blob: 6e484ef8f878636ccff9129b3f3b8072b94709f2 [file] [log] [blame]
philippea501f122013-12-01 12:42:30 +00001Release 3.10.0 (?? ?????? 201?)
2~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
33.10.0 is not yet released.
4
5* ================== PLATFORM CHANGES =================
6
bartb1b01822013-12-01 12:59:32 +00007
philippea501f122013-12-01 12:42:30 +00008* ==================== TOOL CHANGES ====================
9
10* Helgrind:
11
12* ==================== OTHER CHANGES ====================
13
philippe8587b542013-12-15 20:24:43 +000014* New and modified GDB server monitor features:
15
16 - A new monitor command "v.info stats" that shows various valgrind core and tool
17 statistics.
18
philippea501f122013-12-01 12:42:30 +000019* ==================== FIXED BUGS ====================
20
21The following bugs have been fixed or resolved. Note that "n-i-bz"
22stands for "not in bugzilla" -- that is, a bug that was reported to us
23but never got a bugzilla entry. We encourage you to file bugs in
24bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather
25than mailing the developers (or mailing lists) directly -- bugs that
26are not entered into bugzilla tend to get forgotten about or ignored.
27
28To see details of a given bug, visit
29 https://bugs.kde.org/show_bug.cgi?id=XXXXXX
30where XXXXXX is the bug number as listed below.
31
philippe3c761f02013-12-01 14:56:28 +000032326462 Refactor vgdb to isolate invoker stuff into separate module
philippe7553cde2013-12-01 12:52:00 +000033326983 Clear direction flag after tests on amd64.
34327238 Callgrind Assertion 'passed <= last_bb->cjmp_count' failed
35327837 dwz compressed alternate .debug_info and .debug_str not read correctly
36327916 DW_TAG_typedef may have no name
mjw7cd88482013-12-10 09:14:52 +000037328100 XABORT not implemented
philippe6d9e4232013-12-01 14:51:21 +000038328205 Implement additional Xen hypercalls
floriana9c40a32013-12-10 16:53:59 +000039328455 s390x: SIGILL after emitting wrong register pair for ldxbr
mjw4229cbd2013-12-12 21:20:48 +000040328711 valgrind.1 manpage "memcheck options" section is badly generated
sewardj4c7254d2013-11-29 23:08:28 +000041
42Release 3.9.0 (31 October 2013)
43~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
443.9.0 is a feature release with many improvements and the usual
45collection of bug fixes.
46
47This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux,
48PPC64/Linux, S390X/Linux, MIPS32/Linux, MIPS64/Linux, ARM/Android,
49X86/Android, X86/MacOSX 10.7 and AMD64/MacOSX 10.7. Support for
50MacOSX 10.8 is significantly improved relative to the 3.8.0 release.
philippe2d75ea22012-08-16 19:40:52 +000051
52* ================== PLATFORM CHANGES =================
53
sewardj4c7254d2013-11-29 23:08:28 +000054* Support for MIPS64 LE and BE running Linux. Valgrind has been
55 tested on MIPS64 Debian Squeeze and Debian Wheezy distributions.
petarj817232e2013-10-19 02:24:11 +000056
57* Support for MIPS DSP ASE on MIPS32 platforms.
sewardj6fd919d2013-03-04 11:27:25 +000058
sewardj4c7254d2013-11-29 23:08:28 +000059* Support for s390x Decimal Floating Point instructions on hosts that
florian0acaac32013-07-28 16:19:34 +000060 have the DFP facility installed.
61
sewardj4c7254d2013-11-29 23:08:28 +000062* Support for POWER8 (Power ISA 2.07) instructions
sewardjbf267d42013-10-18 15:13:12 +000063
sewardj4c7254d2013-11-29 23:08:28 +000064* Support for Intel AVX2 instructions. This is available only on 64
65 bit code.
sewardjbf267d42013-10-18 15:13:12 +000066
sewardj4c7254d2013-11-29 23:08:28 +000067* Initial support for Intel Transactional Synchronization Extensions,
68 both RTM and HLE.
69
70* Initial support for Hardware Transactional Memory on POWER.
71
72* Improved support for MacOSX 10.8 (64-bit only). Memcheck can now
73 run large GUI apps tolerably well.
sewardjbf267d42013-10-18 15:13:12 +000074
philippe2d75ea22012-08-16 19:40:52 +000075* ==================== TOOL CHANGES ====================
sewardj4c7254d2013-11-29 23:08:28 +000076
philippe2193a7c2012-12-08 17:54:16 +000077* Memcheck:
78
sewardj4c7254d2013-11-29 23:08:28 +000079 - Improvements in handling of vectorised code, leading to
80 significantly fewer false error reports. You need to use the flag
81 --partial-loads-ok=yes to get the benefits of these changes.
philippe2d75ea22012-08-16 19:40:52 +000082
sewardj4c7254d2013-11-29 23:08:28 +000083 - Better control over the leak checker. It is now possible to
84 specify which leak kinds (definite/indirect/possible/reachable)
85 should be displayed, which should be regarded as errors, and which
86 should be suppressed by a given leak suppression. This is done
87 using the options --show-leak-kinds=kind1,kind2,..,
88 --errors-for-leak-kinds=kind1,kind2,.. and an optional
89 "match-leak-kinds:" line in suppression entries, respectively.
philippeab1fce92013-09-29 13:47:32 +000090
sewardj4c7254d2013-11-29 23:08:28 +000091 Note that generated leak suppressions contain this new line and
92 are therefore more specific than in previous releases. To get the
93 same behaviour as previous releases, remove the "match-leak-kinds:"
94 line from generated suppressions before using them.
philippe8617b5b2013-01-12 19:53:08 +000095
sewardj4c7254d2013-11-29 23:08:28 +000096 - Reduced "possible leak" reports from the leak checker by the use
97 of better heuristics. The available heuristics provide detection
98 of valid interior pointers to std::stdstring, to new[] allocated
99 arrays with elements having destructors and to interior pointers
100 pointing to an inner part of a C++ object using multiple
101 inheritance. They can be selected individually using the
102 option --leak-check-heuristics=heur1,heur2,...
103
104 - Better control of stacktrace acquisition for heap-allocated
105 blocks. Using the --keep-stacktraces option, it is possible to
106 control independently whether a stack trace is acquired for each
107 allocation and deallocation. This can be used to create better
108 "use after free" errors or to decrease Valgrind's resource
109 consumption by recording less information.
110
111 - Better reporting of leak suppression usage. The list of used
112 suppressions (shown when the -v option is given) now shows, for
113 each leak suppressions, how many blocks and bytes it suppressed
114 during the last leak search.
115
116* Helgrind:
117
118 - False errors resulting from the use of statically initialised
119 mutexes and condition variables (PTHREAD_MUTEX_INITIALISER, etc)
120 have been removed.
121
122 - False errors resulting from the use of pthread_cond_waits that
123 timeout, have been removed.
philippe4e32d672013-10-17 22:10:41 +0000124
philippe2d75ea22012-08-16 19:40:52 +0000125* ==================== OTHER CHANGES ====================
126
sewardj4c7254d2013-11-29 23:08:28 +0000127* Some attempt to tune Valgrind's space requirements to the expected
128 capabilities of the target:
philippe8e1bee42013-10-18 00:08:20 +0000129
sewardj4c7254d2013-11-29 23:08:28 +0000130 - The default size of the translation cache has been reduced from 8
131 sectors to 6 on Android platforms, since each sector occupies
132 about 40MB when using Memcheck.
philippe46207652013-01-20 17:11:58 +0000133
sewardj4c7254d2013-11-29 23:08:28 +0000134 - The default size of the translation cache has been increased to 16
135 sectors on all other platforms, reflecting the fact that large
136 applications require instrumentation and storage of huge amounts
137 of code. For similar reasons, the number of memory mapped
138 segments that can be tracked has been increased by a factor of 6.
philippe46207652013-01-20 17:11:58 +0000139
sewardj4c7254d2013-11-29 23:08:28 +0000140 - In all cases, the maximum number of sectors in the translation
141 cache can be controlled by the new flag --num-transtab-sectors.
philippe6ec8d632013-01-23 22:10:28 +0000142
sewardj4c7254d2013-11-29 23:08:28 +0000143* Changes in how debug info (line numbers, etc) is read:
philippe02ea4132013-09-04 21:42:43 +0000144
sewardj4c7254d2013-11-29 23:08:28 +0000145 - Valgrind no longer temporarily mmaps the entire object to read
146 from it. Instead, reading is done through a small fixed sized
147 buffer. This avoids virtual memory usage spikes when Valgrind
148 reads debuginfo from large shared objects.
philippe6ec8d632013-01-23 22:10:28 +0000149
sewardj4c7254d2013-11-29 23:08:28 +0000150 - A new experimental remote debug info server. Valgrind can read
151 debug info from a different machine (typically, a build host)
152 where debuginfo objects are stored. This can save a lot of time
153 and hassle when running Valgrind on resource-constrained targets
154 (phones, tablets) when the full debuginfo objects are stored
155 somewhere else. This is enabled by the --debuginfo-server=
156 option.
philippe6ec8d632013-01-23 22:10:28 +0000157
sewardj4c7254d2013-11-29 23:08:28 +0000158 - Consistency checking between main and debug objects can be
159 disabled using the --allow-mismatched-debuginfo option.
philippe4e32d672013-10-17 22:10:41 +0000160
sewardj4c7254d2013-11-29 23:08:28 +0000161* Stack unwinding by stack scanning, on ARM. Unwinding by stack
162 scanning can recover stack traces in some cases when the normal
163 unwind mechanisms fail. Stack scanning is best described as "a
164 nasty, dangerous and misleading hack" and so is disabled by default.
165 Use --unw-stack-scan-thresh and --unw-stack-scan-frames to enable
166 and control it.
sewardj82792172013-09-19 23:04:03 +0000167
sewardj4c7254d2013-11-29 23:08:28 +0000168* Detection and merging of recursive stack frame cycles. When your
169 program has recursive algorithms, this limits the memory used by
170 Valgrind for recorded stack traces and avoids recording
171 uninteresting repeated calls. This is controlled by the command
172 line option --merge-recursive-frame and by the monitor command
173 "v.set merge-recursive-frames".
sewardj82792172013-09-19 23:04:03 +0000174
sewardj4c7254d2013-11-29 23:08:28 +0000175* File name and line numbers for used suppressions. The list of used
176 suppressions (shown when the -v option is given) now shows, for each
177 used suppression, the file name and line number where the suppression
178 is defined.
sewardj82792172013-09-19 23:04:03 +0000179
sewardj4c7254d2013-11-29 23:08:28 +0000180* New and modified GDB server monitor features:
sewardj82792172013-09-19 23:04:03 +0000181
sewardj4c7254d2013-11-29 23:08:28 +0000182 - valgrind.h has a new client request, VALGRIND_MONITOR_COMMAND,
183 that can be used to execute gdbserver monitor commands from the
184 client program.
185
186 - A new monitor command, "v.info open_fds", that gives the list of
187 open file descriptors and additional details.
188
189 - An optional message in the "v.info n_errs_found" monitor command,
190 for example "v.info n_errs_found test 1234 finished", allowing a
191 comment string to be added to the process output, perhaps for the
192 purpose of separating errors of different tests or test phases.
193
194 - A new monitor command "v.info execontext" that shows information
195 about the stack traces recorded by Valgrind.
196
197 - A new monitor command "v.do expensive_sanity_check_general" to run
198 some internal consistency checks.
199
200* New flag --sigill-diagnostics to control whether a diagnostic
201 message is printed when the JIT encounters an instruction it can't
202 translate. The actual behavior -- delivery of SIGILL to the
203 application -- is unchanged.
204
205* The maximum amount of memory that Valgrind can use on 64 bit targets
206 has been increased from 32GB to 64GB. This should make it possible
207 to run applications on Memcheck that natively require up to about 35GB.
sewardj82792172013-09-19 23:04:03 +0000208
philippe2d75ea22012-08-16 19:40:52 +0000209* ==================== FIXED BUGS ====================
210
211The following bugs have been fixed or resolved. Note that "n-i-bz"
212stands for "not in bugzilla" -- that is, a bug that was reported to us
213but never got a bugzilla entry. We encourage you to file bugs in
214bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather
215than mailing the developers (or mailing lists) directly -- bugs that
216are not entered into bugzilla tend to get forgotten about or ignored.
217
218To see details of a given bug, visit
219 https://bugs.kde.org/show_bug.cgi?id=XXXXXX
220where XXXXXX is the bug number as listed below.
221
sewardj4c7254d2013-11-29 23:08:28 +0000222123837 system call: 4th argument is optional, depending on cmd
sewardjdae2cd62013-04-02 08:24:48 +0000223135425 memcheck should tell you where Freed blocks were Mallocd
sewardj146a4352013-09-29 09:59:50 +0000224164485 VG_N_SEGNAMES and VG_N_SEGMENTS are (still) too small
sewardj4c7254d2013-11-29 23:08:28 +0000225207815 Adds some of the drm ioctls to syswrap-linux.c
226251569 vex amd64->IR: 0xF 0x1 0xF9 0xBF 0x90 0xD0 0x3 0x0 (RDTSCP)
227252955 Impossible to compile with ccache
228253519 Memcheck reports auxv pointer accesses as invalid reads.
229263034 Crash when loading some PPC64 binaries
230269599 Increase deepest backtrace
231274695 s390x: Support "compare to/from logical" instructions (z196)
232275800 s390x: Autodetect cache info (part 2)
233280271 Valgrind reports possible memory leaks on still-reachable std::string
234284540 Memcheck shouldn't count suppressions matching still-reachable [..]
sewardjbf267d42013-10-18 15:13:12 +0000235289578 Backtraces with ARM unwind tables (stack scan flags)
sewardj4c7254d2013-11-29 23:08:28 +0000236296311 Wrong stack traces due to -fomit-frame-pointer (x86)
237304832 ppc32: build failure
238305431 Use find_buildid shdr fallback for separate .debug files
239305728 Add support for AVX2 instructions
240305948 ppc64: code generation for ShlD64 / ShrD64 asserts
241306035 s390x: Fix IR generation for LAAG and friends
242306054 s390x: Condition code computation for convert-to-int/logical
243306098 s390x: alternate opcode form for convert to/from fixed
244306587 Fix cache line detection from auxiliary vector for PPC.
245306783 Mips unhandled syscall : 4025 / 4079 / 4182
246307038 DWARF2 CFI reader: unhandled DW_OP_ opcode 0x8 (DW_OP_const1u et al)
247307082 HG false positive: pthread_cond_destroy: destruction of unknown CV
248307101 sys_capget second argument can be NULL
249307103 sys_openat: If pathname is absolute, then dirfd is ignored.
250307106 amd64->IR: f0 0f c0 02 (lock xadd byte)
251307113 s390x: DFP support
252307141 valgrind does't work in mips-linux system
253307155 filter_gdb should filter out syscall-template.S T_PSEUDO
254307285 x86_amd64 feature test for avx in test suite is wrong
255307290 memcheck overlap testcase needs memcpy version filter
256307463 Please add "&limit=0" to the "all open bugs" link
257307465 --show-possibly-lost=no should reduce the error count / exit code
258307557 Leaks on Mac OS X 10.7.5 libraries at ImageLoader::recursiveInit[..]
259307729 pkgconfig support broken valgrind.pc
260307828 Memcheck false errors SSE optimized wcscpy, wcscmp, wcsrchr, wcschr
261307955 Building valgrind 3.7.0-r4 fails in Gentoo AMD64 when using clang
262308089 Unhandled syscall on ppc64: prctl
263308135 PPC32 MPC8xx has 16 bytes cache size
264308321 testsuite memcheck filter interferes with gdb_filter
265308333 == 307106
266308341 vgdb should report process exit (or fatal signal)
267308427 s390 memcheck reports tsearch cjump/cmove depends on uninit
268308495 Remove build dependency on installed Xen headers
269308573 Internal error on 64-bit instruction executed in 32-bit mode
270308626 == 308627
271308627 pmovmskb validity bit propagation is imprecise
272308644 vgdb command for having the info for the track-fds option
273308711 give more info about aspacemgr and arenas in out_of_memory
274308717 ARM: implement fixed-point VCVT.F64.[SU]32
275308718 ARM implement SMLALBB family of instructions
276308886 Missing support for PTRACE_SET/GETREGSET
277308930 syscall name_to_handle_at (303 on amd64) not handled
278309229 V-bit tester does not report number of tests generated
279309323 print unrecognized instuction on MIPS
280309425 Provide a --sigill-diagnostics flag to suppress illegal [..]
281309427 SSE optimized stpncpy trigger uninitialised value [..] errors
sewardjf15231f2013-10-22 14:46:55 +0000282309430 Self hosting ppc64 encounters a vassert error on operand type
sewardj4c7254d2013-11-29 23:08:28 +0000283309600 valgrind is a bit confused about 0-sized sections
284309823 Generate errors for still reachable blocks
285309921 PCMPISTRI validity bit propagation is imprecise
286309922 none/tests/ppc64/test_dfp5 sometimes fails
287310169 The Iop_CmpORD class of Iops is not supported by the vbit checker.
288310424 --read-var-info does not properly describe static variables
289310792 search additional path for debug symbols
290310931 s390x: Message-security assist (MSA) instruction extension [..]
291311100 PPC DFP implementation of the integer operands is inconsistent [..]
292311318 ARM: "128-bit constant is not implemented" error message
293311407 ssse3 bcopy (actually converted memcpy) causes invalid read [..]
