Question

I have a TCP client communicating with a LabVIEW GUI.

My program calls connect() at the start and disconnect() at the end. It will call passCommand(x) to read or write data to the LabVIEW GUI. However, in some cases, I have multiple threads which may be calling passCommand() and somehow the return data will get mixed up.

For example, in the main thread I will ask for the voltage, which should be a number between 300 and 400. In a different thread I will ask for the temperature, which should be a number from 0-100. The voltage will be returned as 25, while the temperature will get 250.

Is this a known issue with TCP communication and threading? Is there a way to solve this such as implementing a queue or unique id or something?

import socket as _socket

# get python major version as integer
from sys import version as pythonVersion
pythonVersionMajor = int(pythonVersion[0])

_serverHost = 'localhost'
_serverPort = 50007
isConnected = 0
_sockobj = None
_error_string = "error:"


def connect():
    'opens a connection to LabVIEW Server'
    global _sockobj, isConnected
    _sockobj = _socket.socket(_socket.AF_INET, _socket.SOCK_STREAM)      # create socket
    _sockobj.connect((_serverHost, _serverPort))   # connect to LV
    isConnected = 1


def disconnect():
    'closes the connection to LabVIEW Server'
    global isConnected
    _sockobj.close()                             # close socket
    isConnected = 0


def passCommand(command):
    'passes a command to LabVIEW Server'

    ## We prepend the command length (8 char long) to the message and send it to LV
    # Compute message length and pad with 0 on the left if required
    commandSize=str(len(command)).rjust(8,'0')
    # Prepend msg size to msg
    completeCommand=commandSize+command
    # python 3 requires data to be encoded
    if (pythonVersionMajor >= 3):
        completeCommand = str.encode(completeCommand)
    # Send complete command
    _sockobj.send(completeCommand)
    data = _sockobj.recv(11565536)
    # python 3 requires data to be decoded
    if (pythonVersionMajor >= 3):
        data = bytes.decode(data)
    if data.rfind(_error_string) == 0:
        error = True
        data = data[len(_error_string):] # get data after "error:" string
    else:
        error = False
    execString = "lvdata = " + data
    exec(execString, globals())
    if error:
        raise _LabVIEWError(lvdata)
    else:
        return lvdata


class _Error(Exception):
    """Base class for exceptions in this module."""
    pass


class _LabVIEWError(_Error):
    """Exception raised for errors generated in LabVIEW.

    Attributes:
        code -- LabVIEW Error Code
        source -- location of the error
        message -- explanation of the error
    """

    def __init__(self, error):
        self.code = error[0]
        self.source = error[1]        
        self.message = error[2]

    def __str__(self):
        return "%s" % (self.message,)

Answer 1

This is an example of one of the most common problems with threading. You are accessing a resource from multiple threads and the resource is not considered thread-safe (if both threads are sending/receiving at the same time, it’s possible for a thread to get the wrong response, or even both responses).

Ideally you should be locking access to passCommand with a mutex so it can only be used with by one thread at a time, or opening one socket per thread, or doing all of your socket operations in a single thread.