| 1 |
Am Mittwoch, 3. November 2021, 16:03:37 CET schrieb Thomas Deutschmann: |
| 2 |
> Hi, |
| 3 |
> |
| 4 |
> it is currently not possible to smoothly run a world upgrade on a 4 |
| 5 |
> months old system which doesn't even have a complicated package list: |
| 6 |
> |
| 7 |
[snip] |
| 8 |
|
| 9 |
Yup. We know. It's actually way worse than you describe [*] and was |
| 10 |
noticed already quite some time ago. Unfortunately this is a situation |
| 11 |
that can IMHO not be easily fixed, and we can only strive to do it |
| 12 |
better next time. |
| 13 |
|
| 14 |
The mistake was to allow the use of EAPI=8 too early. In the future, we |
| 15 |
should have a new EAPI supported by portage for at least some months |
| 16 |
before the EAPI is even used in the main tree. Not even speaking about |
| 17 |
stable here. |
| 18 |
|
| 19 |
From there on all the trouble cascades. And no, disallowing a new EAPI |
| 20 |
for only a part of the tree does not help. (Which part?) |
| 21 |
|
| 22 |
An alternative, which we should seriously consider (and which I've been |
| 23 |
advocating for several months now) is to make Portage more robust, so |
| 24 |
it can more easily upgrade itself, and keep the Portage ebuild at old |
| 25 |
EAPI. This means, |
| 26 |
* making Portage independent of the python eclasses, so it runs as long |
| 27 |
as any python3 interpreter exists |
| 28 |
* and bundling all Python dependencies it needs for functioning in it |
| 29 |
|
| 30 |
|
| 31 |
|
| 32 |
[*] Of course there are ways around this to upgrade the system. However, |
| 33 |
that is not the point. It should work out of the box. |
| 34 |
|
| 35 |
-- |
| 36 |
Andreas K. Hüttel |
| 37 |
dilfridge@g.o |
| 38 |
Gentoo Linux developer |
| 39 |
(council, toolchain, base-system, perl, libreoffice) |
Archives