On Wed, 10 Dec 1997, Jonathan Stone wrote:

> The problem is that all the old 1.2G and earlier kernels turn off the
> write-enable bit for the disklabel area on the `last close' of the
> disk partition.  On these kernels, disklabel -W opens the raw
> partition, tells the kernel to make the disk writable, and then closes
> the raw-disk partition.  If that' the  last close (refcount goes to
> zero) the kernel turns the label-write bit off again. So  dd'ing the 
> label of the diskimage fails.
> 
> Hence my and Michael's suggested workarounds: starting a dd to read
> from the raw-disk partition and suspending it (I forget who first
> suggested this) or Michael's suggestion of mounting the disk
> (works if there's a valid filesystem on the `c' or `a' partitions.)
> 
> This probably deserves to go in the FAQ, (hint, hint :-)...

This was the problem:  as Jonathan and Michael suggested, my "helper"
machine's kernel was vintage enough to screw up closing the partition. 
I got by with Jonathan's suggestion of suspending a dd.  I did not try
Michael's suggestion.

Thanks all for the help.  Sysinst seems to be pretty slick (IMHO--my first
netbsd/pmax install was on a headless 5000/120 using an ultrix
helper--sysinst almost seems like cheating....)
 

-Matthew Schnierle
-pyld@grove.iup.edu
-http://www.iup.edu/~pyld/
-Member, Programmers Local 1036--AFL-CPIO.
-Join the fight.  http://www.cauce.org