NEWS revision 1.10
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-7/index.html 9 10 GCC 7 Release Series 11 12 Dec 6, 2018 13 14 The [1]GNU project and the GCC developers are pleased to announce the 15 release of GCC 7.4. 16 17 This release is a bug-fix release, containing fixes for regressions in 18 GCC 7.3 relative to previous releases of GCC. 19 20 Release History 21 22 GCC 7.4 23 Dec 6, 2018 ([2]changes, [3]documentation) 24 25 GCC 7.3 26 Jan 25, 2018 ([4]changes, [5]documentation) 27 28 GCC 7.2 29 Aug 14, 2017 ([6]changes, [7]documentation) 30 31 GCC 7.1 32 May 2, 2017 ([8]changes, [9]documentation) 33 34 References and Acknowledgements 35 36 GCC used to stand for the GNU C Compiler, but since the compiler 37 supports several other languages aside from C, it now stands for the 38 GNU Compiler Collection. 39 40 A list of [10]successful builds is updated as new information becomes 41 available. 42 43 The GCC developers would like to thank the numerous people that have 44 contributed new features, improvements, bug fixes, and other changes as 45 well as test results to GCC. This [11]amazing group of volunteers is 46 what makes GCC successful. 47 48 For additional information about GCC please refer to the [12]GCC 49 project web site or contact the [13]GCC development mailing list. 50 51 To obtain GCC please use [14]our mirror sites or [15]our SVN server. 52 53 54 For questions related to the use of GCC, please consult these web 55 pages and the [16]GCC manuals. If that fails, the 56 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 57 web pages and the development of GCC are welcome on our developer 58 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 59 archives. 60 61 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 62 distribution of this entire article is permitted in any medium, 63 provided this notice is preserved. 64 65 These pages are [21]maintained by the GCC team. Last modified 66 2018-12-06[22]. 67 68 References 69 70 1. http://www.gnu.org/ 71 2. http://gcc.gnu.org/gcc-7/changes.html 72 3. http://gcc.gnu.org/onlinedocs/7.4.0/ 73 4. http://gcc.gnu.org/gcc-7/changes.html 74 5. http://gcc.gnu.org/onlinedocs/7.3.0/ 75 6. http://gcc.gnu.org/gcc-7/changes.html 76 7. http://gcc.gnu.org/onlinedocs/7.2.0/ 77 8. http://gcc.gnu.org/gcc-7/changes.html 78 9. http://gcc.gnu.org/onlinedocs/7.1.0/ 79 10. http://gcc.gnu.org/gcc-7/buildstat.html 80 11. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 81 12. http://gcc.gnu.org/index.html 82 13. mailto:gcc (a] gcc.gnu.org 83 14. http://gcc.gnu.org/mirrors.html 84 15. http://gcc.gnu.org/svn.html 85 16. https://gcc.gnu.org/onlinedocs/ 86 17. mailto:gcc-help (a] gcc.gnu.org 87 18. mailto:gcc (a] gcc.gnu.org 88 19. https://gcc.gnu.org/lists.html 89 20. https://www.fsf.org/ 90 21. https://gcc.gnu.org/about.html 91 22. http://validator.w3.org/check/referer 92 ====================================================================== 93 http://gcc.gnu.org/gcc-7/changes.html 94 95 GCC 7 Release Series 96 Changes, New Features, and Fixes 97 98 This page is a brief summary of some of the huge number of improvements 99 in GCC 7. For more information, see the [1]Porting to GCC 7 page and 100 the [2]full GCC documentation. 101 102 Caveats 103 104 * GCC now uses [3]LRA (a new local register allocator) by default for 105 new targets. 106 * The non-standard C++0x type traits has_trivial_default_constructor, 107 has_trivial_copy_constructor and has_trivial_copy_assign have been 108 removed. 109 * The libstdc++ [4]Profile Mode has been deprecated and will be 110 removed in a future version. 111 * The Cilk+ extensions to the C and C++ languages have been 112 deprecated. 113 * On ARM targets (arm*-*-*), [5]a bug introduced in GCC 5 that 114 affects conformance to the procedure call standard (AAPCS) has been 115 fixed. The bug affects some C++ code where class objects are passed 116 by value to functions and could result in incorrect or inconsistent 117 code being generated. This is an ABI change. If the option -Wpsabi 118 is enabled (on by default) the compiler will emit a diagnostic note 119 for code that might be affected. 120 121 General Optimizer Improvements 122 123 * GCC 7 can determine the return value or range of return values of 124 some calls to the sprintf family of functions and make it available 125 to other optimization passes. Some calls to the snprintf function 126 with a zero size argument can be folded into constants. This 127 optimization is included in -O1 and can be selectively controlled 128 by the -fprintf-return-value option. 129 * A new store merging pass has been added. It merges constant stores 130 to adjacent memory locations into fewer, wider, stores. It is 131 enabled by the -fstore-merging option and at the -O2 optimization 132 level or higher (and -Os). 133 * A new code hoisting optimization has been added to the partial 134 redundancy elimination pass. It attempts to move evaluation of 135 expressions executed on all paths to the function exit as early as 136 possible. This primarily helps improve code size, but can improve 137 the speed of the generated code as well. It is enabled by the 138 -fcode-hoisting option and at the -O2 optimization level or higher 139 (and -Os). 140 * A new interprocedural bitwise constant propagation optimization has 141 been added, which propagates knowledge about which bits of 142 variables are known to be zero (including pointer alignment 143 information) across the call graph. It is enabled by the 144 -fipa-bit-cp option if -fipa-cp is enabled as well, and is enabled 145 at the -O2 optimization level and higher (and -Os). This 146 optimization supersedes interprocedural alignment propagation of 147 GCC 6, and therefore the option -fipa-cp-alignment is now 148 deprecated and ignored. 149 * A new interprocedural value range propagation optimization has been 150 added, which propagates integral range information across the call 151 graph when variable values can be proven to be within those ranges. 152 It is enabled by the -fipa-vrp option and at the -O2 optimization 153 level and higher (and -Os). 154 * A new loop splitting optimization pass has been added. Certain 155 loops which contain a condition that is always true on one side of 156 the iteration space and always false on the other are split into 157 two loops, such that each of the two new loops iterates on just one 158 side of the iteration space and the condition does not need to be 159 checked inside of the loop. It is enabled by the -fsplit-loops 160 option and at the -O3 optimization level or higher. 161 * The shrink-wrapping optimization can now separate portions of 162 prologues and epilogues to improve performance if some of the work 163 done traditionally by prologues and epilogues is not needed on 164 certain paths. This is controlled by the -fshrink-wrap-separate 165 option, enabled by default. It requires target support, which is 166 currently only implemented in the PowerPC and AArch64 ports. 167 * AddressSanitizer gained a new sanitization option, 168 -fsanitize-address-use-after-scope, which enables sanitization of 169 variables whose address is taken and used after a scope where the 170 variable is defined: 171 172 int 173 main (int argc, char **argv) 174 { 175 char *ptr; 176 { 177 char my_char; 178 ptr = &my_char; 179 } 180 181 *ptr = 123; 182 return *ptr; 183 } 184 185 ==28882==ERROR: AddressSanitizer: stack-use-after-scope on address 0x7fffb8dba99 186 0 at pc 0x0000004006d5 bp 0x7fffb8dba960 sp 0x7fffb8dba958 187 WRITE of size 1 at 0x7fffb8dba990 thread T0 188 #0 0x4006d4 in main /tmp/use-after-scope-1.c:10 189 #1 0x7f9c71943290 in __libc_start_main (/lib64/libc.so.6+0x20290) 190 #2 0x400739 in _start (/tmp/a.out+0x400739) 191 192 Address 0x7fffb8dba990 is located in stack of thread T0 at offset 32 in frame 193 #0 0x40067f in main /tmp/use-after-scope-1.c:3 194 195 This frame has 1 object(s): 196 [32, 33) 'my_char' <== Memory access at offset 32 is inside this variable 197 198 The option is enabled by default with -fsanitize=address and 199 disabled by default with -fsanitize=kernel-address. Compared to the 200 LLVM compiler, where the option already exists, the implementation 201 in the GCC compiler has some improvements and advantages: 202 + Complex uses of gotos and case labels are properly handled and 203 should not report any false positive or false negatives. 204 + C++ temporaries are sanitized. 205 + Sanitization can handle invalid memory stores that are 206 optimized out by the LLVM compiler when optimization is 207 enabled. 208 * The -fsanitize=signed-integer-overflow suboption of the 209 UndefinedBehavior Sanitizer now diagnoses arithmetic overflows even 210 on arithmetic operations with generic vectors. 211 * Version 5 of the [6]DWARF debugging information standard is 212 supported through the -gdwarf-5 option. The DWARF version 4 213 debugging information remains the default until consumers of 214 debugging information are adjusted. 215 216 New Languages and Language specific improvements 217 218 OpenACC support in C, C++, and Fortran continues to be maintained and 219 improved. See the [7]OpenACC and [8]Offloading wiki pages for further 220 information. 221 222 Ada 223 224 * On mainstream native platforms, Ada programs no longer require the 225 stack to be made executable in order to run properly. 226 227 BRIG (HSAIL) 228 229 Support for processing BRIG 1.0 files was added in this release. BRIG 230 is a binary format for HSAIL (Heterogeneous System Architecture 231 Intermediate Language). The BRIG front end can be used for implementing 232 HSAIL "finalizers" (compilation of HSAIL to a native ISA) for 233 GCC-supported targets. An implementation of an HSAIL runtime library, 234 libhsail-rt is also included. 235 236 C family 237 238 * New command-line options have been added for the C and C++ 239 compilers: 240 + -Wimplicit-fallthrough warns when a switch case falls through. 241 This warning has five different levels. The compiler is able 242 to parse a wide range of fallthrough comments, depending on 243 the level. It also handles control-flow statements, such as 244 ifs. It's possible to suppress the warning by either adding a 245 fallthrough comment, or by using a null statement: 246 __attribute__ ((fallthrough)); (C, C++), or [[fallthrough]]; 247 (C++17), or [[gnu::fallthrough]]; (C++11/C++14). This warning 248 is enabled by -Wextra. 249 + -Wpointer-compare warns when a pointer is compared with a zero 250 character constant. Such code is now invalid in C++11 and GCC 251 rejects it. This warning is enabled by default. 252 + -Wduplicated-branches warns when an if-else has identical 253 branches. 254 + -Wrestrict warns when an argument passed to a 255 restrict-qualified parameter aliases with another argument. 256 + -Wmemset-elt-size warns for memset calls, when the first 257 argument references an array, and the third argument is a 258 number equal to the number of elements of the array, but not 259 the size of the array. This warning is enabled by -Wall. 260 + -Wint-in-bool-context warns about suspicious uses of integer 261 values where boolean values are expected. This warning is 262 enabled by -Wall. 263 + -Wswitch-unreachable warns when a switch statement has 264 statements between the controlling expression and the first 265 case label which will never be executed. This warning is 266 enabled by default. 267 + -Wexpansion-to-defined warns when defined is used outside #if. 268 This warning is enabled by -Wextra or -Wpedantic. 269 + -Wregister warns about uses of the register storage specifier. 270 In C++17 this keyword has been removed and for C++17 this is a 271 pedantic warning enabled by default. The warning is not 272 emitted for the GNU Explicit Register Variables extension. 273 + -Wvla-larger-than=N warns about unbounded uses of 274 variable-length arrays, and about bounded uses of 275 variable-length arrays whose bound can be larger than N bytes. 276 + -Wduplicate-decl-specifier warns when a declaration has 277 duplicate const, volatile, restrict or _Atomic specifier. This 278 warning is enabled by -Wall. 279 * GCC 6's C and C++ front ends were able to offer suggestions for 280 misspelled field names: 281 282 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 283 you mean 'color'? 284 return ptr->colour; 285 ^~~~~~ 286 287 GCC 7 greatly expands the scope of these suggestions. Firstly, it 288 adds fix-it hints to such suggestions: 289 290 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 291 you mean 'color'? 292 return ptr->colour; 293 ^~~~~~ 294 color 295 296 The suggestions now cover many other things, such as misspelled 297 function names: 298 299 spellcheck-identifiers.c:11:3: warning: implicit declaration of function 'gtk_wi 300 dget_showall'; did you mean 'gtk_widget_show_all'? [-Wimplicit-function-declarat 301 ion] 302 gtk_widget_showall (w); 303 ^~~~~~~~~~~~~~~~~~ 304 gtk_widget_show_all 305 306 misspelled macro names and enum values: 307 308 spellcheck-identifiers.cc:85:11: error: 'MAX_ITEM' undeclared here (not in a fun 309 ction); did you mean 'MAX_ITEMS'? 310 int array[MAX_ITEM]; 311 ^~~~~~~~ 312 MAX_ITEMS 313 314 misspelled type names: 315 316 spellcheck-typenames.c:7:14: error: unknown type name 'singed'; did you mean 'si 317 gned'? 318 void test (singed char e); 319 ^~~~~~ 320 signed 321 322 and, in the C front end, named initializers: 323 324 test.c:7:20: error: 'struct s' has no member named 'colour'; did you mean 'color 325 '? 326 struct s test = { .colour = 3 }; 327 ^~~~~~ 328 color 329 330 * The preprocessor can now offer suggestions for misspelled 331 directives, e.g.: 332 333 test.c:5:2: error:invalid preprocessing directive #endfi; did you mean #endif? 334 #endfi 335 ^~~~~ 336 endif 337 338 * Warnings about format strings now underline the pertinent part of 339 the string, and can offer suggested fixes. In some cases, the 340 pertinent argument is underlined. 341 342 test.c:51:29: warning: format '%s' expects argument of type 'char *', but argume 343 nt 3 has type 'int' [-Wformat=] 344 printf ("foo: %d bar: %s baz: %d", 100, i + j, 102); 345 ~^ ~~~~~ 346 %d 347 348 * The new -Wdangling-else command-line option has been split out of 349 -Wparentheses and warns about dangling else. 350 * The -Wshadow warning has been split into three variants. 351 -Wshadow=global warns for any shadowing. This is the default when 352 using -Wshadow without any argument. -Wshadow=local only warns for 353 a local variable shadowing another local variable or parameter. 354 -Wshadow=compatible-local only warns for a local variable shadowing 355 another local variable or parameter whose type is compatible (in 356 C++ compatible means that the type of the shadowing variable can be 357 converted to that of the shadowed variable). 358 The following example shows the different kinds of shadow warnings: 359 360 enum operation { add, count }; 361 struct container { int nr; }; 362 363 int 364 container_count (struct container c, int count) 365 { 366 int r = 0; 367 for (int count = 0; count > 0; count--) 368 { 369 struct container count = c; 370 r += count.nr; 371 } 372 return r; 373 } 374 375 -Wshadow=compatible-local will warn for the parameter being 376 shadowed with the same type: 377 378 warn-test.c:8:12: warning: declaration of 'count' shadows a parameter [-Wshadow= 379 compatible-local] 380 for (int count = 0; count > 0; count--) 381 ^~~~~ 382 warn-test.c:5:42: note: shadowed declaration is here 383 container_count (struct container c, int count) 384 ^~~~~ 385 386 -Wshadow=local will warn for the above and for the shadowed 387 declaration with incompatible type: 388 389 warn-test.c:10:24: warning: declaration of 'count' shadows a previous local [-Ws 390 hadow=local] 391 struct container count = c; 392 ^~~~~ 393 warn-test.c:8:12: note: shadowed declaration is here 394 for (int count = 0; count > 0; count--) 395 ^~~~~ 396 397 -Wshadow=global will warn for all of the above and the shadowing of 398 the global declaration: 399 400 warn-test.c:5:42: warning: declaration of 'count' shadows a global declaration [ 401 -Wshadow] 402 container_count (struct container c, int count) 403 ^~~~~ 404 warn-test.c:1:23: note: shadowed declaration is here 405 enum operation { add, count }; 406 ^~~~~ 407 408 * GCC 7 contains a number of enhancements that help detect buffer 409 overflow and other forms of invalid memory accesses. 410 + The -Walloc-size-larger-than=size option detects calls to 411 standard and user-defined memory allocation functions 412 decorated with attribute alloc_size whose argument exceeds the 413 specified size (PTRDIFF_MAX by default). The option also 414 detects arithmetic overflow in the computation of the size in 415 two-argument allocation functions like calloc where the total 416 size is the product of the two arguments. Since calls with an 417 excessive size cannot succeed they are typically the result of 418 programming errors. Such bugs have been known to be the source 419 of security vulnerabilities and a target of exploits. 420 -Walloc-size-larger-than=PTRDIFF_MAX is included in -Wall. 421 For example, the following call to malloc incorrectly tries to 422 avoid passing a negative argument to the function and instead 423 ends up unconditionally invoking it with an argument less than 424 or equal to zero. Since after conversion to the type of the 425 argument of the function (size_t) a negative argument results 426 in a value in excess of the maximum PTRDIFF_MAX the call is 427 diagnosed. 428 429 void* f (int n) 430 { 431 return malloc (n > 0 ? 0 : n); 432 } 433 434 warning: argument 1 range [2147483648, 4294967295] exceeds maximum object size 2 435 147483647 [-Walloc-size-larger-than=] 436 437 + The -Walloc-zero option detects calls to standard and 438 user-defined memory allocation functions decorated with 439 attribute alloc_size with a zero argument. -Walloc-zero is not 440 included in either -Wall or -Wextra and must be explicitly 441 enabled. 442 + The -Walloca option detects all calls to the alloca function 443 in the program. -Walloca is not included in either -Wall or 444 -Wextra and must be explicitly enabled. 445 + The -Walloca-larger-than=size option detects calls to the 446 alloca function whose argument either may exceed the specified 447 size, or that is not known to be sufficiently constrained to 448 avoid exceeding it. -Walloca-larger-than is not included in 449 either -Wall or -Wextra and must be explicitly enabled. 450 For example, compiling the following snippet with 451 -Walloca-larger-than=1024 results in a warning because even 452 though the code appears to call alloca only with sizes of 1kb 453 and less, since n is signed, a negative value would result in 454 a call to the function well in excess of the limit. 455 456 void f (int n) 457 { 458 char *d; 459 if (n < 1025) 460 d = alloca (n); 461 else 462 d = malloc (n); 463 ... 464 } 465 466 warning: argument to 'alloca may be too large due to conversion from 'int' to 'l 467 ong unsigned int' [-Walloca-larger-than=] 468 469 In contrast, a call to alloca that isn't bounded at all such 470 as in the following function will elicit the warning below 471 regardless of the size argument to the option. 472 473 void f (size_t n) 474 { 475 char *d = alloca (n); 476 ... 477 } 478 479 warning: unbounded use of 'alloca' [-Walloca-larger-than=] 480 481 + The -Wformat-overflow=level option detects certain and likely 482 buffer overflow in calls to the sprintf family of formatted 483 output functions. Although the option is enabled even without 484 optimization it works best with -O2 and higher. 485 For example, in the following snippet the call to sprintf is 486 diagnosed because even though its output has been constrained 487 using the modulo operation it could result in as many as three 488 bytes if mday were negative. The solution is to either 489 allocate a larger buffer or make sure the argument is not 490 negative, for example by changing mday's type to unsigned or 491 by making the type of the second operand of the modulo 492 expression unsigned: 100U. 493 494 void* f (int mday) 495 { 496 char *buf = malloc (3); 497 sprintf (buf, "%02i", mday % 100); 498 return buf; 499 } 500 501 warning: 'sprintf may write a terminating nul past the end of the destination [- 502 Wformat-overflow=] 503 note: 'sprintf' output between 3 and 4 bytes into a destination of size 3 504 505 + The -Wformat-truncation=level option detects certain and 506 likely output truncation in calls to the snprintf family of 507 formatted output functions. -Wformat-truncation=1 is included 508 in -Wall and enabled without optimization but works best with 509 -O2 and higher. 510 For example, the following function attempts to format an 511 integer between 0 and 255 in hexadecimal, including the 0x 512 prefix, into a buffer of four characters. But since the 513 function must always terminate output by the null character 514 ('\0') such a buffer is only big enough to fit just one digit 515 plus the prefix. Therefore the snprintf call is diagnosed. To 516 avoid the warning either use a bigger buffer or handle the 517 function's return value which indicates whether or not its 518 output has been truncated. 519 520 void f (unsigned x) 521 { 522 char d[4]; 523 snprintf (d, sizeof d, "%#02x", x & 0xff); 524 ... 525 } 526 527 warning: 'snprintf' output may be truncated before the last format character [-W 528 format-truncation=] 529 note: 'snprintf' output between 3 and 5 bytes into a destination of size 4 530 531 + The -Wnonnull option has been enhanced to detect a broader set 532 of cases of passing null pointers to functions that expect a 533 non-null argument (those decorated with attribute nonnull). By 534 taking advantage of optimizations the option can detect many 535 more cases of the problem than in prior GCC versions. 536 + The -Wstringop-overflow=type option detects buffer overflow in 537 calls to string handling functions like memcpy and strcpy. The 538 option relies on [9]Object Size Checking and has an effect 539 similar to defining the _FORTIFY_SOURCE macro. 540 -Wstringop-overflow=2 is enabled by default. 541 For example, in the following snippet, because the call to 542 strncat specifies a maximum that allows the function to write 543 past the end of the destination, it is diagnosed. To correct 544 the problem and avoid the overflow the function should be 545 called with a size of at most sizeof d - strlen(d) - 1. 546 547 void f (const char *fname) 548 { 549 char d[8]; 550 strncpy (d, "/tmp/", sizeof d); 551 strncat (d, fname, sizeof d); 552 ... 553 } 554 555 warning: specified bound 8 equals the size of the destination [-Wstringop-overfl 556 ow=] 557 558 * The <limits.h> header provided by GCC defines macros such as 559 INT_WIDTH for the width in bits of integer types, if 560 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 561 included. The <stdint.h> header defines such macros as SIZE_WIDTH 562 and INTMAX_WIDTH for the width of some standard typedef names for 563 integer types, again if __STDC_WANT_IEC_60559_BFP_EXT__ is defined 564 before the header is included; note that GCC's implementation of 565 this header is only used for freestanding compilations, not hosted 566 compilations, on most systems. These macros come from ISO/IEC TS 567 18661-1:2014. 568 * The <float.h> header provided by GCC defines the macro 569 CR_DECIMAL_DIG, from ISO/IEC TS 18661-1:2014, if 570 __STDC_WANT_IEC_60559_BFP_EXT__ is defined before the header is 571 included. This represents the number of decimal digits for which 572 conversions between decimal character strings and binary formats, 573 in both directions, are correctly rounded, and currently has the 574 value of UINTMAX_MAX on all systems, reflecting that GCC's 575 compile-time conversions are correctly rounded for any number of 576 digits. 577 * New __builtin_add_overflow_p, __builtin_sub_overflow_p, 578 __builtin_mul_overflow_p built-in functions have been added. These 579 work similarly to their siblings without the _p suffix, but do not 580 actually store the result of the arithmetics anywhere, just return 581 whether the operation would overflow. Calls to these built-ins with 582 integer constant arguments evaluate to integer constants 583 expressions. 584 For example, in the following, c is assigned the result of a * b 585 only if the multiplication does not overflow, otherwise it is 586 assigned the value zero. The multiplication is performed at 587 compile-time and without triggering a -Woverflow warning. 588 589 enum { 590 a = 12345678, 591 b = 87654321, 592 c = __builtin_mul_overflow_p (a, b, a) ? 0 : a * b 593 }; 594 595 C 596 597 * The C front end now supports type names _FloatN for floating-point 598 types with IEEE interchange formats and _FloatNx for floating-point 599 types with IEEE extended formats. These type names come from 600 ISO/IEC TS 18661-3:2015. 601 The set of types supported depends on the target for which GCC is 602 configured. Most targets support _Float32, _Float32x and _Float64. 603 _Float128 is supported on targets where IEEE binary128 encoding was 604 already supported as long double or __float128. _Float64x is 605 supported on targets where a type with either binary128 or Intel 606 extended precision format is available. 607 Constants with these types are supported using suffixes fN, FN, fNx 608 and FNx (e.g., 1.2f128 or 2.3F64x). Macros such as FLT128_MAX are 609 defined in <float.h> if __STDC_WANT_IEC_60559_TYPES_EXT__ is 610 defined before it is included. 611 These new types are always distinct from each other and from float, 612 double and long double, even if they have the same encoding. 613 Complex types such as _Complex _Float128 are also supported. 614 Type-generic built-in functions such as __builtin_isinf support the 615 new types, and the following type-specific built-in functions have 616 versions (suffixed fN or fNx) for the new types: 617 __builtin_copysign, __builtin_fabs, __builtin_huge_val, 618 __builtin_inf, __builtin_nan, __builtin_nans. 619 * Compilation with -fopenmp is now compatible with the C11 _Atomic 620 keyword. 621 622 C++ 623 624 * The C++ front end has experimental support for all of the current 625 C++17 draft with the -std=c++1z or -std=gnu++1z flags, including if 626 constexpr, class template argument deduction, auto template 627 parameters, and structured bindings. For a full list of new 628 features, see [10]the C++ status page. 629 * C++17 support for new of over-aligned types can be enabled in other 630 modes with the -faligned-new flag. 631 * The C++17 evaluation order requirements can be selected in other 632 modes with the -fstrong-eval-order flag, or disabled in C++17 mode 633 with -fno-strong-eval-order. 634 * The default semantics of inherited constructors has changed in all 635 modes, following [11]P0136. Essentially, overload resolution 636 happens as if calling the inherited constructor directly, and the 637 compiler fills in construction of the other bases and members as 638 needed. Most uses should not need any changes. The old behavior can 639 be restored with -fno-new-inheriting-ctors, or -fabi-version less 640 than 11. 641 * The resolution of DR 150 on matching of template template 642 parameters, allowing default template arguments to make a template 643 match a parameter, is currently enabled by default in C++17 mode 644 only. The default can be overridden with -f{no-,}new-ttp-matching. 645 * The C++ front end will now provide fix-it hints for some missing 646 semicolons, allowing for automatic fixes by IDEs: 647 648 test.cc:4:11: error: expected ';' after class definition 649 class a {} 650 ^ 651 ; 652 653 * -Waligned-new has been added to the C++ front end. It warns about 654 new of type with extended alignment without -faligned-new. 655 656 Runtime Library (libstdc++) 657 658 * The type of exception thrown by iostreams, std::ios_base::failure, 659 now uses the [12]cxx11 ABI. 660 * Experimental support for C++17, including the following new 661 features: 662 + std::string_view; 663 + std::any, std::optional, and std::variant; 664 + std::invoke, std::is_invocable, std::is_nothrow_invocable, and 665 invoke_result; 666 + std::is_swappable, and std::is_nothrow_swappable; 667 + std::apply, and std::make_from_tuple; 668 + std::void_t, std::bool_constant, std::conjunction, 669 std::disjunction, and std::negation; 670 + Variable templates for type traits; 671 + Mathematical Special Functions; 672 + std::chrono::floor, std::chrono::ceil, std::chrono::round, and 673 std::chrono::abs; 674 + std::clamp, std::gcd, std::lcm, 3-dimensional std::hypot; 675 + std::scoped_lock, std::shared_mutex, 676 std::atomic<T>::is_always_lock_free; 677 + std::sample, std::default_searcher, std::boyer_moore_searcher 678 and std::boyer_moore_horspool_searcher; 679 + Extraction and re-insertion of map and set nodes, try_emplace 680 members for maps, and functions for accessing containers 681 std::size, std::empty, and std::data; 682 + std::shared_ptr support for arrays, 683 std::shared_ptr<T>::weak_type, 684 std::enable_shared_from_this<T>::weak_from_this(), and 685 std::owner_less<void>; 686 + std::byte; 687 + std::as_const, std::not_fn, 688 std::has_unique_object_representations, constexpr 689 std::addressof. 690 Thanks to Daniel Kruegler, Tim Shen, Edward Smith-Rowland, and 691 Ville Voutilainen for work on the C++17 support. 692 * A new power-of-two rehashing policy for use with the _Hashtable 693 internals, thanks to Franc,ois Dumont. 694 695 Fortran 696 697 * Support for a number of extensions for compatibility with legacy 698 code with new flags: 699 + -fdec-structure Support for DEC STRUCTURE and UNION 700 + -fdec-intrinsic-ints Support for new integer intrinsics with 701 B/I/J/K prefixes such as BABS, JIAND... 702 + -fdec-math Support for additional math intrinsics, including 703 COTAN and degree-valued trigonometric functions such as TAND, 704 ASIND... 705 + -fdec Enable the -fdec-* family of extensions. 706 * New flag -finit-derived to allow default initialization of 707 derived-type variables. 708 * Improved DO loops with step equal to 1 or -1, generates faster code 709 without a loop preheader. A new warning, -Wundefined-do-loop, warns 710 when a loop iterates either to HUGE(i) (with step equal to 1), or 711 to -HUGE(i) (with step equal to -1). Invalid behavior can be caught 712 at run time with -fcheck=do enabled: 713 714 program test 715 implicit none 716 integer(1) :: i 717 do i = -HUGE(i)+10, -HUGE(i)-1, -1 718 print *, i 719 end do 720 end program test 721 722 At line 8 of file do_check_12.f90 723 Fortran runtime error: Loop iterates infinitely 724 725 * Version 4.5 of the [13]OpenMP specification is now partially 726 supported in the Fortran compiler; the largest missing item is 727 structure element mapping. 728 * User-defined derived-type input/output (UDTIO) is added. 729 * Derived type coarrays with allocatable and pointer components are 730 partially supported. 731 * Non-constant stop codes and error stop codes (Fortran 2015 732 feature). 733 * Derived types with allocatable components of recursive type. 734 * Intrinsic assignment to polymorphic variables. 735 * Improved submodule support. 736 * Improved diagnostics (polymorphic results in pure functions). 737 * Coarray: Support for failed images (Fortan 2015 feature). 738 739 Go 740 741 * GCC 7 provides a complete implementation of the Go 1.8.1 user 742 packages. 743 * Compared to the Go 1.8.1 toolchain, the garbage collector is more 744 conservative and less concurrent. 745 * Escape analysis is available for experimental use via the 746 -fgo-optimize-allocs option. The -fgo-debug-escape prints 747 information useful for debugging escape analysis choices. 748 749 Java (GCJ) 750 751 The GCC Java front end and associated libjava runtime library have been 752 removed from GCC. 753 754 libgccjit 755 756 The libgccjit API gained support for marking calls as requiring 757 tail-call optimization via a new entry point: 758 [14]gcc_jit_rvalue_set_bool_require_tail_call. 759 760 libgccjit performs numerous checks at the API boundary, but if these 761 succeed, it previously ignored errors and other diagnostics emitted 762 within the core of GCC, and treated the compile of a gcc_jit_context as 763 having succeeded. As of GCC 7 it now ensures that if any diagnostics 764 are emitted, they are visible from the libgccjit API, and that the the 765 context is flagged as having failed. 766 767 New Targets and Target Specific Improvements 768 769 AArch64 770 771 * GCC has been updated to the latest revision of the procedure call 772 standard (AAPCS64) to provide support for parameter passing when 773 data types have been over-aligned. 774 * The ARMv8.3-A architecture is now supported. It can be used by 775 specifying the -march=armv8.3-a option. 776 * The option -msign-return-address= is supported to enable return 777 address protection using ARMv8.3-A Pointer Authentication 778 Extensions. For more information on the arguments accepted by this 779 option, please refer to [15]AArch64-Options. 780 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 781 Extensions are now supported. They can be used by specifying the 782 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 783 Floating-Point Extensions introduce new half-precision data 784 processing floating-point instructions. 785 * Support has been added for the following processors (GCC 786 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), Broadcom 787 Vulcan (vulcan), Cavium ThunderX CN81xx (thunderxt81), Cavium 788 ThunderX CN83xx (thunderxt83), Cavium ThunderX CN88xx 789 (thunderxt88), Cavium ThunderX CN88xx pass 1.x (thunderxt88p1), 790 Cavium ThunderX 2 CN99xx (thunderx2t99), Qualcomm Falkor (falkor). 791 The GCC identifiers can be used as arguments to the -mcpu or -mtune 792 options, for example: -mcpu=cortex-a73 or -mtune=vulcan or as 793 arguments to the equivalent target attributes and pragmas. 794 795 ARC 796 797 * Added support for ARC HS and ARC EM processors. 798 * Added support for ARC EM variation found in Intel QuarkSE SoCs. 799 * Added support for NPS400 ARC700 based CPUs. 800 * Thread Local Storage is now supported by ARC CPUs. 801 * Fixed errors for ARC600 when using 32x16 multiplier option. 802 * Fixed PIE for ARC CPUs. 803 * New CPU templates are supported via multilib. 804 805 ARM 806 807 * Support for the ARMv5 and ARMv5E architectures has been deprecated 808 (which have no known implementations) and will be removed in a 809 future GCC release. Note that ARMv5T, ARMv5TE and ARMv5TEJ 810 architectures remain supported. The values armv5 and armv5e of 811 -march are thus deprecated. 812 * The ARMv8.2-A architecture and the ARMv8.2-A 16-bit Floating-Point 813 Extensions are now supported. They can be used by specifying the 814 -march=armv8.2-a or -march=armv8.2-a+fp16 options. The 16-bit 815 Floating-Point Extensions introduce new half-precision data 816 processing floating-point instructions. 817 * The ARMv8-M architecture is now supported in its two architecture 818 profiles: ARMv8-M Baseline and ARMv8-M Mainline with its DSP and 819 Floating-Point Extensions. They can be used by specifying the 820 -march=armv8-m.base, armv8-m.main or armv8-m.main+dsp options. 821 * Support has been added for the following processors (GCC 822 identifiers in parentheses): ARM Cortex-A73 (cortex-a73), ARM 823 Cortex-M23 (cortex-m23) and ARM Cortex-M33 (cortex-m33). The GCC 824 identifiers can be used as arguments to the -mcpu or -mtune 825 options, for example: -mcpu=cortex-a73 or -mtune=cortex-m33. 826 * A new command-line option -mpure-code has been added. It does not 827 allow constant data to be placed in code sections. This option is 828 only available when generating non-PIC code for ARMv7-M targets. 829 * Support for the ACLE Coprocessor Intrinsics has been added. This 830 enables the generation of coprocessor instructions through the use 831 of intrinsics such as cdp, ldc, and others. 832 * The configure option --with-multilib-list now accepts the value 833 rmprofile to build multilib libraries for a range of embedded 834 targets. See our [16]installation instructions for details. 835 836 AVR 837 838 * On the reduced Tiny cores, the progmem [17]variable attribute is 839 now properly supported. Respective read-only variables are located 840 in flash memory in section .progmem.data. No special code is needed 841 to access such variables; the compiler automatically adds an offset 842 of 0x4000 to all addresses, which is needed to access variables in 843 flash memory. As opposed to ordinary cores where it is sufficient 844 to specify the progmem attribute with definitions, on the reduced 845 Tiny cores the attribute also has to be specified with (external) 846 declarations: 847 848 extern const int array[] __attribute__((__progmem__)); 849 850 int get_value2 (void) 851 { 852 /* Access via addresses array + 0x4004 and array + 0x4005. */ 853 return array[2]; 854 } 855 856 const int* get_address (unsigned idx) 857 { 858 /* Returns array + 0x4000 + 2 * idx. */ 859 return &array[idx]; 860 } 861 862 * A new command-line option -Wmisspelled-isr has been added. It turns 863 off -- or turns into errors -- warnings that are reported for 864 interrupt service routines (ISRs) which don't follow AVR-LibC's 865 naming convention of prefixing ISR names with __vector. 866 * __builtin_avr_nops(n) is a new [18]built-in function that inserts n 867 NOP instructions into the instruction stream. n must be a value 868 known at compile time. 869 870 IA-32/x86-64 871 872 * Support for the AVX-512 Fused Multiply Accumulation Packed Single 873 precision (4FMAPS), AVX-512 Vector Neural Network Instructions Word 874 variable precision (4VNNIW), AVX-512 Vector Population Count 875 (VPOPCNTDQ) and Software Guard Extensions (SGX) ISA extensions has 876 been added. 877 878 NVPTX 879 880 * OpenMP target regions can now be offloaded to NVidia PTX GPGPUs. 881 See the [19]Offloading Wiki on how to configure it. 882 883 PowerPC / PowerPC64 / RS6000 884 885 * The PowerPC port now uses LRA by default. 886 * GCC now diagnoses inline assembly that clobbers register r2. This 887 has always been invalid code, and is no longer quietly tolerated. 888 * The PowerPC port's support for ISA 3.0 (-mcpu=power9) has been 889 enhanced to generate more of the new instructions by default, and 890 to provide more built-in functions to generate code for other new 891 instructions. 892 * The configuration option --enable-gnu-indirect-function is now 893 enabled by default on PowerPC GNU/Linux builds. 894 * The PowerPC port will now allow 64-bit and 32-bit integer types to 895 be allocated to the VSX vector registers (ISA 2.06 and above). In 896 addition, on ISA 3.0, 16-bit and 8-bit integer types can be 897 allocated in the vector registers. Previously, only 64-bit integer 898 types were allowed in the traditional floating point registers. 899 * New options -mstack-protector-guard=global, 900 -mstack-protector-guard=tls, -mstack-protector-guard-reg=, and 901 -mstack-protector-guard-offset= change how the stack protector gets 902 the value to use as canary. 903 904 S/390, System z, IBM z Systems, IBM Z 905 906 * Support for the IBM z14 processor has been added. When using the 907 -march=z14 option, the compiler will generate code making use of 908 the new instructions introduced with the vector enhancement 909 facility and the miscellaneous instruction extension facility 2. 910 The -mtune=z14 option enables z14 specific instruction scheduling 911 without making use of new instructions. 912 * Builtins for the new vector instructions have been added and can be 913 enabled using the -mzvector option. 914 915 RISC-V 916 917 * Support for the RISC-V instruction set has been added. 918 919 RX 920 921 Basic support for atomic built-in function has been added. It is 922 currently implemented by flipping interrupts off and on as needed. 923 924 SH 925 926 * Support for SH5/SH64 has been removed. 927 * Improved utilization of delay slots on SH2A. 928 * Improved utilization of zero-displacement conditional branches. 929 * The following deprecated options have been removed 930 + -mcbranchdi 931 + -mcmpeqdi 932 + -minvalid-symbols 933 + -msoft-atomic 934 + -mspace 935 + -madjust-unroll 936 * Support for the following SH2A instructions has been added 937 + mov.b @-Rm,R0 938 + mov.w @-Rm,R0 939 + mov.l @-Rm,R0 940 + mov.b R0,@Rn+ 941 + mov.w R0,@Rn+ 942 + mov.l R0,@Rn+ 943 944 SPARC 945 946 * The SPARC port now uses LRA by default. 947 * Support for the new Subtract-Extended-with-Carry instruction 948 available in SPARC M7 (Niagara 7) has been added. 949 950 Operating Systems 951 952 AIX 953 954 * Visibility support has been enabled for AIX 7.1 and above. 955 956 Fuchsia 957 958 * Support has been added for the [20]Fuchsia OS. 959 960 RTEMS 961 962 * The ABI changes on ARM so that no short enums are used by default. 963 964 Other significant improvements 965 966 * -fverbose-asm previously emitted information on the meanings of 967 assembly expressions. This has been extended so that it now also 968 prints comments showing the source lines that correspond to the 969 assembly, making it easier to read the generated assembly 970 (especially with larger functions). For example, given this C 971 source file: 972 973 int test (int n) 974 { 975 int i; 976 int total = 0; 977 978 for (i = 0; i < n; i++) 979 total += i * i; 980 return total; 981 } 982 983 -fverbose-asm now gives output similar to this for the function 984 body (when compiling for x86_64, with -Os): 985 986 .text 987 .globl test 988 .type test, @@function 989 test: 990 .LFB0: 991 .cfi_startproc 992 # example.c:4: int total = 0; 993 xorl %eax, %eax # <retval> 994 # example.c:6: for (i = 0; i < n; i++) 995 xorl %edx, %edx # i 996 .L2: 997 # example.c:6: for (i = 0; i < n; i++) 998 cmpl %edi, %edx # n, i 999 jge .L5 #, 1000 # example.c:7: total += i * i; 1001 movl %edx, %ecx # i, tmp92 1002 imull %edx, %ecx # i, tmp92 1003 # example.c:6: for (i = 0; i < n; i++) 1004 incl %edx # i 1005 # example.c:7: total += i * i; 1006 addl %ecx, %eax # tmp92, <retval> 1007 jmp .L2 # 1008 .L5: 1009 # example.c:10: } 1010 ret 1011 .cfi_endproc 1012 1013 * Two new options have been added for printing fix-it hints: 1014 + -fdiagnostics-parseable-fixits allows for fix-it hints to be 1015 emitted in a machine-readable form, suitable for consumption 1016 by IDEs. For example, given: 1017 1018 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 1019 you mean 'color'? 1020 return ptr->colour; 1021 ^~~~~~ 1022 color 1023 1024 it will emit: 1025 1026 fix-it:"spellcheck-fields.cc":{52:13-52:19}:"color" 1027 1028 + -fdiagnostics-generate-patch will print a patch in "unified" 1029 format after any diagnostics are printed, showing the result 1030 of applying all fix-it hints. For the above example it would 1031 emit: 1032 1033 --- spellcheck-fields.cc 1034 +++ spellcheck-fields.cc 1035 @@ -49,5 +49,5 @@ 1036 1037 color get_color(struct s *ptr) 1038 { 1039 - return ptr->colour; 1040 + return ptr->color; 1041 } 1042 1043 * The gcc and g++ driver programs will now provide suggestions for 1044 misspelled arguments to command-line options. 1045 1046 $ gcc -c test.c -ftls-model=global-dinamic 1047 gcc: error: unknown TLS model 'global-dinamic' 1048 gcc: note: valid arguments to '-ftls-model=' are: global-dynamic initial-exec lo 1049 cal-dynamic local-exec; did you mean 'global-dynamic'? 1050 1051 * The compiler will now provide suggestions for misspelled 1052 parameters. 1053 1054 $ gcc -c test.c --param max-early-inliner-iteration=3 1055 cc1: error: invalid --param name 'max-early-inliner-iteration'; did you mean 'ma 1056 x-early-inliner-iterations'? 1057 1058 * Profile-guided optimization (PGO) instrumentation, as well as test 1059 coverage (GCOV), can newly instrument constructors (functions marks 1060 with __attribute__((constructor))), destructors and C++ 1061 constructors (and destructors) of classes that are used as the type 1062 of a global variable. 1063 * A new option -fprofile-update=atomic prevents creation of corrupted 1064 profiles created during an instrumentation run (-fprofile=generate) 1065 of an application. The downside of the option is a speed penalty. 1066 Providing -pthread on the command line selects atomic profile 1067 updating (when supported by the target). 1068 * GCC's already extensive testsuite has gained some new capabilities, 1069 to further improve the reliability of the compiler: 1070 + GCC now has an internal unit-testing API and a suite of tests 1071 for programmatic self-testing of subsystems. 1072 + GCC's C front end has been extended so that it can parse dumps 1073 of GCC's internal representations, allowing for DejaGnu tests 1074 that more directly exercise specific optimization passes. This 1075 covers both the [21]GIMPLE representation (for testing 1076 higher-level optimizations) and the [22]RTL representation, 1077 allowing for more direct testing of lower-level details, such 1078 as register allocation and instruction selection. 1079 1080 GCC 7.1 1081 1082 This is the [23]list of problem reports (PRs) from GCC's bug tracking 1083 system that are known to be fixed in the 7.1 release. This list might 1084 not be complete (that is, it is possible that some PRs that have been 1085 fixed are not listed here). 1086 1087 GCC 7.2 1088 1089 This is the [24]list of problem reports (PRs) from GCC's bug tracking 1090 system that are known to be fixed in the 7.2 release. This list might 1091 not be complete (that is, it is possible that some PRs that have been 1092 fixed are not listed here). 1093 1094 Target Specific Changes 1095 1096 SPARC 1097 1098 * Support for the SPARC M8 processor has been added. 1099 * The switches -mfix-ut700 and -mfix-gr712rc have been added to work 1100 around an erratum in LEON3FT processors. 1101 * Use of the Floating-point Multiply Single to Double (FsMULd) 1102 instruction can now be controlled by the -mfsmuld and -fno-fsmuld 1103 options. 1104 1105 Operating Systems 1106 1107 RTEMS 1108 1109 * The Ada run-time support uses now thread-local storage (TLS). 1110 * Support for RISC-V has been added. 1111 * Support for 64-bit PowerPC using the ELFv2 ABI with 64-bit long 1112 double has been added. 1113 1114 GCC 7.3 1115 1116 This is the [25]list of problem reports (PRs) from GCC's bug tracking 1117 system that are known to be fixed in the 7.3 release. This list might 1118 not be complete (that is, it is possible that some PRs that have been 1119 fixed are not listed here). 1120 1121 Target Specific Changes 1122 1123 SPARC 1124 1125 * Workarounds for the four [26]LEON3FT errata GRLIB-TN-0010..0013 1126 have been added. Relevant errata are activated by the target 1127 specific -mfix-ut699, -mfix-ut700 and -mfix-gr712rc switches. 1128 1129 Operating Systems 1130 1131 RTEMS 1132 1133 * Support has been added for Epiphany target. 1134 1135 GCC 7.4 1136 1137 This is the [27]list of problem reports (PRs) from GCC's bug tracking 1138 system that are known to be fixed in the 7.4 release. This list might 1139 not be complete (that is, it is possible that some PRs that have been 1140 fixed are not listed here). 1141 1142 1143 For questions related to the use of GCC, please consult these web 1144 pages and the [28]GCC manuals. If that fails, the 1145 [29]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1146 web pages and the development of GCC are welcome on our developer 1147 list at [30]gcc (a] gcc.gnu.org. All of [31]our lists have public 1148 archives. 1149 1150 Copyright (C) [32]Free Software Foundation, Inc. Verbatim copying and 1151 distribution of this entire article is permitted in any medium, 1152 provided this notice is preserved. 1153 1154 These pages are [33]maintained by the GCC team. Last modified 1155 2018-12-06[34]. 1156 1157 References 1158 1159 1. http://gcc.gnu.org/gcc-7/porting_to.html 1160 2. http://gcc.gnu.org/onlinedocs/index.html#current 1161 3. https://gcc.gnu.org/wiki/LRAIsDefault 1162 4. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/manual/profile_mode.html 1163 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77728 1164 6. http://www.dwarfstd.org/Download.php 1165 7. https://gcc.gnu.org/wiki/OpenACC 1166 8. https://gcc.gnu.org/wiki/Offloading 1167 9. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/Object-Size-Checking.html 1168 10. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 1169 11. http://wg21.link/p0136 1170 12. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/libstdc++/manual/using_dual_abi.html 1171 13. https://www.openmp.org/specifications/ 1172 14. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/jit/topics/expressions.html#gcc_jit_rvalue_set_bool_require_tail_call 1173 15. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AArch64-Options.html#AArch64-Options 1174 16. https://gcc.gnu.org/install/configure.html 1175 17. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Variable-Attributes.html 1176 18. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gcc/AVR-Built-in-Functions.html 1177 19. https://gcc.gnu.org/wiki/Offloading 1178 20. https://fuchsia.googlesource.com/ 1179 21. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/GIMPLE-Tests.html 1180 22. https://gcc.gnu.org/onlinedocs/gcc-7.1.0/gccint/RTL-Tests.html 1181 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.0 1182 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.2 1183 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.3 1184 26. https://www.gaisler.com/index.php/information/app-tech-notes 1185 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=7.4 1186 28. https://gcc.gnu.org/onlinedocs/ 1187 29. mailto:gcc-help (a] gcc.gnu.org 1188 30. mailto:gcc (a] gcc.gnu.org 1189 31. https://gcc.gnu.org/lists.html 1190 32. https://www.fsf.org/ 1191 33. https://gcc.gnu.org/about.html 1192 34. http://validator.w3.org/check/referer 1193 ====================================================================== 1194 http://gcc.gnu.org/gcc-6/index.html 1195 1196 GCC 6 Release Series 1197 1198 October 26, 2018 1199 1200 The [1]GNU project and the GCC developers are pleased to announce the 1201 release of GCC 6.5. 1202 1203 This release is a bug-fix release, containing fixes for regressions in 1204 GCC 6.4 relative to previous releases of GCC. 1205 1206 Release History 1207 1208 GCC 6.5 1209 October 26, 2018 ([2]changes, [3]documentation) 1210 1211 GCC 6.4 1212 July 4, 2017 ([4]changes, [5]documentation) 1213 1214 GCC 6.3 1215 December 21, 2016 ([6]changes, [7]documentation) 1216 1217 GCC 6.2 1218 August 22, 2016 ([8]changes, [9]documentation) 1219 1220 GCC 6.1 1221 April 27, 2016 ([10]changes, [11]documentation) 1222 1223 References and Acknowledgements 1224 1225 GCC used to stand for the GNU C Compiler, but since the compiler 1226 supports several other languages aside from C, it now stands for the 1227 GNU Compiler Collection. 1228 1229 A list of [12]successful builds is updated as new information becomes 1230 available. 1231 1232 The GCC developers would like to thank the numerous people that have 1233 contributed new features, improvements, bug fixes, and other changes as 1234 well as test results to GCC. This [13]amazing group of volunteers is 1235 what makes GCC successful. 1236 1237 For additional information about GCC please refer to the [14]GCC 1238 project web site or contact the [15]GCC development mailing list. 1239 1240 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 1241 1242 1243 For questions related to the use of GCC, please consult these web 1244 pages and the [18]GCC manuals. If that fails, the 1245 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1246 web pages and the development of GCC are welcome on our developer 1247 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 1248 archives. 1249 1250 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 1251 distribution of this entire article is permitted in any medium, 1252 provided this notice is preserved. 1253 1254 These pages are [23]maintained by the GCC team. Last modified 1255 2018-10-26[24]. 1256 1257 References 1258 1259 1. http://www.gnu.org/ 1260 2. http://gcc.gnu.org/gcc-6/changes.html 1261 3. http://gcc.gnu.org/onlinedocs/6.5.0/ 1262 4. http://gcc.gnu.org/gcc-6/changes.html 1263 5. http://gcc.gnu.org/onlinedocs/6.4.0/ 1264 6. http://gcc.gnu.org/gcc-6/changes.html 1265 7. http://gcc.gnu.org/onlinedocs/6.3.0/ 1266 8. http://gcc.gnu.org/gcc-6/changes.html 1267 9. http://gcc.gnu.org/onlinedocs/6.2.0/ 1268 10. http://gcc.gnu.org/gcc-6/changes.html 1269 11. http://gcc.gnu.org/onlinedocs/6.1.0/ 1270 12. http://gcc.gnu.org/gcc-6/buildstat.html 1271 13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 1272 14. http://gcc.gnu.org/index.html 1273 15. mailto:gcc (a] gcc.gnu.org 1274 16. http://gcc.gnu.org/mirrors.html 1275 17. http://gcc.gnu.org/svn.html 1276 18. https://gcc.gnu.org/onlinedocs/ 1277 19. mailto:gcc-help (a] gcc.gnu.org 1278 20. mailto:gcc (a] gcc.gnu.org 1279 21. https://gcc.gnu.org/lists.html 1280 22. https://www.fsf.org/ 1281 23. https://gcc.gnu.org/about.html 1282 24. http://validator.w3.org/check/referer 1283 ====================================================================== 1284 http://gcc.gnu.org/gcc-6/changes.html 1285 1286 GCC 6 Release Series 1287 Changes, New Features, and Fixes 1288 1289 This page is a brief summary of some of the huge number of improvements 1290 in GCC 6. For more information, see the [1]Porting to GCC 6 page and 1291 the [2]full GCC documentation. 1292 1293 Caveats 1294 1295 * The default mode for C++ is now -std=gnu++14 instead of 1296 -std=gnu++98. 1297 * Support for a number of older systems and recently unmaintained or 1298 untested target ports of GCC has been declared obsolete in GCC 6. 1299 Unless there is activity to revive them, the next release of GCC 1300 will have their sources permanently removed. 1301 The following ports for individual systems on particular 1302 architectures have been obsoleted: 1303 + SH5 / SH64 (sh64-*-*) as announced [3]here. 1304 * The AVR port requires binutils version 2.26.1 or later for the fix 1305 for [4]PR71151 to work. 1306 * The GCC 6.5 release has an accidental ABI incompatibility for 1307 nested std::pair objects, for more details see [5]PR 87822. The bug 1308 causes a layout change for pairs where the first member is also a 1309 pair, e.g. std::pair<std::pair<X, Y>, Z>. The GCC 6 release series 1310 is closed so the bug in GCC 6.5 will not be fixed upstream, but 1311 there is a patch in the bug report to allow it to be fixed by 1312 anybody packaging GCC 6.5 or installing it themselves. 1313 1314 General Optimizer Improvements 1315 1316 * UndefinedBehaviorSanitizer gained a new sanitization option, 1317 -fsanitize=bounds-strict, which enables strict checking of array 1318 bounds. In particular, it enables -fsanitize=bounds as well as 1319 instrumentation of flexible array member-like arrays. 1320 * Type-based alias analysis now disambiguates accesses to different 1321 pointers. This improves precision of the alias oracle by about 1322 20-30% on higher-level C++ programs. Programs doing invalid type 1323 punning of pointer types may now need -fno-strict-aliasing to work 1324 correctly. 1325 * Alias analysis now correctly supports the weakref and alias 1326 attributes. This allows accessing both a variable and its alias in 1327 one translation unit which is common with link-time optimization. 1328 * Value range propagation now assumes that the this pointer in C++ 1329 member functions is non-null. This eliminates common null pointer 1330 checks but also breaks some non-conforming code-bases (such as 1331 Qt-5, Chromium, KDevelop). As a temporary work-around 1332 -fno-delete-null-pointer-checks can be used. Wrong code can be 1333 identified by using -fsanitize=undefined. 1334 * Link-time optimization improvements: 1335 + warning and error attributes are now correctly preserved by 1336 declaration linking and thus -D_FORTIFY_SOURCE=2 is now 1337 supported with -flto. 1338 + Type merging was fixed to handle C and Fortran 1339 interoperability rules as defined by the Fortran 2008 language 1340 standard. 1341 As an exception, CHARACTER(KIND=C_CHAR) is not inter-operable 1342 with char in all cases because it is an array while char is 1343 scalar. INTEGER(KIND=C_SIGNED_CHAR) should be used instead. In 1344 general, this inter-operability cannot be implemented, for 1345 example on targets where the argument passing convention for 1346 arrays differs from scalars. 1347 + More type information is now preserved at link time, reducing 1348 the loss of accuracy of the type-based alias analysis compared 1349 to builds without link-time optimization. 1350 + Invalid type punning on global variables and declarations is 1351 now reported with -Wodr-type-mismatch. 1352 + The size of LTO object files was reduced by about 11% 1353 (measured by compiling Firefox 46.0). 1354 + Link-time parallelization (enabled using -flto=n) was 1355 significantly improved by decreasing the size of streamed data 1356 when partitioning programs. The size of streamed IL while 1357 compiling Firefox 46.0 was reduced by 66%. 1358 + The linker plugin was extended to pass information about the 1359 type of binary produced to the GCC back end. (That can also be 1360 controlled manually by -flinker-output.) This makes it 1361 possible to properly configure the code generator and support 1362 incremental linking. Incremental linking of LTO objects by gcc 1363 -r is now supported for plugin-enabled setups. 1364 There are two ways to perform incremental linking: 1365 1. Linking by ld -r will result in an object file with all 1366 sections from individual object files mechanically 1367 merged. This delays the actual link-time optimization to 1368 the final linking step and thus permits whole program 1369 optimization. Linking the final binary with such object 1370 files is however slower. 1371 2. Linking by gcc -r will lead to link-time optimization and 1372 emit the final binary into the object file. Linking such 1373 an object file is fast but avoids any benefits from whole 1374 program optimization. 1375 GCC 7 will support incremental link-time optimization with gcc 1376 -r. 1377 * Inter-procedural optimization improvements: 1378 + Basic jump threading is now performed before profile 1379 construction and inline analysis, resulting in more realistic 1380 size and time estimates that drive the heuristics of the 1381 inliner and function cloning passes. 1382 + Function cloning now more aggressively eliminates unused 1383 function parameters. 1384 1385 New Languages and Language specific improvements 1386 1387 Compared to GCC 5, the GCC 6 release series includes a much improved 1388 implementation of the [6]OpenACC 2.0a specification. Highlights are: 1389 * In addition to single-threaded host-fallback execution, offloading 1390 is supported for nvptx (Nvidia GPUs) on x86_64 and PowerPC 64-bit 1391 little-endian GNU/Linux host systems. For nvptx offloading, with 1392 the OpenACC parallel construct, the execution model allows for an 1393 arbitrary number of gangs, up to 32 workers, and 32 vectors. 1394 * Initial support for parallelized execution of OpenACC kernels 1395 constructs: 1396 + Parallelization of a kernels region is switched on by 1397 -fopenacc combined with -O2 or higher. 1398 + Code is offloaded onto multiple gangs, but executes with just 1399 one worker, and a vector length of 1. 1400 + Directives inside a kernels region are not supported. 1401 + Loops with reductions can be parallelized. 1402 + Only kernels regions with one loop nest are parallelized. 1403 + Only the outer-most loop of a loop nest can be parallelized. 1404 + Loop nests containing sibling loops are not parallelized. 1405 Typically, using the OpenACC parallel construct gives much better 1406 performance, compared to the initial support of the OpenACC kernels 1407 construct. 1408 * The device_type clause is not supported. The bind and nohost 1409 clauses are not supported. The host_data directive is not supported 1410 in Fortran. 1411 * Nested parallelism (cf. CUDA dynamic parallelism) is not supported. 1412 * Usage of OpenACC constructs inside multithreaded contexts (such as 1413 created by OpenMP, or pthread programming) is not supported. 1414 * If a call to the acc_on_device function has a compile-time constant 1415 argument, the function call evaluates to a compile-time constant 1416 value only for C and C++ but not for Fortran. 1417 1418 See the [7]OpenACC and [8]Offloading wiki pages for further 1419 information. 1420 1421 C family 1422 1423 * Version 4.5 of the [9]OpenMP specification is now supported in the 1424 C and C++ compilers. 1425 * The C and C++ compilers now support attributes on enumerators. For 1426 instance, it is now possible to mark enumerators as deprecated: 1427 1428 enum { 1429 newval, 1430 oldval __attribute__ ((deprecated ("too old"))) 1431 }; 1432 1433 * Source locations for the C and C++ compilers are now tracked as 1434 ranges, rather than just points, making it easier to identify the 1435 subexpression of interest within a complicated expression. For 1436 example: 1437 1438 test.cc: In function 'int test(int, int, foo, int, int)': 1439 test.cc:5:16: error: no match for 'operator*' (operand types are 'int' and 'foo' 1440 ) 1441 return p + q * r * s + t; 1442 ~~^~~ 1443 1444 In addition, there is now initial support for precise diagnostic 1445 locations within strings: 1446 1447 format-strings.c:3:14: warning: field width specifier '*' expects a matching 'in 1448 t' argument [-Wformat=] 1449 printf("%*d"); 1450 ^ 1451 1452 * Diagnostics can now contain "fix-it hints", which are displayed in 1453 context underneath the relevant source code. For example: 1454 1455 fixits.c: In function 'bad_deref': 1456 fixits.c:11:13: error: 'ptr' is a pointer; did you mean to use '->'? 1457 return ptr.x; 1458 ^ 1459 -> 1460 1461 * The C and C++ compilers now offer suggestions for misspelled field 1462 names: 1463 1464 spellcheck-fields.cc:52:13: error: 'struct s' has no member named 'colour'; did 1465 you mean 'color'? 1466 return ptr->colour; 1467 ^~~~~~ 1468 1469 * New command-line options have been added for the C and C++ 1470 compilers: 1471 + -Wshift-negative-value warns about left shifting a negative 1472 value. 1473 + -Wshift-overflow warns about left shift overflows. This 1474 warning is enabled by default. -Wshift-overflow=2 also warns 1475 about left-shifting 1 into the sign bit. 1476 + -Wtautological-compare warns if a self-comparison always 1477 evaluates to true or false. This warning is enabled by -Wall. 1478 + -Wnull-dereference warns if the compiler detects paths that 1479 trigger erroneous or undefined behavior due to dereferencing a 1480 null pointer. This option is only active when 1481 -fdelete-null-pointer-checks is active, which is enabled by 1482 optimizations in most targets. The precision of the warnings 1483 depends on the optimization options used. 1484 + -Wduplicated-cond warns about duplicated conditions in an 1485 if-else-if chain. 1486 + -Wmisleading-indentation warns about places where the 1487 indentation of the code gives a misleading idea of the block 1488 structure of the code to a human reader. For example, given 1489 [10]CVE-2014-1266: 1490 1491 sslKeyExchange.c: In function 'SSLVerifySignedServerKeyExchange': 1492 sslKeyExchange.c:629:3: warning: this 'if' clause does not guard... [-Wmisleadin 1493 g-indentation] 1494 if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0) 1495 ^~ 1496 sslKeyExchange.c:631:5: note: ...this statement, but the latter is misleadingly 1497 indented as if it is guarded by the 'if' 1498 goto fail; 1499 ^~~~ 1500 1501 This warning is enabled by -Wall. 1502 * The C and C++ compilers now emit saner error messages if 1503 merge-conflict markers are present in a source file. 1504 1505 test.c:3:1: error: version control conflict marker in file 1506 <<<<<<< HEAD 1507 ^~~~~~~ 1508 1509 C 1510 1511 * It is possible to disable warnings when an initialized field of a 1512 structure or a union with side effects is being overridden when 1513 using designated initializers via a new warning option 1514 -Woverride-init-side-effects. 1515 * A new type attribute scalar_storage_order applying to structures 1516 and unions has been introduced. It specifies the storage order (aka 1517 endianness) in memory of scalar fields in structures or unions. 1518 1519 C++ 1520 1521 * The default mode has been changed to -std=gnu++14. 1522 * [11]C++ Concepts are now supported when compiling with -fconcepts. 1523 * -flifetime-dse is more aggressive in dead-store elimination in 1524 situations where a memory store to a location precedes a 1525 constructor to that memory location. 1526 * G++ now supports [12]C++17 fold expressions, u8 character literals, 1527 extended static_assert, and nested namespace definitions. 1528 * G++ now allows constant evaluation for all non-type template 1529 arguments. 1530 * G++ now supports C++ Transactional Memory when compiling with 1531 -fgnu-tm. 1532 1533 Runtime Library (libstdc++) 1534 1535 * Extensions to the C++ Library to support mathematical special 1536 functions (ISO/IEC 29124:2010), thanks to Edward Smith-Rowland. 1537 * Experimental support for C++17, including the following new 1538 features: 1539 + std::uncaught_exceptions function (this is also available for 1540 -std=gnu++NN modes); 1541 + new member functions try_emplace and insert_or_assign for 1542 unique_key maps; 1543 + non-member functions std::size, std::empty, and std::data for 1544 accessing containers and arrays; 1545 + std::invoke; 1546 + std::shared_mutex; 1547 + std::void_t and std::bool_constant metaprogramming utilities. 1548 Thanks to Ville Voutilainen for contributing many of the C++17 1549 features. 1550 * An experimental implementation of the File System TS. 1551 * Experimental support for most features of the second version of the 1552 Library Fundamentals TS. This includes polymorphic memory resources 1553 and array support in shared_ptr, thanks to Fan You. 1554 * Some assertions checked by Debug Mode can now also be enabled by 1555 _GLIBCXX_ASSERTIONS. The subset of checks enabled by the new macro 1556 have less run-time overhead than the full _GLIBCXX_DEBUG checks and 1557 don't affect the library ABI, so can be enabled per-translation 1558 unit. 1559 * Timed mutex types are supported on more targets, including Darwin. 1560 * Improved std::locale support for DragonFly and FreeBSD, thanks to 1561 John Marino and Andreas Tobler. 1562 1563 Fortran 1564 1565 * Fortran 2008 SUBMODULE support. 1566 * Fortran 2015 EVENT_TYPE, EVENT_POST, EVENT_WAIT, and EVENT_QUERY 1567 support. 1568 * Improved support for Fortran 2003 deferred-length character 1569 variables. 1570 * Improved support for OpenMP and OpenACC. 1571 * The MATMUL intrinsic is now inlined for straightforward cases if 1572 front-end optimization is active. The maximum size for inlining can 1573 be set to n with the -finline-matmul-limit=n option and turned off 1574 with -finline-matmul-limit=0. 1575 * The -Wconversion-extra option will warn about REAL constants which 1576 have excess precision for their kind. 1577 * The -Winteger-division option has been added, which warns about 1578 divisions of integer constants which are truncated. This option is 1579 included in -Wall by default. 1580 1581 libgccjit 1582 1583 * The driver code is now run in-process within libgccjit, providing a 1584 small speed-up of the compilation process. 1585 * The API has gained entrypoints for 1586 + [13]timing how long was spent in different parts of code, 1587 + [14]creating switch statements, 1588 + [15]allowing unreachable basic blocks in a function, and 1589 + [16]adding arbitrary command-line options to a compilation. 1590 1591 New Targets and Target Specific Improvements 1592 1593 AArch64 1594 1595 * A number of AArch64-specific options have been added. The most 1596 important ones are summarised in this section; for more detailed 1597 information please refer to the documentation. 1598 * The command-line options -march=native, -mcpu=native and 1599 -mtune=native are now available on native AArch64 GNU/Linux 1600 systems. Specifying these options causes GCC to auto-detect the 1601 host CPU and choose the optimal setting for that system. 1602 * -fpic is now supported when generating code for the small code 1603 model (-mcmodel=small). The size of the global offset table (GOT) 1604 is limited to 28KiB under the LP64 SysV ABI, and 15KiB under the 1605 ILP32 SysV ABI. 1606 * The AArch64 port now supports target attributes and pragmas. Please 1607 refer to the [17]documentation for details of available attributes 1608 and pragmas as well as usage instructions. 1609 * Link-time optimization across translation units with different 1610 target-specific options is now supported. 1611 * The option -mtls-size= is now supported. It can be used to specify 1612 the bit size of TLS offsets, allowing GCC to generate better TLS 1613 instruction sequences. 1614 * The option -fno-plt is now fully functional. 1615 * The ARMv8.1-A architecture and the Large System Extensions are now 1616 supported. They can be used by specifying the -march=armv8.1-a 1617 option. Additionally, the +lse option extension can be used in a 1618 similar fashion to other option extensions. The Large System 1619 Extensions introduce new instructions that are used in the 1620 implementation of atomic operations. 1621 * The ACLE half-precision floating-point type __fp16 is now supported 1622 in the C and C++ languages. 1623 * The ARM Cortex-A35 processor is now supported via the 1624 -mcpu=cortex-a35 and -mtune=cortex-a35 options as well as the 1625 equivalent target attributes and pragmas. 1626 * The Qualcomm QDF24xx processor is now supported via the 1627 -mcpu=qdf24xx and -mtune=qdf24xx options as well as the equivalent 1628 target attributes and pragmas. 1629 * Code generation for the ARM Cortex-A57 processor is improved. Among 1630 general code generation improvements, a better algorithm is added 1631 for allocating registers to floating-point multiply-accumulate 1632 instructions offering increased performance when compiling with 1633 -mcpu=cortex-a57 or -mtune=cortex-a57. 1634 * Code generation for the ARM Cortex-A53 processor is improved. A 1635 more accurate instruction scheduling model for the processor is now 1636 used, and a number of compiler tuning parameters have been set to 1637 offer increased performance when compiling with -mcpu=cortex-a53 or 1638 -mtune=cortex-a53. 1639 * Code generation for the Samsung Exynos M1 processor is improved. A 1640 more accurate instruction scheduling model for the processor is now 1641 used, and a number of compiler tuning parameters have been set to 1642 offer increased performance when compiling with -mcpu=exynos-m1 or 1643 -mtune=exynos-m1. 1644 * Improvements in the generation of conditional branches and literal 1645 pools allow the compiler to compile functions of a large size. 1646 Constant pools are now placed into separate rodata sections. The 1647 new option -mpc-relative-literal-loads generates per-function 1648 literal pools, limiting the maximum size of functions to 1MiB. 1649 * Several correctness issues generating Advanced SIMD instructions 1650 for big-endian targets have been fixed resulting in improved code 1651 generation for ACLE intrinsics with -mbig-endian. 1652 1653 ARM 1654 1655 * Support for revisions of the ARM architecture prior to ARMv4t has 1656 been deprecated and will be removed in a future GCC release. The 1657 -mcpu and -mtune values that are deprecated are: arm2, arm250, 1658 arm3, arm6, arm60, arm600, arm610, arm620, arm7, arm7d, arm7di, 1659 arm70, arm700, arm700i, arm710, arm720, arm710c, arm7100, arm7500, 1660 arm7500fe, arm7m, arm7dm, arm7dmi, arm8, arm810, strongarm, 1661 strongarm110, strongarm1100, strongarm1110, fa526, fa626. The value 1662 arm7tdmi is still supported. The values of -march that are 1663 deprecated are: armv2,armv2a,armv3,armv3m,armv4. 1664 * The ARM port now supports target attributes and pragmas. Please 1665 refer to the [18]documentation for details of available attributes 1666 and pragmas as well as usage instructions. 1667 * Support has been added for the following processors (GCC 1668 identifiers in parentheses): ARM Cortex-A32 (cortex-a32), ARM 1669 Cortex-A35 (cortex-a35) and ARM Cortex-R8 (cortex-r8). The GCC 1670 identifiers can be used as arguments to the -mcpu or -mtune 1671 options, for example: -mcpu=cortex-a32 or -mtune=cortex-a35. 1672 1673 Heterogeneous Systems Architecture 1674 1675 * GCC can now generate HSAIL (Heterogeneous System Architecture 1676 Intermediate Language) for simple OpenMP device constructs if 1677 configured with --enable-offload-targets=hsa. A new libgomp plugin 1678 then runs the HSA GPU kernels implementing these constructs on HSA 1679 capable GPUs via a standard HSA run time. 1680 If the HSA compilation back end determines it cannot output HSAIL 1681 for a particular input, it gives a warning by default. These 1682 warnings can be suppressed with -Wno-hsa. To give a few examples, 1683 the HSA back end does not implement compilation of code using 1684 function pointers, automatic allocation of variable sized arrays, 1685 functions with variadic arguments as well as a number of other less 1686 common programming constructs. 1687 When compilation for HSA is enabled, the compiler attempts to 1688 compile composite OpenMP constructs 1689 1690 #pragma omp target teams distribute parallel for 1691 1692 into parallel HSA GPU kernels. 1693 1694 IA-32/x86-64 1695 1696 * GCC now supports the Intel CPU named Skylake with AVX-512 1697 extensions through -march=skylake-avx512. The switch enables the 1698 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 1699 AVX-512DQ. 1700 * Support for new AMD instructions monitorx and mwaitx has been 1701 added. This includes new intrinsic and built-in support. It is 1702 enabled through option -mmwaitx. The instructions monitorx and 1703 mwaitx implement the same functionality as the old monitor and 1704 mwait instructions. In addition mwaitx adds a configurable timer. 1705 The timer value is received as third argument and stored in 1706 register %ebx. 1707 * x86-64 targets now allow stack realignment from a word-aligned 1708 stack pointer using the command-line option -mstackrealign or 1709 __attribute__ ((force_align_arg_pointer)). This allows functions 1710 compiled with a vector-aligned stack to be invoked from objects 1711 that keep only word-alignment. 1712 * Support for address spaces __seg_fs, __seg_gs, and __seg_tls. These 1713 can be used to access data via the %fs and %gs segments without 1714 having to resort to inline assembly. Please refer to the 1715 [19]documentation for usage instructions. 1716 * Support for AMD Zen (family 17h) processors is now available 1717 through the -march=znver1 and -mtune=znver1 options. 1718 1719 MeP 1720 1721 * Support for the MeP (mep-elf) architecture has been deprecated and 1722 will be removed in a future GCC release. 1723 1724 MSP430 1725 1726 * The MSP430 compiler now has the ability to automatically distribute 1727 code and data between low memory (addresses below 64K) and high 1728 memory. This only applies to parts that actually have both memory 1729 regions and only if the linker script for the part has been 1730 specifically set up to support this feature. 1731 A new attribute of either can be applied to both functions and 1732 data, and this tells the compiler to place the object into low 1733 memory if there is room and into high memory otherwise. Two other 1734 new attributes - lower and upper - can be used to explicitly state 1735 that an object should be placed in the specified memory region. If 1736 there is not enough left in that region the compilation will fail. 1737 Two new command-line options - -mcode-region=[lower|upper|either] 1738 and -mdata-region=[lower|upper|either] - can be used to tell the 1739 compiler what to do with objects that do not have one of these new 1740 attributes. 1741 1742 PowerPC / PowerPC64 / RS6000 1743 1744 * PowerPC64 now supports IEEE 128-bit floating-point using the 1745 __float128 data type. In GCC 6, this is not enabled by default, but 1746 you can enable it with -mfloat128. The IEEE 128-bit floating-point 1747 support requires the use of the VSX instruction set. IEEE 128-bit 1748 floating-point values are passed and returned as a single vector 1749 value. The software emulator for IEEE 128-bit floating-point 1750 support is only built on PowerPC GNU/Linux systems where the 1751 default CPU is at least power7. On future ISA 3.0 systems (POWER 9 1752 and later), you will be able to use the -mfloat128-hardware option 1753 to use the ISA 3.0 instructions that support IEEE 128-bit 1754 floating-point. An additional type (__ibm128) has been added to 1755 refer to the IBM extended double type that normally implements long 1756 double. This will allow for a future transition to implementing 1757 long double with IEEE 128-bit floating-point. 1758 * Basic support has been added for POWER9 hardware that will use the 1759 recently published OpenPOWER ISA 3.0 instructions. The following 1760 new switches are available: 1761 + -mcpu=power9: Implement all of the ISA 3.0 instructions 1762 supported by the compiler. 1763 + -mtune=power9: In the future, apply tuning for POWER9 systems. 1764 Currently, POWER8 tunings are used. 1765 + -mmodulo: Generate code using the ISA 3.0 integer instructions 1766 (modulus, count trailing zeros, array index support, integer 1767 multiply/add). 1768 + -mpower9-fusion: Generate code to suitably fuse instruction 1769 sequences for a POWER9 system. 1770 + -mpower9-dform: Generate code to use the new D-form 1771 (register+offset) memory instructions for the vector 1772 registers. 1773 + -mpower9-vector: Generate code using the new ISA 3.0 vector 1774 (VSX or Altivec) instructions. 1775 + -mpower9-minmax: Reserved for future development. 1776 + -mtoc-fusion: Keep TOC entries together to provide more fusion 1777 opportunities. 1778 * New constraints have been added to support IEEE 128-bit 1779 floating-point and ISA 3.0 instructions: 1780 + wb: Altivec register if -mpower9-dform is enabled. 1781 + we: VSX register if -mpower9-vector is enabled for 64-bit code 1782 generation. 1783 + wo: VSX register if -mpower9-vector is enabled. 1784 + wp: Reserved for future use if long double is implemented with 1785 IEEE 128-bit floating-point instead of IBM extended double. 1786 + wq: VSX register if -mfloat128 is enabled. 1787 + wF: Memory operand suitable for POWER9 fusion load/store. 1788 + wG: Memory operand suitable for TOC fusion memory references. 1789 + wL: Integer constant identifying the element number mfvsrld 1790 accesses within a vector. 1791 * Support has been added for __builtin_cpu_is() and 1792 __builtin_cpu_supports(), allowing for very fast access to 1793 AT_PLATFORM, AT_HWCAP, and AT_HWCAP2 values. This requires use of 1794 glibc 2.23 or later. 1795 * All hardware transactional memory builtins now correctly behave as 1796 memory barriers. Programmers can use #ifdef __TM_FENCE__ to 1797 determine whether their "old" compiler treats the builtins as 1798 barriers. 1799 * Split-stack support has been added for gccgo on PowerPC64 for both 1800 big- and little-endian (but not for 32-bit). The gold linker from 1801 at least binutils 2.25.1 must be available in the PATH when 1802 configuring and building gccgo to enable split stack. (The 1803 requirement for binutils 2.25.1 applies to PowerPC64 only.) The 1804 split-stack feature allows a small initial stack size to be 1805 allocated for each goroutine, which increases as needed. 1806 * GCC on PowerPC now supports the standard lround function. 1807 * A new configuration option ---with-advance-toolchain=at was added 1808 for PowerPC 64-bit GNU/Linux systems to use the header files, 1809 library files, and the dynamic linker from a specific Advance 1810 Toolchain release instead of the default versions that are provided 1811 by the GNU/Linux distribution. In general, this option is intended 1812 for the developers of GCC, and it is not intended for general use. 1813 * The "q", "S", "T", and "t" asm-constraints have been removed. 1814 * The "b", "B", "m", "M", and "W" format modifiers have been removed. 1815 1816 S/390, System z, IBM z Systems 1817 1818 * Support for the IBM z13 processor has been added. When using the 1819 -march=z13 option, the compiler will generate code making use of 1820 the new instructions and registers introduced with the vector 1821 extension facility. The -mtune=z13 option enables z13 specific 1822 instruction scheduling without making use of new instructions. 1823 Compiling code with -march=z13 reduces the default alignment of 1824 vector types bigger than 8 bytes to 8. This is an ABI change and 1825 care must be taken when linking modules compiled with different 1826 arch levels which interchange variables containing vector type 1827 values. For newly compiled code the GNU linker will emit a warning. 1828 * The -mzvector option enables a C/C++ language extension. This 1829 extension provides a new keyword vector which can be used to define 1830 vector type variables. (Note: This is not available when enforcing 1831 strict standard compliance e.g. with -std=c99. Either enable GNU 1832 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 1833 Additionally a set of overloaded builtins is provided which is 1834 partially compatible to the PowerPC Altivec builtins. In order to 1835 make use of these builtins the vecintrin.h header file needs to be 1836 included. 1837 * The new command-line options -march=native, and -mtune=native are 1838 now available on native IBM z Systems. Specifying these options 1839 causes GCC to auto-detect the host CPU and choose the optimal 1840 setting for that system. If GCC is unable to detect the host CPU 1841 these options have no effect. 1842 * The IBM z Systems port now supports target attributes and pragmas. 1843 Please refer to the [20]documentation for details of available 1844 attributes and pragmas as well as usage instructions. 1845 * -fsplit-stack is now supported as part of the IBM z Systems port. 1846 This feature requires a recent gold linker to be used. 1847 * Support for the g5 and g6 -march=/-mtune= CPU level switches has 1848 been deprecated and will be removed in a future GCC release. -m31 1849 from now on defaults to -march=z900 if not specified otherwise. 1850 -march=native on a g5/g6 machine will default to -march=z900. 1851 1852 SH 1853 1854 * Support for SH5 / SH64 has been declared obsolete and will be 1855 removed in future releases. 1856 * Support for the FDPIC ABI has been added. It can be enabled using 1857 the new -mfdpic target option and --enable-fdpic configure option. 1858 1859 SPARC 1860 1861 * An ABI bug has been fixed in 64-bit mode. Unfortunately, this 1862 change will break binary compatibility with earlier releases for 1863 code it affects, but this should be pretty rare in practice. The 1864 conditions are: a 16-byte structure containing a double or a 8-byte 1865 vector in the second half is passed to a subprogram in slot #15, 1866 for example as 16th parameter if the first 15 ones have at most 8 1867 bytes. The double or vector was wrongly passed in floating-point 1868 register %d32 in lieu of on the stack as per the SPARC calling 1869 conventions. 1870 1871 Operating Systems 1872 1873 AIX 1874 1875 * DWARF debugging support for AIX 7.1 has been enabled as an optional 1876 debugging format. A more recent Technology Level (TL) and GCC built 1877 with that level are required for full exploitation of DWARF 1878 debugging capabilities. 1879 1880 Linux 1881 1882 * Support for the [21]musl C library was added for the AArch64, ARM, 1883 MicroBlaze, MIPS, MIPS64, PowerPC, PowerPC64, SH, i386, x32 and 1884 x86_64 targets. It can be selected using the new -mmusl option in 1885 case musl is not the default libc. GCC defaults to musl libc if it 1886 is built with a target triplet matching the *-linux-musl* pattern. 1887 1888 RTEMS 1889 1890 * The RTEMS thread model implementation changed. Mutexes now use 1891 self-contained objects defined in Newlib <sys/lock.h> instead of 1892 Classic API semaphores. The keys for thread specific data and the 1893 once function are directly defined via <pthread.h>. Self-contained 1894 condition variables are provided via Newlib <sys/lock.h>. The RTEMS 1895 thread model also supports C++11 threads. 1896 * OpenMP support now uses self-contained objects provided by Newlib 1897 <sys/lock.h> and offers a significantly better performance compared 1898 to the POSIX configuration of libgomp. It is possible to configure 1899 thread pools for each scheduler instance via the environment 1900 variable GOMP_RTEMS_THREAD_POOLS. 1901 1902 Solaris 1903 1904 * Solaris 12 is now fully supported. Minimal support had already been 1905 present in GCC 5.3. 1906 * Solaris 12 provides a full set of startup files (crt1.o, crti.o, 1907 crtn.o), which GCC now prefers over its own ones. 1908 * Position independent executables (PIE) are now supported on Solaris 1909 12. 1910 * Constructor priority is now supported on Solaris 12 with the system 1911 linker. 1912 * libvtv has been ported to Solaris 11 and up. 1913 1914 Windows 1915 1916 * The option -mstackrealign is now automatically activated in 32-bit 1917 mode whenever the use of SSE instructions is requested. 1918 1919 Other significant improvements 1920 1921 * The gcc and g++ driver programs will now provide suggestions for 1922 misspelled command-line options. 1923 1924 $ gcc -static-libfortran test.f95 1925 gcc: error: unrecognized command line option '-static-libfortran'; did you mean 1926 '-static-libgfortran'? 1927 1928 * The --enable-default-pie configure option enables generation of PIE 1929 by default. 1930 1931 GCC 6.2 1932 1933 This is the [22]list of problem reports (PRs) from GCC's bug tracking 1934 system that are known to be fixed in the 6.2 release. This list might 1935 not be complete (that is, it is possible that some PRs that have been 1936 fixed are not listed here). 1937 1938 Target Specific Changes 1939 1940 SPARC 1941 1942 * Support for --with-cpu-32 and --with-cpu-64 configure options has 1943 been added on bi-architecture platforms. 1944 * Support for the SPARC M7 (Niagara 7) processor has been added. 1945 * Support for the VIS 4.0 instruction set has been added. 1946 1947 GCC 6.3 1948 1949 This is the [23]list of problem reports (PRs) from GCC's bug tracking 1950 system that are known to be fixed in the 6.3 release. This list might 1951 not be complete (that is, it is possible that some PRs that have been 1952 fixed are not listed here). 1953 1954 Target Specific Changes 1955 1956 IA-32/x86-64 1957 1958 * Support for the [24]deprecated pcommit instruction has been 1959 removed. 1960 1961 GCC 6.4 1962 1963 This is the [25]list of problem reports (PRs) from GCC's bug tracking 1964 system that are known to be fixed in the 6.4 release. This list might 1965 not be complete (that is, it is possible that some PRs that have been 1966 fixed are not listed here). 1967 1968 Operating Systems 1969 1970 RTEMS 1971 1972 * The ABI changes on ARM so that no short enums are used by default. 1973 1974 GCC 6.5 1975 1976 This is the [26]list of problem reports (PRs) from GCC's bug tracking 1977 system that are known to be fixed in the 6.5 release. This list might 1978 not be complete (that is, it is possible that some PRs that have been 1979 fixed are not listed here). 1980 1981 1982 For questions related to the use of GCC, please consult these web 1983 pages and the [27]GCC manuals. If that fails, the 1984 [28]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 1985 web pages and the development of GCC are welcome on our developer 1986 list at [29]gcc (a] gcc.gnu.org. All of [30]our lists have public 1987 archives. 1988 1989 Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and 1990 distribution of this entire article is permitted in any medium, 1991 provided this notice is preserved. 1992 1993 These pages are [32]maintained by the GCC team. Last modified 1994 2018-10-31[33]. 1995 1996 References 1997 1998 1. http://gcc.gnu.org/gcc-6/porting_to.html 1999 2. http://gcc.gnu.org/onlinedocs/index.html#current 2000 3. https://gcc.gnu.org/ml/gcc/2015-08/msg00101.html 2001 4. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71151 2002 5. https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87822 2003 6. https://www.openacc.org/ 2004 7. https://gcc.gnu.org/wiki/OpenACC 2005 8. https://gcc.gnu.org/wiki/Offloading 2006 9. https://www.openmp.org/specifications/ 2007 10. https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2014-1266 2008 11. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2015/n4377.pdf 2009 12. https://gcc.gnu.org/projects/cxx-status.html#cxx1z 2010 13. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/performance.html 2011 14. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/functions.html#gcc_jit_block_end_with_switch 2012 15. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_set_bool_allow_unreachable_blocks 2013 16. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/jit/topics/contexts.html#gcc_jit_context_add_command_line_option 2014 17. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/AArch64-Function-Attributes.html#AArch64-Function-Attributes 2015 18. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/ARM-Function-Attributes.html#ARM-Function-Attributes 2016 19. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/Named-Address-Spaces.html#Named-Address-Spaces 2017 20. https://gcc.gnu.org/onlinedocs/gcc-6.1.0/gcc/S_002f390-Function-Attributes.html#S_002f390-Function-Attributes 2018 21. http://www.musl-libc.org/ 2019 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.2 2020 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.3 2021 24. https://software.intel.com/en-us/blogs/2016/09/12/deprecate-pcommit-instruction 2022 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.4 2023 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=6.5 2024 27. https://gcc.gnu.org/onlinedocs/ 2025 28. mailto:gcc-help (a] gcc.gnu.org 2026 29. mailto:gcc (a] gcc.gnu.org 2027 30. https://gcc.gnu.org/lists.html 2028 31. https://www.fsf.org/ 2029 32. https://gcc.gnu.org/about.html 2030 33. http://validator.w3.org/check/referer 2031 ====================================================================== 2032 http://gcc.gnu.org/gcc-5/index.html 2033 2034 GCC 5 Release Series 2035 2036 October 10, 2017 2037 2038 The [1]GNU project and the GCC developers are pleased to announce the 2039 release of GCC 5.5. 2040 2041 This release is a bug-fix release, containing fixes for regressions in 2042 GCC 5.4 relative to previous releases of GCC. 2043 2044 Release History 2045 2046 GCC 5.5 2047 October 10, 2017 ([2]changes, [3]documentation) 2048 2049 GCC 5.4 2050 June 3, 2016 ([4]changes, [5]documentation) 2051 2052 GCC 5.3 2053 December 4, 2015 ([6]changes, [7]documentation) 2054 2055 GCC 5.2 2056 July 16, 2015 ([8]changes, [9]documentation) 2057 2058 GCC 5.1 2059 April 22, 2015 ([10]changes, [11]documentation) 2060 2061 References and Acknowledgements 2062 2063 GCC used to stand for the GNU C Compiler, but since the compiler 2064 supports several other languages aside from C, it now stands for the 2065 GNU Compiler Collection. 2066 2067 A list of [12]successful builds is updated as new information becomes 2068 available. 2069 2070 The GCC developers would like to thank the numerous people that have 2071 contributed new features, improvements, bug fixes, and other changes as 2072 well as test results to GCC. This [13]amazing group of volunteers is 2073 what makes GCC successful. 2074 2075 For additional information about GCC please refer to the [14]GCC 2076 project web site or contact the [15]GCC development mailing list. 2077 2078 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 2079 2080 2081 For questions related to the use of GCC, please consult these web 2082 pages and the [18]GCC manuals. If that fails, the 2083 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2084 web pages and the development of GCC are welcome on our developer 2085 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 2086 archives. 2087 2088 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 2089 distribution of this entire article is permitted in any medium, 2090 provided this notice is preserved. 2091 2092 These pages are [23]maintained by the GCC team. Last modified 2093 2018-09-30[24]. 2094 2095 References 2096 2097 1. http://www.gnu.org/ 2098 2. http://gcc.gnu.org/gcc-5/changes.html 2099 3. http://gcc.gnu.org/onlinedocs/5.5.0/ 2100 4. http://gcc.gnu.org/gcc-5/changes.html 2101 5. http://gcc.gnu.org/onlinedocs/5.4.0/ 2102 6. http://gcc.gnu.org/gcc-5/changes.html 2103 7. http://gcc.gnu.org/onlinedocs/5.3.0/ 2104 8. http://gcc.gnu.org/gcc-5/changes.html 2105 9. http://gcc.gnu.org/onlinedocs/5.2.0/ 2106 10. http://gcc.gnu.org/gcc-5/changes.html 2107 11. http://gcc.gnu.org/onlinedocs/5.1.0/ 2108 12. http://gcc.gnu.org/gcc-5/buildstat.html 2109 13. http://gcc.gnu.org/onlinedocs/gcc/Contributors.html 2110 14. http://gcc.gnu.org/index.html 2111 15. mailto:gcc (a] gcc.gnu.org 2112 16. http://gcc.gnu.org/mirrors.html 2113 17. http://gcc.gnu.org/svn.html 2114 18. https://gcc.gnu.org/onlinedocs/ 2115 19. mailto:gcc-help (a] gcc.gnu.org 2116 20. mailto:gcc (a] gcc.gnu.org 2117 21. https://gcc.gnu.org/lists.html 2118 22. https://www.fsf.org/ 2119 23. https://gcc.gnu.org/about.html 2120 24. http://validator.w3.org/check/referer 2121 ====================================================================== 2122 http://gcc.gnu.org/gcc-5/changes.html 2123 2124 GCC 5 Release Series 2125 Changes, New Features, and Fixes 2126 2127 Caveats 2128 2129 * The default mode for C is now -std=gnu11 instead of -std=gnu89. 2130 * The C++ runtime library (libstdc++) uses a new ABI by default (see 2131 [1]below). 2132 * The Graphite framework for loop optimizations no longer requires 2133 the CLooG library, only ISL version 0.14 (recommended) or 0.12.2. 2134 The installation manual contains more information about 2135 requirements to build GCC. 2136 * The non-standard C++0x type traits has_trivial_default_constructor, 2137 has_trivial_copy_constructor and has_trivial_copy_assign have been 2138 deprecated and will be removed in a future version. The standard 2139 C++11 traits is_trivially_default_constructible, 2140 is_trivially_copy_constructible and is_trivially_copy_assignable 2141 should be used instead. 2142 * On AVR, support has been added for the devices 2143 ATtiny4/5/9/10/20/40. This requires Binutils 2.25 or newer. 2144 * The AVR port uses a new scheme to describe supported devices: For 2145 each supported device the compiler provides a device-specific 2146 [2]spec file. If the compiler is used together with AVR-LibC, this 2147 requires at least GCC 5.2 and a version of AVR-LibC which 2148 implements [3]feature #44574. 2149 2150 General Optimizer Improvements 2151 2152 * Inter-procedural optimization improvements: 2153 + An Identical Code Folding (ICF) pass (controlled via 2154 -fipa-icf) has been added. Compared to the identical code 2155 folding performed by the Gold linker this pass does not 2156 require function sections. It also performs merging before 2157 inlining, so inter-procedural optimizations are aware of the 2158 code re-use. On the other hand not all unifications performed 2159 by a linker are doable by GCC which must honor aliasing 2160 information. During link-time optimization of Firefox, this 2161 pass unifies about 31000 functions, that is 14% overall. 2162 + The devirtualization pass was significantly improved by adding 2163 better support for speculative devirtualization and dynamic 2164 type detection. About 50% of virtual calls in Firefox are now 2165 speculatively devirtualized during link-time optimization. 2166 + A new comdat localization pass allows the linker to eliminate 2167 more dead code in presence of C++ inline functions. 2168 + Virtual tables are now optimized. Local aliases are used to 2169 reduce dynamic linking time of C++ virtual tables on ELF 2170 targets and data alignment has been reduced to limit data 2171 segment bloat. 2172 + A new -fno-semantic-interposition option can be used to 2173 improve code quality of shared libraries where interposition 2174 of exported symbols is not allowed. 2175 + Write-only variables are now detected and optimized out. 2176 + With profile feedback the function inliner can now bypass 2177 --param inline-insns-auto and --param inline-insns-single 2178 limits for hot calls. 2179 + The IPA reference pass was significantly sped up making it 2180 feasible to enable -fipa-reference with -fprofile-generate. 2181 This also solves a bottleneck seen when building Chromium with 2182 link-time optimization. 2183 + The symbol table and call-graph API was reworked to C++ and 2184 simplified. 2185 + The interprocedural propagation of constants now also 2186 propagates alignments of pointer parameters. This for example 2187 means that the vectorizer often does not need to generate loop 2188 prologues and epilogues to make up for potential 2189 misalignments. 2190 * Link-time optimization improvements: 2191 + One Definition Rule based merging of C++ types has been 2192 implemented. Type merging enables better devirtualization and 2193 alias analysis. Streaming extra information needed to merge 2194 types adds about 2-6% of memory size and object size increase. 2195 This can be controlled by -flto-odr-type-merging. 2196 + Command-line optimization and target options are now streamed 2197 on a per-function basis and honored by the link-time 2198 optimizer. This change makes link-time optimization a more 2199 transparent replacement of per-file optimizations. It is now 2200 possible to build projects that require different optimization 2201 settings for different translation units (such as -ffast-math, 2202 -mavx, or -finline). Contrary to earlier GCC releases, the 2203 optimization and target options passed on the link command 2204 line are ignored. 2205 Note that this applies only to those command-line options that 2206 can be passed to optimize and target attributes. Command-line 2207 options affecting global code generation (such as -fpic), 2208 warnings (such as -Wodr), optimizations affecting the way 2209 static variables are optimized (such as -fcommon), debug 2210 output (such as -g), and --param parameters can be applied 2211 only to the whole link-time optimization unit. In these cases, 2212 it is recommended to consistently use the same options at both 2213 compile time and link time. 2214 + GCC bootstrap now uses slim LTO object files. 2215 + Memory usage and link times were improved. Tree merging was 2216 sped up, memory usage of GIMPLE declarations and types was 2217 reduced, and, support for on-demand streaming of variable 2218 constructors was added. 2219 * Feedback directed optimization improvements: 2220 + A new auto-FDO mode uses profiles collected by low overhead 2221 profiling tools (perf) instead of more expensive program 2222 instrumentation (via -fprofile-generate). SPEC2006 benchmarks 2223 on x86-64 improve by 4.7% with auto-FDO and by 7.3% with 2224 traditional feedback directed optimization. 2225 + Profile precision was improved in presence of C++ inline and 2226 extern inline functions. 2227 + The new gcov-tool utility allows manipulating profiles. 2228 + Profiles are now more tolerant to source file changes (this 2229 can be controlled by --param profile-func-internal-id). 2230 * Register allocation improvements: 2231 + A new local register allocator (LRA) sub-pass, controlled by 2232 -flra-remat, implements control-flow sensitive global register 2233 rematerialization. Instead of spilling and restoring a 2234 register value, it is recalculated if it is profitable. The 2235 sub-pass improved SPEC2000 generated code by 1% and 0.5% 2236 correspondingly on ARM and x86-64. 2237 + Reuse of the PIC hard register, instead of using a fixed 2238 register, was implemented on x86/x86-64 targets. This improves 2239 generated PIC code performance as more hard registers can be 2240 used. Shared libraries can significantly benefit from this 2241 optimization. Currently it is switched on only for x86/x86-64 2242 targets. As RA infrastructure is already implemented for PIC 2243 register reuse, other targets might follow this in the future. 2244 + A simple form of inter-procedural RA was implemented. When it 2245 is known that a called function does not use caller-saved 2246 registers, save/restore code is not generated around the call 2247 for such registers. This optimization can be controlled by 2248 -fipa-ra 2249 + LRA is now much more effective at generating spills of general 2250 registers into vector registers instead of memory on 2251 architectures (e.g., modern Intel processors) where this is 2252 profitable. 2253 * UndefinedBehaviorSanitizer gained a few new sanitization options: 2254 + -fsanitize=float-divide-by-zero: detect floating-point 2255 division by zero; 2256 + -fsanitize=float-cast-overflow: check that the result of 2257 floating-point type to integer conversions do not overflow; 2258 + -fsanitize=bounds: enable instrumentation of array bounds and 2259 detect out-of-bounds accesses; 2260 + -fsanitize=alignment: enable alignment checking, detect 2261 various misaligned objects; 2262 + -fsanitize=object-size: enable object size checking, detect 2263 various out-of-bounds accesses. 2264 + -fsanitize=vptr: enable checking of C++ member function calls, 2265 member accesses and some conversions between pointers to base 2266 and derived classes, detect if the referenced object does not 2267 have the correct dynamic type. 2268 * Pointer Bounds Checker, a bounds violation detector, has been added 2269 and can be enabled via -fcheck-pointer-bounds. Memory accesses are 2270 instrumented with run-time checks of used pointers against their 2271 bounds to detect pointer bounds violations (overflows). The Pointer 2272 Bounds Checker is available on x86/x86-64 GNU/Linux targets with a 2273 new ISA extension Intel MPX support. See the Pointer Bounds Checker 2274 [4]Wiki page for more details. 2275 2276 New Languages and Language specific improvements 2277 2278 * [5]OpenMP 4.0 specification offloading features are now supported 2279 by the C, C++, and Fortran compilers. Generic changes: 2280 + Infrastructure (suitable for any vendor). 