Question

I have a class the communicates with a device over TCP.
I send commands, and then issue stream.Read(reply, 0, 1) on the TcpClient’s GetStream().
I read out every single byte, and then combine the read bytes in a string, and compare the string to a termination string which lets me know when to stop reading.

There are some flaws in the device software, so that the termination string is not always sent. I would like to handle this in a proper way and using a period of time to determine whether the read should be stopped.

If I set the stream.ReadTimeout to something other than -1, then the socket will be closed when this timeout is exceeded, which I don’t want it to.

If I use the stream.BeginRead(..) with a stream.ReadTimeout = -1 then I will be sucking dry the thread pool after a while.

What intelligent solution can be constructed to address this issue?

Answer 1

I managed to implement something using stream.DataAvailable.

It goes like this:

do
{
    try
    {
        if (socket.Poll(socketTimeout, SelectMode.SelectRead))
        {
            socket.Receive(reply, 0, 1, SocketFlags.None);
        }
        else
        {
            throw new SocketException((int)SocketError.TimedOut);
        }
        replyString.Append(Encoding.ASCII.GetString(reply, 0, reply.Length));
        // Termination string logic...
    }
    catch (SocketException e)
    {
        // Check on e.SocketErrorCode if necessary.
        break;
    }

} while (!terminationReached);

where socket = _tcpClient.Client;

Edit:
socket.Receive also closes the socket on timeout, socket.Poll does not.
Changed to using socket.Receive(..) instead of stream.Read(..).