[Tarantool-patches] [PATCH v1] Set full testing for all branches

[Alexander V. Tikhonov]

Hi Alexander, thanks for the review, please check comments below.

On Mon, Jun 01, 2020 at 03:39:50PM +0300, Alexander Turenko wrote:
> On Wed, May 20, 2020 at 09:35:31PM +0300, Alexander V. Tikhonov wrote:
> > Set full testing with deploy builds and tests for all branches.
> > ---
> > 
> > Github: https://github.com/tarantool/tarantool/tree/avtikhon/all-tests-in-branches
> > 
> >  .gitlab-ci.yml | 13 ++-----------
> >  1 file changed, 2 insertions(+), 11 deletions(-)
> > 
> > diff --git a/.gitlab-ci.yml b/.gitlab-ci.yml
> > index 256b368c4..55b006ed4 100644
> > --- a/.gitlab-ci.yml
> > +++ b/.gitlab-ci.yml
> > @@ -8,11 +8,6 @@ variables:
> >  
> >  # Jobs templates
> >  
> > -.release_only_template: &release_only_definition
> > -  only:
> > -    - master
> > -    - /^.*-full-ci$/
> > -
> >  .deploy_only_template: &deploy_only_definition
> >    only:
> >      - master
> >    except:
> >      - schedules
> >      - external_pull_requests
> >      - merge_requests
> >
> >  .pack_only_template: &pack_only_definition
> >    only:
> >      - schedules
> >      - external_pull_requests
> >      - merge_requests
> > -    - /^.*-full-ci$/
> > +    - branches
> > +    - tags
> 
> First, there is not reason to list 'schedules', 'external_pull_requests'
> and so if you want to just run it everywhere.
> 

Actually I just wanted to make it more visible for the users, but anyway
if it not really needed I'll change it to "except: master".

> Second, we should not run both 'just packaging' and 'packaging and
> deployment' jobs on master. I would make ''pack_only_template' be just
> 'except: master'. Yep, it is not perfect: 'schedules',
> 'external_pull_requests' will don't run neither 'pack', nor 'deploy'.
> But this change is simple and downsides are acceptable.
> 

Right, it is the same as previous my comment.

> By the way, I don't know anything about pipeline for tagged commits in
> GitLab. What jobs will run on a tagged commit? Are we affected by
> [#3745][4] in context of GitLab CI deployments? Just curious, not part
> of this task.
> 

Right, tagged commit as I understood is the same as have tagged version
in the git, but sure it may be some other meaning, due to we never saw
it running.

> Just side note: I still believe that things would be simpler if we would
> deploy within a packaging job: this way we need only condition 'when not
> to deploy' and don't need its negation. (I proposed this in [1].) Now I
> see that it would be implemented not just like as it is done in Travis
> CI, because GitLab CI is weird about separate build and deploy stages
> (see [2], [3]), but would be implemented inside one jobs. Anyway, I
> don't insist, especially within this task.
>

Current patch is the simple change, while I already prepared another one
with overall changes in gitlab-ci configuration, you can check it in the
new branch avtikhon/branches-gitlab-ci .

> [1]: https://lists.tarantool.org/pipermail/tarantool-patches/2020-February/013893.html
> [2]: https://gitlab.com/gitlab-org/gitlab-runner/-/issues/336
> [3]: https://gitlab.com/gitlab-org/gitlab/-/issues/29265
> [4]: https://github.com/tarantool/tarantool/issues/3745
> 
> WBR, Alexander Turenko.