1 Release 3.10.0 (?? ?????? 201?) 2 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 3 3.10.0 is not yet released. 4 5 * ================== PLATFORM CHANGES ================= 6 7 * Support for Android on MIPS32. 8 * Support for 64bit FPU on MIPS32 platforms. 9 10 * ==================== TOOL CHANGES ==================== 11 12 * Memcheck: 13 - new client requests 14 VALGRIND_DISABLE_ADDR_ERROR_REPORTING_IN_RANGE and 15 VALGRIND_ENABLE_ADDR_ERROR_REPORTING_IN_RANGE 16 17 * Helgrind: 18 - Helgrind GDB server monitor command 'info locks' giving 19 the list of locks, their location, and their status. 20 21 * Callgrind: 22 - callgrind_control now supports the --vgdb-prefix argument, 23 which is needed if valgrind was started with this same argument. 24 25 * ==================== OTHER CHANGES ==================== 26 27 * New and modified GDB server monitor features: 28 29 - The GDB server monitor command 'v.info location <address>' 30 outputs information about an address. The information produced depends 31 on the tool and on the options given to valgrind. 32 Possibly, the following are described: global variables, local (stack) 33 variables, allocated or freed blocks, ... 34 35 - The option "--vgdb-stop-at=event1,event2,..." allows the user 36 to ask GDB server to stop before program execution, at the end 37 of the program execution and on Valgrind internal errors. 38 39 - A new monitor command "v.info stats" that shows various valgrind core and 40 tool statistics. 41 42 - A new monitor command "v.set hostvisibility" that allows GDB server 43 to provide access to Valgrind internal host status/memory. 44 45 * New option --aspace-minaddr=<address> allows to (possibly) make 46 use of some more memory by decreasing the default value above which 47 Valgrind maps memory or solve some conflicts with system libraries 48 by increasing the value. 49 See user manual for details. 50 51 * Minor improvements in dwarf handling with --read-var-info=yes 52 - Ada and C struct containing VLA do not cause a bad DIE error anymore 53 - Code compiled with -ffunction-sections -fdata-sections -Wl,--gc-sections 54 does not cause assert errors anymore. 55 56 * ==================== FIXED BUGS ==================== 57 58 The following bugs have been fixed or resolved. Note that "n-i-bz" 59 stands for "not in bugzilla" -- that is, a bug that was reported to us 60 but never got a bugzilla entry. We encourage you to file bugs in 61 bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather 62 than mailing the developers (or mailing lists) directly -- bugs that 63 are not entered into bugzilla tend to get forgotten about or ignored. 64 65 To see details of a given bug, visit 66 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 67 where XXXXXX is the bug number as listed below. 68 69 175819 Support for ipv6 socket reporting with --track-fds 70 232510 make distcheck fails 71 308729 vex x86->IR: unhandled instruction bytes 0xf 0x5 (syscall) 72 325110 Add test-cases for Power ISA 2.06 insns: divdo/divdo. and divduo/divduo. 73 325124 [MIPSEL] Compilation error 74 325477 Phase 4 support for IBM Power ISA 2.07 75 325628 Phase 5 support for IBM Power ISA 2.07 76 325714 Empty vgcore but RLIMIT_CORE is big enough (too big) 77 325751 Missing the two privileged Power PC Transactional Memory Instructions 78 325816 Phase 6 support for IBM Power ISA 2.07 79 325856 Make SGCheck fail gracefully on unsupported platforms 80 326436 DRD: False positive in libstdc++ std::list::push_back 81 326444 Cavium MIPS Octeon Specific Load Indexed Instructions 82 326462 Refactor vgdb to isolate invoker stuff into separate module 83 326469 amd64->IR: 0x66 0xF 0x3A 0x63 0xC1 0xE (pcmpistri 0x0E) 84 326623 DRD: false positive conflict report in a field assignment 85 326816 Intercept for __strncpy_sse2_unaligned missing? 86 326921 coregrind fails to compile m_trampoline.S with MIPS/Linux port of V 87 326983 Clear direction flag after tests on amd64. 88 327212 Do not prepend the current directory to absolute path names. 89 327238 Callgrind Assertion 'passed <= last_bb->cjmp_count' failed 90 327284 s390x: Fix translation of the risbg instruction 91 327639 vex amd64->IR pcmpestri SSE4.2 instruction is unsupported 0x34 92 327837 dwz compressed alternate .debug_info and .debug_str not read correctly 93 327916 DW_TAG_typedef may have no name 94 328100 XABORT not implemented 95 328205 Implement additional Xen hypercalls 96 328455 s390x: SIGILL after emitting wrong register pair for ldxbr 97 328711 valgrind.1 manpage "memcheck options" section is badly generated 98 328878 vex amd64->IR pcmpestri SSE4.2 instruction is unsupported 0x14 99 329612 Incorrect handling of AT_BASE for image execution 100 329956 valgrind crashes when lmw/stmw instructions are used on ppc64 101 330228 mmap must align to VKI_SHMLBA on mips32 102 330459 --track-fds=yes doesn't track eventfds 103 330469 Add clock_adjtime syscall support 104 330622 Add test to regression suite for POWER instruction: dcbzl 105 330939 Support for AMD's syscall instruction on x86 106 == 308729 107 330941 Typo in PRE(poll) syscall wrapper 108 331057 unhandled instruction: 0xEEE01B20 (vfma.f64) (has patch) 109 331254 Fix expected output for memcheck/tests/dw4 110 331255 Fix race condition in test none/tests/coolo_sigaction 111 331257 Fix type of jump buffer in test none/tests/faultstatus 112 331305 configure uses bash specific syntax 113 331337 s390x WARNING: unhandled syscall: 326 (dup3) 114 331380 Syscall param timer_create(evp) points to uninitialised byte(s) 115 331830 ppc64: WARNING: unhandled syscall: 96/97 116 331839 drd/tests/sem_open specifies invalid semaphore name 117 331847 outcome of drd/tests/thread_name is nondeterministic 118 332037 Valgrind cannot handle Thumb "add pc, reg" 119 332055 drd asserts on platforms with VG_STACK_REDZONE_SZB == 0 and 120 consistency checks enabled 121 332263 intercepts for pthread_rwlock_timedrdlock and 122 pthread_rwlock_timedwrlock are incorrect 123 332265 drd could do with post-rwlock_init and pre-rwlock_destroy 124 client requests 125 332276 Implement additional Xen hypercalls 126 332658 ldrd.w r1, r2, [PC, #imm] does not adjust for 32bit alignment 127 333072 drd: Add semaphore annotations 128 333145 Tests for missaligned PC+#imm access for arm 129 333228 AAarch64 Missing instruction encoding: mrs %[reg], ctr_el0 130 333230 AAarch64 missing instruction encodings: dc, ic, dsb. 131 333428 ldr.w pc [rD, #imm] instruction leads to assertion 132 333666 Recognize MPX instructions and bnd prefix. 133 334049 lzcnt fails silently (x86_32) 134 334705 sendmsg and recvmsg should guard against bogus msghdr fields. 135 334727 Build fails with -Werror=format-security 136 334788 clarify doc about --log-file initial program directory 137 335155 vgdb, fix error print statement. 138 n-i-bz Fix KVM_CREATE_IRQCHIP ioctl handling 139 n-i-bz s390x: Fix memory corruption for multithreaded applications 140 n-i-bz vex arm->IR: allow PC as basereg in some LDRD cases 141 n-i-bz internal error in Valgrind if vgdb transmit signals when ptrace invoked 142 n-i-bz Fix mingw64 support in valgrind.h (dev@, 9 May 2014) 143 144 145 Release 3.9.0 (31 October 2013) 146 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 147 3.9.0 is a feature release with many improvements and the usual 148 collection of bug fixes. 149 150 This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux, 151 PPC64/Linux, S390X/Linux, MIPS32/Linux, MIPS64/Linux, ARM/Android, 152 X86/Android, X86/MacOSX 10.7 and AMD64/MacOSX 10.7. Support for 153 MacOSX 10.8 is significantly improved relative to the 3.8.0 release. 154 155 * ================== PLATFORM CHANGES ================= 156 157 * Support for MIPS64 LE and BE running Linux. Valgrind has been 158 tested on MIPS64 Debian Squeeze and Debian Wheezy distributions. 159 160 * Support for MIPS DSP ASE on MIPS32 platforms. 161 162 * Support for s390x Decimal Floating Point instructions on hosts that 163 have the DFP facility installed. 164 165 * Support for POWER8 (Power ISA 2.07) instructions 166 167 * Support for Intel AVX2 instructions. This is available only on 64 168 bit code. 169 170 * Initial support for Intel Transactional Synchronization Extensions, 171 both RTM and HLE. 172 173 * Initial support for Hardware Transactional Memory on POWER. 174 175 * Improved support for MacOSX 10.8 (64-bit only). Memcheck can now 176 run large GUI apps tolerably well. 177 178 * ==================== TOOL CHANGES ==================== 179 180 * Memcheck: 181 182 - Improvements in handling of vectorised code, leading to 183 significantly fewer false error reports. You need to use the flag 184 --partial-loads-ok=yes to get the benefits of these changes. 185 186 - Better control over the leak checker. It is now possible to 187 specify which leak kinds (definite/indirect/possible/reachable) 188 should be displayed, which should be regarded as errors, and which 189 should be suppressed by a given leak suppression. This is done 190 using the options --show-leak-kinds=kind1,kind2,.., 191 --errors-for-leak-kinds=kind1,kind2,.. and an optional 192 "match-leak-kinds:" line in suppression entries, respectively. 193 194 Note that generated leak suppressions contain this new line and 195 are therefore more specific than in previous releases. To get the 196 same behaviour as previous releases, remove the "match-leak-kinds:" 197 line from generated suppressions before using them. 198 199 - Reduced "possible leak" reports from the leak checker by the use 200 of better heuristics. The available heuristics provide detection 201 of valid interior pointers to std::stdstring, to new[] allocated 202 arrays with elements having destructors and to interior pointers 203 pointing to an inner part of a C++ object using multiple 204 inheritance. They can be selected individually using the 205 option --leak-check-heuristics=heur1,heur2,... 206 207 - Better control of stacktrace acquisition for heap-allocated 208 blocks. Using the --keep-stacktraces option, it is possible to 209 control independently whether a stack trace is acquired for each 210 allocation and deallocation. This can be used to create better 211 "use after free" errors or to decrease Valgrind's resource 212 consumption by recording less information. 213 214 - Better reporting of leak suppression usage. The list of used 215 suppressions (shown when the -v option is given) now shows, for 216 each leak suppressions, how many blocks and bytes it suppressed 217 during the last leak search. 218 219 * Helgrind: 220 221 - False errors resulting from the use of statically initialised 222 mutexes and condition variables (PTHREAD_MUTEX_INITIALISER, etc) 223 have been removed. 224 225 - False errors resulting from the use of pthread_cond_waits that 226 timeout, have been removed. 227 228 * ==================== OTHER CHANGES ==================== 229 230 * Some attempt to tune Valgrind's space requirements to the expected 231 capabilities of the target: 232 233 - The default size of the translation cache has been reduced from 8 234 sectors to 6 on Android platforms, since each sector occupies 235 about 40MB when using Memcheck. 236 237 - The default size of the translation cache has been increased to 16 238 sectors on all other platforms, reflecting the fact that large 239 applications require instrumentation and storage of huge amounts 240 of code. For similar reasons, the number of memory mapped 241 segments that can be tracked has been increased by a factor of 6. 242 243 - In all cases, the maximum number of sectors in the translation 244 cache can be controlled by the new flag --num-transtab-sectors. 245 246 * Changes in how debug info (line numbers, etc) is read: 247 248 - Valgrind no longer temporarily mmaps the entire object to read 249 from it. Instead, reading is done through a small fixed sized 250 buffer. This avoids virtual memory usage spikes when Valgrind 251 reads debuginfo from large shared objects. 252 253 - A new experimental remote debug info server. Valgrind can read 254 debug info from a different machine (typically, a build host) 255 where debuginfo objects are stored. This can save a lot of time 256 and hassle when running Valgrind on resource-constrained targets 257 (phones, tablets) when the full debuginfo objects are stored 258 somewhere else. This is enabled by the --debuginfo-server= 259 option. 260 261 - Consistency checking between main and debug objects can be 262 disabled using the --allow-mismatched-debuginfo option. 263 264 * Stack unwinding by stack scanning, on ARM. Unwinding by stack 265 scanning can recover stack traces in some cases when the normal 266 unwind mechanisms fail. Stack scanning is best described as "a 267 nasty, dangerous and misleading hack" and so is disabled by default. 268 Use --unw-stack-scan-thresh and --unw-stack-scan-frames to enable 269 and control it. 270 271 * Detection and merging of recursive stack frame cycles. When your 272 program has recursive algorithms, this limits the memory used by 273 Valgrind for recorded stack traces and avoids recording 274 uninteresting repeated calls. This is controlled by the command 275 line option --merge-recursive-frame and by the monitor command 276 "v.set merge-recursive-frames". 277 278 * File name and line numbers for used suppressions. The list of used 279 suppressions (shown when the -v option is given) now shows, for each 280 used suppression, the file name and line number where the suppression 281 is defined. 282 283 * New and modified GDB server monitor features: 284 285 - valgrind.h has a new client request, VALGRIND_MONITOR_COMMAND, 286 that can be used to execute gdbserver monitor commands from the 287 client program. 288 289 - A new monitor command, "v.info open_fds", that gives the list of 290 open file descriptors and additional details. 291 292 - An optional message in the "v.info n_errs_found" monitor command, 293 for example "v.info n_errs_found test 1234 finished", allowing a 294 comment string to be added to the process output, perhaps for the 295 purpose of separating errors of different tests or test phases. 296 297 - A new monitor command "v.info execontext" that shows information 298 about the stack traces recorded by Valgrind. 299 300 - A new monitor command "v.do expensive_sanity_check_general" to run 301 some internal consistency checks. 302 303 * New flag --sigill-diagnostics to control whether a diagnostic 304 message is printed when the JIT encounters an instruction it can't 305 translate. The actual behavior -- delivery of SIGILL to the 306 application -- is unchanged. 