Bug#407055: Compiz breaks the X server if the other head doesn't have DRI
Daniel Ruoso wrote:
> Qui, 2007-03-01 às 14:21 +0100, Brice Goglin escreveu:
>> Daniel Ruoso wrote:
>>> Well, the version in experimental also breaks, but it presents a
>>> different backtrace...
>> Do you onyl have compiz in experimental? Or also some other X packages?
>> Which xserver-xorg-core do you have? The following backtrace seems to
>> end up in xserver-xorg-core, you could try this package in experimental
>> too :)
> Only compiz-related packages. But, yes, I can try xserver-xorg-core from
> experimental... Any other package?
Looking at the dependencies, mesa could be upgraded too...
>>> 0: X(xf86SigHandler+0x84) [0x80c4354]
>>> 1: [0xffffe420]
>>> 2: /usr/lib/xorg/modules/extensions/libGLcore.so(_mesa_free+0x1d)
>>> 3: /usr/lib/xorg/modules/extensions/libGLcore.so(XMesaDestroyVisual
>>> +0x1d) [0xa7910e2d]
>>> 4: /usr/lib/xorg/modules/extensions/libGLcore.so [0xa791031f]
>>> 5: /usr/lib/xorg/modules/extensions/libglx.so(__glXResetScreens+0x44)
>>> 6: /usr/lib/xorg/modules/extensions/libglx.so [0xb7c75d2c]
>>> 7: X(CloseDownExtensions+0x42) [0x8093f12]
>>> 8: X(main+0x4a3) [0x806e6b3]
>>> 9: /lib/tls/libc.so.6(__libc_start_main+0xc8) [0xb7db3ea8]
>>> 10: X(FontFileCompleteXLFD+0xa9) [0x806d9d1]
>> How did you get this backtrace? It does not look like a complete
>> backtrace with debugging information that gdb could give.
> This appears in my X log file. I'm not even sure of who prints it out,
> but as the X server breaks, I would think this is printed by the x
> server itself.
> I don't know, but maybe this unveils a xorg bug, more than a compiz bug.
Could you ssh from another machine and attach gdb to X before starting
compiz ? (with "gdb -p $(pidof X)")
If would be great to do this with a server recompiled for debugging. If
you want to do so, see how to build at
http://wiki.debian.org/HowToGetABacktrace (with hello=xserver-xorg-core).