[oe] [PATCH] gst-plugins: fixed pattern for meta package dependencies.

[Phil Blundell]

On Sun, 2009-05-24 at 21:37 +0300, Ihar Hrachyshka wrote:
> On Thu, May 21, 2009 at 6:48 PM, Phil Blundell <pb at reciva.com> wrote:
> > On Thu, 2009-05-21 at 16:47 +0100, Phil Blundell wrote:
> >> This kind of matching on the package name is a bit unwholesome in
> >> general.  It would be better to handle these excluded packages by
> >> maintaining them in the blacklist at the point where they're actually
> >> created, rather than trying to recreate that information again after the
> >> fact.
> >
> > Or, actually, in this specific case, just maintain a positive list of
> > the plugin packages that you _do_ want the metapackage to depend on, and
> > make it depend on those packages alone rather than starting from
> > ${PACKAGES} and filtering things out.
> 
> I think -meta package should depend on *all* the plugins. That's what
> -meta is made for. I think it's not very efficient to test every new
> gst-plugins-* release if there are any new or obsolete plugins there.

Yes, indeed, clearly -meta should depend on all the plugins: as you say,
that's exactly what it's for.  The point I was making was that, rather
than starting with ${PACKAGES} (i.e. all the subpackages) and trying to
find suitable heuristics to take out the things that aren't plugins, you
would be better off just computing a list of packages that _are_ plugins
and using that directly.

p.