult these web
14188 pages and the [10]GCC manuals. If that fails, the
14189 [11]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14190 web pages and the development of GCC are welcome on our developer
14191 list at [12]gcc@gcc.gnu.org. All of [13]our lists have public
14194 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and
14195 distribution of this entire article is permitted in any medium,
14196 provided this notice is preserved.
14198 These pages are [15]maintained by the GCC team. Last modified
14203 1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html
14204 2. http://gcc.gnu.org/news/gcse.html
14205 3. http://gcc.gnu.org/news/alias.html
14206 4. http://gcc.gnu.org/egcs-1.1/c++features.html
14207 5. http://gcc.gnu.org/egcs-1.1/features.html
14208 6. http://gcc.gnu.org/egcs-1.1/buildstat.html
14209 7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html
14210 8. http://gcc.gnu.org/egcs-1.1/caveats.html
14211 9. http://gcc.gnu.org/mirrors.html
14212 10. https://gcc.gnu.org/onlinedocs/
14213 11. mailto:gcc-help@gcc.gnu.org
14214 12. mailto:gcc@gcc.gnu.org
14215 13. https://gcc.gnu.org/lists.html
14216 14. http://www.fsf.org/
14217 15. https://gcc.gnu.org/about.html
14218 16. http://validator.w3.org/check/referer
14219 ======================================================================
14220 http://gcc.gnu.org/egcs-1.1/features.html
14221 EGCS 1.1 new features
14223 * Integrated GNU Fortran (g77) compiler and runtime library with
14224 improvements, based on g77 version 0.5.23.
14225 * Vast improvements in the C++ compiler; so many they have [1]page of
14227 * Compiler implements [2]global common subexpression elimination and
14228 global copy/constant propagation.
14229 * More major improvements in the [3]alias analysis code.
14230 * More major improvements in the exception handling code to improve
14231 performance, lower static overhead and provide the infrastructure
14232 for future improvements.
14233 * The infamous /tmp symlink race security problems have been fixed.
14234 * The regmove optimization pass has been nearly completely rewritten
14235 to improve performance of generated code.
14236 * The compiler now recomputes register usage information before local
14237 register allocation. By providing more accurate information to the
14238 priority based allocator, we get better register allocation.
14239 * The register reloading phase of the compiler optimizes spill code
14240 much better than in previous releases.
14241 * Some bad interactions between the register allocator and
14242 instruction scheduler have been fixed, resulting in much better
14243 code for certain programs. Additionally, we have tuned the
14244 scheduler in various ways to improve performance of generated code
14245 for some architectures.
14246 * The compiler's branch shortening algorithms have been significantly
14247 improved to work better on targets which align jump targets.
14248 * The compiler now supports -Os to prefer optimizing for code space
14249 over optimizing for code speed.
14250 * The compiler will now totally eliminate library calls which compute
14251 constant values. This primarily helps targets with no integer
14252 div/mul support and targets without floating point support.
14253 * The compiler now supports an extensive "--help" option.
14254 * cpplib has been greatly improved and may be suitable for limited
14256 * Memory footprint for the compiler has been significantly reduced
14257 for some pathological cases.
14258 * The time to build EGCS has been improved for certain targets
14259 (particularly the alpha and mips platforms).
14260 * Many infrastructure improvements throughout the compiler, plus the
14261 usual mountain of bugfixes and minor improvements.
14262 * Target dependent improvements:
14263 + SPARC port now includes V8 plus and V9 support as well as
14264 performance tuning for Ultra class machines. The SPARC port
14265 now uses the Haifa scheduler.
14266 + Alpha port has been tuned for the EV6 processor and has an
14267 optimized expansion of memcpy/bzero. The Alpha port now uses
14268 the Haifa scheduler.
14269 + RS6000/PowerPC: support for the Power64 architecture and AIX
14270 4.3. The RS6000/PowerPC port now uses the Haifa scheduler.
14271 + x86: Alignment of static store data and jump targets is per
14272 Intel recommendations now. Various improvements throughout the
14273 x86 port to improve performance on Pentium processors
14274 (including improved epilogue sequences for Pentium chips and
14275 backend improvements which should help register allocation on
14276 all x86 variants. Conditional move support has been fixed and
14277 enabled for PPro processors. The x86 port also better supports
14278 64bit operations now. Unixware 7, a System V Release 5 target,
14279 is now supported and SCO OpenServer targets can support GAS.
