Bug#361180: Debian installer doesn't install kernel
On Fri, Apr 07, 2006 at 08:21:52AM +0200, Geert Stappers wrote:
> Tags 361180 moreinfo
> thanks
>
> On Fri, Apr 07, 2006 at 12:23:00AM -0400, Jack Carroll wrote:
> > Package: installation-reports
> >
> > Debian-installer-version: www.debian.org/devel/debian-installer 2006-0406
> > daily build
> > uname -a: Not available. Boot failed.
> > Date: 20060406 1800
> > Method: <How did you install? Network, with 3C905 card, DHCP
> > What did you boot off? Floppies
> > If network install, from where? ftp.us.debian.org Proxied? No>
> >
> > Install command was:
> > expert hw-detect/start_pcmcia=false
>
> What is the reason for using 'expert'?
I've been doing it that way for so long, it's reflex now. I want to
make sure it doesn't try to do something like partitioning my hard disk
without asking how I want it partitioned. That would be really upsetting.
>
> Please try
> install26 hw-detect/start_pcmcia=false
> or just
> install26
>
> >
> > Machine: HP Vectra VL
> > Processor: Pentium 2
> > Memory: 128 MB
> > Root Device: /dev/hda2
> > Root Size/partition table:
> > / /dev/hda2 5 GB
> > /var /dev/hda6 3 GB
> > /usr/local /dev/hda8 1 GB
> > swap /dev/hda9 512 MB
> > /data /dev/hda10 54 GB
>
> Here I miss /dev/hda1
> Any idea where that partition went?
That's the / partition for Sarge. That's the known good distro,
which I want to keep working, while testing other distros on the other two
sets of partitions. I have the drive partitioned to boot 3 operating
systems, with a common /usr/local partition and a common /data partition.
(I got away from a common /home partition, because different distros
sometimes need different files in their home directories. So now I give
each user a directory in /data, and point to it with a symbolic link in ~,
in each /home.)
>
> [ Output of lspci: ]
>
> > Base System Installation Checklist:
>
> <snip/>
>
> > Install boot loader: [O]
> > Reboot: [E]
> > [O] = OK, [E] = Error (please elaborate below), [ ] = didn't try it
> >
> > Comments/Problems:
> >
> > FIRST RUN:
> > At reboot, the GRUB screen listed only "Other operating systems",
> > showing the kernel previously installed in /dev/hda1. It didn't list the
> > newly installed kernel.
> > Booted the previously installed system in /dev/hda1 (Debian Sarge,
> > kernel 2.6.8-686). Mounted /dev/hda2 and examined /boot on that partition.
> > No kernel, initrd, or System.map files present, only a grub directory. The
> > progress bars said it was installing the kernel, but it's not there.
>
>
> Mmm, interresting.
> The rare case of upgrading a Debian system by a re-install.
> (the usual way to get from Debian Sarge to Debian Etch is
> editting /etc/apt/sources.list and then
> apt-get update
> apt-get dist-upgrade
> )
It didn't occur to me to try exactly that. I still could. What I
did was edit etc/apt/sources.list to include both Sarge and Etch, set Sarge
as the default branch, and installed a 2.6.15 kernel to do some testing. I
hit what appeared to be a bug when loading modules, and wanted to test a
pure Etch distro to see if it was still present. Not wanting to take time
to download and burn a CD, I downloaded the Etch RC2 installer floppies
instead, and found they didn't work. So I figured I'd try the daily build
floppies before reporting a bug. And that's where we are now.
I've tested the floppies, and found a solid bug. The major question
is whether it's because the daily build floppies don't include a boot.img,
and are incompatible with the RC2 boot floppy, and therefore I should stop
testing the floppies until one appears.
>
> Anyway, we still have the Sarge system to invest this situation.
I'm re-installing that now, to get it back to a pristine state after
monkeying with a bunch of config files.
I think what I'll do after that's running is take your suggestion to
do a dist-upgrade to Etch in /dev/hda2, and get back to what I was
originally trying to test, which was installing multiple Ethernet driver
modules on the 2.6.15 kernel. I still have the third set of partitions, to
test the floppies on when I have a known valid set of floppies for testing.
Why apt-get, by the way? I understood that was deprecated a couple
of years ago in favor of aptitude, because of the way RECOMMENDS and
SUGGESTS dependencies are handled.
> Please provide the output of
>
> fdisk -l /dev/hda
I'll get you that in the next message, after the install finishes.
>
>
> >
> >
> > During install:
> [ Selected .... ]
> >
> >
> > SECOND RUN:
> [ Selected .... ]
> >
> > Same result as before. Also checked /var and /tmp. Empty. Grepped the
> > whole partition for vmlinuz. Not there.
>
> Please describe the selections during partitioning.
3 partitions of 5 GB each, used for / in installations #1, #2, and #3
respectively. /dev/hda1, /dev/hda2, /dev/hda3
3 partitions of 3 GB each, used for /var in installations #1, #2, and #3.
/dev/hda5, /dev/hda6, /dev/hda7
1 partition of 1 GB, used for /usr/local in all installations /dev/hda8
1 partition of 512 MB, used for swap in all installations /dev/hda9
1 partition of 54 GB, used for /data in all installations /dev/hda10
/dev/hda3 and /dev/hda7 have never been formatted.
>
>
> Cheers
> Geert Stappers
Reply to: