问题描述：

使用pickle序列化和反序列化进行socket传输时，发送方可以正常执行，接收方报错：_pickle.UnpicklingError: pickle data was truncated

参考：https://blog.csdn.net/qq_42816922/article/details/110678845

通信双方代码：

@发送方
	connection, address = sock.accept()
	message = connection.recv(1024).decode()
    if message == 'helper':
        Helper_pickle = pickle.dumps(Helper)
        helper = Helper_pickle
        connection.sendall(helper)
        connection.close()

@接收方
	sock = socket.socket()
    sock.connect((HOST, PORT)) 
    message = 'helper'
    sock.sendall(message.encode())
    data = sock.recv(1024)
    Helper = pickle.loads(data)

执行结果报错：_pickle.UnpicklingError: pickle data was truncated

原因分析

根据报错提示：数据被截断，反序列化出现错误。
实际上socket.recv的缓冲区是有限的，就算设置4096*4096也会导致长数据的截断。
接收方并没有接收到完整的数据，导致在进行反序列化时，pickle无法正确进行反序列化，进而出错。

解决方案

@接收方
	sock = socket.socket()
    sock.connect((HOST, PORT)) 
    message = 'helper'
    sock.sendall(message.encode())
	data = b""
     	while True:
            packet = client.recv(4096)
            # print("1")
            if packet is not None:
                data += packet
            if packet is None or len(packet) < 4096:
                break
        print("Download weights finished", len(data))
        Helper = pickle.loads(data)