1 Release 3.9.0 (?? ?????? 201?)
2 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4 * ================== PLATFORM CHANGES =================
6 * ==================== TOOL CHANGES ====================
8 * ==================== OTHER CHANGES ====================
10 - Addition of GDB server monitor command 'v.info open_fds' that gives the
11 list of open file descriptors and additional details.
13 * ==================== FIXED BUGS ====================
15 The following bugs have been fixed or resolved. Note that "n-i-bz"
16 stands for "not in bugzilla" -- that is, a bug that was reported to us
17 but never got a bugzilla entry. We encourage you to file bugs in
18 bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather
19 than mailing the developers (or mailing lists) directly -- bugs that
20 are not entered into bugzilla tend to get forgotten about or ignored.
22 To see details of a given bug, visit
23 https://bugs.kde.org/show_bug.cgi?id=XXXXXX
24 where XXXXXX is the bug number as listed below.
26 v = verified fixed in 3_8_BRANCH
27 m = merged into 3_8_BRANCH
28 [390] = fixed in trunk (for 3.9.0)
29 [381] = fixed in trunk and in 3_8_BRANCH, for 3.8.1
30 [382] = fixed in trunk and needs to be made available for 3.8.2 too
32 123837 [390] semctl system call: 4rth argument is optional, depending on cmd
33 252955 [390] Impossible to compile with ccache
34 274695 [390] s390x: Support "compare to/from logical" instructions (z196)
35 275800 [390] s390x: Add support for the ecag instruction (part 1)
36 275800 [390] s390x: Autodetect cache info (part 2)
37 305948 [390] ppc64: code generation for ShlD64 / ShrD64 asserts
38 306054 [390] s390x: Condition code computation for convert-to-int/logical
39 307155 [390] filter_gdb should filter out syscall-template.S T_PSEUDO
40 308321 [390] testsuite memcheck filter interferes with gdb_filter
41 308341 [390] vgdb should report process exit (or fatal signal)
42 308644 [390] vgdb command for having the info for the track-fds option
43 308711 [390] give more info about aspacemgr and arenas in out_of_memory
44 308886 [390] Missing support for PTRACE_SET/GETREGSET
45 n-i-bz [390] report error for vgdb snapshot requested before execution
46 n-i-bz [390] Some wrong command line options could be ignored
47 n-i-bz [390] same as 303624 (fixed in 3.8.0), but for x86 android
49 Release 3.8.1 (19 September 2012)
50 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
51 3.8.1 is a bug fix release. It fixes some assertion failures in 3.8.0
52 that occur moderately frequently in real use cases, adds support for
53 some missing instructions on ARM, and fixes a deadlock condition on
54 MacOSX. If you package or deliver 3.8.0 for others to use, you might
55 want to consider upgrading to 3.8.1 instead.
57 The following bugs have been fixed or resolved. Note that "n-i-bz"
58 stands for "not in bugzilla" -- that is, a bug that was reported to us
59 but never got a bugzilla entry. We encourage you to file bugs in
60 bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather
61 than mailing the developers (or mailing lists) directly -- bugs that
62 are not entered into bugzilla tend to get forgotten about or ignored.
64 To see details of a given bug, visit
65 https://bugs.kde.org/show_bug.cgi?id=XXXXXX
66 where XXXXXX is the bug number as listed below.
69 289584 Unhandled instruction: 0xF 0x29 0xE5 (MOVAPS)
70 295808 amd64->IR: 0xF3 0xF 0xBC 0xC0 (TZCNT)
71 298281 wcslen causes false(?) uninitialised value warnings
72 301281 valgrind hangs on OS X when the process calls system()
73 304035 disInstr(arm): unhandled instruction 0xE1023053
74 304867 implement MOVBE instruction in x86 mode
75 304980 Assertion 'lo <= hi' failed in vgModuleLocal_find_rx_mapping
76 305042 amd64: implement 0F 7F encoding of movq between two registers
77 305199 ARM: implement QDADD and QDSUB
78 305321 amd64->IR: 0xF 0xD 0xC (prefetchw)
79 305513 killed by fatal signal: SIGSEGV
80 305690 DRD reporting invalid semaphore when sem_trywait fails
81 305926 Invalid alignment checks for some AVX instructions
82 306297 disInstr(thumb): unhandled instruction 0xE883 0x000C
83 306310 3.8.0 release tarball missing some files
84 306612 RHEL 6 glibc-2.X default suppressions need /lib*/libc-*patterns
85 306664 vex amd64->IR: 0x66 0xF 0x3A 0x62 0xD1 0x46 0x66 0xF
86 n-i-bz shmat of a segment > 4Gb does not work
87 n-i-bz simulate_control_c script wrong USR1 signal number on mips
88 n-i-bz vgdb ptrace calls wrong on mips [...]
89 n-i-bz Fixes for more MPI false positives
90 n-i-bz exp-sgcheck's memcpy causes programs to segfault
91 n-i-bz OSX build w/ clang: asserts at startup
92 n-i-bz Incorrect undef'dness prop for Iop_DPBtoBCD and Iop_BCDtoDPB
93 n-i-bz fix a couple of union tag-vs-field mixups
94 n-i-bz OSX: use __NR_poll_nocancel rather than __NR_poll
96 The following bugs were fixed in 3.8.0 but not listed in this NEWS
99 254088 Valgrind should know about UD2 instruction
102 304754 NEWS blows TeX's little mind
104 (3.8.1: 19 September 2012, vex rXXXX, valgrind rXXXXX)
108 Release 3.8.0 (10 August 2012)
109 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
110 3.8.0 is a feature release with many improvements and the usual
111 collection of bug fixes.
113 This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux,
114 PPC64/Linux, S390X/Linux, MIPS/Linux, ARM/Android, X86/Android,
115 X86/MacOSX 10.6/10.7 and AMD64/MacOSX 10.6/10.7. Support for recent
116 distros and toolchain components (glibc 2.16, gcc 4.7) has been added.
117 There is initial support for MacOSX 10.8, but it is not usable for
118 serious work at present.
120 * ================== PLATFORM CHANGES =================
122 * Support for MIPS32 platforms running Linux. Valgrind has been
123 tested on MIPS32 and MIPS32r2 platforms running different Debian
124 Squeeze and MeeGo distributions. Both little-endian and big-endian
125 cores are supported. The tools Memcheck, Massif and Lackey have
126 been tested and are known to work. See README.mips for more details.
128 * Preliminary support for Android running on x86.
130 * Preliminary (as-yet largely unusable) support for MacOSX 10.8.
132 * Support for Intel AVX instructions and for AES instructions. This
133 support is available only for 64 bit code.
135 * Support for POWER Decimal Floating Point instructions.
137 * ==================== TOOL CHANGES ====================
139 * Non-libc malloc implementations are now supported. This is useful
140 for tools that replace malloc (Memcheck, Massif, DRD, Helgrind).
141 Using the new option --soname-synonyms, such tools can be informed
142 that the malloc implementation is either linked statically into the
143 executable, or is present in some other shared library different
144 from libc.so. This makes it possible to process statically linked
145 programs, and programs using other malloc libraries, for example
146 TCMalloc or JEMalloc.
148 * For tools that provide their own replacement for malloc et al, the
149 option --redzone-size=<number> allows users to specify the size of
150 the padding blocks (redzones) added before and after each client
151 allocated block. Smaller redzones decrease the memory needed by
152 Valgrind. Bigger redzones increase the chance to detect blocks
153 overrun or underrun. Prior to this change, the redzone size was
154 hardwired to 16 bytes in Memcheck.
158 - The leak_check GDB server monitor command now can
159 control the maximum nr of loss records to output.
161 - Reduction of memory use for applications allocating
162 many blocks and/or having many partially defined bytes.
164 - Addition of GDB server monitor command 'block_list' that lists
165 the addresses/sizes of the blocks of a leak search loss record.
167 - Addition of GDB server monitor command 'who_points_at' that lists
168 the locations pointing at a block.
170 - If a redzone size > 0 is given, VALGRIND_MALLOCLIKE_BLOCK now will
171 detect an invalid access of these redzones, by marking them
172 noaccess. Similarly, if a redzone size is given for a memory
173 pool, VALGRIND_MEMPOOL_ALLOC will mark the redzones no access.
174 This still allows to find some bugs if the user has forgotten to
175 mark the pool superblock noaccess.
177 - Performance of memory leak check has been improved, especially in
178 cases where there are many leaked blocks and/or many suppression
179 rules used to suppress leak reports.
181 - Reduced noise (false positive) level on MacOSX 10.6/10.7, due to
182 more precise analysis, which is important for LLVM/Clang
183 generated code. This is at the cost of somewhat reduced
184 performance. Note there is no change to analysis precision or
185 costs on Linux targets.
189 - Added even more facilities that can help finding the cause of a data
190 race, namely the command-line option --ptrace-addr and the macro
191 DRD_STOP_TRACING_VAR(x). More information can be found in the manual.
193 - Fixed a subtle bug that could cause false positive data race reports.
195 * ==================== OTHER CHANGES ====================
197 * The C++ demangler has been updated so as to work well with C++
198 compiled by up to at least g++ 4.6.
200 * Tool developers can make replacement/wrapping more flexible thanks
201 to the new option --soname-synonyms. This was reported above, but
202 in fact is very general and applies to all function
203 replacement/wrapping, not just to malloc-family functions.
205 * Round-robin scheduling of threads can be selected, using the new
206 option --fair-sched= yes. Prior to this change, the pipe-based
207 thread serialisation mechanism (which is still the default) could
208 give very unfair scheduling. --fair-sched=yes improves
209 responsiveness of interactive multithreaded applications, and
210 improves repeatability of results from the thread checkers Helgrind
213 * For tool developers: support to run Valgrind on Valgrind has been
214 improved. We can now routinely Valgrind on Helgrind or Memcheck.
216 * gdbserver now shows the float shadow registers as integer
217 rather than float values, as the shadow values are mostly
218 used as bit patterns.
220 * Increased limit for the --num-callers command line flag to 500.
222 * Performance improvements for error matching when there are many
223 suppression records in use.
225 * Improved support for DWARF4 debugging information (bug 284184).
227 * Initial support for DWZ compressed Dwarf debug info.
229 * Improved control over the IR optimiser's handling of the tradeoff
230 between performance and precision of exceptions. Specifically,
231 --vex-iropt-precise-memory-exns has been removed and replaced by
232 --vex-iropt-register-updates, with extended functionality. This
233 allows the Valgrind gdbserver to always show up to date register
236 * Modest performance gains through the use of translation chaining for
239 * ==================== FIXED BUGS ====================
241 The following bugs have been fixed or resolved. Note that "n-i-bz"
242 stands for "not in bugzilla" -- that is, a bug that was reported to us
243 but never got a bugzilla entry. We encourage you to file bugs in
244 bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather
245 than mailing the developers (or mailing lists) directly -- bugs that
246 are not entered into bugzilla tend to get forgotten about or ignored.
248 To see details of a given bug, visit
249 https://bugs.kde.org/show_bug.cgi?id=XXXXXX
250 where XXXXXX is the bug number as listed below.
