Samuel Colvin

How about we default to polling if the following returns true? ```py from pathlib import Path import platform import re def auto_poll() -> bool: if 'microsoft-standard' not in platform.uname().release: return...

Sounds like the simplest solution might be just: ```py import platform def auto_poll() -> bool: return 'microsoft-standard' in platform.uname().release ``` WDYT? I favour this as it's much less complex than...

or `'microsoft-standard' in uname.release.lower() and uname.system.lower() == 'linux'`?

Agreed, `_default_force_pulling` should only check for WSL if `force_polling is None`, same as it currently only respected `WATCHFILES_FORCE_POLLING` when `force_polling is None`, so when using watchfiles directly, that should already...

This is (hopefully) fixed in #194, please can someone with windows and WSL (and ideally docker) run watchfiles from that branch and check it's working correctly?

Have it tried `validate_all`? https://pydantic-docs.helpmanual.io/usage/model_config/

You set the default so it generally doesn't make sense to validate it. It only really matters to vision types like urls which can't be instantiated trivially.

Confirmed, we've never replicated `min_anystr_length` and `max_anystr_length` to JSON Schema. I think with the alpha of v1.10 released and V2 dropping `min_anystr_length` and `max_anystr_length`, we probably won't fix this as...

Thanks for reporting, @PrettyWood has plans to improve generics in V2. Until then this is just a limitation of pydantic, better this than the silent memory leak.

This will be fixed in V2, not before.