From: Julian Seward Date: Wed, 4 Jul 2012 20:30:47 +0000 (+0000) Subject: Commit results of triagery so far. X-Git-Tag: svn/VALGRIND_3_8_0~159 X-Git-Url: http://git.ipfire.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=8837568969328a167c44fcf7e457f978f6edbda3;p=thirdparty%2Fvalgrind.git Commit results of triagery so far. git-svn-id: svn://svn.valgrind.org/valgrind/trunk@12709 --- diff --git a/NEWS b/NEWS index 9eae88a67e..a5825ab752 100644 --- a/NEWS +++ b/NEWS @@ -8,6 +8,8 @@ xxx Don't forget to update VALGRIND_MAJOR/MINOR before release * Support for intel AES instructions (AESKEYGENASSIST, AESENC, AESENCLAST, AESDEC, AESDECLAST, AESIMC). Only supported for 64 bit architecture. +* Support for Intel AVX instructions. + * ==================== TOOL CHANGES ==================== * Massif @@ -164,6 +166,7 @@ where XXXXXX is the bug number as listed below. 295428 coregrind/m_main.c has incorrect x86 assembly for darwin 295799 Missing \n with get_vbits in gdbserver when line is % 80 and there are some unaddressable bytes 296422 Add translation chaining support +296457 vex amd64->IR: 0x66 0xF 0x3A 0xDF 0xD1 0x1 0xE8 0x6A (dup of AES) 296983 Fix build issues on x86_64/ppc64 without 32-bit toolchains n-i-bz s390x: Shadow registers can now be examined using vgdb 297078 gdbserver signal handling problems caused by diff vki nr/gdb nr @@ -205,8 +208,6 @@ n-i-bz Fix assert in gdbserver for watchpoints watching the same address Assertion `are_valid_hwcaps(VexArchAMD64, vta->archinfo_host.hwcaps)' failed. 300414 FCOM and FCOMP unimplemented for amd64 guest 301229 dup of 203877, see above. -301967 unhandled instruction bytes 0xC5 0xF9 0x6E 0x40 in /lib32/ld-2.14.1.so - == 273475 (canonical avx) 301984 configure script doesn't detect certain versions of clang 302205 Fix compiler warnings for POWER VEX code and POWER test cases 302536 Fix for the POWER Valgrind regression test: memcheck-ISA2.0. diff --git a/docs/internals/3_7_BUGSTATUS.txt b/docs/internals/3_7_BUGSTATUS.txt index 25b397f3b2..df55c7f730 100644 --- a/docs/internals/3_7_BUGSTATUS.txt +++ b/docs/internals/3_7_BUGSTATUS.txt @@ -8,6 +8,12 @@ been closed as "fixed" are listed in the top level NEWS file instead. In other words they get listed here first and moved to NEWS if they get fixed. +-- reenable amd64_toIR.c commented out insns +-- amd64 toIR memcheck icc -O3 -mAVX probs? +-- OSX 10.8 (alignment) and XCode 4.3 (build, link) problems +-- add arm stack-scan hacks? +-- add allocator-mismatch checking? + 284004 Application under test hangs after call to __posix_spawn (Darwin) open, not started @@ -57,13 +63,11 @@ get fixed. 288756 add '-x' to strip when stripping dynamic libraries on OSX OSX. low prio ---- JRS got this far, tidying up, 30 June 2012. - 289360 parse_type_DIE confused by DW_TAG_enumeration_type - low prio; --read-var-info=yes problem + low prio; --read-var-info=yes problem; OP queried 289470 Loading of large Mach-O thin binaries fails. - w/ patch; apparently harmless; easy fix + **potentially 3.8.0; w/ patch; apparently harmless; easy fix 289578 Backtraces with unwind tables (=without debug symbols) using offline symbol resolving @@ -71,6 +75,7 @@ get fixed. 289584 Unhandled instruction: 0xF 0x29 0xE5 (MOVAPS) movaps reg,reg (w/ patch) + **potentially 3.8.0 289836 Unhanded instructions mfatbu and mfatbl ppc e500 insns @@ -84,83 +89,127 @@ get fixed. 