Merge lp://staging/~ev/apport/handle-thread-crashes into lp://staging/~apport-hackers/apport/trunk
Proposed by
Evan
| Status: | Merged |
|---|---|
| Merged at revision: | 2538 |
| Proposed branch: | lp://staging/~ev/apport/handle-thread-crashes |
| Merge into: | lp://staging/~apport-hackers/apport/trunk |
| Diff against target: |
134 lines (+50/-10) 5 files modified
apport/ui.py (+18/-1) gtk/apport-gtk (+7/-5) kde/apport-kde (+7/-4) test/test_ui_gtk.py (+9/-0) test/test_ui_kde.py (+9/-0) |
| To merge this branch: | bzr merge lp://staging/~ev/apport/handle-thread-crashes |
| Related bugs: |
| Reviewer | Review Type | Date Requested | Status |
|---|---|---|---|
| Martin Pitt (community) | Approve | ||
|
Review via email:
|
|||
Description of the change
As mentioned in bug 1033902 and in the specification <https:/
I've implemented this as a check for the process still existing, as multiple threads will share the same PID. I do not believe this will be racy - a dying PID should be gone by the time apport-{gtk,kde} is run. However, I'm happy to be told otherwise and sent back to the drawing board.
I've implemented this in both the GTK and KDE frontends and added a test that presumes we have PID 1 to fake a crash around.
To post a comment you must log in.
Preview Diff
[H/L] Next/Prev Comment, [J/K] Next/Prev File, [N/P] Next/Prev Hunk
> a dying PID should be gone by the time apport-{gtk,kde} is run.
This is correct. The kernel core pipeline handler relinquishes stdin as soon as possible, i. e. stdin will be closed (and thus the process will die) even before /usr/share/
Thanks for fixing this, this looks great!
When merging, please remember to add a NEWS entry.