294311690 V crashes because it redirects branches inside of a redirected function
295311880 x86_64: make regtest hangs at shell_valid1
296311922 WARNING: unhandled syscall: 170
297311933 == 251569
298312171 ppc: insn selection for DFP
299312571 Rounding mode call wrong for the DFP Iops [..]
300312620 Change to Iop_D32toD64 [..] for s390 DFP support broke ppc [..]
301312913 Dangling pointers error should also report the alloc stack trace
302312980 Building on Mountain Lion generates some compiler warnings
303313267 Adding MIPS64/Linux port to Valgrind
304313348 == 251569
305313354 == 251569
306313811 Buffer overflow in assert_fail
307314099 coverity pointed out error in VEX guest_ppc_toIR.c insn_suffix
308314269 ppc: dead code in insn selection
309314718 ARM: implement integer divide instruction (sdiv and udiv)
310315345 cl-format.xml and callgrind/dump.c don't agree on using cfl= or cfi=
311315441 sendmsg syscall should ignore unset msghdr msg_flags
312315534 msgrcv inside a thread causes valgrind to hang (block)
313315545 Assertion '(UChar*)sec->tt[tteNo].tcptr <= (UChar*)hcode' failed
314315689 disInstr(thumb): unhandled instruction: 0xF852 0x0E10 (LDRT)
315315738 disInstr(arm): unhandled instruction: 0xEEBE0BEE (vcvt.s32.f64)
316315959 valgrind man page has bogus SGCHECK (and no BBV) OPTIONS section
317316144 valgrind.1 manpage contains unknown ??? strings [..]
318316145 callgrind command line options in manpage reference (unknown) [..]
319316145 callgrind command line options in manpage reference [..]
320316181 drd: Fixed a 4x slowdown for certain applications
321316503 Valgrind does not support SSE4 "movntdqa" instruction
322316535 Use of |signed int| instead of |size_t| in valgrind messages
323316696 fluidanimate program of parsec 2.1 stuck
324316761 syscall open_by_handle_at (304 on amd64, 342 on x86) not handled
325317091 Use -Wl,-Ttext-segment when static linking if possible [..]
326317186 "Impossible happens" when occurs VCVT instruction on ARM
327317318 Support for Threading Building Blocks "scalable_malloc"
328317444 amd64->IR: 0xC4 0x41 0x2C 0xC2 0xD2 0x8 (vcmpeq_uqps)
329317461 Fix BMI assembler configure check and avx2/bmi/fma vgtest prereqs
330317463 bmi testcase IR SANITY CHECK FAILURE
331317506 memcheck/tests/vbit-test fails with unknown opcode after [..]
332318050 libmpiwrap fails to compile with out-of-source build
333318203 setsockopt handling needs to handle SOL_SOCKET/SO_ATTACH_FILTER
334318643 annotate_trace_memory tests infinite loop on arm and ppc [..]
335318773 amd64->IR: 0xF3 0x48 0x0F 0xBC 0xC2 0xC3 0x66 0x0F
336318929 Crash with: disInstr(thumb): 0xF321 0x0001 (ssat16)
337318932 Add missing PPC64 and PPC32 system call support
338319235 --db-attach=yes is broken with Yama (ptrace scoping) enabled
339319395 Crash with unhandled instruction on STRT (Thumb) instructions
340319494 VEX Makefile-gcc standalone build update after r2702
341319505 [MIPSEL] Crash: unhandled UNRAY operator.
342319858 disInstr(thumb): unhandled instruction on instruction STRBT
343319932 disInstr(thumb): unhandled instruction on instruction STRHT
344320057 Problems when we try to mmap more than 12 memory pages on MIPS32
345320063 Memory from PTRACE_GET_THREAD_AREA is reported uninitialised
346320083 disInstr(thumb): unhandled instruction on instruction LDRBT
347320116 bind on AF_BLUETOOTH produces warnings because of sockaddr_rc padding
348320131 WARNING: unhandled syscall: 369 on ARM (prlimit64)
349320211 Stack buffer overflow in ./coregrind/m_main.c with huge TMPDIR
350320661 vgModuleLocal_read_elf_debug_info(): "Assertion '!di->soname'
351320895 add fanotify support (patch included)
352320998 vex amd64->IR pcmpestri and pcmpestrm SSE4.2 instruction
353321065 Valgrind updates for Xen 4.3
354321148 Unhandled instruction: PLI (Thumb 1, 2, 3)
355321363 Unhandled instruction: SSAX (ARM + Thumb)
356321364 Unhandled instruction: SXTAB16 (ARM + Thumb)
357321466 Unhandled instruction: SHASX (ARM + Thumb)
358321467 Unhandled instruction: SHSAX (ARM + Thumb)
359321468 Unhandled instruction: SHSUB16 (ARM + Thumb)
360321619 Unhandled instruction: SHSUB8 (ARM + Thumb)
361321620 Unhandled instruction: UASX (ARM + Thumb)
362321621 Unhandled instruction: USAX (ARM + Thumb)
363321692 Unhandled instruction: UQADD16 (ARM + Thumb)
364321693 Unhandled instruction: LDRSBT (Thumb)
365321694 Unhandled instruction: UQASX (ARM + Thumb)
366321696 Unhandled instruction: UQSAX (Thumb + ARM)
367321697 Unhandled instruction: UHASX (ARM + Thumb)
368321703 Unhandled instruction: UHSAX (ARM + Thumb)
369321704 Unhandled instruction: REVSH (ARM + Thumb)
370321730 Add cg_diff and cg_merge man pages
371321738 Add vgdb and valgrind-listener man pages
372321814 == 315545
373321891 Unhandled instruction: LDRHT (Thumb)
374321960 pthread_create() then alloca() causing invalid stack write errors
375321969 ppc32 and ppc64 don't support [lf]setxattr
376322254 Show threadname together with tid if set by application
377322294 Add initial support for IBM Power ISA 2.07
378322368 Assertion failure in wqthread_hijack under OS X 10.8
379322563 vex mips->IR: 0x70 0x83 0xF0 0x3A
380322807 VALGRIND_PRINTF_BACKTRACE writes callstack to xml and text to stderr
381322851 0bXXX binary literal syntax is not standard
382323035 Unhandled instruction: LDRSHT(Thumb)
383323036 Unhandled instruction: SMMLS (ARM and Thumb)
384323116 The memcheck/tests/ppc64/power_ISA2_05.c fails to build [..]
385323175 Unhandled instruction: SMLALD (ARM + Thumb)
386323177 Unhandled instruction: SMLSLD (ARM + Thumb)
387323432 Calling pthread_cond_destroy() or pthread_mutex_destroy() [..]
388323437 Phase 2 support for IBM Power ISA 2.07
389323713 Support mmxext (integer sse) subset on i386 (athlon)
390323803 Transactional memory instructions are not supported for Power
391323893 SSE3 not available on amd cpus in valgrind
392323905 Probable false positive from Valgrind/drd on close()
393323912 valgrind.h header isn't compatible for mingw64
394324047 Valgrind doesn't support [LDR,ST]{S}[B,H]T ARM instructions
395324149 helgrind: When pthread_cond_timedwait returns ETIMEDOUT [..]
396324181 mmap does not handle MAP_32BIT
397324227 memcheck false positive leak when a thread calls exit+block [..]
398324421 Support for fanotify API on ARM architecture
399324514 gdbserver monitor cmd output behaviour consistency [..]
400324518 ppc64: Emulation of dcbt instructions does not handle [..]
401324546 none/tests/ppc32 test_isa_2_07_part2 requests -m64
402324582 When access is made to freed memory, report both allocation [..]
403324594 Fix overflow computation for Power ISA 2.06 insns: mulldo/mulldo.
404324765 ppc64: illegal instruction when executing none/tests/ppc64/jm-misc
405324816 Incorrect VEX implementation for xscvspdp/xvcvspdp for SNaN inputs
406324834 Unhandled instructions in Microsoft C run-time for x86_64
407324894 Phase 3 support for IBM Power ISA 2.07
408326091 drd: Avoid false race reports from optimized strlen() impls
409326113 valgrind libvex hwcaps error on AMD64
410n-i-bz Some wrong command line options could be ignored
411n-i-bz patch to allow fair-sched on android
412n-i-bz report error for vgdb snapshot requested before execution
413n-i-bz same as 303624 (fixed in 3.8.0), but for x86 android
414
415(3.9.0: 31 October 2013, vex r2796, valgrind r13708)
416
sewardjf15231f2013-10-22 14:46:55 +0000417
florian02a42332013-08-03 20:40:05 +0000418
sewardj328f9182012-09-18 07:03:27 +0000419Release 3.8.1 (19 September 2012)
420~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4213.8.1 is a bug fix release. It fixes some assertion failures in 3.8.0
422that occur moderately frequently in real use cases, adds support for
423some missing instructions on ARM, and fixes a deadlock condition on
424MacOSX. If you package or deliver 3.8.0 for others to use, you might
425want to consider upgrading to 3.8.1 instead.
426
427The following bugs have been fixed or resolved. Note that "n-i-bz"
428stands for "not in bugzilla" -- that is, a bug that was reported to us
429but never got a bugzilla entry. We encourage you to file bugs in
430bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather
431than mailing the developers (or mailing lists) directly -- bugs that
432are not entered into bugzilla tend to get forgotten about or ignored.
433
434To see details of a given bug, visit
435 https://bugs.kde.org/show_bug.cgi?id=XXXXXX
436where XXXXXX is the bug number as listed below.
437
438284004 == 301281
439289584 Unhandled instruction: 0xF 0x29 0xE5 (MOVAPS)
440295808 amd64->IR: 0xF3 0xF 0xBC 0xC0 (TZCNT)
441298281 wcslen causes false(?) uninitialised value warnings
442301281 valgrind hangs on OS X when the process calls system()
443304035 disInstr(arm): unhandled instruction 0xE1023053
444304867 implement MOVBE instruction in x86 mode
445304980 Assertion 'lo <= hi' failed in vgModuleLocal_find_rx_mapping
446305042 amd64: implement 0F 7F encoding of movq between two registers
447305199 ARM: implement QDADD and QDSUB
448305321 amd64->IR: 0xF 0xD 0xC (prefetchw)
449305513 killed by fatal signal: SIGSEGV
450305690 DRD reporting invalid semaphore when sem_trywait fails
451305926 Invalid alignment checks for some AVX instructions
452306297 disInstr(thumb): unhandled instruction 0xE883 0x000C
453306310 3.8.0 release tarball missing some files
454306612 RHEL 6 glibc-2.X default suppressions need /lib*/libc-*patterns
455306664 vex amd64->IR: 0x66 0xF 0x3A 0x62 0xD1 0x46 0x66 0xF
456n-i-bz shmat of a segment > 4Gb does not work
457n-i-bz simulate_control_c script wrong USR1 signal number on mips
458n-i-bz vgdb ptrace calls wrong on mips [...]
459n-i-bz Fixes for more MPI false positives
460n-i-bz exp-sgcheck's memcpy causes programs to segfault
461n-i-bz OSX build w/ clang: asserts at startup
462n-i-bz Incorrect undef'dness prop for Iop_DPBtoBCD and Iop_BCDtoDPB
463n-i-bz fix a couple of union tag-vs-field mixups
464n-i-bz OSX: use __NR_poll_nocancel rather than __NR_poll
465
466The following bugs were fixed in 3.8.0 but not listed in this NEWS
467file at the time:
468
469254088 Valgrind should know about UD2 instruction
470301280 == 254088
471301902 == 254088
472304754 NEWS blows TeX's little mind
473
sewardj4c7254d2013-11-29 23:08:28 +0000474(3.8.1: 19 September 2012, vex r2537, valgrind r12996)
sewardj81f92412012-09-17 18:20:29 +0000475
sewardj5036a7a2012-08-24 16:53:02 +0000476
florianff7efa62011-11-16 03:59:49 +0000477
sewardj1c41a1a2012-08-09 15:06:36 +0000478Release 3.8.0 (10 August 2012)
sewardj31e8f542012-08-05 14:57:38 +0000479~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardj8ab8a8e2012-08-07 14:54:17 +00004803.8.0 is a feature release with many improvements and the usual
481collection of bug fixes.
sewardj31e8f542012-08-05 14:57:38 +0000482
483This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux,
484PPC64/Linux, S390X/Linux, MIPS/Linux, ARM/Android, X86/Android,
485X86/MacOSX 10.6/10.7 and AMD64/MacOSX 10.6/10.7. Support for recent
486distros and toolchain components (glibc 2.16, gcc 4.7) has been added.
487There is initial support for MacOSX 10.8, but it is not usable for
488serious work at present.
sewardj7d79e0b2012-05-21 21:57:05 +0000489
florian8dc79ce2011-12-10 16:00:25 +0000490* ================== PLATFORM CHANGES =================
491
sewardj31e8f542012-08-05 14:57:38 +0000492* Support for MIPS32 platforms running Linux. Valgrind has been
493 tested on MIPS32 and MIPS32r2 platforms running different Debian
494 Squeeze and MeeGo distributions. Both little-endian and big-endian
sewardj10856a02012-08-05 15:24:12 +0000495 cores are supported. The tools Memcheck, Massif and Lackey have
496 been tested and are known to work. See README.mips for more details.
497
sewardj8ab8a8e2012-08-07 14:54:17 +0000498* Preliminary support for Android running on x86.
sewardj10856a02012-08-05 15:24:12 +0000499
500* Preliminary (as-yet largely unusable) support for MacOSX 10.8.
sewardj31e8f542012-08-05 14:57:38 +0000501
502* Support for Intel AVX instructions and for AES instructions. This
503 support is available only for 64 bit code.
504
sewardj10856a02012-08-05 15:24:12 +0000505* Support for POWER Decimal Floating Point instructions.
philippe5d5dd8e2012-08-05 00:08:25 +0000506
florian8dc79ce2011-12-10 16:00:25 +0000507* ==================== TOOL CHANGES ====================
508
sewardj10856a02012-08-05 15:24:12 +0000509* Non-libc malloc implementations are now supported. This is useful
510 for tools that replace malloc (Memcheck, Massif, DRD, Helgrind).
511 Using the new option --soname-synonyms, such tools can be informed
512 that the malloc implementation is either linked statically into the
513 executable, or is present in some other shared library different
514 from libc.so. This makes it possible to process statically linked
515 programs, and programs using other malloc libraries, for example
516 TCMalloc or JEMalloc.
517
518* For tools that provide their own replacement for malloc et al, the
519 option --redzone-size=<number> allows users to specify the size of
520 the padding blocks (redzones) added before and after each client
521 allocated block. Smaller redzones decrease the memory needed by
522 Valgrind. Bigger redzones increase the chance to detect blocks
523 overrun or underrun. Prior to this change, the redzone size was
524 hardwired to 16 bytes in Memcheck.
philippe1e470b52012-05-11 19:33:46 +0000525
philippe84234902012-01-14 13:53:13 +0000526* Memcheck:
527
528 - The leak_check GDB server monitor command now can
529 control the maximum nr of loss records to output.
530
philippe1e470b52012-05-11 19:33:46 +0000531 - Reduction of memory use for applications allocating
philippe6643e962012-01-17 21:16:30 +0000532 many blocks and/or having many partially defined bytes.
533
philippea22f59d2012-01-26 23:13:52 +0000534 - Addition of GDB server monitor command 'block_list' that lists
535 the addresses/sizes of the blocks of a leak search loss record.
536
537 - Addition of GDB server monitor command 'who_points_at' that lists
538 the locations pointing at a block.
philippe6643e962012-01-17 21:16:30 +0000539
sewardj10856a02012-08-05 15:24:12 +0000540 - If a redzone size > 0 is given, VALGRIND_MALLOCLIKE_BLOCK now will
541 detect an invalid access of these redzones, by marking them
542 noaccess. Similarly, if a redzone size is given for a memory
543 pool, VALGRIND_MEMPOOL_ALLOC will mark the redzones no access.
544 This still allows to find some bugs if the user has forgotten to
545 mark the pool superblock noaccess.
philippe11cc9ce2012-03-11 17:59:00 +0000546
sewardj10856a02012-08-05 15:24:12 +0000547 - Performance of memory leak check has been improved, especially in
548 cases where there are many leaked blocks and/or many suppression
549 rules used to suppress leak reports.
philippe57a16a22012-07-18 22:26:51 +0000550
sewardje060f6a2012-08-05 16:14:02 +0000551 - Reduced noise (false positive) level on MacOSX 10.6/10.7, due to
philippe46217292012-08-05 21:35:02 +0000552 more precise analysis, which is important for LLVM/Clang
sewardje060f6a2012-08-05 16:14:02 +0000553 generated code. This is at the cost of somewhat reduced
554 performance. Note there is no change to analysis precision or
555 costs on Linux targets.
556
bartc3009f82012-06-16 19:02:25 +0000557* DRD:
558
bartc3009f82012-06-16 19:02:25 +0000559 - Added even more facilities that can help finding the cause of a data
560 race, namely the command-line option --ptrace-addr and the macro
561 DRD_STOP_TRACING_VAR(x). More information can be found in the manual.
562
barta7a57882012-08-04 09:19:11 +0000563 - Fixed a subtle bug that could cause false positive data race reports.
564
florian8dc79ce2011-12-10 16:00:25 +0000565* ==================== OTHER CHANGES ====================
566
567* The C++ demangler has been updated so as to work well with C++
philippe172a3bb2012-03-16 15:03:08 +0000568 compiled by up to at least g++ 4.6.
florianff7efa62011-11-16 03:59:49 +0000569
bart1e451642012-07-28 13:06:45 +0000570* Tool developers can make replacement/wrapping more flexible thanks
sewardj10856a02012-08-05 15:24:12 +0000571 to the new option --soname-synonyms. This was reported above, but
572 in fact is very general and applies to all function
573 replacement/wrapping, not just to malloc-family functions.
philippe1e470b52012-05-11 19:33:46 +0000574
sewardj10856a02012-08-05 15:24:12 +0000575* Round-robin scheduling of threads can be selected, using the new
576 option --fair-sched= yes. Prior to this change, the pipe-based
577 thread serialisation mechanism (which is still the default) could
578 give very unfair scheduling. --fair-sched=yes improves
579 responsiveness of interactive multithreaded applications, and
580 improves repeatability of results from the thread checkers Helgrind
581 and DRD.
philippe236a71a2012-02-22 20:23:29 +0000582
sewardj10856a02012-08-05 15:24:12 +0000583* For tool developers: support to run Valgrind on Valgrind has been
584 improved. We can now routinely Valgrind on Helgrind or Memcheck.
philippe72faf102012-03-11 22:24:03 +0000585
philippeede2d6c2012-03-14 21:27:35 +0000586* gdbserver now shows the float shadow registers as integer
587 rather than float values, as the shadow values are mostly
588 used as bit patterns.
