[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: AC_LIBOBJ with file in subdirectory does not work
From: |
Ralf Wildenhues |
Subject: |
Re: AC_LIBOBJ with file in subdirectory does not work |
Date: |
Wed, 26 May 2010 07:25:07 +0200 |
User-agent: |
Mutt/1.5.20 (2009-10-28) |
* Bruno Haible wrote on Wed, May 26, 2010 at 12:11:45AM CEST:
> > I see a few possible alternative ways out:
> > - Automake requires subdir-objects if you put slashes into $LIBOBJS,
>
> There are cases where the developer does not want 'subdir-objects'.
> I'm having the problem with AC_LIBOBJ in the
> gettext/gettext-tools/libgettextpo
> directory, and when I enable 'subdir-objects' in this directory, all hell
> breaks loose. Impossible.
Understood. While I think that libgettextpo isn't such an ideal
example, and would like to see if it's possible to rewrite in a more
automake-ish style, possibly using a symlink farm for the shared source
files, I agree that similar situations may be unavoidable in general
(actually, we have a project that requires not to use subdir-objects
either).
> > - Autoconf optionally removes dirname parts from $LIBOBJS (i.e., some
> > kind of "subdir-objects" toggle for Autoconf),
>
> Yes, but that requires Autoconf to receive information from Automake about
> its options. Which is hard because there are several Makefile.ams in the
> scope of a single configure file.
Yep. Information never flows that way. For any requirements from
Automake on Autoconf, automake checks whether side conditions are met
and fails otherwise; there are several examples for this (like
requirement of some macro such as AC_PROG_RANLIB), so this is feasible.
But it's ugly, and still won't fix your use case above nicely (unless
you can choose to turn off subdir-objects everywhere).
> > - Automake always enables subdir-objects machinery for the $LIBOBJS
> > files, but not necessarily for the other files.
>
> This would be acceptable, yes.
>
> The other, completely different alternative is to deprecate AC_LIBOBJ.
Yes, I guess. I'll investigate the feasibility of always assuming
subdir-objects for LIBOBJS only.
Cheers,
Ralf