With up2date-2.4.6-1, the lilo.conf updating doesn't seem to be quite correct since it always tries to add an initrd to the lilo.conf even if one isn't needed (and thus isn't created). Thus the test lilo run fails and you have to manually do the lilo.conf updating for the kernel package update.
This defect considered MUST-FIX for Florence Gold
can you attach a copy of the original and the created lilo.conf? up2date should always run `mkinitrd`, but with the "--if-needed" flag.
looks like I was assuming that `mkinitrd` returned a different error code when it existed because an initrd wasnt needed, and when it was need and created it successfully. Looking though the mkinitrd script would seem to indicate this never should of worked, but seemed to work in practice (the error code was determined empirically, apparently incorrectly...). Now the code checks for existense of the initrd after calling `mkinitrd` to see if it did infact create one, since this seems to be the only way to tell. Should be a fix in cvs for testing now...
tested, fixed in 2.4.7.