Albert Steppi

> Are the CUDA tests added by this PR going to be included in the CI workflow? If included, it could be helpful for people without CUDA (like me) to...

> To be clear, we are simply putting things into `std` which are missing on other platforms. Of course no one should just put random custom things into `std`. Yeah,...

> I took a look today. I was able to run the tests which all pass but the tests for cupy are skipped unless run on a gpu device (as...

> Thanks @steppi. I think this is a good idea. And I see a GPU CI job materializing at some point (maybe first on `xsf`, or on this repo). >...

@rgommers, I tried using `conda` instead of `pip + virtual environment` to manage dependencies for the development build and didn't experience the issue you saw. This is with Ubuntu 22.04...

> Okay, feel free to ignore the issue then - I'll figure it out when I have more time. The Ubuntu 22.04 issue should be a bug in the glibc/multiarch...

We have one approval and @rgommers said we could feel free to ignore the issue he observed when using `conda` for now. If no objections, I plan to merge this...

Sorry; I missed the ping somehow. Though I sympathize with @AlexSzatmary's reservations about it, I prefer using `np.allclose` rather than unpacking the definition in multiple places. I think it's reasonable...

> The following warning seems to reveal that a `}` is missing due to misleading indentation: I see, it does look like there's a missing bracket in the final if...

> @steppi I can make a PR for the first warning if you want. It’s an easy fix, though the code will not look pretty as `import_array` and `import_umath` are...