From: softbrek AT POOL DOT Informatik DOT RWTH-Aachen DOT DE
Subject: autoconf for DJGPP (maybe) (was Re: Fileutils for DJGPP (?))
To: eliz AT is DOT elta DOT co DOT il (Eli Zaretskii)
Date: Mon, 17 Oct 1994 16:32:25 +0100 (MET)
Cc: djgpp AT sun DOT soe DOT clarkson DOT edu (djgpp)

> 
> > No, it's not EZ :-P  At least, without a proper shell, building is
> > nowhere near automatic.
> 
> I didn't mean it could ever be.  But as long as we have no proper means to
> make configure scripts runnable, IMHO we shouldn't care.  Until then, we
> *must* craft Makefiles by hand, so we might as well patch it to get away
> with MS-DOS limitations.  Maybe some day somebody will make a decent
> shell port to DOS.  Sigh...
> 

I hope not to be promising too much, but I *might* be able to provide
at least a starting point for running configure scripts on DOS.

Now, well, I cheated a bit: for some time, I used to edit configure
scripts over and over, until they run all the way to the end, using
ms_sh, sed and whatever I have on my private collector's edition /bin
directory. Then, I realized, the main changes were always the same
ones, e.g. having to do with some misunderstanding between ms_sh and
sed about passing HERE-scripts. For non sh-gurus: these are typical
shell-script commands like

	sed foo >bar <<EOF 
	{several sed-lines}
	EOF

Since I once bought a computer to automatize repetitive work (:-), I
started looking for a way to have those tons of utils do it.  I
decided that the only suitable way to do this, at least for (here
comes the cheat) autoconf-generated configure-scripts, as included in
most of today's FSF-packages, was to port autoconf to DOS, and have it
generate *new* configure-scripts from the normally supplied
configure.in, and embed my standard changes in autoconf's
macro-libraries to test for certain features.  Of course, I had to
port FSF's version of m4 to DJGPP to do this, but this was not really
that hard, and even a useful test-case (I just circle-built m4,
autoconf, and m4's configure-script until everything did what my
constrained knowledge on shell-scripts and m4 and ... said it should
do).

As I'm not sure if I did it even nearly right, but there might be some
usability hidden in my work, I'd like to here from anyone who thinks
he knows good enough to alpha-test this. If there is enough interest,
I might try to convince DJ to put it in the distribution.

Whoever is interested should email to me directly, and I'll mail back
what I have ASAP (perhaps including a list of the utilities I used to
do it). Remember: no beginner-serviceable parts inside (you'll hate
m4 macro language if you look at it without sens-o-cool sunglasses
to protect your eyes :-)

See you

Hans-Bernhard

P.S. I've heard somewhere about a kind of caching addition to
autoconf, remembering results of autoconf's tests, aimed to reduce
testing time next time the same test would be run. Could anyone point
me to some more information on this ?