Repository navigation
Description
CURRENT CHECKPOINT — 2026-11-08 (main
c57720ea25585c861af89802e1365ffef7d03d17)Single public/native renderer implemented and protected by current main. Merged #418, #440, #442 and later approved GUI/bridge changes produce ONE
/preview/display with generated Pages/native WebView assets; legacy preview URLs are redirect-only. #393's 106-file competing physical-source migration was archived rather than bulk merged. Still open for acceptance: actual native visual/interaction parity across operator states, authoritative PCM/FX/WIRE readback on hardware, and source maintainability improvements without a second renderer. Retired #270/#397 must not re-impose multi-agent coordination.
P0 — Stitch the only approved /preview/ GUI into native JUCE
FINAL USER DECISION 2026-11-04: https://masarray.github.io/arsonmix/preview/ is the ONLY approved GUI. The earlier /preview-v2/ claim was wrong and is revoked. Only one public preview and one Windows GUI may exist. Do NOT design a new native screen, substitute old web/index.html/web/mixer-layout.js/web/console-native.*, ship bare /preview-v2/, or fork the GUI. The engine follows the finished GUI.
Provenance: preserve exactly what user approved
The currently-approved /preview/ previously loaded ../preview-buttons/index.html in an iframe. It is assembled by scripts/build_pages.py from internal site/preview-v2/index.html plus all its WIRE/FX/fader/meter styles/assets, but specifically overrides hardware button CSS/JS/PNG with site/preview-buttons/assets/mixer-control-keys.css, mixer-control-keys.js, mute-solo-pfl-56x37.png. Bare /preview-v2/ is visually wrong; do not bundle it into Windows. Internal directories are build inputs only, not alternative product UIs.
Deliver one publication and one native GUI
- Pages source consolidation: publish generated approved HTML and assets DIRECTLY at
/preview/(no competing iframe or alternate page); redirect any old/preview-v2/,/preview-buttons/, archived Vxx public routes to that URL, with no second renderable GUI. Add fail-closed test comparing generated /preview/ HTML to current source with exact approved PNG-button revisions and verifying redirect-only legacy URLs. - Native WebView2 exact parity: deterministic reuse of the exact same assembled /preview/ artifact and CSS/JS/SVG/PNG assets, with SHA256 manifest and CMake BinaryData+ResourceProvider mapping. Actual Windows EXE screenshot/gesture parity at 1366x768, 1680x900, 1920x1080 across density, selection, WIRE fold, FX rack and bank states. A static HTML mock is not native proof.
- Thin accepted native adapter: keep renderer/layout; replace simulated authority with real v8 schemaVersion=6
arMixConsoleSubscribe/Snapshot/Intent/Ack, canonical decimal-string uint64 ChannelId/EdgeId/epochs/revisions, matching ACK then later accepted same-revision snapshot, rejection/stale rollback. One accepted graph powers WIRE/OUT/MAIN/Scenes/Undo, meters/FX/faders; no fake LIVE. - One reliable engine: ENGINE PR [ENGINE SOLE OWNER] One native callback → full-fidelity ChannelId cutover (draft) #266 & WIRE WIRE C2: public v8 ACK edits real native PCM and matches accepted EdgeId snapshot #387 keep ONE installed JUCE PCM graph/callback, ONE VST3 lease/editor and hardware pin truth, off-thread graph prep and RT-safe publish; no new GUI. Preserve CH1 vendor, 10CH+4AUX, Show and physical hardware during migration.
- FAT [INTEGRATION/FAT v7 G0] Canonical JUCE WebView ACK/PCM evidence and exact-SHA release gate #270: one exact Windows SHA with approved GUI bundle manifest/hardware button checks + actual screenshot parity and native PCM/VST3/ASIO/WASAPI/0FX-crackle P0 — Crackling with 0 VST3: diagnose WASAPI clock/input/output buffering before engine expansion #15 evidence. No premature UAT. Physical hardware unavailable = pending, not product ready.
Definition of done: installed Windows mixer displays exactly /preview/ including approved PNG buttons; all controls accurately operate real native PCM/VST3, no second GUI or DSP, with one combined exact SHA and verified FAT.
See root AGENTS.md, PRD/ARD WIRE-FIRST and canonical handoff. Source code path site/preview-v2 is temporary internal base and does not authorize a second public page.
Activity
masarray commentedon Nov 4, 2026
Root-cause audit — prior GUI threads rebuilt a competing renderer (2026-11-04)
Code evidence from actual repo (not screenshot inference):
main/CMakeLists.txtcurrently embedsweb/index.html web/styles.css web/app.js web/mixer-layout.js ...;main/src/Main.cpp::assetFor("/")servesBinaryData::index_html. No production executable binding to currentsite/preview-v2/**.web/console-native.html/.css/.jsandMain.cpp::assetFor("/")toBinaryData::consolenative_html. This is a parallel visual renderer, not a native adapter of the approved GUI. Do not ship this UI even if Windows CTest passes; extract only verified engine/bridge logic.web/index.html,web/mixer-layout.js/.css,web/styles.css(legacy workbench DOM); it does not turn current mainsite/preview-v2/index.htmlinto the compiled native entrypoint. Still-open stale draft INTEGRATION (DRAFT): approved M3 Console in native JUCE, real snapshot/intent path — universal still blocked #356 attempted a source copy from an older M3 revision, but cannot substitute for deterministic build from latest approved main source.scripts/verify_native_windows_artifact.ps1demands SHA256 ofweb/index.html,web/app.js,web/styles.css;tests/native_artifact_packaging_contract_test.mjsassertsBinaryData::index_html;scripts/check_console_puzzle.mjs --releasereadsweb/mixer-layout.js. Therefore green Windows/GUI CI is not proof of the operator-approved Windows screen.9d78b453instructed GUI GUI #271 — Native adaptation of approved PR #278 V8 Mixer Console (draft) #271 to reproduce PR V17 — Magnetic OUT→IN Wire Patch Bay: stable controls, soft curves & snap (GUI reference) #278 instead of using the actualsite/preview-v2DOM and assets. Past GUI resume prompts repeated the GUI #271 — Native adaptation of approved PR #278 V8 Mixer Console (draft) #271 Packet C owner without insisting on exact native source identity. Both instructions and agent implementation contributed.site/preview/index.htmluses../preview-buttons/index.htmliframe.scripts/build_pages.pyderives that variant fromsite/preview-v2while replacingmixer-control-keys.css/.jswithsite/preview-buttons/assets/versions. For screenshot acceptance, confirm the user-approved deployed URL + button skin; lockingpreview-v2source without preview variant parity may still differ from the specific /preview/ screen the owner saw.Required fail-closed remediation: treat this as P0 release blocker. The next Windows binary must be generated deterministically from the exact approved Pages GUI and all assets, with SHA256 manifest/ResourceProvider provenance, rendered WebView2 screenshot parity (bank/density/expanded+folded), accepted-v8 snapshot/ACK binding and genuine native PCM + VST3 FAT. Add a release test that rejects compiling either
web/index.htmllegacy orweb/console-native.htmlas the operator UI. Update contradictorynative_artifact_packaging_contract_test.mjs,verify_native_windows_artifact.ps1andcheck_console_puzzle.mjstogether; do not disable audio/driver protections or call Pages SIMULATION LIVE. Keep prior old renderer branches unmerged.Owner clarity: issue #389 alone owns the exact GUI-to-JUCE stitch; #266 remains audio PCM backend; #270 joined FAT. Do not create another GUI renderer or start another alternative PR. This comment records audit findings; native source/binary migration is NOT implemented by the comment.