2281 + Testsuite which covers offloading from the [6]OpenMP 4.0 2282 Examples document. 2283 Specific for upcoming Intel Xeon Phi products: 2284 + Run-time library. 2285 + Card emulator. 2286 * GCC 5 includes a preliminary implementation of the OpenACC 2.0a 2287 specification. OpenACC is intended for programming accelerator 2288 devices such as GPUs. See [7]the OpenACC wiki page for more 2289 information. 2290 2291 C family 2292 2293 * The default setting of the -fdiagnostics-color= command-line option 2294 is now [8]configurable when building GCC using configuration option 2295 --with-diagnostics-color=. The possible values are: never, always, 2296 auto and auto-if-env. The new default auto uses color only when the 2297 standard error is a terminal. The default in GCC 4.9 was 2298 auto-if-env, which is equivalent to auto if there is a non-empty 2299 GCC_COLORS environment variable, and never otherwise. As in GCC 2300 4.9, an empty GCC_COLORS variable in the environment will always 2301 disable colors, no matter what the default is or what command-line 2302 options are used. 2303 * A new command-line option -Wswitch-bool has been added for the C 2304 and C++ compilers, which warns whenever a switch statement has an 2305 index of boolean type. 2306 * A new command-line option -Wlogical-not-parentheses has been added 2307 for the C and C++ compilers, which warns about "logical not" used 2308 on the left hand side operand of a comparison. 2309 * A new command-line option -Wsizeof-array-argument has been added 2310 for the C and C++ compilers, which warns when the sizeof operator 2311 is applied to a parameter that has been declared as an array in a 2312 function definition. 2313 * A new command-line option -Wbool-compare has been added for the C 2314 and C++ compilers, which warns about boolean expressions compared 2315 with an integer value different from true/false. 2316 * Full support for [9]Cilk Plus has been added to the GCC compiler. 2317 Cilk Plus is an extension to the C and C++ languages to support 2318 data and task parallelism. 2319 * A new attribute no_reorder prevents reordering of selected symbols 2320 against other such symbols or inline assembler. This enables to 2321 link-time optimize the Linux kernel without having to resort to 2322 -fno-toplevel-reorder that disables several optimizations. 2323 * New preprocessor constructs, __has_include and __has_include_next, 2324 to test the availability of headers have been added. 2325 This demonstrates a way to include the header <optional> only if it 2326 is available: 2327 2328 #ifdef __has_include 2329 # if __has_include(<optional>) 2330 # include <optional> 2331 # define have_optional 1 2332 # elif __has_include(<experimental/optional>) 2333 # include <experimental/optional> 2334 # define have_optional 1 2335 # define experimental_optional 2336 # else 2337 # define have_optional 0 2338 # endif 2339 #endif 2340 2341 The header search paths for __has_include and __has_include_next 2342 are equivalent to those of the standard directive #include and the 2343 extension #include_next respectively. 2344 * A new built-in function-like macro to determine the existence of an 2345 attribute, __has_attribute, has been added. The equivalent built-in 2346 macro __has_cpp_attribute was added to C++ to support 2347 [10]Feature-testing recommendations for C++. The macro 2348 __has_attribute is added to all C-like languages as an extension: 2349 2350 int 2351 #ifdef __has_attribute 2352 # if __has_attribute(__noinline__) 2353 __attribute__((__noinline__)) 2354 # endif 2355 #endif 2356 foo(int x); 2357 2358 If an attribute exists, a nonzero constant integer is returned. For 2359 standardized C++ attributes a date is returned, otherwise the 2360 constant returned is 1. Both __has_attribute and 2361 __has_cpp_attribute will add underscores to an attribute name if 2362 necessary to resolve the name. For C++11 and onwards the attribute 2363 may be scoped. 2364 * A new set of built-in functions for arithmetics with overflow 2365 checking has been added: __builtin_add_overflow, 2366 __builtin_sub_overflow and __builtin_mul_overflow and for 2367 compatibility with clang also other variants. These builtins have 2368 two integral arguments (which don't need to have the same type), 2369 the arguments are extended to infinite precision signed type, +, - 2370 or * is performed on those, and the result is stored in an integer 2371 variable pointed to by the last argument. If the stored value is 2372 equal to the infinite precision result, the built-in functions 2373 return false, otherwise true. The type of the integer variable that 2374 will hold the result can be different from the types of the first 2375 two arguments. The following snippet demonstrates how this can be 2376 used in computing the size for the calloc function: 2377 2378 void * 2379 calloc (size_t x, size_t y) 2380 { 2381 size_t sz; 2382 if (__builtin_mul_overflow (x, y, &sz)) 2383 return NULL; 2384 void *ret = malloc (sz); 2385 if (ret) memset (res, 0, sz); 2386 return ret; 2387 } 2388 2389 On e.g. i?86 or x86-64 the above will result in a mul instruction 2390 followed by a jump on overflow. 2391 * The option -fextended-identifiers is now enabled by default for 2392 C++, and for C99 and later C versions. Various bugs in the 2393 implementation of extended identifiers have been fixed. 2394 2395 C 2396 2397 * The default mode has been changed to -std=gnu11. 2398 * A new command-line option -Wc90-c99-compat has been added to warn 2399 about features not present in ISO C90, but present in ISO C99. 2400 * A new command-line option -Wc99-c11-compat has been added to warn 2401 about features not present in ISO C99, but present in ISO C11. 2402 * It is possible to disable warnings about conversions between 2403 pointers that have incompatible types via a new warning option 2404 -Wno-incompatible-pointer-types; warnings about implicit 2405 incompatible integer to pointer and pointer to integer conversions 2406 via a new warning option -Wno-int-conversion; and warnings about 2407 qualifiers on pointers being discarded via a new warning option 2408 -Wno-discarded-qualifiers. 2409 * To allow proper use of const qualifiers with multidimensional 2410 arrays, GCC will not warn about incompatible pointer types anymore 2411 for conversions between pointers to arrays with and without const 2412 qualifier (except when using -pedantic). Instead, a new warning is 2413 emitted only if the const qualifier is lost. This can be controlled 2414 with a new warning option -Wno-discarded-array-qualifiers. 2415 * The C front end now generates more precise caret diagnostics. 2416 * The -pg command-line option now only affects the current file in an 2417 LTO build. 2418 2419 C++ 2420 2421 * G++ now supports [11]C++14 variable templates. 2422 * -Wnon-virtual-dtor doesn't warn anymore for final classes. 2423 * Excessive template instantiation depth is now a fatal error. This 2424 prevents excessive diagnostics that usually do not help to identify 2425 the problem. 2426 * G++ and libstdc++ now implement the feature-testing macros from 2427 [12]Feature-testing recommendations for C++. 2428 * G++ now allows typename in a template template parameter. 2429 2430 template<template<typename> typename X> struct D; // OK 2431 2432 * G++ now supports [13]C++14 aggregates with non-static data member 2433 initializers. 2434 2435 struct A { int i, j = i; }; 2436 A a = { 42 }; // a.j is also 42 2437 2438 * G++ now supports [14]C++14 extended constexpr. 2439 2440 constexpr int f (int i) 2441 { 2442 int j = 0; 2443 for (; i > 0; --i) 2444 ++j; 2445 return j; 2446 } 2447 2448 constexpr int i = f(42); // i is 42 2449 2450 * G++ now supports the [15]C++14 sized deallocation functions. 2451 2452 void operator delete (void *, std::size_t) noexcept; 2453 void operator delete[] (void *, std::size_t) noexcept; 2454 2455 * A new One Definition Rule violation warning (controlled by -Wodr) 2456 detects mismatches in type definitions and virtual table contents 2457 during link-time optimization. 2458 * New warnings -Wsuggest-final-types and -Wsuggest-final-methods help 2459 developers to annotate programs with final specifiers (or anonymous 2460 namespaces) to improve code generation. These warnings can be used 2461 at compile time, but they are more useful in combination with 2462 link-time optimization. 2463 * G++ no longer supports [16]N3639 variable length arrays, as they 2464 were removed from the C++14 working paper prior to ratification. 2465 GNU VLAs are still supported, so VLA support is now the same in 2466 C++14 mode as in C++98 and C++11 modes. 2467 * G++ now allows passing a non-trivially-copyable class via C 2468 varargs, which is conditionally-supported with 2469 implementation-defined semantics in the standard. This uses the 2470 same calling convention as a normal value parameter. 2471 * G++ now defaults to -fabi-version=9 and -fabi-compat-version=2. So 2472 various mangling bugs are fixed, but G++ will still emit aliases 2473 with the old, wrong mangling where feasible. -Wabi=2 will warn 2474 about differences between ABI version 2 and the current setting. 2475 * G++ 5.2 fixes the alignment of std::nullptr_t. Most code is likely 2476 to be unaffected, but -Wabi=8 will warn about a non-static data 2477 member with type std::nullptr_t which changes position due to this 2478 change. 2479 2480 Runtime Library (libstdc++) 2481 2482 * A [17]Dual ABI is provided by the library. A new ABI is enabled by 2483 default. The old ABI is still supported and can be used by defining 2484 the macro _GLIBCXX_USE_CXX11_ABI to 0 before including any C++ 2485 standard library headers. 2486 * A new implementation of std::string is enabled by default, using 2487 the small string optimization instead of copy-on-write reference 2488 counting. 2489 * A new implementation of std::list is enabled by default, with an 2490 O(1) size() function; 2491 * [18]Full support for C++11, including the following new features: 2492 + std::deque and std::vector<bool> meet the allocator-aware 2493 container requirements; 2494 + movable and swappable iostream classes; 2495 + support for std::align and std::aligned_union; 2496 + type traits std::is_trivially_copyable, 2497 std::is_trivially_constructible, std::is_trivially_assignable 2498 etc.; 2499 + I/O manipulators std::put_time, std::get_time, std::hexfloat 2500 and std::defaultfloat; 2501 + generic locale-aware std::isblank; 2502 + locale facets for Unicode conversion; 2503 + atomic operations for std::shared_ptr; 2504 + std::notify_all_at_thread_exit() and functions for making 2505 futures ready at thread exit. 2506 * Support for the C++11 hexfloat manipulator changes how the num_put 2507 facet formats floating point types when 2508 ios_base::fixed|ios_base::scientific is set in a stream's fmtflags. 2509 This change affects all language modes, even though the C++98 2510 standard gave no special meaning to that combination of flags. To 2511 prevent the use of hexadecimal notation for floating point types 2512 use str.unsetf(std::ios_base::floatfield) to clear the relevant 2513 bits in str.flags(). 2514 * [19]Full experimental support for C++14, including the following 2515 new features: 2516 + std::is_final type trait; 2517 + heterogeneous comparison lookup in associative containers. 2518 + global functions cbegin, cend, rbegin, rend, crbegin, and 2519 crend for range access to containers, arrays and initializer 2520 lists. 2521 * [20]Improved experimental support for the Library Fundamentals TS, 2522 including: 2523 + class std::experimental::any; 2524 + function template std::experimental::apply; 2525 + function template std::experimental::sample; 2526 + function template std::experimental::search and related 2527 searcher types; 2528 + variable templates for type traits; 2529 + function template std::experimental::not_fn. 2530 * New random number distributions logistic_distribution and 2531 uniform_on_sphere_distribution as extensions. 2532 * [21]GDB Xmethods for containers and std::unique_ptr. 2533 2534 Fortran 2535 2536 * Compatibility notice: 2537 + The version of the module files (.mod) has been incremented. 2538 + For free-form source files [22]-Werror=line-truncation is now 2539 enabled by default. Note that comments exceeding the line 2540 length are not diagnosed. (For fixed-form source code, the 2541 same warning is available but turned off by default, such that 2542 excess characters are ignored. -ffree-line-length-n and 2543 -ffixed-line-length-n can be used to modify the default line 2544 lengths of 132 and 72 columns, respectively.) 2545 + The -Wtabs option is now more sensible: with -Wtabs the 2546 compiler warns if it encounters tabs and with -Wno-tabs this 2547 warning is turned off. Before, -Wno-tabs warned and -Wtabs 2548 disabled the warning. As before, this warning is also enabled 2549 by -Wall, -pedantic and the f95, f2003, f2008 and f2008ts 2550 options of -std=. 2551 * Incomplete support for colorizing diagnostics emitted by gfortran 2552 has been added. The option [23]-fdiagnostics-color controls when 2553 color is used in diagnostics. The default value of this option can 2554 be [24]configured when building GCC. The GCC_COLORS environment 2555 variable can be used to customize the colors or disable coloring 2556 completely. Sample diagnostics output: 2557 $ gfortran -fdiagnostics-color=always -Wuse-without-only test.f90 2558 test.f90:6:1: 2559 2560 0 continue 2561 1 2562 Error: Zero is not a valid statement label at (1) 2563 test.f90:9:6: 2564 2565 USE foo 2566 1 2567 Warning: USE statement at (1) has no ONLY qualifier [-Wuse-without-only] 2568 2569 * The -Wuse-without-only option has been added to warn when a USE 2570 statement has no ONLY qualifier and thus implicitly imports all 2571 public entities of the used module. 2572 * Formatted READ and WRITE statements now work correctly in 2573 locale-aware programs. For more information and potential caveats, 2574 see [25]Section 5.3 Thread-safety of the runtime library in the 2575 manual. 2576 * [26]Fortran 2003: 2577 + The intrinsic IEEE modules (IEEE_FEATURES, IEEE_EXCEPTIONS and 2578 IEEE_ARITHMETIC) are now supported. 2579 * [27]Fortran 2008: 2580 + [28]Coarrays: Full experimental support of Fortran 2008's 2581 coarrays with -fcoarray=lib except for allocatable/pointer 2582 components of derived-type coarrays. GCC currently only ships 2583 with a single-image library (libcaf_single), but multi-image 2584 support based on MPI and GASNet is provided by the libraries 2585 of the [29]OpenCoarrays project. 2586 * TS18508 Additional Parallel Features in Fortran: 2587 + Support for the collective intrinsic subroutines CO_MAX, 2588 CO_MIN, CO_SUM, CO_BROADCAST and CO_REDUCE has been added, 2589 including -fcoarray=lib support. 2590 + Support for the new atomic intrinsics has been added, 2591 including -fcoarray=lib support. 2592 * Fortran 2015: 2593 + Support for IMPLICIT NONE (external, type). 2594 + ERROR STOP is now permitted in pure procedures. 2595 2596 Go 2597 2598 * GCC 5 provides a complete implementation of the Go 1.4.2 release. 2599 * Building GCC 5 with Go enabled will install two new programs: 2600 [30]go and [31]gofmt. 2601 2602 libgccjit 2603 2604 New in GCC 5 is the ability to build GCC as a shared library for 2605 embedding in other processes (such as interpreters), suitable for 2606 Just-In-Time compilation to machine code. 2607 2608 The shared library has a [32]C API and a [33]C++ wrapper API providing 2609 some "syntactic sugar". There are also bindings available from 3rd 2610 parties for [34]Python and for [35]D. 2611 2612 For example, this library can be used by interpreters for [36]compiling 2613 functions from bytecode to machine code. 2614 2615 The library can also be used for ahead-of-time compilation, enabling 2616 GCC to be plugged into a pre-existing front end. An example of using 2617 this to build a compiler for an esoteric language we'll refer to as 2618 "brainf" can be seen [37]here. 2619 2620 libgccjit is licensed under the GPLv3 (or at your option, any later 2621 version) 2622 2623 It should be regarded as experimental at this time. 2624 2625 New Targets and Target Specific Improvements 2626 2627 Reporting stack usage 2628 2629 * The BFIN, FT32, H8300, IQ2000 and M32C targets now support the 2630 -fstack-usage option. 2631 2632 AArch64 2633 2634 * Code generation for the ARM Cortex-A57 processor has been improved. 2635 A more accurate instruction scheduling model for the processor is 2636 now used, and a number of compiler tuning parameters have been set 2637 to offer increased performance when compiling with -mcpu=cortex-a57 2638 or -mtune=cortex-a57. 2639 * A workaround for the ARM Cortex-A53 erratum 835769 has been added 2640 and can be enabled by giving the -mfix-cortex-a53-835769 option. 2641 Alternatively it can be enabled by default by configuring GCC with 2642 the --enable-fix-cortex-a53-835769 option. 2643 * The optional cryptographic extensions to the ARMv8-A architecture 2644 are no longer enabled by default when specifying the 2645 -mcpu=cortex-a53, -mcpu=cortex-a57 or -mcpu=cortex-a57.cortex-a53 2646 options. To enable these extensions add +crypto to the value of 2647 -mcpu or -march e.g. -mcpu=cortex-a53+crypto. 2648 * Support has been added for the following processors (GCC 2649 identifiers in parentheses): ARM Cortex-A72 (cortex-a72) and 2650 initial support for its big.LITTLE combination with the ARM 2651 Cortex-A53 (cortex-a72.cortex-a53), Cavium ThunderX (thunderx), 2652 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 2653 The GCC identifiers can be used as arguments to the -mcpu or -mtune 2654 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 2655 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 2656 support for the Cortex-A72. 2657 * The transitional options -mlra and -mno-lra have been removed. The 2658 AArch64 backend now uses the local register allocator (LRA) only. 2659 2660 ARM 2661 2662 * Thumb-1 assembly code is now generated in unified syntax. The new 2663 option -masm-syntax-unified specifies whether inline assembly code 2664 is using unified syntax. By default the option is off which means 2665 non-unified syntax is used. However this is subject to change in 2666 future releases. Eventually the non-unified syntax will be 2667 deprecated. 2668 * It is now a configure-time error to use the --with-cpu configure 2669 option with either of --with-tune or --with-arch. 2670 * Code generation for the ARM Cortex-A57 processor has been improved. 2671 A more accurate instruction scheduling model for the processor is 2672 now used, and a number of compiler tuning parameters have been set 2673 to offer increased performance when compiling with -mcpu=cortex-a57 2674 or -mtune=cortex-a57. 2675 * Support has been added for the following processors (GCC 2676 identifiers in parentheses): ARM Cortex-A17 (cortex-a17) and 2677 initial support for its big.LITTLE combination with the ARM 2678 Cortex-A7 (cortex-a17.cortex-a7), ARM Cortex-A72 (cortex-a72) and 2679 initial support for its big.LITTLE combination with the ARM 2680 Cortex-A53 (cortex-a72.cortex-a53), ARM Cortex-M7 (cortex-m7), 2681 Applied Micro X-Gene 1 (xgene1), and Samsung Exynos M1 (exynos-m1). 2682 The GCC identifiers can be used as arguments to the -mcpu or -mtune 2683 options, for example: -mcpu=xgene1 or -mtune=cortex-a72.cortex-a53. 2684 Using -mcpu=cortex-a72 requires a version of GNU binutils that has 2685 support for the Cortex-A72. 2686 * The deprecated option -mwords-little-endian has been removed. 2687 * The options -mapcs, -mapcs-frame, -mtpcs-frame and 2688 -mtpcs-leaf-frame which are only applicable to the old ABI have 2689 been deprecated. 2690 * The transitional options -mlra and -mno-lra have been removed. The 2691 ARM backend now uses the local register allocator (LRA) only. 2692 2693 AVR 2694 2695 * The compiler no more supports individual devices like ATmega8. 2696 Specifying, say, -mmcu=atmega8 triggers the usage of the 2697 device-specific [38]spec file specs-atmega8 which is part of the 2698 installation and describes options for the sub-processes like 2699 compiler proper, assembler and linker. You can add support for a 2700 new device -mmcu=mydevice as follows: 2701 1. In an empty directory /someplace, create a new directory 2702 device-specs. 2703 2. Copy a device spec file from the installed device-specs 2704 folder, follow the comments in that file and then save it as 2705 /someplace/device-specs/specs-mydevice. 2706 3. Add -B /someplace -mmcu=mydevice to the compiler's 2707 command-line options. Notice that /someplace must specify an 2708 absolute path and that mydevice must not start with "avr". 2709 4. Provided you have a device-specific library libmydevice.a 2710 available, you can put it at /someplace, dito for a 2711 device-specific startup file crtmydevice.o. 2712 The contents of the device spec files depend on the compiler's 2713 configuration, in particular on --with-avrlibc=no and whether or 2714 not it is configured for RTEMS. 2715 * A new command-line option -nodevicelib has been added. It prevents 2716 the compiler from linking against AVR-LibC's device-specific 2717 library libdevice.a. 2718 * The following three command-line options have been added: 2719 2720 -mrmw 2721 Set if the device supports the read-modify-write 2722 instructions LAC, LAS, LAT and XCH. 2723 2724 -mn-flash=size 2725 Specify the flash size of the device in units of 64 KiB, 2726 rounded up to the next integer as needed. This option 2727 affects the availability of the [39]AVR address-spaces. 2728 2729 -mskip-bug 2730 Set if the device is affected by the respective silicon 2731 bug. 2732 2733 In general, you don't need to set these options by hand. The new 2734 device-specific spec file will set them as needed. 2735 2736 IA-32/x86-64 2737 2738 * New [40]ISA extensions support AVX-512{BW,DQ,VL,IFMA,VBMI} of 2739 Intel's CPU codenamed Skylake Server was added to GCC. That 2740 includes inline assembly support, new intrinsics, and basic 2741 autovectorization. These new AVX-512 extensions are available via 2742 the following GCC switches: AVX-512 Vector Length EVEX feature: 2743 -mavx512vl, AVX-512 Byte and Word instructions: -mavx512bw, AVX-512 2744 Dword and Qword instructions: -mavx512dq, AVX-512 FMA-52 2745 instructions: -mavx512ifma and for AVX-512 Vector Bit Manipulation 2746 Instructions: -mavx512vbmi. 2747 * New ISA extensions support Intel MPX was added to GCC. This new 2748 extension is available via the -mmpx compiler switch. Intel MPX is 2749 a set of processor features which, with compiler, run-time library 2750 and OS support, brings increased robustness to software by run-time 2751 checking pointer references against their bounds. In GCC Intel MPX 2752 is supported by Pointer Bounds Checker and libmpx run-time 2753 libraries. 2754 * The new -mrecord-mcount option for -pg generates a Linux kernel 2755 style table of pointers to mcount or __fentry__ calls at the 2756 beginning of functions. The new -mnop-mcount option in addition 2757 also generates nops in place of the __fentry__ or mcount call, so 2758 that a call per function can be later patched in. This can be used 2759 for low overhead tracing or hot code patching. 2760 * The new -malign-data option controls how GCC aligns variables. 2761 -malign-data=compat uses increased alignment compatible with GCC 2762 4.8 and earlier, -malign-data=abi uses alignment as specified by 2763 the psABI, and -malign-data=cacheline uses increased alignment to 2764 match the cache line size. -malign-data=compat is the default. 2765 * The new -mskip-rax-setup option skips setting up the RAX register 2766 when SSE is disabled and there are no variable arguments passed in 2767 vector registers. This can be used to optimize the Linux kernel. 2768 2769 MIPS 2770 2771 * MIPS Releases 3 and 5 are now directly supported. Use the 2772 command-line options -mips32r3, -mips64r3, -mips32r5 and -mips64r5 2773 to enable code-generation for these processors. 2774 * The Imagination P5600 processor is now supported using the 2775 -march=p5600 command-line option. 2776 * The Cavium Octeon3 processor is now supported using the 2777 -march=octeon3 command-line option. 2778 * MIPS Release 6 is now supported using the -mips32r6 and -mips64r6 2779 command-line options. 2780 * The o32 ABI has been modified and extended. The o32 64-bit 2781 floating-point register support is now obsolete and has been 2782 removed. It has been replaced by three ABI extensions FPXX, FP64A, 2783 and FP64. The meaning of the -mfp64 command-line option has 2784 changed. It is now used to enable the FP64A and FP64 ABI 2785 extensions. 2786 + The FPXX extension requires that code generated to access 2787 double-precision values use even-numbered registers. Code that 2788 adheres to this extension is link-compatible with all other 2789 o32 double-precision ABI variants and will execute correctly 2790 in all hardware FPU modes. The command-line options -mabi=32 2791 -mfpxx can be used to enable this extension. MIPS II is the 2792 minimum processor required. 2793 + The o32 FP64A extension requires that floating-point registers 2794 be 64-bit and odd-numbered single-precision registers are not 2795 allowed. Code that adheres to the o32 FP64A variant is 2796 link-compatible with all other o32 double-precision ABI 2797 variants. The command-line options -mabi=32 -mfp64 2798 -mno-odd-spreg can be used to enable this extension. MIPS32R2 2799 is the minimum processor required. 2800 + The o32 FP64 extension also requires that floating-point 2801 registers be 64-bit, but permits the use of single-precision 2802 registers. Code that adheres to the o32 FP64 variant is 2803 link-compatible with o32 FPXX and o32 FP64A variants only, 2804 i.e. it is not compatible with the original o32 2805 double-precision ABI. The command-line options -mabi=32 -mfp64 2806 -modd-spreg can be used to enable this extension. MIPS32R2 is 2807 the minimum processor required. 2808 The new ABI variants can be enabled by default using the configure 2809 time options --with-fp-32=[32|xx|64] and --with(out)-odd-sp-reg-32. 2810 It is strongly recommended that all vendors begin to set o32 FPXX 2811 as the default ABI. This will be required to run the generated code 2812 on MIPSR5 cores in conjunction with future MIPS SIMD (MSA) code and 2813 MIPSR6 cores. 2814 * GCC will now pass all floating-point options to the assembler if 2815 GNU binutils 2.25 is used. As a result, any inline assembly code 2816 that uses hard-float instructions should be amended to include a 2817 .set directive to override the global assembler options when 2818 compiling for soft-float targets. 2819 2820 NDS32 2821 2822 * The variadic function ABI implementation is now compatible with 2823 past Andes toolchains where the caller uses registers to pass 2824 arguments and the callee is in charge of pushing them on stack. 2825 * The options -mforce-fp-as-gp, -mforbid-fp-as-gp, and -mex9 have 2826 been removed since they are not yet available in the nds32 port of 2827 GNU binutils. 2828 * A new option -mcmodel=[small|medium|large] supports varied code 2829 models on code generation. The -mgp-direct option became 2830 meaningless and can be discarded. 2831 2832 RX 2833 2834 * A new command line option -mno-allow-string-insns can be used to 2835 disable the generation of the SCMPU, SMOVU, SMOVB, SMOVF, SUNTIL, 2836 SWHILE and RMPA instructions. An erratum released by Renesas shows 2837 that it is unsafe to use these instructions on addresses within the 2838 I/O space of the processor. The new option can be used when the 2839 programmer is concerned that the I/O space might be accessed. The 2840 default is still to enable these instructions. 2841 2842 SH 2843 2844 * The compiler will now pass the appropriate --isa= option to the 2845 assembler. 2846 * The default handling for the GBR has been changed from call 2847 clobbered to call preserved. The old behavior can be reinstated by 2848 specifying the option -fcall-used-gbr. 2849 * Support for the SH4A fpchg instruction has been added which will be 2850 utilized when switching between single and double precision FPU 2851 modes. 2852 * The compiler no longer uses the __fpscr_values array for switching 2853 between single and double FPU precision modes on non-SH4A targets. 2854 Instead mode switching will now be performed by storing, modifying 2855 and reloading the FPSCR, so that other FPSCR bits are preserved 2856 across mode switches. The __fpscr_values array that is defined in 2857 libgcc is still present for backwards compatibility, but it will 2858 not be referenced by compiler generated code anymore. 2859 * New builtin functions __builtin_sh_get_fpscr and 2860 __builtin_sh_set_fpscr have been added. The __builtin_sh_set_fpscr 2861 function will mask the specified bits in such a way that the SZ, PR 2862 and FR mode bits will be preserved, while changing the other bits. 2863 These new functions do not reference the __fpscr_values array. The 2864 old functions __set_fpscr and __get_fpscr in libgcc which access 2865 the __fpscr_values array are still present for backwards 2866 compatibility, but their usage is highly discouraged. 2867 * Some improvements to code generated for __atomic built-in 2868 functions. 2869 * When compiling for SH2E the compiler will no longer force the usage 2870 of delay slots for conditional branch instructions bt and bf. The 2871 old behavior can be reinstated (e.g. to work around a hardware bug 2872 in the original SH7055) by specifying the new option 2873 -mcbranch-force-delay-slot. 2874 2875 Operating Systems 2876 2877 AIX 2878 2879 * GCC now supports stabs debugging continuation lines to allow long 2880 stabs debug information without overflow that generates AIX linker 2881 errors. 2882 2883 DragonFly BSD 2884 2885 * GCC now supports the DragonFly BSD operating system. 2886 2887 FreeBSD 2888 2889 * GCC now supports the FreeBSD operating system for the arm port 2890 through the arm*-*-freebsd* target triplets. 2891 2892 VxWorks MILS 2893 2894 * GCC now supports the MILS (Multiple Independent Levels of Security) 2895 variant of WindRiver's VxWorks operating system for PowerPC 2896 targets. 2897 2898 Other significant improvements 2899 2900 * The gcc-ar, gcc-nm, gcc-ranlib wrappers now understand a -B option 2901 to set the compiler to use. 2902 2903 * When the new command-line option -freport-bug is used, GCC 2904 automatically generates a developer-friendly reproducer whenever an 2905 internal compiler error is encountered. 2906 2907 GCC 5.2 2908 2909 This is the [41]list of problem reports (PRs) from GCC's bug tracking 2910 system that are known to be fixed in the 5.2 release. This list might 2911 not be complete (that is, it is possible that some PRs that have been 2912 fixed are not listed here). 2913 2914 Target Specific Changes 2915 2916 IA-32/x86-64 2917 2918 * Support for new AMD instructions monitorx and mwaitx has been 2919 added. This includes new intrinsic and built-in support. It is 2920 enabled through option -mmwaitx. The instructions monitorx and 2921 mwaitx implement the same functionality as the old monitor and 2922 mwait instructions. In addition mwaitx adds a configurable timer. 2923 The timer value is received as third argument and stored in 2924 register %ebx. 2925 2926 S/390, System z, IBM z Systems 2927 2928 * Support for the IBM z13 processor has been added. When using the 2929 -march=z13 option, the compiler will generate code making use of 2930 the new instructions and registers introduced with the vector 2931 extension facility. The -mtune=z13 option enables z13 specific 2932 instruction scheduling without making use of new instructions. 2933 Compiling code with -march=z13 reduces the default alignment of 2934 vector types bigger than 8 bytes to 8. This is an ABI change and 2935 care must be taken when linking modules compiled with different 2936 arch levels which interchange variables containing vector type 2937 values. For newly compiled code the GNU linker will emit a warning. 2938 * The -mzvector option enables a C/C++ language extension. This 2939 extension provides a new keyword vector which can be used to define 2940 vector type variables. (Note: This is not available when enforcing 2941 strict standard compliance e.g. with -std=c99. Either enable GNU 2942 extensions with e.g. -std=gnu99 or use __vector instead of vector.) 2943 Additionally a set of overloaded builtins is provided which is 2944 partially compatible to the PowerPC Altivec builtins. In order to 2945 make use of these builtins the vecintrin.h header file needs to be 2946 included. 2947 2948 GCC 5.3 2949 2950 This is the [42]list of problem reports (PRs) from GCC's bug tracking 2951 system that are known to be fixed in the 5.3 release. This list might 2952 not be complete (that is, it is possible that some PRs that have been 2953 fixed are not listed here). 2954 2955 Target Specific Changes 2956 2957 IA-32/x86-64 2958 2959 * GCC now supports the Intel CPU named Skylake with AVX-512 2960 extensions through -march=skylake-avx512. The switch enables the 2961 following ISA extensions: AVX-512F, AVX512VL, AVX-512CD, AVX-512BW, 2962 AVX-512DQ. 2963 2964 S/390, System z, IBM z Systems 2965 2966 * With this version of GCC IBM z Systems support has been added to 2967 the GO runtime environment. GCC 5.3 has proven to be able to 2968 compile larger GO applications on IBM z Systems. 2969 2970 GCC 5.4 2971 2972 This is the [43]list of problem reports (PRs) from GCC's bug tracking 2973 system that are known to be fixed in the 5.4 release. This list might 2974 not be complete (that is, it is possible that some PRs that have been 2975 fixed are not listed here). 2976 2977 GCC 5.5 2978 2979 This is the [44]list of problem reports (PRs) from GCC's bug tracking 2980 system that are known to be fixed in the 5.5 release. This list might 2981 not be complete (that is, it is possible that some PRs that have been 2982 fixed are not listed here). 2983 2984 Target Specific Changes 2985 2986 IA-32/x86-64 2987 2988 * Support for the [45]deprecated pcommit instruction has been 2989 removed. 2990 2991 2992 For questions related to the use of GCC, please consult these web 2993 pages and the [46]GCC manuals. If that fails, the 2994 [47]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 2995 web pages and the development of GCC are welcome on our developer 2996 list at [48]gcc (a] gcc.gnu.org. All of [49]our lists have public 2997 archives. 2998 2999 Copyright (C) [50]Free Software Foundation, Inc. Verbatim copying and 3000 distribution of this entire article is permitted in any medium, 3001 provided this notice is preserved. 3002 3003 These pages are [51]maintained by the GCC team. Last modified 3004 2018-09-30[52]. 3005 3006 References 3007 3008 1. http://gcc.gnu.org/gcc-5/changes.html#libstdcxx 3009 2. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 3010 3. https://savannah.nongnu.org/bugs/?44574 3011 4. https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler 3012 5. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.pdf 3013 6. https://www.openmp.org/wp-content/uploads/OpenMP4.0.0.Examples.pdf 3014 7. https://gcc.gnu.org/wiki/OpenACC 3015 8. https://gcc.gnu.org/install/configure.html 3016 9. https://www.cilkplus.org/ 3017 10. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 3018 11. http://gcc.gnu.org/projects/cxx1y.html 3019 12. https://isocpp.org/std/standing-documents/sd-6-sg10-feature-test-recommendations 3020 13. http://gcc.gnu.org/projects/cxx1y.html 3021 14. http://gcc.gnu.org/projects/cxx1y.html 3022 15. http://gcc.gnu.org/projects/cxx1y.html 3023 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3639.html 3024 17. http://gcc.gnu.org/onlinedocs/libstdc++/manual/using_dual_abi.html 3025 18. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2011 3026 19. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 3027 20. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/libstdc++/manual/manual/status.html#status.iso.2014 3028 21. https://sourceware.org/gdb/current/onlinedocs/gdb/Xmethods-In-Python.html 3029 22. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Error-and-Warning-Options.html 3030 23. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gcc/Language-Independent-Options.html 3031 24. https://gcc.gnu.org/install/configure.html 3032 25. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/gfortran/Thread-safety-of-the-runtime-library.html 3033 26. https://gcc.gnu.org/wiki/Fortran2003Status 3034 27. https://gcc.gnu.org/wiki/Fortran2008Status 3035 28. https://gcc.gnu.org/wiki/Coarray 3036 29. http://www.opencoarrays.org/ 3037 30. https://golang.org/cmd/go/ 3038 31. https://golang.org/cmd/gofmt/ 3039 32. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/index.html 3040 33. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/cp/index.html 3041 34. https://github.com/davidmalcolm/pygccjit 3042 35. https://github.com/ibuclaw/gccjitd 3043 36. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial04.html 3044 37. https://gcc.gnu.org/onlinedocs/gcc-5.1.0/jit/intro/tutorial05.html 3045 38. https://gcc.gnu.org/onlinedocs/gcc/Spec-Files.html 3046 39. https://gcc.gnu.org/onlinedocs/gcc/Named-Address-Spaces.html 3047 40. https://software.intel.com/sites/default/files/managed/39/c5/325462-sdm-vol-1-2abcd-3abcd.pdf 3048 41. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.2 3049 42. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.3 3050 43. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.4 3051 44. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=5.5 3052 45. https://software.intel.com/en-us/blogs/2016/09/12/deprecate-pcommit-instruction 3053 46. https://gcc.gnu.org/onlinedocs/ 3054 47. mailto:gcc-help (a] gcc.gnu.org 3055 48. mailto:gcc (a] gcc.gnu.org 3056 49. https://gcc.gnu.org/lists.html 3057 50. https://www.fsf.org/ 3058 51. https://gcc.gnu.org/about.html 3059 52. http://validator.w3.org/check/referer 3060 ====================================================================== 3061 http://gcc.gnu.org/gcc-4.9/index.html 3062 3063 GCC 4.9 Release Series 3064 3065 Aug 3, 2016 3066 3067 The [1]GNU project and the GCC developers are pleased to announce the 3068 release of GCC 4.9.4. 3069 3070 This release is a bug-fix release, containing fixes for regressions in 3071 GCC 4.9.3 relative to previous releases of GCC. 3072 3073 This release series is no longer maintained. 3074 3075 Release History 3076 3077 GCC 4.9.4 3078 Aug 3, 2016 ([2]changes, [3]documentation) 3079 3080 GCC 4.9.3 3081 June 26, 2015 ([4]changes, [5]documentation) 3082 3083 GCC 4.9.2 3084 October 30, 2014 ([6]changes, [7]documentation) 3085 3086 GCC 4.9.1 3087 July 16, 2014 ([8]changes, [9]documentation) 3088 3089 GCC 4.9.0 3090 April 22, 2014 ([10]changes, [11]documentation) 3091 3092 References and Acknowledgements 3093 3094 GCC used to stand for the GNU C Compiler, but since the compiler 3095 supports several other languages aside from C, it now stands for the 3096 GNU Compiler Collection. 3097 3098 A list of [12]successful builds is updated as new information becomes 3099 available. 3100 3101 The GCC developers would like to thank the numerous people that have 3102 contributed new features, improvements, bug fixes, and other changes as 3103 well as test results to GCC. This [13]amazing group of volunteers is 3104 what makes GCC successful. 3105 3106 For additional information about GCC please refer to the [14]GCC 3107 project web site or contact the [15]GCC development mailing list. 3108 3109 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 3110 3111 3112 For questions related to the use of GCC, please consult these web 3113 pages and the [18]GCC manuals. If that fails, the 3114 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3115 web pages and the development of GCC are welcome on our developer 3116 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 3117 archives. 3118 3119 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 3120 distribution of this entire article is permitted in any medium, 3121 provided this notice is preserved. 3122 3123 These pages are [23]maintained by the GCC team. Last modified 3124 2018-09-30[24]. 3125 3126 References 3127 3128 1. http://www.gnu.org/ 3129 2. http://gcc.gnu.org/gcc-4.9/changes.html 3130 3. http://gcc.gnu.org/onlinedocs/4.9.4/ 3131 4. http://gcc.gnu.org/gcc-4.9/changes.html 3132 5. http://gcc.gnu.org/onlinedocs/4.9.3/ 3133 6. http://gcc.gnu.org/gcc-4.9/changes.html 3134 7. http://gcc.gnu.org/onlinedocs/4.9.2/ 3135 8. http://gcc.gnu.org/gcc-4.9/changes.html 3136 9. http://gcc.gnu.org/onlinedocs/4.9.1/ 3137 10. http://gcc.gnu.org/gcc-4.9/changes.html 3138 11. https://gcc.gnu.org/onlinedocs/4.9.0/ 3139 12. http://gcc.gnu.org/gcc-4.9/buildstat.html 3140 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 3141 14. http://gcc.gnu.org/index.html 3142 15. mailto:gcc (a] gcc.gnu.org 3143 16. http://gcc.gnu.org/mirrors.html 3144 17. http://gcc.gnu.org/svn.html 3145 18. https://gcc.gnu.org/onlinedocs/ 3146 19. mailto:gcc-help (a] gcc.gnu.org 3147 20. mailto:gcc (a] gcc.gnu.org 3148 21. https://gcc.gnu.org/lists.html 3149 22. https://www.fsf.org/ 3150 23. https://gcc.gnu.org/about.html 3151 24. http://validator.w3.org/check/referer 3152 ====================================================================== 3153 http://gcc.gnu.org/gcc-4.9/changes.html 3154 3155 GCC 4.9 Release Series 3156 Changes, New Features, and Fixes 3157 3158 Caveats 3159 3160 * The mudflap run time checker has been removed. The mudflap options 3161 remain, but do nothing. 3162 * Support for a number of older systems and recently unmaintained or 3163 untested target ports of GCC has been declared obsolete in GCC 4.9. 3164 Unless there is activity to revive them, the next release of GCC 3165 will have their sources permanently removed. 3166 The following ports for individual systems on particular 3167 architectures have been obsoleted: 3168 + Solaris 9 (*-*-solaris2.9). Details can be found in the 3169 [1]announcement. 3170 * On AArch64, the singleton vector types int64x1_t, uint64x1_t and 3171 float64x1_t exported by arm_neon.h are defined to be the same as 3172 their base types. This results in incorrect application of 3173 parameter passing rules to arguments of types int64x1_t and 3174 uint64x1_t, with respect to the AAPCS64 ABI specification. In 3175 addition, names of C++ functions with parameters of these types 3176 (including float64x1_t) are not mangled correctly. The current 3177 typedef declarations also unintentionally allow implicit casting 3178 between singleton vector types and their base types. These issues 3179 will be resolved in a near future release. See [2]PR60825 for more 3180 information. 3181 3182 More information on porting to GCC 4.9 from previous versions of GCC 3183 can be found in the [3]porting guide for this release. 3184 3185 General Optimizer Improvements 3186 3187 * AddressSanitizer, a fast memory error detector, is now available on 3188 ARM. 3189 * UndefinedBehaviorSanitizer (ubsan), a fast undefined behavior 3190 detector, has been added and can be enabled via 3191 -fsanitize=undefined. Various computations will be instrumented to 3192 detect undefined behavior at runtime. UndefinedBehaviorSanitizer is 3193 currently available for the C and C++ languages. 3194 * Link-time optimization (LTO) improvements: 3195 + Type merging was rewritten. The new implementation is 3196 significantly faster and uses less memory. 3197 + Better partitioning algorithm resulting in less streaming 3198 during link time. 3199 + Early removal of virtual methods reduces the size of object 3200 files and improves link-time memory usage and compile time. 3201 + Function bodies are now loaded on-demand and released early 3202 improving overall memory usage at link time. 3203 + C++ hidden keyed methods can now be optimized out. 3204 + When using a linker plugin, compiling with the -flto option 3205 now generates slim object files (.o) which only contain 3206 intermediate language representation for LTO. Use 3207 -ffat-lto-objects to create files which contain additionally 3208 the object code. To generate static libraries suitable for LTO 3209 processing, use gcc-ar and gcc-ranlib; to list symbols from a 3210 slim object file use gcc-nm. (This requires that ar, ranlib 3211 and nm have been compiled with plugin support.) 3212 Memory usage building Firefox with debug enabled was reduced from 3213 15GB to 3.5GB; link time from 1700 seconds to 350 seconds. 3214 * Inter-procedural optimization improvements: 3215 + New type inheritance analysis module improving 3216 devirtualization. Devirtualization now takes into account 3217 anonymous name-spaces and the C++11 final keyword. 3218 + New speculative devirtualization pass (controlled by 3219 -fdevirtualize-speculatively. 3220 + Calls that were speculatively made direct are turned back to 3221 indirect where direct call is not cheaper. 3222 + Local aliases are introduced for symbols that are known to be 3223 semantically equivalent across shared libraries improving 3224 dynamic linking times. 3225 * Feedback directed optimization improvements: 3226 + Profiling of programs using C++ inline functions is now more 3227 reliable. 3228 + New time profiling determines typical order in which functions 3229 are executed. 3230 + A new function reordering pass (controlled by 3231 -freorder-functions) significantly reduces startup time of 3232 large applications. Until binutils support is completed, it is 3233 effective only with link-time optimization. 3234 + Feedback driven indirect call removal and devirtualization now 3235 handle cross-module calls when link-time optimization is 3236 enabled. 3237 3238 New Languages and Language specific improvements 3239 3240 * Version 4.0 of the [4]OpenMP specification is now supported in the 3241 C and C++ compilers and starting with the 4.9.1 release also in the 3242 Fortran compiler. The new -fopenmp-simd option can be used to 3243 enable OpenMP's SIMD directives while ignoring other OpenMP 3244 directives. The new [5]-fsimd-cost-model= option permits to tune 3245 the vectorization cost model for loops annotated with OpenMP and 3246 Cilk Plus simd directives. -Wopenmp-simd warns when the current 3247 cost model overrides simd directives set by the user. 3248 * The -Wdate-time option has been added for the C, C++ and Fortran 3249 compilers, which warns when the __DATE__, __TIME__ or __TIMESTAMP__ 3250 macros are used. Those macros might prevent bit-wise-identical 3251 reproducible compilations. 3252 3253 Ada 3254 3255 * GNAT switched to Ada 2012 instead of Ada 2005 by default. 3256 3257 C family 3258 3259 * Support for colorizing diagnostics emitted by GCC has been added. 3260 The [6]-fdiagnostics-color=auto will enable it when outputting to 3261 terminals, -fdiagnostics-color=always unconditionally. The 3262 GCC_COLORS environment variable can be used to customize the colors 3263 or disable coloring. If GCC_COLORS variable is present in the 3264 environment, the default is -fdiagnostics-color=auto, otherwise 3265 -fdiagnostics-color=never. 3266 Sample diagnostics output: 3267 $ g++ -fdiagnostics-color=always -S -Wall test.C 3268 test.C: In function `int foo()': 3269 test.C:1:14: warning: no return statement in function returning non-void [-W 3270 return-type] 3271 int foo () { } 3272 ^ 3273 test.C:2:46: error: template instantiation depth exceeds maximum of 900 (use 3274 -ftemplate-depth= to increase the maximum) instantiating `struct X<100>' 3275 template <int N> struct X { static const int value = X<N-1>::value; }; temp 3276 late struct X<1000>; 3277 ^ 3278 test.C:2:46: recursively required from `const int X<999>::value' 3279 test.C:2:46: required from `const int X<1000>::value' 3280 test.C:2:88: required from here 3281 3282 test.C:2:46: error: incomplete type `X<100>' used in nested name specifier 3283 3284 * With the new [7]#pragma GCC ivdep, the user can assert that there 3285 are no loop-carried dependencies which would prevent concurrent 3286 execution of consecutive iterations using SIMD (single instruction 3287 multiple data) instructions. 3288 * Support for [8]Cilk Plus has been added and can be enabled with the 3289 -fcilkplus option. Cilk Plus is an extension to the C and C++ 3290 languages to support data and task parallelism. The present 3291 implementation follows ABI version 1.2; all features but _Cilk_for 3292 have been implemented. 3293 3294 C 3295 3296 * ISO C11 atomics (the _Atomic type specifier and qualifier and the 3297 <stdatomic.h> header) are now supported. 3298 * ISO C11 generic selections (_Generic keyword) are now supported. 3299 * ISO C11 thread-local storage (_Thread_local, similar to GNU C 3300 __thread) is now supported. 3301 * ISO C11 support is now at a similar level of completeness to ISO 3302 C99 support: substantially complete modulo bugs, extended 3303 identifiers (supported except for corner cases when 3304 -fextended-identifiers is used), floating-point issues (mainly but 3305 not entirely relating to optional C99 features from Annexes F and 3306 G) and the optional Annexes K (Bounds-checking interfaces) and L 3307 (Analyzability). 3308 * A new C extension __auto_type provides a subset of the 3309 functionality of C++11 auto in GNU C. 3310 3311 C++ 3312 3313 * The G++ implementation of [9]C++1y return type deduction for normal 3314 functions has been updated to conform to [10]N3638, the proposal 3315 accepted into the working paper. Most notably, it adds 3316 decltype(auto) for getting decltype semantics rather than the 3317 template argument deduction semantics of plain auto: 3318 3319 int& f(); 3320 auto i1 = f(); // int 3321 decltype(auto) i2 = f(); // int& 3322 3323 * G++ supports [11]C++1y lambda capture initializers: 3324 3325 [x = 42]{ ... }; 3326 3327 Actually, they have been accepted since GCC 4.5, but now the 3328 compiler doesn't warn about them with -std=c++1y, and supports 3329 parenthesized and brace-enclosed initializers as well. 3330 * G++ supports [12]C++1y variable length arrays. G++ has supported 3331 GNU/C99-style VLAs for a long time, but now additionally supports 3332 initializers and lambda capture by reference. In C++1y mode G++ 3333 will complain about VLA uses that are not permitted by the draft 3334 standard, such as forming a pointer to VLA type or applying sizeof 3335 to a VLA variable. Note that it now appears that VLAs will not be 3336 part of C++14, but will be part of a separate document and then 3337 perhaps C++17. 3338 3339 void f(int n) { 3340 int a[n] = { 1, 2, 3 }; // throws std::bad_array_length if n < 3 3341 [&a]{ for (int i : a) { cout << i << endl; } }(); 3342 &a; // error, taking address of VLA 3343 } 3344 3345 * G++ supports the [13]C++1y [[deprecated]] attribute modulo bugs in 3346 the underlying [[gnu::deprecated]] attribute. Classes and functions 3347 can be marked deprecated and a diagnostic message added: 3348 3349 class A; 3350 int bar(int n); 3351 #if __cplusplus > 201103 3352 class [[deprecated("A is deprecated in C++14; Use B instead")]] A; 3353 [[deprecated("bar is unsafe; use foo() instead")]] 3354 int bar(int n); 3355 3356 int foo(int n); 3357 class B; 3358 #endif 3359 A aa; // warning: 'A' is deprecated : A is deprecated in C++14; Use B instead 3360 int j = bar(2); // warning: 'int bar(int)' is deprecated : bar is unsafe; use fo 3361 o() instead 3362 3363 * G++ supports [14]C++1y digit separators. Long numeric literals can 3364 be subdivided with a single quote ' to enhance readability: 3365 3366 int i = 1048576; 3367 int j = 1'048'576; 3368 int k = 0x10'0000; 3369 int m = 0'004'000'000; 3370 int n = 0b0001'0000'0000'0000'0000'0000; 3371 3372 double x = 1.602'176'565e-19; 3373 double y = 1.602'176'565e-1'9; 3374 3375 * G++ supports [15]C++1y generic (polymorphic) lambdas. 3376 3377 // a functional object that will increment any type 3378 auto incr = [](auto x) { return x++; }; 3379 3380 * As a GNU extension, G++ supports explicit template parameter syntax 3381 for generic lambdas. This can be combined in the expected way with 3382 the standard auto syntax. 3383 3384 // a functional object that will add two like-type objects 3385 auto add = [] <typename T> (T a, T b) { return a + b; }; 3386 3387 * G++ supports unconstrained generic functions as specified by 3388 S:4.1.2 and S:5.1.1 of [16]N3889: Concepts Lite Specification. 3389 Briefly, auto may be used as a type-specifier in a parameter 3390 declaration of any function declarator in order to introduce an 3391 implicit function template parameter, akin to generic lambdas. 3392 3393 // the following two function declarations are equivalent 3394 auto incr(auto x) { return x++; } 3395 template <typename T> 3396 auto incr(T x) { return x++; } 3397 3398 Runtime Library (libstdc++) 3399 3400 * [17]Improved support for C++11, including: 3401 + support for <regex>; 3402 + The associative containers in <map> and <set> and the 3403 unordered associative containers in <unordered_map> and 3404 <unordered_set> meet the allocator-aware container 3405 requirements; 3406 * [18]Improved experimental support for the upcoming ISO C++ 3407 standard, C++14, including: 3408 + fixing constexpr member functions without const; 3409 + implementation of the std::exchange() utility function; 3410 + addressing tuples by type; 3411 + implemention of std::make_unique; 3412 + implemention of std::shared_lock; 3413 + making std::result_of SFINAE-friendly; 3414 + adding operator() to std::integral_constant; 3415 + adding user-defined literals for standard library types 3416 std::basic_string, std::chrono::duration, and std::complex; 3417 + adding two range overloads to non-modifying sequence oprations 3418 std::equal and std::mismatch; 3419 + adding IO manipulators for quoted strings; 3420 + adding constexpr members to <utility>, <complex>, <chrono>, 3421 and some containers; 3422 + adding compile-time std::integer_sequence; 3423 + adding cleaner transformation traits; 3424 + making <functional>s operator functors easier to use and more 3425 generic; 3426 * An implementation of std::experimental::optional. 3427 * An implementation of std::experimental::string_view. 3428 * The non-standard function std::copy_exception has been deprecated 3429 and will be removed in a future version. std::make_exception_ptr 3430 should be used instead. 3431 3432 Fortran 3433 3434 * Compatibility notice: 3435 + Module files: The version of the module files (.mod) has been 3436 incremented; additionally, module files are now compressed. 3437 Fortran MODULEs compiled by earlier GCC versions have to be 3438 recompiled, when they are USEd by files compiled with GCC 4.9. 3439 GCC 4.9 is not able to read .mod files of earlier GCC 3440 versions; attempting to do so gives an error message. Note: 3441 The ABI of the produced assembler data itself has not changed: 3442 object files and libraries are fully compatible with older 3443 versions (except as stated below). 3444 + ABI changes: 3445 o The [19]argument passing ABI has changed for scalar dummy 3446 arguments of type INTEGER, REAL, COMPLEX and LOGICAL, 3447 which have both the VALUE and the OPTIONAL attributes. 3448 o To support finalization the virtual table associated with 3449 polymorphic variables has changed. Code containing CLASS 3450 should be recompiled, including all files which define 3451 derived types involved in the type definition used by 3452 polymorphic variables. (Note: Due to the incremented 3453 module version, trying to mix old code with new code will 3454 usually give an error message.) 3455 + GNU Fortran no longer deallocates allocatable variables or 3456 allocatable components of variables declared in the main 3457 program. Since Fortran 2008, the standard explicitly states 3458 that variables declared in the Fortran main program 3459 automatically have the SAVE attribute. 3460 + When opening files, the close-on-exec flag is set if the 3461 system supports such a feature. This is generally considered 3462 good practice these days, but if there is a need to pass file 3463 descriptors to child processes the parent process must now 3464 remember to clear the close-on-exec flag by calling fcntl(), 3465 e.g. via ISO_C_BINDING, before executing the child process. 3466 * The deprecated command-line option -fno-whole-file has been 3467 removed. (-fwhole-file is the default since GCC 4.6.) 3468 -fwhole-file/-fno-whole-file continue to be accepted but do not 3469 influence the code generation. 3470 * The compiler no longer unconditionally warns about DO loops with 3471 zero iterations. This warning is now controlled by the -Wzerotrip 3472 option, which is implied by -Wall. 3473 * The new NO_ARG_CHECK attribute of the [20]!GCC$ directive can be 3474 used to disable the type-kind-rank (TKR) argument check for a dummy 3475 argument. The feature is similar to ISO/IEC TS 29133:2012's 3476 TYPE(*), except that it additionally also disables the rank check. 3477 Variables with NO_ARG_CHECK have to be dummy arguments and may only 3478 be used as argument to ISO_C_BINDING's C_LOC and as actual argument 3479 to another NO_ARG_CHECK dummy argument; also the other constraints 3480 of TYPE(*) apply. The dummy arguments should be declared as scalar 3481 or assumed-size variable of type type(*) (recommended) - or of type 3482 integer, real, complex or logical. With NO_ARG_CHECK, a pointer to 3483 the data without further type or shape information is passed, 3484 similar to C's void*. Note that also TS 29113's 3485 type(*),dimension(..) accepts arguments of any type and rank; 3486 contrary to NO_ARG_CHECK assumed-rank arguments pass an array 3487 descriptor which contains the array shape and stride of the 3488 argument. 3489 * [21]Fortran 2003: 3490 + Finalization is now supported. It is currently only done for a 3491 subset of those situations in which it should occur. 3492 + Experimental support for scalar character components with 3493 deferred length (i.e. allocatable string length) in derived 3494 types has been added. (Deferred-length character variables are 3495 supported since GCC 4.6.) 3496 * [22]Fortran 2008: 3497 + When STOP or ERROR STOP are used to terminate the execution 3498 and any exception (but inexact) is signaling, a warning is 3499 printed to ERROR_UNIT, indicating which exceptions are 3500 signaling. The [23]-ffpe-summary= command-line option can be 3501 used to fine-tune for which exceptions the warning should be 3502 shown. 3503 + Rounding on input (READ) is now handled on systems where 3504 strtod honours the rounding mode. (For output, rounding is 3505 supported since GCC 4.5.) Note that for input, the compatible 3506 rounding mode is handled as nearest (i.e., rounding to an even 3507 least significant [cf. IEC 60559:1989] for a tie, while 3508 compatible rounds away from zero in that case). 3509 3510 Go 3511 3512 * GCC 4.9 provides a complete implementation of the Go 1.2.1 release. 3513 3514 New Targets and Target Specific Improvements 3515 3516 AArch64 3517 3518 * The ARMv8-A crypto and CRC instructions are now supported through 3519 intrinsics. These are enabled when the architecture supports these 3520 and are available through the -march=armv8-a+crc and 3521 -march=armv8-a+crypto options. 3522 * Initial support for ILP32 has now been added to the compiler. This 3523 is now available through the command-line option -mabi=ilp32. 3524 Support for ILP32 is considered experimental as the ABI 3525 specification is still beta. 3526 * Coverage of more of the ISA including the SIMD extensions has been 3527 added. The Advanced SIMD intrinsics have also been improved. 3528 * The new local register allocator (LRA) is now on by default for the 3529 AArch64 backend. 3530 * The REE (Redundant extension elimination) pass has now been enabled 3531 by default for the AArch64 backend. 3532 * Tuning for the Cortex-A53 and Cortex-A57 has been improved. 3533 * Initial big.LITTLE tuning support for the combination of Cortex-A57 3534 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 3535 option. 3536 * A number of structural changes have been made to both the ARM and 3537 AArch64 backends to facilitate improved code-generation. 3538 * As of GCC 4.9.2 a workaround for the ARM Cortex-A53 erratum 835769 3539 has been added and can be enabled by giving the 3540 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 3541 default by configuring GCC with the --enable-fix-cortex-a53-835769 3542 option. 3543 3544 ARC 3545 3546 * A port for Synopsys Designware ARC has been contributed by Embecosm 3547 and Synopsys Inc. 3548 3549 ARM 3550 3551 * Use of Advanced SIMD (Neon) for 64-bit scalar computations has been 3552 disabled by default. This was found to generate better code in only 3553 a small number of cases. It can be turned back on with the 3554 -mneon-for-64bits option. 3555 * Further support for the ARMv8-A architecture, notably implementing 3556 the restriction around IT blocks in the Thumb32 instruction set has 3557 been added. The -mrestrict-it option can be used with 3558 -march=armv7-a or the -march=armv7ve options to make code 3559 generation fully compatible with the deprecated instructions in 3560 ARMv8-A. 3561 * Support has now been added for the ARMv7ve variant of the 3562 architecture. This can be used by the -march=armv7ve option. 3563 * The ARMv8-A crypto and CRC instructions are now supported through 3564 intrinsics and are available through the -march=armv8-a+crc and 3565 mfpu=crypto-neon-fp-armv8 options. 3566 * LRA is now on by default for the ARM target. This can be turned off 3567 using the -mno-lra option. This option is a purely transitionary 3568 command-line option and will be removed in a future release. We are 3569 interested in any bug reports regarding functional and performance 3570 regressions with LRA. 3571 * A new option -mslow-flash-data to improve performance of programs 3572 fetching data on slow flash memory has now been introduced for the 3573 ARMv7-M profile cores. 3574 * A new option -mpic-data-is-text-relative for targets that allows 3575 data segments to be relative to text segments has been added. This 3576 is on by default for all targets except VxWorks RTP. 3577 * A number of infrastructural changes have been made to both the ARM 3578 and AArch64 backends to facilitate improved code-generation. 3579 * GCC now supports Cortex-A12 and the Cortex-R7 through the 3580 -mcpu=cortex-a12 and -mcpu=cortex-r7 options. 3581 * GCC now has tuning for the Cortex-A57 and Cortex-A53 through the 3582 -mcpu=cortex-a57 and -mcpu=cortex-a53 options. 3583 * Initial big.LITTLE tuning support for the combination of Cortex-A57 3584 and Cortex-A53 was added through the -mcpu=cortex-a57.cortex-a53 3585 option. Similar support was added for the combination of Cortex-A15 3586 and Cortex-A7 through the -mcpu=cortex-a15.cortex-a7 option. 3587 * Further performance optimizations for the Cortex-A15 and the 3588 Cortex-M4 have been added. 3589 * A number of code generation improvements for Thumb2 to reduce code 3590 size when compiling for the M-profile processors. 3591 3592 AVR 3593 3594 * A new command-line option -mfract-convert-truncate has been added. 3595 It allows compiler to use truncation instead of rounding towards 3596 zero for fractional fixed-point types. 3597 3598 IA-32/x86-64 3599 3600 * -mfpmath=sse is now implied by -ffast-math on all targets where 3601 SSE2 is supported. 3602 * Intel AVX-512 support was added to GCC. That includes inline 3603 assembly support, new registers and extending existing ones, new 3604 intrinsics (covered by corresponding testsuite), and basic 3605 autovectorization. AVX-512 instructions are available via the 3606 following GCC switches: AVX-512 foundation instructions: -mavx512f, 3607 AVX-512 prefetch instructions: -mavx512pf, AVX-512 exponential and 3608 reciprocal instructions: -mavx512er, AVX-512 conflict detection 3609 instructions: -mavx512cd. 3610 * It is now possible to call x86 intrinsics from select functions in 3611 a file that are tagged with the corresponding target attribute 3612 without having to compile the entire file with the -mxxx option. 3613 This improves the usability of x86 intrinsics and is particularly 3614 useful when doing [24]Function Multiversioning. 3615 * GCC now supports the new Intel microarchitecture named Silvermont 3616 through -march=silvermont. 3617 * GCC now supports the new Intel microarchitecture named Broadwell 3618 through -march=broadwell. 3619 * Optimizing for other Intel microarchitectures have been renamed to 3620 -march=nehalem, westmere, sandybridge, ivybridge, haswell, bonnell. 3621 * -march=generic has been retuned for better support of Intel core 3622 and AMD Bulldozer architectures. Performance of AMD K7, K8, Intel 3623 Pentium-M, and Pentium4 based CPUs is no longer considered 3624 important for generic. 3625 * -mtune=intel can now be used to generate code running well on the 3626 most current Intel processors, which are Haswell and Silvermont for 3627 GCC 4.9. 3628 * Support to encode 32-bit assembly instructions in 16-bit format is 3629 now available through the -m16 command-line option. 3630 * Better inlining of memcpy and memset that is aware of value ranges 3631 and produces shorter alignment prologues. 3632 * -mno-accumulate-outgoing-args is now honored when unwind 3633 information is output. Argument accumulation is also now turned off 3634 for portions of programs optimized for size. 3635 * Support for new AMD family 15h processors (Excavator core) is now 3636 available through the -march=bdver4 and -mtune=bdver4 options. 3637 3638 MSP430 3639 3640 * A new command-line option -mcpu= has been added to the MSP430 3641 backend. This option is used to specify the ISA to be used. 3642 Accepted values are msp430 (the default), msp430x and msp430xv2. 3643 The ISA is no longer deduced from the -mmcu= option as there are 3644 far too many different MCU names. The -mmcu= option is still 3645 supported, and this is still used to select linker scripts and 3646 generate a C preprocessor symbol that will be recognised by the 3647 msp430.h header file. 3648 3649 NDS32 3650 3651 * A new nds32 port supports the 32-bit architecture from Andes 3652 Technology Corporation. 3653 * The port provides initial support for the V2, V3, V3m instruction 3654 set architectures. 3655 3656 Nios II 3657 3658 * A port for the Altera Nios II has been contributed by Mentor 3659 Graphics. 3660 3661 PowerPC / PowerPC64 / RS6000 3662 3663 * GCC now supports Power ISA 2.07, which includes support for 3664 Hardware Transactional Memory (HTM), Quadword atomics and several 3665 VMX and VSX additions, including Crypto, 64-bit integer, 128-bit 3666 integer and decimal integer operations. 3667 * Support for the POWER8 processor is now available through the 3668 -mcpu=power8 and -mtune=power8 options. 3669 * The libitm library has been modified to add a HTM fastpath that 3670 automatically uses POWER's HTM hardware instructions when it is 3671 executing on a HTM enabled processor. 3672 * Support for the new powerpc64le-linux platform has been added. It 3673 defaults to generating code that conforms to the ELFV2 ABI. 3674 3675 S/390, System z 3676 3677 * Support for the Transactional Execution Facility included with the 3678 IBM zEnterprise zEC12 processor has been added. A set of GCC style 3679 builtins as well as XLC style builtins are provided. The builtins 3680 are enabled by default when using the -march=zEC12 option but can 3681 explicitly be disabled with -mno-htm. Using the GCC builtins also 3682 libitm supports hardware transactions on S/390. 3683 * The hotpatch features allows to prepare functions for hotpatching. 3684 A certain amount of bytes is reserved before the function entry 3685 label plus a NOP is inserted at its very beginning to implement a 3686 backward jump when applying a patch. The feature can either be 3687 enabled per compilation unit via the command-line option -mhotpatch 3688 or per function using the hotpatch attribute. 3689 * The shrink wrap optimization is now supported on S/390 and enabled 3690 by default. 3691 * A major rework of the routines to determine which registers need to 3692 be saved and restored in function prologue/epilogue now allow to 3693 use floating point registers as save slots. This will happen for 3694 certain leaf function with -march=z10 or higher. 3695 * The LRA rtl pass replaces reload by default on S/390. 3696 3697 RX 3698 3699 * The port now allows to specify the RX100, RX200, and RX600 3700 processors with the command-line options -mcpu=rx100, -mcpu=rx200 3701 and -mcpu=rx600. 3702 3703 SH 3704 3705 * Minor improvements to code generated for integer arithmetic and 3706 code that involves the T bit. 3707 * Added support for the SH2A clips and clipu instructions. The 3708 compiler will now try to utilize them for min/max expressions such 3709 as max (-128, min (127, x)). 3710 * Added support for the cmp/str instruction through built-in 3711 functions such as __builtin_strlen. When not optimizing for size, 3712 the compiler will now expand calls to e.g. strlen as an inlined 3713 sequences which utilize the cmp/str instruction. 3714 * Improved code generated around volatile memory loads and stores. 3715 * The option -mcbranchdi has been deprecated. Specifying it will 3716 result in a warning and will not influence code generation. 3717 * The option -mcmpeqdi has been deprecated. Specifying it will result 3718 in a warning and will not influence code generation. 3719 3720 GCC 4.9.1 3721 3722 This is the [25]list of problem reports (PRs) from GCC's bug tracking 3723 system that are known to be fixed in the 4.9.1 release. This list might 3724 not be complete (that is, it is possible that some PRs that have been 3725 fixed are not listed here). 3726 3727 Version 4.0 of the OpenMP specification is supported even in Fortran, 3728 not just C and C++. 3729 3730 GCC 4.9.2 3731 3732 This is the [26]list of problem reports (PRs) from GCC's bug tracking 3733 system that are known to be fixed in the 4.9.2 release. This list might 3734 not be complete (that is, it is possible that some PRs that have been 3735 fixed are not listed here). 3736 3737 GCC 4.9.3 3738 3739 This is the [27]list of problem reports (PRs) from GCC's bug tracking 3740 system that are known to be fixed in the 4.9.3 release. This list might 3741 not be complete (that is, it is possible that some PRs that have been 3742 fixed are not listed here). 3743 3744 GCC 4.9.4 3745 3746 This is the [28]list of problem reports (PRs) from GCC's bug tracking 3747 system that are known to be fixed in the 4.9.4 release. This list might 3748 not be complete (that is, it is possible that some PRs that have been 3749 fixed are not listed here). 3750 3751 3752 For questions related to the use of GCC, please consult these web 3753 pages and the [29]GCC manuals. If that fails, the 3754 [30]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3755 web pages and the development of GCC are welcome on our developer 3756 list at [31]gcc (a] gcc.gnu.org. All of [32]our lists have public 3757 archives. 3758 3759 Copyright (C) [33]Free Software Foundation, Inc. Verbatim copying and 3760 distribution of this entire article is permitted in any medium, 3761 provided this notice is preserved. 3762 3763 These pages are [34]maintained by the GCC team. Last modified 3764 2018-12-01[35]. 3765 3766 References 3767 3768 1. https://gcc.gnu.org/ml/gcc-patches/2013-05/msg00728.html 3769 2. https://gcc.gnu.org/PR60825 3770 3. https://gcc.gnu.org/gcc-4.9/porting_to.html 3771 4. https://www.openmp.org/specifications/ 3772 5. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Optimize-Options.html#index-fsimd-cost-model-908 3773 6. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Language-Independent-Options.html#index-fdiagnostics-color-252 3774 7. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Loop-Specific-Pragmas.html 3775 8. https://www.cilkplus.org/ 3776 9. http://gcc.gnu.org/projects/cxx1y.html 3777 10. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2013/n3638.html 3778 11. http://gcc.gnu.org/projects/cxx1y.html 3779 12. http://gcc.gnu.org/projects/cxx1y.html 3780 13. http://gcc.gnu.org/projects/cxx1y.html 3781 14. http://gcc.gnu.org/projects/cxx1y.html 3782 15. http://gcc.gnu.org/projects/cxx1y.html 3783 16. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2014/n3889.pdf 3784 17. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/libstdc++/manual/manual/status.html#status.iso.2011 3785 18. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.2014 3786 19. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Argument-passing-conventions.html 3787 20. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/GNU-Fortran-Compiler-Directives.html 3788 21. https://gcc.gnu.org/wiki/Fortran2003Status 3789 22. https://gcc.gnu.org/wiki/Fortran2008Status 3790 23. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gfortran/Debugging-Options.html 3791 24. https://gcc.gnu.org/onlinedocs/gcc-4.9.2/gcc/Function-Multiversioning.html 3792 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.1 3793 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.2 3794 27. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.3 3795 28. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.9.4 3796 29. https://gcc.gnu.org/onlinedocs/ 3797 30. mailto:gcc-help (a] gcc.gnu.org 3798 31. mailto:gcc (a] gcc.gnu.org 3799 32. https://gcc.gnu.org/lists.html 3800 33. https://www.fsf.org/ 3801 34. https://gcc.gnu.org/about.html 3802 35. http://validator.w3.org/check/referer 3803 ====================================================================== 3804 http://gcc.gnu.org/gcc-4.8/index.html 3805 3806 GCC 4.8 Release Series 3807 3808 June 23, 2015 3809 3810 The [1]GNU project and the GCC developers are pleased to announce the 3811 release of GCC 4.8.5. 3812 3813 This release is a bug-fix release, containing fixes for regressions in 3814 GCC 4.8.4 relative to previous releases of GCC. 3815 3816 This release series is no longer maintained. 3817 3818 Release History 3819 3820 GCC 4.8.5 3821 June 23, 2015 ([2]changes, [3]documentation) 3822 3823 GCC 4.8.4 3824 December 19, 2014 ([4]changes, [5]documentation) 3825 3826 GCC 4.8.3 3827 May 22, 2014 ([6]changes, [7]documentation) 3828 3829 GCC 4.8.2 3830 October 16, 2013 ([8]changes, [9]documentation) 3831 3832 GCC 4.8.1 3833 May 31, 2013 ([10]changes, [11]documentation) 3834 3835 GCC 4.8.0 3836 March 22, 2013 ([12]changes, [13]documentation) 3837 3838 References and Acknowledgements 3839 3840 GCC used to stand for the GNU C Compiler, but since the compiler 3841 supports several other languages aside from C, it now stands for the 3842 GNU Compiler Collection. 3843 3844 A list of [14]successful builds is updated as new information becomes 3845 available. 3846 3847 The GCC developers would like to thank the numerous people that have 3848 contributed new features, improvements, bug fixes, and other changes as 3849 well as test results to GCC. This [15]amazing group of volunteers is 3850 what makes GCC successful. 3851 3852 For additional information about GCC please refer to the [16]GCC 3853 project web site or contact the [17]GCC development mailing list. 3854 3855 To obtain GCC please use [18]our mirror sites or [19]our SVN server. 3856 3857 3858 For questions related to the use of GCC, please consult these web 3859 pages and the [20]GCC manuals. If that fails, the 3860 [21]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 3861 web pages and the development of GCC are welcome on our developer 3862 list at [22]gcc (a] gcc.gnu.org. All of [23]our lists have public 3863 archives. 3864 3865 Copyright (C) [24]Free Software Foundation, Inc. Verbatim copying and 3866 distribution of this entire article is permitted in any medium, 3867 provided this notice is preserved. 3868 3869 These pages are [25]maintained by the GCC team. Last modified 3870 2018-09-30[26]. 3871 3872 References 3873 3874 1. http://www.gnu.org/ 3875 2. http://gcc.gnu.org/gcc-4.8/changes.html 3876 3. https://gcc.gnu.org/onlinedocs/4.8.5/ 3877 4. http://gcc.gnu.org/gcc-4.8/changes.html 3878 5. https://gcc.gnu.org/onlinedocs/4.8.4/ 3879 6. http://gcc.gnu.org/gcc-4.8/changes.html 3880 7. https://gcc.gnu.org/onlinedocs/4.8.3/ 3881 8. http://gcc.gnu.org/gcc-4.8/changes.html 3882 9. https://gcc.gnu.org/onlinedocs/4.8.2/ 3883 10. http://gcc.gnu.org/gcc-4.8/changes.html 3884 11. https://gcc.gnu.org/onlinedocs/4.8.1/ 3885 12. http://gcc.gnu.org/gcc-4.8/changes.html 3886 13. https://gcc.gnu.org/onlinedocs/4.8.0/ 3887 14. http://gcc.gnu.org/gcc-4.8/buildstat.html 3888 15. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 3889 16. http://gcc.gnu.org/index.html 3890 17. mailto:gcc (a] gcc.gnu.org 3891 18. http://gcc.gnu.org/mirrors.html 3892 19. http://gcc.gnu.org/svn.html 3893 20. https://gcc.gnu.org/onlinedocs/ 3894 21. mailto:gcc-help (a] gcc.gnu.org 3895 22. mailto:gcc (a] gcc.gnu.org 3896 23. https://gcc.gnu.org/lists.html 3897 24. https://www.fsf.org/ 3898 25. https://gcc.gnu.org/about.html 3899 26. http://validator.w3.org/check/referer 3900 ====================================================================== 3901 http://gcc.gnu.org/gcc-4.8/changes.html 3902 3903 GCC 4.8 Release Series 3904 Changes, New Features, and Fixes 3905 3906 Caveats 3907 3908 GCC now uses C++ as its implementation language. This means that to 3909 build GCC from sources, you will need a C++ compiler that understands 3910 C++ 2003. For more details on the rationale and specific changes, 3911 please refer to the [1]C++ conversion page. 3912 3913 To enable the Graphite framework for loop optimizations you now need 3914 CLooG version 0.18.0 and ISL version 0.11.1. Both can be obtained from 3915 the [2]GCC infrastructure directory. The installation manual contains 3916 more information about requirements to build GCC. 3917 3918 GCC now uses a more aggressive analysis to derive an upper bound for 3919 the number of iterations of loops using constraints imposed by language 3920 standards. This may cause non-conforming programs to no longer work as 3921 expected, such as SPEC CPU 2006 464.h264ref and 416.gamess. A new 3922 option, -fno-aggressive-loop-optimizations, was added to disable this 3923 aggressive analysis. In some loops that have known constant number of 3924 iterations, but undefined behavior is known to occur in the loop before 3925 reaching or during the last iteration, GCC will warn about the 3926 undefined behavior in the loop instead of deriving lower upper bound of 3927 the number of iterations for the loop. The warning can be disabled with 3928 -Wno-aggressive-loop-optimizations. 3929 3930 On ARM, a bug has been fixed in GCC's implementation of the AAPCS rules 3931 for the layout of vectors that could lead to wrong code being 3932 generated. Vectors larger than 8 bytes in size are now by default 3933 aligned to an 8-byte boundary. This is an ABI change: code that makes 3934 explicit use of vector types may be incompatible with binary objects 3935 built with older versions of GCC. Auto-vectorized code is not affected 3936 by this change. 3937 3938 On AVR, support has been removed for the command-line option 3939 -mshort-calls deprecated in GCC 4.7. 3940 3941 On AVR, the configure option --with-avrlibc supported since GCC 4.7.2 3942 is turned on per default for all non-RTEMS configurations. This option 3943 arranges for a better integration of [3]AVR Libc with avr-gcc. For 3944 technical details, see [4]PR54461. To turn off the option in non-RTEMS 3945 configurations, use --with-avrlibc=no. If the compiler is configured 3946 for RTEMS, the option is always turned off. 3947 3948 More information on porting to GCC 4.8 from previous versions of GCC 3949 can be found in the [5]porting guide for this release. 3950 3951 General Optimizer Improvements (and Changes) 3952 3953 * DWARF4 is now the default when generating DWARF debug information. 3954 When -g is used on a platform that uses DWARF debugging 3955 information, GCC will now default to -gdwarf-4 3956 -fno-debug-types-section. 3957 GDB 7.5, Valgrind 3.8.0 and elfutils 0.154 debug information 3958 consumers support DWARF4 by default. Before GCC 4.8 the default 3959 version used was DWARF2. To make GCC 4.8 generate an older DWARF 3960 version use -g together with -gdwarf-2 or -gdwarf-3. The default 3961 for Darwin and VxWorks is still -gdwarf-2 -gstrict-dwarf. 3962 * A new general optimization level, -Og, has been introduced. It 3963 addresses the need for fast compilation and a superior debugging 3964 experience while providing a reasonable level of run-time 3965 performance. Overall experience for development should be better 3966 than the default optimization level -O0. 3967 * A new option -ftree-partial-pre was added to control the partial 3968 redundancy elimination (PRE) optimization. This option is enabled 3969 by default at the -O3 optimization level, and it makes PRE more 3970 aggressive. 3971 * The option -fconserve-space has been removed; it was no longer 3972 useful on most targets since GCC supports putting variables into 3973 BSS without making them common. 3974 * The struct reorg and matrix reorg optimizations (command-line 3975 options -fipa-struct-reorg and -fipa-matrix-reorg) have been 3976 removed. They did not always work correctly, nor did they work with 3977 link-time optimization (LTO), hence were only applicable to 3978 programs consisting of a single translation unit. 3979 * Several scalability bottle-necks have been removed from GCC's 3980 optimization passes. Compilation of extremely large functions, e.g. 3981 due to the use of the flatten attribute in the "Eigen" C++ linear 3982 algebra templates library, is significantly faster than previous 3983 releases of GCC. 3984 * Link-time optimization (LTO) improvements: 3985 + LTO partitioning has been rewritten for better reliability and 3986 maintanibility. Several important bugs leading to link 3987 failures have been fixed. 3988 * Interprocedural optimization improvements: 3989 + A new symbol table has been implemented. It builds on existing 3990 callgraph and varpool modules and provide a new API. Unusual 3991 symbol visibilities and aliases are handled more consistently 3992 leading to, for example, more aggressive unreachable code 3993 removal with LTO. 3994 + The inline heuristic can now bypass limits on the size of of 3995 inlined functions when the inlining is particularly 3996 profitable. This happens, for example, when loop bounds or 3997 array strides get propagated. 3998 + Values passed through aggregates (either by value or 3999 reference) are now propagated at the inter-procedural level 4000 leading to better inlining decisions (for example in the case 4001 of Fortran array descriptors) and devirtualization. 4002 * [6]AddressSanitizer , a fast memory error detector, has been added 4003 and can be enabled via -fsanitize=address. Memory access 4004 instructions will be instrumented to detect heap-, stack-, and 4005 global-buffer overflow as well as use-after-free bugs. To get nicer 4006 stacktraces, use -fno-omit-frame-pointer. The AddressSanitizer is 4007 available on IA-32/x86-64/x32/PowerPC/PowerPC64 GNU/Linux and on 4008 x86-64 Darwin. 4009 * [7]ThreadSanitizer has been added and can be enabled via 4010 -fsanitize=thread. Instructions will be instrumented to detect data 4011 races. The ThreadSanitizer is available on x86-64 GNU/Linux. 4012 * A new local register allocator (LRA) has been implemented, which 4013 replaces the 26 year old reload pass and improves generated code 4014 quality. For now it is active on the IA-32 and x86-64 targets. 4015 * Support for transactional memory has been implemented on the 4016 following architectures: IA-32/x86-64, ARM, PowerPC, SH, SPARC, and 4017 Alpha. 4018 4019 New Languages and Language specific improvements 4020 4021 C family 4022 4023 * Each diagnostic emitted now includes the original source line and a 4024 caret '^' indicating the column. The option 4025 -fno-diagnostics-show-caret suppresses this information. 4026 * The option -ftrack-macro-expansion=2 is now enabled by default. 4027 This allows the compiler to display the macro expansion stack in 4028 diagnostics. Combined with the caret information, an example 4029 diagnostic showing these two features is: 4030 4031 t.c:1:94: error: invalid operands to binary < (have `struct mystruct' and `float 4032 ') 4033 #define MYMAX(A,B) __extension__ ({ __typeof__(A) __a = (A); __typeof__(B) _ 4034 _b = (B); __a < __b ? __b : __a; }) 4035 4036 ^ 4037 t.c:7:7: note: in expansion of macro 'MYMAX' 4038 X = MYMAX(P, F); 4039 ^ 4040 4041 * A new -Wsizeof-pointer-memaccess warning has been added (also 4042 enabled by -Wall) to warn about suspicious length parameters to 4043 certain string and memory built-in functions if the argument uses 4044 sizeof. This warning warns e.g. about memset (ptr, 0, sizeof 4045 (ptr)); if ptr is not an array, but a pointer, and suggests a 4046 possible fix, or about memcpy (&foo, ptr, sizeof (&foo));. 4047 * The new option -Wpedantic is an alias for -pedantic, which is now 4048 deprecated. The forms -Wno-pedantic, -Werror=pedantic, and 4049 -Wno-error=pedantic work in the same way as for any other -W 4050 option. One caveat is that -Werror=pedantic is not equivalent to 4051 -pedantic-errors, since the latter makes into errors some warnings 4052 that are not controlled by -Wpedantic, and the former only affects 4053 diagnostics that are disabled when using -Wno-pedantic. 4054 * The option -Wshadow no longer warns if a declaration shadows a 4055 function declaration, unless the former declares a function or 4056 pointer to function, because this is [8]a common and valid case in 4057 real-world code. 4058 4059 C++ 4060 4061 * G++ now implements the [9]C++11 thread_local keyword; this differs 4062 from the GNU __thread keyword primarily in that it allows dynamic 4063 initialization and destruction semantics. Unfortunately, this 4064 support requires a run-time penalty for references to 4065 non-function-local thread_local variables defined in a different 4066 translation unit even if they don't need dynamic initialization, so 4067 users may want to continue to use __thread for TLS variables with 4068 static initialization semantics. 4069 If the programmer can be sure that no use of the variable in a 4070 non-defining TU needs to trigger dynamic initialization (either 4071 because the variable is statically initialized, or a use of the 4072 variable in the defining TU will be executed before any uses in 4073 another TU), they can avoid this overhead with the 4074 -fno-extern-tls-init option. 4075 OpenMP threadprivate variables now also support dynamic 4076 initialization and destruction by the same mechanism. 4077 * G++ now implements the [10]C++11 attribute syntax, e.g. 4078 4079 [[noreturn]] void f(); 4080 4081 and also the alignment specifier, e.g. 4082 4083 alignas(double) int i; 4084 4085 * G++ now implements [11]C++11 inheriting constructors, e.g. 4086 4087 struct A { A(int); }; 4088 struct B: A { using A::A; }; // defines B::B(int) 4089 B b(42); // OK 4090 4091 * As of GCC 4.8.1, G++ implements the change to decltype semantics 4092 from [12]N3276. 4093 4094 struct A f(); 4095 decltype(f()) g(); // OK, return type of f() is not required to be complete. 4096 4097 * As of GCC 4.8.1, G++ implements [13]C++11 ref-qualifiers, e.g. 4098 4099 struct A { int f() &; }; 4100 int i = A().f(); // error, f() requires an lvalue object 4101 4102 * G++ now supports a -std=c++1y option for experimentation with 4103 features proposed for the next revision of the standard, expected 4104 around 2014. Currently the only difference from -std=c++11 is 4105 support for return type deduction in normal functions, as proposed 4106 in [14]N3386. Status of C++1y features in GCC 4.8 can be found 4107 [15]here. 4108 * The G++ namespace association extension, __attribute ((strong)), 4109 has been deprecated. Inline namespaces should be used instead. 4110 * G++ now supports a -fext-numeric-literal option to control whether 4111 GNU numeric literal suffixes are accepted as extensions or 4112 processed as C++11 user-defined numeric literal suffixes. The flag 4113 is on (use suffixes for GNU literals) by default for -std=gnu++*, 4114 and -std=c++98. The flag is off (use suffixes for user-defined 4115 literals) by default for -std=c++11 and later. 4116 4117 Runtime Library (libstdc++) 4118 4119 * [16]Improved experimental support for the new ISO C++ standard, 4120 C++11, including: 4121 + forward_list meets the allocator-aware container requirements; 4122 + this_thread::sleep_for(), this_thread::sleep_until() and 4123 this_thread::yield() are defined without requiring the 4124 configure option --enable-libstdcxx-time; 4125 * Improvements to <random>: 4126 + SSE optimized normal_distribution. 4127 + Use of hardware RNG instruction for random_device on new x86 4128 processors (requires the assembler to support the 4129 instruction.) 4130 and <ext/random>: 4131 + New random number engine simd_fast_mersenne_twister_engine 4132 with an optimized SSE implementation. 4133 + New random number distributions beta_distribution, 4134 normal_mv_distribution, rice_distribution, 4135 nakagami_distribution, pareto_distribution, k_distribution, 4136 arcsine_distribution, hoyt_distribution. 4137 * Added --disable-libstdcxx-verbose configure option to disable 4138 diagnostic messages issued when a process terminates abnormally. 4139 This may be useful for embedded systems to reduce the size of 4140 executables that link statically to the library. 4141 4142 Fortran 4143 4144 * Compatibility notice: 4145 + Module files: The version of module files (.mod) has been 4146 incremented. Fortran MODULEs compiled by earlier GCC versions 4147 have to be recompiled, when they are USEd by files compiled 4148 with GCC 4.8. GCC 4.8 is not able to read .mod files created 4149 by earlier versions; attempting to do so gives an error 4150 message. 4151 Note: The ABI of the produced assembler data itself has not 4152 changed; object files and libraries are fully compatible with 4153 older versions except as noted below. 4154 + ABI: Some internal names (used in the assembler/object file) 4155 have changed for symbols declared in the specification part of 4156 a module. If an affected module - or a file using it via use 4157 association - is recompiled, the module and all files which 4158 directly use such symbols have to be recompiled as well. This 4159 change only affects the following kind of module symbols: 4160 o Procedure pointers. Note: C-interoperable function 4161 pointers (type(c_funptr)) are not affected nor are 4162 procedure-pointer components. 4163 o Deferred-length character strings. 4164 * The [17]BACKTRACE intrinsic subroutine has been added. It shows a 4165 backtrace at an arbitrary place in user code; program execution 4166 continues normally afterwards. 4167 * The [18]-Wc-binding-type warning option has been added (disabled by 4168 default). It warns if the a variable might not be C interoperable; 4169 in particular, if the variable has been declared using an intrinsic 4170 type with default kind instead of using a kind parameter defined 4171 for C interoperability in the intrinsic ISO_C_Binding module. 4172 Before, this warning was always printed. The -Wc-binding-type 4173 option is enabled by -Wall. 4174 * The [19]-Wrealloc-lhs and -Wrealloc-lhs-all warning command-line 4175 options have been added, which diagnose when code is inserted for 4176 automatic (re)allocation of a variable during assignment. This 4177 option can be used to decide whether it is safe to use 4178 [20]-fno-realloc-lhs. Additionally, it can be used to find 4179 automatic (re)allocation in hot loops. (For arrays, replacing 4180 "var=" by "var(:)=" disables the automatic reallocation.) 4181 * The [21]-Wcompare-reals command-line option has been added. When 4182 this is set, warnings are issued when comparing REAL or COMPLEX 4183 types for equality and inequality; consider replacing a == b by 4184 abs(a-b) < eps with a suitable eps. -Wcompare-reals is enabled by 4185 -Wextra. 4186 * The [22]-Wtarget-lifetime command-line option has been added 4187 (enabled with -Wall), which warns if the pointer in a pointer 4188 assignment might outlive its target. 4189 * Reading floating point numbers which use "q" for the exponential 4190 (such as 4.0q0) is now supported as vendor extension for better 4191 compatibility with old data files. It is strongly recommended to 4192 use for I/O the equivalent but standard conforming "e" (such as 4193 4.0e0). 4194 (For Fortran source code, consider replacing the "q" in 4195 floating-point literals by a kind parameter (e.g. 4.0e0_qp with a 4196 suitable qp). Note that - in Fortran source code - replacing "q" by 4197 a simple "e" is not equivalent.) 4198 * The GFORTRAN_TMPDIR environment variable for specifying a 4199 non-default directory for files opened with STATUS="SCRATCH", is 4200 not used anymore. Instead gfortran checks the POSIX/GNU standard 4201 TMPDIR environment variable. If TMPDIR is not defined, gfortran 4202 falls back to other methods to determine the directory for 4203 temporary files as documented in the [23]user manual. 4204 * [24]Fortran 2003: 4205 + Support for unlimited polymorphic variables (CLASS(*)) has 4206 been added. Nonconstant character lengths are not yet 4207 supported. 4208 * [25]TS 29113: 4209 + Assumed types (TYPE(*)) are now supported. 4210 + Experimental support for assumed-rank arrays (dimension(..)) 4211 has been added. Note that currently gfortran's own array 4212 descriptor is used, which is different from the one defined in 4213 TS29113, see [26]gfortran's header file or use the [27]Chasm 4214 Language Interoperability Tools. 4215 4216 Go 4217 4218 * GCC 4.8.2 provides a complete implementation of the Go 1.1.2 4219 release. 4220 * GCC 4.8.0 and 4.8.1 implement a preliminary version of the Go 1.1 4221 release. The library support is not quite complete. 4222 * Go has been tested on GNU/Linux and Solaris platforms for various 4223 processors including x86, x86_64, PowerPC, SPARC, and Alpha. It may 4224 work on other platforms as well. 4225 4226 New Targets and Target Specific Improvements 4227 4228 AArch64 4229 4230 * A new port has been added to support AArch64, the new 64-bit 4231 architecture from ARM. Note that this is a separate port from the 4232 existing 32-bit ARM port. 4233 * The port provides initial support for the Cortex-A53 and the 4234 Cortex-A57 processors with the command line options 4235 -mcpu=cortex-a53 and -mcpu=cortex-a57. 4236 * As of GCC 4.8.4 a workaround for the ARM Cortex-A53 erratum 835769 4237 has been added and can be enabled by giving the 4238 -mfix-cortex-a53-835769 option. Alternatively it can be enabled by 4239 default by configuring GCC with the --enable-fix-cortex-a53-835769 4240 option. 4241 4242 ARM 4243 4244 * Initial support has been added for the AArch32 extensions defined 4245 in the ARMv8 architecture. 4246 * Code generation improvements for the Cortex-A7 and Cortex-A15 CPUs. 4247 * A new option, -mcpu=marvell-pj4, has been added to generate code 4248 for the Marvell PJ4 processor. 4249 * The compiler can now automatically generate the VFMA, VFMS, REVSH 4250 and REV16 instructions. 4251 * A new vectorizer cost model for Advanced SIMD configurations to 4252 improve the auto-vectorization strategies used. 4253 * The scheduler now takes into account the number of live registers 4254 to reduce the amount of spilling that can occur. This should 4255 improve code performance in large functions. The limit can be 4256 removed by using the option -fno-sched-pressure. 4257 * Improvements have been made to the Marvell iWMMX code generation 4258 and support for the iWMMX2 SIMD unit has been added. The option 4259 -mcpu=iwmmxt2 can be used to enable code generation for the latter. 4260 * A number of code generation improvements for Thumb2 to reduce code 4261 size when compiling for the M-profile processors. 4262 * The RTEMS (arm-rtems) port has been updated to use the EABI. 4263 * Code generation support for the old FPA and Maverick floating-point 4264 architectures has been removed. Ports that previously relied on 4265 these features have also been removed. This includes the targets: 4266 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 4267 + arm*-*-elf (use arm*-*-eabi) 4268 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 4269 + arm*-*-ecos-elf (no alternative) 4270 + arm*-*-freebsd (no alternative) 4271 + arm*-wince-pe* (no alternative). 4272 4273 AVR 4274 4275 * Support for the "Embedded C" fixed-point has been added. For 4276 details, see the [28]GCC wiki and the [29]user manual. The support 4277 is not complete. 4278 * A new print modifier %r for register operands in inline assembler 4279 is supported. It will print the raw register number without the 4280 register prefix 'r': 4281 /* Return the most significant byte of 'val', a 64-bit value. */ 4282 4283 unsigned char msb (long long val) 4284 { 4285 unsigned char c; 4286 __asm__ ("mov %0, %r1+7" : "=r" (c) : "r" (val)); 4287 return c; 4288 } 4289 The inline assembler in this example will generate code like 4290 mov r24, 8+7 4291 provided c is allocated to R24 and val is allocated to R8...R15. 4292 This works because the GNU assembler accepts plain register numbers 4293 without register prefix. 4294 * Static initializers with 3-byte symbols are supported now: 4295 extern const __memx char foo; 4296 const __memx void *pfoo = &foo; 4297 This requires at least Binutils 2.23. 4298 4299 IA-32/x86-64 4300 4301 * Allow -mpreferred-stack-boundary=3 for the x86-64 architecture with 4302 SSE extensions disabled. Since the x86-64 ABI requires 16 byte 4303 stack alignment, this is ABI incompatible and intended to be used 4304 in controlled environments where stack space is an important 4305 limitation. This option will lead to wrong code when functions 4306 compiled with 16 byte stack alignment (such as functions from a 4307 standard library) are called with misaligned stack. In this case, 4308 SSE instructions may lead to misaligned memory access traps. In 4309 addition, variable arguments will be handled incorrectly for 16 4310 byte aligned objects (including x87 long double and __int128), 4311 leading to wrong results. You must build all modules with 4312 -mpreferred-stack-boundary=3, including any libraries. This 4313 includes the system libraries and startup modules. 4314 * Support for the new Intel processor codename Broadwell with RDSEED, 4315 ADCX, ADOX, PREFETCHW is available through -madx, -mprfchw, 4316 -mrdseed command-line options. 4317 * Support for the Intel RTM and HLE intrinsics, built-in functions 4318 and code generation is available via -mrtm and -mhle. 4319 * Support for the Intel FXSR, XSAVE and XSAVEOPT instruction sets. 4320 Intrinsics and built-in functions are available via -mfxsr, -mxsave 4321 and -mxsaveopt respectively. 4322 * New -maddress-mode=[short|long] options for x32. 4323 -maddress-mode=short overrides default 64-bit addresses to 32-bit 4324 by emitting the 0x67 address-size override prefix. This is the 4325 default address mode for x32. 4326 * New built-in functions to detect run-time CPU type and ISA: 4327 + A built-in function __builtin_cpu_is has been added to detect 4328 if the run-time CPU is of a particular type. It returns a 4329 positive integer on a match and zero otherwise. It accepts one 4330 string literal argument, the CPU name. For example, 4331 __builtin_cpu_is("westmere") returns a positive integer if the 4332 run-time CPU is an Intel Core i7 Westmere processor. Please 4333 refer to the [30]user manual for the list of valid CPU names 4334 recognized. 4335 + A built-in function __builtin_cpu_supports has been added to 4336 detect if the run-time CPU supports a particular ISA feature. 4337 It returns a positive integer on a match and zero otherwise. 4338 It accepts one string literal argument, the ISA feature. For 4339 example, __builtin_cpu_supports("ssse3") returns a positive 4340 integer if the run-time CPU supports SSSE3 instructions. 4341 Please refer to the [31]user manual for the list of valid ISA 4342 names recognized. 4343 Caveat: If these built-in functions are called before any static 4344 constructors are invoked, like during IFUNC initialization, then 4345 the CPU detection initialization must be explicitly run using this 4346 newly provided built-in function, __builtin_cpu_init. The 4347 initialization needs to be done only once. For example, this is how 4348 the invocation would look like inside an IFUNC initializer: 4349 static void (*some_ifunc_resolver(void))(void) 4350 { 4351 __builtin_cpu_init(); 4352 if (__builtin_cpu_is("amdfam10h") ... 4353 if (__builtin_cpu_supports("popcnt") ... 4354 } 4355 4356 * Function Multiversioning Support with G++: 4357 It is now possible to create multiple function versions each 4358 targeting a specific processor and/or ISA. Function versions have 4359 the same signature but different target attributes. For example, 4360 here is a program with function versions: 4361 __attribute__ ((target ("default"))) 4362 int foo(void) 4363 { 4364 return 1; 4365 } 4366 4367 __attribute__ ((target ("sse4.2"))) 4368 int foo(void) 4369 { 4370 return 2; 4371 } 4372 4373 int main (void) 4374 { 4375 int (*p) = &foo; 4376 assert ((*p)() == foo()); 4377 return 0; 4378 } 4379 4380 Please refer to this [32]wiki for more information. 4381 * The x86 back end has been improved to allow option -fschedule-insns 4382 to work reliably. This option can be used to schedule instructions 4383 better and leads to improved performace in certain cases. 4384 * Windows MinGW-w64 targets (*-w64-mingw*) require at least r5437 4385 from the Mingw-w64 trunk. 4386 * Support for new AMD family 15h processors (Steamroller core) is now 4387 available through the -march=bdver3 and -mtune=bdver3 options. 4388 * Support for new AMD family 16h processors (Jaguar core) is now 4389 available through the -march=btver2 and -mtune=btver2 options. 4390 4391 FRV 4392 4393 * This target now supports the -fstack-usage command-line option. 4394 4395 MIPS 4396 4397 * GCC can now generate code specifically for the R4700, Broadcom XLP 4398 and MIPS 34kn processors. The associated -march options are 4399 -march=r4700, -march=xlp and -march=34kn respectively. 