Skip to content

HANDOFF P1K→P4/Production: canonical ArSonMix 10 INPUT + 4 AUX architecture, latest GUI/CI gate, next-thread execution #137

Closed as duplicate

Description

@masarray

Canonical engineering handoff — START HERE (2026-10-29, Asia/Jakarta)

Purpose. This issue is the complete cross-thread handoff for the next ArSonMix engineer. It supersedes status snapshots in older handoffs (but not technical evidence). Avoid redesigning an already merged feature. First check live main, PRs, and exact-SHA CI again; other threads are active. This is not a request for speculative architecture or tiny documentation PRs: deliver a meaningful executable/usable native milestone with tests and a verified Windows artifact. User requires excellent real-hardware low latency, robustness, maintainability, portable FX, and light premium UI; Studio One / Adobe Audition are quality/feature benchmarks, not achieved parity.

0. LIVE VERIFIED BASELINE — DO NOT ROLL BACK

Repo: https://github.com/masarray/arsonmix

At authoring, main@6077513e7be0d25e370b8e5841aa6edcba473a73 — 6077513

Milestone State Ground truth
P1I native realtime fader/PAN/MUTE/SOLO 10CH MERGED docs/P1I_UNIFIED_10CH_REALTIME_CONSOLE_20261028.md, exact device/graph revision control
P1J stable FX graph lifecycle + native routing ACK + SIGNAL FLOW editor MERGED PR #130, main@bc11edf; unchanged same-slot/same-catalog VST3 processor instance reused across route graph edits
P4 stage 1 native ShowV2/SnapshotV2 JSON validation MERGED main@faee66f, src/NativeShowSchema.h only inert schema; NOT native snapshot recall/capture
P1K ten independent INPUT strips plus four dedicated AUX stereo FX returns MERGED, not pending PR #136, squash main@662b060f7455abdc76e642cafd9b99d913852e91; exact-head Windows SUCCESS, Chromium SUCCESS; post-merge main@662b CI Windows SUCCESS and Chromium SUCCESS
P0 GUI first-open startup, inset mono/stereo VU layout and LUFS/dBTP LCD MERGED PR #133, main@6077513. Preserve this visual baseline, do not cherry-pick older UI over it.
Latest MAIN CI at 6077513 not fully green at handoff Chromium UI smoke FAIL; Windows run still executing at last read. Re-check before acting.
P4 stage 2 native transactional Show storage OPEN parallel PR #134, not merged #134 ; head eb367a5a752846c05e436a0618168d5f85576b12, older main@faee66f ancestry. Branch's PR file list unexpectedly spans P1K/GUI files; diff-reconcile against current main before merging, do not overwite P1K/#133.
Old #128 / #115 / #117 / #113 / #60 / #59 / #25 / #26 / #29 and others OPEN / overlapping or historical Do not merge based on PR title/date. Audit changes and ownership; some contain superseded GUI/engine code.

URGENT P0: latest GUI CI broken

The exact failure from main@6077513 Chromium log (not inferred):

  • tests/ui_smoke.py:951: after calling updateMeterVisual("masterL",99,performance.now()), asserting latched MASTER clip, clicking [data-clip="masterL"] then checking Boolean(peakHold.masterL) is False fails.
  • Related code in web/app.js: peakHold, clipLatch, resetClip(key), updateMeterVisual, shared native/preview meter repaint. Could be event bubbling/capture, asynchronous meter reapplication or reset semantics — root cause NOT proven. Reproduce and inspect state before altering; preserve 30fps single renderer and latched clip behavior. Do not silence/relax a valid test just to make CI green.
  • CI run: https://github.com/masarray/arsonmix/actions/runs/40786734154 . Last known previous-main Chromium success: https://github.com/masarray/arsonmix/actions/runs/40780687265 .

