On Wed, May 22, 2002 at 06:38:41PM +0900, Junichi Uekawa wrote:
> > This is a partially false positive, obviously. Why is it that I always
> > end up adopting packages whose names/versions/whatever break policy and
> > there's never a convenient way to fix it, hmm? Oi. I'm willing to fix
> > this, with some caveats:
>
> Actually, almost all packages listed shouldn't cause any problem at all until
> the soname is upgraded.
In my case, the soname will not change until SDL 1.3 (which is an unstable
version) and be final for SDL 2.0. The current mechanism is effective,
but only because upstram is paranoid about ABI compatibility in stable
versions.
> The largest disadvantage of current situation is that automatic checking
> for upgraded sonames is not easily possible.
>
> It is not instantaneously problematic, but seeing that many new
> maintainers are following bad examples, and many people don't understand
> why these are bad saying "there are so-and-so examples in the archive",
> this would be nice to be fixed, or documented clearly, or if ftpmaster
> could reject such packages with shared libraries which contain packages
> with a different SONAME than the package name.
I think the lintian check is most suitable. For packages like GTK+ which
contain additional libraries tied to the main one, an override is probably
appropriate. Packages such as SDL which use the upstream version in place
of the soname are definitely a practice which should be discouraged when
possible. But that alone isn't enough for me to justify the effort it
would take to recompile everything.
--
Joseph Carter <knghtbrd@bluecherry.net> What're you looking at?
<NeonKttn> I had a friend stick me in her closet during highschool beacuse
I wouldn't believe that her boyfriend knew about foreplay...
<NeonKttn> I shoulda brought popcorn. :)
Attachment:
pgpoZ54hBiubP.pgp
Description: PGP signature