I’ve searched the source code. There is no reference to latency at all in SC_CoreAudio.cpp. So I have to conclude that CoreAudio handles driver latency differently from JACK or PortAudio. But I don’t know anything about CoreAudio internals.
If CoreAudio handles driver latency internally (so that scsynth doesn’t have to do anything about it), then there’s no need for anything to be printed… maybe?
You don’t have to.
If you do server.sendBundle(latency, ... msgs...), the language calculates a timestamp based on logical time now + latency. Then, the server is supposed to schedule the message so that the audio hits the speakers at the timestamp – that is, the synth will start calculating at the time timestamp - driver_latency. This all happens for you – SC user code should not have to compensate for driver latency.
(Windows MME drivers add a couple hundred milliseconds latency but don’t report this to client apps – if the driver lies to scsynth, of course it doesn’t work, so we recommend ASIO drivers in Windows.)