Evgeni Burovski

Does it look similar if you do `spin test` in numpy? (Am headscratching if it's a a generic `spin` issue which gets visible because of the scipy test suite size,...

Interesting. How does `$ python dev.py test` fare?

> so it might very well be a reasonable ask to bisect this. Indeed very much so. Especially given that we're in the rc1 phase. I tentatively added an 1.16.0...

To clarify my comment https://github.com/scipy/scipy/pull/21012#pullrequestreview-2135129312: 1. Scipy tests fix what the SciPy functionality is. 2. When we add new functionality, it needs new tests. 3. When we add a new...

I put some elbow grease to where my mouth was in https://github.com/scipy/scipy/issues/21031#issuecomment-2186180119, and here's a concrete example to evaluate: in https://github.com/scipy/scipy/pull/21091/ converts `ndimage` tests. A short summary: - There are...

> I don't know the background of these work for array API or why we are jumping on the gun before it matures (maybe quansight connection maybe something else whatever)...

> Given how conservative we are typically towards any addition backend or codebase, adopting a standard that is at its infancy, must have come from somewhere else. I just don't...

> Views are nice, but you do have the downside that you don't know whether the result is a view or not Which is exactly the issue with `round`: on...

> can you clarify what you mean here? I can interpret it as "the behavior in 1.2x was to always return a copy", or "this was changed for other functions...

Added a quick fix in https://github.com/data-apis/array-api-tests/pull/325. There's a TODO item in code comments to restore the range of values in a way that does not trigger the hypothesis health check.