Re: conditional compilation (or: makefiles again)

> It works to also add a "include ../config.make" into the Makefile before
> testing LIBNCURSESW.  Is that too clumsy?

I think it is ok.

> Would you fundamentally object against using automake and latest autoconf?
> (I know that Roland sticks with autoconf2.13, otoh, Debian has excellent
> auto-detection of the version of autoconf required plus a wrapper script
> that seems to do the right thing always).  I might be able to spent some
> time in working out the issues with that.

I don't have any fundamental objection at all, I have thought about it
before.  

What matters is the good features that the current setup has, i.e. dealing
with mig and all easily and having the automatic deps work with
mig-generated headers (-MG), and doing all this in a way that keeps the
subdir makefiles (or .am, whatever the source is) small and without
duplication.  There are other nits I am overlooking, I am sure.  But
Makeconf is really fairly short, so it can't be too much work to replace it
and the existing subdir makefiles with an automake thing.

Incidentally, I use autoconf 2.13 because libc needs it and noone has
wanted to contend with all the incompatible changes later autoconf made
that broke libc.  But I also use a system where the default autoconf and
automake installed are 2.13 and 1.4-p5, so it would be nice to have the
hurd sources be compatible with that if there is not a compelling need to
use newer features.