SIGSEGV in _clutter_stage_has_full_redraw_queued

[Hernando Torque]

Binary package hint: unity

Clicking on the CoF to initiate the window overview randomly results in a crash (unity restarts, programs still run). Backtrace attached.

ProblemType: Bug
DistroRelease: Ubuntu 10.10
Package: unity 0.2.8-0ubuntu1
ProcVersionSignature: Ubuntu 2.6.35-2.3-generic 2.6.35-rc2
Uname: Linux 2.6.35-2-generic i686
NonfreeKernelModules: nvidia
Architecture: i386
Date: Mon Jun 14 17:39:41 2010
ProcEnviron:
 PATH=(custom, user)
 LANG=en_US.UTF-8
 SHELL=/bin/bash
SourcePackage: unity

[Hernando Torque (htorque)]

Sorry, frame 10 from the backtrace was missing - here's the complete one.

[Hernando Torque (htorque)]

Hm, although triggered differently, the backtrace looks the same as the one in https://bugs.launchpad.net/unity/+bug/592772

[Hernando Torque (htorque)]

I installed the daily desktop build to make sure there are no conflicts with the PPAs I have in use (I had to install the Nvidia driver from nvidia.com to bypass the current conflict between X and nvidia-current), and I can reproduce the crash.

[Jason Smith (jassmith)]

This bug needs to be filed against clutter. Somehow the clutter actor being redrawn has a junk pointer for its stage. This state is temporary and seemingly quickly fixed (possibly an initialization issue). Should an actor redraw be requested while the stage pointer is junk, this crash will result. I am creating a patch for testing. What version of clutter should I create it for Hernando?

basically all we are doing is going into the clutter_actor_queue_real_redraw method and making this line:

if (stage &&
          _clutter_stage_has_full_redraw_queued (CLUTTER_STAGE (stage)))

look like

if (CLUTTER_IS_STAGE (stage) &&
          _clutter_stage_has_full_redraw_queued (CLUTTER_STAGE (stage)))

[Hernando Torque (htorque)]

I'm currently at 1.2.10 (1.2.10-0ubuntu1).

[Jason Smith (jassmith)]

Please test this patch (line numbers are probably different in your version, should apply still)

[Jason Smith (jassmith)]

Woops, made the patch backwards

[Hernando Torque (htorque)]

Thanks, now there's just one problem: with the current Unity version I cannot initiate the expose mode and I haven't been able to reproduce the crash with the app-expose mode (right click on an "active" launcher).

[Hernando Torque (htorque)]

Nevermind, I just downgraded to unity 0.2.8-0ubuntu1.

[Hernando Torque (htorque)]

Unpatched: six crashes within ten minutes, patched: no crash within 15 minutes. The patch seems to fix it. :-)

[Didier Roche-Tolomelli (didrocks)]

@Jason: can you ensure that this clutter patch is submitted upstream, please? Maybe other softwares or further release will encounter the same issue and ensure that a stage *is* a stage is always good in upstream code :)

[Neil J. Patel (njpatel)]

Thanks for the testing Hernando!

[Launchpad Janitor (janitor)]

This bug was fixed in the package unity - 0.2.18-0ubuntu1

---------------
unity (0.2.18-0ubuntu1) maverick; urgency=low

  * New upstream release.
    - can't drag and drop item to last position in launcher (LP: #595819)
    - set a solid color as background (LP: #594232)
    - select a place by default (LP: #601020)
    - Support offline shortcuts in quicklist (LP: #595842)
    - keyboard focus in places should default to the search entry (LP: #599888)
    - fix various search issues in places (LP: #600732, #604964)
    - contracted/expanded feature doesn't take mouse position into account
      (LP: #595878)
    - Use X-GNOME-FullName when available for launcher hover text (LP: #594285)
    - SIGSEGV in _clutter_stage_has_full_redraw_queued (LP: #594209)
  * debian/control:
    - bump libclutk-dev build-dep to latest
    - add libdbusmenu-glib-dev, libgnome-desktop-dev and libgnomeui-dev deps
  * add debian/patches/01_add_dbusmenu_glib_vapi.patch:
    - not distribute yet, small fix in the vapi to make a method return an
      unowned variable instead of an owned one
 -- Didier Roche <email address hidden> Fri, 16 Jul 2010 14:48:14 +0200