im.pidgin.pidgin: d77adf1b9a4b44f121620f20e2643602e3f6776e

[Mark Doliner]

On Wed, 23 Jan 2008 06:44:08 -0500, Evan Schoenberg wrote
> So given that it appears the code is creating one-and-only-one timer 
>  in one place and removing it as appropriate, do you have thoughts 
> how  the crash log and disassembly I posted in my previous message 
> in this  thread exist?

Not really... maybe bad RAM?  I suppose this problem only happens
sporadically.  Does it happen to EVERYONE sporadically, or just a subset of
your users?  How frequently does it happen?  For the people that experience
the crash, it wouldn't hurt to ask them if they experience other applications
crashing.  Or if they have ever had additional RAM installed in their
computer.  Or maybe even ask test-savvy users to run a RAM test a few times
(apparently there's one that ships with OS-X).

I'm pretty confident you can eliminate the oscar code as the problem.  Which I
guess leaves the possibility that your event loop isn't really canceling
timers when you tell it to (which I think you had suggested as a remote
possibility in an IM conversation we had).  But that seems unlikely to me. 
You could add some purple_debug_error() calls to flap_connection.c and see if
flap_connection_destroy_cb() is getting called twice (I doubt that it is). 
When applications crash in OS-X is there any type of core file you can use to
browse around the state of the application after it has crashed?  You could
try printing some variables and see if their values make sense.

-Mark