589
florian7711f9e2012-06-29 21:20:52 +0000590* Increased limit for the --num-callers command line flag to 500.
591
philippe02ccd562012-08-05 17:35:17 +0000592* Performance improvements for error matching when there are many
sewardj10856a02012-08-05 15:24:12 +0000593 suppression records in use.
philippecbfe62c2012-08-04 09:08:50 +0000594
sewardj10856a02012-08-05 15:24:12 +0000595* Improved support for DWARF4 debugging information (bug 284184).
sewardjd09354b2012-06-30 16:44:45 +0000596
sewardja930ef22012-08-08 20:21:29 +0000597* Initial support for DWZ compressed Dwarf debug info.
598
sewardj10856a02012-08-05 15:24:12 +0000599* Improved control over the IR optimiser's handling of the tradeoff
600 between performance and precision of exceptions. Specifically,
601 --vex-iropt-precise-memory-exns has been removed and replaced by
602 --vex-iropt-register-updates, with extended functionality. This
philippee30542d2012-08-05 17:37:39 +0000603 allows the Valgrind gdbserver to always show up to date register
sewardj10856a02012-08-05 15:24:12 +0000604 values to GDB.
605
606* Modest performance gains through the use of translation chaining for
607 JIT-generated code.
philippe0c0291a2012-08-01 22:03:12 +0000608
florianff7efa62011-11-16 03:59:49 +0000609* ==================== FIXED BUGS ====================
610
611The following bugs have been fixed or resolved. Note that "n-i-bz"
612stands for "not in bugzilla" -- that is, a bug that was reported to us
613but never got a bugzilla entry. We encourage you to file bugs in
sewardj31e8f542012-08-05 14:57:38 +0000614bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather
615than mailing the developers (or mailing lists) directly -- bugs that
616are not entered into bugzilla tend to get forgotten about or ignored.
florianff7efa62011-11-16 03:59:49 +0000617
618To see details of a given bug, visit
sewardj31e8f542012-08-05 14:57:38 +0000619 https://bugs.kde.org/show_bug.cgi?id=XXXXXX
florianff7efa62011-11-16 03:59:49 +0000620where XXXXXX is the bug number as listed below.
621
florianef0bca62012-05-06 03:37:25 +0000622197914 Building valgrind from svn now requires automake-1.10
sewardj31e8f542012-08-05 14:57:38 +0000623203877 increase to 16Mb maximum allowed alignment for memalign et al
624219156 Handle statically linked malloc or other malloc lib (e.g. tcmalloc)
philippe8f927a52011-12-23 11:42:20 +0000625247386 make perf does not run all performance tests
philippe8d674252012-03-20 19:04:39 +0000626270006 Valgrind scheduler unfair
philippeb79add82012-07-01 20:35:26 +0000627270777 Adding MIPS/Linux port to Valgrind
florian722b48e2012-01-16 17:26:48 +0000628270796 s390x: Removed broken support for the TS insn
florianfdfca222012-01-17 13:16:50 +0000629271438 Fix configure for proper SSE4.2 detection
florianda275432012-02-04 17:16:40 +0000630273114 s390x: Support TR, TRE, TROO, TROT, TRTO, and TRTT instructions
sewardj31e8f542012-08-05 14:57:38 +0000631273475 Add support for AVX instructions
philippef7c54332012-05-17 15:32:54 +0000632274078 improved configure logic for mpicc
philippe15e301e2011-12-22 13:25:58 +0000633276993 fix mremap 'no thrash checks'
sewardj5d1d88f2012-07-16 22:39:24 +0000634278313 Fedora 15/x64: err read debug info with --read-var-info=yes flag
sewardj31e8f542012-08-05 14:57:38 +0000635281482 memcheck incorrect byte allocation count in realloc() for silly argument
philippe6643e962012-01-17 21:16:30 +0000636282230 group allocator for small fixed size, use it for MC_Chunk/SEc vbit
florian8dc79ce2011-12-10 16:00:25 +0000637283413 Fix wrong sanity check
florian535bc7d2012-02-26 17:01:22 +0000638283671 Robustize alignment computation in LibVEX_Alloc
sewardjd09354b2012-06-30 16:44:45 +0000639283961 Adding support for some HCI IOCTLs
640284124 parse_type_DIE: confused by: DWARF 4
sewardj31e8f542012-08-05 14:57:38 +0000641284864 == 273475 (Add support for AVX instructions)
sewardjd09354b2012-06-30 16:44:45 +0000642285219 Too-restrictive constraints for Thumb2 "SP plus/minus register"
sewardj31e8f542012-08-05 14:57:38 +0000643285662 (MacOSX): Memcheck needs to replace memcpy/memmove
644285725 == 273475 (Add support for AVX instructions)
645286261 add wrapper for linux I2C_RDWR ioctl
philippe5eb378e2011-12-19 11:03:07 +0000646286270 vgpreload is not friendly to 64->32 bit execs, gives ld.so warnings
647286374 Running cachegrind with --branch-sim=yes on 64-bit PowerPC program fails
sewardj31e8f542012-08-05 14:57:38 +0000648286384 configure fails "checking for a supported version of gcc"
649286497 == 273475 (Add support for AVX instructions)
650286596 == 273475 (Add support for AVX instructions)
sewardj7ceb70c2012-07-11 14:28:13 +0000651286917 disInstr(arm): unhandled instruction: QADD (also QSUB)
sewardj69f799d2012-08-04 08:41:44 +0000652287175 ARM: scalar VFP fixed-point VCVT instructions not handled
floriane3c0a2a2012-01-29 02:23:07 +0000653287260 Incorrect conditional jump or move depends on uninitialised value(s)
sewardjd09354b2012-06-30 16:44:45 +0000654287301 vex amd64->IR: 0x66 0xF 0x38 0x41 0xC0 0xB8 0x0 0x0 (PHMINPOSUW)
sewardj31e8f542012-08-05 14:57:38 +0000655287307 == 273475 (Add support for AVX instructions)
florian59451642011-12-24 21:50:53 +0000656287858 VG_(strerror): unknown error
sewardj69f799d2012-08-04 08:41:44 +0000657288298 (MacOSX) unhandled syscall shm_unlink
sewardj31e8f542012-08-05 14:57:38 +0000658288995 == 273475 (Add support for AVX instructions)
sewardja1c8b4c2012-07-13 14:04:03 +0000659289470 Loading of large Mach-O thin binaries fails.
sewardj31e8f542012-08-05 14:57:38 +0000660289656 == 273475 (Add support for AVX instructions)
philippe1d76a802011-12-26 21:21:37 +0000661289699 vgdb connection in relay mode erroneously closed due to buffer overrun
sewardj31e8f542012-08-05 14:57:38 +0000662289823 == 293754 (PCMPxSTRx not implemented for 16-bit characters)
florian6746c4c2012-08-06 00:10:53 +0000663289839 s390x: Provide support for unicode conversion instructions
sewardj31e8f542012-08-05 14:57:38 +0000664289939 monitor cmd 'leak_check' with details about leaked or reachable blocks
sewardj617f1b22012-07-11 00:06:35 +0000665290006 memcheck doesn't mark %xmm as initialized after "pcmpeqw %xmm %xmm"
philippe6156b772012-02-14 21:35:46 +0000666290655 Add support for AESKEYGENASSIST instruction
philippefaa4f052012-07-01 20:06:04 +0000667290719 valgrind-3.7.0 fails with automake-1.11.2 due to"pkglibdir" usage
philippe53b0d9a2012-02-02 21:33:55 +0000668290974 vgdb must align pages to VKI_SHMLBA (16KB) on ARM
philippefaa4f052012-07-01 20:06:04 +0000669291253 ES register not initialised in valgrind simulation
670291568 Fix 3DNOW-related crashes with baseline x86_64 CPU (w patch)
floriancc58cef2012-06-06 02:27:51 +0000671291865 s390x: Support the "Compare Double and Swap" family of instructions
sewardj31e8f542012-08-05 14:57:38 +0000672292300 == 273475 (Add support for AVX instructions)
673292430 unrecognized instruction in __intel_get_new_mem_ops_cpuid
674292493 == 273475 (Add support for AVX instructions)
philippefaa4f052012-07-01 20:06:04 +0000675292626 Missing fcntl F_SETOWN_EX and F_GETOWN_EX support
676292627 Missing support for some SCSI ioctls
677292628 none/tests/x86/bug125959-x86.c triggers undefined behavior
sewardj31e8f542012-08-05 14:57:38 +0000678292841 == 273475 (Add support for AVX instructions)
philippefaa4f052012-07-01 20:06:04 +0000679292993 implement the getcpu syscall on amd64-linux
680292995 Implement the “cross memory attach” syscalls introduced in Linux 3.2
florian44587792012-02-15 04:05:52 +0000681293088 Add some VEX sanity checks for ppc64 unhandled instructions
sewardj31e8f542012-08-05 14:57:38 +0000682293751 == 290655 (Add support for AESKEYGENASSIST instruction)
philippefaa4f052012-07-01 20:06:04 +0000683293754 PCMPxSTRx not implemented for 16-bit characters
sewardj31e8f542012-08-05 14:57:38 +0000684293755 == 293754 (No tests for PCMPxSTRx on 16-bit characters)
philippefaa4f052012-07-01 20:06:04 +0000685293808 CLFLUSH not supported by latest VEX for amd64
686294047 valgrind does not correctly emulate prlimit64(..., RLIMIT_NOFILE, ...)
687294048 MPSADBW instruction not implemented
florianc6fccd62012-02-15 03:57:27 +0000688294055 regtest none/tests/shell fails when locale is not set to C
philippefaa4f052012-07-01 20:06:04 +0000689294185 INT 0x44 (and others) not supported on x86 guest, but used by Jikes RVM
philippebaf69642012-02-15 22:29:30 +0000690294190 --vgdb-error=xxx can be out of sync with errors shown to the user
sewardj31e8f542012-08-05 14:57:38 +0000691294191 amd64: fnsave/frstor and 0x66 size prefixes on FP instructions
692294260 disInstr_AMD64: disInstr miscalculated next %rip
sewardj8a91f382012-07-05 13:26:07 +0000693294523 --partial-loads-ok=yes causes false negatives
sewardj31e8f542012-08-05 14:57:38 +0000694294617 vex amd64->IR: 0x66 0xF 0x3A 0xDF 0xD1 0x1 0xE8 0x6A
695294736 vex amd64->IR: 0x48 0xF 0xD7 0xD6 0x48 0x83
philippefaa4f052012-07-01 20:06:04 +0000696294812 patch allowing to run (on x86 at least) helgrind/drd on tool.
697295089 can not annotate source for both helgrind and drd
florian2011f752012-07-01 21:55:13 +0000698295221 POWER Processor decimal floating point instruction support missing
sewardj31e8f542012-08-05 14:57:38 +0000699295427 building for i386 with clang on darwin11 requires "-new_linker linker"
florian2011f752012-07-01 21:55:13 +0000700295428 coregrind/m_main.c has incorrect x86 assembly for darwin
sewardj31e8f542012-08-05 14:57:38 +0000701295590 Helgrind: Assertion 'cvi->nWaiters > 0' failed
tomd0b8b372012-07-11 15:42:25 +0000702295617 ARM - Add some missing syscalls
sewardj31e8f542012-08-05 14:57:38 +0000703295799 Missing \n with get_vbits in gdbserver when line is % 80 [...]
sewardja1c8b4c2012-07-13 14:04:03 +0000704296229 Linux user input device ioctls missing wrappers
705296318 ELF Debug info improvements (more than one rx/rw mapping)
philippefaa4f052012-07-01 20:06:04 +0000706296422 Add translation chaining support
sewardj4a2faf12012-07-04 20:30:47 +0000707296457 vex amd64->IR: 0x66 0xF 0x3A 0xDF 0xD1 0x1 0xE8 0x6A (dup of AES)
sewardj31e8f542012-08-05 14:57:38 +0000708296792 valgrind 3.7.0: add SIOCSHWTSTAMP (0x89B0) ioctl wrapper
florian5bdb9ec2012-05-06 03:51:00 +0000709296983 Fix build issues on x86_64/ppc64 without 32-bit toolchains
sewardj31e8f542012-08-05 14:57:38 +0000710297078 gdbserver signal handling problems [..]
bartc3009f82012-06-16 19:02:25 +0000711297147 drd false positives on newly allocated memory
sewardj31e8f542012-08-05 14:57:38 +0000712297329 disallow decoding of IBM Power DFP insns on some machines
florianf5a39372012-07-02 12:14:00 +0000713297497 POWER Processor decimal floating point instruction support missing
sewardja1c8b4c2012-07-13 14:04:03 +0000714297701 Another alias for strncasecmp_l in libc-2.13.so
sewardj31e8f542012-08-05 14:57:38 +0000715297911 'invalid write' not reported when using APIs for custom mem allocators.
florian295bb7f2012-07-01 21:24:53 +0000716297976 s390x: revisit EX implementation
philippe8050bb72012-04-13 23:07:29 +0000717297991 Valgrind interferes with mmap()+ftell()
philippe8fb1b132012-04-13 17:27:40 +0000718297992 Support systems missing WIFCONTINUED (e.g. pre-2.6.10 Linux)
florianf71b35a2012-04-22 19:27:44 +0000719297993 Fix compilation of valgrind with gcc -g3.
sewardj31e8f542012-08-05 14:57:38 +0000720298080 POWER Processor DFP support missing, part 3
721298227 == 273475 (Add support for AVX instructions)
722298335 == 273475 (Add support for AVX instructions)
florianf5a39372012-07-02 12:14:00 +0000723298354 Unhandled ARM Thumb instruction 0xEB0D 0x0585 (streq)
sewardj31e8f542012-08-05 14:57:38 +0000724298394 s390x: Don't bail out on an unknown machine model. [..]
florianf5a39372012-07-02 12:14:00 +0000725298421 accept4() syscall (366) support is missing for ARM
sewardj31e8f542012-08-05 14:57:38 +0000726298718 vex amd64->IR: 0xF 0xB1 0xCB 0x9C 0x8F 0x45
florianf5a39372012-07-02 12:14:00 +0000727298732 valgrind installation problem in ubuntu with kernel version 3.x
sewardj31e8f542012-08-05 14:57:38 +0000728298862 POWER Processor DFP instruction support missing, part 4
florianf5a39372012-07-02 12:14:00 +0000729298864 DWARF reader mis-parses DW_FORM_ref_addr
sewardj31e8f542012-08-05 14:57:38 +0000730298943 massif asserts with --pages-as-heap=yes when brk is changing [..]
florianf5a39372012-07-02 12:14:00 +0000731299053 Support DWARF4 DW_AT_high_pc constant form
sewardj31e8f542012-08-05 14:57:38 +0000732299104 == 273475 (Add support for AVX instructions)
floriand7ebe552012-07-02 15:03:29 +0000733299316 Helgrind: hg_main.c:628 (map_threads_lookup): Assertion 'thr' failed.
sewardj31e8f542012-08-05 14:57:38 +0000734299629 dup3() syscall (358) support is missing for ARM
735299694 POWER Processor DFP instruction support missing, part 5
736299756 Ignore --free-fill for MEMPOOL_FREE and FREELIKE client requests
737299803 == 273475 (Add support for AVX instructions)
738299804 == 273475 (Add support for AVX instructions)
739299805 == 273475 (Add support for AVX instructions)
sewardja1c8b4c2012-07-13 14:04:03 +0000740300140 ARM - Missing (T1) SMMUL
sewardj31e8f542012-08-05 14:57:38 +0000741300195 == 296318 (ELF Debug info improvements (more than one rx/rw mapping))
742300389 Assertion `are_valid_hwcaps(VexArchAMD64, [..])' failed.
floriand7ebe552012-07-02 15:03:29 +0000743300414 FCOM and FCOMP unimplemented for amd64 guest
sewardj8a91f382012-07-05 13:26:07 +0000744301204 infinite loop in canonicaliseSymtab with ifunc symbol
sewardj31e8f542012-08-05 14:57:38 +0000745301229 == 203877 (increase to 16Mb maximum allowed alignment for memalign etc)
philippe5d5dd8e2012-08-05 00:08:25 +0000746301265 add x86 support to Android build
philippefaa4f052012-07-01 20:06:04 +0000747301984 configure script doesn't detect certain versions of clang
748302205 Fix compiler warnings for POWER VEX code and POWER test cases
sewardj5d1d88f2012-07-16 22:39:24 +0000749302287 Unhandled movbe instruction on Atom processors
sewardj31e8f542012-08-05 14:57:38 +0000750302370 PPC: fnmadd, fnmsub, fnmadds, fnmsubs insns always negate the result
floriand7ebe552012-07-02 15:03:29 +0000751302536 Fix for the POWER Valgrind regression test: memcheck-ISA2.0.
sewardj5d1d88f2012-07-16 22:39:24 +0000752302578 Unrecognized isntruction 0xc5 0x32 0xc2 0xca 0x09 vcmpngess
sewardj31e8f542012-08-05 14:57:38 +0000753302656 == 273475 (Add support for AVX instructions)
754302709 valgrind for ARM needs extra tls support for android emulator [..]
tomd0b8b372012-07-11 15:42:25 +0000755302827 add wrapper for CDROM_GET_CAPABILITY
sewardjf7450412012-07-14 10:08:40 +0000756302901 Valgrind crashes with dwz optimized debuginfo
philippeb0e5b692012-07-06 21:42:14 +0000757302918 Enable testing of the vmaddfp and vnsubfp instructions in the testsuite
philippea6be8b12012-07-06 21:59:03 +0000758303116 Add support for the POWER instruction popcntb
philippe32b4a422012-07-06 22:52:09 +0000759303127 Power test suite fixes for frsqrte, vrefp, and vrsqrtefp instructions.
sewardj31e8f542012-08-05 14:57:38 +0000760303250 Assertion `instrs_in->arr_used <= 10000' failed w/ OpenSSL code
761303466 == 273475 (Add support for AVX instructions)
762303624 segmentation fault on Android 4.1 (e.g. on Galaxy Nexus OMAP)
weidendo0ffe10e2012-07-25 12:13:29 +0000763303963 strstr() function produces wrong results under valgrind callgrind
tom0ec79a12012-08-02 09:26:38 +0000764304054 CALL_FN_xx macros need to enforce stack alignment
sewardja930ef22012-08-08 20:21:29 +0000765304561 tee system call not supported
sewardj8ab8a8e2012-08-07 14:54:17 +0000766715750 (MacOSX): Incorrect invalid-address errors near 0xFFFFxxxx (mozbug#)
sewardj31e8f542012-08-05 14:57:38 +0000767n-i-bz Add missing gdbserver xml files for shadow registers for ppc32
768n-i-bz Bypass gcc4.4/4.5 code gen bugs causing out of memory or asserts
769n-i-bz Fix assert in gdbserver for watchpoints watching the same address
770n-i-bz Fix false positive in sys_clone on amd64 when optional args [..]
