sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 1 | |
sewardj | 3ecf8f2 | 2006-06-06 21:41:22 +0000 | [diff] [blame] | 2 | Release 3.2.0 (7 June 2006) |
sewardj | 535a235 | 2006-04-18 02:04:52 +0000 | [diff] [blame] | 3 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 4 | 3.2.0 is a feature release with many significant improvements and the |
| 5 | usual collection of bug fixes. This release supports X86/Linux, |
| 6 | AMD64/Linux, PPC32/Linux and PPC64/Linux. |
njn | a3e9676 | 2006-03-27 12:23:44 +0000 | [diff] [blame] | 7 | |
sewardj | af92499 | 2006-05-26 13:51:37 +0000 | [diff] [blame] | 8 | Performance, especially of Memcheck, is improved, Addrcheck has been |
| 9 | removed, Callgrind has been added, PPC64/Linux support has been added, |
| 10 | Lackey has been improved, and MPI support has been added. In detail: |
njn | a3e9676 | 2006-03-27 12:23:44 +0000 | [diff] [blame] | 11 | |
sewardj | f41db3c | 2006-05-26 16:06:10 +0000 | [diff] [blame] | 12 | - Memcheck has improved speed and reduced memory use. Run times are |
| 13 | typically reduced by 15-30%, averaging about 24% for SPEC CPU2000. |
sewardj | af92499 | 2006-05-26 13:51:37 +0000 | [diff] [blame] | 14 | The other tools have smaller but noticeable speed improvments. We |
| 15 | are interested to hear what improvements users get. |
njn | a3e9676 | 2006-03-27 12:23:44 +0000 | [diff] [blame] | 16 | |
sewardj | 3ecf8f2 | 2006-06-06 21:41:22 +0000 | [diff] [blame] | 17 | Memcheck uses less memory due to the introduction of a compressed |
sewardj | af92499 | 2006-05-26 13:51:37 +0000 | [diff] [blame] | 18 | representation for shadow memory. The space overhead has been |
| 19 | reduced by a factor of up to four, depending on program behaviour. |
| 20 | This means you should be able to run programs that use more memory |
| 21 | than before without hitting problems. |
njn | a3e9676 | 2006-03-27 12:23:44 +0000 | [diff] [blame] | 22 | |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 23 | - Addrcheck has been removed. It has not worked since version 2.4.0, |
| 24 | and the speed and memory improvements to Memcheck make it redundant. |
| 25 | If you liked using Addrcheck because it didn't give undefined value |
| 26 | errors, you can use the new Memcheck option --undef-value-errors=no |
| 27 | to get the same behaviour. |
| 28 | |
sewardj | af92499 | 2006-05-26 13:51:37 +0000 | [diff] [blame] | 29 | - The number of undefined-value errors incorrectly reported by |
| 30 | Memcheck has been reduced (such false reports were already very |
| 31 | rare). In particular, efforts have been made to ensure Memcheck |
| 32 | works really well with gcc 4.0/4.1-generated code on X86/Linux and |
| 33 | AMD64/Linux. |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 34 | |
| 35 | - Josef Weidendorfer's popular Callgrind tool has been added. Folding |
sewardj | af92499 | 2006-05-26 13:51:37 +0000 | [diff] [blame] | 36 | it in was a logical step given its popularity and usefulness, and |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 37 | makes it easier for us to ensure it works "out of the box" on all |
| 38 | supported targets. The associated KDE KCachegrind GUI remains a |
| 39 | separate project. |
njn | eaf0ca9 | 2006-04-09 01:23:29 +0000 | [diff] [blame] | 40 | |
sewardj | 3ecf8f2 | 2006-06-06 21:41:22 +0000 | [diff] [blame] | 41 | - A new release of the Valkyrie GUI for Memcheck, version 1.2.0, |
| 42 | accompanies this release. Improvements over previous releases |
| 43 | include improved robustness, many refinements to the user interface, |
| 44 | and use of a standard autoconf/automake build system. You can get |
| 45 | it from http://www.valgrind.org/downloads/guis.html. |
| 46 | |
sewardj | 535a235 | 2006-04-18 02:04:52 +0000 | [diff] [blame] | 47 | - Valgrind now works on PPC64/Linux. As with the AMD64/Linux port, |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 48 | this supports programs using to 32G of address space. On 64-bit |
| 49 | capable PPC64/Linux setups, you get a dual architecture build so |
| 50 | that both 32-bit and 64-bit executables can be run. Linux on POWER5 |
| 51 | is supported, and POWER4 is also believed to work. Both 32-bit and |
| 52 | 64-bit DWARF2 is supported. This port is known to work well with |
| 53 | both gcc-compiled and xlc/xlf-compiled code. |
njn | a3e9676 | 2006-03-27 12:23:44 +0000 | [diff] [blame] | 54 | |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 55 | - Floating point accuracy has been improved for PPC32/Linux. |
| 56 | Specifically, the floating point rounding mode is observed on all FP |
| 57 | arithmetic operations, and multiply-accumulate instructions are |
| 58 | preserved by the compilation pipeline. This means you should get FP |
| 59 | results which are bit-for-bit identical to a native run. These |
| 60 | improvements are also present in the PPC64/Linux port. |
njn | b293916 | 2006-04-18 00:52:37 +0000 | [diff] [blame] | 61 | |
njn | eaf0ca9 | 2006-04-09 01:23:29 +0000 | [diff] [blame] | 62 | - Lackey, the example tool, has been improved: |
njn | a3e9676 | 2006-03-27 12:23:44 +0000 | [diff] [blame] | 63 | |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 64 | * It has a new option --detailed-counts (off by default) which |
| 65 | causes it to print out a count of loads, stores and ALU operations |
| 66 | done, and their sizes. |
njn | dbf7ca7 | 2006-03-31 11:57:59 +0000 | [diff] [blame] | 67 | |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 68 | * It has a new option --trace-mem (off by default) which causes it |
| 69 | to print out a trace of all memory accesses performed by a |
| 70 | program. It's a good starting point for building Valgrind tools |
| 71 | that need to track memory accesses. Read the comments at the top |
| 72 | of the file lackey/lk_main.c for details. |
sewardj | 42fcf5d | 2006-04-08 13:16:36 +0000 | [diff] [blame] | 73 | |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 74 | * The original instrumentation (counting numbers of instructions, |
| 75 | jumps, etc) is now controlled by a new option --basic-counts. It |
| 76 | is on by default. |
sewardj | 42fcf5d | 2006-04-08 13:16:36 +0000 | [diff] [blame] | 77 | |
| 78 | - MPI support: partial support for debugging distributed applications |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 79 | using the MPI library specification has been added. Valgrind is |
sewardj | 42fcf5d | 2006-04-08 13:16:36 +0000 | [diff] [blame] | 80 | aware of the memory state changes caused by a subset of the MPI |
| 81 | functions, and will carefully check data passed to the (P)MPI_ |
| 82 | interface. |
| 83 | |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 84 | - A new flag, --error-exitcode=, has been added. This allows changing |
| 85 | the exit code in runs where Valgrind reported errors, which is |
| 86 | useful when using Valgrind as part of an automated test suite. |
sewardj | b977908 | 2006-05-12 23:50:15 +0000 | [diff] [blame] | 87 | |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 88 | - Various segfaults when reading old-style "stabs" debug information |
| 89 | have been fixed. |
njn | eaf0ca9 | 2006-04-09 01:23:29 +0000 | [diff] [blame] | 90 | |
sewardj | c9d09cc | 2006-05-27 03:22:04 +0000 | [diff] [blame] | 91 | - A simple performance evaluation suite has been added. See |
njn | 5359b6f | 2006-06-02 23:57:22 +0000 | [diff] [blame] | 92 | perf/README and README_DEVELOPERS for details. There are |
sewardj | c9d09cc | 2006-05-27 03:22:04 +0000 | [diff] [blame] | 93 | various bells and whistles. |
| 94 | |
sewardj | b70a613 | 2006-05-27 21:14:09 +0000 | [diff] [blame] | 95 | - New configuration flags: |
| 96 | --enable-only32bit |
| 97 | --enable-only64bit |
| 98 | By default, on 64 bit platforms (ppc64-linux, amd64-linux) the build |
| 99 | system will attempt to build a Valgrind which supports both 32-bit |
| 100 | and 64-bit executables. This may not be what you want, and you can |
| 101 | override the default behaviour using these flags. |
| 102 | |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 103 | Please note that Helgrind is still not working. We have made an |
| 104 | important step towards making it work again, however, with the |
| 105 | addition of function wrapping (see below). |
njn | eaf0ca9 | 2006-04-09 01:23:29 +0000 | [diff] [blame] | 106 | |
| 107 | Other user-visible changes: |
| 108 | |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 109 | - Valgrind now has the ability to intercept and wrap arbitrary |
| 110 | functions. This is a preliminary step towards making Helgrind work |
| 111 | again, and was required for MPI support. |
njn | eaf0ca9 | 2006-04-09 01:23:29 +0000 | [diff] [blame] | 112 | |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 113 | - There are some changes to Memcheck's client requests. Some of them |
| 114 | have changed names: |
njn | dbf7ca7 | 2006-03-31 11:57:59 +0000 | [diff] [blame] | 115 | |
| 116 | MAKE_NOACCESS --> MAKE_MEM_NOACCESS |
| 117 | MAKE_WRITABLE --> MAKE_MEM_UNDEFINED |
| 118 | MAKE_READABLE --> MAKE_MEM_DEFINED |
| 119 | |
| 120 | CHECK_WRITABLE --> CHECK_MEM_IS_ADDRESSABLE |
| 121 | CHECK_READABLE --> CHECK_MEM_IS_DEFINED |
| 122 | CHECK_DEFINED --> CHECK_VALUE_IS_DEFINED |
| 123 | |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 124 | The reason for the change is that the old names are subtly |
| 125 | misleading. The old names will still work, but they are deprecated |
| 126 | and may be removed in a future release. |
njn | dbf7ca7 | 2006-03-31 11:57:59 +0000 | [diff] [blame] | 127 | |
| 128 | We also added a new client request: |
| 129 | |
| 130 | MAKE_MEM_DEFINED_IF_ADDRESSABLE(a, len) |
| 131 | |
| 132 | which is like MAKE_MEM_DEFINED but only affects a byte if the byte is |
| 133 | already addressable. |
| 134 | |
sewardj | 3ecf8f2 | 2006-06-06 21:41:22 +0000 | [diff] [blame] | 135 | - The way client requests are encoded in the instruction stream has |
| 136 | changed. Unfortunately, this means 3.2.0 will not honour client |
| 137 | requests compiled into binaries using headers from earlier versions |
| 138 | of Valgrind. We will try to keep the client request encodings more |
| 139 | stable in future. |
njn | eaf0ca9 | 2006-04-09 01:23:29 +0000 | [diff] [blame] | 140 | |
njn | a3e9676 | 2006-03-27 12:23:44 +0000 | [diff] [blame] | 141 | BUGS FIXED: |
| 142 | |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 143 | 108258 NPTL pthread cleanup handlers not called |
| 144 | 117290 valgrind is sigKILL'd on startup |
| 145 | 117295 == 117290 |
| 146 | 118703 m_signals.c:1427 Assertion 'tst->status == VgTs_WaitSys' |
| 147 | 118466 add %reg, %reg generates incorrect validity for bit 0 |
| 148 | 123210 New: strlen from ld-linux on amd64 |
| 149 | 123244 DWARF2 CFI reader: unhandled CFI instruction 0:18 |
| 150 | 123248 syscalls in glibc-2.4: openat, fstatat, symlinkat |
| 151 | 123258 socketcall.recvmsg(msg.msg_iov[i] points to uninit |
| 152 | 123535 mremap(new_addr) requires MREMAP_FIXED in 4th arg |
| 153 | 123836 small typo in the doc |
| 154 | 124029 ppc compile failed: `vor' gcc 3.3.5 |
| 155 | 124222 Segfault: @@don't know what type ':' is |
| 156 | 124475 ppc32: crash (syscall?) timer_settime() |
| 157 | 124499 amd64->IR: 0xF 0xE 0x48 0x85 (femms) |
| 158 | 124528 FATAL: aspacem assertion failed: segment_is_sane |
| 159 | 124697 vex x86->IR: 0xF 0x70 0xC9 0x0 (pshufw) |
| 160 | 124892 vex x86->IR: 0xF3 0xAE (REPx SCASB) |
| 161 | 126216 == 124892 |
| 162 | 124808 ppc32: sys_sched_getaffinity() not handled |
| 163 | n-i-bz Very long stabs strings crash m_debuginfo |
