In #339 we have started to be more judicious about running tests in CI. Not all fortran PRs touch the coarse-graining code, so there is often not a need to check that those tests pass upon each push to every PR. One might ask why we can't just remove the coarse-graining tests from docker image since we already have some native regression tests for those features. The answer there is that the docker tests do two things that the nix tests currently do not:
They exercise the code in debug mode. This is useful for catching things like index errors and the like.
They test to make sure the code reproduces across different domain decompositions.
If we added nix tests in these settings, we could then remove these tests from the docker build and mark these tests with the coarse marker, further reducing the default time required in CI.
In #339 we have started to be more judicious about running tests in CI. Not all fortran PRs touch the coarse-graining code, so there is often not a need to check that those tests pass upon each push to every PR. One might ask why we can't just remove the coarse-graining tests from docker image since we already have some native regression tests for those features. The answer there is that the docker tests do two things that the nix tests currently do not:
If we added nix tests in these settings, we could then remove these tests from the docker build and mark these tests with the
coarsemarker, further reducing the default time required in CI.