| 1 |
W dniu 21.01.2017, sob o godzinie 23∶59 +0100, użytkownik Michał Górny |
| 2 |
napisał: |
| 3 |
> The main goal was to reduce the role of features/* subtree that |
| 4 |
> served |
| 5 |
> as some kind of weird all-in-one middleware for multiple profiles. |
| 6 |
> These |
| 7 |
> patches: |
| 8 |
> |
| 9 |
> a. kill features/*bit-* completely -- all stuff is moved to |
| 10 |
> appropriate |
| 11 |
> arch profiles, |
| 12 |
> |
| 13 |
> b. kill LIBDIR declarations out of features/multilib -- also moved to |
| 14 |
> appropriate arch profiles. |
| 15 |
> |
| 16 |
> The main goal is simplicity. I think it's really better when you |
| 17 |
> don't |
| 18 |
> have to guess which of 10 implicitly inherited profiles set the final |
| 19 |
> value, and it's really bad when minor parent order changes in some |
| 20 |
> profile can randomly effect the results of a derived one. |
| 21 |
> |
| 22 |
> Right now this is terrible. For example, the arch/amd64/x32 profile |
| 23 |
> inherited features/multilib a second time just to reset libdir which |
| 24 |
> has |
| 25 |
> been indirectly set by parent profile. |
| 26 |
> |
| 27 |
> I can't stay this is enough to make things perfect but I think it's |
| 28 |
> a step in the right direction. Cleaner profiles make it easier to |
| 29 |
> improve them more in the future. |
| 30 |
|
| 31 |
Merged. |
| 32 |
|
| 33 |
-- |
| 34 |
Best regards, |
| 35 |
Michał Górny |
Archives