jseward | 4323065 | 2003-12-14 00:14:54 +0000 | [diff] [blame] | 1 | |
jseward | 34ed74f | 2004-08-30 18:04:42 +0000 | [diff] [blame] | 2 | Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.0.0 |
| 3 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
jseward | 7d180d9 | 2004-08-30 18:15:37 +0000 | [diff] [blame^] | 4 | 2.2.0 brings nine months worth of improvements and bug fixes. We |
| 5 | believe it to be a worthy successor to 2.0.0. There are literally |
| 6 | hundreds of bug fixes and minor improvements. There are also some |
| 7 | fairly major user-visible changes: |
| 8 | |
| 9 | * A complete overhaul of handling of system calls and signals, and |
| 10 | their interaction with threads. In general, the accuracy of the |
| 11 | system call, thread and signal simulations is much improved: |
| 12 | |
| 13 | - Blocking system calls behave exactly as they do when running |
| 14 | natively (not on valgrind). That is, if a syscall blocks only the |
| 15 | calling thread when running natively, than it behaves the same on |
| 16 | valgrind. No more mysterious hangs because V doesn't know that some |
| 17 | syscall or other, should block only the calling thread. |
| 18 | |
| 19 | - Interrupted syscalls should now give more faithful results. |
| 20 | |
| 21 | - Signal contexts in signal handlers are supported. |
| 22 | |
| 23 | * Improvements to NPTL support to the extent that V now works |
| 24 | properly on NPTL-only setups. |
| 25 | |
| 26 | * Greater isolation between Valgrind and the program being run, so |
| 27 | the program is less likely to inadvertently kill Valgrind by |
| 28 | doing wild writes. |
| 29 | |
| 30 | * Massif: a new space profiling tool. Try it! It's cool, and it'll |
| 31 | tell you in detail where and when your C/C++ code is allocating heap. |
| 32 | Draws pretty .ps pictures of memory use against time. A potentially |
| 33 | powerful tool for making sense of your program's space use. |
| 34 | |
| 35 | * File descriptor leakage checks. When enabled, Valgrind will print out |
| 36 | a list of open file descriptors on exit. |
| 37 | |
| 38 | * Improved SSE2/SSE3 support. |
| 39 | |
jseward | 34ed74f | 2004-08-30 18:04:42 +0000 | [diff] [blame] | 40 | |
| 41 | |
| 42 | Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.1.2 |
| 43 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 44 | 2.2.0 is not much different from 2.1.2, released seven weeks ago. |
| 45 | A number of bugs have been fixed, most notably #85658, which gave |
| 46 | problems for quite a few people. There have been many internal |
| 47 | cleanups, but those are not user visible. |
| 48 | |
| 49 | The following bugs have been fixed since 2.1.2: |
| 50 | |
| 51 | 85658 Assert in coregrind/vg_libpthread.c:2326 (open64) != |
| 52 | (void*)0 failed |
| 53 | This bug was reported multiple times, and so the following |
| 54 | duplicates of it are also fixed: 87620, 85796, 85935, 86065, |
| 55 | 86919, 86988, 87917, 88156 |
| 56 | |
| 57 | 80716 Semaphore mapping bug caused by unmap (sem_destroy) |
| 58 | (Was fixed prior to 2.1.2) |
| 59 | |
| 60 | 86987 semctl and shmctl syscalls family is not handled properly |
| 61 | |
| 62 | 86696 valgrind 2.1.2 + RH AS2.1 + librt |
| 63 | |
| 64 | 86730 valgrind locks up at end of run with assertion failure |
| 65 | in __pthread_unwind |
| 66 | |
| 67 | 86641 memcheck doesn't work with Mesa OpenGL/ATI on Suse 9.1 |
| 68 | (also fixes 74298, a duplicate of this) |
| 69 | |
| 70 | 85947 MMX/SSE unhandled instruction 'sfence' |
| 71 | |
| 72 | 84978 Wrong error "Conditional jump or move depends on |
| 73 | uninitialised value" resulting from "sbbl %reg, %reg" |
| 74 | |
| 75 | 86254 ssort() fails when signed int return type from comparison is |
| 76 | too small to handle result of unsigned int subtraction |
| 77 | |
| 78 | 87089 memalign( 4, xxx) makes valgrind assert |
| 79 | |
| 80 | 86407 Add partial support for the low-level parallel port driver ioctls. |
| 81 | |
| 82 | 70587 Add timestamps to Valgrind output? (wishlist) |
| 83 | |
| 84 | 84937 vg_libpthread.c:2505 (se_remap): Assertion `res == 0' |
| 85 | (fixed prior to 2.1.2) |
| 86 | |
| 87 | 86317 cannot load libSDL-1.2.so.0 using valgrind |
| 88 | |
| 89 | 86989 memcpy from mac_replace_strmem.c complains about |
| 90 | uninitialized pointers passed when length to copy is zero |
| 91 | |
| 92 | 85811 gnu pascal symbol causes segmentation fault; ok in 2.0.0 |
| 93 | |
| 94 | 79138 writing to sbrk()'d memory causes segfault |
| 95 | |
| 96 | 77369 sched deadlock while signal received during pthread_join |
| 97 | and the joined thread exited |
| 98 | |
| 99 | 88115 In signal handler for SIGFPE, siginfo->si_addr is wrong |
| 100 | under Valgrind |
| 101 | |
| 102 | 78765 Massif crashes on app exit if FP exceptions are enabled |
| 103 | |
| 104 | Additionally there are the following changes, which are not |
