1 |
On Mon, 6 Apr 2015 09:59:22 +0200 Michał Górny wrote: |
2 |
[...] |
3 |
> > Hmm, that's a hard question. I tried to consider this issues from a |
4 |
> > point of view of a user of such arch. If package is not used or |
5 |
> > user may delete it and its deps without much harm, it doesn't affect |
6 |
> > user at all. If it is used and needed, then in case of: |
7 |
> > |
8 |
> > - one package with removed stable keyword a user have to add to |
9 |
> > package.keywords only a single package, though it might be |
10 |
> > difficult to locate such package, because portage deptree failure |
11 |
> > events may be really obscure sometimes; |
12 |
> > |
13 |
> > - all subtree of stable keywords is removed; then user have to |
14 |
> > add all these packages to package.keywords, portage messages should |
15 |
> > be clear here (but one never knows), though manual keywording of |
16 |
> > hundred of packages will be irritating at best (even using "cat/*" |
17 |
> > masks). So if number of affected installed packages is large, users |
18 |
> > will likely move to ~arch all their setup. |
19 |
> > |
20 |
> > So from user's perspective stable deptree broken in single point is |
21 |
> > a better solution, but(!) if portage will cleanly suggest this |
22 |
> > point. |
23 |
> > |
24 |
> > Another issue to consider: what if we have one such package that |
25 |
> > broke stable deptree, then after awhile another one and so on. In |
26 |
> > the result stable tree will got corrupted beyond repair. |
27 |
> > |
28 |
> > Maybe some grace period will help here? E.g. remove stable keyword |
29 |
> > from a single package, wait for 30 days (or so) for reaction from a |
30 |
> > team, and then dekeyword all reverse dependencies. |
31 |
> |
32 |
> Which means developers can't commit properly for 30 days. Really |
33 |
> awesome solution, thank you. Please ping QA instantly once you do it, |
34 |
> you'll save us time figuring who to ban for the breakage. |
35 |
|
36 |
I sad nowhere I'm going to do it. We are discussing opportunities |
37 |
to fix current sore situation. Threatening people with bans just |
38 |
for their thoughts reminds me of George Orwell's Thought Police... |
39 |
|
40 |
> Let me remind you: once you break dependency tree on *ANY* stable |
41 |
> architecture, repoman won't let you commit. Travis turns red. All pull |
42 |
> requests are marked as broken. |
43 |
|
44 |
Repoman, travis and other QA tools can be fixed if policy changes. |
45 |
That's not a problem. Right now we have an all-green travis, but |
46 |
outdated, broken and likely insecure packages it tree marked as |
47 |
stable. That is the problem we're trying to discuss here. |
48 |
|
49 |
And as I see this problem has no good solution, so one less painful |
50 |
for users should be chosen. If this will require a QA policy |
51 |
update, then it should be done. |
52 |
|
53 |
Best regards, |
54 |
Andrew Savchenko |