[6943] in testers
Re: prevert.mit.edu: failed Linux 9.4 update
daemon@ATHENA.MIT.EDU (Greg Hudson)
Wed May 4 18:02:17 2005
Date: Wed, 4 May 2005 17:19:20 -0400 (EDT)
Message-Id: <200505042119.j44LJJDF005569@pyramids.mit.edu>
From: Greg Hudson <ghudson@MIT.EDU>
To: "andrew m. boardman" <amb@MIT.EDU>
CC: testers@MIT.EDU
In-reply-to: <200504290118.j3T1IEOg006467@pothole.mit.edu>
> 1) LILO was configured with both the 9.3 and 9.4 kernels, defaulting to
> 9.3, which lost badly for obvious reasons. (/etc/lilo.conf, however,
> listed only the 9.4 kernel; update.log shows what looks like lilo
> failing due to /dev/hda missing.)
I just finished an update on equal-rites and did not immediately
reboot. /dev has been eviscerated, containing only 13 entries (audit,
cdrom, console, gpmctl, initctl, log, logibm;3d64277f, logicalco,
logimouse;3d64277f, mouse, null, pts, shm). So yes, lilo presumably
can't run after the update for lack of /dev/hda.
> 2) The X configuration was pointing at /dev/mouse; udev provides only
> /dev/mouse0 as a pointer device.
fix-mouse-psaux is supposed to handle this case, but the code checks
for /dev/mouse being readable, and since it's a symlink to "psaux"
which no longer exists under /dev, it doesn't.
(Actually, the fix-mouse-psaux handling for this case is a little
bizarre. It repoints /dev/mouse at /dev/input/mice, which is odd
because I think the /dev/mouse symlink will disappear upon reboot.
Since it also changes the X configuration to point directly at
/dev/input/mice, that's not directly relevant.)
I don't know why anaconda upgrades wouldn't run into the same problem,
partly because I don't know what's cleaning out /dev. Perhaps
anaconda is recreating the /dev tree, but it would be weird for it to
recreate /dev/psaux since that device won't exist under the new world
order.
> 3) The X server exits but irreprably wedges the console when it doesn't
> have a valid pointer device.
Oddly, I can find nothing about this in Red Hat's bugzilla database,
even though it seems to happen on all hardware.