1. Product / non-negotiable requirements

ArSonMix: Windows-first JUCE native VST3 digital live mixer for vocals, music, broadcast, karaoke and hardware interfaces; WebView2 for compact premium offline GUI. Product qualities: ultra-low latency, realtime determinism, no crackle, minimal GPU/CPU/allocations, fast hardware setup, self-contained portable user VSTs/config/backup (user plugins not bundled in public download), production-grade failure recovery. Desktop native and Web preview are not the same feature-completeness: normal NATIVE CH1 status still indicates one live CH1 host + other UI-preview strips unless the verified 10CH Routing LAB is ACTIVE. Do not claim all 10 strips have native FX in normal production mode.

UI reference approved by user: ten equal-width input strips + one MASTER, currently last approved #133 lightweight WebView first-paint, nonblurred premium recessed steel/metal, compact mono meter well without vacant stereo space, fixed red/amber/green volume-meter thresholds, MASTER black-glass green LUFS/dBTP LCD only, gray toolbar OFF / green ON, BTN-180 responsive hover without blur, 30-ish FPS single scheduler, low-end Windows GPU responsiveness. Keep all recent Orbit, channel rack and FX drag/drop/declick behaviors. No heavy CSS filter blur(), brightness(), giant box-shadow glow, transform-heavy hover or multiple interval painters. The latest source CSS is web/styles-src/42-mixer-premium.css with generated runtime/standalone parity; NEVER edit only generated HTML and forget source. PR #133 merged the newest approved layout.

2. SINGLE canonical engine topology — accurate today

JUCE Windows ASIO/WASAPI one authorized DIRECT, full-duplex device
   | input active channel mask (real sparse physical ports)
   v
JuceActivePortSnapshot / LiveInputPortRegistry + DirectRouteFence
   |  verified port ordinals, deviceEpoch; unexpected topology => silent
   v
LiveAudioEngine::audioDeviceIOCallbackWithContext
   |-- NORMAL CH1 production path: original max-8 VST3 stereo host,
   |   FX drag/drop and bypass envelopes, Orbit, existing handoff.
   |
   '-- Operator-opt-in LiveDryRoutingLab when ACTIVE, DIRECT, verified:
         RealtimeRoutePublisher<LiveChannelFxGraph>::ReadLease
           -> PreparedMixerBlock on ONE audio callback
           -> 10 physically assignable MONO/STEREO INPUT strips
              [ CH VST3 -> PRE sends -> realtime fader/PAN/MUTE/SOLO
                -> POST sends -> optional MAIN ]
           -> channel-to-channel legacy BUS DAG (backward compatibility)
           -> 4 DEDICATED stereo AUX RETURN lanes, not CH9/CH10
              [ AUX1..AUX4 sum -> own VST3 -> gain/mute/MAIN ]
           -> stereo MASTER, peak telemetry
           -> verified physical OUTPUT buffers (2ch)

Exactly 10 logical source strips are available; actual live capture count depends on the device active mask and must be enabled in AUDIO SETUP. Four AUX are additional processing returns, not 4 hardware outputs, not 4 more physical mic inputs. BUS as a source for a CH is the older internal return system, still supported. New AUX destination IDs: 10,11,12,13 for AUX1..4; physical CH indices 0..9.

GUI / CONTROL thread (non-RT):
  WebView SIGNAL FLOW -> OPEN ROUTING MATRIX
  draft CH[10] + AUX[4] + PRE/POST sends
      |
      -> src/Main.cpp trusted catalog ID resolution for channel/AUX VST3
      -> LiveAudioEngine::applyDryRoutingLab (OUTPUT MUTED + DIRECT fences)
      -> NativeDryRouteCommand::decode typed JSON and validate
      -> ChannelRoutingContract sparse-port map
      -> ChannelSendPlan DAG; AuxBusPlan (input-only one-way AUX sends)
      -> PreparedRenderGraph::compile -> PreparedMixerBlock preallocation
      -> LiveChannelFxGraph preprepare/reuse actual JUCE processors
      -> RealtimeRoutePublisher::publish(epoch,revision) [atomic]
      -> ACK acceptedRoute + counts, then WebView rehydrate + live controls
AUDIO thread: read immutable prepared graph, atomic live controls,
  process no allocations, no file/GUI/VST discovery, no locks/spin.
CONTROL thread: poll retired readers, release old vendor FX ONLY after quiescence.

