NEWS revision 1.11
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-8/index.html 9 GCC 8 Release Series 10 11 Feb 22, 2019 12 13 The [1]GNU project and the GCC developers are pleased to announce the 14 release of GCC 8.3. 15 16 This release is a bug-fix release, containing fixes for regressions in 17 GCC 8.2 relative to previous releases of GCC. 18 19 Release History 20 21 GCC 8.3 22 Feb 22, 2019 ([2]changes, [3]documentation) 23 24 GCC 8.2 25 Jul 14, 2018 ([4]changes, [5]documentation) 26 27 GCC 8.1 28 May 2, 2018 ([6]changes, [7]documentation) 29 30 References and Acknowledgements 31 32 GCC used to stand for the GNU C Compiler, but since the compiler 33 supports several other languages aside from C, it now stands for the 34 GNU Compiler Collection. 35 36 A list of [8]successful builds is updated as new information becomes 37 available. 38 39 The GCC developers would like to thank the numerous people that have 40 contributed new features, improvements, bug fixes, and other changes as 41 well as test results to GCC. This [9]amazing group of volunteers is 42 what makes GCC successful. 43 44 For additional information about GCC please refer to the [10]GCC 45 project web site or contact the [11]GCC development mailing list. 46 47 To obtain GCC please use [12]our mirror sites or [13]our SVN server. 48 49 50 For questions related to the use of GCC, please consult these web 51 pages and the [14]GCC manuals. If that fails, the 52 [15]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 53 web pages and the development of GCC are welcome on our developer 54 list at [16]gcc (a] gcc.gnu.org. All of [17]our lists have public 55 archives. 56 57 Copyright (C) [18]Free Software Foundation, Inc. Verbatim copying and 58 distribution of this entire article is permitted in any medium, 59 provided this notice is preserved. 60 61 These pages are [19]maintained by the GCC team. Last modified 62 2019-02-22[20]. 63 64 References 65 66 1. http://www.gnu.org/ 67 2. http://gcc.gnu.org/gcc-8/changes.html 68 3. http://gcc.gnu.org/onlinedocs/8.3.0/ 69 4. http://gcc.gnu.org/gcc-8/changes.html 70 5. http://gcc.gnu.org/onlinedocs/8.2.0/ 71 6. http://gcc.gnu.org/gcc-8/changes.html 72 7. http://gcc.gnu.org/onlinedocs/8.1.0/ 73 8. http://gcc.gnu.org/gcc-8/buildstat.html 74 9. http://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Contributors.html 75 10. http://gcc.gnu.org/index.html 76 11. mailto:gcc (a] gcc.gnu.org 77 12. http://gcc.gnu.org/mirrors.html 78 13. http://gcc.gnu.org/svn.html 79 14. https://gcc.gnu.org/onlinedocs/ 80 15. mailto:gcc-help (a] gcc.gnu.org 81 16. mailto:gcc (a] gcc.gnu.org 82 17. https://gcc.gnu.org/lists.html 83 18. https://www.fsf.org/ 84 19. https://gcc.gnu.org/about.html 85 20. http://validator.w3.org/check/referer 86 ====================================================================== 87 http://gcc.gnu.org/gcc-8/changes.html 88 GCC 8 Release Series 89 Changes, New Features, and Fixes 90 91 This page is a "brief" summary of some of the huge number of 92 improvements in GCC 8. You may also want to check out our [1]Porting to 93 GCC 8 page and the [2]full GCC documentation. 94 95 Caveats 96 97 * Support for the obsolete SDB/coff debug info format has been 98 removed. The option -gcoff no longer does anything. 99 * The Cilk+ extensions to the C and C++ languages have been removed. 100 * The MPX extensions to the C and C++ languages have been deprecated 101 and will be removed in a future release. 102 * The extension allowing arithmetic on std::atomic<void*> and types 103 like std::atomic<R(*)()> has been deprecated. 104 * The non-standard C++0x std::copy_exception function was removed. 105 std::make_exception_ptr should be used instead. 106 * Support for the powerpc*-*-*spe* target ports which have been 107 recently unmaintained and untested in GCC has been declared 108 obsolete in GCC 8 as announced [3]here. Unless there is activity to 109 revive them, the next release of GCC will have their sources 110 permanently removed. 111 112 General Improvements 113 114 * Inter-procedural optimization improvements: 115 + Reworked run-time estimation metrics leading to more realistic 116 guesses driving inliner and cloning heuristics. 117 + The ipa-pure-const pass is extended to propagate the malloc 118 attribute, and the corresponding warning option 119 -Wsuggest-attribute=malloc emits a diagnostic for functions 120 which can be annotated with the malloc attribute. 121 * Profile driven optimization improvements: 122 + New infrastructure for representing profiles (both statically 123 guessed and profile feedback) which allows propagation of 124 additional information about the reliability of the profile. 125 + A number of improvements in the profile updating code solving 126 problems found by new verification code. 127 + Static detection of code which is not executed in a valid run 128 of the program. This includes paths which trigger undefined 129 behavior as well as calls to functions declared with the cold 130 attribute. Newly the noreturn attribute does not imply all 131 effects of cold to differentiate between exit (which is 132 noreturn) and abort (which is in addition not executed in 133 valid runs). 134 + -freorder-blocks-and-partition, a pass splitting function 135 bodies into hot and cold regions, is now enabled by default at 136 -O2 and higher for x86 and x86-64. 137 * Link-time optimization improvements: 138 + We have significantly improved debug information on ELF 139 targets using DWARF by properly preserving language-specific 140 information. This allows for example the libstdc++ 141 pretty-printers to work with LTO optimized executables. 142 * A new option -fcf-protection=[full|branch|return|none] is 143 introduced to perform code instrumentation to increase program 144 security by checking that target addresses of control-flow transfer 145 instructions (such as indirect function call, function return, 146 indirect jump) are valid. Currently the instrumentation is 147 supported on x86 GNU/Linux targets only. See the user guide for 148 further information about the option syntax and section "New 149 Targets and Target Specific Improvements" for IA-32/x86-64 for more 150 details. 151 * The -gcolumn-info option is now enabled by default. It includes 152 column information in addition to just filenames and line numbers 153 in DWARF debugging information. 154 * The polyhedral-based loop nest optimization pass 155 -floop-nest-optimize has been overhauled. It's still considered 156 experimental and may not result in any runtime improvements. 157 * Two new classical loop nest optimization passes have been added. 158 -floop-unroll-and-jam performs outer loop unrolling and fusing of 159 the inner loop copies. -floop-interchange exchanges loops in a loop 160 nest to improve data locality. Both passes are enabled by default 161 at -O3 and above. 162 * The classic loop nest optimization pass -ftree-loop-distribution 163 has been improved and enabled by default at -O3 and above. It 164 supports loop nest distribution in some restricted scenarios; it 165 also supports cancellable innermost loop distribution with loop 166 versioning under run-time alias checks. 167 * The new option -fstack-clash-protection causes the compiler to 168 insert probes whenever stack space is allocated statically or 169 dynamically to reliably detect stack overflows and thus mitigate 170 the attack vector that relies on jumping over a stack guard page as 171 provided by the operating system. 172 * A new pragma GCC unroll has been implemented in the C family of 173 languages, as well as Fortran and Ada, so as to make it possible 174 for the user to have a finer-grained control over the loop 175 unrolling optimization. 176 * GCC has been enhanced to detect more instances of meaningless or 177 mutually exclusive attribute specifications and handle such 178 conflicts more consistently. Mutually exclusive attribute 179 specifications are ignored with a warning regardless of whether 180 they appear on the same declaration or on distinct declarations of 181 the same entity. For example, because the noreturn attribute on the 182 second declaration below is mutually exclusive with the malloc 183 attribute on the first, it is ignored and a warning is issued. 184 > 185 void* __attribute__ ((malloc)) f (unsigned); 186 void* __attribute__ ((noreturn)) f (unsigned); 187 188 warning: ignoring attribute 'noreturn' because it conflicts with attribute 189 'malloc' [-Wattributes] 190 * The gcov tool can distinguish functions that begin on a same line 191 in a source file. This can be a different template instantiation or 192 a class constructor: 193 194 File 'ins.C' 195 Lines executed:100.00% of 8 196 Creating 'ins.C.gcov' 197 198 -: 0:Source:ins.C 199 -: 0:Graph:ins.gcno 200 -: 0:Data:ins.gcda 201 -: 0:Runs:1 202 -: 0:Programs:1 203 -: 1:template<class T> 204 -: 2:class Foo 205 -: 3:{ 206 -: 4: public: 207 2: 5: Foo(): b (1000) {} 208 ------------------ 209 Foo<char>::Foo(): 210 1: 5: Foo(): b (1000) {} 211 ------------------ 212 Foo<int>::Foo(): 213 1: 5: Foo(): b (1000) {} 214 ------------------ 215 2: 6: void inc () { b++; } 216 ------------------ 217 Foo<char>::inc(): 218 1: 6: void inc () { b++; } 219 ------------------ 220 Foo<int>::inc(): 221 1: 6: void inc () { b++; } 222 ------------------ 223 -: 7: 224 -: 8: private: 225 -: 9: int b; 226 -: 10:}; 227 -: 11: 228 1: 12:int main(int argc, char **argv) 229 -: 13:{ 230 1: 14: Foo<int> a; 231 1: 15: Foo<char> b; 232 -: 16: 233 1: 17: a.inc (); 234 1: 18: b.inc (); 235 1: 19:} 236 237 * The gcov tool has more accurate numbers for execution of lines in a 238 source file. 239 * The gcov tool can use TERM colors to provide more readable output. 240 * AddressSanitizer gained a new pair of sanitization options, 241 -fsanitize=pointer-compare and -fsanitize=pointer-subtract, which 242 warn about subtraction (or comparison) of pointers that point to a 243 different memory object: 244 245 int 246 main () 247 { 248 /* Heap allocated memory. */ 249 char *heap1 = (char *)__builtin_malloc (42); 250 char *heap2 = (char *)__builtin_malloc (42); 251 if (heap1 > heap2) 252 return 1; 253 254 return 0; 255 } 256 257 ==17465==ERROR: AddressSanitizer: invalid-pointer-pair: 0x604000000010 0x6040000 258 00050 259 #0 0x40070f in main /tmp/pointer-compare.c:7 260 #1 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 261 #2 0x400629 in _start (/tmp/a.out+0x400629) 262 263 0x604000000010 is located 0 bytes inside of 42-byte region [0x604000000010,0x604 264 00000003a) 265 allocated by thread T0 here: 266 #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan 267 _malloc_linux.cc:86 268 #1 0x4006ea in main /tmp/pointer-compare.c:5 269 #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 270 271 0x604000000050 is located 0 bytes inside of 42-byte region [0x604000000050,0x604 272 00000007a) 273 allocated by thread T0 here: 274 #0 0x7ffff6efb390 in __interceptor_malloc ../../../../libsanitizer/asan/asan 275 _malloc_linux.cc:86 276 #1 0x4006f8 in main /tmp/pointer-compare.c:6 277 #2 0x7ffff6a72a86 in __libc_start_main (/lib64/libc.so.6+0x21a86) 278 279 SUMMARY: AddressSanitizer: invalid-pointer-pair /tmp/pointer-compare.c:7 in main 280 281 * The store merging pass has been enhanced to handle bit-fields and 282 not just constant stores, but also data copying from adjacent 283 memory locations into other adjacent memory locations, including 284 bitwise logical operations on the data. The pass can also handle 285 byte swapping into memory locations. 286 * The undefined behavior sanitizer gained two new options included in 287 -fsanitize=undefined: -fsanitize=builtin which diagnoses at run 288 time invalid arguments to __builtin_clz or __builtin_ctz prefixed 289 builtins, and -fsanitize=pointer-overflow which performs cheap run 290 time tests for pointer wrapping. 291 * A new attribute no_sanitize can be applied to functions to instruct 292 the compiler not to do sanitization of the options provided as 293 arguments to the attribute. Acceptable values for no_sanitize match 294 those acceptable by the -fsanitize command-line option. 295 296 void __attribute__ ((no_sanitize ("alignment", "object-size"))) 297 f () { /* Do something. */; } 298 299 New Languages and Language specific improvements 300 301 Ada 302 303 * For its internal exception handling used on the host for error 304 recovery in the front-end, the compiler now relies on the native 305 exception handling mechanism of the host platform, which should be 306 more efficient than the former mechanism. 307 308 BRIG (HSAIL) 309 310 In this release cycle, the focus for the BRIGFE was on stabilization 311 and performance improvements. Also a couple of completely new features 312 were added. 313 * Improved support for function and module scope group segment 314 variables. PRM specs define function and module scope group segment 315 variables as an experimental feature. However, PRM test suite uses 316 them. Now group segment is handled by separate book keeping of 317 module scope and function (kernel) offsets. Each function has a 318 "frame" in the group segment offset to which is given as an 319 argument, similar to traditional call stack frame handling. 320 * Reduce the number of type conversions due to the untyped HSAIL 321 registers. Instead of always representing the HSAIL's untyped 322 registers as unsigned int, the gccbrig now pre-analyzes the BRIG 323 code and builds the register variables as a type used the most when 324 storing or reading data to/from each register. This reduces the 325 number of total casts which cannot be always optimized away. 326 * Support for BRIG_KIND_NONE directives. 327 * Made -O3 the default optimization level for BRIGFE. 328 * Fixed illegal addresses generated from address expressions which 329 refer only to offset 0. 330 * Fixed a bug with reg+offset addressing on 32b segments. In 'large' 331 mode, the offset is treated as 32bits unless it's in global, 332 read-only or kernarg address space. 333 * Fixed a crash caused sometimes by calls with more than 4 arguments. 334 * Fixed a mis-execution issue with kernels that have both unexpanded 335 ID functions and calls to subfunctions. 336 * Treat HSAIL barrier builtins as setjmp/longjump style functions to 337 avoid illegal optimizations. 338 * Ensure per WI copies of private variables are aligned correctly. 339 * libhsail-rt: Assume the host runtime allocates the work group 340 memory. 341 342 C family 343 344 * New command-line options have been added for the C and C++ 345 compilers: 346 + [4]-Wmultistatement-macros warns about unsafe macros expanding 347 to multiple statements used as a body of a statement such as 348 if, else, while, switch, or for. 349 + [5]-Wstringop-truncation warns for calls to bounded string 350 manipulation functions such as strncat, strncpy, and stpncpy 351 that might either truncate the copied string or leave the 352 destination unchanged. For example, the following call to 353 strncat is diagnosed because it appends just three of the four 354 characters from the source string. 355 void append (char *buf, size_t bufsize) 356 { 357 strncat (buf, ".txt", 3); 358 } 359 warning: 'strncat' output truncated copying 3 bytes from a string of length 4 [- 360 Wstringop-truncation] 361 Similarly, in the following example, the call to strncpy 362 specifies the size of the destination buffer as the bound. If 363 the length of the source string is equal to or greater than 364 this size the result of the copy will not be NUL-terminated. 365 Therefore, the call is also diagnosed. To avoid the warning, 366 specify sizeof buf - 1 as the bound and set the last element 367 of the buffer to NUL. 368 void copy (const char *s) 369 { 370 char buf[80]; 371 strncpy (buf, s, sizeof buf); 372 373 } 374 warning: 'strncpy' specified bound 80 equals destination size [-Wstringop-trunca 375 tion] 376 The -Wstringop-truncation option is included in -Wall. 377 Note that due to GCC bug [6]82944, defining strncat, strncpy, 378 or stpncpy as a macro in a system header as some 379 implementations do, suppresses the warning. 380 + [7]-Wif-not-aligned controls warnings issued in response to 381 invalid uses of objects declared with attribute 382 [8]warn_if_not_aligned. 383 The -Wif-not-aligned option is included in -Wall. 384 + [9]-Wmissing-attributes warns when a declaration of a function 385 is missing one or more attributes that a related function is 386 declared with and whose absence may adversely affect the 387 correctness or efficiency of generated code. For example, in 388 C++, the warning is issued when an explicit specialization of 389 a primary template declared with attribute alloc_align, 390 alloc_size, assume_aligned, format, format_arg, malloc, or 391 nonnull is declared without it. Attributes deprecated, error, 392 and warning suppress the warning. 393 The -Wmissing-attributes option is included in -Wall. 394 + [10]-Wpacked-not-aligned warns when a struct or union declared 395 with attribute packed defines a member with an explicitly 396 specified alignment greater than 1. Such a member will wind up 397 under-aligned. For example, a warning will be issued for the 398 definition of struct A in the following: 399 struct __attribute__ ((aligned (8))) 400 S8 { char a[8]; }; 401 402 struct __attribute__ ((packed)) A 403 { 404 struct S8 s8; 405 }; 406 warning: alignment 1 of 'struct S' is less than 8 [-Wpacked-not-aligned] 407 The -Wpacked-not-aligned option is included in -Wall. 408 + -Wcast-function-type warns when a function pointer is cast to 409 an incompatible function pointer. This warning is enabled by 410 -Wextra. 411 + -Wsizeof-pointer-div warns for suspicious divisions of the 412 size of a pointer by the size of the elements it points to, 413 which looks like the usual way to compute the array size but 414 won't work out correctly with pointers. This warning is 415 enabled by -Wall. 416 + -Wcast-align=strict warns whenever a pointer is cast such that 417 the required alignment of the target is increased. For 418 example, warn if a char * is cast to an int * regardless of 419 the target machine. 420 + -fprofile-abs-path creates absolute path names in the .gcno 421 files. This allows gcov to find the correct sources in 422 projects where compilations occur with different working 423 directories. 424 * -fno-strict-overflow is now mapped to -fwrapv -fwrapv-pointer and 425 signed integer overflow is now undefined by default at all 426 optimization levels. Using -fsanitize=signed-integer-overflow is 427 now the preferred way to audit code, -Wstrict-overflow is 428 deprecated. 429 * The [11]-Warray-bounds option has been improved to detect more 430 instances of out-of-bounds array indices and pointer offsets. For 431 example, negative or excessive indices into flexible array members 432 and string literals are detected. 433 * The [12]-Wrestrict option introduced in GCC 7 has been enhanced to 434 detect many more instances of overlapping accesses to objects via 435 restrict-qualified arguments to standard memory and string 436 manipulation functions such as memcpy and strcpy. For example, the 437 strcpy call in the function below attempts to truncate the string 438 by replacing its initial characters with the last four. However, 439 because the function writes the terminating NUL into a[4], the 440 copies overlap and the call is diagnosed. 441 void f (void) 442 { 443 char a[] = "abcd1234"; 444 strcpy (a, a + 4); 445 446 } 447 warning: 'strcpy' accessing 5 bytes at offsets 0 and 4 overlaps 1 byte at offset 448 4 [-Wrestrict] 449 The -Wrestrict option is included in -Wall. 450 * Several optimizer enhancements have enabled improvements to the 451 [13]-Wformat-overflow and [14]-Wformat-truncation options. The 452 warnings detect more instances of buffer overflow and truncation 453 than in GCC 7 and are better at avoiding certain kinds of false 454 positives. 455 * When reporting mismatching argument types at a function call, the C 456 and C++ compilers now underline both the argument and the pertinent 457 parameter in the declaration. 458 $ gcc arg-type-mismatch.cc 459 arg-type-mismatch.cc: In function 'int caller(int, int, float)': 460 arg-type-mismatch.cc:5:24: error: invalid conversion from 'int' to 'const char*' 461 [-fpermissive] 462 return callee(first, second, third); 463 ^~~~~~ 464 arg-type-mismatch.cc:1:40: note: initializing argument 2 of 'int callee(int, c 465 onst char*, float)' 466 extern int callee(int one, const char *two, float three); 467 ~~~~~~~~~~~~^~~ 468 469 * When reporting on unrecognized identifiers, the C and C++ compilers 470 will now emit fix-it hints suggesting #include directives for 471 various headers in the C and C++ standard libraries. 472 $ gcc incomplete.c 473 incomplete.c: In function 'test': 474 incomplete.c:3:10: error: 'NULL' undeclared (first use in this function) 475 return NULL; 476 ^~~~ 477 incomplete.c:3:10: note: 'NULL' is defined in header '<stddef.h>'; did you forge 478 t to '#include <stddef.h>'? 479 incomplete.c:1:1: 480 +#include <stddef.h> 481 const char *test(void) 482 incomplete.c:3:10: 483 return NULL; 484 ^~~~ 485 incomplete.c:3:10: note: each undeclared identifier is reported only once for ea 486 ch function it appears in 487 488 $ gcc incomplete.cc 489 incomplete.cc:1:6: error: 'string' in namespace 'std' does not name a type 490 std::string s("hello world"); 491 ^~~~~~ 492 incomplete.cc:1:1: note: 'std::string' is defined in header '<string>'; did you 493 forget to '#include <string>'? 494 +#include <string> 495 std::string s("hello world"); 496 ^~~ 497 498 * The C and C++ compilers now use more intuitive locations when 499 reporting on missing semicolons, and offer fix-it hints: 500 $ gcc t.c 501 t.c: In function 'test': 502 t.c:3:12: error: expected ';' before '}' token 503 return 42 504 ^ 505 ; 506 } 507 ~ 508 509 * When reporting on missing '}' and ')' tokens, the C and C++ 510 compilers will now highlight the corresponding '{' and '(' token, 511 issuing a 'note' if it's on a separate line: 512 $ gcc unclosed.c 513 unclosed.c: In function 'log_when_out_of_range': 514 unclosed.c:12:50: error: expected ')' before '{' token 515 && (temperature < MIN || temperature > MAX) { 516 ^~ 517 ) 518 unclosed.c:11:6: note: to match this '(' 519 if (logging_enabled && check_range () 520 ^ 521 522 or highlighting it directly if it's on the same line: 523 $ gcc unclosed-2.c 524 unclosed-2.c: In function 'test': 525 unclosed-2.c:8:45: error: expected ')' before '{' token 526 if (temperature < MIN || temperature > MAX { 527 ~ ^~ 528 ) 529 530 They will also emit fix-it hints. 531 532 C++ 533 534 * GCC 8 (-fabi-version=12) has a couple of corrections to the calling 535 convention, which changes the ABI for some uncommon code: 536 + Passing an empty class as an argument now takes up no space on 537 x86_64, as required by the psABI. 538 + Passing or returning a class with only deleted copy and move 539 constructors now uses the same calling convention as a class 540 with a non-trivial copy or move constructor. This only affects 541 C++17 mode, as in earlier standards passing or returning such 542 a class was impossible. 543 + WARNING: In GCC 8.1 the second change mistakenly also affects 544 classes with a deleted copy constructor and defaulted trivial 545 move constructor (bug [15]c++/86094). This issue is fixed in 546 GCC 8.2 (-fabi-version=13). 547 You can test whether these changes affect your code with -Wabi=11 548 (or -Wabi=12 in GCC 8.2 for the third issue); if these changes are 549 problematic for your project, the GCC 7 ABI can be selected with 550 -fabi-version=11. 551 * The value of the C++11 alignof operator has been corrected to match 552 C _Alignof (minimum alignment) rather than GNU __alignof__ 553 (preferred alignment); on ia32 targets this means that 554 alignof(double) is now 4 rather than 8. Code that wants the 555 preferred alignment should use __alignof__ instead. 556 * New command-line options have been added for the C++ compiler to 557 control warnings: 558 + [16]-Wclass-memaccess warns when objects of non-trivial class 559 types are manipulated in potentially unsafe ways by raw memory 560 functions such as memcpy, or realloc. The warning helps detect 561 calls that bypass user-defined constructors or copy-assignment 562 operators, corrupt virtual table pointers, data members of 563 const-qualified types or references, or member pointers. The 564 warning also detects calls that would bypass access controls 565 to data members. For example, a call such as: 566 memcpy (&std::cout, &std::cerr, sizeof std::cout); 567 results in 568 warning: 'void* memcpy(void*, const void*, long unsigned int)' writing t 569 o an object of type 'std::ostream' {aka 'class std::basic_ostream<char>'} with n 570 o trivial copy-assignment [-Wclass-memaccess] 571 The -Wclass-memaccess option is included in -Wall. 572 * The C++ front end has experimental support for some of the upcoming 573 C++2a draft features with the -std=c++2a or -std=gnu++2a flags, 574 including designated initializers, default member initializers for 575 bit-fields, __VA_OPT__ (except that #__VA_OPT__ is unsupported), 576 lambda [=, this] captures, etc. For a full list of new features, 577 see [17]the C++ status page. 578 * When reporting on attempts to access private fields of a class or 579 struct, the C++ compiler will now offer fix-it hints showing how to 580 use an accessor function to get at the field in question, if one 581 exists. 582 $ gcc accessor.cc 583 accessor.cc: In function 'void test(foo*)': 584 accessor.cc:12:12: error: 'double foo::m_ratio' is private within this context 585 if (ptr->m_ratio >= 0.5) 586 ^~~~~~~ 587 accessor.cc:7:10: note: declared private here 588 double m_ratio; 589 ^~~~~~~ 590 accessor.cc:12:12: note: field 'double foo::m_ratio' can be accessed via 'double 591 foo::get_ratio() const' 592 if (ptr->m_ratio >= 0.5) 593 ^~~~~~~ 594 get_ratio() 595 596 * The C++ compiler can now give you a hint if you use a macro before 597 it was defined (e.g. if you mess up the order of your #include 598 directives): 599 $ gcc ordering.cc 600 ordering.cc:2:24: error: expected ';' at end of member declaration 601 virtual void clone() const OVERRIDE { } 602 ^~~~~ 603 ; 604 ordering.cc:2:30: error: 'OVERRIDE' does not name a type 605 virtual void clone() const OVERRIDE { } 606 ^~~~~~~~ 607 ordering.cc:2:30: note: the macro 'OVERRIDE' had not yet been defined 608 In file included from ordering.cc:5: 609 c++11-compat.h:2: note: it was later defined here 610 #define OVERRIDE override 611 612 613 * The -Wold-style-cast diagnostic can now emit fix-it hints telling 614 you when you can use a static_cast, const_cast, or 615 reinterpret_cast. 616 $ gcc -c old-style-cast-fixits.cc -Wold-style-cast 617 old-style-cast-fixits.cc: In function 'void test(void*)': 618 old-style-cast-fixits.cc:5:19: warning: use of old-style cast to 'struct foo*' [ 619 -Wold-style-cast] 620 foo *f = (foo *)ptr; 621 ^~~ 622 ---------- 623 static_cast<foo *> (ptr) 624 625 * When reporting on problems within extern "C" linkage 626 specifications, the C++ compiler will now display the location of 627 the start of the extern "C". 628 $ gcc -c extern-c.cc 629 extern-c.cc:3:1: error: template with C linkage 630 template <typename T> void test (void); 631 ^~~~~~~~ 632 In file included from extern-c.cc:1: 633 unclosed.h:1:1: note: 'extern "C"' linkage started here 634 extern "C" { 635 ^~~~~~~~~~ 636 extern-c.cc:3:39: error: expected '}' at end of input 637 template <typename T> void test (void); 638 ^ 639 In file included from extern-c.cc:1: 640 unclosed.h:1:12: note: to match this '{' 641 extern "C" { 642 ^ 643 644 * When reporting on mismatching template types, the C++ compiler will 645 now use color to highlight the mismatching parts of the template, 646 and will elide the parameters that are common between two 647 mismatching templates, printing [...] instead: 648 $ gcc templates.cc 649 templates.cc: In function 'void test()': 650 templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl 651 e>' to 'vector<int>' 652 fn_1(vector<double> ()); 653 ^~~~~~~~~~~~~~~~~ 654 templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<[...] 655 ,double>' to 'map<[...],int>' 656 fn_2(map<int, double>()); 657 ^~~~~~~~~~~~~~~~~~ 658 659 Those [...] elided parameters can be seen using -fno-elide-type: 660 $ gcc templates.cc -fno-elide-type 661 templates.cc: In function 'void test()': 662 templates.cc:9:8: error: could not convert 'vector<double>()' from 'vector<doubl 663 e>' to 'vector<int>' 664 fn_1(vector<double> ()); 665 ^~~~~~~~~~~~~~~~~ 666 templates.cc:10:8: error: could not convert 'map<int, double>()' from 'map<int,d 667 ouble>' to 'map<int,int>' 668 fn_2(map<int, double>()); 669 ^~~~~~~~~~~~~~~~~~ 670 671 The C++ compiler has also gained an option 672 -fdiagnostics-show-template-tree which visualizes such mismatching 673 templates in a hierarchical form: 674 $ gcc templates-2.cc -fdiagnostics-show-template-tree 675 templates-2.cc: In function 'void test()': 676 templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou 677 ble>' to 'vector<int>' 678 vector< 679 [double != int]> 680 fn_1(vector<double> ()); 681 ^~~~~~~~~~~~~~~~~ 682 templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve 683 ctor<double> >()' from 'map<map<[...],vector<double>>,vector<double>>' to 'map<m 684 ap<[...],vector<float>>,vector<float>>' 685 map< 686 map< 687 [...], 688 vector< 689 [double != float]>>, 690 vector< 691 [double != float]>> 692 fn_2(map<map<int, vector<double>>, vector<double>> ()); 693 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 694 695 which again works with -fno-elide-type: 696 $ gcc templates-2.cc -fdiagnostics-show-template-tree -fno-elide-type 697 templates-2.cc: In function 'void test()': 698 templates-2.cc:9:8: error: could not convert 'vector<double>()' from 'vector<dou 699 ble>' to 'vector<int>' 700 vector< 701 [double != int]> 702 fn_1(vector<double> ()); 703 ^~~~~~~~~~~~~~~~~ 704 templates-2.cc:10:8: error: could not convert 'map<map<int, vector<double> >, ve 705 ctor<double> >()' from 'map<map<int,vector<double>>,vector<double>>' to 'map<map 706 <int,vector<float>>,vector<float>>' 707 map< 708 map< 709 int, 710 vector< 711 [double != float]>>, 712 vector< 713 [double != float]>> 714 fn_2(map<map<int, vector<double>>, vector<double>> ()); 715 ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 716 717 * Flowing off the end of a non-void function is considered 718 unreachable and may be subject to optimization on that basis. As a 719 result of this change, -Wreturn-type warnings are enabled by 720 default for C++. 721 722 Runtime Library (libstdc++) 723 724 * Improved experimental support for C++17, including the following 725 features: 726 + Deduction guides to support class template argument deduction. 727 + std::filesystem implementation. 728 + std::char_traits<char> and std::char_traits<wchar_t> are 729 usable in constant expressions. 730 + std::to_chars and std::from_chars (for integers only, not for 731 floating point types). 732 * Experimental support for C++2a: std::to_address (thanks to Glen 733 Fernandes) and std::endian. 734 * On GNU/Linux, std::random_device::entropy() accesses the kernel's 735 entropy count for the random device, if known (thanks to Xi 736 Ruoyao). 737 * Support for std::experimental::source_location. 738 * AddressSanitizer integration for std::vector, detecting 739 out-of-range accesses to the unused capacity of a vector. 740 * Extensions __gnu_cxx::airy_ai and __gnu_cxx::airy_bi added to the 741 Mathematical Special Functions. 742 743 Fortran 744 745 * The main version of libfortran has been changed to 5. 746 * Parameterized derived types, a major feature of Fortran 2003, have 747 been implemented. 748 * The maximum rank for arrays has been increased to 15, conforming to 749 the Fortran 2008 standard. 750 * Transformational intrinsics are now fully supported in 751 initialization expressions. 752 * New flag -fc-prototypes to write C prototypes for BIND(C) 753 procedures and variables. 754 * If -fmax-stack-var-size is honored if given together with -Ofast, 755 -fstack-arrays is no longer set in that case. 756 * New options -fdefault-real-16 and -fdefault-real-10 to control the 757 default kind of REAL variables. 758 * A warning is now issued if an array subscript inside a DO loop 759 could lead to an out-of-bounds-access. The new option 760 -Wdo-subscript, enabled by -Wextra, warns about this even if the 761 compiler can not prove that the code will be executed. 762 * The Fortran front end now attempts to interchange loops if it is 763 deemed profitable. So far, this is restricted to FORALL and DO 764 CONCURRENT statements with multiple indices. This behavior be 765 controlled with the new flag -ffrontend-loop-interchange, which is 766 enabled with optimization by default. The 767 -Wfrontend-loop-interchange option warns about such occurrences. 768 * When an actual argument contains too few elements for a dummy 769 argument, an error is now issued. The -std=legacy option can be 770 used to still compile such code. 771 * The RECL= argument to OPEN and INQUIRE statements now allows 64-bit 772 integers, making records larger than 2GiB possible. 773 * The GFORTRAN_DEFAULT_RECL environment variable no longer has any 774 effect. The record length for preconnected units is now larger than 775 any practical limit, same as for sequential access units opened 776 without an explicit RECL= specifier. 777 * Character variables longer than HUGE(0) elements are now possible 778 on 64-bit targets. Note that this changes the procedure call ABI 779 for all procedures with character arguments on 64-bit targets, as 780 the type of the hidden character length argument has changed. The 781 hidden character length argument is now of type INTEGER(C_SIZE_T). 782 * Partial support is provided for Fortran 2018 teams, which are 783 hierarchical subsets of images that execute independently of other 784 image subsets. 785 786 Go 787 788 * GCC 8 provides a complete implementation of the Go 1.10.1 user 789 packages. 790 * The garbage collector is now fully concurrent. As before, values 791 stored on the stack are scanned conservatively, but value stored in 792 the heap are scanned precisely. 793 * Escape analysis is fully implemented and enabled by default in the 794 Go frontend. This significantly reduces the number of heap 795 allocations by allocating values on the stack instead. 796 797 libgccjit 798 799 The libgccjit API gained four new entry points: 800 * [18]gcc_jit_type_get_vector and 801 * [19]gcc_jit_context_new_rvalue_from_vector for working with 802 vectors, 803 * [20]gcc_jit_type_get_aligned 804 * [21]gcc_jit_function_get_address 805 806 The C code generated by [22]gcc_jit_context_dump_reproducer_to_file is 807 now easier-to-read. 808 809 New Targets and Target Specific Improvements 810 811 AArch64 812 813 * The Armv8.4-A architecture is now supported. It can be used by 814 specifying the -march=armv8.4-a option. 815 * The Dot Product instructions are now supported as an optional 816 extension to the Armv8.2-A architecture and newer and are mandatory 817 on Armv8.4-A. The extension can be used by specifying the +dotprod 818 architecture extension. E.g. -march=armv8.2-a+dotprod. 819 * The Armv8-A +crypto extension has now been split into two 820 extensions for finer grained control: 821 + +aes which contains the Armv8-A AES crytographic instructions. 822 + +sha2 which contains the Armv8-A SHA2 and SHA1 cryptographic 823 instructions. 824 Using +crypto will now enable these two extensions. 825 * New Armv8.4-A FP16 Floating Point Multiplication Variant 826 instructions have been added. These instructions are mandatory in 827 Armv8.4-A but available as an optional extension to Armv8.2-A and 828 Armv8.3-A. The new extension can be used by specifying the +fp16fml 829 architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A 830 the instructions can be enabled by specifying +fp16. 831 * New cryptographic instructions have been added as optional 832 extensions to Armv8.2-A and newer. These instructions can be 833 enabled with: 834 + +sha3 New SHA3 and SHA2 instructions from Armv8.4-A. This 835 implies +sha2. 836 + +sm4 New SM3 and SM4 instructions from Armv8.4-A. 837 * The Scalable Vector Extension (SVE) is now supported as an optional 838 extension to the Armv8.2-A architecture and newer. This support 839 includes automatic vectorization with SVE instructions, but it does 840 not yet include the SVE Arm C Language Extensions (ACLE). It can be 841 enabled by specifying the +sve architecture extension (for example, 842 -march=armv8.2-a+sve). By default, the generated code works with 843 all vector lengths, but it can be made specific to N-bit vectors 844 using -msve-vector-bits=N. 845 * Support has been added for the following processors (GCC 846 identifiers in parentheses): 847 + Arm Cortex-A75 (cortex-a75). 848 + Arm Cortex-A55 (cortex-a55). 849 + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE 850 (cortex-a75.cortex-a55). 851 The GCC identifiers can be used as arguments to the -mcpu or -mtune 852 options, for example: -mcpu=cortex-a75 or -mtune=cortex-a75 or as 853 arguments to the equivalent target attributes and pragmas. 854 855 ARC 856 857 * Added support for: 858 + Fast interrupts. 859 + Naked functions. 860 + aux variable attributes. 861 + uncached type qualifier. 862 + Secure functions via sjli instruction. 863 * New exception handling implementation. 864 * Revamped trampoline implementation. 865 * Refactored small data feature implementation, controlled via -G 866 command line option. 867 * New support for reduced register set ARC architecture 868 configurations, controlled via -mrf16 command line option. 869 * Refurbished and improved support for zero overhead loops. 870 Introduced -mlpc-width command line option to control the width of 871 lp_count register. 872 873 ARM 874 875 * The -mfpu option now takes a new option setting of -mfpu=auto. When 876 set to this the floating-point and SIMD settings are derived from 877 the settings of the -mcpu or -march options. The internal CPU 878 configurations have been updated with information about the 879 permitted floating-point configurations supported. See the user 880 guide for further information about the extended option syntax for 881 controlling architectural extensions via the -march option. 882 -mfpu=auto is now the default setting unless the compiler has been 883 configured with an explicit --with-fpu option. 884 * The -march and -mcpu options now accept optional extensions to the 885 architecture or CPU option, allowing the user to enable or disable 886 any such extensions supported by that architecture or CPU such as 887 (but not limited to) floating-point and AdvancedSIMD. For example: 888 the option -mcpu=cortex-a53+nofp will generate code for the 889 Cortex-A53 processor with no floating-point support. This, in 890 combination with the new -mfpu=auto option, provides a 891 straightforward way of specifying a valid build target through a 892 single -mcpu or -march option. The -mtune option accepts the same 893 arguments as -mcpu but only the CPU name has an effect on tuning. 894 The architecture extensions do not have any effect. For details of 895 what extensions a particular architecture or CPU option supports 896 please refer to the [23]documentation. 897 * The -mstructure-size-boundary option has been deprecated and will 898 be removed in a future release. 899 * The default link behavior for Armv6 and Armv7-R targets has been 900 changed to produce BE8 format when generating big-endian images. A 901 new flag -mbe32 can be used to force the linker to produce legacy 902 BE32 format images. There is no change of behavior for Armv6-M and 903 other Armv7 or later targets: these already defaulted to BE8 904 format. This change brings GCC into alignment with other compilers 905 for the ARM architecture. 906 * The Armv8-R architecture is now supported. It can be used by 907 specifying the -march=armv8-r option. 908 * The Armv8.3-A architecture is now supported. It can be used by 909 specifying the -march=armv8.3-a option. 910 * The Armv8.4-A architecture is now supported. It can be used by 911 specifying the -march=armv8.4-a option. 912 * The Dot Product instructions are now supported as an optional 913 extension to the Armv8.2-A architecture and newer and are mandatory 914 on Armv8.4-A. The extension can be used by specifying the +dotprod 915 architecture extension. E.g. -march=armv8.2-a+dotprod. 916 * Support for setting extensions and architectures using the GCC 917 target pragma and attribute has been added. It can be used by 918 specifying #pragma GCC target ("arch=..."), #pragma GCC target 919 ("+extension"), __attribute__((target("arch=..."))) or 920 __attribute__((target("+extension"))). 921 * New Armv8.4-A FP16 Floating Point Multiplication Variant 922 instructions have been added. These instructions are mandatory in 923 Armv8.4-A but available as an optional extension to Armv8.2-A and 924 Armv8.3-A. The new extension can be used by specifying the +fp16fml 925 architectural extension on Armv8.2-A and Armv8.3-A. On Armv8.4-A 926 the instructions can be enabled by specifying +fp16. 927 * Support has been added for the following processors (GCC 928 identifiers in parentheses): 929 + Arm Cortex-A75 (cortex-a75). 930 + Arm Cortex-A55 (cortex-a55). 931 + Arm Cortex-A55/Cortex-A75 DynamIQ big.LITTLE 932 (cortex-a75.cortex-a55). 933 + Arm Cortex-R52 for Armv8-R (cortex-r52). 934 The GCC identifiers can be used as arguments to the -mcpu or -mtune 935 options, for example: -mcpu=cortex-a75 or -mtune=cortex-r52 or as 936 arguments to the equivalent target attributes and pragmas. 937 938 AVR 939 940 * The AVR port now supports the following XMEGA-like devices: 941 942 ATtiny212, ATtiny214, ATtiny412, ATtiny414, ATtiny416, ATtiny417, 943 ATtiny814, ATtiny816, ATtiny817, ATtiny1614, ATtiny1616, ATtiny1617, 944 ATtiny3214, ATtiny3216, ATtiny3217 945 The new devices are listed under [24]-mmcu=avrxmega3. 946 + These devices see flash memory in the RAM address space, so 947 that features like PROGMEM and __flash are not needed any more 948 (as opposed to other AVR families for which read-only data 949 will be located in RAM except special, non-standard features 950 are used to locate and access such data). This requires that 951 the compiler is used with Binutils 2.29 or newer so that 952 [25]read-only data will be located in flash memory. 953 + A new command-line option -mshort-calls is supported. This 954 option is used internally for multilib selection of the 955 avrxmega3 variants. It is not an optimization option. Do not 956 set it by hand. 957 * The compiler now generates [26]efficient interrupt service routine 958 (ISR) prologues and epilogues. This is achieved by using the new 959 [27]AVR pseudo instruction __gcc_isr which is supported and 960 resolved by the GNU assembler. 961 + As the __gcc_isr pseudo-instruction will be resolved by the 962 assembler, inline assembly is transparent to the process. This 963 means that when inline assembly uses an instruction like INC 964 that clobbers the condition code, then the assembler will 965 detect this and generate an appropriate ISR prologue / 966 epilogue chunk to save / restore SREG as needed. 967 + A new command-line option -mno-gas-isr-prologues disables the 968 generation of the __gcc_isr pseudo instruction. Any non-naked 969 ISR will save and restore SREG, tmp_reg and zero_reg, no 970 matter whether the respective register is clobbered or used. 971 + The feature is turned on per default for all optimization 972 levels except for -O0 and -Og. It is explicitly enabled by 973 means of option -mgas-isr-prologues. 974 + Support has been added for a new [28]AVR function attribute 975 no_gccisr. It can be used to disable __gcc_isr pseudo 976 instruction generation for individual ISRs. 977 + This optimization is only available if GCC is configured with 978 GNU Binutils 2.29 or newer; or at least with a version of 979 Binutils that implements feature [29]PR21683. 980 * The compiler no more saves / restores registers in main; the effect 981 is the same as if attribute OS_task was specified for main. This 982 optimization can be switched off by the new command-line option 983 -mno-main-is-OS_task. 984 985 IA-32/x86-64 986 987 * The x86 port now supports the naked function attribute. 988 * Better tuning for znver1 and Intel Core based CPUs. 989 * Vectorization cost metrics has been reworked leading to significant 990 improvements on some benchmarks. 991 * GCC now supports the Intel CPU named Cannonlake through 992 -march=cannonlake. The switch enables the AVX512VBMI, AVX512IFMA 993 and SHA ISA extensions. 994 * GCC now supports the Intel CPU named Icelake through 995 -march=icelake. The switch enables the AVX512VNNI, GFNI, VAES, 996 AVX512VBMI2, VPCLMULQDQ, AVX512BITALG, RDPID and AVX512VPOPCNTDQ 997 ISA extensions. 998 * GCC now supports the Intel Control-flow Enforcement Technology 999 (CET) extension through -fcf-protection option. 1000 1001 NDS32 1002 1003 * New command-line options -mext-perf, -mext-perf2, and -mext-string 1004 have been added for performance extension instructions. 1005 1006 Nios II 1007 1008 * The Nios II back end has been improved to generate better-optimized 1009 code. Changes include switching to LRA, more accurate cost models, 1010 and more compact code for addressing static variables. 1011 * New command-line options -mgprel-sec= and -mr0rel-sec= have been 1012 added. 1013 * The stack-smashing protection options are now enabled on Nios II. 1014 1015 PA-RISC 1016 1017 * The default call ABI on 32-bit linux has been changed from callee 1018 copies to caller copies. This affects objects larger than eight 1019 bytes passed by value. The goal is to improve compatibility with 1020 x86 and resolve issues with OpenMP. 1021 * Other PA-RISC targets are unchanged. 1022 1023 PowerPC / PowerPC64 / RS6000 1024 1025 * The PowerPC SPE support is split off to a separate powerpcspe port. 1026 The separate port is deprecated and might be removed in a future 1027 release. 1028 * The Paired Single support (as used on some PPC750 CPUs, -mpaired, 1029 powerpc*-*-linux*paired*) is deprecated and will be removed in a 1030 future release. 1031 * The Xilinx floating point support (-mxilinx-fpu, 1032 powerpc-xilinx-eabi*) is deprecated and will be removed in a future 1033 release. 1034 * Support for using big-endian AltiVec intrinsics on a little-endian 1035 target (-maltivec=be) is deprecated and will be removed in a future 1036 release. 1037 1038 Tile 1039 1040 * The TILE-Gx port is deprecated and will be removed in a future 1041 release. 1042 1043 Operating Systems 1044 1045 Windows 1046 1047 * GCC on Microsoft Windows can now be configured via 1048 --enable-mingw-wildcard or --disable-mingw-wildcard to force a 1049 specific behavior for GCC itself with regards to supporting the 1050 wildcard character. Prior versions of GCC would follow the 1051 configuration of the MinGW runtime. This behavior can still be 1052 obtained by not using the above options or by using 1053 --enable-mingw-wildcard=platform. 1054 1055 Improvements for plugin authors 1056 1057 * Plugins can now register a callback hook for when comments are 1058 encountered by the C and C++ compilers, e.g. allowing for plugins 1059 to handle documentation markup in code comments. 1060 * The gdbinit support script for debugging GCC now has a 1061 break-on-diagnostic command, providing an easy way to trigger a 1062 breakpoint whenever a diagnostic is emitted. 1063 * The API for creating fix-it hints now supports newlines, and for 1064 emitting mutually incompatible fix-it hints for one diagnostic. 1065 1066 GCC 8.1 1067 1068 This is the [30]list of problem reports (PRs) from GCC's bug tracking 1069 system that are known to be fixed in the 8.1 release. This list might 1070 not be complete (that is, it is possible that some PRs that have been 1071 fixed are not listed here). 1072 1073 GCC 8.2 1074 1075 This is the [31]list of problem reports (PRs) from GCC's bug tracking 1076 system that are known to be fixed in the 8.2 release. This list might 1077 not be complete (that is, it is possible that some PRs that have been 1078 fixed are not listed here). 1079 1080 General Improvements 1081 1082 * Fixed LTO link-time performance problems caused by an overflow in 1083 the partitioning algorithm while building large binaries. 1084 1085 Language Specific Changes 1086 1087 C++ 1088 1089 GCC 8.2 fixed a bug introduced in GCC 8.1 affecting passing or 1090 returning of classes with a deleted copy constructor and defaulted 1091 trivial move constructor (bug [32]c++/86094). GCC 8.2 introduces 1092 -fabi-version=13 and makes it the default, ABI incompatibilities 1093 between GCC 8.1 and 8.2 can be reported with -Wabi=12. See [33]C++ 1094 changes for more details. 1095 1096 Target Specific Changes 1097 1098 IA-32/x86-64 1099 1100 * -mtune=native performance regression [34]PR84413 on Intel Skylake 1101 processors has been fixed. 1102 1103 GCC 8.3 1104 1105 This is the [35]list of problem reports (PRs) from GCC's bug tracking 1106 system that are known to be fixed in the 8.3 release. This list might 1107 not be complete (that is, it is possible that some PRs that have been 1108 fixed are not listed here). 1109 1110 Windows 1111 1112 * A C++ Microsoft ABI bitfield layout bug, [36]PR87137 has been 1113 fixed. A non-field declaration could cause the current bitfield 1114 allocation unit to be completed, incorrectly placing a following 1115 bitfield into a new allocation unit. The Microsoft ABI is selected 1116 for: 1117 + Mingw targets 1118 + PowerPC, IA-32 or x86-64 targets when the -mms-bitfields 1119 option is specified, or __attribute__((ms_struct)) is used 1120 + SuperH targets when the -mhitachi option is specified, or 1121 __attribute__((renesas)) is used 1122 GCC 8 introduced additional cases of this defect, but rather than 1123 resolve only those regressions, we decided to resolve all the cases 1124 of this defect in single change. 1125 1126 1127 For questions related to the use of GCC, please consult these web 1128 pages and the [37]GCC manuals. If that fails, the 1129 [38]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1130 web pages and the development of GCC are welcome on our developer 1131 list at [39]gcc (a] gcc.gnu.org. All of [40]our lists have public 1132 archives. 1133 1134 Copyright (C) [41]Free Software Foundation, Inc. Verbatim copying and 1135 distribution of this entire article is permitted in any medium, 1136 provided this notice is preserved. 1137 1138 These pages are [42]maintained by the GCC team. Last modified 1139 2019-02-22[43]. 1140 1141 References 1142 1143 1. http://gcc.gnu.org/gcc-8/porting_to.html 1144 2. http://gcc.gnu.org/onlinedocs/index.html#current 1145 3. https://gcc.gnu.org/ml/gcc/2018-04/msg00102.html 1146 4. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmultistatement-macros 1147 5. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wstringop-truncation 1148 6. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82944 1149 7. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wif-not-aligned 1150 8. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Common-Variable-Attributes.html#index-warn_005fif_005fnot_005faligned-variable-attribute 1151 9. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wmissing-attributes 1152 10. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wpacked-not-aligned 1153 11. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Warray-bounds 1154 12. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wrestrict 1155 13. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-overflow 1156 14. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/Warning-Options.html#index-Wformat-truncation 1157 15. https://gcc.gnu.org/PR86094 1158 16. https://gcc.gnu.org/onlinedocs/gcc/C_002b_002b-Dialect-Options.html#index-Wclass-memaccess 1159 17. http://gcc.gnu.org/projects/cxx-status.html#cxx2a 1160 18. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_vector 1161 19. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/expressions.html#gcc_jit_context_new_rvalue_from_vector 1162 20. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/types.html#gcc_jit_type_get_aligned 1163 21. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/function-pointers.html#gcc_jit_function_get_address 1164 22. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/jit/topics/contexts.html#gcc_jit_context_dump_reproducer_to_file 1165 23. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/ARM-Options.html#ARM-Options 1166 24. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Options.html 1167 25. https://sourceware.org/PR21472 1168 26. https://gcc.gnu.org/PR20296 1169 27. https://sourceware.org/binutils/docs-2.29/as/AVR-Pseudo-Instructions.html 1170 28. https://gcc.gnu.org/onlinedocs/gcc-8.1.0/gcc/AVR-Function-Attributes.html 1171 29. https://sourceware.org/PR21683 1172 30. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.0 1173 31. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.2 1174 32. https://gcc.gnu.org/PR86094 1175 33. http://gcc.gnu.org/gcc-8/changes.html#cxx 1176 34. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84413 1177 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=8.3 1178 36. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87137 1179 37. https://gcc.gnu.org/onlinedocs/ 1180 38. mailto:gcc-help (a] gcc.gnu.org 1181 39. mailto:gcc (a] gcc.gnu.org 1182 40. https://gcc.gnu.org/lists.html 1183 41. https://www.fsf.org/ 1184 42. https://gcc.gnu.org/about.html 1185 43. http://validator.w3.org/check/referer 1186 ====================================================================== 1187 http://gcc.gnu.org/gcc-7/index.html 1188 GCC 7 Release Series 1189 1190 Dec 6, 2018 1191 1192 The [1]GNU project and the GCC developers are pleased to announce the 1193 release of GCC 7.4. 1194 1195 This release is a bug-fix release, containing fixes for regressions in 1196 GCC 7.3 relative to previous releases of GCC. 1197 1198 Release History 1199 1200 GCC 7.4 1201 Dec 6, 2018 ([2]changes, [3]documentation) 1202 1203 GCC 7.3 1204 Jan 25, 2018 ([4]changes, [5]documentation) 1205 1206 GCC 7.2 1207 Aug 14, 2017 ([6]changes, [7]documentation) 1208 1209 GCC 7.1 1210 May 2, 2017 ([8]changes, [9]documentation) 1211 1212 References and Acknowledgements 1213 1214 GCC used to stand for the GNU C Compiler, but since the compiler 1215 supports several other languages aside from C, it now stands for the 1216 GNU Compiler Collection. 1217 1218 A list of [10]successful builds is updated as new information becomes 1219 available. 1220 1221 The GCC developers would like to thank the numerous people that have 1222 contributed new features, improvements, bug fixes, and other changes as 1223 well as test results to GCC. This [11]amazing group of volunteers is 1224 what makes GCC successful. 1225 1226 For additional information about GCC please refer to the [12]GCC 1227 project web site or contact the [13]GCC development mailing list. 1228 1229 To obtain GCC please use [14]our mirror sites or [15]our SVN server. 1230 1231 1232 For questions related to the use of GCC, please consult these web 1233 pages and the [16]GCC manuals. If that fails, the 1234 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1235 web pages and the development of GCC are welcome on our developer 1236 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 1237 archives. 1238 1239 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 1240 distribution of this entire article is permitted in any medium, 1241 provided this notice is preserved. 1242 1243 These pages are [21]maintained by the GCC team. Last modified 1244 2018-12-06[22]. 1245 1246 References 1247 1248 1. http://www.gnu.org/ 1249 2. http://gcc.gnu.org/gcc-7/changes.html 1250 3. http://gcc.gnu.org/onlinedocs/7.4.0/ 1251 4. http://gcc.gnu.org/gcc-7/changes.html 1252 5. http://gcc.gnu.org/onlinedocs/7.3.0/ 1253 6. http://gcc.gnu.org/gcc-7/changes.html 1254 7. http://gcc.gnu.org/onlinedocs/7.2.0/ 1255 8. http://gcc.gnu.org/gcc-7/changes.html 1256 9. http://gcc.gnu.org/onlinedocs/7.1.0/ 1257 10. http://gcc.gnu.org/gcc-7/buildstat.html 1258 11. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 1259 12. http://gcc.gnu.org/index.html 1260 13. mailto:gcc (a] gcc.gnu.org 1261 14. http://gcc.gnu.org/mirrors.html 1262 15. http://gcc.gnu.org/svn.html 1263 16. https://gcc.gnu.org/onlinedocs/ 1264 17. mailto:gcc-help (a] gcc.gnu.org 1265 18. mailto:gcc (a] gcc.gnu.org 1266 19. https://gcc.gnu.org/lists.html 1267 20. https://www.fsf.org/ 1268 21. https://gcc.gnu.org/about.html 1269 22. http://validator.w3.org/check/referer 1270 ====================================================================== 1271 http://gcc.gnu.org/gcc-7/changes.html 1272 GCC 7 Release Series 1273 Changes, New Features, and Fixes 1274 1275 This page is a brief summary of some of the huge number of improvements 1276 in GCC 7. For more information, see the [1]Porting to GCC 7 page and 1277 the [2]full GCC documentation. 1278 1279 Caveats 1280 1281 * GCC now uses [3]LRA (a new local register allocator) by default for 1282 new targets. 1283 * The non-standard C++0x type traits has_trivial_default_constructor, 1284 has_trivial_copy_constructor and has_trivial_copy_assign have been 1285 removed. 1286 * The libstdc++ [4]Profile Mode has been deprecated and will be 1287 removed in a future version. 1288 * The Cilk+ extensions to the C and C++ languages have been 1289 deprecated. 1290 * On ARM targets (arm*-*-*), [5]a bug introduced in GCC 5 that 1291 affects conformance to the procedure call standard (AAPCS) has been 1292 fixed. The bug affects some C++ code where class objects are passed 1293 by value to functions and could result in incorrect or inconsistent 1294 code being generated. This is an ABI change. If the option -Wpsabi 1295 is enabled (on by default) the compiler will emit a diagnostic note 1296 for code that might be affected. 1297 1298 General Optimizer Improvements 1299 1300 * GCC 7 can determine the return value or range of return values of 1301 some calls to the sprintf family of functions and make it available 1302 to other optimization passes. Some calls to the snprintf function 1303 with a zero size argument can be folded into constants. This 1304 optimization is included in -O1 and can be selectively controlled 1305 by the -fprintf-return-value option. 1306 * A new store merging pass has been added. It merges constant stores 1307 to adjacent memory locations into fewer, wider, stores. It is 1308 enabled by the -fstore-merging option and at the -O2 optimization 1309 level or higher (and -Os). 1310 * A new code hoisting optimization has been added to the partial 1311 redundancy elimination pass. It attempts to move evaluation of 1312 expressions executed on all paths to the function exit as early as 1313 possible. This primarily helps improve code size, but can improve 1314 the speed of the generated code as well. It is enabled by the 1315 -fcode-hoisting option and at the -O2 optimization level or higher 1316 (and -Os). 1317 * A new interprocedural bitwise constant propagation optimization has 1318 been added, which propagates knowledge about which bits of 1319 variables are known to be zero (including pointer alignment 1320 information) across the call graph. It is enabled by the 1321 -fipa-bit-cp option if -fipa-cp is enabled as well, and is enabled 1322 at the -O2 optimization level and higher (and -Os). This 1323 optimization supersedes interprocedural alignment propagation of 1324 GCC 6, and therefore the option -fipa-cp-alignment is now 1325 deprecated and ignored. 1326 * A new interprocedural value range propagation optimization has been 1327 added, which propagates integral range information across the call 1328 graph when variable values can be proven to be within those ranges. 1329 It is enabled by the -fipa-vrp option and at the -O2 optimization 1330 level and higher (and -Os). 1331 * A new loop splitting optimization pass has been added. Certain 1332 loops which contain a condition that is always true on one side of 1333 the iteration space and always false on the other are split into 1334 two loops, such that each of the two new loops iterates on just one 1335 side of the iteration space and the condition does not need to be 1336 checked inside of the loop. It is enabled by the -fsplit-loops 1337 option and at the -O3 optimization level or higher. 1338 * The shrink-wrapping optimization can now separate portions of 1339 prologues and epilogues to improve performance if some of the work 1340 done traditionally by prologues and epilogues is not needed on 1341 certain paths. This is controlled by the -fshrink-wrap-separate 1342 option, enabled by default. It requires target support, which is 1343 currently only implemented in the PowerPC and AArch64 ports. 1344 * AddressSanitizer gained a new sanitization option, 1345 -fsanitize-address-use-after-scope, which enables sanitization of 1346 variables whose address is taken and used after a scope where the 1347 variable is defined: 1348 1349 int 1350 main (int argc, char **argv) 1351 { 1352 char *ptr; 1353 { 1354 char my_char; 1355 ptr = &my_char; 1356 } 1357 1358 *ptr = 123; 1359 return *ptr; 1360 } 1361 1362 ==28882==ERROR: AddressSanitizer: stack-use-after-scope on address 0x7fffb8dba99 1363 0 at pc 0x0000004006d5 bp 0x7fffb8dba960 sp 0x7fffb8dba958 1364 WRITE of size 1 at 0x7fffb8dba990 thread T0 1365 #0 0x4006d4 in main /tmp/use-after-scope-1.c:10 1366 #1 0x7f9c71943290 in __libc_start_main (/lib64/libc.so.6+0x20290) 1367 #2 0x400739 in _start (/tmp/a.out+0x400739) 1368 1369 Address 0x7fffb8dba990 is located in stack of thread T0 at offset 32 in frame 1370 #0 0x40067f in main /tmp/use-after-scope-1.c:3 1371 1372 This frame has 1 object(s): 1373 [32, 33) 'my_char' <== Memory access at offset 32 is inside this variable 1374 1375 The option is enabled by default with -fsanitize=address and 1376 disabled by default with -fsanitize=kernel-address. Compared to the 1377 LLVM compiler, where the option already exists, the implementation 1378 in the GCC compiler has some improvements and advantages: 1379 + Complex uses of gotos and case labels are properly handled and 1380 should not report any false positive or false negatives. 1381 + C++ temporaries are sanitized. 1382 + Sanitization can handle invalid memory stores that are 1383 optimized out by the LLVM compiler when optimization is 1384 enabled. 1385 * The -fsanitize=signed-integer-overflow suboption of the 1386 UndefinedBehavior Sanitizer now diagnoses arithmetic overflows even 1387 on arithmetic operations with generic vectors. 1388 * Version 5 of the [6]DWARF debugging information standard is 1389 supported through the -gdwarf-5 option. The DWARF version 4 1390 debugging information remains the default until consumers of 1391 debugging information are adjusted. 1392 1393 New Languages and Language specific improvements 1394 1395 OpenACC support in C, C++, and Fortran continues to be maintained and 1396 improved. See the [7]OpenACC and [8]Offloading wiki pages for further 1397 information. 1398 1399 Ada 1400 1401 * On mainstream native platforms, Ada programs no longer require the 1402 stack to be made executable in order to run properly. 1403 1404 BRIG (HSAIL) 1405 1406 Support for processing BRIG 1.0 files was added in this release. BRIG 1407 is a binary format for HSAIL (Heterogeneous System Architecture 1408 Intermediate Language). The BRIG front end can be used for implementing 1409 HSAIL "finalizers" (compilation of HSAIL to a native ISA) for 1410 GCC-supported targets. An implementation of an HSAIL runtime library, 1411 libhsail-rt is also included. 1412 1413 C family 1414 1415 * New command-line options have been added for the C and C++ 1416 compilers: 1417 + -Wimplicit-fallthrough warns when a switch case falls through. 1418 This warning has five different levels. The compiler is able 1419 to parse a wide range of fallthrough comments, depending on 1420 the level. It also handles control-flow statements, such as 1421 ifs. It's possible to suppress the warning by either adding a 1422 fallthrough comment, or by using a null statement: 1423 __attribute__ ((fallthrough)); (C, C++), or [[fallthrough]]; 1424 (C++17), or [[gnu::fallthrough]]; (C++11/C++14). This warning 1425 is enabled by -Wextra. 1426 + -Wpointer-compare warns when a pointer is compared with a zero 1427 character constant. Such code is now invalid in C++11 and GCC 1428 rejects it. This warning is enabled by default. 1429 + -Wduplicated-branches warns when an if-else has identical 1430 branches. 1431 + -Wrestrict warns when an argument passed to a 1432 restrict-qualified parameter aliases with another argument. 1433 + -Wmemset-elt-size warns for memset calls, when the first 1434 argument references an array, and the third argument is a 1435 number equal to the number of elements of the array, but not 1436 the size of the array. This warning is enabled by -Wall. 1437 + -Wint-in-bool-context warns about suspicious uses of integer 1438 values where boolean values are expected. This warning is 1439 enabled by -Wall. 1440 + -Wswitch-unreachable warns when a switch statement has 1441 statements between the controlling expression and the first 1442 case label which will never be executed. This warning is 1443 enabled by default. 1444 + -Wexpansion-to-defined warns when defined is used outside #if. 1445 This warning is enabled by -Wextra or -Wpedantic. 1446 + -Wregister warns about uses of the register storage specifier. 1447 In C++17 this keyword has been removed and for C++17 this is a 1448 pedantic warning enabled by default. The warning is not 1449 emitted for the GNU Explicit Register Variables extension. 1450 + -Wvla-larger-than=N warns about unbounded uses of 1451 variable-length arrays, and about bounded uses of 1452 variable-length arrays whose bound can be larger than N bytes. 1453 + -Wduplicate-decl-specifier warns when a declaration has 1454 duplicate const, volatile, restrict or _Atomic specifier. This 1455 warning is enabled by -Wall. 1456 * GCC 6's C and C++ front ends were able to offer suggestions for 1457 misspelled field names: 1458 1459 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 1460 you mean 'color'? 1461 return ptr->colour; 1462 ^~~~~~ 1463 1464 GCC 7 greatly expands the scope of these suggestions. Firstly, it 1465 adds fix-it hints to such suggestions: 1466 1467 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 1468 you mean 'color'? 1469 return ptr->colour; 1470 ^~~~~~ 1471 color 1472 1473 The suggestions now cover many other things, such as misspelled 1474 function names: 1475 1476 spellcheck-identifiers.c:11:3: warning: implicit declaration of function 'gtk_wi 1477 dget_showall'; did you mean 'gtk_widget_show_all'? [-Wimplicit-function-declarat 1478 ion] 1479 gtk_widget_showall (w); 1480 ^~~~~~~~~~~~~~~~~~ 1481 gtk_widget_show_all 1482 1483 misspelled macro names and enum values: 1484 1485 spellcheck-identifiers.cc:85:11: error: 'MAX_ITEM' undeclared here (not in a fun 1486 ction); did you mean 'MAX_ITEMS'? 1487 int array[MAX_ITEM]; 1488 ^~~~~~~~ 1489 MAX_ITEMS 1490 1491 misspelled type names: 1492 1493 spellcheck-typenames.c:7:14: error: unknown type name 'singed'; did you mean 'si 1494 gned'? 1495 void test (singed char e); 1496 ^~~~~~ 1497 signed 1498 1499 and, in the C front end, named initializers: 1500 1501 test.c:7:20: error: 'struct s' has no member named 'colour'; did you mean 'color 1502 '? 1503 struct s test = { .colour = 3 }; 1504 ^~~~~~ 1505 color 1506 1507 * The preprocessor can now offer suggestions for misspelled 1508 directives, e.g.: 1509 1510 test.c:5:2: error:invalid preprocessing directive #endfi; did you mean #endif? 1511 #endfi 1512 ^~~~~ 1513 endif 1514 1515 * Warnings about format strings now underline the pertinent part of 1516 the string, and can offer suggested fixes. In some cases, the 1517 pertinent argument is underlined. 1518 1519 test.c:51:29: warning: format '%s' expects argument of type 'char *', but argume 1520 nt 3 has type 'int' [-Wformat=] 1521 printf ("foo: %d bar: %s baz: %d", 100, i + j, 102); 1522 ~^ ~~~~~ 1523 %d 1524 1525 * The new -Wdangling-else command-line option has been split out of 1526 -Wparentheses and warns about dangling else. 1527 * The -Wshadow warning has been split into three variants. 1528 -Wshadow=global warns for any shadowing. This is the default when 1529 using -Wshadow without any argument. -Wshadow=local only warns for 1530 a local variable shadowing another local variable or parameter. 1531 -Wshadow=compatible-local only warns for a local variable shadowing 1532 another local variable or parameter whose type is compatible (in 1533 C++ compatible means that the type of the shadowing variable can be 1534 converted to that of the shadowed variable). 1535 The following example shows the different kinds of shadow warnings: 1536 1537 enum operation { add, count }; 1538 struct container { int nr; }; 1539 1540 int 1541 container_count (struct container c, int count) 1542 { 1543 int r = 0; 1544 for (int count = 0; count > 0; count--) 1545 { 1546 struct container count = c; 1547 r += count.nr; 1548 } 1549 return r; 1550 } 1551 1552 -Wshadow=compatible-local will warn for the parameter being 1553 shadowed with the same type: 1554 1555 warn-test.c:8:12: warning: declaration of 'count' shadows a parameter [-Wshadow= 1556 compatible-local] 1557 for (int count = 0; count > 0; count--) 1558 ^~~~~ 1559 warn-test.c:5:42: note: shadowed declaration is here 1560 container_count (struct container c, int count) 1561 ^~~~~ 1562 1563 -Wshadow=local will warn for the above and for the shadowed 1564 declaration with incompatible type: 1565 1566 warn-test.c:10:24: warning: declaration of 'count' shadows a previous local [-Ws 1567 hadow=local] 1568 struct container count = c; 1569 ^~~~~ 1570 warn-test.c:8:12: note: shadowed declaration is here 1571 for (int count = 0; count > 0; count--) 1572 ^~~~~ 1573 1574 -Wshadow=global will warn for all of the above and the shadowing of 1575 the global declaration: 1576 1577 warn-test.c:5:42: warning: declaration of 'count' shadows a global declaration [ 1578 -Wshadow] 1579 container_count (struct container c, int count) 1580 ^~~~~ 1581 warn-test.c:1:23: note: shadowed declaration is here 1582 enum operation { add, count }; 1583 ^~~~~ 1584 1585 * GCC 7 contains a number of enhancements that help detect buffer 1586 overflow and other forms of invalid memory accesses. 1587 + The -Walloc-size-larger-than=size option detects calls to 1588 standard and user-defined memory allocation functions 1589 decorated with attribute alloc_size whose argument exceeds the 1590 specified size (PTRDIFF_MAX by default). The option also 1591 detects arithmetic overflow in the computation of the size in 1592 two-argument allocation functions like calloc where the total 1593 size is the product of the two arguments. Since calls with an 1594 excessive size cannot succeed they are typically the result of 1595 programming errors. Such bugs have been known to be the source 1596 of security vulnerabilities and a target of exploits. 1597 -Walloc-size-larger-than=PTRDIFF_MAX is included in -Wall. 1598 For example, the following call to malloc incorrectly tries to 1599 avoid passing a negative argument to the function and instead 1600 ends up unconditionally invoking it with an argument less than 1601 or equal to zero. Since after conversion to the type of the 1602 argument of the function (size_t) a negative argument results 1603 in a value in excess of the maximum PTRDIFF_MAX the call is 1604 diagnosed. 1605 1606 void* f (int n) 1607 { 1608 return malloc (n > 0 ? 0 : n); 1609 } 1610 1611 warning: argument 1 range [2147483648, 4294967295] exceeds maximum object size 2 1612 147483647 [-Walloc-size-larger-than=] 1613 1614 + The -Walloc-zero option detects calls to standard and 1615 user-defined memory allocation functions decorated with 1616 attribute alloc_size with a zero argument. -Walloc-zero is not 1617 included in either -Wall or -Wextra and must be explicitly 1618 enabled. 1619 + The -Walloca option detects all calls to the alloca function 1620 in the program. -Walloca is not included in either -Wall or 1621 -Wextra and must be explicitly enabled. 1622 + The -Walloca-larger-than=size option detects calls to the 1623 alloca function whose argument either may exceed the specified 1624 size, or that is not known to be sufficiently constrained to 1625 avoid exceeding it. -Walloca-larger-than is not included in 1626 either -Wall or -Wextra and must be explicitly enabled. 1627 For example, compiling the following snippet with 1628 -Walloca-larger-than=1024 results in a warning because even 1629 though the code appears to call alloca only with sizes of 1kb 1630 and less, since n is signed, a negative value would result in 1631 a call to the function well in excess of the limit. 1632 1633 void f (int n) 1634 { 1635 char *d; 1636 if (n < 1025) 1637 d = alloca (n); 1638 else 1639 d = malloc (n); 1640 1641 } 1642 1643 warning: argument to 'alloca may be too large due to conversion from 'int' to 'l 1644 ong unsigned int' [-Walloca-larger-than=] 1645 1646 In contrast, a call to alloca that isn't bounded at all such 1647 as in the following function will elicit the warning below 1648 regardless of the size argument to the option. 1649 1650 void f (size_t n) 1651 { 1652 char *d = alloca (n); 1653 1654 } 1655 1656 warning: unbounded use of 'alloca' [-Walloca-larger-than=] 1657 1658 + The -Wformat-overflow=level option detects certain and likely 1659 buffer overflow in calls to the sprintf family of formatted 1660 output functions. Although the option is enabled even without 1661 optimization it works best with -O2 and higher. 1662 For example, in the following snippet the call to sprintf is 1663 diagnosed because even though its output has been constrained 1664 using the modulo operation it could result in as many as three 1665 bytes if mday were negative. The solution is to either 1666 allocate a larger buffer or make sure the argument is not 1667 negative, for example by changing mday's type to unsigned or 1668 by making the type of the second operand of the modulo 1669 expression unsigned: 100U. 1670 1671 void* f (int mday) 1672 { 1673 char *buf = malloc (3); 1674 sprintf (buf, "%02i", mday % 100); 1675 return buf; 1676 } 1677 1678 warning: 'sprintf may write a terminating nul past the end of the destination [- 1679 Wformat-overflow=] 1680 note: 'sprintf' output between 3 and 4 bytes into a destination of size 3 1681 1682 + The -Wformat-truncation=level option detects certain and 1683 likely output truncation in calls to the snprintf family of 1684 formatted output functions. -Wformat-truncation=1 is included 1685 in -Wall and enabled without optimization but works best with 1686 -O2 and higher. 1687 For example, the following function attempts to format an 1688 integer between 0 and 255 in hexadecimal, including the 0x 1689 prefix, into a buffer of four characters. But since the 1690 function must always terminate output by the null character 1691 ('\0') such a buffer is only big enough to fit just one digit 1692 plus the prefix. Therefore the snprintf call is diagnosed. To 1693 avoid the warning either use a bigger buffer or handle the 1694 function's return value which indicates whether or not its 1695 output has been truncated. 1696 1697 void f (unsigned x) 1698 { 1699 char d[4]; 1700 snprintf (d, sizeof d, "%#02x", x & 0xff); 1701 1702 } 1703 1704 warning: 'snprintf' output may be truncated before the last format character [-W 1705 format-truncation=] 1706 note: 'snprintf' output between 3 and 5 bytes into a destination of size 4 1707 1708 + The -Wnonnull option has been enhanced to detect a broader set 1709 of cases of passing null pointers to functions that expect a 1710 non-null argument (those decorated with attribute nonnull). By 1711 taking advantage of optimizations the option can detect many 1712 more cases of the problem than in prior GCC versions. 1713 + The -Wstringop-overflow=type option detects buffer overflow in 1714 calls to string handling functions like memcpy and strcpy. The 1715 option relies on [9]Object Size Checking and has an effect 1716 similar to defining the _FORTIFY_SOURCE macro. 1717 -Wstringop-overflow=2 is enabled by default. 1718 For example, in the following snippet, because the call to 1719 strncat specifies a maximum that allows the function to write 1720 past the end of the destination, it is diagnosed. To correct 1721 the problem and avoid the overflow the function should be 1722 called with a size of at most sizeof d - strlen(d) - 1. 1723 1724 void f (const char *fname) 1725 { 1726 char d[8]; 1727 strncpy (d, "/tmp/", sizeof d); 1728 strncat (d, fname, sizeof d); 1729 1730 } 1731 1732 warning: specified bound 8 equals the size of the destination [-Wstringop-overfl 1733 ow=] 1734 1735 * The <limits.h> header provided by GCC defines macros such as 1736 INT_WIDTH for the width in bits of integer types, if 1737 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 1738 included. The <stdint.h> header defines such macros as SIZE_WIDTH 1739 and INTMAX_WIDTH for the width of some standard typedef names for 1740 integer types, again if __STDC_WANT_IEC_60559_BFP_EXT__ is defined 1741 before the header is included; note that GCC's implementation of 1742 this header is only used for freestanding compilations, not hosted 1743 compilations, on most systems. These macros come from ISO/IEC TS 1744 18661-1:2014. 1745 * The <float.h> header provided by GCC defines the macro 1746 CR_DECIMAL_DIG, from ISO/IEC TS 18661-1:2014, if 1747 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 1748 included. This represents the number of decimal digits for which 1749 conversions between decimal character strings and binary formats, 1750 in both directions, are correctly rounded, and currently has the 1751 value of UINTMAX_MAX on all systems, reflecting that GCC's 1752 compile-time conversions are correctly rounded for any number of 1753 digits. 1754 * New __builtin_add_overflow_p, __builtin_sub_overflow_p, 1755 __builtin_mul_overflow_p built-in functions have been added. These 1756 work similarly to their siblings without the _p suffix, but do not 1757 actually store the result of the arithmetics anywhere, just return 1758 whether the operation would overflow. Calls to these built-ins with 1759 integer constant arguments evaluate to integer constants 1760 expressions. 1761 For example, in the following, c is assigned the result of a * b 1762 only if the multiplication does not overflow, otherwise it is 1763 assigned the value zero. The multiplication is performed at 1764 compile-time and without triggering a -Woverflow warning. 1765 1766 enum { 1767 a = 12345678, 1768 b = 87654321, 1769 c = __builtin_mul_overflow_p (a, b, a) ? 0 : a * b 1770 }; 1771 1772 C 1773 1774 * The C front end now supports type names _FloatN for floating-point 1775 types with IEEE interchange formats and _FloatNx for floating-point 1776 types with IEEE extended formats. These type names come from 1777 ISO/IEC TS 18661-3:2015. 1778 The set of types supported depends on the target for which GCC is 1779 configured. Most targets support _Float32, _Float32x and _Float64. 1780 _Float128 is supported on targets where IEEE binary128 encoding was 1781 already supported as long double or __float128. _Float64x is 1782 supported on targets where a type with either binary128 or Intel 1783 extended precision format is available. 1784 Constants with these types are supported using suffixes fN, FN, fNx 1785 and FNx (e.g., 1.2f128 or 2.3F64x). Macros such as FLT128_MAX are 1786 defined in <float.h> if __STDC_WANT_IEC_60559_TYPES_EXT__ is 1787 defined before it is included. 1788 These new types are always distinct from each other and from float, 1789 double and long double, even if they have the same encoding. 1790 Complex types such as _Complex _Float128 are also supported. 1791 Type-generic built-in functions such as __builtin_isinf support the 1792 new types, and the following type-specific built-in functions have 1793 versions (suffixed fN or fNx) for the new types: 1794 __builtin_copysign, __builtin_fabs, __builtin_huge_val, 1795 __builtin_inf, __builtin_nan, __builtin_nans. 1796 * Compilation with -fopenmp is now compatible with the C11 _Atomic 1797 keyword. 1798 1799 C++ 1800 1801 * The C++ front end has experimental support for all of the current 1802 C++17 draft with the -std=c++1z or -std=gnu++1z flags, including if 1803 constexpr, class template argument deduction, auto template 1804 parameters, and structured bindings. For a full list of new 1805 features, see [10]the C++ status page. 1806 * C++17 support for new of over-aligned types can be enabled in other 1807 modes with the -faligned-new flag. 1808 * The C++17 evaluation order requirements can be selected in other 1809 modes with the -fstrong-eval-order flag, or disabled in C++17 mode 1810 with -fno-strong-eval-order. 1811 * The default semantics of inherited constructors has changed in all 1812 modes, following [11]P0136. Essentially, overload resolution 1813 happens as if calling the inherited constructor directly, and the 1814 compiler fills in construction of the other bases and members as 1815 needed. Most uses should not need any changes. The old behavior can 1816 be restored with -fno-new-inheriting-ctors, or -fabi-version less 1817 than 11. 1818 * The resolution of DR 150 on matching of template template 1819 parameters, allowing default template arguments to make a template 1820 match a parameter, is currently enabled by default in C++17 mode 1821 only. The default can be overridden with -f{no-,}new-ttp-matching. 1822 * The C++ front end will now provide fix-it hints for some missing 1823 semicolons, allowing for automatic fixes by IDEs: 1824 1825 test.cc:4:11: error: expected ';' after class definition 1826 class a {} 1827 ^ 1828 ; 1829 1830 * -Waligned-new has been added to the C++ front end. It warns about 1831 new of type with extended alignment without -faligned-new. 1832 1833 Runtime Library (libstdc++) 1834 1835 * The type of exception thrown by iostreams, std::ios_base::failure, 1836 now uses the [12]cxx11 ABI. 1837 * Experimental support for C++17, including the following new 1838 features: 1839 + std::string_view; 1840 + std::any, std::optional, and std::variant; 1841 + std::invoke, std::is_invocable, std::is_nothrow_invocable, and 1842 invoke_result; 1843 + std::is_swappable, and std::is_nothrow_swappable; 1844 + std::apply, and std::make_from_tuple; 1845 + std::void_t, std::bool_constant, std::conjunction, 1846 std::disjunction, and std::negation; 1847 + Variable templates for type traits; 1848 + Mathematical Special Functions; 1849 + std::chrono::floor, std::chrono::ceil, std::chrono::round, and 1850 std::chrono::abs; 1851 + std::clamp, std::gcd, std::lcm, 3-dimensional std::hypot; 1852 + std::scoped_lock, std::shared_mutex, 1853 std::atomic<T>::is_always_lock_free; 1854 + std::sample, std::default_searcher, std::boyer_moore_searcher 1855 and std::boyer_moore_horspool_searcher; 1856 + Extraction and re-insertion of map and set nodes, try_emplace 1857 members for maps, and functions for accessing containers 1858 std::size, std::empty, and std::data; 1859 + std::shared_ptr support for arrays, 1860 std::shared_ptr<T>::weak_type, 1861 std::enable_shared_from_this<T>::weak_from_this(), and 1862 std::owner_less<void>; 1863 + std::byte; 1864 + std::as_const, std::not_fn, 1865 std::has_unique_object_representations, constexpr 1866 std::addressof. 1867 Thanks to Daniel Krgler, Tim Shen, Edward Smith-Rowland, and Ville 1868 Voutilainen for work on the C++17 support. 1869 * A new power-of-two rehashing policy for use with the _Hashtable 1870 internals, thanks to Franois Dumont. 1871 1872 Fortran 1873 1874 * Support for a number of extensions for compatibility with legacy 1875 code with new flags: 1876 + -fdec-structure Support for DEC STRUCTURE and UNION 1877 + -fdec-intrinsic-ints Support for new integer intrinsics with 1878 B/I/J/K prefixes such as BABS, JIAND... 1879 + -fdec-math Support for additional math intrinsics, including 1880 COTAN and degree-valued trigonometric functions such as TAND, 1881 ASIND... 1882 + -fdec Enable the -fdec-* family of extensions. 1883 * New flag -finit-derived to allow default initialization of 1884 derived-type variables. 1885 * Improved DO loops with step equal to 1 or -1, generates faster code 1886 without a loop preheader. A new warning, -Wundefined-do-loop, warns 1887 when a loop iterates either to HUGE(i) (with step equal to 1), or 1888 to -HUGE(i) (with step equal to -1). Invalid behavior can be caught 1889 at run time with -fcheck=do enabled: 1890 1891 program test 1892 implicit none 1893 integer(1) :: i 1894 do i = -HUGE(i)+10, -HUGE(i)-1, -1 1895 print *, i 1896 end do 1897 end program test 1898 1899 At line 8 of file do_check_12.f90 1900 Fortran runtime error: Loop iterates infinitely 1901 1902 * Version 4.5 of the [13]OpenMP specification is now partially 1903 supported in the Fortran compiler; the largest missing item is 1904 structure element mapping. 1905 * User-defined derived-type input/output (UDTIO) is added. 1906 * Derived type coarrays with allocatable and pointer components are 1907 partially supported. 1908 * Non-constant stop codes and error stop codes (Fortran 2015 1909 feature). 1910 * Derived types with allocatable components of recursive type. 1911 * Intrinsic assignment to polymorphic variables. 1912 * Improved submodule support. 1913 * Improved diagnostics (polymorphic results in pure functions). 1914 * Coarray: Support for failed images (Fortan 2015 feature). 1915 1916 Go 1917 1918 * GCC 7 provides a complete implementation of the Go 1.8.1 user 1919 packages. 1920 * Compared to the Go 1.8.1 toolchain, the garbage collector is more 1921 conservative and less concurrent. 1922 * Escape analysis is available for experimental use via the 1923 -fgo-optimize-allocs option. The -fgo-debug-escape prints 1924 information useful for debugging escape analysis choices. 1925 1926 Java (GCJ) 1927 1928 The GCC Java front end and associated libjava runtime library have been 1929 removed from GCC. 1930 1931 libgccjit 1932 1933 The libgccjit API gained support for marking calls as requiring 1934 tail-call optimization via a new entry point: 1935 [14]gcc_jit_rvalue_set_bool_require_tail_call. 1936 1937 libgccjit performs numerous checks at the API boundary, but if these 1938 succeed, it previously ignored errors and other diagnostics emitted 1939 within the core of GCC, and treated the compile of a gcc_jit_context as 1940 having succeeded. As of GCC 7 it now ensures that if any diagnostics 1941 are emitted, they are visible from the libgccjit API, and that the the 1942 context is flagged as having failed. 1943 1944 New Targets and Target Specific Improvements 1945 1946 AArch64 1947 1948 * GCC has been updated to the latest revision of the procedure call 1949 standard (AAPCS64) to provide support for parameter passing when 1950 data types have been over-aligned. 1951 * The ARMv8.3-A architecture is now supported. It can be used by 1952 specifying the -march=armv8.3-a option. 1953 * The option -msign-return-address= is supported to enable return 1954 address protection using ARMv8.3-A Pointer Authentication 1955 Extensions. For more information on the arguments accepted by this 1956 option, please refer to [15]AArch64-Options. 1957 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 1958 Extensions are now supported. They can be used by specifying the 1959 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 1960 Floating-Point Extensions introduce new half-precision data 1961 processing floating-point instructions. 1962 * Support has been added for the following processors (GCC 1963 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), Broadcom 1964 Vulcan (vulcan), Cavium ThunderX CN81xx (thunderxt81), Cavium 1965 ThunderX CN83xx (thunderxt83), Cavium ThunderX CN88xx 1966 (thunderxt88), Cavium ThunderX CN88xx pass 1.x (thunderxt88p1), 1967 Cavium ThunderX 2 CN99xx (thunderx2t99), Qualcomm Falkor (falkor). 1968 The GCC identifiers can be used as arguments to the -mcpu or -mtune 1969 options, for example: -mcpu=cortex-a73 or -mtune=vulcan or as 1970 arguments to the equivalent target attributes and pragmas. 1971 1972 ARC 1973 1974 * Added support for ARC HS and ARC EM processors. 1975 * Added support for ARC EM variation found in Intel QuarkSE SoCs. 1976 * Added support for NPS400 ARC700 based CPUs. 1977 * Thread Local Storage is now supported by ARC CPUs. 1978 * Fixed errors for ARC600 when using 32x16 multiplier option. 1979 * Fixed PIE for ARC CPUs. 1980 * New CPU templates are supported via multilib. 1981 1982 ARM 1983 1984 * Support for the ARMv5 and ARMv5E architectures has been deprecated 1985 (which have no known implementations) and will be removed in a 1986 future GCC release. Note that ARMv5T, ARMv5TE and ARMv5TEJ 1987 architectures remain supported. The values armv5 and armv5e of 1988 -march are thus deprecated. 1989 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 1990 Extensions are now supported. They can be used by specifying the 1991 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 1992 Floating-Point Extensions introduce new half-precision data 1993 processing floating-point instructions. 1994 * The ARMv8-M architecture is now supported in its two architecture 1995 profiles: ARMv8-M Baseline and ARMv8-M Mainline with its DSP and 1996 Floating-Point Extensions. They can be used by specifying the 1997 -march=armv8-m.base, armv8-m.main or armv8-m.main+dsp options. 1998 * Support has been added for the following processors (GCC 1999 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), ARM 2000 Cortex-M23 (cortex-m23) and ARM Cortex-M33 (cortex-m33). The GCC 2001 identifiers can be used as arguments to the -mcpu or -mtune 2002 options, for example: -mcpu=cortex-a73 or -mtune=cortex-m33. 2003 * A new command-line option -mpure-code has been added. It does not 2004 allow constant data to be placed in code sections. This option is 2005 only available when generating non-PIC code for ARMv7-M targets. 2006 * Support for the ACLE Coprocessor Intrinsics has been added. This 2007 enables the generation of coprocessor instructions through the use 2008 of intrinsics such as cdp, ldc, and others. 2009 * The configure option --with-multilib-list now accepts the value 2010 rmprofile to build multilib libraries for a range of embedded 2011 targets. See our [16]installation instructions for details. 2012 2013 AVR 2014 2015 * On the reduced Tiny cores, the progmem [17]variable attribute is 2016 now properly supported. Respective read-only variables are located 2017 in flash memory in section .progmem.data. No special code is needed 2018 to access such variables; the compiler automatically adds an offset 2019 of 0x4000 to all addresses, which is needed to access variables in 2020 flash memory. As opposed to ordinary cores where it is sufficient 2021 to specify the progmem attribute with definitions, on the reduced 2022 Tiny cores the attribute also has to be specified with (external) 2023 declarations: 2024 2025 extern const int array[] __attribute__((__progmem__)); 2026 2027 int get_value2 (void) 2028 { 2029 /* Access via addresses array + 0x4004 and array + 0x4005. */ 2030 return array[2]; 2031 } 2032 2033 const int* get_address (unsigned idx) 2034 { 2035 /* Returns array + 0x4000 + 2 * idx. */ 2036 return &array[idx]; 2037 } 2038 2039 * A new command-line option -Wmisspelled-isr has been added. It turns 2040 off or turns into errors warnings that are reported for 2041 interrupt service routines (ISRs) which don't follow AVR-LibC's 2042 naming convention of prefixing ISR names with __vector. 2043 * __builtin_avr_nops(n) is a new [18]built-in function that inserts n 2044 NOP instructions into the instruction stream. n must be a value 2045 known at compile time. 2046 2047 IA-32/x86-64 2048 2049 * Support for the AVX-512 Fused Multiply Accumulation Packed Single 2050 precision (4FMAPS), AVX-512 Vector Neural Network Instructions Word 2051 variable precision (4VNNIW), AVX-512 Vector Population Count 2052 (VPOPCNTDQ) and Software Guard Extensions (SGX) ISA extensions has 2053 been added. 2054 2055 NVPTX 2056 2057 * OpenMP target regions can now be offloaded to NVidia PTX GPGPUs. 2058 See the [19]Offloading Wiki on how to configure it. 2059 2060 PowerPC / PowerPC64 / RS6000 2061 2062 * The PowerPC port now uses LRA by default. 2063 * GCC now diagnoses inline assembly that clobbers register r2. This 2064 has always been invalid code, and is no longer quietly tolerated. 2065 * The PowerPC port's support for ISA 3.0 (-mcpu=power9) has been 2066 enhanced to generate more of the new instructions by default, and 2067 to provide more built-in functions to generate code for other new 2068 instructions. 2069 * The configuration option --enable-gnu-indirect-function is now 2070 enabled by default on PowerPC GNU/Linux builds. 2071 * The PowerPC port will now allow 64-bit and 32-bit integer types to 2072 be allocated to the VSX vector registers (ISA 2.06 and above). In 2073 addition, on ISA 3.0, 16-bit and 8-bit integer types can be 2074 allocated in the vector registers. Previously, only 64-bit integer 2075 types were allowed in the traditional floating point registers. 2076 * New options -mstack-protector-guard=global, 2077 -mstack-protector-guard=tls, -mstack-protector-guard-reg=, and 2078 -mstack-protector-guard-offset= change how the stack protector gets 2079 the value to use as canary. 2080 2081 S/390, System z, IBM z Systems, IBM Z 2082 2083 * Support for the IBM z14 processor has been added. When using the 2084 -march=z14 option, the compiler will generate code making use of 2085 the new instructions introduced with the vector enhancement 2086 facility and the miscellaneous instruction extension facility 2. 2087 The -mtune=z14 option enables z14 specific instruction scheduling 2088 without making use of new instructions. 2089 * Builtins for the new vector instructions have been added and can be 2090 enabled using the -mzvector option. 2091 2092 RISC-V 2093 2094 * Support for the RISC-V instruction set has been added. 2095 2096 RX 2097 2098 Basic support for atomic built-in function has been added. It is 2099 currently implemented by flipping interrupts off and on as needed. 2100 2101 SH 2102 2103 * Support for SH5/SH64 has been removed. 2104 * Improved utilization of delay slots on SH2A. 2105 * Improved utilization of zero-displacement conditional branches. 2106 * The following deprecated options have been removed 2107 + -mcbranchdi 2108 + -mcmpeqdi 2109 + -minvalid-symbols 2110 + -msoft-atomic 2111 + -mspace 2112 + -madjust-unroll 2113 * Support for the following SH2A instructions has been added 2114 + mov.b @-Rm,R0 2115 + mov.w @-Rm,R0 2116 + mov.l @-Rm,R0 2117 + mov.b R0,@Rn+ 2118 + mov.w R0,@Rn+ 2119 + mov.l R0,@Rn+ 2120 2121 SPARC 2122 2123 * The SPARC port now uses LRA by default. 2124 * Support for the new Subtract-Extended-with-Carry instruction 2125 available in SPARC M7 (Niagara 7) has been added. 2126 2127 Operating Systems 2128 2129 AIX 2130 2131 * Visibility support has been enabled for AIX 7.1 and above. 2132 2133 Fuchsia 2134 2135 * Support has been added for the [20]Fuchsia OS. 2136 2137 RTEMS 2138 2139 * The ABI changes on ARM so that no short enums are used by default. 2140 2141 Other significant improvements 2142 2143 * -fverbose-asm previously emitted information on the meanings of 2144 assembly expressions. This has been extended so that it now also 2145 prints comments showing the source lines that correspond to the 2146 assembly, making it easier to read the generated assembly 2147 (especially with larger functions). For example, given this C 2148 source file: 2149 2150 int test (int n) 2151 { 2152 int i; 2153 int total = 0; 2154 2155 for (i = 0; i < n; i++) 2156 total += i * i; 2157 return total; 2158 } 2159 2160 -fverbose-asm now gives output similar to this for the function 2161 body (when compiling for x86_64, with -Os): 2162 2163 .text 2164 .globl test 2165 .type test, @@function 2166 test: 2167 .LFB0: 2168 .cfi_startproc 2169 # example.c:4: int total = 0; 2170 xorl %eax, %eax # <retval> 2171 # example.c:6: for (i = 0; i < n; i++) 2172 xorl %edx, %edx # i 2173 .L2: 2174 # example.c:6: for (i = 0; i < n; i++) 2175 cmpl %edi, %edx # n, i 2176 jge .L5 #, 2177 # example.c:7: total += i * i; 2178 movl %edx, %ecx # i, tmp92 2179 imull %edx, %ecx # i, tmp92 2180 # example.c:6: for (i = 0; i < n; i++) 2181 incl %edx # i 2182 # example.c:7: total += i * i; 2183 addl %ecx, %eax # tmp92, <retval> 2184 jmp .L2 # 2185 .L5: 2186 # example.c:10: } 2187 ret 2188 .cfi_endproc 2189 2190 * Two new options have been added for printing fix-it hints: 2191 + -fdiagnostics-parseable-fixits allows for fix-it hints to be 2192 emitted in a machine-readable form, suitable for consumption 2193 by IDEs. For example, given: 2194 2195 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 2196 you mean 'color'? 2197 return ptr->colour; 2198 ^~~~~~ 2199 color 2200 2201 it will emit: 2202 2203 fix-it:"spellcheck-fields.cc":{52:13-52:19}:"color" 2204 2205 + -fdiagnostics-generate-patch will print a patch in "unified" 2206 format after any diagnostics are printed, showing the result 2207 of applying all fix-it hints. For the above example it would 2208 emit: 2209 2210 --- spellcheck-fields.cc 2211 +++ spellcheck-fields.cc 2212 @@ -49,5 +49,5 @@ 2213 2214 color get_color(struct s *ptr) 2215 { 2216 - return ptr->colour; 2217 + return ptr->color; 2218 } 2219 2220 * The gcc and g++ driver programs will now provide suggestions for 2221 misspelled arguments to command-line options. 2222 2223 $ gcc -c test.c -ftls-model=global-dinamic 2224 gcc: error: unknown TLS model 'global-dinamic' 2225 gcc: note: valid arguments to '-ftls-model=' are: global-dynamic initial-exec lo 2226 cal-dynamic local-exec; did you mean 'global-dynamic'? 2227 2228 * The compiler will now provide suggestions for misspelled 2229 parameters. 2230 2231 $ gcc -c test.c --param max-early-inliner-iteration=3 2232 cc1: error: invalid --param name 'max-early-inliner-iteration'; did you mean 'ma 2233 x-early-inliner-iterations'? 2234 2235 * Profile-guided optimization (PGO) instrumentation, as well as test 2236 coverage (GCOV), can newly instrument constructors (functions marks 2237 with __attribute__((constructor))), destructors and C++ 2238 constructors (and destructors) of classes that are used as the type 2239 of a global variable. 2240 * A new option -fprofile-update=atomic prevents creation of corrupted 2241 profiles created during an instrumentation run (-fprofile=generate) 2242 of an application. The downside of the option is a speed penalty. 2243 Providing -pthread on the command line selects atomic profile 2244 updating (when supported by the target). 2245 * GCC's already extensive testsuite has gained some new capabilities, 2246 to further improve the reliability of the compiler: 2247 + GCC now has an internal unit-testing API and a suite of tests 2248 for programmatic self-testing of subsystems. 2249 + GCC's C front end has been extended so that it can parse dumps 2250 of GCC's internal representations, allowing for DejaGnu tests 2251 that more directly exercise specific optimization passes. This 2252 covers both the [21]GIMPLE representation (for testing 2253 higher-level optimizations) and the [22]RTL representation, 2254 allowing for more direct testing of lower-level details, such 2255 as register allocation and instruction selection. 2256 2257 GCC 7.1 2258 2259 This is the [23]list of problem reports (PRs) from GCC's bug tracking 2260 system that are known to be fixed in the 7.1 release. This list might 2261 not be complete (that is, it is possible that some PRs that have been 2262 fixed are not listed here). 2263 2264 GCC 7.2 2265 2266 This is the [24]list of problem reports (PRs) from GCC's bug tracking 2267 system that are known to be fixed in the 7.2 release. This list might 2268 not be complete (that is, it is possible that some PRs that have been 2269 fixed are not listed here). 2270 2271 Target Specific Changes 2272 2273 SPARC 2274 2275 * Support for the SPARC M8 processor has been added. 2276 * The switches -mfix-ut700 and -mfix-gr712rc have been added to work 2277 around an erratum in LEON3FT processors. 2278 * Use of the Floating-point Multiply Single to Double (FsMULd) 2279 instruction can now be controlled by the -mfsmuld and -fno-fsmuld 2280 options. 2281 2282 Operating Systems 2283 2284 RTEMS 2285 2286 * The Ada run-time support uses now thread-local storage (TLS). 2287 * Support for RISC-V has been added. 2288 * Support for 64-bit PowerPC using the ELFv2 ABI with 64-bit long 2289 double has been added. 2290 2291 GCC 7.3 2292 2293 This is the [25]list of problem reports (PRs) from GCC's bug tracking 2294 system that are known to be fixed in the 7.3 release. This list might 2295 not be complete (that is, it is possible that some PRs that have been 2296 fixed are not listed here). 2297 2298 Target Specific Changes 2299 2300 SPARC 2301 2302 * Workarounds for the four [26]LEON3FT errata GRLIB-TN-0010..0013 2303 have been added. Relevant errata are activated by the target 2304 specific -mfix-ut699, -mfix-ut700 and -mfix-gr712rc switches. 2305 2306 Operating Systems 2307 2308 RTEMS 2309 2310 * Support has been added for Epiphany target. 2311 2312 GCC 7.4 2313 2314 This is the [27]list of problem reports (PRs) from GCC's bug tracking 2315 system that are known to be fixed in the 7.4 release. This list might 2316 not be complete (that is, it is possible that some PRs that have been 2317 fixed are not listed here). 2318 2319 2320 For questions related to the use of GCC, please consult these web 2321 pages and the [28]GCC manuals. If that fails, the 2322 [29]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2323 web pages and the development of GCC are welcome on our developer 2324 list at [30]gcc (a] gcc.gnu.org. All of [31]our lists have public 2325 archives. 2326 2327 Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and 2328 distribution of this entire article is permitted in any medium, 2329 provided this notice is preserved. 2330 2331 These pages are [33]maintained by the GCC team. Last modified 2332 2018-12-06[34]. 2333 2334 References 2335 2336 1. http://gcc.gnu.org/gcc-7/porting_to.html 2337 2. http://gcc.gnu.org/onlinedocs/index.html#current 2338 3. https://gcc.gnu.org/wiki/LRAIsDefault 2339 4. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/manual/profile_mode.html 2340 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77728 2341 6. http://www.dwarfstd.org/Download.php 2342 7. https://gcc.gnu.org/wiki/OpenACC 2343 8. https://gcc.gnu.org/wiki/Offloading 2344 9. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/Object-Size-Checking.html 2345 10. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 2346 11. http://wg21.link/p0136 2347 12. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/using_dual_abi.html 2348 13. https://www.openmp.org/specifications/ 2349 14. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/jit/topics/expressions.html#gcc_jit_rvalue_set_bool_require_tail_call 2350 15. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AArch64-Options.html#AArch64-Options 2351 16. https://gcc.gnu.org/install/configure.html 2352 17. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Variable-Attributes.html 2353 18. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Built-in-Functions.html 2354 19. https://gcc.gnu.org/wiki/Offloading 2355 20. https://fuchsia.googlesource.com/ 2356 21. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/GIMPLE-Tests.html 2357 22. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/RTL-Tests.html 2358 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.0 2359 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.2 2360 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.3 2361 26. https://www.gaisler.com/index.php/information/app-tech-notes 2362 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.4 2363 28. https://gcc.gnu.org/onlinedocs/ 2364 29. mailto:gcc-help (a] gcc.gnu.org 2365 30. mailto:gcc (a] gcc.gnu.org 2366 31. https://gcc.gnu.org/lists.html 2367 32. https://www.fsf.org/ 2368 33. https://gcc.gnu.org/about.html 2369 34. http://validator.w3.org/check/referer 2370 ====================================================================== 2371 http://gcc.gnu.org/gcc-6/index.html 2372 GCC 6 Release Series 2373 2374 October 26, 2018 2375 2376 The [1]GNU project and the GCC developers are pleased to announce the 2377 release of GCC 6.5. 2378 2379 This release is a bug-fix release, containing fixes for regressions in 2380 GCC 6.4 relative to previous releases of GCC. 2381 2382 Release History 2383 2384 GCC 6.5 2385 October 26, 2018 ([2]changes, [3]documentation) 2386 2387 GCC 6.4 2388 July 4, 2017 ([4]changes, [5]documentation) 2389 2390 GCC 6.3 2391 December 21, 2016 ([6]changes, [7]documentation) 2392 2393 GCC 6.2 2394 August 22, 2016 ([8]changes, [9]documentation) 2395 2396 GCC 6.1 2397 April 27, 2016 ([10]changes, [11]documentation) 2398 2399 References and Acknowledgements 2400 2401 GCC used to stand for the GNU C Compiler, but since the compiler 2402 supports several other languages aside from C, it now stands for the 2403 GNU Compiler Collection. 2404 2405 A list of [12]successful builds is updated as new information becomes 2406 available. 2407 2408 The GCC developers would like to thank the numerous people that have 2409 contributed new features, improvements, bug fixes, and other changes as 2410 well as test results to GCC. This [13]amazing group of volunteers is 2411 what makes GCC successful. 2412 2413 For additional information about GCC please refer to the [14]GCC 2414 project web site or contact the [15]GCC development mailing list. 2415 2416 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 2417 2418 2419 For questions related to the use of GCC, please consult these web 2420 pages and the [18]GCC manuals. If that fails, the 2421 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2422 web pages and the development of GCC are welcome on our developer 2423 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 2424 archives. 2425 2426 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 2427 distribution of this entire article is permitted in any medium, 2428 provided this notice is preserved. 2429 2430 These pages are [23]maintained by the GCC team. Last modified 2431 2018-10-26[24]. 2432 2433 References 2434 2435 1. http://www.gnu.org/ 2436 2. http://gcc.gnu.org/gcc-6/changes.html 2437 3. http://gcc.gnu.org/onlinedocs/6.5.0/ 2438 4. http://gcc.gnu.org/gcc-6/changes.html 2439 5. http://gcc.gnu.org/onlinedocs/6.4.0/ 2440 6. http://gcc.gnu.org/gcc-6/changes.html 2441 7. http://gcc.gnu.org/onlinedocs/6.3.0/ 2442 8. http://gcc.gnu.org/gcc-6/changes.html 2443 9. http://gcc.gnu.org/onlinedocs/6.2.0/ 2444 10. http://gcc.gnu.org/gcc-6/changes.html 2445 11. http://gcc.gnu.org/onlinedocs/6.1.0/ 2446 12. http://gcc.gnu.org/gcc-6/buildstat.html 2447 13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 2448 14. http://gcc.gnu.org/index.html 2449 15. mailto:gcc (a] gcc.gnu.org 2450 16. http://gcc.gnu.org/mirrors.html 2451 17. http://gcc.gnu.org/svn.html 2452 18. https://gcc.gnu.org/onlinedocs/ 2453 19. mailto:gcc-help (a] gcc.gnu.org 2454 20. mailto:gcc (a] gcc.gnu.org 2455 21. https://gcc.gnu.org/lists.html 2456 22. https://www.fsf.org/ 2457 23. https://gcc.gnu.org/about.html 2458 24. http://validator.w3.org/check/referer 2459 ====================================================================== 2460 http://gcc.gnu.org/gcc-6/changes.html 2461 GCC 6 Release Series 2462 Changes, New Features, and Fixes 2463 2464 This page is a brief summary of some of the huge number of improvements 2465 in GCC 6. For more information, see the [1]Porting to GCC 6 page and 2466 the [2]full GCC documentation. 2467 2468 Caveats 2469 2470 * The default mode for C++ is now -std=gnu++14 instead of 2471 -std=gnu++98. 2472 * Support for a number of older systems and recently unmaintained or 2473 untested target ports of GCC has been declared obsolete in GCC 6. 2474 Unless there is activity to revive them, the next release of GCC 2475 will have their sources permanently removed. 