Skip to content

[ENGINE SOLE OWNER] One native callback → full-fidelity ChannelId cutover (draft) - #266

Merged
masarray merged 163 commits into
mainfrom
milestone/m5-universal-channelid-vst3-lease-migration-20261101
Nov 5, 2026
Merged

masarray merged 163 commits into
mainfrom
milestone/m5-universal-channelid-vst3-lease-migration-20261101

Conversation

@masarray

@masarray masarray commented Oct 31, 2026 •

Copy link
Copy Markdown
Owner

ENGINE SINGLE-OWNER LOCK — 2026-11-03 (CURRENT). This is the only open production ENGINE implementation branch. Former M1 PR #287 was fully absorbed into this branch by verified fast-forward and closed; no second engine stream. Canonical implementation HEAD at reconciliation: 12fc5a5ad49289bd642ca6198ae6e6916facdf12. Binding ENGINE single-owner handoff, runtime ownership map, Windows exact-HEAD architecture + native CI. tests/engine_runtime_reconciliation_test.mjs now runs before JUCE checkout and enforces one native callback, one accepted M4 PCM path, no diagnostic VST fallback, no test-only mixer in application, no fabricated universal capabilities, preserved CH1 vendor FX/Show worker. Production remains M4 10CH+4AUX while ChannelId universal is prepared but not cut over; GUI #278/#285 visual intent, #271 UI adapter, #270 joined FAT. Do not merge PR #266 blindly into current main: branches diverged significantly; reconcile conflicts line-by-line and qualify exact joined SHA. New ChannelId/WIRE/Group/Return features go only into universal engine, not an extended M4. Hardware FAT and UAT are still pending.


ENGINE CONTRACT FOR APPROVED PR #278 V8 CONSOLE (2026-11-02). GUI #271 implements PR #278's one-window Audition-style live mixer; this ENGINE PR alone must supply truthful native I/O, ChannelId, VST3 lease/editor/bypass/reorder, sends/returns, fader/MUTE/solo, LINK vs real serial Group, independent Monitor, Show and measured PCM. Never create a second browser DSP or advertise demo device/plugins. Keep one JUCE callback and fail-closed supportedOps until actually installed. Same-source integration checkpoint 2f8eb8f5046368991dcdc1ce100bbc50601a4f97; latest design V8 eda57f931ae3cc2872579df29cdf4e1486592b64. Any missing name/color/icon accepted metadata requires one shared v8 contract extension via #192, not ENGINE-only new wire. Earlier PR body contains historical checkpoints; PRD v5.2 §9/ARD v5.2 §K take precedence. Owner handoff. Product still NOT UAT.


ENGINE PR #266 — Native AUX FX-send / return-fader production milestone, 1 Nov 2026

HEAD: 40a059785993cc6a4bb593020bf9749de92746f9; exact-head Windows CI #41303154353 pending. This PR remains DRAFT.

Implemented in actual one JUCE callback: PRE/POST AUX send-level changes now alter genuine native PCM. Prior code incorrectly kept static AUX dispatch after an accepted atomic gain edit. The DSP keeps optimized compiled static dispatch until first send automation, then uses target-driven 64-sample gain ramps across callback blocks; same plugin instance, no restart. A second post-VST3 AUX1–4 return fader is now reachable through native SET_AUX_FADER and smooths over 64 samples, even with short ASIO buffers, before MAIN/MONITOR. Existing native SET_SEND_LEVEL handler is deduplicated.

Installed renderer: LiveAudioEngine::audioDeviceIOCallbackWithContext → LiveDryRoutingLab → LiveChannelFxGraph → PreparedMixerBlock (legacy M4 fixed 10 inputs+4 AUX); preserves occupied CH01 vendor/Orbit/editor, CH2–10 hosted VST3 and Show. No fake JS audio or dual callback.

Universal readiness: FALSE; no stable ChannelId callback cutover, dynamic CH11, real serial Group/wet FX Return; canonical universal supportedOps=[]. S0 fixture hash unchanged: 4cbded2c7ada3be511c0e9a09b1949979638eb274c432f89b673a98ce55c7718. No customer UAT or HW crackle certification. Complete ENGINE #192 handoff.


ENGINE current checkpoint — PRD/ARD v5.2, 2026-11-01

HEAD: 157bf6950b6cf3943b60ea45daf6e142a9f14cb7; base main@fe3f8add; Windows native exact-head run #41286591056 in progress — PASS NOT YET VERIFIED.

Implemented runtime DSP: PreparedMixerBlock::process now carries a fixed 64-sample MUTE/PRE-send transition across variable 4/16/32-frame JUCE callback blocks (zero callback allocations/locks). NativeVerifiedSourcePatch now emits complete schema-v2 SOURCE fields without confusing mono physical capture with stereo processing. LiveDryRoutingLab::repatchPhysicalSource already routes SOURCE through actual accepted M4 FX graph/Show/MAIN/Monitor path with prepared-state rollback and vendor lease reuse.

