Evolution hangs unpredictably, and doesn't restart without a reboot

[Ashwin]

Evolution hangs unpredictably, ranging from a few minutes after start up, to upto a few hours. This frequently manifests when replying to email. Attempting to restart Evolution after the hang fails; the application will not start up. The only way out seems to be a reboot. Similar issues are being reported in the forums:
http://ubuntuforums.org/showthread.php?p=7164806

This started happening after the system was upgraded from 8.10 to 9.04. Compiz is not enabled. Hardware is a Lenovo Z61e. Oh, and Tracker is disabled; in fact, it got so troublesome (for some reason we couldn't turn it off) that we ended up uninstalling Tracker altogether.

[Ben Crisford (bencrisford)]

I get this problem as well, so confirmed :).

And nice bug report :), everything about it is textbook, I certainly don't think any more info is needed (hence why I am confirming it).

Cheers,
Ben

[Sebastien Bacher (seb128)]

Thank you for taking the time to report this bug and helping to make Ubuntu better. Please try to obtain a backtrace following the instructions at http://wiki.ubuntu.com/DebuggingProgramCrash and upload the backtrace (as an attachment) to the bug report. This will greatly help us in tracking down your problem.

[Ashwin (ashwinjmathew)]

Thanks for your attention, Sebastien! This is actually not a crash; Evolution hangs, but does not shut down. After we close it manually, it refuses to restart. From what we can see in the system monitor, the evolution process is still active, which may be why it doesn't start up again.

[Sebastien Bacher (seb128)]

you can get stacktrace of hanging software too, it will give indication on where it's blocked

[Ashwin (ashwinjmathew)]

Regrettably, I can't help now; this turned into such a problem for our day-to-day work that we had to roll back to 8.10. Ben, any chance you could provide a stacktrace?

[Steve Bergman (sbergman27)]

I run an XDMCP/NX server which runs about 60 concurrent jaunty gnome desktops, and I think we are experiencing this problem.

Next time this happens to a user, I'll try to collect more info.

Just to be clear, here is what I intend to do:

Install dbg packages for evolution, evolution-exchange, and evolution-data-server.

gdb -p <procid>
bt

and report that output. And then:

strace -o <output_file> -f -p <procid>

and post the output_file.

Collect details about exactly what the user was doing and any status bar messages, and report that.

Is that sufficient?

This happens more than once per day, so I may have something by the end of the day.

[Steve Bergman (sbergman27)]

OK. I have a gdb and strace output. Since gnome-keyring-daemon is mentioned in the output, and since I'm pretty sure I saw a bug in Fedora's bugzilla in which it was suspected that the cause was gnome-keyring-daemon deadlocking, I collected gdb and strace output for gnome-keyring-daemon, as well.

Unfortunately, when I ran gdb on evolution, gnome-keyring debug symbols were not installed. I did install them before running gdb on gnome-keyring-daemon.

[Steve Bergman (sbergman27)]

Here is what I think looks like the same bug over at RedHat/Fedora:

https://bugzilla.redhat.com/show_bug.cgi?id=498871

I have about 60 users who would love to see a fix. :-)

[Sebastien Bacher (seb128)]

Thank you for all the details, the fedora bug seems similar indeed, there is no gnome-keyring hacker around though and the bug seems to be a tricky one it would perhaps be worth to open it upstream too on bugzilla.gnome.org

[Steve Bergman (sbergman27)]

I will file upstream also. I should note that the original reporter mentions that he upgraded from 8.10 to 9.04. Mine is a fresh install, but with a big twist. Since I have many users, I copied over the ~/.gnome2/keyring directories from the Fedora 8 that we were coming from. Also, I moved their entries in passwd and shadow over into the new Jaunty /etc/passwd and /etc/shadow files. All the entries use md5 hashes in /etc/shadow rather than the new SHA-1 or whatever it is that Jaunty normally uses. Also the UIDs are < 1000, which differs from what would be the case if I had recreated them in jaunty. (Keyring tries their login passwd as the keyring passwd before it asks them for their keyring password. Right? So the shadow file could be implicated.)

I've just finished resetting the passwords for all the users to get the new hash type. And tonight I think I will delete and recreate their keyrings. Since we don't have a flood of people reporting this problem, it stands to reason that there is something different about my setup.

[HippieDave (dalden)]

I've never experienced a hangup while working with Evolution. Hangup for me appears to be when Evolution has been left open for a while. Then I have to force quit, and often have to reboot PC.

[Pedro Villavicencio (pedro)]

It looks similar to upstream report 532771 but it seems it was fixed, could you guys please comment upstream? Thanks.

[Pedro Villavicencio (pedro)]

is this still an issue with latest package on Lucid?

[Ashwin (ashwinjmathew)]

I haven't tried it on Lucid yet, but have had no problems for several months now on 9.10.

[Martin Mai (mrkanister-deactivatedaccount-deactivatedaccount)]

Thanks. Closing the bug due to your last comment.