Repository navigation
refactor(gui G-MV27): one-shot native control authority and no stale fallback - #811
Conversation
…eny stale field fallback
… routing end-to-end
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. |
|
/arsonmix-ci ui 84e3389 |
Deploying arsonmix with
|
| Latest commit: |
84e3389
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://926aa447.arsonmix.pages.dev |
| Branch Preview URL: | https://refactor-g-mv27-unified-cont.arsonmix.pages.dev |
GUI MVVM ↔ future multi-device source inventory handoffPlease continue G-MV27's current pure control authority work independently. Multi-device coordination and IO-2 acceptance: #812. Current IO-0 #770 is merged; IO-1 #771 adds separate JUCE ASIO B producer + existing ASRC but is draft/unmerged, not installed ChannelId PCM and not hardware-FAT qualified. Ownership: G-MV owns The future integration API must expose: immutable per-device qualified input pins (B Input1 != A Input1), physical capability/masks and captureGeneration, primaryEpoch, accepted graphRevision, requestId/negative ACK and later matching installed PCM/readback. GUI Source Picker renders verified active/offline/remap states through existing pure ViewModel; source selection is not itself physical PCM proof. Current G-MV can safely preserve this boundary without implementing IO features now. Exit criteria for eventual GUI slice: SOURCE ONLY meter without unintended MAIN, ChannelId/FX/WIRE routing, unknown/missing B fail silent, .smix offline intent not discarded, exact native /preview/ parity, no extra output auto-arm. Keep G-MV's current exact-head Chromium/required gate separate from IO's native Windows/physical ASIO FAT. Cross-file handoff only once native IO-2 contract is accepted; don't block G-MV until then. |
G-MV27 final technical handoff — merged to main (2026-11-19)MERGED MAIN: PR #811 from reviewed exact head User-visible operator outcome
Changed files
Qualification
Separate existing CI baseline issueIssue #778 documents nondeterministic Stage C screenshot pixel parity: previous G-MV26 main push and one identical-SHA failed-job retry reported mismatches at different hover/default states although the exact source before merge passed. Evidence was added to that issue; do not weaken pixel parity or modify CSS inside G-MV27. This PR full exact-head parity passed; future main push results must be reported separately. Owner handoff: HASIL one unified native command preflight and no unauthorized fallthrough; STATUS MERGED MAIN; AMAN DIPAKAI as source baseline YA with protected checks, physical ASIO Windows installer BELUM DIUJI; TINDAKAN USER TIDAK PERLU MELAKUKAN APA-APA; NEXT: audit remaining one-shot FX/Device/WIRE presentation decisions for true duplicate authority only, and close CSS flaky test #778 in a standalone CI-only diagnostic/fix. Do not create extra PCM owner or duplicate bridge controllers. |
G-MV27 — one-shot channel control authority in one pure ViewModel
This follows merged G-MV26 PR #809. Native MUTE/SOLO, fader double-click unity, PAN center reset, and the final
consumeFieldbranch used separate authority checks. In particular, the final field branch could call genericSET_FADER/SET_PANafterliveScalarReadyhad already rejected a live control.Implemented
stripVM.controlCommandPlanreturnsLOGICAL,CHANNEL_ID,M4,MAIN_HOST, orBLOCKbased only on exact existing native readback/identity, capability, M4 ordinal, physical MAIN admission or offline workspace/read-only/pending status. No second controller, event bus or accepted JS mixer.controlCommandRoute/dispatchControlCommandbridge adapter selects the existing native transports for MUTE, SOLO and double-click fader/PAN reset. Removed prior divergent branches. MAIN remains on the verified physical hostroutingM4Controlscalar, not fake ChannelId main fader; offline .smix usesarMixLogicalIntentwithout audio device or automatic output arm./preview/are unchanged.Qualification
Scoped GUI-only source/tests on
main@b30a59303ad3aed486a141c96acc42dec00d2caa. Pure selector smoke PASS; repository CI not yet assessed on PR head. Require Fast UI/Native, exact-head Full Chromium/preview/, Codex review and Required Merge Gate before merge. No JUCE C++, CSS, ASIO hardware FAT or Windows installer change. Parallel N-MOD #799/#810 and IO work untouched.Known baseline issue independent of this PR
The earlier post-merge
mainUI run #44453976259 reported a hover screenshot-only Stage C CSS parity mismatch despite pre-merge exact-source Chromium success. Its failed job #135324606652 was rerun once to identify transient vs reproducible failure. Do not treat this PR as fixing CSS and do not waive its visual gate.