Expanded native tests: accepted one physical pin → CH07 → MAIN PCM, channel MUTE fade and steady silence, native JUCE GainFx instance lifetime unchanged across pin switch 5→7, old graph retirement, sample amplitudes/zero-heap renderer. This in-memory JUCE processor test is not physical vendor VST3 or hardware evidence.

S0 native four-event fixture SHA-256: 4cbded2c7ada3be511c0e9a09b1949979638eb274c432f89b673a98ce55c7718; ARMixLiveChannelFxTests.exe --emit-console-s0-json; no schema changes. GUI/FAT joint exact-head consumer gate still required.

Critical limitation: LiveAudioEngine::audioDeviceIOCallbackWithContext still invokes M4 dryRoutingLab_.process(...processLegacyChannelOne); native v8 universal supportedOps=[] and productionCallbackLive=false. NO ChannelId callback cutover, CH11, serial Group, wet FX Return, or hardware/click/crackle release PASS. Remain DRAFT. Integration owner #270 must not claim product UAT. Owner handoff #192.


PUZZLE A S0 native producer checkpoint — v8 (1 November 2026)

HEAD 7a3c1e082c5b079056aa69af95ee54f23387470e (reconciled current v8 main at fe3f8add). Exact shared 4-event C++-generated fixture SHA-256 4cbded2c7ada3be511c0e9a09b1949979638eb274c432f89b673a98ce55c7718. engine/core/ConsolePuzzleV8Wire.h is included by real JUCE src/ConsoleLiveBridgeV6.h. Existing native test executable ARMixLiveChannelFxTests.exe --emit-console-s0-json now emits EXACT v8 Snapshot → Intent → Rejected ACK → Snapshot contract for FAT #270, with byte-equality checks against the native adapter. Local GNU C++20 generator negative checks + strict v8 four-event schema/ACK validation passed. Windows MSVC/CTest exact-head #41279178032 in progress; no PASS claim until completion.

Capability boundary: installed LiveAudioEngine::audioDeviceIOCallbackWithContext STILL renders original M4 dryRoutingLab_.process(...processLegacyChannelOne). productionCallbackLive=false, no universal ChannelIds, supportedOps=[]; CH11/Group/wet FX Return/Show parity under ChannelId NOT IMPLEMENTED. No hardware crackle claim, no customer UAT. S0 only becomes GREEN after GUI #271 and FAT #270 consume the exact native-generated trace on same SHA, not mock adapter. Engine S1 remains high-priority blocked by that integration gate and occupied CH01 vendor ownership/Show parity.

Owner handoff (issue #192): #192 (comment)


STOP-SHIP / MANDATORY PRD v5.1 + ARD v5.1 / PUZZLE SOCKET V8 — Product PRD · Architecture ARD · Owner delivery packet. ENGINE must deliver native C++ producer fixture, actual enabled pin masks + PCM telemetry, real arMixConsoleIntent dispatch (not blanket REJECTED), installed one-callback M4 fidelity, then real CH11/Group/wet Return/VST/Show; exact SHA and audio callback tests. Promotion BLOCKED pending S0 actual producer→GUI→FAT same-fixture proof, S1 real callback, S2 complete mixer and S3 exact-SHA Windows/Chromium + independent QA or explicitly pending hardware. This draft PR is NOT customer test/release ready.

HARD PUZZLE-SOCKET BLOCKER — PR #279. Owner ENGINE: Producer: live native callback + schemaVersion6, stable decimal ChannelIds, enabledInput/Output pins, RAW/MAIN PCM, real supportedOps and ACK. Replace unconditional reject. Real VST/Show parity. A green independent PR is NOT an integrated product. Follow shared contract, pass node scripts/check_console_puzzle.mjs; final integration branch must pass --release before heavy CI. Report exact SHA/real runtime evidence in #192. No user UAT before finished G4.

ACTIVE DELIVERY LOCK (2026-11-01): ONE PRODUCT RELEASE TRAIN v7 · OWNER WORK PACKET. G0 schema/branch reconciliation → G1 full native callback and First Sound → G2 real channel/VST/Group/FX Return/Monitor/Show → G3 same-SHA quality/packaging/internal FAT → G4 ONE finished product for customer UAT. No half-built First Sound customer testing. See epic #192 and v7 comment in this PR.

Latest ENGINE checkpoint — 4e80c4b (M-U1 migration fidelity)

  • M4→ChannelId occupied external CH1/Orbit owner lease is retained in the accepted PCM publisher with no duplicate native FX dispatch; missing/invalid owner rejects without changing previous accepted audio. CH2–10 and four AUX/MONITOR lease/PCM parity harness preserved.
  • New native fix: physical MONO source into a STEREO processing rack retains M4's equal-power PAN, not stereo balance merely because the VST channel width is stereo. Stereo capture folded to MONO processing is similarly covered. One off-realtime coefficient computation and one lock-free L/R target in callback; six actual legacy-vs-dynamic PCM cases, zero callback heap and stale-epoch mute test.
  • Prior source SHA 8364d53e passed Windows 41266492864 including MSVC/CTest and verified packaging. Current SHA 4e80c4b561e3625b7515f25d3a6fffb2f719961c has Windows 41267321438 in progress; its outcome has not been inferred.
  • Physical FAT can start with verified main@bb6dff6b 10CH+4AUX Windows artifact 41263965810; [ENGINE SOLE OWNER] One native callback → full-fidelity ChannelId cutover (draft) #266 CI deliberately does not upload a PR artifact. Do not describe the M4 artifact as a universal ChannelId shipping executable.
  • Blocking actual M-U1: native LiveAudioEngine callback cutover, vendor/Orbit editor and Show state lifetime under swaps, M4 live mute/solo/source/FX update parity, secondary output pin/meter fidelity. CH11/Group/wet Return/MONITOR still offline on [ENGINE SOLE OWNER] One native callback → full-fidelity ChannelId cutover (draft) #266, no native capability may advertise them yet. Hardware P0 — Crackling with 0 VST3: diagnose WASAPI clock/input/output buffering before engine expansion #15 remains open.

