Re: [Denemo-devel] midi input

[Richard Shann]

On Thu, 2008-11-06 at 15:35 -0600, Jeremiah Benham wrote:
> On Thu, 2008-11-06 at 16:22 +0000, Richard Shann wrote:
> 
> > 
> > this seems to have no effect on all the above dignostics.
> > You mentioned that midi.c was broken - do you recall in what way it was
> > broken? Was it working at one time?
> 
> I disabled it a while back because it was cause segfaults and have not
> had time to fix it.
I mentioned that by changing the code in denemo/src/midi.c from
accessing /dev/sequencer to /dev/sequencer2 I had got past the first
ioctl() call - I wrote

> Also I have both /dev/sequencer and /dev/sequencer2 but
> only the second responds to the ioctl (sequencer_fd,
> SNDCTL_SYNTH_INFO,
> &card_info) call in midi.c
> it comes back with this:
> 
> Synthesizer detected: Midi Through Port-0
> Synthesizer supports 16 voices.

well, I've now found that by doing
aconnect 14 128
(that is midi through port-0 to timidity) I get immediate-playback
working in Denemo. The latency is again not good enough for seriously
quick note input to Denemo. However, I could use it to test any code I
write.
So far I have a simple input routine reading from /dev/midi1, my
music-keyboard. It gets NOTE_ON (0x90), <midinum>, <velocity> each time
I press the a key, and (strangely) the same but with velocity 0 each
time I release a key. (The MIDI standard says it should be NOTE_OFF
(0x80) with the velocity of the release).
Where to go with all this?
I think we should turn on the immediate playback possibility - the
segfaults may well have been from bugs elsewhere in the code, now fixed
(I haven't encountered one yet after turning it back on).

Richard