1 Release 3.12.0 (?? ????????? 201?) 2 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 3 Release 3.12.0 is under development, not yet released. 4 5 * ================== PLATFORM CHANGES ================= 6 7 * ==================== TOOL CHANGES ==================== 8 9 * Memcheck: 10 11 * Helgrind: 12 13 * Callgrind: 14 15 * DRD: 16 n-i-bz Improved thread startup time significantly on non-Linux platforms. 17 18 * ==================== OTHER CHANGES ==================== 19 20 * ==================== FIXED BUGS ==================== 21 22 The following bugs have been fixed or resolved. Note that "n-i-bz" 23 stands for "not in bugzilla" -- that is, a bug that was reported to us 24 but never got a bugzilla entry. We encourage you to file bugs in 25 bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather 26 than mailing the developers (or mailing lists) directly -- bugs that 27 are not entered into bugzilla tend to get forgotten about or ignored. 28 29 To see details of a given bug, visit 30 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 31 where XXXXXX is the bug number as listed below. 32 33 278744 cvtps2pd with redundant RexW 34 353083 arm64 doesn't implement various xattr system calls 35 353084 arm64 doesn't support sigpending system call 36 353370 don't advertise RDRAND in cpuid for Core-i7-4910-like avx2 machine 37 353398 WARNING: unhandled amd64-solaris syscall: 207 38 353680 s390x: Crash with certain glibc versions due to non-implemented TBEGIN 39 353891 Assert 'bad_scanned_addr < VG_ROUNDDN(start+len, sizeof(Addr))' failed 40 353917 unhandled amd64-solaris syscall fchdir(120) 41 353920 unhandled amd64-solaris syscall: 170 42 43 44 45 Release 3.11.0 (22 September 2015) 46 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 47 48 3.11.0 is a feature release with many improvements and the usual 49 collection of bug fixes. 50 51 This release supports X86/Linux, AMD64/Linux, ARM32/Linux, 52 ARM64/Linux, PPC32/Linux, PPC64BE/Linux, PPC64LE/Linux, S390X/Linux, 53 MIPS32/Linux, MIPS64/Linux, ARM/Android, ARM64/Android, 54 MIPS32/Android, X86/Android, X86/Solaris, AMD64/Solaris, X86/MacOSX 55 10.10 and AMD64/MacOSX 10.10. There is also preliminary support for 56 X86/MacOSX 10.11, AMD64/MacOSX 10.11 and TILEGX/Linux. 57 58 * ================== PLATFORM CHANGES ================= 59 60 * Support for Solaris/x86 and Solaris/amd64 has been added. 61 62 * Preliminary support for Mac OS X 10.11 (El Capitan) has been added. 63 64 * Preliminary support for the Tilera TileGX architecture has been added. 65 66 * s390x: It is now required for the host to have the "long displacement" 67 facility. The oldest supported machine model is z990. 68 69 * x86: on an SSE2 only host, Valgrind in 32 bit mode now claims to be a 70 Pentium 4. 3.10.1 wrongly claimed to be a Core 2, which is SSSE3. 71 72 * The JIT's register allocator is significantly faster, making the JIT 73 as a whole somewhat faster, so JIT-intensive activities, for example 74 program startup, are modestly faster, around 5%. 75 76 * There have been changes to the default settings of several command 77 line flags, as detailed below. 78 79 * Intel AVX2 support is more complete (64 bit targets only). On AVX2 80 capable hosts, the simulated CPUID will now indicate AVX2 support. 81 82 * ==================== TOOL CHANGES ==================== 83 84 * Memcheck: 85 86 - The default value for --leak-check-heuristics has been changed from 87 "none" to "all". This helps to reduce the number of possibly 88 lost blocks, in particular for C++ applications. 89 90 - The default value for --keep-stacktraces has been changed from 91 "malloc-then-free" to "malloc-and-free". This has a small cost in 92 memory (one word per malloc-ed block) but allows Memcheck to show the 93 3 stacktraces of a dangling reference: where the block was allocated, 94 where it was freed, and where it is acccessed after being freed. 95 96 - The default value for --partial-loads-ok has been changed from "no" to 97 "yes", so as to avoid false positive errors resulting from some kinds 98 of vectorised loops. 99 100 - A new monitor command 'xb <addr> <len>' shows the validity bits of 101 <len> bytes at <addr>. The monitor command 'xb' is easier to use 102 than get_vbits when you need to associate byte data value with 103 their corresponding validity bits. 104 105 - The 'block_list' monitor command has been enhanced: 106 o it can print a range of loss records 107 o it now accepts an optional argument 'limited <max_blocks>' 108 to control the number of blocks printed. 109 o if a block has been found using a heuristic, then 110 'block_list' now shows the heuristic after the block size. 111 o the loss records/blocks to print can be limited to the blocks 112 found via specified heuristics. 113 114 - The C helper functions used to instrument loads on 115 x86-{linux,solaris} and arm-linux (both 32-bit only) have been 116 replaced by handwritten assembly sequences. This gives speedups 117 in the region of 0% to 7% for those targets only. 118 119 - A new command line option, --expensive-definedness-checks=yes|no, 120 has been added. This is useful for avoiding occasional invalid 121 uninitialised-value errors in optimised code. Watch out for 122 runtime degradation, as this can be up to 25%. As always, though, 123 the slowdown is highly application specific. The default setting 124 is "no". 125 126 * Massif: 127 128 - A new monitor command 'all_snapshots <filename>' dumps all 129 snapshots taken so far. 130 131 * Helgrind: 132 133 - Significant memory reduction and moderate speedups for 134 --history-level=full for applications accessing a lot of memory 135 with many different stacktraces. 136 137 - The default value for --conflict-cache-size=N has been doubled to 138 2000000. Users that were not using the default value should 139 preferably also double the value they give. 140 141 The default was changed due to the changes in the "full history" 142 implementation. Doubling the value gives on average a slightly more 143 complete history and uses similar memory (or significantly less memory 144 in the worst case) than the previous implementation. 145 146 - The Helgrind monitor command 'info locks' now accepts an optional 147 argument 'lock_addr', which shows information about the lock at the 148 given address only. 149 150 - When using --history-level=full, the new Helgrind monitor command 151 'accesshistory <addr> [<len>]' will show the recorded accesses for 152 <len> (or 1) bytes at <addr>. 153 154 * ==================== OTHER CHANGES ==================== 155 156 * The default value for the --smc-check option has been changed from 157 "stack" to "all-non-file" on targets that provide automatic D-I 158 cache coherence (x86, amd64 and s390x). The result is to provide, 159 by default, transparent support for JIT generated and self-modifying 160 code on all targets. 161 162 * Mac OS X only: the default value for the --dsymutil option has been 163 changed from "no" to "yes", since any serious usage on Mac OS X 164 always required it to be "yes". 165 166 * The command line options --db-attach and --db-command have been removed. 167 They were deprecated in 3.10.0. 168 169 * When a process dies due to a signal, Valgrind now shows the signal 170 and the stacktrace at default verbosity (i.e. verbosity 1). 171 172 * The address description logic used by Memcheck and Helgrind now 173 describes addresses in anonymous segments, file mmap-ed segments, 174 shared memory segments and the brk data segment. 175 176 * The new option --error-markers=<begin>,<end> can be used to mark the 177 begin/end of errors in textual output mode, to facilitate 178 searching/extracting errors in output files that mix valgrind errors 179 with program output. 180 181 * The new option --max-threads=<number> can be used to change the number 182 of threads valgrind can handle. The default is 500 threads which 183 should be more than enough for most applications. 184 185 * The new option --valgrind-stacksize=<number> can be used to change the 186 size of the private thread stacks used by Valgrind. This is useful 187 for reducing memory use or increasing the stack size if Valgrind 188 segfaults due to stack overflow. 189 190 * The new option --avg-transtab-entry-size=<number> can be used to specify 191 the expected instrumented block size, either to reduce memory use or 192 to avoid excessive retranslation. 193 194 * Valgrind can be built with Intel's ICC compiler, version 14.0 or later. 195 196 * New and modified GDB server monitor features: 197 198 - When a signal is reported in GDB, you can now use the GDB convenience 199 variable $_siginfo to examine detailed signal information. 200 201 - Valgrind's gdbserver now allows the user to change the signal 202 to deliver to the process. So, use 'signal SIGNAL' to continue execution 203 with SIGNAL instead of the signal reported to GDB. Use 'signal 0' to 204 continue without passing the signal to the process. 205 206 - With GDB >= 7.10, the command 'target remote' 207 will automatically load the executable file of the process running 208 under Valgrind. This means you do not need to specify the executable 209 file yourself, GDB will discover it itself. See GDB documentation about 210 'qXfer:exec-file:read' packet for more info. 211 212 * ==================== FIXED BUGS ==================== 213 214 The following bugs have been fixed or resolved. Note that "n-i-bz" 215 stands for "not in bugzilla" -- that is, a bug that was reported to us 216 but never got a bugzilla entry. We encourage you to file bugs in 217 bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather 218 than mailing the developers (or mailing lists) directly -- bugs that 219 are not entered into bugzilla tend to get forgotten about or ignored. 220 221 To see details of a given bug, visit 222 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 223 where XXXXXX is the bug number as listed below. 224 225 116002 VG_(printf): Problems with justification of strings and integers 226 155125 avoid cutting away file:lineno after long function name 227 197259 Unsupported arch_prtctl PR_SET_GS option 228 201152 ppc64: Assertion in ppc32g_dirtyhelper_MFSPR_268_269 229 201216 Fix Valgrind does not support pthread_sigmask() on OS X 230 201435 Fix Darwin: -v does not show kernel version 231 208217 "Warning: noted but unhandled ioctl 0x2000747b" on Mac OS X 232 211256 Fixed an outdated comment regarding the default platform. 233 211529 Incomplete call stacks for code compiled by newer versions of MSVC 234 211926 Avoid compilation warnings in valgrind.h with -pedantic 235 212291 Fix unhandled syscall: unix:132 (mkfifo) on OS X 236 == 263119 237 226609 Crediting upstream authors in man page 238 231257 Valgrind omits path when executing script from shebang line 239 254164 OS X task_info: UNKNOWN task message [id 3405, to mach_task_self() [..] 240 294065 Improve the pdb file reader by avoiding hardwired absolute pathnames 241 269360 s390x: Fix addressing mode selection for compare-and-swap 242 302630 Memcheck: Assertion failed: 'sizeof(UWord) == sizeof(UInt)' 243 == 326797 244 312989 ioctl handling needs to do POST handling on generic ioctls and [..] 245 319274 Fix unhandled syscall: unix:410 (sigsuspend_nocancel) on OS X 246 324181 mmap does not handle MAP_32BIT (handle it now, rather than fail it) 247 327745 Fix valgrind 3.9.0 build fails on Mac OS X 10.6.8 248 330147 libmpiwrap PMPI_Get_count returns undefined value 249 333051 mmap of huge pages fails due to incorrect alignment 250 == 339163 251 334802 valgrind does not always explain why a given option is bad 252 335618 mov.w rN, pc/sp (ARM32) 253 335785 amd64->IR 0xC4 0xE2 0x75 0x2F (vmaskmovpd) 254 == 307399 255 == 343175 256 == 342740 257 == 346912 258 335907 segfault when running wine's ddrawex/tests/surface.c under valgrind 259 338602 AVX2 bit in CPUID missing 260 338606 Strange message for scripts with invalid interpreter 261 338731 ppc: Fix testuite build for toolchains not supporting -maltivec 262 338995 shmat with hugepages (SHM_HUGETLB) fails with EINVAL 263 339045 Getting valgrind to compile and run on OS X Yosemite (10.10) 264 == 340252 265 339156 gdbsrv not called for fatal signal 266 339215 Valgrind 3.10.0 contain 2013 in copyrights notice 267 339288 support Cavium Octeon MIPS specific BBIT*32 instructions 268 339636 Use fxsave64 and fxrstor64 mnemonics instead of old-school rex64 prefix 269 339442 Fix testsuite build failure on OS X 10.9 270 339542 Enable compilation with Intel's ICC compiler 271 339563 The DVB demux DMX_STOP ioctl doesn't have a wrapper 272 339688 Mac-specific ASM does not support .version directive (cpuid, 273 tronical and pushfpopf tests) 274 339745 Valgrind crash when check Marmalade app (partial fix) 275 339755 Fix known deliberate memory leak in setenv() on Mac OS X 10.9 276 339778 Linux/TileGx platform support for Valgrind 277 339780 Fix known uninitialised read in pthread_rwlock_init() on Mac OS X 10.9 278 339789 Fix none/tests/execve test on Mac OS X 10.9 279 339808 Fix none/tests/rlimit64_nofile test on Mac OS X 10.9 280 339820 vex amd64->IR: 0x66 0xF 0x3A 0x63 0xA 0x42 0x74 0x9 (pcmpistri $0x42) 281 340115 Fix none/tests/cmdline[1|2] tests on systems which define TMPDIR 282 340392 Allow user to select more accurate definedness checking in memcheck 283 to avoid invalid complaints on optimised code 284 340430 Fix some grammatical weirdness in the manual. 285 341238 Recognize GCC5/DWARFv5 DW_LANG constants (Go, C11, C++11, C++14) 286 341419 Signal handler ucontext_t not filled out correctly on OS X 287 341539 VG_(describe_addr) should not describe address as belonging to client 288 segment if it is past the heap end 289 341613 Enable building of manythreads and thread-exits tests on Mac OS X 290 341615 Fix none/tests/darwin/access_extended test on Mac OS X 291 341698 Valgrind's AESKEYGENASSIST gives wrong result in words 0 and 2 [..] 292 341789 aarch64: shmat fails with valgrind on ARMv8 293 341997 MIPS64: Cavium OCTEON insns - immediate operand handled incorrectly 294 342008 valgrind.h needs type cast [..] for clang/llvm in 64-bit mode 295 342038 Unhandled syscalls on aarch64 (mbind/get/set_mempolicy) 296 342063 wrong format specifier for test mcblocklistsearch in gdbserver_tests 297 342117 Hang when loading PDB file for MSVC compiled Firefox under Wine 298 342221 socket connect false positive uninit memory for unknown af family 299 342353 Allow dumping full massif output while valgrind is still running 300 342571 Valgrind chokes on AVX compare intrinsic with _CMP_GE_QS 301 == 346476 302 == 348387 303 == 350593 304 342603 Add I2C_SMBUS ioctl support 305 342635 OS X 10.10 (Yosemite) - missing system calls and fcntl code 306 342683 Mark memory past the initial brk limit as unaddressable 307 342783 arm: unhandled instruction 0xEEFE1ACA = "vcvt.s32.f32 s3, s3, #12" 308 342795 Internal glibc __GI_mempcpy call should be intercepted 309 342841 s390x: Support instructions fiebr(a) and fidbr(a) 310 343012 Unhandled syscall 319 (memfd_create) 311 343069 Patch updating v4l2 API support 312 343173 helgrind crash during stack unwind 313 343219 fix GET_STARTREGS for arm 314 343303 Fix known deliberate memory leak in setenv() on Mac OS X 10.10 315 343306 OS X 10.10: UNKNOWN mach_msg unhandled MACH_SEND_TRAILER option 316 343332 Unhandled instruction 0x9E310021 (fcvtmu) on aarch64 317 343335 unhandled instruction 0x1E638400 (fccmp) aarch64 318 343523 OS X mach_ports_register: UNKNOWN task message [id 3403, to [..] 319 343525 OS X host_get_special_port: UNKNOWN host message [id 412, to [..] 320 343597 ppc64le: incorrect use of offseof macro 321 343649 OS X host_create_mach_voucher: UNKNOWN host message [id 222, to [..] 322 343663 OS X 10.10 Memchecj always reports a leak regardless of [..] 323 343732 Unhandled syscall 144 (setgid) on aarch64 324 343733 Unhandled syscall 187 (msgctl and related) on aarch64 325 343802 s390x: False positive "conditional jump or move depends on [..] 326 343902 --vgdb=yes doesn't break when --xml=yes is used 327 343967 Don't warn about setuid/setgid/setcap executable for directories 328 343978 Recognize DWARF5/GCC5 DW_LANG_Fortran 2003 and 2008 constants 329 344007 accept4 syscall unhandled on arm64 (242) and ppc64 (344) 330 344033 Helgrind on ARM32 loses track of mutex state in pthread_cond_wait 331 344054 www - update info for Solaris/illumos 332 344416 'make regtest' does not work cleanly on OS X 333 344235 Remove duplicate include of pub_core_aspacemgr.h 334 344279 syscall sendmmsg on arm64 (269) and ppc32/64 (349) unhandled 335 344295 syscall recvmmsg on arm64 (243) and ppc32/64 (343) unhandled 336 344307 2 unhandled syscalls on aarch64/arm64: umount2(39), mount (40) 337 344314 callgrind_annotate ... warnings about commands containing newlines 338 344318 socketcall should wrap recvmmsg and sendmmsg 339 344337 Fix unhandled syscall: mach:41 (_kernelrpc_mach_port_guard_trap) 340 344416 Fix 'make regtest' does not work cleanly on OS X 341 344499 Fix compilation for Linux kernel >= 4.0.0 342 344512 OS X: unhandled syscall: unix:348 (__pthread_chdir), 343 unix:349 (__pthread_fchdir) 344 344559 Garbage collection of unused segment names in address space manager 345 344560 Fix stack traces missing penultimate frame on OS X 346 344621 Fix memcheck/tests/err_disable4 test on OS X 347 344686 Fix suppression for pthread_rwlock_init on OS X 10.10 348 344702 Fix missing libobjc suppressions on OS X 10.10 349 == 344543 350 344936 Fix unhandled syscall: unix:473 (readlinkat) on OS X 10.10 351 344939 Fix memcheck/tests/xml1 on OS X 10.10 352 345016 helgrind/tests/locked_vs_unlocked2 is failing sometimes 353 345079 Fix build problems in VEX/useful/test_main.c 354 345126 Incorrect handling of VIDIOC_G_AUDIO and G_AUDOUT 355 345177 arm64: prfm (reg) not implemented 356 345215 Performance improvements for the register allocator 357 345248 add support for Solaris OS in valgrind 358 345338 TIOCGSERIAL and TIOCSSERIAL ioctl support on Linux 359 345394 Fix memcheck/tests/strchr on OS X 360 345637 Fix memcheck/tests/sendmsg on OS X 361 345695 Add POWERPC support for AT_DCACHESIZE and HWCAP2 362 345824 Fix aspacem segment mismatch: seen with none/tests/bigcode 363 345887 Fix an assertion in the address space manager 364 345928 amd64: callstack only contains current function for small stacks 365 345984 disInstr(arm): unhandled instruction: 0xEE193F1E 366 345987 MIPS64: Implement cavium LHX instruction 367 346031 MIPS: Implement support for the CvmCount register (rhwr %0, 31) 368 346185 Fix typo saving altivec register v24 369 346267 Compiler warnings for PPC64 code on call to LibVEX_GuestPPC64_get_XER() 370 and LibVEX_GuestPPC64_get_CR() 371 346270 Regression tests none/tests/jm_vec/isa_2_07 and 372 none/tests/test_isa_2_07_part2 have failures on PPC64 little endian 373 346307 fuse filesystem syscall deadlocks 374 346324 PPC64 missing support for lbarx, lharx, stbcx and sthcx instructions 375 346411 MIPS: SysRes::_valEx handling is incorrect 376 346416 Add support for LL_IOC_PATH2FID and LL_IOC_GETPARENT Lustre ioctls 377 346474 PPC64 Power 8, spr TEXASRU register not supported 378 346487 Compiler generates "note" about a future ABI change for PPC64 379 346562 MIPS64: lwl/lwr instructions are performing 64bit loads 380 and causing spurious "invalid read of size 8" warnings 381 346801 Fix link error on OS X: _vgModuleLocal_sf_maybe_extend_stack 382 347151 Fix suppression for pthread_rwlock_init on OS X 10.8 383 347233 Fix memcheck/tests/strchr on OS X 10.10 (Haswell) 384 347322 Power PC regression test cleanup 385 347379 valgrind --leak-check=full leak errors from system libs on OS X 10.8 386 == 217236 387 347389 unhandled syscall: 373 (Linux ARM syncfs) 388 347686 Patch set to cleanup PPC64 regtests 389 347978 Remove bash dependencies where not needed 390 347982 OS X: undefined symbols for architecture x86_64: "_global" [..] 