NEWS revision 1.5
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, 5 see ONEWS. 6 7 ====================================================================== 8 http://gcc.gnu.org/gcc-5/index.html 9 10 GCC 5 Release Series 11 12 December 4, 2015 13 14 The [1]GNU project and the GCC developers are pleased to announce the 15 release of GCC 5.3. 16 17 This release is a bug-fix release, containing fixes for regressions in 18 GCC 5.2 relative to previous releases of GCC. 19 20 Release History 21 22 GCC 5.3 23 December 4, 2015 ([2]changes, [3]documentation) 24 25 GCC 5.2 26 July 16, 2015 ([4]changes, [5]documentation) 27 28 GCC 5.1 29 April 22, 2015 ([6]changes, [7]documentation) 30 31 References and Acknowledgements 32 33 GCC used to stand for the GNU C Compiler, but since the compiler 34 supports several other languages aside from C, it now stands for the 35 GNU Compiler Collection. 36 37 A list of [8]successful builds is updated as new information becomes 38 available. 39 40 The GCC developers would like to thank the numerous people that have 41 contributed new features, improvements, bug fixes, and other changes as 42 well as test results to GCC. This [9]amazing group of volunteers is 43 what makes GCC successful. 44 45 For additional information about GCC please refer to the [10]GCC 46 project web site or contact the [11]GCC development mailing list. 47 48 To obtain GCC please use [12]our mirror sites or [13]our SVN server. 49 50 51 For questions related to the use of GCC, please consult these web 52 pages and the [14]GCC manuals. If that fails, the 53 [15]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 54 web pages and the development of GCC are welcome on our developer 55 list at [16]gcc (a] gcc.gnu.org. All of [17]our lists have public 56 archives. 57 58 Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and 59 distribution of this entire article is permitted in any medium, 60 provided this notice is preserved. 61 62 These pages are [19]maintained by the GCC team. Last modified 63 2015-12-04[20]. 64 65 References 66 67 1. http://www.gnu.org/ 68 2. http://gcc.gnu.org/gcc-5/changes.html 69 3. http://gcc.gnu.org/onlinedocs/5.3.0/ 70 4. http://gcc.gnu.org/gcc-5/changes.html 71 5. http://gcc.gnu.org/onlinedocs/5.2.0/ 72 6. http://gcc.gnu.org/gcc-5/changes.html 73 7. http://gcc.gnu.org/onlinedocs/5.1.0/ 74 8. http://gcc.gnu.org/gcc-5/buildstat.html 75 9. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 76 10. http://gcc.gnu.org/index.html 77 11. mailto:gcc (a] gcc.gnu.org 78 12. http://gcc.gnu.org/mirrors.html 79 13. http://gcc.gnu.org/svn.html 80 14. https://gcc.gnu.org/onlinedocs/ 81 15. mailto:gcc-help (a] gcc.gnu.org 82 16. mailto:gcc (a] gcc.gnu.org 83 17. https://gcc.gnu.org/lists.html 84 18. http://www.fsf.org/ 85 19. https://gcc.gnu.org/about.html 86 20. http://validator.w3.org/check/referer 87 ====================================================================== 88 http://gcc.gnu.org/gcc-5/changes.html 89 90 GCC 5 Release Series 91 Changes, New Features, and Fixes 92 93 Caveats 94 95 * The default mode for C is now -std=gnu11 instead of -std=gnu89. 96 * The C++ runtime library (libstdc++) uses a new ABI by default (see 97 [1]below). 98 * The Graphite framework for loop optimizations no longer requires 99 the CLooG library, only ISL version 0.14 (recommended) or 0.12.2. 100 The installation manual contains more information about 101 requirements to build GCC. 102 * The non-standard C++0x type traits has_trivial_default_constructor, 103 has_trivial_copy_constructor and has_trivial_copy_assign have been 104 deprecated and will be removed in a future version. The standard 105 C++11 traits is_trivially_default_constructible, 106 is_trivially_copy_constructible and is_trivially_copy_assignable 107 should be used instead. 108 * On AVR, support has been added for the devices 109 ATtiny4/5/9/10/20/40. This requires Binutils 2.25 or newer. 110 * The AVR port uses a new scheme to describe supported devices: For 111 each supported device the compiler provides a device-specific 112 [2]spec file. If the compiler is used together with AVR-LibC, this 113 requires at least GCC 5.2 and a version of AVR-LibC which 114 implements [3]feature #44574. 115 116 General Optimizer Improvements 117 118 * Inter-procedural optimization improvements: 119 + An Identical Code Folding (ICF) pass (controlled via 120 -fipa-icf) has been added. Compared to the identical code 121 folding performed by the Gold linker this pass does not 122 require function sections. It also performs merging before 123 inlining, so inter-procedural optimizations are aware of the 124 code re-use. On the other hand not all unifications performed 125 by a linker are doable by GCC which must honor aliasing 126 information. During link-time optimization of Firefox, this 127 pass unifies about 31000 functions, that is 14% overall. 128 + The devirtualization pass was significantly improved by adding 129 better support for speculative devirtualization and dynamic 130 type detection. About 50% of virtual calls in Firefox are now 131 speculatively devirtualized during link-time optimization. 132 + A new comdat localization pass allows the linker to eliminate 133 more dead code in presence of C++ inline functions. 134 + Virtual tables are now optimized. Local aliases are used to 135 reduce dynamic linking time of C++ virtual tables on ELF 136 targets and data alignment has been reduced to limit data 137 segment bloat. 138 + A new -fno-semantic-interposition option can be used to 139 improve code quality of shared libraries where interposition 140 of exported symbols is not allowed. 141 + Write-only variables are now detected and optimized out. 142 + With profile feedback the function inliner can now bypass 143 --param inline-insns-auto and --param inline-insns-single 144 limits for hot calls. 145 + The IPA reference pass was significantly sped up making it 146 feasible to enable -fipa-reference with -fprofile-generate. 147 This also solves a bottleneck seen when building Chromium with 148 link-time optimization. 149 + The symbol table and call-graph API was reworked to C++ and 150 simplified. 151 + The interprocedural propagation of constants now also 152 propagates alignments of pointer parameters. This for example 153 means that the vectorizer often does not need to generate loop 154 prologues and epilogues to make up for potential 155 misalignments. 156 * Link-time optimization improvements: 157 + One Definition Rule based merging of C++ types has been 158 implemented. Type merging enables better devirtualization and 159 alias analysis. Streaming extra information needed to merge 160 types adds about 2-6% of memory size and object size increase. 161 This can be controlled by -flto-odr-type-merging. 162 + Command-line optimization and target options are now streamed 163 on a per-function basis and honored by the link-time 164 optimizer. This change makes link-time optimization a more 165 transparent replacement of per-file optimizations. It is now 166 possible to build projects that require different optimization 167 settings for different translation units (such as -ffast-math, 168 -mavx, or -finline). Contrary to earlier GCC releases, the 169 optimization and target options passed on the link command 170 line are ignored. 171 Note that this applies only to those command-line options that 172 can be passed to optimize and target attributes. Command-line 173 options affecting global code generation (such as -fpic), 174 warnings (such as -Wodr), optimizations affecting the way 175 static variables are optimized (such as -fcommon), debug 176 output (such as -g), and --param parameters can be applied 177 only to the whole link-time optimization unit. In these cases, 178 it is recommended to consistently use the same options at both 179 compile time and link time. 180 + GCC bootstrap now uses slim LTO object files. 181 + Memory usage and link times were improved. Tree merging was 182 sped up, memory usage of GIMPLE declarations and types was 183 reduced, and, support for on-demand streaming of variable 184 constructors was added. 185 * Feedback directed optimization improvements: 186 + A new auto-FDO mode uses profiles collected by low overhead 187 profiling tools (perf) instead of more expensive program 188 instrumentation (via -fprofile-generate). SPEC2006 benchmarks 189 on x86-64 improve by 4.7% with auto-FDO and by 7.3% with 190 traditional feedback directed optimization. 191 + Profile precision was improved in presence of C++ inline and 192 extern inline functions. 193 + The new gcov-tool utility allows manipulating profiles. 194 + Profiles are now more tolerant to source file changes (this 195 can be controlled by --param profile-func-internal-id). 196 * Register allocation improvements: 197 + A new local register allocator (LRA) sub-pass, controlled by 198 -flra-remat, implements control-flow sensitive global register 199 rematerialization. Instead of spilling and restoring a 200 register value, it is recalculated if it is profitable. The 201 sub-pass improved SPEC2000 generated code by 1% and 0.5% 202 correspondingly on ARM and x86-64. 203 + Reuse of the PIC hard register, instead of using a fixed 204 register, was implemented on x86/x86-64 targets. This improves 205 generated PIC code performance as more hard registers can be 206 used. Shared libraries can significantly benefit from this 207 optimization. Currently it is switched on only for x86/x86-64 208 targets. As RA infrastructure is already implemented for PIC 209 register reuse, other targets might follow this in the future. 210 + A simple form of inter-procedural RA was implemented. When it 211 is known that a called function does not use caller-saved 212 registers, save/restore code is not generated around the call 213 for such registers. This optimization can be controlled by 214 -fipa-ra 215 + LRA is now much more effective at generating spills of general 216 registers into vector registers instead of memory on 217 architectures (e.g., modern Intel processors) where this is 218 profitable. 219 * UndefinedBehaviorSanitizer gained a few new sanitization options: 220 + -fsanitize=float-divide-by-zero: detect floating-point 221 division by zero; 222 + -fsanitize=float-cast-overflow: check that the result of 223 floating-point type to integer conversions do not overflow; 224 + -fsanitize=bounds: enable instrumentation of array bounds and 225 detect out-of-bounds accesses; 226 + -fsanitize=alignment: enable alignment checking, detect 227 various misaligned objects; 228 + -fsanitize=object-size: enable object size checking, detect 229 various out-of-bounds accesses. 230 + -fsanitize=vptr: enable checking of C++ member function calls, 231 member accesses and some conversions between pointers to base 232 and derived classes, detect if the referenced object does not 233 have the correct dynamic type. 234 * Pointer Bounds Checker, a bounds violation detector, has been added 235 and can be enabled via -fcheck-pointer-bounds. Memory accesses are 236 instrumented with run-time checks of used pointers against their 237 bounds to detect pointer bounds violations (overflows). The Pointer 238 Bounds Checker is available on x86/x86-64 GNU/Linux targets with a 239 new ISA extension Intel MPX support. See the Pointer Bounds Checker 240 [4]Wiki page for more details. 241 242 New Languages and Language specific improvements 243 244 * [5]OpenMP 4.0 specification offloading features are now supported 245 by the C, C++, and Fortran compilers. Generic changes: 246 + Infrastructure (suitable for any vendor). 247 + Testsuite which covers offloading from the [6]OpenMP 4.0 248 Examples document. 249 Specific for upcoming Intel Xeon Phi products: 250 + Run-time library. 251 + Card emulator. 252 * GCC 5 includes a preliminary implementation of the OpenACC 2.0a 253 specification. OpenACC is intended for programming accelerator 254 devices such as GPUs. See [7]the OpenACC wiki page for more 255 information. 256 257 C family 258 259 * The default setting of the -fdiagnostics-color= command-line option 260 is now [8]configurable when building GCC using configuration option 261 --with-diagnostics-color=. The possible values are: never, always, 262 auto and auto-if-env. The new default auto uses color only when the 263 standard error is a terminal. The default in GCC 4.9 was 264 auto-if-env, which is equivalent to auto if there is a non-empty 265 GCC_COLORS environment variable, and never otherwise. As in GCC 266 4.9, an empty GCC_COLORS variable in the environment will always 267 disable colors, no matter what the default is or what command-line 268 options are used. 269 * A new command-line option -Wswitch-bool has been added for the C 270 and C++ compilers, which warns whenever a switch statement has an 271 index of boolean type. 272 * A new command-line option -Wlogical-not-parentheses has been added 273 for the C and C++ compilers, which warns about "logical not" used 274 on the left hand side operand of a comparison. 275 * A new command-line option -Wsizeof-array-argument has been added 276 for the C and C++ compilers, which warns when the sizeof operator 277 is applied to a parameter that has been declared as an array in a 278 function definition. 279 * A new command-line option -Wbool-compare has been added for the C 280 and C++ compilers, which warns about boolean expressions compared 281 with an integer value different from true/false. 282 * Full support for [9]Cilk Plus has been added to the GCC compiler. 283 Cilk Plus is an extension to the C and C++ languages to support 284 data and task parallelism. 285 * A new attribute no_reorder prevents reordering of selected symbols 286 against other such symbols or inline assembler. This enables to 287 link-time optimize the Linux kernel without having to resort to 288 -fno-toplevel-reorder that disables several optimizations. 289 * New preprocessor constructs, __has_include and __has_include_next, 290 to test the availability of headers have been added. 291 This demonstrates a way to include the header <optional> only if it 292 is available: 293 294 #ifdef __has_include 295 # if __has_include(<optional>) 296 # include <optional> 297 # define have_optional 1 298 # elif __has_include(<experimental/optional>) 299 # include <experimental/optional> 300 # define have_optional 1 301 # define experimental_optional 302 # else 303 # define have_optional 0 304 # endif 305 #endif 306 307 The header search paths for __has_include and __has_include_next 308 are equivalent to those of the standard directive #include and the 309 extension #include_next respectively. 310 * A new built-in function-like macro to determine the existence of an 311 attribute, __has_attribute, has been added. The equivalent built-in 312 macro __has_cpp_attribute was added to C++ to support 313 [10]Feature-testing recommendations for C++. The macro 314 __has_attribute is added to all C-like languages as an extension: 315 316 int 317 #ifdef __has_attribute 318 # if __has_attribute(__noinline__) 319 __attribute__((__noinline__)) 320 # endif 321 #endif 322 foo(int x); 323 324 If an attribute exists, a nonzero constant integer is returned. For 325 standardized C++ attributes a date is returned, otherwise the 326 constant returned is 1. Both __has_attribute and 327 __has_cpp_attribute will add underscores to an attribute name if 328 necessary to resolve the name. For C++11 and onwards the attribute 329 may be scoped. 330 * A new set of built-in functions for arithmetics with overflow 331 checking has been added: __builtin_add_overflow, 332 __builtin_sub_overflow and __builtin_mul_overflow and for 333 compatibility with clang also other variants. These builtins have 334 two integral arguments (which don't need to have the same type), 335 the arguments are extended to infinite precision signed type, +, - 336 or * is performed on those, and the result is stored in an integer 337 variable pointed to by the last argument. If the stored value is 338 equal to the infinite precision result, the built-in functions 339 return false, otherwise true. The type of the integer variable that 340 will hold the result can be different from the types of the first 341 two arguments. The following snippet demonstrates how this can be 342 used in computing the size for the calloc function: 343 344 void * 345 calloc (size_t x, size_t y) 346 { 347 size_t sz; 348 if (__builtin_mul_overflow (x, y, &sz)) 349 return NULL; 350 void *ret = malloc (sz); 351 if (ret) memset (res, 0, sz); 352 return ret; 353 } 354 355 On e.g. i?86 or x86-64 the above will result in a mul instruction 356 followed by a jump on overflow. 357 * The option -fextended-identifiers is now enabled by default for 358 C++, and for C99 and later C versions. Various bugs in the 359 implementation of extended identifiers have been fixed. 360 361 C 362 363 * The default mode has been changed to -std=gnu11. 364 * A new command-line option -Wc90-c99-compat has been added to warn 365 about features not present in ISO C90, but present in ISO C99. 366 * A new command-line option -Wc99-c11-compat has been added to warn 367 about features not present in ISO C99, but present in ISO C11. 368 * It is possible to disable warnings about conversions between 369 pointers that have incompatible types via a new warning option 370 -Wno-incompatible-pointer-types; warnings about implicit 371 incompatible integer to pointer and pointer to integer conversions 372 via a new warning option -Wno-int-conversion; and warnings about 373 qualifiers on pointers being discarded via a new warning option 374 -Wno-discarded-qualifiers. 375 * To allow proper use of const qualifiers with multidimensional 376 arrays, GCC will not warn about incompatible pointer types anymore 377 for conversions between pointers to arrays with and without const 378 qualifier (except when using -pedantic). Instead, a new warning is 379 emitted only if the const qualifier is lost. This can be controlled 380 with a new warning option -Wno-discarded-array-qualifiers. 381 * The C front end now generates more precise caret diagnostics. 382 * The -pg command-line option now only affects the current file in an 383 LTO build. 384 385 C++ 386 387 * G++ now supports [11]C++14 variable templates. 388 * -Wnon-virtual-dtor doesn't warn anymore for final classes. 389 * Excessive template instantiation depth is now a fatal error. This 390 prevents excessive diagnostics that usually do not help to identify 391 the problem. 392 * G++ and libstdc++ now implement the feature-testing macros from 393 [12]Feature-testing recommendations for C++. 394 * G++ now allows typename in a template template parameter. 395 396 template<template<typename> typename X> struct D; // OK 397 398 * G++ now supports [13]C++14 aggregates with non-static data member 399 initializers. 400 401 struct A { int i, j = i; }; 402 A a = { 42 }; // a.j is also 42 403 404 * G++ now supports [14]C++14 extended constexpr. 405 406 constexpr int f (int i) 407 { 408 int j = 0; 409 for (; i > 0; --i) 410 ++j; 411 return j; 412 } 413 414 constexpr int i = f(42); // i is 42 415 416 * G++ now supports the [15]C++14 sized deallocation functions. 417 418 void operator delete (void *, std::size_t) noexcept; 419 void operator delete[] (void *, std::size_t) noexcept; 420 421 * A new One Definition Rule violation warning (controlled by -Wodr) 422 detects mismatches in type definitions and virtual table contents 423 during link-time optimization. 424 * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help 425 developers to annotate programs with final specifiers (or anonymous 426 namespaces) to improve code generation. These warnings can be used 427 at compile time, but they are more useful in combination with 428 link-time optimization. 429 * G++ no longer supports [16]N3639 variable length arrays, as they 430 were removed from the C++14 working paper prior to ratification. 431 GNU VLAs are still supported, so VLA support is now the same in 432 C++14 mode as in C++98 and C++11 modes. 433 * G++ now allows passing a non-trivially-copyable class via C 434 varargs, which is conditionally-supported with 435 implementation-defined semantics in the standard. This uses the 436 same calling convention as a normal value parameter. 437 * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So 438 various mangling bugs are fixed, but G++ will still emit aliases 439 with the old, wrong mangling where feasible. -Wabi=2 will warn 440 about differences between ABI version 2 and the current setting. 441 * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely 442 to be unaffected, but -Wabi=8 will warn about a non-static data 443 member with type std::nullptr_t which changes position due to this 444 change. 445 446 Runtime Library (libstdc++) 447 448 * A [17]Dual ABI is provided by the library. A new ABI is enabled by 449 default. The old ABI is still supported and can be used by defining 450 the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++ 451 standard library headers. 452 * A new implementation of std::string is enabled by default, using 453 the small string optimization instead of copy-on-write reference 454 counting. 455 * A new implementation of std::list is enabled by default, with an 456 O(1) size() function; 457 * [18]Full support for C++11, including the following new features: 458 + std::deque and std::vector<bool> meet the allocator-aware 459 container requirements; 460 + movable and swappable iostream classes; 461 + support for std::align and std::aligned_union; 462 + type traits std::is_trivially_copyable, 463 std::is_trivially_constructible, std::is_trivially_assignable 464 etc.; 465 + I/O manipulators std::put_time, std::get_time, std::hexfloat 466 and std::defaultfloat; 467 + generic locale-aware std::isblank; 468 + locale facets for Unicode conversion; 469 + atomic operations for std::shared_ptr; 470 + std::notify_all_at_thread_exit() and functions for making 471 futures ready at thread exit. 472 * Support for the C++11 hexfloat manipulator changes how the num_put 473 facet formats floating point types when 474 ios_base::fixed|ios_base::scientific is set in a stream's fmtflags. 475 This change affects all language modes, even though the C++98 476 standard gave no special meaning to that combination of flags. To 477 prevent the use of hexadecimal notation for floating point types 478 use str.unsetf(std::ios_base::floatfield) to clear the relevant 479 bits in str.flags(). 480 * [19]Full experimental support for C++14, including the following 481 new features: 482 + std::is_final type trait; 483 + heterogeneous comparison lookup in associative containers. 484 + global functions cbegin, cend, rbegin, rend, crbegin, and 485 crend for range access to containers, arrays and initializer 486 lists. 487 * [20]Improved experimental support for the Library Fundamentals TS, 488 including: 489 + class std::experimental::any; 490 + function template std::experimental::apply; 491 + function template std::experimental::sample; 492 + function template std::experimental::search and related 493 searcher types; 494 + variable templates for type traits; 495 + function template std::experimental::not_fn. 496 * New random number distributions logistic_distribution and 497 uniform_on_sphere_distribution as extensions. 498 * [21]GDB Xmethods for containers and std::unique_ptr. 499 500 Fortran 501 502 * Compatibility notice: 503 + The version of the module files (.mod) has been incremented. 504 + For free-form source files, [22]-Werror=line-truncation is now 505 enabled by default; note that comments exceeding the line 506 length are not diagnosed. (For fixed-form source code, the 507 same warning is available but turned off by default, such that 508 excess characters are ignored. -ffree-line-length-n and 509 -ffixed-line-length-n can be used to modify the default line 510 lengths of 132 and 72 columns, respectively.) 511 + The -Wtabs option is now more sensible: with -Wtabs the 512 compiler warns if it encounters tabs and with -Wno-tabs this 513 warning is turned off. Before, -Wno-tabs warned and -Wtabs 514 turned the warning off. As before, the warning is also enabled 515 by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts 516 options of -std=. 517 * Incomplete support for colorizing diagnostics emitted by gfortran 518 has been added. The option [23]-fdiagnostics-color controls when 519 color is used in diagnostics. The default value of this option can 520 be [24]configured when building GCC. The GCC_COLORS environment 521 variable can be used to customize the colors or disable coloring 522 completely. Sample diagnostics output: 523 $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90 524 test.f90:6:1: 525 526 0 continue 527 1 528 Error: Zero is not a valid statement label at (1) 529 test.f90:9:6: 530 531 USE foo 532 1 533 Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only] 534 535 * The -Wuse-without-only option has been added to warn when a USE 536 statement has no ONLY qualifier and thus implicitly imports all 537 public entities of the used module. 538 * Formatted READ and WRITE statements now work correctly in 539 locale-aware programs. For more information and potential caveats, 540 see [25]Section 5.3 Thread-safety of the runtime library in the 541 manual. 542 * [26]Fortran 2003: 543 + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and 544 IEEE_ARITHMETIC) are now supported. 545 * [27]Fortran 2008: 546 + [28]Coarrays: Full experimental support of Fortran 2008's 547 coarrays with -fcoarray=lib except for allocatable/pointer 548 components of derived-type coarrays. GCC currently only ships 549 with a single-image library (libcaf_single), but multi-image 550 support based on MPI and GASNet is provided by the libraries 551 of the [29]OpenCoarrays project. 552 * TS18508 Additional Parallel Features in Fortran: 553 + Support for the collective intrinsic subroutines CO_MAX, 554 CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added, 555 including -fcoarray=lib support. 556 + Support for the new atomic intrinsics has been added, 557 including -fcoarray=lib support. 558 * Fortran 2015: 559 + Support for IMPLICIT NONE (external, type). 560 + ERROR STOP is now permitted in pure procedures. 561 562 Go 563 564 * GCC 5 provides a complete implementation of the Go 1.4.2 release. 565 * Building GCC 5 with Go enabled will install two new programs: 566 [30]go and [31]gofmt. 567 568 libgccjit 569 570 New in GCC 5 is the ability to build GCC as a shared library for 571 embedding in other processes (such as interpreters), suitable for 572 Just-In-Time compilation to machine code. 573 574 The shared library has a [32]C API and a [33]C++ wrapper API providing 575 some "syntactic sugar". There are also bindings available from 3rd 576 parties for [34]Python and for [35]D. 577 578 For example, this library can be used by interpreters for [36]compiling 579 functions from bytecode to machine code. 580 581 The library can also be used for ahead-of-time compilation, enabling 582 GCC to be plugged into a pre-existing frontend. An example of using 583 this to build a compiler for an esoteric language we'll refer to as 584 "brainf" can be seen [37]here. 585 586 libgccjit is licensed under the GPLv3 (or at your option, any later 587 version) 588 589 It should be regarded as experimental at this time. 590 591 New Targets and Target Specific Improvements 592 593 Reporting stack usage 594 595 * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the 596 -fstack-usage option. 597 598 AArch64 599 600 * Code generation for the ARM Cortex-A57 processor has been improved. 601 A more accurate instruction scheduling model for the processor is 602 now used, and a number of compiler tuning parameters have been set 603 to offer increased performance when compiling with -mcpu=cortex-a57 604 or -mtune=cortex-a57. 605 * A workaround for the ARM Cortex-A53 erratum 835769 has been added 606 and can be enabled by giving the -mfix-cortex-a53-835769 option. 607 Alternatively it can be enabled by default by configuring GCC with 608 the --enable-fix-cortex-a53-835769 option. 609 * The optional cryptographic extensions to the ARMv8-A architecture 610 are no longer enabled by default when specifying the 611 -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53 612 options. To enable these extensions add +crypto to the value of 613 -mcpu or -march e.g. -mcpu=cortex-a53+crypto. 614 * Support has been added for the following processors (GCC 615 identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and 616 initial support for its big.LITTLE combination with the ARM 617 Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx), 618 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 619 The GCC identifiers can be used as arguments to the -mcpu or -mtune 620 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 621 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 622 support for the Cortex-A72. 623 * The transitional options -mlra and -mno-lra have been removed. The 624 AArch64 backend now uses the local register allocator (LRA) only. 625 626 ARM 627 628 * Thumb-1 assembly code is now generated in unified syntax. The new 629 option -masm-syntax-unified specifies whether inline assembly code 630 is using unified syntax. By default the option is off which means 631 non-unified syntax is used. However this is subject to change in 632 future releases. Eventually the non-unified syntax will be 633 deprecated. 634 * It is now a configure-time error to use the --with-cpu configure 635 option with either of --with-tune or --with-arch. 636 * Code generation for the ARM Cortex-A57 processor has been improved. 637 A more accurate instruction scheduling model for the processor is 638 now used, and a number of compiler tuning parameters have been set 639 to offer increased performance when compiling with -mcpu=cortex-a57 640 or -mtune=cortex-a57. 641 * Support has been added for the following processors (GCC 642 identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and 643 initial support for its big.LITTLE combination with the ARM 644 Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and 645 initial support for its big.LITTLE combination with the ARM 646 Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7), 647 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 648 The GCC identifiers can be used as arguments to the -mcpu or -mtune 649 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 650 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 651 support for the Cortex-A72. 652 * The deprecated option -mwords-little-endian has been removed. 653 * The options -mapcs, -mapcs-frame, -mtpcs-frame and 654 -mtpcs-leaf-frame which are only applicable to the old ABI have 655 been deprecated. 656 * The transitional options -mlra and -mno-lra have been removed. The 657 ARM backend now uses the local register allocator (LRA) only. 658 659 AVR 660 661 * The compiler no more supports individual devices like ATmega8. 662 Specifying, say, -mmcu=atmega8 triggers the usage of the 663 device-specific [38]spec file specs-atmega8 which is part of the 664 installation and describes options for the sub-processes like 665 compiler proper, assembler and linker. You can add support for a 666 new device -mmcu=mydevice as follows: 667 1. In an empty directory /someplace, create a new directory 668 device-specs. 669 2. Copy a device spec file from the installed device-specs 670 folder, follow the comments in that file and then save it as 671 /someplace/device-specs/specs-mydevice. 672 3. Add -B /someplace -mmcu=mydevice to the compiler's 673 command-line options. Notice that /someplace must specify an 674 absolute path and that mydevice must not start with "avr". 675 4. Provided you have a device-specific library libmydevice.a 676 available, you can put it at /someplace, dito for a 677 device-specific startup file crtmydevice.o. 678 The contents of the device spec files depend on the compiler's 679 configuration, in particular on --with-avrlibc=no and whether or 680 not it is configured for RTEMS. 681 * A new command-line option -nodevicelib has been added. It prevents 682 the compiler from linking against AVR-LibC's device-specific 683 library libdevice.a. 684 * The following three command-line options have been added: 685 686 -mrmw 687 Set if the device supports the read-modify-write 688 instructions LAC, LAS, LAT and XCH. 689 690 -mn-flash=size 691 Specify the flash size of the device in units of 64 KiB, 692 rounded up to the next integer as needed. This option 693 affects the availability of the [39]AVR address-spaces. 694 695 -mskip-bug 696 Set if the device is affected by the respective silicon 697 bug. 698 699 In general, you don't need to set these options by hand. The new 700 device-specific spec file will set them as needed. 701 702 IA-32/x86-64 703 704 * New ISA extensions support [40]AVX-512{BW,DQ,VL,IFMA,VBMI} of 705 Intel's CPU codenamed Skylake Server was added to GCC. That 706 includes inline assembly support, new intrinsics, and basic 707 autovectorization. These new AVX-512 extensions are available via 708 the following GCC switches: AVX-512 Vector Length EVEX feature: 709 -mavx512vl, AVX-512 Byte and Word instructions: -mavx512bw, AVX-512 710 Dword and Qword instructions: -mavx512dq, AVX-512 FMA-52 711 instructions: -mavx512ifma and for AVX-512 Vector Bit Manipulation 712 Instructions: -mavx512vbmi. 713 * New ISA extensions support [41]Intel MPX was added to GCC. This new 714 extension is available via the -mmpx compiler switch. Intel MPX is 715 a set of processor features which, with compiler, run-time library 716 and OS support, brings increased robustness to software by run-time 717 checking pointer references against their bounds. In GCC Intel MPX 718 is supported by Pointer Bounds Checker and libmpx run-time 719 libraries. 720 * The new -mrecord-mcount option for -pg generates a Linux kernel 721 style table of pointers to mcount or __fentry__ calls at the 722 beginning of functions. The new -mnop-mcount option in addition 723 also generates nops in place of the __fentry__ or mcount call, so 724 that a call per function can be later patched in. This can be used 725 for low overhead tracing or hot code patching. 726 * The new -malign-data option controls how GCC aligns variables. 727 -malign-data=compat uses increased alignment compatible with GCC 728 4.8 and earlier, -malign-data=abi uses alignment as specified by 729 the psABI, and -malign-data=cacheline uses increased alignment to 730 match the cache line size. -malign-data=compat is the default. 731 * The new -mskip-rax-setup option skips setting up the RAX register 732 when SSE is disabled and there are no variable arguments passed in 733 vector registers. This can be used to optimize the Linux kernel. 734 735 MIPS 736 737 * MIPS Releases 3 and 5 are now directly supported. Use the 738 command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5 739 to enable code-generation for these processors. 740 * The Imagination P5600 processor is now supported using the 741 -march=p5600 command-line option. 742 * The Cavium Octeon3 processor is now supported using the 743 -march=octeon3 command-line option. 744 * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6 745 command-line options. 746 * The o32 ABI has been modified and extended. The o32 64-bit 747 floating-point register support is now obsolete and has been 748 removed. It has been replaced by three ABI extensions FPXX, FP64A, 749 and FP64. The meaning of the -mfp64 command-line option has 750 changed. It is now used to enable the FP64A and FP64 ABI 751 extensions. 752 + The FPXX extension requires that code generated to access 753 double-precision values use even-numbered registers. Code that 754 adheres to this extension is link-compatible with all other 755 o32 double-precision ABI variants and will execute correctly 756 in all hardware FPU modes. The command-line options -mabi=32 757 -mfpxx can be used to enable this extension. MIPS II is the 758 minimum processor required. 759 + The o32 FP64A extension requires that floating-point registers 760 be 64-bit and odd-numbered single-precision registers are not 761 allowed. Code that adheres to the o32 FP64A variant is 762 link-compatible with all other o32 double-precision ABI 763 variants. The command-line options -mabi=32 -mfp64 764 -mno-odd-spreg can be used to enable this extension. MIPS32R2 765 is the minimum processor required. 766 + The o32 FP64 extension also requires that floating-point 767 registers be 64-bit, but permits the use of single-precision 768 registers. Code that adheres to the o32 FP64 variant is 769 link-compatible with o32 FPXX and o32 FP64A variants only, 770 i.e. it is not compatible with the original o32 771 double-precision ABI. The command-line options -mabi=32 -mfp64 772 -modd-spreg can be used to enable this extension. MIPS32R2 is 773 the minimum processor required. 774 The new ABI variants can be enabled by default using the configure 775 time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32. 776 It is strongly recommended that all vendors begin to set o32 FPXX 777 as the default ABI. This will be required to run the generated code 778 on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and 779 MIPSR6 cores. 780 * GCC will now pass all floating-point options to the assembler if 781 GNU binutils 2.25 is used. As a result, any inline assembly code 782 that uses hard-float instructions should be amended to include a 783 .set directive to override the global assembler options when 784 compiling for soft-float targets. 785 786 NDS32 787 788 * The variadic function ABI implementation is now compatible with 789 past Andes toolchains where the caller uses registers to pass 790 arguments and the callee is in charge of pushing them on stack. 791 * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have 792 been removed since they are not yet available in the nds32 port of 793 GNU binutils. 794 * A new option -mcmodel=[small|medium|large] supports varied code 795 models on code generation. The -mgp-direct option became 796 meaningless and can be discarded. 797 798 RX 799 800 * A new command line option -mno-allow-string-insns can be used to 801 disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL, 802 SWHILE and RMPA instructions. An erratum released by Renesas shows 803 that it is unsafe to use these instructions on addresses within the 804 I/O space of the processor. The new option can be used when the 805 programmer is concerned that the I/O space might be accessed. The 806 default is still to enable these instructions. 807 808 SH 809 810 * The compiler will now pass the appropriate --isa= option to the 811 assembler. 812 * The default handling for the GBR has been changed from call 813 clobbered to call preserved. The old behavior can be reinstated by 814 specifying the option -fcall-used-gbr. 815 * Support for the SH4A fpchg instruction has been added which will be 816 utilized when switching between single and double precision FPU 817 modes. 818 * The compiler no longer uses the __fpscr_values array for switching 819 between single and double FPU precision modes on non-SH4A targets. 820 Instead mode switching will now be performed by storing, modifying 821 and reloading the FPSCR, so that other FPSCR bits are preserved 822 across mode switches. The __fpscr_values array that is defined in 823 libgcc is still present for backwards compatibility, but it will 824 not be referenced by compiler generated code anymore. 825 * New builtin functions __builtin_sh_get_fpscr and 826 __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr 827 function will mask the specified bits in such a way that the SZ, PR 828 and FR mode bits will be preserved, while changing the other bits. 829 These new functions do not reference the __fpscr_values array. The 830 old functions __set_fpscr and __get_fpscr in libgcc which access 831 the __fpscr_values array are still present for backwards 832 compatibility, but their usage is highly discouraged. 833 * Some improvements to code generated for __atomic built-in 834 functions. 835 * When compiling for SH2E the compiler will no longer force the usage 836 of delay slots for conditional branch instructions bt and bf. The 837 old behavior can be reinstated (e.g. to work around a hardware bug 838 in the original SH7055) by specifying the new option 839 -mcbranch-force-delay-slot. 840 841 Operating Systems 842 843 DragonFly BSD 844 845 * GCC now supports the DragonFly BSD operating system. 846 847 FreeBSD 848 849 * GCC now supports the FreeBSD operating system for the arm port 850 through the arm*-*-freebsd* target triplets. 851 852 VxWorks MILS 853 854 * GCC now supports the MILS (Multiple Independent Levels of Security) 855 variant of WindRiver's VxWorks operating system for PowerPC 856 targets. 857 858 Other significant improvements 859 860 * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option 861 to set the compiler to use. 862 863 * When the new command-line option -freport-bug is used, GCC 864 automatically generates a developer-friendly reproducer whenever an 865 internal compiler error is encountered. 866 867 GCC 5.2 868 869 This is the [42]list of problem reports (PRs) from GCC's bug tracking 870 system that are known to be fixed in the 5.2 release. This list might 871 not be complete (that is, it is possible that some PRs that have been 872 fixed are not listed here). 873 874 Target Specific Changes 875 876 IA-32/x86-64 877 878 * Support for new AMD instructions monitorx and mwaitx has been 879 added. This includes new intrinsic and built-in support. It is 880 enabled through option -mmwaitx. The instructions monitorx and 881 mwaitx implement the same functionality as the old monitor and 882 mwait instructions. In addition mwaitx adds a configurable timer. 883 The timer value is received as third argument and stored in 884 register %ebx. 885 886 (Pending) GCC 5.3 887 888 This is the [43]list of problem reports (PRs) from GCC's bug tracking 889 system that are known to be fixed in the 5.3 release. This list might 890 not be complete (that is, it is possible that some PRs that have been 891 fixed are not listed here). 892 893 Target Specific Changes 894 895 IA-32/x86-64 896 897 * GCC now supports the Intel CPU named Skylake with AVX-512 898 extensions through -march=skylake-avx512. The switch enables the 899 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 900 AVX-512DQ. 901 902 903 For questions related to the use of GCC, please consult these web 904 pages and the [44]GCC manuals. If that fails, the 905 [45]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 906 web pages and the development of GCC are welcome on our developer 907 list at [46]gcc (a] gcc.gnu.org. All of [47]our lists have public 908 archives. 909 910 Copyright (C) [48]Free Software Foundation, Inc. Verbatim copying and 911 distribution of this entire article is permitted in any medium, 912 provided this notice is preserved. 913 914 These pages are [49]maintained by the GCC team. Last modified 915 2015-12-04[50]. 916 917 References 918 919 1. http://gcc.gnu.org/gcc-5/changes.html#libstdcxx 920 2. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 921 3. https://savannah.nongnu.org/bugs/?44574 922 4. https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler 923 5. http://www.openmp.org/mp-documents/OpenMP4.0.0.pdf 924 6. http://openmp.org/mp-documents/OpenMP4.0.0.Examples.pdf 925 7. https://gcc.gnu.org/wiki/OpenACC 926 8. https://gcc.gnu.org/install/configure.html 927 9. https://www.cilkplus.org/ 928 10. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 929 11. http://gcc.gnu.org/projects/cxx1y.html 930 12. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 931 13. http://gcc.gnu.org/projects/cxx1y.html 932 14. http://gcc.gnu.org/projects/cxx1y.html 933 15. http://gcc.gnu.org/projects/cxx1y.html 934 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html 935 17. http://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html 936 18. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011 937 19. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 938 20. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 939 21. https://sourceware.org/gdb/current/onlinedocs/gdb/Xmethods-In-Python.html 940 22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html 941 23. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html 942 24. https://gcc.gnu.org/install/configure.html 943 25. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html 944 26. https://gcc.gnu.org/wiki/Fortran2003Status 945 27. https://gcc.gnu.org/wiki/Fortran2008Status 946 28. https://gcc.gnu.org/wiki/Coarray 947 29. http://www.opencoarrays.org/ 948 30. https://golang.org/cmd/go/ 949 31. https://golang.org/cmd/gofmt/ 950 32. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html 951 33. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html 952 34. https://github.com/davidmalcolm/pygccjit 953 35. https://github.com/ibuclaw/gccjitd 954 36. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html 955 37. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html 956 38. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 957 39. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html 958 40. https://software.intel.com/sites/default/files/managed/0d/53/319433-023.pdf 959 41. https://software.intel.com/sites/default/files/managed/0d/53/319433-022.pdf 960 42. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2 961 43. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3 962 44. https://gcc.gnu.org/onlinedocs/ 963 45. mailto:gcc-help (a] gcc.gnu.org 964 46. mailto:gcc (a] gcc.gnu.org 965 47. https://gcc.gnu.org/lists.html 966 48. http://www.fsf.org/ 967 49. https://gcc.gnu.org/about.html 968 50. http://validator.w3.org/check/referer 969 ====================================================================== 970 http://gcc.gnu.org/gcc-4.9/index.html 971 972 GCC 4.9 Release Series 973 974 June 26, 2015 975 976 The [1]GNU project and the GCC developers are pleased to announce the 977 release of GCC 4.9.3. 978 979 This release is a bug-fix release, containing fixes for regressions in 980 GCC 4.9.2 relative to previous releases of GCC. 981 982 Release History 983 984 GCC 4.9.3 985 June 26, 2015 ([2]changes, [3]documentation) 986 987 GCC 4.9.2 988 October 30, 2014 ([4]changes, [5]documentation) 989 990 GCC 4.9.1 991 July 16, 2014 ([6]changes, [7]documentation) 992 993 GCC 4.9.0 994 April 22, 2014 ([8]changes, [9]documentation) 995 996 References and Acknowledgements 997 998 GCC used to stand for the GNU C Compiler, but since the compiler 999 supports several other languages aside from C, it now stands for the 1000 GNU Compiler Collection. 1001 1002 A list of [10]successful builds is updated as new information becomes 1003 available. 1004 1005 The GCC developers would like to thank the numerous people that have 1006 contributed new features, improvements, bug fixes, and other changes as 1007 well as test results to GCC. This [11]amazing group of volunteers is 1008 what makes GCC successful. 1009 1010 For additional information about GCC please refer to the [12]GCC 1011 project web site or contact the [13]GCC development mailing list. 1012 1013 To obtain GCC please use [14]our mirror sites or [15]our SVN server. 1014 1015 1016 For questions related to the use of GCC, please consult these web 1017 pages and the [16]GCC manuals. If that fails, the 1018 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1019 web pages and the development of GCC are welcome on our developer 1020 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 1021 archives. 1022 1023 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 1024 distribution of this entire article is permitted in any medium, 1025 provided this notice is preserved. 1026 1027 These pages are [21]maintained by the GCC team. Last modified 1028 2015-06-26[22]. 1029 1030 References 1031 1032 1. http://www.gnu.org/ 1033 2. http://gcc.gnu.org/gcc-4.9/changes.html 1034 3. http://gcc.gnu.org/onlinedocs/4.9.3/ 1035 4. http://gcc.gnu.org/gcc-4.9/changes.html 1036 5. http://gcc.gnu.org/onlinedocs/4.9.2/ 1037 6. http://gcc.gnu.org/gcc-4.9/changes.html 1038 7. http://gcc.gnu.org/onlinedocs/4.9.1/ 1039 8. http://gcc.gnu.org/gcc-4.9/changes.html 1040 9. https://gcc.gnu.org/onlinedocs/4.9.0/ 1041 10. http://gcc.gnu.org/gcc-4.9/buildstat.html 1042 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 1043 12. http://gcc.gnu.org/index.html 1044 13. mailto:gcc (a] gcc.gnu.org 1045 14. http://gcc.gnu.org/mirrors.html 1046 15. http://gcc.gnu.org/svn.html 1047 16. https://gcc.gnu.org/onlinedocs/ 1048 17. mailto:gcc-help (a] gcc.gnu.org 1049 18. mailto:gcc (a] gcc.gnu.org 1050 19. https://gcc.gnu.org/lists.html 1051 20. http://www.fsf.org/ 1052 21. https://gcc.gnu.org/about.html 1053 22. http://validator.w3.org/check/referer 1054 ====================================================================== 1055 http://gcc.gnu.org/gcc-4.9/changes.html 1056 1057 GCC 4.9 Release Series 1058 Changes, New Features, and Fixes 1059 1060 Caveats 1061 1062 * The mudflap run time checker has been removed. The mudflap options 1063 remain, but do nothing. 1064 * Support for a number of older systems and recently unmaintained or 1065 untested target ports of GCC has been declared obsolete in GCC 4.9. 1066 Unless there is activity to revive them, the next release of GCC 1067 will have their sources permanently removed. 1068 The following ports for individual systems on particular 1069 architectures have been obsoleted: 1070 + Solaris 9 (*-*-solaris2.9). Details can be found in the 1071 [1]announcement. 1072 * On AArch64, the singleton vector types int64x1_t, uint64x1_t and 1073 float64x1_t exported by arm_neon.h are defined to be the same as 1074 their base types. This results in incorrect application of 1075 parameter passing rules to arguments of types int64x1_t and 1076 uint64x1_t, with respect to the AAPCS64 ABI specification. In 1077 addition, names of C++ functions with parameters of these types 1078 (including float64x1_t) are not mangled correctly. The current 1079 typedef declarations also unintentionally allow implicit casting 1080 between singleton vector types and their base types. These issues 1081 will be resolved in a near future release. See [2]PR60825 for more 1082 information. 1083 1084 More information on porting to GCC 4.9 from previous versions of GCC 1085 can be found in the [3]porting guide for this release. 1086 1087 General Optimizer Improvements 1088 1089 * AddressSanitizer, a fast memory error detector, is now available on 1090 ARM. 1091 * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior 1092 detector, has been added and can be enabled via 1093 -fsanitize=undefined. Various computations will be instrumented to 1094 detect undefined behavior at runtime. UndefinedBehaviorSanitizer is 1095 currently available for the C and C++ languages. 1096 * Link-time optimization (LTO) improvements: 1097 + Type merging was rewritten. The new implementation is 1098 significantly faster and uses less memory. 1099 + Better partitioning algorithm resulting in less streaming 1100 during link time. 1101 + Early removal of virtual methods reduces the size of object 1102 files and improves link-time memory usage and compile time. 1103 + Function bodies are now loaded on-demand and released early 1104 improving overall memory usage at link time. 1105 + C++ hidden keyed methods can now be optimized out. 1106 + When using a linker plugin, compiling with the -flto option 1107 now generates slim object files (.o) which only contain 1108 intermediate language representation for LTO. Use 1109 -ffat-lto-objects to create files which contain additionally 1110 the object code. To generate static libraries suitable for LTO 1111 processing, use gcc-ar and gcc-ranlib; to list symbols from a 1112 slim object file use gcc-nm. (This requires that ar, ranlib 1113 and nm have been compiled with plugin support.) 1114 Memory usage building Firefox with debug enabled was reduced from 1115 15GB to 3.5GB; link time from 1700 seconds to 350 seconds. 1116 * Inter-procedural optimization improvements: 1117 + New type inheritance analysis module improving 1118 devirtualization. Devirtualization now takes into account 1119 anonymous name-spaces and the C++11 final keyword. 1120 + New speculative devirtualization pass (controlled by 1121 -fdevirtualize-speculatively. 1122 + Calls that were speculatively made direct are turned back to 1123 indirect where direct call is not cheaper. 1124 + Local aliases are introduced for symbols that are known to be 1125 semantically equivalent across shared libraries improving 1126 dynamic linking times. 1127 * Feedback directed optimization improvements: 1128 + Profiling of programs using C++ inline functions is now more 1129 reliable. 1130 + New time profiling determines typical order in which functions 1131 are executed. 1132 + A new function reordering pass (controlled by 1133 -freorder-functions) significantly reduces startup time of 1134 large applications. Until binutils support is completed, it is 1135 effective only with link-time optimization. 1136 + Feedback driven indirect call removal and devirtualization now 1137 handle cross-module calls when link-time optimization is 1138 enabled. 1139 1140 New Languages and Language specific improvements 1141 1142 * Version 4.0 of the [4]OpenMP specification is now supported in the 1143 C and C++ compilers and starting with the 4.9.1 release also in the 1144 Fortran compiler. The new -fopenmp-simd option can be used to 1145 enable OpenMP's SIMD directives while ignoring other OpenMP 1146 directives. The new [5]-fsimd-cost-model= option permits to tune 1147 the vectorization cost model for loops annotated with OpenMP and 1148 Cilk Plus simd directives. -Wopenmp-simd warns when the current 1149 cost model overrides simd directives set by the user. 1150 * The -Wdate-time option has been added for the C, C++ and Fortran 1151 compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__ 1152 macros are used. Those macros might prevent bit-wise-identical 1153 reproducible compilations. 1154 1155 Ada 1156 1157 * GNAT switched to Ada 2012 instead of Ada 2005 by default. 1158 1159 C family 1160 1161 * Support for colorizing diagnostics emitted by GCC has been added. 1162 The [6]-fdiagnostics-color=auto will enable it when outputting to 1163 terminals, -fdiagnostics-color=always unconditionally. The 1164 GCC_COLORS environment variable can be used to customize the colors 1165 or disable coloring. If GCC_COLORS variable is present in the 1166 environment, the default is -fdiagnostics-color=auto, otherwise 1167 -fdiagnostics-color=never. 1168 Sample diagnostics output: 1169 $ g++ -fdiagnostics-color=always -S -Wall test.C 1170 test.C: In function `int foo()': 1171 test.C:1:14: warning: no return statement in function returning non-void [-W 1172 return-type] 1173 int foo () { } 1174 ^ 1175 test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use 1176 -ftemplate-depth= to increase the maximum) instantiating `struct X<100>' 1177 template <int N> struct X { static const int value = X<N-1>::value; }; temp 1178 late struct X<1000>; 1179 ^ 1180 test.C:2:46: recursively required from `const int X<999>::value' 1181 test.C:2:46: required from `const int X<1000>::value' 1182 test.C:2:88: required from here 1183 1184 test.C:2:46: error: incomplete type `X<100>' used in nested name specifier 1185 1186 * With the new [7]#pragma GCC ivdep, the user can assert that there 1187 are no loop-carried dependencies which would prevent concurrent 1188 execution of consecutive iterations using SIMD (single instruction 1189 multiple data) instructions. 1190 * Support for [8]Cilk Plus has been added and can be enabled with the 1191 -fcilkplus option. Cilk Plus is an extension to the C and C++ 1192 languages to support data and task parallelism. The present 1193 implementation follows ABI version 1.2; all features but _Cilk_for 1194 have been implemented. 1195 1196 C 1197 1198 * ISO C11 atomics (the _Atomic type specifier and qualifier and the 1199 <stdatomic.h> header) are now supported. 1200 * ISO C11 generic selections (_Generic keyword) are now supported. 1201 * ISO C11 thread-local storage (_Thread_local, similar to GNU C 1202 __thread) is now supported. 1203 * ISO C11 support is now at a similar level of completeness to ISO 1204 C99 support: substantially complete modulo bugs, extended 1205 identifiers (supported except for corner cases when 1206 -fextended-identifiers is used), floating-point issues (mainly but 1207 not entirely relating to optional C99 features from Annexes F and 1208 G) and the optional Annexes K (Bounds-checking interfaces) and L 1209 (Analyzability). 1210 * A new C extension __auto_type provides a subset of the 1211 functionality of C++11 auto in GNU C. 1212 1213 C++ 1214 1215 * The G++ implementation of [9]C++1y return type deduction for normal 1216 functions has been updated to conform to [10]N3638, the proposal 1217 accepted into the working paper. Most notably, it adds 1218 decltype(auto) for getting decltype semantics rather than the 1219 template argument deduction semantics of plain auto: 1220 1221 int& f(); 1222 auto i1 = f(); // int 1223 decltype(auto) i2 = f(); // int& 1224 1225 * G++ supports [11]C++1y lambda capture initializers: 1226 1227 [x = 42]{ ... }; 1228 1229 Actually, they have been accepted since GCC 4.5, but now the 1230 compiler doesn't warn about them with -std=c++1y, and supports 1231 parenthesized and brace-enclosed initializers as well. 1232 * G++ supports [12]C++1y variable length arrays. G++ has supported 1233 GNU/C99-style VLAs for a long time, but now additionally supports 1234 initializers and lambda capture by reference. In C++1y mode G++ 1235 will complain about VLA uses that are not permitted by the draft 1236 standard, such as forming a pointer to VLA type or applying sizeof 1237 to a VLA variable. Note that it now appears that VLAs will not be 1238 part of C++14, but will be part of a separate document and then 1239 perhaps C++17. 1240 1241 void f(int n) { 1242 int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3 1243 [&a]{ for (int i : a) { cout << i << endl; } }(); 1244 &a; // error, taking address of VLA 1245 } 1246 1247 * G++ supports the [13]C++1y [[deprecated]] attribute modulo bugs in 1248 the underlying [[gnu::deprecated]] attribute. Classes and functions 1249 can be marked deprecated and a diagnostic message added: 1250 1251 class A; 1252 int bar(int n); 1253 #if __cplusplus > 201103 1254 class [[deprecated("A is deprecated in C++14; Use B instead")]] A; 1255 [[deprecated("bar is unsafe; use foo() instead")]] 1256 int bar(int n); 1257 1258 int foo(int n); 1259 class B; 1260 #endif 1261 A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead 1262 int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo 1263 o() instead 1264 1265 * G++ supports [14]C++1y digit separators. Long numeric literals can 1266 be subdivided with a single quote ' to enhance readability: 1267 1268 int i = 1048576; 1269 int j = 1'048'576; 1270 int k = 0x10'0000; 1271 int m = 0'004'000'000; 1272 int n = 0b0001'0000'0000'0000'0000'0000; 1273 1274 double x = 1.602'176'565e-19; 1275 double y = 1.602'176'565e-1'9; 1276 1277 * G++ supports [15]C++1y generic (polymorphic) lambdas. 1278 1279 // a functional object that will increment any type 1280 auto incr = [](auto x) { return x++; }; 1281 1282 * As a GNU extension, G++ supports explicit template parameter syntax 1283 for generic lambdas. This can be combined in the expected way with 1284 the standard auto syntax. 1285 1286 // a functional object that will add two like-type objects 1287 auto add = [] <typename T> (T a, T b) { return a + b; }; 1288 1289 * G++ supports unconstrained generic functions as specified by 1290 S:4.1.2 and S:5.1.1 of [16]N3889: Concepts Lite Specification. 1291 Briefly, auto may be used as a type-specifier in a parameter 1292 declaration of any function declarator in order to introduce an 1293 implicit function template parameter, akin to generic lambdas. 1294 1295 // the following two function declarations are equivalent 1296 auto incr(auto x) { return x++; } 1297 template <typename T> 1298 auto incr(T x) { return x++; } 1299 1300 Runtime Library (libstdc++) 1301 1302 * [17]Improved support for C++11, including: 1303 + support for <regex>; 1304 + The associative containers in <map> and <set> and the 1305 unordered associative containers in <unordered_map> and 1306 <unordered_set> meet the allocator-aware container 1307 requirements; 1308 * [18]Improved experimental support for the upcoming ISO C++ 1309 standard, C++14, including: 1310 + fixing constexpr member functions without const; 1311 + implementation of the std::exchange() utility function; 1312 + addressing tuples by type; 1313 + implemention of std::make_unique; 1314 + implemention of std::shared_lock; 1315 + making std::result_of SFINAE-friendly; 1316 + adding operator() to std::integral_constant; 1317 + adding user-defined literals for standard library types 1318 std::basic_string, std::chrono::duration, and std::complex; 1319 + adding two range overloads to non-modifying sequence oprations 1320 std::equal and std::mismatch; 1321 + adding IO manipulators for quoted strings; 1322 + adding constexpr members to <utility>, <complex>, <chrono>, 1323 and some containers; 1324 + adding compile-time std::integer_sequence; 1325 + adding cleaner transformation traits; 1326 + making <functional>s operator functors easier to use and more 1327 generic; 1328 * An implementation of std::experimental::optional. 1329 * An implementation of std::experimental::string_view. 1330 * The non-standard function std::copy_exception has been deprecated 1331 and will be removed in a future version. std::make_exception_ptr 1332 should be used instead. 1333 1334 Fortran 1335 1336 * Compatibility notice: 1337 + Module files: The version of the module files (.mod) has been 1338 incremented; additionally, module files are now compressed. 1339 Fortran MODULEs compiled by earlier GCC versions have to be 1340 recompiled, when they are USEd by files compiled with GCC 4.9. 1341 GCC 4.9 is not able to read .mod files of earlier GCC 1342 versions; attempting to do so gives an error message. Note: 1343 The ABI of the produced assembler data itself has not changed: 1344 object files and libraries are fully compatible with older 1345 versions (except as stated below). 1346 + ABI changes: 1347 o The [19]argument passing ABI has changed for scalar dummy 1348 arguments of type INTEGER, REAL, COMPLEX and LOGICAL, 1349 which have both the VALUE and the OPTIONAL attributes. 1350 o To support finalization the virtual table associated with 1351 polymorphic variables has changed. Code containing CLASS 1352 should be recompiled, including all files which define 1353 derived types involved in the type definition used by 1354 polymorphic variables. (Note: Due to the incremented 1355 module version, trying to mix old code with new code will 1356 usually give an error message.) 1357 + GNU Fortran no longer deallocates allocatable variables or 1358 allocatable components of variables declared in the main 1359 program. Since Fortran 2008, the standard explicitly states 1360 that variables declared in the Fortran main program 1361 automatically have the SAVE attribute. 1362 + When opening files, the close-on-exec flag is set if the 1363 system supports such a feature. This is generally considered 1364 good practice these days, but if there is a need to pass file 1365 descriptors to child processes the parent process must now 1366 remember to clear the close-on-exec flag by calling fcntl(), 1367 e.g. via ISO_C_BINDING, before executing the child process. 1368 * The deprecated command-line option -fno-whole-file has been 1369 removed. (-fwhole-file is the default since GCC 4.6.) 1370 -fwhole-file/-fno-whole-file continue to be accepted but do not 1371 influence the code generation. 1372 * The compiler no longer unconditionally warns about DO loops with 1373 zero iterations. This warning is now controlled by the -Wzerotrip 1374 option, which is implied by -Wall. 1375 * The new NO_ARG_CHECK attribute of the [20]!GCC$ directive can be 1376 used to disable the type-kind-rank (TKR) argument check for a dummy 1377 argument. The feature is similar to ISO/IEC TS 29133:2012's 1378 TYPE(*), except that it additionally also disables the rank check. 1379 Variables with NO_ARG_CHECK have to be dummy arguments and may only 1380 be used as argument to ISO_C_BINDING's C_LOC and as actual argument 1381 to another NO_ARG_CHECK dummy argument; also the other constraints 1382 of TYPE(*) apply. The dummy arguments should be declared as scalar 1383 or assumed-size variable of type type(*) (recommended) - or of type 1384 integer, real, complex or logical. With NO_ARG_CHECK, a pointer to 1385 the data without further type or shape information is passed, 1386 similar to C's void*. Note that also TS 29113's 1387 type(*),dimension(..) accepts arguments of any type and rank; 1388 contrary to NO_ARG_CHECK assumed-rank arguments pass an array 1389 descriptor which contains the array shape and stride of the 1390 argument. 1391 * [21]Fortran 2003: 1392 + Finalization is now supported. It is currently only done for a 1393 subset of those situations in which it should occur. 1394 + Experimental support for scalar character components with 1395 deferred length (i.e. allocatable string length) in derived 1396 types has been added. (Deferred-length character variables are 1397 supported since GCC 4.6.) 1398 * [22]Fortran 2008: 1399 + When STOP or ERROR STOP are used to terminate the execution 1400 and any exception (but inexact) is signaling, a warning is 1401 printed to ERROR_UNIT, indicating which exceptions are 1402 signaling. The [23]-ffpe-summary= command-line option can be 1403 used to fine-tune for which exceptions the warning should be 1404 shown. 1405 + Rounding on input (READ) is now handled on systems where 1406 strtod honours the rounding mode. (For output, rounding is 1407 supported since GCC 4.5.) Note that for input, the compatible 1408 rounding mode is handled as nearest (i.e., rounding to an even 1409 least significant [cf. IEC 60559:1989] for a tie, while 1410 compatible rounds away from zero in that case). 1411 1412 Go 1413 1414 * GCC 4.9 provides a complete implementation of the Go 1.2.1 release. 1415 1416 New Targets and Target Specific Improvements 1417 1418 AArch64 1419 1420 * The ARMv8-A crypto and CRC instructions are now supported through 1421 intrinsics. These are enabled when the architecture supports these 1422 and are available through the -march=armv8-a+crc and 1423 -march=armv8-a+crypto options. 1424 * Initial support for ILP32 has now been added to the compiler. This 1425 is now available through the command-line option -mabi=ilp32. 1426 Support for ILP32 is considered experimental as the ABI 1427 specification is still beta. 1428 * Coverage of more of the ISA including the SIMD extensions has been 1429 added. The Advanced SIMD intrinsics have also been improved. 1430 * The new local register allocator (LRA) is now on by default for the 1431 AArch64 backend. 1432 * The REE (Redundant extension elimination) pass has now been enabled 1433 by default for the AArch64 backend. 1434 * Tuning for the Cortex-A53 and Cortex-A57 has been improved. 1435 * Initial big.LITTLE tuning support for the combination of Cortex-A57 1436 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 1437 option. 1438 * A number of structural changes have been made to both the ARM and 1439 AArch64 backends to facilitate improved code-generation. 1440 * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769 1441 has been added and can be enabled by giving the 1442 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 1443 default by configuring GCC with the --enable-fix-cortex-a53-835769 1444 option. 1445 1446 ARC 1447 1448 * A port for Synopsys Designware ARC has been contributed by Embecosm 1449 and Synopsys Inc. 1450 1451 ARM 1452 1453 * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been 1454 disabled by default. This was found to generate better code in only 1455 a small number of cases. It can be turned back on with the 1456 -mneon-for-64bits option. 1457 * Further support for the ARMv8-A architecture, notably implementing 1458 the restriction around IT blocks in the Thumb32 instruction set has 1459 been added. The -mrestrict-it option can be used with 1460 -march=armv7-a or the -march=armv7ve options to make code 1461 generation fully compatible with the deprecated instructions in 1462 ARMv8-A. 1463 * Support has now been added for the ARMv7ve variant of the 1464 architecture. This can be used by the -march=armv7ve option. 1465 * The ARMv8-A crypto and CRC instructions are now supported through 1466 intrinsics and are available through the -march=armv8-a+crc and 1467 mfpu=crypto-neon-fp-armv8 options. 1468 * LRA is now on by default for the ARM target. This can be turned off 1469 using the -mno-lra option. This option is a purely transitionary 1470 command-line option and will be removed in a future release. We are 1471 interested in any bug reports regarding functional and performance 1472 regressions with LRA. 1473 * A new option -mslow-flash-data to improve performance of programs 1474 fetching data on slow flash memory has now been introduced for the 1475 ARMv7-M profile cores. 1476 * A new option -mpic-data-is-text-relative for targets that allows 1477 data segments to be relative to text segments has been added. This 1478 is on by default for all targets except VxWorks RTP. 1479 * A number of infrastructural changes have been made to both the ARM 1480 and AArch64 backends to facilitate improved code-generation. 1481 * GCC now supports Cortex-A12 and the Cortex-R7 through the 1482 -mcpu=cortex-a12 and -mcpu=cortex-r7 options. 1483 * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the 1484 -mcpu=cortex-a57 and -mcpu=cortex-a53 options. 1485 * Initial big.LITTLE tuning support for the combination of Cortex-A57 1486 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 1487 option. Similar support was added for the combination of Cortex-A15 1488 and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option. 1489 * Further performance optimizations for the Cortex-A15 and the 1490 Cortex-M4 have been added. 1491 * A number of code generation improvements for Thumb2 to reduce code 1492 size when compiling for the M-profile processors. 1493 1494 IA-32/x86-64 1495 1496 * -mfpmath=sse is now implied by -ffast-math on all targets where 1497 SSE2 is supported. 1498 * Intel AVX-512 support was added to GCC. That includes inline 1499 assembly support, new registers and extending existing ones, new 1500 intrinsics (covered by corresponding testsuite), and basic 1501 autovectorization. AVX-512 instructions are available via the 1502 following GCC switches: AVX-512 foundation instructions: -mavx512f, 1503 AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and 1504 reciprocal instructions: -mavx512er, AVX-512 conflict detection 1505 instructions: -mavx512cd. 1506 * It is now possible to call x86 intrinsics from select functions in 1507 a file that are tagged with the corresponding target attribute 1508 without having to compile the entire file with the -mxxx option. 1509 This improves the usability of x86 intrinsics and is particularly 1510 useful when doing [24]Function Multiversioning. 1511 * GCC now supports the new Intel microarchitecture named Silvermont 1512 through -march=silvermont. 1513 * GCC now supports the new Intel microarchitecture named Broadwell 1514 through -march=broadwell. 1515 * Optimizing for other Intel microarchitectures have been renamed to 1516 -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell. 1517 * -march=generic has been retuned for better support of Intel core 1518 and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel 1519 Pentium-M, and Pentium4 based CPUs is no longer considered 1520 important for generic. 1521 * -mtune=intel can now be used to generate code running well on the 1522 most current Intel processors, which are Haswell and Silvermont for 1523 GCC 4.9. 1524 * Support to encode 32-bit assembly instructions in 16-bit format is 1525 now available through the -m16 command-line option. 1526 * Better inlining of memcpy and memset that is aware of value ranges 1527 and produces shorter alignment prologues. 1528 * -mno-accumulate-outgoing-args is now honored when unwind 1529 information is output. Argument accumulation is also now turned off 1530 for portions of programs optimized for size. 1531 * Support for new AMD family 15h processors (Excavator core) is now 1532 available through the -march=bdver4 and -mtune=bdver4 options. 1533 1534 MSP430 1535 1536 * A new command-line option -mcpu= has been added to the MSP430 1537 backend. This option is used to specify the ISA to be used. 1538 Accepted values are msp430 (the default), msp430x and msp430xv2. 1539 The ISA is no longer deduced from the -mmcu= option as there are 1540 far too many different MCU names. The -mmcu= option is still 1541 supported, and this is still used to select linker scripts and 1542 generate a C preprocessor symbol that will be recognised by the 1543 msp430.h header file. 1544 1545 NDS32 1546 1547 * A new nds32 port supports the 32-bit architecture from Andes 1548 Technology Corporation. 1549 * The port provides initial support for the V2, V3, V3m instruction 1550 set architectures. 1551 1552 Nios II 1553 1554 * A port for the Altera Nios II has been contributed by Mentor 1555 Graphics. 1556 1557 PowerPC / PowerPC64 / RS6000 1558 1559 * GCC now supports Power ISA 2.07, which includes support for 1560 Hardware Transactional Memory (HTM), Quadword atomics and several 1561 VMX and VSX additions, including Crypto, 64-bit integer, 128-bit 1562 integer and decimal integer operations. 1563 * Support for the POWER8 processor is now available through the 1564 -mcpu=power8 and -mtune=power8 options. 1565 * The libitm library has been modified to add a HTM fastpath that 1566 automatically uses POWER's HTM hardware instructions when it is 1567 executing on a HTM enabled processor. 1568 * Support for the new powerpc64le-linux platform has been added. It 1569 defaults to generating code that conforms to the ELFV2 ABI. 1570 1571 S/390, System z 1572 1573 * Support for the Transactional Execution Facility included with the 1574 IBM zEnterprise zEC12 processor has been added. A set of GCC style 1575 builtins as well as XLC style builtins are provided. The builtins 1576 are enabled by default when using the -march=zEC12 option but can 1577 explicitly be disabled with -mno-htm. Using the GCC builtins also 1578 libitm supports hardware transactions on S/390. 1579 * The hotpatch features allows to prepare functions for hotpatching. 1580 A certain amount of bytes is reserved before the function entry 1581 label plus a NOP is inserted at its very beginning to implement a 1582 backward jump when applying a patch. The feature can either be 1583 enabled per compilation unit via the command-line option -mhotpatch 1584 or per function using the hotpatch attribute. 1585 * The shrink wrap optimization is now supported on S/390 and enabled 1586 by default. 1587 * A major rework of the routines to determine which registers need to 1588 be saved and restored in function prologue/epilogue now allow to 1589 use floating point registers as save slots. This will happen for 1590 certain leaf function with -march=z10 or higher. 1591 * The LRA rtl pass replaces reload by default on S/390. 1592 1593 RX 1594 1595 * The port now allows to specify the RX100, RX200, and RX600 1596 processors with the command-line options -mcpu=rx100, -mcpu=rx200 1597 and -mcpu=rx600. 1598 1599 SH 1600 1601 * Minor improvements to code generated for integer arithmetic and 1602 code that involves the T bit. 1603 * Added support for the SH2A clips and clipu instructions. The 1604 compiler will now try to utilize them for min/max expressions such 1605 as max (-128, min (127, x)). 1606 * Added support for the cmp/str instruction through built-in 1607 functions such as __builtin_strlen. When not optimizing for size, 1608 the compiler will now expand calls to e.g. strlen as an inlined 1609 sequences which utilize the cmp/str instruction. 1610 * Improved code generated around volatile memory loads and stores. 1611 * The option -mcbranchdi has been deprecated. Specifying it will 1612 result in a warning and will not influence code generation. 1613 * The option -mcmpeqdi has been deprecated. Specifying it will result 1614 in a warning and will not influence code generation. 1615 1616 GCC 4.9.1 1617 1618 This is the [25]list of problem reports (PRs) from GCC's bug tracking 1619 system that are known to be fixed in the 4.9.1 release. This list might 1620 not be complete (that is, it is possible that some PRs that have been 1621 fixed are not listed here). 1622 1623 Version 4.0 of the [26]OpenMP specification is supported even in 1624 Fortran, not just C and C++. 1625 1626 GCC 4.9.2 1627 1628 This is the [27]list of problem reports (PRs) from GCC's bug tracking 1629 system that are known to be fixed in the 4.9.2 release. This list might 1630 not be complete (that is, it is possible that some PRs that have been 1631 fixed are not listed here). 1632 1633 GCC 4.9.3 1634 1635 This is the [28]list of problem reports (PRs) from GCC's bug tracking 1636 system that are known to be fixed in the 4.9.3 release. This list might 1637 not be complete (that is, it is possible that some PRs that have been 1638 fixed are not listed here). 1639 1640 1641 For questions related to the use of GCC, please consult these web 1642 pages and the [29]GCC manuals. If that fails, the 1643 [30]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1644 web pages and the development of GCC are welcome on our developer 1645 list at [31]gcc (a] gcc.gnu.org. All of [32]our lists have public 1646 archives. 1647 1648 Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and 1649 distribution of this entire article is permitted in any medium, 1650 provided this notice is preserved. 1651 1652 These pages are [34]maintained by the GCC team. Last modified 1653 2015-06-28[35]. 1654 1655 References 1656 1657 1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html 1658 2. https://gcc.gnu.org/PR60825 1659 3. https://gcc.gnu.org/gcc-4.9/porting_to.html 1660 4. http://openmp.org/wp/openmp-specifications/ 1661 5. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908 1662 6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252 1663 7. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html 1664 8. https://www.cilkplus.org/ 1665 9. http://gcc.gnu.org/projects/cxx1y.html 1666 10. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html 1667 11. http://gcc.gnu.org/projects/cxx1y.html 1668 12. http://gcc.gnu.org/projects/cxx1y.html 1669 13. http://gcc.gnu.org/projects/cxx1y.html 1670 14. http://gcc.gnu.org/projects/cxx1y.html 1671 15. http://gcc.gnu.org/projects/cxx1y.html 1672 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf 1673 17. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011 1674 18. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014 1675 19. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html 1676 20. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html 1677 21. https://gcc.gnu.org/wiki/Fortran2003Status 1678 22. https://gcc.gnu.org/wiki/Fortran2008Status 1679 23. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html 1680 24. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html 1681 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1 1682 26. http://openmp.org/wp/openmp-specifications/ 1683 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2 1684 28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3 1685 29. https://gcc.gnu.org/onlinedocs/ 1686 30. mailto:gcc-help (a] gcc.gnu.org 1687 31. mailto:gcc (a] gcc.gnu.org 1688 32. https://gcc.gnu.org/lists.html 1689 33. http://www.fsf.org/ 1690 34. https://gcc.gnu.org/about.html 1691 35. http://validator.w3.org/check/referer 1692 ====================================================================== 1693 http://gcc.gnu.org/gcc-4.8/index.html 1694 1695 GCC 4.8 Release Series 1696 1697 June 23, 2015 1698 1699 The [1]GNU project and the GCC developers are pleased to announce the 1700 release of GCC 4.8.5. 1701 1702 This release is a bug-fix release, containing fixes for regressions in 1703 GCC 4.8.4 relative to previous releases of GCC. 1704 1705 Release History 1706 1707 GCC 4.8.5 1708 June 23, 2015 ([2]changes, [3]documentation) 1709 1710 GCC 4.8.4 1711 December 19, 2014 ([4]changes, [5]documentation) 1712 1713 GCC 4.8.3 1714 May 22, 2014 ([6]changes, [7]documentation) 1715 1716 GCC 4.8.2 1717 October 16, 2013 ([8]changes, [9]documentation) 1718 1719 GCC 4.8.1 1720 May 31, 2013 ([10]changes, [11]documentation) 1721 1722 GCC 4.8.0 1723 March 22, 2013 ([12]changes, [13]documentation) 1724 1725 References and Acknowledgements 1726 1727 GCC used to stand for the GNU C Compiler, but since the compiler 1728 supports several other languages aside from C, it now stands for the 1729 GNU Compiler Collection. 1730 1731 A list of [14]successful builds is updated as new information becomes 1732 available. 1733 1734 The GCC developers would like to thank the numerous people that have 1735 contributed new features, improvements, bug fixes, and other changes as 1736 well as test results to GCC. This [15]amazing group of volunteers is 1737 what makes GCC successful. 1738 1739 For additional information about GCC please refer to the [16]GCC 1740 project web site or contact the [17]GCC development mailing list. 1741 1742 To obtain GCC please use [18]our mirror sites or [19]our SVN server. 1743 1744 1745 For questions related to the use of GCC, please consult these web 1746 pages and the [20]GCC manuals. If that fails, the 1747 [21]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1748 web pages and the development of GCC are welcome on our developer 1749 list at [22]gcc (a] gcc.gnu.org. All of [23]our lists have public 1750 archives. 1751 1752 Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and 1753 distribution of this entire article is permitted in any medium, 1754 provided this notice is preserved. 1755 1756 These pages are [25]maintained by the GCC team. Last modified 1757 2015-06-23[26]. 1758 1759 References 1760 1761 1. http://www.gnu.org/ 1762 2. http://gcc.gnu.org/gcc-4.8/changes.html 1763 3. https://gcc.gnu.org/onlinedocs/4.8.5/ 1764 4. http://gcc.gnu.org/gcc-4.8/changes.html 1765 5. https://gcc.gnu.org/onlinedocs/4.8.4/ 1766 6. http://gcc.gnu.org/gcc-4.8/changes.html 1767 7. https://gcc.gnu.org/onlinedocs/4.8.3/ 1768 8. http://gcc.gnu.org/gcc-4.8/changes.html 1769 9. https://gcc.gnu.org/onlinedocs/4.8.2/ 1770 10. http://gcc.gnu.org/gcc-4.8/changes.html 1771 11. https://gcc.gnu.org/onlinedocs/4.8.1/ 1772 12. http://gcc.gnu.org/gcc-4.8/changes.html 1773 13. https://gcc.gnu.org/onlinedocs/4.8.0/ 1774 14. http://gcc.gnu.org/gcc-4.8/buildstat.html 1775 15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 1776 16. http://gcc.gnu.org/index.html 1777 17. mailto:gcc (a] gcc.gnu.org 1778 18. http://gcc.gnu.org/mirrors.html 1779 19. http://gcc.gnu.org/svn.html 1780 20. https://gcc.gnu.org/onlinedocs/ 1781 21. mailto:gcc-help (a] gcc.gnu.org 1782 22. mailto:gcc (a] gcc.gnu.org 1783 23. https://gcc.gnu.org/lists.html 1784 24. http://www.fsf.org/ 1785 25. https://gcc.gnu.org/about.html 1786 26. http://validator.w3.org/check/referer 1787 ====================================================================== 1788 http://gcc.gnu.org/gcc-4.8/changes.html 1789 1790 GCC 4.8 Release Series 1791 Changes, New Features, and Fixes 1792 1793 Caveats 1794 1795 GCC now uses C++ as its implementation language. This means that to 1796 build GCC from sources, you will need a C++ compiler that understands 1797 C++ 2003. For more details on the rationale and specific changes, 1798 please refer to the [1]C++ conversion page. 1799 1800 To enable the Graphite framework for loop optimizations you now need 1801 CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from 1802 the [2]GCC infrastructure directory. The installation manual contains 1803 more information about requirements to build GCC. 1804 1805 GCC now uses a more aggressive analysis to derive an upper bound for 1806 the number of iterations of loops using constraints imposed by language 1807 standards. This may cause non-conforming programs to no longer work as 1808 expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new 1809 option, -fno-aggressive-loop-optimizations, was added to disable this 1810 aggressive analysis. In some loops that have known constant number of 1811 iterations, but undefined behavior is known to occur in the loop before 1812 reaching or during the last iteration, GCC will warn about the 1813 undefined behavior in the loop instead of deriving lower upper bound of 1814 the number of iterations for the loop. The warning can be disabled with 1815 -Wno-aggressive-loop-optimizations. 1816 1817 On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules 1818 for the layout of vectors that could lead to wrong code being 1819 generated. Vectors larger than 8 bytes in size are now by default 1820 aligned to an 8-byte boundary. This is an ABI change: code that makes 1821 explicit use of vector types may be incompatible with binary objects 1822 built with older versions of GCC. Auto-vectorized code is not affected 1823 by this change. 1824 1825 On AVR, support has been removed for the command-line option 1826 -mshort-calls deprecated in GCC 4.7. 1827 1828 On AVR, the configure option --with-avrlibc supported since GCC 4.7.2 1829 is turned on per default for all non-RTEMS configurations. This option 1830 arranges for a better integration of [3]AVR Libc with avr-gcc. For 1831 technical details, see [4]PR54461. To turn off the option in non-RTEMS 1832 configurations, use --with-avrlibc=no. If the compiler is configured 1833 for RTEMS, the option is always turned off. 1834 1835 More information on porting to GCC 4.8 from previous versions of GCC 1836 can be found in the [5]porting guide for this release. 1837 1838 General Optimizer Improvements (and Changes) 1839 1840 * DWARF4 is now the default when generating DWARF debug information. 1841 When -g is used on a platform that uses DWARF debugging 1842 information, GCC will now default to -gdwarf-4 1843 -fno-debug-types-section. 1844 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information 1845 consumers support DWARF4 by default. Before GCC 4.8 the default 1846 version used was DWARF2. To make GCC 4.8 generate an older DWARF 1847 version use -g together with -gdwarf-2 or -gdwarf-3. The default 1848 for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf. 1849 * A new general optimization level, -Og, has been introduced. It 1850 addresses the need for fast compilation and a superior debugging 1851 experience while providing a reasonable level of run-time 1852 performance. Overall experience for development should be better 1853 than the default optimization level -O0. 1854 * A new option -ftree-partial-pre was added to control the partial 1855 redundancy elimination (PRE) optimization. This option is enabled 1856 by default at the -O3 optimization level, and it makes PRE more 1857 aggressive. 1858 * The option -fconserve-space has been removed; it was no longer 1859 useful on most targets since GCC supports putting variables into 1860 BSS without making them common. 1861 * The struct reorg and matrix reorg optimizations (command-line 1862 options -fipa-struct-reorg and -fipa-matrix-reorg) have been 1863 removed. They did not always work correctly, nor did they work with 1864 link-time optimization (LTO), hence were only applicable to 1865 programs consisting of a single translation unit. 1866 * Several scalability bottle-necks have been removed from GCC's 1867 optimization passes. Compilation of extremely large functions, e.g. 1868 due to the use of the flatten attribute in the "Eigen" C++ linear 1869 algebra templates library, is significantly faster than previous 1870 releases of GCC. 1871 * Link-time optimization (LTO) improvements: 1872 + LTO partitioning has been rewritten for better reliability and 1873 maintanibility. Several important bugs leading to link 1874 failures have been fixed. 1875 * Interprocedural optimization improvements: 1876 + A new symbol table has been implemented. It builds on existing 1877 callgraph and varpool modules and provide a new API. Unusual 1878 symbol visibilities and aliases are handled more consistently 1879 leading to, for example, more aggressive unreachable code 1880 removal with LTO. 1881 + The inline heuristic can now bypass limits on the size of of 1882 inlined functions when the inlining is particularly 1883 profitable. This happens, for example, when loop bounds or 1884 array strides get propagated. 1885 + Values passed through aggregates (either by value or 1886 reference) are now propagated at the inter-procedural level 1887 leading to better inlining decisions (for example in the case 1888 of Fortran array descriptors) and devirtualization. 1889 * [6]AddressSanitizer , a fast memory error detector, has been added 1890 and can be enabled via -fsanitize=address. Memory access 1891 instructions will be instrumented to detect heap-, stack-, and 1892 global-buffer overflow as well as use-after-free bugs. To get nicer 1893 stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is 1894 available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on 1895 x86-64 Darwin. 