GNOME Bugzilla – Bug 535516
Crash while replying [GConf Bridge]
Last modified: 2009-10-22 08:56:08 UTC
+ Trace 199006
I got this trace, while I was clicking replying. Just before that I got a error dialog that was all corrupted string wrt gconf and pressing OK crashed. I remember seeing the dialog before.
Sounds vaguely composer related, but need more details. Is it repeatable? Can you post of screenshot of the corrupted error dialog next time you see it?
Ok I got it now again. But this time, I did C+R, closed and C+R and closed, and this dialog popped up. Im sure next it is gonna crash.
Created attachment 111742 [details] Screenshot (evolution:11508): e-utils-WARNING **: GConf error: Bad key or directory name: "\u0018U\u0013 _height": Must begin with a slash '/' (evolution:11508): e-utils-WARNING **: GConf error: Bad key or directory name: "\u0018U\u0013 _maximized": Must begin with a slash '/' But it didnt crash this time. But last time, it crashed for me.
Ok. Keep playing with C+R and Escape 4-5 time, you must see the crash/dialog atleast once.
+ Trace 199013
It got it again.
(evolution:11508): GLib-GObject-WARNING **: invalid unclassed pointer in cast to `GtkWidget' was there on the console this time.l
My guess is maybe GConfBridge is trying to manipulate a dead EMsgComposer instance. Maybe I'm missing an unbind() call somewhere in the finalize logic. I've tried this repeatedly myself and can't seem to reproduce it. An Escape keypress is handled by simply activating the "close" action, same as closing via the window manager or doing File -> Close. Can you reproduce this by closing the window some other way?
+ Trace 199023
Different one this time.
I worked on this a bit today but wound up confusing myself even more. window_binding_perform_scheduled_sync() is a timeout callback. When GConfBridge unbinds the window, it cancels the timeout by calling g_source_remove() and then destroys the WindowBinding structure. Everything about it looks correct to me. Yet somehow the callback is still getting invoked after the timeout has been cancelled. So the callback gets a dangling "binding" pointer, as I suspected. I still have no idea what this has to do with maximizing the window, but indeed I can only seem to get this to occur when the composer window is maximized. I'll keep digging tomorrow...
Got the same trace as in comment 7. Evo 2.24.2. Happened whien switching to another folder while downloading mail (not replying). There was a gconf popup about the error too: (evolution:3959): e-utils-WARNING **: GConf error: Bad key or directory name: »�\u0011�_width«: Must begin with a slash (/)
+ Trace 211467
Sounds very same as bug #555663, I believe the fix there fixes this too, but if you are able to reproduce this in some way on 2.24.2, then please try the same steps on 2.24.3.
https://bugs.launchpad.net/evolution/+bug/335787 has a similar crash on 2.24.3
*** Bug 577056 has been marked as a duplicate of this bug. ***
Looking around in those two above bugs, you were so good to timely close your window just when it was writing postponed information about window position and so on, which leads to using already freed memory (thus the garbage reported by GConf). I thought this all runs on main thread, thus locking couldn't help much, but it doesn't seem so now. Or something caused to process events from window within the call inside that function processing, like accessibility features (assistive technologies). Do you have such thing turned on?
(I'm the reporter of the Ubuntu bug linked to in comment #11). > ... like accessibility features (assistive technologies). > Do you have such thing turned on? No - at least, not to the best of my knowledge (i.e. if it is enabled, then it's by accident - but the desktop seems to behave like a standard desktop, so I have to assume it's off).
(In reply to comment #13) > … like accessibility > features (assistive technologies). > > Do you have such thing turned on? No. I do not either. I am just using Ctrl + scrool wheel for zooming. But I probably did not in this case. But if I did, it was by accident. But reading what I did (Bug 577056) that seems unlikely. Thanks, Paul
Created attachment 132454 [details] test patch for evolution; I cannot reproduce this myself, which is pity, because it might be much easier to produce a fix for such a thing. Anyway, is anybody of you seeing this able to give a try to this test patch, please? If you do not compile yourself then please do not try just because of this, it's not as that easy to compile evolution.
*** Bug 578915 has been marked as a duplicate of this bug. ***
bug 563866 could also be a dupe
*** Bug 563866 has been marked as a duplicate of this bug. ***
This bug has been nagging me for the past week. Every time I've tried send an e-mail this happens. I use Ubuntu 8.10 on an Acer AspireOne (mini laptop/netbook) with some added packages from https://launchpad.net/netbook-remix which among other things maximizes windows. The maximizer I'm using is maximus_0.4.8~ppa1_i386.deb.
Yes, this is an already known bug, because it has been filed here. It's likely that some users run into it. There's a request to test the patch in comment 16.
*** Bug 583769 has been marked as a duplicate of this bug. ***
Created commit b33ce8f in evo master. As it's a test patch, to master only. It took to long, thus committing myself. Also closing, but please reopen if you see in 2.27.3+. Thanks.
Milan, would you mind pushing that patch upstream? http://bugzilla.openedhand.com/enter_bug.cgi?product=GConf%20Bridge
Probably not, just that I do not have account there.
So sign up. It's free. http://bugzilla.openedhand.com/createaccount.cgi
*** Bug 585132 has been marked as a duplicate of this bug. ***
Milan, can you please forward this upstream?
(In reply to comment #28) > Milan, can you please forward this upstream? Eh, I mind to create an account there, I'm sorry.
*** Bug 586713 has been marked as a duplicate of this bug. ***
*** Bug 586788 has been marked as a duplicate of this bug. ***
*** Bug 587136 has been marked as a duplicate of this bug. ***
(In reply to comment #29) > (In reply to comment #28) > > Milan, can you please forward this upstream? > > Eh, I mind to create an account there, I'm sorry. Thanks for not getting the concept of upstream and downstream. Filed for lazy people: http://bugzilla.openedhand.com/show_bug.cgi?id=1680
*** Bug 588723 has been marked as a duplicate of this bug. ***
*** Bug 590555 has been marked as a duplicate of this bug. ***
*** Bug 591130 has been marked as a duplicate of this bug. ***
*** Bug 592922 has been marked as a duplicate of this bug. ***
*** Bug 593574 has been marked as a duplicate of this bug. ***
*** Bug 598245 has been marked as a duplicate of this bug. ***
*** Bug 599253 has been marked as a duplicate of this bug. ***