Pavel Kurdikov

I think ignoring is a mistake (IMHO). There is one case when event without paths is expected - when `inotify` queue is overflowed. User should be able to handle this,...

Oh, I was wrong, sorry about that. The checking of `need_rescan` is before the ignoring an event in the PR. But, after some investigation, I think it might happen because...

OK, it was races [This test](https://github.com/notify-rs/notify/pull/700/files#diff-9663de7aadbfa93d9f87e8c0dc5c94855e59fff314c091caa5478b666e404983R709) did not pass before the fix

> it leaves the watcher in an inconsistent state because the paths have been changed The main problem is not having paths changed but is having watcher stopped if we...

> I would like to keep the API lean I agree. User can do it by themself if they want. Ok, I'll delete it

> Also, if the PathsMut compatibility API is the same as the current one, then we can release it as a feature update (v. 8.3.0). It's possible, i'll try to...

Ok, in addition to the renaming and saving API compatibility I changed `PathOp` argument to us own struct - `WatchPathConfig`. It'll help us to keep compatibility when new `watch`-related features...

@branchseer my main idea wasn't to remake it just because I want to, but to make the code more ready to #632 . I want this feature, because it allows...

@dfaust Hi! Sorry to bother you — would you mind taking another look at the PR when you get a chance? I'd really appreciate your thoughts.

Oh, I've just noticed the `Watcher::kind`, there couldn't be any `Watcher` implementation outside the crate, then it's not breaking until previous version of PathsMut is released Of course an outside...