Engine v6 checkpoint — source-grounded (2026-11-01)

Head: 8364d53e387c4a5eb2ffa929bf7bb0791c27efff — DRAFT, NOT PRODUCTION CALLBACK CUTOVER. Native Windows validation: workflow 41266492864 (running at checkpoint; do not infer PASS).

  • Reconciled PR with main@b911c072 as a two-parent commit, retaining main's unpatched FX silent-preparation and FAT-X tests. Main has since moved one docs-only commit; keep reconciling before merge.
  • Added DynamicChannelFxLeaseBank::ExternalLease + M-U1 admission of occupied CH1 external processor owner on the accepted ChannelId graph, rejecting missing/stale/aliased owner and incompatible native CH1 rack. One RT dispatch, no callback refcount or duplicated vendor FX.
  • Added actual in-memory JUCE GainFx PCM golden parity of M4 external vs dynamic publisher, owner survival after old graph retirement, bounded RT allocation and stale-epoch fail-silent test.
  • Existing M5 10CH+4AUX/MAIN/MONITOR/PRE-POST publisher and offline CH11/Group/FX Return tests retained.
  • Exact-head Windows CI first halted at a stale v1 documentation test before C++ compilation (192/193 JS tests passed). Coordinated in issue #192; compatibility test now asserts active v5/v6 architecture and preserves historical v1 content checks.
  • Still not implemented: actual LiveAudioEngine::audioDeviceIOCallbackWithContext switch, original occupied vendor/Orbit host Graph lease construction, realtime controls/Show state wired into one publisher, physical CH11/Group/Return/monitor hardware proof or P0 — Crackling with 0 VST3: diagnose WASAPI clock/input/output buffering before engine expansion #15 crackle FAT. No +CHANNEL capability/LIVE flag is authorized by this PR yet.

Current exact-head milestone (2026-11-01)

M5 + M-U1 full 10CH/4AUX candidate in ONE native PCM+VST3 publisher: UnifiedLegacyProductionAdmission.h consumes the existing legacy graph migration and exact hardware callback pin pairs, transfers real M4 plugin owners (CH1–10+AUX1–4), preserves MAIN/MONITOR PCM and PRE/POST sends, commits an epoch/revision-bound immutable snapshot, and rejects external CH1/Orbit, hotplug, output alias, and armed output without altering previous accepted audio. New native golden comparison uses real JUCE GainFx before/after legacy-owner retirement. Still not yet the actual production callback cutover and not hardware FAT. Exact-head Windows CTest: https://github.com/masarray/arsonmix/actions/runs/41252951831 — pending. The previous M5 core snapshot at f34334a3 passed Windows CTest: https://github.com/masarray/arsonmix/actions/runs/41252379363.

M5 — substantive universal VST3 owner migration toward Studio Console

The real M4 processor state/identity must survive the universal ChannelId graph: karaoke/live Show users must NOT be forced to remove all FX to use a Group (#263 M3E dry-only). This PR replaces that missing seam with a stable native ChannelId → VST3 rack owner conversion, not a fake GUI +BUS or another audio callback.

Changes

  • DynamicChannelFxLeaseBank: one control-thread prepared, immutable owner map across CH1, CH11+, GROUP and FX Return. Retains the actual M4 LiveChannelFxGraph::FxInstance object via its opaque instance token and shared owner only off realtime. Keeps parameter state, reverb tail, bypass, native VST3 editor and device preparation, no catalog reload.
  • Real callback adapter receives ChannelId, looks up immutable rack and calls native JUCE processor(s) once per PCM graph node. No callback refcount, mutex, allocation, VST scan or plugin clone. Reject invalid/duplicate ID, reused simultaneous processor, missing lease, invalid format or nonstereo vendor bus.
  • LiveDryRoutingLab::captureDynamicFxLeases exports M4's actual accepted CH1..10 + four AUX racks guarded by exact device epoch/revision; no effect on live M4 path.
  • Real native PCM CTest: CH1 and CH11 each with native juce::AudioProcessor, then independent Group FX and parallel FX Return POST SEND each with native FX; verifies actual output samples/no doubled dry, Group fader changes, stable M4 opaque token across old graph retirement, zero heap in audio render, stale device fail-silent and duplicate-plugin reject. Uses existing MSVC CTest target to avoid CI fanout.