2476 The following ports for individual systems on particular 2477 architectures have been obsoleted: 2478 + SH5 / SH64 (sh64-*-*) as announced [3]here. 2479 * The AVR port requires binutils version 2.26.1 or later for the fix 2480 for [4]PR71151 to work. 2481 * The GCC 6.5 release has an accidental ABI incompatibility for 2482 nested std::pair objects, for more details see [5]PR 87822. The bug 2483 causes a layout change for pairs where the first member is also a 2484 pair, e.g. std::pair<std::pair<X, Y>, Z>. The GCC 6 release series 2485 is closed so the bug in GCC 6.5 will not be fixed upstream, but 2486 there is a patch in the bug report to allow it to be fixed by 2487 anybody packaging GCC 6.5 or installing it themselves. 2488 2489 General Optimizer Improvements 2490 2491 * UndefinedBehaviorSanitizer gained a new sanitization option, 2492 -fsanitize=bounds-strict, which enables strict checking of array 2493 bounds. In particular, it enables -fsanitize=bounds as well as 2494 instrumentation of flexible array member-like arrays. 2495 * Type-based alias analysis now disambiguates accesses to different 2496 pointers. This improves precision of the alias oracle by about 2497 20-30% on higher-level C++ programs. Programs doing invalid type 2498 punning of pointer types may now need -fno-strict-aliasing to work 2499 correctly. 2500 * Alias analysis now correctly supports the weakref and alias 2501 attributes. This allows accessing both a variable and its alias in 2502 one translation unit which is common with link-time optimization. 2503 * Value range propagation now assumes that the this pointer in C++ 2504 member functions is non-null. This eliminates common null pointer 2505 checks but also breaks some non-conforming code-bases (such as 2506 Qt-5, Chromium, KDevelop). As a temporary work-around 2507 -fno-delete-null-pointer-checks can be used. Wrong code can be 2508 identified by using -fsanitize=undefined. 2509 * Link-time optimization improvements: 2510 + warning and error attributes are now correctly preserved by 2511 declaration linking and thus -D_FORTIFY_SOURCE=2 is now 2512 supported with -flto. 2513 + Type merging was fixed to handle C and Fortran 2514 interoperability rules as defined by the Fortran 2008 language 2515 standard. 2516 As an exception, CHARACTER(KIND=C_CHAR) is not inter-operable 2517 with char in all cases because it is an array while char is 2518 scalar. INTEGER(KIND=C_SIGNED_CHAR) should be used instead. In 2519 general, this inter-operability cannot be implemented, for 2520 example on targets where the argument passing convention for 2521 arrays differs from scalars. 2522 + More type information is now preserved at link time, reducing 2523 the loss of accuracy of the type-based alias analysis compared 2524 to builds without link-time optimization. 2525 + Invalid type punning on global variables and declarations is 2526 now reported with -Wodr-type-mismatch. 2527 + The size of LTO object files was reduced by about 11% 2528 (measured by compiling Firefox 46.0). 2529 + Link-time parallelization (enabled using -flto=n) was 2530 significantly improved by decreasing the size of streamed data 2531 when partitioning programs. The size of streamed IL while 2532 compiling Firefox 46.0 was reduced by 66%. 2533 + The linker plugin was extended to pass information about the 2534 type of binary produced to the GCC back end. (That can also be 2535 controlled manually by -flinker-output.) This makes it 2536 possible to properly configure the code generator and support 2537 incremental linking. Incremental linking of LTO objects by gcc 2538 -r is now supported for plugin-enabled setups. 2539 There are two ways to perform incremental linking: 2540 1. Linking by ld -r will result in an object file with all 2541 sections from individual object files mechanically 2542 merged. This delays the actual link-time optimization to 2543 the final linking step and thus permits whole program 2544 optimization. Linking the final binary with such object 2545 files is however slower. 2546 2. Linking by gcc -r will lead to link-time optimization and 2547 emit the final binary into the object file. Linking such 2548 an object file is fast but avoids any benefits from whole 2549 program optimization. 2550 GCC 7 will support incremental link-time optimization with gcc 2551 -r. 2552 * Inter-procedural optimization improvements: 2553 + Basic jump threading is now performed before profile 2554 construction and inline analysis, resulting in more realistic 2555 size and time estimates that drive the heuristics of the 2556 inliner and function cloning passes. 2557 + Function cloning now more aggressively eliminates unused 2558 function parameters. 2559 2560 New Languages and Language specific improvements 2561 2562 Compared to GCC 5, the GCC 6 release series includes a much improved 2563 implementation of the [6]OpenACC 2.0a specification. Highlights are: 2564 * In addition to single-threaded host-fallback execution, offloading 2565 is supported for nvptx (Nvidia GPUs) on x86_64 and PowerPC 64-bit 2566 little-endian GNU/Linux host systems. For nvptx offloading, with 2567 the OpenACC parallel construct, the execution model allows for an 2568 arbitrary number of gangs, up to 32 workers, and 32 vectors. 2569 * Initial support for parallelized execution of OpenACC kernels 2570 constructs: 2571 + Parallelization of a kernels region is switched on by 2572 -fopenacc combined with -O2 or higher. 2573 + Code is offloaded onto multiple gangs, but executes with just 2574 one worker, and a vector length of 1. 2575 + Directives inside a kernels region are not supported. 2576 + Loops with reductions can be parallelized. 2577 + Only kernels regions with one loop nest are parallelized. 2578 + Only the outer-most loop of a loop nest can be parallelized. 2579 + Loop nests containing sibling loops are not parallelized. 2580 Typically, using the OpenACC parallel construct gives much better 2581 performance, compared to the initial support of the OpenACC kernels 2582 construct. 2583 * The device_type clause is not supported. The bind and nohost 2584 clauses are not supported. The host_data directive is not supported 2585 in Fortran. 2586 * Nested parallelism (cf. CUDA dynamic parallelism) is not supported. 2587 * Usage of OpenACC constructs inside multithreaded contexts (such as 2588 created by OpenMP, or pthread programming) is not supported. 2589 * If a call to the acc_on_device function has a compile-time constant 2590 argument, the function call evaluates to a compile-time constant 2591 value only for C and C++ but not for Fortran. 2592 2593 See the [7]OpenACC and [8]Offloading wiki pages for further 2594 information. 2595 2596 C family 2597 2598 * Version 4.5 of the [9]OpenMP specification is now supported in the 2599 C and C++ compilers. 2600 * The C and C++ compilers now support attributes on enumerators. For 2601 instance, it is now possible to mark enumerators as deprecated: 2602 2603 enum { 2604 newval, 2605 oldval __attribute__ ((deprecated ("too old"))) 2606 }; 2607 2608 * Source locations for the C and C++ compilers are now tracked as 2609 ranges, rather than just points, making it easier to identify the 2610 subexpression of interest within a complicated expression. For 2611 example: 2612 2613 test.cc: In function 'int test(int, int, foo, int, int)': 2614 test.cc:5:16: error: no match for 'operator*' (operand types are 'int' and 'foo' 2615 ) 2616 return p + q * r * s + t; 2617 ~~^~~ 2618 2619 In addition, there is now initial support for precise diagnostic 2620 locations within strings: 2621 2622 format-strings.c:3:14: warning: field width specifier '*' expects a matching 'in 2623 t' argument [-Wformat=] 2624 printf("%*d"); 2625 ^ 2626 2627 * Diagnostics can now contain "fix-it hints", which are displayed in 2628 context underneath the relevant source code. For example: 2629 2630 fixits.c: In function 'bad_deref': 2631 fixits.c:11:13: error: 'ptr' is a pointer; did you mean to use '->'? 2632 return ptr.x; 2633 ^ 2634 -> 2635 2636 * The C and C++ compilers now offer suggestions for misspelled field 2637 names: 2638 2639 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 2640 you mean 'color'? 2641 return ptr->colour; 2642 ^~~~~~ 2643 2644 * New command-line options have been added for the C and C++ 2645 compilers: 2646 + -Wshift-negative-value warns about left shifting a negative 2647 value. 2648 + -Wshift-overflow warns about left shift overflows. This 2649 warning is enabled by default. -Wshift-overflow=2 also warns 2650 about left-shifting 1 into the sign bit. 2651 + -Wtautological-compare warns if a self-comparison always 2652 evaluates to true or false. This warning is enabled by -Wall. 2653 + -Wnull-dereference warns if the compiler detects paths that 2654 trigger erroneous or undefined behavior due to dereferencing a 2655 null pointer. This option is only active when 2656 -fdelete-null-pointer-checks is active, which is enabled by 2657 optimizations in most targets. The precision of the warnings 2658 depends on the optimization options used. 2659 + -Wduplicated-cond warns about duplicated conditions in an 2660 if-else-if chain. 2661 + -Wmisleading-indentation warns about places where the 2662 indentation of the code gives a misleading idea of the block 2663 structure of the code to a human reader. For example, given 2664 [10]CVE-2014-1266: 2665 2666 sslKeyExchange.c: In function 'SSLVerifySignedServerKeyExchange': 2667 sslKeyExchange.c:629:3: warning: this 'if' clause does not guard... [-Wmisleadin 2668 g-indentation] 2669 if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0) 2670 ^~ 2671 sslKeyExchange.c:631:5: note: ...this statement, but the latter is misleadingly 2672 indented as if it is guarded by the 'if' 2673 goto fail; 2674 ^~~~ 2675 2676 This warning is enabled by -Wall. 2677 * The C and C++ compilers now emit saner error messages if 2678 merge-conflict markers are present in a source file. 2679 2680 test.c:3:1: error: version control conflict marker in file 2681 <<<<<<< HEAD 2682 ^~~~~~~ 2683 2684 C 2685 2686 * It is possible to disable warnings when an initialized field of a 2687 structure or a union with side effects is being overridden when 2688 using designated initializers via a new warning option 2689 -Woverride-init-side-effects. 2690 * A new type attribute scalar_storage_order applying to structures 2691 and unions has been introduced. It specifies the storage order (aka 2692 endianness) in memory of scalar fields in structures or unions. 2693 2694 C++ 2695 2696 * The default mode has been changed to -std=gnu++14. 2697 * [11]C++ Concepts are now supported when compiling with -fconcepts. 2698 * -flifetime-dse is more aggressive in dead-store elimination in 2699 situations where a memory store to a location precedes a 2700 constructor to that memory location. 2701 * G++ now supports [12]C++17 fold expressions, u8 character literals, 2702 extended static_assert, and nested namespace definitions. 2703 * G++ now allows constant evaluation for all non-type template 2704 arguments. 2705 * G++ now supports C++ Transactional Memory when compiling with 2706 -fgnu-tm. 2707 2708 Runtime Library (libstdc++) 2709 2710 * Extensions to the C++ Library to support mathematical special 2711 functions (ISO/IEC 29124:2010), thanks to Edward Smith-Rowland. 2712 * Experimental support for C++17, including the following new 2713 features: 2714 + std::uncaught_exceptions function (this is also available for 2715 -std=gnu++NN modes); 2716 + new member functions try_emplace and insert_or_assign for 2717 unique_key maps; 2718 + non-member functions std::size, std::empty, and std::data for 2719 accessing containers and arrays; 2720 + std::invoke; 2721 + std::shared_mutex; 2722 + std::void_t and std::bool_constant metaprogramming utilities. 2723 Thanks to Ville Voutilainen for contributing many of the C++17 2724 features. 2725 * An experimental implementation of the File System TS. 2726 * Experimental support for most features of the second version of the 2727 Library Fundamentals TS. This includes polymorphic memory resources 2728 and array support in shared_ptr, thanks to Fan You. 2729 * Some assertions checked by Debug Mode can now also be enabled by 2730 _GLIBCXX_ASSERTIONS. The subset of checks enabled by the new macro 2731 have less run-time overhead than the full _GLIBCXX_DEBUG checks and 2732 don't affect the library ABI, so can be enabled per-translation 2733 unit. 2734 * Timed mutex types are supported on more targets, including Darwin. 2735 * Improved std::locale support for DragonFly and FreeBSD, thanks to 2736 John Marino and Andreas Tobler. 2737 2738 Fortran 2739 2740 * Fortran 2008 SUBMODULE support. 2741 * Fortran 2015 EVENT_TYPE, EVENT_POST, EVENT_WAIT, and EVENT_QUERY 2742 support. 2743 * Improved support for Fortran 2003 deferred-length character 2744 variables. 2745 * Improved support for OpenMP and OpenACC. 2746 * The MATMUL intrinsic is now inlined for straightforward cases if 2747 front-end optimization is active. The maximum size for inlining can 2748 be set to n with the -finline-matmul-limit=n option and turned off 2749 with -finline-matmul-limit=0. 2750 * The -Wconversion-extra option will warn about REAL constants which 2751 have excess precision for their kind. 2752 * The -Winteger-division option has been added, which warns about 2753 divisions of integer constants which are truncated. This option is 2754 included in -Wall by default. 2755 2756 libgccjit 2757 2758 * The driver code is now run in-process within libgccjit, providing a 2759 small speed-up of the compilation process. 2760 * The API has gained entrypoints for 2761 + [13]timing how long was spent in different parts of code, 2762 + [14]creating switch statements, 2763 + [15]allowing unreachable basic blocks in a function, and 2764 + [16]adding arbitrary command-line options to a compilation. 2765 2766 New Targets and Target Specific Improvements 2767 2768 AArch64 2769 2770 * A number of AArch64-specific options have been added. The most 2771 important ones are summarised in this section; for more detailed 2772 information please refer to the documentation. 2773 * The command-line options -march=native, -mcpu=native and 2774 -mtune=native are now available on native AArch64 GNU/Linux 2775 systems. Specifying these options causes GCC to auto-detect the 2776 host CPU and choose the optimal setting for that system. 2777 * -fpic is now supported when generating code for the small code 2778 model (-mcmodel=small). The size of the global offset table (GOT) 2779 is limited to 28KiB under the LP64 SysV ABI, and 15KiB under the 2780 ILP32 SysV ABI. 2781 * The AArch64 port now supports target attributes and pragmas. Please 2782 refer to the [17]documentation for details of available attributes 2783 and pragmas as well as usage instructions. 2784 * Link-time optimization across translation units with different 2785 target-specific options is now supported. 2786 * The option -mtls-size= is now supported. It can be used to specify 2787 the bit size of TLS offsets, allowing GCC to generate better TLS 2788 instruction sequences. 2789 * The option -fno-plt is now fully functional. 2790 * The ARMv8.1-A architecture and the Large System Extensions are now 2791 supported. They can be used by specifying the -march=armv8.1-a 2792 option. Additionally, the +lse option extension can be used in a 2793 similar fashion to other option extensions. The Large System 2794 Extensions introduce new instructions that are used in the 2795 implementation of atomic operations. 2796 * The ACLE half-precision floating-point type __fp16 is now supported 2797 in the C and C++ languages. 2798 * The ARM Cortex-A35 processor is now supported via the 2799 -mcpu=cortex-a35 and -mtune=cortex-a35 options as well as the 2800 equivalent target attributes and pragmas. 2801 * The Qualcomm QDF24xx processor is now supported via the 2802 -mcpu=qdf24xx and -mtune=qdf24xx options as well as the equivalent 2803 target attributes and pragmas. 2804 * Code generation for the ARM Cortex-A57 processor is improved. Among 2805 general code generation improvements, a better algorithm is added 2806 for allocating registers to floating-point multiply-accumulate 2807 instructions offering increased performance when compiling with 2808 -mcpu=cortex-a57 or -mtune=cortex-a57. 2809 * Code generation for the ARM Cortex-A53 processor is improved. A 2810 more accurate instruction scheduling model for the processor is now 2811 used, and a number of compiler tuning parameters have been set to 2812 offer increased performance when compiling with -mcpu=cortex-a53 or 2813 -mtune=cortex-a53. 2814 * Code generation for the Samsung Exynos M1 processor is improved. A 2815 more accurate instruction scheduling model for the processor is now 2816 used, and a number of compiler tuning parameters have been set to 2817 offer increased performance when compiling with -mcpu=exynos-m1 or 2818 -mtune=exynos-m1. 2819 * Improvements in the generation of conditional branches and literal 2820 pools allow the compiler to compile functions of a large size. 2821 Constant pools are now placed into separate rodata sections. The 2822 new option -mpc-relative-literal-loads generates per-function 2823 literal pools, limiting the maximum size of functions to 1MiB. 2824 * Several correctness issues generating Advanced SIMD instructions 2825 for big-endian targets have been fixed resulting in improved code 2826 generation for ACLE intrinsics with -mbig-endian. 2827 2828 ARM 2829 2830 * Support for revisions of the ARM architecture prior to ARMv4t has 2831 been deprecated and will be removed in a future GCC release. The 2832 -mcpu and -mtune values that are deprecated are: arm2, arm250, 2833 arm3, arm6, arm60, arm600, arm610, arm620, arm7, arm7d, arm7di, 2834 arm70, arm700, arm700i, arm710, arm720, arm710c, arm7100, arm7500, 2835 arm7500fe, arm7m, arm7dm, arm7dmi, arm8, arm810, strongarm, 2836 strongarm110, strongarm1100, strongarm1110, fa526, fa626. The value 2837 arm7tdmi is still supported. The values of -march that are 2838 deprecated are: armv2,armv2a,armv3,armv3m,armv4. 2839 * The ARM port now supports target attributes and pragmas. Please 2840 refer to the [18]documentation for details of available attributes 2841 and pragmas as well as usage instructions. 2842 * Support has been added for the following processors (GCC 2843 identifiers in parentheses): ARM Cortex-A32 (cortex-a32), ARM 2844 Cortex-A35 (cortex-a35) and ARM Cortex-R8 (cortex-r8). The GCC 2845 identifiers can be used as arguments to the -mcpu or -mtune 2846 options, for example: -mcpu=cortex-a32 or -mtune=cortex-a35. 2847 2848 Heterogeneous Systems Architecture 2849 2850 * GCC can now generate HSAIL (Heterogeneous System Architecture 2851 Intermediate Language) for simple OpenMP device constructs if 2852 configured with --enable-offload-targets=hsa. A new libgomp plugin 2853 then runs the HSA GPU kernels implementing these constructs on HSA 2854 capable GPUs via a standard HSA run time. 2855 If the HSA compilation back end determines it cannot output HSAIL 2856 for a particular input, it gives a warning by default. These 2857 warnings can be suppressed with -Wno-hsa. To give a few examples, 2858 the HSA back end does not implement compilation of code using 2859 function pointers, automatic allocation of variable sized arrays, 2860 functions with variadic arguments as well as a number of other less 2861 common programming constructs. 2862 When compilation for HSA is enabled, the compiler attempts to 2863 compile composite OpenMP constructs 2864 2865 #pragma omp target teams distribute parallel for 2866 2867 into parallel HSA GPU kernels. 2868 2869 IA-32/x86-64 2870 2871 * GCC now supports the Intel CPU named Skylake with AVX-512 2872 extensions through -march=skylake-avx512. The switch enables the 2873 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 2874 AVX-512DQ. 2875 * Support for new AMD instructions monitorx and mwaitx has been 2876 added. This includes new intrinsic and built-in support. It is 2877 enabled through option -mmwaitx. The instructions monitorx and 2878 mwaitx implement the same functionality as the old monitor and 2879 mwait instructions. In addition mwaitx adds a configurable timer. 2880 The timer value is received as third argument and stored in 2881 register %ebx. 2882 * x86-64 targets now allow stack realignment from a word-aligned 2883 stack pointer using the command-line option -mstackrealign or 2884 __attribute__ ((force_align_arg_pointer)). This allows functions 2885 compiled with a vector-aligned stack to be invoked from objects 2886 that keep only word-alignment. 2887 * Support for address spaces __seg_fs, __seg_gs, and __seg_tls. These 2888 can be used to access data via the %fs and %gs segments without 2889 having to resort to inline assembly. Please refer to the 2890 [19]documentation for usage instructions. 2891 * Support for AMD Zen (family 17h) processors is now available 2892 through the -march=znver1 and -mtune=znver1 options. 2893 2894 MeP 2895 2896 * Support for the MeP (mep-elf) architecture has been deprecated and 2897 will be removed in a future GCC release. 2898 2899 MSP430 2900 2901 * The MSP430 compiler now has the ability to automatically distribute 2902 code and data between low memory (addresses below 64K) and high 2903 memory. This only applies to parts that actually have both memory 2904 regions and only if the linker script for the part has been 2905 specifically set up to support this feature. 2906 A new attribute of either can be applied to both functions and 2907 data, and this tells the compiler to place the object into low 2908 memory if there is room and into high memory otherwise. Two other 2909 new attributes - lower and upper - can be used to explicitly state 2910 that an object should be placed in the specified memory region. If 2911 there is not enough left in that region the compilation will fail. 2912 Two new command-line options - -mcode-region=[lower|upper|either] 2913 and -mdata-region=[lower|upper|either] - can be used to tell the 2914 compiler what to do with objects that do not have one of these new 2915 attributes. 2916 2917 PowerPC / PowerPC64 / RS6000 2918 2919 * PowerPC64 now supports IEEE 128-bit floating-point using the 2920 __float128 data type. In GCC 6, this is not enabled by default, but 2921 you can enable it with -mfloat128. The IEEE 128-bit floating-point 2922 support requires the use of the VSX instruction set. IEEE 128-bit 2923 floating-point values are passed and returned as a single vector 2924 value. The software emulator for IEEE 128-bit floating-point 2925 support is only built on PowerPC GNU/Linux systems where the 2926 default CPU is at least power7. On future ISA 3.0 systems (POWER 9 2927 and later), you will be able to use the -mfloat128-hardware option 2928 to use the ISA 3.0 instructions that support IEEE 128-bit 2929 floating-point. An additional type (__ibm128) has been added to 2930 refer to the IBM extended double type that normally implements long 2931 double. This will allow for a future transition to implementing 2932 long double with IEEE 128-bit floating-point. 2933 * Basic support has been added for POWER9 hardware that will use the 2934 recently published OpenPOWER ISA 3.0 instructions. The following 2935 new switches are available: 2936 + -mcpu=power9: Implement all of the ISA 3.0 instructions 2937 supported by the compiler. 2938 + -mtune=power9: In the future, apply tuning for POWER9 systems. 2939 Currently, POWER8 tunings are used. 2940 + -mmodulo: Generate code using the ISA 3.0 integer instructions 2941 (modulus, count trailing zeros, array index support, integer 2942 multiply/add). 2943 + -mpower9-fusion: Generate code to suitably fuse instruction 2944 sequences for a POWER9 system. 2945 + -mpower9-dform: Generate code to use the new D-form 2946 (register+offset) memory instructions for the vector 2947 registers. 2948 + -mpower9-vector: Generate code using the new ISA 3.0 vector 2949 (VSX or Altivec) instructions. 2950 + -mpower9-minmax: Reserved for future development. 2951 + -mtoc-fusion: Keep TOC entries together to provide more fusion 2952 opportunities. 2953 * New constraints have been added to support IEEE 128-bit 2954 floating-point and ISA 3.0 instructions: 2955 + wb: Altivec register if -mpower9-dform is enabled. 2956 + we: VSX register if -mpower9-vector is enabled for 64-bit code 2957 generation. 2958 + wo: VSX register if -mpower9-vector is enabled. 2959 + wp: Reserved for future use if long double is implemented with 2960 IEEE 128-bit floating-point instead of IBM extended double. 2961 + wq: VSX register if -mfloat128 is enabled. 2962 + wF: Memory operand suitable for POWER9 fusion load/store. 2963 + wG: Memory operand suitable for TOC fusion memory references. 2964 + wL: Integer constant identifying the element number mfvsrld 2965 accesses within a vector. 2966 * Support has been added for __builtin_cpu_is() and 2967 __builtin_cpu_supports(), allowing for very fast access to 2968 AT_PLATFORM, AT_HWCAP, and AT_HWCAP2 values. This requires use of 2969 glibc 2.23 or later. 2970 * All hardware transactional memory builtins now correctly behave as 2971 memory barriers. Programmers can use #ifdef __TM_FENCE__ to 2972 determine whether their "old" compiler treats the builtins as 2973 barriers. 2974 * Split-stack support has been added for gccgo on PowerPC64 for both 2975 big- and little-endian (but not for 32-bit). The gold linker from 2976 at least binutils 2.25.1 must be available in the PATH when 2977 configuring and building gccgo to enable split stack. (The 2978 requirement for binutils 2.25.1 applies to PowerPC64 only.) The 2979 split-stack feature allows a small initial stack size to be 2980 allocated for each goroutine, which increases as needed. 2981 * GCC on PowerPC now supports the standard lround function. 2982 * A new configuration option ---with-advance-toolchain=at was added 2983 for PowerPC 64-bit GNU/Linux systems to use the header files, 2984 library files, and the dynamic linker from a specific Advance 2985 Toolchain release instead of the default versions that are provided 2986 by the GNU/Linux distribution. In general, this option is intended 2987 for the developers of GCC, and it is not intended for general use. 2988 * The "q", "S", "T", and "t" asm-constraints have been removed. 2989 * The "b", "B", "m", "M", and "W" format modifiers have been removed. 2990 2991 S/390, System z, IBM z Systems 2992 2993 * Support for the IBM z13 processor has been added. When using the 2994 -march=z13 option, the compiler will generate code making use of 2995 the new instructions and registers introduced with the vector 2996 extension facility. The -mtune=z13 option enables z13 specific 2997 instruction scheduling without making use of new instructions. 2998 Compiling code with -march=z13 reduces the default alignment of 2999 vector types bigger than 8 bytes to 8. This is an ABI change and 3000 care must be taken when linking modules compiled with different 3001 arch levels which interchange variables containing vector type 3002 values. For newly compiled code the GNU linker will emit a warning. 3003 * The -mzvector option enables a C/C++ language extension. This 3004 extension provides a new keyword vector which can be used to define 3005 vector type variables. (Note: This is not available when enforcing 3006 strict standard compliance e.g. with -std=c99. Either enable GNU 3007 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 3008 Additionally a set of overloaded builtins is provided which is 3009 partially compatible to the PowerPC Altivec builtins. In order to 3010 make use of these builtins the vecintrin.h header file needs to be 3011 included. 3012 * The new command-line options -march=native, and -mtune=native are 3013 now available on native IBM z Systems. Specifying these options 3014 causes GCC to auto-detect the host CPU and choose the optimal 3015 setting for that system. If GCC is unable to detect the host CPU 3016 these options have no effect. 3017 * The IBM z Systems port now supports target attributes and pragmas. 3018 Please refer to the [20]documentation for details of available 3019 attributes and pragmas as well as usage instructions. 3020 * -fsplit-stack is now supported as part of the IBM z Systems port. 3021 This feature requires a recent gold linker to be used. 3022 * Support for the g5 and g6 -march=/-mtune= CPU level switches has 3023 been deprecated and will be removed in a future GCC release. -m31 3024 from now on defaults to -march=z900 if not specified otherwise. 3025 -march=native on a g5/g6 machine will default to -march=z900. 3026 3027 SH 3028 3029 * Support for SH5 / SH64 has been declared obsolete and will be 3030 removed in future releases. 3031 * Support for the FDPIC ABI has been added. It can be enabled using 3032 the new -mfdpic target option and --enable-fdpic configure option. 3033 3034 SPARC 3035 3036 * An ABI bug has been fixed in 64-bit mode. Unfortunately, this 3037 change will break binary compatibility with earlier releases for 3038 code it affects, but this should be pretty rare in practice. The 3039 conditions are: a 16-byte structure containing a double or a 8-byte 3040 vector in the second half is passed to a subprogram in slot #15, 3041 for example as 16th parameter if the first 15 ones have at most 8 3042 bytes. The double or vector was wrongly passed in floating-point 3043 register %d32 in lieu of on the stack as per the SPARC calling 3044 conventions. 3045 3046 Operating Systems 3047 3048 AIX 3049 3050 * DWARF debugging support for AIX 7.1 has been enabled as an optional 3051 debugging format. A more recent Technology Level (TL) and GCC built 3052 with that level are required for full exploitation of DWARF 3053 debugging capabilities. 3054 3055 Linux 3056 3057 * Support for the [21]musl C library was added for the AArch64, ARM, 3058 MicroBlaze, MIPS, MIPS64, PowerPC, PowerPC64, SH, i386, x32 and 3059 x86_64 targets. It can be selected using the new -mmusl option in 3060 case musl is not the default libc. GCC defaults to musl libc if it 3061 is built with a target triplet matching the *-linux-musl* pattern. 3062 3063 RTEMS 3064 3065 * The RTEMS thread model implementation changed. Mutexes now use 3066 self-contained objects defined in Newlib <sys/lock.h> instead of 3067 Classic API semaphores. The keys for thread specific data and the 3068 once function are directly defined via <pthread.h>. Self-contained 3069 condition variables are provided via Newlib <sys/lock.h>. The RTEMS 3070 thread model also supports C++11 threads. 3071 * OpenMP support now uses self-contained objects provided by Newlib 3072 <sys/lock.h> and offers a significantly better performance compared 3073 to the POSIX configuration of libgomp. It is possible to configure 3074 thread pools for each scheduler instance via the environment 3075 variable GOMP_RTEMS_THREAD_POOLS. 3076 3077 Solaris 3078 3079 * Solaris 12 is now fully supported. Minimal support had already been 3080 present in GCC 5.3. 3081 * Solaris 12 provides a full set of startup files (crt1.o, crti.o, 3082 crtn.o), which GCC now prefers over its own ones. 3083 * Position independent executables (PIE) are now supported on Solaris 3084 12. 3085 * Constructor priority is now supported on Solaris 12 with the system 3086 linker. 3087 * libvtv has been ported to Solaris 11 and up. 3088 3089 Windows 3090 3091 * The option -mstackrealign is now automatically activated in 32-bit 3092 mode whenever the use of SSE instructions is requested. 3093 3094 Other significant improvements 3095 3096 * The gcc and g++ driver programs will now provide suggestions for 3097 misspelled command-line options. 3098 3099 $ gcc -static-libfortran test.f95 3100 gcc: error: unrecognized command line option '-static-libfortran'; did you mean 3101 '-static-libgfortran'? 3102 3103 * The --enable-default-pie configure option enables generation of PIE 3104 by default. 3105 3106 GCC 6.2 3107 3108 This is the [22]list of problem reports (PRs) from GCC's bug tracking 3109 system that are known to be fixed in the 6.2 release. This list might 3110 not be complete (that is, it is possible that some PRs that have been 3111 fixed are not listed here). 3112 3113 Target Specific Changes 3114 3115 SPARC 3116 3117 * Support for --with-cpu-32 and --with-cpu-64 configure options has 3118 been added on bi-architecture platforms. 3119 * Support for the SPARC M7 (Niagara 7) processor has been added. 3120 * Support for the VIS 4.0 instruction set has been added. 3121 3122 GCC 6.3 3123 3124 This is the [23]list of problem reports (PRs) from GCC's bug tracking 3125 system that are known to be fixed in the 6.3 release. This list might 3126 not be complete (that is, it is possible that some PRs that have been 3127 fixed are not listed here). 3128 3129 Target Specific Changes 3130 3131 IA-32/x86-64 3132 3133 * Support for the [24]deprecated pcommit instruction has been 3134 removed. 3135 3136 GCC 6.4 3137 3138 This is the [25]list of problem reports (PRs) from GCC's bug tracking 3139 system that are known to be fixed in the 6.4 release. This list might 3140 not be complete (that is, it is possible that some PRs that have been 3141 fixed are not listed here). 3142 3143 Operating Systems 3144 3145 RTEMS 3146 3147 * The ABI changes on ARM so that no short enums are used by default. 3148 3149 GCC 6.5 3150 3151 This is the [26]list of problem reports (PRs) from GCC's bug tracking 3152 system that are known to be fixed in the 6.5 release. This list might 3153 not be complete (that is, it is possible that some PRs that have been 3154 fixed are not listed here). 3155 3156 3157 For questions related to the use of GCC, please consult these web 3158 pages and the [27]GCC manuals. If that fails, the 3159 [28]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3160 web pages and the development of GCC are welcome on our developer 3161 list at [29]gcc (a] gcc.gnu.org. All of [30]our lists have public 3162 archives. 3163 3164 Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and 3165 distribution of this entire article is permitted in any medium, 3166 provided this notice is preserved. 3167 3168 These pages are [32]maintained by the GCC team. Last modified 3169 2018-10-31[33]. 3170 3171 References 3172 3173 1. http://gcc.gnu.org/gcc-6/porting_to.html 3174 2. http://gcc.gnu.org/onlinedocs/index.html#current 3175 3. https://gcc.gnu.org/ml/gcc/2015-08/msg00101.html 3176 4. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71151 3177 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87822 3178 6. https://www.openacc.org/ 3179 7. https://gcc.gnu.org/wiki/OpenACC 3180 8. https://gcc.gnu.org/wiki/Offloading 3181 9. https://www.openmp.org/specifications/ 3182 10. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2014-1266 3183 11. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4377.pdf 3184 12. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 3185 13. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/performance.html 3186 14. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/functions.html#gcc_jit_block_end_with_switch 3187 15. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_set_bool_allow_unreachable_blocks 3188 16. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_add_command_line_option 3189 17. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/AArch64-Function-Attributes.html#AArch64-Function-Attributes 3190 18. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/ARM-Function-Attributes.html#ARM-Function-Attributes 3191 19. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/Named-Address-Spaces.html#Named-Address-Spaces 3192 20. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/S_002f390-Function-Attributes.html#S_002f390-Function-Attributes 3193 21. http://www.musl-libc.org/ 3194 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.2 3195 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.3 3196 24. https://software.intel.com/en-us/blogs/2016/09/12/deprecate-pcommit-instruction 3197 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.4 3198 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.5 3199 27. https://gcc.gnu.org/onlinedocs/ 3200 28. mailto:gcc-help (a] gcc.gnu.org 3201 29. mailto:gcc (a] gcc.gnu.org 3202 30. https://gcc.gnu.org/lists.html 3203 31. https://www.fsf.org/ 3204 32. https://gcc.gnu.org/about.html 3205 33. http://validator.w3.org/check/referer 3206 ====================================================================== 3207 http://gcc.gnu.org/gcc-5/index.html 3208 GCC 5 Release Series 3209 3210 October 10, 2017 3211 3212 The [1]GNU project and the GCC developers are pleased to announce the 3213 release of GCC 5.5. 3214 3215 This release is a bug-fix release, containing fixes for regressions in 3216 GCC 5.4 relative to previous releases of GCC. 3217 3218 Release History 3219 3220 GCC 5.5 3221 October 10, 2017 ([2]changes, [3]documentation) 3222 3223 GCC 5.4 3224 June 3, 2016 ([4]changes, [5]documentation) 3225 3226 GCC 5.3 3227 December 4, 2015 ([6]changes, [7]documentation) 3228 3229 GCC 5.2 3230 July 16, 2015 ([8]changes, [9]documentation) 3231 3232 GCC 5.1 3233 April 22, 2015 ([10]changes, [11]documentation) 3234 3235 References and Acknowledgements 3236 3237 GCC used to stand for the GNU C Compiler, but since the compiler 3238 supports several other languages aside from C, it now stands for the 3239 GNU Compiler Collection. 3240 3241 A list of [12]successful builds is updated as new information becomes 3242 available. 3243 3244 The GCC developers would like to thank the numerous people that have 3245 contributed new features, improvements, bug fixes, and other changes as 3246 well as test results to GCC. This [13]amazing group of volunteers is 3247 what makes GCC successful. 3248 3249 For additional information about GCC please refer to the [14]GCC 3250 project web site or contact the [15]GCC development mailing list. 3251 3252 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 3253 3254 3255 For questions related to the use of GCC, please consult these web 3256 pages and the [18]GCC manuals. If that fails, the 3257 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3258 web pages and the development of GCC are welcome on our developer 3259 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 3260 archives. 3261 3262 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 3263 distribution of this entire article is permitted in any medium, 3264 provided this notice is preserved. 3265 3266 These pages are [23]maintained by the GCC team. Last modified 3267 2018-09-30[24]. 3268 3269 References 3270 3271 1. http://www.gnu.org/ 3272 2. http://gcc.gnu.org/gcc-5/changes.html 3273 3. http://gcc.gnu.org/onlinedocs/5.5.0/ 3274 4. http://gcc.gnu.org/gcc-5/changes.html 3275 5. http://gcc.gnu.org/onlinedocs/5.4.0/ 3276 6. http://gcc.gnu.org/gcc-5/changes.html 3277 7. http://gcc.gnu.org/onlinedocs/5.3.0/ 3278 8. http://gcc.gnu.org/gcc-5/changes.html 3279 9. http://gcc.gnu.org/onlinedocs/5.2.0/ 3280 10. http://gcc.gnu.org/gcc-5/changes.html 3281 11. http://gcc.gnu.org/onlinedocs/5.1.0/ 3282 12. http://gcc.gnu.org/gcc-5/buildstat.html 3283 13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 3284 14. http://gcc.gnu.org/index.html 3285 15. mailto:gcc (a] gcc.gnu.org 3286 16. http://gcc.gnu.org/mirrors.html 3287 17. http://gcc.gnu.org/svn.html 3288 18. https://gcc.gnu.org/onlinedocs/ 3289 19. mailto:gcc-help (a] gcc.gnu.org 3290 20. mailto:gcc (a] gcc.gnu.org 3291 21. https://gcc.gnu.org/lists.html 3292 22. https://www.fsf.org/ 3293 23. https://gcc.gnu.org/about.html 3294 24. http://validator.w3.org/check/referer 3295 ====================================================================== 3296 http://gcc.gnu.org/gcc-5/changes.html 3297 GCC 5 Release Series 3298 Changes, New Features, and Fixes 3299 3300 Caveats 3301 3302 * The default mode for C is now -std=gnu11 instead of -std=gnu89. 3303 * The C++ runtime library (libstdc++) uses a new ABI by default (see 3304 [1]below). 3305 * The Graphite framework for loop optimizations no longer requires 3306 the CLooG library, only ISL version 0.14 (recommended) or 0.12.2. 3307 The installation manual contains more information about 3308 requirements to build GCC. 3309 * The non-standard C++0x type traits has_trivial_default_constructor, 3310 has_trivial_copy_constructor and has_trivial_copy_assign have been 3311 deprecated and will be removed in a future version. The standard 3312 C++11 traits is_trivially_default_constructible, 3313 is_trivially_copy_constructible and is_trivially_copy_assignable 3314 should be used instead. 3315 * On AVR, support has been added for the devices 3316 ATtiny4/5/9/10/20/40. This requires Binutils 2.25 or newer. 3317 * The AVR port uses a new scheme to describe supported devices: For 3318 each supported device the compiler provides a device-specific 3319 [2]spec file. If the compiler is used together with AVR-LibC, this 3320 requires at least GCC 5.2 and a version of AVR-LibC which 3321 implements [3]feature #44574. 3322 3323 General Optimizer Improvements 3324 3325 * Inter-procedural optimization improvements: 3326 + An Identical Code Folding (ICF) pass (controlled via 3327 -fipa-icf) has been added. Compared to the identical code 3328 folding performed by the Gold linker this pass does not 3329 require function sections. It also performs merging before 3330 inlining, so inter-procedural optimizations are aware of the 3331 code re-use. On the other hand not all unifications performed 3332 by a linker are doable by GCC which must honor aliasing 3333 information. During link-time optimization of Firefox, this 3334 pass unifies about 31000 functions, that is 14% overall. 3335 + The devirtualization pass was significantly improved by adding 3336 better support for speculative devirtualization and dynamic 3337 type detection. About 50% of virtual calls in Firefox are now 3338 speculatively devirtualized during link-time optimization. 3339 + A new comdat localization pass allows the linker to eliminate 3340 more dead code in presence of C++ inline functions. 3341 + Virtual tables are now optimized. Local aliases are used to 3342 reduce dynamic linking time of C++ virtual tables on ELF 3343 targets and data alignment has been reduced to limit data 3344 segment bloat. 3345 + A new -fno-semantic-interposition option can be used to 3346 improve code quality of shared libraries where interposition 3347 of exported symbols is not allowed. 3348 + Write-only variables are now detected and optimized out. 3349 + With profile feedback the function inliner can now bypass 3350 --param inline-insns-auto and --param inline-insns-single 3351 limits for hot calls. 3352 + The IPA reference pass was significantly sped up making it 3353 feasible to enable -fipa-reference with -fprofile-generate. 3354 This also solves a bottleneck seen when building Chromium with 3355 link-time optimization. 3356 + The symbol table and call-graph API was reworked to C++ and 3357 simplified. 3358 + The interprocedural propagation of constants now also 3359 propagates alignments of pointer parameters. This for example 3360 means that the vectorizer often does not need to generate loop 3361 prologues and epilogues to make up for potential 3362 misalignments. 3363 * Link-time optimization improvements: 3364 + One Definition Rule based merging of C++ types has been 3365 implemented. Type merging enables better devirtualization and 3366 alias analysis. Streaming extra information needed to merge 3367 types adds about 2-6% of memory size and object size increase. 3368 This can be controlled by -flto-odr-type-merging. 3369 + Command-line optimization and target options are now streamed 3370 on a per-function basis and honored by the link-time 3371 optimizer. This change makes link-time optimization a more 3372 transparent replacement of per-file optimizations. It is now 3373 possible to build projects that require different optimization 3374 settings for different translation units (such as -ffast-math, 3375 -mavx, or -finline). Contrary to earlier GCC releases, the 3376 optimization and target options passed on the link command 3377 line are ignored. 3378 Note that this applies only to those command-line options that 3379 can be passed to optimize and target attributes. Command-line 3380 options affecting global code generation (such as -fpic), 3381 warnings (such as -Wodr), optimizations affecting the way 3382 static variables are optimized (such as -fcommon), debug 3383 output (such as -g), and --param parameters can be applied 3384 only to the whole link-time optimization unit. In these cases, 3385 it is recommended to consistently use the same options at both 3386 compile time and link time. 3387 + GCC bootstrap now uses slim LTO object files. 3388 + Memory usage and link times were improved. Tree merging was 3389 sped up, memory usage of GIMPLE declarations and types was 3390 reduced, and, support for on-demand streaming of variable 3391 constructors was added. 3392 * Feedback directed optimization improvements: 3393 + A new auto-FDO mode uses profiles collected by low overhead 3394 profiling tools (perf) instead of more expensive program 3395 instrumentation (via -fprofile-generate). SPEC2006 benchmarks 3396 on x86-64 improve by 4.7% with auto-FDO and by 7.3% with 3397 traditional feedback directed optimization. 3398 + Profile precision was improved in presence of C++ inline and 3399 extern inline functions. 3400 + The new gcov-tool utility allows manipulating profiles. 3401 + Profiles are now more tolerant to source file changes (this 3402 can be controlled by --param profile-func-internal-id). 3403 * Register allocation improvements: 3404 + A new local register allocator (LRA) sub-pass, controlled by 3405 -flra-remat, implements control-flow sensitive global register 3406 rematerialization. Instead of spilling and restoring a 3407 register value, it is recalculated if it is profitable. The 3408 sub-pass improved SPEC2000 generated code by 1% and 0.5% 3409 correspondingly on ARM and x86-64. 3410 + Reuse of the PIC hard register, instead of using a fixed 3411 register, was implemented on x86/x86-64 targets. This improves 3412 generated PIC code performance as more hard registers can be 3413 used. Shared libraries can significantly benefit from this 3414 optimization. Currently it is switched on only for x86/x86-64 3415 targets. As RA infrastructure is already implemented for PIC 3416 register reuse, other targets might follow this in the future. 3417 + A simple form of inter-procedural RA was implemented. When it 3418 is known that a called function does not use caller-saved 3419 registers, save/restore code is not generated around the call 3420 for such registers. This optimization can be controlled by 3421 -fipa-ra 3422 + LRA is now much more effective at generating spills of general 3423 registers into vector registers instead of memory on 3424 architectures (e.g., modern Intel processors) where this is 3425 profitable. 3426 * UndefinedBehaviorSanitizer gained a few new sanitization options: 3427 + -fsanitize=float-divide-by-zero: detect floating-point 3428 division by zero; 3429 + -fsanitize=float-cast-overflow: check that the result of 3430 floating-point type to integer conversions do not overflow; 3431 + -fsanitize=bounds: enable instrumentation of array bounds and 3432 detect out-of-bounds accesses; 3433 + -fsanitize=alignment: enable alignment checking, detect 3434 various misaligned objects; 3435 + -fsanitize=object-size: enable object size checking, detect 3436 various out-of-bounds accesses. 3437 + -fsanitize=vptr: enable checking of C++ member function calls, 3438 member accesses and some conversions between pointers to base 3439 and derived classes, detect if the referenced object does not 3440 have the correct dynamic type. 3441 * Pointer Bounds Checker, a bounds violation detector, has been added 3442 and can be enabled via -fcheck-pointer-bounds. Memory accesses are 3443 instrumented with run-time checks of used pointers against their 3444 bounds to detect pointer bounds violations (overflows). The Pointer 3445 Bounds Checker is available on x86/x86-64 GNU/Linux targets with a 3446 new ISA extension Intel MPX support. See the Pointer Bounds Checker 3447 [4]Wiki page for more details. 3448 3449 New Languages and Language specific improvements 3450 3451 * [5]OpenMP 4.0 specification offloading features are now supported 3452 by the C, C++, and Fortran compilers. Generic changes: 3453 + Infrastructure (suitable for any vendor). 3454 + Testsuite which covers offloading from the [6]OpenMP 4.0 3455 Examples document. 3456 Specific for upcoming Intel Xeon Phi products: 3457 + Run-time library. 3458 + Card emulator. 3459 * GCC 5 includes a preliminary implementation of the OpenACC 2.0a 3460 specification. OpenACC is intended for programming accelerator 3461 devices such as GPUs. See [7]the OpenACC wiki page for more 3462 information. 3463 3464 C family 3465 3466 * The default setting of the -fdiagnostics-color= command-line option 3467 is now [8]configurable when building GCC using configuration option 3468 --with-diagnostics-color=. The possible values are: never, always, 3469 auto and auto-if-env. The new default auto uses color only when the 3470 standard error is a terminal. The default in GCC 4.9 was 3471 auto-if-env, which is equivalent to auto if there is a non-empty 3472 GCC_COLORS environment variable, and never otherwise. As in GCC 3473 4.9, an empty GCC_COLORS variable in the environment will always 3474 disable colors, no matter what the default is or what command-line 3475 options are used. 3476 * A new command-line option -Wswitch-bool has been added for the C 3477 and C++ compilers, which warns whenever a switch statement has an 3478 index of boolean type. 3479 * A new command-line option -Wlogical-not-parentheses has been added 3480 for the C and C++ compilers, which warns about "logical not" used 3481 on the left hand side operand of a comparison. 3482 * A new command-line option -Wsizeof-array-argument has been added 3483 for the C and C++ compilers, which warns when the sizeof operator 3484 is applied to a parameter that has been declared as an array in a 3485 function definition. 3486 * A new command-line option -Wbool-compare has been added for the C 3487 and C++ compilers, which warns about boolean expressions compared 3488 with an integer value different from true/false. 3489 * Full support for [9]Cilk Plus has been added to the GCC compiler. 3490 Cilk Plus is an extension to the C and C++ languages to support 3491 data and task parallelism. 3492 * A new attribute no_reorder prevents reordering of selected symbols 3493 against other such symbols or inline assembler. This enables to 3494 link-time optimize the Linux kernel without having to resort to 3495 -fno-toplevel-reorder that disables several optimizations. 3496 * New preprocessor constructs, __has_include and __has_include_next, 3497 to test the availability of headers have been added. 3498 This demonstrates a way to include the header <optional> only if it 3499 is available: 3500 3501 #ifdef __has_include 3502 # if __has_include(<optional>) 3503 # include <optional> 3504 # define have_optional 1 3505 # elif __has_include(<experimental/optional>) 3506 # include <experimental/optional> 3507 # define have_optional 1 3508 # define experimental_optional 3509 # else 3510 # define have_optional 0 3511 # endif 3512 #endif 3513 3514 The header search paths for __has_include and __has_include_next 3515 are equivalent to those of the standard directive #include and the 3516 extension #include_next respectively. 3517 * A new built-in function-like macro to determine the existence of an 3518 attribute, __has_attribute, has been added. The equivalent built-in 3519 macro __has_cpp_attribute was added to C++ to support 3520 [10]Feature-testing recommendations for C++. The macro 3521 __has_attribute is added to all C-like languages as an extension: 3522 3523 int 3524 #ifdef __has_attribute 3525 # if __has_attribute(__noinline__) 3526 __attribute__((__noinline__)) 3527 # endif 3528 #endif 3529 foo(int x); 3530 3531 If an attribute exists, a nonzero constant integer is returned. For 3532 standardized C++ attributes a date is returned, otherwise the 3533 constant returned is 1. Both __has_attribute and 3534 __has_cpp_attribute will add underscores to an attribute name if 3535 necessary to resolve the name. For C++11 and onwards the attribute 3536 may be scoped. 3537 * A new set of built-in functions for arithmetics with overflow 3538 checking has been added: __builtin_add_overflow, 3539 __builtin_sub_overflow and __builtin_mul_overflow and for 3540 compatibility with clang also other variants. These builtins have 3541 two integral arguments (which don't need to have the same type), 3542 the arguments are extended to infinite precision signed type, +, - 3543 or * is performed on those, and the result is stored in an integer 3544 variable pointed to by the last argument. If the stored value is 3545 equal to the infinite precision result, the built-in functions 3546 return false, otherwise true. The type of the integer variable that 3547 will hold the result can be different from the types of the first 3548 two arguments. The following snippet demonstrates how this can be 3549 used in computing the size for the calloc function: 3550 3551 void * 3552 calloc (size_t x, size_t y) 3553 { 3554 size_t sz; 3555 if (__builtin_mul_overflow (x, y, &sz)) 3556 return NULL; 3557 void *ret = malloc (sz); 3558 if (ret) memset (res, 0, sz); 3559 return ret; 3560 } 3561 3562 On e.g. i?86 or x86-64 the above will result in a mul instruction 3563 followed by a jump on overflow. 3564 * The option -fextended-identifiers is now enabled by default for 3565 C++, and for C99 and later C versions. Various bugs in the 3566 implementation of extended identifiers have been fixed. 3567 3568 C 3569 3570 * The default mode has been changed to -std=gnu11. 3571 * A new command-line option -Wc90-c99-compat has been added to warn 3572 about features not present in ISO C90, but present in ISO C99. 3573 * A new command-line option -Wc99-c11-compat has been added to warn 3574 about features not present in ISO C99, but present in ISO C11. 3575 * It is possible to disable warnings about conversions between 3576 pointers that have incompatible types via a new warning option 3577 -Wno-incompatible-pointer-types; warnings about implicit 3578 incompatible integer to pointer and pointer to integer conversions 3579 via a new warning option -Wno-int-conversion; and warnings about 3580 qualifiers on pointers being discarded via a new warning option 3581 -Wno-discarded-qualifiers. 3582 * To allow proper use of const qualifiers with multidimensional 3583 arrays, GCC will not warn about incompatible pointer types anymore 3584 for conversions between pointers to arrays with and without const 3585 qualifier (except when using -pedantic). Instead, a new warning is 3586 emitted only if the const qualifier is lost. This can be controlled 3587 with a new warning option -Wno-discarded-array-qualifiers. 3588 * The C front end now generates more precise caret diagnostics. 3589 * The -pg command-line option now only affects the current file in an 3590 LTO build. 3591 3592 C++ 3593 3594 * G++ now supports [11]C++14 variable templates. 3595 * -Wnon-virtual-dtor doesn't warn anymore for final classes. 3596 * Excessive template instantiation depth is now a fatal error. This 3597 prevents excessive diagnostics that usually do not help to identify 3598 the problem. 3599 * G++ and libstdc++ now implement the feature-testing macros from 3600 [12]Feature-testing recommendations for C++. 3601 * G++ now allows typename in a template template parameter. 3602 3603 template<template<typename> typename X> struct D; // OK 3604 3605 * G++ now supports [13]C++14 aggregates with non-static data member 3606 initializers. 3607 3608 struct A { int i, j = i; }; 3609 A a = { 42 }; // a.j is also 42 3610 3611 * G++ now supports [14]C++14 extended constexpr. 3612 3613 constexpr int f (int i) 3614 { 3615 int j = 0; 3616 for (; i > 0; --i) 3617 ++j; 3618 return j; 3619 } 3620 3621 constexpr int i = f(42); // i is 42 3622 3623 * G++ now supports the [15]C++14 sized deallocation functions. 3624 3625 void operator delete (void *, std::size_t) noexcept; 3626 void operator delete[] (void *, std::size_t) noexcept; 3627 3628 * A new One Definition Rule violation warning (controlled by -Wodr) 3629 detects mismatches in type definitions and virtual table contents 3630 during link-time optimization. 3631 * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help 3632 developers to annotate programs with final specifiers (or anonymous 3633 namespaces) to improve code generation. These warnings can be used 3634 at compile time, but they are more useful in combination with 3635 link-time optimization. 3636 * G++ no longer supports [16]N3639 variable length arrays, as they 3637 were removed from the C++14 working paper prior to ratification. 3638 GNU VLAs are still supported, so VLA support is now the same in 3639 C++14 mode as in C++98 and C++11 modes. 3640 * G++ now allows passing a non-trivially-copyable class via C 3641 varargs, which is conditionally-supported with 3642 implementation-defined semantics in the standard. This uses the 3643 same calling convention as a normal value parameter. 3644 * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So 3645 various mangling bugs are fixed, but G++ will still emit aliases 3646 with the old, wrong mangling where feasible. -Wabi=2 will warn 3647 about differences between ABI version 2 and the current setting. 3648 * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely 3649 to be unaffected, but -Wabi=8 will warn about a non-static data 3650 member with type std::nullptr_t which changes position due to this 3651 change. 3652 3653 Runtime Library (libstdc++) 3654 3655 * A [17]Dual ABI is provided by the library. A new ABI is enabled by 3656 default. The old ABI is still supported and can be used by defining 3657 the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++ 3658 standard library headers. 3659 * A new implementation of std::string is enabled by default, using 3660 the small string optimization instead of copy-on-write reference 3661 counting. 3662 * A new implementation of std::list is enabled by default, with an 3663 O(1) size() function; 3664 * [18]Full support for C++11, including the following new features: 3665 + std::deque and std::vector<bool> meet the allocator-aware 3666 container requirements; 3667 + movable and swappable iostream classes; 3668 + support for std::align and std::aligned_union; 3669 + type traits std::is_trivially_copyable, 3670 std::is_trivially_constructible, std::is_trivially_assignable 3671 etc.; 3672 + I/O manipulators std::put_time, std::get_time, std::hexfloat 3673 and std::defaultfloat; 3674 + generic locale-aware std::isblank; 3675 + locale facets for Unicode conversion; 3676 + atomic operations for std::shared_ptr; 3677 + std::notify_all_at_thread_exit() and functions for making 3678 futures ready at thread exit. 3679 * Support for the C++11 hexfloat manipulator changes how the num_put 3680 facet formats floating point types when 3681 ios_base::fixed|ios_base::scientific is set in a stream's fmtflags. 3682 This change affects all language modes, even though the C++98 3683 standard gave no special meaning to that combination of flags. To 3684 prevent the use of hexadecimal notation for floating point types 3685 use str.unsetf(std::ios_base::floatfield) to clear the relevant 3686 bits in str.flags(). 3687 * [19]Full experimental support for C++14, including the following 3688 new features: 3689 + std::is_final type trait; 3690 + heterogeneous comparison lookup in associative containers. 3691 + global functions cbegin, cend, rbegin, rend, crbegin, and 3692 crend for range access to containers, arrays and initializer 3693 lists. 3694 * [20]Improved experimental support for the Library Fundamentals TS, 3695 including: 3696 + class std::experimental::any; 3697 + function template std::experimental::apply; 3698 + function template std::experimental::sample; 3699 + function template std::experimental::search and related 3700 searcher types; 3701 + variable templates for type traits; 3702 + function template std::experimental::not_fn. 3703 * New random number distributions logistic_distribution and 3704 uniform_on_sphere_distribution as extensions. 3705 * [21]GDB Xmethods for containers and std::unique_ptr. 3706 3707 Fortran 3708 3709 * Compatibility notice: 3710 + The version of the module files (.mod) has been incremented. 3711 + For free-form source files [22]-Werror=line-truncation is now 3712 enabled by default. Note that comments exceeding the line 3713 length are not diagnosed. (For fixed-form source code, the 3714 same warning is available but turned off by default, such that 3715 excess characters are ignored. -ffree-line-length-n and 3716 -ffixed-line-length-n can be used to modify the default line 3717 lengths of 132 and 72 columns, respectively.) 3718 + The -Wtabs option is now more sensible: with -Wtabs the 3719 compiler warns if it encounters tabs and with -Wno-tabs this 3720 warning is turned off. Before, -Wno-tabs warned and -Wtabs 3721 disabled the warning. As before, this warning is also enabled 3722 by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts 3723 options of -std=. 3724 * Incomplete support for colorizing diagnostics emitted by gfortran 3725 has been added. The option [23]-fdiagnostics-color controls when 3726 color is used in diagnostics. The default value of this option can 3727 be [24]configured when building GCC. The GCC_COLORS environment 3728 variable can be used to customize the colors or disable coloring 3729 completely. Sample diagnostics output: 3730 $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90 3731 test.f90:6:1: 3732 3733 0 continue 3734 1 3735 Error: Zero is not a valid statement label at (1) 3736 test.f90:9:6: 3737 3738 USE foo 3739 1 3740 Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only] 3741 3742 * The -Wuse-without-only option has been added to warn when a USE 3743 statement has no ONLY qualifier and thus implicitly imports all 3744 public entities of the used module. 3745 * Formatted READ and WRITE statements now work correctly in 3746 locale-aware programs. For more information and potential caveats, 3747 see [25]Section 5.3 Thread-safety of the runtime library in the 3748 manual. 3749 * [26]Fortran 2003: 3750 + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and 3751 IEEE_ARITHMETIC) are now supported. 3752 * [27]Fortran 2008: 3753 + [28]Coarrays: Full experimental support of Fortran 2008's 3754 coarrays with -fcoarray=lib except for allocatable/pointer 3755 components of derived-type coarrays. GCC currently only ships 3756 with a single-image library (libcaf_single), but multi-image 3757 support based on MPI and GASNet is provided by the libraries 3758 of the [29]OpenCoarrays project. 3759 * TS18508 Additional Parallel Features in Fortran: 3760 + Support for the collective intrinsic subroutines CO_MAX, 3761 CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added, 3762 including -fcoarray=lib support. 3763 + Support for the new atomic intrinsics has been added, 3764 including -fcoarray=lib support. 3765 * Fortran 2015: 3766 + Support for IMPLICIT NONE (external, type). 3767 + ERROR STOP is now permitted in pure procedures. 3768 3769 Go 3770 3771 * GCC 5 provides a complete implementation of the Go 1.4.2 release. 3772 * Building GCC 5 with Go enabled will install two new programs: 3773 [30]go and [31]gofmt. 3774 3775 libgccjit 3776 3777 New in GCC 5 is the ability to build GCC as a shared library for 3778 embedding in other processes (such as interpreters), suitable for 3779 Just-In-Time compilation to machine code. 3780 3781 The shared library has a [32]C API and a [33]C++ wrapper API providing 3782 some "syntactic sugar". There are also bindings available from 3rd 3783 parties for [34]Python and for [35]D. 3784 3785 For example, this library can be used by interpreters for [36]compiling 3786 functions from bytecode to machine code. 3787 3788 The library can also be used for ahead-of-time compilation, enabling 3789 GCC to be plugged into a pre-existing front end. An example of using 3790 this to build a compiler for an esoteric language we'll refer to as 3791 "brainf" can be seen [37]here. 3792 3793 libgccjit is licensed under the GPLv3 (or at your option, any later 3794 version) 3795 3796 It should be regarded as experimental at this time. 3797 3798 New Targets and Target Specific Improvements 3799 3800 Reporting stack usage 3801 3802 * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the 3803 -fstack-usage option. 3804 3805 AArch64 3806 3807 * Code generation for the ARM Cortex-A57 processor has been improved. 3808 A more accurate instruction scheduling model for the processor is 3809 now used, and a number of compiler tuning parameters have been set 3810 to offer increased performance when compiling with -mcpu=cortex-a57 3811 or -mtune=cortex-a57. 3812 * A workaround for the ARM Cortex-A53 erratum 835769 has been added 3813 and can be enabled by giving the -mfix-cortex-a53-835769 option. 3814 Alternatively it can be enabled by default by configuring GCC with 3815 the --enable-fix-cortex-a53-835769 option. 3816 * The optional cryptographic extensions to the ARMv8-A architecture 3817 are no longer enabled by default when specifying the 3818 -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53 3819 options. To enable these extensions add +crypto to the value of 3820 -mcpu or -march e.g. -mcpu=cortex-a53+crypto. 3821 * Support has been added for the following processors (GCC 3822 identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and 3823 initial support for its big.LITTLE combination with the ARM 3824 Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx), 3825 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 3826 The GCC identifiers can be used as arguments to the -mcpu or -mtune 3827 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 3828 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 3829 support for the Cortex-A72. 3830 * The transitional options -mlra and -mno-lra have been removed. The 3831 AArch64 backend now uses the local register allocator (LRA) only. 3832 3833 ARM 3834 3835 * Thumb-1 assembly code is now generated in unified syntax. The new 3836 option -masm-syntax-unified specifies whether inline assembly code 3837 is using unified syntax. By default the option is off which means 3838 non-unified syntax is used. However this is subject to change in 3839 future releases. Eventually the non-unified syntax will be 3840 deprecated. 3841 * It is now a configure-time error to use the --with-cpu configure 3842 option with either of --with-tune or --with-arch. 3843 * Code generation for the ARM Cortex-A57 processor has been improved. 3844 A more accurate instruction scheduling model for the processor is 3845 now used, and a number of compiler tuning parameters have been set 3846 to offer increased performance when compiling with -mcpu=cortex-a57 3847 or -mtune=cortex-a57. 3848 * Support has been added for the following processors (GCC 3849 identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and 3850 initial support for its big.LITTLE combination with the ARM 3851 Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and 3852 initial support for its big.LITTLE combination with the ARM 3853 Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7), 3854 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 3855 The GCC identifiers can be used as arguments to the -mcpu or -mtune 3856 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 3857 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 3858 support for the Cortex-A72. 3859 * The deprecated option -mwords-little-endian has been removed. 3860 * The options -mapcs, -mapcs-frame, -mtpcs-frame and 3861 -mtpcs-leaf-frame which are only applicable to the old ABI have 3862 been deprecated. 3863 * The transitional options -mlra and -mno-lra have been removed. The 3864 ARM backend now uses the local register allocator (LRA) only. 3865 3866 AVR 3867 3868 * The compiler no more supports individual devices like ATmega8. 3869 Specifying, say, -mmcu=atmega8 triggers the usage of the 3870 device-specific [38]spec file specs-atmega8 which is part of the 3871 installation and describes options for the sub-processes like 3872 compiler proper, assembler and linker. You can add support for a 3873 new device -mmcu=mydevice as follows: 3874 1. In an empty directory /someplace, create a new directory 3875 device-specs. 3876 2. Copy a device spec file from the installed device-specs 3877 folder, follow the comments in that file and then save it as 3878 /someplace/device-specs/specs-mydevice. 3879 3. Add -B /someplace -mmcu=mydevice to the compiler's 3880 command-line options. Notice that /someplace must specify an 3881 absolute path and that mydevice must not start with "avr". 3882 4. Provided you have a device-specific library libmydevice.a 3883 available, you can put it at /someplace, dito for a 3884 device-specific startup file crtmydevice.o. 3885 The contents of the device spec files depend on the compiler's 3886 configuration, in particular on --with-avrlibc=no and whether or 3887 not it is configured for RTEMS. 3888 * A new command-line option -nodevicelib has been added. It prevents 3889 the compiler from linking against AVR-LibC's device-specific 3890 library libdevice.a. 3891 * The following three command-line options have been added: 3892 3893 -mrmw 3894 Set if the device supports the read-modify-write 3895 instructions LAC, LAS, LAT and XCH. 3896 3897 -mn-flash=size 3898 Specify the flash size of the device in units of 64 KiB, 3899 rounded up to the next integer as needed. This option 3900 affects the availability of the [39]AVR address-spaces. 3901 3902 -mskip-bug 3903 Set if the device is affected by the respective silicon 3904 bug. 3905 3906 In general, you don't need to set these options by hand. The new 3907 device-specific spec file will set them as needed. 3908 3909 IA-32/x86-64 3910 3911 * New [40]ISA extensions support AVX-512{BW,DQ,VL,IFMA,VBMI} of 3912 Intel's CPU codenamed Skylake Server was added to GCC. That 3913 includes inline assembly support, new intrinsics, and basic 3914 autovectorization. These new AVX-512 extensions are available via 3915 the following GCC switches: AVX-512 Vector Length EVEX feature: 3916 -mavx512vl, AVX-512 Byte and Word instructions: -mavx512bw, AVX-512 3917 Dword and Qword instructions: -mavx512dq, AVX-512 FMA-52 3918 instructions: -mavx512ifma and for AVX-512 Vector Bit Manipulation 3919 Instructions: -mavx512vbmi. 3920 * New ISA extensions support Intel MPX was added to GCC. This new 3921 extension is available via the -mmpx compiler switch. Intel MPX is 3922 a set of processor features which, with compiler, run-time library 3923 and OS support, brings increased robustness to software by run-time 3924 checking pointer references against their bounds. In GCC Intel MPX 3925 is supported by Pointer Bounds Checker and libmpx run-time 3926 libraries. 3927 * The new -mrecord-mcount option for -pg generates a Linux kernel 3928 style table of pointers to mcount or __fentry__ calls at the 3929 beginning of functions. The new -mnop-mcount option in addition 3930 also generates nops in place of the __fentry__ or mcount call, so 3931 that a call per function can be later patched in. This can be used 3932 for low overhead tracing or hot code patching. 3933 * The new -malign-data option controls how GCC aligns variables. 3934 -malign-data=compat uses increased alignment compatible with GCC 3935 4.8 and earlier, -malign-data=abi uses alignment as specified by 3936 the psABI, and -malign-data=cacheline uses increased alignment to 3937 match the cache line size. -malign-data=compat is the default. 3938 * The new -mskip-rax-setup option skips setting up the RAX register 3939 when SSE is disabled and there are no variable arguments passed in 3940 vector registers. This can be used to optimize the Linux kernel. 3941 3942 MIPS 3943 3944 * MIPS Releases 3 and 5 are now directly supported. Use the 3945 command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5 3946 to enable code-generation for these processors. 3947 * The Imagination P5600 processor is now supported using the 3948 -march=p5600 command-line option. 3949 * The Cavium Octeon3 processor is now supported using the 3950 -march=octeon3 command-line option. 3951 * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6 3952 command-line options. 3953 * The o32 ABI has been modified and extended. The o32 64-bit 3954 floating-point register support is now obsolete and has been 3955 removed. It has been replaced by three ABI extensions FPXX, FP64A, 3956 and FP64. The meaning of the -mfp64 command-line option has 3957 changed. It is now used to enable the FP64A and FP64 ABI 3958 extensions. 3959 + The FPXX extension requires that code generated to access 3960 double-precision values use even-numbered registers. Code that 3961 adheres to this extension is link-compatible with all other 3962 o32 double-precision ABI variants and will execute correctly 3963 in all hardware FPU modes. The command-line options -mabi=32 3964 -mfpxx can be used to enable this extension. MIPS II is the 3965 minimum processor required. 3966 + The o32 FP64A extension requires that floating-point registers 3967 be 64-bit and odd-numbered single-precision registers are not 3968 allowed. Code that adheres to the o32 FP64A variant is 3969 link-compatible with all other o32 double-precision ABI 3970 variants. The command-line options -mabi=32 -mfp64 3971 -mno-odd-spreg can be used to enable this extension. MIPS32R2 3972 is the minimum processor required. 3973 + The o32 FP64 extension also requires that floating-point 3974 registers be 64-bit, but permits the use of single-precision 3975 registers. Code that adheres to the o32 FP64 variant is 3976 link-compatible with o32 FPXX and o32 FP64A variants only, 3977 i.e. it is not compatible with the original o32 3978 double-precision ABI. The command-line options -mabi=32 -mfp64 3979 -modd-spreg can be used to enable this extension. MIPS32R2 is 3980 the minimum processor required. 3981 The new ABI variants can be enabled by default using the configure 3982 time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32. 3983 It is strongly recommended that all vendors begin to set o32 FPXX 3984 as the default ABI. This will be required to run the generated code 3985 on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and 3986 MIPSR6 cores. 3987 * GCC will now pass all floating-point options to the assembler if 3988 GNU binutils 2.25 is used. As a result, any inline assembly code 3989 that uses hard-float instructions should be amended to include a 3990 .set directive to override the global assembler options when 3991 compiling for soft-float targets. 3992 3993 NDS32 3994 3995 * The variadic function ABI implementation is now compatible with 3996 past Andes toolchains where the caller uses registers to pass 3997 arguments and the callee is in charge of pushing them on stack. 3998 * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have 3999 been removed since they are not yet available in the nds32 port of 4000 GNU binutils. 4001 * A new option -mcmodel=[small|medium|large] supports varied code 4002 models on code generation. The -mgp-direct option became 4003 meaningless and can be discarded. 4004 4005 RX 4006 4007 * A new command line option -mno-allow-string-insns can be used to 4008 disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL, 4009 SWHILE and RMPA instructions. An erratum released by Renesas shows 4010 that it is unsafe to use these instructions on addresses within the 4011 I/O space of the processor. The new option can be used when the 4012 programmer is concerned that the I/O space might be accessed. The 4013 default is still to enable these instructions. 4014 4015 SH 4016 4017 * The compiler will now pass the appropriate --isa= option to the 4018 assembler. 4019 * The default handling for the GBR has been changed from call 4020 clobbered to call preserved. The old behavior can be reinstated by 4021 specifying the option -fcall-used-gbr. 4022 * Support for the SH4A fpchg instruction has been added which will be 4023 utilized when switching between single and double precision FPU 4024 modes. 4025 * The compiler no longer uses the __fpscr_values array for switching 4026 between single and double FPU precision modes on non-SH4A targets. 4027 Instead mode switching will now be performed by storing, modifying 4028 and reloading the FPSCR, so that other FPSCR bits are preserved 4029 across mode switches. The __fpscr_values array that is defined in 4030 libgcc is still present for backwards compatibility, but it will 4031 not be referenced by compiler generated code anymore. 4032 * New builtin functions __builtin_sh_get_fpscr and 4033 __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr 4034 function will mask the specified bits in such a way that the SZ, PR 4035 and FR mode bits will be preserved, while changing the other bits. 4036 These new functions do not reference the __fpscr_values array. The 4037 old functions __set_fpscr and __get_fpscr in libgcc which access 4038 the __fpscr_values array are still present for backwards 4039 compatibility, but their usage is highly discouraged. 4040 * Some improvements to code generated for __atomic built-in 4041 functions. 4042 * When compiling for SH2E the compiler will no longer force the usage 4043 of delay slots for conditional branch instructions bt and bf. The 4044 old behavior can be reinstated (e.g. to work around a hardware bug 4045 in the original SH7055) by specifying the new option 4046 -mcbranch-force-delay-slot. 4047 4048 Operating Systems 4049 4050 AIX 4051 4052 * GCC now supports stabs debugging continuation lines to allow long 4053 stabs debug information without overflow that generates AIX linker 4054 errors. 4055 4056 DragonFly BSD 4057 4058 * GCC now supports the DragonFly BSD operating system. 4059 4060 FreeBSD 4061 4062 * GCC now supports the FreeBSD operating system for the arm port 4063 through the arm*-*-freebsd* target triplets. 4064 4065 VxWorks MILS 4066 4067 * GCC now supports the MILS (Multiple Independent Levels of Security) 4068 variant of WindRiver's VxWorks operating system for PowerPC 4069 targets. 4070 4071 Other significant improvements 4072 4073 * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option 4074 to set the compiler to use. 4075 4076 * When the new command-line option -freport-bug is used, GCC 4077 automatically generates a developer-friendly reproducer whenever an 4078 internal compiler error is encountered. 4079 4080 GCC 5.2 4081 4082 This is the [41]list of problem reports (PRs) from GCC's bug tracking 4083 system that are known to be fixed in the 5.2 release. This list might 4084 not be complete (that is, it is possible that some PRs that have been 4085 fixed are not listed here). 4086 4087 Target Specific Changes 4088 4089 IA-32/x86-64 4090 4091 * Support for new AMD instructions monitorx and mwaitx has been 4092 added. This includes new intrinsic and built-in support. It is 4093 enabled through option -mmwaitx. The instructions monitorx and 4094 mwaitx implement the same functionality as the old monitor and 4095 mwait instructions. In addition mwaitx adds a configurable timer. 4096 The timer value is received as third argument and stored in 4097 register %ebx. 4098 4099 S/390, System z, IBM z Systems 4100 4101 * Support for the IBM z13 processor has been added. When using the 4102 -march=z13 option, the compiler will generate code making use of 4103 the new instructions and registers introduced with the vector 4104 extension facility. The -mtune=z13 option enables z13 specific 4105 instruction scheduling without making use of new instructions. 4106 Compiling code with -march=z13 reduces the default alignment of 4107 vector types bigger than 8 bytes to 8. This is an ABI change and 4108 care must be taken when linking modules compiled with different 4109 arch levels which interchange variables containing vector type 4110 values. For newly compiled code the GNU linker will emit a warning. 4111 * The -mzvector option enables a C/C++ language extension. This 4112 extension provides a new keyword vector which can be used to define 4113 vector type variables. (Note: This is not available when enforcing 4114 strict standard compliance e.g. with -std=c99. Either enable GNU 4115 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 4116 Additionally a set of overloaded builtins is provided which is 4117 partially compatible to the PowerPC Altivec builtins. In order to 4118 make use of these builtins the vecintrin.h header file needs to be 4119 included. 4120 4121 GCC 5.3 4122 4123 This is the [42]list of problem reports (PRs) from GCC's bug tracking 4124 system that are known to be fixed in the 5.3 release. This list might 4125 not be complete (that is, it is possible that some PRs that have been 4126 fixed are not listed here). 4127 4128 Target Specific Changes 4129 4130 IA-32/x86-64 4131 4132 * GCC now supports the Intel CPU named Skylake with AVX-512 4133 extensions through -march=skylake-avx512. The switch enables the 4134 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 4135 AVX-512DQ. 4136 4137 S/390, System z, IBM z Systems 4138 4139 * With this version of GCC IBM z Systems support has been added to 4140 the GO runtime environment. GCC 5.3 has proven to be able to 4141 compile larger GO applications on IBM z Systems. 4142 4143 GCC 5.4 4144 4145 This is the [43]list of problem reports (PRs) from GCC's bug tracking 4146 system that are known to be fixed in the 5.4 release. This list might 4147 not be complete (that is, it is possible that some PRs that have been 4148 fixed are not listed here). 4149 4150 GCC 5.5 4151 4152 This is the [44]list of problem reports (PRs) from GCC's bug tracking 4153 system that are known to be fixed in the 5.5 release. This list might 4154 not be complete (that is, it is possible that some PRs that have been 4155 fixed are not listed here). 4156 4157 Target Specific Changes 4158 4159 IA-32/x86-64 4160 4161 * Support for the [45]deprecated pcommit instruction has been 4162 removed. 4163 4164 4165 For questions related to the use of GCC, please consult these web 4166 pages and the [46]GCC manuals. If that fails, the 4167 [47]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4168 web pages and the development of GCC are welcome on our developer 4169 list at [48]gcc (a] gcc.gnu.org. All of [49]our lists have public 4170 archives. 4171 4172 Copyright (C) [50]Free Software Foundation, Inc. Verbatim copying and 4173 distribution of this entire article is permitted in any medium, 4174 provided this notice is preserved. 4175 4176 These pages are [51]maintained by the GCC team. Last modified 4177 2018-09-30[52]. 4178 4179 References 4180 4181 1. http://gcc.gnu.org/gcc-5/changes.html#libstdcxx 4182 2. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 4183 3. https://savannah.nongnu.org/bugs/?44574 4184 4. https://gcc.gnu.org/wiki/Intel MPX support in the GCC compiler 4185 5. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.pdf 4186 6. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.Examples.pdf 4187 7. https://gcc.gnu.org/wiki/OpenACC 4188 8. https://gcc.gnu.org/install/configure.html 4189 9. https://www.cilkplus.org/ 4190 10. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 4191 11. http://gcc.gnu.org/projects/cxx1y.html 4192 12. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 4193 13. http://gcc.gnu.org/projects/cxx1y.html 4194 14. http://gcc.gnu.org/projects/cxx1y.html 4195 15. http://gcc.gnu.org/projects/cxx1y.html 4196 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html 4197 17. http://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html 4198 18. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011 4199 19. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 4200 20. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 4201 21. https://sourceware.org/gdb/current/onlinedocs/gdb/Xmethods-In-Python.html 4202 22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html 4203 23. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html 4204 24. https://gcc.gnu.org/install/configure.html 4205 25. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html 4206 26. https://gcc.gnu.org/wiki/Fortran2003Status 4207 27. https://gcc.gnu.org/wiki/Fortran2008Status 4208 28. https://gcc.gnu.org/wiki/Coarray 4209 29. http://www.opencoarrays.org/ 4210 30. https://golang.org/cmd/go/ 4211 31. https://golang.org/cmd/gofmt/ 4212 32. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html 4213 33. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html 4214 34. https://github.com/davidmalcolm/pygccjit 4215 35. https://github.com/ibuclaw/gccjitd 4216 36. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html 4217 37. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html 4218 38. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 4219 39. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html 4220 40. https://software.intel.com/sites/default/files/managed/39/c5/325462-sdm-vol-1-2abcd-3abcd.pdf 4221 41. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2 4222 42. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3 4223 43. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.4 4224 44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.5 4225 45. https://software.intel.com/en-us/blogs/2016/09/12/deprecate-pcommit-instruction 4226 46. https://gcc.gnu.org/onlinedocs/ 4227 47. mailto:gcc-help (a] gcc.gnu.org 4228 48. mailto:gcc (a] gcc.gnu.org 4229 49. https://gcc.gnu.org/lists.html 4230 50. https://www.fsf.org/ 4231 51. https://gcc.gnu.org/about.html 4232 52. http://validator.w3.org/check/referer 4233 ====================================================================== 4234 http://gcc.gnu.org/gcc-4.9/index.html 4235 GCC 4.9 Release Series 4236 4237 Aug 3, 2016 4238 4239 The [1]GNU project and the GCC developers are pleased to announce the 4240 release of GCC 4.9.4. 4241 4242 This release is a bug-fix release, containing fixes for regressions in 4243 GCC 4.9.3 relative to previous releases of GCC. 4244 4245 This release series is no longer maintained. 4246 4247 Release History 4248 4249 GCC 4.9.4 4250 Aug 3, 2016 ([2]changes, [3]documentation) 4251 4252 GCC 4.9.3 4253 June 26, 2015 ([4]changes, [5]documentation) 4254 4255 GCC 4.9.2 4256 October 30, 2014 ([6]changes, [7]documentation) 4257 4258 GCC 4.9.1 4259 July 16, 2014 ([8]changes, [9]documentation) 4260 4261 GCC 4.9.0 4262 April 22, 2014 ([10]changes, [11]documentation) 4263 4264 References and Acknowledgements 4265 4266 GCC used to stand for the GNU C Compiler, but since the compiler 4267 supports several other languages aside from C, it now stands for the 4268 GNU Compiler Collection. 4269 4270 A list of [12]successful builds is updated as new information becomes 4271 available. 4272 4273 The GCC developers would like to thank the numerous people that have 4274 contributed new features, improvements, bug fixes, and other changes as 4275 well as test results to GCC. This [13]amazing group of volunteers is 4276 what makes GCC successful. 4277 4278 For additional information about GCC please refer to the [14]GCC 4279 project web site or contact the [15]GCC development mailing list. 4280 4281 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 4282 4283 4284 For questions related to the use of GCC, please consult these web 4285 pages and the [18]GCC manuals. If that fails, the 4286 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4287 web pages and the development of GCC are welcome on our developer 4288 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 4289 archives. 4290 4291 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 4292 distribution of this entire article is permitted in any medium, 4293 provided this notice is preserved. 4294 4295 These pages are [23]maintained by the GCC team. Last modified 4296 2018-09-30[24]. 4297 4298 References 4299 4300 1. http://www.gnu.org/ 4301 2. http://gcc.gnu.org/gcc-4.9/changes.html 4302 3. http://gcc.gnu.org/onlinedocs/4.9.4/ 4303 4. http://gcc.gnu.org/gcc-4.9/changes.html 4304 5. http://gcc.gnu.org/onlinedocs/4.9.3/ 4305 6. http://gcc.gnu.org/gcc-4.9/changes.html 4306 7. http://gcc.gnu.org/onlinedocs/4.9.2/ 4307 8. http://gcc.gnu.org/gcc-4.9/changes.html 4308 9. http://gcc.gnu.org/onlinedocs/4.9.1/ 4309 10. http://gcc.gnu.org/gcc-4.9/changes.html 4310 11. https://gcc.gnu.org/onlinedocs/4.9.0/ 4311 12. http://gcc.gnu.org/gcc-4.9/buildstat.html 4312 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 4313 14. http://gcc.gnu.org/index.html 4314 15. mailto:gcc (a] gcc.gnu.org 4315 16. http://gcc.gnu.org/mirrors.html 4316 17. http://gcc.gnu.org/svn.html 4317 18. https://gcc.gnu.org/onlinedocs/ 4318 19. mailto:gcc-help (a] gcc.gnu.org 4319 20. mailto:gcc (a] gcc.gnu.org 4320 21. https://gcc.gnu.org/lists.html 4321 22. https://www.fsf.org/ 4322 23. https://gcc.gnu.org/about.html 4323 24. http://validator.w3.org/check/referer 4324 ====================================================================== 4325 http://gcc.gnu.org/gcc-4.9/changes.html 4326 GCC 4.9 Release Series 4327 Changes, New Features, and Fixes 4328 4329 Caveats 4330 4331 * The mudflap run time checker has been removed. The mudflap options 4332 remain, but do nothing. 4333 * Support for a number of older systems and recently unmaintained or 4334 untested target ports of GCC has been declared obsolete in GCC 4.9. 4335 Unless there is activity to revive them, the next release of GCC 4336 will have their sources permanently removed. 4337 The following ports for individual systems on particular 4338 architectures have been obsoleted: 4339 + Solaris 9 (*-*-solaris2.9). Details can be found in the 4340 [1]announcement. 4341 * On AArch64, the singleton vector types int64x1_t, uint64x1_t and 4342 float64x1_t exported by arm_neon.h are defined to be the same as 4343 their base types. This results in incorrect application of 4344 parameter passing rules to arguments of types int64x1_t and 4345 uint64x1_t, with respect to the AAPCS64 ABI specification. In 4346 addition, names of C++ functions with parameters of these types 4347 (including float64x1_t) are not mangled correctly. The current 4348 typedef declarations also unintentionally allow implicit casting 4349 between singleton vector types and their base types. These issues 4350 will be resolved in a near future release. See [2]PR60825 for more 4351 information. 4352 4353 More information on porting to GCC 4.9 from previous versions of GCC 4354 can be found in the [3]porting guide for this release. 4355 4356 General Optimizer Improvements 4357 4358 * AddressSanitizer, a fast memory error detector, is now available on 4359 ARM. 4360 * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior 4361 detector, has been added and can be enabled via 4362 -fsanitize=undefined. Various computations will be instrumented to 4363 detect undefined behavior at runtime. UndefinedBehaviorSanitizer is 4364 currently available for the C and C++ languages. 4365 * Link-time optimization (LTO) improvements: 4366 + Type merging was rewritten. The new implementation is 4367 significantly faster and uses less memory. 4368 + Better partitioning algorithm resulting in less streaming 4369 during link time. 4370 + Early removal of virtual methods reduces the size of object 4371 files and improves link-time memory usage and compile time. 4372 + Function bodies are now loaded on-demand and released early 4373 improving overall memory usage at link time. 4374 + C++ hidden keyed methods can now be optimized out. 4375 + When using a linker plugin, compiling with the -flto option 4376 now generates slim object files (.o) which only contain 4377 intermediate language representation for LTO. Use 4378 -ffat-lto-objects to create files which contain additionally 4379 the object code. To generate static libraries suitable for LTO 4380 processing, use gcc-ar and gcc-ranlib; to list symbols from a 4381 slim object file use gcc-nm. (This requires that ar, ranlib 4382 and nm have been compiled with plugin support.) 4383 Memory usage building Firefox with debug enabled was reduced from 4384 15GB to 3.5GB; link time from 1700 seconds to 350 seconds. 4385 * Inter-procedural optimization improvements: 4386 + New type inheritance analysis module improving 4387 devirtualization. Devirtualization now takes into account 4388 anonymous name-spaces and the C++11 final keyword. 4389 + New speculative devirtualization pass (controlled by 4390 -fdevirtualize-speculatively. 4391 + Calls that were speculatively made direct are turned back to 4392 indirect where direct call is not cheaper. 4393 + Local aliases are introduced for symbols that are known to be 4394 semantically equivalent across shared libraries improving 4395 dynamic linking times. 4396 * Feedback directed optimization improvements: 4397 + Profiling of programs using C++ inline functions is now more 4398 reliable. 4399 + New time profiling determines typical order in which functions 4400 are executed. 4401 + A new function reordering pass (controlled by 4402 -freorder-functions) significantly reduces startup time of 4403 large applications. Until binutils support is completed, it is 4404 effective only with link-time optimization. 4405 + Feedback driven indirect call removal and devirtualization now 4406 handle cross-module calls when link-time optimization is 4407 enabled. 4408 4409 New Languages and Language specific improvements 4410 4411 * Version 4.0 of the [4]OpenMP specification is now supported in the 4412 C and C++ compilers and starting with the 4.9.1 release also in the 4413 Fortran compiler. The new -fopenmp-simd option can be used to 4414 enable OpenMP's SIMD directives while ignoring other OpenMP 4415 directives. The new [5]-fsimd-cost-model= option permits to tune 4416 the vectorization cost model for loops annotated with OpenMP and 4417 Cilk Plus simd directives. -Wopenmp-simd warns when the current 4418 cost model overrides simd directives set by the user. 4419 * The -Wdate-time option has been added for the C, C++ and Fortran 4420 compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__ 4421 macros are used. Those macros might prevent bit-wise-identical 4422 reproducible compilations. 4423 4424 Ada 4425 4426 * GNAT switched to Ada 2012 instead of Ada 2005 by default. 4427 4428 C family 4429 4430 * Support for colorizing diagnostics emitted by GCC has been added. 4431 The [6]-fdiagnostics-color=auto will enable it when outputting to 4432 terminals, -fdiagnostics-color=always unconditionally. The 4433 GCC_COLORS environment variable can be used to customize the colors 4434 or disable coloring. If GCC_COLORS variable is present in the 4435 environment, the default is -fdiagnostics-color=auto, otherwise 4436 -fdiagnostics-color=never. 4437 Sample diagnostics output: 4438 $ g++ -fdiagnostics-color=always -S -Wall test.C 4439 test.C: In function int foo(): 4440 test.C:1:14: warning: no return statement in function returning non-void [-W 4441 return-type] 4442 int foo () { } 4443 ^ 4444 test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use 4445 -ftemplate-depth= to increase the maximum) instantiating struct X<100> 4446 template <int N> struct X { static const int value = X<N-1>::value; }; temp 4447 late struct X<1000>; 4448 ^ 4449 test.C:2:46: recursively required from const int X<999>::value 4450 test.C:2:46: required from const int X<1000>::value 4451 test.C:2:88: required from here 4452 4453 test.C:2:46: error: incomplete type X<100> used in nested name specifier 4454 4455 * With the new [7]#pragma GCC ivdep, the user can assert that there 4456 are no loop-carried dependencies which would prevent concurrent 4457 execution of consecutive iterations using SIMD (single instruction 4458 multiple data) instructions. 4459 * Support for [8]Cilk Plus has been added and can be enabled with the 4460 -fcilkplus option. Cilk Plus is an extension to the C and C++ 4461 languages to support data and task parallelism. The present 4462 implementation follows ABI version 1.2; all features but _Cilk_for 4463 have been implemented. 4464 4465 C 4466 4467 * ISO C11 atomics (the _Atomic type specifier and qualifier and the 4468 <stdatomic.h> header) are now supported. 4469 * ISO C11 generic selections (_Generic keyword) are now supported. 4470 * ISO C11 thread-local storage (_Thread_local, similar to GNU C 4471 __thread) is now supported. 4472 * ISO C11 support is now at a similar level of completeness to ISO 4473 C99 support: substantially complete modulo bugs, extended 4474 identifiers (supported except for corner cases when 4475 -fextended-identifiers is used), floating-point issues (mainly but 4476 not entirely relating to optional C99 features from Annexes F and 4477 G) and the optional Annexes K (Bounds-checking interfaces) and L 4478 (Analyzability). 4479 * A new C extension __auto_type provides a subset of the 4480 functionality of C++11 auto in GNU C. 4481 4482 C++ 4483 4484 * The G++ implementation of [9]C++1y return type deduction for normal 4485 functions has been updated to conform to [10]N3638, the proposal 4486 accepted into the working paper. Most notably, it adds 4487 decltype(auto) for getting decltype semantics rather than the 4488 template argument deduction semantics of plain auto: 4489 4490 int& f(); 4491 auto i1 = f(); // int 4492 decltype(auto) i2 = f(); // int& 4493 4494 * G++ supports [11]C++1y lambda capture initializers: 4495 4496 [x = 42]{ ... }; 4497 4498 Actually, they have been accepted since GCC 4.5, but now the 4499 compiler doesn't warn about them with -std=c++1y, and supports 4500 parenthesized and brace-enclosed initializers as well. 4501 * G++ supports [12]C++1y variable length arrays. G++ has supported 4502 GNU/C99-style VLAs for a long time, but now additionally supports 4503 initializers and lambda capture by reference. In C++1y mode G++ 4504 will complain about VLA uses that are not permitted by the draft 4505 standard, such as forming a pointer to VLA type or applying sizeof 4506 to a VLA variable. Note that it now appears that VLAs will not be 4507 part of C++14, but will be part of a separate document and then 4508 perhaps C++17. 4509 4510 void f(int n) { 4511 int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3 4512 [&a]{ for (int i : a) { cout << i << endl; } }(); 4513 &a; // error, taking address of VLA 4514 } 4515 4516 * G++ supports the [13]C++1y [[deprecated]] attribute modulo bugs in 4517 the underlying [[gnu::deprecated]] attribute. Classes and functions 4518 can be marked deprecated and a diagnostic message added: 4519 4520 class A; 4521 int bar(int n); 4522 #if __cplusplus > 201103 4523 class [[deprecated("A is deprecated in C++14; Use B instead")]] A; 4524 [[deprecated("bar is unsafe; use foo() instead")]] 4525 int bar(int n); 4526 4527 int foo(int n); 4528 class B; 4529 #endif 4530 A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead 4531 int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo 4532 o() instead 4533 4534 * G++ supports [14]C++1y digit separators. Long numeric literals can 4535 be subdivided with a single quote ' to enhance readability: 4536 4537 int i = 1048576; 4538 int j = 1'048'576; 4539 int k = 0x10'0000; 4540 int m = 0'004'000'000; 4541 int n = 0b0001'0000'0000'0000'0000'0000; 4542 4543 double x = 1.602'176'565e-19; 4544 double y = 1.602'176'565e-1'9; 4545 4546 * G++ supports [15]C++1y generic (polymorphic) lambdas. 4547 4548 // a functional object that will increment any type 4549 auto incr = [](auto x) { return x++; }; 4550 4551 * As a GNU extension, G++ supports explicit template parameter syntax 4552 for generic lambdas. This can be combined in the expected way with 4553 the standard auto syntax. 4554 4555 // a functional object that will add two like-type objects 4556 auto add = [] <typename T> (T a, T b) { return a + b; }; 4557 4558 * G++ supports unconstrained generic functions as specified by 4.1.2 4559 and 5.1.1 of [16]N3889: Concepts Lite Specification. Briefly, auto 4560 may be used as a type-specifier in a parameter declaration of any 4561 function declarator in order to introduce an implicit function 4562 template parameter, akin to generic lambdas. 4563 4564 // the following two function declarations are equivalent 4565 auto incr(auto x) { return x++; } 4566 template <typename T> 4567 auto incr(T x) { return x++; } 4568 4569 Runtime Library (libstdc++) 4570 4571 * [17]Improved support for C++11, including: 4572 + support for <regex>; 4573 + The associative containers in <map> and <set> and the 4574 unordered associative containers in <unordered_map> and 4575 <unordered_set> meet the allocator-aware container 4576 requirements; 4577 * [18]Improved experimental support for the upcoming ISO C++ 4578 standard, C++14, including: 4579 + fixing constexpr member functions without const; 4580 + implementation of the std::exchange() utility function; 4581 + addressing tuples by type; 4582 + implemention of std::make_unique; 4583 + implemention of std::shared_lock; 4584 + making std::result_of SFINAE-friendly; 4585 + adding operator() to std::integral_constant; 4586 + adding user-defined literals for standard library types 4587 std::basic_string, std::chrono::duration, and std::complex; 4588 + adding two range overloads to non-modifying sequence oprations 4589 std::equal and std::mismatch; 4590 + adding IO manipulators for quoted strings; 4591 + adding constexpr members to <utility>, <complex>, <chrono>, 4592 and some containers; 4593 + adding compile-time std::integer_sequence; 4594 + adding cleaner transformation traits; 4595 + making <functional>s operator functors easier to use and more 4596 generic; 4597 * An implementation of std::experimental::optional. 4598 * An implementation of std::experimental::string_view. 4599 * The non-standard function std::copy_exception has been deprecated 4600 and will be removed in a future version. std::make_exception_ptr 4601 should be used instead. 4602 4603 Fortran 4604 4605 * Compatibility notice: 4606 + Module files: The version of the module files (.mod) has been 4607 incremented; additionally, module files are now compressed. 4608 Fortran MODULEs compiled by earlier GCC versions have to be 4609 recompiled, when they are USEd by files compiled with GCC 4.9. 4610 GCC 4.9 is not able to read .mod files of earlier GCC 4611 versions; attempting to do so gives an error message. Note: 4612 The ABI of the produced assembler data itself has not changed: 4613 object files and libraries are fully compatible with older 4614 versions (except as stated below). 4615 + ABI changes: 4616 o The [19]argument passing ABI has changed for scalar dummy 4617 arguments of type INTEGER, REAL, COMPLEX and LOGICAL, 4618 which have both the VALUE and the OPTIONAL attributes. 4619 o To support finalization the virtual table associated with 4620 polymorphic variables has changed. Code containing CLASS 4621 should be recompiled, including all files which define 4622 derived types involved in the type definition used by 4623 polymorphic variables. (Note: Due to the incremented 4624 module version, trying to mix old code with new code will 4625 usually give an error message.) 4626 + GNU Fortran no longer deallocates allocatable variables or 4627 allocatable components of variables declared in the main 4628 program. Since Fortran 2008, the standard explicitly states 4629 that variables declared in the Fortran main program 4630 automatically have the SAVE attribute. 4631 + When opening files, the close-on-exec flag is set if the 4632 system supports such a feature. This is generally considered 4633 good practice these days, but if there is a need to pass file 4634 descriptors to child processes the parent process must now 4635 remember to clear the close-on-exec flag by calling fcntl(), 4636 e.g. via ISO_C_BINDING, before executing the child process. 4637 * The deprecated command-line option -fno-whole-file has been 4638 removed. (-fwhole-file is the default since GCC 4.6.) 4639 -fwhole-file/-fno-whole-file continue to be accepted but do not 4640 influence the code generation. 4641 * The compiler no longer unconditionally warns about DO loops with 4642 zero iterations. This warning is now controlled by the -Wzerotrip 4643 option, which is implied by -Wall. 4644 * The new NO_ARG_CHECK attribute of the [20]!GCC$ directive can be 4645 used to disable the type-kind-rank (TKR) argument check for a dummy 4646 argument. The feature is similar to ISO/IEC TS 29133:2012's 4647 TYPE(*), except that it additionally also disables the rank check. 4648 Variables with NO_ARG_CHECK have to be dummy arguments and may only 4649 be used as argument to ISO_C_BINDING's C_LOC and as actual argument 4650 to another NO_ARG_CHECK dummy argument; also the other constraints 4651 of TYPE(*) apply. The dummy arguments should be declared as scalar 4652 or assumed-size variable of type type(*) (recommended) or of type 4653 integer, real, complex or logical. With NO_ARG_CHECK, a pointer to 4654 the data without further type or shape information is passed, 4655 similar to C's void*. Note that also TS 29113's 4656 type(*),dimension(..) accepts arguments of any type and rank; 4657 contrary to NO_ARG_CHECK assumed-rank arguments pass an array 4658 descriptor which contains the array shape and stride of the 4659 argument. 4660 * [21]Fortran 2003: 4661 + Finalization is now supported. It is currently only done for a 4662 subset of those situations in which it should occur. 4663 + Experimental support for scalar character components with 4664 deferred length (i.e. allocatable string length) in derived 4665 types has been added. (Deferred-length character variables are 4666 supported since GCC 4.6.) 4667 * [22]Fortran 2008: 4668 + When STOP or ERROR STOP are used to terminate the execution 4669 and any exception (but inexact) is signaling, a warning is 4670 printed to ERROR_UNIT, indicating which exceptions are 4671 signaling. The [23]-ffpe-summary= command-line option can be 4672 used to fine-tune for which exceptions the warning should be 4673 shown. 4674 + Rounding on input (READ) is now handled on systems where 4675 strtod honours the rounding mode. (For output, rounding is 4676 supported since GCC 4.5.) Note that for input, the compatible 4677 rounding mode is handled as nearest (i.e., rounding to an even 4678 least significant [cf. IEC 60559:1989] for a tie, while 4679 compatible rounds away from zero in that case). 4680 4681 Go 4682 4683 * GCC 4.9 provides a complete implementation of the Go 1.2.1 release. 4684 4685 New Targets and Target Specific Improvements 4686 4687 AArch64 4688 4689 * The ARMv8-A crypto and CRC instructions are now supported through 4690 intrinsics. These are enabled when the architecture supports these 4691 and are available through the -march=armv8-a+crc and 4692 -march=armv8-a+crypto options. 4693 * Initial support for ILP32 has now been added to the compiler. This 4694 is now available through the command-line option -mabi=ilp32. 4695 Support for ILP32 is considered experimental as the ABI 4696 specification is still beta. 4697 * Coverage of more of the ISA including the SIMD extensions has been 4698 added. The Advanced SIMD intrinsics have also been improved. 4699 * The new local register allocator (LRA) is now on by default for the 4700 AArch64 backend. 4701 * The REE (Redundant extension elimination) pass has now been enabled 4702 by default for the AArch64 backend. 4703 * Tuning for the Cortex-A53 and Cortex-A57 has been improved. 4704 * Initial big.LITTLE tuning support for the combination of Cortex-A57 4705 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 4706 option. 4707 * A number of structural changes have been made to both the ARM and 4708 AArch64 backends to facilitate improved code-generation. 4709 * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769 4710 has been added and can be enabled by giving the 4711 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 4712 default by configuring GCC with the --enable-fix-cortex-a53-835769 4713 option. 4714 4715 ARC 4716 4717 * A port for Synopsys Designware ARC has been contributed by Embecosm 4718 and Synopsys Inc. 4719 4720 ARM 4721 4722 * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been 4723 disabled by default. This was found to generate better code in only 4724 a small number of cases. It can be turned back on with the 4725 -mneon-for-64bits option. 