1 2Release 3.6.1 (16 February 2011) 3~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 43.6.1 is a bug fix release. It adds support for some SSE4 5instructions that were omitted in 3.6.0 due to lack of time. Initial 6support for glibc-2.13 has been added. A number of bugs causing 7crashing or assertion failures have been fixed. 8 9The following bugs have been fixed or resolved. Note that "n-i-bz" 10stands for "not in bugzilla" -- that is, a bug that was reported to us 11but never got a bugzilla entry. We encourage you to file bugs in 12bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than 13mailing the developers (or mailing lists) directly -- bugs that are 14not entered into bugzilla tend to get forgotten about or ignored. 15 16To see details of a given bug, visit 17https://bugs.kde.org/show_bug.cgi?id=XXXXXX 18where XXXXXX is the bug number as listed below. 19 20188572 Valgrind on Mac should suppress setenv() mem leak 21194402 vex amd64->IR: 0x48 0xF 0xAE 0x4 (proper FX{SAVE,RSTOR} support) 22210481 vex amd64->IR: Assertion `sz == 2 || sz == 4' failed (REX.W POPQ) 23246152 callgrind internal error after pthread_cancel on 32 Bit Linux 24250038 ppc64: Altivec LVSR and LVSL instructions fail their regtest 25254420 memory pool tracking broken 26254957 Test code failing to compile due to changes in memcheck.h 27255009 helgrind/drd: crash on chmod with invalid parameter 28255130 readdwarf3.c parse_type_DIE confused by GNAT Ada types 29255355 helgrind/drd: crash on threaded programs doing fork 30255358 == 255355 31255418 (SSE4.x) rint call compiled with ICC 32255822 --gen-suppressions can create invalid files: "too many callers [...]" 33255888 closing valgrindoutput tag outputted to log-stream on error 34255963 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x9 0xDB 0x0 (ROUNDPD) 35255966 Slowness when using mempool annotations 36256387 vex x86->IR: 0xD4 0xA 0x2 0x7 (AAD and AAM) 37256600 super-optimized strcasecmp() false positive 38256669 vex amd64->IR: Unhandled LOOPNEL insn on amd64 39256968 (SSE4.x) vex amd64->IR: 0x66 0xF 0x38 0x10 0xD3 0x66 (BLENDVPx) 40257011 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0xE 0xFD 0xA0 (PBLENDW) 41257063 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x8 0xC0 0x0 (ROUNDPS) 42257276 Missing case in memcheck --track-origins=yes 43258870 (SSE4.x) Add support for EXTRACTPS SSE 4.1 instruction 44261966 (SSE4.x) support for CRC32B and CRC32Q is lacking (also CRC32{W,L}) 45262985 VEX regression in valgrind 3.6.0 in handling PowerPC VMX 46262995 (SSE4.x) crash when trying to valgrind gcc-snapshot (PCMPxSTRx $0) 47263099 callgrind_annotate counts Ir improperly [...] 48263877 undefined coprocessor instruction on ARMv7 49265964 configure FAIL with glibc-2.13 50n-i-bz Fix compile error w/ icc-12.x in guest_arm_toIR.c 51n-i-bz Docs: fix bogus descriptions for VALGRIND_CREATE_BLOCK et al 52n-i-bz Massif: don't assert on shmat() with --pages-as-heap=yes 53n-i-bz Bug fixes and major speedups for the exp-DHAT space profiler 54n-i-bz DRD: disable --free-is-write due to implementation difficulties 55 56(3.6.1: 16 February 2011, vex r2103, valgrind r11561). 57 58 59 60Release 3.6.0 (21 October 2010) 61~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 623.6.0 is a feature release with many significant improvements and the 63usual collection of bug fixes. 64 65This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux, 66PPC64/Linux, X86/Darwin and AMD64/Darwin. Support for recent distros 67and toolchain components (glibc 2.12, gcc 4.5, OSX 10.6) has been added. 68 69 ------------------------- 70 71Here are some highlights. Details are shown further down: 72 73* Support for ARM/Linux. 74 75* Support for recent Linux distros: Ubuntu 10.10 and Fedora 14. 76 77* Support for Mac OS X 10.6, both 32- and 64-bit executables. 78 79* Support for the SSE4.2 instruction set. 80 81* Enhancements to the Callgrind profiler, including the ability to 82 handle CPUs with three levels of cache. 83 84* A new experimental heap profiler, DHAT. 85 86* A huge number of bug fixes and small enhancements. 87 88 ------------------------- 89 90Here are details of the above changes, together with descriptions of 91many other changes, and a list of fixed bugs. 92 93* ================== PLATFORM CHANGES ================= 94 95* Support for ARM/Linux. Valgrind now runs on ARMv7 capable CPUs 96 running Linux. It is known to work on Ubuntu 10.04, Ubuntu 10.10, 97 and Maemo 5, so you can run Valgrind on your Nokia N900 if you want. 98 99 This requires a CPU capable of running the ARMv7-A instruction set 100 (Cortex A5, A8 and A9). Valgrind provides fairly complete coverage 101 of the user space instruction set, including ARM and Thumb integer 102 code, VFPv3, NEON and V6 media instructions. The Memcheck, 103 Cachegrind and Massif tools work properly; other tools work to 104 varying degrees. 105 106* Support for recent Linux distros (Ubuntu 10.10 and Fedora 14), along 107 with support for recent releases of the underlying toolchain 108 components, notably gcc-4.5 and glibc-2.12. 109 110* Support for Mac OS X 10.6, both 32- and 64-bit executables. 64-bit 111 support also works much better on OS X 10.5, and is as solid as 112 32-bit support now. 113 114* Support for the SSE4.2 instruction set. SSE4.2 is supported in 115 64-bit mode. In 32-bit mode, support is only available up to and 116 including SSSE3. Some exceptions: SSE4.2 AES instructions are not 117 supported in 64-bit mode, and 32-bit mode does in fact support the 118 bare minimum SSE4 instructions to needed to run programs on Mac OS X 119 10.6 on 32-bit targets. 120 121* Support for IBM POWER6 cpus has been improved. The Power ISA up to 122 and including version 2.05 is supported. 123 124* ==================== TOOL CHANGES ==================== 125 126* Cachegrind has a new processing script, cg_diff, which finds the 127 difference between two profiles. It's very useful for evaluating 128 the performance effects of a change in a program. 129 130 Related to this change, the meaning of cg_annotate's (rarely-used) 131 --threshold option has changed; this is unlikely to affect many 132 people, if you do use it please see the user manual for details. 133 134* Callgrind now can do branch prediction simulation, similar to 135 Cachegrind. In addition, it optionally can count the number of 136 executed global bus events. Both can be used for a better 137 approximation of a "Cycle Estimation" as derived event (you need to 138 update the event formula in KCachegrind yourself). 139 140* Cachegrind and Callgrind now refer to the LL (last-level) cache 141 rather than the L2 cache. This is to accommodate machines with 142 three levels of caches -- if Cachegrind/Callgrind auto-detects the 143 cache configuration of such a machine it will run the simulation as 144 if the L2 cache isn't present. This means the results are less 145 likely to match the true result for the machine, but 146 Cachegrind/Callgrind's results are already only approximate, and 147 should not be considered authoritative. The results are still 148 useful for giving a general idea about a program's locality. 149 150* Massif has a new option, --pages-as-heap, which is disabled by 151 default. When enabled, instead of tracking allocations at the level 152 of heap blocks (as allocated with malloc/new/new[]), it instead 153 tracks memory allocations at the level of memory pages (as mapped by 154 mmap, brk, etc). Each mapped page is treated as its own block. 155 Interpreting the page-level output is harder than the heap-level 156 output, but this option is useful if you want to account for every 157 byte of memory used by a program. 158 159* DRD has two new command-line options: --free-is-write and 160 --trace-alloc. The former allows to detect reading from already freed 161 memory, and the latter allows tracing of all memory allocations and 162 deallocations. 163 164* DRD has several new annotations. Custom barrier implementations can 165 now be annotated, as well as benign races on static variables. 166 167* DRD's happens before / happens after annotations have been made more 168 powerful, so that they can now also be used to annotate e.g. a smart 169 pointer implementation. 170 171* Helgrind's annotation set has also been drastically improved, so as 172 to provide to users a general set of annotations to describe locks, 173 semaphores, barriers and condition variables. Annotations to 174 describe thread-safe reference counted heap objects have also been 175 added. 176 177* Memcheck has a new command-line option, --show-possibly-lost, which 178 is enabled by default. When disabled, the leak detector will not 179 show possibly-lost blocks. 180 181* A new experimental heap profiler, DHAT (Dynamic Heap Analysis Tool), 182 has been added. DHAT keeps track of allocated heap blocks, and also 183 inspects every memory reference to see which block (if any) is being 184 accessed. This gives a lot of insight into block lifetimes, 185 utilisation, turnover, liveness, and the location of hot and cold 186 fields. You can use DHAT to do hot-field profiling. 187 188* ==================== OTHER CHANGES ==================== 189 190* Improved support for unfriendly self-modifying code: the extra 191 overhead incurred by --smc-check=all has been reduced by 192 approximately a factor of 5 as compared with 3.5.0. 193 194* Ability to show directory names for source files in error messages. 195 This is combined with a flexible mechanism for specifying which 196 parts of the paths should be shown. This is enabled by the new flag 197 --fullpath-after. 198 199* A new flag, --require-text-symbol, which will stop the run if a 200 specified symbol is not found it a given shared object when it is 201 loaded into the process. This makes advanced working with function 202 intercepting and wrapping safer and more reliable. 203 204* Improved support for the Valkyrie GUI, version 2.0.0. GUI output 205 and control of Valgrind is now available for the tools Memcheck and 206 Helgrind. XML output from Valgrind is available for Memcheck, 207 Helgrind and exp-Ptrcheck. 208 209* More reliable stack unwinding on amd64-linux, particularly in the 210 presence of function wrappers, and with gcc-4.5 compiled code. 211 212* Modest scalability (performance improvements) for massive 213 long-running applications, particularly for those with huge amounts 214 of code. 215 216* Support for analyzing programs running under Wine with has been 217 improved. The header files <valgrind/valgrind.h>, 218 <valgrind/memcheck.h> and <valgrind/drd.h> can now be used in 219 Windows-programs compiled with MinGW or one of the Microsoft Visual 220 Studio compilers. 221 222* A rare but serious error in the 64-bit x86 CPU simulation was fixed. 223 The 32-bit simulator was not affected. This did not occur often, 224 but when it did would usually crash the program under test. 225 Bug 245925. 226 227* A large number of bugs were fixed. These are shown below. 228 229* A number of bugs were investigated, and were candidates for fixing, 230 but are not fixed in 3.6.0, due to lack of developer time. They may 231 get fixed in later releases. They are: 232 233 194402 vex amd64->IR: 0x48 0xF 0xAE 0x4 0x24 0x49 (FXSAVE64) 234 212419 false positive "lock order violated" (A+B vs A) 235 213685 Undefined value propagates past dependency breaking instruction 236 216837 Incorrect instrumentation of NSOperationQueue on Darwin 237 237920 valgrind segfault on fork failure 238 242137 support for code compiled by LLVM-2.8 239 242423 Another unknown Intel cache config value 240 243232 Inconsistent Lock Orderings report with trylock 241 243483 ppc: callgrind triggers VEX assertion failure 242 243935 Helgrind: implementation of ANNOTATE_HAPPENS_BEFORE() is wrong 243 244677 Helgrind crash hg_main.c:616 (map_threads_lookup): Assertion 244 'thr' failed. 245 246152 callgrind internal error after pthread_cancel on 32 Bit Linux 246 249435 Analyzing wine programs with callgrind triggers a crash 247 250038 ppc64: Altivec lvsr and lvsl instructions fail their regtest 248 250065 Handling large allocations 249 250101 huge "free" memory usage due to m_mallocfree.c 250 "superblocks fragmentation" 251 251569 vex amd64->IR: 0xF 0x1 0xF9 0x8B 0x4C 0x24 (RDTSCP) 252 252091 Callgrind on ARM does not detect function returns correctly 253 252600 [PATCH] Allow lhs to be a pointer for shl/shr 254 254420 memory pool tracking broken 255 n-i-bz support for adding symbols for JIT generated code 256 257 258The following bugs have been fixed or resolved. Note that "n-i-bz" 259stands for "not in bugzilla" -- that is, a bug that was reported to us 260but never got a bugzilla entry. We encourage you to file bugs in 261bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than 262mailing the developers (or mailing lists) directly -- bugs that are 263not entered into bugzilla tend to get forgotten about or ignored. 264 265To see details of a given bug, visit 266https://bugs.kde.org/show_bug.cgi?id=XXXXXX 267where XXXXXX is the bug number as listed below. 268 269135264 dcbzl instruction missing 270142688 == 250799 271153699 Valgrind should report unaligned reads with movdqa 272180217 == 212335 273190429 Valgrind reports lost of errors in ld.so 274 with x86_64 2.9.90 glibc 275197266 valgrind appears to choke on the xmms instruction 276 "roundsd" on x86_64 277197988 Crash when demangling very large symbol names 278202315 unhandled syscall: 332 (inotify_init1) 279203256 Add page-level profiling to Massif 280205093 dsymutil=yes needs quotes, locking (partial fix) 281205241 Snow Leopard 10.6 support (partial fix) 282206600 Leak checker fails to upgrade indirect blocks when their 283 parent becomes reachable 284210935 port valgrind.h (not valgrind) to win32 so apps run under 285 wine can make client requests 286211410 vex amd64->IR: 0x15 0xFF 0xFF 0x0 0x0 0x89 287 within Linux ip-stack checksum functions 288212335 unhandled instruction bytes: 0xF3 0xF 0xBD 0xC0 289 (lzcnt %eax,%eax) 290213685 Undefined value propagates past dependency breaking instruction 291 (partial fix) 292215914 Valgrind inserts bogus empty environment variable 293217863 == 197988 294219538 adjtimex syscall wrapper wrong in readonly adjtime mode 295222545 shmat fails under valgind on some arm targets 296222560 ARM NEON support 297230407 == 202315 298231076 == 202315 299232509 Docs build fails with formatting inside <title></title> elements 300232793 == 202315 301235642 [PATCH] syswrap-linux.c: support evdev EVIOCG* ioctls 302236546 vex x86->IR: 0x66 0xF 0x3A 0xA 303237202 vex amd64->IR: 0xF3 0xF 0xB8 0xC0 0x49 0x3B 304237371 better support for VALGRIND_MALLOCLIKE_BLOCK 305237485 symlink (syscall 57) is not supported on Mac OS 306237723 sysno == 101 exp-ptrcheck: the 'impossible' happened: 307 unhandled syscall 308238208 is_just_below_ESP doesn't take into account red-zone 309238345 valgrind passes wrong $0 when executing a shell script 310238679 mq_timedreceive syscall doesn't flag the reception buffer 311 as "defined" 312238696 fcntl command F_DUPFD_CLOEXEC not supported 313238713 unhandled instruction bytes: 0x66 0xF 0x29 0xC6 314238713 unhandled instruction bytes: 0x66 0xF 0x29 0xC6 315238745 3.5.0 Make fails on PPC Altivec opcodes, though configure 316 says "Altivec off" 317239992 vex amd64->IR: 0x48 0xF 0xC4 0xC1 0x0 0x48 318240488 == 197988 319240639 == 212335 320241377 == 236546 321241903 == 202315 322241920 == 212335 323242606 unhandled syscall: setegid (in Ptrcheck) 324242814 Helgrind "Impossible has happened" during 325 QApplication::initInstance(); 326243064 Valgrind attempting to read debug information from iso 327243270 Make stack unwinding in Valgrind wrappers more reliable 328243884 exp-ptrcheck: the 'impossible happened: unhandled syscall 329 sysno = 277 (mq_open) 330244009 exp-ptrcheck unknown syscalls in analyzing lighttpd 331244493 ARM VFP d16-d31 registers support 332244670 add support for audit_session_self syscall on Mac OS 10.6 333244921 The xml report of helgrind tool is not well format 334244923 In the xml report file, the <preamble> not escape the 335 xml char, eg '<','&','>' 336245535 print full path names in plain text reports 337245925 x86-64 red zone handling problem 338246258 Valgrind not catching integer underruns + new [] s 339246311 reg/reg cmpxchg doesn't work on amd64 340246549 unhandled syscall unix:277 while testing 32-bit Darwin app 341246888 Improve Makefile.vex.am 342247510 [OS X 10.6] Memcheck reports unaddressable bytes passed 343 to [f]chmod_extended 344247526 IBM POWER6 (ISA 2.05) support is incomplete 345247561 Some leak testcases fails due to reachable addresses in 346 caller save regs 347247875 sizeofIRType to handle Ity_I128 348247894 [PATCH] unhandled syscall sys_readahead 349247980 Doesn't honor CFLAGS passed to configure 350248373 darwin10.supp is empty in the trunk 351248822 Linux FIBMAP ioctl has int parameter instead of long 352248893 [PATCH] make readdwarf.c big endianess safe to enable 353 unwinding on big endian systems 354249224 Syscall 336 not supported (SYS_proc_info) 355249359 == 245535 356249775 Incorrect scheme for detecting NEON capabilities of host CPU 357249943 jni JVM init fails when using valgrind 358249991 Valgrind incorrectly declares AESKEYGENASSIST support 359 since VEX r2011 360249996 linux/arm: unhandled syscall: 181 (__NR_pwrite64) 361250799 frexp$fenv_access_off function generates SIGILL 362250998 vex x86->IR: unhandled instruction bytes: 0x66 0x66 0x66 0x2E 363251251 support pclmulqdq insn 364251362 valgrind: ARM: attach to debugger either fails or provokes 365 kernel oops 366251674 Unhandled syscall 294 367251818 == 254550 368 369254257 Add support for debugfiles found by build-id 370254550 [PATCH] Implement DW_ATE_UTF (DWARF4) 371254646 Wrapped functions cause stack misalignment on OS X 372 (and possibly Linux) 373254556 ARM: valgrinding anything fails with SIGSEGV for 0xFFFF0FA0 374 375(3.6.0: 21 October 2010, vex r2068, valgrind r11471). 376 377 378 379Release 3.5.0 (19 August 2009) 380~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 3813.5.0 is a feature release with many significant improvements and the 382usual collection of bug fixes. The main improvement is that Valgrind 383now works on Mac OS X. 384 385This release supports X86/Linux, AMD64/Linux, PPC32/Linux, PPC64/Linux 386and X86/Darwin. Support for recent distros and toolchain components 387(glibc 2.10, gcc 4.5) has been added. 