Integration truth / dependencies

This is the universal native VST3 bridge for M4→M5, not yet the actual replacement of M4's production callback. #263 dry-only pilot stays DRAFT: cannot be merged into a user's normal VST/AUX/MONITOR Show as-is. Next vertical integration must publish this bank as part of the same prepared dynamic PCM snapshot, transfer the special CH1 external FX owner, retain AUX/PRE/POST/MONITOR/secondary output and real Show, then expose dynamic channel strips. Source details & FAT gates in M5 engineering document.

Studio/Fender Console: https://fenderstudiopromanual.fender.com/en/Content/Mixing_Topics/The_Console.htm
Waves eMotion LV1: https://www.waves.com/emotion-lv1-tutorial-4-2-mixer-channels-groups-aux
Related #192 #263 #265 #15.

Merge gate: exact-head Windows native MSVC + CTest pass, no device/GUI regression. Avoid unnecessary new workflows or duplicate CI.

Coordination #397 — workstream ownership

SOLE ENGINE PCM producer; continue engine-owned files. Hand off Main.cpp, CMake and native bridge edits to INTEGRATOR #397.

Copy link
Copy Markdown
Owner Author

M5 exact-head update bbce5818608a: prepared native PCM now retains the exact control-owned M4 VST3 lease bank in the same RealtimeRoutePublisher snapshot, with epoch+revision admission, stable ChannelId dispatch, never a second callback or double FX. Test fixture uses real JUCE gain processors on CH1, CH11, GROUP and parallel FX RETURN, checks no callback heap, Group gain, old owner destruction, invalid pointer/epoch/revision rejection and same instance after next graph re-publication. This is native engine integration; NOT yet user-facing CH11 GUI or physical FAT, and does not supersede M4 or FAST3. Windows exact-head CI: https://github.com/masarray/arsonmix/actions/runs/41251719542. PR #263 dry-only pilot must stay draft until full Show+MONITOR/AUX/CH1 external plugin owner and target device outputs migrate.

Copy link
Copy Markdown
Owner Author

M-U1 integrated preparation update at 2bbf9f08: new UnifiedLegacyProductionAdmission reuses LegacyGraphMigration to build 10CH+4AUX with actual PRE/POST sends, independent MAIN/MONITOR and exact verified hardware callback output pairs; the same accepted native snapshot retains 14 M4 VST3 lane leases with epoch/revision. CTest now A/B compares real JUCE M4 FX PCM before→after old graph destruction for MAIN L/R and MONITOR L/R and checks stale epoch, output alias, output armed, CH1 external legacy owner rejection and old PCM surviving rejected changes. Earlier core M5 revision f34334a3 has all Windows CTest SUCCESS (run 41252379363). Exact-head combined M5/M-U1 run 41252951831 is still in progress; do not merge or claim hardware FAT until green. NOTE: this is the control-thread unified prepublication milestone, not yet runtime full M-U1 callback cutover nor universal Channel11 UI. PRD/ARD v5 compliance is partial. See docs/M5_UNIVERSAL_CHANNELID_VST3_LEASE_MIGRATION_20261101.md.

@masarray masarray changed the title M5: preserve native M4 VST3 instance/state across dynamic ChannelId CH11, Group & FX Return M-U1: atomic 10CH+4AUX/MONITOR + M4 VST3 lease migration into ChannelId PCM Oct 31, 2026
This was referenced Oct 31, 2026

masarray commented Nov 4, 2026

Copy link
Copy Markdown
Owner Author

M3-B native control ABI handoff — REAL code, single INTEGRATOR PR #402

Commit 4163bf3faebf was added to existing PR #402, without starting a second PR/workflow or changing ENGINE source.

  • New integration/ApprovedNativeControlIntent.h: fail-closed conversion of approved /preview/ v8 SET_FADER / SET_PAN / SET_MUTE / SET_SOLO to the actual ENGINE [ENGINE SOLE OWNER] One native callback → full-fidelity ChannelId cutover (draft) #266 native NATIVE_SET_CHANNEL_MIX / NATIVE_SET_CHANNEL_MUTE_SOLO; GROUP/FX_RETURN fader/pan maps to NATIVE_SET_BUS_MIX only when truly supported.
  • Validates canonical stable decimal-string ChannelId, exact device epoch + graph/roster revision, installed native bus revision, exact target/role, both public supportedOps and native supportedOps, accepted fader/pan/mute/solo readback parity, physical [-60,+10] dB bounds, finite gain; preserves sibling PAN/GAIN or MUTE/SOLO and requestId. Rejects G0, missing/stale callback, unknown IDs, false capability, malformed/unsupported operations without touching graph.
  • Extends existing M3BNativeV8StableIdReadbackAndFailClosedProjection CTest with valid fader/pan/mute/solo conversions, zero-side-effect failure cases and stale/missing native capability. Uses existing Windows runner, no new workflow.
  • IMPORTANT: translator itself never invokes audio, publishes a LIVE flag or fakes an ACK. src/Main.cpp G0 still rejects all universal intents. The actual producer must first deliver admitted PCM and this translator must be wired to the same native dispatcher; exact-head Windows CTest pending.