| 164 | n-i-bz amd64->IR: 0x66 0xF 0xF5 (pmaddwd) |
| 165 | 125492 ppc32: support a bunch more syscalls |
| 166 | 121617 ppc32/64: coredumping gives assertion failure |
| 167 | 121814 Coregrind return error as exitcode patch |
| 168 | 126517 == 121814 |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 169 | 125607 amd64->IR: 0x66 0xF 0xA3 0x2 (btw etc) |
| 170 | 125651 amd64->IR: 0xF8 0x49 0xFF 0xE3 (clc?) |
| 171 | 126253 x86 movx is wrong |
| 172 | 126451 3.2 SVN doesn't work on ppc32 CPU's without FPU |
| 173 | 126217 increase # threads |
| 174 | 126243 vex x86->IR: popw mem |
| 175 | 126583 amd64->IR: 0x48 0xF 0xA4 0xC2 (shld $1,%rax,%rdx) |
sewardj | b70a613 | 2006-05-27 21:14:09 +0000 | [diff] [blame] | 176 | 126668 amd64->IR: 0x1C 0xFF (sbb $0xff,%al) |
| 177 | 126696 support for CDROMREADRAW ioctl and CDROMREADTOCENTRY fix |
| 178 | 126722 assertion: segment_is_sane at m_aspacemgr/aspacemgr.c:1624 |
| 179 | 126938 bad checking for syscalls linkat, renameat, symlinkat |
sewardj | 1ede80f | 2006-05-26 00:47:13 +0000 | [diff] [blame] | 180 | |
sewardj | 3ecf8f2 | 2006-06-06 21:41:22 +0000 | [diff] [blame] | 181 | (3.2.0RC1: 27 May 2006, vex r1626, valgrind r5947). |
| 182 | (3.2.0: 7 June 2006, vex r1628, valgrind r5957). |
njn | a3e9676 | 2006-03-27 12:23:44 +0000 | [diff] [blame] | 183 | |
| 184 | |
sewardj | 5065467 | 2006-03-16 11:33:14 +0000 | [diff] [blame] | 185 | Release 3.1.1 (15 March 2006) |
| 186 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 187 | 3.1.1 fixes a bunch of bugs reported in 3.1.0. There is no new |
| 188 | functionality. The fixed bugs are: |
| 189 | |
| 190 | (note: "n-i-bz" means "not in bugzilla" -- this bug does not have |
| 191 | a bugzilla entry). |
| 192 | |
| 193 | n-i-bz ppc32: fsub 3,3,3 in dispatcher doesn't clear NaNs |
| 194 | n-i-bz ppc32: __NR_{set,get}priority |
| 195 | 117332 x86: missing line info with icc 8.1 |
| 196 | 117366 amd64: 0xDD 0x7C fnstsw |
| 197 | 118274 == 117366 |
| 198 | 117367 amd64: 0xD9 0xF4 fxtract |
| 199 | 117369 amd64: __NR_getpriority (140) |
| 200 | 117419 ppc32: lfsu f5, -4(r11) |
| 201 | 117419 ppc32: fsqrt |
| 202 | 117936 more stabs problems (segfaults while reading debug info) |
| 203 | 119914 == 117936 |
| 204 | 120345 == 117936 |
| 205 | 118239 amd64: 0xF 0xAE 0x3F (clflush) |
| 206 | 118939 vm86old system call |
| 207 | n-i-bz memcheck/tests/mempool reads freed memory |
| 208 | n-i-bz AshleyP's custom-allocator assertion |
| 209 | n-i-bz Dirk strict-aliasing stuff |
| 210 | n-i-bz More space for debugger cmd line (Dan Thaler) |
| 211 | n-i-bz Clarified leak checker output message |
| 212 | n-i-bz AshleyP's --gen-suppressions output fix |
| 213 | n-i-bz cg_annotate's --sort option broken |
| 214 | n-i-bz OSet 64-bit fastcmp bug |
| 215 | n-i-bz VG_(getgroups) fix (Shinichi Noda) |
| 216 | n-i-bz ppc32: allocate from callee-saved FP/VMX regs |
| 217 | n-i-bz misaligned path word-size bug in mc_main.c |
| 218 | 119297 Incorrect error message for sse code |
| 219 | 120410 x86: prefetchw (0xF 0xD 0x48 0x4) |
| 220 | 120728 TIOCSERGETLSR, TIOCGICOUNT, HDIO_GET_DMA ioctls |
| 221 | 120658 Build fixes for gcc 2.96 |
| 222 | 120734 x86: Support for changing EIP in signal handler |
| 223 | n-i-bz memcheck/tests/zeropage de-looping fix |
| 224 | n-i-bz x86: fxtract doesn't work reliably |
| 225 | 121662 x86: lock xadd (0xF0 0xF 0xC0 0x2) |
| 226 | 121893 calloc does not always return zeroed memory |
| 227 | 121901 no support for syscall tkill |
| 228 | n-i-bz Suppression update for Debian unstable |
| 229 | 122067 amd64: fcmovnu (0xDB 0xD9) |
| 230 | n-i-bz ppc32: broken signal handling in cpu feature detection |
| 231 | n-i-bz ppc32: rounding mode problems (improved, partial fix only) |
| 232 | 119482 ppc32: mtfsb1 |
| 233 | n-i-bz ppc32: mtocrf/mfocrf |
| 234 | |
| 235 | (3.1.1: 15 March 2006, vex r1597, valgrind r5771). |
| 236 | |
jseward | 4323065 | 2003-12-14 00:14:54 +0000 | [diff] [blame] | 237 | |
sewardj | 4bbe9d7 | 2005-11-16 18:39:25 +0000 | [diff] [blame] | 238 | Release 3.1.0 (25 November 2005) |
njn | f63170c | 2005-11-12 19:09:51 +0000 | [diff] [blame] | 239 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
sewardj | 4bbe9d7 | 2005-11-16 18:39:25 +0000 | [diff] [blame] | 240 | 3.1.0 is a feature release with a number of significant improvements: |
| 241 | AMD64 support is much improved, PPC32 support is good enough to be |
| 242 | usable, and the handling of memory management and address space is |
| 243 | much more robust. In detail: |
njn | f63170c | 2005-11-12 19:09:51 +0000 | [diff] [blame] | 244 | |
sewardj | 4bbe9d7 | 2005-11-16 18:39:25 +0000 | [diff] [blame] | 245 | - AMD64 support is much improved. The 64-bit vs. 32-bit issues in |
| 246 | 3.0.X have been resolved, and it should "just work" now in all |
| 247 | cases. On AMD64 machines both 64-bit and 32-bit versions of |
| 248 | Valgrind are built. The right version will be invoked |
| 249 | automatically, even when using --trace-children and mixing execution |
| 250 | between 64-bit and 32-bit executables. Also, many more instructions |
| 251 | are supported. |
njn | f63170c | 2005-11-12 19:09:51 +0000 | [diff] [blame] | 252 | |
sewardj | 4bbe9d7 | 2005-11-16 18:39:25 +0000 | [diff] [blame] | 253 | - PPC32 support is now good enough to be usable. It should work with |
| 254 | all tools, but please let us know if you have problems. Three |
| 255 | classes of CPUs are supported: integer only (no FP, no Altivec), |
| 256 | which covers embedded PPC uses, integer and FP but no Altivec |
| 257 | (G3-ish), and CPUs capable of Altivec too (G4, G5). |
njn | f63170c | 2005-11-12 19:09:51 +0000 | [diff] [blame] | 258 | |
sewardj | 4bbe9d7 | 2005-11-16 18:39:25 +0000 | [diff] [blame] | 259 | - Valgrind's address space management has been overhauled. As a |
| 260 | result, Valgrind should be much more robust with programs that use |
| 261 | large amounts of memory. There should be many fewer "memory |
| 262 | exhausted" messages, and debug symbols should be read correctly on |
| 263 | large (eg. 300MB+) executables. On 32-bit machines the full address |
| 264 | space available to user programs (usually 3GB or 4GB) can be fully |
| 265 | utilised. On 64-bit machines up to 32GB of space is usable; when |
| 266 | using Memcheck that means your program can use up to about 14GB. |
njn | f63170c | 2005-11-12 19:09:51 +0000 | [diff] [blame] | 267 | |
| 268 | A side effect of this change is that Valgrind is no longer protected |
| 269 | against wild writes by the client. This feature was nice but relied |
| 270 | on the x86 segment registers and so wasn't portable. |
| 271 | |
| 272 | - Most users should not notice, but as part of the address space |
sewardj | 4bbe9d7 | 2005-11-16 18:39:25 +0000 | [diff] [blame] | 273 | manager change, the way Valgrind is built has been changed. Each |
| 274 | tool is now built as a statically linked stand-alone executable, |
| 275 | rather than as a shared object that is dynamically linked with the |
| 276 | core. The "valgrind" program invokes the appropriate tool depending |
| 277 | on the --tool option. This slightly increases the amount of disk |
| 278 | space used by Valgrind, but it greatly simplified many things and |
| 279 | removed Valgrind's dependence on glibc. |
njn | f63170c | 2005-11-12 19:09:51 +0000 | [diff] [blame] | 280 | |
njn | 336c89d | 2005-11-20 18:14:04 +0000 | [diff] [blame] | 281 | Please note that Addrcheck and Helgrind are still not working. Work |
| 282 | is underway to reinstate them (or equivalents). We apologise for the |
| 283 | inconvenience. |
| 284 | |
njn | f63170c | 2005-11-12 19:09:51 +0000 | [diff] [blame] | 285 | Other user-visible changes: |
| 286 | |
| 287 | - The --weird-hacks option has been renamed --sim-hints. |
| 288 | |
| 289 | - The --time-stamp option no longer gives an absolute date and time. |
| 290 | It now prints the time elapsed since the program began. |
| 291 | |
sewardj | 4bbe9d7 | 2005-11-16 18:39:25 +0000 | [diff] [blame] | 292 | - It should build with gcc-2.96. |
| 293 | |
njn | 7cce5b8 | 2005-11-16 20:12:22 +0000 | [diff] [blame] | 294 | - Valgrind can now run itself (see README_DEVELOPERS for how). |
sewardj | 4705861 | 2005-11-16 19:13:13 +0000 | [diff] [blame] | 295 | This is not much use to you, but it means the developers can now |
| 296 | profile Valgrind using Cachegrind. As a result a couple of |
| 297 | performance bad cases have been fixed. |
| 298 | |
sewardj | 6a5a69c | 2005-11-17 00:51:36 +0000 | [diff] [blame] | 299 | - The XML output format has changed slightly. See |
| 300 | docs/internals/xml-output.txt. |
| 301 | |
njn | 14c9add | 2005-11-18 17:36:01 +0000 | [diff] [blame] | 302 | - Core dumping has been reinstated (it was disabled in 3.0.0 and 3.0.1). |
| 303 | If your program crashes while running under Valgrind, a core file with |
| 304 | the name "vgcore.<pid>" will be created (if your settings allow core |
| 305 | file creation). Note that the floating point information is not all |
| 306 | there. If Valgrind itself crashes, the OS will create a normal core |
| 307 | file. |
| 308 | |
njn | f63170c | 2005-11-12 19:09:51 +0000 | [diff] [blame] | 309 | The following are some user-visible changes that occurred in earlier |
| 310 | versions that may not have been announced, or were announced but not |
njn | 7cce5b8 | 2005-11-16 20:12:22 +0000 | [diff] [blame] | 311 | widely noticed. So we're mentioning them now. |
njn | f63170c | 2005-11-12 19:09:51 +0000 | [diff] [blame] | 312 | |
| 313 | - The --tool flag is optional once again; if you omit it, Memcheck |
| 314 | is run by default. |
| 315 | |
| 316 | - The --num-callers flag now has a default value of 12. It was |
| 317 | previously 4. |
| 318 | |
njn | 7cce5b8 | 2005-11-16 20:12:22 +0000 | [diff] [blame] | 319 | - The --xml=yes flag causes Valgrind's output to be produced in XML |
| 320 | format. This is designed to make it easy for other programs to |
| 321 | consume Valgrind's output. The format is described in the file |
njn | f63170c | 2005-11-12 19:09:51 +0000 | [diff] [blame] | 322 | docs/internals/xml-format.txt. |
| 323 | |
njn | 1f50399 | 2005-11-13 16:41:15 +0000 | [diff] [blame] | 324 | - The --gen-suppressions flag supports an "all" value that causes every |
| 325 | suppression to be printed without asking. |
| 326 | |
njn | ee88d28 | 2005-11-20 17:19:28 +0000 | [diff] [blame] | 327 | - The --log-file option no longer puts "pid" in the filename, eg. the |
| 328 | old name "foo.pid12345" is now "foo.12345". |
| 329 | |
njn | c66077e | 2005-11-20 18:02:15 +0000 | [diff] [blame] | 330 | - There are several graphical front-ends for Valgrind, such as Valkyrie, |
| 331 | Alleyoop and Valgui. See http://www.valgrind.org/downloads/guis.html |
| 332 | for a list. |
| 333 | |
njn | f63170c | 2005-11-12 19:09:51 +0000 | [diff] [blame] | 334 | BUGS FIXED: |
| 335 | |
sewardj | 4bbe9d7 | 2005-11-16 18:39:25 +0000 | [diff] [blame] | 336 | 109861 amd64 hangs at startup |
| 337 | 110301 ditto |
| 338 | 111554 valgrind crashes with Cannot allocate memory |
| 339 | 111809 Memcheck tool doesn't start java |
| 340 | 111901 cross-platform run of cachegrind fails on opteron |
| 341 | 113468 (vgPlain_mprotect_range): Assertion 'r != -1' failed. |
| 342 | 92071 Reading debugging info uses too much memory |
| 343 | 109744 memcheck loses track of mmap from direct ld-linux.so.2 |
| 344 | 110183 tail of page with _end |
| 345 | 82301 FV memory layout too rigid |
| 346 | 98278 Infinite recursion possible when allocating memory |
| 347 | 108994 Valgrind runs out of memory due to 133x overhead |