| 105 | connected to any bug report numbers, AFAICS: |
| 106 | |
| 107 | * Fix scary bug causing mis-identification of SSE stores vs |
| 108 | loads and so causing memcheck to sometimes give nonsense results |
| 109 | on SSE code. |
| 110 | |
| 111 | * Add support for the POSIX message queue system calls. |
| 112 | |
| 113 | * Fix to allow 32-bit Valgrind to run on AMD64 boxes. Note: this does |
| 114 | NOT allow Valgrind to work with 64-bit executables - only with 32-bit |
| 115 | executables on an AMD64 box. |
| 116 | |
| 117 | * At configure time, only check whether linux/mii.h can be processed |
| 118 | so that we don't generate ugly warnings by trying to compile it. |
| 119 | |
| 120 | |
| 121 | |
jseward | 9cd978c | 2004-07-17 13:38:12 +0000 | [diff] [blame] | 122 | Developer (cvs head) release 2.1.2 (18 July 2004) |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 123 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 124 | 2.1.2 contains four months worth of bug fixes and refinements. |
jseward | 9cd978c | 2004-07-17 13:38:12 +0000 | [diff] [blame] | 125 | Although officially a developer release, we believe it to be stable |
| 126 | enough for widespread day-to-day use. 2.1.2 is pretty good, so try it |
| 127 | first, although there is a chance it won't work. If so then try 2.0.0 |
| 128 | and tell us what went wrong." 2.1.2 fixes a lot of problems present |
| 129 | in 2.0.0 and is generally a much better product. |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 130 | |
jseward | 9cd978c | 2004-07-17 13:38:12 +0000 | [diff] [blame] | 131 | Relative to 2.1.1, a large number of minor problems with 2.1.1 have |
| 132 | been fixed, and so if you use 2.1.1 you should try 2.1.2. Users of |
| 133 | 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] | 134 | |
| 135 | The following bugs, and probably many more, have been fixed. These |
| 136 | are listed at http://bugs.kde.org. Reporting a bug for valgrind in |
| 137 | the http://bugs.kde.org is much more likely to get you a fix than |
| 138 | mailing developers directly, so please continue to keep sending bugs |
| 139 | there. |
| 140 | |
| 141 | 76869 Crashes when running any tool under Fedora Core 2 test1 |
| 142 | This fixes the problem with returning from a signal handler |
jseward | 0400d4d | 2004-07-17 23:26:46 +0000 | [diff] [blame] | 143 | when VDSOs are turned off in FC2. |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 144 | |
jseward | 0400d4d | 2004-07-17 23:26:46 +0000 | [diff] [blame] | 145 | 69508 java 1.4.2 client fails with erroneous "stack size too small". |
| 146 | This fix makes more of the pthread stack attribute related |
| 147 | functions work properly. Java still doesn't work though. |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 148 | |
| 149 | 71906 malloc alignment should be 8, not 4 |
| 150 | All memory returned by malloc/new etc is now at least |
| 151 | 8-byte aligned. |
| 152 | |
jseward | a6017a4 | 2004-07-17 23:44:35 +0000 | [diff] [blame] | 153 | 81970 vg_alloc_ThreadState: no free slots available |
| 154 | (closed because the workaround is simple: increase |
| 155 | VG_N_THREADS, rebuild and try again.) |
| 156 | |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 157 | 78514 Conditional jump or move depends on uninitialized value(s) |
| 158 | (a slight mishanding of FP code in memcheck) |
| 159 | |
jseward | a6017a4 | 2004-07-17 23:44:35 +0000 | [diff] [blame] | 160 | 77952 pThread Support (crash) (due to initialisation-ordering probs) |
| 161 | (also 85118) |
| 162 | |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 163 | 80942 Addrcheck wasn't doing overlap checking as it should. |
| 164 | 78048 return NULL on malloc/new etc failure, instead of asserting |
| 165 | 73655 operator new() override in user .so files often doesn't get picked up |
| 166 | 83060 Valgrind does not handle native kernel AIO |
| 167 | 69872 Create proper coredumps after fatal signals |
| 168 | 82026 failure with new glibc versions: __libc_* functions are not exported |
| 169 | 70344 UNIMPLEMENTED FUNCTION: tcdrain |
| 170 | 81297 Cancellation of pthread_cond_wait does not require mutex |
| 171 | 82872 Using debug info from additional packages (wishlist) |
| 172 | 83025 Support for ioctls FIGETBSZ and FIBMAP |
| 173 | 83340 Support for ioctl HDIO_GET_IDENTITY |
| 174 | 79714 Support for the semtimedop system call. |
| 175 | 77022 Support for ioctls FBIOGET_VSCREENINFO and FBIOGET_FSCREENINFO |
| 176 | 82098 hp2ps ansification (wishlist) |
| 177 | 83573 Valgrind SIGSEGV on execve |
| 178 | 82999 show which cmdline option was erroneous (wishlist) |
| 179 | 83040 make valgrind VPATH and distcheck-clean (wishlist) |
jseward | a6017a4 | 2004-07-17 23:44:35 +0000 | [diff] [blame] | 180 | 83998 Assertion `newfd > vgPlain_max_fd' failed (see below) |
| 181 | 82722 Unchecked mmap in as_pad leads to mysterious failures later |
