GNOME Bugzilla – Bug 607744
Crash on MAPI notification close
Last modified: 2010-02-08 14:38:29 UTC
evolution 2.29.5 libmapi 0.9 When i logged in into the laptop after suspension, found that evolution had crashed. gdb traces of evolution exchange-mapi-connection.c:2355: Entering exchange_mapi_set_flags libexchangemapi-Message: exchange-mapi-connection.c:2357: exchange_mapi_set_flags: lock(connect_lock) libexchangemapi-Message: exchange-mapi-connection.c:2395: exchange_mapi_set_flags: unlock(connect_lock) exchange-mapi-connection.c:2397: Leaving exchange_mapi_set_flags [Thread 0xacb91b70 (LWP 4358) exited] evolution-shell-Message: Network disconnected. Forced offline. evolution-shell-Message: Preparing for offline mode... [New Thread 0xacb91b70 (LWP 4534)] Program received signal SIGSEGV, Segmentation fault. [Switching to Thread 0xacb91b70 (LWP 4534)] 0xb1383e21 in exchange_mapi_events_monitor_close () at exchange-mapi-connection.c:3075 3075 close (notify_ctx->fd); (gdb) t a a bt
+ Trace 220204
Thread 42 (Thread 0xacb91b70 (LWP 4534))
Thread 1 (Thread 0xb6330760 (LWP 3857))
(gdb)
I got same crash when i tried to go offline and closed the dialog which asks 'do you want to synchronize or not'
I get this crash every time when i go offline.
I recall I saw something similar myself too, though not too often.
Created attachment 153014 [details] [review] proposed ema patch for evolution-mapi; Please give a try to this. While I was changing the notification API there, I fixed also the notification itself, which I broke few days ago, based on investigation by Jony, whom found that the callback isn't called at all, thus newly received messages are not shown "automatically".
Just note that with this patch it will not crash on the same place, because I removed that function completely. Only make sure it's not more unstable than before. Thanks.
I have applied the patch. It doesn't crash as before but i get crash reported in bug 607732 often.
Created commit 006728d in ema master (0.29.90+) Slightly changed patch, to adapt it to recent master changes. With respect of the other bug, let's move there, I'm not seeing the crash itself, but valgrind claims to me.
This change seems to be causing a crash at shutdown for me:
+ Trace 220435
(In reply to comment #8) > This change seems to be causing a crash at shutdown for me Oh, thanks, I do not see it myself. I guess the "t a a bt" would show here that the global session is doing something in some other thread, thus I added session lock to notification functions, which might ensure it'll be run only when the session is not used elsewhere. Please reopen if not. Thanks. Created commit 68a1e00 in ema master (0.29.91+)
Err, and commit 35409ea, as I used session lock where I shouldn't.