Fix inventory layout integration and qualify ordinary UI and radio guidance

2026-09-19 18:17:51 +02:00
parent 5c7719c19c
commit 67de3752fb
125 changed files with 18310 additions and 42 deletions
@@ -31,3 +31,5 @@ One character is the actual local player's body. The second is a newly spawned p
 The isolated service tests ordinary AI and unregistered/hostile rejection, multiple defenders sharing one interval, unconscious/dead defender loss, empty-server reset, live duration changes, stop/terminal cancellation and duplicate or re-entrant completion. Its explicit timestamps test exact timer boundaries; they are not evidence of elapsed multiplayer or real-time transmission duration. The production service and the live mission phase are unchanged by the isolated probes. After successful cleanup, normal Core smoke tunes the live radio and performs the existing real delayed transmission, interruption and recovery sequence.

 Recent native runs exposed fixture timing failures after physical mutation and during recovery. Those failures remain archived. The archived [050515](evidence/native-2026-09-19-050515/report.json) run completes the asynchronous radio fixture and ordinary escape twice, with 340 total assertions and no script/VM exceptions. Seven inherited resource errors still prevent a clean runtime pass. Scripted controller changes are not medical-treatment or revive validation. The fixture does not demonstrate two remote clients, real disconnection/reconnect timing, remote task synchronization or long-duration multiplayer behavior.
+
+The initial native task now distinguishes `WAITING` from an actual interruption: tuning starts the service before its first occupancy sample, so that short interval says "Waiting for a conscious defender" and retains the configured required duration. A later loss of defenders still says "Transmission interrupted" and resets progress. The focused `test_radio_waiting_text.py` executes the actual copy methods across these states; the Core smoke branch also checks the real task description immediately after tuning, before its first occupancy sample. The [180610 Core regression](evidence/native-2026-09-19-180610-core-inventory-radio/report.json) passes that actual task assertion in both escape/restart rounds, with 614 total assertions and zero captured errors. Remote task synchronization remains unqualified.