Repository navigation
refactor(g-mv20): invalidate stale GUI/native requests on same-epoch hardware remap - #801
Conversation
Codex Review Summary: ✅ CompletedThis 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. |
Deploying arsonmix with
|
| Latest commit: |
bffb325
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://e2c9bdb8.arsonmix.pages.dev |
| Branch Preview URL: | https://refactor-g-mv20-hardware-con.arsonmix.pages.dev |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f88d404420
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".
|
/arsonmix-ci ui ccf4572 |
|
/arsonmix-ci ui bffb325 |
G-MV20 — completion / operator handoff (2026-11-19 WIB)HASIL NYATA: Hardware pin/rate/buffer changes at the same epoch no longer keep stale WebView FX/output/source transactions or the single schema-v6 pending control intent alive. One read-only native hardware fingerprint canonically orders actual enabled physical pin SETS. The existing v6 controller, lifecycle Device ViewModel and bridge cleanup share that identity; old ACK is ignored, while same-hardware pin enumeration reorder and MONITOR ON/OFF do not falsely reset pending gestures. STATUS: MERGED MAIN as squash commit BUKTI exact HEAD:
AMAN DIPAKAI? YES as source-level GUI reliability improvement with passing pre-merge code/Chromium; NOT a fresh field-qualified Windows installer, NOT physical two-ASIO/ASIO latency/clock-drift proof. ASIO hardware FAT NOT RUN for this PR. TINDAKAN ANDA: TIDAK PERLU MELAKUKAN APA-APA. NEXT: Continue G-MV on one bounded GUI-only ViewModel/use-case boundary chosen from latest main and actual regression evidence. Preserve single approved |
G-MV20 — One native hardware identity for v6 controller and pure Device ViewModel
Field bug prevented: backend names/epoch alone did not identify the actual installed input pins, output pins, rate or buffer. A same-epoch physical remap could leave a prior pending native v6 intent blocking new accepted snapshots (until its timeout), as well as stranded FX/output WIRE gestures in the WebView. This change makes that transition fail closed without adding a new controller or dispatcher.
web/approved-native-contract.js: one canonical, immutable hardware-configuration token from schema-v6 verified physical input/output pin SETS plus sample rate / buffer. It normalizes ordering, rejects malformed masks and accompanies accepted native snapshot. Controller invalidates its own prior in-flight request when this token changes on the same epoch; the new native readback is then published without accepting an old ACK.web/approved-device-view-model.js: the existing pure lifecycle lease consumes the same normalized token, rather than independently duplicating hardware identity or holding mutable masks. Existing bridge's MV-18 cleanup remains the only transient GUI cancellation path.tests/approved_device_pin_identity_contract.test.cjs: format/port changes, pin order, outputArmed, malformed masks, M4 first-sound, uint64 epoch, and lifecycle fence behavior.tests/approved_native_m2_contract.test.cjs: v6 controller pending FADER edit, independent physical remap before ACK, old-ACK rejection, one native emitter, no success before native readback; actual bridged M4 pending FX slot lease against same-epoch vendor swap AND mask swap.Protected invariants: ONE JUCE PCM/VST3 render graph, ONE schema-v6 reconciler, ONE existing /preview/, no WebAudio, no auto-arm, no extra retry owner, no .smix/schema wire change, no ASIO driver change. Device formats/pins come only from normalized installed native snapshot.
Baseline main:
b6dfd41393618e81f6518f53545926a6f8c90876. Code changes have been pushed; qualify with fast UI/native checks, then one exact-head full Chromium lane and required merge gate before merging. Windows artifact / physical ASIO FAT NOT RUN at branch creation. Nothing for owner to do.