4400 * GCC now generates better DSP code for MIPS 74k cores thanks to 4401 further scheduling optimizations. 4402 * The MIPS port now supports the -fstack-check option. 4403 * GCC now passes the -mmcu and -mno-mcu options to the assembler. 4404 * Previous versions of GCC would silently accept -fpic and -fPIC for 4405 -mno-abicalls targets like mips*-elf. This combination was not 4406 intended or supported, and did not generate position-independent 4407 code. GCC 4.8 now reports an error when this combination is used. 4408 4409 PowerPC / PowerPC64 / RS6000 4410 4411 * SVR4 configurations (GNU/Linux, FreeBSD, NetBSD) no longer save, 4412 restore or update the VRSAVE register by default. The respective 4413 operating systems manage the VRSAVE register directly. 4414 * Large TOC support has been added for AIX through the command line 4415 option -mcmodel=large. 4416 * Native Thread-Local Storage support has been added for AIX. 4417 * VMX (Altivec) and VSX instruction sets now are enabled implicitly 4418 when targetting processors that support those hardware features on 4419 AIX 6.1 and above. 4420 4421 RX 4422 4423 * This target will now issue a warning message whenever multiple fast 4424 interrupt handlers are found in the same compilation unit. This 4425 feature can be turned off by the new 4426 -mno-warn-multiple-fast-interrupts command-line option. 4427 4428 S/390, System z 4429 4430 * Support for the IBM zEnterprise zEC12 processor has been added. 4431 When using the -march=zEC12 option, the compiler will generate code 4432 making use of the following new instructions: 4433 + load and trap instructions 4434 + 2 new compare and trap instructions 4435 + rotate and insert selected bits - without CC clobber 4436 The -mtune=zEC12 option enables zEC12 specific instruction 4437 scheduling without making use of new instructions. 4438 * Register pressure sensitive instruction scheduling is enabled by 4439 default. 4440 * The ifunc function attribute is enabled by default. 4441 * memcpy and memcmp invokations on big memory chunks or with run time 4442 lengths are not generated inline anymore when tuning for z10 or 4443 higher. The purpose is to make use of the IFUNC optimized versions 4444 in Glibc. 4445 4446 SH 4447 4448 * The default alignment settings have been reduced to be less 4449 aggressive. This results in more compact code for optimization 4450 levels other than -Os. 4451 * Improved support for the __atomic built-in functions: 4452 + A new option -matomic-model=model selects the model for the 4453 generated atomic sequences. The following models are 4454 supported: 4455 4456 soft-gusa 4457 Software gUSA sequences (SH3* and SH4* only). On 4458 SH4A targets this will now also partially utilize 4459 the movco.l and movli.l instructions. This is the 4460 default when the target is sh3*-*-linux* or 4461 sh4*-*-linux*. 4462 4463 hard-llcs 4464 Hardware movco.l / movli.l sequences (SH4A only). 4465 4466 soft-tcb 4467 Software thread control block sequences. 4468 4469 soft-imask 4470 Software interrupt flipping sequences (privileged 4471 mode only). This is the default when the target is 4472 sh1*-*-linux* or sh2*-*-linux*. 4473 4474 none 4475 Generates function calls to the respective __atomic 4476 built-in functions. This is the default for SH64 4477 targets or when the target is not sh*-*-linux*. 4478 4479 + The option -msoft-atomic has been deprecated. It is now an 4480 alias for -matomic-model=soft-gusa. 4481 + A new option -mtas makes the compiler generate the tas.b 4482 instruction for the __atomic_test_and_set built-in function 4483 regardless of the selected atomic model. 4484 + The __sync functions in libgcc now reflect the selected atomic 4485 model when building the toolchain. 4486 * Added support for the mov.b and mov.w instructions with 4487 displacement addressing. 4488 * Added support for the SH2A instructions movu.b and movu.w. 4489 * Various improvements to code generated for integer arithmetic. 4490 * Improvements to conditional branches and code that involves the T 4491 bit. A new option -mzdcbranch tells the compiler to favor 4492 zero-displacement branches. This is enabled by default for SH4* 4493 targets. 4494 * The pref instruction will now be emitted by the __builtin_prefetch 4495 built-in function for SH3* targets. 4496 * The fmac instruction will now be emitted by the fmaf standard 4497 function and the __builtin_fmaf built-in function. 4498 * The -mfused-madd option has been deprecated in favor of the 4499 machine-independent -ffp-contract option. Notice that the fmac 4500 instruction will now be generated by default for expressions like a 4501 * b + c. This is due to the compiler default setting 4502 -ffp-contract=fast. 4503 * Added new options -mfsrra and -mfsca to allow the compiler using 4504 the fsrra and fsca instructions on targets other than SH4A (where 4505 they are already enabled by default). 4506 * Added support for the __builtin_bswap32 built-in function. It is 4507 now expanded as a sequence of swap.b and swap.w instructions 4508 instead of a library function call. 4509 * The behavior of the -mieee option has been fixed and the negative 4510 form -mno-ieee has been added to control the IEEE conformance of 4511 floating point comparisons. By default -mieee is now enabled and 4512 the option -ffinite-math-only implicitly sets -mno-ieee. 4513 * Added support for the built-in functions __builtin_thread_pointer 4514 and __builtin_set_thread_pointer. This assumes that GBR is used to 4515 hold the thread pointer of the current thread. Memory loads and 4516 stores relative to the address returned by __builtin_thread_pointer 4517 will now also utilize GBR based displacement address modes. 4518 * The -mdiv= option for targets other than SHmedia has been fixed and 4519 documented. 4520 4521 SPARC 4522 4523 * Added optimized instruction scheduling for Niagara4. 4524 4525 TILE-Gx 4526 4527 * Added support for the -mcmodel=MODEL command-line option. The 4528 models supported are small and large. 4529 4530 V850 4531 4532 * This target now supports the E3V5 architecture via the use of the 4533 new -mv850e3v5 command-line option. It also has experimental 4534 support for the e3v5 LOOP instruction which can be enabled via the 4535 new -mloop command-line option. 4536 4537 XStormy16 4538 4539 * This target now supports the -fstack-usage command-line option. 4540 4541 Operating Systems 4542 4543 OpenBSD 4544 4545 * Support for OpenBSD/amd64 (x86_64-*-openbsd*) has been added and 4546 support for OpenBSD/i386 (i386-*-openbsd*) has been rejuvenated. 4547 4548 Windows (Cygwin) 4549 4550 * Executables are now linked against shared libgcc by default. The 4551 previous default was to link statically, which can still be done by 4552 explicitly specifying -static or static-libgcc on the command line. 4553 However it is strongly advised against, as it will cause problems 4554 for any application that makes use of DLLs compiled by GCC. It 4555 should be alright for a monolithic stand-alone application that 4556 only links against the Windows DLLs, but offers little or no 4557 benefit. 4558 4559 GCC 4.8.1 4560 4561 This is the [33]list of problem reports (PRs) from GCC's bug tracking 4562 system that are known to be fixed in the 4.8.1 release. This list might 4563 not be complete (that is, it is possible that some PRs that have been 4564 fixed are not listed here). 4565 4566 The C++11 <chrono> std::chrono::system_clock and 4567 std::chrono::steady_clock classes have changed ABI in GCC 4.8.1, they 4568 both are now separate (never typedefs of each other), both use 4569 std::chrono::nanoseconds resolution, on most GNU/Linux configurations 4570 std::chrono::steady_clock is now finally monotonic, and both classes 4571 are mangled differently than in the previous GCC releases. 4572 std::chrono::system_clock::now() with std::chrono::microseconds resp. 4573 std::chrono::seconds resolution is still exported for backwards 4574 compatibility with default configured libstdc++. Note that libstdc++ 4575 configured with --enable-libstdcxx-time= used to be ABI incompatible 4576 with default configured libstdc++ for those two classes and no ABI 4577 compatibility can be offered for those configurations, so any C++11 4578 code that uses those classes and has been compiled and linked against 4579 libstdc++ configured with the non-default --enable-libstdcxx-time= 4580 configuration option needs to be recompiled. 4581 4582 GCC 4.8.2 4583 4584 This is the [34]list of problem reports (PRs) from GCC's bug tracking 4585 system that are known to be fixed in the 4.8.2 release. This list might 4586 not be complete (that is, it is possible that some PRs that have been 4587 fixed are not listed here). 4588 4589 GCC 4.8.3 4590 4591 This is the [35]list of problem reports (PRs) from GCC's bug tracking 4592 system that are known to be fixed in the 4.8.3 release. This list might 4593 not be complete (that is, it is possible that some PRs that have been 4594 fixed are not listed here). 4595 4596 Support for the new powerpc64le-linux platform has been added. It 4597 defaults to generating code that conforms to the ELFV2 ABI. 4598 4599 GCC 4.8.4 4600 4601 This is the [36]list of problem reports (PRs) from GCC's bug tracking 4602 system that are known to be fixed in the 4.8.4 release. This list might 4603 not be complete (that is, it is possible that some PRs that have been 4604 fixed are not listed here). 4605 4606 GCC 4.8.5 4607 4608 This is the [37]list of problem reports (PRs) from GCC's bug tracking 4609 system that are known to be fixed in the 4.8.5 release. This list might 4610 not be complete (that is, it is possible that some PRs that have been 4611 fixed are not listed here). 4612 4613 4614 For questions related to the use of GCC, please consult these web 4615 pages and the [38]GCC manuals. If that fails, the 4616 [39]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4617 web pages and the development of GCC are welcome on our developer 4618 list at [40]gcc (a] gcc.gnu.org. All of [41]our lists have public 4619 archives. 4620 4621 Copyright (C) [42]Free Software Foundation, Inc. Verbatim copying and 4622 distribution of this entire article is permitted in any medium, 4623 provided this notice is preserved. 4624 4625 These pages are [43]maintained by the GCC team. Last modified 4626 2018-09-30[44]. 4627 4628 References 4629 4630 1. https://gcc.gnu.org/wiki/cxx-conversion 4631 2. ftp://gcc.gnu.org/pub/gcc/infrastructure/ 4632 3. http://www.nongnu.org/avr-libc/ 4633 4. https://gcc.gnu.org/PR54461 4634 5. https://gcc.gnu.org/gcc-4.8/porting_to.html 4635 6. https://github.com/google/sanitizers 4636 7. https://code.google.com/archive/p/data-race-test/wikis/ThreadSanitizer.wiki 4637 8. https://lkml.org/lkml/2006/11/28/239 4638 9. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 4639 10. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 4640 11. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 4641 12. http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2011/n3276.pdf 4642 13. http://gcc.gnu.org/gcc-4.8/cxx0x_status.html 4643 14. http://www.open-std.org/JTC1/SC22/WG21/docs/papers/2012/n3386.html 4644 15. http://gcc.gnu.org/projects/cxx1y.html 4645 16. https://gcc.gnu.org/onlinedocs/gcc-4.8.4/libstdc++/manual/manual/status.html#status.iso.2011 4646 17. https://gcc.gnu.org/onlinedocs/gfortran/BACKTRACE.html 4647 18. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 4648 19. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 4649 20. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 4650 21. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 4651 22. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html 4652 23. https://gcc.gnu.org/onlinedocs/gfortran/TMPDIR.html 4653 24. https://gcc.gnu.org/wiki/Fortran2003Status 4654 25. https://gcc.gnu.org/wiki/TS29113Status 4655 26. https://gcc.gnu.org/viewcvs/trunk/libgfortran/libgfortran.h?content-type=text%2Fplain&view=co 4656 27. http://chasm-interop.sourceforge.net/ 4657 28. https://gcc.gnu.org/wiki/avr-gcc#Fixed-Point_Support 4658 29. https://gcc.gnu.org/onlinedocs/gcc/Fixed-Point.html 4659 30. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 4660 31. https://gcc.gnu.org/onlinedocs/gcc/x86-Built-in-Functions.html 4661 32. https://gcc.gnu.org/wiki/FunctionMultiVersioning 4662 33. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.1 4663 34. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.2 4664 35. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.3 4665 36. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.4 4666 37. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.8.5 4667 38. https://gcc.gnu.org/onlinedocs/ 4668 39. mailto:gcc-help (a] gcc.gnu.org 4669 40. mailto:gcc (a] gcc.gnu.org 4670 41. https://gcc.gnu.org/lists.html 4671 42. https://www.fsf.org/ 4672 43. https://gcc.gnu.org/about.html 4673 44. http://validator.w3.org/check/referer 4674 ====================================================================== 4675 http://gcc.gnu.org/gcc-4.7/index.html 4676 4677 GCC 4.7 Release Series 4678 4679 June 12, 2014 4680 4681 The [1]GNU project and the GCC developers are pleased to announce the 4682 release of GCC 4.7.4. 4683 4684 This release is a bug-fix release, containing fixes for regressions in 4685 GCC 4.7.3 relative to previous releases of GCC. 4686 4687 This release series is no longer maintained. 4688 4689 Release History 4690 4691 GCC 4.7.4 4692 June 12, 2014 ([2]changes, [3]documentation) 4693 4694 GCC 4.7.3 4695 April 11, 2013 ([4]changes, [5]documentation) 4696 4697 GCC 4.7.2 4698 September 20, 2012 ([6]changes, [7]documentation) 4699 4700 GCC 4.7.1 4701 June 14, 2012 ([8]changes, [9]documentation) 4702 4703 GCC 4.7.0 4704 March 22, 2012 ([10]changes, [11]documentation) 4705 4706 References and Acknowledgements 4707 4708 GCC used to stand for the GNU C Compiler, but since the compiler 4709 supports several other languages aside from C, it now stands for the 4710 GNU Compiler Collection. 4711 4712 A list of [12]successful builds is updated as new information becomes 4713 available. 4714 4715 The GCC developers would like to thank the numerous people that have 4716 contributed new features, improvements, bug fixes, and other changes as 4717 well as test results to GCC. This [13]amazing group of volunteers is 4718 what makes GCC successful. 4719 4720 For additional information about GCC please refer to the [14]GCC 4721 project web site or contact the [15]GCC development mailing list. 4722 4723 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 4724 4725 4726 For questions related to the use of GCC, please consult these web 4727 pages and the [18]GCC manuals. If that fails, the 4728 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 4729 web pages and the development of GCC are welcome on our developer 4730 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 4731 archives. 4732 4733 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 4734 distribution of this entire article is permitted in any medium, 4735 provided this notice is preserved. 4736 4737 These pages are [23]maintained by the GCC team. Last modified 4738 2018-09-30[24]. 4739 4740 References 4741 4742 1. http://www.gnu.org/ 4743 2. http://gcc.gnu.org/gcc-4.7/changes.html 4744 3. https://gcc.gnu.org/onlinedocs/4.7.4/ 4745 4. http://gcc.gnu.org/gcc-4.7/changes.html 4746 5. https://gcc.gnu.org/onlinedocs/4.7.3/ 4747 6. http://gcc.gnu.org/gcc-4.7/changes.html 4748 7. https://gcc.gnu.org/onlinedocs/4.7.2/ 4749 8. http://gcc.gnu.org/gcc-4.7/changes.html 4750 9. https://gcc.gnu.org/onlinedocs/4.7.1/ 4751 10. http://gcc.gnu.org/gcc-4.7/changes.html 4752 11. https://gcc.gnu.org/onlinedocs/4.7.0/ 4753 12. http://gcc.gnu.org/gcc-4.7/buildstat.html 4754 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 4755 14. http://gcc.gnu.org/index.html 4756 15. mailto:gcc (a] gcc.gnu.org 4757 16. http://gcc.gnu.org/mirrors.html 4758 17. http://gcc.gnu.org/svn.html 4759 18. https://gcc.gnu.org/onlinedocs/ 4760 19. mailto:gcc-help (a] gcc.gnu.org 4761 20. mailto:gcc (a] gcc.gnu.org 4762 21. https://gcc.gnu.org/lists.html 4763 22. https://www.fsf.org/ 4764 23. https://gcc.gnu.org/about.html 4765 24. http://validator.w3.org/check/referer 4766 ====================================================================== 4767 http://gcc.gnu.org/gcc-4.7/changes.html 4768 4769 GCC 4.7 Release Series 4770 Changes, New Features, and Fixes 4771 4772 Caveats 4773 4774 * The -fconserve-space flag has been deprecated. The flag had no 4775 effect for most targets: only targets without a global .bss section 4776 and without support for switchable sections. Furthermore, the flag 4777 only had an effect for G++, where it could result in wrong 4778 semantics (please refer to the GCC manual for further details). The 4779 flag will be removed in GCC 4.8 4780 * Support for a number of older systems and recently unmaintained or 4781 untested target ports of GCC has been declared obsolete in GCC 4.7. 4782 Unless there is activity to revive them, the next release of GCC 4783 will have their sources permanently removed. 4784 All GCC ports for the following processor architectures have been 4785 declared obsolete: 4786 + picoChip (picochip-*) 4787 The following ports for individual systems on particular 4788 architectures have been obsoleted: 4789 + IRIX 6.5 (mips-sgi-irix6.5) 4790 + MIPS OpenBSD (mips*-*-openbsd*) 4791 + Solaris 8 (*-*-solaris2.8). Details can be found in the 4792 [1]announcement. 4793 + Tru64 UNIX V5.1 (alpha*-dec-osf5.1*) 4794 * On ARM, when compiling for ARMv6 (but not ARMv6-M), ARMv7-A, 4795 ARMv7-R, or ARMv7-M, the new option -munaligned-access is active by 4796 default, which for some sources generates code that accesses memory 4797 on unaligned addresses. This requires the kernel of those systems 4798 to enable such accesses (controlled by CP15 register c1, refer to 4799 ARM documentation). Alternatively, or for compatibility with 4800 kernels where unaligned accesses are not supported, all code has to 4801 be compiled with -mno-unaligned-access. Upstream Linux kernel 4802 releases have automatically and unconditionally supported unaligned 4803 accesses as emitted by GCC due to this option being active since 4804 version 2.6.28. 4805 * Support on ARM for the legacy floating-point accelerator (FPA) and 4806 the mixed-endian floating-point format that it used has been 4807 obsoleted. The ports that still use this format have been obsoleted 4808 as well. Many legacy ARM ports already provide an alternative that 4809 uses the VFP floating-point format. The obsolete ports will be 4810 deleted in the next release. 4811 The obsolete ports with alternatives are: 4812 + arm*-*-rtems (use arm*-*-rtemseabi) 4813 + arm*-*-linux-gnu (use arm*-*-linux-gnueabi) 4814 + arm*-*-elf (use arm*-*-eabi) 4815 + arm*-*-uclinux* (use arm*-*-uclinux*eabi) 4816 Note, however, that these alternatives are not binary compatible 4817 with their legacy counterparts (although some can support running 4818 legacy applications). 4819 The obsolete ports that currently lack a modern alternative are: 4820 + arm*-*-ecos-elf 4821 + arm*-*-freebsd 4822 + arm*-wince-pe* 4823 New ports that support more recent versions of the architecture are 4824 welcome. 4825 * Support for the Maverick co-processor on ARM has been obsoleted. 4826 Code to support it will be deleted in the next release. 4827 * Support has been removed for Unix International threads on Solaris 4828 2, so the --enable-threads=solaris configure option and the 4829 -threads compiler option don't work any longer. 4830 * Support has been removed for the Solaris BSD Compatibility Package, 4831 which lives in /usr/ucbinclude and /usr/ucblib. It has been removed 4832 from Solaris 11, and was only intended as a migration aid from 4833 SunOS 4 to SunOS 5. The -compat-bsd compiler option is not 4834 recognized any longer. 4835 * The AVR port's libgcc has been improved and its multilib structure 4836 has been enhanced. As a result, all objects contributing to an 4837 application must either be compiled with GCC versions up to 4.6.x 4838 or with GCC versions 4.7.1 or later. If the compiler is used with 4839 AVR Libc, you need a version that supports the new layout, i.e. 4840 implements [2]#35407. 4841 * The AVR port's -mshort-calls command-line option has been 4842 deprecated. It will be removed in the GCC 4.8 release. See -mrelax 4843 for a replacement. 4844 * The AVR port only references startup code that clears .bss and the 4845 common section resp. initializes the .data and .rodata section 4846 provided respective sections (or subsections thereof) are not 4847 empty, see [3]PR18145. Applications that put all static storage 4848 objects into non-standard sections and / or define all static 4849 storage objects in assembler modules, must reference __do_clear_bss 4850 resp. __do_copy_data by hand or undefine the symbol(s) by means of 4851 -Wl,-u,__do_clear_bss resp. -Wl,-u,__do_copy_data. 4852 * The ARM port's -mwords-little-endian option has been deprecated. It 4853 will be removed in a future release. 4854 * Support has been removed for the NetWare x86 configuration 4855 obsoleted in GCC 4.6. 4856 * It is no longer possible to use the "l" constraint in MIPS16 asm 4857 statements. 4858 * GCC versions 4.7.0 and 4.7.1 had changes to the C++ standard 4859 library which affected the ABI in C++11 mode: a data member was 4860 added to std::list changing its size and altering the definitions 4861 of some member functions, and std::pair's move constructor was 4862 non-trivial which altered the calling convention for functions with 4863 std::pair arguments or return types. The ABI incompatibilities have 4864 been fixed for GCC version 4.7.2 but as a result C++11 code 4865 compiled with GCC 4.7.0 or 4.7.1 may be incompatible with C++11 4866 code compiled with different GCC versions and with C++98/C++03 code 4867 compiled with any version. 4868 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 4869 rules for the layout of vectors that could lead to wrong code being 4870 generated. Vectors larger than 8 bytes in size are now by default 4871 aligned to an 8-byte boundary. This is an ABI change: code that 4872 makes explicit use of vector types may be incompatible with binary 4873 objects built with older versions of GCC. Auto-vectorized code is 4874 not affected by this change. (This change affects GCC versions 4875 4.7.2 and later.) 4876 * More information on porting to GCC 4.7 from previous versions of 4877 GCC can be found in the [4]porting guide for this release. 4878 4879 General Optimizer Improvements 4880 4881 * Support for a new parameter --param case-values-threshold=n was 4882 added to allow users to control the cutoff between doing switch 4883 statements as a series of if statements and using a jump table. 4884 * Link-time optimization (LTO) improvements: 4885 + Improved scalability and reduced memory usage. Link time 4886 optimization of Firefox now requires 3GB of RAM on a 64-bit 4887 system, while over 8GB was needed previously. Linking time has 4888 been improved, too. The serial stage of linking Firefox has 4889 been sped up by about a factor of 10. 4890 + Reduced size of object files and temporary storage used during 4891 linking. 4892 + Streaming performance (both outbound and inbound) has been 4893 improved. 4894 + ld -r is now supported with LTO. 4895 + Several bug fixes, especially in symbol table handling and 4896 merging. 4897 * Interprocedural optimization improvements: 4898 + Heuristics now take into account that after inlining code will 4899 be optimized out because of known values (or properties) of 4900 function parameters. For example: 4901 void foo(int a) 4902 { 4903 if (a > 10) 4904 ... huge code ... 4905 } 4906 void bar (void) 4907 { 4908 foo (0); 4909 } 4910 4911 The call of foo will be inlined into bar even when optimizing 4912 for code size. Constructs based on __builtin_constant_p are 4913 now understood by the inliner and code size estimates are 4914 evaluated a lot more realistically. 4915 + The representation of C++ virtual thunks and aliases (both 4916 implicit and defined via the alias attribute) has been 4917 re-engineered. Aliases no longer pose optimization barriers 4918 and calls to an alias can be inlined and otherwise optimized. 4919 + The inter-procedural constant propagation pass has been 4920 rewritten. It now performs generic function specialization. 4921 For example when compiling the following: 4922 void foo(bool flag) 4923 { 4924 if (flag) 4925 ... do something ... 4926 else 4927 ... do something else ... 4928 } 4929 void bar (void) 4930 { 4931 foo (false); 4932 foo (true); 4933 foo (false); 4934 foo (true); 4935 foo (false); 4936 foo (true); 4937 } 4938 4939 GCC will now produce two copies of foo. One with flag being 4940 true, while other with flag being false. This leads to 4941 performance improvements previously possible only by inlining 4942 all calls. Cloning causes a lot less code size growth. 4943 * A string length optimization pass has been added. It attempts to 4944 track string lengths and optimize various standard C string 4945 functions like strlen, strchr, strcpy, strcat, stpcpy and their 4946 _FORTIFY_SOURCE counterparts into faster alternatives. This pass is 4947 enabled by default at -O2 or above, unless optimizing for size, and 4948 can be disabled by the -fno-optimize-strlen option. The pass can 4949 e.g. optimize 4950 char *bar (const char *a) 4951 { 4952 size_t l = strlen (a) + 2; 4953 char *p = malloc (l); if (p == NULL) return p; 4954 strcpy (p, a); strcat (p, "/"); return p; 4955 } 4956 4957 into: 4958 char *bar (const char *a) 4959 { 4960 size_t tmp = strlen (a); 4961 char *p = malloc (tmp + 2); if (p == NULL) return p; 4962 memcpy (p, a, tmp); memcpy (p + tmp, "/", 2); return p; 4963 } 4964 4965 or for hosted compilations where stpcpy is available in the runtime 4966 and headers provide its prototype, e.g. 4967 void foo (char *a, const char *b, const char *c, const char *d) 4968 { 4969 strcpy (a, b); strcat (a, c); strcat (a, d); 4970 } 4971 4972 can be optimized into: 4973 void foo (char *a, const char *b, const char *c, const char *d) 4974 { 4975 strcpy (stpcpy (stpcpy (a, b), c), d); 4976 } 4977 4978 New Languages and Language specific improvements 4979 4980 * Version 3.1 of the [5]OpenMP specification is now supported for the 4981 C, C++, and Fortran compilers. 4982 4983 Ada 4984 4985 * The command-line option -feliminate-unused-debug-types has been 4986 re-enabled by default, as it is for the other languages, leading to 4987 a reduction in debug info size of 12.5% and more for relevant 4988 cases, as well as to a small compilation speedup. 4989 4990 C family 4991 4992 * A new built-in, __builtin_assume_aligned, has been added, through 4993 which the compiler can be hinted about pointer alignment and can 4994 use it to improve generated code. 4995 * A new warning option -Wunused-local-typedefs was added for C, C++, 4996 Objective-C and Objective-C++. This warning diagnoses typedefs 4997 locally defined in a function, and otherwise not used. 4998 * A new experimental command-line option -ftrack-macro-expansion was 4999 added for C, C++, Objective-C, Objective-C++ and Fortran. It allows 5000 the compiler to emit diagnostic about the current macro expansion 5001 stack when a compilation error occurs in a macro expansion. 5002 * Experimental support for transactional memory has been added. It 5003 includes support in the compiler, as well as a supporting runtime 5004 library called libitm. To compile code with transactional memory 5005 constructs, use the -fgnu-tm option. 5006 Support is currently available for Alpha, ARM, PowerPC, SH, SPARC, 5007 and 32-bit/64-bit x86 platforms. 5008 For more details on transactional memory see [6]the GCC WiKi. 5009 * Support for atomic operations specifying the C++11/C11 memory model 5010 has been added. These new __atomic routines replace the existing 5011 __sync built-in routines. 5012 Atomic support is also available for memory blocks. Lock-free 5013 instructions will be used if a memory block is the same size and 5014 alignment as a supported integer type. Atomic operations which do 5015 not have lock-free support are left as function calls. A set of 5016 library functions is available on the GCC atomic wiki in the 5017 "External Atomics Library" section. 5018 For more details on the memory models and features, see the 5019 [7]atomic wiki. 5020 * When a binary operation is performed on vector types and one of the 5021 operands is a uniform vector, it is possible to replace the vector 5022 with the generating element. For example: 5023 typedef int v4si __attribute__ ((vector_size (16))); 5024 v4si res, a = {1,2,3,4}; 5025 int x; 5026 5027 res = 2 + a; /* means {2,2,2,2} + a */ 5028 res = a - x; /* means a - {x,x,x,x} */ 5029 5030 C 5031 5032 * There is support for some more features from the C11 revision of 5033 the ISO C standard. GCC now accepts the options -std=c11 and 5034 -std=gnu11, in addition to the previous -std=c1x and -std=gnu1x. 5035 + Unicode strings (previously supported only with options such 5036 as -std=gnu11, now supported with -std=c11), and the 5037 predefined macros __STDC_UTF_16__ and __STDC_UTF_32__. 5038 + Nonreturning functions (_Noreturn and <stdnoreturn.h>). 5039 + Alignment support (_Alignas, _Alignof, max_align_t, 5040 <stdalign.h>). 5041 + A built-in function __builtin_complex is provided to support C 5042 library implementation of the CMPLX family of macros. 5043 5044 C++ 5045 5046 * G++ now accepts the -std=c++11, -std=gnu++11, and -Wc++11-compat 5047 options, which are equivalent to -std=c++0x, -std=gnu++0x, and 5048 -Wc++0x-compat, respectively. 5049 * G++ now implements [8]C++11 extended friend syntax: 5050 5051 template<class W> 5052 class Q 5053 { 5054 static const int I = 2; 5055 public: 5056 friend W; 5057 }; 5058 5059 struct B 5060 { 5061 int ar[Q<B>::I]; 5062 }; 5063 5064 * Thanks to Ville Voutilainen, G++ now implements [9]C++11 explicit 5065 override control. 5066 5067 struct B { 5068 virtual void f() const final; 5069 virtual void f(int); 5070 }; 5071 5072 struct D : B { 5073 void f() const; // error: D::f attempts to override final B::f 5074 void f(long) override; // error: doesn't override anything 5075 void f(int) override; // ok 5076 }; 5077 5078 struct E final { }; 5079 struct F: E { }; // error: deriving from final class 5080 5081 * G++ now implements [10]C++11 non-static data member initializers. 5082 5083 struct A { 5084 int i = 42; 5085 } a; // initializes a.i to 42 5086 5087 * Thanks to Ed Smith-Rowland, G++ now implements [11]C++11 5088 user-defined literals. 5089 5090 // Not actually a good approximation. :) 5091 constexpr long double operator"" _degrees (long double d) { return d * 0.0175; } 5092 long double pi = 180.0_degrees; 5093 5094 * G++ now implements [12]C++11 alias-declarations. 5095 5096 template <class T> using Ptr = T*; 5097 Ptr<int> ip; // decltype(ip) is int* 5098 5099 * Thanks to Ville Voutilainen and Pedro Lamarao, G++ now implements 5100 [13]C++11 delegating constructors. 5101 5102 struct A { 5103 A(int); 5104 A(): A(42) { } // delegate to the A(int) constructor 5105 }; 5106 5107 * G++ now fully implements C++11 atomic classes rather than just 5108 integer derived classes. 5109 5110 class POD { 5111 int a; 5112 int b; 5113 }; 5114 std::atomic<POD> my_atomic_POD; 5115 5116 * G++ now sets the predefined macro __cplusplus to the correct value, 5117 199711L for C++98/03, and 201103L for C++11. 5118 * G++ now correctly implements the two-phase lookup rules such that 5119 an unqualified name used in a template must have an appropriate 5120 declaration found either in scope at the point of definition of the 5121 template or by argument-dependent lookup at the point of 5122 instantiation. As a result, code that relies on a second 5123 unqualified lookup at the point of instantiation to find functions 5124 declared after the template or in dependent bases will be rejected. 5125 The compiler will suggest ways to fix affected code, and using the 5126 -fpermissive compiler flag will allow the code to compile with a 5127 warning. 5128 5129 template <class T> 5130 void f() { g(T()); } // error, g(int) not found by argument-dependent lookup 5131 void g(int) { } // fix by moving this declaration before the declaration of f 5132 5133 template <class T> 5134 struct A: T { 5135 // error, B::g(B) not found by argument-dependent lookup 5136 void f() { g(T()); } // fix by using this->g or A::g 5137 }; 5138 5139 struct B { void g(B); }; 5140 5141 int main() 5142 { 5143 f<int>(); 5144 A<B>().f(); 5145 } 5146 5147 * G++ now properly re-uses stack space allocated for temporary 5148 objects when their lifetime ends, which can significantly lower 5149 stack consumption for some C++ functions. As a result of this, some 5150 code with undefined behavior will now break: 5151 5152 const int &f(const int &i) { return i; } 5153 .... 5154 const int &x = f(1); 5155 const int &y = f(2); 5156 5157 Here, x refers to the temporary allocated to hold the 1 argument, 5158 which only lives until the end of the initialization; it 5159 immediately becomes a dangling reference. So the next statement 5160 re-uses the stack slot to hold the 2 argument, and users of x get 5161 that value instead. 5162 Note that this should not cause any change of behavior for 5163 temporaries of types with non-trivial destructors, as they are 5164 already destroyed at end of full-expression; the change is that now 5165 the storage is released as well. 5166 * A new command-line option -Wdelete-non-virtual-dtor has been added 5167 to warn when delete is used to destroy an instance of a class which 5168 has virtual functions and non-virtual destructor. It is unsafe to 5169 delete an instance of a derived class through a pointer to a base 5170 class if the base class does not have a virtual destructor. This 5171 warning is enabled by -Wall. 5172 * A new command-line option -Wzero-as-null-pointer-constant has been 5173 added to warn when a literal '0' is used as null pointer constant. 5174 It can be useful to facilitate the conversion to nullptr in C++11. 5175 * As per C++98, access-declarations are now deprecated by G++. 5176 Using-declarations are to be used instead. Furthermore, some 5177 efforts have been made to improve the support of class scope 5178 using-declarations. In particular, using-declarations referring to 5179 a dependent type now work as expected ([14]bug c++/14258). 5180 * The ELF symbol visibility of a template instantiation is now 5181 properly constrained by the visibility of its template arguments 5182 ([15]bug c++/35688). 5183 5184 Runtime Library (libstdc++) 5185 5186 * [16]Improved experimental support for the new ISO C++ standard, 5187 C++11, including: 5188 + using noexcept in most of the library; 5189 + implementations of pointer_traits, allocator_traits and 5190 scoped_allocator_adaptor; 5191 + uses-allocator construction for tuple; 5192 + vector meets the allocator-aware container requirements; 5193 + replacing monotonic_clock with steady_clock; 5194 + enabling the thread support library on most POSIX targets; 5195 + many small improvements to conform to the FDIS. 5196 * Added --enable-clocale=newlib configure option. 5197 * Debug Mode iterators for unordered associative containers. 5198 * Avoid polluting the global namespace and do not include <unistd.h>. 5199 5200 Fortran 5201 5202 * The compile flag [17]-fstack-arrays has been added, which causes 5203 all local arrays to be put on stack memory. For some programs this 5204 will improve the performance significantly. If your program uses 5205 very large local arrays, it is possible that you will have to 5206 extend your runtime limits for stack memory. 5207 * The [18]-Ofast flag now also implies [19]-fno-protect-parens and 5208 [20]-fstack-arrays. 5209 * Front-end optimizations can now be selected by the 5210 [21]-ffrontend-optimize option and deselected by the 5211 -fno-frontend-optimize option. 5212 * When front-end optimization removes a function call, 5213 [22]-Wfunction-elimination warns about that. 5214 * When performing front-end-optimization, the 5215 [23]-faggressive-function-elimination option allows the removal of 5216 duplicate function calls even for impure functions. 5217 * The flag [24]-Wreal-q-constant has been added, which warns if 5218 floating-point literals have been specified using q (such as 5219 1.0q0); the q marker is now supported as a vendor extension to 5220 denote quad precision (REAL(16) or, if not available, REAL(10)). 5221 Consider using a kind parameter (such as in 1.0_qp) instead, which 5222 can be obtained via [25]SELECTED_REAL_KIND. 5223 * The GFORTRAN_USE_STDERR environment variable has been removed. GNU 5224 Fortran now always prints error messages to standard error. If you 5225 wish to redirect standard error, please consult the manual for your 5226 OS, shell, batch environment etc. as appropriate. 5227 * The -fdump-core option and GFORTRAN_ERROR_DUMPCORE environment 5228 variable have been removed. When encountering a serious error, 5229 gfortran will now always abort the program. Whether a core dump is 5230 generated depends on the user environment settings; see the ulimit 5231 -c setting for POSIX shells, limit coredumpsize for C shells, and 5232 the [26]WER user-mode dumps settings on Windows. 5233 * The [27]-fbacktrace option is now enabled by default. When 5234 encountering a fatal error, gfortran will attempt to print a 5235 backtrace to standard error before aborting. It can be disabled 5236 with -fno-backtrace. Note: On POSIX targets with the addr2line 5237 utility from GNU binutils, GNU Fortran can print a backtrace with 5238 function name, file name, line number information in addition to 5239 the addresses; otherwise only the addresses are printed. 5240 * [28]Fortran 2003: 5241 + Generic interface names which have the same name as derived 5242 types are now supported, which allows to write constructor 5243 functions. Note that Fortran does not support static 5244 constructor functions; only default initialization or an 5245 explicit structure-constructor initialization are available. 5246 + [29]Polymorphic (class) arrays are now supported. 5247 * [30]Fortran 2008: 5248 + Support for the DO CONCURRENT construct has been added, which 5249 allows the user to specify that individual loop iterations 5250 have no interdependencies. 5251 + [31]Coarrays: Full single-image support except for polymorphic 5252 coarrays. Additionally, preliminary support for multiple 5253 images via an MPI-based [32]coarray communication library has 5254 been added. Note: The library version is not yet usable as 5255 remote coarray access is not yet possible. 5256 * [33]TS 29113: 5257 + New flag [34]-std=f2008ts permits programs that are expected 5258 to conform to the Fortran 2008 standard and the draft 5259 Technical Specification (TS) 29113 on Further Interoperability 5260 of Fortran with C. 5261 + The OPTIONAL attribute is now allowed for dummy arguments of 5262 BIND(C) procedures. 5263 + The RANK intrinsic has been added. 5264 + The implementation of the ASYNCHRONOUS attribute in GCC is 5265 compatible with the candidate draft of TS 29113 (since GCC 5266 4.6). 5267 5268 Go 5269 5270 * GCC 4.7 implements the [35]Go 1 language standard. The library 5271 support in 4.7.0 is not quite complete, due to release timing. 5272 Release 4.7.1 includes complete support for Go 1. The Go library is 5273 from the Go 1.0.1 release. 5274 * Go has been tested on GNU/Linux and Solaris platforms. It may work 5275 on other platforms as well. 5276 5277 New Targets and Target Specific Improvements 5278 5279 ARM 5280 5281 * GCC now supports the Cortex-A7 processor implementing the v7-a 5282 version of the architecture using the option -mcpu=cortex-a7. 5283 * The default vector size in auto-vectorization for NEON is now 128 5284 bits. If vectorization fails thusly, the vectorizer tries again 5285 with 64-bit vectors. 5286 * A new option -mvectorize-with-neon-double was added to allow users 5287 to change the vector size to 64 bits. 5288 5289 AVR 5290 5291 * GCC now supports the XMEGA architecture. This requires GNU binutils 5292 2.22 or later. 5293 * Support for the [36]named address spaces __flash, __flash1, ..., 5294 __flash5 and __memx has been added. These address spaces locate 5295 read-only data in flash memory and allow reading from flash memory 5296 by means of ordinary C code, i.e. without the need of (inline) 5297 assembler code: 5298 5299 const __flash int values[] = { 42, 31 }; 5300 5301 int add_values (const __flash int *p, int i) 5302 { 5303 return values[i] + *p; 5304 } 5305 5306 * Support has been added for the AVR-specific configure option 5307 --with-avrlibc=yes in order to arrange for better integration of 5308 [37]AVR-Libc. This configure option is supported in avr-gcc 4.7.2 5309 and newer and will only take effect in non-RTEMS configurations. If 5310 avr-gcc is configured for RTEMS, the option will be ignored which 5311 is the same as specifying --with-avrlibc=no. See [38]PR54461 for 5312 more technical details. 5313 * Support for AVR-specific [39]built-in functions has been added. 5314 * Support has been added for the signed and unsigned 24-bit scalar 5315 integer types __int24 and __uint24. 5316 * New command-line options -maccumulate-args, -mbranch-cost=cost and 5317 -mstrict-X were added to allow better fine-tuning of code 5318 optimization. 5319 * The command-line option -fdata-sections now also takes affect on 5320 the section names of variables with the progmem attribute. 5321 * A new inline assembler print modifier %i to print a RAM address as 5322 I/O address has been added: 5323 5324 #include <avr/io.h> /* Port Definitions from AVR-LibC */ 5325 5326 void set_portb (uint8_t value) 5327 { 5328 asm volatile ("out %i0, %1" :: "n" (&PORTB), "r" (value) : "memory"); 5329 } 5330 5331 The offset between an I/O address and the RAM address for that I/O 5332 location is device-specific. This offset is taken into account when 5333 printing a RAM address with the %i modifier so that the address is 5334 suitable to be used as operand in an I/O command. The address must 5335 be a constant integer known at compile time. 5336 * The inline assembler constraint "R" to represent integers in the 5337 range -6 ... 5 has been removed without replacement. 5338 * Many optimizations to: 5339 + 64-bit integer arithmetic 5340 + Widening multiplication 5341 + Integer division by a constant 5342 + Avoid constant reloading in multi-byte instructions. 5343 + Micro-optimizations for special instruction sequences. 5344 + Generic built-in functions like __builtin_ffs*, 5345 __builtin_clz*, etc. 5346 + If-else decision trees generated by switch instructions 5347 + Merging of data located in flash memory 5348 + New libgcc variants for devices with 8-bit wide stack pointer 5349 + ... 5350 * Better documentation: 5351 + Handling of EIND and indirect jumps on devices with more than 5352 128 KiB of program memory. 5353 + Handling of the RAMPD, RAMPX, RAMPY and RAMPZ special function 5354 registers. 5355 + Function attributes OS_main and OS_task. 5356 + AVR-specific built-in macros. 5357 5358 C6X 5359 5360 * Support has been added for the Texas Instruments C6X family of 5361 processors. 5362 5363 CR16 5364 5365 * Support has been added for National Semiconductor's CR16 5366 architecture. 5367 5368 Epiphany 5369 5370 * Support has been added for Adapteva's Epiphany architecture. 5371 5372 IA-32/x86-64 5373 5374 * Support for Intel AVX2 intrinsics, built-in functions and code 5375 generation is available via -mavx2. 5376 * Support for Intel BMI2 intrinsics, built-in functions and code 5377 generation is available via -mbmi2. 5378 * Implementation and automatic generation of __builtin_clz* using the 5379 lzcnt instruction is available via -mlzcnt. 5380 * Support for Intel FMA3 intrinsics and code generation is available 5381 via -mfma. 5382 * A new -mfsgsbase command-line option is available that makes GCC 5383 generate new segment register read/write instructions through 5384 dedicated built-ins. 5385 * Support for the new Intel rdrnd instruction is available via 5386 -mrdrnd. 5387 * Two additional AVX vector conversion instructions are available via 5388 -mf16c. 5389 * Support for new Intel processor codename IvyBridge with RDRND, 5390 FSGSBASE and F16C is available through -march=core-avx-i. 5391 * Support for the new Intel processor codename Haswell with AVX2, 5392 FMA, BMI, BMI2, LZCNT is available through -march=core-avx2. 5393 * Support for new AMD family 15h processors (Piledriver core) is now 5394 available through -march=bdver2 and -mtune=bdver2 options. 5395 * Support for [40]the x32 psABI is now available through the -mx32 5396 option. 5397 * Windows mingw targets are using the -mms-bitfields option by 5398 default. 5399 * Windows x86 targets are using the __thiscall calling convention for 5400 C++ class-member functions. 5401 * Support for the configure option --with-threads=posix for Windows 5402 mingw targets. 5403 5404 MIPS 5405 5406 * GCC now supports thread-local storage (TLS) for MIPS16. This 5407 requires GNU binutils 2.22 or later. 5408 * GCC can now generate code specifically for the Cavium Octeon+ and 5409 Octeon2 processors. The associated command-line options are 5410 -march=octeon+ and -march=octeon2 respectively. Both options 5411 require GNU binutils 2.22 or later. 5412 * GCC can now work around certain 24k errata, under the control of 5413 the command-line option -mfix-24k. These workarounds require GNU 5414 binutils 2.20 or later. 5415 * 32-bit MIPS GNU/Linux targets such as mips-linux-gnu can now build 5416 n32 and n64 multilibs. The result is effectively a 64-bit GNU/Linux 5417 toolchain that generates 32-bit code by default. Use the 5418 configure-time option --enable-targets=all to select these extra 5419 multilibs. 5420 * Passing -fno-delayed-branch now also stops the assembler from 5421 automatically filling delay slots. 5422 5423 PowerPC/PowerPC64 5424 5425 * Vectors of type vector long long or vector long are passed and 5426 returned using the same method as other vectors with the VSX 5427 instruction set. Previously GCC did not adhere to the ABI for 5428 128-bit vectors with 64-bit integer base types (PR 48857). This 5429 will also be fixed in the GCC 4.6.1 and 4.5.4 releases. 5430 * A new option -mno-pointers-to-nested-functions was added to allow 5431 AIX 32-bit/64-bit and GNU/Linux 64-bit PowerPC users to specify 5432 that the compiler should not load up the chain register (r11) 5433 before calling a function through a pointer. If you use this 5434 option, you cannot call nested functions through a pointer, or call 5435 other languages that might use the static chain. 5436 * A new option msave-toc-indirect was added to allow AIX 5437 32-bit/64-bit and GNU/Linux 64-bit PowerPC users control whether we 5438 save the TOC in the prologue for indirect calls or generate the 5439 save inline. This can speed up some programs that call through a 5440 function pointer a lot, but it can slow down other functions that 5441 only call through a function pointer in exceptional cases. 5442 * The PowerPC port will now enable machine-specific built-in 5443 functions when the user switches the target machine using the 5444 #pragma GCC target or __attribute__ ((__target__ ("target"))) code 5445 sequences. In addition, the target macros are updated. However, due 5446 to the way the -save-temps switch is implemented, you won't see the 5447 effect of these additional macros being defined in preprocessor 5448 output. 5449 5450 SH 5451 5452 * A new option -msoft-atomic has been added. When it is specified, 5453 GCC will generate GNU/Linux-compatible gUSA atomic sequences for 5454 the new __atomic routines. 5455 * Since it is neither supported by GAS nor officially documented, 5456 code generation for little endian SH2A has been disabled. 5457 Specifying -ml with -m2a* will now result in a compiler error. 5458 * The defunct -mbranch-cost option has been fixed. 5459 * Some improvements to the generated code of: 5460 + Utilization of the tst #imm,R0 instruction. 5461 + Dynamic shift instructions on SH2A. 5462 + Integer absolute value calculations. 5463 * The -mdiv= option for targets other than SHmedia has been fixed and 5464 documented. 5465 5466 SPARC 5467 5468 * The option -mflat has been reinstated. When it is specified, the 5469 compiler will generate code for a single register window model. 5470 This is essentially a new implementation and the corresponding 5471 debugger support has been added to GDB 7.4. 5472 * Support for the options -mtune=native and -mcpu=native has been 5473 added on selected native platforms (GNU/Linux and Solaris). 5474 * Support for the SPARC T3 (Niagara 3) processor has been added. 5475 * VIS: 5476 + An intrinsics header visintrin.h has been added. 5477 + Builtin intrinsics for the VIS 1.0 edge handling and pixel 5478 compare instructions have been added. 5479 + The little-endian version of alignaddr is now supported. 5480 + When possible, VIS builtins are marked const, which should 5481 increase the compiler's ability to optimize VIS operations. 5482 + The compiler now properly tracks the %gsr register and how it 5483 behaves as an input for various VIS instructions. 5484 + Akin to fzero, the compiler can now generate fone instructions 5485 in order to set all of the bits of a floating-point register 5486 to 1. 5487 + The documentation for the VIS intrinsics in the GCC manual has 5488 been brought up to date and many inaccuracies were fixed. 5489 + Intrinsics for the VIS 2.0 bmask, bshuffle, and 5490 non-condition-code setting edge instructions have been added. 5491 Their availability is controlled by the new -mvis2 and 5492 -mno-vis2 options. They are enabled by default on 5493 UltraSPARC-III and later CPUs. 5494 * Support for UltraSPARC Fused Multiply-Add floating-point extensions 5495 has been added. These instructions are enabled by default on SPARC 5496 T3 (Niagara 3) and later CPUs. 5497 5498 TILE-Gx/TILEPro 5499 5500 * Support has been added for the Tilera TILE-Gx and TILEPro families 5501 of processors. 5502 5503 Other significant improvements 5504 5505 * A new option (-grecord-gcc-switches) was added that appends 5506 compiler command-line options that might affect code generation to 5507 the DW_AT_producer attribute string in the DWARF debugging 5508 information. 5509 * GCC now supports various new GNU extensions to the DWARF debugging 5510 information format, like [41]entry value and [42]call site 5511 information, [43]typed DWARF stack or [44]a more compact macro 5512 representation. Support for these extensions has been added to GDB 5513 7.4. They can be disabled through the -gstrict-dwarf command-line 5514 option. 5515 5516 GCC 4.7.1 5517 5518 This is the [45]list of problem reports (PRs) from GCC's bug tracking 5519 system that are known to be fixed in the 4.7.1 release. This list might 5520 not be complete (that is, it is possible that some PRs that have been 5521 fixed are not listed here). 5522 5523 The Go front end in the 4.7.1 release fully supports the [46]Go 1 5524 language standard. 5525 5526 GCC 4.7.2 5527 5528 This is the [47]list of problem reports (PRs) from GCC's bug tracking 5529 system that are known to be fixed in the 4.7.2 release. This list might 5530 not be complete (that is, it is possible that some PRs that have been 5531 fixed are not listed here). 5532 5533 GCC 4.7.3 5534 5535 This is the [48]list of problem reports (PRs) from GCC's bug tracking 5536 system that are known to be fixed in the 4.7.3 release. This list might 5537 not be complete (that is, it is possible that some PRs that have been 5538 fixed are not listed here). 5539 5540 GCC 4.7.4 5541 5542 This is the [49]list of problem reports (PRs) from GCC's bug tracking 5543 system that are known to be fixed in the 4.7.4 release. This list might 5544 not be complete (that is, it is possible that some PRs that have been 5545 fixed are not listed here). 5546 5547 5548 For questions related to the use of GCC, please consult these web 5549 pages and the [50]GCC manuals. If that fails, the 5550 [51]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5551 web pages and the development of GCC are welcome on our developer 5552 list at [52]gcc (a] gcc.gnu.org. All of [53]our lists have public 5553 archives. 5554 5555 Copyright (C) [54]Free Software Foundation, Inc. Verbatim copying and 5556 distribution of this entire article is permitted in any medium, 5557 provided this notice is preserved. 5558 5559 These pages are [55]maintained by the GCC team. Last modified 5560 2018-09-30[56]. 5561 5562 References 5563 5564 1. https://gcc.gnu.org/ml/gcc-patches/2011-03/msg01263.html 5565 2. http://savannah.nongnu.org/bugs/?35407 5566 3. https://gcc.gnu.org/PR18145 5567 4. https://gcc.gnu.org/gcc-4.7/porting_to.html 5568 5. http://www.openmp.org/specifications/ 5569 6. https://gcc.gnu.org/wiki/TransactionalMemory 5570 7. https://gcc.gnu.org/wiki/Atomic/GCCMM 5571 8. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 5572 9. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 5573 10. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 5574 11. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 5575 12. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 5576 13. http://gcc.gnu.org/gcc-4.7/cxx0x_status.html 5577 14. https://gcc.gnu.org/PR14258 5578 15. https://gcc.gnu.org/PR35688 5579 16. https://gcc.gnu.org/onlinedocs/gcc-4.7.4/libstdc++/manual/manual/status.html#status.iso.2011 5580 17. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 5581 18. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Optimize-Options.html#index-Ofast-689 5582 19. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfno-protect-parens_007d-270 5583 20. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfstack-arrays_007d-254 5584 21. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfrontend-optimize_007d-275 5585 22. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWfunction-elimination_007d-170 5586 23. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfaggressive-function-elimination_007d-270 5587 24. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWreal-q-constant_007d-149 5588 25. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/SELECTED_005fREAL_005fKIND.html 5589 26. https://docs.microsoft.com/en-us/windows/desktop/wer/collecting-user-mode-dumps 5590 27. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Debugging-Options.html#index-g_t_0040code_007bfno-backtrace_007d-183 5591 28. https://gcc.gnu.org/wiki/Fortran2003Status 5592 29. https://gcc.gnu.org/wiki/OOP 5593 30. https://gcc.gnu.org/wiki/Fortran2008Status 5594 31. https://gcc.gnu.org/wiki/Coarray 5595 32. https://gcc.gnu.org/wiki/CoarrayLib 5596 33. https://gcc.gnu.org/wiki/TS29113Status 5597 34. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bstd_003d_007d_0040var_007bstd_007d-option-53 5598 35. https://golang.org/doc/go1 5599 36. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/Named-Address-Spaces.html 5600 37. http://nongnu.org/avr-libc/ 5601 38. https://gcc.gnu.org/PR54461 5602 39. https://gcc.gnu.org/onlinedocs/gcc-4.7.1/gcc/AVR-Built%5f002din-Functions.html 5603 40. https://sites.google.com/site/x32abi/ 5604 41. http://www.dwarfstd.org/ShowIssue.php?issue=100909.1 5605 42. http://www.dwarfstd.org/ShowIssue.php?issue=100909.2 5606 43. http://www.dwarfstd.org/ShowIssue.php?issue=140425.1 5607 44. http://www.dwarfstd.org/ShowIssue.php?issue=110722.1 5608 45. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.1 5609 46. https://golang.org/doc/go1 5610 47. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.2 5611 48. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.3 5612 49. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.7.4 5613 50. https://gcc.gnu.org/onlinedocs/ 5614 51. mailto:gcc-help (a] gcc.gnu.org 5615 52. mailto:gcc (a] gcc.gnu.org 5616 53. https://gcc.gnu.org/lists.html 5617 54. https://www.fsf.org/ 5618 55. https://gcc.gnu.org/about.html 5619 56. http://validator.w3.org/check/referer 5620 ====================================================================== 5621 http://gcc.gnu.org/gcc-4.6/index.html 5622 5623 GCC 4.6 Release Series 5624 5625 April 12, 2013 5626 5627 The [1]GNU project and the GCC developers are pleased to announce the 5628 release of GCC 4.6.4. 5629 5630 This release is a bug-fix release, containing fixes for regressions in 5631 GCC 4.6.3 relative to previous releases of GCC. 5632 5633 This release series is no longer maintained. 5634 5635 Release History 5636 5637 GCC 4.6.4 5638 April 12, 2013 ([2]changes, [3]documentation) 5639 5640 GCC 4.6.3 5641 March 1, 2012 ([4]changes, [5]documentation) 5642 5643 GCC 4.6.2 5644 October 26, 2011 ([6]changes, [7]documentation) 5645 5646 GCC 4.6.1 5647 June 27, 2011 ([8]changes, [9]documentation) 5648 5649 GCC 4.6.0 5650 March 25, 2011 ([10]changes, [11]documentation) 5651 5652 References and Acknowledgements 5653 5654 GCC used to stand for the GNU C Compiler, but since the compiler 5655 supports several other languages aside from C, it now stands for the 5656 GNU Compiler Collection. 5657 5658 A list of [12]successful builds is updated as new information becomes 5659 available. 5660 5661 The GCC developers would like to thank the numerous people that have 5662 contributed new features, improvements, bug fixes, and other changes as 5663 well as test results to GCC. This [13]amazing group of volunteers is 5664 what makes GCC successful. 5665 5666 For additional information about GCC please refer to the [14]GCC 5667 project web site or contact the [15]GCC development mailing list. 5668 5669 To obtain GCC please use [16]our mirror sites or [17]our SVN server. 5670 5671 5672 For questions related to the use of GCC, please consult these web 5673 pages and the [18]GCC manuals. If that fails, the 5674 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 5675 web pages and the development of GCC are welcome on our developer 5676 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 5677 archives. 5678 5679 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 5680 distribution of this entire article is permitted in any medium, 5681 provided this notice is preserved. 5682 5683 These pages are [23]maintained by the GCC team. Last modified 5684 2018-09-30[24]. 5685 5686 References 5687 5688 1. http://www.gnu.org/ 5689 2. http://gcc.gnu.org/gcc-4.6/changes.html 5690 3. https://gcc.gnu.org/onlinedocs/4.6.4/ 5691 4. http://gcc.gnu.org/gcc-4.6/changes.html 5692 5. https://gcc.gnu.org/onlinedocs/4.6.3/ 5693 6. http://gcc.gnu.org/gcc-4.6/changes.html 5694 7. https://gcc.gnu.org/onlinedocs/4.6.2/ 5695 8. http://gcc.gnu.org/gcc-4.6/changes.html 5696 9. https://gcc.gnu.org/onlinedocs/4.6.1/ 5697 10. http://gcc.gnu.org/gcc-4.6/changes.html 5698 11. https://gcc.gnu.org/onlinedocs/4.6.0/ 5699 12. http://gcc.gnu.org/gcc-4.6/buildstat.html 5700 13. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 5701 14. http://gcc.gnu.org/index.html 5702 15. mailto:gcc (a] gcc.gnu.org 5703 16. http://gcc.gnu.org/mirrors.html 5704 17. http://gcc.gnu.org/svn.html 5705 18. https://gcc.gnu.org/onlinedocs/ 5706 19. mailto:gcc-help (a] gcc.gnu.org 5707 20. mailto:gcc (a] gcc.gnu.org 5708 21. https://gcc.gnu.org/lists.html 5709 22. https://www.fsf.org/ 5710 23. https://gcc.gnu.org/about.html 5711 24. http://validator.w3.org/check/referer 5712 ====================================================================== 5713 http://gcc.gnu.org/gcc-4.6/changes.html 5714 5715 GCC 4.6 Release Series 5716 Changes, New Features, and Fixes 5717 5718 Caveats 5719 5720 * The options -b <machine> and -V <version> have been removed because 5721 they were unreliable. Instead, users should directly run 5722 <machine>-gcc when cross-compiling, or <machine>-gcc-<version> to 5723 run a different version of gcc. 5724 * GCC now has stricter checks for invalid command-line options. In 5725 particular, when gcc was called to link object files rather than 5726 compile source code, it would previously accept and ignore all 5727 options starting with --, including linker options such as 5728 --as-needed and --export-dynamic, although such options would 5729 result in errors if any source code was compiled. Such options, if 5730 unknown to the compiler, are now rejected in all cases; if the 5731 intent was to pass them to the linker, options such as 5732 -Wl,--as-needed should be used. 5733 * Versions of the GNU C library up to and including 2.11.1 included 5734 an [1]incorrect implementation of the cproj function. GCC optimizes 5735 its builtin cproj according to the behavior specified and allowed 5736 by the ISO C99 standard. If you want to avoid discrepancies between 5737 the C library and GCC's builtin transformations when using cproj in 5738 your code, use GLIBC 2.12 or later. If you are using an older GLIBC 5739 and actually rely on the incorrect behavior of cproj, then you can 5740 disable GCC's transformations using -fno-builtin-cproj. 5741 * The C-only intermodule optimization framework (IMA, enabled by 5742 -combine) has been removed in favor of the new generic link-time 5743 optimization framework (LTO) introduced in [2]GCC 4.5.0. 5744 * GCC now ships with the LGPL-licensed libquadmath library, which 5745 provides quad-precision mathematical functions for targets with a 5746 __float128 datatype. __float128 is available for targets on 32-bit 5747 x86, x86-64 and Itanium architectures. The libquadmath library is 5748 automatically built on such targets when building the Fortran 5749 compiler. 5750 * New -Wunused-but-set-variable and -Wunused-but-set-parameter 5751 warnings were added for C, C++, Objective-C and Objective-C++. 5752 These warnings diagnose variables respective parameters which are 5753 only set in the code and never otherwise used. Usually such 5754 variables are useless and often even the value assigned to them is 5755 computed needlessly, sometimes expensively. The 5756 -Wunused-but-set-variable warning is enabled by default by -Wall 5757 flag and -Wunused-but-set-parameter by -Wall -Wextra flags. 5758 * On ARM, a bug has been fixed in GCC's implementation of the AAPCS 5759 rules for the layout of vectors that could lead to wrong code being 5760 generated. Vectors larger than 8 bytes in size are now by default 5761 aligned to an 8-byte boundary. This is an ABI change: code that 5762 makes explicit use of vector types may be incompatible with binary 5763 objects built with older versions of GCC. Auto-vectorized code is 5764 not affected by this change. (This change affects GCC versions 5765 4.6.4 and later, with the exception of versions 4.7.0 and 4.7.1.) 5766 * On AVR, variables with the progmem attribute to locate data in 5767 flash memory must be qualified as const. 5768 * Support for a number of older systems and recently unmaintained or 5769 untested target ports of GCC has been declared obsolete in GCC 4.6. 5770 Unless there is activity to revive them, the next release of GCC 5771 will have their sources permanently removed. 5772 All GCC ports for the following processor architectures have been 5773 declared obsolete: 5774 + Argonaut ARC (arc-*) 5775 + National Semiconductor CRX (crx-*) 5776 + Motorola 68HC11 and 68HC12 (m68hc11-*-*, m6811-*-*, 5777 m68hc12-*-*, m6812-*-*) 5778 + Sunplus S+core (score-*) 5779 The following ports for individual systems on particular 5780 architectures have been obsoleted: 5781 + Interix (i[34567]86-*-interix3*) 5782 + NetWare x86 (i[3456x]86-*-netware*) 5783 + Generic ARM PE (arm-*-pe* other than arm*-wince-pe*) 5784 + MCore PE (mcore-*-pe*) 5785 + SH SymbianOS (sh*-*-symbianelf*) 5786 + GNU Hurd on Alpha and PowerPC (alpha*-*-gnu*, powerpc*-*-gnu*) 5787 + M68K uClinux old ABI (m68k-*-uclinuxoldabi*) 5788 + a.out NetBSD (arm*-*-netbsd*, i[34567]86-*-netbsd*, 5789 vax-*-netbsd*, but not *-*-netbsdelf*) 5790 The i[34567]86-*-pe alias for Cygwin targets has also been 5791 obsoleted; users should configure for i[34567]86-*-cygwin* instead. 5792 Certain configure options to control the set of libraries built 5793 with GCC on some targets have been obsoleted. On ARM targets, the 5794 options --disable-fpu, --disable-26bit, --disable-underscore, 5795 --disable-interwork, --disable-biendian and --disable-nofmult have 5796 been obsoleted. On MIPS targets, the options 5797 --disable-single-float, --disable-biendian and --disable-softfloat 5798 have been obsoleted. 5799 * Support has been removed for all the [3]configurations obsoleted in 5800 GCC 4.5. 5801 * More information on porting to GCC 4.6 from previous versions of 5802 GCC can be found in the [4]porting guide for this release. 5803 5804 General Optimizer Improvements 5805 5806 * A new general optimization level, -Ofast, has been introduced. It 5807 combines the existing optimization level -O3 with options that can 5808 affect standards compliance but result in better optimized code. 5809 For example, -Ofast enables -ffast-math. 5810 * Link-time optimization improvements: 5811 + The [5]Scalable Whole Program Optimizer (WHOPR) project has 5812 stabilized to the point of being usable. It has become the 5813 default mode when using the LTO optimization model. Link time 5814 optimization can now split itself into multiple parallel 5815 compilations. Parallelism is controlled with -flto=n (where n 5816 specifies the number of compilations to execute in parallel). 5817 GCC can also cooperate with a GNU make job server by 5818 specifying the -flto=jobserver option and adding + to the 5819 beginning of the Makefile rule executing the linker. 5820 Classical LTO mode can be enforced by -flto-partition=none. 5821 This may result in small code quality improvements. 5822 + A large number of bugs were fixed. GCC itself, Mozilla Firefox 5823 and other large applications can be built with LTO enabled. 5824 + The linker plugin support improvements 5825 o Linker plugin is now enabled by default when the linker 5826 is detected to have plugin support. This is the case for 5827 GNU ld 2.21.51 or newer (on ELF and Cygwin targets) and 5828 the Gold linker on ELF targets. Plugin support of the 5829 Apple linker on Darwin is not compatible with GCC. The 5830 linker plugin can also be controlled by the 5831 -fuse-linker-plugin command-line option. 5832 o Resolution information from the linker plugin is used to 5833 drive whole program assumptions. Use of the linker plugin 5834 results in more aggressive optimization on binaries and 5835 on shared libraries that use the hidden visibility 5836 attribute. Consequently the use of -fwhole-program is not 5837 necessary in addition to LTO. 5838 + Hidden symbols used from non-LTO objects now have to be 5839 explicitly annotated with externally_visible when the linker 5840 plugin is not used. 5841 + C++ inline functions and virtual tables are now privatized 5842 more aggressively, leading to better inter-procedural 5843 optimization and faster dynamic linking. 5844 + Memory usage and intermediate language streaming performance 5845 have been improved. 5846 + Static constructors and destructors from individual units are 5847 inlined into a single function. This can significantly improve 5848 startup times of large C++ applications where static 5849 constructors are very common. For example, static constructors 5850 are used when including the iostream header. 5851 + Support for the Ada language has been added. 5852 * Interprocedural optimization improvements 5853 + The interprocedural framework was re-tuned for link time 5854 optimization. Several scalability issues were resolved. 5855 + Improved auto-detection of const and pure functions. Newly, 5856 noreturn functions are auto-detected. 5857 The [6]-Wsuggest-attribute=[const|pure|noreturn] flag is 5858 available that informs users when adding attributes to headers 5859 might improve code generation. 5860 + A number of inlining heuristic improvements. In particular: 5861 o Partial inlining is now supported and enabled by default 5862 at -O2 and greater. The feature can be controlled via 5863 -fpartial-inlining. 5864 Partial inlining splits functions with short hot path to 5865 return. This allows more aggressive inlining of the hot 5866 path leading to better performance and often to code size 5867 reductions (because cold parts of functions are not 5868 duplicated). 5869 o Scalability for large compilation units was improved 5870 significantly. 5871 o Inlining of callbacks is now more aggressive. 5872 o Virtual methods are considered for inlining when the 5873 caller is inlined and devirtualization is then possible. 5874 o Inlining when optimizing for size (either in cold regions 5875 of a program or when compiling with -Os) was improved to 5876 better handle C++ programs with larger abstraction 5877 penalty, leading to smaller and faster code. 5878 + The IPA reference optimization pass detecting global variables 5879 used or modified by functions was strengthened and sped up. 5880 + Functions whose address was taken are now optimized out when 5881 all references to them are dead. 5882 + A new inter-procedural static profile estimation pass detects 5883 functions that are executed once or unlikely to be executed. 5884 Unlikely executed functions are optimized for size. Functions 5885 executed once are optimized for size except for the inner 5886 loops. 5887 + On most targets with named section support, functions used 5888 only at startup (static constructors and main), functions used 5889 only at exit and functions detected to be cold are placed into 5890 separate text segment subsections. This extends the 5891 -freorder-functions feature and is controlled by the same 5892 switch. The goal is to improve the startup time of large C++ 5893 programs. 5894 Proper function placement requires linker support. GNU ld 5895 2.21.51 on ELF targets was updated to place those functions 5896 together within the text section leading to better code 5897 locality and faster startup times of large C++ programs. The 5898 feature is also supported in the Apple linker. Support in the 5899 gold linker is planned. 5900 * A new switch -fstack-usage has been added. It makes the compiler 5901 output stack usage information for the program, on a per-function 5902 basis, in an auxiliary file. 5903 * A new switch -fcombine-stack-adjustments has been added. It can be 5904 used to enable or disable the compiler's stack-slot combining pass 5905 which before was enabled automatically at -O1 and above, but could 5906 not be controlled on its own. 5907 * A new switch -fstrict-volatile-bitfields has been added. Using it 5908 indicates that accesses to volatile bitfields should use a single 5909 access of the width of the field's type. This option can be useful 5910 for precisely defining and accessing memory-mapped peripheral 5911 registers from C or C++. 5912 5913 Compile time and memory usage improvements 5914 5915 * Datastructures used by the dataflow framework in GCC were 5916 reorganized for better memory usage and more cache locality. 5917 Compile time is improved especially on units with large functions 5918 (possibly resulting from a lot of inlining) not fitting into the 5919 processor cache. The compile time of the GCC C compiler binary with 5920 link-time optimization went down by over 10% (benchmarked on x86-64 5921 target). 5922 5923 New Languages and Language specific improvements 5924 5925 Ada 5926 5927 * Stack checking has been improved on selected architectures (Alpha, 5928 IA-32/x86-64, RS/6000 and SPARC): it now will detect stack 5929 overflows in all cases on these architectures. 5930 * Initial support for Ada 2012 has been added. 5931 5932 C family 5933 5934 * A new warning, enabled by -Wdouble-promotion, has been added that 5935 warns about cases where a value of type float is implicitly 5936 promoted to double. This is especially helpful for CPUs that handle 5937 the former in hardware, but emulate the latter in software. 5938 * A new function attribute leaf was introduced. This attribute allows 5939 better inter-procedural optimization across calls to functions that 5940 return to the current unit only via returning or exception 5941 handling. This is the case for most library functions that have no 5942 callbacks. 5943 * Support for a new data type __int128 for targets having wide enough 5944 machine-mode support. 5945 * The new function attribute callee_pop_aggregate allows to specify 5946 if the caller or callee is responsible for popping the aggregate 5947 return pointer value from the stack. 5948 * Support for selectively enabling and disabling warnings via #pragma 5949 GCC diagnostic has been added. For instance: 5950 #pragma GCC diagnostic error "-Wuninitialized" 5951 foo(a); /* error is given for this one */ 5952 #pragma GCC diagnostic push 5953 #pragma GCC diagnostic ignored "-Wuninitialized" 5954 foo(b); /* no diagnostic for this one */ 5955 #pragma GCC diagnostic pop 5956 foo(c); /* error is given for this one */ 5957 #pragma GCC diagnostic pop 5958 foo(d); /* depends on command-line options */ 5959 5960 * The -fmax-errors=N option is now supported. Using this option 5961 causes the compiler to exit after N errors have been issued. 5962 5963 C 5964 5965 * There is now experimental support for some features from the 5966 upcoming C1X revision of the ISO C standard. This support may be 5967 selected with -std=c1x, or -std=gnu1x for C1X with GNU extensions. 5968 Note that this support is experimental and may change incompatibly 5969 in future releases for consistency with changes to the C1X standard 5970 draft. The following features are newly supported as described in 5971 the N1539 draft of C1X (with changes agreed at the March 2011 WG14 5972 meeting); some other features were already supported with no 5973 compiler changes being needed, or have some support but not in full 5974 accord with N1539 (as amended). 5975 + Static assertions (_Static_assert keyword) 5976 + Typedef redefinition 5977 + New macros in <float.h> 5978 + Anonymous structures and unions 5979 * The new -fplan9-extensions option directs the compiler to support 5980 some extensions for anonymous struct fields which are implemented 5981 by the Plan 9 compiler. A pointer to a struct may be automatically 5982 converted to a pointer to an anonymous field when calling a 5983 function, in order to make the types match. An anonymous struct 5984 field whose type is a typedef name may be referred to using the 5985 typedef name. 5986 5987 C++ 5988 5989 * Improved [7]experimental support for the upcoming C++0x ISO C++ 5990 standard, including support for constexpr (thanks to Gabriel Dos 5991 Reis and Jason Merrill), nullptr (thanks to Magnus Fromreide), 5992 noexcept, unrestricted unions, range-based for loops (thanks to 5993 Rodrigo Rivas Costa), opaque enum declarations (thanks also to 5994 Rodrigo), implicitly deleted functions and implicit move 5995 constructors. 5996 * When an extern declaration within a function does not match a 5997 declaration in the enclosing context, G++ now properly declares the 5998 name within the namespace of the function rather than the namespace 5999 which was open just before the function definition ([8]c++/43145). 6000 * GCC now warns by default when casting integers to larger pointer 6001 types. These warnings can be disabled with the option 6002 -Wno-int-to-pointer-cast, which is now also available in C++. 6003 * G++ no longer optimizes using the assumption that a value of 6004 enumeration type will fall within the range specified by the 6005 standard, since that assumption is easily violated with a 6006 conversion from integer type ([9]c++/43680). The old behavior can 6007 be restored with -fstrict-enums. 6008 * The new -fnothrow-opt flag changes the semantics of a throw() 6009 exception specification to match the proposed semantics of the 6010 noexcept specification: just call terminate if an exception tries 6011 to propagate out of a function with such an exception 6012 specification. This dramatically reduces or eliminates the code 6013 size overhead from adding the exception specification. 6014 * The new -Wnoexcept flag will suggest adding a noexcept qualifier to 6015 a function that the compiler can tell doesn't throw if it would 6016 change the value of a noexcept expression. 6017 * The -Wshadow option now warns if a local variable or type 6018 declaration shadows another type in C++. Note that the compiler 6019 will not warn if a local variable shadows a struct/class/enum, but 6020 will warn if it shadows an explicit typedef. 6021 * When an identifier is not found in the current scope, G++ now 6022 offers suggestions about which identifier might have been intended. 6023 * G++ now issues clearer diagnostics for missing semicolons after 6024 class, struct, and union definitions. 6025 * G++ now issues clearer diagnostics for missing semicolons after 6026 class member declarations. 6027 * G++ now issues clearer diagnostics when a colon is used in a place 6028 where a double-colon was intended. 6029 * G++ no longer accepts mutable on reference members ([10]c++/33558). 6030 Use -fpermissive to allow the old, non-conforming behaviour. 6031 * A few mangling fixes have been made, to attribute const/volatile on 6032 function pointer types, decltype of a plain decl, and use of a 6033 function parameter in the declaration of another parameter. By 6034 default the compiler still uses the old mangling, but emits aliases 6035 with the new mangling on targets that support strong aliases. Users 6036 can switch over entirely to the new mangling with -fabi-version=5 6037 or -fabi-version=0. -Wabi will now warn about code that uses the 6038 old mangling. 6039 * In 4.6.0 and 4.6.1 G++ no longer allows objects of const-qualified 6040 type to be default initialized unless the type has a user-declared 6041 default constructor. In 4.6.2 G++ implements the proposed 6042 resolution of [11]DR 253, so default initialization is allowed if 6043 it initializes all subobjects. Code that fails to compile can be 6044 fixed by providing an initializer e.g. 6045 struct A { A(); }; 6046 struct B : A { int i; }; 6047 const B b = B(); 6048 Use -fpermissive to allow the old, non-conforming behaviour. 6049 6050 Runtime Library (libstdc++) 6051 6052 * [12]Improved experimental support for the upcoming ISO C++ 6053 standard, C++0x, including using constexpr and nullptr. 6054 * Performance improvements to the [13]Debug Mode, thanks to Franc,ois 6055 Dumont. 6056 * Atomic operations used for reference-counting are annotated so that 6057 they can be understood by race detectors such as Helgrind, see 6058 [14]Data Race Hunting. 6059 * Most libstdc++ standard headers have been changed to no longer 6060 include the cstddef header as an implementation detail. Code that 6061 relied on that header being included as side-effect of including 6062 other standard headers will need to include cstddef explicitly. 6063 6064 Fortran 6065 6066 * On systems supporting the libquadmath library, GNU Fortran now also 6067 supports a quad-precision, kind=16 floating-point data type 6068 (REAL(16), COMPLEX(16)). As the data type is not fully supported in 6069 hardware, calculations might be one to two orders of magnitude 6070 slower than with the 4, 8 or 10 bytes floating-point data types. 6071 This change does not affect systems which support REAL(16) in 6072 hardware nor those which do not support libquadmath. 6073 * Much improved compile time for large array constructors. 6074 * In order to reduce execution time and memory consumption, use of 6075 temporary arrays in assignment expressions is avoided for many 6076 cases. The compiler now reverses loops in order to avoid generating 6077 a temporary array where possible. 6078 * Improved diagnostics, especially with -fwhole-file. 6079 * The -fwhole-file flag is now enabled by default. This improves code 6080 generation and diagnostics. It can be disabled using the deprecated 6081 -fno-whole-file flag. 6082 * Support the generation of Makefile dependencies via the [15]-M... 6083 flags of GCC; you may need to specify the -cpp option in addition. 6084 The dependencies take modules, Fortran's include, and CPP's 6085 #include into account. Note: Using -M for the module path is no 6086 longer supported, use -J instead. 6087 * The flag -Wconversion has been modified to only issue warnings 6088 where a conversion leads to information loss. This drastically 6089 reduces the number of warnings; -Wconversion is thus now enabled 6090 with -Wall. The flag -Wconversion-extra has been added and also 6091 warns about other conversions; -Wconversion-extra typically issues 6092 a huge number of warnings, most of which can be ignored. 6093 * A new command-line option -Wunused-dummy-argument warns about 6094 unused dummy arguments and is included in -Wall. Before, 6095 -Wunused-variable also warned about unused dummy arguments. 6096 * Fortran 2003 support has been extended: 6097 + Improved support for polymorphism between libraries and 6098 programs and for complicated inheritance patterns (cf. 6099 [16]object-oriented programming). 6100 + Experimental support of the ASSOCIATE construct. 6101 + In pointer assignments it is now possible to specify the lower 6102 bounds of the pointer and, for a rank-1 or a simply contiguous 6103 data-target, to remap the bounds. 6104 + Automatic (re)allocation: In intrinsic assignments to 6105 allocatable variables the left-hand side will be automatically 6106 allocated (if unallocated) or reallocated (if the shape or 6107 type parameter is different). To avoid the small performance 6108 penalty, you can use a(:) = ... instead of a = ... for arrays 6109 and character strings - or disable the feature using -std=f95 6110 or -fno-realloc-lhs. 6111 + Deferred type parameter: For scalar allocatable and pointer 6112 variables the character length can be deferred. 6113 + Namelist variables with allocatable and pointer attribute and 6114 nonconstant length type parameter are supported. 6115 * Fortran 2008 support has been extended: 6116 + Experimental [17]coarray support (for one image only, i.e. 6117 num_images() == 1); use the [18]-fcoarray=single flag to 6118 enable it. 6119 + The STOP and the new ERROR STOP statements now support all 6120 constant expressions. 6121 + Support for the CONTIGUOUS attribute. 6122 + Support for ALLOCATE with MOLD. 6123 + Support for the STORAGE_SIZE intrinsic inquiry function. 6124 + Support of the NORM2 and PARITY intrinsic functions. 6125 + The following bit intrinsics were added: POPCNT and POPPAR for 6126 counting the number of 1 bits and returning the parity; BGE, 6127 BGT, BLE, and BLT for bitwise comparisons; DSHIFTL and DSHIFTR 6128 for combined left and right shifts, MASKL and MASKR for simple 6129 left and right justified masks, MERGE_BITS for a bitwise merge 6130 using a mask, SHIFTA, SHIFTL and SHIFTR for shift operations, 6131 and the transformational bit intrinsics IALL, IANY and 6132 IPARITY. 6133 + Support of the EXECUTE_COMMAND_LINE intrinsic subroutine. 6134 + Support for the IMPURE attribute for procedures, which allows 6135 for ELEMENTAL procedures without the restrictions of PURE. 6136 + Null pointers (including NULL()) and not allocated variables 6137 can be used as actual argument to optional non-pointer, 6138 non-allocatable dummy arguments, denoting an absent argument. 6139 + Non-pointer variables with TARGET attribute can be used as 6140 actual argument to POINTER dummies with INTENT(IN) 6141 + Pointers including procedure pointers and those in a derived 6142 type (pointer components) can now be initialized by a target 6143 instead of only by NULL. 6144 + The EXIT statement (with construct-name) can now be used to 6145 leave not only the DO but also the ASSOCIATE, BLOCK, IF, 6146 SELECT CASE and SELECT TYPE constructs. 6147 + Internal procedures can now be used as actual argument. 6148 + The named constants INTEGER_KINDS, LOGICAL_KINDS, REAL_KINDS 6149 and CHARACTER_KINDS of the intrinsic module ISO_FORTRAN_ENV 6150 have been added; these arrays contain the supported kind 6151 values for the respective types. 6152 + The module procedures C_SIZEOF of the intrinsic module 6153 ISO_C_BINDINGS and COMPILER_VERSION and COMPILER_OPTIONS of 6154 ISO_FORTRAN_ENV have been implemented. 6155 + Minor changes: obsolescence diagnostics for ENTRY was added 6156 for -std=f2008; a line may start with a semicolon; for 6157 internal and module procedures END can be used instead of END 6158 SUBROUTINE and END FUNCTION; SELECTED_REAL_KIND now also takes 6159 a RADIX argument; intrinsic types are supported for 6160 TYPE(intrinsic-type-spec); multiple type-bound procedures can 6161 be declared in a single PROCEDURE statement; implied-shape 6162 arrays are supported for named constants (PARAMETER). The 6163 transformational, three argument versions of BESSEL_JN and 6164 BESSEL_YN were added - the elemental, two-argument version had 6165 been added in GCC 4.4; note that the transformational 6166 functions use a recurrence algorithm. 6167 6168 Go 6169 6170 Support for the [19]Go programming language has been added to GCC. It 6171 is not enabled by default when you build GCC; use the 6172 --enable-languages configure option to build it. The driver program for 6173 compiling Go code is gccgo. 6174 6175 Go is currently known to work on GNU/Linux and RTEMS. Solaris support 6176 is in progress. It may or may not work on other platforms. 6177 6178 Objective-C and Objective-C++ 6179 6180 * The -fobjc-exceptions flag is now required to enable Objective-C 6181 exception and synchronization syntax (introduced by the keywords 6182 @try, @catch, @finally and @synchronized). 6183 * A number of Objective-C 2.0 features and extensions are now 6184 supported by GCC. These features are enabled by default; you can 6185 disable them by using the new -fobjc-std=objc1 command-line option. 6186 * The Objective-C 2.0 dot-syntax is now supported. It is an 6187 alternative syntax for using getters and setters; object.count is 6188 automatically converted into [object count] or [object setCount: 6189 ...] depending on context; for example if (object.count > 0) is 6190 automatically compiled into the equivalent of if ([object count] > 6191 0) while object.count = 0; is automatically compiled into the 6192 equivalent ot [object setCount: 0];. The dot-syntax can be used 6193 with instance and class objects and with any setters or getters, no 6194 matter if they are part of a declared property or not. 6195 * Objective-C 2.0 declared properties are now supported. They are 6196 declared using the new @property keyword, and are most commonly 6197 used in conjunction with the new Objective-C 2.0 dot-syntax. The 6198 nonatomic, readonly, readwrite, assign, retain, copy, setter and 6199 getter attributes are all supported. Marking declared properties 6200 with __attribute__ ((deprecated)) is supported too. 6201 * The Objective-C 2.0 @synthesize and @dynamic keywords are 6202 supported. @synthesize causes the compiler to automatically 6203 synthesize a declared property, while @dynamic is used to disable 6204 all warnings for a declared property for which no implementation is 6205 provided at compile time. Synthesizing declared properties requires 6206 runtime support in most useful cases; to be able to use it with the 6207 GNU runtime, appropriate helper functions have been added to the 6208 GNU Objective-C runtime ABI, and are implemented by the GNU 6209 Objective-C runtime library shipped with GCC. 6210 * The Objective-C 2.0 fast enumeration syntax is supported in 6211 Objective-C. This is currently not yet available in Objective-C++. 6212 Fast enumeration requires support in the runtime, and such support 6213 has been added to the GNU Objective-C runtime library (shipped with 6214 GCC). 6215 * The Objective-C 2.0 @optional keyword is supported. It allows you 6216 to mark methods or properties in a protocol as optional as opposed 6217 to required. 6218 * The Objective-C 2.0 @package keyword is supported. It has currently 6219 the same effect as the @public keyword. 6220 * Objective-C 2.0 method attributes are supported. Currently the 6221 supported attributes are deprecated, sentinel, noreturn and format. 6222 * Objective-C 2.0 method argument attributes are supported. The most 6223 widely used attribute is unused, to mark an argument as unused in 6224 the implementation. 6225 * Objective-C 2.0 class and protocol attributes are supported. 6226 Currently the only supported attribute is deprecated. 6227 * Objective-C 2.0 class extensions are supported. A class extension 6228 has the same syntax as a category declaration with no category 6229 name, and the methods and properties declared in it are added 6230 directly to the main class. It is mostly used as an alternative to 6231 a category to add methods to a class without advertising them in 6232 the public headers, with the advantage that for class extensions 6233 the compiler checks that all the privately declared methods are 6234 actually implemented. 6235 * As a result of these enhancements, GCC can now be used to build 6236 Objective-C and Objective-C++ software that uses Foundation and 6237 other important system frameworks with the NeXT runtime on Darwin 9 6238 and Darwin 10 (OSX 10.5 and 10.6). 6239 * Many bugs in the compiler have been fixed in this release; in 6240 particular, LTO can now be used when compiling Objective-C and 6241 Objective-C++ and the parser is much more robust in dealing with 6242 invalid code. 6243 6244 Runtime Library (libobjc) 6245 6246 * The GNU Objective-C runtime library now defines the macro 6247 __GNU_LIBOBJC__ (with a value that is increased at every release 6248 where there is any change to the API) in objc/objc.h, making it 6249 easy to determine if the GNU Objective-C runtime library is being 6250 used, and if so, which version. Previous versions of the GNU 6251 Objective-C runtime library (and other Objective-C runtime 6252 libraries such as the Apple one) do not define this macro. 6253 * A new Objective-C 2.0 API, almost identical to the one implemented 6254 by the Apple Objective-C runtime, has been implemented in the GNU 6255 Objective-C runtime library. The new API hides the internals of 6256 most runtime structures but provides a more extensive set of 6257 functions to operate on them. It is much easier, for example, to 6258 create or modify classes at runtime. The new API also makes it 6259 easier to port software from Apple to GNU as almost no changes 6260 should be required. The old API is still supported for backwards 6261 compatibility; including the old objc/objc-api.h header file 6262 automatically selects the old API, while including the new 6263 objc/runtime.h header file automatically selects the new API. 6264 Support for the old API is being phased out and upgrading the 6265 software to use the new API is strongly recommended. To check for 6266 the availability of the new API, the __GNU_LIBOBJC__ macro can be 6267 used as older versions of the GNU Objective-C runtime library, 6268 which do not support the new API, do not define such a macro. 6269 * Runtime support for @synchronized has been added. 6270 * Runtime support for Objective-C 2.0 synthesized property accessors 6271 has been added. 6272 * Runtime support for Objective-C 2.0 fast enumeration has been 6273 added. 6274 6275 New Targets and Target Specific Improvements 6276 6277 ARM 6278 6279 * GCC now supports the Cortex-M4 processor implementing the v7-em 6280 version of the architecture using the option -mcpu=cortex-m4. 6281 * Scheduling descriptions for the Cortex-M4, the Neon and the 6282 floating point units of the Cortex-A9 and a pipeline description 6283 for the Cortex-A5 have been added. 6284 * Synchronization primitives such as __sync_fetch_and_add and friends 6285 are now inlined for supported architectures rather than calling 6286 into a kernel helper function. 6287 * SSA loop prefetching is enabled by default for the Cortex-A9 at 6288 -O3. 6289 * Several improvements were committed to improve code generation for 6290 the ARM architecture including a rewritten implementation for load 6291 and store multiples. 6292 * Several enhancements were committed to improve SIMD code generation 6293 for NEON by adding support for widening instructions, misaligned 6294 loads and stores, vector conditionals and support for 64 bit 6295 arithmetic. 6296 * Support was added for the Faraday cores fa526, fa606te, fa626te, 6297 fmp626te, fmp626 and fa726te and can be used with the respective 6298 names as parameters to the -mcpu= option. 6299 * Basic support was added for Cortex-A15 and is available through 6300 -mcpu=cortex-a15. 6301 * GCC for AAPCS configurations now more closely adheres to the AAPCS 6302 specification by enabling -fstrict-volatile-bitfields by default. 6303 6304 IA-32/x86-64 6305 6306 * The new -fsplit-stack option permits programs to use a 6307 discontiguous stack. This is useful for threaded programs, in that 6308 it is no longer necessary to specify the maximum stack size when 6309 creating a thread. This feature is currently only implemented for 6310 32-bit and 64-bit x86 GNU/Linux targets. 6311 * Support for emitting profiler counter calls before function 6312 prologues. This is enabled via a new command-line option -mfentry. 6313 * Optimization for the Intel Core 2 processors is now available 6314 through the -march=core2 and -mtune=core2 options. 6315 * Support for Intel Core i3/i5/i7 processors is now available through 6316 the -march=corei7 and -mtune=corei7 options. 6317 * Support for Intel Core i3/i5/i7 processors with AVX is now 6318 available through the -march=corei7-avx and -mtune=corei7-avx 6319 options. 6320 * Support for AMD Bobcat (family 14) processors is now available 6321 through the -march=btver1 and -mtune=btver1 options. 6322 * Support for AMD Bulldozer (family 15) processors is now available 6323 through the -march=bdver1 and -mtune=bdver1 options. 6324 * The default setting (when not optimizing for size) for 32-bit 6325 GNU/Linux and Darwin x86 targets has been changed to 6326 -fomit-frame-pointer. The default can be reverted to 6327 -fno-omit-frame-pointer by configuring GCC with the 6328 --enable-frame-pointer configure option. 6329 * Darwin, FreeBSD, Solaris 2, MinGW and Cygwin now all support 6330 __float128 on 32-bit and 64-bit x86 targets. 6331 * AVX floating-point arithmetic can now be enabled by default at 6332 configure time with the new --with-fpmath=avx option. 6333 * The SSA loop prefetching pass is enabled when using -O3 when 6334 optimizing for CPUs where prefetching is beneficial (AMD CPUs newer 6335 than K6). 6336 * Support for TBM (Trailing Bit Manipulation) built-in functions and 6337 code generation is available via -mtbm. 6338 * Support for AMD's BMI (Bit Manipulation) built-in functions and 6339 code generation is available via -mbmi. 6340 6341 MicroBlaze 6342 6343 * Support has been added for the Xilinx MicroBlaze softcore processor 6344 (microblaze-elf) embedded target. This configurable processor is 6345 supported on several Xilinx Spartan and Virtex FPGAs. 6346 6347 MIPS 6348 6349 * GCC now supports the Loongson 3A processor. Its canonical -march= 6350 and -mtune= name is loongson3a. 6351 6352 MN10300 / AM33 6353 6354 * The inline assembly register constraint "A" has been renamed "c". 6355 This constraint is used to select a floating-point register that 6356 can be used as the destination of a multiply-accumulate 6357 instruction. 6358 * New inline assembly register constraints "A" and "D" have been 6359 added. These constraint letters resolve to all general registers 6360 when compiling for AM33, and resolve to address registers only or 6361 data registers only when compiling for MN10300. 6362 * The MDR register is represented in the compiler. One can access the 6363 register via the "z" constraint in inline assembly. It can be 6364 marked as clobbered or used as a local register variable via the 6365 "mdr" name. The compiler uses the RETF instruction if the function 6366 does not modify the MDR register, so it is important that inline 6367 assembly properly annotate any usage of the register. 6368 6369 PowerPC/PowerPC64 6370 6371 * GCC now supports the Applied Micro Titan processor with 6372 -mcpu=titan. 6373 * The -mrecip option has been added, which indicates whether the 6374 reciprocal and reciprocal square root instructions should be used. 6375 * The -mveclibabi=mass option can be used to enable the compiler to 6376 autovectorize mathematical functions using the Mathematical 6377 Acceleration Subsystem library. 6378 * The -msingle-pic-base option has been added, which instructs the 6379 compiler to avoid loading the PIC base register in function 6380 prologues. The PIC base register must be initialized by the runtime 6381 system. 6382 * The -mblock-move-inline-limit option has been added, which enables 6383 the user to control the maximum size of inlined memcpy calls and 6384 similar. 6385 * PowerPC64 GNU/Linux support for applications requiring a large TOC 6386 section has been improved. A new command-line option, 6387 -mcmodel=MODEL, controls this feature; valid values for MODEL are 6388 small, medium, or large. 6389 * The Altivec builtin functions vec_ld and vec_st have been modified 6390 to generate the Altivec memory instructions LVX and STVX, even if 6391 the -mvsx option is used. In the initial GCC 4.5 release, these 6392 builtin functions were changed to generate VSX memory reference 6393 instructions instead of Altivec memory instructions, but there are 6394 differences between the two instructions. If the VSX instruction 6395 set is available, you can now use the new builtin functions 6396 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 6397 instructions. 6398 * The GCC compiler on AIX now defaults to a process layout with a 6399 larger data space allowing larger programs to be compiled. 6400 * The GCC long double type on AIX 6.1 and above has reverted to 64 6401 bit double precision, matching the AIX XL compiler default, because 6402 of missing C99 symbols required by the GCC runtime. 6403 * The default processor scheduling model and tuning for PowerPC64 6404 GNU/Linux and for AIX 6.1 and above now is POWER7. 6405 * Starting with GCC 4.6.1, vectors of type vector long long or vector 6406 long are passed and returned in the same method as other vectors 6407 with the VSX instruction set. Previously the GCC compiler did not 6408 adhere to the ABI for 128-bit vectors with 64-bit integer base 6409 types (PR 48857). This is also fixed in the GCC 4.5.4 release. 6410 6411 S/390, zSeries and System z9/z10, IBM zEnterprise z196 6412 6413 * Support for the zEnterprise z196 processor has been added. When 6414 using the -march=z196 option, the compiler will generate code 6415 making use of the following instruction facilities: 6416 + Conditional load/store 6417 + Distinct-operands 6418 + Floating-point-extension 6419 + Interlocked-access 6420 + Population-count 6421 The -mtune=z196 option avoids the compare and branch instructions 6422 as well as the load address instruction with an index register as 6423 much as possible and performs instruction scheduling appropriate 6424 for the new out-of-order pipeline architecture. 6425 * When using the -m31 -mzarch options the generated code still 6426 conforms to the 32-bit ABI but uses the general purpose registers 6427 as 64-bit registers internally. This requires a Linux kernel saving 6428 the whole 64-bit registers when doing a context switch. Kernels 6429 providing that feature indicate that by the 'highgprs' string in 6430 /proc/cpuinfo. 6431 * The SSA loop prefetching pass is enabled when using -O3. 6432 6433 SPARC 6434 6435 * GCC now supports the LEON series of SPARC V8 processors. The code 6436 generated by the compiler can either be tuned to it by means of the 6437 --with-tune=leon configure option and -mtune=leon compilation 6438 option, or the compiler can be built for the sparc-leon-{elf,linux} 6439 and sparc-leon3-{elf,linux} targets directly. 6440 * GCC has stopped sign/zero-extending parameter registers in the 6441 callee for functions taking parameters with sub-word size in 32-bit 6442 mode, since this is redundant with the specification of the ABI. 6443 GCC has never done so in 64-bit mode since this is also redundant. 6444 * The command-line option -mfix-at697f has been added to enable the 6445 documented workaround for the single erratum of the Atmel AT697F 6446 processor. 6447 6448 Operating Systems 6449 6450 Android 6451 6452 * GCC now supports the Bionic C library and provides a convenient way 6453 of building native libraries and applications for the Android 6454 platform. Refer to the documentation of the -mandroid and -mbionic 6455 options for details on building native code. At the moment, Android 6456 support is enabled only for ARM. 6457 6458 Darwin/Mac OS X 6459 6460 * General 6461 + Initial support for CFString types has been added. 6462 This allows GCC to build projects including the system Core 6463 Foundation frameworks. The GCC Objective-C family supports 6464 CFString "toll-free bridged" as per the Mac OS X system tools. 6465 CFString is also recognized in the context of format 6466 attributes and arguments (see the documentation for format 6467 attributes for limitations). At present, 8-bit character types 6468 are supported. 6469 + Object file size reduction. 6470 The Darwin zeroed memory allocators have been re-written to 6471 make more use of .zerofill sections. For non-debug code, this 6472 can reduce object file size significantly. 6473 + Objective-C family 64-bit support (NeXT ABI 2). 6474 Initial support has been added to support 64-bit Objective-C 6475 code using the Darwin/OS X native (NeXT) runtime. ABI version 6476 2 will be selected automatically when 64-bit code is built. 6477 + Objective-C family 32-bit ABI 1. 6478 For 32-bit code ABI 1 is also now also allowed. At present it 6479 must be selected manually using -fobjc-abi-version=1 where 6480 applicable - i.e. on Darwin 9/10 (OS X 10.5/10.6). 6481 * x86 Architecture 6482 + The -mdynamic-no-pic option has been enabled. 6483 Code supporting -mdynamic-no-pic optimization has been added 6484 and is applicable to -m32 builds. The compiler bootstrap uses 6485 the option where appropriate. 6486 + The default value for -mtune= has been changed. 