https://bugs.kde.org/show_bug.cgi?id=392085
Milian Wolff <m...@milianw.de> changed:
What |Removed |Added
----------------------------------------------------------------------------
Latest Commit| |https://commits.kde.org/hea
| |ptrack/dfb6123a6b1c5a925f4c
| |346e5737d7129a20c683
Status|CONFIRMED |RESOLVED
Resolution|--- |FIXED
--- Comment #12 from Milian Wolff <m...@milianw.de> ---
Git commit a6839d6f36dfd24cdd99205b395d46d8ff7e0185 by Milian Wolff.
Committed on 22/04/2018 at 08:04.
Pushed by mwolff into branch 'master'.
Error out early during recording when receiving child process events
Heaptrack does not yet support recording child processes. But
apparently it can happen that we actually trace child processes when
we use `heaptrack -d`. This then leads to corrupted data files.
This patch catches this scenario early and errors out then.
M +4 -0 src/interpret/heaptrack_interpret.cpp
https://commits.kde.org/heaptrack/a6839d6f36dfd24cdd99205b395d46d8ff7e0185
--- Comment #13 from Milian Wolff <m...@milianw.de> ---
Git commit dfb6123a6b1c5a925f4c346e5737d7129a20c683 by Milian Wolff.
Committed on 22/04/2018 at 08:13.
Pushed by mwolff into branch 'master'.
Do not use shell to launch new programs in debug mode
When we want to debug heaptrack we use GDB to launch the new
process. By default, that actually first starts a shell and then
eventually the debuggee. But both processes would get heaptrack
injected and will then write to the same heaptrack_interpret
process, thereby corrupting the data file.
By using `set startup-with-shell off` in GDB, we can eliminate the
shell and prevent this issue from occurring.
M +1 -0 src/track/heaptrack.sh.cmake
https://commits.kde.org/heaptrack/dfb6123a6b1c5a925f4c346e5737d7129a20c683
--
You are receiving this mail because:
You are watching all bug changes.