252 197914 Building valgrind from svn now requires automake-1.10
253 203877 increase to 16Mb maximum allowed alignment for memalign et al
254 219156 Handle statically linked malloc or other malloc lib (e.g. tcmalloc)
255 247386 make perf does not run all performance tests
256 270006 Valgrind scheduler unfair
257 270777 Adding MIPS/Linux port to Valgrind
258 270796 s390x: Removed broken support for the TS insn
259 271438 Fix configure for proper SSE4.2 detection
260 273114 s390x: Support TR, TRE, TROO, TROT, TRTO, and TRTT instructions
261 273475 Add support for AVX instructions
262 274078 improved configure logic for mpicc
263 276993 fix mremap 'no thrash checks'
264 278313 Fedora 15/x64: err read debug info with --read-var-info=yes flag
265 281482 memcheck incorrect byte allocation count in realloc() for silly argument
266 282230 group allocator for small fixed size, use it for MC_Chunk/SEc vbit
267 283413 Fix wrong sanity check
268 283671 Robustize alignment computation in LibVEX_Alloc
269 283961 Adding support for some HCI IOCTLs
270 284124 parse_type_DIE: confused by: DWARF 4
271 284864 == 273475 (Add support for AVX instructions)
272 285219 Too-restrictive constraints for Thumb2 "SP plus/minus register"
273 285662 (MacOSX): Memcheck needs to replace memcpy/memmove
274 285725 == 273475 (Add support for AVX instructions)
275 286261 add wrapper for linux I2C_RDWR ioctl
276 286270 vgpreload is not friendly to 64->32 bit execs, gives ld.so warnings
277 286374 Running cachegrind with --branch-sim=yes on 64-bit PowerPC program fails
278 286384 configure fails "checking for a supported version of gcc"
279 286497 == 273475 (Add support for AVX instructions)
280 286596 == 273475 (Add support for AVX instructions)
281 286917 disInstr(arm): unhandled instruction: QADD (also QSUB)
282 287175 ARM: scalar VFP fixed-point VCVT instructions not handled
283 287260 Incorrect conditional jump or move depends on uninitialised value(s)
284 287301 vex amd64->IR: 0x66 0xF 0x38 0x41 0xC0 0xB8 0x0 0x0 (PHMINPOSUW)
285 287307 == 273475 (Add support for AVX instructions)
286 287858 VG_(strerror): unknown error
287 288298 (MacOSX) unhandled syscall shm_unlink
288 288995 == 273475 (Add support for AVX instructions)
289 289470 Loading of large Mach-O thin binaries fails.
290 289656 == 273475 (Add support for AVX instructions)
291 289699 vgdb connection in relay mode erroneously closed due to buffer overrun
292 289823 == 293754 (PCMPxSTRx not implemented for 16-bit characters)
293 289839 s390x: Provide support for unicode conversion instructions
294 289939 monitor cmd 'leak_check' with details about leaked or reachable blocks
295 290006 memcheck doesn't mark %xmm as initialized after "pcmpeqw %xmm %xmm"
296 290655 Add support for AESKEYGENASSIST instruction
297 290719 valgrind-3.7.0 fails with automake-1.11.2 due to"pkglibdir" usage
298 290974 vgdb must align pages to VKI_SHMLBA (16KB) on ARM
299 291253 ES register not initialised in valgrind simulation
300 291568 Fix 3DNOW-related crashes with baseline x86_64 CPU (w patch)
301 291865 s390x: Support the "Compare Double and Swap" family of instructions
302 292300 == 273475 (Add support for AVX instructions)
303 292430 unrecognized instruction in __intel_get_new_mem_ops_cpuid
304 292493 == 273475 (Add support for AVX instructions)
305 292626 Missing fcntl F_SETOWN_EX and F_GETOWN_EX support
306 292627 Missing support for some SCSI ioctls
307 292628 none/tests/x86/bug125959-x86.c triggers undefined behavior
308 292841 == 273475 (Add support for AVX instructions)
309 292993 implement the getcpu syscall on amd64-linux
310 292995 Implement the “cross memory attach” syscalls introduced in Linux 3.2
311 293088 Add some VEX sanity checks for ppc64 unhandled instructions
312 293751 == 290655 (Add support for AESKEYGENASSIST instruction)
313 293754 PCMPxSTRx not implemented for 16-bit characters
314 293755 == 293754 (No tests for PCMPxSTRx on 16-bit characters)
315 293808 CLFLUSH not supported by latest VEX for amd64
316 294047 valgrind does not correctly emulate prlimit64(..., RLIMIT_NOFILE, ...)
317 294048 MPSADBW instruction not implemented
318 294055 regtest none/tests/shell fails when locale is not set to C
319 294185 INT 0x44 (and others) not supported on x86 guest, but used by Jikes RVM
320 294190 --vgdb-error=xxx can be out of sync with errors shown to the user
321 294191 amd64: fnsave/frstor and 0x66 size prefixes on FP instructions
322 294260 disInstr_AMD64: disInstr miscalculated next %rip
323 294523 --partial-loads-ok=yes causes false negatives
324 294617 vex amd64->IR: 0x66 0xF 0x3A 0xDF 0xD1 0x1 0xE8 0x6A
325 294736 vex amd64->IR: 0x48 0xF 0xD7 0xD6 0x48 0x83
326 294812 patch allowing to run (on x86 at least) helgrind/drd on tool.
327 295089 can not annotate source for both helgrind and drd
328 295221 POWER Processor decimal floating point instruction support missing
329 295427 building for i386 with clang on darwin11 requires "-new_linker linker"
330 295428 coregrind/m_main.c has incorrect x86 assembly for darwin
331 295590 Helgrind: Assertion 'cvi->nWaiters > 0' failed
332 295617 ARM - Add some missing syscalls
333 295799 Missing \n with get_vbits in gdbserver when line is % 80 [...]
334 296229 Linux user input device ioctls missing wrappers
335 296318 ELF Debug info improvements (more than one rx/rw mapping)
336 296422 Add translation chaining support
337 296457 vex amd64->IR: 0x66 0xF 0x3A 0xDF 0xD1 0x1 0xE8 0x6A (dup of AES)
338 296792 valgrind 3.7.0: add SIOCSHWTSTAMP (0x89B0) ioctl wrapper
339 296983 Fix build issues on x86_64/ppc64 without 32-bit toolchains
340 297078 gdbserver signal handling problems [..]
341 297147 drd false positives on newly allocated memory
342 297329 disallow decoding of IBM Power DFP insns on some machines
343 297497 POWER Processor decimal floating point instruction support missing
344 297701 Another alias for strncasecmp_l in libc-2.13.so
345 297911 'invalid write' not reported when using APIs for custom mem allocators.
346 297976 s390x: revisit EX implementation
347 297991 Valgrind interferes with mmap()+ftell()
348 297992 Support systems missing WIFCONTINUED (e.g. pre-2.6.10 Linux)
349 297993 Fix compilation of valgrind with gcc -g3.
350 298080 POWER Processor DFP support missing, part 3
351 298227 == 273475 (Add support for AVX instructions)
352 298335 == 273475 (Add support for AVX instructions)
353 298354 Unhandled ARM Thumb instruction 0xEB0D 0x0585 (streq)
354 298394 s390x: Don't bail out on an unknown machine model. [..]
355 298421 accept4() syscall (366) support is missing for ARM
356 298718 vex amd64->IR: 0xF 0xB1 0xCB 0x9C 0x8F 0x45
357 298732 valgrind installation problem in ubuntu with kernel version 3.x
358 298862 POWER Processor DFP instruction support missing, part 4
359 298864 DWARF reader mis-parses DW_FORM_ref_addr
360 298943 massif asserts with --pages-as-heap=yes when brk is changing [..]
361 299053 Support DWARF4 DW_AT_high_pc constant form
362 299104 == 273475 (Add support for AVX instructions)
363 299316 Helgrind: hg_main.c:628 (map_threads_lookup): Assertion 'thr' failed.
364 299629 dup3() syscall (358) support is missing for ARM
365 299694 POWER Processor DFP instruction support missing, part 5
366 299756 Ignore --free-fill for MEMPOOL_FREE and FREELIKE client requests
367 299803 == 273475 (Add support for AVX instructions)
368 299804 == 273475 (Add support for AVX instructions)
369 299805 == 273475 (Add support for AVX instructions)
370 300140 ARM - Missing (T1) SMMUL
371 300195 == 296318 (ELF Debug info improvements (more than one rx/rw mapping))
372 300389 Assertion `are_valid_hwcaps(VexArchAMD64, [..])' failed.
373 300414 FCOM and FCOMP unimplemented for amd64 guest
374 301204 infinite loop in canonicaliseSymtab with ifunc symbol
375 301229 == 203877 (increase to 16Mb maximum allowed alignment for memalign etc)
376 301265 add x86 support to Android build
377 301984 configure script doesn't detect certain versions of clang
378 302205 Fix compiler warnings for POWER VEX code and POWER test cases
379 302287 Unhandled movbe instruction on Atom processors
380 302370 PPC: fnmadd, fnmsub, fnmadds, fnmsubs insns always negate the result
381 302536 Fix for the POWER Valgrind regression test: memcheck-ISA2.0.
382 302578 Unrecognized isntruction 0xc5 0x32 0xc2 0xca 0x09 vcmpngess
383 302656 == 273475 (Add support for AVX instructions)
384 302709 valgrind for ARM needs extra tls support for android emulator [..]
385 302827 add wrapper for CDROM_GET_CAPABILITY
386 302901 Valgrind crashes with dwz optimized debuginfo
387 302918 Enable testing of the vmaddfp and vnsubfp instructions in the testsuite
388 303116 Add support for the POWER instruction popcntb
389 303127 Power test suite fixes for frsqrte, vrefp, and vrsqrtefp instructions.
390 303250 Assertion `instrs_in->arr_used <= 10000' failed w/ OpenSSL code
391 303466 == 273475 (Add support for AVX instructions)
392 303624 segmentation fault on Android 4.1 (e.g. on Galaxy Nexus OMAP)
393 303963 strstr() function produces wrong results under valgrind callgrind
394 304054 CALL_FN_xx macros need to enforce stack alignment
395 304561 tee system call not supported
396 715750 (MacOSX): Incorrect invalid-address errors near 0xFFFFxxxx (mozbug#)
397 n-i-bz Add missing gdbserver xml files for shadow registers for ppc32
398 n-i-bz Bypass gcc4.4/4.5 code gen bugs causing out of memory or asserts
399 n-i-bz Fix assert in gdbserver for watchpoints watching the same address
400 n-i-bz Fix false positive in sys_clone on amd64 when optional args [..]
401 n-i-bz s390x: Shadow registers can now be examined using vgdb
403 (3.8.0-TEST3: 9 August 2012, vex r2465, valgrind r12865)
404 (3.8.0: 10 August 2012, vex r2465, valgrind r12866)
408 Release 3.7.0 (5 November 2011)
409 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
410 3.7.0 is a feature release with many significant improvements and the
411 usual collection of bug fixes.
413 This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux,
414 PPC64/Linux, S390X/Linux, ARM/Android, X86/Darwin and AMD64/Darwin.
415 Support for recent distros and toolchain components (glibc 2.14, gcc
416 4.6, MacOSX 10.7) has been added.
418 * ================== PLATFORM CHANGES =================
420 * Support for IBM z/Architecture (s390x) running Linux. Valgrind can
421 analyse 64-bit programs running on z/Architecture. Most user space
422 instructions up to and including z10 are supported. Valgrind has
423 been tested extensively on z9, z10, and z196 machines running SLES
424 10/11, RedHat 5/6m, and Fedora. The Memcheck and Massif tools are
425 known to work well. Callgrind, Helgrind, and DRD work reasonably
426 well on z9 and later models. See README.s390 for more details.
