1 |
On 2021-08-16 18:06, Joonas Niilola wrote: |
2 |
|
3 |
> What's the rush with 8 anyway? We still have eclasses that don't even |
4 |
> support 7. And since 6 is now, sigh, deprecated, any bump to ebuilds |
5 |
> depending on those eclasses will add to the evergrowing CI issue list |
6 |
> right? I'd say it's more important to secure EAPI-7 support there first. |
7 |
|
8 |
...or we could just skip supporting EAPI 7 support in such packages and |
9 |
go from 6 straight to 8. In my own experience with EAPI bumping, I would |
10 |
rank migration difficulty (with the more difficult cases listed first) |
11 |
as follows: |
12 |
|
13 |
1. 5 to 6 - a lot of changes to take into account, all of them adding up |
14 |
to a massive pain in the neck - especially if patches have to be |
15 |
regenerated in order to work with eapply; |
16 |
|
17 |
2. 6 to 7 - the trailing-slash-in-D-and-ED thing can be a nuisance but |
18 |
that's pretty much it, even moving packages from DEPEND to BDEPEND can |
19 |
be considered minor owing to how little breakage not doing it can cause |
20 |
(cross-building packages IS useful, that's how we initially bootstrapped |
21 |
dev-lang/go on riscv for instance, but it's not that widespread); |
22 |
|
23 |
3. 7 to 8 - everything except the bash version bump can be handled |
24 |
quickly with grep and mgorny's checklist. |
25 |
|
26 |
In short, so far I've found the cost of adding EAPI 8 support minimal |
27 |
comparing to support for earlier versions. |
28 |
|
29 |
-- |
30 |
Marecki |