| 1 |
On Sat, 23 Jun 2012 22:09:31 +0200 |
| 2 |
"Andreas K. Huettel" <dilfridge@g.o> wrote: |
| 3 |
> It is certainly possible refine this more. Allowing only EAPI=5 |
| 4 |
> ebuilds in package.use.stable.*, and have the files only take effect |
| 5 |
> there, comes to my mind. However I doubt if these restrictions are |
| 6 |
> really necessary and if a EAPI dependence at this place in the |
| 7 |
> profile makes actually sense. |
| 8 |
|
| 9 |
The way we usually word such things is to have a table of EAPIs where |
| 10 |
support is required if the package mangler accepts indicated EAPIs. Then |
| 11 |
it's an error for ebuilds to rely upon support if they don't use one of |
| 12 |
those EAPIs. This gets you out of the profile EAPI requirement. |
| 13 |
|
| 14 |
One thing that isn't addressed is what "stable" means. PMS doesn't |
| 15 |
currently attach that level of meaning to KEYWORDS. In particular, |
| 16 |
there's nothing implied about the relationship between ~x86 and x86. |
| 17 |
|
| 18 |
This matches how things were when the relevant parts were written. I'm |
| 19 |
not sure if Portage changed since then, but in the good old days, you'd |
| 20 |
get crazy things like amd64 users specifying ACCEPT_KEYWORDS="~amd64 |
| 21 |
~x86", which would be auto-merged with "amd64" from make.defaults. But |
| 22 |
that wouldn't result in x86 being accepted, so if someone stabled a |
| 23 |
package that was keyworded only ~x86, it could result in it becoming |
| 24 |
masked to users with that kind of wonky configuration... |
| 25 |
|
| 26 |
-- |
| 27 |
Ciaran McCreesh |
Archives