Martin S. Weber wrote:
> Hmm, but NetBSD/i386 expects 'c' to be 'NetBSD part of the disk' and
> 'd' to be 'the whole disk from beginning to ending'. Don't do that.
> Try with 'e' and 'f', restoring 'c' and 'd', then writing on ld1e
> shouldn't effect ld1a, and ld1d should effect ld1a as it's supposed
> to do.

   Okay.  If this is the problem, then I accept your solutions as
a workaround.  I'll try to be more careful about this in the future.
But, I see two issues:
   (1) It's been BSD tradition that 'c' is the whole disk, and
oft 'b' is swap, so people tend not to use them.  Was it a good
idea to make 'd' special?!
   (2) Even in older BSD's, 'c' was *usually* the whole disk, but
didn't have to be.  You could make it a real partition, it just
made many other things difficult.  As I said in my original message
(or at least I know I meant to say), the disklabel output of ld1
does show 'd' to *not* be the whole disk.  It shows it to be offset
by the size of 'a'.  So, there's still a bug.  disklabel should never
show something that the driver doesn't honor.  Is it impossible
to override what 'c' or 'd' are on other device types, and/or on
other ports?

   Thanks...

                              - Chris