428 * Preliminary support for MacOSX 10.7 and XCode 4. Both 32- and
429 64-bit processes are supported. Some complex threaded applications
430 (Firefox) are observed to hang when run as 32 bit applications,
431 whereas 64-bit versions run OK. The cause is unknown. Memcheck
432 will likely report some false errors. In general, expect some rough
433 spots. This release also supports MacOSX 10.6, but drops support
436 * Preliminary support for Android (on ARM). Valgrind can now run
437 large applications (eg, Firefox) on (eg) a Samsung Nexus S. See
438 README.android for more details, plus instructions on how to get
441 * Support for the IBM Power ISA 2.06 (Power7 instructions)
443 * General correctness and performance improvements for ARM/Linux, and,
444 by extension, ARM/Android.
446 * Further solidification of support for SSE 4.2 in 64-bit mode. AVX
447 instruction set support is under development but is not available in
450 * Support for AIX5 has been removed.
452 * ==================== TOOL CHANGES ====================
454 * Memcheck: some incremental changes:
456 - reduction of memory use in some circumstances
458 - improved handling of freed memory, which in some circumstances
459 can cause detection of use-after-free that would previously have
462 - fix of a longstanding bug that could cause false negatives (missed
463 errors) in programs doing vector saturated narrowing instructions.
465 * Helgrind: performance improvements and major memory use reductions,
466 particularly for large, long running applications which perform many
467 synchronisation (lock, unlock, etc) events. Plus many smaller
470 - display of locksets for both threads involved in a race
472 - general improvements in formatting/clarity of error messages
474 - addition of facilities and documentation regarding annotation
475 of thread safe reference counted C++ classes
477 - new flag --check-stack-refs=no|yes [yes], to disable race checking
478 on thread stacks (a performance hack)
480 - new flag --free-is-write=no|yes [no], to enable detection of races
481 where one thread accesses heap memory but another one frees it,
482 without any coordinating synchronisation event
484 * DRD: enabled XML output; added support for delayed thread deletion
485 in order to detect races that occur close to the end of a thread
486 (--join-list-vol); fixed a memory leak triggered by repeated client
487 memory allocatation and deallocation; improved Darwin support.
489 * exp-ptrcheck: this tool has been renamed to exp-sgcheck
491 * exp-sgcheck: this tool has been reduced in scope so as to improve
492 performance and remove checking that Memcheck does better.
493 Specifically, the ability to check for overruns for stack and global
494 arrays is unchanged, but the ability to check for overruns of heap
495 blocks has been removed. The tool has accordingly been renamed to
496 exp-sgcheck ("Stack and Global Array Checking").
498 * ==================== OTHER CHANGES ====================
500 * GDB server: Valgrind now has an embedded GDB server. That means it
501 is possible to control a Valgrind run from GDB, doing all the usual
502 things that GDB can do (single stepping, breakpoints, examining
503 data, etc). Tool-specific functionality is also available. For
504 example, it is possible to query the definedness state of variables
505 or memory from within GDB when running Memcheck; arbitrarily large
506 memory watchpoints are supported, etc. To use the GDB server, start
507 Valgrind with the flag --vgdb-error=0 and follow the on-screen
510 * Improved support for unfriendly self-modifying code: a new option
511 --smc-check=all-non-file is available. This adds the relevant
512 consistency checks only to code that originates in non-file-backed
513 mappings. In effect this confines the consistency checking only to
514 code that is or might be JIT generated, and avoids checks on code
515 that must have been compiled ahead of time. This significantly
516 improves performance on applications that generate code at run time.
518 * It is now possible to build a working Valgrind using Clang-2.9 on
521 * new client requests VALGRIND_{DISABLE,ENABLE}_ERROR_REPORTING.
522 These enable and disable error reporting on a per-thread, and
523 nestable, basis. This is useful for hiding errors in particularly
524 troublesome pieces of code. The MPI wrapper library (libmpiwrap.c)
525 now uses this facility.
527 * Added the --mod-funcname option to cg_diff.
529 * ==================== FIXED BUGS ====================
531 The following bugs have been fixed or resolved. Note that "n-i-bz"
532 stands for "not in bugzilla" -- that is, a bug that was reported to us
533 but never got a bugzilla entry. We encourage you to file bugs in
534 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
535 mailing the developers (or mailing lists) directly -- bugs that are
536 not entered into bugzilla tend to get forgotten about or ignored.
538 To see details of a given bug, visit
539 https://bugs.kde.org/show_bug.cgi?id=XXXXXX
540 where XXXXXX is the bug number as listed below.
542 210935 port valgrind.h (not valgrind) to win32 to support client requests
543 214223 valgrind SIGSEGV on startup gcc 4.4.1 ppc32 (G4) Ubuntu 9.10
544 243404 Port to zSeries
545 243935 Helgrind: incorrect handling of ANNOTATE_HAPPENS_BEFORE()/AFTER()
546 247223 non-x86: Suppress warning: 'regparm' attribute directive ignored
547 250101 huge "free" memory usage due to m_mallocfree.c fragmentation
548 253206 Some fixes for the faultstatus testcase
549 255223 capget testcase fails when running as root
550 256703 xlc_dbl_u32.c testcase broken
551 256726 Helgrind tests have broken inline asm
552 259977 == 214223 (Valgrind segfaults doing __builtin_longjmp)
553 264800 testcase compile failure on zseries
554 265762 make public VEX headers compilable by G++ 3.x
555 265771 assertion in jumps.c (r11523) fails with glibc-2.3
556 266753 configure script does not give the user the option to not use QtCore
557 266931 gen_insn_test.pl is broken
558 266961 ld-linux.so.2 i?86-linux strlen issues
559 266990 setns instruction causes false positive
560 267020 Make directory for temporary files configurable at run-time.
561 267342 == 267997 (segmentation fault on Mac OS 10.6)
562 267383 Assertion 'vgPlain_strlen(dir) + vgPlain_strlen(file) + 1 < 256' failed
563 267413 Assertion 'DRD_(g_threadinfo)[tid].synchr_nesting >= 1' failed.
564 267488 regtest: darwin support for 64-bit build
565 267552 SIGSEGV (misaligned_stack_error) with DRD, but not with other tools
566 267630 Add support for IBM Power ISA 2.06 -- stage 1
567 267769 == 267997 (Darwin: memcheck triggers segmentation fault)
568 267819 Add client request for informing the core about reallocation
569 267925 laog data structure quadratic for a single sequence of lock
570 267968 drd: (vgDrd_thread_set_joinable): Assertion '0 <= (int)tid ..' failed
571 267997 MacOSX: 64-bit V segfaults on launch when built with Xcode 4.0.1
572 268513 missed optimizations in fold_Expr
573 268619 s390x: fpr - gpr transfer facility
574 268620 s390x: reconsider "long displacement" requirement
575 268621 s390x: improve IR generation for XC
576 268715 s390x: FLOGR is not universally available
577 268792 == 267997 (valgrind seg faults on startup when compiled with Xcode 4)
578 268930 s390x: MHY is not universally available
579 269078 arm->IR: unhandled instruction SUB (SP minus immediate/register)
580 269079 Support ptrace system call on ARM
581 269144 missing "Bad option" error message
582 269209 conditional load and store facility (z196)
583 269354 Shift by zero on x86 can incorrectly clobber CC_NDEP
584 269641 == 267997 (valgrind segfaults immediately (segmentation fault))
585 269736 s390x: minor code generation tweaks
586 269778 == 272986 (valgrind.h: swap roles of VALGRIND_DO_CLIENT_REQUEST() ..)
587 269863 s390x: remove unused function parameters
588 269864 s390x: tweak s390_emit_load_cc
589 269884 == 250101 (overhead for huge blocks exhausts space too soon)
590 270082 s390x: Make sure to point the PSW address to the next address on SIGILL
591 270115 s390x: rewrite some testcases
592 270309 == 267997 (valgrind crash on startup)
593 270320 add support for Linux FIOQSIZE ioctl() call
594 270326 segfault while trying to sanitize the environment passed to execle
595 270794 IBM POWER7 support patch causes regression in none/tests
596 270851 IBM POWER7 fcfidus instruction causes memcheck to fail
597 270856 IBM POWER7 xsnmaddadp instruction causes memcheck to fail on 32bit app
598 270925 hyper-optimized strspn() in /lib64/libc-2.13.so needs fix
599 270959 s390x: invalid use of R0 as base register
600 271042 VSX configure check fails when it should not
601 271043 Valgrind build fails with assembler error on ppc64 with binutils 2.21
602 271259 s390x: fix code confusion
603 271337 == 267997 (Valgrind segfaults on MacOS X)
604 271385 s390x: Implement Ist_MBE
605 271501 s390x: misc cleanups
606 271504 s390x: promote likely and unlikely
607 271579 ppc: using wrong enum type
608 271615 unhandled instruction "popcnt" (arch=amd10h)
609 271730 Fix bug when checking ioctls: duplicate check
610 271776 s390x: provide STFLE instruction support
611 271779 s390x: provide clock instructions like STCK
612 271799 Darwin: ioctls without an arg report a memory error
613 271820 arm: fix type confusion
614 271917 pthread_cond_timedwait failure leads to not-locked false positive
615 272067 s390x: fix DISP20 macro
616 272615 A typo in debug output in mc_leakcheck.c
617 272661 callgrind_annotate chokes when run from paths containing regex chars
618 272893 amd64->IR: 0x66 0xF 0x38 0x2B 0xC1 0x66 0xF 0x7F == (closed as dup)
619 272955 Unhandled syscall error for pwrite64 on ppc64 arch
620 272967 make documentation build-system more robust
621 272986 Fix gcc-4.6 warnings with valgrind.h
622 273318 amd64->IR: 0x66 0xF 0x3A 0x61 0xC1 0x38 (missing PCMPxSTRx case)
623 273318 unhandled PCMPxSTRx case: vex amd64->IR: 0x66 0xF 0x3A 0x61 0xC1 0x38
624 273431 valgrind segfaults in evalCfiExpr (debuginfo.c:2039)
625 273465 Callgrind: jumps.c:164 (new_jcc): Assertion '(0 <= jmp) && ...'
626 273536 Build error: multiple definition of `vgDrd_pthread_cond_initializer'
627 273640 ppc64-linux: unhandled syscalls setresuid(164) and setresgid(169)
628 273729 == 283000 (Illegal opcode for SSE2 "roundsd" instruction)
629 273778 exp-ptrcheck: unhandled sysno == 259
630 274089 exp-ptrcheck: unhandled sysno == 208
631 274378 s390x: Various dispatcher tweaks
632 274447 WARNING: unhandled syscall: 340
633 274776 amd64->IR: 0x66 0xF 0x38 0x2B 0xC5 0x66
634 274784 == 267997 (valgrind ls -l results in Segmentation Fault)
635 274926 valgrind does not build against linux-3
636 275148 configure FAIL with glibc-2.14
637 275151 Fedora 15 / glibc-2.14 'make regtest' FAIL
638 275168 Make Valgrind work for MacOSX 10.7 Lion
639 275212 == 275284 (lots of false positives from __memcpy_ssse3_back et al)
640 275278 valgrind does not build on Linux kernel 3.0.* due to silly
641 275284 Valgrind memcpy/memmove redirection stopped working in glibc 2.14/x86_64
642 275308 Fix implementation for ppc64 fres instruc
643 275339 s390x: fix testcase compile warnings
644 275517 s390x: Provide support for CKSM instruction
645 275710 s390x: get rid of redundant address mode calculation
646 275815 == 247894 (Valgrind doesn't know about Linux readahead(2) syscall)
647 275852 == 250101 (valgrind uses all swap space and is killed)
648 276784 Add support for IBM Power ISA 2.06 -- stage 3
649 276987 gdbsrv: fix tests following recent commits
650 277045 Valgrind crashes with unhandled DW_OP_ opcode 0x2a
651 277199 The test_isa_2_06_part1.c in none/tests/ppc64 should be a symlink
652 277471 Unhandled syscall: 340
653 277610 valgrind crashes in VG_(lseek)(core_fd, phdrs[idx].p_offset, ...)
