1 |
On Thu, 23 Apr 2009 11:48:08 +0300 |
2 |
Mart Raudsepp <leio@g.o> wrote: |
3 |
> > The best installed version that matches the spec is always picked |
4 |
> > for :=. We originally considered making this more complicated, or |
5 |
> > possibly making ways of saying "all installed slots", but neither |
6 |
> > appears to have a legitimate use case. The latter in particular |
7 |
> > would only encourage abuse (people might mistakenly think it's a |
8 |
> > solution to the Python / Ruby ABIs thing). |
9 |
> |
10 |
> I don't think that really works in all cases, e.g in combination with |
11 |
> PDEPEND or such. |
12 |
|
13 |
It works in all cases where you're guaranteed to know at build time |
14 |
what versions are there -- that is to say, it works for things that are |
15 |
listed both in DEPEND and in RDEPEND. |
16 |
|
17 |
> So to make this easy for everybody in the future, do you perhaps have |
18 |
> a concrete ebuild code example that shows how exactly to do it in an |
19 |
> ebuild pkg_setup or src_configure exactly? |
20 |
|
21 |
Most things that are affected by this will just work automatically. |
22 |
|
23 |
> My main open issue with SLOT-OPERATOR-DEPS is about the :* syntax, and |
24 |
> specifically how that potentially works un-intuituvely with some |
25 |
> future syntax regarding list of slots. The feature concept itself |
26 |
> seems sound and reasonable, but I think we might be able to do better |
27 |
> with the syntax for the long run when slot lists come into play. |
28 |
|
29 |
It fits in fine with the syntax we've been thinking of using for slot |
30 |
groups and version ranges. |
31 |
|
32 |
-- |
33 |
Ciaran McCreesh |