Openoffice on Ubuntu x86 crashes when tables clicked

[Danny Staple]

May be similar to bug #39907.

In OpenOffice Base, simply clicking on "tables" in a newly created database is causing it to crash on my box.

If Oo is started on a command line, I get the following error:

 *** glibc detected *** free(): invalid pointer: 0xb6244678 ***

This is on Dapper 6.06 LTS using openoffice.org 2.0.2-2ubuntu12.1, vanilla gnome and xorg.

To reproduce (100% here):
* Click on OpenOffice Base, and go through the database wizard.
* Select Create new Database
* Leaving defaults click next, then finish.
* Give the database a filename (not changing the format from the default odb).
* You will now be on the forms tab. Click on the tables tab.
* OpenOffice will now hang, and will need to be physically killed. I mean all of it, including any other components you had open, and the quickstarter.

I also tried to reproduce this on Edgy, with openoffice.org 2.0.3-4ubuntu2, but openoffice.org crashed as soon as I clicked finish (the third step in this). It did offer to save my document though. This is aso 100%.

[John Vivirito (gnomefreak)]

Thanks for your bug report. Could you please try to obtain a
backtrace by following the instructions on
       http://wiki.ubuntu.com/DebuggingProgramCrash.
This will greatly aid us in tracking down your problem.

[Danny Staple (danny-orionrobots)]

First, the latter mentioned Edgy issue was reported as defect #60981.
Trying to get openoffice debug in Dapper fails:

$ sudo wajig search openoffice | grep -i dbg
$ sudo wajig search openoffice | grep -i debug

No results searching for a -dbg or -debug openoffice package.

I then tried to get the build-deps:
$ sudo apt-get build-dep openoffice.org2
Reading package lists... Done
Building dependency tree... Done
E: Build-dependencies for openoffice.org2 could not be satisfied.

$ sudo apt-get build-dep openoffice.org
Reading package lists... Done
Building dependency tree... Done
E: Build-dependencies for openoffice.org could not be satisfied.

No joy there. I then went right ahead to see if maybe I could just simply do:
$export DEB_BUILD_OPTIONS="debug nostrip noopt"
$sudo apt-get source -b openoffice.org

After waiting some time:

Reading package lists... Done
Building dependency tree... Done
Need to get 279MB of source archives.
Get: 1 http://security.ubuntu.com dapper-security/main openoffice.org 2.0.2-2ubuntu12.1 (dsc) [3266B]
Get: 2 http://security.ubuntu.com dapper-security/main openoffice.org 2.0.2-2ubuntu12.1 (tar) [206MB]
Get: 3 http://security.ubuntu.com dapper-security/main openoffice.org 2.0.2-2ubuntu12.1 (diff) [73.0MB]
Fetched 279MB in 5m25s (859kB/s)
dpkg-source: extracting openoffice.org in openoffice.org-2.0.2
dpkg-source: unpacking openoffice.org_2.0.2.orig.tar.gz
dpkg-source: applying ./openoffice.org_2.0.2-2ubuntu12.1.diff.gz
dpkg-buildpackage: source package is openoffice.org
dpkg-buildpackage: source version is 2.0.2-2ubuntu12.1
dpkg-buildpackage: source changed by Matthias Klose <email address hidden>
dpkg-buildpackage: host architecture i386
dpkg-checkbuilddeps: Unmet build dependencies: libpam0g-dev | libpam-dev libxaw7-dev tcsh sharutils libsane-dev libcurl3-openssl-dev | libcurl3-dev render-dev libneon25-dev libstartup-notification0-dev dmake (>= 4.3-1) libxkbfile-dev | xlibs-static-dev (<< 4.3.0.dfsg.1-15) libldap2-dev libsndfile1-dev libdb4.3-dev unixodbc-dev fdupes translate-toolkit (>= 0.8.rel) java-gcj-compat-dev (>= 1.0.28-8) libdb4.3-java firefox-dev (>= 1.5.dfsg+1.5.0.1-1ubuntu7) libnspr-dev libnss-dev (>= 1.7.5) libstlport4.6-dev (>= 4.6.2-3) libboost-dev libportaudio-dev libwpd8-dev (>= 0.8.3-3) libicu34-dev libxmlsec1-dev (>= 1.2.9) libglitz-glx1-dev mdbtools-dev (>= 0.5.99.0.6pre1.0.20050409-1)
dpkg-buildpackage: Build dependencies/conflicts unsatisfied; aborting.
dpkg-buildpackage: (Use -d flag to override.)
Build command ‘cd openoffice.org-2.0.2 && dpkg-buildpackage -b -uc’ failed.
E: Child process failed

