Subprocess Popen confusion

[Dick Holmes]

In article <mailman.2.1589442315.11657.python-list at python.org>, 
__peter__ at web.de says...
> 
> Dick Holmes wrote:
> 
> > https://occovid19.ochealthinfo.com/coronavirus-in-oc 
> 
> > I'm trying to
> > communicate using a continuing dialog between two
> > processes on the same system.
> 
> I think pexpect
> 
> https://pexpect.readthedocs.io/en/stable/index.html
> 
> does this naturally, but I don't know if Windows support is sufficient for 
> your needs.
> 
> > I've looked at various mechanisms and the
> > class that seems to fit my needs is Popen in the subprocess module, but
> > I can't seem to get more than a single round-trip message through Popen.
> > I first call Popen then poll using the identifier returned from the call
> > and the poll seems to work. I then call the communicate function passing
> > None as the value to send to the companion process stdin. I get the
> > expected result, but I also get "Exception condition detected on fd 0
> > \\n" and "error detected on stdin\\n". Subsequent attempts to
> > read/write/communicate with the subprocess fail because the file (stdxx
> > PIPE) is closed.
> > 
> > I can't tell from the documentation if the communicate function is a
> > one-time operation. 
> 
> Yes, communicate() is one-off, 
> 
> 
>         """Interact with process: Send data to stdin and close it.
>         Read data from stdout and stderr, until end-of-file is
>         reached.  Wait for process to terminate.
>         ...
>         """
> 
> seems pretty clear. What would you improve?
> 
> > I have tried using read but the read call doesn't
> > return (I'm using winpdb-reborn to monitor the operations).
> 
> Try readline(). Deadlocks may happen ;)
>  
> > I'm using Python 3.7, Windows 10, winpdb-reborn 2.0.0, rpdb2 1.5.0. If
> > it makes any difference, I'm trying to communicate with GDB using the MI
> > interpreter.
> > 

Per Peter's suggestion I tried readline and it works "as expected". I 
also discovered that the reason the read operations were stalling was 
that they followed a write and the write doesn't actually occur until 
"flush" is called even though the function call returns. There are 
undoubtedly some subtleties lurking about, but at least I'm making good 
progress.

Thanks for the help!

Dick