I think this crate leaks threads + fds when you drop futures

[bjackman]

I guess this is not actively maintained but just dropping a note in case it's of use to anyone...

I haven't actually tested this so if this is a concern for you it's worth verifying for yourself. But, I think this create will leak threads and file descriptors if you drop the futures returned by lock_exclusive etc (unless the file successfully get locked, or something causes the flock operation to fail).

Under the hood it works by using tokio::spawn_blocking and then just using a normal flock call in the thread. But as noted in the Tokio docs, if you use that API you need to think about cancellation.

I guess the way to cancel a flock call is to close the file descriptor. But this crate doesn't seem to do that - the fd is in a std::fs::File which gets moved into the thread doing the flock call. I think in a correct implementation the blocking task thread should get at most a RawFd which can be closed from the other thread if the future gets dropped.

Depending on usecase, I think this leakage is very likely to be harmless. But if you expect to often contend for files that are locked for a long time it could be more of a real issue.