blob: f43f8d374a0c5726d11fb1ac818ecea6eaa415a2 [file] [log] [blame]
njn718d3b12006-12-16 00:54:12 +00001
njne5930da2010-12-17 00:45:19 +00002Release 3.7.0 (???)
3~~~~~~~~~~~~~~~~~~~
sewardjdc873c02011-07-24 16:02:33 +00004
5- new client requests VALGRIND_{DISABLE,ENABLE}_ERROR_REPORTING
6
njne5930da2010-12-17 00:45:19 +00007- Added the --mod-funcname option to cg_diff.
sewardj4fd57292011-02-17 12:20:19 +00008- Further reduction in overheads caused by --smc-check=all, especially
9 on 64-bit targets.
sewardj6dbcc632011-06-07 21:39:28 +000010- new variant --smc-check=all-non-file
njne5930da2010-12-17 00:45:19 +000011
sewardj70ceabc2011-06-24 18:23:42 +000012- hg: performance improvements and memory use reductions, particularly
13 for large, long running applications which perform many synch events.
14
15 showing of locksets for both threads involved in a race
16
17 general improvement of formatting/clarity of error messages
18
19 add facilities and documentation regarding annotation of thread safe
20 reference counted C++ classes
21
22 new flag --check-stack-refs=no|yes [yes], to disable race checking
23 on thread stacks (performance hack)
24
25 new flag --free-is-write=no|yes [no], to enable detection of races
26 where one thread accesses heap memory but another one frees it,
27 without any coordinating synchronisation event
28
sewardj4629a732011-05-17 15:54:31 +000029* IBM z/Architecture (s390x) running Linux
30 Valgrind can analyse 64-bit programs running on z/Architecture.
31 Most user space instructions up to and including z10 are supported.
32 Valgrind has been tested extensively on z9, z10, and z196 machines
33 running SLES 10/11, RedHat 5/6m, and Fedora. The Memcheck and Massif
34 tools are known to work well. Callgrind, Helgrind, and DRD work
35 reasonably well on z9 and later models. See README.s390 for more
36 details.
37
sewardj7a9ff4c2011-06-13 13:36:59 +000038bugs fixed (last update 11 June 2011):
39
40* don't be spooked by libxul.so linked with gold (rXXXX)
41* don't be spooked by libraries mashed by elfhack# (rXXXX)
42* cachegrind/callgrind: handle CPUID information for Core iX Intel
43 CPUs that with non-power-of-2 sizes (also AMDs)
44
45243404 Port to zSeries
46Fixed 3.7
47
48265762 make public VEX headers compilable by G++ 3.x
49Fixed 3.7
50
51265771 assertion in jumps.c (r11523) fails with glibc-2.3
52Fixed 3.7
53
54266753 valgrind's configure script does not give the user the option
55 to not use QtCore
56fixed, apparently
57
58266931 gen_insn_test.pl is broken
59fixed
60
61266961 ld-linux.so.2 i?86-linux strlen issues
62fixed
63
64266990 setns instruction causes false positive
65fixed
66
67243935 Helgrind: implementation of ANNOTATE_HAPPENS_BEFORE() /
68 AFTER() is not correct
69fixed, r11624
70
71247223 non-x86: Suppress warning: 'regparm' attribute directive
72 ignored
73fixed
74
75
76267383 Assertion 'vgPlain_strlen(dir) + vgPlain_strlen(file) + 1 <
77 256' failed.
78fixed
79
80267413 Assertion 'DRD_(g_threadinfo)[tid].synchr_nesting >= 1'
81 failed.
82fixed
83
84210935 port valgrind.h (not valgrind) to win32 so apps run under
85 wine can make client requests
86afaict, this was fixed in 3.6.1 but is not listed in NEWS
87
88267488 regtest: darwin support for 64-bit build
89fixed
90
91267552 SIGSEGV (misaligned_stack_error) with DRD, but not with other
92 tools
93fixed, but is the next one also fixed?
94
95267630 Add support for IBM Power ISA 2.06 -- stage 1
96fixed
97
98267819 Add client request for informing the core about reallocation
99fixed
100
101267968 drd: drd_thread.c:567 (vgDrd_thread_set_joinable): Assertion
102 '0 <= (int)tid && tid < DRD_N_THREADS && tid != DRD_INVALID_THREADID'
103 failed.
104fixed
105
106214223 valgrind SIGSEGV on startup gcc 4.4.1 ppc32 (G4) Ubuntu 9.10 ==
107259977 Valgrind segfaults doing __builtin_longjmp
108fixed
109
110268792 - valgrind seg faults on startup when compiled with Xcode 4 compilers...
111267769 - Darwin: memcheck triggers segmentation fault
112274784 - valgrind ls -l or any other valgrind call(even without parameters) results in Segmentation Fault
113267342 - segmentation fault on Mac OS 10.6
114271337 - Valgrind segfaults on MacOS X
115270309 - valgrind crash on startup
116269641 - valgrind segfaults immediately (segmentation fault)
117267997 MacOSX: 64-bit valgrind segfaults on launch when built with
118 Xcode 4.0.1
119fixed
120
121264800 testcase compile failure on zseries
122fixed
123
124265762 - make public VEX headers compilable by G++ 3.x
125fixed
126
127268513] New: missed optimizations in fold_Expr
128fixed
129
130253206 - Some fixes for the faultstatus testcase
131fixed
132
133268619 - s390x: fpr - gpr transfer facility
134fixed
135
136268620 - s390x: reconsider "long displacement" requirement
137fixed
138
139268621 - s390x: improve IR generation for XC
140fixed
141
142255223 - [PATCH] capget testcase fails when running as root
143fixed
144
145268715 - s390x: FLOGR is not universally available
146fixed
147
148268930 - s390x: MHY is not universally available
149fixed
150
151269078 - [PATCH] vex: arm->IR: unhandled instruction SUB (SP minus
152immediate/register)
153fixed
154
155269079 - [PATCH] Support ptrace system call on ARM
156fixed
157
158269144 - missing "Bad option" error message
159fixed
160
161269209] New: [PATCH] conditional load and store facility (z196)
162fixed
163
164269354] New: Shift by zero on x86 can incorrectly clobber CC_NDEP
165(with patch)
166fixed
167
168256726 - Helgrind tests have broken inline asm
169fixed
170
171269736 - s390x: minor code generation tweaks
172fixed
173
174256703 - xlc_dbl_u32.c testcase broken
175fixed
176
177272986 - gcc-4.6 warnings with valgrind.h ==
178269778] New: valgrind.h: swap roles of VALGRIND_DO_CLIENT_REQUEST()
179and VALGRIND_DO_CLIENT_REQUEST_EXPR()
180fixed
181
182269863 - s390x: remove unused function parameters
183fixed
184
185269864 - s390x: tweak s390_emit_load_cc
186fixed
187
188270115] New: s390x: rewrite some testcases
189fixed
190
191270082 - s390x: [PATCH] Make sure to point the PSW address to the next
192address on SIGILL
193fixed
194
195270794 - New IBM POWER7 support patch causes regression in none/tests
196fixed
197
198270851 - New IBM POWER7 fcfidus instruction causes memcheck to fail
199fixed
200
201270856 - New IBM POWER7 xsnmaddadp instruction causes memcheck to fail
202on 32bit app
203fixed
204
205270959 - s390x: invalid use of R0 as base register
206fixed
207
208271042 - VSX configure check fails when it should not
209fixed
210
211271043 - Valgrind build fails with assembler error on ppc64 with
212binutils 2.21
213fixed
214
215271259 - s390x: fix code confusion
216fixed
217
218271385 - s390x: Implement Ist_MBE
219fixed
220
221271501 - s390x : misc cleanups
222fixed
223
224271504 - s390x: promote likely and unlikely
225fixed
226
227271579 - ppc: using wrong enum type
228fixed
229
230271730 - [PATCH] Fix bug when checking ioctls: duplicate check
231fixed
232
233271779 - s390x: provide clock instructions like STCK
234fixed
235
236271799 - Darwin: ioctls without an arg report a memory error
237fixed
238
239271820 - arm: fix type confusion
240fixed
241
242272067 - s390x: fix DISP20 macro
243fixed
244
245272615 - A typo in debug output in mc_leakcheck.c
246fixed
247
248272661 - callgrind_annotate chokes when run from paths containing
249regex metacharacters
250fixed
251
252272955 - Unhandled syscall error for pwrite64 on ppc64 arch
253fixed
254
255274447] New: WARNING: unhandled syscall: 340
256fixed
257
258275148] New: configure FAIL with glibc-2.14
259fixed
260
261275151] New: Fedora 15 / glibc-2.14 'make regtest' FAIL
262fixed
263
florianeb67dbf2011-06-25 02:06:21 +0000264275339 - s390x: fix testcase compile warnings
sewardj7a9ff4c2011-06-13 13:36:59 +0000265
florian3a58b892011-06-25 02:28:00 +0000266275710 - s390x: get rid of redundant address mode calculation
267
florian036113b2011-07-11 01:51:39 +0000268271776 - s390x: Support STFLE instruction
269
florian1763e812011-07-12 19:07:05 +0000270267020 - Make directory for temporary files configurable at run-time.
271
njne5930da2010-12-17 00:45:19 +0000272
sewardj4fd57292011-02-17 12:20:19 +0000273Release 3.6.1 (16 February 2011)
274~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2753.6.1 is a bug fix release. It adds support for some SSE4
276instructions that were omitted in 3.6.0 due to lack of time. Initial
277support for glibc-2.13 has been added. A number of bugs causing
278crashing or assertion failures have been fixed.
njne5930da2010-12-17 00:45:19 +0000279
sewardj4fd57292011-02-17 12:20:19 +0000280The following bugs have been fixed or resolved. Note that "n-i-bz"
281stands for "not in bugzilla" -- that is, a bug that was reported to us
282but never got a bugzilla entry. We encourage you to file bugs in
283bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
284mailing the developers (or mailing lists) directly -- bugs that are
285not entered into bugzilla tend to get forgotten about or ignored.
bart8e0d2c12011-02-04 19:07:11 +0000286
sewardj4fd57292011-02-17 12:20:19 +0000287To see details of a given bug, visit
288https://bugs.kde.org/show_bug.cgi?id=XXXXXX
289where XXXXXX is the bug number as listed below.
290
291188572 Valgrind on Mac should suppress setenv() mem leak
292194402 vex amd64->IR: 0x48 0xF 0xAE 0x4 (proper FX{SAVE,RSTOR} support)
293210481 vex amd64->IR: Assertion `sz == 2 || sz == 4' failed (REX.W POPQ)
294246152 callgrind internal error after pthread_cancel on 32 Bit Linux
295250038 ppc64: Altivec LVSR and LVSL instructions fail their regtest
296254420 memory pool tracking broken
297254957 Test code failing to compile due to changes in memcheck.h
298255009 helgrind/drd: crash on chmod with invalid parameter
299255130 readdwarf3.c parse_type_DIE confused by GNAT Ada types
300255355 helgrind/drd: crash on threaded programs doing fork
301255358 == 255355
302255418 (SSE4.x) rint call compiled with ICC
303255822 --gen-suppressions can create invalid files: "too many callers [...]"
304255888 closing valgrindoutput tag outputted to log-stream on error
305255963 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x9 0xDB 0x0 (ROUNDPD)
306255966 Slowness when using mempool annotations
307256387 vex x86->IR: 0xD4 0xA 0x2 0x7 (AAD and AAM)
308256600 super-optimized strcasecmp() false positive
309256669 vex amd64->IR: Unhandled LOOPNEL insn on amd64
310256968 (SSE4.x) vex amd64->IR: 0x66 0xF 0x38 0x10 0xD3 0x66 (BLENDVPx)
311257011 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0xE 0xFD 0xA0 (PBLENDW)
312257063 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x8 0xC0 0x0 (ROUNDPS)
313257276 Missing case in memcheck --track-origins=yes
314258870 (SSE4.x) Add support for EXTRACTPS SSE 4.1 instruction
315261966 (SSE4.x) support for CRC32B and CRC32Q is lacking (also CRC32{W,L})
316262985 VEX regression in valgrind 3.6.0 in handling PowerPC VMX
317262995 (SSE4.x) crash when trying to valgrind gcc-snapshot (PCMPxSTRx $0)
318263099 callgrind_annotate counts Ir improperly [...]
319263877 undefined coprocessor instruction on ARMv7
320265964 configure FAIL with glibc-2.13
321n-i-bz Fix compile error w/ icc-12.x in guest_arm_toIR.c
322n-i-bz Docs: fix bogus descriptions for VALGRIND_CREATE_BLOCK et al
323n-i-bz Massif: don't assert on shmat() with --pages-as-heap=yes
324n-i-bz Bug fixes and major speedups for the exp-DHAT space profiler
325n-i-bz DRD: disable --free-is-write due to implementation difficulties
326
327(3.6.1: 16 February 2011, vex r2103, valgrind r11561).
328
bart8e0d2c12011-02-04 19:07:11 +0000329
330
sewardj695f13a2010-10-20 17:07:01 +0000331Release 3.6.0 (21 October 2010)
sewardj18a08922010-10-14 00:26:42 +0000332~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3333.6.0 is a feature release with many significant improvements and the
334usual collection of bug fixes.
sewardj00d9d1e2010-10-13 14:05:30 +0000335
sewardj18a08922010-10-14 00:26:42 +0000336This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux,
337PPC64/Linux, X86/Darwin and AMD64/Darwin. Support for recent distros
338and toolchain components (glibc 2.12, gcc 4.5, OSX 10.6) has been added.
sewardj00d9d1e2010-10-13 14:05:30 +0000339
sewardj18a08922010-10-14 00:26:42 +0000340 -------------------------
sewardj00d9d1e2010-10-13 14:05:30 +0000341
sewardj18a08922010-10-14 00:26:42 +0000342Here are some highlights. Details are shown further down:
njn9b919b62010-02-28 23:41:07 +0000343
sewardj18a08922010-10-14 00:26:42 +0000344* Support for ARM/Linux.
njn9b919b62010-02-28 23:41:07 +0000345
sewardj18a08922010-10-14 00:26:42 +0000346* Support for recent Linux distros: Ubuntu 10.10 and Fedora 14.
347
348* Support for Mac OS X 10.6, both 32- and 64-bit executables.
349
350* Support for the SSE4.2 instruction set.
351
352* Enhancements to the Callgrind profiler, including the ability to
353 handle CPUs with three levels of cache.
354
355* A new experimental heap profiler, DHAT.
356
sewardj695f13a2010-10-20 17:07:01 +0000357* A huge number of bug fixes and small enhancements.
sewardj18a08922010-10-14 00:26:42 +0000358
359 -------------------------
360
361Here are details of the above changes, together with descriptions of
362many other changes, and a list of fixed bugs.
363
364* ================== PLATFORM CHANGES =================
365
366* Support for ARM/Linux. Valgrind now runs on ARMv7 capable CPUs
367 running Linux. It is known to work on Ubuntu 10.04, Ubuntu 10.10,
368 and Maemo 5, so you can run Valgrind on your Nokia N900 if you want.
369
370 This requires a CPU capable of running the ARMv7-A instruction set
371 (Cortex A5, A8 and A9). Valgrind provides fairly complete coverage
372 of the user space instruction set, including ARM and Thumb integer
373 code, VFPv3, NEON and V6 media instructions. The Memcheck,
374 Cachegrind and Massif tools work properly; other tools work to
375 varying degrees.
376
377* Support for recent Linux distros (Ubuntu 10.10 and Fedora 14), along
378 with support for recent releases of the underlying toolchain
379 components, notably gcc-4.5 and glibc-2.12.
380
381* Support for Mac OS X 10.6, both 32- and 64-bit executables. 64-bit
382 support also works much better on OS X 10.5, and is as solid as
383 32-bit support now.
384
385* Support for the SSE4.2 instruction set. SSE4.2 is supported in
386 64-bit mode. In 32-bit mode, support is only available up to and
387 including SSSE3. Some exceptions: SSE4.2 AES instructions are not
388 supported in 64-bit mode, and 32-bit mode does in fact support the
sewardj695f13a2010-10-20 17:07:01 +0000389 bare minimum SSE4 instructions to needed to run programs on Mac OS X
sewardj18a08922010-10-14 00:26:42 +0000390 10.6 on 32-bit targets.
391
392* Support for IBM POWER6 cpus has been improved. The Power ISA up to
393 and including version 2.05 is supported.
394
395* ==================== TOOL CHANGES ====================
396
397* Cachegrind has a new processing script, cg_diff, which finds the
398 difference between two profiles. It's very useful for evaluating
399 the performance effects of a change in a program.
njn69d495d2010-06-30 05:23:34 +0000400
401 Related to this change, the meaning of cg_annotate's (rarely-used)
sewardj18a08922010-10-14 00:26:42 +0000402 --threshold option has changed; this is unlikely to affect many
403 people, if you do use it please see the user manual for details.
njn9b919b62010-02-28 23:41:07 +0000404
sewardj18a08922010-10-14 00:26:42 +0000405* Callgrind now can do branch prediction simulation, similar to
406 Cachegrind. In addition, it optionally can count the number of
407 executed global bus events. Both can be used for a better
408 approximation of a "Cycle Estimation" as derived event (you need to
409 update the event formula in KCachegrind yourself).
njn2d853a12010-10-06 22:46:31 +0000410
sewardj18a08922010-10-14 00:26:42 +0000411* Cachegrind and Callgrind now refer to the LL (last-level) cache
412 rather than the L2 cache. This is to accommodate machines with
413 three levels of caches -- if Cachegrind/Callgrind auto-detects the
414 cache configuration of such a machine it will run the simulation as
415 if the L2 cache isn't present. This means the results are less
416 likely to match the true result for the machine, but
417 Cachegrind/Callgrind's results are already only approximate, and
418 should not be considered authoritative. The results are still
419 useful for giving a general idea about a program's locality.
njn2d853a12010-10-06 22:46:31 +0000420
sewardj695f13a2010-10-20 17:07:01 +0000421* Massif has a new option, --pages-as-heap, which is disabled by
sewardj18a08922010-10-14 00:26:42 +0000422 default. When enabled, instead of tracking allocations at the level
423 of heap blocks (as allocated with malloc/new/new[]), it instead
424 tracks memory allocations at the level of memory pages (as mapped by
425 mmap, brk, etc). Each mapped page is treated as its own block.
426 Interpreting the page-level output is harder than the heap-level
427 output, but this option is useful if you want to account for every
428 byte of memory used by a program.
njne323a6b2010-07-01 02:35:03 +0000429
sewardj18a08922010-10-14 00:26:42 +0000430* DRD has two new command-line options: --free-is-write and
431 --trace-alloc. The former allows to detect reading from already freed
432 memory, and the latter allows tracing of all memory allocations and
bart5cf4c392010-10-10 18:54:44 +0000433 deallocations.
434
sewardj18a08922010-10-14 00:26:42 +0000435* DRD has several new annotations. Custom barrier implementations can
436 now be annotated, as well as benign races on static variables.
bart5cf4c392010-10-10 18:54:44 +0000437
sewardj18a08922010-10-14 00:26:42 +0000438* DRD's happens before / happens after annotations have been made more
439 powerful, so that they can now also be used to annotate e.g. a smart
bart5cf4c392010-10-10 18:54:44 +0000440 pointer implementation.
441
sewardj18a08922010-10-14 00:26:42 +0000442* Helgrind's annotation set has also been drastically improved, so as
sewardj695f13a2010-10-20 17:07:01 +0000443 to provide to users a general set of annotations to describe locks,
sewardj18a08922010-10-14 00:26:42 +0000444 semaphores, barriers and condition variables. Annotations to
445 describe thread-safe reference counted heap objects have also been
446 added.
447
448* Memcheck has a new command-line option, --show-possibly-lost, which
449 is enabled by default. When disabled, the leak detector will not
450 show possibly-lost blocks.
451
452* A new experimental heap profiler, DHAT (Dynamic Heap Analysis Tool),
453 has been added. DHAT keeps track of allocated heap blocks, and also
454 inspects every memory reference to see which block (if any) is being
455 accessed. This gives a lot of insight into block lifetimes,
456 utilisation, turnover, liveness, and the location of hot and cold
457 fields. You can use DHAT to do hot-field profiling.
458
459* ==================== OTHER CHANGES ====================
460
461* Improved support for unfriendly self-modifying code: the extra
462 overhead incurred by --smc-check=all has been reduced by
463 approximately a factor of 5 as compared with 3.5.0.
464
465* Ability to show directory names for source files in error messages.
sewardj695f13a2010-10-20 17:07:01 +0000466 This is combined with a flexible mechanism for specifying which
467 parts of the paths should be shown. This is enabled by the new flag
468 --fullpath-after.
sewardj18a08922010-10-14 00:26:42 +0000469
470* A new flag, --require-text-symbol, which will stop the run if a
471 specified symbol is not found it a given shared object when it is
sewardj695f13a2010-10-20 17:07:01 +0000472 loaded into the process. This makes advanced working with function
473 intercepting and wrapping safer and more reliable.
sewardj18a08922010-10-14 00:26:42 +0000474
sewardj695f13a2010-10-20 17:07:01 +0000475* Improved support for the Valkyrie GUI, version 2.0.0. GUI output
476 and control of Valgrind is now available for the tools Memcheck and
477 Helgrind. XML output from Valgrind is available for Memcheck,
478 Helgrind and exp-Ptrcheck.
sewardj18a08922010-10-14 00:26:42 +0000479
480* More reliable stack unwinding on amd64-linux, particularly in the
481 presence of function wrappers, and with gcc-4.5 compiled code.
482
483* Modest scalability (performance improvements) for massive
484 long-running applications, particularly for those with huge amounts
485 of code.
486
487* Support for analyzing programs running under Wine with has been
488 improved. The header files <valgrind/valgrind.h>,
489 <valgrind/memcheck.h> and <valgrind/drd.h> can now be used in
490 Windows-programs compiled with MinGW or one of the Microsoft Visual
491 Studio compilers.
492
493* A rare but serious error in the 64-bit x86 CPU simulation was fixed.
494 The 32-bit simulator was not affected. This did not occur often,
495 but when it did would usually crash the program under test.
496 Bug 245925.
497
498* A large number of bugs were fixed. These are shown below.
499
500* A number of bugs were investigated, and were candidates for fixing,
501 but are not fixed in 3.6.0, due to lack of developer time. They may
502 get fixed in later releases. They are:
503
504 194402 vex amd64->IR: 0x48 0xF 0xAE 0x4 0x24 0x49 (FXSAVE64)
505 212419 false positive "lock order violated" (A+B vs A)
506 213685 Undefined value propagates past dependency breaking instruction
507 216837 Incorrect instrumentation of NSOperationQueue on Darwin
508 237920 valgrind segfault on fork failure
509 242137 support for code compiled by LLVM-2.8
510 242423 Another unknown Intel cache config value
511 243232 Inconsistent Lock Orderings report with trylock
512 243483 ppc: callgrind triggers VEX assertion failure
513 243935 Helgrind: implementation of ANNOTATE_HAPPENS_BEFORE() is wrong
514 244677 Helgrind crash hg_main.c:616 (map_threads_lookup): Assertion
515 'thr' failed.