307 308 * The maximum amount of memory that Valgrind can use on 64 bit targets 309 has been increased from 32GB to 64GB. This should make it possible 310 to run applications on Memcheck that natively require up to about 35GB. 311 312 * ==================== FIXED BUGS ==================== 313 314 The following bugs have been fixed or resolved. Note that "n-i-bz" 315 stands for "not in bugzilla" -- that is, a bug that was reported to us 316 but never got a bugzilla entry. We encourage you to file bugs in 317 bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather 318 than mailing the developers (or mailing lists) directly -- bugs that 319 are not entered into bugzilla tend to get forgotten about or ignored. 320 321 To see details of a given bug, visit 322 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 323 where XXXXXX is the bug number as listed below. 324 325 123837 system call: 4th argument is optional, depending on cmd 326 135425 memcheck should tell you where Freed blocks were Mallocd 327 164485 VG_N_SEGNAMES and VG_N_SEGMENTS are (still) too small 328 207815 Adds some of the drm ioctls to syswrap-linux.c 329 251569 vex amd64->IR: 0xF 0x1 0xF9 0xBF 0x90 0xD0 0x3 0x0 (RDTSCP) 330 252955 Impossible to compile with ccache 331 253519 Memcheck reports auxv pointer accesses as invalid reads. 332 263034 Crash when loading some PPC64 binaries 333 269599 Increase deepest backtrace 334 274695 s390x: Support "compare to/from logical" instructions (z196) 335 275800 s390x: Autodetect cache info (part 2) 336 280271 Valgrind reports possible memory leaks on still-reachable std::string 337 284540 Memcheck shouldn't count suppressions matching still-reachable [..] 338 289578 Backtraces with ARM unwind tables (stack scan flags) 339 296311 Wrong stack traces due to -fomit-frame-pointer (x86) 340 304832 ppc32: build failure 341 305431 Use find_buildid shdr fallback for separate .debug files 342 305728 Add support for AVX2 instructions 343 305948 ppc64: code generation for ShlD64 / ShrD64 asserts 344 306035 s390x: Fix IR generation for LAAG and friends 345 306054 s390x: Condition code computation for convert-to-int/logical 346 306098 s390x: alternate opcode form for convert to/from fixed 347 306587 Fix cache line detection from auxiliary vector for PPC. 348 306783 Mips unhandled syscall : 4025 / 4079 / 4182 349 307038 DWARF2 CFI reader: unhandled DW_OP_ opcode 0x8 (DW_OP_const1u et al) 350 307082 HG false positive: pthread_cond_destroy: destruction of unknown CV 351 307101 sys_capget second argument can be NULL 352 307103 sys_openat: If pathname is absolute, then dirfd is ignored. 353 307106 amd64->IR: f0 0f c0 02 (lock xadd byte) 354 307113 s390x: DFP support 355 307141 valgrind does't work in mips-linux system 356 307155 filter_gdb should filter out syscall-template.S T_PSEUDO 357 307285 x86_amd64 feature test for avx in test suite is wrong 358 307290 memcheck overlap testcase needs memcpy version filter 359 307463 Please add "&limit=0" to the "all open bugs" link 360 307465 --show-possibly-lost=no should reduce the error count / exit code 361 307557 Leaks on Mac OS X 10.7.5 libraries at ImageLoader::recursiveInit[..] 362 307729 pkgconfig support broken valgrind.pc 363 307828 Memcheck false errors SSE optimized wcscpy, wcscmp, wcsrchr, wcschr 364 307955 Building valgrind 3.7.0-r4 fails in Gentoo AMD64 when using clang 365 308089 Unhandled syscall on ppc64: prctl 366 308135 PPC32 MPC8xx has 16 bytes cache size 367 308321 testsuite memcheck filter interferes with gdb_filter 368 308333 == 307106 369 308341 vgdb should report process exit (or fatal signal) 370 308427 s390 memcheck reports tsearch cjump/cmove depends on uninit 371 308495 Remove build dependency on installed Xen headers 372 308573 Internal error on 64-bit instruction executed in 32-bit mode 373 308626 == 308627 374 308627 pmovmskb validity bit propagation is imprecise 375 308644 vgdb command for having the info for the track-fds option 376 308711 give more info about aspacemgr and arenas in out_of_memory 377 308717 ARM: implement fixed-point VCVT.F64.[SU]32 378 308718 ARM implement SMLALBB family of instructions 379 308886 Missing support for PTRACE_SET/GETREGSET 380 308930 syscall name_to_handle_at (303 on amd64) not handled 381 309229 V-bit tester does not report number of tests generated 382 309323 print unrecognized instuction on MIPS 383 309425 Provide a --sigill-diagnostics flag to suppress illegal [..] 384 309427 SSE optimized stpncpy trigger uninitialised value [..] errors 385 309430 Self hosting ppc64 encounters a vassert error on operand type 386 309600 valgrind is a bit confused about 0-sized sections 387 309823 Generate errors for still reachable blocks 388 309921 PCMPISTRI validity bit propagation is imprecise 389 309922 none/tests/ppc64/test_dfp5 sometimes fails 390 310169 The Iop_CmpORD class of Iops is not supported by the vbit checker. 391 310424 --read-var-info does not properly describe static variables 392 310792 search additional path for debug symbols 393 310931 s390x: Message-security assist (MSA) instruction extension [..] 394 311100 PPC DFP implementation of the integer operands is inconsistent [..] 395 311318 ARM: "128-bit constant is not implemented" error message 396 311407 ssse3 bcopy (actually converted memcpy) causes invalid read [..] 397 311690 V crashes because it redirects branches inside of a redirected function 398 311880 x86_64: make regtest hangs at shell_valid1 399 311922 WARNING: unhandled syscall: 170 400 311933 == 251569 401 312171 ppc: insn selection for DFP 402 312571 Rounding mode call wrong for the DFP Iops [..] 403 312620 Change to Iop_D32toD64 [..] for s390 DFP support broke ppc [..] 404 312913 Dangling pointers error should also report the alloc stack trace 405 312980 Building on Mountain Lion generates some compiler warnings 406 313267 Adding MIPS64/Linux port to Valgrind 407 313348 == 251569 408 313354 == 251569 409 313811 Buffer overflow in assert_fail 410 314099 coverity pointed out error in VEX guest_ppc_toIR.c insn_suffix 411 314269 ppc: dead code in insn selection 412 314718 ARM: implement integer divide instruction (sdiv and udiv) 413 315345 cl-format.xml and callgrind/dump.c don't agree on using cfl= or cfi= 414 315441 sendmsg syscall should ignore unset msghdr msg_flags 415 315534 msgrcv inside a thread causes valgrind to hang (block) 416 315545 Assertion '(UChar*)sec->tt[tteNo].tcptr <= (UChar*)hcode' failed 417 315689 disInstr(thumb): unhandled instruction: 0xF852 0x0E10 (LDRT) 418 315738 disInstr(arm): unhandled instruction: 0xEEBE0BEE (vcvt.s32.f64) 419 315959 valgrind man page has bogus SGCHECK (and no BBV) OPTIONS section 420 316144 valgrind.1 manpage contains unknown ??? strings [..] 421 316145 callgrind command line options in manpage reference (unknown) [..] 422 316145 callgrind command line options in manpage reference [..] 423 316181 drd: Fixed a 4x slowdown for certain applications 424 316503 Valgrind does not support SSE4 "movntdqa" instruction 425 316535 Use of |signed int| instead of |size_t| in valgrind messages 426 316696 fluidanimate program of parsec 2.1 stuck 427 316761 syscall open_by_handle_at (304 on amd64, 342 on x86) not handled 428 317091 Use -Wl,-Ttext-segment when static linking if possible [..] 429 317186 "Impossible happens" when occurs VCVT instruction on ARM 430 317318 Support for Threading Building Blocks "scalable_malloc" 431 317444 amd64->IR: 0xC4 0x41 0x2C 0xC2 0xD2 0x8 (vcmpeq_uqps) 432 317461 Fix BMI assembler configure check and avx2/bmi/fma vgtest prereqs 433 317463 bmi testcase IR SANITY CHECK FAILURE 434 317506 memcheck/tests/vbit-test fails with unknown opcode after [..] 435 318050 libmpiwrap fails to compile with out-of-source build 436 318203 setsockopt handling needs to handle SOL_SOCKET/SO_ATTACH_FILTER 437 318643 annotate_trace_memory tests infinite loop on arm and ppc [..] 438 318773 amd64->IR: 0xF3 0x48 0x0F 0xBC 0xC2 0xC3 0x66 0x0F 439 318929 Crash with: disInstr(thumb): 0xF321 0x0001 (ssat16) 440 318932 Add missing PPC64 and PPC32 system call support 441 319235 --db-attach=yes is broken with Yama (ptrace scoping) enabled 442 319395 Crash with unhandled instruction on STRT (Thumb) instructions 443 319494 VEX Makefile-gcc standalone build update after r2702 444 319505 [MIPSEL] Crash: unhandled UNRAY operator. 445 319858 disInstr(thumb): unhandled instruction on instruction STRBT 446 319932 disInstr(thumb): unhandled instruction on instruction STRHT 447 320057 Problems when we try to mmap more than 12 memory pages on MIPS32 448 320063 Memory from PTRACE_GET_THREAD_AREA is reported uninitialised 449 320083 disInstr(thumb): unhandled instruction on instruction LDRBT 450 320116 bind on AF_BLUETOOTH produces warnings because of sockaddr_rc padding 451 320131 WARNING: unhandled syscall: 369 on ARM (prlimit64) 452 320211 Stack buffer overflow in ./coregrind/m_main.c with huge TMPDIR 453 320661 vgModuleLocal_read_elf_debug_info(): "Assertion '!di->soname' 454 320895 add fanotify support (patch included) 455 320998 vex amd64->IR pcmpestri and pcmpestrm SSE4.2 instruction 456 321065 Valgrind updates for Xen 4.3 457 321148 Unhandled instruction: PLI (Thumb 1, 2, 3) 458 321363 Unhandled instruction: SSAX (ARM + Thumb) 459 321364 Unhandled instruction: SXTAB16 (ARM + Thumb) 460 321466 Unhandled instruction: SHASX (ARM + Thumb) 461 321467 Unhandled instruction: SHSAX (ARM + Thumb) 462 321468 Unhandled instruction: SHSUB16 (ARM + Thumb) 463 321619 Unhandled instruction: SHSUB8 (ARM + Thumb) 464 321620 Unhandled instruction: UASX (ARM + Thumb) 465 321621 Unhandled instruction: USAX (ARM + Thumb) 466 321692 Unhandled instruction: UQADD16 (ARM + Thumb) 467 321693 Unhandled instruction: LDRSBT (Thumb) 468 321694 Unhandled instruction: UQASX (ARM + Thumb) 469 321696 Unhandled instruction: UQSAX (Thumb + ARM) 470 321697 Unhandled instruction: UHASX (ARM + Thumb) 471 321703 Unhandled instruction: UHSAX (ARM + Thumb) 472 321704 Unhandled instruction: REVSH (ARM + Thumb) 473 321730 Add cg_diff and cg_merge man pages 474 321738 Add vgdb and valgrind-listener man pages 475 321814 == 315545 476 321891 Unhandled instruction: LDRHT (Thumb) 477 321960 pthread_create() then alloca() causing invalid stack write errors 478 321969 ppc32 and ppc64 don't support [lf]setxattr 479 322254 Show threadname together with tid if set by application 480 322294 Add initial support for IBM Power ISA 2.07 481 322368 Assertion failure in wqthread_hijack under OS X 10.8 482 322563 vex mips->IR: 0x70 0x83 0xF0 0x3A 483 322807 VALGRIND_PRINTF_BACKTRACE writes callstack to xml and text to stderr 484 322851 0bXXX binary literal syntax is not standard 485 323035 Unhandled instruction: LDRSHT(Thumb) 486 323036 Unhandled instruction: SMMLS (ARM and Thumb) 487 323116 The memcheck/tests/ppc64/power_ISA2_05.c fails to build [..] 488 323175 Unhandled instruction: SMLALD (ARM + Thumb) 489 323177 Unhandled instruction: SMLSLD (ARM + Thumb) 490 323432 Calling pthread_cond_destroy() or pthread_mutex_destroy() [..] 491 323437 Phase 2 support for IBM Power ISA 2.07 492 323713 Support mmxext (integer sse) subset on i386 (athlon) 493 323803 Transactional memory instructions are not supported for Power 494 323893 SSE3 not available on amd cpus in valgrind 495 323905 Probable false positive from Valgrind/drd on close() 496 323912 valgrind.h header isn't compatible for mingw64 497 324047 Valgrind doesn't support [LDR,ST]{S}[B,H]T ARM instructions 498 324149 helgrind: When pthread_cond_timedwait returns ETIMEDOUT [..] 499 324181 mmap does not handle MAP_32BIT 500 324227 memcheck false positive leak when a thread calls exit+block [..] 501 324421 Support for fanotify API on ARM architecture 502 324514 gdbserver monitor cmd output behaviour consistency [..] 503 324518 ppc64: Emulation of dcbt instructions does not handle [..] 504 324546 none/tests/ppc32 test_isa_2_07_part2 requests -m64 505 324582 When access is made to freed memory, report both allocation [..] 506 324594 Fix overflow computation for Power ISA 2.06 insns: mulldo/mulldo. 507 324765 ppc64: illegal instruction when executing none/tests/ppc64/jm-misc 508 324816 Incorrect VEX implementation for xscvspdp/xvcvspdp for SNaN inputs 509 324834 Unhandled instructions in Microsoft C run-time for x86_64 510 324894 Phase 3 support for IBM Power ISA 2.07 511 326091 drd: Avoid false race reports from optimized strlen() impls 512 326113 valgrind libvex hwcaps error on AMD64 513 n-i-bz Some wrong command line options could be ignored 514 n-i-bz patch to allow fair-sched on android 515 n-i-bz report error for vgdb snapshot requested before execution 516 n-i-bz same as 303624 (fixed in 3.8.0), but for x86 android 517 518 (3.9.0: 31 October 2013, vex r2796, valgrind r13708) 519 520 521 522 Release 3.8.1 (19 September 2012) 523 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 524 3.8.1 is a bug fix release. It fixes some assertion failures in 3.8.0 525 that occur moderately frequently in real use cases, adds support for 526 some missing instructions on ARM, and fixes a deadlock condition on 527 MacOSX. If you package or deliver 3.8.0 for others to use, you might 528 want to consider upgrading to 3.8.1 instead. 529 530 The following bugs have been fixed or resolved. Note that "n-i-bz" 531 stands for "not in bugzilla" -- that is, a bug that was reported to us 532 but never got a bugzilla entry. We encourage you to file bugs in 533 bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather 534 than mailing the developers (or mailing lists) directly -- bugs that 535 are not entered into bugzilla tend to get forgotten about or ignored. 536 537 To see details of a given bug, visit 538 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 539 where XXXXXX is the bug number as listed below. 540 541 284004 == 301281 542 289584 Unhandled instruction: 0xF 0x29 0xE5 (MOVAPS) 543 295808 amd64->IR: 0xF3 0xF 0xBC 0xC0 (TZCNT) 544 298281 wcslen causes false(?) uninitialised value warnings 545 301281 valgrind hangs on OS X when the process calls system() 546 304035 disInstr(arm): unhandled instruction 0xE1023053 547 304867 implement MOVBE instruction in x86 mode 548 304980 Assertion 'lo <= hi' failed in vgModuleLocal_find_rx_mapping 549 305042 amd64: implement 0F 7F encoding of movq between two registers 550 305199 ARM: implement QDADD and QDSUB 551 305321 amd64->IR: 0xF 0xD 0xC (prefetchw) 552 305513 killed by fatal signal: SIGSEGV 553 305690 DRD reporting invalid semaphore when sem_trywait fails 554 305926 Invalid alignment checks for some AVX instructions 555 306297 disInstr(thumb): unhandled instruction 0xE883 0x000C 556 306310 3.8.0 release tarball missing some files 557 306612 RHEL 6 glibc-2.X default suppressions need /lib*/libc-*patterns 558 306664 vex amd64->IR: 0x66 0xF 0x3A 0x62 0xD1 0x46 0x66 0xF 559 n-i-bz shmat of a segment > 4Gb does not work 560 n-i-bz simulate_control_c script wrong USR1 signal number on mips 561 n-i-bz vgdb ptrace calls wrong on mips [...] 