Architecture module map / source of truth

Component Responsibility / don't break
src/LiveAudioEngine.h Owns one JUCE AudioDeviceManager, original CH1 production FX graph, direct route fence, P0 diagnostics/clock mode; hands verified callback to the opt-in LAB. No second parallel mixer engine.
src/LiveDryRoutingLab.h One live LAB control/RT adapter, device epoch/active input-output masks, atomic publish, acknowledged route and channel/AUX peak/FX status. STOP restores CH1 safely and MUTES output. acceptedRoute_ is JUCE var owned by control thread; never mutate it in device-stopped callback (fixed P1J race).
engine/device/JuceActivePortSnapshot.h and engine/device/LiveInputPortRegistry.h (if present under exact path) Verify actual JUCE enabled channel mask and sparse pin ordinals; reject stale/reordered devices. Check current tree for exact names/paths.
engine/core/ChannelRoutingContract.h Physical pin maps to callback-ordered indices and MONO/STEREO fold. Port number != callback pointer index.
engine/core/ChannelSendPlan.h Original 10-strip, max32 bounded sends, DAG, explicit PRE/POST, duplicate/cycle prevention and solo closure.
engine/core/AuxBusPlan.h P1K 4 dedicated stereo returns with PRE/POST input-only sends; validates enabled target, gain, PAN, source 0..9 and AUX 0..3, duplicate edges. No AUX-to-AUX feedback possible.
engine/core/PreparedMixerBlock.h Allocation-free prepared per-block 10CH + optionally 4 AUX scratch + MASTER. DSP sample ordering defined here; effect seam ids 0..9 for source, 10..13 for AUX, -1 for MASTER. No std::vector resize inside callback.
engine/core/LiveChannelFxGraph.h P1J FX identity/lifetime: FxInstance owns processor/private MIDI scratch, shared ownership only in control thread, same-slot+catalog ID+device epoch/quantum reuse, RT dereferences raw pointer. std::atomic_flag try-enter on unexpected concurrent callback, fail silent rather than spin. 14 total FX slots, one stereo-compatible VST3 per strip/AUX today.
engine/core/RealtimeRoutePublisher.h One control writer + bounded realtime read leases + deferred retirement. Gate writer/readers with single atomic activity state: do not replace with unsafe separate readerCount/pointer scheme.
engine/core/NativeDryRouteCommand.h Untrusted JSON typed decoding, exactly 10 source CH rows, 4 AUX definitions when supplied, up to32 draft sends, native validation before publication; legacy no-AUX request still valid.
src/Main.cpp Native message-thread WebView action bridge, validates VST3 fxId with catalog, maps 10 input FX + 4 AUX FX selections before apply; no arbitrary client file path. PR #133 and future PRs may touch this file: reconcile file-level changes.
web/monitor.js Native-only Signal Check and Routing Matrix, 10 channel draft controls + 4 dedicated AUX rows, target dropdown CH/AUX, Native ACK rehydration and live fader/PAN/M/S, not direct DSP. Latest UI may have a separate visual preview in web/app.js and standalone HTML; don't mistake it for live DSP.
src/NativeShowSchema.h P4 stage1 only: bounds/typed ShowV2 / SnapshotV2 JSON. Exactly ten source records, up to eight FX refs per channel, but does NOT model P1K 4 AUX plus FX slots and does not capture/recall state. Schema migration/versioning is required before claiming complete AUX-enabled portable snapshot.
PR #134 src/NativeShowStore.h (branch only unless merged) P4 stage2 speculative/under-test transactional on-disk generation store with SHA256 VST state blob verification and atomic HEAD. Has not established native live scene capture/recall, device activation or AUX state. Carefully port storage-only changes onto fresh main.
docs/P1K_TEN_INPUTS_FOUR_DEDICATED_AUX_20261029.md and docs/P1J_STABLE_FX_ROUTING_SESSION_20261029.md Detailed sample tests, hardware route usage, FX reuse and unfinished production gates.