771n-i-bz s390x: Shadow registers can now be examined using vgdb
772
sewardj6e374f82012-08-08 23:35:35 +0000773(3.8.0-TEST3: 9 August 2012, vex r2465, valgrind r12865)
sewardj1c41a1a2012-08-09 15:06:36 +0000774(3.8.0: 10 August 2012, vex r2465, valgrind r12866)
sewardj10856a02012-08-05 15:24:12 +0000775
sewardj31e8f542012-08-05 14:57:38 +0000776
florian295bb7f2012-07-01 21:24:53 +0000777
sewardj62b0d402011-11-09 23:42:41 +0000778Release 3.7.0 (5 November 2011)
779~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardj469b8362011-10-23 10:05:47 +00007803.7.0 is a feature release with many significant improvements and the
781usual collection of bug fixes.
sewardjdc873c02011-07-24 16:02:33 +0000782
sewardj469b8362011-10-23 10:05:47 +0000783This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux,
784PPC64/Linux, S390X/Linux, ARM/Android, X86/Darwin and AMD64/Darwin.
785Support for recent distros and toolchain components (glibc 2.14, gcc
7864.6, MacOSX 10.7) has been added.
787
788* ================== PLATFORM CHANGES =================
789
790* Support for IBM z/Architecture (s390x) running Linux. Valgrind can
791 analyse 64-bit programs running on z/Architecture. Most user space
792 instructions up to and including z10 are supported. Valgrind has
793 been tested extensively on z9, z10, and z196 machines running SLES
794 10/11, RedHat 5/6m, and Fedora. The Memcheck and Massif tools are
795 known to work well. Callgrind, Helgrind, and DRD work reasonably
796 well on z9 and later models. See README.s390 for more details.
797
798* Preliminary support for MacOSX 10.7 and XCode 4. Both 32- and
799 64-bit processes are supported. Some complex threaded applications
800 (Firefox) are observed to hang when run as 32 bit applications,
801 whereas 64-bit versions run OK. The cause is unknown. Memcheck
802 will likely report some false errors. In general, expect some rough
803 spots. This release also supports MacOSX 10.6, but drops support
804 for 10.5.
805
806* Preliminary support for Android (on ARM). Valgrind can now run
807 large applications (eg, Firefox) on (eg) a Samsung Nexus S. See
808 README.android for more details, plus instructions on how to get
809 started.
810
811* Support for the IBM Power ISA 2.06 (Power7 instructions)
812
813* General correctness and performance improvements for ARM/Linux, and,
814 by extension, ARM/Android.
815
sewardj3bbaf482011-10-24 08:54:56 +0000816* Further solidification of support for SSE 4.2 in 64-bit mode. AVX
sewardj469b8362011-10-23 10:05:47 +0000817 instruction set support is under development but is not available in
818 this release.
819
820* Support for AIX5 has been removed.
821
822* ==================== TOOL CHANGES ====================
823
824* Memcheck: some incremental changes:
825
826 - reduction of memory use in some circumstances
827
828 - improved handling of freed memory, which in some circumstances
829 can cause detection of use-after-free that would previously have
830 been missed
831
832 - fix of a longstanding bug that could cause false negatives (missed
833 errors) in programs doing vector saturated narrowing instructions.
834
835* Helgrind: performance improvements and major memory use reductions,
836 particularly for large, long running applications which perform many
837 synchronisation (lock, unlock, etc) events. Plus many smaller
838 changes:
839
840 - display of locksets for both threads involved in a race
841
842 - general improvements in formatting/clarity of error messages
843
844 - addition of facilities and documentation regarding annotation
845 of thread safe reference counted C++ classes
846
847 - new flag --check-stack-refs=no|yes [yes], to disable race checking
848 on thread stacks (a performance hack)
849
850 - new flag --free-is-write=no|yes [no], to enable detection of races
851 where one thread accesses heap memory but another one frees it,
852 without any coordinating synchronisation event
853
854* DRD: enabled XML output; added support for delayed thread deletion
855 in order to detect races that occur close to the end of a thread
856 (--join-list-vol); fixed a memory leak triggered by repeated client
857 memory allocatation and deallocation; improved Darwin support.
858
florian0f0ed672011-11-26 04:05:47 +0000859* exp-ptrcheck: this tool has been renamed to exp-sgcheck
860
861* exp-sgcheck: this tool has been reduced in scope so as to improve
sewardj469b8362011-10-23 10:05:47 +0000862 performance and remove checking that Memcheck does better.
863 Specifically, the ability to check for overruns for stack and global
864 arrays is unchanged, but the ability to check for overruns of heap
865 blocks has been removed. The tool has accordingly been renamed to
866 exp-sgcheck ("Stack and Global Array Checking").
867
868* ==================== OTHER CHANGES ====================
869
870* GDB server: Valgrind now has an embedded GDB server. That means it
871 is possible to control a Valgrind run from GDB, doing all the usual
872 things that GDB can do (single stepping, breakpoints, examining
873 data, etc). Tool-specific functionality is also available. For
874 example, it is possible to query the definedness state of variables
875 or memory from within GDB when running Memcheck; arbitrarily large
876 memory watchpoints are supported, etc. To use the GDB server, start
877 Valgrind with the flag --vgdb-error=0 and follow the on-screen
878 instructions.
879
880* Improved support for unfriendly self-modifying code: a new option
881 --smc-check=all-non-file is available. This adds the relevant
882 consistency checks only to code that originates in non-file-backed
883 mappings. In effect this confines the consistency checking only to
884 code that is or might be JIT generated, and avoids checks on code
885 that must have been compiled ahead of time. This significantly
886 improves performance on applications that generate code at run time.
887
888* It is now possible to build a working Valgrind using Clang-2.9 on
sewardj960eef22011-09-30 17:55:13 +0000889 Linux.
890
sewardj469b8362011-10-23 10:05:47 +0000891* new client requests VALGRIND_{DISABLE,ENABLE}_ERROR_REPORTING.
892 These enable and disable error reporting on a per-thread, and
893 nestable, basis. This is useful for hiding errors in particularly
894 troublesome pieces of code. The MPI wrapper library (libmpiwrap.c)
895 now uses this facility.
sewardj960eef22011-09-30 17:55:13 +0000896
sewardj469b8362011-10-23 10:05:47 +0000897* Added the --mod-funcname option to cg_diff.
sewardjdc873c02011-07-24 16:02:33 +0000898
sewardj469b8362011-10-23 10:05:47 +0000899* ==================== FIXED BUGS ====================
sewardj4629a732011-05-17 15:54:31 +0000900
sewardj520501d2011-10-23 09:09:54 +0000901The following bugs have been fixed or resolved. Note that "n-i-bz"
902stands for "not in bugzilla" -- that is, a bug that was reported to us
903but never got a bugzilla entry. We encourage you to file bugs in
904bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
905mailing the developers (or mailing lists) directly -- bugs that are
906not entered into bugzilla tend to get forgotten about or ignored.
sewardjffd37d62011-10-13 16:41:17 +0000907
sewardj520501d2011-10-23 09:09:54 +0000908To see details of a given bug, visit
909https://bugs.kde.org/show_bug.cgi?id=XXXXXX
910where XXXXXX is the bug number as listed below.
sewardjffd37d62011-10-13 16:41:17 +0000911
sewardj520501d2011-10-23 09:09:54 +0000912210935 port valgrind.h (not valgrind) to win32 to support client requests
913214223 valgrind SIGSEGV on startup gcc 4.4.1 ppc32 (G4) Ubuntu 9.10
sewardj7a9ff4c2011-06-13 13:36:59 +0000914243404 Port to zSeries
sewardj520501d2011-10-23 09:09:54 +0000915243935 Helgrind: incorrect handling of ANNOTATE_HAPPENS_BEFORE()/AFTER()
916247223 non-x86: Suppress warning: 'regparm' attribute directive ignored
917250101 huge "free" memory usage due to m_mallocfree.c fragmentation
918253206 Some fixes for the faultstatus testcase
919255223 capget testcase fails when running as root
920256703 xlc_dbl_u32.c testcase broken
921256726 Helgrind tests have broken inline asm
922259977 == 214223 (Valgrind segfaults doing __builtin_longjmp)
sewardj7a9ff4c2011-06-13 13:36:59 +0000923264800 testcase compile failure on zseries
sewardj520501d2011-10-23 09:09:54 +0000924265762 make public VEX headers compilable by G++ 3.x
925265771 assertion in jumps.c (r11523) fails with glibc-2.3
926266753 configure script does not give the user the option to not use QtCore
927266931 gen_insn_test.pl is broken
928266961 ld-linux.so.2 i?86-linux strlen issues
929266990 setns instruction causes false positive
930267020 Make directory for temporary files configurable at run-time.
931267342 == 267997 (segmentation fault on Mac OS 10.6)
932267383 Assertion 'vgPlain_strlen(dir) + vgPlain_strlen(file) + 1 < 256' failed
933267413 Assertion 'DRD_(g_threadinfo)[tid].synchr_nesting >= 1' failed.
934267488 regtest: darwin support for 64-bit build
935267552 SIGSEGV (misaligned_stack_error) with DRD, but not with other tools
936267630 Add support for IBM Power ISA 2.06 -- stage 1
937267769 == 267997 (Darwin: memcheck triggers segmentation fault)
938267819 Add client request for informing the core about reallocation
939267925 laog data structure quadratic for a single sequence of lock
940267968 drd: (vgDrd_thread_set_joinable): Assertion '0 <= (int)tid ..' failed
941267997 MacOSX: 64-bit V segfaults on launch when built with Xcode 4.0.1
942268513 missed optimizations in fold_Expr
943268619 s390x: fpr - gpr transfer facility
944268620 s390x: reconsider "long displacement" requirement
945268621 s390x: improve IR generation for XC
946268715 s390x: FLOGR is not universally available
947268792 == 267997 (valgrind seg faults on startup when compiled with Xcode 4)
948268930 s390x: MHY is not universally available
949269078 arm->IR: unhandled instruction SUB (SP minus immediate/register)
950269079 Support ptrace system call on ARM
951269144 missing "Bad option" error message
952269209 conditional load and store facility (z196)
953269354 Shift by zero on x86 can incorrectly clobber CC_NDEP
954269641 == 267997 (valgrind segfaults immediately (segmentation fault))
955269736 s390x: minor code generation tweaks
956269778 == 272986 (valgrind.h: swap roles of VALGRIND_DO_CLIENT_REQUEST() ..)
957269863 s390x: remove unused function parameters
958269864 s390x: tweak s390_emit_load_cc
959269884 == 250101 (overhead for huge blocks exhausts space too soon)
960270082 s390x: Make sure to point the PSW address to the next address on SIGILL
961270115 s390x: rewrite some testcases
962270309 == 267997 (valgrind crash on startup)
963270320 add support for Linux FIOQSIZE ioctl() call
964270326 segfault while trying to sanitize the environment passed to execle
965270794 IBM POWER7 support patch causes regression in none/tests
966270851 IBM POWER7 fcfidus instruction causes memcheck to fail
967270856 IBM POWER7 xsnmaddadp instruction causes memcheck to fail on 32bit app
968270925 hyper-optimized strspn() in /lib64/libc-2.13.so needs fix
969270959 s390x: invalid use of R0 as base register
970271042 VSX configure check fails when it should not
971271043 Valgrind build fails with assembler error on ppc64 with binutils 2.21
972271259 s390x: fix code confusion
973271337 == 267997 (Valgrind segfaults on MacOS X)
974271385 s390x: Implement Ist_MBE
975271501 s390x: misc cleanups
976271504 s390x: promote likely and unlikely
977271579 ppc: using wrong enum type
978271615 unhandled instruction "popcnt" (arch=amd10h)
979271730 Fix bug when checking ioctls: duplicate check
980271776 s390x: provide STFLE instruction support
981271779 s390x: provide clock instructions like STCK
982271799 Darwin: ioctls without an arg report a memory error
983271820 arm: fix type confusion
984271917 pthread_cond_timedwait failure leads to not-locked false positive
985272067 s390x: fix DISP20 macro
986272615 A typo in debug output in mc_leakcheck.c
987272661 callgrind_annotate chokes when run from paths containing regex chars
988272893 amd64->IR: 0x66 0xF 0x38 0x2B 0xC1 0x66 0xF 0x7F == (closed as dup)
989272955 Unhandled syscall error for pwrite64 on ppc64 arch
990272967 make documentation build-system more robust
991272986 Fix gcc-4.6 warnings with valgrind.h
992273318 amd64->IR: 0x66 0xF 0x3A 0x61 0xC1 0x38 (missing PCMPxSTRx case)
993273318 unhandled PCMPxSTRx case: vex amd64->IR: 0x66 0xF 0x3A 0x61 0xC1 0x38
994273431 valgrind segfaults in evalCfiExpr (debuginfo.c:2039)
995273465 Callgrind: jumps.c:164 (new_jcc): Assertion '(0 <= jmp) && ...'
996273536 Build error: multiple definition of `vgDrd_pthread_cond_initializer'
997273640 ppc64-linux: unhandled syscalls setresuid(164) and setresgid(169)
998273729 == 283000 (Illegal opcode for SSE2 "roundsd" instruction)
999273778 exp-ptrcheck: unhandled sysno == 259
1000274089 exp-ptrcheck: unhandled sysno == 208
1001274378 s390x: Various dispatcher tweaks
1002274447 WARNING: unhandled syscall: 340
1003274776 amd64->IR: 0x66 0xF 0x38 0x2B 0xC5 0x66
1004274784 == 267997 (valgrind ls -l results in Segmentation Fault)
1005274926 valgrind does not build against linux-3
1006275148 configure FAIL with glibc-2.14
1007275151 Fedora 15 / glibc-2.14 'make regtest' FAIL
1008275168 Make Valgrind work for MacOSX 10.7 Lion
1009275212 == 275284 (lots of false positives from __memcpy_ssse3_back et al)
1010275278 valgrind does not build on Linux kernel 3.0.* due to silly
1011275284 Valgrind memcpy/memmove redirection stopped working in glibc 2.14/x86_64
1012275308 Fix implementation for ppc64 fres instruc
1013275339 s390x: fix testcase compile warnings
1014275517 s390x: Provide support for CKSM instruction
1015275710 s390x: get rid of redundant address mode calculation
1016275815 == 247894 (Valgrind doesn't know about Linux readahead(2) syscall)
1017275852 == 250101 (valgrind uses all swap space and is killed)
1018276784 Add support for IBM Power ISA 2.06 -- stage 3
1019276987 gdbsrv: fix tests following recent commits
1020277045 Valgrind crashes with unhandled DW_OP_ opcode 0x2a
1021277199 The test_isa_2_06_part1.c in none/tests/ppc64 should be a symlink
1022277471 Unhandled syscall: 340
1023277610 valgrind crashes in VG_(lseek)(core_fd, phdrs[idx].p_offset, ...)
1024277653 ARM: support Thumb2 PLD instruction
1025277663 ARM: NEON float VMUL by scalar incorrect
1026277689 ARM: tests for VSTn with register post-index are broken
1027277694 ARM: BLX LR instruction broken in ARM mode
1028277780 ARM: VMOV.F32 (immediate) instruction is broken
1029278057 fuse filesystem syscall deadlocks
1030278078 Unimplemented syscall 280 on ppc32
1031278349 F_GETPIPE_SZ and F_SETPIPE_SZ Linux fcntl commands
1032278454 VALGRIND_STACK_DEREGISTER has wrong output type
1033278502 == 275284 (Valgrind confuses memcpy() and memmove())
sewardj3bbaf482011-10-24 08:54:56 +00001034278892 gdbsrv: factorize gdb version handling, fix doc and typos
sewardj520501d2011-10-23 09:09:54 +00001035279027 Support for MVCL and CLCL instruction
1036279027 s390x: Provide support for CLCL and MVCL instructions
1037279062 Remove a redundant check in the insn selector for ppc.
1038279071 JDK creates PTEST with redundant REX.W prefix
1039279212 gdbsrv: add monitor cmd v.info scheduler.
