Comment 16 for bug 476234

Further to my previous comments, I have corrected the silo problem, and created a bootable SPARC CDROM.
The machine now boots, and the installer runs.
However, when it tries to detect the CDROM from which it is running, it fails to find it.
Going to the Alt-F2 window, and looking in /dev reveals no devices for cdrom
or hda/hdb/hdc. Adding these by hand with mknod, using major and minor numbers from the x86 version does nothing, so I assume that the installer reads and caches the contents of its original /dev.
An alternate thought is, that it’s looking for UUID entries in /dev/disk/by_xxx, which also doesn’t exist.
I have repeated this exercise for installation CDROM's for 9.04, 9.10, 10.04 server, and 10.04 KUbuntu,
with identical results (except the 9.04 also fails to load 'kernel modules'.
It seems to be significant, that all of these releases use the UUID approach to disk drives.
For what it's worth, upgrading 6.06 ->8.04 -> 10.04 seems to work, except that the system is left unusable, since SILO can't find vmlinuz and, if you boot from 2/vmlinuz, it runs to the point where it tries to mount the root disk, and neither root=/dev/hda1, nor the gobbledy-gook UUID is acceptable - irrespective of what is entered in /etc/fstab.
I can't believe that so many releases can be put into circulation, when not one could possibly be installed.