| 1 |
On Sat, Aug 28, 2021 at 06:35:06PM +0200, Michał Górny wrote: |
| 2 |
> Hi, |
| 3 |
> |
| 4 |
> I've been informed of a slight inconsistency in package manager behavior |
| 5 |
> that affects combining EXPORT_FUNCTIONS with inherit (by ionic, thanks |
| 6 |
> for the report!). Please consider the three following snippets: |
| 7 |
|
| 8 |
... |
| 9 |
|
| 10 |
> 1. I'd like to propose that we explicitly require all inherits to happen |
| 11 |
> before EXPORT_FUNCTIONS. This will ensure consistent behavior across |
| 12 |
> all package managers. |
| 13 |
> |
| 14 |
> 2. I'd like to ask your opinion whether we should: |
| 15 |
> |
| 16 |
> a. revert the Portage behavior to be consistent with PkgCore/Paludis |
| 17 |
> |
| 18 |
> b. update PMS to identify the behavior as 'undefined', i.e. either |
| 19 |
> solution is correct. |
| 20 |
|
| 21 |
I would go with 1 and 2 b, but I also propose another option for the |
| 22 |
next eapi which may be a bit controversial. |
| 23 |
|
| 24 |
Starting with eapi 9, make export_functions a noop (you can't remove it |
| 25 |
until all eclasses/ebuilds only support eapis that don't require it). |
| 26 |
|
| 27 |
I understand that this is controversial, because it would require a lot |
| 28 |
of work to convert ebuilds to eapi 9, but it would eliminate this |
| 29 |
ambiguity/inconsistency in the future because it would require all |
| 30 |
ebuilds to have phase functions unless they can use the default phases |
| 31 |
the eapis provide. |
| 32 |
|
| 33 |
Thoughts? |
| 34 |
|
| 35 |
William |
| 36 |
|
| 37 |
> |
| 38 |
> WDYT? |
| 39 |
> |
| 40 |
> |
| 41 |
> [1] https://github.com/gentoo/portage/commit/06d4433e8b8be60d606733b9e23f57f8a5869d8f |
| 42 |
> |
| 43 |
> -- |
| 44 |
> Best regards, |
| 45 |
> Michał Górny |
| 46 |
> |
| 47 |
> |
| 48 |
> |
Archives