[-][P0 SINGLE GUI] Stitch approved /preview-v2/ into native JUCE — no alternate renderer[/-][+][P0 ONE GUI] Integrate ONLY approved /preview/ into native JUCE; remove alternate previews[/+]masarray commentedon Nov 4, 2026
P0 #389 IMPLEMENTATION STATUS — approved /preview/ native JUCE source stitch
Real implementation is underway in DRAFT PR #392, current head
a4d2e565205ac5afd124a8e4fe04fe46d59e002b.scripts/build_approved_native_gui.pycalls the productionscripts/build_pages.pyto assemble the exact approved/preview/index.htmlplus its PNG-button CSS/JS/assets, then adds only an invisibleapproved-native-bridge.jsbootstrap withaudioReady:false. No separately designed console, no copied outdated M3 mock.CMakeLists.txtcompiles only generatedapproved_gui/index.htmland its complete resources into JUCE BinaryData;src/Main.cppdispatches only through generatedApprovedGuiResources.h, not oldweb/index.htmlorweb/console-native.html.Do not reintroduce another GUI in engine or GUI PRs; continue M2 binding accepted native control state to this exact existing DOM after M1 compile proof.
masarray commentedon Nov 4, 2026
M1 exact-source Windows native build evidence — VERIFIED @
a4d2e565205ac5afd124a8e4fe04fe46d59e002bscripts/build_pages.pyfor native/preview/, with approved PNG skin. The Windowssrc/Main.cppWebView ResourceProvider uses only generated approved GUI assets; oldweb/index.htmlandweb/console-native.htmlno longer compiled as native product.Next owner work should consume approved GUI as-is; no parallel renderer, leverage ENGINE #266 + #387 rather than duplicating PCM.
feat(native M1): embed exact approved /preview/ into JUCE WebView2
masarray commentedon Nov 4, 2026
ENGINEER AUDIT — actual GUI→native readiness, main@8beeccabdea8 (2026-11-04)
Verdict: M1 visual asset embedding established; native control/state integration M2 is NOT DONE. This is source-read audit (no customer hardware or Windows screenshot claim). Approved frontend is ONLY assembled /preview/ with PNG buttons.
Evidence from actual main
Exact bridge points — implementation owners
Minimal fast product milestone
Deliver ONE IN1 verified physical/source → CH01 fader/mute/PAN → committed CH01→MAIN WIRE → real MAIN PCM + native meter + persistent Show on the actual approved GUI. All other controls remain honestly UNAVAILABLE until capabilities installed. Then add CH→CH WIRE transaction/replug, per-channel VST3, dynamic channels and full FAT. Do not create another renderer, duplicate callback, mock audio or a new suite/workflow.
DoD: actual Windows EXE with identical /preview/ GUI and its PNG keys, v8 accepted graph/ACK/rollback, real measured PCM, screenshot parity, coherent C++ fixture and physical evidence. M1 visual pass by itself is NOT a product release.
masarray commentedon Nov 4, 2026
M2 implementation directly on latest main — DRAFT PR #395
Continues merged M1 #392 without a competing GUI or PCM callback. New
src/ApprovedConsoleM2Snapshot.hprojects REAL JUCE active ASIO/WASAPI mask and verified LiveDryRoutingLab callback ordinals, with empty universal ChannelId graph andproductionCallbackLive=falseuntil engine #266/#387 admits it.src/Main.cppnow registers actualarMixConsoleSubscribe, emitsarMixConsoleSnapshot, and explicitly REJECTS unadmittedarMixConsoleIntentwitharMixConsoleAck; no optimistic UI update or ordinal→ChannelId lie.Native-only
web/approved-native-m2-state.jsvalidates uint64 decimal IDs, real endpoints, capability gates, request/ACK matching, same epoch/graph/roster snapshot readback, stale/rejected/timeout safety, and fail-closed recovery.web/approved-native-bridge.jssubscribes + blocks browser-only simulated audio controls in the approved /preview/ native application (visual DOM/CSS unchanged). Deterministic Pages assembler + binary manifest now embeds both bridge scripts. 11 Node state-machine tests and Windows CI checks added.Browser Chromium PASS on run #41897279469. Windows JS/preflight/CMake PASS; exact-head JUCE native C++/CTest still building #41897279227.