1896 * [7]ThreadSanitizer has been added and can be enabled via 1897 -fsanitize=thread. Instructions will be instrumented to detect data 1898 races. The ThreadSanitizer is available on x86-64 GNU/Linux. 1899 * A new local register allocator (LRA) has been implemented, which 1900 replaces the 26 year old reload pass and improves generated code 1901 quality. For now it is active on the IA-32 and x86-64 targets. 1902 * Support for transactional memory has been implemented on the 1903 following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and 1904 Alpha. 1905 1906 New Languages and Language specific improvements 1907 1908 C family 1909 1910 * Each diagnostic emitted now includes the original source line and a 1911 caret '^' indicating the column. The option 1912 -fno-diagnostics-show-caret suppresses this information. 1913 * The option -ftrack-macro-expansion=2 is now enabled by default. 1914 This allows the compiler to display the macro expansion stack in 1915 diagnostics. Combined with the caret information, an example 1916 diagnostic showing these two features is: 1917 1918 t.c:1:94: error: invalid operands to binary < (have `struct mystruct' and `float 1919 ') 1920 #define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _ 1921 _b = (B); __a < __b ? __b : __a; }) 1922 1923 ^ 1924 t.c:7:7: note: in expansion of macro 'MYMAX' 1925 X = MYMAX(P, F); 1926 ^ 1927 1928 * A new -Wsizeof-pointer-memaccess warning has been added (also 1929 enabled by -Wall) to warn about suspicious length parameters to 1930 certain string and memory built-in functions if the argument uses 1931 sizeof. This warning warns e.g. about memset (ptr, 0, sizeof 1932 (ptr)); if ptr is not an array, but a pointer, and suggests a 1933 possible fix, or about memcpy (&foo, ptr, sizeof (&foo));. 1934 * The new option -Wpedantic is an alias for -pedantic, which is now 1935 deprecated. The forms -Wno-pedantic, -Werror=pedantic, and 1936 -Wno-error=pedantic work in the same way as for any other -W 1937 option. One caveat is that -Werror=pedantic is not equivalent to 1938 -pedantic-errors, since the latter makes into errors some warnings 1939 that are not controlled by -Wpedantic, and the former only affects 1940 diagnostics that are disabled when using -Wno-pedantic. 1941 * The option -Wshadow no longer warns if a declaration shadows a 1942 function declaration, unless the former declares a function or 1943 pointer to function, because this is [8]a common and valid case in 1944 real-world code. 1945 1946 C++ 1947 1948 * G++ now implements the [9]C++11 thread_local keyword; this differs 1949 from the GNU __thread keyword primarily in that it allows dynamic 1950 initialization and destruction semantics. Unfortunately, this 1951 support requires a run-time penalty for references to 1952 non-function-local thread_local variables defined in a different 1953 translation unit even if they don't need dynamic initialization, so 1954 users may want to continue to use __thread for TLS variables with 1955 static initialization semantics. 1956 If the programmer can be sure that no use of the variable in a 1957 non-defining TU needs to trigger dynamic initialization (either 1958 because the variable is statically initialized, or a use of the 1959 variable in the defining TU will be executed before any uses in 1960 another TU), they can avoid this overhead with the 1961 -fno-extern-tls-init option. 1962 OpenMP threadprivate variables now also support dynamic 1963 initialization and destruction by the same mechanism. 1964 * G++ now implements the [10]C++11 attribute syntax, e.g. 1965 1966 [[noreturn]] void f(); 1967 1968 and also the alignment specifier, e.g. 1969 1970 alignas(double) int i; 1971 1972 * G++ now implements [11]C++11 inheriting constructors, e.g. 1973 1974 struct A { A(int); }; 1975 struct B: A { using A::A; }; // defines B::B(int) 1976 B b(42); // OK 1977 1978 * As of GCC 4.8.1, G++ implements the change to decltype semantics 1979 from [12]N3276. 1980 1981 struct A f(); 1982 decltype(f()) g(); // OK, return type of f() is not required to be complete. 1983 1984 * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g. 1985 1986 struct A { int f() &; }; 1987 int i = A().f(); // error, f() requires an lvalue object 1988 1989 * G++ now supports a -std=c++1y option for experimentation with 1990 features proposed for the next revision of the standard, expected 1991 around 2014. Currently the only difference from -std=c++11 is 1992 support for return type deduction in normal functions, as proposed 1993 in [14]N3386. Status of C++1y features in GCC 4.8 can be found 1994 [15]here. 1995 * The G++ namespace association extension, __attribute ((strong)), 1996 has been deprecated. Inline namespaces should be used instead. 1997 * G++ now supports a -fext-numeric-literal option to control whether 1998 GNU numeric literal suffixes are accepted as extensions or 1999 processed as C++11 user-defined numeric literal suffixes. The flag 2000 is on (use suffixes for GNU literals) by default for -std=gnu++*, 2001 and -std=c++98. The flag is off (use suffixes for user-defined 2002 literals) by default for -std=c++11 and later. 2003 2004 Runtime Library (libstdc++) 2005 2006 * [16]Improved experimental support for the new ISO C++ standard, 2007 C++11, including: 2008 + forward_list meets the allocator-aware container requirements; 2009 + this_thread::sleep_for(), this_thread::sleep_until() and 2010 this_thread::yield() are defined without requiring the 2011 configure option --enable-libstdcxx-time; 2012 * Improvements to <random>: 2013 + SSE optimized normal_distribution. 2014 + Use of hardware RNG instruction for random_device on new x86 2015 processors (requires the assembler to support the 2016 instruction.) 2017 and <ext/random>: 2018 + New random number engine simd_fast_mersenne_twister_engine 2019 with an optimized SSE implementation. 2020 + New random number distributions beta_distribution, 2021 normal_mv_distribution, rice_distribution, 2022 nakagami_distribution, pareto_distribution, k_distribution, 2023 arcsine_distribution, hoyt_distribution. 2024 * Added --disable-libstdcxx-verbose configure option to disable 2025 diagnostic messages issued when a process terminates abnormally. 2026 This may be useful for embedded systems to reduce the size of 2027 executables that link statically to the library. 2028 2029 Fortran 2030 2031 * Compatibility notice: 2032 + Module files: The version of module files (.mod) has been 2033 incremented. Fortran MODULEs compiled by earlier GCC versions 2034 have to be recompiled, when they are USEd by files compiled 2035 with GCC 4.8. GCC 4.8 is not able to read .mod files created 2036 by earlier versions; attempting to do so gives an error 2037 message. 2038 Note: The ABI of the produced assembler data itself has not 2039 changed; object files and libraries are fully compatible with 2040 older versions except as noted below. 2041 + ABI: Some internal names (used in the assembler/object file) 2042 have changed for symbols declared in the specification part of 2043 a module. If an affected module - or a file using it via use 2044 association - is recompiled, the module and all files which 2045 directly use such symbols have to be recompiled as well. This 2046 change only affects the following kind of module symbols: 2047 o Procedure pointers. Note: C-interoperable function 2048 pointers (type(c_funptr)) are not affected nor are 2049 procedure-pointer components. 2050 o Deferred-length character strings. 2051 * The [17]BACKTRACE intrinsic subroutine has been added. It shows a 2052 backtrace at an arbitrary place in user code; program execution 2053 continues normally afterwards. 2054 * The [18]-Wc-binding-type warning option has been added (disabled by 2055 default). It warns if the a variable might not be C interoperable; 2056 in particular, if the variable has been declared using an intrinsic 2057 type with default kind instead of using a kind parameter defined 2058 for C interoperability in the intrinsic ISO_C_Binding module. 2059 Before, this warning was always printed. The -Wc-binding-type 2060 option is enabled by -Wall. 2061 * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line 2062 options have been added, which diagnose when code is inserted for 2063 automatic (re)allocation of a variable during assignment. This 2064 option can be used to decide whether it is safe to use 2065 [20]-fno-realloc-lhs. Additionally, it can be used to find 2066 automatic (re)allocation in hot loops. (For arrays, replacing 2067 "var=" by "var(:)=" disables the automatic reallocation.) 2068 * The [21]-Wcompare-reals command-line option has been added. When 2069 this is set, warnings are issued when comparing REAL or COMPLEX 2070 types for equality and inequality; consider replacing a == b by 2071 abs(a -b) < eps with a suitable eps. -Wcompare-reals is enabled by 2072 -Wextra. 2073 * The [22]-Wtarget-lifetime command-line option has been added 2074 (enabled with -Wall), which warns if the pointer in a pointer 2075 assignment might outlive its target. 2076 * Reading floating point numbers which use "q" for the exponential 2077 (such as 4.0q0) is now supported as vendor extension for better 2078 compatibility with old data files. It is strongly recommended to 2079 use for I/O the equivalent but standard conforming "e" (such as 2080 4.0e0). 2081 (For Fortran source code, consider replacing the "q" in 2082 floating-point literals by a kind parameter (e.g. 4.0e0_qp with a 2083 suitable qp). Note that - in Fortran source code - replacing "q" by 2084 a simple "e" is not equivalent.) 2085 * The GFORTRAN_TMPDIR environment variable for specifying a 2086 non-default directory for files opened with STATUS="SCRATCH", is 2087 not used anymore. Instead gfortran checks the POSIX/GNU standard 2088 TMPDIR environment variable. If TMPDIR is not defined, gfortran 2089 falls back to other methods to determine the directory for 2090 temporary files as documented in the [23]user manual. 2091 * [24]Fortran 2003: 2092 + Support for unlimited polymorphic variables (CLASS(*)) has 2093 been added. Nonconstant character lengths are not yet 2094 supported. 2095 * [25]TS 29113: 2096 + Assumed types (TYPE(*)) are now supported. 2097 + Experimental support for assumed-rank arrays (dimension(..)) 2098 has been added. Note that currently gfortran's own array 2099 descriptor is used, which is different from the one defined in 2100 TS29113, see [26]gfortran's header file or use the [27]Chasm 2101 Language Interoperability Tools. 2102 2103 Go 2104 2105 * GCC 4.8.2 provides a complete implementation of the Go 1.1.2 2106 release. 2107 * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1 2108 release. The library support is not quite complete. 2109 * Go has been tested on GNU/Linux and Solaris platforms for various 2110 processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may 2111 work on other platforms as well. 2112 2113 New Targets and Target Specific Improvements 2114 2115 AArch64 2116 2117 * A new port has been added to support AArch64, the new 64-bit 2118 architecture from ARM. Note that this is a separate port from the 2119 existing 32-bit ARM port. 2120 * The port provides initial support for the Cortex-A53 and the 2121 Cortex-A57 processors with the command line options 2122 -mcpu=cortex-a53 and -mcpu=cortex-a57. 2123 * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769 2124 has been added and can be enabled by giving the 2125 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 2126 default by configuring GCC with the --enable-fix-cortex-a53-835769 2127 option. 2128 2129 ARM 2130 2131 * Initial support has been added for the AArch32 extensions defined 2132 in the ARMv8 architecture. 2133 * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs. 2134 * A new option, -mcpu=marvell-pj4, has been added to generate code 2135 for the Marvell PJ4 processor. 2136 * The compiler can now automatically generate the VFMA, VFMS, REVSH 2137 and REV16 instructions. 2138 * A new vectorizer cost model for Advanced SIMD configurations to 2139 improve the auto-vectorization strategies used. 2140 * The scheduler now takes into account the number of live registers 2141 to reduce the amount of spilling that can occur. This should 2142 improve code performance in large functions. The limit can be 2143 removed by using the option -fno-sched-pressure. 2144 * Improvements have been made to the Marvell iWMMX code generation 2145 and support for the iWMMX2 SIMD unit has been added. The option 2146 -mcpu=iwmmxt2 can be used to enable code generation for the latter. 2147 * A number of code generation improvements for Thumb2 to reduce code 2148 size when compiling for the M-profile processors. 2149 * The RTEMS (arm-rtems) port has been updated to use the EABI. 2150 * Code generation support for the old FPA and Maverick floating-point 2151 architectures has been removed. Ports that previously relied on 2152 these features have also been removed. This includes the targets: 2153 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 2154 + arm*-*-elf (use arm*-*-eabi) 2155 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 2156 + arm*-*-ecos-elf (no alternative) 2157 + arm*-*-freebsd (no alternative) 2158 + arm*-wince-pe* (no alternative). 2159 2160 AVR 2161 2162 * Support for the "Embedded C" fixed-point has been added. For 2163 details, see the [28]GCC wiki and the [29]user manual. The support 2164 is not complete. 2165 * A new print modifier %r for register operands in inline assembler 2166 is supported. It will print the raw register number without the 2167 register prefix 'r': 2168 /* Return the most significant byte of 'val', a 64-bit value. */ 2169 2170 unsigned char msb (long long val) 2171 { 2172 unsigned char c; 2173 __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val)); 2174 return c; 2175 } 2176 The inline assembler in this example will generate code like 2177 mov r24, 8+7 2178 provided c is allocated to R24 and val is allocated to R8...R15. 2179 This works because the GNU assembler accepts plain register numbers 2180 without register prefix. 2181 * Static initializers with 3-byte symbols are supported now: 2182 extern const __memx char foo; 2183 const __memx void *pfoo = &foo; 2184 This requires at least Binutils 2.23. 2185 2186 IA-32/x86-64 2187 2188 * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with 2189 SSE extensions disabled. Since the x86-64 ABI requires 16 byte 2190 stack alignment, this is ABI incompatible and intended to be used 2191 in controlled environments where stack space is an important 2192 limitation. This option will lead to wrong code when functions 2193 compiled with 16 byte stack alignment (such as functions from a 2194 standard library) are called with misaligned stack. In this case, 2195 SSE instructions may lead to misaligned memory access traps. In 2196 addition, variable arguments will be handled incorrectly for 16 2197 byte aligned objects (including x87 long double and __int128), 2198 leading to wrong results. You must build all modules with 2199 -mpreferred-stack-boundary=3, including any libraries. This 2200 includes the system libraries and startup modules. 2201 * Support for the new Intel processor codename Broadwell with RDSEED, 2202 ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw, 2203 -mrdseed command-line options. 2204 * Support for the Intel RTM and HLE intrinsics, built-in functions 2205 and code generation is available via -mrtm and -mhle. 2206 * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets. 2207 Intrinsics and built-in functions are available via -mfxsr, -mxsave 2208 and -mxsaveopt respectively. 2209 * New -maddress-mode=[short|long] options for x32. 2210 -maddress-mode=short overrides default 64-bit addresses to 32-bit 2211 by emitting the 0x67 address-size override prefix. This is the 2212 default address mode for x32. 2213 * New built-in functions to detect run-time CPU type and ISA: 2214 + A built-in function __builtin_cpu_is has been added to detect 2215 if the run-time CPU is of a particular type. It returns a 2216 positive integer on a match and zero otherwise. It accepts one 2217 string literal argument, the CPU name. For example, 2218 __builtin_cpu_is("westmere") returns a positive integer if the 2219 run-time CPU is an Intel Core i7 Westmere processor. Please 2220 refer to the [30]user manual for the list of valid CPU names 2221 recognized. 2222 + A built-in function __builtin_cpu_supports has been added to 2223 detect if the run-time CPU supports a particular ISA feature. 2224 It returns a positive integer on a match and zero otherwise. 2225 It accepts one string literal argument, the ISA feature. For 2226 example, __builtin_cpu_supports("ssse3") returns a positive 2227 integer if the run-time CPU supports SSSE3 instructions. 2228 Please refer to the [31]user manual for the list of valid ISA 2229 names recognized. 2230 Caveat: If these built-in functions are called before any static 2231 constructors are invoked, like during IFUNC initialization, then 2232 the CPU detection initialization must be explicitly run using this 2233 newly provided built-in function, __builtin_cpu_init. The 2234 initialization needs to be done only once. For example, this is how 2235 the invocation would look like inside an IFUNC initializer: 2236 static void (*some_ifunc_resolver(void))(void) 2237 { 2238 __builtin_cpu_init(); 2239 if (__builtin_cpu_is("amdfam10h") ... 2240 if (__builtin_cpu_supports("popcnt") ... 2241 } 2242 2243 * Function Multiversioning Support with G++: 2244 It is now possible to create multiple function versions each 2245 targeting a specific processor and/or ISA. Function versions have 2246 the same signature but different target attributes. For example, 2247 here is a program with function versions: 2248 __attribute__ ((target ("default"))) 2249 int foo(void) 2250 { 2251 return 1; 2252 } 2253 2254 __attribute__ ((target ("sse4.2"))) 2255 int foo(void) 2256 { 2257 return 2; 2258 } 2259 2260 int main (void) 2261 { 2262 int (*p) = &foo; 2263 assert ((*p)() == foo()); 2264 return 0; 2265 } 2266 2267 Please refer to this [32]wiki for more information. 2268 * The x86 back end has been improved to allow option -fschedule-insns 2269 to work reliably. This option can be used to schedule instructions 2270 better and leads to improved performace in certain cases. 2271 * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437 2272 from the Mingw-w64 trunk. 2273 * Support for new AMD family 15h processors (Steamroller core) is now 2274 available through the -march=bdver3 and -mtune=bdver3 options. 2275 * Support for new AMD family 16h processors (Jaguar core) is now 2276 available through the -march=btver2 and -mtune=btver2 options. 2277 2278 FRV 2279 2280 * This target now supports the -fstack-usage command-line option. 2281 2282 MIPS 2283 2284 * GCC can now generate code specifically for the R4700, Broadcom XLP 2285 and MIPS 34kn processors. The associated -march options are 2286 -march=r4700, -march=xlp and -march=34kn respectively. 2287 * GCC now generates better DSP code for MIPS 74k cores thanks to 2288 further scheduling optimizations. 2289 * The MIPS port now supports the -fstack-check option. 2290 * GCC now passes the -mmcu and -mno-mcu options to the assembler. 2291 * Previous versions of GCC would silently accept -fpic and -fPIC for 2292 -mno-abicalls targets like mips*-elf. This combination was not 2293 intended or supported, and did not generate position-independent 2294 code. GCC 4.8 now reports an error when this combination is used. 2295 2296 PowerPC / PowerPC64 / RS6000 2297 2298 * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save, 2299 restore or update the VRSAVE register by default. The respective 2300 operating systems manage the VRSAVE register directly. 2301 * Large TOC support has been added for AIX through the command line 2302 option -mcmodel=large. 2303 * Native Thread-Local Storage support has been added for AIX. 2304 * VMX (Altivec) and VSX instruction sets now are enabled implicitly 2305 when targetting processors that support those hardware features on 2306 AIX 6.1 and above. 2307 2308 RX 2309 2310 * This target will now issue a warning message whenever multiple fast 2311 interrupt handlers are found in the same compilation unit. This 2312 feature can be turned off by the new 2313 -mno-warn-multiple-fast-interrupts command-line option. 2314 2315 S/390, System z 2316 2317 * Support for the IBM zEnterprise zEC12 processor has been added. 2318 When using the -march=zEC12 option, the compiler will generate code 2319 making use of the following new instructions: 2320 + load and trap instructions 2321 + 2 new compare and trap instructions 2322 + rotate and insert selected bits - without CC clobber 2323 The -mtune=zEC12 option enables zEC12 specific instruction 2324 scheduling without making use of new instructions. 2325 * Register pressure sensitive instruction scheduling is enabled by 2326 default. 2327 * The ifunc function attribute is enabled by default. 2328 * memcpy and memcmp invokations on big memory chunks or with run time 2329 lengths are not generated inline anymore when tuning for z10 or 2330 higher. The purpose is to make use of the IFUNC optimized versions 2331 in Glibc. 2332 2333 SH 2334 2335 * The default alignment settings have been reduced to be less 2336 aggressive. This results in more compact code for optimization 2337 levels other than -Os. 2338 * Improved support for the __atomic built-in functions: 2339 + A new option -matomic-model=model selects the model for the 2340 generated atomic sequences. The following models are 2341 supported: 2342 2343 soft-gusa 2344 Software gUSA sequences (SH3* and SH4* only). On 2345 SH4A targets this will now also partially utilize 2346 the movco.l and movli.l instructions. This is the 2347 default when the target is sh3*-*-linux* or 2348 sh4*-*-linux*. 2349 2350 hard-llcs 2351 Hardware movco.l / movli.l sequences (SH4A only). 2352 2353 soft-tcb 2354 Software thread control block sequences. 2355 2356 soft-imask 2357 Software interrupt flipping sequences (privileged 2358 mode only). This is the default when the target is 2359 sh1*-*-linux* or sh2*-*-linux*. 2360 2361 none 2362 Generates function calls to the respective __atomic 2363 built-in functions. This is the default for SH64 2364 targets or when the target is not sh*-*-linux*. 2365 2366 + The option -msoft-atomic has been deprecated. It is now an 2367 alias for -matomic-model=soft-gusa. 2368 + A new option -mtas makes the compiler generate the tas.b 2369 instruction for the __atomic_test_and_set built-in function 2370 regardless of the selected atomic model. 2371 + The __sync functions in libgcc now reflect the selected atomic 2372 model when building the toolchain. 2373 * Added support for the mov.b and mov.w instructions with 2374 displacement addressing. 2375 * Added support for the SH2A instructions movu.b and movu.w. 2376 * Various improvements to code generated for integer arithmetic. 2377 * Improvements to conditional branches and code that involves the T 2378 bit. A new option -mzdcbranch tells the compiler to favor 2379 zero-displacement branches. This is enabled by default for SH4* 2380 targets. 2381 * The pref instruction will now be emitted by the __builtin_prefetch 2382 built-in function for SH3* targets. 2383 * The fmac instruction will now be emitted by the fmaf standard 2384 function and the __builtin_fmaf built-in function. 2385 * The -mfused-madd option has been deprecated in favor of the 2386 machine-independent -ffp-contract option. Notice that the fmac 2387 instruction will now be generated by default for expressions like a 2388 * b + c. This is due to the compiler default setting 2389 -ffp-contract=fast. 2390 * Added new options -mfsrra and -mfsca to allow the compiler using 2391 the fsrra and fsca instructions on targets other than SH4A (where 2392 they are already enabled by default). 2393 * Added support for the __builtin_bswap32 built-in function. It is 2394 now expanded as a sequence of swap.b and swap.w instructions 2395 instead of a library function call. 2396 * The behavior of the -mieee option has been fixed and the negative 2397 form -mno-ieee has been added to control the IEEE conformance of 2398 floating point comparisons. By default -mieee is now enabled and 2399 the option -ffinite-math-only implicitly sets -mno-ieee. 2400 * Added support for the built-in functions __builtin_thread_pointer 2401 and __builtin_set_thread_pointer. This assumes that GBR is used to 2402 hold the thread pointer of the current thread. Memory loads and 2403 stores relative to the address returned by __builtin_thread_pointer 2404 will now also utilize GBR based displacement address modes. 2405 * The -mdiv= option for targets other than SHmedia has been fixed and 2406 documented. 2407 2408 SPARC 2409 2410 * Added optimized instruction scheduling for Niagara4. 2411 2412 TILE-Gx 2413 2414 * Added support for the -mcmodel=MODEL command-line option. The 2415 models supported are small and large. 2416 2417 V850 2418 2419 * This target now supports the E3V5 architecture via the use of the 2420 new -mv850e3v5 command-line option. It also has experimental 2421 support for the e3v5 LOOP instruction which can be enabled via the 2422 new -mloop command-line option. 2423 2424 XStormy16 2425 2426 * This target now supports the -fstack-usage command-line option. 2427 2428 Operating Systems 2429 2430 Windows (Cygwin) 2431 2432 * Executables are now linked against shared libgcc by default. The 2433 previous default was to link statically, which can still be done by 2434 explicitly specifying -static or static-libgcc on the command line. 2435 However it is strongly advised against, as it will cause problems 2436 for any application that makes use of DLLs compiled by GCC. It 2437 should be alright for a monolithic stand-alone application that 2438 only links against the Windows DLLs, but offers little or no 2439 benefit. 2440 2441 GCC 4.8.1 2442 2443 This is the [33]list of problem reports (PRs) from GCC's bug tracking 2444 system that are known to be fixed in the 4.8.1 release. This list might 2445 not be complete (that is, it is possible that some PRs that have been 2446 fixed are not listed here). 2447 2448 The C++11 <chrono> std::chrono::system_clock and 2449 std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they 2450 both are now separate (never typedefs of each other), both use 2451 std::chrono::nanoseconds resolution, on most GNU/Linux configurations 2452 std::chrono::steady_clock is now finally monotonic, and both classes 2453 are mangled differently than in the previous GCC releases. 2454 std::chrono::system_clock::now() with std::chrono::microseconds resp. 2455 std::chrono::seconds resolution is still exported for backwards 2456 compatibility with default configured libstdc++. Note that libstdc++ 2457 configured with --enable-libstdcxx-time= used to be ABI incompatible 2458 with default configured libstdc++ for those two classes and no ABI 2459 compatibility can be offered for those configurations, so any C++11 2460 code that uses those classes and has been compiled and linked against 2461 libstdc++ configured with the non-default --enable-libstdcxx-time= 2462 configuration option needs to be recompiled. 2463 2464 GCC 4.8.2 2465 2466 This is the [34]list of problem reports (PRs) from GCC's bug tracking 2467 system that are known to be fixed in the 4.8.2 release. This list might 2468 not be complete (that is, it is possible that some PRs that have been 2469 fixed are not listed here). 2470 2471 GCC 4.8.3 2472 2473 This is the [35]list of problem reports (PRs) from GCC's bug tracking 2474 system that are known to be fixed in the 4.8.3 release. This list might 2475 not be complete (that is, it is possible that some PRs that have been 2476 fixed are not listed here). 2477 2478 Support for the new powerpc64le-linux platform has been added. It 2479 defaults to generating code that conforms to the ELFV2 ABI. 2480 2481 GCC 4.8.4 2482 2483 This is the [36]list of problem reports (PRs) from GCC's bug tracking 2484 system that are known to be fixed in the 4.8.4 release. This list might 2485 not be complete (that is, it is possible that some PRs that have been 2486 fixed are not listed here). 2487 2488 GCC 4.8.5 2489 2490 This is the [37]list of problem reports (PRs) from GCC's bug tracking 2491 system that are known to be fixed in the 4.8.5 release. This list might 2492 not be complete (that is, it is possible that some PRs that have been 2493 fixed are not listed here). 2494 2495 2496 For questions related to the use of GCC, please consult these web 2497 pages and the [38]GCC manuals. If that fails, the 2498 [39]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2499 web pages and the development of GCC are welcome on our developer 2500 list at [40]gcc (a] gcc.gnu.org. All of [41]our lists have public 2501 archives. 2502 2503 Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and 2504 distribution of this entire article is permitted in any medium, 2505 provided this notice is preserved. 2506 2507 These pages are [43]maintained by the GCC team. Last modified 2508 2015-11-14[44]. 2509 2510 References 2511 2512 1. https://gcc.gnu.org/wiki/cxx-conversion 2513 2. ftp://gcc.gnu.org/pub/gcc/infrastructure/ 2514 3. http://www.nongnu.org/avr-libc/ 2515 4. https://gcc.gnu.org/PR54461 2516 5. https://gcc.gnu.org/gcc-4.8/porting_to.html 2517 6. https://github.com/google/sanitizers 2518 7. https://code.google.com/p/data-race-test/wiki/ThreadSanitizer 2519 8. https://lkml.org/lkml/2006/11/28/239 2520 9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 2521 10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 2522 11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 2523 12. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf 2524 13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 2525 14. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html 2526 15. http://gcc.gnu.org/projects/cxx1y.html 2527 16. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011 2528 17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html 2529 18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 2530 19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 2531 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 2532 21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 2533 22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 2534 23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html 2535 24. https://gcc.gnu.org/wiki/Fortran2003Status 2536 25. https://gcc.gnu.org/wiki/TS29113Status 2537 26. https://gcc.gnu.org/viewcvs/trunk/libgfortran/libgfortran.h?content-type=text%2Fplain&view=co 2538 27. http://chasm-interop.sourceforge.net/ 2539 28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support 2540 29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html 2541 30. https://gcc.gnu.org/onlinedocs/gcc/X86-Built-in-Functions.html#X86-Built-in-Functions 2542 31. https://gcc.gnu.org/onlinedocs/gcc/X86-Built-in-Functions.html#X86-Built-in-Functions 2543 32. https://gcc.gnu.org/wiki/FunctionMultiVersioning 2544 33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1 2545 34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2 2546 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3 2547 36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4 2548 37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5 2549 38. https://gcc.gnu.org/onlinedocs/ 2550 39. mailto:gcc-help (a] gcc.gnu.org 2551 40. mailto:gcc (a] gcc.gnu.org 2552 41. https://gcc.gnu.org/lists.html 2553 42. http://www.fsf.org/ 2554 43. https://gcc.gnu.org/about.html 2555 44. http://validator.w3.org/check/referer 2556 ====================================================================== 2557 http://gcc.gnu.org/gcc-4.7/index.html 2558 2559 GCC 4.7 Release Series 2560 2561 June 12, 2014 2562 2563 The [1]GNU project and the GCC developers are pleased to announce the 2564 release of GCC 4.7.4. 2565 2566 This release is a bug-fix release, containing fixes for regressions in 2567 GCC 4.7.3 relative to previous releases of GCC. 2568 2569 Release History 2570 2571 GCC 4.7.4 2572 June 12, 2014 ([2]changes, [3]documentation) 2573 2574 GCC 4.7.3 2575 April 11, 2013 ([4]changes, [5]documentation) 2576 2577 GCC 4.7.2 2578 September 20, 2012 ([6]changes, [7]documentation) 2579 2580 GCC 4.7.1 2581 June 14, 2012 ([8]changes, [9]documentation) 2582 2583 GCC 4.7.0 2584 March 22, 2012 ([10]changes, [11]documentation) 2585 2586 References and Acknowledgements 2587 2588 GCC used to stand for the GNU C Compiler, but since the compiler 2589 supports several other languages aside from C, it now stands for the 2590 GNU Compiler Collection. 2591 2592 A list of [12]successful builds is updated as new information becomes 2593 available. 2594 2595 The GCC developers would like to thank the numerous people that have 2596 contributed new features, improvements, bug fixes, and other changes as 2597 well as test results to GCC. This [13]amazing group of volunteers is 2598 what makes GCC successful. 2599 2600 For additional information about GCC please refer to the [14]GCC 2601 project web site or contact the [15]GCC development mailing list. 2602 2603 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 2604 2605 2606 For questions related to the use of GCC, please consult these web 2607 pages and the [18]GCC manuals. If that fails, the 2608 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2609 web pages and the development of GCC are welcome on our developer 2610 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 2611 archives. 2612 2613 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 2614 distribution of this entire article is permitted in any medium, 2615 provided this notice is preserved. 2616 2617 These pages are [23]maintained by the GCC team. Last modified 2618 2014-06-12[24]. 2619 2620 References 2621 2622 1. http://www.gnu.org/ 2623 2. http://gcc.gnu.org/gcc-4.7/changes.html 2624 3. https://gcc.gnu.org/onlinedocs/4.7.4/ 2625 4. http://gcc.gnu.org/gcc-4.7/changes.html 2626 5. https://gcc.gnu.org/onlinedocs/4.7.3/ 2627 6. http://gcc.gnu.org/gcc-4.7/changes.html 2628 7. https://gcc.gnu.org/onlinedocs/4.7.2/ 2629 8. http://gcc.gnu.org/gcc-4.7/changes.html 2630 9. https://gcc.gnu.org/onlinedocs/4.7.1/ 2631 10. http://gcc.gnu.org/gcc-4.7/changes.html 2632 11. https://gcc.gnu.org/onlinedocs/4.7.0/ 2633 12. http://gcc.gnu.org/gcc-4.7/buildstat.html 2634 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 2635 14. http://gcc.gnu.org/index.html 2636 15. mailto:gcc (a] gcc.gnu.org 2637 16. http://gcc.gnu.org/mirrors.html 2638 17. http://gcc.gnu.org/svn.html 2639 18. https://gcc.gnu.org/onlinedocs/ 2640 19. mailto:gcc-help (a] gcc.gnu.org 2641 20. mailto:gcc (a] gcc.gnu.org 2642 21. https://gcc.gnu.org/lists.html 2643 22. http://www.fsf.org/ 2644 23. https://gcc.gnu.org/about.html 2645 24. http://validator.w3.org/check/referer 2646 ====================================================================== 2647 http://gcc.gnu.org/gcc-4.7/changes.html 2648 2649 GCC 4.7 Release Series 2650 Changes, New Features, and Fixes 2651 2652 Caveats 2653 2654 * The -fconserve-space flag has been deprecated. The flag had no 2655 effect for most targets: only targets without a global .bss section 2656 and without support for switchable sections. Furthermore, the flag 2657 only had an effect for G++, where it could result in wrong 2658 semantics (please refer to the GCC manual for further details). The 2659 flag will be removed in GCC 4.8 2660 * Support for a number of older systems and recently unmaintained or 2661 untested target ports of GCC has been declared obsolete in GCC 4.7. 2662 Unless there is activity to revive them, the next release of GCC 2663 will have their sources permanently removed. 2664 All GCC ports for the following processor architectures have been 2665 declared obsolete: 2666 + picoChip (picochip-*) 2667 The following ports for individual systems on particular 2668 architectures have been obsoleted: 2669 + IRIX 6.5 (mips-sgi-irix6.5) 2670 + MIPS OpenBSD (mips*-*-openbsd*) 2671 + Solaris 8 (*-*-solaris2.8). Details can be found in the 2672 [1]announcement. 2673 + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*) 2674 * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A, 2675 ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by 2676 default, which for some sources generates code that accesses memory 2677 on unaligned addresses. This requires the kernel of those systems 2678 to enable such accesses (controlled by CP15 register c1, refer to 2679 ARM documentation). Alternatively, or for compatibility with 2680 kernels where unaligned accesses are not supported, all code has to 2681 be compiled with -mno-unaligned-access. Upstream Linux kernel 2682 releases have automatically and unconditionally supported unaligned 2683 accesses as emitted by GCC due to this option being active since 2684 version 2.6.28. 2685 * Support on ARM for the legacy floating-point accelerator (FPA) and 2686 the mixed-endian floating-point format that it used has been 2687 obsoleted. The ports that still use this format have been obsoleted 2688 as well. Many legacy ARM ports already provide an alternative that 2689 uses the VFP floating-point format. The obsolete ports will be 2690 deleted in the next release. 2691 The obsolete ports with alternatives are: 2692 + arm*-*-rtems (use arm*-*-rtemseabi) 2693 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 2694 + arm*-*-elf (use arm*-*-eabi) 2695 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 2696 Note, however, that these alternatives are not binary compatible 2697 with their legacy counterparts (although some can support running 2698 legacy applications). 2699 The obsolete ports that currently lack a modern alternative are: 2700 + arm*-*-ecos-elf 2701 + arm*-*-freebsd 2702 + arm*-wince-pe* 2703 New ports that support more recent versions of the architecture are 2704 welcome. 2705 * Support for the Maverick co-processor on ARM has been obsoleted. 2706 Code to support it will be deleted in the next release. 2707 * Support has been removed for Unix International threads on Solaris 2708 2, so the --enable-threads=solaris configure option and the 2709 -threads compiler option don't work any longer. 2710 * Support has been removed for the Solaris BSD Compatibility Package, 2711 which lives in /usr/ucbinclude and /usr/ucblib. It has been removed 2712 from Solaris 11, and was only intended as a migration aid from 2713 SunOS 4 to SunOS 5. The -compat-bsd compiler option is not 2714 recognized any longer. 2715 * The AVR port's libgcc has been improved and its multilib structure 2716 has been enhanced. As a result, all objects contributing to an 2717 application must either be compiled with GCC versions up to 4.6.x 2718 or with GCC versions 4.7.1 or later. If the compiler is used with 2719 AVR Libc, you need a version that supports the new layout, i.e. 2720 implements [2]#35407. 2721 * The AVR port's -mshort-calls command-line option has been 2722 deprecated. It will be removed in the GCC 4.8 release. See -mrelax 2723 for a replacement. 2724 * The AVR port only references startup code that clears .bss and the 2725 common section resp. initializes the .data and .rodata section 2726 provided respective sections (or subsections thereof) are not 2727 empty, see [3]PR18145. Applications that put all static storage 2728 objects into non-standard sections and / or define all static 2729 storage objects in assembler modules, must reference __do_clear_bss 2730 resp. __do_copy_data by hand or undefine the symbol(s) by means of 2731 -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data. 2732 * The ARM port's -mwords-little-endian option has been deprecated. It 2733 will be removed in a future release. 2734 * Support has been removed for the NetWare x86 configuration 2735 obsoleted in GCC 4.6. 2736 * It is no longer possible to use the "l" constraint in MIPS16 asm 2737 statements. 2738 * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard 2739 library which affected the ABI in C++11 mode: a data member was 2740 added to std::list changing its size and altering the definitions 2741 of some member functions, and std::pair's move constructor was 2742 non-trivial which altered the calling convention for functions with 2743 std::pair arguments or return types. The ABI incompatibilities have 2744 been fixed for GCC version 4.7.2 but as a result C++11 code 2745 compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11 2746 code compiled with different GCC versions and with C++98/C++03 code 2747 compiled with any version. 2748 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 2749 rules for the layout of vectors that could lead to wrong code being 2750 generated. Vectors larger than 8 bytes in size are now by default 2751 aligned to an 8-byte boundary. This is an ABI change: code that 2752 makes explicit use of vector types may be incompatible with binary 2753 objects built with older versions of GCC. Auto-vectorized code is 2754 not affected by this change. (This change affects GCC versions 2755 4.7.2 and later.) 2756 * More information on porting to GCC 4.7 from previous versions of 2757 GCC can be found in the [4]porting guide for this release. 2758 2759 General Optimizer Improvements 2760 2761 * Support for a new parameter --param case-values-threshold=n was 2762 added to allow users to control the cutoff between doing switch 2763 statements as a series of if statements and using a jump table. 2764 * Link-time optimization (LTO) improvements: 2765 + Improved scalability and reduced memory usage. Link time 2766 optimization of Firefox now requires 3GB of RAM on a 64-bit 2767 system, while over 8GB was needed previously. Linking time has 2768 been improved, too. The serial stage of linking Firefox has 2769 been sped up by about a factor of 10. 2770 + Reduced size of object files and temporary storage used during 2771 linking. 2772 + Streaming performance (both outbound and inbound) has been 2773 improved. 2774 + ld -r is now supported with LTO. 2775 + Several bug fixes, especially in symbol table handling and 2776 merging. 2777 * Interprocedural optimization improvements: 2778 + Heuristics now take into account that after inlining code will 2779 be optimized out because of known values (or properties) of 2780 function parameters. For example: 2781 void foo(int a) 2782 { 2783 if (a > 10) 2784 ... huge code ... 2785 } 2786 void bar (void) 2787 { 2788 foo (0); 2789 } 2790 2791 The call of foo will be inlined into bar even when optimizing 2792 for code size. Constructs based on __builtin_constant_p are 2793 now understood by the inliner and code size estimates are 2794 evaluated a lot more realistically. 2795 + The representation of C++ virtual thunks and aliases (both 2796 implicit and defined via the alias attribute) has been 2797 re-engineered. Aliases no longer pose optimization barriers 2798 and calls to an alias can be inlined and otherwise optimized. 2799 + The inter-procedural constant propagation pass has been 2800 rewritten. It now performs generic function specialization. 2801 For example when compiling the following: 2802 void foo(bool flag) 2803 { 2804 if (flag) 2805 ... do something ... 2806 else 2807 ... do something else ... 2808 } 2809 void bar (void) 2810 { 2811 foo (false); 2812 foo (true); 2813 foo (false); 2814 foo (true); 2815 foo (false); 2816 foo (true); 2817 } 2818 2819 GCC will now produce two copies of foo. One with flag being 2820 true, while other with flag being false. This leads to 2821 performance improvements previously possible only by inlining 2822 all calls. Cloning causes a lot less code size growth. 2823 * A string length optimization pass has been added. It attempts to 2824 track string lengths and optimize various standard C string 2825 functions like strlen, strchr, strcpy, strcat, stpcpy and their 2826 _FORTIFY_SOURCE counterparts into faster alternatives. This pass is 2827 enabled by default at -O2 or above, unless optimizing for size, and 2828 can be disabled by the -fno-optimize-strlen option. The pass can 2829 e.g. optimize 2830 char *bar (const char *a) 2831 { 2832 size_t l = strlen (a) + 2; 2833 char *p = malloc (l); if (p == NULL) return p; 2834 strcpy (p, a); strcat (p, "/"); return p; 2835 } 2836 2837 into: 2838 char *bar (const char *a) 2839 { 2840 size_t tmp = strlen (a); 2841 char *p = malloc (tmp + 2); if (p == NULL) return p; 2842 memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p; 2843 } 2844 2845 or for hosted compilations where stpcpy is available in the runtime 2846 and headers provide its prototype, e.g. 2847 void foo (char *a, const char *b, const char *c, const char *d) 2848 { 2849 strcpy (a, b); strcat (a, c); strcat (a, d); 2850 } 2851 2852 can be optimized into: 2853 void foo (char *a, const char *b, const char *c, const char *d) 2854 { 2855 strcpy (stpcpy (stpcpy (a, b), c), d); 2856 } 2857 2858 New Languages and Language specific improvements 2859 2860 * Version 3.1 of the [5]OpenMP specification is now supported for the 2861 C, C++, and Fortran compilers. 2862 2863 Ada 2864 2865 * The command-line option -feliminate-unused-debug-types has been 2866 re-enabled by default, as it is for the other languages, leading to 2867 a reduction in debug info size of 12.5% and more for relevant 2868 cases, as well as to a small compilation speedup. 2869 2870 C family 2871 2872 * A new built-in, __builtin_assume_aligned, has been added, through 2873 which the compiler can be hinted about pointer alignment and can 2874 use it to improve generated code. 2875 * A new warning option -Wunused-local-typedefs was added for C, C++, 2876 Objective-C and Objective-C++. This warning diagnoses typedefs 2877 locally defined in a function, and otherwise not used. 2878 * A new experimental command-line option -ftrack-macro-expansion was 2879 added for C, C++, Objective-C, Objective-C++ and Fortran. It allows 2880 the compiler to emit diagnostic about the current macro expansion 2881 stack when a compilation error occurs in a macro expansion. 2882 * Experimental support for transactional memory has been added. It 2883 includes support in the compiler, as well as a supporting runtime 2884 library called libitm. To compile code with transactional memory 2885 constructs, use the -fgnu-tm option. 2886 Support is currently available for Alpha, ARM, PowerPC, SH, SPARC, 2887 and 32-bit/64-bit x86 platforms. 2888 For more details on transactional memory see [6]the GCC WiKi. 2889 * Support for atomic operations specifying the C++11/C11 memory model 2890 has been added. These new __atomic routines replace the existing 2891 __sync built-in routines. 2892 Atomic support is also available for memory blocks. Lock-free 2893 instructions will be used if a memory block is the same size and 2894 alignment as a supported integer type. Atomic operations which do 2895 not have lock-free support are left as function calls. A set of 2896 library functions is available on the GCC atomic wiki in the 2897 "External Atomics Library" section. 2898 For more details on the memory models and features, see the 2899 [7]atomic wiki. 2900 * When a binary operation is performed on vector types and one of the 2901 operands is a uniform vector, it is possible to replace the vector 2902 with the generating element. For example: 2903 typedef int v4si __attribute__ ((vector_size (16))); 2904 v4si res, a = {1,2,3,4}; 2905 int x; 2906 2907 res = 2 + a; /* means {2,2,2,2} + a */ 2908 res = a - x; /* means a - {x,x,x,x} */ 2909 2910 C 2911 2912 * There is support for some more features from the C11 revision of 2913 the ISO C standard. GCC now accepts the options -std=c11 and 2914 -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x. 2915 + Unicode strings (previously supported only with options such 2916 as -std=gnu11, now supported with -std=c11), and the 2917 predefined macros __STDC_UTF_16__ and __STDC_UTF_32__. 2918 + Nonreturning functions (_Noreturn and <stdnoreturn.h>). 2919 + Alignment support (_Alignas, _Alignof, max_align_t, 2920 <stdalign.h>). 2921 + A built-in function __builtin_complex is provided to support C 2922 library implementation of the CMPLX family of macros. 2923 2924 C++ 2925 2926 * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat 2927 options, which are equivalent to -std=c++0x, -std=gnu++0x, and 2928 -Wc++0x-compat, respectively. 2929 * G++ now implements [8]C++11 extended friend syntax: 2930 2931 template<class W> 2932 class Q 2933 { 2934 static const int I = 2; 2935 public: 2936 friend W; 2937 }; 2938 2939 struct B 2940 { 2941 int ar[Q<B>::I]; 2942 }; 2943 2944 * Thanks to Ville Voutilainen, G++ now implements [9]C++11 explicit 2945 override control. 2946 2947 struct B { 2948 virtual void f() const final; 2949 virtual void f(int); 2950 }; 2951 2952 struct D : B { 2953 void f() const; // error: D::f attempts to override final B::f 2954 void f(long) override; // error: doesn't override anything 2955 void f(int) override; // ok 2956 }; 2957 2958 struct E final { }; 2959 struct F: E { }; // error: deriving from final class 2960 2961 * G++ now implements [10]C++11 non-static data member initializers. 2962 2963 struct A { 2964 int i = 42; 2965 } a; // initializes a.i to 42 2966 2967 * Thanks to Ed Smith-Rowland, G++ now implements [11]C++11 2968 user-defined literals. 2969 2970 // Not actually a good approximation. :) 2971 constexpr long double operator"" _degrees (long double d) { return d * 0.0175; } 2972 long double pi = 180.0_degrees; 2973 2974 * G++ now implements [12]C++11 alias-declarations. 2975 2976 template <class T> using Ptr = T*; 2977 Ptr<int> ip; // decltype(ip) is int* 2978 2979 * Thanks to Ville Voutilainen and Pedro Lamarao, G++ now implements 2980 [13]C++11 delegating constructors. 2981 2982 struct A { 2983 A(int); 2984 A(): A(42) { } // delegate to the A(int) constructor 2985 }; 2986 2987 * G++ now fully implements C++11 atomic classes rather than just 2988 integer derived classes. 2989 2990 class POD { 2991 int a; 2992 int b; 2993 }; 2994 std::atomic<POD> my_atomic_POD; 2995 2996 * G++ now sets the predefined macro __cplusplus to the correct value, 2997 199711L for C++98/03, and 201103L for C++11. 2998 * G++ now correctly implements the two-phase lookup rules such that 2999 an unqualified name used in a template must have an appropriate 3000 declaration found either in scope at the point of definition of the 3001 template or by argument-dependent lookup at the point of 3002 instantiation. As a result, code that relies on a second 3003 unqualified lookup at the point of instantiation to find functions 3004 declared after the template or in dependent bases will be rejected. 3005 The compiler will suggest ways to fix affected code, and using the 3006 -fpermissive compiler flag will allow the code to compile with a 3007 warning. 3008 3009 template <class T> 3010 void f() { g(T()); } // error, g(int) not found by argument-dependent lookup 3011 void g(int) { } // fix by moving this declaration before the declaration of f 3012 3013 template <class T> 3014 struct A: T { 3015 // error, B::g(B) not found by argument-dependent lookup 3016 void f() { g(T()); } // fix by using this->g or A::g 3017 }; 3018 3019 struct B { void g(B); }; 3020 3021 int main() 3022 { 3023 f<int>(); 3024 A<B>().f(); 3025 } 3026 3027 * G++ now properly re-uses stack space allocated for temporary 3028 objects when their lifetime ends, which can significantly lower 3029 stack consumption for some C++ functions. As a result of this, some 3030 code with undefined behavior will now break: 3031 3032 const int &f(const int &i) { return i; } 3033 .... 3034 const int &x = f(1); 3035 const int &y = f(2); 3036 3037 Here, x refers to the temporary allocated to hold the 1 argument, 3038 which only lives until the end of the initialization; it 3039 immediately becomes a dangling reference. So the next statement 3040 re-uses the stack slot to hold the 2 argument, and users of x get 3041 that value instead. 3042 Note that this should not cause any change of behavior for 3043 temporaries of types with non-trivial destructors, as they are 3044 already destroyed at end of full-expression; the change is that now 3045 the storage is released as well. 3046 * A new command-line option -Wdelete-non-virtual-dtor has been added 3047 to warn when delete is used to destroy an instance of a class which 3048 has virtual functions and non-virtual destructor. It is unsafe to 3049 delete an instance of a derived class through a pointer to a base 3050 class if the base class does not have a virtual destructor. This 3051 warning is enabled by -Wall. 3052 * A new command-line option -Wzero-as-null-pointer-constant has been 3053 added to warn when a literal '0' is used as null pointer constant. 3054 It can be useful to facilitate the conversion to nullptr in C++11. 3055 * As per C++98, access-declarations are now deprecated by G++. 3056 Using-declarations are to be used instead. Furthermore, some 3057 efforts have been made to improve the support of class scope 3058 using-declarations. In particular, using-declarations referring to 3059 a dependent type now work as expected ([14]bug c++/14258). 3060 * The ELF symbol visibility of a template instantiation is now 3061 properly constrained by the visibility of its template arguments 3062 ([15]bug c++/35688). 3063 3064 Runtime Library (libstdc++) 3065 3066 * [16]Improved experimental support for the new ISO C++ standard, 3067 C++11, including: 3068 + using noexcept in most of the library; 3069 + implementations of pointer_traits, allocator_traits and 3070 scoped_allocator_adaptor; 3071 + uses-allocator construction for tuple; 3072 + vector meets the allocator-aware container requirements; 3073 + replacing monotonic_clock with steady_clock; 3074 + enabling the thread support library on most POSIX targets; 3075 + many small improvements to conform to the FDIS. 3076 * Added --enable-clocale=newlib configure option. 3077 * Debug Mode iterators for unordered associative containers. 3078 * Avoid polluting the global namespace and do not include <unistd.h>. 3079 3080 Fortran 3081 3082 * The compile flag [17]-fstack-arrays has been added, which causes 3083 all local arrays to be put on stack memory. For some programs this 3084 will improve the performance significantly. If your program uses 3085 very large local arrays, it is possible that you will have to 3086 extend your runtime limits for stack memory. 3087 * The [18]-Ofast flag now also implies [19]-fno-protect-parens and 3088 [20]-fstack-arrays. 3089 * Front-end optimizations can now be selected by the 3090 [21]-ffrontend-optimize option and deselected by the 3091 -fno-frontend-optimize option. 3092 * When front-end optimization removes a function call, 3093 [22]-Wfunction-elimination warns about that. 3094 * When performing front-end-optimization, the 3095 [23]-faggressive-function-elimination option allows the removal of 3096 duplicate function calls even for impure functions. 3097 * The flag [24]-Wreal-q-constant has been added, which warns if 3098 floating-point literals have been specified using q (such as 3099 1.0q0); the q marker is now supported as a vendor extension to 3100 denote quad precision (REAL(16) or, if not available, REAL(10)). 3101 Consider using a kind parameter (such as in 1.0_qp) instead, which 3102 can be obtained via [25]SELECTED_REAL_KIND. 3103 * The GFORTRAN_USE_STDERR environment variable has been removed. GNU 3104 Fortran now always prints error messages to standard error. If you 3105 wish to redirect standard error, please consult the manual for your 3106 OS, shell, batch environment etc. as appropriate. 3107 * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment 3108 variable have been removed. When encountering a serious error, 3109 gfortran will now always abort the program. Whether a core dump is 3110 generated depends on the user environment settings; see the ulimit 3111 -c setting for POSIX shells, limit coredumpsize for C shells, and 3112 the [26]WER user-mode dumps settings on Windows. 3113 * The [27]-fbacktrace option is now enabled by default. When 3114 encountering a fatal error, gfortran will attempt to print a 3115 backtrace to standard error before aborting. It can be disabled 3116 with -fno-backtrace. Note: On POSIX targets with the addr2line 3117 utility from GNU binutils, GNU Fortran can print a backtrace with 3118 function name, file name, line number information in addition to 3119 the addresses; otherwise only the addresses are printed. 3120 * [28]Fortran 2003: 3121 + Generic interface names which have the same name as derived 3122 types are now supported, which allows to write constructor 3123 functions. Note that Fortran does not support static 3124 constructor functions; only default initialization or an 3125 explicit structure-constructor initialization are available. 3126 + [29]Polymorphic (class) arrays are now supported. 3127 * [30]Fortran 2008: 3128 + Support for the DO CONCURRENT construct has been added, which 3129 allows the user to specify that individual loop iterations 3130 have no interdependencies. 3131 + [31]Coarrays: Full single-image support except for polymorphic 3132 coarrays. Additionally, preliminary support for multiple 3133 images via an MPI-based [32]coarray communication library has 3134 been added. Note: The library version is not yet usable as 3135 remote coarray access is not yet possible. 3136 * [33]TS 29113: 3137 + New flag [34]-std=f2008ts permits programs that are expected 3138 to conform to the Fortran 2008 standard and the draft 3139 Technical Specification (TS) 29113 on Further Interoperability 3140 of Fortran with C. 3141 + The OPTIONAL attribute is now allowed for dummy arguments of 3142 BIND(C) procedures. 3143 + The RANK intrinsic has been added. 3144 + The implementation of the ASYNCHRONOUS attribute in GCC is 3145 compatible with the candidate draft of TS 29113 (since GCC 3146 4.6). 3147 3148 Go 3149 3150 * GCC 4.7 implements the [35]Go 1 language standard. The library 3151 support in 4.7.0 is not quite complete, due to release timing. 3152 Release 4.7.1 includes complete support for Go 1. The Go library is 3153 from the Go 1.0.1 release. 3154 * Go has been tested on GNU/Linux and Solaris platforms. It may work 3155 on other platforms as well. 3156 3157 New Targets and Target Specific Improvements 3158 3159 ARM 3160 3161 * GCC now supports the Cortex-A7 processor implementing the v7-a 3162 version of the architecture using the option -mcpu=cortex-a7. 3163 * The default vector size in auto-vectorization for NEON is now 128 3164 bits. If vectorization fails thusly, the vectorizer tries again 3165 with 64-bit vectors. 3166 * A new option -mvectorize-with-neon-double was added to allow users 3167 to change the vector size to 64 bits. 3168 3169 AVR 3170 3171 * GCC now supports the XMEGA architecture. This requires GNU binutils 3172 2.22 or later. 3173 * Support for the [36]named address spaces __flash, __flash1, ..., 3174 __flash5 and __memx has been added. These address spaces locate 3175 read-only data in flash memory and allow reading from flash memory 3176 by means of ordinary C code, i.e. without the need of (inline) 3177 assembler code: 3178 3179 const __flash int values[] = { 42, 31 }; 3180 3181 int add_values (const __flash int *p, int i) 3182 { 3183 return values[i] + *p; 3184 } 3185 3186 * Support has been added for the AVR-specific configure option 3187 --with-avrlibc=yes in order to arrange for better integration of 3188 [37]AVR-Libc. This configure option is supported in avr-gcc 4.7.2 3189 and newer and will only take effect in non-RTEMS configurations. If 3190 avr-gcc is configured for RTEMS, the option will be ignored which 3191 is the same as specifying --with-avrlibc=no. See [38]PR54461 for 3192 more technical details. 3193 * Support for AVR-specific [39]built-in functions has been added. 3194 * Support has been added for the signed and unsigned 24-bit scalar 3195 integer types __int24 and __uint24. 3196 * New command-line options -maccumulate-args, -mbranch-cost=cost and 3197 -mstrict-X were added to allow better fine-tuning of code 3198 optimization. 3199 * The command option -fdata-sections now also takes affect on the 3200 section names of variables with the progmem attribute. 3201 * A new inline assembler print modifier %i to print a RAM address as 3202 I/O address has been added: 3203 3204 #include <avr/io.h> /* Port Definitions from AVR-LibC */ 3205 3206 void set_portb (uint8_t value) 3207 { 3208 asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory"); 3209 } 3210 3211 The offset between an I/O address and the RAM address for that I/O 3212 location is device-specific. This offset is taken into account when 3213 printing a RAM address with the %i modifier so that the address is 3214 suitable to be used as operand in an I/O command. The address must 3215 be a constant integer known at compile time. 3216 * The inline assembler constraint "R" to represent integers in the 3217 range -6 ... 5 has been removed without replacement. 3218 * Many optimizations to: 3219 + 64-bit integer arithmetic 3220 + Widening multiplication 3221 + Integer division by a constant 3222 + Avoid constant reloading in multi-byte instructions. 3223 + Micro-optimizations for special instruction sequences. 3224 + Generic built-in functions like __builtin_ffs*, 3225 __builtin_clz*, etc. 3226 + If-else decision trees generated by switch instructions 3227 + Merging of data located in flash memory 3228 + New libgcc variants for devices with 8-bit wide stack pointer 3229 + ... 3230 * Better documentation: 3231 + Handling of EIND and indirect jumps on devices with more than 3232 128 KiB of program memory. 3233 + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function 3234 registers. 3235 + Function attributes OS_main and OS_task. 3236 + AVR-specific built-in macros. 3237 3238 C6X 3239 3240 * Support has been added for the Texas Instruments C6X family of 3241 processors. 3242 3243 CR16 3244 3245 * Support has been added for National Semiconductor's CR16 3246 architecture. 3247 3248 Epiphany 3249 3250 * Support has been added for Adapteva's Epiphany architecture. 3251 3252 IA-32/x86-64 3253 3254 * Support for Intel AVX2 intrinsics, built-in functions and code 3255 generation is available via -mavx2. 3256 * Support for Intel BMI2 intrinsics, built-in functions and code 3257 generation is available via -mbmi2. 3258 * Implementation and automatic generation of __builtin_clz* using the 3259 lzcnt instruction is available via -mlzcnt. 3260 * Support for Intel FMA3 intrinsics and code generation is available 3261 via -mfma. 3262 * A new -mfsgsbase command-line option is available that makes GCC 3263 generate new segment register read/write instructions through 3264 dedicated built-ins. 3265 * Support for the new Intel rdrnd instruction is available via 3266 -mrdrnd. 3267 * Two additional AVX vector conversion instructions are available via 3268 -mf16c. 3269 * Support for new Intel processor codename IvyBridge with RDRND, 3270 FSGSBASE and F16C is available through -march=core-avx-i. 3271 * Support for the new Intel processor codename Haswell with AVX2, 3272 FMA, BMI, BMI2, LZCNT is available through -march=core-avx2. 3273 * Support for new AMD family 15h processors (Piledriver core) is now 3274 available through -march=bdver2 and -mtune=bdver2 options. 3275 * Support for [40]the x32 psABI is now available through the -mx32 3276 option. 3277 * Windows mingw targets are using the -mms-bitfields option by 3278 default. 3279 * Windows x86 targets are using the __thiscall calling convention for 3280 C++ class-member functions. 3281 * Support for the configure option --with-threads=posix for Windows 3282 mingw targets. 3283 3284 MIPS 3285 3286 * GCC now supports thread-local storage (TLS) for MIPS16. This 3287 requires GNU binutils 2.22 or later. 3288 * GCC can now generate code specifically for the Cavium Octeon+ and 3289 Octeon2 processors. The associated command-line options are 3290 -march=octeon+ and -march=octeon2 respectively. Both options 3291 require GNU binutils 2.22 or later. 3292 * GCC can now work around certain 24k errata, under the control of 3293 the command-line option -mfix-24k. These workarounds require GNU 3294 binutils 2.20 or later. 3295 * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build 3296 n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux 3297 toolchain that generates 32-bit code by default. Use the 3298 configure-time option --enable-targets=all to select these extra 3299 multilibs. 3300 * Passing -fno-delayed-branch now also stops the assembler from 3301 automatically filling delay slots. 3302 3303 PowerPC/PowerPC64 3304 3305 * Vectors of type vector long long or vector long are passed and 3306 returned using the same method as other vectors with the VSX 3307 instruction set. Previously GCC did not adhere to the ABI for 3308 128-bit vectors with 64-bit integer base types (PR 48857). This 3309 will also be fixed in the GCC 4.6.1 and 4.5.4 releases. 3310 * A new option -mno-pointers-to-nested-functions was added to allow 3311 AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify 3312 that the compiler should not load up the chain register (r11) 3313 before calling a function through a pointer. If you use this 3314 option, you cannot call nested functions through a pointer, or call 3315 other languages that might use the static chain. 3316 * A new option msave-toc-indirect was added to allow AIX 3317 32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we 3318 save the TOC in the prologue for indirect calls or generate the 3319 save inline. This can speed up some programs that call through a 3320 function pointer a lot, but it can slow down other functions that 3321 only call through a function pointer in exceptional cases. 3322 * The PowerPC port will now enable machine-specific built-in 3323 functions when the user switches the target machine using the 3324 #pragma GCC target or __attribute__ ((__target__ ("target"))) code 3325 sequences. In addition, the target macros are updated. However, due 3326 to the way the -save-temps switch is implemented, you won't see the 3327 effect of these additional macros being defined in preprocessor 3328 output. 3329 3330 SH 3331 3332 * A new option -msoft-atomic has been added. When it is specified, 3333 GCC will generate GNU/Linux-compatible gUSA atomic sequences for 3334 the new __atomic routines. 3335 * Since it is neither supported by GAS nor officially documented, 3336 code generation for little endian SH2A has been disabled. 3337 Specifying -ml with -m2a* will now result in a compiler error. 3338 * The defunct -mbranch-cost option has been fixed. 3339 * Some improvements to the generated code of: 3340 + Utilization of the tst #imm,R0 instruction. 3341 + Dynamic shift instructions on SH2A. 3342 + Integer absolute value calculations. 3343 * The -mdiv= option for targets other than SHmedia has been fixed and 3344 documented. 3345 3346 SPARC 3347 3348 * The option -mflat has been reinstated. When it is specified, the 3349 compiler will generate code for a single register window model. 3350 This is essentially a new implementation and the corresponding 3351 debugger support has been added to GDB 7.4. 3352 * Support for the options -mtune=native and -mcpu=native has been 3353 added on selected native platforms (GNU/Linux and Solaris). 3354 * Support for the SPARC T3 (Niagara 3) processor has been added. 3355 * VIS: 3356 + An intrinsics header visintrin.h has been added. 3357 + Builtin intrinsics for the VIS 1.0 edge handling and pixel 3358 compare instructions have been added. 3359 + The little-endian version of alignaddr is now supported. 3360 + When possible, VIS builtins are marked const, which should 3361 increase the compiler's ability to optimize VIS operations. 3362 + The compiler now properly tracks the %gsr register and how it 3363 behaves as an input for various VIS instructions. 3364 + Akin to fzero, the compiler can now generate fone instructions 3365 in order to set all of the bits of a floating-point register 3366 to 1. 3367 + The documentation for the VIS intrinsics in the GCC manual has 3368 been brought up to date and many inaccuracies were fixed. 3369 + Intrinsics for the VIS 2.0 bmask, bshuffle, and 3370 non-condition-code setting edge instructions have been added. 3371 Their availability is controlled by the new -mvis2 and 3372 -mno-vis2 options. They are enabled by default on 3373 UltraSPARC-III and later CPUs. 3374 * Support for UltraSPARC Fused Multiply-Add floating-point extensions 3375 has been added. These instructions are enabled by default on SPARC 3376 T3 (Niagara 3) and later CPUs. 3377 3378 TILE-Gx/TILEPro 3379 3380 * Support has been added for the Tilera TILE-Gx and TILEPro families 3381 of processors. 3382 3383 Other significant improvements 3384 3385 * A new option (-grecord-gcc-switches) was added that appends 3386 compiler command-line options that might affect code generation to 3387 the DW_AT_producer attribute string in the DWARF debugging 3388 information. 3389 * GCC now supports various new GNU extensions to the DWARF debugging 3390 information format, like [41]entry value and [42]call site 3391 information, [43]typed DWARF stack or [44]a more compact macro 3392 representation. Support for these extensions has been added to GDB 3393 7.4. They can be disabled through the -gstrict-dwarf command-line 3394 option. 3395 3396 GCC 4.7.1 3397 3398 This is the [45]list of problem reports (PRs) from GCC's bug tracking 3399 system that are known to be fixed in the 4.7.1 release. This list might 3400 not be complete (that is, it is possible that some PRs that have been 3401 fixed are not listed here). 3402 3403 The Go frontend in the 4.7.1 release fully supports the [46]Go 1 3404 language standard. 3405 3406 GCC 4.7.2 3407 3408 This is the [47]list of problem reports (PRs) from GCC's bug tracking 3409 system that are known to be fixed in the 4.7.2 release. This list might 3410 not be complete (that is, it is possible that some PRs that have been 3411 fixed are not listed here). 3412 3413 GCC 4.7.3 3414 3415 This is the [48]list of problem reports (PRs) from GCC's bug tracking 3416 system that are known to be fixed in the 4.7.3 release. This list might 3417 not be complete (that is, it is possible that some PRs that have been 3418 fixed are not listed here). 3419 3420 GCC 4.7.4 3421 3422 This is the [49]list of problem reports (PRs) from GCC's bug tracking 3423 system that are known to be fixed in the 4.7.4 release. This list might 3424 not be complete (that is, it is possible that some PRs that have been 3425 fixed are not listed here). 3426 3427 3428 For questions related to the use of GCC, please consult these web 3429 pages and the [50]GCC manuals. If that fails, the 3430 [51]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3431 web pages and the development of GCC are welcome on our developer 3432 list at [52]gcc (a] gcc.gnu.org. All of [53]our lists have public 3433 archives. 3434 3435 Copyright (C) [54]Free Software Foundation, Inc. Verbatim copying and 3436 distribution of this entire article is permitted in any medium, 3437 provided this notice is preserved. 3438 3439 These pages are [55]maintained by the GCC team. Last modified 3440 2015-11-14[56]. 3441 3442 References 3443 3444 1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html 3445 2. http://savannah.nongnu.org/bugs/?35407 3446 3. https://gcc.gnu.org/PR18145 3447 4. https://gcc.gnu.org/gcc-4.7/porting_to.html 3448 5. http://openmp.org/wp/openmp-specifications/ 3449 6. https://gcc.gnu.org/wiki/TransactionalMemory 3450 7. https://gcc.gnu.org/wiki/Atomic/GCCMM 3451 8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 3452 9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 3453 10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 3454 11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 3455 12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 3456 13. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 3457 14. https://gcc.gnu.org/PR14258 3458 15. https://gcc.gnu.org/PR35688 3459 16. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011 3460 17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 3461 18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689 3462 19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270 3463 20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 3464 21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275 3465 22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170 3466 23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270 3467 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 3468 25. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html 3469 26. http://msdn.microsoft.com/en-us/library/bb787181%28v=vs.85%29.aspx 3470 27. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183 3471 28. https://gcc.gnu.org/wiki/Fortran2003Status 3472 29. https://gcc.gnu.org/wiki/OOP 3473 30. https://gcc.gnu.org/wiki/Fortran2008Status 3474 31. https://gcc.gnu.org/wiki/Coarray 3475 32. https://gcc.gnu.org/wiki/CoarrayLib 3476 33. https://gcc.gnu.org/wiki/TS29113Status 3477 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 3478 35. https://golang.org/doc/go1 3479 36. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html 3480 37. http://nongnu.org/avr-libc/ 3481 38. https://gcc.gnu.org/PR54461 3482 39. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built%5f002din-Functions.html 3483 40. https://sites.google.com/site/x32abi/ 3484 41. http://www.dwarfstd.org/ShowIssue.php?issue=100909.1 3485 42. http://www.dwarfstd.org/ShowIssue.php?issue=100909.2 3486 43. http://www.dwarfstd.org/doc/040408.1.html 3487 44. http://www.dwarfstd.org/ShowIssue.php?issue=110722.1 3488 45. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1 3489 46. https://golang.org/doc/go1 3490 47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2 3491 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3 3492 49. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4 3493 50. https://gcc.gnu.org/onlinedocs/ 3494 51. mailto:gcc-help (a] gcc.gnu.org 3495 52. mailto:gcc (a] gcc.gnu.org 3496 53. https://gcc.gnu.org/lists.html 3497 54. http://www.fsf.org/ 3498 55. https://gcc.gnu.org/about.html 3499 56. http://validator.w3.org/check/referer 3500 ====================================================================== 3501 http://gcc.gnu.org/gcc-4.6/index.html 3502 3503 GCC 4.6 Release Series 3504 3505 April 12, 2013 3506 3507 The [1]GNU project and the GCC developers are pleased to announce the 3508 release of GCC 4.6.4. 3509 3510 This release is a bug-fix release, containing fixes for regressions in 3511 GCC 4.6.3 relative to previous releases of GCC. 3512 3513 Release History 3514 3515 GCC 4.6.4 3516 April 12, 2013 ([2]changes, [3]documentation) 3517 3518 GCC 4.6.3 3519 March 1, 2012 ([4]changes, [5]documentation) 3520 3521 GCC 4.6.2 3522 October 26, 2011 ([6]changes, [7]documentation) 3523 3524 GCC 4.6.1 3525 June 27, 2011 ([8]changes, [9]documentation) 3526 3527 GCC 4.6.0 3528 March 25, 2011 ([10]changes, [11]documentation) 3529 3530 References and Acknowledgements 3531 3532 GCC used to stand for the GNU C Compiler, but since the compiler 3533 supports several other languages aside from C, it now stands for the 3534 GNU Compiler Collection. 3535 3536 A list of [12]successful builds is updated as new information becomes 3537 available. 3538 3539 The GCC developers would like to thank the numerous people that have 3540 contributed new features, improvements, bug fixes, and other changes as 3541 well as test results to GCC. This [13]amazing group of volunteers is 3542 what makes GCC successful. 3543 3544 For additional information about GCC please refer to the [14]GCC 3545 project web site or contact the [15]GCC development mailing list. 3546 3547 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 3548 3549 3550 For questions related to the use of GCC, please consult these web 3551 pages and the [18]GCC manuals. If that fails, the 3552 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3553 web pages and the development of GCC are welcome on our developer 3554 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 3555 archives. 3556 3557 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 3558 distribution of this entire article is permitted in any medium, 3559 provided this notice is preserved. 3560 3561 These pages are [23]maintained by the GCC team. Last modified 3562 2014-06-28[24]. 3563 3564 References 3565 3566 1. http://www.gnu.org/ 3567 2. http://gcc.gnu.org/gcc-4.6/changes.html 3568 3. https://gcc.gnu.org/onlinedocs/4.6.4/ 3569 4. http://gcc.gnu.org/gcc-4.6/changes.html 3570 5. https://gcc.gnu.org/onlinedocs/4.6.3/ 3571 6. http://gcc.gnu.org/gcc-4.6/changes.html 3572 7. https://gcc.gnu.org/onlinedocs/4.6.2/ 3573 8. http://gcc.gnu.org/gcc-4.6/changes.html 3574 9. https://gcc.gnu.org/onlinedocs/4.6.1/ 3575 10. http://gcc.gnu.org/gcc-4.6/changes.html 3576 11. https://gcc.gnu.org/onlinedocs/4.6.0/ 3577 12. http://gcc.gnu.org/gcc-4.6/buildstat.html 3578 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 3579 14. http://gcc.gnu.org/index.html 3580 15. mailto:gcc (a] gcc.gnu.org 3581 16. http://gcc.gnu.org/mirrors.html 3582 17. http://gcc.gnu.org/svn.html 3583 18. https://gcc.gnu.org/onlinedocs/ 3584 19. mailto:gcc-help (a] gcc.gnu.org 3585 20. mailto:gcc (a] gcc.gnu.org 3586 21. https://gcc.gnu.org/lists.html 3587 22. http://www.fsf.org/ 3588 23. https://gcc.gnu.org/about.html 3589 24. http://validator.w3.org/check/referer 3590 ====================================================================== 3591 http://gcc.gnu.org/gcc-4.6/changes.html 3592 3593 GCC 4.6 Release Series 3594 Changes, New Features, and Fixes 3595 3596 Caveats 3597 3598 * The options -b <machine> and -V <version> have been removed because 3599 they were unreliable. Instead, users should directly run 3600 <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to 3601 run a different version of gcc. 3602 * GCC now has stricter checks for invalid command-line options. In 3603 particular, when gcc was called to link object files rather than 3604 compile source code, it would previously accept and ignore all 3605 options starting with --, including linker options such as 3606 --as-needed and --export-dynamic, although such options would 3607 result in errors if any source code was compiled. Such options, if 3608 unknown to the compiler, are now rejected in all cases; if the 3609 intent was to pass them to the linker, options such as 3610 -Wl,--as-needed should be used. 3611 * Versions of the GNU C library up to and including 2.11.1 included 3612 an [1]incorrect implementation of the cproj function. GCC optimizes 3613 its builtin cproj according to the behavior specified and allowed 3614 by the ISO C99 standard. If you want to avoid discrepancies between 3615 the C library and GCC's builtin transformations when using cproj in 3616 your code, use GLIBC 2.12 or later. If you are using an older GLIBC 3617 and actually rely on the incorrect behavior of cproj, then you can 3618 disable GCC's transformations using -fno-builtin-cproj. 3619 * The C-only intermodule optimization framework (IMA, enabled by 3620 -combine) has been removed in favor of the new generic link-time 3621 optimization framework (LTO) introduced in [2]GCC 4.5.0. 3622 * GCC now ships with the LGPL-licensed libquadmath library, which 3623 provides quad-precision mathematical functions for targets with a 3624 __float128 datatype. __float128 is available for targets on 32-bit 3625 x86, x86-64 and Itanium architectures. The libquadmath library is 3626 automatically built on such targets when building the Fortran 3627 compiler. 3628 * New -Wunused-but-set-variable and -Wunused-but-set-parameter 3629 warnings were added for C, C++, Objective-C and Objective-C++. 3630 These warnings diagnose variables respective parameters which are 3631 only set in the code and never otherwise used. Usually such 3632 variables are useless and often even the value assigned to them is 3633 computed needlessly, sometimes expensively. The 3634 -Wunused-but-set-variable warning is enabled by default by -Wall 3635 flag and -Wunused-but-set-parameter by -Wall -Wextra flags. 3636 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 3637 rules for the layout of vectors that could lead to wrong code being 3638 generated. Vectors larger than 8 bytes in size are now by default 3639 aligned to an 8-byte boundary. This is an ABI change: code that 3640 makes explicit use of vector types may be incompatible with binary 3641 objects built with older versions of GCC. Auto-vectorized code is 3642 not affected by this change. (This change affects GCC versions 3643 4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.) 3644 * On AVR, variables with the progmem attribute to locate data in 3645 flash memory must be qualified as const. 3646 * Support for a number of older systems and recently unmaintained or 3647 untested target ports of GCC has been declared obsolete in GCC 4.6. 3648 Unless there is activity to revive them, the next release of GCC 3649 will have their sources permanently removed. 3650 All GCC ports for the following processor architectures have been 3651 declared obsolete: 3652 + Argonaut ARC (arc-*) 3653 + National Semiconductor CRX (crx-*) 3654 + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*, 3655 m68hc12-*-*, m6812-*-*) 3656 + Sunplus S+core (score-*) 3657 The following ports for individual systems on particular 3658 architectures have been obsoleted: 3659 + Interix (i[34567]86-*-interix3*) 3660 + NetWare x86 (i[3456x]86-*-netware*) 3661 + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*) 3662 + MCore PE (mcore-*-pe*) 3663 + SH SymbianOS (sh*-*-symbianelf*) 3664 + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*) 3665 + M68K uClinux old ABI (m68k-*-uclinuxoldabi*) 3666 + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*, 3667 vax-*-netbsd*, but not *-*-netbsdelf*) 3668 The i[34567]86-*-pe alias for Cygwin targets has also been 3669 obsoleted; users should configure for i[34567]86-*-cygwin* instead. 3670 Certain configure options to control the set of libraries built 3671 with GCC on some targets have been obsoleted. On ARM targets, the 3672 options --disable-fpu, --disable-26bit, --disable-underscore, 3673 --disable-interwork, --disable-biendian and --disable-nofmult have 3674 been obsoleted. On MIPS targets, the options 3675 --disable-single-float, --disable-biendian and --disable-softfloat 3676 have been obsoleted. 3677 * Support has been removed for all the [3]configurations obsoleted in 3678 GCC 4.5. 3679 * More information on porting to GCC 4.6 from previous versions of 3680 GCC can be found in the [4]porting guide for this release. 3681 3682 General Optimizer Improvements 3683 3684 * A new general optimization level, -Ofast, has been introduced. It 3685 combines the existing optimization level -O3 with options that can 3686 affect standards compliance but result in better optimized code. 3687 For example, -Ofast enables -ffast-math. 3688 * Link-time optimization improvements: 3689 + The [5]Scalable Whole Program Optimizer (WHOPR) project has 3690 stabilized to the point of being usable. It has become the 3691 default mode when using the LTO optimization model. Link time 3692 optimization can now split itself into multiple parallel 3693 compilations. Parallelism is controlled with -flto=n (where n 3694 specifies the number of compilations to execute in parallel). 3695 GCC can also cooperate with a GNU make job server by 3696 specifying the -flto=jobserver option and adding + to the 3697 beginning of the Makefile rule executing the linker. 3698 Classical LTO mode can be enforced by -flto-partition=none. 3699 This may result in small code quality improvements. 3700 + A large number of bugs were fixed. GCC itself, Mozilla Firefox 3701 and other large applications can be built with LTO enabled. 3702 + The linker plugin support improvements 3703 o Linker plugin is now enabled by default when the linker 3704 is detected to have plugin support. This is the case for 3705 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and 3706 the Gold linker on ELF targets. Plugin support of the 3707 Apple linker on Darwin is not compatible with GCC. The 3708 linker plugin can also be controlled by the 3709 -fuse-linker-plugin command line option. 3710 o Resolution information from the linker plugin is used to 3711 drive whole program assumptions. Use of the linker plugin 3712 results in more aggressive optimization on binaries and 3713 on shared libraries that use the hidden visibility 3714 attribute. Consequently the use of -fwhole-program is not 3715 necessary in addition to LTO. 3716 + Hidden symbols used from non-LTO objects now have to be 3717 explicitly annotated with externally_visible when the linker 3718 plugin is not used. 3719 + C++ inline functions and virtual tables are now privatized 3720 more aggressively, leading to better inter-procedural 3721 optimization and faster dynamic linking. 3722 + Memory usage and intermediate language streaming performance 3723 have been improved. 3724 + Static constructors and destructors from individual units are 3725 inlined into a single function. This can significantly improve 3726 startup times of large C++ applications where static 3727 constructors are very common. For example, static constructors 3728 are used when including the iostream header. 3729 + Support for the Ada language has been added. 3730 * Interprocedural optimization improvements 3731 + The interprocedural framework was re-tuned for link time 3732 optimization. Several scalability issues were resolved. 3733 + Improved auto-detection of const and pure functions. Newly, 3734 noreturn functions are auto-detected. 3735 The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is 3736 available that informs users when adding attributes to headers 3737 might improve code generation. 3738 + A number of inlining heuristic improvements. In particular: 3739 o Partial inlining is now supported and enabled by default 3740 at -O2 and greater. The feature can be controlled via 3741 -fpartial-inlining. 3742 Partial inlining splits functions with short hot path to 3743 return. This allows more aggressive inlining of the hot 3744 path leading to better performance and often to code size 3745 reductions (because cold parts of functions are not 3746 duplicated). 3747 o Scalability for large compilation units was improved 3748 significantly. 3749 o Inlining of callbacks is now more aggressive. 3750 o Virtual methods are considered for inlining when the 3751 caller is inlined and devirtualization is then possible. 3752 o Inlining when optimizing for size (either in cold regions 3753 of a program or when compiling with -Os) was improved to 3754 better handle C++ programs with larger abstraction 3755 penalty, leading to smaller and faster code. 3756 + The IPA reference optimization pass detecting global variables 3757 used or modified by functions was strengthened and sped up. 3758 + Functions whose address was taken are now optimized out when 3759 all references to them are dead. 3760 + A new inter-procedural static profile estimation pass detects 3761 functions that are executed once or unlikely to be executed. 3762 Unlikely executed functions are optimized for size. Functions 3763 executed once are optimized for size except for the inner 3764 loops. 3765 + On most targets with named section support, functions used 3766 only at startup (static constructors and main), functions used 3767 only at exit and functions detected to be cold are placed into 3768 separate text segment subsections. This extends the 3769 -freorder-functions feature and is controlled by the same 3770 switch. The goal is to improve the startup time of large C++ 3771 programs. 3772 Proper function placement requires linker support. GNU ld 3773 2.21.51 on ELF targets was updated to place those functions 3774 together within the text section leading to better code 3775 locality and faster startup times of large C++ programs. The 3776 feature is also supported in the Apple linker. Support in the 3777 gold linker is planned. 3778 * A new switch -fstack-usage has been added. It makes the compiler 3779 output stack usage information for the program, on a per-function 3780 basis, in an auxiliary file. 3781 * A new switch -fcombine-stack-adjustments has been added. It can be 3782 used to enable or disable the compiler's stack-slot combining pass 3783 which before was enabled automatically at -O1 and above, but could 3784 not be controlled on its own. 3785 * A new switch -fstrict-volatile-bitfields has been added. Using it 3786 indicates that accesses to volatile bitfields should use a single 3787 access of the width of the field's type. This option can be useful 3788 for precisely defining and accessing memory-mapped peripheral 3789 registers from C or C++. 3790 3791 Compile time and memory usage improvements 3792 3793 * Datastructures used by the dataflow framework in GCC were 3794 reorganized for better memory usage and more cache locality. 3795 Compile time is improved especially on units with large functions 3796 (possibly resulting from a lot of inlining) not fitting into the 3797 processor cache. The compile time of the GCC C compiler binary with 3798 link-time optimization went down by over 10% (benchmarked on x86-64 3799 target). 