388 389 ------------------------- 390 391Here is a short summary of the changes. Details are shown further 392down: 393 394* Support for Mac OS X (10.5.x). 395 396* Improvements and simplifications to Memcheck's leak checker. 397 398* Clarification and simplifications in various aspects of Valgrind's 399 text output. 400 401* XML output for Helgrind and Ptrcheck. 402 403* Performance and stability improvements for Helgrind and DRD. 404 405* Genuinely atomic support for x86/amd64/ppc atomic instructions. 406 407* A new experimental tool, BBV, useful for computer architecture 408 research. 409 410* Improved Wine support, including ability to read Windows PDB 411 debuginfo. 412 413 ------------------------- 414 415Here are details of the above changes, followed by descriptions of 416many other minor changes, and a list of fixed bugs. 417 418 419* Valgrind now runs on Mac OS X. (Note that Mac OS X is sometimes 420 called "Darwin" because that is the name of the OS core, which is the 421 level that Valgrind works at.) 422 423 Supported systems: 424 425 - It requires OS 10.5.x (Leopard). Porting to 10.4.x is not planned 426 because it would require work and 10.4 is only becoming less common. 427 428 - 32-bit programs on x86 and AMD64 (a.k.a x86-64) machines are supported 429 fairly well. For 10.5.x, 32-bit programs are the default even on 430 64-bit machines, so it handles most current programs. 431 432 - 64-bit programs on x86 and AMD64 (a.k.a x86-64) machines are not 433 officially supported, but simple programs at least will probably work. 434 However, start-up is slow. 435 436 - PowerPC machines are not supported. 437 438 Things that don't work: 439 440 - The Ptrcheck tool. 441 442 - Objective-C garbage collection. 443 444 - --db-attach=yes. 445 446 - If you have Rogue Amoeba's "Instant Hijack" program installed, 447 Valgrind will fail with a SIGTRAP at start-up. See 448 https://bugs.kde.org/show_bug.cgi?id=193917 for details and a 449 simple work-around. 450 451 Usage notes: 452 453 - You will likely find --dsymutil=yes a useful option, as error 454 messages may be imprecise without it. 455 456 - Mac OS X support is new and therefore will be less robust than the 457 Linux support. Please report any bugs you find. 458 459 - Threaded programs may run more slowly than on Linux. 460 461 Many thanks to Greg Parker for developing this port over several years. 462 463 464* Memcheck's leak checker has been improved. 465 466 - The results for --leak-check=summary now match the summary results 467 for --leak-check=full. Previously they could differ because 468 --leak-check=summary counted "indirectly lost" blocks and 469 "suppressed" blocks as "definitely lost". 470 471 - Blocks that are only reachable via at least one interior-pointer, 472 but are directly pointed to by a start-pointer, were previously 473 marked as "still reachable". They are now correctly marked as 474 "possibly lost". 475 476 - The default value for the --leak-resolution option has been 477 changed from "low" to "high". In general, this means that more 478 leak reports will be produced, but each leak report will describe 479 fewer leaked blocks. 480 481 - With --leak-check=full, "definitely lost" and "possibly lost" 482 leaks are now considered as proper errors, ie. they are counted 483 for the "ERROR SUMMARY" and affect the behaviour of 484 --error-exitcode. These leaks are not counted as errors if 485 --leak-check=summary is specified, however. 486 487 - Documentation for the leak checker has been improved. 488 489 490* Various aspects of Valgrind's text output have changed. 491 492 - Valgrind's start-up message has changed. It is shorter but also 493 includes the command being run, which makes it easier to use 494 --trace-children=yes. An example: 495 496 - Valgrind's shut-down messages have also changed. This is most 497 noticeable with Memcheck, where the leak summary now occurs before 498 the error summary. This change was necessary to allow leaks to be 499 counted as proper errors (see the description of the leak checker 500 changes above for more details). This was also necessary to fix a 501 longstanding bug in which uses of suppressions against leaks were 502 not "counted", leading to difficulties in maintaining suppression 503 files (see https://bugs.kde.org/show_bug.cgi?id=186790). 504 505 - Behavior of -v has changed. In previous versions, -v printed out 506 a mixture of marginally-user-useful information, and tool/core 507 statistics. The statistics printing has now been moved to its own 508 flag, --stats=yes. This means -v is less verbose and more likely 509 to convey useful end-user information. 510 511 - The format of some (non-XML) stack trace entries has changed a 512 little. Previously there were six possible forms: 513 514 0x80483BF: really (a.c:20) 515 0x80483BF: really (in /foo/a.out) 516 0x80483BF: really 517 0x80483BF: (within /foo/a.out) 518 0x80483BF: ??? (a.c:20) 519 0x80483BF: ??? 520 521 The third and fourth of these forms have been made more consistent 522 with the others. The six possible forms are now: 523 524 0x80483BF: really (a.c:20) 525 0x80483BF: really (in /foo/a.out) 526 0x80483BF: really (in ???) 527 0x80483BF: ??? (in /foo/a.out) 528 0x80483BF: ??? (a.c:20) 529 0x80483BF: ??? 530 531 Stack traces produced when --xml=yes is specified are different 532 and unchanged. 533 534 535* Helgrind and Ptrcheck now support XML output, so they can be used 536 from GUI tools. Also, the XML output mechanism has been 537 overhauled. 538 539 - The XML format has been overhauled and generalised, so it is more 540 suitable for error reporting tools in general. The Memcheck 541 specific aspects of it have been removed. The new format, which 542 is an evolution of the old format, is described in 543 docs/internals/xml-output-protocol4.txt. 544 545 - Memcheck has been updated to use the new format. 546 547 - Helgrind and Ptrcheck are now able to emit output in this format. 548 549 - The XML output mechanism has been overhauled. XML is now output 550 to its own file descriptor, which means that: 551 552 * Valgrind can output text and XML independently. 553 554 * The longstanding problem of XML output being corrupted by 555 unexpected un-tagged text messages is solved. 556 557 As before, the destination for text output is specified using 558 --log-file=, --log-fd= or --log-socket=. 559 560 As before, XML output for a tool is enabled using --xml=yes. 561 562 Because there's a new XML output channel, the XML output 563 destination is now specified by --xml-file=, --xml-fd= or 564 --xml-socket=. 565 566 Initial feedback has shown this causes some confusion. To 567 clarify, the two envisaged usage scenarios are: 568 569 (1) Normal text output. In this case, do not specify --xml=yes 570 nor any of --xml-file=, --xml-fd= or --xml-socket=. 571 572 (2) XML output. In this case, specify --xml=yes, and one of 573 --xml-file=, --xml-fd= or --xml-socket= to select the XML 574 destination, one of --log-file=, --log-fd= or --log-socket= 575 to select the destination for any remaining text messages, 576 and, importantly, -q. 577 578 -q makes Valgrind completely silent on the text channel, 579 except in the case of critical failures, such as Valgrind 580 itself segfaulting, or failing to read debugging information. 581 Hence, in this scenario, it suffices to check whether or not 582 any output appeared on the text channel. If yes, then it is 583 likely to be a critical error which should be brought to the 584 attention of the user. If no (the text channel produced no 585 output) then it can be assumed that the run was successful. 586 587 This allows GUIs to make the critical distinction they need to 588 make (did the run fail or not?) without having to search or 589 filter the text output channel in any way. 590 591 It is also recommended to use --child-silent-after-fork=yes in 592 scenario (2). 593 594 595* Improvements and changes in Helgrind: 596 597 - XML output, as described above 598 599 - Checks for consistent association between pthread condition 600 variables and their associated mutexes are now performed. 601 602 - pthread_spinlock functions are supported. 603 604 - Modest performance improvements. 605 606 - Initial (skeletal) support for describing the behaviour of 607 non-POSIX synchronisation objects through ThreadSanitizer 608 compatible ANNOTATE_* macros. 609 610 - More controllable tradeoffs between performance and the level of 611 detail of "previous" accesses in a race. There are now three 612 settings: 613 614 * --history-level=full. This is the default, and was also the 615 default in 3.4.x. It shows both stacks involved in a race, but 616 requires a lot of memory and can be very slow in programs that 617 do many inter-thread synchronisation events. 618 619 * --history-level=none. This only shows the later stack involved 620 in a race. This can be much faster than --history-level=full, 621 but makes it much more difficult to find the other access 622 involved in the race. 623 624 The new intermediate setting is 625 626 * --history-level=approx 627 628 For the earlier (other) access, two stacks are presented. The 629 earlier access is guaranteed to be somewhere in between the two 630 program points denoted by those stacks. This is not as useful 631 as showing the exact stack for the previous access (as per 632 --history-level=full), but it is better than nothing, and it's 633 almost as fast as --history-level=none. 634 635 636* New features and improvements in DRD: 637 638 - The error messages printed by DRD are now easier to interpret. 639 Instead of using two different numbers to identify each thread 640 (Valgrind thread ID and DRD thread ID), DRD does now identify 641 threads via a single number (the DRD thread ID). Furthermore 642 "first observed at" information is now printed for all error 643 messages related to synchronization objects. 644 645 - Added support for named semaphores (sem_open() and sem_close()). 646 647 - Race conditions between pthread_barrier_wait() and 648 pthread_barrier_destroy() calls are now reported. 649 650 - Added support for custom allocators through the macros 651 VALGRIND_MALLOCLIKE_BLOCK() VALGRIND_FREELIKE_BLOCK() (defined in 652 in <valgrind/valgrind.h>). An alternative for these two macros is 653 the new client request VG_USERREQ__DRD_CLEAN_MEMORY (defined in 654 <valgrind/drd.h>). 655 656 - Added support for annotating non-POSIX synchronization objects 657 through several new ANNOTATE_*() macros. 658 659 - OpenMP: added support for the OpenMP runtime (libgomp) included 660 with gcc versions 4.3.0 and 4.4.0. 661 662 - Faster operation. 663 664 - Added two new command-line options (--first-race-only and 665 --segment-merging-interval). 666 667 668* Genuinely atomic support for x86/amd64/ppc atomic instructions 669 670 Valgrind will now preserve (memory-access) atomicity of LOCK- 671 prefixed x86/amd64 instructions, and any others implying a global 672 bus lock. Ditto for PowerPC l{w,d}arx/st{w,d}cx. instructions. 673 674 This means that Valgrinded processes will "play nicely" in 675 situations where communication with other processes, or the kernel, 676 is done through shared memory and coordinated with such atomic 677 instructions. Prior to this change, such arrangements usually 678 resulted in hangs, races or other synchronisation failures, because 679 Valgrind did not honour atomicity of such instructions. 680 681 682* A new experimental tool, BBV, has been added. BBV generates basic 683 block vectors for use with the SimPoint analysis tool, which allows 684 a program's overall behaviour to be approximated by running only a 685 fraction of it. This is useful for computer architecture 686 researchers. You can run BBV by specifying --tool=exp-bbv (the 687 "exp-" prefix is short for "experimental"). BBV was written by 688 Vince Weaver. 689 690 691* Modestly improved support for running Windows applications under 692 Wine. In particular, initial support for reading Windows .PDB debug 693 information has been added. 694 695 696* A new Memcheck client request VALGRIND_COUNT_LEAK_BLOCKS has been 697 added. It is similar to VALGRIND_COUNT_LEAKS but counts blocks 698 instead of bytes. 699 700 701* The Valgrind client requests VALGRIND_PRINTF and 702 VALGRIND_PRINTF_BACKTRACE have been changed slightly. Previously, 703 the string was always printed immediately on its own line. Now, the 704 string will be added to a buffer but not printed until a newline is 705 encountered, or other Valgrind output is printed (note that for 706 VALGRIND_PRINTF_BACKTRACE, the back-trace itself is considered 707 "other Valgrind output"). This allows you to use multiple 708 VALGRIND_PRINTF calls to build up a single output line, and also to 709 print multiple output lines with a single request (by embedding 710 multiple newlines in the string). 711 712 713* The graphs drawn by Massif's ms_print program have changed slightly: 714 715 - The half-height chars '.' and ',' are no longer drawn, because 716 they are confusing. The --y option can be used if the default 717 y-resolution is not high enough. 718 719 - Horizontal lines are now drawn after the top of a snapshot if 720 there is a gap until the next snapshot. This makes it clear that 721 the memory usage has not dropped to zero between snapshots. 722 723 724* Something that happened in 3.4.0, but wasn't clearly announced: the 725 option --read-var-info=yes can be used by some tools (Memcheck, 726 Helgrind and DRD). When enabled, it causes Valgrind to read DWARF3 727 variable type and location information. This makes those tools 728 start up more slowly and increases memory consumption, but 729 descriptions of data addresses in error messages become more 730 detailed. 731 732 733* exp-Omega, an experimental instantaneous leak-detecting tool, was 734 disabled in 3.4.0 due to a lack of interest and maintenance, 735 although the source code was still in the distribution. The source 736 code has now been removed from the distribution. For anyone 737 interested, the removal occurred in SVN revision r10247. 738 739 740* Some changes have been made to the build system. 741 742 - VEX/ is now integrated properly into the build system. This means 743 that dependency tracking within VEX/ now works properly, "make 744 install" will work without requiring "make" before it, and 745 parallel builds (ie. 'make -j') now work (previously a 746 .NOTPARALLEL directive was used to serialize builds, ie. 'make -j' 747 was effectively ignored). 748 749 - The --with-vex configure option has been removed. It was of 750 little use and removing it simplified the build system. 751 752 - The location of some install files has changed. This should not 753 affect most users. Those who might be affected: 754 755 * For people who use Valgrind with MPI programs, the installed 756 libmpiwrap.so library has moved from 757 $(INSTALL)/<platform>/libmpiwrap.so to 758 $(INSTALL)/libmpiwrap-<platform>.so. 759 760 * For people who distribute standalone Valgrind tools, the 761 installed libraries such as $(INSTALL)/<platform>/libcoregrind.a 762 have moved to $(INSTALL)/libcoregrind-<platform>.a. 763 764 These changes simplify the build system. 765 766 - Previously, all the distributed suppression (*.supp) files were 767 installed. Now, only default.supp is installed. This should not 768 affect users as the other installed suppression files were not 769 read; the fact that they were installed was a mistake. 770 771 772* KNOWN LIMITATIONS: 773 774 - Memcheck is unusable with the Intel compiler suite version 11.1, 775 when it generates code for SSE2-and-above capable targets. This 776 is because of icc's use of highly optimised inlined strlen 777 implementations. It causes Memcheck to report huge numbers of 778 false errors even in simple programs. Helgrind and DRD may also 779 have problems. 780 781 Versions 11.0 and earlier may be OK, but this has not been 782 properly tested. 783 784 785The following bugs have been fixed or resolved. Note that "n-i-bz" 786stands for "not in bugzilla" -- that is, a bug that was reported to us 787but never got a bugzilla entry. We encourage you to file bugs in 788bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than 789mailing the developers (or mailing lists) directly -- bugs that are 790not entered into bugzilla tend to get forgotten about or ignored. 791 792To see details of a given bug, visit 793https://bugs.kde.org/show_bug.cgi?id=XXXXXX 794where XXXXXX is the bug number as listed below. 795 79684303 How about a LockCheck tool? 79791633 dereference of null ptr in vgPlain_st_basetype 79897452 Valgrind doesn't report any pthreads problems 799100628 leak-check gets assertion failure when using 800 VALGRIND_MALLOCLIKE_BLOCK on malloc()ed memory 801108528 NPTL pthread cleanup handlers not called 802110126 Valgrind 2.4.1 configure.in tramples CFLAGS 803110128 mallinfo is not implemented... 804110770 VEX: Generated files not always updated when making valgrind 805111102 Memcheck: problems with large (memory footprint) applications 806115673 Vex's decoder should never assert 807117564 False positive: Syscall param clone(child_tidptr) contains 808 uninitialised byte(s) 809119404 executing ssh from inside valgrind fails 810133679 Callgrind does not write path names to sources with dwarf debug 811 info 812135847 configure.in problem with non gnu compilers (and possible fix) 813136154 threads.c:273 (vgCallgrind_post_signal): Assertion 814 '*(vgCallgrind_current_fn_stack.top) == 0' failed. 