On 25-03-2009 17:45:27 +0100, Fabian Groffen wrote:
> On 25-03-2009 17:16:17 +0100, Michael Haubenwallner wrote:
> > On Wed, 2009-03-25 at 14:44 +0100, Fabian Groffen wrote:
> > 
> > > If we would also change all ED into $D$EPREFIX inheriting prefix.eclass
> > > would do the trick entirely and we could merge back.
> > 
> > I missed the point here:
> > Is this because D (and ED) might be unset during ebuild source-time, so
> > setting ED in prefix.eclass' global scope will go wrong?
> > If yes, then +1.
> 
> You can't expect $D to be available in depend phase/source time, so ED
> will be wrong.  I think especially Paludis will be strict in this, I was
> told Portage just exports it, but that's luck.
> 
> See PMS, page 63.  ROOT is actually also a problem since it is defined
> only to be available in pkg_*.  That leaves us with EPREFIX.

Alternative is to get EPREFIX, ED and EROOT in EAPI=3, which is just as
preparation for Prefix.  Then with prefix.eclass, we can just require
EAPI=3.  That sucks too.


-- 
Fabian Groffen
Gentoo on a different level