Repository navigation
WIRE B2: atomic native ChannelId graph transactions and stable EdgeId - #382
Conversation
|
Consolidated valid B2 work into canonical ENGINE #266 through merged #379 ( |
|
Consolidated valid B2 work into canonical ENGINE #266 through merged #379 ( |
Packet B2 native WIRE — existing ENGINE #266 ownership
Real code: stable
edgeIdon canonicalDynamicChannelRosterlinks, atomic CONNECT/REMOVE/RETARGET/ROUTE_ONLY candidate transactions, whole-graph DAG/physical/budget checks via existingDynamicPreparedMixerBlock, and publish/rollback via existingRealtimeRoutePublisherandDynamicNativeSessionPublisher(no second PCM engine).Installed engine hook:
LiveDryRoutingLab::applyNativeWireTransactionreuses the actually installed ChannelId native PCM publisher and the same VST3 lease bank.LiveAudioEngineexposes the control-only entry point with physical device/clock guards, while existing M4 input/AUX paths are explicitly fail-closed until safe migration. Routine pure WIRE edits can publish while output is armed only when verified same-epoch physical patch remains unchanged; other structural operations remain disarm-only.Accepted truth:
DynamicConsoleRoutingView::edgesandLiveDryRoutingLab::nativeBusStatus().edgesexpose exact decimal-string edgeId/source/target/tap/gain based on the accepted PCM graph, not JavaScript simulation.Native regressions: same-sample direct / two-hop impulse, precise MAIN edge removal, stable retarget/route-only, duplicate/cyclic/stale rollback, armed wire mutation with changed hardware patch rejected. Windows CI at exact head required before promotion.
Not done / blocked: generic v8
APPLY_WIRE_TRANSACTIONintent/ACK and GUI #271 consumer still gated; original M4 CH1–10/occupied vendor+Show routing migration incomplete; universal supportedOps remains fail-closed, no production parity/FAT #15 hardware pass. Do not merge to main or call READY FOR USER TEST independently. Reconcile into single ENGINE owner #266 after exact-head CI.Refs #192 #266 #271 #270 #15. No CSS/GUI visual changes.