[PlanetCCRMA] problem with new redhat glibc

[Fernando Pablo Lopez-Lezcano]

> I've found that by booting with old 2.4.20-8, everything seems to be
> working fine.

Yes, that should work, it is an nptl enabled kernel. 

> There's also a new update from redhat.  Maybe that fixes the problem.  Dare I?

Where do you see it? Looks like it has not been yet officially released
(I'm looking at the errata pages for 9). 

-- Fernando

> > > > Yuck! Did you try with more recent Planet CCRMA kernels? What kind of
> > > > error did you see happening? I wonder if they (by mistake) released a
> > > > version with the non-nptl compatibility stuff...
> > > 
> > > I didn't try it with a more recent kernel. The errors were:
> > > 
> > > Mozilla: segfault right after a clone
> > > rpm: can't open rpm database. rpm does use a bunch of threads
> > > gnome: error creating a process
> > > 
> > > So, yeah, it does seem to be NPTL-related.
> > 
> > We are not alone, see:
> >   http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=109904
> > 
> > He does not make any reference to kernels so this may be a generic
> > problem. Did you try to boot into the RedHat kernel? It would be
> > interesting for RedHat to know if it is working in the RedHat kernel and
> > not working on the Planet CCRMA. 
> > 
> > There is a workaround in the bug report for (at least) getting rpm to
> > work again:
> > 
> > Leonardo Rodrigues wrote:
> > > Doing some research, I found that doing:
> > > export LD_ASSUME_KERNEL=2.4.1
> > > would solve, for example, the 'rpm' problems. Everything worked fine 
> > > with LD_ASSUME_KERNEL but stops working after 'unset 
> > > LD_ASSUME_KERNEL'.
> > > 
> > > Anyway, I've downgraded (rpm -Uvh --force) glibc to glibc-2.3.2-
> > > 27.9.i686.rpm and assume this is not right. I had to use 
> > > LD_ASSUME_KERNEL=2.4.1 for being able to downgrade glibc RPMs.
> > 
> > So, for the moment, don't upgrade to the latest glibc from RedHat!!
> > Again, kudos to Greg for the very fast warning. 
> > -- Fernando