562 n-i-bz Fixes for more MPI false positives 563 n-i-bz exp-sgcheck's memcpy causes programs to segfault 564 n-i-bz OSX build w/ clang: asserts at startup 565 n-i-bz Incorrect undef'dness prop for Iop_DPBtoBCD and Iop_BCDtoDPB 566 n-i-bz fix a couple of union tag-vs-field mixups 567 n-i-bz OSX: use __NR_poll_nocancel rather than __NR_poll 568 569 The following bugs were fixed in 3.8.0 but not listed in this NEWS 570 file at the time: 571 572 254088 Valgrind should know about UD2 instruction 573 301280 == 254088 574 301902 == 254088 575 304754 NEWS blows TeX's little mind 576 577 (3.8.1: 19 September 2012, vex r2537, valgrind r12996) 578 579 580 581 Release 3.8.0 (10 August 2012) 582 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 583 3.8.0 is a feature release with many improvements and the usual 584 collection of bug fixes. 585 586 This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux, 587 PPC64/Linux, S390X/Linux, MIPS/Linux, ARM/Android, X86/Android, 588 X86/MacOSX 10.6/10.7 and AMD64/MacOSX 10.6/10.7. Support for recent 589 distros and toolchain components (glibc 2.16, gcc 4.7) has been added. 590 There is initial support for MacOSX 10.8, but it is not usable for 591 serious work at present. 592 593 * ================== PLATFORM CHANGES ================= 594 595 * Support for MIPS32 platforms running Linux. Valgrind has been 596 tested on MIPS32 and MIPS32r2 platforms running different Debian 597 Squeeze and MeeGo distributions. Both little-endian and big-endian 598 cores are supported. The tools Memcheck, Massif and Lackey have 599 been tested and are known to work. See README.mips for more details. 600 601 * Preliminary support for Android running on x86. 602 603 * Preliminary (as-yet largely unusable) support for MacOSX 10.8. 604 605 * Support for Intel AVX instructions and for AES instructions. This 606 support is available only for 64 bit code. 607 608 * Support for POWER Decimal Floating Point instructions. 609 610 * ==================== TOOL CHANGES ==================== 611 612 * Non-libc malloc implementations are now supported. This is useful 613 for tools that replace malloc (Memcheck, Massif, DRD, Helgrind). 614 Using the new option --soname-synonyms, such tools can be informed 615 that the malloc implementation is either linked statically into the 616 executable, or is present in some other shared library different 617 from libc.so. This makes it possible to process statically linked 618 programs, and programs using other malloc libraries, for example 619 TCMalloc or JEMalloc. 620 621 * For tools that provide their own replacement for malloc et al, the 622 option --redzone-size=<number> allows users to specify the size of 623 the padding blocks (redzones) added before and after each client 624 allocated block. Smaller redzones decrease the memory needed by 625 Valgrind. Bigger redzones increase the chance to detect blocks 626 overrun or underrun. Prior to this change, the redzone size was 627 hardwired to 16 bytes in Memcheck. 628 629 * Memcheck: 630 631 - The leak_check GDB server monitor command now can 632 control the maximum nr of loss records to output. 633 634 - Reduction of memory use for applications allocating 635 many blocks and/or having many partially defined bytes. 636 637 - Addition of GDB server monitor command 'block_list' that lists 638 the addresses/sizes of the blocks of a leak search loss record. 639 640 - Addition of GDB server monitor command 'who_points_at' that lists 641 the locations pointing at a block. 642 643 - If a redzone size > 0 is given, VALGRIND_MALLOCLIKE_BLOCK now will 644 detect an invalid access of these redzones, by marking them 645 noaccess. Similarly, if a redzone size is given for a memory 646 pool, VALGRIND_MEMPOOL_ALLOC will mark the redzones no access. 647 This still allows to find some bugs if the user has forgotten to 648 mark the pool superblock noaccess. 649 650 - Performance of memory leak check has been improved, especially in 651 cases where there are many leaked blocks and/or many suppression 652 rules used to suppress leak reports. 653 654 - Reduced noise (false positive) level on MacOSX 10.6/10.7, due to 655 more precise analysis, which is important for LLVM/Clang 656 generated code. This is at the cost of somewhat reduced 657 performance. Note there is no change to analysis precision or 658 costs on Linux targets. 659 660 * DRD: 661 662 - Added even more facilities that can help finding the cause of a data 663 race, namely the command-line option --ptrace-addr and the macro 664 DRD_STOP_TRACING_VAR(x). More information can be found in the manual. 665 666 - Fixed a subtle bug that could cause false positive data race reports. 667 668 * ==================== OTHER CHANGES ==================== 669 670 * The C++ demangler has been updated so as to work well with C++ 671 compiled by up to at least g++ 4.6. 672 673 * Tool developers can make replacement/wrapping more flexible thanks 674 to the new option --soname-synonyms. This was reported above, but 675 in fact is very general and applies to all function 676 replacement/wrapping, not just to malloc-family functions. 677 678 * Round-robin scheduling of threads can be selected, using the new 679 option --fair-sched= yes. Prior to this change, the pipe-based 680 thread serialisation mechanism (which is still the default) could 681 give very unfair scheduling. --fair-sched=yes improves 682 responsiveness of interactive multithreaded applications, and 683 improves repeatability of results from the thread checkers Helgrind 684 and DRD. 685 686 * For tool developers: support to run Valgrind on Valgrind has been 687 improved. We can now routinely Valgrind on Helgrind or Memcheck. 688 689 * gdbserver now shows the float shadow registers as integer 690 rather than float values, as the shadow values are mostly 691 used as bit patterns. 692 693 * Increased limit for the --num-callers command line flag to 500. 694 695 * Performance improvements for error matching when there are many 696 suppression records in use. 697 698 * Improved support for DWARF4 debugging information (bug 284184). 699 700 * Initial support for DWZ compressed Dwarf debug info. 701 702 * Improved control over the IR optimiser's handling of the tradeoff 703 between performance and precision of exceptions. Specifically, 704 --vex-iropt-precise-memory-exns has been removed and replaced by 705 --vex-iropt-register-updates, with extended functionality. This 706 allows the Valgrind gdbserver to always show up to date register 707 values to GDB. 708 709 * Modest performance gains through the use of translation chaining for 710 JIT-generated code. 711 712 * ==================== FIXED BUGS ==================== 713 714 The following bugs have been fixed or resolved. Note that "n-i-bz" 715 stands for "not in bugzilla" -- that is, a bug that was reported to us 716 but never got a bugzilla entry. We encourage you to file bugs in 717 bugzilla (https://bugs.kde.org/enter_bug.cgi?product=valgrind) rather 718 than mailing the developers (or mailing lists) directly -- bugs that 719 are not entered into bugzilla tend to get forgotten about or ignored. 720 721 To see details of a given bug, visit 722 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 723 where XXXXXX is the bug number as listed below. 724 725 197914 Building valgrind from svn now requires automake-1.10 726 203877 increase to 16Mb maximum allowed alignment for memalign et al 727 219156 Handle statically linked malloc or other malloc lib (e.g. tcmalloc) 728 247386 make perf does not run all performance tests 729 270006 Valgrind scheduler unfair 730 270777 Adding MIPS/Linux port to Valgrind 731 270796 s390x: Removed broken support for the TS insn 732 271438 Fix configure for proper SSE4.2 detection 733 273114 s390x: Support TR, TRE, TROO, TROT, TRTO, and TRTT instructions 734 273475 Add support for AVX instructions 735 274078 improved configure logic for mpicc 736 276993 fix mremap 'no thrash checks' 737 278313 Fedora 15/x64: err read debug info with --read-var-info=yes flag 738 281482 memcheck incorrect byte allocation count in realloc() for silly argument 739 282230 group allocator for small fixed size, use it for MC_Chunk/SEc vbit 740 283413 Fix wrong sanity check 741 283671 Robustize alignment computation in LibVEX_Alloc 742 283961 Adding support for some HCI IOCTLs 743 284124 parse_type_DIE: confused by: DWARF 4 744 284864 == 273475 (Add support for AVX instructions) 745 285219 Too-restrictive constraints for Thumb2 "SP plus/minus register" 746 285662 (MacOSX): Memcheck needs to replace memcpy/memmove 747 285725 == 273475 (Add support for AVX instructions) 748 286261 add wrapper for linux I2C_RDWR ioctl 749 286270 vgpreload is not friendly to 64->32 bit execs, gives ld.so warnings 750 286374 Running cachegrind with --branch-sim=yes on 64-bit PowerPC program fails 751 286384 configure fails "checking for a supported version of gcc" 752 286497 == 273475 (Add support for AVX instructions) 753 286596 == 273475 (Add support for AVX instructions) 754 286917 disInstr(arm): unhandled instruction: QADD (also QSUB) 755 287175 ARM: scalar VFP fixed-point VCVT instructions not handled 756 287260 Incorrect conditional jump or move depends on uninitialised value(s) 757 287301 vex amd64->IR: 0x66 0xF 0x38 0x41 0xC0 0xB8 0x0 0x0 (PHMINPOSUW) 758 287307 == 273475 (Add support for AVX instructions) 759 287858 VG_(strerror): unknown error 760 288298 (MacOSX) unhandled syscall shm_unlink 761 288995 == 273475 (Add support for AVX instructions) 762 289470 Loading of large Mach-O thin binaries fails. 763 289656 == 273475 (Add support for AVX instructions) 764 289699 vgdb connection in relay mode erroneously closed due to buffer overrun 765 289823 == 293754 (PCMPxSTRx not implemented for 16-bit characters) 766 289839 s390x: Provide support for unicode conversion instructions 767 289939 monitor cmd 'leak_check' with details about leaked or reachable blocks 768 290006 memcheck doesn't mark %xmm as initialized after "pcmpeqw %xmm %xmm" 769 290655 Add support for AESKEYGENASSIST instruction 770 290719 valgrind-3.7.0 fails with automake-1.11.2 due to"pkglibdir" usage 771 290974 vgdb must align pages to VKI_SHMLBA (16KB) on ARM 772 291253 ES register not initialised in valgrind simulation 773 291568 Fix 3DNOW-related crashes with baseline x86_64 CPU (w patch) 774 291865 s390x: Support the "Compare Double and Swap" family of instructions 775 292300 == 273475 (Add support for AVX instructions) 776 292430 unrecognized instruction in __intel_get_new_mem_ops_cpuid 777 292493 == 273475 (Add support for AVX instructions) 778 292626 Missing fcntl F_SETOWN_EX and F_GETOWN_EX support 779 292627 Missing support for some SCSI ioctls 780 292628 none/tests/x86/bug125959-x86.c triggers undefined behavior 781 292841 == 273475 (Add support for AVX instructions) 782 292993 implement the getcpu syscall on amd64-linux 783 292995 Implement the cross memory attach syscalls introduced in Linux 3.2 784 293088 Add some VEX sanity checks for ppc64 unhandled instructions 785 293751 == 290655 (Add support for AESKEYGENASSIST instruction) 786 293754 PCMPxSTRx not implemented for 16-bit characters 787 293755 == 293754 (No tests for PCMPxSTRx on 16-bit characters) 788 293808 CLFLUSH not supported by latest VEX for amd64 789 294047 valgrind does not correctly emulate prlimit64(..., RLIMIT_NOFILE, ...) 790 294048 MPSADBW instruction not implemented 791 294055 regtest none/tests/shell fails when locale is not set to C 792 294185 INT 0x44 (and others) not supported on x86 guest, but used by Jikes RVM 793 294190 --vgdb-error=xxx can be out of sync with errors shown to the user 794 294191 amd64: fnsave/frstor and 0x66 size prefixes on FP instructions 795 294260 disInstr_AMD64: disInstr miscalculated next %rip 796 294523 --partial-loads-ok=yes causes false negatives 797 294617 vex amd64->IR: 0x66 0xF 0x3A 0xDF 0xD1 0x1 0xE8 0x6A 798 294736 vex amd64->IR: 0x48 0xF 0xD7 0xD6 0x48 0x83 799 294812 patch allowing to run (on x86 at least) helgrind/drd on tool. 800 295089 can not annotate source for both helgrind and drd 801 295221 POWER Processor decimal floating point instruction support missing 802 295427 building for i386 with clang on darwin11 requires "-new_linker linker" 803 295428 coregrind/m_main.c has incorrect x86 assembly for darwin 804 295590 Helgrind: Assertion 'cvi->nWaiters > 0' failed 805 295617 ARM - Add some missing syscalls 806 295799 Missing \n with get_vbits in gdbserver when line is % 80 [...] 807 296229 Linux user input device ioctls missing wrappers 808 296318 ELF Debug info improvements (more than one rx/rw mapping) 809 296422 Add translation chaining support 810 296457 vex amd64->IR: 0x66 0xF 0x3A 0xDF 0xD1 0x1 0xE8 0x6A (dup of AES) 811 296792 valgrind 3.7.0: add SIOCSHWTSTAMP (0x89B0) ioctl wrapper 812 296983 Fix build issues on x86_64/ppc64 without 32-bit toolchains 813 297078 gdbserver signal handling problems [..] 814 297147 drd false positives on newly allocated memory 815 297329 disallow decoding of IBM Power DFP insns on some machines 816 297497 POWER Processor decimal floating point instruction support missing 817 297701 Another alias for strncasecmp_l in libc-2.13.so 818 297911 'invalid write' not reported when using APIs for custom mem allocators. 819 297976 s390x: revisit EX implementation 820 297991 Valgrind interferes with mmap()+ftell() 821 297992 Support systems missing WIFCONTINUED (e.g. pre-2.6.10 Linux) 822 297993 Fix compilation of valgrind with gcc -g3. 823 298080 POWER Processor DFP support missing, part 3 824 298227 == 273475 (Add support for AVX instructions) 825 298335 == 273475 (Add support for AVX instructions) 826 298354 Unhandled ARM Thumb instruction 0xEB0D 0x0585 (streq) 827 298394 s390x: Don't bail out on an unknown machine model. [..] 828 298421 accept4() syscall (366) support is missing for ARM 829 298718 vex amd64->IR: 0xF 0xB1 0xCB 0x9C 0x8F 0x45 830 298732 valgrind installation problem in ubuntu with kernel version 3.x 831 298862 POWER Processor DFP instruction support missing, part 4 832 298864 DWARF reader mis-parses DW_FORM_ref_addr 833 298943 massif asserts with --pages-as-heap=yes when brk is changing [..] 834 299053 Support DWARF4 DW_AT_high_pc constant form 835 299104 == 273475 (Add support for AVX instructions) 836 299316 Helgrind: hg_main.c:628 (map_threads_lookup): Assertion 'thr' failed. 837 299629 dup3() syscall (358) support is missing for ARM 838 299694 POWER Processor DFP instruction support missing, part 5 839 299756 Ignore --free-fill for MEMPOOL_FREE and FREELIKE client requests 840 299803 == 273475 (Add support for AVX instructions) 841 299804 == 273475 (Add support for AVX instructions) 842 299805 == 273475 (Add support for AVX instructions) 843 300140 ARM - Missing (T1) SMMUL 844 300195 == 296318 (ELF Debug info improvements (more than one rx/rw mapping)) 845 300389 Assertion `are_valid_hwcaps(VexArchAMD64, [..])' failed. 