Is it normal for copy_bidirectional to always return IO error code: 107, kind: NotConnected whenever one of the peers closes

[Banyc]

If that's the case, we can never get the Ok(u64, u64) from copy_bidirectional.

The setup goes like:

• echo server

    let (mut reader, mut writer) = socket.split();
    let mut buf = [0u8; 1024];
    loop {
        let n = reader.read(&mut buf).await?;
        if n == 0 {
            break;
        }
        trace!(?id, ?n, "TCP thread read");
        writer.write_all(&buf[..n]).await?;
        trace!(?id, ?n, "TCP thread wrote");
    }
  
    tokio::time::sleep(std::time::Duration::from_secs(10)).await;
  
    // tokio::io::copy(&mut reader, &mut writer).await?;
    writer.shutdown().await?;

• proxy

  tokio::io::copy_bidirectional(downstream, &mut upstream).await?;

• client

    let msg = b"hello";
    let mut buf = [0; 1024];
    {
        let mut buf = &mut buf[..msg.len()];
        stream.0.write_all(msg).await.unwrap();
        stream.0.read_exact(&mut buf).await.unwrap();
        assert_eq!(msg, buf);
    }
    stream.0.shutdown().await.unwrap();
  
    tokio::time::sleep(tokio::time::Duration::from_secs(10)).await;

Once client runs stream.0.shutdown().await.unwrap();, the copy_bidirectional pops the error immediately.

[Banyc]

Sorry, it is downstream.peer_addr()? or upstream.peer_addr()? that cause the error, not copy_bidirectional .