391 347988 Memcheck: the 'impossible' happened: unexpected size for Addr (OSX/wine) 392 == 345929 393 348102 Patch updating v4l2 API support 394 348247 amd64 front end: jno jumps wrongly when overflow is not set 395 348269 Improve mmap MAP_HUGETLB support. 396 348334 (ppc) valgrind does not simulate dcbfl - then my program terminates 397 348345 Assertion fails for negative lineno 398 348377 Unsupported ARM instruction: yield 399 348565 Fix detection of command line option availability for clang 400 348574 vex amd64->IR pcmpistri SSE4.2 unsupported (pcmpistri $0x18) 401 348728 Fix broken check for VIDIOC_G_ENC_INDEX 402 348748 Fix redundant condition 403 348890 Fix clang warning about unsupported --param inline-unit-growth=900 404 348949 Bogus "ERROR: --ignore-ranges: suspiciously large range" 405 349034 Add Lustre ioctls LL_IOC_GROUP_LOCK and LL_IOC_GROUP_UNLOCK 406 349086 Fix UNKNOWN task message [id 3406, to mach_task_self(), [..] 407 349087 Fix UNKNOWN task message [id 3410, to mach_task_self(), [..] 408 349626 Implemented additional Xen hypercalls 409 349769 Clang/osx: ld: warning: -read_only_relocs cannot be used with x86_64 410 349790 Clean up of the hardware capability checking utilities. 411 349828 memcpy intercepts memmove causing src/dst overlap error (ppc64 ld.so) 412 349874 Fix typos in source code 413 349879 memcheck: add handwritten assembly for helperc_LOADV* 414 349941 di_notify_mmap might create wrong start/size DebugInfoMapping 415 350062 vex x86->IR: 0x66 0xF 0x3A 0xB (ROUNDSD) on OS X 416 350202 Add limited param to 'monitor block_list' 417 350290 s390x: Support instructions fixbr(a) 418 350359 memcheck/tests/x86/fxsave hangs indefinetely on OS X 419 350809 Fix none/tests/async-sigs for Solaris 420 350811 Remove reference to --db-attach which has been removed. 421 350813 Memcheck/x86: enable handwritten assembly helpers for x86/Solaris too 422 350854 hard-to-understand code in VG_(load_ELF)() 423 351140 arm64 syscalls setuid (146) and setresgid (149) not implemented 424 351386 Solaris: Cannot run ld.so.1 under Valgrind 425 351474 Fix VG_(iseqsigset) as obvious 426 351534 Fix incorrect header guard 427 351632 Fix UNKNOWN fcntl 97 on OS X 10.11 428 351756 Intercept platform_memchr$VARIANT$Haswell on OS X 429 351858 ldsoexec support on Solaris 430 351873 Newer gcc doesn't allow __builtin_tabortdc[i] in ppc32 mode 431 352130 helgrind reports false races for printfs using mempcpy on FILE* state 432 352284 s390: Conditional jump depends on uninitialised value(s) in vfprintf 433 352320 arm64 crash on none/tests/nestedfs 434 352765 Vbit test fails on Power 6 435 352768 The mbar instruction is missing from the Power PC support 436 352769 Power PC program priority register (PPR) is not supported 437 n-i-bz Provide implementations of certain compiler builtins to support 438 compilers that may not provide those 439 n-i-bz Old STABS code is still being compiled, but never used. Remove it. 440 n-i-bz Fix compilation on distros with glibc < 2.5 441 n-i-bz (vex 3098) Avoid generation of Neon insns on non-Neon hosts 442 n-i-bz Enable rt_sigpending syscall on ppc64 linux. 443 n-i-bz mremap did not work properly on shared memory 444 n-i-bz Fix incorrect sizeof expression in syswrap-xen.c reported by Coverity 445 n-i-bz In VALGRIND_PRINTF write out thread name, if any, to xml 446 447 (3.11.0.TEST1: 8 September 2015, vex r3187, valgrind r15646) 448 (3.11.0.TEST2: 21 September 2015, vex r3193, valgrind r15667) 449 (3.11.0: 22 September 2015, vex r3195, valgrind r15674) 450 451 452 453 Release 3.10.1 (25 November 2014) 454 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 455 3.10.1 is a bug fix release. It fixes various bugs reported in 3.10.0 456 and backports fixes for all reported missing AArch64 ARMv8 instructions 457 and syscalls from the trunk. If you package or deliver 3.10.0 for others 458 to use, you might want to consider upgrading to 3.10.1 instead. 459 460 The following bugs have been fixed or resolved. Note that "n-i-bz" 461 stands for "not in bugzilla" -- that is, a bug that was reported to us 462 but never got a bugzilla entry. We encourage you to file bugs in 463 bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather 464 than mailing the developers (or mailing lists) directly -- bugs that 465 are not entered into bugzilla tend to get forgotten about or ignored. 466 467 To see details of a given bug, visit 468 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 469 where XXXXXX is the bug number as listed below. 470 471 335440 arm64: ld1 (single structure) is not implemented 472 335713 arm64: unhanded instruction: prfm (immediate) 473 339020 ppc64: memcheck/tests/ppc64/power_ISA2_05 failing in nightly build 474 339182 ppc64: AvSplat ought to load destination vector register with [..] 475 339336 PPC64 store quad instruction (stq) is not supposed to change [..] 476 339433 ppc64 lxvw4x instruction uses four 32-byte loads 477 339645 Use correct tag names in sys_getdents/64 wrappers 478 339706 Fix false positive for ioctl(TIOCSIG) on linux 479 339721 assertion 'check_sibling == sibling' failed in readdwarf3.c ... 480 339853 arm64 times syscall unknown 481 339855 arm64 unhandled getsid/setsid syscalls 482 339858 arm64 dmb sy not implemented 483 339926 Unhandled instruction 0x1E674001 (frintx) on aarm64 484 339927 Unhandled instruction 0x9E7100C6 (fcvtmu) on aarch64 485 339938 disInstr(arm64): unhandled instruction 0x4F8010A4 (fmla) 486 == 339950 487 339940 arm64: unhandled syscall: 83 (sys_fdatasync) + patch 488 340033 arm64: unhandled insn dmb ishld and some other isb-dmb-dsb variants 489 340028 unhandled syscalls for arm64 (msync, pread64, setreuid and setregid) 490 340036 arm64: Unhandled instruction ld4 (multiple structures, no offset) 491 340236 arm64: unhandled syscalls: mknodat, fchdir, chroot, fchownat 492 340509 arm64: unhandled instruction fcvtas 493 340630 arm64: fchmod (52) and fchown (55) syscalls not recognized 494 340632 arm64: unhandled instruction fcvtas 495 340722 Resolve "UNKNOWN attrlist flags 0:0x10000000" 496 340725 AVX2: Incorrect decoding of vpbroadcast{b,w} reg,reg forms 497 340788 warning: unhandled syscall: 318 (getrandom) 498 340807 disInstr(arm): unhandled instruction: 0xEE989B20 499 340856 disInstr(arm64): unhandled instruction 0x1E634C45 (fcsel) 500 340922 arm64: unhandled getgroups/setgroups syscalls 501 350251 Fix typo in VEX utility program (test_main.c). 502 350407 arm64: unhandled instruction ucvtf (vector, integer) 503 350809 none/tests/async-sigs breaks when run under cron on Solaris 504 350811 update README.solaris after r15445 505 350813 Use handwritten memcheck assembly helpers on x86/Solaris [..] 506 350854 strange code in VG_(load_ELF)() 507 351140 arm64 syscalls setuid (146) and setresgid (149) not implemented 508 n-i-bz DRD and Helgrind: Handle Imbe_CancelReservation (clrex on ARM) 509 n-i-bz Add missing ]] to terminate CDATA. 510 n-i-bz Glibc versions prior to 2.5 do not define PTRACE_GETSIGINFO 511 n-i-bz Enable sys_fadvise64_64 on arm32. 512 n-i-bz Add test cases for all remaining AArch64 SIMD, FP and memory insns. 513 n-i-bz Add test cases for all known arm64 load/store instructions. 514 n-i-bz PRE(sys_openat): when checking whether ARG1 == VKI_AT_FDCWD [..] 515 n-i-bz Add detection of old ppc32 magic instructions from bug 278808. 516 n-i-bz exp-dhat: Implement missing function "dh_malloc_usable_size". 517 n-i-bz arm64: Implement "fcvtpu w, s". 518 n-i-bz arm64: implement ADDP and various others 519 n-i-bz arm64: Implement {S,U}CVTF (scalar, fixedpt). 520 n-i-bz arm64: enable FCVT{A,N}S X,S. 521 522 (3.10.1: 25 November 2014, vex r3026, valgrind r14785) 523 524 525 526 Release 3.10.0 (10 September 2014) 527 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 528 529 3.10.0 is a feature release with many improvements and the usual 530 collection of bug fixes. 531 532 This release supports X86/Linux, AMD64/Linux, ARM32/Linux, ARM64/Linux, 533 PPC32/Linux, PPC64BE/Linux, PPC64LE/Linux, S390X/Linux, MIPS32/Linux, 534 MIPS64/Linux, ARM/Android, MIPS32/Android, X86/Android, X86/MacOSX 10.9 535 and AMD64/MacOSX 10.9. Support for MacOSX 10.8 and 10.9 is 536 significantly improved relative to the 3.9.0 release. 537 538 * ================== PLATFORM CHANGES ================= 539 540 * Support for the 64-bit ARM Architecture (AArch64 ARMv8). This port 541 is mostly complete, and is usable, but some SIMD instructions are as 542 yet unsupported. 543 544 * Support for little-endian variant of the 64-bit POWER architecture. 545 546 * Support for Android on MIPS32. 547 548 * Support for 64bit FPU on MIPS32 platforms. 549 550 * Both 32- and 64-bit executables are supported on MacOSX 10.8 and 10.9. 551 552 * Configuration for and running on Android targets has changed. 553 See README.android in the source tree for details. 554 555 * ================== DEPRECATED FEATURES ================= 556 557 * --db-attach is now deprecated and will be removed in the next 558 valgrind feature release. The built-in GDB server capabilities are 559 superior and should be used instead. Learn more here: 560 http://valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.gdbserver 561 562 * ==================== TOOL CHANGES ==================== 563 564 * Memcheck: 565 566 - Client code can now selectively disable and re-enable reporting of 567 invalid address errors in specific ranges using the new client 568 requests VALGRIND_DISABLE_ADDR_ERROR_REPORTING_IN_RANGE and 569 VALGRIND_ENABLE_ADDR_ERROR_REPORTING_IN_RANGE. 570 571 - Leak checker: there is a new leak check heuristic called 572 "length64". This is used to detect interior pointers pointing 8 573 bytes inside a block, on the assumption that the first 8 bytes 574 holds the value "block size - 8". This is used by 575 sqlite3MemMalloc, for example. 576 577 - Checking of system call parameters: if a syscall parameter 578 (e.g. bind struct sockaddr, sendmsg struct msghdr, ...) has 579 several fields not initialised, an error is now reported for each 580 field. Previously, an error was reported only for the first 581 uninitialised field. 582 583 - Mismatched alloc/free checking: a new flag 584 --show-mismatched-frees=no|yes [yes] makes it possible to turn off 585 such checks if necessary. 586 587 * Helgrind: 588 589 - Improvements to error messages: 590 591 o Race condition error message involving heap allocated blocks also 592 show the thread number that allocated the raced-on block. 593 594 o All locks referenced by an error message are now announced. 595 Previously, some error messages only showed the lock addresses. 596 597 o The message indicating where a lock was first observed now also 598 describes the address/location of the lock. 599 600 - Helgrind now understands the Ada task termination rules and 601 creates a happens-before relationship between a terminated task 602 and its master. This avoids some false positives and avoids a big 603 memory leak when a lot of Ada tasks are created and terminated. 604 The interceptions are only activated with forthcoming releases of 605 gnatpro >= 7.3.0w-20140611 and gcc >= 5.0. 606 607 - A new GDB server monitor command "info locks" giving the list of 608 locks, their location, and their status. 609 610 * Callgrind: 611 612 - callgrind_control now supports the --vgdb-prefix argument, 613 which is needed if valgrind was started with this same argument. 614 615 * ==================== OTHER CHANGES ==================== 616 617 * Unwinding through inlined function calls. Stack unwinding can now 618 make use of Dwarf3 inlined-unwind information if it is available. 619 The practical effect is that inlined calls become visible in stack 620 traces. The suppression matching machinery has been adjusted 621 accordingly. This is controlled by the new option 622 --read-inline-info=yes|no. Currently this is enabled by default 623 only on Linux and Android targets and only for the tools Memcheck, 624 Helgrind and DRD. 625 626 * Valgrind can now read EXIDX unwind information on 32-bit ARM 627 targets. If an object contains both CFI and EXIDX unwind 628 information, Valgrind will prefer the CFI over the EXIDX. This 629 facilitates unwinding through system libraries on arm-android 630 targets. 631 632 * Address description logic has been improved and is now common 633 between Memcheck and Helgrind, resulting in better address 634 descriptions for some kinds of error messages. 635 636 * Error messages about dubious arguments (eg, to malloc or calloc) are 637 output like other errors. This means that they can be suppressed 638 and they have a stack trace. 639 640 * The C++ demangler has been updated for better C++11 support. 641 642 * New and modified GDB server monitor features: 643 644 - Thread local variables/storage (__thread) can now be displayed. 645 646 - The GDB server monitor command "v.info location <address>" 647 displays information about an address. The information produced 648 depends on the tool and on the options given to valgrind. 649 Possibly, the following are described: global variables, local 650 (stack) variables, allocated or freed blocks, ... 651 652 - The option "--vgdb-stop-at=event1,event2,..." allows the user to 653 ask the GDB server to stop at the start of program execution, at 654 the end of the program execution and on Valgrind internal errors. 655 656 - A new monitor command "v.info stats" shows various Valgrind core 657 and tool statistics. 658 659 - A new monitor command "v.set hostvisibility" allows the GDB server 660 to provide access to Valgrind internal host status/memory. 661 662 * A new option "--aspace-minaddr=<address>" can in some situations 663 allow the use of more memory by decreasing the address above which 664 Valgrind maps memory. It can also be used to solve address 665 conflicts with system libraries by increasing the default value. 666 See user manual for details. 667 668 * The amount of memory used by Valgrind to store debug info (unwind 669 info, line number information and symbol data) has been 670 significantly reduced, even though Valgrind now reads more 671 information in order to support unwinding of inlined function calls. 672 673 * Dwarf3 handling with --read-var-info=yes has been improved: 674 675 - Ada and C struct containing VLAs no longer cause a "bad DIE" error 676 677 - Code compiled with 678 -ffunction-sections -fdata-sections -Wl,--gc-sections 679 no longer causes assertion failures. 680 681 * Improved checking for the --sim-hints= and --kernel-variant= 682 options. Unknown strings are now detected and reported to the user 683 as a usage error. 684 685 * The semantics of stack start/end boundaries in the valgrind.h 686 VALGRIND_STACK_REGISTER client request has been clarified and 687 documented. The convention is that start and end are respectively 688 the lowest and highest addressable bytes of the stack. 689 690 * ==================== FIXED BUGS ==================== 691 692 The following bugs have been fixed or resolved. Note that "n-i-bz" 693 stands for "not in bugzilla" -- that is, a bug that was reported to us 694 but never got a bugzilla entry. We encourage you to file bugs in 695 bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather 696 than mailing the developers (or mailing lists) directly -- bugs that 697 are not entered into bugzilla tend to get forgotten about or ignored. 698 699 To see details of a given bug, visit 700 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 701 where XXXXXX is the bug number as listed below. 702 703 175819 Support for ipv6 socket reporting with --track-fds 704 232510 make distcheck fails 705 249435 Analyzing wine programs with callgrind triggers a crash 706 278972 support for inlined function calls in stacktraces and suppression 707 == 199144 708 291310 FXSAVE instruction marks memory as undefined on amd64 709 303536 ioctl for SIOCETHTOOL (ethtool(8)) isn't wrapped 710 308729 vex x86->IR: unhandled instruction bytes 0xf 0x5 (syscall) 711 315199 vgcore file for threaded app does not show which thread crashed 712 315952 tun/tap ioctls are not supported 713 323178 Unhandled instruction: PLDW register (ARM) 714 323179 Unhandled instruction: PLDW immediate (ARM) 715 324050 Helgrind: SEGV because of unaligned stack when using movdqa 716 325110 Add test-cases for Power ISA 2.06 insns: divdo/divdo. and divduo/divduo. 717 325124 [MIPSEL] Compilation error 718 325477 Phase 4 support for IBM Power ISA 2.07 719 325538 cavium octeon mips64, valgrind reported "dumping core" [...] 720 325628 Phase 5 support for IBM Power ISA 2.07 721 325714 Empty vgcore but RLIMIT_CORE is big enough (too big) 722 325751 Missing the two privileged Power PC Transactional Memory Instructions 723 325816 Phase 6 support for IBM Power ISA 2.07 724 325856 Make SGCheck fail gracefully on unsupported platforms 725 326026 Iop names for count leading zeros/sign bits incorrectly imply [..] 726 326436 DRD: False positive in libstdc++ std::list::push_back 727 326444 Cavium MIPS Octeon Specific Load Indexed Instructions 728 326462 Refactor vgdb to isolate invoker stuff into separate module 729 326469 amd64->IR: 0x66 0xF 0x3A 0x63 0xC1 0xE (pcmpistri 0x0E) 730 326623 DRD: false positive conflict report in a field assignment 731 326724 Valgrind does not compile on OSX 1.9 Mavericks 732 326816 Intercept for __strncpy_sse2_unaligned missing? 733 326921 coregrind fails to compile m_trampoline.S with MIPS/Linux port of V 734 326983 Clear direction flag after tests on amd64. 735 327212 Do not prepend the current directory to absolute path names. 736 327223 Support for Cavium MIPS Octeon Atomic and Count Instructions 737 327238 Callgrind Assertion 'passed <= last_bb->cjmp_count' failed 738 327284 s390x: Fix translation of the risbg instruction 739 327639 vex amd64->IR pcmpestri SSE4.2 instruction is unsupported 0x34 740 327837 dwz compressed alternate .debug_info and .debug_str not read correctly 741 327916 DW_TAG_typedef may have no name 742 327943 s390x: add a redirection for the 'index' function 743 328100 XABORT not implemented 744 328205 Implement additional Xen hypercalls 745 328454 add support Backtraces with ARM unwind tables (EXIDX) 746 328455 s390x: SIGILL after emitting wrong register pair for ldxbr 747 328711 valgrind.1 manpage "memcheck options" section is badly generated 748 328878 vex amd64->IR pcmpestri SSE4.2 instruction is unsupported 0x14 749 329612 Incorrect handling of AT_BASE for image execution 750 329694 clang warns about using uninitialized variable 751 329956 valgrind crashes when lmw/stmw instructions are used on ppc64 752 330228 mmap must align to VKI_SHMLBA on mips32 753 330257 LLVM does not support `-mno-dynamic-no-pic` option 754 330319 amd64->IR: unhandled instruction bytes: 0xF 0x1 0xD5 (xend) 755 330459 --track-fds=yes doesn't track eventfds 756 330469 Add clock_adjtime syscall support 757 330594 Missing sysalls on PowerPC / uClibc 758 330622 Add test to regression suite for POWER instruction: dcbzl 759 330939 Support for AMD's syscall instruction on x86 760 == 308729 761 330941 Typo in PRE(poll) syscall wrapper 762 331057 unhandled instruction: 0xEEE01B20 (vfma.f64) (has patch) 763 331254 Fix expected output for memcheck/tests/dw4 764 331255 Fix race condition in test none/tests/coolo_sigaction 765 331257 Fix type of jump buffer in test none/tests/faultstatus 766 331305 configure uses bash specific syntax 767 331337 s390x WARNING: unhandled syscall: 326 (dup3) 768 331380 Syscall param timer_create(evp) points to uninitialised byte(s) 769 331476 Patch to handle ioctl 0x5422 on Linux (x86 and amd64) 770 331829 Unexpected ioctl opcode sign extension 771 331830 ppc64: WARNING: unhandled syscall: 96/97 772 331839 drd/tests/sem_open specifies invalid semaphore name 773 331847 outcome of drd/tests/thread_name is nondeterministic 774 332037 Valgrind cannot handle Thumb "add pc, reg" 775 332055 drd asserts on platforms with VG_STACK_REDZONE_SZB == 0 and 776 consistency checks enabled 777 332263 intercepts for pthread_rwlock_timedrdlock and 778 pthread_rwlock_timedwrlock are incorrect 779 332265 drd could do with post-rwlock_init and pre-rwlock_destroy 780 client requests 781 332276 Implement additional Xen hypercalls 782 332658 ldrd.w r1, r2, [PC, #imm] does not adjust for 32bit alignment 783 332765 Fix ms_print to create temporary files in a proper directory 784 333072 drd: Add semaphore annotations 785 333145 Tests for missaligned PC+#imm access for arm 786 333228 AAarch64 Missing instruction encoding: mrs %[reg], ctr_el0 787 333230 AAarch64 missing instruction encodings: dc, ic, dsb. 788 333248 WARNING: unhandled syscall: unix:443 789 333428 ldr.w pc [rD, #imm] instruction leads to assertion 790 333501 cachegrind: assertion: Cache set count is not a power of two. 791 == 336577 792 == 292281 793 333666 Recognize MPX instructions and bnd prefix. 