846 300414 FCOM and FCOMP unimplemented for amd64 guest 847 301204 infinite loop in canonicaliseSymtab with ifunc symbol 848 301229 == 203877 (increase to 16Mb maximum allowed alignment for memalign etc) 849 301265 add x86 support to Android build 850 301984 configure script doesn't detect certain versions of clang 851 302205 Fix compiler warnings for POWER VEX code and POWER test cases 852 302287 Unhandled movbe instruction on Atom processors 853 302370 PPC: fnmadd, fnmsub, fnmadds, fnmsubs insns always negate the result 854 302536 Fix for the POWER Valgrind regression test: memcheck-ISA2.0. 855 302578 Unrecognized isntruction 0xc5 0x32 0xc2 0xca 0x09 vcmpngess 856 302656 == 273475 (Add support for AVX instructions) 857 302709 valgrind for ARM needs extra tls support for android emulator [..] 858 302827 add wrapper for CDROM_GET_CAPABILITY 859 302901 Valgrind crashes with dwz optimized debuginfo 860 302918 Enable testing of the vmaddfp and vnsubfp instructions in the testsuite 861 303116 Add support for the POWER instruction popcntb 862 303127 Power test suite fixes for frsqrte, vrefp, and vrsqrtefp instructions. 863 303250 Assertion `instrs_in->arr_used <= 10000' failed w/ OpenSSL code 864 303466 == 273475 (Add support for AVX instructions) 865 303624 segmentation fault on Android 4.1 (e.g. on Galaxy Nexus OMAP) 866 303963 strstr() function produces wrong results under valgrind callgrind 867 304054 CALL_FN_xx macros need to enforce stack alignment 868 304561 tee system call not supported 869 715750 (MacOSX): Incorrect invalid-address errors near 0xFFFFxxxx (mozbug#) 870 n-i-bz Add missing gdbserver xml files for shadow registers for ppc32 871 n-i-bz Bypass gcc4.4/4.5 code gen bugs causing out of memory or asserts 872 n-i-bz Fix assert in gdbserver for watchpoints watching the same address 873 n-i-bz Fix false positive in sys_clone on amd64 when optional args [..] 874 n-i-bz s390x: Shadow registers can now be examined using vgdb 875 876 (3.8.0-TEST3: 9 August 2012, vex r2465, valgrind r12865) 877 (3.8.0: 10 August 2012, vex r2465, valgrind r12866) 878 879 880 881 Release 3.7.0 (5 November 2011) 882 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 883 3.7.0 is a feature release with many significant improvements and the 884 usual collection of bug fixes. 885 886 This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux, 887 PPC64/Linux, S390X/Linux, ARM/Android, X86/Darwin and AMD64/Darwin. 888 Support for recent distros and toolchain components (glibc 2.14, gcc 889 4.6, MacOSX 10.7) has been added. 890 891 * ================== PLATFORM CHANGES ================= 892 893 * Support for IBM z/Architecture (s390x) running Linux. Valgrind can 894 analyse 64-bit programs running on z/Architecture. Most user space 895 instructions up to and including z10 are supported. Valgrind has 896 been tested extensively on z9, z10, and z196 machines running SLES 897 10/11, RedHat 5/6m, and Fedora. The Memcheck and Massif tools are 898 known to work well. Callgrind, Helgrind, and DRD work reasonably 899 well on z9 and later models. See README.s390 for more details. 900 901 * Preliminary support for MacOSX 10.7 and XCode 4. Both 32- and 902 64-bit processes are supported. Some complex threaded applications 903 (Firefox) are observed to hang when run as 32 bit applications, 904 whereas 64-bit versions run OK. The cause is unknown. Memcheck 905 will likely report some false errors. In general, expect some rough 906 spots. This release also supports MacOSX 10.6, but drops support 907 for 10.5. 908 909 * Preliminary support for Android (on ARM). Valgrind can now run 910 large applications (eg, Firefox) on (eg) a Samsung Nexus S. See 911 README.android for more details, plus instructions on how to get 912 started. 913 914 * Support for the IBM Power ISA 2.06 (Power7 instructions) 915 916 * General correctness and performance improvements for ARM/Linux, and, 917 by extension, ARM/Android. 918 919 * Further solidification of support for SSE 4.2 in 64-bit mode. AVX 920 instruction set support is under development but is not available in 921 this release. 922 923 * Support for AIX5 has been removed. 924 925 * ==================== TOOL CHANGES ==================== 926 927 * Memcheck: some incremental changes: 928 929 - reduction of memory use in some circumstances 930 931 - improved handling of freed memory, which in some circumstances 932 can cause detection of use-after-free that would previously have 933 been missed 934 935 - fix of a longstanding bug that could cause false negatives (missed 936 errors) in programs doing vector saturated narrowing instructions. 937 938 * Helgrind: performance improvements and major memory use reductions, 939 particularly for large, long running applications which perform many 940 synchronisation (lock, unlock, etc) events. Plus many smaller 941 changes: 942 943 - display of locksets for both threads involved in a race 944 945 - general improvements in formatting/clarity of error messages 946 947 - addition of facilities and documentation regarding annotation 948 of thread safe reference counted C++ classes 949 950 - new flag --check-stack-refs=no|yes [yes], to disable race checking 951 on thread stacks (a performance hack) 952 953 - new flag --free-is-write=no|yes [no], to enable detection of races 954 where one thread accesses heap memory but another one frees it, 955 without any coordinating synchronisation event 956 957 * DRD: enabled XML output; added support for delayed thread deletion 958 in order to detect races that occur close to the end of a thread 959 (--join-list-vol); fixed a memory leak triggered by repeated client 960 memory allocatation and deallocation; improved Darwin support. 961 962 * exp-ptrcheck: this tool has been renamed to exp-sgcheck 963 964 * exp-sgcheck: this tool has been reduced in scope so as to improve 965 performance and remove checking that Memcheck does better. 966 Specifically, the ability to check for overruns for stack and global 967 arrays is unchanged, but the ability to check for overruns of heap 968 blocks has been removed. The tool has accordingly been renamed to 969 exp-sgcheck ("Stack and Global Array Checking"). 970 971 * ==================== OTHER CHANGES ==================== 972 973 * GDB server: Valgrind now has an embedded GDB server. That means it 974 is possible to control a Valgrind run from GDB, doing all the usual 975 things that GDB can do (single stepping, breakpoints, examining 976 data, etc). Tool-specific functionality is also available. For 977 example, it is possible to query the definedness state of variables 978 or memory from within GDB when running Memcheck; arbitrarily large 979 memory watchpoints are supported, etc. To use the GDB server, start 980 Valgrind with the flag --vgdb-error=0 and follow the on-screen 981 instructions. 982 983 * Improved support for unfriendly self-modifying code: a new option 984 --smc-check=all-non-file is available. This adds the relevant 985 consistency checks only to code that originates in non-file-backed 986 mappings. In effect this confines the consistency checking only to 987 code that is or might be JIT generated, and avoids checks on code 988 that must have been compiled ahead of time. This significantly 989 improves performance on applications that generate code at run time. 990 991 * It is now possible to build a working Valgrind using Clang-2.9 on 992 Linux. 993 994 * new client requests VALGRIND_{DISABLE,ENABLE}_ERROR_REPORTING. 995 These enable and disable error reporting on a per-thread, and 996 nestable, basis. This is useful for hiding errors in particularly 997 troublesome pieces of code. The MPI wrapper library (libmpiwrap.c) 998 now uses this facility. 999 1000 * Added the --mod-funcname option to cg_diff. 1001 1002 * ==================== FIXED BUGS ==================== 1003 1004 The following bugs have been fixed or resolved. Note that "n-i-bz" 1005 stands for "not in bugzilla" -- that is, a bug that was reported to us 1006 but never got a bugzilla entry. We encourage you to file bugs in 1007 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than 1008 mailing the developers (or mailing lists) directly -- bugs that are 1009 not entered into bugzilla tend to get forgotten about or ignored. 1010 1011 To see details of a given bug, visit 1012 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 1013 where XXXXXX is the bug number as listed below. 1014 1015 210935 port valgrind.h (not valgrind) to win32 to support client requests 1016 214223 valgrind SIGSEGV on startup gcc 4.4.1 ppc32 (G4) Ubuntu 9.10 1017 243404 Port to zSeries 1018 243935 Helgrind: incorrect handling of ANNOTATE_HAPPENS_BEFORE()/AFTER() 1019 247223 non-x86: Suppress warning: 'regparm' attribute directive ignored 1020 250101 huge "free" memory usage due to m_mallocfree.c fragmentation 1021 253206 Some fixes for the faultstatus testcase 1022 255223 capget testcase fails when running as root 1023 256703 xlc_dbl_u32.c testcase broken 1024 256726 Helgrind tests have broken inline asm 1025 259977 == 214223 (Valgrind segfaults doing __builtin_longjmp) 1026 264800 testcase compile failure on zseries 1027 265762 make public VEX headers compilable by G++ 3.x 1028 265771 assertion in jumps.c (r11523) fails with glibc-2.3 1029 266753 configure script does not give the user the option to not use QtCore 1030 266931 gen_insn_test.pl is broken 1031 266961 ld-linux.so.2 i?86-linux strlen issues 1032 266990 setns instruction causes false positive 1033 267020 Make directory for temporary files configurable at run-time. 1034 267342 == 267997 (segmentation fault on Mac OS 10.6) 1035 267383 Assertion 'vgPlain_strlen(dir) + vgPlain_strlen(file) + 1 < 256' failed 1036 267413 Assertion 'DRD_(g_threadinfo)[tid].synchr_nesting >= 1' failed. 1037 267488 regtest: darwin support for 64-bit build 1038 267552 SIGSEGV (misaligned_stack_error) with DRD, but not with other tools 1039 267630 Add support for IBM Power ISA 2.06 -- stage 1 1040 267769 == 267997 (Darwin: memcheck triggers segmentation fault) 1041 267819 Add client request for informing the core about reallocation 1042 267925 laog data structure quadratic for a single sequence of lock 1043 267968 drd: (vgDrd_thread_set_joinable): Assertion '0 <= (int)tid ..' failed 1044 267997 MacOSX: 64-bit V segfaults on launch when built with Xcode 4.0.1 1045 268513 missed optimizations in fold_Expr 1046 268619 s390x: fpr - gpr transfer facility 1047 268620 s390x: reconsider "long displacement" requirement 1048 268621 s390x: improve IR generation for XC 1049 268715 s390x: FLOGR is not universally available 1050 268792 == 267997 (valgrind seg faults on startup when compiled with Xcode 4) 1051 268930 s390x: MHY is not universally available 1052 269078 arm->IR: unhandled instruction SUB (SP minus immediate/register) 1053 269079 Support ptrace system call on ARM 1054 269144 missing "Bad option" error message 1055 269209 conditional load and store facility (z196) 1056 269354 Shift by zero on x86 can incorrectly clobber CC_NDEP 1057 269641 == 267997 (valgrind segfaults immediately (segmentation fault)) 1058 269736 s390x: minor code generation tweaks 1059 269778 == 272986 (valgrind.h: swap roles of VALGRIND_DO_CLIENT_REQUEST() ..) 1060 269863 s390x: remove unused function parameters 1061 269864 s390x: tweak s390_emit_load_cc 1062 269884 == 250101 (overhead for huge blocks exhausts space too soon) 1063 270082 s390x: Make sure to point the PSW address to the next address on SIGILL 1064 270115 s390x: rewrite some testcases 1065 270309 == 267997 (valgrind crash on startup) 1066 270320 add support for Linux FIOQSIZE ioctl() call 1067 270326 segfault while trying to sanitize the environment passed to execle 1068 270794 IBM POWER7 support patch causes regression in none/tests 1069 270851 IBM POWER7 fcfidus instruction causes memcheck to fail 1070 270856 IBM POWER7 xsnmaddadp instruction causes memcheck to fail on 32bit app 1071 270925 hyper-optimized strspn() in /lib64/libc-2.13.so needs fix 1072 270959 s390x: invalid use of R0 as base register 1073 271042 VSX configure check fails when it should not 1074 271043 Valgrind build fails with assembler error on ppc64 with binutils 2.21 1075 271259 s390x: fix code confusion 1076 271337 == 267997 (Valgrind segfaults on MacOS X) 1077 271385 s390x: Implement Ist_MBE 1078 271501 s390x: misc cleanups 1079 271504 s390x: promote likely and unlikely 1080 271579 ppc: using wrong enum type 1081 271615 unhandled instruction "popcnt" (arch=amd10h) 1082 271730 Fix bug when checking ioctls: duplicate check 1083 271776 s390x: provide STFLE instruction support 1084 271779 s390x: provide clock instructions like STCK 1085 271799 Darwin: ioctls without an arg report a memory error 1086 271820 arm: fix type confusion 1087 271917 pthread_cond_timedwait failure leads to not-locked false positive 1088 272067 s390x: fix DISP20 macro 1089 272615 A typo in debug output in mc_leakcheck.c 1090 272661 callgrind_annotate chokes when run from paths containing regex chars 1091 272893 amd64->IR: 0x66 0xF 0x38 0x2B 0xC1 0x66 0xF 0x7F == (closed as dup) 1092 272955 Unhandled syscall error for pwrite64 on ppc64 arch 1093 272967 make documentation build-system more robust 1094 272986 Fix gcc-4.6 warnings with valgrind.h 1095 273318 amd64->IR: 0x66 0xF 0x3A 0x61 0xC1 0x38 (missing PCMPxSTRx case) 1096 273318 unhandled PCMPxSTRx case: vex amd64->IR: 0x66 0xF 0x3A 0x61 0xC1 0x38 1097 273431 valgrind segfaults in evalCfiExpr (debuginfo.c:2039) 1098 273465 Callgrind: jumps.c:164 (new_jcc): Assertion '(0 <= jmp) && ...' 1099 273536 Build error: multiple definition of `vgDrd_pthread_cond_initializer' 1100 273640 ppc64-linux: unhandled syscalls setresuid(164) and setresgid(169) 1101 273729 == 283000 (Illegal opcode for SSE2 "roundsd" instruction) 1102 273778 exp-ptrcheck: unhandled sysno == 259 1103 274089 exp-ptrcheck: unhandled sysno == 208 1104 274378 s390x: Various dispatcher tweaks 1105 274447 WARNING: unhandled syscall: 340 1106 274776 amd64->IR: 0x66 0xF 0x38 0x2B 0xC5 0x66 1107 274784 == 267997 (valgrind ls -l results in Segmentation Fault) 1108 274926 valgrind does not build against linux-3 1109 275148 configure FAIL with glibc-2.14 1110 275151 Fedora 15 / glibc-2.14 'make regtest' FAIL 1111 275168 Make Valgrind work for MacOSX 10.7 Lion 1112 275212 == 275284 (lots of false positives from __memcpy_ssse3_back et al) 1113 275278 valgrind does not build on Linux kernel 3.0.