4726 * Further support for the ARMv8-A architecture, notably implementing 4727 the restriction around IT blocks in the Thumb32 instruction set has 4728 been added. The -mrestrict-it option can be used with 4729 -march=armv7-a or the -march=armv7ve options to make code 4730 generation fully compatible with the deprecated instructions in 4731 ARMv8-A. 4732 * Support has now been added for the ARMv7ve variant of the 4733 architecture. This can be used by the -march=armv7ve option. 4734 * The ARMv8-A crypto and CRC instructions are now supported through 4735 intrinsics and are available through the -march=armv8-a+crc and 4736 mfpu=crypto-neon-fp-armv8 options. 4737 * LRA is now on by default for the ARM target. This can be turned off 4738 using the -mno-lra option. This option is a purely transitionary 4739 command-line option and will be removed in a future release. We are 4740 interested in any bug reports regarding functional and performance 4741 regressions with LRA. 4742 * A new option -mslow-flash-data to improve performance of programs 4743 fetching data on slow flash memory has now been introduced for the 4744 ARMv7-M profile cores. 4745 * A new option -mpic-data-is-text-relative for targets that allows 4746 data segments to be relative to text segments has been added. This 4747 is on by default for all targets except VxWorks RTP. 4748 * A number of infrastructural changes have been made to both the ARM 4749 and AArch64 backends to facilitate improved code-generation. 4750 * GCC now supports Cortex-A12 and the Cortex-R7 through the 4751 -mcpu=cortex-a12 and -mcpu=cortex-r7 options. 4752 * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the 4753 -mcpu=cortex-a57 and -mcpu=cortex-a53 options. 4754 * Initial big.LITTLE tuning support for the combination of Cortex-A57 4755 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 4756 option. Similar support was added for the combination of Cortex-A15 4757 and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option. 4758 * Further performance optimizations for the Cortex-A15 and the 4759 Cortex-M4 have been added. 4760 * A number of code generation improvements for Thumb2 to reduce code 4761 size when compiling for the M-profile processors. 4762 4763 AVR 4764 4765 * A new command-line option -mfract-convert-truncate has been added. 4766 It allows compiler to use truncation instead of rounding towards 4767 zero for fractional fixed-point types. 4768 4769 IA-32/x86-64 4770 4771 * -mfpmath=sse is now implied by -ffast-math on all targets where 4772 SSE2 is supported. 4773 * Intel AVX-512 support was added to GCC. That includes inline 4774 assembly support, new registers and extending existing ones, new 4775 intrinsics (covered by corresponding testsuite), and basic 4776 autovectorization. AVX-512 instructions are available via the 4777 following GCC switches: AVX-512 foundation instructions: -mavx512f, 4778 AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and 4779 reciprocal instructions: -mavx512er, AVX-512 conflict detection 4780 instructions: -mavx512cd. 4781 * It is now possible to call x86 intrinsics from select functions in 4782 a file that are tagged with the corresponding target attribute 4783 without having to compile the entire file with the -mxxx option. 4784 This improves the usability of x86 intrinsics and is particularly 4785 useful when doing [24]Function Multiversioning. 4786 * GCC now supports the new Intel microarchitecture named Silvermont 4787 through -march=silvermont. 4788 * GCC now supports the new Intel microarchitecture named Broadwell 4789 through -march=broadwell. 4790 * Optimizing for other Intel microarchitectures have been renamed to 4791 -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell. 4792 * -march=generic has been retuned for better support of Intel core 4793 and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel 4794 Pentium-M, and Pentium4 based CPUs is no longer considered 4795 important for generic. 4796 * -mtune=intel can now be used to generate code running well on the 4797 most current Intel processors, which are Haswell and Silvermont for 4798 GCC 4.9. 4799 * Support to encode 32-bit assembly instructions in 16-bit format is 4800 now available through the -m16 command-line option. 4801 * Better inlining of memcpy and memset that is aware of value ranges 4802 and produces shorter alignment prologues. 4803 * -mno-accumulate-outgoing-args is now honored when unwind 4804 information is output. Argument accumulation is also now turned off 4805 for portions of programs optimized for size. 4806 * Support for new AMD family 15h processors (Excavator core) is now 4807 available through the -march=bdver4 and -mtune=bdver4 options. 4808 4809 MSP430 4810 4811 * A new command-line option -mcpu= has been added to the MSP430 4812 backend. This option is used to specify the ISA to be used. 4813 Accepted values are msp430 (the default), msp430x and msp430xv2. 4814 The ISA is no longer deduced from the -mmcu= option as there are 4815 far too many different MCU names. The -mmcu= option is still 4816 supported, and this is still used to select linker scripts and 4817 generate a C preprocessor symbol that will be recognised by the 4818 msp430.h header file. 4819 4820 NDS32 4821 4822 * A new nds32 port supports the 32-bit architecture from Andes 4823 Technology Corporation. 4824 * The port provides initial support for the V2, V3, V3m instruction 4825 set architectures. 4826 4827 Nios II 4828 4829 * A port for the Altera Nios II has been contributed by Mentor 4830 Graphics. 4831 4832 PowerPC / PowerPC64 / RS6000 4833 4834 * GCC now supports Power ISA 2.07, which includes support for 4835 Hardware Transactional Memory (HTM), Quadword atomics and several 4836 VMX and VSX additions, including Crypto, 64-bit integer, 128-bit 4837 integer and decimal integer operations. 4838 * Support for the POWER8 processor is now available through the 4839 -mcpu=power8 and -mtune=power8 options. 4840 * The libitm library has been modified to add a HTM fastpath that 4841 automatically uses POWER's HTM hardware instructions when it is 4842 executing on a HTM enabled processor. 4843 * Support for the new powerpc64le-linux platform has been added. It 4844 defaults to generating code that conforms to the ELFV2 ABI. 4845 4846 S/390, System z 4847 4848 * Support for the Transactional Execution Facility included with the 4849 IBM zEnterprise zEC12 processor has been added. A set of GCC style 4850 builtins as well as XLC style builtins are provided. The builtins 4851 are enabled by default when using the -march=zEC12 option but can 4852 explicitly be disabled with -mno-htm. Using the GCC builtins also 4853 libitm supports hardware transactions on S/390. 4854 * The hotpatch features allows to prepare functions for hotpatching. 4855 A certain amount of bytes is reserved before the function entry 4856 label plus a NOP is inserted at its very beginning to implement a 4857 backward jump when applying a patch. The feature can either be 4858 enabled per compilation unit via the command-line option -mhotpatch 4859 or per function using the hotpatch attribute. 4860 * The shrink wrap optimization is now supported on S/390 and enabled 4861 by default. 4862 * A major rework of the routines to determine which registers need to 4863 be saved and restored in function prologue/epilogue now allow to 4864 use floating point registers as save slots. This will happen for 4865 certain leaf function with -march=z10 or higher. 4866 * The LRA rtl pass replaces reload by default on S/390. 4867 4868 RX 4869 4870 * The port now allows to specify the RX100, RX200, and RX600 4871 processors with the command-line options -mcpu=rx100, -mcpu=rx200 4872 and -mcpu=rx600. 4873 4874 SH 4875 4876 * Minor improvements to code generated for integer arithmetic and 4877 code that involves the T bit. 4878 * Added support for the SH2A clips and clipu instructions. The 4879 compiler will now try to utilize them for min/max expressions such 4880 as max (-128, min (127, x)). 4881 * Added support for the cmp/str instruction through built-in 4882 functions such as __builtin_strlen. When not optimizing for size, 4883 the compiler will now expand calls to e.g. strlen as an inlined 4884 sequences which utilize the cmp/str instruction. 4885 * Improved code generated around volatile memory loads and stores. 4886 * The option -mcbranchdi has been deprecated. Specifying it will 4887 result in a warning and will not influence code generation. 4888 * The option -mcmpeqdi has been deprecated. Specifying it will result 4889 in a warning and will not influence code generation. 4890 4891 GCC 4.9.1 4892 4893 This is the [25]list of problem reports (PRs) from GCC's bug tracking 4894 system that are known to be fixed in the 4.9.1 release. This list might 4895 not be complete (that is, it is possible that some PRs that have been 4896 fixed are not listed here). 4897 4898 Version 4.0 of the OpenMP specification is supported even in Fortran, 4899 not just C and C++. 4900 4901 GCC 4.9.2 4902 4903 This is the [26]list of problem reports (PRs) from GCC's bug tracking 4904 system that are known to be fixed in the 4.9.2 release. This list might 4905 not be complete (that is, it is possible that some PRs that have been 4906 fixed are not listed here). 4907 4908 GCC 4.9.3 4909 4910 This is the [27]list of problem reports (PRs) from GCC's bug tracking 4911 system that are known to be fixed in the 4.9.3 release. This list might 4912 not be complete (that is, it is possible that some PRs that have been 4913 fixed are not listed here). 4914 4915 GCC 4.9.4 4916 4917 This is the [28]list of problem reports (PRs) from GCC's bug tracking 4918 system that are known to be fixed in the 4.9.4 release. This list might 4919 not be complete (that is, it is possible that some PRs that have been 4920 fixed are not listed here). 4921 4922 4923 For questions related to the use of GCC, please consult these web 4924 pages and the [29]GCC manuals. If that fails, the 4925 [30]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4926 web pages and the development of GCC are welcome on our developer 4927 list at [31]gcc (a] gcc.gnu.org. All of [32]our lists have public 4928 archives. 4929 4930 Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and 4931 distribution of this entire article is permitted in any medium, 4932 provided this notice is preserved. 4933 4934 These pages are [34]maintained by the GCC team. Last modified 4935 2018-12-01[35]. 4936 4937 References 4938 4939 1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html 4940 2. https://gcc.gnu.org/PR60825 4941 3. https://gcc.gnu.org/gcc-4.9/porting_to.html 4942 4. https://www.openmp.org/specifications/ 4943 5. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908 4944 6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252 4945 7. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html 4946 8. https://www.cilkplus.org/ 4947 9. http://gcc.gnu.org/projects/cxx1y.html 4948 10. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html 4949 11. http://gcc.gnu.org/projects/cxx1y.html 4950 12. http://gcc.gnu.org/projects/cxx1y.html 4951 13. http://gcc.gnu.org/projects/cxx1y.html 4952 14. http://gcc.gnu.org/projects/cxx1y.html 4953 15. http://gcc.gnu.org/projects/cxx1y.html 4954 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf 4955 17. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011 4956 18. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014 4957 19. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html 4958 20. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html 4959 21. https://gcc.gnu.org/wiki/Fortran2003Status 4960 22. https://gcc.gnu.org/wiki/Fortran2008Status 4961 23. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html 4962 24. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html 4963 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1 4964 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2 4965 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3 4966 28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.4 4967 29. https://gcc.gnu.org/onlinedocs/ 4968 30. mailto:gcc-help (a] gcc.gnu.org 4969 31. mailto:gcc (a] gcc.gnu.org 4970 32. https://gcc.gnu.org/lists.html 4971 33. https://www.fsf.org/ 4972 34. https://gcc.gnu.org/about.html 4973 35. http://validator.w3.org/check/referer 4974 ====================================================================== 4975 http://gcc.gnu.org/gcc-4.8/index.html 4976 GCC 4.8 Release Series 4977 4978 June 23, 2015 4979 4980 The [1]GNU project and the GCC developers are pleased to announce the 4981 release of GCC 4.8.5. 4982 4983 This release is a bug-fix release, containing fixes for regressions in 4984 GCC 4.8.4 relative to previous releases of GCC. 4985 4986 This release series is no longer maintained. 4987 4988 Release History 4989 4990 GCC 4.8.5 4991 June 23, 2015 ([2]changes, [3]documentation) 4992 4993 GCC 4.8.4 4994 December 19, 2014 ([4]changes, [5]documentation) 4995 4996 GCC 4.8.3 4997 May 22, 2014 ([6]changes, [7]documentation) 4998 4999 GCC 4.8.2 5000 October 16, 2013 ([8]changes, [9]documentation) 5001 5002 GCC 4.8.1 5003 May 31, 2013 ([10]changes, [11]documentation) 5004 5005 GCC 4.8.0 5006 March 22, 2013 ([12]changes, [13]documentation) 5007 5008 References and Acknowledgements 5009 5010 GCC used to stand for the GNU C Compiler, but since the compiler 5011 supports several other languages aside from C, it now stands for the 5012 GNU Compiler Collection. 5013 5014 A list of [14]successful builds is updated as new information becomes 5015 available. 5016 5017 The GCC developers would like to thank the numerous people that have 5018 contributed new features, improvements, bug fixes, and other changes as 5019 well as test results to GCC. This [15]amazing group of volunteers is 5020 what makes GCC successful. 5021 5022 For additional information about GCC please refer to the [16]GCC 5023 project web site or contact the [17]GCC development mailing list. 5024 5025 To obtain GCC please use [18]our mirror sites or [19]our SVN server. 5026 5027 5028 For questions related to the use of GCC, please consult these web 5029 pages and the [20]GCC manuals. If that fails, the 5030 [21]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5031 web pages and the development of GCC are welcome on our developer 5032 list at [22]gcc (a] gcc.gnu.org. All of [23]our lists have public 5033 archives. 5034 5035 Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and 5036 distribution of this entire article is permitted in any medium, 5037 provided this notice is preserved. 5038 5039 These pages are [25]maintained by the GCC team. Last modified 5040 2018-09-30[26]. 5041 5042 References 5043 5044 1. http://www.gnu.org/ 5045 2. http://gcc.gnu.org/gcc-4.8/changes.html 5046 3. https://gcc.gnu.org/onlinedocs/4.8.5/ 5047 4. http://gcc.gnu.org/gcc-4.8/changes.html 5048 5. https://gcc.gnu.org/onlinedocs/4.8.4/ 5049 6. http://gcc.gnu.org/gcc-4.8/changes.html 5050 7. https://gcc.gnu.org/onlinedocs/4.8.3/ 5051 8. http://gcc.gnu.org/gcc-4.8/changes.html 5052 9. https://gcc.gnu.org/onlinedocs/4.8.2/ 5053 10. http://gcc.gnu.org/gcc-4.8/changes.html 5054 11. https://gcc.gnu.org/onlinedocs/4.8.1/ 5055 12. http://gcc.gnu.org/gcc-4.8/changes.html 5056 13. https://gcc.gnu.org/onlinedocs/4.8.0/ 5057 14. http://gcc.gnu.org/gcc-4.8/buildstat.html 5058 15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 5059 16. http://gcc.gnu.org/index.html 5060 17. mailto:gcc (a] gcc.gnu.org 5061 18. http://gcc.gnu.org/mirrors.html 5062 19. http://gcc.gnu.org/svn.html 5063 20. https://gcc.gnu.org/onlinedocs/ 5064 21. mailto:gcc-help (a] gcc.gnu.org 5065 22. mailto:gcc (a] gcc.gnu.org 5066 23. https://gcc.gnu.org/lists.html 5067 24. https://www.fsf.org/ 5068 25. https://gcc.gnu.org/about.html 5069 26. http://validator.w3.org/check/referer 5070 ====================================================================== 5071 http://gcc.gnu.org/gcc-4.8/changes.html 5072 GCC 4.8 Release Series 5073 Changes, New Features, and Fixes 5074 5075 Caveats 5076 5077 GCC now uses C++ as its implementation language. This means that to 5078 build GCC from sources, you will need a C++ compiler that understands 5079 C++ 2003. For more details on the rationale and specific changes, 5080 please refer to the [1]C++ conversion page. 5081 5082 To enable the Graphite framework for loop optimizations you now need 5083 CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from 5084 the [2]GCC infrastructure directory. The installation manual contains 5085 more information about requirements to build GCC. 5086 5087 GCC now uses a more aggressive analysis to derive an upper bound for 5088 the number of iterations of loops using constraints imposed by language 5089 standards. This may cause non-conforming programs to no longer work as 5090 expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new 5091 option, -fno-aggressive-loop-optimizations, was added to disable this 5092 aggressive analysis. In some loops that have known constant number of 5093 iterations, but undefined behavior is known to occur in the loop before 5094 reaching or during the last iteration, GCC will warn about the 5095 undefined behavior in the loop instead of deriving lower upper bound of 5096 the number of iterations for the loop. The warning can be disabled with 5097 -Wno-aggressive-loop-optimizations. 5098 5099 On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules 5100 for the layout of vectors that could lead to wrong code being 5101 generated. Vectors larger than 8 bytes in size are now by default 5102 aligned to an 8-byte boundary. This is an ABI change: code that makes 5103 explicit use of vector types may be incompatible with binary objects 5104 built with older versions of GCC. Auto-vectorized code is not affected 5105 by this change. 5106 5107 On AVR, support has been removed for the command-line option 5108 -mshort-calls deprecated in GCC 4.7. 5109 5110 On AVR, the configure option --with-avrlibc supported since GCC 4.7.2 5111 is turned on per default for all non-RTEMS configurations. This option 5112 arranges for a better integration of [3]AVR Libc with avr-gcc. For 5113 technical details, see [4]PR54461. To turn off the option in non-RTEMS 5114 configurations, use --with-avrlibc=no. If the compiler is configured 5115 for RTEMS, the option is always turned off. 5116 5117 More information on porting to GCC 4.8 from previous versions of GCC 5118 can be found in the [5]porting guide for this release. 5119 5120 General Optimizer Improvements (and Changes) 5121 5122 * DWARF4 is now the default when generating DWARF debug information. 5123 When -g is used on a platform that uses DWARF debugging 5124 information, GCC will now default to -gdwarf-4 5125 -fno-debug-types-section. 5126 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information 5127 consumers support DWARF4 by default. Before GCC 4.8 the default 5128 version used was DWARF2. To make GCC 4.8 generate an older DWARF 5129 version use -g together with -gdwarf-2 or -gdwarf-3. The default 5130 for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf. 5131 * A new general optimization level, -Og, has been introduced. It 5132 addresses the need for fast compilation and a superior debugging 5133 experience while providing a reasonable level of run-time 5134 performance. Overall experience for development should be better 5135 than the default optimization level -O0. 5136 * A new option -ftree-partial-pre was added to control the partial 5137 redundancy elimination (PRE) optimization. This option is enabled 5138 by default at the -O3 optimization level, and it makes PRE more 5139 aggressive. 5140 * The option -fconserve-space has been removed; it was no longer 5141 useful on most targets since GCC supports putting variables into 5142 BSS without making them common. 5143 * The struct reorg and matrix reorg optimizations (command-line 5144 options -fipa-struct-reorg and -fipa-matrix-reorg) have been 5145 removed. They did not always work correctly, nor did they work with 5146 link-time optimization (LTO), hence were only applicable to 5147 programs consisting of a single translation unit. 5148 * Several scalability bottle-necks have been removed from GCC's 5149 optimization passes. Compilation of extremely large functions, e.g. 5150 due to the use of the flatten attribute in the "Eigen" C++ linear 5151 algebra templates library, is significantly faster than previous 5152 releases of GCC. 5153 * Link-time optimization (LTO) improvements: 5154 + LTO partitioning has been rewritten for better reliability and 5155 maintanibility. Several important bugs leading to link 5156 failures have been fixed. 5157 * Interprocedural optimization improvements: 5158 + A new symbol table has been implemented. It builds on existing 5159 callgraph and varpool modules and provide a new API. Unusual 5160 symbol visibilities and aliases are handled more consistently 5161 leading to, for example, more aggressive unreachable code 5162 removal with LTO. 5163 + The inline heuristic can now bypass limits on the size of of 5164 inlined functions when the inlining is particularly 5165 profitable. This happens, for example, when loop bounds or 5166 array strides get propagated. 5167 + Values passed through aggregates (either by value or 5168 reference) are now propagated at the inter-procedural level 5169 leading to better inlining decisions (for example in the case 5170 of Fortran array descriptors) and devirtualization. 5171 * [6]AddressSanitizer , a fast memory error detector, has been added 5172 and can be enabled via -fsanitize=address. Memory access 5173 instructions will be instrumented to detect heap-, stack-, and 5174 global-buffer overflow as well as use-after-free bugs. To get nicer 5175 stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is 5176 available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on 5177 x86-64 Darwin. 5178 * [7]ThreadSanitizer has been added and can be enabled via 5179 -fsanitize=thread. Instructions will be instrumented to detect data 5180 races. The ThreadSanitizer is available on x86-64 GNU/Linux. 5181 * A new local register allocator (LRA) has been implemented, which 5182 replaces the 26 year old reload pass and improves generated code 5183 quality. For now it is active on the IA-32 and x86-64 targets. 5184 * Support for transactional memory has been implemented on the 5185 following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and 5186 Alpha. 5187 5188 New Languages and Language specific improvements 5189 5190 C family 5191 5192 * Each diagnostic emitted now includes the original source line and a 5193 caret '^' indicating the column. The option 5194 -fno-diagnostics-show-caret suppresses this information. 5195 * The option -ftrack-macro-expansion=2 is now enabled by default. 5196 This allows the compiler to display the macro expansion stack in 5197 diagnostics. Combined with the caret information, an example 5198 diagnostic showing these two features is: 5199 5200 t.c:1:94: error: invalid operands to binary < (have struct mystruct and float 5201 ) 5202 #define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _ 5203 _b = (B); __a < __b ? __b : __a; }) 5204 5205 ^ 5206 t.c:7:7: note: in expansion of macro 'MYMAX' 5207 X = MYMAX(P, F); 5208 ^ 5209 5210 * A new -Wsizeof-pointer-memaccess warning has been added (also 5211 enabled by -Wall) to warn about suspicious length parameters to 5212 certain string and memory built-in functions if the argument uses 5213 sizeof. This warning warns e.g. about memset (ptr, 0, sizeof 5214 (ptr)); if ptr is not an array, but a pointer, and suggests a 5215 possible fix, or about memcpy (&foo, ptr, sizeof (&foo));. 5216 * The new option -Wpedantic is an alias for -pedantic, which is now 5217 deprecated. The forms -Wno-pedantic, -Werror=pedantic, and 5218 -Wno-error=pedantic work in the same way as for any other -W 5219 option. One caveat is that -Werror=pedantic is not equivalent to 5220 -pedantic-errors, since the latter makes into errors some warnings 5221 that are not controlled by -Wpedantic, and the former only affects 5222 diagnostics that are disabled when using -Wno-pedantic. 5223 * The option -Wshadow no longer warns if a declaration shadows a 5224 function declaration, unless the former declares a function or 5225 pointer to function, because this is [8]a common and valid case in 5226 real-world code. 5227 5228 C++ 5229 5230 * G++ now implements the [9]C++11 thread_local keyword; this differs 5231 from the GNU __thread keyword primarily in that it allows dynamic 5232 initialization and destruction semantics. Unfortunately, this 5233 support requires a run-time penalty for references to 5234 non-function-local thread_local variables defined in a different 5235 translation unit even if they don't need dynamic initialization, so 5236 users may want to continue to use __thread for TLS variables with 5237 static initialization semantics. 5238 If the programmer can be sure that no use of the variable in a 5239 non-defining TU needs to trigger dynamic initialization (either 5240 because the variable is statically initialized, or a use of the 5241 variable in the defining TU will be executed before any uses in 5242 another TU), they can avoid this overhead with the 5243 -fno-extern-tls-init option. 5244 OpenMP threadprivate variables now also support dynamic 5245 initialization and destruction by the same mechanism. 5246 * G++ now implements the [10]C++11 attribute syntax, e.g. 5247 5248 [[noreturn]] void f(); 5249 5250 and also the alignment specifier, e.g. 5251 5252 alignas(double) int i; 5253 5254 * G++ now implements [11]C++11 inheriting constructors, e.g. 5255 5256 struct A { A(int); }; 5257 struct B: A { using A::A; }; // defines B::B(int) 5258 B b(42); // OK 5259 5260 * As of GCC 4.8.1, G++ implements the change to decltype semantics 5261 from [12]N3276. 5262 5263 struct A f(); 5264 decltype(f()) g(); // OK, return type of f() is not required to be complete. 5265 5266 * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g. 5267 5268 struct A { int f() &; }; 5269 int i = A().f(); // error, f() requires an lvalue object 5270 5271 * G++ now supports a -std=c++1y option for experimentation with 5272 features proposed for the next revision of the standard, expected 5273 around 2014. Currently the only difference from -std=c++11 is 5274 support for return type deduction in normal functions, as proposed 5275 in [14]N3386. Status of C++1y features in GCC 4.8 can be found 5276 [15]here. 5277 * The G++ namespace association extension, __attribute ((strong)), 5278 has been deprecated. Inline namespaces should be used instead. 5279 * G++ now supports a -fext-numeric-literal option to control whether 5280 GNU numeric literal suffixes are accepted as extensions or 5281 processed as C++11 user-defined numeric literal suffixes. The flag 5282 is on (use suffixes for GNU literals) by default for -std=gnu++*, 5283 and -std=c++98. The flag is off (use suffixes for user-defined 5284 literals) by default for -std=c++11 and later. 5285 5286 Runtime Library (libstdc++) 5287 5288 * [16]Improved experimental support for the new ISO C++ standard, 5289 C++11, including: 5290 + forward_list meets the allocator-aware container requirements; 5291 + this_thread::sleep_for(), this_thread::sleep_until() and 5292 this_thread::yield() are defined without requiring the 5293 configure option --enable-libstdcxx-time; 5294 * Improvements to <random>: 5295 + SSE optimized normal_distribution. 5296 + Use of hardware RNG instruction for random_device on new x86 5297 processors (requires the assembler to support the 5298 instruction.) 5299 and <ext/random>: 5300 + New random number engine simd_fast_mersenne_twister_engine 5301 with an optimized SSE implementation. 5302 + New random number distributions beta_distribution, 5303 normal_mv_distribution, rice_distribution, 5304 nakagami_distribution, pareto_distribution, k_distribution, 5305 arcsine_distribution, hoyt_distribution. 5306 * Added --disable-libstdcxx-verbose configure option to disable 5307 diagnostic messages issued when a process terminates abnormally. 5308 This may be useful for embedded systems to reduce the size of 5309 executables that link statically to the library. 5310 5311 Fortran 5312 5313 * Compatibility notice: 5314 + Module files: The version of module files (.mod) has been 5315 incremented. Fortran MODULEs compiled by earlier GCC versions 5316 have to be recompiled, when they are USEd by files compiled 5317 with GCC 4.8. GCC 4.8 is not able to read .mod files created 5318 by earlier versions; attempting to do so gives an error 5319 message. 5320 Note: The ABI of the produced assembler data itself has not 5321 changed; object files and libraries are fully compatible with 5322 older versions except as noted below. 5323 + ABI: Some internal names (used in the assembler/object file) 5324 have changed for symbols declared in the specification part of 5325 a module. If an affected module or a file using it via use 5326 association is recompiled, the module and all files which 5327 directly use such symbols have to be recompiled as well. This 5328 change only affects the following kind of module symbols: 5329 o Procedure pointers. Note: C-interoperable function 5330 pointers (type(c_funptr)) are not affected nor are 5331 procedure-pointer components. 5332 o Deferred-length character strings. 5333 * The [17]BACKTRACE intrinsic subroutine has been added. It shows a 5334 backtrace at an arbitrary place in user code; program execution 5335 continues normally afterwards. 5336 * The [18]-Wc-binding-type warning option has been added (disabled by 5337 default). It warns if the a variable might not be C interoperable; 5338 in particular, if the variable has been declared using an intrinsic 5339 type with default kind instead of using a kind parameter defined 5340 for C interoperability in the intrinsic ISO_C_Binding module. 5341 Before, this warning was always printed. The -Wc-binding-type 5342 option is enabled by -Wall. 5343 * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line 5344 options have been added, which diagnose when code is inserted for 5345 automatic (re)allocation of a variable during assignment. This 5346 option can be used to decide whether it is safe to use 5347 [20]-fno-realloc-lhs. Additionally, it can be used to find 5348 automatic (re)allocation in hot loops. (For arrays, replacing 5349 "var=" by "var(:)=" disables the automatic reallocation.) 5350 * The [21]-Wcompare-reals command-line option has been added. When 5351 this is set, warnings are issued when comparing REAL or COMPLEX 5352 types for equality and inequality; consider replacing a == b by 5353 abs(ab) < eps with a suitable eps. -Wcompare-reals is enabled by 5354 -Wextra. 5355 * The [22]-Wtarget-lifetime command-line option has been added 5356 (enabled with -Wall), which warns if the pointer in a pointer 5357 assignment might outlive its target. 5358 * Reading floating point numbers which use "q" for the exponential 5359 (such as 4.0q0) is now supported as vendor extension for better 5360 compatibility with old data files. It is strongly recommended to 5361 use for I/O the equivalent but standard conforming "e" (such as 5362 4.0e0). 5363 (For Fortran source code, consider replacing the "q" in 5364 floating-point literals by a kind parameter (e.g. 4.0e0_qp with a 5365 suitable qp). Note that in Fortran source code replacing "q" by 5366 a simple "e" is not equivalent.) 5367 * The GFORTRAN_TMPDIR environment variable for specifying a 5368 non-default directory for files opened with STATUS="SCRATCH", is 5369 not used anymore. Instead gfortran checks the POSIX/GNU standard 5370 TMPDIR environment variable. If TMPDIR is not defined, gfortran 5371 falls back to other methods to determine the directory for 5372 temporary files as documented in the [23]user manual. 5373 * [24]Fortran 2003: 5374 + Support for unlimited polymorphic variables (CLASS(*)) has 5375 been added. Nonconstant character lengths are not yet 5376 supported. 5377 * [25]TS 29113: 5378 + Assumed types (TYPE(*)) are now supported. 5379 + Experimental support for assumed-rank arrays (dimension(..)) 5380 has been added. Note that currently gfortran's own array 5381 descriptor is used, which is different from the one defined in 5382 TS29113, see [26]gfortran's header file or use the [27]Chasm 5383 Language Interoperability Tools. 5384 5385 Go 5386 5387 * GCC 4.8.2 provides a complete implementation of the Go 1.1.2 5388 release. 5389 * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1 5390 release. The library support is not quite complete. 5391 * Go has been tested on GNU/Linux and Solaris platforms for various 5392 processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may 5393 work on other platforms as well. 5394 5395 New Targets and Target Specific Improvements 5396 5397 AArch64 5398 5399 * A new port has been added to support AArch64, the new 64-bit 5400 architecture from ARM. Note that this is a separate port from the 5401 existing 32-bit ARM port. 5402 * The port provides initial support for the Cortex-A53 and the 5403 Cortex-A57 processors with the command line options 5404 -mcpu=cortex-a53 and -mcpu=cortex-a57. 5405 * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769 5406 has been added and can be enabled by giving the 5407 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 5408 default by configuring GCC with the --enable-fix-cortex-a53-835769 5409 option. 5410 5411 ARM 5412 5413 * Initial support has been added for the AArch32 extensions defined 5414 in the ARMv8 architecture. 5415 * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs. 5416 * A new option, -mcpu=marvell-pj4, has been added to generate code 5417 for the Marvell PJ4 processor. 5418 * The compiler can now automatically generate the VFMA, VFMS, REVSH 5419 and REV16 instructions. 5420 * A new vectorizer cost model for Advanced SIMD configurations to 5421 improve the auto-vectorization strategies used. 5422 * The scheduler now takes into account the number of live registers 5423 to reduce the amount of spilling that can occur. This should 5424 improve code performance in large functions. The limit can be 5425 removed by using the option -fno-sched-pressure. 5426 * Improvements have been made to the Marvell iWMMX code generation 5427 and support for the iWMMX2 SIMD unit has been added. The option 5428 -mcpu=iwmmxt2 can be used to enable code generation for the latter. 5429 * A number of code generation improvements for Thumb2 to reduce code 5430 size when compiling for the M-profile processors. 5431 * The RTEMS (arm-rtems) port has been updated to use the EABI. 5432 * Code generation support for the old FPA and Maverick floating-point 5433 architectures has been removed. Ports that previously relied on 5434 these features have also been removed. This includes the targets: 5435 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 5436 + arm*-*-elf (use arm*-*-eabi) 5437 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 5438 + arm*-*-ecos-elf (no alternative) 5439 + arm*-*-freebsd (no alternative) 5440 + arm*-wince-pe* (no alternative). 5441 5442 AVR 5443 5444 * Support for the "Embedded C" fixed-point has been added. For 5445 details, see the [28]GCC wiki and the [29]user manual. The support 5446 is not complete. 5447 * A new print modifier %r for register operands in inline assembler 5448 is supported. It will print the raw register number without the 5449 register prefix 'r': 5450 /* Return the most significant byte of 'val', a 64-bit value. */ 5451 5452 unsigned char msb (long long val) 5453 { 5454 unsigned char c; 5455 __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val)); 5456 return c; 5457 } 5458 The inline assembler in this example will generate code like 5459 mov r24, 8+7 5460 provided c is allocated to R24 and val is allocated to R8R15. This 5461 works because the GNU assembler accepts plain register numbers 5462 without register prefix. 5463 * Static initializers with 3-byte symbols are supported now: 5464 extern const __memx char foo; 5465 const __memx void *pfoo = &foo; 5466 This requires at least Binutils 2.23. 5467 5468 IA-32/x86-64 5469 5470 * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with 5471 SSE extensions disabled. Since the x86-64 ABI requires 16 byte 5472 stack alignment, this is ABI incompatible and intended to be used 5473 in controlled environments where stack space is an important 5474 limitation. This option will lead to wrong code when functions 5475 compiled with 16 byte stack alignment (such as functions from a 5476 standard library) are called with misaligned stack. In this case, 5477 SSE instructions may lead to misaligned memory access traps. In 5478 addition, variable arguments will be handled incorrectly for 16 5479 byte aligned objects (including x87 long double and __int128), 5480 leading to wrong results. You must build all modules with 5481 -mpreferred-stack-boundary=3, including any libraries. This 5482 includes the system libraries and startup modules. 5483 * Support for the new Intel processor codename Broadwell with RDSEED, 5484 ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw, 5485 -mrdseed command-line options. 5486 * Support for the Intel RTM and HLE intrinsics, built-in functions 5487 and code generation is available via -mrtm and -mhle. 5488 * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets. 5489 Intrinsics and built-in functions are available via -mfxsr, -mxsave 5490 and -mxsaveopt respectively. 5491 * New -maddress-mode=[short|long] options for x32. 5492 -maddress-mode=short overrides default 64-bit addresses to 32-bit 5493 by emitting the 0x67 address-size override prefix. This is the 5494 default address mode for x32. 5495 * New built-in functions to detect run-time CPU type and ISA: 5496 + A built-in function __builtin_cpu_is has been added to detect 5497 if the run-time CPU is of a particular type. It returns a 5498 positive integer on a match and zero otherwise. It accepts one 5499 string literal argument, the CPU name. For example, 5500 __builtin_cpu_is("westmere") returns a positive integer if the 5501 run-time CPU is an Intel Core i7 Westmere processor. Please 5502 refer to the [30]user manual for the list of valid CPU names 5503 recognized. 5504 + A built-in function __builtin_cpu_supports has been added to 5505 detect if the run-time CPU supports a particular ISA feature. 5506 It returns a positive integer on a match and zero otherwise. 5507 It accepts one string literal argument, the ISA feature. For 5508 example, __builtin_cpu_supports("ssse3") returns a positive 5509 integer if the run-time CPU supports SSSE3 instructions. 5510 Please refer to the [31]user manual for the list of valid ISA 5511 names recognized. 5512 Caveat: If these built-in functions are called before any static 5513 constructors are invoked, like during IFUNC initialization, then 5514 the CPU detection initialization must be explicitly run using this 5515 newly provided built-in function, __builtin_cpu_init. The 5516 initialization needs to be done only once. For example, this is how 5517 the invocation would look like inside an IFUNC initializer: 5518 static void (*some_ifunc_resolver(void))(void) 5519 { 5520 __builtin_cpu_init(); 5521 if (__builtin_cpu_is("amdfam10h") ... 5522 if (__builtin_cpu_supports("popcnt") ... 5523 } 5524 5525 * Function Multiversioning Support with G++: 5526 It is now possible to create multiple function versions each 5527 targeting a specific processor and/or ISA. Function versions have 5528 the same signature but different target attributes. For example, 5529 here is a program with function versions: 5530 __attribute__ ((target ("default"))) 5531 int foo(void) 5532 { 5533 return 1; 5534 } 5535 5536 __attribute__ ((target ("sse4.2"))) 5537 int foo(void) 5538 { 5539 return 2; 5540 } 5541 5542 int main (void) 5543 { 5544 int (*p) = &foo; 5545 assert ((*p)() == foo()); 5546 return 0; 5547 } 5548 5549 Please refer to this [32]wiki for more information. 5550 * The x86 back end has been improved to allow option -fschedule-insns 5551 to work reliably. This option can be used to schedule instructions 5552 better and leads to improved performace in certain cases. 5553 * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437 5554 from the Mingw-w64 trunk. 5555 * Support for new AMD family 15h processors (Steamroller core) is now 5556 available through the -march=bdver3 and -mtune=bdver3 options. 5557 * Support for new AMD family 16h processors (Jaguar core) is now 5558 available through the -march=btver2 and -mtune=btver2 options. 5559 5560 FRV 5561 5562 * This target now supports the -fstack-usage command-line option. 5563 5564 MIPS 5565 5566 * GCC can now generate code specifically for the R4700, Broadcom XLP 5567 and MIPS 34kn processors. The associated -march options are 5568 -march=r4700, -march=xlp and -march=34kn respectively. 5569 * GCC now generates better DSP code for MIPS 74k cores thanks to 5570 further scheduling optimizations. 5571 * The MIPS port now supports the -fstack-check option. 5572 * GCC now passes the -mmcu and -mno-mcu options to the assembler. 5573 * Previous versions of GCC would silently accept -fpic and -fPIC for 5574 -mno-abicalls targets like mips*-elf. This combination was not 5575 intended or supported, and did not generate position-independent 5576 code. GCC 4.8 now reports an error when this combination is used. 5577 5578 PowerPC / PowerPC64 / RS6000 5579 5580 * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save, 5581 restore or update the VRSAVE register by default. The respective 5582 operating systems manage the VRSAVE register directly. 5583 * Large TOC support has been added for AIX through the command line 5584 option -mcmodel=large. 5585 * Native Thread-Local Storage support has been added for AIX. 5586 * VMX (Altivec) and VSX instruction sets now are enabled implicitly 5587 when targetting processors that support those hardware features on 5588 AIX 6.1 and above. 5589 5590 RX 5591 5592 * This target will now issue a warning message whenever multiple fast 5593 interrupt handlers are found in the same compilation unit. This 5594 feature can be turned off by the new 5595 -mno-warn-multiple-fast-interrupts command-line option. 5596 5597 S/390, System z 5598 5599 * Support for the IBM zEnterprise zEC12 processor has been added. 5600 When using the -march=zEC12 option, the compiler will generate code 5601 making use of the following new instructions: 5602 + load and trap instructions 5603 + 2 new compare and trap instructions 5604 + rotate and insert selected bits - without CC clobber 5605 The -mtune=zEC12 option enables zEC12 specific instruction 5606 scheduling without making use of new instructions. 5607 * Register pressure sensitive instruction scheduling is enabled by 5608 default. 5609 * The ifunc function attribute is enabled by default. 5610 * memcpy and memcmp invokations on big memory chunks or with run time 5611 lengths are not generated inline anymore when tuning for z10 or 5612 higher. The purpose is to make use of the IFUNC optimized versions 5613 in Glibc. 5614 5615 SH 5616 5617 * The default alignment settings have been reduced to be less 5618 aggressive. This results in more compact code for optimization 5619 levels other than -Os. 5620 * Improved support for the __atomic built-in functions: 5621 + A new option -matomic-model=model selects the model for the 5622 generated atomic sequences. The following models are 5623 supported: 5624 5625 soft-gusa 5626 Software gUSA sequences (SH3* and SH4* only). On 5627 SH4A targets this will now also partially utilize 5628 the movco.l and movli.l instructions. This is the 5629 default when the target is sh3*-*-linux* or 5630 sh4*-*-linux*. 5631 5632 hard-llcs 5633 Hardware movco.l / movli.l sequences (SH4A only). 5634 5635 soft-tcb 5636 Software thread control block sequences. 5637 5638 soft-imask 5639 Software interrupt flipping sequences (privileged 5640 mode only). This is the default when the target is 5641 sh1*-*-linux* or sh2*-*-linux*. 5642 5643 none 5644 Generates function calls to the respective __atomic 5645 built-in functions. This is the default for SH64 5646 targets or when the target is not sh*-*-linux*. 5647 5648 + The option -msoft-atomic has been deprecated. It is now an 5649 alias for -matomic-model=soft-gusa. 5650 + A new option -mtas makes the compiler generate the tas.b 5651 instruction for the __atomic_test_and_set built-in function 5652 regardless of the selected atomic model. 5653 + The __sync functions in libgcc now reflect the selected atomic 5654 model when building the toolchain. 5655 * Added support for the mov.b and mov.w instructions with 5656 displacement addressing. 5657 * Added support for the SH2A instructions movu.b and movu.w. 5658 * Various improvements to code generated for integer arithmetic. 5659 * Improvements to conditional branches and code that involves the T 5660 bit. A new option -mzdcbranch tells the compiler to favor 5661 zero-displacement branches. This is enabled by default for SH4* 5662 targets. 5663 * The pref instruction will now be emitted by the __builtin_prefetch 5664 built-in function for SH3* targets. 5665 * The fmac instruction will now be emitted by the fmaf standard 5666 function and the __builtin_fmaf built-in function. 5667 * The -mfused-madd option has been deprecated in favor of the 5668 machine-independent -ffp-contract option. Notice that the fmac 5669 instruction will now be generated by default for expressions like a 5670 * b + c. This is due to the compiler default setting 5671 -ffp-contract=fast. 5672 * Added new options -mfsrra and -mfsca to allow the compiler using 5673 the fsrra and fsca instructions on targets other than SH4A (where 5674 they are already enabled by default). 5675 * Added support for the __builtin_bswap32 built-in function. It is 5676 now expanded as a sequence of swap.b and swap.w instructions 5677 instead of a library function call. 5678 * The behavior of the -mieee option has been fixed and the negative 5679 form -mno-ieee has been added to control the IEEE conformance of 5680 floating point comparisons. By default -mieee is now enabled and 5681 the option -ffinite-math-only implicitly sets -mno-ieee. 5682 * Added support for the built-in functions __builtin_thread_pointer 5683 and __builtin_set_thread_pointer. This assumes that GBR is used to 5684 hold the thread pointer of the current thread. Memory loads and 5685 stores relative to the address returned by __builtin_thread_pointer 5686 will now also utilize GBR based displacement address modes. 5687 * The -mdiv= option for targets other than SHmedia has been fixed and 5688 documented. 5689 5690 SPARC 5691 5692 * Added optimized instruction scheduling for Niagara4. 5693 5694 TILE-Gx 5695 5696 * Added support for the -mcmodel=MODEL command-line option. The 5697 models supported are small and large. 5698 5699 V850 5700 5701 * This target now supports the E3V5 architecture via the use of the 5702 new -mv850e3v5 command-line option. It also has experimental 5703 support for the e3v5 LOOP instruction which can be enabled via the 5704 new -mloop command-line option. 5705 5706 XStormy16 5707 5708 * This target now supports the -fstack-usage command-line option. 5709 5710 Operating Systems 5711 5712 OpenBSD 5713 5714 * Support for OpenBSD/amd64 (x86_64-*-openbsd*) has been added and 5715 support for OpenBSD/i386 (i386-*-openbsd*) has been rejuvenated. 5716 5717 Windows (Cygwin) 5718 5719 * Executables are now linked against shared libgcc by default. The 5720 previous default was to link statically, which can still be done by 5721 explicitly specifying -static or static-libgcc on the command line. 5722 However it is strongly advised against, as it will cause problems 5723 for any application that makes use of DLLs compiled by GCC. It 5724 should be alright for a monolithic stand-alone application that 5725 only links against the Windows DLLs, but offers little or no 5726 benefit. 5727 5728 GCC 4.8.1 5729 5730 This is the [33]list of problem reports (PRs) from GCC's bug tracking 5731 system that are known to be fixed in the 4.8.1 release. This list might 5732 not be complete (that is, it is possible that some PRs that have been 5733 fixed are not listed here). 5734 5735 The C++11 <chrono> std::chrono::system_clock and 5736 std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they 5737 both are now separate (never typedefs of each other), both use 5738 std::chrono::nanoseconds resolution, on most GNU/Linux configurations 5739 std::chrono::steady_clock is now finally monotonic, and both classes 5740 are mangled differently than in the previous GCC releases. 5741 std::chrono::system_clock::now() with std::chrono::microseconds resp. 5742 std::chrono::seconds resolution is still exported for backwards 5743 compatibility with default configured libstdc++. Note that libstdc++ 5744 configured with --enable-libstdcxx-time= used to be ABI incompatible 5745 with default configured libstdc++ for those two classes and no ABI 5746 compatibility can be offered for those configurations, so any C++11 5747 code that uses those classes and has been compiled and linked against 5748 libstdc++ configured with the non-default --enable-libstdcxx-time= 5749 configuration option needs to be recompiled. 5750 5751 GCC 4.8.2 5752 5753 This is the [34]list of problem reports (PRs) from GCC's bug tracking 5754 system that are known to be fixed in the 4.8.2 release. This list might 5755 not be complete (that is, it is possible that some PRs that have been 5756 fixed are not listed here). 5757 5758 GCC 4.8.3 5759 5760 This is the [35]list of problem reports (PRs) from GCC's bug tracking 5761 system that are known to be fixed in the 4.8.3 release. This list might 5762 not be complete (that is, it is possible that some PRs that have been 5763 fixed are not listed here). 5764 5765 Support for the new powerpc64le-linux platform has been added. It 5766 defaults to generating code that conforms to the ELFV2 ABI. 5767 5768 GCC 4.8.4 5769 5770 This is the [36]list of problem reports (PRs) from GCC's bug tracking 5771 system that are known to be fixed in the 4.8.4 release. This list might 5772 not be complete (that is, it is possible that some PRs that have been 5773 fixed are not listed here). 5774 5775 GCC 4.8.5 5776 5777 This is the [37]list of problem reports (PRs) from GCC's bug tracking 5778 system that are known to be fixed in the 4.8.5 release. This list might 5779 not be complete (that is, it is possible that some PRs that have been 5780 fixed are not listed here). 5781 5782 5783 For questions related to the use of GCC, please consult these web 5784 pages and the [38]GCC manuals. If that fails, the 5785 [39]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5786 web pages and the development of GCC are welcome on our developer 5787 list at [40]gcc (a] gcc.gnu.org. All of [41]our lists have public 5788 archives. 5789 5790 Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and 5791 distribution of this entire article is permitted in any medium, 5792 provided this notice is preserved. 5793 5794 These pages are [43]maintained by the GCC team. Last modified 5795 2018-09-30[44]. 5796 5797 References 5798 5799 1. https://gcc.gnu.org/wiki/cxx-conversion 5800 2. ftp://gcc.gnu.org/pub/gcc/infrastructure/ 5801 3. http://www.nongnu.org/avr-libc/ 5802 4. https://gcc.gnu.org/PR54461 5803 5. https://gcc.gnu.org/gcc-4.8/porting_to.html 5804 6. https://github.com/google/sanitizers 5805 7. https://code.google.com/archive/p/data-race-test/wikis/ThreadSanitizer.wiki 5806 8. https://lkml.org/lkml/2006/11/28/239 5807 9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 5808 10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 5809 11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 5810 12. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf 5811 13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 5812 14. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html 5813 15. http://gcc.gnu.org/projects/cxx1y.html 5814 16. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011 5815 17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html 5816 18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 5817 19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 5818 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 5819 21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 5820 22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 5821 23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html 5822 24. https://gcc.gnu.org/wiki/Fortran2003Status 5823 25. https://gcc.gnu.org/wiki/TS29113Status 5824 26. https://gcc.gnu.org/viewcvs/trunk/libgfortran/libgfortran.h?content-type=text/plain&view=co 5825 27. http://chasm-interop.sourceforge.net/ 5826 28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support 5827 29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html 5828 30. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 5829 31. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 5830 32. https://gcc.gnu.org/wiki/FunctionMultiVersioning 5831 33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1 5832 34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2 5833 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3 5834 36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4 5835 37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5 5836 38. https://gcc.gnu.org/onlinedocs/ 5837 39. mailto:gcc-help (a] gcc.gnu.org 5838 40. mailto:gcc (a] gcc.gnu.org 5839 41. https://gcc.gnu.org/lists.html 5840 42. https://www.fsf.org/ 5841 43. https://gcc.gnu.org/about.html 5842 44. http://validator.w3.org/check/referer 5843 ====================================================================== 5844 http://gcc.gnu.org/gcc-4.7/index.html 5845 GCC 4.7 Release Series 5846 5847 June 12, 2014 5848 5849 The [1]GNU project and the GCC developers are pleased to announce the 5850 release of GCC 4.7.4. 5851 5852 This release is a bug-fix release, containing fixes for regressions in 5853 GCC 4.7.3 relative to previous releases of GCC. 5854 5855 This release series is no longer maintained. 5856 5857 Release History 5858 5859 GCC 4.7.4 5860 June 12, 2014 ([2]changes, [3]documentation) 5861 5862 GCC 4.7.3 5863 April 11, 2013 ([4]changes, [5]documentation) 5864 5865 GCC 4.7.2 5866 September 20, 2012 ([6]changes, [7]documentation) 5867 5868 GCC 4.7.1 5869 June 14, 2012 ([8]changes, [9]documentation) 5870 5871 GCC 4.7.0 5872 March 22, 2012 ([10]changes, [11]documentation) 5873 5874 References and Acknowledgements 5875 5876 GCC used to stand for the GNU C Compiler, but since the compiler 5877 supports several other languages aside from C, it now stands for the 5878 GNU Compiler Collection. 5879 5880 A list of [12]successful builds is updated as new information becomes 5881 available. 5882 5883 The GCC developers would like to thank the numerous people that have 5884 contributed new features, improvements, bug fixes, and other changes as 5885 well as test results to GCC. This [13]amazing group of volunteers is 5886 what makes GCC successful. 5887 5888 For additional information about GCC please refer to the [14]GCC 5889 project web site or contact the [15]GCC development mailing list. 5890 5891 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 5892 5893 5894 For questions related to the use of GCC, please consult these web 5895 pages and the [18]GCC manuals. If that fails, the 5896 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5897 web pages and the development of GCC are welcome on our developer 5898 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 5899 archives. 5900 5901 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 5902 distribution of this entire article is permitted in any medium, 5903 provided this notice is preserved. 5904 5905 These pages are [23]maintained by the GCC team. Last modified 5906 2018-09-30[24]. 5907 5908 References 5909 5910 1. http://www.gnu.org/ 5911 2. http://gcc.gnu.org/gcc-4.7/changes.html 5912 3. https://gcc.gnu.org/onlinedocs/4.7.4/ 5913 4. http://gcc.gnu.org/gcc-4.7/changes.html 5914 5. https://gcc.gnu.org/onlinedocs/4.7.3/ 5915 6. http://gcc.gnu.org/gcc-4.7/changes.html 5916 7. https://gcc.gnu.org/onlinedocs/4.7.2/ 5917 8. http://gcc.gnu.org/gcc-4.7/changes.html 5918 9. https://gcc.gnu.org/onlinedocs/4.7.1/ 5919 10. http://gcc.gnu.org/gcc-4.7/changes.html 5920 11. https://gcc.gnu.org/onlinedocs/4.7.0/ 5921 12. http://gcc.gnu.org/gcc-4.7/buildstat.html 5922 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 5923 14. http://gcc.gnu.org/index.html 5924 15. mailto:gcc (a] gcc.gnu.org 5925 16. http://gcc.gnu.org/mirrors.html 5926 17. http://gcc.gnu.org/svn.html 5927 18. https://gcc.gnu.org/onlinedocs/ 5928 19. mailto:gcc-help (a] gcc.gnu.org 5929 20. mailto:gcc (a] gcc.gnu.org 5930 21. https://gcc.gnu.org/lists.html 5931 22. https://www.fsf.org/ 5932 23. https://gcc.gnu.org/about.html 5933 24. http://validator.w3.org/check/referer 5934 ====================================================================== 5935 http://gcc.gnu.org/gcc-4.7/changes.html 5936 GCC 4.7 Release Series 5937 Changes, New Features, and Fixes 5938 5939 Caveats 5940 5941 * The -fconserve-space flag has been deprecated. The flag had no 5942 effect for most targets: only targets without a global .bss section 5943 and without support for switchable sections. Furthermore, the flag 5944 only had an effect for G++, where it could result in wrong 5945 semantics (please refer to the GCC manual for further details). The 5946 flag will be removed in GCC 4.8 5947 * Support for a number of older systems and recently unmaintained or 5948 untested target ports of GCC has been declared obsolete in GCC 4.7. 5949 Unless there is activity to revive them, the next release of GCC 5950 will have their sources permanently removed. 5951 All GCC ports for the following processor architectures have been 5952 declared obsolete: 5953 + picoChip (picochip-*) 5954 The following ports for individual systems on particular 5955 architectures have been obsoleted: 5956 + IRIX 6.5 (mips-sgi-irix6.5) 5957 + MIPS OpenBSD (mips*-*-openbsd*) 5958 + Solaris 8 (*-*-solaris2.8). Details can be found in the 5959 [1]announcement. 5960 + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*) 5961 * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A, 5962 ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by 5963 default, which for some sources generates code that accesses memory 5964 on unaligned addresses. This requires the kernel of those systems 5965 to enable such accesses (controlled by CP15 register c1, refer to 5966 ARM documentation). Alternatively, or for compatibility with 5967 kernels where unaligned accesses are not supported, all code has to 5968 be compiled with -mno-unaligned-access. Upstream Linux kernel 5969 releases have automatically and unconditionally supported unaligned 5970 accesses as emitted by GCC due to this option being active since 5971 version 2.6.28. 5972 * Support on ARM for the legacy floating-point accelerator (FPA) and 5973 the mixed-endian floating-point format that it used has been 5974 obsoleted. The ports that still use this format have been obsoleted 5975 as well. Many legacy ARM ports already provide an alternative that 5976 uses the VFP floating-point format. The obsolete ports will be 5977 deleted in the next release. 5978 The obsolete ports with alternatives are: 5979 + arm*-*-rtems (use arm*-*-rtemseabi) 5980 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 5981 + arm*-*-elf (use arm*-*-eabi) 5982 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 5983 Note, however, that these alternatives are not binary compatible 5984 with their legacy counterparts (although some can support running 5985 legacy applications). 5986 The obsolete ports that currently lack a modern alternative are: 5987 + arm*-*-ecos-elf 5988 + arm*-*-freebsd 5989 + arm*-wince-pe* 5990 New ports that support more recent versions of the architecture are 5991 welcome. 5992 * Support for the Maverick co-processor on ARM has been obsoleted. 5993 Code to support it will be deleted in the next release. 5994 * Support has been removed for Unix International threads on Solaris 5995 2, so the --enable-threads=solaris configure option and the 5996 -threads compiler option don't work any longer. 5997 * Support has been removed for the Solaris BSD Compatibility Package, 5998 which lives in /usr/ucbinclude and /usr/ucblib. It has been removed 5999 from Solaris 11, and was only intended as a migration aid from 6000 SunOS 4 to SunOS 5. The -compat-bsd compiler option is not 6001 recognized any longer. 6002 * The AVR port's libgcc has been improved and its multilib structure 6003 has been enhanced. As a result, all objects contributing to an 6004 application must either be compiled with GCC versions up to 4.6.x 6005 or with GCC versions 4.7.1 or later. If the compiler is used with 6006 AVR Libc, you need a version that supports the new layout, i.e. 6007 implements [2]#35407. 6008 * The AVR port's -mshort-calls command-line option has been 6009 deprecated. It will be removed in the GCC 4.8 release. See -mrelax 6010 for a replacement. 6011 * The AVR port only references startup code that clears .bss and the 6012 common section resp. initializes the .data and .rodata section 6013 provided respective sections (or subsections thereof) are not 6014 empty, see [3]PR18145. Applications that put all static storage 6015 objects into non-standard sections and / or define all static 6016 storage objects in assembler modules, must reference __do_clear_bss 6017 resp. __do_copy_data by hand or undefine the symbol(s) by means of 6018 -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data. 6019 * The ARM port's -mwords-little-endian option has been deprecated. It 6020 will be removed in a future release. 6021 * Support has been removed for the NetWare x86 configuration 6022 obsoleted in GCC 4.6. 6023 * It is no longer possible to use the "l" constraint in MIPS16 asm 6024 statements. 6025 * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard 6026 library which affected the ABI in C++11 mode: a data member was 6027 added to std::list changing its size and altering the definitions 6028 of some member functions, and std::pair's move constructor was 6029 non-trivial which altered the calling convention for functions with 6030 std::pair arguments or return types. The ABI incompatibilities have 6031 been fixed for GCC version 4.7.2 but as a result C++11 code 6032 compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11 6033 code compiled with different GCC versions and with C++98/C++03 code 6034 compiled with any version. 6035 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 6036 rules for the layout of vectors that could lead to wrong code being 6037 generated. Vectors larger than 8 bytes in size are now by default 6038 aligned to an 8-byte boundary. This is an ABI change: code that 6039 makes explicit use of vector types may be incompatible with binary 6040 objects built with older versions of GCC. Auto-vectorized code is 6041 not affected by this change. (This change affects GCC versions 6042 4.7.2 and later.) 6043 * More information on porting to GCC 4.7 from previous versions of 6044 GCC can be found in the [4]porting guide for this release. 6045 6046 General Optimizer Improvements 6047 6048 * Support for a new parameter --param case-values-threshold=n was 6049 added to allow users to control the cutoff between doing switch 6050 statements as a series of if statements and using a jump table. 6051 * Link-time optimization (LTO) improvements: 6052 + Improved scalability and reduced memory usage. Link time 6053 optimization of Firefox now requires 3GB of RAM on a 64-bit 6054 system, while over 8GB was needed previously. Linking time has 6055 been improved, too. The serial stage of linking Firefox has 6056 been sped up by about a factor of 10. 6057 + Reduced size of object files and temporary storage used during 6058 linking. 6059 + Streaming performance (both outbound and inbound) has been 6060 improved. 6061 + ld -r is now supported with LTO. 6062 + Several bug fixes, especially in symbol table handling and 6063 merging. 6064 * Interprocedural optimization improvements: 6065 + Heuristics now take into account that after inlining code will 6066 be optimized out because of known values (or properties) of 6067 function parameters. For example: 6068 void foo(int a) 6069 { 6070 if (a > 10) 6071 ... huge code ... 6072 } 6073 void bar (void) 6074 { 6075 foo (0); 6076 } 6077 6078 The call of foo will be inlined into bar even when optimizing 6079 for code size. Constructs based on __builtin_constant_p are 6080 now understood by the inliner and code size estimates are 6081 evaluated a lot more realistically. 6082 + The representation of C++ virtual thunks and aliases (both 6083 implicit and defined via the alias attribute) has been 6084 re-engineered. Aliases no longer pose optimization barriers 6085 and calls to an alias can be inlined and otherwise optimized. 6086 + The inter-procedural constant propagation pass has been 6087 rewritten. It now performs generic function specialization. 6088 For example when compiling the following: 6089 void foo(bool flag) 6090 { 6091 if (flag) 6092 ... do something ... 6093 else 6094 ... do something else ... 6095 } 6096 void bar (void) 6097 { 6098 foo (false); 6099 foo (true); 6100 foo (false); 6101 foo (true); 6102 foo (false); 6103 foo (true); 6104 } 6105 6106 GCC will now produce two copies of foo. One with flag being 6107 true, while other with flag being false. This leads to 6108 performance improvements previously possible only by inlining 6109 all calls. Cloning causes a lot less code size growth. 6110 * A string length optimization pass has been added. It attempts to 6111 track string lengths and optimize various standard C string 6112 functions like strlen, strchr, strcpy, strcat, stpcpy and their 6113 _FORTIFY_SOURCE counterparts into faster alternatives. This pass is 6114 enabled by default at -O2 or above, unless optimizing for size, and 6115 can be disabled by the -fno-optimize-strlen option. The pass can 6116 e.g. optimize 6117 char *bar (const char *a) 6118 { 6119 size_t l = strlen (a) + 2; 6120 char *p = malloc (l); if (p == NULL) return p; 6121 strcpy (p, a); strcat (p, "/"); return p; 6122 } 6123 6124 into: 6125 char *bar (const char *a) 6126 { 6127 size_t tmp = strlen (a); 6128 char *p = malloc (tmp + 2); if (p == NULL) return p; 6129 memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p; 6130 } 6131 6132 or for hosted compilations where stpcpy is available in the runtime 6133 and headers provide its prototype, e.g. 6134 void foo (char *a, const char *b, const char *c, const char *d) 6135 { 6136 strcpy (a, b); strcat (a, c); strcat (a, d); 6137 } 6138 6139 can be optimized into: 6140 void foo (char *a, const char *b, const char *c, const char *d) 6141 { 6142 strcpy (stpcpy (stpcpy (a, b), c), d); 6143 } 6144 6145 New Languages and Language specific improvements 6146 6147 * Version 3.1 of the OpenMP specification is now supported for the C, 6148 C++, and Fortran compilers. 6149 6150 Ada 6151 6152 * The command-line option -feliminate-unused-debug-types has been 6153 re-enabled by default, as it is for the other languages, leading to 6154 a reduction in debug info size of 12.5% and more for relevant 6155 cases, as well as to a small compilation speedup. 6156 6157 C family 6158 6159 * A new built-in, __builtin_assume_aligned, has been added, through 6160 which the compiler can be hinted about pointer alignment and can 6161 use it to improve generated code. 6162 * A new warning option -Wunused-local-typedefs was added for C, C++, 6163 Objective-C and Objective-C++. This warning diagnoses typedefs 6164 locally defined in a function, and otherwise not used. 6165 * A new experimental command-line option -ftrack-macro-expansion was 6166 added for C, C++, Objective-C, Objective-C++ and Fortran. It allows 6167 the compiler to emit diagnostic about the current macro expansion 6168 stack when a compilation error occurs in a macro expansion. 6169 * Experimental support for transactional memory has been added. It 6170 includes support in the compiler, as well as a supporting runtime 6171 library called libitm. To compile code with transactional memory 6172 constructs, use the -fgnu-tm option. 6173 Support is currently available for Alpha, ARM, PowerPC, SH, SPARC, 6174 and 32-bit/64-bit x86 platforms. 6175 For more details on transactional memory see [5]the GCC WiKi. 6176 * Support for atomic operations specifying the C++11/C11 memory model 6177 has been added. These new __atomic routines replace the existing 6178 __sync built-in routines. 6179 Atomic support is also available for memory blocks. Lock-free 6180 instructions will be used if a memory block is the same size and 6181 alignment as a supported integer type. Atomic operations which do 6182 not have lock-free support are left as function calls. A set of 6183 library functions is available on the GCC atomic wiki in the 6184 "External Atomics Library" section. 6185 For more details on the memory models and features, see the 6186 [6]atomic wiki. 6187 * When a binary operation is performed on vector types and one of the 6188 operands is a uniform vector, it is possible to replace the vector 6189 with the generating element. For example: 6190 typedef int v4si __attribute__ ((vector_size (16))); 6191 v4si res, a = {1,2,3,4}; 6192 int x; 6193 6194 res = 2 + a; /* means {2,2,2,2} + a */ 6195 res = a - x; /* means a - {x,x,x,x} */ 6196 6197 C 6198 6199 * There is support for some more features from the C11 revision of 6200 the ISO C standard. GCC now accepts the options -std=c11 and 6201 -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x. 6202 + Unicode strings (previously supported only with options such 6203 as -std=gnu11, now supported with -std=c11), and the 6204 predefined macros __STDC_UTF_16__ and __STDC_UTF_32__. 6205 + Nonreturning functions (_Noreturn and <stdnoreturn.h>). 6206 + Alignment support (_Alignas, _Alignof, max_align_t, 6207 <stdalign.h>). 6208 + A built-in function __builtin_complex is provided to support C 6209 library implementation of the CMPLX family of macros. 6210 6211 C++ 6212 6213 * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat 6214 options, which are equivalent to -std=c++0x, -std=gnu++0x, and 6215 -Wc++0x-compat, respectively. 6216 * G++ now implements [7]C++11 extended friend syntax: 6217 6218 template<class W> 6219 class Q 6220 { 6221 static const int I = 2; 6222 public: 6223 friend W; 6224 }; 6225 6226 struct B 6227 { 6228 int ar[Q<B>::I]; 6229 }; 6230 6231 * Thanks to Ville Voutilainen, G++ now implements [8]C++11 explicit 6232 override control. 6233 6234 struct B { 6235 virtual void f() const final; 6236 virtual void f(int); 6237 }; 6238 6239 struct D : B { 6240 void f() const; // error: D::f attempts to override final B::f 6241 void f(long) override; // error: doesn't override anything 6242 void f(int) override; // ok 6243 }; 6244 6245 struct E final { }; 6246 struct F: E { }; // error: deriving from final class 6247 6248 * G++ now implements [9]C++11 non-static data member initializers. 6249 6250 struct A { 6251 int i = 42; 6252 } a; // initializes a.i to 42 6253 6254 * Thanks to Ed Smith-Rowland, G++ now implements [10]C++11 6255 user-defined literals. 6256 6257 // Not actually a good approximation. :) 6258 constexpr long double operator"" _degrees (long double d) { return d * 0.0175; } 6259 long double pi = 180.0_degrees; 6260 6261 * G++ now implements [11]C++11 alias-declarations. 6262 6263 template <class T> using Ptr = T*; 6264 Ptr<int> ip; // decltype(ip) is int* 6265 6266 * Thanks to Ville Voutilainen and Pedro Lamaro, G++ now implements 6267 [12]C++11 delegating constructors. 6268 6269 struct A { 6270 A(int); 6271 A(): A(42) { } // delegate to the A(int) constructor 6272 }; 6273 6274 * G++ now fully implements C++11 atomic classes rather than just 6275 integer derived classes. 6276 6277 class POD { 6278 int a; 6279 int b; 6280 }; 6281 std::atomic<POD> my_atomic_POD; 6282 6283 * G++ now sets the predefined macro __cplusplus to the correct value, 6284 199711L for C++98/03, and 201103L for C++11. 6285 * G++ now correctly implements the two-phase lookup rules such that 6286 an unqualified name used in a template must have an appropriate 6287 declaration found either in scope at the point of definition of the 6288 template or by argument-dependent lookup at the point of 6289 instantiation. As a result, code that relies on a second 6290 unqualified lookup at the point of instantiation to find functions 6291 declared after the template or in dependent bases will be rejected. 6292 The compiler will suggest ways to fix affected code, and using the 6293 -fpermissive compiler flag will allow the code to compile with a 6294 warning. 6295 6296 template <class T> 6297 void f() { g(T()); } // error, g(int) not found by argument-dependent lookup 6298 void g(int) { } // fix by moving this declaration before the declaration of f 6299 6300 template <class T> 6301 struct A: T { 6302 // error, B::g(B) not found by argument-dependent lookup 6303 void f() { g(T()); } // fix by using this->g or A::g 6304 }; 6305 6306 struct B { void g(B); }; 6307 6308 int main() 6309 { 6310 f<int>(); 6311 A<B>().f(); 6312 } 6313 6314 * G++ now properly re-uses stack space allocated for temporary 6315 objects when their lifetime ends, which can significantly lower 6316 stack consumption for some C++ functions. As a result of this, some 6317 code with undefined behavior will now break: 6318 6319 const int &f(const int &i) { return i; } 6320 .... 6321 const int &x = f(1); 6322 const int &y = f(2); 6323 6324 Here, x refers to the temporary allocated to hold the 1 argument, 6325 which only lives until the end of the initialization; it 6326 immediately becomes a dangling reference. So the next statement 6327 re-uses the stack slot to hold the 2 argument, and users of x get 6328 that value instead. 6329 Note that this should not cause any change of behavior for 6330 temporaries of types with non-trivial destructors, as they are 6331 already destroyed at end of full-expression; the change is that now 6332 the storage is released as well. 6333 * A new command-line option -Wdelete-non-virtual-dtor has been added 6334 to warn when delete is used to destroy an instance of a class which 6335 has virtual functions and non-virtual destructor. It is unsafe to 6336 delete an instance of a derived class through a pointer to a base 6337 class if the base class does not have a virtual destructor. This 6338 warning is enabled by -Wall. 6339 * A new command-line option -Wzero-as-null-pointer-constant has been 6340 added to warn when a literal '0' is used as null pointer constant. 6341 It can be useful to facilitate the conversion to nullptr in C++11. 6342 * As per C++98, access-declarations are now deprecated by G++. 6343 Using-declarations are to be used instead. Furthermore, some 6344 efforts have been made to improve the support of class scope 6345 using-declarations. In particular, using-declarations referring to 6346 a dependent type now work as expected ([13]bug c++/14258). 6347 * The ELF symbol visibility of a template instantiation is now 6348 properly constrained by the visibility of its template arguments 6349 ([14]bug c++/35688). 6350 6351 Runtime Library (libstdc++) 6352 6353 * [15]Improved experimental support for the new ISO C++ standard, 6354 C++11, including: 6355 + using noexcept in most of the library; 6356 + implementations of pointer_traits, allocator_traits and 6357 scoped_allocator_adaptor; 6358 + uses-allocator construction for tuple; 6359 + vector meets the allocator-aware container requirements; 6360 + replacing monotonic_clock with steady_clock; 6361 + enabling the thread support library on most POSIX targets; 6362 + many small improvements to conform to the FDIS. 6363 * Added --enable-clocale=newlib configure option. 6364 * Debug Mode iterators for unordered associative containers. 6365 * Avoid polluting the global namespace and do not include <unistd.h>. 6366 6367 Fortran 6368 6369 * The compile flag [16]-fstack-arrays has been added, which causes 6370 all local arrays to be put on stack memory. For some programs this 6371 will improve the performance significantly. If your program uses 6372 very large local arrays, it is possible that you will have to 6373 extend your runtime limits for stack memory. 6374 * The [17]-Ofast flag now also implies [18]-fno-protect-parens and 6375 [19]-fstack-arrays. 6376 * Front-end optimizations can now be selected by the 6377 [20]-ffrontend-optimize option and deselected by the 6378 -fno-frontend-optimize option. 6379 * When front-end optimization removes a function call, 6380 [21]-Wfunction-elimination warns about that. 6381 * When performing front-end-optimization, the 6382 [22]-faggressive-function-elimination option allows the removal of 6383 duplicate function calls even for impure functions. 6384 * The flag [23]-Wreal-q-constant has been added, which warns if 6385 floating-point literals have been specified using q (such as 6386 1.0q0); the q marker is now supported as a vendor extension to 6387 denote quad precision (REAL(16) or, if not available, REAL(10)). 6388 Consider using a kind parameter (such as in 1.0_qp) instead, which 6389 can be obtained via [24]SELECTED_REAL_KIND. 6390 * The GFORTRAN_USE_STDERR environment variable has been removed. GNU 6391 Fortran now always prints error messages to standard error. If you 6392 wish to redirect standard error, please consult the manual for your 6393 OS, shell, batch environment etc. as appropriate. 6394 * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment 6395 variable have been removed. When encountering a serious error, 6396 gfortran will now always abort the program. Whether a core dump is 6397 generated depends on the user environment settings; see the ulimit 6398 -c setting for POSIX shells, limit coredumpsize for C shells, and 6399 the [25]WER user-mode dumps settings on Windows. 6400 * The [26]-fbacktrace option is now enabled by default. When 6401 encountering a fatal error, gfortran will attempt to print a 6402 backtrace to standard error before aborting. It can be disabled 6403 with -fno-backtrace. Note: On POSIX targets with the addr2line 6404 utility from GNU binutils, GNU Fortran can print a backtrace with 6405 function name, file name, line number information in addition to 6406 the addresses; otherwise only the addresses are printed. 6407 * [27]Fortran 2003: 6408 + Generic interface names which have the same name as derived 6409 types are now supported, which allows to write constructor 6410 functions. Note that Fortran does not support static 6411 constructor functions; only default initialization or an 6412 explicit structure-constructor initialization are available. 6413 + [28]Polymorphic (class) arrays are now supported. 6414 * [29]Fortran 2008: 6415 + Support for the DO CONCURRENT construct has been added, which 6416 allows the user to specify that individual loop iterations 6417 have no interdependencies. 6418 + [30]Coarrays: Full single-image support except for polymorphic 6419 coarrays. Additionally, preliminary support for multiple 6420 images via an MPI-based [31]coarray communication library has 6421 been added. Note: The library version is not yet usable as 6422 remote coarray access is not yet possible. 6423 * [32]TS 29113: 6424 + New flag [33]-std=f2008ts permits programs that are expected 6425 to conform to the Fortran 2008 standard and the draft 6426 Technical Specification (TS) 29113 on Further Interoperability 6427 of Fortran with C. 6428 + The OPTIONAL attribute is now allowed for dummy arguments of 6429 BIND(C) procedures. 6430 + The RANK intrinsic has been added. 6431 + The implementation of the ASYNCHRONOUS attribute in GCC is 6432 compatible with the candidate draft of TS 29113 (since GCC 6433 4.6). 6434 6435 Go 6436 6437 * GCC 4.7 implements the [34]Go 1 language standard. The library 6438 support in 4.7.0 is not quite complete, due to release timing. 6439 Release 4.7.1 includes complete support for Go 1. The Go library is 6440 from the Go 1.0.1 release. 6441 * Go has been tested on GNU/Linux and Solaris platforms. It may work 6442 on other platforms as well. 6443 6444 New Targets and Target Specific Improvements 6445 6446 ARM 6447 6448 * GCC now supports the Cortex-A7 processor implementing the v7-a 6449 version of the architecture using the option -mcpu=cortex-a7. 6450 * The default vector size in auto-vectorization for NEON is now 128 6451 bits. If vectorization fails thusly, the vectorizer tries again 6452 with 64-bit vectors. 6453 * A new option -mvectorize-with-neon-double was added to allow users 6454 to change the vector size to 64 bits. 6455 6456 AVR 6457 6458 * GCC now supports the XMEGA architecture. This requires GNU binutils 6459 2.22 or later. 6460 * Support for the [35]named address spaces __flash, __flash1, , 6461 __flash5 and __memx has been added. These address spaces locate 6462 read-only data in flash memory and allow reading from flash memory 6463 by means of ordinary C code, i.e. without the need of (inline) 6464 assembler code: 6465 6466 const __flash int values[] = { 42, 31 }; 6467 6468 int add_values (const __flash int *p, int i) 6469 { 6470 return values[i] + *p; 6471 } 6472 6473 * Support has been added for the AVR-specific configure option 6474 --with-avrlibc=yes in order to arrange for better integration of 6475 [36]AVR-Libc. This configure option is supported in avr-gcc 4.7.2 6476 and newer and will only take effect in non-RTEMS configurations. If 6477 avr-gcc is configured for RTEMS, the option will be ignored which 6478 is the same as specifying --with-avrlibc=no. See [37]PR54461 for 6479 more technical details. 6480 * Support for AVR-specific [38]built-in functions has been added. 6481 * Support has been added for the signed and unsigned 24-bit scalar 6482 integer types __int24 and __uint24. 6483 * New command-line options -maccumulate-args, -mbranch-cost=cost and 6484 -mstrict-X were added to allow better fine-tuning of code 6485 optimization. 6486 * The command-line option -fdata-sections now also takes affect on 6487 the section names of variables with the progmem attribute. 6488 * A new inline assembler print modifier %i to print a RAM address as 6489 I/O address has been added: 6490 6491 #include <avr/io.h> /* Port Definitions from AVR-LibC */ 6492 6493 void set_portb (uint8_t value) 6494 { 6495 asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory"); 6496 } 6497 6498 The offset between an I/O address and the RAM address for that I/O 6499 location is device-specific. This offset is taken into account when 6500 printing a RAM address with the %i modifier so that the address is 6501 suitable to be used as operand in an I/O command. The address must 6502 be a constant integer known at compile time. 6503 * The inline assembler constraint "R" to represent integers in the 6504 range 6 5 has been removed without replacement. 6505 * Many optimizations to: 6506 + 64-bit integer arithmetic 6507 + Widening multiplication 6508 + Integer division by a constant 6509 + Avoid constant reloading in multi-byte instructions. 6510 + Micro-optimizations for special instruction sequences. 6511 + Generic built-in functions like __builtin_ffs*, 6512 __builtin_clz*, etc. 6513 + If-else decision trees generated by switch instructions 6514 + Merging of data located in flash memory 6515 + New libgcc variants for devices with 8-bit wide stack pointer 6516 + 6517 * Better documentation: 6518 + Handling of EIND and indirect jumps on devices with more than 6519 128 KiB of program memory. 6520 + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function 6521 registers. 6522 + Function attributes OS_main and OS_task. 6523 + AVR-specific built-in macros. 6524 6525 C6X 6526 6527 * Support has been added for the Texas Instruments C6X family of 6528 processors. 6529 6530 CR16 6531 6532 * Support has been added for National Semiconductor's CR16 6533 architecture. 6534 6535 Epiphany 6536 6537 * Support has been added for Adapteva's Epiphany architecture. 6538 6539 IA-32/x86-64 6540 6541 * Support for Intel AVX2 intrinsics, built-in functions and code 6542 generation is available via -mavx2. 6543 * Support for Intel BMI2 intrinsics, built-in functions and code 6544 generation is available via -mbmi2. 6545 * Implementation and automatic generation of __builtin_clz* using the 6546 lzcnt instruction is available via -mlzcnt. 6547 * Support for Intel FMA3 intrinsics and code generation is available 6548 via -mfma. 6549 * A new -mfsgsbase command-line option is available that makes GCC 6550 generate new segment register read/write instructions through 6551 dedicated built-ins. 6552 * Support for the new Intel rdrnd instruction is available via 6553 -mrdrnd. 6554 * Two additional AVX vector conversion instructions are available via 6555 -mf16c. 6556 * Support for new Intel processor codename IvyBridge with RDRND, 6557 FSGSBASE and F16C is available through -march=core-avx-i. 6558 * Support for the new Intel processor codename Haswell with AVX2, 6559 FMA, BMI, BMI2, LZCNT is available through -march=core-avx2. 6560 * Support for new AMD family 15h processors (Piledriver core) is now 6561 available through -march=bdver2 and -mtune=bdver2 options. 6562 * Support for [39]the x32 psABI is now available through the -mx32 6563 option. 6564 * Windows mingw targets are using the -mms-bitfields option by 6565 default. 6566 * Windows x86 targets are using the __thiscall calling convention for 6567 C++ class-member functions. 6568 * Support for the configure option --with-threads=posix for Windows 6569 mingw targets. 6570 6571 MIPS 6572 6573 * GCC now supports thread-local storage (TLS) for MIPS16. This 6574 requires GNU binutils 2.22 or later. 6575 * GCC can now generate code specifically for the Cavium Octeon+ and 6576 Octeon2 processors. The associated command-line options are 6577 -march=octeon+ and -march=octeon2 respectively. Both options 6578 require GNU binutils 2.22 or later. 6579 * GCC can now work around certain 24k errata, under the control of 6580 the command-line option -mfix-24k. These workarounds require GNU 6581 binutils 2.20 or later. 6582 * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build 6583 n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux 6584 toolchain that generates 32-bit code by default. Use the 6585 configure-time option --enable-targets=all to select these extra 6586 multilibs. 6587 * Passing -fno-delayed-branch now also stops the assembler from 6588 automatically filling delay slots. 6589 6590 PowerPC/PowerPC64 6591 6592 * Vectors of type vector long long or vector long are passed and 6593 returned using the same method as other vectors with the VSX 6594 instruction set. Previously GCC did not adhere to the ABI for 6595 128-bit vectors with 64-bit integer base types (PR 48857). This 6596 will also be fixed in the GCC 4.6.1 and 4.5.4 releases. 6597 * A new option -mno-pointers-to-nested-functions was added to allow 6598 AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify 6599 that the compiler should not load up the chain register (r11) 6600 before calling a function through a pointer. If you use this 6601 option, you cannot call nested functions through a pointer, or call 6602 other languages that might use the static chain. 6603 * A new option msave-toc-indirect was added to allow AIX 6604 32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we 6605 save the TOC in the prologue for indirect calls or generate the 6606 save inline. This can speed up some programs that call through a 6607 function pointer a lot, but it can slow down other functions that 6608 only call through a function pointer in exceptional cases. 6609 * The PowerPC port will now enable machine-specific built-in 6610 functions when the user switches the target machine using the 6611 #pragma GCC target or __attribute__ ((__target__ ("target"))) code 6612 sequences. In addition, the target macros are updated. However, due 6613 to the way the -save-temps switch is implemented, you won't see the 6614 effect of these additional macros being defined in preprocessor 6615 output. 6616 6617 SH 6618 6619 * A new option -msoft-atomic has been added. When it is specified, 6620 GCC will generate GNU/Linux-compatible gUSA atomic sequences for 6621 the new __atomic routines. 6622 * Since it is neither supported by GAS nor officially documented, 6623 code generation for little endian SH2A has been disabled. 6624 Specifying -ml with -m2a* will now result in a compiler error. 6625 * The defunct -mbranch-cost option has been fixed. 6626 * Some improvements to the generated code of: 6627 + Utilization of the tst #imm,R0 instruction. 6628 + Dynamic shift instructions on SH2A. 6629 + Integer absolute value calculations. 6630 * The -mdiv= option for targets other than SHmedia has been fixed and 6631 documented. 6632 6633 SPARC 6634 6635 * The option -mflat has been reinstated. When it is specified, the 6636 compiler will generate code for a single register window model. 6637 This is essentially a new implementation and the corresponding 6638 debugger support has been added to GDB 7.4. 6639 * Support for the options -mtune=native and -mcpu=native has been 6640 added on selected native platforms (GNU/Linux and Solaris). 6641 * Support for the SPARC T3 (Niagara 3) processor has been added. 6642 * VIS: 6643 + An intrinsics header visintrin.h has been added. 6644 + Builtin intrinsics for the VIS 1.0 edge handling and pixel 6645 compare instructions have been added. 6646 + The little-endian version of alignaddr is now supported. 6647 + When possible, VIS builtins are marked const, which should 6648 increase the compiler's ability to optimize VIS operations. 6649 + The compiler now properly tracks the %gsr register and how it 6650 behaves as an input for various VIS instructions. 6651 + Akin to fzero, the compiler can now generate fone instructions 6652 in order to set all of the bits of a floating-point register 6653 to 1. 6654 + The documentation for the VIS intrinsics in the GCC manual has 6655 been brought up to date and many inaccuracies were fixed. 6656 + Intrinsics for the VIS 2.0 bmask, bshuffle, and 6657 non-condition-code setting edge instructions have been added. 6658 Their availability is controlled by the new -mvis2 and 6659 -mno-vis2 options. They are enabled by default on 6660 UltraSPARC-III and later CPUs. 6661 * Support for UltraSPARC Fused Multiply-Add floating-point extensions 6662 has been added. These instructions are enabled by default on SPARC 6663 T3 (Niagara 3) and later CPUs. 6664 6665 TILE-Gx/TILEPro 6666 6667 * Support has been added for the Tilera TILE-Gx and TILEPro families 6668 of processors. 6669 6670 Other significant improvements 6671 6672 * A new option (-grecord-gcc-switches) was added that appends 6673 compiler command-line options that might affect code generation to 6674 the DW_AT_producer attribute string in the DWARF debugging 6675 information. 6676 * GCC now supports various new GNU extensions to the DWARF debugging 6677 information format, like [40]entry value and [41]call site 6678 information, [42]typed DWARF stack or [43]a more compact macro 6679 representation. Support for these extensions has been added to GDB 6680 7.4. They can be disabled through the -gstrict-dwarf command-line 6681 option. 6682 6683 GCC 4.7.1 6684 6685 This is the [44]list of problem reports (PRs) from GCC's bug tracking 6686 system that are known to be fixed in the 4.7.1 release. This list might 6687 not be complete (that is, it is possible that some PRs that have been 6688 fixed are not listed here). 6689 6690 The Go front end in the 4.7.1 release fully supports the [45]Go 1 6691 language standard. 6692 6693 GCC 4.7.2 6694 6695 This is the [46]list of problem reports (PRs) from GCC's bug tracking 6696 system that are known to be fixed in the 4.7.2 release. This list might 6697 not be complete (that is, it is possible that some PRs that have been 6698 fixed are not listed here). 6699 6700 GCC 4.7.3 6701 6702 This is the [47]list of problem reports (PRs) from GCC's bug tracking 6703 system that are known to be fixed in the 4.7.3 release. This list might 6704 not be complete (that is, it is possible that some PRs that have been 6705 fixed are not listed here). 6706 6707 GCC 4.7.4 6708 6709 This is the [48]list of problem reports (PRs) from GCC's bug tracking 6710 system that are known to be fixed in the 4.7.4 release. This list might 6711 not be complete (that is, it is possible that some PRs that have been 6712 fixed are not listed here). 6713 6714 6715 For questions related to the use of GCC, please consult these web 6716 pages and the [49]GCC manuals. If that fails, the 6717 [50]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6718 web pages and the development of GCC are welcome on our developer 6719 list at [51]gcc (a] gcc.gnu.org. All of [52]our lists have public 6720 archives. 6721 6722 Copyright (C) [53]Free Software Foundation, Inc. Verbatim copying and 6723 distribution of this entire article is permitted in any medium, 6724 provided this notice is preserved. 6725 6726 These pages are [54]maintained by the GCC team. Last modified 6727 2018-12-11[55]. 6728 6729 References 6730 6731 1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html 6732 2. http://savannah.nongnu.org/bugs/?35407 6733 3. https://gcc.gnu.org/PR18145 6734 4. https://gcc.gnu.org/gcc-4.7/porting_to.html 6735 5. https://gcc.gnu.org/wiki/TransactionalMemory 6736 6. https://gcc.gnu.org/wiki/Atomic/GCCMM 6737 7. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 6738 8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 6739 9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 6740 10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 6741 11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 6742 12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 6743 13. https://gcc.gnu.org/PR14258 6744 14. https://gcc.gnu.org/PR35688 6745 15. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011 6746 16. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 6747 17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689 6748 18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270 6749 19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 6750 20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275 6751 21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170 6752 22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270 6753 23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWreal-q-constant_007d-149 6754 24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html 6755 25. https://docs.microsoft.com/en-us/windows/desktop/wer/collecting-user-mode-dumps 6756 26. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183 6757 27. https://gcc.gnu.org/wiki/Fortran2003Status 6758 28. https://gcc.gnu.org/wiki/OOP 6759 29. https://gcc.gnu.org/wiki/Fortran2008Status 6760 30. https://gcc.gnu.org/wiki/Coarray 6761 31. https://gcc.gnu.org/wiki/CoarrayLib 6762 32. https://gcc.gnu.org/wiki/TS29113Status 6763 33. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bstd_003d_007d_0040var_007bstd_007d-option-53 6764 34. https://golang.org/doc/go1 6765 35. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html 6766 36. http://nongnu.org/avr-libc/ 6767 37. https://gcc.gnu.org/PR54461 6768 38. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built_002din-Functions.html 6769 39. https://sites.google.com/site/x32abi/ 6770 40. http://www.dwarfstd.org/ShowIssue.php?issue=100909.1 6771 41. http://www.dwarfstd.org/ShowIssue.php?issue=100909.2 6772 42. http://www.dwarfstd.org/ShowIssue.php?issue=140425.1 6773 43. http://www.dwarfstd.org/ShowIssue.php?issue=110722.1 6774 44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1 6775 45. https://golang.org/doc/go1 6776 46. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2 6777 47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3 6778 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4 6779 49. https://gcc.gnu.org/onlinedocs/ 6780 50. mailto:gcc-help (a] gcc.gnu.org 6781 51. mailto:gcc (a] gcc.gnu.org 6782 52. https://gcc.gnu.org/lists.html 6783 53. https://www.fsf.org/ 6784 54. https://gcc.gnu.org/about.html 6785 55. http://validator.w3.org/check/referer 6786 ====================================================================== 6787 http://gcc.gnu.org/gcc-4.6/index.html 6788 GCC 4.6 Release Series 6789 6790 April 12, 2013 6791 6792 The [1]GNU project and the GCC developers are pleased to announce the 6793 release of GCC 4.6.4. 6794 6795 This release is a bug-fix release, containing fixes for regressions in 6796 GCC 4.6.3 relative to previous releases of GCC. 6797 6798 This release series is no longer maintained. 6799 6800 Release History 6801 6802 GCC 4.6.4 6803 April 12, 2013 ([2]changes, [3]documentation) 6804 6805 GCC 4.6.3 6806 March 1, 2012 ([4]changes, [5]documentation) 6807 6808 GCC 4.6.2 6809 October 26, 2011 ([6]changes, [7]documentation) 6810 6811 GCC 4.6.1 6812 June 27, 2011 ([8]changes, [9]documentation) 6813 6814 GCC 4.6.0 6815 March 25, 2011 ([10]changes, [11]documentation) 6816 6817 References and Acknowledgements 6818 6819 GCC used to stand for the GNU C Compiler, but since the compiler 6820 supports several other languages aside from C, it now stands for the 6821 GNU Compiler Collection. 6822 6823 A list of [12]successful builds is updated as new information becomes 6824 available. 6825 6826 The GCC developers would like to thank the numerous people that have 6827 contributed new features, improvements, bug fixes, and other changes as 6828 well as test results to GCC. This [13]amazing group of volunteers is 6829 what makes GCC successful. 6830 6831 For additional information about GCC please refer to the [14]GCC 6832 project web site or contact the [15]GCC development mailing list. 6833 6834 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 6835 6836 6837 For questions related to the use of GCC, please consult these web 6838 pages and the [18]GCC manuals. If that fails, the 6839 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6840 web pages and the development of GCC are welcome on our developer 6841 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 6842 archives. 6843 6844 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 6845 distribution of this entire article is permitted in any medium, 6846 provided this notice is preserved. 6847 6848 These pages are [23]maintained by the GCC team. Last modified 6849 2018-09-30[24]. 6850 6851 References 6852 6853 1. http://www.gnu.org/ 6854 2. http://gcc.gnu.org/gcc-4.6/changes.html 6855 3. https://gcc.gnu.org/onlinedocs/4.6.4/ 6856 4. http://gcc.gnu.org/gcc-4.6/changes.html 6857 5. https://gcc.gnu.org/onlinedocs/4.6.3/ 6858 6. http://gcc.gnu.org/gcc-4.6/changes.html 6859 7. https://gcc.gnu.org/onlinedocs/4.6.2/ 6860 8. http://gcc.gnu.org/gcc-4.6/changes.html 6861 9. https://gcc.gnu.org/onlinedocs/4.6.1/ 6862 10. http://gcc.gnu.org/gcc-4.6/changes.html 6863 11. https://gcc.gnu.org/onlinedocs/4.6.0/ 6864 12. http://gcc.gnu.org/gcc-4.6/buildstat.html 6865 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 6866 14. http://gcc.gnu.org/index.html 6867 15. mailto:gcc (a] gcc.gnu.org 6868 16. http://gcc.gnu.org/mirrors.html 6869 17. http://gcc.gnu.org/svn.html 6870 18. https://gcc.gnu.org/onlinedocs/ 6871 19. mailto:gcc-help (a] gcc.gnu.org 6872 20. mailto:gcc (a] gcc.gnu.org 6873 21. https://gcc.gnu.org/lists.html 6874 22. https://www.fsf.org/ 6875 23. https://gcc.gnu.org/about.html 6876 24. http://validator.w3.org/check/referer 6877 ====================================================================== 6878 http://gcc.gnu.org/gcc-4.6/changes.html 6879 GCC 4.6 Release Series 6880 Changes, New Features, and Fixes 6881 6882 Caveats 6883 6884 * The options -b <machine> and -V <version> have been removed because 6885 they were unreliable. Instead, users should directly run 6886 <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to 6887 run a different version of gcc. 6888 * GCC now has stricter checks for invalid command-line options. In 6889 particular, when gcc was called to link object files rather than 6890 compile source code, it would previously accept and ignore all 6891 options starting with --, including linker options such as 6892 --as-needed and --export-dynamic, although such options would 6893 result in errors if any source code was compiled. Such options, if 6894 unknown to the compiler, are now rejected in all cases; if the 6895 intent was to pass them to the linker, options such as 6896 -Wl,--as-needed should be used. 6897 * Versions of the GNU C library up to and including 2.11.1 included 6898 an [1]incorrect implementation of the cproj function. GCC optimizes 6899 its builtin cproj according to the behavior specified and allowed 6900 by the ISO C99 standard. If you want to avoid discrepancies between 6901 the C library and GCC's builtin transformations when using cproj in 6902 your code, use GLIBC 2.12 or later. If you are using an older GLIBC 6903 and actually rely on the incorrect behavior of cproj, then you can 6904 disable GCC's transformations using -fno-builtin-cproj. 6905 * The C-only intermodule optimization framework (IMA, enabled by 6906 -combine) has been removed in favor of the new generic link-time 6907 optimization framework (LTO) introduced in [2]GCC 4.5.0. 6908 * GCC now ships with the LGPL-licensed libquadmath library, which 6909 provides quad-precision mathematical functions for targets with a 6910 __float128 datatype. __float128 is available for targets on 32-bit 6911 x86, x86-64 and Itanium architectures. The libquadmath library is 6912 automatically built on such targets when building the Fortran 6913 compiler. 6914 * New -Wunused-but-set-variable and -Wunused-but-set-parameter 6915 warnings were added for C, C++, Objective-C and Objective-C++. 6916 These warnings diagnose variables respective parameters which are 6917 only set in the code and never otherwise used. Usually such 6918 variables are useless and often even the value assigned to them is 6919 computed needlessly, sometimes expensively. The 6920 -Wunused-but-set-variable warning is enabled by default by -Wall 6921 flag and -Wunused-but-set-parameter by -Wall -Wextra flags. 6922 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 6923 rules for the layout of vectors that could lead to wrong code being 6924 generated. Vectors larger than 8 bytes in size are now by default 6925 aligned to an 8-byte boundary. This is an ABI change: code that 6926 makes explicit use of vector types may be incompatible with binary 6927 objects built with older versions of GCC. Auto-vectorized code is 6928 not affected by this change. (This change affects GCC versions 6929 4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.) 6930 * On AVR, variables with the progmem attribute to locate data in 6931 flash memory must be qualified as const. 6932 * Support for a number of older systems and recently unmaintained or 6933 untested target ports of GCC has been declared obsolete in GCC 4.6. 6934 Unless there is activity to revive them, the next release of GCC 6935 will have their sources permanently removed. 6936 All GCC ports for the following processor architectures have been 6937 declared obsolete: 6938 + Argonaut ARC (arc-*) 6939 + National Semiconductor CRX (crx-*) 6940 + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*, 6941 m68hc12-*-*, m6812-*-*) 6942 + Sunplus S+core (score-*) 6943 The following ports for individual systems on particular 6944 architectures have been obsoleted: 6945 + Interix (i[34567]86-*-interix3*) 6946 + NetWare x86 (i[3456x]86-*-netware*) 6947 + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*) 6948 + MCore PE (mcore-*-pe*) 6949 + SH SymbianOS (sh*-*-symbianelf*) 6950 + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*) 6951 + M68K uClinux old ABI (m68k-*-uclinuxoldabi*) 6952 + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*, 6953 vax-*-netbsd*, but not *-*-netbsdelf*) 6954 The i[34567]86-*-pe alias for Cygwin targets has also been 6955 obsoleted; users should configure for i[34567]86-*-cygwin* instead. 6956 Certain configure options to control the set of libraries built 6957 with GCC on some targets have been obsoleted. On ARM targets, the 6958 options --disable-fpu, --disable-26bit, --disable-underscore, 6959 --disable-interwork, --disable-biendian and --disable-nofmult have 6960 been obsoleted. On MIPS targets, the options 6961 --disable-single-float, --disable-biendian and --disable-softfloat 6962 have been obsoleted. 6963 * Support has been removed for all the [3]configurations obsoleted in 6964 GCC 4.5. 6965 * More information on porting to GCC 4.6 from previous versions of 6966 GCC can be found in the [4]porting guide for this release. 6967 6968 General Optimizer Improvements 6969 6970 * A new general optimization level, -Ofast, has been introduced. It 6971 combines the existing optimization level -O3 with options that can 6972 affect standards compliance but result in better optimized code. 6973 For example, -Ofast enables -ffast-math. 6974 * Link-time optimization improvements: 6975 + The [5]Scalable Whole Program Optimizer (WHOPR) project has 6976 stabilized to the point of being usable. It has become the 6977 default mode when using the LTO optimization model. Link time 6978 optimization can now split itself into multiple parallel 6979 compilations. Parallelism is controlled with -flto=n (where n 6980 specifies the number of compilations to execute in parallel). 6981 GCC can also cooperate with a GNU make job server by 6982 specifying the -flto=jobserver option and adding + to the 6983 beginning of the Makefile rule executing the linker. 6984 Classical LTO mode can be enforced by -flto-partition=none. 6985 This may result in small code quality improvements. 6986 + A large number of bugs were fixed. GCC itself, Mozilla Firefox 6987 and other large applications can be built with LTO enabled. 6988 + The linker plugin support improvements 6989 o Linker plugin is now enabled by default when the linker 6990 is detected to have plugin support. This is the case for 6991 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and 6992 the Gold linker on ELF targets. Plugin support of the 6993 Apple linker on Darwin is not compatible with GCC. The 6994 linker plugin can also be controlled by the 6995 -fuse-linker-plugin command-line option. 6996 o Resolution information from the linker plugin is used to 6997 drive whole program assumptions. Use of the linker plugin 6998 results in more aggressive optimization on binaries and 6999 on shared libraries that use the hidden visibility 7000 attribute. Consequently the use of -fwhole-program is not 7001 necessary in addition to LTO. 7002 + Hidden symbols used from non-LTO objects now have to be 7003 explicitly annotated with externally_visible when the linker 7004 plugin is not used. 7005 + C++ inline functions and virtual tables are now privatized 7006 more aggressively, leading to better inter-procedural 7007 optimization and faster dynamic linking. 7008 + Memory usage and intermediate language streaming performance 7009 have been improved. 7010 + Static constructors and destructors from individual units are 7011 inlined into a single function. This can significantly improve 7012 startup times of large C++ applications where static 7013 constructors are very common. For example, static constructors 7014 are used when including the iostream header. 7015 + Support for the Ada language has been added. 7016 * Interprocedural optimization improvements 7017 + The interprocedural framework was re-tuned for link time 7018 optimization. Several scalability issues were resolved. 7019 + Improved auto-detection of const and pure functions. Newly, 7020 noreturn functions are auto-detected. 7021 The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is 7022 available that informs users when adding attributes to headers 7023 might improve code generation. 7024 + A number of inlining heuristic improvements. In particular: 7025 o Partial inlining is now supported and enabled by default 7026 at -O2 and greater. The feature can be controlled via 7027 -fpartial-inlining. 7028 Partial inlining splits functions with short hot path to 7029 return. This allows more aggressive inlining of the hot 7030 path leading to better performance and often to code size 7031 reductions (because cold parts of functions are not 7032 duplicated). 7033 o Scalability for large compilation units was improved 7034 significantly. 7035 o Inlining of callbacks is now more aggressive. 