Drain channel immediately after last receiver is dropped

[akhilles]

After the last receiver is dropped, all the items in the channel are effectively abandoned and can never be accessed. In this scenario, I think it's appropriate to drain the channel and drop all the items. One use case where this is relevant is a dispatch mechanism that uses oneshot channels:

let (req_s, req_r) = bounded(10);

// dispatch
for i in 0.. {
    let (s, r) = oneshot();
    req_s.send((i, s)).await.unwrap();
    let res = r.recv().await.unwrap();
}

// worker
while let Ok((i, s)) = req_r.recv().await {
    s.send(do_work(i));
}

Here, if worker terminates while there are still requests in the channel, dispatch can end up waiting for the response forever. This can be addressed with timeouts, but I think automatically draining the channel is a better solution.

[taiki-e]

This should be possible to do by applying a patch similar to crossbeam-rs/crossbeam@06d77d3 to concurrent-queue.

[NingLin-P]

Hi @taiki-e, is there any plan to resolve this issue? we have also encountered the abovementioned issue (with a similar usage).