why not from threading import RLock?

RLock does not have a locked() method.

https://docs.python.org/3/library/threading.html#rlock-objects

However, _PyRLock provides _count which I am using to provide that method to satisfy the rest of the code's usage of the lock.

Look closely at threading, RLock is a factory function that either returns an instance of _CRLock or _PyRLock. The former does not have a _count attribute though the latter does. I am choosing _PyRLock explicitly in order to implement locked().

nice! Sorry I missed that!

No problem. I had to go re-read the threading module source to jog my memory 😄

Unless I am missing something only tests depend on locked(). Could we instead use RLock in the code, and change the test to check the lock in a different manner? E.g. trying to acquire the lock with blocked=false and testing if we acquired the lock or not.

Or maybe more easily just derive the class from RLock and implement locked() method using blocked=False as you suggest.

Well, nevermind, as I mentioned above, RLock is a factory function (can't derive). Could instead have a locked(lock) -> bool function used in tests.

That seems reasonable to me, it's only one place in tests so we don't need anything too fancy.

Unless I am missing something only tests depend on locked(). Could we instead use RLock in the code, and change the test to check the lock in a different manner? E.g. trying to acquire the lock with blocked=false and testing if we acquired the lock or not.