| 1 |
On Sat, 2020-11-07 at 15:18 -0700, Tim Harder wrote: |
| 2 |
> In terms of QA, unintentional transitive eclass usage is generally bad. |
| 3 |
> This occurs when an ebuild uses functionality from an eclass it doesn't |
| 4 |
> directly inherit. It would be useful for eclasses that allow certain |
| 5 |
> transitive usage (e.g. various python eclasses) to be able to tag that |
| 6 |
> relationship internally so tools can make use of that data. |
| 7 |
> |
| 8 |
> Along those lines, pkgcheck now has eclass doc parsing support which |
| 9 |
> allows scanning ebuilds for missing, indirect, or unused eclass inherits |
| 10 |
> as well as internal eclass function usage. In order to more closely |
| 11 |
> report valid indirect inherit results, some tag including this data |
| 12 |
> needs to be included for eclasses allowing this relationship. |
| 13 |
> |
| 14 |
> What do interested parties think about including an optional eclass doc |
| 15 |
> tag such as '@TRANSITIVE_INHERITS:' or other similar name in eclasses |
| 16 |
> that allow this? The tag value would be a space-separated list of |
| 17 |
> allowed transitive inherits for the given eclass. |
| 18 |
> |
| 19 |
|
| 20 |
Technically speaking, I would go even further and say that listing |
| 21 |
python-utils-r1 redundantly is wrong. This inheritance is considered |
| 22 |
an implementation detail of the eclass. |
| 23 |
|
| 24 |
-- |
| 25 |
Best regards, |
| 26 |
Michał Górny |
Archives