On 2006-03-26 02:22:22 -0500, James Ashenhurst wrote:
> Well, you can always install the non-devel emacs port and just not
> activate it (that's what port did for me, though I don't recall
> telling it to do that...).
This is a hack based on a bug in DP, isn't it? Dependencies should
require the ports to be activated, otherwise some needed library
(for instance) provided by a port which isn't activated won't be
On 2006-03-27 17:48:51 +0100, Arnaud Delobelle wrote:
> However the reason why I think that there shouldn't be a variant at
> all is that auctex only depends on the emacs infrastructure, which is
> equally provided by emacs (21) or emacs-devel (22). The proof is in
> the auctex portfile, whose variant +devel only changes the
> dependency. I think ports like auctex should depend on a "virtual
> port" provided by both emacs and emacs-devel (and potentially other
> versions of emacs).
I completely agree.
> I have just thought about another reason why there shouldn't be a
> variant in the case of auctex, and it seems more compelling to me:
> severeal variants of the same port can be installed at the same time,
But in practice, only one can be activated, since there are some
common files (and DP would complain about that when activating a
> So auctex and auctex +devel could both be installed. Now
> imagine I decide to get rid of emacs-devel; in order to do this I
> need to uninstall auctex +devel, but as they are both physically the
> same files at the same location, then *both* will be removed, even
> though port would still believe that auctex is installed.
Since both can't be activated at the same time, there would be no
problem. However, as I've said, a dependency shouldn't be satisfied
when the port is installed, but not activated; the current behavior
does not make sense.
Also, I assume that having to change the variant just because the user
switches from emacs to emacs-devel (or the opposite) is annoying.