All of the errors reported saying that during an rpm upgrade things start segfaulting seem to result from the post scripts in any package that calls ldconfig. This is happening since glibc-2.3.1-26 (after -21) in rawhide. Just issuing "ldconfig" results in ALL binaries crashing. Rebooting from Phoebe CD1 and reinstalling glibc with --noscripts results in a working system. Could this be happening to people who have a modified /etc/ld.so.conf file? Since this is a configuration file it is not replaced during the upgrade. I am in sync with rawhide of Jan 13, 2003.
I certainly cannot reproduce anything like this. What exactly do you have in /etc/ld.so.conf? How do the crashes look like? Do all programs crash, or just some? Say if you run /bin/echo.
Yes. I just did it again and /bin/echo gives "Segmentation fault" as well. Clue may lie in upgrade process. I have a notebook with all the same software and there is no problem there. The difference was: While doing the first upgrade to TLS based glibc/kernel/rpm I did it from the command line and not from anaconda. The upgrade crashed half-way during the glibc-2.3.1 upgrade (glibc-common was fine) and went into the ldconfig problem mentioned above. I than booted up from Phoebe1 CD and tried to install from "linux rescue" mode. This seems to have some incompatibility with the current rawhide rpm databases, so the only way I could do it is to go back to older rpm and do the upgrades that way. The notebook was upgraded straight from anaconda.
I have fixed this problem by reinstalling from an appropriate boot CD. It seems to be a problem if one tries to install from RH8.0 to current glibc without using anaconda and/or proper rpm/kernel. I am closing this as solved in rawhide.
*** Bug 80948 has been marked as a duplicate of this bug. ***