|
Server : Apache/2.2.17 (Unix) mod_ssl/2.2.17 OpenSSL/0.9.8e-fips-rhel5 DAV/2 PHP/5.2.17 System : Linux localhost 2.6.18-419.el5 #1 SMP Fri Feb 24 22:47:42 UTC 2017 x86_64 User : nobody ( 99) PHP Version : 5.2.17 Disable Function : NONE Directory : /proc/21585/root/usr/share/doc/valgrind-3.5.0/html/ |
Upload File : |
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=ISO-8859-1">
<title>2. NEWS</title>
<link rel="stylesheet" href="vg_basic.css" type="text/css">
<meta name="generator" content="DocBook XSL Stylesheets V1.69.1">
<link rel="start" href="index.html" title="Valgrind Documentation">
<link rel="up" href="dist.html" title="Valgrind Distribution Documents">
<link rel="prev" href="dist.authors.html" title="1. AUTHORS">
<link rel="next" href="dist.readme.html" title="3. README">
</head>
<body bgcolor="white" text="black" link="#0000FF" vlink="#840084" alink="#0000FF">
<div><table class="nav" width="100%" cellspacing="3" cellpadding="3" border="0" summary="Navigation header"><tr>
<td width="22px" align="center" valign="middle"><a accesskey="p" href="dist.authors.html"><img src="images/prev.png" width="18" height="21" border="0" alt="Prev"></a></td>
<td width="25px" align="center" valign="middle"><a accesskey="u" href="dist.html"><img src="images/up.png" width="21" height="18" border="0" alt="Up"></a></td>
<td width="31px" align="center" valign="middle"><a accesskey="h" href="index.html"><img src="images/home.png" width="27" height="20" border="0" alt="Up"></a></td>
<th align="center" valign="middle">Valgrind Distribution Documents</th>
<td width="22px" align="center" valign="middle"><a accesskey="n" href="dist.readme.html"><img src="images/next.png" width="18" height="21" border="0" alt="Next"></a></td>
</tr></table></div>
<div class="chapter" lang="en">
<div class="titlepage"><div><div><h2 class="title">
<a name="dist.news"></a>2. NEWS</h2></div></div></div>
<div class="literallayout"><p><br>
<br>
Release 3.5.0 (19 August 2009)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
3.5.0 is a feature release with many significant improvements and the<br>
usual collection of bug fixes. The main improvement is that Valgrind<br>
now works on Mac OS X.<br>
<br>
This release supports X86/Linux, AMD64/Linux, PPC32/Linux, PPC64/Linux<br>
and X86/Darwin. Support for recent distros and toolchain components<br>
(glibc 2.10, gcc 4.5) has been added.<br>
<br>
-------------------------<br>
<br>
Here is a short summary of the changes. Details are shown further<br>
down:<br>
<br>
* Support for Mac OS X (10.5.x).<br>
<br>
* Improvements and simplifications to Memcheck's leak checker.<br>
<br>
* Clarification and simplifications in various aspects of Valgrind's<br>
text output.<br>
<br>
* XML output for Helgrind and Ptrcheck.<br>
<br>
* Performance and stability improvements for Helgrind and DRD.<br>
<br>
* Genuinely atomic support for x86/amd64/ppc atomic instructions.<br>
<br>
* A new experimental tool, BBV, useful for computer architecture<br>
research.<br>
<br>
* Improved Wine support, including ability to read Windows PDB<br>
debuginfo.<br>
<br>
-------------------------<br>
<br>
Here are details of the above changes, followed by descriptions of<br>
many other minor changes, and a list of fixed bugs.<br>
<br>
<br>
* Valgrind now runs on Mac OS X. (Note that Mac OS X is sometimes<br>
called "Darwin" because that is the name of the OS core, which is the<br>
level that Valgrind works at.)<br>
<br>
Supported systems:<br>
<br>
- It requires OS 10.5.x (Leopard). Porting to 10.4.x is not planned<br>
because it would require work and 10.4 is only becoming less common.<br>
<br>
- 32-bit programs on x86 and AMD64 (a.k.a x86-64) machines are supported<br>
fairly well. For 10.5.x, 32-bit programs are the default even on<br>
64-bit machines, so it handles most current programs.<br>
<br>
- 64-bit programs on x86 and AMD64 (a.k.a x86-64) machines are not<br>
officially supported, but simple programs at least will probably work.<br>
However, start-up is slow.<br>
<br>
- PowerPC machines are not supported.<br>
<br>
Things that don't work:<br>
<br>
- The Ptrcheck tool.<br>
<br>
- Objective-C garbage collection.<br>
<br>
- --db-attach=yes.<br>
<br>
- If you have Rogue Amoeba's "Instant Hijack" program installed,<br>
Valgrind will fail with a SIGTRAP at start-up. See<br>
https://bugs.kde.org/show_bug.cgi?id=193917 for details and a<br>
simple work-around.<br>
<br>
Usage notes:<br>
<br>
- You will likely find --dsymutil=yes a useful option, as error<br>
messages may be imprecise without it.<br>
<br>
- Mac OS X support is new and therefore will be less robust than the<br>
Linux support. Please report any bugs you find.<br>
<br>
- Threaded programs may run more slowly than on Linux.<br>
<br>
Many thanks to Greg Parker for developing this port over several years.<br>
<br>
<br>
* Memcheck's leak checker has been improved. <br>
<br>
- The results for --leak-check=summary now match the summary results<br>
for --leak-check=full. Previously they could differ because<br>
--leak-check=summary counted "indirectly lost" blocks and<br>
"suppressed" blocks as "definitely lost".<br>
<br>
- Blocks that are only reachable via at least one interior-pointer,<br>
but are directly pointed to by a start-pointer, were previously<br>
marked as "still reachable". They are now correctly marked as<br>
"possibly lost".<br>
<br>
- The default value for the --leak-resolution option has been<br>
changed from "low" to "high". In general, this means that more<br>
leak reports will be produced, but each leak report will describe<br>
fewer leaked blocks.<br>
<br>
- With --leak-check=full, "definitely lost" and "possibly lost"<br>
leaks are now considered as proper errors, ie. they are counted<br>
for the "ERROR SUMMARY" and affect the behaviour of<br>
--error-exitcode. These leaks are not counted as errors if<br>
--leak-check=summary is specified, however.<br>
<br>
- Documentation for the leak checker has been improved.<br>
<br>
<br>
* Various aspects of Valgrind's text output have changed.<br>
<br>
- Valgrind's start-up message has changed. It is shorter but also<br>
includes the command being run, which makes it easier to use<br>
--trace-children=yes. An example:<br>
<br>
- Valgrind's shut-down messages have also changed. This is most<br>
noticeable with Memcheck, where the leak summary now occurs before<br>
the error summary. This change was necessary to allow leaks to be<br>
counted as proper errors (see the description of the leak checker<br>
changes above for more details). This was also necessary to fix a<br>
longstanding bug in which uses of suppressions against leaks were<br>
not "counted", leading to difficulties in maintaining suppression<br>
files (XXXX bug number).<br>
<br>
- Behavior of -v has changed. In previous versions, -v printed out<br>
a mixture of marginally-user-useful information, and tool/core<br>
statistics. The statistics printing has now been moved to its own<br>
flag, --stats=yes. This means -v is less verbose and more likely<br>
to convey useful end-user information.<br>
<br>
- The format of some (non-XML) stack trace entries has changed a<br>
little. Previously there were six possible forms:<br>
<br>
0x80483BF: really (a.c:20)<br>
0x80483BF: really (in /foo/a.out)<br>
0x80483BF: really<br>
0x80483BF: (within /foo/a.out)<br>
0x80483BF: ??? (a.c:20)<br>
0x80483BF: ???<br>
<br>
The third and fourth of these forms have been made more consistent<br>
with the others. The six possible forms are now:<br>
<br>
0x80483BF: really (a.c:20)<br>
0x80483BF: really (in /foo/a.out)<br>
0x80483BF: really (in ???)<br>
0x80483BF: ??? (in /foo/a.out)<br>
0x80483BF: ??? (a.c:20)<br>
0x80483BF: ???<br>
<br>
Stack traces produced when --xml=yes is specified are different<br>
and unchanged.<br>
<br>
<br>
* Helgrind and Ptrcheck now support XML output, so they can be used<br>
from GUI tools. Also, the XML output mechanism has been<br>
overhauled.<br>
<br>
- The XML format has been overhauled and generalised, so it is more<br>
suitable for error reporting tools in general. The Memcheck<br>
specific aspects of it have been removed. The new format, which<br>
is an evolution of the old format, is described in<br>
docs/internals/xml-output-protocol4.txt.<br>
<br>
- Memcheck has been updated to use the new format.<br>
<br>
- Helgrind and Ptrcheck are now able to emit output in this format.<br>
<br>
- The XML output mechanism has been overhauled. XML is now output<br>
to its own file descriptor, which means that:<br>
<br>
* Valgrind can output text and XML independently.<br>
<br>
* The longstanding problem of XML output being corrupted by <br>
unexpected un-tagged text messages is solved.<br>
<br>
As before, the destination for text output is specified using<br>
--log-file=, --log-fd= or --log-socket=.<br>
<br>
As before, XML output for a tool is enabled using --xml=yes.<br>
<br>
Because there's a new XML output channel, the XML output<br>
destination is now specified by --xml-file=, --xml-fd= or<br>
--xml-socket=.<br>
<br>
Initial feedback has shown this causes some confusion. To<br>
clarify, the two envisaged usage scenarios are:<br>
<br>
(1) Normal text output. In this case, do not specify --xml=yes<br>
nor any of --xml-file=, --xml-fd= or --xml-socket=.<br>
<br>
(2) XML output. In this case, specify --xml=yes, and one of<br>
--xml-file=, --xml-fd= or --xml-socket= to select the XML<br>
destination, one of --log-file=, --log-fd= or --log-socket=<br>
to select the destination for any remaining text messages,<br>
and, importantly, -q.<br>
<br>
-q makes Valgrind completely silent on the text channel,<br>
except in the case of critical failures, such as Valgrind<br>
itself segfaulting, or failing to read debugging information.<br>
Hence, in this scenario, it suffices to check whether or not<br>
any output appeared on the text channel. If yes, then it is<br>
likely to be a critical error which should be brought to the<br>
attention of the user. If no (the text channel produced no<br>
output) then it can be assumed that the run was successful.<br>
<br>
This allows GUIs to make the critical distinction they need to<br>
make (did the run fail or not?) without having to search or<br>
filter the text output channel in any way.<br>
<br>
It is also recommended to use --child-silent-after-fork=yes in<br>
scenario (2).<br>
<br>
<br>
* Improvements and changes in Helgrind:<br>
<br>
- XML output, as described above<br>
<br>
- Checks for consistent association between pthread condition<br>
variables and their associated mutexes are now performed.<br>
<br>
- pthread_spinlock functions are supported.<br>
<br>
- Modest performance improvements.<br>
<br>
- Initial (skeletal) support for describing the behaviour of<br>
non-POSIX synchronisation objects through ThreadSanitizer<br>
compatible ANNOTATE_* macros.<br>
<br>
- More controllable tradeoffs between performance and the level of<br>
detail of "previous" accesses in a race. There are now three<br>
settings:<br>
<br>
* --history-level=full. This is the default, and was also the<br>
default in 3.4.x. It shows both stacks involved in a race, but<br>
requires a lot of memory and can be very slow in programs that<br>
do many inter-thread synchronisation events.<br>
<br>
* --history-level=none. This only shows the later stack involved<br>
in a race. This can be much faster than --history-level=full,<br>
but makes it much more difficult to find the other access<br>
involved in the race.<br>
<br>
The new intermediate setting is<br>
<br>
* --history-level=approx<br>
<br>
For the earlier (other) access, two stacks are presented. The<br>
earlier access is guaranteed to be somewhere in between the two<br>
program points denoted by those stacks. This is not as useful<br>
as showing the exact stack for the previous access (as per<br>
--history-level=full), but it is better than nothing, and it's<br>
almost as fast as --history-level=none.<br>
<br>
<br>
* New features and improvements in DRD:<br>
<br>
- The error messages printed by DRD are now easier to interpret.<br>
Instead of using two different numbers to identify each thread<br>
(Valgrind thread ID and DRD thread ID), DRD does now identify<br>
threads via a single number (the DRD thread ID). Furthermore<br>
"first observed at" information is now printed for all error<br>
messages related to synchronization objects.<br>
<br>
- Added support for named semaphores (sem_open() and sem_close()).<br>
<br>
- Race conditions between pthread_barrier_wait() and<br>
pthread_barrier_destroy() calls are now reported.<br>
<br>
- Added support for custom allocators through the macros<br>
VALGRIND_MALLOCLIKE_BLOCK() VALGRIND_FREELIKE_BLOCK() (defined in<br>
in <valgrind/valgrind.h>). An alternative for these two macros is<br>
the new client request VG_USERREQ__DRD_CLEAN_MEMORY (defined in<br>
<valgrind/drd.h>).<br>
<br>
- Added support for annotating non-POSIX synchronization objects<br>
through several new ANNOTATE_*() macros.<br>
<br>
- OpenMP: added support for the OpenMP runtime (libgomp) included<br>
with gcc versions 4.3.0 and 4.4.0.<br>
<br>
- Faster operation.<br>
<br>
- Added two new command-line options (--first-race-only and<br>
--segment-merging-interval).<br>
<br>
<br>
* Genuinely atomic support for x86/amd64/ppc atomic instructions<br>
<br>
Valgrind will now preserve (memory-access) atomicity of LOCK-<br>
prefixed x86/amd64 instructions, and any others implying a global<br>
bus lock. Ditto for PowerPC l{w,d}arx/st{w,d}cx. instructions.<br>
<br>
This means that Valgrinded processes will "play nicely" in<br>
situations where communication with other processes, or the kernel,<br>
is done through shared memory and coordinated with such atomic<br>
instructions. Prior to this change, such arrangements usually<br>
resulted in hangs, races or other synchronisation failures, because<br>
Valgrind did not honour atomicity of such instructions.<br>
<br>
<br>
* A new experimental tool, BBV, has been added. BBV generates basic<br>
block vectors for use with the SimPoint analysis tool, which allows<br>
a program's overall behaviour to be approximated by running only a<br>
fraction of it. This is useful for computer architecture<br>
researchers. You can run BBV by specifying --tool=exp-bbv (the<br>
"exp-" prefix is short for "experimental"). BBV was written by<br>
Vince Weaver.<br>
<br>
<br>
* Modestly improved support for running Windows applications under<br>
Wine. In particular, initial support for reading Windows .PDB debug<br>
information has been added.<br>
<br>
<br>
* A new Memcheck client request VALGRIND_COUNT_LEAK_BLOCKS has been<br>
added. It is similar to VALGRIND_COUNT_LEAKS but counts blocks<br>
instead of bytes.<br>
<br>
<br>
* The Valgrind client requests VALGRIND_PRINTF and<br>
VALGRIND_PRINTF_BACKTRACE have been changed slightly. Previously,<br>
the string was always printed immediately on its own line. Now, the<br>
string will be added to a buffer but not printed until a newline is<br>
encountered, or other Valgrind output is printed (note that for<br>
VALGRIND_PRINTF_BACKTRACE, the back-trace itself is considered<br>
"other Valgrind output"). This allows you to use multiple<br>
VALGRIND_PRINTF calls to build up a single output line, and also to<br>
print multiple output lines with a single request (by embedding<br>
multiple newlines in the string).<br>
<br>
<br>
* The graphs drawn by Massif's ms_print program have changed slightly:<br>
<br>
- The half-height chars '.' and ',' are no longer drawn, because<br>
they are confusing. The --y option can be used if the default<br>
y-resolution is not high enough.<br>
<br>
- Horizontal lines are now drawn after the top of a snapshot if<br>
there is a gap until the next snapshot. This makes it clear that<br>
the memory usage has not dropped to zero between snapshots.<br>
<br>
<br>
* Something that happened in 3.4.0, but wasn't clearly announced: the<br>
option --read-var-info=yes can be used by some tools (Memcheck,<br>
Helgrind and DRD). When enabled, it causes Valgrind to read DWARF3<br>
variable type and location information. This makes those tools<br>
start up more slowly and increases memory consumption, but<br>
descriptions of data addresses in error messages become more<br>
detailed.<br>
<br>
