blob: c00c41445669b5b1222afb5ee6f198ab192565e7 [file] [log] [blame]
njn718d3b12006-12-16 00:54:12 +00001
njne5930da2010-12-17 00:45:19 +00002Release 3.7.0 (???)
3~~~~~~~~~~~~~~~~~~~
4- Added the --mod-funcname option to cg_diff.
sewardj4fd57292011-02-17 12:20:19 +00005- Further reduction in overheads caused by --smc-check=all, especially
6 on 64-bit targets.
sewardj6dbcc632011-06-07 21:39:28 +00007- new variant --smc-check=all-non-file
njne5930da2010-12-17 00:45:19 +00008
sewardj70ceabc2011-06-24 18:23:42 +00009- hg: performance improvements and memory use reductions, particularly
10 for large, long running applications which perform many synch events.
11
12 showing of locksets for both threads involved in a race
13
14 general improvement of formatting/clarity of error messages
15
16 add facilities and documentation regarding annotation of thread safe
17 reference counted C++ classes
18
19 new flag --check-stack-refs=no|yes [yes], to disable race checking
20 on thread stacks (performance hack)
21
22 new flag --free-is-write=no|yes [no], to enable detection of races
23 where one thread accesses heap memory but another one frees it,
24 without any coordinating synchronisation event
25
sewardj4629a732011-05-17 15:54:31 +000026* IBM z/Architecture (s390x) running Linux
27 Valgrind can analyse 64-bit programs running on z/Architecture.
28 Most user space instructions up to and including z10 are supported.
29 Valgrind has been tested extensively on z9, z10, and z196 machines
30 running SLES 10/11, RedHat 5/6m, and Fedora. The Memcheck and Massif
31 tools are known to work well. Callgrind, Helgrind, and DRD work
32 reasonably well on z9 and later models. See README.s390 for more
33 details.
34
sewardj7a9ff4c2011-06-13 13:36:59 +000035bugs fixed (last update 11 June 2011):
36
37* don't be spooked by libxul.so linked with gold (rXXXX)
38* don't be spooked by libraries mashed by elfhack# (rXXXX)
39* cachegrind/callgrind: handle CPUID information for Core iX Intel
40 CPUs that with non-power-of-2 sizes (also AMDs)
41
42243404 Port to zSeries
43Fixed 3.7
44
45265762 make public VEX headers compilable by G++ 3.x
46Fixed 3.7
47
48265771 assertion in jumps.c (r11523) fails with glibc-2.3
49Fixed 3.7
50
51266753 valgrind's configure script does not give the user the option
52 to not use QtCore
53fixed, apparently
54
55266931 gen_insn_test.pl is broken
56fixed
57
58266961 ld-linux.so.2 i?86-linux strlen issues
59fixed
60
61266990 setns instruction causes false positive
62fixed
63
64243935 Helgrind: implementation of ANNOTATE_HAPPENS_BEFORE() /
65 AFTER() is not correct
66fixed, r11624
67
68247223 non-x86: Suppress warning: 'regparm' attribute directive
69 ignored
70fixed
71
72
73267383 Assertion 'vgPlain_strlen(dir) + vgPlain_strlen(file) + 1 <
74 256' failed.
75fixed
76
77267413 Assertion 'DRD_(g_threadinfo)[tid].synchr_nesting >= 1'
78 failed.
79fixed
80
81210935 port valgrind.h (not valgrind) to win32 so apps run under
82 wine can make client requests
83afaict, this was fixed in 3.6.1 but is not listed in NEWS
84
85267488 regtest: darwin support for 64-bit build
86fixed
87
88267552 SIGSEGV (misaligned_stack_error) with DRD, but not with other
89 tools
90fixed, but is the next one also fixed?
91
92267630 Add support for IBM Power ISA 2.06 -- stage 1
93fixed
94
95267819 Add client request for informing the core about reallocation
96fixed
97
98267968 drd: drd_thread.c:567 (vgDrd_thread_set_joinable): Assertion
99 '0 <= (int)tid && tid < DRD_N_THREADS && tid != DRD_INVALID_THREADID'
100 failed.
101fixed
102
103214223 valgrind SIGSEGV on startup gcc 4.4.1 ppc32 (G4) Ubuntu 9.10 ==
104259977 Valgrind segfaults doing __builtin_longjmp
105fixed
106
107268792 - valgrind seg faults on startup when compiled with Xcode 4 compilers...
108267769 - Darwin: memcheck triggers segmentation fault
109274784 - valgrind ls -l or any other valgrind call(even without parameters) results in Segmentation Fault
110267342 - segmentation fault on Mac OS 10.6
111271337 - Valgrind segfaults on MacOS X
112270309 - valgrind crash on startup
113269641 - valgrind segfaults immediately (segmentation fault)
114267997 MacOSX: 64-bit valgrind segfaults on launch when built with
115 Xcode 4.0.1
116fixed
117
118264800 testcase compile failure on zseries
119fixed
120
121265762 - make public VEX headers compilable by G++ 3.x
122fixed
123
124268513] New: missed optimizations in fold_Expr
125fixed
126
127253206 - Some fixes for the faultstatus testcase
128fixed
129
130268619 - s390x: fpr - gpr transfer facility
131fixed
132
133268620 - s390x: reconsider "long displacement" requirement
134fixed
135
136268621 - s390x: improve IR generation for XC
137fixed
138
139255223 - [PATCH] capget testcase fails when running as root
140fixed
141
142268715 - s390x: FLOGR is not universally available
143fixed
144
145268930 - s390x: MHY is not universally available
146fixed
147
148269078 - [PATCH] vex: arm->IR: unhandled instruction SUB (SP minus
149immediate/register)
150fixed
151
152269079 - [PATCH] Support ptrace system call on ARM
153fixed
154
155269144 - missing "Bad option" error message
156fixed
157
158269209] New: [PATCH] conditional load and store facility (z196)
159fixed
160
161269354] New: Shift by zero on x86 can incorrectly clobber CC_NDEP
162(with patch)
163fixed
164
165256726 - Helgrind tests have broken inline asm
166fixed
167
168269736 - s390x: minor code generation tweaks
169fixed
170
171256703 - xlc_dbl_u32.c testcase broken
172fixed
173
174272986 - gcc-4.6 warnings with valgrind.h ==
175269778] New: valgrind.h: swap roles of VALGRIND_DO_CLIENT_REQUEST()
176and VALGRIND_DO_CLIENT_REQUEST_EXPR()
177fixed
178
179269863 - s390x: remove unused function parameters
180fixed
181
182269864 - s390x: tweak s390_emit_load_cc
183fixed
184
185270115] New: s390x: rewrite some testcases
186fixed
187
188270082 - s390x: [PATCH] Make sure to point the PSW address to the next
189address on SIGILL
190fixed
191
192270794 - New IBM POWER7 support patch causes regression in none/tests
193fixed
194
195270851 - New IBM POWER7 fcfidus instruction causes memcheck to fail
196fixed
197
198270856 - New IBM POWER7 xsnmaddadp instruction causes memcheck to fail
199on 32bit app
200fixed
201
202270959 - s390x: invalid use of R0 as base register
203fixed
204
205271042 - VSX configure check fails when it should not
206fixed
207
208271043 - Valgrind build fails with assembler error on ppc64 with
209binutils 2.21
210fixed
211
212271259 - s390x: fix code confusion
213fixed
214
215271385 - s390x: Implement Ist_MBE
216fixed
217
218271501 - s390x : misc cleanups
219fixed
220
221271504 - s390x: promote likely and unlikely
222fixed
223
224271579 - ppc: using wrong enum type
225fixed
226
227271730 - [PATCH] Fix bug when checking ioctls: duplicate check
228fixed
229
230271779 - s390x: provide clock instructions like STCK
231fixed
232
233271799 - Darwin: ioctls without an arg report a memory error
234fixed
235
236271820 - arm: fix type confusion
237fixed
238
239272067 - s390x: fix DISP20 macro
240fixed
241
242272615 - A typo in debug output in mc_leakcheck.c
243fixed
244
245272661 - callgrind_annotate chokes when run from paths containing
246regex metacharacters
247fixed
248
249272955 - Unhandled syscall error for pwrite64 on ppc64 arch
250fixed
251
252274447] New: WARNING: unhandled syscall: 340
253fixed
254
255275148] New: configure FAIL with glibc-2.14
256fixed
257
258275151] New: Fedora 15 / glibc-2.14 'make regtest' FAIL
259fixed
260
261
njne5930da2010-12-17 00:45:19 +0000262
sewardj4fd57292011-02-17 12:20:19 +0000263Release 3.6.1 (16 February 2011)
264~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2653.6.1 is a bug fix release. It adds support for some SSE4
266instructions that were omitted in 3.6.0 due to lack of time. Initial
267support for glibc-2.13 has been added. A number of bugs causing
268crashing or assertion failures have been fixed.
njne5930da2010-12-17 00:45:19 +0000269
sewardj4fd57292011-02-17 12:20:19 +0000270The following bugs have been fixed or resolved. Note that "n-i-bz"
271stands for "not in bugzilla" -- that is, a bug that was reported to us
272but never got a bugzilla entry. We encourage you to file bugs in
273bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
274mailing the developers (or mailing lists) directly -- bugs that are
275not entered into bugzilla tend to get forgotten about or ignored.
bart8e0d2c12011-02-04 19:07:11 +0000276
sewardj4fd57292011-02-17 12:20:19 +0000277To see details of a given bug, visit
278https://bugs.kde.org/show_bug.cgi?id=XXXXXX
279where XXXXXX is the bug number as listed below.
280
281188572 Valgrind on Mac should suppress setenv() mem leak
282194402 vex amd64->IR: 0x48 0xF 0xAE 0x4 (proper FX{SAVE,RSTOR} support)
283210481 vex amd64->IR: Assertion `sz == 2 || sz == 4' failed (REX.W POPQ)
284246152 callgrind internal error after pthread_cancel on 32 Bit Linux
285250038 ppc64: Altivec LVSR and LVSL instructions fail their regtest
286254420 memory pool tracking broken
287254957 Test code failing to compile due to changes in memcheck.h
288255009 helgrind/drd: crash on chmod with invalid parameter
289255130 readdwarf3.c parse_type_DIE confused by GNAT Ada types
290255355 helgrind/drd: crash on threaded programs doing fork
291255358 == 255355
292255418 (SSE4.x) rint call compiled with ICC
293255822 --gen-suppressions can create invalid files: "too many callers [...]"
294255888 closing valgrindoutput tag outputted to log-stream on error
295255963 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x9 0xDB 0x0 (ROUNDPD)
296255966 Slowness when using mempool annotations
297256387 vex x86->IR: 0xD4 0xA 0x2 0x7 (AAD and AAM)
298256600 super-optimized strcasecmp() false positive
299256669 vex amd64->IR: Unhandled LOOPNEL insn on amd64
300256968 (SSE4.x) vex amd64->IR: 0x66 0xF 0x38 0x10 0xD3 0x66 (BLENDVPx)
301257011 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0xE 0xFD 0xA0 (PBLENDW)
302257063 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x8 0xC0 0x0 (ROUNDPS)
303257276 Missing case in memcheck --track-origins=yes
304258870 (SSE4.x) Add support for EXTRACTPS SSE 4.1 instruction
305261966 (SSE4.x) support for CRC32B and CRC32Q is lacking (also CRC32{W,L})
306262985 VEX regression in valgrind 3.6.0 in handling PowerPC VMX
307262995 (SSE4.x) crash when trying to valgrind gcc-snapshot (PCMPxSTRx $0)
308263099 callgrind_annotate counts Ir improperly [...]
309263877 undefined coprocessor instruction on ARMv7
310265964 configure FAIL with glibc-2.13
311n-i-bz Fix compile error w/ icc-12.x in guest_arm_toIR.c
312n-i-bz Docs: fix bogus descriptions for VALGRIND_CREATE_BLOCK et al
313n-i-bz Massif: don't assert on shmat() with --pages-as-heap=yes
314n-i-bz Bug fixes and major speedups for the exp-DHAT space profiler
315n-i-bz DRD: disable --free-is-write due to implementation difficulties
316
317(3.6.1: 16 February 2011, vex r2103, valgrind r11561).
318
bart8e0d2c12011-02-04 19:07:11 +0000319
320
sewardj695f13a2010-10-20 17:07:01 +0000321Release 3.6.0 (21 October 2010)
sewardj18a08922010-10-14 00:26:42 +0000322~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3233.6.0 is a feature release with many significant improvements and the
324usual collection of bug fixes.
sewardj00d9d1e2010-10-13 14:05:30 +0000325
sewardj18a08922010-10-14 00:26:42 +0000326This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux,
327PPC64/Linux, X86/Darwin and AMD64/Darwin. Support for recent distros
328and toolchain components (glibc 2.12, gcc 4.5, OSX 10.6) has been added.
sewardj00d9d1e2010-10-13 14:05:30 +0000329
sewardj18a08922010-10-14 00:26:42 +0000330 -------------------------
sewardj00d9d1e2010-10-13 14:05:30 +0000331
sewardj18a08922010-10-14 00:26:42 +0000332Here are some highlights. Details are shown further down:
njn9b919b62010-02-28 23:41:07 +0000333
sewardj18a08922010-10-14 00:26:42 +0000334* Support for ARM/Linux.
njn9b919b62010-02-28 23:41:07 +0000335
sewardj18a08922010-10-14 00:26:42 +0000336* Support for recent Linux distros: Ubuntu 10.10 and Fedora 14.
337
338* Support for Mac OS X 10.6, both 32- and 64-bit executables.
339
340* Support for the SSE4.2 instruction set.
341
342* Enhancements to the Callgrind profiler, including the ability to
343 handle CPUs with three levels of cache.
344
345* A new experimental heap profiler, DHAT.
346
sewardj695f13a2010-10-20 17:07:01 +0000347* A huge number of bug fixes and small enhancements.
sewardj18a08922010-10-14 00:26:42 +0000348
349 -------------------------
350
351Here are details of the above changes, together with descriptions of
352many other changes, and a list of fixed bugs.
353
354* ================== PLATFORM CHANGES =================
355
356* Support for ARM/Linux. Valgrind now runs on ARMv7 capable CPUs
357 running Linux. It is known to work on Ubuntu 10.04, Ubuntu 10.10,
358 and Maemo 5, so you can run Valgrind on your Nokia N900 if you want.
359
360 This requires a CPU capable of running the ARMv7-A instruction set
361 (Cortex A5, A8 and A9). Valgrind provides fairly complete coverage
362 of the user space instruction set, including ARM and Thumb integer
363 code, VFPv3, NEON and V6 media instructions. The Memcheck,
364 Cachegrind and Massif tools work properly; other tools work to
365 varying degrees.
366
367* Support for recent Linux distros (Ubuntu 10.10 and Fedora 14), along
368 with support for recent releases of the underlying toolchain
369 components, notably gcc-4.5 and glibc-2.12.
370
371* Support for Mac OS X 10.6, both 32- and 64-bit executables. 64-bit
372 support also works much better on OS X 10.5, and is as solid as
373 32-bit support now.
374
375* Support for the SSE4.2 instruction set. SSE4.2 is supported in
376 64-bit mode. In 32-bit mode, support is only available up to and
377 including SSSE3. Some exceptions: SSE4.2 AES instructions are not
378 supported in 64-bit mode, and 32-bit mode does in fact support the
sewardj695f13a2010-10-20 17:07:01 +0000379 bare minimum SSE4 instructions to needed to run programs on Mac OS X
sewardj18a08922010-10-14 00:26:42 +0000380 10.6 on 32-bit targets.
381
382* Support for IBM POWER6 cpus has been improved. The Power ISA up to
383 and including version 2.05 is supported.
384
385* ==================== TOOL CHANGES ====================
386
387* Cachegrind has a new processing script, cg_diff, which finds the
388 difference between two profiles. It's very useful for evaluating
389 the performance effects of a change in a program.
njn69d495d2010-06-30 05:23:34 +0000390
391 Related to this change, the meaning of cg_annotate's (rarely-used)
sewardj18a08922010-10-14 00:26:42 +0000392 --threshold option has changed; this is unlikely to affect many
393 people, if you do use it please see the user manual for details.
njn9b919b62010-02-28 23:41:07 +0000394
sewardj18a08922010-10-14 00:26:42 +0000395* Callgrind now can do branch prediction simulation, similar to
396 Cachegrind. In addition, it optionally can count the number of
397 executed global bus events. Both can be used for a better
398 approximation of a "Cycle Estimation" as derived event (you need to
399 update the event formula in KCachegrind yourself).
njn2d853a12010-10-06 22:46:31 +0000400
sewardj18a08922010-10-14 00:26:42 +0000401* Cachegrind and Callgrind now refer to the LL (last-level) cache
402 rather than the L2 cache. This is to accommodate machines with
403 three levels of caches -- if Cachegrind/Callgrind auto-detects the
404 cache configuration of such a machine it will run the simulation as
405 if the L2 cache isn't present. This means the results are less
406 likely to match the true result for the machine, but
407 Cachegrind/Callgrind's results are already only approximate, and
408 should not be considered authoritative. The results are still
409 useful for giving a general idea about a program's locality.
njn2d853a12010-10-06 22:46:31 +0000410
sewardj695f13a2010-10-20 17:07:01 +0000411* Massif has a new option, --pages-as-heap, which is disabled by
sewardj18a08922010-10-14 00:26:42 +0000412 default. When enabled, instead of tracking allocations at the level
413 of heap blocks (as allocated with malloc/new/new[]), it instead
414 tracks memory allocations at the level of memory pages (as mapped by
415 mmap, brk, etc). Each mapped page is treated as its own block.
416 Interpreting the page-level output is harder than the heap-level
417 output, but this option is useful if you want to account for every
418 byte of memory used by a program.
njne323a6b2010-07-01 02:35:03 +0000419
sewardj18a08922010-10-14 00:26:42 +0000420* DRD has two new command-line options: --free-is-write and
421 --trace-alloc. The former allows to detect reading from already freed
422 memory, and the latter allows tracing of all memory allocations and
bart5cf4c392010-10-10 18:54:44 +0000423 deallocations.
424
sewardj18a08922010-10-14 00:26:42 +0000425* DRD has several new annotations. Custom barrier implementations can
426 now be annotated, as well as benign races on static variables.
bart5cf4c392010-10-10 18:54:44 +0000427
sewardj18a08922010-10-14 00:26:42 +0000428* DRD's happens before / happens after annotations have been made more
429 powerful, so that they can now also be used to annotate e.g. a smart
bart5cf4c392010-10-10 18:54:44 +0000430 pointer implementation.
431
sewardj18a08922010-10-14 00:26:42 +0000432* Helgrind's annotation set has also been drastically improved, so as
sewardj695f13a2010-10-20 17:07:01 +0000433 to provide to users a general set of annotations to describe locks,
sewardj18a08922010-10-14 00:26:42 +0000434 semaphores, barriers and condition variables. Annotations to
435 describe thread-safe reference counted heap objects have also been
436 added.
437
438* Memcheck has a new command-line option, --show-possibly-lost, which
439 is enabled by default. When disabled, the leak detector will not
440 show possibly-lost blocks.
441
442* A new experimental heap profiler, DHAT (Dynamic Heap Analysis Tool),
443 has been added. DHAT keeps track of allocated heap blocks, and also
444 inspects every memory reference to see which block (if any) is being
445 accessed. This gives a lot of insight into block lifetimes,
446 utilisation, turnover, liveness, and the location of hot and cold
447 fields. You can use DHAT to do hot-field profiling.
448
449* ==================== OTHER CHANGES ====================
450
451* Improved support for unfriendly self-modifying code: the extra
452 overhead incurred by --smc-check=all has been reduced by
453 approximately a factor of 5 as compared with 3.5.0.
454
455* Ability to show directory names for source files in error messages.
sewardj695f13a2010-10-20 17:07:01 +0000456 This is combined with a flexible mechanism for specifying which
457 parts of the paths should be shown. This is enabled by the new flag
458 --fullpath-after.
sewardj18a08922010-10-14 00:26:42 +0000459
460* A new flag, --require-text-symbol, which will stop the run if a
461 specified symbol is not found it a given shared object when it is
sewardj695f13a2010-10-20 17:07:01 +0000462 loaded into the process. This makes advanced working with function
463 intercepting and wrapping safer and more reliable.
sewardj18a08922010-10-14 00:26:42 +0000464
sewardj695f13a2010-10-20 17:07:01 +0000465* Improved support for the Valkyrie GUI, version 2.0.0. GUI output
466 and control of Valgrind is now available for the tools Memcheck and
467 Helgrind. XML output from Valgrind is available for Memcheck,
468 Helgrind and exp-Ptrcheck.
sewardj18a08922010-10-14 00:26:42 +0000469
470* More reliable stack unwinding on amd64-linux, particularly in the
471 presence of function wrappers, and with gcc-4.5 compiled code.
472
473* Modest scalability (performance improvements) for massive
474 long-running applications, particularly for those with huge amounts
475 of code.
476
477* Support for analyzing programs running under Wine with has been
478 improved. The header files <valgrind/valgrind.h>,
479 <valgrind/memcheck.h> and <valgrind/drd.h> can now be used in
480 Windows-programs compiled with MinGW or one of the Microsoft Visual
481 Studio compilers.
482
483* A rare but serious error in the 64-bit x86 CPU simulation was fixed.
484 The 32-bit simulator was not affected. This did not occur often,
485 but when it did would usually crash the program under test.
486 Bug 245925.
487
488* A large number of bugs were fixed. These are shown below.
489
490* A number of bugs were investigated, and were candidates for fixing,
491 but are not fixed in 3.6.0, due to lack of developer time. They may
492 get fixed in later releases. They are:
493
494 194402 vex amd64->IR: 0x48 0xF 0xAE 0x4 0x24 0x49 (FXSAVE64)
495 212419 false positive "lock order violated" (A+B vs A)
496 213685 Undefined value propagates past dependency breaking instruction
497 216837 Incorrect instrumentation of NSOperationQueue on Darwin
498 237920 valgrind segfault on fork failure
499 242137 support for code compiled by LLVM-2.8
500 242423 Another unknown Intel cache config value
501 243232 Inconsistent Lock Orderings report with trylock
502 243483 ppc: callgrind triggers VEX assertion failure
503 243935 Helgrind: implementation of ANNOTATE_HAPPENS_BEFORE() is wrong
504 244677 Helgrind crash hg_main.c:616 (map_threads_lookup): Assertion
505 'thr' failed.
506 246152 callgrind internal error after pthread_cancel on 32 Bit Linux
507 249435 Analyzing wine programs with callgrind triggers a crash
508 250038 ppc64: Altivec lvsr and lvsl instructions fail their regtest
509 250065 Handling large allocations
510 250101 huge "free" memory usage due to m_mallocfree.c
511 "superblocks fragmentation"
512 251569 vex amd64->IR: 0xF 0x1 0xF9 0x8B 0x4C 0x24 (RDTSCP)
sewardj695f13a2010-10-20 17:07:01 +0000513 252091 Callgrind on ARM does not detect function returns correctly
514 252600 [PATCH] Allow lhs to be a pointer for shl/shr
515 254420 memory pool tracking broken
sewardj18a08922010-10-14 00:26:42 +0000516 n-i-bz support for adding symbols for JIT generated code
517
518
519The following bugs have been fixed or resolved. Note that "n-i-bz"
520stands for "not in bugzilla" -- that is, a bug that was reported to us
521but never got a bugzilla entry. We encourage you to file bugs in
522bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
523mailing the developers (or mailing lists) directly -- bugs that are
524not entered into bugzilla tend to get forgotten about or ignored.