3. IMPORTANT cross-thread ownership & traps

  • Do not rewind main. A GitHub branch created on faee66f or bc11edf is STALE for modifying current GUI. Merge/rebase by comparing individual changed filenames and ancestry, especially src/Main.cpp, web/app.js, web/styles.css, web/styles-src/42-mixer-premium.css, standalone HTML and source/derived styles.
  • P4 P4 stage 2: transactional native Shows generations, verified VST3 blobs and backup tests #134 currently has broad PR file list including P1K and P0: GUI-first WebView startup, compact mono/stereo LED wells, LUFS/dBTP LCD #133 GUI files, despite descriptive claim of storage-only scope. The branch diff relative to current main must be reconciled to only storage/CI/native tests before review/merge. No wholesale squash of stale ancestry.
  • PR P0: gray OFF/green ON toolbar, stable JUCE Audio Setup, snapshot manager + native Show ARD #128 open contains older toolbar/snapshot code; P1D: prepared 10-strip dry execution — sparse inputs → FX tap → bus sends → MASTER #115/P1C: runnable 10-strip + 2-AUX offline DSP routing graph with cycle-proof sends #117 older routing plans conflict with already merged P1I/P1J/P1K. Never blindly merge, and do not duplicate active FX hosts.
  • A functioning binary and GUI screenshot do not prove arbitrary VST3 vendor stability, sidechain bus layout, native scene recall or acceptable real-hardware buffer latency. Explicitly distinguish CI vs physical FAT.
  • User explicitly dislikes an endless sequence of tiny PRs: choose one complete user-facing vertical slice per milestone (e.g. native multi-insert FX rack for 10CH+4AUX with true audible processor/controls/scene integration), audit before code, implement, test, release.
  • All structural LAB APPLY currently requires output muted and genuine same-device DIRECT audio (not independent/dual-clock, not unverified ASIO, not while CH1 standalone chain occupied). Do not pretend it is zero-interruption DAW-style patching. No automatic unmute.
  • No VST3 processing while adding/removing plugin on audio RT, no network/file/logging/UI calls from callback, no malloc/heap, no wait/mutex, no background scanning there. Bounded O(strips+sends) DSP with preallocated sample arrays.
  • Thread-specific FX instance reuse depends on catalog ID + stable slot + device epoch/quantum, and vendor state/tails survive while processor is continuously processed; structural APPLY still requires MUTE.
  • GUI must be responsive on standard low-spec Win11 with onboard GPU, no visually expensive blur. Preserve quality of latest recessed meter visual / dual-metric LCD from P0: GUI-first WebView startup, compact mono/stereo LED wells, LUFS/dBTP LCD #133.
  • User portability requirement: the public app never distributes user's private VSTs; any user can manually import self-owned VSTs/presets into app-owned vault, back up to USB, recover on another PC without losing state. Device-specific physical pins must be remapped/validated; not silently restored across hosts.

4. Actual test and release gates

P1K exact PR head f9a6a3bd67b0de4deb2b17e7dd417bbef54ca3b7 Windows+Chromium both green. After squash merge main@662b060f Windows+Chromium green. Latest GUI-only main@6077513 Chromium red at peak-hold reset and Windows still running at the final author check; re-check both before changing anything.