<br>
* exp-Omega, an experimental instantaneous leak-detecting tool, was<br>
disabled in 3.4.0 due to a lack of interest and maintenance,<br>
although the source code was still in the distribution. The source<br>
code has now been removed from the distribution. For anyone<br>
interested, the removal occurred in SVN revision r10247.<br>
<br>
<br>
* Some changes have been made to the build system.<br>
<br>
- VEX/ is now integrated properly into the build system. This means<br>
that dependency tracking within VEX/ now works properly, "make<br>
install" will work without requiring "make" before it, and<br>
parallel builds (ie. 'make -j') now work (previously a<br>
.NOTPARALLEL directive was used to serialize builds, ie. 'make -j'<br>
was effectively ignored).<br>
<br>
- The --with-vex configure option has been removed. It was of<br>
little use and removing it simplified the build system.<br>
<br>
- The location of some install files has changed. This should not<br>
affect most users. Those who might be affected:<br>
<br>
* For people who use Valgrind with MPI programs, the installed<br>
libmpiwrap.so library has moved from<br>
$(INSTALL)/<platform>/libmpiwrap.so to<br>
$(INSTALL)/libmpiwrap-<platform>.so.<br>
<br>
* For people who distribute standalone Valgrind tools, the<br>
installed libraries such as $(INSTALL)/<platform>/libcoregrind.a<br>
have moved to $(INSTALL)/libcoregrind-<platform>.a.<br>
<br>
These changes simplify the build system.<br>
<br>
- Previously, all the distributed suppression (*.supp) files were<br>
installed. Now, only default.supp is installed. This should not<br>
affect users as the other installed suppression files were not<br>
read; the fact that they were installed was a mistake.<br>
<br>
<br>
* KNOWN LIMITATIONS:<br>
<br>
- Memcheck is unusable with the Intel compiler suite version 11.1,<br>
when it generates code for SSE2-and-above capable targets. This<br>
is because of icc's use of highly optimised inlined strlen<br>
implementations. It causes Memcheck to report huge numbers of<br>
false errors even in simple programs. Helgrind and DRD may also<br>
have problems.<br>
<br>
Versions 11.0 and earlier may be OK, but this has not been<br>
properly tested.<br>
<br>
<br>
The following bugs have been fixed or resolved. Note that "n-i-bz"<br>
stands for "not in bugzilla" -- that is, a bug that was reported to us<br>
but never got a bugzilla entry. We encourage you to file bugs in<br>
bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than<br>
mailing the developers (or mailing lists) directly -- bugs that are<br>
not entered into bugzilla tend to get forgotten about or ignored.<br>
<br>
To see details of a given bug, visit<br>
https://bugs.kde.org/show_bug.cgi?id=XXXXXX<br>
where XXXXXX is the bug number as listed below.<br>
<br>
84303 How about a LockCheck tool? <br>
91633 dereference of null ptr in vgPlain_st_basetype <br>
97452 Valgrind doesn't report any pthreads problems <br>
100628 leak-check gets assertion failure when using <br>
VALGRIND_MALLOCLIKE_BLOCK on malloc()ed memory <br>
108528 NPTL pthread cleanup handlers not called <br>
110126 Valgrind 2.4.1 configure.in tramples CFLAGS <br>
110128 mallinfo is not implemented... <br>
110770 VEX: Generated files not always updated when making valgrind<br>
111102 Memcheck: problems with large (memory footprint) applications <br>
115673 Vex's decoder should never assert <br>
117564 False positive: Syscall param clone(child_tidptr) contains<br>
uninitialised byte(s) <br>
119404 executing ssh from inside valgrind fails <br>
133679 Callgrind does not write path names to sources with dwarf debug<br>
info<br>
135847 configure.in problem with non gnu compilers (and possible fix) <br>
136154 threads.c:273 (vgCallgrind_post_signal): Assertion<br>
'*(vgCallgrind_current_fn_stack.top) == 0' failed. <br>
136230 memcheck reports "possibly lost", should be "still reachable" <br>
137073 NULL arg to MALLOCLIKE_BLOCK causes crash <br>
137904 Valgrind reports a memory leak when using POSIX threads,<br>
while it shouldn't <br>
139076 valgrind VT_GETSTATE error <br>
142228 complaint of elf_dynamic_do_rela in trivial usage <br>
145347 spurious warning with USBDEVFS_REAPURB <br>
148441 (wine) can't find memory leak in Wine, win32 binary <br>
executable file.<br>
148742 Leak-check fails assert on exit <br>
149878 add (proper) check for calloc integer overflow <br>
150606 Call graph is broken when using callgrind control <br>
152393 leak errors produce an exit code of 0. I need some way to <br>
cause leak errors to result in a nonzero exit code. <br>
157154 documentation (leak-resolution doc speaks about num-callers<br>
def=4) + what is a loss record<br>
159501 incorrect handling of ALSA ioctls <br>
162020 Valgrinding an empty/zero-byte file crashes valgrind <br>
162482 ppc: Valgrind crashes while reading stabs information <br>
162718 x86: avoid segment selector 0 in sys_set_thread_area() <br>
163253 (wine) canonicaliseSymtab forgot some fields in DiSym <br>
163560 VEX/test_main.c is missing from valgrind-3.3.1 <br>
164353 malloc_usable_size() doesn't return a usable size <br>
165468 Inconsistent formatting in memcheck manual -- please fix <br>
169505 main.c:286 (endOfInstr):<br>
Assertion 'ii->cost_offset == *cost_offset' failed <br>
177206 Generate default.supp during compile instead of configure<br>
177209 Configure valt_load_address based on arch+os <br>
177305 eventfd / syscall 323 patch lost<br>
179731 Tests fail to build because of inlining of non-local asm labels<br>
181394 helgrind: libhb_core.c:3762 (msm_write): Assertion <br>
'ordxx == POrd_EQ || ordxx == POrd_LT' failed. <br>
181594 Bogus warning for empty text segment <br>
181707 dwarf doesn't require enumerations to have name <br>
185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64 <br>
185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree):<br>
Assertion '!already_present' failed.<br>
185359 exp-ptrcheck: unhandled syscall getresuid()<br>
185794 "WARNING: unhandled syscall: 285" (fallocate) on x86_64<br>
185816 Valgrind is unable to handle debug info for files with split<br>
debug info that are prelinked afterwards <br>
185980 [darwin] unhandled syscall: sem_open <br>
186238 bbToIR_AMD64: disInstr miscalculated next %rip<br>
186507 exp-ptrcheck unhandled syscalls prctl, etc. <br>
186790 Suppression pattern used for leaks are not reported <br>
186796 Symbols with length>200 in suppression files are ignored <br>
187048 drd: mutex PTHREAD_PROCESS_SHARED attribute missinterpretation<br>
187416 exp-ptrcheck: support for __NR_{setregid,setreuid,setresuid}<br>
188038 helgrind: hg_main.c:926: mk_SHVAL_fail: the 'impossible' happened<br>
188046 bashisms in the configure script<br>
188127 amd64->IR: unhandled instruction bytes: 0xF0 0xF 0xB0 0xA<br>
188161 memcheck: --track-origins=yes asserts "mc_machine.c:672<br>
(get_otrack_shadow_offset_wrk): the 'impossible' happened."<br>
188248 helgrind: pthread_cleanup_push, pthread_rwlock_unlock, <br>
assertion fail "!lock->heldBy" <br>
188427 Add support for epoll_create1 (with patch) <br>
188530 Support for SIOCGSTAMPNS<br>
188560 Include valgrind.spec in the tarball<br>
188572 Valgrind on Mac should suppress setenv() mem leak <br>
189054 Valgrind fails to build because of duplicate non-local asm labels <br>
189737 vex amd64->IR: unhandled instruction bytes: 0xAC<br>
189762 epoll_create syscall not handled (--tool=exp-ptrcheck)<br>
189763 drd assertion failure: s_threadinfo[tid].is_recording <br>
190219 unhandled syscall: 328 (x86-linux)<br>
190391 dup of 181394; see above<br>
190429 Valgrind reports lots of errors in ld.so with x86_64 2.9.90 glibc <br>
190820 No debug information on powerpc-linux<br>
190820 No debug information on powerpc-linux <br>
191095 PATCH: Improve usbdevfs ioctl handling <br>
191182 memcheck: VALGRIND_LEAK_CHECK quadratic when big nr of chunks<br>
or big nr of errors<br>
191189 --xml=yes should obey --gen-suppressions=all <br>
191192 syslog() needs a suppression on macosx <br>
191271 DARWIN: WARNING: unhandled syscall: 33554697 a.k.a.: 265 <br>
191761 getrlimit on MacOSX <br>
191992 multiple --fn-skip only works sometimes; dependent on order <br>
192634 V. reports "aspacem sync_check_mapping_callback: <br>
segment mismatch" on Darwin<br>
192954 __extension__ missing on 2 client requests <br>
194429 Crash at start-up with glibc-2.10.1 and linux-2.6.29 <br>
194474 "INSTALL" file has different build instructions than "README"<br>
194671 Unhandled syscall (sem_wait?) from mac valgrind <br>
195069 memcheck: reports leak (memory still reachable) for <br>
printf("%d', x) <br>
195169 drd: (vgDrd_barrier_post_wait):<br>
Assertion 'r->sg[p->post_iteration]' failed. <br>
195268 valgrind --log-file doesn't accept ~/...<br>
195838 VEX abort: LibVEX_N_SPILL_BYTES too small for CPUID boilerplate <br>
195860 WARNING: unhandled syscall: unix:223 <br>
196528 need a error suppression for pthread_rwlock_init under os x? <br>
197227 Support aio_* syscalls on Darwin<br>
197456 valgrind should reject --suppressions=(directory) <br>
197512 DWARF2 CFI reader: unhandled CFI instruction 0:10 <br>
197591 unhandled syscall 27 (mincore) <br>
197793 Merge DCAS branch to the trunk == 85756, 142103<br>
197794 Avoid duplicate filenames in Vex <br>
197898 make check fails on current SVN <br>
197901 make check fails also under exp-ptrcheck in current SVN <br>
197929 Make --leak-resolution=high the default <br>
197930 Reduce spacing between leak reports <br>
197933 Print command line of client at start-up, and shorten preamble <br>
197966 unhandled syscall 205 (x86-linux, --tool=exp-ptrcheck)<br>
198395 add BBV to the distribution as an experimental tool <br>
198624 Missing syscalls on Darwin: 82, 167, 281, 347 <br>
198649 callgrind_annotate doesn't cumulate counters <br>
199338 callgrind_annotate sorting/thresholds are broken for all but Ir <br>
199977 Valgrind complains about an unrecognized instruction in the<br>
atomic_incs test program<br>
200029 valgrind isn't able to read Fedora 12 debuginfo <br>
200760 darwin unhandled syscall: unix:284 <br>
200827 DRD doesn't work on Mac OS X <br>
200990 VG_(read_millisecond_timer)() does not work correctly <br>
201016 Valgrind does not support pthread_kill() on Mac OS <br>
201169 Document --read-var-info<br>
201323 Pre-3.5.0 performance sanity checking <br>
201384 Review user manual for the 3.5.0 release <br>
201585 mfpvr not implemented on ppc <br>
201708 tests failing because x86 direction flag is left set <br>
201757 Valgrind doesn't handle any recent sys_futex additions <br>
204377 64-bit valgrind can not start a shell script<br>
(with #!/path/to/shell) if the shell is a 32-bit executable<br>
n-i-bz drd: fixed assertion failure triggered by mutex reinitialization.<br>
n-i-bz drd: fixed a bug that caused incorrect messages to be printed<br>
about memory allocation events with memory access tracing enabled<br>
n-i-bz drd: fixed a memory leak triggered by vector clock deallocation<br>
<br>
(3.5.0: 20 Aug 2009, vex r1913, valgrind r10846).<br>
<br>
<br>
<br>
Release 3.4.1 (28 February 2009)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
3.4.1 is a bug-fix release that fixes some regressions and assertion<br>
failures in debug info reading in 3.4.0, most notably incorrect stack<br>
traces on amd64-linux on older (glibc-2.3 based) systems. Various<br>
other debug info problems are also fixed. A number of bugs in the<br>
exp-ptrcheck tool introduced in 3.4.0 have been fixed.<br>
<br>
In view of the fact that 3.4.0 contains user-visible regressions<br>
relative to 3.3.x, upgrading to 3.4.1 is recommended. Packagers are<br>
encouraged to ship 3.4.1 in preference to 3.4.0.<br>
<br>
The fixed bugs are as follows. Note that "n-i-bz" stands for "not in<br>
bugzilla" -- that is, a bug that was reported to us but never got a<br>
bugzilla entry. We encourage you to file bugs in bugzilla<br>
(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the<br>
developers (or mailing lists) directly -- bugs that are not entered<br>
into bugzilla tend to get forgotten about or ignored.<br>
<br>
n-i-bz Fix various bugs reading icc-11 generated debug info<br>
n-i-bz Fix various bugs reading gcc-4.4 generated debug info<br>
n-i-bz Preliminary support for glibc-2.10 / Fedora 11<br>
n-i-bz Cachegrind and Callgrind: handle non-power-of-two cache sizes,<br>
so as to support (eg) 24k Atom D1 and Core2 with 3/6/12MB L2.<br>
179618 exp-ptrcheck crashed / exit prematurely<br>
179624 helgrind: false positive races with pthread_create and<br>
recv/open/close/read<br>
134207 pkg-config output contains @VG_PLATFORM@<br>
176926 floating point exception at valgrind startup with PPC 440EPX<br>
181594 Bogus warning for empty text segment<br>
173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant rex prefixes)<br>
181707 Dwarf3 doesn't require enumerations to have name<br>
185038 exp-ptrcheck: "unhandled syscall: 285" (fallocate) on x86_64<br>
185050 exp-ptrcheck: sg_main.c:727 (add_block_to_GlobalTree):<br>
Assertion '!already_present' failed.<br>
185359 exp-ptrcheck unhandled syscall getresuid()<br>
<br>
(3.4.1.RC1: 24 Feb 2008, vex r1884, valgrind r9253).<br>
(3.4.1: 28 Feb 2008, vex r1884, valgrind r9293).<br>
<br>
<br>
<br>
Release 3.4.0 (2 January 2009)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
3.4.0 is a feature release with many significant improvements and the<br>
usual collection of bug fixes. This release supports X86/Linux,<br>
AMD64/Linux, PPC32/Linux and PPC64/Linux. Support for recent distros<br>
(using gcc 4.4, glibc 2.8 and 2.9) has been added.<br>
<br>
3.4.0 brings some significant tool improvements. Memcheck can now<br>
report the origin of uninitialised values, the thread checkers<br>
Helgrind and DRD are much improved, and we have a new experimental<br>
tool, exp-Ptrcheck, which is able to detect overruns of stack and<br>
global arrays. In detail:<br>
<br>
* Memcheck is now able to track the origin of uninitialised values.<br>
When it reports an uninitialised value error, it will try to show<br>
the origin of the value, as either a heap or stack allocation.<br>
Origin tracking is expensive and so is not enabled by default. To<br>
use it, specify --track-origins=yes. Memcheck's speed will be<br>
essentially halved, and memory usage will be significantly<br>
increased. Nevertheless it can drastically reduce the effort<br>
required to identify the root cause of uninitialised value errors,<br>
and so is often a programmer productivity win, despite running more<br>
slowly.<br>
<br>
* A version (1.4.0) of the Valkyrie GUI, that works with Memcheck in<br>
3.4.0, will be released shortly.<br>
<br>
* Helgrind's race detection algorithm has been completely redesigned<br>
and reimplemented, to address usability and scalability concerns:<br>
<br>
- The new algorithm has a lower false-error rate: it is much less<br>
likely to report races that do not really exist.<br>
<br>
- Helgrind will display full call stacks for both accesses involved<br>
in a race. This makes it easier to identify the root causes of<br>
races.<br>
<br>
- Limitations on the size of program that can run have been removed.<br>
<br>
- Performance has been modestly improved, although that is very<br>
workload-dependent.<br>
<br>
- Direct support for Qt4 threading has been added.<br>
<br>
- pthread_barriers are now directly supported.<br>
<br>
- Helgrind works well on all supported Linux targets.<br>
<br>
* The DRD thread debugging tool has seen major improvements:<br>
<br>
- Greatly improved performance and significantly reduced memory<br>
usage.<br>
<br>
- Support for several major threading libraries (Boost.Thread, Qt4,<br>
glib, OpenMP) has been added.<br>
<br>
- Support for atomic instructions, POSIX semaphores, barriers and<br>
reader-writer locks has been added.<br>
<br>
- Works now on PowerPC CPUs too.<br>
<br>
- Added support for printing thread stack usage at thread exit time.<br>
<br>
- Added support for debugging lock contention.<br>
<br>
- Added a manual for Drd.<br>
<br>
* A new experimental tool, exp-Ptrcheck, has been added. Ptrcheck<br>
checks for misuses of pointers. In that sense it is a bit like<br>
Memcheck. However, Ptrcheck can do things Memcheck can't: it can<br>
