Re: Separate CI stages

[Toon Claes]

On Thu, Dec 31 2020, Fermin wrote:

> Hello,

Hello Fermin, sorry I'm dropping in quite late. I never found time to
properly reply.

>     I'm working on how we can separate the stages, and seems like it's
> "working", I'm using the cache of the job to provide the test stages with the
> Emacs compiled executable.

Cool stuff. I like the idea of parallelizing.

But for the record, you don't need to add separate stages in
.gitlab-ci.yml for each job. You can use the `parallel:` keyword
instead. It will trigger that number of jobs, with $CI_NODE_INDEX and
$CI_NODE_TOTAL set. So the script can use those to select the correct
set of tests.

>     (I'm using the gitlab.com instance, and testing the built in Dockerfile
> registry and work as expected)

I think we can even use EMBA itself for this. Maybe we should fork
https://emba.gnu.org/emacs/emacs into
https://emba.gnu.org/emba/emacs (or so) and use that project push
branches to directly. This makes more direct for testing, easier for the
rest of use to follow-up on changes. And then we can also track issues
there. When things get more mature, we can ask someone to commit them to
the canonical repo on Savannah.

But if that does not work well, I don't mind using GitLab.com as
alternative. We can add a project to group of the Ansible playbook [1]
and give you all access?

Other question: Is it worth it to also set up Container Registry on EMBA?


--
Toon

[1] https://gitlab.com/emacs-ci/