794 333788 Valgrind does not support the CDROM_DISC_STATUS ioctl (has patch) 795 333817 Valgrind reports the memory areas written to by the SG_IO 796 ioctl as untouched 797 334049 lzcnt fails silently (x86_32) 798 334384 Valgrind does not have support Little Endian support for 799 IBM POWER PPC 64 800 334585 recvmmsg unhandled (+patch) (arm) 801 334705 sendmsg and recvmsg should guard against bogus msghdr fields. 802 334727 Build fails with -Werror=format-security 803 334788 clarify doc about --log-file initial program directory 804 334834 PPC64 Little Endian support, patch 2 805 334836 PPC64 Little Endian support, patch 3 testcase fixes 806 334936 patch to fix false positives on alsa SNDRV_CTL_* ioctls 807 335034 Unhandled ioctl: HCIGETDEVLIST 808 335155 vgdb, fix error print statement. 809 335262 arm64: movi 8bit version is not supported 810 335263 arm64: dmb instruction is not implemented 811 335441 unhandled ioctl 0x8905 (SIOCATMARK) when running wine under valgrind 812 335496 arm64: sbc/abc instructions are not implemented 813 335554 arm64: unhandled instruction: abs 814 335564 arm64: unhandled instruction: fcvtpu Xn, Sn 815 335735 arm64: unhandled instruction: cnt 816 335736 arm64: unhandled instruction: uaddlv 817 335848 arm64: unhandled instruction: {s,u}cvtf 818 335902 arm64: unhandled instruction: sli 819 335903 arm64: unhandled instruction: umull (vector) 820 336055 arm64: unhandled instruction: mov (element) 821 336062 arm64: unhandled instruction: shrn{,2} 822 336139 mip64: [...] valgrind hangs and spins on a single core [...] 823 336189 arm64: unhandled Instruction: mvn 824 336435 Valgrind hangs in pthread_spin_lock consuming 100% CPU 825 336619 valgrind --read-var-info=yes doesn't handle DW_TAG_restrict_type 826 336772 Make moans about unknown ioctls more informative 827 336957 Add a section about the Solaris/illumos port on the webpage 828 337094 ifunc wrapper is broken on ppc64 829 337285 fcntl commands F_OFD_SETLK, F_OFD_SETLKW, and F_OFD_GETLK not supported 830 337528 leak check heuristic for block prefixed by length as 64bit number 831 337740 Implement additional Xen hypercalls 832 337762 guest_arm64_toIR.c:4166 (dis_ARM64_load_store): Assertion `0' failed. 833 337766 arm64-linux: unhandled syscalls mlock (228) and mlockall (230) 834 337871 deprecate --db-attach 835 338023 Add support for all V4L2/media ioctls 836 338024 inlined functions are not shown if DW_AT_ranges is used 837 338106 Add support for 'kcmp' syscall 838 338115 DRD: computed conflict set differs from actual after fork 839 338160 implement display of thread local storage in gdbsrv 840 338205 configure.ac and check for -Wno-tautological-compare 841 338300 coredumps are missing one byte of every segment 842 338445 amd64 vbit-test fails with unknown opcodes used by arm64 VEX 843 338499 --sim-hints parsing broken due to wrong order in tokens 844 338615 suppress glibc 2.20 optimized strcmp implementation for ARMv7 845 338681 Unable to unwind through clone thread created on i386-linux 846 338698 race condition between gdbsrv and vgdb on startup 847 338703 helgrind on arm-linux gets false positives in dynamic loader 848 338791 alt dwz files can be relative of debug/main file 849 338878 on MacOS: assertion 'VG_IS_PAGE_ALIGNED(clstack_end+1)' failed 850 338932 build V-trunk with gcc-trunk 851 338974 glibc 2.20 changed size of struct sigaction sa_flags field on s390 852 345079 Fix build problems in VEX/useful/test_main.c 853 n-i-bz Fix KVM_CREATE_IRQCHIP ioctl handling 854 n-i-bz s390x: Fix memory corruption for multithreaded applications 855 n-i-bz vex arm->IR: allow PC as basereg in some LDRD cases 856 n-i-bz internal error in Valgrind if vgdb transmit signals when ptrace invoked 857 n-i-bz Fix mingw64 support in valgrind.h (dev@, 9 May 2014) 858 n-i-bz drd manual: Document how to C++11 programs that use class "std::thread" 859 n-i-bz Add command-line option --default-suppressions 860 n-i-bz Add support for BLKDISCARDZEROES ioctl 861 n-i-bz ppc32/64: fix a regression with the mtfsb0/mtfsb1 instructions 862 n-i-bz Add support for sys_pivot_root and sys_unshare 863 864 (3.10.0.BETA1: 2 September 2014, vex r2940, valgrind r14428) 865 (3.10.0.BETA2: 8 September 2014, vex r2950, valgrind r14503) 866 (3.10.0: 10 September 2014, vex r2950, valgrind r14514) 867 868 869 870 Release 3.9.0 (31 October 2013) 871 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 872 3.9.0 is a feature release with many improvements and the usual 873 collection of bug fixes. 874 875 This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux, 876 PPC64/Linux, S390X/Linux, MIPS32/Linux, MIPS64/Linux, ARM/Android, 877 X86/Android, X86/MacOSX 10.7 and AMD64/MacOSX 10.7. Support for 878 MacOSX 10.8 is significantly improved relative to the 3.8.0 release. 879 880 * ================== PLATFORM CHANGES ================= 881 882 * Support for MIPS64 LE and BE running Linux. Valgrind has been 883 tested on MIPS64 Debian Squeeze and Debian Wheezy distributions. 884 885 * Support for MIPS DSP ASE on MIPS32 platforms. 886 887 * Support for s390x Decimal Floating Point instructions on hosts that 888 have the DFP facility installed. 889 890 * Support for POWER8 (Power ISA 2.07) instructions 891 892 * Support for Intel AVX2 instructions. This is available only on 64 893 bit code. 894 895 * Initial support for Intel Transactional Synchronization Extensions, 896 both RTM and HLE. 897 898 * Initial support for Hardware Transactional Memory on POWER. 899 900 * Improved support for MacOSX 10.8 (64-bit only). Memcheck can now 901 run large GUI apps tolerably well. 902 903 * ==================== TOOL CHANGES ==================== 904 905 * Memcheck: 906 907 - Improvements in handling of vectorised code, leading to 908 significantly fewer false error reports. You need to use the flag 909 --partial-loads-ok=yes to get the benefits of these changes. 910 911 - Better control over the leak checker. It is now possible to 912 specify which leak kinds (definite/indirect/possible/reachable) 913 should be displayed, which should be regarded as errors, and which 914 should be suppressed by a given leak suppression. This is done 915 using the options --show-leak-kinds=kind1,kind2,.., 916 --errors-for-leak-kinds=kind1,kind2,.. and an optional 917 "match-leak-kinds:" line in suppression entries, respectively. 918 919 Note that generated leak suppressions contain this new line and 920 are therefore more specific than in previous releases. To get the 921 same behaviour as previous releases, remove the "match-leak-kinds:" 922 line from generated suppressions before using them. 923 924 - Reduced "possible leak" reports from the leak checker by the use 925 of better heuristics. The available heuristics provide detection 926 of valid interior pointers to std::stdstring, to new[] allocated 927 arrays with elements having destructors and to interior pointers 928 pointing to an inner part of a C++ object using multiple 929 inheritance. They can be selected individually using the 930 option --leak-check-heuristics=heur1,heur2,... 931 932 - Better control of stacktrace acquisition for heap-allocated 933 blocks. Using the --keep-stacktraces option, it is possible to 934 control independently whether a stack trace is acquired for each 935 allocation and deallocation. This can be used to create better 936 "use after free" errors or to decrease Valgrind's resource 937 consumption by recording less information. 938 939 - Better reporting of leak suppression usage. The list of used 940 suppressions (shown when the -v option is given) now shows, for 941 each leak suppressions, how many blocks and bytes it suppressed 942 during the last leak search. 943 944 * Helgrind: 945 946 - False errors resulting from the use of statically initialised 947 mutexes and condition variables (PTHREAD_MUTEX_INITIALISER, etc) 948 have been removed. 949 950 - False errors resulting from the use of pthread_cond_waits that 951 timeout, have been removed. 952 953 * ==================== OTHER CHANGES ==================== 954 955 * Some attempt to tune Valgrind's space requirements to the expected 956 capabilities of the target: 957 958 - The default size of the translation cache has been reduced from 8 959 sectors to 6 on Android platforms, since each sector occupies 960 about 40MB when using Memcheck. 961 962 - The default size of the translation cache has been increased to 16 963 sectors on all other platforms, reflecting the fact that large 964 applications require instrumentation and storage of huge amounts 965 of code. For similar reasons, the number of memory mapped 966 segments that can be tracked has been increased by a factor of 6. 967 968 - In all cases, the maximum number of sectors in the translation 969 cache can be controlled by the new flag --num-transtab-sectors. 970 971 * Changes in how debug info (line numbers, etc) is read: 972 973 - Valgrind no longer temporarily mmaps the entire object to read 974 from it. Instead, reading is done through a small fixed sized 975 buffer. This avoids virtual memory usage spikes when Valgrind 976 reads debuginfo from large shared objects. 977 978 - A new experimental remote debug info server. Valgrind can read 979 debug info from a different machine (typically, a build host) 980 where debuginfo objects are stored. This can save a lot of time 981 and hassle when running Valgrind on resource-constrained targets 982 (phones, tablets) when the full debuginfo objects are stored 983 somewhere else. This is enabled by the --debuginfo-server= 984 option. 985 986 - Consistency checking between main and debug objects can be 987 disabled using the --allow-mismatched-debuginfo option. 988 989 * Stack unwinding by stack scanning, on ARM. Unwinding by stack 990 scanning can recover stack traces in some cases when the normal 991 unwind mechanisms fail. Stack scanning is best described as "a 992 nasty, dangerous and misleading hack" and so is disabled by default. 993 Use --unw-stack-scan-thresh and --unw-stack-scan-frames to enable 994 and control it. 995 996 * Detection and merging of recursive stack frame cycles. When your 997 program has recursive algorithms, this limits the memory used by 998 Valgrind for recorded stack traces and avoids recording 999 uninteresting repeated calls. This is controlled by the command 1000 line option --merge-recursive-frame and by the monitor command 1001 "v.set merge-recursive-frames". 1002 1003 * File name and line numbers for used suppressions. The list of used 1004 suppressions (shown when the -v option is given) now shows, for each 1005 used suppression, the file name and line number where the suppression 1006 is defined. 1007 1008 * New and modified GDB server monitor features: 1009 1010 - valgrind.h has a new client request, VALGRIND_MONITOR_COMMAND, 1011 that can be used to execute gdbserver monitor commands from the 1012 client program. 1013 1014 - A new monitor command, "v.info open_fds", that gives the list of 1015 open file descriptors and additional details. 1016 1017 - An optional message in the "v.info n_errs_found" monitor command, 1018 for example "v.info n_errs_found test 1234 finished", allowing a 1019 comment string to be added to the process output, perhaps for the 1020 purpose of separating errors of different tests or test phases. 1021 1022 - A new monitor command "v.info execontext" that shows information 1023 about the stack traces recorded by Valgrind. 1024 1025 - A new monitor command "v.do expensive_sanity_check_general" to run 1026 some internal consistency checks. 1027 1028 * New flag --sigill-diagnostics to control whether a diagnostic 1029 message is printed when the JIT encounters an instruction it can't 1030 translate. The actual behavior -- delivery of SIGILL to the 1031 application -- is unchanged. 1032 1033 * The maximum amount of memory that Valgrind can use on 64 bit targets 1034 has been increased from 32GB to 64GB. This should make it possible 1035 to run applications on Memcheck that natively require up to about 35GB. 1036 1037 * ==================== FIXED BUGS ==================== 1038 1039 The following bugs have been fixed or resolved. Note that "n-i-bz" 1040 stands for "not in bugzilla" -- that is, a bug that was reported to us 1041 but never got a bugzilla entry. We encourage you to file bugs in 1042 bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather 1043 than mailing the developers (or mailing lists) directly -- bugs that 1044 are not entered into bugzilla tend to get forgotten about or ignored. 1045 1046 To see details of a given bug, visit 1047 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 1048 where XXXXXX is the bug number as listed below. 1049 1050 123837 system call: 4th argument is optional, depending on cmd 1051 135425 memcheck should tell you where Freed blocks were Mallocd 1052 164485 VG_N_SEGNAMES and VG_N_SEGMENTS are (still) too small 1053 207815 Adds some of the drm ioctls to syswrap-linux.c 1054 251569 vex amd64->IR: 0xF 0x1 0xF9 0xBF 0x90 0xD0 0x3 0x0 (RDTSCP) 1055 252955 Impossible to compile with ccache 1056 253519 Memcheck reports auxv pointer accesses as invalid reads. 1057 263034 Crash when loading some PPC64 binaries 1058 269599 Increase deepest backtrace 1059 274695 s390x: Support "compare to/from logical" instructions (z196) 1060 275800 s390x: Autodetect cache info (part 2) 1061 280271 Valgrind reports possible memory leaks on still-reachable std::string 1062 284540 Memcheck shouldn't count suppressions matching still-reachable [..] 1063 289578 Backtraces with ARM unwind tables (stack scan flags) 1064 296311 Wrong stack traces due to -fomit-frame-pointer (x86) 1065 304832 ppc32: build failure 1066 305431 Use find_buildid shdr fallback for separate .debug files 1067 305728 Add support for AVX2 instructions 1068 305948 ppc64: code generation for ShlD64 / ShrD64 asserts 1069 306035 s390x: Fix IR generation for LAAG and friends 1070 306054 s390x: Condition code computation for convert-to-int/logical 1071 306098 s390x: alternate opcode form for convert to/from fixed 1072 306587 Fix cache line detection from auxiliary vector for PPC. 1073 306783 Mips unhandled syscall : 4025 / 4079 / 4182 1074 307038 DWARF2 CFI reader: unhandled DW_OP_ opcode 0x8 (DW_OP_const1u et al) 1075 307082 HG false positive: pthread_cond_destroy: destruction of unknown CV 1076 307101 sys_capget second argument can be NULL 1077 307103 sys_openat: If pathname is absolute, then dirfd is ignored. 1078 307106 amd64->IR: f0 0f c0 02 (lock xadd byte) 1079 307113 s390x: DFP support 1080 307141 valgrind does't work in mips-linux system 1081 307155 filter_gdb should filter out syscall-template.S T_PSEUDO 1082 307285 x86_amd64 feature test for avx in test suite is wrong 1083 307290 memcheck overlap testcase needs memcpy version filter 1084 307463 Please add "&limit=0" to the "all open bugs" link 1085 307465 --show-possibly-lost=no should reduce the error count / exit code 1086 307557 Leaks on Mac OS X 10.7.5 libraries at ImageLoader::recursiveInit[..] 1087 307729 pkgconfig support broken valgrind.pc 1088 307828 Memcheck false errors SSE optimized wcscpy, wcscmp, wcsrchr, wcschr 1089 307955 Building valgrind 3.7.0-r4 fails in Gentoo AMD64 when using clang 1090 308089 Unhandled syscall on ppc64: prctl 1091 308135 PPC32 MPC8xx has 16 bytes cache size 1092 308321 testsuite memcheck filter interferes with gdb_filter 1093 308333 == 307106 1094 308341 vgdb should report process exit (or fatal signal) 1095 308427 s390 memcheck reports tsearch cjump/cmove depends on uninit 1096 308495 Remove build dependency on installed Xen headers 1097 308573 Internal error on 64-bit instruction executed in 32-bit mode 1098 308626 == 308627 1099 308627 pmovmskb validity bit propagation is imprecise 1100 308644 vgdb command for having the info for the track-fds option 1101 308711 give more info about aspacemgr and arenas in out_of_memory 1102 308717 ARM: implement fixed-point VCVT.F64.[SU]32 1103 308718 ARM implement SMLALBB family of instructions 1104 308886 Missing support for PTRACE_SET/GETREGSET 1105 308930 syscall name_to_handle_at (303 on amd64) not handled 1106 309229 V-bit tester does not report number of tests generated 1107 309323 print unrecognized instuction on MIPS 1108 309425 Provide a --sigill-diagnostics flag to suppress illegal [..] 1109 309427 SSE optimized stpncpy trigger uninitialised value [..] errors 1110 309430 Self hosting ppc64 encounters a vassert error on operand type 1111 309600 valgrind is a bit confused about 0-sized sections 1112 309823 Generate errors for still reachable blocks 1113 309921 PCMPISTRI validity bit propagation is imprecise 1114 309922 none/tests/ppc64/test_dfp5 sometimes fails 1115 310169 The Iop_CmpORD class of Iops is not supported by the vbit checker. 1116 310424 --read-var-info does not properly describe static variables 1117 310792 search additional path for debug symbols 1118 310931 s390x: Message-security assist (MSA) instruction extension [..] 1119 311100 PPC DFP implementation of the integer operands is inconsistent [..] 1120 311318 ARM: "128-bit constant is not implemented" error message 1121 311407 ssse3 bcopy (actually converted memcpy) causes invalid read [..] 1122 311690 V crashes because it redirects branches inside of a redirected function 1123 311880 x86_64: make regtest hangs at shell_valid1 1124 311922 WARNING: unhandled syscall: 170 1125 311933 == 251569 1126 312171 ppc: insn selection for DFP 1127 312571 Rounding mode call wrong for the DFP Iops [..] 1128 312620 Change to Iop_D32toD64 [..] for s390 DFP support broke ppc [..] 1129 312913 Dangling pointers error should also report the alloc stack trace 1130 312980 Building on Mountain Lion generates some compiler warnings 1131 313267 Adding MIPS64/Linux port to Valgrind 1132 313348 == 251569 1133 313354 == 251569 1134 313811 Buffer overflow in assert_fail 1135 314099 coverity pointed out error in VEX guest_ppc_toIR.c insn_suffix 1136 314269 ppc: dead code in insn selection 1137 314718 ARM: implement integer divide instruction (sdiv and udiv) 1138 315345 cl-format.xml and callgrind/dump.c don't agree on using cfl= or cfi= 1139 315441 sendmsg syscall should ignore unset msghdr msg_flags 1140 315534 msgrcv inside a thread causes valgrind to hang (block) 1141 315545 Assertion '(UChar*)sec->tt[tteNo].tcptr <= (UChar*)hcode' failed 1142 315689 disInstr(thumb): unhandled instruction: 0xF852 0x0E10 (LDRT) 1143 315738 disInstr(arm): unhandled instruction: 0xEEBE0BEE (vcvt.s32.f64) 1144 315959 valgrind man page has bogus SGCHECK (and no BBV) OPTIONS section 1145 316144 valgrind.1 manpage contains unknown ??? strings [..] 