Re: update: arlad dumps core under load

david l goodrich <[email protected]> writes:
> Tomas Olsson wrote:
> > If you want to stress it some more you could always play around with
> > arla-obj/tests/run-tests.
> 
> hmm.  any suggestions as to which would put arla under the highest load?
> 
Depends on what you like.  build* or untar-emacs if you like I/O (or just a
grep -r in some tree close to you).  A few getdents-and-unlink* in parallel
is sometimes fun.  I'm not sure what kind of operations are needed, I would
guess just thread intensive things.

> > Just a thought. Could this be the kernel bug mentioned here before? See 
> > https://lists.stacken.kth.se/pipermail/arla-drinkers/2006-January/003774.html
> > and the reply.
> 
> yes, it seems to be the same bug, the registers are, in fact, mostly 0.
>  This is why i have posted this to both arla-drinkers and the netbsd pr
> system.
> 
> But like i said, if i compile it with -g -O0, it doesn't seem to fail.
> That's the part that confuses me.
>
I have no idea, and I don't know in what way the generated code may be
different.  Thread things that don't get inlined?  

Was the crash very easy to trigger before?

/t