[Media_api] portmidi exits, sysex, and code changes

[Roger Dannenberg]

Oron,
	Here are better answers to your 3 or 4 questions/comments:

Yes, there is a call to exit() in *source code* portmidi.c, but this is 
only compiled if PM_CHECK_ERRORS is defined. This is the default in the 
makefiles because it's probably the right setting for development work. 
It's convenient because it catches and reports errors even if you don't 
explicitly check return values in your application. "Serious" 
applications and non-console Win32 apps should compile without 
PM_CHECK_ERRORS, check return values explicitly, and handle them.

portmidi28aug06 is not the latest version. I can't check the website at 
the moment, but I think it's portmidi23oct06, and I restored some code 
that was somehow dropped from the portmidi17nov04 version. If there's 
still stuff missing, please let me know.

The "huge buffer of 1024" should not really be a problem -- the buffers 
are passed around by reference (although perhaps there's a copy when you 
cross the kernel boundary?). I think it would be more of a problem to 
use small buffers in the case of large sysex messages. Allocating the 
right size for each message would incur heap allocation and free 
overhead, and could cause some serious real-time problems acquiring 
locks to a shared heap (as in the default thread-safe malloc()). I think 
there is far more overhead handling byte-by-byte delivery of even short 
sysex messages in the device driver than possibly copying up to 1024 
bytes for nothing, but I'm open to exploring any apparent performance 
problems.

Finally, you said sometimes sysex messages are not delivered to the MIDI 
device. I've experienced synthesizers dropping sysex messages when they 
are sent at the full MIDI data rate, and portmidi does not try to 
throttle MIDI data rates. Another possible problem is internal buffer 
overflow: you should make sure the portmidi output buffer size passed to 
pm_OpenOutput() is large enough for the largest sysex message and you 
should check all return values for errors. If that doesn't help, let me 
know how to get portmidi to drop messages and I'll track down the 
problem. A good way to start is to use two machines -- one to send, and 
one to receive and check. The pm_test/sysex.exe program uses a similar 
idea, but sends to itself, and seems to work reliably.

Thanks,
Roger