14280 + MIPS has improved multiply/multiply-add support and now
14281 includes mips16 ISA support.
14282 + M68k has many micro-optimizations and Coldfire fixes.
14283 * Core compiler is based on the GCC development tree from June 9,
14284 1998, so we have all of the [4]features found in GCC 2.8.
14287 For questions related to the use of GCC, please consult these web
14288 pages and the [5]GCC manuals. If that fails, the
14289 [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14290 web pages and the development of GCC are welcome on our developer
14291 list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
14293 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
14294 distribution of this entire article is permitted in any medium,
14295 provided this notice is preserved.
14297 These pages are [10]maintained by the GCC team. Last modified
14302 1. http://gcc.gnu.org/egcs-1.1/c++features.html
14303 2. http://gcc.gnu.org/news/gcse.html
14304 3. http://gcc.gnu.org/news/alias.html
14305 4. http://gcc.gnu.org/egcs-1.0/features-2.8.html
14306 5. https://gcc.gnu.org/onlinedocs/
14307 6. mailto:gcc-help@gcc.gnu.org
14308 7. mailto:gcc@gcc.gnu.org
14309 8. https://gcc.gnu.org/lists.html
14310 9. http://www.fsf.org/
14311 10. https://gcc.gnu.org/about.html
14312 11. http://validator.w3.org/check/referer
14313 ======================================================================
14314 http://gcc.gnu.org/egcs-1.1/caveats.html
14317 * EGCS has an integrated libstdc++, but does not have an integrated
14318 libg++. Furthermore old libg++ releases will not work with EGCS; HJ
14319 Lu has made a libg++-2.8.1.2 snapshot available which may work with
14321 Note most C++ programs only need libstdc++.
14322 * Exception handling may not work with shared libraries, particularly
14323 on alphas, hppas, rs6000/powerpc and mips based platforms.
14324 Exception handling is known to work on x86-linux platforms with
14326 * Some versions of the Linux kernel have bugs which prevent them from
14327 being compiled or from running when compiled by EGCS. See the FAQ
14328 (as shipped with EGCS 1.1) for additional information.
14329 * In general, EGCS is more rigorous about rejecting invalid C++ code
14330 or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As
14331 a result it may be necessary to fix C++ code before it will compile
14333 * G++ is also converting toward the ISO C++ standard; as a result
14334 code which was previously valid (and thus accepted by other
14335 compilers and older versions of g++) may no longer be accepted.
14336 * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x
14337 or GCC 2.8.x due to changes necessary to support thread safe
14338 exception handling.
14341 For questions related to the use of GCC, please consult these web
14342 pages and the [1]GCC manuals. If that fails, the
14343 [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14344 web pages and the development of GCC are welcome on our developer
14345 list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
14347 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
14348 distribution of this entire article is permitted in any medium,
14349 provided this notice is preserved.
14351 These pages are [6]maintained by the GCC team. Last modified
14356 1. https://gcc.gnu.org/onlinedocs/
14357 2. mailto:gcc-help@gcc.gnu.org
14358 3. mailto:gcc@gcc.gnu.org
14359 4. https://gcc.gnu.org/lists.html
14360 5. http://www.fsf.org/
14361 6. https://gcc.gnu.org/about.html
14362 7. http://validator.w3.org/check/referer
14363 ======================================================================
14364 http://gcc.gnu.org/egcs-1.0/index.html
14367 December 3, 1997: We are pleased to announce the release of EGCS 1.0.
14368 January 6, 1998: We are pleased to announce the release of EGCS 1.0.1.
14369 March 16, 1998: We are pleased to announce the release of EGCS 1.0.2.
14370 May 15, 1998 We are pleased to announce the release of EGCS 1.0.3.
14372 EGCS is a collaborative effort involving several groups of hackers
14373 using an open development model to accelerate development and testing
14374 of GNU compilers and runtime libraries.
14376 An important goal of EGCS is to allow wide scale testing of
14377 experimental features and optimizations; therefore, EGCS contains some
14378 features and optimizations which are still under development. However,
14379 EGCS has been carefully tested and should be comparable in quality to
14382 EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8
14383 development sources; it contains nearly all of the new features found
14386 EGCS 1.0 also contains many improvements and features not found in GCC
14387 2.7 and even the GCC 2.8 series (which was released after the original
14389 * Integrated C++ runtime libraries, including support for most major
14391 * The integrated libstdc++ library includes a verbatim copy of SGI's
14393 * Integrated GNU Fortran compiler.
