Jason Gross

Most of the tests (`python dev.py test -t scipy/special/tests/test_log_softmax.py`) now pass, but the final one is blocked on https://github.com/numpy/numpy/issues/25623, because `np.argmax` doesn't support tuple axes even though `np.max` does.

I've pushed a horrible kludge of a commit that special-cases `tuple` arguments to `axis` and uses the old implementation for those arguments. This makes the test-suite (`python dev.py test -t...

If you approve the CI run again, I expect (fingers crossed) all the tests will pass

I have added a kludge to work around the windows test issues, can someone approve the CI?

All checks except for lint passed, and I've force-pushed again to fix the lint issue. If someone would be so kind as to start the CI again, I believe this...

What's causing the benchmark failure? I'm not sure how to interpret the logs

Windows failure seems spurious as per https://github.com/scipy/scipy/issues/20806#issuecomment-2334335644, still confused about the bench failure.

For future reference, I debugged this with ```python try: await self.page.frame(name=login_frame).press("[placeholder=\"Password\"]", "Enter") await self.page.wait_for_url(re.compile(r"app/trade"), wait_until="domcontentloaded") # Making it more robust than specifying an exact url which may change. except TimeoutError...

I think the strategy I'm suggesting is independent of what's going on in the proof. (But, broadly, there's some pre-processing, then a big `repeat match` that handles all but one...

You are suggesting that it might be related to https://github.com/ocaml/setup-ocaml/issues/815#issuecomment-2204843777 in particular, where perhaps indirecting through `ocaml-env exec -- opam.exe exec` instead of jumping straight to `opam exec` might change...