Re: [Csnd-dev] Art's Issue

On my macOS I tested your .csd and also created my own .csd (attached)
which behaved as follows:

Both files received keystrokes and translated them to MIDI messages that
were sent out from Csound.

Both files recorded valid MIDI files that I could load into Reaper and play.

Only the first keystroke actually played a note on the synth. All the other
keystrokes, though they continued to be translated and to be saved, were
silent.

When the user ended the Csound performance with Control-C, the Midiview app
on the Mac showed all the note off messages arriving at once; no trace of
the note on messages.

This issue describes similar behavior experienced by another user:
https://github.com/csound/csound/issues/1784

Regards,
Mike

-----------------------------------------------------
Michael Gogins
Irreducible Productions
http://michaelgogins.tumblr.com
Michael dot Gogins at gmail dot com


On Thu, Aug 14, 2025 at 7:01 PM Arthur Hunkins <[email protected]> wrote:

> Hi, Mike -
>
> Good and helpful discussion this afternoon.
>
> Glad to have any feedback regarding my issue as demoed in the attached
> .csd.
>
> The problem is simple: run as is, everything works as expected, *except* that
> the MIDI file written is empty. This is associated with the fact that the
> duration of performance (via ASC keyboard, the first 8 number keys) is
> infinite (z) and must be stopped by CTRL-C.
>
> Changing the z infinite duration to a finite number (say 5 or 10) finishes
> the performance normally and writes an appropriate MIDI file. (Try it and
> see/verify.)
>
> According to Victor L., z (CTRL-C) should write a proper MIDI file as
> well. This is the only issue. The audio performance depends on the right -Q
> value, but the MIDI file output seems totally independent of this.
>
> Actually, as long as you have some kind of MIDI device plugged in (or an
> active MIDI cable), you can test this out without any audio at all. Just
> see whether the MIDI file (which if Csound runs, will be created) has any
> bytes!
>
> Thanks for any insight. Victor suggests that "this is a Windows problem" -
> which may well be true. (Do you find the same results with Mac or Linux?)
>
> Art H.
>

<CsoundSynthesizer>
<CsLicense>
Run this .csd after starting your MIDI controlled instrument. I have tested 
this with the standalone Pianoteq.

There is a problem in that all keystroke events are correctly scheduled and 
the MIDI file is correctly recorded for all keystrokes -- but only the first 
keystroke actually makes a sound.

When the Csound performance is ended with Ctrl-C, then the Midiview app 
finally shows events from all keystrokes being turned off. But the initial 
NOTE ON events are never shown (except for the very first).
</CsLicense>
<CsOptions>
-m167 -odac -d -Q2 -+rtmidi=PortMidi --midioutfile="senskey_test.mid"
</CsOptions>
<CsInstruments>
sr = 48000
ksmps = 128
nchnls = 2
0dbfs = 1
gk_channel init 1

instr 1
; The `sensekey` opcode does not return 1 for `k_pressed` as long as the kay 
; is actually held down, then return 0 when the key is released; rather, the 
; opcode 'emulates' this behavior by returning 0 immediately after returning 
; 1. Therefore, `sensekey` cannot provide a duration for key presses without 
; some more elaborate code that 'latches' the keyboard state. Here, we simply 
; send a short fixed duration that represents a struck bell.
k_onset init 0
k_duration init .01
k_velocity init 80
k_key sensekey 
if k_key > 0 then
schedulek 2, k_onset, k_duration, k_key
endif
endin

instr 2
i_channel = 1
i_key = p4
i_velocity init 80
i_duration init .01
noteondur i_channel, i_key, i_velocity, i_duration
prints "MIDI note: channel %9.4f key %9.4f velocity %9.4f duration %9.4f\n", i_channel, i_key, i_velocity, i_duration
endin

</CsInstruments>
<CsScore>
f 0 2000
i 1 0 2000
</CsScore>
</CsoundSynthesizer>