Improvements to <exec> function

[dolk13]

I added only 'exec verify'.
Improving output tracing brings a lot of changes. I will try to do PR later.

resolve #7

[wdoekes]

@rkday : something you want to review?

This may need some docs, as documenting in code isn't enough if we want people to find it.

[rkday]

@dolk13 , thanks for this!

As Walter says, the docs need an update to mention verify:

• https://github.com/SIPp/sipp/blob/master/docs/scenarios/actions.rst#external-commands
• and a DTD update at https://github.com/SIPp/sipp/blob/master/sipp.dtd#L102

Probably the key thing to mention as a difference between exec command and exec verify is that the latter is synchronous rather than asynchronous - SIPp will wait for the verified command to run before proceeding, and nothing can happen (no SIP messages can be handled, new calls set up, etc.) while it is running.

That makes me somewhat nervous, because I think this implementation is only suitable for very low loads (or very fast commands) - I ran a scenario with a verify-command that took 200ms to run, and despite me requesting a call setup rate of 10/sec it only actually achieved 2.7/sec.

I think a better approach would be:

• don't call waitpid() in call.cpp
• instead, pause the call (task.cpp has a setPaused method) and add a pointer to the call to a std::map keyed off the PID
• at the start of each iteration in traffic_thread in sipp.cpp, call waitpid with the WNOHANG option and a PID of -1, to check of any child process has exited
• if it has, unpause the call associated with it, and verify the return code

But I think I'd be OK to accept this without this change as long as the impact on the call rate was clearly documented, and improve performance in future.

[rkday]

Do we not still need this INTCMD section?