| 348 | 115643 valgrind cannot allocate memory |
| 349 | 105974 vg_hashtable.c static hash table |
| 350 | 109323 ppc32: dispatch.S uses Altivec insn, which doesn't work on POWER. |
| 351 | 109345 ptrace_setregs not yet implemented for ppc |
| 352 | 110831 Would like to be able to run against both 32 and 64 bit |
| 353 | binaries on AMD64 |
| 354 | 110829 == 110831 |
| 355 | 111781 compile of valgrind-3.0.0 fails on my linux (gcc 2.X prob) |
| 356 | 112670 Cachegrind: cg_main.c:486 (handleOneStatement ... |
| 357 | 112941 vex x86: 0xD9 0xF4 (fxtract) |
| 358 | 110201 == 112941 |
| 359 | 113015 vex amd64->IR: 0xE3 0x14 0x48 0x83 (jrcxz) |
| 360 | 113126 Crash with binaries built with -gstabs+/-ggdb |
| 361 | 104065 == 113126 |
| 362 | 115741 == 113126 |
| 363 | 113403 Partial SSE3 support on x86 |
| 364 | 113541 vex: Grp5(x86) (alt encoding inc/dec) case 1 |
| 365 | 113642 valgrind crashes when trying to read debug information |
| 366 | 113810 vex x86->IR: 66 0F F6 (66 + PSADBW == SSE PSADBW) |
| 367 | 113796 read() and write() do not work if buffer is in shared memory |
| 368 | 113851 vex x86->IR: (pmaddwd): 0x66 0xF 0xF5 0xC7 |
| 369 | 114366 vex amd64 cannnot handle __asm__( "fninit" ) |
| 370 | 114412 vex amd64->IR: 0xF 0xAD 0xC2 0xD3 (128-bit shift, shrdq?) |
| 371 | 114455 vex amd64->IR: 0xF 0xAC 0xD0 0x1 (also shrdq) |
| 372 | 115590: amd64->IR: 0x67 0xE3 0x9 0xEB (address size override) |
| 373 | 115953 valgrind svn r5042 does not build with parallel make (-j3) |
| 374 | 116057 maximum instruction size - VG_MAX_INSTR_SZB too small? |
| 375 | 116483 shmat failes with invalid argument |
| 376 | 102202 valgrind crashes when realloc'ing until out of memory |
| 377 | 109487 == 102202 |
| 378 | 110536 == 102202 |
| 379 | 112687 == 102202 |
| 380 | 111724 vex amd64->IR: 0x41 0xF 0xAB (more BT{,S,R,C} fun n games) |
| 381 | 111748 vex amd64->IR: 0xDD 0xE2 (fucom) |
| 382 | 111785 make fails if CC contains spaces |
| 383 | 111829 vex x86->IR: sbb AL, Ib |
| 384 | 111851 vex x86->IR: 0x9F 0x89 (lahf/sahf) |
| 385 | 112031 iopl on AMD64 and README_MISSING_SYSCALL_OR_IOCTL update |
| 386 | 112152 code generation for Xin_MFence on x86 with SSE0 subarch |
| 387 | 112167 == 112152 |
| 388 | 112789 == 112152 |
| 389 | 112199 naked ar tool is used in vex makefile |
| 390 | 112501 vex x86->IR: movq (0xF 0x7F 0xC1 0xF) (mmx MOVQ) |
| 391 | 113583 == 112501 |
| 392 | 112538 memalign crash |
| 393 | 113190 Broken links in docs/html/ |
| 394 | 113230 Valgrind sys_pipe on x86-64 wrongly thinks file descriptors |
| 395 | should be 64bit |
| 396 | 113996 vex amd64->IR: fucomp (0xDD 0xE9) |
| 397 | 114196 vex x86->IR: out %eax,(%dx) (0xEF 0xC9 0xC3 0x90) |
| 398 | 114289 Memcheck fails to intercept malloc when used in an uclibc environment |
| 399 | 114756 mbind syscall support |
| 400 | 114757 Valgrind dies with assertion: Assertion 'noLargerThan > 0' failed |
| 401 | 114563 stack tracking module not informed when valgrind switches threads |
| 402 | 114564 clone() and stacks |
| 403 | 114565 == 114564 |
| 404 | 115496 glibc crashes trying to use sysinfo page |
| 405 | 116200 enable fsetxattr, fgetxattr, and fremovexattr for amd64 |
njn | f63170c | 2005-11-12 19:09:51 +0000 | [diff] [blame] | 406 | |
sewardj | ae17e7e | 2005-11-20 19:45:30 +0000 | [diff] [blame] | 407 | (3.1.0RC1: 20 November 2005, vex r1466, valgrind r5224). |
sewardj | fa46202 | 2005-11-26 13:38:58 +0000 | [diff] [blame] | 408 | (3.1.0: 26 November 2005, vex r1471, valgrind r5235). |
sewardj | ae17e7e | 2005-11-20 19:45:30 +0000 | [diff] [blame] | 409 | |
njn | f63170c | 2005-11-12 19:09:51 +0000 | [diff] [blame] | 410 | |
sewardj | cd62fe7 | 2005-08-29 13:24:51 +0000 | [diff] [blame] | 411 | Release 3.0.1 (29 August 2005) |
| 412 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 413 | 3.0.1 fixes a bunch of bugs reported in 3.0.0. There is no new |
| 414 | functionality. Some of the fixed bugs are critical, so if you |
sewardj | fa23389 | 2005-08-30 12:35:36 +0000 | [diff] [blame] | 415 | use/distribute 3.0.0, an upgrade to 3.0.1 is recommended. The fixed |
sewardj | cd62fe7 | 2005-08-29 13:24:51 +0000 | [diff] [blame] | 416 | bugs are: |
| 417 | |
| 418 | (note: "n-i-bz" means "not in bugzilla" -- this bug does not have |
| 419 | a bugzilla entry). |
| 420 | |
| 421 | 109313 (== 110505) x86 cmpxchg8b |
| 422 | n-i-bz x86: track but ignore changes to %eflags.AC (alignment check) |
| 423 | 110102 dis_op2_E_G(amd64) |
| 424 | 110202 x86 sys_waitpid(#286) |
| 425 | 110203 clock_getres(,0) |
| 426 | 110208 execve fail wrong retval |
| 427 | 110274 SSE1 now mandatory for x86 |
| 428 | 110388 amd64 0xDD 0xD1 |
| 429 | 110464 amd64 0xDC 0x1D FCOMP |
| 430 | 110478 amd64 0xF 0xD PREFETCH |
| 431 | n-i-bz XML <unique> printing wrong |
| 432 | n-i-bz Dirk r4359 (amd64 syscalls from trunk) |
| 433 | 110591 amd64 and x86: rdtsc not implemented properly |
| 434 | n-i-bz Nick r4384 (stub implementations of Addrcheck and Helgrind) |
| 435 | 110652 AMD64 valgrind crashes on cwtd instruction |
| 436 | 110653 AMD64 valgrind crashes on sarb $0x4,foo(%rip) instruction |
| 437 | 110656 PATH=/usr/bin::/bin valgrind foobar stats ./fooba |
| 438 | 110657 Small test fixes |
| 439 | 110671 vex x86->IR: unhandled instruction bytes: 0xF3 0xC3 (rep ret) |
| 440 | n-i-bz Nick (Cachegrind should not assert when it encounters a client |
| 441 | request.) |
| 442 | 110685 amd64->IR: unhandled instruction bytes: 0xE1 0x56 (loope Jb) |
| 443 | 110830 configuring with --host fails to build 32 bit on 64 bit target |
| 444 | 110875 Assertion when execve fails |
| 445 | n-i-bz Updates to Memcheck manual |
| 446 | n-i-bz Fixed broken malloc_usable_size() |
| 447 | 110898 opteron instructions missing: btq btsq btrq bsfq |
| 448 | 110954 x86->IR: unhandled instruction bytes: 0xE2 0xF6 (loop Jb) |
| 449 | n-i-bz Make suppressions work for "???" lines in stacktraces. |
| 450 | 111006 bogus warnings from linuxthreads |
| 451 | 111092 x86: dis_Grp2(Reg): unhandled case(x86) |
| 452 | 111231 sctp_getladdrs() and sctp_getpaddrs() returns uninitialized |
| 453 | memory |
| 454 | 111102 (comment #4) Fixed 64-bit unclean "silly arg" message |
| 455 | n-i-bz vex x86->IR: unhandled instruction bytes: 0x14 0x0 |
| 456 | n-i-bz minor umount/fcntl wrapper fixes |
| 457 | 111090 Internal Error running Massif |
| 458 | 101204 noisy warning |
| 459 | 111513 Illegal opcode for SSE instruction (x86 movups) |
| 460 | 111555 VEX/Makefile: CC is set to gcc |
sewardj | fa23389 | 2005-08-30 12:35:36 +0000 | [diff] [blame] | 461 | n-i-bz Fix XML bugs in FAQ |
sewardj | cd62fe7 | 2005-08-29 13:24:51 +0000 | [diff] [blame] | 462 | |
sewardj | fa23389 | 2005-08-30 12:35:36 +0000 | [diff] [blame] | 463 | (3.0.1: 29 August 05, |
| 464 | vex/branches/VEX_3_0_BRANCH r1367, |
| 465 | valgrind/branches/VALGRIND_3_0_BRANCH r4574). |
sewardj | 820c772 | 2005-08-29 13:44:43 +0000 | [diff] [blame] | 466 | |
| 467 | |
sewardj | cd62fe7 | 2005-08-29 13:24:51 +0000 | [diff] [blame] | 468 | |
sewardj | ec39f33 | 2005-08-02 15:28:07 +0000 | [diff] [blame] | 469 | Release 3.0.0 (3 August 2005) |
| 470 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
sewardj | f9929e6 | 2005-08-02 21:05:58 +0000 | [diff] [blame] | 471 | 3.0.0 is a major overhaul of Valgrind. The most significant user |
| 472 | visible change is that Valgrind now supports architectures other than |
| 473 | x86. The new architectures it supports are AMD64 and PPC32, and the |
| 474 | infrastructure is present for other architectures to be added later. |
njn | 4a9f6f4 | 2005-07-19 23:46:08 +0000 | [diff] [blame] | 475 | |
sewardj | f9929e6 | 2005-08-02 21:05:58 +0000 | [diff] [blame] | 476 | AMD64 support works well, but has some shortcomings: |
njn | 4a9f6f4 | 2005-07-19 23:46:08 +0000 | [diff] [blame] | 477 | |
njn | 7b1a6a4 | 2005-07-20 04:20:44 +0000 | [diff] [blame] | 478 | - It generally won't be as solid as the x86 version. For example, |
| 479 | support for more obscure instructions and system calls may be missing. |
| 480 | We will fix these as they arise. |
njn | 4a9f6f4 | 2005-07-19 23:46:08 +0000 | [diff] [blame] | 481 | |
sewardj | c371bd0 | 2005-07-27 09:55:24 +0000 | [diff] [blame] | 482 | - Address space may be limited; see the point about |
njn | 7b1a6a4 | 2005-07-20 04:20:44 +0000 | [diff] [blame] | 483 | position-independent executables below. |
njn | 4a9f6f4 | 2005-07-19 23:46:08 +0000 | [diff] [blame] | 484 | |
njn | 7b1a6a4 | 2005-07-20 04:20:44 +0000 | [diff] [blame] | 485 | - If Valgrind is built on an AMD64 machine, it will only run 64-bit |
| 486 | executables. If you want to run 32-bit x86 executables under Valgrind |
| 487 | on an AMD64, you will need to build Valgrind on an x86 machine and |
| 488 | copy it to the AMD64 machine. And it probably won't work if you do |
| 489 | something tricky like exec'ing a 32-bit program from a 64-bit program |
| 490 | while using --trace-children=yes. We hope to improve this situation |
| 491 | in the future. |
| 492 | |
| 493 | The PPC32 support is very basic. It may not work reliably even for |
sewardj | c371bd0 | 2005-07-27 09:55:24 +0000 | [diff] [blame] | 494 | small programs, but it's a start. Many thanks to Paul Mackerras for |
| 495 | his great work that enabled this support. We are working to make |
| 496 | PPC32 usable as soon as possible. |
njn | 4a9f6f4 | 2005-07-19 23:46:08 +0000 | [diff] [blame] | 497 | |
| 498 | Other user-visible changes: |
| 499 | |
sewardj | ec39f33 | 2005-08-02 15:28:07 +0000 | [diff] [blame] | 500 | - Valgrind is no longer built by default as a position-independent |
| 501 | executable (PIE), as this caused too many problems. |
njn | 4a9f6f4 | 2005-07-19 23:46:08 +0000 | [diff] [blame] | 502 | |
| 503 | Without PIE enabled, AMD64 programs will only be able to access 2GB of |
| 504 | address space. We will fix this eventually, but not for the moment. |
| 505 | |
| 506 | Use --enable-pie at configure-time to turn this on. |
| 507 | |
| 508 | - Support for programs that use stack-switching has been improved. Use |
| 509 | the --max-stackframe flag for simple cases, and the |
| 510 | VALGRIND_STACK_REGISTER, VALGRIND_STACK_DEREGISTER and |
| 511 | VALGRIND_STACK_CHANGE client requests for trickier cases. |
| 512 | |
njn | 7b1a6a4 | 2005-07-20 04:20:44 +0000 | [diff] [blame] | 513 | - Support for programs that use self-modifying code has been improved, |
| 514 | in particular programs that put temporary code fragments on the stack. |
| 515 | This helps for C programs compiled with GCC that use nested functions, |
njn | e772332 | 2005-07-24 17:19:09 +0000 | [diff] [blame] | 516 | and also Ada programs. This is controlled with the --smc-check |
njn | 7b1a6a4 | 2005-07-20 04:20:44 +0000 | [diff] [blame] | 517 | flag, although the default setting should work in most cases. |
njn | 4a9f6f4 | 2005-07-19 23:46:08 +0000 | [diff] [blame] | 518 | |
njn | 7b1a6a4 | 2005-07-20 04:20:44 +0000 | [diff] [blame] | 519 | - Output can now be printed in XML format. This should make it easier |
sewardj | c371bd0 | 2005-07-27 09:55:24 +0000 | [diff] [blame] | 520 | for tools such as GUI front-ends and automated error-processing |
| 521 | schemes to use Valgrind output as input. The --xml flag controls this. |
| 522 | As part of this change, ELF directory information is read from executables, |
| 523 | so absolute source file paths are available if needed. |
njn | 4a9f6f4 | 2005-07-19 23:46:08 +0000 | [diff] [blame] | 524 | |
| 525 | - Programs that allocate many heap blocks may run faster, due to |
| 526 | improvements in certain data structures. |
| 527 | |
