1 |
>>>>> On Thu, 29 Mar 2018, Michał Górny wrote: |
2 |
|
3 |
>> Stable Portage supports EAPI 6 since 2016-01-17, i.e. since 26 |
4 |
>> months. So we would be somewhat on the early side. |
5 |
|
6 |
> Not that it's less than the supported upgrade path. |
7 |
|
8 |
Yes, I don't think that we have a problem there. Just noting that's |
9 |
it would be sooner than for all previous EAPIs. |
10 |
|
11 |
>> What worries me more is that deprecation of EAPI 5 would apply to |
12 |
>> profiles too. However, all profiles are still at EAPI 5 at this |
13 |
>> point, and I don't see any value in upgrading them to EAPI 6. |
14 |
|
15 |
> That's a fair argument. However: |
16 |
|
17 |
> 1. Does deprecation really mean anything in terms of profiles? Even |
18 |
> in the context of EAPI bans we explicitly stated that it affects new |
19 |
> packages and EAPI bumps. I think deprecating it for ebuilds is still |
20 |
> meaningful even if profiles would stay EAPI 5. |
21 |
|
22 |
OK, but then we should clearly state this. |
23 |
|
24 |
> 2. Do we want to keep profiles EAPI 5 indefinitely? If we consider |
25 |
> it a goal to reduce the number of EAPIs in use, I think it would be |
26 |
> reasonable to bump profiles to EAPI 6 proactively, even if it |
27 |
> doesn't change anything. |
28 |
|
29 |
The only effect this has is that it can impede some users' upgrade |
30 |
path. Or is there any feature in EAPI 6 that is needed in profiles? |
31 |
|
32 |
Another way to keep the number of EAPIs limited would be to skip |
33 |
EAPI 6 for profiles. We have done that for EAPIs 3 and 4 previously |
34 |
(i.e., all previous and current profiles were EAPI 0, 1, 2, or 5). |
35 |
|
36 |
Ulrich |