On Sat, Feb 12, 2011 at 08:32:50AM +0100, Marc Espie wrote:
> On Sat, Feb 12, 2011 at 08:25:54AM +0100, Antoine Jacoutot wrote:
> > On Sat, 12 Feb 2011, Marc Espie wrote:
> >
> > > > > on reflection I'm not so sure about @pkgpath for this, but I think
> > > > > conflict is right. in any event, pkg_add updates would needtesting.
> > > >
> > > > Hmm ok, I though adding an entry in quirks would render the @conflict
> > > > useless in this case.
> > >
> > > No, quirks will only extend the packages considered for updates, the usual
> > > rules still apply afterwards.
> >
> > Which is why I don't understand @conflict is needed. If the new pkgname
> > is considered for update and that the pkgpath matches, why do we need a
> > conflict?
>
> Because you don't get the actual algorithm used.
> Normal update
> 1/ derive the stem from the pkgname: foo-1 -> foo-*
> 2/ find all packages that match that stem
> 3/ keep only packages that conflict
> 4/ trim down what doesn't have the right pkgpath
> 5/ only allow updates to go forward (doesn't apply if stem doesn't match)
>
> quirks only changes 1/
Err, 2/
should really try to complete night of sleep.
