Subject: Re: Sysinst coredumps during floppy install
To: David Laight <>
From: walt <>
List: current-users
Date: 07/27/2003 11:41:07
David Laight wrote:
> On Sun, Jul 27, 2003 at 08:40:08AM -0700, walt wrote:

>>To recap, after successfully creating the new fs on the target
>>hard disk, sysinst coredumps with a 'fs full' error when attempting
>>to select the source for the installation 'sets'.

> If you have a hard reproducible error, it is probably worth trying
> hard to track this one down.  Unfortunately, all the evidence so far
> seems to show that traversing a different route through sysinst causes
> the error to go away.  No one has found the path that actually
> makes it happen...

A few things I did which many people may not do:

I installed on wd1, which WAS a primary disk for a long time and
MAY still have other left-over disklabels somewhere on the disk.
The first disk wd0 does have a FreeBSD-current installation on it,
but the NetBSD installer didn't offer any complaints.  I never had
sysinst examine wd0 at any time.

I delclined the offer to install the NetBSD boot selector.

During the editing of the DOS partition table I actually converted
an existing ext2 partition into a NetBSD partition instead of deleting
and re-creating the partition.  No error messages occured.

I created a custom disklabel with only a root and swap partition,
(a and b).  The newfs was created without problems and then I went
to the menu to select the source of the install 'sets' and picked
NFS, which is where the segfault happened.

Now that I have the hard disk set up I can reproduce the segfault
easily just by retracing my steps and accepting the partition table
and disklabel 'as is', then trying to select the source for the 'sets'
again.  I get the segfault with either 'ftp' or 'nfs' which are the
only options I've tried so far.

> If it is an overwrite, knowing the text/data of the
> overwrite might be enough to track it down.

Any suggestions for debugging?

>>I made the install floppies on my working 1.6U desktop machine
>>which is current as of yesterday.
> Does that mean that my new boot code actually works on a true floppy
> install set?

Well, I don't know how it worked before your new code, but these floppies
certainly do boot now and I just made them yesterday.  I cvs'd again
just now but today's build dies while compiling the toolchain so I can't
make any new floppies just yet.  I'd be happy to try any tests you can