6487 Since Darwin systems are primarily Xeon, Core-2 or similar the 6488 default tuning has been changed to -mtune=core2. 6489 + Enable 128-bit long double (__float128) support on Darwin. 6490 * PPC Architecture 6491 + Darwin64 ABI. 6492 Several significant bugs have been fixed, such that GCC now 6493 produces code compatible with the Darwin64 PowerPC ABI. 6494 + libffi and boehm-gc. 6495 The Darwin ports of the libffi and boehm-gc libraries have 6496 been upgraded to include a Darwin64 implementation. This means 6497 that powerpc*-*-darwin9 platforms may now, for example, build 6498 Java applications with -m64 enabled. 6499 + Plug-in support has been enabled. 6500 + The -fsection-anchors option is now available although, 6501 presently, not heavily tested. 6502 6503 Solaris 2 6504 6505 New Features 6506 6507 * Support symbol versioning with the Sun linker. 6508 * Allow libstdc++ to leverage full ISO C99 support on Solaris 10+. 6509 * Support thread-local storage (TLS) with the Sun assembler on 6510 Solaris 2/x86. 6511 * Support TLS on Solaris 8/9 if prerequisites are met. 6512 * Support COMDAT group with the GNU assembler and recent Sun linker. 6513 * Support the Sun assembler visibility syntax. 6514 * Default Solaris 2/x86 to -march=pentium4 (Solaris 10+) resp. 6515 -march=pentiumpro (Solaris 8/9). 6516 * Don't use SSE on Solaris 8/9 x86 by default. 6517 * Enable 128-bit long double (__float128) support on Solaris 2/x86. 6518 6519 ABI Change 6520 6521 * Change the ABI for returning 8-byte vectors like __m64 in MMX 6522 registers on Solaris 10+/x86 to match the Sun Studio 12.1+ 6523 compilers. This is an incompatible change. If you use such types, 6524 you must either recompile all your code with the new compiler or 6525 use the new -mvect8-ret-in-mem option to remain compatible with 6526 previous versions of GCC and Sun Studio. 6527 6528 Windows x86/x86_64 6529 6530 * Initial support for decimal floating point. 6531 * Support for the __thiscall calling-convention. 6532 * Support for hot-patchable function prologues via the 6533 ms_hook_prologue attribute for x86_64 in addition to 32-bit x86. 6534 * Improvements of stack-probing and stack-allocation mechanisms. 6535 * Support of push/pop-macro pragma as preprocessor command. 6536 With #pragma push_macro("macro-name") the current definition of 6537 macro-name is saved and can be restored with #pragma 6538 pop_macro("macro-name") to its saved definition. 6539 * Enable 128-bit long double (__float128) support on MinGW and 6540 Cygwin. 6541 6542 Other significant improvements 6543 6544 Installation changes 6545 6546 * An install-strip make target is provided that installs stripped 6547 executables, and may install libraries with unneeded or debugging 6548 sections stripped. 6549 * On Power7 systems, there is a potential problem if you build the 6550 GCC compiler with a host compiler using options that enable the VSX 6551 instruction set generation. If the host compiler has been patched 6552 so that the vec_ld and vec_st builtin functions generate Altivec 6553 memory instructions instead of VSX memory instructions, then you 6554 should be able to build the compiler with VSX instruction 6555 generation. 6556 6557 Changes for GCC Developers 6558 6559 Note: these changes concern developers that develop GCC itself or 6560 software that integrates with GCC, such as plugins, and not the general 6561 GCC users. 6562 * The gengtype utility, which previously was internal to the GCC 6563 build process, has been enchanced to provide GC root information 6564 for plugins as necessary. 6565 * The old GC allocation interface of ggc_alloc and friends was 6566 replaced with a type-safe alternative. 6567 6568 GCC 4.6.1 6569 6570 This is the [20]list of problem reports (PRs) from GCC's bug tracking 6571 system that are known to be fixed in the 4.6.1 release. This list might 6572 not be complete (that is, it is possible that some PRs that have been 6573 fixed are not listed here). 6574 6575 GCC 4.6.2 6576 6577 This is the [21]list of problem reports (PRs) from GCC's bug tracking 6578 system that are known to be fixed in the 4.6.2 release. This list might 6579 not be complete (that is, it is possible that some PRs that have been 6580 fixed are not listed here). 6581 6582 GCC 4.6.3 6583 6584 This is the [22]list of problem reports (PRs) from GCC's bug tracking 6585 system that are known to be fixed in the 4.6.3 release. This list might 6586 not be complete (that is, it is possible that some PRs that have been 6587 fixed are not listed here). 6588 6589 GCC 4.6.4 6590 6591 This is the [23]list of problem reports (PRs) from GCC's bug tracking 6592 system that are known to be fixed in the 4.6.4 release. This list might 6593 not be complete (that is, it is possible that some PRs that have been 6594 fixed are not listed here). 6595 6596 6597 For questions related to the use of GCC, please consult these web 6598 pages and the [24]GCC manuals. If that fails, the 6599 [25]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6600 web pages and the development of GCC are welcome on our developer 6601 list at [26]gcc (a] gcc.gnu.org. All of [27]our lists have public 6602 archives. 6603 6604 Copyright (C) [28]Free Software Foundation, Inc. Verbatim copying and 6605 distribution of this entire article is permitted in any medium, 6606 provided this notice is preserved. 6607 6608 These pages are [29]maintained by the GCC team. Last modified 6609 2018-09-30[30]. 6610 6611 References 6612 6613 1. https://sourceware.org/bugzilla/show_bug.cgi?id=10401 6614 2. http://gcc.gnu.org/gcc-4.5/changes.html 6615 3. http://gcc.gnu.org/gcc-4.5/changes.html#obsoleted 6616 4. http://gcc.gnu.org/gcc-4.6/porting_to.html 6617 5. http://gcc.gnu.org/projects/lto/whopr.pdf 6618 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 6619 7. http://gcc.gnu.org/gcc-4.6/cxx0x_status.html 6620 8. https://gcc.gnu.org/PR43145 6621 9. https://gcc.gnu.org/PR43680 6622 10. https://gcc.gnu.org/PR33558 6623 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_active.html#253 6624 12. https://gcc.gnu.org/onlinedocs/gcc-4.6.4/libstdc++/manual/manual/status.html#status.iso.200x 6625 13. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug_mode.html 6626 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/debug.html#debug.races 6627 15. https://gcc.gnu.org/onlinedocs/gcc/Preprocessor-Options.html 6628 16. https://gcc.gnu.org/wiki/OOP 6629 17. https://gcc.gnu.org/wiki/Coarray 6630 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcoarray_007d-233 6631 19. https://golang.org/ 6632 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.1 6633 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.2 6634 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.3 6635 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.6.4 6636 24. https://gcc.gnu.org/onlinedocs/ 6637 25. mailto:gcc-help (a] gcc.gnu.org 6638 26. mailto:gcc (a] gcc.gnu.org 6639 27. https://gcc.gnu.org/lists.html 6640 28. https://www.fsf.org/ 6641 29. https://gcc.gnu.org/about.html 6642 30. http://validator.w3.org/check/referer 6643 ====================================================================== 6644 http://gcc.gnu.org/gcc-4.5/index.html 6645 6646 GCC 4.5 Release Series 6647 6648 Jul 2, 2012 6649 6650 The [1]GNU project and the GCC developers are pleased to announce the 6651 release of GCC 4.5.4. 6652 6653 This release is a bug-fix release, containing fixes for regressions in 6654 GCC 4.5.3 relative to previous releases of GCC. 6655 6656 This release series is no longer maintained. 6657 6658 Release History 6659 6660 GCC 4.5.4 6661 Jul 2, 2012 ([2]changes) 6662 6663 GCC 4.5.3 6664 Apr 28, 2011 ([3]changes) 6665 6666 GCC 4.5.2 6667 Dec 16, 2010 ([4]changes) 6668 6669 GCC 4.5.1 6670 Jul 31, 2010 ([5]changes) 6671 6672 GCC 4.5.0 6673 April 14, 2010 ([6]changes) 6674 6675 References and Acknowledgements 6676 6677 GCC used to stand for the GNU C Compiler, but since the compiler 6678 supports several other languages aside from C, it now stands for the 6679 GNU Compiler Collection. 6680 6681 A list of [7]successful builds is updated as new information becomes 6682 available. 6683 6684 The GCC developers would like to thank the numerous people that have 6685 contributed new features, improvements, bug fixes, and other changes as 6686 well as test results to GCC. This [8]amazing group of volunteers is 6687 what makes GCC successful. 6688 6689 For additional information about GCC please refer to the [9]GCC project 6690 web site or contact the [10]GCC development mailing list. 6691 6692 To obtain GCC please use [11]our mirror sites or [12]our SVN server. 6693 6694 6695 For questions related to the use of GCC, please consult these web 6696 pages and the [13]GCC manuals. If that fails, the 6697 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 6698 web pages and the development of GCC are welcome on our developer 6699 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 6700 archives. 6701 6702 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 6703 distribution of this entire article is permitted in any medium, 6704 provided this notice is preserved. 6705 6706 These pages are [18]maintained by the GCC team. Last modified 6707 2018-09-30[19]. 6708 6709 References 6710 6711 1. http://www.gnu.org/ 6712 2. http://gcc.gnu.org/gcc-4.5/changes.html 6713 3. http://gcc.gnu.org/gcc-4.5/changes.html 6714 4. http://gcc.gnu.org/gcc-4.5/changes.html 6715 5. http://gcc.gnu.org/gcc-4.5/changes.html 6716 6. http://gcc.gnu.org/gcc-4.5/changes.html 6717 7. http://gcc.gnu.org/gcc-4.5/buildstat.html 6718 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 6719 9. http://gcc.gnu.org/index.html 6720 10. mailto:gcc (a] gcc.gnu.org 6721 11. http://gcc.gnu.org/mirrors.html 6722 12. http://gcc.gnu.org/svn.html 6723 13. https://gcc.gnu.org/onlinedocs/ 6724 14. mailto:gcc-help (a] gcc.gnu.org 6725 15. mailto:gcc (a] gcc.gnu.org 6726 16. https://gcc.gnu.org/lists.html 6727 17. https://www.fsf.org/ 6728 18. https://gcc.gnu.org/about.html 6729 19. http://validator.w3.org/check/referer 6730 ====================================================================== 6731 http://gcc.gnu.org/gcc-4.5/changes.html 6732 6733 GCC 4.5 Release Series 6734 Changes, New Features, and Fixes 6735 6736 Caveats 6737 6738 * GCC now requires the [1]MPC library in order to build. See the 6739 [2]prerequisites page for version requirements. 6740 * Support for a number of older systems and recently unmaintained or 6741 untested target ports of GCC has been declared obsolete in GCC 4.5. 6742 Unless there is activity to revive them, the next release of GCC 6743 will have their sources permanently removed. 6744 The following ports for individual systems on particular 6745 architectures have been obsoleted: 6746 + IRIX releases before 6.5 (mips-sgi-irix5*, 6747 mips-sgi-irix6.[0-4]) 6748 + Solaris 7 (*-*-solaris2.7) 6749 + Tru64 UNIX releases before V5.1 (alpha*-dec-osf4*, 6750 alpha-dec-osf5.0*) 6751 + Details for the IRIX, Solaris 7, and Tru64 UNIX obsoletions 6752 can be found in the [3]announcement. 6753 Support for the classic POWER architecture implemented in the 6754 original RIOS and RIOS2 processors of the old IBM RS/6000 product 6755 line has been obsoleted in the rs6000 port. This does not affect 6756 the new generation Power and PowerPC architectures. 6757 * Support has been removed for all the [4]configurations obsoleted in 6758 GCC 4.4. 6759 * Support has been removed for the protoize and unprotoize utilities, 6760 obsoleted in GCC 4.4. 6761 * Support has been removed for tuning for Itanium1 (Merced) variants. 6762 Note that code tuned for Itanium2 should also run correctly on 6763 Itanium1. 6764 * GCC now generates unwind info also for epilogues. DWARF debuginfo 6765 generated by GCC now uses more features of DWARF3 than before, and 6766 also some DWARF4 features. GDB older than 7.0 is not able to handle 6767 either of these, so to debug GCC 4.5 generated binaries or 6768 libraries GDB 7.0 or later is needed. You can disable use of DWARF4 6769 features with the -gdwarf-3 -gstrict-dwarf options, or use 6770 -gdwarf-2 -gstrict-dwarf to restrict GCC to just DWARF2, but 6771 epilogue unwind info is emitted unconditionally whenever unwind 6772 info is emitted. 6773 * On x86 targets, code containing floating-point calculations may run 6774 significantly more slowly when compiled with GCC 4.5 in strict C99 6775 conformance mode than they did with earlier GCC versions. This is 6776 due to stricter standard conformance of the compiler and can be 6777 avoided by using the option -fexcess-precision=fast; also see 6778 [5]below. 6779 * The function attribute noinline no longer prevents GCC from cloning 6780 the function. A new attribute noclone has been introduced for this 6781 purpose. Cloning a function means that it is duplicated and the new 6782 copy is specialized for certain contexts (for example when a 6783 parameter is a known constant). 6784 6785 General Optimizer Improvements 6786 6787 * The -save-temps now takes an optional argument. The -save-temps and 6788 -save-temps=cwd switches write the temporary files in the current 6789 working directory based on the original source file. The 6790 -save-temps=obj switch will write files into the directory 6791 specified with the -o option, and the intermediate filenames are 6792 based on the output file. This will allow the user to get the 6793 compiler intermediate files when doing parallel builds without two 6794 builds of the same filename located in different directories from 6795 interfering with each other. 6796 * Debugging dumps are now created in the same directory as the object 6797 file rather than in the current working directory. This allows the 6798 user to get debugging dumps when doing parallel builds without two 6799 builds of the same filename interfering with each other. 6800 * GCC has been integrated with the MPC library. This allows GCC to 6801 evaluate complex arithmetic at compile time [6]more accurately. It 6802 also allows GCC to evaluate calls to complex built-in math 6803 functions having constant arguments and replace them at compile 6804 time with their mathematically equivalent results. In doing so, GCC 6805 can generate correct results regardless of the math library 6806 implementation or floating point precision of the host platform. 6807 This also allows GCC to generate identical results regardless of 6808 whether one compiles in native or cross-compile configurations to a 6809 particular target. The following built-in functions take advantage 6810 of this new capability: cacos, cacosh, casin, casinh, catan, 6811 catanh, ccos, ccosh, cexp, clog, cpow, csin, csinh, csqrt, ctan, 6812 and ctanh. The float and long double variants of these functions 6813 (e.g. csinf and csinl) are also handled. 6814 * A new link-time optimizer has been added ([7]-flto). When this 6815 option is used, GCC generates a bytecode representation of each 6816 input file and writes it to specially-named sections in each object 6817 file. When the object files are linked together, all the function 6818 bodies are read from these named sections and instantiated as if 6819 they had been part of the same translation unit. This enables 6820 interprocedural optimizations to work across different files (and 6821 even different languages), potentially improving the performance of 6822 the generated code. To use the link-timer optimizer, -flto needs to 6823 be specified at compile time and during the final link. If the 6824 program does not require any symbols to be exported, it is possible 6825 to combine -flto and the experimental [8]-fwhopr with 6826 [9]-fwhole-program to allow the interprocedural optimizers to use 6827 more aggressive assumptions. 6828 * The automatic parallelization pass was enhanced to support 6829 parallelization of outer loops. 6830 * Automatic parallelization can be enabled as part of Graphite. In 6831 addition to -ftree-parallelize-loops=, specify 6832 -floop-parallelize-all to enable the Graphite-based optimization. 6833 * The infrastructure for optimizing based on [10]restrict qualified 6834 pointers has been rewritten and should result in code generation 6835 improvements. Optimizations based on restrict qualified pointers 6836 are now also available when using -fno-strict-aliasing. 6837 * There is a new optimization pass that attempts to change prototype 6838 of functions to avoid unused parameters, pass only relevant parts 6839 of structures and turn arguments passed by reference to arguments 6840 passed by value when possible. It is enabled by -O2 and above as 6841 well as -Os and can be manually invoked using the new command-line 6842 switch -fipa-sra. 6843 * GCC now optimize exception handling code. In particular cleanup 6844 regions that are proved to not have any effect are optimized out. 6845 6846 New Languages and Language specific improvements 6847 6848 All languages 6849 6850 * The -fshow-column option is now on by default. This means error 6851 messages now have a column associated with them. 6852 6853 Ada 6854 6855 * Compilation of programs heavily using discriminated record types 6856 with variant parts has been sped up and generates more compact 6857 code. 6858 * Stack checking now works reasonably well on most plaforms. In some 6859 specific cases, stack overflows may still fail to be detected, but 6860 a compile-time warning will be issued for these cases. 6861 6862 C family 6863 6864 * If a header named in a #include directive is not found, the 6865 compiler exits immediately. This avoids a cascade of errors arising 6866 from declarations expected to be found in that header being 6867 missing. 6868 * A new built-in function __builtin_unreachable() has been added that 6869 tells the compiler that control will never reach that point. It may 6870 be used after asm statements that terminate by transferring control 6871 elsewhere, and in other places that are known to be unreachable. 6872 * The -Wlogical-op option now warns for logical expressions such as 6873 (c == 1 && c == 2) and (c != 1 || c != 2), which are likely to be 6874 mistakes. This option is disabled by default. 6875 * An asm goto feature has been added to allow asm statements that 6876 jump to C labels. 6877 * C++0x raw strings are supported for C++ and for C with -std=gnu99. 6878 * The deprecated attribute now takes an optional string argument, for 6879 example, __attribute__((deprecated("text string"))), that will be 6880 printed together with the deprecation warning. 6881 6882 C 6883 6884 * The -Wenum-compare option, which warns when comparing values of 6885 different enum types, now works for C. It formerly only worked for 6886 C++. This warning is enabled by -Wall. It may be avoided by using a 6887 type cast. 6888 * The -Wcast-qual option now warns about casts which are unsafe in 6889 that they permit const-correctness to be violated without further 6890 warnings. Specifically, it warns about cases where a qualifier is 6891 added when all the lower types are not const. For example, it warns 6892 about a cast from char ** to const char **. 6893 * The -Wc++-compat option is significantly improved. It issues new 6894 warnings for: 6895 + Using C++ reserved operator names as identifiers. 6896 + Conversions to enum types without explicit casts. 6897 + Using va_arg with an enum type. 6898 + Using different enum types in the two branches of ?:. 6899 + Using ++ or -- on a variable of enum type. 6900 + Using the same name as both a struct, union or enum tag and a 6901 typedef, unless the typedef refers to the tagged type itself. 6902 + Using a struct, union, or enum which is defined within another 6903 struct or union. 6904 + A struct field defined using a typedef if there is a field in 6905 the struct, or an enclosing struct, whose name is the typedef 6906 name. 6907 + Duplicate definitions at file scope. 6908 + Uninitialized const variables. 6909 + A global variable with an anonymous struct, union, or enum 6910 type. 6911 + Using a string constant to initialize a char array whose size 6912 is the length of the string. 6913 * The new -Wjump-misses-init option warns about cases where a goto or 6914 switch skips the initialization of a variable. This sort of branch 6915 is an error in C++ but not in C. This warning is enabled by 6916 -Wc++-compat. 6917 * GCC now ensures that a C99-conforming <stdint.h> is present on most 6918 targets, and uses information about the types in this header to 6919 implement the Fortran bindings to those types. GCC does not ensure 6920 the presence of such a header, and does not implement the Fortran 6921 bindings, on the following targets: NetBSD, VxWorks, VMS, 6922 SymbianOS, WinCE, LynxOS, Netware, QNX, Interix, TPF. 6923 * GCC now implements C90- and C99-conforming rules for constant 6924 expressions. This may cause warnings or errors for some code using 6925 expressions that can be folded to a constant but are not constant 6926 expressions as defined by ISO C. 6927 * All known target-independent C90 and C90 Amendment 1 conformance 6928 bugs, and all known target-independent C99 conformance bugs not 6929 related to floating point or extended identifiers, have been fixed. 6930 * The C decimal floating point support now includes support for the 6931 FLOAT_CONST_DECIMAL64 pragma. 6932 * The named address space feature from ISO/IEC TR 18037 is now 6933 supported. This is currently only implemented for the SPU 6934 processor. 6935 6936 C++ 6937 6938 * Improved [11]experimental support for the upcoming C++0x ISO C++ 6939 standard, including support for raw strings, lambda expressions and 6940 explicit type conversion operators. 6941 * When printing the name of a class template specialization, G++ will 6942 now omit any template arguments which come from default template 6943 arguments. This behavior (and the pretty-printing of function 6944 template specializations as template signature and arguments) can 6945 be disabled with the -fno-pretty-templates option. 6946 * Access control is now applied to typedef names used in a template, 6947 which may cause G++ to reject some ill-formed code that was 6948 accepted by earlier releases. The -fno-access-control option can be 6949 used as a temporary workaround until the code is corrected. 6950 * Compilation time for code that uses templates should now scale 6951 linearly with the number of instantiations rather than 6952 quadratically, as template instantiations are now looked up using 6953 hash tables. 6954 * Declarations of functions that look like builtin declarations of 6955 library functions are only considered to be redeclarations if they 6956 are declared with extern "C". This may cause problems with code 6957 that omits extern "C" on hand-written declarations of C library 6958 functions such as abort or memcpy. Such code is ill-formed, but was 6959 accepted by earlier releases. 6960 * Diagnostics that used to complain about passing non-POD types to 6961 ... or jumping past the declaration of a non-POD variable now check 6962 for triviality rather than PODness, as per C++0x. 6963 * In C++0x mode local and anonymous classes are now allowed as 6964 template arguments, and in declarations of variables and functions 6965 with linkage, so long as any such declaration that is used is also 6966 defined ([12]DR 757). 6967 * Labels may now have attributes, as has been permitted for a while 6968 in C. This is only permitted when the label definition and the 6969 attribute specifier is followed by a semicolon--i.e., the label 6970 applies to an empty statement. The only useful attribute for a 6971 label is unused. 6972 * G++ now implements [13]DR 176. Previously G++ did not support using 6973 the injected-class-name of a template base class as a type name, 6974 and lookup of the name found the declaration of the template in the 6975 enclosing scope. Now lookup of the name finds the 6976 injected-class-name, which can be used either as a type or as a 6977 template, depending on whether or not the name is followed by a 6978 template argument list. As a result of this change, some code that 6979 was previously accepted may be ill-formed because 6980 1. The injected-class-name is not accessible because it's from a 6981 private base, or 6982 2. The injected-class-name cannot be used as an argument for a 6983 template template parameter. 6984 In either of these cases, the code can be fixed by adding a 6985 nested-name-specifier to explicitly name the template. The first 6986 can be worked around with -fno-access-control; the second is only 6987 rejected with -pedantic. 6988 * A new standard mangling for SIMD vector types has been added, to 6989 avoid name clashes on systems with vectors of varying length. By 6990 default the compiler still uses the old mangling, but emits aliases 6991 with the new mangling on targets that support strong aliases. Users 6992 can switch over entirely to the new mangling with -fabi-version=4 6993 or -fabi-version=0. -Wabi will now warn about code that uses the 6994 old mangling. 6995 * The command-line option -ftemplate-depth-N is now written as 6996 -ftemplate-depth=N and the old form is deprecated. 6997 * Conversions between NULL and non-pointer types are now warned by 6998 default. The new option -Wno-conversion-null disables these 6999 warnings. Previously these warnings were only available when using 7000 -Wconversion explicitly. 7001 7002 Runtime Library (libstdc++) 7003 7004 * Improved experimental support for the upcoming ISO C++ standard, 7005 C++0x, including: 7006 + Support for <future>, <functional>, and <random>. 7007 + Existing facilities now exploit explicit operators and the 7008 newly implemented core C++0x features. 7009 + The header <cstdatomic> has been renamed to <atomic>. 7010 * An experimental [14]profile mode has been added. This is an 7011 implementation of many C++ standard library constructs with an 7012 additional analysis layer that gives performance improvement advice 7013 based on recognition of suboptimal usage patterns. For example, 7014 #include <vector> 7015 int main() 7016 { 7017 std::vector<int> v; 7018 for (int k = 0; k < 1024; ++k) 7019 v.insert(v.begin(), k); 7020 } 7021 7022 When instrumented via the profile mode, can return suggestions 7023 about the initial size and choice of the container used as follows: 7024 vector-to-list: improvement = 5: call stack = 0x804842c ... 7025 : advice = change std::vector to std::list 7026 vector-size: improvement = 3: call stack = 0x804842c ... 7027 : advice = change initial container size from 0 to 1024 7028 7029 These constructs can be substituted for the normal libstdc++ 7030 constructs on a piecemeal basis, or all existing components can be 7031 transformed via the -D_GLIBCXX_PROFILE macro. 7032 * [15]Support for decimal floating-point arithmetic (aka ISO C++ TR 7033 24733) has been added. This support is in header file 7034 <decimal/decimal>, uses namespace std::decimal, and includes 7035 classes decimal32, decimal64, and decimal128. 7036 * Sources have been audited for application of function attributes 7037 nothrow, const, pure, and noreturn. 7038 * Python pretty-printers have been added for many standard library 7039 components that simplify the internal representation and present a 7040 more intuitive view of components when used with 7041 appropriately-advanced versions of GDB. For more information, 7042 please consult the more [16]detailed description. 7043 * The default behavior for comparing typeinfo names has changed, so 7044 in <typeinfo>, __GXX_MERGED_TYPEINFO_NAMES now defaults to zero. 7045 * The new -static-libstdc++ option directs g++ to link the C++ 7046 library statically, even if the default would normally be to link 7047 it dynamically. 7048 7049 Fortran 7050 7051 * The COMMON default padding has been changed - instead of adding the 7052 padding before a variable it is now added afterwards, which 7053 increases the compatibility with other vendors and helps to obtain 7054 the correct output in some cases. Cf. also the -falign-commons 7055 option ([17]added in 4.4). 7056 * The -finit-real= option now also supports the value snan for 7057 signaling not-a-number; to be effective, one additionally needs to 7058 enable trapping (e.g. via -ffpe-trap=). Note: Compile-time 7059 optimizations can turn a signaling NaN into a quiet one. 7060 * The new option -fcheck= has been added with the options bounds, 7061 array-temps, do, pointer, and recursive. The bounds and array-temps 7062 options are equivalent to -fbounds-check and 7063 -fcheck-array-temporaries. The do option checks for invalid 7064 modification of loop iteration variables, and the recursive option 7065 tests for recursive calls to subroutines/functions which are not 7066 marked as recursive. With pointer pointer association checks in 7067 calls are performed; however, neither undefined pointers nor 7068 pointers in expressions are handled. Using -fcheck=all enables all 7069 these run-time checks. 7070 * The run-time checking -fcheck=bounds now warns about invalid string 7071 lengths of character dummy arguments. Additionally, more 7072 compile-time checks have been added. 7073 * The new option [18]-fno-protect-parens has been added; if set, the 7074 compiler may reorder REAL and COMPLEX expressions without regard to 7075 parentheses. 7076 * GNU Fortran no longer links against libgfortranbegin. As before, 7077 MAIN__ (assembler symbol name) is the actual Fortran main program, 7078 which is invoked by the main function. However, main is now 7079 generated and put in the same object file as MAIN__. For the time 7080 being, libgfortranbegin still exists for backward compatibility. 7081 For details see the new [19]Mixed-Language Programming chapter in 7082 the manual. 7083 * The I/O library was restructured for performance and cleaner code. 7084 * Array assignments and WHERE are now run in parallel when OpenMP's 7085 WORKSHARE is used. 7086 * The experimental option -fwhole-file was added. The option allows 7087 whole-file checking of procedure arguments and allows for better 7088 optimizations. It can also be used with -fwhole-program, which is 7089 now also supported in gfortran. 7090 * More Fortran 2003 and Fortran 2008 mathematical functions can now 7091 be used as initialization expressions. 7092 * Some extended attributes such as STDCALL are now supported via the 7093 [20]GCC$ compiler directive. 7094 * For Fortran 77 compatibility: If -fno-sign-zero is used, the SIGN 7095 intrinsic behaves now as if zero were always positive. 7096 * For legacy compatibiliy: On Cygwin and MinGW, the special files 7097 CONOUT$ and CONIN$ (and CONERR$ which maps to CONOUT$) are now 7098 supported. 7099 * Fortran 2003 support has been extended: 7100 + Procedure-pointer function results and procedure-pointer 7101 components (including PASS), 7102 + allocatable scalars (experimental), 7103 + DEFERRED type-bound procedures, 7104 + the ERRMSG= argument of the ALLOCATE and DEALLOCATE statements 7105 have been implemented. 7106 + The ALLOCATE statement supports type-specs and the SOURCE= 7107 argument. 7108 + OPERATOR(*) and ASSIGNMENT(=) are now allowed as GENERIC 7109 type-bound procedure (i.e. as type-bound operators). 7110 + Rounding (ROUND=, RZ, ...) for output is now supported. 7111 + The INT_FAST{8,16,32,64,128}_T kind type parameters of the 7112 intrinsic module ISO_C_BINDING are now supported, except for 7113 the targets listed above as ones where GCC does not have 7114 <stdint.h> type information. 7115 + Extensible derived types with type-bound procedure or 7116 procedure pointer with PASS attribute now have to use CLASS in 7117 line with the Fortran 2003 standard; the workaround to use 7118 TYPE is no longer supported. 7119 + [21]Experimental, incomplete support for polymorphism, 7120 including CLASS, SELECT TYPE and dynamic dispatch of 7121 type-bound procedure calls. Some features do not work yet such 7122 as unlimited polymorphism (CLASS(*)). 7123 * Fortran 2008 support has been extended: 7124 + The OPEN statement now supports the NEWUNIT= option, which 7125 returns a unique file unit, thus preventing inadvertent use of 7126 the same unit in different parts of the program. 7127 + Support for unlimited format items has been added. 7128 + The INT{8,16,32} and REAL{32,64,128} kind type parameters of 7129 the intrinsic module ISO_FORTRAN_ENV are now supported. 7130 + Using complex arguments with TAN, SINH, COSH, TANH, ASIN, 7131 ACOS, and ATAN is now possible; the functions ASINH, ACOSH, 7132 and ATANH have been added (for real and complex arguments) and 7133 ATAN(Y,X) is now an alias for ATAN2(Y,X). 7134 + The BLOCK construct has been implemented. 7135 7136 New Targets and Target Specific Improvements 7137 7138 AIX 7139 7140 * Full cross-toolchain support now available with GNU Binutils 7141 7142 ARM 7143 7144 * GCC now supports the Cortex-M0 and Cortex-A5 processors. 7145 * GCC now supports the ARM v7E-M architecture. 7146 * GCC now supports VFPv4-based FPUs and FPUs with 7147 single-precision-only VFP. 7148 * GCC has many improvements to optimization for other ARM processors, 7149 including scheduling support for the integer pipeline on Cortex-A9. 7150 * GCC now supports the IEEE 754-2008 half-precision floating-point 7151 type, and a variant ARM-specific half-precision type. This type is 7152 specified using __fp16, with the layout determined by 7153 -mfp16-format. With appropriate -mfpu options, the Cortex-A9 and 7154 VFPv4 half-precision instructions will be used. 7155 * GCC now supports the variant of AAPCS that uses VFP registers for 7156 parameter passing and return values. 7157 7158 AVR 7159 7160 * The -mno-tablejump option has been removed because it has the same 7161 effect as the -fno-jump-tables option. 7162 * Added support for these new AVR devices: 7163 + ATmega8U2 7164 + ATmega16U2 7165 + ATmega32U2 7166 7167 IA-32/x86-64 7168 7169 * GCC now will set the default for -march= based on the configure 7170 target. 7171 * GCC now supports handling floating-point excess precision arising 7172 from use of the x87 floating-point unit in a way that conforms to 7173 ISO C99. This is enabled with -fexcess-precision=standard and with 7174 standards conformance options such as -std=c99, and may be disabled 7175 using -fexcess-precision=fast. 7176 * Support for the Intel Atom processor is now available through the 7177 -march=atom and -mtune=atom options. 7178 * A new -mcrc32 option is now available to enable crc32 intrinsics. 7179 * A new -mmovbe option is now available to enable GCC to use the 7180 movbe instruction to implement __builtin_bswap32 and 7181 __builtin_bswap64. 7182 * SSE math now can be enabled by default at configure time with the 7183 new --with-fpmath=sse option. 7184 * There is a new intrinsic header file, <x86intrin.h>. It should be 7185 included before using any IA-32/x86-64 intrinsics. 7186 * Support for the XOP, FMA4, and LWP instruction sets for the AMD 7187 Orochi processors are now available with the -mxop, -mfma4, and 7188 -mlwp options. 7189 * The -mabm option enables GCC to use the popcnt and lzcnt 7190 instructions on AMD processors. 7191 * The -mpopcnt option enables GCC to use the popcnt instructions on 7192 both AMD and Intel processors. 7193 7194 M68K/ColdFire 7195 7196 * GCC now supports ColdFire 51xx, 5221x, 5225x, 52274, 52277, 5301x 7197 and 5441x devices. 7198 * GCC now supports thread-local storage (TLS) on M68K and ColdFire 7199 processors. 7200 7201 MeP 7202 7203 Support has been added for the Toshiba Media embedded Processor (MeP, 7204 or mep-elf) embedded target. 7205 7206 MIPS 7207 7208 * GCC now supports MIPS 1004K processors. 7209 * GCC can now be configured with options --with-arch-32, 7210 --with-arch-64, --with-tune-32 and --with-tune-64 to control the 7211 default optimization separately for 32-bit and 64-bit modes. 7212 * MIPS targets now support an alternative _mcount interface, in which 7213 register $12 points to the function's save slot for register $31. 7214 This interface is selected by the -mcount-ra-address option; see 7215 the documentation for more details. 7216 * GNU/Linux targets can now generate read-only .eh_frame sections. 7217 This optimization requires GNU binutils 2.20 or above, and is only 7218 available if GCC is configured with a suitable version of binutils. 7219 * GNU/Linux targets can now attach special relocations to indirect 7220 calls, so that the linker can turn them into direct jumps or 7221 branches. This optimization requires GNU binutils 2.20 or later, 7222 and is automatically selected if GCC is configured with an 7223 appropriate version of binutils. It can be explicitly enabled or 7224 disabled using the -mrelax-pic-calls command-line option. 7225 * GCC now generates more heavily-optimized atomic operations on 7226 Octeon processors. 7227 * MIPS targets now support the -fstack-protector option. 7228 * GCC now supports an -msynci option, which specifies that synci is 7229 enough to flush the instruction cache, without help from the 7230 operating system. GCC uses this information to optimize 7231 automatically-generated cache flush operations, such as those used 7232 for nested functions in C. There is also a --with-synci 7233 configure-time option, which makes -msynci the default. 7234 * GCC supports four new function attributes for interrupt handlers: 7235 interrupt, use_shadow_register_set, keep_interrupts_masked and 7236 use_debug_exception_return. See the documentation for more details 7237 about these attributes. 7238 7239 RS/6000 (POWER/PowerPC) 7240 7241 * GCC now supports the Power ISA 2.06, which includes the VSX 7242 instructions that add vector 64-bit floating point support, new 7243 population count instructions, and conversions between floating 7244 point and unsigned types. 7245 * Support for the power7 processor is now available through the 7246 -mcpu=power7 and -mtune=power7. 7247 * GCC will now vectorize loops that contain simple math functions 7248 like copysign when generating code for altivec or VSX targets. 7249 * Support for the A2 processor is now available through the -mcpu=a2 7250 and -mtune=a2 options. 7251 * Support for the 476 processor is now available through the 7252 -mcpu={476,476fp} and -mtune={476,476fp} options. 7253 * Support for the e500mc64 processor is now available through the 7254 -mcpu=e500mc64 and -mtune=e500mc64 options. 7255 * GCC can now be configured with options --with-cpu-32, 7256 --with-cpu-64, --with-tune-32 and --with-tune-64 to control the 7257 default optimization separately for 32-bit and 64-bit modes. 7258 * Starting with GCC 4.5.4, vectors of type vector long long or vector 7259 long are passed and returned in the same method as other vectors 7260 with the VSX instruction set. Previously the GCC compiler did not 7261 adhere to the ABI for 128-bit vectors with 64-bit integer base 7262 types (PR 48857). This is also fixed in the GCC 4.6.1 release. 7263 7264 RX 7265 7266 Support has been added for the Renesas RX Processor (rx-elf) target. 7267 7268 Operating Systems 7269 7270 Windows (Cygwin and MinGW) 7271 7272 * GCC now installs all the major language runtime libraries as DLLs 7273 when configured with the --enable-shared option. 7274 * GCC now makes use of the new support for aligned common variables 7275 in versions of binutils >= 2.20 to fix bugs in the support for SSE 7276 data types. 7277 * Improvements to the libffi support library increase the reliability 7278 of code generated by GCJ on all Windows platforms. Libgcj is 7279 enabled by default for the first time. 7280 * Libtool improvements simplify installation by placing the generated 7281 DLLs in the correct binaries directory. 7282 * Numerous other minor bugfixes and improvements, and substantial 7283 enhancements to the Fortran language support library. 7284 7285 > 7286 7287 Other significant improvements 7288 7289 Plugins 7290 7291 * It is now possible to extend the compiler without having to modify 7292 its source code. A new option -fplugin=file.so tells GCC to load 7293 the shared object file.so and execute it as part of the compiler. 7294 The internal documentation describes the details on how plugins can 7295 interact with the compiler. 7296 7297 Installation changes 7298 7299 * The move to newer autotools changed default installation 7300 directories and switches to control them: The --with-datarootdir, 7301 --with-docdir, --with-pdfdir, and --with-htmldir switches are not 7302 used any more. Instead, you can now use --datarootdir, --docdir, 7303 --htmldir, and --pdfdir. The default installation directories have 7304 changed as follows according to the GNU Coding Standards: 7305 7306 datarootdir read-only architecture-independent data root [PREFIX/share] 7307 localedir locale-specific message catalogs [DATAROOTDIR/locale] 7308 docdir documentation root [DATAROOTDIR/doc/PACKAGE] 7309 htmldir html documentation [DOCDIR] 7310 dvidir dvi documentation [DOCDIR] 7311 pdfdir pdf documentation [DOCDIR] 7312 psdir ps documentation [DOCDIR] 7313 The following variables have new default values: 7314 7315 datadir read-only architecture-independent data [DATAROOTDIR] 7316 infodir info documentation [DATAROOTDIR/info] 7317 mandir man documentation [DATAROOTDIR/man] 7318 7319 GCC 4.5.1 7320 7321 This is the [22]list of problem reports (PRs) from GCC's bug tracking 7322 system that are known to be fixed in the 4.5.1 release. This list might 7323 not be complete (that is, it is possible that some PRs that have been 7324 fixed are not listed here). 7325 7326 All languages 7327 7328 * GCC's new link-time optimizer ([23]-flto) now also works on a few 7329 non-ELF targets: 7330 + Cygwin (*-cygwin*) 7331 + MinGW (*-mingw*) 7332 + Darwin on x86-64 (x86_64-apple-darwin*) 7333 LTO is not enabled by default for these targets. To enable LTO, you 7334 should configure with the --enable-lto option. 7335 7336 GCC 4.5.2 7337 7338 This is the [24]list of problem reports (PRs) from GCC's bug tracking 7339 system that are known to be fixed in the 4.5.2 release. This list might 7340 not be complete (that is, it is possible that some PRs that have been 7341 fixed are not listed here). 7342 7343 GCC 4.5.3 7344 7345 This is the [25]list of problem reports (PRs) from GCC's bug tracking 7346 system that are known to be fixed in the 4.5.3 release. This list might 7347 not be complete (that is, it is possible that some PRs that have been 7348 fixed are not listed here). 7349 7350 On the PowerPC compiler, the Altivec builtin functions vec_ld and 7351 vec_st have been modified to generate the Altivec memory instructions 7352 LVX and STVX, even if the -mvsx option is used. In the initial GCC 4.5 7353 release, these builtin functions were changed to generate VSX memory 7354 reference instructions instead of Altivec memory instructions, but 7355 there are differences between the two instructions. If the VSX 7356 instruction set is available, you can now use the new builtin functions 7357 vec_vsx_ld and vec_vsx_st which always generates the VSX memory 7358 instructions. 7359 7360 GCC 4.5.4 7361 7362 This is the [26]list of problem reports (PRs) from GCC's bug tracking 7363 system that are known to be fixed in the 4.5.4 release. This list might 7364 not be complete (that is, it is possible that some PRs that have been 7365 fixed are not listed here). 7366 7367 7368 For questions related to the use of GCC, please consult these web 7369 pages and the [27]GCC manuals. If that fails, the 7370 [28]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7371 web pages and the development of GCC are welcome on our developer 7372 list at [29]gcc (a] gcc.gnu.org. All of [30]our lists have public 7373 archives. 7374 7375 Copyright (C) [31]Free Software Foundation, Inc. Verbatim copying and 7376 distribution of this entire article is permitted in any medium, 7377 provided this notice is preserved. 7378 7379 These pages are [32]maintained by the GCC team. Last modified 7380 2018-09-30[33]. 7381 7382 References 7383 7384 1. http://www.multiprecision.org/mpc/ 7385 2. https://gcc.gnu.org/install/prerequisites.html 7386 3. https://gcc.gnu.org/ml/gcc/2010-01/msg00510.html 7387 4. http://gcc.gnu.org/gcc-4.4/changes.html#obsoleted 7388 5. http://gcc.gnu.org/gcc-4.5/changes.html#x86 7389 6. https://gcc.gnu.org/PR30789 7390 7. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 7391 8. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhopr-802 7392 9. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-fwhole-program-800 7393 10. https://gcc.gnu.org/onlinedocs/gcc/Restricted-Pointers.html 7394 11. http://gcc.gnu.org/gcc-4.5/cxx0x_status.html 7395 12. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#757 7396 13. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#176 7397 14. https://gcc.gnu.org/onlinedocs/libstdc++/manual/profile_mode.html 7398 15. https://gcc.gnu.org/onlinedocs/libstdc++/manual/status.html#status.iso.tr24733 7399 16. https://sourceware.org/gdb/wiki/STLSupport 7400 17. http://gcc.gnu.org/gcc-4.4/changes.html 7401 18. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html 7402 19. https://gcc.gnu.org/onlinedocs/gfortran/Mixed-Language-Programming.html 7403 20. https://gcc.gnu.org/onlinedocs/gfortran/GNU-Fortran-Compiler-Directives.html 7404 21. https://gcc.gnu.org/wiki/OOP 7405 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.1 7406 23. https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-801 7407 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.2 7408 25. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.3 7409 26. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.5.4 7410 27. https://gcc.gnu.org/onlinedocs/ 7411 28. mailto:gcc-help (a] gcc.gnu.org 7412 29. mailto:gcc (a] gcc.gnu.org 7413 30. https://gcc.gnu.org/lists.html 7414 31. https://www.fsf.org/ 7415 32. https://gcc.gnu.org/about.html 7416 33. http://validator.w3.org/check/referer 7417 ====================================================================== 7418 http://gcc.gnu.org/gcc-4.4/index.html 7419 7420 GCC 4.4 Release Series 7421 7422 March 13, 2012 7423 7424 The [1]GNU project and the GCC developers are pleased to announce the 7425 release of GCC 4.4.7. 7426 7427 This release is a bug-fix release, containing fixes for regressions in 7428 GCC 4.4.6 relative to previous releases of GCC. 7429 7430 This release series is no longer maintained. 7431 7432 Release History 7433 7434 GCC 4.4.7 7435 March 13, 2012 ([2]changes) 7436 7437 GCC 4.4.6 7438 April 16, 2011 ([3]changes) 7439 7440 GCC 4.4.5 7441 October 1, 2010 ([4]changes) 7442 7443 GCC 4.4.4 7444 April 29, 2010 ([5]changes) 7445 7446 GCC 4.4.3 7447 January 21, 2010 ([6]changes) 7448 7449 GCC 4.4.2 7450 October 15, 2009 ([7]changes) 7451 7452 GCC 4.4.1 7453 July 22, 2009 ([8]changes) 7454 7455 GCC 4.4.0 7456 April 21, 2009 ([9]changes) 7457 7458 References and Acknowledgements 7459 7460 GCC used to stand for the GNU C Compiler, but since the compiler 7461 supports several other languages aside from C, it now stands for the 7462 GNU Compiler Collection. 7463 7464 A list of [10]successful builds is updated as new information becomes 7465 available. 7466 7467 The GCC developers would like to thank the numerous people that have 7468 contributed new features, improvements, bug fixes, and other changes as 7469 well as test results to GCC. This [11]amazing group of volunteers is 7470 what makes GCC successful. 7471 7472 For additional information about GCC please refer to the [12]GCC 7473 project web site or contact the [13]GCC development mailing list. 7474 7475 To obtain GCC please use [14]our mirror sites or [15]our SVN server. 7476 7477 7478 For questions related to the use of GCC, please consult these web 7479 pages and the [16]GCC manuals. If that fails, the 7480 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 7481 web pages and the development of GCC are welcome on our developer 7482 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 7483 archives. 7484 7485 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 7486 distribution of this entire article is permitted in any medium, 7487 provided this notice is preserved. 7488 7489 These pages are [21]maintained by the GCC team. Last modified 7490 2018-09-30[22]. 7491 7492 References 7493 7494 1. http://www.gnu.org/ 7495 2. http://gcc.gnu.org/gcc-4.4/changes.html 7496 3. http://gcc.gnu.org/gcc-4.4/changes.html 7497 4. http://gcc.gnu.org/gcc-4.4/changes.html 7498 5. http://gcc.gnu.org/gcc-4.4/changes.html 7499 6. http://gcc.gnu.org/gcc-4.4/changes.html 7500 7. http://gcc.gnu.org/gcc-4.4/changes.html 7501 8. http://gcc.gnu.org/gcc-4.4/changes.html 7502 9. http://gcc.gnu.org/gcc-4.4/changes.html 7503 10. http://gcc.gnu.org/gcc-4.4/buildstat.html 7504 11. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 7505 12. http://gcc.gnu.org/index.html 7506 13. mailto:gcc (a] gcc.gnu.org 7507 14. http://gcc.gnu.org/mirrors.html 7508 15. http://gcc.gnu.org/svn.html 7509 16. https://gcc.gnu.org/onlinedocs/ 7510 17. mailto:gcc-help (a] gcc.gnu.org 7511 18. mailto:gcc (a] gcc.gnu.org 7512 19. https://gcc.gnu.org/lists.html 7513 20. https://www.fsf.org/ 7514 21. https://gcc.gnu.org/about.html 7515 22. http://validator.w3.org/check/referer 7516 ====================================================================== 7517 http://gcc.gnu.org/gcc-4.4/changes.html 7518 7519 GCC 4.4 Release Series 7520 Changes, New Features, and Fixes 7521 7522 The latest release in the 4.4 release series is [1]GCC 4.4.7. 7523 7524 Caveats 7525 7526 * __builtin_stdarg_start has been completely removed from GCC. 7527 Support for <varargs.h> had been deprecated since GCC 4.0. Use 7528 __builtin_va_start as a replacement. 7529 * Some of the errors issued by the C++ front end that could be 7530 downgraded to warnings in previous releases by using -fpermissive 7531 are now warnings by default. They can be converted into errors by 7532 using -pedantic-errors. 7533 * Use of the cpp assertion extension will now emit a warning when 7534 -Wdeprecated or -pedantic is used. This extension has been 7535 deprecated for many years, but never warned about. 7536 * Packed bit-fields of type char were not properly bit-packed on many 7537 targets prior to GCC 4.4. On these targets, the fix in GCC 4.4 7538 causes an ABI change. For example there is no longer a 4-bit 7539 padding between field a and b in this structure: 7540 struct foo 7541 { 7542 char a:4; 7543 char b:8; 7544 } __attribute__ ((packed)); 7545 There is a new warning to help identify fields that are affected: 7546 foo.c:5: note: Offset of packed bit-field 'b' has changed in GCC 4.4 7547 The warning can be disabled with -Wno-packed-bitfield-compat. 7548 * On ARM EABI targets, the C++ mangling of the va_list type has been 7549 changed to conform to the current revision of the EABI. This does 7550 not affect the libstdc++ library included with GCC. 7551 * The SCOUNT and POS bits of the MIPS DSP control register are now 7552 treated as global. Previous versions of GCC treated these fields as 7553 call-clobbered instead. 7554 * The MIPS port no longer recognizes the h asm constraint. It was 7555 necessary to remove this constraint in order to avoid generating 7556 unpredictable code sequences. 7557 One of the main uses of the h constraint was to extract the high 7558 part of a multiplication on 64-bit targets. For example: 7559 asm ("dmultu\t%1,%2" : "=h" (result) : "r" (x), "r" (y)); 7560 You can now achieve the same effect using 128-bit types: 7561 typedef unsigned int uint128_t __attribute__((mode(TI))); 7562 result = ((uint128_t) x * y) >> 64; 7563 The second sequence is better in many ways. For example, if x and y 7564 are constants, the compiler can perform the multiplication at 7565 compile time. If x and y are not constants, the compiler can 7566 schedule the runtime multiplication better than it can schedule an 7567 asm statement. 7568 * Support for a number of older systems and recently unmaintained or 7569 untested target ports of GCC has been declared obsolete in GCC 4.4. 7570 Unless there is activity to revive them, the next release of GCC 7571 will have their sources permanently removed. 7572 The following ports for individual systems on particular 7573 architectures have been obsoleted: 7574 + Generic a.out on IA32 and m68k (i[34567]86-*-aout*, 7575 m68k-*-aout*) 7576 + Generic COFF on ARM, H8300, IA32, m68k and SH (arm-*-coff*, 7577 armel-*-coff*, h8300-*-*, i[34567]86-*-coff*, m68k-*-coff*, 7578 sh-*-*). This does not affect other more specific targets 7579 using the COFF object format on those architectures, or the 7580 more specific H8300 and SH targets (h8300-*-rtems*, 7581 h8300-*-elf*, sh-*-elf*, sh-*-symbianelf*, sh-*-linux*, 7582 sh-*-netbsdelf*, sh-*-rtems*, sh-wrs-vxworks). 7583 + 2BSD on PDP-11 (pdp11-*-bsd) 7584 + AIX 4.1 and 4.2 on PowerPC (rs6000-ibm-aix4.[12]*, 7585 powerpc-ibm-aix4.[12]*) 7586 + Tuning support for Itanium1 (Merced) variants. Note that code 7587 tuned for Itanium2 should also run correctly on Itanium1. 7588 * The protoize and unprotoize utilities have been obsoleted and will 7589 be removed in GCC 4.5. These utilities have not been installed by 7590 default since GCC 3.0. 7591 * Support has been removed for all the [2]configurations obsoleted in 7592 GCC 4.3. 7593 * Unknown -Wno-* options are now silently ignored by GCC if no other 7594 diagnostics are issued. If other diagnostics are issued, then GCC 7595 warns about the unknown options. 7596 * More information on porting to GCC 4.4 from previous versions of 7597 GCC can be found in the [3]porting guide for this release. 7598 7599 General Optimizer Improvements 7600 7601 * A new command-line switch -findirect-inlining has been added. When 7602 turned on it allows the inliner to also inline indirect calls that 7603 are discovered to have known targets at compile time thanks to 7604 previous inlining. 7605 * A new command-line switch -ftree-switch-conversion has been added. 7606 This new pass turns simple initializations of scalar variables in 7607 switch statements into initializations from a static array, given 7608 that all the values are known at compile time and the ratio between 7609 the new array size and the original switch branches does not exceed 7610 the parameter --param switch-conversion-max-branch-ratio (default 7611 is eight). 7612 * A new command-line switch -ftree-builtin-call-dce has been added. 7613 This optimization eliminates unnecessary calls to certain builtin 7614 functions when the return value is not used, in cases where the 7615 calls can not be eliminated entirely because the function may set 7616 errno. This optimization is on by default at -O2 and above. 7617 * A new command-line switch -fconserve-stack directs the compiler to 7618 minimize stack usage even if it makes the generated code slower. 7619 This affects inlining decisions. 7620 * When the assembler supports it, the compiler will now emit unwind 7621 information using assembler .cfi directives. This makes it possible 7622 to use such directives in inline assembler code. The new option 7623 -fno-dwarf2-cfi-asm directs the compiler to not use .cfi 7624 directives. 7625 * The [4]Graphite branch has been merged. This merge has brought in a 7626 new framework for loop optimizations based on a polyhedral 7627 intermediate representation. These optimizations apply to all the 7628 languages supported by GCC. The following new code transformations 7629 are available in GCC 4.4: 7630 + -floop-interchange performs loop interchange transformations 7631 on loops. Interchanging two nested loops switches the inner 7632 and outer loops. For example, given a loop like: 7633 DO J = 1, M 7634 DO I = 1, N 7635 A(J, I) = A(J, I) * C 7636 ENDDO 7637 ENDDO 7638 7639 loop interchange will transform the loop as if the user had 7640 written: 7641 DO I = 1, N 7642 DO J = 1, M 7643 A(J, I) = A(J, I) * C 7644 ENDDO 7645 ENDDO 7646 7647 which can be beneficial when N is larger than the caches, 7648 because in Fortran, the elements of an array are stored in 7649 memory contiguously by column, and the original loop iterates 7650 over rows, potentially creating at each access a cache miss. 7651 + -floop-strip-mine performs loop strip mining transformations 7652 on loops. Strip mining splits a loop into two nested loops. 7653 The outer loop has strides equal to the strip size and the 7654 inner loop has strides of the original loop within a strip. 7655 For example, given a loop like: 7656 DO I = 1, N 7657 A(I) = A(I) + C 7658 ENDDO 7659 7660 loop strip mining will transform the loop as if the user had 7661 written: 7662 DO II = 1, N, 4 7663 DO I = II, min (II + 3, N) 7664 A(I) = A(I) + C 7665 ENDDO 7666 ENDDO 7667 7668 + -floop-block performs loop blocking transformations on loops. 7669 Blocking strip mines each loop in the loop nest such that the 7670 memory accesses of the element loops fit inside caches. For 7671 example, given a loop like: 7672 DO I = 1, N 7673 DO J = 1, M 7674 A(J, I) = B(I) + C(J) 7675 ENDDO 7676 ENDDO 7677 7678 loop blocking will transform the loop as if the user had 7679 written: 7680 DO II = 1, N, 64 7681 DO JJ = 1, M, 64 7682 DO I = II, min (II + 63, N) 7683 DO J = JJ, min (JJ + 63, M) 7684 A(J, I) = B(I) + C(J) 7685 ENDDO 7686 ENDDO 7687 ENDDO 7688 ENDDO 7689 7690 which can be beneficial when M is larger than the caches, 7691 because the innermost loop will iterate over a smaller amount 7692 of data that can be kept in the caches. 7693 * A new register allocator has replaced the old one. It is called 7694 integrated register allocator (IRA) because coalescing, register 7695 live range splitting, and hard register preferencing are done 7696 on-the-fly during coloring. It also has better integration with the 7697 reload pass. IRA is a regional register allocator which uses modern 7698 Chaitin-Briggs coloring instead of Chow's priority coloring used in 7699 the old register allocator. More info about IRA internals and 7700 options can be found in the GCC manuals. 7701 * A new instruction scheduler and software pipeliner, based on the 7702 selective scheduling approach, has been added. The new pass 7703 performs instruction unification, register renaming, substitution 7704 through register copies, and speculation during scheduling. The 7705 software pipeliner is able to pipeline non-countable loops. The new 7706 pass is targeted at scheduling-eager in-order platforms. In GCC 4.4 7707 it is available for the Intel Itanium platform working by default 7708 as the second scheduling pass (after register allocation) at the 7709 -O3 optimization level. 7710 * When using -fprofile-generate with a multi-threaded program, the 7711 profile counts may be slightly wrong due to race conditions. The 7712 new -fprofile-correction option directs the compiler to apply 7713 heuristics to smooth out the inconsistencies. By default the 7714 compiler will give an error message when it finds an inconsistent 7715 profile. 7716 * The new -fprofile-dir=PATH option permits setting the directory 7717 where profile data files are stored when using -fprofile-generate 7718 and friends, and the directory used when reading profile data files 7719 using -fprofile-use and friends. 7720 7721 New warning options 7722 7723 * The new -Wframe-larger-than=NUMBER option directs GCC to emit a 7724 warning if any stack frame is larger than NUMBER bytes. This may be 7725 used to help ensure that code fits within a limited amount of stack 7726 space. 7727 * The command-line option -Wlarger-than-N is now written as 7728 -Wlarger-than=N and the old form is deprecated. 7729 * The new -Wno-mudflap option disables warnings about constructs 7730 which can not be instrumented when using -fmudflap. 7731 7732 New Languages and Language specific improvements 7733 7734 * Version 3.0 of the OpenMP specification is now supported for the C, 7735 C++, and Fortran compilers. 7736 * New character data types, per [5]TR 19769: New character types in 7737 C, are now supported for the C compiler in -std=gnu99 mode, as 7738 __CHAR16_TYPE__ and __CHAR32_TYPE__, and for the C++ compiler in 7739 -std=c++0x and -std=gnu++0x modes, as char16_t and char32_t too. 7740 7741 C family 7742 7743 * A new optimize attribute was added to allow programmers to change 7744 the optimization level and particular optimization options for an 7745 individual function. You can also change the optimization options 7746 via the GCC optimize pragma for functions defined after the pragma. 7747 The GCC push_options pragma and the GCC pop_options pragma allow 7748 you temporarily save and restore the options used. The GCC 7749 reset_options pragma restores the options to what was specified on 7750 the command line. 7751 * Uninitialized warnings do not require enabling optimization 7752 anymore, that is, -Wuninitialized can be used together with -O0. 7753 Nonetheless, the warnings given by -Wuninitialized will probably be 7754 more accurate if optimization is enabled. 7755 * -Wparentheses now warns about expressions such as (!x | y) and (!x 7756 & y). Using explicit parentheses, such as in ((!x) | y), silences 7757 this warning. 7758 * -Wsequence-point now warns within if, while,do while and for 7759 conditions, and within for begin/end expressions. 7760 * A new option -dU is available to dump definitions of preprocessor 7761 macros that are tested or expanded. 7762 7763 C++ 7764 7765 * [6]Improved experimental support for the upcoming ISO C++ standard, 7766 C++0x. Including support for auto, inline namespaces, generalized 7767 initializer lists, defaulted and deleted functions, new character 7768 types, and scoped enums. 7769 * Those errors that may be downgraded to warnings to build legacy 7770 code now mention -fpermissive when -fdiagnostics-show-option is 7771 enabled. 7772 * -Wconversion now warns if the result of a static_cast to enumeral 7773 type is unspecified because the value is outside the range of the 7774 enumeral type. 7775 * -Wuninitialized now warns if a non-static reference or non-static 7776 const member appears in a class without constructors. 7777 * G++ now properly implements value-initialization, so objects with 7778 an initializer of () and an implicitly defined default constructor 7779 will be zero-initialized before the default constructor is called. 7780 7781 Runtime Library (libstdc++) 7782 7783 * Improved experimental support for the upcoming ISO C++ standard, 7784 C++0x, including: 7785 + Support for <chrono>, <condition_variable>, <cstdatomic>, 7786 <forward_list>, <initializer_list>, <mutex>, <ratio>, 7787 <system_error>, and <thread>. 7788 + unique_ptr, <algorithm> additions, exception propagation, and 7789 support for the new character types in <string> and <limits>. 7790 + Existing facilities now exploit initializer lists, defaulted 7791 and deleted functions, and the newly implemented core C++0x 7792 features. 7793 + Some standard containers are more efficient together with 7794 stateful allocators, i.e., no allocator is constructed on the 7795 fly at element construction time. 7796 * Experimental support for non-standard pointer types in containers. 7797 * The long standing libstdc++/30928 has been fixed for targets 7798 running glibc 2.10 or later. 7799 * As usual, many small and larger bug fixes, in particular quite a 7800 few corner cases in <locale>. 7801 7802 Fortran 7803 7804 * GNU Fortran now employs libcpp directly instead of using cc1 as an 7805 external preprocessor. The [7]-cpp option was added to allow manual 7806 invocation of the preprocessor without relying on filename 7807 extensions. 7808 * The [8]-Warray-temporaries option warns about array temporaries 7809 generated by the compiler, as an aid to optimization. 7810 * The [9]-fcheck-array-temporaries option has been added, printing a 7811 notification at run time, when an array temporary had to be created 7812 for an function argument. Contrary to -Warray-temporaries the 7813 warning is only printed if the array is noncontiguous. 7814 * Improved generation of DWARF debugging symbols 7815 * If using an intrinsic not part of the selected standard (via -std= 7816 and -fall-intrinsics) gfortran will now treat it as if this 7817 procedure were declared EXTERNAL and try to link to a user-supplied 7818 procedure. -Wintrinsics-std will warn whenever this happens. The 7819 now-useless option -Wnonstd-intrinsic was removed. 7820 * The flag -falign-commons has been added to control the alignment of 7821 variables in COMMON blocks, which is enabled by default in line 7822 with previous GCC version. Using -fno-align-commons one can force 7823 commons to be contiguous in memory as required by the Fortran 7824 standard, however, this slows down the memory access. The option 7825 -Walign-commons, which is enabled by default, warns when padding 7826 bytes were added for alignment. The proper solution is to sort the 7827 common objects by decreasing storage size, which avoids the 7828 alignment problems. 7829 * Fortran 2003 support has been extended: 7830 + Wide characters (ISO 10646, UCS-4, kind=4) and UTF-8 I/O is 7831 now supported (except internal reads from/writes to wide 7832 strings). [10]-fbackslash now supports also \unnnn and 7833 \Unnnnnnnn to enter Unicode characters. 7834 + Asynchronous I/O (implemented as synchronous I/O) and the 7835 decimal=, size=, sign=, pad=, blank=, and delim= specifiers 7836 are now supported in I/O statements. 7837 + Support for Fortran 2003 structure constructors and for array 7838 constructor with typespec has been added. 7839 + Procedure Pointers (but not yet as component in derived types 7840 and as function results) are now supported. 7841 + Abstract types, type extension, and type-bound procedures 7842 (both PROCEDURE and GENERIC but not as operators). Note: As 7843 CLASS/polymorphyic types are not implemented, type-bound 7844 procedures with PASS accept as non-standard extension TYPE 7845 arguments. 7846 * Fortran 2008 support has been added: 7847 + The -std=f2008 option and support for the file extensions 7848 .f2008 and .F2008 has been added. 7849 + The g0 format descriptor is now supported. 7850 + The Fortran 2008 mathematical intrinsics ASINH, ACOSH, ATANH, 7851 ERF, ERFC, GAMMA, LOG_GAMMA, BESSEL_*, HYPOT, and ERFC_SCALED 7852 are now available (some of them existed as GNU extension 7853 before). Note: The hyperbolic functions are not yet supporting 7854 complex arguments and the three- argument version of BESSEL_*N 7855 is not available. 7856 + The bit intrinsics LEADZ and TRAILZ have been added. 7857 7858 Java (GCJ) 7859 7860 Ada 7861 7862 * The Ada runtime now supports multilibs on many platforms including 7863 x86_64, SPARC and PowerPC. Their build is enabled by default. 7864 7865 New Targets and Target Specific Improvements 7866 7867 ARM 7868 7869 * GCC now supports optimizing for the Cortex-A9, Cortex-R4 and 7870 Cortex-R4F processors and has many other improvements to 7871 optimization for ARM processors. 7872 * GCC now supports the VFPv3 variant with 16 double-precision 7873 registers with -mfpu=vfpv3-d16. The option -mfpu=vfp3 has been 7874 renamed to -mfpu=vfpv3. 7875 * GCC now supports the -mfix-cortex-m3-ldrd option to work around an 7876 erratum on Cortex-M3 processors. 7877 * GCC now supports the __sync_* atomic operations for ARM EABI 7878 GNU/Linux. 7879 * The section anchors optimization is now enabled by default when 7880 optimizing for ARM. 7881 * GCC now uses a new EABI-compatible profiling interface for EABI 7882 targets. This requires a function __gnu_mcount_nc, which is 7883 provided by GNU libc versions 2.8 and later. 7884 7885 AVR 7886 7887 * The -mno-tablejump option has been deprecated because it has the 7888 same effect as the -fno-jump-tables option. 7889 * Added support for these new AVR devices: 7890 + ATA6289 7891 + ATtiny13A 7892 + ATtiny87 7893 + ATtiny167 7894 + ATtiny327 7895 + ATmega8C1 7896 + ATmega16C1 7897 + ATmega32C1 7898 + ATmega8M1 7899 + ATmega16M1 7900 + ATmega32M1 7901 + ATmega32U4 7902 + ATmega16HVB 7903 + ATmega4HVD 7904 + ATmega8HVD 7905 + ATmega64C1 7906 + ATmega64M1 7907 + ATmega16U4 7908 + ATmega32U6 7909 + ATmega128RFA1 7910 + AT90PWM81 7911 + AT90SCR100 7912 + M3000F 7913 + M3000S 7914 + M3001B 7915 7916 IA-32/x86-64 7917 7918 * Support for Intel AES built-in functions and code generation is 7919 available via -maes. 7920 * Support for Intel PCLMUL built-in function and code generation is 7921 available via -mpclmul. 7922 * Support for Intel AVX built-in functions and code generation is 7923 available via -mavx. 7924 * Automatically align the stack for local variables with alignment 7925 requirement. 7926 * GCC can now utilize the SVML library for vectorizing calls to a set 7927 of C99 functions if -mveclibabi=svml is specified and you link to 7928 an SVML ABI compatible library. 7929 * On x86-64, the ABI has been changed in the following cases to 7930 conform to the x86-64 ABI: 7931 + Passing/returning structures with flexible array member: 7932 struct foo 7933 { 7934 int i; 7935 int flex[]; 7936 }; 7937 + Passing/returning structures with complex float member: 7938 struct foo 7939 { 7940 int i; 7941 __complex__ float f; 7942 }; 7943 + Passing/returning unions with long double member: 7944 union foo 7945 { 7946 int x; 7947 long double ld; 7948 }; 7949 Code built with previous versions of GCC that uses any of these is 7950 not compatible with code built with GCC 4.4.0 or later. 7951 * A new target attribute was added to allow programmers to change the 7952 target options like -msse2 or -march=k8 for an individual function. 7953 You can also change the target options via the GCC target pragma 7954 for functions defined after the pragma. 7955 * GCC can now be configured with options --with-arch-32, 7956 --with-arch-64, --with-cpu-32, --with-cpu-64, --with-tune-32 and 7957 --with-tune-64 to control the default optimization separately for 7958 32-bit and 64-bit modes. 7959 7960 IA-32/IA64 7961 7962 * Support for __float128 (TFmode) IEEE quad type and corresponding 7963 TCmode IEEE complex quad type is available via the soft-fp library 7964 on IA-32/IA64 targets. This includes basic arithmetic operations 7965 (addition, subtraction, negation, multiplication and division) on 7966 __float128 real and TCmode complex values, the full set of IEEE 7967 comparisons between __float128 values, conversions to and from 7968 float, double and long double floating point types, as well as 7969 conversions to and from signed or unsigned integer, signed or 7970 unsigned long integer and signed or unsigned quad (TImode, IA64 7971 only) integer types. Additionally, all operations generate the full 7972 set of IEEE exceptions and support the full set of IEEE rounding 7973 modes. 7974 7975 M68K/ColdFire 7976 7977 * GCC now supports instruction scheduling for ColdFire V1, V3 and V4 7978 processors. (Scheduling support for ColdFire V2 processors was 7979 added in GCC 4.3.) 7980 * GCC now supports the -mxgot option to support programs requiring 7981 many GOT entries on ColdFire. 7982 * The m68k-*-linux-gnu target now builds multilibs by default. 7983 7984 MIPS 7985 7986 * MIPS Technologies have extended the original MIPS SVR4 ABI to 7987 include support for procedure linkage tables (PLTs) and copy 7988 relocations. These extensions allow GNU/Linux executables to use a 7989 significantly more efficient code model than the one defined by the 7990 original ABI. 7991 GCC support for this code model is available via a new command-line 7992 option, -mplt. There is also a new configure-time option, 7993 --with-mips-plt, to make -mplt the default. 7994 The new code model requires support from the assembler, the linker, 7995 and the runtime C library. This support is available in binutils 7996 2.19 and GLIBC 2.9. 7997 * GCC can now generate MIPS16 code for 32-bit GNU/Linux executables 7998 and 32-bit GNU/Linux shared libraries. This feature requires GNU 7999 binutils 2.19 or above. 8000 * Support for RMI's XLR processor is now available through the 8001 -march=xlr and -mtune=xlr options. 8002 * 64-bit targets can now perform 128-bit multiplications inline, 8003 instead of relying on a libgcc function. 8004 * Native GNU/Linux toolchains now support -march=native and 8005 -mtune=native, which select the host processor. 8006 * GCC now supports the R10K, R12K, R14K and R16K processors. The 8007 canonical -march= and -mtune= names for these processors are 8008 r10000, r12000, r14000 and r16000 respectively. 8009 * GCC can now work around the side effects of speculative execution 8010 on R10K processors. Please see the documentation of the 8011 -mr10k-cache-barrier option for details. 8012 * Support for the MIPS64 Release 2 instruction set has been added. 8013 The option -march=mips64r2 enables generation of these 8014 instructions. 8015 * GCC now supports Cavium Networks' Octeon processor. This support is 8016 available through the -march=octeon and -mtune=octeon options. 8017 * GCC now supports STMicroelectronics' Loongson 2E/2F processors. The 8018 canonical -march= and -mtune= names for these processors are 8019 loongson2e and loongson2f. 8020 8021 picochip 8022 8023 Picochip is a 16-bit processor. A typical picoChip contains over 250 8024 small cores, each with small amounts of memory. There are three 8025 processor variants (STAN, MEM and CTRL) with different instruction sets 8026 and memory configurations and they can be chosen using the -mae option. 8027 8028 This port is intended to be a "C" only port. 8029 8030 Power Architecture and PowerPC 8031 8032 * GCC now supports the e300c2, e300c3 and e500mc processors. 8033 * GCC now supports Xilinx processors with a single-precision FPU. 8034 * Decimal floating point is now supported for e500 processors. 8035 8036 S/390, zSeries and System z9/z10 8037 8038 * Support for the IBM System z10 EC/BC processor has been added. When 8039 using the -march=z10 option, the compiler will generate code making 8040 use of instructions provided by the General-Instruction-Extension 8041 Facility and the Execute-Extension Facility. 8042 8043 VxWorks 8044 8045 * GCC now supports the thread-local storage mechanism used on 8046 VxWorks. 8047 8048 Xtensa 8049 8050 * GCC now supports thread-local storage (TLS) for Xtensa processor 8051 configurations that include the Thread Pointer option. TLS also 8052 requires support from the assembler and linker; this support is 8053 provided in the GNU binutils beginning with version 2.19. 8054 8055 Documentation improvements 8056 8057 Other significant improvements 8058 8059 GCC 4.4.1 8060 8061 This is the [11]list of problem reports (PRs) from GCC's bug tracking 8062 system that are known to be fixed in the 4.4.1 release. This list might 8063 not be complete (that is, it is possible that some PRs that have been 8064 fixed are not listed here). 8065 8066 GCC 4.4.2 8067 8068 This is the [12]list of problem reports (PRs) from GCC's bug tracking 8069 system that are known to be fixed in the 4.4.2 release. This list might 8070 not be complete (that is, it is possible that some PRs that have been 8071 fixed are not listed here). 8072 8073 GCC 4.4.3 8074 8075 This is the [13]list of problem reports (PRs) from GCC's bug tracking 8076 system that are known to be fixed in the 4.4.3 release. This list might 8077 not be complete (that is, it is possible that some PRs that have been 8078 fixed are not listed here). 8079 8080 GCC 4.4.4 8081 8082 This is the [14]list of problem reports (PRs) from GCC's bug tracking 8083 system that are known to be fixed in the 4.4.4 release. This list might 8084 not be complete (that is, it is possible that some PRs that have been 8085 fixed are not listed here). 8086 8087 GCC 4.4.5 8088 8089 This is the [15]list of problem reports (PRs) from GCC's bug tracking 8090 system that are known to be fixed in the 4.4.5 release. This list might 8091 not be complete (that is, it is possible that some PRs that have been 8092 fixed are not listed here). 8093 8094 GCC 4.4.6 8095 8096 This is the [16]list of problem reports (PRs) from GCC's bug tracking 8097 system that are known to be fixed in the 4.4.6 release. This list might 8098 not be complete (that is, it is possible that some PRs that have been 8099 fixed are not listed here). 8100 8101 GCC 4.4.7 8102 8103 This is the [17]list of problem reports (PRs) from GCC's bug tracking 8104 system that are known to be fixed in the 4.4.7 release. This list might 8105 not be complete (that is, it is possible that some PRs that have been 8106 fixed are not listed here). 8107 8108 8109 For questions related to the use of GCC, please consult these web 8110 pages and the [18]GCC manuals. If that fails, the 8111 [19]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8112 web pages and the development of GCC are welcome on our developer 8113 list at [20]gcc (a] gcc.gnu.org. All of [21]our lists have public 8114 archives. 8115 8116 Copyright (C) [22]Free Software Foundation, Inc. Verbatim copying and 8117 distribution of this entire article is permitted in any medium, 8118 provided this notice is preserved. 8119 8120 These pages are [23]maintained by the GCC team. Last modified 8121 2018-09-30[24]. 8122 8123 References 8124 8125 1. http://gcc.gnu.org/gcc-4.4/changes.html#4.4.7 8126 2. http://gcc.gnu.org/gcc-4.3/changes.html#obsoleted 8127 3. http://gcc.gnu.org/gcc-4.4/porting_to.html 8128 4. https://gcc.gnu.org/wiki/Graphite 8129 5. http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1040.pdf 8130 6. http://gcc.gnu.org/gcc-4.4/cxx0x_status.html 8131 7. https://gcc.gnu.org/onlinedocs/gfortran/Preprocessing-Options.html 8132 8. https://gcc.gnu.org/onlinedocs/gfortran/Error-and-Warning-Options.html#index-g_t_0040code_007bWarray-temporaries_007d-125 8133 9. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfcheck-array-temporaries_007d-221 8134 10. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html#index-g_t_0040code_007bbackslash_007d-34 8135 11. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.1 8136 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.2 8137 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.3 8138 14. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.4 8139 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.5 8140 16. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.6 8141 17. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.4.7 8142 18. https://gcc.gnu.org/onlinedocs/ 8143 19. mailto:gcc-help (a] gcc.gnu.org 8144 20. mailto:gcc (a] gcc.gnu.org 8145 21. https://gcc.gnu.org/lists.html 8146 22. https://www.fsf.org/ 8147 23. https://gcc.gnu.org/about.html 8148 24. http://validator.w3.org/check/referer 8149 ====================================================================== 8150 http://gcc.gnu.org/gcc-4.3/index.html 8151 8152 GCC 4.3 Release Series 8153 8154 Jun 27, 2011 8155 8156 The [1]GNU project and the GCC developers are pleased to announce the 8157 release of GCC 4.3.6. 8158 8159 This release is a bug-fix release, containing fixes for regressions in 8160 GCC 4.3.5 relative to previous releases of GCC. 8161 8162 This release series is no longer maintained. 8163 8164 Release History 8165 8166 GCC 4.3.6 8167 Jun 27, 2011 ([2]changes) 8168 8169 GCC 4.3.5 8170 May 22, 2010 ([3]changes) 8171 8172 GCC 4.3.4 8173 August 4, 2009 ([4]changes) 8174 8175 GCC 4.3.3 8176 January 24, 2009 ([5]changes) 8177 8178 GCC 4.3.2 8179 August 27, 2008 ([6]changes) 8180 8181 GCC 4.3.1 8182 June 6, 2008 ([7]changes) 8183 8184 GCC 4.3.0 8185 March 5, 2008 ([8]changes) 8186 8187 References and Acknowledgements 8188 8189 GCC used to stand for the GNU C Compiler, but since the compiler 8190 supports several other languages aside from C, it now stands for the 8191 GNU Compiler Collection. 8192 8193 A list of [9]successful builds is updated as new information becomes 8194 available. 8195 8196 The GCC developers would like to thank the numerous people that have 8197 contributed new features, improvements, bug fixes, and other changes as 8198 well as test results to GCC. This [10]amazing group of volunteers is 8199 what makes GCC successful. 8200 8201 For additional information about GCC please refer to the [11]GCC 8202 project web site or contact the [12]GCC development mailing list. 8203 8204 To obtain GCC please use [13]our mirror sites or [14]our SVN server. 8205 8206 8207 For questions related to the use of GCC, please consult these web 8208 pages and the [15]GCC manuals. If that fails, the 8209 [16]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 8210 web pages and the development of GCC are welcome on our developer 8211 list at [17]gcc (a] gcc.gnu.org. All of [18]our lists have public 8212 archives. 8213 8214 Copyright (C) [19]Free Software Foundation, Inc. Verbatim copying and 8215 distribution of this entire article is permitted in any medium, 8216 provided this notice is preserved. 8217 8218 These pages are [20]maintained by the GCC team. Last modified 8219 2018-09-30[21]. 8220 8221 References 8222 8223 1. http://www.gnu.org/ 8224 2. http://gcc.gnu.org/gcc-4.3/changes.html 8225 3. http://gcc.gnu.org/gcc-4.3/changes.html 8226 4. http://gcc.gnu.org/gcc-4.3/changes.html 8227 5. http://gcc.gnu.org/gcc-4.3/changes.html 8228 6. http://gcc.gnu.org/gcc-4.3/changes.html 8229 7. http://gcc.gnu.org/gcc-4.3/changes.html 8230 8. http://gcc.gnu.org/gcc-4.3/changes.html 8231 9. http://gcc.gnu.org/gcc-4.3/buildstat.html 8232 10. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 8233 11. http://gcc.gnu.org/index.html 8234 12. mailto:gcc (a] gcc.gnu.org 8235 13. http://gcc.gnu.org/mirrors.html 8236 14. http://gcc.gnu.org/svn.html 8237 15. https://gcc.gnu.org/onlinedocs/ 8238 16. mailto:gcc-help (a] gcc.gnu.org 8239 17. mailto:gcc (a] gcc.gnu.org 8240 18. https://gcc.gnu.org/lists.html 8241 19. https://www.fsf.org/ 8242 20. https://gcc.gnu.org/about.html 8243 21. http://validator.w3.org/check/referer 8244 ====================================================================== 8245 http://gcc.gnu.org/gcc-4.3/changes.html 8246 8247 GCC 4.3 Release Series 8248 Changes, New Features, and Fixes 8249 8250 The latest release in the 4.3 release series is [1]GCC 4.3.5. 8251 8252 Caveats 8253 8254 * GCC requires the [2]GMP and [3]MPFR libraries for building all the 8255 various front-end languages it supports. See the [4]prerequisites 8256 page for version requirements. 8257 * ColdFire targets now treat long double as having the same format as 8258 double. In earlier versions of GCC, they used the 68881 long double 8259 format instead. 8260 * The m68k-uclinux target now uses the same calling conventions as 8261 m68k-linux-gnu. You can select the original calling conventions by 8262 configuring for m68k-uclinuxoldabi instead. Note that 8263 m68k-uclinuxoldabi also retains the original 80-bit long double on 8264 ColdFire targets. 8265 * The -fforce-mem option has been removed because it has had no 8266 effect in the last few GCC releases. 8267 * The i386 -msvr3-shlib option has been removed since it is no longer 8268 used. 8269 * Fastcall for i386 has been changed not to pass aggregate arguments 8270 in registers, following Microsoft compilers. 8271 * Support for the AOF assembler has been removed from the ARM back 8272 end; this affects only the targets arm-semi-aof and armel-semi-aof, 8273 which are no longer recognized. We removed these targets without a 8274 deprecation period because we discovered that they have been 8275 unusable since GCC 4.0.0. 8276 * Support for the TMS320C3x/C4x processor (targets c4x-* and tic4x-*) 8277 has been removed. This support had been deprecated since GCC 4.0.0. 8278 * Support for a number of older systems and recently unmaintained or 8279 untested target ports of GCC has been declared obsolete in GCC 4.3. 8280 Unless there is activity to revive them, the next release of GCC 8281 will have their sources permanently removed. 8282 All GCC ports for the following processor architectures have been 8283 declared obsolete: 8284 + Morpho MT (mt-*) 8285 The following aliases for processor architectures have been 8286 declared obsolete. Users should use the indicated generic target 8287 names instead, with compile-time options such as -mcpu or 8288 configure-time options such as --with-cpu to control the 8289 configuration more precisely. 8290 + strongarm*-*-*, ep9312*-*-*, xscale*-*-* (use arm*-*-* 8291 instead). 8292 + parisc*-*-* (use hppa*-*-* instead). 8293 + m680[012]0-*-* (use m68k-*-* instead). 8294 All GCC ports for the following operating systems have been 8295 declared obsolete: 8296 + BeOS (*-*-beos*) 8297 + kaOS (*-*-kaos*) 8298 + GNU/Linux using the a.out object format (*-*-linux*aout*) 8299 + GNU/Linux using version 1 of the GNU C Library 8300 (*-*-linux*libc1*) 8301 + Solaris versions before Solaris 7 (*-*-solaris2.[0-6], 8302 *-*-solaris2.[0-6].*) 8303 + Miscellaneous System V (*-*-sysv*) 8304 + WindISS (*-*-windiss*) 8305 Also, those for some individual systems on particular architectures 8306 have been obsoleted: 8307 + UNICOS/mk on DEC Alpha (alpha*-*-unicosmk*) 8308 + CRIS with a.out object format (cris-*-aout) 8309 + BSD 4.3 on PA-RISC (hppa1.1-*-bsd*) 8310 + OSF/1 on PA-RISC (hppa1.1-*-osf*) 8311 + PRO on PA-RISC (hppa1.1-*-pro*) 8312 + Sequent PTX on IA32 (i[34567]86-sequent-ptx4*, 8313 i[34567]86-sequent-sysv4*) 8314 + SCO Open Server 5 on IA32 (i[34567]86-*-sco3.2v5*) 8315 + UWIN on IA32 (i[34567]86-*-uwin*) (support for UWIN as a host 8316 was previously [5]removed in 2001, leaving only the support 8317 for UWIN as a target now being deprecated) 8318 + ChorusOS on PowerPC (powerpc-*-chorusos*) 8319 + All VAX configurations apart from NetBSD and OpenBSD 8320 (vax-*-bsd*, vax-*-sysv*, vax-*-ultrix*) 8321 * The [6]-Wconversion option has been modified. Its purpose now is to 8322 warn for implicit conversions that may alter a value. This new 8323 behavior is available for both C and C++. Warnings about 8324 conversions between signed and unsigned integers can be disabled by 8325 using -Wno-sign-conversion. In C++, they are disabled by default 8326 unless -Wsign-conversion is explicitly requested. The old behavior 8327 of -Wconversion, that is, warn for prototypes causing a type 8328 conversion that is different from what would happen to the same 8329 argument in the absence of a prototype, has been moved to a new 8330 option -Wtraditional-conversion, which is only available for C. 8331 * The -m386, -m486, -mpentium and -mpentiumpro tuning options have 8332 been removed because they were deprecated for more than 3 GCC major 8333 releases. Use -mtune=i386, -mtune=i486, -mtune=pentium or 8334 -mtune=pentiumpro as a replacement. 8335 * The -funsafe-math-optimizations option now automatically turns on 8336 -fno-trapping-math in addition to -fno-signed-zeros, as it enables 8337 reassociation and thus may introduce or remove traps. 8338 * The -ftree-vectorize option is now on by default under -O3. In 8339 order to generate code for a SIMD extension, it has to be enabled 8340 as well: use -maltivec for PowerPC platforms and -msse/-msse2 for 8341 i?86 and x86_64. 8342 * More information on porting to GCC 4.3 from previous versions of 8343 GCC can be found in the [7]porting guide for this release. 8344 8345 General Optimizer Improvements 8346 8347 * The GCC middle-end has been integrated with the MPFR library. This 8348 allows GCC to evaluate and replace at compile-time calls to 8349 built-in math functions having constant arguments with their 8350 mathematically equivalent results. In making use of MPFR, GCC can 8351 generate correct results regardless of the math library 8352 implementation or floating point precision of the host platform. 8353 This also allows GCC to generate identical results regardless of 8354 whether one compiles in native or cross-compile configurations to a 8355 particular target. The following built-in functions take advantage 8356 of this new capability: acos, acosh, asin, asinh, atan2, atan, 8357 atanh, cbrt, cos, cosh, drem, erf, erfc, exp10, exp2, exp, expm1, 8358 fdim, fma, fmax, fmin, gamma_r, hypot, j0, j1, jn, lgamma_r, log10, 8359 log1p, log2, log, pow10, pow, remainder, remquo, sin, sincos, sinh, 8360 tan, tanh, tgamma, y0, y1 and yn. The float and long double 8361 variants of these functions (e.g. sinf and sinl) are also handled. 8362 The sqrt and cabs functions with constant arguments were already 8363 optimized in prior GCC releases. Now they also use MPFR. 8364 * A new forward propagation pass on RTL was added. The new pass 8365 replaces several slower transformations, resulting in compile-time 8366 improvements as well as better code generation in some cases. 8367 * A new command-line switch -frecord-gcc-switches has been added to 8368 GCC, although it is only enabled for some targets. The switch 8369 causes the command line that was used to invoke the compiler to be 8370 recorded into the object file that is being created. The exact 8371 format of this recording is target and binary file format 8372 dependent, but it usually takes the form of a note section 8373 containing ASCII text. The switch is related to the -fverbose-asm 8374 switch, but that one only records the information in the assembler 8375 output file as comments, so the information never reaches the 8376 object file. 8377 * The inliner heuristic is now aware of stack frame consumption. New 8378 command-line parameters --param large-stack-frame and --param 8379 large-stack-frame-growth can be used to limit stack frame size 8380 growth caused by inlining. 8381 * During feedback directed optimizations, the expected block size the 8382 memcpy, memset and bzero functions operate on is discovered and for 8383 cases of commonly used small sizes, specialized inline code is 8384 generated. 8385 * __builtin_expect no longer requires its argument to be a compile 8386 time constant. 8387 * Interprocedural optimization was reorganized to work on functions 8388 in SSA form. This enables more precise and cheaper dataflow 8389 analysis and makes writing interprocedural optimizations easier. 8390 The following improvements have been implemented on top of this 8391 framework: 8392 + Pre-inline optimization: Selected local optimization passes 8393 are run before the inliner (and other interprocedural passes) 8394 are executed. This significantly improves the accuracy of code 8395 growth estimates used by the inliner and reduces the overall 8396 memory footprint for large compilation units. 8397 + Early inlining (a simple bottom-up inliner pass inlining only 8398 functions whose body is smaller than the expected call 8399 overhead) is now executed with the early optimization passes, 8400 thus inlining already optimized function bodies into an 8401 unoptimized function that is subsequently optimized by early 8402 optimizers. This enables the compiler to quickly eliminate 8403 abstraction penalty in C++ programs. 8404 + Interprocedural constant propagation now operate on SSA form 8405 increasing accuracy of the analysis. 8406 * A new internal representation for GIMPLE statements has been 8407 contributed, resulting in compile-time memory savings. 8408 * The vectorizer was enhanced to support vectorization of outer 8409 loops, intra-iteration parallelism (loop-aware SLP), vectorization 8410 of strided accesses and loops with multiple data-types. Run-time 8411 dependency testing using loop versioning was added. The cost model, 8412 turned on by -fvect-cost-model, was developed. 8413 8414 New Languages and Language specific improvements 8415 8416 * We have added new command-line options 8417 -finstrument-functions-exclude-function-list and 8418 -finstrument-functions-exclude-file-list. They provide more control 8419 over which functions are annotated by the -finstrument-functions 8420 option. 8421 8422 C family 8423 8424 * Implicit conversions between generic vector types are now only 8425 permitted when the two vectors in question have the same number of 8426 elements and compatible element types. (Note that the restriction 8427 involves compatible element types, not implicitly-convertible 8428 element types: thus, a vector type with element type int may not be 8429 implicitly converted to a vector type with element type unsigned 8430 int.) This restriction, which is in line with specifications for 8431 SIMD architectures such as AltiVec, may be relaxed using the flag 8432 -flax-vector-conversions. This flag is intended only as a 8433 compatibility measure and should not be used for new code. 8434 * -Warray-bounds has been added and is now enabled by default for 8435 -Wall . It produces warnings for array subscripts that can be 8436 determined at compile time to be always out of bounds. 8437 -Wno-array-bounds will disable the warning. 8438 * The constructor and destructor function attributes now accept 8439 optional priority arguments which control the order in which the 8440 constructor and destructor functions are run. 8441 * New [8]command-line options -Wtype-limits, -Wold-style-declaration, 8442 -Wmissing-parameter-type, -Wempty-body, -Wclobbered and 8443 -Wignored-qualifiers have been added for finer control of the 8444 diverse warnings enabled by -Wextra. 8445 * A new function attribute alloc_size has been added to mark up 8446 malloc style functions. For constant sized allocations this can be 8447 used to find out the size of the returned pointer using the 8448 __builtin_object_size() function for buffer overflow checking and 8449 similar. This supplements the already built-in malloc and calloc 8450 constant size handling. 8451 * Integer constants written in binary are now supported as a GCC 8452 extension. They consist of a prefix 0b or 0B, followed by a 8453 sequence of 0 and 1 digits. 8454 * A new predefined macro __COUNTER__ has been added. It expands to 8455 sequential integral values starting from 0. In conjunction with the 8456 ## operator, this provides a convenient means to generate unique 8457 identifiers. 8458 * A new command-line option -fdirectives-only has been added. It 8459 enables a special preprocessing mode which improves the performance 8460 of applications like distcc and ccache. 8461 * Fixed-point data types and operators have been added. They are 8462 based on Chapter 4 of the Embedded-C specification (n1169.pdf). 8463 Currently, only MIPS targets are supported. 8464 * Decimal floating-point arithmetic based on draft ISO/IEC TR 24732, 8465 N1241, is now supported as a GCC extension to C for targets 8466 i[34567]86-*-linux-gnu, powerpc*-*-linux-gnu, s390*-ibm-linux-gnu, 8467 and x86_64-*-linux-gnu. The feature introduces new data types 8468 _Decimal32, _Decimal64, and _Decimal128 with constant suffixes DF, 8469 DD, and DL. 8470 8471 C++ 8472 8473 * [9]Experimental support for the upcoming ISO C++ standard, C++0x. 8474 * -Wc++0x-compat has been added and is now enabled by default for 8475 -Wall. It produces warnings for constructs whose meaning differs 8476 between ISO C++ 1998 and C++0x. 8477 * The -Wparentheses option now works for C++ as it does for C. It 8478 warns if parentheses are omitted when operators with confusing 8479 precedence are nested. It also warns about ambiguous else 8480 statements. Since -Wparentheses is enabled by -Wall, this may cause 8481 additional warnings with existing C++ code which uses -Wall. These 8482 new warnings may be disabled by using -Wall -Wno-parentheses. 8483 * The -Wmissing-declarations now works for C++ as it does for C. 8484 * The -fvisibility-ms-compat flag was added, to make it easier to 8485 port larger projects using shared libraries from Microsoft's Visual 8486 Studio to ELF and Mach-O systems. 8487 * C++ attribute handling has been overhauled for template arguments 8488 (ie dependent types). In particular, __attribute__((aligned(T))); 8489 works for C++ types. 8490 8491 Runtime Library (libstdc++) 8492 8493 * [10]Experimental support for the upcoming ISO C++ standard, C++0x. 8494 * Support for TR1 mathematical special functions and regular 8495 expressions. 8496 * Default what implementations give more elaborate exception strings 8497 for bad_cast, bad_typeid, bad_exception, and bad_alloc. 8498 * Header dependencies have been streamlined, reducing unnecessary 8499 includes and pre-processed bloat. 8500 * Variadic template implementations of items in <tuple> and 8501 <functional>. 8502 * An experimental [11]parallel mode has been added. This is a 8503 parallel implementation of many C++ Standard library algorithms, 8504 like std::accumulate, std::for_each, std::transform, or std::sort, 8505 to give but four examples. These algorithms can be substituted for 8506 the normal (sequential) libstdc++ algorithms on a piecemeal basis, 8507 or all existing algorithms can be transformed via the 8508 -D_GLIBCXX_PARALLEL macro. 8509 * Debug mode versions of classes in <unordered_set> and 8510 <unordered_map>. 8511 * Formal deprecation of <ext/hash_set> and <ext/hash_map>, which are 8512 now <backward/hash_set> and <backward/hash_map>. This code: 8513 #include <ext/hash_set> 8514 __gnu_cxx::hash_set<int> s; 8515 8516 Can be transformed (in order of preference) to: 8517 #include <tr1/unordered_set> 8518 std::tr1::unordered_set<int> s; 8519 8520 or 8521 #include <backward/hash_set> 8522 __gnu_cxx::hash_set<int> s; 8523 8524 Similar transformations apply to __gnu_cxx::hash_map, 8525 __gnu_cxx::hash_multimap, __gnu_cxx::hash_set, 8526 __gnu_cxx::hash_multiset. 8527 8528 Fortran 8529 8530 * Due to the fact that the GMP and MPFR libraries are required for 8531 all languages, Fortran is no longer special in this regard and is 8532 available by default. 8533 * The [12]-fexternal-blas option has been added, which generates 8534 calls to BLAS routines for intrinsic matrix operations such as 8535 matmul rather than using the built-in algorithms. 8536 * Support to give a backtrace (compiler flag -fbacktrace or 8537 environment variable GFORTRAN_ERROR_BACKTRACE; on glibc systems 8538 only) or a core dump (-fdump-core, GFORTRAN_ERROR_DUMPCORE) when a 8539 run-time error occured. 8540 * GNU Fortran now defines __GFORTRAN__ when it runs the C 8541 preprocessor (CPP). 8542 * The [13]-finit-local-zero, -finit-real, -finit-integer, 8543 -finit-character, and -finit-logical options have been added, which 8544 can be used to initialize local variables. 8545 * The intrinsic procedures [14]GAMMA and [15]LGAMMA have been added, 8546 which calculate the Gamma function and its logarithm. Use EXTERNAL 8547 gamma if you want to use your own gamma function. 8548 * GNU Fortran now regards the backslash character as literal (as 8549 required by the Fortran 2003 standard); using [16]-fbackslash GNU 8550 Fortran interprets backslashes as C-style escape characters. 8551 * The [17]interpretation of binary, octal and hexadecimal (BOZ) 8552 literal constants has been changed. Before they were always 8553 interpreted as integer; now they are bit-wise transferred as 8554 argument of INT, REAL, DBLE and CMPLX as required by the Fortran 8555 2003 standard, and for real and complex variables in DATA 8556 statements or when directly assigned to real and complex variables. 8557 Everywhere else and especially in expressions they are still 8558 regarded as integer constants. 8559 * Fortran 2003 support has been extended: 8560 + Intrinsic statements IMPORT, PROTECTED, VALUE and VOLATILE 8561 + Pointer intent 8562 + Intrinsic module ISO_ENV_FORTRAN 8563 + Interoperability with C (ISO C Bindings) 8564 + ABSTRACT INTERFACES and PROCEDURE statements (without POINTER 8565 attribute) 8566 + Fortran 2003 BOZ 8567 8568 Java (GCJ) 8569 8570 * GCJ now uses the Eclipse Java compiler for its Java parsing needs. 8571 This enables the use of all 1.5 language features, and fixes most 8572 existing front end bugs. 8573 * libgcj now supports all 1.5 language features which require runtime 8574 support: foreach, enum, annotations, generics, and auto-boxing. 8575 * We've made many changes to the tools shipped with gcj. 8576 + The old jv-scan tool has been removed. This tool never really 8577 worked properly. There is no replacement. 8578 + gcjh has been rewritten. Some of its more obscure options no 8579 longer work, but are still recognized in an attempt at 8580 compatibility. gjavah is a new program with similar 8581 functionality but different command-line options. 8582 + grmic and grmiregistry have been rewritten. grmid has been 8583 added. 8584 + gjar replaces the old fastjar. 8585 + gjarsigner (used for signing jars), gkeytool (used for key 8586 management), gorbd (for CORBA), gserialver (computes 8587 serialization UIDs), and gtnameserv (also for CORBA) are now 8588 installed. 8589 * The ability to dump the contents of the java run time heap to a 8590 file for off-line analysis has been added. The heap dumps may be 8591 analyzed with the new gc-analyze tool. They may be generated on 8592 out-of-memory conditions or on demand and are controlled by the new 8593 run time class gnu.gcj.util.GCInfo. 8594 * java.util.TimeZone can now read files from /usr/share/zoneinfo to 8595 provide correct, updated, timezone information. This means that 8596 packagers no longer have to update libgcj when a time zone change 8597 is published. 8598 8599 New Targets and Target Specific Improvements 8600 8601 IA-32/x86-64 8602 8603 * Tuning for Intel Core 2 processors is available via -mtune=core2 8604 and -march=core2. 8605 * Tuning for AMD Geode processors is available via -mtune=geode and 8606 -march=geode. 8607 * Code generation of block move (memcpy) and block set (memset) was 8608 rewritten. GCC can now pick the best algorithm (loop, unrolled 8609 loop, instruction with rep prefix or a library call) based on the 8610 size of the block being copied and the CPU being optimized for. A 8611 new option -minline-stringops-dynamically has been added. With this 8612 option string operations of unknown size are expanded such that 8613 small blocks are copied by in-line code, while for large blocks a 8614 library call is used. This results in faster code than 8615 -minline-all-stringops when the library implementation is capable 8616 of using cache hierarchy hints. The heuristic choosing the 8617 particular algorithm can be overwritten via -mstringop-strategy. 8618 Newly also memset of values different from 0 is inlined. 8619 * GCC no longer places the cld instruction before string operations. 8620 Both i386 and x86-64 ABI documents mandate the direction flag to be 8621 clear at the entry of a function. It is now invalid to set the flag 8622 in asm statement without reseting it afterward. 8623 * Support for SSSE3 built-in functions and code generation are 8624 available via -mssse3. 8625 * Support for SSE4.1 built-in functions and code generation are 8626 available via -msse4.1. 