654 277653 ARM: support Thumb2 PLD instruction
655 277663 ARM: NEON float VMUL by scalar incorrect
656 277689 ARM: tests for VSTn with register post-index are broken
657 277694 ARM: BLX LR instruction broken in ARM mode
658 277780 ARM: VMOV.F32 (immediate) instruction is broken
659 278057 fuse filesystem syscall deadlocks
660 278078 Unimplemented syscall 280 on ppc32
661 278349 F_GETPIPE_SZ and F_SETPIPE_SZ Linux fcntl commands
662 278454 VALGRIND_STACK_DEREGISTER has wrong output type
663 278502 == 275284 (Valgrind confuses memcpy() and memmove())
664 278892 gdbsrv: factorize gdb version handling, fix doc and typos
665 279027 Support for MVCL and CLCL instruction
666 279027 s390x: Provide support for CLCL and MVCL instructions
667 279062 Remove a redundant check in the insn selector for ppc.
668 279071 JDK creates PTEST with redundant REX.W prefix
669 279212 gdbsrv: add monitor cmd v.info scheduler.
670 279378 exp-ptrcheck: the 'impossible' happened on mkfifo call
671 279698 memcheck discards valid-bits for packuswb
672 279795 memcheck reports uninitialised values for mincore on amd64
673 279994 Add support for IBM Power ISA 2.06 -- stage 3
674 280083 mempolicy syscall check errors
675 280290 vex amd64->IR: 0x66 0xF 0x38 0x28 0xC1 0x66 0xF 0x6F
676 280710 s390x: config files for nightly builds
677 280757 /tmp dir still used by valgrind even if TMPDIR is specified
678 280965 Valgrind breaks fcntl locks when program does mmap
679 281138 WARNING: unhandled syscall: 340
680 281241 == 275168 (valgrind useless on Macos 10.7.1 Lion)
681 281304 == 275168 (Darwin: dyld "cannot load inserted library")
682 281305 == 275168 (unhandled syscall: unix:357 on Darwin 11.1)
683 281468 s390x: handle do_clone and gcc clones in call traces
684 281488 ARM: VFP register corruption
685 281828 == 275284 (false memmove warning: "Source and destination overlap")
686 281883 s390x: Fix system call wrapper for "clone".
687 282105 generalise 'reclaimSuperBlock' to also reclaim splittable superblock
688 282112 Unhandled instruction bytes: 0xDE 0xD9 0x9B 0xDF (fcompp)
689 282238 SLES10: make check fails
690 282979 strcasestr needs replacement with recent(>=2.12) glibc
691 283000 vex amd64->IR: 0x66 0xF 0x3A 0xA 0xC0 0x9 0xF3 0xF
692 283243 Regression in ppc64 memcheck tests
693 283325 == 267997 (Darwin: V segfaults on startup when built with Xcode 4.0)
694 283427 re-connect epoll_pwait syscall on ARM linux
695 283600 gdbsrv: android: port vgdb.c
696 283709 none/tests/faultstatus needs to account for page size
697 284305 filter_gdb needs enhancement to work on ppc64
698 284384 clang 3.1 -Wunused-value warnings in valgrind.h, memcheck.h
699 284472 Thumb2 ROR.W encoding T2 not implemented
700 284621 XML-escape process command line in XML output
701 n-i-bz cachegrind/callgrind: handle CPUID information for Core iX Intel CPUs
702 that have non-power-of-2 sizes (also AMDs)
703 n-i-bz don't be spooked by libraries mashed by elfhack
704 n-i-bz don't be spooked by libxul.so linked with gold
705 n-i-bz improved checking for VALGRIND_CHECK_MEM_IS_DEFINED
707 (3.7.0-TEST1: 27 October 2011, vex r2228, valgrind r12245)
708 (3.7.0.RC1: 1 November 2011, vex r2231, valgrind r12257)
709 (3.7.0: 5 November 2011, vex r2231, valgrind r12258)
713 Release 3.6.1 (16 February 2011)
714 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
715 3.6.1 is a bug fix release. It adds support for some SSE4
716 instructions that were omitted in 3.6.0 due to lack of time. Initial
717 support for glibc-2.13 has been added. A number of bugs causing
718 crashing or assertion failures have been fixed.
720 The following bugs have been fixed or resolved. Note that "n-i-bz"
721 stands for "not in bugzilla" -- that is, a bug that was reported to us
722 but never got a bugzilla entry. We encourage you to file bugs in
723 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
724 mailing the developers (or mailing lists) directly -- bugs that are
725 not entered into bugzilla tend to get forgotten about or ignored.
727 To see details of a given bug, visit
728 https://bugs.kde.org/show_bug.cgi?id=XXXXXX
729 where XXXXXX is the bug number as listed below.
731 188572 Valgrind on Mac should suppress setenv() mem leak
732 194402 vex amd64->IR: 0x48 0xF 0xAE 0x4 (proper FX{SAVE,RSTOR} support)
733 210481 vex amd64->IR: Assertion `sz == 2 || sz == 4' failed (REX.W POPQ)
734 246152 callgrind internal error after pthread_cancel on 32 Bit Linux
735 250038 ppc64: Altivec LVSR and LVSL instructions fail their regtest
736 254420 memory pool tracking broken
737 254957 Test code failing to compile due to changes in memcheck.h
738 255009 helgrind/drd: crash on chmod with invalid parameter
739 255130 readdwarf3.c parse_type_DIE confused by GNAT Ada types
740 255355 helgrind/drd: crash on threaded programs doing fork
742 255418 (SSE4.x) rint call compiled with ICC
743 255822 --gen-suppressions can create invalid files: "too many callers [...]"
744 255888 closing valgrindoutput tag outputted to log-stream on error
745 255963 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x9 0xDB 0x0 (ROUNDPD)
746 255966 Slowness when using mempool annotations
747 256387 vex x86->IR: 0xD4 0xA 0x2 0x7 (AAD and AAM)
748 256600 super-optimized strcasecmp() false positive
749 256669 vex amd64->IR: Unhandled LOOPNEL insn on amd64
750 256968 (SSE4.x) vex amd64->IR: 0x66 0xF 0x38 0x10 0xD3 0x66 (BLENDVPx)
751 257011 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0xE 0xFD 0xA0 (PBLENDW)
752 257063 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x8 0xC0 0x0 (ROUNDPS)
753 257276 Missing case in memcheck --track-origins=yes
754 258870 (SSE4.x) Add support for EXTRACTPS SSE 4.1 instruction
755 261966 (SSE4.x) support for CRC32B and CRC32Q is lacking (also CRC32{W,L})
756 262985 VEX regression in valgrind 3.6.0 in handling PowerPC VMX
757 262995 (SSE4.x) crash when trying to valgrind gcc-snapshot (PCMPxSTRx $0)
758 263099 callgrind_annotate counts Ir improperly [...]
759 263877 undefined coprocessor instruction on ARMv7
760 265964 configure FAIL with glibc-2.13
761 n-i-bz Fix compile error w/ icc-12.x in guest_arm_toIR.c
762 n-i-bz Docs: fix bogus descriptions for VALGRIND_CREATE_BLOCK et al
763 n-i-bz Massif: don't assert on shmat() with --pages-as-heap=yes
764 n-i-bz Bug fixes and major speedups for the exp-DHAT space profiler
765 n-i-bz DRD: disable --free-is-write due to implementation difficulties
767 (3.6.1: 16 February 2011, vex r2103, valgrind r11561).
771 Release 3.6.0 (21 October 2010)
772 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
773 3.6.0 is a feature release with many significant improvements and the
774 usual collection of bug fixes.
776 This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux,
777 PPC64/Linux, X86/Darwin and AMD64/Darwin. Support for recent distros
778 and toolchain components (glibc 2.12, gcc 4.5, OSX 10.6) has been added.
780 -------------------------
782 Here are some highlights. Details are shown further down:
784 * Support for ARM/Linux.
786 * Support for recent Linux distros: Ubuntu 10.10 and Fedora 14.
788 * Support for Mac OS X 10.6, both 32- and 64-bit executables.
790 * Support for the SSE4.2 instruction set.
792 * Enhancements to the Callgrind profiler, including the ability to
793 handle CPUs with three levels of cache.
795 * A new experimental heap profiler, DHAT.
797 * A huge number of bug fixes and small enhancements.
799 -------------------------
801 Here are details of the above changes, together with descriptions of
802 many other changes, and a list of fixed bugs.
804 * ================== PLATFORM CHANGES =================
806 * Support for ARM/Linux. Valgrind now runs on ARMv7 capable CPUs
807 running Linux. It is known to work on Ubuntu 10.04, Ubuntu 10.10,
808 and Maemo 5, so you can run Valgrind on your Nokia N900 if you want.
810 This requires a CPU capable of running the ARMv7-A instruction set
811 (Cortex A5, A8 and A9). Valgrind provides fairly complete coverage
812 of the user space instruction set, including ARM and Thumb integer
813 code, VFPv3, NEON and V6 media instructions. The Memcheck,
814 Cachegrind and Massif tools work properly; other tools work to
817 * Support for recent Linux distros (Ubuntu 10.10 and Fedora 14), along
818 with support for recent releases of the underlying toolchain
819 components, notably gcc-4.5 and glibc-2.12.
821 * Support for Mac OS X 10.6, both 32- and 64-bit executables. 64-bit
822 support also works much better on OS X 10.5, and is as solid as
825 * Support for the SSE4.2 instruction set. SSE4.2 is supported in
826 64-bit mode. In 32-bit mode, support is only available up to and
827 including SSSE3. Some exceptions: SSE4.2 AES instructions are not
828 supported in 64-bit mode, and 32-bit mode does in fact support the
829 bare minimum SSE4 instructions to needed to run programs on Mac OS X
830 10.6 on 32-bit targets.
832 * Support for IBM POWER6 cpus has been improved. The Power ISA up to
833 and including version 2.05 is supported.
835 * ==================== TOOL CHANGES ====================
837 * Cachegrind has a new processing script, cg_diff, which finds the
838 difference between two profiles. It's very useful for evaluating
839 the performance effects of a change in a program.
841 Related to this change, the meaning of cg_annotate's (rarely-used)
842 --threshold option has changed; this is unlikely to affect many
843 people, if you do use it please see the user manual for details.
845 * Callgrind now can do branch prediction simulation, similar to
846 Cachegrind. In addition, it optionally can count the number of
847 executed global bus events. Both can be used for a better
848 approximation of a "Cycle Estimation" as derived event (you need to
849 update the event formula in KCachegrind yourself).
