--- phase: 06.2-gap-closure-d-50-d-53-uat-2026-05-13 status: findings-captured captured: 2026-05-13 operator: decidel verdict: not approved prior_cycle: 06.1-gap-closure-d-39-d-46-uat-2026-05-11 --- # Phase 06.2 — UAT Findings (cycle-3, 2026-05-13) > Re-run of `06.1-HUMAN-UAT.md` after Phase 06.1 plans merged and deploy-staging > redeployed staging on commit `bd93e7f`. Verdict: **not approved**. Findings > below drive Phase 06.2 gap-closure scope. --- ## Cycle-3 prior to UAT Two deploy-blocking CI failures had to be fixed first before staging picked up Phase 06.1: 1. **prediction.test.ts stale assertions** — local vitest passed against a stale `packages/game-logic/dist/step.js` carrying the old `FRICTION=0.85` / accel/decay model. CI rebuilt `dist/` from source (BNO instant-set/stop model from Plan 06.1-02) and the cumulative-vx asserts (`mid.vx>0`, `vx ≈ mid.vx * 0.85`, `afterReplay.vx>0`) all failed. Fixed in commit `bd93e7f` — three asserts rewritten to match the instant-set contract. 2. **87 commits unpushed** — local `main` was 87 commits ahead of `origin/main`. `deploy-staging` workflow fires on `push:main`. No push = no redeploy. After push, deploy-staging actually shipped Phase 06.1. These two are workflow-discipline gotchas, not Phase 06.1 code bugs. Captured here for cycle-3 post-mortem; no Phase 06.2 plan required. --- ## Operator UAT verdict per test ### Test 1 — Alpha-keyed sprites (D-39 closure) — ✅ PASS Sprite outlines clean on staging. No solid white box, no magenta halo. 06.1-test-1-alpha.png artifact captured. ### Test 2 — CLI-08 milestone — ❌ FAIL (compound) Operator findings, verbatim: 1. **Ghost player remains after logout** — `uat_b` sprite continues to render in `uat_a`'s view even though `uat_b` is not logged in at all. Server does not drop the session on WS disconnect. 2. **No floor tiles or TSides visible** — D-40 visually unclosed on the deployed staging build. (Atlas frame `0023-Tile1_000` IS present in `/atlas-mvp.json` on staging — verified by curl. So the frame exists but `RoomRenderer` never spawns tile sprites for it. Likely `layout.tiles` empty or `tileset_sprite_id` mismatch on the wire.) 3. **Background animation OK** — `0064-BKA1` scrolls correctly. 4. **Two sprites at the local player position** — local + duplicate ghost `uat_a`. The ghost has a white (remote) nameplate and animates; the local one does NOT animate. Server allows two simultaneous sessions per account. 5. **Local sprite still does not animate** (carryover from 2026-05-11 UAT — D-41 was claimed closed by Plan 06.1-04 but the visual behavior is unchanged). Sprite frame freezes on stand; remote sprites animate correctly during their movement. 6. **Player position occasionally resets back to spawn** during normal movement. Possibly reconciler snap, server-side teleport, or auth-replay artifact. 7. **Chat still works** — bidirectional message round-trip OK. ### Test 3 — Anim rate consistency (D-41/D-42) — ⚠️ partial Remote-vs-local cadence comparison impossible — local sprite never animates (see Test 2.5). Remote-only observation: animation speed feels correct when remote players are walking around. ### Test 4 — Camera + nameplate + Fixedsys (D-45/D-46) — ⚠️ partial Camera behavior seems correct. Can't verify "no black void past room edge" because there are no floor tiles to delimit the room edge. - **Cannot verify Fixedsys font load** without DevTools dive (operator did not check Network panel; nameplate text rendered but typography hard to judge against alternatives). - **D-45 reopened — nameplate Y too low** — nametag sits low enough to be overlapped by the player sprite head. Citation per CLAUDE.md: `text_y = sprite.y - 16`. Current Nameplate may be offsetting from sprite center y instead of sprite top y, producing a smaller-than-16 px above-head margin. ### Test 5 — Walkable-edge block + bg (D6.1-28/29) — ⏸ NOT TESTABLE No floor tiles → no visible room edge → cannot exercise edge-block. Same root cause as Test 2.2. --- ## Newly surfaced MVP gap (missed in 06.1 discussion) **Spawn-in init delay.** Legacy BNO has a player-spawn init delay before the run animation kicks off. Operator request: mirror this exactly. Citation lookup needed in `extracted/client-5-8/objects/0042-player/events/ Create.gml` and `Other-7.gml` (alarm 0 or Init). File as `D6.2-12` (spawn init delay) in the Phase 06.2 plans. --- ## Post-UAT e2e regression (RC1 + RC2 from triage) After the operator UAT, `deploy-staging` post-deploy Playwright also went RED with 17/21 tests failing. Triage of the 17 failures groups them under 4 root causes: - **RC1 — telemetry env-gate** (`GameScene.ts:753`, `PlayerRenderer.ts:199`): `__rebno.{roomId, cameraScrollX, loadedTextures, localPlayerX/Y, localFrame}` hooks gated by `DEV || MODE==='test'`. Staging build runs `MODE='staging'` → entire hook block silenced → ~7 e2e failures. **Fixed in commit `cd47745` (local, not yet pushed). Drops the gate.** - **RC2 — login-contrast missed inviteSuffix** (`login-contrast.e2e.test.ts:56` used `page.goto('/')`): 7 login-contrast tests timed out on `form#login-form h1` because staging returns 401 without `?invite=$TOKEN`. **Fixed in commit `cd47745`.** - **RC3 — `data-game-ready` never set on some tests** (ws-kill-reconnect, sprite-state, logout, cli-08 first-pass) — to be re-evaluated after RC1 re-deploy; some of these likely have RC1 as the actual cause. - **RC4 — Genuine code regressions** — captured below as Phase 06.2 plans. --- ## Phase 06.2 scope summary | ID | Finding | Source | Status | |---|---|---|---| | D-40 | Floor tiles + TSides not rendering on staging | Test 2.2 | reopened | | D-41 | Local sprite no-animate | Test 2.5, claimed closed in 06.1-04 | reopened | | D-45 | Nameplate Y too low — overlaps sprite | Test 4 | reopened | | D-50 | Server: ghost player remains after WS disconnect | Test 2.1 | new | | D-51 | Server: duplicate account login allowed | Test 2.4 | new | | D-52 | Player position occasionally resets to spawn | Test 2.6 | new | | D-53 | Legacy spawn-in init delay missing | Operator request | new | **Out of cycle-3 scope** (defer Phase 7 or later): everything from the Phase 06.1 deferred list still defers. --- *Cycle-3 UAT captured 2026-05-13. Next: `/gsd-discuss-phase 06.2` to lock implementation decisions for D-40/D-41/D-45 reopenings + D-50..D-53 new findings, then `/gsd-plan-phase 06.2`.*