Maintain existing Windows workflow .github/workflows/p1-windows.yml and Chromium tests/ui_smoke.py:

  • CTest P1KTenIndependentInputsAndFourDedicatedAuxReturns exercises all ten physical samples, two concurrent AUX, exact PRE/POST FX/gain/Master PCM math, AUX mute, rollback, invalid routing and zero allocation in callback.
  • P1HRealNativePerChannelFxAndSafeRetirement extended in P1K with real JUCE AudioProcessor::processBlock on CH1 and AUX1, exact processor pointer identity across revision and no re-prepare/early-release.
  • LiveNativePrePostSendAndBusContract includes native juce::var AUX destination ID 10 decode, disabled invalid AUX and rollback. JS tests/live_routing_ui_contract_test.mjs checks real native integration.
  • Keep all tests for epoch hotplug, stale revision, publisher lease/retirement, bound masks, bypass de-click, Orbit, snapshot/schema, FX order, GUI latency.
  • For any merge require: fresh MAIN, exact PR-head Windows MSVC full app + all CTests + ZIP verification, exact-head Chromium UI smoke and browser screenshots, review of PR diff for unrelated regressions. Then squash merge with expected SHA; monitor postmerge main Windows+Chromium. If failing, fix actual cause, don't edit acceptance tests to hide faults.
  • Real Windows hardware FAT required: 48 kHz, 128/256/480 sample periods (where supported), ASIO and WASAPI, active sparse input masks; test with physical 1,2,10 inputs actually enabled, 0 FX dry, one CH VST3 and AUX reverb, PRE/POST behavior, disable/restart/unplug, mute/crossfade, DPC/xruns/CPU/latency and vendor GUI ownership, safe retry; collect capture xrun vs render underrun separately.
  • Known open P0 P0 — Crackling with 0 VST3: diagnose WASAPI clock/input/output buffering before engine expansion #15: audible crackle can occur with ZERO VST3 in hardware, so never attribute it solely to GUI, VST3 or routing without evidence. Do not claim solved by synthetic test.

5. PROPOSED next meaningful production milestone — priorities

Gate A / P0 stability first: resolve main@6077513 MASTER peak-hold click reset UI smoke failure (without undoing #133). Verify newest Windows CI and actual screenshot baseline. Fix only the defect, not wholesale GUI redesign. If parallel GUI thread is already repairing it, coordinate / avoid overlapping edits.

Gate B / P4 Show storage coordination: read PR #134 exact diff against CURRENT main, split/cherry-pick only relevant transactional store/schema/test code (if correct), preserve newer P1K and #133 GUI. Stage1 schema must be extended/versioned for 4 AUX returns, AUX FX and sends before marketing ShowV2 as mixer-full; physical device remap, vendor state binary capture/restore, preflight and atomic apply are separate real engineering tasks. No raw VST DLL copy into public installer.

Milestone C / production-native routing/rack vertical slice: migrate real native 10CH+4AUX from opt-in LAB to an explicit production mixer mode without a second audio engine. Separate device routes, bus graph, per-strip independent ordered multi-insert VST3 rack with parameter GUI/bypass/reorder, correct mono/stereo bus negotiation, master FX and AUX routing. Compile immutable graph/control transactions offline, preallocate scratch and MIDI, keep plugin identity & state on harmless edits, de-click and zero RT alloc. Implement PDC before allowing latent plugin inserts in parallel bus paths (currently >0 latency FX are rejected). No hardware AUX output patch until endpoint masks and per-bus routing validated. One consistent native SSOT, not UI-local sound settings.

Milestone D / UI and portability complete: wire MIX/FX RACK and SIGNAL FLOW into same acknowledged channel/bus graph: intuitive routing input selector, AUX list/dedicated returns, PRE/POST sends, routing badges and meters; remove misleading CH10 "FX Return" mock label when channel is genuinely a physical INPUT. State persistence: project-owned copies of user-imported VST3 + presets, verified plugin identifiers/ABI, snapshot capture/restore with state blobs, non-destructive recover and endpoint remap, atomic cross-device failures. UI redraw exclusively from native acknowledgment; no hidden simulator path presented as real.

Milestone E / audio hardware qualification: complete #15 zero-FX crackle proof, low-end Windows GPU profiling, latency measurement, session hotplug/recovery, 10CH+4AUX stress with real vendors. CI does not replace physical interface testing.