1040279378 exp-ptrcheck: the 'impossible' happened on mkfifo call
1041279698 memcheck discards valid-bits for packuswb
1042279795 memcheck reports uninitialised values for mincore on amd64
1043279994 Add support for IBM Power ISA 2.06 -- stage 3
1044280083 mempolicy syscall check errors
1045280290 vex amd64->IR: 0x66 0xF 0x38 0x28 0xC1 0x66 0xF 0x6F
1046280710 s390x: config files for nightly builds
1047280757 /tmp dir still used by valgrind even if TMPDIR is specified
sewardj3bbaf482011-10-24 08:54:56 +00001048280965 Valgrind breaks fcntl locks when program does mmap
sewardj520501d2011-10-23 09:09:54 +00001049281138 WARNING: unhandled syscall: 340
1050281241 == 275168 (valgrind useless on Macos 10.7.1 Lion)
1051281304 == 275168 (Darwin: dyld "cannot load inserted library")
1052281305 == 275168 (unhandled syscall: unix:357 on Darwin 11.1)
1053281468 s390x: handle do_clone and gcc clones in call traces
1054281488 ARM: VFP register corruption
1055281828 == 275284 (false memmove warning: "Source and destination overlap")
1056281883 s390x: Fix system call wrapper for "clone".
1057282105 generalise 'reclaimSuperBlock' to also reclaim splittable superblock
1058282112 Unhandled instruction bytes: 0xDE 0xD9 0x9B 0xDF (fcompp)
1059282238 SLES10: make check fails
1060282979 strcasestr needs replacement with recent(>=2.12) glibc
1061283000 vex amd64->IR: 0x66 0xF 0x3A 0xA 0xC0 0x9 0xF3 0xF
1062283243 Regression in ppc64 memcheck tests
1063283325 == 267997 (Darwin: V segfaults on startup when built with Xcode 4.0)
1064283427 re-connect epoll_pwait syscall on ARM linux
1065283600 gdbsrv: android: port vgdb.c
1066283709 none/tests/faultstatus needs to account for page size
1067284305 filter_gdb needs enhancement to work on ppc64
sewardjc2f1d962011-10-25 09:05:53 +00001068284384 clang 3.1 -Wunused-value warnings in valgrind.h, memcheck.h
sewardj520501d2011-10-23 09:09:54 +00001069284472 Thumb2 ROR.W encoding T2 not implemented
sewardj3bbaf482011-10-24 08:54:56 +00001070284621 XML-escape process command line in XML output
sewardj520501d2011-10-23 09:09:54 +00001071n-i-bz cachegrind/callgrind: handle CPUID information for Core iX Intel CPUs
1072 that have non-power-of-2 sizes (also AMDs)
1073n-i-bz don't be spooked by libraries mashed by elfhack
1074n-i-bz don't be spooked by libxul.so linked with gold
sewardj3bbaf482011-10-24 08:54:56 +00001075n-i-bz improved checking for VALGRIND_CHECK_MEM_IS_DEFINED
sewardj7a9ff4c2011-06-13 13:36:59 +00001076
sewardj62b0d402011-11-09 23:42:41 +00001077(3.7.0-TEST1: 27 October 2011, vex r2228, valgrind r12245)
1078(3.7.0.RC1: 1 November 2011, vex r2231, valgrind r12257)
1079(3.7.0: 5 November 2011, vex r2231, valgrind r12258)
sewardj469b8362011-10-23 10:05:47 +00001080
florian0c865b42011-09-30 01:27:26 +00001081
njne5930da2010-12-17 00:45:19 +00001082
sewardj4fd57292011-02-17 12:20:19 +00001083Release 3.6.1 (16 February 2011)
1084~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
10853.6.1 is a bug fix release. It adds support for some SSE4
1086instructions that were omitted in 3.6.0 due to lack of time. Initial
1087support for glibc-2.13 has been added. A number of bugs causing
1088crashing or assertion failures have been fixed.
njne5930da2010-12-17 00:45:19 +00001089
sewardj4fd57292011-02-17 12:20:19 +00001090The following bugs have been fixed or resolved. Note that "n-i-bz"
1091stands for "not in bugzilla" -- that is, a bug that was reported to us
1092but never got a bugzilla entry. We encourage you to file bugs in
1093bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
1094mailing the developers (or mailing lists) directly -- bugs that are
1095not entered into bugzilla tend to get forgotten about or ignored.
bart8e0d2c12011-02-04 19:07:11 +00001096
sewardj4fd57292011-02-17 12:20:19 +00001097To see details of a given bug, visit
1098https://bugs.kde.org/show_bug.cgi?id=XXXXXX
1099where XXXXXX is the bug number as listed below.
1100
1101188572 Valgrind on Mac should suppress setenv() mem leak
1102194402 vex amd64->IR: 0x48 0xF 0xAE 0x4 (proper FX{SAVE,RSTOR} support)
1103210481 vex amd64->IR: Assertion `sz == 2 || sz == 4' failed (REX.W POPQ)
1104246152 callgrind internal error after pthread_cancel on 32 Bit Linux
1105250038 ppc64: Altivec LVSR and LVSL instructions fail their regtest
1106254420 memory pool tracking broken
1107254957 Test code failing to compile due to changes in memcheck.h
1108255009 helgrind/drd: crash on chmod with invalid parameter
1109255130 readdwarf3.c parse_type_DIE confused by GNAT Ada types
1110255355 helgrind/drd: crash on threaded programs doing fork
1111255358 == 255355
1112255418 (SSE4.x) rint call compiled with ICC
1113255822 --gen-suppressions can create invalid files: "too many callers [...]"
1114255888 closing valgrindoutput tag outputted to log-stream on error
1115255963 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x9 0xDB 0x0 (ROUNDPD)
1116255966 Slowness when using mempool annotations
1117256387 vex x86->IR: 0xD4 0xA 0x2 0x7 (AAD and AAM)
1118256600 super-optimized strcasecmp() false positive
1119256669 vex amd64->IR: Unhandled LOOPNEL insn on amd64
1120256968 (SSE4.x) vex amd64->IR: 0x66 0xF 0x38 0x10 0xD3 0x66 (BLENDVPx)
1121257011 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0xE 0xFD 0xA0 (PBLENDW)
1122257063 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x8 0xC0 0x0 (ROUNDPS)
1123257276 Missing case in memcheck --track-origins=yes
1124258870 (SSE4.x) Add support for EXTRACTPS SSE 4.1 instruction
1125261966 (SSE4.x) support for CRC32B and CRC32Q is lacking (also CRC32{W,L})
1126262985 VEX regression in valgrind 3.6.0 in handling PowerPC VMX
1127262995 (SSE4.x) crash when trying to valgrind gcc-snapshot (PCMPxSTRx $0)
1128263099 callgrind_annotate counts Ir improperly [...]
1129263877 undefined coprocessor instruction on ARMv7
1130265964 configure FAIL with glibc-2.13
1131n-i-bz Fix compile error w/ icc-12.x in guest_arm_toIR.c
1132n-i-bz Docs: fix bogus descriptions for VALGRIND_CREATE_BLOCK et al
1133n-i-bz Massif: don't assert on shmat() with --pages-as-heap=yes
1134n-i-bz Bug fixes and major speedups for the exp-DHAT space profiler
1135n-i-bz DRD: disable --free-is-write due to implementation difficulties
1136
1137(3.6.1: 16 February 2011, vex r2103, valgrind r11561).
1138
bart8e0d2c12011-02-04 19:07:11 +00001139
1140
sewardj695f13a2010-10-20 17:07:01 +00001141Release 3.6.0 (21 October 2010)
sewardj18a08922010-10-14 00:26:42 +00001142~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
11433.6.0 is a feature release with many significant improvements and the
1144usual collection of bug fixes.
sewardj00d9d1e2010-10-13 14:05:30 +00001145
sewardj18a08922010-10-14 00:26:42 +00001146This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux,
1147PPC64/Linux, X86/Darwin and AMD64/Darwin. Support for recent distros
1148and toolchain components (glibc 2.12, gcc 4.5, OSX 10.6) has been added.
sewardj00d9d1e2010-10-13 14:05:30 +00001149
sewardj18a08922010-10-14 00:26:42 +00001150 -------------------------
sewardj00d9d1e2010-10-13 14:05:30 +00001151
sewardj18a08922010-10-14 00:26:42 +00001152Here are some highlights. Details are shown further down:
njn9b919b62010-02-28 23:41:07 +00001153
sewardj18a08922010-10-14 00:26:42 +00001154* Support for ARM/Linux.
njn9b919b62010-02-28 23:41:07 +00001155
sewardj18a08922010-10-14 00:26:42 +00001156* Support for recent Linux distros: Ubuntu 10.10 and Fedora 14.
1157
1158* Support for Mac OS X 10.6, both 32- and 64-bit executables.
1159
1160* Support for the SSE4.2 instruction set.
1161
1162* Enhancements to the Callgrind profiler, including the ability to
1163 handle CPUs with three levels of cache.
1164
1165* A new experimental heap profiler, DHAT.
1166
sewardj695f13a2010-10-20 17:07:01 +00001167* A huge number of bug fixes and small enhancements.
sewardj18a08922010-10-14 00:26:42 +00001168
1169 -------------------------
1170
1171Here are details of the above changes, together with descriptions of
1172many other changes, and a list of fixed bugs.
1173
1174* ================== PLATFORM CHANGES =================
1175
1176* Support for ARM/Linux. Valgrind now runs on ARMv7 capable CPUs
1177 running Linux. It is known to work on Ubuntu 10.04, Ubuntu 10.10,
1178 and Maemo 5, so you can run Valgrind on your Nokia N900 if you want.
1179
1180 This requires a CPU capable of running the ARMv7-A instruction set
1181 (Cortex A5, A8 and A9). Valgrind provides fairly complete coverage
1182 of the user space instruction set, including ARM and Thumb integer
1183 code, VFPv3, NEON and V6 media instructions. The Memcheck,
1184 Cachegrind and Massif tools work properly; other tools work to
1185 varying degrees.
1186
1187* Support for recent Linux distros (Ubuntu 10.10 and Fedora 14), along
1188 with support for recent releases of the underlying toolchain
1189 components, notably gcc-4.5 and glibc-2.12.
1190
1191* Support for Mac OS X 10.6, both 32- and 64-bit executables. 64-bit
1192 support also works much better on OS X 10.5, and is as solid as
1193 32-bit support now.
1194
1195* Support for the SSE4.2 instruction set. SSE4.2 is supported in
1196 64-bit mode. In 32-bit mode, support is only available up to and
1197 including SSSE3. Some exceptions: SSE4.2 AES instructions are not
1198 supported in 64-bit mode, and 32-bit mode does in fact support the
sewardj695f13a2010-10-20 17:07:01 +00001199 bare minimum SSE4 instructions to needed to run programs on Mac OS X
sewardj18a08922010-10-14 00:26:42 +00001200 10.6 on 32-bit targets.
1201
1202* Support for IBM POWER6 cpus has been improved. The Power ISA up to
1203 and including version 2.05 is supported.
1204
1205* ==================== TOOL CHANGES ====================
1206
1207* Cachegrind has a new processing script, cg_diff, which finds the
1208 difference between two profiles. It's very useful for evaluating
1209 the performance effects of a change in a program.
njn69d495d2010-06-30 05:23:34 +00001210
1211 Related to this change, the meaning of cg_annotate's (rarely-used)
sewardj18a08922010-10-14 00:26:42 +00001212 --threshold option has changed; this is unlikely to affect many
1213 people, if you do use it please see the user manual for details.
njn9b919b62010-02-28 23:41:07 +00001214
sewardj18a08922010-10-14 00:26:42 +00001215* Callgrind now can do branch prediction simulation, similar to
1216 Cachegrind. In addition, it optionally can count the number of
1217 executed global bus events. Both can be used for a better
1218 approximation of a "Cycle Estimation" as derived event (you need to
1219 update the event formula in KCachegrind yourself).
njn2d853a12010-10-06 22:46:31 +00001220
sewardj18a08922010-10-14 00:26:42 +00001221* Cachegrind and Callgrind now refer to the LL (last-level) cache
1222 rather than the L2 cache. This is to accommodate machines with
1223 three levels of caches -- if Cachegrind/Callgrind auto-detects the
1224 cache configuration of such a machine it will run the simulation as
1225 if the L2 cache isn't present. This means the results are less
1226 likely to match the true result for the machine, but
1227 Cachegrind/Callgrind's results are already only approximate, and
1228 should not be considered authoritative. The results are still
1229 useful for giving a general idea about a program's locality.
njn2d853a12010-10-06 22:46:31 +00001230
sewardj695f13a2010-10-20 17:07:01 +00001231* Massif has a new option, --pages-as-heap, which is disabled by
sewardj18a08922010-10-14 00:26:42 +00001232 default. When enabled, instead of tracking allocations at the level
1233 of heap blocks (as allocated with malloc/new/new[]), it instead
1234 tracks memory allocations at the level of memory pages (as mapped by
1235 mmap, brk, etc). Each mapped page is treated as its own block.
1236 Interpreting the page-level output is harder than the heap-level
1237 output, but this option is useful if you want to account for every
1238 byte of memory used by a program.
njne323a6b2010-07-01 02:35:03 +00001239
sewardj18a08922010-10-14 00:26:42 +00001240* DRD has two new command-line options: --free-is-write and
1241 --trace-alloc. The former allows to detect reading from already freed
1242 memory, and the latter allows tracing of all memory allocations and
bart5cf4c392010-10-10 18:54:44 +00001243 deallocations.
1244
sewardj18a08922010-10-14 00:26:42 +00001245* DRD has several new annotations. Custom barrier implementations can
1246 now be annotated, as well as benign races on static variables.
bart5cf4c392010-10-10 18:54:44 +00001247
sewardj18a08922010-10-14 00:26:42 +00001248* DRD's happens before / happens after annotations have been made more
1249 powerful, so that they can now also be used to annotate e.g. a smart
bart5cf4c392010-10-10 18:54:44 +00001250 pointer implementation.
1251
sewardj18a08922010-10-14 00:26:42 +00001252* Helgrind's annotation set has also been drastically improved, so as
sewardj695f13a2010-10-20 17:07:01 +00001253 to provide to users a general set of annotations to describe locks,
sewardj18a08922010-10-14 00:26:42 +00001254 semaphores, barriers and condition variables. Annotations to
1255 describe thread-safe reference counted heap objects have also been
1256 added.
1257
1258* Memcheck has a new command-line option, --show-possibly-lost, which
1259 is enabled by default. When disabled, the leak detector will not
1260 show possibly-lost blocks.
1261
1262* A new experimental heap profiler, DHAT (Dynamic Heap Analysis Tool),
1263 has been added. DHAT keeps track of allocated heap blocks, and also
1264 inspects every memory reference to see which block (if any) is being
1265 accessed. This gives a lot of insight into block lifetimes,
1266 utilisation, turnover, liveness, and the location of hot and cold
1267 fields. You can use DHAT to do hot-field profiling.
1268
1269* ==================== OTHER CHANGES ====================
1270
1271* Improved support for unfriendly self-modifying code: the extra
1272 overhead incurred by --smc-check=all has been reduced by
1273 approximately a factor of 5 as compared with 3.5.0.
1274
1275* Ability to show directory names for source files in error messages.
sewardj695f13a2010-10-20 17:07:01 +00001276 This is combined with a flexible mechanism for specifying which
1277 parts of the paths should be shown. This is enabled by the new flag
1278 --fullpath-after.
sewardj18a08922010-10-14 00:26:42 +00001279
1280* A new flag, --require-text-symbol, which will stop the run if a
1281 specified symbol is not found it a given shared object when it is
sewardj695f13a2010-10-20 17:07:01 +00001282 loaded into the process. This makes advanced working with function
1283 intercepting and wrapping safer and more reliable.
sewardj18a08922010-10-14 00:26:42 +00001284
sewardj695f13a2010-10-20 17:07:01 +00001285* Improved support for the Valkyrie GUI, version 2.0.0. GUI output
1286 and control of Valgrind is now available for the tools Memcheck and
1287 Helgrind. XML output from Valgrind is available for Memcheck,
1288 Helgrind and exp-Ptrcheck.
sewardj18a08922010-10-14 00:26:42 +00001289
1290* More reliable stack unwinding on amd64-linux, particularly in the
1291 presence of function wrappers, and with gcc-4.5 compiled code.
1292
1293* Modest scalability (performance improvements) for massive
1294 long-running applications, particularly for those with huge amounts
1295 of code.
1296
1297* Support for analyzing programs running under Wine with has been
1298 improved. The header files <valgrind/valgrind.h>,
1299 <valgrind/memcheck.h> and <valgrind/drd.h> can now be used in
1300 Windows-programs compiled with MinGW or one of the Microsoft Visual
1301 Studio compilers.
1302
1303* A rare but serious error in the 64-bit x86 CPU simulation was fixed.
1304 The 32-bit simulator was not affected. This did not occur often,
1305 but when it did would usually crash the program under test.
1306 Bug 245925.
1307
1308* A large number of bugs were fixed. These are shown below.
