1 |
Jeroen Roovers wrote: |
2 |
> For once someone suggests a single good case where git beats CVS for |
3 |
> portage tree changes: easily checking suggested changes ... |
4 |
|
5 |
Did you look at Gerrit one of the many times I mentioned it already? |
6 |
|
7 |
That is what it is for, and it is pretty great. |
8 |
|
9 |
|
10 |
> A shiny new workflow doesn't magically make Gentoo development easier. |
11 |
|
12 |
Actually it does. It doesn't make hard things easy, but it does make |
13 |
things which are currently hard but which *should* actually be easy |
14 |
indeed be easy. Such as contributing commits for tested bumps, and |
15 |
tested bugfixes. |
16 |
|
17 |
Even if you might not think so, there is quite significant overhead |
18 |
for patches which sit in bugzilla compared to a Gerrit in front of a |
19 |
git repo. You really need to try it, and really want to learn Gerrit, |
20 |
to discover just how well it works. |
21 |
|
22 |
End result: |
23 |
|
24 |
Contributors can work independently and then send their work in with |
25 |
*zero* overhead. Do not underestimate the importance of this. |
26 |
|
27 |
Developers with portage tree write access review commits and can |
28 |
apply them to portage, or discard them, with a single click or with |
29 |
a single SSH command. |
30 |
|
31 |
|
32 |
> The hard bits aren't usually related to interactive repo access methods. |
33 |
|
34 |
It was argued that the hard bits are only learned by mentoring. I |
35 |
think this makes sense, and I think it has nothing whatsoever to do |
36 |
with commit workflow. There are PLENTY of TRIVIAL NO-BRAINER |
37 |
contributions which are NOT GETTING INTO PORTAGE. Let's not worry |
38 |
about optimizing the really difficult problems until all the really |
39 |
simple problems have all been solved. |
40 |
|
41 |
|
42 |
> We can't get rid of the quizes just because we all now use $shiny. |
43 |
|
44 |
Actually we can. |
45 |
|
46 |
|
47 |
> Proxy commits should work exactly like recruit/mentor commits: you |
48 |
> review patches and give the nod and then the commits get done. This |
49 |
> workflow is only slightly less convenient with CVS than with git. |
50 |
|
51 |
I really recommend spending some time on learning Gerrit if you |
52 |
haven't used it already. |
53 |
|
54 |
|
55 |
//Peter |