Le dimanche 24 juin 2012 à 16:48 +0800, Ben de Groot a écrit :
> On 24 June 2012 06:50, Gilles Dartiguelongue <eva@g.o> wrote:
> > Le samedi 23 juin 2012 à 18:30 +0100, Ciaran McCreesh a écrit :
> >>
> >> It treats -r300 as being newer than -r200, and so will treat "the gtk3
> >> version" or "the jruby version" as being newer versions of "the gtk2
> >> version" or "the ruby 1.8 version", just as it tries to bring in a
> >> newer GCC and so on.
> >
> > I'm stopping my reading of this thread a minute to answer here.
> >
> > This is actually true when you think of it, gtk3 bindings are newer than
> > gtk2.
> 
> Now you're playing with semantics. In the case of -r200/-r300 we
> are talking about the *exact same* $PV, but for some reason
> the revision numbers are confusingly abused for something
> that we normally use useflags for (toggling support for specific
> toolkits for example).
> 
> Please stop abusing revision numbers for something they are
> not meant to convey. And please stop pushing developers to
> drop perfectly legal usage of the gtk3 useflag.
> 

This is the same codebase, but they really are slotted libs (that
happens to have the same $PV):
 * different include path
 * different pkgconfig files
 * different sonames
 * ...

If the $PV wasn't the same, there would be no question about have a USE
flag or not, the answer would be obvious to anyone. So please stop
pretending this is a good case for USE flag.

Now if this is the only case (lib with support for two gtk+ versions but
slottable/slotted) that is causing a problem to anyone here, I propose
we go with the simplest fix, have a new package name. That will remember
me of debian packaging :)

-- 
Gilles Dartiguelongue <eva@g.o>
Gentoo