1309
1310* A number of bugs were investigated, and were candidates for fixing,
1311 but are not fixed in 3.6.0, due to lack of developer time. They may
1312 get fixed in later releases. They are:
1313
1314 194402 vex amd64->IR: 0x48 0xF 0xAE 0x4 0x24 0x49 (FXSAVE64)
1315 212419 false positive "lock order violated" (A+B vs A)
1316 213685 Undefined value propagates past dependency breaking instruction
1317 216837 Incorrect instrumentation of NSOperationQueue on Darwin
1318 237920 valgrind segfault on fork failure
1319 242137 support for code compiled by LLVM-2.8
1320 242423 Another unknown Intel cache config value
1321 243232 Inconsistent Lock Orderings report with trylock
1322 243483 ppc: callgrind triggers VEX assertion failure
1323 243935 Helgrind: implementation of ANNOTATE_HAPPENS_BEFORE() is wrong
1324 244677 Helgrind crash hg_main.c:616 (map_threads_lookup): Assertion
1325 'thr' failed.
1326 246152 callgrind internal error after pthread_cancel on 32 Bit Linux
1327 249435 Analyzing wine programs with callgrind triggers a crash
1328 250038 ppc64: Altivec lvsr and lvsl instructions fail their regtest
1329 250065 Handling large allocations
1330 250101 huge "free" memory usage due to m_mallocfree.c
1331 "superblocks fragmentation"
1332 251569 vex amd64->IR: 0xF 0x1 0xF9 0x8B 0x4C 0x24 (RDTSCP)
sewardj695f13a2010-10-20 17:07:01 +00001333 252091 Callgrind on ARM does not detect function returns correctly
1334 252600 [PATCH] Allow lhs to be a pointer for shl/shr
1335 254420 memory pool tracking broken
sewardj18a08922010-10-14 00:26:42 +00001336 n-i-bz support for adding symbols for JIT generated code
1337
1338
1339The following bugs have been fixed or resolved. Note that "n-i-bz"
1340stands for "not in bugzilla" -- that is, a bug that was reported to us
1341but never got a bugzilla entry. We encourage you to file bugs in
1342bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
1343mailing the developers (or mailing lists) directly -- bugs that are
1344not entered into bugzilla tend to get forgotten about or ignored.
1345
1346To see details of a given bug, visit
1347https://bugs.kde.org/show_bug.cgi?id=XXXXXX
1348where XXXXXX is the bug number as listed below.
1349
1350135264 dcbzl instruction missing
1351142688 == 250799
1352153699 Valgrind should report unaligned reads with movdqa
1353180217 == 212335
1354190429 Valgrind reports lost of errors in ld.so
1355 with x86_64 2.9.90 glibc
1356197266 valgrind appears to choke on the xmms instruction
1357 "roundsd" on x86_64
1358197988 Crash when demangling very large symbol names
1359202315 unhandled syscall: 332 (inotify_init1)
1360203256 Add page-level profiling to Massif
1361205093 dsymutil=yes needs quotes, locking (partial fix)
1362205241 Snow Leopard 10.6 support (partial fix)
1363206600 Leak checker fails to upgrade indirect blocks when their
1364 parent becomes reachable
1365210935 port valgrind.h (not valgrind) to win32 so apps run under
1366 wine can make client requests
1367211410 vex amd64->IR: 0x15 0xFF 0xFF 0x0 0x0 0x89
1368 within Linux ip-stack checksum functions
1369212335 unhandled instruction bytes: 0xF3 0xF 0xBD 0xC0
1370 (lzcnt %eax,%eax)
1371213685 Undefined value propagates past dependency breaking instruction
1372 (partial fix)
1373215914 Valgrind inserts bogus empty environment variable
1374217863 == 197988
1375219538 adjtimex syscall wrapper wrong in readonly adjtime mode
1376222545 shmat fails under valgind on some arm targets
1377222560 ARM NEON support
1378230407 == 202315
1379231076 == 202315
1380232509 Docs build fails with formatting inside <title></title> elements
1381232793 == 202315
1382235642 [PATCH] syswrap-linux.c: support evdev EVIOCG* ioctls
1383236546 vex x86->IR: 0x66 0xF 0x3A 0xA
1384237202 vex amd64->IR: 0xF3 0xF 0xB8 0xC0 0x49 0x3B
1385237371 better support for VALGRIND_MALLOCLIKE_BLOCK
1386237485 symlink (syscall 57) is not supported on Mac OS
1387237723 sysno == 101 exp-ptrcheck: the 'impossible' happened:
1388 unhandled syscall
1389238208 is_just_below_ESP doesn't take into account red-zone
1390238345 valgrind passes wrong $0 when executing a shell script
1391238679 mq_timedreceive syscall doesn't flag the reception buffer
1392 as "defined"
1393238696 fcntl command F_DUPFD_CLOEXEC not supported
1394238713 unhandled instruction bytes: 0x66 0xF 0x29 0xC6
1395238713 unhandled instruction bytes: 0x66 0xF 0x29 0xC6
1396238745 3.5.0 Make fails on PPC Altivec opcodes, though configure
1397 says "Altivec off"
1398239992 vex amd64->IR: 0x48 0xF 0xC4 0xC1 0x0 0x48
1399240488 == 197988
1400240639 == 212335
1401241377 == 236546
1402241903 == 202315
1403241920 == 212335
1404242606 unhandled syscall: setegid (in Ptrcheck)
1405242814 Helgrind "Impossible has happened" during
1406 QApplication::initInstance();
1407243064 Valgrind attempting to read debug information from iso
1408243270 Make stack unwinding in Valgrind wrappers more reliable
1409243884 exp-ptrcheck: the 'impossible happened: unhandled syscall
1410 sysno = 277 (mq_open)
1411244009 exp-ptrcheck unknown syscalls in analyzing lighttpd
1412244493 ARM VFP d16-d31 registers support
1413244670 add support for audit_session_self syscall on Mac OS 10.6
1414244921 The xml report of helgrind tool is not well format
1415244923 In the xml report file, the <preamble> not escape the
1416 xml char, eg '<','&','>'
1417245535 print full path names in plain text reports
1418245925 x86-64 red zone handling problem
1419246258 Valgrind not catching integer underruns + new [] s
1420246311 reg/reg cmpxchg doesn't work on amd64
1421246549 unhandled syscall unix:277 while testing 32-bit Darwin app
1422246888 Improve Makefile.vex.am
1423247510 [OS X 10.6] Memcheck reports unaddressable bytes passed
1424 to [f]chmod_extended
1425247526 IBM POWER6 (ISA 2.05) support is incomplete
1426247561 Some leak testcases fails due to reachable addresses in
1427 caller save regs
1428247875 sizeofIRType to handle Ity_I128
1429247894 [PATCH] unhandled syscall sys_readahead
1430247980 Doesn't honor CFLAGS passed to configure
1431248373 darwin10.supp is empty in the trunk
1432248822 Linux FIBMAP ioctl has int parameter instead of long
1433248893 [PATCH] make readdwarf.c big endianess safe to enable
1434 unwinding on big endian systems
1435249224 Syscall 336 not supported (SYS_proc_info)
1436249359 == 245535
1437249775 Incorrect scheme for detecting NEON capabilities of host CPU
1438249943 jni JVM init fails when using valgrind
1439249991 Valgrind incorrectly declares AESKEYGENASSIST support
1440 since VEX r2011
1441249996 linux/arm: unhandled syscall: 181 (__NR_pwrite64)
1442250799 frexp$fenv_access_off function generates SIGILL
1443250998 vex x86->IR: unhandled instruction bytes: 0x66 0x66 0x66 0x2E
1444251251 support pclmulqdq insn
1445251362 valgrind: ARM: attach to debugger either fails or provokes
1446 kernel oops
1447251674 Unhandled syscall 294
sewardj695f13a2010-10-20 17:07:01 +00001448251818 == 254550
sewardj18a08922010-10-14 00:26:42 +00001449
sewardj695f13a2010-10-20 17:07:01 +00001450254257 Add support for debugfiles found by build-id
1451254550 [PATCH] Implement DW_ATE_UTF (DWARF4)
1452254646 Wrapped functions cause stack misalignment on OS X
1453 (and possibly Linux)
1454254556 ARM: valgrinding anything fails with SIGSEGV for 0xFFFF0FA0
1455
sewardje07c2b52010-10-22 06:44:59 +00001456(3.6.0: 21 October 2010, vex r2068, valgrind r11471).
sewardj18a08922010-10-14 00:26:42 +00001457
1458
njn9b919b62010-02-28 23:41:07 +00001459
sewardj6bbe0242009-08-19 11:38:54 +00001460Release 3.5.0 (19 August 2009)
1461~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardj570e2442009-08-14 13:42:41 +000014623.5.0 is a feature release with many significant improvements and the
1463usual collection of bug fixes. The main improvement is that Valgrind
1464now works on Mac OS X.
1465
1466This release supports X86/Linux, AMD64/Linux, PPC32/Linux, PPC64/Linux
1467and X86/Darwin. Support for recent distros and toolchain components
1468(glibc 2.10, gcc 4.5) has been added.
1469
1470 -------------------------
1471
1472Here is a short summary of the changes. Details are shown further
1473down:
1474
njnf95c30b2009-08-16 00:20:58 +00001475* Support for Mac OS X (10.5.x).
sewardj570e2442009-08-14 13:42:41 +00001476
njn78b7bb32009-08-14 22:34:30 +00001477* Improvements and simplifications to Memcheck's leak checker.
sewardj570e2442009-08-14 13:42:41 +00001478
njn78b7bb32009-08-14 22:34:30 +00001479* Clarification and simplifications in various aspects of Valgrind's
1480 text output.
sewardj570e2442009-08-14 13:42:41 +00001481
njn78b7bb32009-08-14 22:34:30 +00001482* XML output for Helgrind and Ptrcheck.
sewardj570e2442009-08-14 13:42:41 +00001483
njn78b7bb32009-08-14 22:34:30 +00001484* Performance and stability improvements for Helgrind and DRD.
sewardj570e2442009-08-14 13:42:41 +00001485
njn78b7bb32009-08-14 22:34:30 +00001486* Genuinely atomic support for x86/amd64/ppc atomic instructions.
sewardj570e2442009-08-14 13:42:41 +00001487
njn78b7bb32009-08-14 22:34:30 +00001488* A new experimental tool, BBV, useful for computer architecture
1489 research.
sewardj570e2442009-08-14 13:42:41 +00001490
njn78b7bb32009-08-14 22:34:30 +00001491* Improved Wine support, including ability to read Windows PDB
1492 debuginfo.
sewardj570e2442009-08-14 13:42:41 +00001493
1494 -------------------------
1495
sewardj9176c862009-08-18 13:47:31 +00001496Here are details of the above changes, followed by descriptions of
1497many other minor changes, and a list of fixed bugs.
sewardj570e2442009-08-14 13:42:41 +00001498
1499
1500* Valgrind now runs on Mac OS X. (Note that Mac OS X is sometimes
njnf95c30b2009-08-16 00:20:58 +00001501 called "Darwin" because that is the name of the OS core, which is the
1502 level that Valgrind works at.)
njnf76d27a2009-05-28 01:53:07 +00001503
njnf95c30b2009-08-16 00:20:58 +00001504 Supported systems:
njn6eccab92009-08-14 02:41:37 +00001505
njnf95c30b2009-08-16 00:20:58 +00001506 - It requires OS 10.5.x (Leopard). Porting to 10.4.x is not planned
1507 because it would require work and 10.4 is only becoming less common.
njn6eccab92009-08-14 02:41:37 +00001508
njnf95c30b2009-08-16 00:20:58 +00001509 - 32-bit programs on x86 and AMD64 (a.k.a x86-64) machines are supported
1510 fairly well. For 10.5.x, 32-bit programs are the default even on
1511 64-bit machines, so it handles most current programs.
1512
1513 - 64-bit programs on x86 and AMD64 (a.k.a x86-64) machines are not
1514 officially supported, but simple programs at least will probably work.
1515 However, start-up is slow.
njn6eccab92009-08-14 02:41:37 +00001516
sewardj570e2442009-08-14 13:42:41 +00001517 - PowerPC machines are not supported.
njn6eccab92009-08-14 02:41:37 +00001518
njnc6ab7b62009-06-04 22:47:21 +00001519 Things that don't work:
njn6eccab92009-08-14 02:41:37 +00001520
1521 - The Ptrcheck tool.
1522
1523 - Objective-C garbage collection.
1524
1525 - --db-attach=yes.
1526
sewardj570e2442009-08-14 13:42:41 +00001527 - If you have Rogue Amoeba's "Instant Hijack" program installed,
1528 Valgrind will fail with a SIGTRAP at start-up. See
1529 https://bugs.kde.org/show_bug.cgi?id=193917 for details and a
1530 simple work-around.
njnc6ab7b62009-06-04 22:47:21 +00001531
njn97db7612009-08-04 02:32:55 +00001532 Usage notes:
njn6eccab92009-08-14 02:41:37 +00001533
sewardj570e2442009-08-14 13:42:41 +00001534 - You will likely find --dsymutil=yes a useful option, as error
1535 messages may be imprecise without it.
njn6eccab92009-08-14 02:41:37 +00001536
sewardj570e2442009-08-14 13:42:41 +00001537 - Mac OS X support is new and therefore will be less robust than the
njn1205e022009-08-14 02:12:54 +00001538 Linux support. Please report any bugs you find.
njn97db7612009-08-04 02:32:55 +00001539
sewardj570e2442009-08-14 13:42:41 +00001540 - Threaded programs may run more slowly than on Linux.
1541
njnc6ab7b62009-06-04 22:47:21 +00001542 Many thanks to Greg Parker for developing this port over several years.
1543
sewardj570e2442009-08-14 13:42:41 +00001544
1545* Memcheck's leak checker has been improved.
1546
1547 - The results for --leak-check=summary now match the summary results
1548 for --leak-check=full. Previously they could differ because
1549 --leak-check=summary counted "indirectly lost" blocks and
1550 "suppressed" blocks as "definitely lost".
1551
1552 - Blocks that are only reachable via at least one interior-pointer,
1553 but are directly pointed to by a start-pointer, were previously
1554 marked as "still reachable". They are now correctly marked as
1555 "possibly lost".
1556
1557 - The default value for the --leak-resolution option has been
1558 changed from "low" to "high". In general, this means that more
1559 leak reports will be produced, but each leak report will describe
1560 fewer leaked blocks.
1561
1562 - With --leak-check=full, "definitely lost" and "possibly lost"
1563 leaks are now considered as proper errors, ie. they are counted
1564 for the "ERROR SUMMARY" and affect the behaviour of
1565 --error-exitcode. These leaks are not counted as errors if
1566 --leak-check=summary is specified, however.
1567
1568 - Documentation for the leak checker has been improved.
1569
njnf76d27a2009-05-28 01:53:07 +00001570
njn6eccab92009-08-14 02:41:37 +00001571* Various aspects of Valgrind's text output have changed.
njndbebecc2009-07-14 01:39:54 +00001572
njn6eccab92009-08-14 02:41:37 +00001573 - Valgrind's start-up message has changed. It is shorter but also
1574 includes the command being run, which makes it easier to use
1575 --trace-children=yes. An example:
njn53162bf2009-07-29 23:34:49 +00001576
njn6eccab92009-08-14 02:41:37 +00001577 - Valgrind's shut-down messages have also changed. This is most
sewardj570e2442009-08-14 13:42:41 +00001578 noticeable with Memcheck, where the leak summary now occurs before
1579 the error summary. This change was necessary to allow leaks to be
1580 counted as proper errors (see the description of the leak checker
1581 changes above for more details). This was also necessary to fix a
1582 longstanding bug in which uses of suppressions against leaks were
1583 not "counted", leading to difficulties in maintaining suppression
sewardj783e9d62009-08-19 22:18:35 +00001584 files (see https://bugs.kde.org/show_bug.cgi?id=186790).
njn6eccab92009-08-14 02:41:37 +00001585
sewardj570e2442009-08-14 13:42:41 +00001586 - Behavior of -v has changed. In previous versions, -v printed out
1587 a mixture of marginally-user-useful information, and tool/core
1588 statistics. The statistics printing has now been moved to its own
1589 flag, --stats=yes. This means -v is less verbose and more likely
1590 to convey useful end-user information.
njn6eccab92009-08-14 02:41:37 +00001591
sewardj570e2442009-08-14 13:42:41 +00001592 - The format of some (non-XML) stack trace entries has changed a
1593 little. Previously there were six possible forms:
njn6eccab92009-08-14 02:41:37 +00001594
1595 0x80483BF: really (a.c:20)
1596 0x80483BF: really (in /foo/a.out)
1597 0x80483BF: really
1598 0x80483BF: (within /foo/a.out)
1599 0x80483BF: ??? (a.c:20)
1600 0x80483BF: ???
1601
sewardj570e2442009-08-14 13:42:41 +00001602 The third and fourth of these forms have been made more consistent
1603 with the others. The six possible forms are now:
njn6eccab92009-08-14 02:41:37 +00001604
1605 0x80483BF: really (a.c:20)
1606 0x80483BF: really (in /foo/a.out)
1607 0x80483BF: really (in ???)
1608 0x80483BF: ??? (in /foo/a.out)
1609 0x80483BF: ??? (a.c:20)
1610 0x80483BF: ???