M2 not yet finished: authoritative stable ChannelId→physical GUI strip mapping + admitted PCM/FX/WIRE operations remain dependent on unmerged native engine producer #266/#387. PR remains draft, no M3 product LIVE, no hardware FAT #15 claim. Ownership safety: do not cherry-pick/duplicate ENGINE's PCM source or bring back
web/console-native.html.masarray commentedon Nov 4, 2026
M2 EXACT-HEAD VALIDATED — PR #395
Commit
28862dd4a79d8772bb34b73fd151853b94edb88cis green on BOTH independent gates:/preview/UI remains unchanged visually.Native v6 socket is now installed in
src/Main.cpp;src/ApprovedConsoleM2Snapshot.hreports only ACTUAL physical device pin masks plus epoch-verified callback ordinals;web/approved-native-m2-state.jscommits ONLY after matching native ACK + same-revision/epoch accepted snapshot, including stale/malformed/timeout rejection. Native demos do not control imaginary PCM.M2 transport/safety milestone complete, M2 operator-to-PCM journey NOT complete. Main installed callback currently cannot certify universal ChannelId ownership; it advertises
productionCallbackLive=false, empty accepted ChannelId roster and zero supported operations. Engine #266/#387 still draft/unmerged. Do NOT merge PR #395 as audio-live product or mark UAT ready; it remains DRAFT while the producer, GUI ChannelId binding, M3 WIRE/VST3, Windows EXE visual parity and hardware crackle #15 FAT remain blocked. No duplicate GUI/DSP allowed.masarray commentedon Nov 4, 2026
Milestone A — integration coordination (2026-11-05 WIB)
User authorized parallel execution with other ArSonMix threads. One owner per runtime domain; do not fork GUI or engine. Current main
c1a11afd145380a944c9844937ccb6a97b5e0557already merged M2 #394; treat it as baseline, NOT old #392/#395 checkpoints.src/LiveAudioEngine.h,src/LiveDryRoutingLab.h,engine/core/**, native VST3 leases and installed JUCE callback. Deliver accepted ChannelId PCM cutover with one renderer and preserve CH1/Orbit, 10CH+4AUX, Show. Do NOT promote oldweb/console-native.*renderer. Reconcilesrc/ConsoleLiveBridgeV6.hinto approved M2src/Main.cppevent handler, no parallel socket or blanket-reject once truly admitted.productionCallbackLivemust reflect actual admitted callback; never force true./preview/plus PNG controls. M2 — authoritative native v6 snapshot/intent/ACK seam for the single approved /preview/ mixer #394 is merged. M2 — approved /preview/ native v6 readback + ACK/snapshot gate (no fake LIVE) #395 overlaps with M2 — authoritative native v6 snapshot/intent/ACK seam for the single approved /preview/ mixer #394; selectively reconcile differences, not both controllers simultaneously. Browser never owns PCM or physical pins.scripts/check_console_puzzle.mjsgate that still tests supersededweb/mixer-layout.js+ nonexistent GUI GUI #271 — Native adaptation of approved PR #278 V8 Mixer Console (draft) #271/FAT files rather than actual approved native bridge/asset provenance. Add deterministic negative and positive source tests, preserve STOP-SHIP until integrated native PCM+GUI+FAT evidence. I will use a separate PR based on live main, without changing ENGINE/GUI visual code. Please avoid concurrently changing this gate; integrate my PR after its exact-head checks.Milestone A admission: one approved renderer in Windows, one installed ChannelId-native audio callback, one accepted graph drives real faders/meters/FX/WIRE, exact matched ACK→accepted snapshot, no alternative GUI/DSP, exact-SHA Windows native/Chromium PASS. Physical FAT remains explicitly PENDING until measured. This is a coordination lock, not a claim of completion.
masarray commentedon Nov 4, 2026
M3 FIRST REAL SOUND — direct implementation checkpoint (2026-11-05)
Actual code PR: #396 (DRAFT, no premature UAT).
Merged M2 #394 already provides native-only hooks and contract; M3 audit uncovered two additional integration blockers: (1) the deferred bridge cannot see original
/preview/IIFE-localtracks/render/findTrack/wireEdges, preventing any accepted ChannelId snapshot projection, and (2) the existing externalwindow.pointerupfader callback executes after the original GUI clearsfaderGesture. Also native FX reorder can mutate demo inserts without a JUCE guard.PR #396 adds a deterministic, reversible native-only scope export INSIDE the exact approved Pages GUI, bridges accepted snapshots into that same original model, commits pointer faders at the original gesture owner's release point, prevents native-only demo FX mutations, and forces failure/REJECTED rollback to accepted visuals. Public
/preview/HTML/CSS/PNG assets are unchanged; no new renderer or audio callback. Tests exercise a mocked JUCE socket throughsnapshot → SET_FADER intent → matching ACK → later actual control readback, and rejected state rollback. Native source bundle and exact asset parity are guarded.Truthful milestone status: M3-A gesture/transport integration is in PR; M3-B production native PCM ChannelId cutover and genuine VST3/WIRE through actual callback remain BLOCKED on reconciling engine #266/#387 into approved GUI (NOT on introducing another GUI). Real ASIO/WASAPI hardware and zero-FX crackle #15 remain FAT PENDING. Do not claim First Real Sound or product/UAT ready from the M3-A tests alone. Do not merge until exact PR-head Windows/Chromium tests pass.
masarray commentedon Nov 4, 2026
M2.5 reconciliation update: #396 merged (native approved GUI state/fader), #398 merged as
0d23ff90(release/source admission; exact-head Windows/JUCE/CTest PASS). New central file-ownership handoff and ordering is posted in #397. ENGINE #266 remains merge-dirty and must NOT importweb/console-native.*or overwrite the approved M2.5/preview/bridge. #393 GUI and #270 FAT must be selectively extracted, not raw merged. Real ChannelId PCM, Windows exact joined EXE and physical #15 FAT remain pending.masarray commentedon Nov 4, 2026
M4 GUI-owner handoff, following control-plane PR #404 (no redesign of approved
/preview/):planM4OperatorIntent(approvedController.state().current, {kind,sourceId,destinationId})fromintegration/console-v6/m4-operator-routing.mjs. The returnedintentis a candidate only, and must be issued via existingcontroller.intent(operation,targetChannelId,payload); keep spinner/state frozen until native ACK and the later matching accepted graph readback. Return explicit error on cycle, duplicate, double-dry, stale ID, missing capability.ROUTE_ONLYfrom MIC→GROUP is a safer default than adding a second audible send while MIC→MAIN is still active.SEND_POSTto native-verified 100%-wet FX_RETURN preserves dry path. Do NOT infer wet-only from a visual label.SEND_PRE, variable gain, create Group/FX Return and external VST3 SIDECHAIN remain blocked, not mocked, until [ENGINE SOLE OWNER] One native callback → full-fidelity ChannelId cutover (draft) #266 installs real callback/aux detector bus and exact ACK proof.This is a targeted future GUI wiring contract, NOT a request to edit
site/**simultaneously with the M3 #402 owner. PR #404 operates purely in INTEGRATOR-owned paths and does not change any GUI presentation file. See #397 for cross-thread ownership; hardware #15 FAT still pending.[-][P0 ONE GUI] Integrate ONLY approved /preview/ into native JUCE; remove alternate previews[/-][+]P0 approved /preview/ — qualify native GUI parity and authoritative PCM readback[/+]masarray commentedon Nov 8, 2026
Reconciled 2026-11-08 (
mainb0cea49). The ONE/preview/public/native renderer integration is already in main through #418/#440/#442 and subsequent GUI/bridge work. Historical competing GUI migration #393 was archived. Retaining this issue is only for distinct native screenshot/interaction parity and real authoritative PCM/ACK readback verification; engine signal FAT is #412, channel lifecycle #192. Do not interpret the old title/body as 'GUI not yet integrated', and do not create a second renderer.masarray commentedon Nov 10, 2026
Administrative consolidation (2026-11-10)
The single approved /preview/ native GUI implementation is already integrated; its remaining independent acceptance has now been explicitly transferred to:
All these owners remain OPEN until real acceptance. Audit document: merged PR #524. Closing this redundant tracking issue does not claim the product is ready, and does not authorize a second renderer.