Created attachment 111421 [details] Don't remove /tmp/ks.cfg.new
Description of problem: When defining a custom installclass in the ks.cfg, the Kickstart method unlinks the ks.cfg.new long before setInstallData (hence also readKickstart) is called. This, of course, leads to a traceback when readKickstart tries to open the long-gone file. Version-Release number of selected component (if applicable): anaconda-9.1.4.1-1.RHEL How reproducible: Always Steps to Reproduce: 1. define a basic subclass of KickstartBase using %installclass in ks.cfg 2. try to use it 3. Actual results: Traceback - I can get the exact traceback on request, but all you need to do is look at the code and you'll see it. Expected results: No traceback Additional info:
FWIW, I wouldn't recommend counting on this functionality. It's undocumented from what I remember and very likely to go away (I know for a fact that it's never tested :)