* due to silly 1114 275284 Valgrind memcpy/memmove redirection stopped working in glibc 2.14/x86_64 1115 275308 Fix implementation for ppc64 fres instruc 1116 275339 s390x: fix testcase compile warnings 1117 275517 s390x: Provide support for CKSM instruction 1118 275710 s390x: get rid of redundant address mode calculation 1119 275815 == 247894 (Valgrind doesn't know about Linux readahead(2) syscall) 1120 275852 == 250101 (valgrind uses all swap space and is killed) 1121 276784 Add support for IBM Power ISA 2.06 -- stage 3 1122 276987 gdbsrv: fix tests following recent commits 1123 277045 Valgrind crashes with unhandled DW_OP_ opcode 0x2a 1124 277199 The test_isa_2_06_part1.c in none/tests/ppc64 should be a symlink 1125 277471 Unhandled syscall: 340 1126 277610 valgrind crashes in VG_(lseek)(core_fd, phdrs[idx].p_offset, ...) 1127 277653 ARM: support Thumb2 PLD instruction 1128 277663 ARM: NEON float VMUL by scalar incorrect 1129 277689 ARM: tests for VSTn with register post-index are broken 1130 277694 ARM: BLX LR instruction broken in ARM mode 1131 277780 ARM: VMOV.F32 (immediate) instruction is broken 1132 278057 fuse filesystem syscall deadlocks 1133 278078 Unimplemented syscall 280 on ppc32 1134 278349 F_GETPIPE_SZ and F_SETPIPE_SZ Linux fcntl commands 1135 278454 VALGRIND_STACK_DEREGISTER has wrong output type 1136 278502 == 275284 (Valgrind confuses memcpy() and memmove()) 1137 278892 gdbsrv: factorize gdb version handling, fix doc and typos 1138 279027 Support for MVCL and CLCL instruction 1139 279027 s390x: Provide support for CLCL and MVCL instructions 1140 279062 Remove a redundant check in the insn selector for ppc. 1141 279071 JDK creates PTEST with redundant REX.W prefix 1142 279212 gdbsrv: add monitor cmd v.info scheduler. 1143 279378 exp-ptrcheck: the 'impossible' happened on mkfifo call 1144 279698 memcheck discards valid-bits for packuswb 1145 279795 memcheck reports uninitialised values for mincore on amd64 1146 279994 Add support for IBM Power ISA 2.06 -- stage 3 1147 280083 mempolicy syscall check errors 1148 280290 vex amd64->IR: 0x66 0xF 0x38 0x28 0xC1 0x66 0xF 0x6F 1149 280710 s390x: config files for nightly builds 1150 280757 /tmp dir still used by valgrind even if TMPDIR is specified 1151 280965 Valgrind breaks fcntl locks when program does mmap 1152 281138 WARNING: unhandled syscall: 340 1153 281241 == 275168 (valgrind useless on Macos 10.7.1 Lion) 1154 281304 == 275168 (Darwin: dyld "cannot load inserted library") 1155 281305 == 275168 (unhandled syscall: unix:357 on Darwin 11.1) 1156 281468 s390x: handle do_clone and gcc clones in call traces 1157 281488 ARM: VFP register corruption 1158 281828 == 275284 (false memmove warning: "Source and destination overlap") 1159 281883 s390x: Fix system call wrapper for "clone". 1160 282105 generalise 'reclaimSuperBlock' to also reclaim splittable superblock 1161 282112 Unhandled instruction bytes: 0xDE 0xD9 0x9B 0xDF (fcompp) 1162 282238 SLES10: make check fails 1163 282979 strcasestr needs replacement with recent(>=2.12) glibc 1164 283000 vex amd64->IR: 0x66 0xF 0x3A 0xA 0xC0 0x9 0xF3 0xF 1165 283243 Regression in ppc64 memcheck tests 1166 283325 == 267997 (Darwin: V segfaults on startup when built with Xcode 4.0) 1167 283427 re-connect epoll_pwait syscall on ARM linux 1168 283600 gdbsrv: android: port vgdb.c 1169 283709 none/tests/faultstatus needs to account for page size 1170 284305 filter_gdb needs enhancement to work on ppc64 1171 284384 clang 3.1 -Wunused-value warnings in valgrind.h, memcheck.h 1172 284472 Thumb2 ROR.W encoding T2 not implemented 1173 284621 XML-escape process command line in XML output 1174 n-i-bz cachegrind/callgrind: handle CPUID information for Core iX Intel CPUs 1175 that have non-power-of-2 sizes (also AMDs) 1176 n-i-bz don't be spooked by libraries mashed by elfhack 1177 n-i-bz don't be spooked by libxul.so linked with gold 1178 n-i-bz improved checking for VALGRIND_CHECK_MEM_IS_DEFINED 1179 1180 (3.7.0-TEST1: 27 October 2011, vex r2228, valgrind r12245) 1181 (3.7.0.RC1: 1 November 2011, vex r2231, valgrind r12257) 1182 (3.7.0: 5 November 2011, vex r2231, valgrind r12258) 1183 1184 1185 1186 Release 3.6.1 (16 February 2011) 1187 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1188 3.6.1 is a bug fix release. It adds support for some SSE4 1189 instructions that were omitted in 3.6.0 due to lack of time. Initial 1190 support for glibc-2.13 has been added. A number of bugs causing 1191 crashing or assertion failures have been fixed. 1192 1193 The following bugs have been fixed or resolved. Note that "n-i-bz" 1194 stands for "not in bugzilla" -- that is, a bug that was reported to us 1195 but never got a bugzilla entry. We encourage you to file bugs in 1196 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than 1197 mailing the developers (or mailing lists) directly -- bugs that are 1198 not entered into bugzilla tend to get forgotten about or ignored. 1199 1200 To see details of a given bug, visit 1201 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 1202 where XXXXXX is the bug number as listed below. 1203 1204 188572 Valgrind on Mac should suppress setenv() mem leak 1205 194402 vex amd64->IR: 0x48 0xF 0xAE 0x4 (proper FX{SAVE,RSTOR} support) 1206 210481 vex amd64->IR: Assertion `sz == 2 || sz == 4' failed (REX.W POPQ) 1207 246152 callgrind internal error after pthread_cancel on 32 Bit Linux 1208 250038 ppc64: Altivec LVSR and LVSL instructions fail their regtest 1209 254420 memory pool tracking broken 1210 254957 Test code failing to compile due to changes in memcheck.h 1211 255009 helgrind/drd: crash on chmod with invalid parameter 1212 255130 readdwarf3.c parse_type_DIE confused by GNAT Ada types 1213 255355 helgrind/drd: crash on threaded programs doing fork 1214 255358 == 255355 1215 255418 (SSE4.x) rint call compiled with ICC 1216 255822 --gen-suppressions can create invalid files: "too many callers [...]" 1217 255888 closing valgrindoutput tag outputted to log-stream on error 1218 255963 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x9 0xDB 0x0 (ROUNDPD) 1219 255966 Slowness when using mempool annotations 1220 256387 vex x86->IR: 0xD4 0xA 0x2 0x7 (AAD and AAM) 1221 256600 super-optimized strcasecmp() false positive 1222 256669 vex amd64->IR: Unhandled LOOPNEL insn on amd64 1223 256968 (SSE4.x) vex amd64->IR: 0x66 0xF 0x38 0x10 0xD3 0x66 (BLENDVPx) 1224 257011 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0xE 0xFD 0xA0 (PBLENDW) 1225 257063 (SSE4.x) vex amd64->IR: 0x66 0xF 0x3A 0x8 0xC0 0x0 (ROUNDPS) 1226 257276 Missing case in memcheck --track-origins=yes 1227 258870 (SSE4.x) Add support for EXTRACTPS SSE 4.1 instruction 1228 261966 (SSE4.x) support for CRC32B and CRC32Q is lacking (also CRC32{W,L}) 1229 262985 VEX regression in valgrind 3.6.0 in handling PowerPC VMX 1230 262995 (SSE4.x) crash when trying to valgrind gcc-snapshot (PCMPxSTRx $0) 1231 263099 callgrind_annotate counts Ir improperly [...] 1232 263877 undefined coprocessor instruction on ARMv7 1233 265964 configure FAIL with glibc-2.13 1234 n-i-bz Fix compile error w/ icc-12.x in guest_arm_toIR.c 1235 n-i-bz Docs: fix bogus descriptions for VALGRIND_CREATE_BLOCK et al 1236 n-i-bz Massif: don't assert on shmat() with --pages-as-heap=yes 1237 n-i-bz Bug fixes and major speedups for the exp-DHAT space profiler 1238 n-i-bz DRD: disable --free-is-write due to implementation difficulties 1239 1240 (3.6.1: 16 February 2011, vex r2103, valgrind r11561). 1241 1242 1243 1244 Release 3.6.0 (21 October 2010) 1245 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1246 3.6.0 is a feature release with many significant improvements and the 1247 usual collection of bug fixes. 1248 1249 This release supports X86/Linux, AMD64/Linux, ARM/Linux, PPC32/Linux, 1250 PPC64/Linux, X86/Darwin and AMD64/Darwin. Support for recent distros 1251 and toolchain components (glibc 2.12, gcc 4.5, OSX 10.6) has been added. 1252 1253 ------------------------- 1254 1255 Here are some highlights. Details are shown further down: 1256 1257 * Support for ARM/Linux. 1258 1259 * Support for recent Linux distros: Ubuntu 10.10 and Fedora 14. 1260 1261 * Support for Mac OS X 10.6, both 32- and 64-bit executables. 1262 1263 * Support for the SSE4.2 instruction set. 1264 1265 * Enhancements to the Callgrind profiler, including the ability to 1266 handle CPUs with three levels of cache. 1267 1268 * A new experimental heap profiler, DHAT. 1269 1270 * A huge number of bug fixes and small enhancements. 1271 1272 ------------------------- 1273 1274 Here are details of the above changes, together with descriptions of 1275 many other changes, and a list of fixed bugs. 1276 1277 * ================== PLATFORM CHANGES ================= 1278 1279 * Support for ARM/Linux. Valgrind now runs on ARMv7 capable CPUs 1280 running Linux. It is known to work on Ubuntu 10.04, Ubuntu 10.10, 1281 and Maemo 5, so you can run Valgrind on your Nokia N900 if you want. 1282 1283 This requires a CPU capable of running the ARMv7-A instruction set 1284 (Cortex A5, A8 and A9). Valgrind provides fairly complete coverage 1285 of the user space instruction set, including ARM and Thumb integer 1286 code, VFPv3, NEON and V6 media instructions. The Memcheck, 1287 Cachegrind and Massif tools work properly; other tools work to 1288 varying degrees. 1289 1290 * Support for recent Linux distros (Ubuntu 10.10 and Fedora 14), along 1291 with support for recent releases of the underlying toolchain 1292 components, notably gcc-4.5 and glibc-2.12. 1293 1294 * Support for Mac OS X 10.6, both 32- and 64-bit executables. 64-bit 1295 support also works much better on OS X 10.5, and is as solid as 1296 32-bit support now. 1297 1298 * Support for the SSE4.2 instruction set. SSE4.2 is supported in 1299 64-bit mode. In 32-bit mode, support is only available up to and 1300 including SSSE3. Some exceptions: SSE4.2 AES instructions are not 1301 supported in 64-bit mode, and 32-bit mode does in fact support the 1302 bare minimum SSE4 instructions to needed to run programs on Mac OS X 1303 10.6 on 32-bit targets. 1304 1305 * Support for IBM POWER6 cpus has been improved. The Power ISA up to 1306 and including version 2.05 is supported. 1307 1308 * ==================== TOOL CHANGES ==================== 1309 1310 * Cachegrind has a new processing script, cg_diff, which finds the 1311 difference between two profiles. It's very useful for evaluating 1312 the performance effects of a change in a program. 1313 1314 Related to this change, the meaning of cg_annotate's (rarely-used) 1315 --threshold option has changed; this is unlikely to affect many 1316 people, if you do use it please see the user manual for details. 1317 1318 * Callgrind now can do branch prediction simulation, similar to 1319 Cachegrind. In addition, it optionally can count the number of 1320 executed global bus events. Both can be used for a better 1321 approximation of a "Cycle Estimation" as derived event (you need to 1322 update the event formula in KCachegrind yourself). 1323 1324 * Cachegrind and Callgrind now refer to the LL (last-level) cache 1325 rather than the L2 cache. This is to accommodate machines with 1326 three levels of caches -- if Cachegrind/Callgrind auto-detects the 1327 cache configuration of such a machine it will run the simulation as 1328 if the L2 cache isn't present. This means the results are less 1329 likely to match the true result for the machine, but 1330 Cachegrind/Callgrind's results are already only approximate, and 1331 should not be considered authoritative. The results are still 1332 useful for giving a general idea about a program's locality. 1333 1334 * Massif has a new option, --pages-as-heap, which is disabled by 1335 default. When enabled, instead of tracking allocations at the level 1336 of heap blocks (as allocated with malloc/new/new[]), it instead 1337 tracks memory allocations at the level of memory pages (as mapped by 1338 mmap, brk, etc). Each mapped page is treated as its own block. 1339 Interpreting the page-level output is harder than the heap-level 1340 output, but this option is useful if you want to account for every 1341 byte of memory used by a program. 1342 1343 * DRD has two new command-line options: --free-is-write and 1344 --trace-alloc. The former allows to detect reading from already freed 1345 memory, and the latter allows tracing of all memory allocations and 1346 deallocations. 1347 1348 * DRD has several new annotations. Custom barrier implementations can 1349 now be annotated, as well as benign races on static variables. 1350 1351 * DRD's happens before / happens after annotations have been made more 1352 powerful, so that they can now also be used to annotate e.g. a smart 1353 pointer implementation. 1354 1355 * Helgrind's annotation set has also been drastically improved, so as 1356 to provide to users a general set of annotations to describe locks, 1357 semaphores, barriers and condition variables. Annotations to 1358 describe thread-safe reference counted heap objects have also been 1359 added. 1360 1361 * Memcheck has a new command-line option, --show-possibly-lost, which 1362 is enabled by default. When disabled, the leak detector will not 1363 show possibly-lost blocks. 1364 1365 * A new experimental heap profiler, DHAT (Dynamic Heap Analysis Tool), 1366 has been added. DHAT keeps track of allocated heap blocks, and also 1367 inspects every memory reference to see which block (if any) is being 1368 accessed. This gives a lot of insight into block lifetimes, 1369 utilisation, turnover, liveness, and the location of hot and cold 1370 fields. You can use DHAT to do hot-field profiling. 1371 1372 * ==================== OTHER CHANGES ==================== 1373 1374 * Improved support for unfriendly self-modifying code: the extra 1375 overhead incurred by --smc-check=all has been reduced by 1376 approximately a factor of 5 as compared with 3.5.0. 1377 1378 * Ability to show directory names for source files in error messages. 1379 This is combined with a flexible mechanism for specifying which 1380 parts of the paths should be shown. This is enabled by the new flag 1381 --fullpath-after. 1382 1383 * A new flag, --require-text-symbol, which will stop the run if a 1384 specified symbol is not found it a given shared object when it is 1385 loaded into the process. This makes advanced working with function 1386 intercepting and wrapping safer and more reliable. 1387 1388 * Improved support for the Valkyrie GUI, version 2.0.0. GUI output 1389 and control of Valgrind is now available for the tools Memcheck and 1390 Helgrind. XML output from Valgrind is available for Memcheck, 1391 Helgrind and exp-Ptrcheck. 1392 1393 * More reliable stack unwinding on amd64-linux, particularly in the 1394 presence of function wrappers, and with gcc-4.5 compiled code. 1395 1396 * Modest scalability (performance improvements) for massive 1397 long-running applications, particularly for those with huge amounts 1398 of code. 1399 1400 * Support for analyzing programs running under Wine with has been 1401 improved. The header files <valgrind/valgrind.h>, 1402 <valgrind/memcheck.h> and <valgrind/drd.h> can now be used in 1403 Windows-programs compiled with MinGW or one of the Microsoft Visual 1404 Studio compilers. 