Description of problem: I have two systems with r300-based graphic chipsets. One is and i386 notebook with 01:00.0 VGA compatible controller: ATI Technologies Inc RV350 [Mobility Radeon 9600 M10] 01:00.0 0300: 1002:4e50 the other one a x86_64 desktop machine with 02:00.0 VGA compatible controller: ATI Technologies Inc RV350 AP [Radeon 9600] 02:00.0 0300: 1002:4150 I'm running rawhide since two weeks on both machines (updated via yum from FC5). I enabled dri and started gnome and everything worked fine. Later I enabled compiz and it worked fine, too -- on both machines. I logged in and out some times and it still worked. But after a reboot the whole machines locks up after logging into gnome via gdm in 75% of the cases *on the first log in*. Loggin in and out after the first log in fails only in round about 15% of the cases. And the system only lock up in 15 % of the cases on the first login if I do this: - start the system into runlevel 5 with dri enabled - disable dri in xorg.conf from a console - log into gdm; gnome starts without window decorations/window manager; wait until everything has loaded and the buffers are filed - enable dri in xorg.conf from console - kill X with CTRL - ALT - BACKSPACE - log in via GDM -> gnome starts with compiz and all the effects Version-Release number of selected component (if applicable): $ rpm -q compiz xorg-x11-drv-ati kernel mesa-libGL compiz-0.0.13-0.25.20060817git.fc6 xorg-x11-drv-ati-6.6.2-1.fc6 kernel-2.6.17-1.2630.fc6 kernel-2.6.17-1.2647.fc6 mesa-libGL-6.5.1-0.rc2.fc6 otherwise rawhide up2date as of today Additional info: I know this isn't a proper bug report yet. But could somebody tell me where to look further for the root of this problem? Or is this just another "r300 dri isn't perfect yet" issues?
Found a better way to circumvent the problem in 85% of the cases -- start GNOME with metacity enabled and enable compiz afterwards when all the buffers are filled and the system is otherwise idle (the system still locks up in 15% of the cases shortly after I hit "enable windows effects"). (Note: percentages only wild guesses)
Seems to be a lot better know. Thx guys for all your hard work!