3800 3801 New Languages and Language specific improvements 3802 3803 Ada 3804 3805 * Stack checking has been improved on selected architectures (Alpha, 3806 IA-32/x86-64, RS/6000 and SPARC): it now will detect stack 3807 overflows in all cases on these architectures. 3808 * Initial support for Ada 2012 has been added. 3809 3810 C family 3811 3812 * A new warning, enabled by -Wdouble-promotion, has been added that 3813 warns about cases where a value of type float is implicitly 3814 promoted to double. This is especially helpful for CPUs that handle 3815 the former in hardware, but emulate the latter in software. 3816 * A new function attribute leaf was introduced. This attribute allows 3817 better inter-procedural optimization across calls to functions that 3818 return to the current unit only via returning or exception 3819 handling. This is the case for most library functions that have no 3820 callbacks. 3821 * Support for a new data type __int128 for targets having wide enough 3822 machine-mode support. 3823 * The new function attribute callee_pop_aggregate allows to specify 3824 if the caller or callee is responsible for popping the aggregate 3825 return pointer value from the stack. 3826 * Support for selectively enabling and disabling warnings via #pragma 3827 GCC diagnostic has been added. For instance: 3828 #pragma GCC diagnostic error "-Wuninitialized" 3829 foo(a); /* error is given for this one */ 3830 #pragma GCC diagnostic push 3831 #pragma GCC diagnostic ignored "-Wuninitialized" 3832 foo(b); /* no diagnostic for this one */ 3833 #pragma GCC diagnostic pop 3834 foo(c); /* error is given for this one */ 3835 #pragma GCC diagnostic pop 3836 foo(d); /* depends on command line options */ 3837 3838 * The -fmax-errors=N option is now supported. Using this option 3839 causes the compiler to exit after N errors have been issued. 3840 3841 C 3842 3843 * There is now experimental support for some features from the 3844 upcoming C1X revision of the ISO C standard. This support may be 3845 selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions. 3846 Note that this support is experimental and may change incompatibly 3847 in future releases for consistency with changes to the C1X standard 3848 draft. The following features are newly supported as described in 3849 the N1539 draft of C1X (with changes agreed at the March 2011 WG14 3850 meeting); some other features were already supported with no 3851 compiler changes being needed, or have some support but not in full 3852 accord with N1539 (as amended). 3853 + Static assertions (_Static_assert keyword) 3854 + Typedef redefinition 3855 + New macros in <float.h> 3856 + Anonymous structures and unions 3857 * The new -fplan9-extensions option directs the compiler to support 3858 some extensions for anonymous struct fields which are implemented 3859 by the Plan 9 compiler. A pointer to a struct may be automatically 3860 converted to a pointer to an anonymous field when calling a 3861 function, in order to make the types match. An anonymous struct 3862 field whose type is a typedef name may be referred to using the 3863 typedef name. 3864 3865 C++ 3866 3867 * Improved [7]experimental support for the upcoming C++0x ISO C++ 3868 standard, including support for constexpr (thanks to Gabriel Dos 3869 Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide), 3870 noexcept, unrestricted unions, range-based for loops (thanks to 3871 Rodrigo Rivas Costa), opaque enum declarations (thanks also to 3872 Rodrigo), implicitly deleted functions and implicit move 3873 constructors. 3874 * When an extern declaration within a function does not match a 3875 declaration in the enclosing context, G++ now properly declares the 3876 name within the namespace of the function rather than the namespace 3877 which was open just before the function definition ([8]c++/43145). 3878 * GCC now warns by default when casting integers to larger pointer 3879 types. These warnings can be disabled with the option 3880 -Wno-int-to-pointer-cast, which is now also available in C++. 3881 * G++ no longer optimizes using the assumption that a value of 3882 enumeration type will fall within the range specified by the 3883 standard, since that assumption is easily violated with a 3884 conversion from integer type ([9]c++/43680). The old behavior can 3885 be restored with -fstrict-enums. 3886 * The new -fnothrow-opt flag changes the semantics of a throw() 3887 exception specification to match the proposed semantics of the 3888 noexcept specification: just call terminate if an exception tries 3889 to propagate out of a function with such an exception 3890 specification. This dramatically reduces or eliminates the code 3891 size overhead from adding the exception specification. 3892 * The new -Wnoexcept flag will suggest adding a noexcept qualifier to 3893 a function that the compiler can tell doesn't throw if it would 3894 change the value of a noexcept expression. 3895 * The -Wshadow option now warns if a local variable or type 3896 declaration shadows another type in C++. Note that the compiler 3897 will not warn if a local variable shadows a struct/class/enum, but 3898 will warn if it shadows an explicit typedef. 3899 * When an identifier is not found in the current scope, G++ now 3900 offers suggestions about which identifier might have been intended. 3901 * G++ now issues clearer diagnostics for missing semicolons after 3902 class, struct, and union definitions. 3903 * G++ now issues clearer diagnostics for missing semicolons after 3904 class member declarations. 3905 * G++ now issues clearer diagnostics when a colon is used in a place 3906 where a double-colon was intended. 3907 * G++ no longer accepts mutable on reference members ([10]c++/33558). 3908 Use -fpermissive to allow the old, non-conforming behaviour. 3909 * A few mangling fixes have been made, to attribute const/volatile on 3910 function pointer types, decltype of a plain decl, and use of a 3911 function parameter in the declaration of another parameter. By 3912 default the compiler still uses the old mangling, but emits aliases 3913 with the new mangling on targets that support strong aliases. Users 3914 can switch over entirely to the new mangling with -fabi-version=5 3915 or -fabi-version=0. -Wabi will now warn about code that uses the 3916 old mangling. 3917 * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified 3918 type to be default initialized unless the type has a user-declared 3919 default constructor. In 4.6.2 G++ implements the proposed 3920 resolution of [11]DR 253, so default initialization is allowed if 3921 it initializes all subobjects. Code that fails to compile can be 3922 fixed by providing an initializer e.g. 3923 struct A { A(); }; 3924 struct B : A { int i; }; 3925 const B b = B(); 3926 Use -fpermissive to allow the old, non-conforming behaviour. 3927 3928 Runtime Library (libstdc++) 3929 3930 * [12]Improved experimental support for the upcoming ISO C++ 3931 standard, C++0x, including using constexpr and nullptr. 3932 * Performance improvements to the [13]Debug Mode, thanks to Franc,ois 3933 Dumont. 3934 * Atomic operations used for reference-counting are annotated so that 3935 they can be understood by race detectors such as Helgrind, see 3936 [14]Data Race Hunting. 3937 * Most libstdc++ standard headers have been changed to no longer 3938 include the cstddef header as an implementation detail. Code that 3939 relied on that header being included as side-effect of including 3940 other standard headers will need to include cstddef explicitly. 3941 3942 Fortran 3943 3944 * On systems supporting the libquadmath library, GNU Fortran now also 3945 supports a quad-precision, kind=16 floating-point data type 3946 (REAL(16), COMPLEX(16)). As the data type is not fully supported in 3947 hardware, calculations might be one to two orders of magnitude 3948 slower than with the 4, 8 or 10 bytes floating-point data types. 3949 This change does not affect systems which support REAL(16) in 3950 hardware nor those which do not support libquadmath. 3951 * Much improved compile time for large array constructors. 3952 * In order to reduce execution time and memory consumption, use of 3953 temporary arrays in assignment expressions is avoided for many 3954 cases. The compiler now reverses loops in order to avoid generating 3955 a temporary array where possible. 3956 * Improved diagnostics, especially with -fwhole-file. 3957 * The -fwhole-file flag is now enabled by default. This improves code 3958 generation and diagnostics. It can be disabled using the deprecated 3959 -fno-whole-file flag. 3960 * Support the generation of Makefile dependencies via the [15]-M... 3961 flags of GCC; you may need to specify the -cpp option in addition. 3962 The dependencies take modules, Fortran's include, and CPP's 3963 #include into account. Note: Using -M for the module path is no 3964 longer supported, use -J instead. 3965 * The flag -Wconversion has been modified to only issue warnings 3966 where a conversion leads to information loss. This drastically 3967 reduces the number of warnings; -Wconversion is thus now enabled 3968 with -Wall. The flag -Wconversion-extra has been added and also 3969 warns about other conversions; -Wconversion-extra typically issues 3970 a huge number of warnings, most of which can be ignored. 3971 * A new command-line option -Wunused-dummy-argument warns about 3972 unused dummy arguments and is included in -Wall. Before, 3973 -Wunused-variable also warned about unused dummy arguments. 3974 * Fortran 2003 support has been extended: 3975 + Improved support for polymorphism between libraries and 3976 programs and for complicated inheritance patterns (cf. 3977 [16]object-oriented programming). 3978 + Experimental support of the ASSOCIATE construct. 3979 + In pointer assignments it is now possible to specify the lower 3980 bounds of the pointer and, for a rank-1 or a simply contiguous 3981 data-target, to remap the bounds. 3982 + Automatic (re)allocation: In intrinsic assignments to 3983 allocatable variables the left-hand side will be automatically 3984 allocated (if unallocated) or reallocated (if the shape or 3985 type parameter is different). To avoid the small performance 3986 penalty, you can use a(:) = ... instead of a = ... for arrays 3987 and character strings - or disable the feature using -std=f95 3988 or -fno-realloc-lhs. 3989 + Deferred type parameter: For scalar allocatable and pointer 3990 variables the character length can be deferred. 3991 + Namelist variables with allocatable and pointer attribute and 3992 nonconstant length type parameter are supported. 3993 * Fortran 2008 support has been extended: 3994 + Experimental [17]coarray support (for one image only, i.e. 3995 num_images() == 1); use the [18]-fcoarray=single flag to 3996 enable it. 3997 + The STOP and the new ERROR STOP statements now support all 3998 constant expressions. 3999 + Support for the CONTIGUOUS attribute. 4000 + Support for ALLOCATE with MOLD. 4001 + Support for the STORAGE_SIZE intrinsic inquiry function. 4002 + Support of the NORM2 and PARITY intrinsic functions. 4003 + The following bit intrinsics were added: POPCNT and POPPAR for 4004 counting the number of 1 bits and returning the parity; BGE, 4005 BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR 4006 for combined left and right shifts, MASKL and MASKR for simple 4007 left and right justified masks, MERGE_BITS for a bitwise merge 4008 using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations, 4009 and the transformational bit intrinsics IALL, IANY and 4010 IPARITY. 4011 + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine. 4012 + Support for the IMPURE attribute for procedures, which allows 4013 for ELEMENTAL procedures without the restrictions of PURE. 4014 + Null pointers (including NULL()) and not allocated variables 4015 can be used as actual argument to optional non-pointer, 4016 non-allocatable dummy arguments, denoting an absent argument. 4017 + Non-pointer variables with TARGET attribute can be used as 4018 actual argument to POINTER dummies with INTENT(IN) 4019 + Pointers including procedure pointers and those in a derived 4020 type (pointer components) can now be initialized by a target 4021 instead of only by NULL. 4022 + The EXIT statement (with construct-name) can now be used to 4023 leave not only the DO but also the ASSOCIATE, BLOCK, IF, 4024 SELECT CASE and SELECT TYPE constructs. 4025 + Internal procedures can now be used as actual argument. 4026 + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS 4027 and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV 4028 have been added; these arrays contain the supported kind 4029 values for the respective types. 4030 + The module procedures C_SIZEOF of the intrinsic module 4031 ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of 4032 ISO_FORTRAN_ENV have been implemented. 4033 + Minor changes: obsolescence diagnostics for ENTRY was added 4034 for -std=f2008; a line may start with a semicolon; for 4035 internal and module procedures END can be used instead of END 4036 SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes 4037 a RADIX argument; intrinsic types are supported for 4038 TYPE(intrinsic-type-spec); multiple type-bound procedures can 4039 be declared in a single PROCEDURE statement; implied-shape 4040 arrays are supported for named constants (PARAMETER). The 4041 transformational, three argument versions of BESSEL_JN and 4042 BESSEL_YN were added - the elemental, two-argument version had 4043 been added in GCC 4.4; note that the transformational 4044 functions use a recurrence algorithm. 4045 4046 Go 4047 4048 Support for the [19]Go programming language has been added to GCC. It 4049 is not enabled by default when you build GCC; use the 4050 --enable-languages configure option to build it. The driver program for 4051 compiling Go code is gccgo. 4052 4053 Go is currently known to work on GNU/Linux and RTEMS. Solaris support 4054 is in progress. It may or may not work on other platforms. 4055 4056 Objective-C and Objective-C++ 4057 4058 * The -fobjc-exceptions flag is now required to enable Objective-C 4059 exception and synchronization syntax (introduced by the keywords 4060 @try, @catch, @finally and @synchronized). 4061 * A number of Objective-C 2.0 features and extensions are now 4062 supported by GCC. These features are enabled by default; you can 4063 disable them by using the new -fobjc-std=objc1 command-line option. 4064 * The Objective-C 2.0 dot-syntax is now supported. It is an 4065 alternative syntax for using getters and setters; object.count is 4066 automatically converted into [object count] or [object setCount: 4067 ...] depending on context; for example if (object.count > 0) is 4068 automatically compiled into the equivalent of if ([object count] > 4069 0) while object.count = 0; is automatically compiled into the 4070 equivalent ot [object setCount: 0];. The dot-syntax can be used 4071 with instance and class objects and with any setters or getters, no 4072 matter if they are part of a declared property or not. 4073 * Objective-C 2.0 declared properties are now supported. They are 4074 declared using the new @property keyword, and are most commonly 4075 used in conjunction with the new Objective-C 2.0 dot-syntax. The 4076 nonatomic, readonly, readwrite, assign, retain, copy, setter and 4077 getter attributes are all supported. Marking declared properties 4078 with __attribute__ ((deprecated)) is supported too. 4079 * The Objective-C 2.0 @synthesize and @dynamic keywords are 4080 supported. @synthesize causes the compiler to automatically 4081 synthesize a declared property, while @dynamic is used to disable 4082 all warnings for a declared property for which no implementation is 4083 provided at compile time. Synthesizing declared properties requires 4084 runtime support in most useful cases; to be able to use it with the 4085 GNU runtime, appropriate helper functions have been added to the 4086 GNU Objective-C runtime ABI, and are implemented by the GNU 4087 Objective-C runtime library shipped with GCC. 4088 * The Objective-C 2.0 fast enumeration syntax is supported in 4089 Objective-C. This is currently not yet available in Objective-C++. 4090 Fast enumeration requires support in the runtime, and such support 4091 has been added to the GNU Objective-C runtime library (shipped with 4092 GCC). 4093 * The Objective-C 2.0 @optional keyword is supported. It allows you 4094 to mark methods or properties in a protocol as optional as opposed 4095 to required. 4096 * The Objective-C 2.0 @package keyword is supported. It has currently 4097 the same effect as the @public keyword. 4098 * Objective-C 2.0 method attributes are supported. Currently the 4099 supported attributes are deprecated, sentinel, noreturn and format. 4100 * Objective-C 2.0 method argument attributes are supported. The most 4101 widely used attribute is unused, to mark an argument as unused in 4102 the implementation. 4103 * Objective-C 2.0 class and protocol attributes are supported. 4104 Currently the only supported attribute is deprecated. 4105 * Objective-C 2.0 class extensions are supported. A class extension 4106 has the same syntax as a category declaration with no category 4107 name, and the methods and properties declared in it are added 4108 directly to the main class. It is mostly used as an alternative to 4109 a category to add methods to a class without advertising them in 4110 the public headers, with the advantage that for class extensions 4111 the compiler checks that all the privately declared methods are 4112 actually implemented. 4113 * As a result of these enhancements, GCC can now be used to build 4114 Objective-C and Objective-C++ software that uses Foundation and 4115 other important system frameworks with the NeXT runtime on Darwin 9 4116 and Darwin 10 (OSX 10.5 and 10.6). 4117 * Many bugs in the compiler have been fixed in this release; in 4118 particular, LTO can now be used when compiling Objective-C and 4119 Objective-C++ and the parser is much more robust in dealing with 4120 invalid code. 4121 4122 Runtime Library (libobjc) 4123 4124 * The GNU Objective-C runtime library now defines the macro 4125 __GNU_LIBOBJC__ (with a value that is increased at every release 4126 where there is any change to the API) in objc/objc.h, making it 4127 easy to determine if the GNU Objective-C runtime library is being 4128 used, and if so, which version. Previous versions of the GNU 4129 Objective-C runtime library (and other Objective-C runtime 4130 libraries such as the Apple one) do not define this macro. 4131 * A new Objective-C 2.0 API, almost identical to the one implemented 4132 by the Apple Objective-C runtime, has been implemented in the GNU 4133 Objective-C runtime library. The new API hides the internals of 4134 most runtime structures but provides a more extensive set of 4135 functions to operate on them. It is much easier, for example, to 4136 create or modify classes at runtime. The new API also makes it 4137 easier to port software from Apple to GNU as almost no changes 4138 should be required. The old API is still supported for backwards 4139 compatibility; including the old objc/objc-api.h header file 4140 automatically selects the old API, while including the new 4141 objc/runtime.h header file automatically selects the new API. 4142 Support for the old API is being phased out and upgrading the 4143 software to use the new API is strongly recommended. To check for 4144 the availability of the new API, the __GNU_LIBOBJC__ macro can be 4145 used as older versions of the GNU Objective-C runtime library, 4146 which do not support the new API, do not define such a macro. 4147 * Runtime support for @synchronized has been added. 4148 * Runtime support for Objective-C 2.0 synthesized property accessors 4149 has been added. 4150 * Runtime support for Objective-C 2.0 fast enumeration has been 4151 added. 4152 4153 New Targets and Target Specific Improvements 4154 4155 ARM 4156 4157 * GCC now supports the Cortex-M4 processor implementing the v7-em 4158 version of the architecture using the option -mcpu=cortex-m4. 4159 * Scheduling descriptions for the Cortex-M4, the Neon and the 4160 floating point units of the Cortex-A9 and a pipeline description 4161 for the Cortex-A5 have been added. 4162 * Synchronization primitives such as __sync_fetch_and_add and friends 4163 are now inlined for supported architectures rather than calling 4164 into a kernel helper function. 4165 * SSA loop prefetching is enabled by default for the Cortex-A9 at 4166 -O3. 4167 * Several improvements were committed to improve code generation for 4168 the ARM architecture including a rewritten implementation for load 4169 and store multiples. 4170 * Several enhancements were committed to improve SIMD code generation 4171 for NEON by adding support for widening instructions, misaligned 4172 loads and stores, vector conditionals and support for 64 bit 4173 arithmetic. 4174 * Support was added for the Faraday cores fa526, fa606te, fa626te, 4175 fmp626te, fmp626 and fa726te and can be used with the respective 4176 names as parameters to the -mcpu= option. 4177 * Basic support was added for Cortex-A15 and is available through 4178 -mcpu=cortex-a15. 4179 * GCC for AAPCS configurations now more closely adheres to the AAPCS 4180 specification by enabling -fstrict-volatile-bitfields by default. 4181 4182 IA-32/x86-64 4183 4184 * The new -fsplit-stack option permits programs to use a 4185 discontiguous stack. This is useful for threaded programs, in that 4186 it is no longer necessary to specify the maximum stack size when 4187 creating a thread. This feature is currently only implemented for 4188 32-bit and 64-bit x86 GNU/Linux targets. 4189 * Support for emitting profiler counter calls before function 4190 prologues. This is enabled via a new command-line option -mfentry. 4191 * Optimization for the Intel Core 2 processors is now available 4192 through the -march=core2 and -mtune=core2 options. 4193 * Support for Intel Core i3/i5/i7 processors is now available through 4194 the -march=corei7 and -mtune=corei7 options. 4195 * Support for Intel Core i3/i5/i7 processors with AVX is now 4196 available through the -march=corei7-avx and -mtune=corei7-avx 4197 options. 4198 * Support for AMD Bobcat (family 14) processors is now available 4199 through the -march=btver1 and -mtune=btver1 options. 4200 * Support for AMD Bulldozer (family 15) processors is now available 4201 through the -march=bdver1 and -mtune=bdver1 options. 4202 * The default setting (when not optimizing for size) for 32-bit 4203 GNU/Linux and Darwin x86 targets has been changed to 4204 -fomit-frame-pointer. The default can be reverted to 4205 -fno-omit-frame-pointer by configuring GCC with the 4206 --enable-frame-pointer configure option. 4207 * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support 4208 __float128 on 32-bit and 64-bit x86 targets. 4209 * AVX floating-point arithmetic can now be enabled by default at 4210 configure time with the new --with-fpmath=avx option. 4211 * The SSA loop prefetching pass is enabled when using -O3 when 4212 optimizing for CPUs where prefetching is beneficial (AMD CPUs newer 4213 than K6). 4214 * Support for TBM (Trailing Bit Manipulation) built-in functions and 4215 code generation is available via -mtbm. 4216 * Support for AMD's BMI (Bit Manipulation) built-in functions and 4217 code generation is available via -mbmi. 4218 4219 MicroBlaze 4220 4221 * Support has been added for the Xilinx MicroBlaze softcore processor 4222 (microblaze-elf) embedded target. This configurable processor is 4223 supported on several Xilinx Spartan and Virtex FPGAs. 4224 4225 MIPS 4226 4227 * GCC now supports the Loongson 3A processor. Its canonical -march= 4228 and -mtune= name is loongson3a. 4229 4230 MN10300 / AM33 4231 4232 * The inline assembly register constraint "A" has been renamed "c". 4233 This constraint is used to select a floating-point register that 4234 can be used as the destination of a multiply-accumulate 4235 instruction. 4236 * New inline assembly register constraints "A" and "D" have been 4237 added. These constraint letters resolve to all general registers 4238 when compiling for AM33, and resolve to address registers only or 4239 data registers only when compiling for MN10300. 4240 * The MDR register is represented in the compiler. One can access the 4241 register via the "z" constraint in inline assembly. It can be 4242 marked as clobbered or used as a local register variable via the 4243 "mdr" name. The compiler uses the RETF instruction if the function 4244 does not modify the MDR register, so it is important that inline 4245 assembly properly annotate any usage of the register. 4246 4247 PowerPC/PowerPC64 4248 4249 * GCC now supports the Applied Micro Titan processor with 4250 -mcpu=titan. 4251 * The -mrecip option has been added, which indicates whether the 4252 reciprocal and reciprocal square root instructions should be used. 4253 * The -mveclibabi=mass option can be used to enable the compiler to 4254 autovectorize mathematical functions using the Mathematical 4255 Acceleration Subsystem library. 4256 * The -msingle-pic-base option has been added, which instructs the 4257 compiler to avoid loading the PIC base register in function 4258 prologues. The PIC base register must be initialized by the runtime 4259 system. 4260 * The -mblock-move-inline-limit option has been added, which enables 4261 the user to control the maximum size of inlined memcpy calls and 4262 similar. 4263 * PowerPC64 GNU/Linux support for applications requiring a large TOC 4264 section has been improved. A new command-line option, 4265 -mcmodel=MODEL, controls this feature; valid values for MODEL are 4266 small, medium, or large. 4267 * The Altivec builtin functions vec_ld and vec_st have been modified 4268 to generate the Altivec memory instructions LVX and STVX, even if 4269 the -mvsx option is used. In the initial GCC 4.5 release, these 4270 builtin functions were changed to generate VSX memory reference 4271 instructions instead of Altivec memory instructions, but there are 4272 differences between the two instructions. If the VSX instruction 4273 set is available, you can now use the new builtin functions 4274 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 4275 instructions. 4276 * The GCC compiler on AIX now defaults to a process layout with a 4277 larger data space allowing larger programs to be compiled. 4278 * The GCC long double type on AIX 6.1 and above has reverted to 64 4279 bit double precision, matching the AIX XL compiler default, because 4280 of missing C99 symbols required by the GCC runtime. 4281 * The default processor scheduling model and tuning for PowerPC64 4282 GNU/Linux and for AIX 6.1 and above now is POWER7. 4283 * Starting with GCC 4.6.1, vectors of type vector long long or vector 4284 long are passed and returned in the same method as other vectors 4285 with the VSX instruction set. Previously the GCC compiler did not 4286 adhere to the ABI for 128-bit vectors with 64-bit integer base 4287 types (PR 48857). This is also fixed in the GCC 4.5.4 release. 4288 4289 S/390, zSeries and System z9/z10, IBM zEnterprise z196 4290 4291 * Support for the zEnterprise z196 processor has been added. When 4292 using the -march=z196 option, the compiler will generate code 4293 making use of the following instruction facilities: 4294 + Conditional load/store 4295 + Distinct-operands 4296 + Floating-point-extension 4297 + Interlocked-access 4298 + Population-count 4299 The -mtune=z196 option avoids the compare and branch instructions 4300 as well as the load address instruction with an index register as 4301 much as possible and performs instruction scheduling appropriate 4302 for the new out-of-order pipeline architecture. 4303 * When using the -m31 -mzarch options the generated code still 4304 conforms to the 32-bit ABI but uses the general purpose registers 4305 as 64-bit registers internally. This requires a Linux kernel saving 4306 the whole 64-bit registers when doing a context switch. Kernels 4307 providing that feature indicate that by the 'highgprs' string in 4308 /proc/cpuinfo. 4309 * The SSA loop prefetching pass is enabled when using -O3. 4310 4311 SPARC 4312 4313 * GCC now supports the LEON series of SPARC V8 processors. The code 4314 generated by the compiler can either be tuned to it by means of the 4315 --with-tune=leon configure option and -mtune=leon compilation 4316 option, or the compiler can be built for the sparc-leon-{elf,linux} 4317 and sparc-leon3-{elf,linux} targets directly. 4318 * GCC has stopped sign/zero-extending parameter registers in the 4319 callee for functions taking parameters with sub-word size in 32-bit 4320 mode, since this is redundant with the specification of the ABI. 4321 GCC has never done so in 64-bit mode since this is also redundant. 4322 * The command line option -mfix-at697f has been added to enable the 4323 documented workaround for the single erratum of the Atmel AT697F 4324 processor. 4325 4326 Operating Systems 4327 4328 Android 4329 4330 * GCC now supports the Bionic C library and provides a convenient way 4331 of building native libraries and applications for the Android 4332 platform. Refer to the documentation of the -mandroid and -mbionic 4333 options for details on building native code. At the moment, Android 4334 support is enabled only for ARM. 4335 4336 Darwin/Mac OS X 4337 4338 * General 4339 + Initial support for CFString types has been added. 4340 This allows GCC to build projects including the system Core 4341 Foundation frameworks. The GCC Objective-C family supports 4342 CFString "toll-free bridged" as per the Mac OS X system tools. 4343 CFString is also recognized in the context of format 4344 attributes and arguments (see the documentation for format 4345 attributes for limitations). At present, 8-bit character types 4346 are supported. 4347 + Object file size reduction. 4348 The Darwin zeroed memory allocators have been re-written to 4349 make more use of .zerofill sections. For non-debug code, this 4350 can reduce object file size significantly. 4351 + Objective-C family 64-bit support (NeXT ABI 2). 4352 Initial support has been added to support 64-bit Objective-C 4353 code using the Darwin/OS X native (NeXT) runtime. ABI version 4354 2 will be selected automatically when 64-bit code is built. 4355 + Objective-C family 32-bit ABI 1. 4356 For 32-bit code ABI 1 is also now also allowed. At present it 4357 must be selected manually using -fobjc-abi-version=1 where 4358 applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6). 4359 * x86 Architecture 4360 + The -mdynamic-no-pic option has been enabled. 4361 Code supporting -mdynamic-no-pic optimization has been added 4362 and is applicable to -m32 builds. The compiler bootstrap uses 4363 the option where appropriate. 4364 + The default value for -mtune= has been changed. 4365 Since Darwin systems are primarily Xeon, Core-2 or similar the 4366 default tuning has been changed to -mtune=core2. 4367 + Enable 128-bit long double (__float128) support on Darwin. 4368 * PPC Architecture 4369 + Darwin64 ABI. 4370 Several significant bugs have been fixed, such that GCC now 4371 produces code compatible with the Darwin64 PowerPC ABI. 4372 + libffi and boehm-gc. 4373 The Darwin ports of the libffi and boehm-gc libraries have 4374 been upgraded to include a Darwin64 implementation. This means 4375 that powerpc*-*-darwin9 platforms may now, for example, build 4376 Java applications with -m64 enabled. 4377 + Plug-in support has been enabled. 4378 + The -fsection-anchors option is now available although, 4379 presently, not heavily tested. 4380 4381 Solaris 2 4382 4383 New Features 4384 4385 * Support symbol versioning with the Sun linker. 4386 * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+. 4387 * Support thread-local storage (TLS) with the Sun assembler on 4388 Solaris 2/x86. 4389 * Support TLS on Solaris 8/9 if prerequisites are met. 4390 * Support COMDAT group with the GNU assembler and recent Sun linker. 4391 * Support the Sun assembler visibility syntax. 4392 * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp. 4393 -march=pentiumpro (Solaris 8/9). 4394 * Don't use SSE on Solaris 8/9 x86 by default. 4395 * Enable 128-bit long double (__float128) support on Solaris 2/x86. 4396 4397 ABI Change 4398 4399 * Change the ABI for returning 8-byte vectors like __m64 in MMX 4400 registers on Solaris 10+/x86 to match the Sun Studio 12.1+ 4401 compilers. This is an incompatible change. If you use such types, 4402 you must either recompile all your code with the new compiler or 4403 use the new -mvect8-ret-in-mem option to remain compatible with 4404 previous versions of GCC and Sun Studio. 4405 4406 Windows x86/x86_64 4407 4408 * Initial support for decimal floating point. 4409 * Support for the __thiscall calling-convention. 4410 * Support for hot-patchable function prologues via the 4411 ms_hook_prologue attribute for x86_64 in addition to 32-bit x86. 4412 * Improvements of stack-probing and stack-allocation mechanisms. 4413 * Support of push/pop-macro pragma as preprocessor command. 4414 With #pragma push_macro("macro-name") the current definition of 4415 macro-name is saved and can be restored with #pragma 4416 pop_macro("macro-name") to its saved definition. 4417 * Enable 128-bit long double (__float128) support on MinGW and 4418 Cygwin. 4419 4420 Other significant improvements 4421 4422 Installation changes 4423 4424 * An install-strip make target is provided that installs stripped 4425 executables, and may install libraries with unneeded or debugging 4426 sections stripped. 4427 * On Power7 systems, there is a potential problem if you build the 4428 GCC compiler with a host compiler using options that enable the VSX 4429 instruction set generation. If the host compiler has been patched 4430 so that the vec_ld and vec_st builtin functions generate Altivec 4431 memory instructions instead of VSX memory instructions, then you 4432 should be able to build the compiler with VSX instruction 4433 generation. 4434 4435 Changes for GCC Developers 4436 4437 Note: these changes concern developers that develop GCC itself or 4438 software that integrates with GCC, such as plugins, and not the general 4439 GCC users. 4440 * The gengtype utility, which previously was internal to the GCC 4441 build process, has been enchanced to provide GC root information 4442 for plugins as necessary. 4443 * The old GC allocation interface of ggc_alloc and friends was 4444 replaced with a type-safe alternative. 4445 4446 GCC 4.6.1 4447 4448 This is the [20]list of problem reports (PRs) from GCC's bug tracking 4449 system that are known to be fixed in the 4.6.1 release. This list might 4450 not be complete (that is, it is possible that some PRs that have been 4451 fixed are not listed here). 4452 4453 GCC 4.6.2 4454 4455 This is the [21]list of problem reports (PRs) from GCC's bug tracking 4456 system that are known to be fixed in the 4.6.2 release. This list might 4457 not be complete (that is, it is possible that some PRs that have been 4458 fixed are not listed here). 4459 4460 GCC 4.6.3 4461 4462 This is the [22]list of problem reports (PRs) from GCC's bug tracking 4463 system that are known to be fixed in the 4.6.3 release. This list might 4464 not be complete (that is, it is possible that some PRs that have been 4465 fixed are not listed here). 4466 4467 GCC 4.6.4 4468 4469 This is the [23]list of problem reports (PRs) from GCC's bug tracking 4470 system that are known to be fixed in the 4.6.4 release. This list might 4471 not be complete (that is, it is possible that some PRs that have been 4472 fixed are not listed here). 4473 4474 4475 For questions related to the use of GCC, please consult these web 4476 pages and the [24]GCC manuals. If that fails, the 4477 [25]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4478 web pages and the development of GCC are welcome on our developer 4479 list at [26]gcc (a] gcc.gnu.org. All of [27]our lists have public 4480 archives. 4481 4482 Copyright (C) [28]Free Software Foundation, Inc. Verbatim copying and 4483 distribution of this entire article is permitted in any medium, 4484 provided this notice is preserved. 4485 4486 These pages are [29]maintained by the GCC team. Last modified 4487 2015-11-14[30]. 4488 4489 References 4490 4491 1. http://sourceware.org/bugzilla/show_bug.cgi?id=10401 4492 2. http://gcc.gnu.org/gcc-4.5/changes.html 4493 3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted 4494 4. http://gcc.gnu.org/gcc-4.6/porting_to.html 4495 5. http://gcc.gnu.org/projects/lto/whopr.pdf 4496 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 4497 7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html 4498 8. https://gcc.gnu.org/PR43145 4499 9. https://gcc.gnu.org/PR43680 4500 10. https://gcc.gnu.org/PR33558 4501 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253 4502 12. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x 4503 13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html 4504 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races 4505 15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html 4506 16. https://gcc.gnu.org/wiki/OOP 4507 17. https://gcc.gnu.org/wiki/Coarray 4508 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233 4509 19. https://golang.org/ 4510 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1 4511 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2 4512 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3 4513 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4 4514 24. https://gcc.gnu.org/onlinedocs/ 4515 25. mailto:gcc-help (a] gcc.gnu.org 4516 26. mailto:gcc (a] gcc.gnu.org 4517 27. https://gcc.gnu.org/lists.html 4518 28. http://www.fsf.org/ 4519 29. https://gcc.gnu.org/about.html 4520 30. http://validator.w3.org/check/referer 4521 ====================================================================== 4522 http://gcc.gnu.org/gcc-4.5/index.html 4523 4524 GCC 4.5 Release Series 4525 4526 Jul 2, 2012 4527 4528 The [1]GNU project and the GCC developers are pleased to announce the 4529 release of GCC 4.5.4. 4530 4531 This release is a bug-fix release, containing fixes for regressions in 4532 GCC 4.5.3 relative to previous releases of GCC. 4533 4534 Release History 4535 4536 GCC 4.5.4 4537 Jul 2, 2012 ([2]changes) 4538 4539 GCC 4.5.3 4540 Apr 28, 2011 ([3]changes) 4541 4542 GCC 4.5.2 4543 Dec 16, 2010 ([4]changes) 4544 4545 GCC 4.5.1 4546 Jul 31, 2010 ([5]changes) 4547 4548 GCC 4.5.0 4549 April 14, 2010 ([6]changes) 4550 4551 References and Acknowledgements 4552 4553 GCC used to stand for the GNU C Compiler, but since the compiler 4554 supports several other languages aside from C, it now stands for the 4555 GNU Compiler Collection. 4556 4557 A list of [7]successful builds is updated as new information becomes 4558 available. 4559 4560 The GCC developers would like to thank the numerous people that have 4561 contributed new features, improvements, bug fixes, and other changes as 4562 well as test results to GCC. This [8]amazing group of volunteers is 4563 what makes GCC successful. 4564 4565 For additional information about GCC please refer to the [9]GCC project 4566 web site or contact the [10]GCC development mailing list. 4567 4568 To obtain GCC please use [11]our mirror sites or [12]our SVN server. 4569 4570 4571 For questions related to the use of GCC, please consult these web 4572 pages and the [13]GCC manuals. If that fails, the 4573 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4574 web pages and the development of GCC are welcome on our developer 4575 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 4576 archives. 4577 4578 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 4579 distribution of this entire article is permitted in any medium, 4580 provided this notice is preserved. 4581 4582 These pages are [18]maintained by the GCC team. Last modified 4583 2014-06-28[19]. 4584 4585 References 4586 4587 1. http://www.gnu.org/ 4588 2. http://gcc.gnu.org/gcc-4.5/changes.html 4589 3. http://gcc.gnu.org/gcc-4.5/changes.html 4590 4. http://gcc.gnu.org/gcc-4.5/changes.html 4591 5. http://gcc.gnu.org/gcc-4.5/changes.html 4592 6. http://gcc.gnu.org/gcc-4.5/changes.html 4593 7. http://gcc.gnu.org/gcc-4.5/buildstat.html 4594 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 4595 9. http://gcc.gnu.org/index.html 4596 10. mailto:gcc (a] gcc.gnu.org 4597 11. http://gcc.gnu.org/mirrors.html 4598 12. http://gcc.gnu.org/svn.html 4599 13. https://gcc.gnu.org/onlinedocs/ 4600 14. mailto:gcc-help (a] gcc.gnu.org 4601 15. mailto:gcc (a] gcc.gnu.org 4602 16. https://gcc.gnu.org/lists.html 4603 17. http://www.fsf.org/ 4604 18. https://gcc.gnu.org/about.html 4605 19. http://validator.w3.org/check/referer 4606 ====================================================================== 4607 http://gcc.gnu.org/gcc-4.5/changes.html 4608 4609 GCC 4.5 Release Series 4610 Changes, New Features, and Fixes 4611 4612 Caveats 4613 4614 * GCC now requires the [1]MPC library in order to build. See the 4615 [2]prerequisites page for version requirements. 4616 * Support for a number of older systems and recently unmaintained or 4617 untested target ports of GCC has been declared obsolete in GCC 4.5. 4618 Unless there is activity to revive them, the next release of GCC 4619 will have their sources permanently removed. 4620 The following ports for individual systems on particular 4621 architectures have been obsoleted: 4622 + IRIX releases before 6.5 (mips-sgi-irix5*, 4623 mips-sgi-irix6.[0-4]) 4624 + Solaris 7 (*-*-solaris2.7) 4625 + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*, 4626 alpha-dec-osf5.0*) 4627 + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions 4628 can be found in the [3]announcement. 4629 Support for the classic POWER architecture implemented in the 4630 original RIOS and RIOS2 processors of the old IBM RS/6000 product 4631 line has been obsoleted in the rs6000 port. This does not affect 4632 the new generation Power and PowerPC architectures. 4633 * Support has been removed for all the [4]configurations obsoleted in 4634 GCC 4.4. 4635 * Support has been removed for the protoize and unprotoize utilities, 4636 obsoleted in GCC 4.4. 4637 * Support has been removed for tuning for Itanium1 (Merced) variants. 4638 Note that code tuned for Itanium2 should also run correctly on 4639 Itanium1. 4640 * GCC now generates unwind info also for epilogues. DWARF debuginfo 4641 generated by GCC now uses more features of DWARF3 than before, and 4642 also some DWARF4 features. GDB older than 7.0 is not able to handle 4643 either of these, so to debug GCC 4.5 generated binaries or 4644 libraries GDB 7.0 or later is needed. You can disable use of DWARF4 4645 features with the -gdwarf-3 -gstrict-dwarf options, or use 4646 -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but 4647 epilogue unwind info is emitted unconditionally whenever unwind 4648 info is emitted. 4649 * On x86 targets, code containing floating-point calculations may run 4650 significantly slower when compiled with GCC 4.5 in strict C99 4651 conformance mode than they did with earlier GCC versions. This is 4652 due to stricter standard conformance of the compiler and can be 4653 avoided by using the option -fexcess-precision=fast; also see 4654 [5]below. 4655 * The function attribute noinline no longer prevents GCC from cloning 4656 the function. A new attribute noclone has been introduced for this 4657 purpose. Cloning a function means that it is duplicated and the new 4658 copy is specialized for certain contexts (for example when a 4659 parameter is a known constant). 4660 4661 General Optimizer Improvements 4662 4663 * The -save-temps now takes an optional argument. The -save-temps and 4664 -save-temps=cwd switches write the temporary files in the current 4665 working directory based on the original source file. The 4666 -save-temps=obj switch will write files into the directory 4667 specified with the -o option, and the intermediate filenames are 4668 based on the output file. This will allow the user to get the 4669 compiler intermediate files when doing parallel builds without two 4670 builds of the same filename located in different directories from 4671 interfering with each other. 4672 * Debugging dumps are now created in the same directory as the object 4673 file rather than in the current working directory. This allows the 4674 user to get debugging dumps when doing parallel builds without two 4675 builds of the same filename interfering with each other. 4676 * GCC has been integrated with the [6]MPC library. This allows GCC to 4677 evaluate complex arithmetic at compile time [7]more accurately. It 4678 also allows GCC to evaluate calls to complex built-in math 4679 functions having constant arguments and replace them at compile 4680 time with their mathematically equivalent results. In doing so, GCC 4681 can generate correct results regardless of the math library 4682 implementation or floating point precision of the host platform. 4683 This also allows GCC to generate identical results regardless of 4684 whether one compiles in native or cross-compile configurations to a 4685 particular target. The following built-in functions take advantage 4686 of this new capability: cacos, cacosh, casin, casinh, catan, 4687 catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan, 4688 and ctanh. The float and long double variants of these functions 4689 (e.g. csinf and csinl) are also handled. 4690 * A new link-time optimizer has been added ([8]-flto). When this 4691 option is used, GCC generates a bytecode representation of each 4692 input file and writes it to specially-named sections in each object 4693 file. When the object files are linked together, all the function 4694 bodies are read from these named sections and instantiated as if 4695 they had been part of the same translation unit. This enables 4696 interprocedural optimizations to work across different files (and 4697 even different languages), potentially improving the performance of 4698 the generated code. To use the link-timer optimizer, -flto needs to 4699 be specified at compile time and during the final link. If the 4700 program does not require any symbols to be exported, it is possible 4701 to combine -flto and the experimental [9]-fwhopr with 4702 [10]-fwhole-program to allow the interprocedural optimizers to use 4703 more aggressive assumptions. 4704 * The automatic parallelization pass was enhanced to support 4705 parallelization of outer loops. 4706 * Automatic parallelization can be enabled as part of Graphite. In 4707 addition to -ftree-parallelize-loops=, specify 4708 -floop-parallelize-all to enable the Graphite-based optimization. 4709 * The infrastructure for optimizing based on [11]restrict qualified 4710 pointers has been rewritten and should result in code generation 4711 improvements. Optimizations based on restrict qualified pointers 4712 are now also available when using -fno-strict-aliasing. 4713 * There is a new optimization pass that attempts to change prototype 4714 of functions to avoid unused parameters, pass only relevant parts 4715 of structures and turn arguments passed by reference to arguments 4716 passed by value when possible. It is enabled by -O2 and above as 4717 well as -Os and can be manually invoked using the new command-line 4718 switch -fipa-sra. 4719 * GCC now optimize exception handling code. In particular cleanup 4720 regions that are proved to not have any effect are optimized out. 4721 4722 New Languages and Language specific improvements 4723 4724 All languages 4725 4726 * The -fshow-column option is now on by default. This means error 4727 messages now have a column associated with them. 4728 4729 Ada 4730 4731 * Compilation of programs heavily using discriminated record types 4732 with variant parts has been sped up and generates more compact 4733 code. 4734 * Stack checking now works reasonably well on most plaforms. In some 4735 specific cases, stack overflows may still fail to be detected, but 4736 a compile-time warning will be issued for these cases. 4737 4738 C family 4739 4740 * If a header named in a #include directive is not found, the 4741 compiler exits immediately. This avoids a cascade of errors arising 4742 from declarations expected to be found in that header being 4743 missing. 4744 * A new built-in function __builtin_unreachable() has been added that 4745 tells the compiler that control will never reach that point. It may 4746 be used after asm statements that terminate by transferring control 4747 elsewhere, and in other places that are known to be unreachable. 4748 * The -Wlogical-op option now warns for logical expressions such as 4749 (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be 4750 mistakes. This option is disabled by default. 4751 * An asm goto feature has been added to allow asm statements that 4752 jump to C labels. 4753 * C++0x raw strings are supported for C++ and for C with -std=gnu99. 4754 * The deprecated attribute now takes an optional string argument, for 4755 example, __attribute__((deprecated("text string"))), that will be 4756 printed together with the deprecation warning. 4757 4758 C 4759 4760 * The -Wenum-compare option, which warns when comparing values of 4761 different enum types, now works for C. It formerly only worked for 4762 C++. This warning is enabled by -Wall. It may be avoided by using a 4763 type cast. 4764 * The -Wcast-qual option now warns about casts which are unsafe in 4765 that they permit const-correctness to be violated without further 4766 warnings. Specifically, it warns about cases where a qualifier is 4767 added when all the lower types are not const. For example, it warns 4768 about a cast from char ** to const char **. 4769 * The -Wc++-compat option is significantly improved. It issues new 4770 warnings for: 4771 + Using C++ reserved operator names as identifiers. 4772 + Conversions to enum types without explicit casts. 4773 + Using va_arg with an enum type. 4774 + Using different enum types in the two branches of ?:. 4775 + Using ++ or -- on a variable of enum type. 4776 + Using the same name as both a struct, union or enum tag and a 4777 typedef, unless the typedef refers to the tagged type itself. 4778 + Using a struct, union, or enum which is defined within another 4779 struct or union. 4780 + A struct field defined using a typedef if there is a field in 4781 the struct, or an enclosing struct, whose name is the typedef 4782 name. 4783 + Duplicate definitions at file scope. 4784 + Uninitialized const variables. 4785 + A global variable with an anonymous struct, union, or enum 4786 type. 4787 + Using a string constant to initialize a char array whose size 4788 is the length of the string. 4789 * The new -Wjump-misses-init option warns about cases where a goto or 4790 switch skips the initialization of a variable. This sort of branch 4791 is an error in C++ but not in C. This warning is enabled by 4792 -Wc++-compat. 4793 * GCC now ensures that a C99-conforming <stdint.h> is present on most 4794 targets, and uses information about the types in this header to 4795 implement the Fortran bindings to those types. GCC does not ensure 4796 the presence of such a header, and does not implement the Fortran 4797 bindings, on the following targets: NetBSD, VxWorks, VMS, 4798 SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF. 4799 * GCC now implements C90- and C99-conforming rules for constant 4800 expressions. This may cause warnings or errors for some code using 4801 expressions that can be folded to a constant but are not constant 4802 expressions as defined by ISO C. 4803 * All known target-independent C90 and C90 Amendment 1 conformance 4804 bugs, and all known target-independent C99 conformance bugs not 4805 related to floating point or extended identifiers, have been fixed. 4806 * The C decimal floating point support now includes support for the 4807 FLOAT_CONST_DECIMAL64 pragma. 4808 * The named address space feature from ISO/IEC TR 18037 is now 4809 supported. This is currently only implemented for the SPU 4810 processor. 4811 4812 C++ 4813 4814 * Improved [12]experimental support for the upcoming C++0x ISO C++ 4815 standard, including support for raw strings, lambda expressions and 4816 explicit type conversion operators. 4817 * When printing the name of a class template specialization, G++ will 4818 now omit any template arguments which come from default template 4819 arguments. This behavior (and the pretty-printing of function 4820 template specializations as template signature and arguments) can 4821 be disabled with the -fno-pretty-templates option. 4822 * Access control is now applied to typedef names used in a template, 4823 which may cause G++ to reject some ill-formed code that was 4824 accepted by earlier releases. The -fno-access-control option can be 4825 used as a temporary workaround until the code is corrected. 4826 * Compilation time for code that uses templates should now scale 4827 linearly with the number of instantiations rather than 4828 quadratically, as template instantiations are now looked up using 4829 hash tables. 4830 * Declarations of functions that look like builtin declarations of 4831 library functions are only considered to be redeclarations if they 4832 are declared with extern "C". This may cause problems with code 4833 that omits extern "C" on hand-written declarations of C library 4834 functions such as abort or memcpy. Such code is ill-formed, but was 4835 accepted by earlier releases. 4836 * Diagnostics that used to complain about passing non-POD types to 4837 ... or jumping past the declaration of a non-POD variable now check 4838 for triviality rather than PODness, as per C++0x. 4839 * In C++0x mode local and anonymous classes are now allowed as 4840 template arguments, and in declarations of variables and functions 4841 with linkage, so long as any such declaration that is used is also 4842 defined ([13]DR 757). 4843 * Labels may now have attributes, as has been permitted for a while 4844 in C. This is only permitted when the label definition and the 4845 attribute specifier is followed by a semicolon--i.e., the label 4846 applies to an empty statement. The only useful attribute for a 4847 label is unused. 4848 * G++ now implements [14]DR 176. Previously G++ did not support using 4849 the injected-class-name of a template base class as a type name, 4850 and lookup of the name found the declaration of the template in the 4851 enclosing scope. Now lookup of the name finds the 4852 injected-class-name, which can be used either as a type or as a 4853 template, depending on whether or not the name is followed by a 4854 template argument list. As a result of this change, some code that 4855 was previously accepted may be ill-formed because 4856 1. The injected-class-name is not accessible because it's from a 4857 private base, or 4858 2. The injected-class-name cannot be used as an argument for a 4859 template template parameter. 4860 In either of these cases, the code can be fixed by adding a 4861 nested-name-specifier to explicitly name the template. The first 4862 can be worked around with -fno-access-control; the second is only 4863 rejected with -pedantic. 4864 * A new standard mangling for SIMD vector types has been added, to 4865 avoid name clashes on systems with vectors of varying length. By 4866 default the compiler still uses the old mangling, but emits aliases 4867 with the new mangling on targets that support strong aliases. Users 4868 can switch over entirely to the new mangling with -fabi-version=4 4869 or -fabi-version=0. -Wabi will now warn about code that uses the 4870 old mangling. 4871 * The command-line option -ftemplate-depth-N is now written as 4872 -ftemplate-depth=N and the old form is deprecated. 4873 * Conversions between NULL and non-pointer types are now warned by 4874 default. The new option -Wno-conversion-null disables these 4875 warnings. Previously these warnings were only available when using 4876 -Wconversion explicitly. 4877 4878 Runtime Library (libstdc++) 4879 4880 * Improved experimental support for the upcoming ISO C++ standard, 4881 C++0x, including: 4882 + Support for <future>, <functional>, and <random>. 4883 + Existing facilities now exploit explicit operators and the 4884 newly implemented core C++0x features. 4885 + The header <cstdatomic> has been renamed to <atomic>. 4886 * An experimental [15]profile mode has been added. This is an 4887 implementation of many C++ standard library constructs with an 4888 additional analysis layer that gives performance improvement advice 4889 based on recognition of suboptimal usage patterns. For example, 4890 #include <vector> 4891 int main() 4892 { 4893 std::vector<int> v; 4894 for (int k = 0; k < 1024; ++k) 4895 v.insert(v.begin(), k); 4896 } 4897 4898 When instrumented via the profile mode, can return suggestions 4899 about the initial size and choice of the container used as follows: 4900 vector-to-list: improvement = 5: call stack = 0x804842c ... 4901 : advice = change std::vector to std::list 4902 vector-size: improvement = 3: call stack = 0x804842c ... 4903 : advice = change initial container size from 0 to 1024 4904 4905 These constructs can be substituted for the normal libstdc++ 4906 constructs on a piecemeal basis, or all existing components can be 4907 transformed via the -D_GLIBCXX_PROFILE macro. 4908 * [16]Support for decimal floating-point arithmetic (aka ISO C++ TR 4909 24733) has been added. This support is in header file 4910 <decimal/decimal>, uses namespace std::decimal, and includes 4911 classes decimal32, decimal64, and decimal128. 4912 * Sources have been audited for application of function attributes 4913 nothrow, const, pure, and noreturn. 4914 * Python pretty-printers have been added for many standard library 4915 components that simplify the internal representation and present a 4916 more intuitive view of components when used with 4917 appropriately-advanced versions of GDB. For more information, 4918 please consult the more [17]detailed description. 4919 * The default behavior for comparing typeinfo names has changed, so 4920 in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero. 4921 * The new -static-libstdc++ option directs g++ to link the C++ 4922 library statically, even if the default would normally be to link 4923 it dynamically. 4924 4925 Fortran 4926 4927 * The COMMON default padding has been changed - instead of adding the 4928 padding before a variable it is now added afterwards, which 4929 increases the compatibility with other vendors and helps to obtain 4930 the correct output in some cases. Cf. also the -falign-commons 4931 option ([18]added in 4.4). 4932 * The -finit-real= option now also supports the value snan for 4933 signalling not-a-number; to be effective, one additionally needs to 4934 enable trapping (e.g. via -ffpe-trap=). Note: Compile-time 4935 optimizations can turn a signalling NaN into a quiet one. 4936 * The new option -fcheck= has been added with the options bounds, 4937 array-temps, do, pointer, and recursive. The bounds and array-temps 4938 options are equivalent to -fbounds-check and 4939 -fcheck-array-temporaries. The do option checks for invalid 4940 modification of loop iteration variables, and the recursive option 4941 tests for recursive calls to subroutines/functions which are not 4942 marked as recursive. With pointer pointer association checks in 4943 calls are performed; however, neither undefined pointers nor 4944 pointers in expressions are handled. Using -fcheck=all enables all 4945 these run-time checks. 4946 * The run-time checking -fcheck=bounds now warns about invalid string 4947 lengths of character dummy arguments. Additionally, more 4948 compile-time checks have been added. 4949 * The new option [19]-fno-protect-parens has been added; if set, the 4950 compiler may reorder REAL and COMPLEX expressions without regard to 4951 parentheses. 4952 * GNU Fortran no longer links against libgfortranbegin. As before, 4953 MAIN__ (assembler symbol name) is the actual Fortran main program, 4954 which is invoked by the main function. However, main is now 4955 generated and put in the same object file as MAIN__. For the time 4956 being, libgfortranbegin still exists for backward compatibility. 4957 For details see the new [20]Mixed-Language Programming chapter in 4958 the manual. 4959 * The I/O library was restructured for performance and cleaner code. 4960 * Array assignments and WHERE are now run in parallel when OpenMP's 4961 WORKSHARE is used. 4962 * The experimental option -fwhole-file was added. The option allows 4963 whole-file checking of procedure arguments and allows for better 4964 optimizations. It can also be used with -fwhole-program, which is 4965 now also supported in gfortran. 4966 * More Fortran 2003 and Fortran 2008 mathematical functions can now 4967 be used as initialization expressions. 4968 * Some extended attributes such as STDCALL are now supported via the 4969 [21]GCC$ compiler directive. 4970 * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN 4971 intrinsic behaves now as if zero were always positive. 4972 * For legacy compatibiliy: On Cygwin and MinGW, the special files 4973 CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now 4974 supported. 4975 * Fortran 2003 support has been extended: 4976 + Procedure-pointer function results and procedure-pointer 4977 components (including PASS), 4978 + allocatable scalars (experimental), 4979 + DEFERRED type-bound procedures, 4980 + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements 4981 have been implemented. 4982 + The ALLOCATE statement supports type-specs and the SOURCE= 4983 argument. 4984 + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC 4985 type-bound procedure (i.e. as type-bound operators). 4986 + Rounding (ROUND=, RZ, ...) for output is now supported. 4987 + The INT_FAST{8,16,32,64,128}_T kind type parameters of the 4988 intrinsic module ISO_C_BINDING are now supported, except for 4989 the targets listed above as ones where GCC does not have 4990 <stdint.h> type information. 4991 + Extensible derived types with type-bound procedure or 4992 procedure pointer with PASS attribute now have to use CLASS in 4993 line with the Fortran 2003 standard; the workaround to use 4994 TYPE is no longer supported. 4995 + [22]Experimental, incomplete support for polymorphism, 4996 including CLASS, SELECT TYPE and dynamic dispatch of 4997 type-bound procedure calls. Some features do not work yet such 4998 as unlimited polymorphism (CLASS(*)). 4999 * Fortran 2008 support has been extended: 5000 + The OPEN statement now supports the NEWUNIT= option, which 5001 returns a unique file unit, thus preventing inadvertent use of 5002 the same unit in different parts of the program. 5003 + Support for unlimited format items has been added. 5004 + The INT{8,16,32} and REAL{32,64,128} kind type parameters of 5005 the intrinsic module ISO_FORTRAN_ENV are now supported. 5006 + Using complex arguments with TAN, SINH, COSH, TANH, ASIN, 5007 ACOS, and ATAN is now possible; the functions ASINH, ACOSH, 5008 and ATANH have been added (for real and complex arguments) and 5009 ATAN(Y,X) is now an alias for ATAN2(Y,X). 5010 + The BLOCK construct has been implemented. 5011 5012 New Targets and Target Specific Improvements 5013 5014 AIX 5015 5016 * Full cross-toolchain support now available with GNU Binutils 5017 5018 ARM 5019 5020 * GCC now supports the Cortex-M0 and Cortex-A5 processors. 5021 * GCC now supports the ARM v7E-M architecture. 5022 * GCC now supports VFPv4-based FPUs and FPUs with 5023 single-precision-only VFP. 5024 * GCC has many improvements to optimization for other ARM processors, 5025 including scheduling support for the integer pipeline on Cortex-A9. 5026 * GCC now supports the IEEE 754-2008 half-precision floating-point 5027 type, and a variant ARM-specific half-precision type. This type is 5028 specified using __fp16, with the layout determined by 5029 -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and 5030 VFPv4 half-precision instructions will be used. 5031 * GCC now supports the variant of AAPCS that uses VFP registers for 5032 parameter passing and return values. 5033 5034 AVR 5035 5036 * The -mno-tablejump option has been removed because it has the same 5037 effect as the -fno-jump-tables option. 5038 * Added support for these new AVR devices: 5039 + ATmega8U2 5040 + ATmega16U2 5041 + ATmega32U2 5042 5043 IA-32/x86-64 5044 5045 * GCC now will set the default for -march= based on the configure 5046 target. 5047 * GCC now supports handling floating-point excess precision arising 5048 from use of the x87 floating-point unit in a way that conforms to 5049 ISO C99. This is enabled with -fexcess-precision=standard and with 5050 standards conformance options such as -std=c99, and may be disabled 5051 using -fexcess-precision=fast. 5052 * Support for the Intel Atom processor is now available through the 5053 -march=atom and -mtune=atom options. 5054 * A new -mcrc32 option is now available to enable crc32 intrinsics. 5055 * A new -mmovbe option is now available to enable GCC to use the 5056 movbe instruction to implement __builtin_bswap32 and 5057 __builtin_bswap64. 5058 * SSE math now can be enabled by default at configure time with the 5059 new --with-fpmath=sse option. 5060 * There is a new intrinsic header file, <x86intrin.h>. It should be 5061 included before using any IA-32/x86-64 intrinsics. 5062 * Support for the XOP, FMA4, and LWP instruction sets for the AMD 5063 Orochi processors are now available with the -mxop, -mfma4, and 5064 -mlwp options. 5065 * The -mabm option enables GCC to use the popcnt and lzcnt 5066 instructions on AMD processors. 5067 * The -mpopcnt option enables GCC to use the popcnt instructions on 5068 both AMD and Intel processors. 5069 5070 M68K/ColdFire 5071 5072 * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x 5073 and 5441x devices. 5074 * GCC now supports thread-local storage (TLS) on M68K and ColdFire 5075 processors. 5076 5077 MeP 5078 5079 Support has been added for the Toshiba Media embedded Processor (MeP, 5080 or mep-elf) embedded target. 5081 5082 MIPS 5083 5084 * GCC now supports MIPS 1004K processors. 5085 * GCC can now be configured with options --with-arch-32, 5086 --with-arch-64, --with-tune-32 and --with-tune-64 to control the 5087 default optimization separately for 32-bit and 64-bit modes. 5088 * MIPS targets now support an alternative _mcount interface, in which 5089 register $12 points to the function's save slot for register $31. 5090 This interface is selected by the -mcount-ra-address option; see 5091 the documentation for more details. 5092 * GNU/Linux targets can now generate read-only .eh_frame sections. 5093 This optimization requires GNU binutils 2.20 or above, and is only 5094 available if GCC is configured with a suitable version of binutils. 5095 * GNU/Linux targets can now attach special relocations to indirect 5096 calls, so that the linker can turn them into direct jumps or 5097 branches. This optimization requires GNU binutils 2.20 or later, 5098 and is automatically selected if GCC is configured with an 5099 appropriate version of binutils. It can be explicitly enabled or 5100 disabled using the -mrelax-pic-calls command-line option. 5101 * GCC now generates more heavily-optimized atomic operations on 5102 Octeon processors. 5103 * MIPS targets now support the -fstack-protector option. 5104 * GCC now supports an -msynci option, which specifies that synci is 5105 enough to flush the instruction cache, without help from the 5106 operating system. GCC uses this information to optimize 5107 automatically-generated cache flush operations, such as those used 5108 for nested functions in C. There is also a --with-synci 5109 configure-time option, which makes -msynci the default. 5110 * GCC supports four new function attributes for interrupt handlers: 5111 interrupt, use_shadow_register_set, keep_interrupts_masked and 5112 use_debug_exception_return. See the documentation for more details 5113 about these attributes. 5114 5115 RS/6000 (POWER/PowerPC) 5116 5117 * GCC now supports the Power ISA 2.06, which includes the VSX 5118 instructions that add vector 64-bit floating point support, new 5119 population count instructions, and conversions between floating 5120 point and unsigned types. 5121 * Support for the power7 processor is now available through the 5122 -mcpu=power7 and -mtune=power7. 5123 * GCC will now vectorize loops that contain simple math functions 5124 like copysign when generating code for altivec or VSX targets. 5125 * Support for the A2 processor is now available through the -mcpu=a2 5126 and -mtune=a2 options. 5127 * Support for the 476 processor is now available through the 5128 -mcpu={476,476fp} and -mtune={476,476fp} options. 5129 * Support for the e500mc64 processor is now available through the 5130 -mcpu=e500mc64 and -mtune=e500mc64 options. 5131 * GCC can now be configured with options --with-cpu-32, 5132 --with-cpu-64, --with-tune-32 and --with-tune-64 to control the 5133 default optimization separately for 32-bit and 64-bit modes. 5134 * Starting with GCC 4.5.4, vectors of type vector long long or vector 5135 long are passed and returned in the same method as other vectors 5136 with the VSX instruction set. Previously the GCC compiler did not 5137 adhere to the ABI for 128-bit vectors with 64-bit integer base 5138 types (PR 48857). This is also fixed in the GCC 4.6.1 release. 5139 5140 RX 5141 5142 Support has been added for the Renesas RX Processor (rx-elf) target. 5143 5144 Operating Systems 5145 5146 Windows (Cygwin and MinGW) 5147 5148 * GCC now installs all the major language runtime libraries as DLLs 5149 when configured with the --enable-shared option. 5150 * GCC now makes use of the new support for aligned common variables 5151 in versions of binutils >= 2.20 to fix bugs in the support for SSE 5152 data types. 5153 * Improvements to the libffi support library increase the reliability 5154 of code generated by GCJ on all Windows platforms. Libgcj is 5155 enabled by default for the first time. 5156 * Libtool improvements simplify installation by placing the generated 5157 DLLs in the correct binaries directory. 5158 * Numerous other minor bugfixes and improvements, and substantial 5159 enhancements to the Fortran language support library. 5160 5161 > 5162 5163 Other significant improvements 5164 5165 Plugins 5166 5167 * It is now possible to extend the compiler without having to modify 5168 its source code. A new option -fplugin=file.so tells GCC to load 5169 the shared object file.so and execute it as part of the compiler. 5170 The internal documentation describes the details on how plugins can 5171 interact with the compiler. 5172 5173 Installation changes 5174 5175 * The move to newer autotools changed default installation 5176 directories and switches to control them: The --with-datarootdir, 5177 --with-docdir, --with-pdfdir, and --with-htmldir switches are not 5178 used any more. Instead, you can now use --datarootdir, --docdir, 5179 --htmldir, and --pdfdir. The default installation directories have 5180 changed as follows according to the GNU Coding Standards: 5181 5182 datarootdir read-only architecture-independent data root [PREFIX/share] 5183 localedir locale-specific message catalogs [DATAROOTDIR/locale] 5184 docdir documentation root [DATAROOTDIR/doc/PACKAGE] 5185 htmldir html documentation [DOCDIR] 5186 dvidir dvi documentation [DOCDIR] 5187 pdfdir pdf documentation [DOCDIR] 5188 psdir ps documentation [DOCDIR] 5189 The following variables have new default values: 5190 5191 datadir read-only architecture-independent data [DATAROOTDIR] 5192 infodir info documentation [DATAROOTDIR/info] 5193 mandir man documentation [DATAROOTDIR/man] 5194 5195 GCC 4.5.1 5196 5197 This is the [23]list of problem reports (PRs) from GCC's bug tracking 5198 system that are known to be fixed in the 4.5.1 release. This list might 5199 not be complete (that is, it is possible that some PRs that have been 5200 fixed are not listed here). 5201 5202 All languages 5203 5204 * GCC's new link-time optimizer ([24]-flto) now also works on a few 5205 non-ELF targets: 5206 + Cygwin (*-cygwin*) 5207 + MinGW (*-mingw*) 5208 + Darwin on x86-64 (x86_64-apple-darwin*) 5209 LTO is not enabled by default for these targets. To enable LTO, you 5210 should configure with the --enable-lto option. 5211 5212 GCC 4.5.2 5213 5214 This is the [25]list of problem reports (PRs) from GCC's bug tracking 5215 system that are known to be fixed in the 4.5.2 release. This list might 5216 not be complete (that is, it is possible that some PRs that have been 5217 fixed are not listed here). 5218 5219 GCC 4.5.3 5220 5221 This is the [26]list of problem reports (PRs) from GCC's bug tracking 5222 system that are known to be fixed in the 4.5.3 release. This list might 5223 not be complete (that is, it is possible that some PRs that have been 5224 fixed are not listed here). 5225 5226 On the PowerPC compiler, the Altivec builtin functions vec_ld and 5227 vec_st have been modified to generate the Altivec memory instructions 5228 LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5 5229 release, these builtin functions were changed to generate VSX memory 5230 reference instructions instead of Altivec memory instructions, but 5231 there are differences between the two instructions. If the VSX 5232 instruction set is available, you can now use the new builtin functions 5233 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 5234 instructions. 5235 5236 GCC 4.5.4 5237 5238 This is the [27]list of problem reports (PRs) from GCC's bug tracking 5239 system that are known to be fixed in the 4.5.4 release. This list might 5240 not be complete (that is, it is possible that some PRs that have been 5241 fixed are not listed here). 5242 5243 5244 For questions related to the use of GCC, please consult these web 5245 pages and the [28]GCC manuals. If that fails, the 5246 [29]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5247 web pages and the development of GCC are welcome on our developer 5248 list at [30]gcc (a] gcc.gnu.org. All of [31]our lists have public 5249 archives. 5250 5251 Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and 5252 distribution of this entire article is permitted in any medium, 5253 provided this notice is preserved. 5254 5255 These pages are [33]maintained by the GCC team. Last modified 5256 2014-12-06[34]. 5257 5258 References 5259 5260 1. http://www.multiprecision.org/ 5261 2. https://gcc.gnu.org/install/prerequisites.html 5262 3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html 5263 4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted 5264 5. http://gcc.gnu.org/gcc-4.5/changes.html#x86 5265 6. http://www.multiprecision.org/ 5266 7. https://gcc.gnu.org/PR30789 5267 8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 5268 9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802 5269 10. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800 5270 11. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html 5271 12. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html 5272 13. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757 5273 14. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176 5274 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html 5275 16. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733 5276 17. http://sourceware.org/gdb/wiki/STLSupport 5277 18. http://gcc.gnu.org/gcc-4.4/changes.html 5278 19. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 5279 20. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html 5280 21. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html 5281 22. https://gcc.gnu.org/wiki/OOP 5282 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1 5283 24. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 5284 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2 5285 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3 5286 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4 5287 28. https://gcc.gnu.org/onlinedocs/ 5288 29. mailto:gcc-help (a] gcc.gnu.org 5289 30. mailto:gcc (a] gcc.gnu.org 5290 31. https://gcc.gnu.org/lists.html 5291 32. http://www.fsf.org/ 5292 33. https://gcc.gnu.org/about.html 5293 34. http://validator.w3.org/check/referer 5294 ====================================================================== 5295 http://gcc.gnu.org/gcc-4.4/index.html 5296 5297 GCC 4.4 Release Series 5298 5299 March 13, 2012 5300 5301 The [1]GNU project and the GCC developers are pleased to announce the 5302 release of GCC 4.4.7. 5303 5304 This release is a bug-fix release, containing fixes for regressions in 5305 GCC 4.4.6 relative to previous releases of GCC. 5306 5307 Release History 5308 5309 GCC 4.4.7 5310 March 13, 2012 ([2]changes) 5311 5312 GCC 4.4.6 5313 April 16, 2011 ([3]changes) 5314 5315 GCC 4.4.5 5316 October 1, 2010 ([4]changes) 5317 5318 GCC 4.4.4 5319 April 29, 2010 ([5]changes) 5320 5321 GCC 4.4.3 5322 January 21, 2010 ([6]changes) 5323 5324 GCC 4.4.2 5325 October 15, 2009 ([7]changes) 5326 5327 GCC 4.4.1 5328 July 22, 2009 ([8]changes) 5329 5330 GCC 4.4.0 5331 April 21, 2009 ([9]changes) 5332 5333 References and Acknowledgements 5334 5335 GCC used to stand for the GNU C Compiler, but since the compiler 5336 supports several other languages aside from C, it now stands for the 5337 GNU Compiler Collection. 5338 5339 A list of [10]successful builds is updated as new information becomes 5340 available. 5341 5342 The GCC developers would like to thank the numerous people that have 5343 contributed new features, improvements, bug fixes, and other changes as 5344 well as test results to GCC. This [11]amazing group of volunteers is 5345 what makes GCC successful. 5346 5347 For additional information about GCC please refer to the [12]GCC 5348 project web site or contact the [13]GCC development mailing list. 5349 5350 To obtain GCC please use [14]our mirror sites or [15]our SVN server. 5351 5352 5353 For questions related to the use of GCC, please consult these web 5354 pages and the [16]GCC manuals. If that fails, the 5355 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5356 web pages and the development of GCC are welcome on our developer 5357 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 5358 archives. 5359 5360 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 5361 distribution of this entire article is permitted in any medium, 5362 provided this notice is preserved. 5363 5364 These pages are [21]maintained by the GCC team. Last modified 5365 2014-06-28[22]. 5366 5367 References 5368 5369 1. http://www.gnu.org/ 5370 2. http://gcc.gnu.org/gcc-4.4/changes.html 5371 3. http://gcc.gnu.org/gcc-4.4/changes.html 5372 4. http://gcc.gnu.org/gcc-4.4/changes.html 5373 5. http://gcc.gnu.org/gcc-4.4/changes.html 5374 6. http://gcc.gnu.org/gcc-4.4/changes.html 5375 7. http://gcc.gnu.org/gcc-4.4/changes.html 5376 8. http://gcc.gnu.org/gcc-4.4/changes.html 5377 9. http://gcc.gnu.org/gcc-4.4/changes.html 5378 10. http://gcc.gnu.org/gcc-4.4/buildstat.html 5379 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 5380 12. http://gcc.gnu.org/index.html 5381 13. mailto:gcc (a] gcc.gnu.org 5382 14. http://gcc.gnu.org/mirrors.html 5383 15. http://gcc.gnu.org/svn.html 5384 16. https://gcc.gnu.org/onlinedocs/ 5385 17. mailto:gcc-help (a] gcc.gnu.org 5386 18. mailto:gcc (a] gcc.gnu.org 5387 19. https://gcc.gnu.org/lists.html 5388 20. http://www.fsf.org/ 5389 21. https://gcc.gnu.org/about.html 5390 22. http://validator.w3.org/check/referer 5391 ====================================================================== 5392 http://gcc.gnu.org/gcc-4.4/changes.html 5393 5394 GCC 4.4 Release Series 5395 Changes, New Features, and Fixes 5396 5397 The latest release in the 4.4 release series is [1]GCC 4.4.7. 5398 5399 Caveats 5400 5401 * __builtin_stdarg_start has been completely removed from GCC. 5402 Support for <varargs.h> had been deprecated since GCC 4.0. Use 5403 __builtin_va_start as a replacement. 5404 * Some of the errors issued by the C++ front end that could be 5405 downgraded to warnings in previous releases by using -fpermissive 5406 are now warnings by default. They can be converted into errors by 5407 using -pedantic-errors. 5408 * Use of the cpp assertion extension will now emit a warning when 5409 -Wdeprecated or -pedantic is used. This extension has been 5410 deprecated for many years, but never warned about. 5411 * Packed bit-fields of type char were not properly bit-packed on many 5412 targets prior to GCC 4.4. On these targets, the fix in GCC 4.4 5413 causes an ABI change. For example there is no longer a 4-bit 5414 padding between field a and b in this structure: 5415 struct foo 5416 { 5417 char a:4; 5418 char b:8; 5419 } __attribute__ ((packed)); 5420 There is a new warning to help identify fields that are affected: 5421 foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4 5422 The warning can be disabled with -Wno-packed-bitfield-compat. 5423 * On ARM EABI targets, the C++ mangling of the va_list type has been 5424 changed to conform to the current revision of the EABI. This does 5425 not affect the libstdc++ library included with GCC. 5426 * The SCOUNT and POS bits of the MIPS DSP control register are now 5427 treated as global. Previous versions of GCC treated these fields as 5428 call-clobbered instead. 5429 * The MIPS port no longer recognizes the h asm constraint. It was 5430 necessary to remove this constraint in order to avoid generating 5431 unpredictable code sequences. 5432 One of the main uses of the h constraint was to extract the high 5433 part of a multiplication on 64-bit targets. For example: 5434 asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y)); 5435 You can now achieve the same effect using 128-bit types: 5436 typedef unsigned int uint128_t __attribute__((mode(TI))); 5437 result = ((uint128_t) x * y) >> 64; 5438 The second sequence is better in many ways. For example, if x and y 5439 are constants, the compiler can perform the multiplication at 5440 compile time. If x and y are not constants, the compiler can 5441 schedule the runtime multiplication better than it can schedule an 5442 asm statement. 5443 * Support for a number of older systems and recently unmaintained or 5444 untested target ports of GCC has been declared obsolete in GCC 4.4. 5445 Unless there is activity to revive them, the next release of GCC 5446 will have their sources permanently removed. 5447 The following ports for individual systems on particular 5448 architectures have been obsoleted: 5449 + Generic a.out on IA32 and m68k (i[34567]86-*-aout*, 5450 m68k-*-aout*) 5451 + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*, 5452 armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*, 5453 sh-*-*). This does not affect other more specific targets 5454 using the COFF object format on those architectures, or the 5455 more specific H8300 and SH targets (h8300-*-rtems*, 5456 h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*, 5457 sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks). 5458 + 2BSD on PDP-11 (pdp11-*-bsd) 5459 + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*, 5460 powerpc-ibm-aix4.[12]*) 5461 + Tuning support for Itanium1 (Merced) variants. Note that code 5462 tuned for Itanium2 should also run correctly on Itanium1. 5463 * The protoize and unprotoize utilities have been obsoleted and will 5464 be removed in GCC 4.5. These utilities have not been installed by 5465 default since GCC 3.0. 5466 * Support has been removed for all the [2]configurations obsoleted in 5467 GCC 4.3. 5468 * Unknown -Wno-* options are now silently ignored by GCC if no other 5469 diagnostics are issued. If other diagnostics are issued, then GCC 5470 warns about the unknown options. 5471 * More information on porting to GCC 4.4 from previous versions of 5472 GCC can be found in the [3]porting guide for this release. 5473 5474 General Optimizer Improvements 5475 5476 * A new command-line switch -findirect-inlining has been added. When 5477 turned on it allows the inliner to also inline indirect calls that 5478 are discovered to have known targets at compile time thanks to 5479 previous inlining. 5480 * A new command-line switch -ftree-switch-conversion has been added. 5481 This new pass turns simple initializations of scalar variables in 5482 switch statements into initializations from a static array, given 5483 that all the values are known at compile time and the ratio between 5484 the new array size and the original switch branches does not exceed 5485 the parameter --param switch-conversion-max-branch-ratio (default 5486 is eight). 5487 * A new command-line switch -ftree-builtin-call-dce has been added. 5488 This optimization eliminates unnecessary calls to certain builtin 5489 functions when the return value is not used, in cases where the 5490 calls can not be eliminated entirely because the function may set 5491 errno. This optimization is on by default at -O2 and above. 5492 * A new command-line switch -fconserve-stack directs the compiler to 5493 minimize stack usage even if it makes the generated code slower. 5494 This affects inlining decisions. 5495 * When the assembler supports it, the compiler will now emit unwind 5496 information using assembler .cfi directives. This makes it possible 5497 to use such directives in inline assembler code. The new option 5498 -fno-dwarf2-cfi-asm directs the compiler to not use .cfi 5499 directives. 5500 * The [4]Graphite branch has been merged. This merge has brought in a 5501 new framework for loop optimizations based on a polyhedral 5502 intermediate representation. These optimizations apply to all the 5503 languages supported by GCC. The following new code transformations 5504 are available in GCC 4.4: 5505 + -floop-interchange performs loop interchange transformations 5506 on loops. Interchanging two nested loops switches the inner 5507 and outer loops. For example, given a loop like: 5508 DO J = 1, M 5509 DO I = 1, N 5510 A(J, I) = A(J, I) * C 5511 ENDDO 5512 ENDDO 5513 5514 loop interchange will transform the loop as if the user had 5515 written: 5516 DO I = 1, N 5517 DO J = 1, M 5518 A(J, I) = A(J, I) * C 5519 ENDDO 5520 ENDDO 5521 5522 which can be beneficial when N is larger than the caches, 5523 because in Fortran, the elements of an array are stored in 5524 memory contiguously by column, and the original loop iterates 5525 over rows, potentially creating at each access a cache miss. 5526 + -floop-strip-mine performs loop strip mining transformations 5527 on loops. Strip mining splits a loop into two nested loops. 5528 The outer loop has strides equal to the strip size and the 5529 inner loop has strides of the original loop within a strip. 5530 For example, given a loop like: 5531 DO I = 1, N 5532 A(I) = A(I) + C 5533 ENDDO 5534 5535 loop strip mining will transform the loop as if the user had 5536 written: 5537 DO II = 1, N, 4 5538 DO I = II, min (II + 3, N) 5539 A(I) = A(I) + C 5540 ENDDO 5541 ENDDO 5542 5543 + -floop-block performs loop blocking transformations on loops. 5544 Blocking strip mines each loop in the loop nest such that the 5545 memory accesses of the element loops fit inside caches. For 5546 example, given a loop like: 5547 DO I = 1, N 5548 DO J = 1, M 5549 A(J, I) = B(I) + C(J) 5550 ENDDO 5551 ENDDO 5552 5553 loop blocking will transform the loop as if the user had 5554 written: 5555 DO II = 1, N, 64 5556 DO JJ = 1, M, 64 5557 DO I = II, min (II + 63, N) 5558 DO J = JJ, min (JJ + 63, M) 5559 A(J, I) = B(I) + C(J) 5560 ENDDO 5561 ENDDO 5562 ENDDO 5563 ENDDO 5564 5565 which can be beneficial when M is larger than the caches, 5566 because the innermost loop will iterate over a smaller amount 5567 of data that can be kept in the caches. 5568 * A new register allocator has replaced the old one. It is called 5569 integrated register allocator (IRA) because coalescing, register 5570 live range splitting, and hard register preferencing are done 5571 on-the-fly during coloring. It also has better integration with the 5572 reload pass. IRA is a regional register allocator which uses modern 5573 Chaitin-Briggs coloring instead of Chow's priority coloring used in 5574 the old register allocator. More info about IRA internals and 5575 options can be found in the GCC manuals. 5576 * A new instruction scheduler and software pipeliner, based on the 5577 selective scheduling approach, has been added. The new pass 5578 performs instruction unification, register renaming, substitution 5579 through register copies, and speculation during scheduling. The 5580 software pipeliner is able to pipeline non-countable loops. The new 5581 pass is targeted at scheduling-eager in-order platforms. In GCC 4.4 5582 it is available for the Intel Itanium platform working by default 5583 as the second scheduling pass (after register allocation) at the 5584 -O3 optimization level. 5585 * When using -fprofile-generate with a multi-threaded program, the 5586 profile counts may be slightly wrong due to race conditions. The 5587 new -fprofile-correction option directs the compiler to apply 5588 heuristics to smooth out the inconsistencies. By default the 5589 compiler will give an error message when it finds an inconsistent 5590 profile. 5591 * The new -fprofile-dir=PATH option permits setting the directory 5592 where profile data files are stored when using -fprofile-generate 5593 and friends, and the directory used when reading profile data files 5594 using -fprofile-use and friends. 