1405 1406 * A rare but serious error in the 64-bit x86 CPU simulation was fixed. 1407 The 32-bit simulator was not affected. This did not occur often, 1408 but when it did would usually crash the program under test. 1409 Bug 245925. 1410 1411 * A large number of bugs were fixed. These are shown below. 1412 1413 * A number of bugs were investigated, and were candidates for fixing, 1414 but are not fixed in 3.6.0, due to lack of developer time. They may 1415 get fixed in later releases. They are: 1416 1417 194402 vex amd64->IR: 0x48 0xF 0xAE 0x4 0x24 0x49 (FXSAVE64) 1418 212419 false positive "lock order violated" (A+B vs A) 1419 213685 Undefined value propagates past dependency breaking instruction 1420 216837 Incorrect instrumentation of NSOperationQueue on Darwin 1421 237920 valgrind segfault on fork failure 1422 242137 support for code compiled by LLVM-2.8 1423 242423 Another unknown Intel cache config value 1424 243232 Inconsistent Lock Orderings report with trylock 1425 243483 ppc: callgrind triggers VEX assertion failure 1426 243935 Helgrind: implementation of ANNOTATE_HAPPENS_BEFORE() is wrong 1427 244677 Helgrind crash hg_main.c:616 (map_threads_lookup): Assertion 1428 'thr' failed. 1429 246152 callgrind internal error after pthread_cancel on 32 Bit Linux 1430 249435 Analyzing wine programs with callgrind triggers a crash 1431 250038 ppc64: Altivec lvsr and lvsl instructions fail their regtest 1432 250065 Handling large allocations 1433 250101 huge "free" memory usage due to m_mallocfree.c 1434 "superblocks fragmentation" 1435 251569 vex amd64->IR: 0xF 0x1 0xF9 0x8B 0x4C 0x24 (RDTSCP) 1436 252091 Callgrind on ARM does not detect function returns correctly 1437 252600 [PATCH] Allow lhs to be a pointer for shl/shr 1438 254420 memory pool tracking broken 1439 n-i-bz support for adding symbols for JIT generated code 1440 1441 1442 The following bugs have been fixed or resolved. Note that "n-i-bz" 1443 stands for "not in bugzilla" -- that is, a bug that was reported to us 1444 but never got a bugzilla entry. We encourage you to file bugs in 1445 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than 1446 mailing the developers (or mailing lists) directly -- bugs that are 1447 not entered into bugzilla tend to get forgotten about or ignored. 1448 1449 To see details of a given bug, visit 1450 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 1451 where XXXXXX is the bug number as listed below. 1452 1453 135264 dcbzl instruction missing 1454 142688 == 250799 1455 153699 Valgrind should report unaligned reads with movdqa 1456 180217 == 212335 1457 190429 Valgrind reports lost of errors in ld.so 1458 with x86_64 2.9.90 glibc 1459 197266 valgrind appears to choke on the xmms instruction 1460 "roundsd" on x86_64 1461 197988 Crash when demangling very large symbol names 1462 202315 unhandled syscall: 332 (inotify_init1) 1463 203256 Add page-level profiling to Massif 1464 205093 dsymutil=yes needs quotes, locking (partial fix) 1465 205241 Snow Leopard 10.6 support (partial fix) 1466 206600 Leak checker fails to upgrade indirect blocks when their 1467 parent becomes reachable 1468 210935 port valgrind.h (not valgrind) to win32 so apps run under 1469 wine can make client requests 1470 211410 vex amd64->IR: 0x15 0xFF 0xFF 0x0 0x0 0x89 1471 within Linux ip-stack checksum functions 1472 212335 unhandled instruction bytes: 0xF3 0xF 0xBD 0xC0 1473 (lzcnt %eax,%eax) 1474 213685 Undefined value propagates past dependency breaking instruction 1475 (partial fix) 1476 215914 Valgrind inserts bogus empty environment variable 1477 217863 == 197988 1478 219538 adjtimex syscall wrapper wrong in readonly adjtime mode 1479 222545 shmat fails under valgind on some arm targets 1480 222560 ARM NEON support 1481 230407 == 202315 1482 231076 == 202315 1483 232509 Docs build fails with formatting inside <title></title> elements 1484 232793 == 202315 1485 235642 [PATCH] syswrap-linux.c: support evdev EVIOCG* ioctls 1486 236546 vex x86->IR: 0x66 0xF 0x3A 0xA 1487 237202 vex amd64->IR: 0xF3 0xF 0xB8 0xC0 0x49 0x3B 1488 237371 better support for VALGRIND_MALLOCLIKE_BLOCK 1489 237485 symlink (syscall 57) is not supported on Mac OS 1490 237723 sysno == 101 exp-ptrcheck: the 'impossible' happened: 1491 unhandled syscall 1492 238208 is_just_below_ESP doesn't take into account red-zone 1493 238345 valgrind passes wrong $0 when executing a shell script 1494 238679 mq_timedreceive syscall doesn't flag the reception buffer 1495 as "defined" 1496 238696 fcntl command F_DUPFD_CLOEXEC not supported 1497 238713 unhandled instruction bytes: 0x66 0xF 0x29 0xC6 1498 238713 unhandled instruction bytes: 0x66 0xF 0x29 0xC6 1499 238745 3.5.0 Make fails on PPC Altivec opcodes, though configure 1500 says "Altivec off" 1501 239992 vex amd64->IR: 0x48 0xF 0xC4 0xC1 0x0 0x48 1502 240488 == 197988 1503 240639 == 212335 1504 241377 == 236546 1505 241903 == 202315 1506 241920 == 212335 1507 242606 unhandled syscall: setegid (in Ptrcheck) 1508 242814 Helgrind "Impossible has happened" during 1509 QApplication::initInstance(); 1510 243064 Valgrind attempting to read debug information from iso 1511 243270 Make stack unwinding in Valgrind wrappers more reliable 1512 243884 exp-ptrcheck: the 'impossible happened: unhandled syscall 1513 sysno = 277 (mq_open) 1514 244009 exp-ptrcheck unknown syscalls in analyzing lighttpd 1515 244493 ARM VFP d16-d31 registers support 1516 244670 add support for audit_session_self syscall on Mac OS 10.6 1517 244921 The xml report of helgrind tool is not well format 1518 244923 In the xml report file, the <preamble> not escape the 1519 xml char, eg '<','&','>' 1520 245535 print full path names in plain text reports 1521 245925 x86-64 red zone handling problem 1522 246258 Valgrind not catching integer underruns + new [] s 1523 246311 reg/reg cmpxchg doesn't work on amd64 1524 246549 unhandled syscall unix:277 while testing 32-bit Darwin app 1525 246888 Improve Makefile.vex.am 1526 247510 [OS X 10.6] Memcheck reports unaddressable bytes passed 1527 to [f]chmod_extended 1528 247526 IBM POWER6 (ISA 2.05) support is incomplete 1529 247561 Some leak testcases fails due to reachable addresses in 1530 caller save regs 1531 247875 sizeofIRType to handle Ity_I128 1532 247894 [PATCH] unhandled syscall sys_readahead 1533 247980 Doesn't honor CFLAGS passed to configure 1534 248373 darwin10.supp is empty in the trunk 1535 248822 Linux FIBMAP ioctl has int parameter instead of long 1536 248893 [PATCH] make readdwarf.c big endianess safe to enable 1537 unwinding on big endian systems 1538 249224 Syscall 336 not supported (SYS_proc_info) 1539 249359 == 245535 1540 249775 Incorrect scheme for detecting NEON capabilities of host CPU 1541 249943 jni JVM init fails when using valgrind 1542 249991 Valgrind incorrectly declares AESKEYGENASSIST support 1543 since VEX r2011 1544 249996 linux/arm: unhandled syscall: 181 (__NR_pwrite64) 1545 250799 frexp$fenv_access_off function generates SIGILL 1546 250998 vex x86->IR: unhandled instruction bytes: 0x66 0x66 0x66 0x2E 1547 251251 support pclmulqdq insn 1548 251362 valgrind: ARM: attach to debugger either fails or provokes 1549 kernel oops 1550 251674 Unhandled syscall 294 1551 251818 == 254550 1552 1553 254257 Add support for debugfiles found by build-id 1554 254550 [PATCH] Implement DW_ATE_UTF (DWARF4) 1555 254646 Wrapped functions cause stack misalignment on OS X 1556 (and possibly Linux) 1557 254556 ARM: valgrinding anything fails with SIGSEGV for 0xFFFF0FA0 1558 1559 (3.6.0: 21 October 2010, vex r2068, valgrind r11471). 1560 1561 1562 1563 Release 3.5.0 (19 August 2009) 1564 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1565 3.5.0 is a feature release with many significant improvements and the 1566 usual collection of bug fixes. The main improvement is that Valgrind 1567 now works on Mac OS X. 1568 1569 This release supports X86/Linux, AMD64/Linux, PPC32/Linux, PPC64/Linux 1570 and X86/Darwin. Support for recent distros and toolchain components 1571 (glibc 2.10, gcc 4.5) has been added. 1572 1573 ------------------------- 1574 1575 Here is a short summary of the changes. Details are shown further 1576 down: 1577 1578 * Support for Mac OS X (10.5.x). 1579 1580 * Improvements and simplifications to Memcheck's leak checker. 1581 1582 * Clarification and simplifications in various aspects of Valgrind's 1583 text output. 1584 1585 * XML output for Helgrind and Ptrcheck. 1586 1587 * Performance and stability improvements for Helgrind and DRD. 1588 1589 * Genuinely atomic support for x86/amd64/ppc atomic instructions. 1590 1591 * A new experimental tool, BBV, useful for computer architecture 1592 research. 1593 1594 * Improved Wine support, including ability to read Windows PDB 1595 debuginfo. 1596 1597 ------------------------- 1598 1599 Here are details of the above changes, followed by descriptions of 1600 many other minor changes, and a list of fixed bugs. 1601 1602 1603 * Valgrind now runs on Mac OS X. (Note that Mac OS X is sometimes 1604 called "Darwin" because that is the name of the OS core, which is the 1605 level that Valgrind works at.) 1606 1607 Supported systems: 1608 1609 - It requires OS 10.5.x (Leopard). Porting to 10.4.x is not planned 1610 because it would require work and 10.4 is only becoming less common. 1611 1612 - 32-bit programs on x86 and AMD64 (a.k.a x86-64) machines are supported 1613 fairly well. For 10.5.x, 32-bit programs are the default even on 1614 64-bit machines, so it handles most current programs. 1615 1616 - 64-bit programs on x86 and AMD64 (a.k.a x86-64) machines are not 1617 officially supported, but simple programs at least will probably work. 1618 However, start-up is slow. 1619 1620 - PowerPC machines are not supported. 1621 1622 Things that don't work: 1623 1624 - The Ptrcheck tool. 1625 1626 - Objective-C garbage collection. 1627 1628 - --db-attach=yes. 1629 1630 - If you have Rogue Amoeba's "Instant Hijack" program installed, 1631 Valgrind will fail with a SIGTRAP at start-up. See 1632 https://bugs.kde.org/show_bug.cgi?id=193917 for details and a 1633 simple work-around. 1634 1635 Usage notes: 1636 1637 - You will likely find --dsymutil=yes a useful option, as error 1638 messages may be imprecise without it. 1639 1640 - Mac OS X support is new and therefore will be less robust than the 1641 Linux support. Please report any bugs you find. 1642 1643 - Threaded programs may run more slowly than on Linux. 1644 1645 Many thanks to Greg Parker for developing this port over several years. 1646 1647 1648 * Memcheck's leak checker has been improved. 1649 1650 - The results for --leak-check=summary now match the summary results 1651 for --leak-check=full. Previously they could differ because 1652 --leak-check=summary counted "indirectly lost" blocks and 1653 "suppressed" blocks as "definitely lost". 1654 1655 - Blocks that are only reachable via at least one interior-pointer, 1656 but are directly pointed to by a start-pointer, were previously 1657 marked as "still reachable". They are now correctly marked as 1658 "possibly lost". 1659 1660 - The default value for the --leak-resolution option has been 1661 changed from "low" to "high". In general, this means that more 1662 leak reports will be produced, but each leak report will describe 1663 fewer leaked blocks. 1664 1665 - With --leak-check=full, "definitely lost" and "possibly lost" 1666 leaks are now considered as proper errors, ie. they are counted 1667 for the "ERROR SUMMARY" and affect the behaviour of 1668 --error-exitcode. These leaks are not counted as errors if 1669 --leak-check=summary is specified, however. 1670 1671 - Documentation for the leak checker has been improved. 1672 1673 1674 * Various aspects of Valgrind's text output have changed. 1675 1676 - Valgrind's start-up message has changed. It is shorter but also 1677 includes the command being run, which makes it easier to use 1678 --trace-children=yes. An example: 1679 1680 - Valgrind's shut-down messages have also changed. This is most 1681 noticeable with Memcheck, where the leak summary now occurs before 1682 the error summary. This change was necessary to allow leaks to be 1683 counted as proper errors (see the description of the leak checker 1684 changes above for more details). This was also necessary to fix a 1685 longstanding bug in which uses of suppressions against leaks were 1686 not "counted", leading to difficulties in maintaining suppression 1687 files (see https://bugs.kde.org/show_bug.cgi?id=186790). 1688 1689 - Behavior of -v has changed. In previous versions, -v printed out 1690 a mixture of marginally-user-useful information, and tool/core 1691 statistics. The statistics printing has now been moved to its own 1692 flag, --stats=yes. This means -v is less verbose and more likely 1693 to convey useful end-user information. 1694 1695 - The format of some (non-XML) stack trace entries has changed a 1696 little. Previously there were six possible forms: 1697 1698 0x80483BF: really (a.c:20) 1699 0x80483BF: really (in /foo/a.out) 1700 0x80483BF: really 1701 0x80483BF: (within /foo/a.out) 1702 0x80483BF: ??? (a.c:20) 1703 0x80483BF: ??? 1704 1705 The third and fourth of these forms have been made more consistent 1706 with the others. The six possible forms are now: 1707 1708 0x80483BF: really (a.c:20) 1709 0x80483BF: really (in /foo/a.out) 1710 0x80483BF: really (in ???) 1711 0x80483BF: ??? (in /foo/a.out) 1712 0x80483BF: ??? (a.c:20) 1713 0x80483BF: ??? 1714 1715 Stack traces produced when --xml=yes is specified are different 1716 and unchanged. 1717 1718 1719 * Helgrind and Ptrcheck now support XML output, so they can be used 1720 from GUI tools. Also, the XML output mechanism has been 1721 overhauled. 1722 1723 - The XML format has been overhauled and generalised, so it is more 1724 suitable for error reporting tools in general. The Memcheck 1725 specific aspects of it have been removed. The new format, which 1726 is an evolution of the old format, is described in 1727 docs/internals/xml-output-protocol4.txt. 1728 1729 - Memcheck has been updated to use the new format. 1730 1731 - Helgrind and Ptrcheck are now able to emit output in this format. 1732 1733 - The XML output mechanism has been overhauled. XML is now output 1734 to its own file descriptor, which means that: 1735 1736 * Valgrind can output text and XML independently. 1737 1738 * The longstanding problem of XML output being corrupted by 1739 unexpected un-tagged text messages is solved. 1740 1741 As before, the destination for text output is specified using 1742 --log-file=, --log-fd= or --log-socket=. 