https://bugs.kde.org/show_bug.cgi?id=493433
--- Comment #12 from Mark Wielaard <m...@klomp.org> --- Looking through the results of some of the try builds it looks there are three kind of issues with matching the backtraces. (The testsuite backtrace matching is sadly somewhat flaky in general). First there are internal symbols in the c library because there are multiple calls between the dprintf and the system call. Second there are sometimes versioned symbols. Third the debian (32bit) arches seem to have a totally different call chain (and system call) than the other fedora 64bit ones. I would try to use stderr_filter: filter_fdleak in none/tests/track_new.vgtest And completely filter out any frames with internal symbols starting with _. # Remove "internal" _functions sed '/by 0x........: _/d' | Then remove all symbol versions (those with @ and @@) # Remove @version symbols sed -E 's/ ([a-zA-Z0-9_]+)@@?[A-Z0-9._]+/ \1/' | Finally try replacing the new possible debuginfo (source:line) with in (/...libc...) perl -p -e "s/\(dprintf.c:[0-9]*\)/(in \/...libc...)/" | perl -p -e "s/\(open.c:[0-9]*\)/(in \/...libc...)/" | perl -p -e "s/\(lseek(?:64)?.c:[0-9]*\)/(in \/...libc...)/" | Then have simpler .exp files (one for "regular" and one for "debian 32bit": == none/tests/track_new.stderr.exp == File descriptor was closed already at 0x........: write (in /...libc...) by 0x........: dprintf (in /...libc...) by 0x........: main Previously closed at 0x........: close (in /...libc...) by 0x........: main Originally opened at 0x........: creat (in /...libc...) by 0x........: main == none/tests/track_new.stderr.exp.debian32 == File descriptor was closed already at 0x........: llseek (in /...libc...) by 0x........: dprintf (in /...libc...) by 0x........: main Previously closed at 0x........: close (in /...libc...) by 0x........: main Originally opened at 0x........: creat (in /...libc...) by 0x........: main -- You are receiving this mail because: You are watching all bug changes.
