Andrew Nelson

@j-bowhay I think what @ev-br is saying: ``` import numpy as np from scipy.misc import derivative from scipy.optimize._numdiff import approx_derivative sin = lambda x: np.sin(x) x = np.linspace(0, 1, 10)...

As soon as there are numpy wheels available for musllinux then we can start build musllinux wheel for scipy. i.e. numpy wheels are the blocker at the moment.

We would also need the OpenBLAS toolchain to be updated for musllinux

I've been playing around with building scipy in alpine today, in a docker image. I've created a CI run for musllinux on my fork. I'll ask scipy-dev if it's worthwhile...

I wonder if it's worth running `black` on notebooks that we add to make sure the style is nice from the start. We can ensure that it only runs on...

Black will work on jupyter, but not MD

Is that like Richardson extrapolation?

There was a brief discussion (if you could call it that) on Slack. There was a suggestion to look into cirrus because it offers native `macosx_arm64` and `linux_aarch64`. On GH...

My previous comment was solely directed towards wheel building. It would also be possible to do regular native CI testing on linux_aarch64 and macosx_arm64 (possibly other archs), which we're not...

Thanks @rgommers. I retriggered the wheel build run and all the linux_aarch64 builds worked (including test suite) on cirrus-ci, https://cirrus-ci.com/build/4966426158039040. The only job remaining for this PR is for you...