edgy: ooffice crashing x

Bug #64013 reported by Matthew Carroll
4
Affects Status Importance Assigned to Milestone
openoffice.org (Ubuntu)
Fix Released
Medium
Unassigned
xorg-server (Ubuntu)
Fix Released
Undecided
Unassigned

Bug Description

running 'ooffice' or any of the standard shortcuts (e.g. 'ooffice -writer %U') is crashing x for me. since x is going down i'm not sure how to provide debugging info for this. let me know what you need.

Revision history for this message
Matthew Carroll (m-tthew) wrote :

I don't think this is a duplicate of #63676, since '/usr/lib/openoffice/program/soffice.bin' crashes x instantly for me. However, they may be related. Like I said, let me know what you need.

Revision history for this message
Matthew Carroll (m-tthew) wrote :

still occurring with the latest edgy build (2.0.4~rc3-0ubuntu4)

Revision history for this message
Joey Stanford (joey) wrote :

This is interesting Matthew. I have a gut feeling this is related to your X configuration and not OpenOffice. However the valid question is why does this happen only with OpenOffice. Would you be able to reconfigure X to default and try again?

sudo dpkg-reconfigure -phigh xserver-xorg

Perhaps with a full update as well:

e.g.
sudo apt-get update
sudo apt-get -y -f dist-upgrade
sudo apt-get -y autoremove
sudo apt-get autoclean

If this still happens we'll need backtraces and valgrinds on both X and OpenOffice. https://wiki.ubuntu.com/DebuggingProgramCrash

You might also want to run memtest86 from the Grub menu just to make certain it's not a memory issue.

Changed in openoffice.org:
status: Unconfirmed → Needs Info
Revision history for this message
Matthew Carroll (m-tthew) wrote :

Hi Joey, you were right - reconfiguring X to default made the problem dissapear, although I'm feeling rather short of screen real-estate now! I had a customised X config (widescreen laptop, dual head with external LCD and xinerama - although xinerama isn't working at the moment, see #63050).

Also, I did full updates and ran memtest (2 full passes, 0 errors) so no issue there.

I'll post my old xorg.conf now and then try to work out what the minimum changes required to trigger the bug is - I suspect it's related to dual head. OO is still the only app that has ever triggered this.

Revision history for this message
Matthew Carroll (m-tthew) wrote :
Revision history for this message
Matthew Carroll (m-tthew) wrote :

OK, I reworked my xorg.conf from scratch starting with the default, ending with the same setup as I had before. The bug now isn't happening, even with my dual head setup and the correct resolutions. Here's my new xorg.conf - some difference between the two is triggering the bug (it recurs if I swap back in the original).

I'm trying to pin it down by swapping lines from the old config into the new config one by one, but I haven't yet worked out which is to blame; perhaps more experienced eyes might be able to spot something?... I'm stabbing in the dark somewhat. So far: it doesn't seem to be the modeline, horizsync, vertrefresh or useedidfreqs lines from the original xorg.conf.

Revision history for this message
Matthew Carroll (m-tthew) wrote : bug triggered by 'Load glx

Through trial and error I isolated the line that causes the bug, in the "Module" section:

Load "glx"

...this only triggers the bug when added to my new dual-head setup. Adding this line to a default xorg.conf does not cause the bug.

So, I'm back to my original post. How do I go about

Revision history for this message
Matthew Carroll (m-tthew) wrote :

oops... can you tell I'm using a different keyboard to normal ;-)

How do I go about getting a backtrace when the bug crashes X? I can't run 'gdb openoffice' from a virtual terminal since it's a graphical application, and if I run it from a terminal window within X it dies with my X session.

Help me help you :-)

Revision history for this message
Matthew Carroll (m-tthew) wrote :

I think I managed to successfully backtrace X via ssh from another machine. The wiki doesn't explain how to attach Valgrind to an already running process though, so you'll have to talk me through that if you need Valgrind for X.

Revision history for this message
Matthew Carroll (m-tthew) wrote :

not sure if this worked properly... attempted valgrinds for soffice.bin - I ran valgrind from a terminal window

Revision history for this message
Matthew Carroll (m-tthew) wrote :
Revision history for this message
Joey Stanford (joey) wrote :

Matthew,

So it looks like i2c, GLCore, and GLX are missing in the Good X setup. I've had a look at the OOo logs and the GDB. It's a poser as they say. I've now seen 2 other bugs with issues which seem to manifest in GLX but my hunch is that there is something awry with packaging environment on our end for OOo. This might be a root cause for OOo crashing but it wouldn't take down X. So, given that your X config issue is most likely tied to GLX, I am of the current opinion that you have something awry with your libraries. I'm not qualified to work on those even though I can fake it at times).

Do you have libgl1-mesa-dri (6.5.1) and libgl1-mesa-glx (6.5.1) installed? If not, please install them, put the bad X config back, and try to make it crash again.

Joey

Revision history for this message
Matthew Carroll (m-tthew) wrote :

Hi Joey

Yes, I have both libgl1-mesa-dri (6.5.1) and libgl1-mesa-glx (6.5.1) installed. They've been there at least since I ran updates, before doing all the testing.

Matthew

Revision history for this message
Matthew Carroll (m-tthew) wrote :

Yesterday I thought I could simply switch off glx and sidestep this bug, but just now I tried to run IE6 in wine (I use this for testing websites):

Xlib: extension "GLX" missing on display ":0.0".
err:wgl:X11DRV_WineGL_InitOpenglInfo couldn't initialize OpenGL, expect problems
err:wgl:has_opengl Intialization of OpenGL info failed, disabling OpenGL!

...so it seems I can choose: open office, or wine, but not both. (Time to boot up the old PI for IE testing.) Not being able to do all my work on the go means this bug just became a show stopper for me :-(

Revision history for this message
Matthias Klose (doko) wrote :

crash in /usr/lib/xorg/modules/extensions/libGLcore.so for non-standard setup

Changed in openoffice.org:
importance: Undecided → Medium
status: Needs Info → Confirmed
Revision history for this message
Matthew Carroll (m-tthew) wrote :

Just to confirm, this bug is still present now edgy proper has been released.

Revision history for this message
Matthew Carroll (m-tthew) wrote :

This bug (whatever the root cause was) appears to be fixed in a fresh install of Feisty 20070415.

Thanks,
Matthew

Revision history for this message
Matthias Klose (doko) wrote :

thanks for checking; closing the OOo task

Changed in openoffice.org:
status: Confirmed → Fix Released
Revision history for this message
Timo Aaltonen (tjaalton) wrote :

As it is fixed I'll close the xorg-server task as well.

Changed in xorg-server:
status: Unconfirmed → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.