1 |
On 11/25/10 4:07 PM, Alexis Ballier wrote: |
2 |
> By following closely upstream development you are able to adapt the live |
3 |
> ebuild to the current status, eg, for new deps. Usually with vlc, when a new |
4 |
> major version comes out, there are a couple of new features and a couple of |
5 |
> outdated features removed, doing all these changes at once is more error prone |
6 |
> than doing it once you notice the change for each of them in the live ebuild. |
7 |
> |
8 |
> As for version bumps, the live ebuild is simply copied to an ebuild with a |
9 |
> name matching the version that has just been released. |
10 |
|
11 |
Ah, I see. It makes sense. In the case of www-client/chromium, we have |
12 |
weekly "dev channel" releases, which go through Google's QA etc, and |
13 |
it's useful for me to work based on those and not the trunk for |
14 |
packaging. Everything else seems to work very similarly for me. |
15 |
|
16 |
> What kind of modification do you need to do? The only cases I had to change |
17 |
> anything is when the live ebuild was out of sync. |
18 |
|
19 |
I do adjustments for dev channel releases, and when I need to do them, |
20 |
the live ebuild usually is out of sync, so I forward-port the fixes to |
21 |
the live ebuild. |
22 |
|
23 |
> In case of a responsive upstream, like vlc or ffmpeg here, you can apply the |
24 |
> policy of "backports only" for patches and have absolutely zero patches for |
25 |
> the live version. |
26 |
|
27 |
Right, Chromium is actually very responsive and I'm also a Chromium |
28 |
committer, so upstreaming patches is usually easy. |
29 |
|
30 |
However, my patches related to using system libraries instead of bundled |
31 |
copies are often "controversial", so getting them through code review |
32 |
takes some time. |
33 |
|
34 |
Most of the time when I'm making local patches, they fix _regressions_ |
35 |
in the system library support, i.e. we were using a system library, but |
36 |
an upstream change broke it. |