Memory leakage and high cpu usage for each peer about webrtc

[rainliu]

Thanks for the sharing.

I will make a new release soon.

[rainliu]

webrtc v0.10.0 is released.

[LIMPIX31]

Okay. In the second case I forgot to close the connections, but in the first case there's still something wrong.

peer_connection.close().await hangs

  peer_connection.on_peer_connection_state_change({
    let peer_connection = peer_connection.clone();

    Box::new(move |s| {
      info!("Peer Connection State has changed: {s}");
      let peer_connection = peer_connection.clone();
      Box::pin(async move {
        use RTCPeerConnectionState::*;

        if let Disconnected | Failed = s {
          info!("Attempting to close ended connection");
          if let Err(err) = peer_connection.close().await {
            error!(%err);
          } else {
            info!("Connection closed");
          }
        }
      })
    })
  });

[samguns]

Interestingly, we have an app seemingly the same as you've done. It employs HTTP served by Axum to do the signaling. Everything works fine but a similar scenario especially the high CPU usage was observed in our app as well, even if we made sure the peer connection had been closed explicitly. Did you get any progress further concerning this issue? @LIMPIX31

[LIMPIX31]

Interestingly, we have an app seemingly the same as you've done. It employs HTTP served by Axum to do the signaling. Everything works fine but a similar scenario especially the high CPU usage was observed in our app as well, even if we made sure the peer connection had been closed explicitly. Did you get any progress further concerning this issue? @LIMPIX31

Thanks for your reply. Unfortunately I was not able to find out more. I was in a hurry, so I had to write a broadcast server in golang as a temporary solution.

[samguns]

Some follow-ups to share with you @LIMPIX31 and to all users especially devs of this project.
After browsering through PRs and we found this recently merged #514 seemed highly related to what we experienced. So we changed crate source to pull latest code from master branch instead of v0.8.0 in crate.io. The issue disappeared. Maybe you could try it as well and let's wait for a formal new release in crate.io.