815136230 memcheck reports "possibly lost", should be "still reachable" 816137073 NULL arg to MALLOCLIKE_BLOCK causes crash 817137904 Valgrind reports a memory leak when using POSIX threads, 818 while it shouldn't 819139076 valgrind VT_GETSTATE error 820142228 complaint of elf_dynamic_do_rela in trivial usage 821145347 spurious warning with USBDEVFS_REAPURB 822148441 (wine) can't find memory leak in Wine, win32 binary 823 executable file. 824148742 Leak-check fails assert on exit 825149878 add (proper) check for calloc integer overflow 826150606 Call graph is broken when using callgrind control 827152393 leak errors produce an exit code of 0. I need some way to 828 cause leak errors to result in a nonzero exit code. 829157154 documentation (leak-resolution doc speaks about num-callers 830 def=4) + what is a loss record 831159501 incorrect handling of ALSA ioctls 832162020 Valgrinding an empty/zero-byte file crashes valgrind 833162482 ppc: Valgrind crashes while reading stabs information 834162718 x86: avoid segment selector 0 in sys_set_thread_area() 835163253 (wine) canonicaliseSymtab forgot some fields in DiSym 836163560 VEX/test_main.c is missing from valgrind-3.3.1 837164353 malloc_usable_size() doesn't return a usable size 838165468 Inconsistent formatting in memcheck manual -- please fix 839169505 main.c:286 (endOfInstr): 840 Assertion 'ii->cost_offset == *cost_offset' failed 841177206 Generate default.supp during compile instead of configure 842177209 Configure valt_load_address based on arch+os 843177305 eventfd / syscall 323 patch lost 844179731 Tests fail to build because of inlining of non-local asm labels 845181394 helgrind: libhb_core.c:3762 (msm_write): Assertion 846 'ordxx == POrd_EQ || ordxx == POrd_LT' failed. 847181594 Bogus warning for empty text segment 848181707 dwarf doesn't require enumerations to have name 849185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64 850185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree): 851 Assertion '!already_present' failed. 852185359 exp-ptrcheck: unhandled syscall getresuid() 853185794 "WARNING: unhandled syscall: 285" (fallocate) on x86_64 854185816 Valgrind is unable to handle debug info for files with split 855 debug info that are prelinked afterwards 856185980 [darwin] unhandled syscall: sem_open 857186238 bbToIR_AMD64: disInstr miscalculated next %rip 858186507 exp-ptrcheck unhandled syscalls prctl, etc. 859186790 Suppression pattern used for leaks are not reported 860186796 Symbols with length>200 in suppression files are ignored 861187048 drd: mutex PTHREAD_PROCESS_SHARED attribute missinterpretation 862187416 exp-ptrcheck: support for __NR_{setregid,setreuid,setresuid} 863188038 helgrind: hg_main.c:926: mk_SHVAL_fail: the 'impossible' happened 864188046 bashisms in the configure script 865188127 amd64->IR: unhandled instruction bytes: 0xF0 0xF 0xB0 0xA 866188161 memcheck: --track-origins=yes asserts "mc_machine.c:672 867 (get_otrack_shadow_offset_wrk): the 'impossible' happened." 868188248 helgrind: pthread_cleanup_push, pthread_rwlock_unlock, 869 assertion fail "!lock->heldBy" 870188427 Add support for epoll_create1 (with patch) 871188530 Support for SIOCGSTAMPNS 872188560 Include valgrind.spec in the tarball 873188572 Valgrind on Mac should suppress setenv() mem leak 874189054 Valgrind fails to build because of duplicate non-local asm labels 875189737 vex amd64->IR: unhandled instruction bytes: 0xAC 876189762 epoll_create syscall not handled (--tool=exp-ptrcheck) 877189763 drd assertion failure: s_threadinfo[tid].is_recording 878190219 unhandled syscall: 328 (x86-linux) 879190391 dup of 181394; see above 880190429 Valgrind reports lots of errors in ld.so with x86_64 2.9.90 glibc 881190820 No debug information on powerpc-linux 882191095 PATCH: Improve usbdevfs ioctl handling 883191182 memcheck: VALGRIND_LEAK_CHECK quadratic when big nr of chunks 884 or big nr of errors 885191189 --xml=yes should obey --gen-suppressions=all 886191192 syslog() needs a suppression on macosx 887191271 DARWIN: WARNING: unhandled syscall: 33554697 a.k.a.: 265 888191761 getrlimit on MacOSX 889191992 multiple --fn-skip only works sometimes; dependent on order 890192634 V. reports "aspacem sync_check_mapping_callback: 891 segment mismatch" on Darwin 892192954 __extension__ missing on 2 client requests 893194429 Crash at start-up with glibc-2.10.1 and linux-2.6.29 894194474 "INSTALL" file has different build instructions than "README" 895194671 Unhandled syscall (sem_wait?) from mac valgrind 896195069 memcheck: reports leak (memory still reachable) for 897 printf("%d', x) 898195169 drd: (vgDrd_barrier_post_wait): 899 Assertion 'r->sg[p->post_iteration]' failed. 900195268 valgrind --log-file doesn't accept ~/... 901195838 VEX abort: LibVEX_N_SPILL_BYTES too small for CPUID boilerplate 902195860 WARNING: unhandled syscall: unix:223 903196528 need a error suppression for pthread_rwlock_init under os x? 904197227 Support aio_* syscalls on Darwin 905197456 valgrind should reject --suppressions=(directory) 906197512 DWARF2 CFI reader: unhandled CFI instruction 0:10 907197591 unhandled syscall 27 (mincore) 908197793 Merge DCAS branch to the trunk == 85756, 142103 909197794 Avoid duplicate filenames in Vex 910197898 make check fails on current SVN 911197901 make check fails also under exp-ptrcheck in current SVN 912197929 Make --leak-resolution=high the default 913197930 Reduce spacing between leak reports 914197933 Print command line of client at start-up, and shorten preamble 915197966 unhandled syscall 205 (x86-linux, --tool=exp-ptrcheck) 916198395 add BBV to the distribution as an experimental tool 917198624 Missing syscalls on Darwin: 82, 167, 281, 347 918198649 callgrind_annotate doesn't cumulate counters 919199338 callgrind_annotate sorting/thresholds are broken for all but Ir 920199977 Valgrind complains about an unrecognized instruction in the 921 atomic_incs test program 922200029 valgrind isn't able to read Fedora 12 debuginfo 923200760 darwin unhandled syscall: unix:284 924200827 DRD doesn't work on Mac OS X 925200990 VG_(read_millisecond_timer)() does not work correctly 926201016 Valgrind does not support pthread_kill() on Mac OS 927201169 Document --read-var-info 928201323 Pre-3.5.0 performance sanity checking 929201384 Review user manual for the 3.5.0 release 930201585 mfpvr not implemented on ppc 931201708 tests failing because x86 direction flag is left set 932201757 Valgrind doesn't handle any recent sys_futex additions 933204377 64-bit valgrind can not start a shell script 934 (with #!/path/to/shell) if the shell is a 32-bit executable 935n-i-bz drd: fixed assertion failure triggered by mutex reinitialization. 936n-i-bz drd: fixed a bug that caused incorrect messages to be printed 937 about memory allocation events with memory access tracing enabled 938n-i-bz drd: fixed a memory leak triggered by vector clock deallocation 939 940(3.5.0: 19 Aug 2009, vex r1913, valgrind r10846). 941 942 943 944Release 3.4.1 (28 February 2009) 945~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 9463.4.1 is a bug-fix release that fixes some regressions and assertion 947failures in debug info reading in 3.4.0, most notably incorrect stack 948traces on amd64-linux on older (glibc-2.3 based) systems. Various 949other debug info problems are also fixed. A number of bugs in the 950exp-ptrcheck tool introduced in 3.4.0 have been fixed. 951 952In view of the fact that 3.4.0 contains user-visible regressions 953relative to 3.3.x, upgrading to 3.4.1 is recommended. Packagers are 954encouraged to ship 3.4.1 in preference to 3.4.0. 955 956The fixed bugs are as follows. Note that "n-i-bz" stands for "not in 957bugzilla" -- that is, a bug that was reported to us but never got a 958bugzilla entry. We encourage you to file bugs in bugzilla 959(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the 960developers (or mailing lists) directly -- bugs that are not entered 961into bugzilla tend to get forgotten about or ignored. 962 963n-i-bz Fix various bugs reading icc-11 generated debug info 964n-i-bz Fix various bugs reading gcc-4.4 generated debug info 965n-i-bz Preliminary support for glibc-2.10 / Fedora 11 966n-i-bz Cachegrind and Callgrind: handle non-power-of-two cache sizes, 967 so as to support (eg) 24k Atom D1 and Core2 with 3/6/12MB L2. 968179618 exp-ptrcheck crashed / exit prematurely 969179624 helgrind: false positive races with pthread_create and 970 recv/open/close/read 971134207 pkg-config output contains @VG_PLATFORM@ 972176926 floating point exception at valgrind startup with PPC 440EPX 973181594 Bogus warning for empty text segment 974173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant rex prefixes) 975181707 Dwarf3 doesn't require enumerations to have name 976185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64 977185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree): 978 Assertion '!already_present' failed. 979185359 exp-ptrcheck unhandled syscall getresuid() 980 981(3.4.1.RC1: 24 Feb 2008, vex r1884, valgrind r9253). 982(3.4.1: 28 Feb 2008, vex r1884, valgrind r9293). 983 984 985 986Release 3.4.0 (2 January 2009) 987~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 9883.4.0 is a feature release with many significant improvements and the 989usual collection of bug fixes. This release supports X86/Linux, 990AMD64/Linux, PPC32/Linux and PPC64/Linux. Support for recent distros 991(using gcc 4.4, glibc 2.8 and 2.9) has been added. 992 9933.4.0 brings some significant tool improvements. Memcheck can now 994report the origin of uninitialised values, the thread checkers 995Helgrind and DRD are much improved, and we have a new experimental 996tool, exp-Ptrcheck, which is able to detect overruns of stack and 997global arrays. In detail: 998 999* Memcheck is now able to track the origin of uninitialised values. 1000 When it reports an uninitialised value error, it will try to show 1001 the origin of the value, as either a heap or stack allocation. 1002 Origin tracking is expensive and so is not enabled by default. To 1003 use it, specify --track-origins=yes. Memcheck's speed will be 1004 essentially halved, and memory usage will be significantly 1005 increased. Nevertheless it can drastically reduce the effort 1006 required to identify the root cause of uninitialised value errors, 1007 and so is often a programmer productivity win, despite running more 1008 slowly. 1009 1010* A version (1.4.0) of the Valkyrie GUI, that works with Memcheck in 1011 3.4.0, will be released shortly. 1012 1013* Helgrind's race detection algorithm has been completely redesigned 1014 and reimplemented, to address usability and scalability concerns: 1015 1016 - The new algorithm has a lower false-error rate: it is much less 1017 likely to report races that do not really exist. 1018 1019 - Helgrind will display full call stacks for both accesses involved 1020 in a race. This makes it easier to identify the root causes of 1021 races. 1022 1023 - Limitations on the size of program that can run have been removed. 1024 1025 - Performance has been modestly improved, although that is very 1026 workload-dependent. 1027 1028 - Direct support for Qt4 threading has been added. 1029 1030 - pthread_barriers are now directly supported. 1031 1032 - Helgrind works well on all supported Linux targets. 1033 1034* The DRD thread debugging tool has seen major improvements: 1035 1036 - Greatly improved performance and significantly reduced memory 1037 usage. 1038 1039 - Support for several major threading libraries (Boost.Thread, Qt4, 1040 glib, OpenMP) has been added. 1041 1042 - Support for atomic instructions, POSIX semaphores, barriers and 1043 reader-writer locks has been added. 1044 1045 - Works now on PowerPC CPUs too. 1046 1047 - Added support for printing thread stack usage at thread exit time. 1048 1049 - Added support for debugging lock contention. 1050 1051 - Added a manual for Drd. 1052 1053* A new experimental tool, exp-Ptrcheck, has been added. Ptrcheck 1054 checks for misuses of pointers. In that sense it is a bit like 1055 Memcheck. However, Ptrcheck can do things Memcheck can't: it can 1056 detect overruns of stack and global arrays, it can detect 1057 arbitrarily far out-of-bounds accesses to heap blocks, and it can 1058 detect accesses heap blocks that have been freed a very long time 1059 ago (millions of blocks in the past). 1060 1061 Ptrcheck currently works only on x86-linux and amd64-linux. To use 1062 it, use --tool=exp-ptrcheck. A simple manual is provided, as part 1063 of the main Valgrind documentation. As this is an experimental 1064 tool, we would be particularly interested in hearing about your 1065 experiences with it. 1066 1067* exp-Omega, an experimental instantaneous leak-detecting tool, is no 1068 longer built by default, although the code remains in the repository 1069 and the tarball. This is due to three factors: a perceived lack of 1070 users, a lack of maintenance, and concerns that it may not be 1071 possible to achieve reliable operation using the existing design. 1072 1073* As usual, support for the latest Linux distros and toolchain 1074 components has been added. It should work well on Fedora Core 10, 1075 OpenSUSE 11.1 and Ubuntu 8.10. gcc-4.4 (in its current pre-release 1076 state) is supported, as is glibc-2.9. The C++ demangler has been 1077 updated so as to work well with C++ compiled by even the most recent 1078 g++'s. 1079 1080* You can now use frame-level wildcards in suppressions. This was a 1081 frequently-requested enhancement. A line "..." in a suppression now 1082 matches zero or more frames. This makes it easier to write 1083 suppressions which are precise yet insensitive to changes in 1084 inlining behaviour. 1085 1086* 3.4.0 adds support on x86/amd64 for the SSSE3 instruction set. 1087 1088* Very basic support for IBM Power6 has been added (64-bit processes only). 1089 1090* Valgrind is now cross-compilable. For example, it is possible to 1091 cross compile Valgrind on an x86/amd64-linux host, so that it runs 1092 on a ppc32/64-linux target. 1093 1094* You can set the main thread's stack size at startup using the 1095 new --main-stacksize= flag (subject of course to ulimit settings). 1096 This is useful for running apps that need a lot of stack space. 1097 1098* The limitation that you can't use --trace-children=yes together 1099 with --db-attach=yes has been removed. 1100 1101* The following bugs have been fixed. Note that "n-i-bz" stands for 1102 "not in bugzilla" -- that is, a bug that was reported to us but 1103 never got a bugzilla entry. We encourage you to file bugs in 1104 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than 1105 mailing the developers (or mailing lists) directly. 1106 1107 n-i-bz Make return types for some client requests 64-bit clean 1108 n-i-bz glibc 2.9 support 1109 n-i-bz ignore unsafe .valgrindrc's (CVE-2008-4865) 1110 n-i-bz MPI_Init(0,0) is valid but libmpiwrap.c segfaults 1111 n-i-bz Building in an env without gdb gives bogus gdb attach 1112 92456 Tracing the origin of uninitialised memory 1113 106497 Valgrind does not demangle some C++ template symbols 1114 162222 ==106497 1115 151612 Suppression with "..." (frame-level wildcards in .supp files) 1116 156404 Unable to start oocalc under memcheck on openSUSE 10.3 (64-bit) 1117 159285 unhandled syscall:25 (stime, on x86-linux) 1118 159452 unhandled ioctl 0x8B01 on "valgrind iwconfig" 1119 160954 ppc build of valgrind crashes with illegal instruction (isel) 1120 160956 mallinfo implementation, w/ patch 1121 162092 Valgrind fails to start gnome-system-monitor 1122 162819 malloc_free_fill test doesn't pass on glibc2.8 x86 1123 163794 assertion failure with "--track-origins=yes" 1124 163933 sigcontext.err and .trapno must be set together 1125 163955 remove constraint !(--db-attach=yes && --trace-children=yes) 1126 164476 Missing kernel module loading system calls 1127 164669 SVN regression: mmap() drops posix file locks 1128 166581 Callgrind output corruption when program forks 1129 167288 Patch file for missing system calls on Cell BE 1130 168943 unsupported scas instruction pentium 1131 171645 Unrecognised instruction (MOVSD, non-binutils encoding) 1132 172417 x86->IR: 0x82 ... 1133 172563 amd64->IR: 0xD9 0xF5 - fprem1 1134 173099 .lds linker script generation error 1135 173177 [x86_64] syscalls: 125/126/179 (capget/capset/quotactl) 1136 173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant prefixes) 1137 174532 == 173751 1138 174908 --log-file value not expanded correctly for core file 1139 175044 Add lookup_dcookie for amd64 1140 175150 x86->IR: 0xF2 0xF 0x11 0xC1 (movss non-binutils encoding) 1141 1142Developer-visible changes: 1143 1144* Valgrind's debug-info reading machinery has been majorly overhauled. 1145 It can now correctly establish the addresses for ELF data symbols, 1146 which is something that has never worked properly before now. 1147 1148 Also, Valgrind can now read DWARF3 type and location information for 1149 stack and global variables. This makes it possible to use the 1150 framework to build tools that rely on knowing the type and locations 1151 of stack and global variables, for example exp-Ptrcheck. 1152 1153 Reading of such information is disabled by default, because most 1154 tools don't need it, and because it is expensive in space and time. 1155 However, you can force Valgrind to read it, using the 1156 --read-var-info=yes flag. Memcheck, Helgrind and DRD are able to 1157 make use of such information, if present, to provide source-level 1158 descriptions of data addresses in the error messages they create. 1159 1160(3.4.0.RC1: 24 Dec 2008, vex r1878, valgrind r8882). 1161(3.4.0: 3 Jan 2009, vex r1878, valgrind r8899). 1162 1163 1164 1165Release 3.3.1 (4 June 2008) 1166~~~~~~~~~~~~~~~~~~~~~~~~~~~ 11673.3.1 fixes a bunch of bugs in 3.3.0, adds support for glibc-2.8 based 1168systems (openSUSE 11, Fedora Core 9), improves the existing glibc-2.7 1169support, and adds support for the SSSE3 (Core 2) instruction set. 1170 11713.3.1 will likely be the last release that supports some very old 1172systems. In particular, the next major release, 3.4.0, will drop 1173support for the old LinuxThreads threading library, and for gcc 1174versions prior to 3.0. 1175 1176The fixed bugs are as follows. Note that "n-i-bz" stands for "not in 1177bugzilla" -- that is, a bug that was reported to us but never got a 1178bugzilla entry. We encourage you to file bugs in bugzilla 1179(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the 1180developers (or mailing lists) directly -- bugs that are not entered 1181into bugzilla tend to get forgotten about or ignored. 