So I manually try to see what I can do.
After installing most of the dependancies, and checking their version numbers, I came to the final group, and as a result filed defect #62065.

[Danny Staple (danny-orionrobots)]

Hmm - that link for the defect: bug #62065.

[John Vivirito (gnomefreak)]

you have to run an strace on openoffice packages. backtrace wont work.

[Danny Staple (danny-orionrobots)]

Since I am having ISP issues - posting this via email (gmail is one of
the few things that works!):

nanosleep({0, 20000000}, NULL) = 0
write(4, "C\2\5\0\1\0 \2\2\0 \2\f\0\f\1\240\1\6\0F\0\5\0\1\0 \2\2"...,
40) = 40
ioctl(4, FIONREAD, [0]) = 0
poll([{fd=5, events=POLLIN|POLLPRI}], 1, 0) = 0
nanosleep({0, 20000000}, NULL) = 0
write(4, "C\2\5\0\1\0 \2\2\0 \2\f\0\f\1\240\1\6\0F\0\5\0\1\0 \2\2"...,
40) = 40
ioctl(4, FIONREAD, [0]) = 0
poll([{fd=5, events=POLLIN|POLLPRI}], 1, 0) = 0
nanosleep({0, 20000000}, NULL) = 0
write(4, "C\2\5\0\1\0 \2\2\0 \2\f\0\f\1\240\1\6\0F\0\5\0\1\0 \2\2"...,
40) = 40
ioctl(4, FIONREAD, [0]) = 0
poll([{fd=5, events=POLLIN|POLLPRI}], 1, 0) = 0
nanosleep({0, 20000000}, NULL) = 0
write(4, "C\2\5\0\1\0 \2\2\0 \2\f\0\f\1\240\1\6\0F\0\5\0\1\0 \2\2"...,
40) = 40
ioctl(4, FIONREAD, [0]) = 0
poll([{fd=5, events=POLLIN|POLLPRI}], 1, 0) = 0
nanosleep({0, 20000000}, NULL) = 0
write(4, "C\2\5\0\1\0 \2\2\0 \2\f\0\f\1\240\1\6\0F\0\5\0\1\0 \2\2"...,
40) = 40
ioctl(4, FIONREAD, [0]) = 0
poll([{fd=5, events=POLLIN|POLLPRI}], 1, 0) = 0
nanosleep({0, 20000000}, NULL) = 0
write(4, "C\2\5\0\1\0 \2\2\0 \2\f\0\f\1\240\1\6\0F\0\5\0\1\0 \2\2"...,
40) = 40
ioctl(4, FIONREAD, [0]) = 0
poll([{fd=5, events=POLLIN|POLLPRI}], 1, 0) = 0
nanosleep({0, 20000000}, NULL) = 0
write(4, "C\2\5\0\1\0 \2\2\0 \2\f\0\f\1\240\1\6\0F\0\5\0\1\0 \2\2"...,
40) = 40
ioctl(4, FIONREAD, [0]) = 0
poll([{fd=5, events=POLLIN|POLLPRI}], 1, 0) = 0
nanosleep({0, 20000000}, NULL) = 0
write(4, "C\2\5\0\1\0 \2\2\0 \2\f\0\f\1\240\1\6\0F\0\5\0\1\0 \2\2"...,
40) = 40
ioctl(4, FIONREAD, [0]) = 0
poll([{fd=5, events=POLLIN|POLLPRI}], 1, 0) = 0
nanosleep({0, 20000000}, NULL) = 0
write(4, "C\2\5\0\1\0 \2\2\0 \2\f\0\f\1\240\1\6\0F\0\5\0\1\0 \2\2"...,
40) = 40
ioctl(4, FIONREAD, [0]) = 0
poll([{fd=5, events=POLLIN|POLLPRI}], 1, 0) = 0
nanosleep({0, 20000000}, NULL) = 0
write(4, "C\2\5\0\1\0 \2\2\0 \2\f\0\f\1\240\1\6\0F\0\5\0\1\0 \2\2"...,
40) = 40
ioctl(4, FIONREAD, [0]) = 0
poll([{fd=5, events=POLLIN|POLLPRI}], 1, 0) = 0
nanosleep({0, 20000000}, NULL) = 0
write(4, "C\2\5\0\1\0 \2\2\0 \2\f\0\f\1\240\1\6\0F\0\5\0\1\0 \2\2"...,
40) = 40
ioctl(4, FIONREAD, [0]) = 0
poll([{fd=5, events=POLLIN|POLLPRI}], 1, 0) = 0
nanosleep({0, 20000000}, NULL) = 0
write(4, "C\2\5\0\1\0 \2\2\0 \2\f\0\f\1\240\1\6\0F\0\5\0\1\0 \2\2"...,
40) = 40
ioctl(4, FIONREAD, [0]) = 0
poll([{fd=5, events=POLLIN|POLLPRI}], 1, 0) = 0
nanosleep({0, 20000000}, NULL) = 0
write(4, "C\2\5\0\1\0 \2\2\0 \2\f\0\f\1\240\1\6\0F\0\5\0\1\0 \2\2"...,
40) = 40
ioctl(4, FIONREAD, [0]) = 0
poll([{fd=5, events=POLLIN|POLLPRI}], 1, 0) = 0
nanosleep({0, 20000000}, NULL) = 0
write(4, "C\2\5\0\1\0 \2\2\0 \2\f\0\f\1\240\1\6\0F\0\5\0\1\0 \2\2"...,
40) = 40
ioctl(4, FIONREAD, [0]) = 0
poll([{fd=5, events=POLLIN|POLLPRI}], 1, 0) = 0
nanosleep({0, 20000000}, NULL) = 0
write(4, "C\2\5\0\1\0 \2\2\0 \2...

