1 |
On Sun, May 15, 2016 at 6:46 PM, Duncan <1i5t5.duncan@×××.net> wrote: |
2 |
> |
3 |
> Committing without testing, as long as you don't push, is fine, even |
4 |
> meritorious. It's the push that uploads those commits to the gentoo |
5 |
> reference repo, however, and testing should *definitely* be done before |
6 |
> pushing, with more commits /before/ the push to fix what the tests found |
7 |
> to be broken, should they be necessary. |
8 |
> |
9 |
|
10 |
Of course. In fact, this is often the type of workflow you'd tend to |
11 |
employ in a CI setup. I believe that pull requests submitted on |
12 |
github get automatically tinderboxed, though I have no idea whether |
13 |
that provides any benefits to something like an eclass (if the CI |
14 |
script just tests the ebuilds being modified it obviously would not). |
15 |
Maybe in a perfect world we'd actually have a CI testing package |
16 |
category with dummy packages that do nothing but run tests to cover |
17 |
this sort of thing. |
18 |
|
19 |
Even so, I would imagine that in most organizations CI is intended |
20 |
more as a sanity check than a substitute for testing your own work. |
21 |
Certainly where I work the expectation is that somebody would have at |
22 |
least compiled and run something before putting it into some kind of |
23 |
QA workflow, even with CI. |
24 |
|
25 |
-- |
26 |
Rich |