Re: MIDI parser source (bytestream->MIDI messages)

Matthijs Hollemans wrote:
> > any hints where I can find source code which takes a raw byte stream=20
> > and parses it into MIDI messages?
>=20
> The midi_server does something like that already. The code is in=20
> PortDrivers.cpp. It appears to work on at least some systems, but it is=
=20
> not finished yet. The code to split the raw bytes into MIDI message is=20
> actually pretty simple:
> http://cvs.sourceforge.net/viewcvs.py/open-beos/current/src/servers/mid=
i/PortDrivers.cpp

Thanks! That will help alot -  having the switch() and case statements=20
written out saves a lot of typing. :)

But I think there's a bug, or rather a missing workaround, in the code.=20
IIRC, some MIDI units don't send proper End-Of-SysEx messages (e.g. Yamah=
a=20
DX7), and might simply send normal channel messages after a SYX-dump. So=20
the code should sense that the SYX-dump is complete after reading any byt=
e=20
with 0x80 set - SysEx data bytes are always only 7 bit.

> There is also supposed to be a midi_parser kernel module, but this is=20
> only intended to be used by device drivers. I don't know how it works, =
or=20
> whether we already have it in the OpenBeOS CVS.

Didn't find it, but I only had time for a quick look. Anybody knows where=
=20
it is (supposed to be) located? I currently don't have time to dig throug=
h=20
the CVS tree...
=20
>> [...] there already is serial_midi, but it only does MIDI-Out, not
>> MIDI-In, and this isn't acceptable.
>=20
> Hmm, I didn't know that.

I found out the hard way. With a self-made cable, which I resoldered n=20
times. Desperation. Then I tried out a CBX driver under Windows, which ga=
ve=20
me MIDI-In on the first try... <sob>

> You could always write a replacement driver for=20
> serial_midi, of course :-) Then the midi_server will take care of parsi=
ng=20
> the raw byte stream.

Hmm, later maybe. AFAIK OBOS will get a new driver system anyway, so it=20
might be better to wait 'til then. And I don't have any experience with=20
writing kernel drivers yet, and am still a bit respectful of that...

Bye,
Chris