ENGINE #266 required next

Ensure original CH1–10 + dynamic INPUT ChannelIds actually mutate the one installed PCM callback via setNativeInputMix / setNativeInputMuteSolo and produce updated accepted readback, while preserving CH1 vendor/Show/AUX. Only then advertise corresponding public graph.capabilities.supportedOps. The current #266 nativeBus names and /preview/ public names are intentionally different; never advertise SET_FADER/PAN/MUTE/SOLO merely because NATIVE_SET_CHANNEL_MIX exists. Only INTEGRATOR owns Main.cpp/CMake/approved bridge stitching, and #15 hardware FAT remains PENDING.

This keeps M3 progressing within one owner branch and avoids claiming First Real Sound from C++ fixtures alone.

masarray commented Nov 4, 2026

Copy link
Copy Markdown
Owner Author

ENGINE #266 exact code blocker for approved GUI First Sound (M3):

At current #266 HEAD 066d7b05444a, src/LiveDryRoutingLab.h::setNativeInputMix and setNativeInputMuteSolo both explicitly execute for(const auto original:nativeAccepted_.graph.ids.inputs) if(original==id)return false; even when the native ChannelId callback is installed and active. This blocks native GUI fader/PAN/MUTE/SOLO on the original CH1–10 IDs. The engine's nativeBus.supportedOps nevertheless lists NATIVE_SET_CHANNEL_MIX and NATIVE_SET_CHANNEL_MUTE_SOLO globally.

Required ENGINE-owned meaningful patch: make original CH1–10 ChannelIds and dynamic inputs share one REAL accepted fader/mute/pan control owner, preserving original CH1 VST3/Orbit instance, AUX/MONITOR/Show state, device epoch, 64-sample control ramp and single callback; reject stale IDs and do not advertise public SET_FADER/PAN/MUTE/SOLO before exact PCM readback is proven for original and dynamic IDs. One native CTest must exercise original CH1 and CH2 + a new dynamic input, proving actual sample changes at -12dB, PAN, MUTE, SOLO and rollback after stale command on SAME source/graph owner.

INTEGRATOR adapter is now under #402 / commit 4163bf3, awaiting exact-head CI; it only translates public v8 operations when explicitly admitted. Please do not rewrite Main.cpp/CMake or alternate GUI in #266—provide exact callback/intent ABI and PCM results as a handoff on #397. Hardware issue #15 remains stop-ship.

masarray commented Nov 4, 2026

Copy link
Copy Markdown
Owner Author

M4-B/C precise ENGINE handoff from PR #404 (separate INTEGRATOR control-plane lane, no PCM/GUI edits):

Current actual source blockers: src/VstBusLayout.h::prepareStereoEffectLayout deactivates every auxiliary input bus, so VST3 external sidechain cannot work; #266 src/ConsoleLiveBridgeV6.h WIRE decoder rejects tap != POST_FADER; current public readbackProvesIntent proves connection existence but not PRE/POST/gain, and graph-aware PDC for dual dry/wet + detector paths is not qualified.

ENGINE-owned native extension, same callback/graph (no fork):

  1. During OFF-RT VST3 preparation negotiate plugin's actual main stereo + optional supported aux detector bus. Store stable plugin instance lease and enabled bus index/ch count; make detector input a separate, non-audible signal kind, not an ordinary WIRE send. If unsupported, return explicit unsupported without altering any audible route.
  2. Accept and publish PRE_FADER/POST_FADER sends, per-send gain, and graph revisions atomically. Emit actual tap/gain in accepted edge readback for exact controller ACK proof. Validate cycles on the combined DSP scheduling/dependency DAG; reject double-dry/group feedback.
  3. Declare per-plugin latencySamples + actual graph-aware PDC offsets and monitor-mode policy. Nonzero latency without compensation must remain blocked. Do not silently degrade live low-latency monitor into a delayed studio path.
  4. Deterministic golden tests: Kick → Bass compressor detector only (no kick audio injected in Bass output), external detector toggled vs off; mono/stereo bus layouts, plugins with no sidechain, PRE/POST differentiation, detector/FX latency aligned impulse, no allocations/locks/heap scanning in callback, original CH1+Show/VST3 preserved.
  5. Only after true DSP+readback is installed advertise CONNECT_PLUGIN_SIDECHAIN, SET_SEND_TAP, SET_SEND_GAIN and native PDC capability to the approved v8 GUI contract. Do not mark product LIVE before FAT P0 — Crackling with 0 VST3: diagnose WASAPI clock/input/output buffering before engine expansion #15.

INTEGRATOR #404 currently implements fail-closed operator-intent planning for Route Only and post-fader send and rejects unsupported sidechain/PDC/pre-fader/gain. This leaves #266 as sole PCM owner; GUI owner later maps one/two-click gestures in existing /preview/, no second renderer. CI evidence on #404 will follow after exact-head completed.

masarray commented Nov 4, 2026 •

Copy link
Copy Markdown
Owner Author

ENGINE E6 — Original CH1–10 + Dynamic ChannelId PCM control unification (code committed)

Sole-engine branch/PR: #266 milestone/m5-universal-channelid-vst3-lease-migration-20261101
Exact HEAD: 59d1ef66d62f, advanced from 066d7b05 using an expected-SHA ref lease. ENGINE-owned files only: src/LiveDryRoutingLab.h, tests/dynamic_accepted_session_test.cpp, tests/engine_runtime_reconciliation_test.mjs. No other PR, CMake, Main.cpp, approved GUI, workflow or callback was modified.