detect overruns of stack and global arrays, it can detect<br>
arbitrarily far out-of-bounds accesses to heap blocks, and it can<br>
detect accesses heap blocks that have been freed a very long time<br>
ago (millions of blocks in the past).<br>
<br>
Ptrcheck currently works only on x86-linux and amd64-linux. To use<br>
it, use --tool=exp-ptrcheck. A simple manual is provided, as part<br>
of the main Valgrind documentation. As this is an experimental<br>
tool, we would be particularly interested in hearing about your<br>
experiences with it.<br>
<br>
* exp-Omega, an experimental instantaneous leak-detecting tool, is no<br>
longer built by default, although the code remains in the repository<br>
and the tarball. This is due to three factors: a perceived lack of<br>
users, a lack of maintenance, and concerns that it may not be<br>
possible to achieve reliable operation using the existing design.<br>
<br>
* As usual, support for the latest Linux distros and toolchain<br>
components has been added. It should work well on Fedora Core 10,<br>
OpenSUSE 11.1 and Ubuntu 8.10. gcc-4.4 (in its current pre-release<br>
state) is supported, as is glibc-2.9. The C++ demangler has been<br>
updated so as to work well with C++ compiled by even the most recent<br>
g++'s.<br>
<br>
* You can now use frame-level wildcards in suppressions. This was a<br>
frequently-requested enhancement. A line "..." in a suppression now<br>
matches zero or more frames. This makes it easier to write<br>
suppressions which are precise yet insensitive to changes in<br>
inlining behaviour.<br>
<br>
* 3.4.0 adds support on x86/amd64 for the SSSE3 instruction set.<br>
<br>
* Very basic support for IBM Power6 has been added (64-bit processes only).<br>
<br>
* Valgrind is now cross-compilable. For example, it is possible to<br>
cross compile Valgrind on an x86/amd64-linux host, so that it runs<br>
on a ppc32/64-linux target.<br>
<br>
* You can set the main thread's stack size at startup using the<br>
new --main-stacksize= flag (subject of course to ulimit settings).<br>
This is useful for running apps that need a lot of stack space.<br>
<br>
* The limitation that you can't use --trace-children=yes together<br>
with --db-attach=yes has been removed.<br>
<br>
* The following bugs have been fixed. Note that "n-i-bz" stands for<br>
"not in bugzilla" -- that is, a bug that was reported to us but<br>
never got a bugzilla entry. We encourage you to file bugs in<br>
bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than<br>
mailing the developers (or mailing lists) directly.<br>
<br>
n-i-bz Make return types for some client requests 64-bit clean<br>
n-i-bz glibc 2.9 support<br>
n-i-bz ignore unsafe .valgrindrc's (CVE-2008-4865)<br>
n-i-bz MPI_Init(0,0) is valid but libmpiwrap.c segfaults<br>
n-i-bz Building in an env without gdb gives bogus gdb attach<br>
92456 Tracing the origin of uninitialised memory<br>
106497 Valgrind does not demangle some C++ template symbols<br>
162222 ==106497<br>
151612 Suppression with "..." (frame-level wildcards in .supp files)<br>
156404 Unable to start oocalc under memcheck on openSUSE 10.3 (64-bit)<br>
159285 unhandled syscall:25 (stime, on x86-linux)<br>
159452 unhandled ioctl 0x8B01 on "valgrind iwconfig"<br>
160954 ppc build of valgrind crashes with illegal instruction (isel)<br>
160956 mallinfo implementation, w/ patch<br>
162092 Valgrind fails to start gnome-system-monitor<br>
162819 malloc_free_fill test doesn't pass on glibc2.8 x86<br>
163794 assertion failure with "--track-origins=yes"<br>
163933 sigcontext.err and .trapno must be set together<br>
163955 remove constraint !(--db-attach=yes && --trace-children=yes)<br>
164476 Missing kernel module loading system calls<br>
164669 SVN regression: mmap() drops posix file locks<br>
166581 Callgrind output corruption when program forks<br>
167288 Patch file for missing system calls on Cell BE<br>
168943 unsupported scas instruction pentium<br>
171645 Unrecognised instruction (MOVSD, non-binutils encoding)<br>
172417 x86->IR: 0x82 ...<br>
172563 amd64->IR: 0xD9 0xF5 - fprem1<br>
173099 .lds linker script generation error<br>
173177 [x86_64] syscalls: 125/126/179 (capget/capset/quotactl)<br>
173751 amd64->IR: 0x48 0xF 0x6F 0x45 (even more redundant prefixes)<br>
174532 == 173751<br>
174908 --log-file value not expanded correctly for core file<br>
175044 Add lookup_dcookie for amd64<br>
175150 x86->IR: 0xF2 0xF 0x11 0xC1 (movss non-binutils encoding)<br>
<br>
Developer-visible changes:<br>
<br>
* Valgrind's debug-info reading machinery has been majorly overhauled.<br>
It can now correctly establish the addresses for ELF data symbols,<br>
which is something that has never worked properly before now.<br>
<br>
Also, Valgrind can now read DWARF3 type and location information for<br>
stack and global variables. This makes it possible to use the<br>
framework to build tools that rely on knowing the type and locations<br>
of stack and global variables, for example exp-Ptrcheck.<br>
<br>
Reading of such information is disabled by default, because most<br>
tools don't need it, and because it is expensive in space and time.<br>
However, you can force Valgrind to read it, using the<br>
--read-var-info=yes flag. Memcheck, Helgrind and DRD are able to<br>
make use of such information, if present, to provide source-level<br>
descriptions of data addresses in the error messages they create.<br>
<br>
(3.4.0.RC1: 24 Dec 2008, vex r1878, valgrind r8882).<br>
(3.4.0: 3 Jan 2009, vex r1878, valgrind r8899).<br>
<br>
<br>
<br>
Release 3.3.1 (4 June 2008)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
3.3.1 fixes a bunch of bugs in 3.3.0, adds support for glibc-2.8 based<br>
systems (openSUSE 11, Fedora Core 9), improves the existing glibc-2.7<br>
support, and adds support for the SSSE3 (Core 2) instruction set.<br>
<br>
3.3.1 will likely be the last release that supports some very old<br>
systems. In particular, the next major release, 3.4.0, will drop<br>
support for the old LinuxThreads threading library, and for gcc<br>
versions prior to 3.0.<br>
<br>
The fixed bugs are as follows. Note that "n-i-bz" stands for "not in<br>
bugzilla" -- that is, a bug that was reported to us but never got a<br>
bugzilla entry. We encourage you to file bugs in bugzilla<br>
(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the<br>
developers (or mailing lists) directly -- bugs that are not entered<br>
into bugzilla tend to get forgotten about or ignored.<br>
<br>
n-i-bz Massif segfaults at exit<br>
n-i-bz Memcheck asserts on Altivec code<br>
n-i-bz fix sizeof bug in Helgrind<br>
n-i-bz check fd on sys_llseek<br>
n-i-bz update syscall lists to kernel 2.6.23.1<br>
n-i-bz support sys_sync_file_range<br>
n-i-bz handle sys_sysinfo, sys_getresuid, sys_getresgid on ppc64-linux<br>
n-i-bz intercept memcpy in 64-bit ld.so's<br>
n-i-bz Fix wrappers for sys_{futimesat,utimensat}<br>
n-i-bz Minor false-error avoidance fixes for Memcheck<br>
n-i-bz libmpiwrap.c: add a wrapper for MPI_Waitany<br>
n-i-bz helgrind support for glibc-2.8<br>
n-i-bz partial fix for mc_leakcheck.c:698 assert:<br>
'lc_shadows[i]->data + lc_shadows[i] ...<br>
n-i-bz Massif/Cachegrind output corruption when programs fork<br>
n-i-bz register allocator fix: handle spill stores correctly<br>
n-i-bz add support for PA6T PowerPC CPUs<br>
126389 vex x86->IR: 0xF 0xAE (FXRSTOR)<br>
158525 ==126389<br>
152818 vex x86->IR: 0xF3 0xAC (repz lodsb) <br>
153196 vex x86->IR: 0xF2 0xA6 (repnz cmpsb) <br>
155011 vex x86->IR: 0xCF (iret)<br>
155091 Warning [...] unhandled DW_OP_ opcode 0x23<br>
156960 ==155901<br>
155528 support Core2/SSSE3 insns on x86/amd64<br>
155929 ms_print fails on massif outputs containing long lines<br>
157665 valgrind fails on shmdt(0) after shmat to 0<br>
157748 support x86 PUSHFW/POPFW<br>
158212 helgrind: handle pthread_rwlock_try{rd,wr}lock.<br>
158425 sys_poll incorrectly emulated when RES==0<br>
158744 vex amd64->IR: 0xF0 0x41 0xF 0xC0 (xaddb)<br>
160907 Support for a couple of recent Linux syscalls<br>
161285 Patch -- support for eventfd() syscall<br>
161378 illegal opcode in debug libm (FUCOMPP)<br>
160136 ==161378<br>
161487 number of suppressions files is limited to 10<br>
162386 ms_print typo in milliseconds time unit for massif<br>
161036 exp-drd: client allocated memory was never freed<br>
162663 signalfd_wrapper fails on 64bit linux<br>
<br>
(3.3.1.RC1: 2 June 2008, vex r1854, valgrind r8169).<br>
(3.3.1: 4 June 2008, vex r1854, valgrind r8180).<br>
<br>
<br>
<br>
Release 3.3.0 (7 December 2007)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
3.3.0 is a feature release with many significant improvements and the<br>
usual collection of bug fixes. This release supports X86/Linux,<br>
AMD64/Linux, PPC32/Linux and PPC64/Linux. Support for recent distros<br>
(using gcc 4.3, glibc 2.6 and 2.7) has been added.<br>
<br>
The main excitement in 3.3.0 is new and improved tools. Helgrind<br>
works again, Massif has been completely overhauled and much improved,<br>
Cachegrind now does branch-misprediction profiling, and a new category<br>
of experimental tools has been created, containing two new tools:<br>
Omega and DRD. There are many other smaller improvements. In detail:<br>
<br>
- Helgrind has been completely overhauled and works for the first time<br>
since Valgrind 2.2.0. Supported functionality is: detection of<br>
misuses of the POSIX PThreads API, detection of potential deadlocks<br>
resulting from cyclic lock dependencies, and detection of data<br>
races. Compared to the 2.2.0 Helgrind, the race detection algorithm<br>
has some significant improvements aimed at reducing the false error<br>
rate. Handling of various kinds of corner cases has been improved.<br>
Efforts have been made to make the error messages easier to<br>
understand. Extensive documentation is provided.<br>
<br>
- Massif has been completely overhauled. Instead of measuring<br>
space-time usage -- which wasn't always useful and many people found<br>
confusing -- it now measures space usage at various points in the<br>
execution, including the point of peak memory allocation. Its<br>
output format has also changed: instead of producing PostScript<br>
graphs and HTML text, it produces a single text output (via the new<br>
'ms_print' script) that contains both a graph and the old textual<br>
information, but in a more compact and readable form. Finally, the<br>
new version should be more reliable than the old one, as it has been<br>
tested more thoroughly.<br>
<br>
- Cachegrind has been extended to do branch-misprediction profiling.<br>
Both conditional and indirect branches are profiled. The default<br>
behaviour of Cachegrind is unchanged. To use the new functionality,<br>
give the option --branch-sim=yes.<br>
<br>
- A new category of "experimental tools" has been created. Such tools<br>
may not work as well as the standard tools, but are included because<br>
some people will find them useful, and because exposure to a wider<br>
user group provides tool authors with more end-user feedback. These<br>
tools have a "exp-" prefix attached to their names to indicate their<br>
experimental nature. Currently there are two experimental tools:<br>
<br>
* exp-Omega: an instantaneous leak detector. See<br>
exp-omega/docs/omega_introduction.txt.<br>
<br>
* exp-DRD: a data race detector based on the happens-before<br>
relation. See exp-drd/docs/README.txt.<br>
<br>
- Scalability improvements for very large programs, particularly those<br>
which have a million or more malloc'd blocks in use at once. These<br>
improvements mostly affect Memcheck. Memcheck is also up to 10%<br>
faster for all programs, with x86-linux seeing the largest<br>
improvement.<br>
<br>
- Works well on the latest Linux distros. Has been tested on Fedora<br>
Core 8 (x86, amd64, ppc32, ppc64) and openSUSE 10.3. glibc 2.6 and<br>
2.7 are supported. gcc-4.3 (in its current pre-release state) is<br>
supported. At the same time, 3.3.0 retains support for older<br>
distros.<br>
<br>
- The documentation has been modestly reorganised with the aim of<br>
making it easier to find information on common-usage scenarios.<br>
Some advanced material has been moved into a new chapter in the main<br>
manual, so as to unclutter the main flow, and other tidying up has<br>
been done.<br>
<br>
- There is experimental support for AIX 5.3, both 32-bit and 64-bit<br>
processes. You need to be running a 64-bit kernel to use Valgrind<br>
on a 64-bit executable.<br>
<br>
- There have been some changes to command line options, which may<br>
affect you:<br>
<br>
* --log-file-exactly and <br>
--log-file-qualifier options have been removed.<br>
<br>
To make up for this --log-file option has been made more powerful.<br>
It now accepts a %p format specifier, which is replaced with the<br>
process ID, and a %q{FOO} format specifier, which is replaced with<br>
the contents of the environment variable FOO.<br>
<br>
* --child-silent-after-fork=yes|no [no]<br>
<br>
Causes Valgrind to not show any debugging or logging output for<br>
the child process resulting from a fork() call. This can make the<br>
output less confusing (although more misleading) when dealing with<br>
processes that create children.<br>
<br>
* --cachegrind-out-file, --callgrind-out-file and --massif-out-file<br>
<br>
These control the names of the output files produced by<br>
Cachegrind, Callgrind and Massif. They accept the same %p and %q<br>
format specifiers that --log-file accepts. --callgrind-out-file<br>
replaces Callgrind's old --base option.<br>
<br>
* Cachegrind's 'cg_annotate' script no longer uses the --<pid><br>
option to specify the output file. Instead, the first non-option<br>
argument is taken to be the name of the output file, and any<br>
subsequent non-option arguments are taken to be the names of<br>
source files to be annotated.<br>
<br>
* Cachegrind and Callgrind now use directory names where possible in<br>
their output files. This means that the -I option to<br>
'cg_annotate' and 'callgrind_annotate' should not be needed in<br>
most cases. It also means they can correctly handle the case<br>
where two source files in different directories have the same<br>
name.<br>
<br>
- Memcheck offers a new suppression kind: "Jump". This is for<br>
suppressing jump-to-invalid-address errors. Previously you had to<br>
use an "Addr1" suppression, which didn't make much sense.<br>
<br>
- Memcheck has new flags --malloc-fill=<hexnum> and<br>
--free-fill=<hexnum> which free malloc'd / free'd areas with the<br>
specified byte. This can help shake out obscure memory corruption<br>
problems. The definedness and addressability of these areas is<br>
unchanged -- only the contents are affected.<br>
<br>
- The behaviour of Memcheck's client requests VALGRIND_GET_VBITS and<br>
VALGRIND_SET_VBITS have changed slightly. They no longer issue<br>
addressability errors -- if either array is partially unaddressable,<br>
they just return 3 (as before). Also, SET_VBITS doesn't report<br>
definedness errors if any of the V bits are undefined.<br>
<br>
- The following Memcheck client requests have been removed:<br>
VALGRIND_MAKE_NOACCESS<br>
VALGRIND_MAKE_WRITABLE<br>
VALGRIND_MAKE_READABLE<br>
VALGRIND_CHECK_WRITABLE<br>
VALGRIND_CHECK_READABLE<br>
VALGRIND_CHECK_DEFINED<br>
They were deprecated in 3.2.0, when equivalent but better-named client<br>
requests were added. See the 3.2.0 release notes for more details.<br>
<br>
- The behaviour of the tool Lackey has changed slightly. First, the output<br>
from --trace-mem has been made more compact, to reduce the size of the<br>
traces. Second, a new option --trace-superblocks has been added, which<br>
shows the addresses of superblocks (code blocks) as they are executed.<br>
<br>
- The following bugs have been fixed. Note that "n-i-bz" stands for<br>
"not in bugzilla" -- that is, a bug that was reported to us but<br>
never got a bugzilla entry. We encourage you to file bugs in<br>
bugzilla (http://bugs.kde.org/enter_valgrind_bug.cgi) rather than<br>
mailing the developers (or mailing lists) directly.<br>
<br>
n-i-bz x86_linux_REDIR_FOR_index() broken<br>
n-i-bz guest-amd64/toIR.c:2512 (dis_op2_E_G): Assertion `0' failed.<br>
n-i-bz Support x86 INT insn (INT (0xCD) 0x40 - 0x43)<br>
