1 |
On Fri, 2 Sep 2016 18:13:20 +0200 |
2 |
Kristian Fiskerstrand <k_f@g.o> wrote: |
3 |
|
4 |
> I'm wondering whether it wouldn't make sense to require eclasses (or |
5 |
> strongly encourage) to include an explicit list of EAPIs it has been |
6 |
> tested for in order to ease testing when introducing new EAPIs. |
7 |
> |
8 |
> We have seen some issues already with EAPI6 bump related to get_libdir |
9 |
> and people updating EAPI in ebuild without properly testing the |
10 |
> inherited eclasses. having a whitelist in place and die if eclass is not |
11 |
> updated to handle it solves it. |
12 |
> |
13 |
> Thoughts? comments? cookies? threats? |
14 |
|
15 |
+1. Because: |
16 |
|
17 |
1. it makes it possible to change API safely with EAPI bump, including |
18 |
major refactorings, |
19 |
|
20 |
2. it makes it possible for the eclass maintainer to confirm that |
21 |
the eclass is correctly ported to new EAPI, rather than some random |
22 |
developer with poor knowledge of eclass assuming it works fine, |
23 |
|
24 |
3. it makes it possible to ban the eclass in a new EAPI to more |
25 |
effectively phase it out. |
26 |
|
27 |
This only reminds me of the cases when eclasses weren't calling |
28 |
eapply_user in EAPI 6 and caused ebuilds to fail. Because someone |
29 |
assumed that if his ebuild works in EAPI 6, then the eclass is |
30 |
certainly correct. |
31 |
|
32 |
-- |
33 |
Best regards, |
34 |
Michał Górny |
35 |
<http://dev.gentoo.org/~mgorny/> |