Production source repair: removed explicit rejection of original nativeAccepted_.graph.ids.inputs (CH1–10) from setNativeInputMix and setNativeInputMuteSolo. Both setters now accept all validated INPUT ChannelIds on the SAME installed native publisher, with device epoch, native PCM graph revision, finite parameters, role, and accepted roster readback fencing. The legacy setChannelParameter no longer mutates the inactive M4 publisher before attempting a native control; it publishes to ChannelId first and returns failure without false ACK on rejection. Old ordinal operations reject after structural graph cutover. A changed native control forces fail-silent on callback revocation and on rejected native AUX/mirror updates, never reverting to stale legacy PCM. The latter is hardened in the follow-up commit.

Exact current HEAD: 59d1ef66d62f (two ENGINE-only commits; original 086300c9, hardening 59d1ef66).

Native PCM tests: extends existing M3DAtomicStudioBusFxChannelSessionRollback CTest with 10 original inputs + newly added CH11 sharing ONE accepted session: audible CH1 -12dB, CH2 PAN, CH1 MUTE, CH11 SOLO + PAN/fader, stale epoch/revision/identity rollback, no revision drift. Adds cheap ENGINE architecture contract ensuring legacy-first dual control cannot regress. No extra CI workflow.

Qualification: source committed; exact-head Windows CI on 59d1ef66d62f NOT yet observed (no check runs at last query). No physical hardware proof, no first-sound or product LIVE claim; #15 remains STOP-SHIP. DO NOT merge entire historic #266 branch over current main. INTEGRATOR #402 consumes ONLY actual admitted native PCM/readback and wires one approved /preview/ dispatcher, never duplicates this control path.

masarray commented Nov 4, 2026

Copy link
Copy Markdown
Owner Author

M3-D exact ENGINE → INTEGRATOR handoff (2026-11-05 WIB)

Current main@6b51caf07dd4ae04047c04717adcb3f53e3f673f includes #404 routing contract and #405 always-on coordination policy. Existing INTEGRATOR #402 HEAD cb7a44eb now projects native LiveAudioEngine::status().dryRoutingLab.nativeBus into the approved /preview/ graph only after positive nativeChannelIdPcmActive, matching device/routing epoch, actual direct I/O/ports, no fault and accepted graph. The M4 upstream double-dry and optional edge-primary review bugs are also patched in #402; joined Windows/Chromium exact-head CI underway.

To ENGINE owner #266: your E6 59d1ef66 original CH1–10 + CH11 native parameter unification is useful, but its code remains on a highly diverged historical branch. Please provide a source-only current-main extraction of the ONE installed native PCM+VST3 producer, original CH1 vendor/Orbit ownership and Show/AUX/MONITOR, retaining LiveDryRoutingLab::nativeBusStatus(), nativeChannelIdPcmActive, accepted graph/roster/device epoch and stable ID. Do not merge historic web/console-native.*, src/Main.cpp, CMakeLists.txt or old UI/workflow. Qualify exact-head Windows/MSVC/CTest on the actual source branch, including original CH1/CH2 + CH11 gain -12 dB, PAN, MUTE/SOLO, WIRE ROUTE_ONLY/POST, actual JUCE VST3 and negative stale commands; no fake ACK or double callback.

INTEGRATOR next after real producer handoff: only then connect existing integration/ApprovedNativeControlIntent.h and src/Main.cpp accepted mutating dispatcher, advertise SET_FADER/PAN/MUTE/SOLO/WIRE per ACTUAL supported native operations with exact ACK→later accepted readback. Until then public supportedOps=[] and G0 rejection must remain intact.

Hardware issue #15 FAT PENDING. This is a targeted API/source handoff, not authorization to overwrite another workstream.

masarray commented Nov 4, 2026

Copy link
Copy Markdown
Owner Author

M3-E FAT consumer handoff (FAT PR #406, head 4ca9f24) — ENGINE remains sole native PCM owner.