516 246152 callgrind internal error after pthread_cancel on 32 Bit Linux
517 249435 Analyzing wine programs with callgrind triggers a crash
518 250038 ppc64: Altivec lvsr and lvsl instructions fail their regtest
519 250065 Handling large allocations
520 250101 huge "free" memory usage due to m_mallocfree.c
521 "superblocks fragmentation"
522 251569 vex amd64->IR: 0xF 0x1 0xF9 0x8B 0x4C 0x24 (RDTSCP)
sewardj695f13a2010-10-20 17:07:01 +0000523 252091 Callgrind on ARM does not detect function returns correctly
524 252600 [PATCH] Allow lhs to be a pointer for shl/shr
525 254420 memory pool tracking broken
sewardj18a08922010-10-14 00:26:42 +0000526 n-i-bz support for adding symbols for JIT generated code
527
528
529The following bugs have been fixed or resolved. Note that "n-i-bz"
530stands for "not in bugzilla" -- that is, a bug that was reported to us
531but never got a bugzilla entry. We encourage you to file bugs in
532bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
533mailing the developers (or mailing lists) directly -- bugs that are
534not entered into bugzilla tend to get forgotten about or ignored.
535
536To see details of a given bug, visit
537https://bugs.kde.org/show_bug.cgi?id=XXXXXX
538where XXXXXX is the bug number as listed below.
539
540135264 dcbzl instruction missing
541142688 == 250799
542153699 Valgrind should report unaligned reads with movdqa
543180217 == 212335
544190429 Valgrind reports lost of errors in ld.so
545 with x86_64 2.9.90 glibc
546197266 valgrind appears to choke on the xmms instruction
547 "roundsd" on x86_64
548197988 Crash when demangling very large symbol names
549202315 unhandled syscall: 332 (inotify_init1)
550203256 Add page-level profiling to Massif
551205093 dsymutil=yes needs quotes, locking (partial fix)
552205241 Snow Leopard 10.6 support (partial fix)
553206600 Leak checker fails to upgrade indirect blocks when their
554 parent becomes reachable
555210935 port valgrind.h (not valgrind) to win32 so apps run under
556 wine can make client requests
557211410 vex amd64->IR: 0x15 0xFF 0xFF 0x0 0x0 0x89
558 within Linux ip-stack checksum functions
559212335 unhandled instruction bytes: 0xF3 0xF 0xBD 0xC0
560 (lzcnt %eax,%eax)
561213685 Undefined value propagates past dependency breaking instruction
562 (partial fix)
563215914 Valgrind inserts bogus empty environment variable
564217863 == 197988
565219538 adjtimex syscall wrapper wrong in readonly adjtime mode
566222545 shmat fails under valgind on some arm targets
567222560 ARM NEON support
568230407 == 202315
569231076 == 202315
570232509 Docs build fails with formatting inside <title></title> elements
571232793 == 202315
572235642 [PATCH] syswrap-linux.c: support evdev EVIOCG* ioctls
573236546 vex x86->IR: 0x66 0xF 0x3A 0xA
574237202 vex amd64->IR: 0xF3 0xF 0xB8 0xC0 0x49 0x3B
575237371 better support for VALGRIND_MALLOCLIKE_BLOCK
576237485 symlink (syscall 57) is not supported on Mac OS
577237723 sysno == 101 exp-ptrcheck: the 'impossible' happened:
578 unhandled syscall
579238208 is_just_below_ESP doesn't take into account red-zone
580238345 valgrind passes wrong $0 when executing a shell script
581238679 mq_timedreceive syscall doesn't flag the reception buffer
582 as "defined"
583238696 fcntl command F_DUPFD_CLOEXEC not supported
584238713 unhandled instruction bytes: 0x66 0xF 0x29 0xC6
585238713 unhandled instruction bytes: 0x66 0xF 0x29 0xC6
586238745 3.5.0 Make fails on PPC Altivec opcodes, though configure
587 says "Altivec off"
588239992 vex amd64->IR: 0x48 0xF 0xC4 0xC1 0x0 0x48
589240488 == 197988
590240639 == 212335
591241377 == 236546
592241903 == 202315
593241920 == 212335
594242606 unhandled syscall: setegid (in Ptrcheck)
595242814 Helgrind "Impossible has happened" during
596 QApplication::initInstance();
597243064 Valgrind attempting to read debug information from iso
598243270 Make stack unwinding in Valgrind wrappers more reliable
599243884 exp-ptrcheck: the 'impossible happened: unhandled syscall
600 sysno = 277 (mq_open)
601244009 exp-ptrcheck unknown syscalls in analyzing lighttpd
602244493 ARM VFP d16-d31 registers support
603244670 add support for audit_session_self syscall on Mac OS 10.6
604244921 The xml report of helgrind tool is not well format
605244923 In the xml report file, the <preamble> not escape the
606 xml char, eg '<','&','>'
607245535 print full path names in plain text reports
608245925 x86-64 red zone handling problem
609246258 Valgrind not catching integer underruns + new [] s
610246311 reg/reg cmpxchg doesn't work on amd64
611246549 unhandled syscall unix:277 while testing 32-bit Darwin app
612246888 Improve Makefile.vex.am
613247510 [OS X 10.6] Memcheck reports unaddressable bytes passed
614 to [f]chmod_extended
615247526 IBM POWER6 (ISA 2.05) support is incomplete
616247561 Some leak testcases fails due to reachable addresses in
617 caller save regs
618247875 sizeofIRType to handle Ity_I128
619247894 [PATCH] unhandled syscall sys_readahead
620247980 Doesn't honor CFLAGS passed to configure
621248373 darwin10.supp is empty in the trunk
622248822 Linux FIBMAP ioctl has int parameter instead of long
623248893 [PATCH] make readdwarf.c big endianess safe to enable
624 unwinding on big endian systems
625249224 Syscall 336 not supported (SYS_proc_info)
626249359 == 245535
627249775 Incorrect scheme for detecting NEON capabilities of host CPU
628249943 jni JVM init fails when using valgrind
629249991 Valgrind incorrectly declares AESKEYGENASSIST support
630 since VEX r2011
631249996 linux/arm: unhandled syscall: 181 (__NR_pwrite64)
632250799 frexp$fenv_access_off function generates SIGILL
633250998 vex x86->IR: unhandled instruction bytes: 0x66 0x66 0x66 0x2E
634251251 support pclmulqdq insn
635251362 valgrind: ARM: attach to debugger either fails or provokes
636 kernel oops
637251674 Unhandled syscall 294
sewardj695f13a2010-10-20 17:07:01 +0000638251818 == 254550
sewardj18a08922010-10-14 00:26:42 +0000639
sewardj695f13a2010-10-20 17:07:01 +0000640254257 Add support for debugfiles found by build-id
641254550 [PATCH] Implement DW_ATE_UTF (DWARF4)
642254646 Wrapped functions cause stack misalignment on OS X
643 (and possibly Linux)
644254556 ARM: valgrinding anything fails with SIGSEGV for 0xFFFF0FA0
645
sewardje07c2b52010-10-22 06:44:59 +0000646(3.6.0: 21 October 2010, vex r2068, valgrind r11471).
sewardj18a08922010-10-14 00:26:42 +0000647
648
njn9b919b62010-02-28 23:41:07 +0000649
sewardj6bbe0242009-08-19 11:38:54 +0000650Release 3.5.0 (19 August 2009)
651~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardj570e2442009-08-14 13:42:41 +00006523.5.0 is a feature release with many significant improvements and the
653usual collection of bug fixes. The main improvement is that Valgrind
654now works on Mac OS X.
655
656This release supports X86/Linux, AMD64/Linux, PPC32/Linux, PPC64/Linux
657and X86/Darwin. Support for recent distros and toolchain components
658(glibc 2.10, gcc 4.5) has been added.
659
660 -------------------------
661
662Here is a short summary of the changes. Details are shown further
663down:
664
njnf95c30b2009-08-16 00:20:58 +0000665* Support for Mac OS X (10.5.x).
sewardj570e2442009-08-14 13:42:41 +0000666
njn78b7bb32009-08-14 22:34:30 +0000667* Improvements and simplifications to Memcheck's leak checker.
sewardj570e2442009-08-14 13:42:41 +0000668
njn78b7bb32009-08-14 22:34:30 +0000669* Clarification and simplifications in various aspects of Valgrind's
670 text output.
sewardj570e2442009-08-14 13:42:41 +0000671
njn78b7bb32009-08-14 22:34:30 +0000672* XML output for Helgrind and Ptrcheck.
sewardj570e2442009-08-14 13:42:41 +0000673
njn78b7bb32009-08-14 22:34:30 +0000674* Performance and stability improvements for Helgrind and DRD.
sewardj570e2442009-08-14 13:42:41 +0000675
njn78b7bb32009-08-14 22:34:30 +0000676* Genuinely atomic support for x86/amd64/ppc atomic instructions.
sewardj570e2442009-08-14 13:42:41 +0000677
njn78b7bb32009-08-14 22:34:30 +0000678* A new experimental tool, BBV, useful for computer architecture
679 research.
sewardj570e2442009-08-14 13:42:41 +0000680
njn78b7bb32009-08-14 22:34:30 +0000681* Improved Wine support, including ability to read Windows PDB
682 debuginfo.
sewardj570e2442009-08-14 13:42:41 +0000683
684 -------------------------
685
sewardj9176c862009-08-18 13:47:31 +0000686Here are details of the above changes, followed by descriptions of
687many other minor changes, and a list of fixed bugs.
sewardj570e2442009-08-14 13:42:41 +0000688
689
690* Valgrind now runs on Mac OS X. (Note that Mac OS X is sometimes
njnf95c30b2009-08-16 00:20:58 +0000691 called "Darwin" because that is the name of the OS core, which is the
692 level that Valgrind works at.)
njnf76d27a2009-05-28 01:53:07 +0000693
njnf95c30b2009-08-16 00:20:58 +0000694 Supported systems:
njn6eccab92009-08-14 02:41:37 +0000695
njnf95c30b2009-08-16 00:20:58 +0000696 - It requires OS 10.5.x (Leopard). Porting to 10.4.x is not planned
697 because it would require work and 10.4 is only becoming less common.
njn6eccab92009-08-14 02:41:37 +0000698
njnf95c30b2009-08-16 00:20:58 +0000699 - 32-bit programs on x86 and AMD64 (a.k.a x86-64) machines are supported
700 fairly well. For 10.5.x, 32-bit programs are the default even on
701 64-bit machines, so it handles most current programs.
702
703 - 64-bit programs on x86 and AMD64 (a.k.a x86-64) machines are not
704 officially supported, but simple programs at least will probably work.
705 However, start-up is slow.
njn6eccab92009-08-14 02:41:37 +0000706
sewardj570e2442009-08-14 13:42:41 +0000707 - PowerPC machines are not supported.
njn6eccab92009-08-14 02:41:37 +0000708
njnc6ab7b62009-06-04 22:47:21 +0000709 Things that don't work:
njn6eccab92009-08-14 02:41:37 +0000710
711 - The Ptrcheck tool.
712
713 - Objective-C garbage collection.
714
715 - --db-attach=yes.
716
sewardj570e2442009-08-14 13:42:41 +0000717 - If you have Rogue Amoeba's "Instant Hijack" program installed,
718 Valgrind will fail with a SIGTRAP at start-up. See
719 https://bugs.kde.org/show_bug.cgi?id=193917 for details and a
720 simple work-around.
njnc6ab7b62009-06-04 22:47:21 +0000721
njn97db7612009-08-04 02:32:55 +0000722 Usage notes:
njn6eccab92009-08-14 02:41:37 +0000723
sewardj570e2442009-08-14 13:42:41 +0000724 - You will likely find --dsymutil=yes a useful option, as error
725 messages may be imprecise without it.
njn6eccab92009-08-14 02:41:37 +0000726
sewardj570e2442009-08-14 13:42:41 +0000727 - Mac OS X support is new and therefore will be less robust than the
njn1205e022009-08-14 02:12:54 +0000728 Linux support. Please report any bugs you find.
njn97db7612009-08-04 02:32:55 +0000729
sewardj570e2442009-08-14 13:42:41 +0000730 - Threaded programs may run more slowly than on Linux.
731
njnc6ab7b62009-06-04 22:47:21 +0000732 Many thanks to Greg Parker for developing this port over several years.
733
sewardj570e2442009-08-14 13:42:41 +0000734
735* Memcheck's leak checker has been improved.
736
737 - The results for --leak-check=summary now match the summary results
738 for --leak-check=full. Previously they could differ because
739 --leak-check=summary counted "indirectly lost" blocks and
740 "suppressed" blocks as "definitely lost".
741
742 - Blocks that are only reachable via at least one interior-pointer,
743 but are directly pointed to by a start-pointer, were previously
744 marked as "still reachable". They are now correctly marked as
745 "possibly lost".
746
747 - The default value for the --leak-resolution option has been
748 changed from "low" to "high". In general, this means that more
749 leak reports will be produced, but each leak report will describe
750 fewer leaked blocks.
751
752 - With --leak-check=full, "definitely lost" and "possibly lost"
753 leaks are now considered as proper errors, ie. they are counted
754 for the "ERROR SUMMARY" and affect the behaviour of
755 --error-exitcode. These leaks are not counted as errors if
756 --leak-check=summary is specified, however.
757
758 - Documentation for the leak checker has been improved.
759
njnf76d27a2009-05-28 01:53:07 +0000760
njn6eccab92009-08-14 02:41:37 +0000761* Various aspects of Valgrind's text output have changed.
njndbebecc2009-07-14 01:39:54 +0000762
njn6eccab92009-08-14 02:41:37 +0000763 - Valgrind's start-up message has changed. It is shorter but also
764 includes the command being run, which makes it easier to use
765 --trace-children=yes. An example:
njn53162bf2009-07-29 23:34:49 +0000766
njn6eccab92009-08-14 02:41:37 +0000767 - Valgrind's shut-down messages have also changed. This is most
sewardj570e2442009-08-14 13:42:41 +0000768 noticeable with Memcheck, where the leak summary now occurs before
769 the error summary. This change was necessary to allow leaks to be
770 counted as proper errors (see the description of the leak checker
771 changes above for more details). This was also necessary to fix a
772 longstanding bug in which uses of suppressions against leaks were
773 not "counted", leading to difficulties in maintaining suppression
sewardj783e9d62009-08-19 22:18:35 +0000774 files (see https://bugs.kde.org/show_bug.cgi?id=186790).
njn6eccab92009-08-14 02:41:37 +0000775
sewardj570e2442009-08-14 13:42:41 +0000776 - Behavior of -v has changed. In previous versions, -v printed out
777 a mixture of marginally-user-useful information, and tool/core
778 statistics. The statistics printing has now been moved to its own
779 flag, --stats=yes. This means -v is less verbose and more likely
780 to convey useful end-user information.
njn6eccab92009-08-14 02:41:37 +0000781
sewardj570e2442009-08-14 13:42:41 +0000782 - The format of some (non-XML) stack trace entries has changed a
783 little. Previously there were six possible forms:
njn6eccab92009-08-14 02:41:37 +0000784
785 0x80483BF: really (a.c:20)
786 0x80483BF: really (in /foo/a.out)
787 0x80483BF: really
788 0x80483BF: (within /foo/a.out)
789 0x80483BF: ??? (a.c:20)
790 0x80483BF: ???
791
sewardj570e2442009-08-14 13:42:41 +0000792 The third and fourth of these forms have been made more consistent
793 with the others. The six possible forms are now:
njn6eccab92009-08-14 02:41:37 +0000794
795 0x80483BF: really (a.c:20)
796 0x80483BF: really (in /foo/a.out)
797 0x80483BF: really (in ???)
798 0x80483BF: ??? (in /foo/a.out)
799 0x80483BF: ??? (a.c:20)
800 0x80483BF: ???
801
sewardj570e2442009-08-14 13:42:41 +0000802 Stack traces produced when --xml=yes is specified are different
803 and unchanged.
njn6eccab92009-08-14 02:41:37 +0000804
njn6eccab92009-08-14 02:41:37 +0000805
sewardj570e2442009-08-14 13:42:41 +0000806* Helgrind and Ptrcheck now support XML output, so they can be used
807 from GUI tools. Also, the XML output mechanism has been
808 overhauled.
njn6eccab92009-08-14 02:41:37 +0000809
njn78b7bb32009-08-14 22:34:30 +0000810 - The XML format has been overhauled and generalised, so it is more
sewardj570e2442009-08-14 13:42:41 +0000811 suitable for error reporting tools in general. The Memcheck
812 specific aspects of it have been removed. The new format, which
813 is an evolution of the old format, is described in
814 docs/internals/xml-output-protocol4.txt.
njn6eccab92009-08-14 02:41:37 +0000815
sewardj570e2442009-08-14 13:42:41 +0000816 - Memcheck has been updated to use the new format.
njn6eccab92009-08-14 02:41:37 +0000817
sewardj570e2442009-08-14 13:42:41 +0000818 - Helgrind and Ptrcheck are now able to emit output in this format.
njn6eccab92009-08-14 02:41:37 +0000819
sewardj570e2442009-08-14 13:42:41 +0000820 - The XML output mechanism has been overhauled. XML is now output
821 to its own file descriptor, which means that:
njn6eccab92009-08-14 02:41:37 +0000822
sewardj570e2442009-08-14 13:42:41 +0000823 * Valgrind can output text and XML independently.
njnb4e06872009-08-10 05:11:43 +0000824
sewardj570e2442009-08-14 13:42:41 +0000825 * The longstanding problem of XML output being corrupted by
826 unexpected un-tagged text messages is solved.
njn8df80b22009-03-02 05:11:06 +0000827
sewardj570e2442009-08-14 13:42:41 +0000828 As before, the destination for text output is specified using
829 --log-file=, --log-fd= or --log-socket=.
njnd55f0d92009-08-03 01:38:56 +0000830
sewardj570e2442009-08-14 13:42:41 +0000831 As before, XML output for a tool is enabled using --xml=yes.
njnc6ab7b62009-06-04 22:47:21 +0000832
sewardj570e2442009-08-14 13:42:41 +0000833 Because there's a new XML output channel, the XML output
834 destination is now specified by --xml-file=, --xml-fd= or
835 --xml-socket=.
njn6eccab92009-08-14 02:41:37 +0000836
sewardj570e2442009-08-14 13:42:41 +0000837 Initial feedback has shown this causes some confusion. To
838 clarify, the two envisaged usage scenarios are:
njn6eccab92009-08-14 02:41:37 +0000839
sewardj570e2442009-08-14 13:42:41 +0000840 (1) Normal text output. In this case, do not specify --xml=yes
841 nor any of --xml-file=, --xml-fd= or --xml-socket=.
842
843 (2) XML output. In this case, specify --xml=yes, and one of
844 --xml-file=, --xml-fd= or --xml-socket= to select the XML
845 destination, one of --log-file=, --log-fd= or --log-socket=
846 to select the destination for any remaining text messages,
847 and, importantly, -q.
848
849 -q makes Valgrind completely silent on the text channel,
850 except in the case of critical failures, such as Valgrind
851 itself segfaulting, or failing to read debugging information.
852 Hence, in this scenario, it suffices to check whether or not
853 any output appeared on the text channel. If yes, then it is
854 likely to be a critical error which should be brought to the
855 attention of the user. If no (the text channel produced no
856 output) then it can be assumed that the run was successful.
857
858 This allows GUIs to make the critical distinction they need to
859 make (did the run fail or not?) without having to search or
860 filter the text output channel in any way.
861
862 It is also recommended to use --child-silent-after-fork=yes in
863 scenario (2).
864
865
866* Improvements and changes in Helgrind:
867
868 - XML output, as described above
869
870 - Checks for consistent association between pthread condition
871 variables and their associated mutexes are now performed.
872
873 - pthread_spinlock functions are supported.
874
875 - Modest performance improvements.
876
877 - Initial (skeletal) support for describing the behaviour of
878 non-POSIX synchronisation objects through ThreadSanitizer
879 compatible ANNOTATE_* macros.
880
881 - More controllable tradeoffs between performance and the level of
882 detail of "previous" accesses in a race. There are now three
883 settings:
884
885 * --history-level=full. This is the default, and was also the
886 default in 3.4.x. It shows both stacks involved in a race, but
887 requires a lot of memory and can be very slow in programs that
888 do many inter-thread synchronisation events.
889
890 * --history-level=none. This only shows the later stack involved
891 in a race. This can be much faster than --history-level=full,
892 but makes it much more difficult to find the other access
893 involved in the race.
894
895 The new intermediate setting is
896
897 * --history-level=approx
898
899 For the earlier (other) access, two stacks are presented. The
900 earlier access is guaranteed to be somewhere in between the two
901 program points denoted by those stacks. This is not as useful
902 as showing the exact stack for the previous access (as per
903 --history-level=full), but it is better than nothing, and it's
904 almost as fast as --history-level=none.
905
njnace90d82009-04-27 01:27:58 +0000906
bart6a146fc2009-07-21 07:32:19 +0000907* New features and improvements in DRD:
njn6eccab92009-08-14 02:41:37 +0000908
sewardj570e2442009-08-14 13:42:41 +0000909 - The error messages printed by DRD are now easier to interpret.
910 Instead of using two different numbers to identify each thread
911 (Valgrind thread ID and DRD thread ID), DRD does now identify
912 threads via a single number (the DRD thread ID). Furthermore
913 "first observed at" information is now printed for all error
914 messages related to synchronization objects.
njn6eccab92009-08-14 02:41:37 +0000915
bartfe212db2009-07-25 11:16:51 +0000916 - Added support for named semaphores (sem_open() and sem_close()).
njn6eccab92009-08-14 02:41:37 +0000917
bart6a146fc2009-07-21 07:32:19 +0000918 - Race conditions between pthread_barrier_wait() and
919 pthread_barrier_destroy() calls are now reported.
njn6eccab92009-08-14 02:41:37 +0000920
921 - Added support for custom allocators through the macros
sewardj570e2442009-08-14 13:42:41 +0000922 VALGRIND_MALLOCLIKE_BLOCK() VALGRIND_FREELIKE_BLOCK() (defined in
923 in <valgrind/valgrind.h>). An alternative for these two macros is
924 the new client request VG_USERREQ__DRD_CLEAN_MEMORY (defined in
bart6a146fc2009-07-21 07:32:19 +0000925 <valgrind/drd.h>).
njn6eccab92009-08-14 02:41:37 +0000926
sewardj570e2442009-08-14 13:42:41 +0000927 - Added support for annotating non-POSIX synchronization objects
928 through several new ANNOTATE_*() macros.
njn6eccab92009-08-14 02:41:37 +0000929
sewardj570e2442009-08-14 13:42:41 +0000930 - OpenMP: added support for the OpenMP runtime (libgomp) included
931 with gcc versions 4.3.0 and 4.4.0.
njn6eccab92009-08-14 02:41:37 +0000932
bart6a146fc2009-07-21 07:32:19 +0000933 - Faster operation.
njn6eccab92009-08-14 02:41:37 +0000934
935 - Added two new command-line options (--first-race-only and
bart6a146fc2009-07-21 07:32:19 +0000936 --segment-merging-interval).
937
njn6eccab92009-08-14 02:41:37 +0000938
sewardj570e2442009-08-14 13:42:41 +0000939* Genuinely atomic support for x86/amd64/ppc atomic instructions
940
941 Valgrind will now preserve (memory-access) atomicity of LOCK-
942 prefixed x86/amd64 instructions, and any others implying a global
943 bus lock. Ditto for PowerPC l{w,d}arx/st{w,d}cx. instructions.
