Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 25 Mar 2004 04:23:43 +0100
From:      Michael Nottebrock <michaelnottebrock@gmx.net>
To:        Oliver Eikemeier <eikemeier@fillmore-labs.com>
Cc:        freebsd-ports@FreeBSD.org
Subject:   Re: cvs commit: ports/devel/libvanessa_adt Makefile pkg-plist	ports/devel/libvanessa_adt/files patch-ltmain.sh
Message-ID:  <200403250423.47441.michaelnottebrock@gmx.net>
In-Reply-To: <40624417.5040209@fillmore-labs.com>
References:  <Pine.LNX.4.44.0403231422350.6103-100000@pancho> <200403240109.09430.michaelnottebrock@gmx.net> <40624417.5040209@fillmore-labs.com>

next in thread | previous in thread | raw e-mail | index | archive | help

--Boundary-02=_DDlYA8k8DWFbScU
Content-Type: text/plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: inline

On Thursday 25 March 2004 03:29, Oliver Eikemeier wrote:

=2D---------- [Long monologue following - skip forward to the last paragrap=
h if=20
you're easily annoyed by long mails] --------------------------------------=
=2D-

> So far I followed the former discussions, but I can't remember an example
> where something *really* uses the .la files.

Yes, although I keep repeating it, nobody seems to ever remember: KDE _real=
ly_=20
uses them, as it _really_ uses lt_dlopen(). Also third party KDE applicatio=
ns=20
will use them (if they happen to load modules or plugins - a simple=20
preferences dialog can be such a module for example), since they use kdelib=
s=20
which in turn use lt_dlopen().

> Just `they might be used' is=20
> not really helpful.
> I guess documentation when and why these=20
> should be installed would be nice. Otherwise you have lots of ports where
> they would be needed, but aren't there, or ports where they are useless
> but are installed.

That is exactly the problem.

Imagine this: A programmer, sometime in the future, writes an application i=
n=20
which he uses lt_dlopen() to load a shared-library installed by your port a=
nd=20
somebody makes a FreeBSD port of said application and finds that the=20
application does not work because the .la file is missing. You cannot predi=
ct=20
this. You cannot tell if an .la file will be useless forever or if it will =
be=20
needed by something tomorrow.

This actually happened a short while ago btw. KDE 3.2.1 (kdelibs-3.2.1, mor=
e=20
precisely) can use libidn to translate International Domain Names. It tries=
=20
to lt_dlopen() libidn - and it doesn't work on FreeBSD, because the libidn=
=20
porter made sure the port doesn't install libtool archives, in 'good=20
tradition'.

We at kde@ decided that in this case, we would rather NOT stir up yet anoth=
er=20
debate about libtool archives, since we're all really quite fed up being to=
ld=20
that lt_dlopen() sucks and libltdl should be fixed and that nobody should=20
really need libtool archives. So Andy Fawcett made a patch to kdelibs to us=
e=20
the native FreeBSD dlopen() - unportable of course, but in this case this i=
s=20
bearable since in KDE HEAD libidn will actually be linked into KDE in the=20
traditional way and not be (lt_)dlopen()'d anymore.

As you might notice, I'm really tired of the whole subject. If we could jus=
t=20
drop the 'good tradition' until somebody _really_ changes libltdl to not=20
require libtool archives on the FreeBSD platform (note that I'm talking abo=
ut=20
an upstream fix here, in the GNU source), everybody would have at max one=20
megabyte or so less diskspace less in the future (taken by libtool archives=
)=20
and that's it. Then, with time, as projects adopt the new version of libltd=
l,=20
the libtool archives would simply fade out of existence again.

=2D------------------------------[End of monologue]------------------------=
=2D----


However, I would already be totally happy if we could just add a paragraph =
to=20
the porter's handbook that says

"If a port maintainer requests that your port SHOULD install libtool archiv=
es=20
because another port needs them, you shall comply with his request."

To me, this seems like a workable and fair compromise and it would be a nic=
e=20
thing to point to if anybody feels like beating the dead horse again. We=20
wouldn't even need to change portlint's warnings that way. Sound good?

=2D-=20
   ,_,   | Michael Nottebrock               | lofi@freebsd.org
 (/^ ^\) | FreeBSD - The Power to Serve     | http://www.freebsd.org
   \u/   | K Desktop Environment on FreeBSD | http://freebsd.kde.org

--Boundary-02=_DDlYA8k8DWFbScU
Content-Type: application/pgp-signature
Content-Description: signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (FreeBSD)

iD8DBQBAYlDDXhc68WspdLARAiC9AJkBHpMVg0ESmbQuc/evb8ExI3NYTQCgnpGM
+bTJYeA5NzcP0mN0i7hUg50=
=7wJ6
-----END PGP SIGNATURE-----

--Boundary-02=_DDlYA8k8DWFbScU--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200403250423.47441.michaelnottebrock>