7036 o Virtual methods are considered for inlining when the 7037 caller is inlined and devirtualization is then possible. 7038 o Inlining when optimizing for size (either in cold regions 7039 of a program or when compiling with -Os) was improved to 7040 better handle C++ programs with larger abstraction 7041 penalty, leading to smaller and faster code. 7042 + The IPA reference optimization pass detecting global variables 7043 used or modified by functions was strengthened and sped up. 7044 + Functions whose address was taken are now optimized out when 7045 all references to them are dead. 7046 + A new inter-procedural static profile estimation pass detects 7047 functions that are executed once or unlikely to be executed. 7048 Unlikely executed functions are optimized for size. Functions 7049 executed once are optimized for size except for the inner 7050 loops. 7051 + On most targets with named section support, functions used 7052 only at startup (static constructors and main), functions used 7053 only at exit and functions detected to be cold are placed into 7054 separate text segment subsections. This extends the 7055 -freorder-functions feature and is controlled by the same 7056 switch. The goal is to improve the startup time of large C++ 7057 programs. 7058 Proper function placement requires linker support. GNU ld 7059 2.21.51 on ELF targets was updated to place those functions 7060 together within the text section leading to better code 7061 locality and faster startup times of large C++ programs. The 7062 feature is also supported in the Apple linker. Support in the 7063 gold linker is planned. 7064 * A new switch -fstack-usage has been added. It makes the compiler 7065 output stack usage information for the program, on a per-function 7066 basis, in an auxiliary file. 7067 * A new switch -fcombine-stack-adjustments has been added. It can be 7068 used to enable or disable the compiler's stack-slot combining pass 7069 which before was enabled automatically at -O1 and above, but could 7070 not be controlled on its own. 7071 * A new switch -fstrict-volatile-bitfields has been added. Using it 7072 indicates that accesses to volatile bitfields should use a single 7073 access of the width of the field's type. This option can be useful 7074 for precisely defining and accessing memory-mapped peripheral 7075 registers from C or C++. 7076 7077 Compile time and memory usage improvements 7078 7079 * Datastructures used by the dataflow framework in GCC were 7080 reorganized for better memory usage and more cache locality. 7081 Compile time is improved especially on units with large functions 7082 (possibly resulting from a lot of inlining) not fitting into the 7083 processor cache. The compile time of the GCC C compiler binary with 7084 link-time optimization went down by over 10% (benchmarked on x86-64 7085 target). 7086 7087 New Languages and Language specific improvements 7088 7089 Ada 7090 7091 * Stack checking has been improved on selected architectures (Alpha, 7092 IA-32/x86-64, RS/6000 and SPARC): it now will detect stack 7093 overflows in all cases on these architectures. 7094 * Initial support for Ada 2012 has been added. 7095 7096 C family 7097 7098 * A new warning, enabled by -Wdouble-promotion, has been added that 7099 warns about cases where a value of type float is implicitly 7100 promoted to double. This is especially helpful for CPUs that handle 7101 the former in hardware, but emulate the latter in software. 7102 * A new function attribute leaf was introduced. This attribute allows 7103 better inter-procedural optimization across calls to functions that 7104 return to the current unit only via returning or exception 7105 handling. This is the case for most library functions that have no 7106 callbacks. 7107 * Support for a new data type __int128 for targets having wide enough 7108 machine-mode support. 7109 * The new function attribute callee_pop_aggregate allows to specify 7110 if the caller or callee is responsible for popping the aggregate 7111 return pointer value from the stack. 7112 * Support for selectively enabling and disabling warnings via #pragma 7113 GCC diagnostic has been added. For instance: 7114 #pragma GCC diagnostic error "-Wuninitialized" 7115 foo(a); /* error is given for this one */ 7116 #pragma GCC diagnostic push 7117 #pragma GCC diagnostic ignored "-Wuninitialized" 7118 foo(b); /* no diagnostic for this one */ 7119 #pragma GCC diagnostic pop 7120 foo(c); /* error is given for this one */ 7121 #pragma GCC diagnostic pop 7122 foo(d); /* depends on command-line options */ 7123 7124 * The -fmax-errors=N option is now supported. Using this option 7125 causes the compiler to exit after N errors have been issued. 7126 7127 C 7128 7129 * There is now experimental support for some features from the 7130 upcoming C1X revision of the ISO C standard. This support may be 7131 selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions. 7132 Note that this support is experimental and may change incompatibly 7133 in future releases for consistency with changes to the C1X standard 7134 draft. The following features are newly supported as described in 7135 the N1539 draft of C1X (with changes agreed at the March 2011 WG14 7136 meeting); some other features were already supported with no 7137 compiler changes being needed, or have some support but not in full 7138 accord with N1539 (as amended). 7139 + Static assertions (_Static_assert keyword) 7140 + Typedef redefinition 7141 + New macros in <float.h> 7142 + Anonymous structures and unions 7143 * The new -fplan9-extensions option directs the compiler to support 7144 some extensions for anonymous struct fields which are implemented 7145 by the Plan 9 compiler. A pointer to a struct may be automatically 7146 converted to a pointer to an anonymous field when calling a 7147 function, in order to make the types match. An anonymous struct 7148 field whose type is a typedef name may be referred to using the 7149 typedef name. 7150 7151 C++ 7152 7153 * Improved [7]experimental support for the upcoming C++0x ISO C++ 7154 standard, including support for constexpr (thanks to Gabriel Dos 7155 Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide), 7156 noexcept, unrestricted unions, range-based for loops (thanks to 7157 Rodrigo Rivas Costa), opaque enum declarations (thanks also to 7158 Rodrigo), implicitly deleted functions and implicit move 7159 constructors. 7160 * When an extern declaration within a function does not match a 7161 declaration in the enclosing context, G++ now properly declares the 7162 name within the namespace of the function rather than the namespace 7163 which was open just before the function definition ([8]c++/43145). 7164 * GCC now warns by default when casting integers to larger pointer 7165 types. These warnings can be disabled with the option 7166 -Wno-int-to-pointer-cast, which is now also available in C++. 7167 * G++ no longer optimizes using the assumption that a value of 7168 enumeration type will fall within the range specified by the 7169 standard, since that assumption is easily violated with a 7170 conversion from integer type ([9]c++/43680). The old behavior can 7171 be restored with -fstrict-enums. 7172 * The new -fnothrow-opt flag changes the semantics of a throw() 7173 exception specification to match the proposed semantics of the 7174 noexcept specification: just call terminate if an exception tries 7175 to propagate out of a function with such an exception 7176 specification. This dramatically reduces or eliminates the code 7177 size overhead from adding the exception specification. 7178 * The new -Wnoexcept flag will suggest adding a noexcept qualifier to 7179 a function that the compiler can tell doesn't throw if it would 7180 change the value of a noexcept expression. 7181 * The -Wshadow option now warns if a local variable or type 7182 declaration shadows another type in C++. Note that the compiler 7183 will not warn if a local variable shadows a struct/class/enum, but 7184 will warn if it shadows an explicit typedef. 7185 * When an identifier is not found in the current scope, G++ now 7186 offers suggestions about which identifier might have been intended. 7187 * G++ now issues clearer diagnostics for missing semicolons after 7188 class, struct, and union definitions. 7189 * G++ now issues clearer diagnostics for missing semicolons after 7190 class member declarations. 7191 * G++ now issues clearer diagnostics when a colon is used in a place 7192 where a double-colon was intended. 7193 * G++ no longer accepts mutable on reference members ([10]c++/33558). 7194 Use -fpermissive to allow the old, non-conforming behaviour. 7195 * A few mangling fixes have been made, to attribute const/volatile on 7196 function pointer types, decltype of a plain decl, and use of a 7197 function parameter in the declaration of another parameter. By 7198 default the compiler still uses the old mangling, but emits aliases 7199 with the new mangling on targets that support strong aliases. Users 7200 can switch over entirely to the new mangling with -fabi-version=5 7201 or -fabi-version=0. -Wabi will now warn about code that uses the 7202 old mangling. 7203 * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified 7204 type to be default initialized unless the type has a user-declared 7205 default constructor. In 4.6.2 G++ implements the proposed 7206 resolution of [11]DR 253, so default initialization is allowed if 7207 it initializes all subobjects. Code that fails to compile can be 7208 fixed by providing an initializer e.g. 7209 struct A { A(); }; 7210 struct B : A { int i; }; 7211 const B b = B(); 7212 Use -fpermissive to allow the old, non-conforming behaviour. 7213 7214 Runtime Library (libstdc++) 7215 7216 * [12]Improved experimental support for the upcoming ISO C++ 7217 standard, C++0x, including using constexpr and nullptr. 7218 * Performance improvements to the [13]Debug Mode, thanks to Franois 7219 Dumont. 7220 * Atomic operations used for reference-counting are annotated so that 7221 they can be understood by race detectors such as Helgrind, see 7222 [14]Data Race Hunting. 7223 * Most libstdc++ standard headers have been changed to no longer 7224 include the cstddef header as an implementation detail. Code that 7225 relied on that header being included as side-effect of including 7226 other standard headers will need to include cstddef explicitly. 7227 7228 Fortran 7229 7230 * On systems supporting the libquadmath library, GNU Fortran now also 7231 supports a quad-precision, kind=16 floating-point data type 7232 (REAL(16), COMPLEX(16)). As the data type is not fully supported in 7233 hardware, calculations might be one to two orders of magnitude 7234 slower than with the 4, 8 or 10 bytes floating-point data types. 7235 This change does not affect systems which support REAL(16) in 7236 hardware nor those which do not support libquadmath. 7237 * Much improved compile time for large array constructors. 7238 * In order to reduce execution time and memory consumption, use of 7239 temporary arrays in assignment expressions is avoided for many 7240 cases. The compiler now reverses loops in order to avoid generating 7241 a temporary array where possible. 7242 * Improved diagnostics, especially with -fwhole-file. 7243 * The -fwhole-file flag is now enabled by default. This improves code 7244 generation and diagnostics. It can be disabled using the deprecated 7245 -fno-whole-file flag. 7246 * Support the generation of Makefile dependencies via the [15]-M... 7247 flags of GCC; you may need to specify the -cpp option in addition. 7248 The dependencies take modules, Fortran's include, and CPP's 7249 #include into account. Note: Using -M for the module path is no 7250 longer supported, use -J instead. 7251 * The flag -Wconversion has been modified to only issue warnings 7252 where a conversion leads to information loss. This drastically 7253 reduces the number of warnings; -Wconversion is thus now enabled 7254 with -Wall. The flag -Wconversion-extra has been added and also 7255 warns about other conversions; -Wconversion-extra typically issues 7256 a huge number of warnings, most of which can be ignored. 7257 * A new command-line option -Wunused-dummy-argument warns about 7258 unused dummy arguments and is included in -Wall. Before, 7259 -Wunused-variable also warned about unused dummy arguments. 7260 * Fortran 2003 support has been extended: 7261 + Improved support for polymorphism between libraries and 7262 programs and for complicated inheritance patterns (cf. 7263 [16]object-oriented programming). 7264 + Experimental support of the ASSOCIATE construct. 7265 + In pointer assignments it is now possible to specify the lower 7266 bounds of the pointer and, for a rank-1 or a simply contiguous 7267 data-target, to remap the bounds. 7268 + Automatic (re)allocation: In intrinsic assignments to 7269 allocatable variables the left-hand side will be automatically 7270 allocated (if unallocated) or reallocated (if the shape or 7271 type parameter is different). To avoid the small performance 7272 penalty, you can use a(:) = ... instead of a = ... for arrays 7273 and character strings or disable the feature using -std=f95 7274 or -fno-realloc-lhs. 7275 + Deferred type parameter: For scalar allocatable and pointer 7276 variables the character length can be deferred. 7277 + Namelist variables with allocatable and pointer attribute and 7278 nonconstant length type parameter are supported. 7279 * Fortran 2008 support has been extended: 7280 + Experimental [17]coarray support (for one image only, i.e. 7281 num_images() == 1); use the [18]-fcoarray=single flag to 7282 enable it. 7283 + The STOP and the new ERROR STOP statements now support all 7284 constant expressions. 7285 + Support for the CONTIGUOUS attribute. 7286 + Support for ALLOCATE with MOLD. 7287 + Support for the STORAGE_SIZE intrinsic inquiry function. 7288 + Support of the NORM2 and PARITY intrinsic functions. 7289 + The following bit intrinsics were added: POPCNT and POPPAR for 7290 counting the number of 1 bits and returning the parity; BGE, 7291 BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR 7292 for combined left and right shifts, MASKL and MASKR for simple 7293 left and right justified masks, MERGE_BITS for a bitwise merge 7294 using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations, 7295 and the transformational bit intrinsics IALL, IANY and 7296 IPARITY. 7297 + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine. 7298 + Support for the IMPURE attribute for procedures, which allows 7299 for ELEMENTAL procedures without the restrictions of PURE. 7300 + Null pointers (including NULL()) and not allocated variables 7301 can be used as actual argument to optional non-pointer, 7302 non-allocatable dummy arguments, denoting an absent argument. 7303 + Non-pointer variables with TARGET attribute can be used as 7304 actual argument to POINTER dummies with INTENT(IN) 7305 + Pointers including procedure pointers and those in a derived 7306 type (pointer components) can now be initialized by a target 7307 instead of only by NULL. 7308 + The EXIT statement (with construct-name) can now be used to 7309 leave not only the DO but also the ASSOCIATE, BLOCK, IF, 7310 SELECT CASE and SELECT TYPE constructs. 7311 + Internal procedures can now be used as actual argument. 7312 + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS 7313 and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV 7314 have been added; these arrays contain the supported kind 7315 values for the respective types. 7316 + The module procedures C_SIZEOF of the intrinsic module 7317 ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of 7318 ISO_FORTRAN_ENV have been implemented. 7319 + Minor changes: obsolescence diagnostics for ENTRY was added 7320 for -std=f2008; a line may start with a semicolon; for 7321 internal and module procedures END can be used instead of END 7322 SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes 7323 a RADIX argument; intrinsic types are supported for 7324 TYPE(intrinsic-type-spec); multiple type-bound procedures can 7325 be declared in a single PROCEDURE statement; implied-shape 7326 arrays are supported for named constants (PARAMETER). The 7327 transformational, three argument versions of BESSEL_JN and 7328 BESSEL_YN were added the elemental, two-argument version had 7329 been added in GCC 4.4; note that the transformational 7330 functions use a recurrence algorithm. 7331 7332 Go 7333 7334 Support for the [19]Go programming language has been added to GCC. It 7335 is not enabled by default when you build GCC; use the 7336 --enable-languages configure option to build it. The driver program for 7337 compiling Go code is gccgo. 7338 7339 Go is currently known to work on GNU/Linux and RTEMS. Solaris support 7340 is in progress. It may or may not work on other platforms. 7341 7342 Objective-C and Objective-C++ 7343 7344 * The -fobjc-exceptions flag is now required to enable Objective-C 7345 exception and synchronization syntax (introduced by the keywords 7346 @try, @catch, @finally and @synchronized). 7347 * A number of Objective-C 2.0 features and extensions are now 7348 supported by GCC. These features are enabled by default; you can 7349 disable them by using the new -fobjc-std=objc1 command-line option. 7350 * The Objective-C 2.0 dot-syntax is now supported. It is an 7351 alternative syntax for using getters and setters; object.count is 7352 automatically converted into [object count] or [object setCount: 7353 ...] depending on context; for example if (object.count > 0) is 7354 automatically compiled into the equivalent of if ([object count] > 7355 0) while object.count = 0; is automatically compiled into the 7356 equivalent ot [object setCount: 0];. The dot-syntax can be used 7357 with instance and class objects and with any setters or getters, no 7358 matter if they are part of a declared property or not. 7359 * Objective-C 2.0 declared properties are now supported. They are 7360 declared using the new @property keyword, and are most commonly 7361 used in conjunction with the new Objective-C 2.0 dot-syntax. The 7362 nonatomic, readonly, readwrite, assign, retain, copy, setter and 7363 getter attributes are all supported. Marking declared properties 7364 with __attribute__ ((deprecated)) is supported too. 7365 * The Objective-C 2.0 @synthesize and @dynamic keywords are 7366 supported. @synthesize causes the compiler to automatically 7367 synthesize a declared property, while @dynamic is used to disable 7368 all warnings for a declared property for which no implementation is 7369 provided at compile time. Synthesizing declared properties requires 7370 runtime support in most useful cases; to be able to use it with the 7371 GNU runtime, appropriate helper functions have been added to the 7372 GNU Objective-C runtime ABI, and are implemented by the GNU 7373 Objective-C runtime library shipped with GCC. 7374 * The Objective-C 2.0 fast enumeration syntax is supported in 7375 Objective-C. This is currently not yet available in Objective-C++. 7376 Fast enumeration requires support in the runtime, and such support 7377 has been added to the GNU Objective-C runtime library (shipped with 7378 GCC). 7379 * The Objective-C 2.0 @optional keyword is supported. It allows you 7380 to mark methods or properties in a protocol as optional as opposed 7381 to required. 7382 * The Objective-C 2.0 @package keyword is supported. It has currently 7383 the same effect as the @public keyword. 7384 * Objective-C 2.0 method attributes are supported. Currently the 7385 supported attributes are deprecated, sentinel, noreturn and format. 7386 * Objective-C 2.0 method argument attributes are supported. The most 7387 widely used attribute is unused, to mark an argument as unused in 7388 the implementation. 7389 * Objective-C 2.0 class and protocol attributes are supported. 7390 Currently the only supported attribute is deprecated. 7391 * Objective-C 2.0 class extensions are supported. A class extension 7392 has the same syntax as a category declaration with no category 7393 name, and the methods and properties declared in it are added 7394 directly to the main class. It is mostly used as an alternative to 7395 a category to add methods to a class without advertising them in 7396 the public headers, with the advantage that for class extensions 7397 the compiler checks that all the privately declared methods are 7398 actually implemented. 7399 * As a result of these enhancements, GCC can now be used to build 7400 Objective-C and Objective-C++ software that uses Foundation and 7401 other important system frameworks with the NeXT runtime on Darwin 9 7402 and Darwin 10 (OSX 10.5 and 10.6). 7403 * Many bugs in the compiler have been fixed in this release; in 7404 particular, LTO can now be used when compiling Objective-C and 7405 Objective-C++ and the parser is much more robust in dealing with 7406 invalid code. 7407 7408 Runtime Library (libobjc) 7409 7410 * The GNU Objective-C runtime library now defines the macro 7411 __GNU_LIBOBJC__ (with a value that is increased at every release 7412 where there is any change to the API) in objc/objc.h, making it 7413 easy to determine if the GNU Objective-C runtime library is being 7414 used, and if so, which version. Previous versions of the GNU 7415 Objective-C runtime library (and other Objective-C runtime 7416 libraries such as the Apple one) do not define this macro. 7417 * A new Objective-C 2.0 API, almost identical to the one implemented 7418 by the Apple Objective-C runtime, has been implemented in the GNU 7419 Objective-C runtime library. The new API hides the internals of 7420 most runtime structures but provides a more extensive set of 7421 functions to operate on them. It is much easier, for example, to 7422 create or modify classes at runtime. The new API also makes it 7423 easier to port software from Apple to GNU as almost no changes 7424 should be required. The old API is still supported for backwards 7425 compatibility; including the old objc/objc-api.h header file 7426 automatically selects the old API, while including the new 7427 objc/runtime.h header file automatically selects the new API. 7428 Support for the old API is being phased out and upgrading the 7429 software to use the new API is strongly recommended. To check for 7430 the availability of the new API, the __GNU_LIBOBJC__ macro can be 7431 used as older versions of the GNU Objective-C runtime library, 7432 which do not support the new API, do not define such a macro. 7433 * Runtime support for @synchronized has been added. 7434 * Runtime support for Objective-C 2.0 synthesized property accessors 7435 has been added. 7436 * Runtime support for Objective-C 2.0 fast enumeration has been 7437 added. 7438 7439 New Targets and Target Specific Improvements 7440 7441 ARM 7442 7443 * GCC now supports the Cortex-M4 processor implementing the v7-em 7444 version of the architecture using the option -mcpu=cortex-m4. 7445 * Scheduling descriptions for the Cortex-M4, the Neon and the 7446 floating point units of the Cortex-A9 and a pipeline description 7447 for the Cortex-A5 have been added. 7448 * Synchronization primitives such as __sync_fetch_and_add and friends 7449 are now inlined for supported architectures rather than calling 7450 into a kernel helper function. 7451 * SSA loop prefetching is enabled by default for the Cortex-A9 at 7452 -O3. 7453 * Several improvements were committed to improve code generation for 7454 the ARM architecture including a rewritten implementation for load 7455 and store multiples. 7456 * Several enhancements were committed to improve SIMD code generation 7457 for NEON by adding support for widening instructions, misaligned 7458 loads and stores, vector conditionals and support for 64 bit 7459 arithmetic. 7460 * Support was added for the Faraday cores fa526, fa606te, fa626te, 7461 fmp626te, fmp626 and fa726te and can be used with the respective 7462 names as parameters to the -mcpu= option. 7463 * Basic support was added for Cortex-A15 and is available through 7464 -mcpu=cortex-a15. 7465 * GCC for AAPCS configurations now more closely adheres to the AAPCS 7466 specification by enabling -fstrict-volatile-bitfields by default. 7467 7468 IA-32/x86-64 7469 7470 * The new -fsplit-stack option permits programs to use a 7471 discontiguous stack. This is useful for threaded programs, in that 7472 it is no longer necessary to specify the maximum stack size when 7473 creating a thread. This feature is currently only implemented for 7474 32-bit and 64-bit x86 GNU/Linux targets. 7475 * Support for emitting profiler counter calls before function 7476 prologues. This is enabled via a new command-line option -mfentry. 7477 * Optimization for the Intel Core 2 processors is now available 7478 through the -march=core2 and -mtune=core2 options. 7479 * Support for Intel Core i3/i5/i7 processors is now available through 7480 the -march=corei7 and -mtune=corei7 options. 7481 * Support for Intel Core i3/i5/i7 processors with AVX is now 7482 available through the -march=corei7-avx and -mtune=corei7-avx 7483 options. 7484 * Support for AMD Bobcat (family 14) processors is now available 7485 through the -march=btver1 and -mtune=btver1 options. 7486 * Support for AMD Bulldozer (family 15) processors is now available 7487 through the -march=bdver1 and -mtune=bdver1 options. 7488 * The default setting (when not optimizing for size) for 32-bit 7489 GNU/Linux and Darwin x86 targets has been changed to 7490 -fomit-frame-pointer. The default can be reverted to 7491 -fno-omit-frame-pointer by configuring GCC with the 7492 --enable-frame-pointer configure option. 7493 * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support 7494 __float128 on 32-bit and 64-bit x86 targets. 7495 * AVX floating-point arithmetic can now be enabled by default at 7496 configure time with the new --with-fpmath=avx option. 7497 * The SSA loop prefetching pass is enabled when using -O3 when 7498 optimizing for CPUs where prefetching is beneficial (AMD CPUs newer 7499 than K6). 7500 * Support for TBM (Trailing Bit Manipulation) built-in functions and 7501 code generation is available via -mtbm. 7502 * Support for AMD's BMI (Bit Manipulation) built-in functions and 7503 code generation is available via -mbmi. 7504 7505 MicroBlaze 7506 7507 * Support has been added for the Xilinx MicroBlaze softcore processor 7508 (microblaze-elf) embedded target. This configurable processor is 7509 supported on several Xilinx Spartan and Virtex FPGAs. 7510 7511 MIPS 7512 7513 * GCC now supports the Loongson 3A processor. Its canonical -march= 7514 and -mtune= name is loongson3a. 7515 7516 MN10300 / AM33 7517 7518 * The inline assembly register constraint "A" has been renamed "c". 7519 This constraint is used to select a floating-point register that 7520 can be used as the destination of a multiply-accumulate 7521 instruction. 7522 * New inline assembly register constraints "A" and "D" have been 7523 added. These constraint letters resolve to all general registers 7524 when compiling for AM33, and resolve to address registers only or 7525 data registers only when compiling for MN10300. 7526 * The MDR register is represented in the compiler. One can access the 7527 register via the "z" constraint in inline assembly. It can be 7528 marked as clobbered or used as a local register variable via the 7529 "mdr" name. The compiler uses the RETF instruction if the function 7530 does not modify the MDR register, so it is important that inline 7531 assembly properly annotate any usage of the register. 7532 7533 PowerPC/PowerPC64 7534 7535 * GCC now supports the Applied Micro Titan processor with 7536 -mcpu=titan. 7537 * The -mrecip option has been added, which indicates whether the 7538 reciprocal and reciprocal square root instructions should be used. 7539 * The -mveclibabi=mass option can be used to enable the compiler to 7540 autovectorize mathematical functions using the Mathematical 7541 Acceleration Subsystem library. 7542 * The -msingle-pic-base option has been added, which instructs the 7543 compiler to avoid loading the PIC base register in function 7544 prologues. The PIC base register must be initialized by the runtime 7545 system. 7546 * The -mblock-move-inline-limit option has been added, which enables 7547 the user to control the maximum size of inlined memcpy calls and 7548 similar. 7549 * PowerPC64 GNU/Linux support for applications requiring a large TOC 7550 section has been improved. A new command-line option, 7551 -mcmodel=MODEL, controls this feature; valid values for MODEL are 7552 small, medium, or large. 7553 * The Altivec builtin functions vec_ld and vec_st have been modified 7554 to generate the Altivec memory instructions LVX and STVX, even if 7555 the -mvsx option is used. In the initial GCC 4.5 release, these 7556 builtin functions were changed to generate VSX memory reference 7557 instructions instead of Altivec memory instructions, but there are 7558 differences between the two instructions. If the VSX instruction 7559 set is available, you can now use the new builtin functions 7560 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 7561 instructions. 7562 * The GCC compiler on AIX now defaults to a process layout with a 7563 larger data space allowing larger programs to be compiled. 7564 * The GCC long double type on AIX 6.1 and above has reverted to 64 7565 bit double precision, matching the AIX XL compiler default, because 7566 of missing C99 symbols required by the GCC runtime. 7567 * The default processor scheduling model and tuning for PowerPC64 7568 GNU/Linux and for AIX 6.1 and above now is POWER7. 7569 * Starting with GCC 4.6.1, vectors of type vector long long or vector 7570 long are passed and returned in the same method as other vectors 7571 with the VSX instruction set. Previously the GCC compiler did not 7572 adhere to the ABI for 128-bit vectors with 64-bit integer base 7573 types (PR 48857). This is also fixed in the GCC 4.5.4 release. 7574 7575 S/390, zSeries and System z9/z10, IBM zEnterprise z196 7576 7577 * Support for the zEnterprise z196 processor has been added. When 7578 using the -march=z196 option, the compiler will generate code 7579 making use of the following instruction facilities: 7580 + Conditional load/store 7581 + Distinct-operands 7582 + Floating-point-extension 7583 + Interlocked-access 7584 + Population-count 7585 The -mtune=z196 option avoids the compare and branch instructions 7586 as well as the load address instruction with an index register as 7587 much as possible and performs instruction scheduling appropriate 7588 for the new out-of-order pipeline architecture. 7589 * When using the -m31 -mzarch options the generated code still 7590 conforms to the 32-bit ABI but uses the general purpose registers 7591 as 64-bit registers internally. This requires a Linux kernel saving 7592 the whole 64-bit registers when doing a context switch. Kernels 7593 providing that feature indicate that by the 'highgprs' string in 7594 /proc/cpuinfo. 7595 * The SSA loop prefetching pass is enabled when using -O3. 7596 7597 SPARC 7598 7599 * GCC now supports the LEON series of SPARC V8 processors. The code 7600 generated by the compiler can either be tuned to it by means of the 7601 --with-tune=leon configure option and -mtune=leon compilation 7602 option, or the compiler can be built for the sparc-leon-{elf,linux} 7603 and sparc-leon3-{elf,linux} targets directly. 7604 * GCC has stopped sign/zero-extending parameter registers in the 7605 callee for functions taking parameters with sub-word size in 32-bit 7606 mode, since this is redundant with the specification of the ABI. 7607 GCC has never done so in 64-bit mode since this is also redundant. 7608 * The command-line option -mfix-at697f has been added to enable the 7609 documented workaround for the single erratum of the Atmel AT697F 7610 processor. 7611 7612 Operating Systems 7613 7614 Android 7615 7616 * GCC now supports the Bionic C library and provides a convenient way 7617 of building native libraries and applications for the Android 7618 platform. Refer to the documentation of the -mandroid and -mbionic 7619 options for details on building native code. At the moment, Android 7620 support is enabled only for ARM. 7621 7622 Darwin/Mac OS X 7623 7624 * General 7625 + Initial support for CFString types has been added. 7626 This allows GCC to build projects including the system Core 7627 Foundation frameworks. The GCC Objective-C family supports 7628 CFString "toll-free bridged" as per the Mac OS X system tools. 7629 CFString is also recognized in the context of format 7630 attributes and arguments (see the documentation for format 7631 attributes for limitations). At present, 8-bit character types 7632 are supported. 7633 + Object file size reduction. 7634 The Darwin zeroed memory allocators have been re-written to 7635 make more use of .zerofill sections. For non-debug code, this 7636 can reduce object file size significantly. 7637 + Objective-C family 64-bit support (NeXT ABI 2). 7638 Initial support has been added to support 64-bit Objective-C 7639 code using the Darwin/OS X native (NeXT) runtime. ABI version 7640 2 will be selected automatically when 64-bit code is built. 7641 + Objective-C family 32-bit ABI 1. 7642 For 32-bit code ABI 1 is also now also allowed. At present it 7643 must be selected manually using -fobjc-abi-version=1 where 7644 applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6). 7645 * x86 Architecture 7646 + The -mdynamic-no-pic option has been enabled. 7647 Code supporting -mdynamic-no-pic optimization has been added 7648 and is applicable to -m32 builds. The compiler bootstrap uses 7649 the option where appropriate. 7650 + The default value for -mtune= has been changed. 7651 Since Darwin systems are primarily Xeon, Core-2 or similar the 7652 default tuning has been changed to -mtune=core2. 7653 + Enable 128-bit long double (__float128) support on Darwin. 7654 * PPC Architecture 7655 + Darwin64 ABI. 7656 Several significant bugs have been fixed, such that GCC now 7657 produces code compatible with the Darwin64 PowerPC ABI. 7658 + libffi and boehm-gc. 7659 The Darwin ports of the libffi and boehm-gc libraries have 7660 been upgraded to include a Darwin64 implementation. This means 7661 that powerpc*-*-darwin9 platforms may now, for example, build 7662 Java applications with -m64 enabled. 7663 + Plug-in support has been enabled. 7664 + The -fsection-anchors option is now available although, 7665 presently, not heavily tested. 7666 7667 Solaris 2 7668 7669 New Features 7670 7671 * Support symbol versioning with the Sun linker. 7672 * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+. 7673 * Support thread-local storage (TLS) with the Sun assembler on 7674 Solaris 2/x86. 7675 * Support TLS on Solaris 8/9 if prerequisites are met. 7676 * Support COMDAT group with the GNU assembler and recent Sun linker. 7677 * Support the Sun assembler visibility syntax. 7678 * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp. 7679 -march=pentiumpro (Solaris 8/9). 7680 * Don't use SSE on Solaris 8/9 x86 by default. 7681 * Enable 128-bit long double (__float128) support on Solaris 2/x86. 7682 7683 ABI Change 7684 7685 * Change the ABI for returning 8-byte vectors like __m64 in MMX 7686 registers on Solaris 10+/x86 to match the Sun Studio 12.1+ 7687 compilers. This is an incompatible change. If you use such types, 7688 you must either recompile all your code with the new compiler or 7689 use the new -mvect8-ret-in-mem option to remain compatible with 7690 previous versions of GCC and Sun Studio. 7691 7692 Windows x86/x86_64 7693 7694 * Initial support for decimal floating point. 7695 * Support for the __thiscall calling-convention. 7696 * Support for hot-patchable function prologues via the 7697 ms_hook_prologue attribute for x86_64 in addition to 32-bit x86. 7698 * Improvements of stack-probing and stack-allocation mechanisms. 7699 * Support of push/pop-macro pragma as preprocessor command. 7700 With #pragma push_macro("macro-name") the current definition of 7701 macro-name is saved and can be restored with #pragma 7702 pop_macro("macro-name") to its saved definition. 7703 * Enable 128-bit long double (__float128) support on MinGW and 7704 Cygwin. 7705 7706 Other significant improvements 7707 7708 Installation changes 7709 7710 * An install-strip make target is provided that installs stripped 7711 executables, and may install libraries with unneeded or debugging 7712 sections stripped. 7713 * On Power7 systems, there is a potential problem if you build the 7714 GCC compiler with a host compiler using options that enable the VSX 7715 instruction set generation. If the host compiler has been patched 7716 so that the vec_ld and vec_st builtin functions generate Altivec 7717 memory instructions instead of VSX memory instructions, then you 7718 should be able to build the compiler with VSX instruction 7719 generation. 7720 7721 Changes for GCC Developers 7722 7723 Note: these changes concern developers that develop GCC itself or 7724 software that integrates with GCC, such as plugins, and not the general 7725 GCC users. 7726 * The gengtype utility, which previously was internal to the GCC 7727 build process, has been enchanced to provide GC root information 7728 for plugins as necessary. 7729 * The old GC allocation interface of ggc_alloc and friends was 7730 replaced with a type-safe alternative. 7731 7732 GCC 4.6.1 7733 7734 This is the [20]list of problem reports (PRs) from GCC's bug tracking 7735 system that are known to be fixed in the 4.6.1 release. This list might 7736 not be complete (that is, it is possible that some PRs that have been 7737 fixed are not listed here). 7738 7739 GCC 4.6.2 7740 7741 This is the [21]list of problem reports (PRs) from GCC's bug tracking 7742 system that are known to be fixed in the 4.6.2 release. This list might 7743 not be complete (that is, it is possible that some PRs that have been 7744 fixed are not listed here). 7745 7746 GCC 4.6.3 7747 7748 This is the [22]list of problem reports (PRs) from GCC's bug tracking 7749 system that are known to be fixed in the 4.6.3 release. This list might 7750 not be complete (that is, it is possible that some PRs that have been 7751 fixed are not listed here). 7752 7753 GCC 4.6.4 7754 7755 This is the [23]list of problem reports (PRs) from GCC's bug tracking 7756 system that are known to be fixed in the 4.6.4 release. This list might 7757 not be complete (that is, it is possible that some PRs that have been 7758 fixed are not listed here). 7759 7760 7761 For questions related to the use of GCC, please consult these web 7762 pages and the [24]GCC manuals. If that fails, the 7763 [25]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7764 web pages and the development of GCC are welcome on our developer 7765 list at [26]gcc (a] gcc.gnu.org. All of [27]our lists have public 7766 archives. 7767 7768 Copyright (C) [28]Free Software Foundation, Inc. Verbatim copying and 7769 distribution of this entire article is permitted in any medium, 7770 provided this notice is preserved. 7771 7772 These pages are [29]maintained by the GCC team. Last modified 7773 2018-09-30[30]. 7774 7775 References 7776 7777 1. https://sourceware.org/bugzilla/show_bug.cgi?id=10401 7778 2. http://gcc.gnu.org/gcc-4.5/changes.html 7779 3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted 7780 4. http://gcc.gnu.org/gcc-4.6/porting_to.html 7781 5. http://gcc.gnu.org/projects/lto/whopr.pdf 7782 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 7783 7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html 7784 8. https://gcc.gnu.org/PR43145 7785 9. https://gcc.gnu.org/PR43680 7786 10. https://gcc.gnu.org/PR33558 7787 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253 7788 12. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x 7789 13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html 7790 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races 7791 15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html 7792 16. https://gcc.gnu.org/wiki/OOP 7793 17. https://gcc.gnu.org/wiki/Coarray 7794 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233 7795 19. https://golang.org/ 7796 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1 7797 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2 7798 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3 7799 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4 7800 24. https://gcc.gnu.org/onlinedocs/ 7801 25. mailto:gcc-help (a] gcc.gnu.org 7802 26. mailto:gcc (a] gcc.gnu.org 7803 27. https://gcc.gnu.org/lists.html 7804 28. https://www.fsf.org/ 7805 29. https://gcc.gnu.org/about.html 7806 30. http://validator.w3.org/check/referer 7807 ====================================================================== 7808 http://gcc.gnu.org/gcc-4.5/index.html 7809 GCC 4.5 Release Series 7810 7811 Jul 2, 2012 7812 7813 The [1]GNU project and the GCC developers are pleased to announce the 7814 release of GCC 4.5.4. 7815 7816 This release is a bug-fix release, containing fixes for regressions in 7817 GCC 4.5.3 relative to previous releases of GCC. 7818 7819 This release series is no longer maintained. 7820 7821 Release History 7822 7823 GCC 4.5.4 7824 Jul 2, 2012 ([2]changes) 7825 7826 GCC 4.5.3 7827 Apr 28, 2011 ([3]changes) 7828 7829 GCC 4.5.2 7830 Dec 16, 2010 ([4]changes) 7831 7832 GCC 4.5.1 7833 Jul 31, 2010 ([5]changes) 7834 7835 GCC 4.5.0 7836 April 14, 2010 ([6]changes) 7837 7838 References and Acknowledgements 7839 7840 GCC used to stand for the GNU C Compiler, but since the compiler 7841 supports several other languages aside from C, it now stands for the 7842 GNU Compiler Collection. 7843 7844 A list of [7]successful builds is updated as new information becomes 7845 available. 7846 7847 The GCC developers would like to thank the numerous people that have 7848 contributed new features, improvements, bug fixes, and other changes as 7849 well as test results to GCC. This [8]amazing group of volunteers is 7850 what makes GCC successful. 7851 7852 For additional information about GCC please refer to the [9]GCC project 7853 web site or contact the [10]GCC development mailing list. 7854 7855 To obtain GCC please use [11]our mirror sites or [12]our SVN server. 7856 7857 7858 For questions related to the use of GCC, please consult these web 7859 pages and the [13]GCC manuals. If that fails, the 7860 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7861 web pages and the development of GCC are welcome on our developer 7862 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 7863 archives. 7864 7865 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 7866 distribution of this entire article is permitted in any medium, 7867 provided this notice is preserved. 7868 7869 These pages are [18]maintained by the GCC team. Last modified 7870 2018-09-30[19]. 7871 7872 References 7873 7874 1. http://www.gnu.org/ 7875 2. http://gcc.gnu.org/gcc-4.5/changes.html 7876 3. http://gcc.gnu.org/gcc-4.5/changes.html 7877 4. http://gcc.gnu.org/gcc-4.5/changes.html 7878 5. http://gcc.gnu.org/gcc-4.5/changes.html 7879 6. http://gcc.gnu.org/gcc-4.5/changes.html 7880 7. http://gcc.gnu.org/gcc-4.5/buildstat.html 7881 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7882 9. http://gcc.gnu.org/index.html 7883 10. mailto:gcc (a] gcc.gnu.org 7884 11. http://gcc.gnu.org/mirrors.html 7885 12. http://gcc.gnu.org/svn.html 7886 13. https://gcc.gnu.org/onlinedocs/ 7887 14. mailto:gcc-help (a] gcc.gnu.org 7888 15. mailto:gcc (a] gcc.gnu.org 7889 16. https://gcc.gnu.org/lists.html 7890 17. https://www.fsf.org/ 7891 18. https://gcc.gnu.org/about.html 7892 19. http://validator.w3.org/check/referer 7893 ====================================================================== 7894 http://gcc.gnu.org/gcc-4.5/changes.html 7895 GCC 4.5 Release Series 7896 Changes, New Features, and Fixes 7897 7898 Caveats 7899 7900 * GCC now requires the [1]MPC library in order to build. See the 7901 [2]prerequisites page for version requirements. 7902 * Support for a number of older systems and recently unmaintained or 7903 untested target ports of GCC has been declared obsolete in GCC 4.5. 7904 Unless there is activity to revive them, the next release of GCC 7905 will have their sources permanently removed. 7906 The following ports for individual systems on particular 7907 architectures have been obsoleted: 7908 + IRIX releases before 6.5 (mips-sgi-irix5*, 7909 mips-sgi-irix6.[0-4]) 7910 + Solaris 7 (*-*-solaris2.7) 7911 + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*, 7912 alpha-dec-osf5.0*) 7913 + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions 7914 can be found in the [3]announcement. 7915 Support for the classic POWER architecture implemented in the 7916 original RIOS and RIOS2 processors of the old IBM RS/6000 product 7917 line has been obsoleted in the rs6000 port. This does not affect 7918 the new generation Power and PowerPC architectures. 7919 * Support has been removed for all the [4]configurations obsoleted in 7920 GCC 4.4. 7921 * Support has been removed for the protoize and unprotoize utilities, 7922 obsoleted in GCC 4.4. 7923 * Support has been removed for tuning for Itanium1 (Merced) variants. 7924 Note that code tuned for Itanium2 should also run correctly on 7925 Itanium1. 7926 * GCC now generates unwind info also for epilogues. DWARF debuginfo 7927 generated by GCC now uses more features of DWARF3 than before, and 7928 also some DWARF4 features. GDB older than 7.0 is not able to handle 7929 either of these, so to debug GCC 4.5 generated binaries or 7930 libraries GDB 7.0 or later is needed. You can disable use of DWARF4 7931 features with the -gdwarf-3 -gstrict-dwarf options, or use 7932 -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but 7933 epilogue unwind info is emitted unconditionally whenever unwind 7934 info is emitted. 7935 * On x86 targets, code containing floating-point calculations may run 7936 significantly more slowly when compiled with GCC 4.5 in strict C99 7937 conformance mode than they did with earlier GCC versions. This is 7938 due to stricter standard conformance of the compiler and can be 7939 avoided by using the option -fexcess-precision=fast; also see 7940 [5]below. 7941 * The function attribute noinline no longer prevents GCC from cloning 7942 the function. A new attribute noclone has been introduced for this 7943 purpose. Cloning a function means that it is duplicated and the new 7944 copy is specialized for certain contexts (for example when a 7945 parameter is a known constant). 7946 7947 General Optimizer Improvements 7948 7949 * The -save-temps now takes an optional argument. The -save-temps and 7950 -save-temps=cwd switches write the temporary files in the current 7951 working directory based on the original source file. The 7952 -save-temps=obj switch will write files into the directory 7953 specified with the -o option, and the intermediate filenames are 7954 based on the output file. This will allow the user to get the 7955 compiler intermediate files when doing parallel builds without two 7956 builds of the same filename located in different directories from 7957 interfering with each other. 7958 * Debugging dumps are now created in the same directory as the object 7959 file rather than in the current working directory. This allows the 7960 user to get debugging dumps when doing parallel builds without two 7961 builds of the same filename interfering with each other. 7962 * GCC has been integrated with the MPC library. This allows GCC to 7963 evaluate complex arithmetic at compile time [6]more accurately. It 7964 also allows GCC to evaluate calls to complex built-in math 7965 functions having constant arguments and replace them at compile 7966 time with their mathematically equivalent results. In doing so, GCC 7967 can generate correct results regardless of the math library 7968 implementation or floating point precision of the host platform. 7969 This also allows GCC to generate identical results regardless of 7970 whether one compiles in native or cross-compile configurations to a 7971 particular target. The following built-in functions take advantage 7972 of this new capability: cacos, cacosh, casin, casinh, catan, 7973 catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan, 7974 and ctanh. The float and long double variants of these functions 7975 (e.g. csinf and csinl) are also handled. 7976 * A new link-time optimizer has been added ([7]-flto). When this 7977 option is used, GCC generates a bytecode representation of each 7978 input file and writes it to specially-named sections in each object 7979 file. When the object files are linked together, all the function 7980 bodies are read from these named sections and instantiated as if 7981 they had been part of the same translation unit. This enables 7982 interprocedural optimizations to work across different files (and 7983 even different languages), potentially improving the performance of 7984 the generated code. To use the link-timer optimizer, -flto needs to 7985 be specified at compile time and during the final link. If the 7986 program does not require any symbols to be exported, it is possible 7987 to combine -flto and the experimental [8]-fwhopr with 7988 [9]-fwhole-program to allow the interprocedural optimizers to use 7989 more aggressive assumptions. 7990 * The automatic parallelization pass was enhanced to support 7991 parallelization of outer loops. 7992 * Automatic parallelization can be enabled as part of Graphite. In 7993 addition to -ftree-parallelize-loops=, specify 7994 -floop-parallelize-all to enable the Graphite-based optimization. 7995 * The infrastructure for optimizing based on [10]restrict qualified 7996 pointers has been rewritten and should result in code generation 7997 improvements. Optimizations based on restrict qualified pointers 7998 are now also available when using -fno-strict-aliasing. 7999 * There is a new optimization pass that attempts to change prototype 8000 of functions to avoid unused parameters, pass only relevant parts 8001 of structures and turn arguments passed by reference to arguments 8002 passed by value when possible. It is enabled by -O2 and above as 8003 well as -Os and can be manually invoked using the new command-line 8004 switch -fipa-sra. 8005 * GCC now optimize exception handling code. In particular cleanup 8006 regions that are proved to not have any effect are optimized out. 8007 8008 New Languages and Language specific improvements 8009 8010 All languages 8011 8012 * The -fshow-column option is now on by default. This means error 8013 messages now have a column associated with them. 8014 8015 Ada 8016 8017 * Compilation of programs heavily using discriminated record types 8018 with variant parts has been sped up and generates more compact 8019 code. 8020 * Stack checking now works reasonably well on most plaforms. In some 8021 specific cases, stack overflows may still fail to be detected, but 8022 a compile-time warning will be issued for these cases. 8023 8024 C family 8025 8026 * If a header named in a #include directive is not found, the 8027 compiler exits immediately. This avoids a cascade of errors arising 8028 from declarations expected to be found in that header being 8029 missing. 8030 * A new built-in function __builtin_unreachable() has been added that 8031 tells the compiler that control will never reach that point. It may 8032 be used after asm statements that terminate by transferring control 8033 elsewhere, and in other places that are known to be unreachable. 8034 * The -Wlogical-op option now warns for logical expressions such as 8035 (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be 8036 mistakes. This option is disabled by default. 8037 * An asm goto feature has been added to allow asm statements that 8038 jump to C labels. 8039 * C++0x raw strings are supported for C++ and for C with -std=gnu99. 8040 * The deprecated attribute now takes an optional string argument, for 8041 example, __attribute__((deprecated("text string"))), that will be 8042 printed together with the deprecation warning. 8043 8044 C 8045 8046 * The -Wenum-compare option, which warns when comparing values of 8047 different enum types, now works for C. It formerly only worked for 8048 C++. This warning is enabled by -Wall. It may be avoided by using a 8049 type cast. 8050 * The -Wcast-qual option now warns about casts which are unsafe in 8051 that they permit const-correctness to be violated without further 8052 warnings. Specifically, it warns about cases where a qualifier is 8053 added when all the lower types are not const. For example, it warns 8054 about a cast from char ** to const char **. 8055 * The -Wc++-compat option is significantly improved. It issues new 8056 warnings for: 8057 + Using C++ reserved operator names as identifiers. 8058 + Conversions to enum types without explicit casts. 8059 + Using va_arg with an enum type. 8060 + Using different enum types in the two branches of ?:. 8061 + Using ++ or -- on a variable of enum type. 8062 + Using the same name as both a struct, union or enum tag and a 8063 typedef, unless the typedef refers to the tagged type itself. 8064 + Using a struct, union, or enum which is defined within another 8065 struct or union. 8066 + A struct field defined using a typedef if there is a field in 8067 the struct, or an enclosing struct, whose name is the typedef 8068 name. 8069 + Duplicate definitions at file scope. 8070 + Uninitialized const variables. 8071 + A global variable with an anonymous struct, union, or enum 8072 type. 8073 + Using a string constant to initialize a char array whose size 8074 is the length of the string. 8075 * The new -Wjump-misses-init option warns about cases where a goto or 8076 switch skips the initialization of a variable. This sort of branch 8077 is an error in C++ but not in C. This warning is enabled by 8078 -Wc++-compat. 8079 * GCC now ensures that a C99-conforming <stdint.h> is present on most 8080 targets, and uses information about the types in this header to 8081 implement the Fortran bindings to those types. GCC does not ensure 8082 the presence of such a header, and does not implement the Fortran 8083 bindings, on the following targets: NetBSD, VxWorks, VMS, 8084 SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF. 8085 * GCC now implements C90- and C99-conforming rules for constant 8086 expressions. This may cause warnings or errors for some code using 8087 expressions that can be folded to a constant but are not constant 8088 expressions as defined by ISO C. 8089 * All known target-independent C90 and C90 Amendment 1 conformance 8090 bugs, and all known target-independent C99 conformance bugs not 8091 related to floating point or extended identifiers, have been fixed. 8092 * The C decimal floating point support now includes support for the 8093 FLOAT_CONST_DECIMAL64 pragma. 8094 * The named address space feature from ISO/IEC TR 18037 is now 8095 supported. This is currently only implemented for the SPU 8096 processor. 8097 8098 C++ 8099 8100 * Improved [11]experimental support for the upcoming C++0x ISO C++ 8101 standard, including support for raw strings, lambda expressions and 8102 explicit type conversion operators. 8103 * When printing the name of a class template specialization, G++ will 8104 now omit any template arguments which come from default template 8105 arguments. This behavior (and the pretty-printing of function 8106 template specializations as template signature and arguments) can 8107 be disabled with the -fno-pretty-templates option. 8108 * Access control is now applied to typedef names used in a template, 8109 which may cause G++ to reject some ill-formed code that was 8110 accepted by earlier releases. The -fno-access-control option can be 8111 used as a temporary workaround until the code is corrected. 8112 * Compilation time for code that uses templates should now scale 8113 linearly with the number of instantiations rather than 8114 quadratically, as template instantiations are now looked up using 8115 hash tables. 8116 * Declarations of functions that look like builtin declarations of 8117 library functions are only considered to be redeclarations if they 8118 are declared with extern "C". This may cause problems with code 8119 that omits extern "C" on hand-written declarations of C library 8120 functions such as abort or memcpy. Such code is ill-formed, but was 8121 accepted by earlier releases. 8122 * Diagnostics that used to complain about passing non-POD types to 8123 ... or jumping past the declaration of a non-POD variable now check 8124 for triviality rather than PODness, as per C++0x. 8125 * In C++0x mode local and anonymous classes are now allowed as 8126 template arguments, and in declarations of variables and functions 8127 with linkage, so long as any such declaration that is used is also 8128 defined ([12]DR 757). 8129 * Labels may now have attributes, as has been permitted for a while 8130 in C. This is only permitted when the label definition and the 8131 attribute specifier is followed by a semicoloni.e., the label 8132 applies to an empty statement. The only useful attribute for a 8133 label is unused. 8134 * G++ now implements [13]DR 176. Previously G++ did not support using 8135 the injected-class-name of a template base class as a type name, 8136 and lookup of the name found the declaration of the template in the 8137 enclosing scope. Now lookup of the name finds the 8138 injected-class-name, which can be used either as a type or as a 8139 template, depending on whether or not the name is followed by a 8140 template argument list. As a result of this change, some code that 8141 was previously accepted may be ill-formed because 8142 1. The injected-class-name is not accessible because it's from a 8143 private base, or 8144 2. The injected-class-name cannot be used as an argument for a 8145 template template parameter. 8146 In either of these cases, the code can be fixed by adding a 8147 nested-name-specifier to explicitly name the template. The first 8148 can be worked around with -fno-access-control; the second is only 8149 rejected with -pedantic. 8150 * A new standard mangling for SIMD vector types has been added, to 8151 avoid name clashes on systems with vectors of varying length. By 8152 default the compiler still uses the old mangling, but emits aliases 8153 with the new mangling on targets that support strong aliases. Users 8154 can switch over entirely to the new mangling with -fabi-version=4 8155 or -fabi-version=0. -Wabi will now warn about code that uses the 8156 old mangling. 8157 * The command-line option -ftemplate-depth-N is now written as 8158 -ftemplate-depth=N and the old form is deprecated. 8159 * Conversions between NULL and non-pointer types are now warned by 8160 default. The new option -Wno-conversion-null disables these 8161 warnings. Previously these warnings were only available when using 8162 -Wconversion explicitly. 8163 8164 Runtime Library (libstdc++) 8165 8166 * Improved experimental support for the upcoming ISO C++ standard, 8167 C++0x, including: 8168 + Support for <future>, <functional>, and <random>. 8169 + Existing facilities now exploit explicit operators and the 8170 newly implemented core C++0x features. 8171 + The header <cstdatomic> has been renamed to <atomic>. 8172 * An experimental [14]profile mode has been added. This is an 8173 implementation of many C++ standard library constructs with an 8174 additional analysis layer that gives performance improvement advice 8175 based on recognition of suboptimal usage patterns. For example, 8176 #include <vector> 8177 int main() 8178 { 8179 std::vector<int> v; 8180 for (int k = 0; k < 1024; ++k) 8181 v.insert(v.begin(), k); 8182 } 8183 8184 When instrumented via the profile mode, can return suggestions 8185 about the initial size and choice of the container used as follows: 8186 vector-to-list: improvement = 5: call stack = 0x804842c ... 8187 : advice = change std::vector to std::list 8188 vector-size: improvement = 3: call stack = 0x804842c ... 8189 : advice = change initial container size from 0 to 1024 8190 8191 These constructs can be substituted for the normal libstdc++ 8192 constructs on a piecemeal basis, or all existing components can be 8193 transformed via the -D_GLIBCXX_PROFILE macro. 8194 * [15]Support for decimal floating-point arithmetic (aka ISO C++ TR 8195 24733) has been added. This support is in header file 8196 <decimal/decimal>, uses namespace std::decimal, and includes 8197 classes decimal32, decimal64, and decimal128. 8198 * Sources have been audited for application of function attributes 8199 nothrow, const, pure, and noreturn. 8200 * Python pretty-printers have been added for many standard library 8201 components that simplify the internal representation and present a 8202 more intuitive view of components when used with 8203 appropriately-advanced versions of GDB. For more information, 8204 please consult the more [16]detailed description. 8205 * The default behavior for comparing typeinfo names has changed, so 8206 in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero. 8207 * The new -static-libstdc++ option directs g++ to link the C++ 8208 library statically, even if the default would normally be to link 8209 it dynamically. 8210 8211 Fortran 8212 8213 * The COMMON default padding has been changed instead of adding the 8214 padding before a variable it is now added afterwards, which 8215 increases the compatibility with other vendors and helps to obtain 8216 the correct output in some cases. Cf. also the -falign-commons 8217 option ([17]added in 4.4). 8218 * The -finit-real= option now also supports the value snan for 8219 signaling not-a-number; to be effective, one additionally needs to 8220 enable trapping (e.g. via -ffpe-trap=). Note: Compile-time 8221 optimizations can turn a signaling NaN into a quiet one. 8222 * The new option -fcheck= has been added with the options bounds, 8223 array-temps, do, pointer, and recursive. The bounds and array-temps 8224 options are equivalent to -fbounds-check and 8225 -fcheck-array-temporaries. The do option checks for invalid 8226 modification of loop iteration variables, and the recursive option 8227 tests for recursive calls to subroutines/functions which are not 8228 marked as recursive. With pointer pointer association checks in 8229 calls are performed; however, neither undefined pointers nor 8230 pointers in expressions are handled. Using -fcheck=all enables all 8231 these run-time checks. 8232 * The run-time checking -fcheck=bounds now warns about invalid string 8233 lengths of character dummy arguments. Additionally, more 8234 compile-time checks have been added. 8235 * The new option [18]-fno-protect-parens has been added; if set, the 8236 compiler may reorder REAL and COMPLEX expressions without regard to 8237 parentheses. 8238 * GNU Fortran no longer links against libgfortranbegin. As before, 8239 MAIN__ (assembler symbol name) is the actual Fortran main program, 8240 which is invoked by the main function. However, main is now 8241 generated and put in the same object file as MAIN__. For the time 8242 being, libgfortranbegin still exists for backward compatibility. 8243 For details see the new [19]Mixed-Language Programming chapter in 8244 the manual. 8245 * The I/O library was restructured for performance and cleaner code. 8246 * Array assignments and WHERE are now run in parallel when OpenMP's 8247 WORKSHARE is used. 8248 * The experimental option -fwhole-file was added. The option allows 8249 whole-file checking of procedure arguments and allows for better 8250 optimizations. It can also be used with -fwhole-program, which is 8251 now also supported in gfortran. 8252 * More Fortran 2003 and Fortran 2008 mathematical functions can now 8253 be used as initialization expressions. 8254 * Some extended attributes such as STDCALL are now supported via the 8255 [20]GCC$ compiler directive. 8256 * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN 8257 intrinsic behaves now as if zero were always positive. 8258 * For legacy compatibiliy: On Cygwin and MinGW, the special files 8259 CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now 8260 supported. 8261 * Fortran 2003 support has been extended: 8262 + Procedure-pointer function results and procedure-pointer 8263 components (including PASS), 8264 + allocatable scalars (experimental), 8265 + DEFERRED type-bound procedures, 8266 + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements 8267 have been implemented. 8268 + The ALLOCATE statement supports type-specs and the SOURCE= 8269 argument. 8270 + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC 8271 type-bound procedure (i.e. as type-bound operators). 8272 + Rounding (ROUND=, RZ, ...) for output is now supported. 8273 + The INT_FAST{8,16,32,64,128}_T kind type parameters of the 8274 intrinsic module ISO_C_BINDING are now supported, except for 8275 the targets listed above as ones where GCC does not have 8276 <stdint.h> type information. 8277 + Extensible derived types with type-bound procedure or 8278 procedure pointer with PASS attribute now have to use CLASS in 8279 line with the Fortran 2003 standard; the workaround to use 8280 TYPE is no longer supported. 8281 + [21]Experimental, incomplete support for polymorphism, 8282 including CLASS, SELECT TYPE and dynamic dispatch of 8283 type-bound procedure calls. Some features do not work yet such 8284 as unlimited polymorphism (CLASS(*)). 8285 * Fortran 2008 support has been extended: 8286 + The OPEN statement now supports the NEWUNIT= option, which 8287 returns a unique file unit, thus preventing inadvertent use of 8288 the same unit in different parts of the program. 8289 + Support for unlimited format items has been added. 8290 + The INT{8,16,32} and REAL{32,64,128} kind type parameters of 8291 the intrinsic module ISO_FORTRAN_ENV are now supported. 8292 + Using complex arguments with TAN, SINH, COSH, TANH, ASIN, 8293 ACOS, and ATAN is now possible; the functions ASINH, ACOSH, 8294 and ATANH have been added (for real and complex arguments) and 8295 ATAN(Y,X) is now an alias for ATAN2(Y,X). 8296 + The BLOCK construct has been implemented. 8297 8298 New Targets and Target Specific Improvements 8299 8300 AIX 8301 8302 * Full cross-toolchain support now available with GNU Binutils 8303 8304 ARM 8305 8306 * GCC now supports the Cortex-M0 and Cortex-A5 processors. 8307 * GCC now supports the ARM v7E-M architecture. 8308 * GCC now supports VFPv4-based FPUs and FPUs with 8309 single-precision-only VFP. 8310 * GCC has many improvements to optimization for other ARM processors, 8311 including scheduling support for the integer pipeline on Cortex-A9. 8312 * GCC now supports the IEEE 754-2008 half-precision floating-point 8313 type, and a variant ARM-specific half-precision type. This type is 8314 specified using __fp16, with the layout determined by 8315 -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and 8316 VFPv4 half-precision instructions will be used. 8317 * GCC now supports the variant of AAPCS that uses VFP registers for 8318 parameter passing and return values. 8319 8320 AVR 8321 8322 * The -mno-tablejump option has been removed because it has the same 8323 effect as the -fno-jump-tables option. 8324 * Added support for these new AVR devices: 8325 + ATmega8U2 8326 + ATmega16U2 8327 + ATmega32U2 8328 8329 IA-32/x86-64 8330 8331 * GCC now will set the default for -march= based on the configure 8332 target. 8333 * GCC now supports handling floating-point excess precision arising 8334 from use of the x87 floating-point unit in a way that conforms to 8335 ISO C99. This is enabled with -fexcess-precision=standard and with 8336 standards conformance options such as -std=c99, and may be disabled 8337 using -fexcess-precision=fast. 8338 * Support for the Intel Atom processor is now available through the 8339 -march=atom and -mtune=atom options. 8340 * A new -mcrc32 option is now available to enable crc32 intrinsics. 8341 * A new -mmovbe option is now available to enable GCC to use the 8342 movbe instruction to implement __builtin_bswap32 and 8343 __builtin_bswap64. 8344 * SSE math now can be enabled by default at configure time with the 8345 new --with-fpmath=sse option. 8346 * There is a new intrinsic header file, <x86intrin.h>. It should be 8347 included before using any IA-32/x86-64 intrinsics. 8348 * Support for the XOP, FMA4, and LWP instruction sets for the AMD 8349 Orochi processors are now available with the -mxop, -mfma4, and 8350 -mlwp options. 8351 * The -mabm option enables GCC to use the popcnt and lzcnt 8352 instructions on AMD processors. 8353 * The -mpopcnt option enables GCC to use the popcnt instructions on 8354 both AMD and Intel processors. 8355 8356 M68K/ColdFire 8357 8358 * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x 8359 and 5441x devices. 8360 * GCC now supports thread-local storage (TLS) on M68K and ColdFire 8361 processors. 8362 8363 MeP 8364 8365 Support has been added for the Toshiba Media embedded Processor (MeP, 8366 or mep-elf) embedded target. 8367 8368 MIPS 8369 8370 * GCC now supports MIPS 1004K processors. 8371 * GCC can now be configured with options --with-arch-32, 8372 --with-arch-64, --with-tune-32 and --with-tune-64 to control the 8373 default optimization separately for 32-bit and 64-bit modes. 8374 * MIPS targets now support an alternative _mcount interface, in which 8375 register $12 points to the function's save slot for register $31. 8376 This interface is selected by the -mcount-ra-address option; see 8377 the documentation for more details. 8378 * GNU/Linux targets can now generate read-only .eh_frame sections. 8379 This optimization requires GNU binutils 2.20 or above, and is only 8380 available if GCC is configured with a suitable version of binutils. 8381 * GNU/Linux targets can now attach special relocations to indirect 8382 calls, so that the linker can turn them into direct jumps or 8383 branches. This optimization requires GNU binutils 2.20 or later, 8384 and is automatically selected if GCC is configured with an 8385 appropriate version of binutils. It can be explicitly enabled or 8386 disabled using the -mrelax-pic-calls command-line option. 8387 * GCC now generates more heavily-optimized atomic operations on 8388 Octeon processors. 8389 * MIPS targets now support the -fstack-protector option. 8390 * GCC now supports an -msynci option, which specifies that synci is 8391 enough to flush the instruction cache, without help from the 8392 operating system. GCC uses this information to optimize 8393 automatically-generated cache flush operations, such as those used 8394 for nested functions in C. There is also a --with-synci 8395 configure-time option, which makes -msynci the default. 8396 * GCC supports four new function attributes for interrupt handlers: 8397 interrupt, use_shadow_register_set, keep_interrupts_masked and 8398 use_debug_exception_return. See the documentation for more details 8399 about these attributes. 8400 8401 RS/6000 (POWER/PowerPC) 8402 8403 * GCC now supports the Power ISA 2.06, which includes the VSX 8404 instructions that add vector 64-bit floating point support, new 8405 population count instructions, and conversions between floating 8406 point and unsigned types. 8407 * Support for the power7 processor is now available through the 8408 -mcpu=power7 and -mtune=power7. 8409 * GCC will now vectorize loops that contain simple math functions 8410 like copysign when generating code for altivec or VSX targets. 8411 * Support for the A2 processor is now available through the -mcpu=a2 8412 and -mtune=a2 options. 8413 * Support for the 476 processor is now available through the 8414 -mcpu={476,476fp} and -mtune={476,476fp} options. 8415 * Support for the e500mc64 processor is now available through the 8416 -mcpu=e500mc64 and -mtune=e500mc64 options. 8417 * GCC can now be configured with options --with-cpu-32, 8418 --with-cpu-64, --with-tune-32 and --with-tune-64 to control the 8419 default optimization separately for 32-bit and 64-bit modes. 8420 * Starting with GCC 4.5.4, vectors of type vector long long or vector 8421 long are passed and returned in the same method as other vectors 8422 with the VSX instruction set. Previously the GCC compiler did not 8423 adhere to the ABI for 128-bit vectors with 64-bit integer base 8424 types (PR 48857). This is also fixed in the GCC 4.6.1 release. 8425 8426 RX 8427 8428 Support has been added for the Renesas RX Processor (rx-elf) target. 8429 8430 Operating Systems 8431 8432 Windows (Cygwin and MinGW) 8433 8434 * GCC now installs all the major language runtime libraries as DLLs 8435 when configured with the --enable-shared option. 8436 * GCC now makes use of the new support for aligned common variables 8437 in versions of binutils >= 2.20 to fix bugs in the support for SSE 8438 data types. 8439 * Improvements to the libffi support library increase the reliability 8440 of code generated by GCJ on all Windows platforms. Libgcj is 8441 enabled by default for the first time. 8442 * Libtool improvements simplify installation by placing the generated 8443 DLLs in the correct binaries directory. 8444 * Numerous other minor bugfixes and improvements, and substantial 8445 enhancements to the Fortran language support library. 8446 8447 > 8448 8449 Other significant improvements 8450 8451 Plugins 8452 8453 * It is now possible to extend the compiler without having to modify 8454 its source code. A new option -fplugin=file.so tells GCC to load 8455 the shared object file.so and execute it as part of the compiler. 8456 The internal documentation describes the details on how plugins can 8457 interact with the compiler. 8458 8459 Installation changes 8460 8461 * The move to newer autotools changed default installation 8462 directories and switches to control them: The --with-datarootdir, 8463 --with-docdir, --with-pdfdir, and --with-htmldir switches are not 8464 used any more. Instead, you can now use --datarootdir, --docdir, 8465 --htmldir, and --pdfdir. The default installation directories have 8466 changed as follows according to the GNU Coding Standards: 8467 8468 datarootdir read-only architecture-independent data root [PREFIX/share] 8469 localedir locale-specific message catalogs [DATAROOTDIR/locale] 8470 docdir documentation root [DATAROOTDIR/doc/PACKAGE] 8471 htmldir html documentation [DOCDIR] 8472 dvidir dvi documentation [DOCDIR] 8473 pdfdir pdf documentation [DOCDIR] 8474 psdir ps documentation [DOCDIR] 8475 The following variables have new default values: 8476 8477 datadir read-only architecture-independent data [DATAROOTDIR] 8478 infodir info documentation [DATAROOTDIR/info] 8479 mandir man documentation [DATAROOTDIR/man] 8480 8481 GCC 4.5.1 8482 8483 This is the [22]list of problem reports (PRs) from GCC's bug tracking 8484 system that are known to be fixed in the 4.5.1 release. This list might 8485 not be complete (that is, it is possible that some PRs that have been 8486 fixed are not listed here). 8487 8488 All languages 8489 8490 * GCC's new link-time optimizer ([23]-flto) now also works on a few 8491 non-ELF targets: 8492 + Cygwin (*-cygwin*) 8493 + MinGW (*-mingw*) 8494 + Darwin on x86-64 (x86_64-apple-darwin*) 8495 LTO is not enabled by default for these targets. To enable LTO, you 8496 should configure with the --enable-lto option. 8497 8498 GCC 4.5.2 8499 8500 This is the [24]list of problem reports (PRs) from GCC's bug tracking 8501 system that are known to be fixed in the 4.5.2 release. This list might 8502 not be complete (that is, it is possible that some PRs that have been 8503 fixed are not listed here). 8504 8505 GCC 4.5.3 8506 8507 This is the [25]list of problem reports (PRs) from GCC's bug tracking 8508 system that are known to be fixed in the 4.5.3 release. This list might 8509 not be complete (that is, it is possible that some PRs that have been 8510 fixed are not listed here). 8511 8512 On the PowerPC compiler, the Altivec builtin functions vec_ld and 8513 vec_st have been modified to generate the Altivec memory instructions 8514 LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5 8515 release, these builtin functions were changed to generate VSX memory 8516 reference instructions instead of Altivec memory instructions, but 8517 there are differences between the two instructions. If the VSX 8518 instruction set is available, you can now use the new builtin functions 8519 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 8520 instructions. 8521 8522 GCC 4.5.4 8523 8524 This is the [26]list of problem reports (PRs) from GCC's bug tracking 8525 system that are known to be fixed in the 4.5.4 release. This list might 8526 not be complete (that is, it is possible that some PRs that have been 8527 fixed are not listed here). 8528 8529 8530 For questions related to the use of GCC, please consult these web 8531 pages and the [27]GCC manuals. If that fails, the 8532 [28]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8533 web pages and the development of GCC are welcome on our developer 8534 list at [29]gcc (a] gcc.gnu.org. All of [30]our lists have public 8535 archives. 8536 8537 Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and 8538 distribution of this entire article is permitted in any medium, 8539 provided this notice is preserved. 8540 8541 These pages are [32]maintained by the GCC team. Last modified 8542 2018-09-30[33]. 8543 8544 References 8545 8546 1. http://www.multiprecision.org/mpc/ 8547 2. https://gcc.gnu.org/install/prerequisites.html 8548 3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html 8549 4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted 8550 5. http://gcc.gnu.org/gcc-4.5/changes.html#x86 8551 6. https://gcc.gnu.org/PR30789 8552 7. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 8553 8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802 8554 9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800 8555 10. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html 8556 11. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html 8557 12. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757 8558 13. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176 8559 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html 8560 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733 8561 16. https://sourceware.org/gdb/wiki/STLSupport 8562 17. http://gcc.gnu.org/gcc-4.4/changes.html 8563 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 8564 19. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html 8565 20. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html 8566 21. https://gcc.gnu.org/wiki/OOP 8567 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1 8568 23. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 8569 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2 8570 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3 8571 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4 8572 27. https://gcc.gnu.org/onlinedocs/ 8573 28. mailto:gcc-help (a] gcc.gnu.org 8574 29. mailto:gcc (a] gcc.gnu.org 8575 30. https://gcc.gnu.org/lists.html 8576 31. https://www.fsf.org/ 8577 32. https://gcc.gnu.org/about.html 8578 33. http://validator.w3.org/check/referer 8579 ====================================================================== 8580 http://gcc.gnu.org/gcc-4.4/index.html 8581 GCC 4.4 Release Series 8582 8583 March 13, 2012 8584 8585 The [1]GNU project and the GCC developers are pleased to announce the 8586 release of GCC 4.4.7. 8587 8588 This release is a bug-fix release, containing fixes for regressions in 8589 GCC 4.4.6 relative to previous releases of GCC. 8590 8591 This release series is no longer maintained. 8592 8593 Release History 8594 8595 GCC 4.4.7 8596 March 13, 2012 ([2]changes) 8597 8598 GCC 4.4.6 8599 April 16, 2011 ([3]changes) 8600 8601 GCC 4.4.5 8602 October 1, 2010 ([4]changes) 8603 8604 GCC 4.4.4 8605 April 29, 2010 ([5]changes) 8606 8607 GCC 4.4.3 8608 January 21, 2010 ([6]changes) 8609 8610 GCC 4.4.2 8611 October 15, 2009 ([7]changes) 8612 8613 GCC 4.4.1 8614 July 22, 2009 ([8]changes) 8615 8616 GCC 4.4.0 8617 April 21, 2009 ([9]changes) 8618 8619 References and Acknowledgements 8620 8621 GCC used to stand for the GNU C Compiler, but since the compiler 8622 supports several other languages aside from C, it now stands for the 8623 GNU Compiler Collection. 8624 8625 A list of [10]successful builds is updated as new information becomes 8626 available. 8627 8628 The GCC developers would like to thank the numerous people that have 8629 contributed new features, improvements, bug fixes, and other changes as 8630 well as test results to GCC. This [11]amazing group of volunteers is 8631 what makes GCC successful. 8632 8633 For additional information about GCC please refer to the [12]GCC 8634 project web site or contact the [13]GCC development mailing list. 8635 8636 To obtain GCC please use [14]our mirror sites or [15]our SVN server. 8637 8638 8639 For questions related to the use of GCC, please consult these web 8640 pages and the [16]GCC manuals. If that fails, the 8641 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8642 web pages and the development of GCC are welcome on our developer 8643 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 8644 archives. 8645 8646 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 8647 distribution of this entire article is permitted in any medium, 8648 provided this notice is preserved. 8649 8650 These pages are [21]maintained by the GCC team. Last modified 8651 2018-09-30[22]. 8652 8653 References 8654 8655 1. http://www.gnu.org/ 8656 2. http://gcc.gnu.org/gcc-4.4/changes.html 8657 3. http://gcc.gnu.org/gcc-4.4/changes.html 8658 4. http://gcc.gnu.org/gcc-4.4/changes.html 8659 5. http://gcc.gnu.org/gcc-4.4/changes.html 8660 6. http://gcc.gnu.org/gcc-4.4/changes.html 8661 7. http://gcc.gnu.org/gcc-4.4/changes.html 8662 8. http://gcc.gnu.org/gcc-4.4/changes.html 8663 9. http://gcc.gnu.org/gcc-4.4/changes.html 8664 10. http://gcc.gnu.org/gcc-4.4/buildstat.html 8665 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8666 12. http://gcc.gnu.org/index.html 8667 13. mailto:gcc (a] gcc.gnu.org 8668 14. http://gcc.gnu.org/mirrors.html 8669 15. http://gcc.gnu.org/svn.html 8670 16. https://gcc.gnu.org/onlinedocs/ 8671 17. mailto:gcc-help (a] gcc.gnu.org 8672 18. mailto:gcc (a] gcc.gnu.org 8673 19. https://gcc.gnu.org/lists.html 8674 20. https://www.fsf.org/ 8675 21. https://gcc.gnu.org/about.html 8676 22. http://validator.w3.org/check/referer 8677 ====================================================================== 8678 http://gcc.gnu.org/gcc-4.4/changes.html 8679 GCC 4.4 Release Series 8680 Changes, New Features, and Fixes 8681 8682 The latest release in the 4.4 release series is [1]GCC 4.4.7. 8683 8684 Caveats 8685 8686 * __builtin_stdarg_start has been completely removed from GCC. 8687 Support for <varargs.h> had been deprecated since GCC 4.0. Use 8688 __builtin_va_start as a replacement. 8689 * Some of the errors issued by the C++ front end that could be 8690 downgraded to warnings in previous releases by using -fpermissive 8691 are now warnings by default. They can be converted into errors by 8692 using -pedantic-errors. 8693 * Use of the cpp assertion extension will now emit a warning when 8694 -Wdeprecated or -pedantic is used. This extension has been 8695 deprecated for many years, but never warned about. 8696 * Packed bit-fields of type char were not properly bit-packed on many 8697 targets prior to GCC 4.4. On these targets, the fix in GCC 4.4 8698 causes an ABI change. For example there is no longer a 4-bit 8699 padding between field a and b in this structure: 8700 struct foo 8701 { 8702 char a:4; 8703 char b:8; 8704 } __attribute__ ((packed)); 8705 There is a new warning to help identify fields that are affected: 8706 foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4 8707 The warning can be disabled with -Wno-packed-bitfield-compat. 8708 * On ARM EABI targets, the C++ mangling of the va_list type has been 8709 changed to conform to the current revision of the EABI. This does 8710 not affect the libstdc++ library included with GCC. 8711 * The SCOUNT and POS bits of the MIPS DSP control register are now 8712 treated as global. Previous versions of GCC treated these fields as 8713 call-clobbered instead. 8714 * The MIPS port no longer recognizes the h asm constraint. It was 8715 necessary to remove this constraint in order to avoid generating 8716 unpredictable code sequences. 8717 One of the main uses of the h constraint was to extract the high 8718 part of a multiplication on 64-bit targets. For example: 8719 asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y)); 8720 You can now achieve the same effect using 128-bit types: 8721 typedef unsigned int uint128_t __attribute__((mode(TI))); 8722 result = ((uint128_t) x * y) >> 64; 8723 The second sequence is better in many ways. For example, if x and y 8724 are constants, the compiler can perform the multiplication at 8725 compile time. If x and y are not constants, the compiler can 8726 schedule the runtime multiplication better than it can schedule an 8727 asm statement. 8728 * Support for a number of older systems and recently unmaintained or 8729 untested target ports of GCC has been declared obsolete in GCC 4.4. 8730 Unless there is activity to revive them, the next release of GCC 8731 will have their sources permanently removed. 8732 The following ports for individual systems on particular 8733 architectures have been obsoleted: 8734 + Generic a.out on IA32 and m68k (i[34567]86-*-aout*, 8735 m68k-*-aout*) 8736 + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*, 8737 armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*, 8738 sh-*-*). This does not affect other more specific targets 8739 using the COFF object format on those architectures, or the 8740 more specific H8300 and SH targets (h8300-*-rtems*, 8741 h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*, 8742 sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks). 8743 + 2BSD on PDP-11 (pdp11-*-bsd) 8744 + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*, 8745 powerpc-ibm-aix4.[12]*) 8746 + Tuning support for Itanium1 (Merced) variants. Note that code 8747 tuned for Itanium2 should also run correctly on Itanium1. 8748 * The protoize and unprotoize utilities have been obsoleted and will 8749 be removed in GCC 4.5. These utilities have not been installed by 8750 default since GCC 3.0. 8751 * Support has been removed for all the [2]configurations obsoleted in 8752 GCC 4.3. 8753 * Unknown -Wno-* options are now silently ignored by GCC if no other 8754 diagnostics are issued. If other diagnostics are issued, then GCC 8755 warns about the unknown options. 8756 * More information on porting to GCC 4.4 from previous versions of 8757 GCC can be found in the [3]porting guide for this release. 8758 8759 General Optimizer Improvements 8760 8761 * A new command-line switch -findirect-inlining has been added. When 8762 turned on it allows the inliner to also inline indirect calls that 8763 are discovered to have known targets at compile time thanks to 8764 previous inlining. 8765 * A new command-line switch -ftree-switch-conversion has been added. 8766 This new pass turns simple initializations of scalar variables in 8767 switch statements into initializations from a static array, given 8768 that all the values are known at compile time and the ratio between 8769 the new array size and the original switch branches does not exceed 8770 the parameter --param switch-conversion-max-branch-ratio (default 8771 is eight). 8772 * A new command-line switch -ftree-builtin-call-dce has been added. 8773 This optimization eliminates unnecessary calls to certain builtin 8774 functions when the return value is not used, in cases where the 8775 calls can not be eliminated entirely because the function may set 8776 errno. This optimization is on by default at -O2 and above. 8777 * A new command-line switch -fconserve-stack directs the compiler to 8778 minimize stack usage even if it makes the generated code slower. 8779 This affects inlining decisions. 8780 * When the assembler supports it, the compiler will now emit unwind 8781 information using assembler .cfi directives. This makes it possible 8782 to use such directives in inline assembler code. The new option 8783 -fno-dwarf2-cfi-asm directs the compiler to not use .cfi 8784 directives. 8785 * The [4]Graphite branch has been merged. This merge has brought in a 8786 new framework for loop optimizations based on a polyhedral 8787 intermediate representation. These optimizations apply to all the 8788 languages supported by GCC. The following new code transformations 8789 are available in GCC 4.4: 8790 + -floop-interchange performs loop interchange transformations 8791 on loops. Interchanging two nested loops switches the inner 8792 and outer loops. For example, given a loop like: 8793 DO J = 1, M 8794 DO I = 1, N 8795 A(J, I) = A(J, I) * C 8796 ENDDO 8797 ENDDO 8798 8799 loop interchange will transform the loop as if the user had 8800 written: 8801 DO I = 1, N 8802 DO J = 1, M 8803 A(J, I) = A(J, I) * C 8804 ENDDO 8805 ENDDO 8806 8807 which can be beneficial when N is larger than the caches, 8808 because in Fortran, the elements of an array are stored in 8809 memory contiguously by column, and the original loop iterates 8810 over rows, potentially creating at each access a cache miss. 8811 + -floop-strip-mine performs loop strip mining transformations 8812 on loops. Strip mining splits a loop into two nested loops. 8813 The outer loop has strides equal to the strip size and the 8814 inner loop has strides of the original loop within a strip. 8815 For example, given a loop like: 8816 DO I = 1, N 8817 A(I) = A(I) + C 8818 ENDDO 8819 8820 loop strip mining will transform the loop as if the user had 8821 written: 8822 DO II = 1, N, 4 8823 DO I = II, min (II + 3, N) 8824 A(I) = A(I) + C 8825 ENDDO 8826 ENDDO 8827 8828 + -floop-block performs loop blocking transformations on loops. 8829 Blocking strip mines each loop in the loop nest such that the 8830 memory accesses of the element loops fit inside caches. For 8831 example, given a loop like: 8832 DO I = 1, N 8833 DO J = 1, M 8834 A(J, I) = B(I) + C(J) 8835 ENDDO 8836 ENDDO 8837 8838 loop blocking will transform the loop as if the user had 8839 written: 8840 DO II = 1, N, 64 8841 DO JJ = 1, M, 64 8842 DO I = II, min (II + 63, N) 8843 DO J = JJ, min (JJ + 63, M) 8844 A(J, I) = B(I) + C(J) 8845 ENDDO 8846 ENDDO 8847 ENDDO 8848 ENDDO 8849 8850 which can be beneficial when M is larger than the caches, 8851 because the innermost loop will iterate over a smaller amount 8852 of data that can be kept in the caches. 8853 * A new register allocator has replaced the old one. It is called 8854 integrated register allocator (IRA) because coalescing, register 8855 live range splitting, and hard register preferencing are done 8856 on-the-fly during coloring. It also has better integration with the 8857 reload pass. IRA is a regional register allocator which uses modern 8858 Chaitin-Briggs coloring instead of Chow's priority coloring used in 8859 the old register allocator. More info about IRA internals and 8860 options can be found in the GCC manuals. 8861 * A new instruction scheduler and software pipeliner, based on the 8862 selective scheduling approach, has been added. The new pass 8863 performs instruction unification, register renaming, substitution 8864 through register copies, and speculation during scheduling. The 8865 software pipeliner is able to pipeline non-countable loops. The new 8866 pass is targeted at scheduling-eager in-order platforms. In GCC 4.4 8867 it is available for the Intel Itanium platform working by default 8868 as the second scheduling pass (after register allocation) at the 8869 -O3 optimization level. 8870 * When using -fprofile-generate with a multi-threaded program, the 8871 profile counts may be slightly wrong due to race conditions. The 8872 new -fprofile-correction option directs the compiler to apply 8873 heuristics to smooth out the inconsistencies. By default the 8874 compiler will give an error message when it finds an inconsistent 8875 profile. 8876 * The new -fprofile-dir=PATH option permits setting the directory 8877 where profile data files are stored when using -fprofile-generate 8878 and friends, and the directory used when reading profile data files 8879 using -fprofile-use and friends. 8880 8881 New warning options 8882 8883 * The new -Wframe-larger-than=NUMBER option directs GCC to emit a 8884 warning if any stack frame is larger than NUMBER bytes. This may be 8885 used to help ensure that code fits within a limited amount of stack 8886 space. 8887 * The command-line option -Wlarger-than-N is now written as 8888 -Wlarger-than=N and the old form is deprecated. 8889 * The new -Wno-mudflap option disables warnings about constructs 8890 which can not be instrumented when using -fmudflap. 8891 8892 New Languages and Language specific improvements 8893 8894 * Version 3.0 of the OpenMP specification is now supported for the C, 8895 C++, and Fortran compilers. 8896 * New character data types, per [5]TR 19769: New character types in 8897 C, are now supported for the C compiler in -std=gnu99 mode, as 8898 __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in 8899 -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too. 8900 8901 C family 8902 8903 * A new optimize attribute was added to allow programmers to change 8904 the optimization level and particular optimization options for an 8905 individual function. You can also change the optimization options 8906 via the GCC optimize pragma for functions defined after the pragma. 8907 The GCC push_options pragma and the GCC pop_options pragma allow 8908 you temporarily save and restore the options used. The GCC 8909 reset_options pragma restores the options to what was specified on 8910 the command line. 8911 * Uninitialized warnings do not require enabling optimization 8912 anymore, that is, -Wuninitialized can be used together with -O0. 8913 Nonetheless, the warnings given by -Wuninitialized will probably be 8914 more accurate if optimization is enabled. 8915 * -Wparentheses now warns about expressions such as (!x | y) and (!x 8916 & y). Using explicit parentheses, such as in ((!x) | y), silences 8917 this warning. 8918 * -Wsequence-point now warns within if, while,do while and for 8919 conditions, and within for begin/end expressions. 8920 * A new option -dU is available to dump definitions of preprocessor 8921 macros that are tested or expanded. 8922 8923 C++ 8924 8925 * [6]Improved experimental support for the upcoming ISO C++ standard, 8926 C++0x. Including support for auto, inline namespaces, generalized 8927 initializer lists, defaulted and deleted functions, new character 8928 types, and scoped enums. 8929 * Those errors that may be downgraded to warnings to build legacy 8930 code now mention -fpermissive when -fdiagnostics-show-option is 8931 enabled. 8932 * -Wconversion now warns if the result of a static_cast to enumeral 8933 type is unspecified because the value is outside the range of the 8934 enumeral type. 8935 * -Wuninitialized now warns if a non-static reference or non-static 8936 const member appears in a class without constructors. 8937 * G++ now properly implements value-initialization, so objects with 8938 an initializer of () and an implicitly defined default constructor 8939 will be zero-initialized before the default constructor is called. 8940 8941 Runtime Library (libstdc++) 8942 8943 * Improved experimental support for the upcoming ISO C++ standard, 8944 C++0x, including: 8945 + Support for <chrono>, <condition_variable>, <cstdatomic>, 8946 <forward_list>, <initializer_list>, <mutex>, <ratio>, 8947 <system_error>, and <thread>. 8948 + unique_ptr, <algorithm> additions, exception propagation, and 8949 support for the new character types in <string> and <limits>. 8950 + Existing facilities now exploit initializer lists, defaulted 8951 and deleted functions, and the newly implemented core C++0x 8952 features. 8953 + Some standard containers are more efficient together with 8954 stateful allocators, i.e., no allocator is constructed on the 8955 fly at element construction time. 8956 * Experimental support for non-standard pointer types in containers. 8957 * The long standing libstdc++/30928 has been fixed for targets 8958 running glibc 2.10 or later. 8959 * As usual, many small and larger bug fixes, in particular quite a 8960 few corner cases in <locale>. 8961 8962 Fortran 8963 8964 * GNU Fortran now employs libcpp directly instead of using cc1 as an 8965 external preprocessor. The [7]-cpp option was added to allow manual 8966 invocation of the preprocessor without relying on filename 8967 extensions. 8968 * The [8]-Warray-temporaries option warns about array temporaries 8969 generated by the compiler, as an aid to optimization. 8970 * The [9]-fcheck-array-temporaries option has been added, printing a 8971 notification at run time, when an array temporary had to be created 8972 for an function argument. Contrary to -Warray-temporaries the 8973 warning is only printed if the array is noncontiguous. 8974 * Improved generation of DWARF debugging symbols 8975 * If using an intrinsic not part of the selected standard (via -std= 8976 and -fall-intrinsics) gfortran will now treat it as if this 8977 procedure were declared EXTERNAL and try to link to a user-supplied 8978 procedure. -Wintrinsics-std will warn whenever this happens. The 8979 now-useless option -Wnonstd-intrinsic was removed. 8980 * The flag -falign-commons has been added to control the alignment of 8981 variables in COMMON blocks, which is enabled by default in line 8982 with previous GCC version. Using -fno-align-commons one can force 8983 commons to be contiguous in memory as required by the Fortran 8984 standard, however, this slows down the memory access. The option 8985 -Walign-commons, which is enabled by default, warns when padding 8986 bytes were added for alignment. The proper solution is to sort the 8987 common objects by decreasing storage size, which avoids the 8988 alignment problems. 8989 * Fortran 2003 support has been extended: 8990 + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is 8991 now supported (except internal reads from/writes to wide 8992 strings). [10]-fbackslash now supports also \unnnn and 8993 \Unnnnnnnn to enter Unicode characters. 8994 + Asynchronous I/O (implemented as synchronous I/O) and the 8995 decimal=, size=, sign=, pad=, blank=, and delim= specifiers 8996 are now supported in I/O statements. 8997 + Support for Fortran 2003 structure constructors and for array 8998 constructor with typespec has been added. 8999 + Procedure Pointers (but not yet as component in derived types 9000 and as function results) are now supported. 9001 + Abstract types, type extension, and type-bound procedures 9002 (both PROCEDURE and GENERIC but not as operators). Note: As 9003 CLASS/polymorphyic types are not implemented, type-bound 9004 procedures with PASS accept as non-standard extension TYPE 9005 arguments. 9006 * Fortran 2008 support has been added: 9007 + The -std=f2008 option and support for the file extensions 9008 .f2008 and .F2008 has been added. 9009 + The g0 format descriptor is now supported. 9010 + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH, 9011 ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED 9012 are now available (some of them existed as GNU extension 9013 before). Note: The hyperbolic functions are not yet supporting 9014 complex arguments and the three- argument version of BESSEL_*N 9015 is not available. 9016 + The bit intrinsics LEADZ and TRAILZ have been added. 9017 9018 Java (GCJ) 9019 9020 Ada 9021 9022 * The Ada runtime now supports multilibs on many platforms including 9023 x86_64, SPARC and PowerPC. Their build is enabled by default. 9024 9025 New Targets and Target Specific Improvements 9026 9027 ARM 9028 9029 * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and 9030 Cortex-R4F processors and has many other improvements to 9031 optimization for ARM processors. 9032 * GCC now supports the VFPv3 variant with 16 double-precision 9033 registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been 9034 renamed to -mfpu=vfpv3. 9035 * GCC now supports the -mfix-cortex-m3-ldrd option to work around an 9036 erratum on Cortex-M3 processors. 9037 * GCC now supports the __sync_* atomic operations for ARM EABI 9038 GNU/Linux. 9039 * The section anchors optimization is now enabled by default when 9040 optimizing for ARM. 9041 * GCC now uses a new EABI-compatible profiling interface for EABI 9042 targets. This requires a function __gnu_mcount_nc, which is 9043 provided by GNU libc versions 2.8 and later. 9044 9045 AVR 9046 9047 * The -mno-tablejump option has been deprecated because it has the 9048 same effect as the -fno-jump-tables option. 9049 * Added support for these new AVR devices: 9050 + ATA6289 9051 + ATtiny13A 9052 + ATtiny87 9053 + ATtiny167 9054 + ATtiny327 9055 + ATmega8C1 9056 + ATmega16C1 9057 + ATmega32C1 9058 + ATmega8M1 9059 + ATmega16M1 9060 + ATmega32M1 9061 + ATmega32U4 9062 + ATmega16HVB 9063 + ATmega4HVD 9064 + ATmega8HVD 9065 + ATmega64C1 9066 + ATmega64M1 9067 + ATmega16U4 9068 + ATmega32U6 9069 + ATmega128RFA1 9070 + AT90PWM81 9071 + AT90SCR100 9072 + M3000F 9073 + M3000S 9074 + M3001B 9075 9076 IA-32/x86-64 9077 9078 * Support for Intel AES built-in functions and code generation is 9079 available via -maes. 9080 * Support for Intel PCLMUL built-in function and code generation is 9081 available via -mpclmul. 9082 * Support for Intel AVX built-in functions and code generation is 9083 available via -mavx. 9084 * Automatically align the stack for local variables with alignment 9085 requirement. 9086 * GCC can now utilize the SVML library for vectorizing calls to a set 9087 of C99 functions if -mveclibabi=svml is specified and you link to 9088 an SVML ABI compatible library. 9089 * On x86-64, the ABI has been changed in the following cases to 9090 conform to the x86-64 ABI: 9091 + Passing/returning structures with flexible array member: 9092 struct foo 9093 { 9094 int i; 9095 int flex[]; 9096 }; 9097 + Passing/returning structures with complex float member: 9098 struct foo 9099 { 9100 int i; 9101 __complex__ float f; 9102 }; 9103 + Passing/returning unions with long double member: 9104 union foo 9105 { 9106 int x; 9107 long double ld; 9108 }; 9109 Code built with previous versions of GCC that uses any of these is 9110 not compatible with code built with GCC 4.4.0 or later. 9111 * A new target attribute was added to allow programmers to change the 9112 target options like -msse2 or -march=k8 for an individual function. 9113 You can also change the target options via the GCC target pragma 9114 for functions defined after the pragma. 9115 * GCC can now be configured with options --with-arch-32, 9116 --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and 9117 --with-tune-64 to control the default optimization separately for 9118 32-bit and 64-bit modes. 9119 9120 IA-32/IA64 9121 9122 * Support for __float128 (TFmode) IEEE quad type and corresponding 9123 TCmode IEEE complex quad type is available via the soft-fp library 9124 on IA-32/IA64 targets. This includes basic arithmetic operations 9125 (addition, subtraction, negation, multiplication and division) on 9126 __float128 real and TCmode complex values, the full set of IEEE 9127 comparisons between __float128 values, conversions to and from 9128 float, double and long double floating point types, as well as 9129 conversions to and from signed or unsigned integer, signed or 9130 unsigned long integer and signed or unsigned quad (TImode, IA64 9131 only) integer types. Additionally, all operations generate the full 9132 set of IEEE exceptions and support the full set of IEEE rounding 9133 modes. 9134 9135 M68K/ColdFire 9136 9137 * GCC now supports instruction scheduling for ColdFire V1, V3 and V4 9138 processors. (Scheduling support for ColdFire V2 processors was 9139 added in GCC 4.3.) 9140 * GCC now supports the -mxgot option to support programs requiring 9141 many GOT entries on ColdFire. 9142 * The m68k-*-linux-gnu target now builds multilibs by default. 9143 9144 MIPS 9145 9146 * MIPS Technologies have extended the original MIPS SVR4 ABI to 9147 include support for procedure linkage tables (PLTs) and copy 9148 relocations. These extensions allow GNU/Linux executables to use a 9149 significantly more efficient code model than the one defined by the 9150 original ABI. 9151 GCC support for this code model is available via a new command-line 9152 option, -mplt. There is also a new configure-time option, 9153 --with-mips-plt, to make -mplt the default. 9154 The new code model requires support from the assembler, the linker, 9155 and the runtime C library. This support is available in binutils 9156 2.19 and GLIBC 2.9. 9157 * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables 9158 and 32-bit GNU/Linux shared libraries. This feature requires GNU 9159 binutils 2.19 or above. 9160 * Support for RMI's XLR processor is now available through the 9161 -march=xlr and -mtune=xlr options. 9162 * 64-bit targets can now perform 128-bit multiplications inline, 9163 instead of relying on a libgcc function. 9164 * Native GNU/Linux toolchains now support -march=native and 9165 -mtune=native, which select the host processor. 9166 * GCC now supports the R10K, R12K, R14K and R16K processors. The 9167 canonical -march= and -mtune= names for these processors are 9168 r10000, r12000, r14000 and r16000 respectively. 9169 * GCC can now work around the side effects of speculative execution 9170 on R10K processors. Please see the documentation of the 9171 -mr10k-cache-barrier option for details. 9172 * Support for the MIPS64 Release 2 instruction set has been added. 9173 The option -march=mips64r2 enables generation of these 9174 instructions. 9175 * GCC now supports Cavium Networks' Octeon processor. This support is 9176 available through the -march=octeon and -mtune=octeon options. 9177 * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The 9178 canonical -march= and -mtune= names for these processors are 9179 loongson2e and loongson2f. 9180 9181 picochip 9182 9183 Picochip is a 16-bit processor. A typical picoChip contains over 250 9184 small cores, each with small amounts of memory. There are three 9185 processor variants (STAN, MEM and CTRL) with different instruction sets 9186 and memory configurations and they can be chosen using the -mae option. 9187 9188 This port is intended to be a "C" only port. 9189 9190 Power Architecture and PowerPC 9191 9192 * GCC now supports the e300c2, e300c3 and e500mc processors. 9193 * GCC now supports Xilinx processors with a single-precision FPU. 9194 * Decimal floating point is now supported for e500 processors. 9195 9196 S/390, zSeries and System z9/z10 9197 9198 * Support for the IBM System z10 EC/BC processor has been added. When 9199 using the -march=z10 option, the compiler will generate code making 9200 use of instructions provided by the General-Instruction-Extension 9201 Facility and the Execute-Extension Facility. 9202 9203 VxWorks 9204 9205 * GCC now supports the thread-local storage mechanism used on 9206 VxWorks. 9207 9208 Xtensa 9209 9210 * GCC now supports thread-local storage (TLS) for Xtensa processor 9211 configurations that include the Thread Pointer option. TLS also 9212 requires support from the assembler and linker; this support is 9213 provided in the GNU binutils beginning with version 2.19. 9214 9215 Documentation improvements 9216 9217 Other significant improvements 9218 9219 GCC 4.4.1 9220 9221 This is the [11]list of problem reports (PRs) from GCC's bug tracking 9222 system that are known to be fixed in the 4.4.1 release. This list might 9223 not be complete (that is, it is possible that some PRs that have been 9224 fixed are not listed here). 9225 9226 GCC 4.4.2 9227 9228 This is the [12]list of problem reports (PRs) from GCC's bug tracking 9229 system that are known to be fixed in the 4.4.2 release. This list might 9230 not be complete (that is, it is possible that some PRs that have been 9231 fixed are not listed here). 9232 9233 GCC 4.4.3 9234 9235 This is the [13]list of problem reports (PRs) from GCC's bug tracking 9236 system that are known to be fixed in the 4.4.3 release. This list might 9237 not be complete (that is, it is possible that some PRs that have been 9238 fixed are not listed here). 