We seem to be using dtype=np.float or (assert array.dtype == np.float), should be easy to fix.

We seem to be using dtype=np.float or (assert array.dtype == np.float), should be easy to fix.

OK, trying that out. Can you approve the CI workflow again please?

oh, it seems to be rasterio that is doing this too

@martindurant CI workflow needs approval again.

There is still one failure there, but I can't immediately tell what the issue is.

It looks like test_http_server_files is picking up the changes to glob in the latest fsspec release. I think I've fixed this now.

Needs another workflow approval 😁

With Python 3.8 we are using older numpy (3.8 support dropped recently) but new rasterio. So we are essentially debugging rasterio-numpy combinations.

Hm, the failure in py38 would be because the newest numpy no longer builds on that version? We could change the supported versions to 3.9-3.11, seems reasonable.

Sorry, your comment only just showed up. Let's test 3.9-3.11, but we don't need to change the python_requires as yet (but we could).

I have pinned rasterio in the python 3.8 environment file and the tests pass locally so let's try that. Can you approve the workflow yet again?

@ianthomas23 , is it too late to make you a member of Intake :)

It is too late now, the tests have all passed!