Re: Serial COM port communication problems

On 21 Mai, 03:21, "Scott McPhillips [MVP]" <org-dot-mvps-at-scottmcp>
wrote:
It is not really practical to use the rx flag to detect end of message.
The problem is that your program is subject to latency. It may not
resume execution for 50-100 msec after the CR is received, and in that
time more characters can arrive.

But this should not cause a loss of data, assuming you are reading the
actual number of characters retured by WaitCommEvent. If you are losing
characters you have a bug in your program, or perhaps ClearCommError
will show a hardware timing error.

You are right, I could experience this behavior today. The rx flag is
used to detect the subpackages
of the hole package. As soon as the CR is detected an int event
counter (or subpackage counter)
is incremented. The event is always very accurate detected. So at
least I have some kind of
signal to know how many subpackages arrived. Then the queue size is
requested by ClearCommError,
as some new bytes already arrived the size returned is bigger than the
actual subpackage, including the CR.
Later on Readfile is called. Non OVL implementation clears the
receiving queue, the OVL
implementation doesn't yet.

I suggest that you use overlapped I/O and the SetCommTimeouts to cause
it to return when the buffer is full or a brief silence (a few character
times) is detected. Then parse the returned buffer to find the CR characters yourself.
With this approach I have successfully stress tested the COM ports at extreme speeds
and loads without ever losing a character.
Scott McPhillips [MVP VC++]

How can this be done? (SetCommTimeout association to buffer size?)...I
will try to understand the overlapped
method better. Do you mean the Readfile Buffer that receives the data
read? I'm free to set the size of that buffer,
I will need to find the best size for this buffer, subpackages vary
amoung them as much as 20 bytes.
I'm a bit afraid of adding timing parameters to the problem but it
seems an elegant approach, hope
doesn't get more complex.

The problem by this type of signalization (the new timeouted buffers)
is that there will be CRs embeeded
on them. So the CR is first to be found, and regard the left part as a
subpackage, the right part as a part
of a subpackage an so on...? Is there some type of generic algorithm
to build the subpackages together?

Thank you very much for your advice

.

Relevant Pages

  • Re: xPC Serial Communication Problems
    ... Is the receiving machine using a read call to its system that can ... does it wait until there are at least 15 characters in the receive ... buffer, then return exactly the first 15, or does it return whatever ...
    (comp.soft-sys.matlab)
  • [UNIX] wu-ftpd fb_realpath() Off-by-One Bug
    ... Wu-ftpd FTP server contains remotely exploitable off-by-one bug. ... characters while the size of the buffer is MAXPATHLEN characters only. ... Following FTP commands may be used to cause buffer overflow: ...
    (Securiteam)
  • Re: Serial Port CE_OVERRUN errors
    ... SerialNG component. ... The main Input buffer is set to 32000, and I have a variable which is ... updated with the maximum number of characters waiting there each time the ... I tried boosting the baud rate from 57.6k to 115.2k, ...
    (comp.lang.pascal.delphi.misc)
  • Re: WaitCommEvent and ResetEvent for SerialPort
    ... This is extremely true when you say you are exhausting the input buffer, ... This receiver thread will then signal your main ... >> From the WaitCommEvent documentation, ... Since no characters are currently coming it, ...
    (microsoft.public.win32.programmer.kernel)
  • Re: xPC Serial Communication Problems
    ... Everything about the packet looks just fine, ... reasonable facsimile) so the hardware buffer should be 16 bytes. ... Is the receiving machine executing at the 'same' 50 ms rate? ... does it wait until there are at least 15 characters in the receive ...
    (comp.soft-sys.matlab)