| 1 |
On Sun, 2022-01-09 at 09:15 +0100, Michał Górny wrote: |
| 2 |
> The Python distutils deprecation is for real, and nowadays setuptools |
| 3 |
> installs hacks that override distutils with their own bundled version. |
| 4 |
> For this reason, the check no longer works reliably and even if it were, |
| 5 |
> we will eventually be requiring setuptools on all Python packages. |
| 6 |
> Avoid asking people to set DISTUTILS_USE_SETUPTOOLS=no; allow either |
| 7 |
> "no" or "bdepend" for the time being. |
| 8 |
> |
| 9 |
|
| 10 |
To be clearer: |
| 11 |
|
| 12 |
If you have setuptools-60+ installed (currently p.masked but not for |
| 13 |
very long), the check will always detect setuptools and demand "bdepend" |
| 14 |
from packages using pure distutils. With older setuptools, the current |
| 15 |
version will demand "no" instead -- effectively two devs could get |
| 16 |
conflicting demands. |
| 17 |
|
| 18 |
This change prevents that by allowing "bdepend" or "no" with |
| 19 |
setuptools<60. Setuptools 60+ will still demand "bdepend" and this is |
| 20 |
where we'll all eventually end up. |
| 21 |
|
| 22 |
-- |
| 23 |
Best regards, |
| 24 |
Michał Górny |
Archives