C++ Linux Google Protobuf + boost::asio Cannot Parse

Question

I am trying to send a Google Protobuf message over a boost::asio socket via TCP. I recognize that TCP is a streaming protocol and thus I am performing length-prefixing on the messages before they go through the socket. I have the code working, but it only appears to work some of the time, even though I'm repeating the same calls and not changing the environment. On occasion I will receive the following error:

[libprotobuf ERROR google/protobuf/message_lite.cc:123] Can't parse message of type "xxx" because it is missing required fields: Name, ApplicationType, MessageType

The reason is easy to understand, but I cannot single out why this only occurs sometimes and parses just fine the majority of the time. It is very easy to duplicate the error by just having a single client talking to the server and simply restarting the processes.

Below are the socket code snippets.

const int TCP_HEADER_SIZE = 8;

Sender:

bool Write(const google::protobuf::MessageLite& proto) {
    char header[TCP_HEADER_SIZE];
    int size = proto.ByteSize();
    char data[TCP_HEADER_SIZE + size];
    sprintf(data, "%i", size);
    proto.SerializeToArray(data+TCP_HEADER_SIZE, size);
    boost::asio::async_write(Socket, 
                             boost::asio::buffer(data, TCP_HEADER_SIZE + size),
                             boost::bind(&TCPSender::WriteHandler, 
                                         this, _1, _2));
}

Receiver:

std::array<char, TCP_HEADER_SIZE> Header;
std::array<char, 8192> Bytes;

void ReadHandler(const boost::system::error_code &ec, 
                 std::size_t bytes_transferred) {
    if(!ec) {
        int msgsize = atoi(Header.data());
        if(msgsize > 0) {
            boost::asio::read(Socket, boost::asio::buffer(Bytes,static_cast<std::size_t>(msgsize)));
            ReadFunc(Bytes.data(), msgsize);
        }
        boost::asio::async_read(Socket, boost::asio::buffer(Header, TCP_HEADER_SIZE),
                                boost::bind(&TCPReceiver::ReadHandler, this, _1, _2));
    }
    else {
        std::cerr << "Server::ReadHandler::" << ec.message() << '\n';
    }
}

ReadFunc:

void HandleIncomingData(const char *data, const std::size_t size) {
    xxx::messaging::CMSMessage proto;
    proto.ParseFromArray(data, static_cast<int>(size));
}

I should mention that I need this to be as fast as possible, so any optimizations would be very much appreciated as well.

Answer 1

The program invokes undefined behavior as it fails to meet a lifetime requirement for boost::asio::async_write() 's buffers parameter:

[...] ownership of the underlying memory blocks is retained by the caller, which must guarantee that they remain valid until the handler is called.

Within the Write() function, boost::asio::async_write() will return immediately, and potentially cause data to go out of scope before the asynchronous write operation has completed. To resolve this, consider expanding the life of the underlying buffer, such as by associating the buffer with the operation and performing cleanup in the handler, or making the buffer a data member on TCPSender .