Repository navigation
refactor(mv7): settle native SOURCE only after ACKed exact PCM readback - #783
Conversation
Deploying arsonmix with
|
| Latest commit: |
774c9a8
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://1ed2e232.arsonmix.pages.dev |
| Branch Preview URL: | https://refactor-mv7-native-source-r.arsonmix.pages.dev |
|
/arsonmix-ci ui f1da56e |
… native readback revision
|
/arsonmix-ci ui 774c9a8 |
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. |
OWNER DELIVERY — MV-7 SOURCE readback correctness (2026-11-18)Hasil terlihat oleh pengguna: SOURCE selection stays at its verified installed input while a native ChannelId physical patch is pending; an ACK alone no longer releases the channel queue. The next channel's SOURCE edit waits for accepted native PCM proof (exact pin pair, input MONO/STEREO and graph revision) or a truly rejected/stale/timed-out request. Device reconnect and incompatible session authority are fail-closed. No visible design changes and no audio rerenderers added. STATUS: MERGED MAIN — PR #783, squash commit Tests/proof: Fast UI PASS, Fast Native PASS; exact-head Full Chromium workflow #44252952668 PASS 109/109 steps, including Stage C original/compiled CSS screenshot parity; Ready-state Fast UI PASS, Fast Native PASS, Required Merge Gate SUCCESS "All scoped checks passed" on AMAN DIPAKAI? YA for merged source-GUI correctness qualified by CI. BELUM for freshly built installer, actual vendor ASIO/dual-device hardware FAT, or an audio live release (all NOT RUN/NOT SHIPPED). Single approved Next: evaluate postmerge main jobs; next MVVM milestone reduce pending/readback duplication in channel scalar or device lifecycle, only where true domain policy is removed from the 4k-line bridge. Do not duplicate graph/state/transport or redesign approved GUI. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 774c9a8679
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".
| !decimalId(pending.ackRevision)|| | ||
| !decimalId(next.graph.revision)|| | ||
| BigInt(pending.ackRevision)<=BigInt(pending.beforeRevision)|| | ||
| BigInt(next.graph.revision)<BigInt(pending.ackRevision)) |
There was a problem hiding this comment.
Require the ACK's exact graph revision before confirming SOURCE
If another native edit advances the graph before the ACKed revision is observed, this comparison accepts the later snapshot as proof of the SOURCE request. For example, an ACK for revision 6 followed only by revision 7 with matching pins returns CONFIRMED, causing settleNativeSourceReadback to report verified PCM and release the next queued edit without observing revision 6. The existing controller in web/approved-native-contract.js instead treats readback beyond the ACK revision as conflicting. Require exact revision equality and handle a superseding revision as a conflict/refresh rather than successful settlement; AGENTS.md explicitly requires a matching ACK and same-revision native snapshot.
AGENTS.md reference: AGENTS.md:L178-L178
Useful? React with 👍 / 👎.
MV-7 — Channel Strip SOURCE pending/ACK/readback correctness
Base:
main@1e5df9a3cc91c31b4fd457fb9ce360725afcf2d6(MV-6 #782 merged). No overlapping active approved GUI PR.Actual reliability problem
The ChannelId SOURCE bridge cleared
activeNativeSourceimmediately on a positivearMixNativeSourceAck. A second queued channel could therefore dispatch against a pending/old installed PCM graph, and the first input could appear settled before its exact physical pins had been proven by native readback. A successful command response does not prove callback-applied source.Real code change (three existing files)
web/approved-channel-strip-view-model.js: new puresourceReadbackSettlement(pending,next)checks admitted ChannelId PCM, matching full string uint64 device epoch/ChannelId, reported ACK graph revision (never Number rounding), and exact nativesource.portId,portIdRight,sourceFormator actual absent SOURCE NONE; returns NONE/PENDING_ACK/PENDING/CONFIRMED/STALE. The existingsourceAckResultnow verifies ACK ChannelId and strictly later decimal-string graphRevision instead of accepting any matching requestIdok:true.web/approved-native-bridge.js: single existingactiveNativeSourceremains busy after an accepted ACK, recording publisher revision. One native callback readback settlement releases it only on independently verified native SOURCE pins, or after epoch/roster loss. Pending same-epoch queue drains only after that release. Both a cached readback arriving before ACK and an unchanged snapshot are handled; existing 8-second timeout and no-fake readback discipline are retained.tests/approved_native_m2_contract.test.cjs: positive/negative exact ChannelId/64-bit/epoch/revision, mono/stereo/right physical pin, SOURCE NONE, stale graph, output absence, and end-to-end staged ACK+unchanged PCM+later published readback+next queued channel; prior fail-ACK drain test stays enabled.Boundaries
No new ViewModel/controller/store; no C++/JUCE/ASIO/DSP/plugin/.smix/HTML/CSS/PNG changes; one actual native SOURCE dispatcher/queue, one accepted ChannelId/EdgeId PCM graph and one approved
/preview/GUI. Native ACK is never treated as physical readback or hardware FAT.Status: KODE DIBUAT; exact-head Fast UI + Fast Native + full Chromium pending; hardware FAT and installer NOT RUN. Owner action: TIDAK PERLU MELAKUKAN APA-APA.