Repository navigation
Conversation
…re channel inputs
… and use custom Audio I/O setup
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
…or DAW native input workflow
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f4d0ca5e6a
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".
| // Reuse the ALREADY implemented compact Audio I/O panel from mixer-io. | ||
| // Its window-capture setup handler owns driver inventory and APPLY/ACK. | ||
| // The stock JUCE device popup remains available only via Advanced. | ||
| setup.dataset.nativeAction="audioSetup"; | ||
| setup.title="Configure real input and output devices in the mixer"; |
There was a problem hiding this comment.
Keep Audio Setup connected to a handler in the approved bundle
In the approved native Windows GUI, this removes the working Audio Setup click handler without installing its replacement. scripts/build_approved_native_gui.py bundles the approved Pages assets plus approved-native-contract.js and this bridge; neither it nor scripts/build_pages.py loads web/mixer-io.js, where the referenced capture handler lives. Consequently, More → Audio Setup now does nothing, preventing users from opening device configuration through this control. Wire the custom panel into the actual approved bundle before removing the existing backend action, or retain a working fallback.
AGENTS.md reference: AGENTS.md:L37-L37
Useful? React with 👍 / 👎.
| const strip=el.closest?.(".strip[data-strip]"); | ||
| const channel=Number(strip?.dataset.strip); | ||
| const outcome=window.__arMixChooseChannelHardwareInput?.(channel,el.value); |
There was a problem hiding this comment.
Bind the IN selector to the approved renderer's channel identity
For every channel in the approved native GUI, this lookup returns no strip: site/preview-v2/index.html renders .strip with data-id, not data-strip, so channel becomes NaN. Moreover, the approved bundle does not load mixer-io.js, which defines __arMixChooseChannelHardwareInput and populates the hardware options; that module also immediately exits without the legacy #mixer and .workbench elements. Thus the new IN path always rejects rather than selecting hardware. Integrate the controller with the approved DOM and map its track identity to the native channel instead of relying on the legacy renderer's attributes and initialization.
AGENTS.md reference: AGENTS.md:L36-L36
Useful? React with 👍 / 👎.
…eep native panic controls
…during G0 without fake ChannelId admission
…ual M4 PCM via epoch-fenced native readback
… PCM control in G0
…mute/meter; stale epoch fails closed
…ative IN routing authority
…wns DAW mixer controls
…wns DAW mixer controls
…wns DAW mixer controls
…ps including prepared unpatched graph
… ten-channel M4 source switch
…e native monitor, restore real JUCE device settings action
…t legacy mixer-io.js
…driver inventory and apply ACK
…e the same virtual cable
Partially addresses #412 — P0 DAW-style first-sound recovery for the actual Windows-shipped GUI.
Root cause verified
The EXE embeds
site/preview-v2/index.htmlinto JUCE WebView2 and injects onlyweb/approved-native-contract.js+web/approved-native-bridge.js. The full-featured legacyweb/mixer-io.js/web/monitor.jsare NOT loaded by this approved EXE. Older proposals that only wired legacy selectors were not effective on hardware. The approvedINfield was a demo selector with a G0 blocker while the installed C++ M4 graph already implements real 10-channel physical source and fader/pan/mute/solo controls.Changes on actual runtime path
channelSourcesfrom real accepted / prepared graph and 10 per-channel control+meter readbacks in the existing G0 snapshot, without falsely claiming ChannelId callback admission./preview/GUI, re-use its visible per-strip IN select and actualdata-idtrack identity (not legacydata-strip), enumerate only active JUCE hardware pins, issue epoch+revision guardedroutingSourceSwitchfor any CH1–CH10, and confirm matching physical source from native ACK before enabling output. Keep real existing M4 callback, no second PCM engine.audioDeviceApply/ACK; advanced rate/buffer). Existing generic JUCE selector remains in expert More menu.Safety
ONE native realtime PCM callback and single owner for audio drivers. Driver/source hotplug or stale epoch fails safe; no JS DSP/audio worklet. New source switching requires matched native ACK. Explicit output safety MUTE remains available. Physical first sound, latency and crackle #15 NOT YET hardware qualified.
Qualification
Only merge after exact HEAD successful Chromium/Node plus Windows MSVC/JUCE CTests and provenance gates, then produce fresh exact-SHA hardware FAT ZIP. Never count build green as physical speaker evidence.
2026-11-08 reconciliation: ARCHIVED, NOT MERGED
Superseded first-sound iteration: current main contains the #418 integrated native PCM DAW graph and current approved bridge/IN control. Its older pre-#418 independent M4/native bridge patch is not a safe merge. Real physical workflow regressions remain tracked in #412.
Current approved authority: main, merged native integration #418. Existing branch history remains available; this closure is not evidence of physical audio FAT.