1Release 3.11.0 (?? ????????? 201?)
2~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3Release 3.11.0 is under development, not yet released.
4
5* ================== PLATFORM CHANGES =================
6
7* s390x: It is now required for the host to have the long displacement
8  facility. The oldest supported machine model is z990.
9
10* ==================== TOOL CHANGES ====================
11
12* Memcheck:
13
14* Massif:
15  - New monitor command 'all_snapshots <filename>' that dumps all snapshots
16    taken so far.
17
18* Helgrind:
19  - The default value for --conflict-cache-size=N has been doubled to 2000000.
20    Users that were not using the default value should preferrably also
21    double the value they give.
22    The default was updaded due to the changes in the full history
23    implementation. Doubling the value gives in average a slightly more
24    complete history and uses similar memory (or significantly less memory
25    in the worst case) than the previous Helgrind version.
26
27  - Significant memory improvement and moderate speed improvement for
28    --history-level=full for applications accessing a lot of memory with
29    many different stacktraces.
30
31  - The helgrind monitor command 'info locks' now accepts an optional
32    argument 'lock_addr', to only show information about the lock at the
33    given address.
34
35  - When using --history-level=full, the new helgrind monitor command
36    'accesshistory <addr> [<len>]' will show the recorded accesses for
37    <len> (or 1) bytes at <addr>.
38
39* Callgrind:
40
41* ==================== OTHER CHANGES ====================
42
43* When a process dies due to a signal, Valgrind now shows the signal
44  and the stacktrace at default verbosity (i.e. verbosity 1).
45
46* Address description logic (used by memcheck and helgrind)
47  now describes addresses in anonymous segments, file mmap-ed
48  segments, shared memory segments and the brk data segment.
49
50* Option --error-markers=<begin>,<end> can be used to mark
51  the begin/end of errors in textual output mode, to facilitate
52  searching/extracting errors in output files mixing valgrind
53  errors with program output.
54
55* New Option --max-threads=<number> can be used to change the
56  number of threads valgrind can handle. The default is 500 threads
57  which should be more than enough for most applications.
58
59* New Option --valgrind-stacksize=<number> can be used to change
60  the size of the private thread stacks used by Valgrind.
61  Useful to reduce memory use or increase the stack size if Valgrind
62  segfaults due to stack exhausted.
63
64* New Option --avg-transtab-entry-size=<number> can be used to tune
65  the size of the translation table sectors, either to gain memory
66  or to avoid too many retranslations.
67
68* Valgrind can be built with Intel's ICC compiler. The required
69  compiler version is 14.0 or later.
70
71* New and modified GDB server monitor features:
72  - When a signal is reported in GDB, you can now use the GDB convenience
73    variable $_siginfo to examine detailed signal information.
74
75  - Valgrind gdbserver now allows the user to change the signal
76    to deliver to the process. So, use 'signal SIGNAL' to continue execution
77    with SIGNAL instead of the signal reported to GDB. Use 'signal 0' to
78    continue without passing the signal to the process.
79
80  - With recent GDB (>= 7.9.50.20150514-cvs), the command 'target remote'
81    will automatically load the executable file of the process running
82    under Valgrind. This means you do not need to specify the executable
83    file yourself, GDB will discover it itself.
84    See GDB documentation about 'qXfer:exec-file:read' packet for more info.
85
86* ==================== FIXED BUGS ====================
87
88The following bugs have been fixed or resolved.  Note that "n-i-bz"
89stands for "not in bugzilla" -- that is, a bug that was reported to us
90but never got a bugzilla entry.  We encourage you to file bugs in
91bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather
92than mailing the developers (or mailing lists) directly -- bugs that
93are not entered into bugzilla tend to get forgotten about or ignored.
94
95To see details of a given bug, visit
96  https://bugs.kde.org/show_bug.cgi?id=XXXXXX
97where XXXXXX is the bug number as listed below.
98
99116002  VG_(printf): Problems with justification of strings and integers
100155125  avoid cutting away file:lineno after long function name
101197259  Unsupported arch_prtctl PR_SET_GS option
102201152  ppc64: Assertion in ppc32g_dirtyhelper_MFSPR_268_269
103201435  Fix Darwin: -v does not show kernel version
104208217  "Warning: noted but unhandled ioctl 0x2000747b" on Mac OS X
105211256  Fixed an outdated comment regarding the default platform.
106211926  Avoid compilation warnings in valgrind.h with -pedantic
107212291  Fix unhandled syscall: unix:132 (mkfifo) on OS X
108        == 263119
109226609  Crediting upstream authors in man page
110231257  Valgrind omits path when executing script from shebang line
111254164  OS X task_info: UNKNOWN task message [id 3405, to mach_task_self(),
112        reply 0x........]
113269360  s390x: Fix addressing mode selection for compare-and-swap
114319274  Fix unhandled syscall: unix:410 (sigsuspend_nocancel) on OS X
115333051  mmap of huge pages fails due to incorrect alignment
116        == 339163
117334802  valgrind does not always explain why a given option is bad
118335785  amd64->IR 0xC4 0xE2 0x75 0x2F (vmaskmovpd)
119        == 307399
120        == 343175
121        == 342740
122335907  segfault when running wine's ddrawex/tests/surface.c under valgrind
123338731  ppc: Fix testuite build for toolchains not supporting -maltivec
124338995  shmat with hugepages (SHM_HUGETLB) fails with EINVAL
125339045  Getting valgrind to compile and run on OS X Yosemite (10.10)
126        == 340252
127339156  gdbsrv not called for fatal signal
128339288  support Cavium Octeon MIPS specific BBIT*32 instructions
129339442  Fix testsuite build failure on OS X 10.9
130339542  Enable compilation with Intel's ICC compiler
131339636  Use fxsave64 and fxrstor64 mnemonics instead of old-school rex64 prefix
132339688  Mac-specific ASM does not support .version directive (cpuid,
133        tronical and pushfpopf tests)
134339745  Valgrind crash when check Marmalade app (partial fix)
135339755  Fix known deliberate memory leak in setenv() on Mac OS X 10.9
136339780  Fix known uninitialised read in pthread_rwlock_init() on Mac OS X 10.9
137339789  Fix none/tests/execve test on Mac OS X 10.9
138339808  Fix none/tests/rlimit64_nofile test on Mac OS X 10.9
139340115  Fix none/tests/cmdline[1|2] tests on systems which define TMPDIR
140340430  Fix some grammatical weirdness in the manual.
141341238  Recognize GCC5/DWARFv5 DW_LANG constants (Go, C11, C++11, C++14)
142341539  VG_(describe_addr) should not describe address as belonging to client
143        segment if it is past the heap end
144341613  Enable building of manythreads and thread-exits tests on Mac OS X
145341615  Fix none/tests/darwin/access_extended test on Mac OS X
146341698  Valgrind's AESKEYGENASSIST gives wrong result in words 0 and 2
147        when dest register = source register
148341789  aarch64: shmat fails with valgrind on ARMv8
149342038  Unhandled syscalls on aarch64 (mbind/get/set_mempolicy)
150342063  wrong format specifier for test mcblocklistsearch in gdbserver_tests
151342221  socket connect false positive uninit memory for unknown af family
152342353  Allow dumping full massif output while valgrind is still running
153342603  Add I2C_SMBUS ioctl support
154342635  OS X 10.10 (Yosemite) - missing system calls and fcntl code
155342683  Mark memory past the initial brk limit as unaddressable
156342795  Internal glibc __GI_mempcpy call should be intercepted
157343012  Unhandled syscall 319 (memfd_create)
158343069  Patch updating v4l2 API support
159343173  helgrind crash during stack unwind
160343219  fix GET_STARTREGS for arm
161343303  Fix known deliberate memory leak in setenv() on Mac OS X 10.10
162343306  OS X 10.10: UNKNOWN mach_msg unhandled MACH_SEND_TRAILER option
163343332  Unhandled instruction 0x9E310021 (fcvtmu) on aarch64
164343335  unhandled instruction 0x1E638400 (fccmp) aarch64
165343523  OS X mach_ports_register: UNKNOWN task message [id 3403, to
166        mach_task_self(), reply 0x30f]
167343597  ppc64le: incorrect use of offseof macro
168343732  Unhandled syscall 144 (setgid) on aarch64
169343733  Unhandled syscall 187 (msgctl and related) on aarch64
170343802  s390x: Fix false positives "conditional jump or move depends on
171        unitialised value(s)"
172343902  --vgdb=yes doesn't break when --xml=yes is used
173343967  Don't warn about setuid/setgid/setcap executable for directories
174343978  Recognize DWARF5/GCC5 DW_LANG_Fortran 2003 and 2008 constants
175344007  accept4 syscall unhandled on arm64 (242) and ppc64 (344)
176344033  Helgrind on ARM32 loses track of mutex lockedness state in
177        pthread_cond_wait
178344054  www - update info for Solaris/illumos
179344235  Remove duplicate include of pub_core_aspacemgr.h
180344279  syscall sendmmsg on arm64 (269) and ppc32/64 (349) unhandled
181344295  syscall recvmmsg on arm64 (243) and ppc32/64 (343) unhandled
182344307  2 unhandled syscalls on aarch64/arm64: umount2(39), mount (40)
183344314  callgrind_annotate ... warnings about commands containing newlines
184344318  socketcall should wrap recvmmsg and sendmmsg
185344499  Fix compilation for Linux kernel >= 4. With this, also require
186        a Linux kernel >= 2.6 as 2.4 is mostly untested and might trigger
187        obvious and non-obvious issues
188344512  Fix unhandled syscall: unix:348 (__pthread_chdir) and unhandled
189        syscall: unix:349 (__pthread_fchdir) on OS X
190344559  Garbage collection of unused segment names in address space manager
191344560  Fix stack traces missing penultimate frame on OS X
192344621  Fix memcheck/tests/err_disable4 test on OS X
193344686  Fix suppression for pthread_rwlock_init on OS X 10.10
194344702  Fix missing libobjc suppressions on OS X 10.10
195344936  Fix unhandled syscall: unix:473 (readlinkat) on OS X 10.10
196344939  Fix memcheck/tests/xml1 on OS X 10.10
197345016  helgrind/tests/locked_vs_unlocked2 is failing sometimes
198345079  Fix build problems in VEX/useful/test_main.c
199345215  Performance improvements for the register allocator
200345394  Fix memcheck/tests/strchr on OS X
201345637  Fix memcheck/tests/sendmsg on OS X
202345695  Add POWERPC support for AT_DCACHESIZE and HWCAP2
203345824  Fix aspacem segment mismatch: seen with none/tests/bigcode
204345887  Fix an assertion in the address space manager
205346267  Compiler warnings for PPC64 code on call to LibVEX_GuestPPC64_get_XER()
206        and LibVEX_GuestPPC64_get_CR()
207346270  Regression tests none/tests/jm_vec/isa_2_07 and
208        none/tests/test_isa_2_07_part2 have failures on PPC64 little endian
209346307  fuse filesystem syscall deadlocks
210346324  PPC64 missing support for lbarx, lharx, stbcx and sthcx instructions
211346416  Add support for LL_IOC_PATH2FID and LL_IOC_GETPARENT Lustre ioctls
212346474  PPC64 Power 8, spr TEXASRU register not supported
213346487  Compiler generates "note" about a future ABI change for PPC64
214346801  Fix link error on OS X: _vgModuleLocal_sf_maybe_extend_stack
215347151  Fix suppression for pthread_rwlock_init on OS X 10.8
216347233  Fix memcheck/tests/strchr on OS X 10.10 (Haswell)
217347379  valgrind --leak-check=full memleak errors from system libraries on OS X 10.8
218        == 217236
219347389  unhandled syscall: 373 (Linux ARM syncfs)
220n-i-bz  Provide implementations of certain compiler builtins to support
221        compilers who may not provide those
222n-i-bz  Old STABS code is still being compiled, but never used. Remove it.
223n-i-bz  Fix compilation on distros with glibc < 2.5
224n-i-bz  (vex 3098) Avoid generation of Neon insns on non-Neon hosts
225n-i-bz  Enable rt_sigpending syscall on ppc64 linux.
226n-i-bz  mremap did not work properly on shared memory
227
228
229Release 3.10.1 (25 November 2014)
230~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2313.10.1 is a bug fix release.  It fixes various bugs reported in 3.10.0
232and backports fixes for all reported missing AArch64 ARMv8 instructions
233and syscalls from the trunk.  If you package or deliver 3.10.0 for others
234to use, you might want to consider upgrading to 3.10.1 instead.
235
236The following bugs have been fixed or resolved.  Note that "n-i-bz"
237stands for "not in bugzilla" -- that is, a bug that was reported to us
238but never got a bugzilla entry.  We encourage you to file bugs in
239bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather
240than mailing the developers (or mailing lists) directly -- bugs that
241are not entered into bugzilla tend to get forgotten about or ignored.
242
243To see details of a given bug, visit
244  https://bugs.kde.org/show_bug.cgi?id=XXXXXX
245where XXXXXX is the bug number as listed below.
246
247335440  arm64: ld1 (single structure) is not implemented
248335713  arm64: unhanded instruction: prfm (immediate)
249339020  ppc64: memcheck/tests/ppc64/power_ISA2_05 failing in nightly build
250339182  ppc64: AvSplat ought to load destination vector register with [..]
251339336  PPC64 store quad instruction (stq) is not supposed to change [..]
252339433  ppc64 lxvw4x instruction uses four 32-byte loads
253339645  Use correct tag names in sys_getdents/64 wrappers
254339706  Fix false positive for ioctl(TIOCSIG) on linux
255339721  assertion 'check_sibling == sibling' failed in readdwarf3.c ...
256339853  arm64 times syscall unknown
257339855  arm64 unhandled getsid/setsid syscalls
258339858  arm64 dmb sy not implemented
259339926  Unhandled instruction 0x1E674001 (frintx) on aarm64
260339927  Unhandled instruction 0x9E7100C6 (fcvtmu) on aarch64
261339938  disInstr(arm64): unhandled instruction 0x4F8010A4 (fmla)
262        == 339950
263339940  arm64: unhandled syscall: 83 (sys_fdatasync) + patch
264340033  arm64: unhandled insn dmb ishld and some other isb-dmb-dsb variants
265340028  unhandled syscalls for arm64 (msync, pread64, setreuid and setregid)
266340036  arm64: Unhandled instruction ld4 (multiple structures, no offset)
267340236  arm64: unhandled syscalls: mknodat, fchdir, chroot, fchownat
268340509  arm64: unhandled instruction fcvtas
269340630  arm64: fchmod (52) and fchown (55) syscalls not recognized
270340632  arm64: unhandled instruction fcvtas
271340722  Resolve "UNKNOWN attrlist flags 0:0x10000000"
272340725  AVX2: Incorrect decoding of vpbroadcast{b,w} reg,reg forms
273340788  warning: unhandled syscall: 318 (getrandom)
274340807  disInstr(arm): unhandled instruction: 0xEE989B20
275340856  disInstr(arm64): unhandled instruction 0x1E634C45 (fcsel)
276340922  arm64: unhandled getgroups/setgroups syscalls
277n-i-bz  DRD and Helgrind: Handle Imbe_CancelReservation (clrex on ARM)
278n-i-bz  Add missing ]] to terminate CDATA.
279n-i-bz  Glibc versions prior to 2.5 do not define PTRACE_GETSIGINFO
280n-i-bz  Enable sys_fadvise64_64 on arm32.
281n-i-bz  Add test cases for all remaining AArch64 SIMD, FP and memory insns.
282n-i-bz  Add test cases for all known arm64 load/store instructions.
283n-i-bz  PRE(sys_openat): when checking whether ARG1 == VKI_AT_FDCWD [..]
284n-i-bz  Add detection of old ppc32 magic instructions from bug 278808.
285n-i-bz  exp-dhat: Implement missing function "dh_malloc_usable_size".
286n-i-bz  arm64: Implement "fcvtpu w, s".
287n-i-bz  arm64: implement ADDP and various others
288n-i-bz  arm64: Implement {S,U}CVTF (scalar, fixedpt).
289n-i-bz  arm64: enable FCVT{A,N}S X,S.
290
291(3.10.1: 25 November 2014, vex r3026, valgrind r14785)
292
293
294
295Release 3.10.0 (10 September 2014)
296~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
297
2983.10.0 is a feature release with many improvements and the usual
299collection of bug fixes.
300
301This release supports X86/Linux, AMD64/Linux, ARM32/Linux, ARM64/Linux,
302PPC32/Linux, PPC64BE/Linux, PPC64LE/Linux, S390X/Linux, MIPS32/Linux,
303MIPS64/Linux, ARM/Android, MIPS32/Android, X86/Android, X86/MacOSX 10.9
304and AMD64/MacOSX 10.9.  Support for MacOSX 10.8 and 10.9 is
305significantly improved relative to the 3.9.0 release.
306
307* ================== PLATFORM CHANGES =================
308
309* Support for the 64-bit ARM Architecture (AArch64 ARMv8).  This port
310  is mostly complete, and is usable, but some SIMD instructions are as
311  yet unsupported.
312
313* Support for little-endian variant of the 64-bit POWER architecture.
314
315* Support for Android on MIPS32.
316
317* Support for 64bit FPU on MIPS32 platforms.
318
319* Both 32- and 64-bit executables are supported on MacOSX 10.8 and 10.9.
320
321* Configuration for and running on Android targets has changed.
322  See README.android in the source tree for details.
323
324* ================== DEPRECATED FEATURES =================
325
326* --db-attach is now deprecated and will be removed in the next
327  valgrind feature release.  The built-in GDB server capabilities are
328  superior and should be used instead. Learn more here:
329  http://valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.gdbserver
330
331* ==================== TOOL CHANGES ====================
332
333* Memcheck:
334
335  - Client code can now selectively disable and re-enable reporting of
336    invalid address errors in specific ranges using the new client
337    requests VALGRIND_DISABLE_ADDR_ERROR_REPORTING_IN_RANGE and
338    VALGRIND_ENABLE_ADDR_ERROR_REPORTING_IN_RANGE.
339
340  - Leak checker: there is a new leak check heuristic called
341    "length64".  This is used to detect interior pointers pointing 8
342    bytes inside a block, on the assumption that the first 8 bytes
343    holds the value "block size - 8".  This is used by
344    sqlite3MemMalloc, for example.
