This will need a release note at some point.

@charris could you guide me towards this? Where should I add stuff? (note I already added flame to the documentation).

Where should I add stuff?

If you could add a note on both this PR and gh-13132 to doc/release/1.17.0-notes.rst, that would be very helpful. Under "new features"

@rgommers could you see if 3b5b21c is fine?

@rgommers I can't see why the Windows Python36 fails, but Windows Python37 does not... It seems I don't have permission to re-run the test?

Hmm, that looks unrelated but since it's in linalg I'd like to rerun the tests. Unfortunately I can't trigger a rerun either. @tylerjereddy could you trigger the rerun on Azure, and perhaps also add me ([email protected]) to the account?

I've triggered the re-run.

Failure is still there for 32-bit Python 3.6 Windows, so maybe real. Various ways to debug--normally I'd set up Azure on a fork and play with the CI until I identify the specific test that crashes--maybe using a more verbose output.

@rgommers ok, you're an admin for NumPy Azure account now. Stephan Hoyer, Matti, and Chuck should also be admins.

@tylerjereddy @rgommers I am still very puzzled by the apparent show of linalg error in one build.
I haven't changed any default build-options, only enabled more.

Is there anything I can do?

Hmm the build logs for the failures have disappeared. I've triggered a re-run

All green, in it goes. Thanks @zerothi!