1 |
On Wed, Jan 28, 2015 at 12:56 PM, Ciaran McCreesh |
2 |
<ciaran.mccreesh@××××××××××.com> wrote: |
3 |
> On Wed, 28 Jan 2015 17:22:21 +0100 |
4 |
> Alexey Lapitsky <lex.public@×××××.com> wrote: |
5 |
>> Here's a naive example how it might work: |
6 |
>> |
7 |
>> 1. A user creates a PR |
8 |
>> 2. travis-ci automatically runs QA tests: repoman full on the PR, |
9 |
>> reports the result back |
10 |
>> 3. Github hook is emitted, a "system" does some processing: for |
11 |
>> example, create a bug in bugzilla, notify the maintainer, comment on |
12 |
>> the PR |
13 |
>> 4. The relevant maintainer checks the PR, comments: "looks fine, |
14 |
>> merging" 5. Github hook is emitted to "the system", which checks the |
15 |
>> PR comments, does some sanity checks, merges commit back to cvs, |
16 |
>> rejects the PR |
17 |
>> |
18 |
>> "The system" might be a simple script which has opt-in membership for |
19 |
>> the subsystems / maintainers who wants to use it. |
20 |
>> |
21 |
>> Does it sound sane? |
22 |
> |
23 |
> Exherbo has been doing something along those lines for ages, and not |
24 |
> just for repoman: there's also full build testing (from a clean system, |
25 |
> to check dependencies are correct). It's not very hard to do. |
26 |
> |
27 |
|
28 |
Including having two repositories both being committed to, with a |
29 |
delayed propagation of commits from each to the other? Somehow I |
30 |
doubt that is what you're doing with exherbo. |
31 |
|
32 |
The CI isn't the part I'm picking on here. The idea of still working |
33 |
in cvs while having a separate CI and commit path in git that applies |
34 |
back to cvs is the part that sounds rather prone to breakage. That |
35 |
seems like a massive hack. The real solution is to get the heck off |
36 |
of cvs. |
37 |
|
38 |
-- |
39 |
Rich |