GNOME Bugzilla – Bug 309320
Evolution crashes unpredictably
Last modified: 2013-09-10 14:03:39 UTC
Version details: evolution-2.2.2-5 stock FC4 RPM Distribution/Version: gnome-desktop-2.10.0-5 stock RC4 RPM Steps to reproduce: 1. Setup a maildir which is periodically updated by offlineimap <http://gopher.quux.org:70/devel/offlineimap/SVN/manual.html>. 2. Configure an Evolution account which uses the maildir protocol to periodically (e.g. every 5 minutes) check for changes to this maildir's contents. 3. Start Evolution. Observed behavior: Evolution will eventually crash without me even touching keyboard or mouse. This crash doesn't necessarily coincide with my settings for periodic scans of my maildir for changes to its contents. I'm unsure whether this crash coincides with updates to the contents of the maildir by offlineimap. With Evolution 1.x on FC3, this didn't happen often (0-2 times a day). With Evolution 2.0.x on FC3, the problem was worse, and occurred several (5+) times a day. Now I have two separate computers using the stock evolution-2.2.2-5 RPM from FC4, and the crash occurs _all the time_, ranging from immediately (milliseconds) after starting Evolution (while it's beginning to scan my maildir), to anywhere from 2-30 minutes. I haven't seen Evolution stay up for more than an hour or two. Expected behavior: Evolution absolutely should not segfault, period. Related issues: Bug 308074 and bug 268720 look possibly related to this problem.
Created attachment 48524 [details] The list of Evolution and GNOME RPMs installed on my systems
Created attachment 48525 [details] Debugging details generated by Bug Buddy from Evolution crash
Created attachment 48526 [details] Debugging details generated by Bug Buddy from Evolution crash (now with -debuginfo packages!)
There's some interesting activity from a couple of threads around the time of the last crash (forunately now more visible due to the -debuginfo packages I installed). Thread 8: ...
+ Trace 61570
The lengths on these string buffers look a little odd, but perhaps are simply related to malloc'd data in the buffer which is not currently in use.
looks like you're using exchange too, which you don't mention in your first entry; it is pretty buggy. But anyway just remove the 'ibex' file(s) from the folder(s) crashing, then restart evolution to rebuild the indixes, that should probably fix it. it would be better if you didn't add CC lists (just setting the correct component would serve the same purpose), and the priority and severity are for our use too.
I have three accounts configured both on my computer at the office, and at home. At home, I do not use the Exchange Connector (nor have it configured), and am still experiencing this problem. Thanks for the suggestion WRT the 'ibex' files. I stopped Evolution, removed the following files, then restarted Evolution: $ find ~/.evolution/ -type f -name '*ibex*' /home/dlr/.evolution/mail/local/Outbox.ibex.index.data /home/dlr/.evolution/mail/local/Sent.ibex.index.data /home/dlr/.evolution/mail/local/Sent.ibex.index /home/dlr/.evolution/mail/local/Inbox.ibex.index /home/dlr/.evolution/mail/local/Drafts.ibex.index.data /home/dlr/.evolution/mail/local/Outbox.ibex.index /home/dlr/.evolution/mail/local/Inbox.ibex.index.data /home/dlr/.evolution/mail/local/Drafts.ibex.index These files were re-created after restarting Evolution. However, I do not use the "On This Computer" account, and there were no such files in my maildir: $ find ~/.offlineimap/ -name '*ibex*' # returns nothing This action had no apparent affect on this bug -- Evolution still crashes regularly. Did I take the appropriate steps, and are there any other things I can try?
Well the trace clearly shows code that ONLY exists in exchange-connector. 'camel-stub' is only used by exchange-connector, and only used if it is active ... Anyway the trace was truncated, instead of using bugbuddy which seems to have bugs of its own, try running gdb manually from a terminal. Do "thread apply all bt" when it crashes. That should also show other terminal output which might help too. It might show *which* folder's index is causing the assert (well it can be determined if you know how to drive gdb ... but its hard to explain in a bug log), if indeed it is still the same crash, and then you can turn indexing off of that folder. You could also try using valgrind, it should work quite well with debug packages (but it probably wont help with any assert since by then the problem has probably already occured.) I can't remember where the maildir indices go if they are generated, but i thought they were either in the directory or in the .evolution one. A full trace will tell me which backend is causing the assert at least, so can identify if it is indeed maildir or something else.
Closing this bug report as no further information has been provided. Daniel, please feel free to reopen this bug if you can provide the information Not Zed asked for. Thanks very much!
I had to stop using Evolution due to this and other stability (and scalability) problems (now using mutt and offlineimap), so I'm fine with closing this as DONTCARE.