290006 memcheck doesn't mark %xmm as initialized after "pcmpeqw %xmm %xmm" easy fix; w/ patch + **potentially 3.8.0 290061 pie elf always loaded at 0x108000 w/ patch; looks OK; easy fix ? + not sure this wouldn't break something else -- needs looking into 290066 Warning messages from libraries with Valgrind 3.7.0 ppc elfery wierdness + **revisit -- related to some other bug re ELF problems (296318)? 290384 Valgrind does not show line numbers on 32 bit FC16 + **uninvestigated 290673 wine+valgrind: failed to start tool 'memcheck' for platform 'x86-linux': No such file or directory - probably invalid; build config problem? - -(Sun 8 Jan 2012) + **probably invalid; build config problem? 291310 FXSAVE instruction marks memory as undefined on amd64 x86 vs amd64 inconsistency? + **probably easy to fix; low prio 291924 vex x86->IR: unhandled instruction bytes: 0x66 0xF 0x38 0x31 - invalid (SSE4) + invalid (SSE4) (WONTFIX) 292281 Not supported LL cache size of AMD Opteron 6176 hmm, more cache size kludgery required - -292958 ARM SVC instructions with non-zero immediates are not supported. - invalid + **need access to the machine to fix 278313 Fedora 15/x64: err read debug info with --read-var-info=yes flag seems complex; is reopened - - ---- 10 Feb 2012 + **queried OP 293855 vex amd64->IR: 0x2F 0x55 0x73 0x65 0x72 0x73 0x2F 0x6A some problem with interceptors on Helgrind/OSX -294065 valgrind running wine does not find pdb file because of wrong path to egrep +294065 valgrind running wine does not find pdb file because of wrong + path to egrep w/ somewhat dodgy patch 294147 Exception in thread "main" java probably invalid + not sure what is going on here + +294285 --partial-loads-ok does not work for 16-byte SSE loads + true + +294454 delphi uses 16-bit addressing in the TEB (%fs) + obscure (16 bit (0x67 ?) insns in 32 bit mode) + +294523 --partial-loads-ok=yes causes false negatives + **fixed I think -- should close + +294552 svn trunk compilation stuck in mig + XCode 4.3 problem? + no action so far + **possible OSX 10.8 issue + +294861 ./configure fails to locate Boost (located in /usr/local) + low prio + +295053 Valgrind fails to load some x86_64 Mac OS Lion executables + OSX + **possible 3.8.0, med prio + +295084 Hard-coded /usr/include + OSX, XCode 4.3 problem + +251569 rdtscp not supported: vex amd64->IR: 0xF 0x1 0xF9 0x8B 0x4C 0x24 + has doesn't-really-look-right patch + +295403 Memory access below SP with some STRD instructions. + ARM, fiddly to fix/verify + +295427 building i386 support with clang on darwin11 requires + -new_linker linker option + OSX, XCode 4.3, + **has plausible looking patch though (low risk) + +295472 "executable has wrong endian-ness" when target executable has + correct endianess + cross-compiling x86->ppc32 weirdness + +295514 unhandled instruction bytes : 0x66 0x9C (pushfw on x86_64) + not high prio + **possible 3.8.0 (easy to fix?) + +295590 Helgrind: hg_main.c:2298 (evh__HG_PTHREAD_COND_WAIT_POST): + Assertion 'cvi->nWaiters > 0' failed. + **possible 3.8.0 + +295617 ARM - Add some missing syscalls + **possible 3.8.0, needs landing + +295808 vex amd64->IR: 0xF3 0xF 0xBC 0xC0 0x48 0x1 0xD0 0x48 (TZCNT) + **possible 3.8.0 ? + +295974 vex-x86 omits some SSE4.1 opcodes such as PEXTRD + **check before 3.8.0; maybe some 64-bit insns