1182 1183n-i-bz Massif segfaults at exit 1184n-i-bz Memcheck asserts on Altivec code 1185n-i-bz fix sizeof bug in Helgrind 1186n-i-bz check fd on sys_llseek 1187n-i-bz update syscall lists to kernel 2.6.23.1 1188n-i-bz support sys_sync_file_range 1189n-i-bz handle sys_sysinfo, sys_getresuid, sys_getresgid on ppc64-linux 1190n-i-bz intercept memcpy in 64-bit ld.so's 1191n-i-bz Fix wrappers for sys_{futimesat,utimensat} 1192n-i-bz Minor false-error avoidance fixes for Memcheck 1193n-i-bz libmpiwrap.c: add a wrapper for MPI_Waitany 1194n-i-bz helgrind support for glibc-2.8 1195n-i-bz partial fix for mc_leakcheck.c:698 assert: 1196 'lc_shadows[i]->data + lc_shadows[i] ... 1197n-i-bz Massif/Cachegrind output corruption when programs fork 1198n-i-bz register allocator fix: handle spill stores correctly 1199n-i-bz add support for PA6T PowerPC CPUs 1200126389 vex x86->IR: 0xF 0xAE (FXRSTOR) 1201158525 ==126389 1202152818 vex x86->IR: 0xF3 0xAC (repz lodsb) 1203153196 vex x86->IR: 0xF2 0xA6 (repnz cmpsb) 1204155011 vex x86->IR: 0xCF (iret) 1205155091 Warning [...] unhandled DW_OP_ opcode 0x23 1206156960 ==155901 1207155528 support Core2/SSSE3 insns on x86/amd64 1208155929 ms_print fails on massif outputs containing long lines 1209157665 valgrind fails on shmdt(0) after shmat to 0 1210157748 support x86 PUSHFW/POPFW 1211158212 helgrind: handle pthread_rwlock_try{rd,wr}lock. 1212158425 sys_poll incorrectly emulated when RES==0 1213158744 vex amd64->IR: 0xF0 0x41 0xF 0xC0 (xaddb) 1214160907 Support for a couple of recent Linux syscalls 1215161285 Patch -- support for eventfd() syscall 1216161378 illegal opcode in debug libm (FUCOMPP) 1217160136 ==161378 1218161487 number of suppressions files is limited to 10 1219162386 ms_print typo in milliseconds time unit for massif 1220161036 exp-drd: client allocated memory was never freed 1221162663 signalfd_wrapper fails on 64bit linux 1222 1223(3.3.1.RC1: 2 June 2008, vex r1854, valgrind r8169). 1224(3.3.1: 4 June 2008, vex r1854, valgrind r8180). 1225 1226 1227 1228Release 3.3.0 (7 December 2007) 1229~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 12303.3.0 is a feature release with many significant improvements and the 1231usual collection of bug fixes. This release supports X86/Linux, 1232AMD64/Linux, PPC32/Linux and PPC64/Linux. Support for recent distros 1233(using gcc 4.3, glibc 2.6 and 2.7) has been added. 1234 1235The main excitement in 3.3.0 is new and improved tools. Helgrind 1236works again, Massif has been completely overhauled and much improved, 1237Cachegrind now does branch-misprediction profiling, and a new category 1238of experimental tools has been created, containing two new tools: 1239Omega and DRD. There are many other smaller improvements. In detail: 1240 1241- Helgrind has been completely overhauled and works for the first time 1242 since Valgrind 2.2.0. Supported functionality is: detection of 1243 misuses of the POSIX PThreads API, detection of potential deadlocks 1244 resulting from cyclic lock dependencies, and detection of data 1245 races. Compared to the 2.2.0 Helgrind, the race detection algorithm 1246 has some significant improvements aimed at reducing the false error 1247 rate. Handling of various kinds of corner cases has been improved. 1248 Efforts have been made to make the error messages easier to 1249 understand. Extensive documentation is provided. 1250 1251- Massif has been completely overhauled. Instead of measuring 1252 space-time usage -- which wasn't always useful and many people found 1253 confusing -- it now measures space usage at various points in the 1254 execution, including the point of peak memory allocation. Its 1255 output format has also changed: instead of producing PostScript 1256 graphs and HTML text, it produces a single text output (via the new 1257 'ms_print' script) that contains both a graph and the old textual 1258 information, but in a more compact and readable form. Finally, the 1259 new version should be more reliable than the old one, as it has been 1260 tested more thoroughly. 1261 1262- Cachegrind has been extended to do branch-misprediction profiling. 1263 Both conditional and indirect branches are profiled. The default 1264 behaviour of Cachegrind is unchanged. To use the new functionality, 1265 give the option --branch-sim=yes. 1266 1267- A new category of "experimental tools" has been created. Such tools 1268 may not work as well as the standard tools, but are included because 1269 some people will find them useful, and because exposure to a wider 1270 user group provides tool authors with more end-user feedback. These 1271 tools have a "exp-" prefix attached to their names to indicate their 1272 experimental nature. Currently there are two experimental tools: 1273 1274 * exp-Omega: an instantaneous leak detector. See 1275 exp-omega/docs/omega_introduction.txt. 1276 1277 * exp-DRD: a data race detector based on the happens-before 1278 relation. See exp-drd/docs/README.txt. 1279 1280- Scalability improvements for very large programs, particularly those 1281 which have a million or more malloc'd blocks in use at once. These 1282 improvements mostly affect Memcheck. Memcheck is also up to 10% 1283 faster for all programs, with x86-linux seeing the largest 1284 improvement. 1285 1286- Works well on the latest Linux distros. Has been tested on Fedora 1287 Core 8 (x86, amd64, ppc32, ppc64) and openSUSE 10.3. glibc 2.6 and 1288 2.7 are supported. gcc-4.3 (in its current pre-release state) is 1289 supported. At the same time, 3.3.0 retains support for older 1290 distros. 1291 1292- The documentation has been modestly reorganised with the aim of 1293 making it easier to find information on common-usage scenarios. 1294 Some advanced material has been moved into a new chapter in the main 1295 manual, so as to unclutter the main flow, and other tidying up has 1296 been done. 1297 1298- There is experimental support for AIX 5.3, both 32-bit and 64-bit 1299 processes. You need to be running a 64-bit kernel to use Valgrind 1300 on a 64-bit executable. 1301 1302- There have been some changes to command line options, which may 1303 affect you: 1304 1305 * --log-file-exactly and 1306 --log-file-qualifier options have been removed. 1307 1308 To make up for this --log-file option has been made more powerful. 1309 It now accepts a %p format specifier, which is replaced with the 1310 process ID, and a %q{FOO} format specifier, which is replaced with 1311 the contents of the environment variable FOO. 1312 1313 * --child-silent-after-fork=yes|no [no] 1314 1315 Causes Valgrind to not show any debugging or logging output for 1316 the child process resulting from a fork() call. This can make the 1317 output less confusing (although more misleading) when dealing with 1318 processes that create children. 1319 1320 * --cachegrind-out-file, --callgrind-out-file and --massif-out-file 1321 1322 These control the names of the output files produced by 1323 Cachegrind, Callgrind and Massif. They accept the same %p and %q 1324 format specifiers that --log-file accepts. --callgrind-out-file 1325 replaces Callgrind's old --base option. 1326 1327 * Cachegrind's 'cg_annotate' script no longer uses the --<pid> 1328 option to specify the output file. Instead, the first non-option 1329 argument is taken to be the name of the output file, and any 1330 subsequent non-option arguments are taken to be the names of 1331 source files to be annotated. 1332 1333 * Cachegrind and Callgrind now use directory names where possible in 1334 their output files. This means that the -I option to 1335 'cg_annotate' and 'callgrind_annotate' should not be needed in 1336 most cases. It also means they can correctly handle the case 1337 where two source files in different directories have the same 1338 name. 1339 1340- Memcheck offers a new suppression kind: "Jump". This is for 1341 suppressing jump-to-invalid-address errors. Previously you had to 1342 use an "Addr1" suppression, which didn't make much sense. 1343 1344- Memcheck has new flags --malloc-fill=<hexnum> and 1345 --free-fill=<hexnum> which free malloc'd / free'd areas with the 1346 specified byte. This can help shake out obscure memory corruption 1347 problems. The definedness and addressability of these areas is 1348 unchanged -- only the contents are affected. 1349 1350- The behaviour of Memcheck's client requests VALGRIND_GET_VBITS and 1351 VALGRIND_SET_VBITS have changed slightly. They no longer issue 1352 addressability errors -- if either array is partially unaddressable, 1353 they just return 3 (as before). Also, SET_VBITS doesn't report 1354 definedness errors if any of the V bits are undefined. 1355 1356- The following Memcheck client requests have been removed: 1357 VALGRIND_MAKE_NOACCESS 1358 VALGRIND_MAKE_WRITABLE 1359 VALGRIND_MAKE_READABLE 1360 VALGRIND_CHECK_WRITABLE 1361 VALGRIND_CHECK_READABLE 1362 VALGRIND_CHECK_DEFINED 1363 They were deprecated in 3.2.0, when equivalent but better-named client 1364 requests were added. See the 3.2.0 release notes for more details. 1365 1366- The behaviour of the tool Lackey has changed slightly. First, the output 1367 from --trace-mem has been made more compact, to reduce the size of the 1368 traces. Second, a new option --trace-superblocks has been added, which 1369 shows the addresses of superblocks (code blocks) as they are executed. 1370 1371- The following bugs have been fixed. Note that "n-i-bz" stands for 1372 "not in bugzilla" -- that is, a bug that was reported to us but 1373 never got a bugzilla entry. We encourage you to file bugs in 1374 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than 1375 mailing the developers (or mailing lists) directly. 1376 1377 n-i-bz x86_linux_REDIR_FOR_index() broken 1378 n-i-bz guest-amd64/toIR.c:2512 (dis_op2_E_G): Assertion `0' failed. 1379 n-i-bz Support x86 INT insn (INT (0xCD) 0x40 - 0x43) 1380 n-i-bz Add sys_utimensat system call for Linux x86 platform 1381 79844 Helgrind complains about race condition which does not exist 1382 82871 Massif output function names too short 1383 89061 Massif: ms_main.c:485 (get_XCon): Assertion `xpt->max_chi...' 1384 92615 Write output from Massif at crash 1385 95483 massif feature request: include peak allocation in report 1386 112163 MASSIF crashed with signal 7 (SIGBUS) after running 2 days 1387 119404 problems running setuid executables (partial fix) 1388 121629 add instruction-counting mode for timing 1389 127371 java vm giving unhandled instruction bytes: 0x26 0x2E 0x64 0x65 1390 129937 ==150380 1391 129576 Massif loses track of memory, incorrect graphs 1392 132132 massif --format=html output does not do html entity escaping 1393 132950 Heap alloc/usage summary 1394 133962 unhandled instruction bytes: 0xF2 0x4C 0xF 0x10 1395 134990 use -fno-stack-protector if possible 1396 136382 ==134990 1397 137396 I would really like helgrind to work again... 1398 137714 x86/amd64->IR: 0x66 0xF 0xF7 0xC6 (maskmovq, maskmovdq) 1399 141631 Massif: percentages don't add up correctly 1400 142706 massif numbers don't seem to add up 1401 143062 massif crashes on app exit with signal 8 SIGFPE 1402 144453 (get_XCon): Assertion 'xpt->max_children != 0' failed. 1403 145559 valgrind aborts when malloc_stats is called 1404 145609 valgrind aborts all runs with 'repeated section!' 1405 145622 --db-attach broken again on x86-64 1406 145837 ==149519 1407 145887 PPC32: getitimer() system call is not supported 1408 146252 ==150678 1409 146456 (update_XCon): Assertion 'xpt->curr_space >= -space_delta'... 1410 146701 ==134990 1411 146781 Adding support for private futexes 1412 147325 valgrind internal error on syscall (SYS_io_destroy, 0) 1413 147498 amd64->IR: 0xF0 0xF 0xB0 0xF (lock cmpxchg %cl,(%rdi)) 1414 147545 Memcheck: mc_main.c:817 (get_sec_vbits8): Assertion 'n' failed. 1415 147628 SALC opcode 0xd6 unimplemented 1416 147825 crash on amd64-linux with gcc 4.2 and glibc 2.6 (CFI) 1417 148174 Incorrect type of freed_list_volume causes assertion [...] 1418 148447 x86_64 : new NOP codes: 66 66 66 66 2e 0f 1f 1419 149182 PPC Trap instructions not implemented in valgrind 1420 149504 Assertion hit on alloc_xpt->curr_space >= -space_delta 1421 149519 ppc32: V aborts with SIGSEGV on execution of a signal handler 1422 149892 ==137714 1423 150044 SEGV during stack deregister 1424 150380 dwarf/gcc interoperation (dwarf3 read problems) 1425 150408 ==148447 1426 150678 guest-amd64/toIR.c:3741 (dis_Grp5): Assertion `sz == 4' failed 1427 151209 V unable to execute programs for users with UID > 2^16 1428 151938 help on --db-command= misleading 1429 152022 subw $0x28, %%sp causes assertion failure in memcheck 1430 152357 inb and outb not recognized in 64-bit mode 1431 152501 vex x86->IR: 0x27 0x66 0x89 0x45 (daa) 1432 152818 vex x86->IR: 0xF3 0xAC 0xFC 0x9C (rep lodsb) 1433 1434Developer-visible changes: 1435 1436- The names of some functions and types within the Vex IR have 1437 changed. Run 'svn log -r1689 VEX/pub/libvex_ir.h' for full details. 1438 Any existing standalone tools will have to be updated to reflect 1439 these changes. The new names should be clearer. The file 1440 VEX/pub/libvex_ir.h is also much better commented. 1441 1442- A number of new debugging command line options have been added. 1443 These are mostly of use for debugging the symbol table and line 1444 number readers: 1445 1446 --trace-symtab-patt=<patt> limit debuginfo tracing to obj name <patt> 1447 --trace-cfi=no|yes show call-frame-info details? [no] 1448 --debug-dump=syms mimic /usr/bin/readelf --syms 1449 --debug-dump=line mimic /usr/bin/readelf --debug-dump=line 1450 --debug-dump=frames mimic /usr/bin/readelf --debug-dump=frames 1451 --sym-offsets=yes|no show syms in form 'name+offset' ? [no] 1452 1453- Internally, the code base has been further factorised and 1454 abstractified, particularly with respect to support for non-Linux 1455 OSs. 1456 1457(3.3.0.RC1: 2 Dec 2007, vex r1803, valgrind r7268). 1458(3.3.0.RC2: 5 Dec 2007, vex r1804, valgrind r7282). 1459(3.3.0.RC3: 9 Dec 2007, vex r1804, valgrind r7288). 1460(3.3.0: 10 Dec 2007, vex r1804, valgrind r7290). 1461 1462 1463 1464Release 3.2.3 (29 Jan 2007) 1465~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1466Unfortunately 3.2.2 introduced a regression which can cause an 1467assertion failure ("vex: the `impossible' happened: eqIRConst") when 1468running obscure pieces of SSE code. 3.2.3 fixes this and adds one 1469more glibc-2.5 intercept. In all other respects it is identical to 14703.2.2. Please do not use (or package) 3.2.2; instead use 3.2.3. 1471 1472n-i-bz vex: the `impossible' happened: eqIRConst 1473n-i-bz Add an intercept for glibc-2.5 __stpcpy_chk 1474 1475(3.2.3: 29 Jan 2007, vex r1732, valgrind r6560). 1476 1477 1478Release 3.2.2 (22 Jan 2007) 1479~~~~~~~~~~~~~~~~~~~~~~~~~~~ 14803.2.2 fixes a bunch of bugs in 3.2.1, adds support for glibc-2.5 based 1481systems (openSUSE 10.2, Fedora Core 6), improves support for icc-9.X 1482compiled code, and brings modest performance improvements in some 1483areas, including amd64 floating point, powerpc support, and startup 1484responsiveness on all targets. 1485 1486The fixed bugs are as follows. Note that "n-i-bz" stands for "not in 1487bugzilla" -- that is, a bug that was reported to us but never got a 1488bugzilla entry. We encourage you to file bugs in bugzilla 1489(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the 1490developers (or mailing lists) directly. 1491 1492129390 ppc?->IR: some kind of VMX prefetch (dstt) 1493129968 amd64->IR: 0xF 0xAE 0x0 (fxsave) 1494134319 ==129968 1495133054 'make install' fails with syntax errors 1496118903 ==133054 1497132998 startup fails in when running on UML 1498134207 pkg-config output contains @VG_PLATFORM@ 1499134727 valgrind exits with "Value too large for defined data type" 1500n-i-bz ppc32/64: support mcrfs 1501n-i-bz Cachegrind/Callgrind: Update cache parameter detection 1502135012 x86->IR: 0xD7 0x8A 0xE0 0xD0 (xlat) 1503125959 ==135012 1504126147 x86->IR: 0xF2 0xA5 0xF 0x77 (repne movsw) 1505136650 amd64->IR: 0xC2 0x8 0x0 1506135421 x86->IR: unhandled Grp5(R) case 6 1507n-i-bz Improved documentation of the IR intermediate representation 1508n-i-bz jcxz (x86) (users list, 8 Nov) 1509n-i-bz ExeContext hashing fix 1510n-i-bz fix CFI reading failures ("Dwarf CFI 0:24 0:32 0:48 0:7") 1511n-i-bz fix Cachegrind/Callgrind simulation bug 1512n-i-bz libmpiwrap.c: fix handling of MPI_LONG_DOUBLE 1513n-i-bz make User errors suppressible 1514136844 corrupted malloc line when using --gen-suppressions=yes 1515138507 ==136844 1516n-i-bz Speed up the JIT's register allocator 1517n-i-bz Fix confusing leak-checker flag hints 1518n-i-bz Support recent autoswamp versions 1519n-i-bz ppc32/64 dispatcher speedups 1520n-i-bz ppc64 front end rld/rlw improvements 1521n-i-bz ppc64 back end imm64 improvements 1522136300 support 64K pages on ppc64-linux 1523139124 == 136300 1524n-i-bz fix ppc insn set tests for gcc >= 4.1 1525137493 x86->IR: recent binutils no-ops 1526137714 x86->IR: 0x66 0xF 0xF7 0xC6 (maskmovdqu) 1527138424 "failed in UME with error 22" (produce a better error msg) 1528138856 ==138424 1529138627 Enhancement support for prctl ioctls 1530138896 Add support for usb ioctls 1531136059 ==138896 1532139050 ppc32->IR: mfspr 268/269 instructions not handled 1533n-i-bz ppc32->IR: lvxl/stvxl 1534n-i-bz glibc-2.5 support 1535n-i-bz memcheck: provide replacement for mempcpy 1536n-i-bz memcheck: replace bcmp in ld.so 1537n-i-bz Use 'ifndef' in VEX's Makefile correctly 1538n-i-bz Suppressions for MVL 4.0.1 on ppc32-linux 1539n-i-bz libmpiwrap.c: Fixes for MPICH 1540n-i-bz More robust handling of hinted client mmaps 1541139776 Invalid read in unaligned memcpy with Intel compiler v9 1542n-i-bz Generate valid XML even for very long fn names 1543n-i-bz Don't prompt about suppressions for unshown reachable leaks 1544139910 amd64 rcl is not supported 1545n-i-bz DWARF CFI reader: handle DW_CFA_undefined 1546n-i-bz DWARF CFI reader: handle icc9 generated CFI info better 1547n-i-bz fix false uninit-value errs in icc9 generated FP code 1548n-i-bz reduce extraneous frames in libmpiwrap.c 1549n-i-bz support pselect6 on amd64-linux 1550 1551(3.2.2: 22 Jan 2007, vex r1729, valgrind r6545). 