851 * Cachegrind and Callgrind now refer to the LL (last-level) cache
852 rather than the L2 cache. This is to accommodate machines with
853 three levels of caches -- if Cachegrind/Callgrind auto-detects the
854 cache configuration of such a machine it will run the simulation as
855 if the L2 cache isn't present. This means the results are less
856 likely to match the true result for the machine, but
857 Cachegrind/Callgrind's results are already only approximate, and
858 should not be considered authoritative. The results are still
859 useful for giving a general idea about a program's locality.
861 * Massif has a new option, --pages-as-heap, which is disabled by
862 default. When enabled, instead of tracking allocations at the level
863 of heap blocks (as allocated with malloc/new/new[]), it instead
864 tracks memory allocations at the level of memory pages (as mapped by
865 mmap, brk, etc). Each mapped page is treated as its own block.
866 Interpreting the page-level output is harder than the heap-level
867 output, but this option is useful if you want to account for every
868 byte of memory used by a program.
870 * DRD has two new command-line options: --free-is-write and
871 --trace-alloc. The former allows to detect reading from already freed
872 memory, and the latter allows tracing of all memory allocations and
875 * DRD has several new annotations. Custom barrier implementations can
876 now be annotated, as well as benign races on static variables.
878 * DRD's happens before / happens after annotations have been made more
879 powerful, so that they can now also be used to annotate e.g. a smart
880 pointer implementation.
882 * Helgrind's annotation set has also been drastically improved, so as
883 to provide to users a general set of annotations to describe locks,
884 semaphores, barriers and condition variables. Annotations to
885 describe thread-safe reference counted heap objects have also been
888 * Memcheck has a new command-line option, --show-possibly-lost, which
889 is enabled by default. When disabled, the leak detector will not
890 show possibly-lost blocks.
892 * A new experimental heap profiler, DHAT (Dynamic Heap Analysis Tool),
893 has been added. DHAT keeps track of allocated heap blocks, and also
894 inspects every memory reference to see which block (if any) is being
895 accessed. This gives a lot of insight into block lifetimes,
896 utilisation, turnover, liveness, and the location of hot and cold
897 fields. You can use DHAT to do hot-field profiling.
899 * ==================== OTHER CHANGES ====================
901 * Improved support for unfriendly self-modifying code: the extra
902 overhead incurred by --smc-check=all has been reduced by
903 approximately a factor of 5 as compared with 3.5.0.
905 * Ability to show directory names for source files in error messages.
906 This is combined with a flexible mechanism for specifying which
907 parts of the paths should be shown. This is enabled by the new flag
910 * A new flag, --require-text-symbol, which will stop the run if a
911 specified symbol is not found it a given shared object when it is
912 loaded into the process. This makes advanced working with function
913 intercepting and wrapping safer and more reliable.
915 * Improved support for the Valkyrie GUI, version 2.0.0. GUI output
916 and control of Valgrind is now available for the tools Memcheck and
917 Helgrind. XML output from Valgrind is available for Memcheck,
918 Helgrind and exp-Ptrcheck.
920 * More reliable stack unwinding on amd64-linux, particularly in the
921 presence of function wrappers, and with gcc-4.5 compiled code.
923 * Modest scalability (performance improvements) for massive
924 long-running applications, particularly for those with huge amounts
927 * Support for analyzing programs running under Wine with has been
928 improved. The header files <valgrind/valgrind.h>,
929 <valgrind/memcheck.h> and <valgrind/drd.h> can now be used in
930 Windows-programs compiled with MinGW or one of the Microsoft Visual
933 * A rare but serious error in the 64-bit x86 CPU simulation was fixed.
934 The 32-bit simulator was not affected. This did not occur often,
935 but when it did would usually crash the program under test.
938 * A large number of bugs were fixed. These are shown below.
940 * A number of bugs were investigated, and were candidates for fixing,
941 but are not fixed in 3.6.0, due to lack of developer time. They may
942 get fixed in later releases. They are:
944 194402 vex amd64->IR: 0x48 0xF 0xAE 0x4 0x24 0x49 (FXSAVE64)
945 212419 false positive "lock order violated" (A+B vs A)
946 213685 Undefined value propagates past dependency breaking instruction
947 216837 Incorrect instrumentation of NSOperationQueue on Darwin
948 237920 valgrind segfault on fork failure
949 242137 support for code compiled by LLVM-2.8
950 242423 Another unknown Intel cache config value
951 243232 Inconsistent Lock Orderings report with trylock
952 243483 ppc: callgrind triggers VEX assertion failure
953 243935 Helgrind: implementation of ANNOTATE_HAPPENS_BEFORE() is wrong
954 244677 Helgrind crash hg_main.c:616 (map_threads_lookup): Assertion
956 246152 callgrind internal error after pthread_cancel on 32 Bit Linux
957 249435 Analyzing wine programs with callgrind triggers a crash
958 250038 ppc64: Altivec lvsr and lvsl instructions fail their regtest
959 250065 Handling large allocations
960 250101 huge "free" memory usage due to m_mallocfree.c
961 "superblocks fragmentation"
962 251569 vex amd64->IR: 0xF 0x1 0xF9 0x8B 0x4C 0x24 (RDTSCP)
963 252091 Callgrind on ARM does not detect function returns correctly
964 252600 [PATCH] Allow lhs to be a pointer for shl/shr
965 254420 memory pool tracking broken
966 n-i-bz support for adding symbols for JIT generated code
969 The following bugs have been fixed or resolved. Note that "n-i-bz"
970 stands for "not in bugzilla" -- that is, a bug that was reported to us
971 but never got a bugzilla entry. We encourage you to file bugs in
972 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
973 mailing the developers (or mailing lists) directly -- bugs that are
974 not entered into bugzilla tend to get forgotten about or ignored.
976 To see details of a given bug, visit
977 https://bugs.kde.org/show_bug.cgi?id=XXXXXX
978 where XXXXXX is the bug number as listed below.
980 135264 dcbzl instruction missing
982 153699 Valgrind should report unaligned reads with movdqa
984 190429 Valgrind reports lost of errors in ld.so
985 with x86_64 2.9.90 glibc
986 197266 valgrind appears to choke on the xmms instruction
988 197988 Crash when demangling very large symbol names
989 202315 unhandled syscall: 332 (inotify_init1)
990 203256 Add page-level profiling to Massif
991 205093 dsymutil=yes needs quotes, locking (partial fix)
992 205241 Snow Leopard 10.6 support (partial fix)
993 206600 Leak checker fails to upgrade indirect blocks when their
994 parent becomes reachable
995 210935 port valgrind.h (not valgrind) to win32 so apps run under
996 wine can make client requests
997 211410 vex amd64->IR: 0x15 0xFF 0xFF 0x0 0x0 0x89
998 within Linux ip-stack checksum functions
999 212335 unhandled instruction bytes: 0xF3 0xF 0xBD 0xC0
1001 213685 Undefined value propagates past dependency breaking instruction
1003 215914 Valgrind inserts bogus empty environment variable
1005 219538 adjtimex syscall wrapper wrong in readonly adjtime mode
1006 222545 shmat fails under valgind on some arm targets
1007 222560 ARM NEON support
1010 232509 Docs build fails with formatting inside <title></title> elements
1012 235642 [PATCH] syswrap-linux.c: support evdev EVIOCG* ioctls
1013 236546 vex x86->IR: 0x66 0xF 0x3A 0xA
1014 237202 vex amd64->IR: 0xF3 0xF 0xB8 0xC0 0x49 0x3B
1015 237371 better support for VALGRIND_MALLOCLIKE_BLOCK
1016 237485 symlink (syscall 57) is not supported on Mac OS
1017 237723 sysno == 101 exp-ptrcheck: the 'impossible' happened:
1019 238208 is_just_below_ESP doesn't take into account red-zone
1020 238345 valgrind passes wrong $0 when executing a shell script
1021 238679 mq_timedreceive syscall doesn't flag the reception buffer
1023 238696 fcntl command F_DUPFD_CLOEXEC not supported
1024 238713 unhandled instruction bytes: 0x66 0xF 0x29 0xC6
1025 238713 unhandled instruction bytes: 0x66 0xF 0x29 0xC6
1026 238745 3.5.0 Make fails on PPC Altivec opcodes, though configure
1028 239992 vex amd64->IR: 0x48 0xF 0xC4 0xC1 0x0 0x48
1034 242606 unhandled syscall: setegid (in Ptrcheck)
1035 242814 Helgrind "Impossible has happened" during
1036 QApplication::initInstance();
1037 243064 Valgrind attempting to read debug information from iso
1038 243270 Make stack unwinding in Valgrind wrappers more reliable
1039 243884 exp-ptrcheck: the 'impossible happened: unhandled syscall
1040 sysno = 277 (mq_open)
1041 244009 exp-ptrcheck unknown syscalls in analyzing lighttpd
1042 244493 ARM VFP d16-d31 registers support
1043 244670 add support for audit_session_self syscall on Mac OS 10.6
1044 244921 The xml report of helgrind tool is not well format
1045 244923 In the xml report file, the <preamble> not escape the
1046 xml char, eg '<','&','>'
1047 245535 print full path names in plain text reports
1048 245925 x86-64 red zone handling problem
1049 246258 Valgrind not catching integer underruns + new [] s
1050 246311 reg/reg cmpxchg doesn't work on amd64
1051 246549 unhandled syscall unix:277 while testing 32-bit Darwin app
1052 246888 Improve Makefile.vex.am
1053 247510 [OS X 10.6] Memcheck reports unaddressable bytes passed
1054 to [f]chmod_extended
1055 247526 IBM POWER6 (ISA 2.05) support is incomplete
1056 247561 Some leak testcases fails due to reachable addresses in
1058 247875 sizeofIRType to handle Ity_I128
1059 247894 [PATCH] unhandled syscall sys_readahead
1060 247980 Doesn't honor CFLAGS passed to configure
1061 248373 darwin10.supp is empty in the trunk
1062 248822 Linux FIBMAP ioctl has int parameter instead of long
1063 248893 [PATCH] make readdwarf.c big endianess safe to enable
1064 unwinding on big endian systems
1065 249224 Syscall 336 not supported (SYS_proc_info)
1067 249775 Incorrect scheme for detecting NEON capabilities of host CPU
1068 249943 jni JVM init fails when using valgrind
1069 249991 Valgrind incorrectly declares AESKEYGENASSIST support
1071 249996 linux/arm: unhandled syscall: 181 (__NR_pwrite64)
1072 250799 frexp$fenv_access_off function generates SIGILL
1073 250998 vex x86->IR: unhandled instruction bytes: 0x66 0x66 0x66 0x2E
1074 251251 support pclmulqdq insn
1075 251362 valgrind: ARM: attach to debugger either fails or provokes
1077 251674 Unhandled syscall 294
1080 254257 Add support for debugfiles found by build-id
1081 254550 [PATCH] Implement DW_ATE_UTF (DWARF4)
1082 254646 Wrapped functions cause stack misalignment on OS X
1083 (and possibly Linux)
1084 254556 ARM: valgrinding anything fails with SIGSEGV for 0xFFFF0FA0
1086 (3.6.0: 21 October 2010, vex r2068, valgrind r11471).
1090 Release 3.5.0 (19 August 2009)
1091 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1092 3.5.0 is a feature release with many significant improvements and the
1093 usual collection of bug fixes. The main improvement is that Valgrind
1094 now works on Mac OS X.