6. Immediate first actions for a NEW THREAD

  1. Read this issue, then live GitHub main SHA and newest commits, open PR P4 stage 2: transactional native Shows generations, verified VST3 blobs and backup tests #134 and any parallel PRs, Windows/UI workflows. Do NOT assume this text's 6077513 still heads main.
  2. Verify which commits already merged (P1K: 10 always-available input channels plus four real independent AUX/VST3 return buses #136 P1K and P0: GUI-first WebView startup, compact mono/stereo LED wells, LUFS/dBTP LCD #133 GUI ARE MERGED) and mark old handoffs stale. Read live files src/LiveAudioEngine.h, src/LiveDryRoutingLab.h, engine/core/{PreparedMixerBlock,AuxBusPlan,ChannelSendPlan,LiveChannelFxGraph,RealtimeRoutePublisher,NativeDryRouteCommand}.h, src/NativeShowSchema.h, web/monitor.js, src/Main.cpp, tests/ui_smoke.py.
  3. Investigate the exact UI test failure with console state/event trace and apply the minimal correct fix; run BOTH Windows and Chromium CI on exact head. Coordinate with current GUI owner; avoid duplicated commits.
  4. Audit P4 P4 stage 2: transactional native Shows generations, verified VST3 blobs and backup tests #134 against newest main and its unexpectedly broad PR changed-files list; if parallel Show thread owns it, keep hands off and leave a precise review/handoff comment.
  5. Select a single meaningful end-to-end next milestone, write short in-repo ARD/proof contracts, implement real native observable behavior, native PCM/VST processor/rollback/hotplug tests, and no-RT-allocation instrumentation. Test source tree before writing patches, use a branch off latest main and do not reintroduce old GUI.
  6. Report verifiable links: updated PR, exact head SHA, test results, packaged Windows artifact, what is and is not real hardware verified; link merged main commit only if actually merged.

Ready-to-use new-thread instruction

Continue masarray/arsonmix from [THIS ISSUE]. Treat live main as SSOT, not older screenshots or handoffs. P1I/P1J/P1K and PR #133 GUI are already merged; do not duplicate or regress them. Resolve latest UI smoke failure before feature merge, reconcile P4 #134 carefully. Respect one JUCE DSP engine, immutable prepared routing + AUX, epoch/revision/hardware safety, control-only plugin lifecycle, zero callback alloc/locks, real 10 input + 4 AUX, vendor FX state, latest lightweight GUI/Orbit and backup portability. Deliver a production-grade milestone and Windows/Chromium exact-head acceptance, with truthful hardware caveats. Never naive trial/error or cosmetic-only progress.

Historical references: issue #109 (long-running route requirements and handoff comments), P1K PR #136, P1J PR #130, P4 #134, P0 GUI #133, docs/HANDOFF_ARSONMIX_NEXT_THREAD_20261029.md (earlier, now status-stale), docs/ARD_SMART_AUDIO_ENGINE_V1_2026-10-25.md.

Activity

  1. masarray commented on Oct 28, 2026

    @masarray
    OwnerAuthor

    Execution continuation — 2026-10-29 (Asia/Jakarta)

