This was fun. Well, ok, it wasn’t. But it works now.
The ofa_kernel rpm crashes and burns being rebuilt on FC9. As do sdp, rds, ibutils, and dapl. Fine. Also have to downgrade tcl to 8.4 from 8.5. Because the RPMs hard-link to a specific library in tk (which depends upon tcl). Again, fine.
Run the install.pl script. Select install. Select customize. Select everything but those things. Go get coffee.
Note that the conversion to gcc 4.3 is going to be painful for OFED, and I have the logs to prove it. 🙁
The issue in this case is that they are using RPM as a build environment, and not as a simple packaging environment from a working build (./configure ; make ; make install).
Sounds similar to the basis of my rant from yesterday … And a previous rant on (ab)uses of RPM for OFED, and why it is broken (not OFED, but the build environment). Very similar problems. Very similar results.
This isn’t an HPC issue. It is an issue of reliable building and packaging. And the need to separate the reliable build from the reliable packaging layer. Lest you wind up with a massive layering violation (to quote Andrew Morton and others), and a next-to-impossible-to-fix situation.
Which is where we are now.
Viewed 14721 times by 3269 viewers