9239 9240 GCC 4.4.4 9241 9242 This is the [14]list of problem reports (PRs) from GCC's bug tracking 9243 system that are known to be fixed in the 4.4.4 release. This list might 9244 not be complete (that is, it is possible that some PRs that have been 9245 fixed are not listed here). 9246 9247 GCC 4.4.5 9248 9249 This is the [15]list of problem reports (PRs) from GCC's bug tracking 9250 system that are known to be fixed in the 4.4.5 release. This list might 9251 not be complete (that is, it is possible that some PRs that have been 9252 fixed are not listed here). 9253 9254 GCC 4.4.6 9255 9256 This is the [16]list of problem reports (PRs) from GCC's bug tracking 9257 system that are known to be fixed in the 4.4.6 release. This list might 9258 not be complete (that is, it is possible that some PRs that have been 9259 fixed are not listed here). 9260 9261 GCC 4.4.7 9262 9263 This is the [17]list of problem reports (PRs) from GCC's bug tracking 9264 system that are known to be fixed in the 4.4.7 release. This list might 9265 not be complete (that is, it is possible that some PRs that have been 9266 fixed are not listed here). 9267 9268 9269 For questions related to the use of GCC, please consult these web 9270 pages and the [18]GCC manuals. If that fails, the 9271 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9272 web pages and the development of GCC are welcome on our developer 9273 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 9274 archives. 9275 9276 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 9277 distribution of this entire article is permitted in any medium, 9278 provided this notice is preserved. 9279 9280 These pages are [23]maintained by the GCC team. Last modified 9281 2018-09-30[24]. 9282 9283 References 9284 9285 1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7 9286 2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted 9287 3. http://gcc.gnu.org/gcc-4.4/porting_to.html 9288 4. https://gcc.gnu.org/wiki/Graphite 9289 5. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf 9290 6. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html 9291 7. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html 9292 8. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125 9293 9. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221 9294 10. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34 9295 11. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1 9296 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2 9297 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3 9298 14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4 9299 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5 9300 16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6 9301 17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7 9302 18. https://gcc.gnu.org/onlinedocs/ 9303 19. mailto:gcc-help (a] gcc.gnu.org 9304 20. mailto:gcc (a] gcc.gnu.org 9305 21. https://gcc.gnu.org/lists.html 9306 22. https://www.fsf.org/ 9307 23. https://gcc.gnu.org/about.html 9308 24. http://validator.w3.org/check/referer 9309 ====================================================================== 9310 http://gcc.gnu.org/gcc-4.3/index.html 9311 GCC 4.3 Release Series 9312 9313 Jun 27, 2011 9314 9315 The [1]GNU project and the GCC developers are pleased to announce the 9316 release of GCC 4.3.6. 9317 9318 This release is a bug-fix release, containing fixes for regressions in 9319 GCC 4.3.5 relative to previous releases of GCC. 9320 9321 This release series is no longer maintained. 9322 9323 Release History 9324 9325 GCC 4.3.6 9326 Jun 27, 2011 ([2]changes) 9327 9328 GCC 4.3.5 9329 May 22, 2010 ([3]changes) 9330 9331 GCC 4.3.4 9332 August 4, 2009 ([4]changes) 9333 9334 GCC 4.3.3 9335 January 24, 2009 ([5]changes) 9336 9337 GCC 4.3.2 9338 August 27, 2008 ([6]changes) 9339 9340 GCC 4.3.1 9341 June 6, 2008 ([7]changes) 9342 9343 GCC 4.3.0 9344 March 5, 2008 ([8]changes) 9345 9346 References and Acknowledgements 9347 9348 GCC used to stand for the GNU C Compiler, but since the compiler 9349 supports several other languages aside from C, it now stands for the 9350 GNU Compiler Collection. 9351 9352 A list of [9]successful builds is updated as new information becomes 9353 available. 9354 9355 The GCC developers would like to thank the numerous people that have 9356 contributed new features, improvements, bug fixes, and other changes as 9357 well as test results to GCC. This [10]amazing group of volunteers is 9358 what makes GCC successful. 9359 9360 For additional information about GCC please refer to the [11]GCC 9361 project web site or contact the [12]GCC development mailing list. 9362 9363 To obtain GCC please use [13]our mirror sites or [14]our SVN server. 9364 9365 9366 For questions related to the use of GCC, please consult these web 9367 pages and the [15]GCC manuals. If that fails, the 9368 [16]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9369 web pages and the development of GCC are welcome on our developer 9370 list at [17]gcc (a] gcc.gnu.org. All of [18]our lists have public 9371 archives. 9372 9373 Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and 9374 distribution of this entire article is permitted in any medium, 9375 provided this notice is preserved. 9376 9377 These pages are [20]maintained by the GCC team. Last modified 9378 2018-09-30[21]. 9379 9380 References 9381 9382 1. http://www.gnu.org/ 9383 2. http://gcc.gnu.org/gcc-4.3/changes.html 9384 3. http://gcc.gnu.org/gcc-4.3/changes.html 9385 4. http://gcc.gnu.org/gcc-4.3/changes.html 9386 5. http://gcc.gnu.org/gcc-4.3/changes.html 9387 6. http://gcc.gnu.org/gcc-4.3/changes.html 9388 7. http://gcc.gnu.org/gcc-4.3/changes.html 9389 8. http://gcc.gnu.org/gcc-4.3/changes.html 9390 9. http://gcc.gnu.org/gcc-4.3/buildstat.html 9391 10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 9392 11. http://gcc.gnu.org/index.html 9393 12. mailto:gcc (a] gcc.gnu.org 9394 13. http://gcc.gnu.org/mirrors.html 9395 14. http://gcc.gnu.org/svn.html 9396 15. https://gcc.gnu.org/onlinedocs/ 9397 16. mailto:gcc-help (a] gcc.gnu.org 9398 17. mailto:gcc (a] gcc.gnu.org 9399 18. https://gcc.gnu.org/lists.html 9400 19. https://www.fsf.org/ 9401 20. https://gcc.gnu.org/about.html 9402 21. http://validator.w3.org/check/referer 9403 ====================================================================== 9404 http://gcc.gnu.org/gcc-4.3/changes.html 9405 GCC 4.3 Release Series 9406 Changes, New Features, and Fixes 9407 9408 The latest release in the 4.3 release series is [1]GCC 4.3.5. 9409 9410 Caveats 9411 9412 * GCC requires the [2]GMP and [3]MPFR libraries for building all the 9413 various front-end languages it supports. See the [4]prerequisites 9414 page for version requirements. 9415 * ColdFire targets now treat long double as having the same format as 9416 double. In earlier versions of GCC, they used the 68881 long double 9417 format instead. 9418 * The m68k-uclinux target now uses the same calling conventions as 9419 m68k-linux-gnu. You can select the original calling conventions by 9420 configuring for m68k-uclinuxoldabi instead. Note that 9421 m68k-uclinuxoldabi also retains the original 80-bit long double on 9422 ColdFire targets. 9423 * The -fforce-mem option has been removed because it has had no 9424 effect in the last few GCC releases. 9425 * The i386 -msvr3-shlib option has been removed since it is no longer 9426 used. 9427 * Fastcall for i386 has been changed not to pass aggregate arguments 9428 in registers, following Microsoft compilers. 9429 * Support for the AOF assembler has been removed from the ARM back 9430 end; this affects only the targets arm-semi-aof and armel-semi-aof, 9431 which are no longer recognized. We removed these targets without a 9432 deprecation period because we discovered that they have been 9433 unusable since GCC 4.0.0. 9434 * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*) 9435 has been removed. This support had been deprecated since GCC 4.0.0. 9436 * Support for a number of older systems and recently unmaintained or 9437 untested target ports of GCC has been declared obsolete in GCC 4.3. 9438 Unless there is activity to revive them, the next release of GCC 9439 will have their sources permanently removed. 9440 All GCC ports for the following processor architectures have been 9441 declared obsolete: 9442 + Morpho MT (mt-*) 9443 The following aliases for processor architectures have been 9444 declared obsolete. Users should use the indicated generic target 9445 names instead, with compile-time options such as -mcpu or 9446 configure-time options such as --with-cpu to control the 9447 configuration more precisely. 9448 + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-* 9449 instead). 9450 + parisc*-*-* (use hppa*-*-* instead). 9451 + m680[012]0-*-* (use m68k-*-* instead). 9452 All GCC ports for the following operating systems have been 9453 declared obsolete: 9454 + BeOS (*-*-beos*) 9455 + kaOS (*-*-kaos*) 9456 + GNU/Linux using the a.out object format (*-*-linux*aout*) 9457 + GNU/Linux using version 1 of the GNU C Library 9458 (*-*-linux*libc1*) 9459 + Solaris versions before Solaris 7 (*-*-solaris2.[0-6], 9460 *-*-solaris2.[0-6].*) 9461 + Miscellaneous System V (*-*-sysv*) 9462 + WindISS (*-*-windiss*) 9463 Also, those for some individual systems on particular architectures 9464 have been obsoleted: 9465 + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*) 9466 + CRIS with a.out object format (cris-*-aout) 9467 + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*) 9468 + OSF/1 on PA-RISC (hppa1.1-*-osf*) 9469 + PRO on PA-RISC (hppa1.1-*-pro*) 9470 + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*, 9471 i[34567]86-sequent-sysv4*) 9472 + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*) 9473 + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host 9474 was previously [5]removed in 2001, leaving only the support 9475 for UWIN as a target now being deprecated) 9476 + ChorusOS on PowerPC (powerpc-*-chorusos*) 9477 + All VAX configurations apart from NetBSD and OpenBSD 9478 (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*) 9479 * The [6]-Wconversion option has been modified. Its purpose now is to 9480 warn for implicit conversions that may alter a value. This new 9481 behavior is available for both C and C++. Warnings about 9482 conversions between signed and unsigned integers can be disabled by 9483 using -Wno-sign-conversion. In C++, they are disabled by default 9484 unless -Wsign-conversion is explicitly requested. The old behavior 9485 of -Wconversion, that is, warn for prototypes causing a type 9486 conversion that is different from what would happen to the same 9487 argument in the absence of a prototype, has been moved to a new 9488 option -Wtraditional-conversion, which is only available for C. 9489 * The -m386, -m486, -mpentium and -mpentiumpro tuning options have 9490 been removed because they were deprecated for more than 3 GCC major 9491 releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or 9492 -mtune=pentiumpro as a replacement. 9493 * The -funsafe-math-optimizations option now automatically turns on 9494 -fno-trapping-math in addition to -fno-signed-zeros, as it enables 9495 reassociation and thus may introduce or remove traps. 9496 * The -ftree-vectorize option is now on by default under -O3. In 9497 order to generate code for a SIMD extension, it has to be enabled 9498 as well: use -maltivec for PowerPC platforms and -msse/-msse2 for 9499 i?86 and x86_64. 9500 * More information on porting to GCC 4.3 from previous versions of 9501 GCC can be found in the [7]porting guide for this release. 9502 9503 General Optimizer Improvements 9504 9505 * The GCC middle-end has been integrated with the MPFR library. This 9506 allows GCC to evaluate and replace at compile-time calls to 9507 built-in math functions having constant arguments with their 9508 mathematically equivalent results. In making use of MPFR, GCC can 9509 generate correct results regardless of the math library 9510 implementation or floating point precision of the host platform. 9511 This also allows GCC to generate identical results regardless of 9512 whether one compiles in native or cross-compile configurations to a 9513 particular target. The following built-in functions take advantage 9514 of this new capability: acos, acosh, asin, asinh, atan2, atan, 9515 atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1, 9516 fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10, 9517 log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh, 9518 tan, tanh, tgamma, y0, y1 and yn. The float and long double 9519 variants of these functions (e.g. sinf and sinl) are also handled. 9520 The sqrt and cabs functions with constant arguments were already 9521 optimized in prior GCC releases. Now they also use MPFR. 9522 * A new forward propagation pass on RTL was added. The new pass 9523 replaces several slower transformations, resulting in compile-time 9524 improvements as well as better code generation in some cases. 9525 * A new command-line switch -frecord-gcc-switches has been added to 9526 GCC, although it is only enabled for some targets. The switch 9527 causes the command line that was used to invoke the compiler to be 9528 recorded into the object file that is being created. The exact 9529 format of this recording is target and binary file format 9530 dependent, but it usually takes the form of a note section 9531 containing ASCII text. The switch is related to the -fverbose-asm 9532 switch, but that one only records the information in the assembler 9533 output file as comments, so the information never reaches the 9534 object file. 9535 * The inliner heuristic is now aware of stack frame consumption. New 9536 command-line parameters --param large-stack-frame and --param 9537 large-stack-frame-growth can be used to limit stack frame size 9538 growth caused by inlining. 9539 * During feedback directed optimizations, the expected block size the 9540 memcpy, memset and bzero functions operate on is discovered and for 9541 cases of commonly used small sizes, specialized inline code is 9542 generated. 9543 * __builtin_expect no longer requires its argument to be a compile 9544 time constant. 9545 * Interprocedural optimization was reorganized to work on functions 9546 in SSA form. This enables more precise and cheaper dataflow 9547 analysis and makes writing interprocedural optimizations easier. 9548 The following improvements have been implemented on top of this 9549 framework: 9550 + Pre-inline optimization: Selected local optimization passes 9551 are run before the inliner (and other interprocedural passes) 9552 are executed. This significantly improves the accuracy of code 9553 growth estimates used by the inliner and reduces the overall 9554 memory footprint for large compilation units. 9555 + Early inlining (a simple bottom-up inliner pass inlining only 9556 functions whose body is smaller than the expected call 9557 overhead) is now executed with the early optimization passes, 9558 thus inlining already optimized function bodies into an 9559 unoptimized function that is subsequently optimized by early 9560 optimizers. This enables the compiler to quickly eliminate 9561 abstraction penalty in C++ programs. 9562 + Interprocedural constant propagation now operate on SSA form 9563 increasing accuracy of the analysis. 9564 * A new internal representation for GIMPLE statements has been 9565 contributed, resulting in compile-time memory savings. 9566 * The vectorizer was enhanced to support vectorization of outer 9567 loops, intra-iteration parallelism (loop-aware SLP), vectorization 9568 of strided accesses and loops with multiple data-types. Run-time 9569 dependency testing using loop versioning was added. The cost model, 9570 turned on by -fvect-cost-model, was developed. 9571 9572 New Languages and Language specific improvements 9573 9574 * We have added new command-line options 9575 -finstrument-functions-exclude-function-list and 9576 -finstrument-functions-exclude-file-list. They provide more control 9577 over which functions are annotated by the -finstrument-functions 9578 option. 9579 9580 C family 9581 9582 * Implicit conversions between generic vector types are now only 9583 permitted when the two vectors in question have the same number of 9584 elements and compatible element types. (Note that the restriction 9585 involves compatible element types, not implicitly-convertible 9586 element types: thus, a vector type with element type int may not be 9587 implicitly converted to a vector type with element type unsigned 9588 int.) This restriction, which is in line with specifications for 9589 SIMD architectures such as AltiVec, may be relaxed using the flag 9590 -flax-vector-conversions. This flag is intended only as a 9591 compatibility measure and should not be used for new code. 9592 * -Warray-bounds has been added and is now enabled by default for 9593 -Wall . It produces warnings for array subscripts that can be 9594 determined at compile time to be always out of bounds. 9595 -Wno-array-bounds will disable the warning. 9596 * The constructor and destructor function attributes now accept 9597 optional priority arguments which control the order in which the 9598 constructor and destructor functions are run. 9599 * New [8]command-line options -Wtype-limits, -Wold-style-declaration, 9600 -Wmissing-parameter-type, -Wempty-body, -Wclobbered and 9601 -Wignored-qualifiers have been added for finer control of the 9602 diverse warnings enabled by -Wextra. 9603 * A new function attribute alloc_size has been added to mark up 9604 malloc style functions. For constant sized allocations this can be 9605 used to find out the size of the returned pointer using the 9606 __builtin_object_size() function for buffer overflow checking and 9607 similar. This supplements the already built-in malloc and calloc 9608 constant size handling. 9609 * Integer constants written in binary are now supported as a GCC 9610 extension. They consist of a prefix 0b or 0B, followed by a 9611 sequence of 0 and 1 digits. 9612 * A new predefined macro __COUNTER__ has been added. It expands to 9613 sequential integral values starting from 0. In conjunction with the 9614 ## operator, this provides a convenient means to generate unique 9615 identifiers. 9616 * A new command-line option -fdirectives-only has been added. It 9617 enables a special preprocessing mode which improves the performance 9618 of applications like distcc and ccache. 9619 * Fixed-point data types and operators have been added. They are 9620 based on Chapter 4 of the Embedded-C specification (n1169.pdf). 9621 Currently, only MIPS targets are supported. 9622 * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732, 9623 N1241, is now supported as a GCC extension to C for targets 9624 i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu, 9625 and x86_64-*-linux-gnu. The feature introduces new data types 9626 _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF, 9627 DD, and DL. 9628 9629 C++ 9630 9631 * [9]Experimental support for the upcoming ISO C++ standard, C++0x. 9632 * -Wc++0x-compat has been added and is now enabled by default for 9633 -Wall. It produces warnings for constructs whose meaning differs 9634 between ISO C++ 1998 and C++0x. 9635 * The -Wparentheses option now works for C++ as it does for C. It 9636 warns if parentheses are omitted when operators with confusing 9637 precedence are nested. It also warns about ambiguous else 9638 statements. Since -Wparentheses is enabled by -Wall, this may cause 9639 additional warnings with existing C++ code which uses -Wall. These 9640 new warnings may be disabled by using -Wall -Wno-parentheses. 9641 * The -Wmissing-declarations now works for C++ as it does for C. 9642 * The -fvisibility-ms-compat flag was added, to make it easier to 9643 port larger projects using shared libraries from Microsoft's Visual 9644 Studio to ELF and Mach-O systems. 9645 * C++ attribute handling has been overhauled for template arguments 9646 (ie dependent types). In particular, __attribute__((aligned(T))); 9647 works for C++ types. 9648 9649 Runtime Library (libstdc++) 9650 9651 * [10]Experimental support for the upcoming ISO C++ standard, C++0x. 9652 * Support for TR1 mathematical special functions and regular 9653 expressions. 9654 * Default what implementations give more elaborate exception strings 9655 for bad_cast, bad_typeid, bad_exception, and bad_alloc. 9656 * Header dependencies have been streamlined, reducing unnecessary 9657 includes and pre-processed bloat. 9658 * Variadic template implementations of items in <tuple> and 9659 <functional>. 9660 * An experimental [11]parallel mode has been added. This is a 9661 parallel implementation of many C++ Standard library algorithms, 9662 like std::accumulate, std::for_each, std::transform, or std::sort, 9663 to give but four examples. These algorithms can be substituted for 9664 the normal (sequential) libstdc++ algorithms on a piecemeal basis, 9665 or all existing algorithms can be transformed via the 9666 -D_GLIBCXX_PARALLEL macro. 9667 * Debug mode versions of classes in <unordered_set> and 9668 <unordered_map>. 9669 * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are 9670 now <backward/hash_set> and <backward/hash_map>. This code: 9671 #include <ext/hash_set> 9672 __gnu_cxx::hash_set<int> s; 9673 9674 Can be transformed (in order of preference) to: 9675 #include <tr1/unordered_set> 9676 std::tr1::unordered_set<int> s; 9677 9678 or 9679 #include <backward/hash_set> 9680 __gnu_cxx::hash_set<int> s; 9681 9682 Similar transformations apply to __gnu_cxx::hash_map, 9683 __gnu_cxx::hash_multimap, __gnu_cxx::hash_set, 9684 __gnu_cxx::hash_multiset. 9685 9686 Fortran 9687 9688 * Due to the fact that the GMP and MPFR libraries are required for 9689 all languages, Fortran is no longer special in this regard and is 9690 available by default. 9691 * The [12]-fexternal-blas option has been added, which generates 9692 calls to BLAS routines for intrinsic matrix operations such as 9693 matmul rather than using the built-in algorithms. 9694 * Support to give a backtrace (compiler flag -fbacktrace or 9695 environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems 9696 only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a 9697 run-time error occured. 9698 * GNU Fortran now defines __GFORTRAN__ when it runs the C 9699 preprocessor (CPP). 9700 * The [13]-finit-local-zero, -finit-real, -finit-integer, 9701 -finit-character, and -finit-logical options have been added, which 9702 can be used to initialize local variables. 9703 * The intrinsic procedures [14]GAMMA and [15]LGAMMA have been added, 9704 which calculate the Gamma function and its logarithm. Use EXTERNAL 9705 gamma if you want to use your own gamma function. 9706 * GNU Fortran now regards the backslash character as literal (as 9707 required by the Fortran 2003 standard); using [16]-fbackslash GNU 9708 Fortran interprets backslashes as C-style escape characters. 9709 * The [17]interpretation of binary, octal and hexadecimal (BOZ) 9710 literal constants has been changed. Before they were always 9711 interpreted as integer; now they are bit-wise transferred as 9712 argument of INT, REAL, DBLE and CMPLX as required by the Fortran 9713 2003 standard, and for real and complex variables in DATA 9714 statements or when directly assigned to real and complex variables. 9715 Everywhere else and especially in expressions they are still 9716 regarded as integer constants. 9717 * Fortran 2003 support has been extended: 9718 + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE 9719 + Pointer intent 9720 + Intrinsic module ISO_ENV_FORTRAN 9721 + Interoperability with C (ISO C Bindings) 9722 + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER 9723 attribute) 9724 + Fortran 2003 BOZ 9725 9726 Java (GCJ) 9727 9728 * GCJ now uses the Eclipse Java compiler for its Java parsing needs. 9729 This enables the use of all 1.5 language features, and fixes most 9730 existing front end bugs. 9731 * libgcj now supports all 1.5 language features which require runtime 9732 support: foreach, enum, annotations, generics, and auto-boxing. 9733 * We've made many changes to the tools shipped with gcj. 9734 + The old jv-scan tool has been removed. This tool never really 9735 worked properly. There is no replacement. 9736 + gcjh has been rewritten. Some of its more obscure options no 9737 longer work, but are still recognized in an attempt at 9738 compatibility. gjavah is a new program with similar 9739 functionality but different command-line options. 9740 + grmic and grmiregistry have been rewritten. grmid has been 9741 added. 9742 + gjar replaces the old fastjar. 9743 + gjarsigner (used for signing jars), gkeytool (used for key 9744 management), gorbd (for CORBA), gserialver (computes 9745 serialization UIDs), and gtnameserv (also for CORBA) are now 9746 installed. 9747 * The ability to dump the contents of the java run time heap to a 9748 file for off-line analysis has been added. The heap dumps may be 9749 analyzed with the new gc-analyze tool. They may be generated on 9750 out-of-memory conditions or on demand and are controlled by the new 9751 run time class gnu.gcj.util.GCInfo. 9752 * java.util.TimeZone can now read files from /usr/share/zoneinfo to 9753 provide correct, updated, timezone information. This means that 9754 packagers no longer have to update libgcj when a time zone change 9755 is published. 9756 9757 New Targets and Target Specific Improvements 9758 9759 IA-32/x86-64 9760 9761 * Tuning for Intel Core 2 processors is available via -mtune=core2 9762 and -march=core2. 9763 * Tuning for AMD Geode processors is available via -mtune=geode and 9764 -march=geode. 9765 * Code generation of block move (memcpy) and block set (memset) was 9766 rewritten. GCC can now pick the best algorithm (loop, unrolled 9767 loop, instruction with rep prefix or a library call) based on the 9768 size of the block being copied and the CPU being optimized for. A 9769 new option -minline-stringops-dynamically has been added. With this 9770 option string operations of unknown size are expanded such that 9771 small blocks are copied by in-line code, while for large blocks a 9772 library call is used. This results in faster code than 9773 -minline-all-stringops when the library implementation is capable 9774 of using cache hierarchy hints. The heuristic choosing the 9775 particular algorithm can be overwritten via -mstringop-strategy. 9776 Newly also memset of values different from 0 is inlined. 9777 * GCC no longer places the cld instruction before string operations. 9778 Both i386 and x86-64 ABI documents mandate the direction flag to be 9779 clear at the entry of a function. It is now invalid to set the flag 9780 in asm statement without reseting it afterward. 9781 * Support for SSSE3 built-in functions and code generation are 9782 available via -mssse3. 9783 * Support for SSE4.1 built-in functions and code generation are 9784 available via -msse4.1. 9785 * Support for SSE4.2 built-in functions and code generation are 9786 available via -msse4.2. 9787 * Both SSE4.1 and SSE4.2 support can be enabled via -msse4. 9788 * A new set of options -mpc32, -mpc64 and -mpc80 have been added to 9789 allow explicit control of x87 floating point precision. 9790 * Support for __float128 (TFmode) IEEE quad type and corresponding 9791 TCmode IEEE complex quad type is available via the soft-fp library 9792 on x86_64 targets. This includes basic arithmetic operations 9793 (addition, subtraction, negation, multiplication and division) on 9794 __float128 real and TCmode complex values, the full set of IEEE 9795 comparisons between __float128 values, conversions to and from 9796 float, double and long double floating point types, as well as 9797 conversions to and from signed or unsigned integer, signed or 9798 unsigned long integer and signed or unsigned quad (TImode) integer 9799 types. Additionally, all operations generate the full set of IEEE 9800 exceptions and support the full set of IEEE rounding modes. 9801 * GCC can now utilize the ACML library for vectorizing calls to a set 9802 of C99 functions on x86_64 if -mveclibabi=acml is specified and you 9803 link to an ACML ABI compatible library. 9804 9805 ARM 9806 9807 * Compiler and Library support for Thumb-2 and the ARMv7 architecture 9808 has been added. 9809 9810 CRIS 9811 9812 New features 9813 9814 * Compiler and Library support for the CRIS v32 architecture, as 9815 found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been 9816 added. 9817 9818 Configuration changes 9819 9820 * The cris-*-elf target now includes support for CRIS v32, including 9821 libraries, through the -march=v32 option. 9822 * A new crisv32-*-elf target defaults to generate code for CRIS v32. 9823 * A new crisv32-*-linux* target defaults to generate code for CRIS 9824 v32. 9825 * The cris-*-aout target has been obsoleted. 9826 9827 Improved support for built-in functions 9828 9829 * GCC can now use the lz and swapwbr instructions to implement the 9830 __builtin_clz, __builtin_ctz and __builtin_ffs family of functions. 9831 * __builtin_bswap32 is now implemented using the swapwb instruction, 9832 when available. 9833 9834 m68k and ColdFire 9835 9836 New features 9837 9838 * Support for several new ColdFire processors has been added. You can 9839 generate code for them using the new -mcpu option. 9840 * All targets now support ColdFire processors. 9841 * m68k-uclinux targets have improved support for C++ constructors and 9842 destructors, and for shared libraries. 9843 * It is now possible to set breakpoints on the first or last line of 9844 a function, even if there are no statements on that line. 9845 9846 Optimizations 9847 9848 * Support for sibling calls has been added. 9849 * More use is now made of the ColdFire mov3q instruction. 9850 * __builtin_clz is now implemented using the ff1 ColdFire 9851 instruction, when available. 9852 * GCC now honors the -m68010 option. 68010 code now uses clr rather 9853 than move to zero volatile memory. 9854 * 68020 targets and above can now use symbol(index.size*scale) 9855 addresses for indexed array accesses. Earlier compilers would 9856 always load the symbol into a base register first. 9857 9858 Configuration changes 9859 9860 * All m68k and ColdFire targets now allow the default processor to be 9861 set at configure time using --with-cpu. 9862 * A --with-arch configuration option has been added. This option 9863 allows you to restrict a target to ColdFire or non-ColdFire 9864 processors. 9865 9866 Preprocessor macros 9867 9868 * An __mcfv*__ macro is now defined for all ColdFire targets. 9869 (Earlier versions of GCC only defined __mcfv4e__.) 9870 * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added. 9871 * All targets now define __mc68010 and __mc68010__ when generating 9872 68010 code. 9873 9874 Command-line changes 9875 9876 * New command-line options -march, -mcpu, -mtune and -mhard-float 9877 have been added. These options apply to both m68k and ColdFire 9878 targets. 9879 * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative 9880 versions of -mshort, etc. 9881 * -fforce-addr has been removed. It is now ignored by the compiler. 9882 9883 Other improvements 9884 9885 * ColdFire targets now try to maintain a 4-byte-aligned stack where 9886 possible. 9887 * m68k-uclinux targets now try to avoid situations that lead to the 9888 load-time error: BINFMT_FLAT: reloc outside program. 9889 9890 MIPS 9891 9892 Changes to existing configurations 9893 9894 * libffi and libjava now support all three GNU/Linux ABIs: o32, n32 9895 and n64. Every GNU/Linux configuration now builds these libraries 9896 by default. 9897 * GNU/Linux configurations now generate -mno-shared code unless 9898 overridden by -fpic, -fPIC, -fpie or -fPIE. 9899 * mipsisa32*-linux-gnu configurations now generate hard-float code by 9900 default, just like other mipsisa32* and mips*-linux-gnu 9901 configurations. You can build a soft-float version of any 9902 mips*-linux-gnu configuration by passing --with-float=soft to 9903 configure. 9904 * mips-wrs-vxworks now supports run-time processes (RTPs). 9905 9906 Changes to existing command-line options 9907 9908 * The -march and -mtune options no longer accept 24k as a processor 9909 name. Please use 24kc, 24kf2_1 or 24kf1_1 instead. 9910 * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and 9911 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The 9912 options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for 9913 24kx, 24kex and 34kx. 9914 9915 New configurations 9916 9917 GCC now supports the following configurations: 9918 * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by 9919 default. Earlier releases also recognized this configuration, but 9920 they treated it in the same way as mipsisa32*-linux-gnu*. Note that 9921 you can customize any mips*-linux-gnu* configuration to a 9922 particular ISA or processor by passing an appropriate --with-arch 9923 option to configure. 9924 * mipsisa*-sde-elf*, which provides compatibility with MIPS 9925 Technologies' SDE toolchains. The configuration uses the SDE 9926 libraries by default, but you can use it like other newlib-based 9927 ELF configurations by passing --with-newlib to configure. It is the 9928 only configuration besides mips64vr*-elf* to build MIPS16 as well 9929 as non-MIPS16 libraries. 9930 * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf* 9931 configuration, but uses the o32 and o64 ABIs instead of the 32-bit 9932 and 64-bit forms of the EABI. 9933 9934 New processors and application-specific extensions 9935 9936 * Support for the SmartMIPS ASE is available through the new 9937 -msmartmips option. 9938 * Support for revision 2 of the DSP ASE is available through the new 9939 -mdspr2 option. A new preprocessor macro called __mips_dsp_rev 9940 indicates the revision of the ASE in use. 9941 * Support for the 4KS and 74K families of processors is available 9942 through the -march and -mtune options. 9943 9944 Improved support for built-in functions 9945 9946 * GCC can now use load-linked, store-conditional and sync 9947 instructions to implement atomic built-in functions such as 9948 __sync_fetch_and_add. The memory reference must be 4 bytes wide for 9949 32-bit targets and either 4 or 8 bytes wide for 64-bit targets. 9950 * GCC can now use the clz and dclz instructions to implement the 9951 __builtin_ctz and __builtin_ffs families of functions. 9952 * There is a new __builtin___clear_cache function for flushing the 9953 instruction cache. GCC expands this function inline on MIPS32 9954 revision 2 targets, otherwise it calls the function specified by 9955 -mcache-flush-func. 9956 9957 MIPS16 improvements 9958 9959 * GCC can now compile objects that contain a mixture of MIPS16 and 9960 non-MIPS16 code. There are two new attributes, mips16 and nomips16, 9961 for specifying which mode a function should use. 9962 * A new option called -minterlink-mips16 makes non-MIPS16 code 9963 link-compatible with MIPS16 code. 9964 * After many bug fixes, the long-standing MIPS16 -mhard-float support 9965 should now work fairly reliably. 9966 * GCC can now use the MIPS16e save and restore instructions. 9967 * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled 9968 with -G0 -fsection-anchors is often smaller than code compiled with 9969 -G8. However, please note that you must usually compile all objects 9970 in your application with the same -G option; see the documentation 9971 of -G for details. 9972 * A new option called-mcode-readable specifies which instructions are 9973 allowed to load from the code segment. -mcode-readable=yes is the 9974 default and says that any instruction may load from the code 9975 segment. The other alternatives are -mcode-readable=pcrel, which 9976 says that only PC-relative MIPS16 instructions may load from the 9977 code segment, and -mcode-readable=no, which says that no 9978 instruction may do so. Please see the documentation for more 9979 details, including example uses. 9980 9981 Small-data improvements 9982 9983 There are three new options for controlling small data: 9984 * -mno-extern-sdata, which disables small-data accesses for 9985 externally-defined variables. Code compiled with -Gn 9986 -mno-extern-sdata will be link-compatible with any -G setting 9987 between -G0 and -Gn inclusive. 9988 * -mno-local-sdata, which disables the use of small-data sections for 9989 data that is not externally visible. This option can be a useful 9990 way of reducing small-data usage in less performance-critical parts 9991 of an application. 9992 * -mno-gpopt, which disables the use of the $gp register while still 9993 honoring the -G limit when placing externally-visible data. This 9994 option implies -mno-extern-sdata and -mno-local-sdata and it can be 9995 useful in situations where $gp does not necessarily hold the 9996 expected value. 9997 9998 Miscellaneous improvements 9999 10000 * There is a new option called -mbranch-cost for tweaking the 10001 perceived cost of branches. 10002 * If GCC is configured to use a version of GAS that supports the 10003 .gnu_attribute directive, it will use that directive to record 10004 certain properties of the output code. .gnu_attribute is new to GAS 10005 2.18. 10006 * There are two new function attributes, near and far, for overriding 10007 the command-line setting of -mlong-calls on a function-by-function 10008 basis. 10009 * -mfp64, which previously required a 64-bit target, now works with 10010 MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and 10011 mipsisa*-sde-elf* configurations provide suitable library support. 10012 * GCC now recognizes the -mdmx and -mmt options and passes them down 10013 to the assembler. It does nothing else with the options at present. 10014 10015 SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture 10016 (BEA) 10017 10018 * Support has been added for this new architecture. 10019 10020 RS6000 (POWER/PowerPC) 10021 10022 * Support for the PowerPC 750CL paired-single instructions has been 10023 added with a new powerpc-*-linux*paired* target configuration. It 10024 is enabled by an associated -mpaired option and can be accessed 10025 using new built-in functions. 10026 * Support for auto-detecting architecture and system configuration to 10027 auto-select processor optimization tuning. 10028 * Support for VMX on AIX 5.3 has been added. 10029 * Support for AIX Version 6.1 has been added. 10030 10031 S/390, zSeries and System z9 10032 10033 * Support for the IBM System z9 EC/BC processor (z9 GA3) has been 10034 added. When using the -march=z9-ec option, the compiler will 10035 generate code making use of instructions provided by the decimal 10036 floating point facility and the floating point conversion facility 10037 (pfpo). Besides the instructions used to implement decimal floating 10038 point operations these facilities also contain instructions to move 10039 between general purpose and floating point registers and to modify 10040 and copy the sign-bit of floating point values. 10041 * When the -march=z9-ec option is used the new 10042 -mhard-dfp/-mno-hard-dfp options can be used to specify whether the 10043 decimal floating point hardware instructions will be used or not. 10044 If none of them is given the hardware support is enabled by 10045 default. 10046 * The -mstack-guard option can now be omitted when using stack 10047 checking via -mstack-size in order to let GCC choose a sensible 10048 stack guard value according to the frame size of each function. 10049 * Various changes to improve performance of generated code have been 10050 implemented, including: 10051 + The condition code set by an add logical with carry 10052 instruction is now available for overflow checks like: a + b + 10053 carry < b. 10054 + The test data class instruction is now used to implement 10055 sign-bit and infinity checks of binary and decimal floating 10056 point numbers. 10057 10058 SPARC 10059 10060 * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been 10061 added. 10062 10063 Xtensa 10064 10065 * Stack unwinding for exception handling now uses by default a 10066 specialized version of DWARF unwinding. This is not 10067 binary-compatible with the setjmp/longjmp (sjlj) unwinding used for 10068 Xtensa with previous versions of GCC. 10069 * For Xtensa processors that include the Conditional Store option, 10070 the built-in functions for atomic memory access are now implemented 10071 using S32C1I instructions. 10072 * If the Xtensa NSA option is available, GCC will use it to implement 10073 the __builtin_ctz and __builtin_clz functions. 10074 10075 Documentation improvements 10076 10077 * Existing libstdc++ documentation has been edited and restructured 10078 into a single DocBook XML manual. The results can be viewed online 10079 [18]here. 10080 10081 Other significant improvements 10082 10083 * The compiler's --help command-line option has been extended so that 10084 it now takes an optional set of arguments. These arguments restrict 10085 the information displayed to specific classes of command-line 10086 options, and possibly only a subset of those options. It is also 10087 now possible to replace the descriptive text associated with each 10088 displayed option with an indication of its current value, or for 10089 binary options, whether it has been enabled or disabled. 10090 Here are some examples. The following will display all the options 10091 controlling warning messages: 10092 --help=warnings 10093 10094 Whereas this will display all the undocumented, target specific 10095 options: 10096 --help=target,undocumented 10097 10098 This sequence of commands will display the binary optimizations 10099 that are enabled by -O3: 10100 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 10101 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 10102 diff /tmp/O2-opts /tmp/O3-opts | grep enabled 10103 10104 * The configure options --with-pkgversion and --with-bugurl have been 10105 added. These allow distributors of GCC to include a 10106 distributor-specific string in manuals and --version output and to 10107 specify the URL for reporting bugs in their versions of GCC. 10108 10109 GCC 4.3.1 10110 10111 This is the [19]list of problem reports (PRs) from GCC's bug tracking 10112 system that are known to be fixed in the 4.3.1 release. This list might 10113 not be complete (that is, it is possible that some PRs that have been 10114 fixed are not listed here). 10115 10116 Target Specific Changes 10117 10118 IA-32/x86-64 10119 10120 ABI changes 10121 10122 * Starting with GCC 4.3.1, decimal floating point variables are 10123 aligned to their natural boundaries when they are passed on the 10124 stack for i386. 10125 10126 Command-line changes 10127 10128 * Starting with GCC 4.3.1, the -mcld option has been added to 10129 automatically generate a cld instruction in the prologue of 10130 functions that use string instructions. This option is used for 10131 backward compatibility on some operating systems and can be enabled 10132 by default for 32-bit x86 targets by configuring GCC with the 10133 --enable-cld configure option. 10134 10135 GCC 4.3.2 10136 10137 This is the [20]list of problem reports (PRs) from GCC's bug tracking 10138 system that are known to be fixed in the 4.3.2 release. This list might 10139 not be complete (that is, it is possible that some PRs that have been 10140 fixed are not listed here). 10141 10142 GCC 4.3.3 10143 10144 This is the [21]list of problem reports (PRs) from GCC's bug tracking 10145 system that are known to be fixed in the 4.3.3 release. This list might 10146 not be complete (that is, it is possible that some PRs that have been 10147 fixed are not listed here). 10148 10149 GCC 4.3.4 10150 10151 This is the [22]list of problem reports (PRs) from GCC's bug tracking 10152 system that are known to be fixed in the 4.3.4 release. This list might 10153 not be complete (that is, it is possible that some PRs that have been 10154 fixed are not listed here). 10155 10156 GCC 4.3.5 10157 10158 This is the [23]list of problem reports (PRs) from GCC's bug tracking 10159 system that are known to be fixed in the 4.3.5 release. This list might 10160 not be complete (that is, it is possible that some PRs that have been 10161 fixed are not listed here). 10162 10163 GCC 4.3.6 10164 10165 This is the [24]list of problem reports (PRs) from GCC's bug tracking 10166 system that are known to be fixed in the 4.3.6 release. This list might 10167 not be complete (that is, it is possible that some PRs that have been 10168 fixed are not listed here). 10169 10170 10171 For questions related to the use of GCC, please consult these web 10172 pages and the [25]GCC manuals. If that fails, the 10173 [26]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10174 web pages and the development of GCC are welcome on our developer 10175 list at [27]gcc (a] gcc.gnu.org. All of [28]our lists have public 10176 archives. 10177 10178 Copyright (C) [29]Free Software Foundation, Inc. Verbatim copying and 10179 distribution of this entire article is permitted in any medium, 10180 provided this notice is preserved. 10181 10182 These pages are [30]maintained by the GCC team. Last modified 10183 2018-09-30[31]. 10184 10185 References 10186 10187 1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5 10188 2. https://gmplib.org/ 10189 3. https://www.mpfr.org/ 10190 4. https://gcc.gnu.org/install/prerequisites.html 10191 5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html 10192 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 10193 7. http://gcc.gnu.org/gcc-4.3/porting_to.html 10194 8. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 10195 9. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 10196 10. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 10197 11. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html 10198 12. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options 10199 13. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167 10200 14. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html 10201 15. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html 10202 16. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html 10203 17. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html 10204 18. https://gcc.gnu.org/onlinedocs/libstdc++/ 10205 19. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1 10206 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2 10207 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3 10208 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4 10209 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5 10210 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6 10211 25. https://gcc.gnu.org/onlinedocs/ 10212 26. mailto:gcc-help (a] gcc.gnu.org 10213 27. mailto:gcc (a] gcc.gnu.org 10214 28. https://gcc.gnu.org/lists.html 10215 29. https://www.fsf.org/ 10216 30. https://gcc.gnu.org/about.html 10217 31. http://validator.w3.org/check/referer 10218 ====================================================================== 10219 http://gcc.gnu.org/gcc-4.2/index.html 10220 GCC 4.2 Release Series 10221 10222 May 19, 2008 10223 10224 The [1]GNU project and the GCC developers are pleased to announce the 10225 release of GCC 4.2.4. 10226 10227 This release is a bug-fix release, containing fixes for regressions in 10228 GCC 4.2.3 relative to previous releases of GCC. 10229 10230 This release series is no longer maintained. 10231 10232 Release History 10233 10234 GCC 4.2.4 10235 May 19, 2008 ([2]changes) 10236 10237 GCC 4.2.3 10238 February 1, 2008 ([3]changes) 10239 10240 GCC 4.2.2 10241 October 7, 2007 ([4]changes) 10242 10243 GCC 4.2.1 10244 July 18, 2007 ([5]changes) 10245 10246 GCC 4.2.0 10247 May 13, 2007 ([6]changes) 10248 10249 References and Acknowledgements 10250 10251 GCC used to stand for the GNU C Compiler, but since the compiler 10252 supports several other languages aside from C, it now stands for the 10253 GNU Compiler Collection. 10254 10255 A list of [7]successful builds is updated as new information becomes 10256 available. 10257 10258 The GCC developers would like to thank the numerous people that have 10259 contributed new features, improvements, bug fixes, and other changes as 10260 well as test results to GCC. This [8]amazing group of volunteers is 10261 what makes GCC successful. 10262 10263 For additional information about GCC please refer to the [9]GCC project 10264 web site or contact the [10]GCC development mailing list. 10265 10266 To obtain GCC please use [11]our mirror sites or [12]our SVN server. 10267 10268 10269 For questions related to the use of GCC, please consult these web 10270 pages and the [13]GCC manuals. If that fails, the 10271 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10272 web pages and the development of GCC are welcome on our developer 10273 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 10274 archives. 10275 10276 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 10277 distribution of this entire article is permitted in any medium, 10278 provided this notice is preserved. 10279 10280 These pages are [18]maintained by the GCC team. Last modified 10281 2018-09-30[19]. 10282 10283 References 10284 10285 1. http://www.gnu.org/ 10286 2. http://gcc.gnu.org/gcc-4.2/changes.html 10287 3. http://gcc.gnu.org/gcc-4.2/changes.html 10288 4. http://gcc.gnu.org/gcc-4.2/changes.html 10289 5. http://gcc.gnu.org/gcc-4.2/changes.html 10290 6. http://gcc.gnu.org/gcc-4.2/changes.html 10291 7. http://gcc.gnu.org/gcc-4.2/buildstat.html 10292 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 10293 9. http://gcc.gnu.org/index.html 10294 10. mailto:gcc (a] gcc.gnu.org 10295 11. http://gcc.gnu.org/mirrors.html 10296 12. http://gcc.gnu.org/svn.html 10297 13. https://gcc.gnu.org/onlinedocs/ 10298 14. mailto:gcc-help (a] gcc.gnu.org 10299 15. mailto:gcc (a] gcc.gnu.org 10300 16. https://gcc.gnu.org/lists.html 10301 17. https://www.fsf.org/ 10302 18. https://gcc.gnu.org/about.html 10303 19. http://validator.w3.org/check/referer 10304 ====================================================================== 10305 http://gcc.gnu.org/gcc-4.2/changes.html 10306 GCC 4.2 Release Series 10307 Changes, New Features, and Fixes 10308 10309 Caveats 10310 10311 * GCC no longer accepts the -fshared-data option. This option has had 10312 no effect in any GCC 4 release; the targets to which the option 10313 used to apply had been removed before GCC 4.0. 10314 10315 General Optimizer Improvements 10316 10317 * New command-line options specify the possible relationships among 10318 parameters and between parameters and global data. For example, 10319 -fargument-noalias-anything specifies that arguments do not alias 10320 any other storage. 10321 Each language will automatically use whatever option is required by 10322 the language standard. You should not need to use these options 10323 yourself. 10324 10325 New Languages and Language specific improvements 10326 10327 * [1]OpenMP is now supported for the C, C++ and Fortran compilers. 10328 * New command-line options -fstrict-overflow and -Wstrict-overflow 10329 have been added. -fstrict-overflow tells the compiler that it may 10330 assume that the program follows the strict signed overflow 10331 semantics permitted for the language: for C and C++ this means that 10332 the compiler may assume that signed overflow does not occur. For 10333 example, a loop like 10334 for (i = 1; i > 0; i *= 2) 10335 10336 is presumably intended to continue looping until i overflows. With 10337 -fstrict-overflow, the compiler may assume that signed overflow 10338 will not occur, and transform this into an infinite loop. 10339 -fstrict-overflow is turned on by default at -O2, and may be 10340 disabled via -fno-strict-overflow. The -Wstrict-overflow option may 10341 be used to warn about cases where the compiler assumes that signed 10342 overflow will not occur. It takes five different levels: 10343 -Wstrict-overflow=1 to 5. See the [2]documentation for details. 10344 -Wstrict-overflow=1 is enabled by -Wall. 10345 * The new command-line option -fno-toplevel-reorder directs GCC to 10346 emit top-level functions, variables, and asm statements in the same 10347 order that they appear in the input file. This is intended to 10348 support existing code which relies on a particular ordering (for 10349 example, code which uses top-level asm statements to switch 10350 sections). For new code, it is generally better to use function and 10351 variable attributes. The -fno-toplevel-reorder option may be used 10352 for most cases which currently use -fno-unit-at-a-time. The 10353 -fno-unit-at-a-time option will be removed in some future version 10354 of GCC. If you know of a case which requires -fno-unit-at-a-time 10355 which is not fixed by -fno-toplevel-reorder, please open a bug 10356 report. 10357 10358 C family 10359 10360 * The pragma redefine_extname will now macro expand its tokens for 10361 compatibility with SunPRO. 10362 * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct 10363 GCC to handle inline functions as specified in the C99 standard. In 10364 preparation for this, GCC 4.2 will warn about any use of non-static 10365 inline functions in gnu99 or c99 mode. This new warning may be 10366 disabled with the new gnu_inline function attribute or the new 10367 -fgnu89-inline command-line option. Also, GCC 4.2 and later will 10368 define one of the preprocessor macros __GNUC_GNU_INLINE__ or 10369 __GNUC_STDC_INLINE__ to indicate the semantics of inline functions 10370 in the current compilation. 10371 * A new command-line option -Waddress has been added to warn about 10372 suspicious uses of memory addresses as, for example, using the 10373 address of a function in a conditional expression, and comparisons 10374 against the memory address of a string literal. This warning is 10375 enabled by -Wall. 10376 10377 C++ 10378 10379 * C++ visibility handling has been overhauled. 10380 Restricted visiblity is propagated from classes to members, from 10381 functions to local statics, and from templates and template 10382 arguments to instantiations, unless the latter has explicitly 10383 declared visibility. 10384 The visibility attribute for a class must come between the 10385 class-key and the name, not after the closing brace. 10386 Attributes are now allowed for enums and elaborated-type-specifiers 10387 that only declare a type. 10388 Members of the anonymous namespace are now local to a particular 10389 translation unit, along with any other declarations which use them, 10390 though they are still treated as having external linkage for 10391 language semantics. 10392 * The (undocumented) extension which permitted templates with default 10393 arguments to be bound to template template parameters with fewer 10394 parameters has been removed. For example: 10395 template <template <typename> class C> 10396 void f(C<double>) {} 10397 10398 template <typename T, typename U = int> 10399 struct S {}; 10400 10401 template void f(S<double>); 10402 10403 is no longer accepted by G++. The reason this code is not accepted 10404 is that S is a template with two parameters; therefore, it cannot 10405 be bound to C which has only one parameter. 10406 * The <?, >?, <?=, and >?= operators, deprecated in previous GCC 10407 releases, have been removed. 10408 * The command-line option -fconst-strings, deprecated in previous GCC 10409 releases, has been removed. 10410 * The configure variable enable-__cxa_atexit is now enabled by 10411 default for more targets. Enabling this variable is necessary in 10412 order for static destructors to be executed in the correct order, 10413 but it depends upon the presence of a non-standard C library in the 10414 target library in order to work. The variable is now enabled for 10415 more targets which are known to have suitable C libraries. 10416 * -Wextra will produce warnings for if statements with a semicolon as 10417 the only body, to catch code like: 10418 if (a); 10419 return 1; 10420 return 0; 10421 10422 To suppress the warning in valid cases, use { } instead. 10423 * The C++ front end now also produces strict aliasing warnings when 10424 -fstrict-aliasing -Wstrict-aliasing is in effect. 10425 10426 Runtime Library (libstdc++) 10427 10428 * Added support for TR1 <random>, <complex>, and C compatibility 10429 headers. In addition, a lock-free version of shared_ptr was 10430 contributed as part of Phillip Jordan's Google Summer of Code 10431 project on lock-free containers. 10432 * In association with the Summer of Code work on lock-free 10433 containers, the interface for atomic builtins was adjusted, 10434 creating simpler alternatives for non-threaded code paths. Also, 10435 usage was consolidated and all elements were moved from namespace 10436 std to namespace__gnu_cxx. Affected interfaces are the functions 10437 __exchange_and_add, __atomic_add, and the objects __mutex, 10438 __recursive_mutex, and __scoped_lock. 10439 * Support for versioning weak symbol names via namespace association 10440 was added. However, as this changes the names of exported symbols, 10441 this is turned off by default in the current ABI. Intrepid users 10442 can enable this feature by using 10443 --enable-symvers=gnu-versioned-namespace during configuration. 10444 * Revised, simplified, and expanded policy-based associative 10445 containers, including data types for tree and trie forms 10446 (basic_tree, tree, trie), lists (list_update), and both 10447 collision-chaining and probing hash-based containers 10448 (basic_hash_table, cc_hash_table, gp_hash_table). More details per 10449 the [3]documentation. 10450 * The implementation of the debug mode was modified, whereby the 10451 debug namespaces were nested inside of namespace std and namespace 10452 __gnu_cxx in order to resolve some long standing corner cases 10453 involving name lookup. Debug functionality from the policy-based 10454 data structures was consolidated and enabled with the single macro, 10455 _GLIBCXX_DEBUG. See PR 26142 for more information. 10456 * Added extensions for type traits: __conditional_type, 10457 __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if. 10458 * Added a typelist implementation for compile-time meta-programming. 10459 Elements for typelist construction and operation can be found 10460 within namespace __gnu_cxx::typelist. 10461 * Added a new allocator, __gnu_cxx::throw_allocator, for testing 10462 exception-safety. 10463 * Enabled library-wide visibility control, allowing -fvisibility to 10464 be used. 10465 * Consolidated all nested namespaces and the conversion of 10466 __gnu_internal implementation-private details to anonymous 10467 namespaces whenever possible. 10468 * Implemented LWG resolutions DR 431 and DR 538. 10469 10470 Fortran 10471 10472 * Support for allocatable components has been added (TR 15581 and 10473 Fortran 2003). 10474 * Support for the Fortran 2003 streaming IO extension has been added. 10475 * The GNU Fortran compiler now uses 4-byte record markers by default 10476 for unformatted files to be compatible with g77 and most other 10477 compilers. The implementation allows for records greater than 2 GB 10478 and is compatible with several other compilers. Older versions of 10479 gfortran used 8-byte record markers by default (on most systems). 10480 In order to change the length of the record markers, e.g. to read 10481 unformatted files created by older gfortran versions, the 10482 [4]-frecord-marker=8 option can be used. 10483 10484 Java (GCJ) 10485 10486 * A new command-line option -static-libgcj has been added for targets 10487 that use a linker compatible with GNU Binutils. As its name 10488 implies, this causes libgcj to be linked statically. In some cases 10489 this causes the resulting executable to start faster and use less 10490 memory than if the shared version of libgcj were used. However 10491 caution should be used as it can also cause essential parts of the 10492 library to be omitted. Some of these issues are discussed in: 10493 [5]https://gcc.gnu.org/wiki/Statically_linking_libgcj 10494 * fastjar is no longer bundled with GCC. To build libgcj, you will 10495 need either InfoZIP (both zip and unzip) or an external jar 10496 program. In the former case, the GCC build will install a jar shell 10497 script that is based on InfoZIP and provides the same functionality 10498 as fastjar. 10499 10500 New Targets and Target Specific Improvements 10501 10502 IA-32/x86-64 10503 10504 * -mtune=generic can now be used to generate code running well on 10505 common x86 chips. This includes AMD Athlon, AMD Opteron, Intel 10506 Pentium-M, Intel Pentium 4 and Intel Core 2. 10507 * -mtune=native and -march=native will produce code optimized for the 10508 host architecture as detected using the cpuid instruction. 10509 * Added a new command-line option -fstackrealign and and 10510 __attribute__ ((force_align_arg_pointer)) to realign the stack at 10511 runtime. This allows functions compiled with a vector-aligned stack 10512 to be invoked from legacy objects that keep only word-alignment. 10513 10514 SPARC 10515 10516 * The default CPU setting has been changed from V7 to V9 in 32-bit 10517 mode on Solaris 7 and above. This is already the case in 64-bit 10518 mode. It can be overridden by specifying --with-cpu at configure 10519 time. 10520 * Back-end support of built-in functions for atomic memory access has 10521 been implemented. 10522 * Support for the Sun UltraSPARC T1 (Niagara) processor has been 10523 added. 10524 10525 M32C 10526 10527 * Various bug fixes have made some functions (notably, functions 10528 returning structures) incompatible with previous releases. 10529 Recompiling all libraries is recommended. Note that code quality 10530 has considerably improved since 4.1, making a recompile even more 10531 beneficial. 10532 10533 MIPS 10534 10535 * Added support for the Broadcom SB-1A core. 10536 10537 IA-64 10538 10539 * Added support for IA-64 data and control speculation. By default 10540 speculation is enabled only during second scheduler pass. A number 10541 of machine flags was introduced to control the usage of speculation 10542 for both scheduler passes. 10543 10544 HPPA 10545 10546 * Added Java language support (libffi and libjava) for 32-bit HP-UX 10547 11 target. 10548 10549 Obsolete Systems 10550 10551 Documentation improvements 10552 10553 PDF Documentation 10554 10555 * A make pdf target has been added to the top-level makefile, 10556 enabling automated production of PDF documentation files. 10557 (Front-ends external to GCC should modify their Make-lang.in file 10558 to add a lang.pdf: target.) 10559 10560 Other significant improvements 10561 10562 Build system improvements 10563 10564 * All the components of the compiler are now bootstrapped by default. 10565 This improves the resilience to bugs in the system compiler or 10566 binary compatibility problems, as well as providing better testing 10567 of GCC 4.2 itself. In addition, if you build the compiler from a 10568 combined tree, the assembler, linker, etc. will also be 10569 bootstrapped (i.e. built with themselves). 10570 You can disable this behavior, and go back to the pre-GCC 4.2 set 10571 up, by configuring GCC with --disable-bootstrap. 10572 * The rules that configure follows to find target tools resemble more 10573 closely the locations that the built compiler will search. In 10574 addition, you can use the new configure option --with-target-tools 10575 to specify where to find the target tools used during the build, 10576 without affecting what the built compiler will use. 10577 This can be especially useful when building packages of GCC. For 10578 example, you may want to build GCC with GNU as or ld, even if the 10579 resulting compiler to work with the native assembler and linker. To 10580 do so, you can use --with-target-tools to point to the native 10581 tools. 10582 10583 Incompatible changes to the build system 10584 10585 * Front-ends external to GCC should modify their Make-lang.in file to 10586 replace double-colon rules (e.g. dvi::) with normal rules (like 10587 lang.dvi:). Front-end makefile hooks do not use double-colon rules 10588 anymore. 10589 * Up to GCC 4.1, a popular way to specify the target tools used 10590 during the build was to create directories named gas, binutils, 10591 etc. in the build tree, and create links to the tools from there. 10592 This does not work any more when the compiler is bootstrapped. The 10593 new configure option --with-target-tools provides a better way to 10594 achieve the same effect, and works for all native and cross 10595 settings. 10596 10597 10598 For questions related to the use of GCC, please consult these web 10599 pages and the [6]GCC manuals. If that fails, the 10600 [7]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10601 web pages and the development of GCC are welcome on our developer 10602 list at [8]gcc (a] gcc.gnu.org. All of [9]our lists have public archives. 10603 10604 Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and 10605 distribution of this entire article is permitted in any medium, 10606 provided this notice is preserved. 10607 10608 These pages are [11]maintained by the GCC team. Last modified 10609 2018-09-30[12]. 10610 10611 References 10612 10613 1. http://gcc.gnu.org/projects/gomp/ 10614 2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 10615 3. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html 10616 4. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html 10617 5. https://gcc.gnu.org/wiki/Statically_linking_libgcj 10618 6. https://gcc.gnu.org/onlinedocs/ 10619 7. mailto:gcc-help (a] gcc.gnu.org 10620 8. mailto:gcc (a] gcc.gnu.org 10621 9. https://gcc.gnu.org/lists.html 10622 10. https://www.fsf.org/ 10623 11. https://gcc.gnu.org/about.html 10624 12. http://validator.w3.org/check/referer 10625 ====================================================================== 10626 http://gcc.gnu.org/gcc-4.1/index.html 10627 GCC 4.1 Release Series 10628 10629 February 13, 2007 10630 10631 The [1]GNU project and the GCC developers are pleased to announce the 10632 release of GCC 4.1.2. 10633 10634 This release is a bug-fix release, containing fixes for regressions in 10635 GCC 4.1.1 relative to previous releases of GCC. 10636 10637 This release series is no longer maintained. 10638 10639 Release History 10640 10641 GCC 4.1.2 10642 February 13, 2007 ([2]changes) 10643 10644 GCC 4.1.1 10645 May 24, 2006 ([3]changes) 10646 10647 GCC 4.1.0 10648 February 28, 2006 ([4]changes) 10649 10650 References and Acknowledgements 10651 10652 GCC used to stand for the GNU C Compiler, but since the compiler 10653 supports several other languages aside from C, it now stands for the 10654 GNU Compiler Collection. 10655 10656 A list of [5]successful builds is updated as new information becomes 10657 available. 10658 10659 The GCC developers would like to thank the numerous people that have 10660 contributed new features, improvements, bug fixes, and other changes as 10661 well as test results to GCC. This [6]amazing group of volunteers is 10662 what makes GCC successful. 10663 10664 For additional information about GCC please refer to the [7]GCC project 10665 web site or contact the [8]GCC development mailing list. 10666 10667 To obtain GCC please use [9]our mirror sites or [10]our SVN server. 10668 10669 10670 For questions related to the use of GCC, please consult these web 10671 pages and the [11]GCC manuals. If that fails, the 10672 [12]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10673 web pages and the development of GCC are welcome on our developer 10674 list at [13]gcc (a] gcc.gnu.org. All of [14]our lists have public 10675 archives. 10676 10677 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and 10678 distribution of this entire article is permitted in any medium, 10679 provided this notice is preserved. 10680 10681 These pages are [16]maintained by the GCC team. Last modified 10682 2018-09-30[17]. 10683 10684 References 10685 10686 1. http://www.gnu.org/ 10687 2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 10688 3. http://gcc.gnu.org/gcc-4.1/changes.html 10689 4. http://gcc.gnu.org/gcc-4.1/changes.html 10690 5. http://gcc.gnu.org/gcc-4.1/buildstat.html 10691 6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 10692 7. http://gcc.gnu.org/index.html 10693 8. mailto:gcc (a] gcc.gnu.org 10694 9. http://gcc.gnu.org/mirrors.html 10695 10. http://gcc.gnu.org/svn.html 10696 11. https://gcc.gnu.org/onlinedocs/ 10697 12. mailto:gcc-help (a] gcc.gnu.org 10698 13. mailto:gcc (a] gcc.gnu.org 10699 14. https://gcc.gnu.org/lists.html 10700 15. https://www.fsf.org/ 10701 16. https://gcc.gnu.org/about.html 10702 17. http://validator.w3.org/check/referer 10703 ====================================================================== 10704 http://gcc.gnu.org/gcc-4.1/changes.html 10705 GCC 4.1 Release Series 10706 Changes, New Features, and Fixes 10707 10708 The latest release in the 4.1 release series is [1]GCC 4.1.2. 10709 10710 Caveats 10711 10712 General Optimizer Improvements 10713 10714 * GCC now has infrastructure for inter-procedural optimizations and 10715 the following inter-procedural optimizations are implemented: 10716 + Profile guided inlining. When doing profile feedback guided 10717 optimization, GCC can now use the profile to make better 10718 informed decisions on whether inlining of a function is 10719 profitable or not. This means that GCC will no longer inline 10720 functions at call sites that are not executed very often, and 10721 that functions at hot call sites are more likely to be 10722 inlined. 10723 A new parameter min-inline-recursive-probability is also now 10724 available to throttle recursive inlining of functions with 10725 small average recursive depths. 10726 + Discovery of pure and const functions, a form of side-effects 10727 analysis. While older GCC releases could also discover such 10728 special functions, the new IPA-based pass runs earlier so that 10729 the results are available to more optimizers. The pass is also 10730 simply more powerful than the old one. 10731 + Analysis of references to static variables and type escape 10732 analysis, also forms of side-effects analysis. The results of 10733 these passes allow the compiler to be less conservative about 10734 call-clobbered variables and references. This results in more 10735 redundant loads being eliminated and in making static 10736 variables candidates for register promotion. 10737 + Improvement of RTL-based alias analysis. The results of type 10738 escape analysis are fed to the RTL type-based alias analyzer, 10739 allowing it to disambiguate more memory references. 10740 + Interprocedural constant propagation and function versioning. 10741 This pass looks for functions that are always called with the 10742 same constant value for one or more of the function arguments, 10743 and propagates those constants into those functions. 10744 + GCC will now eliminate static variables whose usage was 10745 optimized out. 10746 + -fwhole-program --combine can now be used to make all 10747 functions in program static allowing whole program 10748 optimization. As an exception, the main function and all 10749 functions marked with the new externally_visible attribute are 10750 kept global so that programs can link with runtime libraries. 10751 * GCC can now do a form of partial dead code elimination (PDCE) that 10752 allows code motion of expressions to the paths where the result of 10753 the expression is actually needed. This is not always a win, so the 10754 pass has been limited to only consider profitable cases. Here is an 10755 example: 10756 int foo (int *, int *); 10757 int 10758 bar (int d) 10759 { 10760 int a, b, c; 10761 b = d + 1; 10762 c = d + 2; 10763 a = b + c; 10764 if (d) 10765 { 10766 foo (&b, &c); 10767 a = b + c; 10768 } 10769 printf ("%d\n", a); 10770 } 10771 10772 The a = b + c can be sunk to right before the printf. Normal code 10773 sinking will not do this, it will sink the first one above into the 10774 else-branch of the conditional jump, which still gives you two 10775 copies of the code. 10776 * GCC now has a value range propagation pass. This allows the 10777 compiler to eliminate bounds checks and branches. The results of 10778 the pass can also be used to accurately compute branch 10779 probabilities. 10780 * The pass to convert PHI nodes to straight-line code (a form of 10781 if-conversion for GIMPLE) has been improved significantly. The two 10782 most significant improvements are an improved algorithm to 10783 determine the order in which the PHI nodes are considered, and an 10784 improvement that allow the pass to consider if-conversions of basic 10785 blocks with more than two predecessors. 10786 * Alias analysis improvements. GCC can now differentiate between 10787 different fields of structures in Tree-SSA's virtual operands form. 10788 This lets stores/loads from non-overlapping structure fields not 10789 conflict. A new algorithm to compute points-to sets was contributed 10790 that can allows GCC to see now that p->a and p->b, where p is a 10791 pointer to a structure, can never point to the same field. 10792 * Various enhancements to auto-vectorization: 10793 + Incrementally preserve SSA form when vectorizing. 10794 + Incrementally preserve loop-closed form when vectorizing. 10795 + Improvements to peeling for alignment: generate better code 10796 when the misalignment of an access is known at compile time, 10797 or when different accesses are known to have the same 10798 misalignment, even if the misalignment amount itself is 10799 unknown. 10800 + Consider dependence distance in the vectorizer. 10801 + Externalize generic parts of data reference analysis to make 10802 this analysis available to other passes. 10803 + Vectorization of conditional code. 10804 + Reduction support. 10805 * GCC can now partition functions in sections of hot and cold code. 10806 This can significantly improve performance due to better 10807 instruction cache locality. This feature works best together with 10808 profile feedback driven optimization. 10809 * A new pass to avoid saving of unneeded arguments to the stack in 10810 vararg functions if the compiler can prove that they will not be 10811 needed. 10812 * Transition of basic block profiling to tree level implementation 10813 has been completed. The new implementation should be considerably 10814 more reliable (hopefully avoiding profile mismatch errors when 10815 using -fprofile-use or -fbranch-probabilities) and can be used to 10816 drive higher level optimizations, such as inlining. 10817 The -ftree-based-profiling command-line option was removed and 10818 -fprofile-use now implies disabling old RTL level loop optimizer 10819 (-fno-loop-optimize). Speculative prefetching optimization 10820 (originally enabled by -fspeculative-prefetching) was removed. 10821 10822 New Languages and Language specific improvements 10823 10824 C and Objective-C 10825 10826 * The old Bison-based C and Objective-C parser has been replaced by a 10827 new, faster hand-written recursive-descent parser. 10828 10829 Ada 10830 10831 * The build infrastructure for the Ada runtime library and tools has 10832 been changed to be better integrated with the rest of the build 10833 infrastructure of GCC. This should make doing cross builds of Ada a 10834 bit easier. 10835 10836 C++ 10837 10838 * ARM-style name-injection of friend declarations is no longer the 10839 default. For example: 10840 struct S { 10841 friend void f(); 10842 }; 10843 10844 void g() { f(); } 10845 will not be accepted; instead a declaration of f will need to be 10846 present outside of the scope of S. The new -ffriend-injection 10847 option will enable the old behavior. 10848 * The (undocumented) extension which permitted templates with default 10849 arguments to be bound to template template parameters with fewer 10850 parameters has been deprecated, and will be removed in the next 10851 major release of G++. For example: 10852 template <template <typename> class C> 10853 void f(C<double>) {} 10854 10855 template <typename T, typename U = int> 10856 struct S {}; 10857 10858 template void f(S<double>); 10859 10860 makes use of the deprecated extension. The reason this code is not 10861 valid ISO C++ is that S is a template with two parameters; 10862 therefore, it cannot be bound to C which has only one parameter. 10863 10864 Runtime Library (libstdc++) 10865 10866 * Optimization work: 10867 + A new implementation of std::search_n is provided, better 10868 performing in case of random access iterators. 10869 + Added further efficient specializations of istream functions, 10870 i.e., character array and string extractors. 10871 + Other smaller improvements throughout. 10872 * Policy-based associative containers, designed for high-performance, 10873 flexibility and semantic safety are delivered in ext/pb_assoc. 10874 * A versatile string class, __gnu_cxx::__versa_string, providing 10875 facilities conforming to the standard requirements for 10876 basic_string, is delivered in <ext/vstring.h>. In particular: 10877 + Two base classes are provided: the default one avoids 10878 reference counting and is optimized for short strings; the 10879 alternate one, still uses it while improving in a few low 10880 level areas (e.g., alignment). See vstring_fwd.h for some 10881 useful typedefs. 10882 + Various algorithms have been rewritten (e.g., replace), the 10883 code streamlined and simple optimizations added. 10884 + Option 3 of DR 431 is implemented for both available bases, 10885 thus improving the support for stateful allocators. 10886 * As usual, many bugs have been fixed (e.g., libstdc++/13583, 10887 libstdc++/23953) and LWG resolutions put into effect for the first 10888 time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1 10889 Issue 6.19). The implementation status of TR1 is now tracked in the 10890 docs in tr1.html. 10891 10892 Objective-C++ 10893 10894 * A new language front end for Objective-C++ has been added. This 10895 language allows users to mix the object oriented features of 10896 Objective-C with those of C++. 10897 10898 Java (GCJ) 10899 10900 * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19 10901 features (plus some 0.20 bug-fixes) 10902 + Networking 10903 o The java.net.HttpURLConnection implementation no longer 10904 buffers the entire response body in memory. This means 10905 that response bodies larger than available memory can now 10906 be handled. 10907 + (N)IO 10908 o NIO FileChannel.map implementation, fast bulk put 10909 implementation for DirectByteBuffer (speeds up this 10910 method 10x). 10911 o FileChannel.lock() and FileChannel.force() implemented. 10912 + XML 10913 o gnu.xml fix for nodes created outside a namespace 10914 context. 10915 o Add support for output indenting and 10916 cdata-section-elements output instruction in 10917 xml.transform. 10918 o xml.xpath corrections for cases where elements/attributes 10919 might have been created in non-namespace-aware mode. 10920 Corrections to handling of XSL variables and minor 10921 conformance updates. 10922 + AWT 10923 o GNU JAWT implementation, the AWT Native Interface, which 10924 allows direct access to native screen resources from 10925 within a Canvas's paint method. GNU Classpath Examples 10926 comes with a Demo, see libjava/classpath/examples/README. 10927 o awt.datatransfer updated to 1.5 with support for 10928 FlavorEvents. The gtk+ awt peers now allow copy/paste of 10929 text, images, URIs/files and serialized objects with 10930 other applications and tracking clipboard change events 10931 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized 10932 objects are supported). A GNU Classpath Examples 10933 datatransfer Demo was added to show the new 10934 functionality. 10935 o Split gtk+ awt peers event handling in two threads and 10936 improve gdk lock handling (solves several awt lock ups). 10937 o Speed up awt Image loading. 10938 o Better gtk+ scrollbar peer implementation when using gtk+ 10939 >= 2.6. 10940 o Handle image loading errors correctly for gdkpixbuf and 10941 MediaTracker. 10942 o Better handle GDK lock. Properly prefix gtkpeer native 10943 functions (cp_gtk). 10944 o GdkGraphics2D has been updated to use Cairo 0.5.x or 10945 higher. 10946 o BufferedImage and GtkImage rewrites. All image drawing 10947 operations should now work correctly (flipping requires 10948 gtk+ >= 2.6) 10949 o Future Graphics2D, image and text work is documented at: 10950 [2]http://developer.classpath.org/mediation/ClasspathGrap 10951 hicsImagesText 10952 o When gtk+ 2.6 or higher is installed the default log 10953 handler will produce stack traces whenever a WARNING, 10954 CRITICAL or ERROR message is produced. 10955 + Free Swing 10956 o The RepaintManager has been reworked for more efficient 10957 painting, especially for large GUIs. 10958 o The layout manager OverlayLayout has been implemented, 10959 the BoxLayout has been rewritten to make use of the 10960 SizeRequirements utility class and caching for more 10961 efficient layout. 10962 o Improved accessibility support. 10963 o Significant progress has been made in the implementation 10964 of the javax.swing.plaf.metal package, with most UI 10965 delegates in a working state now. Please test this with 10966 your own applications and provide feedback that will help 10967 us to improve this package. 10968 o The GUI demo (gnu.classpath.examples.swing.Demo) has been 10969 extended to highlight various features in our Free Swing 10970 implementation. And it includes a look and feel switcher 10971 for Metal (default), Ocean and GNU themes. 10972 o The javax.swing.plaf.multi package is now implemented. 10973 o Editing and several key actions for JTree and JTable were 10974 implemented. 10975 o Lots of icons and look and feel improvements for Free 10976 Swing basic and metal themes were added. Try running the 10977 GNU Classpath Swing Demo in examples 10978 (gnu.classpath.examples.swing.Demo) with: 10979 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee 10980 l or 10981 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee 10982 l 10983 o Start of styled text capabilites for java.swing.text. 10984 o DefaultMutableTreeNode pre-order, post-order, depth-first 10985 and breadth-first traversal enumerations implemented. 10986 o JInternalFrame colors and titlebar draw properly. 10987 o JTree is working up to par (icons, selection and keyboard 10988 traversal). 10989 o JMenus were made more compatible in visual and 10990 programmatic behavior. 10991 o JTable changeSelection and multiple selections 10992 implemented. 10993 o JButton and JToggleButton change states work properly 10994 now. 10995 o JFileChooser fixes. 10996 o revalidate() and repaint() fixes which make Free Swing 10997 much more responsive. 10998 o MetalIconFactory implemented. 10999 o Free Swing Top-Level Compatibility. JFrame, JDialog, 11000 JApplet, JInternalFrame, and JWindow are now 1.5 11001 compatible in the sense that you can call add() and 11002 setLayout() directly on them, which will have the same 11003 effect as calling getContentPane().add() and 11004 getContentPane().setLayout(). 11005 o The JTree interface has been completed. JTrees now 11006 recognizes mouse clicks and selections work. 11007 o BoxLayout works properly now. 11008 o Fixed GrayFilter to actually work. 11009 o Metal SplitPane implemented. 11010 o Lots of Free Swing text and editor stuff work now. 11011 + Free RMI and Corba 11012 o Andrew Watson, Vice President and Technical Director of 11013 the Object Management Group, has officially assigned us 11014 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will 11015 mark remote classpath-specific system exceptions. 11016 Obtaining the VMCID means that GNU Classpath now is a 11017 recogniseable type of node in a highly interoperable 11018 CORBA world. 11019 o GNU Classpath now includes the first working draft to 11020 support the RMI over IIOP protocol. The current 11021 implementation is capable of remote invocations, 11022 transferring various Serializables and Externalizables 11023 via RMI-IIOP protocol. It can flatten graphs and, at 11024 least for the simple cases, is interoperable with 1.5 11025 JDKs. 11026 o org.omg.PortableInterceptor and related functionality in 11027 other packages is now implemented: 11028 # The sever and client interceptors work as required 11029 since 1.4. 11030 # The IOR interceptor works as needed for 1.5. 11031 o The org.omg.DynamicAny package is completed and passes 11032 the prepared tests. 11033 o The Portable Object Adapter should now support the output 11034 of the recent IDL to java compilers. These compilers now 11035 generate servants and not CORBA objects as before, making 11036 the output depend on the existing POA implementation. 11037 Completing POA means that such code can already be tried 11038 to run on Classpath. Our POA is tested for the following 11039 usager scenarios: 11040 # POA converts servant to the CORBA object. 11041 # Servant provides to the CORBA object. 11042 # POA activates new CORBA object with the given Object 11043 Id (byte array) that is later accessible for the 11044 servant. 11045 # During the first call, the ServantActivator provides 11046 servant for this and all subsequent calls on the 11047 current object. 11048 # During each call, the ServantLocator provides 11049 servant for this call only. 11050 # ServantLocator or ServantActivator forwards call to 11051 another server. 11052 # POA has a single servant, responsible for all 11053 objects. 11054 # POA has a default servant, but some objects are 11055 explicitly connected to they specific servants. 11056 The POA is verified using tests from the former 11057 cost.omg.org. 11058 o The CORBA implementation is now a working prototype that 11059 should support features up to 1.3 inclusive. We invite 11060 groups writing CORBA dependent applications to try 11061 Classpath implementation, reporting any possible bugs. 11062 The CORBA prototype is interoperable with Sun's 11063 implementation v 1.4, transferring object references, 11064 primitive types, narrow and wide strings, arrays, 11065 structures, trees, abstract interfaces and value types 11066 (feature of CORBA 2.3) between these two platforms. 11067 Remote exceptions are transferred and handled correctly. 11068 The stringified object references (IORs) from various 11069 sources are parsed as required. The transient (for 11070 current session) and permanent (till jre restart) 11071 redirections work. Both Little and Big Endian encoded 11072 messages are accepted. The implementation is verified 11073 using tests from the former cost.omg.org. The current 11074 release includes working examples (see the examples 11075 directory), demonstrating the client-server 11076 communication, using either CORBA Request or IDL-based 11077 stub (usually generated by a IDL to java compiler). These 11078 examples also show how to use the Classpath CORBA naming 11079 service. The IDL to java compiler is not yet written, but 11080 as our library must be compatible, it naturally accepts 11081 the output of other idlj implementations. 11082 + Misc 11083 o Updated TimeZone data against Olson tzdata2005l. 11084 o Make zip and jar packages UTF-8 clean. 11085 o "native" code builds and compiles (warning free) on 11086 Darwin and Solaris. 11087 o java.util.logging.FileHandler now rotates files. 11088 o Start of a generic JDWP framework in gnu/classpath/jdwp. 11089 This is unfinished, but feedback (at classpath (a] gnu.org) 11090 from runtime hackers is greatly appreciated. Although 11091 most of the work is currently being done around gcj/gij 11092 we want this framework to be as VM neutral as possible. 11093 Early design is described in: 11094 [3]https://gcc.gnu.org/ml/java/2005-05/msg00260.html 11095 o QT4 AWT peers, enable by giving configure 11096 --enable-qt-peer. Included, but not ready for production 11097 yet. They are explicitly disabled and not supported. But 11098 if you want to help with the development of these new 11099 features we are interested in feedback. You will have to 11100 explicitly enable them to try them out (and they will 11101 most likely contain bugs). 11102 o Documentation fixes all over the place. See 11103 [4]http://developer.classpath.org/doc/ 11104 11105 New Targets and Target Specific Improvements 11106 11107 IA-32/x86-64 11108 11109 * The x86-64 medium model (that allows building applications whose 11110 data segment exceeds 4GB) was redesigned to match latest ABI draft. 11111 New implementation split large datastructures into separate segment 11112 improving performance of accesses to small datastructures and also 11113 allows linking of small model libraries into medium model programs 11114 as long as the libraries are not accessing the large datastructures 11115 directly. Medium model is also supported in position independent 11116 code now. 11117 The ABI change results in partial incompatibility among medium 11118 model objects. Linking medium model libraries (or objects) compiled 11119 with new compiler into medium model program compiled with older 11120 will likely result in exceeding ranges of relocations. 11121 Binutils 2.16.91 or newer are required for compiling medium model 11122 now. 11123 11124 RS6000 (POWER/PowerPC) 11125 11126 * The AltiVec vector primitives in <altivec.h> are now implemented in 11127 a way that puts a smaller burden on the preprocessor, instead 11128 processing the "overloading" in the front ends. This should benefit 11129 compilation speed on AltiVec vector code. 11130 * AltiVec initializers now are generated more efficiently. 11131 * The popcountb instruction available on POWER5 now is generated. 11132 * The floating point round to integer instructions available on 11133 POWER5+ now is generated. 11134 * Floating point divides can be synthesized using the floating point 11135 reciprocal estimate instructions. 11136 * Double precision floating point constants are initialized as single 11137 precision values if they can be represented exactly. 11138 11139 S/390, zSeries and System z9 11140 11141 * Support for the IBM System z9 109 processor has been added. When 11142 using the -march=z9-109 option, the compiler will generate code 11143 making use of instructions provided by the extended immediate 11144 facility. 11145 * Support for 128-bit IEEE floating point has been added. When using 11146 the -mlong-double-128 option, the compiler will map the long double 11147 data type to 128-bit IEEE floating point. Using this option 11148 constitutes an ABI change, and requires glibc support. 11149 * Various changes to improve performance of generated code have been 11150 implemented, including: 11151 + In functions that do not require a literal pool, register %r13 11152 (which is traditionally reserved as literal pool pointer), can 11153 now be freely used for other purposes by the compiler. 11154 + More precise tracking of register use allows the compiler to 11155 generate more efficient function prolog and epilog code in 11156 certain cases. 11157 + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING 11158 instructions are now used to implement C string functions. 11159 + The MOVE CHARACTER instruction with single byte overlap is now 11160 used to implement the memset function with non-zero fill byte. 11161 + The LOAD ZERO instructions are now used where appropriate. 11162 + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK, 11163 and INSERT IMMEDIATE instructions are now used more frequently 11164 to optimize bitfield operations. 11165 + The BRANCH ON COUNT instruction is now used more frequently. 11166 In particular, the fact that a loop contains a subroutine call 11167 no longer prevents the compiler from using this instruction. 11168 + The compiler is now aware that all shift and rotate 11169 instructions implicitly truncate the shift count to six bits. 11170 * Back-end support for the following generic features has been 11171 implemented: 11172 + The full set of [5]built-in functions for atomic memory 11173 access. 11174 + The -fstack-protector feature. 11175 + The optimization pass avoiding unnecessary stores of incoming 11176 argument registers in functions with variable argument list. 11177 11178 SPARC 11179 11180 * The default code model in 64-bit mode has been changed from 11181 Medium/Anywhere to Medium/Middle on Solaris. 11182 * TLS support is disabled by default on Solaris prior to release 10. 11183 It can be enabled on TLS-capable Solaris 9 versions (4/04 release 11184 and later) by specifying --enable-tls at configure time. 11185 11186 MorphoSys 11187 11188 * Support has been added for this new architecture. 11189 11190 Obsolete Systems 11191 11192 Documentation improvements 11193 11194 Other significant improvements 11195 11196 * GCC can now emit code for protecting applications from 11197 stack-smashing attacks. The protection is realized by buffer 11198 overflow detection and reordering of stack variables to avoid 11199 pointer corruption. 11200 * Some built-in functions have been fortified to protect them against 11201 various buffer overflow (and format string) vulnerabilities. 11202 Compared to the mudflap bounds checking feature, the safe builtins 11203 have far smaller overhead. This means that programs built using 11204 safe builtins should not experience any measurable slowdown. 11205 11206 GCC 4.1.2 11207 11208 This is the [6]list of problem reports (PRs) from GCC's bug tracking 11209 system that are known to be fixed in the 4.1.2 release. This list might 11210 not be complete (that is, it is possible that some PRs that have been 11211 fixed are not listed here). 11212 11213 When generating code for a shared library, GCC now recognizes that 11214 global functions may be replaced when the program runs. Therefore, it 11215 is now more conservative in deducing information from the bodies of 11216 functions. For example, in this example: 11217 void f() {} 11218 void g() { 11219 try { f(); } 11220 catch (...) { 11221 cout << "Exception"; 11222 } 11223 } 11224 11225 G++ would previously have optimized away the catch clause, since it 11226 would have concluded that f cannot throw exceptions. Because users may 11227 replace f with another function in the main body of the program, this 11228 optimization is unsafe, and is no longer performed. If you wish G++ to 11229 continue to optimize as before, you must add a throw() clause to the 11230 declaration of f to make clear that it does not throw exceptions. 11231 11232 11233 For questions related to the use of GCC, please consult these web 11234 pages and the [7]GCC manuals. If that fails, the 11235 [8]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11236 web pages and the development of GCC are welcome on our developer 11237 list at [9]gcc (a] gcc.gnu.org. All of [10]our lists have public 11238 archives. 11239 11240 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and 11241 distribution of this entire article is permitted in any medium, 11242 provided this notice is preserved. 11243 11244 These pages are [12]maintained by the GCC team. Last modified 11245 2018-09-30[13]. 11246 11247 References 11248 11249 1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 11250 2. http://developer.classpath.org/mediation/ClasspathGraphicsImagesText 11251 3. https://gcc.gnu.org/ml/java/2005-05/msg00260.html 11252 4. http://developer.classpath.org/doc/ 11253 5. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html 11254 6. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2 11255 7. https://gcc.gnu.org/onlinedocs/ 11256 8. mailto:gcc-help (a] gcc.gnu.org 11257 9. mailto:gcc (a] gcc.gnu.org 11258 10. https://gcc.gnu.org/lists.html 11259 11. https://www.fsf.org/ 11260 12. https://gcc.gnu.org/about.html 11261 13. http://validator.w3.org/check/referer 11262 ====================================================================== 11263 http://gcc.gnu.org/gcc-4.0/index.html 11264 GCC 4.0 Release Series 11265 11266 January 31, 2007 11267 11268 The [1]GNU project and the GCC developers are pleased to announce the 11269 release of GCC 4.0.4. 11270 11271 This release is a bug-fix release, containing fixes for regressions in 11272 GCC 4.0.3 relative to previous releases of GCC. 11273 11274 This release series is no longer maintained. 11275 11276 Release History 11277 11278 GCC 4.0.4 11279 January 31, 2007 ([2]changes) 11280 11281 GCC 4.0.3 11282 March 10, 2006 ([3]changes) 11283 11284 GCC 4.0.2 11285 September 28, 2005 ([4]changes) 11286 11287 GCC 4.0.1 11288 July 7, 2005 ([5]changes) 11289 11290 GCC 4.0.0 11291 April 20, 2005 ([6]changes) 11292 11293 References and Acknowledgements 11294 11295 GCC used to stand for the GNU C Compiler, but since the compiler 11296 supports several other languages aside from C, it now stands for the 11297 GNU Compiler Collection. 11298 11299 A list of [7]successful builds is updated as new information becomes 11300 available. 11301 11302 The GCC developers would like to thank the numerous people that have 11303 contributed new features, improvements, bug fixes, and other changes as 11304 well as test results to GCC. This [8]amazing group of volunteers is 11305 what makes GCC successful. 11306 11307 For additional information about GCC please refer to the [9]GCC project 11308 web site or contact the [10]GCC development mailing list. 11309 11310 To obtain GCC please use [11]our mirror sites, or [12]our SVN server. 11311 11312 11313 For questions related to the use of GCC, please consult these web 11314 pages and the [13]GCC manuals. If that fails, the 11315 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11316 web pages and the development of GCC are welcome on our developer 11317 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 11318 archives. 11319 11320 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 11321 distribution of this entire article is permitted in any medium, 11322 provided this notice is preserved. 11323 11324 These pages are [18]maintained by the GCC team. Last modified 11325 2018-09-30[19]. 11326 11327 References 11328 11329 1. http://www.gnu.org/ 11330 2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 11331 3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3 11332 4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2 11333 5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1 11334 6. http://gcc.gnu.org/gcc-4.0/changes.html 11335 7. http://gcc.gnu.org/gcc-4.0/buildstat.html 11336 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 11337 9. http://gcc.gnu.org/index.html 11338 10. mailto:gcc (a] gcc.gnu.org 11339 11. http://gcc.gnu.org/mirrors.html 11340 12. http://gcc.gnu.org/svn.html 11341 13. https://gcc.gnu.org/onlinedocs/ 11342 14. mailto:gcc-help (a] gcc.gnu.org 11343 15. mailto:gcc (a] gcc.gnu.org 11344 16. https://gcc.gnu.org/lists.html 11345 17. https://www.fsf.org/ 11346 18. https://gcc.gnu.org/about.html 11347 19. http://validator.w3.org/check/referer 11348 ====================================================================== 11349 http://gcc.gnu.org/gcc-4.0/changes.html 11350 GCC 4.0 Release Series 11351 Changes, New Features, and Fixes 11352 11353 The latest release in the 4.0 release series is [1]GCC 4.0.4. 11354 11355 Caveats 11356 11357 * GCC now generates location lists by default when compiling with 11358 debug info and optimization. 11359 + GDB 6.0 and older crashes when it sees location lists. GDB 6.1 11360 or later is needed to debug binaries containing location 11361 lists. 11362 + When you are trying to view a value of a variable in a part of 11363 a function where it has no location (for example when the 11364 variable is no longer used and thus its location was used for 11365 something else) GDB will say that it is not available. 11366 You can disable generating location lists by -fno-var-tracking. 11367 * GCC no longer accepts the -fwritable-strings option. Use named 11368 character arrays when you need a writable string. 11369 * The options -freduce-all-givs and -fmove-all-movables have been 11370 discontinued. They were used to circumvent a shortcoming in the 11371 heuristics of the old loop optimization code with respect to common 11372 Fortran constructs. The new (tree) loop optimizer works differently 11373 and doesn't need those work-arounds. 11374 * The graph-coloring register allocator, formerly enabled by the 11375 option -fnew-ra, has been discontinued. 11376 * -I- has been deprecated. -iquote is meant to replace the need for 11377 this option. 11378 * The MIPS -membedded-pic and -mrnames options have been removed. 11379 * All MIPS targets now require the GNU assembler. In particular, IRIX 11380 configurations can no longer use the MIPSpro assemblers, although 11381 they do still support the MIPSpro linkers. 11382 * The SPARC option -mflat has been removed. 11383 * English-language diagnostic messages will now use Unicode quotation 11384 marks in UTF-8 locales. (Non-English messages already used the 11385 quotes appropriate for the language in previous releases.) If your 11386 terminal does not support UTF-8 but you are using a UTF-8 locale 11387 (such locales are the default on many GNU/Linux systems) then you 11388 should set LC_CTYPE=C in the environment to disable that locale. 11389 Programs that parse diagnostics and expect plain ASCII 11390 English-language messages should set LC_ALL=C. See [2]Markus Kuhn's 11391 explanation of Unicode quotation marks for more information. 11392 * The specs file is no longer installed on most platforms. Most users 11393 will be totally unaffected. However, if you are accustomed to 11394 editing the specs file yourself, you will now have to use the 11395 -dumpspecs option to generate the specs file, and then edit the 11396 resulting file. 11397 11398 General Optimizer Improvements 11399 11400 * The [3]tree ssa branch has been merged. This merge has brought in a 11401 completely new optimization framework based on a higher level 11402 intermediate representation than the existing RTL representation. 11403 Numerous new code transformations based on the new framework are 11404 available in GCC 4.0, including: 11405 + Scalar replacement of aggregates 11406 + Constant propagation 11407 + Value range propagation 11408 + Partial redundancy elimination 11409 + Load and store motion 11410 + Strength reduction 11411 + Dead store elimination 11412 + Dead and unreachable code elimination 11413 + [4]Autovectorization 11414 + Loop interchange 11415 + Tail recursion by accumulation 11416 Many of these passes outperform their counterparts from previous 11417 GCC releases. 11418 * [5]Swing Modulo Scheduling (SMS). An RTL level instruction 11419 scheduling optimization intended for loops that perform heavy 11420 computations. 11421 11422 New Languages and Language specific improvements 11423 11424 C family 11425 11426 * The sentinel attribute has been added to GCC. This function 11427 attribute allows GCC to warn when variadic functions such as execl 11428 are not NULL terminated. See the GCC manual for a complete 11429 description of its behavior. 11430 * Given __attribute__((alias("target"))) it is now an error if target 11431 is not a symbol, defined in the same translation unit. This also 11432 applies to aliases created by #pragma weak alias=target. This is 11433 because it's meaningless to define an alias to an undefined symbol. 11434 On Solaris, the native assembler would have caught this error, but 11435 GNU as does not. 11436 11437 C and Objective-C 11438 11439 * The -Wstrict-aliasing=2 option has been added. This warning catches 11440 all unsafe cases, but it may also give a warning for some cases 11441 that are safe. 11442 * The cast-as-lvalue, conditional-expression-as-lvalue and 11443 compound-expression-as-lvalue extensions, which were deprecated in 11444 3.3.4 and 3.4, have been removed. 11445 * The -fwritable-strings option, which was deprecated in 3.4, has 11446 been removed. 11447 * #pragma pack() semantics have been brought closer to those used by 11448 other compilers. This also applies to C++. 11449 * Taking the address of a variable with register storage is invalid 11450 in C. GCC now issues an error instead of a warning. 11451 * Arrays of incomplete element type are invalid in C. GCC now issues 11452 an error for such arrays. Declarations such as extern struct s x[]; 11453 (where struct s has not been defined) can be moved after the 11454 definition of struct s. Function parameters declared as arrays of 11455 incomplete type can instead be declared as pointers. 11456 11457 C++ 11458 11459 * When compiling without optimizations (-O0), the C++ front end is 11460 much faster than in any previous versions of GCC. Independent 11461 testers have measured speed-ups up to 25% in real-world production 11462 code, compared to the 3.4 family (which was already the fastest 11463 version to date). Upgrading from older versions might show even 11464 bigger improvements. 11465 * ELF visibility attributes can now be applied to a class type, so 11466 that it affects every member function of a class at once, without 11467 having to specify each individually: 11468 class __attribute__ ((visibility("hidden"))) Foo 11469 { 11470 int foo1(); 11471 void foo2(); 11472 }; 11473 The syntax is deliberately similar to the __declspec() system used 11474 by Microsoft Windows based compilers, allowing cross-platform 11475 projects to easily reuse their existing macro system for denoting 11476 exports and imports. By explicitly marking internal classes never 11477 used outside a binary as hidden, one can completely avoid PLT 11478 indirection overheads during their usage by the compiler. You can 11479 find out more about the advantages of this at 11480 [6]https://www.akkadia.org/drepper/dsohowto.pdf 11481 * The -fvisibility-inlines-hidden option has been added which marks 11482 all inlineable functions as having hidden ELF visibility, thus 11483 removing their symbol and typeinfo from the exported symbol table 11484 of the output ELF binary. Using this option can reduce the exported 11485 symbol count of template-heavy code by up to 40% with no code 11486 change at all, thus notably improving link and load times for the 11487 binary as well as a reduction in size of up to 10%. Also, check the 11488 new [7]-fvisibility option. 11489 * The compiler now uses the library interface specified by the [8]C++ 11490 ABI for thread-safe initialization of function-scope static 11491 variables. Most users should leave this alone, but embedded 11492 programmers may want to disable this by specifying 11493 -fno-threadsafe-statics for a small savings in code size. 11494 * Taking the address of an explicit register variable is no longer 11495 supported. Note that C++ allows taking the address of variables 11496 with register storage so this will continue to compile with a 11497 warning. For example, assuming that r0 is a machine register: 11498 register int foo asm ("r0"); 11499 register int bar; 11500 &foo; // error, no longer accepted 11501 &bar; // OK, with a warning 11502 * G++ has an undocumented extension to virtual function covariancy 11503 rules that allowed the overrider to return a type that was 11504 implicitly convertable to the overridden function's return type. 11505 For instance a function returning void * could be overridden by a 11506 function returning T *. This is now deprecated and will be removed 11507 in a future release. 11508 * The G++ minimum and maximum operators (<? and >?) and their 11509 compound forms (<?=) and >?=) have been deprecated and will be 11510 removed in a future version. Code using these operators should be 11511 modified to use std::min and std::max instead. 11512 * Declaration of nested classes of class templates as friends are 11513 supported: 11514 template <typename T> struct A { 11515 class B {}; 11516 }; 11517 class C { 11518 template <typename T> friend class A<T>::B; 11519 }; 11520 This complements the feature member functions of class templates as 11521 friends introduced in GCC 3.4.0. 11522 * When declaring a friend class using an unqualified name, classes 11523 outside the innermost non-class scope are not searched: 11524 class A; 11525 namespace N { 11526 class B { 11527 friend class A; // Refer to N::A which has not been declared yet 11528 // because name outside namespace N are not searched 11529 friend class ::A; // Refer to ::A 11530 }; 11531 } 11532 Hiding the friend name until declaration is still not implemented. 11533 * Friends of classes defined outside their namespace are correctly 11534 handled: 11535 namespace N { 11536 class A; 11537 } 11538 class N::A { 11539 friend class B; // Refer to N::B in GCC 4.0.0 11540 // but ::B in earlier versions of GCC 11541 }; 11542 11543 Runtime Library (libstdc++) 11544 11545 * Optimization work: 11546 + Added efficient specializations of istream functions for char 11547 and wchar_t. 11548 + Further performance tuning of strings, in particular wrt 11549 single-char append and getline. 