1552 1553 1554Release 3.2.1 (16 Sept 2006) 1555~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 15563.2.1 adds x86/amd64 support for all SSE3 instructions except monitor 1557and mwait, further reduces memcheck's false error rate on all 1558platforms, adds support for recent binutils (in OpenSUSE 10.2 and 1559Fedora Rawhide) and fixes a bunch of bugs in 3.2.0. Some of the fixed 1560bugs were causing large programs to segfault with --tool=callgrind and 1561--tool=cachegrind, so an upgrade is recommended. 1562 1563In view of the fact that any 3.3.0 release is unlikely to happen until 1564well into 1Q07, we intend to keep the 3.2.X line alive for a while 1565yet, and so we tentatively plan a 3.2.2 release sometime in December 156606. 1567 1568The fixed bugs are as follows. Note that "n-i-bz" stands for "not in 1569bugzilla" -- that is, a bug that was reported to us but never got a 1570bugzilla entry. 1571 1572n-i-bz Expanding brk() into last available page asserts 1573n-i-bz ppc64-linux stack RZ fast-case snafu 1574n-i-bz 'c' in --gen-supps=yes doesn't work 1575n-i-bz VG_N_SEGMENTS too low (users, 28 June) 1576n-i-bz VG_N_SEGNAMES too low (Stu Robinson) 1577106852 x86->IR: fisttp (SSE3) 1578117172 FUTEX_WAKE does not use uaddr2 1579124039 Lacks support for VKI_[GP]IO_UNIMAP* 1580127521 amd64->IR: 0xF0 0x48 0xF 0xC7 (cmpxchg8b) 1581128917 amd64->IR: 0x66 0xF 0xF6 0xC4 (psadbw,SSE2) 1582129246 JJ: ppc32/ppc64 syscalls, w/ patch 1583129358 x86->IR: fisttpl (SSE3) 1584129866 cachegrind/callgrind causes executable to die 1585130020 Can't stat .so/.exe error while reading symbols 1586130388 Valgrind aborts when process calls malloc_trim() 1587130638 PATCH: ppc32 missing system calls 1588130785 amd64->IR: unhandled instruction "pushfq" 1589131481: (HINT_NOP) vex x86->IR: 0xF 0x1F 0x0 0xF 1590131298 ==131481 1591132146 Programs with long sequences of bswap[l,q]s 1592132918 vex amd64->IR: 0xD9 0xF8 (fprem) 1593132813 Assertion at priv/guest-x86/toIR.c:652 fails 1594133051 'cfsi->len > 0 && cfsi->len < 2000000' failed 1595132722 valgrind header files are not standard C 1596n-i-bz Livelocks entire machine (users list, Timothy Terriberry) 1597n-i-bz Alex Bennee mmap problem (9 Aug) 1598n-i-bz BartV: Don't print more lines of a stack-trace than were obtained. 1599n-i-bz ppc32 SuSE 10.1 redir 1600n-i-bz amd64 padding suppressions 1601n-i-bz amd64 insn printing fix. 1602n-i-bz ppc cmp reg,reg fix 1603n-i-bz x86/amd64 iropt e/rflag reduction rules 1604n-i-bz SuSE 10.1 (ppc32) minor fixes 1605133678 amd64->IR: 0x48 0xF 0xC5 0xC0 (pextrw?) 1606133694 aspacem assertion: aspacem_minAddr <= holeStart 1607n-i-bz callgrind: fix warning about malformed creator line 1608n-i-bz callgrind: fix annotate script for data produced with 1609 --dump-instr=yes 1610n-i-bz callgrind: fix failed assertion when toggling 1611 instrumentation mode 1612n-i-bz callgrind: fix annotate script fix warnings with 1613 --collect-jumps=yes 1614n-i-bz docs path hardwired (Dennis Lubert) 1615 1616The following bugs were not fixed, due primarily to lack of developer 1617time, and also because bug reporters did not answer requests for 1618feedback in time for the release: 1619 1620129390 ppc?->IR: some kind of VMX prefetch (dstt) 1621129968 amd64->IR: 0xF 0xAE 0x0 (fxsave) 1622133054 'make install' fails with syntax errors 1623n-i-bz Signal race condition (users list, 13 June, Johannes Berg) 1624n-i-bz Unrecognised instruction at address 0x70198EC2 (users list, 1625 19 July, Bennee) 1626132998 startup fails in when running on UML 1627 1628The following bug was tentatively fixed on the mainline but the fix 1629was considered too risky to push into 3.2.X: 1630 1631133154 crash when using client requests to register/deregister stack 1632 1633(3.2.1: 16 Sept 2006, vex r1658, valgrind r6070). 1634 1635 1636Release 3.2.0 (7 June 2006) 1637~~~~~~~~~~~~~~~~~~~~~~~~~~~ 16383.2.0 is a feature release with many significant improvements and the 1639usual collection of bug fixes. This release supports X86/Linux, 1640AMD64/Linux, PPC32/Linux and PPC64/Linux. 1641 1642Performance, especially of Memcheck, is improved, Addrcheck has been 1643removed, Callgrind has been added, PPC64/Linux support has been added, 1644Lackey has been improved, and MPI support has been added. In detail: 1645 1646- Memcheck has improved speed and reduced memory use. Run times are 1647 typically reduced by 15-30%, averaging about 24% for SPEC CPU2000. 1648 The other tools have smaller but noticeable speed improvments. We 1649 are interested to hear what improvements users get. 1650 1651 Memcheck uses less memory due to the introduction of a compressed 1652 representation for shadow memory. The space overhead has been 1653 reduced by a factor of up to four, depending on program behaviour. 1654 This means you should be able to run programs that use more memory 1655 than before without hitting problems. 1656 1657- Addrcheck has been removed. It has not worked since version 2.4.0, 1658 and the speed and memory improvements to Memcheck make it redundant. 1659 If you liked using Addrcheck because it didn't give undefined value 1660 errors, you can use the new Memcheck option --undef-value-errors=no 1661 to get the same behaviour. 1662 1663- The number of undefined-value errors incorrectly reported by 1664 Memcheck has been reduced (such false reports were already very 1665 rare). In particular, efforts have been made to ensure Memcheck 1666 works really well with gcc 4.0/4.1-generated code on X86/Linux and 1667 AMD64/Linux. 1668 1669- Josef Weidendorfer's popular Callgrind tool has been added. Folding 1670 it in was a logical step given its popularity and usefulness, and 1671 makes it easier for us to ensure it works "out of the box" on all 1672 supported targets. The associated KDE KCachegrind GUI remains a 1673 separate project. 1674 1675- A new release of the Valkyrie GUI for Memcheck, version 1.2.0, 1676 accompanies this release. Improvements over previous releases 1677 include improved robustness, many refinements to the user interface, 1678 and use of a standard autoconf/automake build system. You can get 1679 it from http://www.valgrind.org/downloads/guis.html. 1680 1681- Valgrind now works on PPC64/Linux. As with the AMD64/Linux port, 1682 this supports programs using to 32G of address space. On 64-bit 1683 capable PPC64/Linux setups, you get a dual architecture build so 1684 that both 32-bit and 64-bit executables can be run. Linux on POWER5 1685 is supported, and POWER4 is also believed to work. Both 32-bit and 1686 64-bit DWARF2 is supported. This port is known to work well with 1687 both gcc-compiled and xlc/xlf-compiled code. 1688 1689- Floating point accuracy has been improved for PPC32/Linux. 1690 Specifically, the floating point rounding mode is observed on all FP 1691 arithmetic operations, and multiply-accumulate instructions are 1692 preserved by the compilation pipeline. This means you should get FP 1693 results which are bit-for-bit identical to a native run. These 1694 improvements are also present in the PPC64/Linux port. 1695 1696- Lackey, the example tool, has been improved: 1697 1698 * It has a new option --detailed-counts (off by default) which 1699 causes it to print out a count of loads, stores and ALU operations 1700 done, and their sizes. 1701 1702 * It has a new option --trace-mem (off by default) which causes it 1703 to print out a trace of all memory accesses performed by a 1704 program. It's a good starting point for building Valgrind tools 1705 that need to track memory accesses. Read the comments at the top 1706 of the file lackey/lk_main.c for details. 1707 1708 * The original instrumentation (counting numbers of instructions, 1709 jumps, etc) is now controlled by a new option --basic-counts. It 1710 is on by default. 1711 1712- MPI support: partial support for debugging distributed applications 1713 using the MPI library specification has been added. Valgrind is 1714 aware of the memory state changes caused by a subset of the MPI 1715 functions, and will carefully check data passed to the (P)MPI_ 1716 interface. 1717 1718- A new flag, --error-exitcode=, has been added. This allows changing 1719 the exit code in runs where Valgrind reported errors, which is 1720 useful when using Valgrind as part of an automated test suite. 1721 1722- Various segfaults when reading old-style "stabs" debug information 1723 have been fixed. 1724 1725- A simple performance evaluation suite has been added. See 1726 perf/README and README_DEVELOPERS for details. There are 1727 various bells and whistles. 1728 1729- New configuration flags: 1730 --enable-only32bit 1731 --enable-only64bit 1732 By default, on 64 bit platforms (ppc64-linux, amd64-linux) the build 1733 system will attempt to build a Valgrind which supports both 32-bit 1734 and 64-bit executables. This may not be what you want, and you can 1735 override the default behaviour using these flags. 1736 1737Please note that Helgrind is still not working. We have made an 1738important step towards making it work again, however, with the 1739addition of function wrapping (see below). 1740 1741Other user-visible changes: 1742 1743- Valgrind now has the ability to intercept and wrap arbitrary 1744 functions. This is a preliminary step towards making Helgrind work 1745 again, and was required for MPI support. 1746 1747- There are some changes to Memcheck's client requests. Some of them 1748 have changed names: 1749 1750 MAKE_NOACCESS --> MAKE_MEM_NOACCESS 1751 MAKE_WRITABLE --> MAKE_MEM_UNDEFINED 1752 MAKE_READABLE --> MAKE_MEM_DEFINED 1753 1754 CHECK_WRITABLE --> CHECK_MEM_IS_ADDRESSABLE 1755 CHECK_READABLE --> CHECK_MEM_IS_DEFINED 1756 CHECK_DEFINED --> CHECK_VALUE_IS_DEFINED 1757 1758 The reason for the change is that the old names are subtly 1759 misleading. The old names will still work, but they are deprecated 1760 and may be removed in a future release. 1761 1762 We also added a new client request: 1763 1764 MAKE_MEM_DEFINED_IF_ADDRESSABLE(a, len) 1765 1766 which is like MAKE_MEM_DEFINED but only affects a byte if the byte is 1767 already addressable. 1768 1769- The way client requests are encoded in the instruction stream has 1770 changed. Unfortunately, this means 3.2.0 will not honour client 1771 requests compiled into binaries using headers from earlier versions 1772 of Valgrind. We will try to keep the client request encodings more 1773 stable in future. 1774 1775BUGS FIXED: 1776 1777108258 NPTL pthread cleanup handlers not called 1778117290 valgrind is sigKILL'd on startup 1779117295 == 117290 1780118703 m_signals.c:1427 Assertion 'tst->status == VgTs_WaitSys' 1781118466 add %reg, %reg generates incorrect validity for bit 0 1782123210 New: strlen from ld-linux on amd64 1783123244 DWARF2 CFI reader: unhandled CFI instruction 0:18 1784123248 syscalls in glibc-2.4: openat, fstatat, symlinkat 1785123258 socketcall.recvmsg(msg.msg_iov[i] points to uninit 1786123535 mremap(new_addr) requires MREMAP_FIXED in 4th arg 1787123836 small typo in the doc 1788124029 ppc compile failed: `vor' gcc 3.3.5 1789124222 Segfault: @@don't know what type ':' is 1790124475 ppc32: crash (syscall?) timer_settime() 1791124499 amd64->IR: 0xF 0xE 0x48 0x85 (femms) 1792124528 FATAL: aspacem assertion failed: segment_is_sane 1793124697 vex x86->IR: 0xF 0x70 0xC9 0x0 (pshufw) 1794124892 vex x86->IR: 0xF3 0xAE (REPx SCASB) 1795126216 == 124892 1796124808 ppc32: sys_sched_getaffinity() not handled 1797n-i-bz Very long stabs strings crash m_debuginfo 1798n-i-bz amd64->IR: 0x66 0xF 0xF5 (pmaddwd) 1799125492 ppc32: support a bunch more syscalls 1800121617 ppc32/64: coredumping gives assertion failure 1801121814 Coregrind return error as exitcode patch 1802126517 == 121814 1803125607 amd64->IR: 0x66 0xF 0xA3 0x2 (btw etc) 1804125651 amd64->IR: 0xF8 0x49 0xFF 0xE3 (clc?) 1805126253 x86 movx is wrong 1806126451 3.2 SVN doesn't work on ppc32 CPU's without FPU 1807126217 increase # threads 1808126243 vex x86->IR: popw mem 1809126583 amd64->IR: 0x48 0xF 0xA4 0xC2 (shld $1,%rax,%rdx) 1810126668 amd64->IR: 0x1C 0xFF (sbb $0xff,%al) 1811126696 support for CDROMREADRAW ioctl and CDROMREADTOCENTRY fix 1812126722 assertion: segment_is_sane at m_aspacemgr/aspacemgr.c:1624 1813126938 bad checking for syscalls linkat, renameat, symlinkat 1814 1815(3.2.0RC1: 27 May 2006, vex r1626, valgrind r5947). 1816(3.2.0: 7 June 2006, vex r1628, valgrind r5957). 1817 1818 1819Release 3.1.1 (15 March 2006) 1820~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 18213.1.1 fixes a bunch of bugs reported in 3.1.0. There is no new 1822functionality. The fixed bugs are: 1823 1824(note: "n-i-bz" means "not in bugzilla" -- this bug does not have 1825 a bugzilla entry). 1826 1827n-i-bz ppc32: fsub 3,3,3 in dispatcher doesn't clear NaNs 1828n-i-bz ppc32: __NR_{set,get}priority 1829117332 x86: missing line info with icc 8.1 1830117366 amd64: 0xDD 0x7C fnstsw 1831118274 == 117366 1832117367 amd64: 0xD9 0xF4 fxtract 1833117369 amd64: __NR_getpriority (140) 1834117419 ppc32: lfsu f5, -4(r11) 1835117419 ppc32: fsqrt 1836117936 more stabs problems (segfaults while reading debug info) 1837119914 == 117936 1838120345 == 117936 1839118239 amd64: 0xF 0xAE 0x3F (clflush) 1840118939 vm86old system call 1841n-i-bz memcheck/tests/mempool reads freed memory 1842n-i-bz AshleyP's custom-allocator assertion 1843n-i-bz Dirk strict-aliasing stuff 1844n-i-bz More space for debugger cmd line (Dan Thaler) 1845n-i-bz Clarified leak checker output message 1846n-i-bz AshleyP's --gen-suppressions output fix 1847n-i-bz cg_annotate's --sort option broken 1848n-i-bz OSet 64-bit fastcmp bug 1849n-i-bz VG_(getgroups) fix (Shinichi Noda) 1850n-i-bz ppc32: allocate from callee-saved FP/VMX regs 1851n-i-bz misaligned path word-size bug in mc_main.c 1852119297 Incorrect error message for sse code 1853120410 x86: prefetchw (0xF 0xD 0x48 0x4) 1854120728 TIOCSERGETLSR, TIOCGICOUNT, HDIO_GET_DMA ioctls 1855120658 Build fixes for gcc 2.96 1856120734 x86: Support for changing EIP in signal handler 1857n-i-bz memcheck/tests/zeropage de-looping fix 1858n-i-bz x86: fxtract doesn't work reliably 1859121662 x86: lock xadd (0xF0 0xF 0xC0 0x2) 1860121893 calloc does not always return zeroed memory 1861121901 no support for syscall tkill 1862n-i-bz Suppression update for Debian unstable 1863122067 amd64: fcmovnu (0xDB 0xD9) 1864n-i-bz ppc32: broken signal handling in cpu feature detection 1865n-i-bz ppc32: rounding mode problems (improved, partial fix only) 1866119482 ppc32: mtfsb1 1867n-i-bz ppc32: mtocrf/mfocrf 1868 1869(3.1.1: 15 March 2006, vex r1597, valgrind r5771). 1870 1871 1872Release 3.1.0 (25 November 2005) 1873~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 18743.1.0 is a feature release with a number of significant improvements: 1875AMD64 support is much improved, PPC32 support is good enough to be 1876usable, and the handling of memory management and address space is 1877much more robust. In detail: 1878 1879- AMD64 support is much improved. The 64-bit vs. 32-bit issues in 1880 3.0.X have been resolved, and it should "just work" now in all 1881 cases. On AMD64 machines both 64-bit and 32-bit versions of 1882 Valgrind are built. The right version will be invoked 1883 automatically, even when using --trace-children and mixing execution 1884 between 64-bit and 32-bit executables. Also, many more instructions 1885 are supported. 1886 1887- PPC32 support is now good enough to be usable. It should work with 1888 all tools, but please let us know if you have problems. Three 1889 classes of CPUs are supported: integer only (no FP, no Altivec), 1890 which covers embedded PPC uses, integer and FP but no Altivec 1891 (G3-ish), and CPUs capable of Altivec too (G4, G5). 1892 1893- Valgrind's address space management has been overhauled. As a 1894 result, Valgrind should be much more robust with programs that use 1895 large amounts of memory. There should be many fewer "memory 1896 exhausted" messages, and debug symbols should be read correctly on 1897 large (eg. 300MB+) executables. On 32-bit machines the full address 1898 space available to user programs (usually 3GB or 4GB) can be fully 1899 utilised. On 64-bit machines up to 32GB of space is usable; when 1900 using Memcheck that means your program can use up to about 14GB. 1901 1902 A side effect of this change is that Valgrind is no longer protected 1903 against wild writes by the client. This feature was nice but relied 1904 on the x86 segment registers and so wasn't portable. 1905 1906- Most users should not notice, but as part of the address space 1907 manager change, the way Valgrind is built has been changed. Each 1908 tool is now built as a statically linked stand-alone executable, 1909 rather than as a shared object that is dynamically linked with the 1910 core. The "valgrind" program invokes the appropriate tool depending 1911 on the --tool option. This slightly increases the amount of disk 1912 space used by Valgrind, but it greatly simplified many things and 1913 removed Valgrind's dependence on glibc. 1914 1915Please note that Addrcheck and Helgrind are still not working. Work 1916is underway to reinstate them (or equivalents). We apologise for the 1917inconvenience. 1918 1919Other user-visible changes: 1920 1921- The --weird-hacks option has been renamed --sim-hints. 1922 1923- The --time-stamp option no longer gives an absolute date and time. 1924 It now prints the time elapsed since the program began. 1925 1926- It should build with gcc-2.96. 1927 1928- Valgrind can now run itself (see README_DEVELOPERS for how). 1929 This is not much use to you, but it means the developers can now 1930 profile Valgrind using Cachegrind. As a result a couple of 1931 performance bad cases have been fixed. 1932 1933- The XML output format has changed slightly. See 1934 docs/internals/xml-output.txt. 1935 1936- Core dumping has been reinstated (it was disabled in 3.0.0 and 3.0.1). 1937 If your program crashes while running under Valgrind, a core file with 1938 the name "vgcore.<pid>" will be created (if your settings allow core 1939 file creation). Note that the floating point information is not all 1940 there. If Valgrind itself crashes, the OS will create a normal core 1941 file. 1942 1943The following are some user-visible changes that occurred in earlier 1944versions that may not have been announced, or were announced but not 1945widely noticed. So we're mentioning them now. 1946 1947- The --tool flag is optional once again; if you omit it, Memcheck 1948 is run by default. 