345
346  - Checking of system call parameters: if a syscall parameter
347    (e.g. bind struct sockaddr, sendmsg struct msghdr, ...) has
348    several fields not initialised, an error is now reported for each
349    field. Previously, an error was reported only for the first
350    uninitialised field.
351
352  - Mismatched alloc/free checking: a new flag
353    --show-mismatched-frees=no|yes [yes] makes it possible to turn off
354    such checks if necessary.
355
356* Helgrind:
357
358  - Improvements to error messages:
359
360    o Race condition error message involving heap allocated blocks also
361      show the thread number that allocated the raced-on block.
362
363    o All locks referenced by an error message are now announced.
364      Previously, some error messages only showed the lock addresses.
365
366    o The message indicating where a lock was first observed now also
367      describes the address/location of the lock.
368
369  - Helgrind now understands the Ada task termination rules and
370    creates a happens-before relationship between a terminated task
371    and its master.  This avoids some false positives and avoids a big
372    memory leak when a lot of Ada tasks are created and terminated.
373    The interceptions are only activated with forthcoming releases of
374    gnatpro >= 7.3.0w-20140611 and gcc >= 5.0.
375
376  - A new GDB server monitor command "info locks" giving the list of
377    locks, their location, and their status.
378
379* Callgrind:
380
381  - callgrind_control now supports the --vgdb-prefix argument,
382    which is needed if valgrind was started with this same argument.
383
384* ==================== OTHER CHANGES ====================
385
386* Unwinding through inlined function calls.  Stack unwinding can now
387  make use of Dwarf3 inlined-unwind information if it is available.
388  The practical effect is that inlined calls become visible in stack
389  traces.  The suppression matching machinery has been adjusted
390  accordingly.  This is controlled by the new option
391  --read-inline-info=yes|no.  Currently this is enabled by default
392  only on Linux and Android targets and only for the tools Memcheck,
393  Helgrind and DRD.
394
395* Valgrind can now read EXIDX unwind information on 32-bit ARM
396  targets.  If an object contains both CFI and EXIDX unwind
397  information, Valgrind will prefer the CFI over the EXIDX.  This
398  facilitates unwinding through system libraries on arm-android
399  targets.
400
401* Address description logic has been improved and is now common
402  between Memcheck and Helgrind, resulting in better address
403  descriptions for some kinds of error messages.
404
405* Error messages about dubious arguments (eg, to malloc or calloc) are
406  output like other errors.  This means that they can be suppressed
407  and they have a stack trace.
408
409* The C++ demangler has been updated for better C++11 support.
410
411* New and modified GDB server monitor features:
412
413  - Thread local variables/storage (__thread) can now be displayed.
414
415  - The GDB server monitor command "v.info location <address>"
416    displays information about an address.  The information produced
417    depends on the tool and on the options given to valgrind.
418    Possibly, the following are described: global variables, local
419    (stack) variables, allocated or freed blocks, ...
420
421  - The option "--vgdb-stop-at=event1,event2,..." allows the user to
422    ask the GDB server to stop at the start of program execution, at
423    the end of the program execution and on Valgrind internal errors.
424
425  - A new monitor command "v.info stats" shows various Valgrind core
426    and tool statistics.
427
428  - A new monitor command "v.set hostvisibility" allows the GDB server
429    to provide access to Valgrind internal host status/memory.
430
431* A new option "--aspace-minaddr=<address>" can in some situations
432  allow the use of more memory by decreasing the address above which
433  Valgrind maps memory.  It can also be used to solve address
434  conflicts with system libraries by increasing the default value.
435  See user manual for details.
436
437* The amount of memory used by Valgrind to store debug info (unwind
438  info, line number information and symbol data) has been
439  significantly reduced, even though Valgrind now reads more
440  information in order to support unwinding of inlined function calls.
441
442* Dwarf3 handling with --read-var-info=yes has been improved:
443
444  - Ada and C struct containing VLAs no longer cause a "bad DIE" error
445
446  - Code compiled with
447    -ffunction-sections -fdata-sections -Wl,--gc-sections
448    no longer causes assertion failures.
449
450* Improved checking for the --sim-hints= and --kernel-variant=
451  options.  Unknown strings are now detected and reported to the user
452  as a usage error.
453
454* The semantics of stack start/end boundaries in the valgrind.h
455  VALGRIND_STACK_REGISTER client request has been clarified and
456  documented.  The convention is that start and end are respectively
457  the lowest and highest addressable bytes of the stack.
458
459* ==================== FIXED BUGS ====================
460
461The following bugs have been fixed or resolved.  Note that "n-i-bz"
462stands for "not in bugzilla" -- that is, a bug that was reported to us
463but never got a bugzilla entry.  We encourage you to file bugs in
464bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather
465than mailing the developers (or mailing lists) directly -- bugs that
466are not entered into bugzilla tend to get forgotten about or ignored.
467
468To see details of a given bug, visit
469  https://bugs.kde.org/show_bug.cgi?id=XXXXXX
470where XXXXXX is the bug number as listed below.
471
472175819  Support for ipv6 socket reporting with --track-fds
473232510  make distcheck fails
474249435  Analyzing wine programs with callgrind triggers a crash
475278972  support for inlined function calls in stacktraces and suppression
476        == 199144
477291310  FXSAVE instruction marks memory as undefined on amd64
478303536  ioctl for SIOCETHTOOL (ethtool(8)) isn't wrapped
479308729  vex x86->IR: unhandled instruction bytes 0xf 0x5 (syscall)
480315199  vgcore file for threaded app does not show which thread crashed
481315952  tun/tap ioctls are not supported
482323178  Unhandled instruction: PLDW register (ARM)
483323179  Unhandled instruction: PLDW immediate (ARM)
484324050  Helgrind: SEGV because of unaligned stack when using movdqa
485325110  Add test-cases for Power ISA 2.06 insns: divdo/divdo. and divduo/divduo.
486325124  [MIPSEL] Compilation error
487325477  Phase 4 support for IBM Power ISA 2.07
488325538  cavium octeon mips64, valgrind reported "dumping core" [...]
489325628  Phase 5 support for IBM Power ISA 2.07
490325714  Empty vgcore but RLIMIT_CORE is big enough (too big)
491325751  Missing the two privileged Power PC Transactional Memory Instructions
492325816  Phase 6 support for IBM Power ISA 2.07
493325856  Make SGCheck fail gracefully on unsupported platforms
494326026  Iop names for count leading zeros/sign bits incorrectly imply [..]
495326436  DRD: False positive in libstdc++ std::list::push_back
496326444  Cavium MIPS Octeon Specific Load Indexed Instructions
497326462  Refactor vgdb to isolate invoker stuff into separate module
498326469  amd64->IR: 0x66 0xF 0x3A 0x63 0xC1 0xE (pcmpistri 0x0E)
499326623  DRD: false positive conflict report in a field assignment
500326724  Valgrind does not compile on OSX 1.9 Mavericks
501326816  Intercept for __strncpy_sse2_unaligned missing?
502326921  coregrind fails to compile m_trampoline.S with MIPS/Linux port of V
503326983  Clear direction flag after tests on amd64.
504327212  Do not prepend the current directory to absolute path names.
505327223  Support for Cavium MIPS Octeon Atomic and Count Instructions
506327238  Callgrind Assertion 'passed <= last_bb->cjmp_count' failed
507327284  s390x: Fix translation of the risbg instruction
508327639  vex amd64->IR pcmpestri SSE4.2 instruction is unsupported 0x34
509327837  dwz compressed alternate .debug_info and .debug_str not read correctly
510327916  DW_TAG_typedef may have no name
511327943  s390x: add a redirection for the 'index' function
512328100  XABORT not implemented
513328205  Implement additional Xen hypercalls
514328454  add support Backtraces with ARM unwind tables (EXIDX)
515328455  s390x: SIGILL after emitting wrong register pair for ldxbr
516328711  valgrind.1 manpage "memcheck options" section is badly generated
517328878  vex amd64->IR pcmpestri SSE4.2 instruction is unsupported 0x14
518329612  Incorrect handling of AT_BASE for image execution
519329694  clang warns about using uninitialized variable
520329956  valgrind crashes when lmw/stmw instructions are used on ppc64
521330228  mmap must align to VKI_SHMLBA on mips32
522330257  LLVM does not support `-mno-dynamic-no-pic` option
523330319  amd64->IR: unhandled instruction bytes: 0xF 0x1 0xD5 (xend)
524330459  --track-fds=yes doesn't track eventfds
525330469  Add clock_adjtime syscall support
526330594  Missing sysalls on PowerPC / uClibc
527330622  Add test to regression suite for POWER instruction: dcbzl
528330939  Support for AMD's syscall instruction on x86
529        == 308729
530330941  Typo in PRE(poll) syscall wrapper
531331057  unhandled instruction: 0xEEE01B20 (vfma.f64) (has patch)
532331254  Fix expected output for memcheck/tests/dw4
533331255  Fix race condition in test none/tests/coolo_sigaction
534331257  Fix type of jump buffer in test none/tests/faultstatus
535331305  configure uses bash specific syntax
536331337  s390x WARNING: unhandled syscall: 326 (dup3)
537331380  Syscall param timer_create(evp) points to uninitialised byte(s)
538331476  Patch to handle ioctl 0x5422 on Linux (x86 and amd64)
539331829  Unexpected ioctl opcode sign extension
540331830  ppc64: WARNING: unhandled syscall: 96/97
541331839  drd/tests/sem_open specifies invalid semaphore name
542331847  outcome of drd/tests/thread_name is nondeterministic
543332037  Valgrind cannot handle Thumb "add pc, reg"
544332055  drd asserts on platforms with VG_STACK_REDZONE_SZB == 0 and
545        consistency checks enabled
546332263  intercepts for pthread_rwlock_timedrdlock and
547        pthread_rwlock_timedwrlock are incorrect
548332265  drd could do with post-rwlock_init and pre-rwlock_destroy
549        client requests
550332276  Implement additional Xen hypercalls
551332658  ldrd.w r1, r2, [PC, #imm] does not adjust for 32bit alignment
552332765  Fix ms_print to create temporary files in a proper directory
553333072  drd: Add semaphore annotations
554333145  Tests for missaligned PC+#imm access for arm
555333228  AAarch64 Missing instruction encoding: mrs %[reg], ctr_el0
556333230  AAarch64 missing instruction encodings: dc, ic, dsb.
557333248  WARNING: unhandled syscall: unix:443
558333428  ldr.w pc [rD, #imm] instruction leads to assertion
559333501  cachegrind: assertion: Cache set count is not a power of two.
560        == 336577
561        == 292281
562333666  Recognize MPX instructions and bnd prefix.
563333788  Valgrind does not support the CDROM_DISC_STATUS ioctl (has patch)
564333817  Valgrind reports the memory areas written to by the SG_IO
565        ioctl as untouched
566334049  lzcnt fails silently (x86_32)
567334384  Valgrind does not have support Little Endian support for
568        IBM POWER PPC 64
569334585  recvmmsg unhandled (+patch) (arm)
570334705  sendmsg and recvmsg should guard against bogus msghdr fields.
571334727  Build fails with -Werror=format-security
572334788  clarify doc about --log-file initial program directory
573334834  PPC64 Little Endian support, patch 2
574334836  PPC64 Little Endian support, patch 3 testcase fixes
575334936  patch to fix false positives on alsa SNDRV_CTL_* ioctls
576335034  Unhandled ioctl: HCIGETDEVLIST
577335155  vgdb, fix error print statement.
578335262  arm64: movi 8bit version is not supported
579335263  arm64: dmb instruction is not implemented
580335441  unhandled ioctl 0x8905 (SIOCATMARK) when running wine under valgrind
581335496  arm64: sbc/abc instructions are not implemented
582335554  arm64: unhandled instruction: abs
583335564  arm64: unhandled instruction: fcvtpu  Xn, Sn
584335735  arm64: unhandled instruction: cnt
585335736  arm64: unhandled instruction: uaddlv
586335848  arm64: unhandled instruction: {s,u}cvtf
587335902  arm64: unhandled instruction: sli
588335903  arm64: unhandled instruction: umull (vector)
589336055  arm64: unhandled instruction: mov (element)
590336062  arm64: unhandled instruction: shrn{,2}
591336139  mip64: [...] valgrind hangs and spins on a single core [...]
592336189  arm64: unhandled Instruction: mvn
593336435  Valgrind hangs in pthread_spin_lock consuming 100% CPU
594336619  valgrind --read-var-info=yes doesn't handle DW_TAG_restrict_type
595336772  Make moans about unknown ioctls more informative
596336957  Add a section about the Solaris/illumos port on the webpage
597337094  ifunc wrapper is broken on ppc64
598337285  fcntl commands F_OFD_SETLK, F_OFD_SETLKW, and F_OFD_GETLK not supported
599337528  leak check heuristic for block prefixed by length as 64bit number
600337740  Implement additional Xen hypercalls
601337762  guest_arm64_toIR.c:4166 (dis_ARM64_load_store): Assertion `0' failed.
602337766  arm64-linux: unhandled syscalls mlock (228) and mlockall (230)
603337871  deprecate --db-attach
604338023  Add support for all V4L2/media ioctls
605338024  inlined functions are not shown if DW_AT_ranges is used
606338106  Add support for 'kcmp' syscall
607338115  DRD: computed conflict set differs from actual after fork
608338160  implement display of thread local storage in gdbsrv
609338205  configure.ac and check for -Wno-tautological-compare
610338300  coredumps are missing one byte of every segment
611338445  amd64 vbit-test fails with unknown opcodes used by arm64 VEX
612338499  --sim-hints parsing broken due to wrong order in tokens
613338615  suppress glibc 2.20 optimized strcmp implementation for ARMv7
614338681  Unable to unwind through clone thread created on i386-linux
615338698  race condition between gdbsrv and vgdb on startup
616338703  helgrind on arm-linux gets false positives in dynamic loader
617338791  alt dwz files can be relative of debug/main file
618338878  on MacOS: assertion 'VG_IS_PAGE_ALIGNED(clstack_end+1)' failed
619338932  build V-trunk with gcc-trunk
620338974  glibc 2.20 changed size of struct sigaction sa_flags field on s390
621345079  Fix build problems in VEX/useful/test_main.c
622n-i-bz  Fix KVM_CREATE_IRQCHIP ioctl handling
623n-i-bz  s390x: Fix memory corruption for multithreaded applications
624n-i-bz  vex arm->IR: allow PC as basereg in some LDRD cases
625n-i-bz  internal error in Valgrind if vgdb transmit signals when ptrace invoked
626n-i-bz  Fix mingw64 support in valgrind.h (dev@, 9 May 2014)
627n-i-bz  drd manual: Document how to C++11 programs that use class "std::thread"
628n-i-bz  Add command-line option --default-suppressions
629n-i-bz  Add support for BLKDISCARDZEROES ioctl
630n-i-bz  ppc32/64: fix a regression with the mtfsb0/mtfsb1 instructions
631n-i-bz  Add support for	sys_pivot_root and sys_unshare
632
633(3.10.0.BETA1:  2 September 2014, vex r2940, valgrind r14428)
634(3.10.0.BETA2:  8 September 2014, vex r2950, valgrind r14503)
635(3.10.0:       10 September 2014, vex r2950, valgrind r14514)
636
637
638
639Release 3.9.0 (31 October 2013)
640~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
6413.9.0 is a feature release with many improvements and the usual
642collection of bug fixes.
643
644This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux,
645PPC64/Linux, S390X/Linux, MIPS32/Linux, MIPS64/Linux, ARM/Android,
646X86/Android, X86/MacOSX 10.7 and AMD64/MacOSX 10.7.  Support for
647MacOSX 10.8 is significantly improved relative to the 3.8.0 release.
648
649* ================== PLATFORM CHANGES =================
650
651* Support for MIPS64 LE and BE running Linux.  Valgrind has been
652  tested on MIPS64 Debian Squeeze and Debian Wheezy distributions.
653
654* Support for MIPS DSP ASE on MIPS32 platforms.
655
656* Support for s390x Decimal Floating Point instructions on hosts that
657  have the DFP facility installed.
658
659* Support for POWER8 (Power ISA 2.07) instructions
660
661* Support for Intel AVX2 instructions.  This is available only on 64
662  bit code.
663
664* Initial support for Intel Transactional Synchronization Extensions,
665  both RTM and HLE.
666
667* Initial support for Hardware Transactional Memory on POWER.
668
669* Improved support for MacOSX 10.8 (64-bit only).  Memcheck can now
670  run large GUI apps tolerably well.
671
672* ==================== TOOL CHANGES ====================
673
674* Memcheck:
675
676  - Improvements in handling of vectorised code, leading to
677    significantly fewer false error reports.  You need to use the flag
678    --partial-loads-ok=yes to get the benefits of these changes.
679
680  - Better control over the leak checker.  It is now possible to
681    specify which leak kinds (definite/indirect/possible/reachable)
682    should be displayed, which should be regarded as errors, and which
683    should be suppressed by a given leak suppression.  This is done
684    using the options --show-leak-kinds=kind1,kind2,..,
685    --errors-for-leak-kinds=kind1,kind2,.. and an optional
686    "match-leak-kinds:" line in suppression entries, respectively.
687
688    Note that generated leak suppressions contain this new line and
689    are therefore more specific than in previous releases.  To get the
690    same behaviour as previous releases, remove the "match-leak-kinds:"
691    line from generated suppressions before using them.
692
693  - Reduced "possible leak" reports from the leak checker by the use
694    of better heuristics.  The available heuristics provide detection
695    of valid interior pointers to std::stdstring, to new[] allocated
696    arrays with elements having destructors and to interior pointers
697    pointing to an inner part of a C++ object using multiple
698    inheritance.  They can be selected individually using the
699    option --leak-check-heuristics=heur1,heur2,...
700
701  - Better control of stacktrace acquisition for heap-allocated
702    blocks.  Using the --keep-stacktraces option, it is possible to
703    control independently whether a stack trace is acquired for each
704    allocation and deallocation.  This can be used to create better
705    "use after free" errors or to decrease Valgrind's resource
706    consumption by recording less information.
707
708  - Better reporting of leak suppression usage.  The list of used
709    suppressions (shown when the -v option is given) now shows, for
710    each leak suppressions, how many blocks and bytes it suppressed
711    during the last leak search.