njn | 7b1a6a4 | 2005-07-20 04:20:44 +0000 | [diff] [blame] | 528 | - Addrcheck is currently not working. We hope to get it working again |
| 529 | soon. Helgrind is still not working, as was the case for the 2.4.0 |
| 530 | release. |
njn | 4a9f6f4 | 2005-07-19 23:46:08 +0000 | [diff] [blame] | 531 | |
sewardj | c371bd0 | 2005-07-27 09:55:24 +0000 | [diff] [blame] | 532 | - The JITter has been completely rewritten, and is now in a separate |
| 533 | library, called Vex. This enabled a lot of the user-visible changes, |
| 534 | such as new architecture support. The new JIT unfortunately translates |
| 535 | more slowly than the old one, so programs may take longer to start. |
| 536 | We believe the code quality is produces is about the same, so once |
| 537 | started, programs should run at about the same speed. Feedback about |
| 538 | this would be useful. |
| 539 | |
| 540 | On the plus side, Vex and hence Memcheck tracks value flow properly |
| 541 | through floating point and vector registers, something the 2.X line |
| 542 | could not do. That means that Memcheck is much more likely to be |
| 543 | usably accurate on vectorised code. |
| 544 | |
sewardj | ec39f33 | 2005-08-02 15:28:07 +0000 | [diff] [blame] | 545 | - There is a subtle change to the way exiting of threaded programs |
sewardj | c371bd0 | 2005-07-27 09:55:24 +0000 | [diff] [blame] | 546 | is handled. In 3.0, Valgrind's final diagnostic output (leak check, |
| 547 | etc) is not printed until the last thread exits. If the last thread |
| 548 | to exit was not the original thread which started the program, any |
| 549 | other process wait()-ing on this one to exit may conclude it has |
| 550 | finished before the diagnostic output is printed. This may not be |
| 551 | what you expect. 2.X had a different scheme which avoided this |
| 552 | problem, but caused deadlocks under obscure circumstances, so we |
| 553 | are trying something different for 3.0. |
| 554 | |
| 555 | - Small changes in control log file naming which make it easier to |
sewardj | ec39f33 | 2005-08-02 15:28:07 +0000 | [diff] [blame] | 556 | use valgrind for debugging MPI-based programs. The relevant |
| 557 | new flags are --log-file-exactly= and --log-file-qualifier=. |
sewardj | c371bd0 | 2005-07-27 09:55:24 +0000 | [diff] [blame] | 558 | |
| 559 | - As part of adding AMD64 support, DWARF2 CFI-based stack unwinding |
| 560 | support was added. In principle this means Valgrind can produce |
| 561 | meaningful backtraces on x86 code compiled with -fomit-frame-pointer |
tom | 81868f0 | 2005-07-27 10:33:08 +0000 | [diff] [blame] | 562 | providing you also compile your code with -fasynchronous-unwind-tables. |
sewardj | c371bd0 | 2005-07-27 09:55:24 +0000 | [diff] [blame] | 563 | |
sewardj | ec39f33 | 2005-08-02 15:28:07 +0000 | [diff] [blame] | 564 | - The documentation build system has been completely redone. |
| 565 | The documentation masters are now in XML format, and from that |
| 566 | HTML, PostScript and PDF documentation is generated. As a result |
| 567 | the manual is now available in book form. Note that the |
| 568 | documentation in the source tarballs is pre-built, so you don't need |
| 569 | any XML processing tools to build Valgrind from a tarball. |
njn | 4a9f6f4 | 2005-07-19 23:46:08 +0000 | [diff] [blame] | 570 | |
| 571 | Changes that are not user-visible: |
| 572 | |
sewardj | c371bd0 | 2005-07-27 09:55:24 +0000 | [diff] [blame] | 573 | - The code has been massively overhauled in order to modularise it. |
| 574 | As a result we hope it is easier to navigate and understand. |
njn | 4a9f6f4 | 2005-07-19 23:46:08 +0000 | [diff] [blame] | 575 | |
| 576 | - Lots of code has been rewritten. |
| 577 | |
sewardj | f9929e6 | 2005-08-02 21:05:58 +0000 | [diff] [blame] | 578 | BUGS FIXED: |
njn | 4a9f6f4 | 2005-07-19 23:46:08 +0000 | [diff] [blame] | 579 | |
sewardj | 79bd9b6 | 2005-08-03 16:40:10 +0000 | [diff] [blame] | 580 | 110046 sz == 4 assertion failed |
| 581 | 109810 vex amd64->IR: unhandled instruction bytes: 0xA3 0x4C 0x70 0xD7 |
sewardj | f9929e6 | 2005-08-02 21:05:58 +0000 | [diff] [blame] | 582 | 109802 Add a plausible_stack_size command-line parameter ? |
| 583 | 109783 unhandled ioctl TIOCMGET (running hw detection tool discover) |
| 584 | 109780 unhandled ioctl BLKSSZGET (running fdisk -l /dev/hda) |
| 585 | 109718 vex x86->IR: unhandled instruction: ffreep |
| 586 | 109429 AMD64 unhandled syscall: 127 (sigpending) |
| 587 | 109401 false positive uninit in strchr from ld-linux.so.2 |
| 588 | 109385 "stabs" parse failure |
| 589 | 109378 amd64: unhandled instruction REP NOP |
| 590 | 109376 amd64: unhandled instruction LOOP Jb |
| 591 | 109363 AMD64 unhandled instruction bytes |
| 592 | 109362 AMD64 unhandled syscall: 24 (sched_yield) |
| 593 | 109358 fork() won't work with valgrind-3.0 SVN |
| 594 | 109332 amd64 unhandled instruction: ADC Ev, Gv |
| 595 | 109314 Bogus memcheck report on amd64 |
| 596 | 108883 Crash; vg_memory.c:905 (vgPlain_init_shadow_range): |
| 597 | Assertion `vgPlain_defined_init_shadow_page()' failed. |
| 598 | 108349 mincore syscall parameter checked incorrectly |
| 599 | 108059 build infrastructure: small update |
| 600 | 107524 epoll_ctl event parameter checked on EPOLL_CTL_DEL |
| 601 | 107123 Vex dies with unhandled instructions: 0xD9 0x31 0xF 0xAE |
| 602 | 106841 auxmap & openGL problems |
| 603 | 106713 SDL_Init causes valgrind to exit |
| 604 | 106352 setcontext and makecontext not handled correctly |
| 605 | 106293 addresses beyond initial client stack allocation |
| 606 | not checked in VALGRIND_DO_LEAK_CHECK |
| 607 | 106283 PIE client programs are loaded at address 0 |
| 608 | 105831 Assertion `vgPlain_defined_init_shadow_page()' failed. |
| 609 | 105039 long run-times probably due to memory manager |
| 610 | 104797 valgrind needs to be aware of BLKGETSIZE64 |
| 611 | 103594 unhandled instruction: FICOM |
| 612 | 103320 Valgrind 2.4.0 fails to compile with gcc 3.4.3 and -O0 |
| 613 | 103168 potentially memory leak in coregrind/ume.c |
| 614 | 102039 bad permissions for mapped region at address 0xB7C73680 |
| 615 | 101881 weird assertion problem |
| 616 | 101543 Support fadvise64 syscalls |
| 617 | 75247 x86_64/amd64 support (the biggest "bug" we have ever fixed) |
njn | 4a9f6f4 | 2005-07-19 23:46:08 +0000 | [diff] [blame] | 618 | |
sewardj | 9a91951 | 2005-08-03 17:01:51 +0000 | [diff] [blame] | 619 | (3.0RC1: 27 July 05, vex r1303, valgrind r4283). |
sewardj | 1346e13 | 2005-08-03 20:42:06 +0000 | [diff] [blame] | 620 | (3.0.0: 3 August 05, vex r1313, valgrind r4316). |
sewardj | bfa2b99 | 2005-07-27 17:49:17 +0000 | [diff] [blame] | 621 | |
| 622 | |
| 623 | |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 624 | Stable release 2.4.0 (March 2005) -- CHANGES RELATIVE TO 2.2.0 |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 625 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 626 | 2.4.0 brings many significant changes and bug fixes. The most |
| 627 | significant user-visible change is that we no longer supply our own |
| 628 | pthread implementation. Instead, Valgrind is finally capable of |
| 629 | running the native thread library, either LinuxThreads or NPTL. |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 630 | |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 631 | This means our libpthread has gone, along with the bugs associated |
| 632 | with it. Valgrind now supports the kernel's threading syscalls, and |
| 633 | lets you use your standard system libpthread. As a result: |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 634 | |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 635 | * There are many fewer system dependencies and strange library-related |
| 636 | bugs. There is a small performance improvement, and a large |
| 637 | stability improvement. |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 638 | |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 639 | * On the downside, Valgrind can no longer report misuses of the POSIX |
| 640 | PThreads API. It also means that Helgrind currently does not work. |
| 641 | We hope to fix these problems in a future release. |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 642 | |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 643 | Note that running the native thread libraries does not mean Valgrind |
| 644 | is able to provide genuine concurrent execution on SMPs. We still |
| 645 | impose the restriction that only one thread is running at any given |
| 646 | time. |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 647 | |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 648 | There are many other significant changes too: |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 649 | |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 650 | * Memcheck is (once again) the default tool. |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 651 | |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 652 | * The default stack backtrace is now 12 call frames, rather than 4. |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 653 | |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 654 | * Suppressions can have up to 25 call frame matches, rather than 4. |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 655 | |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 656 | * Memcheck and Addrcheck use less memory. Under some circumstances, |
| 657 | they no longer allocate shadow memory if there are large regions of |
| 658 | memory with the same A/V states - such as an mmaped file. |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 659 | |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 660 | * The memory-leak detector in Memcheck and Addrcheck has been |
| 661 | improved. It now reports more types of memory leak, including |
| 662 | leaked cycles. When reporting leaked memory, it can distinguish |
| 663 | between directly leaked memory (memory with no references), and |
| 664 | indirectly leaked memory (memory only referred to by other leaked |
| 665 | memory). |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 666 | |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 667 | * Memcheck's confusion over the effect of mprotect() has been fixed: |
| 668 | previously mprotect could erroneously mark undefined data as |
| 669 | defined. |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 670 | |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 671 | * Signal handling is much improved and should be very close to what |
| 672 | you get when running natively. |
| 673 | |
| 674 | One result of this is that Valgrind observes changes to sigcontexts |
| 675 | passed to signal handlers. Such modifications will take effect when |
| 676 | the signal returns. You will need to run with --single-step=yes to |
| 677 | make this useful. |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 678 | |
| 679 | * Valgrind is built in Position Independent Executable (PIE) format if |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 680 | your toolchain supports it. This allows it to take advantage of all |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 681 | the available address space on systems with 4Gbyte user address |
| 682 | spaces. |
| 683 | |
| 684 | * Valgrind can now run itself (requires PIE support). |
| 685 | |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 686 | * Syscall arguments are now checked for validity. Previously all |
| 687 | memory used by syscalls was checked, but now the actual values |