14394 * New instruction scheduler.
14395 * New alias analysis code.
14397 See the [1]new features page for a more complete list of new features.
14399 EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few
14400 critical bugs and add support for Red Hat 5.0 Linux. Changes since the
14402 * Add support for Red Hat 5.0 Linux and better support for Linux
14403 systems using glibc2.
14404 Many programs failed to link when compiled with EGCS 1.0 on Red Hat
14405 5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should
14406 fix these problems.
14407 * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception
14408 handling interfaces.
14409 To avoid future compatibility problems, we strongly urge anyone who
14410 is planning on distributing shared libraries that contain C++ code
14411 to upgrade to EGCS 1.0.1 first.
14412 Soon after EGCS 1.0 was released, the GCC developers made some
14413 incompatible changes in libgcc's exception handling interfaces.
14414 These changes were needed to solve problems on some platforms. This
14415 means that GCC 2.8.0, when released, will not be seamlessly
14416 compatible with shared libraries built by EGCS 1.0. The reason is
14417 that the libgcc.a in GCC 2.8.0 will not contain a function needed
14418 by the old interface.
14419 The result of this is that there may be compatibility problems with
14420 shared libraries built by EGCS 1.0 when used with GCC 2.8.0.
14421 With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface,
14422 and libgcc.a has the support routines for both the old and the new
14423 interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed,
14424 and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed).
14425 The maintainers of GCC 2.x have decided against including seamless
14426 support for the old interface in 2.8.0, since it was never
14427 "official", so to avoid future compatibility problems we recommend
14428 against distributing any shared libraries built by EGCS 1.0 that
14429 contain C++ code (upgrade to 1.0.1 and use that).
14430 * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends.
14431 The x86 changes fix code generation errors exposed when building
14432 glibc2 and the usual GNU/Linux dynamic linker (ld.so).
14433 The hppa change fixes a compiler abort when configured for use with
14435 The MIPS changes fix problems with the definition of LONG_MAX on
14436 newer systems, allow for command line selection of the target ABI,
14437 and fix one code generation problem.
14438 The rs6000/ppc change fixes some problems with passing structures
14439 to varargs/stdarg functions.
14440 * A few machine independent bugfixes, mostly to fix code generation
14441 errors when building Linux kernels or glibc.
14442 * Fix a few critical exception handling and template bugs in the C++
14444 * Fix Fortran namelist bug on alphas.
14445 * Fix build problems on x86-solaris systems.
14447 EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several
14448 serious problems in EGCS 1.0.1.
14449 * General improvements and fixes
14450 + Memory consumption significantly reduced, especially for
14451 templates and inline functions.
14452 + Fix various problems with glibc2.1.
14453 + Fix loop optimization bug exposed by rs6000/ppc port.
14454 + Fix to avoid potential code generation problems in jump.c.
14455 + Fix some undefined symbol problems in dwarf1 debug support.
14456 * g++/libstdc++ improvements and fixes
14457 + libstdc++ in the EGCS release has been updated and should be
14458 link compatible with libstdc++-2.8.
14459 + Various fixes in libio/libstdc++ to work better on GNU/Linux
14461 + Fix problems with duplicate symbols on systems that do not
14462 support weak symbols.
14463 + Memory corruption bug and undefined symbols in bastring have
14465 + Various exception handling fixes.
14466 + Fix compiler abort for very long thunk names.
14467 * g77 improvements and fixes
14468 + Fix compiler crash for omitted bound in Fortran CASE
14470 + Add missing entries to g77 lang-options.
14471 + Fix problem with -fpedantic in the g77 compiler.
14472 + Fix "backspace" problem with g77 on alphas.
14473 + Fix x86 backend problem with Fortran literals and -fpic.
14474 + Fix some of the problems with negative subscripts for g77 on
14476 + Fixes for Fortran builds on cygwin32/mingw32.
14477 * platform specific improvements and fixes
14478 + Fix long double problems on x86 (exposed by glibc).
14479 + x86 ports define i386 again to keep imake happy.
14480 + Fix exception handling support on NetBSD ports.