1743 1744 As before, XML output for a tool is enabled using --xml=yes. 1745 1746 Because there's a new XML output channel, the XML output 1747 destination is now specified by --xml-file=, --xml-fd= or 1748 --xml-socket=. 1749 1750 Initial feedback has shown this causes some confusion. To 1751 clarify, the two envisaged usage scenarios are: 1752 1753 (1) Normal text output. In this case, do not specify --xml=yes 1754 nor any of --xml-file=, --xml-fd= or --xml-socket=. 1755 1756 (2) XML output. In this case, specify --xml=yes, and one of 1757 --xml-file=, --xml-fd= or --xml-socket= to select the XML 1758 destination, one of --log-file=, --log-fd= or --log-socket= 1759 to select the destination for any remaining text messages, 1760 and, importantly, -q. 1761 1762 -q makes Valgrind completely silent on the text channel, 1763 except in the case of critical failures, such as Valgrind 1764 itself segfaulting, or failing to read debugging information. 1765 Hence, in this scenario, it suffices to check whether or not 1766 any output appeared on the text channel. If yes, then it is 1767 likely to be a critical error which should be brought to the 1768 attention of the user. If no (the text channel produced no 1769 output) then it can be assumed that the run was successful. 1770 1771 This allows GUIs to make the critical distinction they need to 1772 make (did the run fail or not?) without having to search or 1773 filter the text output channel in any way. 1774 1775 It is also recommended to use --child-silent-after-fork=yes in 1776 scenario (2). 1777 1778 1779 * Improvements and changes in Helgrind: 1780 1781 - XML output, as described above 1782 1783 - Checks for consistent association between pthread condition 1784 variables and their associated mutexes are now performed. 1785 1786 - pthread_spinlock functions are supported. 1787 1788 - Modest performance improvements. 1789 1790 - Initial (skeletal) support for describing the behaviour of 1791 non-POSIX synchronisation objects through ThreadSanitizer 1792 compatible ANNOTATE_* macros. 1793 1794 - More controllable tradeoffs between performance and the level of 1795 detail of "previous" accesses in a race. There are now three 1796 settings: 1797 1798 * --history-level=full. This is the default, and was also the 1799 default in 3.4.x. It shows both stacks involved in a race, but 1800 requires a lot of memory and can be very slow in programs that 1801 do many inter-thread synchronisation events. 1802 1803 * --history-level=none. This only shows the later stack involved 1804 in a race. This can be much faster than --history-level=full, 1805 but makes it much more difficult to find the other access 1806 involved in the race. 1807 1808 The new intermediate setting is 1809 1810 * --history-level=approx 1811 1812 For the earlier (other) access, two stacks are presented. The 1813 earlier access is guaranteed to be somewhere in between the two 1814 program points denoted by those stacks. This is not as useful 1815 as showing the exact stack for the previous access (as per 1816 --history-level=full), but it is better than nothing, and it's 1817 almost as fast as --history-level=none. 1818 1819 1820 * New features and improvements in DRD: 1821 1822 - The error messages printed by DRD are now easier to interpret. 1823 Instead of using two different numbers to identify each thread 1824 (Valgrind thread ID and DRD thread ID), DRD does now identify 1825 threads via a single number (the DRD thread ID). Furthermore 1826 "first observed at" information is now printed for all error 1827 messages related to synchronization objects. 1828 1829 - Added support for named semaphores (sem_open() and sem_close()). 1830 1831 - Race conditions between pthread_barrier_wait() and 1832 pthread_barrier_destroy() calls are now reported. 1833 1834 - Added support for custom allocators through the macros 1835 VALGRIND_MALLOCLIKE_BLOCK() VALGRIND_FREELIKE_BLOCK() (defined in 1836 in <valgrind/valgrind.h>). An alternative for these two macros is 1837 the new client request VG_USERREQ__DRD_CLEAN_MEMORY (defined in 1838 <valgrind/drd.h>). 1839 1840 - Added support for annotating non-POSIX synchronization objects 1841 through several new ANNOTATE_*() macros. 1842 1843 - OpenMP: added support for the OpenMP runtime (libgomp) included 1844 with gcc versions 4.3.0 and 4.4.0. 1845 1846 - Faster operation. 1847 1848 - Added two new command-line options (--first-race-only and 1849 --segment-merging-interval). 1850 1851 1852 * Genuinely atomic support for x86/amd64/ppc atomic instructions 1853 1854 Valgrind will now preserve (memory-access) atomicity of LOCK- 1855 prefixed x86/amd64 instructions, and any others implying a global 1856 bus lock. Ditto for PowerPC l{w,d}arx/st{w,d}cx. instructions. 1857 1858 This means that Valgrinded processes will "play nicely" in 1859 situations where communication with other processes, or the kernel, 1860 is done through shared memory and coordinated with such atomic 1861 instructions. Prior to this change, such arrangements usually 1862 resulted in hangs, races or other synchronisation failures, because 1863 Valgrind did not honour atomicity of such instructions. 1864 1865 1866 * A new experimental tool, BBV, has been added. BBV generates basic 1867 block vectors for use with the SimPoint analysis tool, which allows 1868 a program's overall behaviour to be approximated by running only a 1869 fraction of it. This is useful for computer architecture 1870 researchers. You can run BBV by specifying --tool=exp-bbv (the 1871 "exp-" prefix is short for "experimental"). BBV was written by 1872 Vince Weaver. 1873 1874 1875 * Modestly improved support for running Windows applications under 1876 Wine. In particular, initial support for reading Windows .PDB debug 1877 information has been added. 1878 1879 1880 * A new Memcheck client request VALGRIND_COUNT_LEAK_BLOCKS has been 1881 added. It is similar to VALGRIND_COUNT_LEAKS but counts blocks 1882 instead of bytes. 1883 1884 1885 * The Valgrind client requests VALGRIND_PRINTF and 1886 VALGRIND_PRINTF_BACKTRACE have been changed slightly. Previously, 1887 the string was always printed immediately on its own line. Now, the 1888 string will be added to a buffer but not printed until a newline is 1889 encountered, or other Valgrind output is printed (note that for 1890 VALGRIND_PRINTF_BACKTRACE, the back-trace itself is considered 1891 "other Valgrind output"). This allows you to use multiple 1892 VALGRIND_PRINTF calls to build up a single output line, and also to 1893 print multiple output lines with a single request (by embedding 1894 multiple newlines in the string). 1895 1896 1897 * The graphs drawn by Massif's ms_print program have changed slightly: 1898 1899 - The half-height chars '.' and ',' are no longer drawn, because 1900 they are confusing. The --y option can be used if the default 1901 y-resolution is not high enough. 1902 1903 - Horizontal lines are now drawn after the top of a snapshot if 1904 there is a gap until the next snapshot. This makes it clear that 1905 the memory usage has not dropped to zero between snapshots. 1906 1907 1908 * Something that happened in 3.4.0, but wasn't clearly announced: the 1909 option --read-var-info=yes can be used by some tools (Memcheck, 1910 Helgrind and DRD). When enabled, it causes Valgrind to read DWARF3 1911 variable type and location information. This makes those tools 1912 start up more slowly and increases memory consumption, but 1913 descriptions of data addresses in error messages become more 1914 detailed. 1915 1916 1917 * exp-Omega, an experimental instantaneous leak-detecting tool, was 1918 disabled in 3.4.0 due to a lack of interest and maintenance, 1919 although the source code was still in the distribution. The source 1920 code has now been removed from the distribution. For anyone 1921 interested, the removal occurred in SVN revision r10247. 1922 1923 1924 * Some changes have been made to the build system. 1925 1926 - VEX/ is now integrated properly into the build system. This means 1927 that dependency tracking within VEX/ now works properly, "make 1928 install" will work without requiring "make" before it, and 1929 parallel builds (ie. 'make -j') now work (previously a 1930 .NOTPARALLEL directive was used to serialize builds, ie. 'make -j' 1931 was effectively ignored). 1932 1933 - The --with-vex configure option has been removed. It was of 1934 little use and removing it simplified the build system. 1935 1936 - The location of some install files has changed. This should not 1937 affect most users. Those who might be affected: 1938 1939 * For people who use Valgrind with MPI programs, the installed 1940 libmpiwrap.so library has moved from 1941 $(INSTALL)/<platform>/libmpiwrap.so to 1942 $(INSTALL)/libmpiwrap-<platform>.so. 1943 1944 * For people who distribute standalone Valgrind tools, the 1945 installed libraries such as $(INSTALL)/<platform>/libcoregrind.a 1946 have moved to $(INSTALL)/libcoregrind-<platform>.a. 1947 1948 These changes simplify the build system. 1949 1950 - Previously, all the distributed suppression (*.supp) files were 1951 installed. Now, only default.supp is installed. This should not 1952 affect users as the other installed suppression files were not 1953 read; the fact that they were installed was a mistake. 1954 1955 1956 * KNOWN LIMITATIONS: 1957 1958 - Memcheck is unusable with the Intel compiler suite version 11.1, 1959 when it generates code for SSE2-and-above capable targets. This 1960 is because of icc's use of highly optimised inlined strlen 1961 implementations. It causes Memcheck to report huge numbers of 1962 false errors even in simple programs. Helgrind and DRD may also 1963 have problems. 1964 1965 Versions 11.0 and earlier may be OK, but this has not been 1966 properly tested. 1967 1968 1969 The following bugs have been fixed or resolved. Note that "n-i-bz" 1970 stands for "not in bugzilla" -- that is, a bug that was reported to us 1971 but never got a bugzilla entry. We encourage you to file bugs in 1972 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than 1973 mailing the developers (or mailing lists) directly -- bugs that are 1974 not entered into bugzilla tend to get forgotten about or ignored. 1975 1976 To see details of a given bug, visit 1977 https://bugs.kde.org/show_bug.cgi?id=XXXXXX 1978 where XXXXXX is the bug number as listed below. 1979 1980 84303 How about a LockCheck tool? 1981 91633 dereference of null ptr in vgPlain_st_basetype 1982 97452 Valgrind doesn't report any pthreads problems 1983 100628 leak-check gets assertion failure when using 1984 VALGRIND_MALLOCLIKE_BLOCK on malloc()ed memory 1985 108528 NPTL pthread cleanup handlers not called 1986 110126 Valgrind 2.4.1 configure.in tramples CFLAGS 1987 110128 mallinfo is not implemented... 1988 110770 VEX: Generated files not always updated when making valgrind 1989 111102 Memcheck: problems with large (memory footprint) applications 1990 115673 Vex's decoder should never assert 1991 117564 False positive: Syscall param clone(child_tidptr) contains 1992 uninitialised byte(s) 1993 119404 executing ssh from inside valgrind fails 1994 133679 Callgrind does not write path names to sources with dwarf debug 1995 info 1996 135847 configure.in problem with non gnu compilers (and possible fix) 1997 136154 threads.c:273 (vgCallgrind_post_signal): Assertion 1998 '*(vgCallgrind_current_fn_stack.top) == 0' failed. 1999 136230 memcheck reports "possibly lost", should be "still reachable" 2000 137073 NULL arg to MALLOCLIKE_BLOCK causes crash 2001 137904 Valgrind reports a memory leak when using POSIX threads, 2002 while it shouldn't 2003 139076 valgrind VT_GETSTATE error 2004 142228 complaint of elf_dynamic_do_rela in trivial usage 2005 145347 spurious warning with USBDEVFS_REAPURB 2006 148441 (wine) can't find memory leak in Wine, win32 binary 2007 executable file. 2008 148742 Leak-check fails assert on exit 2009 149878 add (proper) check for calloc integer overflow 2010 150606 Call graph is broken when using callgrind control 2011 152393 leak errors produce an exit code of 0. I need some way to 2012 cause leak errors to result in a nonzero exit code. 2013 157154 documentation (leak-resolution doc speaks about num-callers 2014 def=4) + what is a loss record 2015 159501 incorrect handling of ALSA ioctls 2016 162020 Valgrinding an empty/zero-byte file crashes valgrind 2017 162482 ppc: Valgrind crashes while reading stabs information 2018 162718 x86: avoid segment selector 0 in sys_set_thread_area() 2019 163253 (wine) canonicaliseSymtab forgot some fields in DiSym 2020 163560 VEX/test_main.c is missing from valgrind-3.3.1 2021 164353 malloc_usable_size() doesn't return a usable size 2022 165468 Inconsistent formatting in memcheck manual -- please fix 2023 169505 main.c:286 (endOfInstr): 2024 Assertion 'ii->cost_offset == *cost_offset' failed 2025 177206 Generate default.supp during compile instead of configure 2026 177209 Configure valt_load_address based on arch+os 2027 177305 eventfd / syscall 323 patch lost 2028 179731 Tests fail to build because of inlining of non-local asm labels 2029 181394 helgrind: libhb_core.c:3762 (msm_write): Assertion 2030 'ordxx == POrd_EQ || ordxx == POrd_LT' failed. 2031 181594 Bogus warning for empty text segment 2032 181707 dwarf doesn't require enumerations to have name 2033 185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64 2034 185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree): 2035 Assertion '!already_present' failed. 2036 185359 exp-ptrcheck: unhandled syscall getresuid() 2037 185794 "WARNING: unhandled syscall: 285" (fallocate) on x86_64 2038 185816 Valgrind is unable to handle debug info for files with split 2039 debug info that are prelinked afterwards 2040 185980 [darwin] unhandled syscall: sem_open 2041 186238 bbToIR_AMD64: disInstr miscalculated next %rip 2042 186507 exp-ptrcheck unhandled syscalls prctl, etc. 2043 186790 Suppression pattern used for leaks are not reported 2044 186796 Symbols with length>200 in suppression files are ignored 2045 187048 drd: mutex PTHREAD_PROCESS_SHARED attribute missinterpretation 2046 187416 exp-ptrcheck: support for __NR_{setregid,setreuid,setresuid} 2047 188038 helgrind: hg_main.c:926: mk_SHVAL_fail: the 'impossible' happened 2048 188046 bashisms in the configure script 2049 188127 amd64->IR: unhandled instruction bytes: 0xF0 0xF 0xB0 0xA 2050 188161 memcheck: --track-origins=yes asserts "mc_machine.c:672 2051 (get_otrack_shadow_offset_wrk): the 'impossible' happened." 