FC4 install aborts abnormally after copying install image to drive

[Deron Meranda]

On 7/19/05, Jim Cornette <fc-cornette at insight.rr.com> wrote:
> Chris Ong wrote:
> Did you try to do a text install? How much memory do you have in your
> computer? Are you using acpi?

I got further with a text install, but the install still bombed.

First, the system has 1 GB memory, which I hope is plenty--confirmed
by cat'ing /proc/meminfo.  I had ACPI enabled previously, but I disabled
in the BIOS this time around.  (This is a desktop, not a laptop).

During this text install I tried a few different things.  First I manually
partitioned everything at the ctrl-alt-2 shell.  I deleted all prior partitions,
physical volumes, and volume groups, and then recreated them with
fdisk and lvm commands.

I only used one of my two drives this time.  /boot is sda1 and sda2
is an LVM PV, which contains the other LVs: /, /var, /home and a
4 GB swap.  sdb was left unused with no partitions defined.

I also instructed anaconda to set SELinux to warn mode rather than
enforcing this time.

The text install then proceeded to format, lay down the base image,
and then actually started installing packages.  However after about
50 or so packages the system once again bombed out!

The /root/install.log had left off saying it was installing ghostscript,
then libxslt (I was tailing it), before the install was aborted.  Unfortunately
the shell was locked up so I couldn't type any commands even though
I was at a shell prompt.  The ctrl-alt-[12345] still switched consoles though.

The other virtual consoles had little of use.  The anaconda screen
was trashed with unmounting and ready to reboot messages.  The
install debug had left off with something like these messages:

  moving (1) to step install packages.
  setting file_context_path to /etc/selinux/targed/contexts/file_contexts

and nothing more.  The audit console only had a few avc messages,
mostly saying that the "restoreconn" was denied read on a pipe because
it's scontext was restore_t and tcontext had anaconda_t.  But since I
had SELinux in warning mode, I don't think those mean much.  I found
no other errors.

What can I try next?   If this is a sneaky bug I'd really like to be able
to help find it so it can get fixed.  Can I get anaconda to run
in some sort of debugging mode?
-- 
Deron Meranda