5595 5596 New warning options 5597 5598 * The new -Wframe-larger-than=NUMBER option directs GCC to emit a 5599 warning if any stack frame is larger than NUMBER bytes. This may be 5600 used to help ensure that code fits within a limited amount of stack 5601 space. 5602 * The command-line option -Wlarger-than-N is now written as 5603 -Wlarger-than=N and the old form is deprecated. 5604 * The new -Wno-mudflap option disables warnings about constructs 5605 which can not be instrumented when using -fmudflap. 5606 5607 New Languages and Language specific improvements 5608 5609 * Version 3.0 of the [5]OpenMP specification is now supported for the 5610 C, C++, and Fortran compilers. 5611 * New character data types, per [6]TR 19769: New character types in 5612 C, are now supported for the C compiler in -std=gnu99 mode, as 5613 __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in 5614 -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too. 5615 5616 C family 5617 5618 * A new optimize attribute was added to allow programmers to change 5619 the optimization level and particular optimization options for an 5620 individual function. You can also change the optimization options 5621 via the GCC optimize pragma for functions defined after the pragma. 5622 The GCC push_options pragma and the GCC pop_options pragma allow 5623 you temporarily save and restore the options used. The GCC 5624 reset_options pragma restores the options to what was specified on 5625 the command line. 5626 * Uninitialized warnings do not require enabling optimization 5627 anymore, that is, -Wuninitialized can be used together with -O0. 5628 Nonetheless, the warnings given by -Wuninitialized will probably be 5629 more accurate if optimization is enabled. 5630 * -Wparentheses now warns about expressions such as (!x | y) and (!x 5631 & y). Using explicit parentheses, such as in ((!x) | y), silences 5632 this warning. 5633 * -Wsequence-point now warns within if, while,do while and for 5634 conditions, and within for begin/end expressions. 5635 * A new option -dU is available to dump definitions of preprocessor 5636 macros that are tested or expanded. 5637 5638 C++ 5639 5640 * [7]Improved experimental support for the upcoming ISO C++ standard, 5641 C++0x. Including support for auto, inline namespaces, generalized 5642 initializer lists, defaulted and deleted functions, new character 5643 types, and scoped enums. 5644 * Those errors that may be downgraded to warnings to build legacy 5645 code now mention -fpermissive when -fdiagnostics-show-option is 5646 enabled. 5647 * -Wconversion now warns if the result of a static_cast to enumeral 5648 type is unspecified because the value is outside the range of the 5649 enumeral type. 5650 * -Wuninitialized now warns if a non-static reference or non-static 5651 const member appears in a class without constructors. 5652 * G++ now properly implements value-initialization, so objects with 5653 an initializer of () and an implicitly defined default constructor 5654 will be zero-initialized before the default constructor is called. 5655 5656 Runtime Library (libstdc++) 5657 5658 * Improved experimental support for the upcoming ISO C++ standard, 5659 C++0x, including: 5660 + Support for <chrono>, <condition_variable>, <cstdatomic>, 5661 <forward_list>, <initializer_list>, <mutex>, <ratio>, 5662 <system_error>, and <thread>. 5663 + unique_ptr, <algorithm> additions, exception propagation, and 5664 support for the new character types in <string> and <limits>. 5665 + Existing facilities now exploit initializer lists, defaulted 5666 and deleted functions, and the newly implemented core C++0x 5667 features. 5668 + Some standard containers are more efficient together with 5669 stateful allocators, i.e., no allocator is constructed on the 5670 fly at element construction time. 5671 * Experimental support for non-standard pointer types in containers. 5672 * The long standing libstdc++/30928 has been fixed for targets 5673 running glibc 2.10 or later. 5674 * As usual, many small and larger bug fixes, in particular quite a 5675 few corner cases in <locale>. 5676 5677 Fortran 5678 5679 * GNU Fortran now employs libcpp directly instead of using cc1 as an 5680 external preprocessor. The [8]-cpp option was added to allow manual 5681 invocation of the preprocessor without relying on filename 5682 extensions. 5683 * The [9]-Warray-temporaries option warns about array temporaries 5684 generated by the compiler, as an aid to optimization. 5685 * The [10]-fcheck-array-temporaries option has been added, printing a 5686 notification at run time, when an array temporary had to be created 5687 for an function argument. Contrary to -Warray-temporaries the 5688 warning is only printed if the array is noncontiguous. 5689 * Improved generation of DWARF debugging symbols 5690 * If using an intrinsic not part of the selected standard (via -std= 5691 and -fall-intrinsics) gfortran will now treat it as if this 5692 procedure were declared EXTERNAL and try to link to a user-supplied 5693 procedure. -Wintrinsics-std will warn whenever this happens. The 5694 now-useless option -Wnonstd-intrinsic was removed. 5695 * The flag -falign-commons has been added to control the alignment of 5696 variables in COMMON blocks, which is enabled by default in line 5697 with previous GCC version. Using -fno-align-commons one can force 5698 commons to be contiguous in memory as required by the Fortran 5699 standard, however, this slows down the memory access. The option 5700 -Walign-commons, which is enabled by default, warns when padding 5701 bytes were added for alignment. The proper solution is to sort the 5702 common objects by decreasing storage size, which avoids the 5703 alignment problems. 5704 * Fortran 2003 support has been extended: 5705 + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is 5706 now supported (except internal reads from/writes to wide 5707 strings). [11]-fbackslash now supports also \unnnn and 5708 \Unnnnnnnn to enter Unicode characters. 5709 + Asynchronous I/O (implemented as synchronous I/O) and the 5710 decimal=, size=, sign=, pad=, blank=, and delim= specifiers 5711 are now supported in I/O statements. 5712 + Support for Fortran 2003 structure constructors and for array 5713 constructor with typespec has been added. 5714 + Procedure Pointers (but not yet as component in derived types 5715 and as function results) are now supported. 5716 + Abstract types, type extension, and type-bound procedures 5717 (both PROCEDURE and GENERIC but not as operators). Note: As 5718 CLASS/polymorphyic types are not implemented, type-bound 5719 procedures with PASS accept as non-standard extension TYPE 5720 arguments. 5721 * Fortran 2008 support has been added: 5722 + The -std=f2008 option and support for the file extensions 5723 .f2008 and .F2008 has been added. 5724 + The g0 format descriptor is now supported. 5725 + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH, 5726 ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED 5727 are now available (some of them existed as GNU extension 5728 before). Note: The hyperbolic functions are not yet supporting 5729 complex arguments and the three- argument version of BESSEL_*N 5730 is not available. 5731 + The bit intrinsics LEADZ and TRAILZ have been added. 5732 5733 Java (GCJ) 5734 5735 Ada 5736 5737 * The Ada runtime now supports multilibs on many platforms including 5738 x86_64, SPARC and PowerPC. Their build is enabled by default. 5739 5740 New Targets and Target Specific Improvements 5741 5742 ARM 5743 5744 * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and 5745 Cortex-R4F processors and has many other improvements to 5746 optimization for ARM processors. 5747 * GCC now supports the VFPv3 variant with 16 double-precision 5748 registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been 5749 renamed to -mfpu=vfpv3. 5750 * GCC now supports the -mfix-cortex-m3-ldrd option to work around an 5751 erratum on Cortex-M3 processors. 5752 * GCC now supports the __sync_* atomic operations for ARM EABI 5753 GNU/Linux. 5754 * The section anchors optimization is now enabled by default when 5755 optimizing for ARM. 5756 * GCC now uses a new EABI-compatible profiling interface for EABI 5757 targets. This requires a function __gnu_mcount_nc, which is 5758 provided by GNU libc versions 2.8 and later. 5759 5760 AVR 5761 5762 * The -mno-tablejump option has been deprecated because it has the 5763 same effect as the -fno-jump-tables option. 5764 * Added support for these new AVR devices: 5765 + ATA6289 5766 + ATtiny13A 5767 + ATtiny87 5768 + ATtiny167 5769 + ATtiny327 5770 + ATmega8C1 5771 + ATmega16C1 5772 + ATmega32C1 5773 + ATmega8M1 5774 + ATmega16M1 5775 + ATmega32M1 5776 + ATmega32U4 5777 + ATmega16HVB 5778 + ATmega4HVD 5779 + ATmega8HVD 5780 + ATmega64C1 5781 + ATmega64M1 5782 + ATmega16U4 5783 + ATmega32U6 5784 + ATmega128RFA1 5785 + AT90PWM81 5786 + AT90SCR100 5787 + M3000F 5788 + M3000S 5789 + M3001B 5790 5791 IA-32/x86-64 5792 5793 * Support for Intel AES built-in functions and code generation is 5794 available via -maes. 5795 * Support for Intel PCLMUL built-in function and code generation is 5796 available via -mpclmul. 5797 * Support for Intel AVX built-in functions and code generation is 5798 available via -mavx. 5799 * Automatically align the stack for local variables with alignment 5800 requirement. 5801 * GCC can now utilize the SVML library for vectorizing calls to a set 5802 of C99 functions if -mveclibabi=svml is specified and you link to 5803 an SVML ABI compatible library. 5804 * On x86-64, the ABI has been changed in the following cases to 5805 conform to the x86-64 ABI: 5806 + Passing/returning structures with flexible array member: 5807 struct foo 5808 { 5809 int i; 5810 int flex[]; 5811 }; 5812 + Passing/returning structures with complex float member: 5813 struct foo 5814 { 5815 int i; 5816 __complex__ float f; 5817 }; 5818 + Passing/returning unions with long double member: 5819 union foo 5820 { 5821 int x; 5822 long double ld; 5823 }; 5824 Code built with previous versions of GCC that uses any of these is 5825 not compatible with code built with GCC 4.4.0 or later. 5826 * A new target attribute was added to allow programmers to change the 5827 target options like -msse2 or -march=k8 for an individual function. 5828 You can also change the target options via the GCC target pragma 5829 for functions defined after the pragma. 5830 * GCC can now be configured with options --with-arch-32, 5831 --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and 5832 --with-tune-64 to control the default optimization separately for 5833 32-bit and 64-bit modes. 5834 5835 IA-32/IA64 5836 5837 * Support for __float128 (TFmode) IEEE quad type and corresponding 5838 TCmode IEEE complex quad type is available via the soft-fp library 5839 on IA-32/IA64 targets. This includes basic arithmetic operations 5840 (addition, subtraction, negation, multiplication and division) on 5841 __float128 real and TCmode complex values, the full set of IEEE 5842 comparisons between __float128 values, conversions to and from 5843 float, double and long double floating point types, as well as 5844 conversions to and from signed or unsigned integer, signed or 5845 unsigned long integer and signed or unsigned quad (TImode, IA64 5846 only) integer types. Additionally, all operations generate the full 5847 set of IEEE exceptions and support the full set of IEEE rounding 5848 modes. 5849 5850 M68K/ColdFire 5851 5852 * GCC now supports instruction scheduling for ColdFire V1, V3 and V4 5853 processors. (Scheduling support for ColdFire V2 processors was 5854 added in GCC 4.3.) 5855 * GCC now supports the -mxgot option to support programs requiring 5856 many GOT entries on ColdFire. 5857 * The m68k-*-linux-gnu target now builds multilibs by default. 5858 5859 MIPS 5860 5861 * MIPS Technologies have extended the original MIPS SVR4 ABI to 5862 include support for procedure linkage tables (PLTs) and copy 5863 relocations. These extensions allow GNU/Linux executables to use a 5864 significantly more efficient code model than the one defined by the 5865 original ABI. 5866 GCC support for this code model is available via a new command-line 5867 option, -mplt. There is also a new configure-time option, 5868 --with-mips-plt, to make -mplt the default. 5869 The new code model requires support from the assembler, the linker, 5870 and the runtime C library. This support is available in binutils 5871 2.19 and GLIBC 2.9. 5872 * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables 5873 and 32-bit GNU/Linux shared libraries. This feature requires GNU 5874 binutils 2.19 or above. 5875 * Support for RMI's XLR processor is now available through the 5876 -march=xlr and -mtune=xlr options. 5877 * 64-bit targets can now perform 128-bit multiplications inline, 5878 instead of relying on a libgcc function. 5879 * Native GNU/Linux toolchains now support -march=native and 5880 -mtune=native, which select the host processor. 5881 * GCC now supports the R10K, R12K, R14K and R16K processors. The 5882 canonical -march= and -mtune= names for these processors are 5883 r10000, r12000, r14000 and r16000 respectively. 5884 * GCC can now work around the side effects of speculative execution 5885 on R10K processors. Please see the documentation of the 5886 -mr10k-cache-barrier option for details. 5887 * Support for the MIPS64 Release 2 instruction set has been added. 5888 The option -march=mips64r2 enables generation of these 5889 instructions. 5890 * GCC now supports Cavium Networks' Octeon processor. This support is 5891 available through the -march=octeon and -mtune=octeon options. 5892 * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The 5893 canonical -march= and -mtune= names for these processors are 5894 loongson2e and loongson2f. 5895 5896 picochip 5897 5898 Picochip is a 16-bit processor. A typical picoChip contains over 250 5899 small cores, each with small amounts of memory. There are three 5900 processor variants (STAN, MEM and CTRL) with different instruction sets 5901 and memory configurations and they can be chosen using the -mae option. 5902 5903 This port is intended to be a "C" only port. 5904 5905 Power Architecture and PowerPC 5906 5907 * GCC now supports the e300c2, e300c3 and e500mc processors. 5908 * GCC now supports Xilinx processors with a single-precision FPU. 5909 * Decimal floating point is now supported for e500 processors. 5910 5911 S/390, zSeries and System z9/z10 5912 5913 * Support for the IBM System z10 EC/BC processor has been added. When 5914 using the -march=z10 option, the compiler will generate code making 5915 use of instructions provided by the General-Instruction-Extension 5916 Facility and the Execute-Extension Facility. 5917 5918 VxWorks 5919 5920 * GCC now supports the thread-local storage mechanism used on 5921 VxWorks. 5922 5923 Xtensa 5924 5925 * GCC now supports thread-local storage (TLS) for Xtensa processor 5926 configurations that include the Thread Pointer option. TLS also 5927 requires support from the assembler and linker; this support is 5928 provided in the GNU binutils beginning with version 2.19. 5929 5930 Documentation improvements 5931 5932 Other significant improvements 5933 5934 GCC 4.4.1 5935 5936 This is the [12]list of problem reports (PRs) from GCC's bug tracking 5937 system that are known to be fixed in the 4.4.1 release. This list might 5938 not be complete (that is, it is possible that some PRs that have been 5939 fixed are not listed here). 5940 5941 GCC 4.4.2 5942 5943 This is the [13]list of problem reports (PRs) from GCC's bug tracking 5944 system that are known to be fixed in the 4.4.2 release. This list might 5945 not be complete (that is, it is possible that some PRs that have been 5946 fixed are not listed here). 5947 5948 GCC 4.4.3 5949 5950 This is the [14]list of problem reports (PRs) from GCC's bug tracking 5951 system that are known to be fixed in the 4.4.3 release. This list might 5952 not be complete (that is, it is possible that some PRs that have been 5953 fixed are not listed here). 5954 5955 GCC 4.4.4 5956 5957 This is the [15]list of problem reports (PRs) from GCC's bug tracking 5958 system that are known to be fixed in the 4.4.4 release. This list might 5959 not be complete (that is, it is possible that some PRs that have been 5960 fixed are not listed here). 5961 5962 GCC 4.4.5 5963 5964 This is the [16]list of problem reports (PRs) from GCC's bug tracking 5965 system that are known to be fixed in the 4.4.5 release. This list might 5966 not be complete (that is, it is possible that some PRs that have been 5967 fixed are not listed here). 5968 5969 GCC 4.4.6 5970 5971 This is the [17]list of problem reports (PRs) from GCC's bug tracking 5972 system that are known to be fixed in the 4.4.6 release. This list might 5973 not be complete (that is, it is possible that some PRs that have been 5974 fixed are not listed here). 5975 5976 GCC 4.4.7 5977 5978 This is the [18]list of problem reports (PRs) from GCC's bug tracking 5979 system that are known to be fixed in the 4.4.7 release. This list might 5980 not be complete (that is, it is possible that some PRs that have been 5981 fixed are not listed here). 5982 5983 5984 For questions related to the use of GCC, please consult these web 5985 pages and the [19]GCC manuals. If that fails, the 5986 [20]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5987 web pages and the development of GCC are welcome on our developer 5988 list at [21]gcc (a] gcc.gnu.org. All of [22]our lists have public 5989 archives. 5990 5991 Copyright (C) [23]Free Software Foundation, Inc. Verbatim copying and 5992 distribution of this entire article is permitted in any medium, 5993 provided this notice is preserved. 5994 5995 These pages are [24]maintained by the GCC team. Last modified 5996 2014-12-06[25]. 5997 5998 References 5999 6000 1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7 6001 2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted 6002 3. http://gcc.gnu.org/gcc-4.4/porting_to.html 6003 4. https://gcc.gnu.org/wiki/Graphite 6004 5. http://openmp.org/wp/openmp-specifications/ 6005 6. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf 6006 7. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html 6007 8. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html 6008 9. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125 6009 10. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221 6010 11. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34 6011 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1 6012 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2 6013 14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3 6014 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4 6015 16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5 6016 17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6 6017 18. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7 6018 19. https://gcc.gnu.org/onlinedocs/ 6019 20. mailto:gcc-help (a] gcc.gnu.org 6020 21. mailto:gcc (a] gcc.gnu.org 6021 22. https://gcc.gnu.org/lists.html 6022 23. http://www.fsf.org/ 6023 24. https://gcc.gnu.org/about.html 6024 25. http://validator.w3.org/check/referer 6025 ====================================================================== 6026 http://gcc.gnu.org/gcc-4.3/index.html 6027 6028 GCC 4.3 Release Series 6029 6030 Jun 27, 2011 6031 6032 The [1]GNU project and the GCC developers are pleased to announce the 6033 release of GCC 4.3.6. 6034 6035 This release is a bug-fix release, containing fixes for regressions in 6036 GCC 4.3.5 relative to previous releases of GCC. 6037 6038 Release History 6039 6040 GCC 4.3.6 6041 Jun 27, 2011 ([2]changes) 6042 6043 GCC 4.3.5 6044 May 22, 2010 ([3]changes) 6045 6046 GCC 4.3.4 6047 August 4, 2009 ([4]changes) 6048 6049 GCC 4.3.3 6050 January 24, 2009 ([5]changes) 6051 6052 GCC 4.3.2 6053 August 27, 2008 ([6]changes) 6054 6055 GCC 4.3.1 6056 June 6, 2008 ([7]changes) 6057 6058 GCC 4.3.0 6059 March 5, 2008 ([8]changes) 6060 6061 References and Acknowledgements 6062 6063 GCC used to stand for the GNU C Compiler, but since the compiler 6064 supports several other languages aside from C, it now stands for the 6065 GNU Compiler Collection. 6066 6067 A list of [9]successful builds is updated as new information becomes 6068 available. 6069 6070 The GCC developers would like to thank the numerous people that have 6071 contributed new features, improvements, bug fixes, and other changes as 6072 well as test results to GCC. This [10]amazing group of volunteers is 6073 what makes GCC successful. 6074 6075 For additional information about GCC please refer to the [11]GCC 6076 project web site or contact the [12]GCC development mailing list. 6077 6078 To obtain GCC please use [13]our mirror sites or [14]our SVN server. 6079 6080 6081 For questions related to the use of GCC, please consult these web 6082 pages and the [15]GCC manuals. If that fails, the 6083 [16]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6084 web pages and the development of GCC are welcome on our developer 6085 list at [17]gcc (a] gcc.gnu.org. All of [18]our lists have public 6086 archives. 6087 6088 Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and 6089 distribution of this entire article is permitted in any medium, 6090 provided this notice is preserved. 6091 6092 These pages are [20]maintained by the GCC team. Last modified 6093 2014-06-28[21]. 6094 6095 References 6096 6097 1. http://www.gnu.org/ 6098 2. http://gcc.gnu.org/gcc-4.3/changes.html 6099 3. http://gcc.gnu.org/gcc-4.3/changes.html 6100 4. http://gcc.gnu.org/gcc-4.3/changes.html 6101 5. http://gcc.gnu.org/gcc-4.3/changes.html 6102 6. http://gcc.gnu.org/gcc-4.3/changes.html 6103 7. http://gcc.gnu.org/gcc-4.3/changes.html 6104 8. http://gcc.gnu.org/gcc-4.3/changes.html 6105 9. http://gcc.gnu.org/gcc-4.3/buildstat.html 6106 10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 6107 11. http://gcc.gnu.org/index.html 6108 12. mailto:gcc (a] gcc.gnu.org 6109 13. http://gcc.gnu.org/mirrors.html 6110 14. http://gcc.gnu.org/svn.html 6111 15. https://gcc.gnu.org/onlinedocs/ 6112 16. mailto:gcc-help (a] gcc.gnu.org 6113 17. mailto:gcc (a] gcc.gnu.org 6114 18. https://gcc.gnu.org/lists.html 6115 19. http://www.fsf.org/ 6116 20. https://gcc.gnu.org/about.html 6117 21. http://validator.w3.org/check/referer 6118 ====================================================================== 6119 http://gcc.gnu.org/gcc-4.3/changes.html 6120 6121 GCC 4.3 Release Series 6122 Changes, New Features, and Fixes 6123 6124 The latest release in the 4.3 release series is [1]GCC 4.3.5. 6125 6126 Caveats 6127 6128 * GCC requires the [2]GMP and [3]MPFR libraries for building all the 6129 various front-end languages it supports. See the [4]prerequisites 6130 page for version requirements. 6131 * ColdFire targets now treat long double as having the same format as 6132 double. In earlier versions of GCC, they used the 68881 long double 6133 format instead. 6134 * The m68k-uclinux target now uses the same calling conventions as 6135 m68k-linux-gnu. You can select the original calling conventions by 6136 configuring for m68k-uclinuxoldabi instead. Note that 6137 m68k-uclinuxoldabi also retains the original 80-bit long double on 6138 ColdFire targets. 6139 * The -fforce-mem option has been removed because it has had no 6140 effect in the last few GCC releases. 6141 * The i386 -msvr3-shlib option has been removed since it is no longer 6142 used. 6143 * Fastcall for i386 has been changed not to pass aggregate arguments 6144 in registers, following Microsoft compilers. 6145 * Support for the AOF assembler has been removed from the ARM back 6146 end; this affects only the targets arm-semi-aof and armel-semi-aof, 6147 which are no longer recognized. We removed these targets without a 6148 deprecation period because we discovered that they have been 6149 unusable since GCC 4.0.0. 6150 * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*) 6151 has been removed. This support had been deprecated since GCC 4.0.0. 6152 * Support for a number of older systems and recently unmaintained or 6153 untested target ports of GCC has been declared obsolete in GCC 4.3. 6154 Unless there is activity to revive them, the next release of GCC 6155 will have their sources permanently removed. 6156 All GCC ports for the following processor architectures have been 6157 declared obsolete: 6158 + Morpho MT (mt-*) 6159 The following aliases for processor architectures have been 6160 declared obsolete. Users should use the indicated generic target 6161 names instead, with compile-time options such as -mcpu or 6162 configure-time options such as --with-cpu to control the 6163 configuration more precisely. 6164 + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-* 6165 instead). 6166 + parisc*-*-* (use hppa*-*-* instead). 6167 + m680[012]0-*-* (use m68k-*-* instead). 6168 All GCC ports for the following operating systems have been 6169 declared obsolete: 6170 + BeOS (*-*-beos*) 6171 + kaOS (*-*-kaos*) 6172 + GNU/Linux using the a.out object format (*-*-linux*aout*) 6173 + GNU/Linux using version 1 of the GNU C Library 6174 (*-*-linux*libc1*) 6175 + Solaris versions before Solaris 7 (*-*-solaris2.[0-6], 6176 *-*-solaris2.[0-6].*) 6177 + Miscellaneous System V (*-*-sysv*) 6178 + WindISS (*-*-windiss*) 6179 Also, those for some individual systems on particular architectures 6180 have been obsoleted: 6181 + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*) 6182 + CRIS with a.out object format (cris-*-aout) 6183 + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*) 6184 + OSF/1 on PA-RISC (hppa1.1-*-osf*) 6185 + PRO on PA-RISC (hppa1.1-*-pro*) 6186 + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*, 6187 i[34567]86-sequent-sysv4*) 6188 + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*) 6189 + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host 6190 was previously [5]removed in 2001, leaving only the support 6191 for UWIN as a target now being deprecated) 6192 + ChorusOS on PowerPC (powerpc-*-chorusos*) 6193 + All VAX configurations apart from NetBSD and OpenBSD 6194 (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*) 6195 * The [6]-Wconversion option has been modified. Its purpose now is to 6196 warn for implicit conversions that may alter a value. This new 6197 behavior is available for both C and C++. Warnings about 6198 conversions between signed and unsigned integers can be disabled by 6199 using -Wno-sign-conversion. In C++, they are disabled by default 6200 unless -Wsign-conversion is explicitly requested. The old behavior 6201 of -Wconversion, that is, warn for prototypes causing a type 6202 conversion that is different from what would happen to the same 6203 argument in the absence of a prototype, has been moved to a new 6204 option -Wtraditional-conversion, which is only available for C. 6205 * The -m386, -m486, -mpentium and -mpentiumpro tuning options have 6206 been removed because they were deprecated for more than 3 GCC major 6207 releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or 6208 -mtune=pentiumpro as a replacement. 6209 * The -funsafe-math-optimizations option now automatically turns on 6210 -fno-trapping-math in addition to -fno-signed-zeros, as it enables 6211 reassociation and thus may introduce or remove traps. 6212 * The -ftree-vectorize option is now on by default under -O3. In 6213 order to generate code for a SIMD extension, it has to be enabled 6214 as well: use -maltivec for PowerPC platforms and -msse/-msse2 for 6215 i?86 and x86_64. 6216 * More information on porting to GCC 4.3 from previous versions of 6217 GCC can be found in the [7]porting guide for this release. 6218 6219 General Optimizer Improvements 6220 6221 * The GCC middle-end has been integrated with the [8]MPFR library. 6222 This allows GCC to evaluate and replace at compile-time calls to 6223 built-in math functions having constant arguments with their 6224 mathematically equivalent results. In making use of [9]MPFR, GCC 6225 can generate correct results regardless of the math library 6226 implementation or floating point precision of the host platform. 6227 This also allows GCC to generate identical results regardless of 6228 whether one compiles in native or cross-compile configurations to a 6229 particular target. The following built-in functions take advantage 6230 of this new capability: acos, acosh, asin, asinh, atan2, atan, 6231 atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1, 6232 fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10, 6233 log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh, 6234 tan, tanh, tgamma, y0, y1 and yn. The float and long double 6235 variants of these functions (e.g. sinf and sinl) are also handled. 6236 The sqrt and cabs functions with constant arguments were already 6237 optimized in prior GCC releases. Now they also use [10]MPFR. 6238 * A new forward propagation pass on RTL was added. The new pass 6239 replaces several slower transformations, resulting in compile-time 6240 improvements as well as better code generation in some cases. 6241 * A new command-line switch -frecord-gcc-switches has been added to 6242 GCC, although it is only enabled for some targets. The switch 6243 causes the command line that was used to invoke the compiler to be 6244 recorded into the object file that is being created. The exact 6245 format of this recording is target and binary file format 6246 dependent, but it usually takes the form of a note section 6247 containing ASCII text. The switch is related to the -fverbose-asm 6248 switch, but that one only records the information in the assembler 6249 output file as comments, so the information never reaches the 6250 object file. 6251 * The inliner heuristic is now aware of stack frame consumption. New 6252 command-line parameters --param large-stack-frame and --param 6253 large-stack-frame-growth can be used to limit stack frame size 6254 growth caused by inlining. 6255 * During feedback directed optimizations, the expected block size the 6256 memcpy, memset and bzero functions operate on is discovered and for 6257 cases of commonly used small sizes, specialized inline code is 6258 generated. 6259 * __builtin_expect no longer requires its argument to be a compile 6260 time constant. 6261 * Interprocedural optimization was reorganized to work on functions 6262 in SSA form. This enables more precise and cheaper dataflow 6263 analysis and makes writing interprocedural optimizations easier. 6264 The following improvements have been implemented on top of this 6265 framework: 6266 + Pre-inline optimization: Selected local optimization passes 6267 are run before the inliner (and other interprocedural passes) 6268 are executed. This significantly improves the accuracy of code 6269 growth estimates used by the inliner and reduces the overall 6270 memory footprint for large compilation units. 6271 + Early inlining (a simple bottom-up inliner pass inlining only 6272 functions whose body is smaller than the expected call 6273 overhead) is now executed with the early optimization passes, 6274 thus inlining already optimized function bodies into an 6275 unoptimized function that is subsequently optimized by early 6276 optimizers. This enables the compiler to quickly eliminate 6277 abstraction penalty in C++ programs. 6278 + Interprocedural constant propagation now operate on SSA form 6279 increasing accuracy of the analysis. 6280 * A new internal representation for GIMPLE statements has been 6281 contributed, resulting in compile-time memory savings. 6282 * The vectorizer was enhanced to support vectorization of outer 6283 loops, intra-iteration parallelism (loop-aware SLP), vectorization 6284 of strided accesses and loops with multiple data-types. Run-time 6285 dependency testing using loop versioning was added. The cost model, 6286 turned on by -fvect-cost-model, was developed. 6287 6288 New Languages and Language specific improvements 6289 6290 * We have added new command-line options 6291 -finstrument-functions-exclude-function-list and 6292 -finstrument-functions-exclude-file-list. They provide more control 6293 over which functions are annotated by the -finstrument-functions 6294 option. 6295 6296 C family 6297 6298 * Implicit conversions between generic vector types are now only 6299 permitted when the two vectors in question have the same number of 6300 elements and compatible element types. (Note that the restriction 6301 involves compatible element types, not implicitly-convertible 6302 element types: thus, a vector type with element type int may not be 6303 implicitly converted to a vector type with element type unsigned 6304 int.) This restriction, which is in line with specifications for 6305 SIMD architectures such as AltiVec, may be relaxed using the flag 6306 -flax-vector-conversions. This flag is intended only as a 6307 compatibility measure and should not be used for new code. 6308 * -Warray-bounds has been added and is now enabled by default for 6309 -Wall . It produces warnings for array subscripts that can be 6310 determined at compile time to be always out of bounds. 6311 -Wno-array-bounds will disable the warning. 6312 * The constructor and destructor function attributes now accept 6313 optional priority arguments which control the order in which the 6314 constructor and destructor functions are run. 6315 * New [11]command-line options -Wtype-limits, 6316 -Wold-style-declaration, -Wmissing-parameter-type, -Wempty-body, 6317 -Wclobbered and -Wignored-qualifiers have been added for finer 6318 control of the diverse warnings enabled by -Wextra. 6319 * A new function attribute alloc_size has been added to mark up 6320 malloc style functions. For constant sized allocations this can be 6321 used to find out the size of the returned pointer using the 6322 __builtin_object_size() function for buffer overflow checking and 6323 similar. This supplements the already built-in malloc and calloc 6324 constant size handling. 6325 * Integer constants written in binary are now supported as a GCC 6326 extension. They consist of a prefix 0b or 0B, followed by a 6327 sequence of 0 and 1 digits. 6328 * A new predefined macro __COUNTER__ has been added. It expands to 6329 sequential integral values starting from 0. In conjunction with the 6330 ## operator, this provides a convenient means to generate unique 6331 identifiers. 6332 * A new command-line option -fdirectives-only has been added. It 6333 enables a special preprocessing mode which improves the performance 6334 of applications like distcc and ccache. 6335 * Fixed-point data types and operators have been added. They are 6336 based on Chapter 4 of the Embedded-C specification (n1169.pdf). 6337 Currently, only MIPS targets are supported. 6338 * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732, 6339 N1241, is now supported as a GCC extension to C for targets 6340 i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu, 6341 and x86_64-*-linux-gnu. The feature introduces new data types 6342 _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF, 6343 DD, and DL. 6344 6345 C++ 6346 6347 * [12]Experimental support for the upcoming ISO C++ standard, C++0x. 6348 * -Wc++0x-compat has been added and is now enabled by default for 6349 -Wall. It produces warnings for constructs whose meaning differs 6350 between ISO C++ 1998 and C++0x. 6351 * The -Wparentheses option now works for C++ as it does for C. It 6352 warns if parentheses are omitted when operators with confusing 6353 precedence are nested. It also warns about ambiguous else 6354 statements. Since -Wparentheses is enabled by -Wall, this may cause 6355 additional warnings with existing C++ code which uses -Wall. These 6356 new warnings may be disabled by using -Wall -Wno-parentheses. 6357 * The -Wmissing-declarations now works for C++ as it does for C. 6358 * The -fvisibility-ms-compat flag was added, to make it easier to 6359 port larger projects using shared libraries from Microsoft's Visual 6360 Studio to ELF and Mach-O systems. 6361 * C++ attribute handling has been overhauled for template arguments 6362 (ie dependent types). In particular, __attribute__((aligned(T))); 6363 works for C++ types. 6364 6365 Runtime Library (libstdc++) 6366 6367 * [13]Experimental support for the upcoming ISO C++ standard, C++0x. 6368 * Support for TR1 mathematical special functions and regular 6369 expressions. ([14]Implementation status of TR1) 6370 * Default what implementations give more elaborate exception strings 6371 for bad_cast, bad_typeid, bad_exception, and bad_alloc. 6372 * Header dependencies have been streamlined, reducing unnecessary 6373 includes and pre-processed bloat. 6374 * Variadic template implementations of items in <tuple> and 6375 <functional>. 6376 * An experimental [15]parallel mode has been added. This is a 6377 parallel implementation of many C++ Standard library algorithms, 6378 like std::accumulate, std::for_each, std::transform, or std::sort, 6379 to give but four examples. These algorithms can be substituted for 6380 the normal (sequential) libstdc++ algorithms on a piecemeal basis, 6381 or all existing algorithms can be transformed via the 6382 -D_GLIBCXX_PARALLEL macro. 6383 * Debug mode versions of classes in <unordered_set> and 6384 <unordered_map>. 6385 * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are 6386 now <backward/hash_set> and <backward/hash_map>. This code: 6387 #include <ext/hash_set> 6388 __gnu_cxx::hash_set<int> s; 6389 6390 Can be transformed (in order of preference) to: 6391 #include <tr1/unordered_set> 6392 std::tr1::unordered_set<int> s; 6393 6394 or 6395 #include <backward/hash_set> 6396 __gnu_cxx::hash_set<int> s; 6397 6398 Similar transformations apply to __gnu_cxx::hash_map, 6399 __gnu_cxx::hash_multimap, __gnu_cxx::hash_set, 6400 __gnu_cxx::hash_multiset. 6401 6402 Fortran 6403 6404 * Due to the fact that the GMP and MPFR libraries are required for 6405 all languages, Fortran is no longer special in this regard and is 6406 available by default. 6407 * The [16]-fexternal-blas option has been added, which generates 6408 calls to BLAS routines for intrinsic matrix operations such as 6409 matmul rather than using the built-in algorithms. 6410 * Support to give a backtrace (compiler flag -fbacktrace or 6411 environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems 6412 only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a 6413 run-time error occured. 6414 * GNU Fortran now defines __GFORTRAN__ when it runs the C 6415 preprocessor (CPP). 6416 * The [17]-finit-local-zero, -finit-real, -finit-integer, 6417 -finit-character, and -finit-logical options have been added, which 6418 can be used to initialize local variables. 6419 * The intrinsic procedures [18]GAMMA and [19]LGAMMA have been added, 6420 which calculate the Gamma function and its logarithm. Use EXTERNAL 6421 gamma if you want to use your own gamma function. 6422 * GNU Fortran now regards the backslash character as literal (as 6423 required by the Fortran 2003 standard); using [20]-fbackslash GNU 6424 Fortran interprets backslashes as C-style escape characters. 6425 * The [21]interpretation of binary, octal and hexadecimal (BOZ) 6426 literal constants has been changed. Before they were always 6427 interpreted as integer; now they are bit-wise transferred as 6428 argument of INT, REAL, DBLE and CMPLX as required by the Fortran 6429 2003 standard, and for real and complex variables in DATA 6430 statements or when directly assigned to real and complex variables. 6431 Everywhere else and especially in expressions they are still 6432 regarded as integer constants. 6433 * Fortran 2003 support has been extended: 6434 + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE 6435 + Pointer intent 6436 + Intrinsic module ISO_ENV_FORTRAN 6437 + Interoperability with C (ISO C Bindings) 6438 + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER 6439 attribute) 6440 + Fortran 2003 BOZ 6441 6442 Java (GCJ) 6443 6444 * GCJ now uses the Eclipse Java compiler for its Java parsing needs. 6445 This enables the use of all 1.5 language features, and fixes most 6446 existing front end bugs. 6447 * libgcj now supports all 1.5 language features which require runtime 6448 support: foreach, enum, annotations, generics, and auto-boxing. 6449 * We've made many changes to the tools shipped with gcj. 6450 + The old jv-scan tool has been removed. This tool never really 6451 worked properly. There is no replacement. 6452 + gcjh has been rewritten. Some of its more obscure options no 6453 longer work, but are still recognized in an attempt at 6454 compatibility. gjavah is a new program with similar 6455 functionality but different command-line options. 6456 + grmic and grmiregistry have been rewritten. grmid has been 6457 added. 6458 + gjar replaces the old fastjar. 6459 + gjarsigner (used for signing jars), gkeytool (used for key 6460 management), gorbd (for CORBA), gserialver (computes 6461 serialization UIDs), and gtnameserv (also for CORBA) are now 6462 installed. 6463 * The ability to dump the contents of the java run time heap to a 6464 file for off-line analysis has been added. The heap dumps may be 6465 analyzed with the new gc-analyze tool. They may be generated on 6466 out-of-memory conditions or on demand and are controlled by the new 6467 run time class gnu.gcj.util.GCInfo. 6468 * java.util.TimeZone can now read files from /usr/share/zoneinfo to 6469 provide correct, updated, timezone information. This means that 6470 packagers no longer have to update libgcj when a time zone change 6471 is published. 6472 6473 New Targets and Target Specific Improvements 6474 6475 IA-32/x86-64 6476 6477 * Tuning for Intel Core 2 processors is available via -mtune=core2 6478 and -march=core2. 6479 * Tuning for AMD Geode processors is available via -mtune=geode and 6480 -march=geode. 6481 * Code generation of block move (memcpy) and block set (memset) was 6482 rewritten. GCC can now pick the best algorithm (loop, unrolled 6483 loop, instruction with rep prefix or a library call) based on the 6484 size of the block being copied and the CPU being optimized for. A 6485 new option -minline-stringops-dynamically has been added. With this 6486 option string operations of unknown size are expanded such that 6487 small blocks are copied by in-line code, while for large blocks a 6488 library call is used. This results in faster code than 6489 -minline-all-stringops when the library implementation is capable 6490 of using cache hierarchy hints. The heuristic choosing the 6491 particular algorithm can be overwritten via -mstringop-strategy. 6492 Newly also memset of values different from 0 is inlined. 6493 * GCC no longer places the cld instruction before string operations. 6494 Both i386 and x86-64 ABI documents mandate the direction flag to be 6495 clear at the entry of a function. It is now invalid to set the flag 6496 in asm statement without reseting it afterward. 6497 * Support for SSSE3 built-in functions and code generation are 6498 available via -mssse3. 6499 * Support for SSE4.1 built-in functions and code generation are 6500 available via -msse4.1. 6501 * Support for SSE4.2 built-in functions and code generation are 6502 available via -msse4.2. 6503 * Both SSE4.1 and SSE4.2 support can be enabled via -msse4. 6504 * A new set of options -mpc32, -mpc64 and -mpc80 have been added to 6505 allow explicit control of x87 floating point precision. 6506 * Support for __float128 (TFmode) IEEE quad type and corresponding 6507 TCmode IEEE complex quad type is available via the soft-fp library 6508 on x86_64 targets. This includes basic arithmetic operations 6509 (addition, subtraction, negation, multiplication and division) on 6510 __float128 real and TCmode complex values, the full set of IEEE 6511 comparisons between __float128 values, conversions to and from 6512 float, double and long double floating point types, as well as 6513 conversions to and from signed or unsigned integer, signed or 6514 unsigned long integer and signed or unsigned quad (TImode) integer 6515 types. Additionally, all operations generate the full set of IEEE 6516 exceptions and support the full set of IEEE rounding modes. 6517 * GCC can now utilize the ACML library for vectorizing calls to a set 6518 of C99 functions on x86_64 if -mveclibabi=acml is specified and you 6519 link to an ACML ABI compatible library. 6520 6521 ARM 6522 6523 * Compiler and Library support for Thumb-2 and the ARMv7 architecture 6524 has been added. 6525 6526 CRIS 6527 6528 New features 6529 6530 * Compiler and Library support for the CRIS v32 architecture, as 6531 found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been 6532 added. 6533 6534 Configuration changes 6535 6536 * The cris-*-elf target now includes support for CRIS v32, including 6537 libraries, through the -march=v32 option. 6538 * A new crisv32-*-elf target defaults to generate code for CRIS v32. 6539 * A new crisv32-*-linux* target defaults to generate code for CRIS 6540 v32. 6541 * The cris-*-aout target has been obsoleted. 6542 6543 Improved support for built-in functions 6544 6545 * GCC can now use the lz and swapwbr instructions to implement the 6546 __builtin_clz, __builtin_ctz and __builtin_ffs family of functions. 6547 * __builtin_bswap32 is now implemented using the swapwb instruction, 6548 when available. 6549 6550 m68k and ColdFire 6551 6552 New features 6553 6554 * Support for several new ColdFire processors has been added. You can 6555 generate code for them using the new -mcpu option. 6556 * All targets now support ColdFire processors. 6557 * m68k-uclinux targets have improved support for C++ constructors and 6558 destructors, and for shared libraries. 6559 * It is now possible to set breakpoints on the first or last line of 6560 a function, even if there are no statements on that line. 6561 6562 Optimizations 6563 6564 * Support for sibling calls has been added. 6565 * More use is now made of the ColdFire mov3q instruction. 6566 * __builtin_clz is now implemented using the ff1 ColdFire 6567 instruction, when available. 6568 * GCC now honors the -m68010 option. 68010 code now uses clr rather 6569 than move to zero volatile memory. 6570 * 68020 targets and above can now use symbol(index.size*scale) 6571 addresses for indexed array accesses. Earlier compilers would 6572 always load the symbol into a base register first. 6573 6574 Configuration changes 6575 6576 * All m68k and ColdFire targets now allow the default processor to be 6577 set at configure time using --with-cpu. 6578 * A --with-arch configuration option has been added. This option 6579 allows you to restrict a target to ColdFire or non-ColdFire 6580 processors. 6581 6582 Preprocessor macros 6583 6584 * An __mcfv*__ macro is now defined for all ColdFire targets. 6585 (Earlier versions of GCC only defined __mcfv4e__.) 6586 * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added. 6587 * All targets now define __mc68010 and __mc68010__ when generating 6588 68010 code. 6589 6590 Command-line changes 6591 6592 * New command-line options -march, -mcpu, -mtune and -mhard-float 6593 have been added. These options apply to both m68k and ColdFire 6594 targets. 6595 * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative 6596 versions of -mshort, etc. 6597 * -fforce-addr has been removed. It is now ignored by the compiler. 6598 6599 Other improvements 6600 6601 * ColdFire targets now try to maintain a 4-byte-aligned stack where 6602 possible. 6603 * m68k-uclinux targets now try to avoid situations that lead to the 6604 load-time error: BINFMT_FLAT: reloc outside program. 6605 6606 MIPS 6607 6608 Changes to existing configurations 6609 6610 * libffi and libjava now support all three GNU/Linux ABIs: o32, n32 6611 and n64. Every GNU/Linux configuration now builds these libraries 6612 by default. 6613 * GNU/Linux configurations now generate -mno-shared code unless 6614 overridden by -fpic, -fPIC, -fpie or -fPIE. 6615 * mipsisa32*-linux-gnu configurations now generate hard-float code by 6616 default, just like other mipsisa32* and mips*-linux-gnu 6617 configurations. You can build a soft-float version of any 6618 mips*-linux-gnu configuration by passing --with-float=soft to 6619 configure. 6620 * mips-wrs-vxworks now supports run-time processes (RTPs). 6621 6622 Changes to existing command-line options 6623 6624 * The -march and -mtune options no longer accept 24k as a processor 6625 name. Please use 24kc, 24kf2_1 or 24kf1_1 instead. 6626 * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and 6627 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The 6628 options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for 6629 24kx, 24kex and 34kx. 6630 6631 New configurations 6632 6633 GCC now supports the following configurations: 6634 * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by 6635 default. Earlier releases also recognized this configuration, but 6636 they treated it in the same way as mipsisa32*-linux-gnu*. Note that 6637 you can customize any mips*-linux-gnu* configuration to a 6638 particular ISA or processor by passing an appropriate --with-arch 6639 option to configure. 6640 * mipsisa*-sde-elf*, which provides compatibility with MIPS 6641 Technologies' SDE toolchains. The configuration uses the SDE 6642 libraries by default, but you can use it like other newlib-based 6643 ELF configurations by passing --with-newlib to configure. It is the 6644 only configuration besides mips64vr*-elf* to build MIPS16 as well 6645 as non-MIPS16 libraries. 6646 * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf* 6647 configuration, but uses the o32 and o64 ABIs instead of the 32-bit 6648 and 64-bit forms of the EABI. 6649 6650 New processors and application-specific extensions 6651 6652 * Support for the SmartMIPS ASE is available through the new 6653 -msmartmips option. 6654 * Support for revision 2 of the DSP ASE is available through the new 6655 -mdspr2 option. A new preprocessor macro called __mips_dsp_rev 6656 indicates the revision of the ASE in use. 6657 * Support for the 4KS and 74K families of processors is available 6658 through the -march and -mtune options. 6659 6660 Improved support for built-in functions 6661 6662 * GCC can now use load-linked, store-conditional and sync 6663 instructions to implement atomic built-in functions such as 6664 __sync_fetch_and_add. The memory reference must be 4 bytes wide for 6665 32-bit targets and either 4 or 8 bytes wide for 64-bit targets. 6666 * GCC can now use the clz and dclz instructions to implement the 6667 __builtin_ctz and __builtin_ffs families of functions. 6668 * There is a new __builtin___clear_cache function for flushing the 6669 instruction cache. GCC expands this function inline on MIPS32 6670 revision 2 targets, otherwise it calls the function specified by 6671 -mcache-flush-func. 6672 6673 MIPS16 improvements 6674 6675 * GCC can now compile objects that contain a mixture of MIPS16 and 6676 non-MIPS16 code. There are two new attributes, mips16 and nomips16, 6677 for specifying which mode a function should use. 6678 * A new option called -minterlink-mips16 makes non-MIPS16 code 6679 link-compatible with MIPS16 code. 6680 * After many bug fixes, the long-standing MIPS16 -mhard-float support 6681 should now work fairly reliably. 6682 * GCC can now use the MIPS16e save and restore instructions. 6683 * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled 6684 with -G0 -fsection-anchors is often smaller than code compiled with 6685 -G8. However, please note that you must usually compile all objects 6686 in your application with the same -G option; see the documentation 6687 of -G for details. 6688 * A new option called-mcode-readable specifies which instructions are 6689 allowed to load from the code segment. -mcode-readable=yes is the 6690 default and says that any instruction may load from the code 6691 segment. The other alternatives are -mcode-readable=pcrel, which 6692 says that only PC-relative MIPS16 instructions may load from the 6693 code segment, and -mcode-readable=no, which says that no 6694 instruction may do so. Please see the documentation for more 6695 details, including example uses. 6696 6697 Small-data improvements 6698 6699 There are three new options for controlling small data: 6700 * -mno-extern-sdata, which disables small-data accesses for 6701 externally-defined variables. Code compiled with -Gn 6702 -mno-extern-sdata will be link-compatible with any -G setting 6703 between -G0 and -Gn inclusive. 6704 * -mno-local-sdata, which disables the use of small-data sections for 6705 data that is not externally visible. This option can be a useful 6706 way of reducing small-data usage in less performance-critical parts 6707 of an application. 6708 * -mno-gpopt, which disables the use of the $gp register while still 6709 honoring the -G limit when placing externally-visible data. This 6710 option implies -mno-extern-sdata and -mno-local-sdata and it can be 6711 useful in situations where $gp does not necessarily hold the 6712 expected value. 6713 6714 Miscellaneous improvements 6715 6716 * There is a new option called -mbranch-cost for tweaking the 6717 perceived cost of branches. 6718 * If GCC is configured to use a version of GAS that supports the 6719 .gnu_attribute directive, it will use that directive to record 6720 certain properties of the output code. .gnu_attribute is new to GAS 6721 2.18. 6722 * There are two new function attributes, near and far, for overriding 6723 the command-line setting of -mlong-calls on a function-by-function 6724 basis. 6725 * -mfp64, which previously required a 64-bit target, now works with 6726 MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and 6727 mipsisa*-sde-elf* configurations provide suitable library support. 6728 * GCC now recognizes the -mdmx and -mmt options and passes them down 6729 to the assembler. It does nothing else with the options at present. 6730 6731 SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture 6732 (BEA) 6733 6734 * Support has been added for this new architecture. 6735 6736 RS6000 (POWER/PowerPC) 6737 6738 * Support for the PowerPC 750CL paired-single instructions has been 6739 added with a new powerpc-*-linux*paired* target configuration. It 6740 is enabled by an associated -mpaired option and can be accessed 6741 using new built-in functions. 6742 * Support for auto-detecting architecture and system configuration to 6743 auto-select processor optimization tuning. 6744 * Support for VMX on AIX 5.3 has been added. 6745 * Support for AIX Version 6.1 has been added. 6746 6747 S/390, zSeries and System z9 6748 6749 * Support for the IBM System z9 EC/BC processor (z9 GA3) has been 6750 added. When using the -march=z9-ec option, the compiler will 6751 generate code making use of instructions provided by the decimal 6752 floating point facility and the floating point conversion facility 6753 (pfpo). Besides the instructions used to implement decimal floating 6754 point operations these facilities also contain instructions to move 6755 between general purpose and floating point registers and to modify 6756 and copy the sign-bit of floating point values. 6757 * When the -march=z9-ec option is used the new 6758 -mhard-dfp/-mno-hard-dfp options can be used to specify whether the 6759 decimal floating point hardware instructions will be used or not. 6760 If none of them is given the hardware support is enabled by 6761 default. 6762 * The -mstack-guard option can now be omitted when using stack 6763 checking via -mstack-size in order to let GCC choose a sensible 6764 stack guard value according to the frame size of each function. 6765 * Various changes to improve performance of generated code have been 6766 implemented, including: 6767 + The condition code set by an add logical with carry 6768 instruction is now available for overflow checks like: a + b + 6769 carry < b. 6770 + The test data class instruction is now used to implement 6771 sign-bit and infinity checks of binary and decimal floating 6772 point numbers. 6773 6774 SPARC 6775 6776 * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been 6777 added. 6778 6779 Xtensa 6780 6781 * Stack unwinding for exception handling now uses by default a 6782 specialized version of DWARF unwinding. This is not 6783 binary-compatible with the setjmp/longjmp (sjlj) unwinding used for 6784 Xtensa with previous versions of GCC. 6785 * For Xtensa processors that include the Conditional Store option, 6786 the built-in functions for atomic memory access are now implemented 6787 using S32C1I instructions. 6788 * If the Xtensa NSA option is available, GCC will use it to implement 6789 the __builtin_ctz and __builtin_clz functions. 6790 6791 Documentation improvements 6792 6793 * Existing libstdc++ documentation has been edited and restructured 6794 into a single DocBook XML manual. The results can be viewed online 6795 [22]here. 6796 6797 Other significant improvements 6798 6799 * The compiler's --help command-line option has been extended so that 6800 it now takes an optional set of arguments. These arguments restrict 6801 the information displayed to specific classes of command-line 6802 options, and possibly only a subset of those options. It is also 6803 now possible to replace the descriptive text associated with each 6804 displayed option with an indication of its current value, or for 6805 binary options, whether it has been enabled or disabled. 6806 Here are some examples. The following will display all the options 6807 controlling warning messages: 6808 --help=warnings 6809 6810 Whereas this will display all the undocumented, target specific 6811 options: 6812 --help=target,undocumented 6813 6814 This sequence of commands will display the binary optimizations 6815 that are enabled by -O3: 6816 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 6817 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 6818 diff /tmp/O2-opts /tmp/O3-opts | grep enabled 6819 6820 * The configure options --with-pkgversion and --with-bugurl have been 6821 added. These allow distributors of GCC to include a 6822 distributor-specific string in manuals and --version output and to 6823 specify the URL for reporting bugs in their versions of GCC. 6824 6825 GCC 4.3.1 6826 6827 This is the [23]list of problem reports (PRs) from GCC's bug tracking 6828 system that are known to be fixed in the 4.3.1 release. This list might 6829 not be complete (that is, it is possible that some PRs that have been 6830 fixed are not listed here). 6831 6832 Target Specific Changes 6833 6834 IA-32/x86-64 6835 6836 ABI changes 6837 6838 * Starting with GCC 4.3.1, decimal floating point variables are 6839 aligned to their natural boundaries when they are passed on the 6840 stack for i386. 6841 6842 Command-line changes 6843 6844 * Starting with GCC 4.3.1, the -mcld option has been added to 6845 automatically generate a cld instruction in the prologue of 6846 functions that use string instructions. This option is used for 6847 backward compatibility on some operating systems and can be enabled 6848 by default for 32-bit x86 targets by configuring GCC with the 6849 --enable-cld configure option. 6850 6851 GCC 4.3.2 6852 6853 This is the [24]list of problem reports (PRs) from GCC's bug tracking 6854 system that are known to be fixed in the 4.3.2 release. This list might 6855 not be complete (that is, it is possible that some PRs that have been 6856 fixed are not listed here). 6857 6858 GCC 4.3.3 6859 6860 This is the [25]list of problem reports (PRs) from GCC's bug tracking 6861 system that are known to be fixed in the 4.3.3 release. This list might 6862 not be complete (that is, it is possible that some PRs that have been 6863 fixed are not listed here). 6864 6865 GCC 4.3.4 6866 6867 This is the [26]list of problem reports (PRs) from GCC's bug tracking 6868 system that are known to be fixed in the 4.3.4 release. This list might 6869 not be complete (that is, it is possible that some PRs that have been 6870 fixed are not listed here). 6871 6872 GCC 4.3.5 6873 6874 This is the [27]list of problem reports (PRs) from GCC's bug tracking 6875 system that are known to be fixed in the 4.3.5 release. This list might 6876 not be complete (that is, it is possible that some PRs that have been 6877 fixed are not listed here). 6878 6879 GCC 4.3.6 6880 6881 This is the [28]list of problem reports (PRs) from GCC's bug tracking 6882 system that are known to be fixed in the 4.3.6 release. This list might 6883 not be complete (that is, it is possible that some PRs that have been 6884 fixed are not listed here). 6885 6886 6887 For questions related to the use of GCC, please consult these web 6888 pages and the [29]GCC manuals. If that fails, the 6889 [30]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6890 web pages and the development of GCC are welcome on our developer 6891 list at [31]gcc (a] gcc.gnu.org. All of [32]our lists have public 6892 archives. 6893 6894 Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and 6895 distribution of this entire article is permitted in any medium, 6896 provided this notice is preserved. 6897 6898 These pages are [34]maintained by the GCC team. Last modified 6899 2015-08-27[35]. 6900 6901 References 6902 6903 1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5 6904 2. https://gmplib.org/ 6905 3. http://www.mpfr.org/ 6906 4. https://gcc.gnu.org/install/prerequisites.html 6907 5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html 6908 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 6909 7. http://gcc.gnu.org/gcc-4.3/porting_to.html 6910 8. http://www.mpfr.org/ 6911 9. http://www.mpfr.org/ 6912 10. http://www.mpfr.org/ 6913 11. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 6914 12. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 6915 13. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 6916 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/bk01pt01ch01.html#m anual.intro.status.standard.tr1 6917 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html 6918 16. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options 6919 17. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167 6920 18. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html 6921 19. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html 6922 20. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html 6923 21. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html 6924 22. https://gcc.gnu.org/onlinedocs/libstdc++/ 6925 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1 6926 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2 6927 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3 6928 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4 6929 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5 6930 28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6 6931 29. https://gcc.gnu.org/onlinedocs/ 6932 30. mailto:gcc-help (a] gcc.gnu.org 6933 31. mailto:gcc (a] gcc.gnu.org 6934 32. https://gcc.gnu.org/lists.html 6935 33. http://www.fsf.org/ 6936 34. https://gcc.gnu.org/about.html 6937 35. http://validator.w3.org/check/referer 6938 ====================================================================== 6939 http://gcc.gnu.org/gcc-4.2/index.html 6940 6941 GCC 4.2 Release Series 6942 6943 May 19, 2008 6944 6945 The [1]GNU project and the GCC developers are pleased to announce the 6946 release of GCC 4.2.4. 6947 6948 This release is a bug-fix release, containing fixes for regressions in 6949 GCC 4.2.3 relative to previous releases of GCC. 6950 6951 Release History 6952 6953 GCC 4.2.4 6954 May 19, 2008 ([2]changes) 6955 6956 GCC 4.2.3 6957 February 1, 2008 ([3]changes) 6958 6959 GCC 4.2.2 6960 October 7, 2007 ([4]changes) 6961 6962 GCC 4.2.1 6963 July 18, 2007 ([5]changes) 6964 6965 GCC 4.2.0 6966 May 13, 2007 ([6]changes) 6967 6968 References and Acknowledgements 6969 6970 GCC used to stand for the GNU C Compiler, but since the compiler 6971 supports several other languages aside from C, it now stands for the 6972 GNU Compiler Collection. 6973 6974 A list of [7]successful builds is updated as new information becomes 6975 available. 6976 6977 The GCC developers would like to thank the numerous people that have 6978 contributed new features, improvements, bug fixes, and other changes as 6979 well as test results to GCC. This [8]amazing group of volunteers is 6980 what makes GCC successful. 6981 6982 For additional information about GCC please refer to the [9]GCC project 6983 web site or contact the [10]GCC development mailing list. 6984 6985 To obtain GCC please use [11]our mirror sites or [12]our SVN server. 6986 6987 6988 For questions related to the use of GCC, please consult these web 6989 pages and the [13]GCC manuals. If that fails, the 6990 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6991 web pages and the development of GCC are welcome on our developer 6992 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 6993 archives. 6994 6995 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 6996 distribution of this entire article is permitted in any medium, 6997 provided this notice is preserved. 6998 6999 These pages are [18]maintained by the GCC team. Last modified 7000 2014-06-28[19]. 7001 7002 References 7003 7004 1. http://www.gnu.org/ 7005 2. http://gcc.gnu.org/gcc-4.2/changes.html 7006 3. http://gcc.gnu.org/gcc-4.2/changes.html 7007 4. http://gcc.gnu.org/gcc-4.2/changes.html 7008 5. http://gcc.gnu.org/gcc-4.2/changes.html 7009 6. http://gcc.gnu.org/gcc-4.2/changes.html 7010 7. http://gcc.gnu.org/gcc-4.2/buildstat.html 7011 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7012 9. http://gcc.gnu.org/index.html 7013 10. mailto:gcc (a] gcc.gnu.org 7014 11. http://gcc.gnu.org/mirrors.html 7015 12. http://gcc.gnu.org/svn.html 7016 13. https://gcc.gnu.org/onlinedocs/ 7017 14. mailto:gcc-help (a] gcc.gnu.org 7018 15. mailto:gcc (a] gcc.gnu.org 7019 16. https://gcc.gnu.org/lists.html 7020 17. http://www.fsf.org/ 7021 18. https://gcc.gnu.org/about.html 7022 19. http://validator.w3.org/check/referer 7023 ====================================================================== 7024 http://gcc.gnu.org/gcc-4.2/changes.html 7025 7026 GCC 4.2 Release Series 7027 Changes, New Features, and Fixes 7028 7029 Caveats 7030 7031 * GCC no longer accepts the -fshared-data option. This option has had 7032 no effect in any GCC 4 release; the targets to which the option 7033 used to apply had been removed before GCC 4.0. 7034 7035 General Optimizer Improvements 7036 7037 * New command-line options specify the possible relationships among 7038 parameters and between parameters and global data. For example, 7039 -fargument-noalias-anything specifies that arguments do not alias 7040 any other storage. 7041 Each language will automatically use whatever option is required by 7042 the language standard. You should not need to use these options 7043 yourself. 7044 7045 New Languages and Language specific improvements 7046 7047 * [1]OpenMP is now supported for the C, C++ and Fortran compilers. 7048 * New command-line options -fstrict-overflow and -Wstrict-overflow 7049 have been added. -fstrict-overflow tells the compiler that it may 7050 assume that the program follows the strict signed overflow 7051 semantics permitted for the language: for C and C++ this means that 7052 the compiler may assume that signed overflow does not occur. For 7053 example, a loop like 7054 for (i = 1; i > 0; i *= 2) 7055 7056 is presumably intended to continue looping until i overflows. With 7057 -fstrict-overflow, the compiler may assume that signed overflow 7058 will not occur, and transform this into an infinite loop. 7059 -fstrict-overflow is turned on by default at -O2, and may be 7060 disabled via -fno-strict-overflow. The -Wstrict-overflow option may 7061 be used to warn about cases where the compiler assumes that signed 7062 overflow will not occur. It takes five different levels: 7063 -Wstrict-overflow=1 to 5. See the [2]documentation for details. 7064 -Wstrict-overflow=1 is enabled by -Wall. 7065 * The new command-line option -fno-toplevel-reorder directs GCC to 7066 emit top-level functions, variables, and asm statements in the same 7067 order that they appear in the input file. This is intended to 7068 support existing code which relies on a particular ordering (for 7069 example, code which uses top-level asm statements to switch 7070 sections). For new code, it is generally better to use function and 7071 variable attributes. The -fno-toplevel-reorder option may be used 7072 for most cases which currently use -fno-unit-at-a-time. The 7073 -fno-unit-at-a-time option will be removed in some future version 7074 of GCC. If you know of a case which requires -fno-unit-at-a-time 7075 which is not fixed by -fno-toplevel-reorder, please open a bug 7076 report. 7077 7078 C family 7079 7080 * The pragma redefine_extname will now macro expand its tokens for 7081 compatibility with SunPRO. 7082 * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct 7083 GCC to handle inline functions as specified in the C99 standard. In 7084 preparation for this, GCC 4.2 will warn about any use of non-static 7085 inline functions in gnu99 or c99 mode. This new warning may be 7086 disabled with the new gnu_inline function attribute or the new 7087 -fgnu89-inline command-line option. Also, GCC 4.2 and later will 7088 define one of the preprocessor macros __GNUC_GNU_INLINE__ or 7089 __GNUC_STDC_INLINE__ to indicate the semantics of inline functions 7090 in the current compilation. 7091 * A new command-line option -Waddress has been added to warn about 7092 suspicious uses of memory addresses as, for example, using the 7093 address of a function in a conditional expression, and comparisons 7094 against the memory address of a string literal. This warning is 7095 enabled by -Wall. 7096 7097 C++ 7098 7099 * C++ visibility handling has been overhauled. 7100 Restricted visiblity is propagated from classes to members, from 7101 functions to local statics, and from templates and template 7102 arguments to instantiations, unless the latter has explicitly 7103 declared visibility. 7104 The visibility attribute for a class must come between the 7105 class-key and the name, not after the closing brace. 7106 Attributes are now allowed for enums and elaborated-type-specifiers 7107 that only declare a type. 7108 Members of the anonymous namespace are now local to a particular 7109 translation unit, along with any other declarations which use them, 7110 though they are still treated as having external linkage for 7111 language semantics. 7112 * The (undocumented) extension which permitted templates with default 7113 arguments to be bound to template template parameters with fewer 7114 parameters has been removed. For example: 7115 template <template <typename> class C> 7116 void f(C<double>) {} 7117 7118 template <typename T, typename U = int> 7119 struct S {}; 7120 7121 template void f(S<double>); 7122 7123 is no longer accepted by G++. The reason this code is not accepted 7124 is that S is a template with two parameters; therefore, it cannot 7125 be bound to C which has only one parameter. 7126 * The <?, >?, <?=, and >?= operators, deprecated in previous GCC 7127 releases, have been removed. 7128 * The command-line option -fconst-strings, deprecated in previous GCC 7129 releases, has been removed. 7130 * The configure variable enable-__cxa_atexit is now enabled by 7131 default for more targets. Enabling this variable is necessary in 7132 order for static destructors to be executed in the correct order, 7133 but it depends upon the presence of a non-standard C library in the 7134 target library in order to work. The variable is now enabled for 7135 more targets which are known to have suitable C libraries. 7136 * -Wextra will produce warnings for if statements with a semicolon as 7137 the only body, to catch code like: 7138 if (a); 7139 return 1; 7140 return 0; 7141 7142 To suppress the warning in valid cases, use { } instead. 7143 * The C++ frontend now also produces strict aliasing warnings when 7144 -fstrict-aliasing -Wstrict-aliasing is in effect. 7145 7146 Runtime Library (libstdc++) 7147 7148 * Added support for TR1 <random>, <complex>, and C compatibility 7149 headers. In addition, a lock-free version of shared_ptr was 7150 contributed as part of Phillip Jordan's Google Summer of Code 7151 project on lock-free containers. ([3]Implementation status of TR1) 7152 * In association with the Summer of Code work on lock-free 7153 containers, the interface for atomic builtins was adjusted, 7154 creating simpler alternatives for non-threaded code paths. Also, 7155 usage was consolidated and all elements were moved from namespace 7156 std to namespace__gnu_cxx. Affected interfaces are the functions 7157 __exchange_and_add, __atomic_add, and the objects __mutex, 7158 __recursive_mutex, and __scoped_lock. 7159 * Support for versioning weak symbol names via namespace association 7160 was added. However, as this changes the names of exported symbols, 7161 this is turned off by default in the current ABI. Intrepid users 7162 can enable this feature by using 7163 --enable-symvers=gnu-versioned-namespace during configuration. 7164 * Revised, simplified, and expanded policy-based associative 7165 containers, including data types for tree and trie forms 7166 (basic_tree, tree, trie), lists (list_update), and both 7167 collision-chaining and probing hash-based containers 7168 (basic_hash_table, cc_hash_table, gp_hash_table). More details per 7169 the [4]documentation. 7170 * The implementation of the debug mode was modified, whereby the 7171 debug namespaces were nested inside of namespace std and namespace 7172 __gnu_cxx in order to resolve some long standing corner cases 7173 involving name lookup. Debug functionality from the policy-based 7174 data structures was consolidated and enabled with the single macro, 7175 _GLIBCXX_DEBUG. See PR 26142 for more information. 7176 * Added extensions for type traits: __conditional_type, 7177 __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if. 7178 * Added a typelist implementation for compile-time meta-programming. 7179 Elements for typelist construction and operation can be found 7180 within namespace __gnu_cxx::typelist. 7181 * Added a new allocator, __gnu_cxx::throw_allocator, for testing 7182 exception-safety. 7183 * Enabled library-wide visibility control, allowing -fvisibility to 7184 be used. 7185 * Consolidated all nested namespaces and the conversion of 7186 __gnu_internal implementation-private details to anonymous 7187 namespaces whenever possible. 7188 * Implemented LWG resolutions DR 431 and DR 538. 7189 7190 Fortran 7191 7192 * Support for allocatable components has been added (TR 15581 and 7193 Fortran 2003). 7194 * Support for the Fortran 2003 streaming IO extension has been added. 7195 * The GNU Fortran compiler now uses 4-byte record markers by default 7196 for unformatted files to be compatible with g77 and most other 7197 compilers. The implementation allows for records greater than 2 GB 7198 and is compatible with several other compilers. Older versions of 7199 gfortran used 8-byte record markers by default (on most systems). 7200 In order to change the length of the record markers, e.g. to read 7201 unformatted files created by older gfortran versions, the 7202 [5]-frecord-marker=8 option can be used. 7203 7204 Java (GCJ) 7205 7206 * A new command-line option -static-libgcj has been added for targets 7207 that use a linker compatible with GNU Binutils. As its name 7208 implies, this causes libgcj to be linked statically. In some cases 7209 this causes the resulting executable to start faster and use less 7210 memory than if the shared version of libgcj were used. However 7211 caution should be used as it can also cause essential parts of the 7212 library to be omitted. Some of these issues are discussed in: 7213 [6]https://gcc.gnu.org/wiki/Statically_linking_libgcj 7214 * fastjar is no longer bundled with GCC. To build libgcj, you will 7215 need either InfoZIP (both zip and unzip) or an external jar 7216 program. In the former case, the GCC build will install a jar shell 7217 script that is based on InfoZIP and provides the same functionality 7218 as fastjar. 7219 7220 New Targets and Target Specific Improvements 7221 7222 IA-32/x86-64 7223 7224 * -mtune=generic can now be used to generate code running well on 7225 common x86 chips. This includes AMD Athlon, AMD Opteron, Intel 7226 Pentium-M, Intel Pentium 4 and Intel Core 2. 7227 * -mtune=native and -march=native will produce code optimized for the 7228 host architecture as detected using the cpuid instruction. 7229 * Added a new command-line option -fstackrealign and and 7230 __attribute__ ((force_align_arg_pointer)) to realign the stack at 7231 runtime. This allows functions compiled with a vector-aligned stack 7232 to be invoked from legacy objects that keep only word-alignment. 7233 7234 SPARC 7235 7236 * The default CPU setting has been changed from V7 to V9 in 32-bit 7237 mode on Solaris 7 and above. This is already the case in 64-bit 7238 mode. It can be overridden by specifying --with-cpu at configure 7239 time. 7240 * Back-end support of built-in functions for atomic memory access has 7241 been implemented. 7242 * Support for the Sun UltraSPARC T1 (Niagara) processor has been 7243 added. 7244 7245 M32C 7246 7247 * Various bug fixes have made some functions (notably, functions 7248 returning structures) incompatible with previous releases. 7249 Recompiling all libraries is recommended. Note that code quality 7250 has considerably improved since 4.1, making a recompile even more 7251 beneficial. 7252 7253 MIPS 7254 7255 * Added support for the Broadcom SB-1A core. 7256 7257 IA-64 7258 7259 * Added support for IA-64 data and control speculation. By default 7260 speculation is enabled only during second scheduler pass. A number 7261 of machine flags was introduced to control the usage of speculation 7262 for both scheduler passes. 7263 7264 HPPA 7265 7266 * Added Java language support (libffi and libjava) for 32-bit HP-UX 7267 11 target. 7268 7269 Obsolete Systems 7270 7271 Documentation improvements 7272 7273 PDF Documentation 7274 7275 * A make pdf target has been added to the top-level makefile, 7276 enabling automated production of PDF documentation files. 7277 (Front-ends external to GCC should modify their Make-lang.in file 7278 to add a lang.pdf: target.) 7279 7280 Other significant improvements 7281 7282 Build system improvements 7283 7284 * All the components of the compiler are now bootstrapped by default. 7285 This improves the resilience to bugs in the system compiler or 7286 binary compatibility problems, as well as providing better testing 7287 of GCC 4.2 itself. In addition, if you build the compiler from a 7288 combined tree, the assembler, linker, etc. will also be 7289 bootstrapped (i.e. built with themselves). 7290 You can disable this behavior, and go back to the pre-GCC 4.2 set 7291 up, by configuring GCC with --disable-bootstrap. 7292 * The rules that configure follows to find target tools resemble more 7293 closely the locations that the built compiler will search. In 7294 addition, you can use the new configure option --with-target-tools 7295 to specify where to find the target tools used during the build, 7296 without affecting what the built compiler will use. 7297 This can be especially useful when building packages of GCC. For 7298 example, you may want to build GCC with GNU as or ld, even if the 7299 resulting compiler to work with the native assembler and linker. To 7300 do so, you can use --with-target-tools to point to the native 7301 tools. 7302 7303 Incompatible changes to the build system 7304 7305 * Front-ends external to GCC should modify their Make-lang.in file to 7306 replace double-colon rules (e.g. dvi::) with normal rules (like 7307 lang.dvi:). Front-end makefile hooks do not use double-colon rules 7308 anymore. 7309 * Up to GCC 4.1, a popular way to specify the target tools used 7310 during the build was to create directories named gas, binutils, 7311 etc. in the build tree, and create links to the tools from there. 7312 This does not work any more when the compiler is bootstrapped. The 7313 new configure option --with-target-tools provides a better way to 7314 achieve the same effect, and works for all native and cross 7315 settings. 7316 7317 7318 For questions related to the use of GCC, please consult these web 7319 pages and the [7]GCC manuals. If that fails, the 7320 [8]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7321 web pages and the development of GCC are welcome on our developer 7322 list at [9]gcc (a] gcc.gnu.org. All of [10]our lists have public 7323 archives. 7324 7325 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and 7326 distribution of this entire article is permitted in any medium, 7327 provided this notice is preserved. 7328 7329 These pages are [12]maintained by the GCC team. Last modified 7330 2014-06-28[13]. 7331 7332 References 7333 7334 1. http://gcc.gnu.org/projects/gomp/ 7335 2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 7336 3. https://gcc.gnu.org/onlinedocs/libstdc++/manual/bk01pt01ch01.html#manual.intro.status.standard.tr1 7337 4. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html 7338 5. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html 7339 6. https://gcc.gnu.org/wiki/Statically_linking_libgcj 7340 7. https://gcc.gnu.org/onlinedocs/ 7341 8. mailto:gcc-help (a] gcc.gnu.org 7342 9. mailto:gcc (a] gcc.gnu.org 7343 10. https://gcc.gnu.org/lists.html 7344 11. http://www.fsf.org/ 7345 12. https://gcc.gnu.org/about.html 7346 13. http://validator.w3.org/check/referer 7347 ====================================================================== 7348 http://gcc.gnu.org/gcc-4.1/index.html 7349 7350 GCC 4.1 Release Series 7351 7352 February 13, 2007 7353 7354 The [1]GNU project and the GCC developers are pleased to announce the 7355 release of GCC 4.1.2. 7356 7357 This release is a bug-fix release, containing fixes for regressions in 7358 GCC 4.1.1 relative to previous releases of GCC. 7359 7360 Release History 7361 7362 GCC 4.1.2 7363 February 13, 2007 ([2]changes) 7364 7365 GCC 4.1.1 7366 May 24, 2006 ([3]changes) 7367 7368 GCC 4.1.0 7369 February 28, 2006 ([4]changes) 7370 7371 References and Acknowledgements 7372 7373 GCC used to stand for the GNU C Compiler, but since the compiler 7374 supports several other languages aside from C, it now stands for the 7375 GNU Compiler Collection. 7376 7377 A list of [5]successful builds is updated as new information becomes 7378 available. 7379 7380 The GCC developers would like to thank the numerous people that have 7381 contributed new features, improvements, bug fixes, and other changes as 7382 well as test results to GCC. This [6]amazing group of volunteers is 7383 what makes GCC successful. 7384 7385 For additional information about GCC please refer to the [7]GCC project 7386 web site or contact the [8]GCC development mailing list. 7387 7388 To obtain GCC please use [9]our mirror sites or [10]our SVN server. 7389 7390 7391 For questions related to the use of GCC, please consult these web 7392 pages and the [11]GCC manuals. If that fails, the 7393 [12]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7394 web pages and the development of GCC are welcome on our developer 7395 list at [13]gcc (a] gcc.gnu.org. All of [14]our lists have public 7396 archives. 7397 7398 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and 7399 distribution of this entire article is permitted in any medium, 7400 provided this notice is preserved. 7401 7402 These pages are [16]maintained by the GCC team. Last modified 7403 2014-06-28[17]. 7404 7405 References 7406 7407 1. http://www.gnu.org/ 7408 2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 7409 3. http://gcc.gnu.org/gcc-4.1/changes.html 7410 4. http://gcc.gnu.org/gcc-4.1/changes.html 7411 5. http://gcc.gnu.org/gcc-4.1/buildstat.html 7412 6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7413 7. http://gcc.gnu.org/index.html 7414 8. mailto:gcc (a] gcc.gnu.org 7415 9. http://gcc.gnu.org/mirrors.html 7416 10. http://gcc.gnu.org/svn.html 7417 11. https://gcc.gnu.org/onlinedocs/ 7418 12. mailto:gcc-help (a] gcc.gnu.org 7419 13. mailto:gcc (a] gcc.gnu.org 7420 14. https://gcc.gnu.org/lists.html 7421 15. http://www.fsf.org/ 7422 16. https://gcc.gnu.org/about.html 7423 17. http://validator.w3.org/check/referer 7424 ====================================================================== 7425 http://gcc.gnu.org/gcc-4.1/changes.html 7426 7427 GCC 4.1 Release Series 7428 Changes, New Features, and Fixes 7429 7430 The latest release in the 4.1 release series is [1]GCC 4.1.2. 7431 7432 Caveats 7433 7434 General Optimizer Improvements 7435 7436 * GCC now has infrastructure for inter-procedural optimizations and 7437 the following inter-procedural optimizations are implemented: 7438 + Profile guided inlining. When doing profile feedback guided 7439 optimization, GCC can now use the profile to make better 7440 informed decisions on whether inlining of a function is 7441 profitable or not. This means that GCC will no longer inline 7442 functions at call sites that are not executed very often, and 7443 that functions at hot call sites are more likely to be 7444 inlined. 7445 A new parameter min-inline-recursive-probability is also now 7446 available to throttle recursive inlining of functions with 7447 small average recursive depths. 7448 + Discovery of pure and const functions, a form of side-effects 7449 analysis. While older GCC releases could also discover such 7450 special functions, the new IPA-based pass runs earlier so that 7451 the results are available to more optimizers. The pass is also 7452 simply more powerful than the old one. 7453 + Analysis of references to static variables and type escape 7454 analysis, also forms of side-effects analysis. The results of 7455 these passes allow the compiler to be less conservative about 7456 call-clobbered variables and references. This results in more 7457 redundant loads being eliminated and in making static 7458 variables candidates for register promotion. 7459 + Improvement of RTL-based alias analysis. The results of type 7460 escape analysis are fed to the RTL type-based alias analyzer, 7461 allowing it to disambiguate more memory references. 7462 + Interprocedural constant propagation and function versioning. 7463 This pass looks for functions that are always called with the 7464 same constant value for one or more of the function arguments, 7465 and propagates those constants into those functions. 7466 + GCC will now eliminate static variables whose usage was 7467 optimized out. 7468 + -fwhole-program --combine can now be used to make all 7469 functions in program static allowing whole program 7470 optimization. As an exception, the main function and all 7471 functions marked with the new externally_visible attribute are 7472 kept global so that programs can link with runtime libraries. 7473 * GCC can now do a form of partial dead code elimination (PDCE) that 7474 allows code motion of expressions to the paths where the result of 7475 the expression is actually needed. This is not always a win, so the 7476 pass has been limited to only consider profitable cases. Here is an 7477 example: 7478 int foo (int *, int *); 7479 int 7480 bar (int d) 7481 { 7482 int a, b, c; 7483 b = d + 1; 7484 c = d + 2; 7485 a = b + c; 7486 if (d) 7487 { 7488 foo (&b, &c); 7489 a = b + c; 7490 } 7491 printf ("%d\n", a); 7492 } 7493 7494 The a = b + c can be sunk to right before the printf. Normal code 7495 sinking will not do this, it will sink the first one above into the 7496 else-branch of the conditional jump, which still gives you two 7497 copies of the code. 7498 * GCC now has a value range propagation pass. This allows the 7499 compiler to eliminate bounds checks and branches. The results of 7500 the pass can also be used to accurately compute branch 7501 probabilities. 7502 * The pass to convert PHI nodes to straight-line code (a form of 7503 if-conversion for GIMPLE) has been improved significantly. The two 7504 most significant improvements are an improved algorithm to 7505 determine the order in which the PHI nodes are considered, and an 7506 improvement that allow the pass to consider if-conversions of basic 7507 blocks with more than two predecessors. 7508 * Alias analysis improvements. GCC can now differentiate between 7509 different fields of structures in Tree-SSA's virtual operands form. 7510 This lets stores/loads from non-overlapping structure fields not 7511 conflict. A new algorithm to compute points-to sets was contributed 7512 that can allows GCC to see now that p->a and p->b, where p is a 7513 pointer to a structure, can never point to the same field. 7514 * Various enhancements to auto-vectorization: 7515 + Incrementally preserve SSA form when vectorizing. 7516 + Incrementally preserve loop-closed form when vectorizing. 7517 + Improvements to peeling for alignment: generate better code 7518 when the misalignment of an access is known at compile time, 7519 or when different accesses are known to have the same 7520 misalignment, even if the misalignment amount itself is 7521 unknown. 7522 + Consider dependence distance in the vectorizer. 7523 + Externalize generic parts of data reference analysis to make 7524 this analysis available to other passes. 7525 + Vectorization of conditional code. 7526 + Reduction support. 7527 * GCC can now partition functions in sections of hot and cold code. 7528 This can significantly improve performance due to better 7529 instruction cache locality. This feature works best together with 7530 profile feedback driven optimization. 