| 688 | passed are also checked. |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 689 | |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 690 | * Syscall wrappers are more robust against bad addresses being passed |
| 691 | to syscalls: they will fail with EFAULT rather than killing Valgrind |
| 692 | with SIGSEGV. |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 693 | |
njn | 1aa5ae4 | 2005-03-26 16:04:27 +0000 | [diff] [blame] | 694 | * Because clone() is directly supported, some non-pthread uses of it |
| 695 | will work. Partial sharing (where some resources are shared, and |
| 696 | some are not) is not supported. |
| 697 | |
| 698 | * open() and readlink() on /proc/self/exe are supported. |
| 699 | |
| 700 | BUGS FIXED: |
| 701 | |
| 702 | 88520 pipe+fork+dup2 kills the main program |
| 703 | 88604 Valgrind Aborts when using $VALGRIND_OPTS and user progra... |
| 704 | 88614 valgrind: vg_libpthread.c:2323 (read): Assertion `read_pt... |
| 705 | 88703 Stabs parser fails to handle ";" |
| 706 | 88886 ioctl wrappers for TIOCMBIS and TIOCMBIC |
| 707 | 89032 valgrind pthread_cond_timedwait fails |
| 708 | 89106 the 'impossible' happened |
| 709 | 89139 Missing sched_setaffinity & sched_getaffinity |
| 710 | 89198 valgrind lacks support for SIOCSPGRP and SIOCGPGRP |
| 711 | 89263 Missing ioctl translations for scsi-generic and CD playing |
| 712 | 89440 tests/deadlock.c line endings |
| 713 | 89481 `impossible' happened: EXEC FAILED |
| 714 | 89663 valgrind 2.2.0 crash on Redhat 7.2 |
| 715 | 89792 Report pthread_mutex_lock() deadlocks instead of returnin... |
| 716 | 90111 statvfs64 gives invalid error/warning |
| 717 | 90128 crash+memory fault with stabs generated by gnat for a run... |
| 718 | 90778 VALGRIND_CHECK_DEFINED() not as documented in memcheck.h |
| 719 | 90834 cachegrind crashes at end of program without reporting re... |
| 720 | 91028 valgrind: vg_memory.c:229 (vgPlain_unmap_range): Assertio... |
| 721 | 91162 valgrind crash while debugging drivel 1.2.1 |
| 722 | 91199 Unimplemented function |
| 723 | 91325 Signal routing does not propagate the siginfo structure |
| 724 | 91599 Assertion `cv == ((void *)0)' |
| 725 | 91604 rw_lookup clears orig and sends the NULL value to rw_new |
| 726 | 91821 Small problems building valgrind with $top_builddir ne $t... |
| 727 | 91844 signal 11 (SIGSEGV) at get_tcb (libpthread.c:86) in corec... |
| 728 | 92264 UNIMPLEMENTED FUNCTION: pthread_condattr_setpshared |
| 729 | 92331 per-target flags necessitate AM_PROG_CC_C_O |
| 730 | 92420 valgrind doesn't compile with linux 2.6.8.1/9 |
| 731 | 92513 Valgrind 2.2.0 generates some warning messages |
| 732 | 92528 vg_symtab2.c:170 (addLoc): Assertion `loc->size > 0' failed. |
| 733 | 93096 unhandled ioctl 0x4B3A and 0x5601 |
| 734 | 93117 Tool and core interface versions do not match |
| 735 | 93128 Can't run valgrind --tool=memcheck because of unimplement... |
| 736 | 93174 Valgrind can crash if passed bad args to certain syscalls |
| 737 | 93309 Stack frame in new thread is badly aligned |
| 738 | 93328 Wrong types used with sys_sigprocmask() |
| 739 | 93763 /usr/include/asm/msr.h is missing |
| 740 | 93776 valgrind: vg_memory.c:508 (vgPlain_find_map_space): Asser... |
| 741 | 93810 fcntl() argument checking a bit too strict |
| 742 | 94378 Assertion `tst->sigqueue_head != tst->sigqueue_tail' failed. |
| 743 | 94429 valgrind 2.2.0 segfault with mmap64 in glibc 2.3.3 |
| 744 | 94645 Impossible happened: PINSRW mem |
| 745 | 94953 valgrind: the `impossible' happened: SIGSEGV |
| 746 | 95667 Valgrind does not work with any KDE app |
| 747 | 96243 Assertion 'res==0' failed |
| 748 | 96252 stage2 loader of valgrind fails to allocate memory |
| 749 | 96520 All programs crashing at _dl_start (in /lib/ld-2.3.3.so) ... |
| 750 | 96660 ioctl CDROMREADTOCENTRY causes bogus warnings |
| 751 | 96747 After looping in a segfault handler, the impossible happens |
| 752 | 96923 Zero sized arrays crash valgrind trace back with SIGFPE |
| 753 | 96948 valgrind stops with assertion failure regarding mmap2 |
| 754 | 96966 valgrind fails when application opens more than 16 sockets |
| 755 | 97398 valgrind: vg_libpthread.c:2667 Assertion failed |
| 756 | 97407 valgrind: vg_mylibc.c:1226 (vgPlain_safe_fd): Assertion `... |
| 757 | 97427 "Warning: invalid file descriptor -1 in syscall close()" ... |
| 758 | 97785 missing backtrace |
| 759 | 97792 build in obj dir fails - autoconf / makefile cleanup |
| 760 | 97880 pthread_mutex_lock fails from shared library (special ker... |
| 761 | 97975 program aborts without ang VG messages |
| 762 | 98129 Failed when open and close file 230000 times using stdio |
| 763 | 98175 Crashes when using valgrind-2.2.0 with a program using al... |
| 764 | 98288 Massif broken |
| 765 | 98303 UNIMPLEMENTED FUNCTION pthread_condattr_setpshared |
| 766 | 98630 failed--compilation missing warnings.pm, fails to make he... |
| 767 | 98756 Cannot valgrind signal-heavy kdrive X server |
| 768 | 98966 valgrinding the JVM fails with a sanity check assertion |
| 769 | 99035 Valgrind crashes while profiling |
| 770 | 99142 loops with message "Signal 11 being dropped from thread 0... |
| 771 | 99195 threaded apps crash on thread start (using QThread::start... |
| 772 | 99348 Assertion `vgPlain_lseek(core_fd, 0, 1) == phdrs[i].p_off... |
| 773 | 99568 False negative due to mishandling of mprotect |
| 774 | 99738 valgrind memcheck crashes on program that uses sigitimer |
| 775 | 99923 0-sized allocations are reported as leaks |
| 776 | 99949 program seg faults after exit() |
| 777 | 100036 "newSuperblock's request for 1048576 bytes failed" |
| 778 | 100116 valgrind: (pthread_cond_init): Assertion `sizeof(* cond) ... |
| 779 | 100486 memcheck reports "valgrind: the `impossible' happened: V... |
| 780 | 100833 second call to "mremap" fails with EINVAL |
| 781 | 101156 (vgPlain_find_map_space): Assertion `(addr & ((1 << 12)-1... |
| 782 | 101173 Assertion `recDepth >= 0 && recDepth < 500' failed |
| 783 | 101291 creating threads in a forked process fails |
| 784 | 101313 valgrind causes different behavior when resizing a window... |
| 785 | 101423 segfault for c++ array of floats |
| 786 | 101562 valgrind massif dies on SIGINT even with signal handler r... |
| 787 | |
njn | 8c92730 | 2005-03-12 16:45:01 +0000 | [diff] [blame] | 788 | |
jseward | 34ed74f | 2004-08-30 18:04:42 +0000 | [diff] [blame] | 789 | Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.0.0 |
| 790 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
jseward | 7d180d9 | 2004-08-30 18:15:37 +0000 | [diff] [blame] | 791 | 2.2.0 brings nine months worth of improvements and bug fixes. We |
| 792 | believe it to be a worthy successor to 2.0.0. There are literally |
| 793 | hundreds of bug fixes and minor improvements. There are also some |
| 794 | fairly major user-visible changes: |
| 795 | |
| 796 | * A complete overhaul of handling of system calls and signals, and |
| 797 | their interaction with threads. In general, the accuracy of the |
| 798 | system call, thread and signal simulations is much improved: |
| 799 | |
| 800 | - Blocking system calls behave exactly as they do when running |
| 801 | natively (not on valgrind). That is, if a syscall blocks only the |
| 802 | calling thread when running natively, than it behaves the same on |
| 803 | valgrind. No more mysterious hangs because V doesn't know that some |
| 804 | syscall or other, should block only the calling thread. |
| 805 | |
| 806 | - Interrupted syscalls should now give more faithful results. |
| 807 | |
| 808 | - Signal contexts in signal handlers are supported. |
| 809 | |
| 810 | * Improvements to NPTL support to the extent that V now works |
| 811 | properly on NPTL-only setups. |
| 812 | |
| 813 | * Greater isolation between Valgrind and the program being run, so |
| 814 | the program is less likely to inadvertently kill Valgrind by |
| 815 | doing wild writes. |
| 816 | |
| 817 | * Massif: a new space profiling tool. Try it! It's cool, and it'll |
| 818 | tell you in detail where and when your C/C++ code is allocating heap. |
| 819 | Draws pretty .ps pictures of memory use against time. A potentially |
| 820 | powerful tool for making sense of your program's space use. |
| 821 | |
| 822 | * File descriptor leakage checks. When enabled, Valgrind will print out |
| 823 | a list of open file descriptors on exit. |
| 824 | |
| 825 | * Improved SSE2/SSE3 support. |
| 826 | |
jseward | 2eb002b | 2004-08-31 00:14:02 +0000 | [diff] [blame] | 827 | * Time-stamped output; use --time-stamp=yes |
| 828 | |
jseward | 34ed74f | 2004-08-30 18:04:42 +0000 | [diff] [blame] | 829 | |
| 830 | |
| 831 | Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.1.2 |
| 832 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 833 | 2.2.0 is not much different from 2.1.2, released seven weeks ago. |
| 834 | A number of bugs have been fixed, most notably #85658, which gave |
| 835 | problems for quite a few people. There have been many internal |
| 836 | cleanups, but those are not user visible. |
| 837 | |
| 838 | The following bugs have been fixed since 2.1.2: |
| 839 | |
| 840 | 85658 Assert in coregrind/vg_libpthread.c:2326 (open64) != |
| 841 | (void*)0 failed |
| 842 | This bug was reported multiple times, and so the following |
| 843 | duplicates of it are also fixed: 87620, 85796, 85935, 86065, |
| 844 | 86919, 86988, 87917, 88156 |
| 845 | |
| 846 | 80716 Semaphore mapping bug caused by unmap (sem_destroy) |
| 847 | (Was fixed prior to 2.1.2) |
| 848 | |
| 849 | 86987 semctl and shmctl syscalls family is not handled properly |
| 850 | |
| 851 | 86696 valgrind 2.1.2 + RH AS2.1 + librt |
| 852 | |
| 853 | 86730 valgrind locks up at end of run with assertion failure |
| 854 | in __pthread_unwind |
| 855 | |
| 856 | 86641 memcheck doesn't work with Mesa OpenGL/ATI on Suse 9.1 |
| 857 | (also fixes 74298, a duplicate of this) |
| 858 | |
| 859 | 85947 MMX/SSE unhandled instruction 'sfence' |
| 860 | |
| 861 | 84978 Wrong error "Conditional jump or move depends on |
| 862 | uninitialised value" resulting from "sbbl %reg, %reg" |
| 863 | |
| 864 | 86254 ssort() fails when signed int return type from comparison is |
| 865 | too small to handle result of unsigned int subtraction |
| 866 | |
| 867 | 87089 memalign( 4, xxx) makes valgrind assert |
| 868 | |
jseward | 2eb002b | 2004-08-31 00:14:02 +0000 | [diff] [blame] | 869 | 86407 Add support for low-level parallel port driver ioctls. |
jseward | 34ed74f | 2004-08-30 18:04:42 +0000 | [diff] [blame] | 870 | |
| 871 | 70587 Add timestamps to Valgrind output? (wishlist) |
| 872 | |
| 873 | 84937 vg_libpthread.c:2505 (se_remap): Assertion `res == 0' |
| 874 | (fixed prior to 2.1.2) |
| 875 | |
| 876 | 86317 cannot load libSDL-1.2.so.0 using valgrind |
| 877 | |
| 878 | 86989 memcpy from mac_replace_strmem.c complains about |
| 879 | uninitialized pointers passed when length to copy is zero |
| 880 | |
| 881 | 85811 gnu pascal symbol causes segmentation fault; ok in 2.0.0 |
| 882 | |
| 883 | 79138 writing to sbrk()'d memory causes segfault |
| 884 | |
| 885 | 77369 sched deadlock while signal received during pthread_join |
| 886 | and the joined thread exited |
| 887 | |
| 888 | 88115 In signal handler for SIGFPE, siginfo->si_addr is wrong |
| 889 | under Valgrind |
| 890 | |
| 891 | 78765 Massif crashes on app exit if FP exceptions are enabled |
| 892 | |
| 893 | Additionally there are the following changes, which are not |
| 894 | connected to any bug report numbers, AFAICS: |
| 895 | |
| 896 | * Fix scary bug causing mis-identification of SSE stores vs |
