Hello, I've worked on a patch to correct the memory leak in invoke() function. The patch has been submitted, see: https://bugzilla.gnome.org/show_bug.cgi?id=630271
Le 03/09/2010, Tomeu Vizoso <[email protected]> a écrit : > Hi Damien, we should certainly be doing something like that. I think > that if make check and make check.valgrind do pass without warnings, > your patch is good to push. The make check is fine, and the make check.valgrind shows a lot of issues, but the current HEAD gives: ==32522== LEAK SUMMARY: ==32522== definitely lost: 29,550 bytes in 1,316 blocks ==32522== indirectly lost: 31,713 bytes in 656 blocks ==32522== possibly lost: 2,071,160 bytes in 1,759 blocks while after the patch, we have: ==32541== LEAK SUMMARY: ==32541== definitely lost: 4,895 bytes in 113 blocks ==32541== indirectly lost: 2,640 bytes in 81 blocks ==32541== possibly lost: 2,074,814 bytes in 1,767 blocks But, both still segfault here and there around the testproperty series (at least, the given memory estimations are for the roughly the same number of successful tests)... Beside, I think that this patch corrects the TODOs related to release the argument in case of error since the pygi_argument_release is moved to the free routine. What's your opinion ? Regards, Damien. _______________________________________________ pygtk mailing list [email protected] http://www.daa.com.au/mailman/listinfo/pygtk Read the PyGTK FAQ: http://faq.pygtk.org/