1611
sewardj570e2442009-08-14 13:42:41 +00001612 Stack traces produced when --xml=yes is specified are different
1613 and unchanged.
njn6eccab92009-08-14 02:41:37 +00001614
njn6eccab92009-08-14 02:41:37 +00001615
sewardj570e2442009-08-14 13:42:41 +00001616* Helgrind and Ptrcheck now support XML output, so they can be used
1617 from GUI tools. Also, the XML output mechanism has been
1618 overhauled.
njn6eccab92009-08-14 02:41:37 +00001619
njn78b7bb32009-08-14 22:34:30 +00001620 - The XML format has been overhauled and generalised, so it is more
sewardj570e2442009-08-14 13:42:41 +00001621 suitable for error reporting tools in general. The Memcheck
1622 specific aspects of it have been removed. The new format, which
1623 is an evolution of the old format, is described in
1624 docs/internals/xml-output-protocol4.txt.
njn6eccab92009-08-14 02:41:37 +00001625
sewardj570e2442009-08-14 13:42:41 +00001626 - Memcheck has been updated to use the new format.
njn6eccab92009-08-14 02:41:37 +00001627
sewardj570e2442009-08-14 13:42:41 +00001628 - Helgrind and Ptrcheck are now able to emit output in this format.
njn6eccab92009-08-14 02:41:37 +00001629
sewardj570e2442009-08-14 13:42:41 +00001630 - The XML output mechanism has been overhauled. XML is now output
1631 to its own file descriptor, which means that:
njn6eccab92009-08-14 02:41:37 +00001632
sewardj570e2442009-08-14 13:42:41 +00001633 * Valgrind can output text and XML independently.
njnb4e06872009-08-10 05:11:43 +00001634
sewardj570e2442009-08-14 13:42:41 +00001635 * The longstanding problem of XML output being corrupted by
1636 unexpected un-tagged text messages is solved.
njn8df80b22009-03-02 05:11:06 +00001637
sewardj570e2442009-08-14 13:42:41 +00001638 As before, the destination for text output is specified using
1639 --log-file=, --log-fd= or --log-socket=.
njnd55f0d92009-08-03 01:38:56 +00001640
sewardj570e2442009-08-14 13:42:41 +00001641 As before, XML output for a tool is enabled using --xml=yes.
njnc6ab7b62009-06-04 22:47:21 +00001642
sewardj570e2442009-08-14 13:42:41 +00001643 Because there's a new XML output channel, the XML output
1644 destination is now specified by --xml-file=, --xml-fd= or
1645 --xml-socket=.
njn6eccab92009-08-14 02:41:37 +00001646
sewardj570e2442009-08-14 13:42:41 +00001647 Initial feedback has shown this causes some confusion. To
1648 clarify, the two envisaged usage scenarios are:
njn6eccab92009-08-14 02:41:37 +00001649
sewardj570e2442009-08-14 13:42:41 +00001650 (1) Normal text output. In this case, do not specify --xml=yes
1651 nor any of --xml-file=, --xml-fd= or --xml-socket=.
1652
1653 (2) XML output. In this case, specify --xml=yes, and one of
1654 --xml-file=, --xml-fd= or --xml-socket= to select the XML
1655 destination, one of --log-file=, --log-fd= or --log-socket=
1656 to select the destination for any remaining text messages,
1657 and, importantly, -q.
1658
1659 -q makes Valgrind completely silent on the text channel,
1660 except in the case of critical failures, such as Valgrind
1661 itself segfaulting, or failing to read debugging information.
1662 Hence, in this scenario, it suffices to check whether or not
1663 any output appeared on the text channel. If yes, then it is
1664 likely to be a critical error which should be brought to the
1665 attention of the user. If no (the text channel produced no
1666 output) then it can be assumed that the run was successful.
1667
1668 This allows GUIs to make the critical distinction they need to
1669 make (did the run fail or not?) without having to search or
1670 filter the text output channel in any way.
1671
1672 It is also recommended to use --child-silent-after-fork=yes in
1673 scenario (2).
1674
1675
1676* Improvements and changes in Helgrind:
1677
1678 - XML output, as described above
1679
1680 - Checks for consistent association between pthread condition
1681 variables and their associated mutexes are now performed.
1682
1683 - pthread_spinlock functions are supported.
1684
1685 - Modest performance improvements.
1686
1687 - Initial (skeletal) support for describing the behaviour of
1688 non-POSIX synchronisation objects through ThreadSanitizer
1689 compatible ANNOTATE_* macros.
1690
1691 - More controllable tradeoffs between performance and the level of
1692 detail of "previous" accesses in a race. There are now three
1693 settings:
1694
1695 * --history-level=full. This is the default, and was also the
1696 default in 3.4.x. It shows both stacks involved in a race, but
1697 requires a lot of memory and can be very slow in programs that
1698 do many inter-thread synchronisation events.
1699
1700 * --history-level=none. This only shows the later stack involved
1701 in a race. This can be much faster than --history-level=full,
1702 but makes it much more difficult to find the other access
1703 involved in the race.
1704
1705 The new intermediate setting is
1706
1707 * --history-level=approx
1708
1709 For the earlier (other) access, two stacks are presented. The
1710 earlier access is guaranteed to be somewhere in between the two
1711 program points denoted by those stacks. This is not as useful
1712 as showing the exact stack for the previous access (as per
1713 --history-level=full), but it is better than nothing, and it's
1714 almost as fast as --history-level=none.
1715
njnace90d82009-04-27 01:27:58 +00001716
bart6a146fc2009-07-21 07:32:19 +00001717* New features and improvements in DRD:
njn6eccab92009-08-14 02:41:37 +00001718
sewardj570e2442009-08-14 13:42:41 +00001719 - The error messages printed by DRD are now easier to interpret.
1720 Instead of using two different numbers to identify each thread
1721 (Valgrind thread ID and DRD thread ID), DRD does now identify
1722 threads via a single number (the DRD thread ID). Furthermore
1723 "first observed at" information is now printed for all error
1724 messages related to synchronization objects.
njn6eccab92009-08-14 02:41:37 +00001725
bartfe212db2009-07-25 11:16:51 +00001726 - Added support for named semaphores (sem_open() and sem_close()).
njn6eccab92009-08-14 02:41:37 +00001727
bart6a146fc2009-07-21 07:32:19 +00001728 - Race conditions between pthread_barrier_wait() and
1729 pthread_barrier_destroy() calls are now reported.
njn6eccab92009-08-14 02:41:37 +00001730
1731 - Added support for custom allocators through the macros
sewardj570e2442009-08-14 13:42:41 +00001732 VALGRIND_MALLOCLIKE_BLOCK() VALGRIND_FREELIKE_BLOCK() (defined in
1733 in <valgrind/valgrind.h>). An alternative for these two macros is
1734 the new client request VG_USERREQ__DRD_CLEAN_MEMORY (defined in
bart6a146fc2009-07-21 07:32:19 +00001735 <valgrind/drd.h>).
njn6eccab92009-08-14 02:41:37 +00001736
sewardj570e2442009-08-14 13:42:41 +00001737 - Added support for annotating non-POSIX synchronization objects
1738 through several new ANNOTATE_*() macros.
njn6eccab92009-08-14 02:41:37 +00001739
sewardj570e2442009-08-14 13:42:41 +00001740 - OpenMP: added support for the OpenMP runtime (libgomp) included
1741 with gcc versions 4.3.0 and 4.4.0.
njn6eccab92009-08-14 02:41:37 +00001742
bart6a146fc2009-07-21 07:32:19 +00001743 - Faster operation.
njn6eccab92009-08-14 02:41:37 +00001744
1745 - Added two new command-line options (--first-race-only and
bart6a146fc2009-07-21 07:32:19 +00001746 --segment-merging-interval).
1747
njn6eccab92009-08-14 02:41:37 +00001748
sewardj570e2442009-08-14 13:42:41 +00001749* Genuinely atomic support for x86/amd64/ppc atomic instructions
1750
1751 Valgrind will now preserve (memory-access) atomicity of LOCK-
1752 prefixed x86/amd64 instructions, and any others implying a global
1753 bus lock. Ditto for PowerPC l{w,d}arx/st{w,d}cx. instructions.
1754
1755 This means that Valgrinded processes will "play nicely" in
1756 situations where communication with other processes, or the kernel,
1757 is done through shared memory and coordinated with such atomic
1758 instructions. Prior to this change, such arrangements usually
1759 resulted in hangs, races or other synchronisation failures, because
1760 Valgrind did not honour atomicity of such instructions.
1761
1762
1763* A new experimental tool, BBV, has been added. BBV generates basic
1764 block vectors for use with the SimPoint analysis tool, which allows
1765 a program's overall behaviour to be approximated by running only a
1766 fraction of it. This is useful for computer architecture
1767 researchers. You can run BBV by specifying --tool=exp-bbv (the
1768 "exp-" prefix is short for "experimental"). BBV was written by
1769 Vince Weaver.
1770
1771
1772* Modestly improved support for running Windows applications under
1773 Wine. In particular, initial support for reading Windows .PDB debug
1774 information has been added.
1775
1776
1777* A new Memcheck client request VALGRIND_COUNT_LEAK_BLOCKS has been
1778 added. It is similar to VALGRIND_COUNT_LEAKS but counts blocks
1779 instead of bytes.
1780
1781
1782* The Valgrind client requests VALGRIND_PRINTF and
1783 VALGRIND_PRINTF_BACKTRACE have been changed slightly. Previously,
1784 the string was always printed immediately on its own line. Now, the
1785 string will be added to a buffer but not printed until a newline is
1786 encountered, or other Valgrind output is printed (note that for
1787 VALGRIND_PRINTF_BACKTRACE, the back-trace itself is considered
1788 "other Valgrind output"). This allows you to use multiple
1789 VALGRIND_PRINTF calls to build up a single output line, and also to
1790 print multiple output lines with a single request (by embedding
1791 multiple newlines in the string).
1792
1793
1794* The graphs drawn by Massif's ms_print program have changed slightly:
1795
1796 - The half-height chars '.' and ',' are no longer drawn, because
1797 they are confusing. The --y option can be used if the default
1798 y-resolution is not high enough.
1799
1800 - Horizontal lines are now drawn after the top of a snapshot if
1801 there is a gap until the next snapshot. This makes it clear that
1802 the memory usage has not dropped to zero between snapshots.
1803
1804
1805* Something that happened in 3.4.0, but wasn't clearly announced: the
1806 option --read-var-info=yes can be used by some tools (Memcheck,
1807 Helgrind and DRD). When enabled, it causes Valgrind to read DWARF3
1808 variable type and location information. This makes those tools
1809 start up more slowly and increases memory consumption, but
1810 descriptions of data addresses in error messages become more
1811 detailed.
1812
1813
1814* exp-Omega, an experimental instantaneous leak-detecting tool, was
1815 disabled in 3.4.0 due to a lack of interest and maintenance,
1816 although the source code was still in the distribution. The source
1817 code has now been removed from the distribution. For anyone
1818 interested, the removal occurred in SVN revision r10247.
1819
njnacda1782009-06-04 23:11:50 +00001820
njn8b68b642009-06-24 00:37:09 +00001821* Some changes have been made to the build system.
njn6bf365c2009-02-11 00:35:45 +00001822
sewardj570e2442009-08-14 13:42:41 +00001823 - VEX/ is now integrated properly into the build system. This means
1824 that dependency tracking within VEX/ now works properly, "make
1825 install" will work without requiring "make" before it, and
1826 parallel builds (ie. 'make -j') now work (previously a
1827 .NOTPARALLEL directive was used to serialize builds, ie. 'make -j'
1828 was effectively ignored).
njn6bf365c2009-02-11 00:35:45 +00001829
sewardj570e2442009-08-14 13:42:41 +00001830 - The --with-vex configure option has been removed. It was of
1831 little use and removing it simplified the build system.
njn8b68b642009-06-24 00:37:09 +00001832
sewardj570e2442009-08-14 13:42:41 +00001833 - The location of some install files has changed. This should not
1834 affect most users. Those who might be affected:
njn8b68b642009-06-24 00:37:09 +00001835
1836 * For people who use Valgrind with MPI programs, the installed
sewardj570e2442009-08-14 13:42:41 +00001837 libmpiwrap.so library has moved from
1838 $(INSTALL)/<platform>/libmpiwrap.so to
1839 $(INSTALL)/libmpiwrap-<platform>.so.
njn8b68b642009-06-24 00:37:09 +00001840
sewardj570e2442009-08-14 13:42:41 +00001841 * For people who distribute standalone Valgrind tools, the
1842 installed libraries such as $(INSTALL)/<platform>/libcoregrind.a
1843 have moved to $(INSTALL)/libcoregrind-<platform>.a.
njn8b68b642009-06-24 00:37:09 +00001844
sewardj570e2442009-08-14 13:42:41 +00001845 These changes simplify the build system.
njn8b68b642009-06-24 00:37:09 +00001846
sewardj570e2442009-08-14 13:42:41 +00001847 - Previously, all the distributed suppression (*.supp) files were
1848 installed. Now, only default.supp is installed. This should not
1849 affect users as the other installed suppression files were not
1850 read; the fact that they were installed was a mistake.
1851
1852
1853* KNOWN LIMITATIONS:
1854
1855 - Memcheck is unusable with the Intel compiler suite version 11.1,
1856 when it generates code for SSE2-and-above capable targets. This
1857 is because of icc's use of highly optimised inlined strlen
1858 implementations. It causes Memcheck to report huge numbers of
1859 false errors even in simple programs. Helgrind and DRD may also
1860 have problems.
1861
1862 Versions 11.0 and earlier may be OK, but this has not been
1863 properly tested.
1864
1865
sewardj9176c862009-08-18 13:47:31 +00001866The following bugs have been fixed or resolved. Note that "n-i-bz"
1867stands for "not in bugzilla" -- that is, a bug that was reported to us
1868but never got a bugzilla entry. We encourage you to file bugs in
1869bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
1870mailing the developers (or mailing lists) directly -- bugs that are
1871not entered into bugzilla tend to get forgotten about or ignored.
sewardj570e2442009-08-14 13:42:41 +00001872
sewardj9176c862009-08-18 13:47:31 +00001873To see details of a given bug, visit
1874https://bugs.kde.org/show_bug.cgi?id=XXXXXX
1875where XXXXXX is the bug number as listed below.
sewardj570e2442009-08-14 13:42:41 +00001876
sewardj9176c862009-08-18 13:47:31 +0000187784303 How about a LockCheck tool?
187891633 dereference of null ptr in vgPlain_st_basetype
187997452 Valgrind doesn't report any pthreads problems
1880100628 leak-check gets assertion failure when using
1881 VALGRIND_MALLOCLIKE_BLOCK on malloc()ed memory
1882108528 NPTL pthread cleanup handlers not called
1883110126 Valgrind 2.4.1 configure.in tramples CFLAGS
1884110128 mallinfo is not implemented...
1885110770 VEX: Generated files not always updated when making valgrind
1886111102 Memcheck: problems with large (memory footprint) applications
1887115673 Vex's decoder should never assert
1888117564 False positive: Syscall param clone(child_tidptr) contains
1889 uninitialised byte(s)
1890119404 executing ssh from inside valgrind fails
1891133679 Callgrind does not write path names to sources with dwarf debug
1892 info
1893135847 configure.in problem with non gnu compilers (and possible fix)
1894136154 threads.c:273 (vgCallgrind_post_signal): Assertion
1895 '*(vgCallgrind_current_fn_stack.top) == 0' failed.
1896136230 memcheck reports "possibly lost", should be "still reachable"
1897137073 NULL arg to MALLOCLIKE_BLOCK causes crash
1898137904 Valgrind reports a memory leak when using POSIX threads,
1899 while it shouldn't
1900139076 valgrind VT_GETSTATE error
1901142228 complaint of elf_dynamic_do_rela in trivial usage
1902145347 spurious warning with USBDEVFS_REAPURB
1903148441 (wine) can't find memory leak in Wine, win32 binary
1904 executable file.
1905148742 Leak-check fails assert on exit
1906149878 add (proper) check for calloc integer overflow
1907150606 Call graph is broken when using callgrind control
1908152393 leak errors produce an exit code of 0. I need some way to
1909 cause leak errors to result in a nonzero exit code.
1910157154 documentation (leak-resolution doc speaks about num-callers
1911 def=4) + what is a loss record
1912159501 incorrect handling of ALSA ioctls
1913162020 Valgrinding an empty/zero-byte file crashes valgrind
1914162482 ppc: Valgrind crashes while reading stabs information
1915162718 x86: avoid segment selector 0 in sys_set_thread_area()
1916163253 (wine) canonicaliseSymtab forgot some fields in DiSym
1917163560 VEX/test_main.c is missing from valgrind-3.3.1
1918164353 malloc_usable_size() doesn't return a usable size
1919165468 Inconsistent formatting in memcheck manual -- please fix
1920169505 main.c:286 (endOfInstr):
1921 Assertion 'ii->cost_offset == *cost_offset' failed
1922177206 Generate default.supp during compile instead of configure
1923177209 Configure valt_load_address based on arch+os
1924177305 eventfd / syscall 323 patch lost
1925179731 Tests fail to build because of inlining of non-local asm labels
1926181394 helgrind: libhb_core.c:3762 (msm_write): Assertion
1927 'ordxx == POrd_EQ || ordxx == POrd_LT' failed.
1928181594 Bogus warning for empty text segment
1929181707 dwarf doesn't require enumerations to have name
1930185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64
1931185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree):
1932 Assertion '!already_present' failed.
1933185359 exp-ptrcheck: unhandled syscall getresuid()
1934185794 "WARNING: unhandled syscall: 285" (fallocate) on x86_64
1935185816 Valgrind is unable to handle debug info for files with split
1936 debug info that are prelinked afterwards
1937185980 [darwin] unhandled syscall: sem_open
1938186238 bbToIR_AMD64: disInstr miscalculated next %rip
1939186507 exp-ptrcheck unhandled syscalls prctl, etc.
1940186790 Suppression pattern used for leaks are not reported
1941186796 Symbols with length>200 in suppression files are ignored
1942187048 drd: mutex PTHREAD_PROCESS_SHARED attribute missinterpretation
1943187416 exp-ptrcheck: support for __NR_{setregid,setreuid,setresuid}
1944188038 helgrind: hg_main.c:926: mk_SHVAL_fail: the 'impossible' happened
1945188046 bashisms in the configure script
1946188127 amd64->IR: unhandled instruction bytes: 0xF0 0xF 0xB0 0xA
1947188161 memcheck: --track-origins=yes asserts "mc_machine.c:672
1948 (get_otrack_shadow_offset_wrk): the 'impossible' happened."