525
526To see details of a given bug, visit
527https://bugs.kde.org/show_bug.cgi?id=XXXXXX
528where XXXXXX is the bug number as listed below.
529
530135264 dcbzl instruction missing
531142688 == 250799
532153699 Valgrind should report unaligned reads with movdqa
533180217 == 212335
534190429 Valgrind reports lost of errors in ld.so
535 with x86_64 2.9.90 glibc
536197266 valgrind appears to choke on the xmms instruction
537 "roundsd" on x86_64
538197988 Crash when demangling very large symbol names
539202315 unhandled syscall: 332 (inotify_init1)
540203256 Add page-level profiling to Massif
541205093 dsymutil=yes needs quotes, locking (partial fix)
542205241 Snow Leopard 10.6 support (partial fix)
543206600 Leak checker fails to upgrade indirect blocks when their
544 parent becomes reachable
545210935 port valgrind.h (not valgrind) to win32 so apps run under
546 wine can make client requests
547211410 vex amd64->IR: 0x15 0xFF 0xFF 0x0 0x0 0x89
548 within Linux ip-stack checksum functions
549212335 unhandled instruction bytes: 0xF3 0xF 0xBD 0xC0
550 (lzcnt %eax,%eax)
551213685 Undefined value propagates past dependency breaking instruction
552 (partial fix)
553215914 Valgrind inserts bogus empty environment variable
554217863 == 197988
555219538 adjtimex syscall wrapper wrong in readonly adjtime mode
556222545 shmat fails under valgind on some arm targets
557222560 ARM NEON support
558230407 == 202315
559231076 == 202315
560232509 Docs build fails with formatting inside <title></title> elements
561232793 == 202315
562235642 [PATCH] syswrap-linux.c: support evdev EVIOCG* ioctls
563236546 vex x86->IR: 0x66 0xF 0x3A 0xA
564237202 vex amd64->IR: 0xF3 0xF 0xB8 0xC0 0x49 0x3B
565237371 better support for VALGRIND_MALLOCLIKE_BLOCK
566237485 symlink (syscall 57) is not supported on Mac OS
567237723 sysno == 101 exp-ptrcheck: the 'impossible' happened:
568 unhandled syscall
569238208 is_just_below_ESP doesn't take into account red-zone
570238345 valgrind passes wrong $0 when executing a shell script
571238679 mq_timedreceive syscall doesn't flag the reception buffer
572 as "defined"
573238696 fcntl command F_DUPFD_CLOEXEC not supported
574238713 unhandled instruction bytes: 0x66 0xF 0x29 0xC6
575238713 unhandled instruction bytes: 0x66 0xF 0x29 0xC6
576238745 3.5.0 Make fails on PPC Altivec opcodes, though configure
577 says "Altivec off"
578239992 vex amd64->IR: 0x48 0xF 0xC4 0xC1 0x0 0x48
579240488 == 197988
580240639 == 212335
581241377 == 236546
582241903 == 202315
583241920 == 212335
584242606 unhandled syscall: setegid (in Ptrcheck)
585242814 Helgrind "Impossible has happened" during
586 QApplication::initInstance();
587243064 Valgrind attempting to read debug information from iso
588243270 Make stack unwinding in Valgrind wrappers more reliable
589243884 exp-ptrcheck: the 'impossible happened: unhandled syscall
590 sysno = 277 (mq_open)
591244009 exp-ptrcheck unknown syscalls in analyzing lighttpd
592244493 ARM VFP d16-d31 registers support
593244670 add support for audit_session_self syscall on Mac OS 10.6
594244921 The xml report of helgrind tool is not well format
595244923 In the xml report file, the <preamble> not escape the
596 xml char, eg '<','&','>'
597245535 print full path names in plain text reports
598245925 x86-64 red zone handling problem
599246258 Valgrind not catching integer underruns + new [] s
600246311 reg/reg cmpxchg doesn't work on amd64
601246549 unhandled syscall unix:277 while testing 32-bit Darwin app
602246888 Improve Makefile.vex.am
603247510 [OS X 10.6] Memcheck reports unaddressable bytes passed
604 to [f]chmod_extended
605247526 IBM POWER6 (ISA 2.05) support is incomplete
606247561 Some leak testcases fails due to reachable addresses in
607 caller save regs
608247875 sizeofIRType to handle Ity_I128
609247894 [PATCH] unhandled syscall sys_readahead
610247980 Doesn't honor CFLAGS passed to configure
611248373 darwin10.supp is empty in the trunk
612248822 Linux FIBMAP ioctl has int parameter instead of long
613248893 [PATCH] make readdwarf.c big endianess safe to enable
614 unwinding on big endian systems
615249224 Syscall 336 not supported (SYS_proc_info)
616249359 == 245535
617249775 Incorrect scheme for detecting NEON capabilities of host CPU
618249943 jni JVM init fails when using valgrind
619249991 Valgrind incorrectly declares AESKEYGENASSIST support
620 since VEX r2011
621249996 linux/arm: unhandled syscall: 181 (__NR_pwrite64)
622250799 frexp$fenv_access_off function generates SIGILL
623250998 vex x86->IR: unhandled instruction bytes: 0x66 0x66 0x66 0x2E
624251251 support pclmulqdq insn
625251362 valgrind: ARM: attach to debugger either fails or provokes
626 kernel oops
627251674 Unhandled syscall 294
sewardj695f13a2010-10-20 17:07:01 +0000628251818 == 254550
sewardj18a08922010-10-14 00:26:42 +0000629
sewardj695f13a2010-10-20 17:07:01 +0000630254257 Add support for debugfiles found by build-id
631254550 [PATCH] Implement DW_ATE_UTF (DWARF4)
632254646 Wrapped functions cause stack misalignment on OS X
633 (and possibly Linux)
634254556 ARM: valgrinding anything fails with SIGSEGV for 0xFFFF0FA0
635
sewardje07c2b52010-10-22 06:44:59 +0000636(3.6.0: 21 October 2010, vex r2068, valgrind r11471).
sewardj18a08922010-10-14 00:26:42 +0000637
638
njn9b919b62010-02-28 23:41:07 +0000639
sewardj6bbe0242009-08-19 11:38:54 +0000640Release 3.5.0 (19 August 2009)
641~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardj570e2442009-08-14 13:42:41 +00006423.5.0 is a feature release with many significant improvements and the
643usual collection of bug fixes. The main improvement is that Valgrind
644now works on Mac OS X.
645
646This release supports X86/Linux, AMD64/Linux, PPC32/Linux, PPC64/Linux
647and X86/Darwin. Support for recent distros and toolchain components
648(glibc 2.10, gcc 4.5) has been added.
649
650 -------------------------
651
652Here is a short summary of the changes. Details are shown further
653down:
654
njnf95c30b2009-08-16 00:20:58 +0000655* Support for Mac OS X (10.5.x).
sewardj570e2442009-08-14 13:42:41 +0000656
njn78b7bb32009-08-14 22:34:30 +0000657* Improvements and simplifications to Memcheck's leak checker.
sewardj570e2442009-08-14 13:42:41 +0000658
njn78b7bb32009-08-14 22:34:30 +0000659* Clarification and simplifications in various aspects of Valgrind's
660 text output.
sewardj570e2442009-08-14 13:42:41 +0000661
njn78b7bb32009-08-14 22:34:30 +0000662* XML output for Helgrind and Ptrcheck.
sewardj570e2442009-08-14 13:42:41 +0000663
njn78b7bb32009-08-14 22:34:30 +0000664* Performance and stability improvements for Helgrind and DRD.
sewardj570e2442009-08-14 13:42:41 +0000665
njn78b7bb32009-08-14 22:34:30 +0000666* Genuinely atomic support for x86/amd64/ppc atomic instructions.
sewardj570e2442009-08-14 13:42:41 +0000667
njn78b7bb32009-08-14 22:34:30 +0000668* A new experimental tool, BBV, useful for computer architecture
669 research.
sewardj570e2442009-08-14 13:42:41 +0000670
njn78b7bb32009-08-14 22:34:30 +0000671* Improved Wine support, including ability to read Windows PDB
672 debuginfo.
sewardj570e2442009-08-14 13:42:41 +0000673
674 -------------------------
675
sewardj9176c862009-08-18 13:47:31 +0000676Here are details of the above changes, followed by descriptions of
677many other minor changes, and a list of fixed bugs.
sewardj570e2442009-08-14 13:42:41 +0000678
679
680* Valgrind now runs on Mac OS X. (Note that Mac OS X is sometimes
njnf95c30b2009-08-16 00:20:58 +0000681 called "Darwin" because that is the name of the OS core, which is the
682 level that Valgrind works at.)
njnf76d27a2009-05-28 01:53:07 +0000683
njnf95c30b2009-08-16 00:20:58 +0000684 Supported systems:
njn6eccab92009-08-14 02:41:37 +0000685
njnf95c30b2009-08-16 00:20:58 +0000686 - It requires OS 10.5.x (Leopard). Porting to 10.4.x is not planned
687 because it would require work and 10.4 is only becoming less common.
njn6eccab92009-08-14 02:41:37 +0000688
njnf95c30b2009-08-16 00:20:58 +0000689 - 32-bit programs on x86 and AMD64 (a.k.a x86-64) machines are supported
690 fairly well. For 10.5.x, 32-bit programs are the default even on
691 64-bit machines, so it handles most current programs.
692
693 - 64-bit programs on x86 and AMD64 (a.k.a x86-64) machines are not
694 officially supported, but simple programs at least will probably work.
695 However, start-up is slow.
njn6eccab92009-08-14 02:41:37 +0000696
sewardj570e2442009-08-14 13:42:41 +0000697 - PowerPC machines are not supported.
njn6eccab92009-08-14 02:41:37 +0000698
njnc6ab7b62009-06-04 22:47:21 +0000699 Things that don't work:
njn6eccab92009-08-14 02:41:37 +0000700
701 - The Ptrcheck tool.
702
703 - Objective-C garbage collection.
704
705 - --db-attach=yes.
706
sewardj570e2442009-08-14 13:42:41 +0000707 - If you have Rogue Amoeba's "Instant Hijack" program installed,
708 Valgrind will fail with a SIGTRAP at start-up. See
709 https://bugs.kde.org/show_bug.cgi?id=193917 for details and a
710 simple work-around.
njnc6ab7b62009-06-04 22:47:21 +0000711
njn97db7612009-08-04 02:32:55 +0000712 Usage notes:
njn6eccab92009-08-14 02:41:37 +0000713
sewardj570e2442009-08-14 13:42:41 +0000714 - You will likely find --dsymutil=yes a useful option, as error
715 messages may be imprecise without it.
njn6eccab92009-08-14 02:41:37 +0000716
sewardj570e2442009-08-14 13:42:41 +0000717 - Mac OS X support is new and therefore will be less robust than the
njn1205e022009-08-14 02:12:54 +0000718 Linux support. Please report any bugs you find.
njn97db7612009-08-04 02:32:55 +0000719
sewardj570e2442009-08-14 13:42:41 +0000720 - Threaded programs may run more slowly than on Linux.
721
njnc6ab7b62009-06-04 22:47:21 +0000722 Many thanks to Greg Parker for developing this port over several years.
723
sewardj570e2442009-08-14 13:42:41 +0000724
725* Memcheck's leak checker has been improved.
726
727 - The results for --leak-check=summary now match the summary results
728 for --leak-check=full. Previously they could differ because
729 --leak-check=summary counted "indirectly lost" blocks and
730 "suppressed" blocks as "definitely lost".
731
732 - Blocks that are only reachable via at least one interior-pointer,
733 but are directly pointed to by a start-pointer, were previously
734 marked as "still reachable". They are now correctly marked as
735 "possibly lost".
736
737 - The default value for the --leak-resolution option has been
738 changed from "low" to "high". In general, this means that more
739 leak reports will be produced, but each leak report will describe
740 fewer leaked blocks.
741
742 - With --leak-check=full, "definitely lost" and "possibly lost"
743 leaks are now considered as proper errors, ie. they are counted
744 for the "ERROR SUMMARY" and affect the behaviour of
745 --error-exitcode. These leaks are not counted as errors if
746 --leak-check=summary is specified, however.
747
748 - Documentation for the leak checker has been improved.
749
njnf76d27a2009-05-28 01:53:07 +0000750
njn6eccab92009-08-14 02:41:37 +0000751* Various aspects of Valgrind's text output have changed.
njndbebecc2009-07-14 01:39:54 +0000752
njn6eccab92009-08-14 02:41:37 +0000753 - Valgrind's start-up message has changed. It is shorter but also
754 includes the command being run, which makes it easier to use
755 --trace-children=yes. An example:
njn53162bf2009-07-29 23:34:49 +0000756
njn6eccab92009-08-14 02:41:37 +0000757 - Valgrind's shut-down messages have also changed. This is most
sewardj570e2442009-08-14 13:42:41 +0000758 noticeable with Memcheck, where the leak summary now occurs before
759 the error summary. This change was necessary to allow leaks to be
760 counted as proper errors (see the description of the leak checker
761 changes above for more details). This was also necessary to fix a
762 longstanding bug in which uses of suppressions against leaks were
763 not "counted", leading to difficulties in maintaining suppression
sewardj783e9d62009-08-19 22:18:35 +0000764 files (see https://bugs.kde.org/show_bug.cgi?id=186790).
njn6eccab92009-08-14 02:41:37 +0000765
sewardj570e2442009-08-14 13:42:41 +0000766 - Behavior of -v has changed. In previous versions, -v printed out
767 a mixture of marginally-user-useful information, and tool/core
768 statistics. The statistics printing has now been moved to its own
769 flag, --stats=yes. This means -v is less verbose and more likely
770 to convey useful end-user information.
njn6eccab92009-08-14 02:41:37 +0000771
sewardj570e2442009-08-14 13:42:41 +0000772 - The format of some (non-XML) stack trace entries has changed a
773 little. Previously there were six possible forms:
njn6eccab92009-08-14 02:41:37 +0000774
775 0x80483BF: really (a.c:20)
776 0x80483BF: really (in /foo/a.out)
777 0x80483BF: really
778 0x80483BF: (within /foo/a.out)
779 0x80483BF: ??? (a.c:20)
780 0x80483BF: ???
781
sewardj570e2442009-08-14 13:42:41 +0000782 The third and fourth of these forms have been made more consistent
783 with the others. The six possible forms are now:
njn6eccab92009-08-14 02:41:37 +0000784
785 0x80483BF: really (a.c:20)
786 0x80483BF: really (in /foo/a.out)
787 0x80483BF: really (in ???)
788 0x80483BF: ??? (in /foo/a.out)
789 0x80483BF: ??? (a.c:20)
790 0x80483BF: ???
791
sewardj570e2442009-08-14 13:42:41 +0000792 Stack traces produced when --xml=yes is specified are different
793 and unchanged.
njn6eccab92009-08-14 02:41:37 +0000794
njn6eccab92009-08-14 02:41:37 +0000795
sewardj570e2442009-08-14 13:42:41 +0000796* Helgrind and Ptrcheck now support XML output, so they can be used
797 from GUI tools. Also, the XML output mechanism has been
798 overhauled.
njn6eccab92009-08-14 02:41:37 +0000799
njn78b7bb32009-08-14 22:34:30 +0000800 - The XML format has been overhauled and generalised, so it is more
sewardj570e2442009-08-14 13:42:41 +0000801 suitable for error reporting tools in general. The Memcheck
802 specific aspects of it have been removed. The new format, which
803 is an evolution of the old format, is described in
804 docs/internals/xml-output-protocol4.txt.
njn6eccab92009-08-14 02:41:37 +0000805
sewardj570e2442009-08-14 13:42:41 +0000806 - Memcheck has been updated to use the new format.
njn6eccab92009-08-14 02:41:37 +0000807
sewardj570e2442009-08-14 13:42:41 +0000808 - Helgrind and Ptrcheck are now able to emit output in this format.
njn6eccab92009-08-14 02:41:37 +0000809
sewardj570e2442009-08-14 13:42:41 +0000810 - The XML output mechanism has been overhauled. XML is now output
811 to its own file descriptor, which means that:
njn6eccab92009-08-14 02:41:37 +0000812
sewardj570e2442009-08-14 13:42:41 +0000813 * Valgrind can output text and XML independently.
njnb4e06872009-08-10 05:11:43 +0000814
sewardj570e2442009-08-14 13:42:41 +0000815 * The longstanding problem of XML output being corrupted by
816 unexpected un-tagged text messages is solved.
njn8df80b22009-03-02 05:11:06 +0000817
sewardj570e2442009-08-14 13:42:41 +0000818 As before, the destination for text output is specified using
819 --log-file=, --log-fd= or --log-socket=.
njnd55f0d92009-08-03 01:38:56 +0000820
sewardj570e2442009-08-14 13:42:41 +0000821 As before, XML output for a tool is enabled using --xml=yes.
njnc6ab7b62009-06-04 22:47:21 +0000822
sewardj570e2442009-08-14 13:42:41 +0000823 Because there's a new XML output channel, the XML output
824 destination is now specified by --xml-file=, --xml-fd= or
825 --xml-socket=.
njn6eccab92009-08-14 02:41:37 +0000826
sewardj570e2442009-08-14 13:42:41 +0000827 Initial feedback has shown this causes some confusion. To
828 clarify, the two envisaged usage scenarios are:
njn6eccab92009-08-14 02:41:37 +0000829
sewardj570e2442009-08-14 13:42:41 +0000830 (1) Normal text output. In this case, do not specify --xml=yes
831 nor any of --xml-file=, --xml-fd= or --xml-socket=.
832
833 (2) XML output. In this case, specify --xml=yes, and one of
834 --xml-file=, --xml-fd= or --xml-socket= to select the XML
835 destination, one of --log-file=, --log-fd= or --log-socket=
836 to select the destination for any remaining text messages,
837 and, importantly, -q.
838
839 -q makes Valgrind completely silent on the text channel,
840 except in the case of critical failures, such as Valgrind
841 itself segfaulting, or failing to read debugging information.
842 Hence, in this scenario, it suffices to check whether or not
843 any output appeared on the text channel. If yes, then it is
844 likely to be a critical error which should be brought to the
845 attention of the user. If no (the text channel produced no
846 output) then it can be assumed that the run was successful.
847
848 This allows GUIs to make the critical distinction they need to
849 make (did the run fail or not?) without having to search or
850 filter the text output channel in any way.
851
852 It is also recommended to use --child-silent-after-fork=yes in
853 scenario (2).
854
855
856* Improvements and changes in Helgrind:
857
858 - XML output, as described above
859
860 - Checks for consistent association between pthread condition
861 variables and their associated mutexes are now performed.
862
863 - pthread_spinlock functions are supported.
864
865 - Modest performance improvements.
866
867 - Initial (skeletal) support for describing the behaviour of
868 non-POSIX synchronisation objects through ThreadSanitizer
869 compatible ANNOTATE_* macros.
870
871 - More controllable tradeoffs between performance and the level of
872 detail of "previous" accesses in a race. There are now three
873 settings:
874
875 * --history-level=full. This is the default, and was also the
876 default in 3.4.x. It shows both stacks involved in a race, but
877 requires a lot of memory and can be very slow in programs that
878 do many inter-thread synchronisation events.
879
880 * --history-level=none. This only shows the later stack involved
881 in a race. This can be much faster than --history-level=full,
882 but makes it much more difficult to find the other access
883 involved in the race.
884
885 The new intermediate setting is
886
887 * --history-level=approx
888
889 For the earlier (other) access, two stacks are presented. The
890 earlier access is guaranteed to be somewhere in between the two
891 program points denoted by those stacks. This is not as useful
892 as showing the exact stack for the previous access (as per
893 --history-level=full), but it is better than nothing, and it's
894 almost as fast as --history-level=none.
895
njnace90d82009-04-27 01:27:58 +0000896
bart6a146fc2009-07-21 07:32:19 +0000897* New features and improvements in DRD:
njn6eccab92009-08-14 02:41:37 +0000898
sewardj570e2442009-08-14 13:42:41 +0000899 - The error messages printed by DRD are now easier to interpret.
900 Instead of using two different numbers to identify each thread
901 (Valgrind thread ID and DRD thread ID), DRD does now identify
902 threads via a single number (the DRD thread ID). Furthermore
903 "first observed at" information is now printed for all error
904 messages related to synchronization objects.
njn6eccab92009-08-14 02:41:37 +0000905
bartfe212db2009-07-25 11:16:51 +0000906 - Added support for named semaphores (sem_open() and sem_close()).
njn6eccab92009-08-14 02:41:37 +0000907
bart6a146fc2009-07-21 07:32:19 +0000908 - Race conditions between pthread_barrier_wait() and
909 pthread_barrier_destroy() calls are now reported.
njn6eccab92009-08-14 02:41:37 +0000910
911 - Added support for custom allocators through the macros
sewardj570e2442009-08-14 13:42:41 +0000912 VALGRIND_MALLOCLIKE_BLOCK() VALGRIND_FREELIKE_BLOCK() (defined in
913 in <valgrind/valgrind.h>). An alternative for these two macros is
914 the new client request VG_USERREQ__DRD_CLEAN_MEMORY (defined in
bart6a146fc2009-07-21 07:32:19 +0000915 <valgrind/drd.h>).
njn6eccab92009-08-14 02:41:37 +0000916
sewardj570e2442009-08-14 13:42:41 +0000917 - Added support for annotating non-POSIX synchronization objects
918 through several new ANNOTATE_*() macros.
njn6eccab92009-08-14 02:41:37 +0000919
sewardj570e2442009-08-14 13:42:41 +0000920 - OpenMP: added support for the OpenMP runtime (libgomp) included
921 with gcc versions 4.3.0 and 4.4.0.
njn6eccab92009-08-14 02:41:37 +0000922
bart6a146fc2009-07-21 07:32:19 +0000923 - Faster operation.
njn6eccab92009-08-14 02:41:37 +0000924
925 - Added two new command-line options (--first-race-only and
bart6a146fc2009-07-21 07:32:19 +0000926 --segment-merging-interval).
927
njn6eccab92009-08-14 02:41:37 +0000928
sewardj570e2442009-08-14 13:42:41 +0000929* Genuinely atomic support for x86/amd64/ppc atomic instructions
930
931 Valgrind will now preserve (memory-access) atomicity of LOCK-
932 prefixed x86/amd64 instructions, and any others implying a global
933 bus lock. Ditto for PowerPC l{w,d}arx/st{w,d}cx. instructions.
934
935 This means that Valgrinded processes will "play nicely" in
936 situations where communication with other processes, or the kernel,
937 is done through shared memory and coordinated with such atomic
938 instructions. Prior to this change, such arrangements usually
939 resulted in hangs, races or other synchronisation failures, because
940 Valgrind did not honour atomicity of such instructions.
941
942
943* A new experimental tool, BBV, has been added. BBV generates basic
944 block vectors for use with the SimPoint analysis tool, which allows
945 a program's overall behaviour to be approximated by running only a
946 fraction of it. This is useful for computer architecture
947 researchers. You can run BBV by specifying --tool=exp-bbv (the
948 "exp-" prefix is short for "experimental"). BBV was written by
949 Vince Weaver.
950
951
952* Modestly improved support for running Windows applications under
953 Wine. In particular, initial support for reading Windows .PDB debug
954 information has been added.
