The version 3.2.1 of valgrind (the current ebuild) seems to have problems with suppressions: in particular, if you generate a suppression rule (with --gen-suppressions=yes) and then re-run valgrind on the same program using the generated suppression rule, then this suppression rule is simply ignored. I think this is a problem with this version of valgrind (not with the specific gentoo ebuild, since the same problem with the same version occurs in debian too). In debian, there's no such problem with version 3.2.3 of valgrind. Reproducible: Always Steps to Reproduce: 1. make a C program with a memory leak, and compile it 2. run valgrind on this program with --gen-suppressions=yes (and leak checking enabled) and save the suppression rule in a file, e.g., suppressions.txt 3. rerun valgrind on this program, using --suppressions=suppressions.txt Actual Results: valgrind will still show the leak even if there's a suppresion rule for ignoring it Expected Results: valgrind should show that leak since it should be suppressed by the generated suppression rule there are no problems with suppression rules created by other versions of valgrind
This is bug is present in valgrind's bugzilla and has been fixed in 3.2.3: http://bugs.kde.org/show_bug.cgi?id=136844 Valgrind-3.2.3 has been in portage (~x86) for a while, so you can use that or you can fix your suppressions by replacing "fun:_vgrZU_libcZdsoZa_malloc" with "fun:malloc".