n-i-bz Add sys_utimensat system call for Linux x86 platform<br>
79844 Helgrind complains about race condition which does not exist<br>
82871 Massif output function names too short<br>
89061 Massif: ms_main.c:485 (get_XCon): Assertion `xpt->max_chi...'<br>
92615 Write output from Massif at crash<br>
95483 massif feature request: include peak allocation in report<br>
112163 MASSIF crashed with signal 7 (SIGBUS) after running 2 days<br>
119404 problems running setuid executables (partial fix)<br>
121629 add instruction-counting mode for timing<br>
127371 java vm giving unhandled instruction bytes: 0x26 0x2E 0x64 0x65<br>
129937 ==150380<br>
129576 Massif loses track of memory, incorrect graphs<br>
132132 massif --format=html output does not do html entity escaping<br>
132950 Heap alloc/usage summary<br>
133962 unhandled instruction bytes: 0xF2 0x4C 0xF 0x10<br>
134990 use -fno-stack-protector if possible<br>
136382 ==134990<br>
137396 I would really like helgrind to work again...<br>
137714 x86/amd64->IR: 0x66 0xF 0xF7 0xC6 (maskmovq, maskmovdq)<br>
141631 Massif: percentages don't add up correctly<br>
142706 massif numbers don't seem to add up<br>
143062 massif crashes on app exit with signal 8 SIGFPE<br>
144453 (get_XCon): Assertion 'xpt->max_children != 0' failed.<br>
145559 valgrind aborts when malloc_stats is called<br>
145609 valgrind aborts all runs with 'repeated section!'<br>
145622 --db-attach broken again on x86-64<br>
145837 ==149519<br>
145887 PPC32: getitimer() system call is not supported<br>
146252 ==150678<br>
146456 (update_XCon): Assertion 'xpt->curr_space >= -space_delta'...<br>
146701 ==134990<br>
146781 Adding support for private futexes<br>
147325 valgrind internal error on syscall (SYS_io_destroy, 0)<br>
147498 amd64->IR: 0xF0 0xF 0xB0 0xF (lock cmpxchg %cl,(%rdi))<br>
147545 Memcheck: mc_main.c:817 (get_sec_vbits8): Assertion 'n' failed.<br>
147628 SALC opcode 0xd6 unimplemented<br>
147825 crash on amd64-linux with gcc 4.2 and glibc 2.6 (CFI)<br>
148174 Incorrect type of freed_list_volume causes assertion [...]<br>
148447 x86_64 : new NOP codes: 66 66 66 66 2e 0f 1f<br>
149182 PPC Trap instructions not implemented in valgrind<br>
149504 Assertion hit on alloc_xpt->curr_space >= -space_delta<br>
149519 ppc32: V aborts with SIGSEGV on execution of a signal handler<br>
149892 ==137714<br>
150044 SEGV during stack deregister<br>
150380 dwarf/gcc interoperation (dwarf3 read problems)<br>
150408 ==148447<br>
150678 guest-amd64/toIR.c:3741 (dis_Grp5): Assertion `sz == 4' failed<br>
151209 V unable to execute programs for users with UID > 2^16<br>
151938 help on --db-command= misleading<br>
152022 subw $0x28, %%sp causes assertion failure in memcheck<br>
152357 inb and outb not recognized in 64-bit mode<br>
152501 vex x86->IR: 0x27 0x66 0x89 0x45 (daa) <br>
152818 vex x86->IR: 0xF3 0xAC 0xFC 0x9C (rep lodsb)<br>
<br>
Developer-visible changes:<br>
<br>
- The names of some functions and types within the Vex IR have<br>
changed. Run 'svn log -r1689 VEX/pub/libvex_ir.h' for full details.<br>
Any existing standalone tools will have to be updated to reflect<br>
these changes. The new names should be clearer. The file<br>
VEX/pub/libvex_ir.h is also much better commented.<br>
<br>
- A number of new debugging command line options have been added.<br>
These are mostly of use for debugging the symbol table and line<br>
number readers:<br>
<br>
--trace-symtab-patt=<patt> limit debuginfo tracing to obj name <patt><br>
--trace-cfi=no|yes show call-frame-info details? [no]<br>
--debug-dump=syms mimic /usr/bin/readelf --syms<br>
--debug-dump=line mimic /usr/bin/readelf --debug-dump=line<br>
--debug-dump=frames mimic /usr/bin/readelf --debug-dump=frames<br>
--sym-offsets=yes|no show syms in form 'name+offset' ? [no]<br>
<br>
- Internally, the code base has been further factorised and<br>
abstractified, particularly with respect to support for non-Linux<br>
OSs.<br>
<br>
(3.3.0.RC1: 2 Dec 2007, vex r1803, valgrind r7268).<br>
(3.3.0.RC2: 5 Dec 2007, vex r1804, valgrind r7282).<br>
(3.3.0.RC3: 9 Dec 2007, vex r1804, valgrind r7288).<br>
(3.3.0: 10 Dec 2007, vex r1804, valgrind r7290).<br>
<br>
<br>
<br>
Release 3.2.3 (29 Jan 2007)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
Unfortunately 3.2.2 introduced a regression which can cause an<br>
assertion failure ("vex: the `impossible' happened: eqIRConst") when<br>
running obscure pieces of SSE code. 3.2.3 fixes this and adds one<br>
more glibc-2.5 intercept. In all other respects it is identical to<br>
3.2.2. Please do not use (or package) 3.2.2; instead use 3.2.3.<br>
<br>
n-i-bz vex: the `impossible' happened: eqIRConst<br>
n-i-bz Add an intercept for glibc-2.5 __stpcpy_chk<br>
<br>
(3.2.3: 29 Jan 2007, vex r1732, valgrind r6560).<br>
<br>
<br>
Release 3.2.2 (22 Jan 2007)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
3.2.2 fixes a bunch of bugs in 3.2.1, adds support for glibc-2.5 based<br>
systems (openSUSE 10.2, Fedora Core 6), improves support for icc-9.X<br>
compiled code, and brings modest performance improvements in some<br>
areas, including amd64 floating point, powerpc support, and startup<br>
responsiveness on all targets.<br>
<br>
The fixed bugs are as follows. Note that "n-i-bz" stands for "not in<br>
bugzilla" -- that is, a bug that was reported to us but never got a<br>
bugzilla entry. We encourage you to file bugs in bugzilla<br>
(http://bugs.kde.org/enter_valgrind_bug.cgi) rather than mailing the<br>
developers (or mailing lists) directly.<br>
<br>
129390 ppc?->IR: some kind of VMX prefetch (dstt)<br>
129968 amd64->IR: 0xF 0xAE 0x0 (fxsave)<br>
134319 ==129968<br>
133054 'make install' fails with syntax errors<br>
118903 ==133054<br>
132998 startup fails in when running on UML<br>
134207 pkg-config output contains @VG_PLATFORM@<br>
134727 valgrind exits with "Value too large for defined data type"<br>
n-i-bz ppc32/64: support mcrfs<br>
n-i-bz Cachegrind/Callgrind: Update cache parameter detection<br>
135012 x86->IR: 0xD7 0x8A 0xE0 0xD0 (xlat)<br>
125959 ==135012<br>
126147 x86->IR: 0xF2 0xA5 0xF 0x77 (repne movsw)<br>
136650 amd64->IR: 0xC2 0x8 0x0<br>
135421 x86->IR: unhandled Grp5(R) case 6<br>
n-i-bz Improved documentation of the IR intermediate representation<br>
n-i-bz jcxz (x86) (users list, 8 Nov)<br>
n-i-bz ExeContext hashing fix<br>
n-i-bz fix CFI reading failures ("Dwarf CFI 0:24 0:32 0:48 0:7")<br>
n-i-bz fix Cachegrind/Callgrind simulation bug<br>
n-i-bz libmpiwrap.c: fix handling of MPI_LONG_DOUBLE<br>
n-i-bz make User errors suppressible<br>
136844 corrupted malloc line when using --gen-suppressions=yes<br>
138507 ==136844<br>
n-i-bz Speed up the JIT's register allocator<br>
n-i-bz Fix confusing leak-checker flag hints<br>
n-i-bz Support recent autoswamp versions<br>
n-i-bz ppc32/64 dispatcher speedups<br>
n-i-bz ppc64 front end rld/rlw improvements<br>
n-i-bz ppc64 back end imm64 improvements<br>
136300 support 64K pages on ppc64-linux<br>
139124 == 136300<br>
n-i-bz fix ppc insn set tests for gcc >= 4.1<br>
137493 x86->IR: recent binutils no-ops<br>
137714 x86->IR: 0x66 0xF 0xF7 0xC6 (maskmovdqu)<br>
138424 "failed in UME with error 22" (produce a better error msg)<br>
138856 ==138424<br>
138627 Enhancement support for prctl ioctls<br>
138896 Add support for usb ioctls<br>
136059 ==138896<br>
139050 ppc32->IR: mfspr 268/269 instructions not handled<br>
n-i-bz ppc32->IR: lvxl/stvxl<br>
n-i-bz glibc-2.5 support<br>
n-i-bz memcheck: provide replacement for mempcpy<br>
n-i-bz memcheck: replace bcmp in ld.so<br>
n-i-bz Use 'ifndef' in VEX's Makefile correctly<br>
n-i-bz Suppressions for MVL 4.0.1 on ppc32-linux<br>
n-i-bz libmpiwrap.c: Fixes for MPICH<br>
n-i-bz More robust handling of hinted client mmaps<br>
139776 Invalid read in unaligned memcpy with Intel compiler v9<br>
n-i-bz Generate valid XML even for very long fn names<br>
n-i-bz Don't prompt about suppressions for unshown reachable leaks<br>
139910 amd64 rcl is not supported<br>
n-i-bz DWARF CFI reader: handle DW_CFA_undefined<br>
n-i-bz DWARF CFI reader: handle icc9 generated CFI info better<br>
n-i-bz fix false uninit-value errs in icc9 generated FP code<br>
n-i-bz reduce extraneous frames in libmpiwrap.c<br>
n-i-bz support pselect6 on amd64-linux<br>
<br>
(3.2.2: 22 Jan 2007, vex r1729, valgrind r6545).<br>
<br>
<br>
Release 3.2.1 (16 Sept 2006)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
3.2.1 adds x86/amd64 support for all SSE3 instructions except monitor<br>
and mwait, further reduces memcheck's false error rate on all<br>
platforms, adds support for recent binutils (in OpenSUSE 10.2 and<br>
Fedora Rawhide) and fixes a bunch of bugs in 3.2.0. Some of the fixed<br>
bugs were causing large programs to segfault with --tool=callgrind and<br>
--tool=cachegrind, so an upgrade is recommended.<br>
<br>
In view of the fact that any 3.3.0 release is unlikely to happen until<br>
well into 1Q07, we intend to keep the 3.2.X line alive for a while<br>
yet, and so we tentatively plan a 3.2.2 release sometime in December<br>
06.<br>
<br>
The fixed bugs are as follows. Note that "n-i-bz" stands for "not in<br>
bugzilla" -- that is, a bug that was reported to us but never got a<br>
bugzilla entry.<br>
<br>
n-i-bz Expanding brk() into last available page asserts<br>
n-i-bz ppc64-linux stack RZ fast-case snafu<br>
n-i-bz 'c' in --gen-supps=yes doesn't work<br>
n-i-bz VG_N_SEGMENTS too low (users, 28 June)<br>
n-i-bz VG_N_SEGNAMES too low (Stu Robinson)<br>
106852 x86->IR: fisttp (SSE3)<br>
117172 FUTEX_WAKE does not use uaddr2<br>
124039 Lacks support for VKI_[GP]IO_UNIMAP*<br>
127521 amd64->IR: 0xF0 0x48 0xF 0xC7 (cmpxchg8b)<br>
128917 amd64->IR: 0x66 0xF 0xF6 0xC4 (psadbw,SSE2)<br>
129246 JJ: ppc32/ppc64 syscalls, w/ patch<br>
129358 x86->IR: fisttpl (SSE3)<br>
129866 cachegrind/callgrind causes executable to die<br>
130020 Can't stat .so/.exe error while reading symbols<br>
130388 Valgrind aborts when process calls malloc_trim()<br>
130638 PATCH: ppc32 missing system calls<br>
130785 amd64->IR: unhandled instruction "pushfq"<br>
131481: (HINT_NOP) vex x86->IR: 0xF 0x1F 0x0 0xF<br>
131298 ==131481<br>
132146 Programs with long sequences of bswap[l,q]s<br>
132918 vex amd64->IR: 0xD9 0xF8 (fprem)<br>
132813 Assertion at priv/guest-x86/toIR.c:652 fails<br>
133051 'cfsi->len > 0 && cfsi->len < 2000000' failed<br>
132722 valgrind header files are not standard C<br>
n-i-bz Livelocks entire machine (users list, Timothy Terriberry)<br>
n-i-bz Alex Bennee mmap problem (9 Aug)<br>
n-i-bz BartV: Don't print more lines of a stack-trace than were obtained.<br>
n-i-bz ppc32 SuSE 10.1 redir<br>
n-i-bz amd64 padding suppressions<br>
n-i-bz amd64 insn printing fix.<br>
n-i-bz ppc cmp reg,reg fix<br>
n-i-bz x86/amd64 iropt e/rflag reduction rules<br>
n-i-bz SuSE 10.1 (ppc32) minor fixes<br>
133678 amd64->IR: 0x48 0xF 0xC5 0xC0 (pextrw?)<br>
133694 aspacem assertion: aspacem_minAddr <= holeStart<br>
n-i-bz callgrind: fix warning about malformed creator line <br>
n-i-bz callgrind: fix annotate script for data produced with <br>
--dump-instr=yes<br>
n-i-bz callgrind: fix failed assertion when toggling <br>
instrumentation mode<br>
n-i-bz callgrind: fix annotate script fix warnings with<br>
--collect-jumps=yes<br>
n-i-bz docs path hardwired (Dennis Lubert)<br>
<br>
The following bugs were not fixed, due primarily to lack of developer<br>
time, and also because bug reporters did not answer requests for<br>
feedback in time for the release:<br>
<br>
129390 ppc?->IR: some kind of VMX prefetch (dstt)<br>
129968 amd64->IR: 0xF 0xAE 0x0 (fxsave)<br>
133054 'make install' fails with syntax errors<br>
n-i-bz Signal race condition (users list, 13 June, Johannes Berg)<br>
n-i-bz Unrecognised instruction at address 0x70198EC2 (users list,<br>
19 July, Bennee)<br>
132998 startup fails in when running on UML<br>
<br>
The following bug was tentatively fixed on the mainline but the fix<br>
was considered too risky to push into 3.2.X:<br>
<br>
133154 crash when using client requests to register/deregister stack<br>
<br>
(3.2.1: 16 Sept 2006, vex r1658, valgrind r6070).<br>
<br>
<br>
Release 3.2.0 (7 June 2006)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
3.2.0 is a feature release with many significant improvements and the<br>
usual collection of bug fixes. This release supports X86/Linux,<br>
AMD64/Linux, PPC32/Linux and PPC64/Linux.<br>
<br>
Performance, especially of Memcheck, is improved, Addrcheck has been<br>
removed, Callgrind has been added, PPC64/Linux support has been added,<br>
Lackey has been improved, and MPI support has been added. In detail:<br>
<br>
- Memcheck has improved speed and reduced memory use. Run times are<br>
typically reduced by 15-30%, averaging about 24% for SPEC CPU2000.<br>
The other tools have smaller but noticeable speed improvments. We<br>
are interested to hear what improvements users get.<br>
<br>
Memcheck uses less memory due to the introduction of a compressed<br>
representation for shadow memory. The space overhead has been<br>
reduced by a factor of up to four, depending on program behaviour.<br>
This means you should be able to run programs that use more memory<br>
than before without hitting problems.<br>
<br>
- Addrcheck has been removed. It has not worked since version 2.4.0,<br>
and the speed and memory improvements to Memcheck make it redundant.<br>
If you liked using Addrcheck because it didn't give undefined value<br>
errors, you can use the new Memcheck option --undef-value-errors=no<br>
to get the same behaviour.<br>
<br>
- The number of undefined-value errors incorrectly reported by<br>
Memcheck has been reduced (such false reports were already very<br>
rare). In particular, efforts have been made to ensure Memcheck<br>
works really well with gcc 4.0/4.1-generated code on X86/Linux and<br>
AMD64/Linux.<br>
<br>
- Josef Weidendorfer's popular Callgrind tool has been added. Folding<br>
it in was a logical step given its popularity and usefulness, and<br>
makes it easier for us to ensure it works "out of the box" on all<br>
supported targets. The associated KDE KCachegrind GUI remains a<br>
separate project.<br>
<br>
- A new release of the Valkyrie GUI for Memcheck, version 1.2.0,<br>
accompanies this release. Improvements over previous releases<br>
include improved robustness, many refinements to the user interface,<br>
and use of a standard autoconf/automake build system. You can get<br>
it from http://www.valgrind.org/downloads/guis.html.<br>
<br>
- Valgrind now works on PPC64/Linux. As with the AMD64/Linux port,<br>
this supports programs using to 32G of address space. On 64-bit<br>
capable PPC64/Linux setups, you get a dual architecture build so<br>
that both 32-bit and 64-bit executables can be run. Linux on POWER5<br>
is supported, and POWER4 is also believed to work. Both 32-bit and<br>
64-bit DWARF2 is supported. This port is known to work well with<br>
both gcc-compiled and xlc/xlf-compiled code.<br>
<br>
- Floating point accuracy has been improved for PPC32/Linux.<br>
Specifically, the floating point rounding mode is observed on all FP<br>
arithmetic operations, and multiply-accumulate instructions are<br>
preserved by the compilation pipeline. This means you should get FP<br>
results which are bit-for-bit identical to a native run. These<br>
improvements are also present in the PPC64/Linux port.<br>
<br>
- Lackey, the example tool, has been improved:<br>
<br>
* It has a new option --detailed-counts (off by default) which<br>
causes it to print out a count of loads, stores and ALU operations<br>
done, and their sizes.<br>
<br>
* It has a new option --trace-mem (off by default) which causes it<br>
to print out a trace of all memory accesses performed by a<br>
program. It's a good starting point for building Valgrind tools<br>
that need to track memory accesses. Read the comments at the top<br>
of the file lackey/lk_main.c for details.<br>
<br>
* The original instrumentation (counting numbers of instructions,<br>
jumps, etc) is now controlled by a new option --basic-counts. It<br>
is on by default.<br>
<br>
- MPI support: partial support for debugging distributed applications<br>
using the MPI library specification has been added. Valgrind is<br>
aware of the memory state changes caused by a subset of the MPI<br>