712
713* Helgrind:
714
715  - False errors resulting from the use of statically initialised
716    mutexes and condition variables (PTHREAD_MUTEX_INITIALISER, etc)
717    have been removed.
718
719  - False errors resulting from the use of pthread_cond_waits that
720    timeout, have been removed.
721
722* ==================== OTHER CHANGES ====================
723
724* Some attempt to tune Valgrind's space requirements to the expected
725  capabilities of the target:
726
727  - The default size of the translation cache has been reduced from 8
728    sectors to 6 on Android platforms, since each sector occupies
729    about 40MB when using Memcheck.
730
731  - The default size of the translation cache has been increased to 16
732    sectors on all other platforms, reflecting the fact that large
733    applications require instrumentation and storage of huge amounts
734    of code.  For similar reasons, the number of memory mapped
735    segments that can be tracked has been increased by a factor of 6.
736
737  - In all cases, the maximum number of sectors in the translation
738    cache can be controlled by the new flag --num-transtab-sectors.
739
740* Changes in how debug info (line numbers, etc) is read:
741
742  - Valgrind no longer temporarily mmaps the entire object to read
743    from it.  Instead, reading is done through a small fixed sized
744    buffer.  This avoids virtual memory usage spikes when Valgrind
745    reads debuginfo from large shared objects.
746
747  - A new experimental remote debug info server.  Valgrind can read
748    debug info from a different machine (typically, a build host)
749    where debuginfo objects are stored.  This can save a lot of time
750    and hassle when running Valgrind on resource-constrained targets
751    (phones, tablets) when the full debuginfo objects are stored
752    somewhere else.  This is enabled by the --debuginfo-server=
753    option.
754
755  - Consistency checking between main and debug objects can be
756    disabled using the --allow-mismatched-debuginfo option.
757
758* Stack unwinding by stack scanning, on ARM.  Unwinding by stack
759  scanning can recover stack traces in some cases when the normal
760  unwind mechanisms fail.  Stack scanning is best described as "a
761  nasty, dangerous and misleading hack" and so is disabled by default.
762  Use --unw-stack-scan-thresh and --unw-stack-scan-frames to enable
763  and control it.
764
765* Detection and merging of recursive stack frame cycles.  When your
766  program has recursive algorithms, this limits the memory used by
767  Valgrind for recorded stack traces and avoids recording
768  uninteresting repeated calls.  This is controlled by the command
769  line option --merge-recursive-frame and by the monitor command
770  "v.set merge-recursive-frames".
771
772* File name and line numbers for used suppressions.  The list of used
773  suppressions (shown when the -v option is given) now shows, for each
774  used suppression, the file name and line number where the suppression
775  is defined.
776
777* New and modified GDB server monitor features:
778
779  - valgrind.h has a new client request, VALGRIND_MONITOR_COMMAND,
780    that can be used to execute gdbserver monitor commands from the
781    client program.
782
783  - A new monitor command, "v.info open_fds", that gives the list of
784    open file descriptors and additional details.
785
786  - An optional message in the "v.info n_errs_found" monitor command,
787    for example "v.info n_errs_found test 1234 finished", allowing a
788    comment string to be added to the process output, perhaps for the
789    purpose of separating errors of different tests or test phases.
790
791  - A new monitor command "v.info execontext" that shows information
792    about the stack traces recorded by Valgrind.
793
794  - A new monitor command "v.do expensive_sanity_check_general" to run
795    some internal consistency checks.
796
797* New flag --sigill-diagnostics to control whether a diagnostic
798  message is printed when the JIT encounters an instruction it can't
799  translate.  The actual behavior -- delivery of SIGILL to the
800  application -- is unchanged.
801
802* The maximum amount of memory that Valgrind can use on 64 bit targets
803  has been increased from 32GB to 64GB.  This should make it possible
804  to run applications on Memcheck that natively require up to about 35GB.
805
806* ==================== FIXED BUGS ====================
807
808The following bugs have been fixed or resolved.  Note that "n-i-bz"
809stands for "not in bugzilla" -- that is, a bug that was reported to us
810but never got a bugzilla entry.  We encourage you to file bugs in
811bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather
812than mailing the developers (or mailing lists) directly -- bugs that
813are not entered into bugzilla tend to get forgotten about or ignored.
814
815To see details of a given bug, visit
816  https://bugs.kde.org/show_bug.cgi?id=XXXXXX
817where XXXXXX is the bug number as listed below.
818
819123837  system call: 4th argument is optional, depending on cmd
820135425  memcheck should tell you where Freed blocks were Mallocd
821164485  VG_N_SEGNAMES and VG_N_SEGMENTS are (still) too small
822207815  Adds some of the drm ioctls to syswrap-linux.c
823251569  vex amd64->IR: 0xF 0x1 0xF9 0xBF 0x90 0xD0 0x3 0x0 (RDTSCP)
824252955  Impossible to compile with ccache
825253519  Memcheck reports auxv pointer accesses as invalid reads.
826263034  Crash when loading some PPC64 binaries
827269599  Increase deepest backtrace
828274695  s390x: Support "compare to/from logical" instructions (z196)
829275800  s390x: Autodetect cache info (part 2)
830280271  Valgrind reports possible memory leaks on still-reachable std::string
831284540  Memcheck shouldn't count suppressions matching still-reachable [..]
832289578  Backtraces with ARM unwind tables (stack scan flags)
833296311  Wrong stack traces due to -fomit-frame-pointer (x86)
834304832  ppc32: build failure
835305431  Use find_buildid shdr fallback for separate .debug files
836305728  Add support for AVX2 instructions
837305948  ppc64: code generation for ShlD64 / ShrD64 asserts
838306035  s390x: Fix IR generation for LAAG and friends
839306054  s390x: Condition code computation for convert-to-int/logical
840306098  s390x: alternate opcode form for convert to/from fixed
841306587  Fix cache line detection from auxiliary vector for PPC.
842306783  Mips unhandled syscall :  4025  /  4079  / 4182
843307038  DWARF2 CFI reader: unhandled DW_OP_ opcode 0x8 (DW_OP_const1u et al)
844307082  HG false positive: pthread_cond_destroy: destruction of unknown CV
845307101  sys_capget second argument can be NULL
846307103  sys_openat: If pathname is absolute, then dirfd is ignored.
847307106  amd64->IR: f0 0f c0 02 (lock xadd byte)
848307113  s390x: DFP support
849307141  valgrind does't work in mips-linux system
850307155  filter_gdb should filter out syscall-template.S T_PSEUDO
851307285  x86_amd64 feature test for avx in test suite is wrong
852307290  memcheck overlap testcase needs memcpy version filter
853307463  Please add "&limit=0" to the "all open bugs" link
854307465  --show-possibly-lost=no should reduce the error count / exit code
855307557  Leaks on Mac OS X 10.7.5 libraries at ImageLoader::recursiveInit[..]
856307729  pkgconfig support broken valgrind.pc
857307828  Memcheck false errors SSE optimized wcscpy, wcscmp, wcsrchr, wcschr
858307955  Building valgrind 3.7.0-r4 fails in Gentoo AMD64 when using clang
859308089  Unhandled syscall on ppc64: prctl
860308135  PPC32 MPC8xx has 16 bytes cache size
861308321  testsuite memcheck filter interferes with gdb_filter
862308333 == 307106
863308341  vgdb should report process exit (or fatal signal)
864308427  s390 memcheck reports tsearch cjump/cmove depends on uninit
865308495  Remove build dependency on installed Xen headers
866308573  Internal error on 64-bit instruction executed in 32-bit mode
867308626  == 308627
868308627  pmovmskb validity bit propagation is imprecise
869308644  vgdb command for having the info for the track-fds option
870308711  give more info about aspacemgr and arenas in out_of_memory
871308717  ARM: implement fixed-point VCVT.F64.[SU]32
872308718  ARM implement SMLALBB family of instructions
873308886  Missing support for PTRACE_SET/GETREGSET
874308930  syscall name_to_handle_at (303 on amd64) not handled
875309229  V-bit tester does not report number of tests generated
876309323  print unrecognized instuction on MIPS
877309425  Provide a --sigill-diagnostics flag to suppress illegal [..]
878309427  SSE optimized stpncpy trigger uninitialised value [..] errors
879309430  Self hosting ppc64 encounters a vassert error on operand type
880309600  valgrind is a bit confused about 0-sized sections
881309823  Generate errors for still reachable blocks
882309921  PCMPISTRI validity bit propagation is imprecise
883309922  none/tests/ppc64/test_dfp5 sometimes fails
884310169  The Iop_CmpORD class of Iops is not supported by the vbit checker.
885310424  --read-var-info does not properly describe static variables
886310792  search additional path for debug symbols
887310931  s390x: Message-security assist (MSA) instruction extension [..]
888311100  PPC DFP implementation of the integer operands is inconsistent [..]
889311318  ARM: "128-bit constant is not implemented" error message
890311407  ssse3 bcopy (actually converted memcpy) causes invalid read [..]
891311690  V crashes because it redirects branches inside of a redirected function
892311880  x86_64: make regtest hangs at shell_valid1
893311922  WARNING: unhandled syscall: 170
894311933  == 251569
895312171  ppc: insn selection for DFP
896312571  Rounding mode call wrong for the DFP Iops [..]
897312620  Change to Iop_D32toD64 [..] for s390 DFP support broke ppc [..]
898312913  Dangling pointers error should also report the alloc stack trace
899312980  Building on Mountain Lion generates some compiler warnings
900313267  Adding MIPS64/Linux port to Valgrind
901313348  == 251569
902313354  == 251569
903313811  Buffer overflow in assert_fail
904314099  coverity pointed out error in VEX guest_ppc_toIR.c insn_suffix
905314269  ppc: dead code in insn selection
906314718  ARM: implement integer divide instruction (sdiv and udiv)
907315345  cl-format.xml and callgrind/dump.c don't agree on using cfl= or cfi=
908315441  sendmsg syscall should ignore unset msghdr msg_flags
909315534  msgrcv inside a thread causes valgrind to hang (block)
910315545  Assertion '(UChar*)sec->tt[tteNo].tcptr <= (UChar*)hcode' failed
911315689  disInstr(thumb): unhandled instruction: 0xF852 0x0E10 (LDRT)
912315738  disInstr(arm): unhandled instruction: 0xEEBE0BEE (vcvt.s32.f64)
913315959  valgrind man page has bogus SGCHECK (and no BBV) OPTIONS section
914316144  valgrind.1 manpage contains unknown ??? strings [..]
915316145  callgrind command line options in manpage reference (unknown) [..]
916316145  callgrind command line options in manpage reference [..]
917316181  drd: Fixed a 4x slowdown for certain applications
918316503  Valgrind does not support SSE4 "movntdqa" instruction
919316535  Use of |signed int| instead of |size_t| in valgrind messages
920316696   fluidanimate program of parsec 2.1 stuck
921316761  syscall open_by_handle_at (304 on amd64, 342 on x86) not handled
922317091  Use -Wl,-Ttext-segment when static linking if possible [..]
923317186  "Impossible happens" when occurs VCVT instruction on ARM
924317318  Support for Threading Building Blocks "scalable_malloc"
925317444  amd64->IR: 0xC4 0x41 0x2C 0xC2 0xD2 0x8 (vcmpeq_uqps)
926317461  Fix BMI assembler configure check and avx2/bmi/fma vgtest prereqs
927317463  bmi testcase IR SANITY CHECK FAILURE
928317506  memcheck/tests/vbit-test fails with unknown opcode after [..]
929318050  libmpiwrap fails to compile with out-of-source build
930318203  setsockopt handling needs to handle SOL_SOCKET/SO_ATTACH_FILTER
931318643  annotate_trace_memory tests infinite loop on arm and ppc [..]
932318773  amd64->IR: 0xF3 0x48 0x0F 0xBC 0xC2 0xC3 0x66 0x0F
933318929  Crash with: disInstr(thumb): 0xF321 0x0001 (ssat16)
934318932  Add missing PPC64 and PPC32 system call support
935319235  --db-attach=yes is broken with Yama (ptrace scoping) enabled
936319395  Crash with unhandled instruction on STRT (Thumb) instructions
937319494  VEX Makefile-gcc standalone build update after r2702
938319505  [MIPSEL] Crash: unhandled UNRAY operator.
939319858  disInstr(thumb): unhandled instruction on instruction STRBT
940319932  disInstr(thumb): unhandled instruction on instruction STRHT
941320057  Problems when we try to mmap more than 12 memory pages on MIPS32
942320063  Memory from PTRACE_GET_THREAD_AREA is reported uninitialised
943320083  disInstr(thumb): unhandled instruction on instruction LDRBT
944320116  bind on AF_BLUETOOTH produces warnings because of sockaddr_rc padding
945320131  WARNING: unhandled syscall: 369 on ARM (prlimit64)
946320211  Stack buffer overflow in ./coregrind/m_main.c with huge TMPDIR
947320661  vgModuleLocal_read_elf_debug_info(): "Assertion '!di->soname'
948320895  add fanotify support (patch included)
949320998  vex amd64->IR pcmpestri and pcmpestrm SSE4.2 instruction
950321065  Valgrind updates for Xen 4.3
951321148  Unhandled instruction: PLI (Thumb 1, 2, 3)
952321363  Unhandled instruction: SSAX (ARM + Thumb)
953321364  Unhandled instruction: SXTAB16 (ARM + Thumb)
954321466  Unhandled instruction: SHASX (ARM + Thumb)
955321467  Unhandled instruction: SHSAX (ARM + Thumb)
956321468  Unhandled instruction: SHSUB16 (ARM + Thumb)
957321619  Unhandled instruction: SHSUB8 (ARM + Thumb)
958321620  Unhandled instruction: UASX (ARM + Thumb)
959321621  Unhandled instruction: USAX (ARM + Thumb)
960321692  Unhandled instruction: UQADD16 (ARM + Thumb)
961321693  Unhandled instruction: LDRSBT (Thumb)
962321694  Unhandled instruction: UQASX (ARM + Thumb)
963321696  Unhandled instruction: UQSAX (Thumb + ARM)
964321697  Unhandled instruction: UHASX (ARM + Thumb)
965321703  Unhandled instruction: UHSAX (ARM + Thumb)
966321704  Unhandled instruction: REVSH (ARM + Thumb)
967321730  Add cg_diff and cg_merge man pages
968321738  Add vgdb and valgrind-listener man pages
969321814  == 315545
970321891  Unhandled instruction: LDRHT (Thumb)
971321960  pthread_create() then alloca() causing invalid stack write errors
972321969  ppc32 and ppc64 don't support [lf]setxattr
973322254  Show threadname together with tid if set by application
974322294  Add initial support for IBM Power ISA 2.07
975322368  Assertion failure in wqthread_hijack under OS X 10.8
976322563  vex mips->IR: 0x70 0x83 0xF0 0x3A
977322807  VALGRIND_PRINTF_BACKTRACE writes callstack to xml and text to stderr
978322851  0bXXX binary literal syntax is not standard
979323035  Unhandled instruction: LDRSHT(Thumb)
980323036  Unhandled instruction: SMMLS (ARM and Thumb)
981323116  The memcheck/tests/ppc64/power_ISA2_05.c fails to build [..]
982323175  Unhandled instruction: SMLALD (ARM + Thumb)
983323177  Unhandled instruction: SMLSLD (ARM + Thumb)
984323432  Calling pthread_cond_destroy() or pthread_mutex_destroy() [..]
985323437  Phase 2 support for IBM Power ISA 2.07
986323713  Support mmxext (integer sse) subset on i386 (athlon)
987323803  Transactional memory instructions are not supported for Power
988323893  SSE3 not available on amd cpus in valgrind
989323905  Probable false positive from Valgrind/drd on close()
990323912  valgrind.h header isn't compatible for mingw64
991324047  Valgrind doesn't support [LDR,ST]{S}[B,H]T ARM instructions
992324149  helgrind: When pthread_cond_timedwait returns ETIMEDOUT [..]
993324181  mmap does not handle MAP_32BIT
994324227  memcheck false positive leak when a thread calls exit+block [..]
995324421  Support for fanotify API on ARM architecture
996324514  gdbserver monitor cmd output behaviour consistency [..]
997324518  ppc64: Emulation of dcbt instructions does not handle [..]
998324546  none/tests/ppc32 test_isa_2_07_part2 requests -m64
999324582  When access is made to freed memory, report both allocation [..]
1000324594  Fix overflow computation for Power ISA 2.06 insns: mulldo/mulldo.
1001324765  ppc64: illegal instruction when executing none/tests/ppc64/jm-misc
1002324816  Incorrect VEX implementation for xscvspdp/xvcvspdp for SNaN inputs
1003324834  Unhandled instructions in Microsoft C run-time for x86_64
1004324894  Phase 3 support for IBM Power ISA 2.07
1005326091  drd: Avoid false race reports from optimized strlen() impls
1006326113  valgrind libvex hwcaps error on AMD64
1007n-i-bz  Some wrong command line options could be ignored
1008n-i-bz  patch to allow fair-sched on android
1009n-i-bz  report error for vgdb snapshot requested before execution
1010n-i-bz  same as 303624 (fixed in 3.8.0), but for x86 android
1011
1012(3.9.0: 31 October 2013, vex r2796, valgrind r13708)
1013
1014
1015
1016Release 3.8.1 (19 September 2012)
1017~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
10183.8.1 is a bug fix release.  It fixes some assertion failures in 3.8.0
1019that occur moderately frequently in real use cases, adds support for
1020some missing instructions on ARM, and fixes a deadlock condition on
1021MacOSX.  If you package or deliver 3.8.0 for others to use, you might
1022want to consider upgrading to 3.8.1 instead.
1023
1024The following bugs have been fixed or resolved.  Note that "n-i-bz"
1025stands for "not in bugzilla" -- that is, a bug that was reported to us
1026but never got a bugzilla entry.  We encourage you to file bugs in
1027bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather
1028than mailing the developers (or mailing lists) directly -- bugs that
1029are not entered into bugzilla tend to get forgotten about or ignored.
1030
1031To see details of a given bug, visit
1032  https://bugs.kde.org/show_bug.cgi?id=XXXXXX
1033where XXXXXX is the bug number as listed below.
