Red Hat Bugzilla – Bug 1253243
Firefox 40 randomly crashes
Last modified: 2015-08-14 12:55:28 EDT
Created attachment 1062437 [details]
Probably duplicate of https://bugzilla.redhat.com/show_bug.cgi?id=1253086 but I am unable to attach my back traces to existing bugs.
Description of problem:
Since Firefox 39 I'm experiencing random crashes.
With Firefox 40 it seems a lot worse.
Version-Release number of selected component (if applicable):
Normal browsing (reading pages, changing tabs, watching videos) but very random.
I.e. not (easily) reproducible.
Steps to acquire back traces:
1. Created new profile
2. Ran firefox in gdb: firefox -g -d gdb -P Bug -safe-mode -no-remote
3. Browsed a while until it crashed and followed steps on https://fedoraproject.org/wiki/Debugging_guidelines_for_Mozilla_products
Flash or no flash doesn't seem to matter.
debuginfo was installed via dnf debuginfo install firefox, but gdm mentioned a lot of additional packages, yet dnf wouldn't install them (for some reason it hanged at the downloading stage).
My related mozilla bug reports since 39:
Created attachment 1062438 [details]
back trace 1
Created attachment 1062439 [details]
back trace 2
Created attachment 1062441 [details]
back trace 3
Created attachment 1062442 [details]
back trace 4
Created attachment 1062443 [details]
back trace 5
Created attachment 1062444 [details]
I've browsed a couple of hours on a Sandy Bridge IGP (instead the AMD Bulldozer system with a GTX 650 that was used for the back traces) and haven't had a single crash.
I've read that the crashes are supposedly caused by OMTC which I could have had enabled in FF39 myself which explains the pre FF40 crashes.
Besides the crashes with nouveau I'm fairly sure it also happened with the Nvidia proprietary driver (but I do not use that as often as nouveau so it could have been something else).
The work around of disabling OMTC seems to work and I've now been running a day without a problem (on the AMD Bulldozer system ).
I also tried to only disable the layers acceleration but that didn't work nor did the full disabling of hardware acceleration.
Since the work around seems to work I'll close this report.