| 1 |
W dniu pon, 04.09.2017 o godzinie 13∶52 -0400, użytkownik Michael |
| 2 |
Orlitzky napisał: |
| 3 |
> On 09/04/2017 01:36 PM, Michał Górny wrote: |
| 4 |
> > |
| 5 |
> > I don't like marking profiles non-stable because it disables CI |
| 6 |
> > checks on pull requests, so whatever people do with those keywords, |
| 7 |
> > we no longer verify it. |
| 8 |
> |
| 9 |
> I think that's the idea: I could remove the ebuilds with two-year old |
| 10 |
> security problems that are still waiting for a sparc stabilization and |
| 11 |
> not be accused of breaking the tree. |
| 12 |
> |
| 13 |
|
| 14 |
I would really prefer if our priority would be making a good |
| 15 |
distribution rather than making excuses for breaking stuff purposefully. |
| 16 |
|
| 17 |
As far as I'm concerned, if you consider degrading an architecture to |
| 18 |
non-stable profile, then it's pretty much equivalent to killing |
| 19 |
the support for the architecture altogether. Especially when the sole |
| 20 |
purpose is to let people break everything randomly. |
| 21 |
|
| 22 |
I really do think that it'd be better to drop everything to ~arch first, |
| 23 |
at least for sparc, and see how that works out. If arch teams still |
| 24 |
can't cope with the requests, we can look into removing keywords from |
| 25 |
individual packages along with fixing the dependency graph. Allowing |
| 26 |
broken dependency graph should only be a last resort. |
| 27 |
|
| 28 |
Compare this with arm64, and how much effort it takes for the arm64 team |
| 29 |
to fix their completely broken dependency graph. And until they finish |
| 30 |
that, we can't enable testing for arm64 depgraph by default, so people |
| 31 |
may still commit breakage. |
| 32 |
|
| 33 |
-- |
| 34 |
Best regards, |
| 35 |
Michał Górny |
Archives