| 182 | 78958 memcheck seg faults while running Mozilla |
jseward | a2c0800 | 2004-07-18 09:24:57 +0000 | [diff] [blame] | 183 | 85416 Arguments with colon (e.g. --logsocket) ignored |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 184 | |
| 185 | |
| 186 | Additionally there are the following changes, which are not |
| 187 | connected to any bug report numbers, AFAICS: |
| 188 | |
| 189 | * Rearranged address space layout relative to 2.1.1, so that |
| 190 | Valgrind/tools will run out of memory later than currently in many |
| 191 | circumstances. This is good news esp. for Calltree. It should |
| 192 | be possible for client programs to allocate over 800MB of |
| 193 | memory when using memcheck now. |
| 194 | |
jseward | 9cd978c | 2004-07-17 13:38:12 +0000 | [diff] [blame] | 195 | * Improved checking when laying out memory. Should hopefully avoid |
| 196 | the random segmentation faults that 2.1.1 sometimes caused. |
| 197 | |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 198 | * Support for Fedora Core 2 and SuSE 9.1. Improvements to NPTL |
| 199 | support to the extent that V now works properly on NPTL-only setups. |
| 200 | |
| 201 | * Renamed the following options: |
| 202 | --logfile-fd --> --log-fd |
| 203 | --logfile --> --log-file |
| 204 | --logsocket --> --log-socket |
| 205 | to be consistent with each other and other options (esp. --input-fd). |
| 206 | |
| 207 | * Add support for SIOCGMIIPHY, SIOCGMIIREG and SIOCSMIIREG ioctls and |
| 208 | improve the checking of other interface related ioctls. |
| 209 | |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 210 | * Fix building with gcc-3.4.1. |
| 211 | |
| 212 | * Remove limit on number of semaphores supported. |
| 213 | |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 214 | * Add support for syscalls: set_tid_address (258), acct (51). |
| 215 | |
| 216 | * Support instruction "repne movs" -- not official but seems to occur. |
| 217 | |
| 218 | * Implement an emulated soft limit for file descriptors in addition to |
| 219 | the current reserved area, which effectively acts as a hard limit. The |
| 220 | setrlimit system call now simply updates the emulated limits as best |
| 221 | as possible - the hard limit is not allowed to move at all and just |
| 222 | returns EPERM if you try and change it. This should stop reductions |
| 223 | in the soft limit causing assertions when valgrind tries to allocate |
| 224 | descriptors from the reserved area. |
jseward | a6017a4 | 2004-07-17 23:44:35 +0000 | [diff] [blame] | 225 | (This actually came from bug #83998). |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 226 | |
nethercote | 76e8fd9 | 2004-07-21 10:37:31 +0000 | [diff] [blame] | 227 | * Major overhaul of Cachegrind implementation. First user-visible change |
nethercote | 5e63bfb | 2004-07-17 12:49:53 +0000 | [diff] [blame] | 228 | is that cachegrind.out files are now typically 90% smaller than they |
nethercote | 76e8fd9 | 2004-07-21 10:37:31 +0000 | [diff] [blame] | 229 | used to be; code annotation times are correspondingly much smaller. |
| 230 | Second user-visible change is that hit/miss counts for code that is |
| 231 | unloaded at run-time is no longer dumped into a single "discard" pile, |
| 232 | but accurately preserved. |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 233 | |
| 234 | * Client requests for telling valgrind about memory pools. |
| 235 | |
| 236 | |
| 237 | |
jseward | 9cd978c | 2004-07-17 13:38:12 +0000 | [diff] [blame] | 238 | Developer (cvs head) release 2.1.1 (12 March 2004) |
| 239 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
jseward | bd0a6e4 | 2004-07-17 12:19:43 +0000 | [diff] [blame] | 240 | 2.1.1 contains some internal structural changes needed for V's |
jseward | e3e7704 | 2004-03-12 21:07:05 +0000 | [diff] [blame] | 241 | long-term future. These don't affect end-users. Most notable |
| 242 | user-visible changes are: |
| 243 | |
| 244 | * Greater isolation between Valgrind and the program being run, so |
| 245 | the program is less likely to inadvertently kill Valgrind by |
| 246 | doing wild writes. |
| 247 | |
| 248 | * Massif: a new space profiling tool. Try it! It's cool, and it'll |
| 249 | tell you in detail where and when your C/C++ code is allocating heap. |
| 250 | Draws pretty .ps pictures of memory use against time. A potentially |
| 251 | powerful tool for making sense of your program's space use. |
| 252 | |
| 253 | * Fixes for many bugs, including support for more SSE2/SSE3 instructions, |
| 254 | various signal/syscall things, and various problems with debug |
| 255 | info readers. |
| 256 | |
| 257 | * Support for glibc-2.3.3 based systems. |
| 258 | |
| 259 | We are now doing automatic overnight build-and-test runs on a variety |
| 260 | of distros. As a result, we believe 2.1.1 builds and runs on: |
| 261 | Red Hat 7.2, 7.3, 8.0, 9, Fedora Core 1, SuSE 8.2, SuSE 9. |
