Calling an MFC-Socket vom Java

[gobo]

Hallo,

I have an C++/MFC-Program that creats a socket server via a class
derived from CSocket. In its ::OnReceive()-Method it does its job.

In a Java-Class, this socket is opened (new Socket()) and its Input-
and OutputStream used for communication.

That works so far. However: sometimes(!) the communication is broken.
The Request from the java-Part is not received in the OnReceive()-
Method in the MFC-Server. So, everything hangs, the Java-part is
waiting infinitely for an answer that is not computet, because the
Server did not receive the request.

debug-output has shown, that the previous OnReceive()-method in the C+
+/MFC-part did run to an end - seemingly without any error. But after
that, it does not receive anymore. The writing of data into the
OutputStream also did not result in an Exception.

I thought, one process overtook the other, so i put in a sleep in the
java-part after receiving any answer - but it did not work.

Curiosly enough - the problem does not occur on every PC, but there
are PCs where the error occurs more often that on others. Also we had
a nearly a whole day where the error did not occur at all!

It also does not occur alway at the same call. It is seamingly random
- but I dont think so. There must be a reason.

We are complete at a loss. Is there any important status, that we
could read before sending the request in java? Is there any way to
check if the MFC-Socket is still open and listening? Is there a way to
restart the socket?

Thanx for any help - we really could do with a hint
Goetz

 

[Gordon Beaton]

I have an C++/MFC-Program that creats a socket server via a class
derived from CSocket. In its ::OnReceive()-Method it does its job.

In a Java-Class, this socket is opened (new Socket()) and its Input-
and OutputStream used for communication.

That works so far. However: sometimes(!) the communication is broken.
The Request from the java-Part is not received in the OnReceive()-
Method in the MFC-Server. So, everything hangs, the Java-part is
waiting infinitely for an answer that is not computet, because the
Server did not receive the request.

If the communication were *broken* you would get an indication
indication that fact when either reading or writing.

Possibly your protocol does not account for the fact that TCP does not
preserve message boundaries. Message sometimes arrive in parts, and
two messages can arrive together.

Use a tool like Wireshark to see what's actually happening on the
network.

Probably you should post some example code.

/gordon

--

 

[Christian]

Hallo,

I have an C++/MFC-Program that creats a socket server via a class
derived from CSocket. In its ::OnReceive()-Method it does its job.

In a Java-Class, this socket is opened (new Socket()) and its Input-
and OutputStream used for communication.

That works so far. However: sometimes(!) the communication is broken.
The Request from the java-Part is not received in the OnReceive()-
Method in the MFC-Server. So, everything hangs, the Java-part is
waiting infinitely for an answer that is not computet, because the
Server did not receive the request.

debug-output has shown, that the previous OnReceive()-method in the C+
+/MFC-part did run to an end - seemingly without any error. But after
that, it does not receive anymore. The writing of data into the
OutputStream also did not result in an Exception.

I thought, one process overtook the other, so i put in a sleep in the
java-part after receiving any answer - but it did not work.

Curiosly enough - the problem does not occur on every PC, but there
are PCs where the error occurs more often that on others. Also we had
a nearly a whole day where the error did not occur at all!

It also does not occur alway at the same call. It is seamingly random
- but I dont think so. There must be a reason.

We are complete at a loss. Is there any important status, that we
could read before sending the request in java? Is there any way to
check if the MFC-Socket is still open and listening? Is there a way to
restart the socket?

Thanx for any help - we really could do with a hint
Goetz

amke shure you are calling flush() on the outgoing stream. That might be
the problem. For further guessing it might be better if you posted some
of your code.

Christian

 

[Gordon Beaton]

Hi, thanx for the ideas.

Here's the code:

How long is a typical message? Does the other side loop in a similar
manner when reading the incoming messages?

I the lunch break I have heard rumours about a possible java error in
the actual JRE.

Once again: get Wireshark and look at the actual traffic. All else is
speculation.

/gordon

--

 

[gobo]

If the communication were *broken* you would get an indication
indication that fact when either reading or writing.

Possibly your protocol does not account for the fact that TCP does not
preserve message boundaries. Message sometimes arrive in parts, and
two messages can arrive together.

Use a tool like Wireshark to see what's actually happening on the
network.

Probably you should post some example code.

/gordon

--

Hi, thanx for the ideas.

Here's the code:

Writing the Steam in Java
======
public void sendData(String data)
{
writer.println(startTransmit);
writer.println(data);
writer.println(endTransmit);
writer.flush();
}
======
(startTransmit and endTransmit are begin and end signs to synchronize
the communication.)

Reading in Java
======
public String receiveData() throws IOException
{
String retStr = null;
recvBuffer = new StringBuffer(200);
do
{
retStr = reader.readLine();
if( retStr == null )
return(null); // Connection lost ?

}while(checkReceiveContent(retStr));

return(recvBuffer.toString());
}
======
(checkReceiveContent() checks, if the endTransmit-Code is received and
returns true, false otherwise)

So, the connection is flushed and synchonized.

The MFC-Part uses CSocket::Receive() and CSocket::Send() to
communicate. Up to now, the return code from CSocket::Send is ignored.

I the lunch break I have heard rumours about a possible java error in
the actual JRE. We are using
=======
java version "1.6.0_02"
Java(TM) SE Runtime Environment (build 1.6.0_02-b06)
Java HotSpot(TM) Client VM (build 1.6.0_02-b06, mixed mode)
=======

Thanx, Goetz

 

[Gordon Beaton]

i got and installed wireshark. Can anyone tell me a good filter? The
Communication is between "localhost" on port 4441.

Assuming you mean "capture filter", something like
"port 4441" or "proto tcp and port 4441" should work.

/gordon


--

 

[gobo]

How long is a typical message? Does the other side loop in a similar
manner when reading the incoming messages?


Once again: get Wireshark and look at the actual traffic. All else is
speculation.

/gordon

--

Hi,

i got and installed wireshark. Can anyone tell me a good filter? The
Communication is between "localhost" on port 4441.

Thanx,
Goetz

 

[gobo]

Assuming you mean "capture filter", something like
"port 4441" or "proto tcp and port 4441" should work.

/gordon

--

Sounds good. But how can I capture localhost? The Capture-Dialog only
displays the Interfaces and a local communication does not run through
my network device, right?

How can I capture 'localhost'?

Sorry, but you see, I am to total wireshark-newbie.

Goetz

 

[Gordon Beaton]

Sounds good. But how can I capture localhost? The Capture-Dialog only
displays the Interfaces and a local communication does not run through
my network device, right?

How can I capture 'localhost'?

There is a capture menu where you can specify the interface to capture
from. On some platforms, you might be able to capture from the
localhost device.

Or, instead of connecting via "localhost", have your client connect to
the hosts "real" interface instead by specifying that address instead
when you connect (i.e. 192.168.128.3 instead of 127.0.0.1). That may
be sufficient to make the traffic available to Wireshark, even though
no traffic will actually get sent on the network in either case.

If that doesn't do it, then you need to run one of the processes on a
different host, then run Wireshark on either the client or the server
host.

/gordon

--

 

[Nigel Wade]

Sounds good. But how can I capture localhost? The Capture-Dialog only
displays the Interfaces and a local communication does not run through
my network device, right?

How can I capture 'localhost'?

Sorry, but you see, I am to total wireshark-newbie.

Goetz

You can't on Windows.

You will need to modify your code so it uses the IP address of the interface,
rather than 127.0.0.1, localhost or whatever you use.