7531 * A new pass to avoid saving of unneeded arguments to the stack in 7532 vararg functions if the compiler can prove that they will not be 7533 needed. 7534 * Transition of basic block profiling to tree level implementation 7535 has been completed. The new implementation should be considerably 7536 more reliable (hopefully avoiding profile mismatch errors when 7537 using -fprofile-use or -fbranch-probabilities) and can be used to 7538 drive higher level optimizations, such as inlining. 7539 The -ftree-based-profiling command-line option was removed and 7540 -fprofile-use now implies disabling old RTL level loop optimizer 7541 (-fno-loop-optimize). Speculative prefetching optimization 7542 (originally enabled by -fspeculative-prefetching) was removed. 7543 7544 New Languages and Language specific improvements 7545 7546 C and Objective-C 7547 7548 * The old Bison-based C and Objective-C parser has been replaced by a 7549 new, faster hand-written recursive-descent parser. 7550 7551 Ada 7552 7553 * The build infrastructure for the Ada runtime library and tools has 7554 been changed to be better integrated with the rest of the build 7555 infrastructure of GCC. This should make doing cross builds of Ada a 7556 bit easier. 7557 7558 C++ 7559 7560 * ARM-style name-injection of friend declarations is no longer the 7561 default. For example: 7562 struct S { 7563 friend void f(); 7564 }; 7565 7566 void g() { f(); } 7567 will not be accepted; instead a declaration of f will need to be 7568 present outside of the scope of S. The new -ffriend-injection 7569 option will enable the old behavior. 7570 * The (undocumented) extension which permitted templates with default 7571 arguments to be bound to template template parameters with fewer 7572 parameters has been deprecated, and will be removed in the next 7573 major release of G++. For example: 7574 template <template <typename> class C> 7575 void f(C<double>) {} 7576 7577 template <typename T, typename U = int> 7578 struct S {}; 7579 7580 template void f(S<double>); 7581 7582 makes use of the deprecated extension. The reason this code is not 7583 valid ISO C++ is that S is a template with two parameters; 7584 therefore, it cannot be bound to C which has only one parameter. 7585 7586 Runtime Library (libstdc++) 7587 7588 * Optimization work: 7589 + A new implementation of std::search_n is provided, better 7590 performing in case of random access iterators. 7591 + Added further efficient specializations of istream functions, 7592 i.e., character array and string extractors. 7593 + Other smaller improvements throughout. 7594 * Policy-based associative containers, designed for high-performance, 7595 flexibility and semantic safety are delivered in ext/pb_assoc. 7596 * A versatile string class, __gnu_cxx::__versa_string, providing 7597 facilities conforming to the standard requirements for 7598 basic_string, is delivered in <ext/vstring.h>. In particular: 7599 + Two base classes are provided: the default one avoids 7600 reference counting and is optimized for short strings; the 7601 alternate one, still uses it while improving in a few low 7602 level areas (e.g., alignment). See vstring_fwd.h for some 7603 useful typedefs. 7604 + Various algorithms have been rewritten (e.g., replace), the 7605 code streamlined and simple optimizations added. 7606 + Option 3 of DR 431 is implemented for both available bases, 7607 thus improving the support for stateful allocators. 7608 * As usual, many bugs have been fixed (e.g., libstdc++/13583, 7609 libstdc++/23953) and LWG resolutions put into effect for the first 7610 time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1 7611 Issue 6.19). The implementation status of TR1 is now tracked in the 7612 docs in tr1.html. 7613 7614 Objective-C++ 7615 7616 * A new language front end for Objective-C++ has been added. This 7617 language allows users to mix the object oriented features of 7618 Objective-C with those of C++. 7619 7620 Java (GCJ) 7621 7622 * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19 7623 features (plus some 0.20 bug-fixes) 7624 + Networking 7625 o The java.net.HttpURLConnection implementation no longer 7626 buffers the entire response body in memory. This means 7627 that response bodies larger than available memory can now 7628 be handled. 7629 + (N)IO 7630 o NIO FileChannel.map implementation, fast bulk put 7631 implementation for DirectByteBuffer (speeds up this 7632 method 10x). 7633 o FileChannel.lock() and FileChannel.force() implemented. 7634 + XML 7635 o gnu.xml fix for nodes created outside a namespace 7636 context. 7637 o Add support for output indenting and 7638 cdata-section-elements output instruction in 7639 xml.transform. 7640 o xml.xpath corrections for cases where elements/attributes 7641 might have been created in non-namespace-aware mode. 7642 Corrections to handling of XSL variables and minor 7643 conformance updates. 7644 + AWT 7645 o GNU JAWT implementation, the AWT Native Interface, which 7646 allows direct access to native screen resources from 7647 within a Canvas's paint method. GNU Classpath Examples 7648 comes with a Demo, see libjava/classpath/examples/README. 7649 o awt.datatransfer updated to 1.5 with support for 7650 FlavorEvents. The gtk+ awt peers now allow copy/paste of 7651 text, images, URIs/files and serialized objects with 7652 other applications and tracking clipboard change events 7653 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized 7654 objects are supported). A GNU Classpath Examples 7655 datatransfer Demo was added to show the new 7656 functionality. 7657 o Split gtk+ awt peers event handling in two threads and 7658 improve gdk lock handling (solves several awt lock ups). 7659 o Speed up awt Image loading. 7660 o Better gtk+ scrollbar peer implementation when using gtk+ 7661 >= 2.6. 7662 o Handle image loading errors correctly for gdkpixbuf and 7663 MediaTracker. 7664 o Better handle GDK lock. Properly prefix gtkpeer native 7665 functions (cp_gtk). 7666 o GdkGraphics2D has been updated to use Cairo 0.5.x or 7667 higher. 7668 o BufferedImage and GtkImage rewrites. All image drawing 7669 operations should now work correctly (flipping requires 7670 gtk+ >= 2.6) 7671 o Future Graphics2D, image and text work is documented at: 7672 [2]http://developer.classpath.org/mediation/ClasspathGrap 7673 hicsImagesText 7674 o When gtk+ 2.6 or higher is installed the default log 7675 handler will produce stack traces whenever a WARNING, 7676 CRITICAL or ERROR message is produced. 7677 + Free Swing 7678 o The RepaintManager has been reworked for more efficient 7679 painting, especially for large GUIs. 7680 o The layout manager OverlayLayout has been implemented, 7681 the BoxLayout has been rewritten to make use of the 7682 SizeRequirements utility class and caching for more 7683 efficient layout. 7684 o Improved accessibility support. 7685 o Significant progress has been made in the implementation 7686 of the javax.swing.plaf.metal package, with most UI 7687 delegates in a working state now. Please test this with 7688 your own applications and provide feedback that will help 7689 us to improve this package. 7690 o The GUI demo (gnu.classpath.examples.swing.Demo) has been 7691 extended to highlight various features in our Free Swing 7692 implementation. And it includes a look and feel switcher 7693 for Metal (default), Ocean and GNU themes. 7694 o The javax.swing.plaf.multi package is now implemented. 7695 o Editing and several key actions for JTree and JTable were 7696 implemented. 7697 o Lots of icons and look and feel improvements for Free 7698 Swing basic and metal themes were added. Try running the 7699 GNU Classpath Swing Demo in examples 7700 (gnu.classpath.examples.swing.Demo) with: 7701 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee 7702 l or 7703 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee 7704 l 7705 o Start of styled text capabilites for java.swing.text. 7706 o DefaultMutableTreeNode pre-order, post-order, depth-first 7707 and breadth-first traversal enumerations implemented. 7708 o JInternalFrame colors and titlebar draw properly. 7709 o JTree is working up to par (icons, selection and keyboard 7710 traversal). 7711 o JMenus were made more compatible in visual and 7712 programmatic behavior. 7713 o JTable changeSelection and multiple selections 7714 implemented. 7715 o JButton and JToggleButton change states work properly 7716 now. 7717 o JFileChooser fixes. 7718 o revalidate() and repaint() fixes which make Free Swing 7719 much more responsive. 7720 o MetalIconFactory implemented. 7721 o Free Swing Top-Level Compatibility. JFrame, JDialog, 7722 JApplet, JInternalFrame, and JWindow are now 1.5 7723 compatible in the sense that you can call add() and 7724 setLayout() directly on them, which will have the same 7725 effect as calling getContentPane().add() and 7726 getContentPane().setLayout(). 7727 o The JTree interface has been completed. JTrees now 7728 recognizes mouse clicks and selections work. 7729 o BoxLayout works properly now. 7730 o Fixed GrayFilter to actually work. 7731 o Metal SplitPane implemented. 7732 o Lots of Free Swing text and editor stuff work now. 7733 + Free RMI and Corba 7734 o Andrew Watson, Vice President and Technical Director of 7735 the Object Management Group, has officially assigned us 7736 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will 7737 mark remote classpath-specific system exceptions. 7738 Obtaining the VMCID means that GNU Classpath now is a 7739 recogniseable type of node in a highly interoperable 7740 CORBA world. 7741 o GNU Classpath now includes the first working draft to 7742 support the RMI over IIOP protocol. The current 7743 implementation is capable of remote invocations, 7744 transferring various Serializables and Externalizables 7745 via RMI-IIOP protocol. It can flatten graphs and, at 7746 least for the simple cases, is interoperable with 1.5 7747 JDKs. 7748 o org.omg.PortableInterceptor and related functionality in 7749 other packages is now implemented: 7750 # The sever and client interceptors work as required 7751 since 1.4. 7752 # The IOR interceptor works as needed for 1.5. 7753 o The org.omg.DynamicAny package is completed and passes 7754 the prepared tests. 7755 o The Portable Object Adapter should now support the output 7756 of the recent IDL to java compilers. These compilers now 7757 generate servants and not CORBA objects as before, making 7758 the output depend on the existing POA implementation. 7759 Completing POA means that such code can already be tried 7760 to run on Classpath. Our POA is tested for the following 7761 usager scenarios: 7762 # POA converts servant to the CORBA object. 7763 # Servant provides to the CORBA object. 7764 # POA activates new CORBA object with the given Object 7765 Id (byte array) that is later accessible for the 7766 servant. 7767 # During the first call, the ServantActivator provides 7768 servant for this and all subsequent calls on the 7769 current object. 7770 # During each call, the ServantLocator provides 7771 servant for this call only. 7772 # ServantLocator or ServantActivator forwards call to 7773 another server. 7774 # POA has a single servant, responsible for all 7775 objects. 7776 # POA has a default servant, but some objects are 7777 explicitly connected to they specific servants. 7778 The POA is verified using tests from the former 7779 cost.omg.org. 7780 o The CORBA implementation is now a working prototype that 7781 should support features up to 1.3 inclusive. We invite 7782 groups writing CORBA dependent applications to try 7783 Classpath implementation, reporting any possible bugs. 7784 The CORBA prototype is interoperable with Sun's 7785 implementation v 1.4, transferring object references, 7786 primitive types, narrow and wide strings, arrays, 7787 structures, trees, abstract interfaces and value types 7788 (feature of CORBA 2.3) between these two platforms. 7789 Remote exceptions are transferred and handled correctly. 7790 The stringified object references (IORs) from various 7791 sources are parsed as required. The transient (for 7792 current session) and permanent (till jre restart) 7793 redirections work. Both Little and Big Endian encoded 7794 messages are accepted. The implementation is verified 7795 using tests from the former cost.omg.org. The current 7796 release includes working examples (see the examples 7797 directory), demonstrating the client-server 7798 communication, using either CORBA Request or IDL-based 7799 stub (usually generated by a IDL to java compiler). These 7800 examples also show how to use the Classpath CORBA naming 7801 service. The IDL to java compiler is not yet written, but 7802 as our library must be compatible, it naturally accepts 7803 the output of other idlj implementations. 7804 + Misc 7805 o Updated TimeZone data against Olson tzdata2005l. 7806 o Make zip and jar packages UTF-8 clean. 7807 o "native" code builds and compiles (warning free) on 7808 Darwin and Solaris. 7809 o java.util.logging.FileHandler now rotates files. 7810 o Start of a generic JDWP framework in gnu/classpath/jdwp. 7811 This is unfinished, but feedback (at classpath (a] gnu.org) 7812 from runtime hackers is greatly appreciated. Although 7813 most of the work is currently being done around gcj/gij 7814 we want this framework to be as VM neutral as possible. 7815 Early design is described in: 7816 [3]https://gcc.gnu.org/ml/java/2005-05/msg00260.html 7817 o QT4 AWT peers, enable by giving configure 7818 --enable-qt-peer. Included, but not ready for production 7819 yet. They are explicitly disabled and not supported. But 7820 if you want to help with the development of these new 7821 features we are interested in feedback. You will have to 7822 explicitly enable them to try them out (and they will 7823 most likely contain bugs). 7824 o Documentation fixes all over the place. See 7825 [4]http://developer.classpath.org/doc/ 7826 7827 New Targets and Target Specific Improvements 7828 7829 IA-32/x86-64 7830 7831 * The x86-64 medium model (that allows building applications whose 7832 data segment exceeds 4GB) was redesigned to match latest ABI draft. 7833 New implementation split large datastructures into separate segment 7834 improving performance of accesses to small datastructures and also 7835 allows linking of small model libraries into medium model programs 7836 as long as the libraries are not accessing the large datastructures 7837 directly. Medium model is also supported in position independent 7838 code now. 7839 The ABI change results in partial incompatibility among medium 7840 model objects. Linking medium model libraries (or objects) compiled 7841 with new compiler into medium model program compiled with older 7842 will likely result in exceeding ranges of relocations. 7843 Binutils 2.16.91 or newer are required for compiling medium model 7844 now. 7845 7846 RS6000 (POWER/PowerPC) 7847 7848 * The AltiVec vector primitives in <altivec.h> are now implemented in 7849 a way that puts a smaller burden on the preprocessor, instead 7850 processing the "overloading" in the front ends. This should benefit 7851 compilation speed on AltiVec vector code. 7852 * AltiVec initializers now are generated more efficiently. 7853 * The popcountb instruction available on POWER5 now is generated. 7854 * The floating point round to integer instructions available on 7855 POWER5+ now is generated. 7856 * Floating point divides can be synthesized using the floating point 7857 reciprocal estimate instructions. 7858 * Double precision floating point constants are initialized as single 7859 precision values if they can be represented exactly. 7860 7861 S/390, zSeries and System z9 7862 7863 * Support for the IBM System z9 109 processor has been added. When 7864 using the -march=z9-109 option, the compiler will generate code 7865 making use of instructions provided by the extended immediate 7866 facility. 7867 * Support for 128-bit IEEE floating point has been added. When using 7868 the -mlong-double-128 option, the compiler will map the long double 7869 data type to 128-bit IEEE floating point. Using this option 7870 constitutes an ABI change, and requires glibc support. 7871 * Various changes to improve performance of generated code have been 7872 implemented, including: 7873 + In functions that do not require a literal pool, register %r13 7874 (which is traditionally reserved as literal pool pointer), can 7875 now be freely used for other purposes by the compiler. 7876 + More precise tracking of register use allows the compiler to 7877 generate more efficient function prolog and epilog code in 7878 certain cases. 7879 + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING 7880 instructions are now used to implement C string functions. 7881 + The MOVE CHARACTER instruction with single byte overlap is now 7882 used to implement the memset function with non-zero fill byte. 7883 + The LOAD ZERO instructions are now used where appropriate. 7884 + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK, 7885 and INSERT IMMEDIATE instructions are now used more frequently 7886 to optimize bitfield operations. 7887 + The BRANCH ON COUNT instruction is now used more frequently. 7888 In particular, the fact that a loop contains a subroutine call 7889 no longer prevents the compiler from using this instruction. 7890 + The compiler is now aware that all shift and rotate 7891 instructions implicitly truncate the shift count to six bits. 7892 * Back-end support for the following generic features has been 7893 implemented: 7894 + The full set of [5]built-in functions for atomic memory 7895 access. 7896 + The -fstack-protector feature. 7897 + The optimization pass avoiding unnecessary stores of incoming 7898 argument registers in functions with variable argument list. 7899 7900 SPARC 7901 7902 * The default code model in 64-bit mode has been changed from 7903 Medium/Anywhere to Medium/Middle on Solaris. 7904 * TLS support is disabled by default on Solaris prior to release 10. 7905 It can be enabled on TLS-capable Solaris 9 versions (4/04 release 7906 and later) by specifying --enable-tls at configure time. 7907 7908 MorphoSys 7909 7910 * Support has been added for this new architecture. 7911 7912 Obsolete Systems 7913 7914 Documentation improvements 7915 7916 Other significant improvements 7917 7918 * GCC can now emit code for protecting applications from 7919 stack-smashing attacks. The protection is realized by buffer 7920 overflow detection and reordering of stack variables to avoid 7921 pointer corruption. 7922 * Some built-in functions have been fortified to protect them against 7923 various buffer overflow (and format string) vulnerabilities. 7924 Compared to the mudflap bounds checking feature, the safe builtins 7925 have far smaller overhead. This means that programs built using 7926 safe builtins should not experience any measurable slowdown. 7927 7928 GCC 4.1.2 7929 7930 This is the [6]list of problem reports (PRs) from GCC's bug tracking 7931 system that are known to be fixed in the 4.1.2 release. This list might 7932 not be complete (that is, it is possible that some PRs that have been 7933 fixed are not listed here). 7934 7935 When generating code for a shared library, GCC now recognizes that 7936 global functions may be replaced when the program runs. Therefore, it 7937 is now more conservative in deducing information from the bodies of 7938 functions. For example, in this example: 7939 void f() {} 7940 void g() { 7941 try { f(); } 7942 catch (...) { 7943 cout << "Exception"; 7944 } 7945 } 7946 7947 G++ would previously have optimized away the catch clause, since it 7948 would have concluded that f cannot throw exceptions. Because users may 7949 replace f with another function in the main body of the program, this 7950 optimization is unsafe, and is no longer performed. If you wish G++ to 7951 continue to optimize as before, you must add a throw() clause to the 7952 declaration of f to make clear that it does not throw exceptions. 7953 7954 7955 For questions related to the use of GCC, please consult these web 7956 pages and the [7]GCC manuals. If that fails, the 7957 [8]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7958 web pages and the development of GCC are welcome on our developer 7959 list at [9]gcc (a] gcc.gnu.org. All of [10]our lists have public 7960 archives. 7961 7962 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and 7963 distribution of this entire article is permitted in any medium, 7964 provided this notice is preserved. 7965 7966 These pages are [12]maintained by the GCC team. Last modified 7967 2014-06-28[13]. 7968 7969 References 7970 7971 1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 7972 2. http://developer.classpath.org/mediation/ClasspathGraphicsImagesText 7973 3. https://gcc.gnu.org/ml/java/2005-05/msg00260.html 7974 4. http://developer.classpath.org/doc/ 7975 5. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html 7976 6. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2 7977 7. https://gcc.gnu.org/onlinedocs/ 7978 8. mailto:gcc-help (a] gcc.gnu.org 7979 9. mailto:gcc (a] gcc.gnu.org 7980 10. https://gcc.gnu.org/lists.html 7981 11. http://www.fsf.org/ 7982 12. https://gcc.gnu.org/about.html 7983 13. http://validator.w3.org/check/referer 7984 ====================================================================== 7985 http://gcc.gnu.org/gcc-4.0/index.html 7986 7987 GCC 4.0 Release Series 7988 7989 January 31, 2007 7990 7991 The [1]GNU project and the GCC developers are pleased to announce the 7992 release of GCC 4.0.4. 7993 7994 This release is a bug-fix release, containing fixes for regressions in 7995 GCC 4.0.3 relative to previous releases of GCC. 7996 7997 Release History 7998 7999 GCC 4.0.4 8000 January 31, 2007 ([2]changes) 8001 8002 GCC 4.0.3 8003 March 10, 2006 ([3]changes) 8004 8005 GCC 4.0.2 8006 September 28, 2005 ([4]changes) 8007 8008 GCC 4.0.1 8009 July 7, 2005 ([5]changes) 8010 8011 GCC 4.0.0 8012 April 20, 2005 ([6]changes) 8013 8014 References and Acknowledgements 8015 8016 GCC used to stand for the GNU C Compiler, but since the compiler 8017 supports several other languages aside from C, it now stands for the 8018 GNU Compiler Collection. 8019 8020 A list of [7]successful builds is updated as new information becomes 8021 available. 8022 8023 The GCC developers would like to thank the numerous people that have 8024 contributed new features, improvements, bug fixes, and other changes as 8025 well as test results to GCC. This [8]amazing group of volunteers is 8026 what makes GCC successful. 8027 8028 For additional information about GCC please refer to the [9]GCC project 8029 web site or contact the [10]GCC development mailing list. 8030 8031 To obtain GCC please use [11]our mirror sites, or [12]our SVN server. 8032 8033 8034 For questions related to the use of GCC, please consult these web 8035 pages and the [13]GCC manuals. If that fails, the 8036 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8037 web pages and the development of GCC are welcome on our developer 8038 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 8039 archives. 8040 8041 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 8042 distribution of this entire article is permitted in any medium, 8043 provided this notice is preserved. 8044 8045 These pages are [18]maintained by the GCC team. Last modified 8046 2014-06-28[19]. 8047 8048 References 8049 8050 1. http://www.gnu.org/ 8051 2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 8052 3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3 8053 4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2 8054 5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1 8055 6. http://gcc.gnu.org/gcc-4.0/changes.html 8056 7. http://gcc.gnu.org/gcc-4.0/buildstat.html 8057 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8058 9. http://gcc.gnu.org/index.html 8059 10. mailto:gcc (a] gcc.gnu.org 8060 11. http://gcc.gnu.org/mirrors.html 8061 12. http://gcc.gnu.org/svn.html 8062 13. https://gcc.gnu.org/onlinedocs/ 8063 14. mailto:gcc-help (a] gcc.gnu.org 8064 15. mailto:gcc (a] gcc.gnu.org 8065 16. https://gcc.gnu.org/lists.html 8066 17. http://www.fsf.org/ 8067 18. https://gcc.gnu.org/about.html 8068 19. http://validator.w3.org/check/referer 8069 ====================================================================== 8070 http://gcc.gnu.org/gcc-4.0/changes.html 8071 8072 GCC 4.0 Release Series 8073 Changes, New Features, and Fixes 8074 8075 The latest release in the 4.0 release series is [1]GCC 4.0.4. 8076 8077 Caveats 8078 8079 * GCC now generates location lists by default when compiling with 8080 debug info and optimization. 8081 + GDB 6.0 and older crashes when it sees location lists. GDB 6.1 8082 or later is needed to debug binaries containing location 8083 lists. 8084 + When you are trying to view a value of a variable in a part of 8085 a function where it has no location (for example when the 8086 variable is no longer used and thus its location was used for 8087 something else) GDB will say that it is not available. 8088 You can disable generating location lists by -fno-var-tracking. 8089 * GCC no longer accepts the -fwritable-strings option. Use named 8090 character arrays when you need a writable string. 8091 * The options -freduce-all-givs and -fmove-all-movables have been 8092 discontinued. They were used to circumvent a shortcoming in the 8093 heuristics of the old loop optimization code with respect to common 8094 Fortran constructs. The new (tree) loop optimizer works differently 8095 and doesn't need those work-arounds. 8096 * The graph-coloring register allocator, formerly enabled by the 8097 option -fnew-ra, has been discontinued. 8098 * -I- has been deprecated. -iquote is meant to replace the need for 8099 this option. 8100 * The MIPS -membedded-pic and -mrnames options have been removed. 8101 * All MIPS targets now require the GNU assembler. In particular, IRIX 8102 configurations can no longer use the MIPSpro assemblers, although 8103 they do still support the MIPSpro linkers. 8104 * The SPARC option -mflat has been removed. 8105 * English-language diagnostic messages will now use Unicode quotation 8106 marks in UTF-8 locales. (Non-English messages already used the 8107 quotes appropriate for the language in previous releases.) If your 8108 terminal does not support UTF-8 but you are using a UTF-8 locale 8109 (such locales are the default on many GNU/Linux systems) then you 8110 should set LC_CTYPE=C in the environment to disable that locale. 8111 Programs that parse diagnostics and expect plain ASCII 8112 English-language messages should set LC_ALL=C. See [2]Markus Kuhn's 8113 explanation of Unicode quotation marks for more information. 8114 * The specs file is no longer installed on most platforms. Most users 8115 will be totally unaffected. However, if you are accustomed to 8116 editing the specs file yourself, you will now have to use the 8117 -dumpspecs option to generate the specs file, and then edit the 8118 resulting file. 8119 8120 General Optimizer Improvements 8121 8122 * The [3]tree ssa branch has been merged. This merge has brought in a 8123 completely new optimization framework based on a higher level 8124 intermediate representation than the existing RTL representation. 8125 Numerous new code transformations based on the new framework are 8126 available in GCC 4.0, including: 8127 + Scalar replacement of aggregates 8128 + Constant propagation 8129 + Value range propagation 8130 + Partial redundancy elimination 8131 + Load and store motion 8132 + Strength reduction 8133 + Dead store elimination 8134 + Dead and unreachable code elimination 8135 + [4]Autovectorization 8136 + Loop interchange 8137 + Tail recursion by accumulation 8138 Many of these passes outperform their counterparts from previous 8139 GCC releases. 8140 * [5]Swing Modulo Scheduling (SMS). An RTL level instruction 8141 scheduling optimization intended for loops that perform heavy 8142 computations. 8143 8144 New Languages and Language specific improvements 8145 8146 C family 8147 8148 * The sentinel attribute has been added to GCC. This function 8149 attribute allows GCC to warn when variadic functions such as execl 8150 are not NULL terminated. See the GCC manual for a complete 8151 description of its behavior. 8152 * Given __attribute__((alias("target"))) it is now an error if target 8153 is not a symbol, defined in the same translation unit. This also 8154 applies to aliases created by #pragma weak alias=target. This is 8155 because it's meaningless to define an alias to an undefined symbol. 8156 On Solaris, the native assembler would have caught this error, but 8157 GNU as does not. 8158 8159 C and Objective-C 8160 8161 * The -Wstrict-aliasing=2 option has been added. This warning catches 8162 all unsafe cases, but it may also give a warning for some cases 8163 that are safe. 8164 * The cast-as-lvalue, conditional-expression-as-lvalue and 8165 compound-expression-as-lvalue extensions, which were deprecated in 8166 3.3.4 and 3.4, have been removed. 8167 * The -fwritable-strings option, which was deprecated in 3.4, has 8168 been removed. 8169 * #pragma pack() semantics have been brought closer to those used by 8170 other compilers. This also applies to C++. 8171 * Taking the address of a variable with register storage is invalid 8172 in C. GCC now issues an error instead of a warning. 8173 * Arrays of incomplete element type are invalid in C. GCC now issues 8174 an error for such arrays. Declarations such as extern struct s x[]; 8175 (where struct s has not been defined) can be moved after the 8176 definition of struct s. Function parameters declared as arrays of 8177 incomplete type can instead be declared as pointers. 8178 8179 C++ 8180 8181 * When compiling without optimizations (-O0), the C++ frontend is 8182 much faster than in any previous versions of GCC. Independent 8183 testers have measured speed-ups up to 25% in real-world production 8184 code, compared to the 3.4 family (which was already the fastest 8185 version to date). Upgrading from older versions might show even 8186 bigger improvements. 8187 * ELF visibility attributes can now be applied to a class type, so 8188 that it affects every member function of a class at once, without 8189 having to specify each individually: 8190 class __attribute__ ((visibility("hidden"))) Foo 8191 { 8192 int foo1(); 8193 void foo2(); 8194 }; 8195 The syntax is deliberately similar to the __declspec() system used 8196 by Microsoft Windows based compilers, allowing cross-platform 8197 projects to easily reuse their existing macro system for denoting 8198 exports and imports. By explicitly marking internal classes never 8199 used outside a binary as hidden, one can completely avoid PLT 8200 indirection overheads during their usage by the compiler. You can 8201 find out more about the advantages of this at 8202 [6]http://www.akkadia.org/drepper/dsohowto.pdf 8203 * The -fvisibility-inlines-hidden option has been added which marks 8204 all inlineable functions as having hidden ELF visibility, thus 8205 removing their symbol and typeinfo from the exported symbol table 8206 of the output ELF binary. Using this option can reduce the exported 8207 symbol count of template-heavy code by up to 40% with no code 8208 change at all, thus notably improving link and load times for the 8209 binary as well as a reduction in size of up to 10%. Also, check the 8210 new [7]-fvisibility option. 8211 * The compiler now uses the library interface specified by the [8]C++ 8212 ABI for thread-safe initialization of function-scope static 8213 variables. Most users should leave this alone, but embedded 8214 programmers may want to disable this by specifying 8215 -fno-threadsafe-statics for a small savings in code size. 8216 * Taking the address of an explicit register variable is no longer 8217 supported. Note that C++ allows taking the address of variables 8218 with register storage so this will continue to compile with a 8219 warning. For example, assuming that r0 is a machine register: 8220 register int foo asm ("r0"); 8221 register int bar; 8222 &foo; // error, no longer accepted 8223 &bar; // OK, with a warning 8224 * G++ has an undocumented extension to virtual function covariancy 8225 rules that allowed the overrider to return a type that was 8226 implicitly convertable to the overridden function's return type. 8227 For instance a function returning void * could be overridden by a 8228 function returning T *. This is now deprecated and will be removed 8229 in a future release. 8230 * The G++ minimum and maximum operators (<? and >?) and their 8231 compound forms (<?=) and >?=) have been deprecated and will be 8232 removed in a future version. Code using these operators should be 8233 modified to use std::min and std::max instead. 8234 * Declaration of nested classes of class templates as friends are 8235 supported: 8236 template <typename T> struct A { 8237 class B {}; 8238 }; 8239 class C { 8240 template <typename T> friend class A<T>::B; 8241 }; 8242 This complements the feature member functions of class templates as 8243 friends introduced in GCC 3.4.0. 8244 * When declaring a friend class using an unqualified name, classes 8245 outside the innermost non-class scope are not searched: 8246 class A; 8247 namespace N { 8248 class B { 8249 friend class A; // Refer to N::A which has not been declared yet 8250 // because name outside namespace N are not searched 8251 friend class ::A; // Refer to ::A 8252 }; 8253 } 8254 Hiding the friend name until declaration is still not implemented. 8255 * Friends of classes defined outside their namespace are correctly 8256 handled: 8257 namespace N { 8258 class A; 8259 } 8260 class N::A { 8261 friend class B; // Refer to N::B in GCC 4.0.0 8262 // but ::B in earlier versions of GCC 8263 }; 8264 8265 Runtime Library (libstdc++) 8266 8267 * Optimization work: 8268 + Added efficient specializations of istream functions for char 8269 and wchar_t. 8270 + Further performance tuning of strings, in particular wrt 8271 single-char append and getline. 8272 + iter_swap - and therefore most of the mutating algorithms - 8273 now makes an unqualified call to swap when the value_type of 8274 the two iterators is the same. 8275 * A large subset of the features in Technical Report 1 (TR1 for 8276 short) is experimentally delivered (i.e., no guarantees about the 8277 implementation are provided. In particular it is not promised that 8278 the library will remain link-compatible when code using TR1 is 8279 used): 8280 + General utilities such as reference_wrapper and shared_ptr. 8281 + Function objects, i.e., result_of, mem_fn, bind, function. 8282 + Support for metaprogramming. 8283 + New containers such as tuple, array, unordered_set, 8284 unordered_map, unordered_multiset, unordered_multimap. 8285 * As usual, many bugs have been fixed and LWG resolutions implemented 8286 for the first time (e.g., DR 409). 8287 8288 Java 8289 8290 * In order to prevent naming conflicts with other implementations of 8291 these tools, some GCJ binaries have been renamed: 8292 + rmic is now grmic, 8293 + rmiregistry is now grmiregistry, and 8294 + jar is now fastjar. 8295 In particular, these names were problematic for the jpackage.org 8296 packaging conventions which install symlinks in /usr/bin that point 8297 to the preferred versions of these tools. 8298 * The -findirect-dispatch argument to the compiler now works and 8299 generates code following a new "binary compatibility" ABI. Code 8300 compiled this way follows the binary compatibility rules of the 8301 Java Language Specification. 8302 * libgcj now has support for using GCJ as a JIT, using the 8303 gnu.gcj.jit family of system properties. 8304 * libgcj can now find a shared library corresponding to the bytecode 8305 representation of a class. See the documentation for the new 8306 gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system 8307 property. 8308 * There have been many improvements to the class library. Here are 8309 some highlights: 8310 + Much more of AWT and Swing exist. 8311 + Many new packages and classes were added, including 8312 java.util.regex, java.net.URI, javax.crypto, 8313 javax.crypto.interfaces, javax.crypto.spec, javax.net, 8314 javax.net.ssl, javax.security.auth, 8315 javax.security.auth.callback, javax.security.auth.login, 8316 javax.security.auth.x500, javax.security.sasl, org.ietf.jgss, 8317 javax.imageio, javax.imageio.event, javax.imageio.spi, 8318 javax.print, javax.print.attribute, 8319 javax.print.attribute.standard, javax.print.event, and 8320 javax.xml 8321 + Updated SAX and DOM, and imported GNU JAXP 8322 8323 Fortran 8324 8325 * A new [9]Fortran front end has replaced the aging GNU Fortran 77 8326 front end. The new front end supports Fortran 90 and Fortran 95. It 8327 may not yet be as stable as the old Fortran front end. 8328 8329 Ada 8330 8331 * Ada (with tasking and Zero Cost Exceptions) is now available on 8332 many more targets, including but not limited to: alpha-linux, 8333 hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux, 8334 s390x-linux, sparc-linux. 8335 * Some of the new Ada 2005 features are now implemented like 8336 Wide_Wide_Character and Ada.Containers. 8337 * Many bugs have been fixed, tools and documentation improved. 8338 * To compile Ada from the sources, install an older working Ada 8339 compiler and then use --enable-languages=ada at configuration time, 8340 since the Ada frontend is not currently activated by default. See 8341 the [10]Installing GCC for details. 8342 8343 New Targets and Target Specific Improvements 8344 8345 H8/300 8346 8347 * The frame layout has changed. In the new layout, the prologue of a 8348 function first saves registers and then allocate space for locals, 8349 resulting in an 1% improvement on code size. 8350 8351 IA-32/x86-64 (AMD64) 8352 8353 * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10, 8354 log1p, log2, logb and tan mathematical builtins (and their float 8355 and long double variants) are now implemented as inline x87 8356 intrinsics when using -ffast-math. 8357 * The ceil, floor, nearbyint, rint and trunc mathematical builtins 8358 (and their float and long double variants) are now implemented as 8359 inline x87 intrinsics when using -ffast-math. 8360 * The x87's fsincos instruction is now used automatically with 8361 -ffast-math when calculating both the sin and cos of the same 8362 argument. 8363 * Instruction selection for multiplication and division by constants 8364 has been improved. 8365 8366 IA-64 8367 8368 * Floating point division, integer division and sqrt are now inlined, 8369 resulting in significant performance improvements on some codes. 8370 8371 MIPS 8372 8373 * Division by zero checks now use conditional traps if the target 8374 processor supports them. This decreases code size by one word per 8375 division operation. The old behavior (branch and break) can be 8376 obtained either at configure time by passing --with-divide=breaks 8377 to configure or at runtime by passing -mdivide-breaks to GCC. 8378 * Support for MIPS64 paired-single instructions has been added. It is 8379 enabled by -mpaired-single and can be accessed using both the 8380 target-independent vector extensions and new MIPS-specific built-in 8381 functions. 8382 * Support for the MIPS-3D ASE has been added. It is enabled by 8383 -mips3d and provides new MIPS-3D-specific built-in functions. 8384 * The -mexplicit-relocs option now supports static n64 code (as is 8385 used, for example, in 64-bit linux kernels). -mexplicit-relocs 8386 should now be feature-complete and is enabled by default when GCC 8387 is configured to use a compatible assembler. 8388 * Support for the NEC VR4130 series has been added. This support 8389 includes the use of VR-specific instructions and a new VR4130 8390 scheduler. Full VR4130 support can be selected with -march=vr4130 8391 while code for any ISA can be tuned for the VR4130 using 8392 -mtune=vr4130. There is also a new -mvr4130-align option that 8393 produces better schedules at the cost of increased code size. 8394 * Support for the Broadcom SB-1 has been extended. There is now an 8395 SB-1 scheduler as well as support for the SB-1-specific 8396 paired-single instructions. Full SB-1 support can be selected with 8397 -march=sb1 while code for any ISA can be optimized for the SB-1 8398 using -mtune=sb1. 8399 * The compiler can now work around errata in R4000, R4400, VR4120 and 8400 VR4130 processors. These workarounds are enabled by -mfix-r4000, 8401 -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120 8402 and VR4130 workarounds need binutils 2.16 or above. 8403 * IRIX shared libraries are now installed into the standard library 8404 directories: o32 libraries go into lib/, n32 libraries go into 8405 lib32/ and n64 libraries go into lib64/. 8406 * The compiler supports a new -msym32 option. It can be used to 8407 optimize n64 code in which all symbols are known to have 32-bit 8408 values. 8409 8410 S/390 and zSeries 8411 8412 * New command-line options help to generate code intended to run in 8413 an environment where stack space is restricted, e.g. Linux kernel 8414 code: 8415 + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time 8416 warnings for single functions that require large or dynamic 8417 stack frames. 8418 + -mstack-size and -mstack-guard generate code that checks for 8419 stack overflow at run time. 8420 + -mpacked-stack generates code that reduces the stack frame 8421 size of many functions by reusing unneeded parts of the stack 8422 bias area. 8423 * The -msoft-float option now ensures that generated code never 8424 accesses floating point registers. 8425 * The s390x-ibm-tpf target now fully supports C++, including 8426 exceptions and threads. 8427 * Various changes to improve performance of the generated code have 8428 been implemented, including: 8429 + GCC now uses sibling calls where possible. 8430 + Condition code handling has been optimized, allowing GCC to 8431 omit redundant comparisons in certain cases. 8432 + The cost function guiding many optimizations has been refined 8433 to more accurately represent the z900 and z990 processors. 8434 + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW 8435 instructions are now used to avoid conditional branches in 8436 certain cases. 8437 + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to 8438 optimize address arithmetic required to access large stack 8439 frames. 8440 + GCC now makes more efficient use of memory-to-memory type 8441 instructions (MVC, CLC, ...). 8442 + More precise tracking of special register use allows better 8443 instruction scheduling, in particular of the function prologue 8444 and epilogue sequences. 8445 + The Java front end now generates inline code to implement 8446 integer division, instead of calling library routines. 8447 8448 SPARC 8449 8450 * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and 8451 -mf934 have been removed. They have been replaced with -mcpu=xxx. 8452 * The internal model used to estimate the relative cost of each 8453 instruction has been updated. It is expected to give better results 8454 on recent UltraSPARC processors. 8455 * Code generation for function prologues and epilogues has been 8456 improved, resulting in better scheduling and allowing multiple exit 8457 points in functions. 8458 * Support for Sun's Visual Instruction Set (VIS) has been enhanced. 8459 It is enabled by -mvis and provides new built-in functions for VIS 8460 instructions on UltraSPARC processors. 8461 * The option -mapp-regs has been turned on by default on Solaris too. 8462 8463 NetWare 8464 8465 * Novell NetWare (on ix86, no other hardware platform was ever really 8466 supported by this OS) has been re-enabled and the ABI supported by 8467 GCC has been brought into sync with that of MetroWerks CodeWarrior 8468 (the ABI previously supported was that of some Unix systems, which 8469 NetWare never tried to support). 8470 8471 Obsolete Systems 8472 8473 Support for a number of older systems has been declared obsolete in GCC 8474 4.0. Unless there is activity to revive them, the next release of GCC 8475 will have their sources permanently removed. 8476 8477 All GCC ports for the following processor architectures have been 8478 declared obsolete: 8479 * Intel i860 8480 * Ubicom IP2022 8481 * National Semiconductor NS32K 8482 * Texas Instruments TMS320C[34]x 8483 8484 Also, those for some individual systems have been obsoleted: 8485 * SPARC family 8486 + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf, 8487 sparc86x-*-elf) 8488 + OpenBSD 32-bit (sparc-*-openbsd*) 8489 8490 Documentation improvements 8491 8492 Other significant improvements 8493 8494 * Location lists are now generated by default when compiling with 8495 debug info and optimization. Location lists provide more accurate 8496 debug info about locations of variables and they allow debugging 8497 code compiled with -fomit-frame-pointer. 8498 * The -fvisibility option has been added which allows the default ELF 8499 visibility of all symbols to be set per compilation and the new 8500 #pragma GCC visibility preprocessor command allows the setting of 8501 default ELF visibility for a region of code. Using 8502 -fvisibility=hidden especially in combination with the new 8503 -fvisibility-inlines-hidden can yield substantial improvements in 8504 output binary quality including avoiding PLT indirection overheads, 8505 reduction of the exported symbol count by up to 60% (with resultant 8506 improvements to link and load times), better scope for the 8507 optimizer to improve code and up to a 20% reduction in binary size. 8508 Using these options correctly yields a binary with a similar symbol 8509 count to a Windows DLL. 8510 Perhaps more importantly, this new feature finally allows (with 8511 careful planning) complete avoidance of symbol clashes when 8512 manually loading shared objects with RTLD_GLOBAL, thus finally 8513 solving problems many projects such as python were forced to use 8514 RTLD_LOCAL for (with its resulting issues for C++ correctness). You 8515 can find more information about using these options at 8516 [11]https://gcc.gnu.org/wiki/Visibility. 8517 __________________________________________________________________ 8518 8519 GCC 4.0.1 8520 8521 This is the [12]list of problem reports (PRs) from GCC's bug tracking 8522 system that are known to be fixed in the 4.0.1 release. This list might 8523 not be complete (that is, it is possible that some PRs that have been 8524 fixed are not listed here). 8525 8526 GCC 4.0.2 8527 8528 This is the [13]list of problem reports (PRs) from GCC's bug tracking 8529 system that are known to be fixed in the 4.0.2 release. This list might 8530 not be complete (that is, it is possible that some PRs that have been 8531 fixed are not listed here). 8532 8533 Unfortunately, due to a release engineering failure, this release has a 8534 regression on Solaris that will affect some C++ programs. We suggest 8535 that Solaris users apply a [14]patch that corrects the problem. Users 8536 who do not wish to apply the patch should explicitly link C++ programs 8537 with the -pthreads option, even if they do not use threads. This 8538 problem has been corrected in the current 4.0 branch sources and will 8539 not be present in GCC 4.0.3. 8540 8541 GCC 4.0.3 8542 8543 Starting with this release, the function getcontext is recognized by 8544 the compiler as having the same semantics as the setjmp function. In 8545 particular, the compiler will ensure that all registers are dead before 8546 calling such a function and will emit a warning about the variables 8547 that may be clobbered after the second return from the function. 8548 8549 GCC 4.0.4 8550 8551 This is the [15]list of problem reports (PRs) from GCC's bug tracking 8552 system that are known to be fixed in the 4.0.4 release. This list might 8553 not be complete (that is, it is possible that some PRs that have been 8554 fixed are not listed here). 8555 8556 The 4.0.4 release is provided for those that require a high degree of 8557 binary compatibility with previous 4.0.x releases. For most users, the 8558 GCC team recommends that version 4.1.1 or later be used instead." 8559 8560 8561 For questions related to the use of GCC, please consult these web 8562 pages and the [16]GCC manuals. If that fails, the 8563 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8564 web pages and the development of GCC are welcome on our developer 8565 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 8566 archives. 8567 8568 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 8569 distribution of this entire article is permitted in any medium, 8570 provided this notice is preserved. 8571 8572 These pages are [21]maintained by the GCC team. Last modified 8573 2015-06-27[22]. 8574 8575 References 8576 8577 1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 8578 2. http://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html 8579 3. http://gcc.gnu.org/projects/tree-ssa/ 8580 4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html 8581 5. http://gcc.gnu.org/news/sms.html 8582 6. http://www.akkadia.org/drepper/dsohowto.pdf 8583 7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility 8584 8. http://mentorembedded.github.io/cxx-abi/ 8585 9. http://gcc.gnu.org/fortran/ 8586 10. https://gcc.gnu.org/install/ 8587 11. https://gcc.gnu.org/wiki/Visibility 8588 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1 8589 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2 8590 14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html 8591 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4 8592 16. https://gcc.gnu.org/onlinedocs/ 8593 17. mailto:gcc-help (a] gcc.gnu.org 8594 18. mailto:gcc (a] gcc.gnu.org 8595 19. https://gcc.gnu.org/lists.html 8596 20. http://www.fsf.org/ 8597 21. https://gcc.gnu.org/about.html 8598 22. http://validator.w3.org/check/referer 8599 ====================================================================== 8600 http://gcc.gnu.org/gcc-3.4/index.html 8601 8602 GCC 3.4 Release Series 8603 8604 May 26, 2006 8605 8606 The [1]GNU project and the GCC developers are pleased to announce the 8607 release of GCC 3.4.6. 8608 8609 This release is a bug-fix release, containing fixes for regressions in 8610 GCC 3.4.4 relative to previous releases of GCC. This is the last of the 8611 3.4.x series. 8612 8613 The GCC 3.4 release series includes numerous [2]new features, 8614 improvements, bug fixes, and other changes, thanks to an [3]amazing 8615 group of volunteers. 8616 8617 Release History 8618 8619 GCC 3.4.6 8620 March 6, 2006 ([4]changes) 8621 8622 GCC 3.4.5 8623 November 30, 2005 ([5]changes) 8624 8625 GCC 3.4.4 8626 May 18, 2005 ([6]changes) 8627 8628 GCC 3.4.3 8629 November 4, 2004 ([7]changes) 8630 8631 GCC 3.4.2 8632 September 6, 2004 ([8]changes) 8633 8634 GCC 3.4.1 8635 July 1, 2004 ([9]changes) 8636 8637 GCC 3.4.0 8638 April 18, 2004 ([10]changes) 8639 8640 References and Acknowledgements 8641 8642 GCC used to stand for the GNU C Compiler, but since the compiler 8643 supports several other languages aside from C, it now stands for the 8644 GNU Compiler Collection. 8645 8646 A list of [11]successful builds is updated as new information becomes 8647 available. 8648 8649 The GCC developers would like to thank the numerous people that have 8650 contributed new features, improvements, bug fixes, and other changes as 8651 well as test results to GCC. This [12]amazing group of volunteers is 8652 what makes GCC successful. 8653 8654 For additional information about GCC please refer to the [13]GCC 8655 project web site or contact the [14]GCC development mailing list. 8656 8657 To obtain GCC please use [15]our mirror sites, or [16]our SVN server. 8658 8659 8660 For questions related to the use of GCC, please consult these web 8661 pages and the [17]GCC manuals. If that fails, the 8662 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8663 web pages and the development of GCC are welcome on our developer 8664 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 8665 archives. 8666 8667 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 8668 distribution of this entire article is permitted in any medium, 8669 provided this notice is preserved. 8670 8671 These pages are [22]maintained by the GCC team. Last modified 8672 2014-06-28[23]. 8673 8674 References 8675 8676 1. http://www.gnu.org/ 8677 2. http://gcc.gnu.org/gcc-3.4/changes.html 8678 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8679 4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 8680 5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5 8681 6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4 8682 7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3 8683 8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2 8684 9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1 8685 10. http://gcc.gnu.org/gcc-3.4/changes.html 8686 11. http://gcc.gnu.org/gcc-3.4/buildstat.html 8687 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8688 13. http://gcc.gnu.org/index.html 8689 14. mailto:gcc (a] gcc.gnu.org 8690 15. http://gcc.gnu.org/mirrors.html 8691 16. http://gcc.gnu.org/svn.html 8692 17. https://gcc.gnu.org/onlinedocs/ 8693 18. mailto:gcc-help (a] gcc.gnu.org 8694 19. mailto:gcc (a] gcc.gnu.org 8695 20. https://gcc.gnu.org/lists.html 8696 21. http://www.fsf.org/ 8697 22. https://gcc.gnu.org/about.html 8698 23. http://validator.w3.org/check/referer 8699 ====================================================================== 8700 http://gcc.gnu.org/gcc-3.4/changes.html 8701 8702 GCC 3.4 Release Series 8703 Changes, New Features, and Fixes 8704 8705 The final release in the 3.4 release series is [1]GCC 3.4.6. The series 8706 is now closed. 8707 8708 GCC 3.4 has [2]many improvements in the C++ frontend. Before reporting 8709 a bug, please make sure it's really GCC, and not your code, that is 8710 broken. 8711 8712 Caveats 8713 8714 * GNU Make is now required to build GCC. 8715 * With -nostdinc the preprocessor used to ignore both standard 8716 include paths and include paths contained in environment variables. 8717 It was neither documented nor intended that environment variable 8718 paths be ignored, so this has been corrected. 8719 * GCC no longer accepts the options -fvolatile, -fvolatile-global and 8720 -fvolatile-static. It is unlikely that they worked correctly in any 8721 3.x release. 8722 * GCC no longer ships <varargs.h>. Use <stdarg.h> instead. 8723 * Support for all the systems [3]obsoleted in GCC 3.3 has been 8724 removed from GCC 3.4. See below for a [4]list of systems which are 8725 obsoleted in this release. 8726 * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C 8727 compilers will not work. 8728 * The implementation of the [5]MIPS ABIs has changed. As a result, 8729 the code generated for certain MIPS targets will not be binary 8730 compatible with earlier releases. 8731 * In previous releases, the MIPS port had a fake "hilo" register with 8732 the user-visible name accum. This register has been removed. 8733 * The implementation of the [6]SPARC ABIs has changed. As a result, 8734 the code generated will not be binary compatible with earlier 8735 releases in certain cases. 8736 * The configure option --enable-threads=pthreads has been removed; 8737 use --enable-threads=posix instead, which should have the same 8738 effect. 8739 * Code size estimates used by inlining heuristics for C, Objective-C, 8740 C++ and Java have been redesigned significantly. As a result the 8741 parameters of -finline-insns, --param max-inline-insns-single and 8742 --param max-inline-insns-auto need to be reconsidered. 8743 * --param max-inline-slope and --param min-inline-insns have been 8744 removed; they are not needed for the new bottom-up inlining 8745 heuristics. 8746 * The new unit-at-a-time compilation scheme has several compatibility 8747 issues: 8748 + The order in which functions, variables, and top-level asm 8749 statements are emitted may have changed. Code relying on some 8750 particular ordering needs to be updated. The majority of such 8751 top-level asm statements can be replaced by section 8752 attributes. 8753 + Unreferenced static variables and functions are removed. This 8754 may result in undefined references when an asm statement 8755 refers to the variable/function directly. In that case either 8756 the variable/function shall be listed in asm statement operand 8757 or in the case of top-level asm statements the attribute used 8758 shall be used to force function/variable to be always output 8759 and considered as a possibly used by unknown code. 8760 For variables the attribute is accepted only by GCC 3.4 and 8761 newer, while for earlier versions it is sufficient to use 8762 unused to silence warnings about the variables not being 8763 referenced. To keep code portable across different GCC 8764 versions, you can use appropriate preprocessor conditionals. 8765 + Static functions now can use non-standard passing conventions 8766 that may break asm statements calling functions directly. 8767 Again the attribute used shall be used to prevent this 8768 behavior. 8769 As a temporary workaround, -fno-unit-at-a-time can be used, but 8770 this scheme may not be supported by future releases of GCC. 8771 * GCC 3.4 automatically places zero-initialized variables in the .bss 8772 section on some operating systems. Versions of GNU Emacs up to (and 8773 including) 21.3 will not work correctly when using this 8774 optimization; you can use -fno-zero-initialized-in-bss to disable 8775 it. 8776 * If GCC 3.4 is configured with --enable-threads=posix (the default 8777 on most targets that support pthreads) then _REENTRANT will be 8778 defined unconditionally by some libstdc++ headers. C++ code which 8779 relies on that macro to detect whether multi-threaded code is being 8780 compiled might change in meaning, possibly resulting in linker 8781 errors for single-threaded programs. Affected users of [7]Boost 8782 should compile single-threaded code with -DBOOST_DISABLE_THREADS. 8783 See Bugzilla for [8]more information. 8784 8785 General Optimizer Improvements 8786 8787 * Usability of the profile feedback and coverage testing has been 8788 improved. 8789 + Performance of profiled programs has been improved by faster 8790 profile merging code. 8791 + Better use of the profile feedback for optimization (loop 8792 unrolling and loop peeling). 8793 + File locking support allowing fork() calls and parallel runs 8794 of profiled programs. 8795 + Coverage file format has been redesigned. 8796 + gcov coverage tool has been improved. 8797 + make profiledbootstrap available to build a faster compiler. 8798 Experiments made on i386 hardware showed an 11% speedup on -O0 8799 and a 7.5% speedup on -O2 compilation of a [9]large C++ 8800 testcase. 8801 + New value profiling pass enabled via -fprofile-values 8802 + New value profile transformations pass enabled via -fvpt aims 8803 to optimize some code sequences by exploiting knowledge about 8804 value ranges or other properties of the operands. At the 8805 moment a conversion of expensive divisions into cheaper 8806 operations has been implemented. 8807 + New -fprofile-generate and -fprofile-use command-line options 8808 to simplify the use of profile feedback. 8809 * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and 8810 Java which is enabled via -funit-at-a-time (and implied by -O2). In 8811 this scheme a whole file is parsed first and optimized later. The 8812 following basic inter-procedural optimizations are implemented: 8813 + Removal of unreachable functions and variables 8814 + Discovery of local functions (functions with static linkage 8815 whose address is never taken) 8816 + On i386, these local functions use register parameter passing 8817 conventions. 8818 + Reordering of functions in topological order of the call graph 8819 to enable better propagation of optimizing hints (such as the 8820 stack alignments needed by functions) in the back end. 8821 + Call graph based out-of-order inlining heuristics which allows 8822 to limit overall compilation unit growth (--param 8823 inline-unit-growth). 8824 Overall, the unit-at-a-time scheme produces a 1.3% improvement for 8825 the SPECint2000 benchmark on the i386 architecture (AMD Athlon 8826 CPU). 8827 * More realistic code size estimates used by inlining for C, 8828 Objective-C, C++ and Java. The growth of large functions can now be 8829 limited via --param large-function-insns and --param 8830 large-function-growth. 8831 * A new cfg-level loop optimizer pass replaces the old loop unrolling 8832 pass and adds two other loop transformations -- loop peeling and 8833 loop unswitching -- and also uses the profile feedback to limit 8834 code growth. (The three optimizations are enabled by 8835 -funroll-loops, -fpeel-loops and -funswitch-loops flags, 8836 respectively). 8837 The old loop unroller still can be enabled by -fold-unroll-loops 8838 and may produce better code in some cases, especially when the 8839 webizer optimization pass is not run. 8840 * A new web construction pass enabled via -fweb (and implied by -O3) 8841 improves the quality of register allocation, CSE, first scheduling 8842 pass and some other optimization passes by avoiding re-use of 8843 pseudo registers with non-overlapping live ranges. The pass almost 8844 always improves code quality but does make debugging difficult and 8845 thus is not enabled by default by -O2 8846 The pass is especially effective as cleanup after code duplication 8847 passes, such as the loop unroller or the tracer. 8848 * Experimental implementations of superblock or trace scheduling in 8849 the second scheduling pass can be enabled via 8850 -fsched2-use-superblocks and -fsched2-use-traces, respectively. 8851 8852 New Languages and Language specific improvements 8853 8854 Ada 8855 8856 * The Ada front end has been updated to include numerous bug fixes 8857 and enhancements. These include: 8858 + Improved project file support 8859 + Additional set of warnings about potential wrong code 8860 + Improved error messages 8861 + Improved code generation 8862 + Improved cross reference information 8863 + Improved inlining 8864 + Better run-time check elimination 8865 + Better error recovery 8866 + More efficient implementation of unbounded strings 8867 + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools, 8868 ... 8869 + New GNAT.xxxx packages (e.g. GNAT.Strings, 8870 GNAT.Exception_Action) 8871 + New pragmas 8872 + New -gnatS switch replacing gnatpsta 8873 + Implementation of new Ada features (in particular limited 8874 with, limited aggregates) 8875 8876 C/Objective-C/C++ 8877 8878 * Precompiled headers are now supported. Precompiled headers can 8879 dramatically speed up compilation of some projects. There are some 8880 known defects in the current precompiled header implementation that 8881 will result in compiler crashes in relatively rare situations. 8882 Therefore, precompiled headers should be considered a "technology 8883 preview" in this release. Read the manual for details about how to 8884 use precompiled headers. 8885 * File handling in the preprocessor has been rewritten. GCC no longer 8886 gets confused by symlinks and hardlinks, and now has a correct 8887 implementation of #import and #pragma once. These two directives 8888 have therefore been un-deprecated. 8889 * The undocumented extension that allowed C programs to have a label 8890 at the end of a compound statement, which has been deprecated since 8891 GCC 3.0, has been removed. 8892 * The cast-as-lvalue extension has been removed for C++ and 8893 deprecated for C and Objective-C. In particular, code like this: 8894 int i; 8895 (char) i = 5; 8896 8897 or this: 8898 char *p; 8899 ((int *) p)++; 8900 8901 is no longer accepted for C++ and will not be accepted for C and 8902 Objective-C in a future version. 8903 * The conditional-expression-as-lvalue extension has been deprecated 8904 for C and Objective-C. In particular, code like this: 8905 int a, b, c; 8906 (a ? b : c) = 2; 8907 8908 will not be accepted for C and Objective-C in a future version. 8909 * The compound-expression-as-lvalue extension has been deprecated for 8910 C and Objective-C. In particular, code like this: 8911 int a, b; 8912 (a, b) = 2; 8913 8914 will not be accepted for C and Objective-C in a future version. A 8915 possible non-intrusive workaround is the following: 8916 (*(a, &b)) = 2; 8917 8918 * Several [10]built-in functions such as __builtin_popcount for 8919 counting bits, finding the highest and lowest bit in a word, and 8920 parity have been added. 8921 * The -fwritable-strings option has been deprecated and will be 8922 removed. 8923 * Many C math library functions are now recognized as built-ins and 8924 optimized. 8925 * The C, C++, and Objective-C compilers can now handle source files 8926 written in any character encoding supported by the host C library. 8927 The default input character set is taken from the current locale, 8928 and may be overridden with the -finput-charset command line option. 8929 In the future we will add support for inline encoding markers. 8930 8931 C++ 8932 8933 * G++ is now much closer to full conformance to the ISO/ANSI C++ 8934 standard. This means, among other things, that a lot of invalid 8935 constructs which used to be accepted in previous versions will now 8936 be rejected. It is very likely that existing C++ code will need to 8937 be fixed. This document lists some of the most common issues. 8938 * A hand-written recursive-descent C++ parser has replaced the 8939 YACC-derived C++ parser from previous GCC releases. The new parser 8940 contains much improved infrastructure needed for better parsing of 8941 C++ source codes, handling of extensions, and clean separation 8942 (where possible) between proper semantics analysis and parsing. The 8943 new parser fixes many bugs that were found in the old parser. 8944 * You must now use the typename and template keywords to disambiguate 8945 dependent names, as required by the C++ standard. 8946 struct K { 8947 typedef int mytype_t; 8948 }; 8949 8950 template <class T1> struct A { 8951 template <class T2> struct B { 8952 void callme(void); 8953 }; 8954 8955 template <int N> void bar(void) 8956 { 8957 // Use 'typename' to tell the parser that T1::mytype_t names 8958 // a type. This is needed because the name is dependent (in 8959 // this case, on template parameter T1). 8960 typename T1::mytype_t x; 8961 x = 0; 8962 } 8963 }; 8964 8965 template <class T> void template_func(void) 8966 { 8967 // Use 'template' to prefix member templates within 8968 // dependent types (a has type A<T>, which depends on 8969 // the template parameter T). 8970 A<T> a; 8971 a.template bar<0>(); 8972 8973 // Use 'template' to tell the parser that B is a nested 8974 // template class (dependent on template parameter T), and 8975 // 'typename' because the whole A<T>::B<int> is 8976 // the name of a type (again, dependent). 8977 typename A<T>::template B<int> b; 8978 b.callme(); 8979 } 8980 8981 void non_template_func(void) 8982 { 8983 // Outside of any template class or function, no names can be 8984 // dependent, so the use of the keyword 'typename' and 'template' 8985 // is not needed (and actually forbidden). 8986 A<K> a; 8987 a.bar<0>(); 8988 A<K>::B<float> b; 8989 b.callme(); 8990 } 8991 * In a template definition, unqualified names will no longer find 8992 members of a dependent base (as specified by [temp.dep]/3 in the 8993 C++ standard). For example, 8994 template <typename T> struct B { 8995 int m; 8996 int n; 8997 int f (); 8998 int g (); 8999 }; 9000 int n; 9001 int g (); 9002 template <typename T> struct C : B<T> { 9003 void h () 9004 { 9005 m = 0; // error 9006 f (); // error 9007 n = 0; // ::n is modified 9008 g (); // ::g is called 9009 } 9010 }; 9011 You must make the names dependent, e.g. by prefixing them with 9012 this->. Here is the corrected definition of C<T>::h, 9013 template <typename T> void C<T>::h () 9014 { 9015 this->m = 0; 9016 this->f (); 9017 this->n = 0 9018 this->g (); 9019 } 9020 As an alternative solution (unfortunately not backwards compatible 9021 with GCC 3.3), you may use using declarations instead of this->: 9022 template <typename T> struct C : B<T> { 9023 using B<T>::m; 9024 using B<T>::f; 9025 using B<T>::n; 9026 using B<T>::g; 9027 void h () 9028 { 9029 m = 0; 9030 f (); 9031 n = 0; 9032 g (); 9033 } 9034 }; 9035 * In templates, all non-dependent names are now looked up and bound 9036 at definition time (while parsing the code), instead of later when 9037 the template is instantiated. For instance: 9038 void foo(int); 9039 9040 template <int> struct A { 9041 static void bar(void){ 9042 foo('a'); 9043 } 9044 }; 9045 9046 void foo(char); 9047 9048 int main() 9049 { 9050 A<0>::bar(); // Calls foo(int), used to call foo(char). 9051 } 9052 9053 * In an explicit instantiation of a class template, you must use 9054 class or struct before the template-id: 9055 template <int N> 9056 class A {}; 9057 9058 template A<0>; // error, not accepted anymore 9059 template class A<0>; // OK 9060 * The "named return value" and "implicit typename" extensions have 9061 been removed. 9062 * Default arguments in function types have been deprecated and will 9063 be removed. 9064 * ARM-style name-injection of friend declarations has been deprecated 9065 and will be removed. For example: struct S { friend void f(); }; 9066 void g() { f(); } will not be accepted by future versions of G++; 9067 instead a declaration of "f" will need to be present outside of the 9068 scope of "S". 9069 * Covariant returns are implemented for all but varadic functions 9070 that require an adjustment. 9071 * When -pedantic is used, G++ now issues errors about spurious 9072 semicolons. For example, 9073 namespace N {}; // Invalid semicolon. 9074 void f() {}; // Invalid semicolon. 9075 * G++ no longer accepts attributes for a declarator after the 9076 initializer associated with that declarator. For example, 9077 X x(1) __attribute__((...)); 9078 is no longer accepted. Instead, use: 9079 X x __attribute__((...)) (1); 9080 * Inside the scope of a template class, the name of the class itself 9081 can be treated as either a class or a template. So GCC used to 9082 accept the class name as argument of type template, and template 9083 template parameter. However this is not C++ standard compliant. Now 9084 the name is not treated as a valid template template argument 9085 unless you qualify the name by its scope. For example, the code 9086 below no longer compiles. 9087 template <template <class> class TT> class X {}; 9088 template <class T> class Y { 9089 X<Y> x; // Invalid, Y is always a type template parameter. 9090 }; 9091 The valid code for the above example is 9092 X< ::Y> x; // Valid. 9093 (Notice the space between < and : to prevent GCC to interpret this 9094 as a digraph for [.) 9095 * Friend declarations that refer to template specializations are 9096 rejected if the template has not already been declared. For 9097 example, 9098 template <typename T> 9099 class C { 9100 friend void f<> (C&); 9101 }; 9102 is rejected. You must first declare f as a template, 9103 template <typename T> 9104 void f(T); 9105 * In case of friend declarations, every name used in the friend 9106 declaration must be accessible at the point of that declaration. 9107 Previous versions of G++ used to be less strict about this and 9108 allowed friend declarations for private class members, for example. 9109 See the ISO C++ Standard Committee's [11]defect report #209 for 9110 details. 9111 * Declaration of member functions of class templates as friends are 9112 supported. For example, 9113 template <typename T> struct A { 9114 void f(); 9115 }; 9116 class C { 9117 template <typename T> friend void A<T>::f(); 9118 }; 9119 * You must use template <> to introduce template specializations, as 9120 required by the standard. For example, 9121 template <typename T> 9122 struct S; 9123 9124 struct S<int> { }; 9125 is rejected. You must write, 9126 template <> struct S<int> {}; 9127 * G++ used to accept code like this, 9128 struct S { 9129 int h(); 9130 void f(int i = g()); 9131 int g(int i = h()); 9132 }; 9133 This behavior is not mandated by the standard. Now G++ issues an 9134 error about this code. To avoid the error, you must move the 9135 declaration of g before the declaration of f. The default arguments 9136 for g must be visible at the point where it is called. 9137 * The C++ ABI Section 3.3.3 specifications for the array construction 9138 routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return 9139 NULL when the allocator argument returns NULL. These changes are 9140 incorporated into the libstdc++ runtime library. 9141 * Using a name introduced by a typedef in a friend declaration or in 9142 an explicit instantiation is now rejected, as specified by the ISO 9143 C++ standard. 9144 class A; 9145 typedef A B; 9146 class C { 9147 friend class B; // error, no typedef name here 9148 friend B; // error, friend always needs class/struct/enum 9149 friend class A; // OK 9150 }; 9151 9152 template <int> class Q {}; 9153 typedef Q<0> R; 9154 template class R; // error, no typedef name here 9155 template class Q<0>; // OK 9156 * When allocating an array with a new expression, GCC used to allow 9157 parentheses around the type name. This is actually ill-formed and 9158 it is now rejected: 9159 int* a = new (int)[10]; // error, not accepted anymore 9160 int* a = new int[10]; // OK 9161 * When binding an rvalue of class type to a reference, the copy 9162 constructor of the class must be accessible. For instance, consider 9163 the following code: 9164 class A 9165 { 9166 public: 9167 A(); 9168 9169 private: 9170 A(const A&); // private copy ctor 9171 }; 9172 9173 A makeA(void); 9174 void foo(const A&); 9175 9176 void bar(void) 9177 { 9178 foo(A()); // error, copy ctor is not accessible 9179 foo(makeA()); // error, copy ctor is not accessible 9180 9181 A a1; 9182 foo(a1); // OK, a1 is a lvalue 9183 } 9184 This might be surprising at first sight, especially since most 9185 popular compilers do not correctly implement this rule ([12]further 9186 details). 9187 * When forming a pointer to member or a pointer to member function, 9188 access checks for class visibility (public, protected, private) are 9189 now performed using the qualifying scope of the name itself. This 9190 is better explained with an example: 9191 class A 9192 { 9193 public: 9194 void pub_func(); 9195 protected: 9196 void prot_func(); 9197 private: 9198 void priv_func(); 9199 }; 9200 9201 class B : public A 9202 { 9203 public: 9204 void foo() 9205 { 9206 &A::pub_func; // OK, pub_func is accessible through A 9207 &A::prot_func; // error, cannot access prot_func through A 9208 &A::priv_func; // error, cannot access priv_func through A 9209 9210 &B::pub_func; // OK, pub_func is accessible through B 9211 &B::prot_func; // OK, can access prot_func through B (within B) 9212 &B::priv_func; // error, cannot access priv_func through B 9213 } 9214 }; 9215 9216 Runtime Library (libstdc++) 9217 9218 * Optimization work: 9219 + Streamlined streambuf, filebuf, separate synched with C 9220 Standard I/O streambuf. 9221 + All formatted I/O now uses cached locale information. 9222 + STL optimizations (memory/speed for list, red-black trees as 9223 used by sets and maps). 9224 + More use of GCC builtins. 9225 + String optimizations (avoid contention on 9226 increment/decrement-and-test of the reference count in the 9227 empty-string object, constructor from input_iterators 9228 speedup). 9229 * Static linkage size reductions. 9230 * Large File Support (files larger than 2 GB on 32-bit systems). 9231 * Wide character and variable encoding filebuf work (UTF-8, Unicode). 9232 * Generic character traits. 9233 * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x, 9234 Solaris 2.7 and above, AIX 5.x, Irix 6.5. 9235 * The allocator class is now standard-conformant, and two additional 9236 extension allocators have been added, mt_alloc and 9237 bitmap_allocator. 9238 * PCH support: -include bits/stdc++.h (2x compile speedup). 9239 * Rewrote __cxa_demangle with support for C++ style allocators. 9240 * New debug modes for STL containers and iterators. 9241 * Testsuite rewrite: five times as many tests, plus increasingly 9242 sophisticated tests, including I/O, MT, multi-locale, wide and 9243 narrow characters. 9244 * Use current versions of GNU "autotools" for build/configuration. 9245 9246 Objective-C 9247 9248 * The Objective-C front end has been updated to include the numerous 9249 bug fixes and enhancements previously available only in Apple's 9250 version of GCC. These include: 9251 + Structured exception (@try... @catch... @finally, @throw) and 9252 synchronization (@synchronized) support. These are accessible 9253 via the -fobjc-exceptions switch; as of this writing, they may 9254 only be used in conjunction with -fnext-runtime on Mac OS X 9255 10.3 and later. See [13]Options Controlling Objective-C 9256 Dialect for more information. 9257 + An overhaul of @encode logic. The C99 _Bool and C++ bool type 9258 may now be encoded as 'B'. In addition, the back-end/codegen 9259 dependencies have been removed. 9260 + An overhaul of message dispatch construction, ensuring that 9261 the various receiver types (and casts thereof) are handled 9262 properly, and that correct diagnostics are issued. 9263 + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue" 9264 (-freplace-objc-classes) debugging modes, currently available 9265 on Mac OS X 10.3 and later. See [14]Options Controlling 9266 Objective-C Dialect for more information. 9267 + Access to optimized runtime entry points (-fno-nil-receivers ) 9268 on the assumption that message receivers are never nil. This 9269 is currently available on Mac OS X 10.3 and later. See 9270 [15]Options Controlling Objective-C Dialect for more 9271 information. 9272 9273 Java 9274 9275 * Compiling a .jar file will now cause non-.class entries to be 9276 automatically compiled as resources. 9277 * libgcj has been ported to Darwin. 9278 * Jeff Sturm has adapted Jan Hubicka's call graph optimization code 9279 to gcj. 9280 * libgcj has a new gcjlib URL type; this lets URLClassLoader load 9281 code from shared libraries. 9282 * libgcj has been much more completely merged with [16]GNU Classpath. 9283 * Class loading is now much more correct; in particular the caller's 9284 class loader is now used when that is required. 9285 * [17]Eclipse 2.x will run out of the box using gij. 9286 * Parts of java.nio have been implemented. Direct and indirect 9287 buffers work, as do fundamental file and socket operations. 9288 * java.awt has been improved, though it is still not ready for 9289 general use. 9290 * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST 9291 method. 9292 * The MinGW port has matured. Enhancements include socket timeout 9293 support, thread interruption, improved Runtime.exec() handling and 9294 support for accented characters in filenames. 9295 9296 Fortran 9297 9298 * Fortran improvements are listed in the [18]Fortran documentation. 9299 9300 New Targets and Target Specific Improvements 9301 9302 Alpha 9303 9304 * Several [19]built-in functions have been added such as 9305 __builtin_alpha_zap to allow utilizing the more obscure 9306 instructions of the CPU. 9307 * Parameter passing of complex arguments has changed to match the 9308 [20]ABI. This change is incompatible with previous GCC versions, 9309 but does fix compatibility with the Tru64 compiler and several 9310 corner cases where GCC was incompatible with itself. 9311 9312 ARM 9313 9314 * Nicolas Pitre has contributed his hand-coded floating-point support 9315 code for ARM. It is both significantly smaller and faster than the 9316 existing C-based implementation, even when building applications 9317 for Thumb. The arm-elf configuration has been converted to use the 9318 new code. 9319 * Support for the Intel's iWMMXt architecture, a second generation 9320 XScale processor, has been added. Enabled at run time with the 9321 -mcpu=iwmmxt command line switch. 9322 * A new ARM target has been added: arm-wince-pe. This is similar to 9323 the arm-pe target, but it defaults to using the APCS32 ABI. 9324 * The existing ARM pipeline description has been converted to the use 9325 the [21]DFA processor pipeline model. There is not much change in 9326 code performance, but the description is now [22]easier to 9327 understand. 9328 * Support for the Cirrus EP9312 Maverick floating point co-processor 9329 added. Enabled at run time with the -mcpu=ep9312 command line 9330 switch. Note however that the multilibs to support this chip are 9331 currently disabled in gcc/config/arm/t-arm-elf, so if you want to 9332 enable their production you will have to uncomment the entries in 9333 that file. 9334 9335 H8/300 9336 9337 * Support for long long has been added. 9338 * Support for saveall attribute has been added. 9339 * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code 9340 for H8/300H and H8S, which is much faster than the previous 9341 implementation. 9342 * A lot of small performance improvements. 9343 9344 IA-32/AMD64 (x86-64) 9345 9346 * Tuning for K8 (AMD Opteron/Athlon64) core is available via 9347 -march=k8 and -mcpu=k8. 9348 * Scalar SSE code generation carefully avoids reformatting penalties, 9349 hidden dependencies and minimizes the number of uops generated on 9350 both Intel and AMD CPUs. 9351 * Vector MMX and SSE operands are now passed in registers to improve 9352 performance and match the argument passing convention used by the 9353 Intel C++ Compiler. As a result it is not possible to call 9354 functions accepting vector arguments compiled by older GCC version. 9355 * Conditional jump elimination is now more aggressive on modern CPUs. 9356 * The Athlon ports has been converted to use the DFA processor 9357 pipeline description. 9358 * Optimization of indirect tail calls is now possible in a similar 9359 fashion as direct sibcall optimization. 9360 * Further small performance improvements. 9361 * -m128bit-long-double is now less buggy. 9362 * __float128 support in 64-bit compilation. 9363 * Support for data structures exceeding 2GB in 64-bit mode. 9364 * -mcpu has been renamed to -mtune. 9365 9366 IA-64 9367 9368 * Tuning code for the Itanium 2 processor has been added. The 9369 generation of code tuned for Itanium 2 (option -mtune=itanium2) is 9370 enabled by default now. To generate code tuned for Itanium 1 the 9371 option -mtune=itanium1 should be used. 9372 * [23]DFA processor pipeline descriptions for the IA-64 processors 9373 have been added. This resulted in about 3% improvement on the 9374 SPECInt2000 benchmark for Itanium 2. 9375 * Instruction bundling for the IA-64 processors has been rewritten 9376 using the DFA pipeline hazard recognizer. It resulted in about 60% 9377 compiler speedup on the SPECInt2000 C programs. 9378 9379 M32R 9380 9381 * Support for the M32R/2 processor has been added by Renesas. 9382 * Support for an M32R GNU/Linux target and PIC code generation has 9383 been added by Renesas. 9384 9385 M68000 9386 9387 * Bernardo Innocenti (Develer S.r.l.) has contributed the 9388 m68k-uclinux target, based on former work done by Paul Dale 9389 (SnapGear Inc.). Code generation for the ColdFire processors family 9390 has been enhanced and extended to support the MCF 53xx and MCF 54xx 9391 cores, integrating former work done by Peter Barada (Motorola). 9392 9393 MIPS 9394 9395 Processor-specific changes 9396 9397 * Support for the RM7000 and RM9000 processors has been added. It can 9398 be selected using the -march compiler option and should work with 9399 any MIPS I (mips-*) or MIPS III (mips64-*) configuration. 9400 * Support for revision 2 of the MIPS32 ISA has been added. It can be 9401 selected with the command-line option -march=mips32r2. 9402 * There is a new option, -mfix-sb1, to work around certain SB-1 9403 errata. 9404 9405 Configuration 9406 9407 * It is possible to customize GCC using the following configure-time 9408 options: 9409 + --with-arch, which specifies the default value of the -march 9410 option. 9411 + --with-tune, which specifies the default value of the -mtune 9412 option. 9413 + --with-abi, which specifies the default ABI. 9414 + --with-float=soft, which tells GCC to use software floating 9415 point by default. 9416 + --with-float=hard, which tells GCC to use hardware floating 9417 point by default. 9418 * A 64-bit GNU/Linux port has been added. The associated 9419 configurations are mips64-linux-gnu and mips64el-linux-gnu. 9420 * The 32-bit GNU/Linux port now supports Java. 9421 * The IRIX 6 configuration now supports the o32 ABI and will build 9422 o32 multilibs by default. This support is compatible with both 9423 binutils and the SGI tools, but note that several features, 9424 including debugging information and DWARF2 exception handling, are 9425 only available when using the GNU assembler. Use of the GNU 9426 assembler and linker (version 2.15 or above) is strongly 9427 recommended. 9428 * The IRIX 6 configuration now supports 128-bit long doubles. 9429 * There are two new RTEMS-specific configurations, mips-rtems and 9430 mipsel-rtems. 9431 * There are two new *-elf configurations, mipsisa32r2-elf and 9432 mipsisa32r2el-elf. 9433 9434 General 9435 9436 * Several [24]ABI bugs have been fixed. Unfortunately, these changes 9437 will break binary compatibility with earlier releases. 9438 * GCC can now use explicit relocation operators when generating 9439 -mabicalls code. This behavior is controlled by -mexplicit-relocs 9440 and can have several performance benefits. For example: 9441 + It allows for more optimization of GOT accesses, including 9442 better scheduling and redundancy elimination. 9443 + It allows sibling calls to be implemented as jumps. 9444 + n32 and n64 leaf functions can use a call-clobbered global 9445 pointer instead of $28. 9446 + The code to set up $gp can be removed from functions that 9447 don't need it. 9448 * A new option, -mxgot, allows the GOT to be bigger than 64k. This 9449 option is equivalent to the assembler's -xgot option and should be 9450 used instead of -Wa,-xgot. 9451 * Frame pointer elimination is now supported when generating 64-bit 9452 MIPS16 code. 9453 * Inline block moves have been optimized to take more account of 9454 alignment information. 9455 * Many internal changes have been made to the MIPS port, mostly aimed 9456 at reducing the reliance on assembler macros. 9457 9458 PowerPC 9459 9460 * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64 9461 [25]ABI incompatibilities regarding the way parameters are passed 9462 during functions calls. These changes may result in incompatibility 9463 between code compiled with GCC 3.3 and GCC 3.4. 9464 9465 PowerPC Darwin 9466 9467 * Support for shared/dylib gcc libraries has been added. It is 9468 enabled by default on powerpc-apple-darwin7.0.0 and up. 9469 * Libgcj is enabled by default. On systems older than 9470 powerpc-apple-darwin7.0.0 you need to install dlcompat. 9471 * 128-bit IBM extended precision format support added for long 9472 double. 9473 9474 PowerPC64 GNU/Linux 9475 9476 * By default, PowerPC64 GNU/Linux now uses natural alignment of 9477 structure elements. The old four byte alignment for double, with 9478 special rules for a struct starting with a double, can be chosen 9479 with -malign-power. This change may result in incompatibility 9480 between code compiled with GCC 3.3 and GCC 3.4. 9481 * -mabi=altivec is now the default rather than -mabi=no-altivec. 9482 * 128-bit IBM extended precision format support added for long 9483 double. 9484 9485 S/390 and zSeries 9486 9487 * New command-line options allow to specify the intended execution 9488 environment for generated code: 9489 + -mesa/-mzarch allows to specify whether to generate code 9490 running in ESA/390 mode or in z/Architecture mode (this is 9491 applicable to 31-bit code only). 9492 + -march allows to specify a minimum processor architecture 9493 level (g5, g6, z900, or z990). 9494 + -mtune allows to specify which processor to tune for. 9495 * It is possible to customize GCC using the following configure-time 9496 options: 9497 + --with-mode, which specifies whether to default to assuming 9498 ESA/390 or z/Architecture mode. 9499 + --with-arch, which specifies the default value of the -march 9500 option. 9501 + --with-tune, which specifies the default value of the -mtune 9502 option. 9503 * Support for the z990 processor has been added, and can be selected 9504 using -march=z990 or -mtune=z990. This includes instruction 9505 scheduling tuned for the superscalar instruction pipeline of the 9506 z990 processor as well as support for all new instructions provided 9507 by the long-displacement facility. 9508 * Support to generate 31-bit code optimized for zSeries processors 9509 (running in ESA/390 or in z/Architecture mode) has been added. This 9510 can be selected using -march=z900 and -mzarch respectively. 9511 * Instruction scheduling for the z900 and z990 processors now uses 9512 the DFA pipeline hazard recognizer. 9513 * GCC no longer generates code to maintain a stack backchain, 9514 previously used to generate stack backtraces for debugging 9515 purposes. As replacement that does not incur runtime overhead, 9516 DWARF-2 call frame information is provided by GCC; this is 9517 supported by GDB 6.1. The old behavior can be restored using the 9518 -mbackchain option. 