Version-Release number of selected component (if applicable): glib2-2.10.3-1.x86_64 gtk2-2.8.18-1.x86_64 How reproducible: Always Steps to Reproduce: 1. Turn on accessibility (gconftool-2 --set --type bool /desktop/gnome/interface/accessibility true) 2. Launch a GTK-based application (metacity, evince, eog) 3. Watch the app dump its memory map on the console and freeze. Additional info: This did not happen with the packages included in the FC5 release - it only started happening after I updated to the latest available packages as of yesterday afternoon.
Created attachment 131210 [details] eog backtrace This is a backtrace from eog when I launch it with accessibility turned on.
Created attachment 131211 [details] eog memory map This is the memory map dumped to the console when eog crashes. I get similar dumps from every other GTK program. (If any other program's backtraces or memory maps are required, I can provide them.)
You know what? Yesterday I noticed that I couldn't log in with a11y enabled on my x86_64 workstation. Couldn't find any useful information in logs. Maybe this is why... :)
I spent a week trying to track this down prior to fc5 and gave up. It is filed upstream, but I have not seen any action on the bug.
Matthias, can you add a reference to the upstream bug to this one?
Noting, as zcerza asked me to, that this is in fact a regression from the packages that shipped in FC5-release.
I went and found the upstream bug.
I think we either need to track this down or revert the memleak fix that introduced it.
Fixed in at-spi-1.7.9-3.
Unfortunately, the fix leads to other crashes
*** Bug 193909 has been marked as a duplicate of this bug. ***
I haven't seen a11y-related crashes with the latest version of at-spi, so maybe this is fixed after all.
I played around with things on the box in the megacube and i'm not seeing any crashes. I'm going to close this for now. We can open a new bug if problems pop up.
The megacube box is running RHEL5 beta. I'm glad the bug is fixed there, but this bug was reported against FC5, and to my knowledge it hasn't been fixed there. Reopening.
Moving off FC6Blocker list...
Finally closing this. FC5 is no longer supported, and the bug is fixed in newer releases, I believe.