| 897 | loads and so causing memcheck to sometimes give nonsense results |
| 898 | on SSE code. |
| 899 | |
| 900 | * Add support for the POSIX message queue system calls. |
| 901 | |
| 902 | * Fix to allow 32-bit Valgrind to run on AMD64 boxes. Note: this does |
| 903 | NOT allow Valgrind to work with 64-bit executables - only with 32-bit |
| 904 | executables on an AMD64 box. |
| 905 | |
| 906 | * At configure time, only check whether linux/mii.h can be processed |
| 907 | so that we don't generate ugly warnings by trying to compile it. |
| 908 | |
jseward | 2eb002b | 2004-08-31 00:14:02 +0000 | [diff] [blame] | 909 | * Add support for POSIX clocks and timers. |
| 910 | |
jseward | 34ed74f | 2004-08-30 18:04:42 +0000 | [diff] [blame] | 911 | |
| 912 | |
jseward | 9cd978c | 2004-07-17 13:38:12 +0000 | [diff] [blame] | 913 | Developer (cvs head) release 2.1.2 (18 July 2004) |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 914 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 915 | 2.1.2 contains four months worth of bug fixes and refinements. |
jseward | 9cd978c | 2004-07-17 13:38:12 +0000 | [diff] [blame] | 916 | Although officially a developer release, we believe it to be stable |
| 917 | enough for widespread day-to-day use. 2.1.2 is pretty good, so try it |
| 918 | first, although there is a chance it won't work. If so then try 2.0.0 |
| 919 | and tell us what went wrong." 2.1.2 fixes a lot of problems present |
| 920 | in 2.0.0 and is generally a much better product. |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 921 | |
jseward | 9cd978c | 2004-07-17 13:38:12 +0000 | [diff] [blame] | 922 | Relative to 2.1.1, a large number of minor problems with 2.1.1 have |
| 923 | been fixed, and so if you use 2.1.1 you should try 2.1.2. Users of |
| 924 | the last stable release, 2.0.0, might also want to try this release. |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 925 | |
| 926 | The following bugs, and probably many more, have been fixed. These |
| 927 | are listed at http://bugs.kde.org. Reporting a bug for valgrind in |
| 928 | the http://bugs.kde.org is much more likely to get you a fix than |
| 929 | mailing developers directly, so please continue to keep sending bugs |
| 930 | there. |
| 931 | |
| 932 | 76869 Crashes when running any tool under Fedora Core 2 test1 |
| 933 | This fixes the problem with returning from a signal handler |
jseward | 0400d4d | 2004-07-17 23:26:46 +0000 | [diff] [blame] | 934 | when VDSOs are turned off in FC2. |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 935 | |
jseward | 0400d4d | 2004-07-17 23:26:46 +0000 | [diff] [blame] | 936 | 69508 java 1.4.2 client fails with erroneous "stack size too small". |
| 937 | This fix makes more of the pthread stack attribute related |
| 938 | functions work properly. Java still doesn't work though. |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 939 | |
| 940 | 71906 malloc alignment should be 8, not 4 |
| 941 | All memory returned by malloc/new etc is now at least |
| 942 | 8-byte aligned. |
| 943 | |
jseward | a6017a4 | 2004-07-17 23:44:35 +0000 | [diff] [blame] | 944 | 81970 vg_alloc_ThreadState: no free slots available |
| 945 | (closed because the workaround is simple: increase |
| 946 | VG_N_THREADS, rebuild and try again.) |
| 947 | |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 948 | 78514 Conditional jump or move depends on uninitialized value(s) |
| 949 | (a slight mishanding of FP code in memcheck) |
| 950 | |
jseward | a6017a4 | 2004-07-17 23:44:35 +0000 | [diff] [blame] | 951 | 77952 pThread Support (crash) (due to initialisation-ordering probs) |
| 952 | (also 85118) |
| 953 | |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 954 | 80942 Addrcheck wasn't doing overlap checking as it should. |
| 955 | 78048 return NULL on malloc/new etc failure, instead of asserting |
| 956 | 73655 operator new() override in user .so files often doesn't get picked up |
| 957 | 83060 Valgrind does not handle native kernel AIO |
| 958 | 69872 Create proper coredumps after fatal signals |
| 959 | 82026 failure with new glibc versions: __libc_* functions are not exported |
| 960 | 70344 UNIMPLEMENTED FUNCTION: tcdrain |
| 961 | 81297 Cancellation of pthread_cond_wait does not require mutex |
| 962 | 82872 Using debug info from additional packages (wishlist) |
| 963 | 83025 Support for ioctls FIGETBSZ and FIBMAP |
| 964 | 83340 Support for ioctl HDIO_GET_IDENTITY |
| 965 | 79714 Support for the semtimedop system call. |
| 966 | 77022 Support for ioctls FBIOGET_VSCREENINFO and FBIOGET_FSCREENINFO |
| 967 | 82098 hp2ps ansification (wishlist) |
| 968 | 83573 Valgrind SIGSEGV on execve |
| 969 | 82999 show which cmdline option was erroneous (wishlist) |
| 970 | 83040 make valgrind VPATH and distcheck-clean (wishlist) |
jseward | a6017a4 | 2004-07-17 23:44:35 +0000 | [diff] [blame] | 971 | 83998 Assertion `newfd > vgPlain_max_fd' failed (see below) |
| 972 | 82722 Unchecked mmap in as_pad leads to mysterious failures later |
| 973 | 78958 memcheck seg faults while running Mozilla |
jseward | a2c0800 | 2004-07-18 09:24:57 +0000 | [diff] [blame] | 974 | 85416 Arguments with colon (e.g. --logsocket) ignored |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 975 | |
| 976 | |
| 977 | Additionally there are the following changes, which are not |
| 978 | connected to any bug report numbers, AFAICS: |
| 979 | |
| 980 | * Rearranged address space layout relative to 2.1.1, so that |
| 981 | Valgrind/tools will run out of memory later than currently in many |
| 982 | circumstances. This is good news esp. for Calltree. It should |
| 983 | be possible for client programs to allocate over 800MB of |
| 984 | memory when using memcheck now. |
| 985 | |
jseward | 9cd978c | 2004-07-17 13:38:12 +0000 | [diff] [blame] | 986 | * Improved checking when laying out memory. Should hopefully avoid |
| 987 | the random segmentation faults that 2.1.1 sometimes caused. |
| 988 | |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 989 | * Support for Fedora Core 2 and SuSE 9.1. Improvements to NPTL |
| 990 | support to the extent that V now works properly on NPTL-only setups. |
| 991 | |
| 992 | * Renamed the following options: |
| 993 | --logfile-fd --> --log-fd |
| 994 | --logfile --> --log-file |
| 995 | --logsocket --> --log-socket |
| 996 | to be consistent with each other and other options (esp. --input-fd). |
| 997 | |
| 998 | * Add support for SIOCGMIIPHY, SIOCGMIIREG and SIOCSMIIREG ioctls and |
| 999 | improve the checking of other interface related ioctls. |
| 1000 | |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 1001 | * Fix building with gcc-3.4.1. |
| 1002 | |
| 1003 | * Remove limit on number of semaphores supported. |
| 1004 | |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 1005 | * Add support for syscalls: set_tid_address (258), acct (51). |
| 1006 | |
| 1007 | * Support instruction "repne movs" -- not official but seems to occur. |
| 1008 | |
| 1009 | * Implement an emulated soft limit for file descriptors in addition to |
| 1010 | the current reserved area, which effectively acts as a hard limit. The |
| 1011 | setrlimit system call now simply updates the emulated limits as best |
| 1012 | as possible - the hard limit is not allowed to move at all and just |
| 1013 | returns EPERM if you try and change it. This should stop reductions |
| 1014 | in the soft limit causing assertions when valgrind tries to allocate |
| 1015 | descriptors from the reserved area. |
jseward | a6017a4 | 2004-07-17 23:44:35 +0000 | [diff] [blame] | 1016 | (This actually came from bug #83998). |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 1017 | |
nethercote | 76e8fd9 | 2004-07-21 10:37:31 +0000 | [diff] [blame] | 1018 | * Major overhaul of Cachegrind implementation. First user-visible change |
nethercote | 5e63bfb | 2004-07-17 12:49:53 +0000 | [diff] [blame] | 1019 | is that cachegrind.out files are now typically 90% smaller than they |
nethercote | 76e8fd9 | 2004-07-21 10:37:31 +0000 | [diff] [blame] | 1020 | used to be; code annotation times are correspondingly much smaller. |
| 1021 | Second user-visible change is that hit/miss counts for code that is |
| 1022 | unloaded at run-time is no longer dumped into a single "discard" pile, |
| 1023 | but accurately preserved. |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 1024 | |
| 1025 | * Client requests for telling valgrind about memory pools. |
| 1026 | |
| 1027 | |
| 1028 | |
jseward | 9cd978c | 2004-07-17 13:38:12 +0000 | [diff] [blame] | 1029 | Developer (cvs head) release 2.1.1 (12 March 2004) |
| 1030 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 1031 | 2.1.1 contains some internal structural changes needed for V's |
jseward | e3e7704 | 2004-03-12 21:07:05 +0000 | [diff] [blame] | 1032 | long-term future. These don't affect end-users. Most notable |
| 1033 | user-visible changes are: |
| 1034 | |
| 1035 | * Greater isolation between Valgrind and the program being run, so |
| 1036 | the program is less likely to inadvertently kill Valgrind by |
| 1037 | doing wild writes. |
| 1038 | |
| 1039 | * Massif: a new space profiling tool. Try it! It's cool, and it'll |
| 1040 | tell you in detail where and when your C/C++ code is allocating heap. |
| 1041 | Draws pretty .ps pictures of memory use against time. A potentially |
| 1042 | powerful tool for making sense of your program's space use. |
| 1043 | |
| 1044 | * Fixes for many bugs, including support for more SSE2/SSE3 instructions, |
| 1045 | various signal/syscall things, and various problems with debug |
| 1046 | info readers. |
| 1047 | |
| 1048 | * Support for glibc-2.3.3 based systems. |
| 1049 | |
| 1050 | We are now doing automatic overnight build-and-test runs on a variety |
| 1051 | of distros. As a result, we believe 2.1.1 builds and runs on: |
| 1052 | Red Hat 7.2, 7.3, 8.0, 9, Fedora Core 1, SuSE 8.2, SuSE 9. |
| 1053 | |
| 1054 | |
| 1055 | The following bugs, and probably many more, have been fixed. These |
| 1056 | are listed at http://bugs.kde.org. Reporting a bug for valgrind in |
| 1057 | the http://bugs.kde.org is much more likely to get you a fix than |
| 1058 | mailing developers directly, so please continue to keep sending bugs |
| 1059 | there. |
| 1060 | |
| 1061 | 69616 glibc 2.3.2 w/NPTL is massively different than what valgrind expects |
| 1062 | 69856 I don't know how to instrument MMXish stuff (Helgrind) |
| 1063 | 73892 valgrind segfaults starting with Objective-C debug info |
| 1064 | (fix for S-type stabs) |
| 1065 | 73145 Valgrind complains too much about close(<reserved fd>) |
| 1066 | 73902 Shadow memory allocation seems to fail on RedHat 8.0 |
| 1067 | 68633 VG_N_SEMAPHORES too low (V itself was leaking semaphores) |
| 1068 | 75099 impossible to trace multiprocess programs |
| 1069 | 76839 the `impossible' happened: disInstr: INT but not 0x80 ! |
| 1070 | 76762 vg_to_ucode.c:3748 (dis_push_segreg): Assertion `sz == 4' failed. |
| 1071 | 76747 cannot include valgrind.h in c++ program |
| 1072 | 76223 parsing B(3,10) gave NULL type => impossible happens |
| 1073 | 75604 shmdt handling problem |
| 1074 | 76416 Problems with gcc 3.4 snap 20040225 |
| 1075 | 75614 using -gstabs when building your programs the `impossible' happened |
| 1076 | 75787 Patch for some CDROM ioctls CDORM_GET_MCN, CDROM_SEND_PACKET, |
| 1077 | 75294 gcc 3.4 snapshot's libstdc++ have unsupported instructions. |
| 1078 | (REP RET) |
| 1079 | 73326 vg_symtab2.c:272 (addScopeRange): Assertion `range->size > 0' failed. |
| 1080 | 72596 not recognizing __libc_malloc |