1146 316145 callgrind command line options in manpage reference (unknown) [..] 1147 316145 callgrind command line options in manpage reference [..] 1148 316181 drd: Fixed a 4x slowdown for certain applications 1149 316503 Valgrind does not support SSE4 "movntdqa" instruction 1150 316535 Use of |signed int| instead of |size_t| in valgrind messages 1151 316696 fluidanimate program of parsec 2.1 stuck 1152 316761 syscall open_by_handle_at (304 on amd64, 342 on x86) not handled 1153 317091 Use -Wl,-Ttext-segment when static linking if possible [..] 1154 317186 "Impossible happens" when occurs VCVT instruction on ARM 1155 317318 Support for Threading Building Blocks "scalable_malloc" 1156 317444 amd64->IR: 0xC4 0x41 0x2C 0xC2 0xD2 0x8 (vcmpeq_uqps) 1157 317461 Fix BMI assembler configure check and avx2/bmi/fma vgtest prereqs 1158 317463 bmi testcase IR SANITY CHECK FAILURE 1159 317506 memcheck/tests/vbit-test fails with unknown opcode after [..] 1160 318050 libmpiwrap fails to compile with out-of-source build 1161 318203 setsockopt handling needs to handle SOL_SOCKET/SO_ATTACH_FILTER 1162 318643 annotate_trace_memory tests infinite loop on arm and ppc [..] 1163 318773 amd64->IR: 0xF3 0x48 0x0F 0xBC 0xC2 0xC3 0x66 0x0F 1164 318929 Crash with: disInstr(thumb): 0xF321 0x0001 (ssat16) 1165 318932 Add missing PPC64 and PPC32 system call support 1166 319235 --db-attach=yes is broken with Yama (ptrace scoping) enabled 1167 319395 Crash with unhandled instruction on STRT (Thumb) instructions 1168 319494 VEX Makefile-gcc standalone build update after r2702 1169 319505 [MIPSEL] Crash: unhandled UNRAY operator. 1170 319858 disInstr(thumb): unhandled instruction on instruction STRBT 1171 319932 disInstr(thumb): unhandled instruction on instruction STRHT 1172 320057 Problems when we try to mmap more than 12 memory pages on MIPS32 1173 320063 Memory from PTRACE_GET_THREAD_AREA is reported uninitialised 1174 320083 disInstr(thumb): unhandled instruction on instruction LDRBT 1175 320116 bind on AF_BLUETOOTH produces warnings because of sockaddr_rc padding 1176 320131 WARNING: unhandled syscall: 369 on ARM (prlimit64) 1177 320211 Stack buffer overflow in ./coregrind/m_main.c with huge TMPDIR 1178 320661 vgModuleLocal_read_elf_debug_info(): "Assertion '!di->soname' 1179 320895 add fanotify support (patch included) 1180 320998 vex amd64->IR pcmpestri and pcmpestrm SSE4.2 instruction 1181 321065 Valgrind updates for Xen 4.3 1182 321148 Unhandled instruction: PLI (Thumb 1, 2, 3) 1183 321363 Unhandled instruction: SSAX (ARM + Thumb) 1184 321364 Unhandled instruction: SXTAB16 (ARM + Thumb) 1185 321466 Unhandled instruction: SHASX (ARM + Thumb) 1186 321467 Unhandled instruction: SHSAX (ARM + Thumb) 1187 321468 Unhandled instruction: SHSUB16 (ARM + Thumb) 1188 321619 Unhandled instruction: SHSUB8 (ARM + Thumb) 1189 321620 Unhandled instruction: UASX (ARM + Thumb) 1190 321621 Unhandled instruction: USAX (ARM + Thumb) 1191 321692 Unhandled instruction: UQADD16 (ARM + Thumb) 1192 321693 Unhandled instruction: LDRSBT (Thumb) 1193 321694 Unhandled instruction: UQASX (ARM + Thumb) 1194 321696 Unhandled instruction: UQSAX (Thumb + ARM) 1195 321697 Unhandled instruction: UHASX (ARM + Thumb) 1196 321703 Unhandled instruction: UHSAX (ARM + Thumb) 1197 321704 Unhandled instruction: REVSH (ARM + Thumb) 1198 321730 Add cg_diff and cg_merge man pages 1199 321738 Add vgdb and valgrind-listener man pages 1200 321814 == 315545 1201 321891 Unhandled instruction: LDRHT (Thumb) 1202 321960 pthread_create() then alloca() causing invalid stack write errors 1203 321969 ppc32 and ppc64 don't support [lf]setxattr 1204 322254 Show threadname together with tid if set by application 1205 322294 Add initial support for IBM Power ISA 2.07 1206 322368 Assertion failure in wqthread_hijack under OS X 10.8 1207 322563 vex mips->IR: 0x70 0x83 0xF0 0x3A 1208 322807 VALGRIND_PRINTF_BACKTRACE writes callstack to xml and text to stderr 1209 322851 0bXXX binary literal syntax is not standard 1210 323035 Unhandled instruction: LDRSHT(Thumb) 1211 323036 Unhandled instruction: SMMLS (ARM and Thumb) 1212 323116 The memcheck/tests/ppc64/power_ISA2_05.c fails to build [..] 1213 323175 Unhandled instruction: SMLALD (ARM + Thumb) 1214 323177 Unhandled instruction: SMLSLD (ARM + Thumb) 1215 323432 Calling pthread_cond_destroy() or pthread_mutex_destroy() [..] 1216 323437 Phase 2 support for IBM Power ISA 2.07 1217 323713 Support mmxext (integer sse) subset on i386 (athlon) 1218 323803 Transactional memory instructions are not supported for Power 1219 323893 SSE3 not available on amd cpus in valgrind 1220 323905 Probable false positive from Valgrind/drd on close() 1221 323912 valgrind.h header isn't compatible for mingw64 1222 324047 Valgrind doesn't support [LDR,ST]{S}[B,H]T ARM instructions 1223 324149 helgrind: When pthread_cond_timedwait returns ETIMEDOUT [..] 1224 324181 mmap does not handle MAP_32BIT 1225 324227 memcheck false positive leak when a thread calls exit+block [..] 1226 324421 Support for fanotify API on ARM architecture 1227 324514 gdbserver monitor cmd output behaviour consistency [..] 1228 324518 ppc64: Emulation of dcbt instructions does not handle [..] 1229 324546 none/tests/ppc32 test_isa_2_07_part2 requests -m64 1230 324582 When access is made to freed memory, report both allocation [..] 1231 324594 Fix overflow computation for Power ISA 2.06 insns: mulldo/mulldo. 1232 324765 ppc64: illegal instruction when executing none/tests/ppc64/jm-misc 1233 324816 Incorrect VEX implementation for xscvspdp/xvcvspdp for SNaN inputs 1234 324834 Unhandled instructions in Microsoft C run-time for x86_64 1235 324894 Phase 3 support for IBM Power ISA 2.07 1236 326091 drd: Avoid false race reports from optimized strlen() impls 1237 326113 valgrind libvex hwcaps error on AMD64 1238 n-i-bz Some wrong command line options could be ignored 1239 n-i-bz patch to allow fair-sched on android 1240 n-i-bz report error for vgdb snapshot requested before execution 1241 n-i-bz same as 303624 (fixed in 3.8.0), but for x86 android 1242 1243 (3.9.0: 31 October 2013, vex r2796, valgrind r13708) 1244 1245 1246 1247 Release 3.8.1 (19 September 2012) 1248 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1249 3.8.1 is a bug fix release. It fixes some assertion failures in 3.8.0 1250 that occur moderately frequently in real use cases, adds support for 1251 some missing instructions on ARM, and fixes a deadlock condition on 1252 MacOSX. If you package or deliver 3.8.0 for others to use, you might 1253 want to consider upgrading to 3.8.1 instead. 1254 1255 The following bugs have been fixed or resolved. Note that "n-i-bz" 1256 stands for "not in bugzilla" -- that is, a bug that was reported to us 1257 but never got a bugzilla entry. We encourage you to file bugs in 1258 bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather 1259 than mailing the developers (or mailing lists) directly -- bugs that 1260 are not entered into bugzilla tend to get forgotten about or ignored. 1261 1262 To see details of a given bug, visit 1263 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 1264 where XXXXXX is the bug number as listed below. 1265 1266 284004 == 301281 1267 289584 Unhandled instruction: 0xF 0x29 0xE5 (MOVAPS) 1268 295808 amd64->IR: 0xF3 0xF 0xBC 0xC0 (TZCNT) 1269 298281 wcslen causes false(?) uninitialised value warnings 1270 301281 valgrind hangs on OS X when the process calls system() 1271 304035 disInstr(arm): unhandled instruction 0xE1023053 1272 304867 implement MOVBE instruction in x86 mode 1273 304980 Assertion 'lo <= hi' failed in vgModuleLocal_find_rx_mapping 1274 305042 amd64: implement 0F 7F encoding of movq between two registers 1275 305199 ARM: implement QDADD and QDSUB 1276 305321 amd64->IR: 0xF 0xD 0xC (prefetchw) 1277 305513 killed by fatal signal: SIGSEGV 1278 305690 DRD reporting invalid semaphore when sem_trywait fails 1279 305926 Invalid alignment checks for some AVX instructions 1280 306297 disInstr(thumb): unhandled instruction 0xE883 0x000C 1281 306310 3.8.0 release tarball missing some files 1282 306612 RHEL 6 glibc-2.X default suppressions need /lib*/libc-*patterns 1283 306664 vex amd64->IR: 0x66 0xF 0x3A 0x62 0xD1 0x46 0x66 0xF 1284 n-i-bz shmat of a segment > 4Gb does not work 1285 n-i-bz simulate_control_c script wrong USR1 signal number on mips 1286 n-i-bz vgdb ptrace calls wrong on mips [...] 1287 n-i-bz Fixes for more MPI false positives 1288 n-i-bz exp-sgcheck's memcpy causes programs to segfault 1289 n-i-bz OSX build w/ clang: asserts at startup 1290 n-i-bz Incorrect undef'dness prop for Iop_DPBtoBCD and Iop_BCDtoDPB 1291 n-i-bz fix a couple of union tag-vs-field mixups 1292 n-i-bz OSX: use __NR_poll_nocancel rather than __NR_poll 1293 1294 The following bugs were fixed in 3.8.0 but not listed in this NEWS 1295 file at the time: 1296 1297 254088 Valgrind should know about UD2 instruction 1298 301280 == 254088 1299 301902 == 254088 1300 304754 NEWS blows TeX's little mind 1301 1302 (3.8.1: 19 September 2012, vex r2537, valgrind r12996) 1303 1304 1305 1306 Release 3.8.0 (10 August 2012) 1307 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1308 3.8.0 is a feature release with many improvements and the usual 1309 collection of bug fixes. 1310 1311 This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux, 1312 PPC64/Linux, S390X/Linux, MIPS/Linux, ARM/Android, X86/Android, 1313 X86/MacOSX 10.6/10.7 and AMD64/MacOSX 10.6/10.7. Support for recent 1314 distros and toolchain components (glibc 2.16, gcc 4.7) has been added. 1315 There is initial support for MacOSX 10.8, but it is not usable for 1316 serious work at present. 1317 1318 * ================== PLATFORM CHANGES ================= 1319 1320 * Support for MIPS32 platforms running Linux. Valgrind has been 1321 tested on MIPS32 and MIPS32r2 platforms running different Debian 1322 Squeeze and MeeGo distributions. Both little-endian and big-endian 1323 cores are supported. The tools Memcheck, Massif and Lackey have 1324 been tested and are known to work. See README.mips for more details. 1325 1326 * Preliminary support for Android running on x86. 1327 1328 * Preliminary (as-yet largely unusable) support for MacOSX 10.8. 1329 1330 * Support for Intel AVX instructions and for AES instructions. This 1331 support is available only for 64 bit code. 1332 1333 * Support for POWER Decimal Floating Point instructions. 1334 1335 * ==================== TOOL CHANGES ==================== 1336 1337 * Non-libc malloc implementations are now supported. This is useful 1338 for tools that replace malloc (Memcheck, Massif, DRD, Helgrind). 1339 Using the new option --soname-synonyms, such tools can be informed 1340 that the malloc implementation is either linked statically into the 1341 executable, or is present in some other shared library different 1342 from libc.so. This makes it possible to process statically linked 1343 programs, and programs using other malloc libraries, for example 1344 TCMalloc or JEMalloc. 1345 1346 * For tools that provide their own replacement for malloc et al, the 1347 option --redzone-size=<number> allows users to specify the size of 1348 the padding blocks (redzones) added before and after each client 1349 allocated block. Smaller redzones decrease the memory needed by 1350 Valgrind. Bigger redzones increase the chance to detect blocks 1351 overrun or underrun. Prior to this change, the redzone size was 1352 hardwired to 16 bytes in Memcheck. 1353 1354 * Memcheck: 1355 1356 - The leak_check GDB server monitor command now can 1357 control the maximum nr of loss records to output. 1358 1359 - Reduction of memory use for applications allocating 1360 many blocks and/or having many partially defined bytes. 1361 1362 - Addition of GDB server monitor command 'block_list' that lists 1363 the addresses/sizes of the blocks of a leak search loss record. 1364 1365 - Addition of GDB server monitor command 'who_points_at' that lists 1366 the locations pointing at a block. 1367 1368 - If a redzone size > 0 is given, VALGRIND_MALLOCLIKE_BLOCK now will 1369 detect an invalid access of these redzones, by marking them 1370 noaccess. Similarly, if a redzone size is given for a memory 1371 pool, VALGRIND_MEMPOOL_ALLOC will mark the redzones no access. 1372 This still allows to find some bugs if the user has forgotten to 1373 mark the pool superblock noaccess. 1374 1375 - Performance of memory leak check has been improved, especially in 1376 cases where there are many leaked blocks and/or many suppression 1377 rules used to suppress leak reports. 1378 1379 - Reduced noise (false positive) level on MacOSX 10.6/10.7, due to 1380 more precise analysis, which is important for LLVM/Clang 1381 generated code. This is at the cost of somewhat reduced 1382 performance. Note there is no change to analysis precision or 1383 costs on Linux targets. 1384 1385 * DRD: 1386 1387 - Added even more facilities that can help finding the cause of a data 1388 race, namely the command-line option --ptrace-addr and the macro 1389 DRD_STOP_TRACING_VAR(x). More information can be found in the manual. 1390 1391 - Fixed a subtle bug that could cause false positive data race reports. 1392 1393 * ==================== OTHER CHANGES ==================== 1394 1395 * The C++ demangler has been updated so as to work well with C++ 1396 compiled by up to at least g++ 4.6. 1397 1398 * Tool developers can make replacement/wrapping more flexible thanks 1399 to the new option --soname-synonyms. This was reported above, but 1400 in fact is very general and applies to all function 1401 replacement/wrapping, not just to malloc-family functions. 1402 1403 * Round-robin scheduling of threads can be selected, using the new 1404 option --fair-sched= yes. Prior to this change, the pipe-based 1405 thread serialisation mechanism (which is still the default) could 1406 give very unfair scheduling. --fair-sched=yes improves 1407 responsiveness of interactive multithreaded applications, and 1408 improves repeatability of results from the thread checkers Helgrind 1409 and DRD. 1410 1411 * For tool developers: support to run Valgrind on Valgrind has been 1412 improved. We can now routinely Valgrind on Helgrind or Memcheck. 1413 1414 * gdbserver now shows the float shadow registers as integer 1415 rather than float values, as the shadow values are mostly 1416 used as bit patterns. 1417 1418 * Increased limit for the --num-callers command line flag to 500. 1419 1420 * Performance improvements for error matching when there are many 1421 suppression records in use. 1422 1423 * Improved support for DWARF4 debugging information (bug 284184). 1424 1425 * Initial support for DWZ compressed Dwarf debug info. 1426 1427 * Improved control over the IR optimiser's handling of the tradeoff 1428 between performance and precision of exceptions. Specifically, 1429 --vex-iropt-precise-memory-exns has been removed and replaced by 1430 --vex-iropt-register-updates, with extended functionality. This 1431 allows the Valgrind gdbserver to always show up to date register 1432 values to GDB. 1433 1434 * Modest performance gains through the use of translation chaining for 1435 JIT-generated code. 1436 1437 * ==================== FIXED BUGS ==================== 1438 1439 The following bugs have been fixed or resolved. Note that "n-i-bz" 1440 stands for "not in bugzilla" -- that is, a bug that was reported to us 1441 but never got a bugzilla entry. We encourage you to file bugs in 1442 bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather 1443 than mailing the developers (or mailing lists) directly -- bugs that 1444 are not entered into bugzilla tend to get forgotten about or ignored. 1445 1446 To see details of a given bug, visit 1447 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 1448 where XXXXXX is the bug number as listed below. 1449 1450 197914 Building valgrind from svn now requires automake-1.10 1451 203877 increase to 16Mb maximum allowed alignment for memalign et al 1452 219156 Handle statically linked malloc or other malloc lib (e.g. tcmalloc) 1453 247386 make perf does not run all performance tests 1454 270006 Valgrind scheduler unfair 1455 270777 Adding MIPS/Linux port to Valgrind 1456 270796 s390x: Removed broken support for the TS insn 1457 271438 Fix configure for proper SSE4.2 detection 1458 273114 s390x: Support TR, TRE, TROO, TROT, TRTO, and TRTT instructions 1459 273475 Add support for AVX instructions 1460 274078 improved configure logic for mpicc 1461 276993 fix mremap 'no thrash checks' 1462 278313 Fedora 15/x64: err read debug info with --read-var-info=yes flag 1463 281482 memcheck incorrect byte allocation count in realloc() for silly argument 1464 282230 group allocator for small fixed size, use it for MC_Chunk/SEc vbit 1465 283413 Fix wrong sanity check 1466 283671 Robustize alignment computation in LibVEX_Alloc 1467 283961 Adding support for some HCI IOCTLs 1468 284124 parse_type_DIE: confused by: DWARF 4 1469 284864 == 273475 (Add support for AVX instructions) 1470 285219 Too-restrictive constraints for Thumb2 "SP plus/minus register" 1471 285662 (MacOSX): Memcheck needs to replace memcpy/memmove 1472 285725 == 273475 (Add support for AVX instructions) 1473 286261 add wrapper for linux I2C_RDWR ioctl 1474 286270 vgpreload is not friendly to 64->32 bit execs, gives ld.so warnings 1475 286374 Running cachegrind with --branch-sim=yes on 64-bit PowerPC program fails 1476 286384 configure fails "checking for a supported version of gcc" 1477 286497 == 273475 (Add support for AVX instructions) 1478 286596 == 273475 (Add support for AVX instructions) 1479 286917 disInstr(arm): unhandled instruction: QADD (also QSUB) 1480 287175 ARM: scalar VFP fixed-point VCVT instructions not handled 1481 287260 Incorrect conditional jump or move depends on uninitialised value(s) 1482 287301 vex amd64->IR: 0x66 0xF 0x38 0x41 0xC0 0xB8 0x0 0x0 (PHMINPOSUW) 1483 287307 == 273475 (Add support for AVX instructions) 1484 287858 VG_(strerror): unknown error 1485 288298 (MacOSX) unhandled syscall shm_unlink 1486 288995 == 273475 (Add support for AVX instructions) 1487 289470 Loading of large Mach-O thin binaries fails. 1488 289656 == 273475 (Add support for AVX instructions) 1489 289699 vgdb connection in relay mode erroneously closed due to buffer overrun 1490 289823 == 293754 (PCMPxSTRx not implemented for 16-bit characters) 1491 289839 s390x: Provide support for unicode conversion instructions 1492 289939 monitor cmd 'leak_check' with details about leaked or reachable blocks 1493 290006 memcheck doesn't mark %xmm as initialized after "pcmpeqw %xmm %xmm" 1494 290655 Add support for AESKEYGENASSIST instruction 1495 290719 valgrind-3.7.0 fails with automake-1.11.2 due to"pkglibdir" usage 1496 290974 vgdb must align pages to VKI_SHMLBA (16KB) on ARM 1497 291253 ES register not initialised in valgrind simulation 1498 291568 Fix 3DNOW-related crashes with baseline x86_64 CPU (w patch) 1499 291865 s390x: Support the "Compare Double and Swap" family of instructions 1500 292300 == 273475 (Add support for AVX instructions) 1501 292430 unrecognized instruction in __intel_get_new_mem_ops_cpuid 1502 292493 == 273475 (Add support for AVX instructions) 1503 292626 Missing fcntl F_SETOWN_EX and F_GETOWN_EX support 1504 292627 Missing support for some SCSI ioctls 1505 292628 none/tests/x86/bug125959-x86.c triggers undefined behavior 1506 292841 == 273475 (Add support for AVX instructions) 1507 292993 implement the getcpu syscall on amd64-linux 1508 292995 Implement the cross memory attach syscalls introduced in Linux 3.2 1509 293088 Add some VEX sanity checks for ppc64 unhandled instructions 1510 293751 == 290655 (Add support for AESKEYGENASSIST instruction) 1511 293754 PCMPxSTRx not implemented for 16-bit characters 1512 293755 == 293754 (No tests for PCMPxSTRx on 16-bit characters) 1513 293808 CLFLUSH not supported by latest VEX for amd64 1514 294047 valgrind does not correctly emulate prlimit64(..., RLIMIT_NOFILE, ...) 