functions, and will carefully check data passed to the (P)MPI_<br>
interface.<br>
<br>
- A new flag, --error-exitcode=, has been added. This allows changing<br>
the exit code in runs where Valgrind reported errors, which is<br>
useful when using Valgrind as part of an automated test suite.<br>
<br>
- Various segfaults when reading old-style "stabs" debug information<br>
have been fixed.<br>
<br>
- A simple performance evaluation suite has been added. See<br>
perf/README and README_DEVELOPERS for details. There are<br>
various bells and whistles.<br>
<br>
- New configuration flags:<br>
--enable-only32bit<br>
--enable-only64bit<br>
By default, on 64 bit platforms (ppc64-linux, amd64-linux) the build<br>
system will attempt to build a Valgrind which supports both 32-bit<br>
and 64-bit executables. This may not be what you want, and you can<br>
override the default behaviour using these flags.<br>
<br>
Please note that Helgrind is still not working. We have made an<br>
important step towards making it work again, however, with the<br>
addition of function wrapping (see below).<br>
<br>
Other user-visible changes:<br>
<br>
- Valgrind now has the ability to intercept and wrap arbitrary<br>
functions. This is a preliminary step towards making Helgrind work<br>
again, and was required for MPI support.<br>
<br>
- There are some changes to Memcheck's client requests. Some of them<br>
have changed names:<br>
<br>
MAKE_NOACCESS --> MAKE_MEM_NOACCESS<br>
MAKE_WRITABLE --> MAKE_MEM_UNDEFINED<br>
MAKE_READABLE --> MAKE_MEM_DEFINED<br>
<br>
CHECK_WRITABLE --> CHECK_MEM_IS_ADDRESSABLE<br>
CHECK_READABLE --> CHECK_MEM_IS_DEFINED<br>
CHECK_DEFINED --> CHECK_VALUE_IS_DEFINED<br>
<br>
The reason for the change is that the old names are subtly<br>
misleading. The old names will still work, but they are deprecated<br>
and may be removed in a future release.<br>
<br>
We also added a new client request:<br>
<br>
MAKE_MEM_DEFINED_IF_ADDRESSABLE(a, len)<br>
<br>
which is like MAKE_MEM_DEFINED but only affects a byte if the byte is<br>
already addressable.<br>
<br>
- The way client requests are encoded in the instruction stream has<br>
changed. Unfortunately, this means 3.2.0 will not honour client<br>
requests compiled into binaries using headers from earlier versions<br>
of Valgrind. We will try to keep the client request encodings more <br>
stable in future.<br>
<br>
BUGS FIXED:<br>
<br>
108258 NPTL pthread cleanup handlers not called <br>
117290 valgrind is sigKILL'd on startup<br>
117295 == 117290<br>
118703 m_signals.c:1427 Assertion 'tst->status == VgTs_WaitSys'<br>
118466 add %reg, %reg generates incorrect validity for bit 0<br>
123210 New: strlen from ld-linux on amd64<br>
123244 DWARF2 CFI reader: unhandled CFI instruction 0:18<br>
123248 syscalls in glibc-2.4: openat, fstatat, symlinkat<br>
123258 socketcall.recvmsg(msg.msg_iov[i] points to uninit<br>
123535 mremap(new_addr) requires MREMAP_FIXED in 4th arg<br>
123836 small typo in the doc<br>
124029 ppc compile failed: `vor' gcc 3.3.5<br>
124222 Segfault: @@don't know what type ':' is<br>
124475 ppc32: crash (syscall?) timer_settime()<br>
124499 amd64->IR: 0xF 0xE 0x48 0x85 (femms)<br>
124528 FATAL: aspacem assertion failed: segment_is_sane<br>
124697 vex x86->IR: 0xF 0x70 0xC9 0x0 (pshufw)<br>
124892 vex x86->IR: 0xF3 0xAE (REPx SCASB)<br>
126216 == 124892<br>
124808 ppc32: sys_sched_getaffinity() not handled<br>
n-i-bz Very long stabs strings crash m_debuginfo<br>
n-i-bz amd64->IR: 0x66 0xF 0xF5 (pmaddwd)<br>
125492 ppc32: support a bunch more syscalls<br>
121617 ppc32/64: coredumping gives assertion failure<br>
121814 Coregrind return error as exitcode patch<br>
126517 == 121814<br>
125607 amd64->IR: 0x66 0xF 0xA3 0x2 (btw etc)<br>
125651 amd64->IR: 0xF8 0x49 0xFF 0xE3 (clc?)<br>
126253 x86 movx is wrong<br>
126451 3.2 SVN doesn't work on ppc32 CPU's without FPU<br>
126217 increase # threads<br>
126243 vex x86->IR: popw mem<br>
126583 amd64->IR: 0x48 0xF 0xA4 0xC2 (shld $1,%rax,%rdx)<br>
126668 amd64->IR: 0x1C 0xFF (sbb $0xff,%al)<br>
126696 support for CDROMREADRAW ioctl and CDROMREADTOCENTRY fix<br>
126722 assertion: segment_is_sane at m_aspacemgr/aspacemgr.c:1624<br>
126938 bad checking for syscalls linkat, renameat, symlinkat<br>
<br>
(3.2.0RC1: 27 May 2006, vex r1626, valgrind r5947).<br>
(3.2.0: 7 June 2006, vex r1628, valgrind r5957).<br>
<br>
<br>
Release 3.1.1 (15 March 2006)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
3.1.1 fixes a bunch of bugs reported in 3.1.0. There is no new<br>
functionality. The fixed bugs are:<br>
<br>
(note: "n-i-bz" means "not in bugzilla" -- this bug does not have<br>
a bugzilla entry).<br>
<br>
n-i-bz ppc32: fsub 3,3,3 in dispatcher doesn't clear NaNs<br>
n-i-bz ppc32: __NR_{set,get}priority<br>
117332 x86: missing line info with icc 8.1<br>
117366 amd64: 0xDD 0x7C fnstsw<br>
118274 == 117366<br>
117367 amd64: 0xD9 0xF4 fxtract<br>
117369 amd64: __NR_getpriority (140)<br>
117419 ppc32: lfsu f5, -4(r11)<br>
117419 ppc32: fsqrt<br>
117936 more stabs problems (segfaults while reading debug info)<br>
119914 == 117936<br>
120345 == 117936<br>
118239 amd64: 0xF 0xAE 0x3F (clflush)<br>
118939 vm86old system call<br>
n-i-bz memcheck/tests/mempool reads freed memory<br>
n-i-bz AshleyP's custom-allocator assertion<br>
n-i-bz Dirk strict-aliasing stuff<br>
n-i-bz More space for debugger cmd line (Dan Thaler)<br>
n-i-bz Clarified leak checker output message<br>
n-i-bz AshleyP's --gen-suppressions output fix<br>
n-i-bz cg_annotate's --sort option broken<br>
n-i-bz OSet 64-bit fastcmp bug<br>
n-i-bz VG_(getgroups) fix (Shinichi Noda)<br>
n-i-bz ppc32: allocate from callee-saved FP/VMX regs<br>
n-i-bz misaligned path word-size bug in mc_main.c<br>
119297 Incorrect error message for sse code<br>
120410 x86: prefetchw (0xF 0xD 0x48 0x4)<br>
120728 TIOCSERGETLSR, TIOCGICOUNT, HDIO_GET_DMA ioctls<br>
120658 Build fixes for gcc 2.96<br>
120734 x86: Support for changing EIP in signal handler<br>
n-i-bz memcheck/tests/zeropage de-looping fix<br>
n-i-bz x86: fxtract doesn't work reliably<br>
121662 x86: lock xadd (0xF0 0xF 0xC0 0x2)<br>
121893 calloc does not always return zeroed memory<br>
121901 no support for syscall tkill<br>
n-i-bz Suppression update for Debian unstable<br>
122067 amd64: fcmovnu (0xDB 0xD9)<br>
n-i-bz ppc32: broken signal handling in cpu feature detection<br>
n-i-bz ppc32: rounding mode problems (improved, partial fix only)<br>
119482 ppc32: mtfsb1<br>
n-i-bz ppc32: mtocrf/mfocrf<br>
<br>
(3.1.1: 15 March 2006, vex r1597, valgrind r5771).<br>
<br>
<br>
Release 3.1.0 (25 November 2005)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
3.1.0 is a feature release with a number of significant improvements:<br>
AMD64 support is much improved, PPC32 support is good enough to be<br>
usable, and the handling of memory management and address space is<br>
much more robust. In detail:<br>
<br>
- AMD64 support is much improved. The 64-bit vs. 32-bit issues in<br>
3.0.X have been resolved, and it should "just work" now in all<br>
cases. On AMD64 machines both 64-bit and 32-bit versions of<br>
Valgrind are built. The right version will be invoked<br>
automatically, even when using --trace-children and mixing execution<br>
between 64-bit and 32-bit executables. Also, many more instructions<br>
are supported.<br>
<br>
- PPC32 support is now good enough to be usable. It should work with<br>
all tools, but please let us know if you have problems. Three<br>
classes of CPUs are supported: integer only (no FP, no Altivec),<br>
which covers embedded PPC uses, integer and FP but no Altivec<br>
(G3-ish), and CPUs capable of Altivec too (G4, G5).<br>
<br>
- Valgrind's address space management has been overhauled. As a<br>
result, Valgrind should be much more robust with programs that use<br>
large amounts of memory. There should be many fewer "memory<br>
exhausted" messages, and debug symbols should be read correctly on<br>
large (eg. 300MB+) executables. On 32-bit machines the full address<br>
space available to user programs (usually 3GB or 4GB) can be fully<br>
utilised. On 64-bit machines up to 32GB of space is usable; when<br>
using Memcheck that means your program can use up to about 14GB.<br>
<br>
A side effect of this change is that Valgrind is no longer protected<br>
against wild writes by the client. This feature was nice but relied<br>
on the x86 segment registers and so wasn't portable.<br>
<br>
- Most users should not notice, but as part of the address space<br>
manager change, the way Valgrind is built has been changed. Each<br>
tool is now built as a statically linked stand-alone executable,<br>
rather than as a shared object that is dynamically linked with the<br>
core. The "valgrind" program invokes the appropriate tool depending<br>
on the --tool option. This slightly increases the amount of disk<br>
space used by Valgrind, but it greatly simplified many things and<br>
removed Valgrind's dependence on glibc.<br>
<br>
Please note that Addrcheck and Helgrind are still not working. Work<br>
is underway to reinstate them (or equivalents). We apologise for the<br>
inconvenience.<br>
<br>
Other user-visible changes:<br>
<br>
- The --weird-hacks option has been renamed --sim-hints.<br>
<br>
- The --time-stamp option no longer gives an absolute date and time.<br>
It now prints the time elapsed since the program began.<br>
<br>
- It should build with gcc-2.96.<br>
<br>
- Valgrind can now run itself (see README_DEVELOPERS for how).<br>
This is not much use to you, but it means the developers can now<br>
profile Valgrind using Cachegrind. As a result a couple of<br>
performance bad cases have been fixed.<br>
<br>
- The XML output format has changed slightly. See<br>
docs/internals/xml-output.txt.<br>
<br>
- Core dumping has been reinstated (it was disabled in 3.0.0 and 3.0.1).<br>
If your program crashes while running under Valgrind, a core file with<br>
the name "vgcore.<pid>" will be created (if your settings allow core<br>
file creation). Note that the floating point information is not all<br>
there. If Valgrind itself crashes, the OS will create a normal core<br>
file.<br>
<br>
The following are some user-visible changes that occurred in earlier<br>
versions that may not have been announced, or were announced but not<br>
widely noticed. So we're mentioning them now.<br>
<br>
- The --tool flag is optional once again; if you omit it, Memcheck<br>
is run by default.<br>
<br>
- The --num-callers flag now has a default value of 12. It was<br>
previously 4.<br>
<br>
- The --xml=yes flag causes Valgrind's output to be produced in XML<br>
format. This is designed to make it easy for other programs to<br>
consume Valgrind's output. The format is described in the file<br>
docs/internals/xml-format.txt.<br>
<br>
- The --gen-suppressions flag supports an "all" value that causes every<br>
suppression to be printed without asking.<br>
<br>
- The --log-file option no longer puts "pid" in the filename, eg. the<br>
old name "foo.pid12345" is now "foo.12345".<br>
<br>
- There are several graphical front-ends for Valgrind, such as Valkyrie,<br>
Alleyoop and Valgui. See http://www.valgrind.org/downloads/guis.html<br>
for a list.<br>
<br>
BUGS FIXED:<br>
<br>
109861 amd64 hangs at startup<br>
110301 ditto<br>
111554 valgrind crashes with Cannot allocate memory<br>
111809 Memcheck tool doesn't start java<br>
111901 cross-platform run of cachegrind fails on opteron<br>
113468 (vgPlain_mprotect_range): Assertion 'r != -1' failed.<br>
92071 Reading debugging info uses too much memory<br>
109744 memcheck loses track of mmap from direct ld-linux.so.2<br>
110183 tail of page with _end<br>
82301 FV memory layout too rigid<br>
98278 Infinite recursion possible when allocating memory<br>
108994 Valgrind runs out of memory due to 133x overhead<br>
115643 valgrind cannot allocate memory<br>
105974 vg_hashtable.c static hash table<br>
109323 ppc32: dispatch.S uses Altivec insn, which doesn't work on POWER. <br>
109345 ptrace_setregs not yet implemented for ppc<br>
110831 Would like to be able to run against both 32 and 64 bit <br>
binaries on AMD64<br>
110829 == 110831<br>
111781 compile of valgrind-3.0.0 fails on my linux (gcc 2.X prob)<br>
112670 Cachegrind: cg_main.c:486 (handleOneStatement ...<br>
112941 vex x86: 0xD9 0xF4 (fxtract)<br>
110201 == 112941<br>
113015 vex amd64->IR: 0xE3 0x14 0x48 0x83 (jrcxz)<br>
113126 Crash with binaries built with -gstabs+/-ggdb<br>
104065 == 113126<br>
115741 == 113126<br>
113403 Partial SSE3 support on x86<br>
113541 vex: Grp5(x86) (alt encoding inc/dec) case 1<br>
113642 valgrind crashes when trying to read debug information<br>
113810 vex x86->IR: 66 0F F6 (66 + PSADBW == SSE PSADBW)<br>
113796 read() and write() do not work if buffer is in shared memory<br>
113851 vex x86->IR: (pmaddwd): 0x66 0xF 0xF5 0xC7<br>
114366 vex amd64 cannnot handle __asm__( "fninit" )<br>
114412 vex amd64->IR: 0xF 0xAD 0xC2 0xD3 (128-bit shift, shrdq?)<br>
114455 vex amd64->IR: 0xF 0xAC 0xD0 0x1 (also shrdq)<br>
115590: amd64->IR: 0x67 0xE3 0x9 0xEB (address size override)<br>
115953 valgrind svn r5042 does not build with parallel make (-j3)<br>
116057 maximum instruction size - VG_MAX_INSTR_SZB too small?<br>
116483 shmat failes with invalid argument<br>
102202 valgrind crashes when realloc'ing until out of memory<br>
109487 == 102202<br>
110536 == 102202<br>
112687 == 102202<br>
111724 vex amd64->IR: 0x41 0xF 0xAB (more BT{,S,R,C} fun n games)<br>
111748 vex amd64->IR: 0xDD 0xE2 (fucom)<br>
111785 make fails if CC contains spaces<br>
111829 vex x86->IR: sbb AL, Ib<br>
111851 vex x86->IR: 0x9F 0x89 (lahf/sahf)<br>
112031 iopl on AMD64 and README_MISSING_SYSCALL_OR_IOCTL update<br>
112152 code generation for Xin_MFence on x86 with SSE0 subarch<br>
112167 == 112152<br>
112789 == 112152<br>
112199 naked ar tool is used in vex makefile<br>
112501 vex x86->IR: movq (0xF 0x7F 0xC1 0xF) (mmx MOVQ)<br>
113583 == 112501<br>
112538 memalign crash<br>
113190 Broken links in docs/html/<br>
113230 Valgrind sys_pipe on x86-64 wrongly thinks file descriptors<br>
should be 64bit<br>
113996 vex amd64->IR: fucomp (0xDD 0xE9)<br>
114196 vex x86->IR: out %eax,(%dx) (0xEF 0xC9 0xC3 0x90)<br>
114289 Memcheck fails to intercept malloc when used in an uclibc environment<br>
114756 mbind syscall support<br>
114757 Valgrind dies with assertion: Assertion 'noLargerThan > 0' failed<br>
114563 stack tracking module not informed when valgrind switches threads<br>
114564 clone() and stacks<br>
114565 == 114564<br>
115496 glibc crashes trying to use sysinfo page<br>
116200 enable fsetxattr, fgetxattr, and fremovexattr for amd64<br>
<br>
(3.1.0RC1: 20 November 2005, vex r1466, valgrind r5224).<br>
(3.1.0: 26 November 2005, vex r1471, valgrind r5235).<br>
<br>
<br>
Release 3.0.1 (29 August 2005)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
3.0.1 fixes a bunch of bugs reported in 3.0.0. There is no new<br>
functionality. Some of the fixed bugs are critical, so if you<br>
use/distribute 3.0.0, an upgrade to 3.0.1 is recommended. The fixed<br>
bugs are:<br>
<br>
(note: "n-i-bz" means "not in bugzilla" -- this bug does not have<br>
a bugzilla entry).<br>
<br>
109313 (== 110505) x86 cmpxchg8b<br>
n-i-bz x86: track but ignore changes to %eflags.AC (alignment check)<br>
110102 dis_op2_E_G(amd64)<br>
110202 x86 sys_waitpid(#286)<br>
110203 clock_getres(,0)<br>
110208 execve fail wrong retval<br>
110274 SSE1 now mandatory for x86<br>
110388 amd64 0xDD 0xD1<br>
110464 amd64 0xDC 0x1D FCOMP<br>
110478 amd64 0xF 0xD PREFETCH<br>
n-i-bz XML <unique> printing wrong<br>
n-i-bz Dirk r4359 (amd64 syscalls from trunk)<br>
110591 amd64 and x86: rdtsc not implemented properly<br>
n-i-bz Nick r4384 (stub implementations of Addrcheck and Helgrind)<br>
110652 AMD64 valgrind crashes on cwtd instruction<br>
110653 AMD64 valgrind crashes on sarb $0x4,foo(%rip) instruction<br>
110656 PATH=/usr/bin::/bin valgrind foobar stats ./fooba<br>
110657 Small test fixes<br>
110671 vex x86->IR: unhandled instruction bytes: 0xF3 0xC3 (rep ret)<br>
n-i-bz Nick (Cachegrind should not assert when it encounters a client<br>
request.)<br>
110685 amd64->IR: unhandled instruction bytes: 0xE1 0x56 (loope Jb)<br>
110830 configuring with --host fails to build 32 bit on 64 bit target<br>
