FYI,
we have rolled out pretty much everything of the U5 updates at CERN and
have seen only minor issues:

* a CASTOR plugin compiled against a (proprietary) library apparently
crashed after the update of libstdc++ and had to be recompiled. No
further details.

* firefox updates during a running session will cause problems with
plugins (expected, plugin directory changes underneath the running
application).

* occasional kernel Oopses for SMP farm machines at boot, where the
second CPU takes ages to come alive. Unrelated to U5, as this started
with a security update earlier, and may be related to (long) serial
console cables after all...

* GNOME/gconfd-2 can still cause an AFS Oops/panic thanks to the lock
file creation/removal cycle. An slightly improved version of
SL_startgnome_afs now tells all GNOME applications to use local lock
files, before we only set things up for a full GNOME session.
http://lxnfs6/cern/slc30X/updates/testing/i386/http://lxnfs6/cern/slc30X/updates/testing/i386/SL_startgnome_afs-1.1-1.noarch.rpm
http://lxnfs6/cern/slc30X/updates/testing/SRPMS/SL_startgnome_afs-1.1-1.src.rpm

The libstdc++/CXXABI problem reported earlier ("..POOL-1.x crashes..")
is fixed with this release. Newer POOL versions already didn't suffer
from this, but we had agreed to stay on the U2 version of libstdc++ for
a few more months. This is no longer required, we use the U5 version
now.