| 262 | |
| 263 | |
| 264 | The following bugs, and probably many more, have been fixed. These |
| 265 | are listed at http://bugs.kde.org. Reporting a bug for valgrind in |
| 266 | the http://bugs.kde.org is much more likely to get you a fix than |
| 267 | mailing developers directly, so please continue to keep sending bugs |
| 268 | there. |
| 269 | |
| 270 | 69616 glibc 2.3.2 w/NPTL is massively different than what valgrind expects |
| 271 | 69856 I don't know how to instrument MMXish stuff (Helgrind) |
| 272 | 73892 valgrind segfaults starting with Objective-C debug info |
| 273 | (fix for S-type stabs) |
| 274 | 73145 Valgrind complains too much about close(<reserved fd>) |
| 275 | 73902 Shadow memory allocation seems to fail on RedHat 8.0 |
| 276 | 68633 VG_N_SEMAPHORES too low (V itself was leaking semaphores) |
| 277 | 75099 impossible to trace multiprocess programs |
| 278 | 76839 the `impossible' happened: disInstr: INT but not 0x80 ! |
| 279 | 76762 vg_to_ucode.c:3748 (dis_push_segreg): Assertion `sz == 4' failed. |
| 280 | 76747 cannot include valgrind.h in c++ program |
| 281 | 76223 parsing B(3,10) gave NULL type => impossible happens |
| 282 | 75604 shmdt handling problem |
| 283 | 76416 Problems with gcc 3.4 snap 20040225 |
| 284 | 75614 using -gstabs when building your programs the `impossible' happened |
| 285 | 75787 Patch for some CDROM ioctls CDORM_GET_MCN, CDROM_SEND_PACKET, |
| 286 | 75294 gcc 3.4 snapshot's libstdc++ have unsupported instructions. |
| 287 | (REP RET) |
| 288 | 73326 vg_symtab2.c:272 (addScopeRange): Assertion `range->size > 0' failed. |
| 289 | 72596 not recognizing __libc_malloc |
| 290 | 69489 Would like to attach ddd to running program |
| 291 | 72781 Cachegrind crashes with kde programs |
| 292 | 73055 Illegal operand at DXTCV11CompressBlockSSE2 (more SSE opcodes) |
| 293 | 73026 Descriptor leak check reports port numbers wrongly |
| 294 | 71705 README_MISSING_SYSCALL_OR_IOCTL out of date |
| 295 | 72643 Improve support for SSE/SSE2 instructions |
| 296 | 72484 valgrind leaves it's own signal mask in place when execing |
| 297 | 72650 Signal Handling always seems to restart system calls |
| 298 | 72006 The mmap system call turns all errors in ENOMEM |
| 299 | 71781 gdb attach is pretty useless |
| 300 | 71180 unhandled instruction bytes: 0xF 0xAE 0x85 0xE8 |
| 301 | 69886 writes to zero page cause valgrind to assert on exit |
| 302 | 71791 crash when valgrinding gimp 1.3 (stabs reader problem) |
| 303 | 69783 unhandled syscall: 218 |
| 304 | 69782 unhandled instruction bytes: 0x66 0xF 0x2B 0x80 |
| 305 | 70385 valgrind fails if the soft file descriptor limit is less |
| 306 | than about 828 |
| 307 | 69529 "rep; nop" should do a yield |
| 308 | 70827 programs with lots of shared libraries report "mmap failed" |
| 309 | for some of them when reading symbols |
| 310 | 71028 glibc's strnlen is optimised enough to confuse valgrind |
| 311 | |
| 312 | |
| 313 | |
| 314 | |
jseward | 4323065 | 2003-12-14 00:14:54 +0000 | [diff] [blame] | 315 | Unstable (cvs head) release 2.1.0 (15 December 2003) |
| 316 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 317 | For whatever it's worth, 2.1.0 actually seems pretty darn stable to me |
| 318 | (Julian). It looks eminently usable, and given that it fixes some |
| 319 | significant bugs, may well be worth using on a day-to-day basis. |
| 320 | 2.1.0 is known to build and pass regression tests on: SuSE 9, SuSE |
| 321 | 8.2, RedHat 8. |
| 322 | |
| 323 | 2.1.0 most notably includes Jeremy Fitzhardinge's complete overhaul of |
| 324 | handling of system calls and signals, and their interaction with |
| 325 | threads. In general, the accuracy of the system call, thread and |
| 326 | signal simulations is much improved. Specifically: |
| 327 | |
| 328 | - Blocking system calls behave exactly as they do when running |
| 329 | natively (not on valgrind). That is, if a syscall blocks only the |
| 330 | calling thread when running natively, than it behaves the same on |
| 331 | valgrind. No more mysterious hangs because V doesn't know that some |
| 332 | syscall or other, should block only the calling thread. |
| 333 | |
| 334 | - Interrupted syscalls should now give more faithful results. |
| 335 | |
| 336 | - Finally, signal contexts in signal handlers are supported. As a |
| 337 | result, konqueror on SuSE 9 no longer segfaults when notified of |
| 338 | file changes in directories it is watching. |
| 339 | |
| 340 | Other changes: |
| 341 | |
| 342 | - Robert Walsh's file descriptor leakage checks. When enabled, |
| 343 | Valgrind will print out a list of open file descriptors on |
| 344 | exit. Along with each file descriptor, Valgrind prints out a stack |
| 345 | backtrace of where the file was opened and any details relating to the |
| 346 | file descriptor such as the file name or socket details. |