1096 This release supports X86/Linux, AMD64/Linux, PPC32/Linux, PPC64/Linux
1097 and X86/Darwin. Support for recent distros and toolchain components
1098 (glibc 2.10, gcc 4.5) has been added.
1100 -------------------------
1102 Here is a short summary of the changes. Details are shown further
1105 * Support for Mac OS X (10.5.x).
1107 * Improvements and simplifications to Memcheck's leak checker.
1109 * Clarification and simplifications in various aspects of Valgrind's
1112 * XML output for Helgrind and Ptrcheck.
1114 * Performance and stability improvements for Helgrind and DRD.
1116 * Genuinely atomic support for x86/amd64/ppc atomic instructions.
1118 * A new experimental tool, BBV, useful for computer architecture
1121 * Improved Wine support, including ability to read Windows PDB
1124 -------------------------
1126 Here are details of the above changes, followed by descriptions of
1127 many other minor changes, and a list of fixed bugs.
1130 * Valgrind now runs on Mac OS X. (Note that Mac OS X is sometimes
1131 called "Darwin" because that is the name of the OS core, which is the
1132 level that Valgrind works at.)
1136 - It requires OS 10.5.x (Leopard). Porting to 10.4.x is not planned
1137 because it would require work and 10.4 is only becoming less common.
1139 - 32-bit programs on x86 and AMD64 (a.k.a x86-64) machines are supported
1140 fairly well. For 10.5.x, 32-bit programs are the default even on
1141 64-bit machines, so it handles most current programs.
1143 - 64-bit programs on x86 and AMD64 (a.k.a x86-64) machines are not
1144 officially supported, but simple programs at least will probably work.
1145 However, start-up is slow.
1147 - PowerPC machines are not supported.
1149 Things that don't work:
1151 - The Ptrcheck tool.
1153 - Objective-C garbage collection.
1157 - If you have Rogue Amoeba's "Instant Hijack" program installed,
1158 Valgrind will fail with a SIGTRAP at start-up. See
1159 https://bugs.kde.org/show_bug.cgi?id=193917 for details and a
1164 - You will likely find --dsymutil=yes a useful option, as error
1165 messages may be imprecise without it.
1167 - Mac OS X support is new and therefore will be less robust than the
1168 Linux support. Please report any bugs you find.
1170 - Threaded programs may run more slowly than on Linux.
1172 Many thanks to Greg Parker for developing this port over several years.
1175 * Memcheck's leak checker has been improved.
1177 - The results for --leak-check=summary now match the summary results
1178 for --leak-check=full. Previously they could differ because
1179 --leak-check=summary counted "indirectly lost" blocks and
1180 "suppressed" blocks as "definitely lost".
1182 - Blocks that are only reachable via at least one interior-pointer,
1183 but are directly pointed to by a start-pointer, were previously
1184 marked as "still reachable". They are now correctly marked as
1187 - The default value for the --leak-resolution option has been
1188 changed from "low" to "high". In general, this means that more
1189 leak reports will be produced, but each leak report will describe
1190 fewer leaked blocks.
1192 - With --leak-check=full, "definitely lost" and "possibly lost"
1193 leaks are now considered as proper errors, ie. they are counted
1194 for the "ERROR SUMMARY" and affect the behaviour of
1195 --error-exitcode. These leaks are not counted as errors if
1196 --leak-check=summary is specified, however.
1198 - Documentation for the leak checker has been improved.
1201 * Various aspects of Valgrind's text output have changed.
1203 - Valgrind's start-up message has changed. It is shorter but also
1204 includes the command being run, which makes it easier to use
1205 --trace-children=yes. An example:
1207 - Valgrind's shut-down messages have also changed. This is most
1208 noticeable with Memcheck, where the leak summary now occurs before
1209 the error summary. This change was necessary to allow leaks to be
1210 counted as proper errors (see the description of the leak checker
1211 changes above for more details). This was also necessary to fix a
1212 longstanding bug in which uses of suppressions against leaks were
1213 not "counted", leading to difficulties in maintaining suppression
1214 files (see https://bugs.kde.org/show_bug.cgi?id=186790).
1216 - Behavior of -v has changed. In previous versions, -v printed out
1217 a mixture of marginally-user-useful information, and tool/core
1218 statistics. The statistics printing has now been moved to its own
1219 flag, --stats=yes. This means -v is less verbose and more likely
1220 to convey useful end-user information.
1222 - The format of some (non-XML) stack trace entries has changed a
1223 little. Previously there were six possible forms:
1225 0x80483BF: really (a.c:20)
1226 0x80483BF: really (in /foo/a.out)
1228 0x80483BF: (within /foo/a.out)
1229 0x80483BF: ??? (a.c:20)
1232 The third and fourth of these forms have been made more consistent
1233 with the others. The six possible forms are now:
1235 0x80483BF: really (a.c:20)
1236 0x80483BF: really (in /foo/a.out)
1237 0x80483BF: really (in ???)
1238 0x80483BF: ??? (in /foo/a.out)
1239 0x80483BF: ??? (a.c:20)
1242 Stack traces produced when --xml=yes is specified are different
1246 * Helgrind and Ptrcheck now support XML output, so they can be used
1247 from GUI tools. Also, the XML output mechanism has been
1250 - The XML format has been overhauled and generalised, so it is more
1251 suitable for error reporting tools in general. The Memcheck
1252 specific aspects of it have been removed. The new format, which
1253 is an evolution of the old format, is described in
1254 docs/internals/xml-output-protocol4.txt.
1256 - Memcheck has been updated to use the new format.
1258 - Helgrind and Ptrcheck are now able to emit output in this format.
1260 - The XML output mechanism has been overhauled. XML is now output
1261 to its own file descriptor, which means that:
1263 * Valgrind can output text and XML independently.
1265 * The longstanding problem of XML output being corrupted by
1266 unexpected un-tagged text messages is solved.
1268 As before, the destination for text output is specified using
1269 --log-file=, --log-fd= or --log-socket=.
1271 As before, XML output for a tool is enabled using --xml=yes.
1273 Because there's a new XML output channel, the XML output
1274 destination is now specified by --xml-file=, --xml-fd= or
1277 Initial feedback has shown this causes some confusion. To
1278 clarify, the two envisaged usage scenarios are:
1280 (1) Normal text output. In this case, do not specify --xml=yes
1281 nor any of --xml-file=, --xml-fd= or --xml-socket=.
1283 (2) XML output. In this case, specify --xml=yes, and one of
1284 --xml-file=, --xml-fd= or --xml-socket= to select the XML
1285 destination, one of --log-file=, --log-fd= or --log-socket=
1286 to select the destination for any remaining text messages,
1287 and, importantly, -q.
1289 -q makes Valgrind completely silent on the text channel,
1290 except in the case of critical failures, such as Valgrind
1291 itself segfaulting, or failing to read debugging information.
1292 Hence, in this scenario, it suffices to check whether or not
1293 any output appeared on the text channel. If yes, then it is
1294 likely to be a critical error which should be brought to the
1295 attention of the user. If no (the text channel produced no
1296 output) then it can be assumed that the run was successful.
1298 This allows GUIs to make the critical distinction they need to
1299 make (did the run fail or not?) without having to search or
1300 filter the text output channel in any way.
1302 It is also recommended to use --child-silent-after-fork=yes in
1306 * Improvements and changes in Helgrind:
1308 - XML output, as described above
1310 - Checks for consistent association between pthread condition
1311 variables and their associated mutexes are now performed.
1313 - pthread_spinlock functions are supported.
1315 - Modest performance improvements.
1317 - Initial (skeletal) support for describing the behaviour of
1318 non-POSIX synchronisation objects through ThreadSanitizer
1319 compatible ANNOTATE_* macros.
1321 - More controllable tradeoffs between performance and the level of
1322 detail of "previous" accesses in a race. There are now three
1325 * --history-level=full. This is the default, and was also the
1326 default in 3.4.x. It shows both stacks involved in a race, but
1327 requires a lot of memory and can be very slow in programs that
1328 do many inter-thread synchronisation events.
1330 * --history-level=none. This only shows the later stack involved
1331 in a race. This can be much faster than --history-level=full,
1332 but makes it much more difficult to find the other access
1333 involved in the race.
1335 The new intermediate setting is
1337 * --history-level=approx
1339 For the earlier (other) access, two stacks are presented. The
1340 earlier access is guaranteed to be somewhere in between the two
1341 program points denoted by those stacks. This is not as useful
1342 as showing the exact stack for the previous access (as per
1343 --history-level=full), but it is better than nothing, and it's
1344 almost as fast as --history-level=none.
1347 * New features and improvements in DRD:
1349 - The error messages printed by DRD are now easier to interpret.
1350 Instead of using two different numbers to identify each thread
1351 (Valgrind thread ID and DRD thread ID), DRD does now identify
1352 threads via a single number (the DRD thread ID). Furthermore
1353 "first observed at" information is now printed for all error
1354 messages related to synchronization objects.
1356 - Added support for named semaphores (sem_open() and sem_close()).
1358 - Race conditions between pthread_barrier_wait() and
1359 pthread_barrier_destroy() calls are now reported.
1361 - Added support for custom allocators through the macros
1362 VALGRIND_MALLOCLIKE_BLOCK() VALGRIND_FREELIKE_BLOCK() (defined in
1363 in <valgrind/valgrind.h>). An alternative for these two macros is
1364 the new client request VG_USERREQ__DRD_CLEAN_MEMORY (defined in
1367 - Added support for annotating non-POSIX synchronization objects
1368 through several new ANNOTATE_*() macros.
1370 - OpenMP: added support for the OpenMP runtime (libgomp) included
1371 with gcc versions 4.3.0 and 4.4.0.
1375 - Added two new command-line options (--first-race-only and
1376 --segment-merging-interval).
1379 * Genuinely atomic support for x86/amd64/ppc atomic instructions
1381 Valgrind will now preserve (memory-access) atomicity of LOCK-
1382 prefixed x86/amd64 instructions, and any others implying a global
1383 bus lock. Ditto for PowerPC l{w,d}arx/st{w,d}cx. instructions.
1385 This means that Valgrinded processes will "play nicely" in
1386 situations where communication with other processes, or the kernel,
1387 is done through shared memory and coordinated with such atomic
1388 instructions. Prior to this change, such arrangements usually
1389 resulted in hangs, races or other synchronisation failures, because
1390 Valgrind did not honour atomicity of such instructions.
1393 * A new experimental tool, BBV, has been added. BBV generates basic
1394 block vectors for use with the SimPoint analysis tool, which allows
1395 a program's overall behaviour to be approximated by running only a
1396 fraction of it. This is useful for computer architecture
1397 researchers. You can run BBV by specifying --tool=exp-bbv (the
1398 "exp-" prefix is short for "experimental"). BBV was written by
1402 * Modestly improved support for running Windows applications under
1403 Wine. In particular, initial support for reading Windows .PDB debug
1404 information has been added.
1407 * A new Memcheck client request VALGRIND_COUNT_LEAK_BLOCKS has been
1408 added. It is similar to VALGRIND_COUNT_LEAKS but counts blocks
1412 * The Valgrind client requests VALGRIND_PRINTF and
1413 VALGRIND_PRINTF_BACKTRACE have been changed slightly. Previously,
1414 the string was always printed immediately on its own line. Now, the
1415 string will be added to a buffer but not printed until a newline is
1416 encountered, or other Valgrind output is printed (note that for
1417 VALGRIND_PRINTF_BACKTRACE, the back-trace itself is considered
1418 "other Valgrind output"). This allows you to use multiple
1419 VALGRIND_PRINTF calls to build up a single output line, and also to
1420 print multiple output lines with a single request (by embedding
1421 multiple newlines in the string).
