[ On , March 23, 2000 at 17:45:21 (+0100), Akim Demaille wrote: ]
> Subject: Re: Autoconf and gnu-m4 dependency
>
> Juergen> Now, what to do? Two things: Firstly, *explicit*
> Juergen> documentation of all the stuff in autoconf that depends on
> Juergen> undocumented stuff in GNU m4, and second, get the m4
> Juergen> maintainer(s) to document these undocumenteds.
>
> You are right.
and in th mean time ensuring that autoconf works properly with an
equally correct implementation of M4 ensures that autoconf only uses the
documented features of the 'language'.
In fact I think this should be an ongoing goal for autoconf --
gratuitous dependence on features unique to an implementation of M4, and
especially if those features are themselves either "gratuitous" or
uintended in some way, is generally bad for autoconf and will ultimately
make it harder to maintain, especially in the true free software sense.
It has always bothered me a great deal that autoconf required GNU M4 and
only GNU M4 (though that clearly wasn't the case in the past when I used
NetBSD M4 with it, and clearly isn't the case today in that it now works
with the OpenBSD M4). I can readily accept that there are inferior
implementations of M4 but I cannot accept that something unique called
"GNUM4", and that goes double if what makes it unique is undocumented
magic.
--
Greg A. Woods
+1 416 218-0098 VE3TCP <[EMAIL PROTECTED]> <robohack!woods>
Planix, Inc. <[EMAIL PROTECTED]>; Secrets of the Weird <[EMAIL PROTECTED]>