| 347 | To use, give: --track-fds=yes |
| 348 | |
| 349 | - Implemented a few more SSE/SSE2 instructions. |
| 350 | |
| 351 | - Less crud on the stack when you do 'where' inside a GDB attach. |
| 352 | |
| 353 | - Fixed the following bugs: |
| 354 | 68360: Valgrind does not compile against 2.6.0-testX kernels |
| 355 | 68525: CVS head doesn't compile on C90 compilers |
| 356 | 68566: pkgconfig support (wishlist) |
| 357 | 68588: Assertion `sz == 4' failed in vg_to_ucode.c (disInstr) |
| 358 | 69140: valgrind not able to explicitly specify a path to a binary. |
| 359 | 69432: helgrind asserts encountering a MutexErr when there are |
| 360 | EraserErr suppressions |
| 361 | |
jseward | 0b12daf | 2003-12-14 14:35:28 +0000 | [diff] [blame] | 362 | - Increase the max size of the translation cache from 200k average bbs |
| 363 | to 300k average bbs. Programs on the size of OOo (680m17) are |
| 364 | thrashing the cache at the smaller size, creating large numbers of |
| 365 | retranslations and wasting significant time as a result. |
| 366 | |
jseward | 4323065 | 2003-12-14 00:14:54 +0000 | [diff] [blame] | 367 | |
| 368 | |
| 369 | Stable release 2.0.0 (5 Nov 2003) |
| 370 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 371 | |
| 372 | 2.0.0 improves SSE/SSE2 support, fixes some minor bugs, and |
| 373 | improves support for SuSE 9 and the Red Hat "Severn" beta. |
| 374 | |
| 375 | - Further improvements to SSE/SSE2 support. The entire test suite of |
| 376 | the GNU Scientific Library (gsl-1.4) compiled with Intel Icc 7.1 |
| 377 | 20030307Z '-g -O -xW' now works. I think this gives pretty good |
| 378 | coverage of SSE/SSE2 floating point instructions, or at least the |
| 379 | subset emitted by Icc. |
| 380 | |
| 381 | - Also added support for the following instructions: |
| 382 | MOVNTDQ UCOMISD UNPCKLPS UNPCKHPS SQRTSS |
| 383 | PUSH/POP %{FS,GS}, and PUSH %CS (Nb: there is no POP %CS). |
| 384 | |
| 385 | - CFI support for GDB version 6. Needed to enable newer GDBs |
| 386 | to figure out where they are when using --gdb-attach=yes. |
| 387 | |
| 388 | - Fix this: |
| 389 | mc_translate.c:1091 (memcheck_instrument): Assertion |
| 390 | `u_in->size == 4 || u_in->size == 16' failed. |
| 391 | |
| 392 | - Return an error rather than panicing when given a bad socketcall. |
| 393 | |
| 394 | - Fix checking of syscall rt_sigtimedwait(). |
| 395 | |
| 396 | - Implement __NR_clock_gettime (syscall 265). Needed on Red Hat Severn. |
| 397 | |
| 398 | - Fixed bug in overlap check in strncpy() -- it was assuming the src was 'n' |
| 399 | bytes long, when it could be shorter, which could cause false |
| 400 | positives. |
| 401 | |
| 402 | - Support use of select() for very large numbers of file descriptors. |
| 403 | |
| 404 | - Don't fail silently if the executable is statically linked, or is |
| 405 | setuid/setgid. Print an error message instead. |
| 406 | |
| 407 | - Support for old DWARF-1 format line number info. |
| 408 | |
| 409 | |
| 410 | |
| 411 | Snapshot 20031012 (12 October 2003) |
| 412 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 413 | |
| 414 | Three months worth of bug fixes, roughly. Most significant single |
| 415 | change is improved SSE/SSE2 support, mostly thanks to Dirk Mueller. |
| 416 | |
| 417 | 20031012 builds on Red Hat Fedora ("Severn") but doesn't really work |
| 418 | (curiosly, mozilla runs OK, but a modest "ls -l" bombs). I hope to |
| 419 | get a working version out soon. It may or may not work ok on the |
| 420 | forthcoming SuSE 9; I hear positive noises about it but haven't been |
| 421 | able to verify this myself (not until I get hold of a copy of 9). |
| 422 | |
| 423 | A detailed list of changes, in no particular order: |
| 424 | |
| 425 | - Describe --gen-suppressions in the FAQ. |
| 426 | |
| 427 | - Syscall __NR_waitpid supported. |
| 428 | |
| 429 | - Minor MMX bug fix. |
| 430 | |
| 431 | - -v prints program's argv[] at startup. |
| 432 | |
| 433 | - More glibc-2.3 suppressions. |
| 434 | |
| 435 | - Suppressions for stack underrun bug(s) in the c++ support library |
| 436 | distributed with Intel Icc 7.0. |
| 437 | |
| 438 | - Fix problems reading /proc/self/maps. |
| 439 | |
| 440 | - Fix a couple of messages that should have been suppressed by -q, |
| 441 | but weren't. |
| 442 | |
| 443 | - Make Addrcheck understand "Overlap" suppressions. |
| 444 | |
| 445 | - At startup, check if program is statically linked and bail out if so. |
| 446 | |
| 447 | - Cachegrind: Auto-detect Intel Pentium-M, also VIA Nehemiah |
| 448 | |
| 449 | - Memcheck/addrcheck: minor speed optimisations |
| 450 | |
| 451 | - Handle syscall __NR_brk more correctly than before. |
| 452 | |
| 453 | - Fixed incorrect allocate/free mismatch errors when using |
| 454 | operator new(unsigned, std::nothrow_t const&) |
| 455 | operator new[](unsigned, std::nothrow_t const&) |
| 456 | |
| 457 | - Support POSIX pthread spinlocks. |
| 458 | |
| 459 | - Fixups for clean compilation with gcc-3.3.1. |
daywalker | 5d945de | 2003-09-26 00:32:53 +0000 | [diff] [blame] | 460 | |
| 461 | - Implemented more opcodes: |
| 462 | - push %es |
| 463 | - push %ds |
| 464 | - pop %es |
| 465 | - pop %ds |
| 466 | - movntq |
| 467 | - sfence |
| 468 | - pshufw |
| 469 | - pavgb |
| 470 | - ucomiss |
| 471 | - enter |
daywalker | b18d253 | 2003-09-27 20:15:01 +0000 | [diff] [blame] | 472 | - mov imm32, %esp |
| 473 | - all "in" and "out" opcodes |
daywalker | 79aad84 | 2003-09-30 22:58:12 +0000 | [diff] [blame] | 474 | - inc/dec %esp |
jseward | 4323065 | 2003-12-14 00:14:54 +0000 | [diff] [blame] | 475 | - A whole bunch of SSE/SSE2 instructions |
daywalker | 5d945de | 2003-09-26 00:32:53 +0000 | [diff] [blame] | 476 | |
jseward | 4323065 | 2003-12-14 00:14:54 +0000 | [diff] [blame] | 477 | - Memcheck: don't bomb on SSE/SSE2 code. |
daywalker | 5d945de | 2003-09-26 00:32:53 +0000 | [diff] [blame] | 478 | |
sewardj | 9023879 | 2003-05-05 00:23:42 +0000 | [diff] [blame] | 479 | |
sewardj | 945f39f | 2003-07-25 21:11:40 +0000 | [diff] [blame] | 480 | Snapshot 20030725 (25 July 2003) |
| 481 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 482 | |
| 483 | Fixes some minor problems in 20030716. |
| 484 | |
| 485 | - Fix bugs in overlap checking for strcpy/memcpy etc. |
| 486 | |
| 487 | - Do overlap checking with Addrcheck as well as Memcheck. |
| 488 | |
| 489 | - Fix this: |
| 490 | Memcheck: the `impossible' happened: |
| 491 | get_error_name: unexpected type |
| 492 | |
| 493 | - Install headers needed to compile new skins. |
| 494 | |
sewardj | e3dd2e0 | 2003-07-27 08:04:48 +0000 | [diff] [blame] | 495 | - Remove leading spaces and colon in the LD_LIBRARY_PATH / LD_PRELOAD |
sewardj | 945f39f | 2003-07-25 21:11:40 +0000 | [diff] [blame] | 496 | passed to non-traced children. |
| 497 | |
sewardj | b9eda95 | 2003-07-26 21:39:05 +0000 | [diff] [blame] | 498 | - Fix file descriptor leak in valgrind-listener. |
| 499 | |
sewardj | e3dd2e0 | 2003-07-27 08:04:48 +0000 | [diff] [blame] | 500 | - Fix longstanding bug in which the allocation point of a |
| 501 | block resized by realloc was not correctly set. This may |
| 502 | have caused confusing error messages. |
sewardj | 945f39f | 2003-07-25 21:11:40 +0000 | [diff] [blame] | 503 | |
| 504 | |
sewardj | 626fd89 | 2003-07-16 20:10:26 +0000 | [diff] [blame] | 505 | Snapshot 20030716 (16 July 2003) |
sewardj | 9d916ed | 2003-07-14 23:38:40 +0000 | [diff] [blame] | 506 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 507 | |
sewardj | 626fd89 | 2003-07-16 20:10:26 +0000 | [diff] [blame] | 508 | 20030716 is a snapshot of our current CVS head (development) branch. |
sewardj | 9d916ed | 2003-07-14 23:38:40 +0000 | [diff] [blame] | 509 | This is the branch which will become valgrind-2.0. It contains |
| 510 | significant enhancements over the 1.9.X branch. |
| 511 | |
| 512 | Despite this being a snapshot of the CVS head, it is believed to be |
| 513 | quite stable -- at least as stable as 1.9.6 or 1.0.4, if not more so |
| 514 | -- and therefore suitable for widespread use. Please let us know asap |
| 515 | if it causes problems for you. |
| 516 | |
| 517 | Two reasons for releasing a snapshot now are: |
| 518 | |
| 519 | - It's been a while since 1.9.6, and this snapshot fixes |
| 520 | various problems that 1.9.6 has with threaded programs |
| 521 | on glibc-2.3.X based systems. |
| 522 | |
| 523 | - So as to make available improvements in the 2.0 line. |
| 524 | |
sewardj | 626fd89 | 2003-07-16 20:10:26 +0000 | [diff] [blame] | 525 | Major changes in 20030716, as compared to 1.9.6: |
sewardj | 9d916ed | 2003-07-14 23:38:40 +0000 | [diff] [blame] | 526 | |
| 527 | - More fixes to threading support on glibc-2.3.1 and 2.3.2-based |
| 528 | systems (SuSE 8.2, Red Hat 9). If you have had problems |
| 529 | with inconsistent/illogical behaviour of errno, h_errno or the DNS |
sewardj | 626fd89 | 2003-07-16 20:10:26 +0000 | [diff] [blame] | 530 | resolver functions in threaded programs, 20030716 should improve |
sewardj | 9d916ed | 2003-07-14 23:38:40 +0000 | [diff] [blame] | 531 | matters. This snapshot seems stable enough to run OpenOffice.org |
| 532 | 1.1rc on Red Hat 7.3, SuSE 8.2 and Red Hat 9, and that's a big |
| 533 | threaded app if ever I saw one. |
| 534 | |
| 535 | - Automatic generation of suppression records; you no longer |
| 536 | need to write them by hand. Use --gen-suppressions=yes. |
| 537 | |
sewardj | 2151180 | 2003-07-22 17:47:42 +0000 | [diff] [blame] | 538 | - strcpy/memcpy/etc check their arguments for overlaps, when |
| 539 | running with the Memcheck or Addrcheck skins. |
| 540 | |
| 541 | - malloc_usable_size() is now supported. |
| 542 | |
| 543 | - new client requests: |
| 544 | - VALGRIND_COUNT_ERRORS, VALGRIND_COUNT_LEAKS: |
| 545 | useful with regression testing |
| 546 | - VALGRIND_NON_SIMD_CALL[0123]: for running arbitrary functions |
| 547 | on real CPU (use with caution!) |
| 548 | |
sewardj | 9d916ed | 2003-07-14 23:38:40 +0000 | [diff] [blame] | 549 | - The GDB attach mechanism is more flexible. Allow the GDB to |
| 550 | be run to be specified by --gdb-path=/path/to/gdb, and specify |
| 551 | which file descriptor V will read its input from with |
| 552 | --input-fd=<number>. |
| 553 | |
sewardj | 2151180 | 2003-07-22 17:47:42 +0000 | [diff] [blame] | 554 | - Cachegrind gives more accurate results (wasn't tracking instructions in |
| 555 | malloc() and friends previously, is now). |
| 556 | |
sewardj | 9d916ed | 2003-07-14 23:38:40 +0000 | [diff] [blame] | 557 | - Complete support for the MMX instruction set. |
| 558 | |
| 559 | - Partial support for the SSE and SSE2 instruction sets. Work for this |
| 560 | is ongoing. About half the SSE/SSE2 instructions are done, so |
| 561 | some SSE based programs may work. Currently you need to specify |
| 562 | --skin=addrcheck. Basically not suitable for real use yet. |
| 563 | |
| 564 | - Significant speedups (10%-20%) for standard memory checking. |
| 565 | |
| 566 | - Fix assertion failure in pthread_once(). |
| 567 | |
| 568 | - Fix this: |
| 569 | valgrind: vg_intercept.c:598 (vgAllRoadsLeadToRome_select): |
| 570 | Assertion `ms_end >= ms_now' failed. |
| 571 | |
| 572 | - Implement pthread_mutexattr_setpshared. |
| 573 | |
| 574 | - Understand Pentium 4 branch hints. Also implemented a couple more |
| 575 | obscure x86 instructions. |
| 576 | |
| 577 | - Lots of other minor bug fixes. |
| 578 | |
sewardj | 626fd89 | 2003-07-16 20:10:26 +0000 | [diff] [blame] | 579 | - We have a decent regression test system, for the first time. |
| 580 | This doesn't help you directly, but it does make it a lot easier |
| 581 | for us to track the quality of the system, especially across |
| 582 | multiple linux distributions. |
| 583 | |
| 584 | You can run the regression tests with 'make regtest' after 'make |
| 585 | install' completes. On SuSE 8.2 and Red Hat 9 I get this: |
| 586 | |
| 587 | == 84 tests, 0 stderr failures, 0 stdout failures == |
| 588 | |
| 589 | On Red Hat 8, I get this: |
| 590 | |
| 591 | == 84 tests, 2 stderr failures, 1 stdout failure == |
| 592 | corecheck/tests/res_search (stdout) |
| 593 | memcheck/tests/sigaltstack (stderr) |
| 594 | |
| 595 | sigaltstack is probably harmless. res_search doesn't work |
| 596 | on R H 8 even running natively, so I'm not too worried. |
| 597 | |
| 598 | On Red Hat 7.3, a glibc-2.2.5 system, I get these harmless failures: |
| 599 | |
| 600 | == 84 tests, 2 stderr failures, 1 stdout failure == |
| 601 | corecheck/tests/pth_atfork1 (stdout) |
| 602 | corecheck/tests/pth_atfork1 (stderr) |
| 603 | memcheck/tests/sigaltstack (stderr) |
| 604 | |
| 605 | You need to run on a PII system, at least, since some tests |
| 606 | contain P6-specific instructions, and the test machine needs |
| 607 | access to the internet so that corecheck/tests/res_search |
| 608 | (a test that the DNS resolver works) can function. |
| 609 | |
sewardj | 9d916ed | 2003-07-14 23:38:40 +0000 | [diff] [blame] | 610 | As ever, thanks for the vast amount of feedback :) and bug reports :( |
| 611 | We may not answer all messages, but we do at least look at all of |
| 612 | them, and tend to fix the most frequently reported bugs. |
| 613 | |
| 614 | |
| 615 | |
sewardj | 3791882 | 2003-05-05 01:05:09 +0000 | [diff] [blame] | 616 | Version 1.9.6 (7 May 2003 or thereabouts) |
| 617 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 618 | |
| 619 | Major changes in 1.9.6: |
| 620 | |
| 621 | - Improved threading support for glibc >= 2.3.2 (SuSE 8.2, |
| 622 | RedHat 9, to name but two ...) It turned out that 1.9.5 |
| 623 | had problems with threading support on glibc >= 2.3.2, |
| 624 | usually manifested by threaded programs deadlocking in system calls, |
| 625 | or running unbelievably slowly. Hopefully these are fixed now. 1.9.6 |
| 626 | is the first valgrind which gives reasonable support for |
| 627 | glibc-2.3.2. Also fixed a 2.3.2 problem with pthread_atfork(). |
| 628 | |
| 629 | - Majorly expanded FAQ.txt. We've added workarounds for all |
| 630 | common problems for which a workaround is known. |
| 631 | |
| 632 | Minor changes in 1.9.6: |
| 633 | |
| 634 | - Fix identification of the main thread's stack. Incorrect |
| 635 | identification of it was causing some on-stack addresses to not get |
| 636 | identified as such. This only affected the usefulness of some error |
| 637 | messages; the correctness of the checks made is unchanged. |
| 638 | |
| 639 | - Support for kernels >= 2.5.68. |
| 640 | |
| 641 | - Dummy implementations of __libc_current_sigrtmin, |
| 642 | __libc_current_sigrtmax and __libc_allocate_rtsig, hopefully |