14481 + Several changes to collect2 to fix many problems with AIX.
14482 + Define __ELF__ for GNU/Linux on rs6000.
14483 + Fix -mcall-linux problem on GNU/Linux on rs6000.
14484 + Fix stdarg/vararg problem for GNU/Linux on rs6000.
14485 + Allow autoconf to select a proper install problem on AIX 3.1.
14486 + m68k port support includes -mcpu32 option as well as cpu32
14488 + Fix stdarg bug for irix6.
14489 + Allow EGCS to build on irix5 without the gnu assembler.
14490 + Fix problem with static linking on sco5.
14491 + Fix bootstrap on sco5 with native compiler.
14492 + Fix for abort building newlib on H8 target.
14493 + Fix fixincludes handling of math.h on SunOS.
14494 + Minor fix for Motorola 3300 m68k systems.
14496 EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few
14497 problems reported by Red Hat for builds of Red Hat 5.1.
14498 * Generic bugfixes:
14499 + Fix a typo in the libio library which resulted in incorrect
14500 behavior of istream::get.
14501 + Fix the Fortran negative array index problem.
14502 + Fix a major problem with the ObjC runtime thread support
14504 + Reduce memory consumption of the Haifa scheduler.
14505 * Target specific bugfixes:
14506 + Fix one x86 floating point code generation bug exposed by
14508 + Fix one x86 internal compiler error exposed by glibc2 builds.
14509 + Fix profiling bugs on the Alpha.
14510 + Fix ImageMagick & emacs 20.2 build problems on the Alpha.
14511 + Fix rs6000/ppc bug when converting values from integer types
14512 to floating point types.
14514 The EGCS 1.0 releases include installation instructions in both HTML
14515 and plaintext forms (see the INSTALL directory in the toplevel
14516 directory of the distribution). However, we also keep the most up to
14517 date installation instructions and [2]build/test status on our web
14518 page. We will update those pages as new information becomes available.
14520 And, we can't in good conscience fail to mention some [3]caveats to
14523 Update: Big thanks to Stanford for providing a high speed link for
14524 downloading EGCS (go.cygnus.com)!
14526 Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com
14527 (USA California -- High speed link provided by Stanford).
14529 The EGCS 1.0 release is also available many mirror sites.
14530 [4]Goto mirror list to find a closer site
14532 We'd like to thank the numerous people that have contributed new
14533 features, test results, bugfixes, etc. Unfortunately, they're far too
14534 numerous to mention by name.
14537 For questions related to the use of GCC, please consult these web
14538 pages and the [5]GCC manuals. If that fails, the
14539 [6]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14540 web pages and the development of GCC are welcome on our developer
14541 list at [7]gcc@gcc.gnu.org. All of [8]our lists have public archives.
14543 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and
14544 distribution of this entire article is permitted in any medium,
14545 provided this notice is preserved.
14547 These pages are [10]maintained by the GCC team. Last modified
14552 1. http://gcc.gnu.org/egcs-1.0/features.html
14553 2. http://gcc.gnu.org/egcs-1.0/buildstat.html
14554 3. http://gcc.gnu.org/egcs-1.0/caveats.html
14555 4. http://gcc.gnu.org/mirrors.html
14556 5. https://gcc.gnu.org/onlinedocs/
14557 6. mailto:gcc-help@gcc.gnu.org
14558 7. mailto:gcc@gcc.gnu.org
14559 8. https://gcc.gnu.org/lists.html
14560 9. http://www.fsf.org/
14561 10. https://gcc.gnu.org/about.html
14562 11. http://validator.w3.org/check/referer
14563 ======================================================================
14564 http://gcc.gnu.org/egcs-1.0/features.html
14567 * Core compiler is based on the gcc2 development tree from Aug 2,
14568 1997, so we have most of the [1]features found in GCC 2.8.
14569 * Integrated GNU Fortran compiler based on g77-0.5.22-19970929.
14570 * Vast improvements in the C++ compiler; so many they have [2]page of
14572 * Integrated C++ runtime libraries, including support for most major
14574 * New instruction scheduler from IBM Haifa which includes support for
14575 function wide instruction scheduling as well as superscalar
14577 * Significantly improved alias analysis code.
14578 * Improved register allocation for two address machines.
14579 * Significant code generation improvements for Fortran code on
14581 * Various optimizations from the g77 project as well as improved loop
14583 * Dwarf2 debug format support for some targets.
