1 |
On Fri, 27 Jun 2014 14:41:01 +0000 |
2 |
bugzilla-daemon@g.o wrote: |
3 |
|
4 |
> https://bugs.gentoo.org/show_bug.cgi?id=513882 |
5 |
> |
6 |
> --- Comment #6 from Jeroen Roovers <jer@g.o> --- |
7 |
> The common procedure goes like this: |
8 |
> |
9 |
> 1) Put the new split-off library in the tree and make it block older |
10 |
> versions of the ebuilds that use it "internally". |
11 |
> 2) Put a new ebuild in the tree that depends on that library, and |
12 |
> drop keywords on that ebuild that can't resolve the new dependency. |
13 |
> 3) Ask for keywords to be restored for both the library and the new |
14 |
> ebuild. |
15 |
> |
16 |
> This way people can test the new torque ebuild against the split-off |
17 |
> library ebuild. And you don't need to wait at all to put the new |
18 |
> torque ebuild in the tree. |
19 |
|
20 |
I keep wondering why I need to remind people how this works dozens of |
21 |
times a year. Should we document this better somewhere? |
22 |
|
23 |
devmanual[1] has simply this: |
24 |
|
25 |
"Sometimes you may need to remove a keyword because of new |
26 |
unresolved dependencies. If you do this, you must file a bug |
27 |
notifying the relevant arch teams." |
28 |
|
29 |
and it seems that is not enough since people needlessly wait for the |
30 |
KEYWORDREQ to be resolved before they do anything. |
31 |
|
32 |
|
33 |
jer |
34 |
|
35 |
|
36 |
|
37 |
[1] http://devmanual.gentoo.org/keywording/index.html |