1034
1035284004  == 301281
1036289584  Unhandled instruction: 0xF 0x29 0xE5 (MOVAPS)
1037295808  amd64->IR: 0xF3 0xF 0xBC 0xC0 (TZCNT)
1038298281  wcslen causes false(?) uninitialised value warnings
1039301281  valgrind hangs on OS X when the process calls system()
1040304035  disInstr(arm): unhandled instruction 0xE1023053
1041304867  implement MOVBE instruction in x86 mode
1042304980  Assertion 'lo <= hi' failed in vgModuleLocal_find_rx_mapping
1043305042  amd64: implement 0F 7F encoding of movq between two registers
1044305199  ARM: implement QDADD and QDSUB
1045305321  amd64->IR: 0xF 0xD 0xC (prefetchw)
1046305513  killed by fatal signal: SIGSEGV
1047305690  DRD reporting invalid semaphore when sem_trywait fails
1048305926  Invalid alignment checks for some AVX instructions
1049306297  disInstr(thumb): unhandled instruction 0xE883 0x000C
1050306310  3.8.0 release tarball missing some files
1051306612  RHEL 6 glibc-2.X default suppressions need /lib*/libc-*patterns
1052306664  vex amd64->IR: 0x66 0xF 0x3A 0x62 0xD1 0x46 0x66 0xF
1053n-i-bz  shmat of a segment > 4Gb does not work
1054n-i-bz  simulate_control_c script wrong USR1 signal number on mips
1055n-i-bz  vgdb ptrace calls wrong on mips [...]
1056n-i-bz  Fixes for more MPI false positives
1057n-i-bz  exp-sgcheck's memcpy causes programs to segfault
1058n-i-bz  OSX build w/ clang: asserts at startup
1059n-i-bz  Incorrect undef'dness prop for Iop_DPBtoBCD and Iop_BCDtoDPB
1060n-i-bz  fix a couple of union tag-vs-field mixups
1061n-i-bz  OSX: use __NR_poll_nocancel rather than __NR_poll
1062
1063The following bugs were fixed in 3.8.0 but not listed in this NEWS
1064file at the time:
1065
1066254088  Valgrind should know about UD2 instruction
1067301280  == 254088
1068301902  == 254088
1069304754  NEWS blows TeX's little mind
1070
1071(3.8.1: 19 September 2012, vex r2537, valgrind r12996)
1072
1073
1074
1075Release 3.8.0 (10 August 2012)
1076~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
10773.8.0 is a feature release with many improvements and the usual
1078collection of bug fixes.
1079
1080This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux,
1081PPC64/Linux, S390X/Linux, MIPS/Linux, ARM/Android, X86/Android,
1082X86/MacOSX 10.6/10.7 and AMD64/MacOSX 10.6/10.7.  Support for recent
1083distros and toolchain components (glibc 2.16, gcc 4.7) has been added.
1084There is initial support for MacOSX 10.8, but it is not usable for
1085serious work at present.
1086
1087* ================== PLATFORM CHANGES =================
1088
1089* Support for MIPS32 platforms running Linux.  Valgrind has been
1090  tested on MIPS32 and MIPS32r2 platforms running different Debian
1091  Squeeze and MeeGo distributions.  Both little-endian and big-endian
1092  cores are supported.  The tools Memcheck, Massif and Lackey have
1093  been tested and are known to work. See README.mips for more details.
1094
1095* Preliminary support for Android running on x86.
1096
1097* Preliminary (as-yet largely unusable) support for MacOSX 10.8.
1098
1099* Support for Intel AVX instructions and for AES instructions.  This
1100  support is available only for 64 bit code.
1101
1102* Support for POWER Decimal Floating Point instructions.
1103
1104* ==================== TOOL CHANGES ====================
1105
1106* Non-libc malloc implementations are now supported.  This is useful
1107  for tools that replace malloc (Memcheck, Massif, DRD, Helgrind).
1108  Using the new option --soname-synonyms, such tools can be informed
1109  that the malloc implementation is either linked statically into the
1110  executable, or is present in some other shared library different
1111  from libc.so.  This makes it possible to process statically linked
1112  programs, and programs using other malloc libraries, for example
1113  TCMalloc or JEMalloc.
1114
1115* For tools that provide their own replacement for malloc et al, the
1116  option --redzone-size=<number> allows users to specify the size of
1117  the padding blocks (redzones) added before and after each client
1118  allocated block.  Smaller redzones decrease the memory needed by
1119  Valgrind.  Bigger redzones increase the chance to detect blocks
1120  overrun or underrun.  Prior to this change, the redzone size was
1121  hardwired to 16 bytes in Memcheck.
1122
1123* Memcheck:
1124
1125  - The leak_check GDB server monitor command now can
1126    control the maximum nr of loss records to output.
1127
1128  - Reduction of memory use for applications allocating
1129    many blocks and/or having many partially defined bytes.
1130
1131  - Addition of GDB server monitor command 'block_list' that lists
1132    the addresses/sizes of the blocks of a leak search loss record.
1133
1134  - Addition of GDB server monitor command 'who_points_at' that lists
1135    the locations pointing at a block.
1136
1137  - If a redzone size > 0 is given, VALGRIND_MALLOCLIKE_BLOCK now will
1138    detect an invalid access of these redzones, by marking them
1139    noaccess.  Similarly, if a redzone size is given for a memory
1140    pool, VALGRIND_MEMPOOL_ALLOC will mark the redzones no access.
1141    This still allows to find some bugs if the user has forgotten to
1142    mark the pool superblock noaccess.
1143
1144  - Performance of memory leak check has been improved, especially in
1145    cases where there are many leaked blocks and/or many suppression
1146    rules used to suppress leak reports.
1147
1148  - Reduced noise (false positive) level on MacOSX 10.6/10.7, due to
1149    more precise analysis, which is important for LLVM/Clang
1150    generated code.  This is at the cost of somewhat reduced
1151    performance.  Note there is no change to analysis precision or
1152    costs on Linux targets.
1153
1154* DRD:
1155
1156  - Added even more facilities that can help finding the cause of a data
1157    race, namely the command-line option --ptrace-addr and the macro
1158    DRD_STOP_TRACING_VAR(x). More information can be found in the manual.
1159
1160  - Fixed a subtle bug that could cause false positive data race reports.
1161
1162* ==================== OTHER CHANGES ====================
1163
1164* The C++ demangler has been updated so as to work well with C++
1165  compiled by up to at least g++ 4.6.
1166
1167* Tool developers can make replacement/wrapping more flexible thanks
1168  to the new option --soname-synonyms.  This was reported above, but
1169  in fact is very general and applies to all function
1170  replacement/wrapping, not just to malloc-family functions.
1171
1172* Round-robin scheduling of threads can be selected, using the new
1173  option --fair-sched= yes.  Prior to this change, the pipe-based
1174  thread serialisation mechanism (which is still the default) could
1175  give very unfair scheduling.  --fair-sched=yes improves
1176  responsiveness of interactive multithreaded applications, and
1177  improves repeatability of results from the thread checkers Helgrind
1178  and DRD.
1179
1180* For tool developers: support to run Valgrind on Valgrind has been
1181  improved.  We can now routinely Valgrind on Helgrind or Memcheck.
1182
1183* gdbserver now shows the float shadow registers as integer
1184  rather than float values, as the shadow values are mostly
1185  used as bit patterns.
1186
1187* Increased limit for the --num-callers command line flag to 500.
1188
1189* Performance improvements for error matching when there are many
1190  suppression records in use.
1191
1192* Improved support for DWARF4 debugging information (bug 284184).
1193
1194* Initial support for DWZ compressed Dwarf debug info.
1195
1196* Improved control over the IR optimiser's handling of the tradeoff
1197  between performance and precision of exceptions.  Specifically,
1198  --vex-iropt-precise-memory-exns has been removed and replaced by
1199  --vex-iropt-register-updates, with extended functionality.  This
1200  allows the Valgrind gdbserver to always show up to date register
1201  values to GDB.
1202
1203* Modest performance gains through the use of translation chaining for
1204  JIT-generated code.
1205
1206* ==================== FIXED BUGS ====================
1207
1208The following bugs have been fixed or resolved.  Note that "n-i-bz"
1209stands for "not in bugzilla" -- that is, a bug that was reported to us
1210but never got a bugzilla entry.  We encourage you to file bugs in
1211bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather
1212than mailing the developers (or mailing lists) directly -- bugs that
1213are not entered into bugzilla tend to get forgotten about or ignored.
1214
1215To see details of a given bug, visit
1216  https://bugs.kde.org/show_bug.cgi?id=XXXXXX
1217where XXXXXX is the bug number as listed below.
1218
1219197914  Building valgrind from svn now requires automake-1.10
1220203877  increase to 16Mb maximum allowed alignment for memalign et al
1221219156  Handle statically linked malloc or other malloc lib (e.g. tcmalloc)
1222247386  make perf does not run all performance tests
1223270006  Valgrind scheduler unfair
1224270777  Adding MIPS/Linux port to Valgrind
1225270796  s390x: Removed broken support for the TS insn
1226271438  Fix configure for proper SSE4.2 detection
1227273114  s390x: Support TR, TRE, TROO, TROT, TRTO, and TRTT instructions
1228273475  Add support for AVX instructions
1229274078  improved configure logic for mpicc
1230276993  fix mremap 'no thrash checks'
1231278313  Fedora 15/x64: err read debug info with --read-var-info=yes flag
1232281482  memcheck incorrect byte allocation count in realloc() for silly argument
1233282230  group allocator for small fixed size, use it for MC_Chunk/SEc vbit
1234283413  Fix wrong sanity check
1235283671  Robustize alignment computation in LibVEX_Alloc
1236283961  Adding support for some HCI IOCTLs
1237284124  parse_type_DIE: confused by: DWARF 4
1238284864  == 273475 (Add support for AVX instructions)
1239285219  Too-restrictive constraints for Thumb2 "SP plus/minus register"
1240285662  (MacOSX): Memcheck needs to replace memcpy/memmove
1241285725  == 273475 (Add support for AVX instructions)
1242286261  add wrapper for linux I2C_RDWR ioctl
1243286270  vgpreload is not friendly to 64->32 bit execs, gives ld.so warnings
1244286374  Running cachegrind with --branch-sim=yes on 64-bit PowerPC program fails
1245286384  configure fails "checking for a supported version of gcc"
1246286497  == 273475 (Add support for AVX instructions)
1247286596  == 273475 (Add support for AVX instructions)
1248286917  disInstr(arm): unhandled instruction: QADD (also QSUB)
1249287175  ARM: scalar VFP fixed-point VCVT instructions not handled
1250287260  Incorrect conditional jump or move depends on uninitialised value(s)
1251287301  vex amd64->IR: 0x66 0xF 0x38 0x41 0xC0 0xB8 0x0 0x0 (PHMINPOSUW)
1252287307  == 273475 (Add support for AVX instructions)
1253287858  VG_(strerror): unknown error
1254288298  (MacOSX) unhandled syscall shm_unlink
1255288995  == 273475 (Add support for AVX instructions)
1256289470  Loading of large Mach-O thin binaries fails.
1257289656  == 273475 (Add support for AVX instructions)
1258289699  vgdb connection in relay mode erroneously closed due to buffer overrun
1259289823  ==  293754 (PCMPxSTRx not implemented for 16-bit characters)
1260289839  s390x: Provide support for unicode conversion instructions
1261289939  monitor cmd 'leak_check' with details about leaked or reachable blocks
1262290006  memcheck doesn't mark %xmm as initialized after "pcmpeqw %xmm %xmm"
1263290655  Add support for AESKEYGENASSIST instruction
1264290719  valgrind-3.7.0 fails with automake-1.11.2 due to"pkglibdir" usage
1265290974  vgdb must align pages to VKI_SHMLBA (16KB) on ARM
1266291253  ES register not initialised in valgrind simulation
1267291568  Fix 3DNOW-related crashes with baseline x86_64 CPU (w patch)
1268291865  s390x: Support the "Compare Double and Swap" family of instructions
1269292300  == 273475 (Add support for AVX instructions)
1270292430  unrecognized instruction in __intel_get_new_mem_ops_cpuid
1271292493  == 273475 (Add support for AVX instructions)
1272292626  Missing fcntl F_SETOWN_EX and F_GETOWN_EX support
1273292627  Missing support for some SCSI ioctls
1274292628  none/tests/x86/bug125959-x86.c triggers undefined behavior
1275292841  == 273475 (Add support for AVX instructions)
1276292993  implement the getcpu syscall on amd64-linux
1277292995  Implement the “cross memory attach” syscalls introduced in Linux 3.2
1278293088  Add some VEX sanity checks for ppc64 unhandled instructions
1279293751  == 290655 (Add support for AESKEYGENASSIST instruction)
1280293754  PCMPxSTRx not implemented for 16-bit characters
1281293755  == 293754 (No tests for PCMPxSTRx on 16-bit characters)
1282293808  CLFLUSH not supported by latest VEX for amd64
1283294047  valgrind does not correctly emulate prlimit64(..., RLIMIT_NOFILE, ...)
1284294048  MPSADBW instruction not implemented
1285294055  regtest none/tests/shell fails when locale is not set to C
1286294185  INT 0x44 (and others) not supported on x86 guest, but used by Jikes RVM
1287294190  --vgdb-error=xxx can be out of sync with errors shown to the user
1288294191  amd64: fnsave/frstor and 0x66 size prefixes on FP instructions
1289294260  disInstr_AMD64: disInstr miscalculated next %rip
1290294523  --partial-loads-ok=yes causes false negatives
1291294617  vex amd64->IR: 0x66 0xF 0x3A 0xDF 0xD1 0x1 0xE8 0x6A
1292294736  vex amd64->IR: 0x48 0xF 0xD7 0xD6 0x48 0x83
1293294812  patch allowing to run (on x86 at least) helgrind/drd on tool.
1294295089  can not annotate source for both helgrind and drd
1295295221  POWER Processor decimal floating point instruction support missing
1296295427  building for i386 with clang on darwin11 requires "-new_linker linker"
1297295428  coregrind/m_main.c has incorrect x86 assembly for darwin
1298295590  Helgrind: Assertion 'cvi->nWaiters > 0' failed
1299295617  ARM - Add some missing syscalls
1300295799  Missing \n with get_vbits in gdbserver when line is % 80 [...]
1301296229  Linux user input device ioctls missing wrappers
1302296318  ELF Debug info improvements (more than one rx/rw mapping)
1303296422  Add translation chaining support
1304296457  vex amd64->IR: 0x66 0xF 0x3A 0xDF 0xD1 0x1 0xE8 0x6A (dup of AES)
1305296792  valgrind 3.7.0: add SIOCSHWTSTAMP (0x89B0) ioctl wrapper
1306296983  Fix build issues on x86_64/ppc64 without 32-bit toolchains
1307297078  gdbserver signal handling problems [..]
1308297147  drd false positives on newly allocated memory
1309297329  disallow decoding of IBM Power DFP insns on some machines
1310297497  POWER Processor decimal floating point instruction support missing
1311297701  Another alias for strncasecmp_l in libc-2.13.so
1312297911  'invalid write' not reported when using APIs for custom mem allocators.
1313297976  s390x: revisit EX implementation
1314297991  Valgrind interferes with mmap()+ftell()
1315297992  Support systems missing WIFCONTINUED (e.g. pre-2.6.10 Linux)
1316297993  Fix compilation of valgrind with gcc -g3.
1317298080  POWER Processor DFP support missing, part 3
1318298227  == 273475 (Add support for AVX instructions)
1319298335  == 273475 (Add support for AVX instructions)
1320298354  Unhandled ARM Thumb instruction 0xEB0D 0x0585 (streq)
1321298394  s390x: Don't bail out on an unknown machine model.  [..]
1322298421  accept4() syscall (366) support is missing for ARM
1323298718  vex amd64->IR: 0xF 0xB1 0xCB 0x9C 0x8F 0x45
1324298732  valgrind installation problem in ubuntu with kernel version 3.x
1325298862  POWER Processor DFP instruction support missing, part 4
1326298864  DWARF reader mis-parses DW_FORM_ref_addr
1327298943  massif asserts with --pages-as-heap=yes when brk is changing [..]
1328299053  Support DWARF4 DW_AT_high_pc constant form
1329299104  == 273475 (Add support for AVX instructions)
1330299316  Helgrind: hg_main.c:628 (map_threads_lookup): Assertion 'thr' failed.
1331299629  dup3() syscall (358) support is missing for ARM
1332299694  POWER Processor DFP instruction support missing, part 5
1333299756  Ignore --free-fill for MEMPOOL_FREE and FREELIKE client requests
1334299803  == 273475 (Add support for AVX instructions)
1335299804  == 273475 (Add support for AVX instructions)
1336299805  == 273475 (Add support for AVX instructions)
1337300140  ARM - Missing (T1) SMMUL
1338300195  == 296318 (ELF Debug info improvements (more than one rx/rw mapping))
1339300389  Assertion `are_valid_hwcaps(VexArchAMD64, [..])' failed.
1340300414  FCOM and FCOMP unimplemented for amd64 guest
1341301204  infinite loop in canonicaliseSymtab with ifunc symbol
1342301229  == 203877 (increase to 16Mb maximum allowed alignment for memalign etc)
1343301265  add x86 support to Android build
1344301984  configure script doesn't detect certain versions of clang
1345302205  Fix compiler warnings for POWER VEX code and POWER test cases
1346302287  Unhandled movbe instruction on Atom processors
1347302370  PPC: fnmadd, fnmsub, fnmadds, fnmsubs insns always negate the result
1348302536  Fix for the POWER Valgrind regression test: memcheck-ISA2.0.
1349302578  Unrecognized isntruction 0xc5 0x32 0xc2 0xca 0x09 vcmpngess
1350302656  == 273475 (Add support for AVX instructions)
1351302709  valgrind for ARM needs extra tls support for android emulator [..]
1352302827  add wrapper for CDROM_GET_CAPABILITY
1353302901  Valgrind crashes with dwz optimized debuginfo
1354302918  Enable testing of the vmaddfp and vnsubfp instructions in the testsuite
1355303116  Add support for the POWER instruction popcntb
1356303127  Power test suite fixes for frsqrte, vrefp, and vrsqrtefp instructions.
1357303250  Assertion `instrs_in->arr_used <= 10000' failed w/ OpenSSL code
1358303466  == 273475 (Add support for AVX instructions)
1359303624  segmentation fault on Android 4.1 (e.g. on Galaxy Nexus OMAP)
1360303963  strstr() function produces wrong results under valgrind callgrind
1361304054  CALL_FN_xx macros need to enforce stack alignment
1362304561  tee system call not supported
1363715750  (MacOSX): Incorrect invalid-address errors near 0xFFFFxxxx (mozbug#)
1364n-i-bz  Add missing gdbserver xml files for shadow registers for ppc32
1365n-i-bz  Bypass gcc4.4/4.5 code gen bugs causing out of memory or asserts
1366n-i-bz  Fix assert in gdbserver for watchpoints watching the same address
1367n-i-bz  Fix false positive in sys_clone on amd64 when optional args [..]