1515 294048 MPSADBW instruction not implemented 1516 294055 regtest none/tests/shell fails when locale is not set to C 1517 294185 INT 0x44 (and others) not supported on x86 guest, but used by Jikes RVM 1518 294190 --vgdb-error=xxx can be out of sync with errors shown to the user 1519 294191 amd64: fnsave/frstor and 0x66 size prefixes on FP instructions 1520 294260 disInstr_AMD64: disInstr miscalculated next %rip 1521 294523 --partial-loads-ok=yes causes false negatives 1522 294617 vex amd64->IR: 0x66 0xF 0x3A 0xDF 0xD1 0x1 0xE8 0x6A 1523 294736 vex amd64->IR: 0x48 0xF 0xD7 0xD6 0x48 0x83 1524 294812 patch allowing to run (on x86 at least) helgrind/drd on tool. 1525 295089 can not annotate source for both helgrind and drd 1526 295221 POWER Processor decimal floating point instruction support missing 1527 295427 building for i386 with clang on darwin11 requires "-new_linker linker" 1528 295428 coregrind/m_main.c has incorrect x86 assembly for darwin 1529 295590 Helgrind: Assertion 'cvi->nWaiters > 0' failed 1530 295617 ARM - Add some missing syscalls 1531 295799 Missing \n with get_vbits in gdbserver when line is % 80 [...] 1532 296229 Linux user input device ioctls missing wrappers 1533 296318 ELF Debug info improvements (more than one rx/rw mapping) 1534 296422 Add translation chaining support 1535 296457 vex amd64->IR: 0x66 0xF 0x3A 0xDF 0xD1 0x1 0xE8 0x6A (dup of AES) 1536 296792 valgrind 3.7.0: add SIOCSHWTSTAMP (0x89B0) ioctl wrapper 1537 296983 Fix build issues on x86_64/ppc64 without 32-bit toolchains 1538 297078 gdbserver signal handling problems [..] 1539 297147 drd false positives on newly allocated memory 1540 297329 disallow decoding of IBM Power DFP insns on some machines 1541 297497 POWER Processor decimal floating point instruction support missing 1542 297701 Another alias for strncasecmp_l in libc-2.13.so 1543 297911 'invalid write' not reported when using APIs for custom mem allocators. 1544 297976 s390x: revisit EX implementation 1545 297991 Valgrind interferes with mmap()+ftell() 1546 297992 Support systems missing WIFCONTINUED (e.g. pre-2.6.10 Linux) 1547 297993 Fix compilation of valgrind with gcc -g3. 1548 298080 POWER Processor DFP support missing, part 3 1549 298227 == 273475 (Add support for AVX instructions) 1550 298335 == 273475 (Add support for AVX instructions) 1551 298354 Unhandled ARM Thumb instruction 0xEB0D 0x0585 (streq) 1552 298394 s390x: Don't bail out on an unknown machine model. [..] 1553 298421 accept4() syscall (366) support is missing for ARM 1554 298718 vex amd64->IR: 0xF 0xB1 0xCB 0x9C 0x8F 0x45 1555 298732 valgrind installation problem in ubuntu with kernel version 3.x 1556 298862 POWER Processor DFP instruction support missing, part 4 1557 298864 DWARF reader mis-parses DW_FORM_ref_addr 1558 298943 massif asserts with --pages-as-heap=yes when brk is changing [..] 1559 299053 Support DWARF4 DW_AT_high_pc constant form 1560 299104 == 273475 (Add support for AVX instructions) 1561 299316 Helgrind: hg_main.c:628 (map_threads_lookup): Assertion 'thr' failed. 1562 299629 dup3() syscall (358) support is missing for ARM 1563 299694 POWER Processor DFP instruction support missing, part 5 1564 299756 Ignore --free-fill for MEMPOOL_FREE and FREELIKE client requests 1565 299803 == 273475 (Add support for AVX instructions) 1566 299804 == 273475 (Add support for AVX instructions) 1567 299805 == 273475 (Add support for AVX instructions) 1568 300140 ARM - Missing (T1) SMMUL 1569 300195 == 296318 (ELF Debug info improvements (more than one rx/rw mapping)) 1570 300389 Assertion `are_valid_hwcaps(VexArchAMD64, [..])' failed. 1571 300414 FCOM and FCOMP unimplemented for amd64 guest 1572 301204 infinite loop in canonicaliseSymtab with ifunc symbol 1573 301229 == 203877 (increase to 16Mb maximum allowed alignment for memalign etc) 1574 301265 add x86 support to Android build 1575 301984 configure script doesn't detect certain versions of clang 1576 302205 Fix compiler warnings for POWER VEX code and POWER test cases 1577 302287 Unhandled movbe instruction on Atom processors 1578 302370 PPC: fnmadd, fnmsub, fnmadds, fnmsubs insns always negate the result 1579 302536 Fix for the POWER Valgrind regression test: memcheck-ISA2.0. 1580 302578 Unrecognized isntruction 0xc5 0x32 0xc2 0xca 0x09 vcmpngess 1581 302656 == 273475 (Add support for AVX instructions) 1582 302709 valgrind for ARM needs extra tls support for android emulator [..] 1583 302827 add wrapper for CDROM_GET_CAPABILITY 1584 302901 Valgrind crashes with dwz optimized debuginfo 1585 302918 Enable testing of the vmaddfp and vnsubfp instructions in the testsuite 1586 303116 Add support for the POWER instruction popcntb 1587 303127 Power test suite fixes for frsqrte, vrefp, and vrsqrtefp instructions. 1588 303250 Assertion `instrs_in->arr_used <= 10000' failed w/ OpenSSL code 1589 303466 == 273475 (Add support for AVX instructions) 1590 303624 segmentation fault on Android 4.1 (e.g. on Galaxy Nexus OMAP) 1591 303963 strstr() function produces wrong results under valgrind callgrind 1592 304054 CALL_FN_xx macros need to enforce stack alignment 1593 304561 tee system call not supported 1594 715750 (MacOSX): Incorrect invalid-address errors near 0xFFFFxxxx (mozbug#) 1595 n-i-bz Add missing gdbserver xml files for shadow registers for ppc32 1596 n-i-bz Bypass gcc4.4/4.5 code gen bugs causing out of memory or asserts 1597 n-i-bz Fix assert in gdbserver for watchpoints watching the same address 1598 n-i-bz Fix false positive in sys_clone on amd64 when optional args [..] 1599 n-i-bz s390x: Shadow registers can now be examined using vgdb 1600 1601 (3.8.0-TEST3: 9 August 2012, vex r2465, valgrind r12865) 1602 (3.8.0: 10 August 2012, vex r2465, valgrind r12866) 1603 1604 1605 1606 Release 3.7.0 (5 November 2011) 1607 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1608 3.7.0 is a feature release with many significant improvements and the 1609 usual collection of bug fixes. 1610 1611 This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux, 1612 PPC64/Linux, S390X/Linux, ARM/Android, X86/Darwin and AMD64/Darwin. 1613 Support for recent distros and toolchain components (glibc 2.14, gcc 1614 4.6, MacOSX 10.7) has been added. 1615 1616 * ================== PLATFORM CHANGES ================= 1617 1618 * Support for IBM z/Architecture (s390x) running Linux. Valgrind can 1619 analyse 64-bit programs running on z/Architecture. Most user space 1620 instructions up to and including z10 are supported. Valgrind has 1621 been tested extensively on z9, z10, and z196 machines running SLES 1622 10/11, RedHat 5/6m, and Fedora. The Memcheck and Massif tools are 1623 known to work well. Callgrind, Helgrind, and DRD work reasonably 1624 well on z9 and later models. See README.s390 for more details. 1625 1626 * Preliminary support for MacOSX 10.7 and XCode 4. Both 32- and 1627 64-bit processes are supported. Some complex threaded applications 1628 (Firefox) are observed to hang when run as 32 bit applications, 1629 whereas 64-bit versions run OK. The cause is unknown. Memcheck 1630 will likely report some false errors. In general, expect some rough 1631 spots. This release also supports MacOSX 10.6, but drops support 1632 for 10.5. 1633 1634 * Preliminary support for Android (on ARM). Valgrind can now run 1635 large applications (eg, Firefox) on (eg) a Samsung Nexus S. See 1636 README.android for more details, plus instructions on how to get 1637 started. 1638 1639 * Support for the IBM Power ISA 2.06 (Power7 instructions) 1640 1641 * General correctness and performance improvements for ARM/Linux, and, 1642 by extension, ARM/Android. 1643 1644 * Further solidification of support for SSE 4.2 in 64-bit mode. AVX 1645 instruction set support is under development but is not available in 1646 this release. 1647 1648 * Support for AIX5 has been removed. 1649 1650 * ==================== TOOL CHANGES ==================== 1651 1652 * Memcheck: some incremental changes: 1653 1654 - reduction of memory use in some circumstances 1655 1656 - improved handling of freed memory, which in some circumstances 1657 can cause detection of use-after-free that would previously have 1658 been missed 1659 1660 - fix of a longstanding bug that could cause false negatives (missed 1661 errors) in programs doing vector saturated narrowing instructions. 1662 1663 * Helgrind: performance improvements and major memory use reductions, 1664 particularly for large, long running applications which perform many 1665 synchronisation (lock, unlock, etc) events. Plus many smaller 1666 changes: 1667 1668 - display of locksets for both threads involved in a race 1669 1670 - general improvements in formatting/clarity of error messages 1671 1672 - addition of facilities and documentation regarding annotation 1673 of thread safe reference counted C++ classes 1674 1675 - new flag --check-stack-refs=no|yes [yes], to disable race checking 1676 on thread stacks (a performance hack) 1677 1678 - new flag --free-is-write=no|yes [no], to enable detection of races 1679 where one thread accesses heap memory but another one frees it, 1680 without any coordinating synchronisation event 1681 1682 * DRD: enabled XML output; added support for delayed thread deletion 1683 in order to detect races that occur close to the end of a thread 1684 (--join-list-vol); fixed a memory leak triggered by repeated client 1685 memory allocatation and deallocation; improved Darwin support. 1686 1687 * exp-ptrcheck: this tool has been renamed to exp-sgcheck 1688 1689 * exp-sgcheck: this tool has been reduced in scope so as to improve 1690 performance and remove checking that Memcheck does better. 1691 Specifically, the ability to check for overruns for stack and global 1692 arrays is unchanged, but the ability to check for overruns of heap 1693 blocks has been removed. The tool has accordingly been renamed to 1694 exp-sgcheck ("Stack and Global Array Checking"). 1695 1696 * ==================== OTHER CHANGES ==================== 1697 1698 * GDB server: Valgrind now has an embedded GDB server. That means it 1699 is possible to control a Valgrind run from GDB, doing all the usual 1700 things that GDB can do (single stepping, breakpoints, examining 1701 data, etc). Tool-specific functionality is also available. For 1702 example, it is possible to query the definedness state of variables 1703 or memory from within GDB when running Memcheck; arbitrarily large 1704 memory watchpoints are supported, etc. To use the GDB server, start 1705 Valgrind with the flag --vgdb-error=0 and follow the on-screen 1706 instructions. 1707 1708 * Improved support for unfriendly self-modifying code: a new option 1709 --smc-check=all-non-file is available. This adds the relevant 1710 consistency checks only to code that originates in non-file-backed 1711 mappings. In effect this confines the consistency checking only to 1712 code that is or might be JIT generated, and avoids checks on code 1713 that must have been compiled ahead of time. This significantly 1714 improves performance on applications that generate code at run time. 1715 1716 * It is now possible to build a working Valgrind using Clang-2.9 on 1717 Linux. 1718 1719 * new client requests VALGRIND_{DISABLE,ENABLE}_ERROR_REPORTING. 1720 These enable and disable error reporting on a per-thread, and 1721 nestable, basis. This is useful for hiding errors in particularly 1722 troublesome pieces of code. The MPI wrapper library (libmpiwrap.c) 1723 now uses this facility. 1724 1725 * Added the --mod-funcname option to cg_diff. 1726 1727 * ==================== FIXED BUGS ==================== 1728 1729 The following bugs have been fixed or resolved. Note that "n-i-bz" 1730 stands for "not in bugzilla" -- that is, a bug that was reported to us 1731 but never got a bugzilla entry. We encourage you to file bugs in 1732 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than 1733 mailing the developers (or mailing lists) directly -- bugs that are 1734 not entered into bugzilla tend to get forgotten about or ignored. 1735 1736 To see details of a given bug, visit 1737 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 1738 where XXXXXX is the bug number as listed below. 1739 1740 79311 malloc silly arg warning does not give stack trace 1741 210935 port valgrind.h (not valgrind) to win32 to support client requests 1742 214223 valgrind SIGSEGV on startup gcc 4.4.1 ppc32 (G4) Ubuntu 9.10 1743 243404 Port to zSeries 1744 243935 Helgrind: incorrect handling of ANNOTATE_HAPPENS_BEFORE()/AFTER() 1745 247223 non-x86: Suppress warning: 'regparm' attribute directive ignored 1746 250101 huge "free" memory usage due to m_mallocfree.c fragmentation 1747 253206 Some fixes for the faultstatus testcase 1748 255223 capget testcase fails when running as root 1749 256703 xlc_dbl_u32.c testcase broken 1750 256726 Helgrind tests have broken inline asm 1751 259977 == 214223 (Valgrind segfaults doing __builtin_longjmp) 1752 264800 testcase compile failure on zseries 1753 265762 make public VEX headers compilable by G++ 3.x 1754 265771 assertion in jumps.c (r11523) fails with glibc-2.3 1755 266753 configure script does not give the user the option to not use QtCore 1756 266931 gen_insn_test.pl is broken 1757 266961 ld-linux.so.2 i?86-linux strlen issues 1758 266990 setns instruction causes false positive 1759 267020 Make directory for temporary files configurable at run-time. 1760 267342 == 267997 (segmentation fault on Mac OS 10.6) 1761 267383 Assertion 'vgPlain_strlen(dir) + vgPlain_strlen(file) + 1 < 256' failed 1762 267413 Assertion 'DRD_(g_threadinfo)[tid].synchr_nesting >= 1' failed. 1763 267488 regtest: darwin support for 64-bit build 1764 267552 SIGSEGV (misaligned_stack_error) with DRD, but not with other tools 1765 267630 Add support for IBM Power ISA 2.06 -- stage 1 1766 267769 == 267997 (Darwin: memcheck triggers segmentation fault) 1767 267819 Add client request for informing the core about reallocation 1768 267925 laog data structure quadratic for a single sequence of lock 1769 267968 drd: (vgDrd_thread_set_joinable): Assertion '0 <= (int)tid ..' failed 1770 267997 MacOSX: 64-bit V segfaults on launch when built with Xcode 4.0.1 1771 268513 missed optimizations in fold_Expr 1772 268619 s390x: fpr - gpr transfer facility 1773 268620 s390x: reconsider "long displacement" requirement 1774 268621 s390x: improve IR generation for XC 1775 268715 s390x: FLOGR is not universally available 1776 268792 == 267997 (valgrind seg faults on startup when compiled with Xcode 4) 1777 268930 s390x: MHY is not universally available 1778 269078 arm->IR: unhandled instruction SUB (SP minus immediate/register) 1779 269079 Support ptrace system call on ARM 1780 269144 missing "Bad option" error message 1781 269209 conditional load and store facility (z196) 1782 269354 Shift by zero on x86 can incorrectly clobber CC_NDEP 1783 269641 == 267997 (valgrind segfaults immediately (segmentation fault)) 1784 269736 s390x: minor code generation tweaks 1785 269778 == 272986 (valgrind.h: swap roles of VALGRIND_DO_CLIENT_REQUEST() ..) 1786 269863 s390x: remove unused function parameters 1787 269864 s390x: tweak s390_emit_load_cc 1788 269884 == 250101 (overhead for huge blocks exhausts space too soon) 1789 270082 s390x: Make sure to point the PSW address to the next address on SIGILL 1790 270115 s390x: rewrite some testcases 1791 270309 == 267997 (valgrind crash on startup) 1792 270320 add support for Linux FIOQSIZE ioctl() call 1793 270326 segfault while trying to sanitize the environment passed to execle 1794 270794 IBM POWER7 support patch causes regression in none/tests 1795 270851 IBM POWER7 fcfidus instruction causes memcheck to fail 1796 270856 IBM POWER7 xsnmaddadp instruction causes memcheck to fail on 32bit app 1797 270925 hyper-optimized strspn() in /lib64/libc-2.13.so needs fix 1798 270959 s390x: invalid use of R0 as base register 1799 271042 VSX configure check fails when it should not 1800 271043 Valgrind build fails with assembler error on ppc64 with binutils 2.21 1801 271259 s390x: fix code confusion 1802 271337 == 267997 (Valgrind segfaults on MacOS X) 1803 271385 s390x: Implement Ist_MBE 1804 271501 s390x: misc cleanups 1805 271504 s390x: promote likely and unlikely 1806 271579 ppc: using wrong enum type 1807 271615 unhandled instruction "popcnt" (arch=amd10h) 1808 271730 Fix bug when checking ioctls: duplicate check 1809 271776 s390x: provide STFLE instruction support 1810 271779 s390x: provide clock instructions like STCK 1811 271799 Darwin: ioctls without an arg report a memory error 1812 271820 arm: fix type confusion 1813 271917 pthread_cond_timedwait failure leads to not-locked false positive 1814 272067 s390x: fix DISP20 macro 1815 272615 A typo in debug output in mc_leakcheck.c 1816 272661 callgrind_annotate chokes when run from paths containing regex chars 1817 272893 amd64->IR: 0x66 0xF 0x38 0x2B 0xC1 0x66 0xF 0x7F == (closed as dup) 1818 272955 Unhandled syscall error for pwrite64 on ppc64 arch 1819 272967 make documentation build-system more robust 1820 272986 Fix gcc-4.6 warnings with valgrind.h 1821 273318 amd64->IR: 0x66 0xF 0x3A 0x61 0xC1 0x38 (missing PCMPxSTRx case) 1822 273318 unhandled PCMPxSTRx case: vex amd64->IR: 0x66 0xF 0x3A 0x61 0xC1 0x38 1823 273431 valgrind segfaults in evalCfiExpr (debuginfo.c:2039) 1824 273465 Callgrind: jumps.c:164 (new_jcc): Assertion '(0 <= jmp) && ...' 1825 273536 Build error: multiple definition of `vgDrd_pthread_cond_initializer' 1826 273640 ppc64-linux: unhandled syscalls setresuid(164) and setresgid(169) 1827 273729 == 283000 (Illegal opcode for SSE2 "roundsd" instruction) 1828 273778 exp-ptrcheck: unhandled sysno == 259 1829 274089 exp-ptrcheck: unhandled sysno == 208 1830 274378 s390x: Various dispatcher tweaks 1831 274447 WARNING: unhandled syscall: 340 1832 274776 amd64->IR: 0x66 0xF 0x38 0x2B 0xC5 0x66 1833 274784 == 267997 (valgrind ls -l results in Segmentation Fault) 1834 274926 valgrind does not build against linux-3 1835 275148 configure FAIL with glibc-2.14 1836 275151 Fedora 15 / glibc-2.14 'make regtest' FAIL 1837 275168 Make Valgrind work for MacOSX 10.7 Lion 1838 275212 == 275284 (lots of false positives from __memcpy_ssse3_back et al) 1839 275278 valgrind does not build on Linux kernel 3.0.* due to silly 1840 275284 Valgrind memcpy/memmove redirection stopped working in glibc 2.14/x86_64 1841 275308 Fix implementation for ppc64 fres instruc 1842 275339 s390x: fix testcase compile warnings 1843 275517 s390x: Provide support for CKSM instruction 1844 275710 s390x: get rid of redundant address mode calculation 1845 275815 == 247894 (Valgrind doesn't know about Linux readahead(2) syscall) 1846 275852 == 250101 (valgrind uses all swap space and is killed) 1847 276784 Add support for IBM Power ISA 2.06 -- stage 3 1848 276987 gdbsrv: fix tests following recent commits 1849 277045 Valgrind crashes with unhandled DW_OP_ opcode 0x2a 1850 277199 The test_isa_2_06_part1.c in none/tests/ppc64 should be a symlink 1851 277471 Unhandled syscall: 340 1852 277610 valgrind crashes in VG_(lseek)(core_fd, phdrs[idx].p_offset, ...) 