11550 + iter_swap - and therefore most of the mutating algorithms - 11551 now makes an unqualified call to swap when the value_type of 11552 the two iterators is the same. 11553 * A large subset of the features in Technical Report 1 (TR1 for 11554 short) is experimentally delivered (i.e., no guarantees about the 11555 implementation are provided. In particular it is not promised that 11556 the library will remain link-compatible when code using TR1 is 11557 used): 11558 + General utilities such as reference_wrapper and shared_ptr. 11559 + Function objects, i.e., result_of, mem_fn, bind, function. 11560 + Support for metaprogramming. 11561 + New containers such as tuple, array, unordered_set, 11562 unordered_map, unordered_multiset, unordered_multimap. 11563 * As usual, many bugs have been fixed and LWG resolutions implemented 11564 for the first time (e.g., DR 409). 11565 11566 Java 11567 11568 * In order to prevent naming conflicts with other implementations of 11569 these tools, some GCJ binaries have been renamed: 11570 + rmic is now grmic, 11571 + rmiregistry is now grmiregistry, and 11572 + jar is now fastjar. 11573 In particular, these names were problematic for the jpackage.org 11574 packaging conventions which install symlinks in /usr/bin that point 11575 to the preferred versions of these tools. 11576 * The -findirect-dispatch argument to the compiler now works and 11577 generates code following a new "binary compatibility" ABI. Code 11578 compiled this way follows the binary compatibility rules of the 11579 Java Language Specification. 11580 * libgcj now has support for using GCJ as a JIT, using the 11581 gnu.gcj.jit family of system properties. 11582 * libgcj can now find a shared library corresponding to the bytecode 11583 representation of a class. See the documentation for the new 11584 gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system 11585 property. 11586 * There have been many improvements to the class library. Here are 11587 some highlights: 11588 + Much more of AWT and Swing exist. 11589 + Many new packages and classes were added, including 11590 java.util.regex, java.net.URI, javax.crypto, 11591 javax.crypto.interfaces, javax.crypto.spec, javax.net, 11592 javax.net.ssl, javax.security.auth, 11593 javax.security.auth.callback, javax.security.auth.login, 11594 javax.security.auth.x500, javax.security.sasl, org.ietf.jgss, 11595 javax.imageio, javax.imageio.event, javax.imageio.spi, 11596 javax.print, javax.print.attribute, 11597 javax.print.attribute.standard, javax.print.event, and 11598 javax.xml 11599 + Updated SAX and DOM, and imported GNU JAXP 11600 11601 Fortran 11602 11603 * A new [9]Fortran front end has replaced the aging GNU Fortran 77 11604 front end. The new front end supports Fortran 90 and Fortran 95. It 11605 may not yet be as stable as the old Fortran front end. 11606 11607 Ada 11608 11609 * Ada (with tasking and Zero Cost Exceptions) is now available on 11610 many more targets, including but not limited to: alpha-linux, 11611 hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux, 11612 s390x-linux, sparc-linux. 11613 * Some of the new Ada 2005 features are now implemented like 11614 Wide_Wide_Character and Ada.Containers. 11615 * Many bugs have been fixed, tools and documentation improved. 11616 * To compile Ada from the sources, install an older working Ada 11617 compiler and then use --enable-languages=ada at configuration time, 11618 since the Ada front end is not currently activated by default. See 11619 the [10]Installing GCC for details. 11620 11621 New Targets and Target Specific Improvements 11622 11623 H8/300 11624 11625 * The frame layout has changed. In the new layout, the prologue of a 11626 function first saves registers and then allocate space for locals, 11627 resulting in an 1% improvement on code size. 11628 11629 IA-32/x86-64 (AMD64) 11630 11631 * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10, 11632 log1p, log2, logb and tan mathematical builtins (and their float 11633 and long double variants) are now implemented as inline x87 11634 intrinsics when using -ffast-math. 11635 * The ceil, floor, nearbyint, rint and trunc mathematical builtins 11636 (and their float and long double variants) are now implemented as 11637 inline x87 intrinsics when using -ffast-math. 11638 * The x87's fsincos instruction is now used automatically with 11639 -ffast-math when calculating both the sin and cos of the same 11640 argument. 11641 * Instruction selection for multiplication and division by constants 11642 has been improved. 11643 11644 IA-64 11645 11646 * Floating point division, integer division and sqrt are now inlined, 11647 resulting in significant performance improvements on some codes. 11648 11649 MIPS 11650 11651 * Division by zero checks now use conditional traps if the target 11652 processor supports them. This decreases code size by one word per 11653 division operation. The old behavior (branch and break) can be 11654 obtained either at configure time by passing --with-divide=breaks 11655 to configure or at runtime by passing -mdivide-breaks to GCC. 11656 * Support for MIPS64 paired-single instructions has been added. It is 11657 enabled by -mpaired-single and can be accessed using both the 11658 target-independent vector extensions and new MIPS-specific built-in 11659 functions. 11660 * Support for the MIPS-3D ASE has been added. It is enabled by 11661 -mips3d and provides new MIPS-3D-specific built-in functions. 11662 * The -mexplicit-relocs option now supports static n64 code (as is 11663 used, for example, in 64-bit linux kernels). -mexplicit-relocs 11664 should now be feature-complete and is enabled by default when GCC 11665 is configured to use a compatible assembler. 11666 * Support for the NEC VR4130 series has been added. This support 11667 includes the use of VR-specific instructions and a new VR4130 11668 scheduler. Full VR4130 support can be selected with -march=vr4130 11669 while code for any ISA can be tuned for the VR4130 using 11670 -mtune=vr4130. There is also a new -mvr4130-align option that 11671 produces better schedules at the cost of increased code size. 11672 * Support for the Broadcom SB-1 has been extended. There is now an 11673 SB-1 scheduler as well as support for the SB-1-specific 11674 paired-single instructions. Full SB-1 support can be selected with 11675 -march=sb1 while code for any ISA can be optimized for the SB-1 11676 using -mtune=sb1. 11677 * The compiler can now work around errata in R4000, R4400, VR4120 and 11678 VR4130 processors. These workarounds are enabled by -mfix-r4000, 11679 -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120 11680 and VR4130 workarounds need binutils 2.16 or above. 11681 * IRIX shared libraries are now installed into the standard library 11682 directories: o32 libraries go into lib/, n32 libraries go into 11683 lib32/ and n64 libraries go into lib64/. 11684 * The compiler supports a new -msym32 option. It can be used to 11685 optimize n64 code in which all symbols are known to have 32-bit 11686 values. 11687 11688 S/390 and zSeries 11689 11690 * New command-line options help to generate code intended to run in 11691 an environment where stack space is restricted, e.g. Linux kernel 11692 code: 11693 + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time 11694 warnings for single functions that require large or dynamic 11695 stack frames. 11696 + -mstack-size and -mstack-guard generate code that checks for 11697 stack overflow at run time. 11698 + -mpacked-stack generates code that reduces the stack frame 11699 size of many functions by reusing unneeded parts of the stack 11700 bias area. 11701 * The -msoft-float option now ensures that generated code never 11702 accesses floating point registers. 11703 * The s390x-ibm-tpf target now fully supports C++, including 11704 exceptions and threads. 11705 * Various changes to improve performance of the generated code have 11706 been implemented, including: 11707 + GCC now uses sibling calls where possible. 11708 + Condition code handling has been optimized, allowing GCC to 11709 omit redundant comparisons in certain cases. 11710 + The cost function guiding many optimizations has been refined 11711 to more accurately represent the z900 and z990 processors. 11712 + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW 11713 instructions are now used to avoid conditional branches in 11714 certain cases. 11715 + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to 11716 optimize address arithmetic required to access large stack 11717 frames. 11718 + GCC now makes more efficient use of memory-to-memory type 11719 instructions (MVC, CLC, ...). 11720 + More precise tracking of special register use allows better 11721 instruction scheduling, in particular of the function prologue 11722 and epilogue sequences. 11723 + The Java front end now generates inline code to implement 11724 integer division, instead of calling library routines. 11725 11726 SPARC 11727 11728 * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and 11729 -mf934 have been removed. They have been replaced with -mcpu=xxx. 11730 * The internal model used to estimate the relative cost of each 11731 instruction has been updated. It is expected to give better results 11732 on recent UltraSPARC processors. 11733 * Code generation for function prologues and epilogues has been 11734 improved, resulting in better scheduling and allowing multiple exit 11735 points in functions. 11736 * Support for Sun's Visual Instruction Set (VIS) has been enhanced. 11737 It is enabled by -mvis and provides new built-in functions for VIS 11738 instructions on UltraSPARC processors. 11739 * The option -mapp-regs has been turned on by default on Solaris too. 11740 11741 NetWare 11742 11743 * Novell NetWare (on ix86, no other hardware platform was ever really 11744 supported by this OS) has been re-enabled and the ABI supported by 11745 GCC has been brought into sync with that of MetroWerks CodeWarrior 11746 (the ABI previously supported was that of some Unix systems, which 11747 NetWare never tried to support). 11748 11749 Obsolete Systems 11750 11751 Support for a number of older systems has been declared obsolete in GCC 11752 4.0. Unless there is activity to revive them, the next release of GCC 11753 will have their sources permanently removed. 11754 11755 All GCC ports for the following processor architectures have been 11756 declared obsolete: 11757 * Intel i860 11758 * Ubicom IP2022 11759 * National Semiconductor NS32K (ns32k) 11760 * Texas Instruments TMS320C[34]x 11761 11762 Also, those for some individual systems have been obsoleted: 11763 * SPARC family 11764 + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf, 11765 sparc86x-*-elf) 11766 + OpenBSD 32-bit (sparc-*-openbsd*) 11767 11768 Documentation improvements 11769 11770 Other significant improvements 11771 11772 * Location lists are now generated by default when compiling with 11773 debug info and optimization. Location lists provide more accurate 11774 debug info about locations of variables and they allow debugging 11775 code compiled with -fomit-frame-pointer. 11776 * The -fvisibility option has been added which allows the default ELF 11777 visibility of all symbols to be set per compilation and the new 11778 #pragma GCC visibility preprocessor command allows the setting of 11779 default ELF visibility for a region of code. Using 11780 -fvisibility=hidden especially in combination with the new 11781 -fvisibility-inlines-hidden can yield substantial improvements in 11782 output binary quality including avoiding PLT indirection overheads, 11783 reduction of the exported symbol count by up to 60% (with resultant 11784 improvements to link and load times), better scope for the 11785 optimizer to improve code and up to a 20% reduction in binary size. 11786 Using these options correctly yields a binary with a similar symbol 11787 count to a Windows DLL. 11788 Perhaps more importantly, this new feature finally allows (with 11789 careful planning) complete avoidance of symbol clashes when 11790 manually loading shared objects with RTLD_GLOBAL, thus finally 11791 solving problems many projects such as python were forced to use 11792 RTLD_LOCAL for (with its resulting issues for C++ correctness). You 11793 can find more information about using these options at 11794 [11]https://gcc.gnu.org/wiki/Visibility. 11795 __________________________________________________________________ 11796 11797 GCC 4.0.1 11798 11799 This is the [12]list of problem reports (PRs) from GCC's bug tracking 11800 system that are known to be fixed in the 4.0.1 release. This list might 11801 not be complete (that is, it is possible that some PRs that have been 11802 fixed are not listed here). 11803 11804 GCC 4.0.2 11805 11806 This is the [13]list of problem reports (PRs) from GCC's bug tracking 11807 system that are known to be fixed in the 4.0.2 release. This list might 11808 not be complete (that is, it is possible that some PRs that have been 11809 fixed are not listed here). 11810 11811 Unfortunately, due to a release engineering failure, this release has a 11812 regression on Solaris that will affect some C++ programs. We suggest 11813 that Solaris users apply a [14]patch that corrects the problem. Users 11814 who do not wish to apply the patch should explicitly link C++ programs 11815 with the -pthreads option, even if they do not use threads. This 11816 problem has been corrected in the current 4.0 branch sources and will 11817 not be present in GCC 4.0.3. 11818 11819 GCC 4.0.3 11820 11821 Starting with this release, the function getcontext is recognized by 11822 the compiler as having the same semantics as the setjmp function. In 11823 particular, the compiler will ensure that all registers are dead before 11824 calling such a function and will emit a warning about the variables 11825 that may be clobbered after the second return from the function. 11826 11827 GCC 4.0.4 11828 11829 This is the [15]list of problem reports (PRs) from GCC's bug tracking 11830 system that are known to be fixed in the 4.0.4 release. This list might 11831 not be complete (that is, it is possible that some PRs that have been 11832 fixed are not listed here). 11833 11834 The 4.0.4 release is provided for those that require a high degree of 11835 binary compatibility with previous 4.0.x releases. For most users, the 11836 GCC team recommends that version 4.1.1 or later be used instead." 11837 11838 11839 For questions related to the use of GCC, please consult these web 11840 pages and the [16]GCC manuals. If that fails, the 11841 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11842 web pages and the development of GCC are welcome on our developer 11843 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 11844 archives. 11845 11846 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 11847 distribution of this entire article is permitted in any medium, 11848 provided this notice is preserved. 11849 11850 These pages are [21]maintained by the GCC team. Last modified 11851 2018-09-30[22]. 11852 11853 References 11854 11855 1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 11856 2. https://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html 11857 3. http://gcc.gnu.org/projects/tree-ssa/ 11858 4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html 11859 5. http://gcc.gnu.org/news/sms.html 11860 6. https://www.akkadia.org/drepper/dsohowto.pdf 11861 7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility 11862 8. https://itanium-cxx-abi.github.io/cxx-abi/ 11863 9. http://gcc.gnu.org/fortran/ 11864 10. https://gcc.gnu.org/install/ 11865 11. https://gcc.gnu.org/wiki/Visibility 11866 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1 11867 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2 11868 14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html 11869 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4 11870 16. https://gcc.gnu.org/onlinedocs/ 11871 17. mailto:gcc-help (a] gcc.gnu.org 11872 18. mailto:gcc (a] gcc.gnu.org 11873 19. https://gcc.gnu.org/lists.html 11874 20. https://www.fsf.org/ 11875 21. https://gcc.gnu.org/about.html 11876 22. http://validator.w3.org/check/referer 11877 ====================================================================== 11878 http://gcc.gnu.org/gcc-3.4/index.html 11879 GCC 3.4 Release Series 11880 11881 May 26, 2006 11882 11883 The [1]GNU project and the GCC developers are pleased to announce the 11884 release of GCC 3.4.6. 11885 11886 This release is a bug-fix release, containing fixes for regressions in 11887 GCC 3.4.4 relative to previous releases of GCC. This is the last of the 11888 3.4.x series. 11889 11890 The GCC 3.4 release series includes numerous [2]new features, 11891 improvements, bug fixes, and other changes, thanks to an [3]amazing 11892 group of volunteers. 11893 11894 This release series is no longer maintained. 11895 11896 Release History 11897 11898 GCC 3.4.6 11899 March 6, 2006 ([4]changes) 11900 11901 GCC 3.4.5 11902 November 30, 2005 ([5]changes) 11903 11904 GCC 3.4.4 11905 May 18, 2005 ([6]changes) 11906 11907 GCC 3.4.3 11908 November 4, 2004 ([7]changes) 11909 11910 GCC 3.4.2 11911 September 6, 2004 ([8]changes) 11912 11913 GCC 3.4.1 11914 July 1, 2004 ([9]changes) 11915 11916 GCC 3.4.0 11917 April 18, 2004 ([10]changes) 11918 11919 References and Acknowledgements 11920 11921 GCC used to stand for the GNU C Compiler, but since the compiler 11922 supports several other languages aside from C, it now stands for the 11923 GNU Compiler Collection. 11924 11925 A list of [11]successful builds is updated as new information becomes 11926 available. 11927 11928 The GCC developers would like to thank the numerous people that have 11929 contributed new features, improvements, bug fixes, and other changes as 11930 well as test results to GCC. This [12]amazing group of volunteers is 11931 what makes GCC successful. 11932 11933 For additional information about GCC please refer to the [13]GCC 11934 project web site or contact the [14]GCC development mailing list. 11935 11936 To obtain GCC please use [15]our mirror sites, or [16]our SVN server. 11937 11938 11939 For questions related to the use of GCC, please consult these web 11940 pages and the [17]GCC manuals. If that fails, the 11941 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 11942 web pages and the development of GCC are welcome on our developer 11943 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 11944 archives. 11945 11946 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 11947 distribution of this entire article is permitted in any medium, 11948 provided this notice is preserved. 11949 11950 These pages are [22]maintained by the GCC team. Last modified 11951 2018-09-30[23]. 11952 11953 References 11954 11955 1. http://www.gnu.org/ 11956 2. http://gcc.gnu.org/gcc-3.4/changes.html 11957 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 11958 4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 11959 5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5 11960 6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4 11961 7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3 11962 8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2 11963 9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1 11964 10. http://gcc.gnu.org/gcc-3.4/changes.html 11965 11. http://gcc.gnu.org/gcc-3.4/buildstat.html 11966 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 11967 13. http://gcc.gnu.org/index.html 11968 14. mailto:gcc (a] gcc.gnu.org 11969 15. http://gcc.gnu.org/mirrors.html 11970 16. http://gcc.gnu.org/svn.html 11971 17. https://gcc.gnu.org/onlinedocs/ 11972 18. mailto:gcc-help (a] gcc.gnu.org 11973 19. mailto:gcc (a] gcc.gnu.org 11974 20. https://gcc.gnu.org/lists.html 11975 21. https://www.fsf.org/ 11976 22. https://gcc.gnu.org/about.html 11977 23. http://validator.w3.org/check/referer 11978 ====================================================================== 11979 http://gcc.gnu.org/gcc-3.4/changes.html 11980 GCC 3.4 Release Series 11981 Changes, New Features, and Fixes 11982 11983 The final release in the 3.4 release series is [1]GCC 3.4.6. The series 11984 is now closed. 11985 11986 GCC 3.4 has [2]many improvements in the C++ front end. Before reporting 11987 a bug, please make sure it's really GCC, and not your code, that is 11988 broken. 11989 11990 Caveats 11991 11992 * GNU Make is now required to build GCC. 11993 * With -nostdinc the preprocessor used to ignore both standard 11994 include paths and include paths contained in environment variables. 11995 It was neither documented nor intended that environment variable 11996 paths be ignored, so this has been corrected. 11997 * GCC no longer accepts the options -fvolatile, -fvolatile-global and 11998 -fvolatile-static. It is unlikely that they worked correctly in any 11999 3.x release. 12000 * GCC no longer ships <varargs.h>. Use <stdarg.h> instead. 12001 * Support for all the systems [3]obsoleted in GCC 3.3 has been 12002 removed from GCC 3.4. See below for a [4]list of systems which are 12003 obsoleted in this release. 12004 * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C 12005 compilers will not work. 12006 * The implementation of the [5]MIPS ABIs has changed. As a result, 12007 the code generated for certain MIPS targets will not be binary 12008 compatible with earlier releases. 12009 * In previous releases, the MIPS port had a fake "hilo" register with 12010 the user-visible name accum. This register has been removed. 12011 * The implementation of the [6]SPARC ABIs has changed. As a result, 12012 the code generated will not be binary compatible with earlier 12013 releases in certain cases. 12014 * The configure option --enable-threads=pthreads has been removed; 12015 use --enable-threads=posix instead, which should have the same 12016 effect. 12017 * Code size estimates used by inlining heuristics for C, Objective-C, 12018 C++ and Java have been redesigned significantly. As a result the 12019 parameters of -finline-insns, --param max-inline-insns-single and 12020 --param max-inline-insns-auto need to be reconsidered. 12021 * --param max-inline-slope and --param min-inline-insns have been 12022 removed; they are not needed for the new bottom-up inlining 12023 heuristics. 12024 * The new unit-at-a-time compilation scheme has several compatibility 12025 issues: 12026 + The order in which functions, variables, and top-level asm 12027 statements are emitted may have changed. Code relying on some 12028 particular ordering needs to be updated. The majority of such 12029 top-level asm statements can be replaced by section 12030 attributes. 12031 + Unreferenced static variables and functions are removed. This 12032 may result in undefined references when an asm statement 12033 refers to the variable/function directly. In that case either 12034 the variable/function shall be listed in asm statement operand 12035 or in the case of top-level asm statements the attribute used 12036 shall be used to force function/variable to be always output 12037 and considered as a possibly used by unknown code. 12038 For variables the attribute is accepted only by GCC 3.4 and 12039 newer, while for earlier versions it is sufficient to use 12040 unused to silence warnings about the variables not being 12041 referenced. To keep code portable across different GCC 12042 versions, you can use appropriate preprocessor conditionals. 12043 + Static functions now can use non-standard passing conventions 12044 that may break asm statements calling functions directly. 12045 Again the attribute used shall be used to prevent this 12046 behavior. 12047 As a temporary workaround, -fno-unit-at-a-time can be used, but 12048 this scheme may not be supported by future releases of GCC. 12049 * GCC 3.4 automatically places zero-initialized variables in the .bss 12050 section on some operating systems. Versions of GNU Emacs up to (and 12051 including) 21.3 will not work correctly when using this 12052 optimization; you can use -fno-zero-initialized-in-bss to disable 12053 it. 12054 * If GCC 3.4 is configured with --enable-threads=posix (the default 12055 on most targets that support pthreads) then _REENTRANT will be 12056 defined unconditionally by some libstdc++ headers. C++ code which 12057 relies on that macro to detect whether multi-threaded code is being 12058 compiled might change in meaning, possibly resulting in linker 12059 errors for single-threaded programs. Affected users of [7]Boost 12060 should compile single-threaded code with -DBOOST_DISABLE_THREADS. 12061 See Bugzilla for [8]more information. 12062 12063 General Optimizer Improvements 12064 12065 * Usability of the profile feedback and coverage testing has been 12066 improved. 12067 + Performance of profiled programs has been improved by faster 12068 profile merging code. 12069 + Better use of the profile feedback for optimization (loop 12070 unrolling and loop peeling). 12071 + File locking support allowing fork() calls and parallel runs 12072 of profiled programs. 12073 + Coverage file format has been redesigned. 12074 + gcov coverage tool has been improved. 12075 + make profiledbootstrap available to build a faster compiler. 12076 Experiments made on i386 hardware showed an 11% speedup on -O0 12077 and a 7.5% speedup on -O2 compilation of a [9]large C++ 12078 testcase. 12079 + New value profiling pass enabled via -fprofile-values 12080 + New value profile transformations pass enabled via -fvpt aims 12081 to optimize some code sequences by exploiting knowledge about 12082 value ranges or other properties of the operands. At the 12083 moment a conversion of expensive divisions into cheaper 12084 operations has been implemented. 12085 + New -fprofile-generate and -fprofile-use command-line options 12086 to simplify the use of profile feedback. 12087 * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and 12088 Java which is enabled via -funit-at-a-time (and implied by -O2). In 12089 this scheme a whole file is parsed first and optimized later. The 12090 following basic inter-procedural optimizations are implemented: 12091 + Removal of unreachable functions and variables 12092 + Discovery of local functions (functions with static linkage 12093 whose address is never taken) 12094 + On i386, these local functions use register parameter passing 12095 conventions. 12096 + Reordering of functions in topological order of the call graph 12097 to enable better propagation of optimizing hints (such as the 12098 stack alignments needed by functions) in the back end. 12099 + Call graph based out-of-order inlining heuristics which allows 12100 to limit overall compilation unit growth (--param 12101 inline-unit-growth). 12102 Overall, the unit-at-a-time scheme produces a 1.3% improvement for 12103 the SPECint2000 benchmark on the i386 architecture (AMD Athlon 12104 CPU). 12105 * More realistic code size estimates used by inlining for C, 12106 Objective-C, C++ and Java. The growth of large functions can now be 12107 limited via --param large-function-insns and --param 12108 large-function-growth. 12109 * A new cfg-level loop optimizer pass replaces the old loop unrolling 12110 pass and adds two other loop transformations -- loop peeling and 12111 loop unswitching -- and also uses the profile feedback to limit 12112 code growth. (The three optimizations are enabled by 12113 -funroll-loops, -fpeel-loops and -funswitch-loops flags, 12114 respectively). 12115 The old loop unroller still can be enabled by -fold-unroll-loops 12116 and may produce better code in some cases, especially when the 12117 webizer optimization pass is not run. 12118 * A new web construction pass enabled via -fweb (and implied by -O3) 12119 improves the quality of register allocation, CSE, first scheduling 12120 pass and some other optimization passes by avoiding re-use of 12121 pseudo registers with non-overlapping live ranges. The pass almost 12122 always improves code quality but does make debugging difficult and 12123 thus is not enabled by default by -O2 12124 The pass is especially effective as cleanup after code duplication 12125 passes, such as the loop unroller or the tracer. 12126 * Experimental implementations of superblock or trace scheduling in 12127 the second scheduling pass can be enabled via 12128 -fsched2-use-superblocks and -fsched2-use-traces, respectively. 12129 12130 New Languages and Language specific improvements 12131 12132 Ada 12133 12134 * The Ada front end has been updated to include numerous bug fixes 12135 and enhancements. These include: 12136 + Improved project file support 12137 + Additional set of warnings about potential wrong code 12138 + Improved error messages 12139 + Improved code generation 12140 + Improved cross reference information 12141 + Improved inlining 12142 + Better run-time check elimination 12143 + Better error recovery 12144 + More efficient implementation of unbounded strings 12145 + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools, 12146 ... 12147 + New GNAT.xxxx packages (e.g. GNAT.Strings, 12148 GNAT.Exception_Action) 12149 + New pragmas 12150 + New -gnatS switch replacing gnatpsta 12151 + Implementation of new Ada features (in particular limited 12152 with, limited aggregates) 12153 12154 C/Objective-C/C++ 12155 12156 * Precompiled headers are now supported. Precompiled headers can 12157 dramatically speed up compilation of some projects. There are some 12158 known defects in the current precompiled header implementation that 12159 will result in compiler crashes in relatively rare situations. 12160 Therefore, precompiled headers should be considered a "technology 12161 preview" in this release. Read the manual for details about how to 12162 use precompiled headers. 12163 * File handling in the preprocessor has been rewritten. GCC no longer 12164 gets confused by symlinks and hardlinks, and now has a correct 12165 implementation of #import and #pragma once. These two directives 12166 have therefore been un-deprecated. 12167 * The undocumented extension that allowed C programs to have a label 12168 at the end of a compound statement, which has been deprecated since 12169 GCC 3.0, has been removed. 12170 * The cast-as-lvalue extension has been removed for C++ and 12171 deprecated for C and Objective-C. In particular, code like this: 12172 int i; 12173 (char) i = 5; 12174 12175 or this: 12176 char *p; 12177 ((int *) p)++; 12178 12179 is no longer accepted for C++ and will not be accepted for C and 12180 Objective-C in a future version. 12181 * The conditional-expression-as-lvalue extension has been deprecated 12182 for C and Objective-C. In particular, code like this: 12183 int a, b, c; 12184 (a ? b : c) = 2; 12185 12186 will not be accepted for C and Objective-C in a future version. 12187 * The compound-expression-as-lvalue extension has been deprecated for 12188 C and Objective-C. In particular, code like this: 12189 int a, b; 12190 (a, b) = 2; 12191 12192 will not be accepted for C and Objective-C in a future version. A 12193 possible non-intrusive workaround is the following: 12194 (*(a, &b)) = 2; 12195 12196 * Several [10]built-in functions such as __builtin_popcount for 12197 counting bits, finding the highest and lowest bit in a word, and 12198 parity have been added. 12199 * The -fwritable-strings option has been deprecated and will be 12200 removed. 12201 * Many C math library functions are now recognized as built-ins and 12202 optimized. 12203 * The C, C++, and Objective-C compilers can now handle source files 12204 written in any character encoding supported by the host C library. 12205 The default input character set is taken from the current locale, 12206 and may be overridden with the -finput-charset command line option. 12207 In the future we will add support for inline encoding markers. 12208 12209 C++ 12210 12211 * G++ is now much closer to full conformance to the ISO/ANSI C++ 12212 standard. This means, among other things, that a lot of invalid 12213 constructs which used to be accepted in previous versions will now 12214 be rejected. It is very likely that existing C++ code will need to 12215 be fixed. This document lists some of the most common issues. 12216 * A hand-written recursive-descent C++ parser has replaced the 12217 YACC-derived C++ parser from previous GCC releases. The new parser 12218 contains much improved infrastructure needed for better parsing of 12219 C++ source codes, handling of extensions, and clean separation 12220 (where possible) between proper semantics analysis and parsing. The 12221 new parser fixes many bugs that were found in the old parser. 12222 * You must now use the typename and template keywords to disambiguate 12223 dependent names, as required by the C++ standard. 12224 struct K { 12225 typedef int mytype_t; 12226 }; 12227 12228 template <class T1> struct A { 12229 template <class T2> struct B { 12230 void callme(void); 12231 }; 12232 12233 template <int N> void bar(void) 12234 { 12235 // Use 'typename' to tell the parser that T1::mytype_t names 12236 // a type. This is needed because the name is dependent (in 12237 // this case, on template parameter T1). 12238 typename T1::mytype_t x; 12239 x = 0; 12240 } 12241 }; 12242 12243 template <class T> void template_func(void) 12244 { 12245 // Use 'template' to prefix member templates within 12246 // dependent types (a has type A<T>, which depends on 12247 // the template parameter T). 12248 A<T> a; 12249 a.template bar<0>(); 12250 12251 // Use 'template' to tell the parser that B is a nested 12252 // template class (dependent on template parameter T), and 12253 // 'typename' because the whole A<T>::B<int> is 12254 // the name of a type (again, dependent). 12255 typename A<T>::template B<int> b; 12256 b.callme(); 12257 } 12258 12259 void non_template_func(void) 12260 { 12261 // Outside of any template class or function, no names can be 12262 // dependent, so the use of the keyword 'typename' and 'template' 12263 // is not needed (and actually forbidden). 12264 A<K> a; 12265 a.bar<0>(); 12266 A<K>::B<float> b; 12267 b.callme(); 12268 } 12269 * In a template definition, unqualified names will no longer find 12270 members of a dependent base (as specified by [temp.dep]/3 in the 12271 C++ standard). For example, 12272 template <typename T> struct B { 12273 int m; 12274 int n; 12275 int f (); 12276 int g (); 12277 }; 12278 int n; 12279 int g (); 12280 template <typename T> struct C : B<T> { 12281 void h () 12282 { 12283 m = 0; // error 12284 f (); // error 12285 n = 0; // ::n is modified 12286 g (); // ::g is called 12287 } 12288 }; 12289 You must make the names dependent, e.g. by prefixing them with 12290 this->. Here is the corrected definition of C<T>::h, 12291 template <typename T> void C<T>::h () 12292 { 12293 this->m = 0; 12294 this->f (); 12295 this->n = 0 12296 this->g (); 12297 } 12298 As an alternative solution (unfortunately not backwards compatible 12299 with GCC 3.3), you may use using declarations instead of this->: 12300 template <typename T> struct C : B<T> { 12301 using B<T>::m; 12302 using B<T>::f; 12303 using B<T>::n; 12304 using B<T>::g; 12305 void h () 12306 { 12307 m = 0; 12308 f (); 12309 n = 0; 12310 g (); 12311 } 12312 }; 12313 * In templates, all non-dependent names are now looked up and bound 12314 at definition time (while parsing the code), instead of later when 12315 the template is instantiated. For instance: 12316 void foo(int); 12317 12318 template <int> struct A { 12319 static void bar(void){ 12320 foo('a'); 12321 } 12322 }; 12323 12324 void foo(char); 12325 12326 int main() 12327 { 12328 A<0>::bar(); // Calls foo(int), used to call foo(char). 12329 } 12330 12331 * In an explicit instantiation of a class template, you must use 12332 class or struct before the template-id: 12333 template <int N> 12334 class A {}; 12335 12336 template A<0>; // error, not accepted anymore 12337 template class A<0>; // OK 12338 * The "named return value" and "implicit typename" extensions have 12339 been removed. 12340 * Default arguments in function types have been deprecated and will 12341 be removed. 12342 * ARM-style name-injection of friend declarations has been deprecated 12343 and will be removed. For example: struct S { friend void f(); }; 12344 void g() { f(); } will not be accepted by future versions of G++; 12345 instead a declaration of "f" will need to be present outside of the 12346 scope of "S". 12347 * Covariant returns are implemented for all but varadic functions 12348 that require an adjustment. 12349 * When -pedantic is used, G++ now issues errors about spurious 12350 semicolons. For example, 12351 namespace N {}; // Invalid semicolon. 12352 void f() {}; // Invalid semicolon. 12353 * G++ no longer accepts attributes for a declarator after the 12354 initializer associated with that declarator. For example, 12355 X x(1) __attribute__((...)); 12356 is no longer accepted. Instead, use: 12357 X x __attribute__((...)) (1); 12358 * Inside the scope of a template class, the name of the class itself 12359 can be treated as either a class or a template. So GCC used to 12360 accept the class name as argument of type template, and template 12361 template parameter. However this is not C++ standard compliant. Now 12362 the name is not treated as a valid template template argument 12363 unless you qualify the name by its scope. For example, the code 12364 below no longer compiles. 12365 template <template <class> class TT> class X {}; 12366 template <class T> class Y { 12367 X<Y> x; // Invalid, Y is always a type template parameter. 12368 }; 12369 The valid code for the above example is 12370 X< ::Y> x; // Valid. 12371 (Notice the space between < and : to prevent GCC to interpret this 12372 as a digraph for [.) 12373 * Friend declarations that refer to template specializations are 12374 rejected if the template has not already been declared. For 12375 example, 12376 template <typename T> 12377 class C { 12378 friend void f<> (C&); 12379 }; 12380 is rejected. You must first declare f as a template, 12381 template <typename T> 12382 void f(T); 12383 * In case of friend declarations, every name used in the friend 12384 declaration must be accessible at the point of that declaration. 12385 Previous versions of G++ used to be less strict about this and 12386 allowed friend declarations for private class members, for example. 12387 See the ISO C++ Standard Committee's [11]defect report #209 for 12388 details. 12389 * Declaration of member functions of class templates as friends are 12390 supported. For example, 12391 template <typename T> struct A { 12392 void f(); 12393 }; 12394 class C { 12395 template <typename T> friend void A<T>::f(); 12396 }; 12397 * You must use template <> to introduce template specializations, as 12398 required by the standard. For example, 12399 template <typename T> 12400 struct S; 12401 12402 struct S<int> { }; 12403 is rejected. You must write, 12404 template <> struct S<int> {}; 12405 * G++ used to accept code like this, 12406 struct S { 12407 int h(); 12408 void f(int i = g()); 12409 int g(int i = h()); 12410 }; 12411 This behavior is not mandated by the standard. Now G++ issues an 12412 error about this code. To avoid the error, you must move the 12413 declaration of g before the declaration of f. The default arguments 12414 for g must be visible at the point where it is called. 12415 * The C++ ABI Section 3.3.3 specifications for the array construction 12416 routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return 12417 NULL when the allocator argument returns NULL. These changes are 12418 incorporated into the libstdc++ runtime library. 12419 * Using a name introduced by a typedef in a friend declaration or in 12420 an explicit instantiation is now rejected, as specified by the ISO 12421 C++ standard. 12422 class A; 12423 typedef A B; 12424 class C { 12425 friend class B; // error, no typedef name here 12426 friend B; // error, friend always needs class/struct/enum 12427 friend class A; // OK 12428 }; 12429 12430 template <int> class Q {}; 12431 typedef Q<0> R; 12432 template class R; // error, no typedef name here 12433 template class Q<0>; // OK 12434 * When allocating an array with a new expression, GCC used to allow 12435 parentheses around the type name. This is actually ill-formed and 12436 it is now rejected: 12437 int* a = new (int)[10]; // error, not accepted anymore 12438 int* a = new int[10]; // OK 12439 * When binding an rvalue of class type to a reference, the copy 12440 constructor of the class must be accessible. For instance, consider 12441 the following code: 12442 class A 12443 { 12444 public: 12445 A(); 12446 12447 private: 12448 A(const A&); // private copy ctor 12449 }; 12450 12451 A makeA(void); 12452 void foo(const A&); 12453 12454 void bar(void) 12455 { 12456 foo(A()); // error, copy ctor is not accessible 12457 foo(makeA()); // error, copy ctor is not accessible 12458 12459 A a1; 12460 foo(a1); // OK, a1 is a lvalue 12461 } 12462 This might be surprising at first sight, especially since most 12463 popular compilers do not correctly implement this rule ([12]further 12464 details). 12465 * When forming a pointer to member or a pointer to member function, 12466 access checks for class visibility (public, protected, private) are 12467 now performed using the qualifying scope of the name itself. This 12468 is better explained with an example: 12469 class A 12470 { 12471 public: 12472 void pub_func(); 12473 protected: 12474 void prot_func(); 12475 private: 12476 void priv_func(); 12477 }; 12478 12479 class B : public A 12480 { 12481 public: 12482 void foo() 12483 { 12484 &A::pub_func; // OK, pub_func is accessible through A 12485 &A::prot_func; // error, cannot access prot_func through A 12486 &A::priv_func; // error, cannot access priv_func through A 12487 12488 &B::pub_func; // OK, pub_func is accessible through B 12489 &B::prot_func; // OK, can access prot_func through B (within B) 12490 &B::priv_func; // error, cannot access priv_func through B 12491 } 12492 }; 12493 12494 Runtime Library (libstdc++) 12495 12496 * Optimization work: 12497 + Streamlined streambuf, filebuf, separate synched with C 12498 Standard I/O streambuf. 12499 + All formatted I/O now uses cached locale information. 12500 + STL optimizations (memory/speed for list, red-black trees as 12501 used by sets and maps). 12502 + More use of GCC builtins. 12503 + String optimizations (avoid contention on 12504 increment/decrement-and-test of the reference count in the 12505 empty-string object, constructor from input_iterators 12506 speedup). 12507 * Static linkage size reductions. 12508 * Large File Support (files larger than 2 GB on 32-bit systems). 12509 * Wide character and variable encoding filebuf work (UTF-8, Unicode). 12510 * Generic character traits. 12511 * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x, 12512 Solaris 2.7 and above, AIX 5.x, Irix 6.5. 12513 * The allocator class is now standard-conformant, and two additional 12514 extension allocators have been added, mt_alloc and 12515 bitmap_allocator. 12516 * PCH support: -include bits/stdc++.h (2x compile speedup). 12517 * Rewrote __cxa_demangle with support for C++ style allocators. 12518 * New debug modes for STL containers and iterators. 12519 * Testsuite rewrite: five times as many tests, plus increasingly 12520 sophisticated tests, including I/O, MT, multi-locale, wide and 12521 narrow characters. 12522 * Use current versions of GNU "autotools" for build/configuration. 12523 12524 Objective-C 12525 12526 * The Objective-C front end has been updated to include the numerous 12527 bug fixes and enhancements previously available only in Apple's 12528 version of GCC. These include: 12529 + Structured exception (@try... @catch... @finally, @throw) and 12530 synchronization (@synchronized) support. These are accessible 12531 via the -fobjc-exceptions switch; as of this writing, they may 12532 only be used in conjunction with -fnext-runtime on Mac OS X 12533 10.3 and later. See [13]Options Controlling Objective-C 12534 Dialect for more information. 12535 + An overhaul of @encode logic. The C99 _Bool and C++ bool type 12536 may now be encoded as 'B'. In addition, the back-end/codegen 12537 dependencies have been removed. 12538 + An overhaul of message dispatch construction, ensuring that 12539 the various receiver types (and casts thereof) are handled 12540 properly, and that correct diagnostics are issued. 12541 + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue" 12542 (-freplace-objc-classes) debugging modes, currently available 12543 on Mac OS X 10.3 and later. See [14]Options Controlling 12544 Objective-C Dialect for more information. 12545 + Access to optimized runtime entry points (-fno-nil-receivers ) 12546 on the assumption that message receivers are never nil. This 12547 is currently available on Mac OS X 10.3 and later. See 12548 [15]Options Controlling Objective-C Dialect for more 12549 information. 12550 12551 Java 12552 12553 * Compiling a .jar file will now cause non-.class entries to be 12554 automatically compiled as resources. 12555 * libgcj has been ported to Darwin. 12556 * Jeff Sturm has adapted Jan Hubicka's call graph optimization code 12557 to gcj. 12558 * libgcj has a new gcjlib URL type; this lets URLClassLoader load 12559 code from shared libraries. 12560 * libgcj has been much more completely merged with [16]GNU Classpath. 12561 * Class loading is now much more correct; in particular the caller's 12562 class loader is now used when that is required. 12563 * [17]Eclipse 2.x will run out of the box using gij. 12564 * Parts of java.nio have been implemented. Direct and indirect 12565 buffers work, as do fundamental file and socket operations. 12566 * java.awt has been improved, though it is still not ready for 12567 general use. 12568 * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST 12569 method. 12570 * The MinGW port has matured. Enhancements include socket timeout 12571 support, thread interruption, improved Runtime.exec() handling and 12572 support for accented characters in filenames. 12573 12574 Fortran 12575 12576 * Fortran improvements are listed in the [18]Fortran documentation. 12577 12578 New Targets and Target Specific Improvements 12579 12580 Alpha 12581 12582 * Several [19]built-in functions have been added such as 12583 __builtin_alpha_zap to allow utilizing the more obscure 12584 instructions of the CPU. 12585 * Parameter passing of complex arguments has changed to match the 12586 ABI. This change is incompatible with previous GCC versions, but 12587 does fix compatibility with the Tru64 compiler and several corner 12588 cases where GCC was incompatible with itself. 12589 12590 ARM 12591 12592 * Nicolas Pitre has contributed his hand-coded floating-point support 12593 code for ARM. It is both significantly smaller and faster than the 12594 existing C-based implementation, even when building applications 12595 for Thumb. The arm-elf configuration has been converted to use the 12596 new code. 12597 * Support for the Intel's iWMMXt architecture, a second generation 12598 XScale processor, has been added. Enabled at run time with the 12599 -mcpu=iwmmxt command line switch. 12600 * A new ARM target has been added: arm-wince-pe. This is similar to 12601 the arm-pe target, but it defaults to using the APCS32 ABI. 12602 * The existing ARM pipeline description has been converted to the use 12603 the [20]DFA processor pipeline model. There is not much change in 12604 code performance, but the description is now [21]easier to 12605 understand. 12606 * Support for the Cirrus EP9312 Maverick floating point co-processor 12607 added. Enabled at run time with the -mcpu=ep9312 command line 12608 switch. Note however that the multilibs to support this chip are 12609 currently disabled in gcc/config/arm/t-arm-elf, so if you want to 12610 enable their production you will have to uncomment the entries in 12611 that file. 12612 12613 H8/300 12614 12615 * Support for long long has been added. 12616 * Support for saveall attribute has been added. 12617 * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code 12618 for H8/300H and H8S, which is much faster than the previous 12619 implementation. 12620 * A lot of small performance improvements. 12621 12622 IA-32/AMD64 (x86-64) 12623 12624 * Tuning for K8 (AMD Opteron/Athlon64) core is available via 12625 -march=k8 and -mcpu=k8. 12626 * Scalar SSE code generation carefully avoids reformatting penalties, 12627 hidden dependencies and minimizes the number of uops generated on 12628 both Intel and AMD CPUs. 12629 * Vector MMX and SSE operands are now passed in registers to improve 12630 performance and match the argument passing convention used by the 12631 Intel C++ Compiler. As a result it is not possible to call 12632 functions accepting vector arguments compiled by older GCC version. 12633 * Conditional jump elimination is now more aggressive on modern CPUs. 12634 * The Athlon ports has been converted to use the DFA processor 12635 pipeline description. 12636 * Optimization of indirect tail calls is now possible in a similar 12637 fashion as direct sibcall optimization. 12638 * Further small performance improvements. 12639 * -m128bit-long-double is now less buggy. 12640 * __float128 support in 64-bit compilation. 12641 * Support for data structures exceeding 2GB in 64-bit mode. 12642 * -mcpu has been renamed to -mtune. 12643 12644 IA-64 12645 12646 * Tuning code for the Itanium 2 processor has been added. The 12647 generation of code tuned for Itanium 2 (option -mtune=itanium2) is 12648 enabled by default now. To generate code tuned for Itanium 1 the 12649 option -mtune=itanium1 should be used. 12650 * [22]DFA processor pipeline descriptions for the IA-64 processors 12651 have been added. This resulted in about 3% improvement on the 12652 SPECInt2000 benchmark for Itanium 2. 12653 * Instruction bundling for the IA-64 processors has been rewritten 12654 using the DFA pipeline hazard recognizer. It resulted in about 60% 12655 compiler speedup on the SPECInt2000 C programs. 12656 12657 M32R 12658 12659 * Support for the M32R/2 processor has been added by Renesas. 12660 * Support for an M32R GNU/Linux target and PIC code generation has 12661 been added by Renesas. 12662 12663 M68000 12664 12665 * Bernardo Innocenti (Develer S.r.l.) has contributed the 12666 m68k-uclinux target, based on former work done by Paul Dale 12667 (SnapGear Inc.). Code generation for the ColdFire processors family 12668 has been enhanced and extended to support the MCF 53xx and MCF 54xx 12669 cores, integrating former work done by Peter Barada (Motorola). 12670 12671 MIPS 12672 12673 Processor-specific changes 12674 12675 * Support for the RM7000 and RM9000 processors has been added. It can 12676 be selected using the -march compiler option and should work with 12677 any MIPS I (mips-*) or MIPS III (mips64-*) configuration. 12678 * Support for revision 2 of the MIPS32 ISA has been added. It can be 12679 selected with the command-line option -march=mips32r2. 12680 * There is a new option, -mfix-sb1, to work around certain SB-1 12681 errata. 12682 12683 Configuration 12684 12685 * It is possible to customize GCC using the following configure-time 12686 options: 12687 + --with-arch, which specifies the default value of the -march 12688 option. 12689 + --with-tune, which specifies the default value of the -mtune 12690 option. 12691 + --with-abi, which specifies the default ABI. 12692 + --with-float=soft, which tells GCC to use software floating 12693 point by default. 12694 + --with-float=hard, which tells GCC to use hardware floating 12695 point by default. 12696 * A 64-bit GNU/Linux port has been added. The associated 12697 configurations are mips64-linux-gnu and mips64el-linux-gnu. 12698 * The 32-bit GNU/Linux port now supports Java. 12699 * The IRIX 6 configuration now supports the o32 ABI and will build 12700 o32 multilibs by default. This support is compatible with both 12701 binutils and the SGI tools, but note that several features, 12702 including debugging information and DWARF2 exception handling, are 12703 only available when using the GNU assembler. Use of the GNU 12704 assembler and linker (version 2.15 or above) is strongly 12705 recommended. 12706 * The IRIX 6 configuration now supports 128-bit long doubles. 12707 * There are two new RTEMS-specific configurations, mips-rtems and 12708 mipsel-rtems. 12709 * There are two new *-elf configurations, mipsisa32r2-elf and 12710 mipsisa32r2el-elf. 12711 12712 General 12713 12714 * Several [23]ABI bugs have been fixed. Unfortunately, these changes 12715 will break binary compatibility with earlier releases. 12716 * GCC can now use explicit relocation operators when generating 12717 -mabicalls code. This behavior is controlled by -mexplicit-relocs 12718 and can have several performance benefits. For example: 12719 + It allows for more optimization of GOT accesses, including 12720 better scheduling and redundancy elimination. 12721 + It allows sibling calls to be implemented as jumps. 12722 + n32 and n64 leaf functions can use a call-clobbered global 12723 pointer instead of $28. 12724 + The code to set up $gp can be removed from functions that 12725 don't need it. 12726 * A new option, -mxgot, allows the GOT to be bigger than 64k. This 12727 option is equivalent to the assembler's -xgot option and should be 12728 used instead of -Wa,-xgot. 12729 * Frame pointer elimination is now supported when generating 64-bit 12730 MIPS16 code. 12731 * Inline block moves have been optimized to take more account of 12732 alignment information. 12733 * Many internal changes have been made to the MIPS port, mostly aimed 12734 at reducing the reliance on assembler macros. 12735 12736 PowerPC 12737 12738 * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64 12739 [24]ABI incompatibilities regarding the way parameters are passed 12740 during functions calls. These changes may result in incompatibility 12741 between code compiled with GCC 3.3 and GCC 3.4. 12742 12743 PowerPC Darwin 12744 12745 * Support for shared/dylib gcc libraries has been added. It is 12746 enabled by default on powerpc-apple-darwin7.0.0 and up. 12747 * Libgcj is enabled by default. On systems older than 12748 powerpc-apple-darwin7.0.0 you need to install dlcompat. 12749 * 128-bit IBM extended precision format support added for long 12750 double. 12751 12752 PowerPC64 GNU/Linux 12753 12754 * By default, PowerPC64 GNU/Linux now uses natural alignment of 12755 structure elements. The old four byte alignment for double, with 12756 special rules for a struct starting with a double, can be chosen 12757 with -malign-power. This change may result in incompatibility 12758 between code compiled with GCC 3.3 and GCC 3.4. 12759 * -mabi=altivec is now the default rather than -mabi=no-altivec. 12760 * 128-bit IBM extended precision format support added for long 12761 double. 12762 12763 S/390 and zSeries 12764 12765 * New command-line options allow to specify the intended execution 12766 environment for generated code: 12767 + -mesa/-mzarch allows to specify whether to generate code 12768 running in ESA/390 mode or in z/Architecture mode (this is 12769 applicable to 31-bit code only). 12770 + -march allows to specify a minimum processor architecture 12771 level (g5, g6, z900, or z990). 12772 + -mtune allows to specify which processor to tune for. 12773 * It is possible to customize GCC using the following configure-time 12774 options: 12775 + --with-mode, which specifies whether to default to assuming 12776 ESA/390 or z/Architecture mode. 12777 + --with-arch, which specifies the default value of the -march 12778 option. 12779 + --with-tune, which specifies the default value of the -mtune 12780 option. 12781 * Support for the z990 processor has been added, and can be selected 12782 using -march=z990 or -mtune=z990. This includes instruction 12783 scheduling tuned for the superscalar instruction pipeline of the 12784 z990 processor as well as support for all new instructions provided 12785 by the long-displacement facility. 12786 * Support to generate 31-bit code optimized for zSeries processors 12787 (running in ESA/390 or in z/Architecture mode) has been added. This 12788 can be selected using -march=z900 and -mzarch respectively. 12789 * Instruction scheduling for the z900 and z990 processors now uses 12790 the DFA pipeline hazard recognizer. 12791 * GCC no longer generates code to maintain a stack backchain, 12792 previously used to generate stack backtraces for debugging 12793 purposes. As replacement that does not incur runtime overhead, 12794 DWARF-2 call frame information is provided by GCC; this is 12795 supported by GDB 6.1. The old behavior can be restored using the 12796 -mbackchain option. 12797 * The stack frame size of functions may now exceed 2 GB in 64-bit 12798 code. 12799 * A port for the 64-bit IBM TPF operating system has been added; the 12800 configuration is s390x-ibm-tpf. This configuration is supported as 12801 cross-compilation target only. 12802 * Various changes to improve the generated code have been 12803 implemented, including: 12804 + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT 12805 instructions to significantly speed up many floating-point 12806 applications. 12807 + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL 12808 WITH BORROW instructions to speed up long long arithmetic. 12809 + GCC now uses the SEARCH STRING instruction to implement 12810 strlen(). 12811 + In many cases, function call overhead for 31-bit code has been 12812 reduced by placing the literal pool after the function code 12813 instead of after the function prolog. 12814 + Register 14 is no longer reserved in 64-bit code. 12815 + Handling of global register variables has been improved. 12816 12817 SPARC 12818 12819 * The option -mflat is deprecated. 12820 * Support for large (> 2GB) frames has been added to the 64-bit port. 12821 * Several [25]ABI bugs have been fixed. Unfortunately, these changes 12822 will break binary compatibility with earlier releases. 12823 * The default debugging format has been switched from STABS to 12824 DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already 12825 the default debugging format for 64-bit code on Solaris. 12826 12827 SuperH 12828 12829 * Support for the SH2E processor has been added. Enabled at run time 12830 with the -m2e command line switch, or at configure time by 12831 specifying sh2e as the machine part of the target triple. 12832 12833 V850 12834 12835 * Support for the Mitsubishi V850E1 processor has been added. This is 12836 a variant of the V850E processor with some additional debugging 12837 instructions. 12838 12839 Xtensa 12840 12841 * Several ABI bugs have been fixed. Unfortunately, these changes 12842 break binary compatibility with earlier releases. 12843 + For big-endian processors, the padding of aggregate return 12844 values larger than a word has changed. If the size of an 12845 aggregate return value is not a multiple of 32 bits, previous 12846 versions of GCC inserted padding in the most-significant bytes 12847 of the first return value register. Aggregates larger than a 12848 word are now padded in the least-significant bytes of the last 12849 return value register used. Aggregates smaller than a word are 12850 still padded in the most-significant bytes. The return value 12851 padding has not changed for little-endian processors. 12852 + Function arguments with 16-byte alignment are now properly 12853 aligned. 12854 + The implementation of the va_list type has changed. A va_list 12855 value created by va_start from a previous release cannot be 12856 used with va_arg from this release, or vice versa. 12857 * More processor configuration options for Xtensa processors are 12858 supported: 12859 + the ABS instruction is now optional; 12860 + the ADDX* and SUBX* instructions are now optional; 12861 + an experimental CONST16 instruction can be used to synthesize 12862 constants instead of loading them from constant pools. 12863 These and other Xtensa processor configuration options can no 12864 longer be enabled or disabled by command-line options; the 12865 processor configuration must be specified by the xtensa-config.h 12866 header file when building GCC. Additionally, the 12867 -mno-serialize-volatile option is no longer supported. 12868 12869 Obsolete Systems 12870 12871 Support for a number of older systems has been declared obsolete in GCC 12872 3.4. Unless there is activity to revive them, the next release of GCC 12873 will have their sources permanently removed. 12874 12875 All configurations of the following processor architectures have been 12876 declared obsolete: 12877 * Mitsubishi D30V, d30v-* 12878 * AT&T DSP1600 and DSP1610, dsp16xx-* 12879 * Intel 80960, i960 12880 12881 Also, some individual systems have been obsoleted: 12882 * ARM Family 12883 + Support for generating code for operation in APCS/26 mode 12884 (-mapcs-26). 12885 * IBM ESA/390 12886 + "Bigfoot" port, i370-*. (The other port, s390-*, is actively 12887 maintained and supported.) 12888 * Intel 386 family 12889 + MOSS, i?86-moss-msdos and i?86-*-moss* 12890 + NCR 3000 running System V r.4, i?86-ncr-sysv4* 12891 + FreeBSD with a.out object format, i?86-*-freebsd*aout* and 12892 i?86-*-freebsd2* 12893 + GNU/Linux with a.out object format, i?86-linux*aout* 12894 + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1* 12895 + Interix versions before Interix 3, i?86-*-interix 12896 + Mach microkernel, i?86-mach* 12897 + SCO UnixWare with UDK, i?86-*-udk* 12898 + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]* 12899 + VSTa microkernel, i386-*-vsta 12900 * Motorola M68000 family 12901 + HPUX, m68k-hp-hpux* and m68000-hp-hpux* 12902 + NetBSD with a.out object format (before NetBSD 1.4), 12903 m68k-*-*-netbsd* except m68k-*-*-netbsdelf* 12904 + Generic System V r.4, m68k-*-sysv4* 12905 * VAX 12906 + Generic VAX, vax-*-* (This is generic VAX only; we have not 12907 obsoleted any VAX triples for specific operating systems.) 12908 12909 Documentation improvements 12910 12911 Other significant improvements 12912 12913 * The build system has undergone several significant cleanups. 12914 Subdirectories will only be configured if they are being built, and 12915 all subdirectory configures are run from the make command. The top 12916 level has been autoconfiscated. 12917 * Building GCC no longer writes to its source directory. This should 12918 help those wishing to share a read-only source directory over NFS 12919 or build from a CD. The exceptions to this feature are if you 12920 configure with either --enable-maintainer-mode or 12921 --enable-generated-files-in-srcdir. 12922 * The -W warning option has been renamed to -Wextra, which is more 12923 easily understood. The older spelling will be retained for 12924 backwards compatibility. 12925 * Substantial improvements in compile time have been made, 12926 particularly for non-optimizing compilations. 12927 __________________________________________________________________ 12928 12929 GCC 3.4.0 12930 12931 Bug Fixes 12932 12933 A vast number of bugs have been fixed in 3.4.0, too many to publish a 12934 complete list here. [26]Follow this link to query the Bugzilla database 12935 for the list of over 900 bugs fixed in 3.4.0. This is the list of all 12936 bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4 12937 regressions. 12938 __________________________________________________________________ 12939 12940 GCC 3.4.1 12941 12942 Bug Fixes 12943 12944 This section lists the problem reports (PRs) from GCC's bug tracking 12945 system that are known to be fixed in the 3.4.1 release. This list might 12946 not be complete (that is, it is possible that some PRs that have been 12947 fixed are not listed here). 12948 12949 Bootstrap failures 12950 12951 * [27]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler 12952 emitted - PIC related 12953 * [28]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf 12954 * [29]14760 A bug in configure.in prevents using both 12955 --program-suffix and --program-prefix 12956 * [30]14671 [hppa64] bootstrap fails: ICE in 12957 save_call_clobbered_regs, in caller_save.c 12958 * [31]15093 [alpha][Java] make bootstrap fails to configure libffi on 12959 Alpha 12960 * [32]15178 Solaris 9/x86 fails linking after stage 3 12961 12962 Multi-platform internal compiler errors (ICEs) 12963 12964 * [33]12753 (preprocessor) Memory corruption in preprocessor on bad 12965 input 12966 * [34]13985 ICE in gcc.c-torture/compile/930621-1.c 12967 * [35]14810 (c++) tree check failures with invalid code involving 12968 templates 12969 * [36]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in 12970 cp/parser.c 12971 * [37]15044 (c++) ICE on syntax error, template header 12972 * [38]15057 (c++) Compiling of conditional value throw constructs 12973 cause a segmentation violation 12974 * [39]15064 (c++) typeid of template parameter gives ICE 12975 * [40]15142 (c++) ICE when passing a string where a char* is expected 12976 in a throw statement 12977 * [41]15159 ICE in rtl_verify_flow_info_1 12978 * [42]15165 (c++) ICE in instantiate_template 12979 * [43]15193 Unary minus using pointer to V4SF vector causes 12980 -fforce-mem to exhaust all memory 12981 * [44]15209 (c++) Runs out of memory with packed structs 12982 * [45]15227 (c++) Trouble with invalid function definition 12983 * [46]15285 (c++) instantiate_type ICE when forming pointer to 12984 template function 12985 * [47]15299 (c++) ICE in resolve_overloaded_unification 12986 * [48]15329 (c++) ICE on constructor of member template 12987 * [49]15550 ICE in extract_insn, in recog.c 12988 * [50]15554 (c++) ICE in tsubst_copy, in cp/pt.c 12989 * [51]15640 (c++) ICE on invalid code in arg_assoc, in 12990 cp/name-lookup.c 12991 * [52]15666 [unit-at-a-time] Gcc abort on valid code 12992 * [53]15696 (c++) ICE with bad pointer-to-member code 12993 * [54]15701 (c++) ICE with friends and template template parameter 12994 * [55]15761 ICE in do_SUBST, in combine.c 12995 * [56]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops 12996 12997 Ada 12998 12999 * [57]14538 All RTEMS targets broken for gnat 13000 13001 C front end 13002 13003 * [58]12391 missing warning about assigning to an incomplete type 13004 * [59]14649 atan(1.0) should not be a constant expression 13005 * [60]15004 [unit-at-a-time] no warning for unused paramater in 13006 static function 13007 * [61]15749 --pedantic-errors behaves differently from --pedantic 13008 with C-compiler on GNU/Linux 13009 13010 C++ compiler and library 13011 13012 * [62]10646 non-const reference is incorrectly matched in a "const T" 13013 partial specialization 13014 * [63]12077 wcin.rdbuf()->in_avail() return value too high 13015 * [64]13598 enc_filebuf doesn't work 13016 * [65]14211 const_cast returns lvalue but should be rvalue 13017 * [66]14220 num_put::do_put() undesired float/double behavior 13018 * [67]14245 problem with user-defined allocators in std::basic_string 13019 * [68]14340 libstdc++ Debug mode: failure to convert iterator to 13020 const_iterator 13021 * [69]14600 __gnu_cxx::stdio_sync_filebuf should expose internal 13022 FILE* 13023 * [70]14668 no warning anymore for reevaluation of declaration 13024 * [71]14775 LFS (large file support) tests missing 13025 * [72]14821 Duplicate namespace alias declaration should not conflict 13026 * [73]14930 Friend declaration ignored 13027 * [74]14932 cannot use offsetof to get offsets of array elements in 13028 g++ 3.4.0 13029 * [75]14950 [non unit-at-a-time] always_inline does not mix with 13030 templates and -O0 13031 * [76]14962 g++ ignores #pragma redefine_extname 13032 * [77]14975 Segfault on low-level write error during imbue 13033 * [78]15002 Linewise stream input is unusably slow (std::string slow) 13034 * [79]15025 compiler accepts redeclaration of template as 13035 non-template 13036 * [80]15046 [arm] Math functions misdetected by cross configuration 13037 * [81]15069 a bit test on a variable of enum type is miscompiled 13038 * [82]15074 g++ -lsupc++ still links against libstdc++ 13039 * [83]15083 spurious "statement has no effect" warning 13040 * [84]15096 parse error with templates and pointer to const member 13041 * [85]15287 combination of operator[] and operator .* fails in 13042 templates 13043 * [86]15317 __attribute__ unused in first parameter of constructor 13044 gives error 13045 * [87]15337 sizeof on incomplete type diagnostic 13046 * [88]15361 bitset<>::_Find_next fails 13047 * [89]15412 _GLIBCXX_ symbols symbols defined and used in different 13048 namespaces 13049 * [90]15427 valid code results in incomplete type error 13050 * [91]15471 Incorrect member pointer offsets in anonymous 13051 structs/unions 13052 * [92]15503 nested template problem 13053 * [93]15507 compiler hangs while laying out union 13054 * [94]15542 operator & and template definitions 13055 * [95]15565 SLES9: leading + sign for unsigned int with showpos 13056 * [96]15625 friend defined inside a template fails to find static 13057 function 13058 * [97]15629 Function templates, overloads, and friend name injection 13059 * [98]15742 'noreturn' attribute ignored in method of template 13060 functions. 13061 * [99]15775 Allocator::pointer consistently ignored 13062 * [100]15821 Duplicate namespace alias within namespace rejected 13063 * [101]15862 'enum yn' fails (confict with undeclared builtin) 13064 * [102]15875 rejects pointer to member in template 13065 * [103]15877 valid code using templates and anonymous enums is 13066 rejected 13067 * [104]15947 Puzzling error message for wrong destructor declaration 13068 in template class 13069 * [105]16020 cannot copy __gnu_debug::bitset 13070 * [106]16154 input iterator concept too restrictive 13071 * [107]16174 deducing top-level consts 13072 13073 Java 13074 13075 * [108]14315 Java compiler is not parallel make safe 13076 13077 Fortran 13078 13079 * [109]15151 [g77] incorrect logical i/o in 64-bit mode 13080 13081 Objective-C 13082 13083 * [110]7993 private variables cannot be shadowed in subclasses 13084 13085 Optimization bugs 13086 13087 * [111]15228 useless copies of floating point operands 13088 * [112]15345 [non-unit-at-a-time] unreferenced nested inline 13089 functions not optimized away 13090 * [113]15945 Incorrect floating point optimization 13091 * [114]15526 ftrapv aborts on 0 * (-1) 13092 * [115]14690 Miscompiled POOMA tests 13093 * [116]15112 GCC generates code to write to unchanging memory 13094 13095 Preprocessor 13096 13097 * [117]15067 Minor glitch in the source of cpp 13098 13099 Main driver program bugs 13100 13101 * [118]1963 collect2 interprets -oldstyle_liblookup as -o 13102 ldstyle_liblookup 13103 13104 x86-specific (Intel/AMD) 13105 13106 * [119]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND* 13107 section} 13108 13109 HPPA-specific 13110 13111 * [120]14782 GCC produces an unaligned data access at -O2 13112 * [121]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2 13113 * [122]15202 ICE in reload_cse_simplify_operands, in postreload.c 13114 13115 IA64-specific 13116 13117 * [123]14610 __float80 constants incorrectly emitted 13118 * [124]14813 init_array sections are initialized in the wrong order 13119 * [125]14857 GCC segfault on duplicated asm statement 13120 * [126]15598 Gcc 3.4 ICE on valid code 13121 * [127]15653 Gcc 3.4 ICE on valid code 13122 13123 MIPS-specific 13124 13125 * [128]15189 wrong filling of delay slot with -march=mips1 -G0 13126 -mno-split-addresses -mno-explicit-relocs 13127 * [129]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as 13128 2.14.91 13129 * [130]16144 Bogus reference to __divdf3 when -O1 13130 * [131]16176 Miscompilation of unaligned data in MIPS backend 13131 13132 PowerPC-specific 13133 13134 * [132]11591 ICE in gcc.dg/altivec-5.c 13135 * [133]12028 powerpc-eabispe produces bad sCOND operation 13136 * [134]14478 rs6000 geu/ltu patterns generate incorrect code 13137 * [135]14567 long double and va_arg complex args 13138 * [136]14715 Altivec stack layout may overlap gpr save with stack 13139 temps 13140 * [137]14902 (libstdc++) Stream checking functions fail when -pthread 13141 option is used. 13142 * [138]14924 Compiler ICE on valid code 13143 * [139]14960 -maltivec affects vector return with -mabi=no-altivec 13144 * [140]15106 vector varargs failure passing from altivec to 13145 non-altivec code for -m32 13146 * [141]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 & 13147 half-word operation 13148 * [142]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx 13149 and stvx 13150 * [143]15662 Segmentation fault when an exception is thrown - even if 13151 try and catch are specified 13152 13153 s390-specific 13154 13155 * [144]15054 Bad code due to overlapping stack temporaries 13156 13157 SPARC-specific 13158 13159 * [145]15783 ICE with union assignment in 64-bit mode 13160 * [146]15626 GCC 3.4 emits "ld: warning: relocation error: 13161 R_SPARC_UA32" 13162 13163 x86-64-specific 13164 13165 * [147]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64 13166 * [148]14723 Backported -march=nocona from mainline 13167 * [149]15290 __float128 failed to pass to function properly 13168 13169 Cygwin/Mingw32-specific 13170 13171 * [150]15250 Option -mms-bitfields support on GCC 3.4 is not 13172 conformant to MS layout 13173 * [151]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe 13174 worker on windows32 targets 13175 13176 Bugs specific to embedded processors 13177 13178 * [152]8309 [m68k] -m5200 produces erroneous SImode set of short 13179 varaible on stack 13180 * [153]13250 [SH] Gcc code for rotation clobbers the register, but 13181 gcc continues to use the register as if it was not clobbered 13182 * [154]13803 [coldfire] movqi operand constraints too restrictivefor 13183 TARGET_COLDFIRE 13184 * [155]14093 [SH] ICE for code when using -mhitachi option in SH 13185 * [156]14457 [m6811hc] ICE with simple c++ source 13186 * [157]14542 [m6811hc] ICE on simple source 13187 * [158]15100 [SH] cc1plus got hang-up on 13188 libstdc++-v3/testsuite/abi_check.cc 13189 * [159]15296 [CRIS] Delayed branch scheduling causing invalid code on 13190 cris-* 13191 * [160]15396 [SH] ICE with -O2 -fPIC 13192 * [161]15782 [coldfire] m68k_output_mi_thunk emits wrong code for 13193 ColdFire 13194 13195 Testsuite problems (compiler not affected) 13196 13197 * [162]11610 libstdc++ testcases 27_io/* don't work properly remotely 13198 * [163]15488 (libstdc++) possibly insufficient file permissions for 13199 executing test suite 13200 * [164]15489 (libstdc++) testsuite_files determined incorrectly 13201 13202 Documentation bugs 13203 13204 * [165]13928 (libstdc++) no whatis info in some man pages generated 13205 by doxygen 13206 * [166]14150 Ada documentation out of date 13207 * [167]14949 (c++) Need to document method visibility changes 13208 * [168]15123 libstdc++-doc: Allocators.3 manpage is empty 13209 __________________________________________________________________ 13210 13211 GCC 3.4.2 13212 13213 Bug Fixes 13214 13215 This section lists the problem reports (PRs) from GCC's bug tracking 13216 system that are known to be fixed in the 3.4.2 release. This list might 13217 not be complete (that is, it is possible that some PRs that have been 13218 fixed are not listed here). 13219 13220 Bootstrap failures and issues 13221 13222 * [169]16469 [mips-sgi-irix5.3] bootstrap fails in 13223 libstdc++-v3/testsuite 13224 * [170]16344 [hppa-linux-gnu] libstdc++'s PCH built by 13225 profiledbootstrap does not work with the built compiler 13226 * [171]16842 [Solaris/x86] mkheaders can not find mkheaders.conf 13227 13228 Multi-platform internal compiler errors (ICEs) 13229 13230 * [172]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in 13231 cp_parser_class_specifier, in cp/parser.c 13232 * [173]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 13233 * [174]15461 (c++) ICE due to NRV and inlining 13234 * [175]15890 (c++) ICE in c_expand_expr, in c-common.c 13235 * [176]16180 ICE: segmentation fault in RTL optimization 13236 * [177]16224 (c++) ICE in write_unscoped_name (template/namespace) 13237 * [178]16408 ICE: in delete_insn, in cfgrtl.c 13238 * [179]16529 (c++) ICE for: namespace-alias shall not be declared as 13239 the name of any other entity 13240 * [180]16698 (c++) ICE with exceptions and declaration of __cxa_throw 13241 * [181]16706 (c++) ICE in finish_member_declaration, in 13242 cp/semantics.c 13243 * [182]16810 (c++) Legal C++ program with cast gives ICE in 13244 build_ptrmemfunc 13245 * [183]16851 (c++) ICE when throwing a comma expression 13246 * [184]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c 13247 * [185]16904 (c++) ICE in finish_class_member_access_expr, in 13248 cp/typeck.c 13249 * [186]16905 (c++) ICE (segfault) with exceptions 13250 * [187]16964 (c++) ICE in cp_parser_class_specifier due to 13251 redefinition 13252 * [188]17068 (c++) ICE: tree check: expected class 'd', have 'x' 13253 (identifier_node) in dependent_template_p, in cp/pt.c 13254 13255 Preprocessor bugs 13256 13257 * [189]16366 Preprocessor option -remap causes memory corruption 13258 13259 Optimization 13260 13261 * [190]15345 unreferenced nested inline functions not optimized away 13262 * [191]16590 Incorrect execution when compiling with -O2 13263 * [192]16693 Bitwise AND is lost when used within a cast to an enum 13264 of the same precision 13265 * [193]17078 Jump into if(0) substatement fails 13266 13267 Problems in generated debug information 13268 13269 * [194]13956 incorrect stabs for nested local variables 13270 13271 C front end bugs 13272 13273 * [195]16684 GCC should not warn about redundant redeclarations of 13274 built-ins 13275 13276 C++ compiler and library 13277 13278 * [196]12658 Thread safety problems in locale::global() and 13279 locale::locale() 13280 * [197]13092 g++ accepts invalid pointer-to-member conversion 13281 * [198]15320 Excessive memory consumption 13282 * [199]16246 Incorrect template argument deduction 13283 * [200]16273 Memory exhausted when using nested classes and virtual 13284 functions 13285 * [201]16401 ostringstream in gcc 3.4.x very slow for big data 13286 * [202]16411 undefined reference to 13287 __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char> 13288 >::file() 13289 * [203]16489 G++ incorrectly rejects use of a null constant integral 13290 expression as a null constant pointer 13291 * [204]16618 offsetof fails with constant member 13292 * [205]16637 syntax error reported for valid input code 13293 * [206]16717 __attribute__((constructor)) broken in C++ 13294 * [207]16813 compiler error in DEBUG version of range insertion 13295 std::map::insert 13296 * [208]16853 pointer-to-member initialization from incompatible one 13297 accepted 13298 * [209]16889 ambiguity is not detected 13299 * [210]16959 Segmentation fault in ios_base::sync_with_stdio 13300 13301 Java compiler and library 13302 13303 * [211]7587 direct threaded interpreter not thread-safe 13304 * [212]16473 ServerSocket accept() leaks file descriptors 13305 * [213]16478 Hash synchronization deadlock with finalizers 13306 13307 Alpha-specific 13308 13309 * [214]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c 13310 * [215]16974 could not split insn (ice in final_scan_insn, in 13311 final.c) 13312 13313 x86-specific 13314 13315 * [216]16298 ICE in output_operand 13316 * [217]17113 ICE with SSE2 intrinsics 13317 13318 x86-64 specific 13319 13320 * [218]14697 libstdc++ couldn't find 32bit libgcc_s 13321 13322 MIPS-specific 13323 13324 * [219]15869 [mips64] No NOP after LW (with -mips1 -O0) 13325 * [220]16325 [mips64] value profiling clobbers gp on mips 13326 * [221]16357 [mipsisa64-elf] ICE copying 7 bytes between extern 13327 char[]s 13328 * [222]16380 [mips64] Use of uninitialised register after dbra 13329 conversion 13330 * [223]16407 [mips64] Unaligned access to local variables 13331 * [224]16643 [mips64] verify_local_live_at_start ICE after 13332 crossjumping & cfgcleanup 13333 13334 ARM-specific 13335 13336 * [225]15927 THUMB -O2: strength-reduced iteration variable ends up 13337 off by 1 13338 * [226]15948 THUMB: ICE with non-commutative cbranch 13339 * [227]17019 THUMB: bad switch statement in md code for 13340 addsi3_cbranch_scratch 13341 13342 IA64-specific 13343 13344 * [228]16130 ICE on valid code: in bundling, in config/ia64/ia64.c 13345 (-mtune=merced) 13346 * [229]16142 ICE on valid code: in bundling, in config/ia64/ia64.c 13347 (-mtune=itanium) 13348 * [230]16278 Gcc failed to build Linux kernel with -mtune=merced 13349 * [231]16414 ICE on valid code: typo in comparison of asm_noperands 13350 result 13351 * [232]16445 ICE on valid code: don't count ignored insns 13352 * [233]16490 ICE (segfault) while compiling with -fprofile-use 13353 * [234]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS 13354 13355 PowerPC-specific 13356 13357 * [235]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x 13358 * [236]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions 13359 issue) 13360 13361 SPARC-specific 13362 13363 * [237]16199 ICE while compiling apache 2.0.49 13364 * [238]16416 -m64 doesn't imply -mcpu=v9 anymore 13365 * [239]16430 ICE when returning non-C aggregates larger than 16 bytes 13366 13367 Bugs specific to embedded processors 13368 13369 * [240]16379 [m32r] can't output large model function call of memcpy 13370 * [241]17093 [m32r] ICE with -msdata=use -O0 13371 * [242]17119 [m32r] ICE at switch case 0x8000 13372 13373 DJGPP-specific 13374 13375 * [243]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp 13376 13377 Alpha Tru64-specific 13378 13379 * [244]16210 libstdc++ gratuitously omits "long long" I/O 13380 13381 Testsuite, documentation issues (compiler is not affected): 13382 13383 * [245]15488 (libstdc++) possibly insufficient file permissions for 13384 executing test suite 13385 * [246]16250 ada/doctools runs makeinfo even in release tarball 13386 __________________________________________________________________ 13387 13388 GCC 3.4.3 13389 13390 This is the [247]list of problem reports (PRs) from GCC's bug tracking 13391 system that are known to be fixed in the 3.4.3 release. This list might 13392 not be complete (that is, it is possible that some PRs that have been 13393 fixed are not listed here). 13394 13395 Bootstrap failures 13396 13397 * [248]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1 13398 * [249]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold 13399 when undeclared 13400 13401 Internal compiler errors (ICEs) affecting multiple platforms 13402 13403 * [250]13948 (java) GCJ segmentation fault while compiling GL4Java 13404 .class files 13405 * [251]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 13406 * [252]16301 (c++) ICE when "strong" attribute is attached to a using 13407 directive 13408 * [253]16566 ICE with flexible arrays 13409 * [254]17023 ICE with nested functions in parameter declaration 13410 * [255]17027 ICE with noreturn function in loop at -O2 13411 * [256]17524 ICE in grokdeclarator, in cp/decl.c 13412 * [257]17826 (c++) ICE in cp_tree_equal 13413 13414 C and optimization bugs 13415 13416 * [258]15526 -ftrapv aborts on 0 * (-1) 13417 * [259]16999 #ident stopped working 13418 * [260]17503 quadratic behaviour in invalid_mode_change_p 13419 * [261]17581 Long long arithmetic fails inside a switch/case 13420 statement when compiled with -O2 13421 * [262]18129 -fwritable-strings doesn't work 13422 13423 C++ compiler and library bugs 13424 13425 * [263]10975 incorrect initial ostringstream::tellp() 13426 * [264]11722 Unbuffered filebuf::sgetn is slow 13427 * [265]14534 Unrecognizing static function as a template parameter 13428 when its return value is also templated 13429 * [266]15172 Copy constructor optimization in aggregate 13430 initialization 13431 * [267]15786 Bad error message for frequently occuring error. 13432 * [268]16162 Rejects valid member-template-definition 13433 * [269]16612 empty basic_strings can't live in shared memory 13434 * [270]16715 std::basic_iostream is instantiated when used, even 13435 though instantiations are already contained in libstdc++ 13436 * [271]16848 code in /ext/demangle.h appears broken 13437 * [272]17132 GCC fails to eliminate function template specialization 13438 when argument deduction fails 13439 * [273]17259 One more _S_leaf incorrectly qualified with _RopeRep:: 13440 in ropeimpl.h 13441 * [274]17327 use of `enumeral_type' in template type unification 13442 * [275]17393 "unused variable '._0'" warning with -Wall 13443 * [276]17501 Confusion with member templates 13444 * [277]17537 g++ not passing -lstdc++ to linker when all command line 13445 arguments are libraries 13446 * [278]17585 usage of unqualified name of static member from within 13447 class not allowed 13448 * [279]17821 Poor diagnostic for using "." instead of "->" 13449 * [280]17829 wrong error: call of overloaded function is ambiguous 13450 * [281]17851 Misleading diagnostic for invalid function declarations 13451 with undeclared types 13452 * [282]17976 Destructor is called twice 13453 * [283]18020 rejects valid definition of enum value in template 13454 * [284]18093 bogus conflict in namespace aliasing 13455 * [285]18140 C++ parser bug when using >> in templates 13456 13457 Fortran 13458 13459 * [286]17541 data statements with double precision constants fail 13460 13461 x86-specific 13462 13463 * [287]17853 -O2 ICE for MMX testcase 13464 13465 SPARC-specific 13466 13467 * [288]17245 ICE compiling gsl-1.5 statistics/lag1.c 13468 13469 Darwin-specific 13470 13471 * [289]17167 FATAL:Symbol L_foo$stub already defined. 13472 13473 AIX-specific 13474 13475 * [290]17277 could not catch an exception when specified -maix64 13476 13477 Solaris-specific 13478 13479 * [291]17505 <cmath> calls acosf(), ceilf(), and other functions 13480 missing from system libraries 13481 13482 HP/UX specific: 13483 13484 * [292]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl 13485 13486 ARM-specific 13487 13488 * [293]17384 ICE with mode attribute on structures 13489 13490 MIPS-specific 13491 13492 * [294]17770 No NOP after LWL with -mips1 13493 13494 Other embedded target specific 13495 13496 * [295]11476 [arc-elf] gcc ICE on newlib's vfprintf.c 13497 * [296]14064 [avr-elf] -fdata-sections triggers ICE 13498 * [297]14678 [m68hc11-elf] gcc ICE 13499 * [298]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__ 13500 * [299]15790 [i686-coff] Alignment error building gcc with i686-coff 13501 target 13502 * [300]15886 [SH] Miscompilation with -O2 -fPIC 13503 * [301]16884 [avr-elf] [fweb related] bug while initializing 13504 variables 13505 13506 Bugs relating to debugger support 13507 13508 * [302]13841 missing debug info for _Complex function arguments 13509 * [303]15860 [big-endian targets] No DW_AT_location debug info is 13510 emitted for formal arguments to a function that uses "register" 13511 qualifiers 13512 13513 Testsuite issues (compiler not affected) 13514 13515 * [304]17465 Testsuite in libffi overrides LD_LIBRARY_PATH 13516 * [305]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH 13517 * [306]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit 13518 testsuite 13519 13520 Documentation 13521 13522 * [307]15498 typo in gcc manual: non-existing locale example en_UK, 13523 should be en_GB 13524 * [308]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap: 13525 document broken shell 13526 * [309]16406 USE_LD_AS_NEEDED undocumented 13527 __________________________________________________________________ 13528 13529 GCC 3.4.4 13530 13531 This is the [310]list of problem reports (PRs) from GCC's bug tracking 13532 system that are known to be fixed in the 3.4.4 release. This list might 13533 not be complete (that is, it is possible that some PRs that have been 13534 fixed are not listed here). 13535 __________________________________________________________________ 13536 13537 GCC 3.4.5 13538 13539 This is the [311]list of problem reports (PRs) from GCC's bug tracking 13540 system that are known to be fixed in the 3.4.5 release. This list might 13541 not be complete (that is, it is possible that some PRs that have been 13542 fixed are not listed here). 13543 13544 Bootstrap issues 13545 13546 * [312]24688 sco_math fixincl breaks math.h 13547 13548 C compiler bugs 13549 13550 * [313]17188 struct Foo { } redefinition 13551 * [314]20187 wrong code for ((unsigned char)(unsigned long 13552 long)((a?a:1)&(a*b)))?0:1) 13553 * [315]21873 infinite warning loop on bad array initializer 13554 * [316]21899 enum definition accepts values to be overriden 13555 * [317]22061 ICE in find_function_data, in function.c 13556 * [318]22308 Failure to diagnose violation of constraint 6.516p2 13557 * [319]22458 ICE on missing brace 13558 * [320]22589 ICE casting to long long 13559 * [321]24101 Segfault with preprocessed source 13560 13561 C++ compiler and library bugs 13562 13563 * [322]10611 operations on vector mode not recognized in C++ 13564 * [323]13377 unexpected behavior of namespace usage directive 13565 * [324]16002 Strange error message with new parser 13566 * [325]17413 local classes as template argument 13567 * [326]17609 spurious error message after using keyword 13568 * [327]17618 ICE in cp_convert_to_pointer, in cp/cvt.c 13569 * [328]18124 ICE with invalid template template parameter 13570 * [329]18155 typedef in template declaration not rejected 13571 * [330]18177 ICE with const_cast for undeclared variable 13572 * [331]18368 C++ error message regression 13573 * [332]16378 ICE when returning a copy of a packed member 13574 * [333]18466 int ::i; accepted 13575 * [334]18512 ICE on invalid usage of template base class 13576 * [335]18454 ICE when returning undefined type 13577 * [336]18738 typename not allowed with non-dependent qualified name 13578 * [337]18803 rejects access to operator() in template 13579 * [338]19004 ICE in uses_template_parms, in cp/pt.c 13580 * [339]19208 Spurious error about variably modified type 13581 * [340]18253 bad error message / ICE for invalid template parameter 13582 * [341]19608 ICE after friend function definition in local class 13583 * [342]19884 ICE on explicit instantiation of a non-template 13584 constructor 13585 * [343]20153 ICE when C++ template function contains anonymous union 13586 * [344]20563 Infinite loop in diagnostic (and ice after error 13587 message) 13588 * [345]20789 ICE with incomplete type in template 13589 * [346]21336 Internal compiler error when using custom new operators 13590 * [347]21768 ICE in error message due to violation of coding 13591 conventions 13592 * [348]21853 constness of pointer to data member ignored 13593 * [349]21903 Default argument of template function causes a 13594 compile-time error 13595 * [350]21983 multiple diagnostics 13596 * [351]21987 New testsuite failure 13597 g++.dg/warn/conversion-function-1.C 13598 * [352]22153 ICE on invalid template specialization 13599 * [353]22172 Internal compiler error, seg fault. 13600 * [354]21286 filebuf::xsgetn vs pipes 13601 * [355]22233 ICE with wrong number of template parameters 13602 * [356]22508 ICE after invalid operator new 13603 * [357]22545 ICE with pointer to class member & user defined 13604 conversion operator 13605 * [358]23528 Wrong default allocator in ext/hash_map 13606 * [359]23550 char_traits requirements/1.cc test bad math 13607 * [360]23586 Bad diagnostic for invalid namespace-name 13608 * [361]23624 ICE in invert_truthvalue, in fold-const.c 13609 * [362]23639 Bad error message: not a member of '<declaration error>' 13610 * [363]23797 ICE on typename outside template 13611 * [364]23965 Bogus error message: no matching function for call to 13612 'foo(<type error>)' 13613 * [365]24052 &#`label_decl' not supported by dump_expr#<expression 13614 error> 13615 * [366]24580 virtual base class cause exception not to be caught 13616 13617 Problems in generated debug information 13618 13619 * [367]24267 Bad DWARF for altivec vectors 13620 13621 Optimizations issues 13622 13623 * [368]17810 ICE in verify_local_live_at_start 13624 * [369]17860 Wrong generated code for loop with varying bound 13625 * [370]21709 ICE on compile-time complex NaN 13626 * [371]21964 broken tail call at -O2 or more 13627 * [372]22167 Strange optimization bug when using -Os 13628 * [373]22619 Compilation failure for real_const_1.f and 13629 real_const_2.f90 13630 * [374]23241 Invalid code generated for comparison of uchar to 255 13631 * [375]23478 Miscompilation due to reloading of a var that is also 13632 used in EH pad 13633 * [376]24470 segmentation fault in cc1plus when compiling with -O 13634 * [377]24950 ICE in operand_subword_force 13635 13636 Precompiled headers problems 13637 13638 * [378]14400 Cannot compile qt-x11-free-3.3.0 13639 * [379]14940 PCH largefile test fails on various platforms 13640 13641 Preprocessor bugs 13642 13643 * [380]20239 ICE on empty preprocessed input 13644 * [381]15220 "gcc -E -MM -MG" reports missing system headers in 13645 source directory 13646 13647 Testsuite issues 13648 13649 * [382]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on 13650 i686-pc-linux-gnu 13651 13652 Alpha specific 13653 13654 * [383]21888 bootstrap failure with linker relaxation enabled 13655 13656 ARM specific 13657 13658 * [384]15342 [arm-linux]: ICE in verify_local_live_at_start 13659 * [385]23985 Memory aliasing information incorrect in inlined memcpy 13660 13661 ColdFile specific 13662 13663 * [386]16719 Illegal move of byte into address register causes 13664 compiler to ICE 13665 13666 HPPA specific 13667 13668 * [387]21723 ICE while building libgfortran 13669 * [388]21841 -mhp-ld/-mgnu-ld documentation 13670 13671 IA-64 specific 13672 13673 * [389]23644 IA-64 hardware models and configuration options 13674 documentation error 13675 * [390]24718 Shared libgcc not used for linking by default 13676 13677 M68000 specific 13678 13679 * [391]18421 ICE in reload_cse_simplify_operands, in postreload.c 13680 13681 MIPS specific 13682 13683 * [392]20621 ICE in change_address_1, in emit-rtl.c 13684 13685 PowerPC and PowerPC64 specific 13686 13687 * [393]18583 error on valid code: const 13688 __attribute__((altivec(vector__))) doesn't work in arrays 13689 * [394]20191 ICE in reload_cse_simplify_operands 13690 * [395]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined 13691 * [396]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set 13692 * [397]23404 gij trashes args of functions with more than 8 fp args 13693 * [398]23539 C & C++ compiler generating misaligned references 13694 regardless of compiler flags 13695 * [399]24102 floatdisf2_internal2 broken 13696 * [400]24465 -mminimal-toc miscompilation of __thread vars 13697 13698 Solaris specific 13699 13700 * [401]19933 Problem with define of HUGE_VAL in math_c99 13701 * [402]21889 Native Solaris assembler cannot grok DTP-relative debug 13702 symbols 13703 13704 SPARC specific 13705 13706 * [403]19300 PCH failures on sparc-linux 13707 * [404]20301 Assembler labels have a leading "-" 13708 * [405]20673 C PCH testsuite assembly comparison failure 13709 13710 x86 and x86_64 specific 13711 13712 * [406]18582 ICE with arrays of type V2DF 13713 * [407]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2 13714 -fsched2-use-traces 13715 * [408]21716 ICE in reg-stack.c's swap_rtx_condition 13716 * [409]24315 amd64 fails -fpeephole2 13717 __________________________________________________________________ 13718 13719 GCC 3.4.6 13720 13721 This is the [410]list of problem reports (PRs) from GCC's bug tracking 13722 system that are known to be fixed in the 3.4.6 release. This list might 13723 not be complete (that is, it is possible that some PRs that have been 13724 fixed are not listed here). 13725 13726 13727 For questions related to the use of GCC, please consult these web 13728 pages and the [411]GCC manuals. If that fails, the 13729 [412]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13730 web pages and the development of GCC are welcome on our developer 13731 list at [413]gcc (a] gcc.gnu.org. All of [414]our lists have public 13732 archives. 13733 13734 Copyright (C) [415]Free Software Foundation, Inc. Verbatim copying and 13735 distribution of this entire article is permitted in any medium, 13736 provided this notice is preserved. 13737 13738 These pages are [416]maintained by the GCC team. Last modified 13739 2018-09-30[417]. 13740 13741 References 13742 13743 1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 13744 2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus 13745 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 13746 4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems 13747 5. http://gcc.gnu.org/gcc-3.4/mips-abi.html 13748 6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 13749 7. https://www.boost.org/ 13750 8. https://gcc.gnu.org/PR11953 13751 9. https://gcc.gnu.org/PR8361 13752 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other Builtins 13753 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209 13754 12. http://gcc.gnu.org/bugs/#cxx_rvalbind 13755 13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 13756 14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 13757 15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 13758 16. http://www.gnu.org/software/classpath/ 13759 17. http://www.eclipse.org/ 13760 18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html 13761 19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html 13762 20. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 13763 21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html 13764 22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 13765 23. http://gcc.gnu.org/gcc-3.4/mips-abi.html 13766 24. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html 13767 25. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 13768 26. https://gcc.gnu.org/bugzilla/buglist.cgi?short_desc_type=notregexp&short_desc=\[3\.4.*[Rr]egression&target_milestone=3.4.0&bug_status=RESOLVED&resolution=FIXED 13769 27. https://gcc.gnu.org/PR10129 13770 28. https://gcc.gnu.org/PR14576 13771 29. https://gcc.gnu.org/PR14760 13772 30. https://gcc.gnu.org/PR14671 13773 31. https://gcc.gnu.org/PR15093 13774 32. https://gcc.gnu.org/PR15178 13775 33. https://gcc.gnu.org/PR12753 13776 34. https://gcc.gnu.org/PR13985 13777 35. https://gcc.gnu.org/PR14810 13778 36. https://gcc.gnu.org/PR14883 13779 37. https://gcc.gnu.org/PR15044 13780 38. https://gcc.gnu.org/PR15057 13781 39. https://gcc.gnu.org/PR15064 13782 40. https://gcc.gnu.org/PR15142 13783 41. https://gcc.gnu.org/PR15159 13784 42. https://gcc.gnu.org/PR15165 13785 43. https://gcc.gnu.org/PR15193 13786 44. https://gcc.gnu.org/PR15209 13787 45. https://gcc.gnu.org/PR15227 13788 46. https://gcc.gnu.org/PR15285 13789 47. https://gcc.gnu.org/PR15299 13790 48. https://gcc.gnu.org/PR15329 13791 49. https://gcc.gnu.org/PR15550 13792 50. https://gcc.gnu.org/PR15554 13793 51. https://gcc.gnu.org/PR15640 13794 52. https://gcc.gnu.org/PR15666 13795 53. https://gcc.gnu.org/PR15696 13796 54. https://gcc.gnu.org/PR15701 13797 55. https://gcc.gnu.org/PR15761 13798 56. https://gcc.gnu.org/PR15829 13799 57. https://gcc.gnu.org/PR14538 13800 58. https://gcc.gnu.org/PR12391 13801 59. https://gcc.gnu.org/PR14649 13802 60. https://gcc.gnu.org/PR15004 13803 61. https://gcc.gnu.org/PR15749 13804 62. https://gcc.gnu.org/PR10646 13805 63. https://gcc.gnu.org/PR12077 13806 64. https://gcc.gnu.org/PR13598 13807 65. https://gcc.gnu.org/PR14211 13808 66. https://gcc.gnu.org/PR14220 13809 67. https://gcc.gnu.org/PR14245 13810 68. https://gcc.gnu.org/PR14340 13811 69. https://gcc.gnu.org/PR14600 13812 70. https://gcc.gnu.org/PR14668 13813 71. https://gcc.gnu.org/PR14775 13814 72. https://gcc.gnu.org/PR14821 13815 73. https://gcc.gnu.org/PR14930 13816 74. https://gcc.gnu.org/PR14932 13817 75. https://gcc.gnu.org/PR14950 13818 76. https://gcc.gnu.org/PR14962 13819 77. https://gcc.gnu.org/PR14975 13820 78. https://gcc.gnu.org/PR15002 13821 79. https://gcc.gnu.org/PR15025 13822 80. https://gcc.gnu.org/PR15046 13823 81. https://gcc.gnu.org/PR15069 13824 82. https://gcc.gnu.org/PR15074 13825 83. https://gcc.gnu.org/PR15083 13826 84. https://gcc.gnu.org/PR15096 13827 85. https://gcc.gnu.org/PR15287 13828 86. https://gcc.gnu.org/PR15317 13829 87. https://gcc.gnu.org/PR15337 13830 88. https://gcc.gnu.org/PR15361 13831 89. https://gcc.gnu.org/PR15412 13832 90. https://gcc.gnu.org/PR15427 13833 91. https://gcc.gnu.org/PR15471 13834 92. https://gcc.gnu.org/PR15503 13835 93. https://gcc.gnu.org/PR15507 13836 94. https://gcc.gnu.org/PR15542 13837 95. https://gcc.gnu.org/PR15565 13838 96. https://gcc.gnu.org/PR15625 13839 97. https://gcc.gnu.org/PR15629 13840 98. https://gcc.gnu.org/PR15742 13841 99. https://gcc.gnu.org/PR15775 13842 100. https://gcc.gnu.org/PR15821 13843 101. https://gcc.gnu.org/PR15862 13844 102. https://gcc.gnu.org/PR15875 13845 103. https://gcc.gnu.org/PR15877 13846 104. https://gcc.gnu.org/PR15947 13847 105. https://gcc.gnu.org/PR16020 13848 106. https://gcc.gnu.org/PR16154 13849 107. https://gcc.gnu.org/PR16174 13850 108. https://gcc.gnu.org/PR14315 13851 109. https://gcc.gnu.org/PR15151 13852 110. https://gcc.gnu.org/PR7993 13853 111. https://gcc.gnu.org/PR15228 13854 112. https://gcc.gnu.org/PR15345 13855 113. https://gcc.gnu.org/PR15945 13856 114. https://gcc.gnu.org/PR15526 13857 115. https://gcc.gnu.org/PR14690 13858 116. https://gcc.gnu.org/PR15112 13859 117. https://gcc.gnu.org/PR15067 13860 118. https://gcc.gnu.org/PR1963 13861 119. https://gcc.gnu.org/PR15717 13862 120. https://gcc.gnu.org/PR14782 13863 121. https://gcc.gnu.org/PR14828 13864 122. https://gcc.gnu.org/PR15202 13865 123. https://gcc.gnu.org/PR14610 13866 124. https://gcc.gnu.org/PR14813 13867 125. https://gcc.gnu.org/PR14857 13868 126. https://gcc.gnu.org/PR15598 13869 127. https://gcc.gnu.org/PR15653 13870 128. https://gcc.gnu.org/PR15189 13871 129. https://gcc.gnu.org/PR15331 13872 130. https://gcc.gnu.org/PR16144 13873 131. https://gcc.gnu.org/PR16176 13874 132. https://gcc.gnu.org/PR11591 13875 133. https://gcc.gnu.org/PR12028 13876 134. https://gcc.gnu.org/PR14478 13877 135. https://gcc.gnu.org/PR14567 13878 136. https://gcc.gnu.org/PR14715 13879 137. https://gcc.gnu.org/PR14902 13880 138. https://gcc.gnu.org/PR14924 13881 139. https://gcc.gnu.org/PR14960 13882 140. https://gcc.gnu.org/PR15106 13883 141. https://gcc.gnu.org/PR16026 13884 142. https://gcc.gnu.org/PR15191 13885 143. https://gcc.gnu.org/PR15662 13886 144. https://gcc.gnu.org/PR15054 13887 145. https://gcc.gnu.org/PR15783 13888 146. https://gcc.gnu.org/PR15626 13889 147. https://gcc.gnu.org/PR14326 13890 148. https://gcc.gnu.org/PR14723 13891 149. https://gcc.gnu.org/PR15290 13892 150. https://gcc.gnu.org/PR15250 13893 151. https://gcc.gnu.org/PR15551 13894 152. https://gcc.gnu.org/PR8309 13895 153. https://gcc.gnu.org/PR13250 13896 154. https://gcc.gnu.org/PR13803 13897 155. https://gcc.gnu.org/PR14093 13898 156. https://gcc.gnu.org/PR14457 13899 157. https://gcc.gnu.org/PR14542 13900 158. https://gcc.gnu.org/PR15100 13901 159. https://gcc.gnu.org/PR15296 13902 160. https://gcc.gnu.org/PR15396 13903 161. https://gcc.gnu.org/PR15782 13904 162. https://gcc.gnu.org/PR11610 13905 163. https://gcc.gnu.org/PR15488 13906 164. https://gcc.gnu.org/PR15489 13907 165. https://gcc.gnu.org/PR13928 13908 166. https://gcc.gnu.org/PR14150 13909 167. https://gcc.gnu.org/PR14949 13910 168. https://gcc.gnu.org/PR15123 13911 169. https://gcc.gnu.org/PR16469 13912 170. https://gcc.gnu.org/PR16344 13913 171. https://gcc.gnu.org/PR16842 13914 172. https://gcc.gnu.org/PR12608 13915 173. https://gcc.gnu.org/PR14492 13916 174. https://gcc.gnu.org/PR15461 13917 175. https://gcc.gnu.org/PR15890 13918 176. https://gcc.gnu.org/PR16180 13919 177. https://gcc.gnu.org/PR16224 13920 178. https://gcc.gnu.org/PR16408 13921 179. https://gcc.gnu.org/PR16529 13922 180. https://gcc.gnu.org/PR16698 13923 181. https://gcc.gnu.org/PR16706 13924 182. https://gcc.gnu.org/PR16810 13925 183. https://gcc.gnu.org/PR16851 13926 184. https://gcc.gnu.org/PR16870 13927 185. https://gcc.gnu.org/PR16904 13928 186. https://gcc.gnu.org/PR16905 13929 187. https://gcc.gnu.org/PR16964 13930 188. https://gcc.gnu.org/PR17068 13931 189. https://gcc.gnu.org/PR16366 13932 190. https://gcc.gnu.org/PR15345 13933 191. https://gcc.gnu.org/PR16590 13934 192. https://gcc.gnu.org/PR16693 13935 193. https://gcc.gnu.org/PR17078 13936 194. https://gcc.gnu.org/PR13956 13937 195. https://gcc.gnu.org/PR16684 13938 196. https://gcc.gnu.org/PR12658 13939 197. https://gcc.gnu.org/PR13092 13940 198. https://gcc.gnu.org/PR15320 13941 199. https://gcc.gnu.org/PR16246 13942 200. https://gcc.gnu.org/PR16273 13943 201. https://gcc.gnu.org/PR16401 13944 202. https://gcc.gnu.org/PR16411 13945 203. https://gcc.gnu.org/PR16489 13946 204. https://gcc.gnu.org/PR16618 13947 205. https://gcc.gnu.org/PR16637 13948 206. https://gcc.gnu.org/PR16717 13949 207. https://gcc.gnu.org/PR16813 13950 208. https://gcc.gnu.org/PR16853 13951 209. https://gcc.gnu.org/PR16889 13952 210. https://gcc.gnu.org/PR16959 13953 211. https://gcc.gnu.org/PR7587 13954 212. https://gcc.gnu.org/PR16473 13955 213. https://gcc.gnu.org/PR16478 13956 214. https://gcc.gnu.org/PR10695 13957 215. https://gcc.gnu.org/PR16974 13958 216. https://gcc.gnu.org/PR16298 13959 217. https://gcc.gnu.org/PR17113 13960 218. https://gcc.gnu.org/PR14697 13961 219. https://gcc.gnu.org/PR15869 13962 220. https://gcc.gnu.org/PR16325 13963 221. https://gcc.gnu.org/PR16357 13964 222. https://gcc.gnu.org/PR16380 13965 223. https://gcc.gnu.org/PR16407 13966 224. https://gcc.gnu.org/PR16643 13967 225. https://gcc.gnu.org/PR15927 13968 226. https://gcc.gnu.org/PR15948 13969 227. https://gcc.gnu.org/PR17019 13970 228. https://gcc.gnu.org/PR16130 13971 229. https://gcc.gnu.org/PR16142 13972 230. https://gcc.gnu.org/PR16278 13973 231. https://gcc.gnu.org/PR16414 13974 232. https://gcc.gnu.org/PR16445 13975 233. https://gcc.gnu.org/PR16490 13976 234. https://gcc.gnu.org/PR16683 13977 235. https://gcc.gnu.org/PR16195 13978 236. https://gcc.gnu.org/PR16239 13979 237. https://gcc.gnu.org/PR16199 13980 238. https://gcc.gnu.org/PR16416 13981 239. https://gcc.gnu.org/PR16430 13982 240. https://gcc.gnu.org/PR16379 13983 241. https://gcc.gnu.org/PR17093 13984 242. https://gcc.gnu.org/PR17119 13985 243. https://gcc.gnu.org/PR15928 13986 244. https://gcc.gnu.org/PR16210 13987 245. https://gcc.gnu.org/PR15488 13988 246. https://gcc.gnu.org/PR16250 13989 247. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3 13990 248. https://gcc.gnu.org/PR17369 13991 249. https://gcc.gnu.org/PR17850 13992 250. https://gcc.gnu.org/PR13948 13993 251. https://gcc.gnu.org/PR14492 13994 252. https://gcc.gnu.org/PR16301 13995 253. https://gcc.gnu.org/PR16566 13996 254. https://gcc.gnu.org/PR17023 13997 255. https://gcc.gnu.org/PR17027 13998 256. https://gcc.gnu.org/PR17524 13999 257. https://gcc.gnu.org/PR17826 14000 258. https://gcc.gnu.org/PR15526 14001 259. https://gcc.gnu.org/PR16999 14002 260. https://gcc.gnu.org/PR17503 14003 261. https://gcc.gnu.org/PR17581 14004 262. https://gcc.gnu.org/PR18129 14005 263. https://gcc.gnu.org/PR10975 14006 264. https://gcc.gnu.org/PR11722 14007 265. https://gcc.gnu.org/PR14534 14008 266. https://gcc.gnu.org/PR15172 14009 267. https://gcc.gnu.org/PR15786 14010 268. https://gcc.gnu.org/PR16162 14011 269. https://gcc.gnu.org/PR16612 14012 270. https://gcc.gnu.org/PR16715 14013 271. https://gcc.gnu.org/PR16848 14014 272. https://gcc.gnu.org/PR17132 14015 273. https://gcc.gnu.org/PR17259 14016 274. https://gcc.gnu.org/PR17327 14017 275. https://gcc.gnu.org/PR17393 14018 276. https://gcc.gnu.org/PR17501 14019 277. https://gcc.gnu.org/PR17537 14020 278. https://gcc.gnu.org/PR17585 14021 279. https://gcc.gnu.org/PR17821 14022 280. https://gcc.gnu.org/PR17829 14023 281. https://gcc.gnu.org/PR17851 14024 282. https://gcc.gnu.org/PR17976 14025 283. https://gcc.gnu.org/PR18020 14026 284. https://gcc.gnu.org/PR18093 14027 285. https://gcc.gnu.org/PR18140 14028 286. https://gcc.gnu.org/PR17541 14029 287. https://gcc.gnu.org/PR17853 14030 288. https://gcc.gnu.org/PR17245 14031 289. https://gcc.gnu.org/PR17167 14032 290. https://gcc.gnu.org/PR17277 14033 291. https://gcc.gnu.org/PR17505 14034 292. https://gcc.gnu.org/PR17684 14035 293. https://gcc.gnu.org/PR17384 14036 294. https://gcc.gnu.org/PR17770 14037 295. https://gcc.gnu.org/PR11476 14038 296. https://gcc.gnu.org/PR14064 14039 297. https://gcc.gnu.org/PR14678 14040 298. https://gcc.gnu.org/PR15583 14041 299. https://gcc.gnu.org/PR15790 14042 300. https://gcc.gnu.org/PR15886 14043 301. https://gcc.gnu.org/PR16884 14044 302. https://gcc.gnu.org/PR13841 14045 303. https://gcc.gnu.org/PR15860 14046 304. https://gcc.gnu.org/PR17465 14047 305. https://gcc.gnu.org/PR17469 14048 306. https://gcc.gnu.org/PR18138 14049 307. https://gcc.gnu.org/PR15498 14050 308. https://gcc.gnu.org/PR15747 14051 309. https://gcc.gnu.org/PR16406 14052 310. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4 14053 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5 14054 312. https://gcc.gnu.org/PR24688 14055 313. https://gcc.gnu.org/PR17188 14056 314. https://gcc.gnu.org/PR20187 14057 315. https://gcc.gnu.org/PR21873 14058 316. https://gcc.gnu.org/PR21899 14059 317. https://gcc.gnu.org/PR22061 14060 318. https://gcc.gnu.org/PR22208 14061 319. https://gcc.gnu.org/PR22458 14062 320. https://gcc.gnu.org/PR22589 14063 321. https://gcc.gnu.org/PR24101 14064 322. https://gcc.gnu.org/PR10611 14065 323. https://gcc.gnu.org/PR13377 14066 324. https://gcc.gnu.org/PR16002 14067 325. https://gcc.gnu.org/PR17413 14068 326. https://gcc.gnu.org/PR17609 14069 327. https://gcc.gnu.org/PR17618 14070 328. https://gcc.gnu.org/PR18124 14071 329. https://gcc.gnu.org/PR18155 14072 330. https://gcc.gnu.org/PR18177 14073 331. https://gcc.gnu.org/PR18368 14074 332. https://gcc.gnu.org/PR18378 14075 333. https://gcc.gnu.org/PR18466 14076 334. https://gcc.gnu.org/PR18512 14077 335. https://gcc.gnu.org/PR18545 14078 336. https://gcc.gnu.org/PR18738 14079 337. https://gcc.gnu.org/PR18803 14080 338. https://gcc.gnu.org/PR19004 14081 339. https://gcc.gnu.org/PR19208 14082 340. https://gcc.gnu.org/PR19253 14083 341. https://gcc.gnu.org/PR19608 14084 342. https://gcc.gnu.org/PR19884 14085 343. https://gcc.gnu.org/PR20153 14086 344. https://gcc.gnu.org/PR20563 14087 345. https://gcc.gnu.org/PR20789 14088 346. https://gcc.gnu.org/PR21336 14089 347. https://gcc.gnu.org/PR21768 14090 348. https://gcc.gnu.org/PR21853 14091 349. https://gcc.gnu.org/PR21903 14092 350. https://gcc.gnu.org/PR21983 14093 351. https://gcc.gnu.org/PR21987 14094 352. https://gcc.gnu.org/PR22153 14095 353. https://gcc.gnu.org/PR22172 14096 354. https://gcc.gnu.org/PR21286 14097 355. https://gcc.gnu.org/PR22233 14098 356. https://gcc.gnu.org/PR22508 14099 357. https://gcc.gnu.org/PR22545 14100 358. https://gcc.gnu.org/PR23528 14101 359. https://gcc.gnu.org/PR23550 14102 360. https://gcc.gnu.org/PR23586 14103 361. https://gcc.gnu.org/PR23624 14104 362. https://gcc.gnu.org/PR23639 14105 363. https://gcc.gnu.org/PR23797 14106 364. https://gcc.gnu.org/PR23965 14107 365. https://gcc.gnu.org/PR24052 14108 366. https://gcc.gnu.org/PR24580 14109 367. https://gcc.gnu.org/PR24267 14110 368. https://gcc.gnu.org/PR17810 14111 369. https://gcc.gnu.org/PR17860 14112 370. https://gcc.gnu.org/PR21709 14113 371. https://gcc.gnu.org/PR21964 14114 372. https://gcc.gnu.org/PR22167 14115 373. https://gcc.gnu.org/PR22619 14116 374. https://gcc.gnu.org/PR23241 14117 375. https://gcc.gnu.org/PR23478 14118 376. https://gcc.gnu.org/PR24470 14119 377. https://gcc.gnu.org/PR24950 14120 378. https://gcc.gnu.org/PR14400 14121 379. https://gcc.gnu.org/PR14940 14122 380. https://gcc.gnu.org/PR20239 14123 381. https://gcc.gnu.org/PR15220 14124 382. https://gcc.gnu.org/PR19275 14125 383. https://gcc.gnu.org/PR21888 14126 384. https://gcc.gnu.org/PR15342 14127 385. https://gcc.gnu.org/PR23985 14128 386. https://gcc.gnu.org/PR16719 14129 387. https://gcc.gnu.org/PR21723 14130 388. https://gcc.gnu.org/PR21841 14131 389. https://gcc.gnu.org/PR23644 14132 390. https://gcc.gnu.org/PR24718 14133 391. https://gcc.gnu.org/PR18421 14134 392. https://gcc.gnu.org/PR20621 14135 393. https://gcc.gnu.org/PR18583 14136 394. https://gcc.gnu.org/PR20191 14137 395. https://gcc.gnu.org/PR22083 14138 396. https://gcc.gnu.org/PR23070 14139 397. https://gcc.gnu.org/PR23404 14140 398. https://gcc.gnu.org/PR23539 14141 399. https://gcc.gnu.org/PR24102 14142 400. https://gcc.gnu.org/PR24465 14143 401. https://gcc.gnu.org/PR19933 14144 402. https://gcc.gnu.org/PR21889 14145 403. https://gcc.gnu.org/PR19300 14146 404. https://gcc.gnu.org/PR20301 14147 405. https://gcc.gnu.org/PR20673 14148 406. https://gcc.gnu.org/PR18582 14149 407. https://gcc.gnu.org/PR19340 14150 408. https://gcc.gnu.org/PR21716 14151 409. https://gcc.gnu.org/PR24315 14152 410. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6 14153 411. https://gcc.gnu.org/onlinedocs/ 14154 412. mailto:gcc-help (a] gcc.gnu.org 14155 413. mailto:gcc (a] gcc.gnu.org 14156 414. https://gcc.gnu.org/lists.html 14157 415. https://www.fsf.org/ 14158 416. https://gcc.gnu.org/about.html 14159 417. http://validator.w3.org/check/referer 14160 ====================================================================== 14161 http://gcc.gnu.org/gcc-3.3/index.html 14162 GCC 3.3 Release Series 14163 14164 May 03, 2005 14165 14166 The [1]GNU project and the GCC developers are pleased to announce the 14167 release of GCC 3.3.6. 14168 14169 This release is a bug-fix release, containing fixes for regressions in 14170 GCC 3.3.5 relative to previous releases of GCC. 14171 14172 This release is the last of the series 3.3.x. 14173 14174 The GCC 3.3 release series includes numerous [2]new features, 14175 improvements, bug fixes, and other changes, thanks to an [3]amazing 14176 group of volunteers. 14177 14178 This release series is no longer maintained. 14179 14180 Release History 14181 14182 GCC 3.3.6 14183 May 3, 2005 ([4]changes) 14184 14185 GCC 3.3.5 14186 September 30, 2004 ([5]changes) 14187 14188 GCC 3.3.4 14189 May 31, 2004 ([6]changes) 14190 14191 GCC 3.3.3 14192 February 14, 2004 ([7]changes) 14193 14194 GCC 3.3.2 14195 October 16, 2003 ([8]changes) 14196 14197 GCC 3.3.1 14198 August 8, 2003 ([9]changes) 14199 14200 GCC 3.3 14201 May 14, 2003 ([10]changes) 14202 14203 References and Acknowledgements 14204 14205 GCC used to stand for the GNU C Compiler, but since the compiler 14206 supports several other languages aside from C, it now stands for the 14207 GNU Compiler Collection. 14208 14209 A list of [11]successful builds is updated as new information becomes 14210 available. 14211 14212 The GCC developers would like to thank the numerous people that have 14213 contributed new features, improvements, bug fixes, and other changes as 14214 well as test results to GCC. This [12]amazing group of volunteers is 14215 what makes GCC successful. 14216 14217 For additional information about GCC please refer to the [13]GCC 14218 project web site or contact the [14]GCC development mailing list. 14219 14220 To obtain GCC please use [15]our mirror sites, or our CVS server. 14221 14222 14223 For questions related to the use of GCC, please consult these web 14224 pages and the [16]GCC manuals. If that fails, the 14225 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14226 web pages and the development of GCC are welcome on our developer 14227 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 14228 archives. 14229 14230 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 14231 distribution of this entire article is permitted in any medium, 14232 provided this notice is preserved. 14233 14234 These pages are [21]maintained by the GCC team. Last modified 14235 2018-09-30[22]. 14236 14237 References 14238 14239 1. http://www.gnu.org/ 14240 2. http://gcc.gnu.org/gcc-3.3/changes.html 14241 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14242 4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 14243 5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5 14244 6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4 14245 7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3 14246 8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2 14247 9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1 14248 10. http://gcc.gnu.org/gcc-3.3/changes.html 14249 11. http://gcc.gnu.org/gcc-3.3/buildstat.html 14250 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 14251 13. http://gcc.gnu.org/index.html 14252 14. mailto:gcc (a] gcc.gnu.org 14253 15. http://gcc.gnu.org/mirrors.html 14254 16. https://gcc.gnu.org/onlinedocs/ 14255 17. mailto:gcc-help (a] gcc.gnu.org 14256 18. mailto:gcc (a] gcc.gnu.org 14257 19. https://gcc.gnu.org/lists.html 14258 20. https://www.fsf.org/ 14259 21. https://gcc.gnu.org/about.html 14260 22. http://validator.w3.org/check/referer 14261 ====================================================================== 14262 http://gcc.gnu.org/gcc-3.3/changes.html 14263 GCC 3.3 Release Series 14264 Changes, New Features, and Fixes 14265 14266 The latest release in the 3.3 release series is [1]GCC 3.3.6. 14267 14268 Caveats 14269 14270 * The preprocessor no longer accepts multi-line string literals. They 14271 were deprecated in 3.0, 3.1, and 3.2. 14272 * The preprocessor no longer supports the -A- switch when appearing 14273 alone. -A- followed by an assertion is still supported. 14274 * Support for all the systems [2]obsoleted in GCC 3.1 has been 14275 removed from GCC 3.3. See below for a [3]list of systems which are 14276 obsoleted in this release. 14277 * Checking for null format arguments has been decoupled from the rest 14278 of the format checking mechanism. Programs which use the format 14279 attribute may regain this functionality by using the new [4]nonnull 14280 function attribute. Note that all functions for which GCC has a 14281 built-in format attribute, an appropriate built-in nonnull 14282 attribute is also applied. 14283 * The DWARF (version 1) debugging format has been deprecated and will 14284 be removed in a future version of GCC. Version 2 of the DWARF 14285 debugging format will continue to be supported for the foreseeable 14286 future. 14287 * The C and Objective-C compilers no longer accept the "Naming Types" 14288 extension (typedef foo = bar); it was already unavailable in C++. 14289 Code which uses it will need to be changed to use the "typeof" 14290 extension instead: typedef typeof(bar) foo. (We have removed this 14291 extension without a period of deprecation because it has caused the 14292 compiler to crash since version 3.0 and no one noticed until very 14293 recently. Thus we conclude it is not in widespread use.) 14294 * The -traditional C compiler option has been removed. It was 14295 deprecated in 3.1 and 3.2. (Traditional preprocessing remains 14296 available.) The <varargs.h> header, used for writing variadic 14297 functions in traditional C, still exists but will produce an error 14298 message if used. 14299 * GCC 3.3.1 automatically places zero-initialized variables in the 14300 .bss section on some operating systems. Versions of GNU Emacs up to 14301 (and including) 21.3 will not work correctly when using this 14302 optimization; you can use -fno-zero-initialized-in-bss to disable 14303 it. 14304 14305 General Optimizer Improvements 14306 14307 * A new scheme for accurately describing processor pipelines, the 14308 [5]DFA scheduler, has been added. 14309 * Pavel Nejedly, Charles University Prague, has contributed new file 14310 format used by the edge coverage profiler (-fprofile-arcs). 14311 The new format is robust and diagnoses common mistakes where 14312 profiles from different versions (or compilations) of the program 14313 are combined resulting in nonsensical profiles and slow code to 14314 produced with profile feedback. Additionally this format allows 14315 extra data to be gathered. Currently, overall statistics are 14316 produced helping optimizers to identify hot spots of a program 14317 globally replacing the old intra-procedural scheme and resulting in 14318 better code. Note that the gcov tool from older GCC versions will 14319 not be able to parse the profiles generated by GCC 3.3 and vice 14320 versa. 14321 * Jan Hubicka, SuSE Labs, has contributed a new superblock formation 14322 pass enabled using -ftracer. This pass simplifies the control flow 14323 of functions allowing other optimizations to do better job. 14324 He also contributed the function reordering pass 14325 (-freorder-functions) to optimize function placement using profile 14326 feedback. 14327 14328 New Languages and Language specific improvements 14329 14330 C/ObjC/C++ 14331 14332 * The preprocessor now accepts directives within macro arguments. It 14333 processes them just as if they had not been within macro arguments. 14334 * The separate ISO and traditional preprocessors have been completely 14335 removed. The front end handles either type of preprocessed output 14336 if necessary. 14337 * In C99 mode preprocessor arithmetic is done in the precision of the 14338 target's intmax_t, as required by that standard. 14339 * The preprocessor can now copy comments inside macros to the output 14340 file when the macro is expanded. This feature, enabled using the 14341 -CC option, is intended for use by applications which place 14342 metadata or directives inside comments, such as lint. 14343 * The method of constructing the list of directories to be searched 14344 for header files has been revised. If a directory named by a -I 14345 option is a standard system include directory, the option is 14346 ignored to ensure that the default search order for system 14347 directories and the special treatment of system header files are 14348 not defeated. 14349 * A few more [6]ISO C99 features now work correctly. 14350 * A new function attribute, nonnull, has been added which allows 14351 pointer arguments to functions to be specified as requiring a 14352 non-null value. The compiler currently uses this information to 14353 issue a warning when it detects a null value passed in such an 14354 argument slot. 14355 * A new type attribute, may_alias, has been added. Accesses to 14356 objects with types with this attribute are not subjected to 14357 type-based alias analysis, but are instead assumed to be able to 14358 alias any other type of objects, just like the char type. 14359 14360 C++ 14361 14362 * Type based alias analysis has been implemented for C++ aggregate 14363 types. 14364 14365 Objective-C 14366 14367 * Generate an error if Objective-C objects are passed by value in 14368 function and method calls. 14369 * When -Wselector is used, check the whole list of selectors at the 14370 end of compilation, and emit a warning if a @selector() is not 14371 known. 14372 * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime. 14373 * No longer need to include objc/objc-class.h to compile self calls 14374 in class methods (NeXT runtime only). 14375 * New -Wundeclared-selector option. 14376 * Removed selector bloating which was causing object files to be 10% 14377 bigger on average (GNU runtime only). 14378 * Using at run time @protocol() objects has been fixed in certain 14379 situations (GNU runtime only). 14380 * Type checking has been fixed and improved in many situations 14381 involving protocols. 14382 14383 Java 14384 14385 * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK 14386 1.4) API. 14387 * The JDK 1.4 assert facility has been implemented. 14388 * The bytecode interpreter is now direct threaded and thus faster. 14389 14390 Fortran 14391 14392 * Fortran improvements are listed in [7]the Fortran documentation. 14393 14394 Ada 14395 14396 * Ada tasking now works with glibc 2.3.x threading libraries. 14397 14398 New Targets and Target Specific Improvements 14399 14400 * The following changes have been made to the HP-PA port: 14401 + The port now defaults to scheduling for the PA8000 series of 14402 processors. 14403 + Scheduling support for the PA7300 processor has been added. 14404 + The 32-bit port now supports weak symbols under HP-UX 11. 14405 + The handling of initializers and finalizers has been improved 14406 under HP-UX 11. The 64-bit port no longer uses collect2. 14407 + Dwarf2 EH support has been added to the 32-bit GNU/Linux port. 14408 + ABI fixes to correct the passing of small structures by value. 14409 * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to 14410 use the DFA processor pipeline description. 14411 * The following NetBSD configurations for the SuperH processor family 14412 have been added: 14413 + SH3, big-endian, sh-*-netbsdelf* 14414 + SH3, little-endian, shle-*-netbsdelf* 14415 + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd* 14416 + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd* 14417 + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd* 14418 + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd* 14419 * The following changes have been made to the IA-32/x86-64 port: 14420 + SSE2 and 3dNOW! intrinsics are now supported. 14421 + Support for thread local storage has been added to the IA-32 14422 and x86-64 ports. 14423 + The x86-64 port has been significantly improved. 14424 * The following changes have been made to the MIPS port: 14425 + All configurations now accept the -mabi switch. Note that you 14426 will need appropriate multilibs for this option to work 14427 properly. 14428 + ELF configurations will always pass an ABI flag to the 14429 assembler, except when the MIPS EABI is selected. 14430 + -mabi=64 no longer selects MIPS IV code. 14431 + The -mcpu option, which was deprecated in 3.1 and 3.2, has 14432 been removed from this release. 14433 + -march now changes the core ISA level. In previous releases, 14434 it would change the use of processor-specific extensions, but 14435 would leave the core ISA unchanged. For example, mips64-elf 14436 -march=r8000 will now generate MIPS IV code. 14437 + Under most configurations, -mipsN now acts as a synonym for 14438 -march. 14439 + There are some new preprocessor macros to describe the -march 14440 and -mtune settings. See the documentation of those options 14441 for details. 14442 + Support for the NEC VR-Series processors has been added. This 14443 includes the 54xx, 5500, and 41xx series. 14444 + Support for the Sandcraft sr71k processor has been added. 14445 * The following changes have been made to the S/390 port: 14446 + Support to build the Java runtime libraries has been added. 14447 Java is now enabled by default on s390-*-linux* and 14448 s390x-*-linux* targets. 14449 + Multilib support for the s390x-*-linux* target has been added; 14450 this allows to build 31-bit binaries using the -m31 option. 14451 + Support for thread local storage has been added. 14452 + Inline assembler code may now use the 'Q' constraint to 14453 specify memory operands without index register. 14454 + Various platform-specific performance improvements have been 14455 implemented; in particular, the compiler now uses the BRANCH 14456 ON COUNT family of instructions and makes more frequent use of 14457 the TEST UNDER MASK family of instructions. 14458 * The following changes have been made to the PowerPC port: 14459 + Support for IBM Power4 processor added. 14460 + Support for Motorola e500 SPE added. 14461 + Support for AIX 5.2 added. 14462 + Function and Data sections now supported on AIX. 14463 + Sibcall optimizations added. 14464 * The support for H8 Tiny is added to the H8/300 port with -mn. 14465 14466 Obsolete Systems 14467 14468 Support for a number of older systems has been declared obsolete in GCC 14469 3.3. Unless there is activity to revive them, the next release of GCC 14470 will have their sources permanently removed. 14471 14472 All configurations of the following processor architectures have been 14473 declared obsolete: 14474 * Matsushita MN10200, mn10200-*-* 14475 * Motorola 88000, m88k-*-* 14476 * IBM ROMP, romp-*-* 14477 14478 Also, some individual systems have been obsoleted: 14479 * Alpha 14480 + Interix, alpha*-*-interix* 14481 + Linux libc1, alpha*-*-linux*libc1* 14482 + Linux ECOFF, alpha*-*-linux*ecoff* 14483 * ARM 14484 + Generic a.out, arm*-*-aout* 14485 + Conix, arm*-*-conix* 14486 + "Old ABI," arm*-*-oabi 14487 + StrongARM/COFF, strongarm-*-coff* 14488 * HPPA (PA-RISC) 14489 + Generic OSF, hppa1.0-*-osf* 14490 + Generic BSD, hppa1.0-*-bsd* 14491 + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]* 14492 + HiUX, hppa*-*-hiux* 14493 + Mach Lites, hppa*-*-lites* 14494 * Intel 386 family 14495 + Windows NT 3.x, i?86-*-win32 14496 * MC68000 family 14497 + HP systems, m68000-hp-bsd* and m68k-hp-bsd* 14498 + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and 14499 m68k-sun-mach* 14500 + AT&T systems, m68000-att-sysv* 14501 + Atari systems, m68k-atari-sysv* 14502 + Motorola systems, m68k-motorola-sysv* 14503 + NCR systems, m68k-ncr-sysv* 14504 + Plexus systems, m68k-plexus-sysv* 14505 + Commodore systems, m68k-cbm-sysv* 14506 + Citicorp TTI, m68k-tti-* 14507 + Unos, m68k-crds-unos* 14508 + Concurrent RTU, m68k-ccur-rtu* 14509 + Linux a.out, m68k-*-linux*aout* 14510 + Linux libc1, m68k-*-linux*libc1* 14511 + pSOS, m68k-*-psos* 14512 * MIPS 14513 + Generic ECOFF, mips*-*-ecoff* 14514 + SINIX, mips-sni-sysv4 14515 + Orion RTEMS, mips64orion-*-rtems* 14516 * National Semiconductor 32000 14517 + OpenBSD, ns32k-*-openbsd* 14518 * POWER (aka RS/6000) and PowerPC 14519 + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]* 14520 + Bull BOSX, rs6000-bull-bosx 14521 + Generic Mach, rs6000-*-mach* 14522 + Generic SysV, powerpc*-*-sysv* 14523 + Linux libc1, powerpc*-*-linux*libc1* 14524 * Sun SPARC 14525 + Generic a.out, sparc-*-aout*, sparclet-*-aout*, 14526 sparclite-*-aout*, and sparc86x-*-aout* 14527 + NetBSD a.out, sparc-*-netbsd*aout* 14528 + Generic BSD, sparc-*-bsd* 14529 + ChorusOS, sparc-*-chorusos* 14530 + Linux a.out, sparc-*-linux*aout* 14531 + Linux libc1, sparc-*-linux*libc1* 14532 + LynxOS, sparc-*-lynxos* 14533 + Solaris on HAL hardware, sparc-hal-solaris2* 14534 + SunOS versions 3 and 4, sparc-*-sunos[34]* 14535 * NEC V850 14536 + RTEMS, v850-*-rtems* 14537 * VAX 14538 + VMS, vax-*-vms* 14539 14540 Documentation improvements 14541 14542 Other significant improvements 14543 14544 * Almost all front-end dependencies in the compiler have been 14545 separated out into a set of language hooks. This should make adding 14546 a new front end clearer and easier. 14547 * One effect of removing the separate preprocessor is a small 14548 increase in the robustness of the compiler in general, and the 14549 maintainability of target descriptions. Previously target-specific 14550 built-in macros and others, such as __FAST_MATH__, had to be 14551 handled with so-called specs that were hard to maintain. Often they 14552 would fail to behave properly when conflicting options were 14553 supplied on the command line, and define macros in the user's 14554 namespace even when strict ISO compliance was requested. 14555 Integrating the preprocessor has cleanly solved these issues. 14556 * The Makefile suite now supports redirection of make install by 14557 means of the variable DESTDIR. 14558 __________________________________________________________________ 14559 14560 GCC 3.3 14561 14562 Detailed release notes for the GCC 3.3 release follow. 14563 14564 Bug Fixes 14565 14566 bootstrap failures 14567 14568 * [8]10140 cross compiler build failures: missing __mempcpy (DUP: 14569 [9]10198,[10]10338) 14570 14571 Internal compiler errors (multi-platform) 14572 14573 * [11]3581 large string causes segmentation fault in cc1 14574 * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler 14575 * [13]5533 (c++) ICE when processing std::accumulate(begin, end, 14576 init, invalid_op) 14577 * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out 14578 * [15]6412 (c++) ICE in retrieve_specialization 14579 * [16]6620 (c++) partial template specialization causes an ICE 14580 (segmentation fault) 14581 * [17]6663 (c++) ICE with attribute aligned 14582 * [18]7068 ICE with incomplete types 14583 * [19]7083 (c++) ICE using -gstabs with dodgy class derivation 14584 * [20]7647 (c++) ICE when data member has the name of the enclosing 14585 class 14586 * [21]7675 ICE in fixup_var_refs_1 14587 * [22]7718 'complex' template instantiation causes ICE 14588 * [23]8116 (c++) ICE in member template function 14589 * [24]8358 (ada) Ada compiler accesses freed memory, crashes 14590 * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation 14591 fault 14592 * [26]8564 (c++) ICE in find_function_data, in function.c 14593 * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c 14594 * [28]8766 (c++) ICE after failed initialization of static template 14595 variable 14596 * [29]8803 ICE in instantiate_virtual_regs_1, in function.c 14597 * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set 14598 * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class 14599 definition 14600 * [32]9216 (c++) ICE on missing template parameter 14601 * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c 14602 * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO 14603 loop 14604 * [35]9429 (c++) ICE in template instantiation with a pointered new 14605 operator 14606 * [36]9516 Internal error when using a big array 14607 * [37]9600 (c++) ICE with typedefs in template class 14608 * [38]9629 (c++) virtual inheritance segfault 14609 * [39]9672 (c++) ICE: Error reporting routines re-entered 14610 * [40]9749 (c++) ICE in write_expression on invalid function 14611 prototype 14612 * [41]9794 (fortran) ICE: floating point exception during constant 14613 folding 14614 * [42]9829 (c++) Missing colon in nested namespace usage causes ICE 14615 * [43]9916 (c++) ICE with noreturn function in ?: statement 14616 * [44]9936 ICE with local function and variable-length 2d array 14617 * [45]10262 (c++) cc1plus crashes with large generated code 14618 * [46]10278 (c++) ICE in parser for invalid code 14619 * [47]10446 (c++) ICE on definition of nonexistent member function of 14620 nested class in a class template 14621 * [48]10451 (c++) ICE in grokdeclarator on spurious mutable 14622 declaration 14623 * [49]10506 (c++) ICE in build_new at cp/init.c with 14624 -fkeep-inline-functions and multiple inheritance 14625 * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the 14626 precision of the declared type 14627 14628 Optimization bugs 14629 14630 * [51]2001 Inordinately long compile times in reload CSE regs 14631 * [52]2391 Exponential compilation time explosion in combine 14632 * [53]2960 Duplicate loop conditions even with -Os 14633 * [54]4046 redundant conditional branch 14634 * [55]6405 Loop-unrolling related performance regressions 14635 * [56]6798 very long compile time with large case-statement 14636 * [57]6871 const objects shouldn't be moved to .bss 14637 * [58]6909 problem w/ -Os on modified loop-2c.c test case 14638 * [59]7189 gcc -O2 -Wall does not print ``control reaches end of 14639 non-void function'' warning 14640 * [60]7642 optimization problem with signbit() 14641 * [61]8634 incorrect code for inlining of memcpy under -O2 14642 * [62]8750 Cygwin prolog generation erroneously emitting __alloca as 14643 regular function call 14644 14645 C front end 14646 14647 * [63]2161 long if-else cascade overflows parser stack 14648 * [64]4319 short accepted on typedef'd char 14649 * [65]8602 incorrect line numbers in warning messages when using 14650 inline functions 14651 * [66]9177 -fdump-translation-unit: C front end deletes function_decl 14652 AST nodes and breaks debugging dumps 14653 * [67]9853 miscompilation of non-constant structure initializer 14654 14655 c++ compiler and library 14656 14657 * [68]45 legal template specialization code is rejected (DUP: 14658 [69]3784) 14659 * [70]764 lookup failure: friend operator and dereferencing a pointer 14660 and templates (DUP: [71]5116) 14661 * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP: 14662 2863) 14663 * [73]3663 G++ doesn't check access control during template 14664 instantiation 14665 * [74]3797 gcc fails to emit explicit specialization of a template 14666 member 14667 * [75]3948 Two destructors are called when no copy destructor is 14668 defined (ABI change) 14669 * [76]4137 Conversion operator within template is not accepted 14670 * [77]4361 bogus ambiguity taking the address of a member template 14671 * [78]4802 g++ accepts illegal template code (access to private 14672 member; DUP: [79]5837) 14673 * [80]4803 inline function is used but never defined, and g++ does 14674 not object 14675 * [81]5094 Partial specialization cannot be friend? 14676 * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66 14677 * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run 14678 time 14679 * [84]7015 certain __asm__ constructs rejected 14680 * [85]7086 compile time regression (quadratic behavior in 14681 fixup_var_refs) 14682 * [86]7099 G++ doesn't set the noreturn attribute on std::exit and 14683 std::abort 14684 * [87]7247 copy constructor missing when inlining enabled (invalid 14685 optimization?) 14686 * [88]7441 string array initialization compilation time regression 14687 from seconds to minutes 14688 * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong 14689 * [90]7804 bad printing of floating point constant in warning message 14690 * [91]8099 Friend classes and template specializations 14691 * [92]8117 member function pointers and multiple inheritance 14692 * [93]8205 using declaration and multiple inheritance 14693 * [94]8645 unnecessary non-zero checks in stl_tree.h 14694 * [95]8724 explicit destructor call for incomplete class allowed 14695 * [96]8805 compile time regression with many member variables 14696 * [97]8691 -O3 and -fno-implicit-templates are incompatible 14697 * [98]8700 unhelpful error message for binding temp to reference 14698 * [99]8724 explicit destructor call for incomplete class allowed 14699 * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems 14700 * [101]9016 Failure to consistently constant fold "constant" C++ 14701 objects 14702 * [102]9053 g++ confused about ambiguity of overloaded function 14703 templates 14704 * [103]9152 undefined virtual thunks 14705 * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out 14706 * [105]9297 data corruption due to codegen bug (when copying.) 14707 * [106]9318 i/ostream::operator>>/<<(streambuf*) broken 14708 * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf 14709 * [108]9400 bogus -Wshadow warning: shadowed declaration of this in 14710 local classes 14711 * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters 14712 * [110]9425 filebuf::pbackfail broken (DUP: [111]9439) 14713 * [112]9474 GCC freezes in compiling a weird code mixing <iostream> 14714 and <iostream.h> 14715 * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1) 14716 [114][DR 231] 14717 * [115]9555 ostream inserters fail to set badbit on exception 14718 * [116]9561 ostream inserters rethrow exception of wrong type 14719 * [117]9563 ostream::sentry returns true after a failed preparation 14720 * [118]9582 one-definition rule violation in std::allocator 14721 * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors 14722 * [120]9683 bug in initialization chains for static const variables 14723 from template classes 14724 * [121]9791 -Woverloaded-virtual reports hiding of destructor 14725 * [122]9817 collate::compare doesn't handle nul characters 14726 * [123]9825 filebuf::sputbackc breaks sbumpc 14727 * [124]9826 operator>>(basic_istream, basic_string) fails to compile 14728 with custom traits 14729 * [125]9924 Multiple using statements for builtin functions not 14730 allowed 14731 * [126]9946 destructor is not called for temporary object 14732 * [127]9964 filebuf::close() sometimes fails to close file 14733 * [128]9988 filebuf::overflow writes EOF to file 14734 * [129]10033 optimization breaks polymorphic references w/ typeid 14735 operator 14736 * [130]10097 filebuf::underflow drops characters 14737 * [131]10132 filebuf destructor can throw exceptions 14738 * [132]10180 gcc fails to warn about non-inlined function 14739 * [133]10199 method parametrized by template does not work everywhere 14740 * [134]10300 use of array-new (nothrow) in segfaults on NULL return 14741 * [135]10427 Stack corruption with variable-length automatic arrays 14742 and virtual destructors 14743 * [136]10503 Compilation never stops in fixed_type_or_null 14744 14745 Objective-C 14746 14747 * [137]5956 selectors aren't matched properly when added to the 14748 selector table 14749 14750 Fortran compiler and library 14751 14752 * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't 14753 detect 14754 * [139]3924 g77 generates code that is rejected by GAS if COFF debug 14755 info requested 14756 * [140]5634 doc: explain that configure --prefix=~/... does not work 14757 * [141]6367 multiple repeat counts confuse namelist read into array 14758 * [142]6491 Logical operations error on logicals when using 14759 -fugly-logint 14760 * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C" 14761 * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os 14762 on irix6.5 14763 * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should 14764 assume a direct access file 14765 * [146]7278 g77 "bug"; the executable misbehaves (with -O2 14766 -fno-automatic) 14767 * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows 14768 * [148]7388 Incorrect output with 0-based array of characters 14769 * [149]8587 Double complex zero ** double precision number -> NaN 14770 instead of zero 14771 * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning: 14772 unknown register name line-length-none 14773 * [151]10197 Direct access files not unformatted by default 14774 14775 Java compiler and library 14776 14777 * [152]6005 gcj fails to build rhug on alpha 14778 * [153]6389 System.getProperty("") should always throw an 14779 IllegalArgumentException 14780 * [154]6576 java.util.ResourceBundle.getResource ignores locale 14781 * [155]6652 new java.io.File("").getCanonicalFile() throws exception 14782 * [156]7060 getMethod() doesn't search super interface 14783 * [157]7073 bytecode interpreter gives wrong answer for interface 14784 getSuperclass() 14785 * [158]7180 possible bug in 14786 javax.naming.spi.NamingManager.getPlusPath() 14787 * [159]7416 java.security startup refs "GNU libgcj.security" 14788 * [160]7570 Runtime.exec with null envp: child doesn't inherit parent 14789 env (DUP: [161]7578) 14790 * [162]7611 Internal error while compiling libjava with -O 14791 * [163]7709 NullPointerException in _Jv_ResolvePoolEntry 14792 * [164]7766 ZipInputStream.available returns 0 immediately after 14793 construction 14794 * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public 14795 * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented 14796 * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens' 14797 * [168]8234 ZipInputStream chokes when InputStream.read() returns 14798 small chunks 14799 * [169]8415 reflection bug: exception info for Method 14800 * [170]8481 java.Random.nextInt(int) may return negative 14801 * [171]8593 Error reading GZIPped files with BufferedReader 14802 * [172]8759 java.beans.Introspector has no flushCaches() or 14803 flushFromCaches() methods 14804 * [173]8997 spin() calls Thread.sleep 14805 * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd 14806 instead of the root content of C: 14807 * [175]9254 java::lang::Object::wait(), threads-win32.cc returns 14808 wrong return codes 14809 * [176]9271 Severe bias in java.security.SecureRandom 14810 14811 Ada compiler and library 14812 14813 * [177]6767 make gnatlib-shared fails on -laddr2line 14814 * [178]9911 gnatmake fails to link when GCC configured with 14815 --with-sjlj-exceptions=yes 14816 * [179]10020 Can't bootstrap gcc on AIX with Ada enabled 14817 * [180]10546 Ada tasking not working on Red Hat 9 14818 14819 preprocessor 14820 14821 * [181]7029 preprocessor should ignore #warning with -M 14822 14823 ARM-specific 14824 14825 * [182]2903 [arm] Optimization bug with long long arithmetic 14826 * [183]7873 arm-linux-gcc fails when assigning address to a bit field 14827 14828 FreeBSD-specific 14829 14830 * [184]7680 float functions undefined in math.h/cmath with #define 14831 _XOPEN_SOURCE 14832 14833 HP-UX or HP-PA-specific 14834 14835 * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c 14836 * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to 14837 fputc_unlocked 14838 * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen 14839 14840 m68hc11-specific 14841 14842 * [188]6744 Bad assembler code generated: reference to pseudo 14843 register z 14844 * [189]7361 Internal compiler error in reload_cse_simplify_operands, 14845 in reload1.c 14846 14847 MIPS-specific 14848 14849 * [190]9496 [mips-linux] bug in optimizer? 14850 14851 PowerPC-specific 14852 14853 * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of 14854 space 14855 * [192]8480 reload ICEs for LAPACK code on powerpc64-linux 14856 * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg 14857 * [194]10315 [powerpc] ICE: in extract_insn, in recog.c 14858 14859 SPARC-specific 14860 14861 * [195]10267 (documentation) Wrong build instructions for 14862 *-*-solaris2* 14863 14864 x86-specific (Intel/AMD) 14865 14866 * [196]7916 ICE in instantiate_virtual_register_1 14867 * [197]7926 (c++) i486 instructions in header files make c++ programs 14868 crash on i386 14869 * [198]8555 ICE in gen_split_1231 14870 * [199]8994 ICE with -O -march=pentium4 14871 * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs 14872 * [201]9806 ICE in inline assembly with -fPIC flag 14873 * [202]10077 gcc -msse2 generates movd to move dwords between xmm 14874 regs 14875 * [203]10233 64-bit comparison only comparing bottom 32-bits 14876 * [204]10286 type-punning doesn't work with __m64 and -O 14877 * [205]10308 [x86] ICE with -O -fgcse or -O2 14878 __________________________________________________________________ 14879 14880 GCC 3.3.1 14881 14882 Bug Fixes 14883 14884 This section lists the problem reports (PRs) from GCC's bug tracking 14885 system that are known to be fixed in the 3.3.1 release. This list might 14886 not be complete (that is, it is possible that some PRs that have been 14887 fixed are not listed here). 14888 14889 Bootstrap failures 14890 14891 * [206]11272 [Solaris] make bootstrap fails while building libstdc++ 14892 14893 Internal compiler errors (multi-platform) 14894 14895 * [207]5754 ICE on invalid nested template class 14896 * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64 14897 and --enable-checking 14898 * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c 14899 * [210]7053 (c++) ICE when declaring a function already defined as a 14900 friend method of a template class 14901 * [211]8164 (c++) ICE when using different const expressions as 14902 template parameter 14903 * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c 14904 * [213]9559 (c++) ICE with invalid initialization of a static const 14905 * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c 14906 when redeclaring a static member variable 14907 * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in 14908 dwarfout.c with -g -O -finline-functions 14909 * [216]10432 (c++) ICE in poplevel, in cp/decl.c 14910 * [217]10475 ICE in subreg_highpart_offset for code with long long 14911 * [218]10635 (c++) ICE when dereferencing an incomplete type casted 14912 from a void pointer 14913 * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while 14914 instantiating static member variables 14915 * [220]10700 ICE in copy_to_mode_reg on 64-bit targets 14916 * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c 14917 * [222]10796 (c++) ICE when defining an enum with two values: -1 and 14918 MAX_INT_64BIT 14919 * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x 14920 sched.c 14921 * [224]10939 (c++) ICE with template code 14922 * [225]10956 (c++) ICE when specializing a template member function 14923 of a template class, in tsubst, in cp/pt.c 14924 * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*() 14925 defined) 14926 * [227]11059 (c++) ICE with empty union 14927 * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with 14928 -O2 -fnon-call-exceptions 14929 * [229]11105 (c++) ICE in mangle_conv_op_name_for_type 14930 * [230]11149 (c++) ICE on error when instantiation with call function 14931 of a base type 14932 * [231]11228 (c++) ICE on new-expression using array operator new and 14933 default-initialization 14934 * [232]11282 (c++) Infinite memory usage after syntax error 14935 * [233]11301 (fortran) ICE with -fno-globals 14936 * [234]11308 (c++) ICE when using an enum type name as if it were a 14937 class or namespace 14938 * [235]11473 (c++) ICE with -gstabs when empty struct inherits from 14939 an empty struct 14940 * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR 14941 * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c: 14942 template member functions 14943 14944 Optimization bugs 14945 14946 * [238]11198 -O2 -frename-registers generates wrong code (aliasing 14947 problem) 14948 * [239]11304 Wrong code production with -fomit-frame-pointer 14949 * [240]11381 volatile memory access optimized away 14950 * [241]11536 [strength-reduce] -O2 optimization produces wrong code 14951 * [242]11557 constant folding bug generates wrong code 14952 14953 C front end 14954 14955 * [243]5897 No warning for statement after return 14956 * [244]11279 DWARF-2 output mishandles large enums 14957 14958 Preprocessor bugs 14959 14960 * [245]11022 no warning for non-compatible macro redefinition 14961 14962 C++ compiler and library 14963 14964 * [246]2330 static_cast<>() to a private base is allowed 14965 * [247]5388 Incorrect message "operands to ?: have different types" 14966 * [248]5390 Libiberty fails to demangle multi-digit template 14967 parameters 14968 * [249]7877 Incorrect parameter passing to specializations of member 14969 function templates 14970 * [250]9393 Anonymous namespaces and compiling the same file twice 14971 * [251]10032 -pedantic converts some errors to warnings 14972 * [252]10468 const typeof(x) is non-const, but only in templates 14973 * [253]10527 confused error message with "new int()" parameter 14974 initializer 14975 * [254]10679 parameter MIN_INLINE_INSNS is not honored 14976 * [255]10682 gcc chokes on a typedef for an enum inside a class 14977 template 14978 * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of 14979 0. 14980 * [257]10845 template member function (with nested template as 14981 parameter) cannot be called anymore if another unrelated template 14982 member function is defined 14983 * [258]10849 Cannot define an out-of-class specialization of a 14984 private nested template class 14985 * [259]10888 Suppress -Winline warnings for system headers 14986 * [260]10929 -Winline warns about functions for which no definition 14987 is visible 14988 * [261]10931 valid conversion static_cast<const unsigned 14989 int&>(lvalue-of-type-int) is rejected 14990 * [262]10940 Bad code with explicit specialization 14991 * [263]10968 If member function implicitly instantiated, explicit 14992 instantiation of class fails to instantiate it 14993 * [264]10990 Cannot convert with dynamic_cast<> to a private base 14994 class from within a member function 14995 * [265]11039 Bad interaction between implicit typename deprecation 14996 and friendship 14997 * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say 14998 "__unused__" instead 14999 * [267]11095 C++ iostream manipulator causes segfault when called 15000 with negative argument 15001 * [268]11098 g++ doesn't emit complete debugging information for 15002 local variables in destructors 15003 * [269]11137 GNU/Linux shared library constructors not called unless 15004 there's one global object 15005 * [270]11154 spurious ambiguity report for template class 15006 specialization 15007 * [271]11329 Compiler cannot find user defined implicit typecast 15008 * [272]11332 Spurious error with casts in ?: expression 15009 * [273]11431 static_cast behavior with subclasses when default 15010 constructor available 15011 * [274]11528 money_get facet does not accept "$.00" as valid 15012 * [275]11546 Type lookup problems in out-of-line definition of a 15013 class doubly nested from a template class 15014 * [276]11567 C++ code containing templated member function with same 15015 name as pure virtual member function results in linking failure 15016 * [277]11645 Failure to deal with using and private inheritance 15017 15018 Java compiler and library 15019 15020 * [278]5179 Qualified static field access doesn't initialize its 15021 class 15022 * [279]8204 gcj -O2 to native reorders certain instructions 15023 improperly 15024 * [280]10838 java.io.ObjectInputStream syntax error 15025 * [281]10886 The RMI registry that comes with GCJ does not work 15026 correctly 15027 * [282]11349 JNDI URL context factories not located correctly 15028 15029 x86-specific (Intel/AMD) 15030 15031 * [283]4823 ICE on inline assembly code 15032 * [284]8878 miscompilation with -O and SSE 15033 * [285]9815 (c++ library) atomicity.h - fails to compile with -O3 15034 -masm=intel 15035 * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads, 15036 in reload1.c 15037 * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2 15038 * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source 15039 * [289]11044 [x86] out of range loop instructions for FP code on K6 15040 * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE 15041 built-ins 15042 * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC" 15043 is used 15044 15045 SPARC- or Solaris- specific 15046 15047 * [292]9362 solaris 'as' dies when fed .s and "-gstabs" 15048 * [293]10142 [SPARC64] gcc produces wrong code when passing 15049 structures by value 15050 * [294]10663 New configure check aborts with Sun tools. 15051 * [295]10835 combinatorial explosion in scheduler on HyperSPARC 15052 * [296]10876 ICE in calculate_giv_inc when building KDE 15053 * [297]10955 wrong code at -O3 for structure argument in context of 15054 structure return 15055 * [298]11018 -mcpu=ultrasparc busts tar-1.13.25 15056 * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x 15057 Linux kernel 15058 15059 ia64 specific 15060 15061 * [300]10907 gcc violates the ia64 ABI (GP must be preserved) 15062 * [301]11320 scheduler bug (in machine depended reorganization pass) 15063 * [302]11599 bug with conditional and __builtin_prefetch 15064 15065 PowerPC specific 15066 15067 * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem 15068 during loop) 15069 * [304]10871 error in rs6000_stack_info save_size computation 15070 * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse 15071 cures it 15072 15073 m68k-specific 15074 15075 * [306]7594 [m68k] ICE on legal code associated with simplify-rtx 15076 * [307]10557 [m68k] ICE in subreg_offset_representable_p 15077 * [308]11054 [m68k] ICE in reg_overlap_mentioned_p 15078 15079 ARM-specific 15080 15081 * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for 15082 functions with __attribute__ ((interrupt ("IRQ"))) 15083 * [310]10842 [arm] Clobbered link register is copied to pc under 15084 certain circumstances 15085 * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes 15086 * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno 15087 (3.4) 15088 15089 MIPS-specific 15090 15091 * [313]11084 ICE in propagate_one_insn, in flow.c 15092 15093 SH-specific 15094 15095 * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf 15096 * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c 15097 * [316]11096 i686-linux to sh-linux cross compiler fails to compile 15098 C++ files 15099 15100 GNU/Linux (or Hurd?) specific 15101 15102 * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3 15103 15104 UnixWare specific 15105 15106 * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare 15107 7.1.1 15108 15109 Cygwin (or mingw) specific 15110 15111 * [319]5287 ICE with dllimport attribute 15112 * [320]10148 [MingW/CygWin] Compiler dumps core 15113 15114 DJGPP specific 15115 15116 * [321]8787 GCC fails to emit .intel_syntax when invoked with 15117 -masm=intel on DJGPP 15118 15119 Darwin (and MacOS X) specific 15120 15121 * [322]10900 trampolines crash 15122 15123 Documentation 15124 15125 * [323]1607 (c++) Format attributes on methods undocumented 15126 * [324]4252 Invalid option `-fdump-translation-unit' 15127 * [325]4490 Clarify restrictions on -m96bit-long-double, 15128 -m128bit-long-double 15129 * [326]10355 document an issue with regparm attribute on some systems 15130 (e.g. Solaris) 15131 * [327]10726 (fortran) Documentation for function "IDate Intrinsic 15132 (Unix)" is wrong 15133 * [328]10805 document bug in old version of Sun assembler 15134 * [329]10815 warn against GNU binutils on AIX 15135 * [330]10877 document need for newer binutils on i?86-*-linux-gnu 15136 * [331]11280 Manual incorrect with respect to -freorder-blocks 15137 * [332]11466 Document -mlittle-endian and its restrictions for the 15138 sparc64 port 15139 15140 Testsuite bugs (compiler itself is not affected) 15141 15142 * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly 15143 report failure 15144 * [334]10810 gcc-3.3 fails make check: buffer overrun in 15145 test_demangle.c 15146 __________________________________________________________________ 15147 15148 GCC 3.3.2 15149 15150 Bug Fixes 15151 15152 This section lists the problem reports (PRs) from GCC's bug tracker 15153 that are known to be fixed in the 3.3.2 release. This list might not be 15154 complete (that is, it is possible that some PRs that have been fixed 15155 are not listed here). 15156 15157 Bootstrap failures and problems 15158 15159 * [335]8336 [SCO5] bootstrap config still tries to use COFF options 15160 * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with 15161 --enable-threads=posix 15162 * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap 15163 * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare 15164 7.1.1) 15165 * [339]11687 xstormy16-elf build fails in libf2c 15166 * [340]12263 [SGI IRIX] bootstrap fails during compile of 15167 libf2c/libI77/backspace.c 15168 * [341]12490 buffer overflow in scan-decls.c (during Solaris 9 15169 fix-header processing) 15170 15171 Internal compiler errors (multi-platform) 15172 15173 * [342]7277 Casting integers to vector types causes ICE 15174 * [343]7939 (c++) ICE on invalid function template specialization 15175 * [344]11063 (c++) ICE on parsing initialization list of const array 15176 member 15177 * [345]11207 ICE with negative index in array element designator 15178 * [346]11522 (fortran) g77 dwarf-2 ICE in 15179 add_abstract_origin_attribute 15180 * [347]11595 (c++) ICE on duplicate label definition 15181 * [348]11646 (c++) ICE in commit_one_edge_insertion with 15182 -fnon-call-exceptions -fgcse -O 15183 * [349]11665 ICE in struct initializer when taking address 15184 * [350]11852 (c++) ICE with bad struct initializer. 15185 * [351]11878 (c++) ICE in cp_expr_size 15186 * [352]11883 ICE with any -O on mercury-generated C code 15187 * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in 15188 cp/typeck2.c when applying typeid operator to template template 15189 parameter 15190 * [354]12146 ICE in lookup_template_function, in cp/pt.c 15191 * [355]12215 ICE in make_label_edge with -fnon-call-exceptions 15192 -fno-gcse -O2 15193 * [356]12369 (c++) ICE with templates and friends 15194 * [357]12446 ICE in emit_move_insn on complicated array reference 15195 * [358]12510 ICE in final_scan_insn 15196 * [359]12544 ICE with large parameters used in nested functions 15197 15198 C and optimization bugs 15199 15200 * [360]9862 spurious warnings with -W -finline-functions 15201 * [361]10962 lookup_field is a linear search on a linked list (can be 15202 slow if large struct) 15203 * [362]11370 -Wunreachable-code gives false complaints 15204 * [363]11637 invalid assembly with -fnon-call-exceptions 15205 * [364]11885 Problem with bitfields in packed structs 15206 * [365]12082 Inappropriate unreachable code warnings 15207 * [366]12180 Inline optimization fails for variadic function 15208 * [367]12340 loop unroller + gcse produces wrong code 15209 15210 C++ compiler and library 15211 15212 * [368]3907 nested template parameter collides with member name 15213 * [369]5293 confusing message when binding a temporary to a reference 15214 * [370]5296 [DR115] Pointers to functions and to template functions 15215 behave differently in deduction 15216 * [371]7939 ICE on function template specialization 15217 * [372]8656 Unable to assign function with __attribute__ and pointer 15218 return type to an appropriate variable 15219 * [373]10147 Confusing error message for invalid template function 15220 argument 15221 * [374]11400 std::search_n() makes assumptions about Size parameter 15222 * [375]11409 issues with using declarations, overloading, and 15223 built-in functions 15224 * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle 15225 multiple bits in mask 15226 * [377]11786 operator() call on variable in other namespace not 15227 recognized 15228 * [378]11867 static_cast ignores ambiguity 15229 * [379]11928 bug with conversion operators that are typedefs 15230 * [380]12114 Uninitialized memory accessed in dtor 15231 * [381]12163 static_cast + explicit constructor regression 15232 * [382]12181 Wrong code with comma operator and c++ 15233 * [383]12236 regparm and fastcall messes up parameters 15234 * [384]12266 incorrect instantiation of unneeded template during 15235 overload resolution 15236 * [385]12296 istream::peek() doesn't set eofbit 15237 * [386]12298 [sjlj exceptions] Stack unwind destroys 15238 not-yet-constructed object 15239 * [387]12369 ICE with templates and friends 15240 * [388]12337 apparently infinite loop in g++ 15241 * [389]12344 stdcall attribute ignored if function returns a pointer 15242 * [390]12451 missing(late) class forward declaration in cxxabi.h 15243 * [391]12486 g++ accepts invalid use of a qualified name 15244 15245 x86 specific (Intel/AMD) 15246 15247 * [392]8869 [x86 MMX] ICE with const variable optimization and MMX 15248 builtins 15249 * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions 15250 -O2 15251 * [394]11689 g++3.3 emits un-assembleable code for k6 architecture 15252 * [395]12116 [k6] Invalid assembly output values with X-MAME code 15253 * [396]12070 ICE converting between double and long double with 15254 -msoft-float 15255 15256 ia64-specific 15257 15258 * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc 15259 * [398]11535 __builtin_return_address may not work on ia64 15260 * [399]11693 [ia64] ICE in gen_nop_type 15261 * [400]12224 [ia64] Thread-local storage doesn't work 15262 15263 PowerPC-specific 15264 15265 * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux 15266 kernel 15267 * [402]11319 loop miscompiled on ppc32 15268 * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code 15269 15270 SPARC-specific 15271 15272 * [404]11662 wrong code for expr. with cast to long long and 15273 exclusive or 15274 * [405]11965 invalid assembler code for a shift < 32 operation 15275 * [406]12301 (c++) stack corruption when a returned expression throws 15276 an exception 15277 15278 Alpha-specific 15279 15280 * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of 15281 kernel 2.4.22-pre8 15282 15283 HPUX-specific 15284 15285 * [408]11313 problem with #pragma weak and static inline functions 15286 * [409]11712 __STDC_EXT__ not defined for C++ by default anymore? 15287 15288 Solaris specific 15289 15290 * [410]12166 Profiled programs crash if PROFDIR is set 15291 15292 Solaris-x86 specific 15293 15294 * [411]12101 i386 Solaris no longer works with GNU as? 15295 15296 Miscellaneous embedded target-specific bugs 15297 15298 * [412]10988 [m32r-elf] wrong blockmove code with -O3 15299 * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with 15300 -O2 15301 * [414]11902 [sh4] spec file improperly inserts rpath even when none 15302 needed 15303 * [415]11903 [sh4] -pthread fails to link due to error in spec file 15304 on sh4 15305 __________________________________________________________________ 15306 15307 GCC 3.3.3 15308 15309 Minor features 15310 15311 In addition to the bug fixes documented below, this release contains 15312 few minor features such as: 15313 * Support for --with-sysroot 15314 * Support for automatic detection of executable stacks 15315 * Support for SSE3 instructions 15316 * Support for thread local storage debugging under GDB on S390 15317 15318 Bug Fixes 15319 15320 This section lists the problem reports (PRs) from GCC's bug tracker 15321 that are known to be fixed in the 3.3.3 release. This list might not be 15322 complete (that is, it is possible that some PRs that have been fixed 15323 are not listed here). 15324 15325 Bootstrap failures and issues 15326 15327 * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails 15328 * [417]12399 boehm-gc fails (when building a cross compiler): libtool 15329 unable to infer tagged configuration 15330 * [418]13068 mklibgcc.in doesn't handle multi-level multilib 15331 subdirectories properly 15332 15333 Internal compiler errors (multi-platform) 15334 15335 * [419]10060 ICE (stack overflow) on huge file (300k lines) due to 15336 recursive behaviour of copy_rtx_if_shared, in emit_rtl.c 15337 * [420]10555 (c++) ICE on undefined template argument 15338 * [421]10706 (c++) ICE in mangle_class_name_for_template 15339 * [422]11496 (fortran) error in flow_loops_find when -funroll-loops 15340 active 15341 * [423]11741 ICE in pre_insert_copy_insn, in gcse.c 15342 * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0 15343 * [425]12632 (fortran) -fbounds-check ICE 15344 * [426]12712 (c++) ICE on short legit C++ code fragment with gcc 15345 3.3.2 15346 * [427]12726 (c++) ICE (segfault) on trivial code 15347 * [428]12890 (c++) ICE on compilation of class with throwing method 15348 * [429]12900 (c++) ICE in rtl_verify_flow_info_1 15349 * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on 15350 correct code with -O2 -fno-force-mem 15351 * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive 15352 template 15353 * [432]13318 ICE: floating point exception in the loop optimizer 15354 * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in 15355 except.c 15356 * [434]13574 (c++) invalid array default initializer in class lets 15357 gcc consume all memory and die 15358 * [435]13475 ICE on SIMD variables with partial value initialization 15359 * [436]13797 (c++) ICE on invalid template parameter 15360 * [437]13824 (java) gcj SEGV with simple .java program 15361 15362 C and optimization bugs 15363 15364 * [438]8776 loop invariants are not removed (most likely) 15365 * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing 15366 strncmp by memcmp 15367 * [440]11350 undefined labels with -Os -fPIC 15368 * [441]12826 Optimizer removes reference through volatile pointer 15369 * [442]12500 stabs debug info: void no longer a predefined / builtin 15370 type 15371 * [443]12941 builtin-bitops-1.c miscompilation (latent bug) 15372 * [444]12953 tree inliner bug (in inline_forbidden_p) and fix 15373 * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled 15374 * [446]13507 spurious printf format warning 15375 * [447]13382 Type information for const pointer disappears during 15376 optimization. 15377 * [448]13394 noreturn attribute ignored on recursive invokation 15378 * [449]13400 Compiled code crashes storing to read-only location 15379 * [450]13521 Endless loop in calculate_global_regs_live 15380 15381 C++ compiler and library 15382 15383 Some of the bug fixes in this list were made to implement decisions 15384 that the ISO C++ standards committee has made concerning several defect 15385 reports (DRs). Links in the list below point to detailed discussion of 15386 the relevant defect report. 15387 * [451]2094 unimplemented: use of `ptrmem_cst' in template type 15388 unification 15389 * [452]2294 using declaration confusion 15390 * [453]5050 template instantiation depth exceeds limit: recursion 15391 problem? 15392 * [454]9371 Bad exception handling in 15393 i/ostream::operator>>/<<(streambuf*) 15394 * [455]9546 bad exception handling in ostream members 15395 * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the 15396 face of unknown locales 15397 * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work 15398 * [459]10095 istream::operator>>(int&) sets ios::badbit when 15399 ios::failbit is set. 15400 * [460]11554 Warning about reordering of initializers doesn't mention 15401 location of constructor 15402 * [461]12297 istream::sentry::sentry() handles eof() incorrectly. 15403 * [462]12352 Exception safety problems in src/localename.cc 15404 * [463]12438 Memory leak in locale::combine() 15405 * [464]12540 Memory leak in locale::locale(const char*) 15406 * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented 15407 * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented 15408 * [470]12696 memory eating infinite loop in diagnostics (error 15409 recovery problem) 15410 * [471]12815 Code compiled with optimization behaves unexpectedly 15411 * [472]12862 Conflicts between typedefs/enums and namespace member 15412 declarations 15413 * [473]12926 Wrong value after assignment in initialize list using 15414 bit-fields 15415 * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented 15416 * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented 15417 * [478]13007 basic_streambuf::pubimbue, imbue wrong 15418 * [479]13009 Implicitly-defined assignment operator writes to wrong 15419 memory 15420 * [480]13057 regparm attribute not applied to destructor 15421 * [481]13070 -Wformat option ignored in g++ 15422 * [482]13081 forward template declarations in <complex> let inlining 15423 fail 15424 * [483]13239 Assertion does not seem to work correctly anymore 15425 * [484]13262 "xxx is private within this context" when initializing a 15426 self-contained template class 15427 * [485]13290 simple typo in concept checking for std::generate_n 15428 * [486]13323 Template code does not compile in presence of typedef 15429 * [487]13369 __verify_grouping (and __add_grouping?) not correct 15430 * [488]13371 infinite loop with packed struct and inlining 15431 * [489]13445 Template argument replacement "dereferences" a typedef 15432 * [490]13461 Fails to access protected-ctor from public constant 15433 * [491]13462 Non-standard-conforming type set::pointer 15434 * [492]13478 gcc uses wrong constructor to initialize a const 15435 reference 15436 * [493]13544 "conflicting types" for enums in different scopes 15437 * [494]13650 string::compare should not (always) use 15438 traits_type::length() 15439 * [495]13683 bogus warning about passing non-PODs through ellipsis 15440 * [496]13688 Derived class is denied access to protected base class 15441 member class 15442 * [497]13774 Member variable cleared in virtual multiple inheritance 15443 class 15444 * [498]13884 Protect sstream.tcc from extern template use 15445 15446 Java compiler and library 15447 15448 * [499]10746 [win32] garbage collection crash in GCJ 15449 15450 Objective-C compiler and library 15451 15452 * [500]11433 Crash due to dereferencing null pointer when querying 15453 protocol 15454 15455 Fortran compiler and library 15456 15457 * [501]12633 logical expression gives incorrect result with 15458 -fugly-logint option 15459 * [502]13037 [gcse-lm] g77 generates incorrect code 15460 * [503]13213 Hex constant problem when compiling with -fugly-logint 15461 and -ftypeless-boz 15462 15463 x86-specific (Intel/AMD) 15464 15465 * [504]4490 ICE with -m128bit-long-double 15466 * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have 15467 `reg' in make_field_assignment, in combine.c 15468 * [506]12441 ICE: can't find a register to spill 15469 * [507]12943 array static-init failure under -fpic, -fPIC 15470 * [508]13608 Incorrect code with -O3 -ffast-math 15471 15472 PowerPC-specific 15473 15474 * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of 15475 __attribute__((aligned(16))) 15476 * [510]11793 ICE in extract_insn, in recog.c (const_vector's) 15477 * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in 15478 altivec.md) 15479 * [512]12537 g++ generates writeable text sections 15480 15481 SPARC-specific 15482 15483 * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0 15484 -m64 15485 * [514]12865 mprotect call to make trampoline executable may fail 15486 * [515]13354 ICE in sparc_emit_set_const32 15487 15488 ARM-specific 15489 15490 * [516]10467 [arm] ICE in pre_insert_copy_insn, 15491 15492 ia64-specific 15493 15494 * [517]11226 ICE passing struct arg with two floats 15495 * [518]11227 ICE for _Complex float, _Complex long double args 15496 * [519]12644 GCC 3.3.2 fails to compile glibc on ia64 15497 * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn 15498 * Various fixes for libunwind 15499 15500 Alpha-specific 15501 15502 * [521]12654 Incorrect comparison code generated for Alpha 15503 * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2 15504 * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2 15505 15506 HPPA-specific 15507 15508 * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c 15509 * [525]12158 [hppa] compilation does not terminate at -O1 15510 15511 S390-specific 15512 15513 * [526]11992 Wrong built-in code for memcmp with length 1<<24: only 15514 (1<<24)-1 possible for CLCL-Instruction 15515 15516 SH-specific 15517 15518 * [527]9365 segfault in gen_far_branch (config/sh/sh.c) 15519 * [528]10392 optimizer generates faulty array indexing 15520 * [529]11322 SH profiler outputs multiple definitions of symbol 15521 * [530]13069 gcc/config/sh/rtems.h broken 15522 * [531]13302 Putting a va_list in a struct causes seg fault 15523 * [532]13585 Incorrect optimization of call to sfunc 15524 * Fix inappropriately exported libgcc functions from the shared 15525 library 15526 15527 Other embedded target specific 15528 15529 * [533]8916 [mcore] unsigned char assign gets hosed. 15530 * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c 15531 * [535]13122 [h8300] local variable gets corrupted by function call 15532 when -fomit-frame-pointer is given 15533 * [536]13256 [cris] strict_low_part mistreated in delay slots 15534 * [537]13373 [mcore] optimization with -frerun-cse-after-loop 15535 -fexpensive-optimizations produces wrong code on mcore 15536 15537 GNU HURD-specific 15538 15539 * [538]12561 gcc/config/t-gnu needs updating to work with 15540 --with-sysroot 15541 15542 Tru64 Unix specific 15543 15544 * [539]6243 testsuite fails almost all tests due to no libintl in 15545 LD_LIBRARY_PATH during test. 15546 * [540]11397 weak aliases broken on Tru64 UNIX 15547 15548 AIX-specific 15549 15550 * [541]12505 build failure due to defines of uchar in cpphash.h and 15551 sys/types.h 15552 * [542]13150 WEAK symbols not exported by collect2 15553 15554 IRIX-specific 15555 15556 * [543]12666 fixincludes problem on IRIX 6.5.19m 15557 15558 Solaris-specific 15559 15560 * [544]12969 Including sys/byteorder.h breaks configure checks 15561 15562 Testsuite problems (compiler is not affected) 15563 15564 * [545]10819 testsuite creates CR+LF on compiler version lines in 15565 test summary files 15566 * [546]11612 abi_check not finding correct libgcc_s.so.1 15567 15568 Miscellaneous 15569 15570 * [547]13211 using -###, incorrect warnings about unused linker file 15571 are produced 15572 __________________________________________________________________ 15573 15574 GCC 3.3.4 15575 15576 This is the [548]list of problem reports (PRs) from GCC's bug tracking 15577 system that are known to be fixed in the 3.3.4 release. This list might 15578 not be complete (that is, it is possible that some PRs that have been 15579 fixed are not listed here). 15580 __________________________________________________________________ 15581 15582 GCC 3.3.5 15583 15584 This is the [549]list of problem reports (PRs) from GCC's bug tracking 15585 system that are known to be fixed in the 3.3.5 release. This list might 15586 not be complete (that is, it is possible that some PRs that have been 15587 fixed are not listed here). 15588 __________________________________________________________________ 15589 15590 GCC 3.3.6 15591 15592 This is the [550]list of problem reports (PRs) from GCC's bug tracking 15593 system that are known to be fixed in the 3.3.6 release. This list might 15594 not be complete (that is, it is possible that some PRs that have been 15595 fixed are not listed here). 15596 15597 15598 For questions related to the use of GCC, please consult these web 15599 pages and the [551]GCC manuals. If that fails, the 15600 [552]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15601 web pages and the development of GCC are welcome on our developer 15602 list at [553]gcc (a] gcc.gnu.org. All of [554]our lists have public 15603 archives. 15604 15605 Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and 15606 distribution of this entire article is permitted in any medium, 15607 provided this notice is preserved. 15608 15609 These pages are [556]maintained by the GCC team. Last modified 15610 2018-09-30[557]. 15611 15612 References 15613 15614 1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 15615 2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems 15616 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 15617 4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute 15618 5. http://gcc.gnu.org/news/dfa.html 15619 6. http://gcc.gnu.org/c99status.html 15620 7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html 15621 8. https://gcc.gnu.org/PR10140 15622 9. https://gcc.gnu.org/PR10198 15623 10. https://gcc.gnu.org/PR10338 15624 11. https://gcc.gnu.org/PR3581 15625 12. https://gcc.gnu.org/PR4382 15626 13. https://gcc.gnu.org/PR5533 15627 14. https://gcc.gnu.org/PR6387 15628 15. https://gcc.gnu.org/PR6412 15629 16. https://gcc.gnu.org/PR6620 15630 17. https://gcc.gnu.org/PR6663 15631 18. https://gcc.gnu.org/PR7068 15632 19. https://gcc.gnu.org/PR7083 15633 20. https://gcc.gnu.org/PR7647 15634 21. https://gcc.gnu.org/PR7675 15635 22. https://gcc.gnu.org/PR7718 15636 23. https://gcc.gnu.org/PR8116 15637 24. https://gcc.gnu.org/PR8358 15638 25. https://gcc.gnu.org/PR8511 15639 26. https://gcc.gnu.org/PR8564 15640 27. https://gcc.gnu.org/PR8660 15641 28. https://gcc.gnu.org/PR8766 15642 29. https://gcc.gnu.org/PR8803 15643 30. https://gcc.gnu.org/PR8846 15644 31. https://gcc.gnu.org/PR8906 15645 32. https://gcc.gnu.org/PR9216 15646 33. https://gcc.gnu.org/PR9261 15647 34. https://gcc.gnu.org/PR9263 15648 35. https://gcc.gnu.org/PR9429 15649 36. https://gcc.gnu.org/PR9516 15650 37. https://gcc.gnu.org/PR9600 15651 38. https://gcc.gnu.org/PR9629 15652 39. https://gcc.gnu.org/PR9672 15653 40. https://gcc.gnu.org/PR9749 15654 41. https://gcc.gnu.org/PR9794 15655 42. https://gcc.gnu.org/PR9829 15656 43. https://gcc.gnu.org/PR9916 15657 44. https://gcc.gnu.org/PR9936 15658 45. https://gcc.gnu.org/PR10262 15659 46. https://gcc.gnu.org/PR10278 15660 47. https://gcc.gnu.org/PR10446 15661 48. https://gcc.gnu.org/PR10451 15662 49. https://gcc.gnu.org/PR10506 15663 50. https://gcc.gnu.org/PR10549 15664 51. https://gcc.gnu.org/PR2001 15665 52. https://gcc.gnu.org/PR2391 15666 53. https://gcc.gnu.org/PR2960 15667 54. https://gcc.gnu.org/PR4046 15668 55. https://gcc.gnu.org/PR6405 15669 56. https://gcc.gnu.org/PR6798 15670 57. https://gcc.gnu.org/PR6871 15671 58. https://gcc.gnu.org/PR6909 15672 59. https://gcc.gnu.org/PR7189 15673 60. https://gcc.gnu.org/PR7642 15674 61. https://gcc.gnu.org/PR8634 15675 62. https://gcc.gnu.org/PR8750 15676 63. https://gcc.gnu.org/PR2161 15677 64. https://gcc.gnu.org/PR4319 15678 65. https://gcc.gnu.org/PR8602 15679 66. https://gcc.gnu.org/PR9177 15680 67. https://gcc.gnu.org/PR9853 15681 68. https://gcc.gnu.org/PR45 15682 69. https://gcc.gnu.org/PR3784 15683 70. https://gcc.gnu.org/PR764 15684 71. https://gcc.gnu.org/PR5116 15685 72. https://gcc.gnu.org/PR2862 15686 73. https://gcc.gnu.org/PR3663 15687 74. https://gcc.gnu.org/PR3797 15688 75. https://gcc.gnu.org/PR3948 15689 76. https://gcc.gnu.org/PR4137 15690 77. https://gcc.gnu.org/PR4361 15691 78. https://gcc.gnu.org/PR4802 15692 79. https://gcc.gnu.org/PR5837 15693 80. https://gcc.gnu.org/PR4803 15694 81. https://gcc.gnu.org/PR5094 15695 82. https://gcc.gnu.org/PR5730 15696 83. https://gcc.gnu.org/PR6713 15697 84. https://gcc.gnu.org/PR7015 15698 85. https://gcc.gnu.org/PR7086 15699 86. https://gcc.gnu.org/PR7099 15700 87. https://gcc.gnu.org/PR7247 15701 88. https://gcc.gnu.org/PR7441 15702 89. https://gcc.gnu.org/PR7768 15703 90. https://gcc.gnu.org/PR7804 15704 91. https://gcc.gnu.org/PR8099 15705 92. https://gcc.gnu.org/PR8117 15706 93. https://gcc.gnu.org/PR8205 15707 94. https://gcc.gnu.org/PR8645 15708 95. https://gcc.gnu.org/PR8724 15709 96. https://gcc.gnu.org/PR8805 15710 97. https://gcc.gnu.org/PR8691 15711 98. https://gcc.gnu.org/PR8700 15712 99. https://gcc.gnu.org/PR8724 15713 100. https://gcc.gnu.org/PR8949 15714 101. https://gcc.gnu.org/PR9016 15715 102. https://gcc.gnu.org/PR9053 15716 103. https://gcc.gnu.org/PR9152 15717 104. https://gcc.gnu.org/PR9182 15718 105. https://gcc.gnu.org/PR9297 15719 106. https://gcc.gnu.org/PR9318 15720 107. https://gcc.gnu.org/PR9320 15721 108. https://gcc.gnu.org/PR9400 15722 109. https://gcc.gnu.org/PR9424 15723 110. https://gcc.gnu.org/PR9425 15724 111. https://gcc.gnu.org/PR9439 15725 112. https://gcc.gnu.org/PR9474 15726 113. https://gcc.gnu.org/PR9548 15727 114. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231 15728 115. https://gcc.gnu.org/PR9555 15729 116. https://gcc.gnu.org/PR9561 15730 117. https://gcc.gnu.org/PR9563 15731 118. https://gcc.gnu.org/PR9582 15732 119. https://gcc.gnu.org/PR9622 15733 120. https://gcc.gnu.org/PR9683 15734 121. https://gcc.gnu.org/PR9791 15735 122. https://gcc.gnu.org/PR9817 15736 123. https://gcc.gnu.org/PR9825 15737 124. https://gcc.gnu.org/PR9826 15738 125. https://gcc.gnu.org/PR9924 15739 126. https://gcc.gnu.org/PR9946 15740 127. https://gcc.gnu.org/PR9964 15741 128. https://gcc.gnu.org/PR9988 15742 129. https://gcc.gnu.org/PR10033 15743 130. https://gcc.gnu.org/PR10097 15744 131. https://gcc.gnu.org/PR10132 15745 132. https://gcc.gnu.org/PR10180 15746 133. https://gcc.gnu.org/PR10199 15747 134. https://gcc.gnu.org/PR10300 15748 135. https://gcc.gnu.org/PR10427 15749 136. https://gcc.gnu.org/PR10503 15750 137. https://gcc.gnu.org/PR5956 15751 138. https://gcc.gnu.org/PR1832 15752 139. https://gcc.gnu.org/PR3924 15753 140. https://gcc.gnu.org/PR5634 15754 141. https://gcc.gnu.org/PR6367 15755 142. https://gcc.gnu.org/PR6491 15756 143. https://gcc.gnu.org/PR6742 15757 144. https://gcc.gnu.org/PR7113 15758 145. https://gcc.gnu.org/PR7236 15759 146. https://gcc.gnu.org/PR7278 15760 147. https://gcc.gnu.org/PR7384 15761 148. https://gcc.gnu.org/PR7388 15762 149. https://gcc.gnu.org/PR8587 15763 150. https://gcc.gnu.org/PR9038 15764 151. https://gcc.gnu.org/PR10197 15765 152. https://gcc.gnu.org/PR6005 15766 153. https://gcc.gnu.org/PR6389 15767 154. https://gcc.gnu.org/PR6576 15768 155. https://gcc.gnu.org/PR6652 15769 156. https://gcc.gnu.org/PR7060 15770 157. https://gcc.gnu.org/PR7073 15771 158. https://gcc.gnu.org/PR7180 15772 159. https://gcc.gnu.org/PR7416 15773 160. https://gcc.gnu.org/PR7570 15774 161. https://gcc.gnu.org/PR7578 15775 162. https://gcc.gnu.org/PR7611 15776 163. https://gcc.gnu.org/PR7709 15777 164. https://gcc.gnu.org/PR7766 15778 165. https://gcc.gnu.org/PR7785 15779 166. https://gcc.gnu.org/PR7786 15780 167. https://gcc.gnu.org/PR8142 15781 168. https://gcc.gnu.org/PR8234 15782 169. https://gcc.gnu.org/PR8415 15783 170. https://gcc.gnu.org/PR8481 15784 171. https://gcc.gnu.org/PR8593 15785 172. https://gcc.gnu.org/PR8759 15786 173. https://gcc.gnu.org/PR8997 15787 174. https://gcc.gnu.org/PR9253 15788 175. https://gcc.gnu.org/PR9254 15789 176. https://gcc.gnu.org/PR9271 15790 177. https://gcc.gnu.org/PR6767 15791 178. https://gcc.gnu.org/PR9911 15792 179. https://gcc.gnu.org/PR10020 15793 180. https://gcc.gnu.org/PR10546 15794 181. https://gcc.gnu.org/PR7029 15795 182. https://gcc.gnu.org/PR2903 15796 183. https://gcc.gnu.org/PR7873 15797 184. https://gcc.gnu.org/PR7680 15798 185. https://gcc.gnu.org/PR8705 15799 186. https://gcc.gnu.org/PR9986 15800 187. https://gcc.gnu.org/PR10056 15801 188. https://gcc.gnu.org/PR6744 15802 189. https://gcc.gnu.org/PR7361 15803 190. https://gcc.gnu.org/PR9496 15804 191. https://gcc.gnu.org/PR7067 15805 192. https://gcc.gnu.org/PR8480 15806 193. https://gcc.gnu.org/PR8784 15807 194. https://gcc.gnu.org/PR10315 15808 195. https://gcc.gnu.org/PR10267 15809 196. https://gcc.gnu.org/PR7916 15810 197. https://gcc.gnu.org/PR7926 15811 198. https://gcc.gnu.org/PR8555 15812 199. https://gcc.gnu.org/PR8994 15813 200. https://gcc.gnu.org/PR9426 15814 201. https://gcc.gnu.org/PR9806 15815 202. https://gcc.gnu.org/PR10077 15816 203. https://gcc.gnu.org/PR10233 15817 204. https://gcc.gnu.org/PR10286 15818 205. https://gcc.gnu.org/PR10308 15819 206. https://gcc.gnu.org/PR11272 15820 207. https://gcc.gnu.org/PR5754 15821 208. https://gcc.gnu.org/PR6597 15822 209. https://gcc.gnu.org/PR6949 15823 210. https://gcc.gnu.org/PR7053 15824 211. https://gcc.gnu.org/PR8164 15825 212. https://gcc.gnu.org/PR8384 15826 213. https://gcc.gnu.org/PR9559 15827 214. https://gcc.gnu.org/PR9649 15828 215. https://gcc.gnu.org/PR9864 15829 216. https://gcc.gnu.org/PR10432 15830 217. https://gcc.gnu.org/PR10475 15831 218. https://gcc.gnu.org/PR10635 15832 219. https://gcc.gnu.org/PR10661 15833 220. https://gcc.gnu.org/PR10700 15834 221. https://gcc.gnu.org/PR10712 15835 222. https://gcc.gnu.org/PR10796 15836 223. https://gcc.gnu.org/PR10890 15837 224. https://gcc.gnu.org/PR10939 15838 225. https://gcc.gnu.org/PR10956 15839 226. https://gcc.gnu.org/PR11041 15840 227. https://gcc.gnu.org/PR11059 15841 228. https://gcc.gnu.org/PR11083 15842 229. https://gcc.gnu.org/PR11105 15843 230. https://gcc.gnu.org/PR11149 15844 231. https://gcc.gnu.org/PR11228 15845 232. https://gcc.gnu.org/PR11282 15846 233. https://gcc.gnu.org/PR11301 15847 234. https://gcc.gnu.org/PR11308 15848 235. https://gcc.gnu.org/PR11473 15849 236. https://gcc.gnu.org/PR11503 15850 237. https://gcc.gnu.org/PR11513 15851 238. https://gcc.gnu.org/PR11198 15852 239. https://gcc.gnu.org/PR11304 15853 240. https://gcc.gnu.org/PR11381 15854 241. https://gcc.gnu.org/PR11536 15855 242. https://gcc.gnu.org/PR11557 15856 243. https://gcc.gnu.org/PR5897 15857 244. https://gcc.gnu.org/PR11279 15858 245. https://gcc.gnu.org/PR11022 15859 246. https://gcc.gnu.org/PR2330 15860 247. https://gcc.gnu.org/PR5388 15861 248. https://gcc.gnu.org/PR5390 15862 249. https://gcc.gnu.org/PR7877 15863 250. https://gcc.gnu.org/PR9393 15864 251. https://gcc.gnu.org/PR10032 15865 252. https://gcc.gnu.org/PR10468 15866 253. https://gcc.gnu.org/PR10527 15867 254. https://gcc.gnu.org/PR10679 15868 255. https://gcc.gnu.org/PR10682 15869 256. https://gcc.gnu.org/PR10689 15870 257. https://gcc.gnu.org/PR10845 15871 258. https://gcc.gnu.org/PR10849 15872 259. https://gcc.gnu.org/PR10888 15873 260. https://gcc.gnu.org/PR10929 15874 261. https://gcc.gnu.org/PR10931 15875 262. https://gcc.gnu.org/PR10940 15876 263. https://gcc.gnu.org/PR10968 15877 264. https://gcc.gnu.org/PR10990 15878 265. https://gcc.gnu.org/PR11039 15879 266. https://gcc.gnu.org/PR11062 15880 267. https://gcc.gnu.org/PR11095 15881 268. https://gcc.gnu.org/PR11098 15882 269. https://gcc.gnu.org/PR11137 15883 270. https://gcc.gnu.org/PR11154 15884 271. https://gcc.gnu.org/PR11329 15885 272. https://gcc.gnu.org/PR11332 15886 273. https://gcc.gnu.org/PR11431 15887 274. https://gcc.gnu.org/PR11528 15888 275. https://gcc.gnu.org/PR11546 15889 276. https://gcc.gnu.org/PR11567 15890 277. https://gcc.gnu.org/PR11645 15891 278. https://gcc.gnu.org/PR5179 15892 279. https://gcc.gnu.org/PR8204 15893 280. https://gcc.gnu.org/PR10838 15894 281. https://gcc.gnu.org/PR10886 15895 282. https://gcc.gnu.org/PR11349 15896 283. https://gcc.gnu.org/PR4823 15897 284. https://gcc.gnu.org/PR8878 15898 285. https://gcc.gnu.org/PR9815 15899 286. https://gcc.gnu.org/PR10402 15900 287. https://gcc.gnu.org/PR10504 15901 288. https://gcc.gnu.org/PR10673 15902 289. https://gcc.gnu.org/PR11044 15903 290. https://gcc.gnu.org/PR11089 15904 291. https://gcc.gnu.org/PR11420 15905 292. https://gcc.gnu.org/PR9362 15906 293. https://gcc.gnu.org/PR10142 15907 294. https://gcc.gnu.org/PR10663 15908 295. https://gcc.gnu.org/PR10835 15909 296. https://gcc.gnu.org/PR10876 15910 297. https://gcc.gnu.org/PR10955 15911 298. https://gcc.gnu.org/PR11018 15912 299. https://gcc.gnu.org/PR11556 15913 300. https://gcc.gnu.org/PR10907 15914 301. https://gcc.gnu.org/PR11320 15915 302. https://gcc.gnu.org/PR11599 15916 303. https://gcc.gnu.org/PR9745 15917 304. https://gcc.gnu.org/PR10871 15918 305. https://gcc.gnu.org/PR11440 15919 306. https://gcc.gnu.org/PR7594 15920 307. https://gcc.gnu.org/PR10557 15921 308. https://gcc.gnu.org/PR11054 15922 309. https://gcc.gnu.org/PR10834 15923 310. https://gcc.gnu.org/PR10842 15924 311. https://gcc.gnu.org/PR11052 15925 312. https://gcc.gnu.org/PR11183 15926 313. https://gcc.gnu.org/PR11084 15927 314. https://gcc.gnu.org/PR10331 15928 315. https://gcc.gnu.org/PR10413 15929 316. https://gcc.gnu.org/PR11096 15930 317. https://gcc.gnu.org/PR2873 15931 318. https://gcc.gnu.org/PR3163 15932 319. https://gcc.gnu.org/PR5287 15933 320. https://gcc.gnu.org/PR10148 15934 321. https://gcc.gnu.org/PR8787 15935 322. https://gcc.gnu.org/PR10900 15936 323. https://gcc.gnu.org/PR1607 15937 324. https://gcc.gnu.org/PR4252 15938 325. https://gcc.gnu.org/PR4490 15939 326. https://gcc.gnu.org/PR10355 15940 327. https://gcc.gnu.org/PR10726 15941 328. https://gcc.gnu.org/PR10805 15942 329. https://gcc.gnu.org/PR10815 15943 330. https://gcc.gnu.org/PR10877 15944 331. https://gcc.gnu.org/PR11280 15945 332. https://gcc.gnu.org/PR11466 15946 333. https://gcc.gnu.org/PR10737 15947 334. https://gcc.gnu.org/PR10810 15948 335. https://gcc.gnu.org/PR8336 15949 336. https://gcc.gnu.org/PR9330 15950 337. https://gcc.gnu.org/PR9631 15951 338. https://gcc.gnu.org/PR9877 15952 339. https://gcc.gnu.org/PR11687 15953 340. https://gcc.gnu.org/PR12263 15954 341. https://gcc.gnu.org/PR12490 15955 342. https://gcc.gnu.org/PR7277 15956 343. https://gcc.gnu.org/PR7939 15957 344. https://gcc.gnu.org/PR11063 15958 345. https://gcc.gnu.org/PR11207 15959 346. https://gcc.gnu.org/PR11522 15960 347. https://gcc.gnu.org/PR11595 15961 348. https://gcc.gnu.org/PR11646 15962 349. https://gcc.gnu.org/PR11665 15963 350. https://gcc.gnu.org/PR11852 15964 351. https://gcc.gnu.org/PR11878 15965 352. https://gcc.gnu.org/PR11883 15966 353. https://gcc.gnu.org/PR11991 15967 354. https://gcc.gnu.org/PR12146 15968 355. https://gcc.gnu.org/PR12215 15969 356. https://gcc.gnu.org/PR12369 15970 357. https://gcc.gnu.org/PR12446 15971 358. https://gcc.gnu.org/PR12510 15972 359. https://gcc.gnu.org/PR12544 15973 360. https://gcc.gnu.org/PR9862 15974 361. https://gcc.gnu.org/PR10962 15975 362. https://gcc.gnu.org/PR11370 15976 363. https://gcc.gnu.org/PR11637 15977 364. https://gcc.gnu.org/PR11885 15978 365. https://gcc.gnu.org/PR12082 15979 366. https://gcc.gnu.org/PR12180 15980 367. https://gcc.gnu.org/PR12340 15981 368. https://gcc.gnu.org/PR3907 15982 369. https://gcc.gnu.org/PR5293 15983 370. https://gcc.gnu.org/PR5296 15984 371. https://gcc.gnu.org/PR7939 15985 372. https://gcc.gnu.org/PR8656 15986 373. https://gcc.gnu.org/PR10147 15987 374. https://gcc.gnu.org/PR11400 15988 375. https://gcc.gnu.org/PR11409 15989 376. https://gcc.gnu.org/PR11740 15990 377. https://gcc.gnu.org/PR11786 15991 378. https://gcc.gnu.org/PR11867 15992 379. https://gcc.gnu.org/PR11928 15993 380. https://gcc.gnu.org/PR12114 15994 381. https://gcc.gnu.org/PR12163 15995 382. https://gcc.gnu.org/PR12181 15996 383. https://gcc.gnu.org/PR12236 15997 384. https://gcc.gnu.org/PR12266 15998 385. https://gcc.gnu.org/PR12296 15999 386. https://gcc.gnu.org/PR12298 16000 387. https://gcc.gnu.org/PR12369 16001 388. https://gcc.gnu.org/PR12337 16002 389. https://gcc.gnu.org/PR12344 16003 390. https://gcc.gnu.org/PR12451 16004 391. https://gcc.gnu.org/PR12486 16005 392. https://gcc.gnu.org/PR8869 16006 393. https://gcc.gnu.org/PR9786 16007 394. https://gcc.gnu.org/PR11689 16008 395. https://gcc.gnu.org/PR12116 16009 396. https://gcc.gnu.org/PR12070 16010 397. https://gcc.gnu.org/PR11184 16011 398. https://gcc.gnu.org/PR11535 16012 399. https://gcc.gnu.org/PR11693 16013 400. https://gcc.gnu.org/PR12224 16014 401. https://gcc.gnu.org/PR11087 16015 402. https://gcc.gnu.org/PR11319 16016 403. https://gcc.gnu.org/PR11949 16017 404. https://gcc.gnu.org/PR11662 16018 405. https://gcc.gnu.org/PR11965 16019 406. https://gcc.gnu.org/PR12301 16020 407. https://gcc.gnu.org/PR11717 16021 408. https://gcc.gnu.org/PR11313 16022 409. https://gcc.gnu.org/PR11712 16023 410. https://gcc.gnu.org/PR12166 16024 411. https://gcc.gnu.org/PR12101 16025 412. https://gcc.gnu.org/PR10988 16026 413. https://gcc.gnu.org/PR11805 16027 414. https://gcc.gnu.org/PR11902 16028 415. https://gcc.gnu.org/PR11903 16029 416. https://gcc.gnu.org/PR11890 16030 417. https://gcc.gnu.org/PR12399 16031 418. https://gcc.gnu.org/PR13068 16032 419. https://gcc.gnu.org/PR10060 16033 420. https://gcc.gnu.org/PR10555 16034 421. https://gcc.gnu.org/PR10706 16035 422. https://gcc.gnu.org/PR11496 16036 423. https://gcc.gnu.org/PR11741 16037 424. https://gcc.gnu.org/PR12440 16038 425. https://gcc.gnu.org/PR12632 16039 426. https://gcc.gnu.org/PR12712 16040 427. https://gcc.gnu.org/PR12726 16041 428. https://gcc.gnu.org/PR12890 16042 429. https://gcc.gnu.org/PR12900 16043 430. https://gcc.gnu.org/PR13060 16044 431. https://gcc.gnu.org/PR13289 16045 432. https://gcc.gnu.org/PR13318 16046 433. https://gcc.gnu.org/PR13392 16047 434. https://gcc.gnu.org/PR13574 16048 435. https://gcc.gnu.org/PR13475 16049 436. https://gcc.gnu.org/PR13797 16050 437. https://gcc.gnu.org/PR13824 16051 438. https://gcc.gnu.org/PR8776 16052 439. https://gcc.gnu.org/PR10339 16053 440. https://gcc.gnu.org/PR11350 16054 441. https://gcc.gnu.org/PR12826 16055 442. https://gcc.gnu.org/PR12500 16056 443. https://gcc.gnu.org/PR12941 16057 444. https://gcc.gnu.org/PR12953 16058 445. https://gcc.gnu.org/PR13041 16059 446. https://gcc.gnu.org/PR13507 16060 447. https://gcc.gnu.org/PR13382 16061 448. https://gcc.gnu.org/PR13394 16062 449. https://gcc.gnu.org/PR13400 16063 450. https://gcc.gnu.org/PR13521 16064 451. https://gcc.gnu.org/PR2094 16065 452. https://gcc.gnu.org/PR2294 16066 453. https://gcc.gnu.org/PR5050 16067 454. https://gcc.gnu.org/PR9371 16068 455. https://gcc.gnu.org/PR9546 16069 456. https://gcc.gnu.org/PR10081 16070 457. https://gcc.gnu.org/PR10093 16071 458. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61 16072 459. https://gcc.gnu.org/PR10095 16073 460. https://gcc.gnu.org/PR11554 16074 461. https://gcc.gnu.org/PR12297 16075 462. https://gcc.gnu.org/PR12352 16076 463. https://gcc.gnu.org/PR12438 16077 464. https://gcc.gnu.org/PR12540 16078 465. https://gcc.gnu.org/PR12594 16079 466. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60 16080 467. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63 16081 468. https://gcc.gnu.org/PR12657 16082 469. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292 16083 470. https://gcc.gnu.org/PR12696 16084 471. https://gcc.gnu.org/PR12815 16085 472. https://gcc.gnu.org/PR12862 16086 473. https://gcc.gnu.org/PR12926 16087 474. https://gcc.gnu.org/PR12967 16088 475. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html 16089 476. https://gcc.gnu.org/PR12971 16090 477. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328 16091 478. https://gcc.gnu.org/PR13007 16092 479. https://gcc.gnu.org/PR13009 16093 480. https://gcc.gnu.org/PR13057 16094 481. https://gcc.gnu.org/PR13070 16095 482. https://gcc.gnu.org/PR13081 16096 483. https://gcc.gnu.org/PR13239 16097 484. https://gcc.gnu.org/PR13262 16098 485. https://gcc.gnu.org/PR13290 16099 486. https://gcc.gnu.org/PR13323 16100 487. https://gcc.gnu.org/PR13369 16101 488. https://gcc.gnu.org/PR13371 16102 489. https://gcc.gnu.org/PR13445 16103 490. https://gcc.gnu.org/PR13461 16104 491. https://gcc.gnu.org/PR13462 16105 492. https://gcc.gnu.org/PR13478 16106 493. https://gcc.gnu.org/PR13544 16107 494. https://gcc.gnu.org/PR13650 16108 495. https://gcc.gnu.org/PR13683 16109 496. https://gcc.gnu.org/PR13688 16110 497. https://gcc.gnu.org/PR13774 16111 498. https://gcc.gnu.org/PR13884 16112 499. https://gcc.gnu.org/PR10746 16113 500. https://gcc.gnu.org/PR11433 16114 501. https://gcc.gnu.org/PR12633 16115 502. https://gcc.gnu.org/PR13037 16116 503. https://gcc.gnu.org/PR13213 16117 504. https://gcc.gnu.org/PR4490 16118 505. https://gcc.gnu.org/PR12292 16119 506. https://gcc.gnu.org/PR12441 16120 507. https://gcc.gnu.org/PR12943 16121 508. https://gcc.gnu.org/PR13608 16122 509. https://gcc.gnu.org/PR11598 16123 510. https://gcc.gnu.org/PR11793 16124 511. https://gcc.gnu.org/PR12467 16125 512. https://gcc.gnu.org/PR12537 16126 513. https://gcc.gnu.org/PR12496 16127 514. https://gcc.gnu.org/PR12865 16128 515. https://gcc.gnu.org/PR13354 16129 516. https://gcc.gnu.org/PR10467 16130 517. https://gcc.gnu.org/PR11226 16131 518. https://gcc.gnu.org/PR11227 16132 519. https://gcc.gnu.org/PR12644 16133 520. https://gcc.gnu.org/PR13149 16134 521. https://gcc.gnu.org/PR12654 16135 522. https://gcc.gnu.org/PR12965 16136 523. https://gcc.gnu.org/PR13031 16137 524. https://gcc.gnu.org/PR11634 16138 525. https://gcc.gnu.org/PR12158 16139 526. https://gcc.gnu.org/PR11992 16140 527. https://gcc.gnu.org/PR9365 16141 528. https://gcc.gnu.org/PR10392 16142 529. https://gcc.gnu.org/PR11322 16143 530. https://gcc.gnu.org/PR13069 16144 531. https://gcc.gnu.org/PR13302 16145 532. https://gcc.gnu.org/PR13585 16146 533. https://gcc.gnu.org/PR8916 16147 534. https://gcc.gnu.org/PR11576 16148 535. https://gcc.gnu.org/PR13122 16149 536. https://gcc.gnu.org/PR13256 16150 537. https://gcc.gnu.org/PR13373 16151 538. https://gcc.gnu.org/PR12561 16152 539. https://gcc.gnu.org/PR6243 16153 540. https://gcc.gnu.org/PR11397 16154 541. https://gcc.gnu.org/PR12505 16155 542. https://gcc.gnu.org/PR13150 16156 543. https://gcc.gnu.org/PR12666 16157 544. https://gcc.gnu.org/PR12969 16158 545. https://gcc.gnu.org/PR10819 16159 546. https://gcc.gnu.org/PR11612 16160 547. https://gcc.gnu.org/PR13211 16161 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4 16162 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5 16163 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6 16164 551. https://gcc.gnu.org/onlinedocs/ 16165 552. mailto:gcc-help (a] gcc.gnu.org 16166 553. mailto:gcc (a] gcc.gnu.org 16167 554. https://gcc.gnu.org/lists.html 16168 555. https://www.fsf.org/ 16169 556. https://gcc.gnu.org/about.html 16170 557. http://validator.w3.org/check/referer 16171 ====================================================================== 16172 http://gcc.gnu.org/gcc-3.2/index.html 16173 GCC 3.2 Release Series 16174 16175 April 25, 2003 16176 16177 The [1]GNU project and the GCC developers are pleased to announce the 16178 release of GCC 3.2.3. 16179 16180 The purpose of the GCC 3.2 release series is to provide a stable 16181 platform for OS distributors to use building their next releases. A 16182 primary objective was to stabilize the C++ ABI; we believe that the 16183 interface to the compiler and the C++ standard library are now 16184 relatively stable. 16185 16186 Be aware that C++ code compiled by GCC 3.2.x will (in general) not 16187 interoperate with code compiled by GCC 3.1.1 or earlier. 16188 16189 Please refer to our [2]detailed list of news, caveats, and bug-fixes 16190 for further information. 16191 16192 This release series is no longer maintained. 16193 16194 Release History 16195 16196 GCC 3.2.3 16197 April 25, 2003 ([3]changes) 16198 16199 GCC 3.2.2 16200 February 5, 2003 ([4]changes) 16201 16202 GCC 3.2.1 16203 November 19, 2002 ([5]changes) 16204 16205 GCC 3.2 16206 August 14, 2002 ([6]changes) 16207 16208 References and Acknowledgements 16209 16210 GCC used to stand for the GNU C Compiler, but since the compiler 16211 supports several other languages aside from C, it now stands for the 16212 GNU Compiler Collection. 16213 16214 A list of [7]successful builds is updated as new information becomes 16215 available. 16216 16217 The GCC developers would like to thank the numerous people that have 16218 contributed new features, improvements, bug fixes, and other changes as 16219 well as test results to GCC. This [8]amazing group of volunteers is 16220 what makes GCC successful. 16221 16222 For additional information about GCC please refer to the [9]GCC project 16223 web site or contact the [10]GCC development mailing list. 16224 16225 To obtain GCC please use [11]our mirror sites, or our CVS server. 16226 16227 16228 For questions related to the use of GCC, please consult these web 16229 pages and the [12]GCC manuals. If that fails, the 16230 [13]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16231 web pages and the development of GCC are welcome on our developer 16232 list at [14]gcc (a] gcc.gnu.org. All of [15]our lists have public 16233 archives. 16234 16235 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and 16236 distribution of this entire article is permitted in any medium, 16237 provided this notice is preserved. 16238 16239 These pages are [17]maintained by the GCC team. Last modified 16240 2018-09-30[18]. 16241 16242 References 16243 16244 1. http://www.gnu.org/ 16245 2. http://gcc.gnu.org/gcc-3.2/changes.html 16246 3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 16247 4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2 16248 5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1 16249 6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2 16250 7. http://gcc.gnu.org/gcc-3.2/buildstat.html 16251 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 16252 9. http://gcc.gnu.org/index.html 16253 10. mailto:gcc (a] gcc.gnu.org 16254 11. http://gcc.gnu.org/mirrors.html 16255 12. https://gcc.gnu.org/onlinedocs/ 16256 13. mailto:gcc-help (a] gcc.gnu.org 16257 14. mailto:gcc (a] gcc.gnu.org 16258 15. https://gcc.gnu.org/lists.html 16259 16. https://www.fsf.org/ 16260 17. https://gcc.gnu.org/about.html 16261 18. http://validator.w3.org/check/referer 16262 ====================================================================== 16263 http://gcc.gnu.org/gcc-3.2/changes.html 16264 GCC 3.2 Release Series 16265 Changes, New Features, and Fixes 16266 16267 The latest release in the 3.2 release series is [1]GCC 3.2.3. 16268 16269 Caveats and New Features 16270 16271 Caveats 16272 16273 * The C++ compiler does not correctly zero-initialize 16274 pointers-to-data members. You must explicitly initialize them. For 16275 example: int S::*m(0); will work, but depending on 16276 default-initialization to zero will not work. This bug cannot be 16277 fixed in GCC 3.2 without inducing unacceptable risks. It will be 16278 fixed in GCC 3.3. 16279 * This GCC release is based on the GCC 3.1 sourcebase, and thus has 16280 all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has 16281 a number of C++ ABI fixes which make its C++ compiler generate 16282 binary code which is incompatible with the C++ compilers found in 16283 earlier GCC releases, including GCC 3.1 and GCC 3.1.1. 16284 16285 Frontend Enhancements 16286 16287 C/C++/Objective-C 16288 16289 * The method of constructing the list of directories to be searched 16290 for header files has been revised. If a directory named by a -I 16291 option is a standard system include directory, the option is 16292 ignored to ensure that the default search order for system 16293 directories and the special treatment of system header files are 16294 not defeated. 16295 * The C and Objective-C compilers no longer accept the "Naming Types" 16296 extension (typedef foo = bar); it was already unavailable in C++. 16297 Code which uses it will need to be changed to use the "typeof" 16298 extension instead: typedef typeof(bar) foo. (We have removed this 16299 extension without a period of deprecation because it has caused the 16300 compiler to crash since version 3.0 and no one noticed until very 16301 recently. Thus we conclude it is not in widespread use.) 16302 16303 C++ 16304 16305 * GCC 3.2 fixed serveral differences between the C++ ABI implemented 16306 in GCC and the multi-vendor standard, but more have been found 16307 since the release. 3.2.1 adds a new warning, -Wabi, to warn about 16308 code which is affected by these bugs. We will fix these bugs in 16309 some future release, once we are confident that all have been 16310 found; until then, it is our intention to make changes to the ABI 16311 only if they are necessary for correct compilation of C++, as 16312 opposed to conformance to the ABI documents. 16313 * For details on how to build an ABI compliant compiler for GNU/Linux 16314 systems, check the [3]common C++ ABI page. 16315 16316 New Targets and Target Specific Improvements 16317 16318 IA-32 16319 16320 * Fixed a number of bugs in SSE and MMX intrinsics. 16321 * Fixed common compiler crashes with SSE instruction set enabled 16322 (implied by -march=pentium3, pentium4, athlon-xp) 16323 * __m128 and __m128i is not 128bit aligned when used in structures. 16324 16325 x86-64 16326 16327 * A bug whereby the compiler could generate bad code for bzero has 16328 been fixed. 16329 * ABI fixes (implying ABI incompatibilities with previous version in 16330 some corner cases) 16331 * Fixed prefetch code generation 16332 __________________________________________________________________ 16333 16334 GCC 3.2.3 16335 16336 3.2.3 is a bug fix release only; there are no new features that were 16337 not present in GCC 3.2.2. 16338 16339 Bug Fixes 16340 16341 This section lists the problem reports (PRs) from GCC's bug tracking 16342 system that are known to be fixed in the 3.2.3 release. This list might 16343 not be complete (that is, it is possible that some PRs that have been 16344 fixed are not listed here), and some of the titles have been changed to 16345 make them more clear. 16346 16347 Internal Compiler Errors (multi-platform) 16348 16349 * [4]3782: (c++) -quiet -fstats produces a segmentation fault in 16350 cc1plus 16351 * [5]6440: (c++) template specializations cause ICE 16352 * [6]7050: (c++) ICE on: (i ? get_string() : throw) 16353 * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c) 16354 * [8]7982: (c++) ICE due to infinite recursion (using STL set) 16355 * [9]8068: exceedingly high (infinite) memory usage 16356 * [10]8178: ICE with __builtin_ffs 16357 * [11]8396: ICE in copy_to_mode_reg, in explow.c 16358 * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c 16359 * [13]9768: ICE when optimizing inline code at -O2 16360 * [14]9798: (c++) Infinite recursion (segfault) in 16361 cp/decl.c:push_using_directive with recursive using directives 16362 * [15]9799: mismatching structure initializer with nested flexible 16363 array member: ICE 16364 * [16]9928: ICE on duplicate enum declaration 16365 * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects 16366 sparc, alpha) 16367 * [18]10352: ICE in find_reloads_toplev 16368 * [19]10336: ICE with -Wunreachable-code 16369 16370 C/optimizer bugs: 16371 16372 * [20]8224: Incorrect joining of signed and unsigned division 16373 * [21]8613: -O2 produces wrong code with builtin strlen and 16374 postincrements 16375 * [22]8828: gcc reports some code is unreachable when it is not 16376 * [23]9226: GCSE breaking argument passing 16377 * [24]9853: miscompilation of non-constant structure initializer 16378 * [25]9797: C99-style struct initializers are miscompiled 16379 * [26]9967: Some standard C function calls should not be replaced 16380 when optimizing for size 16381 * [27]10116: ce2: invalid merge of join_bb in the context of switch 16382 statements 16383 * [28]10171: wrong code for inlined function 16384 * [29]10175: -Wunreachable-code doesn't work for single lines 16385 16386 C++ compiler and library: 16387 16388 * [30]8316: Confusing diagnostic for code that misuses conversion 16389 operators 16390 * [31]9169: filebuf output fails if codecvt<>::out returns noconv 16391 * [32]9420: incomplete type incorrectly reported 16392 * [33]9459: typeof in return type specification of template not 16393 supported 16394 * [34]9507: filebuf::open handles ios_base::ate incorrectly 16395 * [35]9538: Out-of-bounds memory access in streambuf::sputbackc 16396 * [36]9602: Total confusion about template/friend/virtual/abstract 16397 * [37]9993: destructor not called for local object created within and 16398 returned from infinite loop 16399 * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2 16400 system 16401 16402 Java compiler and library: 16403 16404 * [39]9652: libgcj build fails on irix6.5.1[78] 16405 * [40]10144: gas on solaris complains about bad .stabs lines for 16406 java, native as unaffected 16407 16408 x86-specific (Intel/AMD): 16409 16410 * [41]8746: gcc miscompiles Linux kernel ppa driver on x86 16411 * [42]9888: -mcpu=k6 -Os produces out of range loop instructions 16412 * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu 16413 failed 16414 * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib) 16415 failed 16416 16417 SPARC-specific: 16418 16419 * [45]7784: [Sparc] ICE in extract_insn, in recog.c 16420 * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in 16421 unroll.c 16422 * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc 16423 * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in 16424 execute/loop-2d.c 16425 * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc 16426 * [50]9414: Scheduling bug on Ultrasparc 16427 * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64 16428 16429 m68k-specific: 16430 16431 * [52]7248: broken "inclusive or" code 16432 * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1 16433 16434 PowerPC-specific: 16435 16436 * [54]9732: Wrong code with -O2 -fPIC 16437 * [55]10073: ICE: powerpc cannot split insn 16438 16439 Alpha-specific: 16440 16441 * [56]7702: optimization problem on a DEC alpha under OSF1 16442 * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system 16443 16444 HP-specific: 16445 16446 * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275) 16447 * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10 16448 (missing symbol) 16449 * [60]10271: Floating point args don't get reloaded across function 16450 calls with -O2 16451 16452 MIPS specific: 16453 16454 * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in 16455 compile/920501-4.c 16456 16457 CRIS specific: 16458 16459 * [62]10377: gcc-3.2.2 creates bad assembler code for cris 16460 16461 Miscellaneous and minor bugs: 16462 16463 * [63]6955: collect2 says "core dumped" when there is no core 16464 __________________________________________________________________ 16465 16466 GCC 3.2.2 16467 16468 Beginning with 3.2.2, GCC's Makefile suite supports redirection of make 16469 install by means of the DESTDIR variable. Parts of the GCC tree have 16470 featured that support long before, but now it is available even from 16471 the top level. 16472 16473 Other than that, GCC 3.2.2 is a bug fix release only; there are no new 16474 features that were not present in GCC 3.2.1. 16475 16476 Bug Fixes 16477 16478 On the following i386-based systems GCC 3.2.1 broke the C ABI wrt. 16479 functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped 16480 with FreeBSD 5.0 does not have this problem), Interix, a.out-based 16481 GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI 16482 change, and thus restores ABI-compatibility with previous releases 16483 (except GCC 3.2.1) on these platforms. 16484 16485 This section lists the problem reports (PRs) from GCC's bug tracking 16486 system that are known to be fixed in the 3.2.2 release. This list might 16487 not be complete (that is, it is possible that some PRs that have been 16488 fixed are not listed here) and some of the titles have been changed to 16489 make them more clear. 16490 16491 Internal Compiler Errors (multi-platform) 16492 16493 * [64]5919: (c++) ICE when passing variable array to template 16494 function 16495 * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=) 16496 * [66]7507: ICE with -O2 when address of called function is a 16497 complicated expression 16498 * [67]7622: ICE with nested inline functions if function's address is 16499 taken 16500 * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR 16501 [69]9258) 16502 * [70]8031: (c++) ICE in code comparing typeids and casting from 16503 virtual base 16504 * [71]8275: ICE in simplify_subreg 16505 * [72]8332: (c++) builtin strlen/template interaction causes ICE 16506 * [73]8372: (c++) ICE on explicit call of destructor 16507 * [74]8439: (c, not c++) empty struct causes ICE 16508 * [75]8442: (c++) ICE with nested template classes 16509 * [76]8518: ICE when compiling mplayer ("extern inline" issue) 16510 * [77]8615: (c++) ICE with out-of-range character constant template 16511 argument 16512 * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307 16513 * [79]8799: (c++) ICE: error reporting routines re-entered 16514 * [80]9328: (c++) ICE with typeof(X) for overloaded X 16515 * [81]9465: (preprocessor) cpp -traditional ICE on null bytes 16516 16517 C++ (compiler and library) bugs 16518 16519 * [82]47: scoping in nested classes is broken 16520 * [83]6745: problems with iostream rdbuf() member function 16521 * [84]8214: conversion from const char* const to char* sometimes 16522 accepted illegally 16523 * [85]8493: builtin strlen and overload resolution (same bug as 16524 [86]8332) 16525 * [87]8503: strange behaviour of function types 16526 * [88]8727: compiler confused by inheritance from an anonymous struct 16527 * [89]7445: poor performance of std::locale::classic() in 16528 multi-threaded applications 16529 * [90]8230: mishandling of overflow in vector<T>::resize 16530 * [91]8399: sync_with_stdio(false) breaks unformatted input 16531 * [92]8662: illegal access of private member of unnamed class is 16532 accepted 16533 * [93]8707: "make distclean" fails in libstdc++-v3 directory 16534 * [94]8708: __USE_MALLOC doesn't work 16535 * [95]8790: Use of non-thread-safe strtok in src/localename.cc 16536 * [96]8887: Bug in date formats with --enable-clocale=generic 16537 * [97]9076: Call Frame Instructions are not handled correctly during 16538 unwind operation 16539 * [98]9151: std::setprecision limited to 16 digits when outputting a 16540 double to a stream 16541 * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers 16542 * [100]9269: libstdc++ headers: explicit specialization of function 16543 must precede its first use 16544 * [101]9322: return value of basic_streambuf<>::getloc affected by 16545 locale::global 16546 * [102]9433: segfault in runtime support for dynamic_cast 16547 16548 C and optimizer bugs 16549 16550 * [103]8032: GCC incorrectly initializes static structs that have 16551 flexible arrays 16552 * [104]8639: simple arithmetic expression broken 16553 * [105]8794: optimization improperly eliminates certain expressions 16554 * [106]8832: traditional "asm volatile" code is illegally optimized 16555 * [107]8988: loop optimizer bug: with -O2, code is generated that 16556 segfaults (found on i386, bug present for all platforms) 16557 * [108]9492: structure copy clobbers subsequent stores to structure 16558 16559 Objective-C bugs 16560 16561 * [109]9267: Objective-C parser won't build with newer bison versions 16562 (e.g. 1.875) 16563 16564 Ada bugs 16565 16566 * [110]8344: Ada build problem due to conflict between gcc/final.o, 16567 gcc/ada/final.o 16568 16569 Preprocessor bugs 16570 16571 * [111]8524: _Pragma within macros is improperly expanded 16572 * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with 16573 -fshort-wchar 16574 16575 ARM-specific 16576 16577 * [113]9090: arm ICE with >= -O2; regression from gcc-2.95 16578 16579 x86-specific (Intel/AMD) 16580 16581 * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction) 16582 * [115]8599: loop unroll bug with -march=k6-3 16583 * [116]9506: ABI breakage in structure return (affects BSD and 16584 Cygwin, but not GNU/Linux) 16585 16586 FreeBSD 5.0 specific 16587 16588 * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0 16589 16590 RTEMS-specific 16591 16592 * [118]9292: hppa1.1-rtems configurery problems 16593 * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug 16594 * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue 16595 * [121]9296: gthr-rtems regression 16596 * [122]9316: powerpc-rtems: extending multilibs 16597 16598 HP-PA specific 16599 16600 * [123]9493: ICE with -O2 when building a simple function 16601 16602 Documentation 16603 16604 * [124]7341: hyperlink to gcov in GCC documentation doesn't work 16605 * [125]8947: Please add a warning about "-malign-double" in docs 16606 * [126]7448, [127]8882: typo cleanups 16607 __________________________________________________________________ 16608 16609 GCC 3.2.1 16610 16611 3.2.1 adds a new warning, -Wabi. This option warns when GNU C++ 16612 generates code that is known not to be binary-compatible with the 16613 vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included 16614 in the distribution, for details. 