944
945 This means that Valgrinded processes will "play nicely" in
946 situations where communication with other processes, or the kernel,
947 is done through shared memory and coordinated with such atomic
948 instructions. Prior to this change, such arrangements usually
949 resulted in hangs, races or other synchronisation failures, because
950 Valgrind did not honour atomicity of such instructions.
951
952
953* A new experimental tool, BBV, has been added. BBV generates basic
954 block vectors for use with the SimPoint analysis tool, which allows
955 a program's overall behaviour to be approximated by running only a
956 fraction of it. This is useful for computer architecture
957 researchers. You can run BBV by specifying --tool=exp-bbv (the
958 "exp-" prefix is short for "experimental"). BBV was written by
959 Vince Weaver.
960
961
962* Modestly improved support for running Windows applications under
963 Wine. In particular, initial support for reading Windows .PDB debug
964 information has been added.
965
966
967* A new Memcheck client request VALGRIND_COUNT_LEAK_BLOCKS has been
968 added. It is similar to VALGRIND_COUNT_LEAKS but counts blocks
969 instead of bytes.
970
971
972* The Valgrind client requests VALGRIND_PRINTF and
973 VALGRIND_PRINTF_BACKTRACE have been changed slightly. Previously,
974 the string was always printed immediately on its own line. Now, the
975 string will be added to a buffer but not printed until a newline is
976 encountered, or other Valgrind output is printed (note that for
977 VALGRIND_PRINTF_BACKTRACE, the back-trace itself is considered
978 "other Valgrind output"). This allows you to use multiple
979 VALGRIND_PRINTF calls to build up a single output line, and also to
980 print multiple output lines with a single request (by embedding
981 multiple newlines in the string).
982
983
984* The graphs drawn by Massif's ms_print program have changed slightly:
985
986 - The half-height chars '.' and ',' are no longer drawn, because
987 they are confusing. The --y option can be used if the default
988 y-resolution is not high enough.
989
990 - Horizontal lines are now drawn after the top of a snapshot if
991 there is a gap until the next snapshot. This makes it clear that
992 the memory usage has not dropped to zero between snapshots.
993
994
995* Something that happened in 3.4.0, but wasn't clearly announced: the
996 option --read-var-info=yes can be used by some tools (Memcheck,
997 Helgrind and DRD). When enabled, it causes Valgrind to read DWARF3
998 variable type and location information. This makes those tools
999 start up more slowly and increases memory consumption, but
1000 descriptions of data addresses in error messages become more
1001 detailed.
1002
1003
1004* exp-Omega, an experimental instantaneous leak-detecting tool, was
1005 disabled in 3.4.0 due to a lack of interest and maintenance,
1006 although the source code was still in the distribution. The source
1007 code has now been removed from the distribution. For anyone
1008 interested, the removal occurred in SVN revision r10247.
1009
njnacda1782009-06-04 23:11:50 +00001010
njn8b68b642009-06-24 00:37:09 +00001011* Some changes have been made to the build system.
njn6bf365c2009-02-11 00:35:45 +00001012
sewardj570e2442009-08-14 13:42:41 +00001013 - VEX/ is now integrated properly into the build system. This means
1014 that dependency tracking within VEX/ now works properly, "make
1015 install" will work without requiring "make" before it, and
1016 parallel builds (ie. 'make -j') now work (previously a
1017 .NOTPARALLEL directive was used to serialize builds, ie. 'make -j'
1018 was effectively ignored).
njn6bf365c2009-02-11 00:35:45 +00001019
sewardj570e2442009-08-14 13:42:41 +00001020 - The --with-vex configure option has been removed. It was of
1021 little use and removing it simplified the build system.
njn8b68b642009-06-24 00:37:09 +00001022
sewardj570e2442009-08-14 13:42:41 +00001023 - The location of some install files has changed. This should not
1024 affect most users. Those who might be affected:
njn8b68b642009-06-24 00:37:09 +00001025
1026 * For people who use Valgrind with MPI programs, the installed
sewardj570e2442009-08-14 13:42:41 +00001027 libmpiwrap.so library has moved from
1028 $(INSTALL)/<platform>/libmpiwrap.so to
1029 $(INSTALL)/libmpiwrap-<platform>.so.
njn8b68b642009-06-24 00:37:09 +00001030
sewardj570e2442009-08-14 13:42:41 +00001031 * For people who distribute standalone Valgrind tools, the
1032 installed libraries such as $(INSTALL)/<platform>/libcoregrind.a
1033 have moved to $(INSTALL)/libcoregrind-<platform>.a.
njn8b68b642009-06-24 00:37:09 +00001034
sewardj570e2442009-08-14 13:42:41 +00001035 These changes simplify the build system.
njn8b68b642009-06-24 00:37:09 +00001036
sewardj570e2442009-08-14 13:42:41 +00001037 - Previously, all the distributed suppression (*.supp) files were
1038 installed. Now, only default.supp is installed. This should not
1039 affect users as the other installed suppression files were not
1040 read; the fact that they were installed was a mistake.
1041
1042
1043* KNOWN LIMITATIONS:
1044
1045 - Memcheck is unusable with the Intel compiler suite version 11.1,
1046 when it generates code for SSE2-and-above capable targets. This
1047 is because of icc's use of highly optimised inlined strlen
1048 implementations. It causes Memcheck to report huge numbers of
1049 false errors even in simple programs. Helgrind and DRD may also
1050 have problems.
1051
1052 Versions 11.0 and earlier may be OK, but this has not been
1053 properly tested.
1054
1055
sewardj9176c862009-08-18 13:47:31 +00001056The following bugs have been fixed or resolved. Note that "n-i-bz"
1057stands for "not in bugzilla" -- that is, a bug that was reported to us
1058but never got a bugzilla entry. We encourage you to file bugs in
1059bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
1060mailing the developers (or mailing lists) directly -- bugs that are
1061not entered into bugzilla tend to get forgotten about or ignored.
sewardj570e2442009-08-14 13:42:41 +00001062
sewardj9176c862009-08-18 13:47:31 +00001063To see details of a given bug, visit
1064https://bugs.kde.org/show_bug.cgi?id=XXXXXX
1065where XXXXXX is the bug number as listed below.
sewardj570e2442009-08-14 13:42:41 +00001066
sewardj9176c862009-08-18 13:47:31 +0000106784303 How about a LockCheck tool?
106891633 dereference of null ptr in vgPlain_st_basetype
106997452 Valgrind doesn't report any pthreads problems
1070100628 leak-check gets assertion failure when using
1071 VALGRIND_MALLOCLIKE_BLOCK on malloc()ed memory
1072108528 NPTL pthread cleanup handlers not called
1073110126 Valgrind 2.4.1 configure.in tramples CFLAGS
1074110128 mallinfo is not implemented...
1075110770 VEX: Generated files not always updated when making valgrind
1076111102 Memcheck: problems with large (memory footprint) applications
1077115673 Vex's decoder should never assert
1078117564 False positive: Syscall param clone(child_tidptr) contains
1079 uninitialised byte(s)
1080119404 executing ssh from inside valgrind fails
1081133679 Callgrind does not write path names to sources with dwarf debug
1082 info
1083135847 configure.in problem with non gnu compilers (and possible fix)
1084136154 threads.c:273 (vgCallgrind_post_signal): Assertion
1085 '*(vgCallgrind_current_fn_stack.top) == 0' failed.
1086136230 memcheck reports "possibly lost", should be "still reachable"
1087137073 NULL arg to MALLOCLIKE_BLOCK causes crash
1088137904 Valgrind reports a memory leak when using POSIX threads,
1089 while it shouldn't
1090139076 valgrind VT_GETSTATE error
1091142228 complaint of elf_dynamic_do_rela in trivial usage
1092145347 spurious warning with USBDEVFS_REAPURB
1093148441 (wine) can't find memory leak in Wine, win32 binary
1094 executable file.
1095148742 Leak-check fails assert on exit
1096149878 add (proper) check for calloc integer overflow
1097150606 Call graph is broken when using callgrind control
1098152393 leak errors produce an exit code of 0. I need some way to
1099 cause leak errors to result in a nonzero exit code.
1100157154 documentation (leak-resolution doc speaks about num-callers
1101 def=4) + what is a loss record
1102159501 incorrect handling of ALSA ioctls
1103162020 Valgrinding an empty/zero-byte file crashes valgrind
1104162482 ppc: Valgrind crashes while reading stabs information
1105162718 x86: avoid segment selector 0 in sys_set_thread_area()
1106163253 (wine) canonicaliseSymtab forgot some fields in DiSym
1107163560 VEX/test_main.c is missing from valgrind-3.3.1
1108164353 malloc_usable_size() doesn't return a usable size
1109165468 Inconsistent formatting in memcheck manual -- please fix
1110169505 main.c:286 (endOfInstr):
1111 Assertion 'ii->cost_offset == *cost_offset' failed
1112177206 Generate default.supp during compile instead of configure
1113177209 Configure valt_load_address based on arch+os
1114177305 eventfd / syscall 323 patch lost
1115179731 Tests fail to build because of inlining of non-local asm labels
1116181394 helgrind: libhb_core.c:3762 (msm_write): Assertion
1117 'ordxx == POrd_EQ || ordxx == POrd_LT' failed.
1118181594 Bogus warning for empty text segment
1119181707 dwarf doesn't require enumerations to have name
1120185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64
1121185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree):
1122 Assertion '!already_present' failed.
1123185359 exp-ptrcheck: unhandled syscall getresuid()
1124185794 "WARNING: unhandled syscall: 285" (fallocate) on x86_64
1125185816 Valgrind is unable to handle debug info for files with split
1126 debug info that are prelinked afterwards
1127185980 [darwin] unhandled syscall: sem_open
1128186238 bbToIR_AMD64: disInstr miscalculated next %rip
1129186507 exp-ptrcheck unhandled syscalls prctl, etc.
1130186790 Suppression pattern used for leaks are not reported
1131186796 Symbols with length>200 in suppression files are ignored
1132187048 drd: mutex PTHREAD_PROCESS_SHARED attribute missinterpretation
1133187416 exp-ptrcheck: support for __NR_{setregid,setreuid,setresuid}
1134188038 helgrind: hg_main.c:926: mk_SHVAL_fail: the 'impossible' happened
1135188046 bashisms in the configure script
1136188127 amd64->IR: unhandled instruction bytes: 0xF0 0xF 0xB0 0xA
1137188161 memcheck: --track-origins=yes asserts "mc_machine.c:672
1138 (get_otrack_shadow_offset_wrk): the 'impossible' happened."
1139188248 helgrind: pthread_cleanup_push, pthread_rwlock_unlock,
1140 assertion fail "!lock->heldBy"
1141188427 Add support for epoll_create1 (with patch)
1142188530 Support for SIOCGSTAMPNS
1143188560 Include valgrind.spec in the tarball
1144188572 Valgrind on Mac should suppress setenv() mem leak
1145189054 Valgrind fails to build because of duplicate non-local asm labels
1146189737 vex amd64->IR: unhandled instruction bytes: 0xAC
1147189762 epoll_create syscall not handled (--tool=exp-ptrcheck)
1148189763 drd assertion failure: s_threadinfo[tid].is_recording
1149190219 unhandled syscall: 328 (x86-linux)
1150190391 dup of 181394; see above
1151190429 Valgrind reports lots of errors in ld.so with x86_64 2.9.90 glibc
1152190820 No debug information on powerpc-linux
sewardj9176c862009-08-18 13:47:31 +00001153191095 PATCH: Improve usbdevfs ioctl handling
1154191182 memcheck: VALGRIND_LEAK_CHECK quadratic when big nr of chunks
1155 or big nr of errors
1156191189 --xml=yes should obey --gen-suppressions=all
1157191192 syslog() needs a suppression on macosx
1158191271 DARWIN: WARNING: unhandled syscall: 33554697 a.k.a.: 265
1159191761 getrlimit on MacOSX
1160191992 multiple --fn-skip only works sometimes; dependent on order
1161192634 V. reports "aspacem sync_check_mapping_callback:
1162 segment mismatch" on Darwin
1163192954 __extension__ missing on 2 client requests
1164194429 Crash at start-up with glibc-2.10.1 and linux-2.6.29
1165194474 "INSTALL" file has different build instructions than "README"
1166194671 Unhandled syscall (sem_wait?) from mac valgrind
1167195069 memcheck: reports leak (memory still reachable) for
1168 printf("%d', x)
1169195169 drd: (vgDrd_barrier_post_wait):
1170 Assertion 'r->sg[p->post_iteration]' failed.
1171195268 valgrind --log-file doesn't accept ~/...
1172195838 VEX abort: LibVEX_N_SPILL_BYTES too small for CPUID boilerplate
1173195860 WARNING: unhandled syscall: unix:223
1174196528 need a error suppression for pthread_rwlock_init under os x?
1175197227 Support aio_* syscalls on Darwin
1176197456 valgrind should reject --suppressions=(directory)
1177197512 DWARF2 CFI reader: unhandled CFI instruction 0:10
1178197591 unhandled syscall 27 (mincore)
1179197793 Merge DCAS branch to the trunk == 85756, 142103
1180197794 Avoid duplicate filenames in Vex
1181197898 make check fails on current SVN
1182197901 make check fails also under exp-ptrcheck in current SVN
1183197929 Make --leak-resolution=high the default
1184197930 Reduce spacing between leak reports
1185197933 Print command line of client at start-up, and shorten preamble
1186197966 unhandled syscall 205 (x86-linux, --tool=exp-ptrcheck)
1187198395 add BBV to the distribution as an experimental tool
1188198624 Missing syscalls on Darwin: 82, 167, 281, 347
1189198649 callgrind_annotate doesn't cumulate counters
1190199338 callgrind_annotate sorting/thresholds are broken for all but Ir
1191199977 Valgrind complains about an unrecognized instruction in the
1192 atomic_incs test program
1193200029 valgrind isn't able to read Fedora 12 debuginfo
1194200760 darwin unhandled syscall: unix:284
1195200827 DRD doesn't work on Mac OS X
1196200990 VG_(read_millisecond_timer)() does not work correctly
1197201016 Valgrind does not support pthread_kill() on Mac OS
1198201169 Document --read-var-info
1199201323 Pre-3.5.0 performance sanity checking
1200201384 Review user manual for the 3.5.0 release
1201201585 mfpvr not implemented on ppc
1202201708 tests failing because x86 direction flag is left set
1203201757 Valgrind doesn't handle any recent sys_futex additions
sewardj6bbe0242009-08-19 11:38:54 +00001204204377 64-bit valgrind can not start a shell script
1205 (with #!/path/to/shell) if the shell is a 32-bit executable
sewardj9176c862009-08-18 13:47:31 +00001206n-i-bz drd: fixed assertion failure triggered by mutex reinitialization.
1207n-i-bz drd: fixed a bug that caused incorrect messages to be printed
1208 about memory allocation events with memory access tracing enabled
1209n-i-bz drd: fixed a memory leak triggered by vector clock deallocation
njn254dafe2009-02-09 03:25:04 +00001210
sewardj48867fb2009-08-19 22:40:53 +00001211(3.5.0: 19 Aug 2009, vex r1913, valgrind r10846).
sewardj570e2442009-08-14 13:42:41 +00001212
njn6eccab92009-08-14 02:41:37 +00001213
njn254dafe2009-02-09 03:25:04 +00001214
njnad91a752009-03-13 19:55:36 +00001215Release 3.4.1 (28 February 2009)
1216~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
12173.4.1 is a bug-fix release that fixes some regressions and assertion
1218failures in debug info reading in 3.4.0, most notably incorrect stack
1219traces on amd64-linux on older (glibc-2.3 based) systems. Various
1220other debug info problems are also fixed. A number of bugs in the
1221exp-ptrcheck tool introduced in 3.4.0 have been fixed.
1222
1223In view of the fact that 3.4.0 contains user-visible regressions
1224relative to 3.3.x, upgrading to 3.4.1 is recommended. Packagers are
1225encouraged to ship 3.4.1 in preference to 3.4.0.
1226
1227The fixed bugs are as follows. Note that "n-i-bz" stands for "not in
1228bugzilla" -- that is, a bug that was reported to us but never got a
1229bugzilla entry. We encourage you to file bugs in bugzilla
1230(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the
1231developers (or mailing lists) directly -- bugs that are not entered
1232into bugzilla tend to get forgotten about or ignored.
1233
1234n-i-bz Fix various bugs reading icc-11 generated debug info
1235n-i-bz Fix various bugs reading gcc-4.4 generated debug info
1236n-i-bz Preliminary support for glibc-2.10 / Fedora 11
1237n-i-bz Cachegrind and Callgrind: handle non-power-of-two cache sizes,
1238 so as to support (eg) 24k Atom D1 and Core2 with 3/6/12MB L2.
1239179618 exp-ptrcheck crashed / exit prematurely
1240179624 helgrind: false positive races with pthread_create and
1241 recv/open/close/read
1242134207 pkg-config output contains @VG_PLATFORM@
1243176926 floating point exception at valgrind startup with PPC 440EPX
1244181594 Bogus warning for empty text segment
1245173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant rex prefixes)
1246181707 Dwarf3 doesn't require enumerations to have name
1247185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64
1248185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree):
1249 Assertion '!already_present' failed.
1250185359 exp-ptrcheck unhandled syscall getresuid()
1251
1252(3.4.1.RC1: 24 Feb 2008, vex r1884, valgrind r9253).
1253(3.4.1: 28 Feb 2008, vex r1884, valgrind r9293).
1254
1255
1256
sewardj565fcf12008-12-23 13:14:06 +00001257Release 3.4.0 (2 January 2009)
1258~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
12593.4.0 is a feature release with many significant improvements and the
1260usual collection of bug fixes. This release supports X86/Linux,
1261AMD64/Linux, PPC32/Linux and PPC64/Linux. Support for recent distros
1262(using gcc 4.4, glibc 2.8 and 2.9) has been added.
sewardj8639f112008-12-17 12:26:24 +00001263
sewardj565fcf12008-12-23 13:14:06 +000012643.4.0 brings some significant tool improvements. Memcheck can now
1265report the origin of uninitialised values, the thread checkers
1266Helgrind and DRD are much improved, and we have a new experimental
1267tool, exp-Ptrcheck, which is able to detect overruns of stack and
1268global arrays. In detail:
sewardj8639f112008-12-17 12:26:24 +00001269
sewardj565fcf12008-12-23 13:14:06 +00001270* Memcheck is now able to track the origin of uninitialised values.
1271 When it reports an uninitialised value error, it will try to show
1272 the origin of the value, as either a heap or stack allocation.
1273 Origin tracking is expensive and so is not enabled by default. To
1274 use it, specify --track-origins=yes. Memcheck's speed will be
1275 essentially halved, and memory usage will be significantly
1276 increased. Nevertheless it can drastically reduce the effort
1277 required to identify the root cause of uninitialised value errors,
1278 and so is often a programmer productivity win, despite running more
1279 slowly.
sewardj8639f112008-12-17 12:26:24 +00001280
sewardj565fcf12008-12-23 13:14:06 +00001281* A version (1.4.0) of the Valkyrie GUI, that works with Memcheck in
sewardjb8b16642009-01-02 23:21:54 +00001282 3.4.0, will be released shortly.
sewardj8639f112008-12-17 12:26:24 +00001283
sewardj565fcf12008-12-23 13:14:06 +00001284* Helgrind's race detection algorithm has been completely redesigned
1285 and reimplemented, to address usability and scalability concerns:
sewardj8639f112008-12-17 12:26:24 +00001286
sewardj565fcf12008-12-23 13:14:06 +00001287 - The new algorithm has a lower false-error rate: it is much less
1288 likely to report races that do not really exist.
sewardj8639f112008-12-17 12:26:24 +00001289
sewardj565fcf12008-12-23 13:14:06 +00001290 - Helgrind will display full call stacks for both accesses involved
sewardjf1fa5ed2008-12-23 17:04:52 +00001291 in a race. This makes it easier to identify the root causes of
1292 races.
sewardj8639f112008-12-17 12:26:24 +00001293
sewardj565fcf12008-12-23 13:14:06 +00001294 - Limitations on the size of program that can run have been removed.
sewardj8639f112008-12-17 12:26:24 +00001295
sewardj565fcf12008-12-23 13:14:06 +00001296 - Performance has been modestly improved, although that is very
1297 workload-dependent.
sewardj8639f112008-12-17 12:26:24 +00001298
sewardj565fcf12008-12-23 13:14:06 +00001299 - Direct support for Qt4 threading has been added.
sewardj8639f112008-12-17 12:26:24 +00001300
sewardj565fcf12008-12-23 13:14:06 +00001301 - pthread_barriers are now directly supported.
njn75010ca2008-02-09 05:10:45 +00001302
sewardj565fcf12008-12-23 13:14:06 +00001303 - Helgrind works well on all supported Linux targets.
sewardj8639f112008-12-17 12:26:24 +00001304
sewardj565fcf12008-12-23 13:14:06 +00001305* The DRD thread debugging tool has seen major improvements:
sewardj8639f112008-12-17 12:26:24 +00001306
sewardj565fcf12008-12-23 13:14:06 +00001307 - Greatly improved performance and significantly reduced memory
1308 usage.
sewardj8639f112008-12-17 12:26:24 +00001309
sewardj565fcf12008-12-23 13:14:06 +00001310 - Support for several major threading libraries (Boost.Thread, Qt4,
1311 glib, OpenMP) has been added.
sewardj8639f112008-12-17 12:26:24 +00001312
sewardj565fcf12008-12-23 13:14:06 +00001313 - Support for atomic instructions, POSIX semaphores, barriers and
1314 reader-writer locks has been added.
1315
1316 - Works now on PowerPC CPUs too.
1317
1318 - Added support for printing thread stack usage at thread exit time.
1319
1320 - Added support for debugging lock contention.
1321
1322 - Added a manual for Drd.
1323
1324* A new experimental tool, exp-Ptrcheck, has been added. Ptrcheck
1325 checks for misuses of pointers. In that sense it is a bit like
1326 Memcheck. However, Ptrcheck can do things Memcheck can't: it can
1327 detect overruns of stack and global arrays, it can detect
1328 arbitrarily far out-of-bounds accesses to heap blocks, and it can
1329 detect accesses heap blocks that have been freed a very long time
1330 ago (millions of blocks in the past).
1331
1332 Ptrcheck currently works only on x86-linux and amd64-linux. To use
1333 it, use --tool=exp-ptrcheck. A simple manual is provided, as part
1334 of the main Valgrind documentation. As this is an experimental
1335 tool, we would be particularly interested in hearing about your
1336 experiences with it.
1337
sewardj974aace2008-12-24 18:40:57 +00001338* exp-Omega, an experimental instantaneous leak-detecting tool, is no
1339 longer built by default, although the code remains in the repository
1340 and the tarball. This is due to three factors: a perceived lack of
1341 users, a lack of maintenance, and concerns that it may not be
1342 possible to achieve reliable operation using the existing design.
sewardj565fcf12008-12-23 13:14:06 +00001343
1344* As usual, support for the latest Linux distros and toolchain
1345 components has been added. It should work well on Fedora Core 10,
1346 OpenSUSE 11.1 and Ubuntu 8.10. gcc-4.4 (in its current pre-release
1347 state) is supported, as is glibc-2.9. The C++ demangler has been
1348 updated so as to work well with C++ compiled by even the most recent
1349 g++'s.
