On Sun, May 25, 2003 at 10:25:19AM -0700, Matt Thomas wrote:
  | At 04:05 AM 5/25/2003, Luke Mewburn wrote:
  | >On Sun, May 25, 2003 at 11:54:35AM +0200, Martin Husemann wrote:
  | >  | On Sat, May 24, 2003 at 10:07:07PM -0400, Christos Zoulas wrote:
  | >  | > Please let me know what you think. It is pretty easy to convert the 
  | >  new
  | >  | > archs to use it, but it might break the script that generates the man
  | >  | > pages.
  | >  |
  | >  | I like the idea.
  | >  |
  | >  | Is mtree(1) available when this script is run? Maybe we could make it 
  | >output
  | >  | an mtree template and pipe that to mtree instead of running 
  | >individual rm/
  | >  | mknod etc. itself?
  | >
  | >mtree is in /usr/bin, and we currently use /dev/MAKEDEV for the "mfs /dev "
  | >hack in /sbin/init.  Therefore, this would cause problems on systems
  | >with a separate /usr.
  | 
  | Use pax -M

Good call; that could work.

Obviously, MAKEDEV would need to output an mtree specfile, but that
could be done with a switch and change all of the rm/mknod (or
${MKNOD} if my last big proposal was applied) into a function call
that switched between outputting an mtree specfile and running ${MKNOD}.

I'd be happy to work with Christos (or whoever) to implement this.

Luke.