1 |
On Sun, 5 Apr 2015 18:54:10 -0400 Rich Freeman wrote: |
2 |
> On Sun, Apr 5, 2015 at 5:27 PM, Andrew Savchenko <bircoph@g.o> wrote: |
3 |
> > |
4 |
> > 2. Otherwise allow developers to drop stable keywords from affected |
5 |
> > package and _all_ its reverse dependencies. This way a part of |
6 |
> > stable tree will be removed, but only a part. With this approach |
7 |
> > arch teams will be freed of an extra burden, while they will be |
8 |
> > still able to maintain a smaller stable tree. |
9 |
> > |
10 |
> > This is a win-win solution: a stable tree will be still kept in a |
11 |
> > maintainable size and developers will not have a long-term blockers |
12 |
> > on their stabilization requests. |
13 |
> > |
14 |
> > 3. And last but not the least: apply the rules above to all arches, |
15 |
> > not just minor teams (though probability that amd64/x86 will be |
16 |
> > slow is lower, of course). |
17 |
> > |
18 |
> |
19 |
> This was some of what I was getting at. My question still stands that |
20 |
> I'm not sure arch teams REALLY want 300 packages to have their stable |
21 |
> keywords removed instead of just having one package break the |
22 |
> depgraph. |
23 |
|
24 |
Hmm, that's a hard question. I tried to consider this issues from a |
25 |
point of view of a user of such arch. If package is not used or |
26 |
user may delete it and its deps without much harm, it doesn't affect |
27 |
user at all. If it is used and needed, then in case of: |
28 |
|
29 |
- one package with removed stable keyword a user have to add to |
30 |
package.keywords only a single package, though it might be |
31 |
difficult to locate such package, because portage deptree failure |
32 |
events may be really obscure sometimes; |
33 |
|
34 |
- all subtree of stable keywords is removed; then user have to |
35 |
add all these packages to package.keywords, portage messages should |
36 |
be clear here (but one never knows), though manual keywording of |
37 |
hundred of packages will be irritating at best (even using "cat/*" |
38 |
masks). So if number of affected installed packages is large, users |
39 |
will likely move to ~arch all their setup. |
40 |
|
41 |
So from user's perspective stable deptree broken in single point is |
42 |
a better solution, but(!) if portage will cleanly suggest this |
43 |
point. |
44 |
|
45 |
Another issue to consider: what if we have one such package that |
46 |
broke stable deptree, then after awhile another one and so on. In |
47 |
the result stable tree will got corrupted beyond repair. |
48 |
|
49 |
Maybe some grace period will help here? E.g. remove stable keyword |
50 |
from a single package, wait for 30 days (or so) for reaction from a |
51 |
team, and then dekeyword all reverse dependencies. |
52 |
|
53 |
> I would prefer to get at a more generic policy that can be applied |
54 |
> everywhere, and not just arch by arch. Being able to keep up or not |
55 |
> isn't really a black/white thing. Or rather, if it is I think it is |
56 |
> more a case that nobody can keep up. I think that it would be better |
57 |
> to have one policy that makes sense on any arch, and as you point out |
58 |
> it probably won't tend to get applied much to amd64/x86 simply because |
59 |
> they are better supported. |
60 |
|
61 |
Agreed, the policy should be generic and for everyone (maybe with |
62 |
reasonable exceptions as toolchan or other core packages). |
63 |
|
64 |
Best regards, |
65 |
Andrew Savchenko |