From Bugzilla Helper: User-Agent: Mozilla/4.73 [en] (X11; U; Linux 2.4.0-XFS i686) Description of problem: I have a RedHat 7.2 system with both the RH7.1 2.4.2 and RH7.2 2.4.7 kernels. Boot into 2.4.2 and gdm works for everyone. Reboot into 2.4.7 and only UID=0 accounts can get passed gdm. Harddrive swaps hard for 60 seconds before Xserver dies and gdm recovers with a login dialog again. All this is AFTER the pam module patch was installed. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1. Install Redhat 7.2 with NIS configured. 2. Update nsswitch.conf with "files nis" for passwd, shadow and group. 3. Apply latest pam_unix patches to fix major login problem. 4. Reboot and login with UID!=0 account and X stalls w/ mad swapping Until gdm kills xserver and starts with login dialog again. Actual Results: While the Xserver is swapping ctrl-alt-backspace the xserver, login at the text console as root and you will see dozens of RunWM -help calls and one RunWM -FVWM95 (Fvwm95, where did that come from???). If /usr/bin is not in your path, then exec gnome-session and exec startkde fail, leaving Expected Results: Users with UID!=0 should be able to login. :) Additional info: /etc/X11/xinit/Xclients needs PATH=/usr/bin:$PATH or gnome-session is not found for non-root users. If you don't have Fvwm95 (the last ditch WM) then RunWM goes into an infinate loop fork()'ing RunWM -help calls. Because of the dual kernel boot setup it is clear that the kernel is the start of the problem. `login` is not setting PATH or something is killing /bin and /usr/bin from the PATH. Regardless of the cause PATH=/bin:/usr/bin:$PATH should be there.
PATH is supposed to come from /etc/profile I believe. I don't have any initial idea on where it is getting lost; no one else has reported this issue. Maybe there are issues with not finding user home dirs or user shells due to the info in NIS... I'm just guessing though.
I'm not going get get anywhere on this without a way to reproduce it. Infinite loop in RunWM should be fixed though, so moving to xinitrc (Probably fvwm95 should be cleaned out of there, and RunWM should just bail to twm.)
Does this problem occur with RHL 7.3 and/or the current Limbo beta, or is it resolved now? Please update.
Assuming the problem is resolved now due to lack of response. If problem continues, reopen bug report and provide the details and answers requested above. Thanks.