On 10.05.99, 18:40:36, Manuel Bouyer wrote:

I re-downloaded the image and doublel checked the floppies, just to avoid
any other sources of problems.

After the above message it says "floating point exception". Note that this
happened on both Cyrix 486 with 1.3.2 installed and on Celeron with 1.4_BETA
installed.

> 
> Did you check if sysinst coredumped ? the / filesystem can be full because
> sysinst crashed and dumped a core in the ram disk.

I'm then left with a shell (which does not echo my typing btw, but still
works), and, yes, there is a sysinst.core. But the / filesystem
is not completely full, of the 1375(?) blocks about 130 remain free.

I also tried it again with logging enabled in sysinst, and the last entry in
the log is "renaming /mnt/boot.bak to /mnt/boot". (which is also the last
message it had written to screen before aborting.)

> 
> > Rebooting from hard disk after this starts up a sysinst, that is not
> > usable at all. Overwriting /sbin/init from the base.tgz and moving
> > etc.old back to etc makes the original 1.4-BETA system usable again. I
> > don't know what would happen if the original system was 1.3.x.
> 
> I guess you didn't need to override init. the installation was just missing
> files from /etc.

This even happens after completely restoring the old etc. The /sbin/init left
on my disk is about 1MB in size, the regular init is only some 170k.

Taking a close look, the file called /sbin/init on my / partition after
the failed installation _is_ sysinst. Now how could that happen?

What's going wrong here?

> 
> --
> Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
> --