The manually started pipelines over the Run pipeline button are currently
blocked. Let's enable them again so one can e.g test Dockerfile changes
without merging them first.
This was an oversight in [1] and causes the pipeline not to run for
release branches, but for any other branch. Fixed by inverting the
condition.
[1] ecac11569e
This prevents running a pipeline for any branch that is not a main or a
release branch, unless a merge request for it is created. It therefore
fixes the double pipeline issue where one pipeline would run for the
branch and one for the merge request itself.
The flatpak@aarch64 is not allowed on other branches than master but
is allowed on forks. It would be nice to allow manual run in those cases
to save resources, but to allow testing aarch64-specific fixes.
The pages job is allowed on forks currently. It is not desired to run it
there and it is going to fail there anyway. Let's explicitly disable this
job on forks.
The pages job is not running currently on master. This is because
`only: refs: $CI_DEFAULT_BRANCH` is not supported (e.g. see
https://gitlab.com/gitlab-org/gitlab/-/issues/29591). Let's fix that
using one of the suggestions from that issue.
The gnome-runtime-images have been recently migrated to Quay. This is already reflected in the template.
Please note this MR has been created semi-automatically. If it doesn't make sense, feel free to close it.
The fedora job was there before the flatpak job. But now we relies on
flatpak job and the fedora job doesn't really give much benefits. It is
rather waste of runners and maintainers time. Let's remove it.
The flatpak master job seems to be redundant currently. It uses the same
image as flatpak devel and their .yml files are equal. The two flatpak jobs
are relict from the times when GNOME SDK was unstable and broke CI often,
which is no more true. Let's remove the flatpak master job.
The style check job currently fails on stable branches since the recent
uncrustify change. We could use older image on stable, or backport the code
changes, but I think that the style check job on stable branches is not
actually needed. Let's disable the style check job on stable branches to
fix pipeline there and avoid this situation in the future.
There are various problems with the rawhide image currently and I don't
know how to workaround them. Let's use fedora:latest (temporarily) to fix
our pipeline.
It was set to flathub because we used to depend on a stable runtime. [1]
We have since changed to depend on the unstable runtime[2], but the
Nightly flatpakref still configures the wrong repo on
installation. If the user didn't have the gnome-nightly repository
configured in advance, then the runtime cannot be found, failing the
installation.[3]
Update the repo URL, which I should have done in [2].
[1] c4afd14696
[2] d95a616116
[3] https://blogs.gnome.org/antoniof/2021/03/05/files-40-beta-more-productivity-and-some-eye-candy/#comment-541
The pipeline currently fails with Fedora rawhide, because g-ir-scanner fails
with failures like: "ldd: error: you do not have read permission for
`/builds/GNOME/nautilus/_build/tmp-introspectgwhh729q/Nautilus-3.0'".
This obviously affects more projects:, e.g. GNOME/grilo!62. Let's use
Fedora latest for now as a workaround.
The triage job is broken, which regularly causes CI failures. I've
made some attemts to fix it but I failed. I don't have capacity to
spent more time on it. Let's remove the job completely for now to
prevent the confusing CI failures.
Fixes: https://gitlab.gnome.org/GNOME/nautilus/-/issues/1625
The GNOME runners are no more privileged and thus it is not possible to
use Docker from the pipeline. Let's use Buildah instead Docker to fix
the image generation.
The style check job currently fails as uncrustify from rawhide obviously
changed its behavior. It tries to change function pointer the following
way:
- gboolean (*each_function)(NautilusCanvasIcon *, gpointer),
+ gboolean ( *each_function )(NautilusCanvasIcon *, gpointer),
I don't think this is right, but don't know how to fix that ellegantly.
Note that this is because of "sp_before_ptr_star = false". Let's
hardcode the previous image varsion to avoid this change for now.
Currently, the nightly job is skipped if fedora rawhide, or flatpak
master job fails. This is not wanted as the nightly should be published
if flatpak devel job succeeds. Let's allow failures of those jobs to
not block the deploy stage.
The update image job is always shown (under the GNOME namespace),
however, fedora rawhide job is used only on master/web. So let's show
this job only on master/web as well.
The fedora job is run only on schedules and tags. The job is pretty
fast now. So let's run it always on master so we know about possible
breakages immediately and not after several days.
Fixes: https://gitlab.gnome.org/GNOME/nautilus/-/issues/1441
This reverts commit e7c2a0182a, because
the stage name doesn't have to be explicitely defined currently as it
was changed directly in the template:
5bffac6dfb
Fedora and style check jobs install dependencies on each run. This
causes that the pipelines take a long time. Let's use a custom docker
image to not block runners for such a long time and to get the pipeline
results faster. This commit also adds a new job to generate the docker
image for us.
The whole code was reformatted using uncrustify script, but MRs are not
checked using the script. Let's add style-check job to prevent wrongly
formated MRs to be merged.
The stage and job names can contain spaces and they are already used
in job names. Let's replace underscore in cross_environemnt stage to
make the pipeline output nicer.
The coverage is not computed since the commit a5a405d, but .gitlab-ci.yml
still contains some coverage related code which is not currently used. Let's
remove that code to make it more readable.