GNOME Bugzilla – Bug 108656
Gnome panel crash
Last modified: 2004-12-22 21:47:04 UTC
Description of Problem: Menu panel crashes when changing windows. Im using a fresh install of GARNOME 0.22 (gnome 2.2.1) using Redhat 8.0 on an Athlon. Steps to reproduce the problem: 1. Have multiple windows open 2. Changed windows by using the builtin window list applet on a menu panel Actual Results: gnome-panel crashes (see trace below) How often does this happen?: Every time Debugging Information: Backtrace was generated from '/achilles/gnome/bin/gnome-panel' [New Thread 8192 (LWP 27458)] 0x420ae169 in wait4 () from /lib/i686/libc.so.6
+ Trace 35014
Thread 1 (Thread 8192 (LWP 27458))
Looks like the stack trace in bug 84082, which was using May 30th 2002 CVS code and was later closed as not being reproducable. Instead of marking this as a duplicate and reopening that bug, I'll just leave this note here. I'm marking priority->high & severity->critical (it's a crasher), and adding the bugsquad keyword.
*** Bug 108722 has been marked as a duplicate of this bug. ***
The reporter of bug 108722 said he used the following steps to obtain this crash (although he didn't test reproducability): Steps to reproduce the problem: 1. Create a new panel (floating one) 2. Right-click on it and chose supress
I'm starting to suspect this is a gtk+ problem. This bug and a few others I've recently seen have the following two functions (even with matchine line numbers) at the top of their stack traces: gtk_widget_get_toplevel at gtkwidget.c:5356 gtk_main_get_window_group at gtkmain.c:1569 (Note that this bug, 108656 is unique to have the g_type_check_instance_is_a function call above those other two--even bug 108722 which has been marked as a duplicate of this one doesn't have that function in the stack trace) Half of these bugs were gnome-panel crashes, the other half are nautilus ones. See bug 108205 for the ones tracked against nautilus. I believe either this should be a duplicate of that one, or that one should be a duplicate of this bug (so long as the product is changes)--but I'm not completely sure.
Sorry for the spam...forgot to add myself to the cc list and I'd like to know whether I was right about this and 108205 being dups...
I looked at the FIXED bugs that had gtk_widget_get_toplevel() in their trace. It seems that bug #78764 (a nautilus one) was due to a GTK+ bug. On the other hand, bug #88427 was due to a bug in gnome-panel. They both have this typical stack trace. Chris : can you give us more information about the configuration of your panel(s) ? IIRC, bug #98274 (still the same trace) happened when the code for transparent panels was checked in. Mark rewrote some stuff and it then became unreproducable. Maybe it is returning now... Arvind : do you have any idea ? Could it be a bug in GTK+ ?
I've tried many panel configs, but it doesn't seem to make a difference. At the moment I'm using a plain menu panel by itself with nothing on it. My original setup also had a sliding transparent panel at the bottom, but I removed it, and the problem still remains. Sometimes you can get away with selecting a window without it crashing, but I've tried looking for some kind of pattern (relating to number of items in the window list etc) but there's nothing. One of those yukky random segfault bugs....
Vincent, Newren: crazy bug heh ? ;) When Luis reported it originally in bug #98274, he did not talk about any transparent panel. And the scerario where he got the crash was after closing bug-buddy. He probably had tasklist on his panel which got updated when bug-buddy was closed which caused the panel to crash. If you look at bug #88427, you could say, the fix could be in panel itself. Since nautilus bugs #108719 and #108205 have also reported similar traces, it could well be a gtk+ bug. Just wondering if Owen is aware of these nautilus and panel crashes. He could throw some light on it.
Something else, don't know if it helps.. The window selected from the list does get the focus before the crash occurs.
Owen: I'm adding you to the cc list (as Arvind suggested might be a good idea) so you can take a look at this and perhaps provide some insight as to whether this might be a gtk+ problem or not. Note that this bug is almost identical (if not identical) to the stack traces being tracked in bug 108205 (which is recording nautilus crashes instead of gnome-panel ones).
Ok guys - bug nailed GARNOME 0.22.0 uses a patched version of GTK+ to add pretty (albeit buggy) drop shadow eyecandy and the like - installing a clean version of GTK+ fixes the problem. I've informed the list so hopefully an updated version will be released. I can't speak for the other bugs with the similar trace - but it fixes it at my end - sorry for wasting your time. Marking bug as NOTGNOME
*** Bug 106857 has been marked as a duplicate of this bug. ***