>>>>> "Stephane" == Stephane Bortzmeyer <[EMAIL PROTECTED]> writes:
Stephane> Since it seems a new version of autoconf is close, could I
Stephane> ask that we integrate the macros in the official archive:
Stephane> http://peti.cys.de/autoconf-archive/
Stephane> They are well tested (I maintain some Java macros and I can
Stephane> testify that I receive bug reports, which is a good sign
Stephane> that the archive is used) and would simplify the work of
Stephane> developers.
Stephane> I regret that Java or Perl is not supported "out of the box"
Stephane> by autoconf.
I'm only expressing my own feeling, and in no way a summary of the
development team's opinion.
Firstly, I think it is too early to integrate too many specialized
macros in Autoconf. Yet the one we have are still buggy, out of
dated, fragile to strict environments which require the prototypes
(because many #includes are missing) etc. We first have to clean up
the whole Autoconf so that we can step backward, look what happened,
and write down the requirements on the wannabe-Autoconf macros.
For me this step is in the medium term (2.17 or so). Nonetheless, of
course, shortcomings of the current acspecific.m4 should be fixed.
Secondly, and more importantly, the issue of other languages. Yet
with the three languages we have, we have problems to solve. Tiny
steps are made, but there are not enough. Again, the problem appears
to be the programmer interface provided by Autoconf. This is, IMHO,
much more urgent, and should be solved either for 2.15, or next to
follow 2.16. Then, at this time, we will be able to integrate support
for other languages.
Since 2.15 is needed, I'd like to avoid adding this requirement to the
spec of 2.15.
Finally, IMHO, it is not the mission of Autoconf to be a constellation
of specialized macros (I don't mean you believed it was, I just want
to state publicly my opinion). The true nature of Autoconf is to
provide the environment. Major improvements were done so that
Autoconf is extendable.
I would like to ask people to help us finalize Autoconf, which means,
concentrate of the problems of today. It is too late to open new
horizons to Autoconf but those which concerns the base programmer
interface. We need it to be tested (yet there are problems on the
test suite which are not fixed, so don't rush onto make check :), we
need the documentation to be criticized, we need acspecific to be
updated etc. Wide spread macros in the style of acspecific will be
included, but only a small set of them (I'm thinking in particular to
Paul's LARGE_FS macros).
Akim
