Re: Can packages rely on AC_PROVIDE_IFELSE?

[Eric Blake]

On 11/26/2012 08:02 AM, Nick Bowler wrote:
> Hi folks,
> 
> Autoconf defines a macro called AC_PROVIDE_IFELSE, which can allow
> one to hook into the AC_REQUIRE machinery in order to do potentially
> interesting things.  For a specific example, I recently wanted to emit a
> (hopefully) helpful warning message if LT_INIT had not been previously
> expanded, and it seems that I can do that very easily with something
> like:
> 
>   AC_PROVIDE_IFELSE([LT_INIT], , [m4_warn([syntax], [blah blah])])

Good question - looks like you found a real hole in the documentation.
It looks like AC_PROVIDE vs. AC_PROVIDE_IFELSE should be documented
together; and also that m4_provide and m4_provide_if should be
documented together.  What's more, the manual mentions that macros can
call AC_PROVIDE but does not document what it does.

> 
> However, AC_PROVIDE_IFELSE is not described in the Autoconf manual,
> which makes me wonder if the macro should be used by packages.  The fact
> that it's defined at all (as an alias for m4_provide_if) seems to
> suggest that it was intended to be used...

Yes, the macros are stable and intended to be usable; we're just missing
the documentation patch.

> 
> Also, on a related note, AC_PROVIDE is referred to several times in the
> Autoconf manual (in a manner that suggests that packages can use it) but
> the manual never actually explains what AC_PROVIDE does or how to invoke
> it properly.

Yep, I just came to the same conclusion after reading just your first
paragraph and searching for myself.

> 
> So, are these macros intended to be used by packages?  If so, can they
> be documented in the manual?

Absolutely.  Want to help write the patch?

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

signature.asc
Description: OpenPGP digital signature