Re: configure dependencies, and version.texi dependencies (PR/358).

[Alexandre Duret-Lutz]

>>> "Akim" == Akim Demaille <address@hidden> writes:

[...]

 Akim> As a result, they ship a modified Autoconf which no longer requires
 Akim> GNU M4.

 >> Is there any way Automake can filter user files from Autoconf's files?

 Akim> Hm.

 Akim> So maybe we want to introduce AC_INCLUDE and trace the latter.

Actually I was rather thinking of something that would work with
existing configure.acs: leave m4_include to the user, and use
something else in Autoconf, e.g.

m4_define([m4_include_private], [m4_builtin([include], [$1])])

Probably AC_INCLUDE is cleaner.  Dunno.

 Akim> OTOH, that's *their* problem.  Not ours.

Not sure :) It'll be our problem if we end up wasting our time
handling bug reports about this.  It's also our problem if we
give people reasons to stay with older Autotools.

 >> Is it enough if Automakes ignore all `m4_include' traces up to the
 >> first occurrence which is either in aclocal.m4 or configure.(in|ac)?
 >> (This assumes Autoconf won't try to m4_include something conditionally.)

 Akim> Which is probably an assumption we can't agree on.

I feared that.

 >> Or any way Autoconf could hide its includes from `--trace m4_include'?

 Akim> Or maybe you just want to filter out everything that has an absolute
 Akim> file name.

Does this mean Autoconf should use absolute paths?

% autoconf -f -i -tm4_include:'$1' /dev/null
m4sugar/version.m4
autoconf/autoupdate.m4
autoconf/general.m4
autoconf/status.m4
autoconf/autoheader.m4
autoconf/autotest.m4
autoconf/programs.m4
autoconf/lang.m4
autoconf/c.m4
autoconf/fortran.m4
autoconf/functions.m4
autoconf/headers.m4
autoconf/types.m4
autoconf/libs.m4
autoconf/specific.m4
autoconf/oldnames.m4

-- 
Alexandre Duret-Lutz