| 1081 | 69489 Would like to attach ddd to running program |
| 1082 | 72781 Cachegrind crashes with kde programs |
| 1083 | 73055 Illegal operand at DXTCV11CompressBlockSSE2 (more SSE opcodes) |
| 1084 | 73026 Descriptor leak check reports port numbers wrongly |
| 1085 | 71705 README_MISSING_SYSCALL_OR_IOCTL out of date |
| 1086 | 72643 Improve support for SSE/SSE2 instructions |
| 1087 | 72484 valgrind leaves it's own signal mask in place when execing |
| 1088 | 72650 Signal Handling always seems to restart system calls |
| 1089 | 72006 The mmap system call turns all errors in ENOMEM |
| 1090 | 71781 gdb attach is pretty useless |
| 1091 | 71180 unhandled instruction bytes: 0xF 0xAE 0x85 0xE8 |
| 1092 | 69886 writes to zero page cause valgrind to assert on exit |
| 1093 | 71791 crash when valgrinding gimp 1.3 (stabs reader problem) |
| 1094 | 69783 unhandled syscall: 218 |
| 1095 | 69782 unhandled instruction bytes: 0x66 0xF 0x2B 0x80 |
| 1096 | 70385 valgrind fails if the soft file descriptor limit is less |
| 1097 | than about 828 |
| 1098 | 69529 "rep; nop" should do a yield |
| 1099 | 70827 programs with lots of shared libraries report "mmap failed" |
| 1100 | for some of them when reading symbols |
| 1101 | 71028 glibc's strnlen is optimised enough to confuse valgrind |
| 1102 | |
| 1103 | |
| 1104 | |
| 1105 | |
jseward | 4323065 | 2003-12-14 00:14:54 +0000 | [diff] [blame] | 1106 | Unstable (cvs head) release 2.1.0 (15 December 2003) |
| 1107 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 1108 | For whatever it's worth, 2.1.0 actually seems pretty darn stable to me |
| 1109 | (Julian). It looks eminently usable, and given that it fixes some |
| 1110 | significant bugs, may well be worth using on a day-to-day basis. |
| 1111 | 2.1.0 is known to build and pass regression tests on: SuSE 9, SuSE |
| 1112 | 8.2, RedHat 8. |
| 1113 | |
| 1114 | 2.1.0 most notably includes Jeremy Fitzhardinge's complete overhaul of |
| 1115 | handling of system calls and signals, and their interaction with |
| 1116 | threads. In general, the accuracy of the system call, thread and |
| 1117 | signal simulations is much improved. Specifically: |
| 1118 | |
| 1119 | - Blocking system calls behave exactly as they do when running |
| 1120 | natively (not on valgrind). That is, if a syscall blocks only the |
| 1121 | calling thread when running natively, than it behaves the same on |
| 1122 | valgrind. No more mysterious hangs because V doesn't know that some |
| 1123 | syscall or other, should block only the calling thread. |
| 1124 | |
| 1125 | - Interrupted syscalls should now give more faithful results. |
| 1126 | |
| 1127 | - Finally, signal contexts in signal handlers are supported. As a |
| 1128 | result, konqueror on SuSE 9 no longer segfaults when notified of |
| 1129 | file changes in directories it is watching. |
| 1130 | |
| 1131 | Other changes: |
| 1132 | |
| 1133 | - Robert Walsh's file descriptor leakage checks. When enabled, |
| 1134 | Valgrind will print out a list of open file descriptors on |
| 1135 | exit. Along with each file descriptor, Valgrind prints out a stack |
| 1136 | backtrace of where the file was opened and any details relating to the |
| 1137 | file descriptor such as the file name or socket details. |
| 1138 | To use, give: --track-fds=yes |
| 1139 | |
| 1140 | - Implemented a few more SSE/SSE2 instructions. |
| 1141 | |
| 1142 | - Less crud on the stack when you do 'where' inside a GDB attach. |
| 1143 | |
| 1144 | - Fixed the following bugs: |
| 1145 | 68360: Valgrind does not compile against 2.6.0-testX kernels |
| 1146 | 68525: CVS head doesn't compile on C90 compilers |
| 1147 | 68566: pkgconfig support (wishlist) |
| 1148 | 68588: Assertion `sz == 4' failed in vg_to_ucode.c (disInstr) |
| 1149 | 69140: valgrind not able to explicitly specify a path to a binary. |
| 1150 | 69432: helgrind asserts encountering a MutexErr when there are |
| 1151 | EraserErr suppressions |
| 1152 | |
jseward | 0b12daf | 2003-12-14 14:35:28 +0000 | [diff] [blame] | 1153 | - Increase the max size of the translation cache from 200k average bbs |
| 1154 | to 300k average bbs. Programs on the size of OOo (680m17) are |
| 1155 | thrashing the cache at the smaller size, creating large numbers of |
| 1156 | retranslations and wasting significant time as a result. |
| 1157 | |
jseward | 4323065 | 2003-12-14 00:14:54 +0000 | [diff] [blame] | 1158 | |
| 1159 | |
| 1160 | Stable release 2.0.0 (5 Nov 2003) |
| 1161 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 1162 | |
| 1163 | 2.0.0 improves SSE/SSE2 support, fixes some minor bugs, and |
| 1164 | improves support for SuSE 9 and the Red Hat "Severn" beta. |
| 1165 | |
| 1166 | - Further improvements to SSE/SSE2 support. The entire test suite of |
| 1167 | the GNU Scientific Library (gsl-1.4) compiled with Intel Icc 7.1 |
| 1168 | 20030307Z '-g -O -xW' now works. I think this gives pretty good |
| 1169 | coverage of SSE/SSE2 floating point instructions, or at least the |
| 1170 | subset emitted by Icc. |
| 1171 | |
| 1172 | - Also added support for the following instructions: |
| 1173 | MOVNTDQ UCOMISD UNPCKLPS UNPCKHPS SQRTSS |
| 1174 | PUSH/POP %{FS,GS}, and PUSH %CS (Nb: there is no POP %CS). |
| 1175 | |
| 1176 | - CFI support for GDB version 6. Needed to enable newer GDBs |
| 1177 | to figure out where they are when using --gdb-attach=yes. |
| 1178 | |
| 1179 | - Fix this: |
| 1180 | mc_translate.c:1091 (memcheck_instrument): Assertion |
| 1181 | `u_in->size == 4 || u_in->size == 16' failed. |
| 1182 | |
| 1183 | - Return an error rather than panicing when given a bad socketcall. |
| 1184 | |
| 1185 | - Fix checking of syscall rt_sigtimedwait(). |
| 1186 | |
| 1187 | - Implement __NR_clock_gettime (syscall 265). Needed on Red Hat Severn. |
| 1188 | |
| 1189 | - Fixed bug in overlap check in strncpy() -- it was assuming the src was 'n' |
| 1190 | bytes long, when it could be shorter, which could cause false |
| 1191 | positives. |
| 1192 | |
| 1193 | - Support use of select() for very large numbers of file descriptors. |
| 1194 | |
| 1195 | - Don't fail silently if the executable is statically linked, or is |
| 1196 | setuid/setgid. Print an error message instead. |
| 1197 | |
| 1198 | - Support for old DWARF-1 format line number info. |
| 1199 | |
| 1200 | |
| 1201 | |
| 1202 | Snapshot 20031012 (12 October 2003) |
| 1203 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 1204 | |
| 1205 | Three months worth of bug fixes, roughly. Most significant single |
| 1206 | change is improved SSE/SSE2 support, mostly thanks to Dirk Mueller. |
| 1207 | |
| 1208 | 20031012 builds on Red Hat Fedora ("Severn") but doesn't really work |
| 1209 | (curiosly, mozilla runs OK, but a modest "ls -l" bombs). I hope to |
| 1210 | get a working version out soon. It may or may not work ok on the |
| 1211 | forthcoming SuSE 9; I hear positive noises about it but haven't been |
| 1212 | able to verify this myself (not until I get hold of a copy of 9). |
| 1213 | |
| 1214 | A detailed list of changes, in no particular order: |
| 1215 | |
| 1216 | - Describe --gen-suppressions in the FAQ. |
| 1217 | |
| 1218 | - Syscall __NR_waitpid supported. |
| 1219 | |
| 1220 | - Minor MMX bug fix. |
| 1221 | |
| 1222 | - -v prints program's argv[] at startup. |
| 1223 | |
| 1224 | - More glibc-2.3 suppressions. |
| 1225 | |
| 1226 | - Suppressions for stack underrun bug(s) in the c++ support library |
| 1227 | distributed with Intel Icc 7.0. |
| 1228 | |
| 1229 | - Fix problems reading /proc/self/maps. |
| 1230 | |
| 1231 | - Fix a couple of messages that should have been suppressed by -q, |
| 1232 | but weren't. |
| 1233 | |
| 1234 | - Make Addrcheck understand "Overlap" suppressions. |
| 1235 | |
| 1236 | - At startup, check if program is statically linked and bail out if so. |
| 1237 | |
| 1238 | - Cachegrind: Auto-detect Intel Pentium-M, also VIA Nehemiah |
| 1239 | |
| 1240 | - Memcheck/addrcheck: minor speed optimisations |
| 1241 | |
| 1242 | - Handle syscall __NR_brk more correctly than before. |
| 1243 | |
| 1244 | - Fixed incorrect allocate/free mismatch errors when using |
| 1245 | operator new(unsigned, std::nothrow_t const&) |
| 1246 | operator new[](unsigned, std::nothrow_t const&) |
| 1247 | |
| 1248 | - Support POSIX pthread spinlocks. |
| 1249 | |
| 1250 | - Fixups for clean compilation with gcc-3.3.1. |
daywalker | 5d945de | 2003-09-26 00:32:53 +0000 | [diff] [blame] | 1251 | |
| 1252 | - Implemented more opcodes: |
| 1253 | - push %es |
| 1254 | - push %ds |
| 1255 | - pop %es |
| 1256 | - pop %ds |
| 1257 | - movntq |
| 1258 | - sfence |
| 1259 | - pshufw |
| 1260 | - pavgb |
| 1261 | - ucomiss |
| 1262 | - enter |
daywalker | b18d253 | 2003-09-27 20:15:01 +0000 | [diff] [blame] | 1263 | - mov imm32, %esp |
| 1264 | - all "in" and "out" opcodes |
daywalker | 79aad84 | 2003-09-30 22:58:12 +0000 | [diff] [blame] | 1265 | - inc/dec %esp |
jseward | 4323065 | 2003-12-14 00:14:54 +0000 | [diff] [blame] | 1266 | - A whole bunch of SSE/SSE2 instructions |
daywalker | 5d945de | 2003-09-26 00:32:53 +0000 | [diff] [blame] | 1267 | |
jseward | 4323065 | 2003-12-14 00:14:54 +0000 | [diff] [blame] | 1268 | - Memcheck: don't bomb on SSE/SSE2 code. |
daywalker | 5d945de | 2003-09-26 00:32:53 +0000 | [diff] [blame] | 1269 | |
sewardj | 9023879 | 2003-05-05 00:23:42 +0000 | [diff] [blame] | 1270 | |
sewardj | 945f39f | 2003-07-25 21:11:40 +0000 | [diff] [blame] | 1271 | Snapshot 20030725 (25 July 2003) |
| 1272 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 1273 | |
| 1274 | Fixes some minor problems in 20030716. |
| 1275 | |
| 1276 | - Fix bugs in overlap checking for strcpy/memcpy etc. |
| 1277 | |
| 1278 | - Do overlap checking with Addrcheck as well as Memcheck. |
| 1279 | |
| 1280 | - Fix this: |
| 1281 | Memcheck: the `impossible' happened: |
| 1282 | get_error_name: unexpected type |
| 1283 | |
| 1284 | - Install headers needed to compile new skins. |
| 1285 | |
sewardj | e3dd2e0 | 2003-07-27 08:04:48 +0000 | [diff] [blame] | 1286 | - Remove leading spaces and colon in the LD_LIBRARY_PATH / LD_PRELOAD |
sewardj | 945f39f | 2003-07-25 21:11:40 +0000 | [diff] [blame] | 1287 | passed to non-traced children. |
| 1288 | |
sewardj | b9eda95 | 2003-07-26 21:39:05 +0000 | [diff] [blame] | 1289 | - Fix file descriptor leak in valgrind-listener. |
| 1290 | |
sewardj | e3dd2e0 | 2003-07-27 08:04:48 +0000 | [diff] [blame] | 1291 | - Fix longstanding bug in which the allocation point of a |
| 1292 | block resized by realloc was not correctly set. This may |
| 1293 | have caused confusing error messages. |
sewardj | 945f39f | 2003-07-25 21:11:40 +0000 | [diff] [blame] | 1294 | |
| 1295 | |
sewardj | 626fd89 | 2003-07-16 20:10:26 +0000 | [diff] [blame] | 1296 | Snapshot 20030716 (16 July 2003) |
sewardj | 9d916ed | 2003-07-14 23:38:40 +0000 | [diff] [blame] | 1297 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 1298 | |
sewardj | 626fd89 | 2003-07-16 20:10:26 +0000 | [diff] [blame] | 1299 | 20030716 is a snapshot of our current CVS head (development) branch. |
sewardj | 9d916ed | 2003-07-14 23:38:40 +0000 | [diff] [blame] | 1300 | This is the branch which will become valgrind-2.0. It contains |
| 1301 | significant enhancements over the 1.9.X branch. |
| 1302 | |
| 1303 | Despite this being a snapshot of the CVS head, it is believed to be |
| 1304 | quite stable -- at least as stable as 1.9.6 or 1.0.4, if not more so |
| 1305 | -- and therefore suitable for widespread use. Please let us know asap |
| 1306 | if it causes problems for you. |
| 1307 | |