1350
1351* You can now use frame-level wildcards in suppressions. This was a
1352 frequently-requested enhancement. A line "..." in a suppression now
1353 matches zero or more frames. This makes it easier to write
1354 suppressions which are precise yet insensitive to changes in
1355 inlining behaviour.
1356
1357* 3.4.0 adds support on x86/amd64 for the SSSE3 instruction set.
1358
1359* Very basic support for IBM Power6 has been added (64-bit processes only).
1360
1361* Valgrind is now cross-compilable. For example, it is possible to
1362 cross compile Valgrind on an x86/amd64-linux host, so that it runs
1363 on a ppc32/64-linux target.
1364
sewardjf1fa5ed2008-12-23 17:04:52 +00001365* You can set the main thread's stack size at startup using the
1366 new --main-stacksize= flag (subject of course to ulimit settings).
1367 This is useful for running apps that need a lot of stack space.
1368
1369* The limitation that you can't use --trace-children=yes together
1370 with --db-attach=yes has been removed.
1371
sewardj565fcf12008-12-23 13:14:06 +00001372* The following bugs have been fixed. Note that "n-i-bz" stands for
1373 "not in bugzilla" -- that is, a bug that was reported to us but
1374 never got a bugzilla entry. We encourage you to file bugs in
1375 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
1376 mailing the developers (or mailing lists) directly.
1377
sewardjf1fa5ed2008-12-23 17:04:52 +00001378 n-i-bz Make return types for some client requests 64-bit clean
1379 n-i-bz glibc 2.9 support
1380 n-i-bz ignore unsafe .valgrindrc's (CVE-2008-4865)
1381 n-i-bz MPI_Init(0,0) is valid but libmpiwrap.c segfaults
1382 n-i-bz Building in an env without gdb gives bogus gdb attach
1383 92456 Tracing the origin of uninitialised memory
1384 106497 Valgrind does not demangle some C++ template symbols
1385 162222 ==106497
sewardjdf1fc9a2008-12-23 23:09:57 +00001386 151612 Suppression with "..." (frame-level wildcards in .supp files)
sewardjf1fa5ed2008-12-23 17:04:52 +00001387 156404 Unable to start oocalc under memcheck on openSUSE 10.3 (64-bit)
1388 159285 unhandled syscall:25 (stime, on x86-linux)
1389 159452 unhandled ioctl 0x8B01 on "valgrind iwconfig"
1390 160954 ppc build of valgrind crashes with illegal instruction (isel)
1391 160956 mallinfo implementation, w/ patch
1392 162092 Valgrind fails to start gnome-system-monitor
1393 162819 malloc_free_fill test doesn't pass on glibc2.8 x86
1394 163794 assertion failure with "--track-origins=yes"
1395 163933 sigcontext.err and .trapno must be set together
1396 163955 remove constraint !(--db-attach=yes && --trace-children=yes)
1397 164476 Missing kernel module loading system calls
1398 164669 SVN regression: mmap() drops posix file locks
1399 166581 Callgrind output corruption when program forks
1400 167288 Patch file for missing system calls on Cell BE
1401 168943 unsupported scas instruction pentium
1402 171645 Unrecognised instruction (MOVSD, non-binutils encoding)
1403 172417 x86->IR: 0x82 ...
1404 172563 amd64->IR: 0xD9 0xF5 - fprem1
1405 173099 .lds linker script generation error
1406 173177 [x86_64] syscalls: 125/126/179 (capget/capset/quotactl)
1407 173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant prefixes)
1408 174532 == 173751
1409 174908 --log-file value not expanded correctly for core file
1410 175044 Add lookup_dcookie for amd64
1411 175150 x86->IR: 0xF2 0xF 0x11 0xC1 (movss non-binutils encoding)
sewardj565fcf12008-12-23 13:14:06 +00001412
1413Developer-visible changes:
1414
1415* Valgrind's debug-info reading machinery has been majorly overhauled.
1416 It can now correctly establish the addresses for ELF data symbols,
1417 which is something that has never worked properly before now.
1418
1419 Also, Valgrind can now read DWARF3 type and location information for
1420 stack and global variables. This makes it possible to use the
1421 framework to build tools that rely on knowing the type and locations
1422 of stack and global variables, for example exp-Ptrcheck.
1423
1424 Reading of such information is disabled by default, because most
1425 tools don't need it, and because it is expensive in space and time.
1426 However, you can force Valgrind to read it, using the
1427 --read-var-info=yes flag. Memcheck, Helgrind and DRD are able to
1428 make use of such information, if present, to provide source-level
1429 descriptions of data addresses in the error messages they create.
1430
sewardj974aace2008-12-24 18:40:57 +00001431(3.4.0.RC1: 24 Dec 2008, vex r1878, valgrind r8882).
sewardjb8b16642009-01-02 23:21:54 +00001432(3.4.0: 3 Jan 2009, vex r1878, valgrind r8899).
sewardj565fcf12008-12-23 13:14:06 +00001433
sewardj8639f112008-12-17 12:26:24 +00001434
sewardj9f42ba32008-06-04 11:47:51 +00001435
1436Release 3.3.1 (4 June 2008)
1437~~~~~~~~~~~~~~~~~~~~~~~~~~~
14383.3.1 fixes a bunch of bugs in 3.3.0, adds support for glibc-2.8 based
1439systems (openSUSE 11, Fedora Core 9), improves the existing glibc-2.7
1440support, and adds support for the SSSE3 (Core 2) instruction set.
1441
14423.3.1 will likely be the last release that supports some very old
1443systems. In particular, the next major release, 3.4.0, will drop
1444support for the old LinuxThreads threading library, and for gcc
1445versions prior to 3.0.
1446
1447The fixed bugs are as follows. Note that "n-i-bz" stands for "not in
1448bugzilla" -- that is, a bug that was reported to us but never got a
1449bugzilla entry. We encourage you to file bugs in bugzilla
1450(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the
1451developers (or mailing lists) directly -- bugs that are not entered
1452into bugzilla tend to get forgotten about or ignored.
1453
1454n-i-bz Massif segfaults at exit
1455n-i-bz Memcheck asserts on Altivec code
1456n-i-bz fix sizeof bug in Helgrind
1457n-i-bz check fd on sys_llseek
1458n-i-bz update syscall lists to kernel 2.6.23.1
1459n-i-bz support sys_sync_file_range
1460n-i-bz handle sys_sysinfo, sys_getresuid, sys_getresgid on ppc64-linux
1461n-i-bz intercept memcpy in 64-bit ld.so's
1462n-i-bz Fix wrappers for sys_{futimesat,utimensat}
1463n-i-bz Minor false-error avoidance fixes for Memcheck
1464n-i-bz libmpiwrap.c: add a wrapper for MPI_Waitany
1465n-i-bz helgrind support for glibc-2.8
1466n-i-bz partial fix for mc_leakcheck.c:698 assert:
1467 'lc_shadows[i]->data + lc_shadows[i] ...
1468n-i-bz Massif/Cachegrind output corruption when programs fork
1469n-i-bz register allocator fix: handle spill stores correctly
1470n-i-bz add support for PA6T PowerPC CPUs
1471126389 vex x86->IR: 0xF 0xAE (FXRSTOR)
1472158525 ==126389
1473152818 vex x86->IR: 0xF3 0xAC (repz lodsb)
1474153196 vex x86->IR: 0xF2 0xA6 (repnz cmpsb)
1475155011 vex x86->IR: 0xCF (iret)
1476155091 Warning [...] unhandled DW_OP_ opcode 0x23
1477156960 ==155901
1478155528 support Core2/SSSE3 insns on x86/amd64
1479155929 ms_print fails on massif outputs containing long lines
1480157665 valgrind fails on shmdt(0) after shmat to 0
1481157748 support x86 PUSHFW/POPFW
1482158212 helgrind: handle pthread_rwlock_try{rd,wr}lock.
1483158425 sys_poll incorrectly emulated when RES==0
1484158744 vex amd64->IR: 0xF0 0x41 0xF 0xC0 (xaddb)
1485160907 Support for a couple of recent Linux syscalls
1486161285 Patch -- support for eventfd() syscall
1487161378 illegal opcode in debug libm (FUCOMPP)
1488160136 ==161378
1489161487 number of suppressions files is limited to 10
1490162386 ms_print typo in milliseconds time unit for massif
1491161036 exp-drd: client allocated memory was never freed
1492162663 signalfd_wrapper fails on 64bit linux
1493
1494(3.3.1.RC1: 2 June 2008, vex r1854, valgrind r8169).
1495(3.3.1: 4 June 2008, vex r1854, valgrind r8180).
1496
1497
1498
sewardjcb2288b2007-12-02 02:08:17 +00001499Release 3.3.0 (7 December 2007)
1500~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardjcb2288b2007-12-02 02:08:17 +000015013.3.0 is a feature release with many significant improvements and the
1502usual collection of bug fixes. This release supports X86/Linux,
1503AMD64/Linux, PPC32/Linux and PPC64/Linux. Support for recent distros
1504(using gcc 4.3, glibc 2.6 and 2.7) has been added.
njnf4f9cd42007-11-23 01:42:10 +00001505
sewardjcb2288b2007-12-02 02:08:17 +00001506The main excitement in 3.3.0 is new and improved tools. Helgrind
1507works again, Massif has been completely overhauled and much improved,
1508Cachegrind now does branch-misprediction profiling, and a new category
1509of experimental tools has been created, containing two new tools:
1510Omega and DRD. There are many other smaller improvements. In detail:
njn374a36d2007-11-23 01:41:32 +00001511
sewardjcb2288b2007-12-02 02:08:17 +00001512- Helgrind has been completely overhauled and works for the first time
1513 since Valgrind 2.2.0. Supported functionality is: detection of
1514 misuses of the POSIX PThreads API, detection of potential deadlocks
1515 resulting from cyclic lock dependencies, and detection of data
1516 races. Compared to the 2.2.0 Helgrind, the race detection algorithm
1517 has some significant improvements aimed at reducing the false error
1518 rate. Handling of various kinds of corner cases has been improved.
1519 Efforts have been made to make the error messages easier to
1520 understand. Extensive documentation is provided.
njn374a36d2007-11-23 01:41:32 +00001521
sewardjcb2288b2007-12-02 02:08:17 +00001522- Massif has been completely overhauled. Instead of measuring
1523 space-time usage -- which wasn't always useful and many people found
1524 confusing -- it now measures space usage at various points in the
1525 execution, including the point of peak memory allocation. Its
1526 output format has also changed: instead of producing PostScript
1527 graphs and HTML text, it produces a single text output (via the new
1528 'ms_print' script) that contains both a graph and the old textual
1529 information, but in a more compact and readable form. Finally, the
1530 new version should be more reliable than the old one, as it has been
1531 tested more thoroughly.
njn718d3b12006-12-16 00:54:12 +00001532
sewardj8badbaa2007-05-08 09:20:25 +00001533- Cachegrind has been extended to do branch-misprediction profiling.
1534 Both conditional and indirect branches are profiled. The default
1535 behaviour of Cachegrind is unchanged. To use the new functionality,
1536 give the option --branch-sim=yes.
1537
sewardjcb2288b2007-12-02 02:08:17 +00001538- A new category of "experimental tools" has been created. Such tools
1539 may not work as well as the standard tools, but are included because
1540 some people will find them useful, and because exposure to a wider
1541 user group provides tool authors with more end-user feedback. These
1542 tools have a "exp-" prefix attached to their names to indicate their
1543 experimental nature. Currently there are two experimental tools:
njn374a36d2007-11-23 01:41:32 +00001544
sewardjcb2288b2007-12-02 02:08:17 +00001545 * exp-Omega: an instantaneous leak detector. See
1546 exp-omega/docs/omega_introduction.txt.
njn374a36d2007-11-23 01:41:32 +00001547
sewardjcb2288b2007-12-02 02:08:17 +00001548 * exp-DRD: a data race detector based on the happens-before
sewardj8b09d4f2007-12-04 21:27:18 +00001549 relation. See exp-drd/docs/README.txt.
njn374a36d2007-11-23 01:41:32 +00001550
sewardjcb2288b2007-12-02 02:08:17 +00001551- Scalability improvements for very large programs, particularly those
1552 which have a million or more malloc'd blocks in use at once. These
1553 improvements mostly affect Memcheck. Memcheck is also up to 10%
1554 faster for all programs, with x86-linux seeing the largest
1555 improvement.
njn374a36d2007-11-23 01:41:32 +00001556
sewardjcb2288b2007-12-02 02:08:17 +00001557- Works well on the latest Linux distros. Has been tested on Fedora
1558 Core 8 (x86, amd64, ppc32, ppc64) and openSUSE 10.3. glibc 2.6 and
1559 2.7 are supported. gcc-4.3 (in its current pre-release state) is
1560 supported. At the same time, 3.3.0 retains support for older
1561 distros.
njn374a36d2007-11-23 01:41:32 +00001562
sewardjcb2288b2007-12-02 02:08:17 +00001563- The documentation has been modestly reorganised with the aim of
1564 making it easier to find information on common-usage scenarios.
1565 Some advanced material has been moved into a new chapter in the main
1566 manual, so as to unclutter the main flow, and other tidying up has
1567 been done.
njn374a36d2007-11-23 01:41:32 +00001568
sewardjcb2288b2007-12-02 02:08:17 +00001569- There is experimental support for AIX 5.3, both 32-bit and 64-bit
1570 processes. You need to be running a 64-bit kernel to use Valgrind
1571 on a 64-bit executable.
njn374a36d2007-11-23 01:41:32 +00001572
sewardjcb2288b2007-12-02 02:08:17 +00001573- There have been some changes to command line options, which may
1574 affect you:
njn374a36d2007-11-23 01:41:32 +00001575
sewardjcb2288b2007-12-02 02:08:17 +00001576 * --log-file-exactly and
1577 --log-file-qualifier options have been removed.
njn374a36d2007-11-23 01:41:32 +00001578
sewardjcb2288b2007-12-02 02:08:17 +00001579 To make up for this --log-file option has been made more powerful.
1580 It now accepts a %p format specifier, which is replaced with the
1581 process ID, and a %q{FOO} format specifier, which is replaced with
1582 the contents of the environment variable FOO.
1583
1584 * --child-silent-after-fork=yes|no [no]
1585
1586 Causes Valgrind to not show any debugging or logging output for
1587 the child process resulting from a fork() call. This can make the
1588 output less confusing (although more misleading) when dealing with
1589 processes that create children.
1590
1591 * --cachegrind-out-file, --callgrind-out-file and --massif-out-file
1592
1593 These control the names of the output files produced by
1594 Cachegrind, Callgrind and Massif. They accept the same %p and %q
1595 format specifiers that --log-file accepts. --callgrind-out-file
1596 replaces Callgrind's old --base option.
1597
1598 * Cachegrind's 'cg_annotate' script no longer uses the --<pid>
1599 option to specify the output file. Instead, the first non-option
1600 argument is taken to be the name of the output file, and any
1601 subsequent non-option arguments are taken to be the names of
1602 source files to be annotated.
1603
1604 * Cachegrind and Callgrind now use directory names where possible in
1605 their output files. This means that the -I option to
1606 'cg_annotate' and 'callgrind_annotate' should not be needed in
1607 most cases. It also means they can correctly handle the case
1608 where two source files in different directories have the same
1609 name.
1610
1611- Memcheck offers a new suppression kind: "Jump". This is for
1612 suppressing jump-to-invalid-address errors. Previously you had to
1613 use an "Addr1" suppression, which didn't make much sense.
1614
1615- Memcheck has new flags --malloc-fill=<hexnum> and
1616 --free-fill=<hexnum> which free malloc'd / free'd areas with the
1617 specified byte. This can help shake out obscure memory corruption
barte7b86c32009-07-21 07:35:23 +00001618 problems. The definedness and addressability of these areas is
sewardjcb2288b2007-12-02 02:08:17 +00001619 unchanged -- only the contents are affected.
njn718d3b12006-12-16 00:54:12 +00001620
1621- The behaviour of Memcheck's client requests VALGRIND_GET_VBITS and
1622 VALGRIND_SET_VBITS have changed slightly. They no longer issue
1623 addressability errors -- if either array is partially unaddressable,
1624 they just return 3 (as before). Also, SET_VBITS doesn't report
1625 definedness errors if any of the V bits are undefined.
sewardj1ede80f2006-05-26 00:47:13 +00001626
njn92676d72007-12-04 21:18:06 +00001627- The following Memcheck client requests have been removed:
1628 VALGRIND_MAKE_NOACCESS
1629 VALGRIND_MAKE_WRITABLE
1630 VALGRIND_MAKE_READABLE
1631 VALGRIND_CHECK_WRITABLE
1632 VALGRIND_CHECK_READABLE
1633 VALGRIND_CHECK_DEFINED
1634 They were deprecated in 3.2.0, when equivalent but better-named client
1635 requests were added. See the 3.2.0 release notes for more details.
1636
njn5a0a7732007-02-12 10:37:37 +00001637- The behaviour of the tool Lackey has changed slightly. First, the output
1638 from --trace-mem has been made more compact, to reduce the size of the
1639 traces. Second, a new option --trace-superblocks has been added, which
1640 shows the addresses of superblocks (code blocks) as they are executed.
1641
sewardjcb2288b2007-12-02 02:08:17 +00001642- The following bugs have been fixed. Note that "n-i-bz" stands for
1643 "not in bugzilla" -- that is, a bug that was reported to us but
1644 never got a bugzilla entry. We encourage you to file bugs in
1645 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
1646 mailing the developers (or mailing lists) directly.
1647
1648 n-i-bz x86_linux_REDIR_FOR_index() broken
1649 n-i-bz guest-amd64/toIR.c:2512 (dis_op2_E_G): Assertion `0' failed.
1650 n-i-bz Support x86 INT insn (INT (0xCD) 0x40 - 0x43)
1651 n-i-bz Add sys_utimensat system call for Linux x86 platform
1652 79844 Helgrind complains about race condition which does not exist
1653 82871 Massif output function names too short
1654 89061 Massif: ms_main.c:485 (get_XCon): Assertion `xpt->max_chi...'
1655 92615 Write output from Massif at crash
1656 95483 massif feature request: include peak allocation in report
1657 112163 MASSIF crashed with signal 7 (SIGBUS) after running 2 days
1658 119404 problems running setuid executables (partial fix)
1659 121629 add instruction-counting mode for timing
1660 127371 java vm giving unhandled instruction bytes: 0x26 0x2E 0x64 0x65
1661 129937 ==150380
1662 129576 Massif loses track of memory, incorrect graphs
1663 132132 massif --format=html output does not do html entity escaping
1664 132950 Heap alloc/usage summary
1665 133962 unhandled instruction bytes: 0xF2 0x4C 0xF 0x10
1666 134990 use -fno-stack-protector if possible
1667 136382 ==134990
1668 137396 I would really like helgrind to work again...
1669 137714 x86/amd64->IR: 0x66 0xF 0xF7 0xC6 (maskmovq, maskmovdq)
1670 141631 Massif: percentages don't add up correctly
1671 142706 massif numbers don't seem to add up
1672 143062 massif crashes on app exit with signal 8 SIGFPE
1673 144453 (get_XCon): Assertion 'xpt->max_children != 0' failed.
1674 145559 valgrind aborts when malloc_stats is called
1675 145609 valgrind aborts all runs with 'repeated section!'
1676 145622 --db-attach broken again on x86-64
1677 145837 ==149519
1678 145887 PPC32: getitimer() system call is not supported
1679 146252 ==150678
1680 146456 (update_XCon): Assertion 'xpt->curr_space >= -space_delta'...
1681 146701 ==134990
1682 146781 Adding support for private futexes
1683 147325 valgrind internal error on syscall (SYS_io_destroy, 0)
1684 147498 amd64->IR: 0xF0 0xF 0xB0 0xF (lock cmpxchg %cl,(%rdi))
sewardjbc39ad22007-12-09 02:24:02 +00001685 147545 Memcheck: mc_main.c:817 (get_sec_vbits8): Assertion 'n' failed.
sewardjcb2288b2007-12-02 02:08:17 +00001686 147628 SALC opcode 0xd6 unimplemented
1687 147825 crash on amd64-linux with gcc 4.2 and glibc 2.6 (CFI)
1688 148174 Incorrect type of freed_list_volume causes assertion [...]
1689 148447 x86_64 : new NOP codes: 66 66 66 66 2e 0f 1f
1690 149182 PPC Trap instructions not implemented in valgrind
1691 149504 Assertion hit on alloc_xpt->curr_space >= -space_delta
1692 149519 ppc32: V aborts with SIGSEGV on execution of a signal handler
1693 149892 ==137714
1694 150044 SEGV during stack deregister
1695 150380 dwarf/gcc interoperation (dwarf3 read problems)
1696 150408 ==148447
1697 150678 guest-amd64/toIR.c:3741 (dis_Grp5): Assertion `sz == 4' failed
1698 151209 V unable to execute programs for users with UID > 2^16
1699 151938 help on --db-command= misleading
1700 152022 subw $0x28, %%sp causes assertion failure in memcheck
1701 152357 inb and outb not recognized in 64-bit mode
1702 152501 vex x86->IR: 0x27 0x66 0x89 0x45 (daa)
1703 152818 vex x86->IR: 0xF3 0xAC 0xFC 0x9C (rep lodsb)
1704
njnd5f4bef2006-12-24 07:54:19 +00001705Developer-visible changes:
1706
sewardjcb2288b2007-12-02 02:08:17 +00001707- The names of some functions and types within the Vex IR have
1708 changed. Run 'svn log -r1689 VEX/pub/libvex_ir.h' for full details.
1709 Any existing standalone tools will have to be updated to reflect
1710 these changes. The new names should be clearer. The file
1711 VEX/pub/libvex_ir.h is also much better commented.
njnd8366ee2007-02-12 22:38:50 +00001712
sewardjcb2288b2007-12-02 02:08:17 +00001713- A number of new debugging command line options have been added.
1714 These are mostly of use for debugging the symbol table and line
1715 number readers:
1716
1717 --trace-symtab-patt=<patt> limit debuginfo tracing to obj name <patt>
1718 --trace-cfi=no|yes show call-frame-info details? [no]
1719 --debug-dump=syms mimic /usr/bin/readelf --syms
1720 --debug-dump=line mimic /usr/bin/readelf --debug-dump=line
1721 --debug-dump=frames mimic /usr/bin/readelf --debug-dump=frames
1722 --sym-offsets=yes|no show syms in form 'name+offset' ? [no]
1723
1724- Internally, the code base has been further factorised and
1725 abstractified, particularly with respect to support for non-Linux
1726 OSs.
1727
sewardj135a7ff2007-12-10 22:17:42 +00001728(3.3.0.RC1: 2 Dec 2007, vex r1803, valgrind r7268).
1729(3.3.0.RC2: 5 Dec 2007, vex r1804, valgrind r7282).
1730(3.3.0.RC3: 9 Dec 2007, vex r1804, valgrind r7288).
sewardj686b5be2007-12-10 22:21:14 +00001731(3.3.0: 10 Dec 2007, vex r1804, valgrind r7290).
sewardjcb2288b2007-12-02 02:08:17 +00001732
njnd5f4bef2006-12-24 07:54:19 +00001733
1734
njn99237752007-01-30 05:18:38 +00001735Release 3.2.3 (29 Jan 2007)
1736~~~~~~~~~~~~~~~~~~~~~~~~~~~
1737Unfortunately 3.2.2 introduced a regression which can cause an
1738assertion failure ("vex: the `impossible' happened: eqIRConst") when
1739running obscure pieces of SSE code. 3.2.3 fixes this and adds one
1740more glibc-2.5 intercept. In all other respects it is identical to
17413.2.2. Please do not use (or package) 3.2.2; instead use 3.2.3.
1742
1743n-i-bz vex: the `impossible' happened: eqIRConst
1744n-i-bz Add an intercept for glibc-2.5 __stpcpy_chk
1745
1746(3.2.3: 29 Jan 2007, vex r1732, valgrind r6560).
