Surprising behavior from mpsc channels when only using try_ methods

[nwoeanhinnogaehr]

Once a bounded channel has been filled and cleared using only try_send and try_next, it remains in a broken state. From here calling try_send always reports that the queue is full, where as try_next always reports that it is empty. This behavior is not found with std::sync::mpsc.

See the following example (0.2.0-alpha):

use futures::channel::mpsc;
fn main() {
    let (mut tx, mut rx) = mpsc::channel(1);
    println!("{:?}", tx.try_send("hello")); // Ok(())
    println!("{:?}", tx.try_send("hello")); // Ok(())
    println!("{:?}", rx.try_next()); // Ok(Some("hello"))
    println!("{:?}", rx.try_next()); // Ok(Some("hello"))
    println!("{:?}", tx.try_send("hello")); // Err(TrySendError { kind: Full })
    println!("{:?}", rx.try_next()); // Err(TryRecvError)
    println!("{:?}", tx.try_send("hello")); // Err(TrySendError { kind: Full })
    println!("{:?}", rx.try_next()); // Err(TryRecvError)
}