9519 * The stack frame size of functions may now exceed 2 GB in 64-bit 9520 code. 9521 * A port for the 64-bit IBM TPF operating system has been added; the 9522 configuration is s390x-ibm-tpf. This configuration is supported as 9523 cross-compilation target only. 9524 * Various changes to improve the generated code have been 9525 implemented, including: 9526 + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT 9527 instructions to significantly speed up many floating-point 9528 applications. 9529 + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL 9530 WITH BORROW instructions to speed up long long arithmetic. 9531 + GCC now uses the SEARCH STRING instruction to implement 9532 strlen(). 9533 + In many cases, function call overhead for 31-bit code has been 9534 reduced by placing the literal pool after the function code 9535 instead of after the function prolog. 9536 + Register 14 is no longer reserved in 64-bit code. 9537 + Handling of global register variables has been improved. 9538 9539 SPARC 9540 9541 * The option -mflat is deprecated. 9542 * Support for large (> 2GB) frames has been added to the 64-bit port. 9543 * Several [26]ABI bugs have been fixed. Unfortunately, these changes 9544 will break binary compatibility with earlier releases. 9545 * The default debugging format has been switched from STABS to 9546 DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already 9547 the default debugging format for 64-bit code on Solaris. 9548 9549 SuperH 9550 9551 * Support for the SH2E processor has been added. Enabled at run time 9552 with the -m2e command line switch, or at configure time by 9553 specifying sh2e as the machine part of the target triple. 9554 9555 V850 9556 9557 * Support for the Mitsubishi V850E1 processor has been added. This is 9558 a variant of the V850E processor with some additional debugging 9559 instructions. 9560 9561 Xtensa 9562 9563 * Several ABI bugs have been fixed. Unfortunately, these changes 9564 break binary compatibility with earlier releases. 9565 + For big-endian processors, the padding of aggregate return 9566 values larger than a word has changed. If the size of an 9567 aggregate return value is not a multiple of 32 bits, previous 9568 versions of GCC inserted padding in the most-significant bytes 9569 of the first return value register. Aggregates larger than a 9570 word are now padded in the least-significant bytes of the last 9571 return value register used. Aggregates smaller than a word are 9572 still padded in the most-significant bytes. The return value 9573 padding has not changed for little-endian processors. 9574 + Function arguments with 16-byte alignment are now properly 9575 aligned. 9576 + The implementation of the va_list type has changed. A va_list 9577 value created by va_start from a previous release cannot be 9578 used with va_arg from this release, or vice versa. 9579 * More processor configuration options for Xtensa processors are 9580 supported: 9581 + the ABS instruction is now optional; 9582 + the ADDX* and SUBX* instructions are now optional; 9583 + an experimental CONST16 instruction can be used to synthesize 9584 constants instead of loading them from constant pools. 9585 These and other Xtensa processor configuration options can no 9586 longer be enabled or disabled by command-line options; the 9587 processor configuration must be specified by the xtensa-config.h 9588 header file when building GCC. Additionally, the 9589 -mno-serialize-volatile option is no longer supported. 9590 9591 Obsolete Systems 9592 9593 Support for a number of older systems has been declared obsolete in GCC 9594 3.4. Unless there is activity to revive them, the next release of GCC 9595 will have their sources permanently removed. 9596 9597 All configurations of the following processor architectures have been 9598 declared obsolete: 9599 * Mitsubishi D30V, d30v-* 9600 * AT&T DSP1600 and DSP1610, dsp16xx-* 9601 * Intel 80960, i960 9602 9603 Also, some individual systems have been obsoleted: 9604 * ARM Family 9605 + Support for generating code for operation in APCS/26 mode 9606 (-mapcs-26). 9607 * IBM ESA/390 9608 + "Bigfoot" port, i370-*. (The other port, s390-*, is actively 9609 maintained and supported.) 9610 * Intel 386 family 9611 + MOSS, i?86-moss-msdos and i?86-*-moss* 9612 + NCR 3000 running System V r.4, i?86-ncr-sysv4* 9613 + FreeBSD with a.out object format, i?86-*-freebsd*aout* and 9614 i?86-*-freebsd2* 9615 + GNU/Linux with a.out object format, i?86-linux*aout* 9616 + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1* 9617 + Interix versions before Interix 3, i?86-*-interix 9618 + Mach microkernel, i?86-mach* 9619 + SCO UnixWare with UDK, i?86-*-udk* 9620 + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]* 9621 + VSTa microkernel, i386-*-vsta 9622 * Motorola M68000 family 9623 + HPUX, m68k-hp-hpux* and m68000-hp-hpux* 9624 + NetBSD with a.out object format (before NetBSD 1.4), 9625 m68k-*-*-netbsd* except m68k-*-*-netbsdelf* 9626 + Generic System V r.4, m68k-*-sysv4* 9627 * VAX 9628 + Generic VAX, vax-*-* (This is generic VAX only; we have not 9629 obsoleted any VAX triples for specific operating systems.) 9630 9631 Documentation improvements 9632 9633 Other significant improvements 9634 9635 * The build system has undergone several significant cleanups. 9636 Subdirectories will only be configured if they are being built, and 9637 all subdirectory configures are run from the make command. The top 9638 level has been autoconfiscated. 9639 * Building GCC no longer writes to its source directory. This should 9640 help those wishing to share a read-only source directory over NFS 9641 or build from a CD. The exceptions to this feature are if you 9642 configure with either --enable-maintainer-mode or 9643 --enable-generated-files-in-srcdir. 9644 * The -W warning option has been renamed to -Wextra, which is more 9645 easily understood. The older spelling will be retained for 9646 backwards compatibility. 9647 * Substantial improvements in compile time have been made, 9648 particularly for non-optimizing compilations. 9649 __________________________________________________________________ 9650 9651 GCC 3.4.0 9652 9653 Bug Fixes 9654 9655 A vast number of bugs have been fixed in 3.4.0, too many to publish a 9656 complete list here. [27]Follow this link to query the Bugzilla database 9657 for the list of over 900 bugs fixed in 3.4.0. This is the list of all 9658 bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4 9659 regressions. 9660 __________________________________________________________________ 9661 9662 GCC 3.4.1 9663 9664 Bug Fixes 9665 9666 This section lists the problem reports (PRs) from GCC's bug tracking 9667 system that are known to be fixed in the 3.4.1 release. This list might 9668 not be complete (that is, it is possible that some PRs that have been 9669 fixed are not listed here). 9670 9671 Bootstrap failures 9672 9673 * [28]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler 9674 emitted - PIC related 9675 * [29]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf 9676 * [30]14760 A bug in configure.in prevents using both 9677 --program-suffix and --program-prefix 9678 * [31]14671 [hppa64] bootstrap fails: ICE in 9679 save_call_clobbered_regs, in caller_save.c 9680 * [32]15093 [alpha][Java] make bootstrap fails to configure libffi on 9681 Alpha 9682 * [33]15178 Solaris 9/x86 fails linking after stage 3 9683 9684 Multi-platform internal compiler errors (ICEs) 9685 9686 * [34]12753 (preprocessor) Memory corruption in preprocessor on bad 9687 input 9688 * [35]13985 ICE in gcc.c-torture/compile/930621-1.c 9689 * [36]14810 (c++) tree check failures with invalid code involving 9690 templates 9691 * [37]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in 9692 cp/parser.c 9693 * [38]15044 (c++) ICE on syntax error, template header 9694 * [39]15057 (c++) Compiling of conditional value throw constructs 9695 cause a segmentation violation 9696 * [40]15064 (c++) typeid of template parameter gives ICE 9697 * [41]15142 (c++) ICE when passing a string where a char* is expected 9698 in a throw statement 9699 * [42]15159 ICE in rtl_verify_flow_info_1 9700 * [43]15165 (c++) ICE in instantiate_template 9701 * [44]15193 Unary minus using pointer to V4SF vector causes 9702 -fforce-mem to exhaust all memory 9703 * [45]15209 (c++) Runs out of memory with packed structs 9704 * [46]15227 (c++) Trouble with invalid function definition 9705 * [47]15285 (c++) instantiate_type ICE when forming pointer to 9706 template function 9707 * [48]15299 (c++) ICE in resolve_overloaded_unification 9708 * [49]15329 (c++) ICE on constructor of member template 9709 * [50]15550 ICE in extract_insn, in recog.c 9710 * [51]15554 (c++) ICE in tsubst_copy, in cp/pt.c 9711 * [52]15640 (c++) ICE on invalid code in arg_assoc, in 9712 cp/name-lookup.c 9713 * [53]15666 [unit-at-a-time] Gcc abort on valid code 9714 * [54]15696 (c++) ICE with bad pointer-to-member code 9715 * [55]15701 (c++) ICE with friends and template template parameter 9716 * [56]15761 ICE in do_SUBST, in combine.c 9717 * [57]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops 9718 9719 Ada 9720 9721 * [58]14538 All RTEMS targets broken for gnat 9722 9723 C front end 9724 9725 * [59]12391 missing warning about assigning to an incomplete type 9726 * [60]14649 atan(1.0) should not be a constant expression 9727 * [61]15004 [unit-at-a-time] no warning for unused paramater in 9728 static function 9729 * [62]15749 --pedantic-errors behaves differently from --pedantic 9730 with C-compiler on GNU/Linux 9731 9732 C++ compiler and library 9733 9734 * [63]10646 non-const reference is incorrectly matched in a "const T" 9735 partial specialization 9736 * [64]12077 wcin.rdbuf()->in_avail() return value too high 9737 * [65]13598 enc_filebuf doesn't work 9738 * [66]14211 const_cast returns lvalue but should be rvalue 9739 * [67]14220 num_put::do_put() undesired float/double behavior 9740 * [68]14245 problem with user-defined allocators in std::basic_string 9741 * [69]14340 libstdc++ Debug mode: failure to convert iterator to 9742 const_iterator 9743 * [70]14600 __gnu_cxx::stdio_sync_filebuf should expose internal 9744 FILE* 9745 * [71]14668 no warning anymore for reevaluation of declaration 9746 * [72]14775 LFS (large file support) tests missing 9747 * [73]14821 Duplicate namespace alias declaration should not conflict 9748 * [74]14930 Friend declaration ignored 9749 * [75]14932 cannot use offsetof to get offsets of array elements in 9750 g++ 3.4.0 9751 * [76]14950 [non unit-at-a-time] always_inline does not mix with 9752 templates and -O0 9753 * [77]14962 g++ ignores #pragma redefine_extname 9754 * [78]14975 Segfault on low-level write error during imbue 9755 * [79]15002 Linewise stream input is unusably slow (std::string slow) 9756 * [80]15025 compiler accepts redeclaration of template as 9757 non-template 9758 * [81]15046 [arm] Math functions misdetected by cross configuration 9759 * [82]15069 a bit test on a variable of enum type is miscompiled 9760 * [83]15074 g++ -lsupc++ still links against libstdc++ 9761 * [84]15083 spurious "statement has no effect" warning 9762 * [85]15096 parse error with templates and pointer to const member 9763 * [86]15287 combination of operator[] and operator .* fails in 9764 templates 9765 * [87]15317 __attribute__ unused in first parameter of constructor 9766 gives error 9767 * [88]15337 sizeof on incomplete type diagnostic 9768 * [89]15361 bitset<>::_Find_next fails 9769 * [90]15412 _GLIBCXX_ symbols symbols defined and used in different 9770 namespaces 9771 * [91]15427 valid code results in incomplete type error 9772 * [92]15471 Incorrect member pointer offsets in anonymous 9773 structs/unions 9774 * [93]15503 nested template problem 9775 * [94]15507 compiler hangs while laying out union 9776 * [95]15542 operator & and template definitions 9777 * [96]15565 SLES9: leading + sign for unsigned int with showpos 9778 * [97]15625 friend defined inside a template fails to find static 9779 function 9780 * [98]15629 Function templates, overloads, and friend name injection 9781 * [99]15742 'noreturn' attribute ignored in method of template 9782 functions. 9783 * [100]15775 Allocator::pointer consistently ignored 9784 * [101]15821 Duplicate namespace alias within namespace rejected 9785 * [102]15862 'enum yn' fails (confict with undeclared builtin) 9786 * [103]15875 rejects pointer to member in template 9787 * [104]15877 valid code using templates and anonymous enums is 9788 rejected 9789 * [105]15947 Puzzling error message for wrong destructor declaration 9790 in template class 9791 * [106]16020 cannot copy __gnu_debug::bitset 9792 * [107]16154 input iterator concept too restrictive 9793 * [108]16174 deducing top-level consts 9794 9795 Java 9796 9797 * [109]14315 Java compiler is not parallel make safe 9798 9799 Fortran 9800 9801 * [110]15151 [g77] incorrect logical i/o in 64-bit mode 9802 9803 Objective-C 9804 9805 * [111]7993 private variables cannot be shadowed in subclasses 9806 9807 Optimization bugs 9808 9809 * [112]15228 useless copies of floating point operands 9810 * [113]15345 [non-unit-at-a-time] unreferenced nested inline 9811 functions not optimized away 9812 * [114]15945 Incorrect floating point optimization 9813 * [115]15526 ftrapv aborts on 0 * (-1) 9814 * [116]14690 Miscompiled POOMA tests 9815 * [117]15112 GCC generates code to write to unchanging memory 9816 9817 Preprocessor 9818 9819 * [118]15067 Minor glitch in the source of cpp 9820 9821 Main driver program bugs 9822 9823 * [119]1963 collect2 interprets -oldstyle_liblookup as -o 9824 ldstyle_liblookup 9825 9826 x86-specific (Intel/AMD) 9827 9828 * [120]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND* 9829 section} 9830 9831 HPPA-specific 9832 9833 * [121]14782 GCC produces an unaligned data access at -O2 9834 * [122]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2 9835 * [123]15202 ICE in reload_cse_simplify_operands, in postreload.c 9836 9837 IA64-specific 9838 9839 * [124]14610 __float80 constants incorrectly emitted 9840 * [125]14813 init_array sections are initialized in the wrong order 9841 * [126]14857 GCC segfault on duplicated asm statement 9842 * [127]15598 Gcc 3.4 ICE on valid code 9843 * [128]15653 Gcc 3.4 ICE on valid code 9844 9845 MIPS-specific 9846 9847 * [129]15189 wrong filling of delay slot with -march=mips1 -G0 9848 -mno-split-addresses -mno-explicit-relocs 9849 * [130]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as 9850 2.14.91 9851 * [131]16144 Bogus reference to __divdf3 when -O1 9852 * [132]16176 Miscompilation of unaligned data in MIPS backend 9853 9854 PowerPC-specific 9855 9856 * [133]11591 ICE in gcc.dg/altivec-5.c 9857 * [134]12028 powerpc-eabispe produces bad sCOND operation 9858 * [135]14478 rs6000 geu/ltu patterns generate incorrect code 9859 * [136]14567 long double and va_arg complex args 9860 * [137]14715 Altivec stack layout may overlap gpr save with stack 9861 temps 9862 * [138]14902 (libstdc++) Stream checking functions fail when -pthread 9863 option is used. 9864 * [139]14924 Compiler ICE on valid code 9865 * [140]14960 -maltivec affects vector return with -mabi=no-altivec 9866 * [141]15106 vector varargs failure passing from altivec to 9867 non-altivec code for -m32 9868 * [142]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 & 9869 half-word operation 9870 * [143]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx 9871 and stvx 9872 * [144]15662 Segmentation fault when an exception is thrown - even if 9873 try and catch are specified 9874 9875 s390-specific 9876 9877 * [145]15054 Bad code due to overlapping stack temporaries 9878 9879 SPARC-specific 9880 9881 * [146]15783 ICE with union assignment in 64-bit mode 9882 * [147]15626 GCC 3.4 emits "ld: warning: relocation error: 9883 R_SPARC_UA32" 9884 9885 x86-64-specific 9886 9887 * [148]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64 9888 * [149]14723 Backported -march=nocona from mainline 9889 * [150]15290 __float128 failed to pass to function properly 9890 9891 Cygwin/Mingw32-specific 9892 9893 * [151]15250 Option -mms-bitfields support on GCC 3.4 is not 9894 conformant to MS layout 9895 * [152]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe 9896 worker on windows32 targets 9897 9898 Bugs specific to embedded processors 9899 9900 * [153]8309 [m68k] -m5200 produces erroneous SImode set of short 9901 varaible on stack 9902 * [154]13250 [SH] Gcc code for rotation clobbers the register, but 9903 gcc continues to use the register as if it was not clobbered 9904 * [155]13803 [coldfire] movqi operand constraints too restrictivefor 9905 TARGET_COLDFIRE 9906 * [156]14093 [SH] ICE for code when using -mhitachi option in SH 9907 * [157]14457 [m6811hc] ICE with simple c++ source 9908 * [158]14542 [m6811hc] ICE on simple source 9909 * [159]15100 [SH] cc1plus got hang-up on 9910 libstdc++-v3/testsuite/abi_check.cc 9911 * [160]15296 [CRIS] Delayed branch scheduling causing invalid code on 9912 cris-* 9913 * [161]15396 [SH] ICE with -O2 -fPIC 9914 * [162]15782 [coldfire] m68k_output_mi_thunk emits wrong code for 9915 ColdFire 9916 9917 Testsuite problems (compiler not affected) 9918 9919 * [163]11610 libstdc++ testcases 27_io/* don't work properly remotely 9920 * [164]15488 (libstdc++) possibly insufficient file permissions for 9921 executing test suite 9922 * [165]15489 (libstdc++) testsuite_files determined incorrectly 9923 9924 Documentation bugs 9925 9926 * [166]13928 (libstdc++) no whatis info in some man pages generated 9927 by doxygen 9928 * [167]14150 Ada documentation out of date 9929 * [168]14949 (c++) Need to document method visibility changes 9930 * [169]15123 libstdc++-doc: Allocators.3 manpage is empty 9931 __________________________________________________________________ 9932 9933 GCC 3.4.2 9934 9935 Bug Fixes 9936 9937 This section lists the problem reports (PRs) from GCC's bug tracking 9938 system that are known to be fixed in the 3.4.2 release. This list might 9939 not be complete (that is, it is possible that some PRs that have been 9940 fixed are not listed here). 9941 9942 Bootstrap failures and issues 9943 9944 * [170]16469 [mips-sgi-irix5.3] bootstrap fails in 9945 libstdc++-v3/testsuite 9946 * [171]16344 [hppa-linux-gnu] libstdc++'s PCH built by 9947 profiledbootstrap does not work with the built compiler 9948 * [172]16842 [Solaris/x86] mkheaders can not find mkheaders.conf 9949 9950 Multi-platform internal compiler errors (ICEs) 9951 9952 * [173]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in 9953 cp_parser_class_specifier, in cp/parser.c 9954 * [174]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 9955 * [175]15461 (c++) ICE due to NRV and inlining 9956 * [176]15890 (c++) ICE in c_expand_expr, in c-common.c 9957 * [177]16180 ICE: segmentation fault in RTL optimization 9958 * [178]16224 (c++) ICE in write_unscoped_name (template/namespace) 9959 * [179]16408 ICE: in delete_insn, in cfgrtl.c 9960 * [180]16529 (c++) ICE for: namespace-alias shall not be declared as 9961 the name of any other entity 9962 * [181]16698 (c++) ICE with exceptions and declaration of __cxa_throw 9963 * [182]16706 (c++) ICE in finish_member_declaration, in 9964 cp/semantics.c 9965 * [183]16810 (c++) Legal C++ program with cast gives ICE in 9966 build_ptrmemfunc 9967 * [184]16851 (c++) ICE when throwing a comma expression 9968 * [185]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c 9969 * [186]16904 (c++) ICE in finish_class_member_access_expr, in 9970 cp/typeck.c 9971 * [187]16905 (c++) ICE (segfault) with exceptions 9972 * [188]16964 (c++) ICE in cp_parser_class_specifier due to 9973 redefinition 9974 * [189]17068 (c++) ICE: tree check: expected class 'd', have 'x' 9975 (identifier_node) in dependent_template_p, in cp/pt.c 9976 9977 Preprocessor bugs 9978 9979 * [190]16366 Preprocessor option -remap causes memory corruption 9980 9981 Optimization 9982 9983 * [191]15345 unreferenced nested inline functions not optimized away 9984 * [192]16590 Incorrect execution when compiling with -O2 9985 * [193]16693 Bitwise AND is lost when used within a cast to an enum 9986 of the same precision 9987 * [194]17078 Jump into if(0) substatement fails 9988 9989 Problems in generated debug information 9990 9991 * [195]13956 incorrect stabs for nested local variables 9992 9993 C front end bugs 9994 9995 * [196]16684 GCC should not warn about redundant redeclarations of 9996 built-ins 9997 9998 C++ compiler and library 9999 10000 * [197]12658 Thread safety problems in locale::global() and 10001 locale::locale() 10002 * [198]13092 g++ accepts invalid pointer-to-member conversion 10003 * [199]15320 Excessive memory consumption 10004 * [200]16246 Incorrect template argument deduction 10005 * [201]16273 Memory exhausted when using nested classes and virtual 10006 functions 10007 * [202]16401 ostringstream in gcc 3.4.x very slow for big data 10008 * [203]16411 undefined reference to 10009 __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char> 10010 >::file() 10011 * [204]16489 G++ incorrectly rejects use of a null constant integral 10012 expression as a null constant pointer 10013 * [205]16618 offsetof fails with constant member 10014 * [206]16637 syntax error reported for valid input code 10015 * [207]16717 __attribute__((constructor)) broken in C++ 10016 * [208]16813 compiler error in DEBUG version of range insertion 10017 std::map::insert 10018 * [209]16853 pointer-to-member initialization from incompatible one 10019 accepted 10020 * [210]16889 ambiguity is not detected 10021 * [211]16959 Segmentation fault in ios_base::sync_with_stdio 10022 10023 Java compiler and library 10024 10025 * [212]7587 direct threaded interpreter not thread-safe 10026 * [213]16473 ServerSocket accept() leaks file descriptors 10027 * [214]16478 Hash synchronization deadlock with finalizers 10028 10029 Alpha-specific 10030 10031 * [215]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c 10032 * [216]16974 could not split insn (ice in final_scan_insn, in 10033 final.c) 10034 10035 x86-specific 10036 10037 * [217]16298 ICE in output_operand 10038 * [218]17113 ICE with SSE2 intrinsics 10039 10040 x86-64 specific 10041 10042 * [219]14697 libstdc++ couldn't find 32bit libgcc_s 10043 10044 MIPS-specific 10045 10046 * [220]15869 [mips64] No NOP after LW (with -mips1 -O0) 10047 * [221]16325 [mips64] value profiling clobbers gp on mips 10048 * [222]16357 [mipsisa64-elf] ICE copying 7 bytes between extern 10049 char[]s 10050 * [223]16380 [mips64] Use of uninitialised register after dbra 10051 conversion 10052 * [224]16407 [mips64] Unaligned access to local variables 10053 * [225]16643 [mips64] verify_local_live_at_start ICE after 10054 crossjumping & cfgcleanup 10055 10056 ARM-specific 10057 10058 * [226]15927 THUMB -O2: strength-reduced iteration variable ends up 10059 off by 1 10060 * [227]15948 THUMB: ICE with non-commutative cbranch 10061 * [228]17019 THUMB: bad switch statement in md code for 10062 addsi3_cbranch_scratch 10063 10064 IA64-specific 10065 10066 * [229]16130 ICE on valid code: in bundling, in config/ia64/ia64.c 10067 (-mtune=merced) 10068 * [230]16142 ICE on valid code: in bundling, in config/ia64/ia64.c 10069 (-mtune=itanium) 10070 * [231]16278 Gcc failed to build Linux kernel with -mtune=merced 10071 * [232]16414 ICE on valid code: typo in comparison of asm_noperands 10072 result 10073 * [233]16445 ICE on valid code: don't count ignored insns 10074 * [234]16490 ICE (segfault) while compiling with -fprofile-use 10075 * [235]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS 10076 10077 PowerPC-specific 10078 10079 * [236]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x 10080 * [237]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions 10081 issue) 10082 10083 SPARC-specific 10084 10085 * [238]16199 ICE while compiling apache 2.0.49 10086 * [239]16416 -m64 doesn't imply -mcpu=v9 anymore 10087 * [240]16430 ICE when returning non-C aggregates larger than 16 bytes 10088 10089 Bugs specific to embedded processors 10090 10091 * [241]16379 [m32r] can't output large model function call of memcpy 10092 * [242]17093 [m32r] ICE with -msdata=use -O0 10093 * [243]17119 [m32r] ICE at switch case 0x8000 10094 10095 DJGPP-specific 10096 10097 * [244]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp 10098 10099 Alpha Tru64-specific 10100 10101 * [245]16210 libstdc++ gratuitously omits "long long" I/O 10102 10103 Testsuite, documentation issues (compiler is not affected): 10104 10105 * [246]15488 (libstdc++) possibly insufficient file permissions for 10106 executing test suite 10107 * [247]16250 ada/doctools runs makeinfo even in release tarball 10108 __________________________________________________________________ 10109 10110 GCC 3.4.3 10111 10112 This is the [248]list of problem reports (PRs) from GCC's bug tracking 10113 system that are known to be fixed in the 3.4.3 release. This list might 10114 not be complete (that is, it is possible that some PRs that have been 10115 fixed are not listed here). 10116 10117 Bootstrap failures 10118 10119 * [249]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1 10120 * [250]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold 10121 when undeclared 10122 10123 Internal compiler errors (ICEs) affecting multiple platforms 10124 10125 * [251]13948 (java) GCJ segmentation fault while compiling GL4Java 10126 .class files 10127 * [252]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 10128 * [253]16301 (c++) ICE when "strong" attribute is attached to a using 10129 directive 10130 * [254]16566 ICE with flexible arrays 10131 * [255]17023 ICE with nested functions in parameter declaration 10132 * [256]17027 ICE with noreturn function in loop at -O2 10133 * [257]17524 ICE in grokdeclarator, in cp/decl.c 10134 * [258]17826 (c++) ICE in cp_tree_equal 10135 10136 C and optimization bugs 10137 10138 * [259]15526 -ftrapv aborts on 0 * (-1) 10139 * [260]16999 #ident stopped working 10140 * [261]17503 quadratic behaviour in invalid_mode_change_p 10141 * [262]17581 Long long arithmetic fails inside a switch/case 10142 statement when compiled with -O2 10143 * [263]18129 -fwritable-strings doesn't work 10144 10145 C++ compiler and library bugs 10146 10147 * [264]10975 incorrect initial ostringstream::tellp() 10148 * [265]11722 Unbuffered filebuf::sgetn is slow 10149 * [266]14534 Unrecognizing static function as a template parameter 10150 when its return value is also templated 10151 * [267]15172 Copy constructor optimization in aggregate 10152 initialization 10153 * [268]15786 Bad error message for frequently occuring error. 10154 * [269]16162 Rejects valid member-template-definition 10155 * [270]16612 empty basic_strings can't live in shared memory 10156 * [271]16715 std::basic_iostream is instantiated when used, even 10157 though instantiations are already contained in libstdc++ 10158 * [272]16848 code in /ext/demangle.h appears broken 10159 * [273]17132 GCC fails to eliminate function template specialization 10160 when argument deduction fails 10161 * [274]17259 One more _S_leaf incorrectly qualified with _RopeRep:: 10162 in ropeimpl.h 10163 * [275]17327 use of `enumeral_type' in template type unification 10164 * [276]17393 "unused variable '._0'" warning with -Wall 10165 * [277]17501 Confusion with member templates 10166 * [278]17537 g++ not passing -lstdc++ to linker when all command line 10167 arguments are libraries 10168 * [279]17585 usage of unqualified name of static member from within 10169 class not allowed 10170 * [280]17821 Poor diagnostic for using "." instead of "->" 10171 * [281]17829 wrong error: call of overloaded function is ambiguous 10172 * [282]17851 Misleading diagnostic for invalid function declarations 10173 with undeclared types 10174 * [283]17976 Destructor is called twice 10175 * [284]18020 rejects valid definition of enum value in template 10176 * [285]18093 bogus conflict in namespace aliasing 10177 * [286]18140 C++ parser bug when using >> in templates 10178 10179 Fortran 10180 10181 * [287]17541 data statements with double precision constants fail 10182 10183 x86-specific 10184 10185 * [288]17853 -O2 ICE for MMX testcase 10186 10187 SPARC-specific 10188 10189 * [289]17245 ICE compiling gsl-1.5 statistics/lag1.c 10190 10191 Darwin-specific 10192 10193 * [290]17167 FATAL:Symbol L_foo$stub already defined. 10194 10195 AIX-specific 10196 10197 * [291]17277 could not catch an exception when specified -maix64 10198 10199 Solaris-specific 10200 10201 * [292]17505 <cmath> calls acosf(), ceilf(), and other functions 10202 missing from system libraries 10203 10204 HP/UX specific: 10205 10206 * [293]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl 10207 10208 ARM-specific 10209 10210 * [294]17384 ICE with mode attribute on structures 10211 10212 MIPS-specific 10213 10214 * [295]17770 No NOP after LWL with -mips1 10215 10216 Other embedded target specific 10217 10218 * [296]11476 [arc-elf] gcc ICE on newlib's vfprintf.c 10219 * [297]14064 [avr-elf] -fdata-sections triggers ICE 10220 * [298]14678 [m68hc11-elf] gcc ICE 10221 * [299]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__ 10222 * [300]15790 [i686-coff] Alignment error building gcc with i686-coff 10223 target 10224 * [301]15886 [SH] Miscompilation with -O2 -fPIC 10225 * [302]16884 [avr-elf] [fweb related] bug while initializing 10226 variables 10227 10228 Bugs relating to debugger support 10229 10230 * [303]13841 missing debug info for _Complex function arguments 10231 * [304]15860 [big-endian targets] No DW_AT_location debug info is 10232 emitted for formal arguments to a function that uses "register" 10233 qualifiers 10234 10235 Testsuite issues (compiler not affected) 10236 10237 * [305]17465 Testsuite in libffi overrides LD_LIBRARY_PATH 10238 * [306]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH 10239 * [307]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit 10240 testsuite 10241 10242 Documentation 10243 10244 * [308]15498 typo in gcc manual: non-existing locale example en_UK, 10245 should be en_GB 10246 * [309]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap: 10247 document broken shell 10248 * [310]16406 USE_LD_AS_NEEDED undocumented 10249 __________________________________________________________________ 10250 10251 GCC 3.4.4 10252 10253 This is the [311]list of problem reports (PRs) from GCC's bug tracking 10254 system that are known to be fixed in the 3.4.4 release. This list might 10255 not be complete (that is, it is possible that some PRs that have been 10256 fixed are not listed here). 10257 __________________________________________________________________ 10258 10259 GCC 3.4.5 10260 10261 This is the [312]list of problem reports (PRs) from GCC's bug tracking 10262 system that are known to be fixed in the 3.4.5 release. This list might 10263 not be complete (that is, it is possible that some PRs that have been 10264 fixed are not listed here). 10265 10266 Bootstrap issues 10267 10268 * [313]24688 sco_math fixincl breaks math.h 10269 10270 C compiler bugs 10271 10272 * [314]17188 struct Foo { } redefinition 10273 * [315]20187 wrong code for ((unsigned char)(unsigned long 10274 long)((a?a:1)&(a*b)))?0:1) 10275 * [316]21873 infinite warning loop on bad array initializer 10276 * [317]21899 enum definition accepts values to be overriden 10277 * [318]22061 ICE in find_function_data, in function.c 10278 * [319]22308 Failure to diagnose violation of constraint 6.516p2 10279 * [320]22458 ICE on missing brace 10280 * [321]22589 ICE casting to long long 10281 * [322]24101 Segfault with preprocessed source 10282 10283 C++ compiler and library bugs 10284 10285 * [323]10611 operations on vector mode not recognized in C++ 10286 * [324]13377 unexpected behavior of namespace usage directive 10287 * [325]16002 Strange error message with new parser 10288 * [326]17413 local classes as template argument 10289 * [327]17609 spurious error message after using keyword 10290 * [328]17618 ICE in cp_convert_to_pointer, in cp/cvt.c 10291 * [329]18124 ICE with invalid template template parameter 10292 * [330]18155 typedef in template declaration not rejected 10293 * [331]18177 ICE with const_cast for undeclared variable 10294 * [332]18368 C++ error message regression 10295 * [333]16378 ICE when returning a copy of a packed member 10296 * [334]18466 int ::i; accepted 10297 * [335]18512 ICE on invalid usage of template base class 10298 * [336]18454 ICE when returning undefined type 10299 * [337]18738 typename not allowed with non-dependent qualified name 10300 * [338]18803 rejects access to operator() in template 10301 * [339]19004 ICE in uses_template_parms, in cp/pt.c 10302 * [340]19208 Spurious error about variably modified type 10303 * [341]18253 bad error message / ICE for invalid template parameter 10304 * [342]19608 ICE after friend function definition in local class 10305 * [343]19884 ICE on explicit instantiation of a non-template 10306 constructor 10307 * [344]20153 ICE when C++ template function contains anonymous union 10308 * [345]20563 Infinite loop in diagnostic (and ice after error 10309 message) 10310 * [346]20789 ICE with incomplete type in template 10311 * [347]21336 Internal compiler error when using custom new operators 10312 * [348]21768 ICE in error message due to violation of coding 10313 conventions 10314 * [349]21853 constness of pointer to data member ignored 10315 * [350]21903 Default argument of template function causes a 10316 compile-time error 10317 * [351]21983 multiple diagnostics 10318 * [352]21987 New testsuite failure 10319 g++.dg/warn/conversion-function-1.C 10320 * [353]22153 ICE on invalid template specialization 10321 * [354]22172 Internal compiler error, seg fault. 10322 * [355]21286 filebuf::xsgetn vs pipes 10323 * [356]22233 ICE with wrong number of template parameters 10324 * [357]22508 ICE after invalid operator new 10325 * [358]22545 ICE with pointer to class member & user defined 10326 conversion operator 10327 * [359]23528 Wrong default allocator in ext/hash_map 10328 * [360]23550 char_traits requirements/1.cc test bad math 10329 * [361]23586 Bad diagnostic for invalid namespace-name 10330 * [362]23624 ICE in invert_truthvalue, in fold-const.c 10331 * [363]23639 Bad error message: not a member of '<declaration error>' 10332 * [364]23797 ICE on typename outside template 10333 * [365]23965 Bogus error message: no matching function for call to 10334 'foo(<type error>)' 10335 * [366]24052 &#`label_decl' not supported by dump_expr#<expression 10336 error> 10337 * [367]24580 virtual base class cause exception not to be caught 10338 10339 Problems in generated debug information 10340 10341 * [368]24267 Bad DWARF for altivec vectors 10342 10343 Optimizations issues 10344 10345 * [369]17810 ICE in verify_local_live_at_start 10346 * [370]17860 Wrong generated code for loop with varying bound 10347 * [371]21709 ICE on compile-time complex NaN 10348 * [372]21964 broken tail call at -O2 or more 10349 * [373]22167 Strange optimization bug when using -Os 10350 * [374]22619 Compilation failure for real_const_1.f and 10351 real_const_2.f90 10352 * [375]23241 Invalid code generated for comparison of uchar to 255 10353 * [376]23478 Miscompilation due to reloading of a var that is also 10354 used in EH pad 10355 * [377]24470 segmentation fault in cc1plus when compiling with -O 10356 * [378]24950 ICE in operand_subword_force 10357 10358 Precompiled headers problems 10359 10360 * [379]14400 Cannot compile qt-x11-free-3.3.0 10361 * [380]14940 PCH largefile test fails on various platforms 10362 10363 Preprocessor bugs 10364 10365 * [381]20239 ICE on empty preprocessed input 10366 * [382]15220 "gcc -E -MM -MG" reports missing system headers in 10367 source directory 10368 10369 Testsuite issues 10370 10371 * [383]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on 10372 i686-pc-linux-gnu 10373 10374 Alpha specific 10375 10376 * [384]21888 bootstrap failure with linker relaxation enabled 10377 10378 ARM specific 10379 10380 * [385]15342 [arm-linux]: ICE in verify_local_live_at_start 10381 * [386]23985 Memory aliasing information incorrect in inlined memcpy 10382 10383 ColdFile specific 10384 10385 * [387]16719 Illegal move of byte into address register causes 10386 compiler to ICE 10387 10388 HPPA specific 10389 10390 * [388]21723 ICE while building libgfortran 10391 * [389]21841 -mhp-ld/-mgnu-ld documentation 10392 10393 IA-64 specific 10394 10395 * [390]23644 IA-64 hardware models and configuration options 10396 documentation error 10397 * [391]24718 Shared libgcc not used for linking by default 10398 10399 M68000 specific 10400 10401 * [392]18421 ICE in reload_cse_simplify_operands, in postreload.c 10402 10403 MIPS specific 10404 10405 * [393]20621 ICE in change_address_1, in emit-rtl.c 10406 10407 PowerPC and PowerPC64 specific 10408 10409 * [394]18583 error on valid code: const 10410 __attribute__((altivec(vector__))) doesn't work in arrays 10411 * [395]20191 ICE in reload_cse_simplify_operands 10412 * [396]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined 10413 * [397]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set 10414 * [398]23404 gij trashes args of functions with more than 8 fp args 10415 * [399]23539 C & C++ compiler generating misaligned references 10416 regardless of compiler flags 10417 * [400]24102 floatdisf2_internal2 broken 10418 * [401]24465 -mminimal-toc miscompilation of __thread vars 10419 10420 Solaris specific 10421 10422 * [402]19933 Problem with define of HUGE_VAL in math_c99 10423 * [403]21889 Native Solaris assembler cannot grok DTP-relative debug 10424 symbols 10425 10426 SPARC specific 10427 10428 * [404]19300 PCH failures on sparc-linux 10429 * [405]20301 Assembler labels have a leading "-" 10430 * [406]20673 C PCH testsuite assembly comparison failure 10431 10432 x86 and x86_64 specific 10433 10434 * [407]18582 ICE with arrays of type V2DF 10435 * [408]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2 10436 -fsched2-use-traces 10437 * [409]21716 ICE in reg-stack.c's swap_rtx_condition 10438 * [410]24315 amd64 fails -fpeephole2 10439 __________________________________________________________________ 10440 10441 GCC 3.4.6 10442 10443 This is the [411]list of problem reports (PRs) from GCC's bug tracking 10444 system that are known to be fixed in the 3.4.6 release. This list might 10445 not be complete (that is, it is possible that some PRs that have been 10446 fixed are not listed here). 10447 10448 10449 For questions related to the use of GCC, please consult these web 10450 pages and the [412]GCC manuals. If that fails, the 10451 [413]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10452 web pages and the development of GCC are welcome on our developer 10453 list at [414]gcc (a] gcc.gnu.org. All of [415]our lists have public 10454 archives. 10455 10456 Copyright (C) [416]Free Software Foundation, Inc. Verbatim copying and 10457 distribution of this entire article is permitted in any medium, 10458 provided this notice is preserved. 10459 10460 These pages are [417]maintained by the GCC team. Last modified 10461 2014-06-28[418]. 10462 10463 References 10464 10465 1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 10466 2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus 10467 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 10468 4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems 10469 5. http://gcc.gnu.org/gcc-3.4/mips-abi.html 10470 6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 10471 7. http://www.boost.org/ 10472 8. https://gcc.gnu.org/PR11953 10473 9. https://gcc.gnu.org/PR8361 10474 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other%20Builtins 10475 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209 10476 12. http://gcc.gnu.org/bugs/#cxx_rvalbind 10477 13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 10478 14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 10479 15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 10480 16. http://www.gnu.org/software/classpath/ 10481 17. http://www.eclipse.org/ 10482 18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html 10483 19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html 10484 20. http://h30097.www3.hp.com/docs/base_doc/DOCUMENTATION/V51A_HTML/ARH9MBTE/DTMNPLTN.HTM#normal-argument-list-structure 10485 21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 10486 22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html 10487 23. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 10488 24. http://gcc.gnu.org/gcc-3.4/mips-abi.html 10489 25. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html 10490 26. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 10491 27. https://gcc.gnu.org/bugzilla/buglist.cgi?short_desc_type=notregexp&short_desc=%5C%5B3%5C.4.*%5BRr%5Degression&target_milestone=3.4.0&bug_status=RESOLVED&resolution=FIXED 10492 28. https://gcc.gnu.org/PR10129 10493 29. https://gcc.gnu.org/PR14576 10494 30. https://gcc.gnu.org/PR14760 10495 31. https://gcc.gnu.org/PR14671 10496 32. https://gcc.gnu.org/PR15093 10497 33. https://gcc.gnu.org/PR15178 10498 34. https://gcc.gnu.org/PR12753 10499 35. https://gcc.gnu.org/PR13985 10500 36. https://gcc.gnu.org/PR14810 10501 37. https://gcc.gnu.org/PR14883 10502 38. https://gcc.gnu.org/PR15044 10503 39. https://gcc.gnu.org/PR15057 10504 40. https://gcc.gnu.org/PR15064 10505 41. https://gcc.gnu.org/PR15142 10506 42. https://gcc.gnu.org/PR15159 10507 43. https://gcc.gnu.org/PR15165 10508 44. https://gcc.gnu.org/PR15193 10509 45. https://gcc.gnu.org/PR15209 10510 46. https://gcc.gnu.org/PR15227 10511 47. https://gcc.gnu.org/PR15285 10512 48. https://gcc.gnu.org/PR15299 10513 49. https://gcc.gnu.org/PR15329 10514 50. https://gcc.gnu.org/PR15550 10515 51. https://gcc.gnu.org/PR15554 10516 52. https://gcc.gnu.org/PR15640 10517 53. https://gcc.gnu.org/PR15666 10518 54. https://gcc.gnu.org/PR15696 10519 55. https://gcc.gnu.org/PR15701 10520 56. https://gcc.gnu.org/PR15761 10521 57. https://gcc.gnu.org/PR15829 10522 58. https://gcc.gnu.org/PR14538 10523 59. https://gcc.gnu.org/PR12391 10524 60. https://gcc.gnu.org/PR14649 10525 61. https://gcc.gnu.org/PR15004 10526 62. https://gcc.gnu.org/PR15749 10527 63. https://gcc.gnu.org/PR10646 10528 64. https://gcc.gnu.org/PR12077 10529 65. https://gcc.gnu.org/PR13598 10530 66. https://gcc.gnu.org/PR14211 10531 67. https://gcc.gnu.org/PR14220 10532 68. https://gcc.gnu.org/PR14245 10533 69. https://gcc.gnu.org/PR14340 10534 70. https://gcc.gnu.org/PR14600 10535 71. https://gcc.gnu.org/PR14668 10536 72. https://gcc.gnu.org/PR14775 10537 73. https://gcc.gnu.org/PR14821 10538 74. https://gcc.gnu.org/PR14930 10539 75. https://gcc.gnu.org/PR14932 10540 76. https://gcc.gnu.org/PR14950 10541 77. https://gcc.gnu.org/PR14962 10542 78. https://gcc.gnu.org/PR14975 10543 79. https://gcc.gnu.org/PR15002 10544 80. https://gcc.gnu.org/PR15025 10545 81. https://gcc.gnu.org/PR15046 10546 82. https://gcc.gnu.org/PR15069 10547 83. https://gcc.gnu.org/PR15074 10548 84. https://gcc.gnu.org/PR15083 10549 85. https://gcc.gnu.org/PR15096 10550 86. https://gcc.gnu.org/PR15287 10551 87. https://gcc.gnu.org/PR15317 10552 88. https://gcc.gnu.org/PR15337 10553 89. https://gcc.gnu.org/PR15361 10554 90. https://gcc.gnu.org/PR15412 10555 91. https://gcc.gnu.org/PR15427 10556 92. https://gcc.gnu.org/PR15471 10557 93. https://gcc.gnu.org/PR15503 10558 94. https://gcc.gnu.org/PR15507 10559 95. https://gcc.gnu.org/PR15542 10560 96. https://gcc.gnu.org/PR15565 10561 97. https://gcc.gnu.org/PR15625 10562 98. https://gcc.gnu.org/PR15629 10563 99. https://gcc.gnu.org/PR15742 10564 100. https://gcc.gnu.org/PR15775 10565 101. https://gcc.gnu.org/PR15821 10566 102. https://gcc.gnu.org/PR15862 10567 103. https://gcc.gnu.org/PR15875 10568 104. https://gcc.gnu.org/PR15877 10569 105. https://gcc.gnu.org/PR15947 10570 106. https://gcc.gnu.org/PR16020 10571 107. https://gcc.gnu.org/PR16154 10572 108. https://gcc.gnu.org/PR16174 10573 109. https://gcc.gnu.org/PR14315 10574 110. https://gcc.gnu.org/PR15151 10575 111. https://gcc.gnu.org/PR7993 10576 112. https://gcc.gnu.org/PR15228 10577 113. https://gcc.gnu.org/PR15345 10578 114. https://gcc.gnu.org/PR15945 10579 115. https://gcc.gnu.org/PR15526 10580 116. https://gcc.gnu.org/PR14690 10581 117. https://gcc.gnu.org/PR15112 10582 118. https://gcc.gnu.org/PR15067 10583 119. https://gcc.gnu.org/PR1963 10584 120. https://gcc.gnu.org/PR15717 10585 121. https://gcc.gnu.org/PR14782 10586 122. https://gcc.gnu.org/PR14828 10587 123. https://gcc.gnu.org/PR15202 10588 124. https://gcc.gnu.org/PR14610 10589 125. https://gcc.gnu.org/PR14813 10590 126. https://gcc.gnu.org/PR14857 10591 127. https://gcc.gnu.org/PR15598 10592 128. https://gcc.gnu.org/PR15653 10593 129. https://gcc.gnu.org/PR15189 10594 130. https://gcc.gnu.org/PR15331 10595 131. https://gcc.gnu.org/PR16144 10596 132. https://gcc.gnu.org/PR16176 10597 133. https://gcc.gnu.org/PR11591 10598 134. https://gcc.gnu.org/PR12028 10599 135. https://gcc.gnu.org/PR14478 10600 136. https://gcc.gnu.org/PR14567 10601 137. https://gcc.gnu.org/PR14715 10602 138. https://gcc.gnu.org/PR14902 10603 139. https://gcc.gnu.org/PR14924 10604 140. https://gcc.gnu.org/PR14960 10605 141. https://gcc.gnu.org/PR15106 10606 142. https://gcc.gnu.org/PR16026 10607 143. https://gcc.gnu.org/PR15191 10608 144. https://gcc.gnu.org/PR15662 10609 145. https://gcc.gnu.org/PR15054 10610 146. https://gcc.gnu.org/PR15783 10611 147. https://gcc.gnu.org/PR15626 10612 148. https://gcc.gnu.org/PR14326 10613 149. https://gcc.gnu.org/PR14723 10614 150. https://gcc.gnu.org/PR15290 10615 151. https://gcc.gnu.org/PR15250 10616 152. https://gcc.gnu.org/PR15551 10617 153. https://gcc.gnu.org/PR8309 10618 154. https://gcc.gnu.org/PR13250 10619 155. https://gcc.gnu.org/PR13803 10620 156. https://gcc.gnu.org/PR14093 10621 157. https://gcc.gnu.org/PR14457 10622 158. https://gcc.gnu.org/PR14542 10623 159. https://gcc.gnu.org/PR15100 10624 160. https://gcc.gnu.org/PR15296 10625 161. https://gcc.gnu.org/PR15396 10626 162. https://gcc.gnu.org/PR15782 10627 163. https://gcc.gnu.org/PR11610 10628 164. https://gcc.gnu.org/PR15488 10629 165. https://gcc.gnu.org/PR15489 10630 166. https://gcc.gnu.org/PR13928 10631 167. https://gcc.gnu.org/PR14150 10632 168. https://gcc.gnu.org/PR14949 10633 169. https://gcc.gnu.org/PR15123 10634 170. https://gcc.gnu.org/PR16469 10635 171. https://gcc.gnu.org/PR16344 10636 172. https://gcc.gnu.org/PR16842 10637 173. https://gcc.gnu.org/PR12608 10638 174. https://gcc.gnu.org/PR14492 10639 175. https://gcc.gnu.org/PR15461 10640 176. https://gcc.gnu.org/PR15890 10641 177. https://gcc.gnu.org/PR16180 10642 178. https://gcc.gnu.org/PR16224 10643 179. https://gcc.gnu.org/PR16408 10644 180. https://gcc.gnu.org/PR16529 10645 181. https://gcc.gnu.org/PR16698 10646 182. https://gcc.gnu.org/PR16706 10647 183. https://gcc.gnu.org/PR16810 10648 184. https://gcc.gnu.org/PR16851 10649 185. https://gcc.gnu.org/PR16870 10650 186. https://gcc.gnu.org/PR16904 10651 187. https://gcc.gnu.org/PR16905 10652 188. https://gcc.gnu.org/PR16964 10653 189. https://gcc.gnu.org/PR17068 10654 190. https://gcc.gnu.org/PR16366 10655 191. https://gcc.gnu.org/PR15345 10656 192. https://gcc.gnu.org/PR16590 10657 193. https://gcc.gnu.org/PR16693 10658 194. https://gcc.gnu.org/PR17078 10659 195. https://gcc.gnu.org/PR13956 10660 196. https://gcc.gnu.org/PR16684 10661 197. https://gcc.gnu.org/PR12658 10662 198. https://gcc.gnu.org/PR13092 10663 199. https://gcc.gnu.org/PR15320 10664 200. https://gcc.gnu.org/PR16246 10665 201. https://gcc.gnu.org/PR16273 10666 202. https://gcc.gnu.org/PR16401 10667 203. https://gcc.gnu.org/PR16411 10668 204. https://gcc.gnu.org/PR16489 10669 205. https://gcc.gnu.org/PR16618 10670 206. https://gcc.gnu.org/PR16637 10671 207. https://gcc.gnu.org/PR16717 10672 208. https://gcc.gnu.org/PR16813 10673 209. https://gcc.gnu.org/PR16853 10674 210. https://gcc.gnu.org/PR16889 10675 211. https://gcc.gnu.org/PR16959 10676 212. https://gcc.gnu.org/PR7587 10677 213. https://gcc.gnu.org/PR16473 10678 214. https://gcc.gnu.org/PR16478 10679 215. https://gcc.gnu.org/PR10695 10680 216. https://gcc.gnu.org/PR16974 10681 217. https://gcc.gnu.org/PR16298 10682 218. https://gcc.gnu.org/PR17113 10683 219. https://gcc.gnu.org/PR14697 10684 220. https://gcc.gnu.org/PR15869 10685 221. https://gcc.gnu.org/PR16325 10686 222. https://gcc.gnu.org/PR16357 10687 223. https://gcc.gnu.org/PR16380 10688 224. https://gcc.gnu.org/PR16407 10689 225. https://gcc.gnu.org/PR16643 10690 226. https://gcc.gnu.org/PR15927 10691 227. https://gcc.gnu.org/PR15948 10692 228. https://gcc.gnu.org/PR17019 10693 229. https://gcc.gnu.org/PR16130 10694 230. https://gcc.gnu.org/PR16142 10695 231. https://gcc.gnu.org/PR16278 10696 232. https://gcc.gnu.org/PR16414 10697 233. https://gcc.gnu.org/PR16445 10698 234. https://gcc.gnu.org/PR16490 10699 235. https://gcc.gnu.org/PR16683 10700 236. https://gcc.gnu.org/PR16195 10701 237. https://gcc.gnu.org/PR16239 10702 238. https://gcc.gnu.org/PR16199 10703 239. https://gcc.gnu.org/PR16416 10704 240. https://gcc.gnu.org/PR16430 10705 241. https://gcc.gnu.org/PR16379 10706 242. https://gcc.gnu.org/PR17093 10707 243. https://gcc.gnu.org/PR17119 10708 244. https://gcc.gnu.org/PR15928 10709 245. https://gcc.gnu.org/PR16210 10710 246. https://gcc.gnu.org/PR15488 10711 247. https://gcc.gnu.org/PR16250 10712 248. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3 10713 249. https://gcc.gnu.org/PR17369 10714 250. https://gcc.gnu.org/PR17850 10715 251. https://gcc.gnu.org/PR13948 10716 252. https://gcc.gnu.org/PR14492 10717 253. https://gcc.gnu.org/PR16301 10718 254. https://gcc.gnu.org/PR16566 10719 255. https://gcc.gnu.org/PR17023 10720 256. https://gcc.gnu.org/PR17027 10721 257. https://gcc.gnu.org/PR17524 10722 258. https://gcc.gnu.org/PR17826 10723 259. https://gcc.gnu.org/PR15526 10724 260. https://gcc.gnu.org/PR16999 10725 261. https://gcc.gnu.org/PR17503 10726 262. https://gcc.gnu.org/PR17581 10727 263. https://gcc.gnu.org/PR18129 10728 264. https://gcc.gnu.org/PR10975 10729 265. https://gcc.gnu.org/PR11722 10730 266. https://gcc.gnu.org/PR14534 10731 267. https://gcc.gnu.org/PR15172 10732 268. https://gcc.gnu.org/PR15786 10733 269. https://gcc.gnu.org/PR16162 10734 270. https://gcc.gnu.org/PR16612 10735 271. https://gcc.gnu.org/PR16715 10736 272. https://gcc.gnu.org/PR16848 10737 273. https://gcc.gnu.org/PR17132 10738 274. https://gcc.gnu.org/PR17259 10739 275. https://gcc.gnu.org/PR17327 10740 276. https://gcc.gnu.org/PR17393 10741 277. https://gcc.gnu.org/PR17501 10742 278. https://gcc.gnu.org/PR17537 10743 279. https://gcc.gnu.org/PR17585 10744 280. https://gcc.gnu.org/PR17821 10745 281. https://gcc.gnu.org/PR17829 10746 282. https://gcc.gnu.org/PR17851 10747 283. https://gcc.gnu.org/PR17976 10748 284. https://gcc.gnu.org/PR18020 10749 285. https://gcc.gnu.org/PR18093 10750 286. https://gcc.gnu.org/PR18140 10751 287. https://gcc.gnu.org/PR17541 10752 288. https://gcc.gnu.org/PR17853 10753 289. https://gcc.gnu.org/PR17245 10754 290. https://gcc.gnu.org/PR17167 10755 291. https://gcc.gnu.org/PR17277 10756 292. https://gcc.gnu.org/PR17505 10757 293. https://gcc.gnu.org/PR17684 10758 294. https://gcc.gnu.org/PR17384 10759 295. https://gcc.gnu.org/PR17770 10760 296. https://gcc.gnu.org/PR11476 10761 297. https://gcc.gnu.org/PR14064 10762 298. https://gcc.gnu.org/PR14678 10763 299. https://gcc.gnu.org/PR15583 10764 300. https://gcc.gnu.org/PR15790 10765 301. https://gcc.gnu.org/PR15886 10766 302. https://gcc.gnu.org/PR16884 10767 303. https://gcc.gnu.org/PR13841 10768 304. https://gcc.gnu.org/PR15860 10769 305. https://gcc.gnu.org/PR17465 10770 306. https://gcc.gnu.org/PR17469 10771 307. https://gcc.gnu.org/PR18138 10772 308. https://gcc.gnu.org/PR15498 10773 309. https://gcc.gnu.org/PR15747 10774 310. https://gcc.gnu.org/PR16406 10775 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4 10776 312. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5 10777 313. https://gcc.gnu.org/PR24688 10778 314. https://gcc.gnu.org/PR17188 10779 315. https://gcc.gnu.org/PR20187 10780 316. https://gcc.gnu.org/PR21873 10781 317. https://gcc.gnu.org/PR21899 10782 318. https://gcc.gnu.org/PR22061 10783 319. https://gcc.gnu.org/PR22208 10784 320. https://gcc.gnu.org/PR22458 10785 321. https://gcc.gnu.org/PR22589 10786 322. https://gcc.gnu.org/PR24101 10787 323. https://gcc.gnu.org/PR10611 10788 324. https://gcc.gnu.org/PR13377 10789 325. https://gcc.gnu.org/PR16002 10790 326. https://gcc.gnu.org/PR17413 10791 327. https://gcc.gnu.org/PR17609 10792 328. https://gcc.gnu.org/PR17618 10793 329. https://gcc.gnu.org/PR18124 10794 330. https://gcc.gnu.org/PR18155 10795 331. https://gcc.gnu.org/PR18177 10796 332. https://gcc.gnu.org/PR18368 10797 333. https://gcc.gnu.org/PR18378 10798 334. https://gcc.gnu.org/PR18466 10799 335. https://gcc.gnu.org/PR18512 10800 336. https://gcc.gnu.org/PR18545 10801 337. https://gcc.gnu.org/PR18738 10802 338. https://gcc.gnu.org/PR18803 10803 339. https://gcc.gnu.org/PR19004 10804 340. https://gcc.gnu.org/PR19208 10805 341. https://gcc.gnu.org/PR19253 10806 342. https://gcc.gnu.org/PR19608 10807 343. https://gcc.gnu.org/PR19884 10808 344. https://gcc.gnu.org/PR20153 10809 345. https://gcc.gnu.org/PR20563 10810 346. https://gcc.gnu.org/PR20789 10811 347. https://gcc.gnu.org/PR21336 10812 348. https://gcc.gnu.org/PR21768 10813 349. https://gcc.gnu.org/PR21853 10814 350. https://gcc.gnu.org/PR21903 10815 351. https://gcc.gnu.org/PR21983 10816 352. https://gcc.gnu.org/PR21987 10817 353. https://gcc.gnu.org/PR22153 10818 354. https://gcc.gnu.org/PR22172 10819 355. https://gcc.gnu.org/PR21286 10820 356. https://gcc.gnu.org/PR22233 10821 357. https://gcc.gnu.org/PR22508 10822 358. https://gcc.gnu.org/PR22545 10823 359. https://gcc.gnu.org/PR23528 10824 360. https://gcc.gnu.org/PR23550 10825 361. https://gcc.gnu.org/PR23586 10826 362. https://gcc.gnu.org/PR23624 10827 363. https://gcc.gnu.org/PR23639 10828 364. https://gcc.gnu.org/PR23797 10829 365. https://gcc.gnu.org/PR23965 10830 366. https://gcc.gnu.org/PR24052 10831 367. https://gcc.gnu.org/PR24580 10832 368. https://gcc.gnu.org/PR24267 10833 369. https://gcc.gnu.org/PR17810 10834 370. https://gcc.gnu.org/PR17860 10835 371. https://gcc.gnu.org/PR21709 10836 372. https://gcc.gnu.org/PR21964 10837 373. https://gcc.gnu.org/PR22167 10838 374. https://gcc.gnu.org/PR22619 10839 375. https://gcc.gnu.org/PR23241 10840 376. https://gcc.gnu.org/PR23478 10841 377. https://gcc.gnu.org/PR24470 10842 378. https://gcc.gnu.org/PR24950 10843 379. https://gcc.gnu.org/PR14400 10844 380. https://gcc.gnu.org/PR14940 10845 381. https://gcc.gnu.org/PR20239 10846 382. https://gcc.gnu.org/PR15220 10847 383. https://gcc.gnu.org/PR19275 10848 384. https://gcc.gnu.org/PR21888 10849 385. https://gcc.gnu.org/PR15342 10850 386. https://gcc.gnu.org/PR23985 10851 387. https://gcc.gnu.org/PR16719 10852 388. https://gcc.gnu.org/PR21723 10853 389. https://gcc.gnu.org/PR21841 10854 390. https://gcc.gnu.org/PR23644 10855 391. https://gcc.gnu.org/PR24718 10856 392. https://gcc.gnu.org/PR18421 10857 393. https://gcc.gnu.org/PR20621 10858 394. https://gcc.gnu.org/PR18583 10859 395. https://gcc.gnu.org/PR20191 10860 396. https://gcc.gnu.org/PR22083 10861 397. https://gcc.gnu.org/PR23070 10862 398. https://gcc.gnu.org/PR23404 10863 399. https://gcc.gnu.org/PR23539 10864 400. https://gcc.gnu.org/PR24102 10865 401. https://gcc.gnu.org/PR24465 10866 402. https://gcc.gnu.org/PR19933 10867 403. https://gcc.gnu.org/PR21889 10868 404. https://gcc.gnu.org/PR19300 10869 405. https://gcc.gnu.org/PR20301 10870 406. https://gcc.gnu.org/PR20673 10871 407. https://gcc.gnu.org/PR18582 10872 408. https://gcc.gnu.org/PR19340 10873 409. https://gcc.gnu.org/PR21716 10874 410. https://gcc.gnu.org/PR24315 10875 411. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6 10876 412. https://gcc.gnu.org/onlinedocs/ 10877 413. mailto:gcc-help (a] gcc.gnu.org 10878 414. mailto:gcc (a] gcc.gnu.org 10879 415. https://gcc.gnu.org/lists.html 10880 416. http://www.fsf.org/ 10881 417. https://gcc.gnu.org/about.html 10882 418. http://validator.w3.org/check/referer 10883 ====================================================================== 10884 http://gcc.gnu.org/gcc-3.3/index.html 10885 10886 GCC 3.3 Release Series 10887 10888 May 03, 2005 10889 10890 The [1]GNU project and the GCC developers are pleased to announce the 10891 release of GCC 3.3.6. 10892 10893 This release is a bug-fix release, containing fixes for regressions in 10894 GCC 3.3.5 relative to previous releases of GCC. 10895 10896 This release is the last of the series 3.3.x. 10897 10898 The GCC 3.3 release series includes numerous [2]new features, 10899 improvements, bug fixes, and other changes, thanks to an [3]amazing 10900 group of volunteers. 10901 10902 Release History 10903 10904 GCC 3.3.6 10905 May 3, 2005 ([4]changes) 10906 10907 GCC 3.3.5 10908 September 30, 2004 ([5]changes) 10909 10910 GCC 3.3.4 10911 May 31, 2004 ([6]changes) 10912 10913 GCC 3.3.3 10914 February 14, 2004 ([7]changes) 10915 10916 GCC 3.3.2 10917 October 16, 2003 ([8]changes) 10918 10919 GCC 3.3.1 10920 August 8, 2003 ([9]changes) 10921 10922 GCC 3.3 10923 May 14, 2003 ([10]changes) 10924 10925 References and Acknowledgements 10926 10927 GCC used to stand for the GNU C Compiler, but since the compiler 10928 supports several other languages aside from C, it now stands for the 10929 GNU Compiler Collection. 10930 10931 A list of [11]successful builds is updated as new information becomes 10932 available. 10933 10934 The GCC developers would like to thank the numerous people that have 10935 contributed new features, improvements, bug fixes, and other changes as 10936 well as test results to GCC. This [12]amazing group of volunteers is 10937 what makes GCC successful. 10938 10939 For additional information about GCC please refer to the [13]GCC 10940 project web site or contact the [14]GCC development mailing list. 10941 10942 To obtain GCC please use [15]our mirror sites, or our CVS server. 10943 10944 10945 For questions related to the use of GCC, please consult these web 10946 pages and the [16]GCC manuals. If that fails, the 10947 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10948 web pages and the development of GCC are welcome on our developer 10949 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 10950 archives. 10951 10952 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 10953 distribution of this entire article is permitted in any medium, 10954 provided this notice is preserved. 10955 10956 These pages are [21]maintained by the GCC team. Last modified 10957 2014-06-28[22]. 10958 10959 References 10960 10961 1. http://www.gnu.org/ 10962 2. http://gcc.gnu.org/gcc-3.3/changes.html 10963 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 10964 4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 10965 5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5 10966 6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4 10967 7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3 10968 8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2 10969 9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1 10970 10. http://gcc.gnu.org/gcc-3.3/changes.html 10971 11. http://gcc.gnu.org/gcc-3.3/buildstat.html 10972 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 10973 13. http://gcc.gnu.org/index.html 10974 14. mailto:gcc (a] gcc.gnu.org 10975 15. http://gcc.gnu.org/mirrors.html 10976 16. https://gcc.gnu.org/onlinedocs/ 10977 17. mailto:gcc-help (a] gcc.gnu.org 10978 18. mailto:gcc (a] gcc.gnu.org 10979 19. https://gcc.gnu.org/lists.html 10980 20. http://www.fsf.org/ 10981 21. https://gcc.gnu.org/about.html 10982 22. http://validator.w3.org/check/referer 10983 ====================================================================== 10984 http://gcc.gnu.org/gcc-3.3/changes.html 10985 10986 GCC 3.3 Release Series 10987 Changes, New Features, and Fixes 10988 10989 The latest release in the 3.3 release series is [1]GCC 3.3.6. 10990 10991 Caveats 10992 10993 * The preprocessor no longer accepts multi-line string literals. They 10994 were deprecated in 3.0, 3.1, and 3.2. 10995 * The preprocessor no longer supports the -A- switch when appearing 10996 alone. -A- followed by an assertion is still supported. 10997 * Support for all the systems [2]obsoleted in GCC 3.1 has been 10998 removed from GCC 3.3. See below for a [3]list of systems which are 10999 obsoleted in this release. 11000 * Checking for null format arguments has been decoupled from the rest 11001 of the format checking mechanism. Programs which use the format 11002 attribute may regain this functionality by using the new [4]nonnull 11003 function attribute. Note that all functions for which GCC has a 11004 built-in format attribute, an appropriate built-in nonnull 11005 attribute is also applied. 11006 * The DWARF (version 1) debugging format has been deprecated and will 11007 be removed in a future version of GCC. Version 2 of the DWARF 11008 debugging format will continue to be supported for the foreseeable 11009 future. 11010 * The C and Objective-C compilers no longer accept the "Naming Types" 11011 extension (typedef foo = bar); it was already unavailable in C++. 11012 Code which uses it will need to be changed to use the "typeof" 11013 extension instead: typedef typeof(bar) foo. (We have removed this 11014 extension without a period of deprecation because it has caused the 11015 compiler to crash since version 3.0 and no one noticed until very 11016 recently. Thus we conclude it is not in widespread use.) 11017 * The -traditional C compiler option has been removed. It was 11018 deprecated in 3.1 and 3.2. (Traditional preprocessing remains 11019 available.) The <varargs.h> header, used for writing variadic 11020 functions in traditional C, still exists but will produce an error 11021 message if used. 11022 * GCC 3.3.1 automatically places zero-initialized variables in the 11023 .bss section on some operating systems. Versions of GNU Emacs up to 11024 (and including) 21.3 will not work correctly when using this 11025 optimization; you can use -fno-zero-initialized-in-bss to disable 11026 it. 11027 11028 General Optimizer Improvements 11029 11030 * A new scheme for accurately describing processor pipelines, the 11031 [5]DFA scheduler, has been added. 11032 * Pavel Nejedly, Charles University Prague, has contributed new file 11033 format used by the edge coverage profiler (-fprofile-arcs). 11034 The new format is robust and diagnoses common mistakes where 11035 profiles from different versions (or compilations) of the program 11036 are combined resulting in nonsensical profiles and slow code to 11037 produced with profile feedback. Additionally this format allows 11038 extra data to be gathered. Currently, overall statistics are 11039 produced helping optimizers to identify hot spots of a program 11040 globally replacing the old intra-procedural scheme and resulting in 11041 better code. Note that the gcov tool from older GCC versions will 11042 not be able to parse the profiles generated by GCC 3.3 and vice 11043 versa. 11044 * Jan Hubicka, SuSE Labs, has contributed a new superblock formation 11045 pass enabled using -ftracer. This pass simplifies the control flow 11046 of functions allowing other optimizations to do better job. 11047 He also contributed the function reordering pass 11048 (-freorder-functions) to optimize function placement using profile 11049 feedback. 11050 11051 New Languages and Language specific improvements 11052 11053 C/ObjC/C++ 11054 11055 * The preprocessor now accepts directives within macro arguments. It 11056 processes them just as if they had not been within macro arguments. 11057 * The separate ISO and traditional preprocessors have been completely 11058 removed. The front end handles either type of preprocessed output 11059 if necessary. 11060 * In C99 mode preprocessor arithmetic is done in the precision of the 11061 target's intmax_t, as required by that standard. 11062 * The preprocessor can now copy comments inside macros to the output 11063 file when the macro is expanded. This feature, enabled using the 11064 -CC option, is intended for use by applications which place 11065 metadata or directives inside comments, such as lint. 11066 * The method of constructing the list of directories to be searched 11067 for header files has been revised. If a directory named by a -I 11068 option is a standard system include directory, the option is 11069 ignored to ensure that the default search order for system 11070 directories and the special treatment of system header files are 11071 not defeated. 11072 * A few more [6]ISO C99 features now work correctly. 11073 * A new function attribute, nonnull, has been added which allows 11074 pointer arguments to functions to be specified as requiring a 11075 non-null value. The compiler currently uses this information to 11076 issue a warning when it detects a null value passed in such an 11077 argument slot. 11078 * A new type attribute, may_alias, has been added. Accesses to 11079 objects with types with this attribute are not subjected to 11080 type-based alias analysis, but are instead assumed to be able to 11081 alias any other type of objects, just like the char type. 11082 11083 C++ 11084 11085 * Type based alias analysis has been implemented for C++ aggregate 11086 types. 11087 11088 Objective-C 11089 11090 * Generate an error if Objective-C objects are passed by value in 11091 function and method calls. 11092 * When -Wselector is used, check the whole list of selectors at the 11093 end of compilation, and emit a warning if a @selector() is not 11094 known. 11095 * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime. 11096 * No longer need to include objc/objc-class.h to compile self calls 11097 in class methods (NeXT runtime only). 11098 * New -Wundeclared-selector option. 11099 * Removed selector bloating which was causing object files to be 10% 11100 bigger on average (GNU runtime only). 11101 * Using at run time @protocol() objects has been fixed in certain 11102 situations (GNU runtime only). 11103 * Type checking has been fixed and improved in many situations 11104 involving protocols. 11105 11106 Java 11107 11108 * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK 11109 1.4) API. 11110 * The JDK 1.4 assert facility has been implemented. 11111 * The bytecode interpreter is now direct threaded and thus faster. 11112 11113 Fortran 11114 11115 * Fortran improvements are listed in [7]the Fortran documentation. 11116 11117 Ada 11118 11119 * Ada tasking now works with glibc 2.3.x threading libraries. 11120 11121 New Targets and Target Specific Improvements 11122 11123 * The following changes have been made to the HP-PA port: 11124 + The port now defaults to scheduling for the PA8000 series of 11125 processors. 11126 + Scheduling support for the PA7300 processor has been added. 11127 + The 32-bit port now supports weak symbols under HP-UX 11. 11128 + The handling of initializers and finalizers has been improved 11129 under HP-UX 11. The 64-bit port no longer uses collect2. 11130 + Dwarf2 EH support has been added to the 32-bit GNU/Linux port. 11131 + ABI fixes to correct the passing of small structures by value. 11132 * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to 11133 use the DFA processor pipeline description. 11134 * The following NetBSD configurations for the SuperH processor family 11135 have been added: 11136 + SH3, big-endian, sh-*-netbsdelf* 11137 + SH3, little-endian, shle-*-netbsdelf* 11138 + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd* 11139 + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd* 11140 + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd* 11141 + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd* 11142 * The following changes have been made to the IA-32/x86-64 port: 11143 + SSE2 and 3dNOW! intrinsics are now supported. 11144 + Support for thread local storage has been added to the IA-32 11145 and x86-64 ports. 11146 + The x86-64 port has been significantly improved. 11147 * The following changes have been made to the MIPS port: 11148 + All configurations now accept the -mabi switch. Note that you 11149 will need appropriate multilibs for this option to work 11150 properly. 11151 + ELF configurations will always pass an ABI flag to the 11152 assembler, except when the MIPS EABI is selected. 11153 + -mabi=64 no longer selects MIPS IV code. 11154 + The -mcpu option, which was deprecated in 3.1 and 3.2, has 11155 been removed from this release. 11156 + -march now changes the core ISA level. In previous releases, 11157 it would change the use of processor-specific extensions, but 11158 would leave the core ISA unchanged. For example, mips64-elf 11159 -march=r8000 will now generate MIPS IV code. 11160 + Under most configurations, -mipsN now acts as a synonym for 11161 -march. 11162 + There are some new preprocessor macros to describe the -march 11163 and -mtune settings. See the documentation of those options 11164 for details. 11165 + Support for the NEC VR-Series processors has been added. This 11166 includes the 54xx, 5500, and 41xx series. 11167 + Support for the Sandcraft sr71k processor has been added. 11168 * The following changes have been made to the S/390 port: 11169 + Support to build the Java runtime libraries has been added. 11170 Java is now enabled by default on s390-*-linux* and 11171 s390x-*-linux* targets. 11172 + Multilib support for the s390x-*-linux* target has been added; 11173 this allows to build 31-bit binaries using the -m31 option. 11174 + Support for thread local storage has been added. 11175 + Inline assembler code may now use the 'Q' constraint to 11176 specify memory operands without index register. 11177 + Various platform-specific performance improvements have been 11178 implemented; in particular, the compiler now uses the BRANCH 11179 ON COUNT family of instructions and makes more frequent use of 11180 the TEST UNDER MASK family of instructions. 11181 * The following changes have been made to the PowerPC port: 11182 + Support for IBM Power4 processor added. 11183 + Support for Motorola e500 SPE added. 11184 + Support for AIX 5.2 added. 11185 + Function and Data sections now supported on AIX. 11186 + Sibcall optimizations added. 11187 * The support for H8 Tiny is added to the H8/300 port with -mn. 11188 11189 Obsolete Systems 11190 11191 Support for a number of older systems has been declared obsolete in GCC 11192 3.3. Unless there is activity to revive them, the next release of GCC 11193 will have their sources permanently removed. 11194 11195 All configurations of the following processor architectures have been 11196 declared obsolete: 11197 * Matsushita MN10200, mn10200-*-* 11198 * Motorola 88000, m88k-*-* 11199 * IBM ROMP, romp-*-* 11200 11201 Also, some individual systems have been obsoleted: 11202 * Alpha 11203 + Interix, alpha*-*-interix* 11204 + Linux libc1, alpha*-*-linux*libc1* 11205 + Linux ECOFF, alpha*-*-linux*ecoff* 11206 * ARM 11207 + Generic a.out, arm*-*-aout* 11208 + Conix, arm*-*-conix* 11209 + "Old ABI," arm*-*-oabi 11210 + StrongARM/COFF, strongarm-*-coff* 11211 * HPPA (PA-RISC) 11212 + Generic OSF, hppa1.0-*-osf* 11213 + Generic BSD, hppa1.0-*-bsd* 11214 + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]* 11215 + HiUX, hppa*-*-hiux* 11216 + Mach Lites, hppa*-*-lites* 11217 * Intel 386 family 11218 + Windows NT 3.x, i?86-*-win32 11219 * MC68000 family 11220 + HP systems, m68000-hp-bsd* and m68k-hp-bsd* 11221 + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and 11222 m68k-sun-mach* 11223 + AT&T systems, m68000-att-sysv* 11224 + Atari systems, m68k-atari-sysv* 11225 + Motorola systems, m68k-motorola-sysv* 11226 + NCR systems, m68k-ncr-sysv* 11227 + Plexus systems, m68k-plexus-sysv* 11228 + Commodore systems, m68k-cbm-sysv* 11229 + Citicorp TTI, m68k-tti-* 11230 + Unos, m68k-crds-unos* 11231 + Concurrent RTU, m68k-ccur-rtu* 11232 + Linux a.out, m68k-*-linux*aout* 11233 + Linux libc1, m68k-*-linux*libc1* 11234 + pSOS, m68k-*-psos* 11235 * MIPS 11236 + Generic ECOFF, mips*-*-ecoff* 11237 + SINIX, mips-sni-sysv4 11238 + Orion RTEMS, mips64orion-*-rtems* 11239 * National Semiconductor 32000 11240 + OpenBSD, ns32k-*-openbsd* 11241 * POWER (aka RS/6000) and PowerPC 11242 + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]* 11243 + Bull BOSX, rs6000-bull-bosx 11244 + Generic Mach, rs6000-*-mach* 11245 + Generic SysV, powerpc*-*-sysv* 11246 + Linux libc1, powerpc*-*-linux*libc1* 11247 * Sun SPARC 11248 + Generic a.out, sparc-*-aout*, sparclet-*-aout*, 11249 sparclite-*-aout*, and sparc86x-*-aout* 11250 + NetBSD a.out, sparc-*-netbsd*aout* 11251 + Generic BSD, sparc-*-bsd* 11252 + ChorusOS, sparc-*-chorusos* 11253 + Linux a.out, sparc-*-linux*aout* 11254 + Linux libc1, sparc-*-linux*libc1* 11255 + LynxOS, sparc-*-lynxos* 11256 + Solaris on HAL hardware, sparc-hal-solaris2* 11257 + SunOS versions 3 and 4, sparc-*-sunos[34]* 11258 * NEC V850 11259 + RTEMS, v850-*-rtems* 11260 * VAX 11261 + VMS, vax-*-vms* 11262 11263 Documentation improvements 11264 11265 Other significant improvements 11266 11267 * Almost all front-end dependencies in the compiler have been 11268 separated out into a set of language hooks. This should make adding 11269 a new front end clearer and easier. 11270 * One effect of removing the separate preprocessor is a small 11271 increase in the robustness of the compiler in general, and the 11272 maintainability of target descriptions. Previously target-specific 11273 built-in macros and others, such as __FAST_MATH__, had to be 11274 handled with so-called specs that were hard to maintain. Often they 11275 would fail to behave properly when conflicting options were 11276 supplied on the command line, and define macros in the user's 11277 namespace even when strict ISO compliance was requested. 11278 Integrating the preprocessor has cleanly solved these issues. 11279 * The Makefile suite now supports redirection of make install by 11280 means of the variable DESTDIR. 11281 __________________________________________________________________ 11282 11283 GCC 3.3 11284 11285 Detailed release notes for the GCC 3.3 release follow. 11286 11287 Bug Fixes 11288 11289 bootstrap failures 11290 11291 * [8]10140 cross compiler build failures: missing __mempcpy (DUP: 11292 [9]10198,[10]10338) 11293 11294 Internal compiler errors (multi-platform) 11295 11296 * [11]3581 large string causes segmentation fault in cc1 11297 * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler 11298 * [13]5533 (c++) ICE when processing std::accumulate(begin, end, 11299 init, invalid_op) 11300 * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out 11301 * [15]6412 (c++) ICE in retrieve_specialization 11302 * [16]6620 (c++) partial template specialization causes an ICE 11303 (segmentation fault) 11304 * [17]6663 (c++) ICE with attribute aligned 11305 * [18]7068 ICE with incomplete types 11306 * [19]7083 (c++) ICE using -gstabs with dodgy class derivation 11307 * [20]7647 (c++) ICE when data member has the name of the enclosing 11308 class 11309 * [21]7675 ICE in fixup_var_refs_1 11310 * [22]7718 'complex' template instantiation causes ICE 11311 * [23]8116 (c++) ICE in member template function 11312 * [24]8358 (ada) Ada compiler accesses freed memory, crashes 11313 * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation 11314 fault 11315 * [26]8564 (c++) ICE in find_function_data, in function.c 11316 * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c 11317 * [28]8766 (c++) ICE after failed initialization of static template 11318 variable 11319 * [29]8803 ICE in instantiate_virtual_regs_1, in function.c 11320 * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set 11321 * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class 11322 definition 11323 * [32]9216 (c++) ICE on missing template parameter 11324 * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c 11325 * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO 11326 loop 11327 * [35]9429 (c++) ICE in template instantiation with a pointered new 11328 operator 11329 * [36]9516 Internal error when using a big array 11330 * [37]9600 (c++) ICE with typedefs in template class 11331 * [38]9629 (c++) virtual inheritance segfault 11332 * [39]9672 (c++) ICE: Error reporting routines re-entered 11333 * [40]9749 (c++) ICE in write_expression on invalid function 11334 prototype 11335 * [41]9794 (fortran) ICE: floating point exception during constant 11336 folding 11337 * [42]9829 (c++) Missing colon in nested namespace usage causes ICE 11338 * [43]9916 (c++) ICE with noreturn function in ?: statement 11339 * [44]9936 ICE with local function and variable-length 2d array 11340 * [45]10262 (c++) cc1plus crashes with large generated code 11341 * [46]10278 (c++) ICE in parser for invalid code 11342 * [47]10446 (c++) ICE on definition of nonexistent member function of 11343 nested class in a class template 11344 * [48]10451 (c++) ICE in grokdeclarator on spurious mutable 11345 declaration 11346 * [49]10506 (c++) ICE in build_new at cp/init.c with 11347 -fkeep-inline-functions and multiple inheritance 11348 * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the 11349 precision of the declared type 11350 11351 Optimization bugs 11352 11353 * [51]2001 Inordinately long compile times in reload CSE regs 11354 * [52]2391 Exponential compilation time explosion in combine 11355 * [53]2960 Duplicate loop conditions even with -Os 11356 * [54]4046 redundant conditional branch 11357 * [55]6405 Loop-unrolling related performance regressions 11358 * [56]6798 very long compile time with large case-statement 11359 * [57]6871 const objects shouldn't be moved to .bss 11360 * [58]6909 problem w/ -Os on modified loop-2c.c test case 11361 * [59]7189 gcc -O2 -Wall does not print ``control reaches end of 11362 non-void function'' warning 11363 * [60]7642 optimization problem with signbit() 11364 * [61]8634 incorrect code for inlining of memcpy under -O2 11365 * [62]8750 Cygwin prolog generation erroneously emitting __alloca as 11366 regular function call 11367 11368 C front end 11369 11370 * [63]2161 long if-else cascade overflows parser stack 11371 * [64]4319 short accepted on typedef'd char 11372 * [65]8602 incorrect line numbers in warning messages when using 11373 inline functions 11374 * [66]9177 -fdump-translation-unit: C front end deletes function_decl 11375 AST nodes and breaks debugging dumps 11376 * [67]9853 miscompilation of non-constant structure initializer 11377 11378 c++ compiler and library 11379 11380 * [68]45 legal template specialization code is rejected (DUP: 11381 [69]3784) 11382 * [70]764 lookup failure: friend operator and dereferencing a pointer 11383 and templates (DUP: [71]5116) 11384 * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP: 11385 2863) 11386 * [73]3663 G++ doesn't check access control during template 11387 instantiation 11388 * [74]3797 gcc fails to emit explicit specialization of a template 11389 member 11390 * [75]3948 Two destructors are called when no copy destructor is 11391 defined (ABI change) 11392 * [76]4137 Conversion operator within template is not accepted 11393 * [77]4361 bogus ambiguity taking the address of a member template 11394 * [78]4802 g++ accepts illegal template code (access to private 11395 member; DUP: [79]5837) 11396 * [80]4803 inline function is used but never defined, and g++ does 11397 not object 11398 * [81]5094 Partial specialization cannot be friend? 11399 * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66 11400 * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run 11401 time 11402 * [84]7015 certain __asm__ constructs rejected 11403 * [85]7086 compile time regression (quadratic behavior in 11404 fixup_var_refs) 11405 * [86]7099 G++ doesn't set the noreturn attribute on std::exit and 11406 std::abort 11407 * [87]7247 copy constructor missing when inlining enabled (invalid 11408 optimization?) 11409 * [88]7441 string array initialization compilation time regression 11410 from seconds to minutes 11411 * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong 11412 * [90]7804 bad printing of floating point constant in warning message 11413 * [91]8099 Friend classes and template specializations 11414 * [92]8117 member function pointers and multiple inheritance 11415 * [93]8205 using declaration and multiple inheritance 11416 * [94]8645 unnecessary non-zero checks in stl_tree.h 11417 * [95]8724 explicit destructor call for incomplete class allowed 11418 * [96]8805 compile time regression with many member variables 11419 * [97]8691 -O3 and -fno-implicit-templates are incompatible 11420 * [98]8700 unhelpful error message for binding temp to reference 11421 * [99]8724 explicit destructor call for incomplete class allowed 11422 * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems 11423 * [101]9016 Failure to consistently constant fold "constant" C++ 11424 objects 11425 * [102]9053 g++ confused about ambiguity of overloaded function 11426 templates 11427 * [103]9152 undefined virtual thunks 11428 * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out 11429 * [105]9297 data corruption due to codegen bug (when copying.) 11430 * [106]9318 i/ostream::operator>>/<<(streambuf*) broken 11431 * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf 11432 * [108]9400 bogus -Wshadow warning: shadowed declaration of this in 11433 local classes 11434 * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters 11435 * [110]9425 filebuf::pbackfail broken (DUP: [111]9439) 11436 * [112]9474 GCC freezes in compiling a weird code mixing <iostream> 11437 and <iostream.h> 11438 * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1) 11439 [114][DR 231] 11440 * [115]9555 ostream inserters fail to set badbit on exception 11441 * [116]9561 ostream inserters rethrow exception of wrong type 11442 * [117]9563 ostream::sentry returns true after a failed preparation 11443 * [118]9582 one-definition rule violation in std::allocator 11444 * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors 11445 * [120]9683 bug in initialization chains for static const variables 11446 from template classes 11447 * [121]9791 -Woverloaded-virtual reports hiding of destructor 11448 * [122]9817 collate::compare doesn't handle nul characters 11449 * [123]9825 filebuf::sputbackc breaks sbumpc 11450 * [124]9826 operator>>(basic_istream, basic_string) fails to compile 11451 with custom traits 11452 * [125]9924 Multiple using statements for builtin functions not 11453 allowed 11454 * [126]9946 destructor is not called for temporary object 11455 * [127]9964 filebuf::close() sometimes fails to close file 11456 * [128]9988 filebuf::overflow writes EOF to file 11457 * [129]10033 optimization breaks polymorphic references w/ typeid 11458 operator 11459 * [130]10097 filebuf::underflow drops characters 11460 * [131]10132 filebuf destructor can throw exceptions 11461 * [132]10180 gcc fails to warn about non-inlined function 11462 * [133]10199 method parametrized by template does not work everywhere 11463 * [134]10300 use of array-new (nothrow) in segfaults on NULL return 11464 * [135]10427 Stack corruption with variable-length automatic arrays 11465 and virtual destructors 11466 * [136]10503 Compilation never stops in fixed_type_or_null 11467 11468 Objective-C 11469 11470 * [137]5956 selectors aren't matched properly when added to the 11471 selector table 11472 11473 Fortran compiler and library 11474 11475 * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't 11476 detect 11477 * [139]3924 g77 generates code that is rejected by GAS if COFF debug 11478 info requested 11479 * [140]5634 doc: explain that configure --prefix=~/... does not work 11480 * [141]6367 multiple repeat counts confuse namelist read into array 11481 * [142]6491 Logical operations error on logicals when using 11482 -fugly-logint 11483 * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C" 11484 * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os 11485 on irix6.5 11486 * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should 11487 assume a direct access file 11488 * [146]7278 g77 "bug"; the executable misbehaves (with -O2 11489 -fno-automatic) 11490 * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows 11491 * [148]7388 Incorrect output with 0-based array of characters 11492 * [149]8587 Double complex zero ** double precision number -> NaN 11493 instead of zero 11494 * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning: 11495 unknown register name line-length-none 11496 * [151]10197 Direct access files not unformatted by default 11497 11498 Java compiler and library 11499 11500 * [152]6005 gcj fails to build rhug on alpha 11501 * [153]6389 System.getProperty("") should always throw an 11502 IllegalArgumentException 11503 * [154]6576 java.util.ResourceBundle.getResource ignores locale 11504 * [155]6652 new java.io.File("").getCanonicalFile() throws exception 11505 * [156]7060 getMethod() doesn't search super interface 11506 * [157]7073 bytecode interpreter gives wrong answer for interface 11507 getSuperclass() 11508 * [158]7180 possible bug in 11509 javax.naming.spi.NamingManager.getPlusPath() 11510 * [159]7416 java.security startup refs "GNU libgcj.security" 11511 * [160]7570 Runtime.exec with null envp: child doesn't inherit parent 11512 env (DUP: [161]7578) 11513 * [162]7611 Internal error while compiling libjava with -O 11514 * [163]7709 NullPointerException in _Jv_ResolvePoolEntry 11515 * [164]7766 ZipInputStream.available returns 0 immediately after 11516 construction 11517 * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public 11518 * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented 11519 * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens' 11520 * [168]8234 ZipInputStream chokes when InputStream.read() returns 11521 small chunks 11522 * [169]8415 reflection bug: exception info for Method 11523 * [170]8481 java.Random.nextInt(int) may return negative 11524 * [171]8593 Error reading GZIPped files with BufferedReader 11525 * [172]8759 java.beans.Introspector has no flushCaches() or 11526 flushFromCaches() methods 11527 * [173]8997 spin() calls Thread.sleep 11528 * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd 11529 instead of the root content of C: 11530 * [175]9254 java::lang::Object::wait(), threads-win32.cc returns 11531 wrong return codes 11532 * [176]9271 Severe bias in java.security.SecureRandom 11533 11534 Ada compiler and library 11535 11536 * [177]6767 make gnatlib-shared fails on -laddr2line 11537 * [178]9911 gnatmake fails to link when GCC configured with 11538 --with-sjlj-exceptions=yes 11539 * [179]10020 Can't bootstrap gcc on AIX with Ada enabled 11540 * [180]10546 Ada tasking not working on Red Hat 9 11541 11542 preprocessor 11543 11544 * [181]7029 preprocessor should ignore #warning with -M 11545 11546 ARM-specific 11547 11548 * [182]2903 [arm] Optimization bug with long long arithmetic 11549 * [183]7873 arm-linux-gcc fails when assigning address to a bit field 11550 11551 FreeBSD-specific 11552 11553 * [184]7680 float functions undefined in math.h/cmath with #define 11554 _XOPEN_SOURCE 11555 11556 HP-UX or HP-PA-specific 11557 11558 * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c 11559 * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to 11560 fputc_unlocked 11561 * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen 11562 11563 m68hc11-specific 11564 11565 * [188]6744 Bad assembler code generated: reference to pseudo 11566 register z 11567 * [189]7361 Internal compiler error in reload_cse_simplify_operands, 11568 in reload1.c 11569 11570 MIPS-specific 11571 11572 * [190]9496 [mips-linux] bug in optimizer? 11573 11574 PowerPC-specific 11575 11576 * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of 11577 space 11578 * [192]8480 reload ICEs for LAPACK code on powerpc64-linux 11579 * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg 11580 * [194]10315 [powerpc] ICE: in extract_insn, in recog.c 11581 11582 SPARC-specific 11583 11584 * [195]10267 (documentation) Wrong build instructions for 11585 *-*-solaris2* 11586 11587 x86-specific (Intel/AMD) 11588 11589 * [196]7916 ICE in instantiate_virtual_register_1 11590 * [197]7926 (c++) i486 instructions in header files make c++ programs 11591 crash on i386 11592 * [198]8555 ICE in gen_split_1231 11593 * [199]8994 ICE with -O -march=pentium4 11594 * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs 11595 * [201]9806 ICE in inline assembly with -fPIC flag 11596 * [202]10077 gcc -msse2 generates movd to move dwords between xmm 11597 regs 11598 * [203]10233 64-bit comparison only comparing bottom 32-bits 11599 * [204]10286 type-punning doesn't work with __m64 and -O 11600 * [205]10308 [x86] ICE with -O -fgcse or -O2 11601 __________________________________________________________________ 11602 11603 GCC 3.3.1 11604 11605 Bug Fixes 11606 11607 This section lists the problem reports (PRs) from GCC's bug tracking 11608 system that are known to be fixed in the 3.3.1 release. This list might 11609 not be complete (that is, it is possible that some PRs that have been 11610 fixed are not listed here). 