OpenOffice.org2 regularly crashing without warning and no option for feedback.

[Caolan McNamara]

On Tue, 2005-11-15 at 15:35 +1100, Rodd Clarkson wrote:
> Hi All,
> 
> My wife uses OpenOffice.org (on an FC4 laptop that is up-to-date with
> udpates) as her main applications and recently (in the last week) she's
> been having problems where OOo just 'disappears' (her words) and offers
> no chance for feedback about what was happening.
> 
> She's loosing 15 minute chunks of work and this is happening a couple of
> times a day.
> 
> I'm on FC-rawhide and don't use OOo enough to have this problem.
> 
> Are others seeing this too?  How can I get some useful backtrace to
> report what the problem is?  gdb maybe?

I've gotten reports of this on FC-4, but have not been able to reproduce
it, I need a real stack trace. One problem is that the auto-stack trace
provider on FC-4 doesn't kick in for a trivial and fixed in rawhide
reason so I'm not getting the stack-traces on crash.

So to debug...

> gdb /usr/lib/openoffice.org/program/soffice.bin
(gdb) handle SIG33 nostop noprint
(gdb) handle SIGPWR nostop noprint
(gdb) handle SIGXCPU nostop noprint
(gdb) run -writer

and on crash

(gdb) bt.

Attaching such crashreports to https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=171873
would be appreciated.

It's crucial to know 

a) GNOME vs KDE
b) if KDE if "klipper" is running
c) if something else changed around the time the problem materialized
d) output of locale

Possibilities:
a) Some XError being caught by OOo and rethrown as an exception
terminating OOo. i.e.
     export SAL_IGNOREXERRORS=true
might make it stop.
b) copy/paste related problem related to klipper, though the patch for
that is in FC-4 OOo and should be fixed
c) gcc cast bug, http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22132,
though FC-4 is built with rh gcc 4.0.1 which has that fixed
d) opening non-existant icu locale data, shouldn't affect FC-4

C.