Serial Communication NT/95 - NONOVERLAPPED does not block execution untill HW end sending message. It seems that the buffer get erased before all bits are sent by the transport.

The methode I used is to send a zero byte in a slow bit rate (9600) before resuming to normal baudrate (19200).
I need to send a syncbreak (at least 13 consecutive bits in zero).
I use NONOVERLAPPED communication.
I wrote a "sendbreak" subroutine that changes baudrate to 9600,
sends a byte, than, changes baudrate back to 19200.
The problem I noticed is that on some computers the break signal is partialy executed, on others it is not executed at all. On all computers each of the lines of the routine performs perfectly when I set a "debug-break" before changing the baudrate back to 19200.
It seems that "WriteFile( ....,NULL) does not block execution before executing next line.
What can I do in order to assert that the message was fully executed on the hardware before executing next line?

Thank you,

Who is Participating?
DarthModConnect With a Mentor Commented:
Submitted to PAQ with no points refunded (of 245)

Community Support Moderator
May be you need wait untill all bytes will be sended ;-)

    while(GetOutQueueSize())  Sleep(0);


UINT GetOutQueueSize()
      COMSTAT Stat ;
            DWORD dwErr = 0;
            ClearCommError(m_hPort, &dwErr, &Stat);
             return (int)dwErr;
      return Stat.cbOutQue ;
ronnyzAuthor Commented:
Hi ,

I tried the method you have sujjested but it remains at the same. It is like a dubble check that proves that the buffer is empty and indeed it is empty although the phisical layer hasn't transfered yet. Therefore when I change baudrate after these checks it cuts the transmition of the syncbreak.

I tried working with a loopback in hardware - here I am trying to assert that the byte was transfered by receiving it into the receive buffer - this method was too slow - about 40 bits time.

An other method was implementing a timer but this is not reliable as they have a latency of 10 to 55 mSec depending on the PC.

Any idea of how to overcome?

Free Tool: IP Lookup

Get more info about an IP address or domain name, such as organization, abuse contacts and geolocation.

One of a set of tools we are providing to everyone as a way of saying thank you for being a part of the community.

The latency is OS intrinsic. Even with another approach the OS might claim time for its own purposes and thus interupting your code. This will affect you espescially on 95.

Quote from MSDN:

Specifies the number of bytes of user data remaining to be transmitted for all write operations. This value will be zero for a nonoverlapped write.

This means you need to go to overlapped IO in order to get the number of bytes in the buffer.
ronnyzAuthor Commented:
cbOutQue must be zero in nonoverlapped as the code is blocked untill all data is moved out from the transmit buffer to the transport layer. One can ask for the status of the buffer only after the write has been finished - buffer is empty.

I looked in MSDN in a technicle article "complition I/O". It says that overlapped events can achieve signal completion as in nonoverlapped. I tried that too but it is no better.

It seems that I should find a way to get a shorter latency with  timers. I need a latency that is shorter from 2 milli seconds.
Any idea how can I achieve that?
Thanks, Ronny
You can use HighResolution (or Multimedia) timers
see timeSetEvent in MSDN
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.