black

black copied to clipboard

Description

This is the last part of #3017. This is effectively unchanged from the original PR, except for the addition of a commit to workaround a Windows shell quoting crash that I can't be bothered to deal with properly.

I don't really have solid test workflow runs to prove this will 100% work, but I've tested similar changes in this PR successfully (don't mind the Windows failure, I messed up). The only differences from the current version of this PR from that is that 1) I added an additional step to upload the wheels as workflow artifacts too (for debugging), 2) tried to ignore the twine command which was going to fail due to missing credentials (the windows job failing shows I didn't get my shell syntax quite right) and 3) mypyc 0.972 was used instead. I've reverted all of these changes, in the particular the mypyc upgrade since I don't feel comfortable updating it so close to the release.

This should work though, I haven't changed the wheel workflow since release 22.3.0 IIRC.

I'll update the release documentation in a separate PR with the changes to both the release CD and release flow.

Checklist - did you ...

• [x] Add a CHANGELOG entry if necessary? -> n/a
• [x] Add / update tests if necessary? -> n/a
• [x] Add new / update outdated documentation? -> n/a

diff-shades reports zero changes comparing this PR (8a21bac441842da63f14a2be8466422d838102e1) to main (507234c47d39f5b1d8289cdd49994e03dd97bcb4).

---

What is this? | Workflow run | diff-shades documentation

For the record, I've verified the wheels built by this new setup behave identically to the pure wheel (via diff-shades) on Ubuntu 20.04.04 for all Python versions we support. @zsol demonstrated that the arm64 wheels don't crash at least which is good. Windows is still relatively untested but I'm not particularly concerned. If I have the time, I'll do a diff-shades run on Windows too.