I'm resending the following, which dropped into the list server's bit bucket this past weekend...
--Bob ----- Forwarded message from Bob Drzyzgula <[EMAIL PROTECTED]> ----- Short version: I haven't yet got it to work. Sorry for the lack of some detail, it took a while for my cooker list subscription to clear through, and I don't have access to the machine again until Tuesday (it's at the office). Hardware: * Shuttle SV24, using all onboard interfaces. http://www.spacewalker.com/sv24.htm * Two Fujitsu UDMA drives, one 30GB (Pri master), one 20GB (Pri slave). * LS120 and LG 12x4x32/8X combo CD-RW/DVD on sec. channel. * 512MB PC133 memory, one stick * Via Cyrix III 866MHz processor ISOs downloaded from Rutgers. Md5sums checked out. Burned to CD-R using xcdroast on a Suse machine and a Plextor SCSI burner. Booted target machine from CD1. First try: ---------- Expert Mode. Blew away everything on the drives, in fact hda was brand new. Broke up the drives extensively, everything XFS. Hda had separate partitions for /, /boot, swap, /usr, /var, /tmp, /home, while hdb had partitions for /opt, /var/ftp, /var/www, and /data (I'm an old-school Unix kinda guy). I selected all top-level package choices, and went through picking a bunch more individual packages; total space requirement listed was something like 2700MB. First problem: I got an error message stating that there was an error installing XFree86-4.2.0-24mdk. I opted to continue. Second problem: When I got past the package installs, mkinitrd failed. I belive that this has been reported by others. I allowed it to continue to completion, but on reboot the BIOS reported "missing operating system". I rebooted into rescue mode using CD1 to see if I could fix the Lilo problems: First rescue try: I had it find and mount my partitions, and then I asked for a console. It in fact found and mounted the partitions and started a console, but I got an "illegal instruction" (SIGILL) error when trying to execute anything from the command line: df, ls, ps -- nothing worked. System was hosed, time to reboot. Second rescue try: I went to the console without mounting the partitions. From there, I could run various commands such as "df". Then I tried mounting the root partition (hda5) on /mnt. From that point on, everything I tried (including umount /mnt) gave another SIGILL. Didn't matter what my cwd was or if I gave relative or absolute paths. The system was again hosed. Third rescue try: Again booting to rescue mode off CD1, again entering the console at first opportunity. I tried doing a fsck.xfs on hda5. It returned immediately. I ran repair_xfs against it. It ran to completion but didn't report any repair activity. I again mounted it on /mnt, and again started getting the SIGILLS, even when using full paths for commands located outside of /mnt. Rebooted again. Fourth rescue try: At the console, I created a new top-level mount point to use instead of /mnt. I called it -- no, you don't want to know what I called it. Say I called it "/foo". I mounted root (hda5) onto /foo. At this point, I was still able to run commands. I looked for a lilo.conf in /foo/etc, but there was none there. I mounted the rest of the filesystems by hand, e.g. /usr on /foo/usr. Then I tried to chroot to /foo, so that I could fix up lilo by hand. No joy -- chroot would SIGILL every time. It didn't matter if I ran the copy of chroot in the ramdisk or the copy on the harddisk. It didn't matter if I used relative or full paths for /foo. It didn't matter if I let the command default to /bin/sh, or if I specified /bin/sh or /foo/bin/sh or /foo/bin/bash, or if my cwd was /foo when I did the chroot, it would SIGILL no matter what I tried. Notice that, as an improvement over the previous three attempts, it was only chroot that was SIGILL-ing -- other commands were working fine. Fifth rescue try: I booted again from CD1, but this time told it to do an upgrade of what was on the hard disk. It ran through the whole thing pretty quickly. It found my old settings for things like the IP address, but it again reported failures installing XFree86-4.2.0-24mdk and running mkinitrd. Second try ---------- I ran through the whole install again, this time changing the type of all the filesystems to ext3 instead of XFS. Again I did an expert install and selected a pile-o-packages. I had the same behavior re XFree86 and mkinitrd. Given my previous luck, I didn't try to repair the install. Third try --------- I ran through a recommended install. I let the installer pick absolutely everything. I let it take over hda and partition it however the hell it wanted to. I took the default choices for packages. I didn't change any defaults unless I had to -- I had to tell it which drive to use, and I had to give it a static IP because I don't have DHCP set up on my local net. But other than those minor things, I just let it go. Again, no joy. Same failures. Hope this helps. I remain utterly frustrated in my attempts to install 9.0Beta4 on that machine. If anyone has any suggestions, I'll try again on Tuesday. BTW, I now see the messages in the archive on the mkinitrd problems, I'll see if any of the information there helps on the bootloader problems. --Bob Drzyzgula ----- End forwarded message -----