8627 * Support for SSE4.2 built-in functions and code generation are 8628 available via -msse4.2. 8629 * Both SSE4.1 and SSE4.2 support can be enabled via -msse4. 8630 * A new set of options -mpc32, -mpc64 and -mpc80 have been added to 8631 allow explicit control of x87 floating point precision. 8632 * Support for __float128 (TFmode) IEEE quad type and corresponding 8633 TCmode IEEE complex quad type is available via the soft-fp library 8634 on x86_64 targets. This includes basic arithmetic operations 8635 (addition, subtraction, negation, multiplication and division) on 8636 __float128 real and TCmode complex values, the full set of IEEE 8637 comparisons between __float128 values, conversions to and from 8638 float, double and long double floating point types, as well as 8639 conversions to and from signed or unsigned integer, signed or 8640 unsigned long integer and signed or unsigned quad (TImode) integer 8641 types. Additionally, all operations generate the full set of IEEE 8642 exceptions and support the full set of IEEE rounding modes. 8643 * GCC can now utilize the ACML library for vectorizing calls to a set 8644 of C99 functions on x86_64 if -mveclibabi=acml is specified and you 8645 link to an ACML ABI compatible library. 8646 8647 ARM 8648 8649 * Compiler and Library support for Thumb-2 and the ARMv7 architecture 8650 has been added. 8651 8652 CRIS 8653 8654 New features 8655 8656 * Compiler and Library support for the CRIS v32 architecture, as 8657 found in Axis Communications ETRAX FS and ARTPEC-3 chips, has been 8658 added. 8659 8660 Configuration changes 8661 8662 * The cris-*-elf target now includes support for CRIS v32, including 8663 libraries, through the -march=v32 option. 8664 * A new crisv32-*-elf target defaults to generate code for CRIS v32. 8665 * A new crisv32-*-linux* target defaults to generate code for CRIS 8666 v32. 8667 * The cris-*-aout target has been obsoleted. 8668 8669 Improved support for built-in functions 8670 8671 * GCC can now use the lz and swapwbr instructions to implement the 8672 __builtin_clz, __builtin_ctz and __builtin_ffs family of functions. 8673 * __builtin_bswap32 is now implemented using the swapwb instruction, 8674 when available. 8675 8676 m68k and ColdFire 8677 8678 New features 8679 8680 * Support for several new ColdFire processors has been added. You can 8681 generate code for them using the new -mcpu option. 8682 * All targets now support ColdFire processors. 8683 * m68k-uclinux targets have improved support for C++ constructors and 8684 destructors, and for shared libraries. 8685 * It is now possible to set breakpoints on the first or last line of 8686 a function, even if there are no statements on that line. 8687 8688 Optimizations 8689 8690 * Support for sibling calls has been added. 8691 * More use is now made of the ColdFire mov3q instruction. 8692 * __builtin_clz is now implemented using the ff1 ColdFire 8693 instruction, when available. 8694 * GCC now honors the -m68010 option. 68010 code now uses clr rather 8695 than move to zero volatile memory. 8696 * 68020 targets and above can now use symbol(index.size*scale) 8697 addresses for indexed array accesses. Earlier compilers would 8698 always load the symbol into a base register first. 8699 8700 Configuration changes 8701 8702 * All m68k and ColdFire targets now allow the default processor to be 8703 set at configure time using --with-cpu. 8704 * A --with-arch configuration option has been added. This option 8705 allows you to restrict a target to ColdFire or non-ColdFire 8706 processors. 8707 8708 Preprocessor macros 8709 8710 * An __mcfv*__ macro is now defined for all ColdFire targets. 8711 (Earlier versions of GCC only defined __mcfv4e__.) 8712 * __mcf_cpu_*, __mcf_family_* and __mcffpu__ macros have been added. 8713 * All targets now define __mc68010 and __mc68010__ when generating 8714 68010 code. 8715 8716 Command-line changes 8717 8718 * New command-line options -march, -mcpu, -mtune and -mhard-float 8719 have been added. These options apply to both m68k and ColdFire 8720 targets. 8721 * -mno-short, -mno-bitfield and -mno-rtd are now accepted as negative 8722 versions of -mshort, etc. 8723 * -fforce-addr has been removed. It is now ignored by the compiler. 8724 8725 Other improvements 8726 8727 * ColdFire targets now try to maintain a 4-byte-aligned stack where 8728 possible. 8729 * m68k-uclinux targets now try to avoid situations that lead to the 8730 load-time error: BINFMT_FLAT: reloc outside program. 8731 8732 MIPS 8733 8734 Changes to existing configurations 8735 8736 * libffi and libjava now support all three GNU/Linux ABIs: o32, n32 8737 and n64. Every GNU/Linux configuration now builds these libraries 8738 by default. 8739 * GNU/Linux configurations now generate -mno-shared code unless 8740 overridden by -fpic, -fPIC, -fpie or -fPIE. 8741 * mipsisa32*-linux-gnu configurations now generate hard-float code by 8742 default, just like other mipsisa32* and mips*-linux-gnu 8743 configurations. You can build a soft-float version of any 8744 mips*-linux-gnu configuration by passing --with-float=soft to 8745 configure. 8746 * mips-wrs-vxworks now supports run-time processes (RTPs). 8747 8748 Changes to existing command-line options 8749 8750 * The -march and -mtune options no longer accept 24k as a processor 8751 name. Please use 24kc, 24kf2_1 or 24kf1_1 instead. 8752 * The -march and -mtune options now accept 24kf2_1, 24kef2_1 and 8753 34kf2_1 as synonyms for 24kf, 24kef and 34kf respectively. The 8754 options also accept 24kf1_1, 24kef1_1 and 34kf1_1 as synonyms for 8755 24kx, 24kex and 34kx. 8756 8757 New configurations 8758 8759 GCC now supports the following configurations: 8760 * mipsisa32r2*-linux-gnu*, which generates MIPS32 revision 2 code by 8761 default. Earlier releases also recognized this configuration, but 8762 they treated it in the same way as mipsisa32*-linux-gnu*. Note that 8763 you can customize any mips*-linux-gnu* configuration to a 8764 particular ISA or processor by passing an appropriate --with-arch 8765 option to configure. 8766 * mipsisa*-sde-elf*, which provides compatibility with MIPS 8767 Technologies' SDE toolchains. The configuration uses the SDE 8768 libraries by default, but you can use it like other newlib-based 8769 ELF configurations by passing --with-newlib to configure. It is the 8770 only configuration besides mips64vr*-elf* to build MIPS16 as well 8771 as non-MIPS16 libraries. 8772 * mipsisa*-elfoabi*, which is similar to the general mipsisa*-elf* 8773 configuration, but uses the o32 and o64 ABIs instead of the 32-bit 8774 and 64-bit forms of the EABI. 8775 8776 New processors and application-specific extensions 8777 8778 * Support for the SmartMIPS ASE is available through the new 8779 -msmartmips option. 8780 * Support for revision 2 of the DSP ASE is available through the new 8781 -mdspr2 option. A new preprocessor macro called __mips_dsp_rev 8782 indicates the revision of the ASE in use. 8783 * Support for the 4KS and 74K families of processors is available 8784 through the -march and -mtune options. 8785 8786 Improved support for built-in functions 8787 8788 * GCC can now use load-linked, store-conditional and sync 8789 instructions to implement atomic built-in functions such as 8790 __sync_fetch_and_add. The memory reference must be 4 bytes wide for 8791 32-bit targets and either 4 or 8 bytes wide for 64-bit targets. 8792 * GCC can now use the clz and dclz instructions to implement the 8793 __builtin_ctz and __builtin_ffs families of functions. 8794 * There is a new __builtin___clear_cache function for flushing the 8795 instruction cache. GCC expands this function inline on MIPS32 8796 revision 2 targets, otherwise it calls the function specified by 8797 -mcache-flush-func. 8798 8799 MIPS16 improvements 8800 8801 * GCC can now compile objects that contain a mixture of MIPS16 and 8802 non-MIPS16 code. There are two new attributes, mips16 and nomips16, 8803 for specifying which mode a function should use. 8804 * A new option called -minterlink-mips16 makes non-MIPS16 code 8805 link-compatible with MIPS16 code. 8806 * After many bug fixes, the long-standing MIPS16 -mhard-float support 8807 should now work fairly reliably. 8808 * GCC can now use the MIPS16e save and restore instructions. 8809 * -fsection-anchors now works in MIPS16 mode. MIPS16 code compiled 8810 with -G0 -fsection-anchors is often smaller than code compiled with 8811 -G8. However, please note that you must usually compile all objects 8812 in your application with the same -G option; see the documentation 8813 of -G for details. 8814 * A new option called-mcode-readable specifies which instructions are 8815 allowed to load from the code segment. -mcode-readable=yes is the 8816 default and says that any instruction may load from the code 8817 segment. The other alternatives are -mcode-readable=pcrel, which 8818 says that only PC-relative MIPS16 instructions may load from the 8819 code segment, and -mcode-readable=no, which says that no 8820 instruction may do so. Please see the documentation for more 8821 details, including example uses. 8822 8823 Small-data improvements 8824 8825 There are three new options for controlling small data: 8826 * -mno-extern-sdata, which disables small-data accesses for 8827 externally-defined variables. Code compiled with -Gn 8828 -mno-extern-sdata will be link-compatible with any -G setting 8829 between -G0 and -Gn inclusive. 8830 * -mno-local-sdata, which disables the use of small-data sections for 8831 data that is not externally visible. This option can be a useful 8832 way of reducing small-data usage in less performance-critical parts 8833 of an application. 8834 * -mno-gpopt, which disables the use of the $gp register while still 8835 honoring the -G limit when placing externally-visible data. This 8836 option implies -mno-extern-sdata and -mno-local-sdata and it can be 8837 useful in situations where $gp does not necessarily hold the 8838 expected value. 8839 8840 Miscellaneous improvements 8841 8842 * There is a new option called -mbranch-cost for tweaking the 8843 perceived cost of branches. 8844 * If GCC is configured to use a version of GAS that supports the 8845 .gnu_attribute directive, it will use that directive to record 8846 certain properties of the output code. .gnu_attribute is new to GAS 8847 2.18. 8848 * There are two new function attributes, near and far, for overriding 8849 the command-line setting of -mlong-calls on a function-by-function 8850 basis. 8851 * -mfp64, which previously required a 64-bit target, now works with 8852 MIPS32 revision 2 targets as well. The mipsisa*-elfoabi* and 8853 mipsisa*-sde-elf* configurations provide suitable library support. 8854 * GCC now recognizes the -mdmx and -mmt options and passes them down 8855 to the assembler. It does nothing else with the options at present. 8856 8857 SPU (Synergistic Processor Unit) of the Cell Broadband Engine Architecture 8858 (BEA) 8859 8860 * Support has been added for this new architecture. 8861 8862 RS6000 (POWER/PowerPC) 8863 8864 * Support for the PowerPC 750CL paired-single instructions has been 8865 added with a new powerpc-*-linux*paired* target configuration. It 8866 is enabled by an associated -mpaired option and can be accessed 8867 using new built-in functions. 8868 * Support for auto-detecting architecture and system configuration to 8869 auto-select processor optimization tuning. 8870 * Support for VMX on AIX 5.3 has been added. 8871 * Support for AIX Version 6.1 has been added. 8872 8873 S/390, zSeries and System z9 8874 8875 * Support for the IBM System z9 EC/BC processor (z9 GA3) has been 8876 added. When using the -march=z9-ec option, the compiler will 8877 generate code making use of instructions provided by the decimal 8878 floating point facility and the floating point conversion facility 8879 (pfpo). Besides the instructions used to implement decimal floating 8880 point operations these facilities also contain instructions to move 8881 between general purpose and floating point registers and to modify 8882 and copy the sign-bit of floating point values. 8883 * When the -march=z9-ec option is used the new 8884 -mhard-dfp/-mno-hard-dfp options can be used to specify whether the 8885 decimal floating point hardware instructions will be used or not. 8886 If none of them is given the hardware support is enabled by 8887 default. 8888 * The -mstack-guard option can now be omitted when using stack 8889 checking via -mstack-size in order to let GCC choose a sensible 8890 stack guard value according to the frame size of each function. 8891 * Various changes to improve performance of generated code have been 8892 implemented, including: 8893 + The condition code set by an add logical with carry 8894 instruction is now available for overflow checks like: a + b + 8895 carry < b. 8896 + The test data class instruction is now used to implement 8897 sign-bit and infinity checks of binary and decimal floating 8898 point numbers. 8899 8900 SPARC 8901 8902 * Support for the Sun UltraSPARC T2 (Niagara 2) processor has been 8903 added. 8904 8905 Xtensa 8906 8907 * Stack unwinding for exception handling now uses by default a 8908 specialized version of DWARF unwinding. This is not 8909 binary-compatible with the setjmp/longjmp (sjlj) unwinding used for 8910 Xtensa with previous versions of GCC. 8911 * For Xtensa processors that include the Conditional Store option, 8912 the built-in functions for atomic memory access are now implemented 8913 using S32C1I instructions. 8914 * If the Xtensa NSA option is available, GCC will use it to implement 8915 the __builtin_ctz and __builtin_clz functions. 8916 8917 Documentation improvements 8918 8919 * Existing libstdc++ documentation has been edited and restructured 8920 into a single DocBook XML manual. The results can be viewed online 8921 [18]here. 8922 8923 Other significant improvements 8924 8925 * The compiler's --help command-line option has been extended so that 8926 it now takes an optional set of arguments. These arguments restrict 8927 the information displayed to specific classes of command-line 8928 options, and possibly only a subset of those options. It is also 8929 now possible to replace the descriptive text associated with each 8930 displayed option with an indication of its current value, or for 8931 binary options, whether it has been enabled or disabled. 8932 Here are some examples. The following will display all the options 8933 controlling warning messages: 8934 --help=warnings 8935 8936 Whereas this will display all the undocumented, target specific 8937 options: 8938 --help=target,undocumented 8939 8940 This sequence of commands will display the binary optimizations 8941 that are enabled by -O3: 8942 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 8943 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 8944 diff /tmp/O2-opts /tmp/O3-opts | grep enabled 8945 8946 * The configure options --with-pkgversion and --with-bugurl have been 8947 added. These allow distributors of GCC to include a 8948 distributor-specific string in manuals and --version output and to 8949 specify the URL for reporting bugs in their versions of GCC. 8950 8951 GCC 4.3.1 8952 8953 This is the [19]list of problem reports (PRs) from GCC's bug tracking 8954 system that are known to be fixed in the 4.3.1 release. This list might 8955 not be complete (that is, it is possible that some PRs that have been 8956 fixed are not listed here). 8957 8958 Target Specific Changes 8959 8960 IA-32/x86-64 8961 8962 ABI changes 8963 8964 * Starting with GCC 4.3.1, decimal floating point variables are 8965 aligned to their natural boundaries when they are passed on the 8966 stack for i386. 8967 8968 Command-line changes 8969 8970 * Starting with GCC 4.3.1, the -mcld option has been added to 8971 automatically generate a cld instruction in the prologue of 8972 functions that use string instructions. This option is used for 8973 backward compatibility on some operating systems and can be enabled 8974 by default for 32-bit x86 targets by configuring GCC with the 8975 --enable-cld configure option. 8976 8977 GCC 4.3.2 8978 8979 This is the [20]list of problem reports (PRs) from GCC's bug tracking 8980 system that are known to be fixed in the 4.3.2 release. This list might 8981 not be complete (that is, it is possible that some PRs that have been 8982 fixed are not listed here). 8983 8984 GCC 4.3.3 8985 8986 This is the [21]list of problem reports (PRs) from GCC's bug tracking 8987 system that are known to be fixed in the 4.3.3 release. This list might 8988 not be complete (that is, it is possible that some PRs that have been 8989 fixed are not listed here). 8990 8991 GCC 4.3.4 8992 8993 This is the [22]list of problem reports (PRs) from GCC's bug tracking 8994 system that are known to be fixed in the 4.3.4 release. This list might 8995 not be complete (that is, it is possible that some PRs that have been 8996 fixed are not listed here). 8997 8998 GCC 4.3.5 8999 9000 This is the [23]list of problem reports (PRs) from GCC's bug tracking 9001 system that are known to be fixed in the 4.3.5 release. This list might 9002 not be complete (that is, it is possible that some PRs that have been 9003 fixed are not listed here). 9004 9005 GCC 4.3.6 9006 9007 This is the [24]list of problem reports (PRs) from GCC's bug tracking 9008 system that are known to be fixed in the 4.3.6 release. This list might 9009 not be complete (that is, it is possible that some PRs that have been 9010 fixed are not listed here). 9011 9012 9013 For questions related to the use of GCC, please consult these web 9014 pages and the [25]GCC manuals. If that fails, the 9015 [26]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9016 web pages and the development of GCC are welcome on our developer 9017 list at [27]gcc (a] gcc.gnu.org. All of [28]our lists have public 9018 archives. 9019 9020 Copyright (C) [29]Free Software Foundation, Inc. Verbatim copying and 9021 distribution of this entire article is permitted in any medium, 9022 provided this notice is preserved. 9023 9024 These pages are [30]maintained by the GCC team. Last modified 9025 2018-09-30[31]. 9026 9027 References 9028 9029 1. http://gcc.gnu.org/gcc-4.3/changes.html#4.3.5 9030 2. https://gmplib.org/ 9031 3. https://www.mpfr.org/ 9032 4. https://gcc.gnu.org/install/prerequisites.html 9033 5. https://gcc.gnu.org/ml/gcc-announce/2001/msg00000.html 9034 6. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#Warning-Options 9035 7. http://gcc.gnu.org/gcc-4.3/porting_to.html 9036 8. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 9037 9. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 9038 10. http://gcc.gnu.org/gcc-4.3/cxx0x_status.html 9039 11. https://gcc.gnu.org/onlinedocs/libstdc++/manual/parallel_mode.html 9040 12. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#Code-Gen-Options 9041 13. https://gcc.gnu.org/onlinedocs/gfortran/Code-Gen-Options.html#index-g_t_0040code_007bfinit-local-zero_007d-167 9042 14. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/GAMMA.html 9043 15. https://gcc.gnu.org/onlinedocs/gcc-4.3.0/gfortran/LGAMMA.html 9044 16. https://gcc.gnu.org/onlinedocs/gfortran/Fortran-Dialect-Options.html 9045 17. https://gcc.gnu.org/onlinedocs/gfortran/BOZ-literal-constants.html 9046 18. https://gcc.gnu.org/onlinedocs/libstdc++/ 9047 19. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.1 9048 20. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.2 9049 21. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.3 9050 22. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.4 9051 23. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.5 9052 24. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.3.6 9053 25. https://gcc.gnu.org/onlinedocs/ 9054 26. mailto:gcc-help (a] gcc.gnu.org 9055 27. mailto:gcc (a] gcc.gnu.org 9056 28. https://gcc.gnu.org/lists.html 9057 29. https://www.fsf.org/ 9058 30. https://gcc.gnu.org/about.html 9059 31. http://validator.w3.org/check/referer 9060 ====================================================================== 9061 http://gcc.gnu.org/gcc-4.2/index.html 9062 9063 GCC 4.2 Release Series 9064 9065 May 19, 2008 9066 9067 The [1]GNU project and the GCC developers are pleased to announce the 9068 release of GCC 4.2.4. 9069 9070 This release is a bug-fix release, containing fixes for regressions in 9071 GCC 4.2.3 relative to previous releases of GCC. 9072 9073 This release series is no longer maintained. 9074 9075 Release History 9076 9077 GCC 4.2.4 9078 May 19, 2008 ([2]changes) 9079 9080 GCC 4.2.3 9081 February 1, 2008 ([3]changes) 9082 9083 GCC 4.2.2 9084 October 7, 2007 ([4]changes) 9085 9086 GCC 4.2.1 9087 July 18, 2007 ([5]changes) 9088 9089 GCC 4.2.0 9090 May 13, 2007 ([6]changes) 9091 9092 References and Acknowledgements 9093 9094 GCC used to stand for the GNU C Compiler, but since the compiler 9095 supports several other languages aside from C, it now stands for the 9096 GNU Compiler Collection. 9097 9098 A list of [7]successful builds is updated as new information becomes 9099 available. 9100 9101 The GCC developers would like to thank the numerous people that have 9102 contributed new features, improvements, bug fixes, and other changes as 9103 well as test results to GCC. This [8]amazing group of volunteers is 9104 what makes GCC successful. 9105 9106 For additional information about GCC please refer to the [9]GCC project 9107 web site or contact the [10]GCC development mailing list. 9108 9109 To obtain GCC please use [11]our mirror sites or [12]our SVN server. 9110 9111 9112 For questions related to the use of GCC, please consult these web 9113 pages and the [13]GCC manuals. If that fails, the 9114 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9115 web pages and the development of GCC are welcome on our developer 9116 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 9117 archives. 9118 9119 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 9120 distribution of this entire article is permitted in any medium, 9121 provided this notice is preserved. 9122 9123 These pages are [18]maintained by the GCC team. Last modified 9124 2018-09-30[19]. 9125 9126 References 9127 9128 1. http://www.gnu.org/ 9129 2. http://gcc.gnu.org/gcc-4.2/changes.html 9130 3. http://gcc.gnu.org/gcc-4.2/changes.html 9131 4. http://gcc.gnu.org/gcc-4.2/changes.html 9132 5. http://gcc.gnu.org/gcc-4.2/changes.html 9133 6. http://gcc.gnu.org/gcc-4.2/changes.html 9134 7. http://gcc.gnu.org/gcc-4.2/buildstat.html 9135 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 9136 9. http://gcc.gnu.org/index.html 9137 10. mailto:gcc (a] gcc.gnu.org 9138 11. http://gcc.gnu.org/mirrors.html 9139 12. http://gcc.gnu.org/svn.html 9140 13. https://gcc.gnu.org/onlinedocs/ 9141 14. mailto:gcc-help (a] gcc.gnu.org 9142 15. mailto:gcc (a] gcc.gnu.org 9143 16. https://gcc.gnu.org/lists.html 9144 17. https://www.fsf.org/ 9145 18. https://gcc.gnu.org/about.html 9146 19. http://validator.w3.org/check/referer 9147 ====================================================================== 9148 http://gcc.gnu.org/gcc-4.2/changes.html 9149 9150 GCC 4.2 Release Series 9151 Changes, New Features, and Fixes 9152 9153 Caveats 9154 9155 * GCC no longer accepts the -fshared-data option. This option has had 9156 no effect in any GCC 4 release; the targets to which the option 9157 used to apply had been removed before GCC 4.0. 9158 9159 General Optimizer Improvements 9160 9161 * New command-line options specify the possible relationships among 9162 parameters and between parameters and global data. For example, 9163 -fargument-noalias-anything specifies that arguments do not alias 9164 any other storage. 9165 Each language will automatically use whatever option is required by 9166 the language standard. You should not need to use these options 9167 yourself. 9168 9169 New Languages and Language specific improvements 9170 9171 * [1]OpenMP is now supported for the C, C++ and Fortran compilers. 9172 * New command-line options -fstrict-overflow and -Wstrict-overflow 9173 have been added. -fstrict-overflow tells the compiler that it may 9174 assume that the program follows the strict signed overflow 9175 semantics permitted for the language: for C and C++ this means that 9176 the compiler may assume that signed overflow does not occur. For 9177 example, a loop like 9178 for (i = 1; i > 0; i *= 2) 9179 9180 is presumably intended to continue looping until i overflows. With 9181 -fstrict-overflow, the compiler may assume that signed overflow 9182 will not occur, and transform this into an infinite loop. 9183 -fstrict-overflow is turned on by default at -O2, and may be 9184 disabled via -fno-strict-overflow. The -Wstrict-overflow option may 9185 be used to warn about cases where the compiler assumes that signed 9186 overflow will not occur. It takes five different levels: 9187 -Wstrict-overflow=1 to 5. See the [2]documentation for details. 9188 -Wstrict-overflow=1 is enabled by -Wall. 9189 * The new command-line option -fno-toplevel-reorder directs GCC to 9190 emit top-level functions, variables, and asm statements in the same 9191 order that they appear in the input file. This is intended to 9192 support existing code which relies on a particular ordering (for 9193 example, code which uses top-level asm statements to switch 9194 sections). For new code, it is generally better to use function and 9195 variable attributes. The -fno-toplevel-reorder option may be used 9196 for most cases which currently use -fno-unit-at-a-time. The 9197 -fno-unit-at-a-time option will be removed in some future version 9198 of GCC. If you know of a case which requires -fno-unit-at-a-time 9199 which is not fixed by -fno-toplevel-reorder, please open a bug 9200 report. 9201 9202 C family 9203 9204 * The pragma redefine_extname will now macro expand its tokens for 9205 compatibility with SunPRO. 9206 * In the next release of GCC, 4.3, -std=c99 or -std=gnu99 will direct 9207 GCC to handle inline functions as specified in the C99 standard. In 9208 preparation for this, GCC 4.2 will warn about any use of non-static 9209 inline functions in gnu99 or c99 mode. This new warning may be 9210 disabled with the new gnu_inline function attribute or the new 9211 -fgnu89-inline command-line option. Also, GCC 4.2 and later will 9212 define one of the preprocessor macros __GNUC_GNU_INLINE__ or 9213 __GNUC_STDC_INLINE__ to indicate the semantics of inline functions 9214 in the current compilation. 9215 * A new command-line option -Waddress has been added to warn about 9216 suspicious uses of memory addresses as, for example, using the 9217 address of a function in a conditional expression, and comparisons 9218 against the memory address of a string literal. This warning is 9219 enabled by -Wall. 9220 9221 C++ 9222 9223 * C++ visibility handling has been overhauled. 9224 Restricted visiblity is propagated from classes to members, from 9225 functions to local statics, and from templates and template 9226 arguments to instantiations, unless the latter has explicitly 9227 declared visibility. 9228 The visibility attribute for a class must come between the 9229 class-key and the name, not after the closing brace. 9230 Attributes are now allowed for enums and elaborated-type-specifiers 9231 that only declare a type. 9232 Members of the anonymous namespace are now local to a particular 9233 translation unit, along with any other declarations which use them, 9234 though they are still treated as having external linkage for 9235 language semantics. 9236 * The (undocumented) extension which permitted templates with default 9237 arguments to be bound to template template parameters with fewer 9238 parameters has been removed. For example: 9239 template <template <typename> class C> 9240 void f(C<double>) {} 9241 9242 template <typename T, typename U = int> 9243 struct S {}; 9244 9245 template void f(S<double>); 9246 9247 is no longer accepted by G++. The reason this code is not accepted 9248 is that S is a template with two parameters; therefore, it cannot 9249 be bound to C which has only one parameter. 9250 * The <?, >?, <?=, and >?= operators, deprecated in previous GCC 9251 releases, have been removed. 9252 * The command-line option -fconst-strings, deprecated in previous GCC 9253 releases, has been removed. 9254 * The configure variable enable-__cxa_atexit is now enabled by 9255 default for more targets. Enabling this variable is necessary in 9256 order for static destructors to be executed in the correct order, 9257 but it depends upon the presence of a non-standard C library in the 9258 target library in order to work. The variable is now enabled for 9259 more targets which are known to have suitable C libraries. 9260 * -Wextra will produce warnings for if statements with a semicolon as 9261 the only body, to catch code like: 9262 if (a); 9263 return 1; 9264 return 0; 9265 9266 To suppress the warning in valid cases, use { } instead. 9267 * The C++ front end now also produces strict aliasing warnings when 9268 -fstrict-aliasing -Wstrict-aliasing is in effect. 9269 9270 Runtime Library (libstdc++) 9271 9272 * Added support for TR1 <random>, <complex>, and C compatibility 9273 headers. In addition, a lock-free version of shared_ptr was 9274 contributed as part of Phillip Jordan's Google Summer of Code 9275 project on lock-free containers. 9276 * In association with the Summer of Code work on lock-free 9277 containers, the interface for atomic builtins was adjusted, 9278 creating simpler alternatives for non-threaded code paths. Also, 9279 usage was consolidated and all elements were moved from namespace 9280 std to namespace__gnu_cxx. Affected interfaces are the functions 9281 __exchange_and_add, __atomic_add, and the objects __mutex, 9282 __recursive_mutex, and __scoped_lock. 9283 * Support for versioning weak symbol names via namespace association 9284 was added. However, as this changes the names of exported symbols, 9285 this is turned off by default in the current ABI. Intrepid users 9286 can enable this feature by using 9287 --enable-symvers=gnu-versioned-namespace during configuration. 9288 * Revised, simplified, and expanded policy-based associative 9289 containers, including data types for tree and trie forms 9290 (basic_tree, tree, trie), lists (list_update), and both 9291 collision-chaining and probing hash-based containers 9292 (basic_hash_table, cc_hash_table, gp_hash_table). More details per 9293 the [3]documentation. 9294 * The implementation of the debug mode was modified, whereby the 9295 debug namespaces were nested inside of namespace std and namespace 9296 __gnu_cxx in order to resolve some long standing corner cases 9297 involving name lookup. Debug functionality from the policy-based 9298 data structures was consolidated and enabled with the single macro, 9299 _GLIBCXX_DEBUG. See PR 26142 for more information. 9300 * Added extensions for type traits: __conditional_type, 9301 __numeric_traits, __add_unsigned, __removed_unsigned, __enable_if. 9302 * Added a typelist implementation for compile-time meta-programming. 9303 Elements for typelist construction and operation can be found 9304 within namespace __gnu_cxx::typelist. 9305 * Added a new allocator, __gnu_cxx::throw_allocator, for testing 9306 exception-safety. 9307 * Enabled library-wide visibility control, allowing -fvisibility to 9308 be used. 9309 * Consolidated all nested namespaces and the conversion of 9310 __gnu_internal implementation-private details to anonymous 9311 namespaces whenever possible. 9312 * Implemented LWG resolutions DR 431 and DR 538. 9313 9314 Fortran 9315 9316 * Support for allocatable components has been added (TR 15581 and 9317 Fortran 2003). 9318 * Support for the Fortran 2003 streaming IO extension has been added. 9319 * The GNU Fortran compiler now uses 4-byte record markers by default 9320 for unformatted files to be compatible with g77 and most other 9321 compilers. The implementation allows for records greater than 2 GB 9322 and is compatible with several other compilers. Older versions of 9323 gfortran used 8-byte record markers by default (on most systems). 9324 In order to change the length of the record markers, e.g. to read 9325 unformatted files created by older gfortran versions, the 9326 [4]-frecord-marker=8 option can be used. 9327 9328 Java (GCJ) 9329 9330 * A new command-line option -static-libgcj has been added for targets 9331 that use a linker compatible with GNU Binutils. As its name 9332 implies, this causes libgcj to be linked statically. In some cases 9333 this causes the resulting executable to start faster and use less 9334 memory than if the shared version of libgcj were used. However 9335 caution should be used as it can also cause essential parts of the 9336 library to be omitted. Some of these issues are discussed in: 9337 [5]https://gcc.gnu.org/wiki/Statically_linking_libgcj 9338 * fastjar is no longer bundled with GCC. To build libgcj, you will 9339 need either InfoZIP (both zip and unzip) or an external jar 9340 program. In the former case, the GCC build will install a jar shell 9341 script that is based on InfoZIP and provides the same functionality 9342 as fastjar. 9343 9344 New Targets and Target Specific Improvements 9345 9346 IA-32/x86-64 9347 9348 * -mtune=generic can now be used to generate code running well on 9349 common x86 chips. This includes AMD Athlon, AMD Opteron, Intel 9350 Pentium-M, Intel Pentium 4 and Intel Core 2. 9351 * -mtune=native and -march=native will produce code optimized for the 9352 host architecture as detected using the cpuid instruction. 9353 * Added a new command-line option -fstackrealign and and 9354 __attribute__ ((force_align_arg_pointer)) to realign the stack at 9355 runtime. This allows functions compiled with a vector-aligned stack 9356 to be invoked from legacy objects that keep only word-alignment. 9357 9358 SPARC 9359 9360 * The default CPU setting has been changed from V7 to V9 in 32-bit 9361 mode on Solaris 7 and above. This is already the case in 64-bit 9362 mode. It can be overridden by specifying --with-cpu at configure 9363 time. 9364 * Back-end support of built-in functions for atomic memory access has 9365 been implemented. 9366 * Support for the Sun UltraSPARC T1 (Niagara) processor has been 9367 added. 9368 9369 M32C 9370 9371 * Various bug fixes have made some functions (notably, functions 9372 returning structures) incompatible with previous releases. 9373 Recompiling all libraries is recommended. Note that code quality 9374 has considerably improved since 4.1, making a recompile even more 9375 beneficial. 9376 9377 MIPS 9378 9379 * Added support for the Broadcom SB-1A core. 9380 9381 IA-64 9382 9383 * Added support for IA-64 data and control speculation. By default 9384 speculation is enabled only during second scheduler pass. A number 9385 of machine flags was introduced to control the usage of speculation 9386 for both scheduler passes. 9387 9388 HPPA 9389 9390 * Added Java language support (libffi and libjava) for 32-bit HP-UX 9391 11 target. 9392 9393 Obsolete Systems 9394 9395 Documentation improvements 9396 9397 PDF Documentation 9398 9399 * A make pdf target has been added to the top-level makefile, 9400 enabling automated production of PDF documentation files. 9401 (Front-ends external to GCC should modify their Make-lang.in file 9402 to add a lang.pdf: target.) 9403 9404 Other significant improvements 9405 9406 Build system improvements 9407 9408 * All the components of the compiler are now bootstrapped by default. 9409 This improves the resilience to bugs in the system compiler or 9410 binary compatibility problems, as well as providing better testing 9411 of GCC 4.2 itself. In addition, if you build the compiler from a 9412 combined tree, the assembler, linker, etc. will also be 9413 bootstrapped (i.e. built with themselves). 9414 You can disable this behavior, and go back to the pre-GCC 4.2 set 9415 up, by configuring GCC with --disable-bootstrap. 9416 * The rules that configure follows to find target tools resemble more 9417 closely the locations that the built compiler will search. In 9418 addition, you can use the new configure option --with-target-tools 9419 to specify where to find the target tools used during the build, 9420 without affecting what the built compiler will use. 9421 This can be especially useful when building packages of GCC. For 9422 example, you may want to build GCC with GNU as or ld, even if the 9423 resulting compiler to work with the native assembler and linker. To 9424 do so, you can use --with-target-tools to point to the native 9425 tools. 9426 9427 Incompatible changes to the build system 9428 9429 * Front-ends external to GCC should modify their Make-lang.in file to 9430 replace double-colon rules (e.g. dvi::) with normal rules (like 9431 lang.dvi:). Front-end makefile hooks do not use double-colon rules 9432 anymore. 9433 * Up to GCC 4.1, a popular way to specify the target tools used 9434 during the build was to create directories named gas, binutils, 9435 etc. in the build tree, and create links to the tools from there. 9436 This does not work any more when the compiler is bootstrapped. The 9437 new configure option --with-target-tools provides a better way to 9438 achieve the same effect, and works for all native and cross 9439 settings. 9440 9441 9442 For questions related to the use of GCC, please consult these web 9443 pages and the [6]GCC manuals. If that fails, the 9444 [7]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9445 web pages and the development of GCC are welcome on our developer 9446 list at [8]gcc (a] gcc.gnu.org. All of [9]our lists have public archives. 9447 9448 Copyright (C) [10]Free Software Foundation, Inc. Verbatim copying and 9449 distribution of this entire article is permitted in any medium, 9450 provided this notice is preserved. 9451 9452 These pages are [11]maintained by the GCC team. Last modified 9453 2018-09-30[12]. 9454 9455 References 9456 9457 1. http://gcc.gnu.org/projects/gomp/ 9458 2. https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html 9459 3. https://gcc.gnu.org/onlinedocs/libstdc++/ext/pb_ds/index.html 9460 4. https://gcc.gnu.org/onlinedocs/gfortran/Runtime-Options.html 9461 5. https://gcc.gnu.org/wiki/Statically_linking_libgcj 9462 6. https://gcc.gnu.org/onlinedocs/ 9463 7. mailto:gcc-help (a] gcc.gnu.org 9464 8. mailto:gcc (a] gcc.gnu.org 9465 9. https://gcc.gnu.org/lists.html 9466 10. https://www.fsf.org/ 9467 11. https://gcc.gnu.org/about.html 9468 12. http://validator.w3.org/check/referer 9469 ====================================================================== 9470 http://gcc.gnu.org/gcc-4.1/index.html 9471 9472 GCC 4.1 Release Series 9473 9474 February 13, 2007 9475 9476 The [1]GNU project and the GCC developers are pleased to announce the 9477 release of GCC 4.1.2. 9478 9479 This release is a bug-fix release, containing fixes for regressions in 9480 GCC 4.1.1 relative to previous releases of GCC. 9481 9482 This release series is no longer maintained. 9483 9484 Release History 9485 9486 GCC 4.1.2 9487 February 13, 2007 ([2]changes) 9488 9489 GCC 4.1.1 9490 May 24, 2006 ([3]changes) 9491 9492 GCC 4.1.0 9493 February 28, 2006 ([4]changes) 9494 9495 References and Acknowledgements 9496 9497 GCC used to stand for the GNU C Compiler, but since the compiler 9498 supports several other languages aside from C, it now stands for the 9499 GNU Compiler Collection. 9500 9501 A list of [5]successful builds is updated as new information becomes 9502 available. 9503 9504 The GCC developers would like to thank the numerous people that have 9505 contributed new features, improvements, bug fixes, and other changes as 9506 well as test results to GCC. This [6]amazing group of volunteers is 9507 what makes GCC successful. 9508 9509 For additional information about GCC please refer to the [7]GCC project 9510 web site or contact the [8]GCC development mailing list. 9511 9512 To obtain GCC please use [9]our mirror sites or [10]our SVN server. 9513 9514 9515 For questions related to the use of GCC, please consult these web 9516 pages and the [11]GCC manuals. If that fails, the 9517 [12]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 9518 web pages and the development of GCC are welcome on our developer 9519 list at [13]gcc (a] gcc.gnu.org. All of [14]our lists have public 9520 archives. 9521 9522 Copyright (C) [15]Free Software Foundation, Inc. Verbatim copying and 9523 distribution of this entire article is permitted in any medium, 9524 provided this notice is preserved. 9525 9526 These pages are [16]maintained by the GCC team. Last modified 9527 2018-09-30[17]. 9528 9529 References 9530 9531 1. http://www.gnu.org/ 9532 2. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 9533 3. http://gcc.gnu.org/gcc-4.1/changes.html 9534 4. http://gcc.gnu.org/gcc-4.1/changes.html 9535 5. http://gcc.gnu.org/gcc-4.1/buildstat.html 9536 6. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 9537 7. http://gcc.gnu.org/index.html 9538 8. mailto:gcc (a] gcc.gnu.org 9539 9. http://gcc.gnu.org/mirrors.html 9540 10. http://gcc.gnu.org/svn.html 9541 11. https://gcc.gnu.org/onlinedocs/ 9542 12. mailto:gcc-help (a] gcc.gnu.org 9543 13. mailto:gcc (a] gcc.gnu.org 9544 14. https://gcc.gnu.org/lists.html 9545 15. https://www.fsf.org/ 9546 16. https://gcc.gnu.org/about.html 9547 17. http://validator.w3.org/check/referer 9548 ====================================================================== 9549 http://gcc.gnu.org/gcc-4.1/changes.html 9550 9551 GCC 4.1 Release Series 9552 Changes, New Features, and Fixes 9553 9554 The latest release in the 4.1 release series is [1]GCC 4.1.2. 9555 9556 Caveats 9557 9558 General Optimizer Improvements 9559 9560 * GCC now has infrastructure for inter-procedural optimizations and 9561 the following inter-procedural optimizations are implemented: 9562 + Profile guided inlining. When doing profile feedback guided 9563 optimization, GCC can now use the profile to make better 9564 informed decisions on whether inlining of a function is 9565 profitable or not. This means that GCC will no longer inline 9566 functions at call sites that are not executed very often, and 9567 that functions at hot call sites are more likely to be 9568 inlined. 9569 A new parameter min-inline-recursive-probability is also now 9570 available to throttle recursive inlining of functions with 9571 small average recursive depths. 9572 + Discovery of pure and const functions, a form of side-effects 9573 analysis. While older GCC releases could also discover such 9574 special functions, the new IPA-based pass runs earlier so that 9575 the results are available to more optimizers. The pass is also 9576 simply more powerful than the old one. 9577 + Analysis of references to static variables and type escape 9578 analysis, also forms of side-effects analysis. The results of 9579 these passes allow the compiler to be less conservative about 9580 call-clobbered variables and references. This results in more 9581 redundant loads being eliminated and in making static 9582 variables candidates for register promotion. 9583 + Improvement of RTL-based alias analysis. The results of type 9584 escape analysis are fed to the RTL type-based alias analyzer, 9585 allowing it to disambiguate more memory references. 9586 + Interprocedural constant propagation and function versioning. 9587 This pass looks for functions that are always called with the 9588 same constant value for one or more of the function arguments, 9589 and propagates those constants into those functions. 9590 + GCC will now eliminate static variables whose usage was 9591 optimized out. 9592 + -fwhole-program --combine can now be used to make all 9593 functions in program static allowing whole program 9594 optimization. As an exception, the main function and all 9595 functions marked with the new externally_visible attribute are 9596 kept global so that programs can link with runtime libraries. 9597 * GCC can now do a form of partial dead code elimination (PDCE) that 9598 allows code motion of expressions to the paths where the result of 9599 the expression is actually needed. This is not always a win, so the 9600 pass has been limited to only consider profitable cases. Here is an 9601 example: 9602 int foo (int *, int *); 9603 int 9604 bar (int d) 9605 { 9606 int a, b, c; 9607 b = d + 1; 9608 c = d + 2; 9609 a = b + c; 9610 if (d) 9611 { 9612 foo (&b, &c); 9613 a = b + c; 9614 } 9615 printf ("%d\n", a); 9616 } 9617 9618 The a = b + c can be sunk to right before the printf. Normal code 9619 sinking will not do this, it will sink the first one above into the 9620 else-branch of the conditional jump, which still gives you two 9621 copies of the code. 9622 * GCC now has a value range propagation pass. This allows the 9623 compiler to eliminate bounds checks and branches. The results of 9624 the pass can also be used to accurately compute branch 9625 probabilities. 9626 * The pass to convert PHI nodes to straight-line code (a form of 9627 if-conversion for GIMPLE) has been improved significantly. The two 9628 most significant improvements are an improved algorithm to 9629 determine the order in which the PHI nodes are considered, and an 9630 improvement that allow the pass to consider if-conversions of basic 9631 blocks with more than two predecessors. 9632 * Alias analysis improvements. GCC can now differentiate between 9633 different fields of structures in Tree-SSA's virtual operands form. 9634 This lets stores/loads from non-overlapping structure fields not 9635 conflict. A new algorithm to compute points-to sets was contributed 9636 that can allows GCC to see now that p->a and p->b, where p is a 9637 pointer to a structure, can never point to the same field. 9638 * Various enhancements to auto-vectorization: 9639 + Incrementally preserve SSA form when vectorizing. 9640 + Incrementally preserve loop-closed form when vectorizing. 9641 + Improvements to peeling for alignment: generate better code 9642 when the misalignment of an access is known at compile time, 9643 or when different accesses are known to have the same 9644 misalignment, even if the misalignment amount itself is 9645 unknown. 9646 + Consider dependence distance in the vectorizer. 9647 + Externalize generic parts of data reference analysis to make 9648 this analysis available to other passes. 9649 + Vectorization of conditional code. 9650 + Reduction support. 9651 * GCC can now partition functions in sections of hot and cold code. 9652 This can significantly improve performance due to better 9653 instruction cache locality. This feature works best together with 9654 profile feedback driven optimization. 9655 * A new pass to avoid saving of unneeded arguments to the stack in 9656 vararg functions if the compiler can prove that they will not be 9657 needed. 9658 * Transition of basic block profiling to tree level implementation 9659 has been completed. The new implementation should be considerably 9660 more reliable (hopefully avoiding profile mismatch errors when 9661 using -fprofile-use or -fbranch-probabilities) and can be used to 9662 drive higher level optimizations, such as inlining. 9663 The -ftree-based-profiling command-line option was removed and 9664 -fprofile-use now implies disabling old RTL level loop optimizer 9665 (-fno-loop-optimize). Speculative prefetching optimization 9666 (originally enabled by -fspeculative-prefetching) was removed. 9667 9668 New Languages and Language specific improvements 9669 9670 C and Objective-C 9671 9672 * The old Bison-based C and Objective-C parser has been replaced by a 9673 new, faster hand-written recursive-descent parser. 9674 9675 Ada 9676 9677 * The build infrastructure for the Ada runtime library and tools has 9678 been changed to be better integrated with the rest of the build 9679 infrastructure of GCC. This should make doing cross builds of Ada a 9680 bit easier. 9681 9682 C++ 9683 9684 * ARM-style name-injection of friend declarations is no longer the 9685 default. For example: 9686 struct S { 9687 friend void f(); 9688 }; 9689 9690 void g() { f(); } 9691 will not be accepted; instead a declaration of f will need to be 9692 present outside of the scope of S. The new -ffriend-injection 9693 option will enable the old behavior. 9694 * The (undocumented) extension which permitted templates with default 9695 arguments to be bound to template template parameters with fewer 9696 parameters has been deprecated, and will be removed in the next 9697 major release of G++. For example: 9698 template <template <typename> class C> 9699 void f(C<double>) {} 9700 9701 template <typename T, typename U = int> 9702 struct S {}; 9703 9704 template void f(S<double>); 9705 9706 makes use of the deprecated extension. The reason this code is not 9707 valid ISO C++ is that S is a template with two parameters; 9708 therefore, it cannot be bound to C which has only one parameter. 9709 9710 Runtime Library (libstdc++) 9711 9712 * Optimization work: 9713 + A new implementation of std::search_n is provided, better 9714 performing in case of random access iterators. 9715 + Added further efficient specializations of istream functions, 9716 i.e., character array and string extractors. 9717 + Other smaller improvements throughout. 9718 * Policy-based associative containers, designed for high-performance, 9719 flexibility and semantic safety are delivered in ext/pb_assoc. 9720 * A versatile string class, __gnu_cxx::__versa_string, providing 9721 facilities conforming to the standard requirements for 9722 basic_string, is delivered in <ext/vstring.h>. In particular: 9723 + Two base classes are provided: the default one avoids 9724 reference counting and is optimized for short strings; the 9725 alternate one, still uses it while improving in a few low 9726 level areas (e.g., alignment). See vstring_fwd.h for some 9727 useful typedefs. 9728 + Various algorithms have been rewritten (e.g., replace), the 9729 code streamlined and simple optimizations added. 9730 + Option 3 of DR 431 is implemented for both available bases, 9731 thus improving the support for stateful allocators. 9732 * As usual, many bugs have been fixed (e.g., libstdc++/13583, 9733 libstdc++/23953) and LWG resolutions put into effect for the first 9734 time (e.g., DR 280, DR 464, N1780 recommendations for DR 233, TR1 9735 Issue 6.19). The implementation status of TR1 is now tracked in the 9736 docs in tr1.html. 9737 9738 Objective-C++ 9739 9740 * A new language front end for Objective-C++ has been added. This 9741 language allows users to mix the object oriented features of 9742 Objective-C with those of C++. 9743 9744 Java (GCJ) 9745 9746 * Core library (libgcj) updates based on GNU Classpath 0.15 - 0.19 9747 features (plus some 0.20 bug-fixes) 9748 + Networking 9749 o The java.net.HttpURLConnection implementation no longer 9750 buffers the entire response body in memory. This means 9751 that response bodies larger than available memory can now 9752 be handled. 9753 + (N)IO 9754 o NIO FileChannel.map implementation, fast bulk put 9755 implementation for DirectByteBuffer (speeds up this 9756 method 10x). 9757 o FileChannel.lock() and FileChannel.force() implemented. 9758 + XML 9759 o gnu.xml fix for nodes created outside a namespace 9760 context. 9761 o Add support for output indenting and 9762 cdata-section-elements output instruction in 9763 xml.transform. 9764 o xml.xpath corrections for cases where elements/attributes 9765 might have been created in non-namespace-aware mode. 9766 Corrections to handling of XSL variables and minor 9767 conformance updates. 9768 + AWT 9769 o GNU JAWT implementation, the AWT Native Interface, which 9770 allows direct access to native screen resources from 9771 within a Canvas's paint method. GNU Classpath Examples 9772 comes with a Demo, see libjava/classpath/examples/README. 9773 o awt.datatransfer updated to 1.5 with support for 9774 FlavorEvents. The gtk+ awt peers now allow copy/paste of 9775 text, images, URIs/files and serialized objects with 9776 other applications and tracking clipboard change events 9777 with gtk+ 2.6 (for gtk+ 2.4 only text and serialized 9778 objects are supported). A GNU Classpath Examples 9779 datatransfer Demo was added to show the new 9780 functionality. 9781 o Split gtk+ awt peers event handling in two threads and 9782 improve gdk lock handling (solves several awt lock ups). 9783 o Speed up awt Image loading. 9784 o Better gtk+ scrollbar peer implementation when using gtk+ 9785 >= 2.6. 9786 o Handle image loading errors correctly for gdkpixbuf and 9787 MediaTracker. 9788 o Better handle GDK lock. Properly prefix gtkpeer native 9789 functions (cp_gtk). 9790 o GdkGraphics2D has been updated to use Cairo 0.5.x or 9791 higher. 9792 o BufferedImage and GtkImage rewrites. All image drawing 9793 operations should now work correctly (flipping requires 9794 gtk+ >= 2.6) 9795 o Future Graphics2D, image and text work is documented at: 9796 [2]http://developer.classpath.org/mediation/ClasspathGrap 9797 hicsImagesText 9798 o When gtk+ 2.6 or higher is installed the default log 9799 handler will produce stack traces whenever a WARNING, 9800 CRITICAL or ERROR message is produced. 9801 + Free Swing 9802 o The RepaintManager has been reworked for more efficient 9803 painting, especially for large GUIs. 9804 o The layout manager OverlayLayout has been implemented, 9805 the BoxLayout has been rewritten to make use of the 9806 SizeRequirements utility class and caching for more 9807 efficient layout. 9808 o Improved accessibility support. 9809 o Significant progress has been made in the implementation 9810 of the javax.swing.plaf.metal package, with most UI 9811 delegates in a working state now. Please test this with 9812 your own applications and provide feedback that will help 9813 us to improve this package. 9814 o The GUI demo (gnu.classpath.examples.swing.Demo) has been 9815 extended to highlight various features in our Free Swing 9816 implementation. And it includes a look and feel switcher 9817 for Metal (default), Ocean and GNU themes. 9818 o The javax.swing.plaf.multi package is now implemented. 9819 o Editing and several key actions for JTree and JTable were 9820 implemented. 9821 o Lots of icons and look and feel improvements for Free 9822 Swing basic and metal themes were added. Try running the 9823 GNU Classpath Swing Demo in examples 9824 (gnu.classpath.examples.swing.Demo) with: 9825 -Dswing.defaultlaf=javax.swing.plaf.basic.BasicLookAndFee 9826 l or 9827 -Dswing.defaultlaf=javax.swing.plaf.metal.MetalLookAndFee 9828 l 9829 o Start of styled text capabilites for java.swing.text. 9830 o DefaultMutableTreeNode pre-order, post-order, depth-first 9831 and breadth-first traversal enumerations implemented. 9832 o JInternalFrame colors and titlebar draw properly. 9833 o JTree is working up to par (icons, selection and keyboard 9834 traversal). 9835 o JMenus were made more compatible in visual and 9836 programmatic behavior. 9837 o JTable changeSelection and multiple selections 9838 implemented. 9839 o JButton and JToggleButton change states work properly 9840 now. 9841 o JFileChooser fixes. 9842 o revalidate() and repaint() fixes which make Free Swing 9843 much more responsive. 9844 o MetalIconFactory implemented. 9845 o Free Swing Top-Level Compatibility. JFrame, JDialog, 9846 JApplet, JInternalFrame, and JWindow are now 1.5 9847 compatible in the sense that you can call add() and 9848 setLayout() directly on them, which will have the same 9849 effect as calling getContentPane().add() and 9850 getContentPane().setLayout(). 9851 o The JTree interface has been completed. JTrees now 9852 recognizes mouse clicks and selections work. 9853 o BoxLayout works properly now. 9854 o Fixed GrayFilter to actually work. 9855 o Metal SplitPane implemented. 9856 o Lots of Free Swing text and editor stuff work now. 9857 + Free RMI and Corba 9858 o Andrew Watson, Vice President and Technical Director of 9859 the Object Management Group, has officially assigned us 9860 20 bit Vendor Minor Code Id: 0x47430 ("GC") that will 9861 mark remote classpath-specific system exceptions. 9862 Obtaining the VMCID means that GNU Classpath now is a 9863 recogniseable type of node in a highly interoperable 9864 CORBA world. 9865 o GNU Classpath now includes the first working draft to 9866 support the RMI over IIOP protocol. The current 9867 implementation is capable of remote invocations, 9868 transferring various Serializables and Externalizables 9869 via RMI-IIOP protocol. It can flatten graphs and, at 9870 least for the simple cases, is interoperable with 1.5 9871 JDKs. 9872 o org.omg.PortableInterceptor and related functionality in 9873 other packages is now implemented: 9874 # The sever and client interceptors work as required 9875 since 1.4. 9876 # The IOR interceptor works as needed for 1.5. 9877 o The org.omg.DynamicAny package is completed and passes 9878 the prepared tests. 9879 o The Portable Object Adapter should now support the output 9880 of the recent IDL to java compilers. These compilers now 9881 generate servants and not CORBA objects as before, making 9882 the output depend on the existing POA implementation. 9883 Completing POA means that such code can already be tried 9884 to run on Classpath. Our POA is tested for the following 9885 usager scenarios: 9886 # POA converts servant to the CORBA object. 9887 # Servant provides to the CORBA object. 9888 # POA activates new CORBA object with the given Object 9889 Id (byte array) that is later accessible for the 9890 servant. 9891 # During the first call, the ServantActivator provides 9892 servant for this and all subsequent calls on the 9893 current object. 9894 # During each call, the ServantLocator provides 9895 servant for this call only. 9896 # ServantLocator or ServantActivator forwards call to 9897 another server. 9898 # POA has a single servant, responsible for all 9899 objects. 9900 # POA has a default servant, but some objects are 9901 explicitly connected to they specific servants. 9902 The POA is verified using tests from the former 9903 cost.omg.org. 9904 o The CORBA implementation is now a working prototype that 9905 should support features up to 1.3 inclusive. We invite 9906 groups writing CORBA dependent applications to try 9907 Classpath implementation, reporting any possible bugs. 9908 The CORBA prototype is interoperable with Sun's 9909 implementation v 1.4, transferring object references, 9910 primitive types, narrow and wide strings, arrays, 9911 structures, trees, abstract interfaces and value types 9912 (feature of CORBA 2.3) between these two platforms. 9913 Remote exceptions are transferred and handled correctly. 9914 The stringified object references (IORs) from various 9915 sources are parsed as required. The transient (for 9916 current session) and permanent (till jre restart) 9917 redirections work. Both Little and Big Endian encoded 9918 messages are accepted. The implementation is verified 9919 using tests from the former cost.omg.org. The current 9920 release includes working examples (see the examples 9921 directory), demonstrating the client-server 9922 communication, using either CORBA Request or IDL-based 9923 stub (usually generated by a IDL to java compiler). These 9924 examples also show how to use the Classpath CORBA naming 9925 service. The IDL to java compiler is not yet written, but 9926 as our library must be compatible, it naturally accepts 9927 the output of other idlj implementations. 9928 + Misc 9929 o Updated TimeZone data against Olson tzdata2005l. 9930 o Make zip and jar packages UTF-8 clean. 9931 o "native" code builds and compiles (warning free) on 9932 Darwin and Solaris. 9933 o java.util.logging.FileHandler now rotates files. 9934 o Start of a generic JDWP framework in gnu/classpath/jdwp. 9935 This is unfinished, but feedback (at classpath (a] gnu.org) 9936 from runtime hackers is greatly appreciated. Although 9937 most of the work is currently being done around gcj/gij 9938 we want this framework to be as VM neutral as possible. 9939 Early design is described in: 9940 [3]https://gcc.gnu.org/ml/java/2005-05/msg00260.html 9941 o QT4 AWT peers, enable by giving configure 9942 --enable-qt-peer. Included, but not ready for production 9943 yet. They are explicitly disabled and not supported. But 9944 if you want to help with the development of these new 9945 features we are interested in feedback. You will have to 9946 explicitly enable them to try them out (and they will 9947 most likely contain bugs). 9948 o Documentation fixes all over the place. See 9949 [4]http://developer.classpath.org/doc/ 9950 9951 New Targets and Target Specific Improvements 9952 9953 IA-32/x86-64 9954 9955 * The x86-64 medium model (that allows building applications whose 9956 data segment exceeds 4GB) was redesigned to match latest ABI draft. 9957 New implementation split large datastructures into separate segment 9958 improving performance of accesses to small datastructures and also 9959 allows linking of small model libraries into medium model programs 9960 as long as the libraries are not accessing the large datastructures 9961 directly. Medium model is also supported in position independent 9962 code now. 9963 The ABI change results in partial incompatibility among medium 9964 model objects. Linking medium model libraries (or objects) compiled 9965 with new compiler into medium model program compiled with older 9966 will likely result in exceeding ranges of relocations. 9967 Binutils 2.16.91 or newer are required for compiling medium model 9968 now. 9969 9970 RS6000 (POWER/PowerPC) 9971 9972 * The AltiVec vector primitives in <altivec.h> are now implemented in 9973 a way that puts a smaller burden on the preprocessor, instead 9974 processing the "overloading" in the front ends. This should benefit 9975 compilation speed on AltiVec vector code. 9976 * AltiVec initializers now are generated more efficiently. 9977 * The popcountb instruction available on POWER5 now is generated. 9978 * The floating point round to integer instructions available on 9979 POWER5+ now is generated. 9980 * Floating point divides can be synthesized using the floating point 9981 reciprocal estimate instructions. 9982 * Double precision floating point constants are initialized as single 9983 precision values if they can be represented exactly. 9984 9985 S/390, zSeries and System z9 9986 9987 * Support for the IBM System z9 109 processor has been added. When 9988 using the -march=z9-109 option, the compiler will generate code 9989 making use of instructions provided by the extended immediate 9990 facility. 9991 * Support for 128-bit IEEE floating point has been added. When using 9992 the -mlong-double-128 option, the compiler will map the long double 9993 data type to 128-bit IEEE floating point. Using this option 9994 constitutes an ABI change, and requires glibc support. 9995 * Various changes to improve performance of generated code have been 9996 implemented, including: 9997 + In functions that do not require a literal pool, register %r13 9998 (which is traditionally reserved as literal pool pointer), can 9999 now be freely used for other purposes by the compiler. 10000 + More precise tracking of register use allows the compiler to 10001 generate more efficient function prolog and epilog code in 10002 certain cases. 10003 + The SEARCH STRING, COMPARE LOGICAL STRING, and MOVE STRING 10004 instructions are now used to implement C string functions. 10005 + The MOVE CHARACTER instruction with single byte overlap is now 10006 used to implement the memset function with non-zero fill byte. 10007 + The LOAD ZERO instructions are now used where appropriate. 10008 + The INSERT CHARACTERS UNDER MASK, STORE CHARACTERS UNDER MASK, 10009 and INSERT IMMEDIATE instructions are now used more frequently 10010 to optimize bitfield operations. 10011 + The BRANCH ON COUNT instruction is now used more frequently. 10012 In particular, the fact that a loop contains a subroutine call 10013 no longer prevents the compiler from using this instruction. 10014 + The compiler is now aware that all shift and rotate 10015 instructions implicitly truncate the shift count to six bits. 10016 * Back-end support for the following generic features has been 10017 implemented: 10018 + The full set of [5]built-in functions for atomic memory 10019 access. 10020 + The -fstack-protector feature. 10021 + The optimization pass avoiding unnecessary stores of incoming 10022 argument registers in functions with variable argument list. 10023 10024 SPARC 10025 10026 * The default code model in 64-bit mode has been changed from 10027 Medium/Anywhere to Medium/Middle on Solaris. 10028 * TLS support is disabled by default on Solaris prior to release 10. 10029 It can be enabled on TLS-capable Solaris 9 versions (4/04 release 10030 and later) by specifying --enable-tls at configure time. 10031 10032 MorphoSys 10033 10034 * Support has been added for this new architecture. 10035 10036 Obsolete Systems 10037 10038 Documentation improvements 10039 10040 Other significant improvements 10041 10042 * GCC can now emit code for protecting applications from 10043 stack-smashing attacks. The protection is realized by buffer 10044 overflow detection and reordering of stack variables to avoid 10045 pointer corruption. 10046 * Some built-in functions have been fortified to protect them against 10047 various buffer overflow (and format string) vulnerabilities. 10048 Compared to the mudflap bounds checking feature, the safe builtins 10049 have far smaller overhead. This means that programs built using 10050 safe builtins should not experience any measurable slowdown. 10051 10052 GCC 4.1.2 10053 10054 This is the [6]list of problem reports (PRs) from GCC's bug tracking 10055 system that are known to be fixed in the 4.1.2 release. This list might 10056 not be complete (that is, it is possible that some PRs that have been 10057 fixed are not listed here). 10058 10059 When generating code for a shared library, GCC now recognizes that 10060 global functions may be replaced when the program runs. Therefore, it 10061 is now more conservative in deducing information from the bodies of 10062 functions. For example, in this example: 10063 void f() {} 10064 void g() { 10065 try { f(); } 10066 catch (...) { 10067 cout << "Exception"; 10068 } 10069 } 10070 10071 G++ would previously have optimized away the catch clause, since it 10072 would have concluded that f cannot throw exceptions. Because users may 10073 replace f with another function in the main body of the program, this 10074 optimization is unsafe, and is no longer performed. If you wish G++ to 10075 continue to optimize as before, you must add a throw() clause to the 10076 declaration of f to make clear that it does not throw exceptions. 10077 10078 10079 For questions related to the use of GCC, please consult these web 10080 pages and the [7]GCC manuals. If that fails, the 10081 [8]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10082 web pages and the development of GCC are welcome on our developer 10083 list at [9]gcc (a] gcc.gnu.org. All of [10]our lists have public 10084 archives. 10085 10086 Copyright (C) [11]Free Software Foundation, Inc. Verbatim copying and 10087 distribution of this entire article is permitted in any medium, 10088 provided this notice is preserved. 10089 10090 These pages are [12]maintained by the GCC team. Last modified 10091 2018-09-30[13]. 10092 10093 References 10094 10095 1. http://gcc.gnu.org/gcc-4.1/changes.html#4.1.2 10096 2. http://developer.classpath.org/mediation/ClasspathGraphicsImagesText 10097 3. https://gcc.gnu.org/ml/java/2005-05/msg00260.html 10098 4. http://developer.classpath.org/doc/ 10099 5. https://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html 10100 6. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.1.2 10101 7. https://gcc.gnu.org/onlinedocs/ 10102 8. mailto:gcc-help (a] gcc.gnu.org 10103 9. mailto:gcc (a] gcc.gnu.org 10104 10. https://gcc.gnu.org/lists.html 10105 11. https://www.fsf.org/ 10106 12. https://gcc.gnu.org/about.html 10107 13. http://validator.w3.org/check/referer 10108 ====================================================================== 10109 http://gcc.gnu.org/gcc-4.0/index.html 10110 10111 GCC 4.0 Release Series 10112 10113 January 31, 2007 10114 10115 The [1]GNU project and the GCC developers are pleased to announce the 10116 release of GCC 4.0.4. 10117 10118 This release is a bug-fix release, containing fixes for regressions in 10119 GCC 4.0.3 relative to previous releases of GCC. 10120 10121 This release series is no longer maintained. 10122 10123 Release History 10124 10125 GCC 4.0.4 10126 January 31, 2007 ([2]changes) 10127 10128 GCC 4.0.3 10129 March 10, 2006 ([3]changes) 10130 10131 GCC 4.0.2 10132 September 28, 2005 ([4]changes) 10133 10134 GCC 4.0.1 10135 July 7, 2005 ([5]changes) 10136 10137 GCC 4.0.0 10138 April 20, 2005 ([6]changes) 10139 10140 References and Acknowledgements 10141 10142 GCC used to stand for the GNU C Compiler, but since the compiler 10143 supports several other languages aside from C, it now stands for the 10144 GNU Compiler Collection. 10145 10146 A list of [7]successful builds is updated as new information becomes 10147 available. 10148 10149 The GCC developers would like to thank the numerous people that have 10150 contributed new features, improvements, bug fixes, and other changes as 10151 well as test results to GCC. This [8]amazing group of volunteers is 10152 what makes GCC successful. 10153 10154 For additional information about GCC please refer to the [9]GCC project 10155 web site or contact the [10]GCC development mailing list. 10156 10157 To obtain GCC please use [11]our mirror sites, or [12]our SVN server. 10158 10159 10160 For questions related to the use of GCC, please consult these web 10161 pages and the [13]GCC manuals. If that fails, the 10162 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10163 web pages and the development of GCC are welcome on our developer 10164 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 10165 archives. 10166 10167 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 10168 distribution of this entire article is permitted in any medium, 10169 provided this notice is preserved. 10170 10171 These pages are [18]maintained by the GCC team. Last modified 10172 2018-09-30[19]. 10173 10174 References 10175 10176 1. http://www.gnu.org/ 10177 2. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 10178 3. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.3 10179 4. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.2 10180 5. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.1 10181 6. http://gcc.gnu.org/gcc-4.0/changes.html 10182 7. http://gcc.gnu.org/gcc-4.0/buildstat.html 10183 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 10184 9. http://gcc.gnu.org/index.html 10185 10. mailto:gcc (a] gcc.gnu.org 10186 11. http://gcc.gnu.org/mirrors.html 10187 12. http://gcc.gnu.org/svn.html 10188 13. https://gcc.gnu.org/onlinedocs/ 10189 14. mailto:gcc-help (a] gcc.gnu.org 10190 15. mailto:gcc (a] gcc.gnu.org 10191 16. https://gcc.gnu.org/lists.html 10192 17. https://www.fsf.org/ 10193 18. https://gcc.gnu.org/about.html 10194 19. http://validator.w3.org/check/referer 10195 ====================================================================== 10196 http://gcc.gnu.org/gcc-4.0/changes.html 10197 10198 GCC 4.0 Release Series 10199 Changes, New Features, and Fixes 10200 10201 The latest release in the 4.0 release series is [1]GCC 4.0.4. 10202 10203 Caveats 10204 10205 * GCC now generates location lists by default when compiling with 10206 debug info and optimization. 10207 + GDB 6.0 and older crashes when it sees location lists. GDB 6.1 10208 or later is needed to debug binaries containing location 10209 lists. 10210 + When you are trying to view a value of a variable in a part of 10211 a function where it has no location (for example when the 10212 variable is no longer used and thus its location was used for 10213 something else) GDB will say that it is not available. 10214 You can disable generating location lists by -fno-var-tracking. 10215 * GCC no longer accepts the -fwritable-strings option. Use named 10216 character arrays when you need a writable string. 10217 * The options -freduce-all-givs and -fmove-all-movables have been 10218 discontinued. They were used to circumvent a shortcoming in the 10219 heuristics of the old loop optimization code with respect to common 10220 Fortran constructs. The new (tree) loop optimizer works differently 10221 and doesn't need those work-arounds. 10222 * The graph-coloring register allocator, formerly enabled by the 10223 option -fnew-ra, has been discontinued. 10224 * -I- has been deprecated. -iquote is meant to replace the need for 10225 this option. 10226 * The MIPS -membedded-pic and -mrnames options have been removed. 10227 * All MIPS targets now require the GNU assembler. In particular, IRIX 10228 configurations can no longer use the MIPSpro assemblers, although 10229 they do still support the MIPSpro linkers. 10230 * The SPARC option -mflat has been removed. 10231 * English-language diagnostic messages will now use Unicode quotation 10232 marks in UTF-8 locales. (Non-English messages already used the 10233 quotes appropriate for the language in previous releases.) If your 10234 terminal does not support UTF-8 but you are using a UTF-8 locale 10235 (such locales are the default on many GNU/Linux systems) then you 10236 should set LC_CTYPE=C in the environment to disable that locale. 10237 Programs that parse diagnostics and expect plain ASCII 10238 English-language messages should set LC_ALL=C. See [2]Markus Kuhn's 10239 explanation of Unicode quotation marks for more information. 10240 * The specs file is no longer installed on most platforms. Most users 10241 will be totally unaffected. However, if you are accustomed to 10242 editing the specs file yourself, you will now have to use the 10243 -dumpspecs option to generate the specs file, and then edit the 10244 resulting file. 10245 10246 General Optimizer Improvements 10247 10248 * The [3]tree ssa branch has been merged. This merge has brought in a 10249 completely new optimization framework based on a higher level 10250 intermediate representation than the existing RTL representation. 10251 Numerous new code transformations based on the new framework are 10252 available in GCC 4.0, including: 10253 + Scalar replacement of aggregates 10254 + Constant propagation 10255 + Value range propagation 10256 + Partial redundancy elimination 10257 + Load and store motion 10258 + Strength reduction 10259 + Dead store elimination 10260 + Dead and unreachable code elimination 10261 + [4]Autovectorization 10262 + Loop interchange 10263 + Tail recursion by accumulation 10264 Many of these passes outperform their counterparts from previous 10265 GCC releases. 10266 * [5]Swing Modulo Scheduling (SMS). An RTL level instruction 10267 scheduling optimization intended for loops that perform heavy 10268 computations. 10269 10270 New Languages and Language specific improvements 10271 10272 C family 10273 10274 * The sentinel attribute has been added to GCC. This function 10275 attribute allows GCC to warn when variadic functions such as execl 10276 are not NULL terminated. See the GCC manual for a complete 10277 description of its behavior. 10278 * Given __attribute__((alias("target"))) it is now an error if target 10279 is not a symbol, defined in the same translation unit. This also 10280 applies to aliases created by #pragma weak alias=target. This is 10281 because it's meaningless to define an alias to an undefined symbol. 10282 On Solaris, the native assembler would have caught this error, but 10283 GNU as does not. 10284 10285 C and Objective-C 10286 10287 * The -Wstrict-aliasing=2 option has been added. This warning catches 10288 all unsafe cases, but it may also give a warning for some cases 10289 that are safe. 10290 * The cast-as-lvalue, conditional-expression-as-lvalue and 10291 compound-expression-as-lvalue extensions, which were deprecated in 10292 3.3.4 and 3.4, have been removed. 10293 * The -fwritable-strings option, which was deprecated in 3.4, has 10294 been removed. 10295 * #pragma pack() semantics have been brought closer to those used by 10296 other compilers. This also applies to C++. 10297 * Taking the address of a variable with register storage is invalid 10298 in C. GCC now issues an error instead of a warning. 10299 * Arrays of incomplete element type are invalid in C. GCC now issues 10300 an error for such arrays. Declarations such as extern struct s x[]; 10301 (where struct s has not been defined) can be moved after the 10302 definition of struct s. Function parameters declared as arrays of 10303 incomplete type can instead be declared as pointers. 10304 10305 C++ 10306 10307 * When compiling without optimizations (-O0), the C++ front end is 10308 much faster than in any previous versions of GCC. Independent 10309 testers have measured speed-ups up to 25% in real-world production 10310 code, compared to the 3.4 family (which was already the fastest 10311 version to date). Upgrading from older versions might show even 10312 bigger improvements. 10313 * ELF visibility attributes can now be applied to a class type, so 10314 that it affects every member function of a class at once, without 10315 having to specify each individually: 10316 class __attribute__ ((visibility("hidden"))) Foo 10317 { 10318 int foo1(); 10319 void foo2(); 10320 }; 10321 The syntax is deliberately similar to the __declspec() system used 10322 by Microsoft Windows based compilers, allowing cross-platform 10323 projects to easily reuse their existing macro system for denoting 10324 exports and imports. By explicitly marking internal classes never 10325 used outside a binary as hidden, one can completely avoid PLT 10326 indirection overheads during their usage by the compiler. You can 10327 find out more about the advantages of this at 10328 [6]https://www.akkadia.org/drepper/dsohowto.pdf 10329 * The -fvisibility-inlines-hidden option has been added which marks 10330 all inlineable functions as having hidden ELF visibility, thus 10331 removing their symbol and typeinfo from the exported symbol table 10332 of the output ELF binary. Using this option can reduce the exported 10333 symbol count of template-heavy code by up to 40% with no code 10334 change at all, thus notably improving link and load times for the 10335 binary as well as a reduction in size of up to 10%. Also, check the 10336 new [7]-fvisibility option. 10337 * The compiler now uses the library interface specified by the [8]C++ 10338 ABI for thread-safe initialization of function-scope static 10339 variables. Most users should leave this alone, but embedded 10340 programmers may want to disable this by specifying 10341 -fno-threadsafe-statics for a small savings in code size. 10342 * Taking the address of an explicit register variable is no longer 10343 supported. Note that C++ allows taking the address of variables 10344 with register storage so this will continue to compile with a 10345 warning. For example, assuming that r0 is a machine register: 10346 register int foo asm ("r0"); 10347 register int bar; 10348 &foo; // error, no longer accepted 10349 &bar; // OK, with a warning 10350 * G++ has an undocumented extension to virtual function covariancy 10351 rules that allowed the overrider to return a type that was 10352 implicitly convertable to the overridden function's return type. 10353 For instance a function returning void * could be overridden by a 10354 function returning T *. This is now deprecated and will be removed 10355 in a future release. 10356 * The G++ minimum and maximum operators (<? and >?) and their 10357 compound forms (<?=) and >?=) have been deprecated and will be 10358 removed in a future version. Code using these operators should be 10359 modified to use std::min and std::max instead. 10360 * Declaration of nested classes of class templates as friends are 10361 supported: 10362 template <typename T> struct A { 10363 class B {}; 10364 }; 10365 class C { 10366 template <typename T> friend class A<T>::B; 10367 }; 10368 This complements the feature member functions of class templates as 10369 friends introduced in GCC 3.4.0. 10370 * When declaring a friend class using an unqualified name, classes 10371 outside the innermost non-class scope are not searched: 10372 class A; 10373 namespace N { 10374 class B { 10375 friend class A; // Refer to N::A which has not been declared yet 10376 // because name outside namespace N are not searched 10377 friend class ::A; // Refer to ::A 10378 }; 10379 } 10380 Hiding the friend name until declaration is still not implemented. 10381 * Friends of classes defined outside their namespace are correctly 10382 handled: 10383 namespace N { 10384 class A; 10385 } 10386 class N::A { 10387 friend class B; // Refer to N::B in GCC 4.0.0 10388 // but ::B in earlier versions of GCC 10389 }; 10390 10391 Runtime Library (libstdc++) 10392 10393 * Optimization work: 10394 + Added efficient specializations of istream functions for char 10395 and wchar_t. 10396 + Further performance tuning of strings, in particular wrt 10397 single-char append and getline. 10398 + iter_swap - and therefore most of the mutating algorithms - 10399 now makes an unqualified call to swap when the value_type of 10400 the two iterators is the same. 10401 * A large subset of the features in Technical Report 1 (TR1 for 10402 short) is experimentally delivered (i.e., no guarantees about the 10403 implementation are provided. In particular it is not promised that 10404 the library will remain link-compatible when code using TR1 is 10405 used): 10406 + General utilities such as reference_wrapper and shared_ptr. 10407 + Function objects, i.e., result_of, mem_fn, bind, function. 10408 + Support for metaprogramming. 10409 + New containers such as tuple, array, unordered_set, 10410 unordered_map, unordered_multiset, unordered_multimap. 10411 * As usual, many bugs have been fixed and LWG resolutions implemented 10412 for the first time (e.g., DR 409). 10413 10414 Java 10415 10416 * In order to prevent naming conflicts with other implementations of 10417 these tools, some GCJ binaries have been renamed: 10418 + rmic is now grmic, 10419 + rmiregistry is now grmiregistry, and 10420 + jar is now fastjar. 10421 In particular, these names were problematic for the jpackage.org 10422 packaging conventions which install symlinks in /usr/bin that point 10423 to the preferred versions of these tools. 10424 * The -findirect-dispatch argument to the compiler now works and 10425 generates code following a new "binary compatibility" ABI. Code 10426 compiled this way follows the binary compatibility rules of the 10427 Java Language Specification. 10428 * libgcj now has support for using GCJ as a JIT, using the 10429 gnu.gcj.jit family of system properties. 10430 * libgcj can now find a shared library corresponding to the bytecode 10431 representation of a class. See the documentation for the new 10432 gcj-dbtool program, and the new gnu.gcj.precompiled.db.path system 10433 property. 10434 * There have been many improvements to the class library. Here are 10435 some highlights: 10436 + Much more of AWT and Swing exist. 10437 + Many new packages and classes were added, including 10438 java.util.regex, java.net.URI, javax.crypto, 10439 javax.crypto.interfaces, javax.crypto.spec, javax.net, 10440 javax.net.ssl, javax.security.auth, 10441 javax.security.auth.callback, javax.security.auth.login, 10442 javax.security.auth.x500, javax.security.sasl, org.ietf.jgss, 10443 javax.imageio, javax.imageio.event, javax.imageio.spi, 10444 javax.print, javax.print.attribute, 10445 javax.print.attribute.standard, javax.print.event, and 10446 javax.xml 10447 + Updated SAX and DOM, and imported GNU JAXP 10448 10449 Fortran 10450 10451 * A new [9]Fortran front end has replaced the aging GNU Fortran 77 10452 front end. The new front end supports Fortran 90 and Fortran 95. It 10453 may not yet be as stable as the old Fortran front end. 10454 10455 Ada 10456 10457 * Ada (with tasking and Zero Cost Exceptions) is now available on 10458 many more targets, including but not limited to: alpha-linux, 10459 hppa-hpux, hppa-linux, powerpc-darwin, powerpc-linux, s390-linux, 10460 s390x-linux, sparc-linux. 10461 * Some of the new Ada 2005 features are now implemented like 10462 Wide_Wide_Character and Ada.Containers. 10463 * Many bugs have been fixed, tools and documentation improved. 10464 * To compile Ada from the sources, install an older working Ada 10465 compiler and then use --enable-languages=ada at configuration time, 10466 since the Ada front end is not currently activated by default. See 10467 the [10]Installing GCC for details. 10468 10469 New Targets and Target Specific Improvements 10470 10471 H8/300 10472 10473 * The frame layout has changed. In the new layout, the prologue of a 10474 function first saves registers and then allocate space for locals, 10475 resulting in an 1% improvement on code size. 10476 10477 IA-32/x86-64 (AMD64) 10478 10479 * The acos, asin, drem, exp10, exp2, expm1, fmod, ilogb, log10, 10480 log1p, log2, logb and tan mathematical builtins (and their float 10481 and long double variants) are now implemented as inline x87 10482 intrinsics when using -ffast-math. 10483 * The ceil, floor, nearbyint, rint and trunc mathematical builtins 10484 (and their float and long double variants) are now implemented as 10485 inline x87 intrinsics when using -ffast-math. 10486 * The x87's fsincos instruction is now used automatically with 10487 -ffast-math when calculating both the sin and cos of the same 10488 argument. 10489 * Instruction selection for multiplication and division by constants 10490 has been improved. 10491 10492 IA-64 10493 10494 * Floating point division, integer division and sqrt are now inlined, 10495 resulting in significant performance improvements on some codes. 10496 10497 MIPS 10498 10499 * Division by zero checks now use conditional traps if the target 10500 processor supports them. This decreases code size by one word per 10501 division operation. The old behavior (branch and break) can be 10502 obtained either at configure time by passing --with-divide=breaks 10503 to configure or at runtime by passing -mdivide-breaks to GCC. 10504 * Support for MIPS64 paired-single instructions has been added. It is 10505 enabled by -mpaired-single and can be accessed using both the 10506 target-independent vector extensions and new MIPS-specific built-in 10507 functions. 10508 * Support for the MIPS-3D ASE has been added. It is enabled by 10509 -mips3d and provides new MIPS-3D-specific built-in functions. 10510 * The -mexplicit-relocs option now supports static n64 code (as is 10511 used, for example, in 64-bit linux kernels). -mexplicit-relocs 10512 should now be feature-complete and is enabled by default when GCC 10513 is configured to use a compatible assembler. 10514 * Support for the NEC VR4130 series has been added. This support 10515 includes the use of VR-specific instructions and a new VR4130 10516 scheduler. Full VR4130 support can be selected with -march=vr4130 10517 while code for any ISA can be tuned for the VR4130 using 10518 -mtune=vr4130. There is also a new -mvr4130-align option that 10519 produces better schedules at the cost of increased code size. 10520 * Support for the Broadcom SB-1 has been extended. There is now an 10521 SB-1 scheduler as well as support for the SB-1-specific 10522 paired-single instructions. Full SB-1 support can be selected with 10523 -march=sb1 while code for any ISA can be optimized for the SB-1 10524 using -mtune=sb1. 10525 * The compiler can now work around errata in R4000, R4400, VR4120 and 10526 VR4130 processors. These workarounds are enabled by -mfix-r4000, 10527 -mfix-r4400, -mfix-vr4120 and -mfix-vr4130 respectively. The VR4120 10528 and VR4130 workarounds need binutils 2.16 or above. 10529 * IRIX shared libraries are now installed into the standard library 10530 directories: o32 libraries go into lib/, n32 libraries go into 10531 lib32/ and n64 libraries go into lib64/. 10532 * The compiler supports a new -msym32 option. It can be used to 10533 optimize n64 code in which all symbols are known to have 32-bit 10534 values. 10535 10536 S/390 and zSeries 10537 10538 * New command-line options help to generate code intended to run in 10539 an environment where stack space is restricted, e.g. Linux kernel 10540 code: 10541 + -mwarn-framesize and -mwarn-dynamicstack trigger compile-time 10542 warnings for single functions that require large or dynamic 10543 stack frames. 10544 + -mstack-size and -mstack-guard generate code that checks for 10545 stack overflow at run time. 10546 + -mpacked-stack generates code that reduces the stack frame 10547 size of many functions by reusing unneeded parts of the stack 10548 bias area. 10549 * The -msoft-float option now ensures that generated code never 10550 accesses floating point registers. 10551 * The s390x-ibm-tpf target now fully supports C++, including 10552 exceptions and threads. 10553 * Various changes to improve performance of the generated code have 10554 been implemented, including: 10555 + GCC now uses sibling calls where possible. 10556 + Condition code handling has been optimized, allowing GCC to 10557 omit redundant comparisons in certain cases. 10558 + The cost function guiding many optimizations has been refined 10559 to more accurately represent the z900 and z990 processors. 10560 + The ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL WITH BORROW 10561 instructions are now used to avoid conditional branches in 10562 certain cases. 10563 + The back end now uses the LEGITIMIZE_RELOAD_ADDRESS feature to 10564 optimize address arithmetic required to access large stack 10565 frames. 10566 + GCC now makes more efficient use of memory-to-memory type 10567 instructions (MVC, CLC, ...). 10568 + More precise tracking of special register use allows better 10569 instruction scheduling, in particular of the function prologue 10570 and epilogue sequences. 10571 + The Java front end now generates inline code to implement 10572 integer division, instead of calling library routines. 10573 10574 SPARC 10575 10576 * The options -mv8, -msparclite, -mcypress, -msupersparc, -mf930 and 10577 -mf934 have been removed. They have been replaced with -mcpu=xxx. 10578 * The internal model used to estimate the relative cost of each 10579 instruction has been updated. It is expected to give better results 10580 on recent UltraSPARC processors. 10581 * Code generation for function prologues and epilogues has been 10582 improved, resulting in better scheduling and allowing multiple exit 10583 points in functions. 10584 * Support for Sun's Visual Instruction Set (VIS) has been enhanced. 10585 It is enabled by -mvis and provides new built-in functions for VIS 10586 instructions on UltraSPARC processors. 10587 * The option -mapp-regs has been turned on by default on Solaris too. 10588 10589 NetWare 10590 10591 * Novell NetWare (on ix86, no other hardware platform was ever really 10592 supported by this OS) has been re-enabled and the ABI supported by 10593 GCC has been brought into sync with that of MetroWerks CodeWarrior 10594 (the ABI previously supported was that of some Unix systems, which 10595 NetWare never tried to support). 10596 10597 Obsolete Systems 10598 10599 Support for a number of older systems has been declared obsolete in GCC 10600 4.0. Unless there is activity to revive them, the next release of GCC 10601 will have their sources permanently removed. 10602 10603 All GCC ports for the following processor architectures have been 10604 declared obsolete: 10605 * Intel i860 10606 * Ubicom IP2022 10607 * National Semiconductor NS32K (ns32k) 10608 * Texas Instruments TMS320C[34]x 10609 10610 Also, those for some individual systems have been obsoleted: 10611 * SPARC family 10612 + SPARClite-based systems (sparclite-*-coff, sparclite-*-elf, 10613 sparc86x-*-elf) 10614 + OpenBSD 32-bit (sparc-*-openbsd*) 10615 10616 Documentation improvements 10617 10618 Other significant improvements 10619 10620 * Location lists are now generated by default when compiling with 10621 debug info and optimization. Location lists provide more accurate 10622 debug info about locations of variables and they allow debugging 10623 code compiled with -fomit-frame-pointer. 10624 * The -fvisibility option has been added which allows the default ELF 10625 visibility of all symbols to be set per compilation and the new 10626 #pragma GCC visibility preprocessor command allows the setting of 10627 default ELF visibility for a region of code. Using 10628 -fvisibility=hidden especially in combination with the new 10629 -fvisibility-inlines-hidden can yield substantial improvements in 10630 output binary quality including avoiding PLT indirection overheads, 10631 reduction of the exported symbol count by up to 60% (with resultant 10632 improvements to link and load times), better scope for the 10633 optimizer to improve code and up to a 20% reduction in binary size. 10634 Using these options correctly yields a binary with a similar symbol 10635 count to a Windows DLL. 10636 Perhaps more importantly, this new feature finally allows (with 10637 careful planning) complete avoidance of symbol clashes when 10638 manually loading shared objects with RTLD_GLOBAL, thus finally 10639 solving problems many projects such as python were forced to use 10640 RTLD_LOCAL for (with its resulting issues for C++ correctness). You 10641 can find more information about using these options at 10642 [11]https://gcc.gnu.org/wiki/Visibility. 10643 __________________________________________________________________ 10644 10645 GCC 4.0.1 10646 10647 This is the [12]list of problem reports (PRs) from GCC's bug tracking 10648 system that are known to be fixed in the 4.0.1 release. This list might 10649 not be complete (that is, it is possible that some PRs that have been 10650 fixed are not listed here). 10651 10652 GCC 4.0.2 10653 10654 This is the [13]list of problem reports (PRs) from GCC's bug tracking 10655 system that are known to be fixed in the 4.0.2 release. This list might 10656 not be complete (that is, it is possible that some PRs that have been 10657 fixed are not listed here). 10658 10659 Unfortunately, due to a release engineering failure, this release has a 10660 regression on Solaris that will affect some C++ programs. We suggest 10661 that Solaris users apply a [14]patch that corrects the problem. Users 10662 who do not wish to apply the patch should explicitly link C++ programs 10663 with the -pthreads option, even if they do not use threads. This 10664 problem has been corrected in the current 4.0 branch sources and will 10665 not be present in GCC 4.0.3. 10666 10667 GCC 4.0.3 10668 10669 Starting with this release, the function getcontext is recognized by 10670 the compiler as having the same semantics as the setjmp function. In 10671 particular, the compiler will ensure that all registers are dead before 10672 calling such a function and will emit a warning about the variables 10673 that may be clobbered after the second return from the function. 10674 10675 GCC 4.0.4 10676 10677 This is the [15]list of problem reports (PRs) from GCC's bug tracking 10678 system that are known to be fixed in the 4.0.4 release. This list might 10679 not be complete (that is, it is possible that some PRs that have been 10680 fixed are not listed here). 10681 10682 The 4.0.4 release is provided for those that require a high degree of 10683 binary compatibility with previous 4.0.x releases. For most users, the 10684 GCC team recommends that version 4.1.1 or later be used instead." 10685 10686 10687 For questions related to the use of GCC, please consult these web 10688 pages and the [16]GCC manuals. If that fails, the 10689 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10690 web pages and the development of GCC are welcome on our developer 10691 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 10692 archives. 10693 10694 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 10695 distribution of this entire article is permitted in any medium, 10696 provided this notice is preserved. 10697 10698 These pages are [21]maintained by the GCC team. Last modified 10699 2018-09-30[22]. 10700 10701 References 10702 10703 1. http://gcc.gnu.org/gcc-4.0/changes.html#4.0.4 10704 2. https://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html 10705 3. http://gcc.gnu.org/projects/tree-ssa/ 10706 4. http://gcc.gnu.org/projects/tree-ssa/vectorization.html 10707 5. http://gcc.gnu.org/news/sms.html 10708 6. https://www.akkadia.org/drepper/dsohowto.pdf 10709 7. http://gcc.gnu.org/gcc-4.0/changes.html#visibility 10710 8. https://itanium-cxx-abi.github.io/cxx-abi/ 10711 9. http://gcc.gnu.org/fortran/ 10712 10. https://gcc.gnu.org/install/ 10713 11. https://gcc.gnu.org/wiki/Visibility 10714 12. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.1 10715 13. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.2 10716 14. https://gcc.gnu.org/ml/gcc-cvs/2005-09/msg00984.html 10717 15. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=4.0.4 10718 16. https://gcc.gnu.org/onlinedocs/ 10719 17. mailto:gcc-help (a] gcc.gnu.org 10720 18. mailto:gcc (a] gcc.gnu.org 10721 19. https://gcc.gnu.org/lists.html 10722 20. https://www.fsf.org/ 10723 21. https://gcc.gnu.org/about.html 10724 22. http://validator.w3.org/check/referer 10725 ====================================================================== 10726 http://gcc.gnu.org/gcc-3.4/index.html 10727 10728 GCC 3.4 Release Series 10729 10730 May 26, 2006 10731 10732 The [1]GNU project and the GCC developers are pleased to announce the 10733 release of GCC 3.4.6. 10734 10735 This release is a bug-fix release, containing fixes for regressions in 10736 GCC 3.4.4 relative to previous releases of GCC. This is the last of the 10737 3.4.x series. 10738 10739 The GCC 3.4 release series includes numerous [2]new features, 10740 improvements, bug fixes, and other changes, thanks to an [3]amazing 10741 group of volunteers. 10742 10743 This release series is no longer maintained. 10744 10745 Release History 10746 10747 GCC 3.4.6 10748 March 6, 2006 ([4]changes) 10749 10750 GCC 3.4.5 10751 November 30, 2005 ([5]changes) 10752 10753 GCC 3.4.4 10754 May 18, 2005 ([6]changes) 10755 10756 GCC 3.4.3 10757 November 4, 2004 ([7]changes) 10758 10759 GCC 3.4.2 10760 September 6, 2004 ([8]changes) 10761 10762 GCC 3.4.1 10763 July 1, 2004 ([9]changes) 10764 10765 GCC 3.4.0 10766 April 18, 2004 ([10]changes) 10767 10768 References and Acknowledgements 10769 10770 GCC used to stand for the GNU C Compiler, but since the compiler 10771 supports several other languages aside from C, it now stands for the 10772 GNU Compiler Collection. 10773 10774 A list of [11]successful builds is updated as new information becomes 10775 available. 10776 10777 The GCC developers would like to thank the numerous people that have 10778 contributed new features, improvements, bug fixes, and other changes as 10779 well as test results to GCC. This [12]amazing group of volunteers is 10780 what makes GCC successful. 10781 10782 For additional information about GCC please refer to the [13]GCC 10783 project web site or contact the [14]GCC development mailing list. 10784 10785 To obtain GCC please use [15]our mirror sites, or [16]our SVN server. 10786 10787 10788 For questions related to the use of GCC, please consult these web 10789 pages and the [17]GCC manuals. If that fails, the 10790 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 10791 web pages and the development of GCC are welcome on our developer 10792 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 10793 archives. 10794 10795 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 10796 distribution of this entire article is permitted in any medium, 10797 provided this notice is preserved. 10798 10799 These pages are [22]maintained by the GCC team. Last modified 10800 2018-09-30[23]. 10801 10802 References 10803 10804 1. http://www.gnu.org/ 10805 2. http://gcc.gnu.org/gcc-3.4/changes.html 10806 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 10807 4. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 10808 5. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.5 10809 6. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.4 10810 7. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.3 10811 8. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.2 10812 9. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.1 10813 10. http://gcc.gnu.org/gcc-3.4/changes.html 10814 11. http://gcc.gnu.org/gcc-3.4/buildstat.html 10815 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 10816 13. http://gcc.gnu.org/index.html 10817 14. mailto:gcc (a] gcc.gnu.org 10818 15. http://gcc.gnu.org/mirrors.html 10819 16. http://gcc.gnu.org/svn.html 10820 17. https://gcc.gnu.org/onlinedocs/ 10821 18. mailto:gcc-help (a] gcc.gnu.org 10822 19. mailto:gcc (a] gcc.gnu.org 10823 20. https://gcc.gnu.org/lists.html 10824 21. https://www.fsf.org/ 10825 22. https://gcc.gnu.org/about.html 10826 23. http://validator.w3.org/check/referer 10827 ====================================================================== 10828 http://gcc.gnu.org/gcc-3.4/changes.html 10829 10830 GCC 3.4 Release Series 10831 Changes, New Features, and Fixes 10832 10833 The final release in the 3.4 release series is [1]GCC 3.4.6. The series 10834 is now closed. 10835 10836 GCC 3.4 has [2]many improvements in the C++ front end. Before reporting 10837 a bug, please make sure it's really GCC, and not your code, that is 10838 broken. 10839 10840 Caveats 10841 10842 * GNU Make is now required to build GCC. 10843 * With -nostdinc the preprocessor used to ignore both standard 10844 include paths and include paths contained in environment variables. 10845 It was neither documented nor intended that environment variable 10846 paths be ignored, so this has been corrected. 10847 * GCC no longer accepts the options -fvolatile, -fvolatile-global and 10848 -fvolatile-static. It is unlikely that they worked correctly in any 10849 3.x release. 10850 * GCC no longer ships <varargs.h>. Use <stdarg.h> instead. 