1853 277653 ARM: support Thumb2 PLD instruction 1854 277663 ARM: NEON float VMUL by scalar incorrect 1855 277689 ARM: tests for VSTn with register post-index are broken 1856 277694 ARM: BLX LR instruction broken in ARM mode 1857 277780 ARM: VMOV.F32 (immediate) instruction is broken 1858 278057 fuse filesystem syscall deadlocks 1859 278078 Unimplemented syscall 280 on ppc32 1860 278349 F_GETPIPE_SZ and F_SETPIPE_SZ Linux fcntl commands 1861 278454 VALGRIND_STACK_DEREGISTER has wrong output type 1862 278502 == 275284 (Valgrind confuses memcpy() and memmove()) 1863 278892 gdbsrv: factorize gdb version handling, fix doc and typos 1864 279027 Support for MVCL and CLCL instruction 1865 279027 s390x: Provide support for CLCL and MVCL instructions 1866 279062 Remove a redundant check in the insn selector for ppc. 1867 279071 JDK creates PTEST with redundant REX.W prefix 1868 279212 gdbsrv: add monitor cmd v.info scheduler. 1869 279378 exp-ptrcheck: the 'impossible' happened on mkfifo call 1870 279698 memcheck discards valid-bits for packuswb 1871 279795 memcheck reports uninitialised values for mincore on amd64 1872 279994 Add support for IBM Power ISA 2.06 -- stage 3 1873 280083 mempolicy syscall check errors 1874 280290 vex amd64->IR: 0x66 0xF 0x38 0x28 0xC1 0x66 0xF 0x6F 1875 280710 s390x: config files for nightly builds 1876 280757 /tmp dir still used by valgrind even if TMPDIR is specified 1877 280965 Valgrind breaks fcntl locks when program does mmap 1878 281138 WARNING: unhandled syscall: 340 1879 281241 == 275168 (valgrind useless on Macos 10.7.1 Lion) 1880 281304 == 275168 (Darwin: dyld "cannot load inserted library") 1881 281305 == 275168 (unhandled syscall: unix:357 on Darwin 11.1) 1882 281468 s390x: handle do_clone and gcc clones in call traces 1883 281488 ARM: VFP register corruption 1884 281828 == 275284 (false memmove warning: "Source and destination overlap") 1885 281883 s390x: Fix system call wrapper for "clone". 1886 282105 generalise 'reclaimSuperBlock' to also reclaim splittable superblock 1887 282112 Unhandled instruction bytes: 0xDE 0xD9 0x9B 0xDF (fcompp) 1888 282238 SLES10: make check fails 1889 282979 strcasestr needs replacement with recent(>=2.12) glibc 1890 283000 vex amd64->IR: 0x66 0xF 0x3A 0xA 0xC0 0x9 0xF3 0xF 1891 283243 Regression in ppc64 memcheck tests 1892 283325 == 267997 (Darwin: V segfaults on startup when built with Xcode 4.0) 1893 283427 re-connect epoll_pwait syscall on ARM linux 1894 283600 gdbsrv: android: port vgdb.c 1895 283709 none/tests/faultstatus needs to account for page size 1896 284305 filter_gdb needs enhancement to work on ppc64 1897 284384 clang 3.1 -Wunused-value warnings in valgrind.h, memcheck.h 1898 284472 Thumb2 ROR.W encoding T2 not implemented 1899 284621 XML-escape process command line in XML output 1900 n-i-bz cachegrind/callgrind: handle CPUID information for Core iX Intel CPUs 1901 that have non-power-of-2 sizes (also AMDs) 1902 n-i-bz don't be spooked by libraries mashed by elfhack 1903 n-i-bz don't be spooked by libxul.so linked with gold 1904 n-i-bz improved checking for VALGRIND_CHECK_MEM_IS_DEFINED 1905 1906 (3.7.0-TEST1: 27 October 2011, vex r2228, valgrind r12245) 1907 (3.7.0.RC1: 1 November 2011, vex r2231, valgrind r12257) 1908 (3.7.0: 5 November 2011, vex r2231, valgrind r12258) 1909 1910 1911 1912 Release 3.6.1 (16 February 2011) 1913 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1914 3.6.1 is a bug fix release. It adds support for some SSE4 1915 instructions that were omitted in 3.6.0 due to lack of time. Initial 1916 support for glibc-2.13 has been added. A number of bugs causing 1917 crashing or assertion failures have been fixed. 1918 1919 The following bugs have been fixed or resolved. Note that "n-i-bz" 1920 stands for "not in bugzilla" -- that is, a bug that was reported to us 1921 but never got a bugzilla entry. We encourage you to file bugs in 1922 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than 1923 mailing the developers (or mailing lists) directly -- bugs that are 1924 not entered into bugzilla tend to get forgotten about or ignored. 1925 1926 To see details of a given bug, visit 1927 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 1928 where XXXXXX is the bug number as listed below. 1929 1930 188572 Valgrind on Mac should suppress setenv() mem leak 1931 194402 vex amd64->IR: 0x48 0xF 0xAE 0x4 (proper FX{SAVE,RSTOR} support) 1932 210481 vex amd64->IR: Assertion `sz == 2 || sz == 4' failed (REX.W POPQ) 1933 246152 callgrind internal error after pthread_cancel on 32 Bit Linux 1934 250038 ppc64: Altivec LVSR and LVSL instructions fail their regtest 1935 254420 memory pool tracking broken 1936 254957 Test code failing to compile due to changes in memcheck.h 1937 255009 helgrind/drd: crash on chmod with invalid parameter 1938 255130 readdwarf3.c parse_type_DIE confused by GNAT Ada types 1939 255355 helgrind/drd: crash on threaded programs doing fork 1940 255358 == 255355 1941 255418 (SSE4.x) rint call compiled with ICC 1942 255822 --gen-suppressions can create invalid files: "too many callers [...]" 1943 255888 closing valgrindoutput tag outputted to log-stream on error 1944 255963 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x9 0xDB 0x0 (ROUNDPD) 1945 255966 Slowness when using mempool annotations 1946 256387 vex x86->IR: 0xD4 0xA 0x2 0x7 (AAD and AAM) 1947 256600 super-optimized strcasecmp() false positive 1948 256669 vex amd64->IR: Unhandled LOOPNEL insn on amd64 1949 256968 (SSE4.x) vex amd64->IR: 0x66 0xF 0x38 0x10 0xD3 0x66 (BLENDVPx) 1950 257011 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0xE 0xFD 0xA0 (PBLENDW) 1951 257063 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x8 0xC0 0x0 (ROUNDPS) 1952 257276 Missing case in memcheck --track-origins=yes 1953 258870 (SSE4.x) Add support for EXTRACTPS SSE 4.1 instruction 1954 261966 (SSE4.x) support for CRC32B and CRC32Q is lacking (also CRC32{W,L}) 1955 262985 VEX regression in valgrind 3.6.0 in handling PowerPC VMX 1956 262995 (SSE4.x) crash when trying to valgrind gcc-snapshot (PCMPxSTRx $0) 1957 263099 callgrind_annotate counts Ir improperly [...] 1958 263877 undefined coprocessor instruction on ARMv7 1959 265964 configure FAIL with glibc-2.13 1960 n-i-bz Fix compile error w/ icc-12.x in guest_arm_toIR.c 1961 n-i-bz Docs: fix bogus descriptions for VALGRIND_CREATE_BLOCK et al 1962 n-i-bz Massif: don't assert on shmat() with --pages-as-heap=yes 1963 n-i-bz Bug fixes and major speedups for the exp-DHAT space profiler 1964 n-i-bz DRD: disable --free-is-write due to implementation difficulties 1965 1966 (3.6.1: 16 February 2011, vex r2103, valgrind r11561). 1967 1968 1969 1970 Release 3.6.0 (21 October 2010) 1971 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1972 3.6.0 is a feature release with many significant improvements and the 1973 usual collection of bug fixes. 1974 1975 This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux, 1976 PPC64/Linux, X86/Darwin and AMD64/Darwin. Support for recent distros 1977 and toolchain components (glibc 2.12, gcc 4.5, OSX 10.6) has been added. 1978 1979 ------------------------- 1980 1981 Here are some highlights. Details are shown further down: 1982 1983 * Support for ARM/Linux. 1984 1985 * Support for recent Linux distros: Ubuntu 10.10 and Fedora 14. 1986 1987 * Support for Mac OS X 10.6, both 32- and 64-bit executables. 1988 1989 * Support for the SSE4.2 instruction set. 1990 1991 * Enhancements to the Callgrind profiler, including the ability to 1992 handle CPUs with three levels of cache. 1993 1994 * A new experimental heap profiler, DHAT. 1995 1996 * A huge number of bug fixes and small enhancements. 1997 1998 ------------------------- 1999 2000 Here are details of the above changes, together with descriptions of 2001 many other changes, and a list of fixed bugs. 2002 2003 * ================== PLATFORM CHANGES ================= 2004 2005 * Support for ARM/Linux. Valgrind now runs on ARMv7 capable CPUs 2006 running Linux. It is known to work on Ubuntu 10.04, Ubuntu 10.10, 2007 and Maemo 5, so you can run Valgrind on your Nokia N900 if you want. 2008 2009 This requires a CPU capable of running the ARMv7-A instruction set 2010 (Cortex A5, A8 and A9). Valgrind provides fairly complete coverage 2011 of the user space instruction set, including ARM and Thumb integer 2012 code, VFPv3, NEON and V6 media instructions. The Memcheck, 2013 Cachegrind and Massif tools work properly; other tools work to 2014 varying degrees. 2015 2016 * Support for recent Linux distros (Ubuntu 10.10 and Fedora 14), along 2017 with support for recent releases of the underlying toolchain 2018 components, notably gcc-4.5 and glibc-2.12. 2019 2020 * Support for Mac OS X 10.6, both 32- and 64-bit executables. 64-bit 2021 support also works much better on OS X 10.5, and is as solid as 2022 32-bit support now. 2023 2024 * Support for the SSE4.2 instruction set. SSE4.2 is supported in 2025 64-bit mode. In 32-bit mode, support is only available up to and 2026 including SSSE3. Some exceptions: SSE4.2 AES instructions are not 2027 supported in 64-bit mode, and 32-bit mode does in fact support the 2028 bare minimum SSE4 instructions to needed to run programs on Mac OS X 2029 10.6 on 32-bit targets. 2030 2031 * Support for IBM POWER6 cpus has been improved. The Power ISA up to 2032 and including version 2.05 is supported. 2033 2034 * ==================== TOOL CHANGES ==================== 2035 2036 * Cachegrind has a new processing script, cg_diff, which finds the 2037 difference between two profiles. It's very useful for evaluating 2038 the performance effects of a change in a program. 2039 2040 Related to this change, the meaning of cg_annotate's (rarely-used) 2041 --threshold option has changed; this is unlikely to affect many 2042 people, if you do use it please see the user manual for details. 2043 2044 * Callgrind now can do branch prediction simulation, similar to 2045 Cachegrind. In addition, it optionally can count the number of 2046 executed global bus events. Both can be used for a better 2047 approximation of a "Cycle Estimation" as derived event (you need to 2048 update the event formula in KCachegrind yourself). 2049 2050 * Cachegrind and Callgrind now refer to the LL (last-level) cache 2051 rather than the L2 cache. This is to accommodate machines with 2052 three levels of caches -- if Cachegrind/Callgrind auto-detects the 2053 cache configuration of such a machine it will run the simulation as 2054 if the L2 cache isn't present. This means the results are less 2055 likely to match the true result for the machine, but 2056 Cachegrind/Callgrind's results are already only approximate, and 2057 should not be considered authoritative. The results are still 2058 useful for giving a general idea about a program's locality. 2059 2060 * Massif has a new option, --pages-as-heap, which is disabled by 2061 default. When enabled, instead of tracking allocations at the level 2062 of heap blocks (as allocated with malloc/new/new[]), it instead 2063 tracks memory allocations at the level of memory pages (as mapped by 2064 mmap, brk, etc). Each mapped page is treated as its own block. 2065 Interpreting the page-level output is harder than the heap-level 2066 output, but this option is useful if you want to account for every 2067 byte of memory used by a program. 2068 2069 * DRD has two new command-line options: --free-is-write and 2070 --trace-alloc. The former allows to detect reading from already freed 2071 memory, and the latter allows tracing of all memory allocations and 2072 deallocations. 2073 2074 * DRD has several new annotations. Custom barrier implementations can 2075 now be annotated, as well as benign races on static variables. 2076 2077 * DRD's happens before / happens after annotations have been made more 2078 powerful, so that they can now also be used to annotate e.g. a smart 2079 pointer implementation. 2080 2081 * Helgrind's annotation set has also been drastically improved, so as 2082 to provide to users a general set of annotations to describe locks, 2083 semaphores, barriers and condition variables. Annotations to 2084 describe thread-safe reference counted heap objects have also been 2085 added. 2086 2087 * Memcheck has a new command-line option, --show-possibly-lost, which 2088 is enabled by default. When disabled, the leak detector will not 2089 show possibly-lost blocks. 2090 2091 * A new experimental heap profiler, DHAT (Dynamic Heap Analysis Tool), 2092 has been added. DHAT keeps track of allocated heap blocks, and also 2093 inspects every memory reference to see which block (if any) is being 2094 accessed. This gives a lot of insight into block lifetimes, 2095 utilisation, turnover, liveness, and the location of hot and cold 2096 fields. You can use DHAT to do hot-field profiling. 2097 2098 * ==================== OTHER CHANGES ==================== 2099 2100 * Improved support for unfriendly self-modifying code: the extra 2101 overhead incurred by --smc-check=all has been reduced by 2102 approximately a factor of 5 as compared with 3.5.0. 2103 2104 * Ability to show directory names for source files in error messages. 2105 This is combined with a flexible mechanism for specifying which 2106 parts of the paths should be shown. This is enabled by the new flag 2107 --fullpath-after. 2108 2109 * A new flag, --require-text-symbol, which will stop the run if a 2110 specified symbol is not found it a given shared object when it is 2111 loaded into the process. This makes advanced working with function 2112 intercepting and wrapping safer and more reliable. 2113 2114 * Improved support for the Valkyrie GUI, version 2.0.0. GUI output 2115 and control of Valgrind is now available for the tools Memcheck and 2116 Helgrind. XML output from Valgrind is available for Memcheck, 2117 Helgrind and exp-Ptrcheck. 2118 2119 * More reliable stack unwinding on amd64-linux, particularly in the 2120 presence of function wrappers, and with gcc-4.5 compiled code. 2121 2122 * Modest scalability (performance improvements) for massive 2123 long-running applications, particularly for those with huge amounts 2124 of code. 2125 2126 * Support for analyzing programs running under Wine with has been 2127 improved. The header files <valgrind/valgrind.h>, 2128 <valgrind/memcheck.h> and <valgrind/drd.h> can now be used in 2129 Windows-programs compiled with MinGW or one of the Microsoft Visual 2130 Studio compilers. 2131 2132 * A rare but serious error in the 64-bit x86 CPU simulation was fixed. 2133 The 32-bit simulator was not affected. This did not occur often, 2134 but when it did would usually crash the program under test. 2135 Bug 245925. 2136 2137 * A large number of bugs were fixed. These are shown below. 2138 2139 * A number of bugs were investigated, and were candidates for fixing, 2140 but are not fixed in 3.6.0, due to lack of developer time. They may 2141 get fixed in later releases. They are: 2142 2143 194402 vex amd64->IR: 0x48 0xF 0xAE 0x4 0x24 0x49 (FXSAVE64) 2144 212419 false positive "lock order violated" (A+B vs A) 2145 213685 Undefined value propagates past dependency breaking instruction 2146 216837 Incorrect instrumentation of NSOperationQueue on Darwin 2147 237920 valgrind segfault on fork failure 2148 242137 support for code compiled by LLVM-2.8 2149 242423 Another unknown Intel cache config value 2150 243232 Inconsistent Lock Orderings report with trylock 2151 243483 ppc: callgrind triggers VEX assertion failure 2152 243935 Helgrind: implementation of ANNOTATE_HAPPENS_BEFORE() is wrong 2153 244677 Helgrind crash hg_main.c:616 (map_threads_lookup): Assertion 2154 'thr' failed. 2155 246152 callgrind internal error after pthread_cancel on 32 Bit Linux 2156 249435 Analyzing wine programs with callgrind triggers a crash 2157 250038 ppc64: Altivec lvsr and lvsl instructions fail their regtest 2158 250065 Handling large allocations 2159 250101 huge "free" memory usage due to m_mallocfree.c 2160 "superblocks fragmentation" 2161 251569 vex amd64->IR: 0xF 0x1 0xF9 0x8B 0x4C 0x24 (RDTSCP) 2162 252091 Callgrind on ARM does not detect function returns correctly 2163 252600 [PATCH] Allow lhs to be a pointer for shl/shr 2164 254420 memory pool tracking broken 2165 n-i-bz support for adding symbols for JIT generated code 2166 2167 2168 The following bugs have been fixed or resolved. Note that "n-i-bz" 2169 stands for "not in bugzilla" -- that is, a bug that was reported to us 2170 but never got a bugzilla entry. We encourage you to file bugs in 2171 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than 2172 mailing the developers (or mailing lists) directly -- bugs that are 2173 not entered into bugzilla tend to get forgotten about or ignored. 2174 2175 To see details of a given bug, visit 2176 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 2177 where XXXXXX is the bug number as listed below. 