Gitlab CI pipeline requires improvement
This issue is forked from the discussion running in !63 (merged).
If we want to run tests extensively to trust our integration pipeline, we need to progress on accelerating the verification pipeline. If we all agree on gitlab CI being used for verification of the development branches and not publishing, as of now we may:
-
Configure the pipeline to retry N times until successful.(increased waste of resources) -
Avoid bundling artifacts until the large file upload issue is sorted out. -
Use the caching mechanism to retain a(cache not created for failed builds, and all builds fail now)ccache
database between builds. -
Divide the KStars build into stages producing smaller shared libraries to reduce artifact footprint. -
Run the build-appimage
job onmaster
only. -
Run the build-and-test
job on merge requests only. -
Use ninja
to build instead ofmake
. -
Add a Coverity analysis job.
Other ideas are welcome.
Edited by Eric Dejouhanet