I was wondering if raising GitlabParsingError is correct for a "valid" response, IIRC we added it for malformed responses and not for real GitLab responses (although here I would expect an empty dict or 404 TBH). Maybe a more specific error or an upstream fix to not return None would be clearer to users.

I doubt it is the correct but it is what we are currently doing. My goal was to fix the test at this stage, not to fix the underlying issue which I think will be harder to fix...

Here is an old issue that I forgot about completely: #1192

And the upstream issue: https://gitlab.com/gitlab-org/gitlab/-/issues/8829

Good info!

Makes sense, I just wanted to track that somewhere which was the purpose of the xfail. I've opened an issue for that :)