Cross-platform serial port access

[David Brown]

It looks like it will be a while before I can do much about this - maybe one
day.  Anyway, thanks for your help.


Grant Edwards wrote in message ...
>In article <99d7g3$qed$1 at news.netpower.no>, David Brown wrote:
>
>>>Currently, serial port stuff in Python is not portable between
>>>Mac/MS/Unix.
>>
>>I suppose it should be easy enough to make a small wrapper around the
>>different platforms' code.
>
>Yup -- somebody should do that someday...
>
>>>>Does anyone know if Linux supports RTS toggle handshaking
>>>>(i.e., the RTS line is high while transmitting, for convenient
>>>>connection to RS-485 convertors),
>>>
>>>Not with the standard PC serial port driver.  I hacked the
>>>16x50 driver to do RTS toggle several years ago, but it never
>>>worked reliably.  It's very difficult to do it reliably in
>>>software at high baud rates.
>
>Back when I tried to do it with the Linux driver, I was trying
>to use extra transmit data bytes (didn't work well) and later
>the shift-register empty bit to shut off RTS.  I should have
>done it using loop-back: enable loopback in the UART and shut
>off RTS in the receive handling code when the end of the tx
>message is seen.  That only works if loopback in the UART
>doesn't disable the transmit data pin...
>
>>>Your best bet is to buy a board that supports RTS toggle in
>>>hardware or use a 232-485 converter that handles it
>>>automatically by detecting the presence of data.
>
>>This is definitely a shame.  The RTS toggle works well under NT
>>(no idea about Win9x) - I have written a Delphi program that
>>uses it to run a RS-485 convertor at 57600.  I don't know for
>>sure how fast RTS is turned off after the last byte - the other
>>nodes on the bus wait 10 ms before answering anyway.
>
>That's the trick.  10ms is long enough that it can be done in
>software.  I had to get it shut off within 1 byte-time of the
>last stop bit.  At 57.6K that's 17us.
>
>>I guess under Linux I would have to toggle RTS manually.
>>Turning it on before sending is easy enough, it is turning it
>>off at the right time that is more of a hassle.
>
>Yup - it's extremely tricky.  I finally gave up and bought a
>board that did it in hardware.  If you've got a converter that
>loops back transmit data, you can use that: do a write(), then
>do a read() -- when you've seen the right number of bytes, shut
>off RTS.
>
>Some of the middle and upper level serial driver layers in
>Linux have a 10ms granularity -- so you might not be able do
>anything from user-space faster than 10ms.
>
>>So-called "intelligent" RS-485 convertors that read the
>>incomming data and change direction appropriately have their
>>own problems.  We have delivered systems that communicated at
>>300 baud, and "intelligent" convertors frequently do not work
>>at such slow rates.  Additionally, they generally need
>>configured with DIP-switches and the like, which is an extra
>>hassle if you need to change the baud rate.
>
>Yup.  The ones without dip-switches (fixed tx data timeout)
>will only work well at certain baud rates.
>
>>Simple RTS-controlled convertors are much easier in that
>>respect.
>
>I agree.
>
>>Although it is not really Python-specific, do you know if there
>>are serial drivers for Linux that simulate 9-bit UARTs (i.e.,
>>swapping the parity bit from space to mark as appropriate)?
>
>I don't think so.
>
>--
>Grant Edwards                   grante             Yow!  .. Do you like
>                                  at               "TENDER VITTLES?"?
>                               visi.com