When remote to 855gm machine use ssh, start x and quit X, the 855gm's display fall into halt, other operations can't effect on 855gm local, but 855gm can be connected and pinged.
Then start x again from remote, 855gm hanged, can't be connected and pinged.
On 855gm local, however start X and quit X many times, the problem can't be reproduced.
Created attachment 11202 [details]
Kernel log after hanged
Created attachment 11203 [details]
Created attachment 11204 [details]
Xorg's config file
try to build xserver with --disable-config-dbus --disable-config-hal
After rebuild Xserver use these options on 855gm, the bug also exists.
so you should try to attach it to gdb, and get backtrack info in seg fault.
In my 945GM machine, when I killed X, it also had below backtrace:
xterm: fatal IO error 32 (Broken pipe) or KillClient on X server ":0.0"
+ Exit 84 xterm
[root@x-945gm ~]# 0: X(xf86SigHandler+0x7e) [0x80b3c9e]
2: X [0x80a020c]
3: X(main+0x4cf) [0x806e56f]
4: /lib/libc.so.6(__libc_start_main+0xdc) [0x9be7e4]
5: X(FontFileCompleteXLFD+0xb9) [0x806d861]
Fatal server error:
Caught signal 11. Server aborting
But X can restart up. So I think the backtrace has nothing to do with the hanged issue on 855GM. The backtrace seems about font library. Maybe we need update some font modules.
Eric suggest to try latest upstream server again..
It seems still dbus problem. Try to remove dbus-devel package, then build xserver with dbus and hal config disabled.
After disable dbus and hal, rebuild xserver, the bug still exists and following is backtrace infor when kill X:
#0 __glXgetActiveScreen (num=0) at glxscreens.c:189
#1 0x080a15e5 in AbortDDX () at xf86Init.c:1102
#2 0x081b4788 in AbortServer () at log.c:406
#3 0x081b4c96 in FatalError (
f=0x81bdd50 "Caught signal %d. Server aborting\n") at log.c:552
#4 0x080b3d0d in xf86SigHandler (signo=11) at xf86Events.c:764
#5 <signal handler called>
#6 0x081aa122 in RemoveGeneralSocket (fd=-1) at connection.c:1071
#7 0x080a020c in teardown () at dbus-core.c:90
#8 0x0806e56f in main (argc=2, argv=0xbfd15414, envp=Cannot access memory at address 0x8000007
) at main.c:471
did you do what I said to remove dbus-devel as a workaround?
dbus-devel is removed, which is dependency of hal.
Then rebuild xserver.
But the bug still exists.
The backtrace is the same as above.
It's ok with the latest code.