11611 11612 Bootstrap failures 11613 11614 * [206]11272 [Solaris] make bootstrap fails while building libstdc++ 11615 11616 Internal compiler errors (multi-platform) 11617 11618 * [207]5754 ICE on invalid nested template class 11619 * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64 11620 and --enable-checking 11621 * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c 11622 * [210]7053 (c++) ICE when declaring a function already defined as a 11623 friend method of a template class 11624 * [211]8164 (c++) ICE when using different const expressions as 11625 template parameter 11626 * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c 11627 * [213]9559 (c++) ICE with invalid initialization of a static const 11628 * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c 11629 when redeclaring a static member variable 11630 * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in 11631 dwarfout.c with -g -O -finline-functions 11632 * [216]10432 (c++) ICE in poplevel, in cp/decl.c 11633 * [217]10475 ICE in subreg_highpart_offset for code with long long 11634 * [218]10635 (c++) ICE when dereferencing an incomplete type casted 11635 from a void pointer 11636 * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while 11637 instantiating static member variables 11638 * [220]10700 ICE in copy_to_mode_reg on 64-bit targets 11639 * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c 11640 * [222]10796 (c++) ICE when defining an enum with two values: -1 and 11641 MAX_INT_64BIT 11642 * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x 11643 sched.c 11644 * [224]10939 (c++) ICE with template code 11645 * [225]10956 (c++) ICE when specializing a template member function 11646 of a template class, in tsubst, in cp/pt.c 11647 * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*() 11648 defined) 11649 * [227]11059 (c++) ICE with empty union 11650 * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with 11651 -O2 -fnon-call-exceptions 11652 * [229]11105 (c++) ICE in mangle_conv_op_name_for_type 11653 * [230]11149 (c++) ICE on error when instantiation with call function 11654 of a base type 11655 * [231]11228 (c++) ICE on new-expression using array operator new and 11656 default-initialization 11657 * [232]11282 (c++) Infinite memory usage after syntax error 11658 * [233]11301 (fortran) ICE with -fno-globals 11659 * [234]11308 (c++) ICE when using an enum type name as if it were a 11660 class or namespace 11661 * [235]11473 (c++) ICE with -gstabs when empty struct inherits from 11662 an empty struct 11663 * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR 11664 * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c: 11665 template member functions 11666 11667 Optimization bugs 11668 11669 * [238]11198 -O2 -frename-registers generates wrong code (aliasing 11670 problem) 11671 * [239]11304 Wrong code production with -fomit-frame-pointer 11672 * [240]11381 volatile memory access optimized away 11673 * [241]11536 [strength-reduce] -O2 optimization produces wrong code 11674 * [242]11557 constant folding bug generates wrong code 11675 11676 C front end 11677 11678 * [243]5897 No warning for statement after return 11679 * [244]11279 DWARF-2 output mishandles large enums 11680 11681 Preprocessor bugs 11682 11683 * [245]11022 no warning for non-compatible macro redefinition 11684 11685 C++ compiler and library 11686 11687 * [246]2330 static_cast<>() to a private base is allowed 11688 * [247]5388 Incorrect message "operands to ?: have different types" 11689 * [248]5390 Libiberty fails to demangle multi-digit template 11690 parameters 11691 * [249]7877 Incorrect parameter passing to specializations of member 11692 function templates 11693 * [250]9393 Anonymous namespaces and compiling the same file twice 11694 * [251]10032 -pedantic converts some errors to warnings 11695 * [252]10468 const typeof(x) is non-const, but only in templates 11696 * [253]10527 confused error message with "new int()" parameter 11697 initializer 11698 * [254]10679 parameter MIN_INLINE_INSNS is not honored 11699 * [255]10682 gcc chokes on a typedef for an enum inside a class 11700 template 11701 * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of 11702 0. 11703 * [257]10845 template member function (with nested template as 11704 parameter) cannot be called anymore if another unrelated template 11705 member function is defined 11706 * [258]10849 Cannot define an out-of-class specialization of a 11707 private nested template class 11708 * [259]10888 Suppress -Winline warnings for system headers 11709 * [260]10929 -Winline warns about functions for which no definition 11710 is visible 11711 * [261]10931 valid conversion static_cast<const unsigned 11712 int&>(lvalue-of-type-int) is rejected 11713 * [262]10940 Bad code with explicit specialization 11714 * [263]10968 If member function implicitly instantiated, explicit 11715 instantiation of class fails to instantiate it 11716 * [264]10990 Cannot convert with dynamic_cast<> to a private base 11717 class from within a member function 11718 * [265]11039 Bad interaction between implicit typename deprecation 11719 and friendship 11720 * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say 11721 "__unused__" instead 11722 * [267]11095 C++ iostream manipulator causes segfault when called 11723 with negative argument 11724 * [268]11098 g++ doesn't emit complete debugging information for 11725 local variables in destructors 11726 * [269]11137 GNU/Linux shared library constructors not called unless 11727 there's one global object 11728 * [270]11154 spurious ambiguity report for template class 11729 specialization 11730 * [271]11329 Compiler cannot find user defined implicit typecast 11731 * [272]11332 Spurious error with casts in ?: expression 11732 * [273]11431 static_cast behavior with subclasses when default 11733 constructor available 11734 * [274]11528 money_get facet does not accept "$.00" as valid 11735 * [275]11546 Type lookup problems in out-of-line definition of a 11736 class doubly nested from a template class 11737 * [276]11567 C++ code containing templated member function with same 11738 name as pure virtual member function results in linking failure 11739 * [277]11645 Failure to deal with using and private inheritance 11740 11741 Java compiler and library 11742 11743 * [278]5179 Qualified static field access doesn't initialize its 11744 class 11745 * [279]8204 gcj -O2 to native reorders certain instructions 11746 improperly 11747 * [280]10838 java.io.ObjectInputStream syntax error 11748 * [281]10886 The RMI registry that comes with GCJ does not work 11749 correctly 11750 * [282]11349 JNDI URL context factories not located correctly 11751 11752 x86-specific (Intel/AMD) 11753 11754 * [283]4823 ICE on inline assembly code 11755 * [284]8878 miscompilation with -O and SSE 11756 * [285]9815 (c++ library) atomicity.h - fails to compile with -O3 11757 -masm=intel 11758 * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads, 11759 in reload1.c 11760 * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2 11761 * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source 11762 * [289]11044 [x86] out of range loop instructions for FP code on K6 11763 * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE 11764 built-ins 11765 * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC" 11766 is used 11767 11768 SPARC- or Solaris- specific 11769 11770 * [292]9362 solaris 'as' dies when fed .s and "-gstabs" 11771 * [293]10142 [SPARC64] gcc produces wrong code when passing 11772 structures by value 11773 * [294]10663 New configure check aborts with Sun tools. 11774 * [295]10835 combinatorial explosion in scheduler on HyperSPARC 11775 * [296]10876 ICE in calculate_giv_inc when building KDE 11776 * [297]10955 wrong code at -O3 for structure argument in context of 11777 structure return 11778 * [298]11018 -mcpu=ultrasparc busts tar-1.13.25 11779 * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x 11780 Linux kernel 11781 11782 ia64 specific 11783 11784 * [300]10907 gcc violates the ia64 ABI (GP must be preserved) 11785 * [301]11320 scheduler bug (in machine depended reorganization pass) 11786 * [302]11599 bug with conditional and __builtin_prefetch 11787 11788 PowerPC specific 11789 11790 * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem 11791 during loop) 11792 * [304]10871 error in rs6000_stack_info save_size computation 11793 * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse 11794 cures it 11795 11796 m68k-specific 11797 11798 * [306]7594 [m68k] ICE on legal code associated with simplify-rtx 11799 * [307]10557 [m68k] ICE in subreg_offset_representable_p 11800 * [308]11054 [m68k] ICE in reg_overlap_mentioned_p 11801 11802 ARM-specific 11803 11804 * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for 11805 functions with __attribute__ ((interrupt ("IRQ"))) 11806 * [310]10842 [arm] Clobbered link register is copied to pc under 11807 certain circumstances 11808 * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes 11809 * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno 11810 (3.4) 11811 11812 MIPS-specific 11813 11814 * [313]11084 ICE in propagate_one_insn, in flow.c 11815 11816 SH-specific 11817 11818 * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf 11819 * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c 11820 * [316]11096 i686-linux to sh-linux cross compiler fails to compile 11821 C++ files 11822 11823 GNU/Linux (or Hurd?) specific 11824 11825 * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3 11826 11827 UnixWare specific 11828 11829 * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare 11830 7.1.1 11831 11832 Cygwin (or mingw) specific 11833 11834 * [319]5287 ICE with dllimport attribute 11835 * [320]10148 [MingW/CygWin] Compiler dumps core 11836 11837 DJGPP specific 11838 11839 * [321]8787 GCC fails to emit .intel_syntax when invoked with 11840 -masm=intel on DJGPP 11841 11842 Darwin (and MacOS X) specific 11843 11844 * [322]10900 trampolines crash 11845 11846 Documentation 11847 11848 * [323]1607 (c++) Format attributes on methods undocumented 11849 * [324]4252 Invalid option `-fdump-translation-unit' 11850 * [325]4490 Clarify restrictions on -m96bit-long-double, 11851 -m128bit-long-double 11852 * [326]10355 document an issue with regparm attribute on some systems 11853 (e.g. Solaris) 11854 * [327]10726 (fortran) Documentation for function "IDate Intrinsic 11855 (Unix)" is wrong 11856 * [328]10805 document bug in old version of Sun assembler 11857 * [329]10815 warn against GNU binutils on AIX 11858 * [330]10877 document need for newer binutils on i?86-*-linux-gnu 11859 * [331]11280 Manual incorrect with respect to -freorder-blocks 11860 * [332]11466 Document -mlittle-endian and its restrictions for the 11861 sparc64 port 11862 11863 Testsuite bugs (compiler itself is not affected) 11864 11865 * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly 11866 report failure 11867 * [334]10810 gcc-3.3 fails make check: buffer overrun in 11868 test_demangle.c 11869 __________________________________________________________________ 11870 11871 GCC 3.3.2 11872 11873 Bug Fixes 11874 11875 This section lists the problem reports (PRs) from GCC's bug tracker 11876 that are known to be fixed in the 3.3.2 release. This list might not be 11877 complete (that is, it is possible that some PRs that have been fixed 11878 are not listed here). 11879 11880 Bootstrap failures and problems 11881 11882 * [335]8336 [SCO5] bootstrap config still tries to use COFF options 11883 * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with 11884 --enable-threads=posix 11885 * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap 11886 * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare 11887 7.1.1) 11888 * [339]11687 xstormy16-elf build fails in libf2c 11889 * [340]12263 [SGI IRIX] bootstrap fails during compile of 11890 libf2c/libI77/backspace.c 11891 * [341]12490 buffer overflow in scan-decls.c (during Solaris 9 11892 fix-header processing) 11893 11894 Internal compiler errors (multi-platform) 11895 11896 * [342]7277 Casting integers to vector types causes ICE 11897 * [343]7939 (c++) ICE on invalid function template specialization 11898 * [344]11063 (c++) ICE on parsing initialization list of const array 11899 member 11900 * [345]11207 ICE with negative index in array element designator 11901 * [346]11522 (fortran) g77 dwarf-2 ICE in 11902 add_abstract_origin_attribute 11903 * [347]11595 (c++) ICE on duplicate label definition 11904 * [348]11646 (c++) ICE in commit_one_edge_insertion with 11905 -fnon-call-exceptions -fgcse -O 11906 * [349]11665 ICE in struct initializer when taking address 11907 * [350]11852 (c++) ICE with bad struct initializer. 11908 * [351]11878 (c++) ICE in cp_expr_size 11909 * [352]11883 ICE with any -O on mercury-generated C code 11910 * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in 11911 cp/typeck2.c when applying typeid operator to template template 11912 parameter 11913 * [354]12146 ICE in lookup_template_function, in cp/pt.c 11914 * [355]12215 ICE in make_label_edge with -fnon-call-exceptions 11915 -fno-gcse -O2 11916 * [356]12369 (c++) ICE with templates and friends 11917 * [357]12446 ICE in emit_move_insn on complicated array reference 11918 * [358]12510 ICE in final_scan_insn 11919 * [359]12544 ICE with large parameters used in nested functions 11920 11921 C and optimization bugs 11922 11923 * [360]9862 spurious warnings with -W -finline-functions 11924 * [361]10962 lookup_field is a linear search on a linked list (can be 11925 slow if large struct) 11926 * [362]11370 -Wunreachable-code gives false complaints 11927 * [363]11637 invalid assembly with -fnon-call-exceptions 11928 * [364]11885 Problem with bitfields in packed structs 11929 * [365]12082 Inappropriate unreachable code warnings 11930 * [366]12180 Inline optimization fails for variadic function 11931 * [367]12340 loop unroller + gcse produces wrong code 11932 11933 C++ compiler and library 11934 11935 * [368]3907 nested template parameter collides with member name 11936 * [369]5293 confusing message when binding a temporary to a reference 11937 * [370]5296 [DR115] Pointers to functions and to template functions 11938 behave differently in deduction 11939 * [371]7939 ICE on function template specialization 11940 * [372]8656 Unable to assign function with __attribute__ and pointer 11941 return type to an appropriate variable 11942 * [373]10147 Confusing error message for invalid template function 11943 argument 11944 * [374]11400 std::search_n() makes assumptions about Size parameter 11945 * [375]11409 issues with using declarations, overloading, and 11946 built-in functions 11947 * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle 11948 multiple bits in mask 11949 * [377]11786 operator() call on variable in other namespace not 11950 recognized 11951 * [378]11867 static_cast ignores ambiguity 11952 * [379]11928 bug with conversion operators that are typedefs 11953 * [380]12114 Uninitialized memory accessed in dtor 11954 * [381]12163 static_cast + explicit constructor regression 11955 * [382]12181 Wrong code with comma operator and c++ 11956 * [383]12236 regparm and fastcall messes up parameters 11957 * [384]12266 incorrect instantiation of unneeded template during 11958 overload resolution 11959 * [385]12296 istream::peek() doesn't set eofbit 11960 * [386]12298 [sjlj exceptions] Stack unwind destroys 11961 not-yet-constructed object 11962 * [387]12369 ICE with templates and friends 11963 * [388]12337 apparently infinite loop in g++ 11964 * [389]12344 stdcall attribute ignored if function returns a pointer 11965 * [390]12451 missing(late) class forward declaration in cxxabi.h 11966 * [391]12486 g++ accepts invalid use of a qualified name 11967 11968 x86 specific (Intel/AMD) 11969 11970 * [392]8869 [x86 MMX] ICE with const variable optimization and MMX 11971 builtins 11972 * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions 11973 -O2 11974 * [394]11689 g++3.3 emits un-assembleable code for k6 architecture 11975 * [395]12116 [k6] Invalid assembly output values with X-MAME code 11976 * [396]12070 ICE converting between double and long double with 11977 -msoft-float 11978 11979 ia64-specific 11980 11981 * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc 11982 * [398]11535 __builtin_return_address may not work on ia64 11983 * [399]11693 [ia64] ICE in gen_nop_type 11984 * [400]12224 [ia64] Thread-local storage doesn't work 11985 11986 PowerPC-specific 11987 11988 * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux 11989 kernel 11990 * [402]11319 loop miscompiled on ppc32 11991 * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code 11992 11993 SPARC-specific 11994 11995 * [404]11662 wrong code for expr. with cast to long long and 11996 exclusive or 11997 * [405]11965 invalid assembler code for a shift < 32 operation 11998 * [406]12301 (c++) stack corruption when a returned expression throws 11999 an exception 12000 12001 Alpha-specific 12002 12003 * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of 12004 kernel 2.4.22-pre8 12005 12006 HPUX-specific 12007 12008 * [408]11313 problem with #pragma weak and static inline functions 12009 * [409]11712 __STDC_EXT__ not defined for C++ by default anymore? 12010 12011 Solaris specific 12012 12013 * [410]12166 Profiled programs crash if PROFDIR is set 12014 12015 Solaris-x86 specific 12016 12017 * [411]12101 i386 Solaris no longer works with GNU as? 12018 12019 Miscellaneous embedded target-specific bugs 12020 12021 * [412]10988 [m32r-elf] wrong blockmove code with -O3 12022 * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with 12023 -O2 12024 * [414]11902 [sh4] spec file improperly inserts rpath even when none 12025 needed 12026 * [415]11903 [sh4] -pthread fails to link due to error in spec file 12027 on sh4 12028 __________________________________________________________________ 12029 12030 GCC 3.3.3 12031 12032 Minor features 12033 12034 In addition to the bug fixes documented below, this release contains 12035 few minor features such as: 12036 * Support for --with-sysroot 12037 * Support for automatic detection of executable stacks 12038 * Support for SSE3 instructions 12039 * Support for thread local storage debugging under GDB on S390 12040 12041 Bug Fixes 12042 12043 This section lists the problem reports (PRs) from GCC's bug tracker 12044 that are known to be fixed in the 3.3.3 release. This list might not be 12045 complete (that is, it is possible that some PRs that have been fixed 12046 are not listed here). 12047 12048 Bootstrap failures and issues 12049 12050 * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails 12051 * [417]12399 boehm-gc fails (when building a cross compiler): libtool 12052 unable to infer tagged configuration 12053 * [418]13068 mklibgcc.in doesn't handle multi-level multilib 12054 subdirectories properly 12055 12056 Internal compiler errors (multi-platform) 12057 12058 * [419]10060 ICE (stack overflow) on huge file (300k lines) due to 12059 recursive behaviour of copy_rtx_if_shared, in emit_rtl.c 12060 * [420]10555 (c++) ICE on undefined template argument 12061 * [421]10706 (c++) ICE in mangle_class_name_for_template 12062 * [422]11496 (fortran) error in flow_loops_find when -funroll-loops 12063 active 12064 * [423]11741 ICE in pre_insert_copy_insn, in gcse.c 12065 * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0 12066 * [425]12632 (fortran) -fbounds-check ICE 12067 * [426]12712 (c++) ICE on short legit C++ code fragment with gcc 12068 3.3.2 12069 * [427]12726 (c++) ICE (segfault) on trivial code 12070 * [428]12890 (c++) ICE on compilation of class with throwing method 12071 * [429]12900 (c++) ICE in rtl_verify_flow_info_1 12072 * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on 12073 correct code with -O2 -fno-force-mem 12074 * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive 12075 template 12076 * [432]13318 ICE: floating point exception in the loop optimizer 12077 * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in 12078 except.c 12079 * [434]13574 (c++) invalid array default initializer in class lets 12080 gcc consume all memory and die 12081 * [435]13475 ICE on SIMD variables with partial value initialization 12082 * [436]13797 (c++) ICE on invalid template parameter 12083 * [437]13824 (java) gcj SEGV with simple .java program 12084 12085 C and optimization bugs 12086 12087 * [438]8776 loop invariants are not removed (most likely) 12088 * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing 12089 strncmp by memcmp 12090 * [440]11350 undefined labels with -Os -fPIC 12091 * [441]12826 Optimizer removes reference through volatile pointer 12092 * [442]12500 stabs debug info: void no longer a predefined / builtin 12093 type 12094 * [443]12941 builtin-bitops-1.c miscompilation (latent bug) 12095 * [444]12953 tree inliner bug (in inline_forbidden_p) and fix 12096 * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled 12097 * [446]13507 spurious printf format warning 12098 * [447]13382 Type information for const pointer disappears during 12099 optimization. 12100 * [448]13394 noreturn attribute ignored on recursive invokation 12101 * [449]13400 Compiled code crashes storing to read-only location 12102 * [450]13521 Endless loop in calculate_global_regs_live 12103 12104 C++ compiler and library 12105 12106 Some of the bug fixes in this list were made to implement decisions 12107 that the ISO C++ standards committee has made concerning several defect 12108 reports (DRs). Links in the list below point to detailed discussion of 12109 the relevant defect report. 12110 * [451]2094 unimplemented: use of `ptrmem_cst' in template type 12111 unification 12112 * [452]2294 using declaration confusion 12113 * [453]5050 template instantiation depth exceeds limit: recursion 12114 problem? 12115 * [454]9371 Bad exception handling in 12116 i/ostream::operator>>/<<(streambuf*) 12117 * [455]9546 bad exception handling in ostream members 12118 * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the 12119 face of unknown locales 12120 * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work 12121 * [459]10095 istream::operator>>(int&) sets ios::badbit when 12122 ios::failbit is set. 12123 * [460]11554 Warning about reordering of initializers doesn't mention 12124 location of constructor 12125 * [461]12297 istream::sentry::sentry() handles eof() incorrectly. 12126 * [462]12352 Exception safety problems in src/localename.cc 12127 * [463]12438 Memory leak in locale::combine() 12128 * [464]12540 Memory leak in locale::locale(const char*) 12129 * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented 12130 * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented 12131 * [470]12696 memory eating infinite loop in diagnostics (error 12132 recovery problem) 12133 * [471]12815 Code compiled with optimization behaves unexpectedly 12134 * [472]12862 Conflicts between typedefs/enums and namespace member 12135 declarations 12136 * [473]12926 Wrong value after assignment in initialize list using 12137 bit-fields 12138 * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented 12139 * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented 12140 * [478]13007 basic_streambuf::pubimbue, imbue wrong 12141 * [479]13009 Implicitly-defined assignment operator writes to wrong 12142 memory 12143 * [480]13057 regparm attribute not applied to destructor 12144 * [481]13070 -Wformat option ignored in g++ 12145 * [482]13081 forward template declarations in <complex> let inlining 12146 fail 12147 * [483]13239 Assertion does not seem to work correctly anymore 12148 * [484]13262 "xxx is private within this context" when initializing a 12149 self-contained template class 12150 * [485]13290 simple typo in concept checking for std::generate_n 12151 * [486]13323 Template code does not compile in presence of typedef 12152 * [487]13369 __verify_grouping (and __add_grouping?) not correct 12153 * [488]13371 infinite loop with packed struct and inlining 12154 * [489]13445 Template argument replacement "dereferences" a typedef 12155 * [490]13461 Fails to access protected-ctor from public constant 12156 * [491]13462 Non-standard-conforming type set::pointer 12157 * [492]13478 gcc uses wrong constructor to initialize a const 12158 reference 12159 * [493]13544 "conflicting types" for enums in different scopes 12160 * [494]13650 string::compare should not (always) use 12161 traits_type::length() 12162 * [495]13683 bogus warning about passing non-PODs through ellipsis 12163 * [496]13688 Derived class is denied access to protected base class 12164 member class 12165 * [497]13774 Member variable cleared in virtual multiple inheritance 12166 class 12167 * [498]13884 Protect sstream.tcc from extern template use 12168 12169 Java compiler and library 12170 12171 * [499]10746 [win32] garbage collection crash in GCJ 12172 12173 Objective-C compiler and library 12174 12175 * [500]11433 Crash due to dereferencing null pointer when querying 12176 protocol 12177 12178 Fortran compiler and library 12179 12180 * [501]12633 logical expression gives incorrect result with 12181 -fugly-logint option 12182 * [502]13037 [gcse-lm] g77 generates incorrect code 12183 * [503]13213 Hex constant problem when compiling with -fugly-logint 12184 and -ftypeless-boz 12185 12186 x86-specific (Intel/AMD) 12187 12188 * [504]4490 ICE with -m128bit-long-double 12189 * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have 12190 `reg' in make_field_assignment, in combine.c 12191 * [506]12441 ICE: can't find a register to spill 12192 * [507]12943 array static-init failure under -fpic, -fPIC 12193 * [508]13608 Incorrect code with -O3 -ffast-math 12194 12195 PowerPC-specific 12196 12197 * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of 12198 __attribute__((aligned(16))) 12199 * [510]11793 ICE in extract_insn, in recog.c (const_vector's) 12200 * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in 12201 altivec.md) 12202 * [512]12537 g++ generates writeable text sections 12203 12204 SPARC-specific 12205 12206 * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0 12207 -m64 12208 * [514]12865 mprotect call to make trampoline executable may fail 12209 * [515]13354 ICE in sparc_emit_set_const32 12210 12211 ARM-specific 12212 12213 * [516]10467 [arm] ICE in pre_insert_copy_insn, 12214 12215 ia64-specific 12216 12217 * [517]11226 ICE passing struct arg with two floats 12218 * [518]11227 ICE for _Complex float, _Complex long double args 12219 * [519]12644 GCC 3.3.2 fails to compile glibc on ia64 12220 * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn 12221 * Various fixes for libunwind 12222 12223 Alpha-specific 12224 12225 * [521]12654 Incorrect comparison code generated for Alpha 12226 * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2 12227 * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2 12228 12229 HPPA-specific 12230 12231 * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c 12232 * [525]12158 [hppa] compilation does not terminate at -O1 12233 12234 S390-specific 12235 12236 * [526]11992 Wrong built-in code for memcmp with length 1<<24: only 12237 (1<<24)-1 possible for CLCL-Instruction 12238 12239 SH-specific 12240 12241 * [527]9365 segfault in gen_far_branch (config/sh/sh.c) 12242 * [528]10392 optimizer generates faulty array indexing 12243 * [529]11322 SH profiler outputs multiple definitions of symbol 12244 * [530]13069 gcc/config/sh/rtems.h broken 12245 * [531]13302 Putting a va_list in a struct causes seg fault 12246 * [532]13585 Incorrect optimization of call to sfunc 12247 * Fix inappropriately exported libgcc functions from the shared 12248 library 12249 12250 Other embedded target specific 12251 12252 * [533]8916 [mcore] unsigned char assign gets hosed. 12253 * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c 12254 * [535]13122 [h8300] local variable gets corrupted by function call 12255 when -fomit-frame-pointer is given 12256 * [536]13256 [cris] strict_low_part mistreated in delay slots 12257 * [537]13373 [mcore] optimization with -frerun-cse-after-loop 12258 -fexpensive-optimizations produces wrong code on mcore 12259 12260 GNU HURD-specific 12261 12262 * [538]12561 gcc/config/t-gnu needs updating to work with 12263 --with-sysroot 12264 12265 Tru64 Unix specific 12266 12267 * [539]6243 testsuite fails almost all tests due to no libintl in 12268 LD_LIBRARY_PATH during test. 12269 * [540]11397 weak aliases broken on Tru64 UNIX 12270 12271 AIX-specific 12272 12273 * [541]12505 build failure due to defines of uchar in cpphash.h and 12274 sys/types.h 12275 * [542]13150 WEAK symbols not exported by collect2 12276 12277 IRIX-specific 12278 12279 * [543]12666 fixincludes problem on IRIX 6.5.19m 12280 12281 Solaris-specific 12282 12283 * [544]12969 Including sys/byteorder.h breaks configure checks 12284 12285 Testsuite problems (compiler is not affected) 12286 12287 * [545]10819 testsuite creates CR+LF on compiler version lines in 12288 test summary files 12289 * [546]11612 abi_check not finding correct libgcc_s.so.1 12290 12291 Miscellaneous 12292 12293 * [547]13211 using -###, incorrect warnings about unused linker file 12294 are produced 12295 __________________________________________________________________ 12296 12297 GCC 3.3.4 12298 12299 This is the [548]list of problem reports (PRs) from GCC's bug tracking 12300 system that are known to be fixed in the 3.3.4 release. This list might 12301 not be complete (that is, it is possible that some PRs that have been 12302 fixed are not listed here). 12303 __________________________________________________________________ 12304 12305 GCC 3.3.5 12306 12307 This is the [549]list of problem reports (PRs) from GCC's bug tracking 12308 system that are known to be fixed in the 3.3.5 release. This list might 12309 not be complete (that is, it is possible that some PRs that have been 12310 fixed are not listed here). 12311 __________________________________________________________________ 12312 12313 GCC 3.3.6 12314 12315 This is the [550]list of problem reports (PRs) from GCC's bug tracking 12316 system that are known to be fixed in the 3.3.6 release. This list might 12317 not be complete (that is, it is possible that some PRs that have been 12318 fixed are not listed here). 12319 12320 12321 For questions related to the use of GCC, please consult these web 12322 pages and the [551]GCC manuals. If that fails, the 12323 [552]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 12324 web pages and the development of GCC are welcome on our developer 12325 list at [553]gcc (a] gcc.gnu.org. All of [554]our lists have public 12326 archives. 12327 12328 Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and 12329 distribution of this entire article is permitted in any medium, 12330 provided this notice is preserved. 12331 12332 These pages are [556]maintained by the GCC team. Last modified 12333 2014-06-28[557]. 12334 12335 References 12336 12337 1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 12338 2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems 12339 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 12340 4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute 12341 5. http://gcc.gnu.org/news/dfa.html 12342 6. http://gcc.gnu.org/c99status.html 12343 7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html 12344 8. https://gcc.gnu.org/PR10140 12345 9. https://gcc.gnu.org/PR10198 12346 10. https://gcc.gnu.org/PR10338 12347 11. https://gcc.gnu.org/PR3581 12348 12. https://gcc.gnu.org/PR4382 12349 13. https://gcc.gnu.org/PR5533 12350 14. https://gcc.gnu.org/PR6387 12351 15. https://gcc.gnu.org/PR6412 12352 16. https://gcc.gnu.org/PR6620 12353 17. https://gcc.gnu.org/PR6663 12354 18. https://gcc.gnu.org/PR7068 12355 19. https://gcc.gnu.org/PR7083 12356 20. https://gcc.gnu.org/PR7647 12357 21. https://gcc.gnu.org/PR7675 12358 22. https://gcc.gnu.org/PR7718 12359 23. https://gcc.gnu.org/PR8116 12360 24. https://gcc.gnu.org/PR8358 12361 25. https://gcc.gnu.org/PR8511 12362 26. https://gcc.gnu.org/PR8564 12363 27. https://gcc.gnu.org/PR8660 12364 28. https://gcc.gnu.org/PR8766 12365 29. https://gcc.gnu.org/PR8803 12366 30. https://gcc.gnu.org/PR8846 12367 31. https://gcc.gnu.org/PR8906 12368 32. https://gcc.gnu.org/PR9216 12369 33. https://gcc.gnu.org/PR9261 12370 34. https://gcc.gnu.org/PR9263 12371 35. https://gcc.gnu.org/PR9429 12372 36. https://gcc.gnu.org/PR9516 12373 37. https://gcc.gnu.org/PR9600 12374 38. https://gcc.gnu.org/PR9629 12375 39. https://gcc.gnu.org/PR9672 12376 40. https://gcc.gnu.org/PR9749 12377 41. https://gcc.gnu.org/PR9794 12378 42. https://gcc.gnu.org/PR9829 12379 43. https://gcc.gnu.org/PR9916 12380 44. https://gcc.gnu.org/PR9936 12381 45. https://gcc.gnu.org/PR10262 12382 46. https://gcc.gnu.org/PR10278 12383 47. https://gcc.gnu.org/PR10446 12384 48. https://gcc.gnu.org/PR10451 12385 49. https://gcc.gnu.org/PR10506 12386 50. https://gcc.gnu.org/PR10549 12387 51. https://gcc.gnu.org/PR2001 12388 52. https://gcc.gnu.org/PR2391 12389 53. https://gcc.gnu.org/PR2960 12390 54. https://gcc.gnu.org/PR4046 12391 55. https://gcc.gnu.org/PR6405 12392 56. https://gcc.gnu.org/PR6798 12393 57. https://gcc.gnu.org/PR6871 12394 58. https://gcc.gnu.org/PR6909 12395 59. https://gcc.gnu.org/PR7189 12396 60. https://gcc.gnu.org/PR7642 12397 61. https://gcc.gnu.org/PR8634 12398 62. https://gcc.gnu.org/PR8750 12399 63. https://gcc.gnu.org/PR2161 12400 64. https://gcc.gnu.org/PR4319 12401 65. https://gcc.gnu.org/PR8602 12402 66. https://gcc.gnu.org/PR9177 12403 67. https://gcc.gnu.org/PR9853 12404 68. https://gcc.gnu.org/PR45 12405 69. https://gcc.gnu.org/PR3784 12406 70. https://gcc.gnu.org/PR764 12407 71. https://gcc.gnu.org/PR5116 12408 72. https://gcc.gnu.org/PR2862 12409 73. https://gcc.gnu.org/PR3663 12410 74. https://gcc.gnu.org/PR3797 12411 75. https://gcc.gnu.org/PR3948 12412 76. https://gcc.gnu.org/PR4137 12413 77. https://gcc.gnu.org/PR4361 12414 78. https://gcc.gnu.org/PR4802 12415 79. https://gcc.gnu.org/PR5837 12416 80. https://gcc.gnu.org/PR4803 12417 81. https://gcc.gnu.org/PR5094 12418 82. https://gcc.gnu.org/PR5730 12419 83. https://gcc.gnu.org/PR6713 12420 84. https://gcc.gnu.org/PR7015 12421 85. https://gcc.gnu.org/PR7086 12422 86. https://gcc.gnu.org/PR7099 12423 87. https://gcc.gnu.org/PR7247 12424 88. https://gcc.gnu.org/PR7441 12425 89. https://gcc.gnu.org/PR7768 12426 90. https://gcc.gnu.org/PR7804 12427 91. https://gcc.gnu.org/PR8099 12428 92. https://gcc.gnu.org/PR8117 12429 93. https://gcc.gnu.org/PR8205 12430 94. https://gcc.gnu.org/PR8645 12431 95. https://gcc.gnu.org/PR8724 12432 96. https://gcc.gnu.org/PR8805 12433 97. https://gcc.gnu.org/PR8691 12434 98. https://gcc.gnu.org/PR8700 12435 99. https://gcc.gnu.org/PR8724 12436 100. https://gcc.gnu.org/PR8949 12437 101. https://gcc.gnu.org/PR9016 12438 102. https://gcc.gnu.org/PR9053 12439 103. https://gcc.gnu.org/PR9152 12440 104. https://gcc.gnu.org/PR9182 12441 105. https://gcc.gnu.org/PR9297 12442 106. https://gcc.gnu.org/PR9318 12443 107. https://gcc.gnu.org/PR9320 12444 108. https://gcc.gnu.org/PR9400 12445 109. https://gcc.gnu.org/PR9424 12446 110. https://gcc.gnu.org/PR9425 12447 111. https://gcc.gnu.org/PR9439 12448 112. https://gcc.gnu.org/PR9474 12449 113. https://gcc.gnu.org/PR9548 12450 114. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231 12451 115. https://gcc.gnu.org/PR9555 12452 116. https://gcc.gnu.org/PR9561 12453 117. https://gcc.gnu.org/PR9563 12454 118. https://gcc.gnu.org/PR9582 12455 119. https://gcc.gnu.org/PR9622 12456 120. https://gcc.gnu.org/PR9683 12457 121. https://gcc.gnu.org/PR9791 12458 122. https://gcc.gnu.org/PR9817 12459 123. https://gcc.gnu.org/PR9825 12460 124. https://gcc.gnu.org/PR9826 12461 125. https://gcc.gnu.org/PR9924 12462 126. https://gcc.gnu.org/PR9946 12463 127. https://gcc.gnu.org/PR9964 12464 128. https://gcc.gnu.org/PR9988 12465 129. https://gcc.gnu.org/PR10033 12466 130. https://gcc.gnu.org/PR10097 12467 131. https://gcc.gnu.org/PR10132 12468 132. https://gcc.gnu.org/PR10180 12469 133. https://gcc.gnu.org/PR10199 12470 134. https://gcc.gnu.org/PR10300 12471 135. https://gcc.gnu.org/PR10427 12472 136. https://gcc.gnu.org/PR10503 12473 137. https://gcc.gnu.org/PR5956 12474 138. https://gcc.gnu.org/PR1832 12475 139. https://gcc.gnu.org/PR3924 12476 140. https://gcc.gnu.org/PR5634 12477 141. https://gcc.gnu.org/PR6367 12478 142. https://gcc.gnu.org/PR6491 12479 143. https://gcc.gnu.org/PR6742 12480 144. https://gcc.gnu.org/PR7113 12481 145. https://gcc.gnu.org/PR7236 12482 146. https://gcc.gnu.org/PR7278 12483 147. https://gcc.gnu.org/PR7384 12484 148. https://gcc.gnu.org/PR7388 12485 149. https://gcc.gnu.org/PR8587 12486 150. https://gcc.gnu.org/PR9038 12487 151. https://gcc.gnu.org/PR10197 12488 152. https://gcc.gnu.org/PR6005 12489 153. https://gcc.gnu.org/PR6389 12490 154. https://gcc.gnu.org/PR6576 12491 155. https://gcc.gnu.org/PR6652 12492 156. https://gcc.gnu.org/PR7060 12493 157. https://gcc.gnu.org/PR7073 12494 158. https://gcc.gnu.org/PR7180 12495 159. https://gcc.gnu.org/PR7416 12496 160. https://gcc.gnu.org/PR7570 12497 161. https://gcc.gnu.org/PR7578 12498 162. https://gcc.gnu.org/PR7611 12499 163. https://gcc.gnu.org/PR7709 12500 164. https://gcc.gnu.org/PR7766 12501 165. https://gcc.gnu.org/PR7785 12502 166. https://gcc.gnu.org/PR7786 12503 167. https://gcc.gnu.org/PR8142 12504 168. https://gcc.gnu.org/PR8234 12505 169. https://gcc.gnu.org/PR8415 12506 170. https://gcc.gnu.org/PR8481 12507 171. https://gcc.gnu.org/PR8593 12508 172. https://gcc.gnu.org/PR8759 12509 173. https://gcc.gnu.org/PR8997 12510 174. https://gcc.gnu.org/PR9253 12511 175. https://gcc.gnu.org/PR9254 12512 176. https://gcc.gnu.org/PR9271 12513 177. https://gcc.gnu.org/PR6767 12514 178. https://gcc.gnu.org/PR9911 12515 179. https://gcc.gnu.org/PR10020 12516 180. https://gcc.gnu.org/PR10546 12517 181. https://gcc.gnu.org/PR7029 12518 182. https://gcc.gnu.org/PR2903 12519 183. https://gcc.gnu.org/PR7873 12520 184. https://gcc.gnu.org/PR7680 12521 185. https://gcc.gnu.org/PR8705 12522 186. https://gcc.gnu.org/PR9986 12523 187. https://gcc.gnu.org/PR10056 12524 188. https://gcc.gnu.org/PR6744 12525 189. https://gcc.gnu.org/PR7361 12526 190. https://gcc.gnu.org/PR9496 12527 191. https://gcc.gnu.org/PR7067 12528 192. https://gcc.gnu.org/PR8480 12529 193. https://gcc.gnu.org/PR8784 12530 194. https://gcc.gnu.org/PR10315 12531 195. https://gcc.gnu.org/PR10267 12532 196. https://gcc.gnu.org/PR7916 12533 197. https://gcc.gnu.org/PR7926 12534 198. https://gcc.gnu.org/PR8555 12535 199. https://gcc.gnu.org/PR8994 12536 200. https://gcc.gnu.org/PR9426 12537 201. https://gcc.gnu.org/PR9806 12538 202. https://gcc.gnu.org/PR10077 12539 203. https://gcc.gnu.org/PR10233 12540 204. https://gcc.gnu.org/PR10286 12541 205. https://gcc.gnu.org/PR10308 12542 206. https://gcc.gnu.org/PR11272 12543 207. https://gcc.gnu.org/PR5754 12544 208. https://gcc.gnu.org/PR6597 12545 209. https://gcc.gnu.org/PR6949 12546 210. https://gcc.gnu.org/PR7053 12547 211. https://gcc.gnu.org/PR8164 12548 212. https://gcc.gnu.org/PR8384 12549 213. https://gcc.gnu.org/PR9559 12550 214. https://gcc.gnu.org/PR9649 12551 215. https://gcc.gnu.org/PR9864 12552 216. https://gcc.gnu.org/PR10432 12553 217. https://gcc.gnu.org/PR10475 12554 218. https://gcc.gnu.org/PR10635 12555 219. https://gcc.gnu.org/PR10661 12556 220. https://gcc.gnu.org/PR10700 12557 221. https://gcc.gnu.org/PR10712 12558 222. https://gcc.gnu.org/PR10796 12559 223. https://gcc.gnu.org/PR10890 12560 224. https://gcc.gnu.org/PR10939 12561 225. https://gcc.gnu.org/PR10956 12562 226. https://gcc.gnu.org/PR11041 12563 227. https://gcc.gnu.org/PR11059 12564 228. https://gcc.gnu.org/PR11083 12565 229. https://gcc.gnu.org/PR11105 12566 230. https://gcc.gnu.org/PR11149 12567 231. https://gcc.gnu.org/PR11228 12568 232. https://gcc.gnu.org/PR11282 12569 233. https://gcc.gnu.org/PR11301 12570 234. https://gcc.gnu.org/PR11308 12571 235. https://gcc.gnu.org/PR11473 12572 236. https://gcc.gnu.org/PR11503 12573 237. https://gcc.gnu.org/PR11513 12574 238. https://gcc.gnu.org/PR11198 12575 239. https://gcc.gnu.org/PR11304 12576 240. https://gcc.gnu.org/PR11381 12577 241. https://gcc.gnu.org/PR11536 12578 242. https://gcc.gnu.org/PR11557 12579 243. https://gcc.gnu.org/PR5897 12580 244. https://gcc.gnu.org/PR11279 12581 245. https://gcc.gnu.org/PR11022 12582 246. https://gcc.gnu.org/PR2330 12583 247. https://gcc.gnu.org/PR5388 12584 248. https://gcc.gnu.org/PR5390 12585 249. https://gcc.gnu.org/PR7877 12586 250. https://gcc.gnu.org/PR9393 12587 251. https://gcc.gnu.org/PR10032 12588 252. https://gcc.gnu.org/PR10468 12589 253. https://gcc.gnu.org/PR10527 12590 254. https://gcc.gnu.org/PR10679 12591 255. https://gcc.gnu.org/PR10682 12592 256. https://gcc.gnu.org/PR10689 12593 257. https://gcc.gnu.org/PR10845 12594 258. https://gcc.gnu.org/PR10849 12595 259. https://gcc.gnu.org/PR10888 12596 260. https://gcc.gnu.org/PR10929 12597 261. https://gcc.gnu.org/PR10931 12598 262. https://gcc.gnu.org/PR10940 12599 263. https://gcc.gnu.org/PR10968 12600 264. https://gcc.gnu.org/PR10990 12601 265. https://gcc.gnu.org/PR11039 12602 266. https://gcc.gnu.org/PR11062 12603 267. https://gcc.gnu.org/PR11095 12604 268. https://gcc.gnu.org/PR11098 12605 269. https://gcc.gnu.org/PR11137 12606 270. https://gcc.gnu.org/PR11154 12607 271. https://gcc.gnu.org/PR11329 12608 272. https://gcc.gnu.org/PR11332 12609 273. https://gcc.gnu.org/PR11431 12610 274. https://gcc.gnu.org/PR11528 12611 275. https://gcc.gnu.org/PR11546 12612 276. https://gcc.gnu.org/PR11567 12613 277. https://gcc.gnu.org/PR11645 12614 278. https://gcc.gnu.org/PR5179 12615 279. https://gcc.gnu.org/PR8204 12616 280. https://gcc.gnu.org/PR10838 12617 281. https://gcc.gnu.org/PR10886 12618 282. https://gcc.gnu.org/PR11349 12619 283. https://gcc.gnu.org/PR4823 12620 284. https://gcc.gnu.org/PR8878 12621 285. https://gcc.gnu.org/PR9815 12622 286. https://gcc.gnu.org/PR10402 12623 287. https://gcc.gnu.org/PR10504 12624 288. https://gcc.gnu.org/PR10673 12625 289. https://gcc.gnu.org/PR11044 12626 290. https://gcc.gnu.org/PR11089 12627 291. https://gcc.gnu.org/PR11420 12628 292. https://gcc.gnu.org/PR9362 12629 293. https://gcc.gnu.org/PR10142 12630 294. https://gcc.gnu.org/PR10663 12631 295. https://gcc.gnu.org/PR10835 12632 296. https://gcc.gnu.org/PR10876 12633 297. https://gcc.gnu.org/PR10955 12634 298. https://gcc.gnu.org/PR11018 12635 299. https://gcc.gnu.org/PR11556 12636 300. https://gcc.gnu.org/PR10907 12637 301. https://gcc.gnu.org/PR11320 12638 302. https://gcc.gnu.org/PR11599 12639 303. https://gcc.gnu.org/PR9745 12640 304. https://gcc.gnu.org/PR10871 12641 305. https://gcc.gnu.org/PR11440 12642 306. https://gcc.gnu.org/PR7594 12643 307. https://gcc.gnu.org/PR10557 12644 308. https://gcc.gnu.org/PR11054 12645 309. https://gcc.gnu.org/PR10834 12646 310. https://gcc.gnu.org/PR10842 12647 311. https://gcc.gnu.org/PR11052 12648 312. https://gcc.gnu.org/PR11183 12649 313. https://gcc.gnu.org/PR11084 12650 314. https://gcc.gnu.org/PR10331 12651 315. https://gcc.gnu.org/PR10413 12652 316. https://gcc.gnu.org/PR11096 12653 317. https://gcc.gnu.org/PR2873 12654 318. https://gcc.gnu.org/PR3163 12655 319. https://gcc.gnu.org/PR5287 12656 320. https://gcc.gnu.org/PR10148 12657 321. https://gcc.gnu.org/PR8787 12658 322. https://gcc.gnu.org/PR10900 12659 323. https://gcc.gnu.org/PR1607 12660 324. https://gcc.gnu.org/PR4252 12661 325. https://gcc.gnu.org/PR4490 12662 326. https://gcc.gnu.org/PR10355 12663 327. https://gcc.gnu.org/PR10726 12664 328. https://gcc.gnu.org/PR10805 12665 329. https://gcc.gnu.org/PR10815 12666 330. https://gcc.gnu.org/PR10877 12667 331. https://gcc.gnu.org/PR11280 12668 332. https://gcc.gnu.org/PR11466 12669 333. https://gcc.gnu.org/PR10737 12670 334. https://gcc.gnu.org/PR10810 12671 335. https://gcc.gnu.org/PR8336 12672 336. https://gcc.gnu.org/PR9330 12673 337. https://gcc.gnu.org/PR9631 12674 338. https://gcc.gnu.org/PR9877 12675 339. https://gcc.gnu.org/PR11687 12676 340. https://gcc.gnu.org/PR12263 12677 341. https://gcc.gnu.org/PR12490 12678 342. https://gcc.gnu.org/PR7277 12679 343. https://gcc.gnu.org/PR7939 12680 344. https://gcc.gnu.org/PR11063 12681 345. https://gcc.gnu.org/PR11207 12682 346. https://gcc.gnu.org/PR11522 12683 347. https://gcc.gnu.org/PR11595 12684 348. https://gcc.gnu.org/PR11646 12685 349. https://gcc.gnu.org/PR11665 12686 350. https://gcc.gnu.org/PR11852 12687 351. https://gcc.gnu.org/PR11878 12688 352. https://gcc.gnu.org/PR11883 12689 353. https://gcc.gnu.org/PR11991 12690 354. https://gcc.gnu.org/PR12146 12691 355. https://gcc.gnu.org/PR12215 12692 356. https://gcc.gnu.org/PR12369 12693 357. https://gcc.gnu.org/PR12446 12694 358. https://gcc.gnu.org/PR12510 12695 359. https://gcc.gnu.org/PR12544 12696 360. https://gcc.gnu.org/PR9862 12697 361. https://gcc.gnu.org/PR10962 12698 362. https://gcc.gnu.org/PR11370 12699 363. https://gcc.gnu.org/PR11637 12700 364. https://gcc.gnu.org/PR11885 12701 365. https://gcc.gnu.org/PR12082 12702 366. https://gcc.gnu.org/PR12180 12703 367. https://gcc.gnu.org/PR12340 12704 368. https://gcc.gnu.org/PR3907 12705 369. https://gcc.gnu.org/PR5293 12706 370. https://gcc.gnu.org/PR5296 12707 371. https://gcc.gnu.org/PR7939 12708 372. https://gcc.gnu.org/PR8656 12709 373. https://gcc.gnu.org/PR10147 12710 374. https://gcc.gnu.org/PR11400 12711 375. https://gcc.gnu.org/PR11409 12712 376. https://gcc.gnu.org/PR11740 12713 377. https://gcc.gnu.org/PR11786 12714 378. https://gcc.gnu.org/PR11867 12715 379. https://gcc.gnu.org/PR11928 12716 380. https://gcc.gnu.org/PR12114 12717 381. https://gcc.gnu.org/PR12163 12718 382. https://gcc.gnu.org/PR12181 12719 383. https://gcc.gnu.org/PR12236 12720 384. https://gcc.gnu.org/PR12266 12721 385. https://gcc.gnu.org/PR12296 12722 386. https://gcc.gnu.org/PR12298 12723 387. https://gcc.gnu.org/PR12369 12724 388. https://gcc.gnu.org/PR12337 12725 389. https://gcc.gnu.org/PR12344 12726 390. https://gcc.gnu.org/PR12451 12727 391. https://gcc.gnu.org/PR12486 12728 392. https://gcc.gnu.org/PR8869 12729 393. https://gcc.gnu.org/PR9786 12730 394. https://gcc.gnu.org/PR11689 12731 395. https://gcc.gnu.org/PR12116 12732 396. https://gcc.gnu.org/PR12070 12733 397. https://gcc.gnu.org/PR11184 12734 398. https://gcc.gnu.org/PR11535 12735 399. https://gcc.gnu.org/PR11693 12736 400. https://gcc.gnu.org/PR12224 12737 401. https://gcc.gnu.org/PR11087 12738 402. https://gcc.gnu.org/PR11319 12739 403. https://gcc.gnu.org/PR11949 12740 404. https://gcc.gnu.org/PR11662 12741 405. https://gcc.gnu.org/PR11965 12742 406. https://gcc.gnu.org/PR12301 12743 407. https://gcc.gnu.org/PR11717 12744 408. https://gcc.gnu.org/PR11313 12745 409. https://gcc.gnu.org/PR11712 12746 410. https://gcc.gnu.org/PR12166 12747 411. https://gcc.gnu.org/PR12101 12748 412. https://gcc.gnu.org/PR10988 12749 413. https://gcc.gnu.org/PR11805 12750 414. https://gcc.gnu.org/PR11902 12751 415. https://gcc.gnu.org/PR11903 12752 416. https://gcc.gnu.org/PR11890 12753 417. https://gcc.gnu.org/PR12399 12754 418. https://gcc.gnu.org/PR13068 12755 419. https://gcc.gnu.org/PR10060 12756 420. https://gcc.gnu.org/PR10555 12757 421. https://gcc.gnu.org/PR10706 12758 422. https://gcc.gnu.org/PR11496 12759 423. https://gcc.gnu.org/PR11741 12760 424. https://gcc.gnu.org/PR12440 12761 425. https://gcc.gnu.org/PR12632 12762 426. https://gcc.gnu.org/PR12712 12763 427. https://gcc.gnu.org/PR12726 12764 428. https://gcc.gnu.org/PR12890 12765 429. https://gcc.gnu.org/PR12900 12766 430. https://gcc.gnu.org/PR13060 12767 431. https://gcc.gnu.org/PR13289 12768 432. https://gcc.gnu.org/PR13318 12769 433. https://gcc.gnu.org/PR13392 12770 434. https://gcc.gnu.org/PR13574 12771 435. https://gcc.gnu.org/PR13475 12772 436. https://gcc.gnu.org/PR13797 12773 437. https://gcc.gnu.org/PR13824 12774 438. https://gcc.gnu.org/PR8776 12775 439. https://gcc.gnu.org/PR10339 12776 440. https://gcc.gnu.org/PR11350 12777 441. https://gcc.gnu.org/PR12826 12778 442. https://gcc.gnu.org/PR12500 12779 443. https://gcc.gnu.org/PR12941 12780 444. https://gcc.gnu.org/PR12953 12781 445. https://gcc.gnu.org/PR13041 12782 446. https://gcc.gnu.org/PR13507 12783 447. https://gcc.gnu.org/PR13382 12784 448. https://gcc.gnu.org/PR13394 12785 449. https://gcc.gnu.org/PR13400 12786 450. https://gcc.gnu.org/PR13521 12787 451. https://gcc.gnu.org/PR2094 12788 452. https://gcc.gnu.org/PR2294 12789 453. https://gcc.gnu.org/PR5050 12790 454. https://gcc.gnu.org/PR9371 12791 455. https://gcc.gnu.org/PR9546 12792 456. https://gcc.gnu.org/PR10081 12793 457. https://gcc.gnu.org/PR10093 12794 458. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61 12795 459. https://gcc.gnu.org/PR10095 12796 460. https://gcc.gnu.org/PR11554 12797 461. https://gcc.gnu.org/PR12297 12798 462. https://gcc.gnu.org/PR12352 12799 463. https://gcc.gnu.org/PR12438 12800 464. https://gcc.gnu.org/PR12540 12801 465. https://gcc.gnu.org/PR12594 12802 466. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60 12803 467. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63 12804 468. https://gcc.gnu.org/PR12657 12805 469. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292 12806 470. https://gcc.gnu.org/PR12696 12807 471. https://gcc.gnu.org/PR12815 12808 472. https://gcc.gnu.org/PR12862 12809 473. https://gcc.gnu.org/PR12926 12810 474. https://gcc.gnu.org/PR12967 12811 475. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html 12812 476. https://gcc.gnu.org/PR12971 12813 477. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328 12814 478. https://gcc.gnu.org/PR13007 12815 479. https://gcc.gnu.org/PR13009 12816 480. https://gcc.gnu.org/PR13057 12817 481. https://gcc.gnu.org/PR13070 12818 482. https://gcc.gnu.org/PR13081 12819 483. https://gcc.gnu.org/PR13239 12820 484. https://gcc.gnu.org/PR13262 12821 485. https://gcc.gnu.org/PR13290 12822 486. https://gcc.gnu.org/PR13323 12823 487. https://gcc.gnu.org/PR13369 12824 488. https://gcc.gnu.org/PR13371 12825 489. https://gcc.gnu.org/PR13445 12826 490. https://gcc.gnu.org/PR13461 12827 491. https://gcc.gnu.org/PR13462 12828 492. https://gcc.gnu.org/PR13478 12829 493. https://gcc.gnu.org/PR13544 12830 494. https://gcc.gnu.org/PR13650 12831 495. https://gcc.gnu.org/PR13683 12832 496. https://gcc.gnu.org/PR13688 12833 497. https://gcc.gnu.org/PR13774 12834 498. https://gcc.gnu.org/PR13884 12835 499. https://gcc.gnu.org/PR10746 12836 500. https://gcc.gnu.org/PR11433 12837 501. https://gcc.gnu.org/PR12633 12838 502. https://gcc.gnu.org/PR13037 12839 503. https://gcc.gnu.org/PR13213 12840 504. https://gcc.gnu.org/PR4490 12841 505. https://gcc.gnu.org/PR12292 12842 506. https://gcc.gnu.org/PR12441 12843 507. https://gcc.gnu.org/PR12943 12844 508. https://gcc.gnu.org/PR13608 12845 509. https://gcc.gnu.org/PR11598 12846 510. https://gcc.gnu.org/PR11793 12847 511. https://gcc.gnu.org/PR12467 12848 512. https://gcc.gnu.org/PR12537 12849 513. https://gcc.gnu.org/PR12496 12850 514. https://gcc.gnu.org/PR12865 12851 515. https://gcc.gnu.org/PR13354 12852 516. https://gcc.gnu.org/PR10467 12853 517. https://gcc.gnu.org/PR11226 12854 518. https://gcc.gnu.org/PR11227 12855 519. https://gcc.gnu.org/PR12644 12856 520. https://gcc.gnu.org/PR13149 12857 521. https://gcc.gnu.org/PR12654 12858 522. https://gcc.gnu.org/PR12965 12859 523. https://gcc.gnu.org/PR13031 12860 524. https://gcc.gnu.org/PR11634 12861 525. https://gcc.gnu.org/PR12158 12862 526. https://gcc.gnu.org/PR11992 12863 527. https://gcc.gnu.org/PR9365 12864 528. https://gcc.gnu.org/PR10392 12865 529. https://gcc.gnu.org/PR11322 12866 530. https://gcc.gnu.org/PR13069 12867 531. https://gcc.gnu.org/PR13302 12868 532. https://gcc.gnu.org/PR13585 12869 533. https://gcc.gnu.org/PR8916 12870 534. https://gcc.gnu.org/PR11576 12871 535. https://gcc.gnu.org/PR13122 12872 536. https://gcc.gnu.org/PR13256 12873 537. https://gcc.gnu.org/PR13373 12874 538. https://gcc.gnu.org/PR12561 12875 539. https://gcc.gnu.org/PR6243 12876 540. https://gcc.gnu.org/PR11397 12877 541. https://gcc.gnu.org/PR12505 12878 542. https://gcc.gnu.org/PR13150 12879 543. https://gcc.gnu.org/PR12666 12880 544. https://gcc.gnu.org/PR12969 12881 545. https://gcc.gnu.org/PR10819 12882 546. https://gcc.gnu.org/PR11612 12883 547. https://gcc.gnu.org/PR13211 12884 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4 12885 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5 12886 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6 12887 551. https://gcc.gnu.org/onlinedocs/ 12888 552. mailto:gcc-help (a] gcc.gnu.org 12889 553. mailto:gcc (a] gcc.gnu.org 12890 554. https://gcc.gnu.org/lists.html 12891 555. http://www.fsf.org/ 12892 556. https://gcc.gnu.org/about.html 12893 557. http://validator.w3.org/check/referer 12894 ====================================================================== 12895 http://gcc.gnu.org/gcc-3.2/index.html 12896 12897 GCC 3.2 Release Series 12898 12899 April 25, 2003 12900 12901 The [1]GNU project and the GCC developers are pleased to announce the 12902 release of GCC 3.2.3. 12903 12904 The purpose of the GCC 3.2 release series is to provide a stable 12905 platform for OS distributors to use building their next releases. A 12906 primary objective was to stabilize the C++ ABI; we believe that the 12907 interface to the compiler and the C++ standard library are now 12908 relatively stable. 12909 12910 Be aware that C++ code compiled by GCC 3.2.x will (in general) not 12911 interoperate with code compiled by GCC 3.1.1 or earlier. 12912 12913 Please refer to our [2]detailed list of news, caveats, and bug-fixes 12914 for further information. 12915 12916 Release History 12917 12918 GCC 3.2.3 12919 April 25, 2003 ([3]changes) 12920 12921 GCC 3.2.2 12922 February 5, 2003 ([4]changes) 12923 12924 GCC 3.2.1 12925 November 19, 2002 ([5]changes) 12926 12927 GCC 3.2 12928 August 14, 2002 ([6]changes) 12929 12930 References and Acknowledgements 12931 12932 GCC used to stand for the GNU C Compiler, but since the compiler 12933 supports several other languages aside from C, it now stands for the 12934 GNU Compiler Collection. 12935 12936 A list of [7]successful builds is updated as new information becomes 12937 available. 12938 12939 The GCC developers would like to thank the numerous people that have 12940 contributed new features, improvements, bug fixes, and other changes as 12941 well as test results to GCC. This [8]amazing group of volunteers is 12942 what makes GCC successful. 12943 12944 For additional information about GCC please refer to the [9]GCC project 12945 web site or contact the [10]GCC development mailing list. 12946 12947 To obtain GCC please use [11]our mirror sites, or our CVS server. 12948 12949 12950 For questions related to the use of GCC, please consult these web 12951 pages and the [12]GCC manuals. If that fails, the 12952 [13]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 12953 web pages and the development of GCC are welcome on our developer 12954 list at [14]gcc (a] gcc.gnu.org. All of [15]our lists have public 12955 archives. 12956 12957 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and 12958 distribution of this entire article is permitted in any medium, 12959 provided this notice is preserved. 12960 12961 These pages are [17]maintained by the GCC team. Last modified 12962 2014-06-28[18]. 12963 12964 References 12965 12966 1. http://www.gnu.org/ 12967 2. http://gcc.gnu.org/gcc-3.2/changes.html 12968 3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 12969 4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2 12970 5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1 12971 6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2 12972 7. http://gcc.gnu.org/gcc-3.2/buildstat.html 12973 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 12974 9. http://gcc.gnu.org/index.html 12975 10. mailto:gcc (a] gcc.gnu.org 12976 11. http://gcc.gnu.org/mirrors.html 12977 12. https://gcc.gnu.org/onlinedocs/ 12978 13. mailto:gcc-help (a] gcc.gnu.org 12979 14. mailto:gcc (a] gcc.gnu.org 12980 15. https://gcc.gnu.org/lists.html 12981 16. http://www.fsf.org/ 12982 17. https://gcc.gnu.org/about.html 12983 18. http://validator.w3.org/check/referer 12984 ====================================================================== 12985 http://gcc.gnu.org/gcc-3.2/changes.html 12986 12987 GCC 3.2 Release Series 12988 Changes, New Features, and Fixes 12989 12990 The latest release in the 3.2 release series is [1]GCC 3.2.3. 12991 12992 Caveats and New Features 12993 12994 Caveats 12995 12996 * The C++ compiler does not correctly zero-initialize 12997 pointers-to-data members. You must explicitly initialize them. For 12998 example: int S::*m(0); will work, but depending on 12999 default-initialization to zero will not work. This bug cannot be 13000 fixed in GCC 3.2 without inducing unacceptable risks. It will be 13001 fixed in GCC 3.3. 13002 * This GCC release is based on the GCC 3.1 sourcebase, and thus has 13003 all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has 13004 a number of C++ ABI fixes which make its C++ compiler generate 13005 binary code which is incompatible with the C++ compilers found in 13006 earlier GCC releases, including GCC 3.1 and GCC 3.1.1. 13007 13008 Frontend Enhancements 13009 13010 C/C++/Objective-C 13011 13012 * The method of constructing the list of directories to be searched 13013 for header files has been revised. If a directory named by a -I 13014 option is a standard system include directory, the option is 13015 ignored to ensure that the default search order for system 13016 directories and the special treatment of system header files are 13017 not defeated. 13018 * The C and Objective-C compilers no longer accept the "Naming Types" 13019 extension (typedef foo = bar); it was already unavailable in C++. 13020 Code which uses it will need to be changed to use the "typeof" 13021 extension instead: typedef typeof(bar) foo. (We have removed this 13022 extension without a period of deprecation because it has caused the 13023 compiler to crash since version 3.0 and no one noticed until very 13024 recently. Thus we conclude it is not in widespread use.) 13025 13026 C++ 13027 13028 * GCC 3.2 fixed serveral differences between the C++ ABI implemented 13029 in GCC and the multi-vendor standard, but more have been found 13030 since the release. 3.2.1 adds a new warning, -Wabi, to warn about 13031 code which is affected by these bugs. We will fix these bugs in 13032 some future release, once we are confident that all have been 13033 found; until then, it is our intention to make changes to the ABI 13034 only if they are necessary for correct compilation of C++, as 13035 opposed to conformance to the ABI documents. 13036 * For details on how to build an ABI compliant compiler for GNU/Linux 13037 systems, check the [3]common C++ ABI page. 13038 13039 New Targets and Target Specific Improvements 13040 13041 IA-32 13042 13043 * Fixed a number of bugs in SSE and MMX intrinsics. 13044 * Fixed common compiler crashes with SSE instruction set enabled 13045 (implied by -march=pentium3, pentium4, athlon-xp) 13046 * __m128 and __m128i is not 128bit aligned when used in structures. 13047 13048 x86-64 13049 13050 * A bug whereby the compiler could generate bad code for bzero has 13051 been fixed. 13052 * ABI fixes (implying ABI incompatibilities with previous version in 13053 some corner cases) 13054 * Fixed prefetch code generation 13055 __________________________________________________________________ 13056 13057 GCC 3.2.3 13058 13059 3.2.3 is a bug fix release only; there are no new features that were 13060 not present in GCC 3.2.2. 13061 13062 Bug Fixes 13063 13064 This section lists the problem reports (PRs) from GCC's bug tracking 13065 system that are known to be fixed in the 3.2.3 release. This list might 13066 not be complete (that is, it is possible that some PRs that have been 13067 fixed are not listed here), and some of the titles have been changed to 13068 make them more clear. 13069 13070 Internal Compiler Errors (multi-platform) 13071 13072 * [4]3782: (c++) -quiet -fstats produces a segmentation fault in 13073 cc1plus 13074 * [5]6440: (c++) template specializations cause ICE 13075 * [6]7050: (c++) ICE on: (i ? get_string() : throw) 13076 * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c) 13077 * [8]7982: (c++) ICE due to infinite recursion (using STL set) 13078 * [9]8068: exceedingly high (infinite) memory usage 13079 * [10]8178: ICE with __builtin_ffs 13080 * [11]8396: ICE in copy_to_mode_reg, in explow.c 13081 * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c 13082 * [13]9768: ICE when optimizing inline code at -O2 13083 * [14]9798: (c++) Infinite recursion (segfault) in 13084 cp/decl.c:push_using_directive with recursive using directives 13085 * [15]9799: mismatching structure initializer with nested flexible 13086 array member: ICE 13087 * [16]9928: ICE on duplicate enum declaration 13088 * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects 13089 sparc, alpha) 13090 * [18]10352: ICE in find_reloads_toplev 13091 * [19]10336: ICE with -Wunreachable-code 13092 13093 C/optimizer bugs: 13094 13095 * [20]8224: Incorrect joining of signed and unsigned division 13096 * [21]8613: -O2 produces wrong code with builtin strlen and 13097 postincrements 13098 * [22]8828: gcc reports some code is unreachable when it is not 13099 * [23]9226: GCSE breaking argument passing 13100 * [24]9853: miscompilation of non-constant structure initializer 13101 * [25]9797: C99-style struct initializers are miscompiled 13102 * [26]9967: Some standard C function calls should not be replaced 13103 when optimizing for size 13104 * [27]10116: ce2: invalid merge of join_bb in the context of switch 13105 statements 13106 * [28]10171: wrong code for inlined function 13107 * [29]10175: -Wunreachable-code doesn't work for single lines 13108 13109 C++ compiler and library: 13110 13111 * [30]8316: Confusing diagnostic for code that misuses conversion 13112 operators 13113 * [31]9169: filebuf output fails if codecvt<>::out returns noconv 13114 * [32]9420: incomplete type incorrectly reported 13115 * [33]9459: typeof in return type specification of template not 13116 supported 13117 * [34]9507: filebuf::open handles ios_base::ate incorrectly 13118 * [35]9538: Out-of-bounds memory access in streambuf::sputbackc 13119 * [36]9602: Total confusion about template/friend/virtual/abstract 13120 * [37]9993: destructor not called for local object created within and 13121 returned from infinite loop 13122 * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2 13123 system 13124 13125 Java compiler and library: 13126 13127 * [39]9652: libgcj build fails on irix6.5.1[78] 13128 * [40]10144: gas on solaris complains about bad .stabs lines for 13129 java, native as unaffected 13130 13131 x86-specific (Intel/AMD): 13132 13133 * [41]8746: gcc miscompiles Linux kernel ppa driver on x86 13134 * [42]9888: -mcpu=k6 -Os produces out of range loop instructions 13135 * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu 13136 failed 13137 * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib) 13138 failed 13139 13140 SPARC-specific: 13141 13142 * [45]7784: [Sparc] ICE in extract_insn, in recog.c 13143 * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in 13144 unroll.c 13145 * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc 13146 * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in 13147 execute/loop-2d.c 13148 * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc 13149 * [50]9414: Scheduling bug on Ultrasparc 13150 * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64 13151 13152 m68k-specific: 13153 13154 * [52]7248: broken "inclusive or" code 13155 * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1 13156 13157 PowerPC-specific: 13158 13159 * [54]9732: Wrong code with -O2 -fPIC 13160 * [55]10073: ICE: powerpc cannot split insn 13161 13162 Alpha-specific: 13163 13164 * [56]7702: optimization problem on a DEC alpha under OSF1 13165 * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system 13166 13167 HP-specific: 13168 13169 * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275) 13170 * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10 13171 (missing symbol) 13172 * [60]10271: Floating point args don't get reloaded across function 13173 calls with -O2 13174 13175 MIPS specific: 13176 13177 * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in 13178 compile/920501-4.c 13179 13180 CRIS specific: 13181 13182 * [62]10377: gcc-3.2.2 creates bad assembler code for cris 13183 13184 Miscellaneous and minor bugs: 13185 13186 * [63]6955: collect2 says "core dumped" when there is no core 13187 __________________________________________________________________ 13188 13189 GCC 3.2.2 13190 13191 Beginning with 3.2.2, GCC's Makefile suite supports redirection of make 13192 install by means of the DESTDIR variable. Parts of the GCC tree have 13193 featured that support long before, but now it is available even from 13194 the top level. 13195 13196 Other than that, GCC 3.2.2 is a bug fix release only; there are no new 13197 features that were not present in GCC 3.2.1. 13198 13199 Bug Fixes 13200 13201 On the following i386-based systems GCC 3.2.1 broke the C ABI wrt. 13202 functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped 13203 with FreeBSD 5.0 does not have this problem), Interix, a.out-based 13204 GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI 13205 change, and thus restores ABI-compatibility with previous releases 13206 (except GCC 3.2.1) on these platforms. 13207 13208 This section lists the problem reports (PRs) from GCC's bug tracking 13209 system that are known to be fixed in the 3.2.2 release. This list might 13210 not be complete (that is, it is possible that some PRs that have been 13211 fixed are not listed here) and some of the titles have been changed to 13212 make them more clear. 13213 13214 Internal Compiler Errors (multi-platform) 13215 13216 * [64]5919: (c++) ICE when passing variable array to template 13217 function 13218 * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=) 13219 * [66]7507: ICE with -O2 when address of called function is a 13220 complicated expression 13221 * [67]7622: ICE with nested inline functions if function's address is 13222 taken 13223 * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR 13224 [69]9258) 13225 * [70]8031: (c++) ICE in code comparing typeids and casting from 13226 virtual base 13227 * [71]8275: ICE in simplify_subreg 13228 * [72]8332: (c++) builtin strlen/template interaction causes ICE 13229 * [73]8372: (c++) ICE on explicit call of destructor 13230 * [74]8439: (c, not c++) empty struct causes ICE 13231 * [75]8442: (c++) ICE with nested template classes 13232 * [76]8518: ICE when compiling mplayer ("extern inline" issue) 13233 * [77]8615: (c++) ICE with out-of-range character constant template 13234 argument 13235 * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307 13236 * [79]8799: (c++) ICE: error reporting routines re-entered 13237 * [80]9328: (c++) ICE with typeof(X) for overloaded X 13238 * [81]9465: (preprocessor) cpp -traditional ICE on null bytes 13239 13240 C++ (compiler and library) bugs 13241 13242 * [82]47: scoping in nested classes is broken 13243 * [83]6745: problems with iostream rdbuf() member function 13244 * [84]8214: conversion from const char* const to char* sometimes 13245 accepted illegally 13246 * [85]8493: builtin strlen and overload resolution (same bug as 13247 [86]8332) 13248 * [87]8503: strange behaviour of function types 13249 * [88]8727: compiler confused by inheritance from an anonymous struct 13250 * [89]7445: poor performance of std::locale::classic() in 13251 multi-threaded applications 13252 * [90]8230: mishandling of overflow in vector<T>::resize 13253 * [91]8399: sync_with_stdio(false) breaks unformatted input 13254 * [92]8662: illegal access of private member of unnamed class is 13255 accepted 13256 * [93]8707: "make distclean" fails in libstdc++-v3 directory 13257 * [94]8708: __USE_MALLOC doesn't work 13258 * [95]8790: Use of non-thread-safe strtok in src/localename.cc 13259 * [96]8887: Bug in date formats with --enable-clocale=generic 13260 * [97]9076: Call Frame Instructions are not handled correctly during 13261 unwind operation 13262 * [98]9151: std::setprecision limited to 16 digits when outputting a 13263 double to a stream 13264 * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers 13265 * [100]9269: libstdc++ headers: explicit specialization of function 13266 must precede its first use 13267 * [101]9322: return value of basic_streambuf<>::getloc affected by 13268 locale::global 13269 * [102]9433: segfault in runtime support for dynamic_cast 13270 13271 C and optimizer bugs 13272 13273 * [103]8032: GCC incorrectly initializes static structs that have 13274 flexible arrays 13275 * [104]8639: simple arithmetic expression broken 13276 * [105]8794: optimization improperly eliminates certain expressions 13277 * [106]8832: traditional "asm volatile" code is illegally optimized 13278 * [107]8988: loop optimizer bug: with -O2, code is generated that 13279 segfaults (found on i386, bug present for all platforms) 13280 * [108]9492: structure copy clobbers subsequent stores to structure 13281 13282 Objective-C bugs 13283 13284 * [109]9267: Objective-C parser won't build with newer bison versions 13285 (e.g. 1.875) 13286 13287 Ada bugs 13288 13289 * [110]8344: Ada build problem due to conflict between gcc/final.o, 13290 gcc/ada/final.o 13291 13292 Preprocessor bugs 13293 13294 * [111]8524: _Pragma within macros is improperly expanded 13295 * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with 13296 -fshort-wchar 13297 13298 ARM-specific 13299 13300 * [113]9090: arm ICE with >= -O2; regression from gcc-2.95 13301 13302 x86-specific (Intel/AMD) 13303 13304 * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction) 13305 * [115]8599: loop unroll bug with -march=k6-3 13306 * [116]9506: ABI breakage in structure return (affects BSD and 13307 Cygwin, but not GNU/Linux) 13308 13309 FreeBSD 5.0 specific 13310 13311 * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0 13312 13313 RTEMS-specific 13314 13315 * [118]9292: hppa1.1-rtems configurery problems 13316 * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug 13317 * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue 13318 * [121]9296: gthr-rtems regression 13319 * [122]9316: powerpc-rtems: extending multilibs 13320 13321 HP-PA specific 13322 13323 * [123]9493: ICE with -O2 when building a simple function 13324 13325 Documentation 13326 13327 * [124]7341: hyperlink to gcov in GCC documentation doesn't work 13328 * [125]8947: Please add a warning about "-malign-double" in docs 13329 * [126]7448, [127]8882: typo cleanups 13330 __________________________________________________________________ 13331 13332 GCC 3.2.1 13333 13334 3.2.1 adds a new warning, -Wabi. This option warns when GNU C++ 13335 generates code that is known not to be binary-compatible with the 13336 vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included 13337 in the distribution, for details. 13338 13339 This release also removes an old GCC extension, "naming types", and the 13340 documentation now directs users to use a different GCC extension, 13341 __typeof__, instead. The feature had evidently been broken for a while. 13342 13343 Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and 13344 the new warning there are no new features that were not present in GCC 13345 3.2. 13346 13347 In addition, the previous fix for [128]PR 7445 (poor performance of 13348 std::locale::classic() in multi-threaded applications) was reverted 13349 ("unfixed"), because the "fix" was not thread-safe. 13350 13351 Bug Fixes 13352 13353 This section lists the problem reports (PRs) from GCC's bug tracking 13354 system that are known to be fixed in the 3.2.1 release. This list might 13355 not be complete (that is, it is possible that some PRs that have been 13356 fixed are not listed here). As you can see, the number of bug fixes is 13357 quite large, so it is strongly recommended that users of earlier GCC 13358 3.x releases upgrade to GCC 3.2.1. 13359 13360 Internal Compiler Errors (multi-platform) 13361 13362 * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c 13363 * [130]5661: (c++) ICE instantiating template on array of unknown 13364 size (bad code) 13365 * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on 13366 64-bit platforms 13367 * [132]6994: (c++) ICE in find_function_data 13368 * [133]7150: preprocessor: GCC -dM -E gives an ICE 13369 * [134]7160: ICE when optimizing branches without a return value 13370 * [135]7228: (c++) ICE when using member template and template 13371 function 13372 * [136]7266: (c++) ICE with -pedantic on missing typename 13373 * [137]7353: ICE from use of "Naming Types" extension, see above 13374 * [138]7411: ICE in instantiate_virtual_regs_1, in function.c 13375 * [139]7478: (c++) ICE on static_cast inside template 13376 * [140]7526: preprocessor core dump when _Pragma implies #pragma 13377 dependency 13378 * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803 13379 is a duplicate) 13380 * [143]7754: (c++) ICE on union with template parameter 13381 * [144]7788: (c++) redeclaring a definition as an incomplete class 13382 causes ICE 13383 * [145]8031: (c++) ICE in comptypes, in cp/typeck.c 13384 * [146]8055: preprocessor dies with SIG11 when building FreeBSD 13385 kernel 13386 * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related 13387 variables 13388 * [148]8134: (c++) ICE in force_store_init_value on legal code 13389 * [149]8149: (c++) ICE on incomplete type 13390 * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array 13391 initialization 13392 13393 C++ (compiler and library) bugs 13394 13395 * [151]5607: No pointer adjustment in covariant return types 13396 * [152]6579: Infinite loop with statement expressions in member 13397 initialization 13398 * [153]6803: Default copy constructor bug in GCC 3.1 13399 * [154]7176: g++ confused by friend and static member with same name 13400 * [155]7188: Segfault with template class and recursive (incorrect) 13401 initializer list 13402 * [156]7306: Regression: GCC 3.x fails to compile code with virtual 13403 inheritance if a method has a variable number of arguments 13404 * [157]7461: ctype<char>::classic_table() returns offset array on 13405 Cygwin 13406 * [158]7524: f(const float arg[3]) fails 13407 * [159]7584: Erroneous ambiguous base error on using declaration 13408 * [160]7676: Member template overloading problem 13409 * [161]7679: infinite loop when a right parenthesis is missing 13410 * [162]7811: default locale not taken from environment 13411 * [163]7961: compare( char *) implemented incorrectly in 13412 basic_string<> 13413 * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if 13414 streambuf::underflow() leaves gptr() NULL (dups: [165]8127, 13415 [166]6745) 13416 * [167]8096: deque::at() throws std::range_error instead of 13417 std::out_of_range 13418 * [168]8127: cout << cin.rdbuf() infinite loop 13419 * [169]8218: Excessively large memory consumed for classes with large 13420 array members 13421 * [170]8287: GCC 3.2: Destructor called for non-constructed local 13422 object 13423 * [171]8347: empty vector range used in string construction causes 13424 core dump 13425 * [172]8348: fail() flag is set in istringstream when eof() flag is 13426 set 13427 * [173]8391: regression: infinite loop in cp/decl2.c(finish_file) 13428 13429 C and optimizer bugs 13430 13431 * [174]6627: -fno-align-functions doesn't seem to disable function 13432 alignment 13433 * [175]6631: life_analysis misoptimizes code to initialize fields of 13434 a structure 13435 * [176]7102: unsigned char division results in floating exception 13436 * [177]7120: Run once loop should *always* be unrolled 13437 (pessimization) 13438 * [178]7209: Bug involving array referencing and ?: operator 13439 * [179]7515: invalid inlining of global function with -O3 13440 * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test 13441 * [181]8467: bug in sibling call optimization 13442 13443 Preprocessor bugs 13444 13445 * [182]4890: incorrect line markers from the traditional preprocessor 13446 * [183]7357: -M option omits system headers files (making it the same 13447 as -MM) 13448 * [184]7358: Changes to Sun's make Dependencies 13449 * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as 13450 C headers 13451 * [186]7862: Interrupting GCC -MD removes .d file but not .o 13452 * [187]8190: Failed compilation deletes -MD dependency file 13453 * [188]8524: _Pragma within macro is improperly expanded 13454 13455 x86 specific (Intel/AMD) 13456 13457 * [189]5351: (i686-only) function pass-by-value structure copy 13458 corrupts stack ([190]7591 is a duplicate) 13459 * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with 13460 -march=pentium3/pentium2/athlon (these are all the same underlying 13461 bug, in MMX register use) 13462 * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same 13463 as above?) 13464 * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken 13465 * [199]6981: wrong code in 64-bit manipulation on x86 13466 * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__ 13467 macro 13468 * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE 13469 intrinsics are broken 13470 * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with 13471 -march=pentium4 13472 * [203]7693: Typo in i386 mmintrin.h header 13473 * [204]7723: ICE - Pentium3 sse - GCC 3.2 13474 * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse 13475 * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3 13476 13477 PowerPC specific 13478 13479 * [207]5967: GCC bug when profiling nested functions on powerpc 13480 * [208]6984: wrong code generated with -O2, -O3, -Os for do-while 13481 loop on PowerPC 13482 * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5 13483 * [210]7130: miscompiled code for GCC-3.1 on 13484 powerpc-unknown-linux-gnu with -funroll-all-loops 13485 * [211]7133: PowerPC ICE: unrecognizable insn 13486 * [212]7380: ICE in extract_insn, at recog.c:2148 13487 * [213]8252: ICE on Altivec code with optimization turned on 13488 * [214]8451: Altivec ICE in GCC 3.2 13489 13490 HP/PA specific 13491 13492 * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa 13493 13494 SPARC specific 13495 13496 * [216]6668: when using --disable-multilib, libgcc_s.so is installed 13497 in the wrong place on sparc-solaris 13498 * [217]7151: ICE when compiling for UltraSPARC 13499 * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long 13500 double and -O1 13501 * [219]7842: [REGRESSION] SPARC code gen bug 13502 13503 ARM specific 13504 13505 * [220]7856: [arm] invalid offset in constant pool reference 13506 * [221]7967: optimization produces wrong code (ARM) 13507 13508 Alpha specific 13509 13510 * [222]7374: __builtin_fabsl broken on alpha 13511 13512 IBM s390 specific 13513 13514 * [223]7370: ICE in fixup_var_refs_1 on s390x 13515 * [224]7409: loop optimization bug on s390x-linux-gnu 13516 * [225]8232: s390x: ICE when using bcmp with int length argument 13517 13518 SCO specific 13519 13520 * [226]7623: SCO OpenServer build fails with machmode.def: undefined 13521 symbol: BITS_PER_UNIT 13522 13523 m68k/Coldfire specific 13524 13525 * [227]8314: crtbegin, crtend need to be multilib'ed for this 13526 platform 13527 13528 Documentation 13529 13530 * [228]761: Document some undocumented options 13531 * [229]5610: Fix documentation about invoking SSE instructions 13532 (-mfpmath=sse) 13533 * [230]7484: List -Wmissing-declarations as C-only option 13534 * [231]7531: -mcmodel not documented for x86-64 13535 * [232]8120: Update documentation of bad use of ## 13536 __________________________________________________________________ 13537 13538 GCC 3.2 13539 13540 3.2 is a small bug fix release, but there is a change to the 13541 application binary interface (ABI), hence the change to the second part 13542 of the version number. 13543 13544 The main purpose of the 3.2 release is to correct a couple of problems 13545 in the C++ ABI, with the intention of providing a stable interface 13546 going forward. Accordingly, 3.2 is only a small change to 3.1.1. 13547 13548 Bug Fixes 13549 13550 C++ 13551 13552 * [233]7320: g++ 3.2 relocation problem 13553 * [234]7470: vtable: virtual function pointers not in declaration 13554 order 13555 13556 libstdc++ 13557 13558 * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t 13559 * [236]6503, [237]6642, [238]7186: Problems with comparing or 13560 subtracting various types of const and non-const iterators 13561 * [239]7216: ambiguity with basic_iostream::traits_type 13562 * [240]7220: problem with basic_istream::ignore(0,delimiter) 13563 * [241]7222: locale::operator==() doesn't work on std::locale("") 13564 * [242]7286: placement operator delete issue 13565 * [243]7442: cxxabi.h does not match the C++ ABI 13566 * [244]7445: poor performance of std::locale::classic() in 13567 multi-threaded applications 13568 13569 x86-64 specific 13570 13571 * [245]7291: off-by-one in generated inline bzero code for x86-64 13572 13573 13574 For questions related to the use of GCC, please consult these web 13575 pages and the [246]GCC manuals. If that fails, the 13576 [247]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13577 web pages and the development of GCC are welcome on our developer 13578 list at [248]gcc (a] gcc.gnu.org. All of [249]our lists have public 13579 archives. 13580 13581 Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and 13582 distribution of this entire article is permitted in any medium, 13583 provided this notice is preserved. 13584 13585 These pages are [251]maintained by the GCC team. Last modified 13586 2014-06-27[252]. 13587 13588 References 13589 13590 1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 13591 2. http://gcc.gnu.org/gcc-3.1/changes.html 13592 3. http://gcc.gnu.org/gcc-3.2/c++-abi.html 13593 4. https://gcc.gnu.org/PR3782 13594 5. https://gcc.gnu.org/PR6440 13595 6. https://gcc.gnu.org/PR7050 13596 7. https://gcc.gnu.org/PR7741 13597 8. https://gcc.gnu.org/PR7982 13598 9. https://gcc.gnu.org/PR8068 13599 10. https://gcc.gnu.org/PR8178 13600 11. https://gcc.gnu.org/PR8396 13601 12. https://gcc.gnu.org/PR8674 13602 13. https://gcc.gnu.org/PR9768 13603 14. https://gcc.gnu.org/PR9798 13604 15. https://gcc.gnu.org/PR9799 13605 16. https://gcc.gnu.org/PR9928 13606 17. https://gcc.gnu.org/PR10114 13607 18. https://gcc.gnu.org/PR10352 13608 19. https://gcc.gnu.org/PR10336 13609 20. https://gcc.gnu.org/PR8224 13610 21. https://gcc.gnu.org/PR8613 13611 22. https://gcc.gnu.org/PR8828 13612 23. https://gcc.gnu.org/PR9226 13613 24. https://gcc.gnu.org/PR9853 13614 25. https://gcc.gnu.org/PR9797 13615 26. https://gcc.gnu.org/PR9967 13616 27. https://gcc.gnu.org/PR10116 13617 28. https://gcc.gnu.org/PR10171 13618 29. https://gcc.gnu.org/PR10175 13619 30. https://gcc.gnu.org/PR8316 13620 31. https://gcc.gnu.org/PR9169 13621 32. https://gcc.gnu.org/PR9420 13622 33. https://gcc.gnu.org/PR9459 13623 34. https://gcc.gnu.org/PR9507 13624 35. https://gcc.gnu.org/PR9538 13625 36. https://gcc.gnu.org/PR9602 13626 37. https://gcc.gnu.org/PR9993 13627 38. https://gcc.gnu.org/PR10167 13628 39. https://gcc.gnu.org/PR9652 13629 40. https://gcc.gnu.org/PR10144 13630 41. https://gcc.gnu.org/PR8746 13631 42. https://gcc.gnu.org/PR9888 13632 43. https://gcc.gnu.org/PR9638 13633 44. https://gcc.gnu.org/PR9954 13634 45. https://gcc.gnu.org/PR7784 13635 46. https://gcc.gnu.org/PR7796 13636 47. https://gcc.gnu.org/PR8281 13637 48. https://gcc.gnu.org/PR8366 13638 49. https://gcc.gnu.org/PR8726 13639 50. https://gcc.gnu.org/PR9414 13640 51. https://gcc.gnu.org/PR10067 13641 52. https://gcc.gnu.org/PR7248 13642 53. https://gcc.gnu.org/PR8343 13643 54. https://gcc.gnu.org/PR9732 13644 55. https://gcc.gnu.org/PR10073 13645 56. https://gcc.gnu.org/PR7702 13646 57. https://gcc.gnu.org/PR9671 13647 58. https://gcc.gnu.org/PR8694 13648 59. https://gcc.gnu.org/PR9953 13649 60. https://gcc.gnu.org/PR10271 13650 61. https://gcc.gnu.org/PR6362 13651 62. https://gcc.gnu.org/PR10377 13652 63. https://gcc.gnu.org/PR6955 13653 64. https://gcc.gnu.org/PR5919 13654 65. https://gcc.gnu.org/PR7129 13655 66. https://gcc.gnu.org/PR7507 13656 67. https://gcc.gnu.org/PR7622 13657 68. https://gcc.gnu.org/PR7681 13658 69. https://gcc.gnu.org/PR9528 13659 70. https://gcc.gnu.org/PR8031 13660 71. https://gcc.gnu.org/PR8275 13661 72. https://gcc.gnu.org/PR8332 13662 73. https://gcc.gnu.org/PR8372 13663 74. https://gcc.gnu.org/PR8439 13664 75. https://gcc.gnu.org/PR8442 13665 76. https://gcc.gnu.org/PR8518 13666 77. https://gcc.gnu.org/PR8615 13667 78. https://gcc.gnu.org/PR8663 13668 79. https://gcc.gnu.org/PR8799 13669 80. https://gcc.gnu.org/PR9328 13670 81. https://gcc.gnu.org/PR9465 13671 82. https://gcc.gnu.org/PR47 13672 83. https://gcc.gnu.org/PR6745 13673 84. https://gcc.gnu.org/PR8214 13674 85. https://gcc.gnu.org/PR8493 13675 86. https://gcc.gnu.org/PR8332 13676 87. https://gcc.gnu.org/PR8503 13677 88. https://gcc.gnu.org/PR8727 13678 89. https://gcc.gnu.org/PR7445 13679 90. https://gcc.gnu.org/PR8230 13680 91. https://gcc.gnu.org/PR8399 13681 92. https://gcc.gnu.org/PR8662 13682 93. https://gcc.gnu.org/PR8707 13683 94. https://gcc.gnu.org/PR8708 13684 95. https://gcc.gnu.org/PR8790 13685 96. https://gcc.gnu.org/PR8887 13686 97. https://gcc.gnu.org/PR9076 13687 98. https://gcc.gnu.org/PR9151 13688 99. https://gcc.gnu.org/PR9168 13689 100. https://gcc.gnu.org/PR9269 13690 101. https://gcc.gnu.org/PR9322 13691 102. https://gcc.gnu.org/PR9433 13692 103. https://gcc.gnu.org/PR8032 13693 104. https://gcc.gnu.org/PR8639 13694 105. https://gcc.gnu.org/PR8794 13695 106. https://gcc.gnu.org/PR8832 13696 107. https://gcc.gnu.org/PR8988 13697 108. https://gcc.gnu.org/PR9492 13698 109. https://gcc.gnu.org/PR9267 13699 110. https://gcc.gnu.org/PR8344 13700 111. https://gcc.gnu.org/PR8524 13701 112. https://gcc.gnu.org/PR8880 13702 113. https://gcc.gnu.org/PR9090 13703 114. https://gcc.gnu.org/PR8588 13704 115. https://gcc.gnu.org/PR8599 13705 116. https://gcc.gnu.org/PR9506 13706 117. https://gcc.gnu.org/PR9484 13707 118. https://gcc.gnu.org/PR9292 13708 119. https://gcc.gnu.org/PR9293 13709 120. https://gcc.gnu.org/PR9295 13710 121. https://gcc.gnu.org/PR9296 13711 122. https://gcc.gnu.org/PR9316 13712 123. https://gcc.gnu.org/PR9493 13713 124. https://gcc.gnu.org/PR7341 13714 125. https://gcc.gnu.org/PR8947 13715 126. https://gcc.gnu.org/PR7448 13716 127. https://gcc.gnu.org/PR8882 13717 128. https://gcc.gnu.org/PR7445 13718 129. https://gcc.gnu.org/PR2521 13719 130. https://gcc.gnu.org/PR5661 13720 131. https://gcc.gnu.org/PR6419 13721 132. https://gcc.gnu.org/PR6994 13722 133. https://gcc.gnu.org/PR7150 13723 134. https://gcc.gnu.org/PR7160 13724 135. https://gcc.gnu.org/PR7228 13725 136. https://gcc.gnu.org/PR7266 13726 137. https://gcc.gnu.org/PR7353 13727 138. https://gcc.gnu.org/PR7411 13728 139. https://gcc.gnu.org/PR7478 13729 140. https://gcc.gnu.org/PR7526 13730 141. https://gcc.gnu.org/PR7721 13731 142. https://gcc.gnu.org/PR7803 13732 143. https://gcc.gnu.org/PR7754 13733 144. https://gcc.gnu.org/PR7788 13734 145. https://gcc.gnu.org/PR8031 13735 146. https://gcc.gnu.org/PR8055 13736 147. https://gcc.gnu.org/PR8067 13737 148. https://gcc.gnu.org/PR8134 13738 149. https://gcc.gnu.org/PR8149 13739 150. https://gcc.gnu.org/PR8160 13740 151. https://gcc.gnu.org/PR5607 13741 152. https://gcc.gnu.org/PR6579 13742 153. https://gcc.gnu.org/PR6803 13743 154. https://gcc.gnu.org/PR7176 13744 155. https://gcc.gnu.org/PR7188 13745 156. https://gcc.gnu.org/PR7306 13746 157. https://gcc.gnu.org/PR7461 13747 158. https://gcc.gnu.org/PR7524 13748 159. https://gcc.gnu.org/PR7584 13749 160. https://gcc.gnu.org/PR7676 13750 161. https://gcc.gnu.org/PR7679 13751 162. https://gcc.gnu.org/PR7811 13752 163. https://gcc.gnu.org/PR7961 13753 164. https://gcc.gnu.org/PR8071 13754 165. https://gcc.gnu.org/PR8127 13755 166. https://gcc.gnu.org/PR6745 13756 167. https://gcc.gnu.org/PR8096 13757 168. https://gcc.gnu.org/PR8127 13758 169. https://gcc.gnu.org/PR8218 13759 170. https://gcc.gnu.org/PR8287 13760 171. https://gcc.gnu.org/PR8347 13761 172. https://gcc.gnu.org/PR8348 13762 173. https://gcc.gnu.org/PR8391 13763 174. https://gcc.gnu.org/PR6627 13764 175. https://gcc.gnu.org/PR6631 13765 176. https://gcc.gnu.org/PR7102 13766 177. https://gcc.gnu.org/PR7120 13767 178. https://gcc.gnu.org/PR7209 13768 179. https://gcc.gnu.org/PR7515 13769 180. https://gcc.gnu.org/PR7814 13770 181. https://gcc.gnu.org/PR8467 13771 182. https://gcc.gnu.org/PR4890 13772 183. https://gcc.gnu.org/PR7357 13773 184. https://gcc.gnu.org/PR7358 13774 185. https://gcc.gnu.org/PR7602 13775 186. https://gcc.gnu.org/PR7862 13776 187. https://gcc.gnu.org/PR8190 13777 188. https://gcc.gnu.org/PR8524 13778 189. https://gcc.gnu.org/PR5351 13779 190. https://gcc.gnu.org/PR7591 13780 191. https://gcc.gnu.org/PR6845 13781 192. https://gcc.gnu.org/PR7034 13782 193. https://gcc.gnu.org/PR7124 13783 194. https://gcc.gnu.org/PR7174 13784 195. https://gcc.gnu.org/PR7134 13785 196. https://gcc.gnu.org/PR7375 13786 197. https://gcc.gnu.org/PR7390 13787 198. https://gcc.gnu.org/PR6890 13788 199. https://gcc.gnu.org/PR6981 13789 200. https://gcc.gnu.org/PR7242 13790 201. https://gcc.gnu.org/PR7396 13791 202. https://gcc.gnu.org/PR7630 13792 203. https://gcc.gnu.org/PR7693 13793 204. https://gcc.gnu.org/PR7723 13794 205. https://gcc.gnu.org/PR7951 13795 206. https://gcc.gnu.org/PR8146 13796 207. https://gcc.gnu.org/PR5967 13797 208. https://gcc.gnu.org/PR6984 13798 209. https://gcc.gnu.org/PR7114 13799 210. https://gcc.gnu.org/PR7130 13800 211. https://gcc.gnu.org/PR7133 13801 212. https://gcc.gnu.org/PR7380 13802 213. https://gcc.gnu.org/PR8252 13803 214. https://gcc.gnu.org/PR8451 13804 215. https://gcc.gnu.org/PR7250 13805 216. https://gcc.gnu.org/PR6668 13806 217. https://gcc.gnu.org/PR7151 13807 218. https://gcc.gnu.org/PR7335 13808 219. https://gcc.gnu.org/PR7842 13809 220. https://gcc.gnu.org/PR7856 13810 221. https://gcc.gnu.org/PR7967 13811 222. https://gcc.gnu.org/PR7374 13812 223. https://gcc.gnu.org/PR7370 13813 224. https://gcc.gnu.org/PR7409 13814 225. https://gcc.gnu.org/PR8232 13815 226. https://gcc.gnu.org/PR7623 13816 227. https://gcc.gnu.org/PR8314 13817 228. https://gcc.gnu.org/PR761 13818 229. https://gcc.gnu.org/PR5610 13819 230. https://gcc.gnu.org/PR7484 13820 231. https://gcc.gnu.org/PR7531 13821 232. https://gcc.gnu.org/PR8120 13822 233. https://gcc.gnu.org/PR7320 13823 234. https://gcc.gnu.org/PR7470 13824 235. https://gcc.gnu.org/PR6410 13825 236. https://gcc.gnu.org/PR6503 13826 237. https://gcc.gnu.org/PR6642 13827 238. https://gcc.gnu.org/PR7186 13828 239. https://gcc.gnu.org/PR7216 13829 240. https://gcc.gnu.org/PR7220 13830 241. https://gcc.gnu.org/PR7222 13831 242. https://gcc.gnu.org/PR7286 13832 243. https://gcc.gnu.org/PR7442 13833 244. https://gcc.gnu.org/PR7445 13834 245. https://gcc.gnu.org/PR7291 13835 246. https://gcc.gnu.org/onlinedocs/ 13836 247. mailto:gcc-help (a] gcc.gnu.org 13837 248. mailto:gcc (a] gcc.gnu.org 13838 249. https://gcc.gnu.org/lists.html 13839 250. http://www.fsf.org/ 13840 251. https://gcc.gnu.org/about.html 13841 252. http://validator.w3.org/check/referer 13842 ====================================================================== 13843 http://gcc.gnu.org/gcc-3.1/index.html 13844 13845 GCC 3.1 13846 13847 July 27, 2002 13848 13849 The [1]GNU project and the GCC developers are pleased to announce the 13850 release of GCC 3.1.1. 