missing? OP queried + +296166 Add support for ARM in ptrcheck + **no patch yet + +296229 Linux user input device ioctls missing wrappers + **possible 3.8.0, but wrong patch; OP queried + +296308 helgrind tests fail on Linux/x86 due to clone syscall + missing unwind info + **possible 3.8.0; has low risk patch + +296311 Wrong stack traces due to -fomit-frame-pointer (x86) + **not sure what to do about this. Is it serious? + +296318 Debug info improvements (more than one rx/rw mapping) + **this is probably important for 3.8.0 + +296480 valgrind hangs early when run on Python + OSX. Not started. + +--- JRS got this far, tidying up, 4 July 2012. --- 16 Feb 2012 - -294285] New: --partial-loads-ok does not work for 16-byte SSE loads -294454] New: delphi uses 16-bit addressing in the TEB (%fs) -294523] New: --partial-loads-ok=yes causes false negatives -294552] New: svn trunk compilation stuck in mig -294861] New: ./configure fails to locate Boost (located in /usr/local) - 295053] New: Valgrind fails to load some x86_64 Mac OS Lion executables -295084] New: Hard-coded /usr/include -251569] rdtscp not supported: vex amd64->IR: unhandled instruction bytes: 0xF 0x1 0xF9 0x8B 0x4C 0x24 -295403] New: Memory access below SP with some STRD instructions. -295427] New: building i386 support with clang on darwin11 requires -new_linker linker option -295472] New: "executable has wrong endian-ness" when target executable has correct endianess -295514] New: unhandled instruction bytes : 0x66 0x9C (pushfw on x86_64) -295590] New: Helgrind: hg_main.c:2298 (evh__HG_PTHREAD_COND_WAIT_POST): Assertion 'cvi->nWaiters > 0' failed. -295617] New: ARM - Add some missing syscalls -295808] New: vex amd64->IR: unhandled instruction bytes: 0xF3 0xF 0xBC 0xC0 0x48 0x1 0xD0 0x48 - --- 11 March 2012 - -Please consider applying the patch contained in bz #274078. It should make these sorts of questions go away (or at least make them substantially rarer). -https://bugs.kde.org/show_bug.cgi?id=274078 - -295974] New: vex-x86 omits some SSE4.1 opcodes such as PEXTRD -296166] New: Add support for ARM in ptrcheck -296229] New: Linux user input device ioctls missing wrappers -296308] New: helgrind tests fail on Linux/x86 due to clone syscall -missing unwind info -296311] New: Wrong stack traces due to -fomit-frame-pointer (x86) - -296318] New: Debug info improvements (more than one rx/rw mapping) -296457] New: vex amd64->IR: unhandled instruction bytes: 0x66 0xF 0x3A -0xDF 0xD1 0x1 0xE8 0x6A -296480] New: valgrind hangs early when run on Python 296577] New: vex x86->IR: unhandled instruction bytes: 0x66 0xF 0x3A 0x17 296578] New: vex amd64->IR: unhandled instruction bytes: 0x60 0x0 0x0 0x0 0x0 0x0 0x0 0x0 296792] New: [PATCH] valgrind 3.7.0: add SIOCSHWTSTAMP (0x89B0) ioctl wrapper @@ -193,6 +242,8 @@ missing unwind info 301830] New: mtV : improve Valgrind so that it runs multiple threads in parallel 301902] New: vex amd64->IR: unhandled instruction bytes: 0xF 0xB 0xF 0x1F 0x0 0x40 0x38 0xFE 301922] New: gcc -O2 problem with SSE optimization? +301967 unhandled instruction bytes 0xC5 0xF9 0x6E 0x40 in /lib32/ld-2.14.1.so + == 273475 (canonical avx) 302234] New: fesetround() function does not affect FPU calculations 302271] New: Bug with Callgrind dump-instr and Trace-children 302287] New: Unhandled movbe instruction on Atom processors