sushi-start crashed with SIGSEGV in __pthread_mutex_lock()

[gogo]

sushi-start crashed

ProblemType: Crash
DistroRelease: Ubuntu 12.04
Package: overlay-scrollbar 0.2.14-0ubuntu1
ProcVersionSignature: Ubuntu 3.2.0-15.24-generic 3.2.5
Uname: Linux 3.2.0-15-generic x86_64
ApportVersion: 1.91-0ubuntu1
Architecture: amd64
Date: Mon Feb 13 12:23:35 2012
ExecutablePath: /usr/lib/gnome-sushi/sushi-start
InstallationMedia: Ubuntu 12.04 LTS "Precise Pangolin" - Alpha amd64 (20120207)
PackageArchitecture: all
ProcCmdline: /usr/lib/gnome-sushi/sushi-start
ProcEnviron:
 SHELL=/bin/bash
 PATH=(custom, no user)
 LANG=hr_HR.UTF-8
SegvAnalysis:
 Segfault happened at: 0x7f71566e1e84 <pthread_mutex_lock+4>: mov 0x10(%rdi),%esi
 PC (0x7f71566e1e84) ok
 source "0x10(%rdi)" (0x656d656863732d82) not located in a known VMA region (needed readable region)!
 destination "%esi" ok
SegvReason: reading unknown VMA
Signal: 11
SourcePackage: overlay-scrollbar
StacktraceTop:
 pthread_mutex_lock () from /lib/x86_64-linux-gnu/libpthread.so.0
 XrmQGetResource () from /usr/lib/x86_64-linux-gnu/libX11.so.6
 XGetDefault () from /usr/lib/x86_64-linux-gnu/libX11.so.6
 ?? () from /usr/lib/x86_64-linux-gnu/libcairo.so.2
 ?? () from /usr/lib/x86_64-linux-gnu/libcairo.so.2
Title: sushi-start crashed with SIGSEGV in pthread_mutex_lock()
UpgradeStatus: Upgraded to precise on 2012-02-12 (1 days ago)
UserGroups: adm cdrom debian-tor dip disk lpadmin plugdev sambashare sudo

[Apport retracing service (apport)]

StacktraceTop:
 __pthread_mutex_lock (mutex=0x7fff36fcab30) at pthread_mutex_lock.c:50
 XrmQGetResource (db=0x1b899e0, names=<optimized out>, classes=<optimized out>, pType=0x7fff36fcab5c, pValue=0x7fff36fcab30) at ../../src/Xrm.c:2543
 XGetDefault (dpy=0x1b295a0, prog=0x7f7156ed0a27 "Xft", name=0x7f7156ed0a2b "antialias") at ../../src/GetDflt.c:254
 get_boolean_default (value=<synthetic pointer>, option=0x7f7156ed0a2b "antialias", dpy=0x1b295a0) at /build/buildd/cairo-1.10.2/src/cairo-xlib-screen.c:95
 _cairo_xlib_init_screen_font_options (info=0x2204c00, dpy=0x1b295a0) at /build/buildd/cairo-1.10.2/src/cairo-xlib-screen.c:143

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Jeroen Ost (jeroen-ost-8)]

I ran into the same problem with a self written libwebkitgtk program, also for the 'antialias' property which is asked in the same path XGetDefault call from cairo-xlib-screen.c .

The segfault seems to always occur in the calls from libcairo2, more precisely cairo-xlib-screen.c function _cairo_xlib_init_screen_font_options (Display *dpy, cairo_xlib_screen_t *info)

it's always one of the get_integer_default or get_boolean_default calls.
If I write down the values in a successful run (for my system that is
    cairo_bool_t xft_hinting=1;
    cairo_bool_t xft_antialias=TRUE;
    int xft_hintstyle=0;
    int xft_lcdfilter=1;
)
and prevent the calls from happening, the segfault goes away. (I create a patched libcairo2)
Anyone else who wants to try this (on oneiric):
sudo apt-get source libcairo2
sudo apt-get build-dep libcairo2
cd cairo-1.10.2/
modify src/cairo-xlib-screen.c so that values above are used (line 135) and comment out all 3 function calls for get_integer_default or get_boolean_default.

sudo dpkg-buildpackage -rfakeroot -b -nc
cd ..
sudo dpkg -i libcairo2_1.10.2-6ubuntu3_i386.deb libcairo2-dbg_1.10.2-6ubuntu3_i386.deb libcairo2-dev_1.10.2-6ubuntu3_i386.deb libcairo-gobject2_1.10.2-6ubuntu3_i386.deb libcairo-script-interpreter2_1.10.2-6ubuntu3_i386.deb

As to what is the clean fix I'm not sure. Is there some requirement on pthread synchronization that cairo doesn't meet when calling XGetDefault, or is it a regular Xorg bug ?

I also noticed /etc/xdg/xfce4/Xft.xrdb allows you to set defaults but that didn't seem to help me (even though I'm running xfce4/xubuntu)