Thanks for picking this up, I'm very grateful this is landing

I know that the py.path hook arguments deprecation didn't work (for implementers, it did work for callers). But are we sure that the py.path node constructor arguments deprecation didn't work? I don't remember seeing any complaints about that one but maybe I missed it.

I know that the py.path hook arguments deprecation didn't work (for implementers, it did work for callers). But are we sure that the py.path node constructor arguments deprecation didn't work? I don't remember seeing any complaints about that one but maybe I missed it.

Good question, I decided to revert both py.path deprecations to avoid confusion -- better to deprecate the py.path arguments (hooks + nodes) in one fell swoop I think.

I think it's better to keep the removal if it was properly deprecated, this way we don't hold up progress in this area until 9.1. But it's OK if you think otherwise.

This PR just undoes the removal, doesn't fix the silent deprecation right?

This PR just undoes the removal, doesn't fix the silent deprecation right?

No it doesn't, we should create a new issue for that.

From #12069 (comment):

the oversight was that warnings of certain hooks where not displayed as expected

@RonnyPfannschmidt do we know exactly for which hooks we were not showing the warnings?

I think it's better to keep the removal if it was properly deprecated, this way we don't hold up progress in this area until 9.1. But it's OK if you think otherwise.

I think it will be less confusing for users, also as I see it the benefit we want is to remove the LEGACY_PATH alias and related code, but having to still support it for hooks we cannot really remove it, so it won't help much removing one but not the other.