1747
1748
njna7e48802007-01-23 05:58:52 +00001749Release 3.2.2 (22 Jan 2007)
1750~~~~~~~~~~~~~~~~~~~~~~~~~~~
17513.2.2 fixes a bunch of bugs in 3.2.1, adds support for glibc-2.5 based
1752systems (openSUSE 10.2, Fedora Core 6), improves support for icc-9.X
1753compiled code, and brings modest performance improvements in some
1754areas, including amd64 floating point, powerpc support, and startup
1755responsiveness on all targets.
1756
1757The fixed bugs are as follows. Note that "n-i-bz" stands for "not in
1758bugzilla" -- that is, a bug that was reported to us but never got a
1759bugzilla entry. We encourage you to file bugs in bugzilla
1760(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the
1761developers (or mailing lists) directly.
1762
1763129390 ppc?->IR: some kind of VMX prefetch (dstt)
1764129968 amd64->IR: 0xF 0xAE 0x0 (fxsave)
1765134319 ==129968
1766133054 'make install' fails with syntax errors
1767118903 ==133054
1768132998 startup fails in when running on UML
1769134207 pkg-config output contains @VG_PLATFORM@
1770134727 valgrind exits with "Value too large for defined data type"
1771n-i-bz ppc32/64: support mcrfs
1772n-i-bz Cachegrind/Callgrind: Update cache parameter detection
1773135012 x86->IR: 0xD7 0x8A 0xE0 0xD0 (xlat)
1774125959 ==135012
1775126147 x86->IR: 0xF2 0xA5 0xF 0x77 (repne movsw)
1776136650 amd64->IR: 0xC2 0x8 0x0
1777135421 x86->IR: unhandled Grp5(R) case 6
1778n-i-bz Improved documentation of the IR intermediate representation
1779n-i-bz jcxz (x86) (users list, 8 Nov)
1780n-i-bz ExeContext hashing fix
1781n-i-bz fix CFI reading failures ("Dwarf CFI 0:24 0:32 0:48 0:7")
1782n-i-bz fix Cachegrind/Callgrind simulation bug
1783n-i-bz libmpiwrap.c: fix handling of MPI_LONG_DOUBLE
1784n-i-bz make User errors suppressible
1785136844 corrupted malloc line when using --gen-suppressions=yes
1786138507 ==136844
1787n-i-bz Speed up the JIT's register allocator
1788n-i-bz Fix confusing leak-checker flag hints
1789n-i-bz Support recent autoswamp versions
1790n-i-bz ppc32/64 dispatcher speedups
1791n-i-bz ppc64 front end rld/rlw improvements
1792n-i-bz ppc64 back end imm64 improvements
1793136300 support 64K pages on ppc64-linux
1794139124 == 136300
1795n-i-bz fix ppc insn set tests for gcc >= 4.1
1796137493 x86->IR: recent binutils no-ops
1797137714 x86->IR: 0x66 0xF 0xF7 0xC6 (maskmovdqu)
1798138424 "failed in UME with error 22" (produce a better error msg)
1799138856 ==138424
1800138627 Enhancement support for prctl ioctls
1801138896 Add support for usb ioctls
1802136059 ==138896
1803139050 ppc32->IR: mfspr 268/269 instructions not handled
1804n-i-bz ppc32->IR: lvxl/stvxl
1805n-i-bz glibc-2.5 support
1806n-i-bz memcheck: provide replacement for mempcpy
1807n-i-bz memcheck: replace bcmp in ld.so
1808n-i-bz Use 'ifndef' in VEX's Makefile correctly
1809n-i-bz Suppressions for MVL 4.0.1 on ppc32-linux
1810n-i-bz libmpiwrap.c: Fixes for MPICH
1811n-i-bz More robust handling of hinted client mmaps
1812139776 Invalid read in unaligned memcpy with Intel compiler v9
1813n-i-bz Generate valid XML even for very long fn names
1814n-i-bz Don't prompt about suppressions for unshown reachable leaks
1815139910 amd64 rcl is not supported
1816n-i-bz DWARF CFI reader: handle DW_CFA_undefined
1817n-i-bz DWARF CFI reader: handle icc9 generated CFI info better
1818n-i-bz fix false uninit-value errs in icc9 generated FP code
1819n-i-bz reduce extraneous frames in libmpiwrap.c
1820n-i-bz support pselect6 on amd64-linux
1821
1822(3.2.2: 22 Jan 2007, vex r1729, valgrind r6545).
1823
1824
sewardj172bce82006-09-16 01:00:35 +00001825Release 3.2.1 (16 Sept 2006)
1826~~~~~~~~~~~~~~~~~~~~~~~~~~~~
18273.2.1 adds x86/amd64 support for all SSE3 instructions except monitor
1828and mwait, further reduces memcheck's false error rate on all
1829platforms, adds support for recent binutils (in OpenSUSE 10.2 and
1830Fedora Rawhide) and fixes a bunch of bugs in 3.2.0. Some of the fixed
1831bugs were causing large programs to segfault with --tool=callgrind and
1832--tool=cachegrind, so an upgrade is recommended.
1833
1834In view of the fact that any 3.3.0 release is unlikely to happen until
1835well into 1Q07, we intend to keep the 3.2.X line alive for a while
1836yet, and so we tentatively plan a 3.2.2 release sometime in December
183706.
1838
1839The fixed bugs are as follows. Note that "n-i-bz" stands for "not in
1840bugzilla" -- that is, a bug that was reported to us but never got a
1841bugzilla entry.
1842
1843n-i-bz Expanding brk() into last available page asserts
1844n-i-bz ppc64-linux stack RZ fast-case snafu
1845n-i-bz 'c' in --gen-supps=yes doesn't work
1846n-i-bz VG_N_SEGMENTS too low (users, 28 June)
1847n-i-bz VG_N_SEGNAMES too low (Stu Robinson)
1848106852 x86->IR: fisttp (SSE3)
1849117172 FUTEX_WAKE does not use uaddr2
1850124039 Lacks support for VKI_[GP]IO_UNIMAP*
1851127521 amd64->IR: 0xF0 0x48 0xF 0xC7 (cmpxchg8b)
1852128917 amd64->IR: 0x66 0xF 0xF6 0xC4 (psadbw,SSE2)
1853129246 JJ: ppc32/ppc64 syscalls, w/ patch
1854129358 x86->IR: fisttpl (SSE3)
1855129866 cachegrind/callgrind causes executable to die
1856130020 Can't stat .so/.exe error while reading symbols
1857130388 Valgrind aborts when process calls malloc_trim()
1858130638 PATCH: ppc32 missing system calls
1859130785 amd64->IR: unhandled instruction "pushfq"
1860131481: (HINT_NOP) vex x86->IR: 0xF 0x1F 0x0 0xF
1861131298 ==131481
1862132146 Programs with long sequences of bswap[l,q]s
1863132918 vex amd64->IR: 0xD9 0xF8 (fprem)
1864132813 Assertion at priv/guest-x86/toIR.c:652 fails
1865133051 'cfsi->len > 0 && cfsi->len < 2000000' failed
1866132722 valgrind header files are not standard C
1867n-i-bz Livelocks entire machine (users list, Timothy Terriberry)
1868n-i-bz Alex Bennee mmap problem (9 Aug)
1869n-i-bz BartV: Don't print more lines of a stack-trace than were obtained.
1870n-i-bz ppc32 SuSE 10.1 redir
1871n-i-bz amd64 padding suppressions
1872n-i-bz amd64 insn printing fix.
1873n-i-bz ppc cmp reg,reg fix
1874n-i-bz x86/amd64 iropt e/rflag reduction rules
1875n-i-bz SuSE 10.1 (ppc32) minor fixes
1876133678 amd64->IR: 0x48 0xF 0xC5 0xC0 (pextrw?)
1877133694 aspacem assertion: aspacem_minAddr <= holeStart
1878n-i-bz callgrind: fix warning about malformed creator line
1879n-i-bz callgrind: fix annotate script for data produced with
1880 --dump-instr=yes
1881n-i-bz callgrind: fix failed assertion when toggling
1882 instrumentation mode
1883n-i-bz callgrind: fix annotate script fix warnings with
1884 --collect-jumps=yes
1885n-i-bz docs path hardwired (Dennis Lubert)
1886
1887The following bugs were not fixed, due primarily to lack of developer
1888time, and also because bug reporters did not answer requests for
1889feedback in time for the release:
1890
1891129390 ppc?->IR: some kind of VMX prefetch (dstt)
1892129968 amd64->IR: 0xF 0xAE 0x0 (fxsave)
1893133054 'make install' fails with syntax errors
1894n-i-bz Signal race condition (users list, 13 June, Johannes Berg)
1895n-i-bz Unrecognised instruction at address 0x70198EC2 (users list,
1896 19 July, Bennee)
1897132998 startup fails in when running on UML
1898
1899The following bug was tentatively fixed on the mainline but the fix
1900was considered too risky to push into 3.2.X:
1901
1902133154 crash when using client requests to register/deregister stack
1903
1904(3.2.1: 16 Sept 2006, vex r1658, valgrind r6070).
1905
1906
sewardj3ecf8f22006-06-06 21:41:22 +00001907Release 3.2.0 (7 June 2006)
sewardj535a2352006-04-18 02:04:52 +00001908~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardj1ede80f2006-05-26 00:47:13 +000019093.2.0 is a feature release with many significant improvements and the
1910usual collection of bug fixes. This release supports X86/Linux,
1911AMD64/Linux, PPC32/Linux and PPC64/Linux.
njna3e96762006-03-27 12:23:44 +00001912
sewardjaf924992006-05-26 13:51:37 +00001913Performance, especially of Memcheck, is improved, Addrcheck has been
1914removed, Callgrind has been added, PPC64/Linux support has been added,
1915Lackey has been improved, and MPI support has been added. In detail:
njna3e96762006-03-27 12:23:44 +00001916
sewardjf41db3c2006-05-26 16:06:10 +00001917- Memcheck has improved speed and reduced memory use. Run times are
1918 typically reduced by 15-30%, averaging about 24% for SPEC CPU2000.
sewardjaf924992006-05-26 13:51:37 +00001919 The other tools have smaller but noticeable speed improvments. We
1920 are interested to hear what improvements users get.
njna3e96762006-03-27 12:23:44 +00001921
sewardj3ecf8f22006-06-06 21:41:22 +00001922 Memcheck uses less memory due to the introduction of a compressed
sewardjaf924992006-05-26 13:51:37 +00001923 representation for shadow memory. The space overhead has been
1924 reduced by a factor of up to four, depending on program behaviour.
1925 This means you should be able to run programs that use more memory
1926 than before without hitting problems.
njna3e96762006-03-27 12:23:44 +00001927
sewardj1ede80f2006-05-26 00:47:13 +00001928- Addrcheck has been removed. It has not worked since version 2.4.0,
1929 and the speed and memory improvements to Memcheck make it redundant.
1930 If you liked using Addrcheck because it didn't give undefined value
1931 errors, you can use the new Memcheck option --undef-value-errors=no
1932 to get the same behaviour.
1933
sewardjaf924992006-05-26 13:51:37 +00001934- The number of undefined-value errors incorrectly reported by
1935 Memcheck has been reduced (such false reports were already very
1936 rare). In particular, efforts have been made to ensure Memcheck
1937 works really well with gcc 4.0/4.1-generated code on X86/Linux and
1938 AMD64/Linux.
sewardj1ede80f2006-05-26 00:47:13 +00001939
1940- Josef Weidendorfer's popular Callgrind tool has been added. Folding
sewardjaf924992006-05-26 13:51:37 +00001941 it in was a logical step given its popularity and usefulness, and
sewardj1ede80f2006-05-26 00:47:13 +00001942 makes it easier for us to ensure it works "out of the box" on all
1943 supported targets. The associated KDE KCachegrind GUI remains a
1944 separate project.
njneaf0ca92006-04-09 01:23:29 +00001945
sewardj3ecf8f22006-06-06 21:41:22 +00001946- A new release of the Valkyrie GUI for Memcheck, version 1.2.0,
1947 accompanies this release. Improvements over previous releases
1948 include improved robustness, many refinements to the user interface,
1949 and use of a standard autoconf/automake build system. You can get
1950 it from http://www.valgrind.org/downloads/guis.html.
1951
sewardj535a2352006-04-18 02:04:52 +00001952- Valgrind now works on PPC64/Linux. As with the AMD64/Linux port,
sewardj1ede80f2006-05-26 00:47:13 +00001953 this supports programs using to 32G of address space. On 64-bit
1954 capable PPC64/Linux setups, you get a dual architecture build so
1955 that both 32-bit and 64-bit executables can be run. Linux on POWER5
1956 is supported, and POWER4 is also believed to work. Both 32-bit and
1957 64-bit DWARF2 is supported. This port is known to work well with
1958 both gcc-compiled and xlc/xlf-compiled code.
njna3e96762006-03-27 12:23:44 +00001959
sewardj1ede80f2006-05-26 00:47:13 +00001960- Floating point accuracy has been improved for PPC32/Linux.
1961 Specifically, the floating point rounding mode is observed on all FP
1962 arithmetic operations, and multiply-accumulate instructions are
1963 preserved by the compilation pipeline. This means you should get FP
1964 results which are bit-for-bit identical to a native run. These
1965 improvements are also present in the PPC64/Linux port.
njnb2939162006-04-18 00:52:37 +00001966
njneaf0ca92006-04-09 01:23:29 +00001967- Lackey, the example tool, has been improved:
njna3e96762006-03-27 12:23:44 +00001968
sewardj1ede80f2006-05-26 00:47:13 +00001969 * It has a new option --detailed-counts (off by default) which
1970 causes it to print out a count of loads, stores and ALU operations
1971 done, and their sizes.
njndbf7ca72006-03-31 11:57:59 +00001972
sewardj1ede80f2006-05-26 00:47:13 +00001973 * It has a new option --trace-mem (off by default) which causes it
1974 to print out a trace of all memory accesses performed by a
1975 program. It's a good starting point for building Valgrind tools
1976 that need to track memory accesses. Read the comments at the top
1977 of the file lackey/lk_main.c for details.
sewardj42fcf5d2006-04-08 13:16:36 +00001978
sewardj1ede80f2006-05-26 00:47:13 +00001979 * The original instrumentation (counting numbers of instructions,
1980 jumps, etc) is now controlled by a new option --basic-counts. It
1981 is on by default.
sewardj42fcf5d2006-04-08 13:16:36 +00001982
1983- MPI support: partial support for debugging distributed applications
sewardj1ede80f2006-05-26 00:47:13 +00001984 using the MPI library specification has been added. Valgrind is
sewardj42fcf5d2006-04-08 13:16:36 +00001985 aware of the memory state changes caused by a subset of the MPI
1986 functions, and will carefully check data passed to the (P)MPI_
1987 interface.
1988
sewardj1ede80f2006-05-26 00:47:13 +00001989- A new flag, --error-exitcode=, has been added. This allows changing
1990 the exit code in runs where Valgrind reported errors, which is
1991 useful when using Valgrind as part of an automated test suite.
sewardjb9779082006-05-12 23:50:15 +00001992
sewardj1ede80f2006-05-26 00:47:13 +00001993- Various segfaults when reading old-style "stabs" debug information
1994 have been fixed.
njneaf0ca92006-04-09 01:23:29 +00001995
sewardjc9d09cc2006-05-27 03:22:04 +00001996- A simple performance evaluation suite has been added. See
njn5359b6f2006-06-02 23:57:22 +00001997 perf/README and README_DEVELOPERS for details. There are
sewardjc9d09cc2006-05-27 03:22:04 +00001998 various bells and whistles.
1999
sewardjb70a6132006-05-27 21:14:09 +00002000- New configuration flags:
2001 --enable-only32bit
2002 --enable-only64bit
2003 By default, on 64 bit platforms (ppc64-linux, amd64-linux) the build
2004 system will attempt to build a Valgrind which supports both 32-bit
2005 and 64-bit executables. This may not be what you want, and you can
2006 override the default behaviour using these flags.
2007
sewardj1ede80f2006-05-26 00:47:13 +00002008Please note that Helgrind is still not working. We have made an
2009important step towards making it work again, however, with the
2010addition of function wrapping (see below).
njneaf0ca92006-04-09 01:23:29 +00002011
2012Other user-visible changes:
2013
sewardj1ede80f2006-05-26 00:47:13 +00002014- Valgrind now has the ability to intercept and wrap arbitrary
2015 functions. This is a preliminary step towards making Helgrind work
2016 again, and was required for MPI support.
njneaf0ca92006-04-09 01:23:29 +00002017
sewardj1ede80f2006-05-26 00:47:13 +00002018- There are some changes to Memcheck's client requests. Some of them
2019 have changed names:
njndbf7ca72006-03-31 11:57:59 +00002020
2021 MAKE_NOACCESS --> MAKE_MEM_NOACCESS
2022 MAKE_WRITABLE --> MAKE_MEM_UNDEFINED
2023 MAKE_READABLE --> MAKE_MEM_DEFINED
2024
2025 CHECK_WRITABLE --> CHECK_MEM_IS_ADDRESSABLE
2026 CHECK_READABLE --> CHECK_MEM_IS_DEFINED
2027 CHECK_DEFINED --> CHECK_VALUE_IS_DEFINED
2028
sewardj1ede80f2006-05-26 00:47:13 +00002029 The reason for the change is that the old names are subtly
2030 misleading. The old names will still work, but they are deprecated
2031 and may be removed in a future release.
njndbf7ca72006-03-31 11:57:59 +00002032
2033 We also added a new client request:
2034
2035 MAKE_MEM_DEFINED_IF_ADDRESSABLE(a, len)
2036
2037 which is like MAKE_MEM_DEFINED but only affects a byte if the byte is
2038 already addressable.
2039
sewardj3ecf8f22006-06-06 21:41:22 +00002040- The way client requests are encoded in the instruction stream has
2041 changed. Unfortunately, this means 3.2.0 will not honour client
2042 requests compiled into binaries using headers from earlier versions
2043 of Valgrind. We will try to keep the client request encodings more
2044 stable in future.
njneaf0ca92006-04-09 01:23:29 +00002045
njna3e96762006-03-27 12:23:44 +00002046BUGS FIXED:
2047
sewardj1ede80f2006-05-26 00:47:13 +00002048108258 NPTL pthread cleanup handlers not called
2049117290 valgrind is sigKILL'd on startup
2050117295 == 117290
2051118703 m_signals.c:1427 Assertion 'tst->status == VgTs_WaitSys'
2052118466 add %reg, %reg generates incorrect validity for bit 0
2053123210 New: strlen from ld-linux on amd64
2054123244 DWARF2 CFI reader: unhandled CFI instruction 0:18
2055123248 syscalls in glibc-2.4: openat, fstatat, symlinkat
2056123258 socketcall.recvmsg(msg.msg_iov[i] points to uninit
2057123535 mremap(new_addr) requires MREMAP_FIXED in 4th arg
2058123836 small typo in the doc
2059124029 ppc compile failed: `vor' gcc 3.3.5
2060124222 Segfault: @@don't know what type ':' is
2061124475 ppc32: crash (syscall?) timer_settime()
2062124499 amd64->IR: 0xF 0xE 0x48 0x85 (femms)
2063124528 FATAL: aspacem assertion failed: segment_is_sane
2064124697 vex x86->IR: 0xF 0x70 0xC9 0x0 (pshufw)
2065124892 vex x86->IR: 0xF3 0xAE (REPx SCASB)
2066126216 == 124892
2067124808 ppc32: sys_sched_getaffinity() not handled
2068n-i-bz Very long stabs strings crash m_debuginfo
2069n-i-bz amd64->IR: 0x66 0xF 0xF5 (pmaddwd)
2070125492 ppc32: support a bunch more syscalls
2071121617 ppc32/64: coredumping gives assertion failure
2072121814 Coregrind return error as exitcode patch
2073126517 == 121814
sewardj1ede80f2006-05-26 00:47:13 +00002074125607 amd64->IR: 0x66 0xF 0xA3 0x2 (btw etc)
2075125651 amd64->IR: 0xF8 0x49 0xFF 0xE3 (clc?)
2076126253 x86 movx is wrong
2077126451 3.2 SVN doesn't work on ppc32 CPU's without FPU
2078126217 increase # threads
2079126243 vex x86->IR: popw mem
2080126583 amd64->IR: 0x48 0xF 0xA4 0xC2 (shld $1,%rax,%rdx)
sewardjb70a6132006-05-27 21:14:09 +00002081126668 amd64->IR: 0x1C 0xFF (sbb $0xff,%al)
2082126696 support for CDROMREADRAW ioctl and CDROMREADTOCENTRY fix
2083126722 assertion: segment_is_sane at m_aspacemgr/aspacemgr.c:1624
2084126938 bad checking for syscalls linkat, renameat, symlinkat
sewardj1ede80f2006-05-26 00:47:13 +00002085
sewardj3ecf8f22006-06-06 21:41:22 +00002086(3.2.0RC1: 27 May 2006, vex r1626, valgrind r5947).
2087(3.2.0: 7 June 2006, vex r1628, valgrind r5957).
njna3e96762006-03-27 12:23:44 +00002088
2089
sewardj50654672006-03-16 11:33:14 +00002090Release 3.1.1 (15 March 2006)
2091~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
20923.1.1 fixes a bunch of bugs reported in 3.1.0. There is no new
2093functionality. The fixed bugs are:
2094
2095(note: "n-i-bz" means "not in bugzilla" -- this bug does not have
2096 a bugzilla entry).
2097
2098n-i-bz ppc32: fsub 3,3,3 in dispatcher doesn't clear NaNs
2099n-i-bz ppc32: __NR_{set,get}priority
2100117332 x86: missing line info with icc 8.1
2101117366 amd64: 0xDD 0x7C fnstsw
2102118274 == 117366
2103117367 amd64: 0xD9 0xF4 fxtract
2104117369 amd64: __NR_getpriority (140)
2105117419 ppc32: lfsu f5, -4(r11)
2106117419 ppc32: fsqrt
2107117936 more stabs problems (segfaults while reading debug info)
2108119914 == 117936
2109120345 == 117936
2110118239 amd64: 0xF 0xAE 0x3F (clflush)
2111118939 vm86old system call
2112n-i-bz memcheck/tests/mempool reads freed memory
2113n-i-bz AshleyP's custom-allocator assertion
2114n-i-bz Dirk strict-aliasing stuff
2115n-i-bz More space for debugger cmd line (Dan Thaler)
2116n-i-bz Clarified leak checker output message
2117n-i-bz AshleyP's --gen-suppressions output fix
2118n-i-bz cg_annotate's --sort option broken
2119n-i-bz OSet 64-bit fastcmp bug
2120n-i-bz VG_(getgroups) fix (Shinichi Noda)
2121n-i-bz ppc32: allocate from callee-saved FP/VMX regs
2122n-i-bz misaligned path word-size bug in mc_main.c
2123119297 Incorrect error message for sse code
2124120410 x86: prefetchw (0xF 0xD 0x48 0x4)
2125120728 TIOCSERGETLSR, TIOCGICOUNT, HDIO_GET_DMA ioctls
2126120658 Build fixes for gcc 2.96
2127120734 x86: Support for changing EIP in signal handler
2128n-i-bz memcheck/tests/zeropage de-looping fix
2129n-i-bz x86: fxtract doesn't work reliably
2130121662 x86: lock xadd (0xF0 0xF 0xC0 0x2)
2131121893 calloc does not always return zeroed memory
2132121901 no support for syscall tkill
2133n-i-bz Suppression update for Debian unstable
2134122067 amd64: fcmovnu (0xDB 0xD9)
2135n-i-bz ppc32: broken signal handling in cpu feature detection
2136n-i-bz ppc32: rounding mode problems (improved, partial fix only)
2137119482 ppc32: mtfsb1
2138n-i-bz ppc32: mtocrf/mfocrf
2139
2140(3.1.1: 15 March 2006, vex r1597, valgrind r5771).
