1 |
AFAIK, we could manage the entire thing via cvs - no incoming dir, no |
2 |
commit-request emails. |
3 |
|
4 |
We would have several 'trees' in cvs: stable, unstable (=devel), external |
5 |
(any better names?), incoming (=user-submitted), trees for past releases |
6 |
which could get bugfixes, etc. |
7 |
|
8 |
A developer who wanted to change another developer's ebuild would have to |
9 |
commit a new revision of that ebuild to the 'external' tree. (That's what |
10 |
he'd have permissions to do.) The maintaining developer could then look over |
11 |
his changes (and the commit log message), and if he accepted them, merge the |
12 |
changes into his main devel tree. Once in there, the changes needed to be |
13 |
proven stable and good (like the maintainer's own work) and would then be |
14 |
merged into the stable tree. Non-developers who did 'emere rsync' could only |
15 |
get things from the stable tree. |
16 |
|
17 |
A similar procedure wuold be used for 'incoming' user ebuilds, with another |
18 |
cvs tree. |
19 |
|
20 |
What do you think? |
21 |
|
22 |
|
23 |
On Friday 03 August 2001 18:08, you wrote: |
24 |
> Hi! |
25 |
> |
26 |
> I renamed the thread since it's not about xchat 1.8.2 anymore. |
27 |
> |
28 |
> > I think if we choose to do this, then we had better rethink the |
29 |
> > incoming directory right now. While it may be okay for a few |
30 |
> > outstanding user-submitted ebuilds, this scheme has the potential to |
31 |
> > cause a lot problems. What we really need is an unstable cvs tree; |
32 |
> > then user-submitted ebuilds and non-team developer-hacked ebuilds can |
33 |
> > go there until they are verified by team members. |
34 |
> |
35 |
> If we have maintainer of ebuilds (which I think we shall have) I don't |
36 |
> think that other people should commit into the CVS tree without his |
37 |
> aproval (neither unstable nor stable). It will be very hard for a |
38 |
> maintainer of many packages to keep track of what's happening if |
39 |
> people just commit at will. |
40 |
> |
41 |
> > I think all we need is maintainer. This is the guy who "puts his rep |
42 |
> > on the line." cvs log keeps all the other information. The person |
43 |
> > you really want to complain to about a bad ebuild is the maintainer |
44 |
> > anyway. Then he can do a cvs log to see who the culprit is and get in |
45 |
> > touch. |
46 |
> |
47 |
> This will not be a problem if only the maintainer are allowed to |
48 |
> commit (or aprove commits). I think this is the best way to do it. |
49 |
> |
50 |
> If you want to make a commit to an ebuild that you are not maintainer |
51 |
> of: |
52 |
> |
53 |
> If you are a developer: |
54 |
> 1) Send the patch to the maintainer asking if you may commit. |
55 |
> 2) The maintainer might commit or aprove of you commiting or tell you |
56 |
> why it won't go in (and perhaps what you need to fix first). |
57 |
> |
58 |
> Otherwise: |
59 |
> 1) Send the patch to the maintainer which will either commit or tell |
60 |
> you to change the patch in some way. |
61 |
> |
62 |
> This will make it harder for people to commit to all kinds of ebuilds |
63 |
> but I think that this is the way we have to do it in the future. |
64 |
> |
65 |
> Of course a maintainer might give a certain person commit rights if he |
66 |
> trusts the other developer to do it right. |
67 |
> |
68 |
> Regards, |
69 |
> Mikael Hallendal |
70 |
|
71 |
-- |
72 |
|
73 |
Dan Armak |
74 |
Gentoo Linux Developer, Desktop Team |
75 |
Matan, Israel |