955
956
957* A new Memcheck client request VALGRIND_COUNT_LEAK_BLOCKS has been
958 added. It is similar to VALGRIND_COUNT_LEAKS but counts blocks
959 instead of bytes.
960
961
962* The Valgrind client requests VALGRIND_PRINTF and
963 VALGRIND_PRINTF_BACKTRACE have been changed slightly. Previously,
964 the string was always printed immediately on its own line. Now, the
965 string will be added to a buffer but not printed until a newline is
966 encountered, or other Valgrind output is printed (note that for
967 VALGRIND_PRINTF_BACKTRACE, the back-trace itself is considered
968 "other Valgrind output"). This allows you to use multiple
969 VALGRIND_PRINTF calls to build up a single output line, and also to
970 print multiple output lines with a single request (by embedding
971 multiple newlines in the string).
972
973
974* The graphs drawn by Massif's ms_print program have changed slightly:
975
976 - The half-height chars '.' and ',' are no longer drawn, because
977 they are confusing. The --y option can be used if the default
978 y-resolution is not high enough.
979
980 - Horizontal lines are now drawn after the top of a snapshot if
981 there is a gap until the next snapshot. This makes it clear that
982 the memory usage has not dropped to zero between snapshots.
983
984
985* Something that happened in 3.4.0, but wasn't clearly announced: the
986 option --read-var-info=yes can be used by some tools (Memcheck,
987 Helgrind and DRD). When enabled, it causes Valgrind to read DWARF3
988 variable type and location information. This makes those tools
989 start up more slowly and increases memory consumption, but
990 descriptions of data addresses in error messages become more
991 detailed.
992
993
994* exp-Omega, an experimental instantaneous leak-detecting tool, was
995 disabled in 3.4.0 due to a lack of interest and maintenance,
996 although the source code was still in the distribution. The source
997 code has now been removed from the distribution. For anyone
998 interested, the removal occurred in SVN revision r10247.
999
njnacda1782009-06-04 23:11:50 +00001000
njn8b68b642009-06-24 00:37:09 +00001001* Some changes have been made to the build system.
njn6bf365c2009-02-11 00:35:45 +00001002
sewardj570e2442009-08-14 13:42:41 +00001003 - VEX/ is now integrated properly into the build system. This means
1004 that dependency tracking within VEX/ now works properly, "make
1005 install" will work without requiring "make" before it, and
1006 parallel builds (ie. 'make -j') now work (previously a
1007 .NOTPARALLEL directive was used to serialize builds, ie. 'make -j'
1008 was effectively ignored).
njn6bf365c2009-02-11 00:35:45 +00001009
sewardj570e2442009-08-14 13:42:41 +00001010 - The --with-vex configure option has been removed. It was of
1011 little use and removing it simplified the build system.
njn8b68b642009-06-24 00:37:09 +00001012
sewardj570e2442009-08-14 13:42:41 +00001013 - The location of some install files has changed. This should not
1014 affect most users. Those who might be affected:
njn8b68b642009-06-24 00:37:09 +00001015
1016 * For people who use Valgrind with MPI programs, the installed
sewardj570e2442009-08-14 13:42:41 +00001017 libmpiwrap.so library has moved from
1018 $(INSTALL)/<platform>/libmpiwrap.so to
1019 $(INSTALL)/libmpiwrap-<platform>.so.
njn8b68b642009-06-24 00:37:09 +00001020
sewardj570e2442009-08-14 13:42:41 +00001021 * For people who distribute standalone Valgrind tools, the
1022 installed libraries such as $(INSTALL)/<platform>/libcoregrind.a
1023 have moved to $(INSTALL)/libcoregrind-<platform>.a.
njn8b68b642009-06-24 00:37:09 +00001024
sewardj570e2442009-08-14 13:42:41 +00001025 These changes simplify the build system.
njn8b68b642009-06-24 00:37:09 +00001026
sewardj570e2442009-08-14 13:42:41 +00001027 - Previously, all the distributed suppression (*.supp) files were
1028 installed. Now, only default.supp is installed. This should not
1029 affect users as the other installed suppression files were not
1030 read; the fact that they were installed was a mistake.
1031
1032
1033* KNOWN LIMITATIONS:
1034
1035 - Memcheck is unusable with the Intel compiler suite version 11.1,
1036 when it generates code for SSE2-and-above capable targets. This
1037 is because of icc's use of highly optimised inlined strlen
1038 implementations. It causes Memcheck to report huge numbers of
1039 false errors even in simple programs. Helgrind and DRD may also
1040 have problems.
1041
1042 Versions 11.0 and earlier may be OK, but this has not been
1043 properly tested.
1044
1045
sewardj9176c862009-08-18 13:47:31 +00001046The following bugs have been fixed or resolved. Note that "n-i-bz"
1047stands for "not in bugzilla" -- that is, a bug that was reported to us
1048but never got a bugzilla entry. We encourage you to file bugs in
1049bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
1050mailing the developers (or mailing lists) directly -- bugs that are
1051not entered into bugzilla tend to get forgotten about or ignored.
sewardj570e2442009-08-14 13:42:41 +00001052
sewardj9176c862009-08-18 13:47:31 +00001053To see details of a given bug, visit
1054https://bugs.kde.org/show_bug.cgi?id=XXXXXX
1055where XXXXXX is the bug number as listed below.
sewardj570e2442009-08-14 13:42:41 +00001056
sewardj9176c862009-08-18 13:47:31 +0000105784303 How about a LockCheck tool?
105891633 dereference of null ptr in vgPlain_st_basetype
105997452 Valgrind doesn't report any pthreads problems
1060100628 leak-check gets assertion failure when using
1061 VALGRIND_MALLOCLIKE_BLOCK on malloc()ed memory
1062108528 NPTL pthread cleanup handlers not called
1063110126 Valgrind 2.4.1 configure.in tramples CFLAGS
1064110128 mallinfo is not implemented...
1065110770 VEX: Generated files not always updated when making valgrind
1066111102 Memcheck: problems with large (memory footprint) applications
1067115673 Vex's decoder should never assert
1068117564 False positive: Syscall param clone(child_tidptr) contains
1069 uninitialised byte(s)
1070119404 executing ssh from inside valgrind fails
1071133679 Callgrind does not write path names to sources with dwarf debug
1072 info
1073135847 configure.in problem with non gnu compilers (and possible fix)
1074136154 threads.c:273 (vgCallgrind_post_signal): Assertion
1075 '*(vgCallgrind_current_fn_stack.top) == 0' failed.
1076136230 memcheck reports "possibly lost", should be "still reachable"
1077137073 NULL arg to MALLOCLIKE_BLOCK causes crash
1078137904 Valgrind reports a memory leak when using POSIX threads,
1079 while it shouldn't
1080139076 valgrind VT_GETSTATE error
1081142228 complaint of elf_dynamic_do_rela in trivial usage
1082145347 spurious warning with USBDEVFS_REAPURB
1083148441 (wine) can't find memory leak in Wine, win32 binary
1084 executable file.
1085148742 Leak-check fails assert on exit
1086149878 add (proper) check for calloc integer overflow
1087150606 Call graph is broken when using callgrind control
1088152393 leak errors produce an exit code of 0. I need some way to
1089 cause leak errors to result in a nonzero exit code.
1090157154 documentation (leak-resolution doc speaks about num-callers
1091 def=4) + what is a loss record
1092159501 incorrect handling of ALSA ioctls
1093162020 Valgrinding an empty/zero-byte file crashes valgrind
1094162482 ppc: Valgrind crashes while reading stabs information
1095162718 x86: avoid segment selector 0 in sys_set_thread_area()
1096163253 (wine) canonicaliseSymtab forgot some fields in DiSym
1097163560 VEX/test_main.c is missing from valgrind-3.3.1
1098164353 malloc_usable_size() doesn't return a usable size
1099165468 Inconsistent formatting in memcheck manual -- please fix
1100169505 main.c:286 (endOfInstr):
1101 Assertion 'ii->cost_offset == *cost_offset' failed
1102177206 Generate default.supp during compile instead of configure
1103177209 Configure valt_load_address based on arch+os
1104177305 eventfd / syscall 323 patch lost
1105179731 Tests fail to build because of inlining of non-local asm labels
1106181394 helgrind: libhb_core.c:3762 (msm_write): Assertion
1107 'ordxx == POrd_EQ || ordxx == POrd_LT' failed.
1108181594 Bogus warning for empty text segment
1109181707 dwarf doesn't require enumerations to have name
1110185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64
1111185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree):
1112 Assertion '!already_present' failed.
1113185359 exp-ptrcheck: unhandled syscall getresuid()
1114185794 "WARNING: unhandled syscall: 285" (fallocate) on x86_64
1115185816 Valgrind is unable to handle debug info for files with split
1116 debug info that are prelinked afterwards
1117185980 [darwin] unhandled syscall: sem_open
1118186238 bbToIR_AMD64: disInstr miscalculated next %rip
1119186507 exp-ptrcheck unhandled syscalls prctl, etc.
1120186790 Suppression pattern used for leaks are not reported
1121186796 Symbols with length>200 in suppression files are ignored
1122187048 drd: mutex PTHREAD_PROCESS_SHARED attribute missinterpretation
1123187416 exp-ptrcheck: support for __NR_{setregid,setreuid,setresuid}
1124188038 helgrind: hg_main.c:926: mk_SHVAL_fail: the 'impossible' happened
1125188046 bashisms in the configure script
1126188127 amd64->IR: unhandled instruction bytes: 0xF0 0xF 0xB0 0xA
1127188161 memcheck: --track-origins=yes asserts "mc_machine.c:672
1128 (get_otrack_shadow_offset_wrk): the 'impossible' happened."
1129188248 helgrind: pthread_cleanup_push, pthread_rwlock_unlock,
1130 assertion fail "!lock->heldBy"
1131188427 Add support for epoll_create1 (with patch)
1132188530 Support for SIOCGSTAMPNS
1133188560 Include valgrind.spec in the tarball
1134188572 Valgrind on Mac should suppress setenv() mem leak
1135189054 Valgrind fails to build because of duplicate non-local asm labels
1136189737 vex amd64->IR: unhandled instruction bytes: 0xAC
1137189762 epoll_create syscall not handled (--tool=exp-ptrcheck)
1138189763 drd assertion failure: s_threadinfo[tid].is_recording
1139190219 unhandled syscall: 328 (x86-linux)
1140190391 dup of 181394; see above
1141190429 Valgrind reports lots of errors in ld.so with x86_64 2.9.90 glibc
1142190820 No debug information on powerpc-linux
sewardj9176c862009-08-18 13:47:31 +00001143191095 PATCH: Improve usbdevfs ioctl handling
1144191182 memcheck: VALGRIND_LEAK_CHECK quadratic when big nr of chunks
1145 or big nr of errors
1146191189 --xml=yes should obey --gen-suppressions=all
1147191192 syslog() needs a suppression on macosx
1148191271 DARWIN: WARNING: unhandled syscall: 33554697 a.k.a.: 265
1149191761 getrlimit on MacOSX
1150191992 multiple --fn-skip only works sometimes; dependent on order
1151192634 V. reports "aspacem sync_check_mapping_callback:
1152 segment mismatch" on Darwin
1153192954 __extension__ missing on 2 client requests
1154194429 Crash at start-up with glibc-2.10.1 and linux-2.6.29
1155194474 "INSTALL" file has different build instructions than "README"
1156194671 Unhandled syscall (sem_wait?) from mac valgrind
1157195069 memcheck: reports leak (memory still reachable) for
1158 printf("%d', x)
1159195169 drd: (vgDrd_barrier_post_wait):
1160 Assertion 'r->sg[p->post_iteration]' failed.
1161195268 valgrind --log-file doesn't accept ~/...
1162195838 VEX abort: LibVEX_N_SPILL_BYTES too small for CPUID boilerplate
1163195860 WARNING: unhandled syscall: unix:223
1164196528 need a error suppression for pthread_rwlock_init under os x?
1165197227 Support aio_* syscalls on Darwin
1166197456 valgrind should reject --suppressions=(directory)
1167197512 DWARF2 CFI reader: unhandled CFI instruction 0:10
1168197591 unhandled syscall 27 (mincore)
1169197793 Merge DCAS branch to the trunk == 85756, 142103
1170197794 Avoid duplicate filenames in Vex
1171197898 make check fails on current SVN
1172197901 make check fails also under exp-ptrcheck in current SVN
1173197929 Make --leak-resolution=high the default
1174197930 Reduce spacing between leak reports
1175197933 Print command line of client at start-up, and shorten preamble
1176197966 unhandled syscall 205 (x86-linux, --tool=exp-ptrcheck)
1177198395 add BBV to the distribution as an experimental tool
1178198624 Missing syscalls on Darwin: 82, 167, 281, 347
1179198649 callgrind_annotate doesn't cumulate counters
1180199338 callgrind_annotate sorting/thresholds are broken for all but Ir
1181199977 Valgrind complains about an unrecognized instruction in the
1182 atomic_incs test program
1183200029 valgrind isn't able to read Fedora 12 debuginfo
1184200760 darwin unhandled syscall: unix:284
1185200827 DRD doesn't work on Mac OS X
1186200990 VG_(read_millisecond_timer)() does not work correctly
1187201016 Valgrind does not support pthread_kill() on Mac OS
1188201169 Document --read-var-info
1189201323 Pre-3.5.0 performance sanity checking
1190201384 Review user manual for the 3.5.0 release
1191201585 mfpvr not implemented on ppc
1192201708 tests failing because x86 direction flag is left set
1193201757 Valgrind doesn't handle any recent sys_futex additions
sewardj6bbe0242009-08-19 11:38:54 +00001194204377 64-bit valgrind can not start a shell script
1195 (with #!/path/to/shell) if the shell is a 32-bit executable
sewardj9176c862009-08-18 13:47:31 +00001196n-i-bz drd: fixed assertion failure triggered by mutex reinitialization.
1197n-i-bz drd: fixed a bug that caused incorrect messages to be printed
1198 about memory allocation events with memory access tracing enabled
1199n-i-bz drd: fixed a memory leak triggered by vector clock deallocation
njn254dafe2009-02-09 03:25:04 +00001200
sewardj48867fb2009-08-19 22:40:53 +00001201(3.5.0: 19 Aug 2009, vex r1913, valgrind r10846).
sewardj570e2442009-08-14 13:42:41 +00001202
njn6eccab92009-08-14 02:41:37 +00001203
njn254dafe2009-02-09 03:25:04 +00001204
njnad91a752009-03-13 19:55:36 +00001205Release 3.4.1 (28 February 2009)
1206~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
12073.4.1 is a bug-fix release that fixes some regressions and assertion
1208failures in debug info reading in 3.4.0, most notably incorrect stack
1209traces on amd64-linux on older (glibc-2.3 based) systems. Various
1210other debug info problems are also fixed. A number of bugs in the
1211exp-ptrcheck tool introduced in 3.4.0 have been fixed.
1212
1213In view of the fact that 3.4.0 contains user-visible regressions
1214relative to 3.3.x, upgrading to 3.4.1 is recommended. Packagers are
1215encouraged to ship 3.4.1 in preference to 3.4.0.
1216
1217The fixed bugs are as follows. Note that "n-i-bz" stands for "not in
1218bugzilla" -- that is, a bug that was reported to us but never got a
1219bugzilla entry. We encourage you to file bugs in bugzilla
1220(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the
1221developers (or mailing lists) directly -- bugs that are not entered
1222into bugzilla tend to get forgotten about or ignored.
1223
1224n-i-bz Fix various bugs reading icc-11 generated debug info
1225n-i-bz Fix various bugs reading gcc-4.4 generated debug info
1226n-i-bz Preliminary support for glibc-2.10 / Fedora 11
1227n-i-bz Cachegrind and Callgrind: handle non-power-of-two cache sizes,
1228 so as to support (eg) 24k Atom D1 and Core2 with 3/6/12MB L2.
1229179618 exp-ptrcheck crashed / exit prematurely
1230179624 helgrind: false positive races with pthread_create and
1231 recv/open/close/read
1232134207 pkg-config output contains @VG_PLATFORM@
1233176926 floating point exception at valgrind startup with PPC 440EPX
1234181594 Bogus warning for empty text segment
1235173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant rex prefixes)
1236181707 Dwarf3 doesn't require enumerations to have name
1237185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64
1238185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree):
1239 Assertion '!already_present' failed.
1240185359 exp-ptrcheck unhandled syscall getresuid()
1241
1242(3.4.1.RC1: 24 Feb 2008, vex r1884, valgrind r9253).
1243(3.4.1: 28 Feb 2008, vex r1884, valgrind r9293).
1244
1245
1246
sewardj565fcf12008-12-23 13:14:06 +00001247Release 3.4.0 (2 January 2009)
1248~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
12493.4.0 is a feature release with many significant improvements and the
1250usual collection of bug fixes. This release supports X86/Linux,
1251AMD64/Linux, PPC32/Linux and PPC64/Linux. Support for recent distros
1252(using gcc 4.4, glibc 2.8 and 2.9) has been added.
sewardj8639f112008-12-17 12:26:24 +00001253
sewardj565fcf12008-12-23 13:14:06 +000012543.4.0 brings some significant tool improvements. Memcheck can now
1255report the origin of uninitialised values, the thread checkers
1256Helgrind and DRD are much improved, and we have a new experimental
1257tool, exp-Ptrcheck, which is able to detect overruns of stack and
1258global arrays. In detail:
sewardj8639f112008-12-17 12:26:24 +00001259
sewardj565fcf12008-12-23 13:14:06 +00001260* Memcheck is now able to track the origin of uninitialised values.
1261 When it reports an uninitialised value error, it will try to show
1262 the origin of the value, as either a heap or stack allocation.
1263 Origin tracking is expensive and so is not enabled by default. To
1264 use it, specify --track-origins=yes. Memcheck's speed will be
1265 essentially halved, and memory usage will be significantly
1266 increased. Nevertheless it can drastically reduce the effort
1267 required to identify the root cause of uninitialised value errors,
1268 and so is often a programmer productivity win, despite running more
1269 slowly.
sewardj8639f112008-12-17 12:26:24 +00001270
sewardj565fcf12008-12-23 13:14:06 +00001271* A version (1.4.0) of the Valkyrie GUI, that works with Memcheck in
sewardjb8b16642009-01-02 23:21:54 +00001272 3.4.0, will be released shortly.
sewardj8639f112008-12-17 12:26:24 +00001273
sewardj565fcf12008-12-23 13:14:06 +00001274* Helgrind's race detection algorithm has been completely redesigned
1275 and reimplemented, to address usability and scalability concerns:
sewardj8639f112008-12-17 12:26:24 +00001276
sewardj565fcf12008-12-23 13:14:06 +00001277 - The new algorithm has a lower false-error rate: it is much less
1278 likely to report races that do not really exist.
sewardj8639f112008-12-17 12:26:24 +00001279
sewardj565fcf12008-12-23 13:14:06 +00001280 - Helgrind will display full call stacks for both accesses involved
sewardjf1fa5ed2008-12-23 17:04:52 +00001281 in a race. This makes it easier to identify the root causes of
1282 races.
sewardj8639f112008-12-17 12:26:24 +00001283
sewardj565fcf12008-12-23 13:14:06 +00001284 - Limitations on the size of program that can run have been removed.
sewardj8639f112008-12-17 12:26:24 +00001285
sewardj565fcf12008-12-23 13:14:06 +00001286 - Performance has been modestly improved, although that is very
1287 workload-dependent.
sewardj8639f112008-12-17 12:26:24 +00001288
sewardj565fcf12008-12-23 13:14:06 +00001289 - Direct support for Qt4 threading has been added.
sewardj8639f112008-12-17 12:26:24 +00001290
sewardj565fcf12008-12-23 13:14:06 +00001291 - pthread_barriers are now directly supported.
njn75010ca2008-02-09 05:10:45 +00001292
sewardj565fcf12008-12-23 13:14:06 +00001293 - Helgrind works well on all supported Linux targets.
sewardj8639f112008-12-17 12:26:24 +00001294
sewardj565fcf12008-12-23 13:14:06 +00001295* The DRD thread debugging tool has seen major improvements:
sewardj8639f112008-12-17 12:26:24 +00001296
sewardj565fcf12008-12-23 13:14:06 +00001297 - Greatly improved performance and significantly reduced memory
1298 usage.
sewardj8639f112008-12-17 12:26:24 +00001299
sewardj565fcf12008-12-23 13:14:06 +00001300 - Support for several major threading libraries (Boost.Thread, Qt4,
1301 glib, OpenMP) has been added.
sewardj8639f112008-12-17 12:26:24 +00001302
sewardj565fcf12008-12-23 13:14:06 +00001303 - Support for atomic instructions, POSIX semaphores, barriers and
1304 reader-writer locks has been added.
1305
1306 - Works now on PowerPC CPUs too.
1307
1308 - Added support for printing thread stack usage at thread exit time.
1309
1310 - Added support for debugging lock contention.
1311
1312 - Added a manual for Drd.
1313
1314* A new experimental tool, exp-Ptrcheck, has been added. Ptrcheck
1315 checks for misuses of pointers. In that sense it is a bit like
1316 Memcheck. However, Ptrcheck can do things Memcheck can't: it can
1317 detect overruns of stack and global arrays, it can detect
1318 arbitrarily far out-of-bounds accesses to heap blocks, and it can
1319 detect accesses heap blocks that have been freed a very long time
1320 ago (millions of blocks in the past).
1321
1322 Ptrcheck currently works only on x86-linux and amd64-linux. To use
1323 it, use --tool=exp-ptrcheck. A simple manual is provided, as part
1324 of the main Valgrind documentation. As this is an experimental
1325 tool, we would be particularly interested in hearing about your
1326 experiences with it.
1327
sewardj974aace2008-12-24 18:40:57 +00001328* exp-Omega, an experimental instantaneous leak-detecting tool, is no
1329 longer built by default, although the code remains in the repository
1330 and the tarball. This is due to three factors: a perceived lack of
1331 users, a lack of maintenance, and concerns that it may not be
1332 possible to achieve reliable operation using the existing design.
sewardj565fcf12008-12-23 13:14:06 +00001333
1334* As usual, support for the latest Linux distros and toolchain
1335 components has been added. It should work well on Fedora Core 10,
1336 OpenSUSE 11.1 and Ubuntu 8.10. gcc-4.4 (in its current pre-release
1337 state) is supported, as is glibc-2.9. The C++ demangler has been
1338 updated so as to work well with C++ compiled by even the most recent
1339 g++'s.
1340
1341* You can now use frame-level wildcards in suppressions. This was a
1342 frequently-requested enhancement. A line "..." in a suppression now
1343 matches zero or more frames. This makes it easier to write
1344 suppressions which are precise yet insensitive to changes in
1345 inlining behaviour.
1346
1347* 3.4.0 adds support on x86/amd64 for the SSSE3 instruction set.
1348
1349* Very basic support for IBM Power6 has been added (64-bit processes only).
1350
1351* Valgrind is now cross-compilable. For example, it is possible to
1352 cross compile Valgrind on an x86/amd64-linux host, so that it runs
1353 on a ppc32/64-linux target.
