Qualify route gain and observe native gate context through ordinary input

This commit is contained in:
2026-09-19 19:44:22 +02:00
parent ec1071595d
commit 53eaf3894d
72 changed files with 71413 additions and 36 deletions
+5
View File
@@ -440,3 +440,8 @@ The successor-only [layout adapter](inventory-layout.md) preserves the complete
## Ordinary prison approach, 19 September19:17
[191750](evidence/game-2026-09-19-191750-prison-approach/runner.json.txt) passes independent native camera/input/gate validation:14turns,9W pulses,4.741m cumulative path,27.199s interval and0.490m final waypoint distance. Original gate remains closed; player stopped; zero captured errors and verified cleanup. The explicit disposable viewport is2560x720, with actual owned window/root dimensions verified in archived display evidence. All physical display and original-mission settings remain unchanged. Earlier190119(pointer limit),190707(mixed body APIs),191122(calibration return0.363° beyond0.35°), and191342(legacy inventory summary) remain failures. The191342 route separately passes raw geometry/input review. Read-only LOOK now uses actor.GetOrigin just like INPUT/GATE; body-facing and camera transform APIs remain unchanged. The approach runner now requires its dedicated raw movement validator, rather than inventory-open-dependent movement summaries. Full checks and focused integration tests pass. Lock selection/hold, exit, full escape,12clients and FPS remain pending.
## Route gain and current Core qualification,19September19:39
The [192238 aircraft run](evidence/native-2026-09-19-192238-aircraft-route-gain/report.json) passes38assertions and the [193223 Core run](evidence/native-2026-09-19-193223-core-current/report.json) passes two rounds/614assertions. Both retain zero captured engine/script errors, immutable195-file source provenance and verified private cleanup. The [hashed comparison](evidence/route-comparison-2026-09-19-192238/comparison.json) records one run per gain,55.4% less time from translation start through arrival hold and unchanged landing constraints; it is not controlled FPS or longer-route evidence. Full repository checks pass after adding the explicitly requested yaw-only alignment experiment; its native result is separate. Original workspace guard:533reference files unchanged. Release remains pending17cases.