audiodynamics.Dynamics has no deinit() on the native builds, so no audioif-tier dynamics class can release one

[bdbarnett]

Stakes. Every audioeffects class of the audioif tier that builds a
audiodynamics.Dynamics — Compressor, Limiter, Expander, NoiseGate,
DeEsser, TransientShaper, MultibandCompressor, DynamicEQ — cannot
release it on a board. The class gate's Tier 1 asks that deinit() release
every node the class built; on the two native builds
_component.Component.deinit() walks the class's node list, looks for
deinit with getattr(node, "deinit", None), and finds nothing to call. A
host that builds and tears down effects in a session leaks every Dynamics
it ever made. audioroute.Splitter is the same shape and is #58; this is its
sibling and the wider one — Limiter alone builds two per instance.

What the surface is. Measured this session, audiocomponents at
AUDIOIF_PIN 2f6cbc3, 2026-09-07:

$ .venv/bin/python -c "import audiodynamics; n = audiodynamics.Dynamics(
      audiodynamics.DYN_LIMIT, sample_rate=48000, channel_count=2)
  print(hasattr(n, 'deinit'), sorted(x for x in dir(n) if not x.startswith('_')))"
True  ['bits_per_sample', 'channel_count', 'deinit', 'gain_reduction_db',
       'key', 'max_buffer_length', 'play', 'playing', 'sample_rate',
       'samples_signed', 'set', 'single_buffer', 'stop']

$ MICROPYPATH=lib:../cmods/micropython/lib ../cmods/bin/micropython <same>
False ['bits_per_sample', 'channel_count', 'gain_reduction_db', 'key',
       'play', 'sample_rate', 'set']

$ MICROPYPATH=lib:../cmods/micropython/lib ../cmods/bin/circuitpython-effects <same>
False ['bits_per_sample', 'channel_count', 'gain_reduction_db', 'key',
       'play', 'sample_rate', 'set']

So the CPython shim has deinit, stop, playing, max_buffer_length and
single_buffer; the two native builds have none of them. The gap is the
native binding's, not the shared C's — audioif_dynamics.c has the state to
free.

What it costs the gate. docs/effects/Limiter-evidence.md §2 records six
partial cells for exactly this: the class's own surface still closes and
the borrowed source still renders, both measured, but "every node the class
built was released" is unmeasured on MicroPython and on
circuitpython-effects, because there is nothing to observe. Every
audioif-tier dynamics pack in Phase 2 carries the same six cells, and the
gate audit rules them Y* with the cause below the class
(audiocomponents/docs/effects-phase2-gate-audit.md §3, G2).

Ask. deinit() on the native audiodynamics.Dynamics binding, with the
deinitialised guard #59 asks for generally. stop() and playing would
close the rest of the shim/native divergence, but deinit is the one a gate
item turns on.

[bdbarnett]

Fixed in 2742489, together with #58 — the deinit and the guard had to land in one change.

The commit's Closes #58, #59, #60, #63 only closed the first: GitHub needs the keyword before each number. Closing the rest by hand.

What landed. deinit() on all thirteen node types that had none (every type audioif wrote itself, as against every type it ported from CircuitPython), and the deinitialised guard on audiosample_get_buffer / audiosample_reset_buffer — the funnel every pull and rewind in the palette passes through, whoever called it.

The crash surface was ten node types, not one. audiosample_mark_deinit() sets channel_count = 0 and a node's get_buffer divides by the channel count, so on the pre-change build pulling a released node dumped core for ten of the thirteen types that already had deinit():

node          before                  after
Mixer         SIGFPE  (core dumped)   raises
Chorus        SIGFPE  (core dumped)   raises
Echo          SIGFPE  (core dumped)   raises
MultiTapDelay SIGFPE  (core dumped)   raises
PitchShift    SIGFPE  (core dumped)   raises
Filter        SIGFPE  (core dumped)   raises
Distortion    SIGFPE  (core dumped)   raises
Phaser        SIGFPE  (core dumped)   raises
Freeverb      SIGFPE  (core dumped)   raises
SpeedChanger  SIGSEGV (core dumped)   raises
RawSample     returned stale audio    raises
Synthesizer   returned stale audio    raises

That is why the two halves could not be split: having deinit() without the guard was strictly more dangerous than having neither, and adding the twelve missing deinit()s on their own would have turned twelve safe types into twelve core dumps.

Gate. tests/parity/deinit_surface_probe.py, over all 25 node types, running unchanged on all three targets. --fault skips the deinit() call and must exit nonzero; CI runs the fault first and fails if it passes. The native half runs in clean-build.yml on a real MicroPython build.

No rendered bytes move: all 23 DSP probes byte-identical on MicroPython before and after.

[bdbarnett]

Correcting the record: this was closed on two of three targets, not three.

The commit closing it described tests/parity/deinit_surface_probe.py as "running unchanged on all three targets". The probe does run on all three. I only ran it on two — the CPython target and desktop MicroPython — and said so as though I had run three. Running it on the patched CircuitPython build, once that build existed later the same day, fails:

$ MICROPYPATH=. .../build-coverage/micropython tests/parity/deinit_surface_probe.py
  audiobiquad.Biquad has no deinit()
  audiodynamics.Dynamics has no deinit()
  audioroute.Splitter has no deinit()
  ...
exit=1

Measured: 0 of 13 type files under src/circuitpython_spike/shared-bindings/ contain MP_QSTR_deinit. audioif keeps a second, hand-written binding per module for CircuitPython, and it never received this work.

The MicroPython and CPython halves stand and are gated. The CircuitPython half is tracked in audioif#75, together with the two other instances of the same lag found the same day and the check that should have caught all three. Leaving this issue closed and the remaining work in #75 rather than reopening four issues for one cause.