1354
sewardjf1fa5ed2008-12-23 17:04:52 +00001355* You can set the main thread's stack size at startup using the
1356 new --main-stacksize= flag (subject of course to ulimit settings).
1357 This is useful for running apps that need a lot of stack space.
1358
1359* The limitation that you can't use --trace-children=yes together
1360 with --db-attach=yes has been removed.
1361
sewardj565fcf12008-12-23 13:14:06 +00001362* The following bugs have been fixed. Note that "n-i-bz" stands for
1363 "not in bugzilla" -- that is, a bug that was reported to us but
1364 never got a bugzilla entry. We encourage you to file bugs in
1365 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
1366 mailing the developers (or mailing lists) directly.
1367
sewardjf1fa5ed2008-12-23 17:04:52 +00001368 n-i-bz Make return types for some client requests 64-bit clean
1369 n-i-bz glibc 2.9 support
1370 n-i-bz ignore unsafe .valgrindrc's (CVE-2008-4865)
1371 n-i-bz MPI_Init(0,0) is valid but libmpiwrap.c segfaults
1372 n-i-bz Building in an env without gdb gives bogus gdb attach
1373 92456 Tracing the origin of uninitialised memory
1374 106497 Valgrind does not demangle some C++ template symbols
1375 162222 ==106497
sewardjdf1fc9a2008-12-23 23:09:57 +00001376 151612 Suppression with "..." (frame-level wildcards in .supp files)
sewardjf1fa5ed2008-12-23 17:04:52 +00001377 156404 Unable to start oocalc under memcheck on openSUSE 10.3 (64-bit)
1378 159285 unhandled syscall:25 (stime, on x86-linux)
1379 159452 unhandled ioctl 0x8B01 on "valgrind iwconfig"
1380 160954 ppc build of valgrind crashes with illegal instruction (isel)
1381 160956 mallinfo implementation, w/ patch
1382 162092 Valgrind fails to start gnome-system-monitor
1383 162819 malloc_free_fill test doesn't pass on glibc2.8 x86
1384 163794 assertion failure with "--track-origins=yes"
1385 163933 sigcontext.err and .trapno must be set together
1386 163955 remove constraint !(--db-attach=yes && --trace-children=yes)
1387 164476 Missing kernel module loading system calls
1388 164669 SVN regression: mmap() drops posix file locks
1389 166581 Callgrind output corruption when program forks
1390 167288 Patch file for missing system calls on Cell BE
1391 168943 unsupported scas instruction pentium
1392 171645 Unrecognised instruction (MOVSD, non-binutils encoding)
1393 172417 x86->IR: 0x82 ...
1394 172563 amd64->IR: 0xD9 0xF5 - fprem1
1395 173099 .lds linker script generation error
1396 173177 [x86_64] syscalls: 125/126/179 (capget/capset/quotactl)
1397 173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant prefixes)
1398 174532 == 173751
1399 174908 --log-file value not expanded correctly for core file
1400 175044 Add lookup_dcookie for amd64
1401 175150 x86->IR: 0xF2 0xF 0x11 0xC1 (movss non-binutils encoding)
sewardj565fcf12008-12-23 13:14:06 +00001402
1403Developer-visible changes:
1404
1405* Valgrind's debug-info reading machinery has been majorly overhauled.
1406 It can now correctly establish the addresses for ELF data symbols,
1407 which is something that has never worked properly before now.
1408
1409 Also, Valgrind can now read DWARF3 type and location information for
1410 stack and global variables. This makes it possible to use the
1411 framework to build tools that rely on knowing the type and locations
1412 of stack and global variables, for example exp-Ptrcheck.
1413
1414 Reading of such information is disabled by default, because most
1415 tools don't need it, and because it is expensive in space and time.
1416 However, you can force Valgrind to read it, using the
1417 --read-var-info=yes flag. Memcheck, Helgrind and DRD are able to
1418 make use of such information, if present, to provide source-level
1419 descriptions of data addresses in the error messages they create.
1420
sewardj974aace2008-12-24 18:40:57 +00001421(3.4.0.RC1: 24 Dec 2008, vex r1878, valgrind r8882).
sewardjb8b16642009-01-02 23:21:54 +00001422(3.4.0: 3 Jan 2009, vex r1878, valgrind r8899).
sewardj565fcf12008-12-23 13:14:06 +00001423
sewardj8639f112008-12-17 12:26:24 +00001424
sewardj9f42ba32008-06-04 11:47:51 +00001425
1426Release 3.3.1 (4 June 2008)
1427~~~~~~~~~~~~~~~~~~~~~~~~~~~
14283.3.1 fixes a bunch of bugs in 3.3.0, adds support for glibc-2.8 based
1429systems (openSUSE 11, Fedora Core 9), improves the existing glibc-2.7
1430support, and adds support for the SSSE3 (Core 2) instruction set.
1431
14323.3.1 will likely be the last release that supports some very old
1433systems. In particular, the next major release, 3.4.0, will drop
1434support for the old LinuxThreads threading library, and for gcc
1435versions prior to 3.0.
1436
1437The fixed bugs are as follows. Note that "n-i-bz" stands for "not in
1438bugzilla" -- that is, a bug that was reported to us but never got a
1439bugzilla entry. We encourage you to file bugs in bugzilla
1440(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the
1441developers (or mailing lists) directly -- bugs that are not entered
1442into bugzilla tend to get forgotten about or ignored.
1443
1444n-i-bz Massif segfaults at exit
1445n-i-bz Memcheck asserts on Altivec code
1446n-i-bz fix sizeof bug in Helgrind
1447n-i-bz check fd on sys_llseek
1448n-i-bz update syscall lists to kernel 2.6.23.1
1449n-i-bz support sys_sync_file_range
1450n-i-bz handle sys_sysinfo, sys_getresuid, sys_getresgid on ppc64-linux
1451n-i-bz intercept memcpy in 64-bit ld.so's
1452n-i-bz Fix wrappers for sys_{futimesat,utimensat}
1453n-i-bz Minor false-error avoidance fixes for Memcheck
1454n-i-bz libmpiwrap.c: add a wrapper for MPI_Waitany
1455n-i-bz helgrind support for glibc-2.8
1456n-i-bz partial fix for mc_leakcheck.c:698 assert:
1457 'lc_shadows[i]->data + lc_shadows[i] ...
1458n-i-bz Massif/Cachegrind output corruption when programs fork
1459n-i-bz register allocator fix: handle spill stores correctly
1460n-i-bz add support for PA6T PowerPC CPUs
1461126389 vex x86->IR: 0xF 0xAE (FXRSTOR)
1462158525 ==126389
1463152818 vex x86->IR: 0xF3 0xAC (repz lodsb)
1464153196 vex x86->IR: 0xF2 0xA6 (repnz cmpsb)
1465155011 vex x86->IR: 0xCF (iret)
1466155091 Warning [...] unhandled DW_OP_ opcode 0x23
1467156960 ==155901
1468155528 support Core2/SSSE3 insns on x86/amd64
1469155929 ms_print fails on massif outputs containing long lines
1470157665 valgrind fails on shmdt(0) after shmat to 0
1471157748 support x86 PUSHFW/POPFW
1472158212 helgrind: handle pthread_rwlock_try{rd,wr}lock.
1473158425 sys_poll incorrectly emulated when RES==0
1474158744 vex amd64->IR: 0xF0 0x41 0xF 0xC0 (xaddb)
1475160907 Support for a couple of recent Linux syscalls
1476161285 Patch -- support for eventfd() syscall
1477161378 illegal opcode in debug libm (FUCOMPP)
1478160136 ==161378
1479161487 number of suppressions files is limited to 10
1480162386 ms_print typo in milliseconds time unit for massif
1481161036 exp-drd: client allocated memory was never freed
1482162663 signalfd_wrapper fails on 64bit linux
1483
1484(3.3.1.RC1: 2 June 2008, vex r1854, valgrind r8169).
1485(3.3.1: 4 June 2008, vex r1854, valgrind r8180).
1486
1487
1488
sewardjcb2288b2007-12-02 02:08:17 +00001489Release 3.3.0 (7 December 2007)
1490~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardjcb2288b2007-12-02 02:08:17 +000014913.3.0 is a feature release with many significant improvements and the
1492usual collection of bug fixes. This release supports X86/Linux,
1493AMD64/Linux, PPC32/Linux and PPC64/Linux. Support for recent distros
1494(using gcc 4.3, glibc 2.6 and 2.7) has been added.
njnf4f9cd42007-11-23 01:42:10 +00001495
sewardjcb2288b2007-12-02 02:08:17 +00001496The main excitement in 3.3.0 is new and improved tools. Helgrind
1497works again, Massif has been completely overhauled and much improved,
1498Cachegrind now does branch-misprediction profiling, and a new category
1499of experimental tools has been created, containing two new tools:
1500Omega and DRD. There are many other smaller improvements. In detail:
njn374a36d2007-11-23 01:41:32 +00001501
sewardjcb2288b2007-12-02 02:08:17 +00001502- Helgrind has been completely overhauled and works for the first time
1503 since Valgrind 2.2.0. Supported functionality is: detection of
1504 misuses of the POSIX PThreads API, detection of potential deadlocks
1505 resulting from cyclic lock dependencies, and detection of data
1506 races. Compared to the 2.2.0 Helgrind, the race detection algorithm
1507 has some significant improvements aimed at reducing the false error
1508 rate. Handling of various kinds of corner cases has been improved.
1509 Efforts have been made to make the error messages easier to
1510 understand. Extensive documentation is provided.
njn374a36d2007-11-23 01:41:32 +00001511
sewardjcb2288b2007-12-02 02:08:17 +00001512- Massif has been completely overhauled. Instead of measuring
1513 space-time usage -- which wasn't always useful and many people found
1514 confusing -- it now measures space usage at various points in the
1515 execution, including the point of peak memory allocation. Its
1516 output format has also changed: instead of producing PostScript
1517 graphs and HTML text, it produces a single text output (via the new
1518 'ms_print' script) that contains both a graph and the old textual
1519 information, but in a more compact and readable form. Finally, the
1520 new version should be more reliable than the old one, as it has been
1521 tested more thoroughly.
njn718d3b12006-12-16 00:54:12 +00001522
sewardj8badbaa2007-05-08 09:20:25 +00001523- Cachegrind has been extended to do branch-misprediction profiling.
1524 Both conditional and indirect branches are profiled. The default
1525 behaviour of Cachegrind is unchanged. To use the new functionality,
1526 give the option --branch-sim=yes.
1527
sewardjcb2288b2007-12-02 02:08:17 +00001528- A new category of "experimental tools" has been created. Such tools
1529 may not work as well as the standard tools, but are included because
1530 some people will find them useful, and because exposure to a wider
1531 user group provides tool authors with more end-user feedback. These
1532 tools have a "exp-" prefix attached to their names to indicate their
1533 experimental nature. Currently there are two experimental tools:
njn374a36d2007-11-23 01:41:32 +00001534
sewardjcb2288b2007-12-02 02:08:17 +00001535 * exp-Omega: an instantaneous leak detector. See
1536 exp-omega/docs/omega_introduction.txt.
njn374a36d2007-11-23 01:41:32 +00001537
sewardjcb2288b2007-12-02 02:08:17 +00001538 * exp-DRD: a data race detector based on the happens-before
sewardj8b09d4f2007-12-04 21:27:18 +00001539 relation. See exp-drd/docs/README.txt.
njn374a36d2007-11-23 01:41:32 +00001540
sewardjcb2288b2007-12-02 02:08:17 +00001541- Scalability improvements for very large programs, particularly those
1542 which have a million or more malloc'd blocks in use at once. These
1543 improvements mostly affect Memcheck. Memcheck is also up to 10%
1544 faster for all programs, with x86-linux seeing the largest
1545 improvement.
njn374a36d2007-11-23 01:41:32 +00001546
sewardjcb2288b2007-12-02 02:08:17 +00001547- Works well on the latest Linux distros. Has been tested on Fedora
1548 Core 8 (x86, amd64, ppc32, ppc64) and openSUSE 10.3. glibc 2.6 and
1549 2.7 are supported. gcc-4.3 (in its current pre-release state) is
1550 supported. At the same time, 3.3.0 retains support for older
1551 distros.
njn374a36d2007-11-23 01:41:32 +00001552
sewardjcb2288b2007-12-02 02:08:17 +00001553- The documentation has been modestly reorganised with the aim of
1554 making it easier to find information on common-usage scenarios.
1555 Some advanced material has been moved into a new chapter in the main
1556 manual, so as to unclutter the main flow, and other tidying up has
1557 been done.
njn374a36d2007-11-23 01:41:32 +00001558
sewardjcb2288b2007-12-02 02:08:17 +00001559- There is experimental support for AIX 5.3, both 32-bit and 64-bit
1560 processes. You need to be running a 64-bit kernel to use Valgrind
1561 on a 64-bit executable.
njn374a36d2007-11-23 01:41:32 +00001562
sewardjcb2288b2007-12-02 02:08:17 +00001563- There have been some changes to command line options, which may
1564 affect you:
njn374a36d2007-11-23 01:41:32 +00001565
sewardjcb2288b2007-12-02 02:08:17 +00001566 * --log-file-exactly and
1567 --log-file-qualifier options have been removed.
njn374a36d2007-11-23 01:41:32 +00001568
sewardjcb2288b2007-12-02 02:08:17 +00001569 To make up for this --log-file option has been made more powerful.
1570 It now accepts a %p format specifier, which is replaced with the
1571 process ID, and a %q{FOO} format specifier, which is replaced with
1572 the contents of the environment variable FOO.
1573
1574 * --child-silent-after-fork=yes|no [no]
1575
1576 Causes Valgrind to not show any debugging or logging output for
1577 the child process resulting from a fork() call. This can make the
1578 output less confusing (although more misleading) when dealing with
1579 processes that create children.
1580
1581 * --cachegrind-out-file, --callgrind-out-file and --massif-out-file
1582
1583 These control the names of the output files produced by
1584 Cachegrind, Callgrind and Massif. They accept the same %p and %q
1585 format specifiers that --log-file accepts. --callgrind-out-file
1586 replaces Callgrind's old --base option.
1587
1588 * Cachegrind's 'cg_annotate' script no longer uses the --<pid>
1589 option to specify the output file. Instead, the first non-option
1590 argument is taken to be the name of the output file, and any
1591 subsequent non-option arguments are taken to be the names of
1592 source files to be annotated.
1593
1594 * Cachegrind and Callgrind now use directory names where possible in
1595 their output files. This means that the -I option to
1596 'cg_annotate' and 'callgrind_annotate' should not be needed in
1597 most cases. It also means they can correctly handle the case
1598 where two source files in different directories have the same
1599 name.
1600
1601- Memcheck offers a new suppression kind: "Jump". This is for
1602 suppressing jump-to-invalid-address errors. Previously you had to
1603 use an "Addr1" suppression, which didn't make much sense.
1604
1605- Memcheck has new flags --malloc-fill=<hexnum> and
1606 --free-fill=<hexnum> which free malloc'd / free'd areas with the
1607 specified byte. This can help shake out obscure memory corruption
barte7b86c32009-07-21 07:35:23 +00001608 problems. The definedness and addressability of these areas is
sewardjcb2288b2007-12-02 02:08:17 +00001609 unchanged -- only the contents are affected.
njn718d3b12006-12-16 00:54:12 +00001610
1611- The behaviour of Memcheck's client requests VALGRIND_GET_VBITS and
1612 VALGRIND_SET_VBITS have changed slightly. They no longer issue
1613 addressability errors -- if either array is partially unaddressable,
1614 they just return 3 (as before). Also, SET_VBITS doesn't report
1615 definedness errors if any of the V bits are undefined.
sewardj1ede80f2006-05-26 00:47:13 +00001616
njn92676d72007-12-04 21:18:06 +00001617- The following Memcheck client requests have been removed:
1618 VALGRIND_MAKE_NOACCESS
1619 VALGRIND_MAKE_WRITABLE
1620 VALGRIND_MAKE_READABLE
1621 VALGRIND_CHECK_WRITABLE
1622 VALGRIND_CHECK_READABLE
1623 VALGRIND_CHECK_DEFINED
1624 They were deprecated in 3.2.0, when equivalent but better-named client
1625 requests were added. See the 3.2.0 release notes for more details.
1626
njn5a0a7732007-02-12 10:37:37 +00001627- The behaviour of the tool Lackey has changed slightly. First, the output
1628 from --trace-mem has been made more compact, to reduce the size of the
1629 traces. Second, a new option --trace-superblocks has been added, which
1630 shows the addresses of superblocks (code blocks) as they are executed.
1631
sewardjcb2288b2007-12-02 02:08:17 +00001632- The following bugs have been fixed. Note that "n-i-bz" stands for
1633 "not in bugzilla" -- that is, a bug that was reported to us but
1634 never got a bugzilla entry. We encourage you to file bugs in
1635 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
1636 mailing the developers (or mailing lists) directly.
1637
1638 n-i-bz x86_linux_REDIR_FOR_index() broken
1639 n-i-bz guest-amd64/toIR.c:2512 (dis_op2_E_G): Assertion `0' failed.
1640 n-i-bz Support x86 INT insn (INT (0xCD) 0x40 - 0x43)
1641 n-i-bz Add sys_utimensat system call for Linux x86 platform
1642 79844 Helgrind complains about race condition which does not exist
1643 82871 Massif output function names too short
1644 89061 Massif: ms_main.c:485 (get_XCon): Assertion `xpt->max_chi...'
1645 92615 Write output from Massif at crash
1646 95483 massif feature request: include peak allocation in report
1647 112163 MASSIF crashed with signal 7 (SIGBUS) after running 2 days
1648 119404 problems running setuid executables (partial fix)
1649 121629 add instruction-counting mode for timing
1650 127371 java vm giving unhandled instruction bytes: 0x26 0x2E 0x64 0x65
1651 129937 ==150380
1652 129576 Massif loses track of memory, incorrect graphs
1653 132132 massif --format=html output does not do html entity escaping
1654 132950 Heap alloc/usage summary
1655 133962 unhandled instruction bytes: 0xF2 0x4C 0xF 0x10
1656 134990 use -fno-stack-protector if possible
1657 136382 ==134990
1658 137396 I would really like helgrind to work again...
1659 137714 x86/amd64->IR: 0x66 0xF 0xF7 0xC6 (maskmovq, maskmovdq)
1660 141631 Massif: percentages don't add up correctly
1661 142706 massif numbers don't seem to add up
1662 143062 massif crashes on app exit with signal 8 SIGFPE
1663 144453 (get_XCon): Assertion 'xpt->max_children != 0' failed.
1664 145559 valgrind aborts when malloc_stats is called
1665 145609 valgrind aborts all runs with 'repeated section!'
1666 145622 --db-attach broken again on x86-64
1667 145837 ==149519
1668 145887 PPC32: getitimer() system call is not supported
1669 146252 ==150678
1670 146456 (update_XCon): Assertion 'xpt->curr_space >= -space_delta'...
1671 146701 ==134990
1672 146781 Adding support for private futexes
1673 147325 valgrind internal error on syscall (SYS_io_destroy, 0)
1674 147498 amd64->IR: 0xF0 0xF 0xB0 0xF (lock cmpxchg %cl,(%rdi))
sewardjbc39ad22007-12-09 02:24:02 +00001675 147545 Memcheck: mc_main.c:817 (get_sec_vbits8): Assertion 'n' failed.
sewardjcb2288b2007-12-02 02:08:17 +00001676 147628 SALC opcode 0xd6 unimplemented
1677 147825 crash on amd64-linux with gcc 4.2 and glibc 2.6 (CFI)
1678 148174 Incorrect type of freed_list_volume causes assertion [...]
1679 148447 x86_64 : new NOP codes: 66 66 66 66 2e 0f 1f
1680 149182 PPC Trap instructions not implemented in valgrind
1681 149504 Assertion hit on alloc_xpt->curr_space >= -space_delta
1682 149519 ppc32: V aborts with SIGSEGV on execution of a signal handler
1683 149892 ==137714
1684 150044 SEGV during stack deregister
1685 150380 dwarf/gcc interoperation (dwarf3 read problems)
1686 150408 ==148447
1687 150678 guest-amd64/toIR.c:3741 (dis_Grp5): Assertion `sz == 4' failed
1688 151209 V unable to execute programs for users with UID > 2^16
1689 151938 help on --db-command= misleading
1690 152022 subw $0x28, %%sp causes assertion failure in memcheck
1691 152357 inb and outb not recognized in 64-bit mode
1692 152501 vex x86->IR: 0x27 0x66 0x89 0x45 (daa)
1693 152818 vex x86->IR: 0xF3 0xAC 0xFC 0x9C (rep lodsb)
1694
njnd5f4bef2006-12-24 07:54:19 +00001695Developer-visible changes:
1696
sewardjcb2288b2007-12-02 02:08:17 +00001697- The names of some functions and types within the Vex IR have
1698 changed. Run 'svn log -r1689 VEX/pub/libvex_ir.h' for full details.
1699 Any existing standalone tools will have to be updated to reflect
1700 these changes. The new names should be clearer. The file
1701 VEX/pub/libvex_ir.h is also much better commented.
njnd8366ee2007-02-12 22:38:50 +00001702
sewardjcb2288b2007-12-02 02:08:17 +00001703- A number of new debugging command line options have been added.
1704 These are mostly of use for debugging the symbol table and line
1705 number readers:
1706
1707 --trace-symtab-patt=<patt> limit debuginfo tracing to obj name <patt>
1708 --trace-cfi=no|yes show call-frame-info details? [no]
1709 --debug-dump=syms mimic /usr/bin/readelf --syms
1710 --debug-dump=line mimic /usr/bin/readelf --debug-dump=line
1711 --debug-dump=frames mimic /usr/bin/readelf --debug-dump=frames
1712 --sym-offsets=yes|no show syms in form 'name+offset' ? [no]
1713
1714- Internally, the code base has been further factorised and
1715 abstractified, particularly with respect to support for non-Linux
1716 OSs.
1717
sewardj135a7ff2007-12-10 22:17:42 +00001718(3.3.0.RC1: 2 Dec 2007, vex r1803, valgrind r7268).
1719(3.3.0.RC2: 5 Dec 2007, vex r1804, valgrind r7282).
1720(3.3.0.RC3: 9 Dec 2007, vex r1804, valgrind r7288).
sewardj686b5be2007-12-10 22:21:14 +00001721(3.3.0: 10 Dec 2007, vex r1804, valgrind r7290).
sewardjcb2288b2007-12-02 02:08:17 +00001722
njnd5f4bef2006-12-24 07:54:19 +00001723
1724
njn99237752007-01-30 05:18:38 +00001725Release 3.2.3 (29 Jan 2007)
1726~~~~~~~~~~~~~~~~~~~~~~~~~~~
1727Unfortunately 3.2.2 introduced a regression which can cause an
1728assertion failure ("vex: the `impossible' happened: eqIRConst") when
1729running obscure pieces of SSE code. 3.2.3 fixes this and adds one
1730more glibc-2.5 intercept. In all other respects it is identical to
17313.2.2. Please do not use (or package) 3.2.2; instead use 3.2.3.
1732
1733n-i-bz vex: the `impossible' happened: eqIRConst
1734n-i-bz Add an intercept for glibc-2.5 __stpcpy_chk
1735
1736(3.2.3: 29 Jan 2007, vex r1732, valgrind r6560).
