Bug 150517 - OpenOffice.org Package Manager segfaults with gcj
Summary: OpenOffice.org Package Manager segfaults with gcj
Status: RESOLVED DUPLICATE of bug 153386
Alias: None
Product: SUSE Linux 10.1
Classification: openSUSE
Component: OpenOffice.org (show other bugs)
Version: Beta 4
Hardware: All Other
: P5 - None : Normal (vote)
Target Milestone: ---
Assignee: Michael Meeks
QA Contact: E-mail List
URL:
Whiteboard:
Keywords:
Depends on: 153386
Blocks:
  Show dependency treegraph
 
Reported: 2006-02-13 17:06 UTC by Petr Mladek
Modified: 2006-03-29 09:26 UTC (History)
2 users (show)

See Also:
Found By: Other
Services Priority:
Business Priority:
Blocker: ---
Marketing QA Status: ---
IT Deployment: ---


Attachments
Backtrace from our package. (8.38 KB, text/plain)
2006-02-13 17:07 UTC, Petr Mladek
Details
Problematic add-on package (2.20 KB, application/octet-stream)
2006-02-13 17:08 UTC, Petr Mladek
Details
Working add-on package. (2.07 KB, application/zip)
2006-02-13 17:09 UTC, Petr Mladek
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Petr Mladek 2006-02-13 17:06:33 UTC
OpenOffice.org has its own package manager. It helps user to add extra UNO-based OOo extensions. The Package Manager seqfaults on one package from the qatesttool.

How to reproduce the problem:
1. start ooffice
2. Select Tools/Package Manager...
3. Press the button Add
4. Select the attached capitalise.uno.pkg
5. Press the button Open

Then I get segfault. It is 100% reproducible with fresh OOo configuration and when gcj/gij is used as the JRE.

It does not happen with another package Basic_Addon.zip. Though, after you try to add the problematic capitalise.uno.pkg, the OOo UNO-package database is somewhat broken and OOo crashes when you try to access Tools/Package Manager.../My Packages.

Note: I am not sure. I think that this is the root of many errors reported by the qatesttool.

It happen also with the upstream package. However, it is specific to the gcj/gij that we use by default.

I'll attach a backtrace from our package.

Michael, how is it crytical?

What do you think? Is it a bug in OOo or in gcj?
Comment 1 Petr Mladek 2006-02-13 17:07:58 UTC
Created attachment 68128 [details]
Backtrace from our package.
Comment 2 Petr Mladek 2006-02-13 17:08:40 UTC
Created attachment 68130 [details]
Problematic add-on package
Comment 3 Petr Mladek 2006-02-13 17:09:05 UTC
Created attachment 68131 [details]
Working add-on package.
Comment 4 Richard Biener 2006-02-13 17:13:35 UTC
We don't support gcj and/or the java runtime.  Please try to reduce this to a sane testcase and report it upstream.
Comment 5 Michael Matz 2006-02-13 17:35:43 UTC
Well, Petr, it clearly shows that segfault being in the garbage collector.
At the point where it segfaults it can normally only happen if something
did randomly overwrite memory.  Perhaps valgrind is of some help?
Comment 6 Petr Mladek 2006-02-13 17:58:50 UTC
Just note that this bug is not as crytical for the testtool as I thought. The many errors reported by the testtool are caused by the test qatesttool/framework/basic/f_basic_protected_libraries.bas. It is broken and is ignored even by upstream.

This does not mean that we should ignore this bug :-)
Comment 7 Richard Biener 2006-02-14 11:26:26 UTC
Current stable (gcc-4.1.0_20060210-4) has a fix for gcc.gnu.org/PR13212, an issue that looks maybe related to your problem.  Can you make sure the problem reproduces with gcj from STABLE?
Comment 8 Petr Mladek 2006-02-14 14:57:15 UTC
I have found another crash that looks similar to gcc.gnu.org/PR13212.

Do you have the new gcc rpms available anywhere? They are still not available in autobuild. They probably were not checked in because of beta4.
Comment 9 Richard Biener 2006-02-14 15:56:58 UTC
They should be used by autobuild - at least I see packages being build with that version.
Comment 10 Petr Mladek 2006-02-15 19:00:44 UTC
They were in STABLE-BETA. I rather rebuilt the packages for STABLE via autobuild. Unfortunately, it did not help.

We should investigate it more. I could look at it later. Though, I am still investigating the other errors found by the qatesttool. I want to be sure that there is not any other more urgent bug. This could take some days.
Comment 11 Petr Mladek 2006-03-01 11:30:05 UTC
It is probably the same problem as in the bug #153386
Comment 12 Petr Mladek 2006-03-29 09:26:12 UTC
It helps to preload libgcj from the latest libgcj package => it is a duplicate of the bug #153386 => it will work on SL10.1-rc1/ NLD9-beta9.

*** This bug has been marked as a duplicate of 153386 ***