[libvirt-jenkins-ci PATCH 3/5] guests: Introduce the new 'gitlab' flavor

Andrea Bolognani abologna at redhat.com
Mon Apr 6 16:35:03 UTC 2020


On Mon, 2020-04-06 at 13:28 +0200, Erik Skultety wrote:
> On Fri, Apr 03, 2020 at 05:08:59PM +0200, Andrea Bolognani wrote:
> > Other than that, storing the token in ~/.config/lcitool is fine I
> > guess. The number of files in that directory is getting fairly
> > ridiculous, though, so Someone™ should really look into throwing
> > away the current, extremely crude configuration system that's been
> > inherited from the original shell implementation and replace it
> > with something more sensible like a config.toml.
> 
> Let me spend some time on that improvement :).

It was not my intention to volunteer you for this effort, but okay,
I'll take it :)

> > Anyway, it's simpler than that: all of our Linux jobs currently
> > expect to run in container images, and we can only have that using
> > the Docker executor.
> 
> With the functional tests in mind, the Docker executor doesn't really solve
> anything, for builds, there are already a bunch of containers taking care of
> that, now we need to cover cases where container technology cannot be utilized
> as well as environments for functional tests - of course, for that, these
> patches are only sort of a prequel.

Okay, so the target for the VMs built this way is

  * functional tests, for all of them;

  * unit tests, for the FreeBSD ones.

Is that correct? Because I was under the impression that, aside of
filling the FreeBSD-sized gap in our GitLab CI coverage, we were
aiming at increasing the capacity of the existing container-based
unit tests beyond what the shared runners provide. If that's not
what we're aiming for, then of course using the Docker executor
makes little sense and would probably only get in the way.

-- 
Andrea Bolognani / Red Hat / Virtualization




More information about the libvir-list mailing list