The problem is actually that we don't have nice `plug-n-play' patches
  for rpc2 decoding. Our current setup depends on the protocol opcodes
  (i.e. headers) generated by rp2gen, and the resulting tcpdump can only
  pretty-print Coda traffic.

If the coda on-the-wire protocol is stable (it seems to be at the
moment, but I'm not sure), one could submit the files already
generated, and teach tcpdump about the port numbers in use.  That
would be an improvement, even though it might need changing later.

  Ok, I've applied your patch, actually testing ${prefix} first. I don't
  think we need CFLAGS during the test at all. We are only attempting to
  link;

Thanks.  Actually, the test should probably be to include the header,
use some function, and then link.  That way it will fail unless the
includes are there too.  autoconf has macros for testing for functions
in libraries, and this is probably a good place.

It would be cool if lwp (etc.) install autoconf macros to find
themselves for aclocal to use.

I realize that going to automake is hard; I meant only to remove
configure, since it can be generated from configure.in.  I'd be very
surprised if this caused trouble for people using anoncvs.  (I check
out things with CVSREAD=t to make emacs vc mode work nicely, and
running autoconf got me a flame about configure being read-only, so
the current scheme caused me trouble.  Plus, 'cvs diff' would have
given me a configure diff, not just configure.in.)

        Greg Troxel <gdt_at_ir.bbn.com>