When we released 6.3 there were some failing tests reported by api-test-runner for the updated "alpha" image: https://github.com/gdcc/api-test-runner/actions/runs/9775864881
I didn't worry too much about it because test were passing in Jenkins.
That said, it would be great to have the "alpha" workflow show green again, like it did for previous releases.
Looks like timing related things to me. I've said it before, let me say it again: the way we setup and run our E2E tests is flawed to be suffering from timing issues.
Should we create an issue?
Dunno. This is not a simple thing. I suspect it will require restructuring these tests and that is a lot of work. I sincerely doubt anyway wants to put these hours in.
But we all want passing tests, right? :grinning: Especially at release time.
Right now we call this step of checking the "alpha" image optional: https://guides.dataverse.org/en/6.3/developers/making-releases.html#optional-test-docker-images
Well what we should take a look at, if we really want to change things: don't create everything over and over again for tests and just assume it will be done in time. We put a lot of stress on the test runner and waste CPU cycles to re-create things like collections, users and datasets. And then we often just wait for a few seconds, hope all is good and light the fuse on the test canon...
Sounds like good advice.
Right now it says this:
"If there are failures, additional dependencies or settings may have been added to the βdevelopβ workflow. Copy them over and try again."
But what if you try again and again and it keeps failing? That's where I was during release. Probably an issue should be created, right?
Last updated: Oct 30 2025 at 05:14 UTC