1949 1950- The --num-callers flag now has a default value of 12. It was 1951 previously 4. 1952 1953- The --xml=yes flag causes Valgrind's output to be produced in XML 1954 format. This is designed to make it easy for other programs to 1955 consume Valgrind's output. The format is described in the file 1956 docs/internals/xml-format.txt. 1957 1958- The --gen-suppressions flag supports an "all" value that causes every 1959 suppression to be printed without asking. 1960 1961- The --log-file option no longer puts "pid" in the filename, eg. the 1962 old name "foo.pid12345" is now "foo.12345". 1963 1964- There are several graphical front-ends for Valgrind, such as Valkyrie, 1965 Alleyoop and Valgui. See http://www.valgrind.org/downloads/guis.html 1966 for a list. 1967 1968BUGS FIXED: 1969 1970109861 amd64 hangs at startup 1971110301 ditto 1972111554 valgrind crashes with Cannot allocate memory 1973111809 Memcheck tool doesn't start java 1974111901 cross-platform run of cachegrind fails on opteron 1975113468 (vgPlain_mprotect_range): Assertion 'r != -1' failed. 1976 92071 Reading debugging info uses too much memory 1977109744 memcheck loses track of mmap from direct ld-linux.so.2 1978110183 tail of page with _end 1979 82301 FV memory layout too rigid 1980 98278 Infinite recursion possible when allocating memory 1981108994 Valgrind runs out of memory due to 133x overhead 1982115643 valgrind cannot allocate memory 1983105974 vg_hashtable.c static hash table 1984109323 ppc32: dispatch.S uses Altivec insn, which doesn't work on POWER. 1985109345 ptrace_setregs not yet implemented for ppc 1986110831 Would like to be able to run against both 32 and 64 bit 1987 binaries on AMD64 1988110829 == 110831 1989111781 compile of valgrind-3.0.0 fails on my linux (gcc 2.X prob) 1990112670 Cachegrind: cg_main.c:486 (handleOneStatement ... 1991112941 vex x86: 0xD9 0xF4 (fxtract) 1992110201 == 112941 1993113015 vex amd64->IR: 0xE3 0x14 0x48 0x83 (jrcxz) 1994113126 Crash with binaries built with -gstabs+/-ggdb 1995104065 == 113126 1996115741 == 113126 1997113403 Partial SSE3 support on x86 1998113541 vex: Grp5(x86) (alt encoding inc/dec) case 1 1999113642 valgrind crashes when trying to read debug information 2000113810 vex x86->IR: 66 0F F6 (66 + PSADBW == SSE PSADBW) 2001113796 read() and write() do not work if buffer is in shared memory 2002113851 vex x86->IR: (pmaddwd): 0x66 0xF 0xF5 0xC7 2003114366 vex amd64 cannnot handle __asm__( "fninit" ) 2004114412 vex amd64->IR: 0xF 0xAD 0xC2 0xD3 (128-bit shift, shrdq?) 2005114455 vex amd64->IR: 0xF 0xAC 0xD0 0x1 (also shrdq) 2006115590: amd64->IR: 0x67 0xE3 0x9 0xEB (address size override) 2007115953 valgrind svn r5042 does not build with parallel make (-j3) 2008116057 maximum instruction size - VG_MAX_INSTR_SZB too small? 2009116483 shmat failes with invalid argument 2010102202 valgrind crashes when realloc'ing until out of memory 2011109487 == 102202 2012110536 == 102202 2013112687 == 102202 2014111724 vex amd64->IR: 0x41 0xF 0xAB (more BT{,S,R,C} fun n games) 2015111748 vex amd64->IR: 0xDD 0xE2 (fucom) 2016111785 make fails if CC contains spaces 2017111829 vex x86->IR: sbb AL, Ib 2018111851 vex x86->IR: 0x9F 0x89 (lahf/sahf) 2019112031 iopl on AMD64 and README_MISSING_SYSCALL_OR_IOCTL update 2020112152 code generation for Xin_MFence on x86 with SSE0 subarch 2021112167 == 112152 2022112789 == 112152 2023112199 naked ar tool is used in vex makefile 2024112501 vex x86->IR: movq (0xF 0x7F 0xC1 0xF) (mmx MOVQ) 2025113583 == 112501 2026112538 memalign crash 2027113190 Broken links in docs/html/ 2028113230 Valgrind sys_pipe on x86-64 wrongly thinks file descriptors 2029 should be 64bit 2030113996 vex amd64->IR: fucomp (0xDD 0xE9) 2031114196 vex x86->IR: out %eax,(%dx) (0xEF 0xC9 0xC3 0x90) 2032114289 Memcheck fails to intercept malloc when used in an uclibc environment 2033114756 mbind syscall support 2034114757 Valgrind dies with assertion: Assertion 'noLargerThan > 0' failed 2035114563 stack tracking module not informed when valgrind switches threads 2036114564 clone() and stacks 2037114565 == 114564 2038115496 glibc crashes trying to use sysinfo page 2039116200 enable fsetxattr, fgetxattr, and fremovexattr for amd64 2040 2041(3.1.0RC1: 20 November 2005, vex r1466, valgrind r5224). 2042(3.1.0: 26 November 2005, vex r1471, valgrind r5235). 2043 2044 2045Release 3.0.1 (29 August 2005) 2046~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 20473.0.1 fixes a bunch of bugs reported in 3.0.0. There is no new 2048functionality. Some of the fixed bugs are critical, so if you 2049use/distribute 3.0.0, an upgrade to 3.0.1 is recommended. The fixed 2050bugs are: 2051 2052(note: "n-i-bz" means "not in bugzilla" -- this bug does not have 2053 a bugzilla entry). 2054 2055109313 (== 110505) x86 cmpxchg8b 2056n-i-bz x86: track but ignore changes to %eflags.AC (alignment check) 2057110102 dis_op2_E_G(amd64) 2058110202 x86 sys_waitpid(#286) 2059110203 clock_getres(,0) 2060110208 execve fail wrong retval 2061110274 SSE1 now mandatory for x86 2062110388 amd64 0xDD 0xD1 2063110464 amd64 0xDC 0x1D FCOMP 2064110478 amd64 0xF 0xD PREFETCH 2065n-i-bz XML <unique> printing wrong 2066n-i-bz Dirk r4359 (amd64 syscalls from trunk) 2067110591 amd64 and x86: rdtsc not implemented properly 2068n-i-bz Nick r4384 (stub implementations of Addrcheck and Helgrind) 2069110652 AMD64 valgrind crashes on cwtd instruction 2070110653 AMD64 valgrind crashes on sarb $0x4,foo(%rip) instruction 2071110656 PATH=/usr/bin::/bin valgrind foobar stats ./fooba 2072110657 Small test fixes 2073110671 vex x86->IR: unhandled instruction bytes: 0xF3 0xC3 (rep ret) 2074n-i-bz Nick (Cachegrind should not assert when it encounters a client 2075 request.) 2076110685 amd64->IR: unhandled instruction bytes: 0xE1 0x56 (loope Jb) 2077110830 configuring with --host fails to build 32 bit on 64 bit target 2078110875 Assertion when execve fails 2079n-i-bz Updates to Memcheck manual 2080n-i-bz Fixed broken malloc_usable_size() 2081110898 opteron instructions missing: btq btsq btrq bsfq 2082110954 x86->IR: unhandled instruction bytes: 0xE2 0xF6 (loop Jb) 2083n-i-bz Make suppressions work for "???" lines in stacktraces. 2084111006 bogus warnings from linuxthreads 2085111092 x86: dis_Grp2(Reg): unhandled case(x86) 2086111231 sctp_getladdrs() and sctp_getpaddrs() returns uninitialized 2087 memory 2088111102 (comment #4) Fixed 64-bit unclean "silly arg" message 2089n-i-bz vex x86->IR: unhandled instruction bytes: 0x14 0x0 2090n-i-bz minor umount/fcntl wrapper fixes 2091111090 Internal Error running Massif 2092101204 noisy warning 2093111513 Illegal opcode for SSE instruction (x86 movups) 2094111555 VEX/Makefile: CC is set to gcc 2095n-i-bz Fix XML bugs in FAQ 2096 2097(3.0.1: 29 August 05, 2098 vex/branches/VEX_3_0_BRANCH r1367, 2099 valgrind/branches/VALGRIND_3_0_BRANCH r4574). 2100 2101 2102 2103Release 3.0.0 (3 August 2005) 2104~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 21053.0.0 is a major overhaul of Valgrind. The most significant user 2106visible change is that Valgrind now supports architectures other than 2107x86. The new architectures it supports are AMD64 and PPC32, and the 2108infrastructure is present for other architectures to be added later. 2109 2110AMD64 support works well, but has some shortcomings: 2111 2112- It generally won't be as solid as the x86 version. For example, 2113 support for more obscure instructions and system calls may be missing. 2114 We will fix these as they arise. 2115 2116- Address space may be limited; see the point about 2117 position-independent executables below. 2118 2119- If Valgrind is built on an AMD64 machine, it will only run 64-bit 2120 executables. If you want to run 32-bit x86 executables under Valgrind 2121 on an AMD64, you will need to build Valgrind on an x86 machine and 2122 copy it to the AMD64 machine. And it probably won't work if you do 2123 something tricky like exec'ing a 32-bit program from a 64-bit program 2124 while using --trace-children=yes. We hope to improve this situation 2125 in the future. 2126 2127The PPC32 support is very basic. It may not work reliably even for 2128small programs, but it's a start. Many thanks to Paul Mackerras for 2129his great work that enabled this support. We are working to make 2130PPC32 usable as soon as possible. 2131 2132Other user-visible changes: 2133 2134- Valgrind is no longer built by default as a position-independent 2135 executable (PIE), as this caused too many problems. 2136 2137 Without PIE enabled, AMD64 programs will only be able to access 2GB of 2138 address space. We will fix this eventually, but not for the moment. 2139 2140 Use --enable-pie at configure-time to turn this on. 2141 2142- Support for programs that use stack-switching has been improved. Use 2143 the --max-stackframe flag for simple cases, and the 2144 VALGRIND_STACK_REGISTER, VALGRIND_STACK_DEREGISTER and 2145 VALGRIND_STACK_CHANGE client requests for trickier cases. 2146 2147- Support for programs that use self-modifying code has been improved, 2148 in particular programs that put temporary code fragments on the stack. 2149 This helps for C programs compiled with GCC that use nested functions, 2150 and also Ada programs. This is controlled with the --smc-check 2151 flag, although the default setting should work in most cases. 2152 2153- Output can now be printed in XML format. This should make it easier 2154 for tools such as GUI front-ends and automated error-processing 2155 schemes to use Valgrind output as input. The --xml flag controls this. 2156 As part of this change, ELF directory information is read from executables, 2157 so absolute source file paths are available if needed. 2158 2159- Programs that allocate many heap blocks may run faster, due to 2160 improvements in certain data structures. 2161 2162- Addrcheck is currently not working. We hope to get it working again 2163 soon. Helgrind is still not working, as was the case for the 2.4.0 2164 release. 2165 2166- The JITter has been completely rewritten, and is now in a separate 2167 library, called Vex. This enabled a lot of the user-visible changes, 2168 such as new architecture support. The new JIT unfortunately translates 2169 more slowly than the old one, so programs may take longer to start. 2170 We believe the code quality is produces is about the same, so once 2171 started, programs should run at about the same speed. Feedback about 2172 this would be useful. 2173 2174 On the plus side, Vex and hence Memcheck tracks value flow properly 2175 through floating point and vector registers, something the 2.X line 2176 could not do. That means that Memcheck is much more likely to be 2177 usably accurate on vectorised code. 2178 2179- There is a subtle change to the way exiting of threaded programs 2180 is handled. In 3.0, Valgrind's final diagnostic output (leak check, 2181 etc) is not printed until the last thread exits. If the last thread 2182 to exit was not the original thread which started the program, any 2183 other process wait()-ing on this one to exit may conclude it has 2184 finished before the diagnostic output is printed. This may not be 2185 what you expect. 2.X had a different scheme which avoided this 2186 problem, but caused deadlocks under obscure circumstances, so we 2187 are trying something different for 3.0. 2188 2189- Small changes in control log file naming which make it easier to 2190 use valgrind for debugging MPI-based programs. The relevant 2191 new flags are --log-file-exactly= and --log-file-qualifier=. 2192 2193- As part of adding AMD64 support, DWARF2 CFI-based stack unwinding 2194 support was added. In principle this means Valgrind can produce 2195 meaningful backtraces on x86 code compiled with -fomit-frame-pointer 2196 providing you also compile your code with -fasynchronous-unwind-tables. 2197 2198- The documentation build system has been completely redone. 2199 The documentation masters are now in XML format, and from that 2200 HTML, PostScript and PDF documentation is generated. As a result 2201 the manual is now available in book form. Note that the 2202 documentation in the source tarballs is pre-built, so you don't need 2203 any XML processing tools to build Valgrind from a tarball. 2204 2205Changes that are not user-visible: 2206 2207- The code has been massively overhauled in order to modularise it. 2208 As a result we hope it is easier to navigate and understand. 2209 2210- Lots of code has been rewritten. 2211 2212BUGS FIXED: 2213 2214110046 sz == 4 assertion failed 2215109810 vex amd64->IR: unhandled instruction bytes: 0xA3 0x4C 0x70 0xD7 2216109802 Add a plausible_stack_size command-line parameter ? 2217109783 unhandled ioctl TIOCMGET (running hw detection tool discover) 2218109780 unhandled ioctl BLKSSZGET (running fdisk -l /dev/hda) 2219109718 vex x86->IR: unhandled instruction: ffreep 2220109429 AMD64 unhandled syscall: 127 (sigpending) 2221109401 false positive uninit in strchr from ld-linux.so.2 2222109385 "stabs" parse failure 2223109378 amd64: unhandled instruction REP NOP 2224109376 amd64: unhandled instruction LOOP Jb 2225109363 AMD64 unhandled instruction bytes 2226109362 AMD64 unhandled syscall: 24 (sched_yield) 2227109358 fork() won't work with valgrind-3.0 SVN 2228109332 amd64 unhandled instruction: ADC Ev, Gv 2229109314 Bogus memcheck report on amd64 2230108883 Crash; vg_memory.c:905 (vgPlain_init_shadow_range): 2231 Assertion `vgPlain_defined_init_shadow_page()' failed. 2232108349 mincore syscall parameter checked incorrectly 2233108059 build infrastructure: small update 2234107524 epoll_ctl event parameter checked on EPOLL_CTL_DEL 2235107123 Vex dies with unhandled instructions: 0xD9 0x31 0xF 0xAE 2236106841 auxmap & openGL problems 2237106713 SDL_Init causes valgrind to exit 2238106352 setcontext and makecontext not handled correctly 2239106293 addresses beyond initial client stack allocation 2240 not checked in VALGRIND_DO_LEAK_CHECK 2241106283 PIE client programs are loaded at address 0 2242105831 Assertion `vgPlain_defined_init_shadow_page()' failed. 2243105039 long run-times probably due to memory manager 2244104797 valgrind needs to be aware of BLKGETSIZE64 2245103594 unhandled instruction: FICOM 2246103320 Valgrind 2.4.0 fails to compile with gcc 3.4.3 and -O0 2247103168 potentially memory leak in coregrind/ume.c 2248102039 bad permissions for mapped region at address 0xB7C73680 2249101881 weird assertion problem 2250101543 Support fadvise64 syscalls 225175247 x86_64/amd64 support (the biggest "bug" we have ever fixed) 2252 2253(3.0RC1: 27 July 05, vex r1303, valgrind r4283). 2254(3.0.0: 3 August 05, vex r1313, valgrind r4316). 2255 2256 2257 2258Stable release 2.4.1 (1 August 2005) 2259~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2260(The notes for this release have been lost. Sorry! It would have 2261contained various bug fixes but no new features.) 2262 2263 2264 2265Stable release 2.4.0 (March 2005) -- CHANGES RELATIVE TO 2.2.0 2266~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 22672.4.0 brings many significant changes and bug fixes. The most 2268significant user-visible change is that we no longer supply our own 2269pthread implementation. Instead, Valgrind is finally capable of 2270running the native thread library, either LinuxThreads or NPTL. 2271 2272This means our libpthread has gone, along with the bugs associated 2273with it. Valgrind now supports the kernel's threading syscalls, and 2274lets you use your standard system libpthread. As a result: 2275 2276* There are many fewer system dependencies and strange library-related 2277 bugs. There is a small performance improvement, and a large 2278 stability improvement. 2279 2280* On the downside, Valgrind can no longer report misuses of the POSIX 2281 PThreads API. It also means that Helgrind currently does not work. 2282 We hope to fix these problems in a future release. 2283 2284Note that running the native thread libraries does not mean Valgrind 2285is able to provide genuine concurrent execution on SMPs. We still 2286impose the restriction that only one thread is running at any given 2287time. 2288 2289There are many other significant changes too: 2290 2291* Memcheck is (once again) the default tool. 2292 2293* The default stack backtrace is now 12 call frames, rather than 4. 2294 2295* Suppressions can have up to 25 call frame matches, rather than 4. 2296 2297* Memcheck and Addrcheck use less memory. Under some circumstances, 2298 they no longer allocate shadow memory if there are large regions of 2299 memory with the same A/V states - such as an mmaped file. 2300 2301* The memory-leak detector in Memcheck and Addrcheck has been 2302 improved. It now reports more types of memory leak, including 2303 leaked cycles. When reporting leaked memory, it can distinguish 2304 between directly leaked memory (memory with no references), and 2305 indirectly leaked memory (memory only referred to by other leaked 2306 memory). 2307 2308* Memcheck's confusion over the effect of mprotect() has been fixed: 2309 previously mprotect could erroneously mark undefined data as 2310 defined. 2311 2312* Signal handling is much improved and should be very close to what 2313 you get when running natively. 2314 2315 One result of this is that Valgrind observes changes to sigcontexts 2316 passed to signal handlers. Such modifications will take effect when 2317 the signal returns. You will need to run with --single-step=yes to 2318 make this useful. 2319 2320* Valgrind is built in Position Independent Executable (PIE) format if 2321 your toolchain supports it. This allows it to take advantage of all 2322 the available address space on systems with 4Gbyte user address 2323 spaces. 2324 2325* Valgrind can now run itself (requires PIE support). 2326 2327* Syscall arguments are now checked for validity. Previously all 2328 memory used by syscalls was checked, but now the actual values 2329 passed are also checked. 2330 2331* Syscall wrappers are more robust against bad addresses being passed 2332 to syscalls: they will fail with EFAULT rather than killing Valgrind 2333 with SIGSEGV. 