compiz crashed with SIGSEGV in __libc_free()

[Martin Pitt]

After today's upgrade and a reboot, compiz immediately crashed twice once I started opening windows, i. e. a few seconds after logging in.

ProblemType: Crash
DistroRelease: Ubuntu 11.04
Package: libnux-0.9-0 0.9.34-0ubuntu1
ProcVersionSignature: Ubuntu 2.6.38-7.38-generic 2.6.38
Uname: Linux 2.6.38-7-generic x86_64
Architecture: amd64
Date: Wed Mar 23 08:00:13 2011
EcryptfsInUse: Yes
ExecutablePath: /usr/bin/compiz
InstallationMedia: Ubuntu 11.04 "Natty Narwhal" - Alpha amd64 (20110302)
ProcCmdline: compiz
ProcEnviron:
 LANGUAGE=de:en
 PATH=(custom, user)
 LANG=de_DE.UTF-8
 SHELL=/bin/bash
SegvAnalysis:
 Segfault happened at: 0x7f1549c4e8c3 <__libc_free+83>: lock cmpxchg %esi,(%rbx)
 PC (0x7f1549c4e8c3) ok
 source "%esi" ok
 destination "(%rbx)" (0x7f1568000012) not located in a known VMA region (needed writable region)!
SegvReason: writing unknown VMA
Signal: 11
SourcePackage: nux
StacktraceTop:
 __libc_free (mem=0x7f1528209470) at malloc.c:3736
 ?? () from /lib/x86_64-linux-gnu/libdrm_intel.so.1
 ?? () from /lib/x86_64-linux-gnu/libdrm_intel.so.1
 intel_region_release () from /usr/lib/dri/i965_dri.so
 intel_miptree_release () from /usr/lib/dri/i965_dri.so
Title: compiz crashed with SIGSEGV in __libc_free()
UpgradeStatus: No upgrade log present (probably fresh install)
UserGroups: adm admin dialout lpadmin sambashare

[Martin Pitt (pitti)]

I just noticed the (probably) duplicate bug 739434, where Brian said he used Alt+Arrow keys quickly. My session launch script does the same with a lot of wmctrl calls.

[Apport retracing service (apport)]

StacktraceTop:
 __libc_free (mem=0x7f1528209470) at malloc.c:3736
 ?? () from /lib/x86_64-linux-gnu/libdrm_intel.so.1
 ?? () from /lib/x86_64-linux-gnu/libdrm_intel.so.1
 intel_region_release (region_handle=0x2dc7aa8) at intel_regions.c:294
 intel_miptree_release (intel=<value optimized out>, mt=0x2785668) at intel_mipmap_tree.c:204

[Jay Taoko (jaytaoko)]

I cannot reproduce it on my intel system with nux 0.9.36. Try to reproduce with the latest Unity release.
If this bug still happens with a stack that traces back to the driver code, then it should be reported as a driver bug.

[Didier Roche-Tolomelli (didrocks)]

@Jay: can you think to revert all tasks? (you didn't revert the nux downstream one) when reverting to a previous state? You fooled my script ;) Setting back to Incomplete :)