    • Rechecked live baseline: prior main@6077513 failed Chromium MASTER peak-reset because browser demo's timer/RAF could overwrite peakHold between reset and assertion. The deterministic smoke isolation was incorporated by parallel P4 PR P4 stage 2: transactional native Shows generations, verified VST3 blobs and backup tests #134; P4 stage 2: transactional native Shows generations, verified VST3 blobs and backup tests #134 is now merged into main@e24f9169eea5. Its exact-main Chromium smoke is SUCCESS. The main Windows CI was still running at this check.
    • Before integrating, a SHA guard detected the parallel main move. No stale writes, force pushes or duplicate fix were made; P4 stage 2: transactional native Shows generations, verified VST3 blobs and backup tests #134 branch is left untouched.
    • New substantive storage milestone PR #138 on fresh main@e24f916, exact head fd54118e91c151e6c2195fc1218af2cf61a0c92c. Only 5 files: native Show codec, native transaction store, two native CTests, P4 doc. No web/UI/Orbit/RT callback modifications.
    • P4 V3 snapshot records exactly 10 INPUT + 4 dedicated stereo AUX return metadata/FX and 32 input-only PRE/POST AUX sends; validates IDs, tap, gain, pan, enabled destinations, duplicates and bounded FX references. Existing V2 snapshot (no AUX) decodes with 4 returns disabled. V3 required for AUX so older executables must reject, rather than silently discard AUX on re-save.
    • Native Show store commit/read now requires and verifies actual supplied AUX VST3 state blobs with SHA256 alongside input FX; missing/corrupt references cause atomic failure. CTest checks roundtrip, version migration, malformed routes, HEAD retention and no partial restore.
    • PR #138 Windows MSVC + all CTest + ZIP in progress at initial check. Chromium PR workflow intentionally does not trigger for storage-only changes; main exact baseline Chrome green. Do not merge until Windows native test/artifact verified and exact-head Chromium gate addressed per release policy.
    • No claim of live plugin capture/scene recall, plugin delay compensation, physical ASIO/WASAPI qualification, or resolution of zero-FX crackle P0 — Crackling with 0 VST3: diagnose WASAPI clock/input/output buffering before engine expansion #15. FAT gates still required at 48k, 128/256/480 frames, 1/2/10 active inputs, AUX VST3, sparse port masks and hotplug, separate capture/render xruns.

    CI update: main Windows run 40788501293 SUCCESS, 47/47 CTest, verified ZIP uploaded artifact ID 12861340936. PR #138 Windows still executing; do not conflate artifacts.

  2. masarray commented on Oct 29, 2026

    @masarray
    OwnerAuthor

    P4 continuation — actual GitHub code and CI gates (29 Oct 2026)

    • Merged PR #138 at main c66649059169428385f3804f4281df3df931ca8e: AUX-aware Show snapshot v3 with verified SHA256 VST3 state references and legacy V2 compatibility. PR Windows CTest 47/47 PASS, verified native ZIP; main Chromium already passed.
    • New end-to-end functional milestone PR #139, initial head 366fead5be75af023c2d9f1acc8e383769647931, hardened final head d69df4691c333374e68cb256e4c87c9ff4030323, fresh parent c666490. Adds operator-visible SAVE DRY SHOW from live Routing LAB native ACK. Captures real verified 10 INPUT / 4 dedicated AUX, sparse physical mask, live fader linear 0 exact, PAN/MUTE/SOLO, PRE/POST internal and AUX sends, NONE vs BUS, master from native state. It rejects stale epoch/revision, unmuted output, any VST3, active independent clock, missing actual ports; no UI draft/localStorage trusted. Saves on owned non-audio IO worker, native transaction/HEAD, post-save verification, mutually excludes Studio USB backup.
    • PR P4 stage 3: live native DRY Show SAVE from ACK routing + off-RT transactional storage #139 changes only new native capture/worker and CTest, tiny native status fields, Routing LAB button/event, CMake/one existing Windows workflow, and stage3 ARD. Never touches main mixer GUI CSS, app.js, Orbit, processor processBlock, or route publisher lifecycle.
    • Chromium exact PR head SUCCESS. Windows MSVC / new CTest / ZIP PR run launched and still running at initial report. Keep PR unmerged until it passes; minimize CI to one commit/head.
    • Honest boundaries: This is the first actual live-session native SAVE for DRY / zero-VST3 only, not plugin-equipped capture, not scene recall/undo, not live production mode. Vendor getStateInformation still needs a qualified quiescence/affinity handshake; PDC, P0 — Crackling with 0 VST3: diagnose WASAPI clock/input/output buffering before engine expansion #15 zero-FX crackle, physical 48 kHz ASIO/WASAPI 128/256/480 FAT and device remap remain open. No hardware qualification is inferred from CI.

