| 1 |
On Mon, 08 Dec 2008 19:25:44 -0500 |
| 2 |
Olivier Crête <tester@g.o> wrote: |
| 3 |
> > The "can be tested properly" phase is when it's in ~arch... |
| 4 |
> |
| 5 |
> That also means that to pull a significant number of ebuilds it forces |
| 6 |
> mostly everyone to test it.. and that part is annoying.. |
| 7 |
|
| 8 |
If you don't like it, don't run ~arch. |
| 9 |
|
| 10 |
> The testing should be two phased, the first for regression (against |
| 11 |
> existing ebuilds), and once thats stable, then we can test with new |
| 12 |
> ebuilds... |
| 13 |
|
| 14 |
Uh, regression testing's handled by the package manager's extensive set |
| 15 |
of unit tests, which can cover this with targetted accuracy with much |
| 16 |
more reliability than making sure random ebuilds still work. |
| 17 |
|
| 18 |
What you're suggesting here is making everyone wait four more months |
| 19 |
for no increase in safety. |
| 20 |
|
| 21 |
-- |
| 22 |
Ciaran McCreesh |
Archives