[Ns-bugs] [Bug 631] RealtimeSimulatorImpl not compatible with python bindings

[code@nsnam.ece.gatech.edu]

http://www.nsnam.org/bugzilla/show_bug.cgi?id=631





--- Comment #8 from Gustavo J. A. M. Carneiro <gjcarneiro at gmail.com>  2009-07-13 07:32:05 EDT ---
(In reply to comment #6)
> (In reply to comment #5)
> > IMHO we should avoid threads in non-realtime simulator due to performance
> > issues and portability issues.  I think the RunOne solution is so simple, why
> > don't you like it?  It is surely a lot simpler than any threaded solution you
> > may want to throw at it.
> 
> Because it makes it impossible to support the realtime scheduler, the mpi
> parallel scheduler, and, the threaded parallel scheduler. To summarize, nothing
> but the simplest default simulator works with RunOne. So, yes, it's simple,
> but, it kills our ability to do fancy things underneath: it is exposing too
> much from the event loop so, I would rather kill it.

It does not make it impossible to do anything, you are exaggerating.  For the
other schedulers we can use Simulator::Run and not catch signals.  Not being
able to catching signals in Python is not a great loss IMHO.

Maybe what we want is to remove RunOne as a generic API and make it specific to
the default simulator implementation.  That's what I would do, if not for your
mania of having huge ABCs with many pure virtual methods that have to be
implemented even if they make no sense for some subclasses (see also class
NetDevice).


-- 
Configure bugmail: http://www.nsnam.org/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.