2334 2335* Because clone() is directly supported, some non-pthread uses of it 2336 will work. Partial sharing (where some resources are shared, and 2337 some are not) is not supported. 2338 2339* open() and readlink() on /proc/self/exe are supported. 2340 2341BUGS FIXED: 2342 234388520 pipe+fork+dup2 kills the main program 234488604 Valgrind Aborts when using $VALGRIND_OPTS and user progra... 234588614 valgrind: vg_libpthread.c:2323 (read): Assertion `read_pt... 234688703 Stabs parser fails to handle ";" 234788886 ioctl wrappers for TIOCMBIS and TIOCMBIC 234889032 valgrind pthread_cond_timedwait fails 234989106 the 'impossible' happened 235089139 Missing sched_setaffinity & sched_getaffinity 235189198 valgrind lacks support for SIOCSPGRP and SIOCGPGRP 235289263 Missing ioctl translations for scsi-generic and CD playing 235389440 tests/deadlock.c line endings 235489481 `impossible' happened: EXEC FAILED 235589663 valgrind 2.2.0 crash on Redhat 7.2 235689792 Report pthread_mutex_lock() deadlocks instead of returnin... 235790111 statvfs64 gives invalid error/warning 235890128 crash+memory fault with stabs generated by gnat for a run... 235990778 VALGRIND_CHECK_DEFINED() not as documented in memcheck.h 236090834 cachegrind crashes at end of program without reporting re... 236191028 valgrind: vg_memory.c:229 (vgPlain_unmap_range): Assertio... 236291162 valgrind crash while debugging drivel 1.2.1 236391199 Unimplemented function 236491325 Signal routing does not propagate the siginfo structure 236591599 Assertion `cv == ((void *)0)' 236691604 rw_lookup clears orig and sends the NULL value to rw_new 236791821 Small problems building valgrind with $top_builddir ne $t... 236891844 signal 11 (SIGSEGV) at get_tcb (libpthread.c:86) in corec... 236992264 UNIMPLEMENTED FUNCTION: pthread_condattr_setpshared 237092331 per-target flags necessitate AM_PROG_CC_C_O 237192420 valgrind doesn't compile with linux 2.6.8.1/9 237292513 Valgrind 2.2.0 generates some warning messages 237392528 vg_symtab2.c:170 (addLoc): Assertion `loc->size > 0' failed. 237493096 unhandled ioctl 0x4B3A and 0x5601 237593117 Tool and core interface versions do not match 237693128 Can't run valgrind --tool=memcheck because of unimplement... 237793174 Valgrind can crash if passed bad args to certain syscalls 237893309 Stack frame in new thread is badly aligned 237993328 Wrong types used with sys_sigprocmask() 238093763 /usr/include/asm/msr.h is missing 238193776 valgrind: vg_memory.c:508 (vgPlain_find_map_space): Asser... 238293810 fcntl() argument checking a bit too strict 238394378 Assertion `tst->sigqueue_head != tst->sigqueue_tail' failed. 238494429 valgrind 2.2.0 segfault with mmap64 in glibc 2.3.3 238594645 Impossible happened: PINSRW mem 238694953 valgrind: the `impossible' happened: SIGSEGV 238795667 Valgrind does not work with any KDE app 238896243 Assertion 'res==0' failed 238996252 stage2 loader of valgrind fails to allocate memory 239096520 All programs crashing at _dl_start (in /lib/ld-2.3.3.so) ... 239196660 ioctl CDROMREADTOCENTRY causes bogus warnings 239296747 After looping in a segfault handler, the impossible happens 239396923 Zero sized arrays crash valgrind trace back with SIGFPE 239496948 valgrind stops with assertion failure regarding mmap2 239596966 valgrind fails when application opens more than 16 sockets 239697398 valgrind: vg_libpthread.c:2667 Assertion failed 239797407 valgrind: vg_mylibc.c:1226 (vgPlain_safe_fd): Assertion `... 239897427 "Warning: invalid file descriptor -1 in syscall close()" ... 239997785 missing backtrace 240097792 build in obj dir fails - autoconf / makefile cleanup 240197880 pthread_mutex_lock fails from shared library (special ker... 240297975 program aborts without ang VG messages 240398129 Failed when open and close file 230000 times using stdio 240498175 Crashes when using valgrind-2.2.0 with a program using al... 240598288 Massif broken 240698303 UNIMPLEMENTED FUNCTION pthread_condattr_setpshared 240798630 failed--compilation missing warnings.pm, fails to make he... 240898756 Cannot valgrind signal-heavy kdrive X server 240998966 valgrinding the JVM fails with a sanity check assertion 241099035 Valgrind crashes while profiling 241199142 loops with message "Signal 11 being dropped from thread 0... 241299195 threaded apps crash on thread start (using QThread::start... 241399348 Assertion `vgPlain_lseek(core_fd, 0, 1) == phdrs[i].p_off... 241499568 False negative due to mishandling of mprotect 241599738 valgrind memcheck crashes on program that uses sigitimer 241699923 0-sized allocations are reported as leaks 241799949 program seg faults after exit() 2418100036 "newSuperblock's request for 1048576 bytes failed" 2419100116 valgrind: (pthread_cond_init): Assertion `sizeof(* cond) ... 2420100486 memcheck reports "valgrind: the `impossible' happened: V... 2421100833 second call to "mremap" fails with EINVAL 2422101156 (vgPlain_find_map_space): Assertion `(addr & ((1 << 12)-1... 2423101173 Assertion `recDepth >= 0 && recDepth < 500' failed 2424101291 creating threads in a forked process fails 2425101313 valgrind causes different behavior when resizing a window... 2426101423 segfault for c++ array of floats 2427101562 valgrind massif dies on SIGINT even with signal handler r... 2428 2429 2430Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.0.0 2431~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 24322.2.0 brings nine months worth of improvements and bug fixes. We 2433believe it to be a worthy successor to 2.0.0. There are literally 2434hundreds of bug fixes and minor improvements. There are also some 2435fairly major user-visible changes: 2436 2437* A complete overhaul of handling of system calls and signals, and 2438 their interaction with threads. In general, the accuracy of the 2439 system call, thread and signal simulations is much improved: 2440 2441 - Blocking system calls behave exactly as they do when running 2442 natively (not on valgrind). That is, if a syscall blocks only the 2443 calling thread when running natively, than it behaves the same on 2444 valgrind. No more mysterious hangs because V doesn't know that some 2445 syscall or other, should block only the calling thread. 2446 2447 - Interrupted syscalls should now give more faithful results. 2448 2449 - Signal contexts in signal handlers are supported. 2450 2451* Improvements to NPTL support to the extent that V now works 2452 properly on NPTL-only setups. 2453 2454* Greater isolation between Valgrind and the program being run, so 2455 the program is less likely to inadvertently kill Valgrind by 2456 doing wild writes. 2457 2458* Massif: a new space profiling tool. Try it! It's cool, and it'll 2459 tell you in detail where and when your C/C++ code is allocating heap. 2460 Draws pretty .ps pictures of memory use against time. A potentially 2461 powerful tool for making sense of your program's space use. 2462 2463* File descriptor leakage checks. When enabled, Valgrind will print out 2464 a list of open file descriptors on exit. 2465 2466* Improved SSE2/SSE3 support. 2467 2468* Time-stamped output; use --time-stamp=yes 2469 2470 2471 2472Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.1.2 2473~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 24742.2.0 is not much different from 2.1.2, released seven weeks ago. 2475A number of bugs have been fixed, most notably #85658, which gave 2476problems for quite a few people. There have been many internal 2477cleanups, but those are not user visible. 2478 2479The following bugs have been fixed since 2.1.2: 2480 248185658 Assert in coregrind/vg_libpthread.c:2326 (open64) != 2482 (void*)0 failed 2483 This bug was reported multiple times, and so the following 2484 duplicates of it are also fixed: 87620, 85796, 85935, 86065, 2485 86919, 86988, 87917, 88156 2486 248780716 Semaphore mapping bug caused by unmap (sem_destroy) 2488 (Was fixed prior to 2.1.2) 2489 249086987 semctl and shmctl syscalls family is not handled properly 2491 249286696 valgrind 2.1.2 + RH AS2.1 + librt 2493 249486730 valgrind locks up at end of run with assertion failure 2495 in __pthread_unwind 2496 249786641 memcheck doesn't work with Mesa OpenGL/ATI on Suse 9.1 2498 (also fixes 74298, a duplicate of this) 2499 250085947 MMX/SSE unhandled instruction 'sfence' 2501 250284978 Wrong error "Conditional jump or move depends on 2503 uninitialised value" resulting from "sbbl %reg, %reg" 2504 250586254 ssort() fails when signed int return type from comparison is 2506 too small to handle result of unsigned int subtraction 2507 250887089 memalign( 4, xxx) makes valgrind assert 2509 251086407 Add support for low-level parallel port driver ioctls. 2511 251270587 Add timestamps to Valgrind output? (wishlist) 2513 251484937 vg_libpthread.c:2505 (se_remap): Assertion `res == 0' 2515 (fixed prior to 2.1.2) 2516 251786317 cannot load libSDL-1.2.so.0 using valgrind 2518 251986989 memcpy from mac_replace_strmem.c complains about 2520 uninitialized pointers passed when length to copy is zero 2521 252285811 gnu pascal symbol causes segmentation fault; ok in 2.0.0 2523 252479138 writing to sbrk()'d memory causes segfault 2525 252677369 sched deadlock while signal received during pthread_join 2527 and the joined thread exited 2528 252988115 In signal handler for SIGFPE, siginfo->si_addr is wrong 2530 under Valgrind 2531 253278765 Massif crashes on app exit if FP exceptions are enabled 2533 2534Additionally there are the following changes, which are not 2535connected to any bug report numbers, AFAICS: 2536 2537* Fix scary bug causing mis-identification of SSE stores vs 2538 loads and so causing memcheck to sometimes give nonsense results 2539 on SSE code. 2540 2541* Add support for the POSIX message queue system calls. 2542 2543* Fix to allow 32-bit Valgrind to run on AMD64 boxes. Note: this does 2544 NOT allow Valgrind to work with 64-bit executables - only with 32-bit 2545 executables on an AMD64 box. 2546 2547* At configure time, only check whether linux/mii.h can be processed 2548 so that we don't generate ugly warnings by trying to compile it. 2549 2550* Add support for POSIX clocks and timers. 2551 2552 2553 2554Developer (cvs head) release 2.1.2 (18 July 2004) 2555~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 25562.1.2 contains four months worth of bug fixes and refinements. 2557Although officially a developer release, we believe it to be stable 2558enough for widespread day-to-day use. 2.1.2 is pretty good, so try it 2559first, although there is a chance it won't work. If so then try 2.0.0 2560and tell us what went wrong." 2.1.2 fixes a lot of problems present 2561in 2.0.0 and is generally a much better product. 2562 2563Relative to 2.1.1, a large number of minor problems with 2.1.1 have 2564been fixed, and so if you use 2.1.1 you should try 2.1.2. Users of 2565the last stable release, 2.0.0, might also want to try this release. 2566 2567The following bugs, and probably many more, have been fixed. These 2568are listed at http://bugs.kde.org. Reporting a bug for valgrind in 2569the http://bugs.kde.org is much more likely to get you a fix than 2570mailing developers directly, so please continue to keep sending bugs 2571there. 2572 257376869 Crashes when running any tool under Fedora Core 2 test1 2574 This fixes the problem with returning from a signal handler 2575 when VDSOs are turned off in FC2. 2576 257769508 java 1.4.2 client fails with erroneous "stack size too small". 2578 This fix makes more of the pthread stack attribute related 2579 functions work properly. Java still doesn't work though. 2580 258171906 malloc alignment should be 8, not 4 2582 All memory returned by malloc/new etc is now at least 2583 8-byte aligned. 2584 258581970 vg_alloc_ThreadState: no free slots available 2586 (closed because the workaround is simple: increase 2587 VG_N_THREADS, rebuild and try again.) 2588 258978514 Conditional jump or move depends on uninitialized value(s) 2590 (a slight mishanding of FP code in memcheck) 2591 259277952 pThread Support (crash) (due to initialisation-ordering probs) 2593 (also 85118) 2594 259580942 Addrcheck wasn't doing overlap checking as it should. 259678048 return NULL on malloc/new etc failure, instead of asserting 259773655 operator new() override in user .so files often doesn't get picked up 259883060 Valgrind does not handle native kernel AIO 259969872 Create proper coredumps after fatal signals 260082026 failure with new glibc versions: __libc_* functions are not exported 260170344 UNIMPLEMENTED FUNCTION: tcdrain 260281297 Cancellation of pthread_cond_wait does not require mutex 260382872 Using debug info from additional packages (wishlist) 260483025 Support for ioctls FIGETBSZ and FIBMAP 260583340 Support for ioctl HDIO_GET_IDENTITY 260679714 Support for the semtimedop system call. 260777022 Support for ioctls FBIOGET_VSCREENINFO and FBIOGET_FSCREENINFO 260882098 hp2ps ansification (wishlist) 260983573 Valgrind SIGSEGV on execve 261082999 show which cmdline option was erroneous (wishlist) 261183040 make valgrind VPATH and distcheck-clean (wishlist) 261283998 Assertion `newfd > vgPlain_max_fd' failed (see below) 261382722 Unchecked mmap in as_pad leads to mysterious failures later 261478958 memcheck seg faults while running Mozilla 261585416 Arguments with colon (e.g. --logsocket) ignored 2616 2617 2618Additionally there are the following changes, which are not 2619connected to any bug report numbers, AFAICS: 2620 2621* Rearranged address space layout relative to 2.1.1, so that 2622 Valgrind/tools will run out of memory later than currently in many 2623 circumstances. This is good news esp. for Calltree. It should 2624 be possible for client programs to allocate over 800MB of 2625 memory when using memcheck now. 2626 2627* Improved checking when laying out memory. Should hopefully avoid 2628 the random segmentation faults that 2.1.1 sometimes caused. 2629 2630* Support for Fedora Core 2 and SuSE 9.1. Improvements to NPTL 2631 support to the extent that V now works properly on NPTL-only setups. 2632 2633* Renamed the following options: 2634 --logfile-fd --> --log-fd 2635 --logfile --> --log-file 2636 --logsocket --> --log-socket 2637 to be consistent with each other and other options (esp. --input-fd). 2638 2639* Add support for SIOCGMIIPHY, SIOCGMIIREG and SIOCSMIIREG ioctls and 2640 improve the checking of other interface related ioctls. 2641 2642* Fix building with gcc-3.4.1. 2643 2644* Remove limit on number of semaphores supported. 2645 2646* Add support for syscalls: set_tid_address (258), acct (51). 2647 2648* Support instruction "repne movs" -- not official but seems to occur. 2649 2650* Implement an emulated soft limit for file descriptors in addition to 2651 the current reserved area, which effectively acts as a hard limit. The 2652 setrlimit system call now simply updates the emulated limits as best 2653 as possible - the hard limit is not allowed to move at all and just 2654 returns EPERM if you try and change it. This should stop reductions 2655 in the soft limit causing assertions when valgrind tries to allocate 2656 descriptors from the reserved area. 2657 (This actually came from bug #83998). 2658 2659* Major overhaul of Cachegrind implementation. First user-visible change 2660 is that cachegrind.out files are now typically 90% smaller than they 2661 used to be; code annotation times are correspondingly much smaller. 2662 Second user-visible change is that hit/miss counts for code that is 2663 unloaded at run-time is no longer dumped into a single "discard" pile, 2664 but accurately preserved. 2665 2666* Client requests for telling valgrind about memory pools. 2667 2668 2669 2670Developer (cvs head) release 2.1.1 (12 March 2004) 2671~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 26722.1.1 contains some internal structural changes needed for V's 2673long-term future. These don't affect end-users. Most notable 2674user-visible changes are: 2675 2676* Greater isolation between Valgrind and the program being run, so 2677 the program is less likely to inadvertently kill Valgrind by 2678 doing wild writes. 2679 2680* Massif: a new space profiling tool. Try it! It's cool, and it'll 2681 tell you in detail where and when your C/C++ code is allocating heap. 2682 Draws pretty .ps pictures of memory use against time. A potentially 2683 powerful tool for making sense of your program's space use. 2684 2685* Fixes for many bugs, including support for more SSE2/SSE3 instructions, 2686 various signal/syscall things, and various problems with debug 2687 info readers. 2688 2689* Support for glibc-2.3.3 based systems. 2690 2691We are now doing automatic overnight build-and-test runs on a variety 2692of distros. As a result, we believe 2.1.1 builds and runs on: 2693Red Hat 7.2, 7.3, 8.0, 9, Fedora Core 1, SuSE 8.2, SuSE 9. 2694 2695 2696The following bugs, and probably many more, have been fixed. These 2697are listed at http://bugs.kde.org. Reporting a bug for valgrind in 2698the http://bugs.kde.org is much more likely to get you a fix than 2699mailing developers directly, so please continue to keep sending bugs 2700there. 2701 270269616 glibc 2.3.2 w/NPTL is massively different than what valgrind expects 270369856 I don't know how to instrument MMXish stuff (Helgrind) 270473892 valgrind segfaults starting with Objective-C debug info 2705 (fix for S-type stabs) 270673145 Valgrind complains too much about close(<reserved fd>) 270773902 Shadow memory allocation seems to fail on RedHat 8.0 270868633 VG_N_SEMAPHORES too low (V itself was leaking semaphores) 270975099 impossible to trace multiprocess programs 271076839 the `impossible' happened: disInstr: INT but not 0x80 ! 271176762 vg_to_ucode.c:3748 (dis_push_segreg): Assertion `sz == 4' failed. 271276747 cannot include valgrind.h in c++ program 271376223 parsing B(3,10) gave NULL type => impossible happens 271475604 shmdt handling problem 271576416 Problems with gcc 3.4 snap 20040225 271675614 using -gstabs when building your programs the `impossible' happened 271775787 Patch for some CDROM ioctls CDORM_GET_MCN, CDROM_SEND_PACKET, 271875294 gcc 3.4 snapshot's libstdc++ have unsupported instructions. 