13851 13852 The links below still apply to GCC 3.1.1. 13853 13854 May 15, 2002 13855 13856 The [2]GNU project and the GCC developers are pleased to announce the 13857 release of GCC 3.1. 13858 13859 GCC used to stand for the GNU C Compiler, but since the compiler 13860 supports several other languages aside from C, it now stands for the 13861 GNU Compiler Collection. 13862 13863 A list of [3]successful builds is updated as new information becomes 13864 available. 13865 13866 The GCC developers would like to thank the numerous people that have 13867 contributed [4]new features, improvements, bug fixes, and other changes 13868 as well as test results to GCC. This [5]amazing group of volunteers is 13869 what makes GCC successful. 13870 13871 For additional information about GCC please refer to the [6]GCC project 13872 web site or contact the [7]GCC development mailing list. 13873 13874 To obtain GCC please use [8]our mirror sites, or our CVS server. 13875 __________________________________________________________________ 13876 13877 13878 For questions related to the use of GCC, please consult these web 13879 pages and the [9]GCC manuals. If that fails, the 13880 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13881 web pages and the development of GCC are welcome on our developer 13882 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 13883 archives. 13884 13885 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 13886 distribution of this entire article is permitted in any medium, 13887 provided this notice is preserved. 13888 13889 These pages are [14]maintained by the GCC team. Last modified 13890 2014-06-28[15]. 13891 13892 References 13893 13894 1. http://www.gnu.org/ 13895 2. http://www.gnu.org/ 13896 3. http://gcc.gnu.org/gcc-3.1/buildstat.html 13897 4. http://gcc.gnu.org/gcc-3.1/changes.html 13898 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 13899 6. http://gcc.gnu.org/index.html 13900 7. mailto:gcc (a] gcc.gnu.org 13901 8. http://gcc.gnu.org/mirrors.html 13902 9. https://gcc.gnu.org/onlinedocs/ 13903 10. mailto:gcc-help (a] gcc.gnu.org 13904 11. mailto:gcc (a] gcc.gnu.org 13905 12. https://gcc.gnu.org/lists.html 13906 13. http://www.fsf.org/ 13907 14. https://gcc.gnu.org/about.html 13908 15. http://validator.w3.org/check/referer 13909 ====================================================================== 13910 http://gcc.gnu.org/gcc-3.1/changes.html 13911 13912 GCC 3.1 Release Series 13913 Changes, New Features, and Fixes 13914 13915 Additional changes in GCC 3.1.1 13916 13917 * A bug related to how structures and unions are returned has been 13918 fixed for powerpc-*-netbsd*. 13919 * An important bug in the implementation of -fprefetch-loop-arrays 13920 has been fixed. Previously the optimization prefetched random 13921 blocks of memory for most targets except for i386. 13922 * The Java compiler now compiles Java programs much faster and also 13923 works with parallel make. 13924 * Nested functions have been fixed for mips*-*-netbsd*. 13925 * Some missing floating point support routines have beed added for 13926 mips*-*-netbsd*. 13927 * This [1]message gives additional information about the bugs fixed 13928 in this release. 13929 13930 Caveats 13931 13932 * The -traditional C compiler option has been deprecated and will be 13933 removed in GCC 3.3. (It remains possible to preprocess non-C code 13934 with the traditional preprocessor.) 13935 * The default debugging format for most ELF platforms (including 13936 GNU/Linux and FreeBSD; notable exception is Solaris) has changed 13937 from stabs to DWARF2. This requires GDB 5.1.1 or later. 13938 13939 General Optimizer Improvements 13940 13941 * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat, 13942 and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure 13943 for profile driven optimizations. 13944 Options -fprofile-arcs and -fbranch-probabilities can now be used 13945 to improve speed of the generated code by profiling the actual 13946 program behaviour on typical runs. In the absence of profile info 13947 the compiler attempts to guess the profile statically. 13948 * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to 13949 monitor performance of the generated code. 13950 According to the SPECInt2000 results on an AMD Athlon CPU, the code 13951 generated by GCC 3.1 is 6% faster on the average (8.2% faster with 13952 profile feedback) compared to GCC 3.0. The code produced by GCC 3.0 13953 is about 2.1% faster compared to 2.95.3. Tests were done using the 13954 -O2 -march=athlon command-line options. 13955 * Alexandre Oliva, of Red Hat, has generalized the tree inlining 13956 infrastructure developed by CodeSourcery, LLC for the C++ front 13957 end, so that it is now used in the C front end too. Inlining 13958 functions as trees exposes them earlier to the compiler, giving it 13959 more opportunities for optimization. 13960 * Support for data prefetching instructions has been added to the GCC 13961 back end and several targets. A new __builtin_prefetch intrinsic is 13962 available to explicitly insert prefetch instructions and 13963 experimental support for loop array prefetching has been added (see 13964 -fprefetch-loop-array documentation). 13965 * Support for emitting debugging information for macros has been 13966 added for DWARF2. It is activated using -g3. 13967 13968 New Languages and Language specific improvements 13969 13970 C/C++ 13971 13972 * A few more [4]ISO C99 features. 13973 * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0. 13974 * The preprocessor's symbol table has been merged with the symbol 13975 table of the C, C++ and Objective-C front ends. 13976 * The preprocessor consumes less memory than the preprocessor in GCC 13977 3.0, often significantly so. On normal input files, it typically 13978 consumes less memory than pre-3.0 cccp-based GCC, too. 13979 13980 C++ 13981 13982 * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std 13983 was a workaround to allow std compliant code to work with the 13984 non-std compliant libstdc++-v2. libstdc++-v3 is std compliant. 13985 * The C++ ABI has been fixed so that void (A::*)() const is mangled 13986 as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only 13987 affects pointer to cv-qualified member function types. 13988 * The C++ ABI has been changed to correctly handle this code: 13989 struct A { 13990 void operator delete[] (void *, size_t); 13991 }; 13992 13993 struct B : public A { 13994 }; 13995 13996 new B[10]; 13997 13998 The amount of storage allocated for the array will be greater than 13999 it was in 3.0, in order to store the number of elements in the 14000 array, so that the correct size can be passed to operator delete[] 14001 when the array is deleted. Previously, the value passed to operator 14002 delete[] was unpredictable. 14003 This change will only affect code that declares a two-argument 14004 operator delete[] with a second parameter of type size_t in a base 14005 class, and does not override that definition in a derived class. 14006 * The C++ ABI has been changed so that: 14007 struct A { 14008 void operator delete[] (void *, size_t); 14009 void operator delete[] (void *); 14010 }; 14011 14012 does not cause unnecessary storage to be allocated when an array of 14013 A objects is allocated. 14014 This change will only affect code that declares both of these forms 14015 of operator delete[], and declared the two-argument form before the 14016 one-argument form. 14017 * The C++ ABI has been changed so that when a parameter is passed by 14018 value, any cleanup for that parameter is performed in the caller, 14019 as specified by the ia64 C++ ABI, rather than the called function 14020 as before. As a result, classes with a non-trivial destructor but a 14021 trivial copy constructor will be passed and returned by invisible 14022 reference, rather than by bitwise copy as before. 14023 * G++ now supports the "named return value optimization": for code 14024 like 14025 A f () { 14026 A a; 14027 ... 14028 return a; 14029 } 14030 14031 G++ will allocate a in the return value slot, so that the return 14032 becomes a no-op. For this to work, all return statements in the 14033 function must return the same variable. 14034 * Improvements to the C++ library are listed in [5]the libstdc++-v3 14035 FAQ. 14036 14037 Objective-C 14038 14039 * Annoying linker warnings (due to incorrect code being generated) 14040 have been fixed. 14041 * If a class method cannot be found, the compiler no longer issues a 14042 warning if a corresponding instance method exists in the root 14043 class. 14044 * Forward @protocol declarations have been fixed. 14045 * Loading of categories has been fixed in certain situations (GNU run 14046 time only). 14047 * The class lookup in the run-time library has been rewritten so that 14048 class method dispatch is more than twice as fast as it used to be 14049 (GNU run time only). 14050 14051 Java 14052 14053 * libgcj now includes RMI, java.lang.ref.*, javax.naming, and 14054 javax.transaction. 14055 * Property files and other system resources can be compiled into 14056 executables which use libgcj using the new gcj --resource feature. 14057 * libgcj has been ported to more platforms. In particular there is 14058 now a mostly-functional mingw32 (Windows) target port. 14059 * JNI and CNI invocation interfaces were implemented, so gcj-compiled 14060 Java code can now be called from a C/C++ application. 14061 * gcj can now use builtin functions for certain known methods, for 14062 instance Math.cos. 14063 * gcj can now automatically remove redundant array-store checks in 14064 some common cases. 14065 * The --no-store-checks optimization option was added. This can be 14066 used to omit runtime store checks for code which is known not to 14067 throw ArrayStoreException 14068 * The following third party interface standards were added to libgcj: 14069 org.w3c.dom and org.xml.sax. 14070 * java.security has been merged with GNU Classpath. The new package 14071 is now JDK 1.2 compliant, and much more complete. 14072 * A bytecode verifier was added to the libgcj interpreter. 14073 * java.lang.Character was rewritten to comply with the Unicode 3.0 14074 standard, and improve performance. 14075 * Partial support for many more locales was added to libgcj. 14076 * Socket timeouts have been implemented. 14077 * libgcj has been merged into a single shared library. There are no 14078 longer separate shared libraries for the garbage collector and 14079 zlib. 14080 * Several performance improvements were made to gcj and libgcj: 14081 + Hash synchronization (thin locks) 14082 + A special allocation path for finalizer-free objects 14083 + Thread-local allocation 14084 + Parallel GC, and other GC tweaks 14085 14086 Fortran 14087 14088 Fortran improvements are listed in [6]the Fortran documentation. 14089 14090 Ada 14091 14092 [7]Ada Core Technologies, Inc, has contributed its GNAT Ada 95 front 14093 end and associated tools. The GNAT compiler fully implements the Ada 14094 language as defined by the ISO/IEC 8652 standard. 14095 14096 Please note that the integration of the Ada front end is still work in 14097 progress. 14098 14099 New Targets and Target Specific Improvements 14100 14101 * Hans-Peter Nilsson has contributed a port to [8]MMIX, the CPU 14102 architecture used in new editions of Donald E. Knuth's The Art of 14103 Computer Programming. 14104 * [9]Axis Communications has contributed its port to the CRIS CPU 14105 architecture, used in the ETRAX system-on-a-chip series. See 14106 [10]Axis' developer site for technical information. 14107 * Alexandre Oliva, of Red Hat, has contributed a port to the 14108 [11]SuperH SH5 64-bit RISC microprocessor architecture, extending 14109 the existing SH port. 14110 * UltraSPARC is fully supported in 64-bit mode. The option -m64 14111 enables it. 14112 * For compatibility with the Sun compiler #pragma redefine_extname 14113 has been implemented on Solaris. 14114 * The x86 back end has had some noticeable work done to it. 14115 + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas 14116 Jaeger have contributed a port to the AMD x86-64 architecture. 14117 For more information on x86-64 see [12]http://www.x86-64.org. 14118 + The compiler now supports MMX, 3DNow!, SSE, and SSE2 14119 instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will 14120 enable the respective instruction sets. Intel C++ compatible 14121 MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics 14122 will be added in next major release. 14123 + Following those improvements, targets for Pentium MMX, K6-2, 14124 K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were 14125 added. Refer to the documentation on -march= and -mcpu= 14126 options for details. 14127 + For those targets that support it, -mfpmath=sse will cause the 14128 compiler to generate SSE/SSE2 instructions for floating point 14129 math instead of x87 instructions. Usually, this will lead to 14130 quicker code -- especially on the Pentium 4. Note that only 14131 scalar floating point instructions are used and GCC does not 14132 exploit SIMD features yet. 14133 + Prefetch support has been added to the Pentium III, Pentium 4, 14134 K6-2, K6-3, and Athlon series. 14135 + Code generated for floating point to integer conversions has 14136 been improved leading to better performance of many 3D 14137 applications. 14138 * The PowerPC back end has added 64-bit PowerPC GNU/Linux support. 14139 * C++ support for AIX has been improved. 14140 * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the 14141 PowerPC port supporting the AltiVec programming model (SIMD). The 14142 support, though presently useful, is experimental and is expected 14143 to stabilize for 3.2. The support is written to conform to 14144 Motorola's AltiVec specs. See -maltivec. 14145 14146 Obsolete Systems 14147 14148 Support for a number of older systems has been declared obsolete in GCC 14149 3.1. Unless there is activity to revive them, the next release of GCC 14150 will have their sources permanently removed. 14151 14152 All configurations of the following processor architectures have been 14153 declared obsolete: 14154 * MIL-STD-1750A, 1750a-*-* 14155 * AMD A29k, a29k-*-* 14156 * Convex, c*-convex-* 14157 * Clipper, clipper-*-* 14158 * Elxsi, elxsi-*-* 14159 * Intel i860, i860-*-* 14160 * Sun picoJava, pj-*-* and pjl-*-* 14161 * Western Electric 32000, we32k-*-* 14162 14163 Most configurations of the following processor architectures have been 14164 declared obsolete, but we are preserving a few systems which may have 14165 active developers. It is unlikely that the remaining systems will 14166 survive much longer unless we see definite signs of port activity. 14167 * Motorola 88000 except 14168 + Generic a.out, m88k-*-aout* 14169 + Generic SVR4, m88k-*-sysv4 14170 + OpenBSD, m88k-*-openbsd* 14171 * NS32k except 14172 + NetBSD, ns32k-*-netbsd* 14173 + OpenBSD, ns32k-*-openbsd*. 14174 * ROMP except 14175 + OpenBSD, romp-*-openbsd*. 14176 14177 Finally, only some configurations of these processor architectures are 14178 being obsoleted. 14179 * Alpha: 14180 + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka 14181 alpha*-*-osf[45], are still supported.) 14182 * ARM: 14183 + RISCiX, arm-*-riscix*. 14184 * i386: 14185 + 386BSD, i?86-*-bsd* 14186 + Chorus, i?86-*-chorusos* 14187 + DG/UX, i?86-*-dgux* 14188 + FreeBSD 1.x, i?86-*-freebsd1.* 14189 + IBM AIX, i?86-*-aix* 14190 + ISC UNIX, i?86-*-isc* 14191 + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld* 14192 + NEXTstep, i?86-next-* 14193 + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose* 14194 + RTEMS/coff, i?86-*-rtemscoff* 14195 + RTEMS/go32, i?86-go32-rtems* 14196 + Sequent/BSD, i?86-sequent-bsd* 14197 + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and 14198 i?86-sequent-sysv3* 14199 + SunOS, i?86-*-sunos* 14200 * Motorola 68000: 14201 + Altos, m68[k0]*-altos-* 14202 + Apollo, m68[k0]*-apollo-* 14203 + Apple A/UX, m68[k0]*-apple-* 14204 + Bull, m68[k0]*-bull-* 14205 + Convergent, m68[k0]*-convergent-* 14206 + Generic SVR3, m68[k0]*-*-sysv3* 14207 + ISI, m68[k0]*-isi-* 14208 + LynxOS, m68[k0]*-*-lynxos* 14209 + NEXT, m68[k0]*-next-* 14210 + RTEMS/coff, m68[k0]*-*-rtemscoff* 14211 + Sony, m68[k0]*-sony-* 14212 * MIPS: 14213 + DEC Ultrix, mips-*-ultrix* and mips-dec-* 14214 + Generic BSD, mips-*-bsd* 14215 + Generic System V, mips-*-sysv* 14216 + IRIX before version 5, mips-sgi-irix[1234]* 14217 + RiscOS, mips-*-riscos* 14218 + Sony, mips-sony-* 14219 + Tandem, mips-tandem-* 14220 * SPARC: 14221 + RTEMS/a.out, sparc-*-rtemsaout*. 14222 14223 Documentation improvements 14224 14225 * The old manual ("Using and Porting the GNU Compiler Collection") 14226 has been replaced by a users manual ("Using the GNU Compiler 14227 Collection") and a separate internals reference manual ("GNU 14228 Compiler Collection Internals"). 14229 * More complete and much improved documentation about GCC's internal 14230 representation used by the C and C++ front ends. 14231 * Many cleanups and improvements in general. 14232 14233 14234 For questions related to the use of GCC, please consult these web 14235 pages and the [13]GCC manuals. If that fails, the 14236 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14237 web pages and the development of GCC are welcome on our developer 14238 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 14239 archives. 14240 14241 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 14242 distribution of this entire article is permitted in any medium, 14243 provided this notice is preserved. 14244 14245 These pages are [18]maintained by the GCC team. Last modified 14246 2015-06-29[19]. 14247 14248 References 14249 14250 1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html 14251 2. http://gcc.gnu.org/news/profiledriven.html 14252 3. http://gcc.gnu.org/benchmarks/ 14253 4. http://gcc.gnu.org/c99status.html 14254 5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html 14255 6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html 14256 7. http://www.adacore.com/ 14257 8. http://www-cs-faculty.stanford.edu/~uno/mmix.html 14258 9. http://www.axis.com/ 14259 10. http://developer.axis.com/ 14260 11. http://www.superh.com/ 14261 12. http://www.x86-64.org/ 14262 13. https://gcc.gnu.org/onlinedocs/ 14263 14. mailto:gcc-help (a] gcc.gnu.org 14264 15. mailto:gcc (a] gcc.gnu.org 14265 16. https://gcc.gnu.org/lists.html 14266 17. http://www.fsf.org/ 14267 18. https://gcc.gnu.org/about.html 14268 19. http://validator.w3.org/check/referer 14269 ====================================================================== 14270 http://gcc.gnu.org/gcc-3.0/index.html 14271 14272 GCC 3.0.4 14273 14274 February 20, 2002 14275 14276 The [1]GNU project and the GCC developers are pleased to announce the 14277 release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0 14278 series. 14279 14280 GCC used to stand for the GNU C Compiler, but since the compiler 14281 supports several other languages aside from C, it now stands for the 14282 GNU Compiler Collection. 14283 14284 GCC 3.0.x has several new optimizations, new targets, new languages and 14285 many other new features, relative to GCC 2.95.x. See the [2]new 14286 features page for a more complete list. 14287 14288 A list of [3]successful builds is updated as new information becomes 14289 available. 14290 14291 The GCC developers would like to thank the numerous people that have 14292 contributed new features, test results, bug fixes, etc to GCC. This 14293 [4]amazing group of volunteers is what makes GCC successful. 14294 14295 And finally, we can't in good conscience fail to mention some 14296 [5]caveats to using GCC 3.0.x. 14297 14298 For additional information about GCC please refer to the [6]GCC project 14299 web site or contact the [7]GCC development mailing list. 14300 14301 To obtain GCC please use [8]our mirror sites, or our CVS server. 14302 __________________________________________________________________ 14303 14304 Previous 3.0.x Releases 14305 14306 December 20, 2001: GCC 3.0.3 has been released. 14307 October 25, 2001: GCC 3.0.2 has been released. 14308 August 20, 2001: GCC 3.0.1 has been released. 14309 June 18, 2001: GCC 3.0 has been released. 14310 14311 14312 For questions related to the use of GCC, please consult these web 14313 pages and the [9]GCC manuals. If that fails, the 14314 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14315 web pages and the development of GCC are welcome on our developer 14316 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 14317 archives. 14318 14319 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 14320 distribution of this entire article is permitted in any medium, 14321 provided this notice is preserved. 14322 14323 These pages are [14]maintained by the GCC team. Last modified 14324 2014-06-28[15]. 14325 14326 References 14327 14328 1. http://www.gnu.org/ 14329 2. http://gcc.gnu.org/gcc-3.0/features.html 14330 3. http://gcc.gnu.org/gcc-3.0/buildstat.html 14331 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14332 5. http://gcc.gnu.org/gcc-3.0/caveats.html 14333 6. http://gcc.gnu.org/index.html 14334 7. mailto:gcc (a] gcc.gnu.org 14335 8. http://gcc.gnu.org/mirrors.html 14336 9. https://gcc.gnu.org/onlinedocs/ 14337 10. mailto:gcc-help (a] gcc.gnu.org 14338 11. mailto:gcc (a] gcc.gnu.org 14339 12. https://gcc.gnu.org/lists.html 14340 13. http://www.fsf.org/ 14341 14. https://gcc.gnu.org/about.html 14342 15. http://validator.w3.org/check/referer 14343 ====================================================================== 14344 http://gcc.gnu.org/gcc-3.0/features.html 14345 14346 GCC 3.0 New Features 14347 14348 Additional changes in GCC 3.0.4 14349 14350 * GCC 3.0 now supports newer versions of the [1]NetBSD operating 14351 system, which use the ELF object file format, on x86 processors. 14352 * Correct debugging information is generated from functions that have 14353 lines from multiple files (e.g. yacc output). 14354 * A fix for whitespace handling in the -traditional preprocessor, 14355 which can affect Fortran. 14356 * Fixes to the exception handling runtime. 14357 * More fixes for bad code generation in C++. 14358 * A fix for shared library generation under AIX 4.3. 14359 * Documentation updates. 14360 * Port of GCC to Tensilica's Xtensa processor contributed. 14361 * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link). 14362 14363 Additional changes in GCC 3.0.3 14364 14365 * A fix to correct an accidental change to the PowerPC ABI. 14366 * Fixes for bad code generation on a variety of architectures. 14367 * Improvements to the debugging information generated for C++ 14368 classes. 14369 * Fixes for bad code generation in C++. 14370 * A fix to avoid crashes in the C++ demangler. 14371 * A fix to the C++ standard library to avoid buffer overflows. 14372 * Miscellaneous improvements for a variety of architectures. 14373 14374 Additional changes in GCC 3.0.2 14375 14376 * Fixes for bad code generation during loop unrolling. 14377 * Fixes for bad code generation by the sibling call optimization. 14378 * Minor improvements to x86 code generation. 14379 * Implementation of function descriptors in C++ vtables for IA64. 14380 * Numerous minor bug-fixes. 14381 14382 Additional changes in GCC 3.0.1 14383 14384 * C++ fixes for incorrect code-generation. 14385 * Improved cross-compiling support for the C++ standard library. 14386 * Fixes for some embedded targets that worked in GCC 2.95.3, but not 14387 in GCC 3.0. 14388 * Fixes for various exception-handling bugs. 14389 * A port to the S/390 architecture. 14390 14391 General Optimizer Improvements 14392 14393 * [2]Basic block reordering pass. 14394 * New if-conversion pass with support for conditional (predicated) 14395 execution. 14396 * New tail call and sibling call elimination optimizations. 14397 * New register renaming pass. 14398 * New (experimental) [3]static single assignment (SSA) representation 14399 support. 14400 * New dead-code elimination pass implemented using the SSA 14401 representation. 14402 * [4]Global null pointer test elimination. 14403 * [5]Global code hoisting/unification. 14404 * More builtins and optimizations for stdio.h, string.h and old BSD 14405 functions, as well as for ISO C99 functions. 14406 * New builtin __builtin_expect for giving hints to the branch 14407 predictor. 14408 14409 New Languages and Language specific improvements 14410 14411 * The GNU Compiler for the Java(TM) language (GCJ) is now integrated 14412 and supported, including the run-time library containing most 14413 common non-GUI Java classes, a bytecode interpreter, and the Boehm 14414 conservative garbage collector. Many bugs have been fixed. GCJ can 14415 compile Java source or Java bytecodes to either native code or Java 14416 class files, and supports native methods written in either the 14417 standard JNI or the more efficient and convenient CNI. 14418 * Here is a [6]partial list of C++ improvements, both new features 14419 and those no longer supported. 14420 * New C++ ABI. On the IA-64 platform GCC is capable of 14421 inter-operating with other IA-64 compilers. 14422 * The new ABI also significantly reduces the size of symbol and debug 14423 information. 14424 * New [7]C++ support library and many C++ bug fixes, vastly improving 14425 our conformance to the ISO C++ standard. 14426 * New [8]inliner for C++. 14427 * Rewritten C preprocessor, integrated into the C, C++ and Objective 14428 C compilers, with very many improvements including ISO C99 support 14429 and [9]improvements to dependency generation. 14430 * Support for more [10]ISO C99 features. 14431 * Many improvements to support for checking calls to format functions 14432 such as printf and scanf, including support for ISO C99 format 14433 features, extensions from the Single Unix Specification and GNU 14434 libc 2.2, checking of strfmon formats and features to assist in 14435 auditing for format string security bugs. 14436 * New warnings for C code that may have undefined semantics because 14437 of violations of sequence point rules in the C standard (such as a 14438 = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall. 14439 * Additional warning option -Wfloat-equal. 14440 * Improvements to -Wtraditional. 14441 * Fortran improvements are listed in [11]the Fortran documentation. 14442 14443 New Targets and Target Specific Improvements 14444 14445 * New x86 back end, generating much improved code. 14446 * Support for a generic i386-elf target contributed. 14447 * New option to emit x86 assembly code using Intel style syntax 14448 (-mintel-syntax). 14449 * HPUX 11 support contributed. 14450 * Improved PowerPC code generation, including scheduled prologue and 14451 epilogue. 14452 * Port of GCC to Intel's IA-64 processor contributed. 14453 * Port of GCC to Motorola's MCore 210 and 340 contributed. 14454 * New unified back-end for Arm, Thumb and StrongArm contributed. 14455 * Port of GCC to Intel's XScale processor contributed. 14456 * Port of GCC to Atmel's AVR microcontrollers contributed. 14457 * Port of GCC to Mitsubishi's D30V processor contributed. 14458 * Port of GCC to Matsushita's AM33 processor (a member of the MN10300 14459 processor family) contributed. 14460 * Port of GCC to Fujitsu's FR30 processor contributed. 14461 * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed. 14462 * Port of GCC to Sun's picoJava processor core contributed. 14463 14464 Documentation improvements 14465 14466 * Substantially rewritten and improved C preprocessor manual. 14467 * Many improvements to other documentation. 14468 * Manpages for gcc, cpp and gcov are now generated automatically from 14469 the master Texinfo manual, eliminating the problem of manpages 14470 being out of date. (The generated manpages are only extracts from 14471 the full manual, which is provided in Texinfo form, from which 14472 info, HTML, other formats and a printed manual can be generated.) 14473 * Generated info files are included in the release tarballs alongside 14474 their Texinfo sources, avoiding problems on some platforms with 14475 building makeinfo as part of the GCC distribution. 14476 14477 Other significant improvements 14478 14479 * Garbage collection used internally by the compiler for most memory 14480 allocation instead of obstacks. 14481 * Lengauer and Tarjan algorithm used for computing dominators in the 14482 CFG. This algorithm can be significantly faster and more space 14483 efficient than our older algorithm. 14484 * gccbug script provided to assist in submitting bug reports to our 14485 bug tracking system. (Bug reports previously submitted directly to 14486 our mailing lists, for which you received no bug tracking number, 14487 should be submitted again using gccbug if you can reproduce the 14488 problem with GCC 3.0.) 14489 * The internal libgcc library is [12]built as a shared library on 14490 systems that support it. 14491 * Extensive testsuite included with GCC, with many new tests. In 14492 addition to tests for GCC bugs that have been fixed, many tests 14493 have been added for language features, compiler warnings and 14494 builtin functions. 14495 * Additional language-independent warning options -Wpacked, -Wpadded, 14496 -Wunreachable-code and -Wdisabled-optimization. 14497 * Target-independent options -falign-functions, -falign-loops and 14498 -falign-jumps. 14499 14500 Plus a great many bug fixes and almost all the [13]features found in 14501 GCC 2.95. 14502 14503 14504 For questions related to the use of GCC, please consult these web 14505 pages and the [14]GCC manuals. If that fails, the 14506 [15]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14507 web pages and the development of GCC are welcome on our developer 14508 list at [16]gcc (a] gcc.gnu.org. All of [17]our lists have public 14509 archives. 14510 14511 Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and 14512 distribution of this entire article is permitted in any medium, 14513 provided this notice is preserved. 14514 14515 These pages are [19]maintained by the GCC team. Last modified 14516 2014-06-28[20]. 14517 14518 References 14519 14520 1. http://www.netbsd.org/ 14521 2. http://gcc.gnu.org/news/reorder.html 14522 3. http://gcc.gnu.org/news/ssa.html 14523 4. http://gcc.gnu.org/news/null.html 14524 5. http://gcc.gnu.org/news/unify.html 14525 6. http://gcc.gnu.org/gcc-3.0/c++features.html 14526 7. http://gcc.gnu.org/libstdc++/ 14527 8. http://gcc.gnu.org/news/inlining.html 14528 9. http://gcc.gnu.org/news/dependencies.html 14529 10. http://gcc.gnu.org/c99status.html 14530 11. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 14531 12. http://gcc.gnu.org/gcc-3.0/libgcc.html 14532 13. http://gcc.gnu.org/gcc-2.95/features.html 14533 14. https://gcc.gnu.org/onlinedocs/ 14534 15. mailto:gcc-help (a] gcc.gnu.org 14535 16. mailto:gcc (a] gcc.gnu.org 14536 17. https://gcc.gnu.org/lists.html 14537 18. http://www.fsf.org/ 14538 19. https://gcc.gnu.org/about.html 14539 20. http://validator.w3.org/check/referer 14540 ====================================================================== 14541 http://gcc.gnu.org/gcc-3.0/caveats.html 14542 14543 GCC 3.0 Caveats 14544 14545 * -fstrict-aliasing is now part of -O2 and higher optimization 14546 levels. This allows the compiler to assume the strictest aliasing 14547 rules applicable to the language being compiled. For C and C++, 14548 this activates optimizations based on the type of expressions. This 14549 optimization may thus break old, non-compliant code. 14550 * Enumerations are now properly promoted to int in function 14551 parameters and function returns. Normally this change is not 14552 visible, but when using -fshort-enums this is an ABI change. 14553 * The undocumented extension that allowed C programs to have a label 14554 at the end of a compound statement has been deprecated and may be 14555 removed in a future version. Programs that now generate a warning 14556 about this may be fixed by adding a null statement (a single 14557 semicolon) after the label. 14558 * The poorly documented extension that allowed string constants in C, 14559 C++ and Objective C to contain unescaped newlines has been 14560 deprecated and may be removed in a future version. Programs using 14561 this extension may be fixed in several ways: the bare newline may 14562 be replaced by \n, or preceded by \n\, or string concatenation may 14563 be used with the bare newline preceded by \n" and " placed at the 14564 start of the next line. 14565 * The Chill compiler is not included in GCC 3.0, because of the lack 14566 of a volunteer to convert it to use garbage collection. 14567 * Certain non-standard iostream methods from earlier versions of 14568 libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach, 14569 ostream::form, and istream::gets. 14570 * The new C++ ABI is not yet fully supported by current (as of 14571 2001-07-01) releases and development versions of GDB, or any 14572 earlier versions. There is a problem setting breakpoints by line 14573 number, and other related issues that have been fixed in GCC 3.0 14574 but not yet handled in GDB: 14575 [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 14576 14577 14578 For questions related to the use of GCC, please consult these web 14579 pages and the [2]GCC manuals. If that fails, the 14580 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14581 web pages and the development of GCC are welcome on our developer 14582 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 14583 14584 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 14585 distribution of this entire article is permitted in any medium, 14586 provided this notice is preserved. 14587 14588 These pages are [7]maintained by the GCC team. Last modified 14589 2014-06-27[8]. 14590 14591 References 14592 14593 1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 14594 2. https://gcc.gnu.org/onlinedocs/ 14595 3. mailto:gcc-help (a] gcc.gnu.org 14596 4. mailto:gcc (a] gcc.gnu.org 14597 5. https://gcc.gnu.org/lists.html 14598 6. http://www.fsf.org/ 14599 7. https://gcc.gnu.org/about.html 14600 8. http://validator.w3.org/check/referer 14601 ====================================================================== 14602 http://gcc.gnu.org/gcc-2.95/index.html 14603 14604 GCC 2.95 14605 14606 March 16, 2001: The GNU project and the GCC developers are pleased to 14607 announce the release of GCC version 2.95.3. 14608 14609 Release History 14610 14611 GCC 2.95.3 14612 March 16, 2001 14613 14614 GCC 2.95.2 14615 October 27, 1999 14616 14617 GCC 2.95.1 14618 August 19, 1999 14619 14620 GCC 2.95 14621 July 31, 1999. This is the first release of GCC since the April 14622 1999 GCC/EGCS reunification and includes nearly a year's worth 14623 of new development and bugfixes. 14624 14625 References and Acknowledgements 14626 14627 GCC used to stand for the GNU C Compiler, but since the compiler 14628 supports several other languages aside from C, it now stands for the 14629 GNU Compiler Collection. 14630 14631 The whole suite has been extensively [1]regression tested and 14632 [2]package tested. It should be reliable and suitable for widespread 14633 use. 14634 14635 The compiler has several new optimizations, new targets, new languages 14636 and other new features. See the [3]new features page for a more 14637 complete list of new features found in the GCC 2.95 releases. 14638 14639 The sources include installation instructions in both HTML and 14640 plaintext forms in the install directory in the distribution. However, 14641 the most up to date installation instructions and [4]build/test status 14642 are on the web pages. We will update those pages as new information 14643 becomes available. 14644 14645 The GCC developers would like to thank the numerous people that have 14646 contributed new features, test results, bugfixes, etc to GCC. This 14647 [5]amazing group of volunteers is what makes GCC successful. 14648 14649 And finally, we can't in good conscience fail to mention some 14650 [6]caveats to using GCC 2.95. 14651 14652 Download GCC 2.95 from one of our many [7]mirror sites. 14653 14654 For additional information about GCC please see the [8]GCC project web 14655 server or contact the [9]GCC development mailing list. 14656 14657 14658 For questions related to the use of GCC, please consult these web 14659 pages and the [10]GCC manuals. If that fails, the 14660 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14661 web pages and the development of GCC are welcome on our developer 14662 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 14663 archives. 14664 14665 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 14666 distribution of this entire article is permitted in any medium, 14667 provided this notice is preserved. 14668 14669 These pages are [15]maintained by the GCC team. Last modified 14670 2014-06-28[16]. 14671 14672 References 14673 14674 1. http://gcc.gnu.org/gcc-2.95/regress.html 14675 2. http://gcc.gnu.org/gcc-2.95/othertest.html 14676 3. http://gcc.gnu.org/gcc-2.95/features.html 14677 4. http://gcc.gnu.org/gcc-2.95/buildstat.html 14678 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14679 6. http://gcc.gnu.org/gcc-2.95/caveats.html 14680 7. http://gcc.gnu.org/mirrors.html 14681 8. http://gcc.gnu.org/index.html 14682 9. mailto:gcc (a] gcc.gnu.org 14683 10. https://gcc.gnu.org/onlinedocs/ 14684 11. mailto:gcc-help (a] gcc.gnu.org 14685 12. mailto:gcc (a] gcc.gnu.org 14686 13. https://gcc.gnu.org/lists.html 14687 14. http://www.fsf.org/ 14688 15. https://gcc.gnu.org/about.html 14689 16. http://validator.w3.org/check/referer 14690 ====================================================================== 14691 http://gcc.gnu.org/gcc-2.95/features.html 14692 14693 GCC 2.95 New Features 14694 14695 * General Optimizer Improvements: 14696 + [1]Localized register spilling to improve speed and code 14697 density especially on small register class machines. 14698 + [2]Global CSE using lazy code motion algorithms. 14699 + [3]Improved global constant/copy propagation. 14700 + [4]Improved control flow graph analysis and manipulation. 14701 + [5]Local dead store elimination. 14702 + [6]Memory Load hoisting/store sinking in loops. 14703 + [7]Type based alias analysis is enabled by default. Note this 14704 feature will expose bugs in the Linux kernel. Please refer to 14705 the FAQ (as shipped with GCC 2.95) for additional information 14706 on this issue. 14707 + Major revamp of GIV detection, combination and simplification 14708 to improve loop performance. 14709 + Major improvements to register allocation and reloading. 14710 * New Languages and Language specific improvements 14711 + [8]Many C++ improvements. 14712 + [9]Many Fortran improvements. 14713 + [10]Java front-end has been integrated. [11]runtime library is 14714 available separately. 14715 + [12]ISO C99 support 14716 + [13]Chill front-end and runtime has been integrated. 14717 + Boehm garbage collector support in libobjc. 14718 + More support for various pragmas which appear in vendor 14719 include files 14720 * New Targets and Target Specific Improvements 14721 + [14]SPARC backend rewrite. 14722 + -mschedule=8000 will optimize code for PA8000 class 14723 processors; -mpa-risc-2-0 will generate code for PA2.0 14724 processors 14725 + Various micro-optimizations for the ia32 port. K6 14726 optimizations 14727 + Compiler will attempt to align doubles in the stack on the 14728 ia32 port 14729 + Alpha EV6 support 14730 + PowerPC 750 14731 + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403. 14732 -mcpu=e603e was added to do -mcpu=603e and -msoft-float. 14733 + c3x, c4x 14734 + HyperSPARC 14735 + SparcLite86x 14736 + sh4 14737 + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix, 14738 arm-linux) 14739 + vxWorks targets include support for vxWorks threads 14740 + StrongARM 110 and ARM9 support added. ARM Scheduling 14741 parameters rewritten. 14742 + Various changes to the MIPS port to avoid assembler macros, 14743 which in turn improves performance 14744 + Various performance improvements to the i960 port. 14745 + Major rewrite of ns32k port 14746 * Other significant improvements 14747 + [15]Ability to dump cfg information and display it using vcg. 14748 + The new faster scheme for fixing vendor header files is 14749 enabled by default. 14750 + Experimental internationalization support. 14751 + multibyte character support 14752 + Some compile-time speedups for pathological problems 14753 + Better support for complex types 14754 * Plus the usual mountain of bugfixes 14755 * Core compiler is based on the gcc2 development tree from Sept 30, 14756 1998, so we have all of the [16]features found in GCC 2.8. 14757 14758 Additional Changes in GCC 2.95.1 14759 14760 * Generic bugfixes and improvements 14761 + Various documentation fixes related to the GCC/EGCS merger. 14762 + Fix memory management bug which could lead to spurious aborts, 14763 core dumps or random parsing errors in the compiler. 14764 + Fix a couple bugs in the dwarf1 and dwarf2 debug record 14765 support. 14766 + Fix infinite loop in the CSE optimizer. 14767 + Avoid undefined behavior in compiler FP emulation code 14768 + Fix install problem when prefix is overridden on the make 14769 install command. 14770 + Fix problem with unwanted installation of assert.h on some 14771 systems. 14772 + Fix problem with finding the wrong assembler in a single tree 14773 build. 14774 + Avoid increasing the known alignment of a register that is 14775 already known to be a pointer. 14776 * Platform specific bugfixes and improvements 14777 + Codegen bugfix for prologue/epilogue for cpu32 target. 14778 + Fix long long code generation bug for the Coldfire target. 14779 + Fix various aborts in the SH compiler. 14780 + Fix bugs in libgcc support library for the SH. 14781 + Fix alpha ev6 code generation bug. 14782 + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on 14783 AIX platforms. 14784 + Fix -fpic code generation bug for rs6000/ppc svr4 targets. 14785 + Fix varargs/stdarg code generation bug for rs6000/ppc svr4 14786 targets. 14787 + Fix weak symbol handling for rs6000/ppc svr4 targets. 14788 + Fix various problems with 64bit code generation for the 14789 rs6000/ppc port. 14790 + Fix codegen bug which caused tetex to be mis-compiled on the 14791 x86. 14792 + Fix compiler abort in new cfg code exposed by x86 port. 14793 + Fix out of range array reference in code convert flat 14794 registers to the x87 stacked FP register file. 14795 + Fix minor vxworks configuration bug. 14796 + Fix return type of bsearch for SunOS 4.x. 14797 * Language & Runtime specific fixes. 14798 + The G++ signature extension has been deprecated. It will be 14799 removed in the next major release of G++. Use of signatures 14800 will result in a warning from the compiler. 14801 + Several bugs relating to templates and namespaces were fixed. 14802 + A bug that caused crashes when combining templates with -g on 14803 DWARF1 platforms was fixed. 14804 + Pointers-to-members, virtual functions, and multiple 14805 inheritance should now work together correctly. 14806 + Some code-generation bugs relating to function try blocks were 14807 fixed. 14808 + G++ is a little bit more lenient with certain archaic 14809 constructs than in GCC 2.95. 14810 + Fix to prevent shared library version #s from bring truncated 14811 to 1 digit 14812 + Fix missing std:: in the libstdc++ library. 14813 + Fix stream locking problems in libio. 14814 + Fix problem in java compiler driver. 14815 14816 Additional Changes in GCC 2.95.2 14817 14818 The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While 14819 the optimizations performed by -fstrict-aliasing are valid according to 14820 the C and C++ standards, the optimization have caused some problems, 14821 particularly with old non-conforming code. 14822 14823 The GCC developers are experimenting with ways to warn users about code 14824 which violates the C/C++ standards, but those warnings are not ready 14825 for widespread use at this time. Rather than wait for those warnings 14826 the GCC developers have chosen to disable -fstrict-aliasing by default 14827 for the GCC 2.95.2 release. 14828 14829 We strongly encourage developers to find and fix code which violates 14830 the C/C++ standards as -fstrict-aliasing may be enabled by default in 14831 future releases. Use the option -fstrict-aliasing to re-enable these 14832 optimizations. 14833 * Generic bugfixes and improvements 14834 + Fix incorrectly optimized memory reference in global common 14835 subexpression elimination (GCSE) optimization pass. 14836 + Fix code generation bug in regmove.c in which it could 14837 incorrectly change a "const" value. 14838 + Fix bug in optimization of conditionals involving volatile 14839 memory references. 14840 + Avoid over-allocation of stack space for some procedures. 14841 + Fixed bug in the compiler which caused incorrect optimization 14842 of an obscure series of bit manipulations, shifts and 14843 arithmetic. 14844 + Fixed register allocator bug which caused teTeX to be 14845 mis-compiled on SPARC targets. 14846 + Avoid incorrect optimization of degenerate case statements for 14847 certain targets such as the ARM. 14848 + Fix out of range memory reference in the jump optimizer. 14849 + Avoid dereferencing null pointer in fix-header. 14850 + Fix test for GCC specific features so that it is possible to 14851 bootstrap with gcc-2.6.2 and older versions of GCC. 14852 + Fix typo in scheduler which could potentially cause out of 14853 range memory accesses. 14854 + Avoid incorrect loop reversal which caused incorrect code for 14855 certain loops on PowerPC targets. 14856 + Avoid incorrect optimization of switch statements on certain 14857 targets (for example the ARM). 14858 * Platform specific bugfixes and improvements 14859 + Work around bug in Sun V5.0 compilers which caused bootstrap 14860 comparison failures on SPARC targets. 14861 + Fix SPARC backend bug which caused aborts in final.c. 14862 + Fix sparc-hal-solaris2* configuration fragments. 14863 + Fix bug in sparc block profiling. 14864 + Fix obscure code generation bug for the PARISC targets. 14865 + Define __STDC_EXT__ for HPUX configurations. 14866 + Various POWERPC64 code generation bugfixes. 14867 + Fix abort for PPC targets using ELF (ex GNU/Linux). 14868 + Fix collect2 problems for AIX targets. 14869 + Correct handling of .file directive for PPC targets. 14870 + Fix bug in fix_trunc x86 patterns. 14871 + Fix x86 port to correctly pop the FP stack for functions that 14872 return structures in memory. 14873 + Fix minor bug in strlen x86 pattern. 14874 + Use stabs debugging instead of dwarf1 for x86-solaris targets. 14875 + Fix template repository code to handle leading underscore in 14876 mangled names. 14877 + Fix weak/weak alias support for OpenBSD. 14878 + GNU/Linux for the ARM has C++ compatible include files. 14879 * Language & Runtime specific fixes. 14880 + Fix handling of constructor attribute in the C front-end which 14881 caused problems building the Chill runtime library on some 14882 targets. 14883 + Fix minor problem merging type qualifiers in the C front-end. 14884 + Fix aliasing bug for pointers and references (C/C++). 14885 + Fix incorrect "non-constant initializer bug" when -traditional 14886 or -fwritable-strings is enabled. 14887 + Fix build error for Chill front-end on SunOS. 14888 + Do not complain about duplicate instantiations when using 14889 -frepo (C++). 14890 + Fix array bounds handling in C++ front-end which caused 14891 problems with dwarf debugging information in some 14892 circumstances. 14893 + Fix minor namespace problem. 14894 + Fix problem linking java programs. 14895 14896 Additional Changes in GCC 2.95.3 14897 14898 * Generic bugfixes and improvements 14899 + Fix numerous problems that caused incorrect optimization in 14900 the register reloading code. 14901 + Fix numerous problems that caused incorrect optimization in 14902 the loop optimizer. 14903 + Fix aborts in the functions build_insn_chain and scan_loops 14904 under some circumstances. 14905 + Fix an alias analysis bug. 14906 + Fix an infinite compilation bug in the combiner. 14907 + A few problems with complex number support have been fixed. 14908 + It is no longer possible for gcc to act as a fork bomb when 14909 installed incorrectly. 14910 + The -fpack-struct option should be recognized now. 14911 + Fixed a bug that caused incorrect code to be generated due to 14912 a lost stack adjustment. 14913 * Platform specific bugfixes and improvements 14914 + Support building ARM toolchains hosted on Windows. 14915 + Fix attribute calculations in ARM toolchains. 14916 + arm-linux support has been improved. 14917 + Fix a PIC failure on sparc targets. 14918 + On ix86 targets, the regparm attribute should now work 14919 reliably. 14920 + Several updates for the h8300 port. 14921 + Fix problem building libio with glibc 2.2. 14922 14923 14924 For questions related to the use of GCC, please consult these web 14925 pages and the [17]GCC manuals. If that fails, the 14926 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14927 web pages and the development of GCC are welcome on our developer 14928 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 14929 archives. 14930 14931 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 14932 distribution of this entire article is permitted in any medium, 14933 provided this notice is preserved. 14934 14935 These pages are [22]maintained by the GCC team. Last modified 14936 2014-06-28[23]. 14937 14938 References 14939 14940 1. http://gcc.gnu.org/news/spill.html 14941 2. http://gcc.gnu.org/news/lcm.html 14942 3. http://gcc.gnu.org/news/cprop.html 14943 4. http://gcc.gnu.org/news/cfg.html 14944 5. http://gcc.gnu.org/news/dse.html 14945 6. http://gcc.gnu.org/news/hoist.html 14946 7. http://gcc.gnu.org/news/alias.html 14947 8. http://gcc.gnu.org/gcc-2.95/c++features.html 14948 9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 14949 10. http://gcc.gnu.org/java/gcj-announce.txt 14950 11. http://gcc.gnu.org/news/javaannounce.html 14951 12. http://gcc.gnu.org/c99status.html 14952 13. http://gcc.gnu.org/news/chill.html 14953 14. http://gcc.gnu.org/news/sparc.html 14954 15. http://gcc.gnu.org/news/egcs-vcg.html 14955 16. http://gcc.gnu.org/egcs-1.0/features-2.8.html 14956 17. https://gcc.gnu.org/onlinedocs/ 14957 18. mailto:gcc-help (a] gcc.gnu.org 14958 19. mailto:gcc (a] gcc.gnu.org 14959 20. https://gcc.gnu.org/lists.html 14960 21. http://www.fsf.org/ 14961 22. https://gcc.gnu.org/about.html 14962 23. http://validator.w3.org/check/referer 14963 ====================================================================== 14964 http://gcc.gnu.org/gcc-2.95/caveats.html 14965 14966 GCC 2.95 Caveats 14967 14968 * GCC 2.95 will issue an error for invalid asm statements that had 14969 been silently accepted by earlier versions of the compiler. This is 14970 particularly noticeable when compiling older versions of the Linux 14971 kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95) 14972 for more information on this issue. 14973 * GCC 2.95 implements type based alias analysis to disambiguate 14974 memory references. Some programs, particularly the Linux kernel 14975 violate ANSI/ISO aliasing rules and therefore may not operate 14976 correctly when compiled with GCC 2.95. Please refer to the FAQ (as 14977 shipped with GCC 2.95) for more information on this issue. 14978 * GCC 2.95 has a known bug in its handling of complex variables for 14979 64bit targets. Instead of silently generating incorrect code, GCC 14980 2.95 will issue a fatal error for situations it can not handle. 14981 This primarily affects the Fortran community as Fortran makes more 14982 use of complex variables than C or C++. 14983 * GCC 2.95 has an integrated libstdc++, but does not have an 14984 integrated libg++. Furthermore old libg++ releases will not work 14985 with GCC 2.95. You can retrieve a recent copy of libg++ from the 14986 [1]GCC ftp server. 14987 Note most C++ programs only need libstdc++. 14988 * Exception handling may not work with shared libraries, particularly 14989 on alphas, hppas, rs6000/powerpc and mips based platforms. 14990 Exception handling is known to work on x86 GNU/Linux platforms with 14991 shared libraries. 14992 * In general, GCC 2.95 is more rigorous about rejecting invalid C++ 14993 code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0, 14994 or EGCS 1.1. As a result it may be necessary to fix C++ code before 14995 it will compile with GCC 2.95. 14996 * G++ is also converting toward the ISO C++ standard; as a result 14997 code which was previously valid (and thus accepted by other 14998 compilers and older versions of g++) may no longer be accepted. The 14999 flag -fpermissive may allow some non-conforming code to compile 15000 with GCC 2.95. 15001 * GCC 2.95 compiled C++ code is not binary compatible with EGCS 15002 1.1.x, EGCS 1.0.x or GCC 2.8.x. 15003 * GCC 2.95 does not have changes from the GCC 2.8 tree that were made 15004 between Sept 30, 1998 and April 30, 1999 (the official end of the 15005 GCC 2.8 project). Future GCC releases will include all the changes 15006 from the defunct GCC 2.8 sources. 15007 15008 15009 For questions related to the use of GCC, please consult these web 15010 pages and the [2]GCC manuals. If that fails, the 15011 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15012 web pages and the development of GCC are welcome on our developer 15013 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 15014 15015 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 15016 distribution of this entire article is permitted in any medium, 15017 provided this notice is preserved. 15018 15019 These pages are [7]maintained by the GCC team. Last modified 15020 2014-06-12[8]. 15021 15022 References 15023 15024 1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz 15025 2. https://gcc.gnu.org/onlinedocs/ 15026 3. mailto:gcc-help (a] gcc.gnu.org 15027 4. mailto:gcc (a] gcc.gnu.org 15028 5. https://gcc.gnu.org/lists.html 15029 6. http://www.fsf.org/ 15030 7. https://gcc.gnu.org/about.html 15031 8. http://validator.w3.org/check/referer 15032 ====================================================================== 15033 http://gcc.gnu.org/egcs-1.1/index.html 15034 15035 EGCS 1.1 15036 15037 September 3, 1998: We are pleased to announce the release of EGCS 1.1. 15038 December 1, 1998: We are pleased to announce the release of EGCS 1.1.1. 15039 March 15, 1999: We are pleased to announce the release of EGCS 1.1.2. 15040 15041 EGCS is a free software project to further the development of the GNU 15042 compilers using an open development environment. 15043 15044 EGCS 1.1 is a major new release of the EGCS compiler system. It has 15045 been [1]extensively tested and is believed to be stable and suitable 15046 for widespread use. 15047 15048 EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8 15049 development sources; it contains all of the new features found in GCC 15050 2.8.1 as well as all new development from GCC up to June 6, 1998. 15051 15052 EGCS 1.1 also contains many improvements and features not found in GCC 15053 or in older versions of EGCS: 15054 * Global common subexpression elimination and global constant/copy 15055 propagation (aka [2]gcse) 15056 * Ongoing improvements to the [3]alias analysis support to allow for 15057 better optimizations throughout the compiler. 15058 * Vastly improved [4]C++ compiler and integrated C++ runtime 15059 libraries. 15060 * Fixes for the /tmp symlink race security problems. 15061 * New targets including mips16, arm-thumb and 64 bit PowerPC. 15062 * Improvements to GNU Fortran (g77) compiler and runtime library made 15063 since g77 version 0.5.23. 15064 15065 See the [5]new features page for a more complete list of new features 15066 found in EGCS 1.1 releases. 15067 15068 EGCS 1.1.1 is a minor update to fix several serious problems in EGCS 15069 1.1: 15070 * General improvements and fixes 15071 + Avoid some stack overflows when compiling large functions. 15072 + Avoid incorrect loop invariant code motions. 15073 + Fix some core dumps on Linux kernel code. 15074 + Bring back the imake -Di386 and friends fix from EGCS 1.0.2. 15075 + Fix code generation problem in gcse. 15076 + Various documentation related fixes. 15077 * g++/libstdc++ improvements and fixes 15078 + MT safe EH fix for setjmp/longjmp based exception handling. 15079 + Fix a few bad interactions between optimization and exception 15080 handling. 15081 + Fixes for demangling of template names starting with "__". 15082 + Fix a bug that would fail to run destructors in some cases 15083 with -O2. 15084 + Fix 'new' of classes with virtual bases. 15085 + Fix crash building Qt on the Alpha. 15086 + Fix failure compiling WIFEXITED macro on GNU/Linux. 15087 + Fix some -frepo failures. 15088 * g77 and libf2c improvements and fixes 15089 + Various documentation fixes. 15090 + Avoid compiler crash on RAND intrinsic. 15091 + Fix minor bugs in makefiles exposed by BSD make programs. 15092 + Define _XOPEN_SOURCE for libI77 build to avoid potential 15093 problems on some 64-bit systems. 15094 + Fix problem with implicit endfile on rewind. 15095 + Fix spurious recursive I/O errors. 15096 * platform specific improvements and fixes 15097 + Match all versions of UnixWare7. 15098 + Do not assume x86 SVR4 or UnixWare targets can handle stabs. 15099 + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion 15100 from unsigned ints to double precision floats. 15101 + Fix ARM ABI issue with NetBSD. 15102 + Fix a few arm code generation bugs. 15103 + Fixincludes will fix additional broken SCO OpenServer header 15104 files. 15105 + Fix a m68k backend bug which caused invalid offsets in reg+d 15106 addresses. 15107 + Fix problems with 64bit AIX 4.3 support. 15108 + Fix handling of long longs for varargs/stdarg functions on the 15109 ppc. 15110 + Minor fixes to CPP predefines for Windows. 15111 + Fix code generation problems with gpr<->fpr copies for 64bit 15112 ppc. 15113 + Fix a few coldfire code generation bugs. 15114 + Fix some more header file problems on SunOS 4.x. 15115 + Fix assert.h handling for RTEMS. 15116 + Fix Windows handling of TREE_SYMBOL_REFERENCED. 15117 + Fix x86 compiler abort in reg-stack pass. 15118 + Fix cygwin/windows problem with section attributes. 15119 + Fix Alpha code generation problem exposed by SMP Linux 15120 kernels. 15121 + Fix typo in m68k 32->64bit integer conversion. 15122 + Make sure target libraries build with -fPIC for PPC & Alpha 15123 targets. 15124 15125 EGCS 1.1.2 is a minor update to fix several serious problems in EGCS 15126 1.1.1: 15127 * General improvements and fixes 15128 + Fix bug in loop optimizer which caused the SPARC (and 15129 potentially other) ports to segfault. 15130 + Fix infinite recursion in alias analysis and combiner code. 15131 + Fix bug in regclass preferencing. 15132 + Fix incorrect loop reversal which caused incorrect code to be 15133 generated for several targets. 15134 + Fix return value for builtin memcpy. 15135 + Reduce compile time for certain loops which exposed quadratic 15136 behavior in the loop optimizer. 15137 + Fix bug which caused volatile memory to be written multiple 15138 times when only one write was needed/desired. 15139 + Fix compiler abort in caller-save.c 15140 + Fix combiner bug which caused incorrect code generation for 15141 certain division by constant operations. 15142 + Fix incorrect code generation due to a bug in range check 15143 optimizations. 15144 + Fix incorrect code generation due to mis-handling of clobbered 15145 values in CSE. 15146 + Fix compiler abort/segfault due to incorrect register 15147 splitting when unrolling loops. 15148 + Fix code generation involving autoincremented addresses with 15149 ternary operators. 15150 + Work around bug in the scheduler which caused qt to be 15151 mis-compiled on some platforms. 15152 + Fix code generation problems with -fshort-enums. 15153 + Tighten security for temporary files. 15154 + Improve compile time for codes which make heavy use of 15155 overloaded functions. 15156 + Fix multiply defined constructor/destructor symbol problems. 15157 + Avoid setting bogus RPATH environment variable during 15158 bootstrap. 15159 + Avoid GNU-make dependencies in the texinfo subdir. 15160 + Install CPP wrapper script in $(prefix)/bin if --enable-cpp. 15161 --enable-cpp=<dirname> can be used to specify an additional 15162 install directory for the cpp wrapper script. 15163 + Fix CSE bug which caused incorrect label-label refs to appear 15164 on some platforms. 15165 + Avoid linking in EH routines from libgcc if they are not 15166 needed. 15167 + Avoid obscure bug in aliasing code. 15168 + Fix bug in weak symbol handling. 15169 * Platform-specific improvements and fixes 15170 + Fix detection of PPro/PII on Unixware 7. 15171 + Fix compiler segfault when building spec99 and other programs 15172 for SPARC targets. 15173 + Fix code-generation bugs for integer and floating point 15174 conditional move instructions on the PPro/PII. 15175 + Use fixincludes to fix byteorder problems on i?86-*-sysv. 15176 + Fix build failure for the arc port. 15177 + Fix floating point format configuration for i?86-gnu port. 15178 + Fix problems with hppa1.0-hp-hpux10.20 configuration when 15179 threads are enabled. 15180 + Fix coldfire code generation bugs. 15181 + Fix "unrecognized insn" problems for Alpha and PPC ports. 15182 + Fix h8/300 code generation problem with floating point values 15183 in memory. 15184 + Fix unrecognized insn problems for the m68k port. 15185 + Fix namespace-pollution problem for the x86 port. 15186 + Fix problems with old assembler on x86 NeXT systems. 15187 + Fix PIC code-generation problems for the SPARC port. 15188 + Fix minor bug with LONG_CALLS in PowerPC SVR4 support. 15189 + Fix minor ISO namespace violation in Alpha varargs/stdarg 15190 support. 15191 + Fix incorrect "braf" instruction usage for the SH port. 15192 + Fix minor bug in va-sh which prevented its use with -ansi. 15193 + Fix problems recognizing and supporting FreeBSD. 15194 + Handle OpenBSD systems correctly. 15195 + Minor fixincludes fix for Digital UNIX 4.0B. 15196 + Fix problems with ctors/dtors in SCO shared libraries. 15197 + Abort instead of generating incorrect code for PPro/PII 15198 floating point conditional moves. 15199 + Avoid multiply defined symbols on GNU/Linux systems using 15200 libc-5.4.xx. 15201 + Fix abort in alpha compiler. 15202 * Fortran-specific fixes 15203 + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year 15204 is in the documented, non-Y2K-compliant range of 0-99, instead 15205 of being returned as 100 in the year 2000. 15206 + Fix the `Date_and_Time' intrinsic (in libg2c) to return the 15207 milliseconds value properly in Values(8). 15208 + Fix the `LStat' intrinsic (in libg2c) to return device-ID 15209 information properly in SArray(7). 15210 15211 Each release includes installation instructions in both HTML and 15212 plaintext forms (see the INSTALL directory in the toplevel directory of 15213 the distribution). However, we also keep the most up to date 15214 installation instructions and [6]build/test status on our web page. We 15215 will update those pages as new information becomes available. 15216 15217 The EGCS project would like to thank the numerous people that have 15218 contributed new features, test results, bugfixes, etc. This [7]amazing 15219 group of volunteers is what makes EGCS successful. 15220 15221 And finally, we can't in good conscience fail to mention some 15222 [8]caveats to using EGCS 1.1. 15223 15224 Download EGCS from egcs.cygnus.com (USA California). 15225 15226 The EGCS 1.1 release is also available on many mirror sites. 15227 [9]Goto mirror list to find a closer site. 15228 15229 15230 For questions related to the use of GCC, please consult these web 15231 pages and the [10]GCC manuals. If that fails, the 15232 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15233 web pages and the development of GCC are welcome on our developer 15234 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 15235 archives. 15236 15237 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 15238 distribution of this entire article is permitted in any medium, 15239 provided this notice is preserved. 15240 15241 These pages are [15]maintained by the GCC team. Last modified 15242 2014-06-28[16]. 15243 15244 References 15245 15246 1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html 15247 2. http://gcc.gnu.org/news/gcse.html 15248 3. http://gcc.gnu.org/news/alias.html 15249 4. http://gcc.gnu.org/egcs-1.1/c++features.html 15250 5. http://gcc.gnu.org/egcs-1.1/features.html 15251 6. http://gcc.gnu.org/egcs-1.1/buildstat.html 15252 7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 15253 8. http://gcc.gnu.org/egcs-1.1/caveats.html 15254 9. http://gcc.gnu.org/mirrors.html 15255 10. https://gcc.gnu.org/onlinedocs/ 15256 11. mailto:gcc-help (a] gcc.gnu.org 15257 12. mailto:gcc (a] gcc.gnu.org 15258 13. https://gcc.gnu.org/lists.html 15259 14. http://www.fsf.org/ 15260 15. https://gcc.gnu.org/about.html 15261 16. http://validator.w3.org/check/referer 15262 ====================================================================== 15263 http://gcc.gnu.org/egcs-1.1/features.html 15264 15265 EGCS 1.1 new features 15266 15267 * Integrated GNU Fortran (g77) compiler and runtime library with 15268 improvements, based on g77 version 0.5.23. 15269 * Vast improvements in the C++ compiler; so many they have [1]page of 15270 their own! 15271 * Compiler implements [2]global common subexpression elimination and 15272 global copy/constant propagation. 15273 * More major improvements in the [3]alias analysis code. 15274 * More major improvements in the exception handling code to improve 15275 performance, lower static overhead and provide the infrastructure 15276 for future improvements. 15277 * The infamous /tmp symlink race security problems have been fixed. 15278 * The regmove optimization pass has been nearly completely rewritten 15279 to improve performance of generated code. 15280 * The compiler now recomputes register usage information before local 15281 register allocation. By providing more accurate information to the 15282 priority based allocator, we get better register allocation. 15283 * The register reloading phase of the compiler optimizes spill code 15284 much better than in previous releases. 15285 * Some bad interactions between the register allocator and 15286 instruction scheduler have been fixed, resulting in much better 15287 code for certain programs. Additionally, we have tuned the 15288 scheduler in various ways to improve performance of generated code 15289 for some architectures. 15290 * The compiler's branch shortening algorithms have been significantly 15291 improved to work better on targets which align jump targets. 15292 * The compiler now supports -Os to prefer optimizing for code space 15293 over optimizing for code speed. 15294 * The compiler will now totally eliminate library calls which compute 15295 constant values. This primarily helps targets with no integer 15296 div/mul support and targets without floating point support. 15297 * The compiler now supports an extensive "--help" option. 15298 * cpplib has been greatly improved and may be suitable for limited 15299 use. 15300 * Memory footprint for the compiler has been significantly reduced 15301 for some pathological cases. 15302 * The time to build EGCS has been improved for certain targets 15303 (particularly the alpha and mips platforms). 15304 * Many infrastructure improvements throughout the compiler, plus the 15305 usual mountain of bugfixes and minor improvements. 15306 * Target dependent improvements: 15307 + SPARC port now includes V8 plus and V9 support as well as 15308 performance tuning for Ultra class machines. The SPARC port 15309 now uses the Haifa scheduler. 15310 + Alpha port has been tuned for the EV6 processor and has an 15311 optimized expansion of memcpy/bzero. The Alpha port now uses 15312 the Haifa scheduler. 15313 + RS6000/PowerPC: support for the Power64 architecture and AIX 15314 4.3. The RS6000/PowerPC port now uses the Haifa scheduler. 15315 + x86: Alignment of static store data and jump targets is per 15316 Intel recommendations now. Various improvements throughout the 15317 x86 port to improve performance on Pentium processors 15318 (including improved epilogue sequences for Pentium chips and 15319 backend improvements which should help register allocation on 15320 all x86 variants. Conditional move support has been fixed and 15321 enabled for PPro processors. The x86 port also better supports 15322 64bit operations now. Unixware 7, a System V Release 5 target, 15323 is now supported and SCO OpenServer targets can support GAS. 15324 + MIPS has improved multiply/multiply-add support and now 15325 includes mips16 ISA support. 15326 + M68k has many micro-optimizations and Coldfire fixes. 15327 * Core compiler is based on the GCC development tree from June 9, 15328 1998, so we have all of the [4]features found in GCC 2.8. 15329 15330 15331 For questions related to the use of GCC, please consult these web 15332 pages and the [5]GCC manuals. If that fails, the 15333 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15334 web pages and the development of GCC are welcome on our developer 15335 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 15336 15337 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 15338 distribution of this entire article is permitted in any medium, 15339 provided this notice is preserved. 15340 15341 These pages are [10]maintained by the GCC team. Last modified 15342 2014-06-12[11]. 15343 15344 References 15345 15346 1. http://gcc.gnu.org/egcs-1.1/c++features.html 15347 2. http://gcc.gnu.org/news/gcse.html 15348 3. http://gcc.gnu.org/news/alias.html 15349 4. http://gcc.gnu.org/egcs-1.0/features-2.8.html 15350 5. https://gcc.gnu.org/onlinedocs/ 15351 6. mailto:gcc-help (a] gcc.gnu.org 15352 7. mailto:gcc (a] gcc.gnu.org 15353 8. https://gcc.gnu.org/lists.html 15354 9. http://www.fsf.org/ 15355 10. https://gcc.gnu.org/about.html 15356 11. http://validator.w3.org/check/referer 15357 ====================================================================== 15358 http://gcc.gnu.org/egcs-1.1/caveats.html 15359 15360 EGCS 1.1 Caveats 15361 15362 * EGCS has an integrated libstdc++, but does not have an integrated 15363 libg++. Furthermore old libg++ releases will not work with EGCS; HJ 15364 Lu has made a libg++-2.8.1.2 snapshot available which may work with 15365 EGCS. 15366 Note most C++ programs only need libstdc++. 15367 * Exception handling may not work with shared libraries, particularly 15368 on alphas, hppas, rs6000/powerpc and mips based platforms. 15369 Exception handling is known to work on x86-linux platforms with 15370 shared libraries. 15371 * Some versions of the Linux kernel have bugs which prevent them from 15372 being compiled or from running when compiled by EGCS. See the FAQ 15373 (as shipped with EGCS 1.1) for additional information. 15374 * In general, EGCS is more rigorous about rejecting invalid C++ code 15375 or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As 15376 a result it may be necessary to fix C++ code before it will compile 15377 with EGCS. 15378 * G++ is also converting toward the ISO C++ standard; as a result 15379 code which was previously valid (and thus accepted by other 15380 compilers and older versions of g++) may no longer be accepted. 15381 * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x 15382 or GCC 2.8.x due to changes necessary to support thread safe 15383 exception handling. 15384 15385 15386 For questions related to the use of GCC, please consult these web 15387 pages and the [1]GCC manuals. If that fails, the 15388 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15389 web pages and the development of GCC are welcome on our developer 15390 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 15391 15392 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 15393 distribution of this entire article is permitted in any medium, 15394 provided this notice is preserved. 15395 15396 These pages are [6]maintained by the GCC team. Last modified 15397 2014-06-12[7]. 15398 15399 References 15400 15401 1. https://gcc.gnu.org/onlinedocs/ 15402 2. mailto:gcc-help (a] gcc.gnu.org 15403 3. mailto:gcc (a] gcc.gnu.org 15404 4. https://gcc.gnu.org/lists.html 15405 5. http://www.fsf.org/ 15406 6. https://gcc.gnu.org/about.html 15407 7. http://validator.w3.org/check/referer 15408 ====================================================================== 15409 http://gcc.gnu.org/egcs-1.0/index.html 15410 15411 EGCS 1.0 15412 15413 December 3, 1997: We are pleased to announce the release of EGCS 1.0. 15414 January 6, 1998: We are pleased to announce the release of EGCS 1.0.1. 15415 March 16, 1998: We are pleased to announce the release of EGCS 1.0.2. 15416 May 15, 1998 We are pleased to announce the release of EGCS 1.0.3. 15417 15418 EGCS is a collaborative effort involving several groups of hackers 15419 using an open development model to accelerate development and testing 15420 of GNU compilers and runtime libraries. 15421 15422 An important goal of EGCS is to allow wide scale testing of 15423 experimental features and optimizations; therefore, EGCS contains some 15424 features and optimizations which are still under development. However, 15425 EGCS has been carefully tested and should be comparable in quality to 15426 most GCC releases. 15427 15428 EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8 15429 development sources; it contains nearly all of the new features found 15430 in GCC 2.8. 15431 15432 EGCS 1.0 also contains many improvements and features not found in GCC 15433 2.7 and even the GCC 2.8 series (which was released after the original 15434 EGCS 1.0 release). 15435 * Integrated C++ runtime libraries, including support for most major 15436 GNU/Linux systems! 15437 * The integrated libstdc++ library includes a verbatim copy of SGI's 15438 STL release. 15439 * Integrated GNU Fortran compiler. 15440 * New instruction scheduler. 15441 * New alias analysis code. 15442 15443 See the [1]new features page for a more complete list of new features. 15444 15445 EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few 15446 critical bugs and add support for Red Hat 5.0 Linux. Changes since the 15447 EGCS 1.0 release: 15448 * Add support for Red Hat 5.0 Linux and better support for Linux 15449 systems using glibc2. 15450 Many programs failed to link when compiled with EGCS 1.0 on Red Hat 15451 5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should 15452 fix these problems. 15453 * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception 15454 handling interfaces. 15455 To avoid future compatibility problems, we strongly urge anyone who 15456 is planning on distributing shared libraries that contain C++ code 15457 to upgrade to EGCS 1.0.1 first. 15458 Soon after EGCS 1.0 was released, the GCC developers made some 15459 incompatible changes in libgcc's exception handling interfaces. 15460 These changes were needed to solve problems on some platforms. This 15461 means that GCC 2.8.0, when released, will not be seamlessly 15462 compatible with shared libraries built by EGCS 1.0. The reason is 15463 that the libgcc.a in GCC 2.8.0 will not contain a function needed 15464 by the old interface. 15465 The result of this is that there may be compatibility problems with 15466 shared libraries built by EGCS 1.0 when used with GCC 2.8.0. 15467 With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface, 15468 and libgcc.a has the support routines for both the old and the new 15469 interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed, 15470 and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed). 15471 The maintainers of GCC 2.x have decided against including seamless 15472 support for the old interface in 2.8.0, since it was never 15473 "official", so to avoid future compatibility problems we recommend 15474 against distributing any shared libraries built by EGCS 1.0 that 15475 contain C++ code (upgrade to 1.0.1 and use that). 15476 * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends. 15477 The x86 changes fix code generation errors exposed when building 15478 glibc2 and the usual GNU/Linux dynamic linker (ld.so). 15479 The hppa change fixes a compiler abort when configured for use with 15480 RTEMS. 15481 The MIPS changes fix problems with the definition of LONG_MAX on 15482 newer systems, allow for command line selection of the target ABI, 15483 and fix one code generation problem. 15484 The rs6000/ppc change fixes some problems with passing structures 15485 to varargs/stdarg functions. 15486 * A few machine independent bugfixes, mostly to fix code generation 15487 errors when building Linux kernels or glibc. 15488 * Fix a few critical exception handling and template bugs in the C++ 15489 compiler. 15490 * Fix Fortran namelist bug on alphas. 15491 * Fix build problems on x86-solaris systems. 15492 15493 EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several 15494 serious problems in EGCS 1.0.1. 15495 * General improvements and fixes 15496 + Memory consumption significantly reduced, especially for 15497 templates and inline functions. 15498 + Fix various problems with glibc2.1. 15499 + Fix loop optimization bug exposed by rs6000/ppc port. 15500 + Fix to avoid potential code generation problems in jump.c. 15501 + Fix some undefined symbol problems in dwarf1 debug support. 15502 * g++/libstdc++ improvements and fixes 15503 + libstdc++ in the EGCS release has been updated and should be 15504 link compatible with libstdc++-2.8. 15505 + Various fixes in libio/libstdc++ to work better on GNU/Linux 15506 systems. 15507 + Fix problems with duplicate symbols on systems that do not 15508 support weak symbols. 15509 + Memory corruption bug and undefined symbols in bastring have 15510 been fixed. 15511 + Various exception handling fixes. 15512 + Fix compiler abort for very long thunk names. 15513 * g77 improvements and fixes 15514 + Fix compiler crash for omitted bound in Fortran CASE 15515 statement. 15516 + Add missing entries to g77 lang-options. 15517 + Fix problem with -fpedantic in the g77 compiler. 15518 + Fix "backspace" problem with g77 on alphas. 15519 + Fix x86 backend problem with Fortran literals and -fpic. 15520 + Fix some of the problems with negative subscripts for g77 on 15521 alphas. 15522 + Fixes for Fortran builds on cygwin32/mingw32. 15523 * platform specific improvements and fixes 15524 + Fix long double problems on x86 (exposed by glibc). 15525 + x86 ports define i386 again to keep imake happy. 15526 + Fix exception handling support on NetBSD ports. 15527 + Several changes to collect2 to fix many problems with AIX. 15528 + Define __ELF__ for GNU/Linux on rs6000. 15529 + Fix -mcall-linux problem on GNU/Linux on rs6000. 15530 + Fix stdarg/vararg problem for GNU/Linux on rs6000. 15531 + Allow autoconf to select a proper install problem on AIX 3.1. 15532 + m68k port support includes -mcpu32 option as well as cpu32 15533 multilibs. 15534 + Fix stdarg bug for irix6. 15535 + Allow EGCS to build on irix5 without the gnu assembler. 15536 + Fix problem with static linking on sco5. 15537 + Fix bootstrap on sco5 with native compiler. 15538 + Fix for abort building newlib on H8 target. 15539 + Fix fixincludes handling of math.h on SunOS. 15540 + Minor fix for Motorola 3300 m68k systems. 15541 15542 EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few 15543 problems reported by Red Hat for builds of Red Hat 5.1. 15544 * Generic bugfixes: 15545 + Fix a typo in the libio library which resulted in incorrect 15546 behavior of istream::get. 15547 + Fix the Fortran negative array index problem. 15548 + Fix a major problem with the ObjC runtime thread support 15549 exposed by glibc2. 15550 + Reduce memory consumption of the Haifa scheduler. 15551 * Target specific bugfixes: 15552 + Fix one x86 floating point code generation bug exposed by 15553 glibc2 builds. 15554 + Fix one x86 internal compiler error exposed by glibc2 builds. 15555 + Fix profiling bugs on the Alpha. 15556 + Fix ImageMagick & emacs 20.2 build problems on the Alpha. 15557 + Fix rs6000/ppc bug when converting values from integer types 15558 to floating point types. 15559 15560 The EGCS 1.0 releases include installation instructions in both HTML 15561 and plaintext forms (see the INSTALL directory in the toplevel 15562 directory of the distribution). However, we also keep the most up to 15563 date installation instructions and [2]build/test status on our web 15564 page. We will update those pages as new information becomes available. 15565 15566 And, we can't in good conscience fail to mention some [3]caveats to 15567 using EGCS. 15568 15569 Update: Big thanks to Stanford for providing a high speed link for 15570 downloading EGCS (go.cygnus.com)! 15571 15572 Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com 15573 (USA California -- High speed link provided by Stanford). 15574 15575 The EGCS 1.0 release is also available many mirror sites. 15576 [4]Goto mirror list to find a closer site 15577 15578 We'd like to thank the numerous people that have contributed new 15579 features, test results, bugfixes, etc. Unfortunately, they're far too 15580 numerous to mention by name. 15581 15582 15583 For questions related to the use of GCC, please consult these web 15584 pages and the [5]GCC manuals. If that fails, the 15585 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15586 web pages and the development of GCC are welcome on our developer 15587 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 15588 15589 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 15590 distribution of this entire article is permitted in any medium, 15591 provided this notice is preserved. 15592 15593 These pages are [10]maintained by the GCC team. Last modified 15594 2014-06-28[11]. 15595 15596 References 15597 15598 1. http://gcc.gnu.org/egcs-1.0/features.html 15599 2. http://gcc.gnu.org/egcs-1.0/buildstat.html 15600 3. http://gcc.gnu.org/egcs-1.0/caveats.html 15601 4. http://gcc.gnu.org/mirrors.html 15602 5. https://gcc.gnu.org/onlinedocs/ 15603 6. mailto:gcc-help (a] gcc.gnu.org 15604 7. mailto:gcc (a] gcc.gnu.org 15605 8. https://gcc.gnu.org/lists.html 15606 9. http://www.fsf.org/ 15607 10. https://gcc.gnu.org/about.html 15608 11. http://validator.w3.org/check/referer 15609 ====================================================================== 15610 http://gcc.gnu.org/egcs-1.0/features.html 15611 15612 EGCS 1.0 features 15613 15614 * Core compiler is based on the gcc2 development tree from Aug 2, 15615 1997, so we have most of the [1]features found in GCC 2.8. 15616 * Integrated GNU Fortran compiler based on g77-0.5.22-19970929. 15617 * Vast improvements in the C++ compiler; so many they have [2]page of 15618 their own! 15619 * Integrated C++ runtime libraries, including support for most major 15620 GNU/Linux systems! 15621 * New instruction scheduler from IBM Haifa which includes support for 15622 function wide instruction scheduling as well as superscalar 15623 scheduling. 15624 * Significantly improved alias analysis code. 15625 * Improved register allocation for two address machines. 15626 * Significant code generation improvements for Fortran code on 15627 Alphas. 15628 * Various optimizations from the g77 project as well as improved loop 15629 optimizations. 15630 * Dwarf2 debug format support for some targets. 15631 * egcs libstdc++ includes the SGI STL implementation without changes. 15632 * As a result of these and other changes, egcs libstc++ is not binary 15633 compatible with previous releases of libstdc++. 15634 * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO 15635 Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and 15636 1.1), Support for RTEMS on several embedded targets, Support for 15637 arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and 15638 MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc. 15639 * Integrated testsuites for gcc, g++, g77, libstdc++ and libio. 15640 * RS6000/PowerPC ports generate code which can run on all 15641 RS6000/PowerPC variants by default. 15642 * -mcpu= and -march= switches for the x86 port to allow better 15643 control over how the x86 port generates code. 15644 * Includes the template repository patch (aka repo patch); note the 15645 new template code makes repo obsolete for ELF systems using gnu-ld 15646 such as GNU/Linux. 15647 * Plus the usual assortment of bugfixes and improvements. 15648 15649 15650 For questions related to the use of GCC, please consult these web 15651 pages and the [3]GCC manuals. If that fails, the 15652 [4]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15653 web pages and the development of GCC are welcome on our developer 15654 list at [5]gcc (a] gcc.gnu.org. All of [6]our lists have public archives. 15655 15656 Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and 15657 distribution of this entire article is permitted in any medium, 15658 provided this notice is preserved. 15659 15660 These pages are [8]maintained by the GCC team. Last modified 15661 2014-06-12[9]. 15662 15663 References 15664 15665 1. http://gcc.gnu.org/egcs-1.0/features-2.8.html 15666 2. http://gcc.gnu.org/egcs-1.0/c++features.html 15667 3. https://gcc.gnu.org/onlinedocs/ 15668 4. mailto:gcc-help (a] gcc.gnu.org 15669 5. mailto:gcc (a] gcc.gnu.org 15670 6. https://gcc.gnu.org/lists.html 15671 7. http://www.fsf.org/ 15672 8. https://gcc.gnu.org/about.html 15673 9. http://validator.w3.org/check/referer 15674 ====================================================================== 15675 http://gcc.gnu.org/egcs-1.0/caveats.html 15676 15677 EGCS 1.0 Caveats 15678 15679 * EGCS has an integrated libstdc++, but does not have an integrated 15680 libg++. Furthermore old libg++ releases will not work with egc; HJ 15681 Lu has made a libg++-2.8.1.2 available which may work with EGCS. 15682 Note most C++ programs only need libstdc++. 15683 * Note that using -pedantic or -Wreturn-type can cause an explosion 15684 in the amount of memory needed for template-heavy C++ code, such as 15685 code that uses STL. Also note that -Wall includes -Wreturn-type, so 15686 if you use -Wall you will need to specify -Wno-return-type to turn 15687 it off. 15688 * Exception handling may not work with shared libraries, particularly 15689 on alphas, hppas, and mips based platforms. Exception handling is 15690 known to work on x86-linux platforms with shared libraries. 15691 * Some versions of the Linux kernel have bugs which prevent them from 15692 being compiled or from running when compiled by EGCS. See the FAQ 15693 (as shipped with EGCS 1.0) for additional information. 15694 * In general, EGCS is more rigorous about rejecting invalid C++ code 15695 or deprecated C++ constructs than G++ 2.7. As a result it may be 15696 necessary to fix C++ code before it will compile with EGCS. 15697 * G++ is also aggressively tracking the C++ standard; as a result 15698 code which was previously valid (and thus accepted by other 15699 compilers and older versions of G++) may no longer be accepted. 15700 * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS 15701 1.0.x and later releases should work with Red Hat Linux 5.0. 15702 15703 15704 For questions related to the use of GCC, please consult these web 15705 pages and the [1]GCC manuals. If that fails, the 15706 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15707 web pages and the development of GCC are welcome on our developer 15708 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 15709 15710 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 15711 distribution of this entire article is permitted in any medium, 15712 provided this notice is preserved. 15713 15714 These pages are [6]maintained by the GCC team. Last modified 15715 2014-06-12[7]. 15716 15717 References 15718 15719 1. https://gcc.gnu.org/onlinedocs/ 15720 2. mailto:gcc-help (a] gcc.gnu.org 15721 3. mailto:gcc (a] gcc.gnu.org 15722 4. https://gcc.gnu.org/lists.html 15723 5. http://www.fsf.org/ 15724 6. https://gcc.gnu.org/about.html 15725 7. http://validator.w3.org/check/referer 15726 ====================================================================== 15727
Indexes created Tue Oct 27 00:25:24 UTC 2026