1424 * The graphs drawn by Massif's ms_print program have changed slightly:
1426 - The half-height chars '.' and ',' are no longer drawn, because
1427 they are confusing. The --y option can be used if the default
1428 y-resolution is not high enough.
1430 - Horizontal lines are now drawn after the top of a snapshot if
1431 there is a gap until the next snapshot. This makes it clear that
1432 the memory usage has not dropped to zero between snapshots.
1435 * Something that happened in 3.4.0, but wasn't clearly announced: the
1436 option --read-var-info=yes can be used by some tools (Memcheck,
1437 Helgrind and DRD). When enabled, it causes Valgrind to read DWARF3
1438 variable type and location information. This makes those tools
1439 start up more slowly and increases memory consumption, but
1440 descriptions of data addresses in error messages become more
1444 * exp-Omega, an experimental instantaneous leak-detecting tool, was
1445 disabled in 3.4.0 due to a lack of interest and maintenance,
1446 although the source code was still in the distribution. The source
1447 code has now been removed from the distribution. For anyone
1448 interested, the removal occurred in SVN revision r10247.
1451 * Some changes have been made to the build system.
1453 - VEX/ is now integrated properly into the build system. This means
1454 that dependency tracking within VEX/ now works properly, "make
1455 install" will work without requiring "make" before it, and
1456 parallel builds (ie. 'make -j') now work (previously a
1457 .NOTPARALLEL directive was used to serialize builds, ie. 'make -j'
1458 was effectively ignored).
1460 - The --with-vex configure option has been removed. It was of
1461 little use and removing it simplified the build system.
1463 - The location of some install files has changed. This should not
1464 affect most users. Those who might be affected:
1466 * For people who use Valgrind with MPI programs, the installed
1467 libmpiwrap.so library has moved from
1468 $(INSTALL)/<platform>/libmpiwrap.so to
1469 $(INSTALL)/libmpiwrap-<platform>.so.
1471 * For people who distribute standalone Valgrind tools, the
1472 installed libraries such as $(INSTALL)/<platform>/libcoregrind.a
1473 have moved to $(INSTALL)/libcoregrind-<platform>.a.
1475 These changes simplify the build system.
1477 - Previously, all the distributed suppression (*.supp) files were
1478 installed. Now, only default.supp is installed. This should not
1479 affect users as the other installed suppression files were not
1480 read; the fact that they were installed was a mistake.
1483 * KNOWN LIMITATIONS:
1485 - Memcheck is unusable with the Intel compiler suite version 11.1,
1486 when it generates code for SSE2-and-above capable targets. This
1487 is because of icc's use of highly optimised inlined strlen
1488 implementations. It causes Memcheck to report huge numbers of
1489 false errors even in simple programs. Helgrind and DRD may also
1492 Versions 11.0 and earlier may be OK, but this has not been
1496 The following bugs have been fixed or resolved. Note that "n-i-bz"
1497 stands for "not in bugzilla" -- that is, a bug that was reported to us
1498 but never got a bugzilla entry. We encourage you to file bugs in
1499 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
1500 mailing the developers (or mailing lists) directly -- bugs that are
1501 not entered into bugzilla tend to get forgotten about or ignored.
1503 To see details of a given bug, visit
1504 https://bugs.kde.org/show_bug.cgi?id=XXXXXX
1505 where XXXXXX is the bug number as listed below.
1507 84303 How about a LockCheck tool?
1508 91633 dereference of null ptr in vgPlain_st_basetype
1509 97452 Valgrind doesn't report any pthreads problems
1510 100628 leak-check gets assertion failure when using
1511 VALGRIND_MALLOCLIKE_BLOCK on malloc()ed memory
1512 108528 NPTL pthread cleanup handlers not called
1513 110126 Valgrind 2.4.1 configure.in tramples CFLAGS
1514 110128 mallinfo is not implemented...
1515 110770 VEX: Generated files not always updated when making valgrind
1516 111102 Memcheck: problems with large (memory footprint) applications
1517 115673 Vex's decoder should never assert
1518 117564 False positive: Syscall param clone(child_tidptr) contains
1519 uninitialised byte(s)
1520 119404 executing ssh from inside valgrind fails
1521 133679 Callgrind does not write path names to sources with dwarf debug
1523 135847 configure.in problem with non gnu compilers (and possible fix)
1524 136154 threads.c:273 (vgCallgrind_post_signal): Assertion
1525 '*(vgCallgrind_current_fn_stack.top) == 0' failed.
1526 136230 memcheck reports "possibly lost", should be "still reachable"
1527 137073 NULL arg to MALLOCLIKE_BLOCK causes crash
1528 137904 Valgrind reports a memory leak when using POSIX threads,
1530 139076 valgrind VT_GETSTATE error
1531 142228 complaint of elf_dynamic_do_rela in trivial usage
1532 145347 spurious warning with USBDEVFS_REAPURB
1533 148441 (wine) can't find memory leak in Wine, win32 binary
1535 148742 Leak-check fails assert on exit
1536 149878 add (proper) check for calloc integer overflow
1537 150606 Call graph is broken when using callgrind control
1538 152393 leak errors produce an exit code of 0. I need some way to
1539 cause leak errors to result in a nonzero exit code.
1540 157154 documentation (leak-resolution doc speaks about num-callers
1541 def=4) + what is a loss record
1542 159501 incorrect handling of ALSA ioctls
1543 162020 Valgrinding an empty/zero-byte file crashes valgrind
1544 162482 ppc: Valgrind crashes while reading stabs information
1545 162718 x86: avoid segment selector 0 in sys_set_thread_area()
1546 163253 (wine) canonicaliseSymtab forgot some fields in DiSym
1547 163560 VEX/test_main.c is missing from valgrind-3.3.1
1548 164353 malloc_usable_size() doesn't return a usable size
1549 165468 Inconsistent formatting in memcheck manual -- please fix
1550 169505 main.c:286 (endOfInstr):
1551 Assertion 'ii->cost_offset == *cost_offset' failed
1552 177206 Generate default.supp during compile instead of configure
1553 177209 Configure valt_load_address based on arch+os
1554 177305 eventfd / syscall 323 patch lost
1555 179731 Tests fail to build because of inlining of non-local asm labels
1556 181394 helgrind: libhb_core.c:3762 (msm_write): Assertion
1557 'ordxx == POrd_EQ || ordxx == POrd_LT' failed.
1558 181594 Bogus warning for empty text segment
1559 181707 dwarf doesn't require enumerations to have name
1560 185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64
1561 185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree):
1562 Assertion '!already_present' failed.
1563 185359 exp-ptrcheck: unhandled syscall getresuid()
1564 185794 "WARNING: unhandled syscall: 285" (fallocate) on x86_64
1565 185816 Valgrind is unable to handle debug info for files with split
1566 debug info that are prelinked afterwards
1567 185980 [darwin] unhandled syscall: sem_open
1568 186238 bbToIR_AMD64: disInstr miscalculated next %rip
1569 186507 exp-ptrcheck unhandled syscalls prctl, etc.
1570 186790 Suppression pattern used for leaks are not reported
1571 186796 Symbols with length>200 in suppression files are ignored
1572 187048 drd: mutex PTHREAD_PROCESS_SHARED attribute missinterpretation
1573 187416 exp-ptrcheck: support for __NR_{setregid,setreuid,setresuid}
1574 188038 helgrind: hg_main.c:926: mk_SHVAL_fail: the 'impossible' happened
1575 188046 bashisms in the configure script
1576 188127 amd64->IR: unhandled instruction bytes: 0xF0 0xF 0xB0 0xA
1577 188161 memcheck: --track-origins=yes asserts "mc_machine.c:672
1578 (get_otrack_shadow_offset_wrk): the 'impossible' happened."
1579 188248 helgrind: pthread_cleanup_push, pthread_rwlock_unlock,
1580 assertion fail "!lock->heldBy"
1581 188427 Add support for epoll_create1 (with patch)
1582 188530 Support for SIOCGSTAMPNS
1583 188560 Include valgrind.spec in the tarball
1584 188572 Valgrind on Mac should suppress setenv() mem leak
1585 189054 Valgrind fails to build because of duplicate non-local asm labels
1586 189737 vex amd64->IR: unhandled instruction bytes: 0xAC
1587 189762 epoll_create syscall not handled (--tool=exp-ptrcheck)
1588 189763 drd assertion failure: s_threadinfo[tid].is_recording
1589 190219 unhandled syscall: 328 (x86-linux)
1590 190391 dup of 181394; see above
1591 190429 Valgrind reports lots of errors in ld.so with x86_64 2.9.90 glibc
1592 190820 No debug information on powerpc-linux
1593 191095 PATCH: Improve usbdevfs ioctl handling
1594 191182 memcheck: VALGRIND_LEAK_CHECK quadratic when big nr of chunks
1596 191189 --xml=yes should obey --gen-suppressions=all
1597 191192 syslog() needs a suppression on macosx
1598 191271 DARWIN: WARNING: unhandled syscall: 33554697 a.k.a.: 265
1599 191761 getrlimit on MacOSX
1600 191992 multiple --fn-skip only works sometimes; dependent on order
1601 192634 V. reports "aspacem sync_check_mapping_callback:
1602 segment mismatch" on Darwin
1603 192954 __extension__ missing on 2 client requests
1604 194429 Crash at start-up with glibc-2.10.1 and linux-2.6.29
1605 194474 "INSTALL" file has different build instructions than "README"
1606 194671 Unhandled syscall (sem_wait?) from mac valgrind
1607 195069 memcheck: reports leak (memory still reachable) for
1609 195169 drd: (vgDrd_barrier_post_wait):
1610 Assertion 'r->sg[p->post_iteration]' failed.
1611 195268 valgrind --log-file doesn't accept ~/...
1612 195838 VEX abort: LibVEX_N_SPILL_BYTES too small for CPUID boilerplate
1613 195860 WARNING: unhandled syscall: unix:223
1614 196528 need a error suppression for pthread_rwlock_init under os x?
1615 197227 Support aio_* syscalls on Darwin
1616 197456 valgrind should reject --suppressions=(directory)
1617 197512 DWARF2 CFI reader: unhandled CFI instruction 0:10
1618 197591 unhandled syscall 27 (mincore)
1619 197793 Merge DCAS branch to the trunk == 85756, 142103
1620 197794 Avoid duplicate filenames in Vex
1621 197898 make check fails on current SVN
1622 197901 make check fails also under exp-ptrcheck in current SVN
1623 197929 Make --leak-resolution=high the default
1624 197930 Reduce spacing between leak reports
1625 197933 Print command line of client at start-up, and shorten preamble
1626 197966 unhandled syscall 205 (x86-linux, --tool=exp-ptrcheck)
1627 198395 add BBV to the distribution as an experimental tool
1628 198624 Missing syscalls on Darwin: 82, 167, 281, 347
1629 198649 callgrind_annotate doesn't cumulate counters
1630 199338 callgrind_annotate sorting/thresholds are broken for all but Ir
1631 199977 Valgrind complains about an unrecognized instruction in the
1632 atomic_incs test program
1633 200029 valgrind isn't able to read Fedora 12 debuginfo
1634 200760 darwin unhandled syscall: unix:284
1635 200827 DRD doesn't work on Mac OS X
1636 200990 VG_(read_millisecond_timer)() does not work correctly
1637 201016 Valgrind does not support pthread_kill() on Mac OS
1638 201169 Document --read-var-info
1639 201323 Pre-3.5.0 performance sanity checking
1640 201384 Review user manual for the 3.5.0 release
1641 201585 mfpvr not implemented on ppc
1642 201708 tests failing because x86 direction flag is left set
1643 201757 Valgrind doesn't handle any recent sys_futex additions
1644 204377 64-bit valgrind can not start a shell script
1645 (with #!/path/to/shell) if the shell is a 32-bit executable
1646 n-i-bz drd: fixed assertion failure triggered by mutex reinitialization.