2719 (REP RET) 272073326 vg_symtab2.c:272 (addScopeRange): Assertion `range->size > 0' failed. 272172596 not recognizing __libc_malloc 272269489 Would like to attach ddd to running program 272372781 Cachegrind crashes with kde programs 272473055 Illegal operand at DXTCV11CompressBlockSSE2 (more SSE opcodes) 272573026 Descriptor leak check reports port numbers wrongly 272671705 README_MISSING_SYSCALL_OR_IOCTL out of date 272772643 Improve support for SSE/SSE2 instructions 272872484 valgrind leaves it's own signal mask in place when execing 272972650 Signal Handling always seems to restart system calls 273072006 The mmap system call turns all errors in ENOMEM 273171781 gdb attach is pretty useless 273271180 unhandled instruction bytes: 0xF 0xAE 0x85 0xE8 273369886 writes to zero page cause valgrind to assert on exit 273471791 crash when valgrinding gimp 1.3 (stabs reader problem) 273569783 unhandled syscall: 218 273669782 unhandled instruction bytes: 0x66 0xF 0x2B 0x80 273770385 valgrind fails if the soft file descriptor limit is less 2738 than about 828 273969529 "rep; nop" should do a yield 274070827 programs with lots of shared libraries report "mmap failed" 2741 for some of them when reading symbols 274271028 glibc's strnlen is optimised enough to confuse valgrind 2743 2744 2745 2746 2747Unstable (cvs head) release 2.1.0 (15 December 2003) 2748~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2749For whatever it's worth, 2.1.0 actually seems pretty darn stable to me 2750(Julian). It looks eminently usable, and given that it fixes some 2751significant bugs, may well be worth using on a day-to-day basis. 27522.1.0 is known to build and pass regression tests on: SuSE 9, SuSE 27538.2, RedHat 8. 2754 27552.1.0 most notably includes Jeremy Fitzhardinge's complete overhaul of 2756handling of system calls and signals, and their interaction with 2757threads. In general, the accuracy of the system call, thread and 2758signal simulations is much improved. Specifically: 2759 2760- Blocking system calls behave exactly as they do when running 2761 natively (not on valgrind). That is, if a syscall blocks only the 2762 calling thread when running natively, than it behaves the same on 2763 valgrind. No more mysterious hangs because V doesn't know that some 2764 syscall or other, should block only the calling thread. 2765 2766- Interrupted syscalls should now give more faithful results. 2767 2768- Finally, signal contexts in signal handlers are supported. As a 2769 result, konqueror on SuSE 9 no longer segfaults when notified of 2770 file changes in directories it is watching. 2771 2772Other changes: 2773 2774- Robert Walsh's file descriptor leakage checks. When enabled, 2775 Valgrind will print out a list of open file descriptors on 2776 exit. Along with each file descriptor, Valgrind prints out a stack 2777 backtrace of where the file was opened and any details relating to the 2778 file descriptor such as the file name or socket details. 2779 To use, give: --track-fds=yes 2780 2781- Implemented a few more SSE/SSE2 instructions. 2782 2783- Less crud on the stack when you do 'where' inside a GDB attach. 2784 2785- Fixed the following bugs: 2786 68360: Valgrind does not compile against 2.6.0-testX kernels 2787 68525: CVS head doesn't compile on C90 compilers 2788 68566: pkgconfig support (wishlist) 2789 68588: Assertion `sz == 4' failed in vg_to_ucode.c (disInstr) 2790 69140: valgrind not able to explicitly specify a path to a binary. 2791 69432: helgrind asserts encountering a MutexErr when there are 2792 EraserErr suppressions 2793 2794- Increase the max size of the translation cache from 200k average bbs 2795 to 300k average bbs. Programs on the size of OOo (680m17) are 2796 thrashing the cache at the smaller size, creating large numbers of 2797 retranslations and wasting significant time as a result. 2798 2799 2800 2801Stable release 2.0.0 (5 Nov 2003) 2802~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2803 28042.0.0 improves SSE/SSE2 support, fixes some minor bugs, and 2805improves support for SuSE 9 and the Red Hat "Severn" beta. 2806 2807- Further improvements to SSE/SSE2 support. The entire test suite of 2808 the GNU Scientific Library (gsl-1.4) compiled with Intel Icc 7.1 2809 20030307Z '-g -O -xW' now works. I think this gives pretty good 2810 coverage of SSE/SSE2 floating point instructions, or at least the 2811 subset emitted by Icc. 2812 2813- Also added support for the following instructions: 2814 MOVNTDQ UCOMISD UNPCKLPS UNPCKHPS SQRTSS 2815 PUSH/POP %{FS,GS}, and PUSH %CS (Nb: there is no POP %CS). 2816 2817- CFI support for GDB version 6. Needed to enable newer GDBs 2818 to figure out where they are when using --gdb-attach=yes. 2819 2820- Fix this: 2821 mc_translate.c:1091 (memcheck_instrument): Assertion 2822 `u_in->size == 4 || u_in->size == 16' failed. 2823 2824- Return an error rather than panicing when given a bad socketcall. 2825 2826- Fix checking of syscall rt_sigtimedwait(). 2827 2828- Implement __NR_clock_gettime (syscall 265). Needed on Red Hat Severn. 2829 2830- Fixed bug in overlap check in strncpy() -- it was assuming the src was 'n' 2831 bytes long, when it could be shorter, which could cause false 2832 positives. 2833 2834- Support use of select() for very large numbers of file descriptors. 2835 2836- Don't fail silently if the executable is statically linked, or is 2837 setuid/setgid. Print an error message instead. 2838 2839- Support for old DWARF-1 format line number info. 2840 2841 2842 2843Snapshot 20031012 (12 October 2003) 2844~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2845 2846Three months worth of bug fixes, roughly. Most significant single 2847change is improved SSE/SSE2 support, mostly thanks to Dirk Mueller. 2848 284920031012 builds on Red Hat Fedora ("Severn") but doesn't really work 2850(curiosly, mozilla runs OK, but a modest "ls -l" bombs). I hope to 2851get a working version out soon. It may or may not work ok on the 2852forthcoming SuSE 9; I hear positive noises about it but haven't been 2853able to verify this myself (not until I get hold of a copy of 9). 2854 2855A detailed list of changes, in no particular order: 2856 2857- Describe --gen-suppressions in the FAQ. 2858 2859- Syscall __NR_waitpid supported. 2860 2861- Minor MMX bug fix. 2862 2863- -v prints program's argv[] at startup. 2864 2865- More glibc-2.3 suppressions. 2866 2867- Suppressions for stack underrun bug(s) in the c++ support library 2868 distributed with Intel Icc 7.0. 2869 2870- Fix problems reading /proc/self/maps. 2871 2872- Fix a couple of messages that should have been suppressed by -q, 2873 but weren't. 2874 2875- Make Addrcheck understand "Overlap" suppressions. 2876 2877- At startup, check if program is statically linked and bail out if so. 2878 2879- Cachegrind: Auto-detect Intel Pentium-M, also VIA Nehemiah 2880 2881- Memcheck/addrcheck: minor speed optimisations 2882 2883- Handle syscall __NR_brk more correctly than before. 2884 2885- Fixed incorrect allocate/free mismatch errors when using 2886 operator new(unsigned, std::nothrow_t const&) 2887 operator new[](unsigned, std::nothrow_t const&) 2888 2889- Support POSIX pthread spinlocks. 2890 2891- Fixups for clean compilation with gcc-3.3.1. 2892 2893- Implemented more opcodes: 2894 - push %es 2895 - push %ds 2896 - pop %es 2897 - pop %ds 2898 - movntq 2899 - sfence 2900 - pshufw 2901 - pavgb 2902 - ucomiss 2903 - enter 2904 - mov imm32, %esp 2905 - all "in" and "out" opcodes 2906 - inc/dec %esp 2907 - A whole bunch of SSE/SSE2 instructions 2908 2909- Memcheck: don't bomb on SSE/SSE2 code. 2910 2911 2912Snapshot 20030725 (25 July 2003) 2913~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2914 2915Fixes some minor problems in 20030716. 2916 2917- Fix bugs in overlap checking for strcpy/memcpy etc. 2918 2919- Do overlap checking with Addrcheck as well as Memcheck. 2920 2921- Fix this: 2922 Memcheck: the `impossible' happened: 2923 get_error_name: unexpected type 2924 2925- Install headers needed to compile new skins. 2926 2927- Remove leading spaces and colon in the LD_LIBRARY_PATH / LD_PRELOAD 2928 passed to non-traced children. 2929 2930- Fix file descriptor leak in valgrind-listener. 2931 2932- Fix longstanding bug in which the allocation point of a 2933 block resized by realloc was not correctly set. This may 2934 have caused confusing error messages. 2935 2936 2937Snapshot 20030716 (16 July 2003) 2938~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2939 294020030716 is a snapshot of our current CVS head (development) branch. 2941This is the branch which will become valgrind-2.0. It contains 2942significant enhancements over the 1.9.X branch. 2943 2944Despite this being a snapshot of the CVS head, it is believed to be 2945quite stable -- at least as stable as 1.9.6 or 1.0.4, if not more so 2946-- and therefore suitable for widespread use. Please let us know asap 2947if it causes problems for you. 2948 2949Two reasons for releasing a snapshot now are: 2950 2951- It's been a while since 1.9.6, and this snapshot fixes 2952 various problems that 1.9.6 has with threaded programs 2953 on glibc-2.3.X based systems. 2954 2955- So as to make available improvements in the 2.0 line. 2956 2957Major changes in 20030716, as compared to 1.9.6: 2958 2959- More fixes to threading support on glibc-2.3.1 and 2.3.2-based 2960 systems (SuSE 8.2, Red Hat 9). If you have had problems 2961 with inconsistent/illogical behaviour of errno, h_errno or the DNS 2962 resolver functions in threaded programs, 20030716 should improve 2963 matters. This snapshot seems stable enough to run OpenOffice.org 2964 1.1rc on Red Hat 7.3, SuSE 8.2 and Red Hat 9, and that's a big 2965 threaded app if ever I saw one. 2966 2967- Automatic generation of suppression records; you no longer 2968 need to write them by hand. Use --gen-suppressions=yes. 2969 2970- strcpy/memcpy/etc check their arguments for overlaps, when 2971 running with the Memcheck or Addrcheck skins. 2972 2973- malloc_usable_size() is now supported. 2974 2975- new client requests: 2976 - VALGRIND_COUNT_ERRORS, VALGRIND_COUNT_LEAKS: 2977 useful with regression testing 2978 - VALGRIND_NON_SIMD_CALL[0123]: for running arbitrary functions 2979 on real CPU (use with caution!) 2980 2981- The GDB attach mechanism is more flexible. Allow the GDB to 2982 be run to be specified by --gdb-path=/path/to/gdb, and specify 2983 which file descriptor V will read its input from with 2984 --input-fd=<number>. 2985 2986- Cachegrind gives more accurate results (wasn't tracking instructions in 2987 malloc() and friends previously, is now). 2988 2989- Complete support for the MMX instruction set. 2990 2991- Partial support for the SSE and SSE2 instruction sets. Work for this 2992 is ongoing. About half the SSE/SSE2 instructions are done, so 2993 some SSE based programs may work. Currently you need to specify 2994 --skin=addrcheck. Basically not suitable for real use yet. 2995 2996- Significant speedups (10%-20%) for standard memory checking. 2997 2998- Fix assertion failure in pthread_once(). 2999 3000- Fix this: 3001 valgrind: vg_intercept.c:598 (vgAllRoadsLeadToRome_select): 3002 Assertion `ms_end >= ms_now' failed. 3003 3004- Implement pthread_mutexattr_setpshared. 3005 3006- Understand Pentium 4 branch hints. Also implemented a couple more 3007 obscure x86 instructions. 3008 3009- Lots of other minor bug fixes. 3010 3011- We have a decent regression test system, for the first time. 3012 This doesn't help you directly, but it does make it a lot easier 3013 for us to track the quality of the system, especially across 3014 multiple linux distributions. 3015 3016 You can run the regression tests with 'make regtest' after 'make 3017 install' completes. On SuSE 8.2 and Red Hat 9 I get this: 3018 3019 == 84 tests, 0 stderr failures, 0 stdout failures == 3020 3021 On Red Hat 8, I get this: 3022 3023 == 84 tests, 2 stderr failures, 1 stdout failure == 3024 corecheck/tests/res_search (stdout) 3025 memcheck/tests/sigaltstack (stderr) 3026 3027 sigaltstack is probably harmless. res_search doesn't work 3028 on R H 8 even running natively, so I'm not too worried. 3029 3030 On Red Hat 7.3, a glibc-2.2.5 system, I get these harmless failures: 3031 3032 == 84 tests, 2 stderr failures, 1 stdout failure == 3033 corecheck/tests/pth_atfork1 (stdout) 3034 corecheck/tests/pth_atfork1 (stderr) 3035 memcheck/tests/sigaltstack (stderr) 3036 3037 You need to run on a PII system, at least, since some tests 3038 contain P6-specific instructions, and the test machine needs 3039 access to the internet so that corecheck/tests/res_search 3040 (a test that the DNS resolver works) can function. 3041 3042As ever, thanks for the vast amount of feedback :) and bug reports :( 3043We may not answer all messages, but we do at least look at all of 3044them, and tend to fix the most frequently reported bugs. 3045 3046 3047 3048Version 1.9.6 (7 May 2003 or thereabouts) 3049~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 3050 3051Major changes in 1.9.6: 3052 3053- Improved threading support for glibc >= 2.3.2 (SuSE 8.2, 3054 RedHat 9, to name but two ...) It turned out that 1.9.5 3055 had problems with threading support on glibc >= 2.3.2, 3056 usually manifested by threaded programs deadlocking in system calls, 3057 or running unbelievably slowly. Hopefully these are fixed now. 1.9.6 3058 is the first valgrind which gives reasonable support for 3059 glibc-2.3.2. Also fixed a 2.3.2 problem with pthread_atfork(). 3060 3061- Majorly expanded FAQ.txt. We've added workarounds for all 3062 common problems for which a workaround is known. 3063 3064Minor changes in 1.9.6: 3065 3066- Fix identification of the main thread's stack. Incorrect 3067 identification of it was causing some on-stack addresses to not get 3068 identified as such. This only affected the usefulness of some error 3069 messages; the correctness of the checks made is unchanged. 3070 3071- Support for kernels >= 2.5.68. 3072 3073- Dummy implementations of __libc_current_sigrtmin, 3074 __libc_current_sigrtmax and __libc_allocate_rtsig, hopefully 3075 good enough to keep alive programs which previously died for lack of 3076 them. 3077 3078- Fix bug in the VALGRIND_DISCARD_TRANSLATIONS client request. 3079 3080- Fix bug in the DWARF2 debug line info loader, when instructions 3081 following each other have source lines far from each other 3082 (e.g. with inlined functions). 3083 3084- Debug info reading: read symbols from both "symtab" and "dynsym" 3085 sections, rather than merely from the one that comes last in the 3086 file. 3087 3088- New syscall support: prctl(), creat(), lookup_dcookie(). 3089 3090- When checking calls to accept(), recvfrom(), getsocketopt(), 3091 don't complain if buffer values are NULL. 3092 3093- Try and avoid assertion failures in 3094 mash_LD_PRELOAD_and_LD_LIBRARY_PATH. 3095 3096- Minor bug fixes in cg_annotate. 3097 3098 3099 3100Version 1.9.5 (7 April 2003) 3101~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 3102 3103It occurs to me that it would be helpful for valgrind users to record 3104in the source distribution the changes in each release. So I now 3105attempt to mend my errant ways :-) Changes in this and future releases 3106will be documented in the NEWS file in the source distribution. 3107 3108Major changes in 1.9.5: 3109 3110- (Critical bug fix): Fix a bug in the FPU simulation. This was 3111 causing some floating point conditional tests not to work right. 3112 Several people reported this. If you had floating point code which 3113 didn't work right on 1.9.1 to 1.9.4, it's worth trying 1.9.5. 3114 3115- Partial support for Red Hat 9. RH9 uses the new Native Posix 3116 Threads Library (NPTL), instead of the older LinuxThreads. 3117 This potentially causes problems with V which will take some 3118 time to correct. In the meantime we have partially worked around 3119 this, and so 1.9.5 works on RH9. Threaded programs still work, 3120 but they may deadlock, because some system calls (accept, read, 3121 write, etc) which should be nonblocking, in fact do block. This 3122 is a known bug which we are looking into. 3123 3124 If you can, your best bet (unfortunately) is to avoid using 3125 1.9.5 on a Red Hat 9 system, or on any NPTL-based distribution. 3126 If your glibc is 2.3.1 or earlier, you're almost certainly OK. 3127 3128Minor changes in 1.9.5: 3129 3130- Added some #errors to valgrind.h to ensure people don't include 3131 it accidentally in their sources. This is a change from 1.0.X 3132 which was never properly documented. The right thing to include 3133 is now memcheck.h. Some people reported problems and strange 3134 behaviour when (incorrectly) including valgrind.h in code with 3135 1.9.1 -- 1.9.4. This is no longer possible. 3136 3137- Add some __extension__ bits and pieces so that gcc configured 3138 for valgrind-checking compiles even with -Werror. If you 3139 don't understand this, ignore it. Of interest to gcc developers 3140 only. 3141 3142- Removed a pointless check which caused problems interworking 3143 with Clearcase. V would complain about shared objects whose 3144 names did not end ".so", and refuse to run. This is now fixed. 3145 In fact it was fixed in 1.9.4 but not documented. 3146 3147- Fixed a bug causing an assertion failure of "waiters == 1" 3148 somewhere in vg_scheduler.c, when running large threaded apps, 3149 notably MySQL. 3150 3151- Add support for the munlock system call (124). 3152 3153Some comments about future releases: 3154 31551.9.5 is, we hope, the most stable Valgrind so far. It pretty much 3156supersedes the 1.0.X branch. If you are a valgrind packager, please 3157consider making 1.9.5 available to your users. You can regard the 31581.0.X branch as obsolete: 1.9.5 is stable and vastly superior. There 3159are no plans at all for further releases of the 1.0.X branch. 3160 3161If you want a leading-edge valgrind, consider building the cvs head 3162(from SourceForge), or getting a snapshot of it. Current cool stuff 3163going in includes MMX support (done); SSE/SSE2 support (in progress), 3164a significant (10-20%) performance improvement (done), and the usual 3165large collection of minor changes. Hopefully we will be able to 3166improve our NPTL support, but no promises. 3167 3168