Losing logs

[maxgttph]

Some logs got lost.
How to reproduce:
-> Change tcp_server.rs handle_message to add some logs:

        module_handle_messages! {
            on_bus self.bus,

            Ok((tcp_stream, addr)) = tcp_listener.accept() => {
                tracing::error!("New connection from: {:?}", addr);
                println!("New connection from: {:?}", addr);
                let r = rand::random::<u64>();
                let sender: &tokio::sync::broadcast::Sender<TcpServerMessage> = self.bus.get();
                let sender = sender.clone();
                readers.spawn(async move {
                    let mut framed = Framed::new(tcp_stream, LengthDelimitedCodec::new());
                    loop {
                        match read_stream(&mut framed).await {
                            Ok(TcpServerNetMessage::NewTx(tx)) => {
                                tracing::error!("Received new transaction from: {:?}, r = {}", addr, r);
                                let tx_hash = tx.hash();
                                sender.send(TcpServerMessage::NewTx(tx))?;
                                // TODO: Wrap resp in a TcpServerMessageResponse enum ?
                                framed.get_mut().write_all(tx_hash.0.as_bytes()).await?;
                            },
                            Ok(TcpServerNetMessage::Ping) => {
                                framed.get_mut().write_all(b"Pong").await?;
                            },
                            Err(e) => { bail!("Error reading stream: {}", e); }
                        };
                    }
                });
            }
        }

-> Comment the tokio::time::sleep in loadtest lib file
-> In loadtest, change number_of_task to only keep 2 (for readability)
-> Run load test. You should now be able to see that multiple transactions have been received, while there is only one log of new connection. You should also see the println of the connection without its associated log

[Supa-mega]

Mind if I try this one?

[hhalex]

For information, most of this code has been extracted in a connection pool implementation, but may be the log problem is still there