https://bugs.kde.org/show_bug.cgi?id=439897
Milian Wolff <m...@milianw.de> changed: What |Removed |Added ---------------------------------------------------------------------------- Latest Commit| |https://invent.kde.org/sdk/ | |heaptrack/commit/7531b88076 | |42f4e265d57f2dd6bdcbafca85f | |ccf Resolution|--- |FIXED Status|REPORTED |RESOLVED --- Comment #4 from Milian Wolff <m...@milianw.de> --- Git commit 7531b8807642f4e265d57f2dd6bdcbafca85fccf by Milian Wolff. Committed on 16/08/2021 at 18:42. Pushed by mwolff into branch 'master'. Fix backtrace symbolication for optimized code For some reason, we actually get an IP address that points to the instruction _after_ the one we need to use for symbolication. I'm quite frankly totally flabbergasted by this breakage, as it seems to be around for years, yet no-one ever complained? And testing this on some of my code, I believe it used to work in the past without this workaround... If anyone knows more about this, please chime in over at: https://github.com/libunwind/libunwind/issues/287 For now, manually subtracting one from the addresses we get from unwinding is enough to workaround this and get sane backtraces once more. M +9 -2 src/track/libheaptrack.cpp https://invent.kde.org/sdk/heaptrack/commit/7531b8807642f4e265d57f2dd6bdcbafca85fccf -- You are receiving this mail because: You are watching all bug changes.