1949188248 helgrind: pthread_cleanup_push, pthread_rwlock_unlock,
1950 assertion fail "!lock->heldBy"
1951188427 Add support for epoll_create1 (with patch)
1952188530 Support for SIOCGSTAMPNS
1953188560 Include valgrind.spec in the tarball
1954188572 Valgrind on Mac should suppress setenv() mem leak
1955189054 Valgrind fails to build because of duplicate non-local asm labels
1956189737 vex amd64->IR: unhandled instruction bytes: 0xAC
1957189762 epoll_create syscall not handled (--tool=exp-ptrcheck)
1958189763 drd assertion failure: s_threadinfo[tid].is_recording
1959190219 unhandled syscall: 328 (x86-linux)
1960190391 dup of 181394; see above
1961190429 Valgrind reports lots of errors in ld.so with x86_64 2.9.90 glibc
1962190820 No debug information on powerpc-linux
sewardj9176c862009-08-18 13:47:31 +00001963191095 PATCH: Improve usbdevfs ioctl handling
1964191182 memcheck: VALGRIND_LEAK_CHECK quadratic when big nr of chunks
1965 or big nr of errors
1966191189 --xml=yes should obey --gen-suppressions=all
1967191192 syslog() needs a suppression on macosx
1968191271 DARWIN: WARNING: unhandled syscall: 33554697 a.k.a.: 265
1969191761 getrlimit on MacOSX
1970191992 multiple --fn-skip only works sometimes; dependent on order
1971192634 V. reports "aspacem sync_check_mapping_callback:
1972 segment mismatch" on Darwin
1973192954 __extension__ missing on 2 client requests
1974194429 Crash at start-up with glibc-2.10.1 and linux-2.6.29
1975194474 "INSTALL" file has different build instructions than "README"
1976194671 Unhandled syscall (sem_wait?) from mac valgrind
1977195069 memcheck: reports leak (memory still reachable) for
1978 printf("%d', x)
1979195169 drd: (vgDrd_barrier_post_wait):
1980 Assertion 'r->sg[p->post_iteration]' failed.
1981195268 valgrind --log-file doesn't accept ~/...
1982195838 VEX abort: LibVEX_N_SPILL_BYTES too small for CPUID boilerplate
1983195860 WARNING: unhandled syscall: unix:223
1984196528 need a error suppression for pthread_rwlock_init under os x?
1985197227 Support aio_* syscalls on Darwin
1986197456 valgrind should reject --suppressions=(directory)
1987197512 DWARF2 CFI reader: unhandled CFI instruction 0:10
1988197591 unhandled syscall 27 (mincore)
1989197793 Merge DCAS branch to the trunk == 85756, 142103
1990197794 Avoid duplicate filenames in Vex
1991197898 make check fails on current SVN
1992197901 make check fails also under exp-ptrcheck in current SVN
1993197929 Make --leak-resolution=high the default
1994197930 Reduce spacing between leak reports
1995197933 Print command line of client at start-up, and shorten preamble
1996197966 unhandled syscall 205 (x86-linux, --tool=exp-ptrcheck)
1997198395 add BBV to the distribution as an experimental tool
1998198624 Missing syscalls on Darwin: 82, 167, 281, 347
1999198649 callgrind_annotate doesn't cumulate counters
2000199338 callgrind_annotate sorting/thresholds are broken for all but Ir
2001199977 Valgrind complains about an unrecognized instruction in the
2002 atomic_incs test program
2003200029 valgrind isn't able to read Fedora 12 debuginfo
2004200760 darwin unhandled syscall: unix:284
2005200827 DRD doesn't work on Mac OS X
2006200990 VG_(read_millisecond_timer)() does not work correctly
2007201016 Valgrind does not support pthread_kill() on Mac OS
2008201169 Document --read-var-info
2009201323 Pre-3.5.0 performance sanity checking
2010201384 Review user manual for the 3.5.0 release
2011201585 mfpvr not implemented on ppc
2012201708 tests failing because x86 direction flag is left set
2013201757 Valgrind doesn't handle any recent sys_futex additions
sewardj6bbe0242009-08-19 11:38:54 +00002014204377 64-bit valgrind can not start a shell script
2015 (with #!/path/to/shell) if the shell is a 32-bit executable
sewardj9176c862009-08-18 13:47:31 +00002016n-i-bz drd: fixed assertion failure triggered by mutex reinitialization.
2017n-i-bz drd: fixed a bug that caused incorrect messages to be printed
2018 about memory allocation events with memory access tracing enabled
2019n-i-bz drd: fixed a memory leak triggered by vector clock deallocation
njn254dafe2009-02-09 03:25:04 +00002020
sewardj48867fb2009-08-19 22:40:53 +00002021(3.5.0: 19 Aug 2009, vex r1913, valgrind r10846).
sewardj570e2442009-08-14 13:42:41 +00002022
njn6eccab92009-08-14 02:41:37 +00002023
njn254dafe2009-02-09 03:25:04 +00002024
njnad91a752009-03-13 19:55:36 +00002025Release 3.4.1 (28 February 2009)
2026~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
20273.4.1 is a bug-fix release that fixes some regressions and assertion
2028failures in debug info reading in 3.4.0, most notably incorrect stack
2029traces on amd64-linux on older (glibc-2.3 based) systems. Various
2030other debug info problems are also fixed. A number of bugs in the
2031exp-ptrcheck tool introduced in 3.4.0 have been fixed.
2032
2033In view of the fact that 3.4.0 contains user-visible regressions
2034relative to 3.3.x, upgrading to 3.4.1 is recommended. Packagers are
2035encouraged to ship 3.4.1 in preference to 3.4.0.
2036
2037The fixed bugs are as follows. Note that "n-i-bz" stands for "not in
2038bugzilla" -- that is, a bug that was reported to us but never got a
2039bugzilla entry. We encourage you to file bugs in bugzilla
2040(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the
2041developers (or mailing lists) directly -- bugs that are not entered
2042into bugzilla tend to get forgotten about or ignored.
2043
2044n-i-bz Fix various bugs reading icc-11 generated debug info
2045n-i-bz Fix various bugs reading gcc-4.4 generated debug info
2046n-i-bz Preliminary support for glibc-2.10 / Fedora 11
2047n-i-bz Cachegrind and Callgrind: handle non-power-of-two cache sizes,
2048 so as to support (eg) 24k Atom D1 and Core2 with 3/6/12MB L2.
2049179618 exp-ptrcheck crashed / exit prematurely
2050179624 helgrind: false positive races with pthread_create and
2051 recv/open/close/read
2052134207 pkg-config output contains @VG_PLATFORM@
2053176926 floating point exception at valgrind startup with PPC 440EPX
2054181594 Bogus warning for empty text segment
2055173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant rex prefixes)
2056181707 Dwarf3 doesn't require enumerations to have name
2057185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64
2058185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree):
2059 Assertion '!already_present' failed.
2060185359 exp-ptrcheck unhandled syscall getresuid()
2061
2062(3.4.1.RC1: 24 Feb 2008, vex r1884, valgrind r9253).
2063(3.4.1: 28 Feb 2008, vex r1884, valgrind r9293).
2064
2065
2066
sewardj565fcf12008-12-23 13:14:06 +00002067Release 3.4.0 (2 January 2009)
2068~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
20693.4.0 is a feature release with many significant improvements and the
2070usual collection of bug fixes. This release supports X86/Linux,
2071AMD64/Linux, PPC32/Linux and PPC64/Linux. Support for recent distros
2072(using gcc 4.4, glibc 2.8 and 2.9) has been added.
sewardj8639f112008-12-17 12:26:24 +00002073
sewardj565fcf12008-12-23 13:14:06 +000020743.4.0 brings some significant tool improvements. Memcheck can now
2075report the origin of uninitialised values, the thread checkers
2076Helgrind and DRD are much improved, and we have a new experimental
2077tool, exp-Ptrcheck, which is able to detect overruns of stack and
2078global arrays. In detail:
sewardj8639f112008-12-17 12:26:24 +00002079
sewardj565fcf12008-12-23 13:14:06 +00002080* Memcheck is now able to track the origin of uninitialised values.
2081 When it reports an uninitialised value error, it will try to show
2082 the origin of the value, as either a heap or stack allocation.
2083 Origin tracking is expensive and so is not enabled by default. To
2084 use it, specify --track-origins=yes. Memcheck's speed will be
2085 essentially halved, and memory usage will be significantly
2086 increased. Nevertheless it can drastically reduce the effort
2087 required to identify the root cause of uninitialised value errors,
2088 and so is often a programmer productivity win, despite running more
2089 slowly.
sewardj8639f112008-12-17 12:26:24 +00002090
sewardj565fcf12008-12-23 13:14:06 +00002091* A version (1.4.0) of the Valkyrie GUI, that works with Memcheck in
sewardjb8b16642009-01-02 23:21:54 +00002092 3.4.0, will be released shortly.
sewardj8639f112008-12-17 12:26:24 +00002093
sewardj565fcf12008-12-23 13:14:06 +00002094* Helgrind's race detection algorithm has been completely redesigned
2095 and reimplemented, to address usability and scalability concerns:
sewardj8639f112008-12-17 12:26:24 +00002096
sewardj565fcf12008-12-23 13:14:06 +00002097 - The new algorithm has a lower false-error rate: it is much less
2098 likely to report races that do not really exist.
sewardj8639f112008-12-17 12:26:24 +00002099
sewardj565fcf12008-12-23 13:14:06 +00002100 - Helgrind will display full call stacks for both accesses involved
sewardjf1fa5ed2008-12-23 17:04:52 +00002101 in a race. This makes it easier to identify the root causes of
2102 races.
sewardj8639f112008-12-17 12:26:24 +00002103
sewardj565fcf12008-12-23 13:14:06 +00002104 - Limitations on the size of program that can run have been removed.
sewardj8639f112008-12-17 12:26:24 +00002105
sewardj565fcf12008-12-23 13:14:06 +00002106 - Performance has been modestly improved, although that is very
2107 workload-dependent.
sewardj8639f112008-12-17 12:26:24 +00002108
sewardj565fcf12008-12-23 13:14:06 +00002109 - Direct support for Qt4 threading has been added.
sewardj8639f112008-12-17 12:26:24 +00002110
sewardj565fcf12008-12-23 13:14:06 +00002111 - pthread_barriers are now directly supported.
njn75010ca2008-02-09 05:10:45 +00002112
sewardj565fcf12008-12-23 13:14:06 +00002113 - Helgrind works well on all supported Linux targets.
sewardj8639f112008-12-17 12:26:24 +00002114
sewardj565fcf12008-12-23 13:14:06 +00002115* The DRD thread debugging tool has seen major improvements:
sewardj8639f112008-12-17 12:26:24 +00002116
sewardj565fcf12008-12-23 13:14:06 +00002117 - Greatly improved performance and significantly reduced memory
2118 usage.
sewardj8639f112008-12-17 12:26:24 +00002119
sewardj565fcf12008-12-23 13:14:06 +00002120 - Support for several major threading libraries (Boost.Thread, Qt4,
2121 glib, OpenMP) has been added.
sewardj8639f112008-12-17 12:26:24 +00002122
sewardj565fcf12008-12-23 13:14:06 +00002123 - Support for atomic instructions, POSIX semaphores, barriers and
2124 reader-writer locks has been added.
2125
2126 - Works now on PowerPC CPUs too.
2127
2128 - Added support for printing thread stack usage at thread exit time.
2129
2130 - Added support for debugging lock contention.
2131
2132 - Added a manual for Drd.
2133
2134* A new experimental tool, exp-Ptrcheck, has been added. Ptrcheck
2135 checks for misuses of pointers. In that sense it is a bit like
2136 Memcheck. However, Ptrcheck can do things Memcheck can't: it can
2137 detect overruns of stack and global arrays, it can detect
2138 arbitrarily far out-of-bounds accesses to heap blocks, and it can
2139 detect accesses heap blocks that have been freed a very long time
2140 ago (millions of blocks in the past).
2141
2142 Ptrcheck currently works only on x86-linux and amd64-linux. To use
2143 it, use --tool=exp-ptrcheck. A simple manual is provided, as part
2144 of the main Valgrind documentation. As this is an experimental
2145 tool, we would be particularly interested in hearing about your
2146 experiences with it.
2147
sewardj974aace2008-12-24 18:40:57 +00002148* exp-Omega, an experimental instantaneous leak-detecting tool, is no
2149 longer built by default, although the code remains in the repository
2150 and the tarball. This is due to three factors: a perceived lack of
2151 users, a lack of maintenance, and concerns that it may not be
2152 possible to achieve reliable operation using the existing design.
sewardj565fcf12008-12-23 13:14:06 +00002153
2154* As usual, support for the latest Linux distros and toolchain
2155 components has been added. It should work well on Fedora Core 10,
2156 OpenSUSE 11.1 and Ubuntu 8.10. gcc-4.4 (in its current pre-release
2157 state) is supported, as is glibc-2.9. The C++ demangler has been
2158 updated so as to work well with C++ compiled by even the most recent
2159 g++'s.
2160
2161* You can now use frame-level wildcards in suppressions. This was a
2162 frequently-requested enhancement. A line "..." in a suppression now
2163 matches zero or more frames. This makes it easier to write
2164 suppressions which are precise yet insensitive to changes in
2165 inlining behaviour.
2166
2167* 3.4.0 adds support on x86/amd64 for the SSSE3 instruction set.
2168
2169* Very basic support for IBM Power6 has been added (64-bit processes only).
2170
2171* Valgrind is now cross-compilable. For example, it is possible to
2172 cross compile Valgrind on an x86/amd64-linux host, so that it runs
2173 on a ppc32/64-linux target.
2174
sewardjf1fa5ed2008-12-23 17:04:52 +00002175* You can set the main thread's stack size at startup using the
2176 new --main-stacksize= flag (subject of course to ulimit settings).
2177 This is useful for running apps that need a lot of stack space.
2178
2179* The limitation that you can't use --trace-children=yes together
2180 with --db-attach=yes has been removed.
2181
sewardj565fcf12008-12-23 13:14:06 +00002182* The following bugs have been fixed. Note that "n-i-bz" stands for
2183 "not in bugzilla" -- that is, a bug that was reported to us but
2184 never got a bugzilla entry. We encourage you to file bugs in
2185 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
2186 mailing the developers (or mailing lists) directly.
2187
sewardjf1fa5ed2008-12-23 17:04:52 +00002188 n-i-bz Make return types for some client requests 64-bit clean
2189 n-i-bz glibc 2.9 support
2190 n-i-bz ignore unsafe .valgrindrc's (CVE-2008-4865)
2191 n-i-bz MPI_Init(0,0) is valid but libmpiwrap.c segfaults
2192 n-i-bz Building in an env without gdb gives bogus gdb attach
2193 92456 Tracing the origin of uninitialised memory
2194 106497 Valgrind does not demangle some C++ template symbols
2195 162222 ==106497
sewardjdf1fc9a2008-12-23 23:09:57 +00002196 151612 Suppression with "..." (frame-level wildcards in .supp files)
sewardjf1fa5ed2008-12-23 17:04:52 +00002197 156404 Unable to start oocalc under memcheck on openSUSE 10.3 (64-bit)
2198 159285 unhandled syscall:25 (stime, on x86-linux)
2199 159452 unhandled ioctl 0x8B01 on "valgrind iwconfig"
2200 160954 ppc build of valgrind crashes with illegal instruction (isel)
2201 160956 mallinfo implementation, w/ patch
2202 162092 Valgrind fails to start gnome-system-monitor
2203 162819 malloc_free_fill test doesn't pass on glibc2.8 x86
2204 163794 assertion failure with "--track-origins=yes"
2205 163933 sigcontext.err and .trapno must be set together
2206 163955 remove constraint !(--db-attach=yes && --trace-children=yes)
2207 164476 Missing kernel module loading system calls
2208 164669 SVN regression: mmap() drops posix file locks
2209 166581 Callgrind output corruption when program forks
2210 167288 Patch file for missing system calls on Cell BE
2211 168943 unsupported scas instruction pentium
2212 171645 Unrecognised instruction (MOVSD, non-binutils encoding)
2213 172417 x86->IR: 0x82 ...
2214 172563 amd64->IR: 0xD9 0xF5 - fprem1
2215 173099 .lds linker script generation error
2216 173177 [x86_64] syscalls: 125/126/179 (capget/capset/quotactl)
2217 173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant prefixes)
2218 174532 == 173751
2219 174908 --log-file value not expanded correctly for core file
2220 175044 Add lookup_dcookie for amd64
2221 175150 x86->IR: 0xF2 0xF 0x11 0xC1 (movss non-binutils encoding)
sewardj565fcf12008-12-23 13:14:06 +00002222
2223Developer-visible changes:
2224
2225* Valgrind's debug-info reading machinery has been majorly overhauled.
2226 It can now correctly establish the addresses for ELF data symbols,
2227 which is something that has never worked properly before now.
2228
2229 Also, Valgrind can now read DWARF3 type and location information for
2230 stack and global variables. This makes it possible to use the
2231 framework to build tools that rely on knowing the type and locations
2232 of stack and global variables, for example exp-Ptrcheck.
2233
2234 Reading of such information is disabled by default, because most
2235 tools don't need it, and because it is expensive in space and time.
2236 However, you can force Valgrind to read it, using the
2237 --read-var-info=yes flag. Memcheck, Helgrind and DRD are able to
2238 make use of such information, if present, to provide source-level
2239 descriptions of data addresses in the error messages they create.
2240
sewardj974aace2008-12-24 18:40:57 +00002241(3.4.0.RC1: 24 Dec 2008, vex r1878, valgrind r8882).
sewardjb8b16642009-01-02 23:21:54 +00002242(3.4.0: 3 Jan 2009, vex r1878, valgrind r8899).