2178 2179 135264 dcbzl instruction missing 2180 142688 == 250799 2181 153699 Valgrind should report unaligned reads with movdqa 2182 180217 == 212335 2183 190429 Valgrind reports lost of errors in ld.so 2184 with x86_64 2.9.90 glibc 2185 197266 valgrind appears to choke on the xmms instruction 2186 "roundsd" on x86_64 2187 197988 Crash when demangling very large symbol names 2188 202315 unhandled syscall: 332 (inotify_init1) 2189 203256 Add page-level profiling to Massif 2190 205093 dsymutil=yes needs quotes, locking (partial fix) 2191 205241 Snow Leopard 10.6 support (partial fix) 2192 206600 Leak checker fails to upgrade indirect blocks when their 2193 parent becomes reachable 2194 210935 port valgrind.h (not valgrind) to win32 so apps run under 2195 wine can make client requests 2196 211410 vex amd64->IR: 0x15 0xFF 0xFF 0x0 0x0 0x89 2197 within Linux ip-stack checksum functions 2198 212335 unhandled instruction bytes: 0xF3 0xF 0xBD 0xC0 2199 (lzcnt %eax,%eax) 2200 213685 Undefined value propagates past dependency breaking instruction 2201 (partial fix) 2202 215914 Valgrind inserts bogus empty environment variable 2203 217863 == 197988 2204 219538 adjtimex syscall wrapper wrong in readonly adjtime mode 2205 222545 shmat fails under valgind on some arm targets 2206 222560 ARM NEON support 2207 230407 == 202315 2208 231076 == 202315 2209 232509 Docs build fails with formatting inside <title></title> elements 2210 232793 == 202315 2211 235642 [PATCH] syswrap-linux.c: support evdev EVIOCG* ioctls 2212 236546 vex x86->IR: 0x66 0xF 0x3A 0xA 2213 237202 vex amd64->IR: 0xF3 0xF 0xB8 0xC0 0x49 0x3B 2214 237371 better support for VALGRIND_MALLOCLIKE_BLOCK 2215 237485 symlink (syscall 57) is not supported on Mac OS 2216 237723 sysno == 101 exp-ptrcheck: the 'impossible' happened: 2217 unhandled syscall 2218 238208 is_just_below_ESP doesn't take into account red-zone 2219 238345 valgrind passes wrong $0 when executing a shell script 2220 238679 mq_timedreceive syscall doesn't flag the reception buffer 2221 as "defined" 2222 238696 fcntl command F_DUPFD_CLOEXEC not supported 2223 238713 unhandled instruction bytes: 0x66 0xF 0x29 0xC6 2224 238713 unhandled instruction bytes: 0x66 0xF 0x29 0xC6 2225 238745 3.5.0 Make fails on PPC Altivec opcodes, though configure 2226 says "Altivec off" 2227 239992 vex amd64->IR: 0x48 0xF 0xC4 0xC1 0x0 0x48 2228 240488 == 197988 2229 240639 == 212335 2230 241377 == 236546 2231 241903 == 202315 2232 241920 == 212335 2233 242606 unhandled syscall: setegid (in Ptrcheck) 2234 242814 Helgrind "Impossible has happened" during 2235 QApplication::initInstance(); 2236 243064 Valgrind attempting to read debug information from iso 2237 243270 Make stack unwinding in Valgrind wrappers more reliable 2238 243884 exp-ptrcheck: the 'impossible happened: unhandled syscall 2239 sysno = 277 (mq_open) 2240 244009 exp-ptrcheck unknown syscalls in analyzing lighttpd 2241 244493 ARM VFP d16-d31 registers support 2242 244670 add support for audit_session_self syscall on Mac OS 10.6 2243 244921 The xml report of helgrind tool is not well format 2244 244923 In the xml report file, the <preamble> not escape the 2245 xml char, eg '<','&','>' 2246 245535 print full path names in plain text reports 2247 245925 x86-64 red zone handling problem 2248 246258 Valgrind not catching integer underruns + new [] s 2249 246311 reg/reg cmpxchg doesn't work on amd64 2250 246549 unhandled syscall unix:277 while testing 32-bit Darwin app 2251 246888 Improve Makefile.vex.am 2252 247510 [OS X 10.6] Memcheck reports unaddressable bytes passed 2253 to [f]chmod_extended 2254 247526 IBM POWER6 (ISA 2.05) support is incomplete 2255 247561 Some leak testcases fails due to reachable addresses in 2256 caller save regs 2257 247875 sizeofIRType to handle Ity_I128 2258 247894 [PATCH] unhandled syscall sys_readahead 2259 247980 Doesn't honor CFLAGS passed to configure 2260 248373 darwin10.supp is empty in the trunk 2261 248822 Linux FIBMAP ioctl has int parameter instead of long 2262 248893 [PATCH] make readdwarf.c big endianess safe to enable 2263 unwinding on big endian systems 2264 249224 Syscall 336 not supported (SYS_proc_info) 2265 249359 == 245535 2266 249775 Incorrect scheme for detecting NEON capabilities of host CPU 2267 249943 jni JVM init fails when using valgrind 2268 249991 Valgrind incorrectly declares AESKEYGENASSIST support 2269 since VEX r2011 2270 249996 linux/arm: unhandled syscall: 181 (__NR_pwrite64) 2271 250799 frexp$fenv_access_off function generates SIGILL 2272 250998 vex x86->IR: unhandled instruction bytes: 0x66 0x66 0x66 0x2E 2273 251251 support pclmulqdq insn 2274 251362 valgrind: ARM: attach to debugger either fails or provokes 2275 kernel oops 2276 251674 Unhandled syscall 294 2277 251818 == 254550 2278 2279 254257 Add support for debugfiles found by build-id 2280 254550 [PATCH] Implement DW_ATE_UTF (DWARF4) 2281 254646 Wrapped functions cause stack misalignment on OS X 2282 (and possibly Linux) 2283 254556 ARM: valgrinding anything fails with SIGSEGV for 0xFFFF0FA0 2284 2285 (3.6.0: 21 October 2010, vex r2068, valgrind r11471). 2286 2287 2288 2289 Release 3.5.0 (19 August 2009) 2290 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2291 3.5.0 is a feature release with many significant improvements and the 2292 usual collection of bug fixes. The main improvement is that Valgrind 2293 now works on Mac OS X. 2294 2295 This release supports X86/Linux, AMD64/Linux, PPC32/Linux, PPC64/Linux 2296 and X86/Darwin. Support for recent distros and toolchain components 2297 (glibc 2.10, gcc 4.5) has been added. 2298 2299 ------------------------- 2300 2301 Here is a short summary of the changes. Details are shown further 2302 down: 2303 2304 * Support for Mac OS X (10.5.x). 2305 2306 * Improvements and simplifications to Memcheck's leak checker. 2307 2308 * Clarification and simplifications in various aspects of Valgrind's 2309 text output. 2310 2311 * XML output for Helgrind and Ptrcheck. 2312 2313 * Performance and stability improvements for Helgrind and DRD. 2314 2315 * Genuinely atomic support for x86/amd64/ppc atomic instructions. 2316 2317 * A new experimental tool, BBV, useful for computer architecture 2318 research. 2319 2320 * Improved Wine support, including ability to read Windows PDB 2321 debuginfo. 2322 2323 ------------------------- 2324 2325 Here are details of the above changes, followed by descriptions of 2326 many other minor changes, and a list of fixed bugs. 2327 2328 2329 * Valgrind now runs on Mac OS X. (Note that Mac OS X is sometimes 2330 called "Darwin" because that is the name of the OS core, which is the 2331 level that Valgrind works at.) 2332 2333 Supported systems: 2334 2335 - It requires OS 10.5.x (Leopard). Porting to 10.4.x is not planned 2336 because it would require work and 10.4 is only becoming less common. 2337 2338 - 32-bit programs on x86 and AMD64 (a.k.a x86-64) machines are supported 2339 fairly well. For 10.5.x, 32-bit programs are the default even on 2340 64-bit machines, so it handles most current programs. 2341 2342 - 64-bit programs on x86 and AMD64 (a.k.a x86-64) machines are not 2343 officially supported, but simple programs at least will probably work. 2344 However, start-up is slow. 2345 2346 - PowerPC machines are not supported. 2347 2348 Things that don't work: 2349 2350 - The Ptrcheck tool. 2351 2352 - Objective-C garbage collection. 2353 2354 - --db-attach=yes. 2355 2356 - If you have Rogue Amoeba's "Instant Hijack" program installed, 2357 Valgrind will fail with a SIGTRAP at start-up. See 2358 https://bugs.kde.org/show_bug.cgi?id=193917 for details and a 2359 simple work-around. 2360 2361 Usage notes: 2362 2363 - You will likely find --dsymutil=yes a useful option, as error 2364 messages may be imprecise without it. 2365 2366 - Mac OS X support is new and therefore will be less robust than the 2367 Linux support. Please report any bugs you find. 2368 2369 - Threaded programs may run more slowly than on Linux. 2370 2371 Many thanks to Greg Parker for developing this port over several years. 2372 2373 2374 * Memcheck's leak checker has been improved. 2375 2376 - The results for --leak-check=summary now match the summary results 2377 for --leak-check=full. Previously they could differ because 2378 --leak-check=summary counted "indirectly lost" blocks and 2379 "suppressed" blocks as "definitely lost". 2380 2381 - Blocks that are only reachable via at least one interior-pointer, 2382 but are directly pointed to by a start-pointer, were previously 2383 marked as "still reachable". They are now correctly marked as 2384 "possibly lost". 2385 2386 - The default value for the --leak-resolution option has been 2387 changed from "low" to "high". In general, this means that more 2388 leak reports will be produced, but each leak report will describe 2389 fewer leaked blocks. 2390 2391 - With --leak-check=full, "definitely lost" and "possibly lost" 2392 leaks are now considered as proper errors, ie. they are counted 2393 for the "ERROR SUMMARY" and affect the behaviour of 2394 --error-exitcode. These leaks are not counted as errors if 2395 --leak-check=summary is specified, however. 2396 2397 - Documentation for the leak checker has been improved. 2398 2399 2400 * Various aspects of Valgrind's text output have changed. 2401 2402 - Valgrind's start-up message has changed. It is shorter but also 2403 includes the command being run, which makes it easier to use 2404 --trace-children=yes. An example: 2405 2406 - Valgrind's shut-down messages have also changed. This is most 2407 noticeable with Memcheck, where the leak summary now occurs before 2408 the error summary. This change was necessary to allow leaks to be 2409 counted as proper errors (see the description of the leak checker 2410 changes above for more details). This was also necessary to fix a 2411 longstanding bug in which uses of suppressions against leaks were 2412 not "counted", leading to difficulties in maintaining suppression 2413 files (see https://bugs.kde.org/show_bug.cgi?id=186790). 2414 2415 - Behavior of -v has changed. In previous versions, -v printed out 2416 a mixture of marginally-user-useful information, and tool/core 2417 statistics. The statistics printing has now been moved to its own 2418 flag, --stats=yes. This means -v is less verbose and more likely 2419 to convey useful end-user information. 2420 2421 - The format of some (non-XML) stack trace entries has changed a 2422 little. Previously there were six possible forms: 2423 2424 0x80483BF: really (a.c:20) 2425 0x80483BF: really (in /foo/a.out) 2426 0x80483BF: really 2427 0x80483BF: (within /foo/a.out) 2428 0x80483BF: ??? (a.c:20) 2429 0x80483BF: ??? 2430 2431 The third and fourth of these forms have been made more consistent 2432 with the others. The six possible forms are now: 2433 2434 0x80483BF: really (a.c:20) 2435 0x80483BF: really (in /foo/a.out) 2436 0x80483BF: really (in ???) 2437 0x80483BF: ??? (in /foo/a.out) 2438 0x80483BF: ??? (a.c:20) 2439 0x80483BF: ??? 2440 2441 Stack traces produced when --xml=yes is specified are different 2442 and unchanged. 2443 2444 2445 * Helgrind and Ptrcheck now support XML output, so they can be used 2446 from GUI tools. Also, the XML output mechanism has been 2447 overhauled. 2448 2449 - The XML format has been overhauled and generalised, so it is more 2450 suitable for error reporting tools in general. The Memcheck 2451 specific aspects of it have been removed. The new format, which 2452 is an evolution of the old format, is described in 2453 docs/internals/xml-output-protocol4.txt. 2454 2455 - Memcheck has been updated to use the new format. 2456 2457 - Helgrind and Ptrcheck are now able to emit output in this format. 2458 2459 - The XML output mechanism has been overhauled. XML is now output 2460 to its own file descriptor, which means that: 2461 2462 * Valgrind can output text and XML independently. 2463 2464 * The longstanding problem of XML output being corrupted by 2465 unexpected un-tagged text messages is solved. 2466 2467 As before, the destination for text output is specified using 2468 --log-file=, --log-fd= or --log-socket=. 2469 2470 As before, XML output for a tool is enabled using --xml=yes. 2471 2472 Because there's a new XML output channel, the XML output 2473 destination is now specified by --xml-file=, --xml-fd= or 2474 --xml-socket=. 2475 2476 Initial feedback has shown this causes some confusion. To 2477 clarify, the two envisaged usage scenarios are: 2478 2479 (1) Normal text output. In this case, do not specify --xml=yes 2480 nor any of --xml-file=, --xml-fd= or --xml-socket=. 2481 2482 (2) XML output. In this case, specify --xml=yes, and one of 2483 --xml-file=, --xml-fd= or --xml-socket= to select the XML 2484 destination, one of --log-file=, --log-fd= or --log-socket= 2485 to select the destination for any remaining text messages, 2486 and, importantly, -q. 2487 2488 -q makes Valgrind completely silent on the text channel, 2489 except in the case of critical failures, such as Valgrind 2490 itself segfaulting, or failing to read debugging information. 2491 Hence, in this scenario, it suffices to check whether or not 2492 any output appeared on the text channel. If yes, then it is 2493 likely to be a critical error which should be brought to the 2494 attention of the user. If no (the text channel produced no 2495 output) then it can be assumed that the run was successful. 2496 2497 This allows GUIs to make the critical distinction they need to 2498 make (did the run fail or not?) without having to search or 2499 filter the text output channel in any way. 2500 2501 It is also recommended to use --child-silent-after-fork=yes in 2502 scenario (2). 2503 2504 2505 * Improvements and changes in Helgrind: 2506 2507 - XML output, as described above 2508 2509 - Checks for consistent association between pthread condition 2510 variables and their associated mutexes are now performed. 2511 2512 - pthread_spinlock functions are supported. 2513 2514 - Modest performance improvements. 2515 2516 - Initial (skeletal) support for describing the behaviour of 2517 non-POSIX synchronisation objects through ThreadSanitizer 2518 compatible ANNOTATE_* macros. 2519 2520 - More controllable tradeoffs between performance and the level of 2521 detail of "previous" accesses in a race. There are now three 2522 settings: 2523 2524 * --history-level=full. This is the default, and was also the 2525 default in 3.4.x. It shows both stacks involved in a race, but 2526 requires a lot of memory and can be very slow in programs that 2527 do many inter-thread synchronisation events. 2528 2529 * --history-level=none. This only shows the later stack involved 2530 in a race. This can be much faster than --history-level=full, 2531 but makes it much more difficult to find the other access 2532 involved in the race. 2533 2534 The new intermediate setting is 2535 2536 * --history-level=approx 2537 2538 For the earlier (other) access, two stacks are presented. The 2539 earlier access is guaranteed to be somewhere in between the two 2540 program points denoted by those stacks. This is not as useful 2541 as showing the exact stack for the previous access (as per 2542 --history-level=full), but it is better than nothing, and it's 2543 almost as fast as --history-level=none. 2544 2545 2546 * New features and improvements in DRD: 2547 2548 - The error messages printed by DRD are now easier to interpret. 2549 Instead of using two different numbers to identify each thread 2550 (Valgrind thread ID and DRD thread ID), DRD does now identify 2551 threads via a single number (the DRD thread ID). Furthermore 2552 "first observed at" information is now printed for all error 2553 messages related to synchronization objects. 2554 2555 - Added support for named semaphores (sem_open() and sem_close()). 2556 2557 - Race conditions between pthread_barrier_wait() and 2558 pthread_barrier_destroy() calls are now reported. 2559 2560 - Added support for custom allocators through the macros 2561 VALGRIND_MALLOCLIKE_BLOCK() VALGRIND_FREELIKE_BLOCK() (defined in 2562 in <valgrind/valgrind.h>). An alternative for these two macros is 2563 the new client request VG_USERREQ__DRD_CLEAN_MEMORY (defined in 2564 <valgrind/drd.h>). 2565 2566 - Added support for annotating non-POSIX synchronization objects 2567 through several new ANNOTATE_*() macros. 2568 2569 - OpenMP: added support for the OpenMP runtime (libgomp) included 2570 with gcc versions 4.3.0 and 4.4.0. 2571 2572 - Faster operation. 2573 2574 - Added two new command-line options (--first-race-only and 2575 --segment-merging-interval). 2576 2577 2578 * Genuinely atomic support for x86/amd64/ppc atomic instructions 2579 2580 Valgrind will now preserve (memory-access) atomicity of LOCK- 2581 prefixed x86/amd64 instructions, and any others implying a global 2582 bus lock. Ditto for PowerPC l{w,d}arx/st{w,d}cx. instructions. 2583 2584 This means that Valgrinded processes will "play nicely" in 2585 situations where communication with other processes, or the kernel, 2586 is done through shared memory and coordinated with such atomic 2587 instructions. Prior to this change, such arrangements usually 2588 resulted in hangs, races or other synchronisation failures, because 2589 Valgrind did not honour atomicity of such instructions. 2590 2591 2592 * A new experimental tool, BBV, has been added. BBV generates basic 2593 block vectors for use with the SimPoint analysis tool, which allows 2594 a program's overall behaviour to be approximated by running only a 2595 fraction of it. This is useful for computer architecture 2596 researchers. You can run BBV by specifying --tool=exp-bbv (the 2597 "exp-" prefix is short for "experimental"). BBV was written by 2598 Vince Weaver. 2599 2600 2601 * Modestly improved support for running Windows applications under 2602 Wine. In particular, initial support for reading Windows .PDB debug 2603 information has been added. 2604 2605 2606 * A new Memcheck client request VALGRIND_COUNT_LEAK_BLOCKS has been 2607 added. It is similar to VALGRIND_COUNT_LEAKS but counts blocks 2608 instead of bytes. 2609 2610 2611 * The Valgrind client requests VALGRIND_PRINTF and 2612 VALGRIND_PRINTF_BACKTRACE have been changed slightly. Previously, 2613 the string was always printed immediately on its own line. Now, the 2614 string will be added to a buffer but not printed until a newline is 2615 encountered, or other Valgrind output is printed (note that for 2616 VALGRIND_PRINTF_BACKTRACE, the back-trace itself is considered 2617 "other Valgrind output"). This allows you to use multiple 2618 VALGRIND_PRINTF calls to build up a single output line, and also to 2619 print multiple output lines with a single request (by embedding 2620 multiple newlines in the string). 2621 2622 2623 * The graphs drawn by Massif's ms_print program have changed slightly: 2624 2625 - The half-height chars '.' and ',' are no longer drawn, because 2626 they are confusing. The --y option can be used if the default 2627 y-resolution is not high enough. 2628 2629 - Horizontal lines are now drawn after the top of a snapshot if 2630 there is a gap until the next snapshot. This makes it clear that 2631 the memory usage has not dropped to zero between snapshots. 2632 2633 2634 * Something that happened in 3.4.0, but wasn't clearly announced: the 2635 option --read-var-info=yes can be used by some tools (Memcheck, 2636 Helgrind and DRD). When enabled, it causes Valgrind to read DWARF3 2637 variable type and location information. This makes those tools 2638 start up more slowly and increases memory consumption, but 2639 descriptions of data addresses in error messages become more 2640 detailed. 2641 2642 2643 * exp-Omega, an experimental instantaneous leak-detecting tool, was 2644 disabled in 3.4.0 due to a lack of interest and maintenance, 2645 although the source code was still in the distribution. The source 2646 code has now been removed from the distribution. For anyone 2647 interested, the removal occurred in SVN revision r10247. 2648 2649 2650 * Some changes have been made to the build system. 2651 2652 - VEX/ is now integrated properly into the build system. This means 2653 that dependency tracking within VEX/ now works properly, "make 2654 install" will work without requiring "make" before it, and 2655 parallel builds (ie. 'make -j') now work (previously a 2656 .NOTPARALLEL directive was used to serialize builds, ie. 'make -j' 2657 was effectively ignored). 2658 2659 - The --with-vex configure option has been removed. It was of 2660 little use and removing it simplified the build system. 2661 2662 - The location of some install files has changed. This should not 2663 affect most users. Those who might be affected: 2664 2665 * For people who use Valgrind with MPI programs, the installed 2666 libmpiwrap.so library has moved from 2667 $(INSTALL)/<platform>/libmpiwrap.so to 2668 $(INSTALL)/libmpiwrap-<platform>.so. 2669 2670 * For people who distribute standalone Valgrind tools, the 2671 installed libraries such as $(INSTALL)/<platform>/libcoregrind.a 2672 have moved to $(INSTALL)/libcoregrind-<platform>.a. 2673 2674 These changes simplify the build system. 2675 2676 - Previously, all the distributed suppression (*.supp) files were 2677 installed. Now, only default.supp is installed. This should not 2678 affect users as the other installed suppression files were not 2679 read; the fact that they were installed was a mistake. 2680 2681 2682 * KNOWN LIMITATIONS: 2683 2684 - Memcheck is unusable with the Intel compiler suite version 11.1, 2685 when it generates code for SSE2-and-above capable targets. This 2686 is because of icc's use of highly optimised inlined strlen 2687 implementations. It causes Memcheck to report huge numbers of 2688 false errors even in simple programs. Helgrind and DRD may also 2689 have problems. 2690 2691 Versions 11.0 and earlier may be OK, but this has not been 2692 properly tested. 