Lars Grüter

I am hoping to find a solution that allows for parallelism in this example because the long build time without even using all resources offends me personally. :sweat_smile: So I...

Aha, I just noticed that https://github.com/scikit-image/scikit-image/pull/7140/commits/0079f69b2134c060b520a646d7db6e1cf3dbabf5 is actually passing fine everywhere except for Some digging with git blame shows that this has been an issue before: - https://github.com/scikit-image/scikit-image/pull/3873 - https://github.com/scikit-image/scikit-image/pull/3854#issuecomment-485239125...

I'm unsure how to properly debug this without convenient access to that arch. :/

Thank you so much for investigating and reproducing this! 🙏

See https://github.com/scikit-image/scikit-image/issues/3079 for earlier similar issues with this test on Python 3.4. 🤷

+1 from me. I think the fact that there is no feature parity between both implementations (regarding dtype) makes this a good idea.

Relying on the footprint to achieve the same effect might have worse performance, right? Because it required zero-padding the footprint until the center is in the desired place, while shifting...

Completely forgot about this and created https://github.com/scikit-image/scikit-image/pull/7966. But I still like the idea to let a decorator handle setting `__doctest_requires__`. Is that too hacky or a good idea?

> I think it's a bit odd to have version specifiers in there, when we already set minimums in the pyproject.toml, but that's a minor concern. The version specifiers weren't...

I've updated this. Feel welcome to review, merge or ignore. I'm happy either way. :)