2141
jseward43230652003-12-14 00:14:54 +00002142
sewardj4bbe9d72005-11-16 18:39:25 +00002143Release 3.1.0 (25 November 2005)
njnf63170c2005-11-12 19:09:51 +00002144~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardj4bbe9d72005-11-16 18:39:25 +000021453.1.0 is a feature release with a number of significant improvements:
2146AMD64 support is much improved, PPC32 support is good enough to be
2147usable, and the handling of memory management and address space is
2148much more robust. In detail:
njnf63170c2005-11-12 19:09:51 +00002149
sewardj4bbe9d72005-11-16 18:39:25 +00002150- AMD64 support is much improved. The 64-bit vs. 32-bit issues in
2151 3.0.X have been resolved, and it should "just work" now in all
2152 cases. On AMD64 machines both 64-bit and 32-bit versions of
2153 Valgrind are built. The right version will be invoked
2154 automatically, even when using --trace-children and mixing execution
2155 between 64-bit and 32-bit executables. Also, many more instructions
2156 are supported.
njnf63170c2005-11-12 19:09:51 +00002157
sewardj4bbe9d72005-11-16 18:39:25 +00002158- PPC32 support is now good enough to be usable. It should work with
2159 all tools, but please let us know if you have problems. Three
2160 classes of CPUs are supported: integer only (no FP, no Altivec),
2161 which covers embedded PPC uses, integer and FP but no Altivec
2162 (G3-ish), and CPUs capable of Altivec too (G4, G5).
njnf63170c2005-11-12 19:09:51 +00002163
sewardj4bbe9d72005-11-16 18:39:25 +00002164- Valgrind's address space management has been overhauled. As a
2165 result, Valgrind should be much more robust with programs that use
2166 large amounts of memory. There should be many fewer "memory
2167 exhausted" messages, and debug symbols should be read correctly on
2168 large (eg. 300MB+) executables. On 32-bit machines the full address
2169 space available to user programs (usually 3GB or 4GB) can be fully
2170 utilised. On 64-bit machines up to 32GB of space is usable; when
2171 using Memcheck that means your program can use up to about 14GB.
njnf63170c2005-11-12 19:09:51 +00002172
2173 A side effect of this change is that Valgrind is no longer protected
2174 against wild writes by the client. This feature was nice but relied
2175 on the x86 segment registers and so wasn't portable.
2176
2177- Most users should not notice, but as part of the address space
sewardj4bbe9d72005-11-16 18:39:25 +00002178 manager change, the way Valgrind is built has been changed. Each
2179 tool is now built as a statically linked stand-alone executable,
2180 rather than as a shared object that is dynamically linked with the
2181 core. The "valgrind" program invokes the appropriate tool depending
2182 on the --tool option. This slightly increases the amount of disk
2183 space used by Valgrind, but it greatly simplified many things and
2184 removed Valgrind's dependence on glibc.
njnf63170c2005-11-12 19:09:51 +00002185
njn336c89d2005-11-20 18:14:04 +00002186Please note that Addrcheck and Helgrind are still not working. Work
2187is underway to reinstate them (or equivalents). We apologise for the
2188inconvenience.
2189
njnf63170c2005-11-12 19:09:51 +00002190Other user-visible changes:
2191
2192- The --weird-hacks option has been renamed --sim-hints.
2193
2194- The --time-stamp option no longer gives an absolute date and time.
2195 It now prints the time elapsed since the program began.
2196
sewardj4bbe9d72005-11-16 18:39:25 +00002197- It should build with gcc-2.96.
2198
njn7cce5b82005-11-16 20:12:22 +00002199- Valgrind can now run itself (see README_DEVELOPERS for how).
sewardj47058612005-11-16 19:13:13 +00002200 This is not much use to you, but it means the developers can now
2201 profile Valgrind using Cachegrind. As a result a couple of
2202 performance bad cases have been fixed.
2203
sewardj6a5a69c2005-11-17 00:51:36 +00002204- The XML output format has changed slightly. See
2205 docs/internals/xml-output.txt.
2206
njn14c9add2005-11-18 17:36:01 +00002207- Core dumping has been reinstated (it was disabled in 3.0.0 and 3.0.1).
2208 If your program crashes while running under Valgrind, a core file with
2209 the name "vgcore.<pid>" will be created (if your settings allow core
2210 file creation). Note that the floating point information is not all
2211 there. If Valgrind itself crashes, the OS will create a normal core
2212 file.
2213
njnf63170c2005-11-12 19:09:51 +00002214The following are some user-visible changes that occurred in earlier
2215versions that may not have been announced, or were announced but not
njn7cce5b82005-11-16 20:12:22 +00002216widely noticed. So we're mentioning them now.
njnf63170c2005-11-12 19:09:51 +00002217
2218- The --tool flag is optional once again; if you omit it, Memcheck
2219 is run by default.
2220
2221- The --num-callers flag now has a default value of 12. It was
2222 previously 4.
2223
njn7cce5b82005-11-16 20:12:22 +00002224- The --xml=yes flag causes Valgrind's output to be produced in XML
2225 format. This is designed to make it easy for other programs to
2226 consume Valgrind's output. The format is described in the file
njnf63170c2005-11-12 19:09:51 +00002227 docs/internals/xml-format.txt.
2228
njn1f503992005-11-13 16:41:15 +00002229- The --gen-suppressions flag supports an "all" value that causes every
2230 suppression to be printed without asking.
2231
njnee88d282005-11-20 17:19:28 +00002232- The --log-file option no longer puts "pid" in the filename, eg. the
2233 old name "foo.pid12345" is now "foo.12345".
2234
njnc66077e2005-11-20 18:02:15 +00002235- There are several graphical front-ends for Valgrind, such as Valkyrie,
2236 Alleyoop and Valgui. See http://www.valgrind.org/downloads/guis.html
2237 for a list.
2238
njnf63170c2005-11-12 19:09:51 +00002239BUGS FIXED:
2240
sewardj4bbe9d72005-11-16 18:39:25 +00002241109861 amd64 hangs at startup
2242110301 ditto
2243111554 valgrind crashes with Cannot allocate memory
2244111809 Memcheck tool doesn't start java
2245111901 cross-platform run of cachegrind fails on opteron
2246113468 (vgPlain_mprotect_range): Assertion 'r != -1' failed.
2247 92071 Reading debugging info uses too much memory
2248109744 memcheck loses track of mmap from direct ld-linux.so.2
2249110183 tail of page with _end
2250 82301 FV memory layout too rigid
2251 98278 Infinite recursion possible when allocating memory
2252108994 Valgrind runs out of memory due to 133x overhead
2253115643 valgrind cannot allocate memory
2254105974 vg_hashtable.c static hash table
2255109323 ppc32: dispatch.S uses Altivec insn, which doesn't work on POWER.
2256109345 ptrace_setregs not yet implemented for ppc
2257110831 Would like to be able to run against both 32 and 64 bit
2258 binaries on AMD64
2259110829 == 110831
2260111781 compile of valgrind-3.0.0 fails on my linux (gcc 2.X prob)
2261112670 Cachegrind: cg_main.c:486 (handleOneStatement ...
2262112941 vex x86: 0xD9 0xF4 (fxtract)
2263110201 == 112941
2264113015 vex amd64->IR: 0xE3 0x14 0x48 0x83 (jrcxz)
2265113126 Crash with binaries built with -gstabs+/-ggdb
2266104065 == 113126
2267115741 == 113126
2268113403 Partial SSE3 support on x86
2269113541 vex: Grp5(x86) (alt encoding inc/dec) case 1
2270113642 valgrind crashes when trying to read debug information
2271113810 vex x86->IR: 66 0F F6 (66 + PSADBW == SSE PSADBW)
2272113796 read() and write() do not work if buffer is in shared memory
2273113851 vex x86->IR: (pmaddwd): 0x66 0xF 0xF5 0xC7
2274114366 vex amd64 cannnot handle __asm__( "fninit" )
2275114412 vex amd64->IR: 0xF 0xAD 0xC2 0xD3 (128-bit shift, shrdq?)
2276114455 vex amd64->IR: 0xF 0xAC 0xD0 0x1 (also shrdq)
2277115590: amd64->IR: 0x67 0xE3 0x9 0xEB (address size override)
2278115953 valgrind svn r5042 does not build with parallel make (-j3)
2279116057 maximum instruction size - VG_MAX_INSTR_SZB too small?
2280116483 shmat failes with invalid argument
2281102202 valgrind crashes when realloc'ing until out of memory
2282109487 == 102202
2283110536 == 102202
2284112687 == 102202
2285111724 vex amd64->IR: 0x41 0xF 0xAB (more BT{,S,R,C} fun n games)
2286111748 vex amd64->IR: 0xDD 0xE2 (fucom)
2287111785 make fails if CC contains spaces
2288111829 vex x86->IR: sbb AL, Ib
2289111851 vex x86->IR: 0x9F 0x89 (lahf/sahf)
2290112031 iopl on AMD64 and README_MISSING_SYSCALL_OR_IOCTL update
2291112152 code generation for Xin_MFence on x86 with SSE0 subarch
2292112167 == 112152
2293112789 == 112152
2294112199 naked ar tool is used in vex makefile
2295112501 vex x86->IR: movq (0xF 0x7F 0xC1 0xF) (mmx MOVQ)
2296113583 == 112501
2297112538 memalign crash
2298113190 Broken links in docs/html/
2299113230 Valgrind sys_pipe on x86-64 wrongly thinks file descriptors
2300 should be 64bit
2301113996 vex amd64->IR: fucomp (0xDD 0xE9)
2302114196 vex x86->IR: out %eax,(%dx) (0xEF 0xC9 0xC3 0x90)
2303114289 Memcheck fails to intercept malloc when used in an uclibc environment
2304114756 mbind syscall support
2305114757 Valgrind dies with assertion: Assertion 'noLargerThan > 0' failed
2306114563 stack tracking module not informed when valgrind switches threads
2307114564 clone() and stacks
2308114565 == 114564
2309115496 glibc crashes trying to use sysinfo page
2310116200 enable fsetxattr, fgetxattr, and fremovexattr for amd64
njnf63170c2005-11-12 19:09:51 +00002311
sewardjae17e7e2005-11-20 19:45:30 +00002312(3.1.0RC1: 20 November 2005, vex r1466, valgrind r5224).
sewardjfa462022005-11-26 13:38:58 +00002313(3.1.0: 26 November 2005, vex r1471, valgrind r5235).
sewardjae17e7e2005-11-20 19:45:30 +00002314
njnf63170c2005-11-12 19:09:51 +00002315
sewardjcd62fe72005-08-29 13:24:51 +00002316Release 3.0.1 (29 August 2005)
2317~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
23183.0.1 fixes a bunch of bugs reported in 3.0.0. There is no new
2319functionality. Some of the fixed bugs are critical, so if you
sewardjfa233892005-08-30 12:35:36 +00002320use/distribute 3.0.0, an upgrade to 3.0.1 is recommended. The fixed
sewardjcd62fe72005-08-29 13:24:51 +00002321bugs are:
2322
2323(note: "n-i-bz" means "not in bugzilla" -- this bug does not have
2324 a bugzilla entry).
2325
2326109313 (== 110505) x86 cmpxchg8b
2327n-i-bz x86: track but ignore changes to %eflags.AC (alignment check)
2328110102 dis_op2_E_G(amd64)
2329110202 x86 sys_waitpid(#286)
2330110203 clock_getres(,0)
2331110208 execve fail wrong retval
2332110274 SSE1 now mandatory for x86
2333110388 amd64 0xDD 0xD1
2334110464 amd64 0xDC 0x1D FCOMP
2335110478 amd64 0xF 0xD PREFETCH
2336n-i-bz XML <unique> printing wrong
2337n-i-bz Dirk r4359 (amd64 syscalls from trunk)
2338110591 amd64 and x86: rdtsc not implemented properly
2339n-i-bz Nick r4384 (stub implementations of Addrcheck and Helgrind)
2340110652 AMD64 valgrind crashes on cwtd instruction
2341110653 AMD64 valgrind crashes on sarb $0x4,foo(%rip) instruction
2342110656 PATH=/usr/bin::/bin valgrind foobar stats ./fooba
2343110657 Small test fixes
2344110671 vex x86->IR: unhandled instruction bytes: 0xF3 0xC3 (rep ret)
2345n-i-bz Nick (Cachegrind should not assert when it encounters a client
2346 request.)
2347110685 amd64->IR: unhandled instruction bytes: 0xE1 0x56 (loope Jb)
2348110830 configuring with --host fails to build 32 bit on 64 bit target
2349110875 Assertion when execve fails
2350n-i-bz Updates to Memcheck manual
2351n-i-bz Fixed broken malloc_usable_size()
2352110898 opteron instructions missing: btq btsq btrq bsfq
2353110954 x86->IR: unhandled instruction bytes: 0xE2 0xF6 (loop Jb)
2354n-i-bz Make suppressions work for "???" lines in stacktraces.
2355111006 bogus warnings from linuxthreads
2356111092 x86: dis_Grp2(Reg): unhandled case(x86)
2357111231 sctp_getladdrs() and sctp_getpaddrs() returns uninitialized
2358 memory
2359111102 (comment #4) Fixed 64-bit unclean "silly arg" message
2360n-i-bz vex x86->IR: unhandled instruction bytes: 0x14 0x0
2361n-i-bz minor umount/fcntl wrapper fixes
2362111090 Internal Error running Massif
2363101204 noisy warning
2364111513 Illegal opcode for SSE instruction (x86 movups)
2365111555 VEX/Makefile: CC is set to gcc
sewardjfa233892005-08-30 12:35:36 +00002366n-i-bz Fix XML bugs in FAQ
sewardjcd62fe72005-08-29 13:24:51 +00002367
sewardjfa233892005-08-30 12:35:36 +00002368(3.0.1: 29 August 05,
2369 vex/branches/VEX_3_0_BRANCH r1367,
2370 valgrind/branches/VALGRIND_3_0_BRANCH r4574).
sewardj820c7722005-08-29 13:44:43 +00002371
2372
sewardjcd62fe72005-08-29 13:24:51 +00002373
sewardjec39f332005-08-02 15:28:07 +00002374Release 3.0.0 (3 August 2005)
2375~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardjf9929e62005-08-02 21:05:58 +000023763.0.0 is a major overhaul of Valgrind. The most significant user
2377visible change is that Valgrind now supports architectures other than
2378x86. The new architectures it supports are AMD64 and PPC32, and the
2379infrastructure is present for other architectures to be added later.
njn4a9f6f42005-07-19 23:46:08 +00002380
sewardjf9929e62005-08-02 21:05:58 +00002381AMD64 support works well, but has some shortcomings:
njn4a9f6f42005-07-19 23:46:08 +00002382
njn7b1a6a42005-07-20 04:20:44 +00002383- It generally won't be as solid as the x86 version. For example,
2384 support for more obscure instructions and system calls may be missing.
2385 We will fix these as they arise.
njn4a9f6f42005-07-19 23:46:08 +00002386
sewardjc371bd02005-07-27 09:55:24 +00002387- Address space may be limited; see the point about
njn7b1a6a42005-07-20 04:20:44 +00002388 position-independent executables below.
njn4a9f6f42005-07-19 23:46:08 +00002389
njn7b1a6a42005-07-20 04:20:44 +00002390- If Valgrind is built on an AMD64 machine, it will only run 64-bit
2391 executables. If you want to run 32-bit x86 executables under Valgrind
2392 on an AMD64, you will need to build Valgrind on an x86 machine and
2393 copy it to the AMD64 machine. And it probably won't work if you do
2394 something tricky like exec'ing a 32-bit program from a 64-bit program
2395 while using --trace-children=yes. We hope to improve this situation
2396 in the future.
2397
2398The PPC32 support is very basic. It may not work reliably even for
sewardjc371bd02005-07-27 09:55:24 +00002399small programs, but it's a start. Many thanks to Paul Mackerras for
2400his great work that enabled this support. We are working to make
2401PPC32 usable as soon as possible.
njn4a9f6f42005-07-19 23:46:08 +00002402
2403Other user-visible changes:
2404
sewardjec39f332005-08-02 15:28:07 +00002405- Valgrind is no longer built by default as a position-independent
2406 executable (PIE), as this caused too many problems.
njn4a9f6f42005-07-19 23:46:08 +00002407
2408 Without PIE enabled, AMD64 programs will only be able to access 2GB of
2409 address space. We will fix this eventually, but not for the moment.
2410
2411 Use --enable-pie at configure-time to turn this on.
2412
2413- Support for programs that use stack-switching has been improved. Use
2414 the --max-stackframe flag for simple cases, and the
2415 VALGRIND_STACK_REGISTER, VALGRIND_STACK_DEREGISTER and
2416 VALGRIND_STACK_CHANGE client requests for trickier cases.
2417
njn7b1a6a42005-07-20 04:20:44 +00002418- Support for programs that use self-modifying code has been improved,
2419 in particular programs that put temporary code fragments on the stack.
2420 This helps for C programs compiled with GCC that use nested functions,
njne7723322005-07-24 17:19:09 +00002421 and also Ada programs. This is controlled with the --smc-check
njn7b1a6a42005-07-20 04:20:44 +00002422 flag, although the default setting should work in most cases.
njn4a9f6f42005-07-19 23:46:08 +00002423
njn7b1a6a42005-07-20 04:20:44 +00002424- Output can now be printed in XML format. This should make it easier
sewardjc371bd02005-07-27 09:55:24 +00002425 for tools such as GUI front-ends and automated error-processing
2426 schemes to use Valgrind output as input. The --xml flag controls this.
2427 As part of this change, ELF directory information is read from executables,
2428 so absolute source file paths are available if needed.
njn4a9f6f42005-07-19 23:46:08 +00002429
2430- Programs that allocate many heap blocks may run faster, due to
2431 improvements in certain data structures.
2432
njn7b1a6a42005-07-20 04:20:44 +00002433- Addrcheck is currently not working. We hope to get it working again
2434 soon. Helgrind is still not working, as was the case for the 2.4.0
2435 release.
njn4a9f6f42005-07-19 23:46:08 +00002436
sewardjc371bd02005-07-27 09:55:24 +00002437- The JITter has been completely rewritten, and is now in a separate
2438 library, called Vex. This enabled a lot of the user-visible changes,
2439 such as new architecture support. The new JIT unfortunately translates
2440 more slowly than the old one, so programs may take longer to start.
2441 We believe the code quality is produces is about the same, so once
2442 started, programs should run at about the same speed. Feedback about
2443 this would be useful.
2444
2445 On the plus side, Vex and hence Memcheck tracks value flow properly
2446 through floating point and vector registers, something the 2.X line
2447 could not do. That means that Memcheck is much more likely to be
2448 usably accurate on vectorised code.
2449
sewardjec39f332005-08-02 15:28:07 +00002450- There is a subtle change to the way exiting of threaded programs
sewardjc371bd02005-07-27 09:55:24 +00002451 is handled. In 3.0, Valgrind's final diagnostic output (leak check,
2452 etc) is not printed until the last thread exits. If the last thread
2453 to exit was not the original thread which started the program, any
2454 other process wait()-ing on this one to exit may conclude it has
2455 finished before the diagnostic output is printed. This may not be
2456 what you expect. 2.X had a different scheme which avoided this
2457 problem, but caused deadlocks under obscure circumstances, so we
2458 are trying something different for 3.0.
2459
2460- Small changes in control log file naming which make it easier to
sewardjec39f332005-08-02 15:28:07 +00002461 use valgrind for debugging MPI-based programs. The relevant
2462 new flags are --log-file-exactly= and --log-file-qualifier=.
sewardjc371bd02005-07-27 09:55:24 +00002463
2464- As part of adding AMD64 support, DWARF2 CFI-based stack unwinding
2465 support was added. In principle this means Valgrind can produce
2466 meaningful backtraces on x86 code compiled with -fomit-frame-pointer
tom81868f02005-07-27 10:33:08 +00002467 providing you also compile your code with -fasynchronous-unwind-tables.
sewardjc371bd02005-07-27 09:55:24 +00002468
sewardjec39f332005-08-02 15:28:07 +00002469- The documentation build system has been completely redone.
2470 The documentation masters are now in XML format, and from that
2471 HTML, PostScript and PDF documentation is generated. As a result
2472 the manual is now available in book form. Note that the
2473 documentation in the source tarballs is pre-built, so you don't need
2474 any XML processing tools to build Valgrind from a tarball.
njn4a9f6f42005-07-19 23:46:08 +00002475
2476Changes that are not user-visible:
2477
sewardjc371bd02005-07-27 09:55:24 +00002478- The code has been massively overhauled in order to modularise it.
2479 As a result we hope it is easier to navigate and understand.
njn4a9f6f42005-07-19 23:46:08 +00002480
2481- Lots of code has been rewritten.
2482
sewardjf9929e62005-08-02 21:05:58 +00002483BUGS FIXED:
njn4a9f6f42005-07-19 23:46:08 +00002484
sewardj79bd9b62005-08-03 16:40:10 +00002485110046 sz == 4 assertion failed
2486109810 vex amd64->IR: unhandled instruction bytes: 0xA3 0x4C 0x70 0xD7
sewardjf9929e62005-08-02 21:05:58 +00002487109802 Add a plausible_stack_size command-line parameter ?
