Not sure why github is linking to the wrong CircleCI build, the real one is https://app.circleci.com/pipelines/github/numpy/numpy/3679/workflows/bbd2257f-2d41-4cb9-882a-48511446daa7/jobs/15768

@larsoner a ping about the build artifact link. It seems it is sometimes broken (such as here), although it is possible that this only happens when circle CI gives an error.

@seberg: In this case even the first-party CircleCI link is broken and points to the wrong build (of a commit that does not exist)

Oh, OK, then maybe that type of thing caused it to fail somewhere. It was just a single PR where I had noticed a mismatch, thought it was repeating now.

There can be problems when the fork and upstream both have CircleCI enabled:

scientific-python/circleci-artifacts-redirector-action#7

But indeed these builds are just failing themselves

The improved issubdtype docstring with examples looks much better, thanks @eric-wieser

This needs the fix in gh-17304 to pin pygments version

Will rebase on master if this needs any other changes - otherwise I'd be inclined to merge as is rather than starting another CI run.

The artifact for the "See also" link seems correct and the rest also seems fine and is definitely an improvement, so I will just put this in (there are also no warnings beyond the gh-17304 one). Thanks.