1 |
On Sun, Jul 21, 2013 at 6:33 PM, Michał Górny <mgorny@g.o> wrote: |
2 |
> Dnia 2013-07-22, o godz. 00:16:31 |
3 |
> hasufell <hasufell@g.o> napisał(a): |
4 |
>> - users have to run "layman -a foo" ...I hope they will manage (and the |
5 |
>> masking reason will be updated to explain where to look for googleearth |
6 |
>> ebuilds) |
7 |
> |
8 |
> Then to get *a single package* they start using the whole overlay. If |
9 |
> it's a sane overlay, fine. But some overlays really replace a lot of |
10 |
> stuff silently and trigger failures we didn't even imagine before. |
11 |
> |
12 |
|
13 |
We're starting to drift off topic here, but I've always felt that this |
14 |
is something that could be improved on. I'm not saying that any of |
15 |
this should just be thrown together, but some of the following might |
16 |
be useful: |
17 |
|
18 |
1. Overlay dependencies (depends on a particular package from a |
19 |
particular overlay). |
20 |
2. Overlay package.* (accept one version of one package from a |
21 |
particular overlay, mask all packages in an overlay that aren't |
22 |
explicitly unmasked, don't apply package.(un)mask from one overlay to |
23 |
packages in another unless the mask references that overlay |
24 |
specifically, etc). |
25 |
3. Repository priority (paludis handles this fairly well I think) - |
26 |
where you can control which overlays override which other ones. |
27 |
|
28 |
I've never really liked the all-or-nothing design of overlays as they |
29 |
currently stand. Even simply prioritizing them is a pretty crude |
30 |
approach. It makes them fairly useless for general-purpose |
31 |
distribution of packages. |
32 |
|
33 |
Rich |