On Sat, May 08, 2004 at 07:07:56PM -0500, Steve Langasek wrote: > Hi Alexandre, > > Thanks for writing. It would be appreciated if you could file an > installation-reports bug about your experiences, following the > directions at > <http://www.debian.org/devel/debian-installer/report-template>. In the > meantime, I'll try to address some of your concerns below. I will. But since I'll be working on-site next week, this is going to be postponed until I come back. > > On Sun, May 09, 2004 at 01:09:54AM +0200, Alexandre Fayolle wrote: > > I got hold of an Alphastation 500 a few days ago and I'm trying to > > install Sarge on it (I thought I could maybe help testing the > > installer). While I'm quite at ease with Debian on intel machines, I'm a > > total newbie when it comes to Alpha machines, SRM and the like, though > > I've read the SRM howto. (I should probably read it a couple more times > > to get things going). > > > I grabbed the 5 mb cdrom iso image from > > /debian/dists/testing/main/installer-alpha/current/images/ tonight, > > booted, and started the installation process. > > > First problem was during hardware detection: sr_mod.o was not found, and > > I got an error message. > > The cdrom-mini.iso you grabbed is not the normal way of installing; this > image is essentially "netboot on a CD". The standard netinst and > businesscard images (which should both contain the sr_mod.o module) can > be found at > <http://cdimage.debian.org/pub/cdimage-testing/sarge_d-i/alpha/beta4/>. Ok, I'll try this one. This is a very minor annoyance, anyway. > > I got back to the partitionning step, removed the / partition, inserted > > a small 2MB partition formated as ext2, not mounted at the beginning of > > the disk, put back / to fit the remaining space, and relaunched the > > installation of the base system which failed twice. > > > The first failure was "dpkg: syntax error: unknown user 'root' in > > statusoverride file", and was caused by exim4. > > > The second error because /target/usr/bin/awk was an existing symling > > on mawk and ln is not called with the -f option. > > This seems to indicate that you did not re-format the partitions before > trying to reinstall? Now that you ask, I have a doubt. But I'm pretty sure I *did* ask partman to reformat. I'll have to check the logs when I reinstall. > > I removed the statoverride file and the symlink and relaunched the > > installation (twice) and finally got back to the installation of aboot > > which failed again: the installer asked me on which partition the > > bootloader should be installed, and proposed the only ext2 partition on > > my machine, namely /dev/sda1, I sayed ok, and it failed because "bootcode > > overlaps with partition #1. If you really want this, use -f1". > > > I read some doc, and issued manually "/target/sbin/swriteboot -f1 > > /dev/scsi/host1/bus0/target0/lun0/part1 /target/boot/bootlx" and this > > failed because "existing disk label is corrupt. Couldn't get a valid > > disk label, exiting". > > You don't really want this. The reason aboot gives you the option of > overlapping a partition is because Tru64 partition tables traditionally > have partition 3 as "whole disk", so it overlaps no matter *where* > you're writing. If you use the -f option pointing at a partition that > you have data on, you *will* corrupt the data on that partition. > > What you need to do is leave a small amount of space at the beginning of > your disk, unpartitioned, where aboot will write the bootloader. For > d-i, 1MB or so should be sufficient; if you plan to use aboot's rawboot > option, you would need to leave more space for a kernel. I'm not a partman expert, but leaving this "small amount of space" at the begining of the disk is not an obvious procedure, since partman doesn't enable the user to start at an offset, as far as I can remember. So, unless support for this is added in d-i, the user has to create a partition at the beginning of the disk, create the other partitions afterwards and the remove the first one, leaving the space unpartitionned. > It would be nice if the d-i partitioner (partman) also warned about > this, but I haven't figured out yet how to detect this problem. At the > very least, I intend for this to be mentioned in the install notes. Would be very nice to have both points covered, yes. I have to go and catch the train, but I'll be back next week to report on my progress. Thanks for your quick answer. -- Alexandre Fayolle http://alexandre.fayolle.free.fr/blog/
Attachment:
signature.asc
Description: Digital signature