[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: FYI: update convenience.at
From: |
Ralf Wildenhues |
Subject: |
Re: FYI: update convenience.at |
Date: |
Sun, 18 Dec 2005 17:29:51 +0100 |
User-agent: |
Mutt/1.5.9i |
Hi Olly,
Sorry for the delay.
* Olly Betts wrote on Sat, Dec 10, 2005 at 05:11:14AM CET:
> On 2005-12-03, Ralf Wildenhues <address@hidden> wrote:
> > There'll be quite a bit of systems/compilers where this will fail.
> > I wonder whether libtool should just go ahead and generate a fake
> > extra object in the case where only convenience archives are given,
> > and 'whole_archive_flag_spec' is nonempty. Thoughts?
>
> This (5 year old!) thread seems relevant:
>
> http://lists.gnu.org/archive/html/automake/2000-11/msg00127.html
Yep!
> The main problem was (and I don't think this has changed) that when
> using libtool from automake you need a source file to infer the tag
> from. For C++ this often matters, for example.
Right.
> AIUI, this is really a particular case of the more general issue that
> automake+libtool doesn't consider the languages that objects in
> convenience libraries were compiled from when deciding how to link.
>
> Fixing this seems rather hard - you'd need to store source information
> in .la files I think. And it's easy to work around (e.g. just add in a
> dummy C++ source when creating a library from C++ objects in convenience
> libraries).
Right again. We could think of putting the source language(s) into .lo
files. But really I think the workaround is much more appropriate.
I will fix convenience.at to remove those tests, and prepare a doc
update for HEAD to mention this (eventually).
Thank you for looking this up!
Cheers,
Ralf