2052 188248 helgrind: pthread_cleanup_push, pthread_rwlock_unlock, 2053 assertion fail "!lock->heldBy" 2054 188427 Add support for epoll_create1 (with patch) 2055 188530 Support for SIOCGSTAMPNS 2056 188560 Include valgrind.spec in the tarball 2057 188572 Valgrind on Mac should suppress setenv() mem leak 2058 189054 Valgrind fails to build because of duplicate non-local asm labels 2059 189737 vex amd64->IR: unhandled instruction bytes: 0xAC 2060 189762 epoll_create syscall not handled (--tool=exp-ptrcheck) 2061 189763 drd assertion failure: s_threadinfo[tid].is_recording 2062 190219 unhandled syscall: 328 (x86-linux) 2063 190391 dup of 181394; see above 2064 190429 Valgrind reports lots of errors in ld.so with x86_64 2.9.90 glibc 2065 190820 No debug information on powerpc-linux 2066 191095 PATCH: Improve usbdevfs ioctl handling 2067 191182 memcheck: VALGRIND_LEAK_CHECK quadratic when big nr of chunks 2068 or big nr of errors 2069 191189 --xml=yes should obey --gen-suppressions=all 2070 191192 syslog() needs a suppression on macosx 2071 191271 DARWIN: WARNING: unhandled syscall: 33554697 a.k.a.: 265 2072 191761 getrlimit on MacOSX 2073 191992 multiple --fn-skip only works sometimes; dependent on order 2074 192634 V. reports "aspacem sync_check_mapping_callback: 2075 segment mismatch" on Darwin 2076 192954 __extension__ missing on 2 client requests 2077 194429 Crash at start-up with glibc-2.10.1 and linux-2.6.29 2078 194474 "INSTALL" file has different build instructions than "README" 2079 194671 Unhandled syscall (sem_wait?) from mac valgrind 2080 195069 memcheck: reports leak (memory still reachable) for 2081 printf("%d', x) 2082 195169 drd: (vgDrd_barrier_post_wait): 2083 Assertion 'r->sg[p->post_iteration]' failed. 2084 195268 valgrind --log-file doesn't accept ~/... 2085 195838 VEX abort: LibVEX_N_SPILL_BYTES too small for CPUID boilerplate 2086 195860 WARNING: unhandled syscall: unix:223 2087 196528 need a error suppression for pthread_rwlock_init under os x? 2088 197227 Support aio_* syscalls on Darwin 2089 197456 valgrind should reject --suppressions=(directory) 2090 197512 DWARF2 CFI reader: unhandled CFI instruction 0:10 2091 197591 unhandled syscall 27 (mincore) 2092 197793 Merge DCAS branch to the trunk == 85756, 142103 2093 197794 Avoid duplicate filenames in Vex 2094 197898 make check fails on current SVN 2095 197901 make check fails also under exp-ptrcheck in current SVN 2096 197929 Make --leak-resolution=high the default 2097 197930 Reduce spacing between leak reports 2098 197933 Print command line of client at start-up, and shorten preamble 2099 197966 unhandled syscall 205 (x86-linux, --tool=exp-ptrcheck) 2100 198395 add BBV to the distribution as an experimental tool 2101 198624 Missing syscalls on Darwin: 82, 167, 281, 347 2102 198649 callgrind_annotate doesn't cumulate counters 2103 199338 callgrind_annotate sorting/thresholds are broken for all but Ir 2104 199977 Valgrind complains about an unrecognized instruction in the 2105 atomic_incs test program 2106 200029 valgrind isn't able to read Fedora 12 debuginfo 2107 200760 darwin unhandled syscall: unix:284 2108 200827 DRD doesn't work on Mac OS X 2109 200990 VG_(read_millisecond_timer)() does not work correctly 2110 201016 Valgrind does not support pthread_kill() on Mac OS 2111 201169 Document --read-var-info 2112 201323 Pre-3.5.0 performance sanity checking 2113 201384 Review user manual for the 3.5.0 release 2114 201585 mfpvr not implemented on ppc 2115 201708 tests failing because x86 direction flag is left set 2116 201757 Valgrind doesn't handle any recent sys_futex additions 2117 204377 64-bit valgrind can not start a shell script 2118 (with #!/path/to/shell) if the shell is a 32-bit executable 2119 n-i-bz drd: fixed assertion failure triggered by mutex reinitialization. 2120 n-i-bz drd: fixed a bug that caused incorrect messages to be printed 2121 about memory allocation events with memory access tracing enabled 2122 n-i-bz drd: fixed a memory leak triggered by vector clock deallocation 2123 2124 (3.5.0: 19 Aug 2009, vex r1913, valgrind r10846). 2125 2126 2127 2128 Release 3.4.1 (28 February 2009) 2129 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2130 3.4.1 is a bug-fix release that fixes some regressions and assertion 2131 failures in debug info reading in 3.4.0, most notably incorrect stack 2132 traces on amd64-linux on older (glibc-2.3 based) systems. Various 2133 other debug info problems are also fixed. A number of bugs in the 2134 exp-ptrcheck tool introduced in 3.4.0 have been fixed. 2135 2136 In view of the fact that 3.4.0 contains user-visible regressions 2137 relative to 3.3.x, upgrading to 3.4.1 is recommended. Packagers are 2138 encouraged to ship 3.4.1 in preference to 3.4.0. 2139 2140 The fixed bugs are as follows. Note that "n-i-bz" stands for "not in 2141 bugzilla" -- that is, a bug that was reported to us but never got a 2142 bugzilla entry. We encourage you to file bugs in bugzilla 2143 (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the 2144 developers (or mailing lists) directly -- bugs that are not entered 2145 into bugzilla tend to get forgotten about or ignored. 2146 2147 n-i-bz Fix various bugs reading icc-11 generated debug info 2148 n-i-bz Fix various bugs reading gcc-4.4 generated debug info 2149 n-i-bz Preliminary support for glibc-2.10 / Fedora 11 2150 n-i-bz Cachegrind and Callgrind: handle non-power-of-two cache sizes, 2151 so as to support (eg) 24k Atom D1 and Core2 with 3/6/12MB L2. 2152 179618 exp-ptrcheck crashed / exit prematurely 2153 179624 helgrind: false positive races with pthread_create and 2154 recv/open/close/read 2155 134207 pkg-config output contains @VG_PLATFORM@ 2156 176926 floating point exception at valgrind startup with PPC 440EPX 2157 181594 Bogus warning for empty text segment 2158 173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant rex prefixes) 2159 181707 Dwarf3 doesn't require enumerations to have name 2160 185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64 2161 185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree): 2162 Assertion '!already_present' failed. 2163 185359 exp-ptrcheck unhandled syscall getresuid() 2164 2165 (3.4.1.RC1: 24 Feb 2008, vex r1884, valgrind r9253). 2166 (3.4.1: 28 Feb 2008, vex r1884, valgrind r9293). 2167 2168 2169 2170 Release 3.4.0 (2 January 2009) 2171 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 2172 3.4.0 is a feature release with many significant improvements and the 2173 usual collection of bug fixes. This release supports X86/Linux, 2174 AMD64/Linux, PPC32/Linux and PPC64/Linux. Support for recent distros 2175 (using gcc 4.4, glibc 2.8 and 2.9) has been added. 2176 2177 3.4.0 brings some significant tool improvements. Memcheck can now 2178 report the origin of uninitialised values, the thread checkers 2179 Helgrind and DRD are much improved, and we have a new experimental 2180 tool, exp-Ptrcheck, which is able to detect overruns of stack and 2181 global arrays. In detail: 2182 2183 * Memcheck is now able to track the origin of uninitialised values. 2184 When it reports an uninitialised value error, it will try to show 2185 the origin of the value, as either a heap or stack allocation. 2186 Origin tracking is expensive and so is not enabled by default. To 2187 use it, specify --track-origins=yes. Memcheck's speed will be 2188 essentially halved, and memory usage will be significantly 2189 increased. Nevertheless it can drastically reduce the effort 2190 required to identify the root cause of uninitialised value errors, 2191 and so is often a programmer productivity win, despite running more 2192 slowly. 2193 2194 * A version (1.4.0) of the Valkyrie GUI, that works with Memcheck in 2195 3.4.0, will be released shortly. 2196 2197 * Helgrind's race detection algorithm has been completely redesigned 2198 and reimplemented, to address usability and scalability concerns: 2199 2200 - The new algorithm has a lower false-error rate: it is much less 2201 likely to report races that do not really exist. 2202 2203 - Helgrind will display full call stacks for both accesses involved 2204 in a race. This makes it easier to identify the root causes of 2205 races. 2206 2207 - Limitations on the size of program that can run have been removed. 2208 2209 - Performance has been modestly improved, although that is very 2210 workload-dependent. 2211 2212 - Direct support for Qt4 threading has been added. 2213 2214 - pthread_barriers are now directly supported. 2215 2216 - Helgrind works well on all supported Linux targets. 2217 2218 * The DRD thread debugging tool has seen major improvements: 2219 2220 - Greatly improved performance and significantly reduced memory 2221 usage. 2222 2223 - Support for several major threading libraries (Boost.Thread, Qt4, 2224 glib, OpenMP) has been added. 2225 2226 - Support for atomic instructions, POSIX semaphores, barriers and 2227 reader-writer locks has been added. 2228 2229 - Works now on PowerPC CPUs too. 2230 2231 - Added support for printing thread stack usage at thread exit time. 2232 2233 - Added support for debugging lock contention. 2234 2235 - Added a manual for Drd. 2236 2237 * A new experimental tool, exp-Ptrcheck, has been added. Ptrcheck 2238 checks for misuses of pointers. In that sense it is a bit like 2239 Memcheck. However, Ptrcheck can do things Memcheck can't: it can 2240 detect overruns of stack and global arrays, it can detect 2241 arbitrarily far out-of-bounds accesses to heap blocks, and it can 2242 detect accesses heap blocks that have been freed a very long time 2243 ago (millions of blocks in the past). 2244 2245 Ptrcheck currently works only on x86-linux and amd64-linux. To use 2246 it, use --tool=exp-ptrcheck. A simple manual is provided, as part 2247 of the main Valgrind documentation. As this is an experimental 2248 tool, we would be particularly interested in hearing about your 2249 experiences with it. 2250 2251 * exp-Omega, an experimental instantaneous leak-detecting tool, is no 2252 longer built by default, although the code remains in the repository 2253 and the tarball. This is due to three factors: a perceived lack of 2254 users, a lack of maintenance, and concerns that it may not be 2255 possible to achieve reliable operation using the existing design. 2256 2257 * As usual, support for the latest Linux distros and toolchain 2258 components has been added. It should work well on Fedora Core 10, 2259 OpenSUSE 11.1 and Ubuntu 8.10. gcc-4.4 (in its current pre-release 2260 state) is supported, as is glibc-2.9. The C++ demangler has been 2261 updated so as to work well with C++ compiled by even the most recent 2262 g++'s. 2263 2264 * You can now use frame-level wildcards in suppressions. This was a 2265 frequently-requested enhancement. A line "..." in a suppression now 2266 matches zero or more frames. This makes it easier to write 2267 suppressions which are precise yet insensitive to changes in 2268 inlining behaviour. 2269 2270 * 3.4.0 adds support on x86/amd64 for the SSSE3 instruction set. 2271 2272 * Very basic support for IBM Power6 has been added (64-bit processes only). 2273 2274 * Valgrind is now cross-compilable. For example, it is possible to 2275 cross compile Valgrind on an x86/amd64-linux host, so that it runs 2276 on a ppc32/64-linux target. 2277 2278 * You can set the main thread's stack size at startup using the 2279 new --main-stacksize= flag (subject of course to ulimit settings). 2280 This is useful for running apps that need a lot of stack space. 2281 2282 * The limitation that you can't use --trace-children=yes together 2283 with --db-attach=yes has been removed. 2284 2285 * The following bugs have been fixed. Note that "n-i-bz" stands for 2286 "not in bugzilla" -- that is, a bug that was reported to us but 2287 never got a bugzilla entry. We encourage you to file bugs in 2288 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than 2289 mailing the developers (or mailing lists) directly. 2290 2291 n-i-bz Make return types for some client requests 64-bit clean 2292 n-i-bz glibc 2.9 support 2293 n-i-bz ignore unsafe .valgrindrc's (CVE-2008-4865) 2294 n-i-bz MPI_Init(0,0) is valid but libmpiwrap.c segfaults 2295 n-i-bz Building in an env without gdb gives bogus gdb attach 2296 92456 Tracing the origin of uninitialised memory 2297 106497 Valgrind does not demangle some C++ template symbols 2298 162222 ==106497 2299 151612 Suppression with "..." (frame-level wildcards in .supp files) 2300 156404 Unable to start oocalc under memcheck on openSUSE 10.3 (64-bit) 2301 159285 unhandled syscall:25 (stime, on x86-linux) 2302 159452 unhandled ioctl 0x8B01 on "valgrind iwconfig" 2303 160954 ppc build of valgrind crashes with illegal instruction (isel) 2304 160956 mallinfo implementation, w/ patch 2305 162092 Valgrind fails to start gnome-system-monitor 2306 162819 malloc_free_fill test doesn't pass on glibc2.8 x86 2307 163794 assertion failure with "--track-origins=yes" 2308 163933 sigcontext.err and .trapno must be set together 2309 163955 remove constraint !(--db-attach=yes && --trace-children=yes) 2310 164476 Missing kernel module loading system calls 2311 164669 SVN regression: mmap() drops posix file locks 2312 166581 Callgrind output corruption when program forks 2313 167288 Patch file for missing system calls on Cell BE 2314 168943 unsupported scas instruction pentium 2315 171645 Unrecognised instruction (MOVSD, non-binutils encoding) 2316 172417 x86->IR: 0x82 ... 2317 172563 amd64->IR: 0xD9 0xF5 - fprem1 2318 173099 .lds linker script generation error 2319 173177 [x86_64] syscalls: 125/126/179 (capget/capset/quotactl) 2320 173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant prefixes) 2321 174532 == 173751 2322 174908 --log-file value not expanded correctly for core file 2323 175044 Add lookup_dcookie for amd64 2324 175150 x86->IR: 0xF2 0xF 0x11 0xC1 (movss non-binutils encoding) 2325 2326 Developer-visible changes: 2327 2328 * Valgrind's debug-info reading machinery has been majorly overhauled. 2329 It can now correctly establish the addresses for ELF data symbols, 2330 which is something that has never worked properly before now. 2331 2332 Also, Valgrind can now read DWARF3 type and location information for 2333 stack and global variables. This makes it possible to use the 2334 framework to build tools that rely on knowing the type and locations 2335 of stack and global variables, for example exp-Ptrcheck. 2336 2337 Reading of such information is disabled by default, because most 2338 tools don't need it, and because it is expensive in space and time. 2339 However, you can force Valgrind to read it, using the 2340 --read-var-info=yes flag. Memcheck, Helgrind and DRD are able to 2341 make use of such information, if present, to provide source-level 2342 descriptions of data addresses in the error messages they create. 2343 2344 (3.4.0.RC1: 24 Dec 2008, vex r1878, valgrind r8882). 2345 (3.4.0: 3 Jan 2009, vex r1878, valgrind r8899). 2346
