Shawn Walker wrote:

> On 09/17/12 14:16, Brock Pytlik wrote:
> >Everything you have seems reasonable to me. I don't like the idea of
> >using -P with a dummy variant to pkgmerge as a way to republish packages
> >with an existing publisher. I think we just need to bite the bullet and
> >put that functionality into pkgrecv, not try to shove it into pkgmerge
> >where it doesn't belong. It'd be really nice though if the two commands
> >could share their implementation though.
> 
> I think I disagree.  In fact, merging across publishers would have been
> useful in a few cases I know of internally in the past.

I don't think it's a common need, though; not nearly as common as simply
wanting to republish the packages in a given repo under a new publisher.
I'm happy requiring merging across publishers to be a two-step process.
(We can likely make this perform pretty well, if we can support having the
source and destination repos the same, but that's another thread.)

> The other reason I've been reticent to put such functionality into pkgrecv
> is because it implicitly alters the packages, invalidating any signatures.
> Placing that functionality into pkgmerge instead seems desirable so that we
> can continue to keep our 'publication' and 'administrative' tools separate.

Except that pkgmerge is really just a specialty tool for creating
multivariant packages.  I don't think it should be anything more than that.

I appreciate the idea behind preventing pkgrecv from making any changes,
but fundamentally what we need to be able to do is transform published
packages, and I think that pkgrecv is the best place for that functionality
(especially because I also really don't want to introduce yet another
commandline tool for this).  I'm just not committed to a side-effect-less
pkgrecv.  I don't think you should be able to do it by accident, but I
think pkgrecv is the right place.  (And, I still think, through the use of
pkgmogrify transforms.)

That said, I think Tim's -p is fine.

Danek
_______________________________________________
pkg-discuss mailing list
pkg-discuss@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to