10851 * Support for all the systems [3]obsoleted in GCC 3.3 has been 10852 removed from GCC 3.4. See below for a [4]list of systems which are 10853 obsoleted in this release. 10854 * GCC now requires an ISO C90 (ANSI C89) C compiler to build. K&R C 10855 compilers will not work. 10856 * The implementation of the [5]MIPS ABIs has changed. As a result, 10857 the code generated for certain MIPS targets will not be binary 10858 compatible with earlier releases. 10859 * In previous releases, the MIPS port had a fake "hilo" register with 10860 the user-visible name accum. This register has been removed. 10861 * The implementation of the [6]SPARC ABIs has changed. As a result, 10862 the code generated will not be binary compatible with earlier 10863 releases in certain cases. 10864 * The configure option --enable-threads=pthreads has been removed; 10865 use --enable-threads=posix instead, which should have the same 10866 effect. 10867 * Code size estimates used by inlining heuristics for C, Objective-C, 10868 C++ and Java have been redesigned significantly. As a result the 10869 parameters of -finline-insns, --param max-inline-insns-single and 10870 --param max-inline-insns-auto need to be reconsidered. 10871 * --param max-inline-slope and --param min-inline-insns have been 10872 removed; they are not needed for the new bottom-up inlining 10873 heuristics. 10874 * The new unit-at-a-time compilation scheme has several compatibility 10875 issues: 10876 + The order in which functions, variables, and top-level asm 10877 statements are emitted may have changed. Code relying on some 10878 particular ordering needs to be updated. The majority of such 10879 top-level asm statements can be replaced by section 10880 attributes. 10881 + Unreferenced static variables and functions are removed. This 10882 may result in undefined references when an asm statement 10883 refers to the variable/function directly. In that case either 10884 the variable/function shall be listed in asm statement operand 10885 or in the case of top-level asm statements the attribute used 10886 shall be used to force function/variable to be always output 10887 and considered as a possibly used by unknown code. 10888 For variables the attribute is accepted only by GCC 3.4 and 10889 newer, while for earlier versions it is sufficient to use 10890 unused to silence warnings about the variables not being 10891 referenced. To keep code portable across different GCC 10892 versions, you can use appropriate preprocessor conditionals. 10893 + Static functions now can use non-standard passing conventions 10894 that may break asm statements calling functions directly. 10895 Again the attribute used shall be used to prevent this 10896 behavior. 10897 As a temporary workaround, -fno-unit-at-a-time can be used, but 10898 this scheme may not be supported by future releases of GCC. 10899 * GCC 3.4 automatically places zero-initialized variables in the .bss 10900 section on some operating systems. Versions of GNU Emacs up to (and 10901 including) 21.3 will not work correctly when using this 10902 optimization; you can use -fno-zero-initialized-in-bss to disable 10903 it. 10904 * If GCC 3.4 is configured with --enable-threads=posix (the default 10905 on most targets that support pthreads) then _REENTRANT will be 10906 defined unconditionally by some libstdc++ headers. C++ code which 10907 relies on that macro to detect whether multi-threaded code is being 10908 compiled might change in meaning, possibly resulting in linker 10909 errors for single-threaded programs. Affected users of [7]Boost 10910 should compile single-threaded code with -DBOOST_DISABLE_THREADS. 10911 See Bugzilla for [8]more information. 10912 10913 General Optimizer Improvements 10914 10915 * Usability of the profile feedback and coverage testing has been 10916 improved. 10917 + Performance of profiled programs has been improved by faster 10918 profile merging code. 10919 + Better use of the profile feedback for optimization (loop 10920 unrolling and loop peeling). 10921 + File locking support allowing fork() calls and parallel runs 10922 of profiled programs. 10923 + Coverage file format has been redesigned. 10924 + gcov coverage tool has been improved. 10925 + make profiledbootstrap available to build a faster compiler. 10926 Experiments made on i386 hardware showed an 11% speedup on -O0 10927 and a 7.5% speedup on -O2 compilation of a [9]large C++ 10928 testcase. 10929 + New value profiling pass enabled via -fprofile-values 10930 + New value profile transformations pass enabled via -fvpt aims 10931 to optimize some code sequences by exploiting knowledge about 10932 value ranges or other properties of the operands. At the 10933 moment a conversion of expensive divisions into cheaper 10934 operations has been implemented. 10935 + New -fprofile-generate and -fprofile-use command-line options 10936 to simplify the use of profile feedback. 10937 * A new unit-at-a-time compilation scheme for C, Objective-C, C++ and 10938 Java which is enabled via -funit-at-a-time (and implied by -O2). In 10939 this scheme a whole file is parsed first and optimized later. The 10940 following basic inter-procedural optimizations are implemented: 10941 + Removal of unreachable functions and variables 10942 + Discovery of local functions (functions with static linkage 10943 whose address is never taken) 10944 + On i386, these local functions use register parameter passing 10945 conventions. 10946 + Reordering of functions in topological order of the call graph 10947 to enable better propagation of optimizing hints (such as the 10948 stack alignments needed by functions) in the back end. 10949 + Call graph based out-of-order inlining heuristics which allows 10950 to limit overall compilation unit growth (--param 10951 inline-unit-growth). 10952 Overall, the unit-at-a-time scheme produces a 1.3% improvement for 10953 the SPECint2000 benchmark on the i386 architecture (AMD Athlon 10954 CPU). 10955 * More realistic code size estimates used by inlining for C, 10956 Objective-C, C++ and Java. The growth of large functions can now be 10957 limited via --param large-function-insns and --param 10958 large-function-growth. 10959 * A new cfg-level loop optimizer pass replaces the old loop unrolling 10960 pass and adds two other loop transformations -- loop peeling and 10961 loop unswitching -- and also uses the profile feedback to limit 10962 code growth. (The three optimizations are enabled by 10963 -funroll-loops, -fpeel-loops and -funswitch-loops flags, 10964 respectively). 10965 The old loop unroller still can be enabled by -fold-unroll-loops 10966 and may produce better code in some cases, especially when the 10967 webizer optimization pass is not run. 10968 * A new web construction pass enabled via -fweb (and implied by -O3) 10969 improves the quality of register allocation, CSE, first scheduling 10970 pass and some other optimization passes by avoiding re-use of 10971 pseudo registers with non-overlapping live ranges. The pass almost 10972 always improves code quality but does make debugging difficult and 10973 thus is not enabled by default by -O2 10974 The pass is especially effective as cleanup after code duplication 10975 passes, such as the loop unroller or the tracer. 10976 * Experimental implementations of superblock or trace scheduling in 10977 the second scheduling pass can be enabled via 10978 -fsched2-use-superblocks and -fsched2-use-traces, respectively. 10979 10980 New Languages and Language specific improvements 10981 10982 Ada 10983 10984 * The Ada front end has been updated to include numerous bug fixes 10985 and enhancements. These include: 10986 + Improved project file support 10987 + Additional set of warnings about potential wrong code 10988 + Improved error messages 10989 + Improved code generation 10990 + Improved cross reference information 10991 + Improved inlining 10992 + Better run-time check elimination 10993 + Better error recovery 10994 + More efficient implementation of unbounded strings 10995 + Added features in GNAT.Sockets, GNAT.OS_Lib, GNAT.Debug_Pools, 10996 ... 10997 + New GNAT.xxxx packages (e.g. GNAT.Strings, 10998 GNAT.Exception_Action) 10999 + New pragmas 11000 + New -gnatS switch replacing gnatpsta 11001 + Implementation of new Ada features (in particular limited 11002 with, limited aggregates) 11003 11004 C/Objective-C/C++ 11005 11006 * Precompiled headers are now supported. Precompiled headers can 11007 dramatically speed up compilation of some projects. There are some 11008 known defects in the current precompiled header implementation that 11009 will result in compiler crashes in relatively rare situations. 11010 Therefore, precompiled headers should be considered a "technology 11011 preview" in this release. Read the manual for details about how to 11012 use precompiled headers. 11013 * File handling in the preprocessor has been rewritten. GCC no longer 11014 gets confused by symlinks and hardlinks, and now has a correct 11015 implementation of #import and #pragma once. These two directives 11016 have therefore been un-deprecated. 11017 * The undocumented extension that allowed C programs to have a label 11018 at the end of a compound statement, which has been deprecated since 11019 GCC 3.0, has been removed. 11020 * The cast-as-lvalue extension has been removed for C++ and 11021 deprecated for C and Objective-C. In particular, code like this: 11022 int i; 11023 (char) i = 5; 11024 11025 or this: 11026 char *p; 11027 ((int *) p)++; 11028 11029 is no longer accepted for C++ and will not be accepted for C and 11030 Objective-C in a future version. 11031 * The conditional-expression-as-lvalue extension has been deprecated 11032 for C and Objective-C. In particular, code like this: 11033 int a, b, c; 11034 (a ? b : c) = 2; 11035 11036 will not be accepted for C and Objective-C in a future version. 11037 * The compound-expression-as-lvalue extension has been deprecated for 11038 C and Objective-C. In particular, code like this: 11039 int a, b; 11040 (a, b) = 2; 11041 11042 will not be accepted for C and Objective-C in a future version. A 11043 possible non-intrusive workaround is the following: 11044 (*(a, &b)) = 2; 11045 11046 * Several [10]built-in functions such as __builtin_popcount for 11047 counting bits, finding the highest and lowest bit in a word, and 11048 parity have been added. 11049 * The -fwritable-strings option has been deprecated and will be 11050 removed. 11051 * Many C math library functions are now recognized as built-ins and 11052 optimized. 11053 * The C, C++, and Objective-C compilers can now handle source files 11054 written in any character encoding supported by the host C library. 11055 The default input character set is taken from the current locale, 11056 and may be overridden with the -finput-charset command line option. 11057 In the future we will add support for inline encoding markers. 11058 11059 C++ 11060 11061 * G++ is now much closer to full conformance to the ISO/ANSI C++ 11062 standard. This means, among other things, that a lot of invalid 11063 constructs which used to be accepted in previous versions will now 11064 be rejected. It is very likely that existing C++ code will need to 11065 be fixed. This document lists some of the most common issues. 11066 * A hand-written recursive-descent C++ parser has replaced the 11067 YACC-derived C++ parser from previous GCC releases. The new parser 11068 contains much improved infrastructure needed for better parsing of 11069 C++ source codes, handling of extensions, and clean separation 11070 (where possible) between proper semantics analysis and parsing. The 11071 new parser fixes many bugs that were found in the old parser. 11072 * You must now use the typename and template keywords to disambiguate 11073 dependent names, as required by the C++ standard. 11074 struct K { 11075 typedef int mytype_t; 11076 }; 11077 11078 template <class T1> struct A { 11079 template <class T2> struct B { 11080 void callme(void); 11081 }; 11082 11083 template <int N> void bar(void) 11084 { 11085 // Use 'typename' to tell the parser that T1::mytype_t names 11086 // a type. This is needed because the name is dependent (in 11087 // this case, on template parameter T1). 11088 typename T1::mytype_t x; 11089 x = 0; 11090 } 11091 }; 11092 11093 template <class T> void template_func(void) 11094 { 11095 // Use 'template' to prefix member templates within 11096 // dependent types (a has type A<T>, which depends on 11097 // the template parameter T). 11098 A<T> a; 11099 a.template bar<0>(); 11100 11101 // Use 'template' to tell the parser that B is a nested 11102 // template class (dependent on template parameter T), and 11103 // 'typename' because the whole A<T>::B<int> is 11104 // the name of a type (again, dependent). 11105 typename A<T>::template B<int> b; 11106 b.callme(); 11107 } 11108 11109 void non_template_func(void) 11110 { 11111 // Outside of any template class or function, no names can be 11112 // dependent, so the use of the keyword 'typename' and 'template' 11113 // is not needed (and actually forbidden). 11114 A<K> a; 11115 a.bar<0>(); 11116 A<K>::B<float> b; 11117 b.callme(); 11118 } 11119 * In a template definition, unqualified names will no longer find 11120 members of a dependent base (as specified by [temp.dep]/3 in the 11121 C++ standard). For example, 11122 template <typename T> struct B { 11123 int m; 11124 int n; 11125 int f (); 11126 int g (); 11127 }; 11128 int n; 11129 int g (); 11130 template <typename T> struct C : B<T> { 11131 void h () 11132 { 11133 m = 0; // error 11134 f (); // error 11135 n = 0; // ::n is modified 11136 g (); // ::g is called 11137 } 11138 }; 11139 You must make the names dependent, e.g. by prefixing them with 11140 this->. Here is the corrected definition of C<T>::h, 11141 template <typename T> void C<T>::h () 11142 { 11143 this->m = 0; 11144 this->f (); 11145 this->n = 0 11146 this->g (); 11147 } 11148 As an alternative solution (unfortunately not backwards compatible 11149 with GCC 3.3), you may use using declarations instead of this->: 11150 template <typename T> struct C : B<T> { 11151 using B<T>::m; 11152 using B<T>::f; 11153 using B<T>::n; 11154 using B<T>::g; 11155 void h () 11156 { 11157 m = 0; 11158 f (); 11159 n = 0; 11160 g (); 11161 } 11162 }; 11163 * In templates, all non-dependent names are now looked up and bound 11164 at definition time (while parsing the code), instead of later when 11165 the template is instantiated. For instance: 11166 void foo(int); 11167 11168 template <int> struct A { 11169 static void bar(void){ 11170 foo('a'); 11171 } 11172 }; 11173 11174 void foo(char); 11175 11176 int main() 11177 { 11178 A<0>::bar(); // Calls foo(int), used to call foo(char). 11179 } 11180 11181 * In an explicit instantiation of a class template, you must use 11182 class or struct before the template-id: 11183 template <int N> 11184 class A {}; 11185 11186 template A<0>; // error, not accepted anymore 11187 template class A<0>; // OK 11188 * The "named return value" and "implicit typename" extensions have 11189 been removed. 11190 * Default arguments in function types have been deprecated and will 11191 be removed. 11192 * ARM-style name-injection of friend declarations has been deprecated 11193 and will be removed. For example: struct S { friend void f(); }; 11194 void g() { f(); } will not be accepted by future versions of G++; 11195 instead a declaration of "f" will need to be present outside of the 11196 scope of "S". 11197 * Covariant returns are implemented for all but varadic functions 11198 that require an adjustment. 11199 * When -pedantic is used, G++ now issues errors about spurious 11200 semicolons. For example, 11201 namespace N {}; // Invalid semicolon. 11202 void f() {}; // Invalid semicolon. 11203 * G++ no longer accepts attributes for a declarator after the 11204 initializer associated with that declarator. For example, 11205 X x(1) __attribute__((...)); 11206 is no longer accepted. Instead, use: 11207 X x __attribute__((...)) (1); 11208 * Inside the scope of a template class, the name of the class itself 11209 can be treated as either a class or a template. So GCC used to 11210 accept the class name as argument of type template, and template 11211 template parameter. However this is not C++ standard compliant. Now 11212 the name is not treated as a valid template template argument 11213 unless you qualify the name by its scope. For example, the code 11214 below no longer compiles. 11215 template <template <class> class TT> class X {}; 11216 template <class T> class Y { 11217 X<Y> x; // Invalid, Y is always a type template parameter. 11218 }; 11219 The valid code for the above example is 11220 X< ::Y> x; // Valid. 11221 (Notice the space between < and : to prevent GCC to interpret this 11222 as a digraph for [.) 11223 * Friend declarations that refer to template specializations are 11224 rejected if the template has not already been declared. For 11225 example, 11226 template <typename T> 11227 class C { 11228 friend void f<> (C&); 11229 }; 11230 is rejected. You must first declare f as a template, 11231 template <typename T> 11232 void f(T); 11233 * In case of friend declarations, every name used in the friend 11234 declaration must be accessible at the point of that declaration. 11235 Previous versions of G++ used to be less strict about this and 11236 allowed friend declarations for private class members, for example. 11237 See the ISO C++ Standard Committee's [11]defect report #209 for 11238 details. 11239 * Declaration of member functions of class templates as friends are 11240 supported. For example, 11241 template <typename T> struct A { 11242 void f(); 11243 }; 11244 class C { 11245 template <typename T> friend void A<T>::f(); 11246 }; 11247 * You must use template <> to introduce template specializations, as 11248 required by the standard. For example, 11249 template <typename T> 11250 struct S; 11251 11252 struct S<int> { }; 11253 is rejected. You must write, 11254 template <> struct S<int> {}; 11255 * G++ used to accept code like this, 11256 struct S { 11257 int h(); 11258 void f(int i = g()); 11259 int g(int i = h()); 11260 }; 11261 This behavior is not mandated by the standard. Now G++ issues an 11262 error about this code. To avoid the error, you must move the 11263 declaration of g before the declaration of f. The default arguments 11264 for g must be visible at the point where it is called. 11265 * The C++ ABI Section 3.3.3 specifications for the array construction 11266 routines __cxa_vec_new2 and __cxa_vec_new3 were changed to return 11267 NULL when the allocator argument returns NULL. These changes are 11268 incorporated into the libstdc++ runtime library. 11269 * Using a name introduced by a typedef in a friend declaration or in 11270 an explicit instantiation is now rejected, as specified by the ISO 11271 C++ standard. 11272 class A; 11273 typedef A B; 11274 class C { 11275 friend class B; // error, no typedef name here 11276 friend B; // error, friend always needs class/struct/enum 11277 friend class A; // OK 11278 }; 11279 11280 template <int> class Q {}; 11281 typedef Q<0> R; 11282 template class R; // error, no typedef name here 11283 template class Q<0>; // OK 11284 * When allocating an array with a new expression, GCC used to allow 11285 parentheses around the type name. This is actually ill-formed and 11286 it is now rejected: 11287 int* a = new (int)[10]; // error, not accepted anymore 11288 int* a = new int[10]; // OK 11289 * When binding an rvalue of class type to a reference, the copy 11290 constructor of the class must be accessible. For instance, consider 11291 the following code: 11292 class A 11293 { 11294 public: 11295 A(); 11296 11297 private: 11298 A(const A&); // private copy ctor 11299 }; 11300 11301 A makeA(void); 11302 void foo(const A&); 11303 11304 void bar(void) 11305 { 11306 foo(A()); // error, copy ctor is not accessible 11307 foo(makeA()); // error, copy ctor is not accessible 11308 11309 A a1; 11310 foo(a1); // OK, a1 is a lvalue 11311 } 11312 This might be surprising at first sight, especially since most 11313 popular compilers do not correctly implement this rule ([12]further 11314 details). 11315 * When forming a pointer to member or a pointer to member function, 11316 access checks for class visibility (public, protected, private) are 11317 now performed using the qualifying scope of the name itself. This 11318 is better explained with an example: 11319 class A 11320 { 11321 public: 11322 void pub_func(); 11323 protected: 11324 void prot_func(); 11325 private: 11326 void priv_func(); 11327 }; 11328 11329 class B : public A 11330 { 11331 public: 11332 void foo() 11333 { 11334 &A::pub_func; // OK, pub_func is accessible through A 11335 &A::prot_func; // error, cannot access prot_func through A 11336 &A::priv_func; // error, cannot access priv_func through A 11337 11338 &B::pub_func; // OK, pub_func is accessible through B 11339 &B::prot_func; // OK, can access prot_func through B (within B) 11340 &B::priv_func; // error, cannot access priv_func through B 11341 } 11342 }; 11343 11344 Runtime Library (libstdc++) 11345 11346 * Optimization work: 11347 + Streamlined streambuf, filebuf, separate synched with C 11348 Standard I/O streambuf. 11349 + All formatted I/O now uses cached locale information. 11350 + STL optimizations (memory/speed for list, red-black trees as 11351 used by sets and maps). 11352 + More use of GCC builtins. 11353 + String optimizations (avoid contention on 11354 increment/decrement-and-test of the reference count in the 11355 empty-string object, constructor from input_iterators 11356 speedup). 11357 * Static linkage size reductions. 11358 * Large File Support (files larger than 2 GB on 32-bit systems). 11359 * Wide character and variable encoding filebuf work (UTF-8, Unicode). 11360 * Generic character traits. 11361 * Also support wchar_t specializations on Mac OS 10.3.x, FreeBSD 5.x, 11362 Solaris 2.7 and above, AIX 5.x, Irix 6.5. 11363 * The allocator class is now standard-conformant, and two additional 11364 extension allocators have been added, mt_alloc and 11365 bitmap_allocator. 11366 * PCH support: -include bits/stdc++.h (2x compile speedup). 11367 * Rewrote __cxa_demangle with support for C++ style allocators. 11368 * New debug modes for STL containers and iterators. 11369 * Testsuite rewrite: five times as many tests, plus increasingly 11370 sophisticated tests, including I/O, MT, multi-locale, wide and 11371 narrow characters. 11372 * Use current versions of GNU "autotools" for build/configuration. 11373 11374 Objective-C 11375 11376 * The Objective-C front end has been updated to include the numerous 11377 bug fixes and enhancements previously available only in Apple's 11378 version of GCC. These include: 11379 + Structured exception (@try... @catch... @finally, @throw) and 11380 synchronization (@synchronized) support. These are accessible 11381 via the -fobjc-exceptions switch; as of this writing, they may 11382 only be used in conjunction with -fnext-runtime on Mac OS X 11383 10.3 and later. See [13]Options Controlling Objective-C 11384 Dialect for more information. 11385 + An overhaul of @encode logic. The C99 _Bool and C++ bool type 11386 may now be encoded as 'B'. In addition, the back-end/codegen 11387 dependencies have been removed. 11388 + An overhaul of message dispatch construction, ensuring that 11389 the various receiver types (and casts thereof) are handled 11390 properly, and that correct diagnostics are issued. 11391 + Support for "Zero-Link" (-fzero-link) and "Fix-and-Continue" 11392 (-freplace-objc-classes) debugging modes, currently available 11393 on Mac OS X 10.3 and later. See [14]Options Controlling 11394 Objective-C Dialect for more information. 11395 + Access to optimized runtime entry points (-fno-nil-receivers ) 11396 on the assumption that message receivers are never nil. This 11397 is currently available on Mac OS X 10.3 and later. See 11398 [15]Options Controlling Objective-C Dialect for more 11399 information. 11400 11401 Java 11402 11403 * Compiling a .jar file will now cause non-.class entries to be 11404 automatically compiled as resources. 11405 * libgcj has been ported to Darwin. 11406 * Jeff Sturm has adapted Jan Hubicka's call graph optimization code 11407 to gcj. 11408 * libgcj has a new gcjlib URL type; this lets URLClassLoader load 11409 code from shared libraries. 11410 * libgcj has been much more completely merged with [16]GNU Classpath. 11411 * Class loading is now much more correct; in particular the caller's 11412 class loader is now used when that is required. 11413 * [17]Eclipse 2.x will run out of the box using gij. 11414 * Parts of java.nio have been implemented. Direct and indirect 11415 buffers work, as do fundamental file and socket operations. 11416 * java.awt has been improved, though it is still not ready for 11417 general use. 11418 * The HTTP protocol handler now uses HTTP/1.1 and can handle the POST 11419 method. 11420 * The MinGW port has matured. Enhancements include socket timeout 11421 support, thread interruption, improved Runtime.exec() handling and 11422 support for accented characters in filenames. 11423 11424 Fortran 11425 11426 * Fortran improvements are listed in the [18]Fortran documentation. 11427 11428 New Targets and Target Specific Improvements 11429 11430 Alpha 11431 11432 * Several [19]built-in functions have been added such as 11433 __builtin_alpha_zap to allow utilizing the more obscure 11434 instructions of the CPU. 11435 * Parameter passing of complex arguments has changed to match the 11436 ABI. This change is incompatible with previous GCC versions, but 11437 does fix compatibility with the Tru64 compiler and several corner 11438 cases where GCC was incompatible with itself. 11439 11440 ARM 11441 11442 * Nicolas Pitre has contributed his hand-coded floating-point support 11443 code for ARM. It is both significantly smaller and faster than the 11444 existing C-based implementation, even when building applications 11445 for Thumb. The arm-elf configuration has been converted to use the 11446 new code. 11447 * Support for the Intel's iWMMXt architecture, a second generation 11448 XScale processor, has been added. Enabled at run time with the 11449 -mcpu=iwmmxt command line switch. 11450 * A new ARM target has been added: arm-wince-pe. This is similar to 11451 the arm-pe target, but it defaults to using the APCS32 ABI. 11452 * The existing ARM pipeline description has been converted to the use 11453 the [20]DFA processor pipeline model. There is not much change in 11454 code performance, but the description is now [21]easier to 11455 understand. 11456 * Support for the Cirrus EP9312 Maverick floating point co-processor 11457 added. Enabled at run time with the -mcpu=ep9312 command line 11458 switch. Note however that the multilibs to support this chip are 11459 currently disabled in gcc/config/arm/t-arm-elf, so if you want to 11460 enable their production you will have to uncomment the entries in 11461 that file. 11462 11463 H8/300 11464 11465 * Support for long long has been added. 11466 * Support for saveall attribute has been added. 11467 * Pavel Pisa contributed hand-written 32-bit-by-32-bit division code 11468 for H8/300H and H8S, which is much faster than the previous 11469 implementation. 11470 * A lot of small performance improvements. 11471 11472 IA-32/AMD64 (x86-64) 11473 11474 * Tuning for K8 (AMD Opteron/Athlon64) core is available via 11475 -march=k8 and -mcpu=k8. 11476 * Scalar SSE code generation carefully avoids reformatting penalties, 11477 hidden dependencies and minimizes the number of uops generated on 11478 both Intel and AMD CPUs. 11479 * Vector MMX and SSE operands are now passed in registers to improve 11480 performance and match the argument passing convention used by the 11481 Intel C++ Compiler. As a result it is not possible to call 11482 functions accepting vector arguments compiled by older GCC version. 11483 * Conditional jump elimination is now more aggressive on modern CPUs. 11484 * The Athlon ports has been converted to use the DFA processor 11485 pipeline description. 11486 * Optimization of indirect tail calls is now possible in a similar 11487 fashion as direct sibcall optimization. 11488 * Further small performance improvements. 11489 * -m128bit-long-double is now less buggy. 11490 * __float128 support in 64-bit compilation. 11491 * Support for data structures exceeding 2GB in 64-bit mode. 11492 * -mcpu has been renamed to -mtune. 11493 11494 IA-64 11495 11496 * Tuning code for the Itanium 2 processor has been added. The 11497 generation of code tuned for Itanium 2 (option -mtune=itanium2) is 11498 enabled by default now. To generate code tuned for Itanium 1 the 11499 option -mtune=itanium1 should be used. 11500 * [22]DFA processor pipeline descriptions for the IA-64 processors 11501 have been added. This resulted in about 3% improvement on the 11502 SPECInt2000 benchmark for Itanium 2. 11503 * Instruction bundling for the IA-64 processors has been rewritten 11504 using the DFA pipeline hazard recognizer. It resulted in about 60% 11505 compiler speedup on the SPECInt2000 C programs. 11506 11507 M32R 11508 11509 * Support for the M32R/2 processor has been added by Renesas. 11510 * Support for an M32R GNU/Linux target and PIC code generation has 11511 been added by Renesas. 11512 11513 M68000 11514 11515 * Bernardo Innocenti (Develer S.r.l.) has contributed the 11516 m68k-uclinux target, based on former work done by Paul Dale 11517 (SnapGear Inc.). Code generation for the ColdFire processors family 11518 has been enhanced and extended to support the MCF 53xx and MCF 54xx 11519 cores, integrating former work done by Peter Barada (Motorola). 11520 11521 MIPS 11522 11523 Processor-specific changes 11524 11525 * Support for the RM7000 and RM9000 processors has been added. It can 11526 be selected using the -march compiler option and should work with 11527 any MIPS I (mips-*) or MIPS III (mips64-*) configuration. 11528 * Support for revision 2 of the MIPS32 ISA has been added. It can be 11529 selected with the command-line option -march=mips32r2. 11530 * There is a new option, -mfix-sb1, to work around certain SB-1 11531 errata. 11532 11533 Configuration 11534 11535 * It is possible to customize GCC using the following configure-time 11536 options: 11537 + --with-arch, which specifies the default value of the -march 11538 option. 11539 + --with-tune, which specifies the default value of the -mtune 11540 option. 11541 + --with-abi, which specifies the default ABI. 11542 + --with-float=soft, which tells GCC to use software floating 11543 point by default. 11544 + --with-float=hard, which tells GCC to use hardware floating 11545 point by default. 11546 * A 64-bit GNU/Linux port has been added. The associated 11547 configurations are mips64-linux-gnu and mips64el-linux-gnu. 11548 * The 32-bit GNU/Linux port now supports Java. 11549 * The IRIX 6 configuration now supports the o32 ABI and will build 11550 o32 multilibs by default. This support is compatible with both 11551 binutils and the SGI tools, but note that several features, 11552 including debugging information and DWARF2 exception handling, are 11553 only available when using the GNU assembler. Use of the GNU 11554 assembler and linker (version 2.15 or above) is strongly 11555 recommended. 11556 * The IRIX 6 configuration now supports 128-bit long doubles. 11557 * There are two new RTEMS-specific configurations, mips-rtems and 11558 mipsel-rtems. 11559 * There are two new *-elf configurations, mipsisa32r2-elf and 11560 mipsisa32r2el-elf. 11561 11562 General 11563 11564 * Several [23]ABI bugs have been fixed. Unfortunately, these changes 11565 will break binary compatibility with earlier releases. 11566 * GCC can now use explicit relocation operators when generating 11567 -mabicalls code. This behavior is controlled by -mexplicit-relocs 11568 and can have several performance benefits. For example: 11569 + It allows for more optimization of GOT accesses, including 11570 better scheduling and redundancy elimination. 11571 + It allows sibling calls to be implemented as jumps. 11572 + n32 and n64 leaf functions can use a call-clobbered global 11573 pointer instead of $28. 11574 + The code to set up $gp can be removed from functions that 11575 don't need it. 11576 * A new option, -mxgot, allows the GOT to be bigger than 64k. This 11577 option is equivalent to the assembler's -xgot option and should be 11578 used instead of -Wa,-xgot. 11579 * Frame pointer elimination is now supported when generating 64-bit 11580 MIPS16 code. 11581 * Inline block moves have been optimized to take more account of 11582 alignment information. 11583 * Many internal changes have been made to the MIPS port, mostly aimed 11584 at reducing the reliance on assembler macros. 11585 11586 PowerPC 11587 11588 * GCC 3.4 releases have a number of fixes for PowerPC and PowerPC64 11589 [24]ABI incompatibilities regarding the way parameters are passed 11590 during functions calls. These changes may result in incompatibility 11591 between code compiled with GCC 3.3 and GCC 3.4. 11592 11593 PowerPC Darwin 11594 11595 * Support for shared/dylib gcc libraries has been added. It is 11596 enabled by default on powerpc-apple-darwin7.0.0 and up. 11597 * Libgcj is enabled by default. On systems older than 11598 powerpc-apple-darwin7.0.0 you need to install dlcompat. 11599 * 128-bit IBM extended precision format support added for long 11600 double. 11601 11602 PowerPC64 GNU/Linux 11603 11604 * By default, PowerPC64 GNU/Linux now uses natural alignment of 11605 structure elements. The old four byte alignment for double, with 11606 special rules for a struct starting with a double, can be chosen 11607 with -malign-power. This change may result in incompatibility 11608 between code compiled with GCC 3.3 and GCC 3.4. 11609 * -mabi=altivec is now the default rather than -mabi=no-altivec. 11610 * 128-bit IBM extended precision format support added for long 11611 double. 11612 11613 S/390 and zSeries 11614 11615 * New command-line options allow to specify the intended execution 11616 environment for generated code: 11617 + -mesa/-mzarch allows to specify whether to generate code 11618 running in ESA/390 mode or in z/Architecture mode (this is 11619 applicable to 31-bit code only). 11620 + -march allows to specify a minimum processor architecture 11621 level (g5, g6, z900, or z990). 11622 + -mtune allows to specify which processor to tune for. 11623 * It is possible to customize GCC using the following configure-time 11624 options: 11625 + --with-mode, which specifies whether to default to assuming 11626 ESA/390 or z/Architecture mode. 11627 + --with-arch, which specifies the default value of the -march 11628 option. 11629 + --with-tune, which specifies the default value of the -mtune 11630 option. 11631 * Support for the z990 processor has been added, and can be selected 11632 using -march=z990 or -mtune=z990. This includes instruction 11633 scheduling tuned for the superscalar instruction pipeline of the 11634 z990 processor as well as support for all new instructions provided 11635 by the long-displacement facility. 11636 * Support to generate 31-bit code optimized for zSeries processors 11637 (running in ESA/390 or in z/Architecture mode) has been added. This 11638 can be selected using -march=z900 and -mzarch respectively. 11639 * Instruction scheduling for the z900 and z990 processors now uses 11640 the DFA pipeline hazard recognizer. 11641 * GCC no longer generates code to maintain a stack backchain, 11642 previously used to generate stack backtraces for debugging 11643 purposes. As replacement that does not incur runtime overhead, 11644 DWARF-2 call frame information is provided by GCC; this is 11645 supported by GDB 6.1. The old behavior can be restored using the 11646 -mbackchain option. 11647 * The stack frame size of functions may now exceed 2 GB in 64-bit 11648 code. 11649 * A port for the 64-bit IBM TPF operating system has been added; the 11650 configuration is s390x-ibm-tpf. This configuration is supported as 11651 cross-compilation target only. 11652 * Various changes to improve the generated code have been 11653 implemented, including: 11654 + GCC now uses the MULTIPLY AND ADD and MULTIPLY AND SUBTRACT 11655 instructions to significantly speed up many floating-point 11656 applications. 11657 + GCC now uses the ADD LOGICAL WITH CARRY and SUBTRACT LOGICAL 11658 WITH BORROW instructions to speed up long long arithmetic. 11659 + GCC now uses the SEARCH STRING instruction to implement 11660 strlen(). 11661 + In many cases, function call overhead for 31-bit code has been 11662 reduced by placing the literal pool after the function code 11663 instead of after the function prolog. 11664 + Register 14 is no longer reserved in 64-bit code. 11665 + Handling of global register variables has been improved. 11666 11667 SPARC 11668 11669 * The option -mflat is deprecated. 11670 * Support for large (> 2GB) frames has been added to the 64-bit port. 11671 * Several [25]ABI bugs have been fixed. Unfortunately, these changes 11672 will break binary compatibility with earlier releases. 11673 * The default debugging format has been switched from STABS to 11674 DWARF-2 for 32-bit code on Solaris 7 and later. DWARF-2 is already 11675 the default debugging format for 64-bit code on Solaris. 11676 11677 SuperH 11678 11679 * Support for the SH2E processor has been added. Enabled at run time 11680 with the -m2e command line switch, or at configure time by 11681 specifying sh2e as the machine part of the target triple. 11682 11683 V850 11684 11685 * Support for the Mitsubishi V850E1 processor has been added. This is 11686 a variant of the V850E processor with some additional debugging 11687 instructions. 11688 11689 Xtensa 11690 11691 * Several ABI bugs have been fixed. Unfortunately, these changes 11692 break binary compatibility with earlier releases. 11693 + For big-endian processors, the padding of aggregate return 11694 values larger than a word has changed. If the size of an 11695 aggregate return value is not a multiple of 32 bits, previous 11696 versions of GCC inserted padding in the most-significant bytes 11697 of the first return value register. Aggregates larger than a 11698 word are now padded in the least-significant bytes of the last 11699 return value register used. Aggregates smaller than a word are 11700 still padded in the most-significant bytes. The return value 11701 padding has not changed for little-endian processors. 11702 + Function arguments with 16-byte alignment are now properly 11703 aligned. 11704 + The implementation of the va_list type has changed. A va_list 11705 value created by va_start from a previous release cannot be 11706 used with va_arg from this release, or vice versa. 11707 * More processor configuration options for Xtensa processors are 11708 supported: 11709 + the ABS instruction is now optional; 11710 + the ADDX* and SUBX* instructions are now optional; 11711 + an experimental CONST16 instruction can be used to synthesize 11712 constants instead of loading them from constant pools. 11713 These and other Xtensa processor configuration options can no 11714 longer be enabled or disabled by command-line options; the 11715 processor configuration must be specified by the xtensa-config.h 11716 header file when building GCC. Additionally, the 11717 -mno-serialize-volatile option is no longer supported. 11718 11719 Obsolete Systems 11720 11721 Support for a number of older systems has been declared obsolete in GCC 11722 3.4. Unless there is activity to revive them, the next release of GCC 11723 will have their sources permanently removed. 11724 11725 All configurations of the following processor architectures have been 11726 declared obsolete: 11727 * Mitsubishi D30V, d30v-* 11728 * AT&T DSP1600 and DSP1610, dsp16xx-* 11729 * Intel 80960, i960 11730 11731 Also, some individual systems have been obsoleted: 11732 * ARM Family 11733 + Support for generating code for operation in APCS/26 mode 11734 (-mapcs-26). 11735 * IBM ESA/390 11736 + "Bigfoot" port, i370-*. (The other port, s390-*, is actively 11737 maintained and supported.) 11738 * Intel 386 family 11739 + MOSS, i?86-moss-msdos and i?86-*-moss* 11740 + NCR 3000 running System V r.4, i?86-ncr-sysv4* 11741 + FreeBSD with a.out object format, i?86-*-freebsd*aout* and 11742 i?86-*-freebsd2* 11743 + GNU/Linux with a.out object format, i?86-linux*aout* 11744 + GNU/Linux with libc5, a.k.a. glibc1, i?86-linux*libc1* 11745 + Interix versions before Interix 3, i?86-*-interix 11746 + Mach microkernel, i?86-mach* 11747 + SCO UnixWare with UDK, i?86-*-udk* 11748 + Generic System V releases 1, 2, and 3, i?86-*-sysv[123]* 11749 + VSTa microkernel, i386-*-vsta 11750 * Motorola M68000 family 11751 + HPUX, m68k-hp-hpux* and m68000-hp-hpux* 11752 + NetBSD with a.out object format (before NetBSD 1.4), 11753 m68k-*-*-netbsd* except m68k-*-*-netbsdelf* 11754 + Generic System V r.4, m68k-*-sysv4* 11755 * VAX 11756 + Generic VAX, vax-*-* (This is generic VAX only; we have not 11757 obsoleted any VAX triples for specific operating systems.) 11758 11759 Documentation improvements 11760 11761 Other significant improvements 11762 11763 * The build system has undergone several significant cleanups. 11764 Subdirectories will only be configured if they are being built, and 11765 all subdirectory configures are run from the make command. The top 11766 level has been autoconfiscated. 11767 * Building GCC no longer writes to its source directory. This should 11768 help those wishing to share a read-only source directory over NFS 11769 or build from a CD. The exceptions to this feature are if you 11770 configure with either --enable-maintainer-mode or 11771 --enable-generated-files-in-srcdir. 11772 * The -W warning option has been renamed to -Wextra, which is more 11773 easily understood. The older spelling will be retained for 11774 backwards compatibility. 11775 * Substantial improvements in compile time have been made, 11776 particularly for non-optimizing compilations. 11777 __________________________________________________________________ 11778 11779 GCC 3.4.0 11780 11781 Bug Fixes 11782 11783 A vast number of bugs have been fixed in 3.4.0, too many to publish a 11784 complete list here. [26]Follow this link to query the Bugzilla database 11785 for the list of over 900 bugs fixed in 3.4.0. This is the list of all 11786 bugs marked as resolved and fixed in 3.4.0 that are not flagged as 3.4 11787 regressions. 11788 __________________________________________________________________ 11789 11790 GCC 3.4.1 11791 11792 Bug Fixes 11793 11794 This section lists the problem reports (PRs) from GCC's bug tracking 11795 system that are known to be fixed in the 3.4.1 release. This list might 11796 not be complete (that is, it is possible that some PRs that have been 11797 fixed are not listed here). 11798 11799 Bootstrap failures 11800 11801 * [27]10129 Ada bootstrap fails on PPC-Darwin - invalid assembler 11802 emitted - PIC related 11803 * [28]14576 [ARM] ICE in libiberty when building gcc-3.4 for arm-elf 11804 * [29]14760 A bug in configure.in prevents using both 11805 --program-suffix and --program-prefix 11806 * [30]14671 [hppa64] bootstrap fails: ICE in 11807 save_call_clobbered_regs, in caller_save.c 11808 * [31]15093 [alpha][Java] make bootstrap fails to configure libffi on 11809 Alpha 11810 * [32]15178 Solaris 9/x86 fails linking after stage 3 11811 11812 Multi-platform internal compiler errors (ICEs) 11813 11814 * [33]12753 (preprocessor) Memory corruption in preprocessor on bad 11815 input 11816 * [34]13985 ICE in gcc.c-torture/compile/930621-1.c 11817 * [35]14810 (c++) tree check failures with invalid code involving 11818 templates 11819 * [36]14883 (c++) ICE on invalid code, in cp_parser_lookup_name, in 11820 cp/parser.c 11821 * [37]15044 (c++) ICE on syntax error, template header 11822 * [38]15057 (c++) Compiling of conditional value throw constructs 11823 cause a segmentation violation 11824 * [39]15064 (c++) typeid of template parameter gives ICE 11825 * [40]15142 (c++) ICE when passing a string where a char* is expected 11826 in a throw statement 11827 * [41]15159 ICE in rtl_verify_flow_info_1 11828 * [42]15165 (c++) ICE in instantiate_template 11829 * [43]15193 Unary minus using pointer to V4SF vector causes 11830 -fforce-mem to exhaust all memory 11831 * [44]15209 (c++) Runs out of memory with packed structs 11832 * [45]15227 (c++) Trouble with invalid function definition 11833 * [46]15285 (c++) instantiate_type ICE when forming pointer to 11834 template function 11835 * [47]15299 (c++) ICE in resolve_overloaded_unification 11836 * [48]15329 (c++) ICE on constructor of member template 11837 * [49]15550 ICE in extract_insn, in recog.c 11838 * [50]15554 (c++) ICE in tsubst_copy, in cp/pt.c 11839 * [51]15640 (c++) ICE on invalid code in arg_assoc, in 11840 cp/name-lookup.c 11841 * [52]15666 [unit-at-a-time] Gcc abort on valid code 11842 * [53]15696 (c++) ICE with bad pointer-to-member code 11843 * [54]15701 (c++) ICE with friends and template template parameter 11844 * [55]15761 ICE in do_SUBST, in combine.c 11845 * [56]15829 (c++) ICE on Botan-1.3.13 due to -funroll-loops 11846 11847 Ada 11848 11849 * [57]14538 All RTEMS targets broken for gnat 11850 11851 C front end 11852 11853 * [58]12391 missing warning about assigning to an incomplete type 11854 * [59]14649 atan(1.0) should not be a constant expression 11855 * [60]15004 [unit-at-a-time] no warning for unused paramater in 11856 static function 11857 * [61]15749 --pedantic-errors behaves differently from --pedantic 11858 with C-compiler on GNU/Linux 11859 11860 C++ compiler and library 11861 11862 * [62]10646 non-const reference is incorrectly matched in a "const T" 11863 partial specialization 11864 * [63]12077 wcin.rdbuf()->in_avail() return value too high 11865 * [64]13598 enc_filebuf doesn't work 11866 * [65]14211 const_cast returns lvalue but should be rvalue 11867 * [66]14220 num_put::do_put() undesired float/double behavior 11868 * [67]14245 problem with user-defined allocators in std::basic_string 11869 * [68]14340 libstdc++ Debug mode: failure to convert iterator to 11870 const_iterator 11871 * [69]14600 __gnu_cxx::stdio_sync_filebuf should expose internal 11872 FILE* 11873 * [70]14668 no warning anymore for reevaluation of declaration 11874 * [71]14775 LFS (large file support) tests missing 11875 * [72]14821 Duplicate namespace alias declaration should not conflict 11876 * [73]14930 Friend declaration ignored 11877 * [74]14932 cannot use offsetof to get offsets of array elements in 11878 g++ 3.4.0 11879 * [75]14950 [non unit-at-a-time] always_inline does not mix with 11880 templates and -O0 11881 * [76]14962 g++ ignores #pragma redefine_extname 11882 * [77]14975 Segfault on low-level write error during imbue 11883 * [78]15002 Linewise stream input is unusably slow (std::string slow) 11884 * [79]15025 compiler accepts redeclaration of template as 11885 non-template 11886 * [80]15046 [arm] Math functions misdetected by cross configuration 11887 * [81]15069 a bit test on a variable of enum type is miscompiled 11888 * [82]15074 g++ -lsupc++ still links against libstdc++ 11889 * [83]15083 spurious "statement has no effect" warning 11890 * [84]15096 parse error with templates and pointer to const member 11891 * [85]15287 combination of operator[] and operator .* fails in 11892 templates 11893 * [86]15317 __attribute__ unused in first parameter of constructor 11894 gives error 11895 * [87]15337 sizeof on incomplete type diagnostic 11896 * [88]15361 bitset<>::_Find_next fails 11897 * [89]15412 _GLIBCXX_ symbols symbols defined and used in different 11898 namespaces 11899 * [90]15427 valid code results in incomplete type error 11900 * [91]15471 Incorrect member pointer offsets in anonymous 11901 structs/unions 11902 * [92]15503 nested template problem 11903 * [93]15507 compiler hangs while laying out union 11904 * [94]15542 operator & and template definitions 11905 * [95]15565 SLES9: leading + sign for unsigned int with showpos 11906 * [96]15625 friend defined inside a template fails to find static 11907 function 11908 * [97]15629 Function templates, overloads, and friend name injection 11909 * [98]15742 'noreturn' attribute ignored in method of template 11910 functions. 11911 * [99]15775 Allocator::pointer consistently ignored 11912 * [100]15821 Duplicate namespace alias within namespace rejected 11913 * [101]15862 'enum yn' fails (confict with undeclared builtin) 11914 * [102]15875 rejects pointer to member in template 11915 * [103]15877 valid code using templates and anonymous enums is 11916 rejected 11917 * [104]15947 Puzzling error message for wrong destructor declaration 11918 in template class 11919 * [105]16020 cannot copy __gnu_debug::bitset 11920 * [106]16154 input iterator concept too restrictive 11921 * [107]16174 deducing top-level consts 11922 11923 Java 11924 11925 * [108]14315 Java compiler is not parallel make safe 11926 11927 Fortran 11928 11929 * [109]15151 [g77] incorrect logical i/o in 64-bit mode 11930 11931 Objective-C 11932 11933 * [110]7993 private variables cannot be shadowed in subclasses 11934 11935 Optimization bugs 11936 11937 * [111]15228 useless copies of floating point operands 11938 * [112]15345 [non-unit-at-a-time] unreferenced nested inline 11939 functions not optimized away 11940 * [113]15945 Incorrect floating point optimization 11941 * [114]15526 ftrapv aborts on 0 * (-1) 11942 * [115]14690 Miscompiled POOMA tests 11943 * [116]15112 GCC generates code to write to unchanging memory 11944 11945 Preprocessor 11946 11947 * [117]15067 Minor glitch in the source of cpp 11948 11949 Main driver program bugs 11950 11951 * [118]1963 collect2 interprets -oldstyle_liblookup as -o 11952 ldstyle_liblookup 11953 11954 x86-specific (Intel/AMD) 11955 11956 * [119]15717 Error: can't resolve `L0' {*ABS* section} - `xx' {*UND* 11957 section} 11958 11959 HPPA-specific 11960 11961 * [120]14782 GCC produces an unaligned data access at -O2 11962 * [121]14828 FAIL: gcc.c-torture/execute/20030408-1.c execution, -O2 11963 * [122]15202 ICE in reload_cse_simplify_operands, in postreload.c 11964 11965 IA64-specific 11966 11967 * [123]14610 __float80 constants incorrectly emitted 11968 * [124]14813 init_array sections are initialized in the wrong order 11969 * [125]14857 GCC segfault on duplicated asm statement 11970 * [126]15598 Gcc 3.4 ICE on valid code 11971 * [127]15653 Gcc 3.4 ICE on valid code 11972 11973 MIPS-specific 11974 11975 * [128]15189 wrong filling of delay slot with -march=mips1 -G0 11976 -mno-split-addresses -mno-explicit-relocs 11977 * [129]15331 Assembler error building gnatlib on IRIX 6.5 with GNU as 11978 2.14.91 11979 * [130]16144 Bogus reference to __divdf3 when -O1 11980 * [131]16176 Miscompilation of unaligned data in MIPS backend 11981 11982 PowerPC-specific 11983 11984 * [132]11591 ICE in gcc.dg/altivec-5.c 11985 * [133]12028 powerpc-eabispe produces bad sCOND operation 11986 * [134]14478 rs6000 geu/ltu patterns generate incorrect code 11987 * [135]14567 long double and va_arg complex args 11988 * [136]14715 Altivec stack layout may overlap gpr save with stack 11989 temps 11990 * [137]14902 (libstdc++) Stream checking functions fail when -pthread 11991 option is used. 11992 * [138]14924 Compiler ICE on valid code 11993 * [139]14960 -maltivec affects vector return with -mabi=no-altivec 11994 * [140]15106 vector varargs failure passing from altivec to 11995 non-altivec code for -m32 11996 * [141]16026 ICE in function.c:4804, assign_parms, when -mpowerpc64 & 11997 half-word operation 11998 * [142]15191 -maltivec -mabi=no-altivec results in mis-aligned lvx 11999 and stvx 12000 * [143]15662 Segmentation fault when an exception is thrown - even if 12001 try and catch are specified 12002 12003 s390-specific 12004 12005 * [144]15054 Bad code due to overlapping stack temporaries 12006 12007 SPARC-specific 12008 12009 * [145]15783 ICE with union assignment in 64-bit mode 12010 * [146]15626 GCC 3.4 emits "ld: warning: relocation error: 12011 R_SPARC_UA32" 12012 12013 x86-64-specific 12014 12015 * [147]14326 boehm-gc hardcodes to 3DNow! prefetch for x86_64 12016 * [148]14723 Backported -march=nocona from mainline 12017 * [149]15290 __float128 failed to pass to function properly 12018 12019 Cygwin/Mingw32-specific 12020 12021 * [150]15250 Option -mms-bitfields support on GCC 3.4 is not 12022 conformant to MS layout 12023 * [151]15551 -mtune=pentium4 -O2 with sjlj EH breaks stack probe 12024 worker on windows32 targets 12025 12026 Bugs specific to embedded processors 12027 12028 * [152]8309 [m68k] -m5200 produces erroneous SImode set of short 12029 varaible on stack 12030 * [153]13250 [SH] Gcc code for rotation clobbers the register, but 12031 gcc continues to use the register as if it was not clobbered 12032 * [154]13803 [coldfire] movqi operand constraints too restrictivefor 12033 TARGET_COLDFIRE 12034 * [155]14093 [SH] ICE for code when using -mhitachi option in SH 12035 * [156]14457 [m6811hc] ICE with simple c++ source 12036 * [157]14542 [m6811hc] ICE on simple source 12037 * [158]15100 [SH] cc1plus got hang-up on 12038 libstdc++-v3/testsuite/abi_check.cc 12039 * [159]15296 [CRIS] Delayed branch scheduling causing invalid code on 12040 cris-* 12041 * [160]15396 [SH] ICE with -O2 -fPIC 12042 * [161]15782 [coldfire] m68k_output_mi_thunk emits wrong code for 12043 ColdFire 12044 12045 Testsuite problems (compiler not affected) 12046 12047 * [162]11610 libstdc++ testcases 27_io/* don't work properly remotely 12048 * [163]15488 (libstdc++) possibly insufficient file permissions for 12049 executing test suite 12050 * [164]15489 (libstdc++) testsuite_files determined incorrectly 12051 12052 Documentation bugs 12053 12054 * [165]13928 (libstdc++) no whatis info in some man pages generated 12055 by doxygen 12056 * [166]14150 Ada documentation out of date 12057 * [167]14949 (c++) Need to document method visibility changes 12058 * [168]15123 libstdc++-doc: Allocators.3 manpage is empty 12059 __________________________________________________________________ 12060 12061 GCC 3.4.2 12062 12063 Bug Fixes 12064 12065 This section lists the problem reports (PRs) from GCC's bug tracking 12066 system that are known to be fixed in the 3.4.2 release. This list might 12067 not be complete (that is, it is possible that some PRs that have been 12068 fixed are not listed here). 12069 12070 Bootstrap failures and issues 12071 12072 * [169]16469 [mips-sgi-irix5.3] bootstrap fails in 12073 libstdc++-v3/testsuite 12074 * [170]16344 [hppa-linux-gnu] libstdc++'s PCH built by 12075 profiledbootstrap does not work with the built compiler 12076 * [171]16842 [Solaris/x86] mkheaders can not find mkheaders.conf 12077 12078 Multi-platform internal compiler errors (ICEs) 12079 12080 * [172]12608 (c++) ICE: expected class 't', have 'x' (error_mark) in 12081 cp_parser_class_specifier, in cp/parser.c 12082 * [173]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 12083 * [174]15461 (c++) ICE due to NRV and inlining 12084 * [175]15890 (c++) ICE in c_expand_expr, in c-common.c 12085 * [176]16180 ICE: segmentation fault in RTL optimization 12086 * [177]16224 (c++) ICE in write_unscoped_name (template/namespace) 12087 * [178]16408 ICE: in delete_insn, in cfgrtl.c 12088 * [179]16529 (c++) ICE for: namespace-alias shall not be declared as 12089 the name of any other entity 12090 * [180]16698 (c++) ICE with exceptions and declaration of __cxa_throw 12091 * [181]16706 (c++) ICE in finish_member_declaration, in 12092 cp/semantics.c 12093 * [182]16810 (c++) Legal C++ program with cast gives ICE in 12094 build_ptrmemfunc 12095 * [183]16851 (c++) ICE when throwing a comma expression 12096 * [184]16870 (c++) Boost.Spirit causes ICE in tsubst, in cp/pt.c 12097 * [185]16904 (c++) ICE in finish_class_member_access_expr, in 12098 cp/typeck.c 12099 * [186]16905 (c++) ICE (segfault) with exceptions 12100 * [187]16964 (c++) ICE in cp_parser_class_specifier due to 12101 redefinition 12102 * [188]17068 (c++) ICE: tree check: expected class 'd', have 'x' 12103 (identifier_node) in dependent_template_p, in cp/pt.c 12104 12105 Preprocessor bugs 12106 12107 * [189]16366 Preprocessor option -remap causes memory corruption 12108 12109 Optimization 12110 12111 * [190]15345 unreferenced nested inline functions not optimized away 12112 * [191]16590 Incorrect execution when compiling with -O2 12113 * [192]16693 Bitwise AND is lost when used within a cast to an enum 12114 of the same precision 12115 * [193]17078 Jump into if(0) substatement fails 12116 12117 Problems in generated debug information 12118 12119 * [194]13956 incorrect stabs for nested local variables 12120 12121 C front end bugs 12122 12123 * [195]16684 GCC should not warn about redundant redeclarations of 12124 built-ins 12125 12126 C++ compiler and library 12127 12128 * [196]12658 Thread safety problems in locale::global() and 12129 locale::locale() 12130 * [197]13092 g++ accepts invalid pointer-to-member conversion 12131 * [198]15320 Excessive memory consumption 12132 * [199]16246 Incorrect template argument deduction 12133 * [200]16273 Memory exhausted when using nested classes and virtual 12134 functions 12135 * [201]16401 ostringstream in gcc 3.4.x very slow for big data 12136 * [202]16411 undefined reference to 12137 __gnu_cxx::stdio_sync_filebuf<char, std::char_traits<char> 12138 >::file() 12139 * [203]16489 G++ incorrectly rejects use of a null constant integral 12140 expression as a null constant pointer 12141 * [204]16618 offsetof fails with constant member 12142 * [205]16637 syntax error reported for valid input code 12143 * [206]16717 __attribute__((constructor)) broken in C++ 12144 * [207]16813 compiler error in DEBUG version of range insertion 12145 std::map::insert 12146 * [208]16853 pointer-to-member initialization from incompatible one 12147 accepted 12148 * [209]16889 ambiguity is not detected 12149 * [210]16959 Segmentation fault in ios_base::sync_with_stdio 12150 12151 Java compiler and library 12152 12153 * [211]7587 direct threaded interpreter not thread-safe 12154 * [212]16473 ServerSocket accept() leaks file descriptors 12155 * [213]16478 Hash synchronization deadlock with finalizers 12156 12157 Alpha-specific 12158 12159 * [214]10695 ICE in dwarf2out_frame_debug_expr, in dwarf2out.c 12160 * [215]16974 could not split insn (ice in final_scan_insn, in 12161 final.c) 12162 12163 x86-specific 12164 12165 * [216]16298 ICE in output_operand 12166 * [217]17113 ICE with SSE2 intrinsics 12167 12168 x86-64 specific 12169 12170 * [218]14697 libstdc++ couldn't find 32bit libgcc_s 12171 12172 MIPS-specific 12173 12174 * [219]15869 [mips64] No NOP after LW (with -mips1 -O0) 12175 * [220]16325 [mips64] value profiling clobbers gp on mips 12176 * [221]16357 [mipsisa64-elf] ICE copying 7 bytes between extern 12177 char[]s 12178 * [222]16380 [mips64] Use of uninitialised register after dbra 12179 conversion 12180 * [223]16407 [mips64] Unaligned access to local variables 12181 * [224]16643 [mips64] verify_local_live_at_start ICE after 12182 crossjumping & cfgcleanup 12183 12184 ARM-specific 12185 12186 * [225]15927 THUMB -O2: strength-reduced iteration variable ends up 12187 off by 1 12188 * [226]15948 THUMB: ICE with non-commutative cbranch 12189 * [227]17019 THUMB: bad switch statement in md code for 12190 addsi3_cbranch_scratch 12191 12192 IA64-specific 12193 12194 * [228]16130 ICE on valid code: in bundling, in config/ia64/ia64.c 12195 (-mtune=merced) 12196 * [229]16142 ICE on valid code: in bundling, in config/ia64/ia64.c 12197 (-mtune=itanium) 12198 * [230]16278 Gcc failed to build Linux kernel with -mtune=merced 12199 * [231]16414 ICE on valid code: typo in comparison of asm_noperands 12200 result 12201 * [232]16445 ICE on valid code: don't count ignored insns 12202 * [233]16490 ICE (segfault) while compiling with -fprofile-use 12203 * [234]16683 ia64 does not honor SUBTARGET_EXTRA_SPECS 12204 12205 PowerPC-specific 12206 12207 * [235]16195 (ppc64): Miscompilation of GCC 3.3.x by 3.4.x 12208 * [236]16239 ICE on ppc64 (mozilla 1.7 compile, -O1 -fno-exceptions 12209 issue) 12210 12211 SPARC-specific 12212 12213 * [237]16199 ICE while compiling apache 2.0.49 12214 * [238]16416 -m64 doesn't imply -mcpu=v9 anymore 12215 * [239]16430 ICE when returning non-C aggregates larger than 16 bytes 12216 12217 Bugs specific to embedded processors 12218 12219 * [240]16379 [m32r] can't output large model function call of memcpy 12220 * [241]17093 [m32r] ICE with -msdata=use -O0 12221 * [242]17119 [m32r] ICE at switch case 0x8000 12222 12223 DJGPP-specific 12224 12225 * [243]15928 libstdc++ in 3.4.x doesn't cross-compile for djgpp 12226 12227 Alpha Tru64-specific 12228 12229 * [244]16210 libstdc++ gratuitously omits "long long" I/O 12230 12231 Testsuite, documentation issues (compiler is not affected): 12232 12233 * [245]15488 (libstdc++) possibly insufficient file permissions for 12234 executing test suite 12235 * [246]16250 ada/doctools runs makeinfo even in release tarball 12236 __________________________________________________________________ 12237 12238 GCC 3.4.3 12239 12240 This is the [247]list of problem reports (PRs) from GCC's bug tracking 12241 system that are known to be fixed in the 3.4.3 release. This list might 12242 not be complete (that is, it is possible that some PRs that have been 12243 fixed are not listed here). 12244 12245 Bootstrap failures 12246 12247 * [248]17369 [ia64] Bootstrap failure with binutils-2.15.90.0.1.1 12248 * [249]17850 [arm-elf] bootstrap failure - libstdc++ uses strtold 12249 when undeclared 12250 12251 Internal compiler errors (ICEs) affecting multiple platforms 12252 12253 * [250]13948 (java) GCJ segmentation fault while compiling GL4Java 12254 .class files 12255 * [251]14492 ICE in loc_descriptor_from_tree, in dwarf2out.c 12256 * [252]16301 (c++) ICE when "strong" attribute is attached to a using 12257 directive 12258 * [253]16566 ICE with flexible arrays 12259 * [254]17023 ICE with nested functions in parameter declaration 12260 * [255]17027 ICE with noreturn function in loop at -O2 12261 * [256]17524 ICE in grokdeclarator, in cp/decl.c 12262 * [257]17826 (c++) ICE in cp_tree_equal 12263 12264 C and optimization bugs 12265 12266 * [258]15526 -ftrapv aborts on 0 * (-1) 12267 * [259]16999 #ident stopped working 12268 * [260]17503 quadratic behaviour in invalid_mode_change_p 12269 * [261]17581 Long long arithmetic fails inside a switch/case 12270 statement when compiled with -O2 12271 * [262]18129 -fwritable-strings doesn't work 12272 12273 C++ compiler and library bugs 12274 12275 * [263]10975 incorrect initial ostringstream::tellp() 12276 * [264]11722 Unbuffered filebuf::sgetn is slow 12277 * [265]14534 Unrecognizing static function as a template parameter 12278 when its return value is also templated 12279 * [266]15172 Copy constructor optimization in aggregate 12280 initialization 12281 * [267]15786 Bad error message for frequently occuring error. 12282 * [268]16162 Rejects valid member-template-definition 12283 * [269]16612 empty basic_strings can't live in shared memory 12284 * [270]16715 std::basic_iostream is instantiated when used, even 12285 though instantiations are already contained in libstdc++ 12286 * [271]16848 code in /ext/demangle.h appears broken 12287 * [272]17132 GCC fails to eliminate function template specialization 12288 when argument deduction fails 12289 * [273]17259 One more _S_leaf incorrectly qualified with _RopeRep:: 12290 in ropeimpl.h 12291 * [274]17327 use of `enumeral_type' in template type unification 12292 * [275]17393 "unused variable '._0'" warning with -Wall 12293 * [276]17501 Confusion with member templates 12294 * [277]17537 g++ not passing -lstdc++ to linker when all command line 12295 arguments are libraries 12296 * [278]17585 usage of unqualified name of static member from within 12297 class not allowed 12298 * [279]17821 Poor diagnostic for using "." instead of "->" 12299 * [280]17829 wrong error: call of overloaded function is ambiguous 12300 * [281]17851 Misleading diagnostic for invalid function declarations 12301 with undeclared types 12302 * [282]17976 Destructor is called twice 12303 * [283]18020 rejects valid definition of enum value in template 12304 * [284]18093 bogus conflict in namespace aliasing 12305 * [285]18140 C++ parser bug when using >> in templates 12306 12307 Fortran 12308 12309 * [286]17541 data statements with double precision constants fail 12310 12311 x86-specific 12312 12313 * [287]17853 -O2 ICE for MMX testcase 12314 12315 SPARC-specific 12316 12317 * [288]17245 ICE compiling gsl-1.5 statistics/lag1.c 12318 12319 Darwin-specific 12320 12321 * [289]17167 FATAL:Symbol L_foo$stub already defined. 12322 12323 AIX-specific 12324 12325 * [290]17277 could not catch an exception when specified -maix64 12326 12327 Solaris-specific 12328 12329 * [291]17505 <cmath> calls acosf(), ceilf(), and other functions 12330 missing from system libraries 12331 12332 HP/UX specific: 12333 12334 * [292]17684 /usr/ccs/bin/ld: Can't create libgcc_s.sl 12335 12336 ARM-specific 12337 12338 * [293]17384 ICE with mode attribute on structures 12339 12340 MIPS-specific 12341 12342 * [294]17770 No NOP after LWL with -mips1 12343 12344 Other embedded target specific 12345 12346 * [295]11476 [arc-elf] gcc ICE on newlib's vfprintf.c 12347 * [296]14064 [avr-elf] -fdata-sections triggers ICE 12348 * [297]14678 [m68hc11-elf] gcc ICE 12349 * [298]15583 [powerpc-rtems] powerpc-rtems lacks __USE_INIT_FINI__ 12350 * [299]15790 [i686-coff] Alignment error building gcc with i686-coff 12351 target 12352 * [300]15886 [SH] Miscompilation with -O2 -fPIC 12353 * [301]16884 [avr-elf] [fweb related] bug while initializing 12354 variables 12355 12356 Bugs relating to debugger support 12357 12358 * [302]13841 missing debug info for _Complex function arguments 12359 * [303]15860 [big-endian targets] No DW_AT_location debug info is 12360 emitted for formal arguments to a function that uses "register" 12361 qualifiers 12362 12363 Testsuite issues (compiler not affected) 12364 12365 * [304]17465 Testsuite in libffi overrides LD_LIBRARY_PATH 12366 * [305]17469 Testsuite in libstdc++ overrides LD_LIBRARY_PATH 12367 * [306]18138 [mips-sgi-irix6.5] libgcc_s.so.1 not found by 64-bit 12368 testsuite 12369 12370 Documentation 12371 12372 * [307]15498 typo in gcc manual: non-existing locale example en_UK, 12373 should be en_GB 12374 * [308]15747 [mips-sgi-irix5.3] /bin/sh hangs during bootstrap: 12375 document broken shell 12376 * [309]16406 USE_LD_AS_NEEDED undocumented 12377 __________________________________________________________________ 12378 12379 GCC 3.4.4 12380 12381 This is the [310]list of problem reports (PRs) from GCC's bug tracking 12382 system that are known to be fixed in the 3.4.4 release. This list might 12383 not be complete (that is, it is possible that some PRs that have been 12384 fixed are not listed here). 12385 __________________________________________________________________ 12386 12387 GCC 3.4.5 12388 12389 This is the [311]list of problem reports (PRs) from GCC's bug tracking 12390 system that are known to be fixed in the 3.4.5 release. This list might 12391 not be complete (that is, it is possible that some PRs that have been 12392 fixed are not listed here). 12393 12394 Bootstrap issues 12395 12396 * [312]24688 sco_math fixincl breaks math.h 12397 12398 C compiler bugs 12399 12400 * [313]17188 struct Foo { } redefinition 12401 * [314]20187 wrong code for ((unsigned char)(unsigned long 12402 long)((a?a:1)&(a*b)))?0:1) 12403 * [315]21873 infinite warning loop on bad array initializer 12404 * [316]21899 enum definition accepts values to be overriden 12405 * [317]22061 ICE in find_function_data, in function.c 12406 * [318]22308 Failure to diagnose violation of constraint 6.516p2 12407 * [319]22458 ICE on missing brace 12408 * [320]22589 ICE casting to long long 12409 * [321]24101 Segfault with preprocessed source 12410 12411 C++ compiler and library bugs 12412 12413 * [322]10611 operations on vector mode not recognized in C++ 12414 * [323]13377 unexpected behavior of namespace usage directive 12415 * [324]16002 Strange error message with new parser 12416 * [325]17413 local classes as template argument 12417 * [326]17609 spurious error message after using keyword 12418 * [327]17618 ICE in cp_convert_to_pointer, in cp/cvt.c 12419 * [328]18124 ICE with invalid template template parameter 12420 * [329]18155 typedef in template declaration not rejected 12421 * [330]18177 ICE with const_cast for undeclared variable 12422 * [331]18368 C++ error message regression 12423 * [332]16378 ICE when returning a copy of a packed member 12424 * [333]18466 int ::i; accepted 12425 * [334]18512 ICE on invalid usage of template base class 12426 * [335]18454 ICE when returning undefined type 12427 * [336]18738 typename not allowed with non-dependent qualified name 12428 * [337]18803 rejects access to operator() in template 12429 * [338]19004 ICE in uses_template_parms, in cp/pt.c 12430 * [339]19208 Spurious error about variably modified type 12431 * [340]18253 bad error message / ICE for invalid template parameter 12432 * [341]19608 ICE after friend function definition in local class 12433 * [342]19884 ICE on explicit instantiation of a non-template 12434 constructor 12435 * [343]20153 ICE when C++ template function contains anonymous union 12436 * [344]20563 Infinite loop in diagnostic (and ice after error 12437 message) 12438 * [345]20789 ICE with incomplete type in template 12439 * [346]21336 Internal compiler error when using custom new operators 12440 * [347]21768 ICE in error message due to violation of coding 12441 conventions 12442 * [348]21853 constness of pointer to data member ignored 12443 * [349]21903 Default argument of template function causes a 12444 compile-time error 12445 * [350]21983 multiple diagnostics 12446 * [351]21987 New testsuite failure 12447 g++.dg/warn/conversion-function-1.C 12448 * [352]22153 ICE on invalid template specialization 12449 * [353]22172 Internal compiler error, seg fault. 12450 * [354]21286 filebuf::xsgetn vs pipes 12451 * [355]22233 ICE with wrong number of template parameters 12452 * [356]22508 ICE after invalid operator new 12453 * [357]22545 ICE with pointer to class member & user defined 12454 conversion operator 12455 * [358]23528 Wrong default allocator in ext/hash_map 12456 * [359]23550 char_traits requirements/1.cc test bad math 12457 * [360]23586 Bad diagnostic for invalid namespace-name 12458 * [361]23624 ICE in invert_truthvalue, in fold-const.c 12459 * [362]23639 Bad error message: not a member of '<declaration error>' 12460 * [363]23797 ICE on typename outside template 12461 * [364]23965 Bogus error message: no matching function for call to 12462 'foo(<type error>)' 12463 * [365]24052 &#`label_decl' not supported by dump_expr#<expression 12464 error> 12465 * [366]24580 virtual base class cause exception not to be caught 12466 12467 Problems in generated debug information 12468 12469 * [367]24267 Bad DWARF for altivec vectors 12470 12471 Optimizations issues 12472 12473 * [368]17810 ICE in verify_local_live_at_start 12474 * [369]17860 Wrong generated code for loop with varying bound 12475 * [370]21709 ICE on compile-time complex NaN 12476 * [371]21964 broken tail call at -O2 or more 12477 * [372]22167 Strange optimization bug when using -Os 12478 * [373]22619 Compilation failure for real_const_1.f and 12479 real_const_2.f90 12480 * [374]23241 Invalid code generated for comparison of uchar to 255 12481 * [375]23478 Miscompilation due to reloading of a var that is also 12482 used in EH pad 12483 * [376]24470 segmentation fault in cc1plus when compiling with -O 12484 * [377]24950 ICE in operand_subword_force 12485 12486 Precompiled headers problems 12487 12488 * [378]14400 Cannot compile qt-x11-free-3.3.0 12489 * [379]14940 PCH largefile test fails on various platforms 12490 12491 Preprocessor bugs 12492 12493 * [380]20239 ICE on empty preprocessed input 12494 * [381]15220 "gcc -E -MM -MG" reports missing system headers in 12495 source directory 12496 12497 Testsuite issues 12498 12499 * [382]19275 gcc.dg/20020919-1.c fails with -fpic/-fPIC on 12500 i686-pc-linux-gnu 12501 12502 Alpha specific 12503 12504 * [383]21888 bootstrap failure with linker relaxation enabled 12505 12506 ARM specific 12507 12508 * [384]15342 [arm-linux]: ICE in verify_local_live_at_start 12509 * [385]23985 Memory aliasing information incorrect in inlined memcpy 12510 12511 ColdFile specific 12512 12513 * [386]16719 Illegal move of byte into address register causes 12514 compiler to ICE 12515 12516 HPPA specific 12517 12518 * [387]21723 ICE while building libgfortran 12519 * [388]21841 -mhp-ld/-mgnu-ld documentation 12520 12521 IA-64 specific 12522 12523 * [389]23644 IA-64 hardware models and configuration options 12524 documentation error 12525 * [390]24718 Shared libgcc not used for linking by default 12526 12527 M68000 specific 12528 12529 * [391]18421 ICE in reload_cse_simplify_operands, in postreload.c 12530 12531 MIPS specific 12532 12533 * [392]20621 ICE in change_address_1, in emit-rtl.c 12534 12535 PowerPC and PowerPC64 specific 12536 12537 * [393]18583 error on valid code: const 12538 __attribute__((altivec(vector__))) doesn't work in arrays 12539 * [394]20191 ICE in reload_cse_simplify_operands 12540 * [395]22083 AIX: TARGET_C99_FUNCTIONS is wrongly defined 12541 * [396]23070 CALL_V4_CLEAR_FP_ARGS flag not properly set 12542 * [397]23404 gij trashes args of functions with more than 8 fp args 12543 * [398]23539 C & C++ compiler generating misaligned references 12544 regardless of compiler flags 12545 * [399]24102 floatdisf2_internal2 broken 12546 * [400]24465 -mminimal-toc miscompilation of __thread vars 12547 12548 Solaris specific 12549 12550 * [401]19933 Problem with define of HUGE_VAL in math_c99 12551 * [402]21889 Native Solaris assembler cannot grok DTP-relative debug 12552 symbols 12553 12554 SPARC specific 12555 12556 * [403]19300 PCH failures on sparc-linux 12557 * [404]20301 Assembler labels have a leading "-" 12558 * [405]20673 C PCH testsuite assembly comparison failure 12559 12560 x86 and x86_64 specific 12561 12562 * [406]18582 ICE with arrays of type V2DF 12563 * [407]19340 Compilation SEGFAULTs with -O1 -fschedule-insns2 12564 -fsched2-use-traces 12565 * [408]21716 ICE in reg-stack.c's swap_rtx_condition 12566 * [409]24315 amd64 fails -fpeephole2 12567 __________________________________________________________________ 12568 12569 GCC 3.4.6 12570 12571 This is the [410]list of problem reports (PRs) from GCC's bug tracking 12572 system that are known to be fixed in the 3.4.6 release. This list might 12573 not be complete (that is, it is possible that some PRs that have been 12574 fixed are not listed here). 12575 12576 12577 For questions related to the use of GCC, please consult these web 12578 pages and the [411]GCC manuals. If that fails, the 12579 [412]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 12580 web pages and the development of GCC are welcome on our developer 12581 list at [413]gcc (a] gcc.gnu.org. All of [414]our lists have public 12582 archives. 12583 12584 Copyright (C) [415]Free Software Foundation, Inc. Verbatim copying and 12585 distribution of this entire article is permitted in any medium, 12586 provided this notice is preserved. 12587 12588 These pages are [416]maintained by the GCC team. Last modified 12589 2018-09-30[417]. 12590 12591 References 12592 12593 1. http://gcc.gnu.org/gcc-3.4/changes.html#3.4.6 12594 2. http://gcc.gnu.org/gcc-3.4/changes.html#cplusplus 12595 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 12596 4. http://gcc.gnu.org/gcc-3.4/changes.html#obsolete_systems 12597 5. http://gcc.gnu.org/gcc-3.4/mips-abi.html 12598 6. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 12599 7. https://www.boost.org/ 12600 8. https://gcc.gnu.org/PR11953 12601 9. https://gcc.gnu.org/PR8361 12602 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Other-Builtins.html#Other%20Builtins 12603 11. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_closed.html#209 12604 12. http://gcc.gnu.org/bugs/#cxx_rvalbind 12605 13. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 12606 14. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 12607 15. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Objective-C-Dialect-Options.html 12608 16. http://www.gnu.org/software/classpath/ 12609 17. http://www.eclipse.org/ 12610 18. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/g77/News.html 12611 19. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gcc/Alpha-Built-in-Functions.html 12612 20. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 12613 21. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Comparison-of-the-two-descriptions.html 12614 22. https://gcc.gnu.org/onlinedocs/gcc-3.4.3/gccint/Processor-pipeline-description.html 12615 23. http://gcc.gnu.org/gcc-3.4/mips-abi.html 12616 24. http://gcc.gnu.org/gcc-3.4/powerpc-abi.html 12617 25. http://gcc.gnu.org/gcc-3.4/sparc-abi.html 12618 26. https://gcc.gnu.org/bugzilla/buglist.cgi?short_desc_type=notregexp&short_desc=%5C%5B3%5C.4.*%5BRr%5Degression&target_milestone=3.4.0&bug_status=RESOLVED&resolution=FIXED 12619 27. https://gcc.gnu.org/PR10129 12620 28. https://gcc.gnu.org/PR14576 12621 29. https://gcc.gnu.org/PR14760 12622 30. https://gcc.gnu.org/PR14671 12623 31. https://gcc.gnu.org/PR15093 12624 32. https://gcc.gnu.org/PR15178 12625 33. https://gcc.gnu.org/PR12753 12626 34. https://gcc.gnu.org/PR13985 12627 35. https://gcc.gnu.org/PR14810 12628 36. https://gcc.gnu.org/PR14883 12629 37. https://gcc.gnu.org/PR15044 12630 38. https://gcc.gnu.org/PR15057 12631 39. https://gcc.gnu.org/PR15064 12632 40. https://gcc.gnu.org/PR15142 12633 41. https://gcc.gnu.org/PR15159 12634 42. https://gcc.gnu.org/PR15165 12635 43. https://gcc.gnu.org/PR15193 12636 44. https://gcc.gnu.org/PR15209 12637 45. https://gcc.gnu.org/PR15227 12638 46. https://gcc.gnu.org/PR15285 12639 47. https://gcc.gnu.org/PR15299 12640 48. https://gcc.gnu.org/PR15329 12641 49. https://gcc.gnu.org/PR15550 12642 50. https://gcc.gnu.org/PR15554 12643 51. https://gcc.gnu.org/PR15640 12644 52. https://gcc.gnu.org/PR15666 12645 53. https://gcc.gnu.org/PR15696 12646 54. https://gcc.gnu.org/PR15701 12647 55. https://gcc.gnu.org/PR15761 12648 56. https://gcc.gnu.org/PR15829 12649 57. https://gcc.gnu.org/PR14538 12650 58. https://gcc.gnu.org/PR12391 12651 59. https://gcc.gnu.org/PR14649 12652 60. https://gcc.gnu.org/PR15004 12653 61. https://gcc.gnu.org/PR15749 12654 62. https://gcc.gnu.org/PR10646 12655 63. https://gcc.gnu.org/PR12077 12656 64. https://gcc.gnu.org/PR13598 12657 65. https://gcc.gnu.org/PR14211 12658 66. https://gcc.gnu.org/PR14220 12659 67. https://gcc.gnu.org/PR14245 12660 68. https://gcc.gnu.org/PR14340 12661 69. https://gcc.gnu.org/PR14600 12662 70. https://gcc.gnu.org/PR14668 12663 71. https://gcc.gnu.org/PR14775 12664 72. https://gcc.gnu.org/PR14821 12665 73. https://gcc.gnu.org/PR14930 12666 74. https://gcc.gnu.org/PR14932 12667 75. https://gcc.gnu.org/PR14950 12668 76. https://gcc.gnu.org/PR14962 12669 77. https://gcc.gnu.org/PR14975 12670 78. https://gcc.gnu.org/PR15002 12671 79. https://gcc.gnu.org/PR15025 12672 80. https://gcc.gnu.org/PR15046 12673 81. https://gcc.gnu.org/PR15069 12674 82. https://gcc.gnu.org/PR15074 12675 83. https://gcc.gnu.org/PR15083 12676 84. https://gcc.gnu.org/PR15096 12677 85. https://gcc.gnu.org/PR15287 12678 86. https://gcc.gnu.org/PR15317 12679 87. https://gcc.gnu.org/PR15337 12680 88. https://gcc.gnu.org/PR15361 12681 89. https://gcc.gnu.org/PR15412 12682 90. https://gcc.gnu.org/PR15427 12683 91. https://gcc.gnu.org/PR15471 12684 92. https://gcc.gnu.org/PR15503 12685 93. https://gcc.gnu.org/PR15507 12686 94. https://gcc.gnu.org/PR15542 12687 95. https://gcc.gnu.org/PR15565 12688 96. https://gcc.gnu.org/PR15625 12689 97. https://gcc.gnu.org/PR15629 12690 98. https://gcc.gnu.org/PR15742 12691 99. https://gcc.gnu.org/PR15775 12692 100. https://gcc.gnu.org/PR15821 12693 101. https://gcc.gnu.org/PR15862 12694 102. https://gcc.gnu.org/PR15875 12695 103. https://gcc.gnu.org/PR15877 12696 104. https://gcc.gnu.org/PR15947 12697 105. https://gcc.gnu.org/PR16020 12698 106. https://gcc.gnu.org/PR16154 12699 107. https://gcc.gnu.org/PR16174 12700 108. https://gcc.gnu.org/PR14315 12701 109. https://gcc.gnu.org/PR15151 12702 110. https://gcc.gnu.org/PR7993 12703 111. https://gcc.gnu.org/PR15228 12704 112. https://gcc.gnu.org/PR15345 12705 113. https://gcc.gnu.org/PR15945 12706 114. https://gcc.gnu.org/PR15526 12707 115. https://gcc.gnu.org/PR14690 12708 116. https://gcc.gnu.org/PR15112 12709 117. https://gcc.gnu.org/PR15067 12710 118. https://gcc.gnu.org/PR1963 12711 119. https://gcc.gnu.org/PR15717 12712 120. https://gcc.gnu.org/PR14782 12713 121. https://gcc.gnu.org/PR14828 12714 122. https://gcc.gnu.org/PR15202 12715 123. https://gcc.gnu.org/PR14610 12716 124. https://gcc.gnu.org/PR14813 12717 125. https://gcc.gnu.org/PR14857 12718 126. https://gcc.gnu.org/PR15598 12719 127. https://gcc.gnu.org/PR15653 12720 128. https://gcc.gnu.org/PR15189 12721 129. https://gcc.gnu.org/PR15331 12722 130. https://gcc.gnu.org/PR16144 12723 131. https://gcc.gnu.org/PR16176 12724 132. https://gcc.gnu.org/PR11591 12725 133. https://gcc.gnu.org/PR12028 12726 134. https://gcc.gnu.org/PR14478 12727 135. https://gcc.gnu.org/PR14567 12728 136. https://gcc.gnu.org/PR14715 12729 137. https://gcc.gnu.org/PR14902 12730 138. https://gcc.gnu.org/PR14924 12731 139. https://gcc.gnu.org/PR14960 12732 140. https://gcc.gnu.org/PR15106 12733 141. https://gcc.gnu.org/PR16026 12734 142. https://gcc.gnu.org/PR15191 12735 143. https://gcc.gnu.org/PR15662 12736 144. https://gcc.gnu.org/PR15054 12737 145. https://gcc.gnu.org/PR15783 12738 146. https://gcc.gnu.org/PR15626 12739 147. https://gcc.gnu.org/PR14326 12740 148. https://gcc.gnu.org/PR14723 12741 149. https://gcc.gnu.org/PR15290 12742 150. https://gcc.gnu.org/PR15250 12743 151. https://gcc.gnu.org/PR15551 12744 152. https://gcc.gnu.org/PR8309 12745 153. https://gcc.gnu.org/PR13250 12746 154. https://gcc.gnu.org/PR13803 12747 155. https://gcc.gnu.org/PR14093 12748 156. https://gcc.gnu.org/PR14457 12749 157. https://gcc.gnu.org/PR14542 12750 158. https://gcc.gnu.org/PR15100 12751 159. https://gcc.gnu.org/PR15296 12752 160. https://gcc.gnu.org/PR15396 12753 161. https://gcc.gnu.org/PR15782 12754 162. https://gcc.gnu.org/PR11610 12755 163. https://gcc.gnu.org/PR15488 12756 164. https://gcc.gnu.org/PR15489 12757 165. https://gcc.gnu.org/PR13928 12758 166. https://gcc.gnu.org/PR14150 12759 167. https://gcc.gnu.org/PR14949 12760 168. https://gcc.gnu.org/PR15123 12761 169. https://gcc.gnu.org/PR16469 12762 170. https://gcc.gnu.org/PR16344 12763 171. https://gcc.gnu.org/PR16842 12764 172. https://gcc.gnu.org/PR12608 12765 173. https://gcc.gnu.org/PR14492 12766 174. https://gcc.gnu.org/PR15461 12767 175. https://gcc.gnu.org/PR15890 12768 176. https://gcc.gnu.org/PR16180 12769 177. https://gcc.gnu.org/PR16224 12770 178. https://gcc.gnu.org/PR16408 12771 179. https://gcc.gnu.org/PR16529 12772 180. https://gcc.gnu.org/PR16698 12773 181. https://gcc.gnu.org/PR16706 12774 182. https://gcc.gnu.org/PR16810 12775 183. https://gcc.gnu.org/PR16851 12776 184. https://gcc.gnu.org/PR16870 12777 185. https://gcc.gnu.org/PR16904 12778 186. https://gcc.gnu.org/PR16905 12779 187. https://gcc.gnu.org/PR16964 12780 188. https://gcc.gnu.org/PR17068 12781 189. https://gcc.gnu.org/PR16366 12782 190. https://gcc.gnu.org/PR15345 12783 191. https://gcc.gnu.org/PR16590 12784 192. https://gcc.gnu.org/PR16693 12785 193. https://gcc.gnu.org/PR17078 12786 194. https://gcc.gnu.org/PR13956 12787 195. https://gcc.gnu.org/PR16684 12788 196. https://gcc.gnu.org/PR12658 12789 197. https://gcc.gnu.org/PR13092 12790 198. https://gcc.gnu.org/PR15320 12791 199. https://gcc.gnu.org/PR16246 12792 200. https://gcc.gnu.org/PR16273 12793 201. https://gcc.gnu.org/PR16401 12794 202. https://gcc.gnu.org/PR16411 12795 203. https://gcc.gnu.org/PR16489 12796 204. https://gcc.gnu.org/PR16618 12797 205. https://gcc.gnu.org/PR16637 12798 206. https://gcc.gnu.org/PR16717 12799 207. https://gcc.gnu.org/PR16813 12800 208. https://gcc.gnu.org/PR16853 12801 209. https://gcc.gnu.org/PR16889 12802 210. https://gcc.gnu.org/PR16959 12803 211. https://gcc.gnu.org/PR7587 12804 212. https://gcc.gnu.org/PR16473 12805 213. https://gcc.gnu.org/PR16478 12806 214. https://gcc.gnu.org/PR10695 12807 215. https://gcc.gnu.org/PR16974 12808 216. https://gcc.gnu.org/PR16298 12809 217. https://gcc.gnu.org/PR17113 12810 218. https://gcc.gnu.org/PR14697 12811 219. https://gcc.gnu.org/PR15869 12812 220. https://gcc.gnu.org/PR16325 12813 221. https://gcc.gnu.org/PR16357 12814 222. https://gcc.gnu.org/PR16380 12815 223. https://gcc.gnu.org/PR16407 12816 224. https://gcc.gnu.org/PR16643 12817 225. https://gcc.gnu.org/PR15927 12818 226. https://gcc.gnu.org/PR15948 12819 227. https://gcc.gnu.org/PR17019 12820 228. https://gcc.gnu.org/PR16130 12821 229. https://gcc.gnu.org/PR16142 12822 230. https://gcc.gnu.org/PR16278 12823 231. https://gcc.gnu.org/PR16414 12824 232. https://gcc.gnu.org/PR16445 12825 233. https://gcc.gnu.org/PR16490 12826 234. https://gcc.gnu.org/PR16683 12827 235. https://gcc.gnu.org/PR16195 12828 236. https://gcc.gnu.org/PR16239 12829 237. https://gcc.gnu.org/PR16199 12830 238. https://gcc.gnu.org/PR16416 12831 239. https://gcc.gnu.org/PR16430 12832 240. https://gcc.gnu.org/PR16379 12833 241. https://gcc.gnu.org/PR17093 12834 242. https://gcc.gnu.org/PR17119 12835 243. https://gcc.gnu.org/PR15928 12836 244. https://gcc.gnu.org/PR16210 12837 245. https://gcc.gnu.org/PR15488 12838 246. https://gcc.gnu.org/PR16250 12839 247. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.3 12840 248. https://gcc.gnu.org/PR17369 12841 249. https://gcc.gnu.org/PR17850 12842 250. https://gcc.gnu.org/PR13948 12843 251. https://gcc.gnu.org/PR14492 12844 252. https://gcc.gnu.org/PR16301 12845 253. https://gcc.gnu.org/PR16566 12846 254. https://gcc.gnu.org/PR17023 12847 255. https://gcc.gnu.org/PR17027 12848 256. https://gcc.gnu.org/PR17524 12849 257. https://gcc.gnu.org/PR17826 12850 258. https://gcc.gnu.org/PR15526 12851 259. https://gcc.gnu.org/PR16999 12852 260. https://gcc.gnu.org/PR17503 12853 261. https://gcc.gnu.org/PR17581 12854 262. https://gcc.gnu.org/PR18129 12855 263. https://gcc.gnu.org/PR10975 12856 264. https://gcc.gnu.org/PR11722 12857 265. https://gcc.gnu.org/PR14534 12858 266. https://gcc.gnu.org/PR15172 12859 267. https://gcc.gnu.org/PR15786 12860 268. https://gcc.gnu.org/PR16162 12861 269. https://gcc.gnu.org/PR16612 12862 270. https://gcc.gnu.org/PR16715 12863 271. https://gcc.gnu.org/PR16848 12864 272. https://gcc.gnu.org/PR17132 12865 273. https://gcc.gnu.org/PR17259 12866 274. https://gcc.gnu.org/PR17327 12867 275. https://gcc.gnu.org/PR17393 12868 276. https://gcc.gnu.org/PR17501 12869 277. https://gcc.gnu.org/PR17537 12870 278. https://gcc.gnu.org/PR17585 12871 279. https://gcc.gnu.org/PR17821 12872 280. https://gcc.gnu.org/PR17829 12873 281. https://gcc.gnu.org/PR17851 12874 282. https://gcc.gnu.org/PR17976 12875 283. https://gcc.gnu.org/PR18020 12876 284. https://gcc.gnu.org/PR18093 12877 285. https://gcc.gnu.org/PR18140 12878 286. https://gcc.gnu.org/PR17541 12879 287. https://gcc.gnu.org/PR17853 12880 288. https://gcc.gnu.org/PR17245 12881 289. https://gcc.gnu.org/PR17167 12882 290. https://gcc.gnu.org/PR17277 12883 291. https://gcc.gnu.org/PR17505 12884 292. https://gcc.gnu.org/PR17684 12885 293. https://gcc.gnu.org/PR17384 12886 294. https://gcc.gnu.org/PR17770 12887 295. https://gcc.gnu.org/PR11476 12888 296. https://gcc.gnu.org/PR14064 12889 297. https://gcc.gnu.org/PR14678 12890 298. https://gcc.gnu.org/PR15583 12891 299. https://gcc.gnu.org/PR15790 12892 300. https://gcc.gnu.org/PR15886 12893 301. https://gcc.gnu.org/PR16884 12894 302. https://gcc.gnu.org/PR13841 12895 303. https://gcc.gnu.org/PR15860 12896 304. https://gcc.gnu.org/PR17465 12897 305. https://gcc.gnu.org/PR17469 12898 306. https://gcc.gnu.org/PR18138 12899 307. https://gcc.gnu.org/PR15498 12900 308. https://gcc.gnu.org/PR15747 12901 309. https://gcc.gnu.org/PR16406 12902 310. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.4 12903 311. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.5 12904 312. https://gcc.gnu.org/PR24688 12905 313. https://gcc.gnu.org/PR17188 12906 314. https://gcc.gnu.org/PR20187 12907 315. https://gcc.gnu.org/PR21873 12908 316. https://gcc.gnu.org/PR21899 12909 317. https://gcc.gnu.org/PR22061 12910 318. https://gcc.gnu.org/PR22208 12911 319. https://gcc.gnu.org/PR22458 12912 320. https://gcc.gnu.org/PR22589 12913 321. https://gcc.gnu.org/PR24101 12914 322. https://gcc.gnu.org/PR10611 12915 323. https://gcc.gnu.org/PR13377 12916 324. https://gcc.gnu.org/PR16002 12917 325. https://gcc.gnu.org/PR17413 12918 326. https://gcc.gnu.org/PR17609 12919 327. https://gcc.gnu.org/PR17618 12920 328. https://gcc.gnu.org/PR18124 12921 329. https://gcc.gnu.org/PR18155 12922 330. https://gcc.gnu.org/PR18177 12923 331. https://gcc.gnu.org/PR18368 12924 332. https://gcc.gnu.org/PR18378 12925 333. https://gcc.gnu.org/PR18466 12926 334. https://gcc.gnu.org/PR18512 12927 335. https://gcc.gnu.org/PR18545 12928 336. https://gcc.gnu.org/PR18738 12929 337. https://gcc.gnu.org/PR18803 12930 338. https://gcc.gnu.org/PR19004 12931 339. https://gcc.gnu.org/PR19208 12932 340. https://gcc.gnu.org/PR19253 12933 341. https://gcc.gnu.org/PR19608 12934 342. https://gcc.gnu.org/PR19884 12935 343. https://gcc.gnu.org/PR20153 12936 344. https://gcc.gnu.org/PR20563 12937 345. https://gcc.gnu.org/PR20789 12938 346. https://gcc.gnu.org/PR21336 12939 347. https://gcc.gnu.org/PR21768 12940 348. https://gcc.gnu.org/PR21853 12941 349. https://gcc.gnu.org/PR21903 12942 350. https://gcc.gnu.org/PR21983 12943 351. https://gcc.gnu.org/PR21987 12944 352. https://gcc.gnu.org/PR22153 12945 353. https://gcc.gnu.org/PR22172 12946 354. https://gcc.gnu.org/PR21286 12947 355. https://gcc.gnu.org/PR22233 12948 356. https://gcc.gnu.org/PR22508 12949 357. https://gcc.gnu.org/PR22545 12950 358. https://gcc.gnu.org/PR23528 12951 359. https://gcc.gnu.org/PR23550 12952 360. https://gcc.gnu.org/PR23586 12953 361. https://gcc.gnu.org/PR23624 12954 362. https://gcc.gnu.org/PR23639 12955 363. https://gcc.gnu.org/PR23797 12956 364. https://gcc.gnu.org/PR23965 12957 365. https://gcc.gnu.org/PR24052 12958 366. https://gcc.gnu.org/PR24580 12959 367. https://gcc.gnu.org/PR24267 12960 368. https://gcc.gnu.org/PR17810 12961 369. https://gcc.gnu.org/PR17860 12962 370. https://gcc.gnu.org/PR21709 12963 371. https://gcc.gnu.org/PR21964 12964 372. https://gcc.gnu.org/PR22167 12965 373. https://gcc.gnu.org/PR22619 12966 374. https://gcc.gnu.org/PR23241 12967 375. https://gcc.gnu.org/PR23478 12968 376. https://gcc.gnu.org/PR24470 12969 377. https://gcc.gnu.org/PR24950 12970 378. https://gcc.gnu.org/PR14400 12971 379. https://gcc.gnu.org/PR14940 12972 380. https://gcc.gnu.org/PR20239 12973 381. https://gcc.gnu.org/PR15220 12974 382. https://gcc.gnu.org/PR19275 12975 383. https://gcc.gnu.org/PR21888 12976 384. https://gcc.gnu.org/PR15342 12977 385. https://gcc.gnu.org/PR23985 12978 386. https://gcc.gnu.org/PR16719 12979 387. https://gcc.gnu.org/PR21723 12980 388. https://gcc.gnu.org/PR21841 12981 389. https://gcc.gnu.org/PR23644 12982 390. https://gcc.gnu.org/PR24718 12983 391. https://gcc.gnu.org/PR18421 12984 392. https://gcc.gnu.org/PR20621 12985 393. https://gcc.gnu.org/PR18583 12986 394. https://gcc.gnu.org/PR20191 12987 395. https://gcc.gnu.org/PR22083 12988 396. https://gcc.gnu.org/PR23070 12989 397. https://gcc.gnu.org/PR23404 12990 398. https://gcc.gnu.org/PR23539 12991 399. https://gcc.gnu.org/PR24102 12992 400. https://gcc.gnu.org/PR24465 12993 401. https://gcc.gnu.org/PR19933 12994 402. https://gcc.gnu.org/PR21889 12995 403. https://gcc.gnu.org/PR19300 12996 404. https://gcc.gnu.org/PR20301 12997 405. https://gcc.gnu.org/PR20673 12998 406. https://gcc.gnu.org/PR18582 12999 407. https://gcc.gnu.org/PR19340 13000 408. https://gcc.gnu.org/PR21716 13001 409. https://gcc.gnu.org/PR24315 13002 410. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.4.6 13003 411. https://gcc.gnu.org/onlinedocs/ 13004 412. mailto:gcc-help (a] gcc.gnu.org 13005 413. mailto:gcc (a] gcc.gnu.org 13006 414. https://gcc.gnu.org/lists.html 13007 415. https://www.fsf.org/ 13008 416. https://gcc.gnu.org/about.html 13009 417. http://validator.w3.org/check/referer 13010 ====================================================================== 13011 http://gcc.gnu.org/gcc-3.3/index.html 13012 13013 GCC 3.3 Release Series 13014 13015 May 03, 2005 13016 13017 The [1]GNU project and the GCC developers are pleased to announce the 13018 release of GCC 3.3.6. 13019 13020 This release is a bug-fix release, containing fixes for regressions in 13021 GCC 3.3.5 relative to previous releases of GCC. 13022 13023 This release is the last of the series 3.3.x. 13024 13025 The GCC 3.3 release series includes numerous [2]new features, 13026 improvements, bug fixes, and other changes, thanks to an [3]amazing 13027 group of volunteers. 13028 13029 This release series is no longer maintained. 13030 13031 Release History 13032 13033 GCC 3.3.6 13034 May 3, 2005 ([4]changes) 13035 13036 GCC 3.3.5 13037 September 30, 2004 ([5]changes) 13038 13039 GCC 3.3.4 13040 May 31, 2004 ([6]changes) 13041 13042 GCC 3.3.3 13043 February 14, 2004 ([7]changes) 13044 13045 GCC 3.3.2 13046 October 16, 2003 ([8]changes) 13047 13048 GCC 3.3.1 13049 August 8, 2003 ([9]changes) 13050 13051 GCC 3.3 13052 May 14, 2003 ([10]changes) 13053 13054 References and Acknowledgements 13055 13056 GCC used to stand for the GNU C Compiler, but since the compiler 13057 supports several other languages aside from C, it now stands for the 13058 GNU Compiler Collection. 13059 13060 A list of [11]successful builds is updated as new information becomes 13061 available. 13062 13063 The GCC developers would like to thank the numerous people that have 13064 contributed new features, improvements, bug fixes, and other changes as 13065 well as test results to GCC. This [12]amazing group of volunteers is 13066 what makes GCC successful. 13067 13068 For additional information about GCC please refer to the [13]GCC 13069 project web site or contact the [14]GCC development mailing list. 13070 13071 To obtain GCC please use [15]our mirror sites, or our CVS server. 13072 13073 13074 For questions related to the use of GCC, please consult these web 13075 pages and the [16]GCC manuals. If that fails, the 13076 [17]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 13077 web pages and the development of GCC are welcome on our developer 13078 list at [18]gcc (a] gcc.gnu.org. All of [19]our lists have public 13079 archives. 13080 13081 Copyright (C) [20]Free Software Foundation, Inc. Verbatim copying and 13082 distribution of this entire article is permitted in any medium, 13083 provided this notice is preserved. 13084 13085 These pages are [21]maintained by the GCC team. Last modified 13086 2018-09-30[22]. 13087 13088 References 13089 13090 1. http://www.gnu.org/ 13091 2. http://gcc.gnu.org/gcc-3.3/changes.html 13092 3. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 13093 4. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 13094 5. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.5 13095 6. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.4 13096 7. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.3 13097 8. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.2 13098 9. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.1 13099 10. http://gcc.gnu.org/gcc-3.3/changes.html 13100 11. http://gcc.gnu.org/gcc-3.3/buildstat.html 13101 12. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 13102 13. http://gcc.gnu.org/index.html 13103 14. mailto:gcc (a] gcc.gnu.org 13104 15. http://gcc.gnu.org/mirrors.html 13105 16. https://gcc.gnu.org/onlinedocs/ 13106 17. mailto:gcc-help (a] gcc.gnu.org 13107 18. mailto:gcc (a] gcc.gnu.org 13108 19. https://gcc.gnu.org/lists.html 13109 20. https://www.fsf.org/ 13110 21. https://gcc.gnu.org/about.html 13111 22. http://validator.w3.org/check/referer 13112 ====================================================================== 13113 http://gcc.gnu.org/gcc-3.3/changes.html 13114 13115 GCC 3.3 Release Series 13116 Changes, New Features, and Fixes 13117 13118 The latest release in the 3.3 release series is [1]GCC 3.3.6. 13119 13120 Caveats 13121 13122 * The preprocessor no longer accepts multi-line string literals. They 13123 were deprecated in 3.0, 3.1, and 3.2. 13124 * The preprocessor no longer supports the -A- switch when appearing 13125 alone. -A- followed by an assertion is still supported. 13126 * Support for all the systems [2]obsoleted in GCC 3.1 has been 13127 removed from GCC 3.3. See below for a [3]list of systems which are 13128 obsoleted in this release. 13129 * Checking for null format arguments has been decoupled from the rest 13130 of the format checking mechanism. Programs which use the format 13131 attribute may regain this functionality by using the new [4]nonnull 13132 function attribute. Note that all functions for which GCC has a 13133 built-in format attribute, an appropriate built-in nonnull 13134 attribute is also applied. 13135 * The DWARF (version 1) debugging format has been deprecated and will 13136 be removed in a future version of GCC. Version 2 of the DWARF 13137 debugging format will continue to be supported for the foreseeable 13138 future. 13139 * The C and Objective-C compilers no longer accept the "Naming Types" 13140 extension (typedef foo = bar); it was already unavailable in C++. 13141 Code which uses it will need to be changed to use the "typeof" 13142 extension instead: typedef typeof(bar) foo. (We have removed this 13143 extension without a period of deprecation because it has caused the 13144 compiler to crash since version 3.0 and no one noticed until very 13145 recently. Thus we conclude it is not in widespread use.) 13146 * The -traditional C compiler option has been removed. It was 13147 deprecated in 3.1 and 3.2. (Traditional preprocessing remains 13148 available.) The <varargs.h> header, used for writing variadic 13149 functions in traditional C, still exists but will produce an error 13150 message if used. 13151 * GCC 3.3.1 automatically places zero-initialized variables in the 13152 .bss section on some operating systems. Versions of GNU Emacs up to 13153 (and including) 21.3 will not work correctly when using this 13154 optimization; you can use -fno-zero-initialized-in-bss to disable 13155 it. 13156 13157 General Optimizer Improvements 13158 13159 * A new scheme for accurately describing processor pipelines, the 13160 [5]DFA scheduler, has been added. 13161 * Pavel Nejedly, Charles University Prague, has contributed new file 13162 format used by the edge coverage profiler (-fprofile-arcs). 13163 The new format is robust and diagnoses common mistakes where 13164 profiles from different versions (or compilations) of the program 13165 are combined resulting in nonsensical profiles and slow code to 13166 produced with profile feedback. Additionally this format allows 13167 extra data to be gathered. Currently, overall statistics are 13168 produced helping optimizers to identify hot spots of a program 13169 globally replacing the old intra-procedural scheme and resulting in 13170 better code. Note that the gcov tool from older GCC versions will 13171 not be able to parse the profiles generated by GCC 3.3 and vice 13172 versa. 13173 * Jan Hubicka, SuSE Labs, has contributed a new superblock formation 13174 pass enabled using -ftracer. This pass simplifies the control flow 13175 of functions allowing other optimizations to do better job. 13176 He also contributed the function reordering pass 13177 (-freorder-functions) to optimize function placement using profile 13178 feedback. 13179 13180 New Languages and Language specific improvements 13181 13182 C/ObjC/C++ 13183 13184 * The preprocessor now accepts directives within macro arguments. It 13185 processes them just as if they had not been within macro arguments. 13186 * The separate ISO and traditional preprocessors have been completely 13187 removed. The front end handles either type of preprocessed output 13188 if necessary. 13189 * In C99 mode preprocessor arithmetic is done in the precision of the 13190 target's intmax_t, as required by that standard. 13191 * The preprocessor can now copy comments inside macros to the output 13192 file when the macro is expanded. This feature, enabled using the 13193 -CC option, is intended for use by applications which place 13194 metadata or directives inside comments, such as lint. 13195 * The method of constructing the list of directories to be searched 13196 for header files has been revised. If a directory named by a -I 13197 option is a standard system include directory, the option is 13198 ignored to ensure that the default search order for system 13199 directories and the special treatment of system header files are 13200 not defeated. 13201 * A few more [6]ISO C99 features now work correctly. 13202 * A new function attribute, nonnull, has been added which allows 13203 pointer arguments to functions to be specified as requiring a 13204 non-null value. The compiler currently uses this information to 13205 issue a warning when it detects a null value passed in such an 13206 argument slot. 13207 * A new type attribute, may_alias, has been added. Accesses to 13208 objects with types with this attribute are not subjected to 13209 type-based alias analysis, but are instead assumed to be able to 13210 alias any other type of objects, just like the char type. 13211 13212 C++ 13213 13214 * Type based alias analysis has been implemented for C++ aggregate 13215 types. 13216 13217 Objective-C 13218 13219 * Generate an error if Objective-C objects are passed by value in 13220 function and method calls. 13221 * When -Wselector is used, check the whole list of selectors at the 13222 end of compilation, and emit a warning if a @selector() is not 13223 known. 13224 * Define __NEXT_RUNTIME__ when compiling for the NeXT runtime. 13225 * No longer need to include objc/objc-class.h to compile self calls 13226 in class methods (NeXT runtime only). 13227 * New -Wundeclared-selector option. 