"Jared D. McNeill" <jmcneill%invisible.ca@localhost> writes:
> On Mon, 16 Mar 2009, Greg Troxel wrote:
>>
>> I ran pkg_rolling-replace on a machine which shouldn't have any gnome
>> goop on it (web server for trac which has graphviz). It failed with:
>>
>> ===> Installing for libgamin-0.1.10
>> ERROR: libgamin-0.1.10 conflicts with installed package(s):
>> ERROR: fam-2.7.0nb9
>> ERROR: They install the same files into the same place.
>> ERROR: Please remove conflicts first with pkg_delete(1).
>> *** Error code 1
>
> I'm not sure why you think gamin/fam == GNOME. libgamin is a libfam
Fair enough, but it has in my experience come along for the ride. I use
gnome on desktop machines, and was trying to keep my trac/svn server
minimal.
I think my real beef is with upstream glib as it seems like something
which shouldn't have this sort of dependency.
> reimplementation used for file change notifications (in glib among
> other things). I did recently change the default, not sure how pkgsrc
> manages that. One thing I suppose you could do is maybe alter the
Well, basically it doesn't, but this isn't your fault - it's just the
nature of swapping a dependency that conflicts. The higher-level
package, glib in this case, needs the new one, but the new one and the
old one conflict, and you can't delete the old one because glib2 depends
on it.
> logic in mk/fam.buildlink3.mk to accept either libgamin or fam, as the
> former is supposed to be source and binary compatible with the latter,
> at least for a subset of functionality.
That would probably be good, as it avoids forcing most people building
From source down this path, letting them choose when to deal.
I am not good enough at pkgsrc makefiles to do this. I would think it
goes like
Change
_FAM_TYPE?= ${_FAM_DEFAULT}
to
if one of fam and gamin is both accepted and installed, set _FAM_TYPE
to that one, else set to _FAM_DEFAULT
That leaves systems with fam alone, but I'm not sure that's good long
term.
I don't mean to complain about what you're doing - having the latest of
all these things is definitely the right thing and appreciated - it's
just pointing out some inadequacies of pkgsrc to cope with the
transitions.
Attachment:
pgpqYhug2FcGZ.pgp
Description: PGP signature