| 1 |
On Thu, Jul 27, 2017 at 07:03:25PM -0500, Denis Dupeyron wrote: |
| 2 |
> On Thu, Jul 27, 2017 at 6:41 PM, Rich Freeman <rich0@g.o> wrote: |
| 3 |
> > |
| 4 |
> > When in the last 16 years was this 2 year period of running stable? |
| 5 |
> > The general state of QA has varied quite a bit over that time. |
| 6 |
> > |
| 7 |
> |
| 8 |
> I would say 3 or 4 years ago, roughly. |
| 9 |
> |
| 10 |
> |
| 11 |
> > running unstable systemd has been |
| 12 |
> |
| 13 |
> |
| 14 |
> Running unstable doesn't mean being stupid. |
| 15 |
|
| 16 |
Exactly. If you are running unstable, you are expected to know how to |
| 17 |
pick up the pieces if something breaks. Unstable is not meant for |
| 18 |
people who aren't comfortable with occasional breakage. |
| 19 |
|
| 20 |
> > If unstable never breaks chances are we aren't actually using it for its |
| 21 |
> > intended purpose, not that we |
| 22 |
> > should be deliberately breaking things. |
| 23 |
> |
| 24 |
> There's this idea that unstable should break. But the initial idea was that |
| 25 |
> unstable is what should be sent straight to stable, barring the occasional |
| 26 |
> mistake. Unstable was never meant for ebuilds in development and very much |
| 27 |
> in flux because of that. That's what package masks are for. |
| 28 |
|
| 29 |
package masks are specifically for things that are known to cause major |
| 30 |
system breakages. |
| 31 |
|
| 32 |
Maintainers test things to the best of their |
| 33 |
ability before putting them in the tree, but may not cover all possible |
| 34 |
test cases, so packages go to the ~arch tree to get much wider coverage |
| 35 |
before they are deamed stable. That is why there is a recommended delay |
| 36 |
of 30 days before the package moves to stable. |
| 37 |
|
| 38 |
William |
Archives