13228 * Removed selector bloating which was causing object files to be 10% 13229 bigger on average (GNU runtime only). 13230 * Using at run time @protocol() objects has been fixed in certain 13231 situations (GNU runtime only). 13232 * Type checking has been fixed and improved in many situations 13233 involving protocols. 13234 13235 Java 13236 13237 * The java.sql and javax.sql packages now implement the JDBC 3.0 (JDK 13238 1.4) API. 13239 * The JDK 1.4 assert facility has been implemented. 13240 * The bytecode interpreter is now direct threaded and thus faster. 13241 13242 Fortran 13243 13244 * Fortran improvements are listed in [7]the Fortran documentation. 13245 13246 Ada 13247 13248 * Ada tasking now works with glibc 2.3.x threading libraries. 13249 13250 New Targets and Target Specific Improvements 13251 13252 * The following changes have been made to the HP-PA port: 13253 + The port now defaults to scheduling for the PA8000 series of 13254 processors. 13255 + Scheduling support for the PA7300 processor has been added. 13256 + The 32-bit port now supports weak symbols under HP-UX 11. 13257 + The handling of initializers and finalizers has been improved 13258 under HP-UX 11. The 64-bit port no longer uses collect2. 13259 + Dwarf2 EH support has been added to the 32-bit GNU/Linux port. 13260 + ABI fixes to correct the passing of small structures by value. 13261 * The SPARC, HP-PA, SH4, and x86/pentium ports have been converted to 13262 use the DFA processor pipeline description. 13263 * The following NetBSD configurations for the SuperH processor family 13264 have been added: 13265 + SH3, big-endian, sh-*-netbsdelf* 13266 + SH3, little-endian, shle-*-netbsdelf* 13267 + SH5, SHmedia, big-endian, 32-bit default, sh5-*-netbsd* 13268 + SH5, SHmedia, little-endian, 32-bit default, sh5le-*-netbsd* 13269 + SH5, SHmedia, big-endian, 64-bit default, sh64-*-netbsd* 13270 + SH5, SHmedia, little-endian, 64-bit default, sh64le-*-netbsd* 13271 * The following changes have been made to the IA-32/x86-64 port: 13272 + SSE2 and 3dNOW! intrinsics are now supported. 13273 + Support for thread local storage has been added to the IA-32 13274 and x86-64 ports. 13275 + The x86-64 port has been significantly improved. 13276 * The following changes have been made to the MIPS port: 13277 + All configurations now accept the -mabi switch. Note that you 13278 will need appropriate multilibs for this option to work 13279 properly. 13280 + ELF configurations will always pass an ABI flag to the 13281 assembler, except when the MIPS EABI is selected. 13282 + -mabi=64 no longer selects MIPS IV code. 13283 + The -mcpu option, which was deprecated in 3.1 and 3.2, has 13284 been removed from this release. 13285 + -march now changes the core ISA level. In previous releases, 13286 it would change the use of processor-specific extensions, but 13287 would leave the core ISA unchanged. For example, mips64-elf 13288 -march=r8000 will now generate MIPS IV code. 13289 + Under most configurations, -mipsN now acts as a synonym for 13290 -march. 13291 + There are some new preprocessor macros to describe the -march 13292 and -mtune settings. See the documentation of those options 13293 for details. 13294 + Support for the NEC VR-Series processors has been added. This 13295 includes the 54xx, 5500, and 41xx series. 13296 + Support for the Sandcraft sr71k processor has been added. 13297 * The following changes have been made to the S/390 port: 13298 + Support to build the Java runtime libraries has been added. 13299 Java is now enabled by default on s390-*-linux* and 13300 s390x-*-linux* targets. 13301 + Multilib support for the s390x-*-linux* target has been added; 13302 this allows to build 31-bit binaries using the -m31 option. 13303 + Support for thread local storage has been added. 13304 + Inline assembler code may now use the 'Q' constraint to 13305 specify memory operands without index register. 13306 + Various platform-specific performance improvements have been 13307 implemented; in particular, the compiler now uses the BRANCH 13308 ON COUNT family of instructions and makes more frequent use of 13309 the TEST UNDER MASK family of instructions. 13310 * The following changes have been made to the PowerPC port: 13311 + Support for IBM Power4 processor added. 13312 + Support for Motorola e500 SPE added. 13313 + Support for AIX 5.2 added. 13314 + Function and Data sections now supported on AIX. 13315 + Sibcall optimizations added. 13316 * The support for H8 Tiny is added to the H8/300 port with -mn. 13317 13318 Obsolete Systems 13319 13320 Support for a number of older systems has been declared obsolete in GCC 13321 3.3. Unless there is activity to revive them, the next release of GCC 13322 will have their sources permanently removed. 13323 13324 All configurations of the following processor architectures have been 13325 declared obsolete: 13326 * Matsushita MN10200, mn10200-*-* 13327 * Motorola 88000, m88k-*-* 13328 * IBM ROMP, romp-*-* 13329 13330 Also, some individual systems have been obsoleted: 13331 * Alpha 13332 + Interix, alpha*-*-interix* 13333 + Linux libc1, alpha*-*-linux*libc1* 13334 + Linux ECOFF, alpha*-*-linux*ecoff* 13335 * ARM 13336 + Generic a.out, arm*-*-aout* 13337 + Conix, arm*-*-conix* 13338 + "Old ABI," arm*-*-oabi 13339 + StrongARM/COFF, strongarm-*-coff* 13340 * HPPA (PA-RISC) 13341 + Generic OSF, hppa1.0-*-osf* 13342 + Generic BSD, hppa1.0-*-bsd* 13343 + HP/UX versions 7, 8, and 9, hppa1.[01]-*-hpux[789]* 13344 + HiUX, hppa*-*-hiux* 13345 + Mach Lites, hppa*-*-lites* 13346 * Intel 386 family 13347 + Windows NT 3.x, i?86-*-win32 13348 * MC68000 family 13349 + HP systems, m68000-hp-bsd* and m68k-hp-bsd* 13350 + Sun systems, m68000-sun-sunos*, m68k-sun-sunos*, and 13351 m68k-sun-mach* 13352 + AT&T systems, m68000-att-sysv* 13353 + Atari systems, m68k-atari-sysv* 13354 + Motorola systems, m68k-motorola-sysv* 13355 + NCR systems, m68k-ncr-sysv* 13356 + Plexus systems, m68k-plexus-sysv* 13357 + Commodore systems, m68k-cbm-sysv* 13358 + Citicorp TTI, m68k-tti-* 13359 + Unos, m68k-crds-unos* 13360 + Concurrent RTU, m68k-ccur-rtu* 13361 + Linux a.out, m68k-*-linux*aout* 13362 + Linux libc1, m68k-*-linux*libc1* 13363 + pSOS, m68k-*-psos* 13364 * MIPS 13365 + Generic ECOFF, mips*-*-ecoff* 13366 + SINIX, mips-sni-sysv4 13367 + Orion RTEMS, mips64orion-*-rtems* 13368 * National Semiconductor 32000 13369 + OpenBSD, ns32k-*-openbsd* 13370 * POWER (aka RS/6000) and PowerPC 13371 + AIX versions 1, 2, and 3, rs6000-ibm-aix[123]* 13372 + Bull BOSX, rs6000-bull-bosx 13373 + Generic Mach, rs6000-*-mach* 13374 + Generic SysV, powerpc*-*-sysv* 13375 + Linux libc1, powerpc*-*-linux*libc1* 13376 * Sun SPARC 13377 + Generic a.out, sparc-*-aout*, sparclet-*-aout*, 13378 sparclite-*-aout*, and sparc86x-*-aout* 13379 + NetBSD a.out, sparc-*-netbsd*aout* 13380 + Generic BSD, sparc-*-bsd* 13381 + ChorusOS, sparc-*-chorusos* 13382 + Linux a.out, sparc-*-linux*aout* 13383 + Linux libc1, sparc-*-linux*libc1* 13384 + LynxOS, sparc-*-lynxos* 13385 + Solaris on HAL hardware, sparc-hal-solaris2* 13386 + SunOS versions 3 and 4, sparc-*-sunos[34]* 13387 * NEC V850 13388 + RTEMS, v850-*-rtems* 13389 * VAX 13390 + VMS, vax-*-vms* 13391 13392 Documentation improvements 13393 13394 Other significant improvements 13395 13396 * Almost all front-end dependencies in the compiler have been 13397 separated out into a set of language hooks. This should make adding 13398 a new front end clearer and easier. 13399 * One effect of removing the separate preprocessor is a small 13400 increase in the robustness of the compiler in general, and the 13401 maintainability of target descriptions. Previously target-specific 13402 built-in macros and others, such as __FAST_MATH__, had to be 13403 handled with so-called specs that were hard to maintain. Often they 13404 would fail to behave properly when conflicting options were 13405 supplied on the command line, and define macros in the user's 13406 namespace even when strict ISO compliance was requested. 13407 Integrating the preprocessor has cleanly solved these issues. 13408 * The Makefile suite now supports redirection of make install by 13409 means of the variable DESTDIR. 13410 __________________________________________________________________ 13411 13412 GCC 3.3 13413 13414 Detailed release notes for the GCC 3.3 release follow. 13415 13416 Bug Fixes 13417 13418 bootstrap failures 13419 13420 * [8]10140 cross compiler build failures: missing __mempcpy (DUP: 13421 [9]10198,[10]10338) 13422 13423 Internal compiler errors (multi-platform) 13424 13425 * [11]3581 large string causes segmentation fault in cc1 13426 * [12]4382 __builtin_{set,long}jmp with -O3 can crash the compiler 13427 * [13]5533 (c++) ICE when processing std::accumulate(begin, end, 13428 init, invalid_op) 13429 * [14]6387 -fpic -gdwarf-2 -g1 combination gives ICE in dwarf2out 13430 * [15]6412 (c++) ICE in retrieve_specialization 13431 * [16]6620 (c++) partial template specialization causes an ICE 13432 (segmentation fault) 13433 * [17]6663 (c++) ICE with attribute aligned 13434 * [18]7068 ICE with incomplete types 13435 * [19]7083 (c++) ICE using -gstabs with dodgy class derivation 13436 * [20]7647 (c++) ICE when data member has the name of the enclosing 13437 class 13438 * [21]7675 ICE in fixup_var_refs_1 13439 * [22]7718 'complex' template instantiation causes ICE 13440 * [23]8116 (c++) ICE in member template function 13441 * [24]8358 (ada) Ada compiler accesses freed memory, crashes 13442 * [25]8511 (c++) ICE: (hopefully) reproducible cc1plus segmentation 13443 fault 13444 * [26]8564 (c++) ICE in find_function_data, in function.c 13445 * [27]8660 (c++) template overloading ICE in tsubst_expr, in cp/pt.c 13446 * [28]8766 (c++) ICE after failed initialization of static template 13447 variable 13448 * [29]8803 ICE in instantiate_virtual_regs_1, in function.c 13449 * [30]8846 (c++) ICE after diagnostic if fr_FR@euro locale is set 13450 * [31]8906 (c++) ICE (Segmentation fault) when parsing nested-class 13451 definition 13452 * [32]9216 (c++) ICE on missing template parameter 13453 * [33]9261 (c++) ICE in arg_assoc, in cp/decl2.c 13454 * [34]9263 (fortran) ICE caused by invalid PARAMETER in implied DO 13455 loop 13456 * [35]9429 (c++) ICE in template instantiation with a pointered new 13457 operator 13458 * [36]9516 Internal error when using a big array 13459 * [37]9600 (c++) ICE with typedefs in template class 13460 * [38]9629 (c++) virtual inheritance segfault 13461 * [39]9672 (c++) ICE: Error reporting routines re-entered 13462 * [40]9749 (c++) ICE in write_expression on invalid function 13463 prototype 13464 * [41]9794 (fortran) ICE: floating point exception during constant 13465 folding 13466 * [42]9829 (c++) Missing colon in nested namespace usage causes ICE 13467 * [43]9916 (c++) ICE with noreturn function in ?: statement 13468 * [44]9936 ICE with local function and variable-length 2d array 13469 * [45]10262 (c++) cc1plus crashes with large generated code 13470 * [46]10278 (c++) ICE in parser for invalid code 13471 * [47]10446 (c++) ICE on definition of nonexistent member function of 13472 nested class in a class template 13473 * [48]10451 (c++) ICE in grokdeclarator on spurious mutable 13474 declaration 13475 * [49]10506 (c++) ICE in build_new at cp/init.c with 13476 -fkeep-inline-functions and multiple inheritance 13477 * [50]10549 (c++) ICE in store_bit_field on bitfields that exceed the 13478 precision of the declared type 13479 13480 Optimization bugs 13481 13482 * [51]2001 Inordinately long compile times in reload CSE regs 13483 * [52]2391 Exponential compilation time explosion in combine 13484 * [53]2960 Duplicate loop conditions even with -Os 13485 * [54]4046 redundant conditional branch 13486 * [55]6405 Loop-unrolling related performance regressions 13487 * [56]6798 very long compile time with large case-statement 13488 * [57]6871 const objects shouldn't be moved to .bss 13489 * [58]6909 problem w/ -Os on modified loop-2c.c test case 13490 * [59]7189 gcc -O2 -Wall does not print ``control reaches end of 13491 non-void function'' warning 13492 * [60]7642 optimization problem with signbit() 13493 * [61]8634 incorrect code for inlining of memcpy under -O2 13494 * [62]8750 Cygwin prolog generation erroneously emitting __alloca as 13495 regular function call 13496 13497 C front end 13498 13499 * [63]2161 long if-else cascade overflows parser stack 13500 * [64]4319 short accepted on typedef'd char 13501 * [65]8602 incorrect line numbers in warning messages when using 13502 inline functions 13503 * [66]9177 -fdump-translation-unit: C front end deletes function_decl 13504 AST nodes and breaks debugging dumps 13505 * [67]9853 miscompilation of non-constant structure initializer 13506 13507 c++ compiler and library 13508 13509 * [68]45 legal template specialization code is rejected (DUP: 13510 [69]3784) 13511 * [70]764 lookup failure: friend operator and dereferencing a pointer 13512 and templates (DUP: [71]5116) 13513 * [72]2862 gcc accepts invalid explicit instantiation syntax (DUP: 13514 2863) 13515 * [73]3663 G++ doesn't check access control during template 13516 instantiation 13517 * [74]3797 gcc fails to emit explicit specialization of a template 13518 member 13519 * [75]3948 Two destructors are called when no copy destructor is 13520 defined (ABI change) 13521 * [76]4137 Conversion operator within template is not accepted 13522 * [77]4361 bogus ambiguity taking the address of a member template 13523 * [78]4802 g++ accepts illegal template code (access to private 13524 member; DUP: [79]5837) 13525 * [80]4803 inline function is used but never defined, and g++ does 13526 not object 13527 * [81]5094 Partial specialization cannot be friend? 13528 * [82]5730 complex<double>::norm() -- huge slowdown from egcs-2.91.66 13529 * [83]6713 Regression wrt 3.0.4: g++ -O2 leads to seg fault at run 13530 time 13531 * [84]7015 certain __asm__ constructs rejected 13532 * [85]7086 compile time regression (quadratic behavior in 13533 fixup_var_refs) 13534 * [86]7099 G++ doesn't set the noreturn attribute on std::exit and 13535 std::abort 13536 * [87]7247 copy constructor missing when inlining enabled (invalid 13537 optimization?) 13538 * [88]7441 string array initialization compilation time regression 13539 from seconds to minutes 13540 * [89]7768 __PRETTY_FUNCTION__ for template destructor is wrong 13541 * [90]7804 bad printing of floating point constant in warning message 13542 * [91]8099 Friend classes and template specializations 13543 * [92]8117 member function pointers and multiple inheritance 13544 * [93]8205 using declaration and multiple inheritance 13545 * [94]8645 unnecessary non-zero checks in stl_tree.h 13546 * [95]8724 explicit destructor call for incomplete class allowed 13547 * [96]8805 compile time regression with many member variables 13548 * [97]8691 -O3 and -fno-implicit-templates are incompatible 13549 * [98]8700 unhelpful error message for binding temp to reference 13550 * [99]8724 explicit destructor call for incomplete class allowed 13551 * [100]8949 numeric_limits<>::denorm_min() and is_iec559 problems 13552 * [101]9016 Failure to consistently constant fold "constant" C++ 13553 objects 13554 * [102]9053 g++ confused about ambiguity of overloaded function 13555 templates 13556 * [103]9152 undefined virtual thunks 13557 * [104]9182 basic_filebuf<> does not report errors in codecvt<>::out 13558 * [105]9297 data corruption due to codegen bug (when copying.) 13559 * [106]9318 i/ostream::operator>>/<<(streambuf*) broken 13560 * [107]9320 Incorrect usage of traits_type::int_type in stdio_filebuf 13561 * [108]9400 bogus -Wshadow warning: shadowed declaration of this in 13562 local classes 13563 * [109]9424 i/ostream::operator>>/<<(streambuf*) drops characters 13564 * [110]9425 filebuf::pbackfail broken (DUP: [111]9439) 13565 * [112]9474 GCC freezes in compiling a weird code mixing <iostream> 13566 and <iostream.h> 13567 * [113]9548 Incorrect results from setf(ios::fixed) and precision(-1) 13568 [114][DR 231] 13569 * [115]9555 ostream inserters fail to set badbit on exception 13570 * [116]9561 ostream inserters rethrow exception of wrong type 13571 * [117]9563 ostream::sentry returns true after a failed preparation 13572 * [118]9582 one-definition rule violation in std::allocator 13573 * [119]9622 __PRETTY_FUNCTION__ incorrect in template destructors 13574 * [120]9683 bug in initialization chains for static const variables 13575 from template classes 13576 * [121]9791 -Woverloaded-virtual reports hiding of destructor 13577 * [122]9817 collate::compare doesn't handle nul characters 13578 * [123]9825 filebuf::sputbackc breaks sbumpc 13579 * [124]9826 operator>>(basic_istream, basic_string) fails to compile 13580 with custom traits 13581 * [125]9924 Multiple using statements for builtin functions not 13582 allowed 13583 * [126]9946 destructor is not called for temporary object 13584 * [127]9964 filebuf::close() sometimes fails to close file 13585 * [128]9988 filebuf::overflow writes EOF to file 13586 * [129]10033 optimization breaks polymorphic references w/ typeid 13587 operator 13588 * [130]10097 filebuf::underflow drops characters 13589 * [131]10132 filebuf destructor can throw exceptions 13590 * [132]10180 gcc fails to warn about non-inlined function 13591 * [133]10199 method parametrized by template does not work everywhere 13592 * [134]10300 use of array-new (nothrow) in segfaults on NULL return 13593 * [135]10427 Stack corruption with variable-length automatic arrays 13594 and virtual destructors 13595 * [136]10503 Compilation never stops in fixed_type_or_null 13596 13597 Objective-C 13598 13599 * [137]5956 selectors aren't matched properly when added to the 13600 selector table 13601 13602 Fortran compiler and library 13603 13604 * [138]1832 list directed i/o overflow hangs, -fbounds-check doesn't 13605 detect 13606 * [139]3924 g77 generates code that is rejected by GAS if COFF debug 13607 info requested 13608 * [140]5634 doc: explain that configure --prefix=~/... does not work 13609 * [141]6367 multiple repeat counts confuse namelist read into array 13610 * [142]6491 Logical operations error on logicals when using 13611 -fugly-logint 13612 * [143]6742 Generation of C++ Prototype for FORTRAN and extern "C" 13613 * [144]7113 Failure of g77.f-torture/execute/f90-intrinsic-bit.f -Os 13614 on irix6.5 13615 * [145]7236 OPEN(...,RECL=nnn,...) without ACCESS='DIRECT' should 13616 assume a direct access file 13617 * [146]7278 g77 "bug"; the executable misbehaves (with -O2 13618 -fno-automatic) 13619 * [147]7384 DATE_AND_TIME milliseconds field inactive on Windows 13620 * [148]7388 Incorrect output with 0-based array of characters 13621 * [149]8587 Double complex zero ** double precision number -> NaN 13622 instead of zero 13623 * [150]9038 -ffixed-line-length-none -x f77-cpp-input gives: Warning: 13624 unknown register name line-length-none 13625 * [151]10197 Direct access files not unformatted by default 13626 13627 Java compiler and library 13628 13629 * [152]6005 gcj fails to build rhug on alpha 13630 * [153]6389 System.getProperty("") should always throw an 13631 IllegalArgumentException 13632 * [154]6576 java.util.ResourceBundle.getResource ignores locale 13633 * [155]6652 new java.io.File("").getCanonicalFile() throws exception 13634 * [156]7060 getMethod() doesn't search super interface 13635 * [157]7073 bytecode interpreter gives wrong answer for interface 13636 getSuperclass() 13637 * [158]7180 possible bug in 13638 javax.naming.spi.NamingManager.getPlusPath() 13639 * [159]7416 java.security startup refs "GNU libgcj.security" 13640 * [160]7570 Runtime.exec with null envp: child doesn't inherit parent 13641 env (DUP: [161]7578) 13642 * [162]7611 Internal error while compiling libjava with -O 13643 * [163]7709 NullPointerException in _Jv_ResolvePoolEntry 13644 * [164]7766 ZipInputStream.available returns 0 immediately after 13645 construction 13646 * [165]7785 Calendar.getTimeInMillis/setTimeInMillis should be public 13647 * [166]7786 TimeZone.getDSTSavings() from JDK1.4 not implemented 13648 * [167]8142 '$' in class names vs. dlopen 'dynamic string tokens' 13649 * [168]8234 ZipInputStream chokes when InputStream.read() returns 13650 small chunks 13651 * [169]8415 reflection bug: exception info for Method 13652 * [170]8481 java.Random.nextInt(int) may return negative 13653 * [171]8593 Error reading GZIPped files with BufferedReader 13654 * [172]8759 java.beans.Introspector has no flushCaches() or 13655 flushFromCaches() methods 13656 * [173]8997 spin() calls Thread.sleep 13657 * [174]9253 on win32, java.io.File.listFiles("C:\\") returns pwd 13658 instead of the root content of C: 13659 * [175]9254 java::lang::Object::wait(), threads-win32.cc returns 13660 wrong return codes 13661 * [176]9271 Severe bias in java.security.SecureRandom 13662 13663 Ada compiler and library 13664 13665 * [177]6767 make gnatlib-shared fails on -laddr2line 13666 * [178]9911 gnatmake fails to link when GCC configured with 13667 --with-sjlj-exceptions=yes 13668 * [179]10020 Can't bootstrap gcc on AIX with Ada enabled 13669 * [180]10546 Ada tasking not working on Red Hat 9 13670 13671 preprocessor 13672 13673 * [181]7029 preprocessor should ignore #warning with -M 13674 13675 ARM-specific 13676 13677 * [182]2903 [arm] Optimization bug with long long arithmetic 13678 * [183]7873 arm-linux-gcc fails when assigning address to a bit field 13679 13680 FreeBSD-specific 13681 13682 * [184]7680 float functions undefined in math.h/cmath with #define 13683 _XOPEN_SOURCE 13684 13685 HP-UX or HP-PA-specific 13686 13687 * [185]8705 [HP-PA] ICE in emit_move_insn_1, in expr.c 13688 * [186]9986 [HP-UX] Incorrect transformation of fputs_unlocked to 13689 fputc_unlocked 13690 * [187]10056 [HP-PA] ICE at -O2 when building c++ code from doxygen 13691 13692 m68hc11-specific 13693 13694 * [188]6744 Bad assembler code generated: reference to pseudo 13695 register z 13696 * [189]7361 Internal compiler error in reload_cse_simplify_operands, 13697 in reload1.c 13698 13699 MIPS-specific 13700 13701 * [190]9496 [mips-linux] bug in optimizer? 13702 13703 PowerPC-specific 13704 13705 * [191]7067 -Os with -mcpu=powerpc optimizes for speed (?) instead of 13706 space 13707 * [192]8480 reload ICEs for LAPACK code on powerpc64-linux 13708 * [193]8784 [AIX] Internal compiler error in simplify_gen_subreg 13709 * [194]10315 [powerpc] ICE: in extract_insn, in recog.c 13710 13711 SPARC-specific 13712 13713 * [195]10267 (documentation) Wrong build instructions for 13714 *-*-solaris2* 13715 13716 x86-specific (Intel/AMD) 13717 13718 * [196]7916 ICE in instantiate_virtual_register_1 13719 * [197]7926 (c++) i486 instructions in header files make c++ programs 13720 crash on i386 13721 * [198]8555 ICE in gen_split_1231 13722 * [199]8994 ICE with -O -march=pentium4 13723 * [200]9426 ICE with -fssa -funroll-loops -fprofile-arcs 13724 * [201]9806 ICE in inline assembly with -fPIC flag 13725 * [202]10077 gcc -msse2 generates movd to move dwords between xmm 13726 regs 13727 * [203]10233 64-bit comparison only comparing bottom 32-bits 13728 * [204]10286 type-punning doesn't work with __m64 and -O 13729 * [205]10308 [x86] ICE with -O -fgcse or -O2 13730 __________________________________________________________________ 13731 13732 GCC 3.3.1 13733 13734 Bug Fixes 13735 13736 This section lists the problem reports (PRs) from GCC's bug tracking 13737 system that are known to be fixed in the 3.3.1 release. This list might 13738 not be complete (that is, it is possible that some PRs that have been 13739 fixed are not listed here). 13740 13741 Bootstrap failures 13742 13743 * [206]11272 [Solaris] make bootstrap fails while building libstdc++ 13744 13745 Internal compiler errors (multi-platform) 13746 13747 * [207]5754 ICE on invalid nested template class 13748 * [208]6597 ICE in set_mem_alias_set compiling Qt with -O2 on ia64 13749 and --enable-checking 13750 * [209]6949 (c++) ICE in tsubst_decl, in cp/pt.c 13751 * [210]7053 (c++) ICE when declaring a function already defined as a 13752 friend method of a template class 13753 * [211]8164 (c++) ICE when using different const expressions as 13754 template parameter 13755 * [212]8384 (c++) ICE in is_base_type, in dwarf2out.c 13756 * [213]9559 (c++) ICE with invalid initialization of a static const 13757 * [214]9649 (c++) ICE in finish_member_declaration, in cp/semantics.c 13758 when redeclaring a static member variable 13759 * [215]9864 (fortran) ICE in add_abstract_origin_attribute, in 13760 dwarfout.c with -g -O -finline-functions 13761 * [216]10432 (c++) ICE in poplevel, in cp/decl.c 13762 * [217]10475 ICE in subreg_highpart_offset for code with long long 13763 * [218]10635 (c++) ICE when dereferencing an incomplete type casted 13764 from a void pointer 13765 * [219]10661 (c++) ICE in instantiate_decl, in cp/pt.c while 13766 instantiating static member variables 13767 * [220]10700 ICE in copy_to_mode_reg on 64-bit targets 13768 * [221]10712 (c++) ICE in constructor_name_full, in cp/decl2.c 13769 * [222]10796 (c++) ICE when defining an enum with two values: -1 and 13770 MAX_INT_64BIT 13771 * [223]10890 ICE in merge_assigned_reloads building Linux 2.4.2x 13772 sched.c 13773 * [224]10939 (c++) ICE with template code 13774 * [225]10956 (c++) ICE when specializing a template member function 13775 of a template class, in tsubst, in cp/pt.c 13776 * [226]11041 (c++) ICE: const myclass &x = *x; (when operator*() 13777 defined) 13778 * [227]11059 (c++) ICE with empty union 13779 * [228]11083 (c++) ICE in commit_one_edge_insertion, in cfgrtl.c with 13780 -O2 -fnon-call-exceptions 13781 * [229]11105 (c++) ICE in mangle_conv_op_name_for_type 13782 * [230]11149 (c++) ICE on error when instantiation with call function 13783 of a base type 13784 * [231]11228 (c++) ICE on new-expression using array operator new and 13785 default-initialization 13786 * [232]11282 (c++) Infinite memory usage after syntax error 13787 * [233]11301 (fortran) ICE with -fno-globals 13788 * [234]11308 (c++) ICE when using an enum type name as if it were a 13789 class or namespace 13790 * [235]11473 (c++) ICE with -gstabs when empty struct inherits from 13791 an empty struct 13792 * [236]11503 (c++) ICE when instantiating template with ADDR_EXPR 13793 * [237]11513 (c++) ICE in push_template_decl_real, in cp/pt.c: 13794 template member functions 13795 13796 Optimization bugs 13797 13798 * [238]11198 -O2 -frename-registers generates wrong code (aliasing 13799 problem) 13800 * [239]11304 Wrong code production with -fomit-frame-pointer 13801 * [240]11381 volatile memory access optimized away 13802 * [241]11536 [strength-reduce] -O2 optimization produces wrong code 13803 * [242]11557 constant folding bug generates wrong code 13804 13805 C front end 13806 13807 * [243]5897 No warning for statement after return 13808 * [244]11279 DWARF-2 output mishandles large enums 13809 13810 Preprocessor bugs 13811 13812 * [245]11022 no warning for non-compatible macro redefinition 13813 13814 C++ compiler and library 13815 13816 * [246]2330 static_cast<>() to a private base is allowed 13817 * [247]5388 Incorrect message "operands to ?: have different types" 13818 * [248]5390 Libiberty fails to demangle multi-digit template 13819 parameters 13820 * [249]7877 Incorrect parameter passing to specializations of member 13821 function templates 13822 * [250]9393 Anonymous namespaces and compiling the same file twice 13823 * [251]10032 -pedantic converts some errors to warnings 13824 * [252]10468 const typeof(x) is non-const, but only in templates 13825 * [253]10527 confused error message with "new int()" parameter 13826 initializer 13827 * [254]10679 parameter MIN_INLINE_INSNS is not honored 13828 * [255]10682 gcc chokes on a typedef for an enum inside a class 13829 template 13830 * [256]10689 pow(std::complex(0),1/3) returns (nan, nan) instead of 13831 0. 13832 * [257]10845 template member function (with nested template as 13833 parameter) cannot be called anymore if another unrelated template 13834 member function is defined 13835 * [258]10849 Cannot define an out-of-class specialization of a 13836 private nested template class 13837 * [259]10888 Suppress -Winline warnings for system headers 13838 * [260]10929 -Winline warns about functions for which no definition 13839 is visible 13840 * [261]10931 valid conversion static_cast<const unsigned 13841 int&>(lvalue-of-type-int) is rejected 13842 * [262]10940 Bad code with explicit specialization 13843 * [263]10968 If member function implicitly instantiated, explicit 13844 instantiation of class fails to instantiate it 13845 * [264]10990 Cannot convert with dynamic_cast<> to a private base 13846 class from within a member function 13847 * [265]11039 Bad interaction between implicit typename deprecation 13848 and friendship 13849 * [266]11062 (libstdc++) avoid __attribute__ ((unused)); say 13850 "__unused__" instead 13851 * [267]11095 C++ iostream manipulator causes segfault when called 13852 with negative argument 13853 * [268]11098 g++ doesn't emit complete debugging information for 13854 local variables in destructors 13855 * [269]11137 GNU/Linux shared library constructors not called unless 13856 there's one global object 13857 * [270]11154 spurious ambiguity report for template class 13858 specialization 13859 * [271]11329 Compiler cannot find user defined implicit typecast 13860 * [272]11332 Spurious error with casts in ?: expression 13861 * [273]11431 static_cast behavior with subclasses when default 13862 constructor available 13863 * [274]11528 money_get facet does not accept "$.00" as valid 13864 * [275]11546 Type lookup problems in out-of-line definition of a 13865 class doubly nested from a template class 13866 * [276]11567 C++ code containing templated member function with same 13867 name as pure virtual member function results in linking failure 13868 * [277]11645 Failure to deal with using and private inheritance 13869 13870 Java compiler and library 13871 13872 * [278]5179 Qualified static field access doesn't initialize its 13873 class 13874 * [279]8204 gcj -O2 to native reorders certain instructions 13875 improperly 13876 * [280]10838 java.io.ObjectInputStream syntax error 13877 * [281]10886 The RMI registry that comes with GCJ does not work 13878 correctly 13879 * [282]11349 JNDI URL context factories not located correctly 13880 13881 x86-specific (Intel/AMD) 13882 13883 * [283]4823 ICE on inline assembly code 13884 * [284]8878 miscompilation with -O and SSE 13885 * [285]9815 (c++ library) atomicity.h - fails to compile with -O3 13886 -masm=intel 13887 * [286]10402 (inline assembly) [x86] ICE in merge_assigned_reloads, 13888 in reload1.c 13889 * [287]10504 ICE with SSE2 code and -O3 -mcpu=pentium4 -msse2 13890 * [288]10673 ICE for x86-64 on freebsd libc vfprintf.c source 13891 * [289]11044 [x86] out of range loop instructions for FP code on K6 13892 * [290]11089 ICE: instantiate_virtual_regs_lossage while using SSE 13893 built-ins 13894 * [291]11420 [x86_64] gcc generates invalid asm code when "-O -fPIC" 13895 is used 13896 13897 SPARC- or Solaris- specific 13898 13899 * [292]9362 solaris 'as' dies when fed .s and "-gstabs" 13900 * [293]10142 [SPARC64] gcc produces wrong code when passing 13901 structures by value 13902 * [294]10663 New configure check aborts with Sun tools. 13903 * [295]10835 combinatorial explosion in scheduler on HyperSPARC 13904 * [296]10876 ICE in calculate_giv_inc when building KDE 13905 * [297]10955 wrong code at -O3 for structure argument in context of 13906 structure return 13907 * [298]11018 -mcpu=ultrasparc busts tar-1.13.25 13908 * [299]11556 [sparc64] ICE in gen_reg_rtx() while compiling 2.6.x 13909 Linux kernel 13910 13911 ia64 specific 13912 13913 * [300]10907 gcc violates the ia64 ABI (GP must be preserved) 13914 * [301]11320 scheduler bug (in machine depended reorganization pass) 13915 * [302]11599 bug with conditional and __builtin_prefetch 13916 13917 PowerPC specific 13918 13919 * [303]9745 [powerpc] gcc mis-compiles libmcrypt (alias problem 13920 during loop) 13921 * [304]10871 error in rs6000_stack_info save_size computation 13922 * [305]11440 gcc mis-compiles c++ code (libkhtml) with -O2, -fno-gcse 13923 cures it 13924 13925 m68k-specific 13926 13927 * [306]7594 [m68k] ICE on legal code associated with simplify-rtx 13928 * [307]10557 [m68k] ICE in subreg_offset_representable_p 13929 * [308]11054 [m68k] ICE in reg_overlap_mentioned_p 13930 13931 ARM-specific 13932 13933 * [309]10834 [arm] GCC 3.3 still generates incorrect instructions for 13934 functions with __attribute__ ((interrupt ("IRQ"))) 13935 * [310]10842 [arm] Clobbered link register is copied to pc under 13936 certain circumstances 13937 * [311]11052 [arm] noce_process_if_block() can lose REG_INC notes 13938 * [312]11183 [arm] ICE in change_address_1 (3.3) / subreg_hard_regno 13939 (3.4) 13940 13941 MIPS-specific 13942 13943 * [313]11084 ICE in propagate_one_insn, in flow.c 13944 13945 SH-specific 13946 13947 * [314]10331 can't compile c++ part of gcc cross compiler for sh-elf 13948 * [315]10413 [SH] ICE in reload_cse_simplify_operands, in reload1.c 13949 * [316]11096 i686-linux to sh-linux cross compiler fails to compile 13950 C++ files 13951 13952 GNU/Linux (or Hurd?) specific 13953 13954 * [317]2873 Bogus fixinclude of stdio.h from glibc 2.2.3 13955 13956 UnixWare specific 13957 13958 * [318]3163 configure bug: gcc/aclocal.m4 mmap test fails on UnixWare 13959 7.1.1 13960 13961 Cygwin (or mingw) specific 13962 13963 * [319]5287 ICE with dllimport attribute 13964 * [320]10148 [MingW/CygWin] Compiler dumps core 13965 13966 DJGPP specific 13967 13968 * [321]8787 GCC fails to emit .intel_syntax when invoked with 13969 -masm=intel on DJGPP 13970 13971 Darwin (and MacOS X) specific 13972 13973 * [322]10900 trampolines crash 13974 13975 Documentation 13976 13977 * [323]1607 (c++) Format attributes on methods undocumented 13978 * [324]4252 Invalid option `-fdump-translation-unit' 13979 * [325]4490 Clarify restrictions on -m96bit-long-double, 13980 -m128bit-long-double 13981 * [326]10355 document an issue with regparm attribute on some systems 13982 (e.g. Solaris) 13983 * [327]10726 (fortran) Documentation for function "IDate Intrinsic 13984 (Unix)" is wrong 13985 * [328]10805 document bug in old version of Sun assembler 13986 * [329]10815 warn against GNU binutils on AIX 13987 * [330]10877 document need for newer binutils on i?86-*-linux-gnu 13988 * [331]11280 Manual incorrect with respect to -freorder-blocks 13989 * [332]11466 Document -mlittle-endian and its restrictions for the 13990 sparc64 port 13991 13992 Testsuite bugs (compiler itself is not affected) 13993 13994 * [333]10737 newer bison causes g++.dg/parse/crash2.C to incorrectly 13995 report failure 13996 * [334]10810 gcc-3.3 fails make check: buffer overrun in 13997 test_demangle.c 13998 __________________________________________________________________ 13999 14000 GCC 3.3.2 14001 14002 Bug Fixes 14003 14004 This section lists the problem reports (PRs) from GCC's bug tracker 14005 that are known to be fixed in the 3.3.2 release. This list might not be 14006 complete (that is, it is possible that some PRs that have been fixed 14007 are not listed here). 14008 14009 Bootstrap failures and problems 14010 14011 * [335]8336 [SCO5] bootstrap config still tries to use COFF options 14012 * [336]9330 [alpha-osf] Bootstrap failure on Compaq Tru64 with 14013 --enable-threads=posix 14014 * [337]9631 [hppa64-linux] gcc-3.3 fails to bootstrap 14015 * [338]9877 fixincludes makes a bad sys/byteorder.h on svr5 (UnixWare 14016 7.1.1) 14017 * [339]11687 xstormy16-elf build fails in libf2c 14018 * [340]12263 [SGI IRIX] bootstrap fails during compile of 14019 libf2c/libI77/backspace.c 14020 * [341]12490 buffer overflow in scan-decls.c (during Solaris 9 14021 fix-header processing) 14022 14023 Internal compiler errors (multi-platform) 14024 14025 * [342]7277 Casting integers to vector types causes ICE 14026 * [343]7939 (c++) ICE on invalid function template specialization 14027 * [344]11063 (c++) ICE on parsing initialization list of const array 14028 member 14029 * [345]11207 ICE with negative index in array element designator 14030 * [346]11522 (fortran) g77 dwarf-2 ICE in 14031 add_abstract_origin_attribute 14032 * [347]11595 (c++) ICE on duplicate label definition 14033 * [348]11646 (c++) ICE in commit_one_edge_insertion with 14034 -fnon-call-exceptions -fgcse -O 14035 * [349]11665 ICE in struct initializer when taking address 14036 * [350]11852 (c++) ICE with bad struct initializer. 14037 * [351]11878 (c++) ICE in cp_expr_size 14038 * [352]11883 ICE with any -O on mercury-generated C code 14039 * [353]11991 (c++) ICE in cxx_incomplete_type_diagnostic, in 14040 cp/typeck2.c when applying typeid operator to template template 14041 parameter 14042 * [354]12146 ICE in lookup_template_function, in cp/pt.c 14043 * [355]12215 ICE in make_label_edge with -fnon-call-exceptions 14044 -fno-gcse -O2 14045 * [356]12369 (c++) ICE with templates and friends 14046 * [357]12446 ICE in emit_move_insn on complicated array reference 14047 * [358]12510 ICE in final_scan_insn 14048 * [359]12544 ICE with large parameters used in nested functions 14049 14050 C and optimization bugs 14051 14052 * [360]9862 spurious warnings with -W -finline-functions 14053 * [361]10962 lookup_field is a linear search on a linked list (can be 14054 slow if large struct) 14055 * [362]11370 -Wunreachable-code gives false complaints 14056 * [363]11637 invalid assembly with -fnon-call-exceptions 14057 * [364]11885 Problem with bitfields in packed structs 14058 * [365]12082 Inappropriate unreachable code warnings 14059 * [366]12180 Inline optimization fails for variadic function 14060 * [367]12340 loop unroller + gcse produces wrong code 14061 14062 C++ compiler and library 14063 14064 * [368]3907 nested template parameter collides with member name 14065 * [369]5293 confusing message when binding a temporary to a reference 14066 * [370]5296 [DR115] Pointers to functions and to template functions 14067 behave differently in deduction 14068 * [371]7939 ICE on function template specialization 14069 * [372]8656 Unable to assign function with __attribute__ and pointer 14070 return type to an appropriate variable 14071 * [373]10147 Confusing error message for invalid template function 14072 argument 14073 * [374]11400 std::search_n() makes assumptions about Size parameter 14074 * [375]11409 issues with using declarations, overloading, and 14075 built-in functions 14076 * [376]11740 ctype<wchar_t>::do_is(mask, wchar_t) doesn't handle 14077 multiple bits in mask 14078 * [377]11786 operator() call on variable in other namespace not 14079 recognized 14080 * [378]11867 static_cast ignores ambiguity 14081 * [379]11928 bug with conversion operators that are typedefs 14082 * [380]12114 Uninitialized memory accessed in dtor 14083 * [381]12163 static_cast + explicit constructor regression 14084 * [382]12181 Wrong code with comma operator and c++ 14085 * [383]12236 regparm and fastcall messes up parameters 14086 * [384]12266 incorrect instantiation of unneeded template during 14087 overload resolution 14088 * [385]12296 istream::peek() doesn't set eofbit 14089 * [386]12298 [sjlj exceptions] Stack unwind destroys 14090 not-yet-constructed object 14091 * [387]12369 ICE with templates and friends 14092 * [388]12337 apparently infinite loop in g++ 14093 * [389]12344 stdcall attribute ignored if function returns a pointer 14094 * [390]12451 missing(late) class forward declaration in cxxabi.h 14095 * [391]12486 g++ accepts invalid use of a qualified name 14096 14097 x86 specific (Intel/AMD) 14098 14099 * [392]8869 [x86 MMX] ICE with const variable optimization and MMX 14100 builtins 14101 * [393]9786 ICE in fixup_abnormal_edges with -fnon-call-exceptions 14102 -O2 14103 * [394]11689 g++3.3 emits un-assembleable code for k6 architecture 14104 * [395]12116 [k6] Invalid assembly output values with X-MAME code 14105 * [396]12070 ICE converting between double and long double with 14106 -msoft-float 14107 14108 ia64-specific 14109 14110 * [397]11184 [ia64 hpux] ICE on __builtin_apply building libobjc 14111 * [398]11535 __builtin_return_address may not work on ia64 14112 * [399]11693 [ia64] ICE in gen_nop_type 14113 * [400]12224 [ia64] Thread-local storage doesn't work 14114 14115 PowerPC-specific 14116 14117 * [401]11087 [powerpc64-linux] GCC miscompiles raid1.c from linux 14118 kernel 14119 * [402]11319 loop miscompiled on ppc32 14120 * [403]11949 ICE Compiler segfault with ffmpeg -maltivec code 14121 14122 SPARC-specific 14123 14124 * [404]11662 wrong code for expr. with cast to long long and 14125 exclusive or 14126 * [405]11965 invalid assembler code for a shift < 32 operation 14127 * [406]12301 (c++) stack corruption when a returned expression throws 14128 an exception 14129 14130 Alpha-specific 14131 14132 * [407]11717 [alpha-linux] unrecognizable insn compiling for.c of 14133 kernel 2.4.22-pre8 14134 14135 HPUX-specific 14136 14137 * [408]11313 problem with #pragma weak and static inline functions 14138 * [409]11712 __STDC_EXT__ not defined for C++ by default anymore? 14139 14140 Solaris specific 14141 14142 * [410]12166 Profiled programs crash if PROFDIR is set 14143 14144 Solaris-x86 specific 14145 14146 * [411]12101 i386 Solaris no longer works with GNU as? 14147 14148 Miscellaneous embedded target-specific bugs 14149 14150 * [412]10988 [m32r-elf] wrong blockmove code with -O3 14151 * [413]11805 [h8300-unknown-coff] [H8300] ICE for simple code with 14152 -O2 14153 * [414]11902 [sh4] spec file improperly inserts rpath even when none 14154 needed 14155 * [415]11903 [sh4] -pthread fails to link due to error in spec file 14156 on sh4 14157 __________________________________________________________________ 14158 14159 GCC 3.3.3 14160 14161 Minor features 14162 14163 In addition to the bug fixes documented below, this release contains 14164 few minor features such as: 14165 * Support for --with-sysroot 14166 * Support for automatic detection of executable stacks 14167 * Support for SSE3 instructions 14168 * Support for thread local storage debugging under GDB on S390 14169 14170 Bug Fixes 14171 14172 This section lists the problem reports (PRs) from GCC's bug tracker 14173 that are known to be fixed in the 3.3.3 release. This list might not be 14174 complete (that is, it is possible that some PRs that have been fixed 14175 are not listed here). 14176 14177 Bootstrap failures and issues 14178 14179 * [416]11890 Building cross gcc-3.3.1 for sparc-sun-solaris2.6 fails 14180 * [417]12399 boehm-gc fails (when building a cross compiler): libtool 14181 unable to infer tagged configuration 14182 * [418]13068 mklibgcc.in doesn't handle multi-level multilib 14183 subdirectories properly 14184 14185 Internal compiler errors (multi-platform) 14186 14187 * [419]10060 ICE (stack overflow) on huge file (300k lines) due to 14188 recursive behaviour of copy_rtx_if_shared, in emit_rtl.c 14189 * [420]10555 (c++) ICE on undefined template argument 14190 * [421]10706 (c++) ICE in mangle_class_name_for_template 14191 * [422]11496 (fortran) error in flow_loops_find when -funroll-loops 14192 active 14193 * [423]11741 ICE in pre_insert_copy_insn, in gcse.c 14194 * [424]12440 GCC crashes during compilation of quicktime4linux 2.0.0 14195 * [425]12632 (fortran) -fbounds-check ICE 14196 * [426]12712 (c++) ICE on short legit C++ code fragment with gcc 14197 3.3.2 14198 * [427]12726 (c++) ICE (segfault) on trivial code 14199 * [428]12890 (c++) ICE on compilation of class with throwing method 14200 * [429]12900 (c++) ICE in rtl_verify_flow_info_1 14201 * [430]13060 (fortran) ICE in fixup_var_refs_1, in function.c on 14202 correct code with -O2 -fno-force-mem 14203 * [431]13289 (c++) ICE in regenerate_decl_from_template on recursive 14204 template 14205 * [432]13318 ICE: floating point exception in the loop optimizer 14206 * [433]13392 (c++) ICE in convert_from_eh_region_ranges_1, in 14207 except.c 14208 * [434]13574 (c++) invalid array default initializer in class lets 14209 gcc consume all memory and die 14210 * [435]13475 ICE on SIMD variables with partial value initialization 14211 * [436]13797 (c++) ICE on invalid template parameter 14212 * [437]13824 (java) gcj SEGV with simple .java program 14213 14214 C and optimization bugs 14215 14216 * [438]8776 loop invariants are not removed (most likely) 14217 * [439]10339 [sparc,ppc,ppc64] Invalid optimization: replacing 14218 strncmp by memcmp 14219 * [440]11350 undefined labels with -Os -fPIC 14220 * [441]12826 Optimizer removes reference through volatile pointer 14221 * [442]12500 stabs debug info: void no longer a predefined / builtin 14222 type 14223 * [443]12941 builtin-bitops-1.c miscompilation (latent bug) 14224 * [444]12953 tree inliner bug (in inline_forbidden_p) and fix 14225 * [445]13041 linux-2.6/sound/core/oss/rate.c miscompiled 14226 * [446]13507 spurious printf format warning 14227 * [447]13382 Type information for const pointer disappears during 14228 optimization. 14229 * [448]13394 noreturn attribute ignored on recursive invokation 14230 * [449]13400 Compiled code crashes storing to read-only location 14231 * [450]13521 Endless loop in calculate_global_regs_live 14232 14233 C++ compiler and library 14234 14235 Some of the bug fixes in this list were made to implement decisions 14236 that the ISO C++ standards committee has made concerning several defect 14237 reports (DRs). Links in the list below point to detailed discussion of 14238 the relevant defect report. 14239 * [451]2094 unimplemented: use of `ptrmem_cst' in template type 14240 unification 14241 * [452]2294 using declaration confusion 14242 * [453]5050 template instantiation depth exceeds limit: recursion 14243 problem? 14244 * [454]9371 Bad exception handling in 14245 i/ostream::operator>>/<<(streambuf*) 14246 * [455]9546 bad exception handling in ostream members 14247 * [456]10081 basic_ios::_M_cache_locale leaves NULL members in the 14248 face of unknown locales 14249 * [457]10093 [458][DR 61] Setting failbit in exceptions doesn't work 14250 * [459]10095 istream::operator>>(int&) sets ios::badbit when 14251 ios::failbit is set. 14252 * [460]11554 Warning about reordering of initializers doesn't mention 14253 location of constructor 14254 * [461]12297 istream::sentry::sentry() handles eof() incorrectly. 14255 * [462]12352 Exception safety problems in src/localename.cc 14256 * [463]12438 Memory leak in locale::combine() 14257 * [464]12540 Memory leak in locale::locale(const char*) 14258 * [465]12594 DRs [466]60 [TC] and [467]63 [TC] not implemented 14259 * [468]12657 Resolution of [469]DR 292 (WP) still unimplemented 14260 * [470]12696 memory eating infinite loop in diagnostics (error 14261 recovery problem) 14262 * [471]12815 Code compiled with optimization behaves unexpectedly 14263 * [472]12862 Conflicts between typedefs/enums and namespace member 14264 declarations 14265 * [473]12926 Wrong value after assignment in initialize list using 14266 bit-fields 14267 * [474]12967 Resolution of [475]DR 300 [WP] still unimplemented 14268 * [476]12971 Resolution of [477]DR 328 [WP] still unimplemented 14269 * [478]13007 basic_streambuf::pubimbue, imbue wrong 14270 * [479]13009 Implicitly-defined assignment operator writes to wrong 14271 memory 14272 * [480]13057 regparm attribute not applied to destructor 14273 * [481]13070 -Wformat option ignored in g++ 14274 * [482]13081 forward template declarations in <complex> let inlining 14275 fail 14276 * [483]13239 Assertion does not seem to work correctly anymore 14277 * [484]13262 "xxx is private within this context" when initializing a 14278 self-contained template class 14279 * [485]13290 simple typo in concept checking for std::generate_n 14280 * [486]13323 Template code does not compile in presence of typedef 14281 * [487]13369 __verify_grouping (and __add_grouping?) not correct 14282 * [488]13371 infinite loop with packed struct and inlining 14283 * [489]13445 Template argument replacement "dereferences" a typedef 14284 * [490]13461 Fails to access protected-ctor from public constant 14285 * [491]13462 Non-standard-conforming type set::pointer 14286 * [492]13478 gcc uses wrong constructor to initialize a const 14287 reference 14288 * [493]13544 "conflicting types" for enums in different scopes 14289 * [494]13650 string::compare should not (always) use 14290 traits_type::length() 14291 * [495]13683 bogus warning about passing non-PODs through ellipsis 14292 * [496]13688 Derived class is denied access to protected base class 14293 member class 14294 * [497]13774 Member variable cleared in virtual multiple inheritance 14295 class 14296 * [498]13884 Protect sstream.tcc from extern template use 14297 14298 Java compiler and library 14299 14300 * [499]10746 [win32] garbage collection crash in GCJ 14301 14302 Objective-C compiler and library 14303 14304 * [500]11433 Crash due to dereferencing null pointer when querying 14305 protocol 14306 14307 Fortran compiler and library 14308 14309 * [501]12633 logical expression gives incorrect result with 14310 -fugly-logint option 14311 * [502]13037 [gcse-lm] g77 generates incorrect code 14312 * [503]13213 Hex constant problem when compiling with -fugly-logint 14313 and -ftypeless-boz 14314 14315 x86-specific (Intel/AMD) 14316 14317 * [504]4490 ICE with -m128bit-long-double 14318 * [505]12292 [x86_64] ICE: RTL check: expected code `const_int', have 14319 `reg' in make_field_assignment, in combine.c 14320 * [506]12441 ICE: can't find a register to spill 14321 * [507]12943 array static-init failure under -fpic, -fPIC 14322 * [508]13608 Incorrect code with -O3 -ffast-math 14323 14324 PowerPC-specific 14325 14326 * [509]11598 testcase gcc.dg/20020118-1.c fails runtime check of 14327 __attribute__((aligned(16))) 14328 * [510]11793 ICE in extract_insn, in recog.c (const_vector's) 14329 * [511]12467 vmsumubm emitted when vmsummbm appropriate (typo in 14330 altivec.md) 14331 * [512]12537 g++ generates writeable text sections 14332 14333 SPARC-specific 14334 14335 * [513]12496 wrong result for __atomic_add(&value, -1) when using -O0 14336 -m64 14337 * [514]12865 mprotect call to make trampoline executable may fail 14338 * [515]13354 ICE in sparc_emit_set_const32 14339 14340 ARM-specific 14341 14342 * [516]10467 [arm] ICE in pre_insert_copy_insn, 14343 14344 ia64-specific 14345 14346 * [517]11226 ICE passing struct arg with two floats 14347 * [518]11227 ICE for _Complex float, _Complex long double args 14348 * [519]12644 GCC 3.3.2 fails to compile glibc on ia64 14349 * [520]13149 build gcc-3.3.2 1305 error:unrecognizable insn 14350 * Various fixes for libunwind 14351 14352 Alpha-specific 14353 14354 * [521]12654 Incorrect comparison code generated for Alpha 14355 * [522]12965 SEGV+ICE in cc1plus on alpha-linux with -O2 14356 * [523]13031 ICE (unrecognizable insn) when building gnome-libs-1.4.2 14357 14358 HPPA-specific 14359 14360 * [524]11634 [hppa] ICE in verify_local_live_at_start, in flow.c 14361 * [525]12158 [hppa] compilation does not terminate at -O1 14362 14363 S390-specific 14364 14365 * [526]11992 Wrong built-in code for memcmp with length 1<<24: only 14366 (1<<24)-1 possible for CLCL-Instruction 14367 14368 SH-specific 14369 14370 * [527]9365 segfault in gen_far_branch (config/sh/sh.c) 14371 * [528]10392 optimizer generates faulty array indexing 14372 * [529]11322 SH profiler outputs multiple definitions of symbol 14373 * [530]13069 gcc/config/sh/rtems.h broken 14374 * [531]13302 Putting a va_list in a struct causes seg fault 14375 * [532]13585 Incorrect optimization of call to sfunc 14376 * Fix inappropriately exported libgcc functions from the shared 14377 library 14378 14379 Other embedded target specific 14380 14381 * [533]8916 [mcore] unsigned char assign gets hosed. 14382 * [534]11576 [h8300] ICE in change_address_1, in emit-rtl.c 14383 * [535]13122 [h8300] local variable gets corrupted by function call 14384 when -fomit-frame-pointer is given 14385 * [536]13256 [cris] strict_low_part mistreated in delay slots 14386 * [537]13373 [mcore] optimization with -frerun-cse-after-loop 14387 -fexpensive-optimizations produces wrong code on mcore 14388 14389 GNU HURD-specific 14390 14391 * [538]12561 gcc/config/t-gnu needs updating to work with 14392 --with-sysroot 14393 14394 Tru64 Unix specific 14395 14396 * [539]6243 testsuite fails almost all tests due to no libintl in 14397 LD_LIBRARY_PATH during test. 14398 * [540]11397 weak aliases broken on Tru64 UNIX 14399 14400 AIX-specific 14401 14402 * [541]12505 build failure due to defines of uchar in cpphash.h and 14403 sys/types.h 14404 * [542]13150 WEAK symbols not exported by collect2 14405 14406 IRIX-specific 14407 14408 * [543]12666 fixincludes problem on IRIX 6.5.19m 14409 14410 Solaris-specific 14411 14412 * [544]12969 Including sys/byteorder.h breaks configure checks 14413 14414 Testsuite problems (compiler is not affected) 14415 14416 * [545]10819 testsuite creates CR+LF on compiler version lines in 14417 test summary files 14418 * [546]11612 abi_check not finding correct libgcc_s.so.1 14419 14420 Miscellaneous 14421 14422 * [547]13211 using -###, incorrect warnings about unused linker file 14423 are produced 14424 __________________________________________________________________ 14425 14426 GCC 3.3.4 14427 14428 This is the [548]list of problem reports (PRs) from GCC's bug tracking 14429 system that are known to be fixed in the 3.3.4 release. This list might 14430 not be complete (that is, it is possible that some PRs that have been 14431 fixed are not listed here). 14432 __________________________________________________________________ 14433 14434 GCC 3.3.5 14435 14436 This is the [549]list of problem reports (PRs) from GCC's bug tracking 14437 system that are known to be fixed in the 3.3.5 release. This list might 14438 not be complete (that is, it is possible that some PRs that have been 14439 fixed are not listed here). 14440 __________________________________________________________________ 14441 14442 GCC 3.3.6 14443 14444 This is the [550]list of problem reports (PRs) from GCC's bug tracking 14445 system that are known to be fixed in the 3.3.6 release. This list might 14446 not be complete (that is, it is possible that some PRs that have been 14447 fixed are not listed here). 14448 14449 14450 For questions related to the use of GCC, please consult these web 14451 pages and the [551]GCC manuals. If that fails, the 14452 [552]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 14453 web pages and the development of GCC are welcome on our developer 14454 list at [553]gcc (a] gcc.gnu.org. All of [554]our lists have public 14455 archives. 14456 14457 Copyright (C) [555]Free Software Foundation, Inc. Verbatim copying and 14458 distribution of this entire article is permitted in any medium, 14459 provided this notice is preserved. 14460 14461 These pages are [556]maintained by the GCC team. Last modified 14462 2018-09-30[557]. 14463 14464 References 14465 14466 1. http://gcc.gnu.org/gcc-3.3/changes.html#3.3.6 14467 2. http://gcc.gnu.org/gcc-3.1/changes.html#obsolete_systems 14468 3. http://gcc.gnu.org/gcc-3.3/changes.html#obsolete_systems 14469 4. http://gcc.gnu.org/gcc-3.3/changes.html#nonnull_attribute 14470 5. http://gcc.gnu.org/news/dfa.html 14471 6. http://gcc.gnu.org/c99status.html 14472 7. https://gcc.gnu.org/onlinedocs/gcc-3.3.6/g77/News.html 14473 8. https://gcc.gnu.org/PR10140 14474 9. https://gcc.gnu.org/PR10198 14475 10. https://gcc.gnu.org/PR10338 14476 11. https://gcc.gnu.org/PR3581 14477 12. https://gcc.gnu.org/PR4382 14478 13. https://gcc.gnu.org/PR5533 14479 14. https://gcc.gnu.org/PR6387 14480 15. https://gcc.gnu.org/PR6412 14481 16. https://gcc.gnu.org/PR6620 14482 17. https://gcc.gnu.org/PR6663 14483 18. https://gcc.gnu.org/PR7068 14484 19. https://gcc.gnu.org/PR7083 14485 20. https://gcc.gnu.org/PR7647 14486 21. https://gcc.gnu.org/PR7675 14487 22. https://gcc.gnu.org/PR7718 14488 23. https://gcc.gnu.org/PR8116 14489 24. https://gcc.gnu.org/PR8358 14490 25. https://gcc.gnu.org/PR8511 14491 26. https://gcc.gnu.org/PR8564 14492 27. https://gcc.gnu.org/PR8660 14493 28. https://gcc.gnu.org/PR8766 14494 29. https://gcc.gnu.org/PR8803 14495 30. https://gcc.gnu.org/PR8846 14496 31. https://gcc.gnu.org/PR8906 14497 32. https://gcc.gnu.org/PR9216 14498 33. https://gcc.gnu.org/PR9261 14499 34. https://gcc.gnu.org/PR9263 14500 35. https://gcc.gnu.org/PR9429 14501 36. https://gcc.gnu.org/PR9516 14502 37. https://gcc.gnu.org/PR9600 14503 38. https://gcc.gnu.org/PR9629 14504 39. https://gcc.gnu.org/PR9672 14505 40. https://gcc.gnu.org/PR9749 14506 41. https://gcc.gnu.org/PR9794 14507 42. https://gcc.gnu.org/PR9829 14508 43. https://gcc.gnu.org/PR9916 14509 44. https://gcc.gnu.org/PR9936 14510 45. https://gcc.gnu.org/PR10262 14511 46. https://gcc.gnu.org/PR10278 14512 47. https://gcc.gnu.org/PR10446 14513 48. https://gcc.gnu.org/PR10451 14514 49. https://gcc.gnu.org/PR10506 14515 50. https://gcc.gnu.org/PR10549 14516 51. https://gcc.gnu.org/PR2001 14517 52. https://gcc.gnu.org/PR2391 14518 53. https://gcc.gnu.org/PR2960 14519 54. https://gcc.gnu.org/PR4046 14520 55. https://gcc.gnu.org/PR6405 14521 56. https://gcc.gnu.org/PR6798 14522 57. https://gcc.gnu.org/PR6871 14523 58. https://gcc.gnu.org/PR6909 14524 59. https://gcc.gnu.org/PR7189 14525 60. https://gcc.gnu.org/PR7642 14526 61. https://gcc.gnu.org/PR8634 14527 62. https://gcc.gnu.org/PR8750 14528 63. https://gcc.gnu.org/PR2161 14529 64. https://gcc.gnu.org/PR4319 14530 65. https://gcc.gnu.org/PR8602 14531 66. https://gcc.gnu.org/PR9177 14532 67. https://gcc.gnu.org/PR9853 14533 68. https://gcc.gnu.org/PR45 14534 69. https://gcc.gnu.org/PR3784 14535 70. https://gcc.gnu.org/PR764 14536 71. https://gcc.gnu.org/PR5116 14537 72. https://gcc.gnu.org/PR2862 14538 73. https://gcc.gnu.org/PR3663 14539 74. https://gcc.gnu.org/PR3797 14540 75. https://gcc.gnu.org/PR3948 14541 76. https://gcc.gnu.org/PR4137 14542 77. https://gcc.gnu.org/PR4361 14543 78. https://gcc.gnu.org/PR4802 14544 79. https://gcc.gnu.org/PR5837 14545 80. https://gcc.gnu.org/PR4803 14546 81. https://gcc.gnu.org/PR5094 14547 82. https://gcc.gnu.org/PR5730 14548 83. https://gcc.gnu.org/PR6713 14549 84. https://gcc.gnu.org/PR7015 14550 85. https://gcc.gnu.org/PR7086 14551 86. https://gcc.gnu.org/PR7099 14552 87. https://gcc.gnu.org/PR7247 14553 88. https://gcc.gnu.org/PR7441 14554 89. https://gcc.gnu.org/PR7768 14555 90. https://gcc.gnu.org/PR7804 14556 91. https://gcc.gnu.org/PR8099 14557 92. https://gcc.gnu.org/PR8117 14558 93. https://gcc.gnu.org/PR8205 14559 94. https://gcc.gnu.org/PR8645 14560 95. https://gcc.gnu.org/PR8724 14561 96. https://gcc.gnu.org/PR8805 14562 97. https://gcc.gnu.org/PR8691 14563 98. https://gcc.gnu.org/PR8700 14564 99. https://gcc.gnu.org/PR8724 14565 100. https://gcc.gnu.org/PR8949 14566 101. https://gcc.gnu.org/PR9016 14567 102. https://gcc.gnu.org/PR9053 14568 103. https://gcc.gnu.org/PR9152 14569 104. https://gcc.gnu.org/PR9182 14570 105. https://gcc.gnu.org/PR9297 14571 106. https://gcc.gnu.org/PR9318 14572 107. https://gcc.gnu.org/PR9320 14573 108. https://gcc.gnu.org/PR9400 14574 109. https://gcc.gnu.org/PR9424 14575 110. https://gcc.gnu.org/PR9425 14576 111. https://gcc.gnu.org/PR9439 14577 112. https://gcc.gnu.org/PR9474 14578 113. https://gcc.gnu.org/PR9548 14579 114. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#231 14580 115. https://gcc.gnu.org/PR9555 14581 116. https://gcc.gnu.org/PR9561 14582 117. https://gcc.gnu.org/PR9563 14583 118. https://gcc.gnu.org/PR9582 14584 119. https://gcc.gnu.org/PR9622 14585 120. https://gcc.gnu.org/PR9683 14586 121. https://gcc.gnu.org/PR9791 14587 122. https://gcc.gnu.org/PR9817 14588 123. https://gcc.gnu.org/PR9825 14589 124. https://gcc.gnu.org/PR9826 14590 125. https://gcc.gnu.org/PR9924 14591 126. https://gcc.gnu.org/PR9946 14592 127. https://gcc.gnu.org/PR9964 14593 128. https://gcc.gnu.org/PR9988 14594 129. https://gcc.gnu.org/PR10033 14595 130. https://gcc.gnu.org/PR10097 14596 131. https://gcc.gnu.org/PR10132 14597 132. https://gcc.gnu.org/PR10180 14598 133. https://gcc.gnu.org/PR10199 14599 134. https://gcc.gnu.org/PR10300 14600 135. https://gcc.gnu.org/PR10427 14601 136. https://gcc.gnu.org/PR10503 14602 137. https://gcc.gnu.org/PR5956 14603 138. https://gcc.gnu.org/PR1832 14604 139. https://gcc.gnu.org/PR3924 14605 140. https://gcc.gnu.org/PR5634 14606 141. https://gcc.gnu.org/PR6367 14607 142. https://gcc.gnu.org/PR6491 14608 143. https://gcc.gnu.org/PR6742 14609 144. https://gcc.gnu.org/PR7113 14610 145. https://gcc.gnu.org/PR7236 14611 146. https://gcc.gnu.org/PR7278 14612 147. https://gcc.gnu.org/PR7384 14613 148. https://gcc.gnu.org/PR7388 14614 149. https://gcc.gnu.org/PR8587 14615 150. https://gcc.gnu.org/PR9038 14616 151. https://gcc.gnu.org/PR10197 14617 152. https://gcc.gnu.org/PR6005 14618 153. https://gcc.gnu.org/PR6389 14619 154. https://gcc.gnu.org/PR6576 14620 155. https://gcc.gnu.org/PR6652 14621 156. https://gcc.gnu.org/PR7060 14622 157. https://gcc.gnu.org/PR7073 14623 158. https://gcc.gnu.org/PR7180 14624 159. https://gcc.gnu.org/PR7416 14625 160. https://gcc.gnu.org/PR7570 14626 161. https://gcc.gnu.org/PR7578 14627 162. https://gcc.gnu.org/PR7611 14628 163. https://gcc.gnu.org/PR7709 14629 164. https://gcc.gnu.org/PR7766 14630 165. https://gcc.gnu.org/PR7785 14631 166. https://gcc.gnu.org/PR7786 14632 167. https://gcc.gnu.org/PR8142 14633 168. https://gcc.gnu.org/PR8234 14634 169. https://gcc.gnu.org/PR8415 14635 170. https://gcc.gnu.org/PR8481 14636 171. https://gcc.gnu.org/PR8593 14637 172. https://gcc.gnu.org/PR8759 14638 173. https://gcc.gnu.org/PR8997 14639 174. https://gcc.gnu.org/PR9253 14640 175. https://gcc.gnu.org/PR9254 14641 176. https://gcc.gnu.org/PR9271 14642 177. https://gcc.gnu.org/PR6767 14643 178. https://gcc.gnu.org/PR9911 14644 179. https://gcc.gnu.org/PR10020 14645 180. https://gcc.gnu.org/PR10546 14646 181. https://gcc.gnu.org/PR7029 14647 182. https://gcc.gnu.org/PR2903 14648 183. https://gcc.gnu.org/PR7873 14649 184. https://gcc.gnu.org/PR7680 14650 185. https://gcc.gnu.org/PR8705 14651 186. https://gcc.gnu.org/PR9986 14652 187. https://gcc.gnu.org/PR10056 14653 188. https://gcc.gnu.org/PR6744 14654 189. https://gcc.gnu.org/PR7361 14655 190. https://gcc.gnu.org/PR9496 14656 191. https://gcc.gnu.org/PR7067 14657 192. https://gcc.gnu.org/PR8480 14658 193. https://gcc.gnu.org/PR8784 14659 194. https://gcc.gnu.org/PR10315 14660 195. https://gcc.gnu.org/PR10267 14661 196. https://gcc.gnu.org/PR7916 14662 197. https://gcc.gnu.org/PR7926 14663 198. https://gcc.gnu.org/PR8555 14664 199. https://gcc.gnu.org/PR8994 14665 200. https://gcc.gnu.org/PR9426 14666 201. https://gcc.gnu.org/PR9806 14667 202. https://gcc.gnu.org/PR10077 14668 203. https://gcc.gnu.org/PR10233 14669 204. https://gcc.gnu.org/PR10286 14670 205. https://gcc.gnu.org/PR10308 14671 206. https://gcc.gnu.org/PR11272 14672 207. https://gcc.gnu.org/PR5754 14673 208. https://gcc.gnu.org/PR6597 14674 209. https://gcc.gnu.org/PR6949 14675 210. https://gcc.gnu.org/PR7053 14676 211. https://gcc.gnu.org/PR8164 14677 212. https://gcc.gnu.org/PR8384 14678 213. https://gcc.gnu.org/PR9559 14679 214. https://gcc.gnu.org/PR9649 14680 215. https://gcc.gnu.org/PR9864 14681 216. https://gcc.gnu.org/PR10432 14682 217. https://gcc.gnu.org/PR10475 14683 218. https://gcc.gnu.org/PR10635 14684 219. https://gcc.gnu.org/PR10661 14685 220. https://gcc.gnu.org/PR10700 14686 221. https://gcc.gnu.org/PR10712 14687 222. https://gcc.gnu.org/PR10796 14688 223. https://gcc.gnu.org/PR10890 14689 224. https://gcc.gnu.org/PR10939 14690 225. https://gcc.gnu.org/PR10956 14691 226. https://gcc.gnu.org/PR11041 14692 227. https://gcc.gnu.org/PR11059 14693 228. https://gcc.gnu.org/PR11083 14694 229. https://gcc.gnu.org/PR11105 14695 230. https://gcc.gnu.org/PR11149 14696 231. https://gcc.gnu.org/PR11228 14697 232. https://gcc.gnu.org/PR11282 14698 233. https://gcc.gnu.org/PR11301 14699 234. https://gcc.gnu.org/PR11308 14700 235. https://gcc.gnu.org/PR11473 14701 236. https://gcc.gnu.org/PR11503 14702 237. https://gcc.gnu.org/PR11513 14703 238. https://gcc.gnu.org/PR11198 14704 239. https://gcc.gnu.org/PR11304 14705 240. https://gcc.gnu.org/PR11381 14706 241. https://gcc.gnu.org/PR11536 14707 242. https://gcc.gnu.org/PR11557 14708 243. https://gcc.gnu.org/PR5897 14709 244. https://gcc.gnu.org/PR11279 14710 245. https://gcc.gnu.org/PR11022 14711 246. https://gcc.gnu.org/PR2330 14712 247. https://gcc.gnu.org/PR5388 14713 248. https://gcc.gnu.org/PR5390 14714 249. https://gcc.gnu.org/PR7877 14715 250. https://gcc.gnu.org/PR9393 14716 251. https://gcc.gnu.org/PR10032 14717 252. https://gcc.gnu.org/PR10468 14718 253. https://gcc.gnu.org/PR10527 14719 254. https://gcc.gnu.org/PR10679 14720 255. https://gcc.gnu.org/PR10682 14721 256. https://gcc.gnu.org/PR10689 14722 257. https://gcc.gnu.org/PR10845 14723 258. https://gcc.gnu.org/PR10849 14724 259. https://gcc.gnu.org/PR10888 14725 260. https://gcc.gnu.org/PR10929 14726 261. https://gcc.gnu.org/PR10931 14727 262. https://gcc.gnu.org/PR10940 14728 263. https://gcc.gnu.org/PR10968 14729 264. https://gcc.gnu.org/PR10990 14730 265. https://gcc.gnu.org/PR11039 14731 266. https://gcc.gnu.org/PR11062 14732 267. https://gcc.gnu.org/PR11095 14733 268. https://gcc.gnu.org/PR11098 14734 269. https://gcc.gnu.org/PR11137 14735 270. https://gcc.gnu.org/PR11154 14736 271. https://gcc.gnu.org/PR11329 14737 272. https://gcc.gnu.org/PR11332 14738 273. https://gcc.gnu.org/PR11431 14739 274. https://gcc.gnu.org/PR11528 14740 275. https://gcc.gnu.org/PR11546 14741 276. https://gcc.gnu.org/PR11567 14742 277. https://gcc.gnu.org/PR11645 14743 278. https://gcc.gnu.org/PR5179 14744 279. https://gcc.gnu.org/PR8204 14745 280. https://gcc.gnu.org/PR10838 14746 281. https://gcc.gnu.org/PR10886 14747 282. https://gcc.gnu.org/PR11349 14748 283. https://gcc.gnu.org/PR4823 14749 284. https://gcc.gnu.org/PR8878 14750 285. https://gcc.gnu.org/PR9815 14751 286. https://gcc.gnu.org/PR10402 14752 287. https://gcc.gnu.org/PR10504 14753 288. https://gcc.gnu.org/PR10673 14754 289. https://gcc.gnu.org/PR11044 14755 290. https://gcc.gnu.org/PR11089 14756 291. https://gcc.gnu.org/PR11420 14757 292. https://gcc.gnu.org/PR9362 14758 293. https://gcc.gnu.org/PR10142 14759 294. https://gcc.gnu.org/PR10663 14760 295. https://gcc.gnu.org/PR10835 14761 296. https://gcc.gnu.org/PR10876 14762 297. https://gcc.gnu.org/PR10955 14763 298. https://gcc.gnu.org/PR11018 14764 299. https://gcc.gnu.org/PR11556 14765 300. https://gcc.gnu.org/PR10907 14766 301. https://gcc.gnu.org/PR11320 14767 302. https://gcc.gnu.org/PR11599 14768 303. https://gcc.gnu.org/PR9745 14769 304. https://gcc.gnu.org/PR10871 14770 305. https://gcc.gnu.org/PR11440 14771 306. https://gcc.gnu.org/PR7594 14772 307. https://gcc.gnu.org/PR10557 14773 308. https://gcc.gnu.org/PR11054 14774 309. https://gcc.gnu.org/PR10834 14775 310. https://gcc.gnu.org/PR10842 14776 311. https://gcc.gnu.org/PR11052 14777 312. https://gcc.gnu.org/PR11183 14778 313. https://gcc.gnu.org/PR11084 14779 314. https://gcc.gnu.org/PR10331 14780 315. https://gcc.gnu.org/PR10413 14781 316. https://gcc.gnu.org/PR11096 14782 317. https://gcc.gnu.org/PR2873 14783 318. https://gcc.gnu.org/PR3163 14784 319. https://gcc.gnu.org/PR5287 14785 320. https://gcc.gnu.org/PR10148 14786 321. https://gcc.gnu.org/PR8787 14787 322. https://gcc.gnu.org/PR10900 14788 323. https://gcc.gnu.org/PR1607 14789 324. https://gcc.gnu.org/PR4252 14790 325. https://gcc.gnu.org/PR4490 14791 326. https://gcc.gnu.org/PR10355 14792 327. https://gcc.gnu.org/PR10726 14793 328. https://gcc.gnu.org/PR10805 14794 329. https://gcc.gnu.org/PR10815 14795 330. https://gcc.gnu.org/PR10877 14796 331. https://gcc.gnu.org/PR11280 14797 332. https://gcc.gnu.org/PR11466 14798 333. https://gcc.gnu.org/PR10737 14799 334. https://gcc.gnu.org/PR10810 14800 335. https://gcc.gnu.org/PR8336 14801 336. https://gcc.gnu.org/PR9330 14802 337. https://gcc.gnu.org/PR9631 14803 338. https://gcc.gnu.org/PR9877 14804 339. https://gcc.gnu.org/PR11687 14805 340. https://gcc.gnu.org/PR12263 14806 341. https://gcc.gnu.org/PR12490 14807 342. https://gcc.gnu.org/PR7277 14808 343. https://gcc.gnu.org/PR7939 14809 344. https://gcc.gnu.org/PR11063 14810 345. https://gcc.gnu.org/PR11207 14811 346. https://gcc.gnu.org/PR11522 14812 347. https://gcc.gnu.org/PR11595 14813 348. https://gcc.gnu.org/PR11646 14814 349. https://gcc.gnu.org/PR11665 14815 350. https://gcc.gnu.org/PR11852 14816 351. https://gcc.gnu.org/PR11878 14817 352. https://gcc.gnu.org/PR11883 14818 353. https://gcc.gnu.org/PR11991 14819 354. https://gcc.gnu.org/PR12146 14820 355. https://gcc.gnu.org/PR12215 14821 356. https://gcc.gnu.org/PR12369 14822 357. https://gcc.gnu.org/PR12446 14823 358. https://gcc.gnu.org/PR12510 14824 359. https://gcc.gnu.org/PR12544 14825 360. https://gcc.gnu.org/PR9862 14826 361. https://gcc.gnu.org/PR10962 14827 362. https://gcc.gnu.org/PR11370 14828 363. https://gcc.gnu.org/PR11637 14829 364. https://gcc.gnu.org/PR11885 14830 365. https://gcc.gnu.org/PR12082 14831 366. https://gcc.gnu.org/PR12180 14832 367. https://gcc.gnu.org/PR12340 14833 368. https://gcc.gnu.org/PR3907 14834 369. https://gcc.gnu.org/PR5293 14835 370. https://gcc.gnu.org/PR5296 14836 371. https://gcc.gnu.org/PR7939 14837 372. https://gcc.gnu.org/PR8656 14838 373. https://gcc.gnu.org/PR10147 14839 374. https://gcc.gnu.org/PR11400 14840 375. https://gcc.gnu.org/PR11409 14841 376. https://gcc.gnu.org/PR11740 14842 377. https://gcc.gnu.org/PR11786 14843 378. https://gcc.gnu.org/PR11867 14844 379. https://gcc.gnu.org/PR11928 14845 380. https://gcc.gnu.org/PR12114 14846 381. https://gcc.gnu.org/PR12163 14847 382. https://gcc.gnu.org/PR12181 14848 383. https://gcc.gnu.org/PR12236 14849 384. https://gcc.gnu.org/PR12266 14850 385. https://gcc.gnu.org/PR12296 14851 386. https://gcc.gnu.org/PR12298 14852 387. https://gcc.gnu.org/PR12369 14853 388. https://gcc.gnu.org/PR12337 14854 389. https://gcc.gnu.org/PR12344 14855 390. https://gcc.gnu.org/PR12451 14856 391. https://gcc.gnu.org/PR12486 14857 392. https://gcc.gnu.org/PR8869 14858 393. https://gcc.gnu.org/PR9786 14859 394. https://gcc.gnu.org/PR11689 14860 395. https://gcc.gnu.org/PR12116 14861 396. https://gcc.gnu.org/PR12070 14862 397. https://gcc.gnu.org/PR11184 14863 398. https://gcc.gnu.org/PR11535 14864 399. https://gcc.gnu.org/PR11693 14865 400. https://gcc.gnu.org/PR12224 14866 401. https://gcc.gnu.org/PR11087 14867 402. https://gcc.gnu.org/PR11319 14868 403. https://gcc.gnu.org/PR11949 14869 404. https://gcc.gnu.org/PR11662 14870 405. https://gcc.gnu.org/PR11965 14871 406. https://gcc.gnu.org/PR12301 14872 407. https://gcc.gnu.org/PR11717 14873 408. https://gcc.gnu.org/PR11313 14874 409. https://gcc.gnu.org/PR11712 14875 410. https://gcc.gnu.org/PR12166 14876 411. https://gcc.gnu.org/PR12101 14877 412. https://gcc.gnu.org/PR10988 14878 413. https://gcc.gnu.org/PR11805 14879 414. https://gcc.gnu.org/PR11902 14880 415. https://gcc.gnu.org/PR11903 14881 416. https://gcc.gnu.org/PR11890 14882 417. https://gcc.gnu.org/PR12399 14883 418. https://gcc.gnu.org/PR13068 14884 419. https://gcc.gnu.org/PR10060 14885 420. https://gcc.gnu.org/PR10555 14886 421. https://gcc.gnu.org/PR10706 14887 422. https://gcc.gnu.org/PR11496 14888 423. https://gcc.gnu.org/PR11741 14889 424. https://gcc.gnu.org/PR12440 14890 425. https://gcc.gnu.org/PR12632 14891 426. https://gcc.gnu.org/PR12712 14892 427. https://gcc.gnu.org/PR12726 14893 428. https://gcc.gnu.org/PR12890 14894 429. https://gcc.gnu.org/PR12900 14895 430. https://gcc.gnu.org/PR13060 14896 431. https://gcc.gnu.org/PR13289 14897 432. https://gcc.gnu.org/PR13318 14898 433. https://gcc.gnu.org/PR13392 14899 434. https://gcc.gnu.org/PR13574 14900 435. https://gcc.gnu.org/PR13475 14901 436. https://gcc.gnu.org/PR13797 14902 437. https://gcc.gnu.org/PR13824 14903 438. https://gcc.gnu.org/PR8776 14904 439. https://gcc.gnu.org/PR10339 14905 440. https://gcc.gnu.org/PR11350 14906 441. https://gcc.gnu.org/PR12826 14907 442. https://gcc.gnu.org/PR12500 14908 443. https://gcc.gnu.org/PR12941 14909 444. https://gcc.gnu.org/PR12953 14910 445. https://gcc.gnu.org/PR13041 14911 446. https://gcc.gnu.org/PR13507 14912 447. https://gcc.gnu.org/PR13382 14913 448. https://gcc.gnu.org/PR13394 14914 449. https://gcc.gnu.org/PR13400 14915 450. https://gcc.gnu.org/PR13521 14916 451. https://gcc.gnu.org/PR2094 14917 452. https://gcc.gnu.org/PR2294 14918 453. https://gcc.gnu.org/PR5050 14919 454. https://gcc.gnu.org/PR9371 14920 455. https://gcc.gnu.org/PR9546 14921 456. https://gcc.gnu.org/PR10081 14922 457. https://gcc.gnu.org/PR10093 14923 458. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#61 14924 459. https://gcc.gnu.org/PR10095 14925 460. https://gcc.gnu.org/PR11554 14926 461. https://gcc.gnu.org/PR12297 14927 462. https://gcc.gnu.org/PR12352 14928 463. https://gcc.gnu.org/PR12438 14929 464. https://gcc.gnu.org/PR12540 14930 465. https://gcc.gnu.org/PR12594 14931 466. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#60 14932 467. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#63 14933 468. https://gcc.gnu.org/PR12657 14934 469. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#292 14935 470. https://gcc.gnu.org/PR12696 14936 471. https://gcc.gnu.org/PR12815 14937 472. https://gcc.gnu.org/PR12862 14938 473. https://gcc.gnu.org/PR12926 14939 474. https://gcc.gnu.org/PR12967 14940 475. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html 14941 476. https://gcc.gnu.org/PR12971 14942 477. http://www.open-std.org/jtc1/sc22/wg21/docs/cwg_defects.html#328 14943 478. https://gcc.gnu.org/PR13007 14944 479. https://gcc.gnu.org/PR13009 14945 480. https://gcc.gnu.org/PR13057 14946 481. https://gcc.gnu.org/PR13070 14947 482. https://gcc.gnu.org/PR13081 14948 483. https://gcc.gnu.org/PR13239 14949 484. https://gcc.gnu.org/PR13262 14950 485. https://gcc.gnu.org/PR13290 14951 486. https://gcc.gnu.org/PR13323 14952 487. https://gcc.gnu.org/PR13369 14953 488. https://gcc.gnu.org/PR13371 14954 489. https://gcc.gnu.org/PR13445 14955 490. https://gcc.gnu.org/PR13461 14956 491. https://gcc.gnu.org/PR13462 14957 492. https://gcc.gnu.org/PR13478 14958 493. https://gcc.gnu.org/PR13544 14959 494. https://gcc.gnu.org/PR13650 14960 495. https://gcc.gnu.org/PR13683 14961 496. https://gcc.gnu.org/PR13688 14962 497. https://gcc.gnu.org/PR13774 14963 498. https://gcc.gnu.org/PR13884 14964 499. https://gcc.gnu.org/PR10746 14965 500. https://gcc.gnu.org/PR11433 14966 501. https://gcc.gnu.org/PR12633 14967 502. https://gcc.gnu.org/PR13037 14968 503. https://gcc.gnu.org/PR13213 14969 504. https://gcc.gnu.org/PR4490 14970 505. https://gcc.gnu.org/PR12292 14971 506. https://gcc.gnu.org/PR12441 14972 507. https://gcc.gnu.org/PR12943 14973 508. https://gcc.gnu.org/PR13608 14974 509. https://gcc.gnu.org/PR11598 14975 510. https://gcc.gnu.org/PR11793 14976 511. https://gcc.gnu.org/PR12467 14977 512. https://gcc.gnu.org/PR12537 14978 513. https://gcc.gnu.org/PR12496 14979 514. https://gcc.gnu.org/PR12865 14980 515. https://gcc.gnu.org/PR13354 14981 516. https://gcc.gnu.org/PR10467 14982 517. https://gcc.gnu.org/PR11226 14983 518. https://gcc.gnu.org/PR11227 14984 519. https://gcc.gnu.org/PR12644 14985 520. https://gcc.gnu.org/PR13149 14986 521. https://gcc.gnu.org/PR12654 14987 522. https://gcc.gnu.org/PR12965 14988 523. https://gcc.gnu.org/PR13031 14989 524. https://gcc.gnu.org/PR11634 14990 525. https://gcc.gnu.org/PR12158 14991 526. https://gcc.gnu.org/PR11992 14992 527. https://gcc.gnu.org/PR9365 14993 528. https://gcc.gnu.org/PR10392 14994 529. https://gcc.gnu.org/PR11322 14995 530. https://gcc.gnu.org/PR13069 14996 531. https://gcc.gnu.org/PR13302 14997 532. https://gcc.gnu.org/PR13585 14998 533. https://gcc.gnu.org/PR8916 14999 534. https://gcc.gnu.org/PR11576 15000 535. https://gcc.gnu.org/PR13122 15001 536. https://gcc.gnu.org/PR13256 15002 537. https://gcc.gnu.org/PR13373 15003 538. https://gcc.gnu.org/PR12561 15004 539. https://gcc.gnu.org/PR6243 15005 540. https://gcc.gnu.org/PR11397 15006 541. https://gcc.gnu.org/PR12505 15007 542. https://gcc.gnu.org/PR13150 15008 543. https://gcc.gnu.org/PR12666 15009 544. https://gcc.gnu.org/PR12969 15010 545. https://gcc.gnu.org/PR10819 15011 546. https://gcc.gnu.org/PR11612 15012 547. https://gcc.gnu.org/PR13211 15013 548. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.4 15014 549. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.5 15015 550. https://gcc.gnu.org/bugzilla/buglist.cgi?bug_status=RESOLVED&resolution=FIXED&target_milestone=3.3.6 15016 551. https://gcc.gnu.org/onlinedocs/ 15017 552. mailto:gcc-help (a] gcc.gnu.org 15018 553. mailto:gcc (a] gcc.gnu.org 15019 554. https://gcc.gnu.org/lists.html 15020 555. https://www.fsf.org/ 15021 556. https://gcc.gnu.org/about.html 15022 557. http://validator.w3.org/check/referer 15023 ====================================================================== 15024 http://gcc.gnu.org/gcc-3.2/index.html 15025 15026 GCC 3.2 Release Series 15027 15028 April 25, 2003 15029 15030 The [1]GNU project and the GCC developers are pleased to announce the 15031 release of GCC 3.2.3. 15032 15033 The purpose of the GCC 3.2 release series is to provide a stable 15034 platform for OS distributors to use building their next releases. A 15035 primary objective was to stabilize the C++ ABI; we believe that the 15036 interface to the compiler and the C++ standard library are now 15037 relatively stable. 15038 15039 Be aware that C++ code compiled by GCC 3.2.x will (in general) not 15040 interoperate with code compiled by GCC 3.1.1 or earlier. 15041 15042 Please refer to our [2]detailed list of news, caveats, and bug-fixes 15043 for further information. 15044 15045 This release series is no longer maintained. 15046 15047 Release History 15048 15049 GCC 3.2.3 15050 April 25, 2003 ([3]changes) 15051 15052 GCC 3.2.2 15053 February 5, 2003 ([4]changes) 15054 15055 GCC 3.2.1 15056 November 19, 2002 ([5]changes) 15057 15058 GCC 3.2 15059 August 14, 2002 ([6]changes) 15060 15061 References and Acknowledgements 15062 15063 GCC used to stand for the GNU C Compiler, but since the compiler 15064 supports several other languages aside from C, it now stands for the 15065 GNU Compiler Collection. 15066 15067 A list of [7]successful builds is updated as new information becomes 15068 available. 15069 15070 The GCC developers would like to thank the numerous people that have 15071 contributed new features, improvements, bug fixes, and other changes as 15072 well as test results to GCC. This [8]amazing group of volunteers is 15073 what makes GCC successful. 15074 15075 For additional information about GCC please refer to the [9]GCC project 15076 web site or contact the [10]GCC development mailing list. 15077 15078 To obtain GCC please use [11]our mirror sites, or our CVS server. 15079 15080 15081 For questions related to the use of GCC, please consult these web 15082 pages and the [12]GCC manuals. If that fails, the 15083 [13]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15084 web pages and the development of GCC are welcome on our developer 15085 list at [14]gcc (a] gcc.gnu.org. All of [15]our lists have public 15086 archives. 15087 15088 Copyright (C) [16]Free Software Foundation, Inc. Verbatim copying and 15089 distribution of this entire article is permitted in any medium, 15090 provided this notice is preserved. 15091 15092 These pages are [17]maintained by the GCC team. Last modified 15093 2018-09-30[18]. 15094 15095 References 15096 15097 1. http://www.gnu.org/ 15098 2. http://gcc.gnu.org/gcc-3.2/changes.html 15099 3. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 15100 4. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.2 15101 5. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.1 15102 6. http://gcc.gnu.org/gcc-3.2/changes.html#3.2 15103 7. http://gcc.gnu.org/gcc-3.2/buildstat.html 15104 8. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 15105 9. http://gcc.gnu.org/index.html 15106 10. mailto:gcc (a] gcc.gnu.org 15107 11. http://gcc.gnu.org/mirrors.html 15108 12. https://gcc.gnu.org/onlinedocs/ 15109 13. mailto:gcc-help (a] gcc.gnu.org 15110 14. mailto:gcc (a] gcc.gnu.org 15111 15. https://gcc.gnu.org/lists.html 15112 16. https://www.fsf.org/ 15113 17. https://gcc.gnu.org/about.html 15114 18. http://validator.w3.org/check/referer 15115 ====================================================================== 15116 http://gcc.gnu.org/gcc-3.2/changes.html 15117 15118 GCC 3.2 Release Series 15119 Changes, New Features, and Fixes 15120 15121 The latest release in the 3.2 release series is [1]GCC 3.2.3. 15122 15123 Caveats and New Features 15124 15125 Caveats 15126 15127 * The C++ compiler does not correctly zero-initialize 15128 pointers-to-data members. You must explicitly initialize them. For 15129 example: int S::*m(0); will work, but depending on 15130 default-initialization to zero will not work. This bug cannot be 15131 fixed in GCC 3.2 without inducing unacceptable risks. It will be 15132 fixed in GCC 3.3. 15133 * This GCC release is based on the GCC 3.1 sourcebase, and thus has 15134 all the [2]changes in the GCC 3.1 series. In addition, GCC 3.2 has 15135 a number of C++ ABI fixes which make its C++ compiler generate 15136 binary code which is incompatible with the C++ compilers found in 15137 earlier GCC releases, including GCC 3.1 and GCC 3.1.1. 15138 15139 Frontend Enhancements 15140 15141 C/C++/Objective-C 15142 15143 * The method of constructing the list of directories to be searched 15144 for header files has been revised. If a directory named by a -I 15145 option is a standard system include directory, the option is 15146 ignored to ensure that the default search order for system 15147 directories and the special treatment of system header files are 15148 not defeated. 15149 * The C and Objective-C compilers no longer accept the "Naming Types" 15150 extension (typedef foo = bar); it was already unavailable in C++. 15151 Code which uses it will need to be changed to use the "typeof" 15152 extension instead: typedef typeof(bar) foo. (We have removed this 15153 extension without a period of deprecation because it has caused the 15154 compiler to crash since version 3.0 and no one noticed until very 15155 recently. Thus we conclude it is not in widespread use.) 15156 15157 C++ 15158 15159 * GCC 3.2 fixed serveral differences between the C++ ABI implemented 15160 in GCC and the multi-vendor standard, but more have been found 15161 since the release. 3.2.1 adds a new warning, -Wabi, to warn about 15162 code which is affected by these bugs. We will fix these bugs in 15163 some future release, once we are confident that all have been 15164 found; until then, it is our intention to make changes to the ABI 15165 only if they are necessary for correct compilation of C++, as 15166 opposed to conformance to the ABI documents. 15167 * For details on how to build an ABI compliant compiler for GNU/Linux 15168 systems, check the [3]common C++ ABI page. 15169 15170 New Targets and Target Specific Improvements 15171 15172 IA-32 15173 15174 * Fixed a number of bugs in SSE and MMX intrinsics. 15175 * Fixed common compiler crashes with SSE instruction set enabled 15176 (implied by -march=pentium3, pentium4, athlon-xp) 15177 * __m128 and __m128i is not 128bit aligned when used in structures. 15178 15179 x86-64 15180 15181 * A bug whereby the compiler could generate bad code for bzero has 15182 been fixed. 15183 * ABI fixes (implying ABI incompatibilities with previous version in 15184 some corner cases) 15185 * Fixed prefetch code generation 15186 __________________________________________________________________ 15187 15188 GCC 3.2.3 15189 15190 3.2.3 is a bug fix release only; there are no new features that were 15191 not present in GCC 3.2.2. 15192 15193 Bug Fixes 15194 15195 This section lists the problem reports (PRs) from GCC's bug tracking 15196 system that are known to be fixed in the 3.2.3 release. This list might 15197 not be complete (that is, it is possible that some PRs that have been 15198 fixed are not listed here), and some of the titles have been changed to 15199 make them more clear. 15200 15201 Internal Compiler Errors (multi-platform) 15202 15203 * [4]3782: (c++) -quiet -fstats produces a segmentation fault in 15204 cc1plus 15205 * [5]6440: (c++) template specializations cause ICE 15206 * [6]7050: (c++) ICE on: (i ? get_string() : throw) 15207 * [7]7741: ICE on conflicting types (make_decl_rtl in varasm.c) 15208 * [8]7982: (c++) ICE due to infinite recursion (using STL set) 15209 * [9]8068: exceedingly high (infinite) memory usage 15210 * [10]8178: ICE with __builtin_ffs 15211 * [11]8396: ICE in copy_to_mode_reg, in explow.c 15212 * [12]8674: (c++) ICE in cp_expr_size, in cp/cp-lang.c 15213 * [13]9768: ICE when optimizing inline code at -O2 15214 * [14]9798: (c++) Infinite recursion (segfault) in 15215 cp/decl.c:push_using_directive with recursive using directives 15216 * [15]9799: mismatching structure initializer with nested flexible 15217 array member: ICE 15218 * [16]9928: ICE on duplicate enum declaration 15219 * [17]10114: ICE in mem_loc_descriptor, in dwarf2out.c (affects 15220 sparc, alpha) 15221 * [18]10352: ICE in find_reloads_toplev 15222 * [19]10336: ICE with -Wunreachable-code 15223 15224 C/optimizer bugs: 15225 15226 * [20]8224: Incorrect joining of signed and unsigned division 15227 * [21]8613: -O2 produces wrong code with builtin strlen and 15228 postincrements 15229 * [22]8828: gcc reports some code is unreachable when it is not 15230 * [23]9226: GCSE breaking argument passing 15231 * [24]9853: miscompilation of non-constant structure initializer 15232 * [25]9797: C99-style struct initializers are miscompiled 15233 * [26]9967: Some standard C function calls should not be replaced 15234 when optimizing for size 15235 * [27]10116: ce2: invalid merge of join_bb in the context of switch 15236 statements 15237 * [28]10171: wrong code for inlined function 15238 * [29]10175: -Wunreachable-code doesn't work for single lines 15239 15240 C++ compiler and library: 15241 15242 * [30]8316: Confusing diagnostic for code that misuses conversion 15243 operators 15244 * [31]9169: filebuf output fails if codecvt<>::out returns noconv 15245 * [32]9420: incomplete type incorrectly reported 15246 * [33]9459: typeof in return type specification of template not 15247 supported 15248 * [34]9507: filebuf::open handles ios_base::ate incorrectly 15249 * [35]9538: Out-of-bounds memory access in streambuf::sputbackc 15250 * [36]9602: Total confusion about template/friend/virtual/abstract 15251 * [37]9993: destructor not called for local object created within and 15252 returned from infinite loop 15253 * [38]10167: ieee_1003.1-2001 locale specialisations on a glibc-2.3.2 15254 system 15255 15256 Java compiler and library: 15257 15258 * [39]9652: libgcj build fails on irix6.5.1[78] 15259 * [40]10144: gas on solaris complains about bad .stabs lines for 15260 java, native as unaffected 15261 15262 x86-specific (Intel/AMD): 15263 15264 * [41]8746: gcc miscompiles Linux kernel ppa driver on x86 15265 * [42]9888: -mcpu=k6 -Os produces out of range loop instructions 15266 * [43]9638: Cross-build for target i386-elf and i586-pc-linux-gnu 15267 failed 15268 * [44]9954: Cross-build for target i586-pc-linux-gnu (--with-newlib) 15269 failed 15270 15271 SPARC-specific: 15272 15273 * [45]7784: [Sparc] ICE in extract_insn, in recog.c 15274 * [46]7796: sparc extra failure with -m64 on execute/930921-1.c in 15275 unroll.c 15276 * [47]8281: ICE when compiling with -O2 -fPIC for Ultrasparc 15277 * [48]8366: [Sparc] C testsuite failure with -m64 -fpic -O in 15278 execute/loop-2d.c 15279 * [49]8726: gcc -O2 miscompiles Samba 2.2.7 on 32-bit sparc 15280 * [50]9414: Scheduling bug on Ultrasparc 15281 * [51]10067: GCC-3.2.2 outputs invalid asm on sparc64 15282 15283 m68k-specific: 15284 15285 * [52]7248: broken "inclusive or" code 15286 * [53]8343: m68k-elf/rtems ICE at instantiate_virtual_regs_1 15287 15288 PowerPC-specific: 15289 15290 * [54]9732: Wrong code with -O2 -fPIC 15291 * [55]10073: ICE: powerpc cannot split insn 15292 15293 Alpha-specific: 15294 15295 * [56]7702: optimization problem on a DEC alpha under OSF1 15296 * [57]9671: gcc.3.2.2 does not build on a HP Tru64 Unix v5.1B system 15297 15298 HP-specific: 15299 15300 * [58]8694: <string> breaks <ctype.h> on HP-UX 10.20 (DUP: 9275) 15301 * [59]9953: (ada) gcc 3.2.x can't build 3.3-branch ada on HP-UX 10 15302 (missing symbol) 15303 * [60]10271: Floating point args don't get reloaded across function 15304 calls with -O2 15305 15306 MIPS specific: 15307 15308 * [61]6362: mips-irix6 gcc-3.1 C testsuite failure with -mips4 in 15309 compile/920501-4.c 15310 15311 CRIS specific: 15312 15313 * [62]10377: gcc-3.2.2 creates bad assembler code for cris 15314 15315 Miscellaneous and minor bugs: 15316 15317 * [63]6955: collect2 says "core dumped" when there is no core 15318 __________________________________________________________________ 15319 15320 GCC 3.2.2 15321 15322 Beginning with 3.2.2, GCC's Makefile suite supports redirection of make 15323 install by means of the DESTDIR variable. Parts of the GCC tree have 15324 featured that support long before, but now it is available even from 15325 the top level. 15326 15327 Other than that, GCC 3.2.2 is a bug fix release only; there are no new 15328 features that were not present in GCC 3.2.1. 15329 15330 Bug Fixes 15331 15332 On the following i386-based systems GCC 3.2.1 broke the C ABI wrt. 15333 functions returning structures: Cygwin, FreeBSD (GCC 3.2.1 as shipped 15334 with FreeBSD 5.0 does not have this problem), Interix, a.out-based 15335 GNU/Linux and NetBSD, OpenBSD, and Darwin. GCC 3.2.2 reverts this ABI 15336 change, and thus restores ABI-compatibility with previous releases 15337 (except GCC 3.2.1) on these platforms. 15338 15339 This section lists the problem reports (PRs) from GCC's bug tracking 15340 system that are known to be fixed in the 3.2.2 release. This list might 15341 not be complete (that is, it is possible that some PRs that have been 15342 fixed are not listed here) and some of the titles have been changed to 15343 make them more clear. 15344 15345 Internal Compiler Errors (multi-platform) 15346 15347 * [64]5919: (c++) ICE when passing variable array to template 15348 function 15349 * [65]7129: (c++) ICE with min/max assignment operators (<?= and >?=) 15350 * [66]7507: ICE with -O2 when address of called function is a 15351 complicated expression 15352 * [67]7622: ICE with nested inline functions if function's address is 15353 taken 15354 * [68]7681: (fortran) ICE in compensate_edge, in reg-stack.c (also PR 15355 [69]9258) 15356 * [70]8031: (c++) ICE in code comparing typeids and casting from 15357 virtual base 15358 * [71]8275: ICE in simplify_subreg 15359 * [72]8332: (c++) builtin strlen/template interaction causes ICE 15360 * [73]8372: (c++) ICE on explicit call of destructor 15361 * [74]8439: (c, not c++) empty struct causes ICE 15362 * [75]8442: (c++) ICE with nested template classes 15363 * [76]8518: ICE when compiling mplayer ("extern inline" issue) 15364 * [77]8615: (c++) ICE with out-of-range character constant template 15365 argument 15366 * [78]8663: (c++) ICE in cp_expr_size, at cp-lang.c:307 15367 * [79]8799: (c++) ICE: error reporting routines re-entered 15368 * [80]9328: (c++) ICE with typeof(X) for overloaded X 15369 * [81]9465: (preprocessor) cpp -traditional ICE on null bytes 15370 15371 C++ (compiler and library) bugs 15372 15373 * [82]47: scoping in nested classes is broken 15374 * [83]6745: problems with iostream rdbuf() member function 15375 * [84]8214: conversion from const char* const to char* sometimes 15376 accepted illegally 15377 * [85]8493: builtin strlen and overload resolution (same bug as 15378 [86]8332) 15379 * [87]8503: strange behaviour of function types 15380 * [88]8727: compiler confused by inheritance from an anonymous struct 15381 * [89]7445: poor performance of std::locale::classic() in 15382 multi-threaded applications 15383 * [90]8230: mishandling of overflow in vector<T>::resize 15384 * [91]8399: sync_with_stdio(false) breaks unformatted input 15385 * [92]8662: illegal access of private member of unnamed class is 15386 accepted 15387 * [93]8707: "make distclean" fails in libstdc++-v3 directory 15388 * [94]8708: __USE_MALLOC doesn't work 15389 * [95]8790: Use of non-thread-safe strtok in src/localename.cc 15390 * [96]8887: Bug in date formats with --enable-clocale=generic 15391 * [97]9076: Call Frame Instructions are not handled correctly during 15392 unwind operation 15393 * [98]9151: std::setprecision limited to 16 digits when outputting a 15394 double to a stream 15395 * [99]9168: codecvt<char, char, mbstate_t> overwrites output buffers 15396 * [100]9269: libstdc++ headers: explicit specialization of function 15397 must precede its first use 15398 * [101]9322: return value of basic_streambuf<>::getloc affected by 15399 locale::global 15400 * [102]9433: segfault in runtime support for dynamic_cast 15401 15402 C and optimizer bugs 15403 15404 * [103]8032: GCC incorrectly initializes static structs that have 15405 flexible arrays 15406 * [104]8639: simple arithmetic expression broken 15407 * [105]8794: optimization improperly eliminates certain expressions 15408 * [106]8832: traditional "asm volatile" code is illegally optimized 15409 * [107]8988: loop optimizer bug: with -O2, code is generated that 15410 segfaults (found on i386, bug present for all platforms) 15411 * [108]9492: structure copy clobbers subsequent stores to structure 15412 15413 Objective-C bugs 15414 15415 * [109]9267: Objective-C parser won't build with newer bison versions 15416 (e.g. 1.875) 15417 15418 Ada bugs 15419 15420 * [110]8344: Ada build problem due to conflict between gcc/final.o, 15421 gcc/ada/final.o 15422 15423 Preprocessor bugs 15424 15425 * [111]8524: _Pragma within macros is improperly expanded 15426 * [112]8880: __WCHAR_TYPE__ macro incorrectly set to "long int" with 15427 -fshort-wchar 15428 15429 ARM-specific 15430 15431 * [113]9090: arm ICE with >= -O2; regression from gcc-2.95 15432 15433 x86-specific (Intel/AMD) 15434 15435 * [114]8588: ICE in extract_insn, at recog.c:NNNN (shift instruction) 15436 * [115]8599: loop unroll bug with -march=k6-3 15437 * [116]9506: ABI breakage in structure return (affects BSD and 15438 Cygwin, but not GNU/Linux) 15439 15440 FreeBSD 5.0 specific 15441 15442 * [117]9484: GCC 3.2.1 Bootstrap failure on FreeBSD 5.0 15443 15444 RTEMS-specific 15445 15446 * [118]9292: hppa1.1-rtems configurery problems 15447 * [119]9293: [m68k-elf/rtems] config/m68k/t-crtstuff bug 15448 * [120]9295: [mips-rtems] config/mips/rtems.h init/fini issue 15449 * [121]9296: gthr-rtems regression 15450 * [122]9316: powerpc-rtems: extending multilibs 15451 15452 HP-PA specific 15453 15454 * [123]9493: ICE with -O2 when building a simple function 15455 15456 Documentation 15457 15458 * [124]7341: hyperlink to gcov in GCC documentation doesn't work 15459 * [125]8947: Please add a warning about "-malign-double" in docs 15460 * [126]7448, [127]8882: typo cleanups 15461 __________________________________________________________________ 15462 15463 GCC 3.2.1 15464 15465 3.2.1 adds a new warning, -Wabi. This option warns when GNU C++ 15466 generates code that is known not to be binary-compatible with the 15467 vendor-neutral ia32/ia64 ABI. Please consult the GCC manual, included 15468 in the distribution, for details. 15469 15470 This release also removes an old GCC extension, "naming types", and the 15471 documentation now directs users to use a different GCC extension, 15472 __typeof__, instead. The feature had evidently been broken for a while. 15473 15474 Otherwise, 3.2.1 is a bug fix release only; other than bug fixes and 15475 the new warning there are no new features that were not present in GCC 15476 3.2. 15477 15478 In addition, the previous fix for [128]PR 7445 (poor performance of 15479 std::locale::classic() in multi-threaded applications) was reverted 15480 ("unfixed"), because the "fix" was not thread-safe. 15481 15482 Bug Fixes 15483 15484 This section lists the problem reports (PRs) from GCC's bug tracking 15485 system that are known to be fixed in the 3.2.1 release. This list might 15486 not be complete (that is, it is possible that some PRs that have been 15487 fixed are not listed here). As you can see, the number of bug fixes is 15488 quite large, so it is strongly recommended that users of earlier GCC 15489 3.x releases upgrade to GCC 3.2.1. 15490 15491 Internal Compiler Errors (multi-platform) 15492 15493 * [129]2521: (c++) ICE in build_ptrmemfunc, in cp/typeck.c 15494 * [130]5661: (c++) ICE instantiating template on array of unknown 15495 size (bad code) 15496 * [131]6419: (c++) ICE in make_decl_rtl for "longest" attribute on 15497 64-bit platforms 15498 * [132]6994: (c++) ICE in find_function_data 15499 * [133]7150: preprocessor: GCC -dM -E gives an ICE 15500 * [134]7160: ICE when optimizing branches without a return value 15501 * [135]7228: (c++) ICE when using member template and template 15502 function 15503 * [136]7266: (c++) ICE with -pedantic on missing typename 15504 * [137]7353: ICE from use of "Naming Types" extension, see above 15505 * [138]7411: ICE in instantiate_virtual_regs_1, in function.c 15506 * [139]7478: (c++) ICE on static_cast inside template 15507 * [140]7526: preprocessor core dump when _Pragma implies #pragma 15508 dependency 15509 * [141]7721: (c++) ICE on simple (but incorrect) template ([142]7803 15510 is a duplicate) 15511 * [143]7754: (c++) ICE on union with template parameter 15512 * [144]7788: (c++) redeclaring a definition as an incomplete class 15513 causes ICE 15514 * [145]8031: (c++) ICE in comptypes, in cp/typeck.c 15515 * [146]8055: preprocessor dies with SIG11 when building FreeBSD 15516 kernel 15517 * [147]8067: (c++) ICE due to mishandling of __FUNCTION__ and related 15518 variables 15519 * [148]8134: (c++) ICE in force_store_init_value on legal code 15520 * [149]8149: (c++) ICE on incomplete type 15521 * [150]8160: (c++) ICE in build_modify_expr, in cp/typeck.c: array 15522 initialization 15523 15524 C++ (compiler and library) bugs 15525 15526 * [151]5607: No pointer adjustment in covariant return types 15527 * [152]6579: Infinite loop with statement expressions in member 15528 initialization 15529 * [153]6803: Default copy constructor bug in GCC 3.1 15530 * [154]7176: g++ confused by friend and static member with same name 15531 * [155]7188: Segfault with template class and recursive (incorrect) 15532 initializer list 15533 * [156]7306: Regression: GCC 3.x fails to compile code with virtual 15534 inheritance if a method has a variable number of arguments 15535 * [157]7461: ctype<char>::classic_table() returns offset array on 15536 Cygwin 15537 * [158]7524: f(const float arg[3]) fails 15538 * [159]7584: Erroneous ambiguous base error on using declaration 15539 * [160]7676: Member template overloading problem 15540 * [161]7679: infinite loop when a right parenthesis is missing 15541 * [162]7811: default locale not taken from environment 15542 * [163]7961: compare( char *) implemented incorrectly in 15543 basic_string<> 15544 * [164]8071: basic_ostream::operator<<(streambuf*) loops forever if 15545 streambuf::underflow() leaves gptr() NULL (dups: [165]8127, 15546 [166]6745) 15547 * [167]8096: deque::at() throws std::range_error instead of 15548 std::out_of_range 15549 * [168]8127: cout << cin.rdbuf() infinite loop 15550 * [169]8218: Excessively large memory consumed for classes with large 15551 array members 15552 * [170]8287: GCC 3.2: Destructor called for non-constructed local 15553 object 15554 * [171]8347: empty vector range used in string construction causes 15555 core dump 15556 * [172]8348: fail() flag is set in istringstream when eof() flag is 15557 set 15558 * [173]8391: regression: infinite loop in cp/decl2.c(finish_file) 15559 15560 C and optimizer bugs 15561 15562 * [174]6627: -fno-align-functions doesn't seem to disable function 15563 alignment 15564 * [175]6631: life_analysis misoptimizes code to initialize fields of 15565 a structure 15566 * [176]7102: unsigned char division results in floating exception 15567 * [177]7120: Run once loop should *always* be unrolled 15568 (pessimization) 15569 * [178]7209: Bug involving array referencing and ?: operator 15570 * [179]7515: invalid inlining of global function with -O3 15571 * [180]7814: incorrect scheduling for glibc-2.2.92 strcpy test 15572 * [181]8467: bug in sibling call optimization 15573 15574 Preprocessor bugs 15575 15576 * [182]4890: incorrect line markers from the traditional preprocessor 15577 * [183]7357: -M option omits system headers files (making it the same 15578 as -MM) 15579 * [184]7358: Changes to Sun's make Dependencies 15580 * [185]7602: C++ header files found in CPLUS_INCLUDE_PATH treated as 15581 C headers 15582 * [186]7862: Interrupting GCC -MD removes .d file but not .o 15583 * [187]8190: Failed compilation deletes -MD dependency file 15584 * [188]8524: _Pragma within macro is improperly expanded 15585 15586 x86 specific (Intel/AMD) 15587 15588 * [189]5351: (i686-only) function pass-by-value structure copy 15589 corrupts stack ([190]7591 is a duplicate) 15590 * [191]6845, [192]7034, [193]7124, [194]7174: ICE's with 15591 -march=pentium3/pentium2/athlon (these are all the same underlying 15592 bug, in MMX register use) 15593 * [195]7134, [196]7375, [197]7390: ICE with -march=athlon (maybe same 15594 as above?) 