Subject: restore doesn't report filesystems on null hostname
To: BSD Current Users <current-users@netbsd.org>
From: Greywolf <greywolf@starwolf.com>
List: current-users
Date: 01/23/2000 14:15:37
[I have send-prd this already -- this is a heads up.]

Does anyone else here do dumps upon just having brought the machine up?

On occasion, I do.

Typically, when the machine jumps into single-user mode on bootup,
you are in the following state:
	/ mounted read-only
	hostname not set
	network not configured - not even lo0

I have discovered with much chagrin that this leaves an empty hostname
portion in the dump header, even though the filesystem and device and
disk volume label (not the disklabel) are transferred without incident.

This is not a problem with dump, though.  It's a problem with restore
that it chooses to punt the header printing on a 'restore v.+' if the
hostname is empty.

The grief this causes me is not on a grand scale.  To be frank, it's
an annoyance.  But if I happen NOT to have a table of filesystems at
the beginning of the tape, I lose, because that nice little line that
reads

	Level 0 dump of /export on somehost:/dev/ccd0c

doesn't get printed.  And I get to divine which dump goes where.

Even if I *do* have a table of contents, it's kind of painful due to
the st driver not working properly (it counts the EOF mark as a separate
file, but that's another can of worms).

I have send-prd (sent-pr?) a fix which will not cause the tape reader
to immediately return if the hostname on the tape is the null string.

If anyone wants the fix to this more immediately than the send-pr will
get processed (it's marked as non-serious high priority :-), let me
know and I'll send you a diff.

				--*greywolf;
--
NetBSD: My Computer Runs!