14584 * egcs libstdc++ includes the SGI STL implementation without changes.
14585 * As a result of these and other changes, egcs libstc++ is not binary
14586 compatible with previous releases of libstdc++.
14587 * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO
14588 Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and
14589 1.1), Support for RTEMS on several embedded targets, Support for
14590 arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and
14591 MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc.
14592 * Integrated testsuites for gcc, g++, g77, libstdc++ and libio.
14593 * RS6000/PowerPC ports generate code which can run on all
14594 RS6000/PowerPC variants by default.
14595 * -mcpu= and -march= switches for the x86 port to allow better
14596 control over how the x86 port generates code.
14597 * Includes the template repository patch (aka repo patch); note the
14598 new template code makes repo obsolete for ELF systems using gnu-ld
14600 * Plus the usual assortment of bugfixes and improvements.
14603 For questions related to the use of GCC, please consult these web
14604 pages and the [3]GCC manuals. If that fails, the
14605 [4]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14606 web pages and the development of GCC are welcome on our developer
14607 list at [5]gcc@gcc.gnu.org. All of [6]our lists have public archives.
14609 Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and
14610 distribution of this entire article is permitted in any medium,
14611 provided this notice is preserved.
14613 These pages are [8]maintained by the GCC team. Last modified
14618 1. http://gcc.gnu.org/egcs-1.0/features-2.8.html
14619 2. http://gcc.gnu.org/egcs-1.0/c++features.html
14620 3. https://gcc.gnu.org/onlinedocs/
14621 4. mailto:gcc-help@gcc.gnu.org
14622 5. mailto:gcc@gcc.gnu.org
14623 6. https://gcc.gnu.org/lists.html
14624 7. http://www.fsf.org/
14625 8. https://gcc.gnu.org/about.html
14626 9. http://validator.w3.org/check/referer
14627 ======================================================================
14628 http://gcc.gnu.org/egcs-1.0/caveats.html
14631 * EGCS has an integrated libstdc++, but does not have an integrated
14632 libg++. Furthermore old libg++ releases will not work with egc; HJ
14633 Lu has made a libg++-2.8.1.2 available which may work with EGCS.
14634 Note most C++ programs only need libstdc++.
14635 * Note that using -pedantic or -Wreturn-type can cause an explosion
14636 in the amount of memory needed for template-heavy C++ code, such as
14637 code that uses STL. Also note that -Wall includes -Wreturn-type, so
14638 if you use -Wall you will need to specify -Wno-return-type to turn
14640 * Exception handling may not work with shared libraries, particularly
14641 on alphas, hppas, and mips based platforms. Exception handling is
14642 known to work on x86-linux platforms with shared libraries.
14643 * Some versions of the Linux kernel have bugs which prevent them from
14644 being compiled or from running when compiled by EGCS. See the FAQ
14645 (as shipped with EGCS 1.0) for additional information.
14646 * In general, EGCS is more rigorous about rejecting invalid C++ code
14647 or deprecated C++ constructs than G++ 2.7. As a result it may be
14648 necessary to fix C++ code before it will compile with EGCS.
14649 * G++ is also aggressively tracking the C++ standard; as a result
14650 code which was previously valid (and thus accepted by other
14651 compilers and older versions of G++) may no longer be accepted.
14652 * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS
14653 1.0.x and later releases should work with Red Hat Linux 5.0.
14656 For questions related to the use of GCC, please consult these web
14657 pages and the [1]GCC manuals. If that fails, the
14658 [2]gcc-help@gcc.gnu.org mailing list might help. Comments on these
14659 web pages and the development of GCC are welcome on our developer
14660 list at [3]gcc@gcc.gnu.org. All of [4]our lists have public archives.
14662 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and
14663 distribution of this entire article is permitted in any medium,
14664 provided this notice is preserved.
14666 These pages are [6]maintained by the GCC team. Last modified
14671 1. https://gcc.gnu.org/onlinedocs/
14672 2. mailto:gcc-help@gcc.gnu.org
14673 3. mailto:gcc@gcc.gnu.org
14674 4. https://gcc.gnu.org/lists.html
14675 5. http://www.fsf.org/
14676 6. https://gcc.gnu.org/about.html
14677 7. http://validator.w3.org/check/referer
14678 ======================================================================