1368n-i-bz  s390x: Shadow registers can now be examined using vgdb
1369
1370(3.8.0-TEST3:  9 August 2012, vex r2465, valgrind r12865)
1371(3.8.0:       10 August 2012, vex r2465, valgrind r12866)
1372
1373
1374
1375Release 3.7.0 (5 November 2011)
1376~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
13773.7.0 is a feature release with many significant improvements and the
1378usual collection of bug fixes.
1379
1380This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux,
1381PPC64/Linux, S390X/Linux, ARM/Android, X86/Darwin and AMD64/Darwin.
1382Support for recent distros and toolchain components (glibc 2.14, gcc
13834.6, MacOSX 10.7) has been added.
1384
1385* ================== PLATFORM CHANGES =================
1386
1387* Support for IBM z/Architecture (s390x) running Linux.  Valgrind can
1388  analyse 64-bit programs running on z/Architecture.  Most user space
1389  instructions up to and including z10 are supported.  Valgrind has
1390  been tested extensively on z9, z10, and z196 machines running SLES
1391  10/11, RedHat 5/6m, and Fedora. The Memcheck and Massif tools are
1392  known to work well. Callgrind, Helgrind, and DRD work reasonably
1393  well on z9 and later models. See README.s390 for more details.
1394
1395* Preliminary support for MacOSX 10.7 and XCode 4.  Both 32- and
1396  64-bit processes are supported.  Some complex threaded applications
1397  (Firefox) are observed to hang when run as 32 bit applications,
1398  whereas 64-bit versions run OK.  The cause is unknown.  Memcheck
1399  will likely report some false errors.  In general, expect some rough
1400  spots.  This release also supports MacOSX 10.6, but drops support
1401  for 10.5.
1402
1403* Preliminary support for Android (on ARM).  Valgrind can now run
1404  large applications (eg, Firefox) on (eg) a Samsung Nexus S.  See
1405  README.android for more details, plus instructions on how to get
1406  started.
1407
1408* Support for the IBM Power ISA 2.06 (Power7 instructions)
1409
1410* General correctness and performance improvements for ARM/Linux, and,
1411  by extension, ARM/Android.
1412
1413* Further solidification of support for SSE 4.2 in 64-bit mode.  AVX
1414  instruction set support is under development but is not available in
1415  this release.
1416
1417* Support for AIX5 has been removed.
1418
1419* ==================== TOOL CHANGES ====================
1420
1421* Memcheck: some incremental changes:
1422
1423  - reduction of memory use in some circumstances
1424
1425  - improved handling of freed memory, which in some circumstances
1426    can cause detection of use-after-free that would previously have
1427    been missed
1428
1429  - fix of a longstanding bug that could cause false negatives (missed
1430    errors) in programs doing vector saturated narrowing instructions.
1431
1432* Helgrind: performance improvements and major memory use reductions,
1433  particularly for large, long running applications which perform many
1434  synchronisation (lock, unlock, etc) events.  Plus many smaller
1435  changes:
1436
1437  - display of locksets for both threads involved in a race
1438
1439  - general improvements in formatting/clarity of error messages
1440
1441  - addition of facilities and documentation regarding annotation
1442    of thread safe reference counted C++ classes
1443
1444  - new flag --check-stack-refs=no|yes [yes], to disable race checking
1445    on thread stacks (a performance hack)
1446
1447  - new flag --free-is-write=no|yes [no], to enable detection of races
1448    where one thread accesses heap memory but another one frees it,
1449    without any coordinating synchronisation event
1450
1451* DRD: enabled XML output; added support for delayed thread deletion
1452  in order to detect races that occur close to the end of a thread
1453  (--join-list-vol); fixed a memory leak triggered by repeated client
1454  memory allocatation and deallocation; improved Darwin support.
1455
1456* exp-ptrcheck: this tool has been renamed to exp-sgcheck
1457
1458* exp-sgcheck: this tool has been reduced in scope so as to improve
1459  performance and remove checking that Memcheck does better.
1460  Specifically, the ability to check for overruns for stack and global
1461  arrays is unchanged, but the ability to check for overruns of heap
1462  blocks has been removed.  The tool has accordingly been renamed to
1463  exp-sgcheck ("Stack and Global Array Checking").
1464
1465* ==================== OTHER CHANGES ====================
1466
1467* GDB server: Valgrind now has an embedded GDB server.  That means it
1468  is possible to control a Valgrind run from GDB, doing all the usual
1469  things that GDB can do (single stepping, breakpoints, examining
1470  data, etc).  Tool-specific functionality is also available.  For
1471  example, it is possible to query the definedness state of variables
1472  or memory from within GDB when running Memcheck; arbitrarily large
1473  memory watchpoints are supported, etc.  To use the GDB server, start
1474  Valgrind with the flag --vgdb-error=0 and follow the on-screen
1475  instructions.
1476
1477* Improved support for unfriendly self-modifying code: a new option
1478  --smc-check=all-non-file is available.  This adds the relevant
1479  consistency checks only to code that originates in non-file-backed
1480  mappings.  In effect this confines the consistency checking only to
1481  code that is or might be JIT generated, and avoids checks on code
1482  that must have been compiled ahead of time.  This significantly
1483  improves performance on applications that generate code at run time.
1484
1485* It is now possible to build a working Valgrind using Clang-2.9 on
1486  Linux.
1487
1488* new client requests VALGRIND_{DISABLE,ENABLE}_ERROR_REPORTING.
1489  These enable and disable error reporting on a per-thread, and
1490  nestable, basis.  This is useful for hiding errors in particularly
1491  troublesome pieces of code.  The MPI wrapper library (libmpiwrap.c)
1492  now uses this facility.
1493
1494* Added the --mod-funcname option to cg_diff.
1495
1496* ==================== FIXED BUGS ====================
1497
1498The following bugs have been fixed or resolved.  Note that "n-i-bz"
1499stands for "not in bugzilla" -- that is, a bug that was reported to us
1500but never got a bugzilla entry.  We encourage you to file bugs in
1501bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
1502mailing the developers (or mailing lists) directly -- bugs that are
1503not entered into bugzilla tend to get forgotten about or ignored.
1504
1505To see details of a given bug, visit
1506https://bugs.kde.org/show_bug.cgi?id=XXXXXX
1507where XXXXXX is the bug number as listed below.
1508
1509 79311  malloc silly arg warning does not give stack trace
1510210935  port valgrind.h (not valgrind) to win32 to support client requests
1511214223  valgrind SIGSEGV on startup gcc 4.4.1 ppc32 (G4) Ubuntu 9.10
1512243404  Port to zSeries
1513243935  Helgrind: incorrect handling of ANNOTATE_HAPPENS_BEFORE()/AFTER()
1514247223  non-x86: Suppress warning: 'regparm' attribute directive ignored
1515250101  huge "free" memory usage due to m_mallocfree.c fragmentation
1516253206  Some fixes for the faultstatus testcase
1517255223  capget testcase fails when running as root
1518256703  xlc_dbl_u32.c testcase broken
1519256726  Helgrind tests have broken inline asm
1520259977  == 214223 (Valgrind segfaults doing __builtin_longjmp)
1521264800  testcase compile failure on zseries
1522265762  make public VEX headers compilable by G++ 3.x
1523265771  assertion in jumps.c (r11523) fails with glibc-2.3
1524266753  configure script does not give the user the option to not use QtCore
1525266931  gen_insn_test.pl is broken
1526266961  ld-linux.so.2 i?86-linux strlen issues
1527266990  setns instruction causes false positive
1528267020  Make directory for temporary files configurable at run-time.
1529267342  == 267997 (segmentation fault on Mac OS 10.6)
1530267383  Assertion 'vgPlain_strlen(dir) + vgPlain_strlen(file) + 1 < 256' failed
1531267413  Assertion 'DRD_(g_threadinfo)[tid].synchr_nesting >= 1' failed.
1532267488  regtest: darwin support for 64-bit build
1533267552  SIGSEGV (misaligned_stack_error) with DRD, but not with other tools
1534267630  Add support for IBM Power ISA 2.06 -- stage 1
1535267769  == 267997 (Darwin: memcheck triggers segmentation fault)
1536267819  Add client request for informing the core about reallocation
1537267925  laog data structure quadratic for a single sequence of lock
1538267968  drd: (vgDrd_thread_set_joinable): Assertion '0 <= (int)tid ..' failed
1539267997  MacOSX: 64-bit V segfaults on launch when built with Xcode 4.0.1
1540268513  missed optimizations in fold_Expr
1541268619  s390x: fpr - gpr transfer facility
1542268620  s390x: reconsider "long displacement" requirement
1543268621  s390x: improve IR generation for XC
1544268715  s390x: FLOGR is not universally available
1545268792  == 267997 (valgrind seg faults on startup when compiled with Xcode 4)
1546268930  s390x: MHY is not universally available
1547269078  arm->IR: unhandled instruction SUB (SP minus immediate/register)
1548269079  Support ptrace system call on ARM
1549269144  missing "Bad option" error message
1550269209  conditional load and store facility (z196)
1551269354  Shift by zero on x86 can incorrectly clobber CC_NDEP
1552269641  == 267997 (valgrind segfaults immediately (segmentation fault))
1553269736  s390x: minor code generation tweaks
1554269778  == 272986 (valgrind.h: swap roles of VALGRIND_DO_CLIENT_REQUEST() ..)
1555269863  s390x: remove unused function parameters
1556269864  s390x: tweak s390_emit_load_cc
1557269884  == 250101 (overhead for huge blocks exhausts space too soon)
1558270082  s390x: Make sure to point the PSW address to the next address on SIGILL
1559270115  s390x: rewrite some testcases
1560270309  == 267997 (valgrind crash on startup)
1561270320  add support for Linux FIOQSIZE ioctl() call
1562270326  segfault while trying to sanitize the environment passed to execle
1563270794  IBM POWER7 support patch causes regression in none/tests
1564270851  IBM POWER7 fcfidus instruction causes memcheck to fail
1565270856  IBM POWER7 xsnmaddadp instruction causes memcheck to fail on 32bit app
1566270925  hyper-optimized strspn() in /lib64/libc-2.13.so needs fix
1567270959  s390x: invalid use of R0 as base register
1568271042  VSX configure check fails when it should not
1569271043  Valgrind build fails with assembler error on ppc64 with binutils 2.21
1570271259  s390x: fix code confusion
1571271337  == 267997 (Valgrind segfaults on MacOS X)
1572271385  s390x: Implement Ist_MBE
1573271501  s390x: misc cleanups
1574271504  s390x: promote likely and unlikely
1575271579  ppc: using wrong enum type
1576271615  unhandled instruction "popcnt" (arch=amd10h)
1577271730  Fix bug when checking ioctls: duplicate check
1578271776  s390x: provide STFLE instruction support
1579271779  s390x: provide clock instructions like STCK
1580271799  Darwin: ioctls without an arg report a memory error
1581271820  arm: fix type confusion
1582271917  pthread_cond_timedwait failure leads to not-locked false positive
1583272067  s390x: fix DISP20 macro
1584272615  A typo in debug output in mc_leakcheck.c
1585272661  callgrind_annotate chokes when run from paths containing regex chars
1586272893  amd64->IR: 0x66 0xF 0x38 0x2B 0xC1 0x66 0xF 0x7F == (closed as dup)
1587272955  Unhandled syscall error for pwrite64 on ppc64 arch
1588272967  make documentation build-system more robust
1589272986  Fix gcc-4.6 warnings with valgrind.h
1590273318  amd64->IR: 0x66 0xF 0x3A 0x61 0xC1 0x38 (missing PCMPxSTRx case)
1591273318  unhandled PCMPxSTRx case: vex amd64->IR: 0x66 0xF 0x3A 0x61 0xC1 0x38
1592273431  valgrind segfaults in evalCfiExpr (debuginfo.c:2039)
1593273465  Callgrind: jumps.c:164 (new_jcc): Assertion '(0 <= jmp) && ...'
1594273536  Build error: multiple definition of `vgDrd_pthread_cond_initializer'
1595273640  ppc64-linux: unhandled syscalls setresuid(164) and setresgid(169)
1596273729  == 283000 (Illegal opcode for SSE2 "roundsd" instruction)
1597273778  exp-ptrcheck: unhandled sysno == 259
1598274089  exp-ptrcheck: unhandled sysno == 208
1599274378  s390x: Various dispatcher tweaks
1600274447  WARNING: unhandled syscall: 340
1601274776  amd64->IR: 0x66 0xF 0x38 0x2B 0xC5 0x66
1602274784  == 267997 (valgrind ls -l results in Segmentation Fault)
1603274926  valgrind does not build against linux-3
1604275148  configure FAIL with glibc-2.14
1605275151  Fedora 15 / glibc-2.14 'make regtest' FAIL
1606275168  Make Valgrind work for MacOSX 10.7 Lion
1607275212  == 275284 (lots of false positives from __memcpy_ssse3_back et al)
1608275278  valgrind does not build on Linux kernel 3.0.* due to silly
1609275284  Valgrind memcpy/memmove redirection stopped working in glibc 2.14/x86_64
1610275308  Fix implementation for ppc64 fres instruc
1611275339  s390x: fix testcase compile warnings
1612275517  s390x: Provide support for CKSM instruction
1613275710  s390x: get rid of redundant address mode calculation
1614275815  == 247894 (Valgrind doesn't know about Linux readahead(2) syscall)
1615275852  == 250101 (valgrind uses all swap space and is killed)
1616276784  Add support for IBM Power ISA 2.06 -- stage 3
1617276987  gdbsrv: fix tests following recent commits
1618277045  Valgrind crashes with  unhandled DW_OP_ opcode 0x2a
1619277199  The test_isa_2_06_part1.c in none/tests/ppc64 should be a symlink
1620277471  Unhandled syscall: 340
1621277610  valgrind crashes in VG_(lseek)(core_fd, phdrs[idx].p_offset, ...)
1622277653  ARM: support Thumb2 PLD instruction
1623277663  ARM: NEON float VMUL by scalar incorrect
1624277689  ARM: tests for VSTn with register post-index are broken
1625277694  ARM: BLX LR instruction broken in ARM mode
1626277780  ARM: VMOV.F32 (immediate) instruction is broken
1627278057  fuse filesystem syscall deadlocks
1628278078  Unimplemented syscall 280 on ppc32
1629278349  F_GETPIPE_SZ and  F_SETPIPE_SZ Linux fcntl commands
1630278454  VALGRIND_STACK_DEREGISTER has wrong output type
1631278502  == 275284 (Valgrind confuses memcpy() and memmove())
1632278892  gdbsrv: factorize gdb version handling, fix doc and typos
1633279027  Support for MVCL and CLCL instruction
1634279027  s390x: Provide support for CLCL and MVCL instructions
1635279062  Remove a redundant check in the insn selector for ppc.
1636279071  JDK creates PTEST with redundant REX.W prefix
1637279212  gdbsrv: add monitor cmd v.info scheduler.
1638279378  exp-ptrcheck: the 'impossible' happened on mkfifo call
1639279698  memcheck discards valid-bits for packuswb
1640279795  memcheck reports uninitialised values for mincore on amd64
1641279994  Add support for IBM Power ISA 2.06 -- stage 3
1642280083  mempolicy syscall check errors
1643280290  vex amd64->IR: 0x66 0xF 0x38 0x28 0xC1 0x66 0xF 0x6F
1644280710  s390x: config files for nightly builds
1645280757  /tmp dir still used by valgrind even if TMPDIR is specified
1646280965  Valgrind breaks fcntl locks when program does mmap
1647281138  WARNING: unhandled syscall: 340
1648281241  == 275168 (valgrind useless on Macos 10.7.1 Lion)
1649281304  == 275168 (Darwin: dyld "cannot load inserted library")
1650281305  == 275168 (unhandled syscall: unix:357 on Darwin 11.1)
1651281468  s390x: handle do_clone and gcc clones in call traces
1652281488  ARM: VFP register corruption
1653281828  == 275284 (false memmove warning: "Source and destination overlap")
1654281883  s390x: Fix system call wrapper for "clone".
1655282105  generalise 'reclaimSuperBlock' to also reclaim splittable superblock
1656282112  Unhandled instruction bytes: 0xDE 0xD9 0x9B 0xDF (fcompp)
1657282238  SLES10: make check fails
1658282979  strcasestr needs replacement with recent(>=2.12) glibc
1659283000  vex amd64->IR: 0x66 0xF 0x3A 0xA 0xC0 0x9 0xF3 0xF
1660283243  Regression in ppc64 memcheck tests
1661283325  == 267997 (Darwin: V segfaults on startup when built with Xcode 4.0)
1662283427  re-connect epoll_pwait syscall on ARM linux
1663283600  gdbsrv: android: port vgdb.c
1664283709  none/tests/faultstatus needs to account for page size
1665284305  filter_gdb needs enhancement to work on ppc64
1666284384  clang 3.1 -Wunused-value warnings in valgrind.h, memcheck.h
1667284472  Thumb2 ROR.W encoding T2 not implemented
1668284621  XML-escape process command line in XML output
1669n-i-bz  cachegrind/callgrind: handle CPUID information for Core iX Intel CPUs
1670        that have non-power-of-2 sizes (also AMDs)
1671n-i-bz  don't be spooked by libraries mashed by elfhack
1672n-i-bz  don't be spooked by libxul.so linked with gold
1673n-i-bz  improved checking for VALGRIND_CHECK_MEM_IS_DEFINED
1674
1675(3.7.0-TEST1: 27  October 2011, vex r2228, valgrind r12245)
1676(3.7.0.RC1:    1 November 2011, vex r2231, valgrind r12257)
1677(3.7.0:        5 November 2011, vex r2231, valgrind r12258)
1678
1679
1680
1681Release 3.6.1 (16 February 2011)
1682~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
16833.6.1 is a bug fix release.  It adds support for some SSE4
1684instructions that were omitted in 3.6.0 due to lack of time.  Initial
1685support for glibc-2.13 has been added.  A number of bugs causing
1686crashing or assertion failures have been fixed.