    Latest final-head gate: P4 #139 added explicit rejection ACK to the native panel and IO-worker exception containment; no further UI/DSP rewrites. Exact-final-head Chromium #40796310269 PASS. Exact-final-head Windows #40796310111 queued at verification; old PR build #40795833394 may still be draining. Do not claim CTest/ZIP PASS or merge until the final-head Windows run actually completes.

  3. masarray commented on Oct 29, 2026

    @masarray
    OwnerAuthor

    Progress update — 2026-10-29, P4 Native Shows

    P4 #141 review hardening — final SHA gate

  4. masarray commented on Oct 29, 2026

    @masarray
    OwnerAuthor

    Product-owner routing direction supersedes operator LAB workflow — 2026-10-29

    New canonical P0 routing roadmap: #143. Reconciled main@81c892f39996689c63b5704dcc060a9c42f0339d, merged #140 splash and #142 measured MASTER metering, latest main Windows/Chromium/Pages PASS. The owner explicitly rejects MODE/LAB-based operation; requires real always-live 10 INPUT + 4 AUX + MASTER, with a Waves eMotion LV1-inspired low-overhead INPUT PATCH / INTERNAL / OUTPUT PATCH grid (original ArSonMix artwork, not proprietary Waves assets).

    Native LiveAudioEngine still has CH1 legacy FX vs opt-in LiveDryRoutingLab graph; current web/monitor.js 760px dropdown editor only exposes the latter. Do not cosmetically rename this LAB. The R0 gate in #143 migrates CH1 FX/Orbit into the single production graph, preserving VST3 identity and zero-alloc RT; R1/R2 then build product-native patch grids from trusted device port mask, ACK and epoch/revision; R3 adds actual physical output patching only when hardware engine supports it. #141 is OPEN, green on an older base but diverged 2 commits; its IO/DRY recall/mono/master code should be reconciled, not bulk merged as LAB UX. Legacy PR #128 likewise not a GUI source. Hardware #15 remains open. CI gate per tested milestone, no docs-only PRs.

  5. masarray commented on Oct 29, 2026

    @masarray
    OwnerAuthor

    R1 product routing now merged: PR #146, main@3ab35d3f3a8f8afef2a5909de531ab98618d61ba. Native 10CH INPUT PATCH matrix is functional (real port/epoch SSOT, MONO/STEREO L/R, draft/ACK/revert, 8 columns/page, 1680/1440 Chromium PASS); detailed handoff on #143. R0B still mandatory for the single default live graph; R2 internal matrix and R3 physical output patch outstanding, hardware FAT #15 open. Postmerge Windows 40849670196 pending. Preserve #145 lightweight hover optimization and main GUI baseline; do not merge stale #141/#128 over this new patch UI.

  6. masarray commented on Oct 29, 2026

    @masarray
    OwnerAuthor

    LIVE HANDOFF CORRECTION — 2026-10-29 18:00 WIB

    This issue's embedded status tables are a historical snapshot; DO NOT treat their old main SHA, CI failures, P1K/R1 pending labels or #134 status as current.

    Canonical reconciled status has been merged to main in PR #150 (commit 831ae2ebb94a83585a73b05bba8cbe7facea8110):
    https://github.com/masarray/arsonmix/blob/main/docs/CANONICAL_PROGRESS_RECONCILIATION_20261029.md

    Verified immediately before the doc merge:

    Future threads: read AGENTS.md first line, the linked canonical doc, then refresh GitHub head/PRs. This is a reconciliation pointer, not evidence of new realtime functionality.

  7. masarray commented on Oct 31, 2026

    @masarray
    OwnerAuthor

    ISSUE RECONCILIATION — historical/overlap closed, NOT declared solved. Historical October 29 engineer handoff is not an executable feature issue and is superseded by PRD/ARD v5 and live progress ledger. Keep its completed-state evidence as archival.

    Current authority: PRD v5, ARD v5, native epic #192, GUI L2 #176, and hardware #15. Existing original comments and links remain.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions