| 1 |
On Mon, 2006-06-12 at 14:15 -0700, Brian Harring wrote: |
| 2 |
> On Mon, Jun 12, 2006 at 09:58:01PM +0100, Stephen Bennett wrote: |
| 3 |
> > Many things were discussed in the last round of this thread (Paludis |
| 4 |
> > and Profiles, in case anyone missed it), and many useful points raised. |
| 5 |
> > One of these, which seems to have been largely missed in amongst the |
| 6 |
> > other noise, forms the basis of this proposal. It is in some ways more |
| 7 |
> > and in some ways less intrusive than the previous proposal, |
| 8 |
> > and is also completely package-manager-agnostic. |
| 9 |
> > |
| 10 |
> > In short, I would like to suggest replacing sys-apps/portage atoms in |
| 11 |
> > the base and default-linux profiles with virtual/portage, and removing |
| 12 |
> > the python dependencies from them. For most users this would have an |
| 13 |
> > effective zero change, since the default provider for virtual/portage |
| 14 |
> > is sys-apps/portage, and the python dependency will be pulled in by |
| 15 |
> > Portage when calculating system deps. According to my understanding, |
| 16 |
> > this should also produce no change when building release media, due to |
| 17 |
> > both Portage and Python being in packages.build. |
| 18 |
> > |
| 19 |
> > The only change introduced by this is that it becomes possible to |
| 20 |
> > bootstrap a system with a different package manager simply by |
| 21 |
> > installing it before 'system'. There are a couple more changes needed |
| 22 |
> > to allow this -- namely that a few system packages have old |
| 23 |
> > dependencies on >=portage-2.0.49, but these can be resolved seperately. |
| 24 |
> > Any problems caused by packages depending implicitly upon Python will |
| 25 |
> > show up only on systems not using Portage, and can be easily fixed with |
| 26 |
> > the cooperation of package maintainers. |
| 27 |
> > |
| 28 |
> > I would like to think that this proposal addresses most of the concerns |
| 29 |
> > raised in the last thread -- it implies nothing about support for any |
| 30 |
> > other package manager, and introduces nothing that could cause problems |
| 31 |
> > for Portage users, while still allowing alternative package managers to |
| 32 |
> > use the tree without needing Portage installed. |
| 33 |
> > |
| 34 |
> > I am also aware that this falls roughly under what the Council was |
| 35 |
> > asked to discuss in its June meeting, but since that seems to have not |
| 36 |
> > happened, I'm bringing it up anyway, since I would like to get |
| 37 |
> > something done here. |
| 38 |
> |
| 39 |
> +1, dependant on A) catalyst folk not poking holes in it, B) council |
| 40 |
> outcome tomorrow (no point in changing it till they've weighed in on |
| 41 |
> the whole enchilada). |
| 42 |
|
| 43 |
As the "catalyst folk" I can say that it shouldn't affect us in any way. |
| 44 |
So long as packages.build still has sys-apps/portage, packages has |
| 45 |
virtual/portage, and virtuals has sys-apps/portage as the default for |
| 46 |
the virtual, it won't affect us. |
| 47 |
|
| 48 |
My only request is that this work gets held off on until after we make |
| 49 |
our 2006.1 snapshot, to reduce the possibility of us having to hunt down |
| 50 |
possibly hundreds of potential problems in release-building. |
| 51 |
|
| 52 |
-- |
| 53 |
Chris Gianelloni |
| 54 |
Release Engineering - Strategic Lead |
| 55 |
x86 Architecture Team |
| 56 |
Games - Developer |
| 57 |
Gentoo Linux |
Archives