1737
1738
njna7e48802007-01-23 05:58:52 +00001739Release 3.2.2 (22 Jan 2007)
1740~~~~~~~~~~~~~~~~~~~~~~~~~~~
17413.2.2 fixes a bunch of bugs in 3.2.1, adds support for glibc-2.5 based
1742systems (openSUSE 10.2, Fedora Core 6), improves support for icc-9.X
1743compiled code, and brings modest performance improvements in some
1744areas, including amd64 floating point, powerpc support, and startup
1745responsiveness on all targets.
1746
1747The fixed bugs are as follows. Note that "n-i-bz" stands for "not in
1748bugzilla" -- that is, a bug that was reported to us but never got a
1749bugzilla entry. We encourage you to file bugs in bugzilla
1750(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the
1751developers (or mailing lists) directly.
1752
1753129390 ppc?->IR: some kind of VMX prefetch (dstt)
1754129968 amd64->IR: 0xF 0xAE 0x0 (fxsave)
1755134319 ==129968
1756133054 'make install' fails with syntax errors
1757118903 ==133054
1758132998 startup fails in when running on UML
1759134207 pkg-config output contains @VG_PLATFORM@
1760134727 valgrind exits with "Value too large for defined data type"
1761n-i-bz ppc32/64: support mcrfs
1762n-i-bz Cachegrind/Callgrind: Update cache parameter detection
1763135012 x86->IR: 0xD7 0x8A 0xE0 0xD0 (xlat)
1764125959 ==135012
1765126147 x86->IR: 0xF2 0xA5 0xF 0x77 (repne movsw)
1766136650 amd64->IR: 0xC2 0x8 0x0
1767135421 x86->IR: unhandled Grp5(R) case 6
1768n-i-bz Improved documentation of the IR intermediate representation
1769n-i-bz jcxz (x86) (users list, 8 Nov)
1770n-i-bz ExeContext hashing fix
1771n-i-bz fix CFI reading failures ("Dwarf CFI 0:24 0:32 0:48 0:7")
1772n-i-bz fix Cachegrind/Callgrind simulation bug
1773n-i-bz libmpiwrap.c: fix handling of MPI_LONG_DOUBLE
1774n-i-bz make User errors suppressible
1775136844 corrupted malloc line when using --gen-suppressions=yes
1776138507 ==136844
1777n-i-bz Speed up the JIT's register allocator
1778n-i-bz Fix confusing leak-checker flag hints
1779n-i-bz Support recent autoswamp versions
1780n-i-bz ppc32/64 dispatcher speedups
1781n-i-bz ppc64 front end rld/rlw improvements
1782n-i-bz ppc64 back end imm64 improvements
1783136300 support 64K pages on ppc64-linux
1784139124 == 136300
1785n-i-bz fix ppc insn set tests for gcc >= 4.1
1786137493 x86->IR: recent binutils no-ops
1787137714 x86->IR: 0x66 0xF 0xF7 0xC6 (maskmovdqu)
1788138424 "failed in UME with error 22" (produce a better error msg)
1789138856 ==138424
1790138627 Enhancement support for prctl ioctls
1791138896 Add support for usb ioctls
1792136059 ==138896
1793139050 ppc32->IR: mfspr 268/269 instructions not handled
1794n-i-bz ppc32->IR: lvxl/stvxl
1795n-i-bz glibc-2.5 support
1796n-i-bz memcheck: provide replacement for mempcpy
1797n-i-bz memcheck: replace bcmp in ld.so
1798n-i-bz Use 'ifndef' in VEX's Makefile correctly
1799n-i-bz Suppressions for MVL 4.0.1 on ppc32-linux
1800n-i-bz libmpiwrap.c: Fixes for MPICH
1801n-i-bz More robust handling of hinted client mmaps
1802139776 Invalid read in unaligned memcpy with Intel compiler v9
1803n-i-bz Generate valid XML even for very long fn names
1804n-i-bz Don't prompt about suppressions for unshown reachable leaks
1805139910 amd64 rcl is not supported
1806n-i-bz DWARF CFI reader: handle DW_CFA_undefined
1807n-i-bz DWARF CFI reader: handle icc9 generated CFI info better
1808n-i-bz fix false uninit-value errs in icc9 generated FP code
1809n-i-bz reduce extraneous frames in libmpiwrap.c
1810n-i-bz support pselect6 on amd64-linux
1811
1812(3.2.2: 22 Jan 2007, vex r1729, valgrind r6545).
1813
1814
sewardj172bce82006-09-16 01:00:35 +00001815Release 3.2.1 (16 Sept 2006)
1816~~~~~~~~~~~~~~~~~~~~~~~~~~~~
18173.2.1 adds x86/amd64 support for all SSE3 instructions except monitor
1818and mwait, further reduces memcheck's false error rate on all
1819platforms, adds support for recent binutils (in OpenSUSE 10.2 and
1820Fedora Rawhide) and fixes a bunch of bugs in 3.2.0. Some of the fixed
1821bugs were causing large programs to segfault with --tool=callgrind and
1822--tool=cachegrind, so an upgrade is recommended.
1823
1824In view of the fact that any 3.3.0 release is unlikely to happen until
1825well into 1Q07, we intend to keep the 3.2.X line alive for a while
1826yet, and so we tentatively plan a 3.2.2 release sometime in December
182706.
1828
1829The fixed bugs are as follows. Note that "n-i-bz" stands for "not in
1830bugzilla" -- that is, a bug that was reported to us but never got a
1831bugzilla entry.
1832
1833n-i-bz Expanding brk() into last available page asserts
1834n-i-bz ppc64-linux stack RZ fast-case snafu
1835n-i-bz 'c' in --gen-supps=yes doesn't work
1836n-i-bz VG_N_SEGMENTS too low (users, 28 June)
1837n-i-bz VG_N_SEGNAMES too low (Stu Robinson)
1838106852 x86->IR: fisttp (SSE3)
1839117172 FUTEX_WAKE does not use uaddr2
1840124039 Lacks support for VKI_[GP]IO_UNIMAP*
1841127521 amd64->IR: 0xF0 0x48 0xF 0xC7 (cmpxchg8b)
1842128917 amd64->IR: 0x66 0xF 0xF6 0xC4 (psadbw,SSE2)
1843129246 JJ: ppc32/ppc64 syscalls, w/ patch
1844129358 x86->IR: fisttpl (SSE3)
1845129866 cachegrind/callgrind causes executable to die
1846130020 Can't stat .so/.exe error while reading symbols
1847130388 Valgrind aborts when process calls malloc_trim()
1848130638 PATCH: ppc32 missing system calls
1849130785 amd64->IR: unhandled instruction "pushfq"
1850131481: (HINT_NOP) vex x86->IR: 0xF 0x1F 0x0 0xF
1851131298 ==131481
1852132146 Programs with long sequences of bswap[l,q]s
1853132918 vex amd64->IR: 0xD9 0xF8 (fprem)
1854132813 Assertion at priv/guest-x86/toIR.c:652 fails
1855133051 'cfsi->len > 0 && cfsi->len < 2000000' failed
1856132722 valgrind header files are not standard C
1857n-i-bz Livelocks entire machine (users list, Timothy Terriberry)
1858n-i-bz Alex Bennee mmap problem (9 Aug)
1859n-i-bz BartV: Don't print more lines of a stack-trace than were obtained.
1860n-i-bz ppc32 SuSE 10.1 redir
1861n-i-bz amd64 padding suppressions
1862n-i-bz amd64 insn printing fix.
1863n-i-bz ppc cmp reg,reg fix
1864n-i-bz x86/amd64 iropt e/rflag reduction rules
1865n-i-bz SuSE 10.1 (ppc32) minor fixes
1866133678 amd64->IR: 0x48 0xF 0xC5 0xC0 (pextrw?)
1867133694 aspacem assertion: aspacem_minAddr <= holeStart
1868n-i-bz callgrind: fix warning about malformed creator line
1869n-i-bz callgrind: fix annotate script for data produced with
1870 --dump-instr=yes
1871n-i-bz callgrind: fix failed assertion when toggling
1872 instrumentation mode
1873n-i-bz callgrind: fix annotate script fix warnings with
1874 --collect-jumps=yes
1875n-i-bz docs path hardwired (Dennis Lubert)
1876
1877The following bugs were not fixed, due primarily to lack of developer
1878time, and also because bug reporters did not answer requests for
1879feedback in time for the release:
1880
1881129390 ppc?->IR: some kind of VMX prefetch (dstt)
1882129968 amd64->IR: 0xF 0xAE 0x0 (fxsave)
1883133054 'make install' fails with syntax errors
1884n-i-bz Signal race condition (users list, 13 June, Johannes Berg)
1885n-i-bz Unrecognised instruction at address 0x70198EC2 (users list,
1886 19 July, Bennee)
1887132998 startup fails in when running on UML
1888
1889The following bug was tentatively fixed on the mainline but the fix
1890was considered too risky to push into 3.2.X:
1891
1892133154 crash when using client requests to register/deregister stack
1893
1894(3.2.1: 16 Sept 2006, vex r1658, valgrind r6070).
1895
1896
sewardj3ecf8f22006-06-06 21:41:22 +00001897Release 3.2.0 (7 June 2006)
sewardj535a2352006-04-18 02:04:52 +00001898~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardj1ede80f2006-05-26 00:47:13 +000018993.2.0 is a feature release with many significant improvements and the
1900usual collection of bug fixes. This release supports X86/Linux,
1901AMD64/Linux, PPC32/Linux and PPC64/Linux.
njna3e96762006-03-27 12:23:44 +00001902
sewardjaf924992006-05-26 13:51:37 +00001903Performance, especially of Memcheck, is improved, Addrcheck has been
1904removed, Callgrind has been added, PPC64/Linux support has been added,
1905Lackey has been improved, and MPI support has been added. In detail:
njna3e96762006-03-27 12:23:44 +00001906
sewardjf41db3c2006-05-26 16:06:10 +00001907- Memcheck has improved speed and reduced memory use. Run times are
1908 typically reduced by 15-30%, averaging about 24% for SPEC CPU2000.
sewardjaf924992006-05-26 13:51:37 +00001909 The other tools have smaller but noticeable speed improvments. We
1910 are interested to hear what improvements users get.
njna3e96762006-03-27 12:23:44 +00001911
sewardj3ecf8f22006-06-06 21:41:22 +00001912 Memcheck uses less memory due to the introduction of a compressed
sewardjaf924992006-05-26 13:51:37 +00001913 representation for shadow memory. The space overhead has been
1914 reduced by a factor of up to four, depending on program behaviour.
1915 This means you should be able to run programs that use more memory
1916 than before without hitting problems.
njna3e96762006-03-27 12:23:44 +00001917
sewardj1ede80f2006-05-26 00:47:13 +00001918- Addrcheck has been removed. It has not worked since version 2.4.0,
1919 and the speed and memory improvements to Memcheck make it redundant.
1920 If you liked using Addrcheck because it didn't give undefined value
1921 errors, you can use the new Memcheck option --undef-value-errors=no
1922 to get the same behaviour.
1923
sewardjaf924992006-05-26 13:51:37 +00001924- The number of undefined-value errors incorrectly reported by
1925 Memcheck has been reduced (such false reports were already very
1926 rare). In particular, efforts have been made to ensure Memcheck
1927 works really well with gcc 4.0/4.1-generated code on X86/Linux and
1928 AMD64/Linux.
sewardj1ede80f2006-05-26 00:47:13 +00001929
1930- Josef Weidendorfer's popular Callgrind tool has been added. Folding
sewardjaf924992006-05-26 13:51:37 +00001931 it in was a logical step given its popularity and usefulness, and
sewardj1ede80f2006-05-26 00:47:13 +00001932 makes it easier for us to ensure it works "out of the box" on all
1933 supported targets. The associated KDE KCachegrind GUI remains a
1934 separate project.
njneaf0ca92006-04-09 01:23:29 +00001935
sewardj3ecf8f22006-06-06 21:41:22 +00001936- A new release of the Valkyrie GUI for Memcheck, version 1.2.0,
1937 accompanies this release. Improvements over previous releases
1938 include improved robustness, many refinements to the user interface,
1939 and use of a standard autoconf/automake build system. You can get
1940 it from http://www.valgrind.org/downloads/guis.html.
1941
sewardj535a2352006-04-18 02:04:52 +00001942- Valgrind now works on PPC64/Linux. As with the AMD64/Linux port,
sewardj1ede80f2006-05-26 00:47:13 +00001943 this supports programs using to 32G of address space. On 64-bit
1944 capable PPC64/Linux setups, you get a dual architecture build so
1945 that both 32-bit and 64-bit executables can be run. Linux on POWER5
1946 is supported, and POWER4 is also believed to work. Both 32-bit and
1947 64-bit DWARF2 is supported. This port is known to work well with
1948 both gcc-compiled and xlc/xlf-compiled code.
njna3e96762006-03-27 12:23:44 +00001949
sewardj1ede80f2006-05-26 00:47:13 +00001950- Floating point accuracy has been improved for PPC32/Linux.
1951 Specifically, the floating point rounding mode is observed on all FP
1952 arithmetic operations, and multiply-accumulate instructions are
1953 preserved by the compilation pipeline. This means you should get FP
1954 results which are bit-for-bit identical to a native run. These
1955 improvements are also present in the PPC64/Linux port.
njnb2939162006-04-18 00:52:37 +00001956
njneaf0ca92006-04-09 01:23:29 +00001957- Lackey, the example tool, has been improved:
njna3e96762006-03-27 12:23:44 +00001958
sewardj1ede80f2006-05-26 00:47:13 +00001959 * It has a new option --detailed-counts (off by default) which
1960 causes it to print out a count of loads, stores and ALU operations
1961 done, and their sizes.
njndbf7ca72006-03-31 11:57:59 +00001962
sewardj1ede80f2006-05-26 00:47:13 +00001963 * It has a new option --trace-mem (off by default) which causes it
1964 to print out a trace of all memory accesses performed by a
1965 program. It's a good starting point for building Valgrind tools
1966 that need to track memory accesses. Read the comments at the top
1967 of the file lackey/lk_main.c for details.
sewardj42fcf5d2006-04-08 13:16:36 +00001968
sewardj1ede80f2006-05-26 00:47:13 +00001969 * The original instrumentation (counting numbers of instructions,
1970 jumps, etc) is now controlled by a new option --basic-counts. It
1971 is on by default.
sewardj42fcf5d2006-04-08 13:16:36 +00001972
1973- MPI support: partial support for debugging distributed applications
sewardj1ede80f2006-05-26 00:47:13 +00001974 using the MPI library specification has been added. Valgrind is
sewardj42fcf5d2006-04-08 13:16:36 +00001975 aware of the memory state changes caused by a subset of the MPI
1976 functions, and will carefully check data passed to the (P)MPI_
1977 interface.
1978
sewardj1ede80f2006-05-26 00:47:13 +00001979- A new flag, --error-exitcode=, has been added. This allows changing
1980 the exit code in runs where Valgrind reported errors, which is
1981 useful when using Valgrind as part of an automated test suite.
sewardjb9779082006-05-12 23:50:15 +00001982
sewardj1ede80f2006-05-26 00:47:13 +00001983- Various segfaults when reading old-style "stabs" debug information
1984 have been fixed.
njneaf0ca92006-04-09 01:23:29 +00001985
sewardjc9d09cc2006-05-27 03:22:04 +00001986- A simple performance evaluation suite has been added. See
njn5359b6f2006-06-02 23:57:22 +00001987 perf/README and README_DEVELOPERS for details. There are
sewardjc9d09cc2006-05-27 03:22:04 +00001988 various bells and whistles.
1989
sewardjb70a6132006-05-27 21:14:09 +00001990- New configuration flags:
1991 --enable-only32bit
1992 --enable-only64bit
1993 By default, on 64 bit platforms (ppc64-linux, amd64-linux) the build
1994 system will attempt to build a Valgrind which supports both 32-bit
1995 and 64-bit executables. This may not be what you want, and you can
1996 override the default behaviour using these flags.
1997
sewardj1ede80f2006-05-26 00:47:13 +00001998Please note that Helgrind is still not working. We have made an
1999important step towards making it work again, however, with the
2000addition of function wrapping (see below).
njneaf0ca92006-04-09 01:23:29 +00002001
2002Other user-visible changes:
2003
sewardj1ede80f2006-05-26 00:47:13 +00002004- Valgrind now has the ability to intercept and wrap arbitrary
2005 functions. This is a preliminary step towards making Helgrind work
2006 again, and was required for MPI support.
njneaf0ca92006-04-09 01:23:29 +00002007
sewardj1ede80f2006-05-26 00:47:13 +00002008- There are some changes to Memcheck's client requests. Some of them
2009 have changed names:
njndbf7ca72006-03-31 11:57:59 +00002010
2011 MAKE_NOACCESS --> MAKE_MEM_NOACCESS
2012 MAKE_WRITABLE --> MAKE_MEM_UNDEFINED
2013 MAKE_READABLE --> MAKE_MEM_DEFINED
2014
2015 CHECK_WRITABLE --> CHECK_MEM_IS_ADDRESSABLE
2016 CHECK_READABLE --> CHECK_MEM_IS_DEFINED
2017 CHECK_DEFINED --> CHECK_VALUE_IS_DEFINED
2018
sewardj1ede80f2006-05-26 00:47:13 +00002019 The reason for the change is that the old names are subtly
2020 misleading. The old names will still work, but they are deprecated
2021 and may be removed in a future release.
njndbf7ca72006-03-31 11:57:59 +00002022
2023 We also added a new client request:
2024
2025 MAKE_MEM_DEFINED_IF_ADDRESSABLE(a, len)
2026
2027 which is like MAKE_MEM_DEFINED but only affects a byte if the byte is
2028 already addressable.
2029
sewardj3ecf8f22006-06-06 21:41:22 +00002030- The way client requests are encoded in the instruction stream has
2031 changed. Unfortunately, this means 3.2.0 will not honour client
2032 requests compiled into binaries using headers from earlier versions
2033 of Valgrind. We will try to keep the client request encodings more
2034 stable in future.
njneaf0ca92006-04-09 01:23:29 +00002035
njna3e96762006-03-27 12:23:44 +00002036BUGS FIXED:
2037
sewardj1ede80f2006-05-26 00:47:13 +00002038108258 NPTL pthread cleanup handlers not called
2039117290 valgrind is sigKILL'd on startup
2040117295 == 117290
2041118703 m_signals.c:1427 Assertion 'tst->status == VgTs_WaitSys'
2042118466 add %reg, %reg generates incorrect validity for bit 0
2043123210 New: strlen from ld-linux on amd64
2044123244 DWARF2 CFI reader: unhandled CFI instruction 0:18
2045123248 syscalls in glibc-2.4: openat, fstatat, symlinkat
2046123258 socketcall.recvmsg(msg.msg_iov[i] points to uninit
2047123535 mremap(new_addr) requires MREMAP_FIXED in 4th arg
2048123836 small typo in the doc
2049124029 ppc compile failed: `vor' gcc 3.3.5
2050124222 Segfault: @@don't know what type ':' is
2051124475 ppc32: crash (syscall?) timer_settime()
2052124499 amd64->IR: 0xF 0xE 0x48 0x85 (femms)
2053124528 FATAL: aspacem assertion failed: segment_is_sane
2054124697 vex x86->IR: 0xF 0x70 0xC9 0x0 (pshufw)
2055124892 vex x86->IR: 0xF3 0xAE (REPx SCASB)
2056126216 == 124892
2057124808 ppc32: sys_sched_getaffinity() not handled
2058n-i-bz Very long stabs strings crash m_debuginfo
2059n-i-bz amd64->IR: 0x66 0xF 0xF5 (pmaddwd)
2060125492 ppc32: support a bunch more syscalls
2061121617 ppc32/64: coredumping gives assertion failure
2062121814 Coregrind return error as exitcode patch
2063126517 == 121814
sewardj1ede80f2006-05-26 00:47:13 +00002064125607 amd64->IR: 0x66 0xF 0xA3 0x2 (btw etc)
2065125651 amd64->IR: 0xF8 0x49 0xFF 0xE3 (clc?)
2066126253 x86 movx is wrong
2067126451 3.2 SVN doesn't work on ppc32 CPU's without FPU
2068126217 increase # threads
2069126243 vex x86->IR: popw mem
2070126583 amd64->IR: 0x48 0xF 0xA4 0xC2 (shld $1,%rax,%rdx)
sewardjb70a6132006-05-27 21:14:09 +00002071126668 amd64->IR: 0x1C 0xFF (sbb $0xff,%al)
2072126696 support for CDROMREADRAW ioctl and CDROMREADTOCENTRY fix
2073126722 assertion: segment_is_sane at m_aspacemgr/aspacemgr.c:1624
2074126938 bad checking for syscalls linkat, renameat, symlinkat
sewardj1ede80f2006-05-26 00:47:13 +00002075
sewardj3ecf8f22006-06-06 21:41:22 +00002076(3.2.0RC1: 27 May 2006, vex r1626, valgrind r5947).
2077(3.2.0: 7 June 2006, vex r1628, valgrind r5957).
njna3e96762006-03-27 12:23:44 +00002078
2079
sewardj50654672006-03-16 11:33:14 +00002080Release 3.1.1 (15 March 2006)
2081~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
20823.1.1 fixes a bunch of bugs reported in 3.1.0. There is no new
2083functionality. The fixed bugs are:
2084
2085(note: "n-i-bz" means "not in bugzilla" -- this bug does not have
2086 a bugzilla entry).
2087
2088n-i-bz ppc32: fsub 3,3,3 in dispatcher doesn't clear NaNs
2089n-i-bz ppc32: __NR_{set,get}priority
2090117332 x86: missing line info with icc 8.1
2091117366 amd64: 0xDD 0x7C fnstsw
2092118274 == 117366
2093117367 amd64: 0xD9 0xF4 fxtract
2094117369 amd64: __NR_getpriority (140)
2095117419 ppc32: lfsu f5, -4(r11)
2096117419 ppc32: fsqrt
2097117936 more stabs problems (segfaults while reading debug info)
2098119914 == 117936
2099120345 == 117936
2100118239 amd64: 0xF 0xAE 0x3F (clflush)
2101118939 vm86old system call
2102n-i-bz memcheck/tests/mempool reads freed memory
2103n-i-bz AshleyP's custom-allocator assertion
2104n-i-bz Dirk strict-aliasing stuff
2105n-i-bz More space for debugger cmd line (Dan Thaler)
2106n-i-bz Clarified leak checker output message
2107n-i-bz AshleyP's --gen-suppressions output fix
2108n-i-bz cg_annotate's --sort option broken
2109n-i-bz OSet 64-bit fastcmp bug
2110n-i-bz VG_(getgroups) fix (Shinichi Noda)
2111n-i-bz ppc32: allocate from callee-saved FP/VMX regs
2112n-i-bz misaligned path word-size bug in mc_main.c
2113119297 Incorrect error message for sse code
2114120410 x86: prefetchw (0xF 0xD 0x48 0x4)
2115120728 TIOCSERGETLSR, TIOCGICOUNT, HDIO_GET_DMA ioctls
2116120658 Build fixes for gcc 2.96
2117120734 x86: Support for changing EIP in signal handler
2118n-i-bz memcheck/tests/zeropage de-looping fix
2119n-i-bz x86: fxtract doesn't work reliably
2120121662 x86: lock xadd (0xF0 0xF 0xC0 0x2)
2121121893 calloc does not always return zeroed memory
2122121901 no support for syscall tkill
2123n-i-bz Suppression update for Debian unstable
2124122067 amd64: fcmovnu (0xDB 0xD9)
2125n-i-bz ppc32: broken signal handling in cpu feature detection
2126n-i-bz ppc32: rounding mode problems (improved, partial fix only)
2127119482 ppc32: mtfsb1
2128n-i-bz ppc32: mtocrf/mfocrf
2129
2130(3.1.1: 15 March 2006, vex r1597, valgrind r5771).
