X server 1.9.0 crashes when launching Emacs 23.2.1 (GeForce 310 ; 260.19.06 driver)

[mr.zor]

I'm running Arch Linux, using the community supported nvidia-beta package (to dodge the slow-font-rendering issue in latest stable drivers). My system is 64bit (which somewhat shows in the backtrace).

This text comes from /var/log/slim.log

Backtrace:
0: /usr/bin/X (xorg_backtrace+0x28) [0x49f558]
1: /usr/bin/X (0x400000+0x67939) [0x467939]
2: /lib/libpthread.so.0 (0x7fa97bc45000+0xf1c0) [0x7fa97bc541c0]
3: /usr/bin/X (0x400000+0x130355) [0x530355]
4: /usr/bin/X (0x400000+0x2a2e1) [0x42a2e1]
5: /usr/bin/X (0x400000+0x2123e) [0x42123e]
6: /lib/libc.so.6 (__libc_start_main+0xfd) [0x7fa97abd7c4d]
7: /usr/bin/X (0x400000+0x20de9) [0x420de9]
Segmentation fault at address 0x4

Fatal server error:
Caught signal 11 (Segmentation fault). Server aborting

[mr.zor]

attachment !!

[AaronP]

There's nothing in that backtrace that implicates the NVIDIA driver. Can you please connect to the system remotely via SSH, attach a debugger by running "sudo gdb -p `pidof X`", reproduce the crash, and then capture the output of "backtrace full" from the debugger?

[mr.zor]

Unfortunately I haven't been able to access the computer remotely so I've tried some juggling on ttys.

I managed to eventually get a gimped coredump (using "gcore"). Reading it again, here's a (useless??) different backtrace :

Core was generated by `/usr/bin/Xorg'.
Program terminated with signal 11, Segmentation fault.
#0 0x0000000000530355 in ?? ()
(gdb) bt
#0 0x0000000000530355 in ?? ()
#1 0x000000000042a2e1 in ?? ()
#2 0x000000000042123e in _start ()

Given the pthread references in the Xorg.log , I tried "info threads" before that, but alas only one thread (detected?), waiting on a select() at first, then continued till crash.

For the record: http://www.x.org/wiki/Development/Do...erverDebugging
If I had that information prior starting, that would have saved me a few reboots because of the freezes.

No package provides symbols for Arch builds, I'll probably go with brewing a package myself with optimizations off and symbols on with the hope of getting a more useful backtrace.

Meanwhile I've also explored "X -core" option, but this involves running X manually (for the experiment sake, I kept running it under "slim"). I don't believe the coredump would be very different from what I've got without debug symbols. If you think it's worth exploring that way I might spend a bit of time to get that -core option work

I'm attaching full Xorg.log.0.old + a somewhat edited `ps auxf | grep Xorg -A3 -B1` with the hope it will help you get a clearer picture of my setup.

I can provide the compressed coredump I got, 12M Bz2 file, 150M or so uncompressed, but I doubt it could be of any use.

Last but not least, I could reproduce the exact same crash by running kcolorchooser - I'm afraid the list could go on...

[mr.zor]

More information (with some symbols).
More info there (+links to other bugtrackers about my issue) : https://bugs.launchpad.net/ubuntu/+s...rs/+bug/650599

(gdb) bt
#0 0x000000000052fcb5 in PanoramiXTranslateCoords ()
#1 0x000000000044fb11 in Dispatch ()
#2 0x000000000042129e in main ()

(gdb) info registers
rax 0x224 548
rbx 0x1051890 17111184
rcx 0xb 11
rdx 0xa26e30 10645040
rsi 0x1273480 19346560
rdi 0x0 0
rbp 0x0 0x0
rsp 0x7fff1dd2c6d0 0x7fff1dd2c6d0
r8 0x7e74e0 8287456
r9 0x7fff1dd2c700 140733693740800
r10 0xc 12
r11 0x1 1
r12 0x0 0
r13 0x1 1
r14 0x1 1
r15 0x0 0
rip 0x52fcb5 0x52fcb5 <PanoramiXTranslateCoords+533>
eflags 0x10206 [ PF IF RF ]
cs 0x33 51
ss 0x2b 43
ds 0x0 0
es 0x0 0
fs 0x0 0
gs 0x0 0

EDIT: Confirmed this is nvidia unrelated but rather some obscure Xorg+Xinerama bug. working workaround, patch is available at https://bbs.archlinux.org/viewtopic....840787#p840787