Re: AC_REQUIRE problems

[Stepan Kasal]

Hi,

On Thu, Feb 24, 2005 at 03:25:38PM -0800, Paul Eggert wrote:
> Dan Manthey <address@hidden> writes:
> > Of course, if we decide that functions "are portable", we can skip the
> > step of designing the hairy workaround.
> 
> I think we can decide that for Autoconf version 3, but not for
> Autoconf 2.60.

I agree with this.  I wouldn't waste time with the hairy workaround you
mean.

Autoconf 2.60 will have m4_defun'd AS_IF, which is a good step forward.
The advantage of using AS_IF should be probably explained in the manual
and in the NEWS file, but I excercise the excuse of not being a native
speaker here.  ;-)

As far as my proposal to modify AC_REQUIRE to place the code near the
beginning of the script (similarily to AS_REQUIRE), I think it's not
worth it.  Introducing this change would mean we would have to add
many AC_REQUIREs to various macros which currently get out without it.
The interaction of Automake and Autoconf might be a good source of
this type of bugs.

Autoconf 3 with functions will have a more elegant solution to the
problem, without raising such problems.

I think that "Autoconf with functions" will use the same configure.ac,
which is very good for users.  I also guess that great deal of the
AC_DEFUNd macros will be the same, and that macros brought by aclocal
will also work without modification.

Thusly we could manage to place both alternatives to one tree,
controlable by a command-line option to autoconf, or perhaps by a
directive in configure.ac.

The only change, which would come with Autoconf 3, would be that we
make the "functions" option the default.

Have a nice day,
        Stepan