2131
jseward43230652003-12-14 00:14:54 +00002132
sewardj4bbe9d72005-11-16 18:39:25 +00002133Release 3.1.0 (25 November 2005)
njnf63170c2005-11-12 19:09:51 +00002134~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardj4bbe9d72005-11-16 18:39:25 +000021353.1.0 is a feature release with a number of significant improvements:
2136AMD64 support is much improved, PPC32 support is good enough to be
2137usable, and the handling of memory management and address space is
2138much more robust. In detail:
njnf63170c2005-11-12 19:09:51 +00002139
sewardj4bbe9d72005-11-16 18:39:25 +00002140- AMD64 support is much improved. The 64-bit vs. 32-bit issues in
2141 3.0.X have been resolved, and it should "just work" now in all
2142 cases. On AMD64 machines both 64-bit and 32-bit versions of
2143 Valgrind are built. The right version will be invoked
2144 automatically, even when using --trace-children and mixing execution
2145 between 64-bit and 32-bit executables. Also, many more instructions
2146 are supported.
njnf63170c2005-11-12 19:09:51 +00002147
sewardj4bbe9d72005-11-16 18:39:25 +00002148- PPC32 support is now good enough to be usable. It should work with
2149 all tools, but please let us know if you have problems. Three
2150 classes of CPUs are supported: integer only (no FP, no Altivec),
2151 which covers embedded PPC uses, integer and FP but no Altivec
2152 (G3-ish), and CPUs capable of Altivec too (G4, G5).
njnf63170c2005-11-12 19:09:51 +00002153
sewardj4bbe9d72005-11-16 18:39:25 +00002154- Valgrind's address space management has been overhauled. As a
2155 result, Valgrind should be much more robust with programs that use
2156 large amounts of memory. There should be many fewer "memory
2157 exhausted" messages, and debug symbols should be read correctly on
2158 large (eg. 300MB+) executables. On 32-bit machines the full address
2159 space available to user programs (usually 3GB or 4GB) can be fully
2160 utilised. On 64-bit machines up to 32GB of space is usable; when
2161 using Memcheck that means your program can use up to about 14GB.
njnf63170c2005-11-12 19:09:51 +00002162
2163 A side effect of this change is that Valgrind is no longer protected
2164 against wild writes by the client. This feature was nice but relied
2165 on the x86 segment registers and so wasn't portable.
2166
2167- Most users should not notice, but as part of the address space
sewardj4bbe9d72005-11-16 18:39:25 +00002168 manager change, the way Valgrind is built has been changed. Each
2169 tool is now built as a statically linked stand-alone executable,
2170 rather than as a shared object that is dynamically linked with the
2171 core. The "valgrind" program invokes the appropriate tool depending
2172 on the --tool option. This slightly increases the amount of disk
2173 space used by Valgrind, but it greatly simplified many things and
2174 removed Valgrind's dependence on glibc.
njnf63170c2005-11-12 19:09:51 +00002175
njn336c89d2005-11-20 18:14:04 +00002176Please note that Addrcheck and Helgrind are still not working. Work
2177is underway to reinstate them (or equivalents). We apologise for the
2178inconvenience.
2179
njnf63170c2005-11-12 19:09:51 +00002180Other user-visible changes:
2181
2182- The --weird-hacks option has been renamed --sim-hints.
2183
2184- The --time-stamp option no longer gives an absolute date and time.
2185 It now prints the time elapsed since the program began.
2186
sewardj4bbe9d72005-11-16 18:39:25 +00002187- It should build with gcc-2.96.
2188
njn7cce5b82005-11-16 20:12:22 +00002189- Valgrind can now run itself (see README_DEVELOPERS for how).
sewardj47058612005-11-16 19:13:13 +00002190 This is not much use to you, but it means the developers can now
2191 profile Valgrind using Cachegrind. As a result a couple of
2192 performance bad cases have been fixed.
2193
sewardj6a5a69c2005-11-17 00:51:36 +00002194- The XML output format has changed slightly. See
2195 docs/internals/xml-output.txt.
2196
njn14c9add2005-11-18 17:36:01 +00002197- Core dumping has been reinstated (it was disabled in 3.0.0 and 3.0.1).
2198 If your program crashes while running under Valgrind, a core file with
2199 the name "vgcore.<pid>" will be created (if your settings allow core
2200 file creation). Note that the floating point information is not all
2201 there. If Valgrind itself crashes, the OS will create a normal core
2202 file.
2203
njnf63170c2005-11-12 19:09:51 +00002204The following are some user-visible changes that occurred in earlier
2205versions that may not have been announced, or were announced but not
njn7cce5b82005-11-16 20:12:22 +00002206widely noticed. So we're mentioning them now.
njnf63170c2005-11-12 19:09:51 +00002207
2208- The --tool flag is optional once again; if you omit it, Memcheck
2209 is run by default.
2210
2211- The --num-callers flag now has a default value of 12. It was
2212 previously 4.
2213
njn7cce5b82005-11-16 20:12:22 +00002214- The --xml=yes flag causes Valgrind's output to be produced in XML
2215 format. This is designed to make it easy for other programs to
2216 consume Valgrind's output. The format is described in the file
njnf63170c2005-11-12 19:09:51 +00002217 docs/internals/xml-format.txt.
2218
njn1f503992005-11-13 16:41:15 +00002219- The --gen-suppressions flag supports an "all" value that causes every
2220 suppression to be printed without asking.
2221
njnee88d282005-11-20 17:19:28 +00002222- The --log-file option no longer puts "pid" in the filename, eg. the
2223 old name "foo.pid12345" is now "foo.12345".
2224
njnc66077e2005-11-20 18:02:15 +00002225- There are several graphical front-ends for Valgrind, such as Valkyrie,
2226 Alleyoop and Valgui. See http://www.valgrind.org/downloads/guis.html
2227 for a list.
2228
njnf63170c2005-11-12 19:09:51 +00002229BUGS FIXED:
2230
sewardj4bbe9d72005-11-16 18:39:25 +00002231109861 amd64 hangs at startup
2232110301 ditto
2233111554 valgrind crashes with Cannot allocate memory
2234111809 Memcheck tool doesn't start java
2235111901 cross-platform run of cachegrind fails on opteron
2236113468 (vgPlain_mprotect_range): Assertion 'r != -1' failed.
2237 92071 Reading debugging info uses too much memory
2238109744 memcheck loses track of mmap from direct ld-linux.so.2
2239110183 tail of page with _end
2240 82301 FV memory layout too rigid
2241 98278 Infinite recursion possible when allocating memory
2242108994 Valgrind runs out of memory due to 133x overhead
2243115643 valgrind cannot allocate memory
2244105974 vg_hashtable.c static hash table
2245109323 ppc32: dispatch.S uses Altivec insn, which doesn't work on POWER.
2246109345 ptrace_setregs not yet implemented for ppc
2247110831 Would like to be able to run against both 32 and 64 bit
2248 binaries on AMD64
2249110829 == 110831
2250111781 compile of valgrind-3.0.0 fails on my linux (gcc 2.X prob)
2251112670 Cachegrind: cg_main.c:486 (handleOneStatement ...
2252112941 vex x86: 0xD9 0xF4 (fxtract)
2253110201 == 112941
2254113015 vex amd64->IR: 0xE3 0x14 0x48 0x83 (jrcxz)
2255113126 Crash with binaries built with -gstabs+/-ggdb
2256104065 == 113126
2257115741 == 113126
2258113403 Partial SSE3 support on x86
2259113541 vex: Grp5(x86) (alt encoding inc/dec) case 1
2260113642 valgrind crashes when trying to read debug information
2261113810 vex x86->IR: 66 0F F6 (66 + PSADBW == SSE PSADBW)
2262113796 read() and write() do not work if buffer is in shared memory
2263113851 vex x86->IR: (pmaddwd): 0x66 0xF 0xF5 0xC7
2264114366 vex amd64 cannnot handle __asm__( "fninit" )
2265114412 vex amd64->IR: 0xF 0xAD 0xC2 0xD3 (128-bit shift, shrdq?)
2266114455 vex amd64->IR: 0xF 0xAC 0xD0 0x1 (also shrdq)
2267115590: amd64->IR: 0x67 0xE3 0x9 0xEB (address size override)
2268115953 valgrind svn r5042 does not build with parallel make (-j3)
2269116057 maximum instruction size - VG_MAX_INSTR_SZB too small?
2270116483 shmat failes with invalid argument
2271102202 valgrind crashes when realloc'ing until out of memory
2272109487 == 102202
2273110536 == 102202
2274112687 == 102202
2275111724 vex amd64->IR: 0x41 0xF 0xAB (more BT{,S,R,C} fun n games)
2276111748 vex amd64->IR: 0xDD 0xE2 (fucom)
2277111785 make fails if CC contains spaces
2278111829 vex x86->IR: sbb AL, Ib
2279111851 vex x86->IR: 0x9F 0x89 (lahf/sahf)
2280112031 iopl on AMD64 and README_MISSING_SYSCALL_OR_IOCTL update
2281112152 code generation for Xin_MFence on x86 with SSE0 subarch
2282112167 == 112152
2283112789 == 112152
2284112199 naked ar tool is used in vex makefile
2285112501 vex x86->IR: movq (0xF 0x7F 0xC1 0xF) (mmx MOVQ)
2286113583 == 112501
2287112538 memalign crash
2288113190 Broken links in docs/html/
2289113230 Valgrind sys_pipe on x86-64 wrongly thinks file descriptors
2290 should be 64bit
2291113996 vex amd64->IR: fucomp (0xDD 0xE9)
2292114196 vex x86->IR: out %eax,(%dx) (0xEF 0xC9 0xC3 0x90)
2293114289 Memcheck fails to intercept malloc when used in an uclibc environment
2294114756 mbind syscall support
2295114757 Valgrind dies with assertion: Assertion 'noLargerThan > 0' failed
2296114563 stack tracking module not informed when valgrind switches threads
2297114564 clone() and stacks
2298114565 == 114564
2299115496 glibc crashes trying to use sysinfo page
2300116200 enable fsetxattr, fgetxattr, and fremovexattr for amd64
njnf63170c2005-11-12 19:09:51 +00002301
sewardjae17e7e2005-11-20 19:45:30 +00002302(3.1.0RC1: 20 November 2005, vex r1466, valgrind r5224).
sewardjfa462022005-11-26 13:38:58 +00002303(3.1.0: 26 November 2005, vex r1471, valgrind r5235).
sewardjae17e7e2005-11-20 19:45:30 +00002304
njnf63170c2005-11-12 19:09:51 +00002305
sewardjcd62fe72005-08-29 13:24:51 +00002306Release 3.0.1 (29 August 2005)
2307~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
23083.0.1 fixes a bunch of bugs reported in 3.0.0. There is no new
2309functionality. Some of the fixed bugs are critical, so if you
sewardjfa233892005-08-30 12:35:36 +00002310use/distribute 3.0.0, an upgrade to 3.0.1 is recommended. The fixed
sewardjcd62fe72005-08-29 13:24:51 +00002311bugs are:
2312
2313(note: "n-i-bz" means "not in bugzilla" -- this bug does not have
2314 a bugzilla entry).
2315
2316109313 (== 110505) x86 cmpxchg8b
2317n-i-bz x86: track but ignore changes to %eflags.AC (alignment check)
2318110102 dis_op2_E_G(amd64)
2319110202 x86 sys_waitpid(#286)
2320110203 clock_getres(,0)
2321110208 execve fail wrong retval
2322110274 SSE1 now mandatory for x86
2323110388 amd64 0xDD 0xD1
2324110464 amd64 0xDC 0x1D FCOMP
2325110478 amd64 0xF 0xD PREFETCH
2326n-i-bz XML <unique> printing wrong
2327n-i-bz Dirk r4359 (amd64 syscalls from trunk)
2328110591 amd64 and x86: rdtsc not implemented properly
2329n-i-bz Nick r4384 (stub implementations of Addrcheck and Helgrind)
2330110652 AMD64 valgrind crashes on cwtd instruction
2331110653 AMD64 valgrind crashes on sarb $0x4,foo(%rip) instruction
2332110656 PATH=/usr/bin::/bin valgrind foobar stats ./fooba
2333110657 Small test fixes
2334110671 vex x86->IR: unhandled instruction bytes: 0xF3 0xC3 (rep ret)
2335n-i-bz Nick (Cachegrind should not assert when it encounters a client
2336 request.)
2337110685 amd64->IR: unhandled instruction bytes: 0xE1 0x56 (loope Jb)
2338110830 configuring with --host fails to build 32 bit on 64 bit target
2339110875 Assertion when execve fails
2340n-i-bz Updates to Memcheck manual
2341n-i-bz Fixed broken malloc_usable_size()
2342110898 opteron instructions missing: btq btsq btrq bsfq
2343110954 x86->IR: unhandled instruction bytes: 0xE2 0xF6 (loop Jb)
2344n-i-bz Make suppressions work for "???" lines in stacktraces.
2345111006 bogus warnings from linuxthreads
2346111092 x86: dis_Grp2(Reg): unhandled case(x86)
2347111231 sctp_getladdrs() and sctp_getpaddrs() returns uninitialized
2348 memory
2349111102 (comment #4) Fixed 64-bit unclean "silly arg" message
2350n-i-bz vex x86->IR: unhandled instruction bytes: 0x14 0x0
2351n-i-bz minor umount/fcntl wrapper fixes
2352111090 Internal Error running Massif
2353101204 noisy warning
2354111513 Illegal opcode for SSE instruction (x86 movups)
2355111555 VEX/Makefile: CC is set to gcc
sewardjfa233892005-08-30 12:35:36 +00002356n-i-bz Fix XML bugs in FAQ
sewardjcd62fe72005-08-29 13:24:51 +00002357
sewardjfa233892005-08-30 12:35:36 +00002358(3.0.1: 29 August 05,
2359 vex/branches/VEX_3_0_BRANCH r1367,
2360 valgrind/branches/VALGRIND_3_0_BRANCH r4574).
sewardj820c7722005-08-29 13:44:43 +00002361
2362
sewardjcd62fe72005-08-29 13:24:51 +00002363
sewardjec39f332005-08-02 15:28:07 +00002364Release 3.0.0 (3 August 2005)
2365~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardjf9929e62005-08-02 21:05:58 +000023663.0.0 is a major overhaul of Valgrind. The most significant user
2367visible change is that Valgrind now supports architectures other than
2368x86. The new architectures it supports are AMD64 and PPC32, and the
2369infrastructure is present for other architectures to be added later.
njn4a9f6f42005-07-19 23:46:08 +00002370
sewardjf9929e62005-08-02 21:05:58 +00002371AMD64 support works well, but has some shortcomings:
njn4a9f6f42005-07-19 23:46:08 +00002372
njn7b1a6a42005-07-20 04:20:44 +00002373- It generally won't be as solid as the x86 version. For example,
2374 support for more obscure instructions and system calls may be missing.
2375 We will fix these as they arise.
njn4a9f6f42005-07-19 23:46:08 +00002376
sewardjc371bd02005-07-27 09:55:24 +00002377- Address space may be limited; see the point about
njn7b1a6a42005-07-20 04:20:44 +00002378 position-independent executables below.
njn4a9f6f42005-07-19 23:46:08 +00002379
njn7b1a6a42005-07-20 04:20:44 +00002380- If Valgrind is built on an AMD64 machine, it will only run 64-bit
2381 executables. If you want to run 32-bit x86 executables under Valgrind
2382 on an AMD64, you will need to build Valgrind on an x86 machine and
2383 copy it to the AMD64 machine. And it probably won't work if you do
2384 something tricky like exec'ing a 32-bit program from a 64-bit program
2385 while using --trace-children=yes. We hope to improve this situation
2386 in the future.
2387
2388The PPC32 support is very basic. It may not work reliably even for
sewardjc371bd02005-07-27 09:55:24 +00002389small programs, but it's a start. Many thanks to Paul Mackerras for
2390his great work that enabled this support. We are working to make
2391PPC32 usable as soon as possible.
njn4a9f6f42005-07-19 23:46:08 +00002392
2393Other user-visible changes:
2394
sewardjec39f332005-08-02 15:28:07 +00002395- Valgrind is no longer built by default as a position-independent
2396 executable (PIE), as this caused too many problems.
njn4a9f6f42005-07-19 23:46:08 +00002397
2398 Without PIE enabled, AMD64 programs will only be able to access 2GB of
2399 address space. We will fix this eventually, but not for the moment.
2400
2401 Use --enable-pie at configure-time to turn this on.
2402
2403- Support for programs that use stack-switching has been improved. Use
2404 the --max-stackframe flag for simple cases, and the
2405 VALGRIND_STACK_REGISTER, VALGRIND_STACK_DEREGISTER and
2406 VALGRIND_STACK_CHANGE client requests for trickier cases.
2407
njn7b1a6a42005-07-20 04:20:44 +00002408- Support for programs that use self-modifying code has been improved,
2409 in particular programs that put temporary code fragments on the stack.
2410 This helps for C programs compiled with GCC that use nested functions,
njne7723322005-07-24 17:19:09 +00002411 and also Ada programs. This is controlled with the --smc-check
njn7b1a6a42005-07-20 04:20:44 +00002412 flag, although the default setting should work in most cases.
njn4a9f6f42005-07-19 23:46:08 +00002413
njn7b1a6a42005-07-20 04:20:44 +00002414- Output can now be printed in XML format. This should make it easier
sewardjc371bd02005-07-27 09:55:24 +00002415 for tools such as GUI front-ends and automated error-processing
2416 schemes to use Valgrind output as input. The --xml flag controls this.
2417 As part of this change, ELF directory information is read from executables,
2418 so absolute source file paths are available if needed.
njn4a9f6f42005-07-19 23:46:08 +00002419
2420- Programs that allocate many heap blocks may run faster, due to
2421 improvements in certain data structures.
2422
njn7b1a6a42005-07-20 04:20:44 +00002423- Addrcheck is currently not working. We hope to get it working again
2424 soon. Helgrind is still not working, as was the case for the 2.4.0
2425 release.
njn4a9f6f42005-07-19 23:46:08 +00002426
sewardjc371bd02005-07-27 09:55:24 +00002427- The JITter has been completely rewritten, and is now in a separate
2428 library, called Vex. This enabled a lot of the user-visible changes,
2429 such as new architecture support. The new JIT unfortunately translates
2430 more slowly than the old one, so programs may take longer to start.
2431 We believe the code quality is produces is about the same, so once
2432 started, programs should run at about the same speed. Feedback about
2433 this would be useful.
2434
2435 On the plus side, Vex and hence Memcheck tracks value flow properly
2436 through floating point and vector registers, something the 2.X line
2437 could not do. That means that Memcheck is much more likely to be
2438 usably accurate on vectorised code.
2439
sewardjec39f332005-08-02 15:28:07 +00002440- There is a subtle change to the way exiting of threaded programs
sewardjc371bd02005-07-27 09:55:24 +00002441 is handled. In 3.0, Valgrind's final diagnostic output (leak check,
2442 etc) is not printed until the last thread exits. If the last thread
2443 to exit was not the original thread which started the program, any
2444 other process wait()-ing on this one to exit may conclude it has
2445 finished before the diagnostic output is printed. This may not be
2446 what you expect. 2.X had a different scheme which avoided this
2447 problem, but caused deadlocks under obscure circumstances, so we
2448 are trying something different for 3.0.
2449
2450- Small changes in control log file naming which make it easier to
sewardjec39f332005-08-02 15:28:07 +00002451 use valgrind for debugging MPI-based programs. The relevant
2452 new flags are --log-file-exactly= and --log-file-qualifier=.
sewardjc371bd02005-07-27 09:55:24 +00002453
2454- As part of adding AMD64 support, DWARF2 CFI-based stack unwinding
2455 support was added. In principle this means Valgrind can produce
2456 meaningful backtraces on x86 code compiled with -fomit-frame-pointer
tom81868f02005-07-27 10:33:08 +00002457 providing you also compile your code with -fasynchronous-unwind-tables.
sewardjc371bd02005-07-27 09:55:24 +00002458
sewardjec39f332005-08-02 15:28:07 +00002459- The documentation build system has been completely redone.
2460 The documentation masters are now in XML format, and from that
2461 HTML, PostScript and PDF documentation is generated. As a result
2462 the manual is now available in book form. Note that the
2463 documentation in the source tarballs is pre-built, so you don't need
2464 any XML processing tools to build Valgrind from a tarball.
njn4a9f6f42005-07-19 23:46:08 +00002465
2466Changes that are not user-visible:
2467
sewardjc371bd02005-07-27 09:55:24 +00002468- The code has been massively overhauled in order to modularise it.
2469 As a result we hope it is easier to navigate and understand.
njn4a9f6f42005-07-19 23:46:08 +00002470
2471- Lots of code has been rewritten.
2472
sewardjf9929e62005-08-02 21:05:58 +00002473BUGS FIXED:
njn4a9f6f42005-07-19 23:46:08 +00002474
sewardj79bd9b62005-08-03 16:40:10 +00002475110046 sz == 4 assertion failed
2476109810 vex amd64->IR: unhandled instruction bytes: 0xA3 0x4C 0x70 0xD7
sewardjf9929e62005-08-02 21:05:58 +00002477109802 Add a plausible_stack_size command-line parameter ?