[Matthias Klose (doko)]

still needs info. you have to upgrade edgy and recheck with the current openoffice.org package

[Danny Staple (danny-orionrobots)]

Matthias, as stated on the original report. This cannot be reproduced in Edgy, as the Edgy version of office crashes at an earlier point. See bug #39907.

Please ask if there is furthar info from the Dapper version that I can provide though, as the issue is still 100% reproducible there.

[Huw (huw-synapticsilence)]

I don't have the time to provide technical info right now, but I just want to say that this is 100% reproducible on my Dapper system with OOo 2.0.2 as well. In fact, it's what brought me here.

[John Vivirito (gnomefreak)]

can you reproduce this in OO.o 2.0.3 its in the dapper-proposed repos?

[Huw (huw-synapticsilence)]

Yes, I'm sorry to report that exactly the same thing happens in 2.0.3.

[Danny Staple (danny-orionrobots)]

I can confirm Huws report. I also still see that in 2.0.3.

[Huw (huw-synapticsilence)]

[Views -> database objects -> tables] also crashes the application.

[Matthias Klose (doko)]

I'm unable to reproduce this in the current 2.0.4~rc3 packages in edgy. are you able to recheck with a Live-CD?

[Huw (huw-synapticsilence)]

I will do so tomorrow. However, I just tried switching to the nv driver and tables now works fine (although it takes about 10 seconds to bring up the table view). I was previously using the nvidia driver. My card is a GeForce FX5200.

[John Vivirito (gnomefreak)]

I cant reproduce this on my dapper with 2.0.3 installed i cant with edgy version fully up-to-date nor can i reproduce this with ubuntu edgy livecd beta1

[Huw (huw-synapticsilence)]

Which graphics driver are you using, John? Did you try the nvidia driver? That's what seems to be the problem here, or at least part of the problem.

[Danny Staple (danny-orionrobots)]

Ah, now that is interesting. I am also using an NVidia driver.

Though what is it that OpenOffice should be doing that is so low lever/graphics intensive that this should be at all relevant?

