openoffice build loops - seemingly only on SMP machines

Hi all,

instead of just posting an "mee too" into freebsd-openoffice...
This is a thing that might interest you people here, so why shouldn't
you get the info and at least think about it a bit?

http://lists.freebsd.org/pipermail/freebsd-openoffice/2008-February/003544.html

This is the discussion, and as it happens to a couple of people, it
seems (more or less) reproducible.
I have compiled openoffice 2.3.1 on my system without problems - at
that time the system had only 1 cpu (and also no SMP kernel) and was
at RELEASE-5.5. 
When compiling it again for RELEASE-6.3, and now with a 2nd CPU,
I got exactly the problems as described in the link (and its
followups). So, I do not say this is an SMP problem, I only say it
could be - with some likelihood.

And three things in this I find interesting. That "idlc" seems to be
just some kind of compiler. One of the posters mentions that it
needed a "kill -9" to stop it, and I experienced the same thing -
this is unusual for a compiler.
The other thing is, when my build did loop, and I looked at the
loadavg icon, it showed one cpu being fully occupied with only 
system(!) time.
And the third is, as the other posters mentioned, the problem
happens at some random point. 

Now I am rather new to SMP operation - but when putting this picture 
together, it looks not so much like a problem of openoffice or of 
the idlc, and more like some strangeness happening in the kernel 
and only getting exploited by the openoffice build.

So, as I said,  this is no complaint - it is just an invitation to 
think about the phenomen.

rgds,
PMc
_______________________________________________
[e-mail] mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-smp
To unsubscribe, send any mail to "[e-mail]"