16615 16616 This release also removes an old GCC extension, "naming types", and the 16617 documentation now directs users to use a different GCC extension, 16618 __typeof__, instead. The feature had evidently been broken for a while. 16619 16620 Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and 16621 the new warning there are no new features that were not present in GCC 16622 3.2. 16623 16624 In addition, the previous fix for [128]PR 7445 (poor performance of 16625 std::locale::classic() in multi-threaded applications) was reverted 16626 ("unfixed"), because the "fix" was not thread-safe. 16627 16628 Bug Fixes 16629 16630 This section lists the problem reports (PRs) from GCC's bug tracking 16631 system that are known to be fixed in the 3.2.1 release. This list might 16632 not be complete (that is, it is possible that some PRs that have been 16633 fixed are not listed here). As you can see, the number of bug fixes is 16634 quite large, so it is strongly recommended that users of earlier GCC 16635 3.x releases upgrade to GCC 3.2.1. 16636 16637 Internal Compiler Errors (multi-platform) 16638 16639 * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c 16640 * [130]5661: (c++) ICE instantiating template on array of unknown 16641 size (bad code) 16642 * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on 16643 64-bit platforms 16644 * [132]6994: (c++) ICE in find_function_data 16645 * [133]7150: preprocessor: GCC -dM -E gives an ICE 16646 * [134]7160: ICE when optimizing branches without a return value 16647 * [135]7228: (c++) ICE when using member template and template 16648 function 16649 * [136]7266: (c++) ICE with -pedantic on missing typename 16650 * [137]7353: ICE from use of "Naming Types" extension, see above 16651 * [138]7411: ICE in instantiate_virtual_regs_1, in function.c 16652 * [139]7478: (c++) ICE on static_cast inside template 16653 * [140]7526: preprocessor core dump when _Pragma implies #pragma 16654 dependency 16655 * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803 16656 is a duplicate) 16657 * [143]7754: (c++) ICE on union with template parameter 16658 * [144]7788: (c++) redeclaring a definition as an incomplete class 16659 causes ICE 16660 * [145]8031: (c++) ICE in comptypes, in cp/typeck.c 16661 * [146]8055: preprocessor dies with SIG11 when building FreeBSD 16662 kernel 16663 * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related 16664 variables 16665 * [148]8134: (c++) ICE in force_store_init_value on legal code 16666 * [149]8149: (c++) ICE on incomplete type 16667 * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array 16668 initialization 16669 16670 C++ (compiler and library) bugs 16671 16672 * [151]5607: No pointer adjustment in covariant return types 16673 * [152]6579: Infinite loop with statement expressions in member 16674 initialization 16675 * [153]6803: Default copy constructor bug in GCC 3.1 16676 * [154]7176: g++ confused by friend and static member with same name 16677 * [155]7188: Segfault with template class and recursive (incorrect) 16678 initializer list 16679 * [156]7306: Regression: GCC 3.x fails to compile code with virtual 16680 inheritance if a method has a variable number of arguments 16681 * [157]7461: ctype<char>::classic_table() returns offset array on 16682 Cygwin 16683 * [158]7524: f(const float arg[3]) fails 16684 * [159]7584: Erroneous ambiguous base error on using declaration 16685 * [160]7676: Member template overloading problem 16686 * [161]7679: infinite loop when a right parenthesis is missing 16687 * [162]7811: default locale not taken from environment 16688 * [163]7961: compare( char *) implemented incorrectly in 16689 basic_string<> 16690 * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if 16691 streambuf::underflow() leaves gptr() NULL (dups: [165]8127, 16692 [166]6745) 16693 * [167]8096: deque::at() throws std::range_error instead of 16694 std::out_of_range 16695 * [168]8127: cout << cin.rdbuf() infinite loop 16696 * [169]8218: Excessively large memory consumed for classes with large 16697 array members 16698 * [170]8287: GCC 3.2: Destructor called for non-constructed local 16699 object 16700 * [171]8347: empty vector range used in string construction causes 16701 core dump 16702 * [172]8348: fail() flag is set in istringstream when eof() flag is 16703 set 16704 * [173]8391: regression: infinite loop in cp/decl2.c(finish_file) 16705 16706 C and optimizer bugs 16707 16708 * [174]6627: -fno-align-functions doesn't seem to disable function 16709 alignment 16710 * [175]6631: life_analysis misoptimizes code to initialize fields of 16711 a structure 16712 * [176]7102: unsigned char division results in floating exception 16713 * [177]7120: Run once loop should *always* be unrolled 16714 (pessimization) 16715 * [178]7209: Bug involving array referencing and ?: operator 16716 * [179]7515: invalid inlining of global function with -O3 16717 * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test 16718 * [181]8467: bug in sibling call optimization 16719 16720 Preprocessor bugs 16721 16722 * [182]4890: incorrect line markers from the traditional preprocessor 16723 * [183]7357: -M option omits system headers files (making it the same 16724 as -MM) 16725 * [184]7358: Changes to Sun's make Dependencies 16726 * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as 16727 C headers 16728 * [186]7862: Interrupting GCC -MD removes .d file but not .o 16729 * [187]8190: Failed compilation deletes -MD dependency file 16730 * [188]8524: _Pragma within macro is improperly expanded 16731 16732 x86 specific (Intel/AMD) 16733 16734 * [189]5351: (i686-only) function pass-by-value structure copy 16735 corrupts stack ([190]7591 is a duplicate) 16736 * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with 16737 -march=pentium3/pentium2/athlon (these are all the same underlying 16738 bug, in MMX register use) 16739 * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same 16740 as above?) 16741 * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken 16742 * [199]6981: wrong code in 64-bit manipulation on x86 16743 * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__ 16744 macro 16745 * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE 16746 intrinsics are broken 16747 * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with 16748 -march=pentium4 16749 * [203]7693: Typo in i386 mmintrin.h header 16750 * [204]7723: ICE - Pentium3 sse - GCC 3.2 16751 * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse 16752 * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3 16753 16754 PowerPC specific 16755 16756 * [207]5967: GCC bug when profiling nested functions on powerpc 16757 * [208]6984: wrong code generated with -O2, -O3, -Os for do-while 16758 loop on PowerPC 16759 * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5 16760 * [210]7130: miscompiled code for GCC-3.1 on 16761 powerpc-unknown-linux-gnu with -funroll-all-loops 16762 * [211]7133: PowerPC ICE: unrecognizable insn 16763 * [212]7380: ICE in extract_insn, at recog.c:2148 16764 * [213]8252: ICE on Altivec code with optimization turned on 16765 * [214]8451: Altivec ICE in GCC 3.2 16766 16767 HP/PA specific 16768 16769 * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa 16770 16771 SPARC specific 16772 16773 * [216]6668: when using --disable-multilib, libgcc_s.so is installed 16774 in the wrong place on sparc-solaris 16775 * [217]7151: ICE when compiling for UltraSPARC 16776 * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long 16777 double and -O1 16778 * [219]7842: [REGRESSION] SPARC code gen bug 16779 16780 ARM specific 16781 16782 * [220]7856: [arm] invalid offset in constant pool reference 16783 * [221]7967: optimization produces wrong code (ARM) 16784 16785 Alpha specific 16786 16787 * [222]7374: __builtin_fabsl broken on alpha 16788 16789 IBM s390 specific 16790 16791 * [223]7370: ICE in fixup_var_refs_1 on s390x 16792 * [224]7409: loop optimization bug on s390x-linux-gnu 16793 * [225]8232: s390x: ICE when using bcmp with int length argument 16794 16795 SCO specific 16796 16797 * [226]7623: SCO OpenServer build fails with machmode.def: undefined 16798 symbol: BITS_PER_UNIT 16799 16800 m68k/Coldfire specific 16801 16802 * [227]8314: crtbegin, crtend need to be multilib'ed for this 16803 platform 16804 16805 Documentation 16806 16807 * [228]761: Document some undocumented options 16808 * [229]5610: Fix documentation about invoking SSE instructions 16809 (-mfpmath=sse) 16810 * [230]7484: List -Wmissing-declarations as C-only option 16811 * [231]7531: -mcmodel not documented for x86-64 16812 * [232]8120: Update documentation of bad use of ## 16813 __________________________________________________________________ 16814 16815 GCC 3.2 16816 16817 3.2 is a small bug fix release, but there is a change to the 16818 application binary interface (ABI), hence the change to the second part 16819 of the version number. 16820 16821 The main purpose of the 3.2 release is to correct a couple of problems 16822 in the C++ ABI, with the intention of providing a stable interface 16823 going forward. Accordingly, 3.2 is only a small change to 3.1.1. 16824 16825 Bug Fixes 16826 16827 C++ 16828 16829 * [233]7320: g++ 3.2 relocation problem 16830 * [234]7470: vtable: virtual function pointers not in declaration 16831 order 16832 16833 libstdc++ 16834 16835 * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t 16836 * [236]6503, [237]6642, [238]7186: Problems with comparing or 16837 subtracting various types of const and non-const iterators 16838 * [239]7216: ambiguity with basic_iostream::traits_type 16839 * [240]7220: problem with basic_istream::ignore(0,delimiter) 16840 * [241]7222: locale::operator==() doesn't work on std::locale("") 16841 * [242]7286: placement operator delete issue 16842 * [243]7442: cxxabi.h does not match the C++ ABI 16843 * [244]7445: poor performance of std::locale::classic() in 16844 multi-threaded applications 16845 16846 x86-64 specific 16847 16848 * [245]7291: off-by-one in generated inline bzero code for x86-64 16849 16850 16851 For questions related to the use of GCC, please consult these web 16852 pages and the [246]GCC manuals. If that fails, the 16853 [247]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16854 web pages and the development of GCC are welcome on our developer 16855 list at [248]gcc (a] gcc.gnu.org. All of [249]our lists have public 16856 archives. 16857 16858 Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and 16859 distribution of this entire article is permitted in any medium, 16860 provided this notice is preserved. 16861 16862 These pages are [251]maintained by the GCC team. Last modified 16863 2018-09-30[252]. 16864 16865 References 16866 16867 1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 16868 2. http://gcc.gnu.org/gcc-3.1/changes.html 16869 3. http://gcc.gnu.org/gcc-3.2/c++-abi.html 16870 4. https://gcc.gnu.org/PR3782 16871 5. https://gcc.gnu.org/PR6440 16872 6. https://gcc.gnu.org/PR7050 16873 7. https://gcc.gnu.org/PR7741 16874 8. https://gcc.gnu.org/PR7982 16875 9. https://gcc.gnu.org/PR8068 16876 10. https://gcc.gnu.org/PR8178 16877 11. https://gcc.gnu.org/PR8396 16878 12. https://gcc.gnu.org/PR8674 16879 13. https://gcc.gnu.org/PR9768 16880 14. https://gcc.gnu.org/PR9798 16881 15. https://gcc.gnu.org/PR9799 16882 16. https://gcc.gnu.org/PR9928 16883 17. https://gcc.gnu.org/PR10114 16884 18. https://gcc.gnu.org/PR10352 16885 19. https://gcc.gnu.org/PR10336 16886 20. https://gcc.gnu.org/PR8224 16887 21. https://gcc.gnu.org/PR8613 16888 22. https://gcc.gnu.org/PR8828 16889 23. https://gcc.gnu.org/PR9226 16890 24. https://gcc.gnu.org/PR9853 16891 25. https://gcc.gnu.org/PR9797 16892 26. https://gcc.gnu.org/PR9967 16893 27. https://gcc.gnu.org/PR10116 16894 28. https://gcc.gnu.org/PR10171 16895 29. https://gcc.gnu.org/PR10175 16896 30. https://gcc.gnu.org/PR8316 16897 31. https://gcc.gnu.org/PR9169 16898 32. https://gcc.gnu.org/PR9420 16899 33. https://gcc.gnu.org/PR9459 16900 34. https://gcc.gnu.org/PR9507 16901 35. https://gcc.gnu.org/PR9538 16902 36. https://gcc.gnu.org/PR9602 16903 37. https://gcc.gnu.org/PR9993 16904 38. https://gcc.gnu.org/PR10167 16905 39. https://gcc.gnu.org/PR9652 16906 40. https://gcc.gnu.org/PR10144 16907 41. https://gcc.gnu.org/PR8746 16908 42. https://gcc.gnu.org/PR9888 16909 43. https://gcc.gnu.org/PR9638 16910 44. https://gcc.gnu.org/PR9954 16911 45. https://gcc.gnu.org/PR7784 16912 46. https://gcc.gnu.org/PR7796 16913 47. https://gcc.gnu.org/PR8281 16914 48. https://gcc.gnu.org/PR8366 16915 49. https://gcc.gnu.org/PR8726 16916 50. https://gcc.gnu.org/PR9414 16917 51. https://gcc.gnu.org/PR10067 16918 52. https://gcc.gnu.org/PR7248 16919 53. https://gcc.gnu.org/PR8343 16920 54. https://gcc.gnu.org/PR9732 16921 55. https://gcc.gnu.org/PR10073 16922 56. https://gcc.gnu.org/PR7702 16923 57. https://gcc.gnu.org/PR9671 16924 58. https://gcc.gnu.org/PR8694 16925 59. https://gcc.gnu.org/PR9953 16926 60. https://gcc.gnu.org/PR10271 16927 61. https://gcc.gnu.org/PR6362 16928 62. https://gcc.gnu.org/PR10377 16929 63. https://gcc.gnu.org/PR6955 16930 64. https://gcc.gnu.org/PR5919 16931 65. https://gcc.gnu.org/PR7129 16932 66. https://gcc.gnu.org/PR7507 16933 67. https://gcc.gnu.org/PR7622 16934 68. https://gcc.gnu.org/PR7681 16935 69. https://gcc.gnu.org/PR9528 16936 70. https://gcc.gnu.org/PR8031 16937 71. https://gcc.gnu.org/PR8275 16938 72. https://gcc.gnu.org/PR8332 16939 73. https://gcc.gnu.org/PR8372 16940 74. https://gcc.gnu.org/PR8439 16941 75. https://gcc.gnu.org/PR8442 16942 76. https://gcc.gnu.org/PR8518 16943 77. https://gcc.gnu.org/PR8615 16944 78. https://gcc.gnu.org/PR8663 16945 79. https://gcc.gnu.org/PR8799 16946 80. https://gcc.gnu.org/PR9328 16947 81. https://gcc.gnu.org/PR9465 16948 82. https://gcc.gnu.org/PR47 16949 83. https://gcc.gnu.org/PR6745 16950 84. https://gcc.gnu.org/PR8214 16951 85. https://gcc.gnu.org/PR8493 16952 86. https://gcc.gnu.org/PR8332 16953 87. https://gcc.gnu.org/PR8503 16954 88. https://gcc.gnu.org/PR8727 16955 89. https://gcc.gnu.org/PR7445 16956 90. https://gcc.gnu.org/PR8230 16957 91. https://gcc.gnu.org/PR8399 16958 92. https://gcc.gnu.org/PR8662 16959 93. https://gcc.gnu.org/PR8707 16960 94. https://gcc.gnu.org/PR8708 16961 95. https://gcc.gnu.org/PR8790 16962 96. https://gcc.gnu.org/PR8887 16963 97. https://gcc.gnu.org/PR9076 16964 98. https://gcc.gnu.org/PR9151 16965 99. https://gcc.gnu.org/PR9168 16966 100. https://gcc.gnu.org/PR9269 16967 101. https://gcc.gnu.org/PR9322 16968 102. https://gcc.gnu.org/PR9433 16969 103. https://gcc.gnu.org/PR8032 16970 104. https://gcc.gnu.org/PR8639 16971 105. https://gcc.gnu.org/PR8794 16972 106. https://gcc.gnu.org/PR8832 16973 107. https://gcc.gnu.org/PR8988 16974 108. https://gcc.gnu.org/PR9492 16975 109. https://gcc.gnu.org/PR9267 16976 110. https://gcc.gnu.org/PR8344 16977 111. https://gcc.gnu.org/PR8524 16978 112. https://gcc.gnu.org/PR8880 16979 113. https://gcc.gnu.org/PR9090 16980 114. https://gcc.gnu.org/PR8588 16981 115. https://gcc.gnu.org/PR8599 16982 116. https://gcc.gnu.org/PR9506 16983 117. https://gcc.gnu.org/PR9484 16984 118. https://gcc.gnu.org/PR9292 16985 119. https://gcc.gnu.org/PR9293 16986 120. https://gcc.gnu.org/PR9295 16987 121. https://gcc.gnu.org/PR9296 16988 122. https://gcc.gnu.org/PR9316 16989 123. https://gcc.gnu.org/PR9493 16990 124. https://gcc.gnu.org/PR7341 16991 125. https://gcc.gnu.org/PR8947 16992 126. https://gcc.gnu.org/PR7448 16993 127. https://gcc.gnu.org/PR8882 16994 128. https://gcc.gnu.org/PR7445 16995 129. https://gcc.gnu.org/PR2521 16996 130. https://gcc.gnu.org/PR5661 16997 131. https://gcc.gnu.org/PR6419 16998 132. https://gcc.gnu.org/PR6994 16999 133. https://gcc.gnu.org/PR7150 17000 134. https://gcc.gnu.org/PR7160 17001 135. https://gcc.gnu.org/PR7228 17002 136. https://gcc.gnu.org/PR7266 17003 137. https://gcc.gnu.org/PR7353 17004 138. https://gcc.gnu.org/PR7411 17005 139. https://gcc.gnu.org/PR7478 17006 140. https://gcc.gnu.org/PR7526 17007 141. https://gcc.gnu.org/PR7721 17008 142. https://gcc.gnu.org/PR7803 17009 143. https://gcc.gnu.org/PR7754 17010 144. https://gcc.gnu.org/PR7788 17011 145. https://gcc.gnu.org/PR8031 17012 146. https://gcc.gnu.org/PR8055 17013 147. https://gcc.gnu.org/PR8067 17014 148. https://gcc.gnu.org/PR8134 17015 149. https://gcc.gnu.org/PR8149 17016 150. https://gcc.gnu.org/PR8160 17017 151. https://gcc.gnu.org/PR5607 17018 152. https://gcc.gnu.org/PR6579 17019 153. https://gcc.gnu.org/PR6803 17020 154. https://gcc.gnu.org/PR7176 17021 155. https://gcc.gnu.org/PR7188 17022 156. https://gcc.gnu.org/PR7306 17023 157. https://gcc.gnu.org/PR7461 17024 158. https://gcc.gnu.org/PR7524 17025 159. https://gcc.gnu.org/PR7584 17026 160. https://gcc.gnu.org/PR7676 17027 161. https://gcc.gnu.org/PR7679 17028 162. https://gcc.gnu.org/PR7811 17029 163. https://gcc.gnu.org/PR7961 17030 164. https://gcc.gnu.org/PR8071 17031 165. https://gcc.gnu.org/PR8127 17032 166. https://gcc.gnu.org/PR6745 17033 167. https://gcc.gnu.org/PR8096 17034 168. https://gcc.gnu.org/PR8127 17035 169. https://gcc.gnu.org/PR8218 17036 170. https://gcc.gnu.org/PR8287 17037 171. https://gcc.gnu.org/PR8347 17038 172. https://gcc.gnu.org/PR8348 17039 173. https://gcc.gnu.org/PR8391 17040 174. https://gcc.gnu.org/PR6627 17041 175. https://gcc.gnu.org/PR6631 17042 176. https://gcc.gnu.org/PR7102 17043 177. https://gcc.gnu.org/PR7120 17044 178. https://gcc.gnu.org/PR7209 17045 179. https://gcc.gnu.org/PR7515 17046 180. https://gcc.gnu.org/PR7814 17047 181. https://gcc.gnu.org/PR8467 17048 182. https://gcc.gnu.org/PR4890 17049 183. https://gcc.gnu.org/PR7357 17050 184. https://gcc.gnu.org/PR7358 17051 185. https://gcc.gnu.org/PR7602 17052 186. https://gcc.gnu.org/PR7862 17053 187. https://gcc.gnu.org/PR8190 17054 188. https://gcc.gnu.org/PR8524 17055 189. https://gcc.gnu.org/PR5351 17056 190. https://gcc.gnu.org/PR7591 17057 191. https://gcc.gnu.org/PR6845 17058 192. https://gcc.gnu.org/PR7034 17059 193. https://gcc.gnu.org/PR7124 17060 194. https://gcc.gnu.org/PR7174 17061 195. https://gcc.gnu.org/PR7134 17062 196. https://gcc.gnu.org/PR7375 17063 197. https://gcc.gnu.org/PR7390 17064 198. https://gcc.gnu.org/PR6890 17065 199. https://gcc.gnu.org/PR6981 17066 200. https://gcc.gnu.org/PR7242 17067 201. https://gcc.gnu.org/PR7396 17068 202. https://gcc.gnu.org/PR7630 17069 203. https://gcc.gnu.org/PR7693 17070 204. https://gcc.gnu.org/PR7723 17071 205. https://gcc.gnu.org/PR7951 17072 206. https://gcc.gnu.org/PR8146 17073 207. https://gcc.gnu.org/PR5967 17074 208. https://gcc.gnu.org/PR6984 17075 209. https://gcc.gnu.org/PR7114 17076 210. https://gcc.gnu.org/PR7130 17077 211. https://gcc.gnu.org/PR7133 17078 212. https://gcc.gnu.org/PR7380 17079 213. https://gcc.gnu.org/PR8252 17080 214. https://gcc.gnu.org/PR8451 17081 215. https://gcc.gnu.org/PR7250 17082 216. https://gcc.gnu.org/PR6668 17083 217. https://gcc.gnu.org/PR7151 17084 218. https://gcc.gnu.org/PR7335 17085 219. https://gcc.gnu.org/PR7842 17086 220. https://gcc.gnu.org/PR7856 17087 221. https://gcc.gnu.org/PR7967 17088 222. https://gcc.gnu.org/PR7374 17089 223. https://gcc.gnu.org/PR7370 17090 224. https://gcc.gnu.org/PR7409 17091 225. https://gcc.gnu.org/PR8232 17092 226. https://gcc.gnu.org/PR7623 17093 227. https://gcc.gnu.org/PR8314 17094 228. https://gcc.gnu.org/PR761 17095 229. https://gcc.gnu.org/PR5610 17096 230. https://gcc.gnu.org/PR7484 17097 231. https://gcc.gnu.org/PR7531 17098 232. https://gcc.gnu.org/PR8120 17099 233. https://gcc.gnu.org/PR7320 17100 234. https://gcc.gnu.org/PR7470 17101 235. https://gcc.gnu.org/PR6410 17102 236. https://gcc.gnu.org/PR6503 17103 237. https://gcc.gnu.org/PR6642 17104 238. https://gcc.gnu.org/PR7186 17105 239. https://gcc.gnu.org/PR7216 17106 240. https://gcc.gnu.org/PR7220 17107 241. https://gcc.gnu.org/PR7222 17108 242. https://gcc.gnu.org/PR7286 17109 243. https://gcc.gnu.org/PR7442 17110 244. https://gcc.gnu.org/PR7445 17111 245. https://gcc.gnu.org/PR7291 17112 246. https://gcc.gnu.org/onlinedocs/ 17113 247. mailto:gcc-help (a] gcc.gnu.org 17114 248. mailto:gcc (a] gcc.gnu.org 17115 249. https://gcc.gnu.org/lists.html 17116 250. https://www.fsf.org/ 17117 251. https://gcc.gnu.org/about.html 17118 252. http://validator.w3.org/check/referer 17119 ====================================================================== 17120 http://gcc.gnu.org/gcc-3.1/index.html 17121 GCC 3.1 17122 17123 July 27, 2002 17124 17125 The [1]GNU project and the GCC developers are pleased to announce the 17126 release of GCC 3.1.1. 17127 17128 The links below still apply to GCC 3.1.1. 17129 17130 This release series is no longer maintained. 17131 17132 May 15, 2002 17133 17134 The [2]GNU project and the GCC developers are pleased to announce the 17135 release of GCC 3.1. 17136 17137 GCC used to stand for the GNU C Compiler, but since the compiler 17138 supports several other languages aside from C, it now stands for the 17139 GNU Compiler Collection. 17140 17141 A list of [3]successful builds is updated as new information becomes 17142 available. 17143 17144 The GCC developers would like to thank the numerous people that have 17145 contributed [4]new features, improvements, bug fixes, and other changes 17146 as well as test results to GCC. This [5]amazing group of volunteers is 17147 what makes GCC successful. 17148 17149 For additional information about GCC please refer to the [6]GCC project 17150 web site or contact the [7]GCC development mailing list. 17151 17152 To obtain GCC please use [8]our mirror sites, or our CVS server. 17153 __________________________________________________________________ 17154 17155 17156 For questions related to the use of GCC, please consult these web 17157 pages and the [9]GCC manuals. If that fails, the 17158 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17159 web pages and the development of GCC are welcome on our developer 17160 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 17161 archives. 17162 17163 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 17164 distribution of this entire article is permitted in any medium, 17165 provided this notice is preserved. 17166 17167 These pages are [14]maintained by the GCC team. Last modified 17168 2018-09-30[15]. 17169 17170 References 17171 17172 1. http://www.gnu.org/ 17173 2. http://www.gnu.org/ 17174 3. http://gcc.gnu.org/gcc-3.1/buildstat.html 17175 4. http://gcc.gnu.org/gcc-3.1/changes.html 17176 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 17177 6. http://gcc.gnu.org/index.html 17178 7. mailto:gcc (a] gcc.gnu.org 17179 8. http://gcc.gnu.org/mirrors.html 17180 9. https://gcc.gnu.org/onlinedocs/ 17181 10. mailto:gcc-help (a] gcc.gnu.org 17182 11. mailto:gcc (a] gcc.gnu.org 17183 12. https://gcc.gnu.org/lists.html 17184 13. https://www.fsf.org/ 17185 14. https://gcc.gnu.org/about.html 17186 15. http://validator.w3.org/check/referer 17187 ====================================================================== 17188 http://gcc.gnu.org/gcc-3.1/changes.html 17189 GCC 3.1 Release Series 17190 Changes, New Features, and Fixes 17191 17192 Additional changes in GCC 3.1.1 17193 17194 * A bug related to how structures and unions are returned has been 17195 fixed for powerpc-*-netbsd*. 17196 * An important bug in the implementation of -fprefetch-loop-arrays 17197 has been fixed. Previously the optimization prefetched random 17198 blocks of memory for most targets except for i386. 17199 * The Java compiler now compiles Java programs much faster and also 17200 works with parallel make. 17201 * Nested functions have been fixed for mips*-*-netbsd*. 17202 * Some missing floating point support routines have beed added for 17203 mips*-*-netbsd*. 17204 * This [1]message gives additional information about the bugs fixed 17205 in this release. 17206 17207 Caveats 17208 17209 * The -traditional C compiler option has been deprecated and will be 17210 removed in GCC 3.3. (It remains possible to preprocess non-C code 17211 with the traditional preprocessor.) 17212 * The default debugging format for most ELF platforms (including 17213 GNU/Linux and FreeBSD; notable exception is Solaris) has changed 17214 from stabs to DWARF2. This requires GDB 5.1.1 or later. 17215 17216 General Optimizer Improvements 17217 17218 * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat, 17219 and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure 17220 for profile driven optimizations. 17221 Options -fprofile-arcs and -fbranch-probabilities can now be used 17222 to improve speed of the generated code by profiling the actual 17223 program behaviour on typical runs. In the absence of profile info 17224 the compiler attempts to guess the profile statically. 17225 * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to 17226 monitor performance of the generated code. 17227 According to the SPECInt2000 results on an AMD Athlon CPU, the code 17228 generated by GCC 3.1 is 6% faster on the average (8.2% faster with 17229 profile feedback) compared to GCC 3.0. The code produced by GCC 3.0 17230 is about 2.1% faster compared to 2.95.3. Tests were done using the 17231 -O2 -march=athlon command-line options. 17232 * Alexandre Oliva, of Red Hat, has generalized the tree inlining 17233 infrastructure developed by CodeSourcery, LLC for the C++ front 17234 end, so that it is now used in the C front end too. Inlining 17235 functions as trees exposes them earlier to the compiler, giving it 17236 more opportunities for optimization. 17237 * Support for data prefetching instructions has been added to the GCC 17238 back end and several targets. A new __builtin_prefetch intrinsic is 17239 available to explicitly insert prefetch instructions and 17240 experimental support for loop array prefetching has been added (see 17241 -fprefetch-loop-array documentation). 17242 * Support for emitting debugging information for macros has been 17243 added for DWARF2. It is activated using -g3. 17244 17245 New Languages and Language specific improvements 17246 17247 C/C++ 17248 17249 * A few more [4]ISO C99 features. 17250 * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0. 17251 * The preprocessor's symbol table has been merged with the symbol 17252 table of the C, C++ and Objective-C front ends. 17253 * The preprocessor consumes less memory than the preprocessor in GCC 17254 3.0, often significantly so. On normal input files, it typically 17255 consumes less memory than pre-3.0 cccp-based GCC, too. 17256 17257 C++ 17258 17259 * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std 17260 was a workaround to allow std compliant code to work with the 17261 non-std compliant libstdc++-v2. libstdc++-v3 is std compliant. 17262 * The C++ ABI has been fixed so that void (A::*)() const is mangled 17263 as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only 17264 affects pointer to cv-qualified member function types. 17265 * The C++ ABI has been changed to correctly handle this code: 17266 struct A { 17267 void operator delete[] (void *, size_t); 17268 }; 17269 17270 struct B : public A { 17271 }; 17272 17273 new B[10]; 17274 17275 The amount of storage allocated for the array will be greater than 17276 it was in 3.0, in order to store the number of elements in the 17277 array, so that the correct size can be passed to operator delete[] 17278 when the array is deleted. Previously, the value passed to operator 17279 delete[] was unpredictable. 17280 This change will only affect code that declares a two-argument 17281 operator delete[] with a second parameter of type size_t in a base 17282 class, and does not override that definition in a derived class. 17283 * The C++ ABI has been changed so that: 17284 struct A { 17285 void operator delete[] (void *, size_t); 17286 void operator delete[] (void *); 17287 }; 17288 17289 does not cause unnecessary storage to be allocated when an array of 17290 A objects is allocated. 17291 This change will only affect code that declares both of these forms 17292 of operator delete[], and declared the two-argument form before the 17293 one-argument form. 17294 * The C++ ABI has been changed so that when a parameter is passed by 17295 value, any cleanup for that parameter is performed in the caller, 17296 as specified by the ia64 C++ ABI, rather than the called function 17297 as before. As a result, classes with a non-trivial destructor but a 17298 trivial copy constructor will be passed and returned by invisible 17299 reference, rather than by bitwise copy as before. 17300 * G++ now supports the "named return value optimization": for code 17301 like 17302 A f () { 17303 A a; 17304 ... 17305 return a; 17306 } 17307 17308 G++ will allocate a in the return value slot, so that the return 17309 becomes a no-op. For this to work, all return statements in the 17310 function must return the same variable. 17311 * Improvements to the C++ library are listed in [5]the libstdc++-v3 17312 FAQ. 17313 17314 Objective-C 17315 17316 * Annoying linker warnings (due to incorrect code being generated) 17317 have been fixed. 17318 * If a class method cannot be found, the compiler no longer issues a 17319 warning if a corresponding instance method exists in the root 17320 class. 17321 * Forward @protocol declarations have been fixed. 17322 * Loading of categories has been fixed in certain situations (GNU run 17323 time only). 17324 * The class lookup in the run-time library has been rewritten so that 17325 class method dispatch is more than twice as fast as it used to be 17326 (GNU run time only). 17327 17328 Java 17329 17330 * libgcj now includes RMI, java.lang.ref.*, javax.naming, and 17331 javax.transaction. 17332 * Property files and other system resources can be compiled into 17333 executables which use libgcj using the new gcj --resource feature. 17334 * libgcj has been ported to more platforms. In particular there is 17335 now a mostly-functional mingw32 (Windows) target port. 17336 * JNI and CNI invocation interfaces were implemented, so gcj-compiled 17337 Java code can now be called from a C/C++ application. 17338 * gcj can now use builtin functions for certain known methods, for 17339 instance Math.cos. 17340 * gcj can now automatically remove redundant array-store checks in 17341 some common cases. 17342 * The --no-store-checks optimization option was added. This can be 17343 used to omit runtime store checks for code which is known not to 17344 throw ArrayStoreException 17345 * The following third party interface standards were added to libgcj: 17346 org.w3c.dom and org.xml.sax. 17347 * java.security has been merged with GNU Classpath. The new package 17348 is now JDK 1.2 compliant, and much more complete. 17349 * A bytecode verifier was added to the libgcj interpreter. 17350 * java.lang.Character was rewritten to comply with the Unicode 3.0 17351 standard, and improve performance. 17352 * Partial support for many more locales was added to libgcj. 17353 * Socket timeouts have been implemented. 17354 * libgcj has been merged into a single shared library. There are no 17355 longer separate shared libraries for the garbage collector and 17356 zlib. 17357 * Several performance improvements were made to gcj and libgcj: 17358 + Hash synchronization (thin locks) 17359 + A special allocation path for finalizer-free objects 17360 + Thread-local allocation 17361 + Parallel GC, and other GC tweaks 17362 17363 Fortran 17364 17365 Fortran improvements are listed in [6]the Fortran documentation. 17366 17367 Ada 17368 17369 [7]AdaCore, has contributed its GNAT Ada 95 front end and associated 17370 tools. The GNAT compiler fully implements the Ada language as defined 17371 by the ISO/IEC 8652 standard. 17372 17373 Please note that the integration of the Ada front end is still work in 17374 progress. 17375 17376 New Targets and Target Specific Improvements 17377 17378 * Hans-Peter Nilsson has contributed a port to MMIX, the CPU 17379 architecture used in new editions of Donald E. Knuth's The Art of 17380 Computer Programming. 17381 * Axis Communications has contributed its port to the CRIS CPU 17382 architecture, used in the ETRAX system-on-a-chip series. 17383 * Alexandre Oliva, of Red Hat, has contributed a port to the SuperH 17384 SH5 64-bit RISC microprocessor architecture, extending the existing 17385 SH port. 17386 * UltraSPARC is fully supported in 64-bit mode. The option -m64 17387 enables it. 17388 * For compatibility with the Sun compiler #pragma redefine_extname 17389 has been implemented on Solaris. 17390 * The x86 back end has had some noticeable work done to it. 17391 + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas 17392 Jaeger have contributed a port to the AMD x86-64 architecture. 17393 For more information on x86-64 see http://www.x86-64.org. 17394 + The compiler now supports MMX, 3DNow!, SSE, and SSE2 17395 instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will 17396 enable the respective instruction sets. Intel C++ compatible 17397 MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics 17398 will be added in next major release. 17399 + Following those improvements, targets for Pentium MMX, K6-2, 17400 K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were 17401 added. Refer to the documentation on -march= and -mcpu= 17402 options for details. 17403 + For those targets that support it, -mfpmath=sse will cause the 17404 compiler to generate SSE/SSE2 instructions for floating point 17405 math instead of x87 instructions. Usually, this will lead to 17406 quicker code especially on the Pentium 4. Note that only 17407 scalar floating point instructions are used and GCC does not 17408 exploit SIMD features yet. 17409 + Prefetch support has been added to the Pentium III, Pentium 4, 17410 K6-2, K6-3, and Athlon series. 17411 + Code generated for floating point to integer conversions has 17412 been improved leading to better performance of many 3D 17413 applications. 17414 * The PowerPC back end has added 64-bit PowerPC GNU/Linux support. 17415 * C++ support for AIX has been improved. 17416 * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the 17417 PowerPC port supporting the AltiVec programming model (SIMD). The 17418 support, though presently useful, is experimental and is expected 17419 to stabilize for 3.2. The support is written to conform to 17420 Motorola's AltiVec specs. See -maltivec. 17421 17422 Obsolete Systems 17423 17424 Support for a number of older systems has been declared obsolete in GCC 17425 3.1. Unless there is activity to revive them, the next release of GCC 17426 will have their sources permanently removed. 17427 17428 All configurations of the following processor architectures have been 17429 declared obsolete: 17430 * MIL-STD-1750A, 1750a-*-* 17431 * AMD A29k, a29k-*-* 17432 * Convex, c*-convex-* 17433 * Clipper, clipper-*-* 17434 * Elxsi, elxsi-*-* 17435 * Intel i860, i860-*-* 17436 * Sun picoJava, pj-*-* and pjl-*-* 17437 * Western Electric 32000, we32k-*-* 17438 17439 Most configurations of the following processor architectures have been 17440 declared obsolete, but we are preserving a few systems which may have 17441 active developers. It is unlikely that the remaining systems will 17442 survive much longer unless we see definite signs of port activity. 17443 * Motorola 88000 except 17444 + Generic a.out, m88k-*-aout* 17445 + Generic SVR4, m88k-*-sysv4 17446 + OpenBSD, m88k-*-openbsd* 17447 * NS32k except 17448 + NetBSD, ns32k-*-netbsd* 17449 + OpenBSD, ns32k-*-openbsd*. 17450 * ROMP except 17451 + OpenBSD, romp-*-openbsd*. 17452 17453 Finally, only some configurations of these processor architectures are 17454 being obsoleted. 17455 * Alpha: 17456 + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka 17457 alpha*-*-osf[45], are still supported.) 17458 * ARM: 17459 + RISCiX, arm-*-riscix*. 17460 * i386: 17461 + 386BSD, i?86-*-bsd* 17462 + Chorus, i?86-*-chorusos* 17463 + DG/UX, i?86-*-dgux* 17464 + FreeBSD 1.x, i?86-*-freebsd1.* 17465 + IBM AIX, i?86-*-aix* 17466 + ISC UNIX, i?86-*-isc* 17467 + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld* 17468 + NEXTstep, i?86-next-* 17469 + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose* 17470 + RTEMS/coff, i?86-*-rtemscoff* 17471 + RTEMS/go32, i?86-go32-rtems* 17472 + Sequent/BSD, i?86-sequent-bsd* 17473 + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and 17474 i?86-sequent-sysv3* 17475 + SunOS, i?86-*-sunos* 17476 * Motorola 68000: 17477 + Altos, m68[k0]*-altos-* 17478 + Apollo, m68[k0]*-apollo-* 17479 + Apple A/UX, m68[k0]*-apple-* 17480 + Bull, m68[k0]*-bull-* 17481 + Convergent, m68[k0]*-convergent-* 17482 + Generic SVR3, m68[k0]*-*-sysv3* 17483 + ISI, m68[k0]*-isi-* 17484 + LynxOS, m68[k0]*-*-lynxos* 17485 + NEXT, m68[k0]*-next-* 17486 + RTEMS/coff, m68[k0]*-*-rtemscoff* 17487 + Sony, m68[k0]*-sony-* 17488 * MIPS: 17489 + DEC Ultrix, mips-*-ultrix* and mips-dec-* 17490 + Generic BSD, mips-*-bsd* 17491 + Generic System V, mips-*-sysv* 17492 + IRIX before version 5, mips-sgi-irix[1234]* 17493 + RiscOS, mips-*-riscos* 17494 + Sony, mips-sony-* 17495 + Tandem, mips-tandem-* 17496 * SPARC: 17497 + RTEMS/a.out, sparc-*-rtemsaout*. 17498 17499 Documentation improvements 17500 17501 * The old manual ("Using and Porting the GNU Compiler Collection") 17502 has been replaced by a users manual ("Using the GNU Compiler 17503 Collection") and a separate internals reference manual ("GNU 17504 Compiler Collection Internals"). 17505 * More complete and much improved documentation about GCC's internal 17506 representation used by the C and C++ front ends. 17507 * Many cleanups and improvements in general. 17508 17509 17510 For questions related to the use of GCC, please consult these web 17511 pages and the [8]GCC manuals. If that fails, the 17512 [9]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17513 web pages and the development of GCC are welcome on our developer 17514 list at [10]gcc (a] gcc.gnu.org. All of [11]our lists have public 17515 archives. 17516 17517 Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and 17518 distribution of this entire article is permitted in any medium, 17519 provided this notice is preserved. 17520 17521 These pages are [13]maintained by the GCC team. Last modified 17522 2018-09-30[14]. 17523 17524 References 17525 17526 1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html 17527 2. http://gcc.gnu.org/news/profiledriven.html 17528 3. http://gcc.gnu.org/benchmarks/ 17529 4. http://gcc.gnu.org/c99status.html 17530 5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html 17531 6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html 17532 7. https://www.adacore.com/ 17533 8. https://gcc.gnu.org/onlinedocs/ 17534 9. mailto:gcc-help (a] gcc.gnu.org 17535 10. mailto:gcc (a] gcc.gnu.org 17536 11. https://gcc.gnu.org/lists.html 17537 12. https://www.fsf.org/ 17538 13. https://gcc.gnu.org/about.html 17539 14. http://validator.w3.org/check/referer 17540 ====================================================================== 17541 http://gcc.gnu.org/gcc-3.0/index.html 17542 GCC 3.0.4 17543 17544 February 20, 2002 17545 17546 The [1]GNU project and the GCC developers are pleased to announce the 17547 release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0 17548 series. 17549 17550 This release series is no longer maintained. 17551 17552 GCC used to stand for the GNU C Compiler, but since the compiler 17553 supports several other languages aside from C, it now stands for the 17554 GNU Compiler Collection. 17555 17556 GCC 3.0.x has several new optimizations, new targets, new languages and 17557 many other new features, relative to GCC 2.95.x. See the [2]new 17558 features page for a more complete list. 17559 17560 A list of [3]successful builds is updated as new information becomes 17561 available. 17562 17563 The GCC developers would like to thank the numerous people that have 17564 contributed new features, test results, bug fixes, etc to GCC. This 17565 [4]amazing group of volunteers is what makes GCC successful. 17566 17567 And finally, we can't in good conscience fail to mention some 17568 [5]caveats to using GCC 3.0.x. 17569 17570 For additional information about GCC please refer to the [6]GCC project 17571 web site or contact the [7]GCC development mailing list. 17572 17573 To obtain GCC please use [8]our mirror sites, or our CVS server. 17574 __________________________________________________________________ 17575 17576 Previous 3.0.x Releases 17577 17578 December 20, 2001: GCC 3.0.3 has been released. 17579 October 25, 2001: GCC 3.0.2 has been released. 17580 August 20, 2001: GCC 3.0.1 has been released. 17581 June 18, 2001: GCC 3.0 has been released. 17582 17583 17584 For questions related to the use of GCC, please consult these web 17585 pages and the [9]GCC manuals. If that fails, the 17586 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17587 web pages and the development of GCC are welcome on our developer 17588 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 17589 archives. 17590 17591 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 17592 distribution of this entire article is permitted in any medium, 17593 provided this notice is preserved. 17594 17595 These pages are [14]maintained by the GCC team. Last modified 17596 2018-09-30[15]. 17597 17598 References 17599 17600 1. http://www.gnu.org/ 17601 2. http://gcc.gnu.org/gcc-3.0/features.html 17602 3. http://gcc.gnu.org/gcc-3.0/buildstat.html 17603 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 17604 5. http://gcc.gnu.org/gcc-3.0/caveats.html 17605 6. http://gcc.gnu.org/index.html 17606 7. mailto:gcc (a] gcc.gnu.org 17607 8. http://gcc.gnu.org/mirrors.html 17608 9. https://gcc.gnu.org/onlinedocs/ 17609 10. mailto:gcc-help (a] gcc.gnu.org 17610 11. mailto:gcc (a] gcc.gnu.org 17611 12. https://gcc.gnu.org/lists.html 17612 13. https://www.fsf.org/ 17613 14. https://gcc.gnu.org/about.html 17614 15. http://validator.w3.org/check/referer 17615 ====================================================================== 17616 http://gcc.gnu.org/gcc-3.0/features.html 17617 GCC 3.0 New Features 17618 17619 Additional changes in GCC 3.0.4 17620 17621 * GCC 3.0 now supports newer versions of the [1]NetBSD operating 17622 system, which use the ELF object file format, on x86 processors. 17623 * Correct debugging information is generated from functions that have 17624 lines from multiple files (e.g. yacc output). 17625 * A fix for whitespace handling in the -traditional preprocessor, 17626 which can affect Fortran. 17627 * Fixes to the exception handling runtime. 17628 * More fixes for bad code generation in C++. 17629 * A fix for shared library generation under AIX 4.3. 17630 * Documentation updates. 17631 * Port of GCC to Tensilica's Xtensa processor contributed. 17632 * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link). 17633 17634 Additional changes in GCC 3.0.3 17635 17636 * A fix to correct an accidental change to the PowerPC ABI. 17637 * Fixes for bad code generation on a variety of architectures. 17638 * Improvements to the debugging information generated for C++ 17639 classes. 17640 * Fixes for bad code generation in C++. 17641 * A fix to avoid crashes in the C++ demangler. 17642 * A fix to the C++ standard library to avoid buffer overflows. 17643 * Miscellaneous improvements for a variety of architectures. 17644 17645 Additional changes in GCC 3.0.2 17646 17647 * Fixes for bad code generation during loop unrolling. 17648 * Fixes for bad code generation by the sibling call optimization. 17649 * Minor improvements to x86 code generation. 17650 * Implementation of function descriptors in C++ vtables for IA64. 17651 * Numerous minor bug-fixes. 17652 17653 Additional changes in GCC 3.0.1 17654 17655 * C++ fixes for incorrect code-generation. 17656 * Improved cross-compiling support for the C++ standard library. 17657 * Fixes for some embedded targets that worked in GCC 2.95.3, but not 17658 in GCC 3.0. 17659 * Fixes for various exception-handling bugs. 17660 * A port to the S/390 architecture. 17661 17662 General Optimizer Improvements 17663 17664 * [2]Basic block reordering pass. 17665 * New if-conversion pass with support for conditional (predicated) 17666 execution. 17667 * New tail call and sibling call elimination optimizations. 17668 * New register renaming pass. 17669 * New (experimental) [3]static single assignment (SSA) representation 17670 support. 17671 * New dead-code elimination pass implemented using the SSA 17672 representation. 17673 * [4]Global null pointer test elimination. 17674 * [5]Global code hoisting/unification. 17675 * More builtins and optimizations for stdio.h, string.h and old BSD 17676 functions, as well as for ISO C99 functions. 17677 * New builtin __builtin_expect for giving hints to the branch 17678 predictor. 17679 17680 New Languages and Language specific improvements 17681 17682 * The GNU Compiler for the Java(TM) language (GCJ) is now integrated 17683 and supported, including the run-time library containing most 17684 common non-GUI Java classes, a bytecode interpreter, and the Boehm 17685 conservative garbage collector. Many bugs have been fixed. GCJ can 17686 compile Java source or Java bytecodes to either native code or Java 17687 class files, and supports native methods written in either the 17688 standard JNI or the more efficient and convenient CNI. 17689 * Here is a [6]partial list of C++ improvements, both new features 17690 and those no longer supported. 17691 * New C++ ABI. On the IA-64 platform GCC is capable of 17692 inter-operating with other IA-64 compilers. 17693 * The new ABI also significantly reduces the size of symbol and debug 17694 information. 17695 * New C++ support library and many C++ bug fixes, vastly improving 17696 our conformance to the ISO C++ standard. 17697 * New [7]inliner for C++. 17698 * Rewritten C preprocessor, integrated into the C, C++ and Objective 17699 C compilers, with very many improvements including ISO C99 support 17700 and [8]improvements to dependency generation. 17701 * Support for more [9]ISO C99 features. 17702 * Many improvements to support for checking calls to format functions 17703 such as printf and scanf, including support for ISO C99 format 17704 features, extensions from the Single Unix Specification and GNU 17705 libc 2.2, checking of strfmon formats and features to assist in 17706 auditing for format string security bugs. 17707 * New warnings for C code that may have undefined semantics because 17708 of violations of sequence point rules in the C standard (such as a 17709 = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall. 17710 * Additional warning option -Wfloat-equal. 17711 * Improvements to -Wtraditional. 17712 * Fortran improvements are listed in [10]the Fortran documentation. 17713 17714 New Targets and Target Specific Improvements 17715 17716 * New x86 back end, generating much improved code. 17717 * Support for a generic i386-elf target contributed. 17718 * New option to emit x86 assembly code using Intel style syntax 17719 (-mintel-syntax). 17720 * HPUX 11 support contributed. 17721 * Improved PowerPC code generation, including scheduled prologue and 17722 epilogue. 17723 * Port of GCC to Intel's IA-64 processor contributed. 17724 * Port of GCC to Motorola's MCore 210 and 340 contributed. 17725 * New unified back-end for Arm, Thumb and StrongArm contributed. 17726 * Port of GCC to Intel's XScale processor contributed. 17727 * Port of GCC to Atmel's AVR microcontrollers contributed. 17728 * Port of GCC to Mitsubishi's D30V processor contributed. 17729 * Port of GCC to Matsushita's AM33 processor (a member of the MN10300 17730 processor family) contributed. 17731 * Port of GCC to Fujitsu's FR30 processor contributed. 17732 * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed. 17733 * Port of GCC to Sun's picoJava processor core contributed. 17734 17735 Documentation improvements 17736 17737 * Substantially rewritten and improved C preprocessor manual. 17738 * Many improvements to other documentation. 17739 * Manpages for gcc, cpp and gcov are now generated automatically from 17740 the master Texinfo manual, eliminating the problem of manpages 17741 being out of date. (The generated manpages are only extracts from 17742 the full manual, which is provided in Texinfo form, from which 17743 info, HTML, other formats and a printed manual can be generated.) 17744 * Generated info files are included in the release tarballs alongside 17745 their Texinfo sources, avoiding problems on some platforms with 17746 building makeinfo as part of the GCC distribution. 17747 17748 Other significant improvements 17749 17750 * Garbage collection used internally by the compiler for most memory 17751 allocation instead of obstacks. 17752 * Lengauer and Tarjan algorithm used for computing dominators in the 17753 CFG. This algorithm can be significantly faster and more space 17754 efficient than our older algorithm. 17755 * gccbug script provided to assist in submitting bug reports to our 17756 bug tracking system. (Bug reports previously submitted directly to 17757 our mailing lists, for which you received no bug tracking number, 17758 should be submitted again using gccbug if you can reproduce the 17759 problem with GCC 3.0.) 17760 * The internal libgcc library is [11]built as a shared library on 17761 systems that support it. 17762 * Extensive testsuite included with GCC, with many new tests. In 17763 addition to tests for GCC bugs that have been fixed, many tests 17764 have been added for language features, compiler warnings and 17765 builtin functions. 17766 * Additional language-independent warning options -Wpacked, -Wpadded, 17767 -Wunreachable-code and -Wdisabled-optimization. 17768 * Target-independent options -falign-functions, -falign-loops and 17769 -falign-jumps. 17770 17771 Plus a great many bug fixes and almost all the [12]features found in 17772 GCC 2.95. 17773 17774 17775 For questions related to the use of GCC, please consult these web 17776 pages and the [13]GCC manuals. If that fails, the 17777 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17778 web pages and the development of GCC are welcome on our developer 17779 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 17780 archives. 17781 17782 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 17783 distribution of this entire article is permitted in any medium, 17784 provided this notice is preserved. 17785 17786 These pages are [18]maintained by the GCC team. Last modified 17787 2018-09-30[19]. 17788 17789 References 17790 17791 1. http://www.netbsd.org/ 17792 2. http://gcc.gnu.org/news/reorder.html 17793 3. http://gcc.gnu.org/news/ssa.html 17794 4. http://gcc.gnu.org/news/null.html 17795 5. http://gcc.gnu.org/news/unify.html 17796 6. http://gcc.gnu.org/gcc-3.0/c++features.html 17797 7. http://gcc.gnu.org/news/inlining.html 17798 8. http://gcc.gnu.org/news/dependencies.html 17799 9. http://gcc.gnu.org/c99status.html 17800 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 17801 11. http://gcc.gnu.org/gcc-3.0/libgcc.html 17802 12. http://gcc.gnu.org/gcc-2.95/features.html 17803 13. https://gcc.gnu.org/onlinedocs/ 17804 14. mailto:gcc-help (a] gcc.gnu.org 17805 15. mailto:gcc (a] gcc.gnu.org 17806 16. https://gcc.gnu.org/lists.html 17807 17. https://www.fsf.org/ 17808 18. https://gcc.gnu.org/about.html 17809 19. http://validator.w3.org/check/referer 17810 ====================================================================== 17811 http://gcc.gnu.org/gcc-3.0/caveats.html 17812 GCC 3.0 Caveats 17813 17814 * -fstrict-aliasing is now part of -O2 and higher optimization 17815 levels. This allows the compiler to assume the strictest aliasing 17816 rules applicable to the language being compiled. For C and C++, 17817 this activates optimizations based on the type of expressions. This 17818 optimization may thus break old, non-compliant code. 17819 * Enumerations are now properly promoted to int in function 17820 parameters and function returns. Normally this change is not 17821 visible, but when using -fshort-enums this is an ABI change. 17822 * The undocumented extension that allowed C programs to have a label 17823 at the end of a compound statement has been deprecated and may be 17824 removed in a future version. Programs that now generate a warning 17825 about this may be fixed by adding a null statement (a single 17826 semicolon) after the label. 17827 * The poorly documented extension that allowed string constants in C, 17828 C++ and Objective C to contain unescaped newlines has been 17829 deprecated and may be removed in a future version. Programs using 17830 this extension may be fixed in several ways: the bare newline may 17831 be replaced by \n, or preceded by \n\, or string concatenation may 17832 be used with the bare newline preceded by \n" and " placed at the 17833 start of the next line. 17834 * The Chill compiler is not included in GCC 3.0, because of the lack 17835 of a volunteer to convert it to use garbage collection. 17836 * Certain non-standard iostream methods from earlier versions of 17837 libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach, 17838 ostream::form, and istream::gets. 17839 * The new C++ ABI is not yet fully supported by current (as of 17840 2001-07-01) releases and development versions of GDB, or any 17841 earlier versions. There is a problem setting breakpoints by line 17842 number, and other related issues that have been fixed in GCC 3.0 17843 but not yet handled in GDB: 17844 [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 17845 17846 17847 For questions related to the use of GCC, please consult these web 17848 pages and the [2]GCC manuals. If that fails, the 17849 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17850 web pages and the development of GCC are welcome on our developer 17851 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 17852 17853 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 17854 distribution of this entire article is permitted in any medium, 17855 provided this notice is preserved. 17856 17857 These pages are [7]maintained by the GCC team. Last modified 17858 2018-09-30[8]. 17859 17860 References 17861 17862 1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 17863 2. https://gcc.gnu.org/onlinedocs/ 17864 3. mailto:gcc-help (a] gcc.gnu.org 17865 4. mailto:gcc (a] gcc.gnu.org 17866 5. https://gcc.gnu.org/lists.html 17867 6. https://www.fsf.org/ 17868 7. https://gcc.gnu.org/about.html 17869 8. http://validator.w3.org/check/referer 17870 ====================================================================== 17871 http://gcc.gnu.org/gcc-2.95/index.html 17872 GCC 2.95 17873 17874 March 16, 2001: The GNU project and the GCC developers are pleased to 17875 announce the release of GCC version 2.95.3. 17876 17877 This release is no longer maintained. 17878 17879 Release History 17880 17881 GCC 2.95.3 17882 March 16, 2001 17883 17884 GCC 2.95.2 17885 October 27, 1999 17886 17887 GCC 2.95.1 17888 August 19, 1999 17889 17890 GCC 2.95 17891 July 31, 1999. This is the first release of GCC since the April 17892 1999 GCC/EGCS reunification and includes nearly a year's worth 17893 of new development and bugfixes. 17894 17895 References and Acknowledgements 17896 17897 GCC used to stand for the GNU C Compiler, but since the compiler 17898 supports several other languages aside from C, it now stands for the 17899 GNU Compiler Collection. 17900 17901 The whole suite has been extensively [1]regression tested and 17902 [2]package tested. It should be reliable and suitable for widespread 17903 use. 17904 17905 The compiler has several new optimizations, new targets, new languages 17906 and other new features. See the [3]new features page for a more 17907 complete list of new features found in the GCC 2.95 releases. 17908 17909 The sources include installation instructions in both HTML and 17910 plaintext forms in the install directory in the distribution. However, 17911 the most up to date installation instructions and [4]build/test status 17912 are on the web pages. We will update those pages as new information 17913 becomes available. 17914 17915 The GCC developers would like to thank the numerous people that have 17916 contributed new features, test results, bugfixes, etc to GCC. This 17917 [5]amazing group of volunteers is what makes GCC successful. 17918 17919 And finally, we can't in good conscience fail to mention some 17920 [6]caveats to using GCC 2.95. 17921 17922 Download GCC 2.95 from one of our many [7]mirror sites. 17923 17924 For additional information about GCC please see the [8]GCC project web 17925 server or contact the [9]GCC development mailing list. 17926 17927 17928 For questions related to the use of GCC, please consult these web 17929 pages and the [10]GCC manuals. If that fails, the 17930 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17931 web pages and the development of GCC are welcome on our developer 17932 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 17933 archives. 17934 17935 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 17936 distribution of this entire article is permitted in any medium, 17937 provided this notice is preserved. 17938 17939 These pages are [15]maintained by the GCC team. Last modified 17940 2018-09-30[16]. 17941 17942 References 17943 17944 1. http://gcc.gnu.org/gcc-2.95/regress.html 17945 2. http://gcc.gnu.org/gcc-2.95/othertest.html 17946 3. http://gcc.gnu.org/gcc-2.95/features.html 17947 4. http://gcc.gnu.org/gcc-2.95/buildstat.html 17948 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 17949 6. http://gcc.gnu.org/gcc-2.95/caveats.html 17950 7. http://gcc.gnu.org/mirrors.html 17951 8. http://gcc.gnu.org/index.html 17952 9. mailto:gcc (a] gcc.gnu.org 17953 10. https://gcc.gnu.org/onlinedocs/ 17954 11. mailto:gcc-help (a] gcc.gnu.org 17955 12. mailto:gcc (a] gcc.gnu.org 17956 13. https://gcc.gnu.org/lists.html 17957 14. https://www.fsf.org/ 17958 15. https://gcc.gnu.org/about.html 17959 16. http://validator.w3.org/check/referer 17960 ====================================================================== 17961 http://gcc.gnu.org/gcc-2.95/features.html 17962 GCC 2.95 New Features 17963 17964 * General Optimizer Improvements: 17965 + [1]Localized register spilling to improve speed and code 17966 density especially on small register class machines. 17967 + [2]Global CSE using lazy code motion algorithms. 17968 + [3]Improved global constant/copy propagation. 17969 + [4]Improved control flow graph analysis and manipulation. 17970 + [5]Local dead store elimination. 17971 + [6]Memory Load hoisting/store sinking in loops. 17972 + [7]Type based alias analysis is enabled by default. Note this 17973 feature will expose bugs in the Linux kernel. Please refer to 17974 the FAQ (as shipped with GCC 2.95) for additional information 17975 on this issue. 17976 + Major revamp of GIV detection, combination and simplification 17977 to improve loop performance. 17978 + Major improvements to register allocation and reloading. 17979 * New Languages and Language specific improvements 17980 + [8]Many C++ improvements. 17981 + [9]Many Fortran improvements. 17982 + [10]Java front-end has been integrated. A [11]runtime library 17983 is available separately. 17984 + [12]ISO C99 support 17985 + [13]Chill front-end and runtime has been integrated. 17986 + Boehm garbage collector support in libobjc. 17987 + More support for various pragmas which appear in vendor 17988 include files 17989 * New Targets and Target Specific Improvements 17990 + [14]SPARC backend rewrite. 17991 + -mschedule=8000 will optimize code for PA8000 class 17992 processors; -mpa-risc-2-0 will generate code for PA2.0 17993 processors 17994 + Various micro-optimizations for the ia32 port. K6 17995 optimizations 17996 + Compiler will attempt to align doubles in the stack on the 17997 ia32 port 17998 + Alpha EV6 support 17999 + PowerPC 750 18000 + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403. 18001 -mcpu=e603e was added to do -mcpu=603e and -msoft-float. 18002 + c3x, c4x 18003 + HyperSPARC 18004 + SparcLite86x 18005 + sh4 18006 + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix, 18007 arm-linux) 18008 + vxWorks targets include support for vxWorks threads 18009 + StrongARM 110 and ARM9 support added. ARM Scheduling 18010 parameters rewritten. 18011 + Various changes to the MIPS port to avoid assembler macros, 18012 which in turn improves performance 18013 + Various performance improvements to the i960 port. 18014 + Major rewrite of ns32k port 18015 * Other significant improvements 18016 + [15]Ability to dump cfg information and display it using vcg. 18017 + The new faster scheme for fixing vendor header files is 18018 enabled by default. 18019 + Experimental internationalization support. 18020 + multibyte character support 18021 + Some compile-time speedups for pathological problems 18022 + Better support for complex types 18023 * Plus the usual mountain of bugfixes 18024 * Core compiler is based on the gcc2 development tree from Sept 30, 18025 1998, so we have all of the [16]features found in GCC 2.8. 18026 18027 Additional Changes in GCC 2.95.1 18028 18029 * Generic bugfixes and improvements 18030 + Various documentation fixes related to the GCC/EGCS merger. 18031 + Fix memory management bug which could lead to spurious aborts, 18032 core dumps or random parsing errors in the compiler. 18033 + Fix a couple bugs in the dwarf1 and dwarf2 debug record 18034 support. 18035 + Fix infinite loop in the CSE optimizer. 18036 + Avoid undefined behavior in compiler FP emulation code 18037 + Fix install problem when prefix is overridden on the make 18038 install command. 18039 + Fix problem with unwanted installation of assert.h on some 18040 systems. 18041 + Fix problem with finding the wrong assembler in a single tree 18042 build. 18043 + Avoid increasing the known alignment of a register that is 18044 already known to be a pointer. 18045 * Platform specific bugfixes and improvements 18046 + Codegen bugfix for prologue/epilogue for cpu32 target. 18047 + Fix long long code generation bug for the Coldfire target. 18048 + Fix various aborts in the SH compiler. 18049 + Fix bugs in libgcc support library for the SH. 18050 + Fix alpha ev6 code generation bug. 18051 + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on 18052 AIX platforms. 18053 + Fix -fpic code generation bug for rs6000/ppc svr4 targets. 18054 + Fix varargs/stdarg code generation bug for rs6000/ppc svr4 18055 targets. 18056 + Fix weak symbol handling for rs6000/ppc svr4 targets. 18057 + Fix various problems with 64bit code generation for the 18058 rs6000/ppc port. 18059 + Fix codegen bug which caused tetex to be mis-compiled on the 18060 x86. 18061 + Fix compiler abort in new cfg code exposed by x86 port. 18062 + Fix out of range array reference in code convert flat 18063 registers to the x87 stacked FP register file. 18064 + Fix minor vxworks configuration bug. 18065 + Fix return type of bsearch for SunOS 4.x. 18066 * Language & Runtime specific fixes. 18067 + The G++ signature extension has been deprecated. It will be 18068 removed in the next major release of G++. Use of signatures 18069 will result in a warning from the compiler. 18070 + Several bugs relating to templates and namespaces were fixed. 18071 + A bug that caused crashes when combining templates with -g on 18072 DWARF1 platforms was fixed. 18073 + Pointers-to-members, virtual functions, and multiple 18074 inheritance should now work together correctly. 18075 + Some code-generation bugs relating to function try blocks were 18076 fixed. 18077 + G++ is a little bit more lenient with certain archaic 18078 constructs than in GCC 2.95. 18079 + Fix to prevent shared library version #s from bring truncated 18080 to 1 digit 18081 + Fix missing std:: in the libstdc++ library. 18082 + Fix stream locking problems in libio. 18083 + Fix problem in java compiler driver. 18084 18085 Additional Changes in GCC 2.95.2 18086 18087 The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While 18088 the optimizations performed by -fstrict-aliasing are valid according to 18089 the C and C++ standards, the optimization have caused some problems, 18090 particularly with old non-conforming code. 18091 18092 The GCC developers are experimenting with ways to warn users about code 18093 which violates the C/C++ standards, but those warnings are not ready 18094 for widespread use at this time. Rather than wait for those warnings 18095 the GCC developers have chosen to disable -fstrict-aliasing by default 18096 for the GCC 2.95.2 release. 18097 18098 We strongly encourage developers to find and fix code which violates 18099 the C/C++ standards as -fstrict-aliasing may be enabled by default in 18100 future releases. Use the option -fstrict-aliasing to re-enable these 18101 optimizations. 18102 * Generic bugfixes and improvements 18103 + Fix incorrectly optimized memory reference in global common 18104 subexpression elimination (GCSE) optimization pass. 18105 + Fix code generation bug in regmove.c in which it could 18106 incorrectly change a "const" value. 18107 + Fix bug in optimization of conditionals involving volatile 18108 memory references. 18109 + Avoid over-allocation of stack space for some procedures. 18110 + Fixed bug in the compiler which caused incorrect optimization 18111 of an obscure series of bit manipulations, shifts and 18112 arithmetic. 18113 + Fixed register allocator bug which caused teTeX to be 18114 mis-compiled on SPARC targets. 18115 + Avoid incorrect optimization of degenerate case statements for 18116 certain targets such as the ARM. 18117 + Fix out of range memory reference in the jump optimizer. 18118 + Avoid dereferencing null pointer in fix-header. 18119 + Fix test for GCC specific features so that it is possible to 18120 bootstrap with gcc-2.6.2 and older versions of GCC. 18121 + Fix typo in scheduler which could potentially cause out of 18122 range memory accesses. 18123 + Avoid incorrect loop reversal which caused incorrect code for 18124 certain loops on PowerPC targets. 18125 + Avoid incorrect optimization of switch statements on certain 18126 targets (for example the ARM). 18127 * Platform specific bugfixes and improvements 18128 + Work around bug in Sun V5.0 compilers which caused bootstrap 18129 comparison failures on SPARC targets. 18130 + Fix SPARC backend bug which caused aborts in final.c. 18131 + Fix sparc-hal-solaris2* configuration fragments. 18132 + Fix bug in sparc block profiling. 18133 + Fix obscure code generation bug for the PARISC targets. 18134 + Define __STDC_EXT__ for HPUX configurations. 18135 + Various POWERPC64 code generation bugfixes. 18136 + Fix abort for PPC targets using ELF (ex GNU/Linux). 18137 + Fix collect2 problems for AIX targets. 18138 + Correct handling of .file directive for PPC targets. 18139 + Fix bug in fix_trunc x86 patterns. 18140 + Fix x86 port to correctly pop the FP stack for functions that 18141 return structures in memory. 18142 + Fix minor bug in strlen x86 pattern. 18143 + Use stabs debugging instead of dwarf1 for x86-solaris targets. 18144 + Fix template repository code to handle leading underscore in 18145 mangled names. 18146 + Fix weak/weak alias support for OpenBSD. 18147 + GNU/Linux for the ARM has C++ compatible include files. 18148 * Language & Runtime specific fixes. 18149 + Fix handling of constructor attribute in the C front-end which 18150 caused problems building the Chill runtime library on some 18151 targets. 18152 + Fix minor problem merging type qualifiers in the C front-end. 18153 + Fix aliasing bug for pointers and references (C/C++). 18154 + Fix incorrect "non-constant initializer bug" when -traditional 18155 or -fwritable-strings is enabled. 18156 + Fix build error for Chill front-end on SunOS. 18157 + Do not complain about duplicate instantiations when using 18158 -frepo (C++). 18159 + Fix array bounds handling in C++ front-end which caused 18160 problems with dwarf debugging information in some 18161 circumstances. 18162 + Fix minor namespace problem. 18163 + Fix problem linking java programs. 18164 18165 Additional Changes in GCC 2.95.3 18166 18167 * Generic bugfixes and improvements 18168 + Fix numerous problems that caused incorrect optimization in 18169 the register reloading code. 18170 + Fix numerous problems that caused incorrect optimization in 18171 the loop optimizer. 18172 + Fix aborts in the functions build_insn_chain and scan_loops 18173 under some circumstances. 18174 + Fix an alias analysis bug. 18175 + Fix an infinite compilation bug in the combiner. 18176 + A few problems with complex number support have been fixed. 18177 + It is no longer possible for gcc to act as a fork bomb when 18178 installed incorrectly. 18179 + The -fpack-struct option should be recognized now. 18180 + Fixed a bug that caused incorrect code to be generated due to 18181 a lost stack adjustment. 18182 * Platform specific bugfixes and improvements 18183 + Support building ARM toolchains hosted on Windows. 18184 + Fix attribute calculations in ARM toolchains. 18185 + arm-linux support has been improved. 18186 + Fix a PIC failure on sparc targets. 18187 + On ix86 targets, the regparm attribute should now work 18188 reliably. 18189 + Several updates for the h8300 port. 18190 + Fix problem building libio with glibc 2.2. 18191 18192 18193 For questions related to the use of GCC, please consult these web 18194 pages and the [17]GCC manuals. If that fails, the 18195 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18196 web pages and the development of GCC are welcome on our developer 18197 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 18198 archives. 18199 18200 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 18201 distribution of this entire article is permitted in any medium, 18202 provided this notice is preserved. 18203 18204 These pages are [22]maintained by the GCC team. Last modified 18205 2018-09-30[23]. 18206 18207 References 18208 18209 1. http://gcc.gnu.org/news/spill.html 18210 2. http://gcc.gnu.org/news/lcm.html 18211 3. http://gcc.gnu.org/news/cprop.html 18212 4. http://gcc.gnu.org/news/cfg.html 18213 5. http://gcc.gnu.org/news/dse.html 18214 6. http://gcc.gnu.org/news/hoist.html 18215 7. http://gcc.gnu.org/news/alias.html 18216 8. http://gcc.gnu.org/gcc-2.95/c++features.html 18217 9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 18218 10. http://gcc.gnu.org/news/gcj-announce.txt 18219 11. http://gcc.gnu.org/news/javaannounce.html 18220 12. http://gcc.gnu.org/c99status.html 18221 13. http://gcc.gnu.org/news/chill.html 18222 14. http://gcc.gnu.org/news/sparc.html 18223 15. http://gcc.gnu.org/news/egcs-vcg.html 18224 16. http://gcc.gnu.org/egcs-1.0/features-2.8.html 18225 17. https://gcc.gnu.org/onlinedocs/ 18226 18. mailto:gcc-help (a] gcc.gnu.org 18227 19. mailto:gcc (a] gcc.gnu.org 18228 20. https://gcc.gnu.org/lists.html 18229 21. https://www.fsf.org/ 18230 22. https://gcc.gnu.org/about.html 18231 23. http://validator.w3.org/check/referer 18232 ====================================================================== 18233 http://gcc.gnu.org/gcc-2.95/caveats.html 18234 GCC 2.95 Caveats 18235 18236 * GCC 2.95 will issue an error for invalid asm statements that had 18237 been silently accepted by earlier versions of the compiler. This is 18238 particularly noticeable when compiling older versions of the Linux 18239 kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95) 18240 for more information on this issue. 18241 * GCC 2.95 implements type based alias analysis to disambiguate 18242 memory references. Some programs, particularly the Linux kernel 18243 violate ANSI/ISO aliasing rules and therefore may not operate 18244 correctly when compiled with GCC 2.95. Please refer to the FAQ (as 18245 shipped with GCC 2.95) for more information on this issue. 18246 * GCC 2.95 has a known bug in its handling of complex variables for 18247 64bit targets. Instead of silently generating incorrect code, GCC 18248 2.95 will issue a fatal error for situations it can not handle. 18249 This primarily affects the Fortran community as Fortran makes more 18250 use of complex variables than C or C++. 18251 * GCC 2.95 has an integrated libstdc++, but does not have an 18252 integrated libg++. Furthermore old libg++ releases will not work 18253 with GCC 2.95. You can retrieve a recent copy of libg++ from the 18254 [1]GCC ftp server. 18255 Note most C++ programs only need libstdc++. 18256 * Exception handling may not work with shared libraries, particularly 18257 on alphas, hppas, rs6000/powerpc and mips based platforms. 18258 Exception handling is known to work on x86 GNU/Linux platforms with 18259 shared libraries. 18260 * In general, GCC 2.95 is more rigorous about rejecting invalid C++ 18261 code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0, 18262 or EGCS 1.1. As a result it may be necessary to fix C++ code before 18263 it will compile with GCC 2.95. 18264 * G++ is also converting toward the ISO C++ standard; as a result 18265 code which was previously valid (and thus accepted by other 18266 compilers and older versions of g++) may no longer be accepted. The 18267 flag -fpermissive may allow some non-conforming code to compile 18268 with GCC 2.95. 18269 * GCC 2.95 compiled C++ code is not binary compatible with EGCS 18270 1.1.x, EGCS 1.0.x or GCC 2.8.x. 18271 * GCC 2.95 does not have changes from the GCC 2.8 tree that were made 18272 between Sept 30, 1998 and April 30, 1999 (the official end of the 18273 GCC 2.8 project). Future GCC releases will include all the changes 18274 from the defunct GCC 2.8 sources. 18275 18276 18277 For questions related to the use of GCC, please consult these web 18278 pages and the [2]GCC manuals. If that fails, the 18279 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18280 web pages and the development of GCC are welcome on our developer 18281 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 18282 18283 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 18284 distribution of this entire article is permitted in any medium, 18285 provided this notice is preserved. 18286 18287 These pages are [7]maintained by the GCC team. Last modified 18288 2018-09-30[8]. 18289 18290 References 18291 18292 1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz 18293 2. https://gcc.gnu.org/onlinedocs/ 18294 3. mailto:gcc-help (a] gcc.gnu.org 18295 4. mailto:gcc (a] gcc.gnu.org 18296 5. https://gcc.gnu.org/lists.html 18297 6. https://www.fsf.org/ 18298 7. https://gcc.gnu.org/about.html 18299 8. http://validator.w3.org/check/referer 18300 ====================================================================== 18301 http://gcc.gnu.org/egcs-1.1/index.html 18302 EGCS 1.1 18303 18304 September 3, 1998: We are pleased to announce the release of EGCS 1.1. 18305 December 1, 1998: We are pleased to announce the release of EGCS 1.1.1. 18306 March 15, 1999: We are pleased to announce the release of EGCS 1.1.2. 18307 18308 EGCS is a free software project to further the development of the GNU 18309 compilers using an open development environment. 18310 18311 EGCS 1.1 is a major new release of the EGCS compiler system. It has 18312 been [1]extensively tested and is believed to be stable and suitable 18313 for widespread use. 18314 18315 EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8 18316 development sources; it contains all of the new features found in GCC 18317 2.8.1 as well as all new development from GCC up to June 6, 1998. 18318 18319 EGCS 1.1 also contains many improvements and features not found in GCC 18320 or in older versions of EGCS: 18321 * Global common subexpression elimination and global constant/copy 18322 propagation (aka [2]gcse) 18323 * Ongoing improvements to the [3]alias analysis support to allow for 18324 better optimizations throughout the compiler. 18325 * Vastly improved [4]C++ compiler and integrated C++ runtime 18326 libraries. 18327 * Fixes for the /tmp symlink race security problems. 18328 * New targets including mips16, arm-thumb and 64 bit PowerPC. 18329 * Improvements to GNU Fortran (g77) compiler and runtime library made 18330 since g77 version 0.5.23. 18331 18332 See the [5]new features page for a more complete list of new features 18333 found in EGCS 1.1 releases. 18334 18335 EGCS 1.1.1 is a minor update to fix several serious problems in EGCS 18336 1.1: 18337 * General improvements and fixes 18338 + Avoid some stack overflows when compiling large functions. 18339 + Avoid incorrect loop invariant code motions. 18340 + Fix some core dumps on Linux kernel code. 18341 + Bring back the imake -Di386 and friends fix from EGCS 1.0.2. 18342 + Fix code generation problem in gcse. 18343 + Various documentation related fixes. 18344 * g++/libstdc++ improvements and fixes 18345 + MT safe EH fix for setjmp/longjmp based exception handling. 18346 + Fix a few bad interactions between optimization and exception 18347 handling. 18348 + Fixes for demangling of template names starting with "__". 18349 + Fix a bug that would fail to run destructors in some cases 18350 with -O2. 18351 + Fix 'new' of classes with virtual bases. 18352 + Fix crash building Qt on the Alpha. 18353 + Fix failure compiling WIFEXITED macro on GNU/Linux. 18354 + Fix some -frepo failures. 18355 * g77 and libf2c improvements and fixes 18356 + Various documentation fixes. 18357 + Avoid compiler crash on RAND intrinsic. 18358 + Fix minor bugs in makefiles exposed by BSD make programs. 18359 + Define _XOPEN_SOURCE for libI77 build to avoid potential 18360 problems on some 64-bit systems. 18361 + Fix problem with implicit endfile on rewind. 18362 + Fix spurious recursive I/O errors. 18363 * platform specific improvements and fixes 18364 + Match all versions of UnixWare7. 18365 + Do not assume x86 SVR4 or UnixWare targets can handle stabs. 18366 + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion 18367 from unsigned ints to double precision floats. 18368 + Fix ARM ABI issue with NetBSD. 18369 + Fix a few arm code generation bugs. 18370 + Fixincludes will fix additional broken SCO OpenServer header 18371 files. 18372 + Fix a m68k backend bug which caused invalid offsets in reg+d 18373 addresses. 18374 + Fix problems with 64bit AIX 4.3 support. 18375 + Fix handling of long longs for varargs/stdarg functions on the 18376 ppc. 18377 + Minor fixes to CPP predefines for Windows. 18378 + Fix code generation problems with gpr<->fpr copies for 64bit 18379 ppc. 18380 + Fix a few coldfire code generation bugs. 18381 + Fix some more header file problems on SunOS 4.x. 18382 + Fix assert.h handling for RTEMS. 18383 + Fix Windows handling of TREE_SYMBOL_REFERENCED. 18384 + Fix x86 compiler abort in reg-stack pass. 18385 + Fix cygwin/windows problem with section attributes. 18386 + Fix Alpha code generation problem exposed by SMP Linux 18387 kernels. 18388 + Fix typo in m68k 32->64bit integer conversion. 18389 + Make sure target libraries build with -fPIC for PPC & Alpha 18390 targets. 18391 18392 EGCS 1.1.2 is a minor update to fix several serious problems in EGCS 18393 1.1.1: 18394 * General improvements and fixes 18395 + Fix bug in loop optimizer which caused the SPARC (and 18396 potentially other) ports to segfault. 18397 + Fix infinite recursion in alias analysis and combiner code. 18398 + Fix bug in regclass preferencing. 18399 + Fix incorrect loop reversal which caused incorrect code to be 18400 generated for several targets. 18401 + Fix return value for builtin memcpy. 18402 + Reduce compile time for certain loops which exposed quadratic 18403 behavior in the loop optimizer. 18404 + Fix bug which caused volatile memory to be written multiple 18405 times when only one write was needed/desired. 18406 + Fix compiler abort in caller-save.c 18407 + Fix combiner bug which caused incorrect code generation for 18408 certain division by constant operations. 18409 + Fix incorrect code generation due to a bug in range check 18410 optimizations. 18411 + Fix incorrect code generation due to mis-handling of clobbered 18412 values in CSE. 18413 + Fix compiler abort/segfault due to incorrect register 18414 splitting when unrolling loops. 18415 + Fix code generation involving autoincremented addresses with 18416 ternary operators. 18417 + Work around bug in the scheduler which caused qt to be 18418 mis-compiled on some platforms. 18419 + Fix code generation problems with -fshort-enums. 18420 + Tighten security for temporary files. 18421 + Improve compile time for codes which make heavy use of 18422 overloaded functions. 18423 + Fix multiply defined constructor/destructor symbol problems. 18424 + Avoid setting bogus RPATH environment variable during 18425 bootstrap. 18426 + Avoid GNU-make dependencies in the texinfo subdir. 18427 + Install CPP wrapper script in $(prefix)/bin if --enable-cpp. 18428 --enable-cpp=<dirname> can be used to specify an additional 18429 install directory for the cpp wrapper script. 18430 + Fix CSE bug which caused incorrect label-label refs to appear 18431 on some platforms. 18432 + Avoid linking in EH routines from libgcc if they are not 18433 needed. 18434 + Avoid obscure bug in aliasing code. 18435 + Fix bug in weak symbol handling. 18436 * Platform-specific improvements and fixes 18437 + Fix detection of PPro/PII on Unixware 7. 18438 + Fix compiler segfault when building spec99 and other programs 18439 for SPARC targets. 18440 + Fix code-generation bugs for integer and floating point 18441 conditional move instructions on the PPro/PII. 18442 + Use fixincludes to fix byteorder problems on i?86-*-sysv. 18443 + Fix build failure for the arc port. 18444 + Fix floating point format configuration for i?86-gnu port. 18445 + Fix problems with hppa1.0-hp-hpux10.20 configuration when 18446 threads are enabled. 18447 + Fix coldfire code generation bugs. 18448 + Fix "unrecognized insn" problems for Alpha and PPC ports. 18449 + Fix h8/300 code generation problem with floating point values 18450 in memory. 18451 + Fix unrecognized insn problems for the m68k port. 18452 + Fix namespace-pollution problem for the x86 port. 18453 + Fix problems with old assembler on x86 NeXT systems. 18454 + Fix PIC code-generation problems for the SPARC port. 18455 + Fix minor bug with LONG_CALLS in PowerPC SVR4 support. 18456 + Fix minor ISO namespace violation in Alpha varargs/stdarg 18457 support. 18458 + Fix incorrect "braf" instruction usage for the SH port. 18459 + Fix minor bug in va-sh which prevented its use with -ansi. 18460 + Fix problems recognizing and supporting FreeBSD. 18461 + Handle OpenBSD systems correctly. 18462 + Minor fixincludes fix for Digital UNIX 4.0B. 18463 + Fix problems with ctors/dtors in SCO shared libraries. 18464 + Abort instead of generating incorrect code for PPro/PII 18465 floating point conditional moves. 18466 + Avoid multiply defined symbols on GNU/Linux systems using 18467 libc-5.4.xx. 18468 + Fix abort in alpha compiler. 18469 * Fortran-specific fixes 18470 + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year 18471 is in the documented, non-Y2K-compliant range of 0-99, instead 18472 of being returned as 100 in the year 2000. 18473 + Fix the `Date_and_Time' intrinsic (in libg2c) to return the 18474 milliseconds value properly in Values(8). 18475 + Fix the `LStat' intrinsic (in libg2c) to return device-ID 18476 information properly in SArray(7). 18477 18478 Each release includes installation instructions in both HTML and 18479 plaintext forms (see the INSTALL directory in the toplevel directory of 18480 the distribution). However, we also keep the most up to date 18481 installation instructions and [6]build/test status on our web page. We 18482 will update those pages as new information becomes available. 18483 18484 The EGCS project would like to thank the numerous people that have 18485 contributed new features, test results, bugfixes, etc. This [7]amazing 18486 group of volunteers is what makes EGCS successful. 18487 18488 And finally, we can't in good conscience fail to mention some 18489 [8]caveats to using EGCS 1.1. 18490 18491 Download EGCS from egcs.cygnus.com (USA California). 18492 18493 The EGCS 1.1 release is also available on many mirror sites. 18494 [9]Goto mirror list to find a closer site. 18495 18496 18497 For questions related to the use of GCC, please consult these web 18498 pages and the [10]GCC manuals. If that fails, the 18499 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18500 web pages and the development of GCC are welcome on our developer 18501 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 18502 archives. 18503 18504 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 18505 distribution of this entire article is permitted in any medium, 18506 provided this notice is preserved. 18507 18508 These pages are [15]maintained by the GCC team. Last modified 18509 2018-09-30[16]. 18510 18511 References 18512 18513 1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html 18514 2. http://gcc.gnu.org/news/gcse.html 18515 3. http://gcc.gnu.org/news/alias.html 18516 4. http://gcc.gnu.org/egcs-1.1/c++features.html 18517 5. http://gcc.gnu.org/egcs-1.1/features.html 18518 6. http://gcc.gnu.org/egcs-1.1/buildstat.html 18519 7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 18520 8. http://gcc.gnu.org/egcs-1.1/caveats.html 18521 9. http://gcc.gnu.org/mirrors.html 18522 10. https://gcc.gnu.org/onlinedocs/ 18523 11. mailto:gcc-help (a] gcc.gnu.org 18524 12. mailto:gcc (a] gcc.gnu.org 18525 13. https://gcc.gnu.org/lists.html 18526 14. https://www.fsf.org/ 18527 15. https://gcc.gnu.org/about.html 18528 16. http://validator.w3.org/check/referer 18529 ====================================================================== 18530 http://gcc.gnu.org/egcs-1.1/features.html 18531 EGCS 1.1 new features 18532 18533 * Integrated GNU Fortran (g77) compiler and runtime library with 18534 improvements, based on g77 version 0.5.23. 18535 * Vast improvements in the C++ compiler; so many they have [1]page of 18536 their own! 18537 * Compiler implements [2]global common subexpression elimination and 18538 global copy/constant propagation. 18539 * More major improvements in the [3]alias analysis code. 18540 * More major improvements in the exception handling code to improve 18541 performance, lower static overhead and provide the infrastructure 18542 for future improvements. 18543 * The infamous /tmp symlink race security problems have been fixed. 18544 * The regmove optimization pass has been nearly completely rewritten 18545 to improve performance of generated code. 18546 * The compiler now recomputes register usage information before local 18547 register allocation. By providing more accurate information to the 18548 priority based allocator, we get better register allocation. 18549 * The register reloading phase of the compiler optimizes spill code 18550 much better than in previous releases. 18551 * Some bad interactions between the register allocator and 18552 instruction scheduler have been fixed, resulting in much better 18553 code for certain programs. Additionally, we have tuned the 18554 scheduler in various ways to improve performance of generated code 18555 for some architectures. 18556 * The compiler's branch shortening algorithms have been significantly 18557 improved to work better on targets which align jump targets. 18558 * The compiler now supports -Os to prefer optimizing for code space 18559 over optimizing for code speed. 18560 * The compiler will now totally eliminate library calls which compute 18561 constant values. This primarily helps targets with no integer 18562 div/mul support and targets without floating point support. 18563 * The compiler now supports an extensive "--help" option. 18564 * cpplib has been greatly improved and may be suitable for limited 18565 use. 18566 * Memory footprint for the compiler has been significantly reduced 18567 for some pathological cases. 18568 * The time to build EGCS has been improved for certain targets 18569 (particularly the alpha and mips platforms). 18570 * Many infrastructure improvements throughout the compiler, plus the 18571 usual mountain of bugfixes and minor improvements. 18572 * Target dependent improvements: 18573 + SPARC port now includes V8 plus and V9 support as well as 18574 performance tuning for Ultra class machines. The SPARC port 18575 now uses the Haifa scheduler. 18576 + Alpha port has been tuned for the EV6 processor and has an 18577 optimized expansion of memcpy/bzero. The Alpha port now uses 18578 the Haifa scheduler. 18579 + RS6000/PowerPC: support for the Power64 architecture and AIX 18580 4.3. The RS6000/PowerPC port now uses the Haifa scheduler. 18581 + x86: Alignment of static store data and jump targets is per 18582 Intel recommendations now. Various improvements throughout the 18583 x86 port to improve performance on Pentium processors 18584 (including improved epilogue sequences for Pentium chips and 18585 backend improvements which should help register allocation on 18586 all x86 variants. Conditional move support has been fixed and 18587 enabled for PPro processors. The x86 port also better supports 18588 64bit operations now. Unixware 7, a System V Release 5 target, 18589 is now supported and SCO OpenServer targets can support GAS. 18590 + MIPS has improved multiply/multiply-add support and now 18591 includes mips16 ISA support. 18592 + M68k has many micro-optimizations and Coldfire fixes. 18593 * Core compiler is based on the GCC development tree from June 9, 18594 1998, so we have all of the [4]features found in GCC 2.8. 18595 18596 18597 For questions related to the use of GCC, please consult these web 18598 pages and the [5]GCC manuals. If that fails, the 18599 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18600 web pages and the development of GCC are welcome on our developer 18601 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 18602 18603 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 18604 distribution of this entire article is permitted in any medium, 18605 provided this notice is preserved. 18606 18607 These pages are [10]maintained by the GCC team. Last modified 18608 2018-09-30[11]. 18609 18610 References 18611 18612 1. http://gcc.gnu.org/egcs-1.1/c++features.html 18613 2. http://gcc.gnu.org/news/gcse.html 18614 3. http://gcc.gnu.org/news/alias.html 18615 4. http://gcc.gnu.org/egcs-1.0/features-2.8.html 18616 5. https://gcc.gnu.org/onlinedocs/ 18617 6. mailto:gcc-help (a] gcc.gnu.org 18618 7. mailto:gcc (a] gcc.gnu.org 18619 8. https://gcc.gnu.org/lists.html 18620 9. https://www.fsf.org/ 18621 10. https://gcc.gnu.org/about.html 18622 11. http://validator.w3.org/check/referer 18623 ====================================================================== 18624 http://gcc.gnu.org/egcs-1.1/caveats.html 18625 EGCS 1.1 Caveats 18626 18627 * EGCS has an integrated libstdc++, but does not have an integrated 18628 libg++. Furthermore old libg++ releases will not work with EGCS; HJ 18629 Lu has made a libg++-2.8.1.2 snapshot available which may work with 18630 EGCS. 18631 Note most C++ programs only need libstdc++. 18632 * Exception handling may not work with shared libraries, particularly 18633 on alphas, hppas, rs6000/powerpc and mips based platforms. 18634 Exception handling is known to work on x86-linux platforms with 18635 shared libraries. 18636 * Some versions of the Linux kernel have bugs which prevent them from 18637 being compiled or from running when compiled by EGCS. See the FAQ 18638 (as shipped with EGCS 1.1) for additional information. 18639 * In general, EGCS is more rigorous about rejecting invalid C++ code 18640 or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As 18641 a result it may be necessary to fix C++ code before it will compile 18642 with EGCS. 18643 * G++ is also converting toward the ISO C++ standard; as a result 18644 code which was previously valid (and thus accepted by other 18645 compilers and older versions of g++) may no longer be accepted. 18646 * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x 18647 or GCC 2.8.x due to changes necessary to support thread safe 18648 exception handling. 18649 18650 18651 For questions related to the use of GCC, please consult these web 18652 pages and the [1]GCC manuals. If that fails, the 18653 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18654 web pages and the development of GCC are welcome on our developer 18655 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 18656 18657 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 18658 distribution of this entire article is permitted in any medium, 18659 provided this notice is preserved. 18660 18661 These pages are [6]maintained by the GCC team. Last modified 18662 2018-09-30[7]. 18663 18664 References 18665 18666 1. https://gcc.gnu.org/onlinedocs/ 18667 2. mailto:gcc-help (a] gcc.gnu.org 18668 3. mailto:gcc (a] gcc.gnu.org 18669 4. https://gcc.gnu.org/lists.html 18670 5. https://www.fsf.org/ 18671 6. https://gcc.gnu.org/about.html 18672 7. http://validator.w3.org/check/referer 18673 ====================================================================== 18674 http://gcc.gnu.org/egcs-1.0/index.html 18675 EGCS 1.0 18676 18677 December 3, 1997: We are pleased to announce the release of EGCS 1.0. 18678 January 6, 1998: We are pleased to announce the release of EGCS 1.0.1. 18679 March 16, 1998: We are pleased to announce the release of EGCS 1.0.2. 18680 May 15, 1998 We are pleased to announce the release of EGCS 1.0.3. 18681 18682 EGCS is a collaborative effort involving several groups of hackers 18683 using an open development model to accelerate development and testing 18684 of GNU compilers and runtime libraries. 18685 18686 An important goal of EGCS is to allow wide scale testing of 18687 experimental features and optimizations; therefore, EGCS contains some 18688 features and optimizations which are still under development. However, 18689 EGCS has been carefully tested and should be comparable in quality to 18690 most GCC releases. 18691 18692 EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8 18693 development sources; it contains nearly all of the new features found 18694 in GCC 2.8. 18695 18696 EGCS 1.0 also contains many improvements and features not found in GCC 18697 2.7 and even the GCC 2.8 series (which was released after the original 18698 EGCS 1.0 release). 18699 * Integrated C++ runtime libraries, including support for most major 18700 GNU/Linux systems! 18701 * The integrated libstdc++ library includes a verbatim copy of SGI's 18702 STL release. 18703 * Integrated GNU Fortran compiler. 18704 * New instruction scheduler. 18705 * New alias analysis code. 18706 18707 See the [1]new features page for a more complete list of new features. 18708 18709 EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few 18710 critical bugs and add support for Red Hat 5.0 Linux. Changes since the 18711 EGCS 1.0 release: 18712 * Add support for Red Hat 5.0 Linux and better support for Linux 18713 systems using glibc2. 18714 Many programs failed to link when compiled with EGCS 1.0 on Red Hat 18715 5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should 18716 fix these problems. 18717 * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception 18718 handling interfaces. 18719 To avoid future compatibility problems, we strongly urge anyone who 18720 is planning on distributing shared libraries that contain C++ code 18721 to upgrade to EGCS 1.0.1 first. 18722 Soon after EGCS 1.0 was released, the GCC developers made some 18723 incompatible changes in libgcc's exception handling interfaces. 18724 These changes were needed to solve problems on some platforms. This 18725 means that GCC 2.8.0, when released, will not be seamlessly 18726 compatible with shared libraries built by EGCS 1.0. The reason is 18727 that the libgcc.a in GCC 2.8.0 will not contain a function needed 18728 by the old interface. 18729 The result of this is that there may be compatibility problems with 18730 shared libraries built by EGCS 1.0 when used with GCC 2.8.0. 18731 With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface, 18732 and libgcc.a has the support routines for both the old and the new 18733 interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed, 18734 and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed). 18735 The maintainers of GCC 2.x have decided against including seamless 18736 support for the old interface in 2.8.0, since it was never 18737 "official", so to avoid future compatibility problems we recommend 18738 against distributing any shared libraries built by EGCS 1.0 that 18739 contain C++ code (upgrade to 1.0.1 and use that). 18740 * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends. 18741 The x86 changes fix code generation errors exposed when building 18742 glibc2 and the usual GNU/Linux dynamic linker (ld.so). 18743 The hppa change fixes a compiler abort when configured for use with 18744 RTEMS. 18745 The MIPS changes fix problems with the definition of LONG_MAX on 18746 newer systems, allow for command line selection of the target ABI, 18747 and fix one code generation problem. 18748 The rs6000/ppc change fixes some problems with passing structures 18749 to varargs/stdarg functions. 18750 * A few machine independent bugfixes, mostly to fix code generation 18751 errors when building Linux kernels or glibc. 18752 * Fix a few critical exception handling and template bugs in the C++ 18753 compiler. 18754 * Fix Fortran namelist bug on alphas. 18755 * Fix build problems on x86-solaris systems. 18756 18757 EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several 18758 serious problems in EGCS 1.0.1. 18759 * General improvements and fixes 18760 + Memory consumption significantly reduced, especially for 18761 templates and inline functions. 18762 + Fix various problems with glibc2.1. 18763 + Fix loop optimization bug exposed by rs6000/ppc port. 18764 + Fix to avoid potential code generation problems in jump.c. 18765 + Fix some undefined symbol problems in dwarf1 debug support. 18766 * g++/libstdc++ improvements and fixes 18767 + libstdc++ in the EGCS release has been updated and should be 18768 link compatible with libstdc++-2.8. 18769 + Various fixes in libio/libstdc++ to work better on GNU/Linux 18770 systems. 18771 + Fix problems with duplicate symbols on systems that do not 18772 support weak symbols. 18773 + Memory corruption bug and undefined symbols in bastring have 18774 been fixed. 18775 + Various exception handling fixes. 18776 + Fix compiler abort for very long thunk names. 18777 * g77 improvements and fixes 18778 + Fix compiler crash for omitted bound in Fortran CASE 18779 statement. 18780 + Add missing entries to g77 lang-options. 18781 + Fix problem with -fpedantic in the g77 compiler. 18782 + Fix "backspace" problem with g77 on alphas. 18783 + Fix x86 backend problem with Fortran literals and -fpic. 18784 + Fix some of the problems with negative subscripts for g77 on 18785 alphas. 18786 + Fixes for Fortran builds on cygwin32/mingw32. 18787 * platform specific improvements and fixes 18788 + Fix long double problems on x86 (exposed by glibc). 18789 + x86 ports define i386 again to keep imake happy. 18790 + Fix exception handling support on NetBSD ports. 18791 + Several changes to collect2 to fix many problems with AIX. 18792 + Define __ELF__ for GNU/Linux on rs6000. 18793 + Fix -mcall-linux problem on GNU/Linux on rs6000. 18794 + Fix stdarg/vararg problem for GNU/Linux on rs6000. 18795 + Allow autoconf to select a proper install problem on AIX 3.1. 18796 + m68k port support includes -mcpu32 option as well as cpu32 18797 multilibs. 18798 + Fix stdarg bug for irix6. 18799 + Allow EGCS to build on irix5 without the gnu assembler. 18800 + Fix problem with static linking on sco5. 18801 + Fix bootstrap on sco5 with native compiler. 18802 + Fix for abort building newlib on H8 target. 18803 + Fix fixincludes handling of math.h on SunOS. 18804 + Minor fix for Motorola 3300 m68k systems. 18805 18806 EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few 18807 problems reported by Red Hat for builds of Red Hat 5.1. 18808 * Generic bugfixes: 18809 + Fix a typo in the libio library which resulted in incorrect 18810 behavior of istream::get. 18811 + Fix the Fortran negative array index problem. 18812 + Fix a major problem with the ObjC runtime thread support 18813 exposed by glibc2. 18814 + Reduce memory consumption of the Haifa scheduler. 18815 * Target specific bugfixes: 18816 + Fix one x86 floating point code generation bug exposed by 18817 glibc2 builds. 18818 + Fix one x86 internal compiler error exposed by glibc2 builds. 18819 + Fix profiling bugs on the Alpha. 18820 + Fix ImageMagick & emacs 20.2 build problems on the Alpha. 18821 + Fix rs6000/ppc bug when converting values from integer types 18822 to floating point types. 18823 18824 The EGCS 1.0 releases include installation instructions in both HTML 18825 and plaintext forms (see the INSTALL directory in the toplevel 18826 directory of the distribution). However, we also keep the most up to 18827 date installation instructions and [2]build/test status on our web 18828 page. We will update those pages as new information becomes available. 18829 18830 And, we can't in good conscience fail to mention some [3]caveats to 18831 using EGCS. 18832 18833 Update: Big thanks to Stanford for providing a high speed link for 18834 downloading EGCS (go.cygnus.com)! 18835 18836 Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com 18837 (USA California -- High speed link provided by Stanford). 18838 18839 The EGCS 1.0 release is also available many mirror sites. 18840 [4]Goto mirror list to find a closer site 18841 18842 We'd like to thank the numerous people that have contributed new 18843 features, test results, bugfixes, etc. Unfortunately, they're far too 18844 numerous to mention by name. 18845 18846 18847 For questions related to the use of GCC, please consult these web 18848 pages and the [5]GCC manuals. If that fails, the 18849 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18850 web pages and the development of GCC are welcome on our developer 18851 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 18852 18853 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 18854 distribution of this entire article is permitted in any medium, 18855 provided this notice is preserved. 18856 18857 These pages are [10]maintained by the GCC team. Last modified 18858 2018-09-30[11]. 18859 18860 References 18861 18862 1. http://gcc.gnu.org/egcs-1.0/features.html 18863 2. http://gcc.gnu.org/egcs-1.0/buildstat.html 18864 3. http://gcc.gnu.org/egcs-1.0/caveats.html 18865 4. http://gcc.gnu.org/mirrors.html 18866 5. https://gcc.gnu.org/onlinedocs/ 18867 6. mailto:gcc-help (a] gcc.gnu.org 18868 7. mailto:gcc (a] gcc.gnu.org 18869 8. https://gcc.gnu.org/lists.html 18870 9. https://www.fsf.org/ 18871 10. https://gcc.gnu.org/about.html 18872 11. http://validator.w3.org/check/referer 18873 ====================================================================== 18874 http://gcc.gnu.org/egcs-1.0/features.html 18875 EGCS 1.0 features 18876 18877 * Core compiler is based on the gcc2 development tree from Aug 2, 18878 1997, so we have most of the [1]features found in GCC 2.8. 18879 * Integrated GNU Fortran compiler based on g77-0.5.22-19970929. 18880 * Vast improvements in the C++ compiler; so many they have [2]page of 18881 their own! 18882 * Integrated C++ runtime libraries, including support for most major 18883 GNU/Linux systems! 18884 * New instruction scheduler from IBM Haifa which includes support for 18885 function wide instruction scheduling as well as superscalar 18886 scheduling. 18887 * Significantly improved alias analysis code. 18888 * Improved register allocation for two address machines. 18889 * Significant code generation improvements for Fortran code on 18890 Alphas. 18891 * Various optimizations from the g77 project as well as improved loop 18892 optimizations. 18893 * Dwarf2 debug format support for some targets. 18894 * egcs libstdc++ includes the SGI STL implementation without changes. 18895 * As a result of these and other changes, egcs libstc++ is not binary 18896 compatible with previous releases of libstdc++. 18897 * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO 18898 Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and 18899 1.1), Support for RTEMS on several embedded targets, Support for 18900 arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and 18901 MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc. 18902 * Integrated testsuites for gcc, g++, g77, libstdc++ and libio. 18903 * RS6000/PowerPC ports generate code which can run on all 18904 RS6000/PowerPC variants by default. 18905 * -mcpu= and -march= switches for the x86 port to allow better 18906 control over how the x86 port generates code. 18907 * Includes the template repository patch (aka repo patch); note the 18908 new template code makes repo obsolete for ELF systems using gnu-ld 18909 such as GNU/Linux. 18910 * Plus the usual assortment of bugfixes and improvements. 18911 18912 18913 For questions related to the use of GCC, please consult these web 18914 pages and the [3]GCC manuals. If that fails, the 18915 [4]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18916 web pages and the development of GCC are welcome on our developer 18917 list at [5]gcc (a] gcc.gnu.org. All of [6]our lists have public archives. 18918 18919 Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and 18920 distribution of this entire article is permitted in any medium, 18921 provided this notice is preserved. 18922 18923 These pages are [8]maintained by the GCC team. Last modified 18924 2018-09-30[9]. 18925 18926 References 18927 18928 1. http://gcc.gnu.org/egcs-1.0/features-2.8.html 18929 2. http://gcc.gnu.org/egcs-1.0/c++features.html 18930 3. https://gcc.gnu.org/onlinedocs/ 18931 4. mailto:gcc-help (a] gcc.gnu.org 18932 5. mailto:gcc (a] gcc.gnu.org 18933 6. https://gcc.gnu.org/lists.html 18934 7. https://www.fsf.org/ 18935 8. https://gcc.gnu.org/about.html 18936 9. http://validator.w3.org/check/referer 18937 ====================================================================== 18938 http://gcc.gnu.org/egcs-1.0/caveats.html 18939 EGCS 1.0 Caveats 18940 18941 * EGCS has an integrated libstdc++, but does not have an integrated 18942 libg++. Furthermore old libg++ releases will not work with egc; HJ 18943 Lu has made a libg++-2.8.1.2 available which may work with EGCS. 18944 Note most C++ programs only need libstdc++. 18945 * Note that using -pedantic or -Wreturn-type can cause an explosion 18946 in the amount of memory needed for template-heavy C++ code, such as 18947 code that uses STL. Also note that -Wall includes -Wreturn-type, so 18948 if you use -Wall you will need to specify -Wno-return-type to turn 18949 it off. 18950 * Exception handling may not work with shared libraries, particularly 18951 on alphas, hppas, and mips based platforms. Exception handling is 18952 known to work on x86-linux platforms with shared libraries. 18953 * Some versions of the Linux kernel have bugs which prevent them from 18954 being compiled or from running when compiled by EGCS. See the FAQ 18955 (as shipped with EGCS 1.0) for additional information. 18956 * In general, EGCS is more rigorous about rejecting invalid C++ code 18957 or deprecated C++ constructs than G++ 2.7. As a result it may be 18958 necessary to fix C++ code before it will compile with EGCS. 18959 * G++ is also aggressively tracking the C++ standard; as a result 18960 code which was previously valid (and thus accepted by other 18961 compilers and older versions of G++) may no longer be accepted. 18962 * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS 18963 1.0.x and later releases should work with Red Hat Linux 5.0. 18964 18965 18966 For questions related to the use of GCC, please consult these web 18967 pages and the [1]GCC manuals. If that fails, the 18968 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 18969 web pages and the development of GCC are welcome on our developer 18970 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 18971 18972 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 18973 distribution of this entire article is permitted in any medium, 18974 provided this notice is preserved. 18975 18976 These pages are [6]maintained by the GCC team. Last modified 18977 2018-09-30[7]. 18978 18979 References 18980 18981 1. https://gcc.gnu.org/onlinedocs/ 18982 2. mailto:gcc-help (a] gcc.gnu.org 18983 3. mailto:gcc (a] gcc.gnu.org 18984 4. https://gcc.gnu.org/lists.html 18985 5. https://www.fsf.org/ 18986 6. https://gcc.gnu.org/about.html 18987 7. http://validator.w3.org/check/referer 18988 ====================================================================== 18989
Indexes created Tue Oct 27 00:25:24 UTC 2026