Hmm, oh ye. Pass argument "--no-build" with the pytest call. Does that make it to go successfully? If not, I think you can ignore that 👍

I basically made that as a sanity check so that the package gets version dynamically when built (changed versioneer to setuptools_scm and the latter suggests to ignore _version.py). I'm not sure if that's no longer as necessary as every master merge goes to test PyPI anyways.

Perhaps I could change it the other way round: CI passes argument "--check-build" to Pytest and only then this sanity check test is run.

pylint: error: Unrecognized option found: no-build - not for me, but I see you've done a PR to fix this so will take a look

with new code, still getting:

but if it's ok to ignore I will for now

Ye, you can ignore it 👍 I'm thinking of removing that test altogether, or at least remove the command-line argument. The reason for it was a sanity check the version is generated but I'm not sure if it's fruitful anymore test setuptools_scm itself works correctly.

The PR has issues with pytest. It seems you cannot pass arguments to pytest argparser in built packages without hacks. I possibly change the package layout a bit in the future by putting tests outside the library code but I'll inform well in advance if/when I'm doing that.