2488109783 unhandled ioctl TIOCMGET (running hw detection tool discover)
2489109780 unhandled ioctl BLKSSZGET (running fdisk -l /dev/hda)
2490109718 vex x86->IR: unhandled instruction: ffreep
2491109429 AMD64 unhandled syscall: 127 (sigpending)
2492109401 false positive uninit in strchr from ld-linux.so.2
2493109385 "stabs" parse failure
2494109378 amd64: unhandled instruction REP NOP
2495109376 amd64: unhandled instruction LOOP Jb
2496109363 AMD64 unhandled instruction bytes
2497109362 AMD64 unhandled syscall: 24 (sched_yield)
2498109358 fork() won't work with valgrind-3.0 SVN
2499109332 amd64 unhandled instruction: ADC Ev, Gv
2500109314 Bogus memcheck report on amd64
2501108883 Crash; vg_memory.c:905 (vgPlain_init_shadow_range):
2502 Assertion `vgPlain_defined_init_shadow_page()' failed.
2503108349 mincore syscall parameter checked incorrectly
2504108059 build infrastructure: small update
2505107524 epoll_ctl event parameter checked on EPOLL_CTL_DEL
2506107123 Vex dies with unhandled instructions: 0xD9 0x31 0xF 0xAE
2507106841 auxmap & openGL problems
2508106713 SDL_Init causes valgrind to exit
2509106352 setcontext and makecontext not handled correctly
2510106293 addresses beyond initial client stack allocation
2511 not checked in VALGRIND_DO_LEAK_CHECK
2512106283 PIE client programs are loaded at address 0
2513105831 Assertion `vgPlain_defined_init_shadow_page()' failed.
2514105039 long run-times probably due to memory manager
2515104797 valgrind needs to be aware of BLKGETSIZE64
2516103594 unhandled instruction: FICOM
2517103320 Valgrind 2.4.0 fails to compile with gcc 3.4.3 and -O0
2518103168 potentially memory leak in coregrind/ume.c
2519102039 bad permissions for mapped region at address 0xB7C73680
2520101881 weird assertion problem
2521101543 Support fadvise64 syscalls
252275247 x86_64/amd64 support (the biggest "bug" we have ever fixed)
njn4a9f6f42005-07-19 23:46:08 +00002523
sewardj9a919512005-08-03 17:01:51 +00002524(3.0RC1: 27 July 05, vex r1303, valgrind r4283).
sewardj1346e132005-08-03 20:42:06 +00002525(3.0.0: 3 August 05, vex r1313, valgrind r4316).
sewardjbfa2b992005-07-27 17:49:17 +00002526
2527
2528
njnccbe3b12009-08-20 00:22:10 +00002529Stable release 2.4.1 (1 August 2005)
2530~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2531(The notes for this release have been lost. Sorry! It would have
2532contained various bug fixes but no new features.)
2533
2534
2535
njn8c927302005-03-12 16:45:01 +00002536Stable release 2.4.0 (March 2005) -- CHANGES RELATIVE TO 2.2.0
njn1aa5ae42005-03-26 16:04:27 +00002537~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
25382.4.0 brings many significant changes and bug fixes. The most
2539significant user-visible change is that we no longer supply our own
2540pthread implementation. Instead, Valgrind is finally capable of
2541running the native thread library, either LinuxThreads or NPTL.
njn8c927302005-03-12 16:45:01 +00002542
njn1aa5ae42005-03-26 16:04:27 +00002543This means our libpthread has gone, along with the bugs associated
2544with it. Valgrind now supports the kernel's threading syscalls, and
2545lets you use your standard system libpthread. As a result:
njn8c927302005-03-12 16:45:01 +00002546
njn1aa5ae42005-03-26 16:04:27 +00002547* There are many fewer system dependencies and strange library-related
2548 bugs. There is a small performance improvement, and a large
2549 stability improvement.
njn8c927302005-03-12 16:45:01 +00002550
njn1aa5ae42005-03-26 16:04:27 +00002551* On the downside, Valgrind can no longer report misuses of the POSIX
2552 PThreads API. It also means that Helgrind currently does not work.
2553 We hope to fix these problems in a future release.
njn8c927302005-03-12 16:45:01 +00002554
njn1aa5ae42005-03-26 16:04:27 +00002555Note that running the native thread libraries does not mean Valgrind
2556is able to provide genuine concurrent execution on SMPs. We still
2557impose the restriction that only one thread is running at any given
2558time.
njn8c927302005-03-12 16:45:01 +00002559
njn1aa5ae42005-03-26 16:04:27 +00002560There are many other significant changes too:
njn8c927302005-03-12 16:45:01 +00002561
njn1aa5ae42005-03-26 16:04:27 +00002562* Memcheck is (once again) the default tool.
njn8c927302005-03-12 16:45:01 +00002563
njn1aa5ae42005-03-26 16:04:27 +00002564* The default stack backtrace is now 12 call frames, rather than 4.
njn8c927302005-03-12 16:45:01 +00002565
njn1aa5ae42005-03-26 16:04:27 +00002566* Suppressions can have up to 25 call frame matches, rather than 4.
njn8c927302005-03-12 16:45:01 +00002567
njn1aa5ae42005-03-26 16:04:27 +00002568* Memcheck and Addrcheck use less memory. Under some circumstances,
2569 they no longer allocate shadow memory if there are large regions of
2570 memory with the same A/V states - such as an mmaped file.
njn8c927302005-03-12 16:45:01 +00002571
njn1aa5ae42005-03-26 16:04:27 +00002572* The memory-leak detector in Memcheck and Addrcheck has been
2573 improved. It now reports more types of memory leak, including
2574 leaked cycles. When reporting leaked memory, it can distinguish
2575 between directly leaked memory (memory with no references), and
2576 indirectly leaked memory (memory only referred to by other leaked
2577 memory).
njn8c927302005-03-12 16:45:01 +00002578
njn1aa5ae42005-03-26 16:04:27 +00002579* Memcheck's confusion over the effect of mprotect() has been fixed:
2580 previously mprotect could erroneously mark undefined data as
2581 defined.
njn8c927302005-03-12 16:45:01 +00002582
njn1aa5ae42005-03-26 16:04:27 +00002583* Signal handling is much improved and should be very close to what
2584 you get when running natively.
2585
2586 One result of this is that Valgrind observes changes to sigcontexts
2587 passed to signal handlers. Such modifications will take effect when
2588 the signal returns. You will need to run with --single-step=yes to
2589 make this useful.
njn8c927302005-03-12 16:45:01 +00002590
2591* Valgrind is built in Position Independent Executable (PIE) format if
njn1aa5ae42005-03-26 16:04:27 +00002592 your toolchain supports it. This allows it to take advantage of all
njn8c927302005-03-12 16:45:01 +00002593 the available address space on systems with 4Gbyte user address
2594 spaces.
2595
2596* Valgrind can now run itself (requires PIE support).
2597
njn1aa5ae42005-03-26 16:04:27 +00002598* Syscall arguments are now checked for validity. Previously all
2599 memory used by syscalls was checked, but now the actual values
2600 passed are also checked.
njn8c927302005-03-12 16:45:01 +00002601
njn1aa5ae42005-03-26 16:04:27 +00002602* Syscall wrappers are more robust against bad addresses being passed
2603 to syscalls: they will fail with EFAULT rather than killing Valgrind
2604 with SIGSEGV.
njn8c927302005-03-12 16:45:01 +00002605
njn1aa5ae42005-03-26 16:04:27 +00002606* Because clone() is directly supported, some non-pthread uses of it
2607 will work. Partial sharing (where some resources are shared, and
2608 some are not) is not supported.
2609
2610* open() and readlink() on /proc/self/exe are supported.
2611
2612BUGS FIXED:
2613
261488520 pipe+fork+dup2 kills the main program
261588604 Valgrind Aborts when using $VALGRIND_OPTS and user progra...
261688614 valgrind: vg_libpthread.c:2323 (read): Assertion `read_pt...
261788703 Stabs parser fails to handle ";"
261888886 ioctl wrappers for TIOCMBIS and TIOCMBIC
261989032 valgrind pthread_cond_timedwait fails
262089106 the 'impossible' happened
262189139 Missing sched_setaffinity & sched_getaffinity
262289198 valgrind lacks support for SIOCSPGRP and SIOCGPGRP
262389263 Missing ioctl translations for scsi-generic and CD playing
262489440 tests/deadlock.c line endings
262589481 `impossible' happened: EXEC FAILED
262689663 valgrind 2.2.0 crash on Redhat 7.2
262789792 Report pthread_mutex_lock() deadlocks instead of returnin...
262890111 statvfs64 gives invalid error/warning
262990128 crash+memory fault with stabs generated by gnat for a run...
263090778 VALGRIND_CHECK_DEFINED() not as documented in memcheck.h
263190834 cachegrind crashes at end of program without reporting re...
263291028 valgrind: vg_memory.c:229 (vgPlain_unmap_range): Assertio...
263391162 valgrind crash while debugging drivel 1.2.1
263491199 Unimplemented function
263591325 Signal routing does not propagate the siginfo structure
263691599 Assertion `cv == ((void *)0)'
263791604 rw_lookup clears orig and sends the NULL value to rw_new
263891821 Small problems building valgrind with $top_builddir ne $t...
263991844 signal 11 (SIGSEGV) at get_tcb (libpthread.c:86) in corec...
264092264 UNIMPLEMENTED FUNCTION: pthread_condattr_setpshared
264192331 per-target flags necessitate AM_PROG_CC_C_O
264292420 valgrind doesn't compile with linux 2.6.8.1/9
264392513 Valgrind 2.2.0 generates some warning messages
264492528 vg_symtab2.c:170 (addLoc): Assertion `loc->size > 0' failed.
264593096 unhandled ioctl 0x4B3A and 0x5601
264693117 Tool and core interface versions do not match
264793128 Can't run valgrind --tool=memcheck because of unimplement...
264893174 Valgrind can crash if passed bad args to certain syscalls
264993309 Stack frame in new thread is badly aligned
265093328 Wrong types used with sys_sigprocmask()
265193763 /usr/include/asm/msr.h is missing
265293776 valgrind: vg_memory.c:508 (vgPlain_find_map_space): Asser...
265393810 fcntl() argument checking a bit too strict
265494378 Assertion `tst->sigqueue_head != tst->sigqueue_tail' failed.
265594429 valgrind 2.2.0 segfault with mmap64 in glibc 2.3.3
265694645 Impossible happened: PINSRW mem
265794953 valgrind: the `impossible' happened: SIGSEGV
265895667 Valgrind does not work with any KDE app
265996243 Assertion 'res==0' failed
266096252 stage2 loader of valgrind fails to allocate memory
266196520 All programs crashing at _dl_start (in /lib/ld-2.3.3.so) ...
266296660 ioctl CDROMREADTOCENTRY causes bogus warnings
266396747 After looping in a segfault handler, the impossible happens
266496923 Zero sized arrays crash valgrind trace back with SIGFPE
266596948 valgrind stops with assertion failure regarding mmap2
266696966 valgrind fails when application opens more than 16 sockets
266797398 valgrind: vg_libpthread.c:2667 Assertion failed
266897407 valgrind: vg_mylibc.c:1226 (vgPlain_safe_fd): Assertion `...
266997427 "Warning: invalid file descriptor -1 in syscall close()" ...
267097785 missing backtrace
267197792 build in obj dir fails - autoconf / makefile cleanup
267297880 pthread_mutex_lock fails from shared library (special ker...
267397975 program aborts without ang VG messages
267498129 Failed when open and close file 230000 times using stdio
267598175 Crashes when using valgrind-2.2.0 with a program using al...
267698288 Massif broken
267798303 UNIMPLEMENTED FUNCTION pthread_condattr_setpshared
267898630 failed--compilation missing warnings.pm, fails to make he...
267998756 Cannot valgrind signal-heavy kdrive X server
268098966 valgrinding the JVM fails with a sanity check assertion
268199035 Valgrind crashes while profiling
268299142 loops with message "Signal 11 being dropped from thread 0...
268399195 threaded apps crash on thread start (using QThread::start...
268499348 Assertion `vgPlain_lseek(core_fd, 0, 1) == phdrs[i].p_off...
268599568 False negative due to mishandling of mprotect
268699738 valgrind memcheck crashes on program that uses sigitimer
268799923 0-sized allocations are reported as leaks
268899949 program seg faults after exit()
2689100036 "newSuperblock's request for 1048576 bytes failed"
2690100116 valgrind: (pthread_cond_init): Assertion `sizeof(* cond) ...
2691100486 memcheck reports "valgrind: the `impossible' happened: V...
2692100833 second call to "mremap" fails with EINVAL
2693101156 (vgPlain_find_map_space): Assertion `(addr & ((1 << 12)-1...
2694101173 Assertion `recDepth >= 0 && recDepth < 500' failed
2695101291 creating threads in a forked process fails
2696101313 valgrind causes different behavior when resizing a window...
2697101423 segfault for c++ array of floats
2698101562 valgrind massif dies on SIGINT even with signal handler r...
2699
njn8c927302005-03-12 16:45:01 +00002700
jseward34ed74f2004-08-30 18:04:42 +00002701Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.0.0
2702~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
jseward7d180d92004-08-30 18:15:37 +000027032.2.0 brings nine months worth of improvements and bug fixes. We
2704believe it to be a worthy successor to 2.0.0. There are literally
2705hundreds of bug fixes and minor improvements. There are also some
2706fairly major user-visible changes:
2707
2708* A complete overhaul of handling of system calls and signals, and
2709 their interaction with threads. In general, the accuracy of the
2710 system call, thread and signal simulations is much improved:
2711
2712 - Blocking system calls behave exactly as they do when running
2713 natively (not on valgrind). That is, if a syscall blocks only the
2714 calling thread when running natively, than it behaves the same on
2715 valgrind. No more mysterious hangs because V doesn't know that some
2716 syscall or other, should block only the calling thread.
2717
2718 - Interrupted syscalls should now give more faithful results.
2719
2720 - Signal contexts in signal handlers are supported.
2721
2722* Improvements to NPTL support to the extent that V now works
2723 properly on NPTL-only setups.
2724
2725* Greater isolation between Valgrind and the program being run, so
2726 the program is less likely to inadvertently kill Valgrind by
2727 doing wild writes.
2728
2729* Massif: a new space profiling tool. Try it! It's cool, and it'll
2730 tell you in detail where and when your C/C++ code is allocating heap.
2731 Draws pretty .ps pictures of memory use against time. A potentially
2732 powerful tool for making sense of your program's space use.
2733
2734* File descriptor leakage checks. When enabled, Valgrind will print out
2735 a list of open file descriptors on exit.
2736
2737* Improved SSE2/SSE3 support.
2738
jseward2eb002b2004-08-31 00:14:02 +00002739* Time-stamped output; use --time-stamp=yes
2740
jseward34ed74f2004-08-30 18:04:42 +00002741
2742
2743Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.1.2
2744~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
27452.2.0 is not much different from 2.1.2, released seven weeks ago.
2746A number of bugs have been fixed, most notably #85658, which gave
2747problems for quite a few people. There have been many internal
2748cleanups, but those are not user visible.
2749
2750The following bugs have been fixed since 2.1.2:
2751
275285658 Assert in coregrind/vg_libpthread.c:2326 (open64) !=
2753 (void*)0 failed
2754 This bug was reported multiple times, and so the following
2755 duplicates of it are also fixed: 87620, 85796, 85935, 86065,
2756 86919, 86988, 87917, 88156
2757
275880716 Semaphore mapping bug caused by unmap (sem_destroy)
2759 (Was fixed prior to 2.1.2)
2760
276186987 semctl and shmctl syscalls family is not handled properly
2762
276386696 valgrind 2.1.2 + RH AS2.1 + librt
2764
276586730 valgrind locks up at end of run with assertion failure
2766 in __pthread_unwind
2767
276886641 memcheck doesn't work with Mesa OpenGL/ATI on Suse 9.1
2769 (also fixes 74298, a duplicate of this)
2770
277185947 MMX/SSE unhandled instruction 'sfence'
2772
277384978 Wrong error "Conditional jump or move depends on
2774 uninitialised value" resulting from "sbbl %reg, %reg"
2775
277686254 ssort() fails when signed int return type from comparison is
2777 too small to handle result of unsigned int subtraction
2778
277987089 memalign( 4, xxx) makes valgrind assert
2780
jseward2eb002b2004-08-31 00:14:02 +0000278186407 Add support for low-level parallel port driver ioctls.
jseward34ed74f2004-08-30 18:04:42 +00002782
278370587 Add timestamps to Valgrind output? (wishlist)
2784
278584937 vg_libpthread.c:2505 (se_remap): Assertion `res == 0'
2786 (fixed prior to 2.1.2)
2787
278886317 cannot load libSDL-1.2.so.0 using valgrind
2789
279086989 memcpy from mac_replace_strmem.c complains about
2791 uninitialized pointers passed when length to copy is zero
2792
279385811 gnu pascal symbol causes segmentation fault; ok in 2.0.0
2794
279579138 writing to sbrk()'d memory causes segfault
2796
279777369 sched deadlock while signal received during pthread_join
2798 and the joined thread exited
2799
280088115 In signal handler for SIGFPE, siginfo->si_addr is wrong
2801 under Valgrind
2802
280378765 Massif crashes on app exit if FP exceptions are enabled
2804
2805Additionally there are the following changes, which are not
2806connected to any bug report numbers, AFAICS:
2807
2808* Fix scary bug causing mis-identification of SSE stores vs
2809 loads and so causing memcheck to sometimes give nonsense results
2810 on SSE code.
2811
2812* Add support for the POSIX message queue system calls.
2813
2814* Fix to allow 32-bit Valgrind to run on AMD64 boxes. Note: this does
2815 NOT allow Valgrind to work with 64-bit executables - only with 32-bit
2816 executables on an AMD64 box.
2817
2818* At configure time, only check whether linux/mii.h can be processed
2819 so that we don't generate ugly warnings by trying to compile it.
2820
jseward2eb002b2004-08-31 00:14:02 +00002821* Add support for POSIX clocks and timers.
2822
jseward34ed74f2004-08-30 18:04:42 +00002823
2824
jseward9cd978c2004-07-17 13:38:12 +00002825Developer (cvs head) release 2.1.2 (18 July 2004)
jsewardbd0a6e42004-07-17 12:19:43 +00002826~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
28272.1.2 contains four months worth of bug fixes and refinements.
jseward9cd978c2004-07-17 13:38:12 +00002828Although officially a developer release, we believe it to be stable
2829enough for widespread day-to-day use. 2.1.2 is pretty good, so try it
2830first, although there is a chance it won't work. If so then try 2.0.0
2831and tell us what went wrong." 2.1.2 fixes a lot of problems present
2832in 2.0.0 and is generally a much better product.
jsewardbd0a6e42004-07-17 12:19:43 +00002833
jseward9cd978c2004-07-17 13:38:12 +00002834Relative to 2.1.1, a large number of minor problems with 2.1.1 have
2835been fixed, and so if you use 2.1.1 you should try 2.1.2. Users of
2836the last stable release, 2.0.0, might also want to try this release.
jsewardbd0a6e42004-07-17 12:19:43 +00002837
2838The following bugs, and probably many more, have been fixed. These
2839are listed at http://bugs.kde.org. Reporting a bug for valgrind in
2840the http://bugs.kde.org is much more likely to get you a fix than
2841mailing developers directly, so please continue to keep sending bugs
2842there.
2843
284476869 Crashes when running any tool under Fedora Core 2 test1
2845 This fixes the problem with returning from a signal handler
jseward0400d4d2004-07-17 23:26:46 +00002846 when VDSOs are turned off in FC2.
jsewardbd0a6e42004-07-17 12:19:43 +00002847
jseward0400d4d2004-07-17 23:26:46 +0000284869508 java 1.4.2 client fails with erroneous "stack size too small".
2849 This fix makes more of the pthread stack attribute related
2850 functions work properly. Java still doesn't work though.
jsewardbd0a6e42004-07-17 12:19:43 +00002851
285271906 malloc alignment should be 8, not 4
2853 All memory returned by malloc/new etc is now at least
2854 8-byte aligned.
2855
jsewarda6017a42004-07-17 23:44:35 +0000285681970 vg_alloc_ThreadState: no free slots available
2857 (closed because the workaround is simple: increase
2858 VG_N_THREADS, rebuild and try again.)
2859
jsewardbd0a6e42004-07-17 12:19:43 +0000286078514 Conditional jump or move depends on uninitialized value(s)
2861 (a slight mishanding of FP code in memcheck)
2862
jsewarda6017a42004-07-17 23:44:35 +0000286377952 pThread Support (crash) (due to initialisation-ordering probs)
2864 (also 85118)
2865
jsewardbd0a6e42004-07-17 12:19:43 +0000286680942 Addrcheck wasn't doing overlap checking as it should.
286778048 return NULL on malloc/new etc failure, instead of asserting
286873655 operator new() override in user .so files often doesn't get picked up
286983060 Valgrind does not handle native kernel AIO
287069872 Create proper coredumps after fatal signals
287182026 failure with new glibc versions: __libc_* functions are not exported
287270344 UNIMPLEMENTED FUNCTION: tcdrain
287381297 Cancellation of pthread_cond_wait does not require mutex
287482872 Using debug info from additional packages (wishlist)
287583025 Support for ioctls FIGETBSZ and FIBMAP
287683340 Support for ioctl HDIO_GET_IDENTITY
287779714 Support for the semtimedop system call.
287877022 Support for ioctls FBIOGET_VSCREENINFO and FBIOGET_FSCREENINFO
287982098 hp2ps ansification (wishlist)
288083573 Valgrind SIGSEGV on execve
288182999 show which cmdline option was erroneous (wishlist)
288283040 make valgrind VPATH and distcheck-clean (wishlist)
jsewarda6017a42004-07-17 23:44:35 +0000288383998 Assertion `newfd > vgPlain_max_fd' failed (see below)
288482722 Unchecked mmap in as_pad leads to mysterious failures later
288578958 memcheck seg faults while running Mozilla
jsewarda2c08002004-07-18 09:24:57 +0000288685416 Arguments with colon (e.g. --logsocket) ignored
jsewardbd0a6e42004-07-17 12:19:43 +00002887
2888
2889Additionally there are the following changes, which are not
2890connected to any bug report numbers, AFAICS:
2891
2892* Rearranged address space layout relative to 2.1.1, so that
2893 Valgrind/tools will run out of memory later than currently in many
2894 circumstances. This is good news esp. for Calltree. It should
2895 be possible for client programs to allocate over 800MB of
2896 memory when using memcheck now.
2897
jseward9cd978c2004-07-17 13:38:12 +00002898* Improved checking when laying out memory. Should hopefully avoid
2899 the random segmentation faults that 2.1.1 sometimes caused.
2900
jsewardbd0a6e42004-07-17 12:19:43 +00002901* Support for Fedora Core 2 and SuSE 9.1. Improvements to NPTL
2902 support to the extent that V now works properly on NPTL-only setups.
2903
2904* Renamed the following options:
2905 --logfile-fd --> --log-fd
2906 --logfile --> --log-file
2907 --logsocket --> --log-socket
2908 to be consistent with each other and other options (esp. --input-fd).
2909
2910* Add support for SIOCGMIIPHY, SIOCGMIIREG and SIOCSMIIREG ioctls and
2911 improve the checking of other interface related ioctls.
2912
jsewardbd0a6e42004-07-17 12:19:43 +00002913* Fix building with gcc-3.4.1.
2914
2915* Remove limit on number of semaphores supported.
2916
jsewardbd0a6e42004-07-17 12:19:43 +00002917* Add support for syscalls: set_tid_address (258), acct (51).
2918
2919* Support instruction "repne movs" -- not official but seems to occur.
2920
2921* Implement an emulated soft limit for file descriptors in addition to
2922 the current reserved area, which effectively acts as a hard limit. The
2923 setrlimit system call now simply updates the emulated limits as best
2924 as possible - the hard limit is not allowed to move at all and just
2925 returns EPERM if you try and change it. This should stop reductions
2926 in the soft limit causing assertions when valgrind tries to allocate
2927 descriptors from the reserved area.
jsewarda6017a42004-07-17 23:44:35 +00002928 (This actually came from bug #83998).
jsewardbd0a6e42004-07-17 12:19:43 +00002929
nethercote76e8fd92004-07-21 10:37:31 +00002930* Major overhaul of Cachegrind implementation. First user-visible change
nethercote5e63bfb2004-07-17 12:49:53 +00002931 is that cachegrind.out files are now typically 90% smaller than they
nethercote76e8fd92004-07-21 10:37:31 +00002932 used to be; code annotation times are correspondingly much smaller.