| 1308 | Two reasons for releasing a snapshot now are: |
| 1309 | |
| 1310 | - It's been a while since 1.9.6, and this snapshot fixes |
| 1311 | various problems that 1.9.6 has with threaded programs |
| 1312 | on glibc-2.3.X based systems. |
| 1313 | |
| 1314 | - So as to make available improvements in the 2.0 line. |
| 1315 | |
sewardj | 626fd89 | 2003-07-16 20:10:26 +0000 | [diff] [blame] | 1316 | Major changes in 20030716, as compared to 1.9.6: |
sewardj | 9d916ed | 2003-07-14 23:38:40 +0000 | [diff] [blame] | 1317 | |
| 1318 | - More fixes to threading support on glibc-2.3.1 and 2.3.2-based |
| 1319 | systems (SuSE 8.2, Red Hat 9). If you have had problems |
| 1320 | with inconsistent/illogical behaviour of errno, h_errno or the DNS |
sewardj | 626fd89 | 2003-07-16 20:10:26 +0000 | [diff] [blame] | 1321 | resolver functions in threaded programs, 20030716 should improve |
sewardj | 9d916ed | 2003-07-14 23:38:40 +0000 | [diff] [blame] | 1322 | matters. This snapshot seems stable enough to run OpenOffice.org |
| 1323 | 1.1rc on Red Hat 7.3, SuSE 8.2 and Red Hat 9, and that's a big |
| 1324 | threaded app if ever I saw one. |
| 1325 | |
| 1326 | - Automatic generation of suppression records; you no longer |
| 1327 | need to write them by hand. Use --gen-suppressions=yes. |
| 1328 | |
sewardj | 2151180 | 2003-07-22 17:47:42 +0000 | [diff] [blame] | 1329 | - strcpy/memcpy/etc check their arguments for overlaps, when |
| 1330 | running with the Memcheck or Addrcheck skins. |
| 1331 | |
| 1332 | - malloc_usable_size() is now supported. |
| 1333 | |
| 1334 | - new client requests: |
| 1335 | - VALGRIND_COUNT_ERRORS, VALGRIND_COUNT_LEAKS: |
| 1336 | useful with regression testing |
| 1337 | - VALGRIND_NON_SIMD_CALL[0123]: for running arbitrary functions |
| 1338 | on real CPU (use with caution!) |
| 1339 | |
sewardj | 9d916ed | 2003-07-14 23:38:40 +0000 | [diff] [blame] | 1340 | - The GDB attach mechanism is more flexible. Allow the GDB to |
| 1341 | be run to be specified by --gdb-path=/path/to/gdb, and specify |
| 1342 | which file descriptor V will read its input from with |
| 1343 | --input-fd=<number>. |
| 1344 | |
sewardj | 2151180 | 2003-07-22 17:47:42 +0000 | [diff] [blame] | 1345 | - Cachegrind gives more accurate results (wasn't tracking instructions in |
| 1346 | malloc() and friends previously, is now). |
| 1347 | |
sewardj | 9d916ed | 2003-07-14 23:38:40 +0000 | [diff] [blame] | 1348 | - Complete support for the MMX instruction set. |
| 1349 | |
| 1350 | - Partial support for the SSE and SSE2 instruction sets. Work for this |
| 1351 | is ongoing. About half the SSE/SSE2 instructions are done, so |
| 1352 | some SSE based programs may work. Currently you need to specify |
| 1353 | --skin=addrcheck. Basically not suitable for real use yet. |
| 1354 | |
| 1355 | - Significant speedups (10%-20%) for standard memory checking. |
| 1356 | |
| 1357 | - Fix assertion failure in pthread_once(). |
| 1358 | |
| 1359 | - Fix this: |
| 1360 | valgrind: vg_intercept.c:598 (vgAllRoadsLeadToRome_select): |
| 1361 | Assertion `ms_end >= ms_now' failed. |
| 1362 | |
| 1363 | - Implement pthread_mutexattr_setpshared. |
| 1364 | |
| 1365 | - Understand Pentium 4 branch hints. Also implemented a couple more |
| 1366 | obscure x86 instructions. |
| 1367 | |
| 1368 | - Lots of other minor bug fixes. |
| 1369 | |
sewardj | 626fd89 | 2003-07-16 20:10:26 +0000 | [diff] [blame] | 1370 | - We have a decent regression test system, for the first time. |
| 1371 | This doesn't help you directly, but it does make it a lot easier |
| 1372 | for us to track the quality of the system, especially across |
| 1373 | multiple linux distributions. |
| 1374 | |
| 1375 | You can run the regression tests with 'make regtest' after 'make |
| 1376 | install' completes. On SuSE 8.2 and Red Hat 9 I get this: |
| 1377 | |
| 1378 | == 84 tests, 0 stderr failures, 0 stdout failures == |
| 1379 | |
| 1380 | On Red Hat 8, I get this: |
| 1381 | |
| 1382 | == 84 tests, 2 stderr failures, 1 stdout failure == |
| 1383 | corecheck/tests/res_search (stdout) |
| 1384 | memcheck/tests/sigaltstack (stderr) |
| 1385 | |
| 1386 | sigaltstack is probably harmless. res_search doesn't work |
| 1387 | on R H 8 even running natively, so I'm not too worried. |
| 1388 | |
| 1389 | On Red Hat 7.3, a glibc-2.2.5 system, I get these harmless failures: |
| 1390 | |
| 1391 | == 84 tests, 2 stderr failures, 1 stdout failure == |
| 1392 | corecheck/tests/pth_atfork1 (stdout) |
| 1393 | corecheck/tests/pth_atfork1 (stderr) |
| 1394 | memcheck/tests/sigaltstack (stderr) |
| 1395 | |
| 1396 | You need to run on a PII system, at least, since some tests |
| 1397 | contain P6-specific instructions, and the test machine needs |
| 1398 | access to the internet so that corecheck/tests/res_search |
| 1399 | (a test that the DNS resolver works) can function. |
| 1400 | |
sewardj | 9d916ed | 2003-07-14 23:38:40 +0000 | [diff] [blame] | 1401 | As ever, thanks for the vast amount of feedback :) and bug reports :( |
| 1402 | We may not answer all messages, but we do at least look at all of |
| 1403 | them, and tend to fix the most frequently reported bugs. |
| 1404 | |
| 1405 | |
| 1406 | |
sewardj | 3791882 | 2003-05-05 01:05:09 +0000 | [diff] [blame] | 1407 | Version 1.9.6 (7 May 2003 or thereabouts) |
| 1408 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 1409 | |
| 1410 | Major changes in 1.9.6: |
| 1411 | |
| 1412 | - Improved threading support for glibc >= 2.3.2 (SuSE 8.2, |
| 1413 | RedHat 9, to name but two ...) It turned out that 1.9.5 |
| 1414 | had problems with threading support on glibc >= 2.3.2, |
| 1415 | usually manifested by threaded programs deadlocking in system calls, |
| 1416 | or running unbelievably slowly. Hopefully these are fixed now. 1.9.6 |
| 1417 | is the first valgrind which gives reasonable support for |
| 1418 | glibc-2.3.2. Also fixed a 2.3.2 problem with pthread_atfork(). |
| 1419 | |
| 1420 | - Majorly expanded FAQ.txt. We've added workarounds for all |
| 1421 | common problems for which a workaround is known. |
| 1422 | |
| 1423 | Minor changes in 1.9.6: |
| 1424 | |
| 1425 | - Fix identification of the main thread's stack. Incorrect |
| 1426 | identification of it was causing some on-stack addresses to not get |
| 1427 | identified as such. This only affected the usefulness of some error |
| 1428 | messages; the correctness of the checks made is unchanged. |
| 1429 | |
| 1430 | - Support for kernels >= 2.5.68. |
| 1431 | |
| 1432 | - Dummy implementations of __libc_current_sigrtmin, |
| 1433 | __libc_current_sigrtmax and __libc_allocate_rtsig, hopefully |
| 1434 | good enough to keep alive programs which previously died for lack of |
| 1435 | them. |
| 1436 | |
| 1437 | - Fix bug in the VALGRIND_DISCARD_TRANSLATIONS client request. |
| 1438 | |
sewardj | 3d47b79 | 2003-05-05 22:15:35 +0000 | [diff] [blame] | 1439 | - Fix bug in the DWARF2 debug line info loader, when instructions |
| 1440 | following each other have source lines far from each other |
| 1441 | (e.g. with inlined functions). |
| 1442 | |
sewardj | 3791882 | 2003-05-05 01:05:09 +0000 | [diff] [blame] | 1443 | - Debug info reading: read symbols from both "symtab" and "dynsym" |
| 1444 | sections, rather than merely from the one that comes last in the |
| 1445 | file. |
| 1446 | |
| 1447 | - New syscall support: prctl(), creat(), lookup_dcookie(). |
| 1448 | |
| 1449 | - When checking calls to accept(), recvfrom(), getsocketopt(), |
| 1450 | don't complain if buffer values are NULL. |
| 1451 | |
| 1452 | - Try and avoid assertion failures in |
| 1453 | mash_LD_PRELOAD_and_LD_LIBRARY_PATH. |
| 1454 | |
| 1455 | - Minor bug fixes in cg_annotate. |
| 1456 | |
| 1457 | |
| 1458 | |
sewardj | 9023879 | 2003-05-05 00:23:42 +0000 | [diff] [blame] | 1459 | Version 1.9.5 (7 April 2003) |
| 1460 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 1461 | |
| 1462 | It occurs to me that it would be helpful for valgrind users to record |
| 1463 | in the source distribution the changes in each release. So I now |
| 1464 | attempt to mend my errant ways :-) Changes in this and future releases |
| 1465 | will be documented in the NEWS file in the source distribution. |
| 1466 | |
| 1467 | Major changes in 1.9.5: |
| 1468 | |
| 1469 | - (Critical bug fix): Fix a bug in the FPU simulation. This was |
| 1470 | causing some floating point conditional tests not to work right. |
| 1471 | Several people reported this. If you had floating point code which |
| 1472 | didn't work right on 1.9.1 to 1.9.4, it's worth trying 1.9.5. |
| 1473 | |
| 1474 | - Partial support for Red Hat 9. RH9 uses the new Native Posix |
| 1475 | Threads Library (NPTL), instead of the older LinuxThreads. |
| 1476 | This potentially causes problems with V which will take some |
| 1477 | time to correct. In the meantime we have partially worked around |
| 1478 | this, and so 1.9.5 works on RH9. Threaded programs still work, |
| 1479 | but they may deadlock, because some system calls (accept, read, |
| 1480 | write, etc) which should be nonblocking, in fact do block. This |
| 1481 | is a known bug which we are looking into. |
| 1482 | |
| 1483 | If you can, your best bet (unfortunately) is to avoid using |
| 1484 | 1.9.5 on a Red Hat 9 system, or on any NPTL-based distribution. |
| 1485 | If your glibc is 2.3.1 or earlier, you're almost certainly OK. |
| 1486 | |
| 1487 | Minor changes in 1.9.5: |
| 1488 | |
| 1489 | - Added some #errors to valgrind.h to ensure people don't include |
| 1490 | it accidentally in their sources. This is a change from 1.0.X |
| 1491 | which was never properly documented. The right thing to include |
| 1492 | is now memcheck.h. Some people reported problems and strange |
| 1493 | behaviour when (incorrectly) including valgrind.h in code with |
| 1494 | 1.9.1 -- 1.9.4. This is no longer possible. |
| 1495 | |
| 1496 | - Add some __extension__ bits and pieces so that gcc configured |
| 1497 | for valgrind-checking compiles even with -Werror. If you |
| 1498 | don't understand this, ignore it. Of interest to gcc developers |
| 1499 | only. |
| 1500 | |
| 1501 | - Removed a pointless check which caused problems interworking |
| 1502 | with Clearcase. V would complain about shared objects whose |
| 1503 | names did not end ".so", and refuse to run. This is now fixed. |
| 1504 | In fact it was fixed in 1.9.4 but not documented. |
| 1505 | |
| 1506 | - Fixed a bug causing an assertion failure of "waiters == 1" |
| 1507 | somewhere in vg_scheduler.c, when running large threaded apps, |
| 1508 | notably MySQL. |
| 1509 | |
| 1510 | - Add support for the munlock system call (124). |
| 1511 | |
| 1512 | Some comments about future releases: |
| 1513 | |
| 1514 | 1.9.5 is, we hope, the most stable Valgrind so far. It pretty much |
| 1515 | supersedes the 1.0.X branch. If you are a valgrind packager, please |
| 1516 | consider making 1.9.5 available to your users. You can regard the |
| 1517 | 1.0.X branch as obsolete: 1.9.5 is stable and vastly superior. There |
| 1518 | are no plans at all for further releases of the 1.0.X branch. |
| 1519 | |
| 1520 | If you want a leading-edge valgrind, consider building the cvs head |
| 1521 | (from SourceForge), or getting a snapshot of it. Current cool stuff |
| 1522 | going in includes MMX support (done); SSE/SSE2 support (in progress), |
| 1523 | a significant (10-20%) performance improvement (done), and the usual |
| 1524 | large collection of minor changes. Hopefully we will be able to |
| 1525 | improve our NPTL support, but no promises. |
| 1526 | |