settled-reach/docs/test-reports/sprint-11-522-preliminary.md

title, description, type, status, ticket, created, updated

title	description	type	status	ticket	created	updated
Sprint 11 #522 Preliminary Review	Preliminary QA review of the OQ-07 no-insert interaction implementation (#522). Confirms option (a) cursor shape changes with verb labels suppressed. Approved pending final test run.	test-report	archived	#522	2026-02-19	2026-02-19

Preliminary Review: #522 OQ-07 Implementation

• Date: 2026-02-19
• Build: working tree (uncommitted, pending #507 + Tyre arch review)
• Spec reference: D-056, D-057
• Reviewer: Hoshe (QA Engineer)
• Status: PRELIMINARY — full test report follows after task #4 unblocks

---

Summary

The OQ-07 resolution implements option (a): cursor shape still changes, verb labels suppressed. Implementation is correct. No critical issues. Two minor observations worth tracking.

---

Files Changed

File	Change
client/scripts/autoloads/game_state.gd	Added insert_active: bool = true, snapshot field wiring
client/scripts/rendering/cursor_renderer.gd	Added insert_active var + set_insert_active() + modified should_show_interactions()
client/ui/interaction_prompt.gd	Added _insert_active var + set_insert_active() + modified _process()
client/ui/interaction_list.gd	Already had _insert_active + set_insert_active() (prior sprint) — no change needed
client/scripts/main.gd	Propagates GameState.insert_active to cursor + list + prompt each snapshot
decisions/perception.md	OQ-07 resolution amendment appended to D-056 and D-057
client/tests/test_cursor_states.gd	3 new tests for insert-off behavior

---

Spec Compliance

D-056 diegetic test

"Interaction labels render on z-layer 6 (insert overlay). If the insert is off, labels disappear."

Status: PASS

• cursor_renderer.should_show_interactions() returns false when insert_active == false.
• interaction_prompt._process() hides prompt when not _insert_active.
• interaction_list.update_from_state() hides list when not _insert_active.

D-057 diegetic test

"Labels render on z-layer 6. If insert is off, labels disappear."

Status: PASS

• interaction_list.get_z_layer() returns Constants.CANVAS_INSERT. ✓
• interaction_prompt is on $InsertOverlay/InteractionPrompt (verified in main.gd line 8). ✓

OQ-07 option (a): cursor shape changes

"The cursor state machine still fires (shape changes: default → entity hover bracket or X-shape on object hover)"

Status: PASS

• cursor_renderer._detect_hover() is unchanged — it does not check insert_active.
• cursor_renderer.set_hover_target() is unchanged — state transitions still fire.
• Confirmed: hovering over an NPC with insert_active == false sets state to EntityHover. ✓

Decision amendment

Status: PASS

• decisions/perception.md has OQ-07 resolution appended to D-056 and D-057 entries. ✓
• Rationale documented: "the body reacts to proximity; the insert reacts to commands."

---

Observations (Non-blocking)

OBS-1: Insert state propagated every snapshot, not only on change

Location: main.gd lines 86-92 Severity: Low (v0.1 harmless — insert is always true)

main.gd propagates GameState.insert_active to three nodes on every snapshot, even when the value has not changed. In v0.1 this means set_insert_active(true) is called ~10 times per second on cursor, interaction_list, and interaction_prompt.

This is harmless now — each setter is a simple bool write with a conditional. If insert state becomes server-driven in a future sprint (a character without an insert?), adding a change-gate would avoid unnecessary _hide() tween calls.

Recommendation: Note in a ticket or code comment for v0.2. Not a blocking issue.

OBS-2: insert_active response timing inconsistency between components

Location: main.gd inside if snapshot != null: block (lines 65-127)

interaction_prompt._process() runs every frame and checks _insert_active directly. interaction_list.update_from_state() and cursor_renderer.set_insert_active() only get called when a new snapshot arrives.

In v0.1 (insert always true, no server-driven changes), this is invisible. If a future sprint allows toggling insert state without a new snapshot (e.g., a UI action), interaction_prompt would respond immediately while interaction_list and cursor_renderer would lag by up to one tick.

Recommendation: Acceptable for v0.1. For future server-driven insert toggling, consider propagating insert state every frame rather than per-snapshot. Track as technical debt.

OBS-3: set_insert_active(false) calls _hide() redundantly in next _process()

Location: interaction_prompt.gd lines 81-84 and 29-32

set_insert_active(false) immediately calls _hide_prompt() (sets _is_showing = false). Then _process() fires, sees not _insert_active, checks if _is_showing: — which is now false — and skips the second _hide_prompt() call. Correct behavior, slightly redundant guard. No bug.

---

Test Coverage for #522

Tests added by Stig (in test_cursor_states.gd)

• test_insert_off_cursor_still_changes_shape — option (a) confirmation ✓
• test_insert_off_suppresses_interactions — should_show_interactions() false ✓
• test_insert_on_restores_interaction_display — re-enable ✓

Tests pre-written by Hoshe (in test_insert_off_behavior.gd)

17 tests covering interaction list, cursor, edge cases, cross-system consistency, and regressions. These will run as part of final task #4 verification.

Coverage estimate for #522

• Happy path: 95%
• Edge cases: 85% (insert-off + weapon mode, shift override, rapid toggle)
• Integration (cursor + list agree): 80%
• Regression: 100% (all prior cursor and interaction list tests unchanged)

---

Preliminary Verdict

APPROVE pending:

1. Task #3 (Tyre arch review) — no architectural red flags found in preliminary scan
2. Final test run (task #4) once #507 is also complete

No blocking issues identified. Implementation is clean, well-commented, and diegetically consistent.