2478109783 unhandled ioctl TIOCMGET (running hw detection tool discover)
2479109780 unhandled ioctl BLKSSZGET (running fdisk -l /dev/hda)
2480109718 vex x86->IR: unhandled instruction: ffreep
2481109429 AMD64 unhandled syscall: 127 (sigpending)
2482109401 false positive uninit in strchr from ld-linux.so.2
2483109385 "stabs" parse failure
2484109378 amd64: unhandled instruction REP NOP
2485109376 amd64: unhandled instruction LOOP Jb
2486109363 AMD64 unhandled instruction bytes
2487109362 AMD64 unhandled syscall: 24 (sched_yield)
2488109358 fork() won't work with valgrind-3.0 SVN
2489109332 amd64 unhandled instruction: ADC Ev, Gv
2490109314 Bogus memcheck report on amd64
2491108883 Crash; vg_memory.c:905 (vgPlain_init_shadow_range):
2492 Assertion `vgPlain_defined_init_shadow_page()' failed.
2493108349 mincore syscall parameter checked incorrectly
2494108059 build infrastructure: small update
2495107524 epoll_ctl event parameter checked on EPOLL_CTL_DEL
2496107123 Vex dies with unhandled instructions: 0xD9 0x31 0xF 0xAE
2497106841 auxmap & openGL problems
2498106713 SDL_Init causes valgrind to exit
2499106352 setcontext and makecontext not handled correctly
2500106293 addresses beyond initial client stack allocation
2501 not checked in VALGRIND_DO_LEAK_CHECK
2502106283 PIE client programs are loaded at address 0
2503105831 Assertion `vgPlain_defined_init_shadow_page()' failed.
2504105039 long run-times probably due to memory manager
2505104797 valgrind needs to be aware of BLKGETSIZE64
2506103594 unhandled instruction: FICOM
2507103320 Valgrind 2.4.0 fails to compile with gcc 3.4.3 and -O0
2508103168 potentially memory leak in coregrind/ume.c
2509102039 bad permissions for mapped region at address 0xB7C73680
2510101881 weird assertion problem
2511101543 Support fadvise64 syscalls
251275247 x86_64/amd64 support (the biggest "bug" we have ever fixed)
njn4a9f6f42005-07-19 23:46:08 +00002513
sewardj9a919512005-08-03 17:01:51 +00002514(3.0RC1: 27 July 05, vex r1303, valgrind r4283).
sewardj1346e132005-08-03 20:42:06 +00002515(3.0.0: 3 August 05, vex r1313, valgrind r4316).
sewardjbfa2b992005-07-27 17:49:17 +00002516
2517
2518
njnccbe3b12009-08-20 00:22:10 +00002519Stable release 2.4.1 (1 August 2005)
2520~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2521(The notes for this release have been lost. Sorry! It would have
2522contained various bug fixes but no new features.)
2523
2524
2525
njn8c927302005-03-12 16:45:01 +00002526Stable release 2.4.0 (March 2005) -- CHANGES RELATIVE TO 2.2.0
njn1aa5ae42005-03-26 16:04:27 +00002527~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
25282.4.0 brings many significant changes and bug fixes. The most
2529significant user-visible change is that we no longer supply our own
2530pthread implementation. Instead, Valgrind is finally capable of
2531running the native thread library, either LinuxThreads or NPTL.
njn8c927302005-03-12 16:45:01 +00002532
njn1aa5ae42005-03-26 16:04:27 +00002533This means our libpthread has gone, along with the bugs associated
2534with it. Valgrind now supports the kernel's threading syscalls, and
2535lets you use your standard system libpthread. As a result:
njn8c927302005-03-12 16:45:01 +00002536
njn1aa5ae42005-03-26 16:04:27 +00002537* There are many fewer system dependencies and strange library-related
2538 bugs. There is a small performance improvement, and a large
2539 stability improvement.
njn8c927302005-03-12 16:45:01 +00002540
njn1aa5ae42005-03-26 16:04:27 +00002541* On the downside, Valgrind can no longer report misuses of the POSIX
2542 PThreads API. It also means that Helgrind currently does not work.
2543 We hope to fix these problems in a future release.
njn8c927302005-03-12 16:45:01 +00002544
njn1aa5ae42005-03-26 16:04:27 +00002545Note that running the native thread libraries does not mean Valgrind
2546is able to provide genuine concurrent execution on SMPs. We still
2547impose the restriction that only one thread is running at any given
2548time.
njn8c927302005-03-12 16:45:01 +00002549
njn1aa5ae42005-03-26 16:04:27 +00002550There are many other significant changes too:
njn8c927302005-03-12 16:45:01 +00002551
njn1aa5ae42005-03-26 16:04:27 +00002552* Memcheck is (once again) the default tool.
njn8c927302005-03-12 16:45:01 +00002553
njn1aa5ae42005-03-26 16:04:27 +00002554* The default stack backtrace is now 12 call frames, rather than 4.
njn8c927302005-03-12 16:45:01 +00002555
njn1aa5ae42005-03-26 16:04:27 +00002556* Suppressions can have up to 25 call frame matches, rather than 4.
njn8c927302005-03-12 16:45:01 +00002557
njn1aa5ae42005-03-26 16:04:27 +00002558* Memcheck and Addrcheck use less memory. Under some circumstances,
2559 they no longer allocate shadow memory if there are large regions of
2560 memory with the same A/V states - such as an mmaped file.
njn8c927302005-03-12 16:45:01 +00002561
njn1aa5ae42005-03-26 16:04:27 +00002562* The memory-leak detector in Memcheck and Addrcheck has been
2563 improved. It now reports more types of memory leak, including
2564 leaked cycles. When reporting leaked memory, it can distinguish
2565 between directly leaked memory (memory with no references), and
2566 indirectly leaked memory (memory only referred to by other leaked
2567 memory).
njn8c927302005-03-12 16:45:01 +00002568
njn1aa5ae42005-03-26 16:04:27 +00002569* Memcheck's confusion over the effect of mprotect() has been fixed:
2570 previously mprotect could erroneously mark undefined data as
2571 defined.
njn8c927302005-03-12 16:45:01 +00002572
njn1aa5ae42005-03-26 16:04:27 +00002573* Signal handling is much improved and should be very close to what
2574 you get when running natively.
2575
2576 One result of this is that Valgrind observes changes to sigcontexts
2577 passed to signal handlers. Such modifications will take effect when
2578 the signal returns. You will need to run with --single-step=yes to
2579 make this useful.
njn8c927302005-03-12 16:45:01 +00002580
2581* Valgrind is built in Position Independent Executable (PIE) format if
njn1aa5ae42005-03-26 16:04:27 +00002582 your toolchain supports it. This allows it to take advantage of all
njn8c927302005-03-12 16:45:01 +00002583 the available address space on systems with 4Gbyte user address
2584 spaces.
2585
2586* Valgrind can now run itself (requires PIE support).
2587
njn1aa5ae42005-03-26 16:04:27 +00002588* Syscall arguments are now checked for validity. Previously all
2589 memory used by syscalls was checked, but now the actual values
2590 passed are also checked.
njn8c927302005-03-12 16:45:01 +00002591
njn1aa5ae42005-03-26 16:04:27 +00002592* Syscall wrappers are more robust against bad addresses being passed
2593 to syscalls: they will fail with EFAULT rather than killing Valgrind
2594 with SIGSEGV.
njn8c927302005-03-12 16:45:01 +00002595
njn1aa5ae42005-03-26 16:04:27 +00002596* Because clone() is directly supported, some non-pthread uses of it
2597 will work. Partial sharing (where some resources are shared, and
2598 some are not) is not supported.
2599
2600* open() and readlink() on /proc/self/exe are supported.
2601
2602BUGS FIXED:
2603
260488520 pipe+fork+dup2 kills the main program
260588604 Valgrind Aborts when using $VALGRIND_OPTS and user progra...
260688614 valgrind: vg_libpthread.c:2323 (read): Assertion `read_pt...
260788703 Stabs parser fails to handle ";"
260888886 ioctl wrappers for TIOCMBIS and TIOCMBIC
260989032 valgrind pthread_cond_timedwait fails
261089106 the 'impossible' happened
261189139 Missing sched_setaffinity & sched_getaffinity
261289198 valgrind lacks support for SIOCSPGRP and SIOCGPGRP
261389263 Missing ioctl translations for scsi-generic and CD playing
261489440 tests/deadlock.c line endings
261589481 `impossible' happened: EXEC FAILED
261689663 valgrind 2.2.0 crash on Redhat 7.2
261789792 Report pthread_mutex_lock() deadlocks instead of returnin...
261890111 statvfs64 gives invalid error/warning
261990128 crash+memory fault with stabs generated by gnat for a run...
262090778 VALGRIND_CHECK_DEFINED() not as documented in memcheck.h
262190834 cachegrind crashes at end of program without reporting re...
262291028 valgrind: vg_memory.c:229 (vgPlain_unmap_range): Assertio...
262391162 valgrind crash while debugging drivel 1.2.1
262491199 Unimplemented function
262591325 Signal routing does not propagate the siginfo structure
262691599 Assertion `cv == ((void *)0)'
262791604 rw_lookup clears orig and sends the NULL value to rw_new
262891821 Small problems building valgrind with $top_builddir ne $t...
262991844 signal 11 (SIGSEGV) at get_tcb (libpthread.c:86) in corec...
263092264 UNIMPLEMENTED FUNCTION: pthread_condattr_setpshared
263192331 per-target flags necessitate AM_PROG_CC_C_O
263292420 valgrind doesn't compile with linux 2.6.8.1/9
263392513 Valgrind 2.2.0 generates some warning messages
263492528 vg_symtab2.c:170 (addLoc): Assertion `loc->size > 0' failed.
263593096 unhandled ioctl 0x4B3A and 0x5601
263693117 Tool and core interface versions do not match
263793128 Can't run valgrind --tool=memcheck because of unimplement...
263893174 Valgrind can crash if passed bad args to certain syscalls
263993309 Stack frame in new thread is badly aligned
264093328 Wrong types used with sys_sigprocmask()
264193763 /usr/include/asm/msr.h is missing
264293776 valgrind: vg_memory.c:508 (vgPlain_find_map_space): Asser...
264393810 fcntl() argument checking a bit too strict
264494378 Assertion `tst->sigqueue_head != tst->sigqueue_tail' failed.
264594429 valgrind 2.2.0 segfault with mmap64 in glibc 2.3.3
264694645 Impossible happened: PINSRW mem
264794953 valgrind: the `impossible' happened: SIGSEGV
264895667 Valgrind does not work with any KDE app
264996243 Assertion 'res==0' failed
265096252 stage2 loader of valgrind fails to allocate memory
265196520 All programs crashing at _dl_start (in /lib/ld-2.3.3.so) ...
265296660 ioctl CDROMREADTOCENTRY causes bogus warnings
265396747 After looping in a segfault handler, the impossible happens
265496923 Zero sized arrays crash valgrind trace back with SIGFPE
265596948 valgrind stops with assertion failure regarding mmap2
265696966 valgrind fails when application opens more than 16 sockets
265797398 valgrind: vg_libpthread.c:2667 Assertion failed
265897407 valgrind: vg_mylibc.c:1226 (vgPlain_safe_fd): Assertion `...
265997427 "Warning: invalid file descriptor -1 in syscall close()" ...
266097785 missing backtrace
266197792 build in obj dir fails - autoconf / makefile cleanup
266297880 pthread_mutex_lock fails from shared library (special ker...
266397975 program aborts without ang VG messages
266498129 Failed when open and close file 230000 times using stdio
266598175 Crashes when using valgrind-2.2.0 with a program using al...
266698288 Massif broken
266798303 UNIMPLEMENTED FUNCTION pthread_condattr_setpshared
266898630 failed--compilation missing warnings.pm, fails to make he...
266998756 Cannot valgrind signal-heavy kdrive X server
267098966 valgrinding the JVM fails with a sanity check assertion
267199035 Valgrind crashes while profiling
267299142 loops with message "Signal 11 being dropped from thread 0...
267399195 threaded apps crash on thread start (using QThread::start...
267499348 Assertion `vgPlain_lseek(core_fd, 0, 1) == phdrs[i].p_off...
267599568 False negative due to mishandling of mprotect
267699738 valgrind memcheck crashes on program that uses sigitimer
267799923 0-sized allocations are reported as leaks
267899949 program seg faults after exit()
2679100036 "newSuperblock's request for 1048576 bytes failed"
2680100116 valgrind: (pthread_cond_init): Assertion `sizeof(* cond) ...
2681100486 memcheck reports "valgrind: the `impossible' happened: V...
2682100833 second call to "mremap" fails with EINVAL
2683101156 (vgPlain_find_map_space): Assertion `(addr & ((1 << 12)-1...
2684101173 Assertion `recDepth >= 0 && recDepth < 500' failed
2685101291 creating threads in a forked process fails
2686101313 valgrind causes different behavior when resizing a window...
2687101423 segfault for c++ array of floats
2688101562 valgrind massif dies on SIGINT even with signal handler r...
2689
njn8c927302005-03-12 16:45:01 +00002690
jseward34ed74f2004-08-30 18:04:42 +00002691Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.0.0
2692~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
jseward7d180d92004-08-30 18:15:37 +000026932.2.0 brings nine months worth of improvements and bug fixes. We
2694believe it to be a worthy successor to 2.0.0. There are literally
2695hundreds of bug fixes and minor improvements. There are also some
2696fairly major user-visible changes:
2697
2698* A complete overhaul of handling of system calls and signals, and
2699 their interaction with threads. In general, the accuracy of the
2700 system call, thread and signal simulations is much improved:
2701
2702 - Blocking system calls behave exactly as they do when running
2703 natively (not on valgrind). That is, if a syscall blocks only the
2704 calling thread when running natively, than it behaves the same on
2705 valgrind. No more mysterious hangs because V doesn't know that some
2706 syscall or other, should block only the calling thread.
2707
2708 - Interrupted syscalls should now give more faithful results.
2709
2710 - Signal contexts in signal handlers are supported.
2711
2712* Improvements to NPTL support to the extent that V now works
2713 properly on NPTL-only setups.
2714
2715* Greater isolation between Valgrind and the program being run, so
2716 the program is less likely to inadvertently kill Valgrind by
2717 doing wild writes.
2718
2719* Massif: a new space profiling tool. Try it! It's cool, and it'll
2720 tell you in detail where and when your C/C++ code is allocating heap.
2721 Draws pretty .ps pictures of memory use against time. A potentially
2722 powerful tool for making sense of your program's space use.
2723
2724* File descriptor leakage checks. When enabled, Valgrind will print out
2725 a list of open file descriptors on exit.
2726
2727* Improved SSE2/SSE3 support.
2728
jseward2eb002b2004-08-31 00:14:02 +00002729* Time-stamped output; use --time-stamp=yes
2730
jseward34ed74f2004-08-30 18:04:42 +00002731
2732
2733Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.1.2
2734~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
27352.2.0 is not much different from 2.1.2, released seven weeks ago.
2736A number of bugs have been fixed, most notably #85658, which gave
2737problems for quite a few people. There have been many internal
2738cleanups, but those are not user visible.
2739
2740The following bugs have been fixed since 2.1.2:
2741
274285658 Assert in coregrind/vg_libpthread.c:2326 (open64) !=
2743 (void*)0 failed
2744 This bug was reported multiple times, and so the following
2745 duplicates of it are also fixed: 87620, 85796, 85935, 86065,
2746 86919, 86988, 87917, 88156
2747
274880716 Semaphore mapping bug caused by unmap (sem_destroy)
2749 (Was fixed prior to 2.1.2)
2750
275186987 semctl and shmctl syscalls family is not handled properly
2752
275386696 valgrind 2.1.2 + RH AS2.1 + librt
2754
275586730 valgrind locks up at end of run with assertion failure
2756 in __pthread_unwind
2757
275886641 memcheck doesn't work with Mesa OpenGL/ATI on Suse 9.1
2759 (also fixes 74298, a duplicate of this)
2760
276185947 MMX/SSE unhandled instruction 'sfence'
2762
276384978 Wrong error "Conditional jump or move depends on
2764 uninitialised value" resulting from "sbbl %reg, %reg"
2765
276686254 ssort() fails when signed int return type from comparison is
2767 too small to handle result of unsigned int subtraction
2768
276987089 memalign( 4, xxx) makes valgrind assert
2770
jseward2eb002b2004-08-31 00:14:02 +0000277186407 Add support for low-level parallel port driver ioctls.
jseward34ed74f2004-08-30 18:04:42 +00002772
277370587 Add timestamps to Valgrind output? (wishlist)
2774
277584937 vg_libpthread.c:2505 (se_remap): Assertion `res == 0'
2776 (fixed prior to 2.1.2)
2777
277886317 cannot load libSDL-1.2.so.0 using valgrind
2779
278086989 memcpy from mac_replace_strmem.c complains about
2781 uninitialized pointers passed when length to copy is zero
2782
278385811 gnu pascal symbol causes segmentation fault; ok in 2.0.0
2784
278579138 writing to sbrk()'d memory causes segfault
2786
278777369 sched deadlock while signal received during pthread_join
2788 and the joined thread exited
2789
279088115 In signal handler for SIGFPE, siginfo->si_addr is wrong
2791 under Valgrind
2792
279378765 Massif crashes on app exit if FP exceptions are enabled
2794
2795Additionally there are the following changes, which are not
2796connected to any bug report numbers, AFAICS:
2797
2798* Fix scary bug causing mis-identification of SSE stores vs
2799 loads and so causing memcheck to sometimes give nonsense results
2800 on SSE code.
2801
2802* Add support for the POSIX message queue system calls.
2803
2804* Fix to allow 32-bit Valgrind to run on AMD64 boxes. Note: this does
2805 NOT allow Valgrind to work with 64-bit executables - only with 32-bit
2806 executables on an AMD64 box.
2807
2808* At configure time, only check whether linux/mii.h can be processed
2809 so that we don't generate ugly warnings by trying to compile it.
2810
jseward2eb002b2004-08-31 00:14:02 +00002811* Add support for POSIX clocks and timers.
2812
jseward34ed74f2004-08-30 18:04:42 +00002813
2814
jseward9cd978c2004-07-17 13:38:12 +00002815Developer (cvs head) release 2.1.2 (18 July 2004)
jsewardbd0a6e42004-07-17 12:19:43 +00002816~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
28172.1.2 contains four months worth of bug fixes and refinements.
jseward9cd978c2004-07-17 13:38:12 +00002818Although officially a developer release, we believe it to be stable
2819enough for widespread day-to-day use. 2.1.2 is pretty good, so try it
2820first, although there is a chance it won't work. If so then try 2.0.0
2821and tell us what went wrong." 2.1.2 fixes a lot of problems present
2822in 2.0.0 and is generally a much better product.
jsewardbd0a6e42004-07-17 12:19:43 +00002823
jseward9cd978c2004-07-17 13:38:12 +00002824Relative to 2.1.1, a large number of minor problems with 2.1.1 have
2825been fixed, and so if you use 2.1.1 you should try 2.1.2. Users of
2826the last stable release, 2.0.0, might also want to try this release.
jsewardbd0a6e42004-07-17 12:19:43 +00002827
2828The following bugs, and probably many more, have been fixed. These
2829are listed at http://bugs.kde.org. Reporting a bug for valgrind in
2830the http://bugs.kde.org is much more likely to get you a fix than
2831mailing developers directly, so please continue to keep sending bugs
2832there.
2833
283476869 Crashes when running any tool under Fedora Core 2 test1
2835 This fixes the problem with returning from a signal handler
jseward0400d4d2004-07-17 23:26:46 +00002836 when VDSOs are turned off in FC2.
jsewardbd0a6e42004-07-17 12:19:43 +00002837
jseward0400d4d2004-07-17 23:26:46 +0000283869508 java 1.4.2 client fails with erroneous "stack size too small".
2839 This fix makes more of the pthread stack attribute related
2840 functions work properly. Java still doesn't work though.
jsewardbd0a6e42004-07-17 12:19:43 +00002841
284271906 malloc alignment should be 8, not 4
2843 All memory returned by malloc/new etc is now at least
2844 8-byte aligned.
2845
jsewarda6017a42004-07-17 23:44:35 +0000284681970 vg_alloc_ThreadState: no free slots available
2847 (closed because the workaround is simple: increase
2848 VG_N_THREADS, rebuild and try again.)
2849
jsewardbd0a6e42004-07-17 12:19:43 +0000285078514 Conditional jump or move depends on uninitialized value(s)
2851 (a slight mishanding of FP code in memcheck)
2852
jsewarda6017a42004-07-17 23:44:35 +0000285377952 pThread Support (crash) (due to initialisation-ordering probs)
2854 (also 85118)
2855
jsewardbd0a6e42004-07-17 12:19:43 +0000285680942 Addrcheck wasn't doing overlap checking as it should.
285778048 return NULL on malloc/new etc failure, instead of asserting
285873655 operator new() override in user .so files often doesn't get picked up
285983060 Valgrind does not handle native kernel AIO
286069872 Create proper coredumps after fatal signals
286182026 failure with new glibc versions: __libc_* functions are not exported
286270344 UNIMPLEMENTED FUNCTION: tcdrain
286381297 Cancellation of pthread_cond_wait does not require mutex
286482872 Using debug info from additional packages (wishlist)
286583025 Support for ioctls FIGETBSZ and FIBMAP
286683340 Support for ioctl HDIO_GET_IDENTITY
286779714 Support for the semtimedop system call.
286877022 Support for ioctls FBIOGET_VSCREENINFO and FBIOGET_FSCREENINFO
286982098 hp2ps ansification (wishlist)
287083573 Valgrind SIGSEGV on execve
287182999 show which cmdline option was erroneous (wishlist)
287283040 make valgrind VPATH and distcheck-clean (wishlist)
jsewarda6017a42004-07-17 23:44:35 +0000287383998 Assertion `newfd > vgPlain_max_fd' failed (see below)
287482722 Unchecked mmap in as_pad leads to mysterious failures later
287578958 memcheck seg faults while running Mozilla
jsewarda2c08002004-07-18 09:24:57 +0000287685416 Arguments with colon (e.g. --logsocket) ignored
jsewardbd0a6e42004-07-17 12:19:43 +00002877
2878
2879Additionally there are the following changes, which are not
2880connected to any bug report numbers, AFAICS:
2881
2882* Rearranged address space layout relative to 2.1.1, so that
2883 Valgrind/tools will run out of memory later than currently in many
2884 circumstances. This is good news esp. for Calltree. It should
2885 be possible for client programs to allocate over 800MB of
2886 memory when using memcheck now.
2887
jseward9cd978c2004-07-17 13:38:12 +00002888* Improved checking when laying out memory. Should hopefully avoid
2889 the random segmentation faults that 2.1.1 sometimes caused.
2890
jsewardbd0a6e42004-07-17 12:19:43 +00002891* Support for Fedora Core 2 and SuSE 9.1. Improvements to NPTL
2892 support to the extent that V now works properly on NPTL-only setups.
2893
2894* Renamed the following options:
2895 --logfile-fd --> --log-fd
2896 --logfile --> --log-file
2897 --logsocket --> --log-socket
2898 to be consistent with each other and other options (esp. --input-fd).
2899
2900* Add support for SIOCGMIIPHY, SIOCGMIIREG and SIOCSMIIREG ioctls and
2901 improve the checking of other interface related ioctls.
2902
jsewardbd0a6e42004-07-17 12:19:43 +00002903* Fix building with gcc-3.4.1.
2904
2905* Remove limit on number of semaphores supported.
2906
jsewardbd0a6e42004-07-17 12:19:43 +00002907* Add support for syscalls: set_tid_address (258), acct (51).
2908
2909* Support instruction "repne movs" -- not official but seems to occur.
2910
2911* Implement an emulated soft limit for file descriptors in addition to
2912 the current reserved area, which effectively acts as a hard limit. The
2913 setrlimit system call now simply updates the emulated limits as best
2914 as possible - the hard limit is not allowed to move at all and just
2915 returns EPERM if you try and change it. This should stop reductions
2916 in the soft limit causing assertions when valgrind tries to allocate
2917 descriptors from the reserved area.
jsewarda6017a42004-07-17 23:44:35 +00002918 (This actually came from bug #83998).
jsewardbd0a6e42004-07-17 12:19:43 +00002919
nethercote76e8fd92004-07-21 10:37:31 +00002920* Major overhaul of Cachegrind implementation. First user-visible change
nethercote5e63bfb2004-07-17 12:49:53 +00002921 is that cachegrind.out files are now typically 90% smaller than they
nethercote76e8fd92004-07-21 10:37:31 +00002922 used to be; code annotation times are correspondingly much smaller.
