The kit action was triggered in my own repo (where it was a push trigger, rather than a pull_request); This is the run producing a py3-none-any wheel.

Thanks!

But in your action I see:

> List non-binary artifacts
-rw-r--r--  1 runner docker 200525 Jan 10 01:49 coverage-7.6.11a0.dev1-py3-none-any.whl
-rw-r--r--  1 runner docker 804612 Jan 10 01:49 coverage-7.6.11a0.dev1.tar.gz

and in my last run of the action I see:

> List sdist
-rw-r--r--  1 runner docker 235413 Dec 26 16:45 coverage-7.6.10-cp39-cp39-linux_x86_64.whl
-rw-r--r--  1 runner docker 803868 Dec 26 16:45 coverage-7.6.10.tar.gz

Was my .whl there useless? Is it ok that it has a different name?

The old sdist build rule ran python -m build, which will produce both a wheel (native to the local platform) and an sdist by default - that's what produced the cp39-cp39-linux_x86_64 wheel. However, the action then excluded all but the .tar.gz file from the uploaded artefacts.

This new config forces the "native" wheel to be py3-none-any, and includes it in the uploaded artefacts.

The old configuration could have been optimised slightly by invoking python -m build -s - that would have only produced the sdist. However, it's a moot point now, as we want the wheel as well.

Sorry for all the questions. You say "The old sdist build rule ran python -m build", but that is still the command being run. The new config seems to be the environment variable disabling the compilation of the extension. That's enough to change the name of the result?

Invoking python -m build produces a wheel and an sdist. The command is running on Linux; so in the old configuration, a binary wheel is produced, because the extension module can be compiled.

In the new configuration, the compilation of the binary extension is disabled with the environment variable, but a wheel is still being produced; so the wheel identifies as py3-none-any.

This is now released as part of coverage 7.6.11.