midi_cc can name an AMY parameter directly (#1175, part 1)

[dpwe]

Part 1 of #1175: the direct-parameter form of midi_cc, which midi_cc_output will mirror in a follow-up.

A mapping payload that starts with a digit is a parameter list instead of a wire command template (the forms can't collide, since no wire command starts with a digit):

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

amy.send(synth=1, midi_cc='74,1,100,6400,0,%d' % amy.PARAM_FILTER_FREQ)   # osc 0 of every voice
amy.send(synth=1, midi_cc='71,0,0.7,8,0,%d,0,%d,1' % (amy.PARAM_RESONANCE, amy.PARAM_RESONANCE))  # oscs 0 and 1

How it's done

• amy.PARAM_* constants for every enum params entry. scripts/gen_param_constants.py (run from the existing constants.py rule) compiles a tiny program against amy.h, so the computed enum values are the compiler's rather than parsed from comments, which are stale (NO_PARAM // 210 is really 234). The PARAM_ prefix keeps FREQ, BUS, MODE etc. from shadowing anything in amy.. It uses $(HOSTCC) (default cc), not $(CC), so a cross build still works. amy_api.generated.js is regenerated to match.
• A direct mapping doesn't build deltas itself. It sets the matching amy_event field (a param_fields table in midi_mappings.c) and sends the event to the synth down the ordinary path. So:
  • values are in amy.send() units (Hz for filter_freq, linear drive, …), with no second copy of the logfreq/log2/logdrive conversions;
  • they reach every voice, at the voice-relative OSC;
  • bus params (PARAM_VOLUME, PARAM_REVERB_LEVEL, PARAM_BUS_DIST_*, …) go to the synth's bus.
• Only params with a plain event field can be mapped. MIDI_NOTE/VELOCITY, osc references (CHAINED_OSC, MOD_SOURCE_*, ALGO_SOURCE_*), breakpoints, resets, LATENCY and BUS are refused. So is MODE, because event_addresses_oscs() doesn't count mode, so a synth-addressed mode change is dropped today (template form too; left alone here).
• Integer params are rounded, not truncated, so a CC scaled to 0..6 steps through filter types cleanly.
• A refused payload no longer deletes the mapping already on that CC: it's parsed before the old one is freed.
• The payload text is kept as the template string, so the state dump and restore round-trip unchanged.
• Also fixed: midi_mapping_init didn't check its malloc_caps.

Tests

• New tests/test_midi_cc_param.c, added to make ctest. It covers:
  • linear and log transforms
  • every voice, and only the named osc
  • multiple P,OSC targets
  • integer rounding
  • a refused payload leaving the existing mapping intact
  • a state dump fed back in as a restore
  • template and direct forms side by side
• make ctest: all pass.
• make test: 90 pass / 43 "fail" at ~-99 dB error. That result is identical on unmodified main in this environment, so it's float noise, not this change.

Follow-up (part 2): midi_cc_output (iC) with the same C,L,N,X,O,P[,OSC] form, hooked beside note_output_handle_event in patches_event_has_voices. Decisions:

• channel follows the synth's note_output MIDI channel, otherwise the synth number;
• timing is event time, like note_output;
• stacking several mappings on one CC is deferred.

It will only send when the target queue is the live delta queue. note_output looks like it has the same issue when events are stored into a patch; that's worth a separate small 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 @ 44bf70d	this PR	Δ
1	981	991	+10
2	1136	1152	+16
3	1706	1723	+17
4	1875	1896	+21
5	2465	2498	+33
6	2600	2631	+31

Full chord settled render μs: 2633 (was 2600, Δ +1.3%) (peak 2640, 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#1379

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