15595 * [198]6890: xmmintrin.h, _MM_TRANSPOSE4_PS is broken 15596 * [199]6981: wrong code in 64-bit manipulation on x86 15597 * [200]7242: GCC -mcpu=pentium[23] doesn't define __tune_pentiumpro__ 15598 macro 15599 * [201]7396: ix86: cmpgt_ss, cmpge_ss, cmpngt_ss, and cmpnge_ss SSE 15600 intrinsics are broken 15601 * [202]7630: GCC 3.2 breaks on Mozilla 1.0's JS sources with 15602 -march=pentium4 15603 * [203]7693: Typo in i386 mmintrin.h header 15604 * [204]7723: ICE - Pentium3 sse - GCC 3.2 15605 * [205]7951: ICE on -march=pentium4 -O2 -mfpmath=sse 15606 * [206]8146: (i686 only) gcc 3.2 miscompiles gcc 2.95.3 15607 15608 PowerPC specific 15609 15610 * [207]5967: GCC bug when profiling nested functions on powerpc 15611 * [208]6984: wrong code generated with -O2, -O3, -Os for do-while 15612 loop on PowerPC 15613 * [209]7114: PowerPC: ICE building strcoll.op from glibc-2.2.5 15614 * [210]7130: miscompiled code for GCC-3.1 on 15615 powerpc-unknown-linux-gnu with -funroll-all-loops 15616 * [211]7133: PowerPC ICE: unrecognizable insn 15617 * [212]7380: ICE in extract_insn, at recog.c:2148 15618 * [213]8252: ICE on Altivec code with optimization turned on 15619 * [214]8451: Altivec ICE in GCC 3.2 15620 15621 HP/PA specific 15622 15623 * [215]7250: __ashrdi3 returns wrong value on 32 bit hppa 15624 15625 SPARC specific 15626 15627 * [216]6668: when using --disable-multilib, libgcc_s.so is installed 15628 in the wrong place on sparc-solaris 15629 * [217]7151: ICE when compiling for UltraSPARC 15630 * [218]7335: SPARC: ICE in verify_wide_reg (flow.c:557) with long 15631 double and -O1 15632 * [219]7842: [REGRESSION] SPARC code gen bug 15633 15634 ARM specific 15635 15636 * [220]7856: [arm] invalid offset in constant pool reference 15637 * [221]7967: optimization produces wrong code (ARM) 15638 15639 Alpha specific 15640 15641 * [222]7374: __builtin_fabsl broken on alpha 15642 15643 IBM s390 specific 15644 15645 * [223]7370: ICE in fixup_var_refs_1 on s390x 15646 * [224]7409: loop optimization bug on s390x-linux-gnu 15647 * [225]8232: s390x: ICE when using bcmp with int length argument 15648 15649 SCO specific 15650 15651 * [226]7623: SCO OpenServer build fails with machmode.def: undefined 15652 symbol: BITS_PER_UNIT 15653 15654 m68k/Coldfire specific 15655 15656 * [227]8314: crtbegin, crtend need to be multilib'ed for this 15657 platform 15658 15659 Documentation 15660 15661 * [228]761: Document some undocumented options 15662 * [229]5610: Fix documentation about invoking SSE instructions 15663 (-mfpmath=sse) 15664 * [230]7484: List -Wmissing-declarations as C-only option 15665 * [231]7531: -mcmodel not documented for x86-64 15666 * [232]8120: Update documentation of bad use of ## 15667 __________________________________________________________________ 15668 15669 GCC 3.2 15670 15671 3.2 is a small bug fix release, but there is a change to the 15672 application binary interface (ABI), hence the change to the second part 15673 of the version number. 15674 15675 The main purpose of the 3.2 release is to correct a couple of problems 15676 in the C++ ABI, with the intention of providing a stable interface 15677 going forward. Accordingly, 3.2 is only a small change to 3.1.1. 15678 15679 Bug Fixes 15680 15681 C++ 15682 15683 * [233]7320: g++ 3.2 relocation problem 15684 * [234]7470: vtable: virtual function pointers not in declaration 15685 order 15686 15687 libstdc++ 15688 15689 * [235]6410: Trouble with non-ASCII monetary symbols and wchar_t 15690 * [236]6503, [237]6642, [238]7186: Problems with comparing or 15691 subtracting various types of const and non-const iterators 15692 * [239]7216: ambiguity with basic_iostream::traits_type 15693 * [240]7220: problem with basic_istream::ignore(0,delimiter) 15694 * [241]7222: locale::operator==() doesn't work on std::locale("") 15695 * [242]7286: placement operator delete issue 15696 * [243]7442: cxxabi.h does not match the C++ ABI 15697 * [244]7445: poor performance of std::locale::classic() in 15698 multi-threaded applications 15699 15700 x86-64 specific 15701 15702 * [245]7291: off-by-one in generated inline bzero code for x86-64 15703 15704 15705 For questions related to the use of GCC, please consult these web 15706 pages and the [246]GCC manuals. If that fails, the 15707 [247]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 15708 web pages and the development of GCC are welcome on our developer 15709 list at [248]gcc (a] gcc.gnu.org. All of [249]our lists have public 15710 archives. 15711 15712 Copyright (C) [250]Free Software Foundation, Inc. Verbatim copying and 15713 distribution of this entire article is permitted in any medium, 15714 provided this notice is preserved. 15715 15716 These pages are [251]maintained by the GCC team. Last modified 15717 2018-09-30[252]. 15718 15719 References 15720 15721 1. http://gcc.gnu.org/gcc-3.2/changes.html#3.2.3 15722 2. http://gcc.gnu.org/gcc-3.1/changes.html 15723 3. http://gcc.gnu.org/gcc-3.2/c++-abi.html 15724 4. https://gcc.gnu.org/PR3782 15725 5. https://gcc.gnu.org/PR6440 15726 6. https://gcc.gnu.org/PR7050 15727 7. https://gcc.gnu.org/PR7741 15728 8. https://gcc.gnu.org/PR7982 15729 9. https://gcc.gnu.org/PR8068 15730 10. https://gcc.gnu.org/PR8178 15731 11. https://gcc.gnu.org/PR8396 15732 12. https://gcc.gnu.org/PR8674 15733 13. https://gcc.gnu.org/PR9768 15734 14. https://gcc.gnu.org/PR9798 15735 15. https://gcc.gnu.org/PR9799 15736 16. https://gcc.gnu.org/PR9928 15737 17. https://gcc.gnu.org/PR10114 15738 18. https://gcc.gnu.org/PR10352 15739 19. https://gcc.gnu.org/PR10336 15740 20. https://gcc.gnu.org/PR8224 15741 21. https://gcc.gnu.org/PR8613 15742 22. https://gcc.gnu.org/PR8828 15743 23. https://gcc.gnu.org/PR9226 15744 24. https://gcc.gnu.org/PR9853 15745 25. https://gcc.gnu.org/PR9797 15746 26. https://gcc.gnu.org/PR9967 15747 27. https://gcc.gnu.org/PR10116 15748 28. https://gcc.gnu.org/PR10171 15749 29. https://gcc.gnu.org/PR10175 15750 30. https://gcc.gnu.org/PR8316 15751 31. https://gcc.gnu.org/PR9169 15752 32. https://gcc.gnu.org/PR9420 15753 33. https://gcc.gnu.org/PR9459 15754 34. https://gcc.gnu.org/PR9507 15755 35. https://gcc.gnu.org/PR9538 15756 36. https://gcc.gnu.org/PR9602 15757 37. https://gcc.gnu.org/PR9993 15758 38. https://gcc.gnu.org/PR10167 15759 39. https://gcc.gnu.org/PR9652 15760 40. https://gcc.gnu.org/PR10144 15761 41. https://gcc.gnu.org/PR8746 15762 42. https://gcc.gnu.org/PR9888 15763 43. https://gcc.gnu.org/PR9638 15764 44. https://gcc.gnu.org/PR9954 15765 45. https://gcc.gnu.org/PR7784 15766 46. https://gcc.gnu.org/PR7796 15767 47. https://gcc.gnu.org/PR8281 15768 48. https://gcc.gnu.org/PR8366 15769 49. https://gcc.gnu.org/PR8726 15770 50. https://gcc.gnu.org/PR9414 15771 51. https://gcc.gnu.org/PR10067 15772 52. https://gcc.gnu.org/PR7248 15773 53. https://gcc.gnu.org/PR8343 15774 54. https://gcc.gnu.org/PR9732 15775 55. https://gcc.gnu.org/PR10073 15776 56. https://gcc.gnu.org/PR7702 15777 57. https://gcc.gnu.org/PR9671 15778 58. https://gcc.gnu.org/PR8694 15779 59. https://gcc.gnu.org/PR9953 15780 60. https://gcc.gnu.org/PR10271 15781 61. https://gcc.gnu.org/PR6362 15782 62. https://gcc.gnu.org/PR10377 15783 63. https://gcc.gnu.org/PR6955 15784 64. https://gcc.gnu.org/PR5919 15785 65. https://gcc.gnu.org/PR7129 15786 66. https://gcc.gnu.org/PR7507 15787 67. https://gcc.gnu.org/PR7622 15788 68. https://gcc.gnu.org/PR7681 15789 69. https://gcc.gnu.org/PR9528 15790 70. https://gcc.gnu.org/PR8031 15791 71. https://gcc.gnu.org/PR8275 15792 72. https://gcc.gnu.org/PR8332 15793 73. https://gcc.gnu.org/PR8372 15794 74. https://gcc.gnu.org/PR8439 15795 75. https://gcc.gnu.org/PR8442 15796 76. https://gcc.gnu.org/PR8518 15797 77. https://gcc.gnu.org/PR8615 15798 78. https://gcc.gnu.org/PR8663 15799 79. https://gcc.gnu.org/PR8799 15800 80. https://gcc.gnu.org/PR9328 15801 81. https://gcc.gnu.org/PR9465 15802 82. https://gcc.gnu.org/PR47 15803 83. https://gcc.gnu.org/PR6745 15804 84. https://gcc.gnu.org/PR8214 15805 85. https://gcc.gnu.org/PR8493 15806 86. https://gcc.gnu.org/PR8332 15807 87. https://gcc.gnu.org/PR8503 15808 88. https://gcc.gnu.org/PR8727 15809 89. https://gcc.gnu.org/PR7445 15810 90. https://gcc.gnu.org/PR8230 15811 91. https://gcc.gnu.org/PR8399 15812 92. https://gcc.gnu.org/PR8662 15813 93. https://gcc.gnu.org/PR8707 15814 94. https://gcc.gnu.org/PR8708 15815 95. https://gcc.gnu.org/PR8790 15816 96. https://gcc.gnu.org/PR8887 15817 97. https://gcc.gnu.org/PR9076 15818 98. https://gcc.gnu.org/PR9151 15819 99. https://gcc.gnu.org/PR9168 15820 100. https://gcc.gnu.org/PR9269 15821 101. https://gcc.gnu.org/PR9322 15822 102. https://gcc.gnu.org/PR9433 15823 103. https://gcc.gnu.org/PR8032 15824 104. https://gcc.gnu.org/PR8639 15825 105. https://gcc.gnu.org/PR8794 15826 106. https://gcc.gnu.org/PR8832 15827 107. https://gcc.gnu.org/PR8988 15828 108. https://gcc.gnu.org/PR9492 15829 109. https://gcc.gnu.org/PR9267 15830 110. https://gcc.gnu.org/PR8344 15831 111. https://gcc.gnu.org/PR8524 15832 112. https://gcc.gnu.org/PR8880 15833 113. https://gcc.gnu.org/PR9090 15834 114. https://gcc.gnu.org/PR8588 15835 115. https://gcc.gnu.org/PR8599 15836 116. https://gcc.gnu.org/PR9506 15837 117. https://gcc.gnu.org/PR9484 15838 118. https://gcc.gnu.org/PR9292 15839 119. https://gcc.gnu.org/PR9293 15840 120. https://gcc.gnu.org/PR9295 15841 121. https://gcc.gnu.org/PR9296 15842 122. https://gcc.gnu.org/PR9316 15843 123. https://gcc.gnu.org/PR9493 15844 124. https://gcc.gnu.org/PR7341 15845 125. https://gcc.gnu.org/PR8947 15846 126. https://gcc.gnu.org/PR7448 15847 127. https://gcc.gnu.org/PR8882 15848 128. https://gcc.gnu.org/PR7445 15849 129. https://gcc.gnu.org/PR2521 15850 130. https://gcc.gnu.org/PR5661 15851 131. https://gcc.gnu.org/PR6419 15852 132. https://gcc.gnu.org/PR6994 15853 133. https://gcc.gnu.org/PR7150 15854 134. https://gcc.gnu.org/PR7160 15855 135. https://gcc.gnu.org/PR7228 15856 136. https://gcc.gnu.org/PR7266 15857 137. https://gcc.gnu.org/PR7353 15858 138. https://gcc.gnu.org/PR7411 15859 139. https://gcc.gnu.org/PR7478 15860 140. https://gcc.gnu.org/PR7526 15861 141. https://gcc.gnu.org/PR7721 15862 142. https://gcc.gnu.org/PR7803 15863 143. https://gcc.gnu.org/PR7754 15864 144. https://gcc.gnu.org/PR7788 15865 145. https://gcc.gnu.org/PR8031 15866 146. https://gcc.gnu.org/PR8055 15867 147. https://gcc.gnu.org/PR8067 15868 148. https://gcc.gnu.org/PR8134 15869 149. https://gcc.gnu.org/PR8149 15870 150. https://gcc.gnu.org/PR8160 15871 151. https://gcc.gnu.org/PR5607 15872 152. https://gcc.gnu.org/PR6579 15873 153. https://gcc.gnu.org/PR6803 15874 154. https://gcc.gnu.org/PR7176 15875 155. https://gcc.gnu.org/PR7188 15876 156. https://gcc.gnu.org/PR7306 15877 157. https://gcc.gnu.org/PR7461 15878 158. https://gcc.gnu.org/PR7524 15879 159. https://gcc.gnu.org/PR7584 15880 160. https://gcc.gnu.org/PR7676 15881 161. https://gcc.gnu.org/PR7679 15882 162. https://gcc.gnu.org/PR7811 15883 163. https://gcc.gnu.org/PR7961 15884 164. https://gcc.gnu.org/PR8071 15885 165. https://gcc.gnu.org/PR8127 15886 166. https://gcc.gnu.org/PR6745 15887 167. https://gcc.gnu.org/PR8096 15888 168. https://gcc.gnu.org/PR8127 15889 169. https://gcc.gnu.org/PR8218 15890 170. https://gcc.gnu.org/PR8287 15891 171. https://gcc.gnu.org/PR8347 15892 172. https://gcc.gnu.org/PR8348 15893 173. https://gcc.gnu.org/PR8391 15894 174. https://gcc.gnu.org/PR6627 15895 175. https://gcc.gnu.org/PR6631 15896 176. https://gcc.gnu.org/PR7102 15897 177. https://gcc.gnu.org/PR7120 15898 178. https://gcc.gnu.org/PR7209 15899 179. https://gcc.gnu.org/PR7515 15900 180. https://gcc.gnu.org/PR7814 15901 181. https://gcc.gnu.org/PR8467 15902 182. https://gcc.gnu.org/PR4890 15903 183. https://gcc.gnu.org/PR7357 15904 184. https://gcc.gnu.org/PR7358 15905 185. https://gcc.gnu.org/PR7602 15906 186. https://gcc.gnu.org/PR7862 15907 187. https://gcc.gnu.org/PR8190 15908 188. https://gcc.gnu.org/PR8524 15909 189. https://gcc.gnu.org/PR5351 15910 190. https://gcc.gnu.org/PR7591 15911 191. https://gcc.gnu.org/PR6845 15912 192. https://gcc.gnu.org/PR7034 15913 193. https://gcc.gnu.org/PR7124 15914 194. https://gcc.gnu.org/PR7174 15915 195. https://gcc.gnu.org/PR7134 15916 196. https://gcc.gnu.org/PR7375 15917 197. https://gcc.gnu.org/PR7390 15918 198. https://gcc.gnu.org/PR6890 15919 199. https://gcc.gnu.org/PR6981 15920 200. https://gcc.gnu.org/PR7242 15921 201. https://gcc.gnu.org/PR7396 15922 202. https://gcc.gnu.org/PR7630 15923 203. https://gcc.gnu.org/PR7693 15924 204. https://gcc.gnu.org/PR7723 15925 205. https://gcc.gnu.org/PR7951 15926 206. https://gcc.gnu.org/PR8146 15927 207. https://gcc.gnu.org/PR5967 15928 208. https://gcc.gnu.org/PR6984 15929 209. https://gcc.gnu.org/PR7114 15930 210. https://gcc.gnu.org/PR7130 15931 211. https://gcc.gnu.org/PR7133 15932 212. https://gcc.gnu.org/PR7380 15933 213. https://gcc.gnu.org/PR8252 15934 214. https://gcc.gnu.org/PR8451 15935 215. https://gcc.gnu.org/PR7250 15936 216. https://gcc.gnu.org/PR6668 15937 217. https://gcc.gnu.org/PR7151 15938 218. https://gcc.gnu.org/PR7335 15939 219. https://gcc.gnu.org/PR7842 15940 220. https://gcc.gnu.org/PR7856 15941 221. https://gcc.gnu.org/PR7967 15942 222. https://gcc.gnu.org/PR7374 15943 223. https://gcc.gnu.org/PR7370 15944 224. https://gcc.gnu.org/PR7409 15945 225. https://gcc.gnu.org/PR8232 15946 226. https://gcc.gnu.org/PR7623 15947 227. https://gcc.gnu.org/PR8314 15948 228. https://gcc.gnu.org/PR761 15949 229. https://gcc.gnu.org/PR5610 15950 230. https://gcc.gnu.org/PR7484 15951 231. https://gcc.gnu.org/PR7531 15952 232. https://gcc.gnu.org/PR8120 15953 233. https://gcc.gnu.org/PR7320 15954 234. https://gcc.gnu.org/PR7470 15955 235. https://gcc.gnu.org/PR6410 15956 236. https://gcc.gnu.org/PR6503 15957 237. https://gcc.gnu.org/PR6642 15958 238. https://gcc.gnu.org/PR7186 15959 239. https://gcc.gnu.org/PR7216 15960 240. https://gcc.gnu.org/PR7220 15961 241. https://gcc.gnu.org/PR7222 15962 242. https://gcc.gnu.org/PR7286 15963 243. https://gcc.gnu.org/PR7442 15964 244. https://gcc.gnu.org/PR7445 15965 245. https://gcc.gnu.org/PR7291 15966 246. https://gcc.gnu.org/onlinedocs/ 15967 247. mailto:gcc-help (a] gcc.gnu.org 15968 248. mailto:gcc (a] gcc.gnu.org 15969 249. https://gcc.gnu.org/lists.html 15970 250. https://www.fsf.org/ 15971 251. https://gcc.gnu.org/about.html 15972 252. http://validator.w3.org/check/referer 15973 ====================================================================== 15974 http://gcc.gnu.org/gcc-3.1/index.html 15975 15976 GCC 3.1 15977 15978 July 27, 2002 15979 15980 The [1]GNU project and the GCC developers are pleased to announce the 15981 release of GCC 3.1.1. 15982 15983 The links below still apply to GCC 3.1.1. 15984 15985 This release series is no longer maintained. 15986 15987 May 15, 2002 15988 15989 The [2]GNU project and the GCC developers are pleased to announce the 15990 release of GCC 3.1. 15991 15992 GCC used to stand for the GNU C Compiler, but since the compiler 15993 supports several other languages aside from C, it now stands for the 15994 GNU Compiler Collection. 15995 15996 A list of [3]successful builds is updated as new information becomes 15997 available. 15998 15999 The GCC developers would like to thank the numerous people that have 16000 contributed [4]new features, improvements, bug fixes, and other changes 16001 as well as test results to GCC. This [5]amazing group of volunteers is 16002 what makes GCC successful. 16003 16004 For additional information about GCC please refer to the [6]GCC project 16005 web site or contact the [7]GCC development mailing list. 16006 16007 To obtain GCC please use [8]our mirror sites, or our CVS server. 16008 __________________________________________________________________ 16009 16010 16011 For questions related to the use of GCC, please consult these web 16012 pages and the [9]GCC manuals. If that fails, the 16013 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16014 web pages and the development of GCC are welcome on our developer 16015 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 16016 archives. 16017 16018 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 16019 distribution of this entire article is permitted in any medium, 16020 provided this notice is preserved. 16021 16022 These pages are [14]maintained by the GCC team. Last modified 16023 2018-09-30[15]. 16024 16025 References 16026 16027 1. http://www.gnu.org/ 16028 2. http://www.gnu.org/ 16029 3. http://gcc.gnu.org/gcc-3.1/buildstat.html 16030 4. http://gcc.gnu.org/gcc-3.1/changes.html 16031 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 16032 6. http://gcc.gnu.org/index.html 16033 7. mailto:gcc (a] gcc.gnu.org 16034 8. http://gcc.gnu.org/mirrors.html 16035 9. https://gcc.gnu.org/onlinedocs/ 16036 10. mailto:gcc-help (a] gcc.gnu.org 16037 11. mailto:gcc (a] gcc.gnu.org 16038 12. https://gcc.gnu.org/lists.html 16039 13. https://www.fsf.org/ 16040 14. https://gcc.gnu.org/about.html 16041 15. http://validator.w3.org/check/referer 16042 ====================================================================== 16043 http://gcc.gnu.org/gcc-3.1/changes.html 16044 16045 GCC 3.1 Release Series 16046 Changes, New Features, and Fixes 16047 16048 Additional changes in GCC 3.1.1 16049 16050 * A bug related to how structures and unions are returned has been 16051 fixed for powerpc-*-netbsd*. 16052 * An important bug in the implementation of -fprefetch-loop-arrays 16053 has been fixed. Previously the optimization prefetched random 16054 blocks of memory for most targets except for i386. 16055 * The Java compiler now compiles Java programs much faster and also 16056 works with parallel make. 16057 * Nested functions have been fixed for mips*-*-netbsd*. 16058 * Some missing floating point support routines have beed added for 16059 mips*-*-netbsd*. 16060 * This [1]message gives additional information about the bugs fixed 16061 in this release. 16062 16063 Caveats 16064 16065 * The -traditional C compiler option has been deprecated and will be 16066 removed in GCC 3.3. (It remains possible to preprocess non-C code 16067 with the traditional preprocessor.) 16068 * The default debugging format for most ELF platforms (including 16069 GNU/Linux and FreeBSD; notable exception is Solaris) has changed 16070 from stabs to DWARF2. This requires GDB 5.1.1 or later. 16071 16072 General Optimizer Improvements 16073 16074 * Jan Hubicka, SuSE Labs, together with Richard Henderson, Red Hat, 16075 and Andreas Jaeger, SuSE Labs, has contributed [2]infrastructure 16076 for profile driven optimizations. 16077 Options -fprofile-arcs and -fbranch-probabilities can now be used 16078 to improve speed of the generated code by profiling the actual 16079 program behaviour on typical runs. In the absence of profile info 16080 the compiler attempts to guess the profile statically. 16081 * [3]SPEC2000 and SPEC95 benchmark suites are now used daily to 16082 monitor performance of the generated code. 16083 According to the SPECInt2000 results on an AMD Athlon CPU, the code 16084 generated by GCC 3.1 is 6% faster on the average (8.2% faster with 16085 profile feedback) compared to GCC 3.0. The code produced by GCC 3.0 16086 is about 2.1% faster compared to 2.95.3. Tests were done using the 16087 -O2 -march=athlon command-line options. 16088 * Alexandre Oliva, of Red Hat, has generalized the tree inlining 16089 infrastructure developed by CodeSourcery, LLC for the C++ front 16090 end, so that it is now used in the C front end too. Inlining 16091 functions as trees exposes them earlier to the compiler, giving it 16092 more opportunities for optimization. 16093 * Support for data prefetching instructions has been added to the GCC 16094 back end and several targets. A new __builtin_prefetch intrinsic is 16095 available to explicitly insert prefetch instructions and 16096 experimental support for loop array prefetching has been added (see 16097 -fprefetch-loop-array documentation). 16098 * Support for emitting debugging information for macros has been 16099 added for DWARF2. It is activated using -g3. 16100 16101 New Languages and Language specific improvements 16102 16103 C/C++ 16104 16105 * A few more [4]ISO C99 features. 16106 * The preprocessor is 10-50% faster than the preprocessor in GCC 3.0. 16107 * The preprocessor's symbol table has been merged with the symbol 16108 table of the C, C++ and Objective-C front ends. 16109 * The preprocessor consumes less memory than the preprocessor in GCC 16110 3.0, often significantly so. On normal input files, it typically 16111 consumes less memory than pre-3.0 cccp-based GCC, too. 16112 16113 C++ 16114 16115 * -fhonor-std and -fno-honor-std have been removed. -fno-honor-std 16116 was a workaround to allow std compliant code to work with the 16117 non-std compliant libstdc++-v2. libstdc++-v3 is std compliant. 16118 * The C++ ABI has been fixed so that void (A::*)() const is mangled 16119 as "M1AKFvvE", rather than "MK1AFvvE" as before. This change only 16120 affects pointer to cv-qualified member function types. 16121 * The C++ ABI has been changed to correctly handle this code: 16122 struct A { 16123 void operator delete[] (void *, size_t); 16124 }; 16125 16126 struct B : public A { 16127 }; 16128 16129 new B[10]; 16130 16131 The amount of storage allocated for the array will be greater than 16132 it was in 3.0, in order to store the number of elements in the 16133 array, so that the correct size can be passed to operator delete[] 16134 when the array is deleted. Previously, the value passed to operator 16135 delete[] was unpredictable. 16136 This change will only affect code that declares a two-argument 16137 operator delete[] with a second parameter of type size_t in a base 16138 class, and does not override that definition in a derived class. 16139 * The C++ ABI has been changed so that: 16140 struct A { 16141 void operator delete[] (void *, size_t); 16142 void operator delete[] (void *); 16143 }; 16144 16145 does not cause unnecessary storage to be allocated when an array of 16146 A objects is allocated. 16147 This change will only affect code that declares both of these forms 16148 of operator delete[], and declared the two-argument form before the 16149 one-argument form. 16150 * The C++ ABI has been changed so that when a parameter is passed by 16151 value, any cleanup for that parameter is performed in the caller, 16152 as specified by the ia64 C++ ABI, rather than the called function 16153 as before. As a result, classes with a non-trivial destructor but a 16154 trivial copy constructor will be passed and returned by invisible 16155 reference, rather than by bitwise copy as before. 16156 * G++ now supports the "named return value optimization": for code 16157 like 16158 A f () { 16159 A a; 16160 ... 16161 return a; 16162 } 16163 16164 G++ will allocate a in the return value slot, so that the return 16165 becomes a no-op. For this to work, all return statements in the 16166 function must return the same variable. 16167 * Improvements to the C++ library are listed in [5]the libstdc++-v3 16168 FAQ. 16169 16170 Objective-C 16171 16172 * Annoying linker warnings (due to incorrect code being generated) 16173 have been fixed. 16174 * If a class method cannot be found, the compiler no longer issues a 16175 warning if a corresponding instance method exists in the root 16176 class. 16177 * Forward @protocol declarations have been fixed. 16178 * Loading of categories has been fixed in certain situations (GNU run 16179 time only). 16180 * The class lookup in the run-time library has been rewritten so that 16181 class method dispatch is more than twice as fast as it used to be 16182 (GNU run time only). 16183 16184 Java 16185 16186 * libgcj now includes RMI, java.lang.ref.*, javax.naming, and 16187 javax.transaction. 16188 * Property files and other system resources can be compiled into 16189 executables which use libgcj using the new gcj --resource feature. 16190 * libgcj has been ported to more platforms. In particular there is 16191 now a mostly-functional mingw32 (Windows) target port. 16192 * JNI and CNI invocation interfaces were implemented, so gcj-compiled 16193 Java code can now be called from a C/C++ application. 16194 * gcj can now use builtin functions for certain known methods, for 16195 instance Math.cos. 16196 * gcj can now automatically remove redundant array-store checks in 16197 some common cases. 16198 * The --no-store-checks optimization option was added. This can be 16199 used to omit runtime store checks for code which is known not to 16200 throw ArrayStoreException 16201 * The following third party interface standards were added to libgcj: 16202 org.w3c.dom and org.xml.sax. 16203 * java.security has been merged with GNU Classpath. The new package 16204 is now JDK 1.2 compliant, and much more complete. 16205 * A bytecode verifier was added to the libgcj interpreter. 16206 * java.lang.Character was rewritten to comply with the Unicode 3.0 16207 standard, and improve performance. 16208 * Partial support for many more locales was added to libgcj. 16209 * Socket timeouts have been implemented. 16210 * libgcj has been merged into a single shared library. There are no 16211 longer separate shared libraries for the garbage collector and 16212 zlib. 16213 * Several performance improvements were made to gcj and libgcj: 16214 + Hash synchronization (thin locks) 16215 + A special allocation path for finalizer-free objects 16216 + Thread-local allocation 16217 + Parallel GC, and other GC tweaks 16218 16219 Fortran 16220 16221 Fortran improvements are listed in [6]the Fortran documentation. 16222 16223 Ada 16224 16225 [7]AdaCore, has contributed its GNAT Ada 95 front end and associated 16226 tools. The GNAT compiler fully implements the Ada language as defined 16227 by the ISO/IEC 8652 standard. 16228 16229 Please note that the integration of the Ada front end is still work in 16230 progress. 16231 16232 New Targets and Target Specific Improvements 16233 16234 * Hans-Peter Nilsson has contributed a port to MMIX, the CPU 16235 architecture used in new editions of Donald E. Knuth's The Art of 16236 Computer Programming. 16237 * Axis Communications has contributed its port to the CRIS CPU 16238 architecture, used in the ETRAX system-on-a-chip series. 16239 * Alexandre Oliva, of Red Hat, has contributed a port to the SuperH 16240 SH5 64-bit RISC microprocessor architecture, extending the existing 16241 SH port. 16242 * UltraSPARC is fully supported in 64-bit mode. The option -m64 16243 enables it. 16244 * For compatibility with the Sun compiler #pragma redefine_extname 16245 has been implemented on Solaris. 16246 * The x86 back end has had some noticeable work done to it. 16247 + SuSE Labs developers Jan Hubicka, Bo Thorsen and Andreas 16248 Jaeger have contributed a port to the AMD x86-64 architecture. 16249 For more information on x86-64 see http://www.x86-64.org. 16250 + The compiler now supports MMX, 3DNow!, SSE, and SSE2 16251 instructions. Options -mmmx, -m3dnow, -msse, and -msse2 will 16252 enable the respective instruction sets. Intel C++ compatible 16253 MMX/3DNow!/SSE intrinsics are implemented. SSE2 intrinsics 16254 will be added in next major release. 16255 + Following those improvements, targets for Pentium MMX, K6-2, 16256 K6-3, Pentium III, Pentium 4, and Athlon 4 Mobile/XP/MP were 16257 added. Refer to the documentation on -march= and -mcpu= 16258 options for details. 16259 + For those targets that support it, -mfpmath=sse will cause the 16260 compiler to generate SSE/SSE2 instructions for floating point 16261 math instead of x87 instructions. Usually, this will lead to 16262 quicker code -- especially on the Pentium 4. Note that only 16263 scalar floating point instructions are used and GCC does not 16264 exploit SIMD features yet. 16265 + Prefetch support has been added to the Pentium III, Pentium 4, 16266 K6-2, K6-3, and Athlon series. 16267 + Code generated for floating point to integer conversions has 16268 been improved leading to better performance of many 3D 16269 applications. 16270 * The PowerPC back end has added 64-bit PowerPC GNU/Linux support. 16271 * C++ support for AIX has been improved. 16272 * Aldy Hernandez, of Red Hat, Inc, has contributed extensions to the 16273 PowerPC port supporting the AltiVec programming model (SIMD). The 16274 support, though presently useful, is experimental and is expected 16275 to stabilize for 3.2. The support is written to conform to 16276 Motorola's AltiVec specs. See -maltivec. 16277 16278 Obsolete Systems 16279 16280 Support for a number of older systems has been declared obsolete in GCC 16281 3.1. Unless there is activity to revive them, the next release of GCC 16282 will have their sources permanently removed. 16283 16284 All configurations of the following processor architectures have been 16285 declared obsolete: 16286 * MIL-STD-1750A, 1750a-*-* 16287 * AMD A29k, a29k-*-* 16288 * Convex, c*-convex-* 16289 * Clipper, clipper-*-* 16290 * Elxsi, elxsi-*-* 16291 * Intel i860, i860-*-* 16292 * Sun picoJava, pj-*-* and pjl-*-* 16293 * Western Electric 32000, we32k-*-* 16294 16295 Most configurations of the following processor architectures have been 16296 declared obsolete, but we are preserving a few systems which may have 16297 active developers. It is unlikely that the remaining systems will 16298 survive much longer unless we see definite signs of port activity. 16299 * Motorola 88000 except 16300 + Generic a.out, m88k-*-aout* 16301 + Generic SVR4, m88k-*-sysv4 16302 + OpenBSD, m88k-*-openbsd* 16303 * NS32k except 16304 + NetBSD, ns32k-*-netbsd* 16305 + OpenBSD, ns32k-*-openbsd*. 16306 * ROMP except 16307 + OpenBSD, romp-*-openbsd*. 16308 16309 Finally, only some configurations of these processor architectures are 16310 being obsoleted. 16311 * Alpha: 16312 + OSF/1, alpha*-*-osf[123]*. (Digital Unix and Tru64 Unix, aka 16313 alpha*-*-osf[45], are still supported.) 16314 * ARM: 16315 + RISCiX, arm-*-riscix*. 16316 * i386: 16317 + 386BSD, i?86-*-bsd* 16318 + Chorus, i?86-*-chorusos* 16319 + DG/UX, i?86-*-dgux* 16320 + FreeBSD 1.x, i?86-*-freebsd1.* 16321 + IBM AIX, i?86-*-aix* 16322 + ISC UNIX, i?86-*-isc* 16323 + GNU/Linux with pre-BFD linker, i?86-*-linux*oldld* 16324 + NEXTstep, i?86-next-* 16325 + OSF UNIX, i?86-*-osf1* and i?86-*-osfrose* 16326 + RTEMS/coff, i?86-*-rtemscoff* 16327 + RTEMS/go32, i?86-go32-rtems* 16328 + Sequent/BSD, i?86-sequent-bsd* 16329 + Sequent/ptx before version 3, i?86-sequent-ptx[12]* and 16330 i?86-sequent-sysv3* 16331 + SunOS, i?86-*-sunos* 16332 * Motorola 68000: 16333 + Altos, m68[k0]*-altos-* 16334 + Apollo, m68[k0]*-apollo-* 16335 + Apple A/UX, m68[k0]*-apple-* 16336 + Bull, m68[k0]*-bull-* 16337 + Convergent, m68[k0]*-convergent-* 16338 + Generic SVR3, m68[k0]*-*-sysv3* 16339 + ISI, m68[k0]*-isi-* 16340 + LynxOS, m68[k0]*-*-lynxos* 16341 + NEXT, m68[k0]*-next-* 16342 + RTEMS/coff, m68[k0]*-*-rtemscoff* 16343 + Sony, m68[k0]*-sony-* 16344 * MIPS: 16345 + DEC Ultrix, mips-*-ultrix* and mips-dec-* 16346 + Generic BSD, mips-*-bsd* 16347 + Generic System V, mips-*-sysv* 16348 + IRIX before version 5, mips-sgi-irix[1234]* 16349 + RiscOS, mips-*-riscos* 16350 + Sony, mips-sony-* 16351 + Tandem, mips-tandem-* 16352 * SPARC: 16353 + RTEMS/a.out, sparc-*-rtemsaout*. 16354 16355 Documentation improvements 16356 16357 * The old manual ("Using and Porting the GNU Compiler Collection") 16358 has been replaced by a users manual ("Using the GNU Compiler 16359 Collection") and a separate internals reference manual ("GNU 16360 Compiler Collection Internals"). 16361 * More complete and much improved documentation about GCC's internal 16362 representation used by the C and C++ front ends. 16363 * Many cleanups and improvements in general. 16364 16365 16366 For questions related to the use of GCC, please consult these web 16367 pages and the [8]GCC manuals. If that fails, the 16368 [9]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16369 web pages and the development of GCC are welcome on our developer 16370 list at [10]gcc (a] gcc.gnu.org. All of [11]our lists have public 16371 archives. 16372 16373 Copyright (C) [12]Free Software Foundation, Inc. Verbatim copying and 16374 distribution of this entire article is permitted in any medium, 16375 provided this notice is preserved. 16376 16377 These pages are [13]maintained by the GCC team. Last modified 16378 2018-09-30[14]. 16379 16380 References 16381 16382 1. https://gcc.gnu.org/ml/gcc/2002-07/msg01208.html 16383 2. http://gcc.gnu.org/news/profiledriven.html 16384 3. http://gcc.gnu.org/benchmarks/ 16385 4. http://gcc.gnu.org/c99status.html 16386 5. https://gcc.gnu.org/onlinedocs/libstdc++/faq.html 16387 6. https://gcc.gnu.org/onlinedocs/gcc-3.1.1/g77/News.html 16388 7. https://www.adacore.com/ 16389 8. https://gcc.gnu.org/onlinedocs/ 16390 9. mailto:gcc-help (a] gcc.gnu.org 16391 10. mailto:gcc (a] gcc.gnu.org 16392 11. https://gcc.gnu.org/lists.html 16393 12. https://www.fsf.org/ 16394 13. https://gcc.gnu.org/about.html 16395 14. http://validator.w3.org/check/referer 16396 ====================================================================== 16397 http://gcc.gnu.org/gcc-3.0/index.html 16398 16399 GCC 3.0.4 16400 16401 February 20, 2002 16402 16403 The [1]GNU project and the GCC developers are pleased to announce the 16404 release of GCC 3.0.4, which is a bug-fix release for the GCC 3.0 16405 series. 16406 16407 This release series is no longer maintained. 16408 16409 GCC used to stand for the GNU C Compiler, but since the compiler 16410 supports several other languages aside from C, it now stands for the 16411 GNU Compiler Collection. 16412 16413 GCC 3.0.x has several new optimizations, new targets, new languages and 16414 many other new features, relative to GCC 2.95.x. See the [2]new 16415 features page for a more complete list. 16416 16417 A list of [3]successful builds is updated as new information becomes 16418 available. 16419 16420 The GCC developers would like to thank the numerous people that have 16421 contributed new features, test results, bug fixes, etc to GCC. This 16422 [4]amazing group of volunteers is what makes GCC successful. 16423 16424 And finally, we can't in good conscience fail to mention some 16425 [5]caveats to using GCC 3.0.x. 16426 16427 For additional information about GCC please refer to the [6]GCC project 16428 web site or contact the [7]GCC development mailing list. 16429 16430 To obtain GCC please use [8]our mirror sites, or our CVS server. 16431 __________________________________________________________________ 16432 16433 Previous 3.0.x Releases 16434 16435 December 20, 2001: GCC 3.0.3 has been released. 16436 October 25, 2001: GCC 3.0.2 has been released. 16437 August 20, 2001: GCC 3.0.1 has been released. 16438 June 18, 2001: GCC 3.0 has been released. 16439 16440 16441 For questions related to the use of GCC, please consult these web 16442 pages and the [9]GCC manuals. If that fails, the 16443 [10]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16444 web pages and the development of GCC are welcome on our developer 16445 list at [11]gcc (a] gcc.gnu.org. All of [12]our lists have public 16446 archives. 16447 16448 Copyright (C) [13]Free Software Foundation, Inc. Verbatim copying and 16449 distribution of this entire article is permitted in any medium, 16450 provided this notice is preserved. 16451 16452 These pages are [14]maintained by the GCC team. Last modified 16453 2018-09-30[15]. 16454 16455 References 16456 16457 1. http://www.gnu.org/ 16458 2. http://gcc.gnu.org/gcc-3.0/features.html 16459 3. http://gcc.gnu.org/gcc-3.0/buildstat.html 16460 4. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 16461 5. http://gcc.gnu.org/gcc-3.0/caveats.html 16462 6. http://gcc.gnu.org/index.html 16463 7. mailto:gcc (a] gcc.gnu.org 16464 8. http://gcc.gnu.org/mirrors.html 16465 9. https://gcc.gnu.org/onlinedocs/ 16466 10. mailto:gcc-help (a] gcc.gnu.org 16467 11. mailto:gcc (a] gcc.gnu.org 16468 12. https://gcc.gnu.org/lists.html 16469 13. https://www.fsf.org/ 16470 14. https://gcc.gnu.org/about.html 16471 15. http://validator.w3.org/check/referer 16472 ====================================================================== 16473 http://gcc.gnu.org/gcc-3.0/features.html 16474 16475 GCC 3.0 New Features 16476 16477 Additional changes in GCC 3.0.4 16478 16479 * GCC 3.0 now supports newer versions of the [1]NetBSD operating 16480 system, which use the ELF object file format, on x86 processors. 16481 * Correct debugging information is generated from functions that have 16482 lines from multiple files (e.g. yacc output). 16483 * A fix for whitespace handling in the -traditional preprocessor, 16484 which can affect Fortran. 16485 * Fixes to the exception handling runtime. 16486 * More fixes for bad code generation in C++. 16487 * A fix for shared library generation under AIX 4.3. 16488 * Documentation updates. 16489 * Port of GCC to Tensilica's Xtensa processor contributed. 16490 * A fix for compiling the PPC Linux kernel (FAT fs wouldn't link). 16491 16492 Additional changes in GCC 3.0.3 16493 16494 * A fix to correct an accidental change to the PowerPC ABI. 16495 * Fixes for bad code generation on a variety of architectures. 16496 * Improvements to the debugging information generated for C++ 16497 classes. 16498 * Fixes for bad code generation in C++. 16499 * A fix to avoid crashes in the C++ demangler. 16500 * A fix to the C++ standard library to avoid buffer overflows. 16501 * Miscellaneous improvements for a variety of architectures. 16502 16503 Additional changes in GCC 3.0.2 16504 16505 * Fixes for bad code generation during loop unrolling. 16506 * Fixes for bad code generation by the sibling call optimization. 16507 * Minor improvements to x86 code generation. 16508 * Implementation of function descriptors in C++ vtables for IA64. 16509 * Numerous minor bug-fixes. 16510 16511 Additional changes in GCC 3.0.1 16512 16513 * C++ fixes for incorrect code-generation. 16514 * Improved cross-compiling support for the C++ standard library. 16515 * Fixes for some embedded targets that worked in GCC 2.95.3, but not 16516 in GCC 3.0. 16517 * Fixes for various exception-handling bugs. 16518 * A port to the S/390 architecture. 16519 16520 General Optimizer Improvements 16521 16522 * [2]Basic block reordering pass. 16523 * New if-conversion pass with support for conditional (predicated) 16524 execution. 16525 * New tail call and sibling call elimination optimizations. 16526 * New register renaming pass. 16527 * New (experimental) [3]static single assignment (SSA) representation 16528 support. 16529 * New dead-code elimination pass implemented using the SSA 16530 representation. 16531 * [4]Global null pointer test elimination. 16532 * [5]Global code hoisting/unification. 16533 * More builtins and optimizations for stdio.h, string.h and old BSD 16534 functions, as well as for ISO C99 functions. 16535 * New builtin __builtin_expect for giving hints to the branch 16536 predictor. 16537 16538 New Languages and Language specific improvements 16539 16540 * The GNU Compiler for the Java(TM) language (GCJ) is now integrated 16541 and supported, including the run-time library containing most 16542 common non-GUI Java classes, a bytecode interpreter, and the Boehm 16543 conservative garbage collector. Many bugs have been fixed. GCJ can 16544 compile Java source or Java bytecodes to either native code or Java 16545 class files, and supports native methods written in either the 16546 standard JNI or the more efficient and convenient CNI. 16547 * Here is a [6]partial list of C++ improvements, both new features 16548 and those no longer supported. 16549 * New C++ ABI. On the IA-64 platform GCC is capable of 16550 inter-operating with other IA-64 compilers. 16551 * The new ABI also significantly reduces the size of symbol and debug 16552 information. 16553 * New C++ support library and many C++ bug fixes, vastly improving 16554 our conformance to the ISO C++ standard. 16555 * New [7]inliner for C++. 16556 * Rewritten C preprocessor, integrated into the C, C++ and Objective 16557 C compilers, with very many improvements including ISO C99 support 16558 and [8]improvements to dependency generation. 16559 * Support for more [9]ISO C99 features. 16560 * Many improvements to support for checking calls to format functions 16561 such as printf and scanf, including support for ISO C99 format 16562 features, extensions from the Single Unix Specification and GNU 16563 libc 2.2, checking of strfmon formats and features to assist in 16564 auditing for format string security bugs. 16565 * New warnings for C code that may have undefined semantics because 16566 of violations of sequence point rules in the C standard (such as a 16567 = a++;, a[n] = b[n++]; and a[i++] = i;), included in -Wall. 16568 * Additional warning option -Wfloat-equal. 16569 * Improvements to -Wtraditional. 16570 * Fortran improvements are listed in [10]the Fortran documentation. 16571 16572 New Targets and Target Specific Improvements 16573 16574 * New x86 back end, generating much improved code. 16575 * Support for a generic i386-elf target contributed. 16576 * New option to emit x86 assembly code using Intel style syntax 16577 (-mintel-syntax). 16578 * HPUX 11 support contributed. 16579 * Improved PowerPC code generation, including scheduled prologue and 16580 epilogue. 16581 * Port of GCC to Intel's IA-64 processor contributed. 16582 * Port of GCC to Motorola's MCore 210 and 340 contributed. 16583 * New unified back-end for Arm, Thumb and StrongArm contributed. 16584 * Port of GCC to Intel's XScale processor contributed. 16585 * Port of GCC to Atmel's AVR microcontrollers contributed. 16586 * Port of GCC to Mitsubishi's D30V processor contributed. 16587 * Port of GCC to Matsushita's AM33 processor (a member of the MN10300 16588 processor family) contributed. 16589 * Port of GCC to Fujitsu's FR30 processor contributed. 16590 * Port of GCC to Motorola's 68HC11 and 68HC12 processors contributed. 16591 * Port of GCC to Sun's picoJava processor core contributed. 16592 16593 Documentation improvements 16594 16595 * Substantially rewritten and improved C preprocessor manual. 16596 * Many improvements to other documentation. 16597 * Manpages for gcc, cpp and gcov are now generated automatically from 16598 the master Texinfo manual, eliminating the problem of manpages 16599 being out of date. (The generated manpages are only extracts from 16600 the full manual, which is provided in Texinfo form, from which 16601 info, HTML, other formats and a printed manual can be generated.) 16602 * Generated info files are included in the release tarballs alongside 16603 their Texinfo sources, avoiding problems on some platforms with 16604 building makeinfo as part of the GCC distribution. 16605 16606 Other significant improvements 16607 16608 * Garbage collection used internally by the compiler for most memory 16609 allocation instead of obstacks. 16610 * Lengauer and Tarjan algorithm used for computing dominators in the 16611 CFG. This algorithm can be significantly faster and more space 16612 efficient than our older algorithm. 16613 * gccbug script provided to assist in submitting bug reports to our 16614 bug tracking system. (Bug reports previously submitted directly to 16615 our mailing lists, for which you received no bug tracking number, 16616 should be submitted again using gccbug if you can reproduce the 16617 problem with GCC 3.0.) 16618 * The internal libgcc library is [11]built as a shared library on 16619 systems that support it. 16620 * Extensive testsuite included with GCC, with many new tests. In 16621 addition to tests for GCC bugs that have been fixed, many tests 16622 have been added for language features, compiler warnings and 16623 builtin functions. 16624 * Additional language-independent warning options -Wpacked, -Wpadded, 16625 -Wunreachable-code and -Wdisabled-optimization. 16626 * Target-independent options -falign-functions, -falign-loops and 16627 -falign-jumps. 16628 16629 Plus a great many bug fixes and almost all the [12]features found in 16630 GCC 2.95. 16631 16632 16633 For questions related to the use of GCC, please consult these web 16634 pages and the [13]GCC manuals. If that fails, the 16635 [14]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16636 web pages and the development of GCC are welcome on our developer 16637 list at [15]gcc (a] gcc.gnu.org. All of [16]our lists have public 16638 archives. 16639 16640 Copyright (C) [17]Free Software Foundation, Inc. Verbatim copying and 16641 distribution of this entire article is permitted in any medium, 16642 provided this notice is preserved. 16643 16644 These pages are [18]maintained by the GCC team. Last modified 16645 2018-09-30[19]. 16646 16647 References 16648 16649 1. http://www.netbsd.org/ 16650 2. http://gcc.gnu.org/news/reorder.html 16651 3. http://gcc.gnu.org/news/ssa.html 16652 4. http://gcc.gnu.org/news/null.html 16653 5. http://gcc.gnu.org/news/unify.html 16654 6. http://gcc.gnu.org/gcc-3.0/c++features.html 16655 7. http://gcc.gnu.org/news/inlining.html 16656 8. http://gcc.gnu.org/news/dependencies.html 16657 9. http://gcc.gnu.org/c99status.html 16658 10. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 16659 11. http://gcc.gnu.org/gcc-3.0/libgcc.html 16660 12. http://gcc.gnu.org/gcc-2.95/features.html 16661 13. https://gcc.gnu.org/onlinedocs/ 16662 14. mailto:gcc-help (a] gcc.gnu.org 16663 15. mailto:gcc (a] gcc.gnu.org 16664 16. https://gcc.gnu.org/lists.html 16665 17. https://www.fsf.org/ 16666 18. https://gcc.gnu.org/about.html 16667 19. http://validator.w3.org/check/referer 16668 ====================================================================== 16669 http://gcc.gnu.org/gcc-3.0/caveats.html 16670 16671 GCC 3.0 Caveats 16672 16673 * -fstrict-aliasing is now part of -O2 and higher optimization 16674 levels. This allows the compiler to assume the strictest aliasing 16675 rules applicable to the language being compiled. For C and C++, 16676 this activates optimizations based on the type of expressions. This 16677 optimization may thus break old, non-compliant code. 16678 * Enumerations are now properly promoted to int in function 16679 parameters and function returns. Normally this change is not 16680 visible, but when using -fshort-enums this is an ABI change. 16681 * The undocumented extension that allowed C programs to have a label 16682 at the end of a compound statement has been deprecated and may be 16683 removed in a future version. Programs that now generate a warning 16684 about this may be fixed by adding a null statement (a single 16685 semicolon) after the label. 16686 * The poorly documented extension that allowed string constants in C, 16687 C++ and Objective C to contain unescaped newlines has been 16688 deprecated and may be removed in a future version. Programs using 16689 this extension may be fixed in several ways: the bare newline may 16690 be replaced by \n, or preceded by \n\, or string concatenation may 16691 be used with the bare newline preceded by \n" and " placed at the 16692 start of the next line. 16693 * The Chill compiler is not included in GCC 3.0, because of the lack 16694 of a volunteer to convert it to use garbage collection. 16695 * Certain non-standard iostream methods from earlier versions of 16696 libstdc++ are not included in libstdc++ v3, i.e. filebuf::attach, 16697 ostream::form, and istream::gets. 16698 * The new C++ ABI is not yet fully supported by current (as of 16699 2001-07-01) releases and development versions of GDB, or any 16700 earlier versions. There is a problem setting breakpoints by line 16701 number, and other related issues that have been fixed in GCC 3.0 16702 but not yet handled in GDB: 16703 [1]https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 16704 16705 16706 For questions related to the use of GCC, please consult these web 16707 pages and the [2]GCC manuals. If that fails, the 16708 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16709 web pages and the development of GCC are welcome on our developer 16710 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 16711 16712 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 16713 distribution of this entire article is permitted in any medium, 16714 provided this notice is preserved. 16715 16716 These pages are [7]maintained by the GCC team. Last modified 16717 2018-09-30[8]. 16718 16719 References 16720 16721 1. https://gcc.gnu.org/ml/gcc-bugs/2001-06/msg00421.html 16722 2. https://gcc.gnu.org/onlinedocs/ 16723 3. mailto:gcc-help (a] gcc.gnu.org 16724 4. mailto:gcc (a] gcc.gnu.org 16725 5. https://gcc.gnu.org/lists.html 16726 6. https://www.fsf.org/ 16727 7. https://gcc.gnu.org/about.html 16728 8. http://validator.w3.org/check/referer 16729 ====================================================================== 16730 http://gcc.gnu.org/gcc-2.95/index.html 16731 16732 GCC 2.95 16733 16734 March 16, 2001: The GNU project and the GCC developers are pleased to 16735 announce the release of GCC version 2.95.3. 16736 16737 This release is no longer maintained. 16738 16739 Release History 16740 16741 GCC 2.95.3 16742 March 16, 2001 16743 16744 GCC 2.95.2 16745 October 27, 1999 16746 16747 GCC 2.95.1 16748 August 19, 1999 16749 16750 GCC 2.95 16751 July 31, 1999. This is the first release of GCC since the April 16752 1999 GCC/EGCS reunification and includes nearly a year's worth 16753 of new development and bugfixes. 16754 16755 References and Acknowledgements 16756 16757 GCC used to stand for the GNU C Compiler, but since the compiler 16758 supports several other languages aside from C, it now stands for the 16759 GNU Compiler Collection. 16760 16761 The whole suite has been extensively [1]regression tested and 16762 [2]package tested. It should be reliable and suitable for widespread 16763 use. 16764 16765 The compiler has several new optimizations, new targets, new languages 16766 and other new features. See the [3]new features page for a more 16767 complete list of new features found in the GCC 2.95 releases. 16768 16769 The sources include installation instructions in both HTML and 16770 plaintext forms in the install directory in the distribution. However, 16771 the most up to date installation instructions and [4]build/test status 16772 are on the web pages. We will update those pages as new information 16773 becomes available. 16774 16775 The GCC developers would like to thank the numerous people that have 16776 contributed new features, test results, bugfixes, etc to GCC. This 16777 [5]amazing group of volunteers is what makes GCC successful. 16778 16779 And finally, we can't in good conscience fail to mention some 16780 [6]caveats to using GCC 2.95. 16781 16782 Download GCC 2.95 from one of our many [7]mirror sites. 16783 16784 For additional information about GCC please see the [8]GCC project web 16785 server or contact the [9]GCC development mailing list. 16786 16787 16788 For questions related to the use of GCC, please consult these web 16789 pages and the [10]GCC manuals. If that fails, the 16790 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 16791 web pages and the development of GCC are welcome on our developer 16792 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 16793 archives. 16794 16795 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 16796 distribution of this entire article is permitted in any medium, 16797 provided this notice is preserved. 16798 16799 These pages are [15]maintained by the GCC team. Last modified 16800 2018-09-30[16]. 16801 16802 References 16803 16804 1. http://gcc.gnu.org/gcc-2.95/regress.html 16805 2. http://gcc.gnu.org/gcc-2.95/othertest.html 16806 3. http://gcc.gnu.org/gcc-2.95/features.html 16807 4. http://gcc.gnu.org/gcc-2.95/buildstat.html 16808 5. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 16809 6. http://gcc.gnu.org/gcc-2.95/caveats.html 16810 7. http://gcc.gnu.org/mirrors.html 16811 8. http://gcc.gnu.org/index.html 16812 9. mailto:gcc (a] gcc.gnu.org 16813 10. https://gcc.gnu.org/onlinedocs/ 16814 11. mailto:gcc-help (a] gcc.gnu.org 16815 12. mailto:gcc (a] gcc.gnu.org 16816 13. https://gcc.gnu.org/lists.html 16817 14. https://www.fsf.org/ 16818 15. https://gcc.gnu.org/about.html 16819 16. http://validator.w3.org/check/referer 16820 ====================================================================== 16821 http://gcc.gnu.org/gcc-2.95/features.html 16822 16823 GCC 2.95 New Features 16824 16825 * General Optimizer Improvements: 16826 + [1]Localized register spilling to improve speed and code 16827 density especially on small register class machines. 16828 + [2]Global CSE using lazy code motion algorithms. 16829 + [3]Improved global constant/copy propagation. 16830 + [4]Improved control flow graph analysis and manipulation. 16831 + [5]Local dead store elimination. 16832 + [6]Memory Load hoisting/store sinking in loops. 16833 + [7]Type based alias analysis is enabled by default. Note this 16834 feature will expose bugs in the Linux kernel. Please refer to 16835 the FAQ (as shipped with GCC 2.95) for additional information 16836 on this issue. 16837 + Major revamp of GIV detection, combination and simplification 16838 to improve loop performance. 16839 + Major improvements to register allocation and reloading. 16840 * New Languages and Language specific improvements 16841 + [8]Many C++ improvements. 16842 + [9]Many Fortran improvements. 16843 + [10]Java front-end has been integrated. A [11]runtime library 16844 is available separately. 16845 + [12]ISO C99 support 16846 + [13]Chill front-end and runtime has been integrated. 16847 + Boehm garbage collector support in libobjc. 16848 + More support for various pragmas which appear in vendor 16849 include files 16850 * New Targets and Target Specific Improvements 16851 + [14]SPARC backend rewrite. 16852 + -mschedule=8000 will optimize code for PA8000 class 16853 processors; -mpa-risc-2-0 will generate code for PA2.0 16854 processors 16855 + Various micro-optimizations for the ia32 port. K6 16856 optimizations 16857 + Compiler will attempt to align doubles in the stack on the 16858 ia32 port 16859 + Alpha EV6 support 16860 + PowerPC 750 16861 + RS6000/PowerPC: -mcpu=401 was added as an alias for -mcpu=403. 16862 -mcpu=e603e was added to do -mcpu=603e and -msoft-float. 16863 + c3x, c4x 16864 + HyperSPARC 16865 + SparcLite86x 16866 + sh4 16867 + Support for new systems (OpenBSD, FreeBSD, UWIN, Interix, 16868 arm-linux) 16869 + vxWorks targets include support for vxWorks threads 16870 + StrongARM 110 and ARM9 support added. ARM Scheduling 16871 parameters rewritten. 16872 + Various changes to the MIPS port to avoid assembler macros, 16873 which in turn improves performance 16874 + Various performance improvements to the i960 port. 16875 + Major rewrite of ns32k port 16876 * Other significant improvements 16877 + [15]Ability to dump cfg information and display it using vcg. 16878 + The new faster scheme for fixing vendor header files is 16879 enabled by default. 16880 + Experimental internationalization support. 16881 + multibyte character support 16882 + Some compile-time speedups for pathological problems 16883 + Better support for complex types 16884 * Plus the usual mountain of bugfixes 16885 * Core compiler is based on the gcc2 development tree from Sept 30, 16886 1998, so we have all of the [16]features found in GCC 2.8. 16887 16888 Additional Changes in GCC 2.95.1 16889 16890 * Generic bugfixes and improvements 16891 + Various documentation fixes related to the GCC/EGCS merger. 16892 + Fix memory management bug which could lead to spurious aborts, 16893 core dumps or random parsing errors in the compiler. 16894 + Fix a couple bugs in the dwarf1 and dwarf2 debug record 16895 support. 16896 + Fix infinite loop in the CSE optimizer. 16897 + Avoid undefined behavior in compiler FP emulation code 16898 + Fix install problem when prefix is overridden on the make 16899 install command. 16900 + Fix problem with unwanted installation of assert.h on some 16901 systems. 16902 + Fix problem with finding the wrong assembler in a single tree 16903 build. 16904 + Avoid increasing the known alignment of a register that is 16905 already known to be a pointer. 16906 * Platform specific bugfixes and improvements 16907 + Codegen bugfix for prologue/epilogue for cpu32 target. 16908 + Fix long long code generation bug for the Coldfire target. 16909 + Fix various aborts in the SH compiler. 16910 + Fix bugs in libgcc support library for the SH. 16911 + Fix alpha ev6 code generation bug. 16912 + Fix problems with EXIT_SUCCESS/EXIT_FAILURE redefinitions on 16913 AIX platforms. 16914 + Fix -fpic code generation bug for rs6000/ppc svr4 targets. 16915 + Fix varargs/stdarg code generation bug for rs6000/ppc svr4 16916 targets. 16917 + Fix weak symbol handling for rs6000/ppc svr4 targets. 16918 + Fix various problems with 64bit code generation for the 16919 rs6000/ppc port. 16920 + Fix codegen bug which caused tetex to be mis-compiled on the 16921 x86. 16922 + Fix compiler abort in new cfg code exposed by x86 port. 16923 + Fix out of range array reference in code convert flat 16924 registers to the x87 stacked FP register file. 16925 + Fix minor vxworks configuration bug. 16926 + Fix return type of bsearch for SunOS 4.x. 16927 * Language & Runtime specific fixes. 16928 + The G++ signature extension has been deprecated. It will be 16929 removed in the next major release of G++. Use of signatures 16930 will result in a warning from the compiler. 16931 + Several bugs relating to templates and namespaces were fixed. 16932 + A bug that caused crashes when combining templates with -g on 16933 DWARF1 platforms was fixed. 16934 + Pointers-to-members, virtual functions, and multiple 16935 inheritance should now work together correctly. 16936 + Some code-generation bugs relating to function try blocks were 16937 fixed. 16938 + G++ is a little bit more lenient with certain archaic 16939 constructs than in GCC 2.95. 16940 + Fix to prevent shared library version #s from bring truncated 16941 to 1 digit 16942 + Fix missing std:: in the libstdc++ library. 16943 + Fix stream locking problems in libio. 16944 + Fix problem in java compiler driver. 16945 16946 Additional Changes in GCC 2.95.2 16947 16948 The -fstrict-aliasing is not enabled by default for GCC 2.95.2. While 16949 the optimizations performed by -fstrict-aliasing are valid according to 16950 the C and C++ standards, the optimization have caused some problems, 16951 particularly with old non-conforming code. 16952 16953 The GCC developers are experimenting with ways to warn users about code 16954 which violates the C/C++ standards, but those warnings are not ready 16955 for widespread use at this time. Rather than wait for those warnings 16956 the GCC developers have chosen to disable -fstrict-aliasing by default 16957 for the GCC 2.95.2 release. 16958 16959 We strongly encourage developers to find and fix code which violates 16960 the C/C++ standards as -fstrict-aliasing may be enabled by default in 16961 future releases. Use the option -fstrict-aliasing to re-enable these 16962 optimizations. 16963 * Generic bugfixes and improvements 16964 + Fix incorrectly optimized memory reference in global common 16965 subexpression elimination (GCSE) optimization pass. 16966 + Fix code generation bug in regmove.c in which it could 16967 incorrectly change a "const" value. 16968 + Fix bug in optimization of conditionals involving volatile 16969 memory references. 16970 + Avoid over-allocation of stack space for some procedures. 16971 + Fixed bug in the compiler which caused incorrect optimization 16972 of an obscure series of bit manipulations, shifts and 16973 arithmetic. 16974 + Fixed register allocator bug which caused teTeX to be 16975 mis-compiled on SPARC targets. 16976 + Avoid incorrect optimization of degenerate case statements for 16977 certain targets such as the ARM. 16978 + Fix out of range memory reference in the jump optimizer. 16979 + Avoid dereferencing null pointer in fix-header. 16980 + Fix test for GCC specific features so that it is possible to 16981 bootstrap with gcc-2.6.2 and older versions of GCC. 16982 + Fix typo in scheduler which could potentially cause out of 16983 range memory accesses. 16984 + Avoid incorrect loop reversal which caused incorrect code for 16985 certain loops on PowerPC targets. 16986 + Avoid incorrect optimization of switch statements on certain 16987 targets (for example the ARM). 16988 * Platform specific bugfixes and improvements 16989 + Work around bug in Sun V5.0 compilers which caused bootstrap 16990 comparison failures on SPARC targets. 16991 + Fix SPARC backend bug which caused aborts in final.c. 16992 + Fix sparc-hal-solaris2* configuration fragments. 16993 + Fix bug in sparc block profiling. 16994 + Fix obscure code generation bug for the PARISC targets. 16995 + Define __STDC_EXT__ for HPUX configurations. 16996 + Various POWERPC64 code generation bugfixes. 16997 + Fix abort for PPC targets using ELF (ex GNU/Linux). 16998 + Fix collect2 problems for AIX targets. 16999 + Correct handling of .file directive for PPC targets. 17000 + Fix bug in fix_trunc x86 patterns. 17001 + Fix x86 port to correctly pop the FP stack for functions that 17002 return structures in memory. 17003 + Fix minor bug in strlen x86 pattern. 17004 + Use stabs debugging instead of dwarf1 for x86-solaris targets. 17005 + Fix template repository code to handle leading underscore in 17006 mangled names. 17007 + Fix weak/weak alias support for OpenBSD. 17008 + GNU/Linux for the ARM has C++ compatible include files. 17009 * Language & Runtime specific fixes. 17010 + Fix handling of constructor attribute in the C front-end which 17011 caused problems building the Chill runtime library on some 17012 targets. 17013 + Fix minor problem merging type qualifiers in the C front-end. 17014 + Fix aliasing bug for pointers and references (C/C++). 17015 + Fix incorrect "non-constant initializer bug" when -traditional 17016 or -fwritable-strings is enabled. 17017 + Fix build error for Chill front-end on SunOS. 17018 + Do not complain about duplicate instantiations when using 17019 -frepo (C++). 17020 + Fix array bounds handling in C++ front-end which caused 17021 problems with dwarf debugging information in some 17022 circumstances. 17023 + Fix minor namespace problem. 17024 + Fix problem linking java programs. 17025 17026 Additional Changes in GCC 2.95.3 17027 17028 * Generic bugfixes and improvements 17029 + Fix numerous problems that caused incorrect optimization in 17030 the register reloading code. 17031 + Fix numerous problems that caused incorrect optimization in 17032 the loop optimizer. 17033 + Fix aborts in the functions build_insn_chain and scan_loops 17034 under some circumstances. 17035 + Fix an alias analysis bug. 17036 + Fix an infinite compilation bug in the combiner. 17037 + A few problems with complex number support have been fixed. 17038 + It is no longer possible for gcc to act as a fork bomb when 17039 installed incorrectly. 17040 + The -fpack-struct option should be recognized now. 17041 + Fixed a bug that caused incorrect code to be generated due to 17042 a lost stack adjustment. 17043 * Platform specific bugfixes and improvements 17044 + Support building ARM toolchains hosted on Windows. 17045 + Fix attribute calculations in ARM toolchains. 17046 + arm-linux support has been improved. 17047 + Fix a PIC failure on sparc targets. 17048 + On ix86 targets, the regparm attribute should now work 17049 reliably. 17050 + Several updates for the h8300 port. 17051 + Fix problem building libio with glibc 2.2. 17052 17053 17054 For questions related to the use of GCC, please consult these web 17055 pages and the [17]GCC manuals. If that fails, the 17056 [18]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17057 web pages and the development of GCC are welcome on our developer 17058 list at [19]gcc (a] gcc.gnu.org. All of [20]our lists have public 17059 archives. 17060 17061 Copyright (C) [21]Free Software Foundation, Inc. Verbatim copying and 17062 distribution of this entire article is permitted in any medium, 17063 provided this notice is preserved. 17064 17065 These pages are [22]maintained by the GCC team. Last modified 17066 2018-09-30[23]. 17067 17068 References 17069 17070 1. http://gcc.gnu.org/news/spill.html 17071 2. http://gcc.gnu.org/news/lcm.html 17072 3. http://gcc.gnu.org/news/cprop.html 17073 4. http://gcc.gnu.org/news/cfg.html 17074 5. http://gcc.gnu.org/news/dse.html 17075 6. http://gcc.gnu.org/news/hoist.html 17076 7. http://gcc.gnu.org/news/alias.html 17077 8. http://gcc.gnu.org/gcc-2.95/c++features.html 17078 9. https://gcc.gnu.org/onlinedocs/gcc-3.4.6/g77/News.html 17079 10. http://gcc.gnu.org/news/gcj-announce.txt 17080 11. http://gcc.gnu.org/news/javaannounce.html 17081 12. http://gcc.gnu.org/c99status.html 17082 13. http://gcc.gnu.org/news/chill.html 17083 14. http://gcc.gnu.org/news/sparc.html 17084 15. http://gcc.gnu.org/news/egcs-vcg.html 17085 16. http://gcc.gnu.org/egcs-1.0/features-2.8.html 17086 17. https://gcc.gnu.org/onlinedocs/ 17087 18. mailto:gcc-help (a] gcc.gnu.org 17088 19. mailto:gcc (a] gcc.gnu.org 17089 20. https://gcc.gnu.org/lists.html 17090 21. https://www.fsf.org/ 17091 22. https://gcc.gnu.org/about.html 17092 23. http://validator.w3.org/check/referer 17093 ====================================================================== 17094 http://gcc.gnu.org/gcc-2.95/caveats.html 17095 17096 GCC 2.95 Caveats 17097 17098 * GCC 2.95 will issue an error for invalid asm statements that had 17099 been silently accepted by earlier versions of the compiler. This is 17100 particularly noticeable when compiling older versions of the Linux 17101 kernel (2.0.xx). Please refer to the FAQ (as shipped with GCC 2.95) 17102 for more information on this issue. 17103 * GCC 2.95 implements type based alias analysis to disambiguate 17104 memory references. Some programs, particularly the Linux kernel 17105 violate ANSI/ISO aliasing rules and therefore may not operate 17106 correctly when compiled with GCC 2.95. Please refer to the FAQ (as 17107 shipped with GCC 2.95) for more information on this issue. 17108 * GCC 2.95 has a known bug in its handling of complex variables for 17109 64bit targets. Instead of silently generating incorrect code, GCC 17110 2.95 will issue a fatal error for situations it can not handle. 17111 This primarily affects the Fortran community as Fortran makes more 17112 use of complex variables than C or C++. 17113 * GCC 2.95 has an integrated libstdc++, but does not have an 17114 integrated libg++. Furthermore old libg++ releases will not work 17115 with GCC 2.95. You can retrieve a recent copy of libg++ from the 17116 [1]GCC ftp server. 17117 Note most C++ programs only need libstdc++. 17118 * Exception handling may not work with shared libraries, particularly 17119 on alphas, hppas, rs6000/powerpc and mips based platforms. 17120 Exception handling is known to work on x86 GNU/Linux platforms with 17121 shared libraries. 17122 * In general, GCC 2.95 is more rigorous about rejecting invalid C++ 17123 code or deprecated C++ constructs than G++ 2.7, G++ 2.8, EGCS 1.0, 17124 or EGCS 1.1. As a result it may be necessary to fix C++ code before 17125 it will compile with GCC 2.95. 17126 * G++ is also converting toward the ISO C++ standard; as a result 17127 code which was previously valid (and thus accepted by other 17128 compilers and older versions of g++) may no longer be accepted. The 17129 flag -fpermissive may allow some non-conforming code to compile 17130 with GCC 2.95. 17131 * GCC 2.95 compiled C++ code is not binary compatible with EGCS 17132 1.1.x, EGCS 1.0.x or GCC 2.8.x. 17133 * GCC 2.95 does not have changes from the GCC 2.8 tree that were made 17134 between Sept 30, 1998 and April 30, 1999 (the official end of the 17135 GCC 2.8 project). Future GCC releases will include all the changes 17136 from the defunct GCC 2.8 sources. 17137 17138 17139 For questions related to the use of GCC, please consult these web 17140 pages and the [2]GCC manuals. If that fails, the 17141 [3]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17142 web pages and the development of GCC are welcome on our developer 17143 list at [4]gcc (a] gcc.gnu.org. All of [5]our lists have public archives. 17144 17145 Copyright (C) [6]Free Software Foundation, Inc. Verbatim copying and 17146 distribution of this entire article is permitted in any medium, 17147 provided this notice is preserved. 17148 17149 These pages are [7]maintained by the GCC team. Last modified 17150 2018-09-30[8]. 17151 17152 References 17153 17154 1. ftp://gcc.gnu.org/pub/gcc/infrastructure/libg++-2.8.1.3.tar.gz 17155 2. https://gcc.gnu.org/onlinedocs/ 17156 3. mailto:gcc-help (a] gcc.gnu.org 17157 4. mailto:gcc (a] gcc.gnu.org 17158 5. https://gcc.gnu.org/lists.html 17159 6. https://www.fsf.org/ 17160 7. https://gcc.gnu.org/about.html 17161 8. http://validator.w3.org/check/referer 17162 ====================================================================== 17163 http://gcc.gnu.org/egcs-1.1/index.html 17164 17165 EGCS 1.1 17166 17167 September 3, 1998: We are pleased to announce the release of EGCS 1.1. 17168 December 1, 1998: We are pleased to announce the release of EGCS 1.1.1. 17169 March 15, 1999: We are pleased to announce the release of EGCS 1.1.2. 17170 17171 EGCS is a free software project to further the development of the GNU 17172 compilers using an open development environment. 17173 17174 EGCS 1.1 is a major new release of the EGCS compiler system. It has 17175 been [1]extensively tested and is believed to be stable and suitable 17176 for widespread use. 17177 17178 EGCS 1.1 is based on an June 6, 1998 snapshot of the GCC 2.8 17179 development sources; it contains all of the new features found in GCC 17180 2.8.1 as well as all new development from GCC up to June 6, 1998. 17181 17182 EGCS 1.1 also contains many improvements and features not found in GCC 17183 or in older versions of EGCS: 17184 * Global common subexpression elimination and global constant/copy 17185 propagation (aka [2]gcse) 17186 * Ongoing improvements to the [3]alias analysis support to allow for 17187 better optimizations throughout the compiler. 17188 * Vastly improved [4]C++ compiler and integrated C++ runtime 17189 libraries. 17190 * Fixes for the /tmp symlink race security problems. 17191 * New targets including mips16, arm-thumb and 64 bit PowerPC. 17192 * Improvements to GNU Fortran (g77) compiler and runtime library made 17193 since g77 version 0.5.23. 17194 17195 See the [5]new features page for a more complete list of new features 17196 found in EGCS 1.1 releases. 17197 17198 EGCS 1.1.1 is a minor update to fix several serious problems in EGCS 17199 1.1: 17200 * General improvements and fixes 17201 + Avoid some stack overflows when compiling large functions. 17202 + Avoid incorrect loop invariant code motions. 17203 + Fix some core dumps on Linux kernel code. 17204 + Bring back the imake -Di386 and friends fix from EGCS 1.0.2. 17205 + Fix code generation problem in gcse. 17206 + Various documentation related fixes. 17207 * g++/libstdc++ improvements and fixes 17208 + MT safe EH fix for setjmp/longjmp based exception handling. 17209 + Fix a few bad interactions between optimization and exception 17210 handling. 17211 + Fixes for demangling of template names starting with "__". 17212 + Fix a bug that would fail to run destructors in some cases 17213 with -O2. 17214 + Fix 'new' of classes with virtual bases. 17215 + Fix crash building Qt on the Alpha. 17216 + Fix failure compiling WIFEXITED macro on GNU/Linux. 17217 + Fix some -frepo failures. 17218 * g77 and libf2c improvements and fixes 17219 + Various documentation fixes. 17220 + Avoid compiler crash on RAND intrinsic. 17221 + Fix minor bugs in makefiles exposed by BSD make programs. 17222 + Define _XOPEN_SOURCE for libI77 build to avoid potential 17223 problems on some 64-bit systems. 17224 + Fix problem with implicit endfile on rewind. 17225 + Fix spurious recursive I/O errors. 17226 * platform specific improvements and fixes 17227 + Match all versions of UnixWare7. 17228 + Do not assume x86 SVR4 or UnixWare targets can handle stabs. 17229 + Fix PPC/RS6000 LEGITIMIZE_ADDRESS macro and bug in conversion 17230 from unsigned ints to double precision floats. 17231 + Fix ARM ABI issue with NetBSD. 17232 + Fix a few arm code generation bugs. 17233 + Fixincludes will fix additional broken SCO OpenServer header 17234 files. 17235 + Fix a m68k backend bug which caused invalid offsets in reg+d 17236 addresses. 17237 + Fix problems with 64bit AIX 4.3 support. 17238 + Fix handling of long longs for varargs/stdarg functions on the 17239 ppc. 17240 + Minor fixes to CPP predefines for Windows. 17241 + Fix code generation problems with gpr<->fpr copies for 64bit 17242 ppc. 17243 + Fix a few coldfire code generation bugs. 17244 + Fix some more header file problems on SunOS 4.x. 17245 + Fix assert.h handling for RTEMS. 17246 + Fix Windows handling of TREE_SYMBOL_REFERENCED. 17247 + Fix x86 compiler abort in reg-stack pass. 17248 + Fix cygwin/windows problem with section attributes. 17249 + Fix Alpha code generation problem exposed by SMP Linux 17250 kernels. 17251 + Fix typo in m68k 32->64bit integer conversion. 17252 + Make sure target libraries build with -fPIC for PPC & Alpha 17253 targets. 17254 17255 EGCS 1.1.2 is a minor update to fix several serious problems in EGCS 17256 1.1.1: 17257 * General improvements and fixes 17258 + Fix bug in loop optimizer which caused the SPARC (and 17259 potentially other) ports to segfault. 17260 + Fix infinite recursion in alias analysis and combiner code. 17261 + Fix bug in regclass preferencing. 17262 + Fix incorrect loop reversal which caused incorrect code to be 17263 generated for several targets. 17264 + Fix return value for builtin memcpy. 17265 + Reduce compile time for certain loops which exposed quadratic 17266 behavior in the loop optimizer. 17267 + Fix bug which caused volatile memory to be written multiple 17268 times when only one write was needed/desired. 17269 + Fix compiler abort in caller-save.c 17270 + Fix combiner bug which caused incorrect code generation for 17271 certain division by constant operations. 17272 + Fix incorrect code generation due to a bug in range check 17273 optimizations. 17274 + Fix incorrect code generation due to mis-handling of clobbered 17275 values in CSE. 17276 + Fix compiler abort/segfault due to incorrect register 17277 splitting when unrolling loops. 17278 + Fix code generation involving autoincremented addresses with 17279 ternary operators. 17280 + Work around bug in the scheduler which caused qt to be 17281 mis-compiled on some platforms. 17282 + Fix code generation problems with -fshort-enums. 17283 + Tighten security for temporary files. 17284 + Improve compile time for codes which make heavy use of 17285 overloaded functions. 17286 + Fix multiply defined constructor/destructor symbol problems. 17287 + Avoid setting bogus RPATH environment variable during 17288 bootstrap. 17289 + Avoid GNU-make dependencies in the texinfo subdir. 17290 + Install CPP wrapper script in $(prefix)/bin if --enable-cpp. 17291 --enable-cpp=<dirname> can be used to specify an additional 17292 install directory for the cpp wrapper script. 17293 + Fix CSE bug which caused incorrect label-label refs to appear 17294 on some platforms. 17295 + Avoid linking in EH routines from libgcc if they are not 17296 needed. 17297 + Avoid obscure bug in aliasing code. 17298 + Fix bug in weak symbol handling. 17299 * Platform-specific improvements and fixes 17300 + Fix detection of PPro/PII on Unixware 7. 17301 + Fix compiler segfault when building spec99 and other programs 17302 for SPARC targets. 17303 + Fix code-generation bugs for integer and floating point 17304 conditional move instructions on the PPro/PII. 17305 + Use fixincludes to fix byteorder problems on i?86-*-sysv. 17306 + Fix build failure for the arc port. 17307 + Fix floating point format configuration for i?86-gnu port. 17308 + Fix problems with hppa1.0-hp-hpux10.20 configuration when 17309 threads are enabled. 17310 + Fix coldfire code generation bugs. 17311 + Fix "unrecognized insn" problems for Alpha and PPC ports. 17312 + Fix h8/300 code generation problem with floating point values 17313 in memory. 17314 + Fix unrecognized insn problems for the m68k port. 17315 + Fix namespace-pollution problem for the x86 port. 17316 + Fix problems with old assembler on x86 NeXT systems. 17317 + Fix PIC code-generation problems for the SPARC port. 17318 + Fix minor bug with LONG_CALLS in PowerPC SVR4 support. 17319 + Fix minor ISO namespace violation in Alpha varargs/stdarg 17320 support. 17321 + Fix incorrect "braf" instruction usage for the SH port. 17322 + Fix minor bug in va-sh which prevented its use with -ansi. 17323 + Fix problems recognizing and supporting FreeBSD. 17324 + Handle OpenBSD systems correctly. 17325 + Minor fixincludes fix for Digital UNIX 4.0B. 17326 + Fix problems with ctors/dtors in SCO shared libraries. 17327 + Abort instead of generating incorrect code for PPro/PII 17328 floating point conditional moves. 17329 + Avoid multiply defined symbols on GNU/Linux systems using 17330 libc-5.4.xx. 17331 + Fix abort in alpha compiler. 17332 * Fortran-specific fixes 17333 + Fix the IDate intrinsic (VXT) (in libg2c) so the returned year 17334 is in the documented, non-Y2K-compliant range of 0-99, instead 17335 of being returned as 100 in the year 2000. 17336 + Fix the `Date_and_Time' intrinsic (in libg2c) to return the 17337 milliseconds value properly in Values(8). 17338 + Fix the `LStat' intrinsic (in libg2c) to return device-ID 17339 information properly in SArray(7). 17340 17341 Each release includes installation instructions in both HTML and 17342 plaintext forms (see the INSTALL directory in the toplevel directory of 17343 the distribution). However, we also keep the most up to date 17344 installation instructions and [6]build/test status on our web page. We 17345 will update those pages as new information becomes available. 17346 17347 The EGCS project would like to thank the numerous people that have 17348 contributed new features, test results, bugfixes, etc. This [7]amazing 17349 group of volunteers is what makes EGCS successful. 17350 17351 And finally, we can't in good conscience fail to mention some 17352 [8]caveats to using EGCS 1.1. 17353 17354 Download EGCS from egcs.cygnus.com (USA California). 17355 17356 The EGCS 1.1 release is also available on many mirror sites. 17357 [9]Goto mirror list to find a closer site. 17358 17359 17360 For questions related to the use of GCC, please consult these web 17361 pages and the [10]GCC manuals. If that fails, the 17362 [11]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17363 web pages and the development of GCC are welcome on our developer 17364 list at [12]gcc (a] gcc.gnu.org. All of [13]our lists have public 17365 archives. 17366 17367 Copyright (C) [14]Free Software Foundation, Inc. Verbatim copying and 17368 distribution of this entire article is permitted in any medium, 17369 provided this notice is preserved. 17370 17371 These pages are [15]maintained by the GCC team. Last modified 17372 2018-09-30[16]. 17373 17374 References 17375 17376 1. http://gcc.gnu.org/egcs-1.1/egcs-1.1-test.html 17377 2. http://gcc.gnu.org/news/gcse.html 17378 3. http://gcc.gnu.org/news/alias.html 17379 4. http://gcc.gnu.org/egcs-1.1/c++features.html 17380 5. http://gcc.gnu.org/egcs-1.1/features.html 17381 6. http://gcc.gnu.org/egcs-1.1/buildstat.html 17382 7. https://gcc.gnu.org/onlinedocs/gcc/Contributors.html 17383 8. http://gcc.gnu.org/egcs-1.1/caveats.html 17384 9. http://gcc.gnu.org/mirrors.html 17385 10. https://gcc.gnu.org/onlinedocs/ 17386 11. mailto:gcc-help (a] gcc.gnu.org 17387 12. mailto:gcc (a] gcc.gnu.org 17388 13. https://gcc.gnu.org/lists.html 17389 14. https://www.fsf.org/ 17390 15. https://gcc.gnu.org/about.html 17391 16. http://validator.w3.org/check/referer 17392 ====================================================================== 17393 http://gcc.gnu.org/egcs-1.1/features.html 17394 17395 EGCS 1.1 new features 17396 17397 * Integrated GNU Fortran (g77) compiler and runtime library with 17398 improvements, based on g77 version 0.5.23. 17399 * Vast improvements in the C++ compiler; so many they have [1]page of 17400 their own! 17401 * Compiler implements [2]global common subexpression elimination and 17402 global copy/constant propagation. 17403 * More major improvements in the [3]alias analysis code. 17404 * More major improvements in the exception handling code to improve 17405 performance, lower static overhead and provide the infrastructure 17406 for future improvements. 17407 * The infamous /tmp symlink race security problems have been fixed. 17408 * The regmove optimization pass has been nearly completely rewritten 17409 to improve performance of generated code. 17410 * The compiler now recomputes register usage information before local 17411 register allocation. By providing more accurate information to the 17412 priority based allocator, we get better register allocation. 17413 * The register reloading phase of the compiler optimizes spill code 17414 much better than in previous releases. 17415 * Some bad interactions between the register allocator and 17416 instruction scheduler have been fixed, resulting in much better 17417 code for certain programs. Additionally, we have tuned the 17418 scheduler in various ways to improve performance of generated code 17419 for some architectures. 17420 * The compiler's branch shortening algorithms have been significantly 17421 improved to work better on targets which align jump targets. 17422 * The compiler now supports -Os to prefer optimizing for code space 17423 over optimizing for code speed. 17424 * The compiler will now totally eliminate library calls which compute 17425 constant values. This primarily helps targets with no integer 17426 div/mul support and targets without floating point support. 17427 * The compiler now supports an extensive "--help" option. 17428 * cpplib has been greatly improved and may be suitable for limited 17429 use. 17430 * Memory footprint for the compiler has been significantly reduced 17431 for some pathological cases. 17432 * The time to build EGCS has been improved for certain targets 17433 (particularly the alpha and mips platforms). 17434 * Many infrastructure improvements throughout the compiler, plus the 17435 usual mountain of bugfixes and minor improvements. 17436 * Target dependent improvements: 17437 + SPARC port now includes V8 plus and V9 support as well as 17438 performance tuning for Ultra class machines. The SPARC port 17439 now uses the Haifa scheduler. 17440 + Alpha port has been tuned for the EV6 processor and has an 17441 optimized expansion of memcpy/bzero. The Alpha port now uses 17442 the Haifa scheduler. 17443 + RS6000/PowerPC: support for the Power64 architecture and AIX 17444 4.3. The RS6000/PowerPC port now uses the Haifa scheduler. 17445 + x86: Alignment of static store data and jump targets is per 17446 Intel recommendations now. Various improvements throughout the 17447 x86 port to improve performance on Pentium processors 17448 (including improved epilogue sequences for Pentium chips and 17449 backend improvements which should help register allocation on 17450 all x86 variants. Conditional move support has been fixed and 17451 enabled for PPro processors. The x86 port also better supports 17452 64bit operations now. Unixware 7, a System V Release 5 target, 17453 is now supported and SCO OpenServer targets can support GAS. 17454 + MIPS has improved multiply/multiply-add support and now 17455 includes mips16 ISA support. 17456 + M68k has many micro-optimizations and Coldfire fixes. 17457 * Core compiler is based on the GCC development tree from June 9, 17458 1998, so we have all of the [4]features found in GCC 2.8. 17459 17460 17461 For questions related to the use of GCC, please consult these web 17462 pages and the [5]GCC manuals. If that fails, the 17463 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17464 web pages and the development of GCC are welcome on our developer 17465 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 17466 17467 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 17468 distribution of this entire article is permitted in any medium, 17469 provided this notice is preserved. 17470 17471 These pages are [10]maintained by the GCC team. Last modified 17472 2018-09-30[11]. 17473 17474 References 17475 17476 1. http://gcc.gnu.org/egcs-1.1/c++features.html 17477 2. http://gcc.gnu.org/news/gcse.html 17478 3. http://gcc.gnu.org/news/alias.html 17479 4. http://gcc.gnu.org/egcs-1.0/features-2.8.html 17480 5. https://gcc.gnu.org/onlinedocs/ 17481 6. mailto:gcc-help (a] gcc.gnu.org 17482 7. mailto:gcc (a] gcc.gnu.org 17483 8. https://gcc.gnu.org/lists.html 17484 9. https://www.fsf.org/ 17485 10. https://gcc.gnu.org/about.html 17486 11. http://validator.w3.org/check/referer 17487 ====================================================================== 17488 http://gcc.gnu.org/egcs-1.1/caveats.html 17489 17490 EGCS 1.1 Caveats 17491 17492 * EGCS has an integrated libstdc++, but does not have an integrated 17493 libg++. Furthermore old libg++ releases will not work with EGCS; HJ 17494 Lu has made a libg++-2.8.1.2 snapshot available which may work with 17495 EGCS. 17496 Note most C++ programs only need libstdc++. 17497 * Exception handling may not work with shared libraries, particularly 17498 on alphas, hppas, rs6000/powerpc and mips based platforms. 17499 Exception handling is known to work on x86-linux platforms with 17500 shared libraries. 17501 * Some versions of the Linux kernel have bugs which prevent them from 17502 being compiled or from running when compiled by EGCS. See the FAQ 17503 (as shipped with EGCS 1.1) for additional information. 17504 * In general, EGCS is more rigorous about rejecting invalid C++ code 17505 or deprecated C++ constructs than g++-2.7, g++-2.8 or EGCS 1.0. As 17506 a result it may be necessary to fix C++ code before it will compile 17507 with EGCS. 17508 * G++ is also converting toward the ISO C++ standard; as a result 17509 code which was previously valid (and thus accepted by other 17510 compilers and older versions of g++) may no longer be accepted. 17511 * EGCS 1.1 compiled C++ code is not binary compatible with EGCS 1.0.x 17512 or GCC 2.8.x due to changes necessary to support thread safe 17513 exception handling. 17514 17515 17516 For questions related to the use of GCC, please consult these web 17517 pages and the [1]GCC manuals. If that fails, the 17518 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17519 web pages and the development of GCC are welcome on our developer 17520 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 17521 17522 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 17523 distribution of this entire article is permitted in any medium, 17524 provided this notice is preserved. 17525 17526 These pages are [6]maintained by the GCC team. Last modified 17527 2018-09-30[7]. 17528 17529 References 17530 17531 1. https://gcc.gnu.org/onlinedocs/ 17532 2. mailto:gcc-help (a] gcc.gnu.org 17533 3. mailto:gcc (a] gcc.gnu.org 17534 4. https://gcc.gnu.org/lists.html 17535 5. https://www.fsf.org/ 17536 6. https://gcc.gnu.org/about.html 17537 7. http://validator.w3.org/check/referer 17538 ====================================================================== 17539 http://gcc.gnu.org/egcs-1.0/index.html 17540 17541 EGCS 1.0 17542 17543 December 3, 1997: We are pleased to announce the release of EGCS 1.0. 17544 January 6, 1998: We are pleased to announce the release of EGCS 1.0.1. 17545 March 16, 1998: We are pleased to announce the release of EGCS 1.0.2. 17546 May 15, 1998 We are pleased to announce the release of EGCS 1.0.3. 17547 17548 EGCS is a collaborative effort involving several groups of hackers 17549 using an open development model to accelerate development and testing 17550 of GNU compilers and runtime libraries. 17551 17552 An important goal of EGCS is to allow wide scale testing of 17553 experimental features and optimizations; therefore, EGCS contains some 17554 features and optimizations which are still under development. However, 17555 EGCS has been carefully tested and should be comparable in quality to 17556 most GCC releases. 17557 17558 EGCS 1.0 is based on an August 2, 1997 snapshot of the GCC 2.8 17559 development sources; it contains nearly all of the new features found 17560 in GCC 2.8. 17561 17562 EGCS 1.0 also contains many improvements and features not found in GCC 17563 2.7 and even the GCC 2.8 series (which was released after the original 17564 EGCS 1.0 release). 17565 * Integrated C++ runtime libraries, including support for most major 17566 GNU/Linux systems! 17567 * The integrated libstdc++ library includes a verbatim copy of SGI's 17568 STL release. 17569 * Integrated GNU Fortran compiler. 17570 * New instruction scheduler. 17571 * New alias analysis code. 17572 17573 See the [1]new features page for a more complete list of new features. 17574 17575 EGCS 1.0.1 is a minor update to the EGCS 1.0 compiler to fix a few 17576 critical bugs and add support for Red Hat 5.0 Linux. Changes since the 17577 EGCS 1.0 release: 17578 * Add support for Red Hat 5.0 Linux and better support for Linux 17579 systems using glibc2. 17580 Many programs failed to link when compiled with EGCS 1.0 on Red Hat 17581 5.0 or on systems with newer versions of glibc2. EGCS 1.0.1 should 17582 fix these problems. 17583 * Compatibility with both EGCS 1.0 and GCC 2.8 libgcc exception 17584 handling interfaces. 17585 To avoid future compatibility problems, we strongly urge anyone who 17586 is planning on distributing shared libraries that contain C++ code 17587 to upgrade to EGCS 1.0.1 first. 17588 Soon after EGCS 1.0 was released, the GCC developers made some 17589 incompatible changes in libgcc's exception handling interfaces. 17590 These changes were needed to solve problems on some platforms. This 17591 means that GCC 2.8.0, when released, will not be seamlessly 17592 compatible with shared libraries built by EGCS 1.0. The reason is 17593 that the libgcc.a in GCC 2.8.0 will not contain a function needed 17594 by the old interface. 17595 The result of this is that there may be compatibility problems with 17596 shared libraries built by EGCS 1.0 when used with GCC 2.8.0. 17597 With EGCS 1.0.1, generated code uses the new (GCC 2.8.0) interface, 17598 and libgcc.a has the support routines for both the old and the new 17599 interfaces (so EGCS 1.0.1 and EGCS 1.0 code can be freely mixed, 17600 and EGCS 1.0.1 and GCC 2.8.0 code can be freely mixed). 17601 The maintainers of GCC 2.x have decided against including seamless 17602 support for the old interface in 2.8.0, since it was never 17603 "official", so to avoid future compatibility problems we recommend 17604 against distributing any shared libraries built by EGCS 1.0 that 17605 contain C++ code (upgrade to 1.0.1 and use that). 17606 * Various bugfixes in the x86, hppa, mips, and rs6000/ppc back ends. 17607 The x86 changes fix code generation errors exposed when building 17608 glibc2 and the usual GNU/Linux dynamic linker (ld.so). 17609 The hppa change fixes a compiler abort when configured for use with 17610 RTEMS. 17611 The MIPS changes fix problems with the definition of LONG_MAX on 17612 newer systems, allow for command line selection of the target ABI, 17613 and fix one code generation problem. 17614 The rs6000/ppc change fixes some problems with passing structures 17615 to varargs/stdarg functions. 17616 * A few machine independent bugfixes, mostly to fix code generation 17617 errors when building Linux kernels or glibc. 17618 * Fix a few critical exception handling and template bugs in the C++ 17619 compiler. 17620 * Fix Fortran namelist bug on alphas. 17621 * Fix build problems on x86-solaris systems. 17622 17623 EGCS 1.0.2 is a minor update to the EGCS 1.0.1 compiler to fix several 17624 serious problems in EGCS 1.0.1. 17625 * General improvements and fixes 17626 + Memory consumption significantly reduced, especially for 17627 templates and inline functions. 17628 + Fix various problems with glibc2.1. 17629 + Fix loop optimization bug exposed by rs6000/ppc port. 17630 + Fix to avoid potential code generation problems in jump.c. 17631 + Fix some undefined symbol problems in dwarf1 debug support. 17632 * g++/libstdc++ improvements and fixes 17633 + libstdc++ in the EGCS release has been updated and should be 17634 link compatible with libstdc++-2.8. 17635 + Various fixes in libio/libstdc++ to work better on GNU/Linux 17636 systems. 17637 + Fix problems with duplicate symbols on systems that do not 17638 support weak symbols. 17639 + Memory corruption bug and undefined symbols in bastring have 17640 been fixed. 17641 + Various exception handling fixes. 17642 + Fix compiler abort for very long thunk names. 17643 * g77 improvements and fixes 17644 + Fix compiler crash for omitted bound in Fortran CASE 17645 statement. 17646 + Add missing entries to g77 lang-options. 17647 + Fix problem with -fpedantic in the g77 compiler. 17648 + Fix "backspace" problem with g77 on alphas. 17649 + Fix x86 backend problem with Fortran literals and -fpic. 17650 + Fix some of the problems with negative subscripts for g77 on 17651 alphas. 17652 + Fixes for Fortran builds on cygwin32/mingw32. 17653 * platform specific improvements and fixes 17654 + Fix long double problems on x86 (exposed by glibc). 17655 + x86 ports define i386 again to keep imake happy. 17656 + Fix exception handling support on NetBSD ports. 17657 + Several changes to collect2 to fix many problems with AIX. 17658 + Define __ELF__ for GNU/Linux on rs6000. 17659 + Fix -mcall-linux problem on GNU/Linux on rs6000. 17660 + Fix stdarg/vararg problem for GNU/Linux on rs6000. 17661 + Allow autoconf to select a proper install problem on AIX 3.1. 17662 + m68k port support includes -mcpu32 option as well as cpu32 17663 multilibs. 17664 + Fix stdarg bug for irix6. 17665 + Allow EGCS to build on irix5 without the gnu assembler. 17666 + Fix problem with static linking on sco5. 17667 + Fix bootstrap on sco5 with native compiler. 17668 + Fix for abort building newlib on H8 target. 17669 + Fix fixincludes handling of math.h on SunOS. 17670 + Minor fix for Motorola 3300 m68k systems. 17671 17672 EGCS 1.0.3 is a minor update to the EGCS 1.0.2 compiler to fix a few 17673 problems reported by Red Hat for builds of Red Hat 5.1. 17674 * Generic bugfixes: 17675 + Fix a typo in the libio library which resulted in incorrect 17676 behavior of istream::get. 17677 + Fix the Fortran negative array index problem. 17678 + Fix a major problem with the ObjC runtime thread support 17679 exposed by glibc2. 17680 + Reduce memory consumption of the Haifa scheduler. 17681 * Target specific bugfixes: 17682 + Fix one x86 floating point code generation bug exposed by 17683 glibc2 builds. 17684 + Fix one x86 internal compiler error exposed by glibc2 builds. 17685 + Fix profiling bugs on the Alpha. 17686 + Fix ImageMagick & emacs 20.2 build problems on the Alpha. 17687 + Fix rs6000/ppc bug when converting values from integer types 17688 to floating point types. 17689 17690 The EGCS 1.0 releases include installation instructions in both HTML 17691 and plaintext forms (see the INSTALL directory in the toplevel 17692 directory of the distribution). However, we also keep the most up to 17693 date installation instructions and [2]build/test status on our web 17694 page. We will update those pages as new information becomes available. 17695 17696 And, we can't in good conscience fail to mention some [3]caveats to 17697 using EGCS. 17698 17699 Update: Big thanks to Stanford for providing a high speed link for 17700 downloading EGCS (go.cygnus.com)! 17701 17702 Download EGCS from ftp.cygnus.com (USA California) or go.cygnus.com 17703 (USA California -- High speed link provided by Stanford). 17704 17705 The EGCS 1.0 release is also available many mirror sites. 17706 [4]Goto mirror list to find a closer site 17707 17708 We'd like to thank the numerous people that have contributed new 17709 features, test results, bugfixes, etc. Unfortunately, they're far too 17710 numerous to mention by name. 17711 17712 17713 For questions related to the use of GCC, please consult these web 17714 pages and the [5]GCC manuals. If that fails, the 17715 [6]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17716 web pages and the development of GCC are welcome on our developer 17717 list at [7]gcc (a] gcc.gnu.org. All of [8]our lists have public archives. 17718 17719 Copyright (C) [9]Free Software Foundation, Inc. Verbatim copying and 17720 distribution of this entire article is permitted in any medium, 17721 provided this notice is preserved. 17722 17723 These pages are [10]maintained by the GCC team. Last modified 17724 2018-09-30[11]. 17725 17726 References 17727 17728 1. http://gcc.gnu.org/egcs-1.0/features.html 17729 2. http://gcc.gnu.org/egcs-1.0/buildstat.html 17730 3. http://gcc.gnu.org/egcs-1.0/caveats.html 17731 4. http://gcc.gnu.org/mirrors.html 17732 5. https://gcc.gnu.org/onlinedocs/ 17733 6. mailto:gcc-help (a] gcc.gnu.org 17734 7. mailto:gcc (a] gcc.gnu.org 17735 8. https://gcc.gnu.org/lists.html 17736 9. https://www.fsf.org/ 17737 10. https://gcc.gnu.org/about.html 17738 11. http://validator.w3.org/check/referer 17739 ====================================================================== 17740 http://gcc.gnu.org/egcs-1.0/features.html 17741 17742 EGCS 1.0 features 17743 17744 * Core compiler is based on the gcc2 development tree from Aug 2, 17745 1997, so we have most of the [1]features found in GCC 2.8. 17746 * Integrated GNU Fortran compiler based on g77-0.5.22-19970929. 17747 * Vast improvements in the C++ compiler; so many they have [2]page of 17748 their own! 17749 * Integrated C++ runtime libraries, including support for most major 17750 GNU/Linux systems! 17751 * New instruction scheduler from IBM Haifa which includes support for 17752 function wide instruction scheduling as well as superscalar 17753 scheduling. 17754 * Significantly improved alias analysis code. 17755 * Improved register allocation for two address machines. 17756 * Significant code generation improvements for Fortran code on 17757 Alphas. 17758 * Various optimizations from the g77 project as well as improved loop 17759 optimizations. 17760 * Dwarf2 debug format support for some targets. 17761 * egcs libstdc++ includes the SGI STL implementation without changes. 17762 * As a result of these and other changes, egcs libstc++ is not binary 17763 compatible with previous releases of libstdc++. 17764 * Various new ports -- UltraSPARC, Irix6.2 & Irix6.3 support, The SCO 17765 Openserver 5 family (5.0.{0,2,4} and Internet FastStart 1.0 and 17766 1.1), Support for RTEMS on several embedded targets, Support for 17767 arm-linux, Mitsubishi M32R, Hitachi H8/S, Matsushita MN102 and 17768 MN103, NEC V850, Sparclet, Solaris & GNU/Linux on PowerPCs, etc. 17769 * Integrated testsuites for gcc, g++, g77, libstdc++ and libio. 17770 * RS6000/PowerPC ports generate code which can run on all 17771 RS6000/PowerPC variants by default. 17772 * -mcpu= and -march= switches for the x86 port to allow better 17773 control over how the x86 port generates code. 17774 * Includes the template repository patch (aka repo patch); note the 17775 new template code makes repo obsolete for ELF systems using gnu-ld 17776 such as GNU/Linux. 17777 * Plus the usual assortment of bugfixes and improvements. 17778 17779 17780 For questions related to the use of GCC, please consult these web 17781 pages and the [3]GCC manuals. If that fails, the 17782 [4]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17783 web pages and the development of GCC are welcome on our developer 17784 list at [5]gcc (a] gcc.gnu.org. All of [6]our lists have public archives. 17785 17786 Copyright (C) [7]Free Software Foundation, Inc. Verbatim copying and 17787 distribution of this entire article is permitted in any medium, 17788 provided this notice is preserved. 17789 17790 These pages are [8]maintained by the GCC team. Last modified 17791 2018-09-30[9]. 17792 17793 References 17794 17795 1. http://gcc.gnu.org/egcs-1.0/features-2.8.html 17796 2. http://gcc.gnu.org/egcs-1.0/c++features.html 17797 3. https://gcc.gnu.org/onlinedocs/ 17798 4. mailto:gcc-help (a] gcc.gnu.org 17799 5. mailto:gcc (a] gcc.gnu.org 17800 6. https://gcc.gnu.org/lists.html 17801 7. https://www.fsf.org/ 17802 8. https://gcc.gnu.org/about.html 17803 9. http://validator.w3.org/check/referer 17804 ====================================================================== 17805 http://gcc.gnu.org/egcs-1.0/caveats.html 17806 17807 EGCS 1.0 Caveats 17808 17809 * EGCS has an integrated libstdc++, but does not have an integrated 17810 libg++. Furthermore old libg++ releases will not work with egc; HJ 17811 Lu has made a libg++-2.8.1.2 available which may work with EGCS. 17812 Note most C++ programs only need libstdc++. 17813 * Note that using -pedantic or -Wreturn-type can cause an explosion 17814 in the amount of memory needed for template-heavy C++ code, such as 17815 code that uses STL. Also note that -Wall includes -Wreturn-type, so 17816 if you use -Wall you will need to specify -Wno-return-type to turn 17817 it off. 17818 * Exception handling may not work with shared libraries, particularly 17819 on alphas, hppas, and mips based platforms. Exception handling is 17820 known to work on x86-linux platforms with shared libraries. 17821 * Some versions of the Linux kernel have bugs which prevent them from 17822 being compiled or from running when compiled by EGCS. See the FAQ 17823 (as shipped with EGCS 1.0) for additional information. 17824 * In general, EGCS is more rigorous about rejecting invalid C++ code 17825 or deprecated C++ constructs than G++ 2.7. As a result it may be 17826 necessary to fix C++ code before it will compile with EGCS. 17827 * G++ is also aggressively tracking the C++ standard; as a result 17828 code which was previously valid (and thus accepted by other 17829 compilers and older versions of G++) may no longer be accepted. 17830 * EGCS 1.0 may not work with Red Hat Linux 5.0 on all targets. EGCS 17831 1.0.x and later releases should work with Red Hat Linux 5.0. 17832 17833 17834 For questions related to the use of GCC, please consult these web 17835 pages and the [1]GCC manuals. If that fails, the 17836 [2]gcc-help (a] gcc.gnu.org mailing list might help. Comments on these 17837 web pages and the development of GCC are welcome on our developer 17838 list at [3]gcc (a] gcc.gnu.org. All of [4]our lists have public archives. 17839 17840 Copyright (C) [5]Free Software Foundation, Inc. Verbatim copying and 17841 distribution of this entire article is permitted in any medium, 17842 provided this notice is preserved. 17843 17844 These pages are [6]maintained by the GCC team. Last modified 17845 2018-09-30[7]. 17846 17847 References 17848 17849 1. https://gcc.gnu.org/onlinedocs/ 17850 2. mailto:gcc-help (a] gcc.gnu.org 17851 3. mailto:gcc (a] gcc.gnu.org 17852 4. https://gcc.gnu.org/lists.html 17853 5. https://www.fsf.org/ 17854 6. https://gcc.gnu.org/about.html 17855 7. http://validator.w3.org/check/referer 17856 ====================================================================== 17857
Indexes created Tue Oct 27 00:25:24 UTC 2026