1687
1688The following bugs have been fixed or resolved.  Note that "n-i-bz"
1689stands for "not in bugzilla" -- that is, a bug that was reported to us
1690but never got a bugzilla entry.  We encourage you to file bugs in
1691bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
1692mailing the developers (or mailing lists) directly -- bugs that are
1693not entered into bugzilla tend to get forgotten about or ignored.
1694
1695To see details of a given bug, visit
1696https://bugs.kde.org/show_bug.cgi?id=XXXXXX
1697where XXXXXX is the bug number as listed below.
1698
1699188572  Valgrind on Mac should suppress setenv() mem leak
1700194402  vex amd64->IR: 0x48 0xF 0xAE 0x4 (proper FX{SAVE,RSTOR} support)
1701210481  vex amd64->IR: Assertion `sz == 2 || sz == 4' failed (REX.W POPQ)
1702246152  callgrind internal error after pthread_cancel on 32 Bit Linux
1703250038  ppc64: Altivec LVSR and LVSL instructions fail their regtest
1704254420  memory pool tracking broken
1705254957  Test code failing to compile due to changes in memcheck.h
1706255009  helgrind/drd: crash on chmod with invalid parameter
1707255130  readdwarf3.c parse_type_DIE confused by GNAT Ada types
1708255355  helgrind/drd: crash on threaded programs doing fork
1709255358  == 255355
1710255418  (SSE4.x) rint call compiled with ICC
1711255822  --gen-suppressions can create invalid files: "too many callers [...]"
1712255888  closing valgrindoutput tag outputted to log-stream on error
1713255963  (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x9 0xDB 0x0 (ROUNDPD)
1714255966  Slowness when using mempool annotations
1715256387  vex x86->IR: 0xD4 0xA 0x2 0x7 (AAD and AAM)
1716256600  super-optimized strcasecmp() false positive
1717256669  vex amd64->IR: Unhandled LOOPNEL insn on amd64
1718256968  (SSE4.x) vex amd64->IR: 0x66 0xF 0x38 0x10 0xD3 0x66 (BLENDVPx)
1719257011  (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0xE 0xFD 0xA0 (PBLENDW)
1720257063  (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x8 0xC0 0x0 (ROUNDPS)
1721257276  Missing case in memcheck --track-origins=yes
1722258870  (SSE4.x) Add support for EXTRACTPS SSE 4.1 instruction
1723261966  (SSE4.x) support for CRC32B and CRC32Q is lacking (also CRC32{W,L})
1724262985  VEX regression in valgrind 3.6.0 in handling PowerPC VMX
1725262995  (SSE4.x) crash when trying to valgrind gcc-snapshot (PCMPxSTRx $0)
1726263099  callgrind_annotate counts Ir improperly [...]
1727263877  undefined coprocessor instruction on ARMv7
1728265964  configure FAIL with glibc-2.13
1729n-i-bz  Fix compile error w/ icc-12.x in guest_arm_toIR.c
1730n-i-bz  Docs: fix bogus descriptions for VALGRIND_CREATE_BLOCK et al
1731n-i-bz  Massif: don't assert on shmat() with --pages-as-heap=yes
1732n-i-bz  Bug fixes and major speedups for the exp-DHAT space profiler
1733n-i-bz  DRD: disable --free-is-write due to implementation difficulties
1734
1735(3.6.1: 16 February 2011, vex r2103, valgrind r11561).
1736
1737
1738
1739Release 3.6.0 (21 October 2010)
1740~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
17413.6.0 is a feature release with many significant improvements and the
1742usual collection of bug fixes.
1743
1744This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux,
1745PPC64/Linux, X86/Darwin and AMD64/Darwin.  Support for recent distros
1746and toolchain components (glibc 2.12, gcc 4.5, OSX 10.6) has been added.
1747
1748                    -------------------------
1749
1750Here are some highlights.  Details are shown further down:
1751
1752* Support for ARM/Linux.
1753
1754* Support for recent Linux distros: Ubuntu 10.10 and Fedora 14.
1755
1756* Support for Mac OS X 10.6, both 32- and 64-bit executables.
1757
1758* Support for the SSE4.2 instruction set.
1759
1760* Enhancements to the Callgrind profiler, including the ability to
1761  handle CPUs with three levels of cache.
1762
1763* A new experimental heap profiler, DHAT.
1764
1765* A huge number of bug fixes and small enhancements.
1766
1767                    -------------------------
1768
1769Here are details of the above changes, together with descriptions of
1770many other changes, and a list of fixed bugs.
1771
1772* ================== PLATFORM CHANGES =================
1773
1774* Support for ARM/Linux.  Valgrind now runs on ARMv7 capable CPUs
1775  running Linux.  It is known to work on Ubuntu 10.04, Ubuntu 10.10,
1776  and Maemo 5, so you can run Valgrind on your Nokia N900 if you want.
1777
1778  This requires a CPU capable of running the ARMv7-A instruction set
1779  (Cortex A5, A8 and A9).  Valgrind provides fairly complete coverage
1780  of the user space instruction set, including ARM and Thumb integer
1781  code, VFPv3, NEON and V6 media instructions.  The Memcheck,
1782  Cachegrind and Massif tools work properly; other tools work to
1783  varying degrees.
1784
1785* Support for recent Linux distros (Ubuntu 10.10 and Fedora 14), along
1786  with support for recent releases of the underlying toolchain
1787  components, notably gcc-4.5 and glibc-2.12.
1788
1789* Support for Mac OS X 10.6, both 32- and 64-bit executables.  64-bit
1790  support also works much better on OS X 10.5, and is as solid as
1791  32-bit support now.
1792
1793* Support for the SSE4.2 instruction set.  SSE4.2 is supported in
1794  64-bit mode.  In 32-bit mode, support is only available up to and
1795  including SSSE3.  Some exceptions: SSE4.2 AES instructions are not
1796  supported in 64-bit mode, and 32-bit mode does in fact support the
1797  bare minimum SSE4 instructions to needed to run programs on Mac OS X
1798  10.6 on 32-bit targets.
1799
1800* Support for IBM POWER6 cpus has been improved.  The Power ISA up to
1801  and including version 2.05 is supported.
1802
1803* ==================== TOOL CHANGES ====================
1804
1805* Cachegrind has a new processing script, cg_diff, which finds the
1806  difference between two profiles.  It's very useful for evaluating
1807  the performance effects of a change in a program.
1808
1809  Related to this change, the meaning of cg_annotate's (rarely-used)
1810  --threshold option has changed; this is unlikely to affect many
1811  people, if you do use it please see the user manual for details.
1812
1813* Callgrind now can do branch prediction simulation, similar to
1814  Cachegrind.  In addition, it optionally can count the number of
1815  executed global bus events.  Both can be used for a better
1816  approximation of a "Cycle Estimation" as derived event (you need to
1817  update the event formula in KCachegrind yourself).
1818
1819* Cachegrind and Callgrind now refer to the LL (last-level) cache
1820  rather than the L2 cache.  This is to accommodate machines with
1821  three levels of caches -- if Cachegrind/Callgrind auto-detects the
1822  cache configuration of such a machine it will run the simulation as
1823  if the L2 cache isn't present.  This means the results are less
1824  likely to match the true result for the machine, but
1825  Cachegrind/Callgrind's results are already only approximate, and
1826  should not be considered authoritative.  The results are still
1827  useful for giving a general idea about a program's locality.
1828
1829* Massif has a new option, --pages-as-heap, which is disabled by
1830  default.  When enabled, instead of tracking allocations at the level
1831  of heap blocks (as allocated with malloc/new/new[]), it instead
1832  tracks memory allocations at the level of memory pages (as mapped by
1833  mmap, brk, etc).  Each mapped page is treated as its own block.
1834  Interpreting the page-level output is harder than the heap-level
1835  output, but this option is useful if you want to account for every
1836  byte of memory used by a program.
1837
1838* DRD has two new command-line options: --free-is-write and
1839  --trace-alloc.  The former allows to detect reading from already freed
1840  memory, and the latter allows tracing of all memory allocations and
1841  deallocations.
1842
1843* DRD has several new annotations.  Custom barrier implementations can
1844  now be annotated, as well as benign races on static variables.
1845
1846* DRD's happens before / happens after annotations have been made more
1847  powerful, so that they can now also be used to annotate e.g. a smart
1848  pointer implementation.
1849
1850* Helgrind's annotation set has also been drastically improved, so as
1851  to provide to users a general set of annotations to describe locks,
1852  semaphores, barriers and condition variables.  Annotations to
1853  describe thread-safe reference counted heap objects have also been
1854  added.
1855
1856* Memcheck has a new command-line option, --show-possibly-lost, which
1857  is enabled by default.  When disabled, the leak detector will not
1858  show possibly-lost blocks.
1859
1860* A new experimental heap profiler, DHAT (Dynamic Heap Analysis Tool),
1861  has been added.  DHAT keeps track of allocated heap blocks, and also
1862  inspects every memory reference to see which block (if any) is being
1863  accessed.  This gives a lot of insight into block lifetimes,
1864  utilisation, turnover, liveness, and the location of hot and cold
1865  fields.  You can use DHAT to do hot-field profiling.
1866
1867* ==================== OTHER CHANGES ====================
1868
1869* Improved support for unfriendly self-modifying code: the extra
1870  overhead incurred by --smc-check=all has been reduced by
1871  approximately a factor of 5 as compared with 3.5.0.
1872
1873* Ability to show directory names for source files in error messages.
1874  This is combined with a flexible mechanism for specifying which
1875  parts of the paths should be shown.  This is enabled by the new flag
1876  --fullpath-after.
1877
1878* A new flag, --require-text-symbol, which will stop the run if a
1879  specified symbol is not found it a given shared object when it is
1880  loaded into the process.  This makes advanced working with function
1881  intercepting and wrapping safer and more reliable.
1882
1883* Improved support for the Valkyrie GUI, version 2.0.0.  GUI output
1884  and control of Valgrind is now available for the tools Memcheck and
1885  Helgrind.  XML output from Valgrind is available for Memcheck,
1886  Helgrind and exp-Ptrcheck.
1887
1888* More reliable stack unwinding on amd64-linux, particularly in the
1889  presence of function wrappers, and with gcc-4.5 compiled code.
1890
1891* Modest scalability (performance improvements) for massive
1892  long-running applications, particularly for those with huge amounts
1893  of code.
1894
1895* Support for analyzing programs running under Wine with has been
1896  improved.  The header files <valgrind/valgrind.h>,
1897  <valgrind/memcheck.h> and <valgrind/drd.h> can now be used in
1898  Windows-programs compiled with MinGW or one of the Microsoft Visual
1899  Studio compilers.
1900
1901* A rare but serious error in the 64-bit x86 CPU simulation was fixed.
1902  The 32-bit simulator was not affected.  This did not occur often,
1903  but when it did would usually crash the program under test.
1904  Bug 245925.
1905
1906* A large number of bugs were fixed.  These are shown below.
1907
1908* A number of bugs were investigated, and were candidates for fixing,
1909  but are not fixed in 3.6.0, due to lack of developer time.  They may
1910  get fixed in later releases.  They are:
1911
1912  194402  vex amd64->IR: 0x48 0xF 0xAE 0x4 0x24 0x49  (FXSAVE64)
1913  212419  false positive "lock order violated" (A+B vs A)
1914  213685  Undefined value propagates past dependency breaking instruction
1915  216837  Incorrect instrumentation of NSOperationQueue on Darwin
1916  237920  valgrind segfault on fork failure
1917  242137  support for code compiled by LLVM-2.8
1918  242423  Another unknown Intel cache config value
1919  243232  Inconsistent Lock Orderings report with trylock
1920  243483  ppc: callgrind triggers VEX assertion failure
1921  243935  Helgrind: implementation of ANNOTATE_HAPPENS_BEFORE() is wrong
1922  244677  Helgrind crash hg_main.c:616 (map_threads_lookup): Assertion
1923          'thr' failed.
1924  246152  callgrind internal error after pthread_cancel on 32 Bit Linux
1925  249435  Analyzing wine programs with callgrind triggers a crash
1926  250038  ppc64: Altivec lvsr and lvsl instructions fail their regtest
1927  250065  Handling large allocations
1928  250101  huge "free" memory usage due to m_mallocfree.c
1929          "superblocks fragmentation"
1930  251569  vex amd64->IR: 0xF 0x1 0xF9 0x8B 0x4C 0x24 (RDTSCP)
1931  252091  Callgrind on ARM does not detect function returns correctly
1932  252600  [PATCH] Allow lhs to be a pointer for shl/shr
1933  254420  memory pool tracking broken
1934  n-i-bz  support for adding symbols for JIT generated code
1935
1936
1937The following bugs have been fixed or resolved.  Note that "n-i-bz"
1938stands for "not in bugzilla" -- that is, a bug that was reported to us
1939but never got a bugzilla entry.  We encourage you to file bugs in
1940bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
1941mailing the developers (or mailing lists) directly -- bugs that are
1942not entered into bugzilla tend to get forgotten about or ignored.
1943
1944To see details of a given bug, visit
1945https://bugs.kde.org/show_bug.cgi?id=XXXXXX
1946where XXXXXX is the bug number as listed below.
1947
1948135264  dcbzl instruction missing
1949142688  == 250799
1950153699  Valgrind should report unaligned reads with movdqa
1951180217  == 212335
1952190429  Valgrind reports lost of errors in ld.so
1953        with x86_64 2.9.90 glibc
1954197266  valgrind appears to choke on the xmms instruction
1955        "roundsd" on x86_64
1956197988  Crash when demangling very large symbol names
1957202315  unhandled syscall: 332 (inotify_init1)
1958203256  Add page-level profiling to Massif
1959205093  dsymutil=yes needs quotes, locking (partial fix)
1960205241  Snow Leopard 10.6 support (partial fix)
1961206600  Leak checker fails to upgrade indirect blocks when their
1962        parent becomes reachable
1963210935  port valgrind.h (not valgrind) to win32 so apps run under
1964        wine can make client requests
1965211410  vex amd64->IR: 0x15 0xFF 0xFF 0x0 0x0 0x89
1966        within Linux ip-stack checksum functions
1967212335  unhandled instruction bytes: 0xF3 0xF 0xBD 0xC0
1968        (lzcnt %eax,%eax)
1969213685  Undefined value propagates past dependency breaking instruction
1970        (partial fix)
1971215914  Valgrind inserts bogus empty environment variable
1972217863  == 197988
1973219538  adjtimex syscall wrapper wrong in readonly adjtime mode
1974222545  shmat fails under valgind on some arm targets
1975222560  ARM NEON support
1976230407  == 202315
1977231076  == 202315
1978232509  Docs build fails with formatting inside <title></title> elements
1979232793  == 202315
1980235642  [PATCH] syswrap-linux.c: support evdev EVIOCG* ioctls
1981236546  vex x86->IR: 0x66 0xF 0x3A 0xA
1982237202  vex amd64->IR: 0xF3 0xF 0xB8 0xC0 0x49 0x3B
1983237371  better support for VALGRIND_MALLOCLIKE_BLOCK
1984237485  symlink (syscall 57) is not supported on Mac OS
1985237723  sysno == 101 exp-ptrcheck: the 'impossible' happened:
1986        unhandled syscall
1987238208  is_just_below_ESP doesn't take into account red-zone
1988238345  valgrind passes wrong $0 when executing a shell script
1989238679  mq_timedreceive syscall doesn't flag the reception buffer
1990        as "defined"
1991238696  fcntl command F_DUPFD_CLOEXEC not supported
1992238713  unhandled instruction bytes: 0x66 0xF 0x29 0xC6
1993238713  unhandled instruction bytes: 0x66 0xF 0x29 0xC6
1994238745  3.5.0 Make fails on PPC Altivec opcodes, though configure
1995        says "Altivec off"
1996239992  vex amd64->IR: 0x48 0xF 0xC4 0xC1 0x0 0x48
1997240488  == 197988
1998240639  == 212335
1999241377  == 236546
2000241903  == 202315
2001241920  == 212335
2002242606  unhandled syscall: setegid (in Ptrcheck)
2003242814  Helgrind "Impossible has happened" during
2004        QApplication::initInstance();
2005243064  Valgrind attempting to read debug information from iso
2006243270  Make stack unwinding in Valgrind wrappers more reliable
2007243884  exp-ptrcheck: the 'impossible happened: unhandled syscall
2008        sysno = 277 (mq_open)
2009244009  exp-ptrcheck unknown syscalls in analyzing lighttpd
2010244493  ARM VFP d16-d31 registers support
2011244670  add support for audit_session_self syscall on Mac OS 10.6
2012244921  The xml report of helgrind tool is not well format
2013244923  In the xml report file, the <preamble> not escape the
2014        xml char, eg '<','&','>'
2015245535  print full path names in plain text reports
2016245925  x86-64 red zone handling problem
2017246258  Valgrind not catching integer underruns + new [] s
2018246311  reg/reg cmpxchg doesn't work on amd64
2019246549  unhandled syscall unix:277 while testing 32-bit Darwin app
2020246888  Improve Makefile.vex.am
2021247510  [OS X 10.6] Memcheck reports unaddressable bytes passed
2022        to [f]chmod_extended
2023247526  IBM POWER6 (ISA 2.05) support is incomplete
2024247561  Some leak testcases fails due to reachable addresses in
2025        caller save regs
2026247875  sizeofIRType to handle Ity_I128
2027247894  [PATCH] unhandled syscall sys_readahead
2028247980  Doesn't honor CFLAGS passed to configure
2029248373  darwin10.supp is empty in the trunk
2030248822  Linux FIBMAP ioctl has int parameter instead of long
2031248893  [PATCH] make readdwarf.c big endianess safe to enable
2032        unwinding on big endian systems
2033249224  Syscall 336 not supported (SYS_proc_info)
2034249359  == 245535
2035249775  Incorrect scheme for detecting NEON capabilities of host CPU
2036249943  jni JVM init fails when using valgrind
2037249991  Valgrind incorrectly declares AESKEYGENASSIST support
2038        since VEX r2011
2039249996  linux/arm: unhandled syscall: 181 (__NR_pwrite64)
2040250799  frexp$fenv_access_off function generates SIGILL
2041250998  vex x86->IR: unhandled instruction bytes: 0x66 0x66 0x66 0x2E
2042251251  support pclmulqdq insn
2043251362  valgrind: ARM: attach to debugger either fails or provokes
2044        kernel oops
2045251674  Unhandled syscall 294
2046251818  == 254550
2047
2048254257  Add support for debugfiles found by build-id
2049254550  [PATCH] Implement DW_ATE_UTF (DWARF4)
2050254646  Wrapped functions cause stack misalignment on OS X
2051        (and possibly Linux)
2052254556  ARM: valgrinding anything fails with SIGSEGV for 0xFFFF0FA0
2053
2054(3.6.0: 21 October 2010, vex r2068, valgrind r11471).
