gerritbot should be interruptable

[Andrew Hutchings]

at the moment the only way to stop gerritbot is a kill -9. It should have a proper signal handler to take normal kills and nohups and clean up it's pid file. This may require having everything in threads so the main thread can handle signals

[Clark Boylan (cboylan)]

I think this will be fixed by pushing and using a new gerritlib release. Lifeless, made the gerritlib listener thread a daemon thread so that if the main thread exits so will the entire program. We should work on getting a new gerritlib release out and test this theory.

[Jeremy Stanley (fungi)]

We released gerritlib 0.3.0 as of 2013-12-02 18:01:44 http://git.openstack.org/cgit/openstack-infra/gerritlib/tag/?id=0.3.0 so I'll test that theory.

[Clark Boylan (cboylan)]

I just tested this with gerritlib 0.3.0 and it is still not interruptable. I am guessing there is another thread somewhere not willing to give up control (perhaps in the irc lib?)

[Tim Kelsey (tim-kelsey)]

I would like to work on this.

[Tim Kelsey (tim-kelsey)]

I can't seem to reproduce this exactly, `kill -TERM pid` will exit gerritbot and delete the pid file. Interestingly using any other signal will exit gerritbot but leave the pid file, I'm thinking its probably due to an uncaught exception. Pip freeze lists the following notable package versions in my virtualenv:

gerritbot==0.2.0.1.g6208bd9
gerritlib==0.4.0
irc==8.9.1
lockfile==0.9.1
python-daemon==1.6