[lxml-dev] is valgrind still used?
Stefan Behnel
behnel_ml at gkec.informatik.tu-darmstadt.de
Fri Feb 17 12:21:45 CET 2006
Victor Ng wrote:
> Is valgrind still being used anymore? I just ran :
>
>> valgrind --tool=memcheck --suppressions=valgrind-python.supp python
> test.py
>
> and got:
>
> ==13242==
> ==13242== ERROR SUMMARY: 138 errors from 16 contexts (suppressed: 10026
> from 6)
> ==13242== malloc/free: in use at exit: 2848232 bytes in 2981 blocks.
> ==13242== malloc/free: 53700 allocs, 50719 frees, 21000875 bytes allocated.
> ==13242== For counts of detected errors, rerun with: -v
> ==13242== searching for pointers to 2981 not-freed blocks.
> ==13242== checked 3490012 bytes.
> ==13242==
> ==13242== LEAK SUMMARY:
> ==13242== definitely lost: 4759 bytes in 168 blocks.
> ==13242== possibly lost: 0 bytes in 0 blocks.
> ==13242== still reachable: 2843473 bytes in 2813 blocks.
> ==13242== suppressed: 0 bytes in 0 blocks.
> ==13242== Use --leak-check=full to see details of leaked memory.
> victorng at mbdev1:~/dev/lxml$
Thanks for testing. However, there have been a number of fixes for memory
leaks applied to the latest SVN (I applied the last one today), which
dramatically reduce the number of lost blocks. It's not zero, though, but it
looks like the remaining problems are more related to Python itself.
You didn't say which version you are running (0.8, I assume), but since there
were many, many changes to SVN since the last version, you should really use
the current SVN version when testing. I currently get this result for the
trunk version:
==24747== LEAK SUMMARY:
==24747== definitely lost: 7,760 bytes in 121 blocks.
==24747== possibly lost: 16,320 bytes in 60 blocks.
==24747== still reachable: 4,521,357 bytes in 3,331 blocks.
==24747== suppressed: 0 bytes in 0 blocks.
> Is this expected? Is there a Trac instance or some other bug tracker
> that I can look at to see what the open bugs are?
No, there is no trac support currently. Bugs are commonly reported to the
mailing list, just as you did.
Anyway, thanks for testing. Narayan Desai has just reported (and fixed) a few
leaks this week, so, if you find anything else that should be done, I'd be
glad to hear about it.
Stefan
More information about the lxml-dev
mailing list