The independent exact-SHA Windows field kit and 44.1/48-kHz native REPORT collector are now implemented under FAT ownership; all 11 new Node tests pass, hardware #15 remains pending. To qualify actual first-sound, please provide one native callback/source-owned PCM evidence packet, not browser samples: sourceCommit, installed deviceEpoch, accepted graphRevision/rosterRevision, canonical ChannelId/EdgeId for original CH1–10 plus dynamic CH11, actual input pin→MAIN L/R PCM, fader/PAN/MUTE/SOLO numerical deltas and mute silence, GROUP/FX-return/monitor isolation, plugin VST3 lease/bypass/real audio and safe hotplug reconnect. The FAT work must consume the SAME publisher as #402 readback, with no second callback/JS DSP. Avoid treating simulated CTest vectors as physical DAC evidence; physical 0FX crackling stop-ship #15 remains open.

masarray commented Nov 4, 2026

Copy link
Copy Markdown
Owner Author

M3-E INTEGRATOR active execution — protected main + exact owner handoffs (2026-11-05 WIB)

WORKSTREAM: INTEGRATOR. Baseline main@4a40e7132871ace9fb7ec4b8c0af07aa42630a39 now includes merged #407 (CRLF-safe Node policy + precise Product Qualification), #405 coordination and #404 routing safety. The sole owned #402 has been reconciled on top of this protected main by expected-SHA fast-forward merge commit c99f60233a1a. Only 9 INTEGRATOR paths changed: approved native readback/translator/MAIN seam, anti-double-dry, tests, CMake. No GUI/DSP/FAT/workflow edits. Branch is DRAFT pending exact-head Product Qualification (new PR UI job + reusable native Windows build), review, then selective merge. Prior joined Windows+Chromium green at cb7a44eb; earlier failure at a2a025d was cross-platform #405 LF test, fixed by owner #407. DO NOT run parallel duplicate native CI.

ENGINE #266 owner handoff REQUIRED for actual FIRST SOUND: Branch still 59d1ef66d62fb66f030feca5ede8e47b77ca959d, 157 ahead / 115 behind main and not raw-mergeable. Source-only reconcile the ONE LiveAudioEngine/LiveDryRoutingLab ChannelId PCM callback + public setNativeInputMix/setNativeInputMuteSolo and WIRE authority, original CH1–10 + dynamic CH11, native VST3 leases and original CH1/Orbit/AUX/MONITOR/Show parity, then exact-head Windows/CTest. Put only ENGINE-owned paths in your changes, hand src/Main.cpp/CMake/approved bridge changes to #402 through #397. No historical GUI/workflow/parallel callback. Without installed producer #402 MUST keep public supportedOps=[] and arMixConsoleIntent REJECTED (not fake LIVE).

FAT #406 owner handoff: reconcile FAT-only ZIP/collector/evidence kit from 4ca9f242 onto corrected main; verify exact software SHA with Windows real PCM/VST3 and 44.1/48k 0FX/1FX 30min matched device runs. Physical crackle #15 remains STOP-SHIP, report UNKNOWN hardware counters honestly.

Owner / merger must never replace frozen approved https://masarray.github.io/arsonmix/preview/ skin. End goal: one accepted EXE commit, one native callback, ACK→later accepted PCM snapshot, measured evidence.

masarray commented Nov 5, 2026

Copy link
Copy Markdown
Owner Author

M3-F CRITICAL PRODUCER HANDOFF — native readback NOW ON MAIN

Accepted main now f7d2993e4973cf8944cedac570dc87451beed43f via merged INTEGRATOR #402, exact PR head 039b8746215349517556616343ad48dd839a8cc8, Product Qualification #41928672962 SUCCESS (owner admission, Chromium, Windows JUCE/MSVC/CTest, Product Qualification); Codex latest-head review no major issues.

P0 actual source blocker verified: accepted main src/LiveAudioEngine.h and src/LiveDryRoutingLab.h contain NO setNativeInputMix, setNativeInputMuteSolo, nativeChannelIdPcmActive, or nativeBus output. Those exist on YOUR historical #266@59d1ef66d62fb66f030feca5ede8e47b77ca959d, which is 157 ahead / 117 behind new main. Main's src/Main.cpp still intentionally rejects arMixConsoleIntent and integration/ApprovedNativeRuntimeSnapshot.h keeps public supportedOps=[]. So do not claim First Real Sound from GUI/CTest and do not raw-merge #266.

Next ENGINE-only execution required NOW: selectively reconcile the same installed native ChannelId callback publisher / PCM graph into current main, bringing engine-owned engine/core/**, src/LiveAudioEngine.h, src/LiveDryRoutingLab.h (and any minimal ENGINE-owned helpers actually needed) while preserving approved GUI, M4/INTEGRATOR seam, vendor CH1/Orbit VST3, original AUX/MONITOR/Show and their real source ownership. DO NOT alter src/Main.cpp, CMakeLists.txt, integration/**, web/**, .github/workflows/**, FAT. Existing #266 must be the sole owner branch/workstream. Avoid rebasing raw old integrated code over new main.

API requirements from existing #266 source (verified):

  • LiveAudioEngine::setNativeInputMix(uint64_t epoch,uint64_t pcmRevision,ChannelId id,float gain,float pan) returning actual bool accepted;
  • LiveAudioEngine::setNativeInputMuteSolo(uint64_t epoch,uint64_t pcmRevision,ChannelId id,bool mute,bool solo);
  • LiveAudioEngine::setNativeBusMix(...) where supported;
  • LiveAudioEngine::status().dryRoutingLab.nativeBus and nativeChannelIdPcmActive only when actual physical callback owns accepted graph with native epoch/graphRevision/rosterRevision, accepted channels & supportedOps.

