🚀 go-pugleaf

On Thu, 24 Oct 2002 11:16:55 -0500, "Brian Cook"
<bcook@rowdydogsoftware[REMOVE].com> wrote:

Brian,

>I used a dual processor computer for about four years to develop serial port
>(and TCP/IP) communications software.  Not once did I have a problem related
>to the fact that the computer was dual processor.  A dual processor
>computer, in my opinion, should do better at handling communictions than an
>equivalent single processor computer for the simple reason that at any given
>moment in time it is roughly twice as likely that a processor is available
>to handle an interrupt.

I now found, that it regularly happens when the message is about 10
bytes long. If it is longer than 20..30 bytes it never happens.

>
>>    SetCommMask(hCom, EV_TXEMPTY);
>
>Really should be...
>
>    SetCommMask(hCom, EV_TXEMPTY or EV_BREAK or EV_ERR);
>
>And somewhere you should be making calls to...
>
>    GetCommMask
>    ClearCommError
>
>To determine what happened.

It did not report any errors.

However, I noticed, that cbOutQue always reports 0 bytes when calling
this right after WriteFile() :

   if (!WaitCommEvent(hCom, &evMask, &os1))
    if (GetLastError() == ERROR_IO_PENDING)
    {
       ClearCommError(hCom, &Errors, &Stat)
      ....
    }

However when the message length is 20, cbOutQue is != 0 
(some bytes still in the queue)

Perhaps with 10 bytes all bytes go into the UART TX buffer at once and
an EV_TXEMPTY interrupt will randomly be missed ?

Jurgen

Jurgen Bartels, Taunusstein near Wiesbaden, Germany
jurgenb@rheingau.netsurf00.de  ( remove zeros in adr )