| 1 |
Am Mittwoch, 18. November 2015, 12:12:05 schrieb Alexander Berntsen: |
| 2 |
> On 18/11/15 12:05, Ulrich Mueller wrote: |
| 3 |
> > Only that there is no real difference to the existing situation |
| 4 |
> > when mixing stable and unstable. It is not guaranteed that all |
| 5 |
> > dependencies of an unstable package are stable, so already now |
| 6 |
> > users may have to accept the ~ keyword for dependencies in some |
| 7 |
> > cases. Similarly, such users may have to accept EAPI 6 for some |
| 8 |
> > dependencies, which implies that they install a package manager |
| 9 |
> > supporting EAPI 6. |
| 10 |
> |
| 11 |
> There's a difference between some packages being troublesome, and |
| 12 |
> encouraging everyone to rewrite their eclasses and ebuilds, if the end |
| 13 |
> result is a huge portion of ebuilds causing headaches. |
| 14 |
|
| 15 |
Well, at some point it has to be introduced in the main tree. |
| 16 |
Can you prove at any point that portage is 100% correct? |
| 17 |
|
| 18 |
Also, adding EAPI=6 support to eclasses mostly consists of adding branches to |
| 19 |
case statements. I.e. the new code paths will never run on old EAPI. |
| 20 |
|
| 21 |
> > And on what basis would you stabilise Portage, when there are no |
| 22 |
> > ebuilds in the tree to test its EAPI 6 code? |
| 23 |
> |
| 24 |
> When I do QA in projects I'm involved with (at least outside of |
| 25 |
> Gentoo), we don't do it live on end-user systems. I'll leave the |
| 26 |
> details as an exercise for the Gentoo developer. |
| 27 |
|
| 28 |
So, I suggest you branch gentoo.git, start adding some new ebuilds to it |
| 29 |
(don't forget to use a random combination of eclasses, like perl-module, |
| 30 |
python-r1, kde4-base ...), update your system and check for all possible |
| 31 |
resolver oddities... Too much work? Tough. |
| 32 |
|
| 33 |
-- |
| 34 |
Andreas K. Huettel |
| 35 |
Gentoo Linux developer (council, perl, libreoffice) |
| 36 |
dilfridge@g.o |
| 37 |
http://www.akhuettel.de/ |
Archives