[pony-build] Some initial timeout stuff in _run_command

[C. Titus Brown]

On Tue, Feb 23, 2010 at 07:26:18PM -0400, Max Laite wrote:
> http://github.com/mlaite/pony-build/blob/subprocess/client/pony_client.py#L93
> 
> In the process of writing a test for this......

Why not just do a 'time.sleep(timeout)'?

And note that _run_command needs to return a triple.  What happens if
instead of returning None, you just let it fall through to 'communicate'?

Otherwise I like, thanks!

You'll probably need to do a merge with my latest master, where I've posted
some changes to _run_command.

> One thing I need to discuss is, do we split the timeout off to its own
> func?(not sure if its even really possible like that with current setup of
> _run_command) or do we just add timeout argument to all the *Clones so that
> you can specify in build scripts, then when the *Clones call _runPcommand
> they pass timeout like that?
> 
> Do we do two separate versions of _run_command?

Hmm.  I think it belongs in _run_command, being passed in as a kw arg.  Also
have a global TIMEOUT variable in pony_client.py that is overridden if a
timeout is passed into _run_command.

If 'timeout' is set to None, don't poll at all and just fall through to
communicate().

Sound good?

--titus
-- 
C. Titus Brown, ctb at msu.edu