110875 Assertion when execve fails<br>
n-i-bz Updates to Memcheck manual<br>
n-i-bz Fixed broken malloc_usable_size()<br>
110898 opteron instructions missing: btq btsq btrq bsfq<br>
110954 x86->IR: unhandled instruction bytes: 0xE2 0xF6 (loop Jb)<br>
n-i-bz Make suppressions work for "???" lines in stacktraces.<br>
111006 bogus warnings from linuxthreads<br>
111092 x86: dis_Grp2(Reg): unhandled case(x86) <br>
111231 sctp_getladdrs() and sctp_getpaddrs() returns uninitialized<br>
memory<br>
111102 (comment #4) Fixed 64-bit unclean "silly arg" message<br>
n-i-bz vex x86->IR: unhandled instruction bytes: 0x14 0x0<br>
n-i-bz minor umount/fcntl wrapper fixes<br>
111090 Internal Error running Massif<br>
101204 noisy warning<br>
111513 Illegal opcode for SSE instruction (x86 movups)<br>
111555 VEX/Makefile: CC is set to gcc<br>
n-i-bz Fix XML bugs in FAQ<br>
<br>
(3.0.1: 29 August 05,<br>
vex/branches/VEX_3_0_BRANCH r1367,<br>
valgrind/branches/VALGRIND_3_0_BRANCH r4574).<br>
<br>
<br>
<br>
Release 3.0.0 (3 August 2005)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
3.0.0 is a major overhaul of Valgrind. The most significant user<br>
visible change is that Valgrind now supports architectures other than<br>
x86. The new architectures it supports are AMD64 and PPC32, and the<br>
infrastructure is present for other architectures to be added later.<br>
<br>
AMD64 support works well, but has some shortcomings:<br>
<br>
- It generally won't be as solid as the x86 version. For example,<br>
support for more obscure instructions and system calls may be missing.<br>
We will fix these as they arise.<br>
<br>
- Address space may be limited; see the point about<br>
position-independent executables below.<br>
<br>
- If Valgrind is built on an AMD64 machine, it will only run 64-bit<br>
executables. If you want to run 32-bit x86 executables under Valgrind<br>
on an AMD64, you will need to build Valgrind on an x86 machine and<br>
copy it to the AMD64 machine. And it probably won't work if you do<br>
something tricky like exec'ing a 32-bit program from a 64-bit program<br>
while using --trace-children=yes. We hope to improve this situation<br>
in the future.<br>
<br>
The PPC32 support is very basic. It may not work reliably even for<br>
small programs, but it's a start. Many thanks to Paul Mackerras for<br>
his great work that enabled this support. We are working to make<br>
PPC32 usable as soon as possible.<br>
<br>
Other user-visible changes:<br>
<br>
- Valgrind is no longer built by default as a position-independent<br>
executable (PIE), as this caused too many problems.<br>
<br>
Without PIE enabled, AMD64 programs will only be able to access 2GB of<br>
address space. We will fix this eventually, but not for the moment.<br>
<br>
Use --enable-pie at configure-time to turn this on.<br>
<br>
- Support for programs that use stack-switching has been improved. Use<br>
the --max-stackframe flag for simple cases, and the<br>
VALGRIND_STACK_REGISTER, VALGRIND_STACK_DEREGISTER and<br>
VALGRIND_STACK_CHANGE client requests for trickier cases.<br>
<br>
- Support for programs that use self-modifying code has been improved,<br>
in particular programs that put temporary code fragments on the stack.<br>
This helps for C programs compiled with GCC that use nested functions,<br>
and also Ada programs. This is controlled with the --smc-check<br>
flag, although the default setting should work in most cases.<br>
<br>
- Output can now be printed in XML format. This should make it easier<br>
for tools such as GUI front-ends and automated error-processing<br>
schemes to use Valgrind output as input. The --xml flag controls this.<br>
As part of this change, ELF directory information is read from executables,<br>
so absolute source file paths are available if needed.<br>
<br>
- Programs that allocate many heap blocks may run faster, due to<br>
improvements in certain data structures.<br>
<br>
- Addrcheck is currently not working. We hope to get it working again<br>
soon. Helgrind is still not working, as was the case for the 2.4.0<br>
release.<br>
<br>
- The JITter has been completely rewritten, and is now in a separate<br>
library, called Vex. This enabled a lot of the user-visible changes,<br>
such as new architecture support. The new JIT unfortunately translates<br>
more slowly than the old one, so programs may take longer to start.<br>
We believe the code quality is produces is about the same, so once<br>
started, programs should run at about the same speed. Feedback about<br>
this would be useful.<br>
<br>
On the plus side, Vex and hence Memcheck tracks value flow properly<br>
through floating point and vector registers, something the 2.X line<br>
could not do. That means that Memcheck is much more likely to be<br>
usably accurate on vectorised code.<br>
<br>
- There is a subtle change to the way exiting of threaded programs<br>
is handled. In 3.0, Valgrind's final diagnostic output (leak check,<br>
etc) is not printed until the last thread exits. If the last thread<br>
to exit was not the original thread which started the program, any<br>
other process wait()-ing on this one to exit may conclude it has<br>
finished before the diagnostic output is printed. This may not be<br>
what you expect. 2.X had a different scheme which avoided this<br>
problem, but caused deadlocks under obscure circumstances, so we<br>
are trying something different for 3.0.<br>
<br>
- Small changes in control log file naming which make it easier to<br>
use valgrind for debugging MPI-based programs. The relevant<br>
new flags are --log-file-exactly= and --log-file-qualifier=.<br>
<br>
- As part of adding AMD64 support, DWARF2 CFI-based stack unwinding<br>
support was added. In principle this means Valgrind can produce<br>
meaningful backtraces on x86 code compiled with -fomit-frame-pointer<br>
providing you also compile your code with -fasynchronous-unwind-tables.<br>
<br>
- The documentation build system has been completely redone.<br>
The documentation masters are now in XML format, and from that<br>
HTML, PostScript and PDF documentation is generated. As a result<br>
the manual is now available in book form. Note that the<br>
documentation in the source tarballs is pre-built, so you don't need<br>
any XML processing tools to build Valgrind from a tarball.<br>
<br>
Changes that are not user-visible:<br>
<br>
- The code has been massively overhauled in order to modularise it.<br>
As a result we hope it is easier to navigate and understand.<br>
<br>
- Lots of code has been rewritten.<br>
<br>
BUGS FIXED:<br>
<br>
110046 sz == 4 assertion failed <br>
109810 vex amd64->IR: unhandled instruction bytes: 0xA3 0x4C 0x70 0xD7<br>
109802 Add a plausible_stack_size command-line parameter ?<br>
109783 unhandled ioctl TIOCMGET (running hw detection tool discover) <br>
109780 unhandled ioctl BLKSSZGET (running fdisk -l /dev/hda)<br>
109718 vex x86->IR: unhandled instruction: ffreep <br>
109429 AMD64 unhandled syscall: 127 (sigpending)<br>
109401 false positive uninit in strchr from ld-linux.so.2<br>
109385 "stabs" parse failure <br>
109378 amd64: unhandled instruction REP NOP<br>
109376 amd64: unhandled instruction LOOP Jb <br>
109363 AMD64 unhandled instruction bytes <br>
109362 AMD64 unhandled syscall: 24 (sched_yield)<br>
109358 fork() won't work with valgrind-3.0 SVN<br>
109332 amd64 unhandled instruction: ADC Ev, Gv<br>
109314 Bogus memcheck report on amd64<br>
108883 Crash; vg_memory.c:905 (vgPlain_init_shadow_range):<br>
Assertion `vgPlain_defined_init_shadow_page()' failed.<br>
108349 mincore syscall parameter checked incorrectly <br>
108059 build infrastructure: small update<br>
107524 epoll_ctl event parameter checked on EPOLL_CTL_DEL<br>
107123 Vex dies with unhandled instructions: 0xD9 0x31 0xF 0xAE<br>
106841 auxmap & openGL problems<br>
106713 SDL_Init causes valgrind to exit<br>
106352 setcontext and makecontext not handled correctly <br>
106293 addresses beyond initial client stack allocation <br>
not checked in VALGRIND_DO_LEAK_CHECK<br>
106283 PIE client programs are loaded at address 0<br>
105831 Assertion `vgPlain_defined_init_shadow_page()' failed.<br>
105039 long run-times probably due to memory manager <br>
104797 valgrind needs to be aware of BLKGETSIZE64<br>
103594 unhandled instruction: FICOM<br>
103320 Valgrind 2.4.0 fails to compile with gcc 3.4.3 and -O0<br>
103168 potentially memory leak in coregrind/ume.c <br>
102039 bad permissions for mapped region at address 0xB7C73680<br>
101881 weird assertion problem<br>
101543 Support fadvise64 syscalls<br>
75247 x86_64/amd64 support (the biggest "bug" we have ever fixed)<br>
<br>
(3.0RC1: 27 July 05, vex r1303, valgrind r4283).<br>
(3.0.0: 3 August 05, vex r1313, valgrind r4316).<br>
<br>
<br>
<br>
Stable release 2.4.0 (March 2005) -- CHANGES RELATIVE TO 2.2.0<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
2.4.0 brings many significant changes and bug fixes. The most<br>
significant user-visible change is that we no longer supply our own<br>
pthread implementation. Instead, Valgrind is finally capable of<br>
running the native thread library, either LinuxThreads or NPTL.<br>
<br>
This means our libpthread has gone, along with the bugs associated<br>
with it. Valgrind now supports the kernel's threading syscalls, and<br>
lets you use your standard system libpthread. As a result:<br>
<br>
* There are many fewer system dependencies and strange library-related<br>
bugs. There is a small performance improvement, and a large<br>
stability improvement.<br>
<br>
* On the downside, Valgrind can no longer report misuses of the POSIX<br>
PThreads API. It also means that Helgrind currently does not work.<br>
We hope to fix these problems in a future release.<br>
<br>
Note that running the native thread libraries does not mean Valgrind<br>
is able to provide genuine concurrent execution on SMPs. We still<br>
impose the restriction that only one thread is running at any given<br>
time.<br>
<br>
There are many other significant changes too:<br>
<br>
* Memcheck is (once again) the default tool.<br>
<br>
* The default stack backtrace is now 12 call frames, rather than 4.<br>
<br>
* Suppressions can have up to 25 call frame matches, rather than 4.<br>
<br>
* Memcheck and Addrcheck use less memory. Under some circumstances,<br>
they no longer allocate shadow memory if there are large regions of<br>
memory with the same A/V states - such as an mmaped file.<br>
<br>
* The memory-leak detector in Memcheck and Addrcheck has been<br>
improved. It now reports more types of memory leak, including<br>
leaked cycles. When reporting leaked memory, it can distinguish<br>
between directly leaked memory (memory with no references), and<br>
indirectly leaked memory (memory only referred to by other leaked<br>
memory).<br>
<br>
* Memcheck's confusion over the effect of mprotect() has been fixed:<br>
previously mprotect could erroneously mark undefined data as<br>
defined.<br>
<br>
* Signal handling is much improved and should be very close to what<br>
you get when running natively. <br>
<br>
One result of this is that Valgrind observes changes to sigcontexts<br>
passed to signal handlers. Such modifications will take effect when<br>
the signal returns. You will need to run with --single-step=yes to<br>
make this useful.<br>
<br>
* Valgrind is built in Position Independent Executable (PIE) format if<br>
your toolchain supports it. This allows it to take advantage of all<br>
the available address space on systems with 4Gbyte user address<br>
spaces.<br>
<br>
* Valgrind can now run itself (requires PIE support).<br>
<br>
* Syscall arguments are now checked for validity. Previously all<br>
memory used by syscalls was checked, but now the actual values<br>
passed are also checked.<br>
<br>
* Syscall wrappers are more robust against bad addresses being passed<br>
to syscalls: they will fail with EFAULT rather than killing Valgrind<br>
with SIGSEGV.<br>
<br>
* Because clone() is directly supported, some non-pthread uses of it<br>
will work. Partial sharing (where some resources are shared, and<br>
some are not) is not supported.<br>
<br>
* open() and readlink() on /proc/self/exe are supported.<br>
<br>
BUGS FIXED:<br>
<br>
88520 pipe+fork+dup2 kills the main program<br>
88604 Valgrind Aborts when using $VALGRIND_OPTS and user progra...<br>
88614 valgrind: vg_libpthread.c:2323 (read): Assertion `read_pt...<br>
88703 Stabs parser fails to handle ";"<br>
88886 ioctl wrappers for TIOCMBIS and TIOCMBIC<br>
89032 valgrind pthread_cond_timedwait fails<br>
89106 the 'impossible' happened<br>
89139 Missing sched_setaffinity & sched_getaffinity<br>
89198 valgrind lacks support for SIOCSPGRP and SIOCGPGRP<br>
89263 Missing ioctl translations for scsi-generic and CD playing<br>
89440 tests/deadlock.c line endings<br>
89481 `impossible' happened: EXEC FAILED<br>
89663 valgrind 2.2.0 crash on Redhat 7.2<br>
89792 Report pthread_mutex_lock() deadlocks instead of returnin...<br>
90111 statvfs64 gives invalid error/warning<br>
90128 crash+memory fault with stabs generated by gnat for a run...<br>
90778 VALGRIND_CHECK_DEFINED() not as documented in memcheck.h<br>
90834 cachegrind crashes at end of program without reporting re...<br>
91028 valgrind: vg_memory.c:229 (vgPlain_unmap_range): Assertio...<br>
91162 valgrind crash while debugging drivel 1.2.1<br>
91199 Unimplemented function<br>
91325 Signal routing does not propagate the siginfo structure<br>
91599 Assertion `cv == ((void *)0)'<br>
91604 rw_lookup clears orig and sends the NULL value to rw_new<br>
91821 Small problems building valgrind with $top_builddir ne $t...<br>
91844 signal 11 (SIGSEGV) at get_tcb (libpthread.c:86) in corec...<br>
92264 UNIMPLEMENTED FUNCTION: pthread_condattr_setpshared<br>
92331 per-target flags necessitate AM_PROG_CC_C_O<br>
92420 valgrind doesn't compile with linux 2.6.8.1/9<br>
92513 Valgrind 2.2.0 generates some warning messages<br>
92528 vg_symtab2.c:170 (addLoc): Assertion `loc->size > 0' failed.<br>
93096 unhandled ioctl 0x4B3A and 0x5601<br>
93117 Tool and core interface versions do not match<br>
93128 Can't run valgrind --tool=memcheck because of unimplement...<br>
93174 Valgrind can crash if passed bad args to certain syscalls<br>
93309 Stack frame in new thread is badly aligned<br>
93328 Wrong types used with sys_sigprocmask()<br>
93763 /usr/include/asm/msr.h is missing<br>
93776 valgrind: vg_memory.c:508 (vgPlain_find_map_space): Asser...<br>
93810 fcntl() argument checking a bit too strict<br>
94378 Assertion `tst->sigqueue_head != tst->sigqueue_tail' failed.<br>
94429 valgrind 2.2.0 segfault with mmap64 in glibc 2.3.3<br>
94645 Impossible happened: PINSRW mem<br>
94953 valgrind: the `impossible' happened: SIGSEGV<br>
95667 Valgrind does not work with any KDE app<br>
96243 Assertion 'res==0' failed<br>
96252 stage2 loader of valgrind fails to allocate memory<br>
96520 All programs crashing at _dl_start (in /lib/ld-2.3.3.so) ...<br>
96660 ioctl CDROMREADTOCENTRY causes bogus warnings<br>
96747 After looping in a segfault handler, the impossible happens<br>
96923 Zero sized arrays crash valgrind trace back with SIGFPE<br>
96948 valgrind stops with assertion failure regarding mmap2<br>
96966 valgrind fails when application opens more than 16 sockets<br>
97398 valgrind: vg_libpthread.c:2667 Assertion failed<br>
97407 valgrind: vg_mylibc.c:1226 (vgPlain_safe_fd): Assertion `...<br>
97427 "Warning: invalid file descriptor -1 in syscall close()" ...<br>
97785 missing backtrace<br>
97792 build in obj dir fails - autoconf / makefile cleanup<br>
97880 pthread_mutex_lock fails from shared library (special ker...<br>
97975 program aborts without ang VG messages<br>
98129 Failed when open and close file 230000 times using stdio<br>
98175 Crashes when using valgrind-2.2.0 with a program using al...<br>
98288 Massif broken<br>
98303 UNIMPLEMENTED FUNCTION pthread_condattr_setpshared<br>
98630 failed--compilation missing warnings.pm, fails to make he...<br>
98756 Cannot valgrind signal-heavy kdrive X server<br>
98966 valgrinding the JVM fails with a sanity check assertion<br>
99035 Valgrind crashes while profiling<br>
99142 loops with message "Signal 11 being dropped from thread 0...<br>
99195 threaded apps crash on thread start (using QThread::start...<br>
99348 Assertion `vgPlain_lseek(core_fd, 0, 1) == phdrs[i].p_off...<br>
99568 False negative due to mishandling of mprotect<br>
99738 valgrind memcheck crashes on program that uses sigitimer<br>
99923 0-sized allocations are reported as leaks<br>
99949 program seg faults after exit()<br>
100036 "newSuperblock's request for 1048576 bytes failed"<br>
100116 valgrind: (pthread_cond_init): Assertion `sizeof(* cond) ...<br>
100486 memcheck reports "valgrind: the `impossible' happened: V...<br>
100833 second call to "mremap" fails with EINVAL<br>
101156 (vgPlain_find_map_space): Assertion `(addr & ((1 << 12)-1...<br>
101173 Assertion `recDepth >= 0 && recDepth < 500' failed<br>
101291 creating threads in a forked process fails<br>
101313 valgrind causes different behavior when resizing a window...<br>
101423 segfault for c++ array of floats<br>
101562 valgrind massif dies on SIGINT even with signal handler r...<br>
<br>
<br>
Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.0.0<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
2.2.0 brings nine months worth of improvements and bug fixes. We<br>
believe it to be a worthy successor to 2.0.0. There are literally<br>
hundreds of bug fixes and minor improvements. There are also some<br>
fairly major user-visible changes:<br>
<br>
* A complete overhaul of handling of system calls and signals, and <br>
their interaction with threads. In general, the accuracy of the <br>
system call, thread and signal simulations is much improved:<br>
<br>
- Blocking system calls behave exactly as they do when running<br>
natively (not on valgrind). That is, if a syscall blocks only the<br>
calling thread when running natively, than it behaves the same on<br>
valgrind. No more mysterious hangs because V doesn't know that some<br>
syscall or other, should block only the calling thread.<br>
<br>
- Interrupted syscalls should now give more faithful results.<br>
<br>
- Signal contexts in signal handlers are supported.<br>
<br>
* Improvements to NPTL support to the extent that V now works <br>
properly on NPTL-only setups.<br>
<br>
* Greater isolation between Valgrind and the program being run, so<br>
the program is less likely to inadvertently kill Valgrind by<br>
doing wild writes.<br>
<br>
* Massif: a new space profiling tool. Try it! It's cool, and it'll<br>
tell you in detail where and when your C/C++ code is allocating heap.<br>
Draws pretty .ps pictures of memory use against time. A potentially<br>
powerful tool for making sense of your program's space use.<br>
<br>
* File descriptor leakage checks. When enabled, Valgrind will print out<br>
a list of open file descriptors on exit.<br>
<br>
* Improved SSE2/SSE3 support.<br>
<br>
* Time-stamped output; use --time-stamp=yes<br>
<br>
<br>
<br>
Stable release 2.2.0 (31 August 2004) -- CHANGES RELATIVE TO 2.1.2<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
2.2.0 is not much different from 2.1.2, released seven weeks ago.<br>
A number of bugs have been fixed, most notably #85658, which gave<br>
problems for quite a few people. There have been many internal<br>
cleanups, but those are not user visible.<br>
<br>
The following bugs have been fixed since 2.1.2:<br>
<br>
85658 Assert in coregrind/vg_libpthread.c:2326 (open64) !=<br>
(void*)0 failed<br>
This bug was reported multiple times, and so the following<br>
duplicates of it are also fixed: 87620, 85796, 85935, 86065, <br>
86919, 86988, 87917, 88156<br>
<br>
80716 Semaphore mapping bug caused by unmap (sem_destroy)<br>
(Was fixed prior to 2.1.2)<br>
<br>
86987 semctl and shmctl syscalls family is not handled properly<br>
<br>
86696 valgrind 2.1.2 + RH AS2.1 + librt<br>
<br>
86730 valgrind locks up at end of run with assertion failure <br>
in __pthread_unwind<br>
<br>
86641 memcheck doesn't work with Mesa OpenGL/ATI on Suse 9.1<br>
(also fixes 74298, a duplicate of this)<br>
<br>
85947 MMX/SSE unhandled instruction 'sfence'<br>
<br>
84978 Wrong error "Conditional jump or move depends on<br>
uninitialised value" resulting from "sbbl %reg, %reg"<br>
<br>
86254 ssort() fails when signed int return type from comparison is <br>
too small to handle result of unsigned int subtraction<br>
<br>
87089 memalign( 4, xxx) makes valgrind assert<br>
<br>
86407 Add support for low-level parallel port driver ioctls.<br>
<br>
70587 Add timestamps to Valgrind output? (wishlist)<br>
<br>
84937 vg_libpthread.c:2505 (se_remap): Assertion `res == 0'<br>
(fixed prior to 2.1.2)<br>
<br>
86317 cannot load libSDL-1.2.so.0 using valgrind<br>
<br>
86989 memcpy from mac_replace_strmem.c complains about<br>
uninitialized pointers passed when length to copy is zero<br>
<br>
85811 gnu pascal symbol causes segmentation fault; ok in 2.0.0<br>
<br>
79138 writing to sbrk()'d memory causes segfault<br>
<br>
77369 sched deadlock while signal received during pthread_join<br>
and the joined thread exited<br>
<br>
88115 In signal handler for SIGFPE, siginfo->si_addr is wrong <br>
under Valgrind<br>
<br>
78765 Massif crashes on app exit if FP exceptions are enabled<br>
<br>
Additionally there are the following changes, which are not <br>
connected to any bug report numbers, AFAICS:<br>
<br>
* Fix scary bug causing mis-identification of SSE stores vs<br>
loads and so causing memcheck to sometimes give nonsense results<br>
on SSE code.<br>
<br>
* Add support for the POSIX message queue system calls.<br>
<br>
* Fix to allow 32-bit Valgrind to run on AMD64 boxes. Note: this does<br>
NOT allow Valgrind to work with 64-bit executables - only with 32-bit<br>
executables on an AMD64 box.<br>
<br>
* At configure time, only check whether linux/mii.h can be processed <br>
so that we don't generate ugly warnings by trying to compile it.<br>
<br>
* Add support for POSIX clocks and timers.<br>
<br>
<br>
<br>
Developer (cvs head) release 2.1.2 (18 July 2004)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
2.1.2 contains four months worth of bug fixes and refinements.<br>
Although officially a developer release, we believe it to be stable<br>
enough for widespread day-to-day use. 2.1.2 is pretty good, so try it<br>
first, although there is a chance it won't work. If so then try 2.0.0<br>
and tell us what went wrong." 2.1.2 fixes a lot of problems present<br>
in 2.0.0 and is generally a much better product.<br>
<br>
Relative to 2.1.1, a large number of minor problems with 2.1.1 have<br>
been fixed, and so if you use 2.1.1 you should try 2.1.2. Users of<br>
the last stable release, 2.0.0, might also want to try this release.<br>
<br>
The following bugs, and probably many more, have been fixed. These<br>
are listed at http://bugs.kde.org. Reporting a bug for valgrind in<br>
the http://bugs.kde.org is much more likely to get you a fix than<br>
mailing developers directly, so please continue to keep sending bugs<br>
there.<br>
<br>
76869 Crashes when running any tool under Fedora Core 2 test1<br>
This fixes the problem with returning from a signal handler <br>
when VDSOs are turned off in FC2.<br>
<br>
69508 java 1.4.2 client fails with erroneous "stack size too small".<br>
This fix makes more of the pthread stack attribute related <br>
functions work properly. Java still doesn't work though.<br>
<br>
71906 malloc alignment should be 8, not 4<br>
All memory returned by malloc/new etc is now at least<br>
8-byte aligned.<br>
<br>
81970 vg_alloc_ThreadState: no free slots available<br>
(closed because the workaround is simple: increase<br>
VG_N_THREADS, rebuild and try again.)<br>
<br>
78514 Conditional jump or move depends on uninitialized value(s)<br>
(a slight mishanding of FP code in memcheck)<br>
<br>
77952 pThread Support (crash) (due to initialisation-ordering probs)<br>
(also 85118)<br>
<br>
80942 Addrcheck wasn't doing overlap checking as it should.<br>
78048 return NULL on malloc/new etc failure, instead of asserting<br>
73655 operator new() override in user .so files often doesn't get picked up<br>
83060 Valgrind does not handle native kernel AIO<br>
69872 Create proper coredumps after fatal signals<br>
82026 failure with new glibc versions: __libc_* functions are not exported<br>
70344 UNIMPLEMENTED FUNCTION: tcdrain <br>
81297 Cancellation of pthread_cond_wait does not require mutex<br>
82872 Using debug info from additional packages (wishlist)<br>
83025 Support for ioctls FIGETBSZ and FIBMAP<br>
83340 Support for ioctl HDIO_GET_IDENTITY<br>
79714 Support for the semtimedop system call.<br>
77022 Support for ioctls FBIOGET_VSCREENINFO and FBIOGET_FSCREENINFO<br>
82098 hp2ps ansification (wishlist)<br>
83573 Valgrind SIGSEGV on execve<br>
82999 show which cmdline option was erroneous (wishlist)<br>
83040 make valgrind VPATH and distcheck-clean (wishlist)<br>
83998 Assertion `newfd > vgPlain_max_fd' failed (see below)<br>
82722 Unchecked mmap in as_pad leads to mysterious failures later<br>
78958 memcheck seg faults while running Mozilla <br>
85416 Arguments with colon (e.g. --logsocket) ignored<br>
<br>
<br>
Additionally there are the following changes, which are not <br>
connected to any bug report numbers, AFAICS:<br>
<br>
* Rearranged address space layout relative to 2.1.1, so that<br>
Valgrind/tools will run out of memory later than currently in many<br>
circumstances. This is good news esp. for Calltree. It should<br>
be possible for client programs to allocate over 800MB of<br>
memory when using memcheck now.<br>
<br>
* Improved checking when laying out memory. Should hopefully avoid<br>
the random segmentation faults that 2.1.1 sometimes caused.<br>
<br>
* Support for Fedora Core 2 and SuSE 9.1. Improvements to NPTL<br>
support to the extent that V now works properly on NPTL-only setups.<br>
<br>
* Renamed the following options:<br>
--logfile-fd --> --log-fd<br>
--logfile --> --log-file<br>
--logsocket --> --log-socket<br>
to be consistent with each other and other options (esp. --input-fd).<br>
<br>
* Add support for SIOCGMIIPHY, SIOCGMIIREG and SIOCSMIIREG ioctls and<br>
improve the checking of other interface related ioctls.<br>
<br>
* Fix building with gcc-3.4.1.<br>
<br>
* Remove limit on number of semaphores supported.<br>
<br>
* Add support for syscalls: set_tid_address (258), acct (51).<br>
<br>
* Support instruction "repne movs" -- not official but seems to occur.<br>
<br>
* Implement an emulated soft limit for file descriptors in addition to<br>
the current reserved area, which effectively acts as a hard limit. The<br>
setrlimit system call now simply updates the emulated limits as best<br>
as possible - the hard limit is not allowed to move at all and just<br>
returns EPERM if you try and change it. This should stop reductions<br>
in the soft limit causing assertions when valgrind tries to allocate<br>
descriptors from the reserved area.<br>
(This actually came from bug #83998).<br>
<br>
* Major overhaul of Cachegrind implementation. First user-visible change<br>
is that cachegrind.out files are now typically 90% smaller than they<br>
used to be; code annotation times are correspondingly much smaller.<br>
Second user-visible change is that hit/miss counts for code that is<br>
unloaded at run-time is no longer dumped into a single "discard" pile,<br>
but accurately preserved.<br>
<br>
* Client requests for telling valgrind about memory pools.<br>
<br>
<br>
<br>
Developer (cvs head) release 2.1.1 (12 March 2004)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
2.1.1 contains some internal structural changes needed for V's<br>
long-term future. These don't affect end-users. Most notable<br>
user-visible changes are:<br>
<br>
* Greater isolation between Valgrind and the program being run, so<br>
the program is less likely to inadvertently kill Valgrind by<br>
doing wild writes.<br>
<br>
* Massif: a new space profiling tool. Try it! It's cool, and it'll<br>
tell you in detail where and when your C/C++ code is allocating heap.<br>
Draws pretty .ps pictures of memory use against time. A potentially<br>
powerful tool for making sense of your program's space use.<br>
<br>
* Fixes for many bugs, including support for more SSE2/SSE3 instructions,<br>
various signal/syscall things, and various problems with debug<br>
info readers.<br>
<br>
* Support for glibc-2.3.3 based systems.<br>
<br>
We are now doing automatic overnight build-and-test runs on a variety<br>
of distros. As a result, we believe 2.1.1 builds and runs on:<br>
Red Hat 7.2, 7.3, 8.0, 9, Fedora Core 1, SuSE 8.2, SuSE 9.<br>
<br>
<br>
The following bugs, and probably many more, have been fixed. These<br>
are listed at http://bugs.kde.org. Reporting a bug for valgrind in<br>
the http://bugs.kde.org is much more likely to get you a fix than<br>
mailing developers directly, so please continue to keep sending bugs<br>
there.<br>
<br>
69616 glibc 2.3.2 w/NPTL is massively different than what valgrind expects <br>
69856 I don't know how to instrument MMXish stuff (Helgrind)<br>
73892 valgrind segfaults starting with Objective-C debug info <br>
(fix for S-type stabs)<br>
73145 Valgrind complains too much about close(<reserved fd>) <br>
73902 Shadow memory allocation seems to fail on RedHat 8.0 <br>
68633 VG_N_SEMAPHORES too low (V itself was leaking semaphores)<br>
75099 impossible to trace multiprocess programs <br>
76839 the `impossible' happened: disInstr: INT but not 0x80 ! <br>
76762 vg_to_ucode.c:3748 (dis_push_segreg): Assertion `sz == 4' failed. <br>
76747 cannot include valgrind.h in c++ program <br>
76223 parsing B(3,10) gave NULL type => impossible happens <br>
75604 shmdt handling problem <br>
76416 Problems with gcc 3.4 snap 20040225 <br>
75614 using -gstabs when building your programs the `impossible' happened<br>
75787 Patch for some CDROM ioctls CDORM_GET_MCN, CDROM_SEND_PACKET,<br>
75294 gcc 3.4 snapshot's libstdc++ have unsupported instructions. <br>
(REP RET)<br>
73326 vg_symtab2.c:272 (addScopeRange): Assertion `range->size > 0' failed. <br>
72596 not recognizing __libc_malloc <br>
69489 Would like to attach ddd to running program <br>
72781 Cachegrind crashes with kde programs <br>
73055 Illegal operand at DXTCV11CompressBlockSSE2 (more SSE opcodes)<br>
73026 Descriptor leak check reports port numbers wrongly <br>
71705 README_MISSING_SYSCALL_OR_IOCTL out of date <br>
72643 Improve support for SSE/SSE2 instructions <br>
72484 valgrind leaves it's own signal mask in place when execing <br>
72650 Signal Handling always seems to restart system calls <br>
72006 The mmap system call turns all errors in ENOMEM <br>
71781 gdb attach is pretty useless <br>
71180 unhandled instruction bytes: 0xF 0xAE 0x85 0xE8 <br>
69886 writes to zero page cause valgrind to assert on exit <br>
71791 crash when valgrinding gimp 1.3 (stabs reader problem)<br>
69783 unhandled syscall: 218 <br>
69782 unhandled instruction bytes: 0x66 0xF 0x2B 0x80 <br>
70385 valgrind fails if the soft file descriptor limit is less <br>
than about 828<br>
69529 "rep; nop" should do a yield <br>
70827 programs with lots of shared libraries report "mmap failed" <br>
for some of them when reading symbols <br>
71028 glibc's strnlen is optimised enough to confuse valgrind <br>
<br>
<br>
<br>
<br>
Unstable (cvs head) release 2.1.0 (15 December 2003)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
For whatever it's worth, 2.1.0 actually seems pretty darn stable to me<br>
(Julian). It looks eminently usable, and given that it fixes some<br>
significant bugs, may well be worth using on a day-to-day basis.<br>
2.1.0 is known to build and pass regression tests on: SuSE 9, SuSE<br>
8.2, RedHat 8.<br>
<br>
2.1.0 most notably includes Jeremy Fitzhardinge's complete overhaul of<br>
handling of system calls and signals, and their interaction with<br>
threads. In general, the accuracy of the system call, thread and<br>
signal simulations is much improved. Specifically:<br>
<br>
- Blocking system calls behave exactly as they do when running<br>
natively (not on valgrind). That is, if a syscall blocks only the<br>
calling thread when running natively, than it behaves the same on<br>
valgrind. No more mysterious hangs because V doesn't know that some<br>
syscall or other, should block only the calling thread.<br>
<br>
- Interrupted syscalls should now give more faithful results.<br>
