1 |
Dnia 2014-09-14, o godz. 10:33:03 |
2 |
Rich Freeman <rich0@g.o> napisał(a): |
3 |
|
4 |
> > Of course, that assumes infra is |
5 |
> > going to cooperate quickly or someone else is willing to provide the |
6 |
> > infra for it. |
7 |
> |
8 |
> The infra components to a git infrastructure are one of the main |
9 |
> blockers at this point. I don't really see cooperation as the issue - |
10 |
> just lack of manpower or interest. |
11 |
|
12 |
By 'cooperating' I simply meant offering the necessary resources |
13 |
in a reasonable time. |
14 |
|
15 |
> > |
16 |
> > 1. send announcement to devs to explain how to use git, |
17 |
> |
18 |
> This is one of the blockers. We haven't actually decided how we want |
19 |
> to use git. |
20 |
> |
21 |
> Sure, everybody knows how to use git. The problem is that there are a |
22 |
> dozen different ways we COULD use git, and nobody has picked the ONE |
23 |
> way we WILL use it. |
24 |
> |
25 |
> This isn't as trivial as you might think. We have a fairly high |
26 |
> commit rate and with a single repository that means that in-between a |
27 |
> pull-merge/rebase-push there is a decent chance of another commit that |
28 |
> will make the resulting push a non-fast-forward. |
29 |
> |
30 |
> People love to point out linux and its insane commit rate. The thing |
31 |
> is, the mainline git repo with all those commits has exactly one |
32 |
> committer - Linus himself. They don't have one big repo with one |
33 |
> master branch that everybody pushes to. At least, that is my |
34 |
> understanding (and there are certainly others here who are more |
35 |
> involved with kernel development). |
36 |
|
37 |
It's hard to talk about commit rate when we combine crippled CVS with |
38 |
awfully stupid two-part repoman committing. This forces us to commit |
39 |
everything immediately, and makes some of us not committing anything |
40 |
at all anymore... |
41 |
|
42 |
With git, we can finally do stuff like preparing everything and pushing |
43 |
in one go. Rebasing or merging will be much easier then, since |
44 |
the effective push rate will be smaller than current commit rate. |
45 |
|
46 |
> > On top of user sync repo rsync is propagated. The rsync tree is populated |
47 |
> > with all old ChangeLogs copied from CVS (stored in 30M git repo), new |
48 |
> > ChangeLogs are generated from git logs and Manifests are expanded. |
49 |
> |
50 |
> So, I don't really have a problem with your design. I still question |
51 |
> whether we still need to be generating changelogs - they seem |
52 |
> incredibly redundant. But, if people really want a redundant copy of |
53 |
> the git log, whatever... |
54 |
|
55 |
I don't want them too. However, I'm pretty sure people will bikeshed |
56 |
this to death if we kill them... Especially that rsync has no git log. |
57 |
Not that many users make real use of ChangeLogs, esp. considering |
58 |
how useless messages often are there... |
59 |
|
60 |
> > Main developer repo |
61 |
> > ------------------- |
62 |
> > |
63 |
> > I was able to create a start git repository that takes around 66M |
64 |
> > as a git pack (this is how much you will have to fetch to start working |
65 |
> > with it). The repository is stripped clean of history and ChangeLogs, |
66 |
> > and has thin Manifests only. |
67 |
> > |
68 |
> > This means we don't have to wait till someone figures out the perfect |
69 |
> > way of converting the old CVS repository. You don't need that history |
70 |
> > most of the time, and you can play with CVS to get it if you really do. |
71 |
> > In any case, we would likely strip the history anyway to get a small |
72 |
> > repo to work with. |
73 |
> |
74 |
> We already have a migration process that coverts the old CVS |
75 |
> repository, generating both a shallow repository that lacks history |
76 |
> and a full repository that contains all of history. Additionally, |
77 |
> these two are consistent - that is the last branch of the full |
78 |
> repository has the same commit ID as the base of the shallow |
79 |
> repository. Basically we generate the full history and then trim out |
80 |
> 99% of it so that the commit in the shallow repository points to a |
81 |
> parent that isn't in the packed repository. |
82 |
> |
83 |
> Actually doing the conversion is basically a solved problem. If this |
84 |
> were actually the blocker I'd be all for just sticking the history in |
85 |
> a different repo and starting from scratch with a new one. |
86 |
|
87 |
Was the resulting tree actually verified? How long does the conversion |
88 |
take? Can it be incremental, i.e. convert most of it, lock CVS, convert |
89 |
the remaining new commits? |
90 |
|
91 |
> > I think we should also merge gentoo-news & glsa & herds.xml into |
92 |
> > the repository. They all reference Gentoo packages at a particular |
93 |
> > state in time, and it would be much nicer to have them synced properly. |
94 |
> > |
95 |
> |
96 |
> I can see the pros/cons here, but I don't personally have an issue |
97 |
> with merging them. As has been brought up elsewhere herds.xml may |
98 |
> just go away. |
99 |
> |
100 |
> If somebody can come up with a set of hooks/scripts that will create |
101 |
> the various trees and the only thing that is left is to get infra to |
102 |
> host them, I think we can make real progress. I don't think this is |
103 |
> something that needs to take a long time. The pieces are mostly there |
104 |
> - they just have to be assembled. |
105 |
|
106 |
Are you willing to champion that, then? :) |
107 |
|
108 |
-- |
109 |
Best regards, |
110 |
Michał Górny |