2055
2056
2057
2058Release 3.5.0 (19 August 2009)
2059~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
20603.5.0 is a feature release with many significant improvements and the
2061usual collection of bug fixes.  The main improvement is that Valgrind
2062now works on Mac OS X.
2063
2064This release supports X86/Linux, AMD64/Linux, PPC32/Linux, PPC64/Linux
2065and X86/Darwin.  Support for recent distros and toolchain components
2066(glibc 2.10, gcc 4.5) has been added.
2067
2068                    -------------------------
2069
2070Here is a short summary of the changes.  Details are shown further
2071down:
2072
2073* Support for Mac OS X (10.5.x).
2074
2075* Improvements and simplifications to Memcheck's leak checker.
2076
2077* Clarification and simplifications in various aspects of Valgrind's
2078  text output.
2079
2080* XML output for Helgrind and Ptrcheck.
2081
2082* Performance and stability improvements for Helgrind and DRD.
2083
2084* Genuinely atomic support for x86/amd64/ppc atomic instructions.
2085
2086* A new experimental tool, BBV, useful for computer architecture
2087  research.
2088
2089* Improved Wine support, including ability to read Windows PDB
2090  debuginfo.
2091
2092                    -------------------------
2093
2094Here are details of the above changes, followed by descriptions of
2095many other minor changes, and a list of fixed bugs.
2096
2097
2098* Valgrind now runs on Mac OS X.  (Note that Mac OS X is sometimes
2099  called "Darwin" because that is the name of the OS core, which is the
2100  level that Valgrind works at.)
2101
2102  Supported systems:
2103
2104  - It requires OS 10.5.x (Leopard).  Porting to 10.4.x is not planned
2105    because it would require work and 10.4 is only becoming less common.
2106
2107  - 32-bit programs on x86 and AMD64 (a.k.a x86-64) machines are supported
2108    fairly well.  For 10.5.x, 32-bit programs are the default even on
2109    64-bit machines, so it handles most current programs.
2110
2111  - 64-bit programs on x86 and AMD64 (a.k.a x86-64) machines are not
2112    officially supported, but simple programs at least will probably work.
2113    However, start-up is slow.
2114
2115  - PowerPC machines are not supported.
2116
2117  Things that don't work:
2118
2119  - The Ptrcheck tool.
2120
2121  - Objective-C garbage collection.
2122
2123  - --db-attach=yes.
2124
2125  - If you have Rogue Amoeba's "Instant Hijack" program installed,
2126    Valgrind will fail with a SIGTRAP at start-up.  See
2127    https://bugs.kde.org/show_bug.cgi?id=193917 for details and a
2128    simple work-around.
2129
2130  Usage notes:
2131
2132  - You will likely find --dsymutil=yes a useful option, as error
2133    messages may be imprecise without it.
2134
2135  - Mac OS X support is new and therefore will be less robust than the
2136    Linux support.  Please report any bugs you find.
2137
2138  - Threaded programs may run more slowly than on Linux.
2139
2140  Many thanks to Greg Parker for developing this port over several years.
2141
2142
2143* Memcheck's leak checker has been improved.
2144
2145  - The results for --leak-check=summary now match the summary results
2146    for --leak-check=full.  Previously they could differ because
2147    --leak-check=summary counted "indirectly lost" blocks and
2148    "suppressed" blocks as "definitely lost".
2149
2150  - Blocks that are only reachable via at least one interior-pointer,
2151    but are directly pointed to by a start-pointer, were previously
2152    marked as "still reachable".  They are now correctly marked as
2153    "possibly lost".
2154
2155  - The default value for the --leak-resolution option has been
2156    changed from "low" to "high".  In general, this means that more
2157    leak reports will be produced, but each leak report will describe
2158    fewer leaked blocks.
2159
2160  - With --leak-check=full, "definitely lost" and "possibly lost"
2161    leaks are now considered as proper errors, ie. they are counted
2162    for the "ERROR SUMMARY" and affect the behaviour of
2163    --error-exitcode.  These leaks are not counted as errors if
2164    --leak-check=summary is specified, however.
2165
2166  - Documentation for the leak checker has been improved.
2167
2168
2169* Various aspects of Valgrind's text output have changed.
2170
2171  - Valgrind's start-up message has changed.  It is shorter but also
2172    includes the command being run, which makes it easier to use
2173    --trace-children=yes.  An example:
2174
2175  - Valgrind's shut-down messages have also changed.  This is most
2176    noticeable with Memcheck, where the leak summary now occurs before
2177    the error summary.  This change was necessary to allow leaks to be
2178    counted as proper errors (see the description of the leak checker
2179    changes above for more details).  This was also necessary to fix a
2180    longstanding bug in which uses of suppressions against leaks were
2181    not "counted", leading to difficulties in maintaining suppression
2182    files (see https://bugs.kde.org/show_bug.cgi?id=186790).
2183
2184  - Behavior of -v has changed.  In previous versions, -v printed out
2185    a mixture of marginally-user-useful information, and tool/core
2186    statistics.  The statistics printing has now been moved to its own
2187    flag, --stats=yes.  This means -v is less verbose and more likely
2188    to convey useful end-user information.
2189
2190  - The format of some (non-XML) stack trace entries has changed a
2191    little.  Previously there were six possible forms:
2192
2193      0x80483BF: really (a.c:20)
2194      0x80483BF: really (in /foo/a.out)
2195      0x80483BF: really
2196      0x80483BF: (within /foo/a.out)
2197      0x80483BF: ??? (a.c:20)
2198      0x80483BF: ???
2199
2200    The third and fourth of these forms have been made more consistent
2201    with the others.  The six possible forms are now:
2202
2203      0x80483BF: really (a.c:20)
2204      0x80483BF: really (in /foo/a.out)
2205      0x80483BF: really (in ???)
2206      0x80483BF: ??? (in /foo/a.out)
2207      0x80483BF: ??? (a.c:20)
2208      0x80483BF: ???
2209
2210    Stack traces produced when --xml=yes is specified are different
2211    and unchanged.
2212
2213
2214* Helgrind and Ptrcheck now support XML output, so they can be used
2215  from GUI tools.  Also, the XML output mechanism has been
2216  overhauled.
2217
2218  - The XML format has been overhauled and generalised, so it is more
2219    suitable for error reporting tools in general.  The Memcheck
2220    specific aspects of it have been removed.  The new format, which
2221    is an evolution of the old format, is described in
2222    docs/internals/xml-output-protocol4.txt.
2223
2224  - Memcheck has been updated to use the new format.
2225
2226  - Helgrind and Ptrcheck are now able to emit output in this format.
2227
2228  - The XML output mechanism has been overhauled.  XML is now output
2229    to its own file descriptor, which means that:
2230
2231    * Valgrind can output text and XML independently.
2232
2233    * The longstanding problem of XML output being corrupted by
2234      unexpected un-tagged text messages  is solved.
2235
2236    As before, the destination for text output is specified using
2237    --log-file=, --log-fd= or --log-socket=.
2238
2239    As before, XML output for a tool is enabled using --xml=yes.
2240
2241    Because there's a new XML output channel, the XML output
2242    destination is now specified by --xml-file=, --xml-fd= or
2243    --xml-socket=.
2244
2245    Initial feedback has shown this causes some confusion.  To
2246    clarify, the two envisaged usage scenarios are:
2247
2248    (1) Normal text output.  In this case, do not specify --xml=yes
2249        nor any of --xml-file=, --xml-fd= or --xml-socket=.
2250
2251    (2) XML output.  In this case, specify --xml=yes, and one of
2252        --xml-file=, --xml-fd= or --xml-socket= to select the XML
2253        destination, one of --log-file=, --log-fd= or --log-socket=
2254        to select the destination for any remaining text messages,
2255        and, importantly, -q.
2256
2257        -q makes Valgrind completely silent on the text channel,
2258        except in the case of critical failures, such as Valgrind
2259        itself segfaulting, or failing to read debugging information.
2260        Hence, in this scenario, it suffices to check whether or not
2261        any output appeared on the text channel.  If yes, then it is
2262        likely to be a critical error which should be brought to the
2263        attention of the user.  If no (the text channel produced no
2264        output) then it can be assumed that the run was successful.
2265
2266        This allows GUIs to make the critical distinction they need to
2267        make (did the run fail or not?) without having to search or
2268        filter the text output channel in any way.
2269
2270    It is also recommended to use --child-silent-after-fork=yes in
2271    scenario (2).
2272
2273
2274* Improvements and changes in Helgrind:
2275
2276  - XML output, as described above
2277
2278  - Checks for consistent association between pthread condition
2279    variables and their associated mutexes are now performed.
2280
2281  - pthread_spinlock functions are supported.
2282
2283  - Modest performance improvements.
2284
2285  - Initial (skeletal) support for describing the behaviour of
2286    non-POSIX synchronisation objects through ThreadSanitizer
2287    compatible ANNOTATE_* macros.
2288
2289  - More controllable tradeoffs between performance and the level of
2290    detail of "previous" accesses in a race.  There are now three
2291    settings:
2292
2293    * --history-level=full.  This is the default, and was also the
2294      default in 3.4.x.  It shows both stacks involved in a race, but
2295      requires a lot of memory and can be very slow in programs that
2296      do many inter-thread synchronisation events.
2297
2298    * --history-level=none.  This only shows the later stack involved
2299      in a race.  This can be much faster than --history-level=full,
2300      but makes it much more difficult to find the other access
2301      involved in the race.
2302
2303    The new intermediate setting is
2304
2305    * --history-level=approx
2306
2307      For the earlier (other) access, two stacks are presented.  The
2308      earlier access is guaranteed to be somewhere in between the two
2309      program points denoted by those stacks.  This is not as useful
2310      as showing the exact stack for the previous access (as per
2311      --history-level=full), but it is better than nothing, and it's
2312      almost as fast as --history-level=none.
2313
2314
2315* New features and improvements in DRD:
2316
2317  - The error messages printed by DRD are now easier to interpret.
2318    Instead of using two different numbers to identify each thread
2319    (Valgrind thread ID and DRD thread ID), DRD does now identify
2320    threads via a single number (the DRD thread ID).  Furthermore
2321    "first observed at" information is now printed for all error
2322    messages related to synchronization objects.
2323
2324  - Added support for named semaphores (sem_open() and sem_close()).
2325
2326  - Race conditions between pthread_barrier_wait() and
2327    pthread_barrier_destroy() calls are now reported.
2328
2329  - Added support for custom allocators through the macros
2330    VALGRIND_MALLOCLIKE_BLOCK() VALGRIND_FREELIKE_BLOCK() (defined in
2331    in <valgrind/valgrind.h>). An alternative for these two macros is
2332    the new client request VG_USERREQ__DRD_CLEAN_MEMORY (defined in
2333    <valgrind/drd.h>).
2334
2335  - Added support for annotating non-POSIX synchronization objects
2336    through several new ANNOTATE_*() macros.
2337
2338  - OpenMP: added support for the OpenMP runtime (libgomp) included
2339    with gcc versions 4.3.0 and 4.4.0.
2340
2341  - Faster operation.
2342
2343  - Added two new command-line options (--first-race-only and
2344    --segment-merging-interval).
2345
2346
2347* Genuinely atomic support for x86/amd64/ppc atomic instructions
2348
2349  Valgrind will now preserve (memory-access) atomicity of LOCK-
2350  prefixed x86/amd64 instructions, and any others implying a global
2351  bus lock.  Ditto for PowerPC l{w,d}arx/st{w,d}cx. instructions.
2352
2353  This means that Valgrinded processes will "play nicely" in
2354  situations where communication with other processes, or the kernel,
2355  is done through shared memory and coordinated with such atomic
2356  instructions.  Prior to this change, such arrangements usually
2357  resulted in hangs, races or other synchronisation failures, because
2358  Valgrind did not honour atomicity of such instructions.
2359
2360
2361* A new experimental tool, BBV, has been added.  BBV generates basic
2362  block vectors for use with the SimPoint analysis tool, which allows
2363  a program's overall behaviour to be approximated by running only a
2364  fraction of it.  This is useful for computer architecture
2365  researchers.  You can run BBV by specifying --tool=exp-bbv (the
2366  "exp-" prefix is short for "experimental").  BBV was written by
2367  Vince Weaver.
2368
2369
2370* Modestly improved support for running Windows applications under
2371  Wine.  In particular, initial support for reading Windows .PDB debug
2372  information has been added.
2373
2374
2375* A new Memcheck client request VALGRIND_COUNT_LEAK_BLOCKS has been
2376  added.  It is similar to VALGRIND_COUNT_LEAKS but counts blocks
2377  instead of bytes.
2378
2379
2380* The Valgrind client requests VALGRIND_PRINTF and
2381  VALGRIND_PRINTF_BACKTRACE have been changed slightly.  Previously,
2382  the string was always printed immediately on its own line.  Now, the
2383  string will be added to a buffer but not printed until a newline is
2384  encountered, or other Valgrind output is printed (note that for
2385  VALGRIND_PRINTF_BACKTRACE, the back-trace itself is considered
2386  "other Valgrind output").  This allows you to use multiple
2387  VALGRIND_PRINTF calls to build up a single output line, and also to
2388  print multiple output lines with a single request (by embedding
2389  multiple newlines in the string).
2390
2391
2392* The graphs drawn by Massif's ms_print program have changed slightly:
2393
2394  - The half-height chars '.' and ',' are no longer drawn, because
2395    they are confusing.  The --y option can be used if the default
2396    y-resolution is not high enough.
2397
2398  - Horizontal lines are now drawn after the top of a snapshot if
2399    there is a gap until the next snapshot.  This makes it clear that
2400    the memory usage has not dropped to zero between snapshots.
2401
2402
2403* Something that happened in 3.4.0, but wasn't clearly announced: the
2404  option --read-var-info=yes can be used by some tools (Memcheck,
2405  Helgrind and DRD).  When enabled, it causes Valgrind to read DWARF3
2406  variable type and location information.  This makes those tools
2407  start up more slowly and increases memory consumption, but
2408  descriptions of data addresses in error messages become more
2409  detailed.
2410
2411
2412* exp-Omega, an experimental instantaneous leak-detecting tool, was
2413  disabled in 3.4.0 due to a lack of interest and maintenance,
2414  although the source code was still in the distribution.  The source
2415  code has now been removed from the distribution.  For anyone
2416  interested, the removal occurred in SVN revision r10247.
2417
2418
2419* Some changes have been made to the build system.
2420
2421  - VEX/ is now integrated properly into the build system.  This means
2422    that dependency tracking within VEX/ now works properly, "make
2423    install" will work without requiring "make" before it, and
2424    parallel builds (ie. 'make -j') now work (previously a
2425    .NOTPARALLEL directive was used to serialize builds, ie. 'make -j'
2426    was effectively ignored).
2427
2428  - The --with-vex configure option has been removed.  It was of
2429    little use and removing it simplified the build system.
2430
2431  - The location of some install files has changed.  This should not
2432    affect most users.  Those who might be affected:
2433
2434    * For people who use Valgrind with MPI programs, the installed
2435      libmpiwrap.so library has moved from
2436      $(INSTALL)/<platform>/libmpiwrap.so to
2437      $(INSTALL)/libmpiwrap-<platform>.so.
2438
2439    * For people who distribute standalone Valgrind tools, the
2440      installed libraries such as $(INSTALL)/<platform>/libcoregrind.a
2441      have moved to $(INSTALL)/libcoregrind-<platform>.a.
2442
2443    These changes simplify the build system.
2444
2445  - Previously, all the distributed suppression (*.supp) files were
2446    installed.  Now, only default.supp is installed.  This should not
2447    affect users as the other installed suppression files were not
2448    read; the fact that they were installed was a mistake.
2449
2450
2451* KNOWN LIMITATIONS:
2452
2453  - Memcheck is unusable with the Intel compiler suite version 11.1,
2454    when it generates code for SSE2-and-above capable targets.  This
2455    is because of icc's use of highly optimised inlined strlen
2456    implementations.  It causes Memcheck to report huge numbers of
2457    false errors even in simple programs.  Helgrind and DRD may also
2458    have problems.
2459
2460    Versions 11.0 and earlier may be OK, but this has not been
2461    properly tested.
2462
2463
2464The following bugs have been fixed or resolved.  Note that "n-i-bz"
2465stands for "not in bugzilla" -- that is, a bug that was reported to us
2466but never got a bugzilla entry.  We encourage you to file bugs in
2467bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
2468mailing the developers (or mailing lists) directly -- bugs that are
2469not entered into bugzilla tend to get forgotten about or ignored.
2470
2471To see details of a given bug, visit
2472https://bugs.kde.org/show_bug.cgi?id=XXXXXX
2473where XXXXXX is the bug number as listed below.
2474
247584303   How about a LockCheck tool?
247691633   dereference of null ptr in vgPlain_st_basetype
247797452   Valgrind doesn't report any pthreads problems
2478100628  leak-check gets assertion failure when using
2479        VALGRIND_MALLOCLIKE_BLOCK on malloc()ed memory
2480108528  NPTL pthread cleanup handlers not called
2481110126  Valgrind 2.4.1 configure.in tramples CFLAGS
2482110128  mallinfo is not implemented...
2483110770  VEX: Generated files not always updated when making valgrind
2484111102  Memcheck: problems with large (memory footprint) applications
2485115673  Vex's decoder should never assert
2486117564  False positive: Syscall param clone(child_tidptr) contains
2487        uninitialised byte(s)
2488119404  executing ssh from inside valgrind fails
2489133679  Callgrind does not write path names to sources with dwarf debug
2490        info
2491135847  configure.in problem with non gnu compilers (and possible fix)
2492136154  threads.c:273 (vgCallgrind_post_signal): Assertion
2493        '*(vgCallgrind_current_fn_stack.top) == 0' failed.