2933 Second user-visible change is that hit/miss counts for code that is
2934 unloaded at run-time is no longer dumped into a single "discard" pile,
2935 but accurately preserved.
jsewardbd0a6e42004-07-17 12:19:43 +00002936
2937* Client requests for telling valgrind about memory pools.
2938
2939
2940
jseward9cd978c2004-07-17 13:38:12 +00002941Developer (cvs head) release 2.1.1 (12 March 2004)
2942~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
jsewardbd0a6e42004-07-17 12:19:43 +000029432.1.1 contains some internal structural changes needed for V's
jsewarde3e77042004-03-12 21:07:05 +00002944long-term future. These don't affect end-users. Most notable
2945user-visible changes are:
2946
2947* Greater isolation between Valgrind and the program being run, so
2948 the program is less likely to inadvertently kill Valgrind by
2949 doing wild writes.
2950
2951* Massif: a new space profiling tool. Try it! It's cool, and it'll
2952 tell you in detail where and when your C/C++ code is allocating heap.
2953 Draws pretty .ps pictures of memory use against time. A potentially
2954 powerful tool for making sense of your program's space use.
2955
2956* Fixes for many bugs, including support for more SSE2/SSE3 instructions,
2957 various signal/syscall things, and various problems with debug
2958 info readers.
2959
2960* Support for glibc-2.3.3 based systems.
2961
2962We are now doing automatic overnight build-and-test runs on a variety
2963of distros. As a result, we believe 2.1.1 builds and runs on:
2964Red Hat 7.2, 7.3, 8.0, 9, Fedora Core 1, SuSE 8.2, SuSE 9.
2965
2966
2967The following bugs, and probably many more, have been fixed. These
2968are listed at http://bugs.kde.org. Reporting a bug for valgrind in
2969the http://bugs.kde.org is much more likely to get you a fix than
2970mailing developers directly, so please continue to keep sending bugs
2971there.
2972
297369616 glibc 2.3.2 w/NPTL is massively different than what valgrind expects
297469856 I don't know how to instrument MMXish stuff (Helgrind)
297573892 valgrind segfaults starting with Objective-C debug info
2976 (fix for S-type stabs)
297773145 Valgrind complains too much about close(<reserved fd>)
297873902 Shadow memory allocation seems to fail on RedHat 8.0
297968633 VG_N_SEMAPHORES too low (V itself was leaking semaphores)
298075099 impossible to trace multiprocess programs
298176839 the `impossible' happened: disInstr: INT but not 0x80 !
298276762 vg_to_ucode.c:3748 (dis_push_segreg): Assertion `sz == 4' failed.
298376747 cannot include valgrind.h in c++ program
298476223 parsing B(3,10) gave NULL type => impossible happens
298575604 shmdt handling problem
298676416 Problems with gcc 3.4 snap 20040225
298775614 using -gstabs when building your programs the `impossible' happened
298875787 Patch for some CDROM ioctls CDORM_GET_MCN, CDROM_SEND_PACKET,
298975294 gcc 3.4 snapshot's libstdc++ have unsupported instructions.
2990 (REP RET)
299173326 vg_symtab2.c:272 (addScopeRange): Assertion `range->size > 0' failed.
299272596 not recognizing __libc_malloc
299369489 Would like to attach ddd to running program
299472781 Cachegrind crashes with kde programs
299573055 Illegal operand at DXTCV11CompressBlockSSE2 (more SSE opcodes)
299673026 Descriptor leak check reports port numbers wrongly
299771705 README_MISSING_SYSCALL_OR_IOCTL out of date
299872643 Improve support for SSE/SSE2 instructions
299972484 valgrind leaves it's own signal mask in place when execing
300072650 Signal Handling always seems to restart system calls
300172006 The mmap system call turns all errors in ENOMEM
300271781 gdb attach is pretty useless
300371180 unhandled instruction bytes: 0xF 0xAE 0x85 0xE8
300469886 writes to zero page cause valgrind to assert on exit
300571791 crash when valgrinding gimp 1.3 (stabs reader problem)
300669783 unhandled syscall: 218
300769782 unhandled instruction bytes: 0x66 0xF 0x2B 0x80
300870385 valgrind fails if the soft file descriptor limit is less
3009 than about 828
301069529 "rep; nop" should do a yield
301170827 programs with lots of shared libraries report "mmap failed"
3012 for some of them when reading symbols
301371028 glibc's strnlen is optimised enough to confuse valgrind
3014
3015
3016
3017
jseward43230652003-12-14 00:14:54 +00003018Unstable (cvs head) release 2.1.0 (15 December 2003)
3019~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3020For whatever it's worth, 2.1.0 actually seems pretty darn stable to me
3021(Julian). It looks eminently usable, and given that it fixes some
3022significant bugs, may well be worth using on a day-to-day basis.
30232.1.0 is known to build and pass regression tests on: SuSE 9, SuSE
30248.2, RedHat 8.
3025
30262.1.0 most notably includes Jeremy Fitzhardinge's complete overhaul of
3027handling of system calls and signals, and their interaction with
3028threads. In general, the accuracy of the system call, thread and
3029signal simulations is much improved. Specifically:
3030
3031- Blocking system calls behave exactly as they do when running
3032 natively (not on valgrind). That is, if a syscall blocks only the
3033 calling thread when running natively, than it behaves the same on
3034 valgrind. No more mysterious hangs because V doesn't know that some
3035 syscall or other, should block only the calling thread.
3036
3037- Interrupted syscalls should now give more faithful results.
3038
3039- Finally, signal contexts in signal handlers are supported. As a
3040 result, konqueror on SuSE 9 no longer segfaults when notified of
3041 file changes in directories it is watching.
3042
3043Other changes:
3044
3045- Robert Walsh's file descriptor leakage checks. When enabled,
3046 Valgrind will print out a list of open file descriptors on
3047 exit. Along with each file descriptor, Valgrind prints out a stack
3048 backtrace of where the file was opened and any details relating to the
3049 file descriptor such as the file name or socket details.
3050 To use, give: --track-fds=yes
3051
3052- Implemented a few more SSE/SSE2 instructions.
3053
3054- Less crud on the stack when you do 'where' inside a GDB attach.
3055
3056- Fixed the following bugs:
3057 68360: Valgrind does not compile against 2.6.0-testX kernels
3058 68525: CVS head doesn't compile on C90 compilers
3059 68566: pkgconfig support (wishlist)
3060 68588: Assertion `sz == 4' failed in vg_to_ucode.c (disInstr)
3061 69140: valgrind not able to explicitly specify a path to a binary.
3062 69432: helgrind asserts encountering a MutexErr when there are
3063 EraserErr suppressions
3064
jseward0b12daf2003-12-14 14:35:28 +00003065- Increase the max size of the translation cache from 200k average bbs
3066 to 300k average bbs. Programs on the size of OOo (680m17) are
3067 thrashing the cache at the smaller size, creating large numbers of
3068 retranslations and wasting significant time as a result.
3069
jseward43230652003-12-14 00:14:54 +00003070
3071
3072Stable release 2.0.0 (5 Nov 2003)
3073~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3074
30752.0.0 improves SSE/SSE2 support, fixes some minor bugs, and
3076improves support for SuSE 9 and the Red Hat "Severn" beta.
3077
3078- Further improvements to SSE/SSE2 support. The entire test suite of
3079 the GNU Scientific Library (gsl-1.4) compiled with Intel Icc 7.1
3080 20030307Z '-g -O -xW' now works. I think this gives pretty good
3081 coverage of SSE/SSE2 floating point instructions, or at least the
3082 subset emitted by Icc.
3083
3084- Also added support for the following instructions:
3085 MOVNTDQ UCOMISD UNPCKLPS UNPCKHPS SQRTSS
3086 PUSH/POP %{FS,GS}, and PUSH %CS (Nb: there is no POP %CS).
3087
3088- CFI support for GDB version 6. Needed to enable newer GDBs
3089 to figure out where they are when using --gdb-attach=yes.
3090
3091- Fix this:
3092 mc_translate.c:1091 (memcheck_instrument): Assertion
3093 `u_in->size == 4 || u_in->size == 16' failed.
3094
3095- Return an error rather than panicing when given a bad socketcall.
3096
3097- Fix checking of syscall rt_sigtimedwait().
3098
3099- Implement __NR_clock_gettime (syscall 265). Needed on Red Hat Severn.
3100
3101- Fixed bug in overlap check in strncpy() -- it was assuming the src was 'n'
3102 bytes long, when it could be shorter, which could cause false
3103 positives.
3104
3105- Support use of select() for very large numbers of file descriptors.
3106
3107- Don't fail silently if the executable is statically linked, or is
3108 setuid/setgid. Print an error message instead.
3109
3110- Support for old DWARF-1 format line number info.
3111
3112
3113
3114Snapshot 20031012 (12 October 2003)
3115~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3116
3117Three months worth of bug fixes, roughly. Most significant single
3118change is improved SSE/SSE2 support, mostly thanks to Dirk Mueller.
3119
312020031012 builds on Red Hat Fedora ("Severn") but doesn't really work
3121(curiosly, mozilla runs OK, but a modest "ls -l" bombs). I hope to
3122get a working version out soon. It may or may not work ok on the
3123forthcoming SuSE 9; I hear positive noises about it but haven't been
3124able to verify this myself (not until I get hold of a copy of 9).
3125
3126A detailed list of changes, in no particular order:
3127
3128- Describe --gen-suppressions in the FAQ.
3129
3130- Syscall __NR_waitpid supported.
3131
3132- Minor MMX bug fix.
3133
3134- -v prints program's argv[] at startup.
3135
3136- More glibc-2.3 suppressions.
3137
3138- Suppressions for stack underrun bug(s) in the c++ support library
3139 distributed with Intel Icc 7.0.
3140
3141- Fix problems reading /proc/self/maps.
3142
3143- Fix a couple of messages that should have been suppressed by -q,
3144 but weren't.
3145
3146- Make Addrcheck understand "Overlap" suppressions.
3147
3148- At startup, check if program is statically linked and bail out if so.
3149
3150- Cachegrind: Auto-detect Intel Pentium-M, also VIA Nehemiah
3151
3152- Memcheck/addrcheck: minor speed optimisations
3153
3154- Handle syscall __NR_brk more correctly than before.
3155
3156- Fixed incorrect allocate/free mismatch errors when using
3157 operator new(unsigned, std::nothrow_t const&)
3158 operator new[](unsigned, std::nothrow_t const&)
3159
3160- Support POSIX pthread spinlocks.
3161
3162- Fixups for clean compilation with gcc-3.3.1.
daywalker5d945de2003-09-26 00:32:53 +00003163
3164- Implemented more opcodes:
3165 - push %es
3166 - push %ds
3167 - pop %es
3168 - pop %ds
3169 - movntq
3170 - sfence
3171 - pshufw
3172 - pavgb
3173 - ucomiss
3174 - enter
daywalkerb18d2532003-09-27 20:15:01 +00003175 - mov imm32, %esp
3176 - all "in" and "out" opcodes
daywalker79aad842003-09-30 22:58:12 +00003177 - inc/dec %esp
jseward43230652003-12-14 00:14:54 +00003178 - A whole bunch of SSE/SSE2 instructions
daywalker5d945de2003-09-26 00:32:53 +00003179
jseward43230652003-12-14 00:14:54 +00003180- Memcheck: don't bomb on SSE/SSE2 code.
daywalker5d945de2003-09-26 00:32:53 +00003181
sewardj90238792003-05-05 00:23:42 +00003182
sewardj945f39f2003-07-25 21:11:40 +00003183Snapshot 20030725 (25 July 2003)
3184~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3185
3186Fixes some minor problems in 20030716.
3187
3188- Fix bugs in overlap checking for strcpy/memcpy etc.
3189
3190- Do overlap checking with Addrcheck as well as Memcheck.
3191
3192- Fix this:
3193 Memcheck: the `impossible' happened:
3194 get_error_name: unexpected type
3195
3196- Install headers needed to compile new skins.
3197
sewardje3dd2e02003-07-27 08:04:48 +00003198- Remove leading spaces and colon in the LD_LIBRARY_PATH / LD_PRELOAD
sewardj945f39f2003-07-25 21:11:40 +00003199 passed to non-traced children.
3200
sewardjb9eda952003-07-26 21:39:05 +00003201- Fix file descriptor leak in valgrind-listener.
3202
sewardje3dd2e02003-07-27 08:04:48 +00003203- Fix longstanding bug in which the allocation point of a
3204 block resized by realloc was not correctly set. This may
3205 have caused confusing error messages.
sewardj945f39f2003-07-25 21:11:40 +00003206
3207
sewardj626fd892003-07-16 20:10:26 +00003208Snapshot 20030716 (16 July 2003)
sewardj9d916ed2003-07-14 23:38:40 +00003209~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3210
sewardj626fd892003-07-16 20:10:26 +0000321120030716 is a snapshot of our current CVS head (development) branch.
sewardj9d916ed2003-07-14 23:38:40 +00003212This is the branch which will become valgrind-2.0. It contains
3213significant enhancements over the 1.9.X branch.
3214
3215Despite this being a snapshot of the CVS head, it is believed to be
3216quite stable -- at least as stable as 1.9.6 or 1.0.4, if not more so
3217-- and therefore suitable for widespread use. Please let us know asap
3218if it causes problems for you.
3219
3220Two reasons for releasing a snapshot now are:
3221
3222- It's been a while since 1.9.6, and this snapshot fixes
3223 various problems that 1.9.6 has with threaded programs
3224 on glibc-2.3.X based systems.
3225
3226- So as to make available improvements in the 2.0 line.
3227
sewardj626fd892003-07-16 20:10:26 +00003228Major changes in 20030716, as compared to 1.9.6:
sewardj9d916ed2003-07-14 23:38:40 +00003229
3230- More fixes to threading support on glibc-2.3.1 and 2.3.2-based
3231 systems (SuSE 8.2, Red Hat 9). If you have had problems
3232 with inconsistent/illogical behaviour of errno, h_errno or the DNS
sewardj626fd892003-07-16 20:10:26 +00003233 resolver functions in threaded programs, 20030716 should improve
sewardj9d916ed2003-07-14 23:38:40 +00003234 matters. This snapshot seems stable enough to run OpenOffice.org
3235 1.1rc on Red Hat 7.3, SuSE 8.2 and Red Hat 9, and that's a big
3236 threaded app if ever I saw one.
3237
3238- Automatic generation of suppression records; you no longer
3239 need to write them by hand. Use --gen-suppressions=yes.
3240
sewardj21511802003-07-22 17:47:42 +00003241- strcpy/memcpy/etc check their arguments for overlaps, when
3242 running with the Memcheck or Addrcheck skins.
3243
3244- malloc_usable_size() is now supported.
3245
3246- new client requests:
3247 - VALGRIND_COUNT_ERRORS, VALGRIND_COUNT_LEAKS:
3248 useful with regression testing
3249 - VALGRIND_NON_SIMD_CALL[0123]: for running arbitrary functions
3250 on real CPU (use with caution!)
3251
sewardj9d916ed2003-07-14 23:38:40 +00003252- The GDB attach mechanism is more flexible. Allow the GDB to
3253 be run to be specified by --gdb-path=/path/to/gdb, and specify
3254 which file descriptor V will read its input from with
3255 --input-fd=<number>.
3256
sewardj21511802003-07-22 17:47:42 +00003257- Cachegrind gives more accurate results (wasn't tracking instructions in
3258 malloc() and friends previously, is now).
3259
sewardj9d916ed2003-07-14 23:38:40 +00003260- Complete support for the MMX instruction set.
3261
3262- Partial support for the SSE and SSE2 instruction sets. Work for this
3263 is ongoing. About half the SSE/SSE2 instructions are done, so
3264 some SSE based programs may work. Currently you need to specify
3265 --skin=addrcheck. Basically not suitable for real use yet.
3266
3267- Significant speedups (10%-20%) for standard memory checking.
3268
3269- Fix assertion failure in pthread_once().
3270
3271- Fix this:
3272 valgrind: vg_intercept.c:598 (vgAllRoadsLeadToRome_select):
3273 Assertion `ms_end >= ms_now' failed.
3274
3275- Implement pthread_mutexattr_setpshared.
3276
3277- Understand Pentium 4 branch hints. Also implemented a couple more
3278 obscure x86 instructions.
3279
3280- Lots of other minor bug fixes.
3281
sewardj626fd892003-07-16 20:10:26 +00003282- We have a decent regression test system, for the first time.
3283 This doesn't help you directly, but it does make it a lot easier
3284 for us to track the quality of the system, especially across
3285 multiple linux distributions.
3286
3287 You can run the regression tests with 'make regtest' after 'make
3288 install' completes. On SuSE 8.2 and Red Hat 9 I get this:
3289
3290 == 84 tests, 0 stderr failures, 0 stdout failures ==
3291
3292 On Red Hat 8, I get this:
3293
3294 == 84 tests, 2 stderr failures, 1 stdout failure ==
3295 corecheck/tests/res_search (stdout)
3296 memcheck/tests/sigaltstack (stderr)
3297
3298 sigaltstack is probably harmless. res_search doesn't work
3299 on R H 8 even running natively, so I'm not too worried.
3300
3301 On Red Hat 7.3, a glibc-2.2.5 system, I get these harmless failures:
3302
3303 == 84 tests, 2 stderr failures, 1 stdout failure ==
3304 corecheck/tests/pth_atfork1 (stdout)
3305 corecheck/tests/pth_atfork1 (stderr)
3306 memcheck/tests/sigaltstack (stderr)
3307
3308 You need to run on a PII system, at least, since some tests
3309 contain P6-specific instructions, and the test machine needs
3310 access to the internet so that corecheck/tests/res_search
3311 (a test that the DNS resolver works) can function.
3312
sewardj9d916ed2003-07-14 23:38:40 +00003313As ever, thanks for the vast amount of feedback :) and bug reports :(
3314We may not answer all messages, but we do at least look at all of
3315them, and tend to fix the most frequently reported bugs.
3316
3317
3318
sewardj37918822003-05-05 01:05:09 +00003319Version 1.9.6 (7 May 2003 or thereabouts)
3320~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3321
3322Major changes in 1.9.6:
3323
3324- Improved threading support for glibc >= 2.3.2 (SuSE 8.2,
3325 RedHat 9, to name but two ...) It turned out that 1.9.5
3326 had problems with threading support on glibc >= 2.3.2,
3327 usually manifested by threaded programs deadlocking in system calls,
3328 or running unbelievably slowly. Hopefully these are fixed now. 1.9.6
3329 is the first valgrind which gives reasonable support for
3330 glibc-2.3.2. Also fixed a 2.3.2 problem with pthread_atfork().
3331
3332- Majorly expanded FAQ.txt. We've added workarounds for all
3333 common problems for which a workaround is known.
3334
3335Minor changes in 1.9.6:
3336
3337- Fix identification of the main thread's stack. Incorrect
3338 identification of it was causing some on-stack addresses to not get
3339 identified as such. This only affected the usefulness of some error
3340 messages; the correctness of the checks made is unchanged.
3341
3342- Support for kernels >= 2.5.68.
3343
3344- Dummy implementations of __libc_current_sigrtmin,
3345 __libc_current_sigrtmax and __libc_allocate_rtsig, hopefully
3346 good enough to keep alive programs which previously died for lack of
3347 them.
3348
3349- Fix bug in the VALGRIND_DISCARD_TRANSLATIONS client request.
3350
sewardj3d47b792003-05-05 22:15:35 +00003351- Fix bug in the DWARF2 debug line info loader, when instructions
3352 following each other have source lines far from each other
3353 (e.g. with inlined functions).
3354
sewardj37918822003-05-05 01:05:09 +00003355- Debug info reading: read symbols from both "symtab" and "dynsym"
3356 sections, rather than merely from the one that comes last in the
3357 file.
3358
3359- New syscall support: prctl(), creat(), lookup_dcookie().
3360
3361- When checking calls to accept(), recvfrom(), getsocketopt(),
3362 don't complain if buffer values are NULL.
3363
3364- Try and avoid assertion failures in
3365 mash_LD_PRELOAD_and_LD_LIBRARY_PATH.
3366
3367- Minor bug fixes in cg_annotate.
3368
3369
3370
sewardj90238792003-05-05 00:23:42 +00003371Version 1.9.5 (7 April 2003)
3372~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3373
3374It occurs to me that it would be helpful for valgrind users to record
3375in the source distribution the changes in each release. So I now
3376attempt to mend my errant ways :-) Changes in this and future releases
3377will be documented in the NEWS file in the source distribution.
3378
3379Major changes in 1.9.5:
3380
3381- (Critical bug fix): Fix a bug in the FPU simulation. This was
3382 causing some floating point conditional tests not to work right.
3383 Several people reported this. If you had floating point code which
3384 didn't work right on 1.9.1 to 1.9.4, it's worth trying 1.9.5.
3385
3386- Partial support for Red Hat 9. RH9 uses the new Native Posix
3387 Threads Library (NPTL), instead of the older LinuxThreads.
3388 This potentially causes problems with V which will take some
3389 time to correct. In the meantime we have partially worked around
3390 this, and so 1.9.5 works on RH9. Threaded programs still work,
3391 but they may deadlock, because some system calls (accept, read,
3392 write, etc) which should be nonblocking, in fact do block. This
3393 is a known bug which we are looking into.
3394
3395 If you can, your best bet (unfortunately) is to avoid using
3396 1.9.5 on a Red Hat 9 system, or on any NPTL-based distribution.
3397 If your glibc is 2.3.1 or earlier, you're almost certainly OK.
3398
3399Minor changes in 1.9.5:
3400
3401- Added some #errors to valgrind.h to ensure people don't include
3402 it accidentally in their sources. This is a change from 1.0.X
3403 which was never properly documented. The right thing to include
3404 is now memcheck.h. Some people reported problems and strange
3405 behaviour when (incorrectly) including valgrind.h in code with
3406 1.9.1 -- 1.9.4. This is no longer possible.
3407
3408- Add some __extension__ bits and pieces so that gcc configured
3409 for valgrind-checking compiles even with -Werror. If you
3410 don't understand this, ignore it. Of interest to gcc developers
3411 only.
3412
3413- Removed a pointless check which caused problems interworking
3414 with Clearcase. V would complain about shared objects whose
3415 names did not end ".so", and refuse to run. This is now fixed.
3416 In fact it was fixed in 1.9.4 but not documented.
3417
3418- Fixed a bug causing an assertion failure of "waiters == 1"
3419 somewhere in vg_scheduler.c, when running large threaded apps,
3420 notably MySQL.
3421
3422- Add support for the munlock system call (124).
3423
3424Some comments about future releases:
3425
34261.9.5 is, we hope, the most stable Valgrind so far. It pretty much
3427supersedes the 1.0.X branch. If you are a valgrind packager, please
3428consider making 1.9.5 available to your users. You can regard the
34291.0.X branch as obsolete: 1.9.5 is stable and vastly superior. There
3430are no plans at all for further releases of the 1.0.X branch.
3431
3432If you want a leading-edge valgrind, consider building the cvs head
3433(from SourceForge), or getting a snapshot of it. Current cool stuff
3434going in includes MMX support (done); SSE/SSE2 support (in progress),
3435a significant (10-20%) performance improvement (done), and the usual
3436large collection of minor changes. Hopefully we will be able to
3437improve our NPTL support, but no promises.
3438