[Yum] Re: [Beowulf] Athlon64 / Opteron test (fwd)

Jason L Tibbitts III tibbs at math.uh.edu
Sat May 15 06:06:38 UTC 2004


>>>>> "RGB" == Robert G Brown <rgb at phy.duke.edu> writes:

RGB> Obviously after FC1, which we're still using in production on our
RGB> Opterons.

Really?  For me it's so unstable as to be useful.  I just went to test
this out on the one Opteron I still have running FC1 only to find it
dead.

RGB> Maybe the solution is just to wait for FC2?

You could just try pulling the latest Yum package from Rawhide
(2.0.7-1.1, I think).  I'm running it on RH9 and FC1 without problems.

RGB> This naming scheme would work fine, especially if there were a
RGB> package group named something pithy like Compat_i386 that had all
RGB> the package.i386 libraries in it.

It's pretty weird.  I have a complete set of i386 packages installed
on my FC2T3 Opterons and, well, I'm not completely sure if everything
I did was proper.  Under FC1 you simply can't bring in all of the X
libraries because the different arch packages conflict.  This worked
better under FC2T3 but some things still give me hives:

> rpm -qf /usr/share/terminfo
ncurses-5.4-4.1
ncurses-5.4-4.1

Most packages aren't split into library-only packages, and so you have
various files installed twice.  But I can at least run pretty
complicated packages, like Matlab (which brings in Java and a bunch of
weird choices in what to link dynamically and what to link
statically), and they run very quickly.  Matlab still needs
LD_ASSUME_KERNEL=2.4.1, but it's needed that since Red Hat 9.

Sorry, this has drifted off-topic from the Yum list.

 - J<



More information about the Yum mailing list