(I know there are other trackers for this bug, but I'm unable to comment on them)

When using the Java dashboard with a USB camera, CameraServer will refuse connections (saying "choose USB Camera HW" on 
the labview dashboard) until you open the labview dashboard, choose the option, then close the dashboard.

We debugged this issue with Wireshark and discovered that the Java dashboard actually sends its request data in 2 
packets most of the time. CameraServer was only picking up the first packet, leaving the last 2 fields of Request zeroed
 out.

Our fix is attached. Here are the relevant lines of code
---
		Request req;
		// the problem with the default implementation of receiving the Request in
		// CameraServer was that it required the entire buffer to already be there
		// by the time it got to read. Sometimes the buffer would not be there, which
		// caused a loop of SmartDashboard repeatedly trying to connect. The code
		// below makes sure we have the entire contents of the Request (12 bytes)
		// before we try to interpret the contents
		char requestBuffer[sizeof(Request)];
		unsigned int index = 0;

		bool connFailed = false;

		while (index < sizeof(Request)) {
			char next;
			int sizeRead = read(conn, &next, sizeof(next));
			if (sizeRead == -1) {
				wpi_setErrnoError();
				close(conn);
				connFailed = true;
				break;
			}
			if (sizeRead < 1)
				continue;
			requestBuffer[index] = next;
			index++;
		}

		if (connFailed) {
			continue;
		}

		// slightly unsafe cast. We use sizeof(Request) so it should be ok
		memcpy(&req, &requestBuffer, sizeof(Request));

		req.fps = ntohl(req.fps);
		req.compression = ntohl(req.compression);
		req.size = ntohl(req.size);
---