acdream/docs/research/2026-07-25-slice-g1-scene-query-candidate-gate.md

3 KiB

Slice G1 Scene-Query Candidate Gate

Date: 2026-07-25

Candidate: e346f8bbaf5c6d73498d74da32bfc37caf8c4e44

Result: accepted; G1 complete

Purpose

G1 builds the exact current render frame's candidate routes from the incremental render scene instead of walking LandblockEntries. It remains a compare-only source: production still draws the accepted InteriorEntityPartition/WorldEntity path.

The comparison preserved route, route index, cell, duplicate membership, and projection identity. At this gate it canonicalized identities within those dimensions because retained scene indices were unordered.

G2 correction: this proved membership and multiplicity, not exact source traversal. Stable equal-distance alpha ordering makes traversal observable. G2 moved that order into GpuWorldState, removed comparator canonicalization, and passed the stronger ordered candidate/packed/output gate. See docs/research/2026-07-25-slice-g2-packed-dispatcher-gate.md.

Exact physical evidence

Route Report Final comparisons Mismatch Candidate parity Exit
Capped nine-stop logs/connected-r6-soak-20260725-001333.report.json 29,392 / 29,392 0 exact at all nine checkpoints graceful
Uncapped nine-stop logs/connected-r6-soak-20260725-002328.report.json 29,025 / 29,025 0 exact at all nine checkpoints graceful
Uncapped dense Arwic logs/connected-dense-town-20260725-003231.report.json 1,522 / 1,522 0 4,274 / 4,274 at checkpoint graceful

Every route reported:

  • Passed=true, ExitCode=0, and GracefulExit=true;
  • source commit, embedded binary commit, and candidate commit equal to e346f8bbaf5c6d73498d74da32bfc37caf8c4e44;
  • product and current PView frame sequences equal;
  • expected and actual candidate count equal;
  • expected and actual candidate digest equal;
  • firstMismatch=null.

The nine-stop routes covered Caul, Sawato, Rynthid, Aerlinthe, Holtburg, same-location Sawato revisit, Caul return, Sawato plateau, and Caul plateau. They exercised outdoor and interior roots, door/outside routes, nested building look-ins, non-flooded retained cell data, portal transitions, equipped children, movement, jump, combat, generation replacement, and graceful logout. Dense Arwic exercised 15,530 resident world entities and 111 active animated statics during its stationary turn sample.

Deterministic gate

  • 24 focused current-oracle, shadow-scene, frame-product, and G1 route tests;
  • 3,745 App tests / 3 skips;
  • 8,229 complete-solution tests / 5 skips;
  • exact hidden-spatial versus portal-withdrawn candidate behavior;
  • EnvCell shell projections excluded from object routes;
  • exact mismatch reporting after a retained projection is removed.

Rollback

G1 does not change the production draw source. Commit e346f8bb is safe to retain if a later visual cutover is reverted. The exact pre-cutover runtime rollback anchor remains e7d9d6fa. G4 must record its own cutover commit before connected visual verification.