[fpc-pascal] Sending break on (Windows) serial comms

[Mark Morgan Lloyd]

Mark Morgan Lloyd wrote:
> Marco van de Voort wrote:
>> In our previous episode, Mark Morgan Lloyd said:
>>> Marco, I see you had a StackOverflow question on this a year or so 
>>> ago. Did you ever get anywhere with it in practice?
>>
>> It was an idea to lower the overhead and improve the reliability of our
>> serial protocol. (because of the out of band information is guaranteed
>> unique and can't happen in the datastream)
>>
>> But if you can't insert the break cleanly into the character stream, 
>> it is
>> useless for protocol purposes I guess. And it was not worth centering the
>> rest of the application around it.
>>
>> I did't have any confidence in the solutions presented (just reset the
>> connection and pauze) because both the pauze is not reliable in length on
>> preemptive systems like Windows, and the damage to the FIFOs is a 
>> problem,
>> since that would negate the size advantages.
>>
>> I never pursued it further.
> 
> OK. I'll start off with the standard API, with an explicit check that 
> the output has drained by default, and a mSec parameter (which will get 
> rounded up to something gross on unix).
> 
> I'm still hoping that at some point the units and test code will be 
> bundled in place of the existing unix-only serial.pp.

I've added it, but it's best avoided except for the classic use of 
getting a remote system's attention. On Linux the only working parameter 
is zero indicating a break of around 250 mSec, on Solaris it doesn't 
work at all, and on Windows I've not found a way of preventing a break 
from trashing anything that's sitting in the Tx buffer.

Modified library appended to issue 18946 in Mantis.

-- 
Mark Morgan Lloyd
markMLl .AT. telemetry.co .DOT. uk

[Opinions above are the author's, not those of his employers or colleagues]