midi_cc_output: echo parameter changes out as MIDI CCs (#1175, part 2)

[dpwe]

Part 2 of #1175, following #1191: midi_cc_output, the converse of midi_cc's direct-parameter form.

iC<C>,<L>,<N>,<X>,<O>,<P>[,<OSC>][,<P>,<OSC>...]

amy.send(synth=1, midi_cc_output='74,1,100,6400,0,%d' % amy.PARAM_FILTER_FREQ)

When parameter P of the synth changes at voice-relative OSC, the new value is sent out as CC C, with the L,N,X,O mapping inverted. The same arguments on ic and iC round-trip every 0..127 value unchanged; the test asserts this for linear and log (with offset) maps. The synth's own voices still get the change.

Behaviour, as agreed on the issue

• Channel: the synth's note_output MIDI channel if it has one, otherwise the synth number. A synth outside 1..16 with no MIDI note output sends nothing, and making such a mapping prints a warning rather than guessing a channel.
• Timing: the CC goes out when the event is added, the same moment note_output sends notes. Sequenced changes go out on their step; an event scheduled for later with time= sends early.
• Stacking several mappings on one CC is deferred.

Details

• Where it hooks: at the top of amy_event_to_deltas_queue for events with a synth. It has to be there because:
  • bus-only events never reach patches_event_has_voices;
  • bus distortion fields are cleared once they've been turned into deltas.
• Stored events: only the live delta queue counts, so an event being stored into a patch sends nothing.
• Reading the value: amy_event_get_param reads the value back from the same param_fields table midi_cc uses to write it, so values are compared in amy.send() units. Matching rules:
  • an event naming no osc counts for every osc;
  • the distortion fields shared between osc and bus scope are matched by whether an osc is named, so PARAM_DIST_MIX and PARAM_BUS_DIST_MIX are told apart.
• Feedback guard: a change that arrived over MIDI isn't echoed unless the note output sets forward_midi_in. So ic and iC on the same CC don't form a loop.
• Sent on change only, per mapping. With several P,OSC pairs, one event sends at most one CC.
• Separate list: output mappings live in their own list. They never answer an incoming CC or mark a MIDI channel as active.
• Template form refused: iC with a wire-command template is refused, since there's nothing for an output to run.
• Clearing and state: iC<C> and iC255 clear mappings. yield_synth_commands emits them (codes 256..383) so they survive a state dump.
• Generated files: Python kwarg midi_cc_output; godot/amy.gd and amy_api.generated.js regenerated.

Tests

• New tests/test_midi_cc_output.c, 29 checks. It covers:
  • basic sends and clamping
  • dedupe
  • the full round trip
  • osc matching
  • multiple targets
  • bus params and the shared distortion scope
  • all three channel cases
  • the feedback guard, with and without forward_midi_in
  • output mappings ignoring incoming CCs
  • the template form being refused
  • clearing
  • state dump and restore
• make ctest: all pass.
• make test: 90 / 43 at ~-99 dB, identical to unchanged main in this environment.

Not in this PR: the note_output counterpart of the stored-patch guard (it sends notes from events being stored into a patch). That's a separate one-line fix.

🤖 Generated with Claude Code

https://claude.ai/code/session_01PykXkeeWw2aRpC6PLfpqTQ

---

Generated by Claude Code

[github-actions]

🎛️ AMY HW CI (AMYboard bench)

Flashed this PR's AMY (LoadTestChord: 6-voice Juno patch=1, one held note every 2 s) onto the physical AMYboard and measured the smoothed render load as the chord grows — back-to-back with the same sketch built at the PR's merge base, so Δ is this PR's own cost.

✅ PASS — the bench ran the test to completion.

notes held	main @ 19bd93f	this PR	Δ
1	996	991	-5
2	1147	1149	+2
3	1727	1717	-10
4	1885	1880	-5
5	2504	2489	-15
6	2631	2627	-4

Full chord settled render μs: 2628 (was 2626, Δ +0.1%) (peak 2630, 39 samples)

⬇️ Artifacts: serial log · load trace · report

_{Self-hosted bench (amyboardci). FAIL means only that the test could not run — the load values are informational, with no threshold and no audio compare. See tools/arduino_loadsweep/.}

[bwhitman]

⛓️ tulipcc integration PR opened

This merge was pinned into tulipcc for full-system CI: shorepine/tulipcc#1381

Test it there and merge that PR to move tulipcc onto this AMY.