Proof on ONE accepted JUCE PCM callback: source→MAIN delta for original CH1 -12 dB, CH2 PAN, CH1 MUTE, CH11 SOLO/gain/PAN; stale epoch/revision/ChannelId reject with no revision drift; original VST3 leases and Show/AUX/MONITOR preserved; mute silence and physical output mask correctness; no second processor, legacy fallback, fake meter or ACK. Post exact source SHA + FULL Windows/JUCE/CTest qualification URL here and in #397, and explicit API contract handoff to INTEGRATOR. Hardware #15 remains STOP-SHIP for stable beta.

Integration will resume immediately on genuine producer handoff using already merged integration/ApprovedNativeControlIntent.h and single src/Main.cpp WebView2 authority. Until then current fail-closed intent path is correct.

masarray commented Nov 5, 2026

Copy link
Copy Markdown
Owner Author

ENGINE owner handoff (FAT #406 independent; no production edits from this thread): exact #266 HEAD e15aefec648f7450e6f2d2afa1c6397dd5f68395 fails Windows build Actions #41941850022, Build executable and native RT tests. C1083 at tests/live_channel_fx_graph_test.cpp:8: #include "../src/ConsoleLiveBridgeV6.h" missing from the selectively reconciled source tree. This is a real missing test/bridge dependency, not FAT, #402 or GUI. Please resolve with ENGINE-owned explicit dependency/source or a genuine current-producer test adaptation; do not raw-merge obsolete CMake/Main/web/GUI. INTEGRATOR M3F #408 is based on exact ENGINE commit e15aefec, so it inherits this compile blocker until ENGINE owner reconciles it. No false PCM or release PASS.

@masarray
masarray marked this pull request as ready for review November 5, 2026 01:05
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 5, 2026 •

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-05T01:09:20.421428Z efe8f49 Draft marked ready
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: efe8f490d5

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment thread src/LiveDryRoutingLab.h
Comment on lines +1791 to +1792
if(nativeStructureLive_ || nativeControlDirty_)
topologyFault_.store(true,std::memory_order_release);

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Preserve native controls through original CH1 FX handoffs

After an admitted session receives any fader, PAN, MUTE or SOLO edit, nativeControlDirty_ becomes true. Adding or removing an original CH1 insert then calls publishFxEdit(), which invokes this method and permanently sets topologyFault_, even with outputs disarmed and no dynamic buses. The callback subsequently silences the entire mixer; finalizeGraphSwitch() cannot re-admit it because hasExplicitRoute() requires ready(), and APPLY also rejects the topology fault. Preserve or synchronize the accepted controls through the FX handoff instead of turning a normal rack edit into a fault requiring device reinitialization.

AGENTS.md reference: AGENTS.md:L30-L30

Useful? React with 👍 / 👎.

Comment thread src/LiveDryRoutingLab.h
Comment on lines +876 to +878
if(!nativePublisher_.setParallelSendGain(
expectedEpoch,nativeAccepted_.pcmRevision,
id,target,false,gain))rejectedNativeMirror();

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Retain zero-level MONITOR sends for live gain changes

When a MONITOR send was zero at admission, LegacyGraphMigration::compile() omits its edge, so this scalar update cannot succeed: setParallelSendGain() only updates existing edges. I reproduced a successful migration with a zero MONITOR gain followed by a rejected zero-to-positive update. If any native channel control has already been edited, rejectedNativeMirror() sets topologyFault_, silencing the entire mix while this method still returns success. Retain zero-gain MONITOR edges during admission, or perform an accepted structural transaction when enabling one, rather than acknowledging a failed mirror.

AGENTS.md reference: AGENTS.md:L28-L29

Useful? React with 👍 / 👎.

masarray commented Nov 5, 2026

Copy link
Copy Markdown
Owner Author

P0 AUDIO USER ESCALATION (ENGINE owner) — direct Windows screenshot reports no sound.

Operator Windows screenshot 2026-11-05 08:12 WIB: top native AR MIX Studio, bottom "Stopped | NATIVE G0 - perangkat aktual terbaca; ChannelId PCM/WIRE belum diterima. Edit audio diblokir", right "SIMULATED METERS - NO REAL AUDIO". Operator explicitly reports NO AUDIO. This is a REAL PRODUCT FAILURE, not an optional FAT paperwork item.

Main 5f4915680b7a1406c8565620041e517f568a5d7d intentionally lacks your producer (all ChannelId intents REJECTED); a green exact-main package is a G0-only diagnostic, not first sound. Your current source efe8f490d5f49906b7a982a443d3123c6893a96a is not merged. Latest Codex review exposes two unresolved P1 silent-mixer hazards in src/LiveDryRoutingLab.h: an accepted fader change followed by CH1 VST3 FX handoff can set permanent topologyFault, and a zero-gain MONITOR send changing to positive can reject and fault the entire native graph. Please fix these in your existing #266 branch, with real native PCM/CTest regression, without touching Main.cpp/GUI/FAT. Then hand your exact verified producer HEAD to #408. Need actual source→MAIN samples, dynamic CH11, plugin/Show parity, not a UI mock. Hardware #15 remains pending.

@masarray
masarray merged commit 7ee0383 into main Nov 5, 2026
6 of 8 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant