blob: 5cbe373860e80c9632450322871237f994f5dd02 [file] [log] [blame]
njn718d3b12006-12-16 00:54:12 +00001
njn254dafe2009-02-09 03:25:04 +00002Release 3.5.0 (???)
3~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
njn8df80b22009-03-02 05:11:06 +00004* A new Memcheck client request VALGRIND_COUNT_LEAK_BLOCKS has been added.
5 It is similar to VALGRIND_COUNT_LEAKS but counts blocks instead of bytes.
6 [XXX: consider adding VALGRIND_COUNT_LEAK_BYTES as a synonym and
7 deprecating VALGRIND_COUNT_LEAKS, which wasn't a good name to begin with]
8
njn8225cc02009-03-09 22:52:24 +00009* Memcheck's leak checker has been improved.
10 - The results for --leak-check=summary now match the summary results for
11 --leak-check=full. Previously they could differ because
12 --leak-check=summary counted "indirectly lost" blocks and "suppressed"
13 blocks as "definitely lost".
14 - Blocks that are only reachable via at least one interior-pointer, but
15 are directly pointed to by a start-pointer, were previously marked as
16 "still reachable". They are now correctly marked as "possibly lost".
17
njn5e40aba2009-03-16 22:11:31 +000018* The format of some (non-XML) stack trace entries has changed a little.
19 Previously there were six possible forms:
20
21 0x80483BF: really (a.c:20)
22 0x80483BF: really (in /foo/a.out)
23 0x80483BF: really
24 0x80483BF: (within /foo/a.out)
25 0x80483BF: ??? (a.c:20)
26 0x80483BF: ???
27
28 The third and fourth of these forms have been made more consistent with
29 the others. The six possible forms are now:
30
31 0x80483BF: really (a.c:20)
32 0x80483BF: really (in /foo/a.out)
33 0x80483BF: really (in ???)
34 0x80483BF: ??? (in /foo/a.out)
35 0x80483BF: ??? (a.c:20)
36 0x80483BF: ???
37
38 Stack traces produced when --xml=yes is specified are different and
39 unchanged.
40
njnace90d82009-04-27 01:27:58 +000041* The graphs drawn by Massif's ms_print program have changed slightly:
42 - The half-height chars '.' and ',' are no longer drawn, because they are
43 confusing. The --y option can be used if the default y-resolution is
44 not high enough.
45 - Horizontal lines are now drawn after the top of a snapshot if there is a
46 gap until the next snapshot. This makes it clear that the memory
47 usage has not dropped to zero between snapshots.
48
njn6bf365c2009-02-11 00:35:45 +000049* The location of some install files has changed. This should not affect
50 most users. Those who might be affected:
njn6bf365c2009-02-11 00:35:45 +000051 - For people who use Valgrind with MPI programs, the installed
52 libmpiwrap.so library has moved from $(INSTALL)/<platform>/libmpiwrap.so
53 to $(INSTALL)/libmpiwrap-<platform>.so.
njn6bf365c2009-02-11 00:35:45 +000054 - For people who distribute standalone Valgrind tools, the installed
55 libraries such as $(INSTALL)/<platform>/libcoregrind.a have moved to
56 $(INSTALL)/libcoregrind-<platform>.a.
57
58 These changes were made to simplify the build system.
59
njn254dafe2009-02-09 03:25:04 +000060* Previously, all the distributed suppression (*.supp) files were installed.
61 Now, only default.supp is installed. This should not affect users as the
62 other installed suppression files were not read.
63
bart076d2b52009-03-22 10:14:39 +000064n-i-bz DRD - race conditions between pthread_barrier_wait() and
65 pthread_barrier_destroy() calls are now reported.
66n-i-bz DRD - added "first observed at" information in error
67 messages related to synchronization objects.
68n-i-bz DRD - added new client request, VG_USERREQ__DRD_CLEAN_MEMORY.
69187048 DRD - the mutex attribute PTHREAD_PROCESS_SHARED is now
70 interpreted correctly.
bart56174ad2009-04-05 11:01:19 +000071188046 Removed a bashism from the configure script.
bartcd99d942009-04-25 16:25:45 +000072189763 DRD - pthread_create@GLIBC_2.0() no longer triggers an assertion
73 failure.
bart076d2b52009-03-22 10:14:39 +000074
njn254dafe2009-02-09 03:25:04 +000075
76
njnad91a752009-03-13 19:55:36 +000077Release 3.4.1 (28 February 2009)
78~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
793.4.1 is a bug-fix release that fixes some regressions and assertion
80failures in debug info reading in 3.4.0, most notably incorrect stack
81traces on amd64-linux on older (glibc-2.3 based) systems. Various
82other debug info problems are also fixed. A number of bugs in the
83exp-ptrcheck tool introduced in 3.4.0 have been fixed.
84
85In view of the fact that 3.4.0 contains user-visible regressions
86relative to 3.3.x, upgrading to 3.4.1 is recommended. Packagers are
87encouraged to ship 3.4.1 in preference to 3.4.0.
88
89The fixed bugs are as follows. Note that "n-i-bz" stands for "not in
90bugzilla" -- that is, a bug that was reported to us but never got a
91bugzilla entry. We encourage you to file bugs in bugzilla
92(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the
93developers (or mailing lists) directly -- bugs that are not entered
94into bugzilla tend to get forgotten about or ignored.
95
96n-i-bz Fix various bugs reading icc-11 generated debug info
97n-i-bz Fix various bugs reading gcc-4.4 generated debug info
98n-i-bz Preliminary support for glibc-2.10 / Fedora 11
99n-i-bz Cachegrind and Callgrind: handle non-power-of-two cache sizes,
100 so as to support (eg) 24k Atom D1 and Core2 with 3/6/12MB L2.
101179618 exp-ptrcheck crashed / exit prematurely
102179624 helgrind: false positive races with pthread_create and
103 recv/open/close/read
104134207 pkg-config output contains @VG_PLATFORM@
105176926 floating point exception at valgrind startup with PPC 440EPX
106181594 Bogus warning for empty text segment
107173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant rex prefixes)
108181707 Dwarf3 doesn't require enumerations to have name
109185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64
110185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree):
111 Assertion '!already_present' failed.
112185359 exp-ptrcheck unhandled syscall getresuid()
113
114(3.4.1.RC1: 24 Feb 2008, vex r1884, valgrind r9253).
115(3.4.1: 28 Feb 2008, vex r1884, valgrind r9293).
116
117
118
sewardj565fcf12008-12-23 13:14:06 +0000119Release 3.4.0 (2 January 2009)
120~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1213.4.0 is a feature release with many significant improvements and the
122usual collection of bug fixes. This release supports X86/Linux,
123AMD64/Linux, PPC32/Linux and PPC64/Linux. Support for recent distros
124(using gcc 4.4, glibc 2.8 and 2.9) has been added.
sewardj8639f112008-12-17 12:26:24 +0000125
sewardj565fcf12008-12-23 13:14:06 +00001263.4.0 brings some significant tool improvements. Memcheck can now
127report the origin of uninitialised values, the thread checkers
128Helgrind and DRD are much improved, and we have a new experimental
129tool, exp-Ptrcheck, which is able to detect overruns of stack and
130global arrays. In detail:
sewardj8639f112008-12-17 12:26:24 +0000131
sewardj565fcf12008-12-23 13:14:06 +0000132* Memcheck is now able to track the origin of uninitialised values.
133 When it reports an uninitialised value error, it will try to show
134 the origin of the value, as either a heap or stack allocation.
135 Origin tracking is expensive and so is not enabled by default. To
136 use it, specify --track-origins=yes. Memcheck's speed will be
137 essentially halved, and memory usage will be significantly
138 increased. Nevertheless it can drastically reduce the effort
139 required to identify the root cause of uninitialised value errors,
140 and so is often a programmer productivity win, despite running more
141 slowly.
sewardj8639f112008-12-17 12:26:24 +0000142
sewardj565fcf12008-12-23 13:14:06 +0000143* A version (1.4.0) of the Valkyrie GUI, that works with Memcheck in
sewardjb8b16642009-01-02 23:21:54 +0000144 3.4.0, will be released shortly.
sewardj8639f112008-12-17 12:26:24 +0000145
sewardj565fcf12008-12-23 13:14:06 +0000146* Helgrind's race detection algorithm has been completely redesigned
147 and reimplemented, to address usability and scalability concerns:
sewardj8639f112008-12-17 12:26:24 +0000148
sewardj565fcf12008-12-23 13:14:06 +0000149 - The new algorithm has a lower false-error rate: it is much less
150 likely to report races that do not really exist.
sewardj8639f112008-12-17 12:26:24 +0000151
sewardj565fcf12008-12-23 13:14:06 +0000152 - Helgrind will display full call stacks for both accesses involved
sewardjf1fa5ed2008-12-23 17:04:52 +0000153 in a race. This makes it easier to identify the root causes of
154 races.
sewardj8639f112008-12-17 12:26:24 +0000155
sewardj565fcf12008-12-23 13:14:06 +0000156 - Limitations on the size of program that can run have been removed.
sewardj8639f112008-12-17 12:26:24 +0000157
sewardj565fcf12008-12-23 13:14:06 +0000158 - Performance has been modestly improved, although that is very
159 workload-dependent.
sewardj8639f112008-12-17 12:26:24 +0000160
sewardj565fcf12008-12-23 13:14:06 +0000161 - Direct support for Qt4 threading has been added.
sewardj8639f112008-12-17 12:26:24 +0000162
sewardj565fcf12008-12-23 13:14:06 +0000163 - pthread_barriers are now directly supported.
njn75010ca2008-02-09 05:10:45 +0000164
sewardj565fcf12008-12-23 13:14:06 +0000165 - Helgrind works well on all supported Linux targets.
sewardj8639f112008-12-17 12:26:24 +0000166
sewardj565fcf12008-12-23 13:14:06 +0000167* The DRD thread debugging tool has seen major improvements:
sewardj8639f112008-12-17 12:26:24 +0000168
sewardj565fcf12008-12-23 13:14:06 +0000169 - Greatly improved performance and significantly reduced memory
170 usage.
sewardj8639f112008-12-17 12:26:24 +0000171
sewardj565fcf12008-12-23 13:14:06 +0000172 - Support for several major threading libraries (Boost.Thread, Qt4,
173 glib, OpenMP) has been added.
sewardj8639f112008-12-17 12:26:24 +0000174
sewardj565fcf12008-12-23 13:14:06 +0000175 - Support for atomic instructions, POSIX semaphores, barriers and
176 reader-writer locks has been added.
177
178 - Works now on PowerPC CPUs too.
179
180 - Added support for printing thread stack usage at thread exit time.
181
182 - Added support for debugging lock contention.
183
184 - Added a manual for Drd.
185
186* A new experimental tool, exp-Ptrcheck, has been added. Ptrcheck
187 checks for misuses of pointers. In that sense it is a bit like
188 Memcheck. However, Ptrcheck can do things Memcheck can't: it can
189 detect overruns of stack and global arrays, it can detect
190 arbitrarily far out-of-bounds accesses to heap blocks, and it can
191 detect accesses heap blocks that have been freed a very long time
192 ago (millions of blocks in the past).
193
194 Ptrcheck currently works only on x86-linux and amd64-linux. To use
195 it, use --tool=exp-ptrcheck. A simple manual is provided, as part
196 of the main Valgrind documentation. As this is an experimental
197 tool, we would be particularly interested in hearing about your
198 experiences with it.
199
sewardj974aace2008-12-24 18:40:57 +0000200* exp-Omega, an experimental instantaneous leak-detecting tool, is no
201 longer built by default, although the code remains in the repository
202 and the tarball. This is due to three factors: a perceived lack of
203 users, a lack of maintenance, and concerns that it may not be
204 possible to achieve reliable operation using the existing design.
sewardj565fcf12008-12-23 13:14:06 +0000205
206* As usual, support for the latest Linux distros and toolchain
207 components has been added. It should work well on Fedora Core 10,
208 OpenSUSE 11.1 and Ubuntu 8.10. gcc-4.4 (in its current pre-release
209 state) is supported, as is glibc-2.9. The C++ demangler has been
210 updated so as to work well with C++ compiled by even the most recent
211 g++'s.
212
213* You can now use frame-level wildcards in suppressions. This was a
214 frequently-requested enhancement. A line "..." in a suppression now
215 matches zero or more frames. This makes it easier to write
216 suppressions which are precise yet insensitive to changes in
217 inlining behaviour.
218
219* 3.4.0 adds support on x86/amd64 for the SSSE3 instruction set.
220
221* Very basic support for IBM Power6 has been added (64-bit processes only).
222
223* Valgrind is now cross-compilable. For example, it is possible to
224 cross compile Valgrind on an x86/amd64-linux host, so that it runs
225 on a ppc32/64-linux target.
226
sewardjf1fa5ed2008-12-23 17:04:52 +0000227* You can set the main thread's stack size at startup using the
228 new --main-stacksize= flag (subject of course to ulimit settings).
229 This is useful for running apps that need a lot of stack space.
230
231* The limitation that you can't use --trace-children=yes together
232 with --db-attach=yes has been removed.
233
sewardj565fcf12008-12-23 13:14:06 +0000234* The following bugs have been fixed. Note that "n-i-bz" stands for
235 "not in bugzilla" -- that is, a bug that was reported to us but
236 never got a bugzilla entry. We encourage you to file bugs in
237 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
238 mailing the developers (or mailing lists) directly.
239
sewardjf1fa5ed2008-12-23 17:04:52 +0000240 n-i-bz Make return types for some client requests 64-bit clean
241 n-i-bz glibc 2.9 support
242 n-i-bz ignore unsafe .valgrindrc's (CVE-2008-4865)
243 n-i-bz MPI_Init(0,0) is valid but libmpiwrap.c segfaults
244 n-i-bz Building in an env without gdb gives bogus gdb attach
245 92456 Tracing the origin of uninitialised memory
246 106497 Valgrind does not demangle some C++ template symbols
247 162222 ==106497
sewardjdf1fc9a2008-12-23 23:09:57 +0000248 151612 Suppression with "..." (frame-level wildcards in .supp files)
sewardjf1fa5ed2008-12-23 17:04:52 +0000249 156404 Unable to start oocalc under memcheck on openSUSE 10.3 (64-bit)
250 159285 unhandled syscall:25 (stime, on x86-linux)
251 159452 unhandled ioctl 0x8B01 on "valgrind iwconfig"
252 160954 ppc build of valgrind crashes with illegal instruction (isel)
253 160956 mallinfo implementation, w/ patch
254 162092 Valgrind fails to start gnome-system-monitor
255 162819 malloc_free_fill test doesn't pass on glibc2.8 x86
256 163794 assertion failure with "--track-origins=yes"
257 163933 sigcontext.err and .trapno must be set together
258 163955 remove constraint !(--db-attach=yes && --trace-children=yes)
259 164476 Missing kernel module loading system calls
260 164669 SVN regression: mmap() drops posix file locks
261 166581 Callgrind output corruption when program forks
262 167288 Patch file for missing system calls on Cell BE
263 168943 unsupported scas instruction pentium
264 171645 Unrecognised instruction (MOVSD, non-binutils encoding)
265 172417 x86->IR: 0x82 ...
266 172563 amd64->IR: 0xD9 0xF5 - fprem1
267 173099 .lds linker script generation error
268 173177 [x86_64] syscalls: 125/126/179 (capget/capset/quotactl)
269 173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant prefixes)
270 174532 == 173751
271 174908 --log-file value not expanded correctly for core file
272 175044 Add lookup_dcookie for amd64
273 175150 x86->IR: 0xF2 0xF 0x11 0xC1 (movss non-binutils encoding)
sewardj565fcf12008-12-23 13:14:06 +0000274
275Developer-visible changes:
276
277* Valgrind's debug-info reading machinery has been majorly overhauled.
278 It can now correctly establish the addresses for ELF data symbols,
279 which is something that has never worked properly before now.
280
281 Also, Valgrind can now read DWARF3 type and location information for
282 stack and global variables. This makes it possible to use the
283 framework to build tools that rely on knowing the type and locations
284 of stack and global variables, for example exp-Ptrcheck.
285
286 Reading of such information is disabled by default, because most
287 tools don't need it, and because it is expensive in space and time.
288 However, you can force Valgrind to read it, using the
289 --read-var-info=yes flag. Memcheck, Helgrind and DRD are able to
290 make use of such information, if present, to provide source-level
291 descriptions of data addresses in the error messages they create.
292
sewardj974aace2008-12-24 18:40:57 +0000293(3.4.0.RC1: 24 Dec 2008, vex r1878, valgrind r8882).
sewardjb8b16642009-01-02 23:21:54 +0000294(3.4.0: 3 Jan 2009, vex r1878, valgrind r8899).
sewardj565fcf12008-12-23 13:14:06 +0000295
sewardj8639f112008-12-17 12:26:24 +0000296
sewardj9f42ba32008-06-04 11:47:51 +0000297
298Release 3.3.1 (4 June 2008)
299~~~~~~~~~~~~~~~~~~~~~~~~~~~
3003.3.1 fixes a bunch of bugs in 3.3.0, adds support for glibc-2.8 based
301systems (openSUSE 11, Fedora Core 9), improves the existing glibc-2.7
302support, and adds support for the SSSE3 (Core 2) instruction set.
303
3043.3.1 will likely be the last release that supports some very old
305systems. In particular, the next major release, 3.4.0, will drop
306support for the old LinuxThreads threading library, and for gcc
307versions prior to 3.0.
308
309The fixed bugs are as follows. Note that "n-i-bz" stands for "not in
310bugzilla" -- that is, a bug that was reported to us but never got a
311bugzilla entry. We encourage you to file bugs in bugzilla
312(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the
313developers (or mailing lists) directly -- bugs that are not entered
314into bugzilla tend to get forgotten about or ignored.
315
316n-i-bz Massif segfaults at exit
317n-i-bz Memcheck asserts on Altivec code
318n-i-bz fix sizeof bug in Helgrind
319n-i-bz check fd on sys_llseek
320n-i-bz update syscall lists to kernel 2.6.23.1
321n-i-bz support sys_sync_file_range
322n-i-bz handle sys_sysinfo, sys_getresuid, sys_getresgid on ppc64-linux
323n-i-bz intercept memcpy in 64-bit ld.so's
324n-i-bz Fix wrappers for sys_{futimesat,utimensat}
325n-i-bz Minor false-error avoidance fixes for Memcheck
326n-i-bz libmpiwrap.c: add a wrapper for MPI_Waitany
327n-i-bz helgrind support for glibc-2.8
328n-i-bz partial fix for mc_leakcheck.c:698 assert:
329 'lc_shadows[i]->data + lc_shadows[i] ...
330n-i-bz Massif/Cachegrind output corruption when programs fork
331n-i-bz register allocator fix: handle spill stores correctly
332n-i-bz add support for PA6T PowerPC CPUs
333126389 vex x86->IR: 0xF 0xAE (FXRSTOR)
334158525 ==126389
335152818 vex x86->IR: 0xF3 0xAC (repz lodsb)
336153196 vex x86->IR: 0xF2 0xA6 (repnz cmpsb)
337155011 vex x86->IR: 0xCF (iret)
338155091 Warning [...] unhandled DW_OP_ opcode 0x23
339156960 ==155901
340155528 support Core2/SSSE3 insns on x86/amd64
341155929 ms_print fails on massif outputs containing long lines
342157665 valgrind fails on shmdt(0) after shmat to 0
343157748 support x86 PUSHFW/POPFW
344158212 helgrind: handle pthread_rwlock_try{rd,wr}lock.
345158425 sys_poll incorrectly emulated when RES==0
346158744 vex amd64->IR: 0xF0 0x41 0xF 0xC0 (xaddb)
347160907 Support for a couple of recent Linux syscalls
348161285 Patch -- support for eventfd() syscall
349161378 illegal opcode in debug libm (FUCOMPP)
350160136 ==161378
351161487 number of suppressions files is limited to 10
352162386 ms_print typo in milliseconds time unit for massif
353161036 exp-drd: client allocated memory was never freed
354162663 signalfd_wrapper fails on 64bit linux
355
356(3.3.1.RC1: 2 June 2008, vex r1854, valgrind r8169).
357(3.3.1: 4 June 2008, vex r1854, valgrind r8180).
358
359
360
sewardjcb2288b2007-12-02 02:08:17 +0000361Release 3.3.0 (7 December 2007)
362~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardjcb2288b2007-12-02 02:08:17 +00003633.3.0 is a feature release with many significant improvements and the
364usual collection of bug fixes. This release supports X86/Linux,
365AMD64/Linux, PPC32/Linux and PPC64/Linux. Support for recent distros
366(using gcc 4.3, glibc 2.6 and 2.7) has been added.
njnf4f9cd42007-11-23 01:42:10 +0000367
sewardjcb2288b2007-12-02 02:08:17 +0000368The main excitement in 3.3.0 is new and improved tools. Helgrind
369works again, Massif has been completely overhauled and much improved,
370Cachegrind now does branch-misprediction profiling, and a new category
371of experimental tools has been created, containing two new tools:
372Omega and DRD. There are many other smaller improvements. In detail:
njn374a36d2007-11-23 01:41:32 +0000373
sewardjcb2288b2007-12-02 02:08:17 +0000374- Helgrind has been completely overhauled and works for the first time
375 since Valgrind 2.2.0. Supported functionality is: detection of
376 misuses of the POSIX PThreads API, detection of potential deadlocks
377 resulting from cyclic lock dependencies, and detection of data
378 races. Compared to the 2.2.0 Helgrind, the race detection algorithm
379 has some significant improvements aimed at reducing the false error
380 rate. Handling of various kinds of corner cases has been improved.
381 Efforts have been made to make the error messages easier to
382 understand. Extensive documentation is provided.
njn374a36d2007-11-23 01:41:32 +0000383
sewardjcb2288b2007-12-02 02:08:17 +0000384- Massif has been completely overhauled. Instead of measuring
385 space-time usage -- which wasn't always useful and many people found
386 confusing -- it now measures space usage at various points in the
387 execution, including the point of peak memory allocation. Its
388 output format has also changed: instead of producing PostScript
389 graphs and HTML text, it produces a single text output (via the new
390 'ms_print' script) that contains both a graph and the old textual
391 information, but in a more compact and readable form. Finally, the
392 new version should be more reliable than the old one, as it has been
393 tested more thoroughly.
njn718d3b12006-12-16 00:54:12 +0000394
sewardj8badbaa2007-05-08 09:20:25 +0000395- Cachegrind has been extended to do branch-misprediction profiling.
396 Both conditional and indirect branches are profiled. The default
397 behaviour of Cachegrind is unchanged. To use the new functionality,
398 give the option --branch-sim=yes.
399
sewardjcb2288b2007-12-02 02:08:17 +0000400- A new category of "experimental tools" has been created. Such tools
401 may not work as well as the standard tools, but are included because
402 some people will find them useful, and because exposure to a wider
403 user group provides tool authors with more end-user feedback. These
404 tools have a "exp-" prefix attached to their names to indicate their
405 experimental nature. Currently there are two experimental tools:
njn374a36d2007-11-23 01:41:32 +0000406
sewardjcb2288b2007-12-02 02:08:17 +0000407 * exp-Omega: an instantaneous leak detector. See
408 exp-omega/docs/omega_introduction.txt.
njn374a36d2007-11-23 01:41:32 +0000409
sewardjcb2288b2007-12-02 02:08:17 +0000410 * exp-DRD: a data race detector based on the happens-before
sewardj8b09d4f2007-12-04 21:27:18 +0000411 relation. See exp-drd/docs/README.txt.
njn374a36d2007-11-23 01:41:32 +0000412
sewardjcb2288b2007-12-02 02:08:17 +0000413- Scalability improvements for very large programs, particularly those
414 which have a million or more malloc'd blocks in use at once. These
415 improvements mostly affect Memcheck. Memcheck is also up to 10%
416 faster for all programs, with x86-linux seeing the largest
417 improvement.
njn374a36d2007-11-23 01:41:32 +0000418
sewardjcb2288b2007-12-02 02:08:17 +0000419- Works well on the latest Linux distros. Has been tested on Fedora
420 Core 8 (x86, amd64, ppc32, ppc64) and openSUSE 10.3. glibc 2.6 and
421 2.7 are supported. gcc-4.3 (in its current pre-release state) is
422 supported. At the same time, 3.3.0 retains support for older
423 distros.
njn374a36d2007-11-23 01:41:32 +0000424
sewardjcb2288b2007-12-02 02:08:17 +0000425- The documentation has been modestly reorganised with the aim of
426 making it easier to find information on common-usage scenarios.
427 Some advanced material has been moved into a new chapter in the main
428 manual, so as to unclutter the main flow, and other tidying up has
429 been done.
njn374a36d2007-11-23 01:41:32 +0000430
sewardjcb2288b2007-12-02 02:08:17 +0000431- There is experimental support for AIX 5.3, both 32-bit and 64-bit
432 processes. You need to be running a 64-bit kernel to use Valgrind
433 on a 64-bit executable.
njn374a36d2007-11-23 01:41:32 +0000434
sewardjcb2288b2007-12-02 02:08:17 +0000435- There have been some changes to command line options, which may
436 affect you:
njn374a36d2007-11-23 01:41:32 +0000437
sewardjcb2288b2007-12-02 02:08:17 +0000438 * --log-file-exactly and
439 --log-file-qualifier options have been removed.
njn374a36d2007-11-23 01:41:32 +0000440
sewardjcb2288b2007-12-02 02:08:17 +0000441 To make up for this --log-file option has been made more powerful.
442 It now accepts a %p format specifier, which is replaced with the
443 process ID, and a %q{FOO} format specifier, which is replaced with
444 the contents of the environment variable FOO.
445
446 * --child-silent-after-fork=yes|no [no]
447
448 Causes Valgrind to not show any debugging or logging output for
449 the child process resulting from a fork() call. This can make the
450 output less confusing (although more misleading) when dealing with
451 processes that create children.
452
453 * --cachegrind-out-file, --callgrind-out-file and --massif-out-file
454
455 These control the names of the output files produced by
456 Cachegrind, Callgrind and Massif. They accept the same %p and %q
457 format specifiers that --log-file accepts. --callgrind-out-file
458 replaces Callgrind's old --base option.
459
460 * Cachegrind's 'cg_annotate' script no longer uses the --<pid>
461 option to specify the output file. Instead, the first non-option
462 argument is taken to be the name of the output file, and any
463 subsequent non-option arguments are taken to be the names of
464 source files to be annotated.
465
466 * Cachegrind and Callgrind now use directory names where possible in
467 their output files. This means that the -I option to
468 'cg_annotate' and 'callgrind_annotate' should not be needed in
469 most cases. It also means they can correctly handle the case
470 where two source files in different directories have the same
471 name.
472
473- Memcheck offers a new suppression kind: "Jump". This is for
474 suppressing jump-to-invalid-address errors. Previously you had to
475 use an "Addr1" suppression, which didn't make much sense.
476
477- Memcheck has new flags --malloc-fill=<hexnum> and
478 --free-fill=<hexnum> which free malloc'd / free'd areas with the
479 specified byte. This can help shake out obscure memory corruption
480 problems. The definedness and addressibility of these areas is
481 unchanged -- only the contents are affected.
njn718d3b12006-12-16 00:54:12 +0000482
483- The behaviour of Memcheck's client requests VALGRIND_GET_VBITS and
484 VALGRIND_SET_VBITS have changed slightly. They no longer issue
485 addressability errors -- if either array is partially unaddressable,
486 they just return 3 (as before). Also, SET_VBITS doesn't report
487 definedness errors if any of the V bits are undefined.
sewardj1ede80f2006-05-26 00:47:13 +0000488
njn92676d72007-12-04 21:18:06 +0000489- The following Memcheck client requests have been removed:
490 VALGRIND_MAKE_NOACCESS
491 VALGRIND_MAKE_WRITABLE
492 VALGRIND_MAKE_READABLE
493 VALGRIND_CHECK_WRITABLE
494 VALGRIND_CHECK_READABLE
495 VALGRIND_CHECK_DEFINED
496 They were deprecated in 3.2.0, when equivalent but better-named client
497 requests were added. See the 3.2.0 release notes for more details.
498
njn5a0a7732007-02-12 10:37:37 +0000499- The behaviour of the tool Lackey has changed slightly. First, the output
500 from --trace-mem has been made more compact, to reduce the size of the
501 traces. Second, a new option --trace-superblocks has been added, which
502 shows the addresses of superblocks (code blocks) as they are executed.
503
sewardjcb2288b2007-12-02 02:08:17 +0000504- The following bugs have been fixed. Note that "n-i-bz" stands for
505 "not in bugzilla" -- that is, a bug that was reported to us but
506 never got a bugzilla entry. We encourage you to file bugs in
507 bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than
508 mailing the developers (or mailing lists) directly.
509
510 n-i-bz x86_linux_REDIR_FOR_index() broken
511 n-i-bz guest-amd64/toIR.c:2512 (dis_op2_E_G): Assertion `0' failed.
512 n-i-bz Support x86 INT insn (INT (0xCD) 0x40 - 0x43)
513 n-i-bz Add sys_utimensat system call for Linux x86 platform
514 79844 Helgrind complains about race condition which does not exist
515 82871 Massif output function names too short
516 89061 Massif: ms_main.c:485 (get_XCon): Assertion `xpt->max_chi...'
517 92615 Write output from Massif at crash
518 95483 massif feature request: include peak allocation in report
519 112163 MASSIF crashed with signal 7 (SIGBUS) after running 2 days
520 119404 problems running setuid executables (partial fix)
521 121629 add instruction-counting mode for timing
522 127371 java vm giving unhandled instruction bytes: 0x26 0x2E 0x64 0x65
523 129937 ==150380
524 129576 Massif loses track of memory, incorrect graphs
525 132132 massif --format=html output does not do html entity escaping
526 132950 Heap alloc/usage summary
527 133962 unhandled instruction bytes: 0xF2 0x4C 0xF 0x10
528 134990 use -fno-stack-protector if possible
529 136382 ==134990
530 137396 I would really like helgrind to work again...
531 137714 x86/amd64->IR: 0x66 0xF 0xF7 0xC6 (maskmovq, maskmovdq)
532 141631 Massif: percentages don't add up correctly
533 142706 massif numbers don't seem to add up
534 143062 massif crashes on app exit with signal 8 SIGFPE
535 144453 (get_XCon): Assertion 'xpt->max_children != 0' failed.
536 145559 valgrind aborts when malloc_stats is called
537 145609 valgrind aborts all runs with 'repeated section!'
538 145622 --db-attach broken again on x86-64
539 145837 ==149519
540 145887 PPC32: getitimer() system call is not supported
541 146252 ==150678
542 146456 (update_XCon): Assertion 'xpt->curr_space >= -space_delta'...
543 146701 ==134990
544 146781 Adding support for private futexes
545 147325 valgrind internal error on syscall (SYS_io_destroy, 0)
546 147498 amd64->IR: 0xF0 0xF 0xB0 0xF (lock cmpxchg %cl,(%rdi))
sewardjbc39ad22007-12-09 02:24:02 +0000547 147545 Memcheck: mc_main.c:817 (get_sec_vbits8): Assertion 'n' failed.
sewardjcb2288b2007-12-02 02:08:17 +0000548 147628 SALC opcode 0xd6 unimplemented
549 147825 crash on amd64-linux with gcc 4.2 and glibc 2.6 (CFI)
550 148174 Incorrect type of freed_list_volume causes assertion [...]
551 148447 x86_64 : new NOP codes: 66 66 66 66 2e 0f 1f
552 149182 PPC Trap instructions not implemented in valgrind
553 149504 Assertion hit on alloc_xpt->curr_space >= -space_delta
554 149519 ppc32: V aborts with SIGSEGV on execution of a signal handler
555 149892 ==137714
556 150044 SEGV during stack deregister
557 150380 dwarf/gcc interoperation (dwarf3 read problems)
558 150408 ==148447
559 150678 guest-amd64/toIR.c:3741 (dis_Grp5): Assertion `sz == 4' failed
560 151209 V unable to execute programs for users with UID > 2^16
561 151938 help on --db-command= misleading
562 152022 subw $0x28, %%sp causes assertion failure in memcheck
563 152357 inb and outb not recognized in 64-bit mode
564 152501 vex x86->IR: 0x27 0x66 0x89 0x45 (daa)
565 152818 vex x86->IR: 0xF3 0xAC 0xFC 0x9C (rep lodsb)
566
njnd5f4bef2006-12-24 07:54:19 +0000567Developer-visible changes:
568
sewardjcb2288b2007-12-02 02:08:17 +0000569- The names of some functions and types within the Vex IR have
570 changed. Run 'svn log -r1689 VEX/pub/libvex_ir.h' for full details.
571 Any existing standalone tools will have to be updated to reflect
572 these changes. The new names should be clearer. The file
573 VEX/pub/libvex_ir.h is also much better commented.
njnd8366ee2007-02-12 22:38:50 +0000574
sewardjcb2288b2007-12-02 02:08:17 +0000575- A number of new debugging command line options have been added.
576 These are mostly of use for debugging the symbol table and line
577 number readers:
578
579 --trace-symtab-patt=<patt> limit debuginfo tracing to obj name <patt>
580 --trace-cfi=no|yes show call-frame-info details? [no]
581 --debug-dump=syms mimic /usr/bin/readelf --syms
582 --debug-dump=line mimic /usr/bin/readelf --debug-dump=line
583 --debug-dump=frames mimic /usr/bin/readelf --debug-dump=frames
584 --sym-offsets=yes|no show syms in form 'name+offset' ? [no]
585
586- Internally, the code base has been further factorised and
587 abstractified, particularly with respect to support for non-Linux
588 OSs.
589
sewardj135a7ff2007-12-10 22:17:42 +0000590(3.3.0.RC1: 2 Dec 2007, vex r1803, valgrind r7268).
591(3.3.0.RC2: 5 Dec 2007, vex r1804, valgrind r7282).
592(3.3.0.RC3: 9 Dec 2007, vex r1804, valgrind r7288).
sewardj686b5be2007-12-10 22:21:14 +0000593(3.3.0: 10 Dec 2007, vex r1804, valgrind r7290).
sewardjcb2288b2007-12-02 02:08:17 +0000594
njnd5f4bef2006-12-24 07:54:19 +0000595
596
njn99237752007-01-30 05:18:38 +0000597Release 3.2.3 (29 Jan 2007)
598~~~~~~~~~~~~~~~~~~~~~~~~~~~
599Unfortunately 3.2.2 introduced a regression which can cause an
600assertion failure ("vex: the `impossible' happened: eqIRConst") when
601running obscure pieces of SSE code. 3.2.3 fixes this and adds one
602more glibc-2.5 intercept. In all other respects it is identical to
6033.2.2. Please do not use (or package) 3.2.2; instead use 3.2.3.
604
605n-i-bz vex: the `impossible' happened: eqIRConst
606n-i-bz Add an intercept for glibc-2.5 __stpcpy_chk
607
608(3.2.3: 29 Jan 2007, vex r1732, valgrind r6560).
609
610
njna7e48802007-01-23 05:58:52 +0000611Release 3.2.2 (22 Jan 2007)
612~~~~~~~~~~~~~~~~~~~~~~~~~~~
6133.2.2 fixes a bunch of bugs in 3.2.1, adds support for glibc-2.5 based
614systems (openSUSE 10.2, Fedora Core 6), improves support for icc-9.X
615compiled code, and brings modest performance improvements in some
616areas, including amd64 floating point, powerpc support, and startup
617responsiveness on all targets.
618
619The fixed bugs are as follows. Note that "n-i-bz" stands for "not in
620bugzilla" -- that is, a bug that was reported to us but never got a
621bugzilla entry. We encourage you to file bugs in bugzilla
622(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the
623developers (or mailing lists) directly.
624
625129390 ppc?->IR: some kind of VMX prefetch (dstt)
626129968 amd64->IR: 0xF 0xAE 0x0 (fxsave)
627134319 ==129968
628133054 'make install' fails with syntax errors
629118903 ==133054
630132998 startup fails in when running on UML
631134207 pkg-config output contains @VG_PLATFORM@
632134727 valgrind exits with "Value too large for defined data type"
633n-i-bz ppc32/64: support mcrfs
634n-i-bz Cachegrind/Callgrind: Update cache parameter detection
635135012 x86->IR: 0xD7 0x8A 0xE0 0xD0 (xlat)
636125959 ==135012
637126147 x86->IR: 0xF2 0xA5 0xF 0x77 (repne movsw)
638136650 amd64->IR: 0xC2 0x8 0x0
639135421 x86->IR: unhandled Grp5(R) case 6
640n-i-bz Improved documentation of the IR intermediate representation
641n-i-bz jcxz (x86) (users list, 8 Nov)
642n-i-bz ExeContext hashing fix
643n-i-bz fix CFI reading failures ("Dwarf CFI 0:24 0:32 0:48 0:7")
644n-i-bz fix Cachegrind/Callgrind simulation bug
645n-i-bz libmpiwrap.c: fix handling of MPI_LONG_DOUBLE
646n-i-bz make User errors suppressible
647136844 corrupted malloc line when using --gen-suppressions=yes
648138507 ==136844
649n-i-bz Speed up the JIT's register allocator
650n-i-bz Fix confusing leak-checker flag hints
651n-i-bz Support recent autoswamp versions
652n-i-bz ppc32/64 dispatcher speedups
653n-i-bz ppc64 front end rld/rlw improvements
654n-i-bz ppc64 back end imm64 improvements
655136300 support 64K pages on ppc64-linux
656139124 == 136300
657n-i-bz fix ppc insn set tests for gcc >= 4.1
658137493 x86->IR: recent binutils no-ops
659137714 x86->IR: 0x66 0xF 0xF7 0xC6 (maskmovdqu)
660138424 "failed in UME with error 22" (produce a better error msg)
661138856 ==138424
662138627 Enhancement support for prctl ioctls
663138896 Add support for usb ioctls
664136059 ==138896
665139050 ppc32->IR: mfspr 268/269 instructions not handled
666n-i-bz ppc32->IR: lvxl/stvxl
667n-i-bz glibc-2.5 support
668n-i-bz memcheck: provide replacement for mempcpy
669n-i-bz memcheck: replace bcmp in ld.so
670n-i-bz Use 'ifndef' in VEX's Makefile correctly
671n-i-bz Suppressions for MVL 4.0.1 on ppc32-linux
672n-i-bz libmpiwrap.c: Fixes for MPICH
673n-i-bz More robust handling of hinted client mmaps
674139776 Invalid read in unaligned memcpy with Intel compiler v9
675n-i-bz Generate valid XML even for very long fn names
676n-i-bz Don't prompt about suppressions for unshown reachable leaks
677139910 amd64 rcl is not supported
678n-i-bz DWARF CFI reader: handle DW_CFA_undefined
679n-i-bz DWARF CFI reader: handle icc9 generated CFI info better
680n-i-bz fix false uninit-value errs in icc9 generated FP code
681n-i-bz reduce extraneous frames in libmpiwrap.c
682n-i-bz support pselect6 on amd64-linux
683
684(3.2.2: 22 Jan 2007, vex r1729, valgrind r6545).
685
686
sewardj172bce82006-09-16 01:00:35 +0000687Release 3.2.1 (16 Sept 2006)
688~~~~~~~~~~~~~~~~~~~~~~~~~~~~
6893.2.1 adds x86/amd64 support for all SSE3 instructions except monitor
690and mwait, further reduces memcheck's false error rate on all
691platforms, adds support for recent binutils (in OpenSUSE 10.2 and
692Fedora Rawhide) and fixes a bunch of bugs in 3.2.0. Some of the fixed
693bugs were causing large programs to segfault with --tool=callgrind and
694--tool=cachegrind, so an upgrade is recommended.
695
696In view of the fact that any 3.3.0 release is unlikely to happen until
697well into 1Q07, we intend to keep the 3.2.X line alive for a while
698yet, and so we tentatively plan a 3.2.2 release sometime in December
69906.
700
701The fixed bugs are as follows. Note that "n-i-bz" stands for "not in
702bugzilla" -- that is, a bug that was reported to us but never got a
703bugzilla entry.
704
705n-i-bz Expanding brk() into last available page asserts
706n-i-bz ppc64-linux stack RZ fast-case snafu
707n-i-bz 'c' in --gen-supps=yes doesn't work
708n-i-bz VG_N_SEGMENTS too low (users, 28 June)
709n-i-bz VG_N_SEGNAMES too low (Stu Robinson)
710106852 x86->IR: fisttp (SSE3)
711117172 FUTEX_WAKE does not use uaddr2
712124039 Lacks support for VKI_[GP]IO_UNIMAP*
713127521 amd64->IR: 0xF0 0x48 0xF 0xC7 (cmpxchg8b)
714128917 amd64->IR: 0x66 0xF 0xF6 0xC4 (psadbw,SSE2)
715129246 JJ: ppc32/ppc64 syscalls, w/ patch
716129358 x86->IR: fisttpl (SSE3)
717129866 cachegrind/callgrind causes executable to die
718130020 Can't stat .so/.exe error while reading symbols
719130388 Valgrind aborts when process calls malloc_trim()
720130638 PATCH: ppc32 missing system calls
721130785 amd64->IR: unhandled instruction "pushfq"
722131481: (HINT_NOP) vex x86->IR: 0xF 0x1F 0x0 0xF
723131298 ==131481
724132146 Programs with long sequences of bswap[l,q]s
725132918 vex amd64->IR: 0xD9 0xF8 (fprem)
726132813 Assertion at priv/guest-x86/toIR.c:652 fails
727133051 'cfsi->len > 0 && cfsi->len < 2000000' failed
728132722 valgrind header files are not standard C
729n-i-bz Livelocks entire machine (users list, Timothy Terriberry)
730n-i-bz Alex Bennee mmap problem (9 Aug)
731n-i-bz BartV: Don't print more lines of a stack-trace than were obtained.
732n-i-bz ppc32 SuSE 10.1 redir
733n-i-bz amd64 padding suppressions
734n-i-bz amd64 insn printing fix.
735n-i-bz ppc cmp reg,reg fix
736n-i-bz x86/amd64 iropt e/rflag reduction rules
737n-i-bz SuSE 10.1 (ppc32) minor fixes
738133678 amd64->IR: 0x48 0xF 0xC5 0xC0 (pextrw?)
739133694 aspacem assertion: aspacem_minAddr <= holeStart
740n-i-bz callgrind: fix warning about malformed creator line
741n-i-bz callgrind: fix annotate script for data produced with
742 --dump-instr=yes
743n-i-bz callgrind: fix failed assertion when toggling
744 instrumentation mode
745n-i-bz callgrind: fix annotate script fix warnings with
746 --collect-jumps=yes
747n-i-bz docs path hardwired (Dennis Lubert)
748
749The following bugs were not fixed, due primarily to lack of developer
750time, and also because bug reporters did not answer requests for
751feedback in time for the release:
752
753129390 ppc?->IR: some kind of VMX prefetch (dstt)
754129968 amd64->IR: 0xF 0xAE 0x0 (fxsave)
755133054 'make install' fails with syntax errors
756n-i-bz Signal race condition (users list, 13 June, Johannes Berg)
757n-i-bz Unrecognised instruction at address 0x70198EC2 (users list,
758 19 July, Bennee)
759132998 startup fails in when running on UML
760
761The following bug was tentatively fixed on the mainline but the fix
762was considered too risky to push into 3.2.X:
763
764133154 crash when using client requests to register/deregister stack
765
766(3.2.1: 16 Sept 2006, vex r1658, valgrind r6070).
767
768
sewardj3ecf8f22006-06-06 21:41:22 +0000769Release 3.2.0 (7 June 2006)
sewardj535a2352006-04-18 02:04:52 +0000770~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardj1ede80f2006-05-26 00:47:13 +00007713.2.0 is a feature release with many significant improvements and the
772usual collection of bug fixes. This release supports X86/Linux,
773AMD64/Linux, PPC32/Linux and PPC64/Linux.
njna3e96762006-03-27 12:23:44 +0000774
sewardjaf924992006-05-26 13:51:37 +0000775Performance, especially of Memcheck, is improved, Addrcheck has been
776removed, Callgrind has been added, PPC64/Linux support has been added,
777Lackey has been improved, and MPI support has been added. In detail:
njna3e96762006-03-27 12:23:44 +0000778
sewardjf41db3c2006-05-26 16:06:10 +0000779- Memcheck has improved speed and reduced memory use. Run times are
780 typically reduced by 15-30%, averaging about 24% for SPEC CPU2000.
sewardjaf924992006-05-26 13:51:37 +0000781 The other tools have smaller but noticeable speed improvments. We
782 are interested to hear what improvements users get.
njna3e96762006-03-27 12:23:44 +0000783
sewardj3ecf8f22006-06-06 21:41:22 +0000784 Memcheck uses less memory due to the introduction of a compressed
sewardjaf924992006-05-26 13:51:37 +0000785 representation for shadow memory. The space overhead has been
786 reduced by a factor of up to four, depending on program behaviour.
787 This means you should be able to run programs that use more memory
788 than before without hitting problems.
njna3e96762006-03-27 12:23:44 +0000789
sewardj1ede80f2006-05-26 00:47:13 +0000790- Addrcheck has been removed. It has not worked since version 2.4.0,
791 and the speed and memory improvements to Memcheck make it redundant.
792 If you liked using Addrcheck because it didn't give undefined value
793 errors, you can use the new Memcheck option --undef-value-errors=no
794 to get the same behaviour.
795
sewardjaf924992006-05-26 13:51:37 +0000796- The number of undefined-value errors incorrectly reported by
797 Memcheck has been reduced (such false reports were already very
798 rare). In particular, efforts have been made to ensure Memcheck
799 works really well with gcc 4.0/4.1-generated code on X86/Linux and
800 AMD64/Linux.
sewardj1ede80f2006-05-26 00:47:13 +0000801
802- Josef Weidendorfer's popular Callgrind tool has been added. Folding
sewardjaf924992006-05-26 13:51:37 +0000803 it in was a logical step given its popularity and usefulness, and
sewardj1ede80f2006-05-26 00:47:13 +0000804 makes it easier for us to ensure it works "out of the box" on all
805 supported targets. The associated KDE KCachegrind GUI remains a
806 separate project.
njneaf0ca92006-04-09 01:23:29 +0000807
sewardj3ecf8f22006-06-06 21:41:22 +0000808- A new release of the Valkyrie GUI for Memcheck, version 1.2.0,
809 accompanies this release. Improvements over previous releases
810 include improved robustness, many refinements to the user interface,
811 and use of a standard autoconf/automake build system. You can get
812 it from http://www.valgrind.org/downloads/guis.html.
813
sewardj535a2352006-04-18 02:04:52 +0000814- Valgrind now works on PPC64/Linux. As with the AMD64/Linux port,
sewardj1ede80f2006-05-26 00:47:13 +0000815 this supports programs using to 32G of address space. On 64-bit
816 capable PPC64/Linux setups, you get a dual architecture build so
817 that both 32-bit and 64-bit executables can be run. Linux on POWER5
818 is supported, and POWER4 is also believed to work. Both 32-bit and
819 64-bit DWARF2 is supported. This port is known to work well with
820 both gcc-compiled and xlc/xlf-compiled code.
njna3e96762006-03-27 12:23:44 +0000821
sewardj1ede80f2006-05-26 00:47:13 +0000822- Floating point accuracy has been improved for PPC32/Linux.
823 Specifically, the floating point rounding mode is observed on all FP
824 arithmetic operations, and multiply-accumulate instructions are
825 preserved by the compilation pipeline. This means you should get FP
826 results which are bit-for-bit identical to a native run. These
827 improvements are also present in the PPC64/Linux port.
njnb2939162006-04-18 00:52:37 +0000828
njneaf0ca92006-04-09 01:23:29 +0000829- Lackey, the example tool, has been improved:
njna3e96762006-03-27 12:23:44 +0000830
sewardj1ede80f2006-05-26 00:47:13 +0000831 * It has a new option --detailed-counts (off by default) which
832 causes it to print out a count of loads, stores and ALU operations
833 done, and their sizes.
njndbf7ca72006-03-31 11:57:59 +0000834
sewardj1ede80f2006-05-26 00:47:13 +0000835 * It has a new option --trace-mem (off by default) which causes it
836 to print out a trace of all memory accesses performed by a
837 program. It's a good starting point for building Valgrind tools
838 that need to track memory accesses. Read the comments at the top
839 of the file lackey/lk_main.c for details.
sewardj42fcf5d2006-04-08 13:16:36 +0000840
sewardj1ede80f2006-05-26 00:47:13 +0000841 * The original instrumentation (counting numbers of instructions,
842 jumps, etc) is now controlled by a new option --basic-counts. It
843 is on by default.
sewardj42fcf5d2006-04-08 13:16:36 +0000844
845- MPI support: partial support for debugging distributed applications
sewardj1ede80f2006-05-26 00:47:13 +0000846 using the MPI library specification has been added. Valgrind is
sewardj42fcf5d2006-04-08 13:16:36 +0000847 aware of the memory state changes caused by a subset of the MPI
848 functions, and will carefully check data passed to the (P)MPI_
849 interface.
850
sewardj1ede80f2006-05-26 00:47:13 +0000851- A new flag, --error-exitcode=, has been added. This allows changing
852 the exit code in runs where Valgrind reported errors, which is
853 useful when using Valgrind as part of an automated test suite.
sewardjb9779082006-05-12 23:50:15 +0000854
sewardj1ede80f2006-05-26 00:47:13 +0000855- Various segfaults when reading old-style "stabs" debug information
856 have been fixed.
njneaf0ca92006-04-09 01:23:29 +0000857
sewardjc9d09cc2006-05-27 03:22:04 +0000858- A simple performance evaluation suite has been added. See
njn5359b6f2006-06-02 23:57:22 +0000859 perf/README and README_DEVELOPERS for details. There are
sewardjc9d09cc2006-05-27 03:22:04 +0000860 various bells and whistles.
861
sewardjb70a6132006-05-27 21:14:09 +0000862- New configuration flags:
863 --enable-only32bit
864 --enable-only64bit
865 By default, on 64 bit platforms (ppc64-linux, amd64-linux) the build
866 system will attempt to build a Valgrind which supports both 32-bit
867 and 64-bit executables. This may not be what you want, and you can
868 override the default behaviour using these flags.
869
sewardj1ede80f2006-05-26 00:47:13 +0000870Please note that Helgrind is still not working. We have made an
871important step towards making it work again, however, with the
872addition of function wrapping (see below).
njneaf0ca92006-04-09 01:23:29 +0000873
874Other user-visible changes:
875
sewardj1ede80f2006-05-26 00:47:13 +0000876- Valgrind now has the ability to intercept and wrap arbitrary
877 functions. This is a preliminary step towards making Helgrind work
878 again, and was required for MPI support.
njneaf0ca92006-04-09 01:23:29 +0000879
sewardj1ede80f2006-05-26 00:47:13 +0000880- There are some changes to Memcheck's client requests. Some of them
881 have changed names:
njndbf7ca72006-03-31 11:57:59 +0000882
883 MAKE_NOACCESS --> MAKE_MEM_NOACCESS
884 MAKE_WRITABLE --> MAKE_MEM_UNDEFINED
885 MAKE_READABLE --> MAKE_MEM_DEFINED
886
887 CHECK_WRITABLE --> CHECK_MEM_IS_ADDRESSABLE
888 CHECK_READABLE --> CHECK_MEM_IS_DEFINED
889 CHECK_DEFINED --> CHECK_VALUE_IS_DEFINED
890
sewardj1ede80f2006-05-26 00:47:13 +0000891 The reason for the change is that the old names are subtly
892 misleading. The old names will still work, but they are deprecated
893 and may be removed in a future release.
njndbf7ca72006-03-31 11:57:59 +0000894
895 We also added a new client request:
896
897 MAKE_MEM_DEFINED_IF_ADDRESSABLE(a, len)
898
899 which is like MAKE_MEM_DEFINED but only affects a byte if the byte is
900 already addressable.
901
sewardj3ecf8f22006-06-06 21:41:22 +0000902- The way client requests are encoded in the instruction stream has
903 changed. Unfortunately, this means 3.2.0 will not honour client
904 requests compiled into binaries using headers from earlier versions
905 of Valgrind. We will try to keep the client request encodings more
906 stable in future.
njneaf0ca92006-04-09 01:23:29 +0000907
njna3e96762006-03-27 12:23:44 +0000908BUGS FIXED:
909
sewardj1ede80f2006-05-26 00:47:13 +0000910108258 NPTL pthread cleanup handlers not called
911117290 valgrind is sigKILL'd on startup
912117295 == 117290
913118703 m_signals.c:1427 Assertion 'tst->status == VgTs_WaitSys'
914118466 add %reg, %reg generates incorrect validity for bit 0
915123210 New: strlen from ld-linux on amd64
916123244 DWARF2 CFI reader: unhandled CFI instruction 0:18
917123248 syscalls in glibc-2.4: openat, fstatat, symlinkat
918123258 socketcall.recvmsg(msg.msg_iov[i] points to uninit
919123535 mremap(new_addr) requires MREMAP_FIXED in 4th arg
920123836 small typo in the doc
921124029 ppc compile failed: `vor' gcc 3.3.5
922124222 Segfault: @@don't know what type ':' is
923124475 ppc32: crash (syscall?) timer_settime()
924124499 amd64->IR: 0xF 0xE 0x48 0x85 (femms)
925124528 FATAL: aspacem assertion failed: segment_is_sane
926124697 vex x86->IR: 0xF 0x70 0xC9 0x0 (pshufw)
927124892 vex x86->IR: 0xF3 0xAE (REPx SCASB)
928126216 == 124892
929124808 ppc32: sys_sched_getaffinity() not handled
930n-i-bz Very long stabs strings crash m_debuginfo
931n-i-bz amd64->IR: 0x66 0xF 0xF5 (pmaddwd)
932125492 ppc32: support a bunch more syscalls
933121617 ppc32/64: coredumping gives assertion failure
934121814 Coregrind return error as exitcode patch
935126517 == 121814
sewardj1ede80f2006-05-26 00:47:13 +0000936125607 amd64->IR: 0x66 0xF 0xA3 0x2 (btw etc)
937125651 amd64->IR: 0xF8 0x49 0xFF 0xE3 (clc?)
938126253 x86 movx is wrong
939126451 3.2 SVN doesn't work on ppc32 CPU's without FPU
940126217 increase # threads
941126243 vex x86->IR: popw mem
942126583 amd64->IR: 0x48 0xF 0xA4 0xC2 (shld $1,%rax,%rdx)
sewardjb70a6132006-05-27 21:14:09 +0000943126668 amd64->IR: 0x1C 0xFF (sbb $0xff,%al)
944126696 support for CDROMREADRAW ioctl and CDROMREADTOCENTRY fix
945126722 assertion: segment_is_sane at m_aspacemgr/aspacemgr.c:1624
946126938 bad checking for syscalls linkat, renameat, symlinkat
sewardj1ede80f2006-05-26 00:47:13 +0000947
sewardj3ecf8f22006-06-06 21:41:22 +0000948(3.2.0RC1: 27 May 2006, vex r1626, valgrind r5947).
949(3.2.0: 7 June 2006, vex r1628, valgrind r5957).
njna3e96762006-03-27 12:23:44 +0000950
951
sewardj50654672006-03-16 11:33:14 +0000952Release 3.1.1 (15 March 2006)
953~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
9543.1.1 fixes a bunch of bugs reported in 3.1.0. There is no new
955functionality. The fixed bugs are:
956
957(note: "n-i-bz" means "not in bugzilla" -- this bug does not have
958 a bugzilla entry).
959
960n-i-bz ppc32: fsub 3,3,3 in dispatcher doesn't clear NaNs
961n-i-bz ppc32: __NR_{set,get}priority
962117332 x86: missing line info with icc 8.1
963117366 amd64: 0xDD 0x7C fnstsw
964118274 == 117366
965117367 amd64: 0xD9 0xF4 fxtract
966117369 amd64: __NR_getpriority (140)
967117419 ppc32: lfsu f5, -4(r11)
968117419 ppc32: fsqrt
969117936 more stabs problems (segfaults while reading debug info)
970119914 == 117936
971120345 == 117936
972118239 amd64: 0xF 0xAE 0x3F (clflush)
973118939 vm86old system call
974n-i-bz memcheck/tests/mempool reads freed memory
975n-i-bz AshleyP's custom-allocator assertion
976n-i-bz Dirk strict-aliasing stuff
977n-i-bz More space for debugger cmd line (Dan Thaler)
978n-i-bz Clarified leak checker output message
979n-i-bz AshleyP's --gen-suppressions output fix
980n-i-bz cg_annotate's --sort option broken
981n-i-bz OSet 64-bit fastcmp bug
982n-i-bz VG_(getgroups) fix (Shinichi Noda)
983n-i-bz ppc32: allocate from callee-saved FP/VMX regs
984n-i-bz misaligned path word-size bug in mc_main.c
985119297 Incorrect error message for sse code
986120410 x86: prefetchw (0xF 0xD 0x48 0x4)
987120728 TIOCSERGETLSR, TIOCGICOUNT, HDIO_GET_DMA ioctls
988120658 Build fixes for gcc 2.96
989120734 x86: Support for changing EIP in signal handler
990n-i-bz memcheck/tests/zeropage de-looping fix
991n-i-bz x86: fxtract doesn't work reliably
992121662 x86: lock xadd (0xF0 0xF 0xC0 0x2)
993121893 calloc does not always return zeroed memory
994121901 no support for syscall tkill
995n-i-bz Suppression update for Debian unstable
996122067 amd64: fcmovnu (0xDB 0xD9)
997n-i-bz ppc32: broken signal handling in cpu feature detection
998n-i-bz ppc32: rounding mode problems (improved, partial fix only)
999119482 ppc32: mtfsb1
1000n-i-bz ppc32: mtocrf/mfocrf
1001
1002(3.1.1: 15 March 2006, vex r1597, valgrind r5771).
1003
jseward43230652003-12-14 00:14:54 +00001004
sewardj4bbe9d72005-11-16 18:39:25 +00001005Release 3.1.0 (25 November 2005)
njnf63170c2005-11-12 19:09:51 +00001006~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardj4bbe9d72005-11-16 18:39:25 +000010073.1.0 is a feature release with a number of significant improvements:
1008AMD64 support is much improved, PPC32 support is good enough to be
1009usable, and the handling of memory management and address space is
1010much more robust. In detail:
njnf63170c2005-11-12 19:09:51 +00001011
sewardj4bbe9d72005-11-16 18:39:25 +00001012- AMD64 support is much improved. The 64-bit vs. 32-bit issues in
1013 3.0.X have been resolved, and it should "just work" now in all
1014 cases. On AMD64 machines both 64-bit and 32-bit versions of
1015 Valgrind are built. The right version will be invoked
1016 automatically, even when using --trace-children and mixing execution
1017 between 64-bit and 32-bit executables. Also, many more instructions
1018 are supported.
njnf63170c2005-11-12 19:09:51 +00001019
sewardj4bbe9d72005-11-16 18:39:25 +00001020- PPC32 support is now good enough to be usable. It should work with
1021 all tools, but please let us know if you have problems. Three
1022 classes of CPUs are supported: integer only (no FP, no Altivec),
1023 which covers embedded PPC uses, integer and FP but no Altivec
1024 (G3-ish), and CPUs capable of Altivec too (G4, G5).
njnf63170c2005-11-12 19:09:51 +00001025
sewardj4bbe9d72005-11-16 18:39:25 +00001026- Valgrind's address space management has been overhauled. As a
1027 result, Valgrind should be much more robust with programs that use
1028 large amounts of memory. There should be many fewer "memory
1029 exhausted" messages, and debug symbols should be read correctly on
1030 large (eg. 300MB+) executables. On 32-bit machines the full address
1031 space available to user programs (usually 3GB or 4GB) can be fully
1032 utilised. On 64-bit machines up to 32GB of space is usable; when
1033 using Memcheck that means your program can use up to about 14GB.
njnf63170c2005-11-12 19:09:51 +00001034
1035 A side effect of this change is that Valgrind is no longer protected
1036 against wild writes by the client. This feature was nice but relied
1037 on the x86 segment registers and so wasn't portable.
1038
1039- Most users should not notice, but as part of the address space
sewardj4bbe9d72005-11-16 18:39:25 +00001040 manager change, the way Valgrind is built has been changed. Each
1041 tool is now built as a statically linked stand-alone executable,
1042 rather than as a shared object that is dynamically linked with the
1043 core. The "valgrind" program invokes the appropriate tool depending
1044 on the --tool option. This slightly increases the amount of disk
1045 space used by Valgrind, but it greatly simplified many things and
1046 removed Valgrind's dependence on glibc.
njnf63170c2005-11-12 19:09:51 +00001047
njn336c89d2005-11-20 18:14:04 +00001048Please note that Addrcheck and Helgrind are still not working. Work
1049is underway to reinstate them (or equivalents). We apologise for the
1050inconvenience.
1051
njnf63170c2005-11-12 19:09:51 +00001052Other user-visible changes:
1053
1054- The --weird-hacks option has been renamed --sim-hints.
1055
1056- The --time-stamp option no longer gives an absolute date and time.
1057 It now prints the time elapsed since the program began.
1058
sewardj4bbe9d72005-11-16 18:39:25 +00001059- It should build with gcc-2.96.
1060
njn7cce5b82005-11-16 20:12:22 +00001061- Valgrind can now run itself (see README_DEVELOPERS for how).
sewardj47058612005-11-16 19:13:13 +00001062 This is not much use to you, but it means the developers can now
1063 profile Valgrind using Cachegrind. As a result a couple of
1064 performance bad cases have been fixed.
1065
sewardj6a5a69c2005-11-17 00:51:36 +00001066- The XML output format has changed slightly. See
1067 docs/internals/xml-output.txt.
1068
njn14c9add2005-11-18 17:36:01 +00001069- Core dumping has been reinstated (it was disabled in 3.0.0 and 3.0.1).
1070 If your program crashes while running under Valgrind, a core file with
1071 the name "vgcore.<pid>" will be created (if your settings allow core
1072 file creation). Note that the floating point information is not all
1073 there. If Valgrind itself crashes, the OS will create a normal core
1074 file.
1075
njnf63170c2005-11-12 19:09:51 +00001076The following are some user-visible changes that occurred in earlier
1077versions that may not have been announced, or were announced but not
njn7cce5b82005-11-16 20:12:22 +00001078widely noticed. So we're mentioning them now.
njnf63170c2005-11-12 19:09:51 +00001079
1080- The --tool flag is optional once again; if you omit it, Memcheck
1081 is run by default.
1082
1083- The --num-callers flag now has a default value of 12. It was
1084 previously 4.
1085
njn7cce5b82005-11-16 20:12:22 +00001086- The --xml=yes flag causes Valgrind's output to be produced in XML
1087 format. This is designed to make it easy for other programs to
1088 consume Valgrind's output. The format is described in the file
njnf63170c2005-11-12 19:09:51 +00001089 docs/internals/xml-format.txt.
1090
njn1f503992005-11-13 16:41:15 +00001091- The --gen-suppressions flag supports an "all" value that causes every
1092 suppression to be printed without asking.
1093
njnee88d282005-11-20 17:19:28 +00001094- The --log-file option no longer puts "pid" in the filename, eg. the
1095 old name "foo.pid12345" is now "foo.12345".
1096
njnc66077e2005-11-20 18:02:15 +00001097- There are several graphical front-ends for Valgrind, such as Valkyrie,
1098 Alleyoop and Valgui. See http://www.valgrind.org/downloads/guis.html
1099 for a list.
1100
njnf63170c2005-11-12 19:09:51 +00001101BUGS FIXED:
1102
sewardj4bbe9d72005-11-16 18:39:25 +00001103109861 amd64 hangs at startup
1104110301 ditto
1105111554 valgrind crashes with Cannot allocate memory
1106111809 Memcheck tool doesn't start java
1107111901 cross-platform run of cachegrind fails on opteron
1108113468 (vgPlain_mprotect_range): Assertion 'r != -1' failed.
1109 92071 Reading debugging info uses too much memory
1110109744 memcheck loses track of mmap from direct ld-linux.so.2
1111110183 tail of page with _end
1112 82301 FV memory layout too rigid
1113 98278 Infinite recursion possible when allocating memory
1114108994 Valgrind runs out of memory due to 133x overhead
1115115643 valgrind cannot allocate memory
1116105974 vg_hashtable.c static hash table
1117109323 ppc32: dispatch.S uses Altivec insn, which doesn't work on POWER.
1118109345 ptrace_setregs not yet implemented for ppc
1119110831 Would like to be able to run against both 32 and 64 bit
1120 binaries on AMD64
1121110829 == 110831
1122111781 compile of valgrind-3.0.0 fails on my linux (gcc 2.X prob)
1123112670 Cachegrind: cg_main.c:486 (handleOneStatement ...
1124112941 vex x86: 0xD9 0xF4 (fxtract)
1125110201 == 112941
1126113015 vex amd64->IR: 0xE3 0x14 0x48 0x83 (jrcxz)
1127113126 Crash with binaries built with -gstabs+/-ggdb
1128104065 == 113126
1129115741 == 113126
1130113403 Partial SSE3 support on x86
1131113541 vex: Grp5(x86) (alt encoding inc/dec) case 1
1132113642 valgrind crashes when trying to read debug information
1133113810 vex x86->IR: 66 0F F6 (66 + PSADBW == SSE PSADBW)
1134113796 read() and write() do not work if buffer is in shared memory
1135113851 vex x86->IR: (pmaddwd): 0x66 0xF 0xF5 0xC7
1136114366 vex amd64 cannnot handle __asm__( "fninit" )
1137114412 vex amd64->IR: 0xF 0xAD 0xC2 0xD3 (128-bit shift, shrdq?)
1138114455 vex amd64->IR: 0xF 0xAC 0xD0 0x1 (also shrdq)
1139115590: amd64->IR: 0x67 0xE3 0x9 0xEB (address size override)
1140115953 valgrind svn r5042 does not build with parallel make (-j3)
1141116057 maximum instruction size - VG_MAX_INSTR_SZB too small?
1142116483 shmat failes with invalid argument
1143102202 valgrind crashes when realloc'ing until out of memory
1144109487 == 102202
1145110536 == 102202
1146112687 == 102202
1147111724 vex amd64->IR: 0x41 0xF 0xAB (more BT{,S,R,C} fun n games)
1148111748 vex amd64->IR: 0xDD 0xE2 (fucom)
1149111785 make fails if CC contains spaces
1150111829 vex x86->IR: sbb AL, Ib
1151111851 vex x86->IR: 0x9F 0x89 (lahf/sahf)
1152112031 iopl on AMD64 and README_MISSING_SYSCALL_OR_IOCTL update
1153112152 code generation for Xin_MFence on x86 with SSE0 subarch
1154112167 == 112152
1155112789 == 112152
1156112199 naked ar tool is used in vex makefile
1157112501 vex x86->IR: movq (0xF 0x7F 0xC1 0xF) (mmx MOVQ)
1158113583 == 112501
1159112538 memalign crash
1160113190 Broken links in docs/html/
1161113230 Valgrind sys_pipe on x86-64 wrongly thinks file descriptors
1162 should be 64bit
1163113996 vex amd64->IR: fucomp (0xDD 0xE9)
1164114196 vex x86->IR: out %eax,(%dx) (0xEF 0xC9 0xC3 0x90)
1165114289 Memcheck fails to intercept malloc when used in an uclibc environment
1166114756 mbind syscall support
1167114757 Valgrind dies with assertion: Assertion 'noLargerThan > 0' failed
1168114563 stack tracking module not informed when valgrind switches threads
1169114564 clone() and stacks
1170114565 == 114564
1171115496 glibc crashes trying to use sysinfo page
1172116200 enable fsetxattr, fgetxattr, and fremovexattr for amd64
njnf63170c2005-11-12 19:09:51 +00001173
sewardjae17e7e2005-11-20 19:45:30 +00001174(3.1.0RC1: 20 November 2005, vex r1466, valgrind r5224).
sewardjfa462022005-11-26 13:38:58 +00001175(3.1.0: 26 November 2005, vex r1471, valgrind r5235).
sewardjae17e7e2005-11-20 19:45:30 +00001176
njnf63170c2005-11-12 19:09:51 +00001177
sewardjcd62fe72005-08-29 13:24:51 +00001178Release 3.0.1 (29 August 2005)
1179~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
11803.0.1 fixes a bunch of bugs reported in 3.0.0. There is no new
1181functionality. Some of the fixed bugs are critical, so if you
sewardjfa233892005-08-30 12:35:36 +00001182use/distribute 3.0.0, an upgrade to 3.0.1 is recommended. The fixed
sewardjcd62fe72005-08-29 13:24:51 +00001183bugs are:
1184
1185(note: "n-i-bz" means "not in bugzilla" -- this bug does not have
1186 a bugzilla entry).
1187
1188109313 (== 110505) x86 cmpxchg8b
1189n-i-bz x86: track but ignore changes to %eflags.AC (alignment check)
1190110102 dis_op2_E_G(amd64)
1191110202 x86 sys_waitpid(#286)
1192110203 clock_getres(,0)
1193110208 execve fail wrong retval
1194110274 SSE1 now mandatory for x86
1195110388 amd64 0xDD 0xD1
1196110464 amd64 0xDC 0x1D FCOMP
1197110478 amd64 0xF 0xD PREFETCH
1198n-i-bz XML <unique> printing wrong
1199n-i-bz Dirk r4359 (amd64 syscalls from trunk)
1200110591 amd64 and x86: rdtsc not implemented properly
1201n-i-bz Nick r4384 (stub implementations of Addrcheck and Helgrind)
1202110652 AMD64 valgrind crashes on cwtd instruction
1203110653 AMD64 valgrind crashes on sarb $0x4,foo(%rip) instruction
1204110656 PATH=/usr/bin::/bin valgrind foobar stats ./fooba
1205110657 Small test fixes
1206110671 vex x86->IR: unhandled instruction bytes: 0xF3 0xC3 (rep ret)
1207n-i-bz Nick (Cachegrind should not assert when it encounters a client
1208 request.)
1209110685 amd64->IR: unhandled instruction bytes: 0xE1 0x56 (loope Jb)
1210110830 configuring with --host fails to build 32 bit on 64 bit target
1211110875 Assertion when execve fails
1212n-i-bz Updates to Memcheck manual
1213n-i-bz Fixed broken malloc_usable_size()
1214110898 opteron instructions missing: btq btsq btrq bsfq
1215110954 x86->IR: unhandled instruction bytes: 0xE2 0xF6 (loop Jb)
1216n-i-bz Make suppressions work for "???" lines in stacktraces.
1217111006 bogus warnings from linuxthreads
1218111092 x86: dis_Grp2(Reg): unhandled case(x86)
1219111231 sctp_getladdrs() and sctp_getpaddrs() returns uninitialized
1220 memory
1221111102 (comment #4) Fixed 64-bit unclean "silly arg" message
1222n-i-bz vex x86->IR: unhandled instruction bytes: 0x14 0x0
1223n-i-bz minor umount/fcntl wrapper fixes
1224111090 Internal Error running Massif
1225101204 noisy warning
1226111513 Illegal opcode for SSE instruction (x86 movups)
1227111555 VEX/Makefile: CC is set to gcc
sewardjfa233892005-08-30 12:35:36 +00001228n-i-bz Fix XML bugs in FAQ
sewardjcd62fe72005-08-29 13:24:51 +00001229
sewardjfa233892005-08-30 12:35:36 +00001230(3.0.1: 29 August 05,
1231 vex/branches/VEX_3_0_BRANCH r1367,
1232 valgrind/branches/VALGRIND_3_0_BRANCH r4574).
sewardj820c7722005-08-29 13:44:43 +00001233
1234
sewardjcd62fe72005-08-29 13:24:51 +00001235
sewardjec39f332005-08-02 15:28:07 +00001236Release 3.0.0 (3 August 2005)
1237~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sewardjf9929e62005-08-02 21:05:58 +000012383.0.0 is a major overhaul of Valgrind. The most significant user
1239visible change is that Valgrind now supports architectures other than
1240x86. The new architectures it supports are AMD64 and PPC32, and the
1241infrastructure is present for other architectures to be added later.
njn4a9f6f42005-07-19 23:46:08 +00001242
sewardjf9929e62005-08-02 21:05:58 +00001243AMD64 support works well, but has some shortcomings:
njn4a9f6f42005-07-19 23:46:08 +00001244
njn7b1a6a42005-07-20 04:20:44 +00001245- It generally won't be as solid as the x86 version. For example,
1246 support for more obscure instructions and system calls may be missing.
1247 We will fix these as they arise.
njn4a9f6f42005-07-19 23:46:08 +00001248
sewardjc371bd02005-07-27 09:55:24 +00001249- Address space may be limited; see the point about
njn7b1a6a42005-07-20 04:20:44 +00001250 position-independent executables below.
njn4a9f6f42005-07-19 23:46:08 +00001251
njn7b1a6a42005-07-20 04:20:44 +00001252- If Valgrind is built on an AMD64 machine, it will only run 64-bit
1253 executables. If you want to run 32-bit x86 executables under Valgrind
1254 on an AMD64, you will need to build Valgrind on an x86 machine and
1255 copy it to the AMD64 machine. And it probably won't work if you do
1256 something tricky like exec'ing a 32-bit program from a 64-bit program
1257 while using --trace-children=yes. We hope to improve this situation
1258 in the future.
1259
1260The PPC32 support is very basic. It may not work reliably even for
sewardjc371bd02005-07-27 09:55:24 +00001261small programs, but it's a start. Many thanks to Paul Mackerras for
1262his great work that enabled this support. We are working to make
1263PPC32 usable as soon as possible.
njn4a9f6f42005-07-19 23:46:08 +00001264
1265Other user-visible changes:
1266
sewardjec39f332005-08-02 15:28:07 +00001267- Valgrind is no longer built by default as a position-independent
1268 executable (PIE), as this caused too many problems.
njn4a9f6f42005-07-19 23:46:08 +00001269
1270 Without PIE enabled, AMD64 programs will only be able to access 2GB of
1271 address space. We will fix this eventually, but not for the moment.
1272
1273 Use --enable-pie at configure-time to turn this on.
1274
1275- Support for programs that use stack-switching has been improved. Use
1276 the --max-stackframe flag for simple cases, and the
1277 VALGRIND_STACK_REGISTER, VALGRIND_STACK_DEREGISTER and
1278 VALGRIND_STACK_CHANGE client requests for trickier cases.
1279
njn7b1a6a42005-07-20 04:20:44 +00001280- Support for programs that use self-modifying code has been improved,
1281 in particular programs that put temporary code fragments on the stack.
1282 This helps for C programs compiled with GCC that use nested functions,
njne7723322005-07-24 17:19:09 +00001283 and also Ada programs. This is controlled with the --smc-check
njn7b1a6a42005-07-20 04:20:44 +00001284 flag, although the default setting should work in most cases.
njn4a9f6f42005-07-19 23:46:08 +00001285
njn7b1a6a42005-07-20 04:20:44 +00001286- Output can now be printed in XML format. This should make it easier
sewardjc371bd02005-07-27 09:55:24 +00001287 for tools such as GUI front-ends and automated error-processing
1288 schemes to use Valgrind output as input. The --xml flag controls this.
1289 As part of this change, ELF directory information is read from executables,
1290 so absolute source file paths are available if needed.
njn4a9f6f42005-07-19 23:46:08 +00001291
1292- Programs that allocate many heap blocks may run faster, due to
1293 improvements in certain data structures.
1294
njn7b1a6a42005-07-20 04:20:44 +00001295- Addrcheck is currently not working. We hope to get it working again
1296 soon. Helgrind is still not working, as was the case for the 2.4.0
1297 release.
njn4a9f6f42005-07-19 23:46:08 +00001298
sewardjc371bd02005-07-27 09:55:24 +00001299- The JITter has been completely rewritten, and is now in a separate
1300 library, called Vex. This enabled a lot of the user-visible changes,
1301 such as new architecture support. The new JIT unfortunately translates
1302 more slowly than the old one, so programs may take longer to start.
1303 We believe the code quality is produces is about the same, so once
1304 started, programs should run at about the same speed. Feedback about
1305 this would be useful.
1306
1307 On the plus side, Vex and hence Memcheck tracks value flow properly
1308 through floating point and vector registers, something the 2.X line
1309 could not do. That means that Memcheck is much more likely to be
1310 usably accurate on vectorised code.
1311
sewardjec39f332005-08-02 15:28:07 +00001312- There is a subtle change to the way exiting of threaded programs
sewardjc371bd02005-07-27 09:55:24 +00001313 is handled. In 3.0, Valgrind's final diagnostic output (leak check,
1314 etc) is not printed until the last thread exits. If the last thread
1315 to exit was not the original thread which started the program, any
1316 other process wait()-ing on this one to exit may conclude it has
1317 finished before the diagnostic output is printed. This may not be
1318 what you expect. 2.X had a different scheme which avoided this
1319 problem, but caused deadlocks under obscure circumstances, so we
1320 are trying something different for 3.0.
1321
1322- Small changes in control log file naming which make it easier to
sewardjec39f332005-08-02 15:28:07 +00001323 use valgrind for debugging MPI-based programs. The relevant
1324 new flags are --log-file-exactly= and --log-file-qualifier=.
sewardjc371bd02005-07-27 09:55:24 +00001325
1326- As part of adding AMD64 support, DWARF2 CFI-based stack unwinding
1327 support was added. In principle this means Valgrind can produce
1328 meaningful backtraces on x86 code compiled with -fomit-frame-pointer
tom81868f02005-07-27 10:33:08 +00001329 providing you also compile your code with -fasynchronous-unwind-tables.
sewardjc371bd02005-07-27 09:55:24 +00001330
sewardjec39f332005-08-02 15:28:07 +00001331- The documentation build system has been completely redone.
1332 The documentation masters are now in XML format, and from that
1333 HTML, PostScript and PDF documentation is generated. As a result
1334 the manual is now available in book form. Note that the
1335 documentation in the source tarballs is pre-built, so you don't need
1336 any XML processing tools to build Valgrind from a tarball.
njn4a9f6f42005-07-19 23:46:08 +00001337
1338Changes that are not user-visible:
1339
sewardjc371bd02005-07-27 09:55:24 +00001340- The code has been massively overhauled in order to modularise it.
1341 As a result we hope it is easier to navigate and understand.
njn4a9f6f42005-07-19 23:46:08 +00001342
1343- Lots of code has been rewritten.
1344
sewardjf9929e62005-08-02 21:05:58 +00001345BUGS FIXED:
njn4a9f6f42005-07-19 23:46:08 +00001346
sewardj79bd9b62005-08-03 16:40:10 +00001347110046 sz == 4 assertion failed
1348109810 vex amd64->IR: unhandled instruction bytes: 0xA3 0x4C 0x70 0xD7
sewardjf9929e62005-08-02 21:05:58 +00001349109802 Add a plausible_stack_size command-line parameter ?
1350109783 unhandled ioctl TIOCMGET (running hw detection tool discover)
1351109780 unhandled ioctl BLKSSZGET (running fdisk -l /dev/hda)
1352109718 vex x86->IR: unhandled instruction: ffreep
1353109429 AMD64 unhandled syscall: 127 (sigpending)
1354109401 false positive uninit in strchr from ld-linux.so.2
1355109385 "stabs" parse failure
1356109378 amd64: unhandled instruction REP NOP
1357109376 amd64: unhandled instruction LOOP Jb
1358109363 AMD64 unhandled instruction bytes
1359109362 AMD64 unhandled syscall: 24 (sched_yield)
1360109358 fork() won't work with valgrind-3.0 SVN
1361109332 amd64 unhandled instruction: ADC Ev, Gv
1362109314 Bogus memcheck report on amd64
1363108883 Crash; vg_memory.c:905 (vgPlain_init_shadow_range):
1364 Assertion `vgPlain_defined_init_shadow_page()' failed.
1365108349 mincore syscall parameter checked incorrectly
1366108059 build infrastructure: small update
1367107524 epoll_ctl event parameter checked on EPOLL_CTL_DEL
1368107123 Vex dies with unhandled instructions: 0xD9 0x31 0xF 0xAE
1369106841 auxmap & openGL problems
1370106713 SDL_Init causes valgrind to exit
1371106352 setcontext and makecontext not handled correctly
1372106293 addresses beyond initial client stack allocation
1373 not checked in VALGRIND_DO_LEAK_CHECK
1374106283 PIE client programs are loaded at address 0
1375105831 Assertion `vgPlain_defined_init_shadow_page()' failed.
1376105039 long run-times probably due to memory manager
1377104797 valgrind needs to be aware of BLKGETSIZE64
1378103594 unhandled instruction: FICOM
1379103320 Valgrind 2.4.0 fails to compile with gcc 3.4.3 and -O0
1380103168 potentially memory leak in coregrind/ume.c
1381102039 bad permissions for mapped region at address 0xB7C73680
1382101881 weird assertion problem
1383101543 Support fadvise64 syscalls
138475247 x86_64/amd64 support (the biggest "bug" we have ever fixed)
njn4a9f6f42005-07-19 23:46:08 +00001385
sewardj9a919512005-08-03 17:01:51 +00001386(3.0RC1: 27 July 05, vex r1303, valgrind r4283).
sewardj1346e132005-08-03 20:42:06 +00001387(3.0.0: 3 August 05, vex r1313, valgrind r4316).
sewardjbfa2b992005-07-27 17:49:17 +00001388
1389
1390
njn8c927302005-03-12 16:45:01 +00001391Stable release 2.4.0 (March 2005) -- CHANGES RELATIVE TO 2.2.0
njn1aa5ae42005-03-26 16:04:27 +00001392~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
13932.4.0 brings many significant changes and bug fixes. The most
1394significant user-visible change is that we no longer supply our own
1395pthread implementation. Instead, Valgrind is finally capable of
1396running the native thread library, either LinuxThreads or NPTL.
njn8c927302005-03-12 16:45:01 +00001397
njn1aa5ae42005-03-26 16:04:27 +00001398This means our libpthread has gone, along with the bugs associated
1399with it. Valgrind now supports the kernel's threading syscalls, and
1400lets you use your standard system libpthread. As a result:
njn8c927302005-03-12 16:45:01 +00001401
njn1aa5ae42005-03-26 16:04:27 +00001402* There are many fewer system dependencies and strange library-related
1403 bugs. There is a small performance improvement, and a large
1404 stability improvement.
njn8c927302005-03-12 16:45:01 +00001405
njn1aa5ae42005-03-26 16:04:27 +00001406* On the downside, Valgrind can no longer report misuses of the POSIX
1407 PThreads API. It also means that Helgrind currently does not work.
1408 We hope to fix these problems in a future release.
njn8c927302005-03-12 16:45:01 +00001409
njn1aa5ae42005-03-26 16:04:27 +00001410Note that running the native thread libraries does not mean Valgrind
1411is able to provide genuine concurrent execution on SMPs. We still
1412impose the restriction that only one thread is running at any given
1413time.
njn8c927302005-03-12 16:45:01 +00001414
njn1aa5ae42005-03-26 16:04:27 +00001415There are many other significant changes too:
njn8c927302005-03-12 16:45:01 +00001416
njn1aa5ae42005-03-26 16:04:27 +00001417* Memcheck is (once again) the default tool.
njn8c927302005-03-12 16:45:01 +00001418
njn1aa5ae42005-03-26 16:04:27 +00001419* The default stack backtrace is now 12 call frames, rather than 4.
njn8c927302005-03-12 16:45:01 +00001420
njn1aa5ae42005-03-26 16:04:27 +00001421* Suppressions can have up to 25 call frame matches, rather than 4.
njn8c927302005-03-12 16:45:01 +00001422
njn1aa5ae42005-03-26 16:04:27 +00001423* Memcheck and Addrcheck use less memory. Under some circumstances,
1424 they no longer allocate shadow memory if there are large regions of
1425 memory with the same A/V states - such as an mmaped file.
njn8c927302005-03-12 16:45:01 +00001426
njn1aa5ae42005-03-26 16:04:27 +00001427* The memory-leak detector in Memcheck and Addrcheck has been
1428 improved. It now reports more types of memory leak, including
1429 leaked cycles. When reporting leaked memory, it can distinguish
1430 between directly leaked memory (memory with no references), and
1431 indirectly leaked memory (memory only referred to by other leaked
1432 memory).
njn8c927302005-03-12 16:45:01 +00001433
njn1aa5ae42005-03-26 16:04:27 +00001434* Memcheck's confusion over the effect of mprotect() has been fixed:
1435 previously mprotect could erroneously mark undefined data as
1436 defined.
njn8c927302005-03-12 16:45:01 +00001437
njn1aa5ae42005-03-26 16:04:27 +00001438* Signal handling is much improved and should be very close to what
1439 you get when running natively.
1440
1441 One result of this is that Valgrind observes changes to sigcontexts
1442 passed to signal handlers. Such modifications will take effect when
1443 the signal returns. You will need to run with --single-step=yes to
1444 make this useful.
njn8c927302005-03-12 16:45:01 +00001445
1446* Valgrind is built in Position Independent Executable (PIE) format if
njn1aa5ae42005-03-26 16:04:27 +00001447 your toolchain supports it. This allows it to take advantage of all
njn8c927302005-03-12 16:45:01 +00001448 the available address space on systems with 4Gbyte user address
1449 spaces.
1450
1451* Valgrind can now run itself (requires PIE support).
1452
njn1aa5ae42005-03-26 16:04:27 +00001453* Syscall arguments are now checked for validity. Previously all
1454 memory used by syscalls was checked, but now the actual values
1455 passed are also checked.
njn8c927302005-03-12 16:45:01 +00001456
njn1aa5ae42005-03-26 16:04:27 +00001457* Syscall wrappers are more robust against bad addresses being passed
1458 to syscalls: they will fail with EFAULT rather than killing Valgrind
1459 with SIGSEGV.
njn8c927302005-03-12 16:45:01 +00001460
njn1aa5ae42005-03-26 16:04:27 +00001461* Because clone() is directly supported, some non-pthread uses of it
1462 will work. Partial sharing (where some resources are shared, and
1463 some are not) is not supported.
1464
1465* open() and readlink() on /proc/self/exe are supported.
1466
1467BUGS FIXED:
1468
146988520 pipe+fork+dup2 kills the main program
147088604 Valgrind Aborts when using $VALGRIND_OPTS and user progra...
147188614 valgrind: vg_libpthread.c:2323 (read): Assertion `read_pt...
147288703 Stabs parser fails to handle ";"
147388886 ioctl wrappers for TIOCMBIS and TIOCMBIC
147489032 valgrind pthread_cond_timedwait fails
147589106 the 'impossible' happened
147689139 Missing sched_setaffinity & sched_getaffinity
147789198 valgrind lacks support for SIOCSPGRP and SIOCGPGRP
147889263 Missing ioctl translations for scsi-generic and CD playing
147989440 tests/deadlock.c line endings
148089481 `impossible' happened: EXEC FAILED
148189663 valgrind 2.2.0 crash on Redhat 7.2
148289792 Report pthread_mutex_lock() deadlocks instead of returnin...
148390111 statvfs64 gives invalid error/warning
148490128 crash+memory fault with stabs generated by gnat for a run...
148590778 VALGRIND_CHECK_DEFINED() not as documented in memcheck.h
148690834 cachegrind crashes at end of program without reporting re...
148791028 valgrind: vg_memory.c:229 (vgPlain_unmap_range): Assertio...
148891162 valgrind crash while debugging drivel 1.2.1
148991199 Unimplemented function
149091325 Signal routing does not propagate the siginfo structure
149191599 Assertion `cv == ((void *)0)'
149291604 rw_lookup clears orig and sends the NULL value to rw_new
149391821 Small problems building valgrind with $top_builddir ne $t...
149491844 signal 11 (SIGSEGV) at get_tcb (libpthread.c:86) in corec...
149592264 UNIMPLEMENTED FUNCTION: pthread_condattr_setpshared
149692331 per-target flags necessitate AM_PROG_CC_C_O
149792420 valgrind doesn't compile with linux 2.6.8.1/9
149892513 Valgrind 2.2.0 generates some warning messages
149992528 vg_symtab2.c:170 (addLoc): Assertion `loc->size > 0' failed.
150093096 unhandled ioctl 0x4B3A and 0x5601
150193117 Tool and core interface versions do not match
150293128 Can't run valgrind --tool=memcheck because of unimplement...
150393174 Valgrind can crash if passed bad args to certain syscalls
150493309 Stack frame in new thread is badly aligned
150593328 Wrong types used with sys_sigprocmask()
150693763 /usr/include/asm/msr.h is missing
150793776 valgrind: vg_memory.c:508 (vgPlain_find_map_space): Asser...
150893810 fcntl() argument checking a bit too strict
150994378 Assertion `tst->sigqueue_head != tst->sigqueue_tail' failed.
151094429 valgrind 2.2.0 segfault with mmap64 in glibc 2.3.3
151194645 Impossible happened: PINSRW mem
151294953 valgrind: the `impossible' happened: SIGSEGV
151395667 Valgrind does not work with any KDE app
151496243 Assertion 'res==0' failed
151596252 stage2 loader of valgrind fails to allocate memory
151696520 All programs crashing at _dl_start (in /lib/ld-2.3.3.so) ...
151796660 ioctl CDROMREADTOCENTRY causes bogus warnings
151896747 After looping in a segfault handler, the impossible happens
151996923 Zero sized arrays crash valgrind trace back with SIGFPE
152096948 valgrind stops with assertion failure regarding mmap2
152196966 valgrind fails when application opens more than 16 sockets
152297398 valgrind: vg_libpthread.c:2667 Assertion failed
152397407 valgrind: vg_mylibc.c:1226 (vgPlain_safe_fd): Assertion `...
152497427 "Warning: invalid file descriptor -1 in syscall close()" ...
152597785 missing backtrace
152697792 build in obj dir fails - autoconf / makefile cleanup
152797880 pthread_mutex_lock fails from shared library (special ker...
152897975 program aborts without ang VG messages
152998129 Failed when open and close file 230000 times using stdio
153098175 Crashes when using valgrind-2.2.0 with a program using al...
153198288 Massif broken
153298303 UNIMPLEMENTED FUNCTION pthread_condattr_setpshared
153398630 failed--compilation missing warnings.pm, fails to make he...
153498756 Cannot valgrind signal-heavy kdrive X server
153598966 valgrinding the JVM fails with a sanity check assertion
153699035 Valgrind crashes while profiling
153799142 loops with message "Signal 11 being dropped from thread 0...
153899195 threaded apps crash on thread start (using QThread::start...
153999348 Assertion `vgPlain_lseek(core_fd, 0, 1) == phdrs[i].p_off...
154099568 False negative due to mishandling of mprotect
154199738 valgrind memcheck crashes on program that uses sigitimer
154299923 0-sized allocations are reported as leaks
154399949 program seg faults after exit()
1544100036 "newSuperblock's request for 1048576 bytes failed"
1545100116 valgrind: (pthread_cond_init): Assertion `sizeof(* cond) ...
1546100486 memcheck reports "valgrind: the `impossible' happened: V...
1547100833 second call to "mremap" fails with EINVAL
1548101156 (vgPlain_find_map_space): Assertion `(addr & ((1 << 12)-1...
1549101173 Assertion `recDepth >= 0 && recDepth < 500' failed
1550101291 creating threads in a forked process fails
1551101313 valgrind causes different behavior when resizing a window...
1552101423 segfault for c++ array of floats
1553101562 valgrind massif dies on SIGINT even with signal handler r...
1554
njn8c927302005-03-12 16:45:01 +00001555
jseward34ed74f2004-08-30 18:04:42 +00001556Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.0.0
1557~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
jseward7d180d92004-08-30 18:15:37 +000015582.2.0 brings nine months worth of improvements and bug fixes. We
1559believe it to be a worthy successor to 2.0.0. There are literally
1560hundreds of bug fixes and minor improvements. There are also some
1561fairly major user-visible changes:
1562
1563* A complete overhaul of handling of system calls and signals, and
1564 their interaction with threads. In general, the accuracy of the
1565 system call, thread and signal simulations is much improved:
1566
1567 - Blocking system calls behave exactly as they do when running
1568 natively (not on valgrind). That is, if a syscall blocks only the
1569 calling thread when running natively, than it behaves the same on
1570 valgrind. No more mysterious hangs because V doesn't know that some
1571 syscall or other, should block only the calling thread.
1572
1573 - Interrupted syscalls should now give more faithful results.
1574
1575 - Signal contexts in signal handlers are supported.
1576
1577* Improvements to NPTL support to the extent that V now works
1578 properly on NPTL-only setups.
1579
1580* Greater isolation between Valgrind and the program being run, so
1581 the program is less likely to inadvertently kill Valgrind by
1582 doing wild writes.
1583
1584* Massif: a new space profiling tool. Try it! It's cool, and it'll
1585 tell you in detail where and when your C/C++ code is allocating heap.
1586 Draws pretty .ps pictures of memory use against time. A potentially
1587 powerful tool for making sense of your program's space use.
1588
1589* File descriptor leakage checks. When enabled, Valgrind will print out
1590 a list of open file descriptors on exit.
1591
1592* Improved SSE2/SSE3 support.
1593
jseward2eb002b2004-08-31 00:14:02 +00001594* Time-stamped output; use --time-stamp=yes
1595
jseward34ed74f2004-08-30 18:04:42 +00001596
1597
1598Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.1.2
1599~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
16002.2.0 is not much different from 2.1.2, released seven weeks ago.
1601A number of bugs have been fixed, most notably #85658, which gave
1602problems for quite a few people. There have been many internal
1603cleanups, but those are not user visible.
1604
1605The following bugs have been fixed since 2.1.2:
1606
160785658 Assert in coregrind/vg_libpthread.c:2326 (open64) !=
1608 (void*)0 failed
1609 This bug was reported multiple times, and so the following
1610 duplicates of it are also fixed: 87620, 85796, 85935, 86065,
1611 86919, 86988, 87917, 88156
1612
161380716 Semaphore mapping bug caused by unmap (sem_destroy)
1614 (Was fixed prior to 2.1.2)
1615
161686987 semctl and shmctl syscalls family is not handled properly
1617
161886696 valgrind 2.1.2 + RH AS2.1 + librt
1619
162086730 valgrind locks up at end of run with assertion failure
1621 in __pthread_unwind
1622
162386641 memcheck doesn't work with Mesa OpenGL/ATI on Suse 9.1
1624 (also fixes 74298, a duplicate of this)
1625
162685947 MMX/SSE unhandled instruction 'sfence'
1627
162884978 Wrong error "Conditional jump or move depends on
1629 uninitialised value" resulting from "sbbl %reg, %reg"
1630
163186254 ssort() fails when signed int return type from comparison is
1632 too small to handle result of unsigned int subtraction
1633
163487089 memalign( 4, xxx) makes valgrind assert
1635
jseward2eb002b2004-08-31 00:14:02 +0000163686407 Add support for low-level parallel port driver ioctls.
jseward34ed74f2004-08-30 18:04:42 +00001637
163870587 Add timestamps to Valgrind output? (wishlist)
1639
164084937 vg_libpthread.c:2505 (se_remap): Assertion `res == 0'
1641 (fixed prior to 2.1.2)
1642
164386317 cannot load libSDL-1.2.so.0 using valgrind
1644
164586989 memcpy from mac_replace_strmem.c complains about
1646 uninitialized pointers passed when length to copy is zero
1647
164885811 gnu pascal symbol causes segmentation fault; ok in 2.0.0
1649
165079138 writing to sbrk()'d memory causes segfault
1651
165277369 sched deadlock while signal received during pthread_join
1653 and the joined thread exited
1654
165588115 In signal handler for SIGFPE, siginfo->si_addr is wrong
1656 under Valgrind
1657
165878765 Massif crashes on app exit if FP exceptions are enabled
1659
1660Additionally there are the following changes, which are not
1661connected to any bug report numbers, AFAICS:
1662
1663* Fix scary bug causing mis-identification of SSE stores vs
1664 loads and so causing memcheck to sometimes give nonsense results
1665 on SSE code.
1666
1667* Add support for the POSIX message queue system calls.
1668
1669* Fix to allow 32-bit Valgrind to run on AMD64 boxes. Note: this does
1670 NOT allow Valgrind to work with 64-bit executables - only with 32-bit
1671 executables on an AMD64 box.
1672
1673* At configure time, only check whether linux/mii.h can be processed
1674 so that we don't generate ugly warnings by trying to compile it.
1675
jseward2eb002b2004-08-31 00:14:02 +00001676* Add support for POSIX clocks and timers.
1677
jseward34ed74f2004-08-30 18:04:42 +00001678
1679
jseward9cd978c2004-07-17 13:38:12 +00001680Developer (cvs head) release 2.1.2 (18 July 2004)
jsewardbd0a6e42004-07-17 12:19:43 +00001681~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
16822.1.2 contains four months worth of bug fixes and refinements.
jseward9cd978c2004-07-17 13:38:12 +00001683Although officially a developer release, we believe it to be stable
1684enough for widespread day-to-day use. 2.1.2 is pretty good, so try it
1685first, although there is a chance it won't work. If so then try 2.0.0
1686and tell us what went wrong." 2.1.2 fixes a lot of problems present
1687in 2.0.0 and is generally a much better product.
jsewardbd0a6e42004-07-17 12:19:43 +00001688
jseward9cd978c2004-07-17 13:38:12 +00001689Relative to 2.1.1, a large number of minor problems with 2.1.1 have
1690been fixed, and so if you use 2.1.1 you should try 2.1.2. Users of
1691the last stable release, 2.0.0, might also want to try this release.
jsewardbd0a6e42004-07-17 12:19:43 +00001692
1693The following bugs, and probably many more, have been fixed. These
1694are listed at http://bugs.kde.org. Reporting a bug for valgrind in
1695the http://bugs.kde.org is much more likely to get you a fix than
1696mailing developers directly, so please continue to keep sending bugs
1697there.
1698
169976869 Crashes when running any tool under Fedora Core 2 test1
1700 This fixes the problem with returning from a signal handler
jseward0400d4d2004-07-17 23:26:46 +00001701 when VDSOs are turned off in FC2.
jsewardbd0a6e42004-07-17 12:19:43 +00001702
jseward0400d4d2004-07-17 23:26:46 +0000170369508 java 1.4.2 client fails with erroneous "stack size too small".
1704 This fix makes more of the pthread stack attribute related
1705 functions work properly. Java still doesn't work though.
jsewardbd0a6e42004-07-17 12:19:43 +00001706
170771906 malloc alignment should be 8, not 4
1708 All memory returned by malloc/new etc is now at least
1709 8-byte aligned.
1710
jsewarda6017a42004-07-17 23:44:35 +0000171181970 vg_alloc_ThreadState: no free slots available
1712 (closed because the workaround is simple: increase
1713 VG_N_THREADS, rebuild and try again.)
1714
jsewardbd0a6e42004-07-17 12:19:43 +0000171578514 Conditional jump or move depends on uninitialized value(s)
1716 (a slight mishanding of FP code in memcheck)
1717
jsewarda6017a42004-07-17 23:44:35 +0000171877952 pThread Support (crash) (due to initialisation-ordering probs)
1719 (also 85118)
1720
jsewardbd0a6e42004-07-17 12:19:43 +0000172180942 Addrcheck wasn't doing overlap checking as it should.
172278048 return NULL on malloc/new etc failure, instead of asserting
172373655 operator new() override in user .so files often doesn't get picked up
172483060 Valgrind does not handle native kernel AIO
172569872 Create proper coredumps after fatal signals
172682026 failure with new glibc versions: __libc_* functions are not exported
172770344 UNIMPLEMENTED FUNCTION: tcdrain
172881297 Cancellation of pthread_cond_wait does not require mutex
172982872 Using debug info from additional packages (wishlist)
173083025 Support for ioctls FIGETBSZ and FIBMAP
173183340 Support for ioctl HDIO_GET_IDENTITY
173279714 Support for the semtimedop system call.
173377022 Support for ioctls FBIOGET_VSCREENINFO and FBIOGET_FSCREENINFO
173482098 hp2ps ansification (wishlist)
173583573 Valgrind SIGSEGV on execve
173682999 show which cmdline option was erroneous (wishlist)
173783040 make valgrind VPATH and distcheck-clean (wishlist)
jsewarda6017a42004-07-17 23:44:35 +0000173883998 Assertion `newfd > vgPlain_max_fd' failed (see below)
173982722 Unchecked mmap in as_pad leads to mysterious failures later
174078958 memcheck seg faults while running Mozilla
jsewarda2c08002004-07-18 09:24:57 +0000174185416 Arguments with colon (e.g. --logsocket) ignored
jsewardbd0a6e42004-07-17 12:19:43 +00001742
1743
1744Additionally there are the following changes, which are not
1745connected to any bug report numbers, AFAICS:
1746
1747* Rearranged address space layout relative to 2.1.1, so that
1748 Valgrind/tools will run out of memory later than currently in many
1749 circumstances. This is good news esp. for Calltree. It should
1750 be possible for client programs to allocate over 800MB of
1751 memory when using memcheck now.
1752
jseward9cd978c2004-07-17 13:38:12 +00001753* Improved checking when laying out memory. Should hopefully avoid
1754 the random segmentation faults that 2.1.1 sometimes caused.
1755
jsewardbd0a6e42004-07-17 12:19:43 +00001756* Support for Fedora Core 2 and SuSE 9.1. Improvements to NPTL
1757 support to the extent that V now works properly on NPTL-only setups.
1758
1759* Renamed the following options:
1760 --logfile-fd --> --log-fd
1761 --logfile --> --log-file
1762 --logsocket --> --log-socket
1763 to be consistent with each other and other options (esp. --input-fd).
1764
1765* Add support for SIOCGMIIPHY, SIOCGMIIREG and SIOCSMIIREG ioctls and
1766 improve the checking of other interface related ioctls.
1767
jsewardbd0a6e42004-07-17 12:19:43 +00001768* Fix building with gcc-3.4.1.
1769
1770* Remove limit on number of semaphores supported.
1771
jsewardbd0a6e42004-07-17 12:19:43 +00001772* Add support for syscalls: set_tid_address (258), acct (51).
1773
1774* Support instruction "repne movs" -- not official but seems to occur.
1775
1776* Implement an emulated soft limit for file descriptors in addition to
1777 the current reserved area, which effectively acts as a hard limit. The
1778 setrlimit system call now simply updates the emulated limits as best
1779 as possible - the hard limit is not allowed to move at all and just
1780 returns EPERM if you try and change it. This should stop reductions
1781 in the soft limit causing assertions when valgrind tries to allocate
1782 descriptors from the reserved area.
jsewarda6017a42004-07-17 23:44:35 +00001783 (This actually came from bug #83998).
jsewardbd0a6e42004-07-17 12:19:43 +00001784
nethercote76e8fd92004-07-21 10:37:31 +00001785* Major overhaul of Cachegrind implementation. First user-visible change
nethercote5e63bfb2004-07-17 12:49:53 +00001786 is that cachegrind.out files are now typically 90% smaller than they
nethercote76e8fd92004-07-21 10:37:31 +00001787 used to be; code annotation times are correspondingly much smaller.
1788 Second user-visible change is that hit/miss counts for code that is
1789 unloaded at run-time is no longer dumped into a single "discard" pile,
1790 but accurately preserved.
jsewardbd0a6e42004-07-17 12:19:43 +00001791
1792* Client requests for telling valgrind about memory pools.
1793
1794
1795
jseward9cd978c2004-07-17 13:38:12 +00001796Developer (cvs head) release 2.1.1 (12 March 2004)
1797~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
jsewardbd0a6e42004-07-17 12:19:43 +000017982.1.1 contains some internal structural changes needed for V's
jsewarde3e77042004-03-12 21:07:05 +00001799long-term future. These don't affect end-users. Most notable
1800user-visible changes are:
1801
1802* Greater isolation between Valgrind and the program being run, so
1803 the program is less likely to inadvertently kill Valgrind by
1804 doing wild writes.
1805
1806* Massif: a new space profiling tool. Try it! It's cool, and it'll
1807 tell you in detail where and when your C/C++ code is allocating heap.
1808 Draws pretty .ps pictures of memory use against time. A potentially
1809 powerful tool for making sense of your program's space use.
1810
1811* Fixes for many bugs, including support for more SSE2/SSE3 instructions,
1812 various signal/syscall things, and various problems with debug
1813 info readers.
1814
1815* Support for glibc-2.3.3 based systems.
1816
1817We are now doing automatic overnight build-and-test runs on a variety
1818of distros. As a result, we believe 2.1.1 builds and runs on:
1819Red Hat 7.2, 7.3, 8.0, 9, Fedora Core 1, SuSE 8.2, SuSE 9.
1820
1821
1822The following bugs, and probably many more, have been fixed. These
1823are listed at http://bugs.kde.org. Reporting a bug for valgrind in
1824the http://bugs.kde.org is much more likely to get you a fix than
1825mailing developers directly, so please continue to keep sending bugs
1826there.
1827
182869616 glibc 2.3.2 w/NPTL is massively different than what valgrind expects
182969856 I don't know how to instrument MMXish stuff (Helgrind)
183073892 valgrind segfaults starting with Objective-C debug info
1831 (fix for S-type stabs)
183273145 Valgrind complains too much about close(<reserved fd>)
183373902 Shadow memory allocation seems to fail on RedHat 8.0
183468633 VG_N_SEMAPHORES too low (V itself was leaking semaphores)
183575099 impossible to trace multiprocess programs
183676839 the `impossible' happened: disInstr: INT but not 0x80 !
183776762 vg_to_ucode.c:3748 (dis_push_segreg): Assertion `sz == 4' failed.
183876747 cannot include valgrind.h in c++ program
183976223 parsing B(3,10) gave NULL type => impossible happens
184075604 shmdt handling problem
184176416 Problems with gcc 3.4 snap 20040225
184275614 using -gstabs when building your programs the `impossible' happened
184375787 Patch for some CDROM ioctls CDORM_GET_MCN, CDROM_SEND_PACKET,
184475294 gcc 3.4 snapshot's libstdc++ have unsupported instructions.
1845 (REP RET)
184673326 vg_symtab2.c:272 (addScopeRange): Assertion `range->size > 0' failed.
184772596 not recognizing __libc_malloc
184869489 Would like to attach ddd to running program
184972781 Cachegrind crashes with kde programs
185073055 Illegal operand at DXTCV11CompressBlockSSE2 (more SSE opcodes)
185173026 Descriptor leak check reports port numbers wrongly
185271705 README_MISSING_SYSCALL_OR_IOCTL out of date
185372643 Improve support for SSE/SSE2 instructions
185472484 valgrind leaves it's own signal mask in place when execing
185572650 Signal Handling always seems to restart system calls
185672006 The mmap system call turns all errors in ENOMEM
185771781 gdb attach is pretty useless
185871180 unhandled instruction bytes: 0xF 0xAE 0x85 0xE8
185969886 writes to zero page cause valgrind to assert on exit
186071791 crash when valgrinding gimp 1.3 (stabs reader problem)
186169783 unhandled syscall: 218
186269782 unhandled instruction bytes: 0x66 0xF 0x2B 0x80
186370385 valgrind fails if the soft file descriptor limit is less
1864 than about 828
186569529 "rep; nop" should do a yield
186670827 programs with lots of shared libraries report "mmap failed"
1867 for some of them when reading symbols
186871028 glibc's strnlen is optimised enough to confuse valgrind
1869
1870
1871
1872
jseward43230652003-12-14 00:14:54 +00001873Unstable (cvs head) release 2.1.0 (15 December 2003)
1874~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1875For whatever it's worth, 2.1.0 actually seems pretty darn stable to me
1876(Julian). It looks eminently usable, and given that it fixes some
1877significant bugs, may well be worth using on a day-to-day basis.
18782.1.0 is known to build and pass regression tests on: SuSE 9, SuSE
18798.2, RedHat 8.
1880
18812.1.0 most notably includes Jeremy Fitzhardinge's complete overhaul of
1882handling of system calls and signals, and their interaction with
1883threads. In general, the accuracy of the system call, thread and
1884signal simulations is much improved. Specifically:
1885
1886- Blocking system calls behave exactly as they do when running
1887 natively (not on valgrind). That is, if a syscall blocks only the
1888 calling thread when running natively, than it behaves the same on
1889 valgrind. No more mysterious hangs because V doesn't know that some
1890 syscall or other, should block only the calling thread.
1891
1892- Interrupted syscalls should now give more faithful results.
1893
1894- Finally, signal contexts in signal handlers are supported. As a
1895 result, konqueror on SuSE 9 no longer segfaults when notified of
1896 file changes in directories it is watching.
1897
1898Other changes:
1899
1900- Robert Walsh's file descriptor leakage checks. When enabled,
1901 Valgrind will print out a list of open file descriptors on
1902 exit. Along with each file descriptor, Valgrind prints out a stack
1903 backtrace of where the file was opened and any details relating to the
1904 file descriptor such as the file name or socket details.
1905 To use, give: --track-fds=yes
1906
1907- Implemented a few more SSE/SSE2 instructions.
1908
1909- Less crud on the stack when you do 'where' inside a GDB attach.
1910
1911- Fixed the following bugs:
1912 68360: Valgrind does not compile against 2.6.0-testX kernels
1913 68525: CVS head doesn't compile on C90 compilers
1914 68566: pkgconfig support (wishlist)
1915 68588: Assertion `sz == 4' failed in vg_to_ucode.c (disInstr)
1916 69140: valgrind not able to explicitly specify a path to a binary.
1917 69432: helgrind asserts encountering a MutexErr when there are
1918 EraserErr suppressions
1919
jseward0b12daf2003-12-14 14:35:28 +00001920- Increase the max size of the translation cache from 200k average bbs
1921 to 300k average bbs. Programs on the size of OOo (680m17) are
1922 thrashing the cache at the smaller size, creating large numbers of
1923 retranslations and wasting significant time as a result.
1924
jseward43230652003-12-14 00:14:54 +00001925
1926
1927Stable release 2.0.0 (5 Nov 2003)
1928~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1929
19302.0.0 improves SSE/SSE2 support, fixes some minor bugs, and
1931improves support for SuSE 9 and the Red Hat "Severn" beta.
1932
1933- Further improvements to SSE/SSE2 support. The entire test suite of
1934 the GNU Scientific Library (gsl-1.4) compiled with Intel Icc 7.1
1935 20030307Z '-g -O -xW' now works. I think this gives pretty good
1936 coverage of SSE/SSE2 floating point instructions, or at least the
1937 subset emitted by Icc.
1938
1939- Also added support for the following instructions:
1940 MOVNTDQ UCOMISD UNPCKLPS UNPCKHPS SQRTSS
1941 PUSH/POP %{FS,GS}, and PUSH %CS (Nb: there is no POP %CS).
1942
1943- CFI support for GDB version 6. Needed to enable newer GDBs
1944 to figure out where they are when using --gdb-attach=yes.
1945
1946- Fix this:
1947 mc_translate.c:1091 (memcheck_instrument): Assertion
1948 `u_in->size == 4 || u_in->size == 16' failed.
1949
1950- Return an error rather than panicing when given a bad socketcall.
1951
1952- Fix checking of syscall rt_sigtimedwait().
1953
1954- Implement __NR_clock_gettime (syscall 265). Needed on Red Hat Severn.
1955
1956- Fixed bug in overlap check in strncpy() -- it was assuming the src was 'n'
1957 bytes long, when it could be shorter, which could cause false
1958 positives.
1959
1960- Support use of select() for very large numbers of file descriptors.
1961
1962- Don't fail silently if the executable is statically linked, or is
1963 setuid/setgid. Print an error message instead.
1964
1965- Support for old DWARF-1 format line number info.
1966
1967
1968
1969Snapshot 20031012 (12 October 2003)
1970~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1971
1972Three months worth of bug fixes, roughly. Most significant single
1973change is improved SSE/SSE2 support, mostly thanks to Dirk Mueller.
1974
197520031012 builds on Red Hat Fedora ("Severn") but doesn't really work
1976(curiosly, mozilla runs OK, but a modest "ls -l" bombs). I hope to
1977get a working version out soon. It may or may not work ok on the
1978forthcoming SuSE 9; I hear positive noises about it but haven't been
1979able to verify this myself (not until I get hold of a copy of 9).
1980
1981A detailed list of changes, in no particular order:
1982
1983- Describe --gen-suppressions in the FAQ.
1984
1985- Syscall __NR_waitpid supported.
1986
1987- Minor MMX bug fix.
1988
1989- -v prints program's argv[] at startup.
1990
1991- More glibc-2.3 suppressions.
1992
1993- Suppressions for stack underrun bug(s) in the c++ support library
1994 distributed with Intel Icc 7.0.
1995
1996- Fix problems reading /proc/self/maps.
1997
1998- Fix a couple of messages that should have been suppressed by -q,
1999 but weren't.
2000
2001- Make Addrcheck understand "Overlap" suppressions.
2002
2003- At startup, check if program is statically linked and bail out if so.
2004
2005- Cachegrind: Auto-detect Intel Pentium-M, also VIA Nehemiah
2006
2007- Memcheck/addrcheck: minor speed optimisations
2008
2009- Handle syscall __NR_brk more correctly than before.
2010
2011- Fixed incorrect allocate/free mismatch errors when using
2012 operator new(unsigned, std::nothrow_t const&)
2013 operator new[](unsigned, std::nothrow_t const&)
2014
2015- Support POSIX pthread spinlocks.
2016
2017- Fixups for clean compilation with gcc-3.3.1.
daywalker5d945de2003-09-26 00:32:53 +00002018
2019- Implemented more opcodes:
2020 - push %es
2021 - push %ds
2022 - pop %es
2023 - pop %ds
2024 - movntq
2025 - sfence
2026 - pshufw
2027 - pavgb
2028 - ucomiss
2029 - enter
daywalkerb18d2532003-09-27 20:15:01 +00002030 - mov imm32, %esp
2031 - all "in" and "out" opcodes
daywalker79aad842003-09-30 22:58:12 +00002032 - inc/dec %esp
jseward43230652003-12-14 00:14:54 +00002033 - A whole bunch of SSE/SSE2 instructions
daywalker5d945de2003-09-26 00:32:53 +00002034
jseward43230652003-12-14 00:14:54 +00002035- Memcheck: don't bomb on SSE/SSE2 code.
daywalker5d945de2003-09-26 00:32:53 +00002036
sewardj90238792003-05-05 00:23:42 +00002037
sewardj945f39f2003-07-25 21:11:40 +00002038Snapshot 20030725 (25 July 2003)
2039~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2040
2041Fixes some minor problems in 20030716.
2042
2043- Fix bugs in overlap checking for strcpy/memcpy etc.
2044
2045- Do overlap checking with Addrcheck as well as Memcheck.
2046
2047- Fix this:
2048 Memcheck: the `impossible' happened:
2049 get_error_name: unexpected type
2050
2051- Install headers needed to compile new skins.
2052
sewardje3dd2e02003-07-27 08:04:48 +00002053- Remove leading spaces and colon in the LD_LIBRARY_PATH / LD_PRELOAD
sewardj945f39f2003-07-25 21:11:40 +00002054 passed to non-traced children.
2055
sewardjb9eda952003-07-26 21:39:05 +00002056- Fix file descriptor leak in valgrind-listener.
2057
sewardje3dd2e02003-07-27 08:04:48 +00002058- Fix longstanding bug in which the allocation point of a
2059 block resized by realloc was not correctly set. This may
2060 have caused confusing error messages.
sewardj945f39f2003-07-25 21:11:40 +00002061
2062
sewardj626fd892003-07-16 20:10:26 +00002063Snapshot 20030716 (16 July 2003)
sewardj9d916ed2003-07-14 23:38:40 +00002064~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2065
sewardj626fd892003-07-16 20:10:26 +0000206620030716 is a snapshot of our current CVS head (development) branch.
sewardj9d916ed2003-07-14 23:38:40 +00002067This is the branch which will become valgrind-2.0. It contains
2068significant enhancements over the 1.9.X branch.
2069
2070Despite this being a snapshot of the CVS head, it is believed to be
2071quite stable -- at least as stable as 1.9.6 or 1.0.4, if not more so
2072-- and therefore suitable for widespread use. Please let us know asap
2073if it causes problems for you.
2074
2075Two reasons for releasing a snapshot now are:
2076
2077- It's been a while since 1.9.6, and this snapshot fixes
2078 various problems that 1.9.6 has with threaded programs
2079 on glibc-2.3.X based systems.
2080
2081- So as to make available improvements in the 2.0 line.
2082
sewardj626fd892003-07-16 20:10:26 +00002083Major changes in 20030716, as compared to 1.9.6:
sewardj9d916ed2003-07-14 23:38:40 +00002084
2085- More fixes to threading support on glibc-2.3.1 and 2.3.2-based
2086 systems (SuSE 8.2, Red Hat 9). If you have had problems
2087 with inconsistent/illogical behaviour of errno, h_errno or the DNS
sewardj626fd892003-07-16 20:10:26 +00002088 resolver functions in threaded programs, 20030716 should improve
sewardj9d916ed2003-07-14 23:38:40 +00002089 matters. This snapshot seems stable enough to run OpenOffice.org
2090 1.1rc on Red Hat 7.3, SuSE 8.2 and Red Hat 9, and that's a big
2091 threaded app if ever I saw one.
2092
2093- Automatic generation of suppression records; you no longer
2094 need to write them by hand. Use --gen-suppressions=yes.
2095
sewardj21511802003-07-22 17:47:42 +00002096- strcpy/memcpy/etc check their arguments for overlaps, when
2097 running with the Memcheck or Addrcheck skins.
2098
2099- malloc_usable_size() is now supported.
2100
2101- new client requests:
2102 - VALGRIND_COUNT_ERRORS, VALGRIND_COUNT_LEAKS:
2103 useful with regression testing
2104 - VALGRIND_NON_SIMD_CALL[0123]: for running arbitrary functions
2105 on real CPU (use with caution!)
2106
sewardj9d916ed2003-07-14 23:38:40 +00002107- The GDB attach mechanism is more flexible. Allow the GDB to
2108 be run to be specified by --gdb-path=/path/to/gdb, and specify
2109 which file descriptor V will read its input from with
2110 --input-fd=<number>.
2111
sewardj21511802003-07-22 17:47:42 +00002112- Cachegrind gives more accurate results (wasn't tracking instructions in
2113 malloc() and friends previously, is now).
2114
sewardj9d916ed2003-07-14 23:38:40 +00002115- Complete support for the MMX instruction set.
2116
2117- Partial support for the SSE and SSE2 instruction sets. Work for this
2118 is ongoing. About half the SSE/SSE2 instructions are done, so
2119 some SSE based programs may work. Currently you need to specify
2120 --skin=addrcheck. Basically not suitable for real use yet.
2121
2122- Significant speedups (10%-20%) for standard memory checking.
2123
2124- Fix assertion failure in pthread_once().
2125
2126- Fix this:
2127 valgrind: vg_intercept.c:598 (vgAllRoadsLeadToRome_select):
2128 Assertion `ms_end >= ms_now' failed.
2129
2130- Implement pthread_mutexattr_setpshared.
2131
2132- Understand Pentium 4 branch hints. Also implemented a couple more
2133 obscure x86 instructions.
2134
2135- Lots of other minor bug fixes.
2136
sewardj626fd892003-07-16 20:10:26 +00002137- We have a decent regression test system, for the first time.
2138 This doesn't help you directly, but it does make it a lot easier
2139 for us to track the quality of the system, especially across
2140 multiple linux distributions.
2141
2142 You can run the regression tests with 'make regtest' after 'make
2143 install' completes. On SuSE 8.2 and Red Hat 9 I get this:
2144
2145 == 84 tests, 0 stderr failures, 0 stdout failures ==
2146
2147 On Red Hat 8, I get this:
2148
2149 == 84 tests, 2 stderr failures, 1 stdout failure ==
2150 corecheck/tests/res_search (stdout)
2151 memcheck/tests/sigaltstack (stderr)
2152
2153 sigaltstack is probably harmless. res_search doesn't work
2154 on R H 8 even running natively, so I'm not too worried.
2155
2156 On Red Hat 7.3, a glibc-2.2.5 system, I get these harmless failures:
2157
2158 == 84 tests, 2 stderr failures, 1 stdout failure ==
2159 corecheck/tests/pth_atfork1 (stdout)
2160 corecheck/tests/pth_atfork1 (stderr)
2161 memcheck/tests/sigaltstack (stderr)
2162
2163 You need to run on a PII system, at least, since some tests
2164 contain P6-specific instructions, and the test machine needs
2165 access to the internet so that corecheck/tests/res_search
2166 (a test that the DNS resolver works) can function.
2167
sewardj9d916ed2003-07-14 23:38:40 +00002168As ever, thanks for the vast amount of feedback :) and bug reports :(
2169We may not answer all messages, but we do at least look at all of
2170them, and tend to fix the most frequently reported bugs.
2171
2172
2173
sewardj37918822003-05-05 01:05:09 +00002174Version 1.9.6 (7 May 2003 or thereabouts)
2175~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2176
2177Major changes in 1.9.6:
2178
2179- Improved threading support for glibc >= 2.3.2 (SuSE 8.2,
2180 RedHat 9, to name but two ...) It turned out that 1.9.5
2181 had problems with threading support on glibc >= 2.3.2,
2182 usually manifested by threaded programs deadlocking in system calls,
2183 or running unbelievably slowly. Hopefully these are fixed now. 1.9.6
2184 is the first valgrind which gives reasonable support for
2185 glibc-2.3.2. Also fixed a 2.3.2 problem with pthread_atfork().
2186
2187- Majorly expanded FAQ.txt. We've added workarounds for all
2188 common problems for which a workaround is known.
2189
2190Minor changes in 1.9.6:
2191
2192- Fix identification of the main thread's stack. Incorrect
2193 identification of it was causing some on-stack addresses to not get
2194 identified as such. This only affected the usefulness of some error
2195 messages; the correctness of the checks made is unchanged.
2196
2197- Support for kernels >= 2.5.68.
2198
2199- Dummy implementations of __libc_current_sigrtmin,
2200 __libc_current_sigrtmax and __libc_allocate_rtsig, hopefully
2201 good enough to keep alive programs which previously died for lack of
2202 them.
2203
2204- Fix bug in the VALGRIND_DISCARD_TRANSLATIONS client request.
2205
sewardj3d47b792003-05-05 22:15:35 +00002206- Fix bug in the DWARF2 debug line info loader, when instructions
2207 following each other have source lines far from each other
2208 (e.g. with inlined functions).
2209
sewardj37918822003-05-05 01:05:09 +00002210- Debug info reading: read symbols from both "symtab" and "dynsym"
2211 sections, rather than merely from the one that comes last in the
2212 file.
2213
2214- New syscall support: prctl(), creat(), lookup_dcookie().
2215
2216- When checking calls to accept(), recvfrom(), getsocketopt(),
2217 don't complain if buffer values are NULL.
2218
2219- Try and avoid assertion failures in
2220 mash_LD_PRELOAD_and_LD_LIBRARY_PATH.
2221
2222- Minor bug fixes in cg_annotate.
2223
2224
2225
sewardj90238792003-05-05 00:23:42 +00002226Version 1.9.5 (7 April 2003)
2227~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2228
2229It occurs to me that it would be helpful for valgrind users to record
2230in the source distribution the changes in each release. So I now
2231attempt to mend my errant ways :-) Changes in this and future releases
2232will be documented in the NEWS file in the source distribution.
2233
2234Major changes in 1.9.5:
2235
2236- (Critical bug fix): Fix a bug in the FPU simulation. This was
2237 causing some floating point conditional tests not to work right.
2238 Several people reported this. If you had floating point code which
2239 didn't work right on 1.9.1 to 1.9.4, it's worth trying 1.9.5.
2240
2241- Partial support for Red Hat 9. RH9 uses the new Native Posix
2242 Threads Library (NPTL), instead of the older LinuxThreads.
2243 This potentially causes problems with V which will take some
2244 time to correct. In the meantime we have partially worked around
2245 this, and so 1.9.5 works on RH9. Threaded programs still work,
2246 but they may deadlock, because some system calls (accept, read,
2247 write, etc) which should be nonblocking, in fact do block. This
2248 is a known bug which we are looking into.
2249
2250 If you can, your best bet (unfortunately) is to avoid using
2251 1.9.5 on a Red Hat 9 system, or on any NPTL-based distribution.
2252 If your glibc is 2.3.1 or earlier, you're almost certainly OK.
2253
2254Minor changes in 1.9.5:
2255
2256- Added some #errors to valgrind.h to ensure people don't include
2257 it accidentally in their sources. This is a change from 1.0.X
2258 which was never properly documented. The right thing to include
2259 is now memcheck.h. Some people reported problems and strange
2260 behaviour when (incorrectly) including valgrind.h in code with
2261 1.9.1 -- 1.9.4. This is no longer possible.
2262
2263- Add some __extension__ bits and pieces so that gcc configured
2264 for valgrind-checking compiles even with -Werror. If you
2265 don't understand this, ignore it. Of interest to gcc developers
2266 only.
2267
2268- Removed a pointless check which caused problems interworking
2269 with Clearcase. V would complain about shared objects whose
2270 names did not end ".so", and refuse to run. This is now fixed.
2271 In fact it was fixed in 1.9.4 but not documented.
2272
2273- Fixed a bug causing an assertion failure of "waiters == 1"
2274 somewhere in vg_scheduler.c, when running large threaded apps,
2275 notably MySQL.
2276
2277- Add support for the munlock system call (124).
2278
2279Some comments about future releases:
2280
22811.9.5 is, we hope, the most stable Valgrind so far. It pretty much
2282supersedes the 1.0.X branch. If you are a valgrind packager, please
2283consider making 1.9.5 available to your users. You can regard the
22841.0.X branch as obsolete: 1.9.5 is stable and vastly superior. There
2285are no plans at all for further releases of the 1.0.X branch.
2286
2287If you want a leading-edge valgrind, consider building the cvs head
2288(from SourceForge), or getting a snapshot of it. Current cool stuff
2289going in includes MMX support (done); SSE/SSE2 support (in progress),
2290a significant (10-20%) performance improvement (done), and the usual
2291large collection of minor changes. Hopefully we will be able to
2292improve our NPTL support, but no promises.
2293