The good news is the release is fine. Just download the artifacts manually, make sure there are no duplicates, then upload them with twine. It doesn't matter if you take the macOS 11 or 13 downloads, just don't take both. (14 will be ARM, 11-13 was Intel). Also don't take the wheel output from build-and-inspect-python-package, only the SDist.

I think we need a twine check on the wheels as well, to avoid this issue happening again. Thanks for the quick fix :)

Simply trying to unzip the file would catch it, actually, since actions/download-artifact#298 garbles the bits together instead of erroring out, making invalid zips. :) (yes, twine check should be doing this. In my experience it is really limited in what it can catch, but I'd think this is one thing it could do!)

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 83.80%. Comparing base (273ed90) to head (08f5d94).

@@           Coverage Diff            @@
##           develop     #471   +/-   ##
========================================
  Coverage    83.80%   83.80%           
========================================
  Files           25       25           
  Lines         8164     8164           
  Branches      1705     1705           
========================================
  Hits          6842     6842           
  Misses        1322     1322           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Yey, issue resolved, just had to delete one of the artifacts and re-run the CI. Still would be good to try and catch these issues in this PR. Could hynek/build-and-inspect-python-package use the artifacts from cibuildwheel?

I've added a simple twine check sanity check. @henryiii do you have anything more here or 🟢 to merge?