The latest pd test version seems to have a few small bugs. (but even on linux the latency seems to be much improved)
For some reason the midi port assignment substracts 2 to get the device, instead of 1, so -midiindev 3 is mapped to /dev/midi1
The jackdriver tries to reconnect to the jack-graph if you switch on audio processing after switching it off the first time, but it hasn't removed itself, so this fails because there is allready a client with that name.
The alsadriver only uses the first two channels untill you reset it in the "media" menu, even if the device has more channels. Regardless what you configure it to, pd will then use all available channels.
I can probably fix most of this myself and send a patch, but I won't have time for the next two weeks I think. BTW should I sent patch here or somewhere else?
Gerard