Matt Haberland

One more change I'll make is to make `complex_func=False` the default. Just as we shouldn't change the default function to the complex exponential, this PR probably shouldn't change the `dtype`...

Looks like the doc build failures were related. The last PR to go in before this (gh-18301) was green, and the first PR to post after this merged (gh-18470) is...

If the ksone error still fails for the same reason, it looks like a small tolerance error. Is the resolution to bump the tolerance so that the test passes on...

It does look like there is a typo. Page 2 is pretty clear, but just to confirm: ```python3 import numpy as np from scipy import stats import matplotlib.pyplot as plt...

I don't see the problem, but it probably isn't a coincidence that the doc build stops with `resample`. To be sure, you could merge the latest `main` into this PR....

@HaoZeke are you able to try this problem out with pure HiGHS code (no SciPy wrapper) and (assuming the problem persists) submit an upstream bug report?

Thanks @Matryoskas! Can you also try gh-20336?

No problem. You don't have to even try to access HiGHS' calculation of the residual. Just manually calculating the residual to show that HiGHS is not respecting the constraint would...

Thanks @xuefeng-xu! Since this is modifying the formulation for all values (not just in cases of overflow), I'd like a `special` regular to take a look. Did you benchmark the...

Thanks for your patience @xuefeng-xu. @steppi is a `scipy.special` maintainer, and he will take a look shortly.