exit status on communication failures

[danwinship]

Currently omping only exits with non-0 status if there are problems with the command line.

When using -c or -T it seems like it should also exit with an error if either (a) it doesn't manage to communicate with a given host before the timeout, or (b) it manages to communicate with the host via unicast but not via multicast.

(I'm working on an end-to-end test of multicast connectivity between containers in OpenShift. omping is nice for this because omping -c 1 will exit quickly in the case where unicast works but multicast doesn't, rather than forcing us to guess an appropriate timeout.)

[jfriesse]

@danwinship First of all, good catch, thanks.

I can fully agree with (a).

(b) is little more problematic because building multicast tree can take a longer time and it's not unusual that first few multicast packets are lost. But I believe it's possible to use adjusted_loss and if it is 0, it's perfectly legal to return success.

I'll try to make some patch.