ww8export tinderbox failure ?

[Lubos Lunak]

On Wednesday 26 of September 2012, Miklos Vajna wrote:
> On Tue, Sep 25, 2012 at 12:16:25PM +0100, Caolán McNamara 
<caolanm at redhat.com> wrote:
> > The continual crash on the ww8 export on two of the Linux tinderboxes,
> > what's up with that ? I think llunak mentioned he was able to reproduce
> > it sometimes ?
> >
> > I guess reproducing the Gentoo env locally is probably impossible, but
> > is there e.g. a virtualbox image available of the right OpenSUSE version
> > that the failure can be seen on ?
>
> I'm almost sure it's a problem with a highly-paralellized build (i.e.
> some library or component files are always there when I build with -j4,
> but with more threads some missing dependency problem can be triggered).

 No. 
Plain 'make /local/home/tinderbox/master/workdir/unxlngx6.pro/CppunitTest/sw_subsequent_ww8export.test' 
in sw/ randomly does or does not trigger the problem. The race condition is 
very likely in the execution.

 One thing that is special about those two tinderboxes is that they both have 
a relatively old gcc version, so I'd expect using any system with similarly 
old components might be useful for reproducing it.

> Getting a backtrace with symbols would indeed help, I guess.

 Didn't help me. IIRC the crash is because a component somewhere is not 
loaded, leading to a NULL pointer dereference later. And the component is not 
loaded, because the GUID read for it from some OLE stream is rubbish. I got 
rather lost at that point and gave up in the hope somebody else would fix it.

-- 
 Lubos Lunak
 l.lunak at suse.cz