On Fri, Dec 20, 2002 at 09:57:38AM +1100, Daniel Carosone wrote:
> Bill's proposal (or a number of other options) allow one type of
> label to update the other each boot (even if the latter is now only
> in-core). Sounds good to me.
> 
> > For MBR-using ports, mbrlabel(8) is useful to find the partitions.
> 
> Hot damn!  I wish I'd known about that a few weeks ago. Neat.

This reminds me.  It would be a very nice feature (whatever the
disklabelling scheme) to be able to update *only* the in-core label.

NetBSD synthesises disklabel partitions from native ones (mbr, etc)
when no real disklabel is found, such as for a new disk, or for
disks that are wholly used by other OS's.

In the latter case, there may be nowhere to put a netbsd label, in
fact you might not want to write to the disk at all.  You might
still need to change the label (such as in the case where the kernel
doesn't create synthetic label entries for extended subpartitions).

Now, it looks like mbrlabel(8) does exactly what I was looking for
last week, at least in the case of mbr disks - it can populate
in-core labels without affecting (or requiring) an on-disk netbsd
label. But I've wanted this capability for other cases in the past.

It seems disklabel(8) can't do the same. It has -r to change the
on-disk label directly, for installation - but no inverse option
to change only the in-core label.  Have I missed something in
disklabel(8)?

--
Dan.