<br>
- Finally, signal contexts in signal handlers are supported. As a<br>
result, konqueror on SuSE 9 no longer segfaults when notified of<br>
file changes in directories it is watching.<br>
<br>
Other changes:<br>
<br>
- Robert Walsh's file descriptor leakage checks. When enabled,<br>
Valgrind will print out a list of open file descriptors on<br>
exit. Along with each file descriptor, Valgrind prints out a stack<br>
backtrace of where the file was opened and any details relating to the<br>
file descriptor such as the file name or socket details.<br>
To use, give: --track-fds=yes<br>
<br>
- Implemented a few more SSE/SSE2 instructions.<br>
<br>
- Less crud on the stack when you do 'where' inside a GDB attach.<br>
<br>
- Fixed the following bugs:<br>
68360: Valgrind does not compile against 2.6.0-testX kernels<br>
68525: CVS head doesn't compile on C90 compilers<br>
68566: pkgconfig support (wishlist)<br>
68588: Assertion `sz == 4' failed in vg_to_ucode.c (disInstr)<br>
69140: valgrind not able to explicitly specify a path to a binary. <br>
69432: helgrind asserts encountering a MutexErr when there are <br>
EraserErr suppressions<br>
<br>
- Increase the max size of the translation cache from 200k average bbs<br>
to 300k average bbs. Programs on the size of OOo (680m17) are<br>
thrashing the cache at the smaller size, creating large numbers of<br>
retranslations and wasting significant time as a result.<br>
<br>
<br>
<br>
Stable release 2.0.0 (5 Nov 2003)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
<br>
2.0.0 improves SSE/SSE2 support, fixes some minor bugs, and<br>
improves support for SuSE 9 and the Red Hat "Severn" beta.<br>
<br>
- Further improvements to SSE/SSE2 support. The entire test suite of<br>
the GNU Scientific Library (gsl-1.4) compiled with Intel Icc 7.1<br>
20030307Z '-g -O -xW' now works. I think this gives pretty good<br>
coverage of SSE/SSE2 floating point instructions, or at least the<br>
subset emitted by Icc.<br>
<br>
- Also added support for the following instructions:<br>
MOVNTDQ UCOMISD UNPCKLPS UNPCKHPS SQRTSS<br>
PUSH/POP %{FS,GS}, and PUSH %CS (Nb: there is no POP %CS).<br>
<br>
- CFI support for GDB version 6. Needed to enable newer GDBs<br>
to figure out where they are when using --gdb-attach=yes.<br>
<br>
- Fix this:<br>
mc_translate.c:1091 (memcheck_instrument): Assertion<br>
`u_in->size == 4 || u_in->size == 16' failed.<br>
<br>
- Return an error rather than panicing when given a bad socketcall.<br>
<br>
- Fix checking of syscall rt_sigtimedwait().<br>
<br>
- Implement __NR_clock_gettime (syscall 265). Needed on Red Hat Severn.<br>
<br>
- Fixed bug in overlap check in strncpy() -- it was assuming the src was 'n'<br>
bytes long, when it could be shorter, which could cause false<br>
positives.<br>
<br>
- Support use of select() for very large numbers of file descriptors.<br>
<br>
- Don't fail silently if the executable is statically linked, or is<br>
setuid/setgid. Print an error message instead.<br>
<br>
- Support for old DWARF-1 format line number info.<br>
<br>
<br>
<br>
Snapshot 20031012 (12 October 2003)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
<br>
Three months worth of bug fixes, roughly. Most significant single<br>
change is improved SSE/SSE2 support, mostly thanks to Dirk Mueller.<br>
<br>
20031012 builds on Red Hat Fedora ("Severn") but doesn't really work<br>
(curiosly, mozilla runs OK, but a modest "ls -l" bombs). I hope to<br>
get a working version out soon. It may or may not work ok on the<br>
forthcoming SuSE 9; I hear positive noises about it but haven't been<br>
able to verify this myself (not until I get hold of a copy of 9).<br>
<br>
A detailed list of changes, in no particular order:<br>
<br>
- Describe --gen-suppressions in the FAQ.<br>
<br>
- Syscall __NR_waitpid supported.<br>
<br>
- Minor MMX bug fix.<br>
<br>
- -v prints program's argv[] at startup.<br>
<br>
- More glibc-2.3 suppressions.<br>
<br>
- Suppressions for stack underrun bug(s) in the c++ support library<br>
distributed with Intel Icc 7.0.<br>
<br>
- Fix problems reading /proc/self/maps.<br>
<br>
- Fix a couple of messages that should have been suppressed by -q, <br>
but weren't.<br>
<br>
- Make Addrcheck understand "Overlap" suppressions.<br>
<br>
- At startup, check if program is statically linked and bail out if so.<br>
<br>
- Cachegrind: Auto-detect Intel Pentium-M, also VIA Nehemiah<br>
<br>
- Memcheck/addrcheck: minor speed optimisations<br>
<br>
- Handle syscall __NR_brk more correctly than before.<br>
<br>
- Fixed incorrect allocate/free mismatch errors when using<br>
operator new(unsigned, std::nothrow_t const&)<br>
operator new[](unsigned, std::nothrow_t const&)<br>
<br>
- Support POSIX pthread spinlocks.<br>
<br>
- Fixups for clean compilation with gcc-3.3.1.<br>
<br>
- Implemented more opcodes: <br>
- push %es<br>
- push %ds<br>
- pop %es<br>
- pop %ds<br>
- movntq<br>
- sfence<br>
- pshufw<br>
- pavgb<br>
- ucomiss<br>
- enter<br>
- mov imm32, %esp<br>
- all "in" and "out" opcodes<br>
- inc/dec %esp<br>
- A whole bunch of SSE/SSE2 instructions<br>
<br>
- Memcheck: don't bomb on SSE/SSE2 code.<br>
<br>
<br>
Snapshot 20030725 (25 July 2003)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
<br>
Fixes some minor problems in 20030716.<br>
<br>
- Fix bugs in overlap checking for strcpy/memcpy etc.<br>
<br>
- Do overlap checking with Addrcheck as well as Memcheck.<br>
<br>
- Fix this:<br>
Memcheck: the `impossible' happened:<br>
get_error_name: unexpected type<br>
<br>
- Install headers needed to compile new skins.<br>
<br>
- Remove leading spaces and colon in the LD_LIBRARY_PATH / LD_PRELOAD<br>
passed to non-traced children.<br>
<br>
- Fix file descriptor leak in valgrind-listener.<br>
<br>
- Fix longstanding bug in which the allocation point of a <br>
block resized by realloc was not correctly set. This may<br>
have caused confusing error messages.<br>
<br>
<br>
Snapshot 20030716 (16 July 2003)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
<br>
20030716 is a snapshot of our current CVS head (development) branch.<br>
This is the branch which will become valgrind-2.0. It contains<br>
significant enhancements over the 1.9.X branch.<br>
<br>
Despite this being a snapshot of the CVS head, it is believed to be<br>
quite stable -- at least as stable as 1.9.6 or 1.0.4, if not more so<br>
-- and therefore suitable for widespread use. Please let us know asap<br>
if it causes problems for you.<br>
<br>
Two reasons for releasing a snapshot now are:<br>
<br>
- It's been a while since 1.9.6, and this snapshot fixes<br>
various problems that 1.9.6 has with threaded programs <br>
on glibc-2.3.X based systems.<br>
<br>
- So as to make available improvements in the 2.0 line.<br>
<br>
Major changes in 20030716, as compared to 1.9.6:<br>
<br>
- More fixes to threading support on glibc-2.3.1 and 2.3.2-based<br>
systems (SuSE 8.2, Red Hat 9). If you have had problems<br>
with inconsistent/illogical behaviour of errno, h_errno or the DNS<br>
resolver functions in threaded programs, 20030716 should improve<br>
matters. This snapshot seems stable enough to run OpenOffice.org<br>
1.1rc on Red Hat 7.3, SuSE 8.2 and Red Hat 9, and that's a big<br>
threaded app if ever I saw one.<br>
<br>
- Automatic generation of suppression records; you no longer<br>
need to write them by hand. Use --gen-suppressions=yes.<br>
<br>
- strcpy/memcpy/etc check their arguments for overlaps, when<br>
running with the Memcheck or Addrcheck skins.<br>
<br>
- malloc_usable_size() is now supported.<br>
<br>
- new client requests:<br>
- VALGRIND_COUNT_ERRORS, VALGRIND_COUNT_LEAKS: <br>
useful with regression testing<br>
- VALGRIND_NON_SIMD_CALL[0123]: for running arbitrary functions <br>
on real CPU (use with caution!)<br>
<br>
- The GDB attach mechanism is more flexible. Allow the GDB to<br>
be run to be specified by --gdb-path=/path/to/gdb, and specify<br>
which file descriptor V will read its input from with<br>
--input-fd=<number>.<br>
<br>
- Cachegrind gives more accurate results (wasn't tracking instructions in<br>
malloc() and friends previously, is now).<br>
<br>
- Complete support for the MMX instruction set.<br>
<br>
- Partial support for the SSE and SSE2 instruction sets. Work for this<br>
is ongoing. About half the SSE/SSE2 instructions are done, so<br>
some SSE based programs may work. Currently you need to specify<br>
--skin=addrcheck. Basically not suitable for real use yet.<br>
<br>
- Significant speedups (10%-20%) for standard memory checking.<br>
<br>
- Fix assertion failure in pthread_once().<br>
<br>
- Fix this:<br>
valgrind: vg_intercept.c:598 (vgAllRoadsLeadToRome_select): <br>
Assertion `ms_end >= ms_now' failed.<br>
<br>
- Implement pthread_mutexattr_setpshared.<br>
<br>
- Understand Pentium 4 branch hints. Also implemented a couple more<br>
obscure x86 instructions.<br>
<br>
- Lots of other minor bug fixes.<br>
<br>
- We have a decent regression test system, for the first time.<br>
This doesn't help you directly, but it does make it a lot easier<br>
for us to track the quality of the system, especially across<br>
multiple linux distributions. <br>
<br>
You can run the regression tests with 'make regtest' after 'make<br>
install' completes. On SuSE 8.2 and Red Hat 9 I get this:<br>
<br>
== 84 tests, 0 stderr failures, 0 stdout failures ==<br>
<br>
On Red Hat 8, I get this:<br>
<br>
== 84 tests, 2 stderr failures, 1 stdout failure ==<br>
corecheck/tests/res_search (stdout)<br>
memcheck/tests/sigaltstack (stderr)<br>
<br>
sigaltstack is probably harmless. res_search doesn't work<br>
on R H 8 even running natively, so I'm not too worried. <br>
<br>
On Red Hat 7.3, a glibc-2.2.5 system, I get these harmless failures:<br>
<br>
== 84 tests, 2 stderr failures, 1 stdout failure ==<br>
corecheck/tests/pth_atfork1 (stdout)<br>
corecheck/tests/pth_atfork1 (stderr)<br>
memcheck/tests/sigaltstack (stderr)<br>
<br>
You need to run on a PII system, at least, since some tests<br>
contain P6-specific instructions, and the test machine needs<br>
access to the internet so that corecheck/tests/res_search<br>
(a test that the DNS resolver works) can function.<br>
<br>
As ever, thanks for the vast amount of feedback :) and bug reports :(<br>
We may not answer all messages, but we do at least look at all of<br>
them, and tend to fix the most frequently reported bugs.<br>
<br>
<br>
<br>
Version 1.9.6 (7 May 2003 or thereabouts)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
<br>
Major changes in 1.9.6:<br>
<br>
- Improved threading support for glibc >= 2.3.2 (SuSE 8.2,<br>
RedHat 9, to name but two ...) It turned out that 1.9.5<br>
had problems with threading support on glibc >= 2.3.2,<br>
usually manifested by threaded programs deadlocking in system calls,<br>
or running unbelievably slowly. Hopefully these are fixed now. 1.9.6<br>
is the first valgrind which gives reasonable support for<br>
glibc-2.3.2. Also fixed a 2.3.2 problem with pthread_atfork().<br>
<br>
- Majorly expanded FAQ.txt. We've added workarounds for all<br>
common problems for which a workaround is known.<br>
<br>
Minor changes in 1.9.6:<br>
<br>
- Fix identification of the main thread's stack. Incorrect<br>
identification of it was causing some on-stack addresses to not get<br>
identified as such. This only affected the usefulness of some error<br>
messages; the correctness of the checks made is unchanged.<br>
<br>
- Support for kernels >= 2.5.68.<br>
<br>
- Dummy implementations of __libc_current_sigrtmin, <br>
__libc_current_sigrtmax and __libc_allocate_rtsig, hopefully<br>
good enough to keep alive programs which previously died for lack of<br>
them.<br>
<br>
- Fix bug in the VALGRIND_DISCARD_TRANSLATIONS client request.<br>
<br>
- Fix bug in the DWARF2 debug line info loader, when instructions <br>
following each other have source lines far from each other <br>
(e.g. with inlined functions).<br>
<br>
- Debug info reading: read symbols from both "symtab" and "dynsym"<br>
sections, rather than merely from the one that comes last in the<br>
file.<br>
<br>
- New syscall support: prctl(), creat(), lookup_dcookie().<br>
<br>
- When checking calls to accept(), recvfrom(), getsocketopt(),<br>
don't complain if buffer values are NULL.<br>
<br>
- Try and avoid assertion failures in<br>
mash_LD_PRELOAD_and_LD_LIBRARY_PATH.<br>
<br>
- Minor bug fixes in cg_annotate.<br>
<br>
<br>
<br>
Version 1.9.5 (7 April 2003)<br>
~~~~~~~~~~~~~~~~~~~~~~~~~~~~<br>
<br>
It occurs to me that it would be helpful for valgrind users to record<br>
in the source distribution the changes in each release. So I now<br>
attempt to mend my errant ways :-) Changes in this and future releases<br>
will be documented in the NEWS file in the source distribution.<br>
<br>
Major changes in 1.9.5:<br>
<br>
- (Critical bug fix): Fix a bug in the FPU simulation. This was<br>
causing some floating point conditional tests not to work right.<br>
Several people reported this. If you had floating point code which<br>
didn't work right on 1.9.1 to 1.9.4, it's worth trying 1.9.5.<br>
<br>
- Partial support for Red Hat 9. RH9 uses the new Native Posix <br>
Threads Library (NPTL), instead of the older LinuxThreads. <br>
This potentially causes problems with V which will take some<br>
time to correct. In the meantime we have partially worked around<br>
this, and so 1.9.5 works on RH9. Threaded programs still work,<br>
but they may deadlock, because some system calls (accept, read,<br>
write, etc) which should be nonblocking, in fact do block. This<br>
is a known bug which we are looking into.<br>
<br>
If you can, your best bet (unfortunately) is to avoid using <br>
1.9.5 on a Red Hat 9 system, or on any NPTL-based distribution.<br>
If your glibc is 2.3.1 or earlier, you're almost certainly OK.<br>
<br>
Minor changes in 1.9.5:<br>
<br>
- Added some #errors to valgrind.h to ensure people don't include<br>
it accidentally in their sources. This is a change from 1.0.X<br>
which was never properly documented. The right thing to include<br>
is now memcheck.h. Some people reported problems and strange<br>
behaviour when (incorrectly) including valgrind.h in code with <br>
1.9.1 -- 1.9.4. This is no longer possible.<br>
<br>
- Add some __extension__ bits and pieces so that gcc configured<br>
for valgrind-checking compiles even with -Werror. If you<br>
don't understand this, ignore it. Of interest to gcc developers<br>
only.<br>
<br>
- Removed a pointless check which caused problems interworking <br>
with Clearcase. V would complain about shared objects whose<br>
names did not end ".so", and refuse to run. This is now fixed.<br>
In fact it was fixed in 1.9.4 but not documented.<br>
<br>
- Fixed a bug causing an assertion failure of "waiters == 1"<br>
somewhere in vg_scheduler.c, when running large threaded apps,<br>
notably MySQL.<br>
<br>
- Add support for the munlock system call (124).<br>
<br>
Some comments about future releases:<br>
<br>
1.9.5 is, we hope, the most stable Valgrind so far. It pretty much<br>
supersedes the 1.0.X branch. If you are a valgrind packager, please<br>
consider making 1.9.5 available to your users. You can regard the<br>
1.0.X branch as obsolete: 1.9.5 is stable and vastly superior. There<br>
are no plans at all for further releases of the 1.0.X branch.<br>
<br>
If you want a leading-edge valgrind, consider building the cvs head<br>
(from SourceForge), or getting a snapshot of it. Current cool stuff<br>
going in includes MMX support (done); SSE/SSE2 support (in progress),<br>
a significant (10-20%) performance improvement (done), and the usual<br>
large collection of minor changes. Hopefully we will be able to<br>
improve our NPTL support, but no promises.<br>
<br>
<br>
</p></div>
</div>
<div>
<br><table class="nav" width="100%" cellspacing="3" cellpadding="2" border="0" summary="Navigation footer">
<tr>
<td rowspan="2" width="40%" align="left">
<a accesskey="p" href="dist.authors.html"><< 1. AUTHORS</a> </td>
<td width="20%" align="center"><a accesskey="u" href="dist.html">Up</a></td>
<td rowspan="2" width="40%" align="right"> <a accesskey="n" href="dist.readme.html">3. README >></a>
</td>
</tr>
<tr><td width="20%" align="center"><a accesskey="h" href="index.html">Home</a></td></tr>
</table>
</div>
</body>
</html>