Bug#297743: Two machines, oocalc OK on one, crashes on the other

[Mark Zimmerman <mark@foresthaven.com>]

Greetings:

[ Philipp Lohmann: please refer to the following to see why you were CC'd ]

<http://www.openoffice.org/servlets/ReadMsg?listName=users&msgNo=86626>

I have two machines, both running debian sarge and kernel 2.6.8, and
the following openoffice.org packages:

openoffice.org 1.1.3-9
openoffice.org-bin 1.1.3-9
openoffice.org-debian-files 1.1.3-8+1
openoffice.org-help-en 1.1+20040420-3
openoffice.org-hyphenation-en 20030813-3
openoffice.org-l10n-en 1.1.3-9

One of them works fine; the other crashes every time. Here is the
result of the crash:

----------------------------------------------------------------------
sh: line 1: crash_report: command not found
Xlib: unexpected async reply (sequence 0x395d)!


Fatal exception: Signal 11
Stack:
/usr/lib/openoffice/program/libsal.so.3[0x40bad3ec]
/usr/lib/openoffice/program/libsal.so.3[0x40bad579]
/usr/lib/openoffice/program/libsal.so.3[0x40bad644]
[0xffffe420]
/usr/lib/openoffice/program/libsal.so.3[0x40ba275f]
/lib/tls/libpthread.so.0[0x410ffb63]
/lib/tls/libc.so.6(__clone+0x5a)[0x4138dc4a]
----------------------------------------------------------------------

All I do to trigger the problem is bring up a blank spreadsheet in
oocalc and enter numbers into a few cells. The crash occurs after the
third or fourth number.

Here are some details about my testing environment:

1. I am using both of these machines (let's call them Working and
   Broken) remotely from a third machine.

2. I log into them via "ssh -X".

3. I deleted .sversionrc and .openoffice on both so as to start clean.

Since Working was installed with sarge from scratch and Broken started
out as potato, was upgraded to woody, and now is sarge, it was not quite
identical to the Working machine so I gradually tweaked it until
everything that I could imagine was relevant was identical. I verified
this state of identicalness as follows:

1. On both machines, I ran oocalc under strace like this:
   "OOO_DEBUG="strace -ostrace.out" oocalc
2. I extracted all of the shared library and font file names from the
   strace.out files and checked that both executions used the same
   files. (exception: Working has gnome installed so it pulled in a few
   more things. I am assuming that the absence of gnome is not the cause
   of the problem.)
3. I did md5sums of all of the referenced shared library and font files
   on the two systems and verified that all were identical.

I think I have ruled out any differences in the software environment as
causes of the problem. Working, however is a 2.4 GHz machine while
Broken is only a 500 Mhz machine. Since the problem is exhibited as an
async error and since the crash occurs in a call from libpthread to
libc, I am beginning to suspect a race condition in openoffice.

If there is anything else I can do to help track this down, let me know.
I may be in a unique situation here with one working machine and one
broken one running the same software.

-- Mark