GNOME Bugzilla – Bug 85553
crash in gdk_event_translate()
Last modified: 2009-08-15 18:40:50 UTC
Package: gnome-panel Severity: normal Version: 2.0.0 Synopsis: crash in gdk_event_translate() Bugzilla-Product: gnome-panel Bugzilla-Component: clock applet BugBuddy-GnomeVersion: 2.0 (2.0.1) Description: I've seen this several times, this time with a clean compile out of CVS from Jun 14. Looks like memory corruption? Debugging Information: Backtrace was generated from '/unst/bin/gnome-panel' [New Thread 1024 (LWP 16284)] 0x420b4769 in wait4 () from /lib/i686/libc.so.6
+ Trace 23932
Thread 1 (Thread 1024 (LWP 16284))
------- Bug moved to this database by unknown@bugzilla.gnome.org 2002-06-17 01:00 ------- Reassigning to the default owner of the component, gnome-panel-maint@bugzilla.gnome.org.
Hmm. Scary...I remember seeing these in #59500 too :/ I hope I'm wrong here.
Ah, crap :/ Someone from wipro or sun should purify this.
Anyone around with purify on a 1.4.x box somewhere!?!?! :)
hp: were there any particular steps you took to trigger this, in case wipro can purify it? Kjartan: you might want to poke fejj and see if he can purify a 1.4 panel for you. Not sure if we have it installed on our 1.4 Sun box or not.
No, every time I've seen this it's been "out of the blue" - _possibly_ when opening a new window? That could be libwnck's fault if so, I think it uses an event filter. Or it may just be that something happens earlier that crashes libwnck later. Suggest building GLib with --enable-gc-friendly --disable-mem-pools and all that when debugging this.
Do those work in 1.2.x too?
Havoc: This is gtk+-2.0.x right ? I spent a while eyeballing the panel and libwnck code to see if there was anything obvious to no avail. Was this a KeyPress do you think ? That's all the panel uses a filter for. libwnck uses it for PropertyNotify, ConfigureNotify and SelectionClear, so the culprit may lie there if it happened when opening a new window ? You don't happen to have any local changes with mismatched signal handlers in libwnck or anything ? I'm marking as NEEDINFO for now since I can't really proceed further without more info. Also, I'm suspicious that we haven't seen any dups ...
Let's just leave this but watch for dups. it doesn't happen _often_ though, just from time to time.
Apologies for the spam. What's the status of this one? Still seeing it, Havoc?
Just adding some data from my #59500 investigation: It seems that the xid hashtable is being corrupted and sometimes gdk_window_lookup()/gdk_xid_table_lookup() returns a window with corrupted data. Could this really be a gdk problem that still hasn't been fixed?
It seems that this bug is back in gnome-panel 2.1.2. See bug #97487. I've never seen this before gnome-panel 2.1.2, though.
Havoc: is this bug still useful or can we close it ? There were no duplicates in 5 months.
I haven't seen it lately. Resolved->MAGICALLYWENTAWAY