|
From: Elijah P N. <ne...@ma...> - 2004-06-09 16:15:14
|
Hi, I would like to get valgrind to run under Fedora Core 2 (uses kernel-2.6.5, glibc-2.3.3, and gcc-3.3.3), but now I'm wondering whether it's possible. Is it? I thought it should be possible since Dag Wieers packaged it (http://dag.wieers.com/home-made/apt/) for that. I have also found in my search some bugs in bugzilla and emails in the archives that seem to imply others are trying to work with valgrind on distros that use the 2.6 kernel. However, I'm getting several problems: 1. Using the version packaged by Dag Wieers (valgrind-2.1.1) and running 'valgrind --tool=memcheck ls' all I see is 'Segmentation fault'. 2. So, I try downloading CVS HEAD. I build it and get a segfault when trying to run valgrind on ls just as with Dag's package. 3. So, I download the 2.0.0 tarball and build it. Trying to run valgrind on ls gives me the bug described at http://bugs.kde.org/show_bug.cgi?id=82026. 4. So, I try out the VALGRIND_2_0_BRANCH in CVS. When trying to configure, I get the response, 'configure: error: Valgrind works on kernels 2.2 and 2.4'. Until the last step above, I hadn't found anything that specifically says that valgrind doesn't work on a 2.6 kernel. Now I'm wondering if that's my problem, though. Can anyone shed some light on this? I seem to be coming up blank in my searches through bugzilla and my searches of the email list archives as to what I can try. If it should be possible, what information could I provide to help pinpoint the problem? I can get valgrind to work just fine on the older systems (running the 2.4 kernel) I've tried, both using packages from Dag or compiling from tarballs. But those machines are really slow and don't have as much memory, among other problems... Thanks in advance for any pointers, Elijah P.S. Output of 'uname -a': 'Linux amr.math.utah.edu 2.6.5-1.358smp #1 SMP Sat May 8 09:25:36 EDT 2004 i686 athlon i386 GNU/Linux'. Output of 'valgrind -v' is identical to that of 'valgrind --help'. |
|
From: Tom H. <th...@cy...> - 2004-06-09 16:41:37
|
In message <108...@am...>
Elijah P Newren <ne...@ma...> wrote:
> 2. So, I try downloading CVS HEAD. I build it and get a segfault when
> trying to run valgrind on ls just as with Dag's package.
This should work find so long as you turn off vdso support, which you
can do with this command:
sysctl -w kernel.vdso=0
Tom
--
Tom Hughes (th...@cy...)
Software Engineer, Cyberscience Corporation
http://www.cyberscience.com/
|
|
From: Elijah P N. <ne...@ma...> - 2004-06-09 17:14:48
|
On Wed, 2004-06-09 at 10:41, Tom Hughes wrote: > In message <108...@am...> > Elijah P Newren <ne...@ma...> wrote: > > > 2. So, I try downloading CVS HEAD. I build it and get a segfault when > > trying to run valgrind on ls just as with Dag's package. > > This should work find so long as you turn off vdso support, which you > can do with this command: > > sysctl -w kernel.vdso=0 Works like a charm. Thanks! Elijah |
|
From: David H. <dw...@ov...> - 2004-11-28 00:15:45
|
Hi, I was valgrind'ing some C++ classes that use vectors and vector <float> data; data.resize(10); Reports 'still reachable' memory. FAQ 4.3 offers a reasonable explanation. If I wanted to suppress such a warning, whats the usual technique (I haven't had to use suppressions other than the defaults)? Thanks, Dave Hawkins Caltech. P.S. To the Valgrind developers: Thanks for developing such a great tool. I now take the time to write Linux-based test cases for my non-x86 destined C code and run it through Valgrind as a sanity check. |