1647 n-i-bz drd: fixed a bug that caused incorrect messages to be printed
1648 about memory allocation events with memory access tracing enabled
1649 n-i-bz drd: fixed a memory leak triggered by vector clock deallocation
1651 (3.5.0: 19 Aug 2009, vex r1913, valgrind r10846).
1655 Release 3.4.1 (28 February 2009)
1656 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1657 3.4.1 is a bug-fix release that fixes some regressions and assertion
1658 failures in debug info reading in 3.4.0, most notably incorrect stack
1659 traces on amd64-linux on older (glibc-2.3 based) systems. Various
1660 other debug info problems are also fixed. A number of bugs in the
1661 exp-ptrcheck tool introduced in 3.4.0 have been fixed.
1663 In view of the fact that 3.4.0 contains user-visible regressions
1664 relative to 3.3.x, upgrading to 3.4.1 is recommended. Packagers are
1665 encouraged to ship 3.4.1 in preference to 3.4.0.
1667 The fixed bugs are as follows. Note that "n-i-bz" stands for "not in
1668 bugzilla" -- that is, a bug that was reported to us but never got a
1669 bugzilla entry. We encourage you to file bugs in bugzilla
1670 (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the
1671 developers (or mailing lists) directly -- bugs that are not entered
1672 into bugzilla tend to get forgotten about or ignored.
1674 n-i-bz Fix various bugs reading icc-11 generated debug info
1675 n-i-bz Fix various bugs reading gcc-4.4 generated debug info
1676 n-i-bz Preliminary support for glibc-2.10 / Fedora 11
1677 n-i-bz Cachegrind and Callgrind: handle non-power-of-two cache sizes,
1678 so as to support (eg) 24k Atom D1 and Core2 with 3/6/12MB L2.
1679 179618 exp-ptrcheck crashed / exit prematurely
1680 179624 helgrind: false positive races with pthread_create and
1681 recv/open/close/read
1682 134207 pkg-config output contains @VG_PLATFORM@
1683 176926 floating point exception at valgrind startup with PPC 440EPX
1684 181594 Bogus warning for empty text segment
1685 173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant rex prefixes)
1686 181707 Dwarf3 doesn't require enumerations to have name
1687 185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64
1688 185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree):
1689 Assertion '!already_present' failed.
1690 185359 exp-ptrcheck unhandled syscall getresuid()
1692 (3.4.1.RC1: 24 Feb 2008, vex r1884, valgrind r9253).
1693 (3.4.1: 28 Feb 2008, vex r1884, valgrind r9293).
1697 Release 3.4.0 (2 January 2009)
1698 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1699 3.4.0 is a feature release with many significant improvements and the
1700 usual collection of bug fixes. This release supports X86/Linux,
1701 AMD64/Linux, PPC32/Linux and PPC64/Linux. Support for recent distros
1702 (using gcc 4.4, glibc 2.8 and 2.9) has been added.
1704 3.4.0 brings some significant tool improvements. Memcheck can now
1705 report the origin of uninitialised values, the thread checkers
1706 Helgrind and DRD are much improved, and we have a new experimental
1707 tool, exp-Ptrcheck, which is able to detect overruns of stack and
1708 global arrays. In detail:
1710 * Memcheck is now able to track the origin of uninitialised values.
1711 When it reports an uninitialised value error, it will try to show
1712 the origin of the value, as either a heap or stack allocation.
1713 Origin tracking is expensive and so is not enabled by default. To
1714 use it, specify --track-origins=yes. Memcheck's speed will be
1715 essentially halved, and memory usage will be significantly
1716 increased. Nevertheless it can drastically reduce the effort
1717 required to identify the root cause of uninitialised value errors,
1718 and so is often a programmer productivity win, despite running more
1721 * A version (1.4.0) of the Valkyrie GUI, that works with Memcheck in
1722 3.4.0, will be released shortly.
1724 * Helgrind's race detection algorithm has been completely redesigned
1725 and reimplemented, to address usability and scalability concerns:
1727 - The new algorithm has a lower false-error rate: it is much less
1728 likely to report races that do not really exist.
1730 - Helgrind will display full call stacks for both accesses involved
1731 in a race. This makes it easier to identify the root causes of
1734 - Limitations on the size of program that can run have been removed.
1736 - Performance has been modestly improved, although that is very
1739 - Direct support for Qt4 threading has been added.
1741 - pthread_barriers are now directly supported.
1743 - Helgrind works well on all supported Linux targets.
1745 * The DRD thread debugging tool has seen major improvements:
1747 - Greatly improved performance and significantly reduced memory
1750 - Support for several major threading libraries (Boost.Thread, Qt4,
1751 glib, OpenMP) has been added.
1753 - Support for atomic instructions, POSIX semaphores, barriers and
1754 reader-writer locks has been added.
1756 - Works now on PowerPC CPUs too.
1758 - Added support for printing thread stack usage at thread exit time.
1760 - Added support for debugging lock contention.
1762 - Added a manual for Drd.
1764 * A new experimental tool, exp-Ptrcheck, has been added. Ptrcheck
1765 checks for misuses of pointers. In that sense it is a bit like
1766 Memcheck. However, Ptrcheck can do things Memcheck can't: it can
1767 detect overruns of stack and global arrays, it can detect
1768 arbitrarily far out-of-bounds accesses to heap blocks, and it can
1769 detect accesses heap blocks that have been freed a very long time
1770 ago (millions of blocks in the past).
1772 Ptrcheck currently works only on x86-linux and amd64-linux. To use
1773 it, use --tool=exp-ptrcheck. A simple manual is provided, as part
1774 of the main Valgrind documentation. As this is an experimental
1775 tool, we would be particularly interested in hearing about your
1776 experiences with it.
1778 * exp-Omega, an experimental instantaneous leak-detecting tool, is no
1779 longer built by default, although the code remains in the repository
1780 and the tarball. This is due to three factors: a perceived lack of
1781 users, a lack of maintenance, and concerns that it may not be
1782 possible to achieve reliable operation using the existing design.
1784 * As usual, support for the latest Linux distros and toolchain
1785 components has been added. It should work well on Fedora Core 10,
1786 OpenSUSE 11.1 and Ubuntu 8.10. gcc-4.4 (in its current pre-release
1787 state) is supported, as is glibc-2.9. The C++ demangler has been
1788 updated so as to work well with C++ compiled by even the most recent
1791 * You can now use frame-level wildcards in suppressions. This was a
1792 frequently-requested enhancement. A line "..." in a suppression now
1793 matches zero or more frames. This makes it easier to write
1794 suppressions which are precise yet insensitive to changes in
1797 * 3.4.0 adds support on x86/amd64 for the SSSE3 instruction set.
1799 * Very basic support for IBM Power6 has been added (64-bit processes only).
1801 * Valgrind is now cross-compilable. For example, it is possible to
1802 cross compile Valgrind on an x86/amd64-linux host, so that it runs
1803 on a ppc32/64-linux target.
1805 * You can set the main thread's stack size at startup using the
1806 new --main-stacksize= flag (subject of course to ulimit settings).
1807 This is useful for running apps that need a lot of stack space.
1809 * The limitation that you can't use --trace-children=yes together
1810 with --db-attach=yes has been removed.
1812 * The following bugs have been fixed. Note that "n-i-bz" stands for
1813 "not in bugzilla" -- that is, a bug that was reported to us but
1814 never got a bugzilla entry. We encourage you to file bugs in
1815 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
1816 mailing the developers (or mailing lists) directly.
1818 n-i-bz Make return types for some client requests 64-bit clean
1819 n-i-bz glibc 2.9 support
1820 n-i-bz ignore unsafe .valgrindrc's (CVE-2008-4865)
1821 n-i-bz MPI_Init(0,0) is valid but libmpiwrap.c segfaults
1822 n-i-bz Building in an env without gdb gives bogus gdb attach
1823 92456 Tracing the origin of uninitialised memory
1824 106497 Valgrind does not demangle some C++ template symbols
1826 151612 Suppression with "..." (frame-level wildcards in .supp files)
1827 156404 Unable to start oocalc under memcheck on openSUSE 10.3 (64-bit)
1828 159285 unhandled syscall:25 (stime, on x86-linux)
1829 159452 unhandled ioctl 0x8B01 on "valgrind iwconfig"
1830 160954 ppc build of valgrind crashes with illegal instruction (isel)
1831 160956 mallinfo implementation, w/ patch
1832 162092 Valgrind fails to start gnome-system-monitor
1833 162819 malloc_free_fill test doesn't pass on glibc2.8 x86
1834 163794 assertion failure with "--track-origins=yes"
1835 163933 sigcontext.err and .trapno must be set together
1836 163955 remove constraint !(--db-attach=yes && --trace-children=yes)
1837 164476 Missing kernel module loading system calls
1838 164669 SVN regression: mmap() drops posix file locks
1839 166581 Callgrind output corruption when program forks
1840 167288 Patch file for missing system calls on Cell BE
1841 168943 unsupported scas instruction pentium
1842 171645 Unrecognised instruction (MOVSD, non-binutils encoding)
1843 172417 x86->IR: 0x82 ...
1844 172563 amd64->IR: 0xD9 0xF5 - fprem1
1845 173099 .lds linker script generation error
1846 173177 [x86_64] syscalls: 125/126/179 (capget/capset/quotactl)
1847 173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant prefixes)
1849 174908 --log-file value not expanded correctly for core file
1850 175044 Add lookup_dcookie for amd64
1851 175150 x86->IR: 0xF2 0xF 0x11 0xC1 (movss non-binutils encoding)
1853 Developer-visible changes:
1855 * Valgrind's debug-info reading machinery has been majorly overhauled.
1856 It can now correctly establish the addresses for ELF data symbols,
1857 which is something that has never worked properly before now.
1859 Also, Valgrind can now read DWARF3 type and location information for
1860 stack and global variables. This makes it possible to use the
1861 framework to build tools that rely on knowing the type and locations
1862 of stack and global variables, for example exp-Ptrcheck.
1864 Reading of such information is disabled by default, because most
1865 tools don't need it, and because it is expensive in space and time.
1866 However, you can force Valgrind to read it, using the
1867 --read-var-info=yes flag. Memcheck, Helgrind and DRD are able to
1868 make use of such information, if present, to provide source-level
1869 descriptions of data addresses in the error messages they create.
1871 (3.4.0.RC1: 24 Dec 2008, vex r1878, valgrind r8882).
1872 (3.4.0: 3 Jan 2009, vex r1878, valgrind r8899).