On Wed, 12 Jul 2006 21:54:28 +0200, Nicolas Mailhot wrote:
> Le mercredi 12 juillet 2006 à 21:44 +0200, Michael Schwendt a écrit :
> > On Wed, 12 Jul 2006 13:30:42 +0200 (CEST), Nicolas Mailhot wrote:
> >
> > >
> > > Le Mer 12 juillet 2006 12:45, Michael Schwendt a écrit :
> > >
> > > > %{?dist} is a variable, giving the false impression that the source
> > > > package may be valid for arbitrary dist releases and that filling it in
> > > > would be enough and that no other package updates are required. Even
> > > > worse, the dist tag enters the file names of the built packages.
> > >
> > > If you know your package will break on another release because foo package
> > > was updated and changed its behaviour, you should require or BR foo with
> > > the right version
> >
> > Who said there exists such a dependency or BR?
>
> If you know it will break with some other FC/FE version there is such a
> dependency or BR
No. Maybe, maybe sometimes it is possible to put such a requirement into
explicit+versioned BR. Ugly. Explicit dependencies on specific versions
(or max.versions, which are often guess-work) are ugly.
> If you don't know it will break, restricting the package in any way is
> 100% useless silliness
No, it's packager's freedom to make clear that the src.rpm was created for
a specific distribution. And if rebuilt for a different dist, without any
changes in the src.rpm contents (e.g. dist-specific patches or options),
the resulting binaries should not automatically look like they are for the
different dist.
