Question

I have a program that performs overlapped I/O on a serial port and I am seeing cases where data that has been received is not being read by the application. My code is using SetCommMask() and WaitCommEvent() as follows:

if(!waiting_for_comm_event)
{
   comm_event_ready = false;
   if(!SetCommMask(port_handle, EV_RXCHAR | EV_ERR | EV_RLSD))
      throw OsException(my_strings[strid_set_comm_mask_failed].c_str());
   rcd = WaitCommEvent(port_handle, &event_mask, &event_control);
   last_error = GetLastError();
   if(rcd && event_mask != 0)
       comm_event_ready = true;
   else if(last_error != ERROR_IO_PENDING)
       throw OsException(my_strings[strid_wait_comm_event_failed].c_str());
   else
       waiting_for_comm_event = true;
  }

This code is invoked after I have written any data in my output buffer. There is a possibility that data could be received while this writing is taking place. Will WaitCommEvent() report the condition where there is already data waiting to be read?

Answer 1

If you have to ask here, you don’t want to know the answer.

No, really — it means you’re relying on undocumented behavior that could change in future Windows versions.

You can instead use overlapped I/O to start tracking incoming data and events before you send the data that triggers them.

Two approaches (I use a ReadFileEx variant of the first in my own serial code):

• Issue a background (overlapped) read.
• Issue a background (overlapped) WaitCommEvent.
• Write the data.
• Wait on both background operations… completion of either one will cause WaitForMultipleEvents to proceed.

or

• Issue a background (overlapped) WaitCommEvent.
• Write the data.
• Wait on the background operation.
• If the wake event was receipt of data, read it.

In either case, incoming data and errors, even before WriteFile completes, will be detected.