Red Hat Bugzilla – Bug 462447
Xorg crashes in exaCompositeRects
Last modified: 2008-10-07 16:11:29 EDT
Created attachment 316840 [details]
Description of problem:
From time to time (like twice day, but there are exceptions) Xorg crashes and I get into console. Backtrace points to /usr/lib64/xorg/modules//libexa.so(exaCompositeRects+0x12f) [0x2272d2f] (see attaached)
Version-Release number of selected component (if applicable):
from time to time
sometimes it happens just without any effort on my side, and sometimes the computer gets to the state that even after relogin, it was killed 3 times out of 3 just by running screen -r in the gnome-terminal. Then I rebooted whole computer and everything works smoothly.
Created attachment 316841 [details]
I don't have /etc/X11/xorg.conf
Created attachment 316842 [details]
1: /lib64/libc.so.6 [0x3bcac33130]
2: /usr/bin/Xorg [0x511a81]
3: /usr/bin/Xorg(ValidatePicture+0x9) [0x511ad9]
$ addr2line -e /usr/lib/debug/usr/bin/Xorg.debug 0x511a81
$ head -1463 /usr/src/debug/xorg-server-1.5.0/render/picture.c | tail -1
if (pPicture->pDrawable && pPicture->serialNumber != pPicture->pDrawable->serialNumber)
That seems... unlikely. The only way you could crash here is if pPicture wasn't a real pointer in the first place.
I have a similar problem, but I can reproduce it quite easily.
If I log into Gnome, start a gnome-terminal and run "top" X will crash and throw me back to GDM-login.
Same happens if I run a MySql-query - even on a remote machine
mysql -h thirdhost
select * from table;
-> instant local X crash
Created attachment 319208 [details]
Xorg.log from a crashed X-session
After a bit of experimenting, it seems like the crash only happens if the gnome-terminal window is bigger than default.
So if I start a new gnome-terminal I can run top just fine. If I either resize the window to fill most of the screen, or maximize it X will crash.
should be fixed in xorg-x11-server-1.5.1-5.fc10
which will land in rawhide soon.
1.5.1-6 does not have xserver-1.5.1-exa-fix-glyph-segfault.patch,
although changelog of the src.rpm lies so.