2923 Second user-visible change is that hit/miss counts for code that is
2924 unloaded at run-time is no longer dumped into a single "discard" pile,
2925 but accurately preserved.
jsewardbd0a6e42004-07-17 12:19:43 +00002926
2927* Client requests for telling valgrind about memory pools.
2928
2929
2930
jseward9cd978c2004-07-17 13:38:12 +00002931Developer (cvs head) release 2.1.1 (12 March 2004)
2932~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
jsewardbd0a6e42004-07-17 12:19:43 +000029332.1.1 contains some internal structural changes needed for V's
jsewarde3e77042004-03-12 21:07:05 +00002934long-term future. These don't affect end-users. Most notable
2935user-visible changes are:
2936
2937* Greater isolation between Valgrind and the program being run, so
2938 the program is less likely to inadvertently kill Valgrind by
2939 doing wild writes.
2940
2941* Massif: a new space profiling tool. Try it! It's cool, and it'll
2942 tell you in detail where and when your C/C++ code is allocating heap.
2943 Draws pretty .ps pictures of memory use against time. A potentially
2944 powerful tool for making sense of your program's space use.
2945
2946* Fixes for many bugs, including support for more SSE2/SSE3 instructions,
2947 various signal/syscall things, and various problems with debug
2948 info readers.
2949
2950* Support for glibc-2.3.3 based systems.
2951
2952We are now doing automatic overnight build-and-test runs on a variety
2953of distros. As a result, we believe 2.1.1 builds and runs on:
2954Red Hat 7.2, 7.3, 8.0, 9, Fedora Core 1, SuSE 8.2, SuSE 9.
2955
2956
2957The following bugs, and probably many more, have been fixed. These
2958are listed at http://bugs.kde.org. Reporting a bug for valgrind in
2959the http://bugs.kde.org is much more likely to get you a fix than
2960mailing developers directly, so please continue to keep sending bugs
2961there.
2962
296369616 glibc 2.3.2 w/NPTL is massively different than what valgrind expects
296469856 I don't know how to instrument MMXish stuff (Helgrind)
296573892 valgrind segfaults starting with Objective-C debug info
2966 (fix for S-type stabs)
296773145 Valgrind complains too much about close(<reserved fd>)
296873902 Shadow memory allocation seems to fail on RedHat 8.0
296968633 VG_N_SEMAPHORES too low (V itself was leaking semaphores)
297075099 impossible to trace multiprocess programs
297176839 the `impossible' happened: disInstr: INT but not 0x80 !
297276762 vg_to_ucode.c:3748 (dis_push_segreg): Assertion `sz == 4' failed.
297376747 cannot include valgrind.h in c++ program
297476223 parsing B(3,10) gave NULL type => impossible happens
297575604 shmdt handling problem
297676416 Problems with gcc 3.4 snap 20040225
297775614 using -gstabs when building your programs the `impossible' happened
297875787 Patch for some CDROM ioctls CDORM_GET_MCN, CDROM_SEND_PACKET,
297975294 gcc 3.4 snapshot's libstdc++ have unsupported instructions.
2980 (REP RET)
298173326 vg_symtab2.c:272 (addScopeRange): Assertion `range->size > 0' failed.
298272596 not recognizing __libc_malloc
298369489 Would like to attach ddd to running program
298472781 Cachegrind crashes with kde programs
298573055 Illegal operand at DXTCV11CompressBlockSSE2 (more SSE opcodes)
298673026 Descriptor leak check reports port numbers wrongly
298771705 README_MISSING_SYSCALL_OR_IOCTL out of date
298872643 Improve support for SSE/SSE2 instructions
298972484 valgrind leaves it's own signal mask in place when execing
299072650 Signal Handling always seems to restart system calls
299172006 The mmap system call turns all errors in ENOMEM
299271781 gdb attach is pretty useless
299371180 unhandled instruction bytes: 0xF 0xAE 0x85 0xE8
299469886 writes to zero page cause valgrind to assert on exit
299571791 crash when valgrinding gimp 1.3 (stabs reader problem)
299669783 unhandled syscall: 218
299769782 unhandled instruction bytes: 0x66 0xF 0x2B 0x80
299870385 valgrind fails if the soft file descriptor limit is less
2999 than about 828
300069529 "rep; nop" should do a yield
300170827 programs with lots of shared libraries report "mmap failed"
3002 for some of them when reading symbols
300371028 glibc's strnlen is optimised enough to confuse valgrind
3004
3005
3006
3007
jseward43230652003-12-14 00:14:54 +00003008Unstable (cvs head) release 2.1.0 (15 December 2003)
3009~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3010For whatever it's worth, 2.1.0 actually seems pretty darn stable to me
3011(Julian). It looks eminently usable, and given that it fixes some
3012significant bugs, may well be worth using on a day-to-day basis.
30132.1.0 is known to build and pass regression tests on: SuSE 9, SuSE
30148.2, RedHat 8.
3015
30162.1.0 most notably includes Jeremy Fitzhardinge's complete overhaul of
3017handling of system calls and signals, and their interaction with
3018threads. In general, the accuracy of the system call, thread and
3019signal simulations is much improved. Specifically:
3020
3021- Blocking system calls behave exactly as they do when running
3022 natively (not on valgrind). That is, if a syscall blocks only the
3023 calling thread when running natively, than it behaves the same on
3024 valgrind. No more mysterious hangs because V doesn't know that some
3025 syscall or other, should block only the calling thread.
3026
3027- Interrupted syscalls should now give more faithful results.
3028
3029- Finally, signal contexts in signal handlers are supported. As a
3030 result, konqueror on SuSE 9 no longer segfaults when notified of
3031 file changes in directories it is watching.
3032
3033Other changes:
3034
3035- Robert Walsh's file descriptor leakage checks. When enabled,
3036 Valgrind will print out a list of open file descriptors on
3037 exit. Along with each file descriptor, Valgrind prints out a stack
3038 backtrace of where the file was opened and any details relating to the
3039 file descriptor such as the file name or socket details.
3040 To use, give: --track-fds=yes
3041
3042- Implemented a few more SSE/SSE2 instructions.
3043
3044- Less crud on the stack when you do 'where' inside a GDB attach.
3045
3046- Fixed the following bugs:
3047 68360: Valgrind does not compile against 2.6.0-testX kernels
3048 68525: CVS head doesn't compile on C90 compilers
3049 68566: pkgconfig support (wishlist)
3050 68588: Assertion `sz == 4' failed in vg_to_ucode.c (disInstr)
3051 69140: valgrind not able to explicitly specify a path to a binary.
3052 69432: helgrind asserts encountering a MutexErr when there are
3053 EraserErr suppressions
3054
jseward0b12daf2003-12-14 14:35:28 +00003055- Increase the max size of the translation cache from 200k average bbs
3056 to 300k average bbs. Programs on the size of OOo (680m17) are
3057 thrashing the cache at the smaller size, creating large numbers of
3058 retranslations and wasting significant time as a result.
3059
jseward43230652003-12-14 00:14:54 +00003060
3061
3062Stable release 2.0.0 (5 Nov 2003)
3063~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3064
30652.0.0 improves SSE/SSE2 support, fixes some minor bugs, and
3066improves support for SuSE 9 and the Red Hat "Severn" beta.
3067
3068- Further improvements to SSE/SSE2 support. The entire test suite of
3069 the GNU Scientific Library (gsl-1.4) compiled with Intel Icc 7.1
3070 20030307Z '-g -O -xW' now works. I think this gives pretty good
3071 coverage of SSE/SSE2 floating point instructions, or at least the
3072 subset emitted by Icc.
3073
3074- Also added support for the following instructions:
3075 MOVNTDQ UCOMISD UNPCKLPS UNPCKHPS SQRTSS
3076 PUSH/POP %{FS,GS}, and PUSH %CS (Nb: there is no POP %CS).
3077
3078- CFI support for GDB version 6. Needed to enable newer GDBs
3079 to figure out where they are when using --gdb-attach=yes.
3080
3081- Fix this:
3082 mc_translate.c:1091 (memcheck_instrument): Assertion
3083 `u_in->size == 4 || u_in->size == 16' failed.
3084
3085- Return an error rather than panicing when given a bad socketcall.
3086
3087- Fix checking of syscall rt_sigtimedwait().
3088
3089- Implement __NR_clock_gettime (syscall 265). Needed on Red Hat Severn.
3090
3091- Fixed bug in overlap check in strncpy() -- it was assuming the src was 'n'
3092 bytes long, when it could be shorter, which could cause false
3093 positives.
3094
3095- Support use of select() for very large numbers of file descriptors.
3096
3097- Don't fail silently if the executable is statically linked, or is
3098 setuid/setgid. Print an error message instead.
3099
3100- Support for old DWARF-1 format line number info.
3101
3102
3103
3104Snapshot 20031012 (12 October 2003)
3105~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3106
3107Three months worth of bug fixes, roughly. Most significant single
3108change is improved SSE/SSE2 support, mostly thanks to Dirk Mueller.
3109
311020031012 builds on Red Hat Fedora ("Severn") but doesn't really work
3111(curiosly, mozilla runs OK, but a modest "ls -l" bombs). I hope to
3112get a working version out soon. It may or may not work ok on the
3113forthcoming SuSE 9; I hear positive noises about it but haven't been
3114able to verify this myself (not until I get hold of a copy of 9).
3115
3116A detailed list of changes, in no particular order:
3117
3118- Describe --gen-suppressions in the FAQ.
3119
3120- Syscall __NR_waitpid supported.
3121
3122- Minor MMX bug fix.
3123
3124- -v prints program's argv[] at startup.
3125
3126- More glibc-2.3 suppressions.
3127
3128- Suppressions for stack underrun bug(s) in the c++ support library
3129 distributed with Intel Icc 7.0.
3130
3131- Fix problems reading /proc/self/maps.
3132
3133- Fix a couple of messages that should have been suppressed by -q,
3134 but weren't.
3135
3136- Make Addrcheck understand "Overlap" suppressions.
3137
3138- At startup, check if program is statically linked and bail out if so.
3139
3140- Cachegrind: Auto-detect Intel Pentium-M, also VIA Nehemiah
3141
3142- Memcheck/addrcheck: minor speed optimisations
3143
3144- Handle syscall __NR_brk more correctly than before.
3145
3146- Fixed incorrect allocate/free mismatch errors when using
3147 operator new(unsigned, std::nothrow_t const&)
3148 operator new[](unsigned, std::nothrow_t const&)
3149
3150- Support POSIX pthread spinlocks.
3151
3152- Fixups for clean compilation with gcc-3.3.1.
daywalker5d945de2003-09-26 00:32:53 +00003153
3154- Implemented more opcodes:
3155 - push %es
3156 - push %ds
3157 - pop %es
3158 - pop %ds
3159 - movntq
3160 - sfence
3161 - pshufw
3162 - pavgb
3163 - ucomiss
3164 - enter
daywalkerb18d2532003-09-27 20:15:01 +00003165 - mov imm32, %esp
3166 - all "in" and "out" opcodes
daywalker79aad842003-09-30 22:58:12 +00003167 - inc/dec %esp
jseward43230652003-12-14 00:14:54 +00003168 - A whole bunch of SSE/SSE2 instructions
daywalker5d945de2003-09-26 00:32:53 +00003169
jseward43230652003-12-14 00:14:54 +00003170- Memcheck: don't bomb on SSE/SSE2 code.
daywalker5d945de2003-09-26 00:32:53 +00003171
sewardj90238792003-05-05 00:23:42 +00003172
sewardj945f39f2003-07-25 21:11:40 +00003173Snapshot 20030725 (25 July 2003)
3174~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3175
3176Fixes some minor problems in 20030716.
3177
3178- Fix bugs in overlap checking for strcpy/memcpy etc.
3179
3180- Do overlap checking with Addrcheck as well as Memcheck.
3181
3182- Fix this:
3183 Memcheck: the `impossible' happened:
3184 get_error_name: unexpected type
3185
3186- Install headers needed to compile new skins.
3187
sewardje3dd2e02003-07-27 08:04:48 +00003188- Remove leading spaces and colon in the LD_LIBRARY_PATH / LD_PRELOAD
sewardj945f39f2003-07-25 21:11:40 +00003189 passed to non-traced children.
3190
sewardjb9eda952003-07-26 21:39:05 +00003191- Fix file descriptor leak in valgrind-listener.
3192
sewardje3dd2e02003-07-27 08:04:48 +00003193- Fix longstanding bug in which the allocation point of a
3194 block resized by realloc was not correctly set. This may
3195 have caused confusing error messages.
sewardj945f39f2003-07-25 21:11:40 +00003196
3197
sewardj626fd892003-07-16 20:10:26 +00003198Snapshot 20030716 (16 July 2003)
sewardj9d916ed2003-07-14 23:38:40 +00003199~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3200
sewardj626fd892003-07-16 20:10:26 +0000320120030716 is a snapshot of our current CVS head (development) branch.
sewardj9d916ed2003-07-14 23:38:40 +00003202This is the branch which will become valgrind-2.0. It contains
3203significant enhancements over the 1.9.X branch.
3204
3205Despite this being a snapshot of the CVS head, it is believed to be
3206quite stable -- at least as stable as 1.9.6 or 1.0.4, if not more so
3207-- and therefore suitable for widespread use. Please let us know asap
3208if it causes problems for you.
3209
3210Two reasons for releasing a snapshot now are:
3211
3212- It's been a while since 1.9.6, and this snapshot fixes
3213 various problems that 1.9.6 has with threaded programs
3214 on glibc-2.3.X based systems.
3215
3216- So as to make available improvements in the 2.0 line.
3217
sewardj626fd892003-07-16 20:10:26 +00003218Major changes in 20030716, as compared to 1.9.6:
sewardj9d916ed2003-07-14 23:38:40 +00003219
3220- More fixes to threading support on glibc-2.3.1 and 2.3.2-based
3221 systems (SuSE 8.2, Red Hat 9). If you have had problems
3222 with inconsistent/illogical behaviour of errno, h_errno or the DNS
sewardj626fd892003-07-16 20:10:26 +00003223 resolver functions in threaded programs, 20030716 should improve
sewardj9d916ed2003-07-14 23:38:40 +00003224 matters. This snapshot seems stable enough to run OpenOffice.org
3225 1.1rc on Red Hat 7.3, SuSE 8.2 and Red Hat 9, and that's a big
3226 threaded app if ever I saw one.
3227
3228- Automatic generation of suppression records; you no longer
3229 need to write them by hand. Use --gen-suppressions=yes.
3230
sewardj21511802003-07-22 17:47:42 +00003231- strcpy/memcpy/etc check their arguments for overlaps, when
3232 running with the Memcheck or Addrcheck skins.
3233
3234- malloc_usable_size() is now supported.
3235
3236- new client requests:
3237 - VALGRIND_COUNT_ERRORS, VALGRIND_COUNT_LEAKS:
3238 useful with regression testing
3239 - VALGRIND_NON_SIMD_CALL[0123]: for running arbitrary functions
3240 on real CPU (use with caution!)
3241
sewardj9d916ed2003-07-14 23:38:40 +00003242- The GDB attach mechanism is more flexible. Allow the GDB to
3243 be run to be specified by --gdb-path=/path/to/gdb, and specify
3244 which file descriptor V will read its input from with
3245 --input-fd=<number>.
3246
sewardj21511802003-07-22 17:47:42 +00003247- Cachegrind gives more accurate results (wasn't tracking instructions in
3248 malloc() and friends previously, is now).
3249
sewardj9d916ed2003-07-14 23:38:40 +00003250- Complete support for the MMX instruction set.
3251
3252- Partial support for the SSE and SSE2 instruction sets. Work for this
3253 is ongoing. About half the SSE/SSE2 instructions are done, so
3254 some SSE based programs may work. Currently you need to specify
3255 --skin=addrcheck. Basically not suitable for real use yet.
3256
3257- Significant speedups (10%-20%) for standard memory checking.
3258
3259- Fix assertion failure in pthread_once().
3260
3261- Fix this:
3262 valgrind: vg_intercept.c:598 (vgAllRoadsLeadToRome_select):
3263 Assertion `ms_end >= ms_now' failed.
3264
3265- Implement pthread_mutexattr_setpshared.
3266
3267- Understand Pentium 4 branch hints. Also implemented a couple more
3268 obscure x86 instructions.
3269
3270- Lots of other minor bug fixes.
3271
sewardj626fd892003-07-16 20:10:26 +00003272- We have a decent regression test system, for the first time.
3273 This doesn't help you directly, but it does make it a lot easier
3274 for us to track the quality of the system, especially across
3275 multiple linux distributions.
3276
3277 You can run the regression tests with 'make regtest' after 'make
3278 install' completes. On SuSE 8.2 and Red Hat 9 I get this:
3279
3280 == 84 tests, 0 stderr failures, 0 stdout failures ==
3281
3282 On Red Hat 8, I get this:
3283
3284 == 84 tests, 2 stderr failures, 1 stdout failure ==
3285 corecheck/tests/res_search (stdout)
3286 memcheck/tests/sigaltstack (stderr)
3287
3288 sigaltstack is probably harmless. res_search doesn't work
3289 on R H 8 even running natively, so I'm not too worried.
3290
3291 On Red Hat 7.3, a glibc-2.2.5 system, I get these harmless failures:
3292
3293 == 84 tests, 2 stderr failures, 1 stdout failure ==
3294 corecheck/tests/pth_atfork1 (stdout)
3295 corecheck/tests/pth_atfork1 (stderr)
3296 memcheck/tests/sigaltstack (stderr)
3297
3298 You need to run on a PII system, at least, since some tests
3299 contain P6-specific instructions, and the test machine needs
3300 access to the internet so that corecheck/tests/res_search
3301 (a test that the DNS resolver works) can function.
3302
sewardj9d916ed2003-07-14 23:38:40 +00003303As ever, thanks for the vast amount of feedback :) and bug reports :(
3304We may not answer all messages, but we do at least look at all of
3305them, and tend to fix the most frequently reported bugs.
3306
3307
3308
sewardj37918822003-05-05 01:05:09 +00003309Version 1.9.6 (7 May 2003 or thereabouts)
3310~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3311
3312Major changes in 1.9.6:
3313
3314- Improved threading support for glibc >= 2.3.2 (SuSE 8.2,
3315 RedHat 9, to name but two ...) It turned out that 1.9.5
3316 had problems with threading support on glibc >= 2.3.2,
3317 usually manifested by threaded programs deadlocking in system calls,
3318 or running unbelievably slowly. Hopefully these are fixed now. 1.9.6
3319 is the first valgrind which gives reasonable support for
3320 glibc-2.3.2. Also fixed a 2.3.2 problem with pthread_atfork().
3321
3322- Majorly expanded FAQ.txt. We've added workarounds for all
3323 common problems for which a workaround is known.
3324
3325Minor changes in 1.9.6:
3326
3327- Fix identification of the main thread's stack. Incorrect
3328 identification of it was causing some on-stack addresses to not get
3329 identified as such. This only affected the usefulness of some error
3330 messages; the correctness of the checks made is unchanged.
3331
3332- Support for kernels >= 2.5.68.
3333
3334- Dummy implementations of __libc_current_sigrtmin,
3335 __libc_current_sigrtmax and __libc_allocate_rtsig, hopefully
3336 good enough to keep alive programs which previously died for lack of
3337 them.
3338
3339- Fix bug in the VALGRIND_DISCARD_TRANSLATIONS client request.
3340
sewardj3d47b792003-05-05 22:15:35 +00003341- Fix bug in the DWARF2 debug line info loader, when instructions
3342 following each other have source lines far from each other
3343 (e.g. with inlined functions).
3344
sewardj37918822003-05-05 01:05:09 +00003345- Debug info reading: read symbols from both "symtab" and "dynsym"
3346 sections, rather than merely from the one that comes last in the
3347 file.
3348
3349- New syscall support: prctl(), creat(), lookup_dcookie().
3350
3351- When checking calls to accept(), recvfrom(), getsocketopt(),
3352 don't complain if buffer values are NULL.
3353
3354- Try and avoid assertion failures in
3355 mash_LD_PRELOAD_and_LD_LIBRARY_PATH.
3356
3357- Minor bug fixes in cg_annotate.
3358
3359
3360
sewardj90238792003-05-05 00:23:42 +00003361Version 1.9.5 (7 April 2003)
3362~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3363
3364It occurs to me that it would be helpful for valgrind users to record
3365in the source distribution the changes in each release. So I now
3366attempt to mend my errant ways :-) Changes in this and future releases
3367will be documented in the NEWS file in the source distribution.
3368
3369Major changes in 1.9.5:
3370
3371- (Critical bug fix): Fix a bug in the FPU simulation. This was
3372 causing some floating point conditional tests not to work right.
3373 Several people reported this. If you had floating point code which
3374 didn't work right on 1.9.1 to 1.9.4, it's worth trying 1.9.5.
3375
3376- Partial support for Red Hat 9. RH9 uses the new Native Posix
3377 Threads Library (NPTL), instead of the older LinuxThreads.
3378 This potentially causes problems with V which will take some
3379 time to correct. In the meantime we have partially worked around
3380 this, and so 1.9.5 works on RH9. Threaded programs still work,
3381 but they may deadlock, because some system calls (accept, read,
3382 write, etc) which should be nonblocking, in fact do block. This
3383 is a known bug which we are looking into.
3384
3385 If you can, your best bet (unfortunately) is to avoid using
3386 1.9.5 on a Red Hat 9 system, or on any NPTL-based distribution.
3387 If your glibc is 2.3.1 or earlier, you're almost certainly OK.
3388
3389Minor changes in 1.9.5:
3390
3391- Added some #errors to valgrind.h to ensure people don't include
3392 it accidentally in their sources. This is a change from 1.0.X
3393 which was never properly documented. The right thing to include
3394 is now memcheck.h. Some people reported problems and strange
3395 behaviour when (incorrectly) including valgrind.h in code with
3396 1.9.1 -- 1.9.4. This is no longer possible.
3397
3398- Add some __extension__ bits and pieces so that gcc configured
3399 for valgrind-checking compiles even with -Werror. If you
3400 don't understand this, ignore it. Of interest to gcc developers
3401 only.
3402
3403- Removed a pointless check which caused problems interworking
3404 with Clearcase. V would complain about shared objects whose
3405 names did not end ".so", and refuse to run. This is now fixed.
3406 In fact it was fixed in 1.9.4 but not documented.
3407
3408- Fixed a bug causing an assertion failure of "waiters == 1"
3409 somewhere in vg_scheduler.c, when running large threaded apps,
3410 notably MySQL.
3411
3412- Add support for the munlock system call (124).
3413
3414Some comments about future releases:
3415
34161.9.5 is, we hope, the most stable Valgrind so far. It pretty much
3417supersedes the 1.0.X branch. If you are a valgrind packager, please
3418consider making 1.9.5 available to your users. You can regard the
34191.0.X branch as obsolete: 1.9.5 is stable and vastly superior. There
3420are no plans at all for further releases of the 1.0.X branch.
3421
3422If you want a leading-edge valgrind, consider building the cvs head
3423(from SourceForge), or getting a snapshot of it. Current cool stuff
3424going in includes MMX support (done); SSE/SSE2 support (in progress),
3425a significant (10-20%) performance improvement (done), and the usual
3426large collection of minor changes. Hopefully we will be able to
3427improve our NPTL support, but no promises.
3428