| 643 | good enough to keep alive programs which previously died for lack of |
| 644 | them. |
| 645 | |
| 646 | - Fix bug in the VALGRIND_DISCARD_TRANSLATIONS client request. |
| 647 | |
sewardj | 3d47b79 | 2003-05-05 22:15:35 +0000 | [diff] [blame] | 648 | - Fix bug in the DWARF2 debug line info loader, when instructions |
| 649 | following each other have source lines far from each other |
| 650 | (e.g. with inlined functions). |
| 651 | |
sewardj | 3791882 | 2003-05-05 01:05:09 +0000 | [diff] [blame] | 652 | - Debug info reading: read symbols from both "symtab" and "dynsym" |
| 653 | sections, rather than merely from the one that comes last in the |
| 654 | file. |
| 655 | |
| 656 | - New syscall support: prctl(), creat(), lookup_dcookie(). |
| 657 | |
| 658 | - When checking calls to accept(), recvfrom(), getsocketopt(), |
| 659 | don't complain if buffer values are NULL. |
| 660 | |
| 661 | - Try and avoid assertion failures in |
| 662 | mash_LD_PRELOAD_and_LD_LIBRARY_PATH. |
| 663 | |
| 664 | - Minor bug fixes in cg_annotate. |
| 665 | |
| 666 | |
| 667 | |
sewardj | 9023879 | 2003-05-05 00:23:42 +0000 | [diff] [blame] | 668 | Version 1.9.5 (7 April 2003) |
| 669 | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~ |
| 670 | |
| 671 | It occurs to me that it would be helpful for valgrind users to record |
| 672 | in the source distribution the changes in each release. So I now |
| 673 | attempt to mend my errant ways :-) Changes in this and future releases |
| 674 | will be documented in the NEWS file in the source distribution. |
| 675 | |
| 676 | Major changes in 1.9.5: |
| 677 | |
| 678 | - (Critical bug fix): Fix a bug in the FPU simulation. This was |
| 679 | causing some floating point conditional tests not to work right. |
| 680 | Several people reported this. If you had floating point code which |
| 681 | didn't work right on 1.9.1 to 1.9.4, it's worth trying 1.9.5. |
| 682 | |
| 683 | - Partial support for Red Hat 9. RH9 uses the new Native Posix |
| 684 | Threads Library (NPTL), instead of the older LinuxThreads. |
| 685 | This potentially causes problems with V which will take some |
| 686 | time to correct. In the meantime we have partially worked around |
| 687 | this, and so 1.9.5 works on RH9. Threaded programs still work, |
| 688 | but they may deadlock, because some system calls (accept, read, |
| 689 | write, etc) which should be nonblocking, in fact do block. This |
| 690 | is a known bug which we are looking into. |
| 691 | |
| 692 | If you can, your best bet (unfortunately) is to avoid using |
| 693 | 1.9.5 on a Red Hat 9 system, or on any NPTL-based distribution. |
| 694 | If your glibc is 2.3.1 or earlier, you're almost certainly OK. |
| 695 | |
| 696 | Minor changes in 1.9.5: |
| 697 | |
| 698 | - Added some #errors to valgrind.h to ensure people don't include |
| 699 | it accidentally in their sources. This is a change from 1.0.X |
| 700 | which was never properly documented. The right thing to include |
| 701 | is now memcheck.h. Some people reported problems and strange |
| 702 | behaviour when (incorrectly) including valgrind.h in code with |
| 703 | 1.9.1 -- 1.9.4. This is no longer possible. |
| 704 | |
| 705 | - Add some __extension__ bits and pieces so that gcc configured |
| 706 | for valgrind-checking compiles even with -Werror. If you |
| 707 | don't understand this, ignore it. Of interest to gcc developers |
| 708 | only. |
| 709 | |
| 710 | - Removed a pointless check which caused problems interworking |
| 711 | with Clearcase. V would complain about shared objects whose |
| 712 | names did not end ".so", and refuse to run. This is now fixed. |
| 713 | In fact it was fixed in 1.9.4 but not documented. |
| 714 | |
| 715 | - Fixed a bug causing an assertion failure of "waiters == 1" |
| 716 | somewhere in vg_scheduler.c, when running large threaded apps, |
| 717 | notably MySQL. |
| 718 | |
| 719 | - Add support for the munlock system call (124). |
| 720 | |
| 721 | Some comments about future releases: |
| 722 | |
| 723 | 1.9.5 is, we hope, the most stable Valgrind so far. It pretty much |
| 724 | supersedes the 1.0.X branch. If you are a valgrind packager, please |
| 725 | consider making 1.9.5 available to your users. You can regard the |
| 726 | 1.0.X branch as obsolete: 1.9.5 is stable and vastly superior. There |
| 727 | are no plans at all for further releases of the 1.0.X branch. |
| 728 | |
| 729 | If you want a leading-edge valgrind, consider building the cvs head |
| 730 | (from SourceForge), or getting a snapshot of it. Current cool stuff |
| 731 | going in includes MMX support (done); SSE/SSE2 support (in progress), |
| 732 | a significant (10-20%) performance improvement (done), and the usual |
| 733 | large collection of minor changes. Hopefully we will be able to |
| 734 | improve our NPTL support, but no promises. |
| 735 | |