Using the default panel layout in the latest Rawhide snapshot, I navigate to the
'Actions' menu item in the menu bar on the top panel and select "Run
Application..." from the menu.
The "Run Application" window pops up, and along with it I get an error dialog
notifying me that gnome-panel has crashed.
Steps to Reproduce:
1. Log into GNOME.
2. Click on the 'Actions' menu
3. Select 'Run Applications...'
The panel crashes.
Updating severity to note that this is in fact a crasher.
Oh, also - please note that this bug does seem to be IA64-specific.
Adding this to the GNOME-breaks-on-ia64 blocker list
I'm having a look at this since I believe markmc is on vacation right now.
Seems to be dying inside 3rd invocation of get_all_applications within
panel_run_dialog_add_items_idle, with a corrupt GSList of
all_applications (using GSList code from /usr/lib/libglib-2.0.so.0)
(gdb) print all_applications
$28 = (GSList *) 0x60000000002ffe68
(gdb) print all_applications->next
$29 = (GSList *) 0x60000000003011c0
(gdb) print all_applications->next->next
$30 = (GSList *) 0x6000000000301cd0
(gdb) print all_applications->next->next->next
$31 = (GSList *) 0x63412f3a736e6f69
The "data" field in each node is NULL, until:
(gdb) print all_applications->next->next->data
$35 = 0x746163696c707061
Am investigating further... some kind of 32-bit vs 64-bit borkage?
Stupid get_all_applications function supposedly returns a GSList*, but
never actually has a "return" statement. Looks like it's managing to
work on every other architecture, but ia64 is fussy about such things.
I'll try to patch things...
I had a look for this in CVS, it turns out get_all_applications is
actually supplied by one of our patches, rather than upstream
(use-correct-applications.patch); am experimenting with an updated
version of the patch that has a fix.
Should just be a "return all_applications;" at the end of the
function. I've no idea how that ever worked.
OK - after much pain learning the mysteries of multiarch-RPM I've got
a patch to the patch that I've verified fixes things on an actual ia64
box. markmc - where do you want me to apply this: FC-3, RHEL-4, devel?
Created attachment 105875 [details]
Patch to add the return statement to the patch, and to update spec file accordingly
You're probably more aware of what's appropriate than I am - but I'm
guessing FC-3, RHEL-4 and devel would all by appropriate.
If you're swamped, feel free to bounce this back at me. Thanks Dave ...
markmc, can you see this gets done please? Reassigning, thanks.
Fixed and built for rawhide and RHEL4, fix will be in next FC-3 update.