Should we remove the doctest run from benchmarking? There was a reason we added it there but I can't remember it...

Should we remove the doctest run from benchmarking? There was a reason we added it there but I can't remember it...

It adds a second architecture to it, but I am not sure that was the reason. So yeah, I am seems OK to remove it (or rename for clarity).

FWIW, I don't expect much activity on scipy-doctest going forward. So far two recent releases were hotfixes to numpy 2.2 repr change and now recently discovered issues with dtype methods. That said, I did not anticipate either of these either :-).

Anyhow, the majority (or all of) gh-28019 needs docstring fixes the numpy side, there's not much scipy-doctest can do. IOW, if you pin it now, there's no bugfix scipy-doctest release to fix the current breakage in numpy.

I'd like a pin in the maintenance/2.2.x branch as I don't want to deal with lots of doc fixes during its lifetime.

I'm not worried about the lint failures, they are an artifact of recent updates and I expect they will get fixed in the near future.

Good point, for maintenance I think the hypothesis pin also makes sense definitely.

Yes, we still should clean out these failures ASAP anyway, but let's put it in for now. Thanks Matti!