Anirudh Dagar

Hey @RPGillespie6 thanks for the report! @rkern, if possible, I'd like to take this issue and send a workaround fix. Thanks for the detailed repro discussion above.

Thanks all for the feedback, completely understand why it makes sense to have pooch as an optional dependency. Does this mean that `scipy` should pass all the tests without pooch...

> Anyway, what we did was to skip any test where pooch failed for whatever reason. This sounds like a good idea to me if we decide to skip only...

Thanks Ralf, yes I was able to reproduce this at a small isolated demo scale for pooch [here](https://github.com/AnirudhDagar/random-tests/runs/7219790510?check_suite_focus=true). It raises on the pooch import which as you suggested can be...

For the lint errors: ``` scipy/datasets/__init__.py:33:1: E402 module level import not at top of file scipy/datasets/_fetchers.py:85:80: E501 line too long (91 > 79 characters) scipy/datasets/_registry.py:9:80: E501 line too long (85...

> Eventually, it would be useful to have prepared a set of instructions for manually loading the cache of datasets, for our friends who work in remote or secure areas...

@KuanHaoHuang could you please rebase this PR to accommodate latest changes from master? That would also trigger the github actions for the same. Thanks!

Hi @KuanHaoHuang, it looks like the merge was not made correctly. Request you to please fix that.

Hi @KuanHaoHuang thanks for the update, yes that's not your fault. I've made another merge from master to include the recent changes. CI should be happy now!

Yes, it seems to be a issue with the cache not being cleared out. I'll fix that and rerun the PR.