We are simply talking about standard forms and GUI widgets, not a graphical application with 3D or extensive low level use of graphics hardware.

Now OpenOffice is Java based, as is Eclipse, and I have seen a problem where Eclipse on Ubuntu crashes when scrolling horizontally which was also related to the NVidia driver. Is there something that the java interpreter does specifically with the NVidia driver to trigger this?

Now I know I have installed the sun-java5-jre 1.5.0-06-1 package and java-gcj-compat 1.0.56-0ubuntu1. Which would OpenOffice be using?

[Danny Staple (danny-orionrobots)]

Sorry - my /etc/alternatives/java is pointing to the sun jre, although sun-java5-jre 1.5.0-06-1 is not in the dependancies for openoffice.org-base, while gcj-compat or j2re are.

[John Vivirito (gnomefreak)]

I use nvidia on all my pcs. i use the beta drivers on one edgy pc and still dont have this issue all othere drivers are the nvidia-glx

[Matthias Klose (doko)]

DannyStaple schrieb:
> Now OpenOffice is Java based, as is Eclipse, and I have seen a problem
> where Eclipse on Ubuntu crashes when scrolling horizontally which was
> also related to the NVidia driver. Is there something that the java
> interpreter does specifically with the NVidia driver to trigger this?

java is only used in some places; in some of the wizards to create
documents, and in oobase. other places are not affected by java problems.

> Now I know I have installed the sun-java5-jre 1.5.0-06-1 package and
> java-gcj-compat 1.0.56-0ubuntu1. Which would OpenOffice be using?

it uses gcj as the preferred alternative; you can set another runtime in
the OOo preferences.

  Matthias

[Danny Staple (danny-orionrobots)]

Tested this today on Edgy, and while it took some time to hit the tables screen, it did not crash.

However, this does not confirm if there is an NVidia issue as I have Egdy running in VMWare at the moment.

Note that OOBase is specifically the area that is crashing here - so a Java issue really could be the problem here.

If I change the runtime to a different one on the dapper box (the sun jre - wow that is faster!) with the NVidia card, then it works albeit slowly.

So the package with the fault is actually java-gcj-compat.

[Matthias Klose (doko)]

reassigning to openoffice.org again, until we have a more concrete bug report.

[Matthias Klose (doko)]

> If I change the runtime to a different one on the dapper box
> (the sun jre - wow that is faster!) with the NVidia card,
> then it works albeit slowly.

please could you install openoffice.org-gcj an recheck the time? make sure that you reenable the GNU runtime.

[John Vivirito (gnomefreak)]

Im thinking if it was a nvidia issue X would crash not just OO.o

[Danny Staple (danny-orionrobots)]

Now that is interesting. i do have a package openoffice.org-gcj which is not installed. If I install it, I first hit problems with having enabled the "proposed" version 2.0.3 since this package only suggests - so first the lengthy process of removing it, and the language files, and then removing the proposed repo, then installing the openoffice packages again with the version (from security in fact).

If there is a better way to downgrade the whole set of packages (taking into account that openoffice.org is just a metapackage for all of its components), please let me know.

I then installed openoffice.org-gcj, and started up the database wizard - on dapper, on a machine with an nvidia card and driver.

Ho hum - slooooow again...
Create a new database, go through the wizard, click on tables, POP! Its dead, even with openoffice.org-gcj installed.

Having killed the hung version, I then started it again with a writer document (safe), set the jvm to the sun jre again, and restarted it. Cue speed increase... Create Database, go through wizard, click on tables, and after a 30 second pause - no crash.

I would say that is fairly conclusive that either gcj is at fault, or the way that openoffice handles it.

[Matthias Klose (doko)]

please recheck with the 2.0.4-0ubuntu2 version; openoffice.org-gcj is installable.

[jon latorre (moebius-etxea)]

I have similar problem (selectin tables in Base makes OOo to crash). Reading the comments about java I chekout what java runtime was using OOo. I have instaled SUNs jre an OOo was using it. Changing it to the GNU jre solved the problem.

[Matthias Klose (doko)]

39907 was fixed in feisty; please reopen if that is not fixed in the feisty build