2693 2694 2695 The following bugs have been fixed or resolved. Note that "n-i-bz" 2696 stands for "not in bugzilla" -- that is, a bug that was reported to us 2697 but never got a bugzilla entry. We encourage you to file bugs in 2698 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than 2699 mailing the developers (or mailing lists) directly -- bugs that are 2700 not entered into bugzilla tend to get forgotten about or ignored. 2701 2702 To see details of a given bug, visit 2703 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 2704 where XXXXXX is the bug number as listed below. 2705 2706 84303 How about a LockCheck tool? 2707 91633 dereference of null ptr in vgPlain_st_basetype 2708 97452 Valgrind doesn't report any pthreads problems 2709 100628 leak-check gets assertion failure when using 2710 VALGRIND_MALLOCLIKE_BLOCK on malloc()ed memory 2711 108528 NPTL pthread cleanup handlers not called 2712 110126 Valgrind 2.4.1 configure.in tramples CFLAGS 2713 110128 mallinfo is not implemented... 2714 110770 VEX: Generated files not always updated when making valgrind 2715 111102 Memcheck: problems with large (memory footprint) applications 2716 115673 Vex's decoder should never assert 2717 117564 False positive: Syscall param clone(child_tidptr) contains 2718 uninitialised byte(s) 2719 119404 executing ssh from inside valgrind fails 2720 133679 Callgrind does not write path names to sources with dwarf debug 2721 info 2722 135847 configure.in problem with non gnu compilers (and possible fix) 2723 136154 threads.c:273 (vgCallgrind_post_signal): Assertion 2724 '*(vgCallgrind_current_fn_stack.top) == 0' failed. 2725 136230 memcheck reports "possibly lost", should be "still reachable" 2726 137073 NULL arg to MALLOCLIKE_BLOCK causes crash 2727 137904 Valgrind reports a memory leak when using POSIX threads, 2728 while it shouldn't 2729 139076 valgrind VT_GETSTATE error 2730 142228 complaint of elf_dynamic_do_rela in trivial usage 2731 145347 spurious warning with USBDEVFS_REAPURB 2732 148441 (wine) can't find memory leak in Wine, win32 binary 2733 executable file. 2734 148742 Leak-check fails assert on exit 2735 149878 add (proper) check for calloc integer overflow 2736 150606 Call graph is broken when using callgrind control 2737 152393 leak errors produce an exit code of 0. I need some way to 2738 cause leak errors to result in a nonzero exit code. 2739 157154 documentation (leak-resolution doc speaks about num-callers 2740 def=4) + what is a loss record 2741 159501 incorrect handling of ALSA ioctls 2742 162020 Valgrinding an empty/zero-byte file crashes valgrind 2743 162482 ppc: Valgrind crashes while reading stabs information 2744 162718 x86: avoid segment selector 0 in sys_set_thread_area() 2745 163253 (wine) canonicaliseSymtab forgot some fields in DiSym 2746 163560 VEX/test_main.c is missing from valgrind-3.3.1 2747 164353 malloc_usable_size() doesn't return a usable size 2748 165468 Inconsistent formatting in memcheck manual -- please fix 2749 169505 main.c:286 (endOfInstr): 2750 Assertion 'ii->cost_offset == *cost_offset' failed 2751 177206 Generate default.supp during compile instead of configure 2752 177209 Configure valt_load_address based on arch+os 2753 177305 eventfd / syscall 323 patch lost 2754 179731 Tests fail to build because of inlining of non-local asm labels 2755 181394 helgrind: libhb_core.c:3762 (msm_write): Assertion 2756 'ordxx == POrd_EQ || ordxx == POrd_LT' failed. 2757 181594 Bogus warning for empty text segment 2758 181707 dwarf doesn't require enumerations to have name 2759 185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64 2760 185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree): 2761 Assertion '!already_present' failed. 2762 185359 exp-ptrcheck: unhandled syscall getresuid() 2763 185794 "WARNING: unhandled syscall: 285" (fallocate) on x86_64 2764 185816 Valgrind is unable to handle debug info for files with split 2765 debug info that are prelinked afterwards 2766 185980 [darwin] unhandled syscall: sem_open 2767 186238 bbToIR_AMD64: disInstr miscalculated next %rip 2768 186507 exp-ptrcheck unhandled syscalls prctl, etc. 2769 186790 Suppression pattern used for leaks are not reported 2770 186796 Symbols with length>200 in suppression files are ignored 2771 187048 drd: mutex PTHREAD_PROCESS_SHARED attribute missinterpretation 2772 187416 exp-ptrcheck: support for __NR_{setregid,setreuid,setresuid} 2773 188038 helgrind: hg_main.c:926: mk_SHVAL_fail: the 'impossible' happened 2774 188046 bashisms in the configure script 2775 188127 amd64->IR: unhandled instruction bytes: 0xF0 0xF 0xB0 0xA 2776 188161 memcheck: --track-origins=yes asserts "mc_machine.c:672 2777 (get_otrack_shadow_offset_wrk): the 'impossible' happened." 2778 188248 helgrind: pthread_cleanup_push, pthread_rwlock_unlock, 2779 assertion fail "!lock->heldBy" 2780 188427 Add support for epoll_create1 (with patch) 2781 188530 Support for SIOCGSTAMPNS 2782 188560 Include valgrind.spec in the tarball 2783 188572 Valgrind on Mac should suppress setenv() mem leak 2784 189054 Valgrind fails to build because of duplicate non-local asm labels 2785 189737 vex amd64->IR: unhandled instruction bytes: 0xAC 2786 189762 epoll_create syscall not handled (--tool=exp-ptrcheck) 2787 189763 drd assertion failure: s_threadinfo[tid].is_recording 2788 190219 unhandled syscall: 328 (x86-linux) 2789 190391 dup of 181394; see above 2790 190429 Valgrind reports lots of errors in ld.so with x86_64 2.9.90 glibc 2791 190820 No debug information on powerpc-linux 2792 191095 PATCH: Improve usbdevfs ioctl handling 2793 191182 memcheck: VALGRIND_LEAK_CHECK quadratic when big nr of chunks 2794 or big nr of errors 2795 191189 --xml=yes should obey --gen-suppressions=all 2796 191192 syslog() needs a suppression on macosx 2797 191271 DARWIN: WARNING: unhandled syscall: 33554697 a.k.a.: 265 2798 191761 getrlimit on MacOSX 2799 191992 multiple --fn-skip only works sometimes; dependent on order 2800 192634 V. reports "aspacem sync_check_mapping_callback: 2801 segment mismatch" on Darwin 2802 192954 __extension__ missing on 2 client requests 2803 194429 Crash at start-up with glibc-2.10.1 and linux-2.6.29 2804 194474 "INSTALL" file has different build instructions than "README" 2805 194671 Unhandled syscall (sem_wait?) from mac valgrind 2806 195069 memcheck: reports leak (memory still reachable) for 2807 printf("%d', x) 2808 195169 drd: (vgDrd_barrier_post_wait): 2809 Assertion 'r->sg[p->post_iteration]' failed. 2810 195268 valgrind --log-file doesn't accept ~/... 2811 195838 VEX abort: LibVEX_N_SPILL_BYTES too small for CPUID boilerplate 2812 195860 WARNING: unhandled syscall: unix:223 2813 196528 need a error suppression for pthread_rwlock_init under os x? 2814 197227 Support aio_* syscalls on Darwin 2815 197456 valgrind should reject --suppressions=(directory) 2816 197512 DWARF2 CFI reader: unhandled CFI instruction 0:10 2817 197591 unhandled syscall 27 (mincore) 2818 197793 Merge DCAS branch to the trunk == 85756, 142103 2819 197794 Avoid duplicate filenames in Vex 2820 197898 make check fails on current SVN 2821 197901 make check fails also under exp-ptrcheck in current SVN 2822 197929 Make --leak-resolution=high the default 2823 197930 Reduce spacing between leak reports 2824 197933 Print command line of client at start-up, and shorten preamble 2825 197966 unhandled syscall 205 (x86-linux, --tool=exp-ptrcheck) 2826 198395 add BBV to the distribution as an experimental tool 2827 198624 Missing syscalls on Darwin: 82, 167, 281, 347 2828 198649 callgrind_annotate doesn't cumulate counters 2829 199338 callgrind_annotate sorting/thresholds are broken for all but Ir 2830 199977 Valgrind complains about an unrecognized instruction in the 2831 atomic_incs test program 2832 200029 valgrind isn't able to read Fedora 12 debuginfo 2833 200760 darwin unhandled syscall: unix:284 2834 200827 DRD doesn't work on Mac OS X 2835 200990 VG_(read_millisecond_timer)() does not work correctly 2836 201016 Valgrind does not support pthread_kill() on Mac OS 2837 201169 Document --read-var-info 2838 201323 Pre-3.5.0 performance sanity checking 2839 201384 Review user manual for the 3.5.0 release 2840 201585 mfpvr not implemented on ppc 2841 201708 tests failing because x86 direction flag is left set 2842 201757 Valgrind doesn't handle any recent sys_futex additions 2843 204377 64-bit valgrind can not start a shell script 2844 (with #!/path/to/shell) if the shell is a 32-bit executable 2845 n-i-bz drd: fixed assertion failure triggered by mutex reinitialization. 2846 n-i-bz drd: fixed a bug that caused incorrect messages to be printed 2847 about memory allocation events with memory access tracing enabled 2848 n-i-bz drd: fixed a memory leak triggered by vector clock deallocation 2849 2850 (3.5.0: 19 Aug 2009, vex r1913, valgrind r10846). 2851 2852 2853 2854 Release 3.4.1 (28 February 2009) 2855 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2856 3.4.1 is a bug-fix release that fixes some regressions and assertion 2857 failures in debug info reading in 3.4.0, most notably incorrect stack 2858 traces on amd64-linux on older (glibc-2.3 based) systems. Various 2859 other debug info problems are also fixed. A number of bugs in the 2860 exp-ptrcheck tool introduced in 3.4.0 have been fixed. 2861 2862 In view of the fact that 3.4.0 contains user-visible regressions 2863 relative to 3.3.x, upgrading to 3.4.1 is recommended. Packagers are 2864 encouraged to ship 3.4.1 in preference to 3.4.0. 2865 2866 The fixed bugs are as follows. Note that "n-i-bz" stands for "not in 2867 bugzilla" -- that is, a bug that was reported to us but never got a 2868 bugzilla entry. We encourage you to file bugs in bugzilla 2869 (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the 2870 developers (or mailing lists) directly -- bugs that are not entered 2871 into bugzilla tend to get forgotten about or ignored. 2872 2873 n-i-bz Fix various bugs reading icc-11 generated debug info 2874 n-i-bz Fix various bugs reading gcc-4.4 generated debug info 2875 n-i-bz Preliminary support for glibc-2.10 / Fedora 11 2876 n-i-bz Cachegrind and Callgrind: handle non-power-of-two cache sizes, 2877 so as to support (eg) 24k Atom D1 and Core2 with 3/6/12MB L2. 2878 179618 exp-ptrcheck crashed / exit prematurely 2879 179624 helgrind: false positive races with pthread_create and 2880 recv/open/close/read 2881 134207 pkg-config output contains @VG_PLATFORM@ 2882 176926 floating point exception at valgrind startup with PPC 440EPX 2883 181594 Bogus warning for empty text segment 2884 173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant rex prefixes) 2885 181707 Dwarf3 doesn't require enumerations to have name 2886 185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64 2887 185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree): 2888 Assertion '!already_present' failed. 2889 185359 exp-ptrcheck unhandled syscall getresuid() 2890 2891 (3.4.1.RC1: 24 Feb 2008, vex r1884, valgrind r9253). 2892 (3.4.1: 28 Feb 2008, vex r1884, valgrind r9293). 2893 2894 2895 2896 Release 3.4.0 (2 January 2009) 2897 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2898 3.4.0 is a feature release with many significant improvements and the 2899 usual collection of bug fixes. This release supports X86/Linux, 2900 AMD64/Linux, PPC32/Linux and PPC64/Linux. Support for recent distros 2901 (using gcc 4.4, glibc 2.8 and 2.9) has been added. 2902 2903 3.4.0 brings some significant tool improvements. Memcheck can now 2904 report the origin of uninitialised values, the thread checkers 2905 Helgrind and DRD are much improved, and we have a new experimental 2906 tool, exp-Ptrcheck, which is able to detect overruns of stack and 2907 global arrays. In detail: 2908 2909 * Memcheck is now able to track the origin of uninitialised values. 2910 When it reports an uninitialised value error, it will try to show 2911 the origin of the value, as either a heap or stack allocation. 2912 Origin tracking is expensive and so is not enabled by default. To 2913 use it, specify --track-origins=yes. Memcheck's speed will be 2914 essentially halved, and memory usage will be significantly 2915 increased. Nevertheless it can drastically reduce the effort 2916 required to identify the root cause of uninitialised value errors, 2917 and so is often a programmer productivity win, despite running more 2918 slowly. 2919 2920 * A version (1.4.0) of the Valkyrie GUI, that works with Memcheck in 2921 3.4.0, will be released shortly. 2922 2923 * Helgrind's race detection algorithm has been completely redesigned 2924 and reimplemented, to address usability and scalability concerns: 2925 2926 - The new algorithm has a lower false-error rate: it is much less 2927 likely to report races that do not really exist. 2928 2929 - Helgrind will display full call stacks for both accesses involved 2930 in a race. This makes it easier to identify the root causes of 2931 races. 2932 2933 - Limitations on the size of program that can run have been removed. 2934 2935 - Performance has been modestly improved, although that is very 2936 workload-dependent. 2937 2938 - Direct support for Qt4 threading has been added. 2939 2940 - pthread_barriers are now directly supported. 2941 2942 - Helgrind works well on all supported Linux targets. 2943 2944 * The DRD thread debugging tool has seen major improvements: 2945 2946 - Greatly improved performance and significantly reduced memory 2947 usage. 2948 2949 - Support for several major threading libraries (Boost.Thread, Qt4, 2950 glib, OpenMP) has been added. 2951 2952 - Support for atomic instructions, POSIX semaphores, barriers and 2953 reader-writer locks has been added. 2954 2955 - Works now on PowerPC CPUs too. 2956 2957 - Added support for printing thread stack usage at thread exit time. 2958 2959 - Added support for debugging lock contention. 2960 2961 - Added a manual for Drd. 2962 2963 * A new experimental tool, exp-Ptrcheck, has been added. Ptrcheck 2964 checks for misuses of pointers. In that sense it is a bit like 2965 Memcheck. However, Ptrcheck can do things Memcheck can't: it can 2966 detect overruns of stack and global arrays, it can detect 2967 arbitrarily far out-of-bounds accesses to heap blocks, and it can 2968 detect accesses heap blocks that have been freed a very long time 2969 ago (millions of blocks in the past). 2970 2971 Ptrcheck currently works only on x86-linux and amd64-linux. To use 2972 it, use --tool=exp-ptrcheck. A simple manual is provided, as part 2973 of the main Valgrind documentation. As this is an experimental 2974 tool, we would be particularly interested in hearing about your 2975 experiences with it. 2976 2977 * exp-Omega, an experimental instantaneous leak-detecting tool, is no 2978 longer built by default, although the code remains in the repository 2979 and the tarball. This is due to three factors: a perceived lack of 2980 users, a lack of maintenance, and concerns that it may not be 2981 possible to achieve reliable operation using the existing design. 2982 2983 * As usual, support for the latest Linux distros and toolchain 2984 components has been added. It should work well on Fedora Core 10, 2985 OpenSUSE 11.1 and Ubuntu 8.10. gcc-4.4 (in its current pre-release 2986 state) is supported, as is glibc-2.9. The C++ demangler has been 2987 updated so as to work well with C++ compiled by even the most recent 2988 g++'s. 2989 2990 * You can now use frame-level wildcards in suppressions. This was a 2991 frequently-requested enhancement. A line "..." in a suppression now 2992 matches zero or more frames. This makes it easier to write 2993 suppressions which are precise yet insensitive to changes in 2994 inlining behaviour. 2995 2996 * 3.4.0 adds support on x86/amd64 for the SSSE3 instruction set. 2997 2998 * Very basic support for IBM Power6 has been added (64-bit processes only). 2999 3000 * Valgrind is now cross-compilable. For example, it is possible to 3001 cross compile Valgrind on an x86/amd64-linux host, so that it runs 3002 on a ppc32/64-linux target. 3003 3004 * You can set the main thread's stack size at startup using the 3005 new --main-stacksize= flag (subject of course to ulimit settings). 3006 This is useful for running apps that need a lot of stack space. 3007 3008 * The limitation that you can't use --trace-children=yes together 3009 with --db-attach=yes has been removed. 3010 3011 * The following bugs have been fixed. Note that "n-i-bz" stands for 3012 "not in bugzilla" -- that is, a bug that was reported to us but 3013 never got a bugzilla entry. We encourage you to file bugs in 3014 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than 3015 mailing the developers (or mailing lists) directly. 3016 3017 n-i-bz Make return types for some client requests 64-bit clean 3018 n-i-bz glibc 2.9 support 3019 n-i-bz ignore unsafe .valgrindrc's (CVE-2008-4865) 3020 n-i-bz MPI_Init(0,0) is valid but libmpiwrap.c segfaults 3021 n-i-bz Building in an env without gdb gives bogus gdb attach 3022 92456 Tracing the origin of uninitialised memory 3023 106497 Valgrind does not demangle some C++ template symbols 3024 162222 ==106497 3025 151612 Suppression with "..." (frame-level wildcards in .supp files) 3026 156404 Unable to start oocalc under memcheck on openSUSE 10.3 (64-bit) 3027 159285 unhandled syscall:25 (stime, on x86-linux) 3028 159452 unhandled ioctl 0x8B01 on "valgrind iwconfig" 3029 160954 ppc build of valgrind crashes with illegal instruction (isel) 3030 160956 mallinfo implementation, w/ patch 3031 162092 Valgrind fails to start gnome-system-monitor 3032 162819 malloc_free_fill test doesn't pass on glibc2.8 x86 3033 163794 assertion failure with "--track-origins=yes" 3034 163933 sigcontext.err and .trapno must be set together 3035 163955 remove constraint !(--db-attach=yes && --trace-children=yes) 3036 164476 Missing kernel module loading system calls 3037 164669 SVN regression: mmap() drops posix file locks 3038 166581 Callgrind output corruption when program forks 3039 167288 Patch file for missing system calls on Cell BE 3040 168943 unsupported scas instruction pentium 3041 171645 Unrecognised instruction (MOVSD, non-binutils encoding) 3042 172417 x86->IR: 0x82 ... 3043 172563 amd64->IR: 0xD9 0xF5 - fprem1 3044 173099 .lds linker script generation error 3045 173177 [x86_64] syscalls: 125/126/179 (capget/capset/quotactl) 3046 173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant prefixes) 3047 174532 == 173751 3048 174908 --log-file value not expanded correctly for core file 3049 175044 Add lookup_dcookie for amd64 3050 175150 x86->IR: 0xF2 0xF 0x11 0xC1 (movss non-binutils encoding) 3051 3052 Developer-visible changes: 3053 3054 * Valgrind's debug-info reading machinery has been majorly overhauled. 3055 It can now correctly establish the addresses for ELF data symbols, 3056 which is something that has never worked properly before now. 3057 3058 Also, Valgrind can now read DWARF3 type and location information for 3059 stack and global variables. This makes it possible to use the 3060 framework to build tools that rely on knowing the type and locations 3061 of stack and global variables, for example exp-Ptrcheck. 3062 3063 Reading of such information is disabled by default, because most 3064 tools don't need it, and because it is expensive in space and time. 3065 However, you can force Valgrind to read it, using the 3066 --read-var-info=yes flag. Memcheck, Helgrind and DRD are able to 3067 make use of such information, if present, to provide source-level 3068 descriptions of data addresses in the error messages they create. 3069 3070 (3.4.0.RC1: 24 Dec 2008, vex r1878, valgrind r8882). 3071 (3.4.0: 3 Jan 2009, vex r1878, valgrind r8899). 3072