2494136230  memcheck reports "possibly lost", should be "still reachable"
2495137073  NULL arg to MALLOCLIKE_BLOCK causes crash
2496137904  Valgrind reports a memory leak when using POSIX threads,
2497        while it shouldn't
2498139076  valgrind VT_GETSTATE error
2499142228  complaint of elf_dynamic_do_rela in trivial usage
2500145347  spurious warning with USBDEVFS_REAPURB
2501148441  (wine) can't find memory leak in Wine, win32 binary
2502        executable file.
2503148742  Leak-check fails assert on exit
2504149878  add (proper) check for calloc integer overflow
2505150606  Call graph is broken when using callgrind control
2506152393  leak errors produce an exit code of 0. I need some way to
2507        cause leak errors to result in a nonzero exit code.
2508157154  documentation (leak-resolution doc speaks about num-callers
2509        def=4) + what is a loss record
2510159501  incorrect handling of ALSA ioctls
2511162020  Valgrinding an empty/zero-byte file crashes valgrind
2512162482  ppc: Valgrind crashes while reading stabs information
2513162718  x86: avoid segment selector 0 in sys_set_thread_area()
2514163253  (wine) canonicaliseSymtab forgot some fields in DiSym
2515163560  VEX/test_main.c is missing from valgrind-3.3.1
2516164353  malloc_usable_size() doesn't return a usable size
2517165468  Inconsistent formatting in memcheck manual -- please fix
2518169505  main.c:286 (endOfInstr):
2519        Assertion 'ii->cost_offset == *cost_offset' failed
2520177206  Generate default.supp during compile instead of configure
2521177209  Configure valt_load_address based on arch+os
2522177305  eventfd / syscall 323 patch lost
2523179731  Tests fail to build because of inlining of non-local asm labels
2524181394  helgrind: libhb_core.c:3762 (msm_write): Assertion
2525        'ordxx == POrd_EQ || ordxx == POrd_LT' failed.
2526181594  Bogus warning for empty text segment
2527181707  dwarf doesn't require enumerations to have name
2528185038  exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64
2529185050  exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree):
2530        Assertion '!already_present' failed.
2531185359  exp-ptrcheck: unhandled syscall getresuid()
2532185794  "WARNING: unhandled syscall: 285" (fallocate) on x86_64
2533185816  Valgrind is unable to handle debug info for files with split
2534        debug info that are prelinked afterwards
2535185980  [darwin] unhandled syscall: sem_open
2536186238  bbToIR_AMD64: disInstr miscalculated next %rip
2537186507  exp-ptrcheck unhandled syscalls prctl, etc.
2538186790  Suppression pattern used for leaks are not reported
2539186796  Symbols with length>200 in suppression files are ignored
2540187048  drd: mutex PTHREAD_PROCESS_SHARED attribute missinterpretation
2541187416  exp-ptrcheck: support for __NR_{setregid,setreuid,setresuid}
2542188038  helgrind: hg_main.c:926: mk_SHVAL_fail: the 'impossible' happened
2543188046  bashisms in the configure script
2544188127  amd64->IR: unhandled instruction bytes: 0xF0 0xF 0xB0 0xA
2545188161  memcheck: --track-origins=yes asserts "mc_machine.c:672
2546        (get_otrack_shadow_offset_wrk): the 'impossible' happened."
2547188248  helgrind: pthread_cleanup_push, pthread_rwlock_unlock,
2548        assertion fail "!lock->heldBy"
2549188427  Add support for epoll_create1 (with patch)
2550188530  Support for SIOCGSTAMPNS
2551188560  Include valgrind.spec in the tarball
2552188572  Valgrind on Mac should suppress setenv() mem leak
2553189054  Valgrind fails to build because of duplicate non-local asm labels
2554189737  vex amd64->IR: unhandled instruction bytes: 0xAC
2555189762  epoll_create syscall not handled (--tool=exp-ptrcheck)
2556189763  drd assertion failure: s_threadinfo[tid].is_recording
2557190219  unhandled syscall: 328 (x86-linux)
2558190391  dup of 181394; see above
2559190429  Valgrind reports lots of errors in ld.so with x86_64 2.9.90 glibc
2560190820  No debug information on powerpc-linux
2561191095  PATCH: Improve usbdevfs ioctl handling
2562191182  memcheck: VALGRIND_LEAK_CHECK quadratic when big nr of chunks
2563        or big nr of errors
2564191189  --xml=yes should obey --gen-suppressions=all
2565191192  syslog() needs a suppression on macosx
2566191271  DARWIN: WARNING: unhandled syscall: 33554697 a.k.a.: 265
2567191761  getrlimit on MacOSX
2568191992  multiple --fn-skip only works sometimes; dependent on order
2569192634  V. reports "aspacem sync_check_mapping_callback:
2570        segment mismatch" on Darwin
2571192954  __extension__ missing on 2 client requests
2572194429  Crash at start-up with glibc-2.10.1 and linux-2.6.29
2573194474  "INSTALL" file has different build instructions than "README"
2574194671  Unhandled syscall (sem_wait?) from mac valgrind
2575195069  memcheck: reports leak (memory still reachable) for
2576        printf("%d', x)
2577195169  drd: (vgDrd_barrier_post_wait):
2578        Assertion 'r->sg[p->post_iteration]' failed.
2579195268  valgrind --log-file doesn't accept ~/...
2580195838  VEX abort: LibVEX_N_SPILL_BYTES too small for CPUID boilerplate
2581195860  WARNING: unhandled syscall: unix:223
2582196528  need a error suppression for pthread_rwlock_init under os x?
2583197227  Support aio_* syscalls on Darwin
2584197456  valgrind should reject --suppressions=(directory)
2585197512  DWARF2 CFI reader: unhandled CFI instruction 0:10
2586197591  unhandled syscall 27 (mincore)
2587197793  Merge DCAS branch to the trunk == 85756, 142103
2588197794  Avoid duplicate filenames in Vex
2589197898  make check fails on current SVN
2590197901  make check fails also under exp-ptrcheck in current SVN
2591197929  Make --leak-resolution=high the default
2592197930  Reduce spacing between leak reports
2593197933  Print command line of client at start-up, and shorten preamble
2594197966  unhandled syscall 205 (x86-linux, --tool=exp-ptrcheck)
2595198395  add BBV to the distribution as an experimental tool
2596198624  Missing syscalls on Darwin: 82, 167, 281, 347
2597198649  callgrind_annotate doesn't cumulate counters
2598199338  callgrind_annotate sorting/thresholds are broken for all but Ir
2599199977  Valgrind complains about an unrecognized instruction in the
2600        atomic_incs test program
2601200029  valgrind isn't able to read Fedora 12 debuginfo
2602200760  darwin unhandled syscall: unix:284
2603200827  DRD doesn't work on Mac OS X
2604200990  VG_(read_millisecond_timer)() does not work correctly
2605201016  Valgrind does not support pthread_kill() on Mac OS
2606201169  Document --read-var-info
2607201323  Pre-3.5.0 performance sanity checking
2608201384  Review user manual for the 3.5.0 release
2609201585  mfpvr not implemented on ppc
2610201708  tests failing because x86 direction flag is left set
2611201757  Valgrind doesn't handle any recent sys_futex additions
2612204377  64-bit valgrind can not start a shell script
2613        (with #!/path/to/shell) if the shell is a 32-bit executable
2614n-i-bz  drd: fixed assertion failure triggered by mutex reinitialization.
2615n-i-bz  drd: fixed a bug that caused incorrect messages to be printed
2616        about memory allocation events with memory access tracing enabled
2617n-i-bz  drd: fixed a memory leak triggered by vector clock deallocation
2618
2619(3.5.0: 19 Aug 2009, vex r1913, valgrind r10846).
2620
2621
2622
2623Release 3.4.1 (28 February 2009)
2624~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
26253.4.1 is a bug-fix release that fixes some regressions and assertion
2626failures in debug info reading in 3.4.0, most notably incorrect stack
2627traces on amd64-linux on older (glibc-2.3 based) systems. Various
2628other debug info problems are also fixed.  A number of bugs in the
2629exp-ptrcheck tool introduced in 3.4.0 have been fixed.
2630
2631In view of the fact that 3.4.0 contains user-visible regressions
2632relative to 3.3.x, upgrading to 3.4.1 is recommended.  Packagers are
2633encouraged to ship 3.4.1 in preference to 3.4.0.
2634
2635The fixed bugs are as follows.  Note that "n-i-bz" stands for "not in
2636bugzilla" -- that is, a bug that was reported to us but never got a
2637bugzilla entry.  We encourage you to file bugs in bugzilla
2638(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the
2639developers (or mailing lists) directly -- bugs that are not entered
2640into bugzilla tend to get forgotten about or ignored.
2641
2642n-i-bz  Fix various bugs reading icc-11 generated debug info
2643n-i-bz  Fix various bugs reading gcc-4.4 generated debug info
2644n-i-bz  Preliminary support for glibc-2.10 / Fedora 11
2645n-i-bz  Cachegrind and Callgrind: handle non-power-of-two cache sizes,
2646        so as to support (eg) 24k Atom D1 and Core2 with 3/6/12MB L2.
2647179618  exp-ptrcheck crashed / exit prematurely
2648179624  helgrind: false positive races with pthread_create and
2649        recv/open/close/read
2650134207  pkg-config output contains @VG_PLATFORM@
2651176926  floating point exception at valgrind startup with PPC 440EPX
2652181594  Bogus warning for empty text segment
2653173751  amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant rex prefixes)
2654181707  Dwarf3 doesn't require enumerations to have name
2655185038  exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64
2656185050  exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree):
2657        Assertion '!already_present' failed.
2658185359  exp-ptrcheck unhandled syscall getresuid()
2659
2660(3.4.1.RC1:  24 Feb 2008, vex r1884, valgrind r9253).
2661(3.4.1:      28 Feb 2008, vex r1884, valgrind r9293).
2662
2663
2664
2665Release 3.4.0 (2 January 2009)
2666~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
26673.4.0 is a feature release with many significant improvements and the
2668usual collection of bug fixes.  This release supports X86/Linux,
2669AMD64/Linux, PPC32/Linux and PPC64/Linux.  Support for recent distros
2670(using gcc 4.4, glibc 2.8 and 2.9) has been added.
2671
26723.4.0 brings some significant tool improvements.  Memcheck can now
2673report the origin of uninitialised values, the thread checkers
2674Helgrind and DRD are much improved, and we have a new experimental
2675tool, exp-Ptrcheck, which is able to detect overruns of stack and
2676global arrays.  In detail:
2677
2678* Memcheck is now able to track the origin of uninitialised values.
2679  When it reports an uninitialised value error, it will try to show
2680  the origin of the value, as either a heap or stack allocation.
2681  Origin tracking is expensive and so is not enabled by default.  To
2682  use it, specify --track-origins=yes.  Memcheck's speed will be
2683  essentially halved, and memory usage will be significantly
2684  increased.  Nevertheless it can drastically reduce the effort
2685  required to identify the root cause of uninitialised value errors,
2686  and so is often a programmer productivity win, despite running more
2687  slowly.
2688
2689* A version (1.4.0) of the Valkyrie GUI, that works with Memcheck in
2690  3.4.0, will be released shortly.
2691
2692* Helgrind's race detection algorithm has been completely redesigned
2693  and reimplemented, to address usability and scalability concerns:
2694
2695  - The new algorithm has a lower false-error rate: it is much less
2696    likely to report races that do not really exist.
2697
2698  - Helgrind will display full call stacks for both accesses involved
2699    in a race.  This makes it easier to identify the root causes of
2700    races.
2701
2702  - Limitations on the size of program that can run have been removed.
2703
2704  - Performance has been modestly improved, although that is very
2705    workload-dependent.
2706
2707  - Direct support for Qt4 threading has been added.
2708
2709  - pthread_barriers are now directly supported.
2710
2711  - Helgrind works well on all supported Linux targets.
2712
2713* The DRD thread debugging tool has seen major improvements:
2714
2715  - Greatly improved performance and significantly reduced memory
2716    usage.
2717
2718  - Support for several major threading libraries (Boost.Thread, Qt4,
2719    glib, OpenMP) has been added.
2720
2721  - Support for atomic instructions, POSIX semaphores, barriers and
2722    reader-writer locks has been added.
2723
2724  - Works now on PowerPC CPUs too.
2725
2726  - Added support for printing thread stack usage at thread exit time.
2727
2728  - Added support for debugging lock contention.
2729
2730  - Added a manual for Drd.
2731
2732* A new experimental tool, exp-Ptrcheck, has been added.  Ptrcheck
2733  checks for misuses of pointers.  In that sense it is a bit like
2734  Memcheck.  However, Ptrcheck can do things Memcheck can't: it can
2735  detect overruns of stack and global arrays, it can detect
2736  arbitrarily far out-of-bounds accesses to heap blocks, and it can
2737  detect accesses heap blocks that have been freed a very long time
2738  ago (millions of blocks in the past).
2739
2740  Ptrcheck currently works only on x86-linux and amd64-linux.  To use
2741  it, use --tool=exp-ptrcheck.  A simple manual is provided, as part
2742  of the main Valgrind documentation.  As this is an experimental
2743  tool, we would be particularly interested in hearing about your
2744  experiences with it.
2745
2746* exp-Omega, an experimental instantaneous leak-detecting tool, is no
2747  longer built by default, although the code remains in the repository
2748  and the tarball.  This is due to three factors: a perceived lack of
2749  users, a lack of maintenance, and concerns that it may not be
2750  possible to achieve reliable operation using the existing design.
2751
2752* As usual, support for the latest Linux distros and toolchain
2753  components has been added.  It should work well on Fedora Core 10,
2754  OpenSUSE 11.1 and Ubuntu 8.10.  gcc-4.4 (in its current pre-release
2755  state) is supported, as is glibc-2.9.  The C++ demangler has been
2756  updated so as to work well with C++ compiled by even the most recent
2757  g++'s.
2758
2759* You can now use frame-level wildcards in suppressions.  This was a
2760  frequently-requested enhancement.  A line "..." in a suppression now
2761  matches zero or more frames.  This makes it easier to write
2762  suppressions which are precise yet insensitive to changes in
2763  inlining behaviour.
2764
2765* 3.4.0 adds support on x86/amd64 for the SSSE3 instruction set.
2766
2767* Very basic support for IBM Power6 has been added (64-bit processes only).
2768
2769* Valgrind is now cross-compilable.  For example, it is possible to
2770  cross compile Valgrind on an x86/amd64-linux host, so that it runs
2771  on a ppc32/64-linux target.
2772
2773* You can set the main thread's stack size at startup using the
2774  new --main-stacksize= flag (subject of course to ulimit settings).
2775  This is useful for running apps that need a lot of stack space.
2776
2777* The limitation that you can't use --trace-children=yes together
2778  with --db-attach=yes has been removed.
2779
2780* The following bugs have been fixed.  Note that "n-i-bz" stands for
2781  "not in bugzilla" -- that is, a bug that was reported to us but
2782  never got a bugzilla entry.  We encourage you to file bugs in
2783  bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
2784  mailing the developers (or mailing lists) directly.
2785
2786  n-i-bz  Make return types for some client requests 64-bit clean
2787  n-i-bz  glibc 2.9 support
2788  n-i-bz  ignore unsafe .valgrindrc's (CVE-2008-4865)
2789  n-i-bz  MPI_Init(0,0) is valid but libmpiwrap.c segfaults
2790  n-i-bz  Building in an env without gdb gives bogus gdb attach
2791  92456   Tracing the origin of uninitialised memory
2792  106497  Valgrind does not demangle some C++ template symbols
2793  162222  ==106497
2794  151612  Suppression with "..." (frame-level wildcards in .supp files)
2795  156404  Unable to start oocalc under memcheck on openSUSE 10.3 (64-bit)
2796  159285  unhandled syscall:25 (stime, on x86-linux)
2797  159452  unhandled ioctl 0x8B01 on "valgrind iwconfig"
2798  160954  ppc build of valgrind crashes with illegal instruction (isel)
2799  160956  mallinfo implementation, w/ patch
2800  162092  Valgrind fails to start gnome-system-monitor
2801  162819  malloc_free_fill test doesn't pass on glibc2.8 x86
2802  163794  assertion failure with "--track-origins=yes"
2803  163933  sigcontext.err and .trapno must be set together
2804  163955  remove constraint !(--db-attach=yes && --trace-children=yes)
2805  164476  Missing kernel module loading system calls
2806  164669  SVN regression: mmap() drops posix file locks
2807  166581  Callgrind output corruption when program forks
2808  167288  Patch file for missing system calls on Cell BE
2809  168943  unsupported scas instruction pentium
2810  171645  Unrecognised instruction (MOVSD, non-binutils encoding)
2811  172417  x86->IR: 0x82 ...
2812  172563  amd64->IR: 0xD9 0xF5  -  fprem1
2813  173099  .lds linker script generation error
2814  173177  [x86_64] syscalls: 125/126/179 (capget/capset/quotactl)
2815  173751  amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant prefixes)
2816  174532  == 173751
2817  174908  --log-file value not expanded correctly for core file
2818  175044  Add lookup_dcookie for amd64
2819  175150  x86->IR: 0xF2 0xF 0x11 0xC1 (movss non-binutils encoding)
2820
2821Developer-visible changes:
2822
2823* Valgrind's debug-info reading machinery has been majorly overhauled.
2824  It can now correctly establish the addresses for ELF data symbols,
2825  which is something that has never worked properly before now.
2826
2827  Also, Valgrind can now read DWARF3 type and location information for
2828  stack and global variables.  This makes it possible to use the
2829  framework to build tools that rely on knowing the type and locations
2830  of stack and global variables, for example exp-Ptrcheck.
2831
2832  Reading of such information is disabled by default, because most
2833  tools don't need it, and because it is expensive in space and time.
2834  However, you can force Valgrind to read it, using the
2835  --read-var-info=yes flag.  Memcheck, Helgrind and DRD are able to
2836  make use of such information, if present, to provide source-level
2837  descriptions of data addresses in the error messages they create.
2838
2839(3.4.0.RC1:  24 Dec 2008, vex r1878, valgrind r8882).
2840(3.4.0:       3 Jan 2009, vex r1878, valgrind r8899).
2841