Open-MSS
scheduled test should create issues
our scheduled test should create an issue automatically when it founds a problem.
I think we need to have a scheduled test daily that tests the stable release as installed via mamba (i.e. a plain mamba install mss and then run the test suite). The scheduled test we have right now only runs for develop, but that does not ensure that the latest published version still works. The issue should then be created in the mss-feedstock repository.
Something similar to what is described here: https://www.aldertvaandering.com/blog/triggering-and-waiting-for-an-external-github-workflow/ to trigger the testing workflows and wait for their result, followed by opening an issue if necessary (can also be done in "github-script") should work. This is a bit annoying to test though and it isn't a blocker for v9, so I'll push it back to v10.
Doing it here in the MSS repository for stable and develop is one thing, it helps us detect breakage with the development dependencies we use here on both branches. Testing that the package as installable from conda-forge still works is another issue though, and is the responsibility of the conda-forge package distribution. So that should probably happen in the mss-feedstock or however else it should preferably be done over in conda-forge land.
With the new pixi development environment and CI overhaul a lot has changed regarding this issue. There are no longer "scheduled tests" in that sense, but rather a scheduled lock file update PR for which the usual PR test suite will run.
We will probably want to get notified when those PR checks fail for some reason, but I think we should first wait a bit and see how the workflow works out in practice at all.
I think the user opening the update PR (which would be open-mss-build) will automatically get notified about failed workflows. I don't think there is a native way to subscribe other users to those notifications, but it might e.g. be possible to add another workflow that adds a comment mentioning interested people on CI failure.
Anyway, I'll push this to v11 for now.